Transcriptional Regulation of Metastatic Progression
in Lung Adenocarcinoma
by
Carman Man-Chung Li
A.B., Molecular Biology
Princeton University (2009)
Submitted to the Department of Biology
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2015
© 2015 Carman Li. All rights reserved.
The author hereby grants to MIT permission to reproduce and to distribute publicly
paper and electronic copies of this thesis document in whole or in part
in any medium now known or hereafter created.
Signature of Author ......................................................................................................
Department of Biology
May 22, 2015
Certified by .................................................................................................................
Tyler Jacks
Professor of Biology
Thesis Supervisor
Accepted by.................................................................................................................
Michael Hemann
Professor of Biology
Chair, Committee for Graduate Students
Transcriptional Regulation of Metastatic Progression
in Lung Adenocarcinoma
by
Carman Man-Chung Li
Submitted to the Department of Biology on May 22, 2015 in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy in Biology
ABSTRACT
Lung cancer is the most prevalent cancer type, leading to more than one million
deaths per year worldwide. The vast majority of these mortalities were attributed to
metastasis, which is the dissemination of tumor cells from the lungs to other organs. The
molecular mechanisms for metastasis is complex and not well understood. In this thesis, I
investigated the gene expression changes in tumor cells that contribute to metastasis of
lung adenocarcinoma, the major subtype of lung cancer.
Using a genetically-engineered mouse model and derivative cell lines, we showed
that metastatic lung adenocarcinoma cells are capable of forming proteolytic membrane
protrusions known as invadopodia to degrade the extracellular matrix. The formation and
function of invadopodia are dependent on an isoform switch of the adaptor protein Tks5.
The Tks5long isoform, which is upregulated in metastatic cells, is capable of localizing to the
cell membrane and activating invadopodia formation. In contrast, the Tks5short isoform, which
is transcribed from a promoter independent of Tks5long, is the predominant isoform in nonmetastatic cells, and functions to inhibit invadopodia-mediated matrix degradation by
destabilizing these protrusions. We demonstrated that an increased ratio of Tks5long-toTks5short promoted invadopodia activity in vitro and metastasis in vivo. Furthermore, a high
Tks5long-to-Tks5short ratio in human tumors correlated with advanced stage and worse
survival. These data strongly suggest that a balance between Tks5long and Tks5short
expression is critical for metastasis.
In addition, we found that the expression of the pro-metastatic Tks5long isoform is
synergistically inhibited by three transcription factors – Nkx2-1, Foxa2, and Cdx2. These
three factors were highly expressed in non-metastatic cells, and downregulated in
metastatic cells. Altered expression of these factors led to commensurate changes in
Tks5long levels.
Finally, we demonstrated that Nkx2-1, Foxa2, and Cdx2 function cooperatively to
inhibit metastasis by suppressing a network of target genes. Silencing of all three factors in
non-metastatic cells activated a program of metastasis-related genes, and increased
metastasis in a transplantation model. Furthermore, the expression patterns of these factors
strongly correlated with tumor progression in an autochthonous model of lung
adenocarcinoma, and were closely associated with disease stage and survival outcomes of
human patients. Collectively, these findings strongly argue that Nkx2-1, Foxa2, and Cdx2
synergize to restrain metastatic progression.
Taken together, this study provides insights on some of the key molecular regulators
of lung cancer metastasis. Our findings contribute to a better understanding of metastasis,
and potentially to the development of better therapeutic strategies in the future.
Thesis Supervisor: Tyler Jacks
Title: Professor of Biology
2
ACKNOWLEDGEMENTS
As I reflect on my graduate career, I feel very fortunate to have worked with and learned
from a wonderful group of people at MIT. They have significantly enriched my research
experience and scientific development, as well as personal growth. Here I will take a
moment to acknowledge them individually.
First and foremost, I am extremely grateful to Tyler for his unceasing support and guidance.
I highly appreciate his confidence in me, and the freedom he has given me to explore and
develop my projects. Tyler has been a great role model in many different ways – not only in
how to think like an esteemed scholar, but also how to be an effective leader and a great
mentor. From Tyler I have learned a tremendous amount over the past five years. It was
also a great pleasure to have worked in the Jacks lab because of the creative and collegial
research environment that Tyler has fostered. Thank you for everything!
I also wanted to thank my thesis committee, Phil Sharp and Frank Gertler, for their valuable
inputs in my projects over the past five years, and to David Barbie for his willingness to
participate in my thesis defense as an external faculty member.
My special gratitude to Judy Teixeira, Anne Deconinck, Ines Baptista, Kate Anderson,
Margaret Magendantz, and Kim Mercer who often go out of their ways to make the lab a
wonderful work environment on a day-to-day basis. Judy in particular has given me great
help in printing this thesis.
To all other members of the Jacks lab, past and present, I cherish your friendship, advice,
and assistance. I am thankful to have worked with such a dynamic group of aspiring
scientists who are dedicated to pushing the boundaries of science. Importantly, their
perseverance in forging forward during difficult times in research and personal lives is
admirable and inspiring. Specifically, I am fortunate to be companied by many brilliant
baymates over the past years, including Trudy Oliver, Greg Chang, Thales
Papagiannakopoulos, Megan Heimannann, Mandar Muzumdar, Rebecca Robbins, Kim
Dorans, and my most kind-hearted bench-share partner, Kim Mercer. Mandar was
especially helpful for sharing his medical knowledge and giving me valuable career advice. I
also wanted to acknowledge Nik Joshi, Michel DuPage, David McFadden, Keara Lane,
Alison Dooley, and Francisco Sanchez-Rivera for showing me the ropes when I first joined
the lab, as well as Monte Winslow, Eric Snyder, David Feldser, Wen Xue, Tuomas
Tammela, and Irene Blat, for helping me on my projects and generously sharing reagents
and data. To all my fellow graduate students in the lab: thank you all for your camaraderie
and support. I wanted to particularly give a shout-out to my classmates Leah Schmidt and
Talya Dayton, for the special bonds we share by joining the lab together and developing our
graduate careers side-by-side. I also wanted to thank my UROPs, Alice Wang and Saya
Date, as well as Sheng Rong Ng, who did his rotation with me, not only for their hard work
that have contributed to this thesis, but also importantly for giving me the opportunity to
learn how to become a better mentor.
Among all the lab members, I particularly wanted to thank Leny Gocheva, Nadya Dimitrova,
and Anna Farago for being my source of inspirations, my mentors and friends. I
tremendously enjoyed the chats we had about our projects and other topics in science and
things-not-science, and I truly value their advice on my research and career development.
Leny was especially helpful in giving me advice on my presentation and writing skills,
including this thesis. Nadya, in her own unique way, has always helped me feel positive and
confident even in difficult times.
3
Outside of the lab, I owe my gratitude to Amy Keating, who has become my informal faculty
mentor and a role model at MIT over the past five years. Our regular lunch chats about
research, science philosophy, and professional development have added a special
dimension to my experience at MIT. Thank you for sharing with me your wisdoms and
insights. Amy’s can-do spirits and great sense of humor has also given me much
encouragement over the years.
Last but not least, to my parents and my husband Lawson, thank you for your unconditional
love and support in every step of the way in pursuing my dreams.
4
TABLE OF CONTENTS
ABSTRACT ............................................................................................................................. 2
ACKNOWLEDGEMENTS ....................................................................................................... 3
TABLE OF CONTENTS .......................................................................................................... 5
CHAPTER 1: INTRODUCTION .............................................................................................. 7
I. AN OVERVIEW OF METASTASIS .................................................................................. 9
1. The metastatic cascade .............................................................................................. 9
2. Clonal selection in metastasis ................................................................................... 11
II. MECHANISMS OF TUMOR MIGRATION AND INVASION ......................................... 14
1. Cytoskeletal changes during cellular migration ......................................................... 14
2. Modes of migration in tumor cells in vivo .................................................................. 15
3. Environmental cues: extracellular matrix, chemokines, and growth factors ............. 17
4. Remodeling the environment: matrix proteases ....................................................... 19
5. Specialized cell-membrane protrusions for migration and invasion .......................... 19
6. Invadopodia............................................................................................................... 20
6.1 Discovery of invadopodia ........................................................................................ 20
6.2 Invadopodia and podosomes: similarities and differences ..................................... 21
6.3 Formation and signal transduction of invadosomes ................................................ 22
6.4 Key components of invadosomes ........................................................................... 26
6.5 The roles of podosomes in normal development and physiology ........................... 35
6.7 The roles of invadopodia in tumor invasion and metastasis ................................... 38
III. ROLES OF DEVELOPMENTAL TRANSCRIPTION FACTORS IN METASTASIS ..... 41
1. Dedifferentiation in tumor progression ...................................................................... 41
1.1 Embryonal transcription factors that promote tumor progression ........................... 42
1.2 Lineage-specific transcription factors that suppress tumor progression ................. 45
2. Lineage survival oncogenes ..................................................................................... 46
3. Context-dependent functions of Nkx2-1, Cdx2, and Foxa2 ...................................... 47
3.1 The roles of Nkx2-1 in lung adenocarcinoma ......................................................... 48
3.2 The roles of Cdx2 in colorectal cancer. ................................................................... 50
3.3 The roles of Foxa2 in lung cancer and neuroendocrine prostate cancer. ............... 52
3.4 Explaining the diverse roles of developmental transcription factors in cancer ........ 54
REFERENCES ................................................................................................................. 57
CHAPTER 2: DIFFERENTIAL TKS5 ISOFORM EXPRESSION CONTRIBUTES TO
METASTATIC INVASION OF LUNG ADENOCARCINOMA ................................................ 79
Abstract ............................................................................................................................. 80
Introduction ....................................................................................................................... 81
Results .............................................................................................................................. 84
Conclusions .................................................................................................................... 101
Materials and methods .................................................................................................... 105
Acknowledgements ......................................................................................................... 114
Supplemental figures ...................................................................................................... 115
References ...................................................................................................................... 121
5
CHAPTER 3: FOXA2 AND CDX2 COOPERATE WITH NKX2-1 TO INHIBIT LUNG
ADENOCARCINOMA METASTASIS ................................................................................. 124
Abstract ........................................................................................................................... 125
Introduction ..................................................................................................................... 126
Results ............................................................................................................................ 128
Conclusion ...................................................................................................................... 151
Materials and methods .................................................................................................... 154
Acknowledgements ......................................................................................................... 162
Supplemental figures ...................................................................................................... 163
References ...................................................................................................................... 170
CHAPTER 4: DISCUSSION AND FUTURE DIRECTIONS ................................................ 173
Tks5 isoforms regulate invadopodia and metastatic spread of lung adenocarcinoma ... 174
Tks5long expression is suppressed by cooperation between Nkx2-1, Foxa2, and Cdx2 . 177
Nkx2-1, Foxa2, and Cdx2 synergistically suppress lung adenocarcinoma metastasis ... 178
Tumor dedifferentiation versus tumor stem cell expansion ............................................. 180
Dedifferentiation and activation of alternative state in metastasis .................................. 181
Potential mechanisms for loss of lineage transcription factors in metastasis ................. 184
Implications on therapeutic strategies ............................................................................. 185
A unifying theme for differentiation and metastasis? ...................................................... 186
References ...................................................................................................................... 187
6
CHAPTER 1
INTRODUCTION
7
Metastasis, or the spread of cancer cells from the primary tumor of origin to other
parts of the body, is the leading cause of cancer morbidity and mortality. The World Health
Organization has estimated that about 8 million cancer deaths occur worldwide per year
(Ferlay et al., 2014), and the vast majority of patients died of metastatic disease (Gupta and
Massague, 2006). While many tumors that are restricted to their primary sites can be
effectively treated by surgery and radiation, cancer that has systemically spread to the rest
of the body is often incurable. In the context of lung cancer, which is the most prevalent
cancer type worldwide, the five-year survival rate drops precipitously from 50% in stage I
patients with localized lung lesions to less than 5% in stage IV patients with metastatic
disease. The reasons for such high mortality include organ failures due to damages caused
by tumor growth, systemic paraneoplastic syndromes induced by hormonal and
immunostimulatory secretions from tumor cells, and complications that arise from
aggressive chemotherapy treatments (Steeg, 2006).
Metastasis is a complex process that draws on numerous aspects of biology. The
development of metastasis involves a succession of interrelated steps, through which the
cancer cells disseminate from the primary tumor of origin, travel in the lymphatic and
cardiovascular circulation, and form secondary outgrowths in distant organs. This metastatic
cascade involves intricate interactions between the tumor and its surrounding environments,
and imposes tremendous selective pressure on molecular alterations and cell state changes
within the tumor cells that favor motility, invasion, and proliferation.
This thesis will focus on two important properties of metastasis in lung cancer,
namely the molecular regulation of cellular invasion, and how alterations in transcriptional
programs affect invasion as well as other aspects of tumor dissemination. The findings from
this work will provide a better understanding of metastasis, and may contribute to the
development of more effective therapeutic strategies for metastatic disease.
8
I. AN OVERVIEW OF METASTASIS
The first part of this introduction will give a brief outline of the metastatic cascade,
which describes the multiple steps involved in the dissemination of tumor cells. I will also
provide a short overview of the clonal selection process in the outgrowth of metastatic cells.
The paradigms of metastatic cascade and clonal selection are important as they have given
rise to many important concepts of metastasis in the past decades, and have shaped our
understanding of metastasis in the present day.
1. The metastatic cascade
The process of metastasis is composed of a sequence of interrelated steps. The first
step begins with the invasion of cancer cells from the primary tumor into the surrounding
tissue (Roussos et al., 2011b). In this process, tumor cells partially lose their cell-to-cell and
cell-to-basement membrane adhesions, which are maintained by various structures
including tight junctions, adherens junctions, desmosomes, and hemi-desmosomes. In
addition, metastatic tumor cells gain the motility required to migrate out from the primary
tumor, often times guided by environmental cues from the extracellular matrix, chemokines,
and growth factors. Furthermore, metastatic tumor cells are able to transverse through their
surrounding physical barriers, such as the basement membrane and the interstitial matrix,
by either squeezing through existing gaps in the extracellular matrix via amoeboid
movement, or by secreting proteases to degrade the extracellular matrix components.
The second step of metastasis is dissemination of tumor cells into the lymphatic or
blood vasculature. Tumor cells are thought to enter systemic circulation either by actively
invading into pre-existing lymphatic or blood vessels in the local microenvironment, or by
stimulating formation of new lymphatic and blood vessels that are prone to leakiness and
9
thus favor tumor intravasation (Carmeliet and Jain, 2011; Tammela and Alitalo, 2010). In the
circulation, metastatic tumor cells undergo several layers of selection, including resistance
to the shear force of vascular circulation, suppression of anoikis (a form of apoptosis
induced by lack of anchorage to substratum), and the ability to evade detection and
elimination by immune cells. Several mechanisms have been proposed to facilitate survival
of circulating tumor cells, including formation of tumor cell clusters in circulation, and
association with blood platelets (Aceto et al., 2014; Gay and Felding-Habermann, 2011).
Finally, circulating tumor cells are thought to be arrested in the distant sites either by
passive entrapment in the narrow passages of the capillaries, or by selective interaction
between cell surface receptors on tumor cells (such as adhesion molecules and cytokine
receptors) and their cognate ligands expressed in the target tissues (Ruoslahti and Rajotte,
2000; Zlotnik et al., 2011). A subset of these arrested cells may gain access to the
parenchyma of the distant organ by either proliferating within the capillaries and physically
rupturing the vasculature, or by extravasating out of the vasculature before establishing an
outgrowth. In the distant organ, metastatic tumor cells undergo selection for adaptive growth
in an environment different from their primary site of origin. A subset of tumor cells may
undergo cell death, enter a quiescent state, or proliferate at a slow rate. However, rare
outgrowth of the metastatic cells that are capable of adapting to and proliferating in the new
environment leads to the establishment of macrometastases in the distant organ (Giancotti,
2013). Furthermore, these established metastases may re-enter the metastatic cascade,
thereby disseminating to a new location in the body and forming secondary metastases
(Hoover and Ketcham, 1975).
10
2. Clonal selection in metastasis
The metastatic cascade leads to clonal selection and outgrowth of only a
subpopulation of tumor cells that have the ability to accomplish the multiple steps of
metastasis. The first compelling experimental evidence for the selective nature of metastasis
comes from Isaiah Fidler in 1970, who reported that metastasis could result from the
survival of only a few tumor cells (Fidler, 1970). By Injecting radiolabeled melanoma cells
intravenously into mice, Fidler found that the vast majority of tumor cells in circulation were
cleared soon after injection, and less than 0.01% of the injected cells were able to produce
experimental lung metastases. Subsequent study by Fidler and Kripke demonstrated that
different subclones of tumor cells had varied ability to form lung nodules following
intravenous injection into syngeneic mice, suggesting that the tumor cells are a
heterogeneous population, and those with metastatic potential can be selected for their
outgrowth (Fidler and Kripke, 1977) .
What regulates the successful formation of metastasis in this highly selective
process? A century before Fidler’s time, Stephen Paget proposed the “seed and soil”
hypothesis to explain the determining factors for metastasis formation (Paget, 1989). Paget
postulated that both the tumor cells (the seeds) and their surrounding environments (the
soil) are critical in determining metastasis outgrowth. Paget formulated the hypothesis based
on the observations from patients that tumors of certain tissues of origin were more prone to
disseminating metastasis in distant organs compared to other tumor types, and that within a
certain cancer type, specific organs were more susceptible to metastasis formation than
other organs. For example, patients with breast cancer developed metastases more
frequently than patients with uterus and intestinal cancers, and breast tumors disseminated
more frequently to the liver and the bones than to the spleen. Thus, both the tumor cells and
11
the target organs appeared to have special properties for shaping the dynamics of
metastasis.
Extending upon the seed-and-soil hypothesis, several major concepts have
developed over the past decades to explain the pro-metastatic properties of the tumors and
their environment. Although many of these topics are outside the scope of this thesis, I will
highlight a number of important areas. The majority of the studies have focused on the
tumor cells themselves, in order to dissect the molecular mechanisms that enable migration,
survival, and outgrowth. Gene expression profiling analysis of tumor cells has revealed that
distinct tropism-specific gene expression signatures in subpopulations of tumor cells can
mediate their dissemination to certain organs (Bos et al., 2009; Kang et al., 2003; Minn et
al., 2005). Furthermore, the cancer stem cell theory has been put forth by Irving Weissman
(Reya et al., 2001). It argues that cancer is propagated by a small subpopulation of
malignant cells that possess stem-like characteristics, including self-renewal, resistance to
apoptosis, independent growth, tumorigenicity, and metastatic potential. In addition, the
epithelial-to-mesenchymal transition (EMT) theory has been proposed by Paul Thiery and
further supported by studies from the laboratories of Robert Weinberg and others (Thiery,
2002; Yang et al., 2004). It contends that tumor cells of epithelial origin may gain metastatic
ability by usurping a developmental process that converts cells from an epithelial state into a
mesenchymal state by activating transcription factors such as Snail, Slug, and Twist.
Besides the intrinsic properties of the tumor cells, the stromal components
surrounding the tumors have also been studied to better understand metastasis. These
stromal components, including the extracellular matrix, immune cells, fibroblasts and
endothelial cells, have been shown to play regulatory roles in promoting or inhibiting
metastasis of various cancer types (Hanahan and Coussens, 2012; Joyce and Pollard,
2009). It has been further proposed that metastasis can be promoted by the formation of the
12
“pre-metastatic niche”, a permissive microenvironment in the secondary site that is induced
by factors secreted by cells in the primary tumor (Kaplan et al., 2006).
For the rest of this introduction, I will focus on two important aspects of metastasis
directly related to this thesis work. First, I will review the molecular mechanisms that regulate
the migration and invasion properties of metastatic tumor cells, as they relate to Chapter 2
of this thesis. Second, I will discuss the roles of developmental transcription factors and
differentiation states in metastatic tumor progression, as they are relevant to Chapter 3 of
this thesis. While these two topics appear to associate with different properties of
metastasis, they are both part of the bigger picture of how alterations in cell states may
confer selective advantage in the process of metastasis.
13
II. MECHANISMS OF TUMOR MIGRATION AND INVASION
The ability of cells to migrate and invade is critical for the metastatic process. During
metastasis, malignant cells are selected for their capacity to move through the surrounding
normal tissues and transverse the endothelial basement membranes both during
intravasation into lymphatic and cardiovascular vessels at the primary site and extravasation
out of the circulation at the distant site. Below I will review the current understanding of the
molecular mechanisms that regulate migration and invasion of cancer cells.
1. Cytoskeletal changes during cellular migration
The migration of tumor cells is thought to follow a common intracellular mechanism
that can be divided into three steps (Friedl and Wolf, 2003). The first step is protrusion of the
leading edge of the cell, which determines the directionality of the migration. Upon
stimulation of cell-surface receptors by chemokines and growth factors, extension of actin
filaments pushes the cell membrane outward, forming various types of cellular protrusions.
The process of actin polymerization is mediated by the ARP2/3-WASP complex, which
promotes the nucleation and branching of actin filaments.
The next step is formation of cellular adhesion to the extracellular matrix via clusters
of integrins at the leading edge of the migrating cell. Integrins are heterodimers of one of 18
α and one of 8 β transmembrane proteins (Desgrosellier and Cheresh, 2010). The
extracellular domain of the integrin heterodimer binds to the extracellular matrix, thus
allowing the integrins to cluster and thereby recruit adaptor and signaling proteins via their
cytoplasmic domain. The cytoplasmic domain of integrin interacts with α-actinin, talin, and
the focal adhesion kinase (FAK), which together recruit actin-binding proteins (vinculin,
14
paxillin and α-actinin) and regulatory molecules (PI3K and RHO-family GTPases) to focal
contacts.
The final step of cellular translocation involves contraction of the cell body and
retraction of the trailing end of the cell. Throughout the cell body, contraction of the actin
filament bundles is mediated by movement of myosin II. This actomyosin contraction pulls
the cell body towards the anchored leading edge. At the trailing end of the cell, focal
adhesions dissemble via severing of actin filaments, degradation of focal contact
components, and cleavage of adhesion receptors. Upon disassembly of focal adhesion,
integrins detach from the extracellular matrix to free the trailing edge of the cell. The
integrins are recycled by endocytotic vesicles. This repeated cycle of protrusion, adhesion,
and detachment results in cell movement.
2. Modes of migration in tumor cells in vivo
Variations from the above general model of migration due to differences in tumor cell
types and the extracellular environment can lead to distinct modes of migration. Broadly, the
modes of tumor cell migration can be categorized into individual-cell migration and collective
migration. Individual-cell migration involves mesenchymal or amoeboid movement of
dissociated cells, whereas collective migration involves translocation of sheets or clusters of
cells that retain cell-cell adhesion.
In mesenchymal individual-cell migration, tumor cells move with an elongated
spindle-like morphology in a crawling motion. It has been proposed that tumor cells that
have undergone epithelial-to-mesenchymal transition adopt this mode of translocation
(Thiery et al., 2009). In addition to following the protrusion-adhesion-retraction model
described above, mesenchymal migration is also facilitated by degradation of the
extracellular matrix by various proteases, such as metalloproteinases and serine proteases.
15
In amoeboid individual-cell migration, tumor cells move through the extracellular
matrix by adopting a short, ellipsoid morphology and squeezing through the interstitial space
(Roussos et al., 2011b). Instead of relying on protease digestion of the extracellular matrix,
amoeboid migration utilizes actomyosin-mediated contractile forces which allow constrictioncompression of the cytoplasm. Therefore, the ability of the cells to squeeze through the
extracellular matrix is limited by the deformability of the nuclei, which in turn is determined
by the nuclear lamin intermediate filaments and chromatin structure. Furthermore, amoeboid
migration involves weaker attachment of the cell to the extracellular matrix and relatively
diffuse organization of focal adhesions. Thus amoeboid migration occurs at relatively high
speed (up to ~4 um min-1) compared to mesenchymal migration (ranging at 0.1-1 um min-1)
(Roussos et al., 2011b). This model of movement has been observed in a number of
carcinoma cells in vivo in transplantation models.
Individual cells in mesenchymal or amoeboid migration may translocate alone or in
groups. The latter case is called multicellular streaming, and involves individual migrating
cells following one another in the same path. The leading cell positioned at the front of the
group can be either a tumor cell or a stromal cell co-opted by the tumor, and is thought to
actively degrade the extracellular matrix to create a path for migration. Cells positioned
downstream of the leading cell co-migrate by following the path of degraded extracellular
matrix (Roussos et al., 2011b).
Tumor cells have also been observed to migrate collectively in sheets or clusters
held together by cell-cell junction (Friedl et al., 2012). One or more cells in the group are
thought to act as leading cells by actively forming leading-edge protrusions and degrading
the extracellular matrix, while the successor cells physically coupled to the leading cells are
being pulled by the movement of the leading cells along the path created.
16
These different modes of migration are plastic and exchangeable. For example,
inhibition of proteolysis can induce a transition from mesenchymal to amoeboid migration
(Wolf et al., 2003), whereas enhanced paracrine chemokine loop between tumor cells and
stromal cells can induce amoeboid movement (Roussos et al., 2011a). Furthermore,
antibody-mediated blockage of integrins has been observed to abolish collective migration
and favor amoeboid movement (Hegerfeldt et al., 2002).
3. Environmental cues: extracellular matrix, chemokines, and growth factors
A wide variety of environmental cues can stimulate cell migration and invasion,
including the extracellular matrix (ECM) components, chemokines, and growth factors.
Tumor cells are capable of sensing these environmental cues via their adhesion molecules,
as well as cell surface receptors.
Cell movement can be affected by physical differences in the ECM, which includes
the basement membranes, interstitial connective tissues, and aligned ECM bundles. The
basement membrane lies at the interface between epithelial cells in an organ and the
interstitial connective tissue, and is composed of a dense layer of laminins, crosslinked
collagen type IV, and proteoglycans. The connective tissues consist of loose fibrillar
meshworks of crosslinked collage types I and III, with interstitial pores of variable size and
shapes. Finally, aligned ECM bundles include packed myofibers, blood vessels, and
neuronal strands decorated with basement membranes. Tumor cells can bind and interact
with these ECM components via their cell surface integrins and non-integrin receptors, such
as CD44, CD26, discoidin receptors, immunoglobulin superfamily receptors, and surface
proteoglycans. In addition to the composition of the ECM, its stiffness is another important
determinant of cellular translocation. Increased substrate stiffness promotes leading-edge
protrusions and cellular movement, while soft matrix supports cell rounding and inhibits
17
locomotion (Peyton et al., 2008; Ulrich et al., 2009). Thus, cells have a tendency to migrate
towards substrates of greater stiffness, which is a behavior called durotaxis. Porosity of the
ECM also affects cell migration: the speed of movement is optimal if the pore diameter of
the ECM is similar or slightly smaller than the diameter of the cells (Harley et al., 2008).
Finally, orientation of ECM fibers can regulate migration. Migrating cells tend to align in
parallel along oriented structures, such as muscle fibers, blood vessels, and neuronal
strands (Petrie et al., 2009; Provenzano et al., 2008). These structures guide the tumor cells
to migrate in a linear direction.
In addition to the ECM itself, migration of tumor cells can be affected by the
chemokines and growth factors that are tethered to or diffuse across the matrix.
Chemokines and growth factors induce tumor cell migration by binding to cell-surface G
protein-coupled receptors and receptor tyrosine kinases, respectively. Gradients of
chemokines and growth factors in the tumor environment could provide directionality for the
intravasation step of metastasis. For example, CCR7 can guide tumor cells to migrate
towards the lymphatics (Shields et al., 2007). Gradients of chemokines and growth factors
could also explain the tissue tropism of metastatic tumor cells. It has been proposed that
chemokines and growth factors produced in specific organs can provide survival and
proliferation cues to subsets of receptor-expressing metastatic cells that are entrapped in
the capillary bed during circulation in the bloodstream (Chambers et al., 2002; Joyce and
Pollard, 2009). Finally, it has been proposed that chemokines and growth factors could
guide metastatic cells in a long-range manner, such that metastatic tumors cells in the
primary site could be attracted to a “metastatic niche” in a distant organ that highly express
specific chemokines or growth factors. Examples include CXCR4-CXCL12 interaction in
bone metastasis of breast and prostate cancer, CCL27-CCR10 interaction in skin
metastasis of melanoma, and CCL19/21-CCR7 interactions in lymph node metastasis of
18
various solid and hematopoietic tumors (Ben-Baruch, 2008; Koizumi et al., 2007; Lazennec
and Richmond, 2010; Müller et al., 2001).
4. Remodeling the environment: matrix proteases
Various proteases are involved in the degradation and remodeling of the ECM during
migration and invasion of tumor cells. These proteases include metalloproteases (MMPs
and ADAMs), serine proteases (seprase and uPAR), and cysteine proteases (cathepsins B,
L, and S). These enzymes can be delivered to the ECM either in their active form or as
proenzymes to become activated by other proteases at the cell surface (Mason and Joyce,
2011).
Together these proteases contribute to the degradation of the ECM via two
mechanisms: diffuse proteolysis mediated by secreted proteases, and focal proteolysis
executed by cell surface-associated proteases. Secreted proteases (e.g. MMPs and
cathepsins) are released by tumor cells into a widely distributed area in the ECM to induce
gradual degradation of the matrix (Wolf and Friedl, 2011). Cell surface-associated proteases
are either inserted into the plasma membrane via a transmembrane domain (in the case of
MT-MMPs and seprase, for instance) or tethered to the cell surface via receptors (for
example, in the case of MMP2, which is bound to αVβ3 integrins or to MMP14). These
proteases induce localized degradation of the ECM in the direct vicinity of the tumor cells.
Finally, solubilized ECM particles generated by these degradation mechanisms can be
further removed by endocytic uptake and lysomal degradation in the tumor cells.
5. Specialized cell-membrane protrusions for migration and invasion
Several types of specialized cell-membrane protrusions have been found to mediate
tumor cell migration and invasion, including lamellipodia, filopodia, and invadopodia (Ridley,
19
2011). Lamellipodia are broad and flat cellular extensions that protrude from the leading
edge of a cell. They are composed of large agglomerations of short, branched actin
filaments. Lamellipodia are thought to be a major driving force for cellular migration.
Filopodia are rod-like protrusions extending from the leading edge of the cells.
Architecturally, they are composed of tight bundles of long actin filaments. Filopodia are
thought to act as sensory organelles, providing the cells with a means to probe the
extracellular matrix for guidance cues. In contrast to lamellipodia and filopodia that are found
in the leading edge, invadopodia are cylindrical cellular protrusions observed on the ventral
side of cells. They are composed of actin filament bundles and meshwork. Unlike
lamellipodia and filopodia, invadopodia are capable of secreting proteases such as MMPs,
ADAMs, cathepsins, and serine proteases to induce focal proteolysis of the ECM. Thus they
are thought to be a critical mechanism for cellular invasion during metastasis.
6. Invadopodia
As invadopodia are related to Chapter 2 of this thesis, I will give a brief review of
these invasive structures and their roles in metastasis below.
6.1 Discovery of invadopodia
The proteolytic cellular protrusions that we now term invadopodia were first
described in the early 1980s in fibroblasts transformed by Rous sarcoma virus. They
appeared as adhesive structures that were capable of degrading the extracellular matrix
(Chen et al., 1985; David-Pfeuty and Singer, 1980; Tarone et al., 1985). These structures
were initially called “rosettes” or “podosomes” for their ringed structure and adhesive nature.
Similar proteolytic protrusions were later observed in human cancer cells in culture (Chen,
1989), and were termed “invadopodia” (for invasive feet) to emphasize their invasive nature.
20
Subsequent studies in the past decades have identified these proteolytic protrusions in
numerous cell types, including normal cells and cancerous cells, even though the
architecture of these protrusions vary slightly depending on the specific types of cells and
the experimental conditions used for their observation.
The name invadopodia was sometimes interchangeably used with podosomes to
describe these protrusions, and this issue of nomenclature has caused confusions and
complications in comparison of published data in the literature. The current consensus on
the nomenclature is to use the term “invadopodia” for protrusions found in cancer cells, and
the term “podosomes” for protrusions found in untransformed cells (Murphy and
Courtneidge, 2011). The term “invadosome” has been coined to collectively refer to both
podosomes and invadopodia when the distinction between the two structures is not critical.
To date, invadopodia have been observed in tumor cells derived from lung cancer, breast
cancer, prostate cancer, melanoma, and head and neck squamous cell carcinoma, while
podosomes have been identified in macrophages, dendritic cells, lymphocytes, osteoclasts,
endothelial cells, and neurons (Murphy and Courtneidge, 2011). We will discuss the
similarities and differences between invadopodia and podosomes in further details below.
6.2 Invadopodia and podosomes: similarities and differences
Invadopodia are closely related to podosomes, as they share numerous similarities
in terms of structure and function. Invadopodia and podosomes are both actin-rich punctate
protrusions that extend from the ventral membrane of the cell into the extracellular matrix.
They both contain actin-regulatory proteins (such as cortactin, WASP, and Arp2/3) and
signaling proteins (such as kinases and adaptor proteins). Furthermore, they can both
deliver proteases focally to degrade the extracellular matrix. As such, both invadopodia and
podosomes are considered to provide an important mechanism for cellular invasion and
21
migration across physical barriers (Buccione et al., 2009; Linder, 2007; Murphy and
Courtneidge, 2011).
Despite their similarities, invadopodia and podosomes can be distinguished by a
number of different properties apart from the cell types in which they are found. The first
difference is their size and abundance. Invadopodia are relatively large (approximately 0.5-2
µm in diameter, and >2 µm in height), and are present in smaller numbers (1-10 per cell). In
contrast, podosomes are smaller (about 0.5-2 µm in diameter and height), and are more
abundant (20-100 per cell) (Sibony-Benyamini and Gil-Henn, 2012). Second, invadopodia
and podosomes differ in their dynamics. Invadopodia are relatively stable and can persist for
hours, while podosomes are more short-lived and have a turnover rate of several minutes.
As a result, the pattern of degradation induced by invadopodia is more focal and penetrates
deeper, while the degradation induced by podosomes are more shallow and widespread
(Gimona et al., 2008; Linder, 2009). Finally, there are protein markers that are uniquely
found in invadopodia or podosomes. For example, Nck1 has been shown to be invadopodiaspecific, while vinculin has been shown to be podosome-specific (Chan et al., 2009; Oser et
al., 2011).
6.3 Formation and signal transduction of invadosomes
Initiation
The formation of invadopodia in cancer cells and the formation of podosomes in
untransformed cells can be induced by a variety of external stimuli. The most well
characterized category of stimuli is growth factors, which bind and activate their respective
receptors on the cell surface. In cancer cells, invadopodia formation can be promoted by
treatment with EGF (epidermal growth factor), TGF-β (transforming growth factor β), PDGF
22
(platelet-derived growth factor), HB-EGF (heparin-binding EGF), and HGF (hepatocyte
growth factor/scatter factor) (Díaz et al., 2013; Eckert et al., 2011; Hayes et al., 2013; Lucas
et al., 2010; Mader et al., 2011; Mandal et al., 2008; Pignatelli et al., 2012; Rajadurai et al.,
2012; Yamaguchi et al., 2005b). Similarly, in untransformed cells, podosomes can be
induced by treating various cell types with growth factors, such as macrophages with CSF-1
(colony stimulating factor-1), endothelial cells with TGF-β (transforming growth factor β), and
vascular smooth muscle cells with PDGF (platelet-derived growth factor) (Daubon et al.,
2011; Osiak et al., 2005; Quintavalle et al., 2010; Rottiers et al., 2009; Varon et al., 2006;
Wang et al., 2009; Wheeler et al., 2006). These growth factors could be secreted from tumor
cells themselves in an autocrine manner, or from tumor-associated stromal cells in a
paracrine fashion. For example, paracrine growth factor induction of
invadopodia/podosomes has been found in breast carcinoma cells and macrophages. The
two cell types form a feedback loop, in which macrophages secrete EGF to activate
invadopodia formation in breast cancer cells and promote tumor invasion, while the cancer
cells secrete CSF-1 to stimulate podosomes formation in macrophages, thus augmenting
the degradation of the extracellular matrix to further facilitate tumor invasion (Wyckoff et al.,
2004) .
Another category of stimuli for invadopodia and podosome formation is ECMdependent stimulation of the cell-surface integrin receptors. Evidence for a role of integrin in
invadopodia and podosome formation comes from observations that certain integrin
subtypes, such as β1 and β3, are found in invadopodia and podosomes. Genetic deletion of
β1 integrin inhibited the formation of invadopodia and podosomes (Destaing et al., 2010).
Similarly, inhibition of β3 integrin impaired podosome function in osteoclasts (Nakamura et
al., 1999). Although not much details are known to-date about the mechanism of integrinmediated induction of invadopodia formation, one possibility is through mechanosensing of
23
the rigidity of the ECM. Several studies have demonstrated that higher stiffness of the ECM
promotes the number and activity of invadopodia in breast and bladder cancer cells
(Alexander et al., 2008; Parekh et al., 2011). The signal of ECM rigidity is thought to be
transmitted from integrin to the actin cytoskeleton via the motor protein myosin II, and
integrin-associated mechanosensing proteins p130Cas and FAK (focal adhesion kinase)
(Alexander et al., 2008).
Assembly
Stimulation of growth factor receptors and integrins on the cell surface activates
canonical pathways that converge on the Src kinase, leading to the phosphorylation and
activation of downstream invadopodium/podosome-associated proteins. The end results are
actin polymerization and protrusion of the cellular membrane at the invadopodia/podosome
foci.
Among these phosphorylated invadosome components is a key adaptor protein,
Tks5, which plays a critical role in mediating the assembly of the invadosome machinery.
Tks5 is a cytoplasmic protein, and is recruited to the sites of early invadosome foci upon
phosphorylation by Src to mediate invadosome formation (Abram and Courtneidge, 2003).
Tks5 directly or indirectly recruits multiple adaptor proteins, including Nck1, Nck2, and
Cortactin (Crimaldi et al., 2009; Stylli et al., 2009). These adaptor proteins bind and activate
actin regulatory proteins WASP and the Arp2/3 complex, inducing actin polymerization
(Oikawa et al., 2008). Furthermore, Cortactin in its phosphorylated form also releases cofilin
from sequestration, thus allowing cofilin to sever actin filaments and generate free barbed
ends, thus further promoting actin polymerization (Oser et al., 2009). As a result of actin
filament growth at the invadopodia foci, the cell membrane protrudes into the extracellular
matrix.
24
Maturation
The maturation phase of invadosomes is characterized by stabilization of actin
filaments and secretion of proteases into the extracellular matrix. The stabilization of actin
filaments is mediated via several mechanisms. First, dephosphorylation of Cortactin allows it
to bind and sequester cofilin, thus preventing cofilin from further severing actin polymers
(Oser et al., 2009; Yamaguchi et al., 2005b). Second, the actin filaments at the tips of
invadosomes are bundled by actin-bundling proteins such as fascin and T-fimbrin, providing
stabilization of the actin core (Schoumacher et al., 2010). Third, microtubules and
intermediate filaments are inserted into invadosome protrusions, thus conferring additional
mechanical support for the structure (Kikuchi and Takahashi, 2008; Schoumacher et al.,
2010).
Proteases are delivered to the ECM by exocytosis along the microtubule network.
The types of proteases that are present at invadosomes include MMPs (e.g. MT1-MMP,
MMP2, MMP9), ADAMs, cysteine cathepsin proteases, and serine proteases (Murphy and
Courtneidge, 2011). The delivery of proteases to the invadosome foci can activate other
zymogens to further promote degradation of the extracellular matrix. For example, MT1MMP activates MMP2 by cleaving the N-terminal prodomain of pro-MMP2. Furthermore,
MT1-MMP and MMP2 can mediate a cleavage cascade that leads to the activation of
MMP9.
The consequences of targeting proteases to invadosomes are two fold. First, these
proteases allow invadosomes to degrade a wide range of basement membrane and ECM
components, including fibronectin, laminin, and collagen type I and IV (Kelly et al., 1994),
thus promoting invasion and migration through the physical barriers. Second, proteases
delivered to the invadosomes may also activate other growth factors to further stimulate
invadosome formation. For example, ADAM12 localized to invadosomes promotes the
25
ectodomain shedding of the growth factor HB-EGF (heparin-binding epidermal growth
factor), which in turn induces invadosome formation and cellular invasion (Díaz et al., 2013).
Disassembly
Efficient turnover of invadosomes are found to be required for efficient invasion (Goto
et al., 2002). Examination of actin dynamics showed that invadopodia and podosomes have
half-lives of hours and minutes, respectively. Although the regulatory mechanisms for the
turnover of invadopodia and podosomes are much less well characterized compared to their
formation, a few proteins have been identified to regulate their disassembly. For instance,
the actin-binding and crosslinking protein AFAP-110 has been implicated in promoting
invadosome disassembly in a phosphorylation-dependent manner, although the detailed
mechanism is not well understood (Dorfleutner et al., 2008). Another example is the
cysteine protease calpain, which has been shown to promote the turnover of invadosomes
by cleaving the invadosome components talin, Pyk2 and WASP (Calle et al., 2006).
Furthermore, phosphorylation of paxillin and activation of Erk are found to be required for
calpain-driven disassembly of invadosomes and efficient degradation of the ECM (Badowski
et al., 2008). Further studies are required to dissect the molecular mechanisms in regulating
the turnover of invadosomes.
6.4 Key components of invadosomes
Src
Src plays a central role in the formation of invadosomes. As a membrane-associated
non-receptor kinase, Src integrates stimulating signals from transmembrane growth factor
receptors and integrins, and is responsible for the tyrosine phosphorylation of a number of
26
downstream components of invadosomes including Tks5, Cortactin, p190RhoGAP, AFAP110, p130Cas, N-WASP, and paxillin (Bowden et al., 1999; Brábek et al., 2004; Gatesman
et al., 2004; Oser et al., 2009; Seals et al., 2005; Stylli et al., 2009; Yamaguchi et al.,
2005b). The phosphorylation is mediated either directly by Src or via other kinases such as
Arg (Mader et al., 2011).
The activity of Src is determined by its phosphorylation state and structural
conformation. The Src protein has four Src homology domains (SH1-4), with the SH1
domain containing the kinase activity. Src is inactive when its C-terminal tyrosine residue
(Tyr 530 in human, Tyr 529 in mouse) is phosphorylated. This phosphorylated tyrosine
induces a closed conformation of Src that is mediated by intramolecular interactions
between the C-terminal phosphotyrosine and the SH2 domain, and between the SH3
domain and the SH1 domain. This closed conformation prevents access of substrates to the
kinase pocket. Conversely, Src becomes active when the negative-regulatory tyrosine
residue is dephosphorylated and its activating tyrosine residue (Tyr 419 in human, Tyr 418
in mouse) is auto-phosphorylated by the SH1 kinase domain. This open conformation of Src
is able to bind and phosphorylate its substrates (Abram and Courtneidge, 2000).
The role of Src in invadosome formation was identified as early as when these
invasive protrusions were first reported in 1985. Src was found to localize to the sites of
podosomes in Rous sarcoma virus-transformed chicken embryonic fibroblasts (Chen et al.,
1985). Numerous studies have subsequently shown that the level of tyrosine
phosphorylation at invadopodia positively correlates with the degree of ECM degradation,
and that Src activation is required for the formation of podosomes and invadopodia in
various cell types (Artym et al., 2006; Balzer et al., 2010; Bowden et al., 2006; Mader et al.,
2011; Quintavalle et al., 2011). Overexpression of wild-type or activate mutants of Src
promoted invadosome formation and matrix degradation. Conversely, loss of Src activity by
27
overexpression of a dominant-negative form of Src, treatment with Src inhibitors, or
knockdown of Src by RNAi reduced invadopodia foci and activity. These observations
underscore the critical role of Src kinase in invadosome formation.
Tks5
Tks5 (formerly known as FISH for Five SH3 domain-containing protein) is an adaptor
protein that plays an important role in invadosome formation and function. Tks5 was initially
identified in a Src substrate screen using cDNA libraries, and was further characterized to be
a component of invadosome in Src-transformed fibroblasts (Lock et al., 1998). Tks5 has
multiple functional domains to mediate its action, including a phox homology (PX) domain in
the N-terminus, and five Src homology 3 (SH3) domains in the C-terminus.
Cytoplasmic Tks5 is recruited to the invadosome foci via its PX domain upon
phosphorylation by Src. Upon phosphorylation of Tks5, the PX domain of Tks5 is released
from intramolecular interaction and binds to phosphatidylinositol-3,4-bisphosphate
(PI(3,4)P2) on cell membrane, thereby localizing Tks5 to the site of invadosome initiation
(Abram and Courtneidge, 2003; Lock et al., 1998). Once localized to the cell membrane,
Tks5 recruits multiple downstream effector proteins of invadosomes, either via its SH3
domains or via other adaptor proteins, in order to mediate invadosome formation. The first
category of proteins recruited by Tks5 involves regulation of actin cytoskeleton. These are
adaptor proteins, including Nck1, Nck2, Grb2, and Cortactin, as well as the actin regulatory
proteins, N-WASP, the Arp2/3 complex, and p190RhoGAP (Crimaldi et al., 2009; Oikawa et
al., 2008; Stylli et al., 2009). N-WASP and the Arp2/3 complex promote actin filament
nucleation and branching, thus allowing invadosome formation. P190RhoGAP further
promotes invadosome formation by inducing local downregulation of RhoA activity and
subsequent dissolution of actin stress fiber and focal adhesion. Many of the these proteins,
28
including Cortactin and p190RhoGAP, are activated by Src phosphorylation, and Tks5 is
known to recruit the adaptor protein AFAP-110 to activate Src in invadosomes (Crimaldi et
al., 2009). Second, Tks5 has been shown to interact with NoxA1 and p22phox, two
components of the NADPH oxidase complex, to facilitate the production of reactive oxygen
species (ROS) by Nox enzymes at the invadosome foci (Diaz et al., 2009; Gianni et al.,
2010; 2009). ROS are found to promote invadosome formation by maintaining or amplifying
the tyrosine phosphorylation of Tks5, potentially by direct activation of Src or by inactivation
of the phosphatase PTP-PEST, which may in turn dephosphorylate Tks5. In this manner,
Tks5 can promote invadosome formation via ROS in a positive feedback loop. Finally, Tks5
has also been found to interact with ADAM family metalloproteases, including ADAMs 12,
15, 19 (Abram and Courtneidge, 2003). It is believed that Tks5 recruits theses proteases to
the invadosome foci to mediate degradation of the ECM and mediate the release of growth
factors to further stimulate invadosome activity.
Functional experiments have demonstrated an indispensible role for Tks5 in
invadosome formation in vitro and in vivo. Knockdown of Tks5 impaired invadosome
formation and ECM degradation in a variety of human cancer cells and untransformed cells
in culture (including melanoma, breast, and prostate cancers, as well as macrophages,
osteoclasts, and neurons), while overexpression of Tks5 promoted invadosome formation
(Burger et al., 2011; 2014; Oikawa et al., 2012; Santiago-Medina et al., 2015; Seals et al.,
2005). In vivo, knockdown of Tks5 has been demonstrated to reduce growth of Srctransformed fibroblasts in a transplantation setting (Blouw et al., 2008). Furthermore, Tks5
has been implicated in epithelial-to-mesenchymal transition, as knockdown of Tks5 in breast
cancer cells inhibited Twist-induced invadopodia formation in vitro and metastasis in vivo
(Eckert et al., 2011). Finally, germline genetic deletion of the Tks5-encoding gene Sh3pxd2a
29
in mice disrupted podosome-mediated migration of cranial neural crest cells in vivo, leading
to complete cleft of the secondary palate and neonatal death (Cejudo-Martin et al., 2014).
Interestingly, multiple isoforms of Tks5 exist, including a 150 kDa long isoform and
one or more 130-140 kDa short isoforms (Cejudo-Martin et al., 2014; Li et al., 2013; Lock et
al., 1998). Structurally, the two isoforms share the same C-terminal sequence, but differ in
the presence/ absence of the N-terminal PX domain, which is required for proper localization
of Tks5 to the cell membrane. This structural difference suggests that the two isoforms may
have different cellular localization and functions. Genetically, the long and short isoforms are
transcribed from independent promoters, as indicated by H3K4me3 chromatinimmunoprecipitation analysis of the promoter DNA and 5’RACE analysis of the isoform
transcripts, as will be discussed in Chapter 2, as well as by genetic deletion experiment in
mice targeting the long isoform without affecting expression of the short isoform (CejudoMartin et al., 2014). A recent study has shown that the expression of the long and short
forms of Tks5 can be regulated post-translationally by Src, as Src phosphorylation increases
the abundance of Tks5 long form, but induced degradation of the short form (Cejudo-Martin
et al., 2014). However, the exact mechanism of this isoform-specific regulation by Src
remains to be examined. More details of the functional difference and transcriptional
regulation of the long and short Tks5 isoforms will be discussed in Chapters 2 and 3.
The paralog of Tks5, called Tks4, is also required for invadosome activity in a
manner that partially overlaps with Tks5 (Buschman et al., 2009). Tks4 is structurally similar
to Tks5, containing a PX domain and four (instead of five) SH3 domains. It is a substrate of
Src phosphorylation. Similar to Tks5, Tks4 interacts with NoxA1 and p22phox to promote
ROS generation in invadopodia (Diaz et al., 2009; Gianni et al., 2009; 2010).
However, Tks4 also has a non-overlapping role with Tks5 in promoting the matrix
proteolysis activity of invadosome by mediating the localization of MT1-MMP to invadosome
30
foci. Silencing of Tks4 expression impaired invadosome formation and matrix degradation,
and introduction of Tks5 rescued the former defect but not the latter (Buschman et al.,
2009). Together, these studies showed that both Tks4 and Tks5 play important roles in
invadosome activity.
Cortactin
Cortactin localizes to invadosomes in cancer cells via recruitment by Tks5 (Artym et
al., 2006; Clark et al., 2007; Crimaldi et al., 2009), where it regulates actin polymerization
and protease secretion of invadosomes.
The actin-regulatory role of Cortactin is two-fold. First, tyrosine phosphorylation of
Cortactin at residues 421, 466 and 482 allows it to recruit actin-regulating complexes, such
as Nck1, N-WASp, and the Arp2/3 complex, to induce actin polymerization (Oser et al.,
2009; 2010; Tehrani et al., 2007). Second, Cortactin regulates the activity of the actinsevering protein cofilin (Magalhaes et al., 2011; Oser et al., 2009). Phosphorylated Cortactin
recruits to the invadosome foci a sodium-hydrogen exchanger NHE1, which locally
increases the pH to cause the release of cofilin from Cortactin. The released cofilin is thus
able to sever actin filaments, generating free barbed ends that promote actin polymerization
in the presence of excess G-actin monomers. Subsequent dephosphorylation of Cortactin,
potentially by the phosphatase PTP1B, allows it to re-sequester cofilin, leading to
stabilization of actin filaments in mature invadosomes.
In addition to regulating actin polymerization, other studies suggest a role for
Cortactin in promoting the secretion of matrix metalloproteases MT1-MMP, MMP2 and
MMP9 (Clark and Weaver, 2008; Clark et al., 2007). This is based on the observation that
Cortactin is required to target matrix metalloproteases to the actin puncta of invadopodia in
head and neck squamous cell carcinoma cells. Furthermore, knockdown of Cortactin in
31
these cells abolished the matrix degradation ability to an extent much greater than the
decrease in invadopodia foci formation, an effect similar to the inhibition of metalloproteases
by small molecules. Future studies are required to dissect the mechanism by which
Cortactin regulates protease secretion in invadosomes.
The tyrosine phosphorylation of Cortactin is mediated primarily by the protein
kinases Src and Arg. The kinase Src has been shown to promote phosphorylation of
Cortactin when overexpressed (Oser et al., 2009; Wu et al., 1991). The Arg kinase has also
been shown to be required for phosphorylation of Cortactin in fibroblasts (Boyle et al., 2007;
Lapetina et al., 2009). It has been proposed that Arg acts to directly phosphorylate Cortactin
under activation of Src (Sibony-Benyamini and Gil-Henn, 2012).
WASP proteins and the Arp2/3 complex
The WASP family proteins and the Arp2/3 complex are essential for actin filament
polymerization during invadosome formation. The WASP family of proteins consists of
WASP, which is expressed exclusively in hematopoietic cells, and N-WASP, which is nearly
ubiquitously expressed in all other cell types. The WASP proteins function as scaffolding
adaptors that bind and activate the Arp2/3 complex via its C-terminal VCA domain by
promoting interactions between the complex and actin (Insall and Machesky, 2009).
The Arp2/3 complex is an assembly of seven subunits, including ARPC1-5, and two
actin-related subunits, Arp2 and Arp3. Upon binding to N-WASP and actin filaments, the
Arp2/3 complex undergoes structural changes and adopts an active conformation, allowing
the Arp2 and Arp 3 subunits form an active dimer for nucleating actin filaments. The new
actin filaments are formed as branches that extend off the sides of preexisting filaments at
an angle of 70° (Insall and Machesky, 2009). Given that the roles of WASP and the Arp2/3
complex in actin polymerization, they are indispensible components of invadosomes.
32
Mena
Mena is a member of the Ena/VASP (Enabled/vasodilator-stimulated phosphoprotein) family, which regulates actin polymerization. It has several functional domains,
including the EVH1 domain, EVH2 domain, and a proline-rich core, which bind FP4
consensus motif, G- and F-actin, and profilin, respectively (Bachmann et al., 1999; Gertler et
al., 1996; Hüttelmaier et al., 1999; Krause et al., 2003). As an Ena/VASP protein, Mena
promotes actin filament elongation by binding to the barbed ends of actin filaments and
preventing them from being blocked by capping proteins (Barzik et al., 2005; Bear et al.,
2000; 2002). Furthermore, Mena also stabilizes actin filaments by bundling the filaments
and clustering the barbed ends (Applewhite et al., 2007; Bachmann et al., 1999; Barzik et
al., 2005).
Mena plays a critical role in regulating invadopodia formation in tumor cells,
potentially by promoting the elongation or stabilization of actin filaments. Mena was found to
co-localize with Cortactin and F-actin at invadopodia foci (Philippar et al., 2008). Expression
of Mena and a specific invasion-associated splice isoform MenaINV lengthened invadopodia
lifetime and promoted degradation activity of rat MTLn3 mammary tumor cells, and
increased the formation of micrometastases in the lungs (Philippar et al., 2008). In contrast,
loss of Mena expression in mouse mammary tumors inhibited their invasion into the
surrounding stroma, and reduced the number of circulating tumor cells and metastases
(Roussos et al., 2010). These data highlight the important role of Mena in regulating
invadopodia.
MT1-MMP
Among the proteases identified in invadopodia, the MMP family member MT1-MMP
(also known as MMP14) is most well characterized within the context of invadopodia. MT1-
33
MMP is a membrane-anchored metalloproteinase. It has been shown to cleave a wide
variety of ECM components in vitro, such as fibronectin, type I, II and III collagen, laminins,
vitronectin and aggrecans (d'Ortho et al., 1997; Fosang et al., 1998; Koshikawa et al., 2000;
Ohuchi et al., 1997). MT1-MMP accumulates at invadosome foci and is required for the
proteolytic function of invadosome, as knockdown of MT1-MMP strongly inhibited matrix
degradation, even though the initial stages of invadosome formation were not dramatically
affected (Artym et al., 2006; Nakahara et al., 1997).
MT1-MMP can be delivered to invadosome through multiple routes. First, nascently
translated MT1-MMP is expressed as a 64 kDa proMT1-MMP, and undergoes Furinmediated proteolytic cleavage in the Golgi apparatus to produce an enzymatically active 54kDa transmembrane fragment that is presented on the plasma membrane in invadosomes
(Mazzone et al., 2004; Sato et al., 1996; Yana and Weiss, 2000). Second, MT1-MMP can be
mobilized from intracellular storage compartments and delivered to invadosomes by
exocytosis via regulation of the Rab8 GTPase (Bravo-Cordero et al., 2007). Third, MT1MMP can be mobilized from other regions of the plasma membrane to the invadosome foci
by endocytic recycling. Membrane-associated MT1-MMP can be internalized by clathrinand caveolae-mediated endocytosis, and subsequently trafficked from endosomes to
invadosome foci via the regulation of N-WASP (Frittoli et al., 2011; Yu et al., 2012). Given its
localization to invadosomes and its ability to mediate matrix degradation, MT1-MMP is an
essential component of invadosomes.
34
6.5 The roles of podosomes in normal development and physiology
Embryonic development
Podosomes have been shown to play critical roles in mediating cell migration in the
developmental process. One example is the dorsal-ventral migration of neural crest cells
during embryogenesis. Using zebrafish as a model, Murphy et al. demonstrated that neural
crest cells produce podosome-like protrusions that enable them to migrate from the dorsal
neural tubes to the ventral compartment, where they differentiate into neurons, pigment
cells, as well as bone and connective tissues (Murphy et al., 2011). The formation and
function of these podosome-like protrusions are dependent on the podosome component
Tks5 and the Src kinase pathway. Inhibition of Tk5 expression or Src kinase activity in these
neural crest cells reduced podosome foci and cellular migration in vitro, and led to impaired
dorsal-ventral migration of neural crest cells as well as developmental defects in neural
crest-derived tissues in vivo. Consistent with the role of podosomes in mediating migration
of neural crest cells during embryogenesis, genetic disruption of the Tks5-encoding gene
Sh3pxd2a in mice led to complete cleft of the secondary palate and neonatal death (CejudoMartin et al., 2014). Additionally, in human, germline mutation or reduced expression of the
invadopodia component Tks4 (encoded by the gene Sh3pxd2b) is associated with two
highly similar disorders in development, the Frank-Ter Haar Syndrome and the Borrone
dermato-cardio-skeletal syndrome, both of which are characterized by craniofacial and other
skeletal abnormalities, as well as eye and heart defects, reminiscent of the effects of
defective neural crest cell migration in zebra fish (Iqbal et al., 2010; Wilson et al., 2014).
Furthermore, these developmental abnormalities were recapitulated in mice with germline
deletion of Tks4 (Iqbal et al., 2010). Together, these studies underscore the contribution of
podosomes in development.
35
Another example for podosomes in embryonic development comes from the axon
guidance of neuronal growth cones (Santiago-Medina et al., 2015). During development of
the nervous system, growth cones are responsible for guiding neurites to their proper
synaptic partners. A recent study on human and Xenopus neurons identified proteolytic
membrane protrusions at the growth cones that are highly similar to podosomes in terms of
their molecular components and degradative functions. These podosome-like structures
contain F-actin foci, Src, Tks5, Cortactin, N-WASP, Mena, and MMPs, and are capable of
degrading a variety of extracellular matrix components in a Src- and Tks5-dependent
manner. Disruption of these podosomes in developing embryos by targeting Tks5 inhibited
the motoneurons from exiting the spinal cord and extending into the periphery. Thus these
neuronal podosomes are proposed to create a passage for axonal outgrowth in the
developing nervous system by degrading the surrounding matrix.
Two additional examples for a role of podosomes in embryogenesis can be found in
invertebrates. In the formation of body wall muscle in Drosophila embryos, fusion-competent
myoblasts are found to produce podosome-like structures to mediate invasion into the
opposing muscle-founder cell (Sens et al., 2010). These protrusions are similar to
podosomes in terms of their morphology, size, dynamics, structural components, and
invasive nature. In addition, during the development of uterine-vulval attachment in C.
elegans larva, the anchor cell in the gonad forms a podosome-like structure that invades
through the gonadal and ventral basement membranes to penetrate the vulval epithelium
(Hagedorn et al., 2014; 2013). These protrusions share many similarities with podosomes in
terms of their structure, size and turnover rate. The presence of podosome-like structures in
lower organisms suggest that podosomes serve as an evolutionarily conserved mechanism
for the cellular invasion and migration processes during development.
36
Normal physiology
In adults, podosomes have also been shown to mediate normal physiological
functions of numerous cell types, including macrophages, dendritic cells, lymphocytes,
osteoclasts, and endothelial cells.
Podosomes are important for proper function of osteoclasts (Kanehisa et al., 1990;
Lakkakorpi and Väänänen, 1991; Sato et al., 1997; Teti et al., 1989; Zambonin-Zallone et
al., 1989; Zhang et al., 1995). Osteoclast podosomes are unique compared to podosomes in
other cell types, as they merge to form a superstructure called the sealing zone. This
podosome-based sealing zone allows osteoclasts to attach to the bone surface, and form an
enclosed lacuna into which protons and lytic enzymes are secreted to enable bone
remodeling. In animal experiments, osteoclasts derived from mice that harbored
homozygous germline mutations for the podosome component WASP showed defective
bone resorption in vitro on bone slices as well as in vivo in animals with bone damage (Calle
et al., 2004), demonstrating the essential role of podosomes in mediating proper bone
resorption activity of osteoclasts.
Immune cells, such as macrophages, dendritic cells, and lymphocytes, have also
been found to form podosomes (Burns et al., 2001; Calle et al., 2006; Carman et al., 2007;
Cougoule et al., 2010; Linder et al., 1999). In these cell types, podosomes have been
proposed to mediate a wide variety of functions, including matrix degradation, migration,
rigidity sensing, topography sensing, and antigen sampling (Baranov et al., 2014; Carman et
al., 2007; Gawden-Bone et al., 2010; Linder and Wiesner, 2015; Sage et al., 2012).
Interestingly, in patients with Wiskott–Aldrich syndrome, which is an X-linked recessive
disease caused by genetic mutations of the podosome component WASP, the dendritic cells
and macrophages are unable to form podosomes, and the patients suffer from severe
37
immune deficiencies. This observation provides indirect evidence that podosomes are
required for the proper function of immune cells {Binks:1998ve, Linder:1999vj}.
Finally, podosomes have also been observed in endothelial cells as large ring- or
crescent-shaped structures that are capable of degrading the extracellular matrix in vitro
(Osiak et al., 2005; Rottiers et al., 2009; Tatin et al., 2006; Varon et al., 2006). It has been
proposed that endothelial podosomes serve as adhesion structures for migration of
endothelial cells and provide a means to remodel the basement membrane during vessel
sprouting and vasculogenesis, but functional experiments in animals remain to be done to
support this hypothesis.
6.7 The roles of invadopodia in tumor invasion and metastasis
Invadopodia are thought to be aberrant derivatives of podosomes that tumor cells
have usurped for promoting cellular invasion and migration (Murphy and Courtneidge,
2011). Because of their ability to degrade the ECM, invadopodia have been proposed to
provide a mechanism for tumor cells to overcome the physical barriers presented by the
basement membrane, the interstitial matrix, and the endothelial cells during metastasis.
Invadopodia are thought to promote multiple steps of the metastatic cascade, including local
invasion into the stromal tissues at the primary tumor, intravasation into the vasculature,
extravasation at distant sites, and colonization of distant organs.
Numerous studies have presented evidence for a role of invadopodia in cancer cells
to mediate degradation of the ECM in vitro. Invadopodia with proteolytic activity have been
observed in a variety of tumor cells in culture, including melanoma, breast cancer, prostate
cancer, and head and neck squamous cell carcinoma (Burger et al., 2014; Clark et al., 2007;
Seals et al., 2005). Furthermore, cancer cells that are capable of forming invadopodia
demonstrated higher invasiveness in vitro (Coopman et al., 1998). Knockdown of
38
invadopodia components, such as Tks5 or Cortactin, in these cells inhibited invadopodia
formation and matrix degradation, while overexpression of these components had the
opposite effect.
There are also in vivo evidence from animal models supporting that tumor cells can
form invadopodia to promote invasion and metastasis. For example, Jing Yang and
colleagues have used a transplantation model of breast cancer to demonstrate that tumor
cells that have undergone Twist1-mediated epithelial-to-mesenchymal transition
metastasized to distant organs by forming invadopodia (Eckert et al., 2011). Mechanistically,
the authors demonstrated that Twist1 induced transcription of PDGFRα, which led to
activation of Src, thus inducing invadopodia formation. Efficient metastasis of these tumor
cells required invadopodia activity, as knockdown of the invadopodia component Tks5
inhibited metastasis to the lungs from subcutaneous tumors formed by transplanted cancer
cells. A second example for an in vivo role of invadopodia in metastasis comes from the
studies by John Condeelis and colleagues. Using multiphoton intravital imaging, the authors
detected invadopodia-like protrusions in breast carcinoma cells that extended from the
primary tumors and penetrated into blood vessels in a mouse xenograft model (Gligorijevic
et al., 2012; Yamaguchi et al., 2005a). Immunofluorescence staining of tissue sections
showed that these invasive protrusions were enriched in F-actin, Cortactin, and N-WASP,
three important components of invadopodia, and were capable of degrading collagen in the
tumor stroma. Furthermore, knockdown of N-WASP or overexpression of a dominantnegative N-WASP mutant inhibited the formation of these invasion protrusions, and reduced
the number of circulating tumor cells as well as lung metastases in the xenograft model. A
final example comes from a collaborative study between Sara Courtneidge, Ann Chambers,
John Lewis and colleagues. Using intravital imaging of a chorioallantoic membrane system,
which is a network of capillaries and stromal cells found in the chicken embryo, the authors
39
monitored the behavior of human tumor cells injected into the vasculature. They observed
that human breast cancer cells, fibrosarcoma cells, and epidermoid carcinoma cells formed
invadopodia protrusions when extravasating from the capillaries into the surrounding
stroma. These protrusions were enriched for invadopodia components, including Tks5,
Tks4, Cortactin, and MT1-MMP. Furthermore, knockdown of these components or inhibition
of Src by small molecules abrogated the invasive protrusions and diminished the rate of
extravasation. Collectively, these data strongly argue that invadopodia promote invasion and
metastasis of tumor cells in vivo.
In summary, invadopodia have been demonstrated to play significant roles in
metastasis in many cancer types. However their role in lung cancer specifically has not been
previously characterized. Furthermore, there is still a lot to be learned about the regulatory
mechanisms of the formation and function of invadopodia. In Chapter 2, I will present
evidence that alteration in the isoform expression of one of the key invadopodia
components, Tks5, can significantly affect invadopodia activity in lung adenocarcinoma
cells. Furthermore, I will provide data demonstrating that Tks5-mediated invadopodia activity
is required for promoting lung adenocarcinoma metastasis.
40
III. ROLES OF DEVELOPMENTAL TRANSCRIPTION FACTORS IN METASTASIS
Cancer is often considered an aberration of the normal developmental program. A
major piece of evidence for this argument is the observation that numerous transcription
factors that are critical for normal differentiation are dysregulated during cancer progression
and metastasis. Interestingly, the contributions of these developmental transcription factors
to cancer are diverse. While some developmental transcription factors inhibit tumor
progression, others promote this process. Here I will review the literature to highlight a few
transcription factors as examples. For the purpose of organization, I will divide these
transcription factors into two categories based on their effect on the differentiation state of
the tumors. The first category includes transcription factors whose expression alterations
correlate with dedifferentiation during tumor progression. This includes transcription factors
that are expressed in embryonic tissues and are upregulated in cancer, as well as those
factors that are expressed in adult tissues and are downregulated in cancer. In contrast, the
second category, called lineage survival oncogenes, includes transcription factors that are
expressed in differentiated adult tissues and, instead of being lost to promote
dedifferentiation in tumor, are further overexpressed in cancer cells to promote tumor
progression.
1. Dedifferentiation in tumor progression
Dedifferentiation is frequently associated with tumor progression. In the field of
surgical pathology, dedifferentiation is often observed in tumors, and histologically poorly
differentiated lesions are classified as high grade and are strongly correlated with disease
progression and poor prognosis. From the perspective of cell biology, many parallels are
found between tumor cells and stem cells. These similarities include unlimited proliferation,
41
self-renewal, resistance to apoptosis, and capacity for independent growth. In terms of gene
expression, transcriptome profiling analysis on a wide variety of tumor types, including lung
cancer, breast cancer, glioblastoma, and bladder carcinoma have revealed a striking
overlap between genes involved in developmental pathways and those that are altered in
cancer (Ben-Porath et al., 2008; Kho et al., 2004; Kopantzev et al., 2008; Liu et al., 2006).
These observations led to the hypothesis that many of the biological networks
involved in developmental organogenesis are also those that go awry in tumor initiation and
progression. In particular, transcription factors critical for normal development are
dysregulated in tumor progression. Such dysregulated expression of developmental
transcription factors can promote tumor dedifferentiation in two major ways. The first type
includes transcription factors whose expression is primarily restricted to embryonic tissues
and is not detected in differentiated tissues. Re-expression of these embryonal transcription
factors drives tissue dedifferentiation and promotes tumor development. The second type
includes transcription factors whose expression is induced during cell-fate specification and
is maintained in differentiated adult tissues, but is downregulated during natural tumor
progression. These tissue-specific transcription factors promote differentiation in normal
tissues, and loss of their expression favors tumor progression. Below I will review examples
for both types of transcription factors associated with tumor dedifferentiation.
1.1 Embryonal transcription factors that promote tumor progression
Numerous studies showed that developmental transcription factors that are primarily
expressed during early embryogenesis and subsequently downregulated in adult tissues
have an oncogenic role in tumor progression when re-activated in malignant cells. The first,
and perhaps the most extreme, example of embryonal transcription factor driving
tumorigenesis comes from the reprogramming factors Klf4, Oct4, Sox2, and c-Myc. In
42
chimeric mice generated using embryonic stem cells in which these reprogramming factors
were inducible under the control of doxycycline, continuous expression of these
reprogramming factors led to the development of teratomas in various organs, while
transient induction of these factors caused the development of undifferentiated dysplastic
cells that showed invasion into surrounding tissues and were distinct from teratomas (Abad
et al., 2013).
A second example in this category is the transcription factor Twist, which drives
epithelial-to-mesenchymal transition (EMT) in normal development and tumor metastasis.
Twist is a basic helix–loop–helix transcription factor. Its expression is induced during
embryonal morphogenesis of the cranial neural tube to convert neural crest cells from an
epithelial state to a mesenchymal state in order to mediate neural crest migration (Chen and
Behringer, 1995; Soo et al., 2002). In adults, Twist is not expressed in most epithelial
tissues, except the kidneys and pancreas (Castanon and Baylies, 2002). In epithelial
tumors, Twist, together with other EMT transcription factors, such as Snail, Slug, Zeb1 and
Zeb2, have been proposed to mediate EMT in tumor cells, leading to increased metastatic
migration and invasion. Because this change in cell state is thought to be transient and
reversible, clinical evidence of epithelial tumor cells being converted to a mesenchymal state
is scarce, and it remains controversial whether EMT actually occurs in metastasis in
patients. Nonetheless, functional studies in cell lines and animal models have demonstrated
that Twist can promote metastasis in vivo (Eckert et al., 2011; Yang et al., 2004).
Furthermore, activated expression of Twist has been detected in human tumors and is
associated with poor survival of patients (Caramel et al., 2013). These studies strongly
argue for a role for this embryonal transcription factor in promoting metastasis.
A third example for a reactivated embryonal gene during tumor progression is Pax2
in kidney cancer. In kidney organogenesis, the expression of Pax2, a paired-box
43
transcription factor, is restricted to the early phase of kidney development in condensing
kidney mesenchyme and its early epithelial derivatives. However, upon further development,
downregulation of Pax2 expression is required for differentiation into mature tubular
epithelium. Pax2 expression is silenced in normal kidney epithelial cells in adults (Dressler
et al., 1990; 1993; Torres et al., 1995). In contrast, in various kidney cancers, including
Wilms tumor, renal cell carcinoma and polycystic kidneys, Pax2 expression is reactivated
(Daniel et al., 2001; Dressler and Douglass, 1992; Gnarra and Dressler, 1995; Ostrom et al.,
2000). Knockdown of Pax2 in human renal cell carcinoma cells resulted in growth inhibition
(Gnarra and Dressler, 1995), while reducing Pax2 expression in a mouse model of
polycystic kidney disease induced apoptosis and impeded disease progression (Ostrom et
al., 2000). Collectively, these findings suggest that Pax2 reactivation in kidney tumors
promotes their progression by inducing an embryonic, undifferentiated, and proliferative
state.
A final example in this category is the gene Hsix1 in breast cancer. Hsix1 is a
homeobox gene expressed during embryonic development of the mammary glands, but is
lowly expressed or absent in adult mammary glands (Ford et al., 1998; Kobayashi et al.,
2008; Laclef et al., 2003; Xu et al., 2003). However, in breast cancer Hsix1 is frequently
overexpressed in both primary tumors and metastases (Ford et al., 1998). Increased
expression of Hsix1 is associated with poor prognosis of these patients (Iwanaga et al.,
2012; Micalizzi et al., 2009). Furthermore, exogenous overexpression of Hsix1 is sufficient
to induce malignant transformation of mammary cells in vitro and increase tumorigenicity in
vivo (Coletta et al., 2008). Mechanistically, Hsix1 has been shown to attenuate DNA
damage-induced G2 cell cycle checkpoint (Ford et al., 1998; 2000), upregulate expression
of the embryonic mammary gland-specific cyclin A1 (Coletta et al., 2008; 2004), and induce
expansion of a stem-like population within mammary carcinoma cells (Iwanaga et al., 2012).
44
Thus, overexpression of Hsix1 is thought to reestablish an embryonic state of proliferation to
promote breast tumorigenesis.
1.2 Lineage-specific transcription factors that suppress tumor progression
While reactivation of embryonal transcription factors can induce dedifferentiation in
tumors, another mechanism for dedifferentiation-mediated cancer progression is the loss of
expression of lineage-specific transcription factors. Numerous studies showed that
developmental transcription factors that are required to maintain adult tissue differentiation
have a suppressive role in cancer progression. One example is Nkx3-1. This homeoboxcontaining transcription factor regulates differentiation of the prostate epithelium (BhatiaGaur et al., 1999). The gene is normally expressed during prostate organogenesis to
mediate proper branching morphogenesis, glandular secretion and growth, and is
subsequently maintained in differentiated prostate tissues in adulthood. However, loss of
Nkx3-1 expression is frequently observed in human prostatic intraepithelial neoplasia and
prostate carcinomas, and correlates with tumor progression (Bowen et al., 2000).
Homozygous and heterozygous mutation of Nkx3-1 in mice was sufficient to induce
development of prostatic intraepithelial neoplasia, which are precursors to prostate
carcinoma (Abdulkadir et al., 2002; Bhatia-Gaur et al., 1999), while genetic inactivation of
Nkx3-1 was found to synergize with Pten loss-of-function mutation to accelerate
tumorigenesis in a mouse model of prostate carcinoma (Kim et al., 2002b). Furthermore,
overexpression of Nkx3-1 inhibited growth of prostate carcinoma cells (Kim et al., 2002a).
Collectively, these data demonstrate that Nkx3-1 is a suppressor of prostate tumor initiation
and progression.
A second example is Elf5 in breast cancer. In normal development, the Ets-domain
transcription factor Elf5 is required for alveolar morphogenesis of the mammary glands and
45
regulating the function of mammary stem cells (Chakrabarti et al., 2012b; Choi et al., 2009;
Oakes et al., 2008; Zhou et al., 2005). In human breast cancer, Elf5 expression is frequently
lost (Ma et al., 2003; Zhou et al., 1998), and is associated with poor prognosis of patients
(Chakrabarti et al., 2012a). Mechanistically, reduced Elf5 expression led to activation of the
EMT mediator Slug, induced EMT, and promoted metastasis in mouse models of breast
cancer (Chakrabarti et al., 2012a). Thus, loss of Elf5-mediated differentiation state in breast
cancer promotes progression of this tumor type.
2. Lineage survival oncogenes
While dysregulation of some developmental regulators discussed above can promote
dedifferentiation and tumor progression, there is also evidence from numerous studies that
support a different role for these lineage-specific factors in tumorigenesis, where gain of
expression of transcription factors that are required for terminal differentiation favors rather
than inhibits tumor progression. These transcription factors are termed “lineage survival
oncogenes”.
The prototype example of lineage survival oncogenes comes from studies of MITF in
melanoma. MITF (microphthalmia-associated transcription factor) encodes the master
transcription factor that regulates melanocyte survival and differentiation (Levy et al., 2006;
Opdecamp et al., 1997). In melanoma, decreased MITF expression generally correlates with
metastatic tumors and poor patient survival (Salti et al., 2000); however in about 10% of
primary melanomas and 15%-20% of metastases, MITF is genetically amplified and thus
overexpressed, and was associated with poor survival outcomes (Garraway et al., 2005).
Mechanistically, increased MITF expression cooperates with BRAF mutation to transform
immortalized melanocytes and promote survival of melanoma cells in the context of aberrant
MAPK pathway activation and cell-cycle deregulation (Garraway et al., 2005). Silencing of
46
MITF in cell lines that harbored copy number gain led to growth inhibition, suggesting that
MITF has a lineage survival function in melanoma (Garraway et al., 2005). Paradoxically, in
non-transformed melanocytes, MITF expression induced cell cycle arrest and differentiation
(Carreira et al., 2005; Loercher et al., 2005). This observation suggests that melanoma cells
have additional genetic or epigenetic alterations that allow MITF to induce proliferation in the
context of malignant cells.
A second example of lineage survival oncogene is ETV1 in gastrointestinal stromal
tumors (GIST). In normal intestine, ETV1 is required for the differentiation of ICC-MY cells
(myenteric interstitial cells of Cajal), which are cells that form a network between the circular
muscle and longitudinal muscle layers surrounding the neuronal myenteric plexus (Arber et
al., 2000). In the context of GIST, ETV1 is highly overexpressed in the tumor cells (Chi et
al., 2010; Zhang et al., 2014), and is found to be required for cell cycle progression and
survival of GIST cancer cells (Chi et al., 2010). Furthermore, ETV1 has been found to
cooperate with KIT for transformation and GIST development (Chi et al., 2010). Thus, these
data suggest that gain of ETV1 expression promotes the formation of GIST.
3. Context-dependent functions of Nkx2-1, Cdx2, and Foxa2
The effect of differentiation on tumor development is complex. While some
developmental transcription factors, such as the ones discussed above, fall neatly into
categories of tumor suppressive or oncogenic roles, other developmental factors have been
found to affect tumor progression in both ways in a context-dependent manner. Here we will
consider three examples: Nkx2-1, Cdx2, and Foxa2.
47
3.1 The roles of Nkx2-1 in lung adenocarcinoma
Nkx2-1 (also known as TTF1, TITF1, or T/EBP) is a homeobox-containing
transcription factor that has been studied extensively in the context of lung development and
lung adenocarcinoma. In normal embryonic development, Nkx2-1 expression is initially
detected in the ventral foregut endoderm and its derivative tracheal progenitor arising from
the lung primordium. Later in lung organogenesis, Nkx2-1 expression becomes
progressively restricted to distal airway cells (Stahlman et al., 1996; Yatabe et al., 2002).
Animal studies have demonstrated that Nkx2-1 expression is required for morphogenesis of
the distal lung (Yuan et al., 2000), but is dispensable for specification of the lung primordium
and for proximal lung morphogenesis (Minoo et al., 1999). Mice with germline homozygous
deletion of Nkx2-1 died immediately after birth due to hypoplastic lung development, as the
lungs exhibited severely impaired branching morphogenesis, lacked lung parenchyma, and
showed abnormal bronchial epithelium (Kimura et al., 1996). Mechanistically, Nkx2-1
transduces morphogenic signals (including fibroblast growth factors, Sonic hedgehog, and
bone morphogenetic proteins) from the surrounding mesenchyme into transcriptional
regulation of lung-specific genes, including Sftp-A, -B, -C, and CCSP (Bohinski et al., 1994;
Bruno et al., 1995; Kelly et al., 1996; Yan et al., 1995)
In lung adenocarcinoma, the role of Nkx2-1 is complex. On one hand, ample
evidence indicates that Nkx2-1 has a tumor suppressor role. Immunohistochemical analysis
of human lung adenocarcinoma showed that strong expression of Nkx2-1 was mostly
detected in well-differentiated tumors and less frequently found in poorly differentiated
lesions (Stenhouse et al., 2004). Patients with tumors expressing high levels of Nkx2-1 had
better prognosis for survival (Barletta et al., 2009; Berghmans et al., 2006). This tumor
suppressive role of Nkx2-1 is further supported by several lines of evidence from
genetically-engineered mouse models of lung adenocarcinoma. In a study by Kang et al,
48
Nkx2-1 expression was detected in a progressively decreasing pattern from wild-type lung
tissues to adenomas to adenocarcinoma in a Tgfβ1+/- mouse model treated with the
carcinogenic ethyl carbamate (Kang et al., 2004). Two recent genetic studies from our group
and Maeda et al. also demonstrated a tumor suppressive role of Nkx2-1 in mouse models
(Maeda et al., 2012; Snyder et al., 2013). Genetic deletion of Nkx2-1 in KrasG12D-driven lung
adenocarcinomas promoted primary tumor growth. Furthermore, overexpression of Nkx2-1
in KrasG12D-driven lung adenocarcinomas inhibited tumor initiation and growth. An additional
study from our group by Winslow et al. showed that Nkx2-1 suppresses metastasis of lung
adenocarcinoma. Nkx2-1 expression was consistently downregulated in metastatic primary
lung tumors and metastases in a KrasLSL-G12D/+; p53fl/fl conditional mouse model, and
silencing of Nkx2-1 expression by shRNA in non-metastatic cell lines promoted metastasis
(Winslow et al., 2011). The mechanisms by which Nkx2-1 suppresses metastasis is
multifaceted. First, Nkx2-1 can repress the expression of the embryonal proto-oncogene
Hmga2 via direct upregulation of miR-33a, a microRNA that binds to the 3’UTR of Hmga2
and inhibits Hmga2 expression (Rice et al., 2013; Winslow et al., 2011). Second, Nkx2-1
can activate the expression of a number of cell adhesion molecules, including E-cadherin,
Occludin, Claudin-1, and Claudin-18, which suppress cellular motility (Niimi et al., 2001;
Runkle et al., 2012; Saito et al., 2009). Third, Nkx2-1 also inhibits the expression of MYBPH
(myosin-binding protein H), which has been found to impair cellular migration by
suppressing actomyosin organization (Hosono et al., 2012). Finally, Nkx2-1 has been found
to repress epithelial-to-mesenchymal transition by reducing TGFβ-mediated induction of
Snail and Slug (Saito et al., 2009). Collectively, these data demonstrate that Nkx2-1
suppresses tumor progression and metastasis.
Paradoxically, other studies have argued that Nkx2-1 functions as a lineage-survival
oncogene in lung adenocarcinoma. Nkx2-1 is one of the genes in a 14q13.3 cytoband
49
amplification that has been found to be the most frequent focal amplification in lung cancer
that is not associated with a known lung oncogene (Barletta et al., 2009; Kwei et al., 2008;
Lee et al., 2013; Tanaka et al., 2007; Weir et al., 2007). Furthermore, these studies
identified that Nkx2-1 amplification was associated with poor prognosis among patients that
had Nkx2-1-expressing tumors. Tumor cells with Nkx2-1 amplification appeared to be reliant
or “addicted” to Nkx2-1 expression for proliferation, as RNAi-mediated knockdown of Nkx2-1
in these cells disrupted cell cycle progression and induced apoptosis (Kendall et al., 2007;
Kwei et al., 2008; Tanaka et al., 2007; Weir et al., 2007). Mechanistic studies demonstrated
that the pro-survival effect of Nkx2-1 is mediated at least in part through ROR1 (receptor
tyrosine kinase-like orphan receptor 1) and LMO3 (LIM domain only 3) (Watanabe et al.,
2013; Yamaguchi et al., 2012). The oncogenic role of Nkx2-1 in human lung
adenocarcinoma is corroborated by animal studies of EGFR-driven lung adenocarcinoma. In
contras to mutant Kras-driven lung adenocarcinoma, Nkx2-1 appeared to enhance
tumorigenesis of EGFR-driven lung adenocarcinoma, as EgfrL858R; Nkx2-1+/- mice
showed reduced lung tumor number and volume compared to EgfrL858R; Nkx2-1+/+ mice
(Maeda et al., 2012). Taken together, these data show that Nkx2-1 has an oncogenic role in
addition to a tumor suppressive role in lung adenocarcinoma.
3.2 The roles of Cdx2 in colorectal cancer
A second example for the complex roles of developmental transcription factors in
tumor progression comes from studies of Cdx2 in the context of intestinal development and
colorectal tumorigenesis. Cdx2 is a caudal type homeobox transcription factor. In normal
physiology, Cdx2 is expressed in the intestinal epithelium during embryonic development
and in adult intestines (James et al., 1994). It controls morphogenesis of intestinal cells
during development, and maintains the differentiated phenotype in adulthood by supporting
50
transcription of intestinal-specific genes (Guo et al., 2004). Conditional disruption of Cdx2 in
early endoderm led to grossly abnormal development of the intestine (Gao et al., 2009),
while acute ablation of Cdx2 in adult intestinal cells led to severe loss of intestinal
differentiation (Hryniuk et al., 2012).
In colorectal cancer, numerous studies have demonstrated that Cdx2 can assert a
tumor suppressive role in tumor development. Cdx2 expression is frequently reduced in
colorectal tumors, especially in high-grade, invasive, dedifferentiated carcinomas.
Furthermore, reduced Cdx2 expression was found to correlate with poor survival of patients
(Baba et al., 2009; EE et al., 1995; Kim et al., 2013; Mallo et al., 1997). Consistent with
these observations in human patients, animal studies have shown that heterozygous Cdx2+/mutation predisposed mice to develop adenomatous intestinal polyps, which notably had
complete loss of Cdx2 expression even though there was no loss of heterozygosity
(Chawengsaksophak et al., 1997). Heterozygous Cdx2 mutation also sensitized mice to
chemically induced colorectal cancer (Bonhomme et al., 2003), and was shown to cooperate
with ApcΔ716/+ mutation to accelerate the formation of colonic polyps (Aoki et al., 2003).
Finally, aberrant expression of Cdx2 in colorectal cancer cell lines was found to suppress
their proliferation (Mallo et al., 1998). Taken together, these studies argue that Cdx2
suppresses initiation and progression of colorectal cancer.
However, evidence from several studies argued that Cdx2 can also act as a lineage
survival oncogene in a subset of colorectal cancer (Douglas et al., 2004; Salari et al., 2012).
These studies found that Cdx2 is genomically amplified in 30%-50% of human colorectal
tumor samples. For these Cdx2 amplified cells, the Cdx2 protein is overexpressed, and
knockdown of Cdx2 induced apoptosis and inhibited cell-cycle progression, at least in part
via the Wnt/β-catenin signaling pathway. Thus, Cdx2 appears to be able to act in both
oncogenic and tumor suppressive manners in intestinal cancer.
51
3.3 The roles of Foxa2 in lung cancer and neuroendocrine prostate cancer
Foxa2 (also known as Hnf3β) is a forkhead box-containing transcription factor
expressed in early embryogenesis for formation of the node, notochord, and definitive
endoderm. Foxa2-/- mice are embryonic lethal (Ang et al., 1993; Dufort et al., 1998;
Weinstein et al., 1994). Foxa2 is also required for differentiation specification and mature
function of various endoderm-derived organs, including the lungs, stomach, intestine, liver,
pancreas, bladder, brain, and prostate (Besnard et al., 2004; Bochkis et al., 2008; Ferri et
al., 2007; Gao et al., 2008; 2007; Lantz et al., 2004; Lee et al., 2005; Lin et al., 2009;
Mirosevich et al., 2005). Here I will focus on the development of the lungs and prostate
specifically, as they are relevant for our subsequent discussion of the roles of Foxa2 in
tumor progression. For lung development, studies from mice bearing conditional mutations
of Fox-family proteins showed that Foxa2 cooperates with its paralog Foxa1 (also known as
Hnf3 α) in a redundant manner to induce the branching morphogenesis (Wan et al., 2005),
and is required for alveolarization (Wan et al., 2004; Zhou et al., 1997). In mature lungs,
Foxa2 has been shown to maintain lung differentiation by regulating the expression of lungspecific genes, including Nkx2-1, SftpB, and Scgblal (Bingle and Gitlin, 1993; Bingle et al.,
1995; Bohinski et al., 1994; Ikeda et al., 1996). In prostate development, Foxa2 is expressed
early during embryonic prostate epithelial bud formation, and later in a subpopulation of
basal neuroendocrine epithelial cells within the periurethral ducts of the adult prostate
(Mirosevich et al., 2005).
Mechanistically, Foxa2 and other Fox-family members are thought to differ from
classical transcription factors as they can also function as pioneer transcription factors and
transcription cofactors. As a classical transcription factor, Foxa2 can promote transcription of
its target genes by recruiting co-activators such as CBP/p300 that promote assembly of the
general transcriptional machinery and RNA polymerase II holoenzyme. Alternatively, Foxa2
52
can inhibit transcription by recruiting histone deacetylaces (Lam et al., 2013). As a pioneer
factor, Foxa2 is also able to open condense chromatin upon binding to forkhead response
elements and allow access for other transcription factors (Li et al., 2012a; Zaret et al., 2010).
Finally, Foxa2 can function as a cofactor to recruit other transcription factors such as
estrogen receptor-α (ERα) and androgen receptor (AR) to regulate transcription of target
genes (Li et al., 2012b).
In lung cancer, several lines of evidence suggest that Foxa2 has a suppressive role
in tumor development. Expression of Foxa2 was found to be silenced by promoter
hypermethylation in a subset of human lung adenocarcinoma and squamous cell
carcinomas, and decreased expression of Foxa2 was associated with poor survival of
patients (Basseres et al., 2012; Halmos et al., 2004). Loss of Foxa2 led to expression of
Slug, a major mediator of epithelial-to-mesenchymal transition, upon TGFβ1 stimulation, and
was shown to promote cellular invasion and migration (Tang et al., 2011). Furthermore,
forced expression of Foxa2 in a metastatic human lung adenocarcinoma cell line led to
proliferation arrest and apoptosis (Halmos et al., 2004). Besides lung cancer, similar
evidence for a tumor suppressive role for Foxa2 has been reported in pancreatic (Song et
al., 2010) and gastric cancer (Zhu et al., 2015).
In contrast to the above reports, studies on neuroendocrine prostate cancer
demonstrated an oncogenic role for Foxa2. Foxa2 expression was found to be highly
upregulated in primary human neuroendocrine carcinomas and metastases (Mirosevich et
al., 2006; Qi et al., 2010). In mouse models of neuroendocrine prostate carcinomas driven
by simian virus 40 large T antigen, Foxa2 was highly upregulated in the primary tumors and
metastases compared to untransformed prostate neuroendocrine cells (Hu et al., 2002;
Mirosevich et al., 2006). Furthermore, Foxa2 was shown to be capable of regulating
prostatic gene expression in a ligand and androgen receptor independent fashion,
53
suggesting that Foxa2 may play an important role in proliferation and the switch to androgen
independence growth of these tumor cells (Mirosevich et al., 2006). Consistent with these
observations, in a mouse model of neuroendocrine prostate cancer driven by HIF-1α, Foxa2
is required to transactivate a collection of hypoxia-responsive genes required for the
neuroendocrine phenotype and the metastatic propensity of this tumor type (Qi et al., 2010).
Taken together, these data support that Foxa2 plays an oncogenic role in neuroendocrine
prostate cancer.
3.4 Explaining the diverse roles of developmental transcription factors in cancer
It is curious and perhaps perplexing to consider that developmental transcription
factors can play such diverse, and even opposing, roles in tumor progression. Nonetheless,
this complexity may be better understood in the context of the metastatic cascade and
clonal selection. Cancer cells undergo constant selective pressure during tumor
progression, and those with properties that allow them to propagate through the metastatic
cascade will be selected for their dissemination and outgrowth. In this context, alterations in
the differentiation state of the tumor cells may confer selective advantages in different ways.
In some situations, loss of differentiation states may promote tumor metastasis, while in
other situations, increased expression of the differentiation transcription factors may benefit
tumor progression. One potential reason that loss of differentiation states may favor
metastasis is the gain of motility. Differentiated epithelial cells are held together by various
cell-to-cell and cell-to-basement adhesions. Loss of these adhesions through tumor
dedifferentiation can promote cellular migration and invasion. Another potential selective
advantage conferred by tumor dedifferentiation is the gain of stem-like properties. Normal
stem cells are known to have special characteristics such as self-renewal, resistance of
apoptosis, and independent growth. Thus by adopting a more stem-like state,
54
dedifferentiated tumor cells may be selected for adaptive survival after they are
disseminated to the foreign environment of a distant organ. Conversely, lineage survival
oncogenes may provide selective advantage for tumor progression in other contexts by
reinforcing the proliferation and survival signals that are programmed into tumor cells of the
specific lineage. In this context, the cellular mechanisms that promote lineage-specific
growth and survival during normal differentiation may be exploited by tumor cells to promote
tumor progression.
For those transcription factors that appear to have both tumor suppressive and
oncogenic role in tumor progression, such as the aforementioned transcription factors Nkx21, Cdx2, and Foxa2, whether their loss or gain of expression favors tumor progression is
likely context dependent. The most obvious determining factors is the tissue type that the
tumor arises in, as exemplified by the case of Foxa2 in lung cancer and neuroendocrine
prostate cancer discussed above. These two tissue types are likely different in their
signaling pathway, thus resulting in different effects of Foxa2 expression in tumor
progression. A second determining factor is the background of genetic mutations and the
activation state of signaling pathways in the cells. For example, loss of Nkx2-1 expression
has been found to promote progression of lung adenocarcinoma driven by Kras, but the
same Nkx2-1 expression alteration inhibits lung adenocarcinoma driven by EGFR (Maeda et
al., 2012; Snyder et al., 2013; Winslow et al., 2011). This discrepancy is likely due to
differences in the downstream pathways that are activated by Kras and EGFR mutations. A
third potential determining factor is the state of tumor progression. While early tumors may
benefit from the proliferation effect of a lineage survival oncogene, more advanced tumors
may have acquired additional genetic or epigenetic alterations to sustain proliferation and/or
avoid apoptosis, and become independent of the proliferation effect of the lineage survival
55
oncogene. In the latter situation, the benefits of increased motility and stemness upon loss
of differentiation may be selected for tumor progression and metastasis.
In summary, the roles of developmental transcription factors in tumor progression
and metastasis are diverse and complex. There is a lot to be learned about the specific
effect of developmental transcription factors in metastatic progression in different cancer
types. In Chapter 3, I will present data showing that loss of expression of the transcription
factors Nkx2-1, Foxa2, and Cdx2 in lung adenocarcinoma can lead to dedifferentiation of
tumor cells and promote progression to metastasis.
56
REFERENCES
Abad, M., Mosteiro, L., Pantoja, C., Cañamero, M., Rayon, T., Ors, I., Graña, O., Megías,
D., Domínguez, O., Martínez, D., et al. (2013). Reprogramming in vivo produces teratomas
and iPS cells with totipotency features. Nature 502, 340–345.
Abdulkadir, S.A., Magee, J.A., Peters, T.J., Kaleem, Z., Naughton, C.K., Humphrey, P.A.,
and Milbrandt, J. (2002). Conditional loss of Nkx3.1 in adult mice induces prostatic
intraepithelial neoplasia. Mol Cell Biol 22, 1495–1503.
Abram, C., and Courtneidge, S. (2003). The adaptor protein fish associates with members of
the ADAMs family and localizes to podosomes of Src-transformed cells. Jbc 278, 16844–
16851.
Abram, C., and Courtneidge, S. (2000). Src family tyrosine kinases and growth factor
signaling. Exp Cell Res 254, 1–13.
Aceto, N., Bardia, A., Miyamoto, D.T., Donaldson, M.C., Wittner, B.S., Spencer, J.A., Yu, M.,
Pely, A., Engstrom, A., Zhu, H., et al. (2014). Circulating tumor cell clusters are oligoclonal
precursors of breast cancer metastasis. Cell 158, 1110–1122.
Alexander, N.R., Branch, K.M., Parekh, A., Clark, E.S., Iwueke, I.C., Guelcher, S.A., and
Weaver, A.M. (2008). Extracellular matrix rigidity promotes invadopodia activity. Curr Biol
18, 1295–1299.
Ang, S.L., Wierda, A., Wong, D., Stevens, K.A., Cascio, S., Rossant, J., and Zaret, K.S.
(1993). The formation and maintenance of the definitive endoderm lineage in the mouse:
involvement of HNF3/forkhead proteins. Development 119, 1301–1315.
Aoki, K., Tamai, Y., Horiike, S., Oshima, M., and Taketo, M.M. (2003). Colonic polyposis
caused by mTOR-mediated chromosomal instability in Apc(+/Delta 716) Cdx(2+/-)
compound mutant mice. Nat Genet 35, 323–330.
Applewhite, D.A., Barzik, M., Kojima, S.-I., Svitkina, T.M., Gertler, F.B., and Borisy, G.G.
(2007). Ena/VASP proteins have an anti-capping independent function in filopodia
formation. Mol Biol Cell 18, 2579–2591.
Arber, S., Ladle, D.R., Lin, J.H., Frank, E., and Jessell, T.M. (2000). ETS gene Er81 controls
the formation of functional connections between group Ia sensory afferents and motor
neurons. Cell 101, 485–498.
Artym, V., Zhang, Y., Seillier-Moiseiwitsch, F., Yamada, K., and Mueller, S. (2006). Dynamic
interactions of cortactin and membrane type 1 matrix metalloproteinase at invadopodia:
Defining the stages of invadopodia formation and function. Cancer Res 66, 3034–3043.
Baba, Y., Nosho, K., Shima, K., Freed, E., Irahara, N., Philips, J., Meyerhardt, J.A., Hornick,
J.L., Shivdasani, R.A., Fuchs, C.S., et al. (2009). Relationship of CDX2 loss with molecular
features and prognosis in colorectal cancer. Clin Cancer Res 15, 4665–4673.
Bachmann, C., Fischer, L., Walter, U., and Reinhard, M. (1999). The EVH2 domain of the
57
vasodilator-stimulated phosphoprotein mediates tetramerization, F-actin binding, and actin
bundle formation. Jbc 274, 23549–23557.
Badowski, C., Pawlak, G., Grichine, A., Chabadel, A., Oddou, C., Jurdic, P., Pfaff, M.,
Albiges-Rizo, C., and Block, M.R. (2008). Paxillin phosphorylation controls
invadopodia/podosomes spatiotemporal organization. Mol Biol Cell 19, 633–645.
Balzer, E.M., Whipple, R.A., Thompson, K., Boggs, A.E., Slovic, J., Cho, E.H., Matrone,
M.A., Yoneda, T., Mueller, S.C., and Martin, S.S. (2010). c-Src differentially regulates the
functions of microtentacles and invadopodia. Oncogene 29, 6402–6408.
Baranov, M.V., Beest, Ter, M., Reinieren-Beeren, I., Cambi, A., Figdor, C.G., and van den
Bogaart, G. (2014). Podosomes of dendritic cells facilitate antigen sampling. J Cell Sci 127,
1052–1064.
Barletta, J.A., Perner, S., Iafrate, A.J., Yeap, B.Y., Weir, B.A., Johnson, L.A., Johnson, B.E.,
Meyerson, M., Rubin, M.A., Travis, W.D., et al. (2009). Clinical significance of TTF-1 protein
expression and TTF-1 gene amplification in lung adenocarcinoma. J Cell Mol Med 13,
1977–1986.
Barzik, M., Kotova, T.I., Higgs, H.N., Hazelwood, L., Hanein, D., Gertler, F.B., and Schafer,
D.A. (2005). Ena/VASP proteins enhance actin polymerization in the presence of barbed
end capping proteins. Jbc 280, 28653–28662.
Basseres, D.S., D'Alò, F., Yeap, B.Y., Loewenberg, E.C., Gonzalez, D.A., Yasuda, H.,
Dayaram, T., Kocher, O.N., Godleski, J.J., Richards, W.G., et al. (2012). Frequent
downregulation of the transcription factor Foxa2 in lung cancer through epigenetic silencing.
Lung Cancer 77, 31–37.
Bear, J.E., Loureiro, J.J., Libova, I., Fässler, R., Wehland, J., and Gertler, F.B. (2000).
Negative regulation of fibroblast motility by Ena/VASP proteins. Cell 101, 717–728.
Bear, J.E., Svitkina, T.M., Krause, M., Schafer, D.A., Loureiro, J.J., Strasser, G.A., Maly,
I.V., Chaga, O.Y., Cooper, J.A., Borisy, G.G., et al. (2002). Antagonism between Ena/VASP
proteins and actin filament capping regulates fibroblast motility. Cell 109, 509–521.
Ben-Baruch, A. (2008). Organ selectivity in metastasis: regulation by chemokines and their
receptors. Clin Exp Metastas 25, 345–356.
Ben-Porath, I., Thomson, M.W., Carey, V.J., Ge, R., Bell, G.W., Regev, A., and Weinberg,
R.A. (2008). An embryonic stem cell-like gene expression signature in poorly differentiated
aggressive human tumors. Nat Genet 40, 499–507.
Berghmans, T., Paesmans, M., Mascaux, C., Martin, B., Meert, A.-P., Haller, A., Lafitte, J.J., and Sculier, J.-P. (2006). Thyroid transcription factor 1--a new prognostic factor in lung
cancer: a meta-analysis. Ann Oncol 17, 1673–1676.
Besnard, V., Wert, S.E., Hull, W.M., and Whitsett, J.A. (2004). Immunohistochemical
localization of Foxa1 and Foxa2 in mouse embryos and adult tissues. Gene Expr Patterns 5,
193–208.
58
Bhatia-Gaur, R., Donjacour, A.A., Sciavolino, P.J., Kim, M., Desai, N., Young, P., Norton,
C.R., Gridley, T., Cardiff, R.D., Cunha, G.R., et al. (1999). Roles for Nkx3.1 in prostate
development and cancer. Gene Dev 13, 966–977.
Bingle, C.D., and Gitlin, J.D. (1993). Identification of hepatocyte nuclear factor-3 binding
sites in the Clara cell secretory protein gene. Biochem J 295 ( Pt 1), 227–232.
Bingle, C.D., Hackett, B.P., Moxley, M., Longmore, W., and Gitlin, J.D. (1995). Role of
hepatocyte nuclear factor-3 alpha and hepatocyte nuclear factor-3 beta in Clara cell
secretory protein gene expression in the bronchiolar epithelium. Biochem J 308 ( Pt 1), 197–
202.
Binks, M., Jones, G., Brickell, P., Kinnon, C., Katz, D., and Thrasher, A. (1998). Intrinsic
dendritic cell abnormalities in Wiskott-Aldrich syndrome. Eur J Immunol 28, 3259–3267.
Blouw, B., Seals, D.F., Pass, I., Diaz, B., and Courtneidge, S.A. (2008). A role for the
podosome/invadopodia scaffold protein Tks5 in tumor growth in vivo. Eur J Cell Biol 87,
555–567.
Bochkis, I.M., Rubins, N.E., White, P., Furth, E.E., Friedman, J.R., and Kaestner, K.H.
(2008). Hepatocyte-specific ablation of Foxa2 alters bile acid homeostasis and results in
endoplasmic reticulum stress. Nat Med 14, 828–836.
Bohinski, R.J., Di Lauro, R., and Whitsett, J.A. (1994). The lung-specific surfactant protein B
gene promoter is a target for thyroid transcription factor 1 and hepatocyte nuclear factor 3,
indicating common factors for organ-specific gene expression along the foregut axis. Mol
Cell Biol 14, 5671–5681.
Bonhomme, C., Duluc, I., Martin, E., Chawengsaksophak, K., Chenard, M.P., Kedinger, M.,
Beck, F., Freund, J.N., and Domon-Dell, C. (2003). The Cdx2 homeobox gene has a tumour
suppressor function in the distal colon in addition to a homeotic role during gut development.
Gut 52, 1465–1471.
Bos, P.D., Zhang, X.H.F., Nadal, C., Shu, W., Gomis, R.R., Nguyen, D.X., Minn, A.J., van de
Vijver, M.J., Gerald, W.L., Foekens, J.A., et al. (2009). Genes that mediate breast cancer
metastasis to the brain. Nature 459, 1005–1009.
Bowden, E., Barth, M., Thomas, D., Glazer, R., and Mueller, S. (1999). An invasion-related
complex of cortactin, paxillin and PKC mu associates with invadopodia at sites of
extracellular matrix degradation. Oncogene 18, 4440–4449.
Bowden, E., Onikoyi, E., Slack, R., Myoui, A., Yoneda, T., Yamada, K., and Mueller, S.
(2006). Co-localization of cortactin and phosphotyrosine identifies active invadopodia in
human breast cancer cells. Exp Cell Res 312, 1240–1253.
Bowen, C., Bubendorf, L., Voeller, H.J., Slack, R., Willi, N., Sauter, G., Gasser, T.C.,
Koivisto, P., Lack, E.E., Kononen, J., et al. (2000). Loss of NKX3.1 expression in human
prostate cancers correlates with tumor progression. Cancer Res 60, 6111–6115.
Boyle, S.N., Michaud, G.A., Schweitzer, B., Predki, P.F., and Koleske, A.J. (2007). A Critical
59
Role for Cortactin Phosphorylation by Abl-Family Kinases in PDGF-Induced Dorsal-Wave
Formation. Curr Biol 17, 445–451.
Bravo-Cordero, J.J., Marrero-Diaz, R., Megías, D., Genís, L., García-Grande, A., García,
M.A., Arroyo, A.G., and Montoya, M.C. (2007). MT1-MMP proinvasive activity is regulated
by a novel Rab8-dependent exocytic pathway. Embo J 26, 1499–1510.
Brábek, J., Constancio, S.S., Shin, N.-Y., Pozzi, A., Weaver, A.M., and Hanks, S.K. (2004).
CAS promotes invasiveness of Src-transformed cells. Oncogene 23, 7406–7415.
Bruno, M.D., Bohinski, R.J., Huelsman, K.M., Whitsett, J.A., and Korfhagen, T.R. (1995).
Lung cell-specific expression of the murine surfactant protein A (SP-A) gene is mediated by
interactions between the SP-A promoter and thyroid transcription factor-1. Jbc 270, 6531–
6536.
Buccione, R., Caldieri, G., and Ayala, I. (2009). Invadopodia: specialized tumor cell
structures for the focal degradation of the extracellular matrix. Cancer Metastasis Rev 28,
137–149.
Burger, K.L., Davis, A.L., Isom, S., Mishra, N., and Seals, D.F. (2011). The podosome
marker protein Tks5 regulates macrophage invasive behavior. Cytoskeleton (Hoboken).
Burger, K.L., Learman, B.S., Boucherle, A.K., Sirintrapun, S.J., Isom, S., Diaz, B.,
Courtneidge, S.A., and Seals, D.F. (2014). Src-dependent Tks5 phosphorylation regulates
invadopodia-associated invasion in prostate cancer cells. Prostate 74, 134–148.
Burns, S., Thrasher, A., Blundell, M., Machesky, L., and Jones, G. (2001). Configuration of
human dendritic cell cytoskeleton by Rho GTPases, the WAS protein, and differentiation.
Blood 98, 1142–1149.
Buschman, M.D., Bromann, P.A., Cejudo-Martin, P., Wen, F., Pass, I., and Courtneidge,
S.A. (2009). The Novel Adaptor Protein Tks4 (SH3PXD2B) Is Required for Functional
Podosome Formation. Mol Biol Cell 20, 1302–1311.
Calle, Y., Carragher, N.O., Thrasher, A.J., and Jones, G.E. (2006). Inhibition of calpain
stabilises podosomes and impairs dendritic cell motility. J Cell Sci 119, 2375–2385.
Calle, Y., Jones, G.E., Jagger, C., Fuller, K., Blundell, M.P., Chow, J., Chambers, T., and
Thrasher, A.J. (2004). WASp deficiency in mice results in failure to form osteoclast sealing
zones and defects in bone resorption. Blood 103, 3552–3561.
Caramel, J., Papadogeorgakis, E., Hill, L., Browne, G.J., Richard, G., Wierinckx, A.,
Saldanha, G., Osborne, J., Hutchinson, P., Tse, G., et al. (2013). A switch in the expression
of embryonic EMT-inducers drives the development of malignant melanoma. Cancer Cell
24, 466–480.
Carman, C.V., Sage, P.T., Sciuto, T.E., la Fuente, de, M.A., Geha, R.S., Ochs, H.D.,
Dvorak, H.F., Dvorak, A.M., and Springer, T.A. (2007). Transcellular diapedesis is initiated
by invasive podosomes. Immunity 26, 784–797.
60
Carmeliet, P., and Jain, R.K. (2011). Principles and mechanisms of vessel normalization for
cancer and other angiogenic diseases. Nat Rev Drug Discov 10, 417–427.
Carreira, S., Goodall, J., Aksan, I., La Rocca, S.A., Galibert, M.-D., Denat, L., Larue, L., and
Goding, C.R. (2005). Mitf cooperates with Rb1 and activates p21Cip1 expression to regulate
cell cycle progression. Nature 433, 764–769.
Castanon, I., and Baylies, M.K. (2002). A Twist in fate: evolutionary comparison of Twist
structure and function. Gene 287, 11–22.
Cejudo-Martin, P., Yuen, A., Vlahovich, N., Lock, P., Courtneidge, S.A., and Diaz, B. (2014).
Genetic Disruption of the Sh3pxd2a Gene Reveals an Essential Role in Mouse
Development and the Existence of a Novel Isoform of Tks5. PLoS ONE 9, e107674.
Chakrabarti, R., Hwang, J., Andres Blanco, M., Wei, Y., Lukačišin, M., Romano, R.-A.,
Smalley, K., Liu, S., Yang, Q., Ibrahim, T., et al. (2012a). Elf5 inhibits the epithelialmesenchymal transition in mammary gland development and breast cancer metastasis by
transcriptionally repressing Snail2. Nat Cell Biol 14, 1212–1222.
Chakrabarti, R., Wei, Y., Romano, R.-A., DeCoste, C., Kang, Y., and Sinha, S. (2012b). Elf5
regulates mammary gland stem/progenitor cell fate by influencing notch signaling. Stem
Cells 30, 1496–1508.
Chambers, A., Groom, A., and MacDonald, I. (2002). Dissemination and growth of cancer
cells in metastatic sites. Nat Rev Cancer 2, 563–572.
Chan, K.T., Cortesio, C.L., and Huttenlocher, A. (2009). FAK alters invadopodia and focal
adhesion composition and dynamics to regulate breast cancer invasion. J Cell Biol 185,
357–370.
Chawengsaksophak, K., James, R., Hammond, V.E., Köntgen, F., and Beck, F. (1997).
Homeosis and intestinal tumours in Cdx2 mutant mice. Nature 386, 84–87.
Chen, W. (1989). Proteolytic activity of specialized surface protrusions formed at rosette
contact sites of transformed-cells. J Exp Zool 251, 167–185.
Chen, W., Chen, J., Parsons, S., and Parsons, J. (1985). Local degradation of fibronectin at
sites of expression of the transforming gene-product pp60Src
. Nature 316, 156–158.
Chen, Z.F., and Behringer, R.R. (1995). twist is required in head mesenchyme for cranial
neural tube morphogenesis. Gene Dev 9, 686–699.
Chi, P., Chen, Y., Zhang, L., Guo, X., Wongvipat, J., Shamu, T., Fletcher, J.A., Dewell, S.,
Maki, R.G., Zheng, D., et al. (2010). ETV1 is a lineage survival factor that cooperates with
KIT in gastrointestinal stromal tumours. Nature 467, 849–853.
Choi, Y.S., Chakrabarti, R., Escamilla-Hernandez, R., and Sinha, S. (2009). Elf5 conditional
knockout mice reveal its role as a master regulator in mammary alveolar development:
61
failure of Stat5 activation and functional differentiation in the absence of Elf5. Dev Biol 329,
227–241.
Clark, E.S., and Weaver, A.M. (2008). A new role for cortactin in invadopodia: regulation of
protease secretion. Eur J Cell Biol 87, 581–590.
Clark, E.S., Whigham, A.S., Yarbrough, W.G., and Weaver, A.M. (2007). Cortactin is an
essential regulator of matrix metalloproteinase secretion and extracellular matrix
degradation in invadopodia. Cancer Res 67, 4227–4235.
Coletta, R.D., Christensen, K.L., Micalizzi, D.S., Jedlicka, P., Varella-Garcia, M., and Ford,
H.L. (2008). Six1 overexpression in mammary cells induces genomic instability and is
sufficient for malignant transformation. Cancer Res 68, 2204–2213.
Coletta, R.D., Christensen, K., Reichenberger, K.J., Lamb, J., Micomonaco, D., Huang, L.,
Wolf, D.M., Müller-Tidow, C., Golub, T.R., Kawakami, K., et al. (2004). The Six1
homeoprotein stimulates tumorigenesis by reactivation of cyclin A1. P Natl Acad Sci Usa
101, 6478–6483.
Coopman, P., Do, M., Thompson, E., and Mueller, S. (1998). Phagocytosis of cross-linked
gelatin matrix by human breast carcinoma cells correlates with their invasive capacity. Clin
Cancer Res 4, 507–515.
Cougoule, C., Le Cabec, V., Poincloux, R., Saati, Al, T., Mège, J.-L., Tabouret, G., Lowell,
C.A., Laviolette-Malirat, N., and Maridonneau-Parini, I. (2010). Three-dimensional migration
of macrophages requires Hck for podosome organization and extracellular matrix
proteolysis. Blood 115, 1444–1452.
Crimaldi, L., Courtneidge, S.A., and Gimona, M. (2009). Tks5 recruits AFAP-110,
p190RhoGAP, and cortactin for podosome formation. Exp Cell Res 315, 2581–2592.
d'Ortho, M.P., Will, H., Atkinson, S., Butler, G., Messent, A., Gavrilovic, J., Smith, B., Timpl,
R., Zardi, L., and Murphy, G. (1997). Membrane-type matrix metalloproteinases 1 and 2
exhibit broad-spectrum proteolytic capacities comparable to many matrix
metalloproteinases. Eur. J. Biochem. 250, 751–757.
Daniel, L., Lechevallier, E., Giorgi, R., Sichez, H., Zattara-Cannoni, H., Figarella-Branger,
D., and Coulange, C. (2001). Pax-2 expression in adult renal tumors. Hum. Pathol. 32, 282–
287.
Daubon, T., Buccione, R., and Génot, E. (2011). The Aarskog-Scott syndrome protein Fgd1
regulates podosome formation and extracellular matrix remodeling in transforming growth
factor β-stimulated aortic endothelial cells. Mol Cell Biol 31, 4430–4441.
David-Pfeuty, T., and Singer, S. (1980). Altered distributions of the cytoskeletal proteins
vinculin and alpha-actinin in cultured fibroblasts transformed by rous-sarcoma virus. P Natl
Acad Sci-Biol 77, 6687–6691.
Desgrosellier, J.S., and Cheresh, D.A. (2010). Integrins in cancer: biological implications
and therapeutic opportunities. Nat Rev Cancer 10, 9–22.
62
Destaing, O., Planus, E., Bouvard, D., Oddou, C., Badowski, C., Bossy, V., Raducanu, A.,
Fourcade, B., Albiges-Rizo, C., and Block, M.R. (2010). β1A integrin is a master regulator of
invadosome organization and function. Mol Biol Cell 21, 4108–4119.
Diaz, B., Shani, G., Pass, I., Anderson, D., Quintavalle, M., and Courtneidge, S.A. (2009).
Tks5-Dependent, Nox-Mediated Generation of Reactive Oxygen Species Is Necessary for
Invadopodia Formation. Sci Signal 2, ra53.
Díaz, B., Yuen, A., Iizuka, S., Higashiyama, S., and Courtneidge, S.A. (2013). Notch
increases the shedding of HB-EGF by ADAM12 to potentiate invadopodia formation in
hypoxia. J Cell Biol 201, 279–292.
Dorfleutner, A., Cho, Y., Vincent, D., Cunnick, J., Lin, H., Weed, S.A., Stehlik, C., and Flynn,
D.C. (2008). Phosphorylation of AFAP-110 affects podosome lifespan in A7r5 cells. J Cell
Sci 121, 2394–2405.
Douglas, E.J., Fiegler, H., Rowan, A., Halford, S., Bicknell, D.C., Bodmer, W., Tomlinson,
I.P.M., and Carter, N.P. (2004). Array comparative genomic hybridization analysis of
colorectal cancer cell lines and primary carcinomas. Cancer Res 64, 4817–4825.
Dressler, G.R., and Douglass, E.C. (1992). Pax-2 is a DNA-binding protein expressed in
embryonic kidney and Wilms tumor. P Natl Acad Sci Usa 89, 1179–1183.
Dressler, G.R., Deutsch, U., Chowdhury, K., Nornes, H.O., and Gruss, P. (1990). Pax2, a
new murine paired-box-containing gene and its expression in the developing excretory
system. Development 109, 787–795.
Dressler, G.R., Wilkinson, J.E., Rothenpieler, U.W., Patterson, L.T., Williams-Simons, L.,
and Westphal, H. (1993). Deregulation of Pax-2 expression in transgenic mice generates
severe kidney abnormalities. Nature 362, 65–67.
Dufort, D., Schwartz, L., HARPAL, K., and Rossant, J. (1998). The transcription factor
HNF3beta is required in visceral endoderm for normal primitive streak morphogenesis.
Development 125, 3015–3025.
Eckert, M.A., Lwin, T.M., Chang, A.T., Kim, J., Danis, E., Ohno-Machado, L., and Yang, J.
(2011). Twist1-induced invadopodia formation promotes tumor metastasis. Cancer Cell 19,
372–386.
EE, H.C., ERLER, T., BHATHAL, P.S., YOUNG, G.P., and JAMES, R.J. (1995). Cdx-2
Homeodomain Protein Expression in Human and Rat Colorectal Adenoma and Carcinoma.
Am J Pathol 147, 586–592.
Ferlay, J., Soerjomataram, I., Ervik, M., Dikshit, R., and Eser, S. (2014). GLOBOCAN 2012
v1. 0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11. Lyon, France:
International Agency for Research on Cancer; 2013. Available from: http://globocan.iarc.fr.
Ferri, A.L.M., Lin, W., Mavromatakis, Y.E., Wang, J.C., Sasaki, H., Whitsett, J.A., and Ang,
S.-L. (2007). Foxa1 and Foxa2 regulate multiple phases of midbrain dopaminergic neuron
development in a dosage-dependent manner. Development 134, 2761–2769.
63
Fidler, I.J. (1970). Metastasis: guantitative analysis of distribution and fate of tumor
embolilabeled with 125 I-5-iodo-2'-deoxyuridine. J Natl Cancer I 45, 773–782.
Fidler, I.J., and Kripke, M.L. (1977). Metastasis results from preexisting variant cells within a
malignant tumor. Science 197, 893–895.
Ford, H.L., Kabingu, E.N., Bump, E.A., Mutter, G.L., and Pardee, A.B. (1998). Abrogation of
the G2 cell cycle checkpoint associated with overexpression of HSIX1: a possible
mechanism of breast carcinogenesis. P Natl Acad Sci Usa 95, 12608–12613.
Ford, H.L., Landesman-Bollag, E., Dacwag, C.S., Stukenberg, P.T., Pardee, A.B., and
Seldin, D.C. (2000). Cell cycle-regulated phosphorylation of the human SIX1 homeodomain
protein. Jbc 275, 22245–22254.
Fosang, A.J., Last, K., Fujii, Y., Seiki, M., and Okada, Y. (1998). Membrane-type 1 MMP
(MMP-14) cleaves at three sites in the aggrecan interglobular domain. FEBS Lett 430, 186–
190.
Friedl, P., and Wolf, K. (2003). Tumour-cell invasion and migration: diversity and escape
mechanisms. Nat Rev Cancer 3, 362–374.
Friedl, P., Locker, J., Sahai, E., and Segall, J.E. (2012). Classifying collective cancer cell
invasion. Nat Cell Biol 14, 777–783.
Frittoli, E., Palamidessi, A., Disanza, A., and Scita, G. (2011). Secretory and endo/exocytic
trafficking in invadopodia formation: the MT1-MMP paradigm. Eur J Cell Biol 90, 108–114.
Gao, N., LeLay, J., Vatamaniuk, M.Z., Rieck, S., Friedman, J.R., and Kaestner, K.H. (2008).
Dynamic regulation of Pdx1 enhancers by Foxa1 and Foxa2 is essential for pancreas
development. Gene Dev 22, 3435–3448.
Gao, N., White, P., and Kaestner, K.H. (2009). Establishment of intestinal identity and
epithelial-mesenchymal signaling by Cdx2. Dev Cell 16, 588–599.
Gao, N., White, P., Doliba, N., Golson, M.L., Matschinsky, F.M., and Kaestner, K.H. (2007).
Foxa2 controls vesicle docking and insulin secretion in mature Beta cells. Cell Metab. 6,
267–279.
Garraway, L.A., Widlund, H.R., Rubin, M.A., Getz, G., Berger, A.J., Ramaswamy, S.,
Beroukhim, R., Milner, D.A., Granter, S.R., Du, J., et al. (2005). Integrative genomic
analyses identify MITF as a lineage survival oncogene amplified in malignant melanoma.
Nature 436, 117–122.
Gatesman, A., Walker, V., Baisden, J., Weed, S., and Flynn, D. (2004). Protein kinase C
alpha activates c-Src and induces podosome formation via AFAP-110. Mol Cell Biol 24,
7578–7597.
Gawden-Bone, C., Zhou, Z., King, E., Prescott, A., Watts, C., and Lucocq, J. (2010).
Dendritic cell podosomes are protrusive and invade the extracellular matrix using
metalloproteinase MMP-14. J Cell Sci 123, 1427–1437.
64
Gay, L.J., and Felding-Habermann, B. (2011). Contribution of platelets to tumour
metastasis. Nat Rev Cancer 11, 123–134.
Gertler, F.B., Niebuhr, K., Reinhard, M., Wehland, J., and Soriano, P. (1996). Mena, a
relative of VASP and Drosophila Enabled, is implicated in the control of microfilament
dynamics. Cell 87, 227–239.
Giancotti, F.G. (2013). Mechanisms governing metastatic dormancy and reactivation. Cell
155, 750–764.
Gianni, D., Dermardirossian, C., and Bokoch, G.M. (2010). Direct interaction between Tks
proteins and the N-terminal proline-rich region (PRR) of NoxA1 mediates Nox1-dependent
ROS generation. Eur J Cell Biol.
Gianni, D., Diaz, B., Taulet, N., Fowler, B., Courtneidge, S.A., and Bokoch, G.M. (2009).
Novel p47(phox)-Related Organizers Regulate Localized NADPH Oxidase 1 (Nox1) Activity.
Sci Signal 2, ra54.
Gimona, M., Buccione, R., Courtneidge, S.A., and Linder, S. (2008). Assembly and
biological role of podosomes and invadopodia. Curr Opin Cell Biol 20, 235–241.
Gligorijevic, B., Wyckoff, J., Yamaguchi, H., Wang, Y., Roussos, E.T., and Condeelis, J.
(2012). N-WASP-mediated invadopodium formation is involved in intravasation and lung
metastasis of mammary tumors. J Cell Sci 125, 724–734.
Gnarra, J.R., and Dressler, G.R. (1995). Expression of Pax-2 in human renal cell carcinoma
and growth inhibition by antisense oligonucleotides. Cancer Res 55, 4092–4098.
Goto, T., Maeda, H., and Tanaka, T. (2002). A selective inhibitor of matrix
metalloproteinases inhibits the migration of isolated osteoclasts by increasing the life span
of podosomes. J. Bone Miner. Metab. 20, 98–105.
Guo, R.J., Suh, E.R., and Lynch, J.P. (2004). The role of Cdx proteins in intestinal
development and cancer. Cancer Biol Ther 3, 593–601.
Gupta, G.P., and Massague, J. (2006). Cancer metastasis: Building a framework. Cell 127,
679–695.
Hagedorn, E.J., Kelley, L.C., Naegeli, K.M., Wang, Z., Chi, Q., and Sherwood, D.R. (2014).
ADF/cofilin promotes invadopodial membrane recycling during cell invasion in vivo. J Cell
Biol 204, 1209–1218.
Hagedorn, E.J., Ziel, J.W., Morrissey, M.A., Linden, L.M., Wang, Z., Chi, Q., Johnson, S.A.,
and Sherwood, D.R. (2013). The netrin receptor DCC focuses invadopodia-driven basement
membrane transmigration in vivo. J Cell Biol 201, 903–913.
Halmos, B., Basseres, D.S., Monti, S., D'Alò, F., Dayaram, T., Ferenczi, K., Wouters, B.J.,
Huettner, C.S., Golub, T.R., and Tenen, D.G. (2004). A transcriptional profiling study of
CCAAT/enhancer binding protein targets identifies hepatocyte nuclear factor 3 beta as a
novel tumor suppressor in lung cancer. Cancer Res 64, 4137–4147.
65
Hanahan, D., and Coussens, L.M. (2012). Accessories to the crime: functions of cells
recruited to the tumor microenvironment. Cancer Cell 21, 309–322.
Harley, B.A.C., Kim, H.-D., Zaman, M.H., Yannas, I.V., Lauffenburger, D.A., and Gibson,
L.J. (2008). Microarchitecture of three-dimensional scaffolds influences cell migration
behavior via junction interactions. Biophys. J. 95, 4013–4024.
Hayes, K.E., Walk, E.L., Ammer, A.G., Kelley, L.C., Martin, K.H., and Weed, S.A. (2013).
Ableson kinases negatively regulate invadopodia function and invasion in head and neck
squamous cell carcinoma by inhibiting an HB-EGF autocrine loop. Oncogene 32, 4766–
4777.
Hegerfeldt, Y., Tusch, M., Bröcker, E.-B., and Friedl, P. (2002). Collective cell movement in
primary melanoma explants: plasticity of cell-cell interaction, beta1-integrin function, and
migration strategies. Cancer Res 62, 2125–2130.
Hoover, H.C., and Ketcham, A.S. (1975). Metastasis of metastases. Am. J. Surg. 130, 405–
411.
Hosono, Y., Yamaguchi, T., Mizutani, E., Yanagisawa, K., Arima, C., Tomida, S., Shimada,
Y., Hiraoka, M., Kato, S., Yokoi, K., et al. (2012). MYBPH, a transcriptional target of TTF-1,
inhibits ROCK1, and reduces cell motility and metastasis. Embo J 31, 481–493.
Hryniuk, A., Grainger, S., Savory, J.G.A., and Lohnes, D. (2012). Cdx function is required for
maintenance of intestinal identity in the adult. Dev Biol 363, 426–437.
Hu, Y., Ippolito, J.E., Garabedian, E.M., Humphrey, P.A., and Gordon, J.I. (2002). Molecular
characterization of a metastatic neuroendocrine cell cancer arising in the prostates of
transgenic mice. Jbc 277, 44462–44474.
Hüttelmaier, S., Harbeck, B., Steffens, O., Messerschmidt, T., Illenberger, S., and Jockusch,
B.M. (1999). Characterization of the actin binding properties of the vasodilator-stimulated
phosphoprotein VASP. FEBS Lett 451, 68–74.
Ikeda, K., Shaw-White, J.R., Wert, S.E., and Whitsett, J.A. (1996). Hepatocyte nuclear factor
3 activates transcription of thyroid transcription factor 1 in respiratory epithelial cells. Mol
Cell Biol 16, 3626–3636.
Insall, R.H., and Machesky, L.M. (2009). Actin dynamics at the leading edge: from simple
machinery to complex networks. Dev Cell 17, 310–322.
Iqbal, Z., Cejudo-Martin, P., de Brouwer, A., van der Zwaag, B., Ruiz-Lozano, P., Scimia,
M.C., Lindsey, J.D., Weinreb, R., Albrecht, B., Megarbane, A., et al. (2010). Disruption of the
podosome adaptor protein TKS4 (SH3PXD2B) causes the skeletal dysplasia, eye, and
cardiac abnormalities of Frank-Ter Haar Syndrome. Am. J. Hum. Genet. 86, 254–261.
Iwanaga, R., Wang, C.-A., Micalizzi, D.S., Harrell, J.C., Jedlicka, P., Sartorius, C.A., Kabos,
P., Farabaugh, S.M., Bradford, A.P., and Ford, H.L. (2012). Expression of Six1 in luminal
breast cancers predicts poor prognosis and promotes increases in tumor initiating cells by
activation of extracellular signal-regulated kinase and transforming growth factor-beta
66
signaling pathways. Breast Cancer Res 14, R100.
James, R., ERLER, T., and Kazenwadel, J. (1994). Structure of the murine homeobox gene
cdx-2. Expression in embryonic and adult intestinal epithelium. Jbc 269, 15229–15237.
Joyce, J.A., and Pollard, J.W. (2009). Microenvironmental regulation of metastasis. Nat Rev
Cancer 9, 239–252.
Kanehisa, J., Yamanaka, T., Doi, S., Turksen, K., Heersche, J.N., Aubin, J.E., and Takeuchi,
H. (1990). A band of F-actin containing podosomes is involved in bone resorption by
osteoclasts. Bone 11, 287–293.
Kang, Y., Hebron, H., Ozbun, L., Mariano, J., Minoo, P., and Jakowlew, S.B. (2004). Nkx2.1
transcription factor in lung cells and a transforming growth factor-beta1 heterozygous mouse
model of lung carcinogenesis. Mol Carcinogen 40, 212–231.
Kang, Y., Siegel, P., Shu, W., Drobnjak, M., Kakonen, S., Cordon-Cardo, C., Guise, T., and
Massague, J. (2003). A multigenic program mediating breast cancer metastasis to bone.
Cancer Cell 3, 537–549.
Kaplan, R.N., Rafii, S., and Lyden, D. (2006). Preparing the “soil”: the premetastatic niche.
Cancer Res 66, 11089–11093.
Kelly, S.E., Bachurski, C.J., Burhans, M.S., and Glasser, S.W. (1996). Transcription of the
lung-specific surfactant protein C gene is mediated by thyroid transcription factor 1. Jbc 271,
6881–6888.
Kelly, T., Mueller, S., Yeh, Y., and Chen, W. (1994). Invadopodia promote proteolysis of a
wide variety of extracellular-matrix proteins. J Cell Physiol 158, 299–308.
Kendall, J., Liu, Q., Bakleh, A., Krasnitz, A., Nguyen, K.C.Q., Lakshmi, B., Gerald, W.L.,
Powers, S., and Mu, D. (2007). Oncogenic cooperation and coamplification of
developmental transcription factor genes in lung cancer. P Natl Acad Sci Usa 104, 16663–
16668.
Kho, A.T., Zhao, Q., Cai, Z., Butte, A.J., Kim, J.Y.H., Pomeroy, S.L., Rowitch, D.H., and
Kohane, I.S. (2004). Conserved mechanisms across development and tumorigenesis
revealed by a mouse development perspective of human cancers. Gene Dev 18, 629–640.
Kikuchi, K., and Takahashi, K. (2008). WAVE2- and microtubule-dependent formation of
long protrusions and invasion of cancer cells cultured on three-dimensional extracellular
matrices. Cancer Sci 99, 2252–2259.
Kim, J.H., Rhee, Y.-Y., Bae, J.M., Cho, N.-Y., and Kang, G.H. (2013). Loss of CDX2/CK20
expression is associated with poorly differentiated carcinoma, the CpG island methylator
phenotype, and adverse prognosis in microsatellite-unstable colorectal cancer. Am J Surg
Pathol 37, 1532–1541.
Kim, M.J., Bhatia-Gaur, R., Banach-Petrosky, W.A., Desai, N., Wang, Y., Hayward, S.W.,
Cunha, G.R., Cardiff, R.D., Shen, M.M., and Abate-Shen, C. (2002a). Nkx3.1 mutant mice
67
recapitulate early stages of prostate carcinogenesis. Cancer Res 62, 2999–3004.
Kim, M.J., Cardiff, R.D., Desai, N., Banach-Petrosky, W.A., Parsons, R., Shen, M.M., and
Abate-Shen, C. (2002b). Cooperativity of Nkx3.1 and Pten loss of function in a mouse model
of prostate carcinogenesis. P Natl Acad Sci Usa 99, 2884–2889.
Kimura, S., Hara, Y., Pineau, T., Fernandez-Salguero, P., Fox, C.H., Ward, J.M., and
Gonzalez, F.J. (1996). The T/ebp null mouse: thyroid-specific enhancer-binding protein is
essential for the organogenesis of the thyroid, lung, ventral forebrain, and pituitary. Gene
Dev 10, 60–69.
Kobayashi, A., Valerius, M.T., Mugford, J.W., Carroll, T.J., Self, M., Oliver, G., and
McMahon, A.P. (2008). Six2 defines and regulates a multipotent self-renewing nephron
progenitor population throughout mammalian kidney development. Cell Stem Cell 3, 169–
181.
Koizumi, K., Hojo, S., Akashi, T., Yasumoto, K., and Saiki, I. (2007). Chemokine receptors in
cancer metastasis and cancer cell-derived chemokines in host immune response. Cancer
Sci 98, 1652–1658.
Kopantzev, E.P., Monastyrskaya, G.S., Vinogradova, T.V., Zinovyeva, M.V., Kostina, M.B.,
Filyukova, O.B., Tonevitsky, A.G., Sukhikh, G.T., and Sverdlov, E.D. (2008). Differences in
gene expression levels between early and later stages of human lung development are
opposite to those between normal lung tissue and non-small lung cell carcinoma. Lung
Cancer 62, 23–34.
Koshikawa, N., Giannelli, G., Cirulli, V., Miyazaki, K., and Quaranta, V. (2000). Role of cell
surface metalloprotease MT1-MMP in epithelial cell migration over laminin-5. J Cell Biol 148,
615–624.
Krause, M., Dent, E.W., Bear, J.E., Loureiro, J.J., and Gertler, F.B. (2003). Ena/VASP
proteins: regulators of the actin cytoskeleton and cell migration. Annu Rev Cell Dev Bi 19,
541–564.
Kwei, K.A., Kim, Y.H., Girard, L., Kao, J., Pacyna-Gengelbach, M., Salari, K., Lee, J., Choi,
Y.-L., Sato, M., Wang, P., et al. (2008). Genomic profiling identifies TITF1 as a lineagespecific oncogene amplified in lung cancer. Oncogene 27, 3635–3640.
Laclef, C., Hamard, G., Demignon, J., Souil, E., Houbron, C., and Maire, P. (2003). Altered
myogenesis in Six1-deficient mice. Development 130, 2239–2252.
Lakkakorpi, P.T., and Väänänen, H.K. (1991). Kinetics of the osteoclast cytoskeleton during
the resorption cycle in vitro. J. Bone Miner. Res. 6, 817–826.
Lam, E.W.-F., Brosens, J.J., Gomes, A.R., and Koo, C.-Y. (2013). Forkhead box proteins:
tuning forks for transcriptional harmony. Nat Rev Cancer 13, 482–495.
Lantz, K.A., Vatamaniuk, M.Z., Brestelli, J.E., Friedman, J.R., Matschinsky, F.M., and
Kaestner, K.H. (2004). Foxa2 regulates multiple pathways of insulin secretion. J Clin Invest
114, 512–520.
68
Lapetina, S., Mader, C.C., Machida, K., Mayer, B.J., and Koleske, A.J. (2009). Arg interacts
with cortactin to promote adhesion-dependent cell edge protrusion. J Cell Biol 185, 503–
519.
Lazennec, G., and Richmond, A. (2010). Chemokines and chemokine receptors: new
insights into cancer-related inflammation. Trends Mol Med 16, 133–144.
Lee, C.S., Friedman, J.R., Fulmer, J.T., and Kaestner, K.H. (2005). The initiation of liver
development is dependent on Foxa transcription factors. Nature 435, 944–947.
Lee, J.S., Kim, H.R., Lee, C.Y., Shin, M., and Shim, H.S. (2013). EGFR and TTF-1 gene
amplification in surgically resected lung adenocarcinomas: clinicopathologic significance and
effect on response to EGFR-tyrosine kinase inhibitors in recurred cases. Ann. Surg. Oncol.
20, 3015–3022.
Levy, C., Khaled, M., and Fisher, D.E. (2006). MITF: master regulator of melanocyte
development and melanoma oncogene. Trends Mol Med 12, 406–414.
Li, C.M.-C., Chen, G., Dayton, T.L., Kim-Kiselak, C., Hoersch, S., Whittaker, C.A., Bronson,
R.T., Beer, D.G., Winslow, M.M., and Jacks, T. (2013). Differential Tks5 isoform expression
contributes to metastatic invasion of lung adenocarcinoma. Gene Dev 27, 1557–1567.
Li, Z., Gadue, P., Chen, K., Jiao, Y., Tuteja, G., Schug, J., Li, W., and Kaestner, K.H.
(2012a). Foxa2 and H2A.Z mediate nucleosome depletion during embryonic stem cell
differentiation. Cell 151, 1608–1616.
Li, Z., Tuteja, G., Schug, J., and Kaestner, K.H. (2012b). Foxa1 and Foxa2 are essential for
sexual dimorphism in liver cancer. Cell 148, 72–83.
Lin, W., Metzakopian, E., Mavromatakis, Y.E., Gao, N., Balaskas, N., Sasaki, H., Briscoe, J.,
Whitsett, J.A., Goulding, M., Kaestner, K.H., et al. (2009). Foxa1 and Foxa2 function both
upstream of and cooperatively with Lmx1a and Lmx1b in a feedforward loop promoting
mesodiencephalic dopaminergic neuron development. Dev Biol 333, 386–396.
Linder, S., Nelson, D., Weiss, M., and Aepfelbacher, M. (1999). Wiskott-Aldrich syndrome
protein regulates podosomes in primary human macrophages. P Natl Acad Sci Usa 96,
9648–9653.
Linder, S. (2007). The matrix corroded: podosomes and invadopodia in extracellular matrix
degradation. Trends Cell Biol 17, 107–117.
Linder, S. (2009). Invadosomes at a glance. J Cell Sci 122, 3009–3013.
Linder, S., and Wiesner, C. (2015). Tools of the trade: podosomes as multipurpose
organelles of monocytic cells. Cell Mol Life Sci 72, 121–135.
Liu, H., Kho, A.T., Kohane, I.S., and Sun, Y. (2006). Predicting survival within the lung
cancer histopathological hierarchy using a multi-scale genomic model of development. PLoS
Med. 3, e232.
69
Lock, P., Abram, C., Gibson, T., and Courtneidge, S. (1998). A new method for isolating
tyrosine kinase substrates used to identify Fish, an SH3 and PX domain-containing protein,
and Src substrate. Embo J 17, 4346–4357.
Loercher, A.E., Tank, E.M.H., Delston, R.B., and Harbour, J.W. (2005). MITF links
differentiation with cell cycle arrest in melanocytes by transcriptional activation of INK4A. J
Cell Biol 168, 35–40.
Lucas, J.T., Salimath, B.P., Slomiany, M.G., and Rosenzweig, S.A. (2010). Regulation of
invasive behavior by vascular endothelial growth factor is HEF1-dependent. Oncogene 29,
4449–4459.
Ma, X.-J., Salunga, R., Tuggle, J.T., Gaudet, J., Enright, E., McQuary, P., Payette, T.,
Pistone, M., Stecker, K., Zhang, B.M., et al. (2003). Gene expression profiles of human
breast cancer progression. P Natl Acad Sci Usa 100, 5974–5979.
Mader, C.C., Oser, M., Magalhaes, M.A.O., Bravo-Cordero, J.J., Condeelis, J., Koleske,
A.J., and Gil-Henn, H. (2011). An EGFR-Src-Arg-cortactin pathway mediates functional
maturation of invadopodia and breast cancer cell invasion. Cancer Res 71, 1730–1741.
Maeda, Y., Tsuchiya, T., Hao, H., Tompkins, D.H., Xu, Y., Mucenski, M.L., Du, L., Keiser,
A.R., Fukazawa, T., Naomoto, Y., et al. (2012). KrasG12D and Nkx2-1 haploinsufficiency
induce mucinous adenocarcinoma of the lung
. J Clin Invest 122, 4388–4400.
Magalhaes, M.A.O., Larson, D.R., Mader, C.C., Bravo-Cordero, J.J., Gil-Henn, H., Oser, M.,
Chen, X., Koleske, A.J., and Condeelis, J. (2011). Cortactin phosphorylation regulates cell
invasion through a pH-dependent pathway. J Cell Biol 195, 903–920.
Mallo, G.V., Reckreche, H., Frigerio, J.M., Rocha, D., Zweibaum, A., Lacasa, M., Jordan,
B.R., Dusetti, N.J., Dagorn, J.C., and Iovanna, J.L. (1997). Molecular cloning, sequencing
and expression of the mRNA encoding human Cdx1 and Cdx2 homeobox. Down-regulation
of Cdx1 and Cdx2 mRNA expression during colorectal carcinogenesis. Int J Cancer 74, 35–
44.
Mallo, G.V., Soubeyran, P., Lissitzky, J.C., André, F., Farnarier, C., Marvaldi, J., Dagorn,
J.C., and Iovanna, J.L. (1998). Expression of the Cdx1 and Cdx2 homeotic genes leads to
reduced malignancy in colon cancer-derived cells. Jbc 273, 14030–14036.
Mandal, S., Johnson, K.R., and Wheelock, M.J. (2008). TGF-beta induces formation of Factin cores and matrix degradation in human breast cancer cells via distinct signaling
pathways. Exp Cell Res 314, 3478–3493.
Mason, S.D., and Joyce, J.A. (2011). Proteolytic networks in cancer. Trends Cell Biol 21,
228–237.
Mazzone, M., Baldassarre, M., Beznoussenko, G., Giacchetti, G., Cao, J., Zucker, S., Luini,
A., and Buccione, R. (2004). Intracellular processing and activation of membrane type 1
matrix metalloprotease depends on its partitioning into lipid domains. J Cell Sci 117, 6275–
70
6287.
Micalizzi, D.S., Christensen, K.L., Jedlicka, P., Coletta, R.D., Barón, A.E., Harrell, J.C.,
Horwitz, K.B., Billheimer, D., Heichman, K.A., Welm, A.L., et al. (2009). The Six1
homeoprotein induces human mammary carcinoma cells to undergo epithelial-mesenchymal
transition and metastasis in mice through increasing TGF-β signaling. J Clin Invest 119,
2678–2690.
Minn, A., Gupta, G., Siegel, P., Bos, P., Shu, W., Giri, D., Viale, A., Olshen, A., Gerald, W.,
and Massague, J. (2005). Genes that mediate breast cancer metastasis to lung. Nature 436,
518–524.
Mirosevich, J., Gao, N., and Matusik, R.J. (2005). Expression of Foxa transcription factors in
the developing and adult murine prostate. Prostate 62, 339–352.
Mirosevich, J., Gao, N., Gupta, A., Shappell, S.B., Jove, R., and Matusik, R.J. (2006).
Expression and role of Foxa proteins in prostate cancer. Prostate 66, 1013–1028.
Murphy, D.A., and Courtneidge, S.A. (2011). The 'ins' and “outs” of podosomes and
invadopodia: characteristics, formation and function. Nat Rev Mol Cell Bio 12, 413–426.
Murphy, D.A., Diaz, B., Bromann, P.A., Tsai, J.H., Kawakami, Y., Maurer, J., Stewart, R.A.,
Izpisúa-Belmonte, J.C., and Courtneidge, S.A. (2011). A Src-Tks5 Pathway Is Required for
Neural Crest Cell Migration during Embryonic Development. PLoS ONE 6, e22499.
Müller, A., Homey, B., Soto, H., Ge, N., Catron, D., Buchanan, M.E., McClanahan, T.,
Murphy, E., Yuan, W., Wagner, S.N., et al. (2001). Involvement of chemokine receptors in
breast cancer metastasis. Nature 410, 50–56.
Nakahara, H., Howard, L., Thompson, E., Sato, H., Seiki, M., Yeh, Y., and Chen, W. (1997).
Transmembrane/cytoplasmic domain-mediated membrane type 1-matrix metalloprotease
docking to invadopodia is required for cell invasion. P Natl Acad Sci Usa 94, 7959–7964.
Nakamura, I., Pilkington, M.F., Lakkakorpi, P.T., Lipfert, L., Sims, S.M., Dixon, S.J., Rodan,
G.A., and Duong, L.T. (1999). Role of alpha(v)beta(3) integrin in osteoclast migration and
formation of the sealing zone. J Cell Sci 112 ( Pt 22), 3985–3993.
Niimi, T., Nagashima, K., Ward, J.M., Minoo, P., Zimonjic, D.B., Popescu, N.C., and Kimura,
S. (2001). claudin-18, a novel downstream target gene for the T/EBP/NKX2.1 homeodomain
transcription factor, encodes lung- and stomach-specific isoforms through alternative
splicing. Mol Cell Biol 21, 7380–7390.
Oakes, S.R., Naylor, M.J., Asselin-Labat, M.-L., Blazek, K.D., Gardiner-Garden, M., Hilton,
H.N., Kazlauskas, M., Pritchard, M.A., Chodosh, L.A., Pfeffer, P.L., et al. (2008). The Ets
transcription factor Elf5 specifies mammary alveolar cell fate. Gene Dev 22, 581–586.
Ohuchi, E., Imai, K., Fujii, Y., Sato, H., Seiki, M., and Okada, Y. (1997). Membrane type 1
matrix metalloproteinase digests interstitial collagens and other extracellular matrix
macromolecules. Jbc 272, 2446–2451.
71
Oikawa, T., Itoh, T., and Takenawa, T. (2008). Sequential signals toward podosome
formation in NIH-src cells. J Cell Biol 182, 157–169.
Oikawa, T., Oyama, M., Kozuka-Hata, H., Uehara, S., Udagawa, N., Saya, H., and Matsuo,
K. (2012). Tks5-dependent formation of circumferential podosomes/invadopodia mediates
cell-cell fusion. J Cell Biol 197, 553–568.
Opdecamp, K., Nakayama, A., Nguyen, M.T., Hodgkinson, C.A., Pavan, W.J., and
Arnheiter, H. (1997). Melanocyte development in vivo and in neural crest cell cultures:
crucial dependence on the Mitf basic-helix-loop-helix-zipper transcription factor.
Development 124, 2377–2386.
Oser, M., Dovas, A., Cox, D., and Condeelis, J. (2011). Nck1 and Grb2 localization patterns
can distinguish invadopodia from podosomes. Eur J Cell Biol 90, 181–188.
Oser, M., Mader, C.C., Gil-Henn, H., Magalhaes, M., Bravo-Cordero, J.J., Koleske, A.J., and
Condeelis, J. (2010). Specific tyrosine phosphorylation sites on cortactin regulate Nck1dependent actin polymerization in invadopodia. J Cell Sci 123, 3662–3673.
Oser, M., Yamaguchi, H., Mader, C.C., Bravo-Cordero, J.J., Arias, M., Chen, X., DesMarais,
V., van Rheenen, J., Koleske, A.J., and Condeelis, J. (2009). Cortactin regulates cofilin and
N-WASp activities to control the stages of invadopodium assembly and maturation. J Cell
Biol 186, 571–587.
Osiak, A.-E., Zenner, G., and Linder, S. (2005). Subconfluent endothelial cells form
podosomes downstream of cytokine and RhoGTPase signaling. Exp Cell Res 307, 342–
353.
Ostrom, L., Tang, M.J., Gruss, P., and Dressler, G.R. (2000). Reduced Pax2 gene dosage
increases apoptosis and slows the progression of renal cystic disease. Dev Biol 219, 250–
258.
Paget, S. (1989). The distribution of secondary growths in cancer of the breast. 1889.
Parekh, A., Ruppender, N.S., Branch, K.M., Sewell-Loftin, M.K., Lin, J., Boyer, P.D.,
Candiello, J.E., Merryman, W.D., Guelcher, S.A., and Weaver, A.M. (2011). Sensing and
modulation of invadopodia across a wide range of rigidities. Biophys. J. 100, 573–582.
Petrie, R.J., Doyle, A.D., and Yamada, K.M. (2009). Random versus directionally persistent
cell migration. Nat Rev Mol Cell Bio 10, 538–549.
Peyton, S.R., Kim, P.D., Ghajar, C.M., Seliktar, D., and Putnam, A.J. (2008). The effects of
matrix stiffness and RhoA on the phenotypic plasticity of smooth muscle cells in a 3-D
biosynthetic hydrogel system. Biomaterials 29, 2597–2607.
Philippar, U., Roussos, E.T., Oser, M., Yamaguchi, H., Kim, H.-D., Giampieri, S., Wang, Y.,
Goswami, S., Wyckoff, J.B., Lauffenburger, D.A., et al. (2008). A Mena Invasion Isoform
Potentiates EGF-Induced Carcinoma Cell Invasion and Metastasis. Dev Cell 15, 813–828.
Pignatelli, J., Tumbarello, D.A., Schmidt, R.P., and Turner, C.E. (2012). Hic-5 promotes
72
invadopodia formation and invasion during TGF-β-induced epithelial-mesenchymal
transition. J Cell Biol 197, 421–437.
Provenzano, P.P., Inman, D.R., Eliceiri, K.W., Trier, S.M., and Keely, P.J. (2008). Contact
guidance mediated three-dimensional cell migration is regulated by Rho/ROCK-dependent
matrix reorganization. Biophys. J. 95, 5374–5384.
Qi, J., Nakayama, K., Cardiff, R.D., Borowsky, A.D., Kaul, K., Williams, R., Krajewski, S.,
Mercola, D., Carpenter, P.M., Bowtell, D., et al. (2010). Siah2-Dependent Concerted Activity
of HIF and FoxA2 Regulates Formation of Neuroendocrine Phenotype and Neuroendocrine
Prostate Tumors. Cancer Cell 18, 23–38.
Quintavalle, M., Elia, L., Condorelli, G., and Courtneidge, S.A. (2010). MicroRNA control of
podosome formation in vascular smooth muscle cells in vivo and in vitro. J Cell Biol 189,
13–22.
Quintavalle, M., Elia, L., Price, J.H., Heynen-Genel, S., and Courtneidge, S.A. (2011). A cellbased high-content screening assay reveals activators and inhibitors of cancer cell invasion.
Sci Signal 4, ra49.
Rajadurai, C.V., Havrylov, S., Zaoui, K., Vaillancourt, R., Stuible, M., Naujokas, M., Zuo, D.,
Tremblay, M.L., and Park, M. (2012). Met receptor tyrosine kinase signals through a
cortactin-Gab1 scaffold complex, to mediate invadopodia. J Cell Sci 125, 2940–2953.
Reya, T., Morrison, S.J., Clarke, M.F., and Weissman, I.L. (2001). Stem cells, cancer, and
cancer stem cells. Nature 414, 105–111.
Rice, S.J., Lai, S.-C., Wood, L.W., Helsley, K.R., Runkle, E.A., Winslow, M.M., and Mu, D.
(2013). MicroRNA-33a mediates the regulation of high mobility group AT-hook 2 gene
(HMGA2) by thyroid transcription factor 1 (TTF-1/NKX2-1). Jbc 288, 16348–16360.
Ridley, A.J. (2011). Life at the leading edge. Cell 145, 1012–1022.
Rottiers, P., Saltel, F., Daubon, T., Chaigne-Delalande, B., Tridon, V., Billottet, C., Reuzeau,
E., and Génot, E. (2009). TGFbeta-induced endothelial podosomes mediate basement
membrane collagen degradation in arterial vessels. J Cell Sci 122, 4311–4318.
Roussos, E.T., Balsamo, M., Alford, S.K., Wyckoff, J.B., Gligorijevic, B., Wang, Y., Pozzuto,
M., Stobezki, R., Goswami, S., Segall, J.E., et al. (2011a). Mena invasive (MenaINV)
promotes multicellular streaming motility and transendothelial migration in a mouse model of
breast cancer. J Cell Sci 124, 2120–2131.
Roussos, E.T., Condeelis, J.S., and Patsialou, A. (2011b). Chemotaxis in cancer. Nat Rev
Cancer 11, 573–587.
Roussos, E.T., Wang, Y., Wyckoff, J.B., Sellers, R.S., Wang, W., Li, J., Pollard, J.W.,
Gertler, F.B., and Condeelis, J.S. (2010). Mena deficiency delays tumor progression and
decreases metastasis in polyoma middle-T transgenic mouse mammary tumors. Breast
Cancer Res 12, R101.
73
Runkle, E.A., Rice, S.J., Qi, J., Masser, D., Antonetti, D.A., Winslow, M.M., and Mu, D.
(2012). Occludin is a direct target of thyroid transcription factor-1 (TTF-1/NKX2-1). Jbc 287,
28790–28801.
Ruoslahti, E., and Rajotte, D. (2000). An address system in the vasculature of normal
tissues and tumors. Annu. Rev. Immunol. 18, 813–827.
Sage, P.T., Varghese, L.M., Martinelli, R., Sciuto, T.E., Kamei, M., Dvorak, A.M., Springer,
T.A., Sharpe, A.H., and Carman, C.V. (2012). Antigen recognition is facilitated by
invadosome-like protrusions formed by memory/effector T cells. J Immunol 188, 3686–3699.
Saito, R.-A., Watabe, T., Horiguchi, K., Kohyama, T., Saitoh, M., Nagase, T., and Miyazono,
K. (2009). Thyroid transcription factor-1 inhibits transforming growth factor-beta-mediated
epithelial-to-mesenchymal transition in lung adenocarcinoma cells. Cancer Res 69, 2783–
2791.
Salari, K., Spulak, M.E., Cuff, J., Forster, A.D., Giacomini, C.P., Huang, S., Ko, M.E., Lin,
A.Y., van de Rijn, M., and Pollack, J.R. (2012). CDX2 is an amplified lineage-survival
oncogene in colorectal cancer. P Natl Acad Sci Usa 109, E3196–E3205.
Salti, G.I., Manougian, T., Farolan, M., Shilkaitis, A., Majumdar, D., and Gupta, Das, T.K.
(2000). Micropthalmia transcription factor: a new prognostic marker in intermediatethickness cutaneous malignant melanoma. Cancer Res 60, 5012–5016.
Santiago-Medina, M., Gregus, K.A., Nichol, R.H., O'Toole, S.M., and Gomez, T.M. (2015).
Regulation of ECM degradation and axon guidance by growth cone invadosomes.
Development 142, 486–496.
Sato, H., Kinoshita, T., Takino, T., Nakayama, K., and Seiki, M. (1996). Activation of a
recombinant membrane type 1-matrix metalloproteinase (MT1-MMP) by furin and its
interaction with tissue inhibitor of metalloproteinases (TIMP)-2. FEBS Lett 393, 101–104.
Sato, T., del Carmen Ovejero, M., Hou, P., Heegaard, A.M., Kumegawa, M., Foged, N.T.,
and Delaissé, J.M. (1997). Identification of the membrane-type matrix metalloproteinase
MT1-MMP in osteoclasts. J Cell Sci 110 ( Pt 5), 589–596.
Schoumacher, M., Goldman, R.D., Louvard, D., and Vignjevic, D.M. (2010). Actin,
microtubules, and vimentin intermediate filaments cooperate for elongation of invadopodia. J
Cell Biol 189, 541–556.
Seals, D., Azucena, E., Pass, I., Tesfay, L., Gordon, R., Woodrow, M., Resau, J., and
Courtneidge, S. (2005). The adaptor protein Tks5/Fish is required for podosome formation
and function, and for the protease-driven invasion of cancer cells. Cancer Cell 7, 155–165.
Sens, K.L., Zhang, S., Jin, P., Duan, R., Zhang, G., Luo, F., Parachini, L., and Chen, E.H.
(2010). An invasive podosome-like structure promotes fusion pore formation during
myoblast fusion. J Cell Biol 191, 1013–1027.
Shields, J.D., Fleury, M.E., Yong, C., Tomei, A.A., Randolph, G.J., and Swartz, M.A. (2007).
Autologous chemotaxis as a mechanism of tumor cell homing to lymphatics via interstitial
74
flow and autocrine CCR7 signaling. Cancer Cell 11, 526–538.
Sibony-Benyamini, H., and Gil-Henn, H. (2012). Invadopodia: The leading force. Eur J Cell
Biol 91, –901.
Snyder, E.L., Watanabe, H., Magendantz, M., Hoersch, S., Chen, T.A., Wang, D.G.,
Crowley, D., Whittaker, C.A., Meyerson, M., Kimura, S., et al. (2013). Nkx2-1 represses a
latent gastric differentiation program in lung adenocarcinoma. Mol Cell 50, 185–199.
Song, Y., Washington, M.K., and Crawford, H.C. (2010). Loss of FOXA1/2 is essential for
the epithelial-to-mesenchymal transition in pancreatic cancer. Cancer Res 70, 2115–2125.
Soo, K., O'Rourke, M.P., Khoo, P.-L., Steiner, K.A., Wong, N., Behringer, R.R., and Tam,
P.P.L. (2002). Twist function is required for the morphogenesis of the cephalic neural tube
and the differentiation of the cranial neural crest cells in the mouse embryo. Dev Biol 247,
251–270.
Steeg, P.S. (2006). Tumor metastasis: mechanistic insights and clinical challenges. Nat Med
12, 895–904.
Stenhouse, G., Fyfe, N., King, G., Chapman, A., and Kerr, K.M. (2004). Thyroid transcription
factor 1 in pulmonary adenocarcinoma. J Clin Pathol 57, 383–387.
Stylli, S.S., Stacey, T.T., Verhagen, A.M., Xu, S.S., Pass, I., Courtneidge, S.A., and Lock, P.
(2009). Nck adaptor proteins link Tks5 to invadopodia actin regulation and ECM
degradation. J Cell Sci 122, 2727–2740.
Tammela, T., and Alitalo, K. (2010). Lymphangiogenesis: Molecular mechanisms and future
promise. Cell 140, 460–476.
Tanaka, H., Yanagisawa, K., Shinjo, K., Taguchi, A., Maeno, K., Tomida, S., Shimada, Y.,
Osada, H., Kosaka, T., Matsubara, H., et al. (2007). Lineage-specific dependency of lung
adenocarcinomas on the lung development regulator TTF-1. Cancer Res 67, 6007–6011.
Tang, Y., Shu, G., Yuan, X., Jing, N., and Song, J. (2011). FOXA2 functions as a
suppressor of tumor metastasis by inhibition of epithelial-to-mesenchymal transition in
human lung cancers. Cell Res 21, 316–326.
Tarone, G., Cirillo, D., Giancotti, F., Comoglio, P., and Marchisio, P. (1985). Rous-sarcoma
virus-transformed fibroblasts adhere primarily at discrete protrusions of the ventral
membrane called podosomes. Exp Cell Res 159, 141–157.
Tatin, F., Varon, C., Genot, E., and Moreau, V. (2006). A signalling cascade involving PKC,
Src and Cdc42 regulates podosome assembly in cultured endothelial cells in response to
phorbol ester. J Cell Sci 119, 769–781.
Tehrani, S., Tomasevic, N., Weed, S., Sakowicz, R., and Cooper, J.A. (2007). Src
phosphorylation of cortactin enhances actin assembly. P Natl Acad Sci Usa 104, 11933–
11938.
75
Teti, A., Grano, M., Carano, A., Colucci, S., and Zambonin-Zallone, A. (1989).
Immunolocalization of beta 3 subunit of integrins in osteoclast membrane. Boll. Soc. Ital.
Biol. Sper. 65, 1031–1037.
Thiery, J.P. (2002). Epithelial-mesenchymal transitions in tumour progression. Nat Rev
Cancer 2, 442–454.
Thiery, J.P., Acloque, H., Huang, R.Y.J., and Nieto, M.A. (2009). Epithelial-mesenchymal
transitions in development and disease. Cell 139, 871–890.
Torres, M., Gómez-Pardo, E., Dressler, G.R., and Gruss, P. (1995). Pax-2 controls multiple
steps of urogenital development. Development 121, 4057–4065.
Ulrich, T.A., de Juan Pardo, E.M., and Kumar, S. (2009). The mechanical rigidity of the
extracellular matrix regulates the structure, motility, and proliferation of glioma cells. Cancer
Res 69, 4167–4174.
Varon, C., Tatin, F., Moreau, V., Van Obberghen-Schilling, E., Fernandez-Sauze, S.,
Reuzeau, E., Kramer, I., and Génot, E. (2006). Transforming growth factor beta induces
rosettes of podosomes in primary aortic endothelial cells. Mol Cell Biol 26, 3582–3594.
Wan, H., Dingle, S., Xu, Y., Besnard, V., Kaestner, K.H., Ang, S.-L., Wert, S., Stahlman,
M.T., and Whitsett, J.A. (2005). Compensatory roles of Foxa1 and Foxa2 during lung
morphogenesis. Jbc 280, 13809–13816.
Wan, H., Kaestner, K.H., Ang, S.-L., Ikegami, M., Finkelman, F.D., Stahlman, M.T.,
Fulkerson, P.C., Rothenberg, M.E., and Whitsett, J.A. (2004). Foxa2 regulates
alveolarization and goblet cell hyperplasia. Development 131, 953–964.
Wang, J., Taba, Y., Pang, J., Yin, G., Yan, C., and Berk, B.C. (2009). GIT1 mediates VEGFinduced podosome formation in endothelial cells: critical role for PLCgamma. Arterioscler.
Thromb. Vasc. Biol. 29, 202–208.
Watanabe, H., Francis, J.M., Woo, M.S., Etemad, B., Lin, W., Fries, D.F., Peng, S., Snyder,
E.L., Tata, P.R., Izzo, F., et al. (2013). Integrated cistromic and expression analysis of
amplified NKX2-1 in lung adenocarcinoma identifies LMO3 as a functional transcriptional
target. Gene Dev 27, 197–210.
Weinstein, D.C., Ruiz i Altaba, A., Chen, W.S., Hoodless, P., Prezioso, V.R., Jessell, T.M.,
and Darnell, J.E. (1994). The winged-helix transcription factor HNF-3 beta is required for
notochord development in the mouse embryo. Cell 78, 575–588.
Weir, B.A., Woo, M.S., Getz, G., Perner, S., Ding, L., Beroukhim, R., Lin, W.M., Province,
M.A., Kraja, A., Johnson, L.A., et al. (2007). Characterizing the cancer genome in lung
adenocarcinoma. Nature 450, 893–U22.
Wheeler, A.P., Smith, S.D., and Ridley, A.J. (2006). CSF-1 and PI 3-kinase regulate
podosome distribution and assembly in macrophages. Cell Motil Cytoskel 63, 132–140.
Wilson, G.R., Sunley, J., Smith, K.R., Pope, K., Bromhead, C.J., Fitzpatrick, E., Di Rocco,
76
M., van Steensel, M., Coman, D.J., Leventer, R.J., et al. (2014). Mutations in SH3PXD2B
cause Borrone dermato-cardio-skeletal syndrome. Eur. J. Hum. Genet. 22, 741–747.
Winslow, M.M., Dayton, T.L., Verhaak, R.G.W., Kim-Kiselak, C., Snyder, E.L., Feldser, D.M.,
Hubbard, D.D., Dupage, M.J., Whittaker, C.A., Hoersch, S., et al. (2011). Suppression of
lung adenocarcinoma progression by Nkx2-1. Nature 473, 101–104.
Wolf, K., Mazo, I., Leung, H., Engelke, K., Andrian, von, U., Deryugina, E., Strongin, A.,
Brocker, E., and Friedl, P. (2003). Compensation mechanism in tumor cell migration:
mesenchymal-amoeboid transition after blocking of pericellular proteolysis. J Cell Biol 160,
267–277.
Wolf, K., and Friedl, P. (2011). Extracellular matrix determinants of proteolytic and nonproteolytic cell migration. Trends Cell Biol 21, 736–744.
Wu, H., Reynolds, A.B., Kanner, S.B., Vines, R.R., and Parsons, J.T. (1991). Identification
and characterization of a novel cytoskeleton-associated pp60src substrate. Mol Cell Biol 11,
5113–5124.
Wyckoff, J., Wang, W., Lin, E.Y., Wang, Y., Pixley, F., Stanley, E.R., Graf, T., Pollard, J.W.,
Segall, J., and Condeelis, J. (2004). A paracrine loop between tumor cells and macrophages
is required for tumor cell migration in mammary tumors. Cancer Res 64, 7022–7029.
Xu, P.-X., Zheng, W., Huang, L., Maire, P., Laclef, C., and Silvius, D. (2003). Six1 is
required for the early organogenesis of mammalian kidney. Development 130, 3085–3094.
Yamaguchi, H., Wyckoff, J., and Condeelis, J. (2005a). Cell migration in tumors. Curr Opin
Cell Biol 17, 559–564.
Yamaguchi, H., Lorenz, M., Kempiak, S., Sarmiento, C., Coniglio, S., Symons, M., Segall, J.,
Eddy, R., Miki, H., Takenawa, T., et al. (2005b). Molecular mechanisms of invadopodium
formation: the role of the N-WASP-Arp2/3 complex pathway and cofilin. J Cell Biol 168,
441–452.
Yamaguchi, T., Yanagisawa, K., Sugiyama, R., Hosono, Y., Shimada, Y., Arima, C., Kato,
S., Tomida, S., Suzuki, M., Osada, H., et al. (2012). NKX2-1/TITF1/TTF-1-Induced ROR1 is
required to sustain EGFR survival signaling in lung adenocarcinoma. Cancer Cell 21, 348–
361.
Yan, C., Sever, Z., and Whitsett, J.A. (1995). Upstream enhancer activity in the human
surfactant protein B gene is mediated by thyroid transcription factor 1. Jbc 270, 24852–
24857.
Yana, I., and Weiss, S.J. (2000). Regulation of membrane type-1 matrix metalloproteinase
activation by proprotein convertases. Mol Biol Cell 11, 2387–2401.
Yang, J., Mani, S.A., Donaher, J.L., Ramaswamy, S., Itzykson, R.A., Come, C., Savagner,
P., Gitelman, I., Richardson, A., and Weinberg, R.A. (2004). Twist, a master regulator of
morphogenesis, plays an essential role in tumor metastasis. Cell 117, 927–939.
77
Yu, X., Zech, T., McDonald, L., Gonzalez, E.G., Li, A., Macpherson, I., Schwarz, J.P.,
Spence, H., Futó, K., Timpson, P., et al. (2012). N-WASP coordinates the delivery and Factin-mediated capture of MT1-MMP at invasive pseudopods. J Cell Biol 199, 527–544.
Zambonin-Zallone, A., Teti, A., Grano, M., RUBINACCI, A., ABBADINI, M., GABOLI, M., and
Marchisio, P.C. (1989). Immunocytochemical distribution of extracellular matrix receptors in
human osteoclasts: a beta 3 integrin is colocalized with vinculin and talin in the podosomes
of osteoclastoma giant cells. Exp Cell Res 182, 645–652.
Zaret, K.S., Caravaca, J.M., Tulin, A., and Sekiya, T. (2010). Nuclear mobility and mitotic
chromosome binding: similarities between pioneer transcription factor FoxA and linker
histone H1. Cold Spring Harb Sym 75, 219–226.
Zhang, D., Udagawa, N., Nakamura, I., Murakami, H., Saito, S., Yamasaki, K., Shibasaki,
Y., Morii, N., Narumiya, S., and Takahashi, N. (1995). The small GTP-binding protein, rho
p21, is involved in bone resorption by regulating cytoskeletal organization in osteoclasts. J
Cell Sci 108 ( Pt 6), 2285–2292.
Zhang, Y., Gu, M.-L., Zhou, X.-X., Ma, H., Yao, H.-P., and Ji, F. (2014). Altered expression
of ETV1 and its contribution to tumorigenic phenotypes in gastrointestinal stromal tumors.
Oncol Rep 32, 927–934.
Zhou, J., Ng, A.Y., Tymms, M.J., Jermiin, L.S., Seth, A.K., Thomas, R.S., and Kola, I.
(1998). A novel transcription factor, ELF5, belongs to the ELF subfamily of ETS genes and
maps to human chromosome 11p13-15, a region subject to LOH and rearrangement in
human carcinoma cell lines. Oncogene 17, 2719–2732.
Zhou, J., Chehab, R., Tkalcevic, J., Naylor, M.J., Harris, J., Wilson, T.J., Tsao, S., Tellis, I.,
Zavarsek, S., Xu, D., et al. (2005). Elf5 is essential for early embryogenesis and mammary
gland development during pregnancy and lactation. Embo J 24, 635–644.
Zhou, L., Dey, C.R., Wert, S.E., Yan, C., Costa, R.H., and Whitsett, J.A. (1997). Hepatocyte
nuclear factor-3beta limits cellular diversity in the developing respiratory epithelium and
alters lung morphogenesis in vivo. Dev. Dyn. 210, 305–314.
Zhu, C.-P., Wang, J., Shi, B., Hu, P.-F., Ning, B.-F., Zhang, Q., Chen, F., Chen, W.-S.,
Zhang, X., and Xie, W.-F. (2015). The transcription factor FOXA2 suppresses gastric
tumorigenesis in vitro and in vivo. Dig. Dis. Sci. 60, 109–117.
Zlotnik, A., Burkhardt, A.M., and Homey, B. (2011). Homeostatic chemokine receptors and
organ-specific metastasis. Nat Rev Immunol 11, 597–606.
78
CHAPTER 2
Differential Tks5 isoform expression contributes to
metastatic invasion of lung adenocarcinoma
Carman Man-Chung Li1, Guoan Chen2, Talya L. Dayton1, Caroline Kim-Kiselak1, Sebastian
Hoersch1, Charles A. Whittaker1, Roderick T. Bronson3, David G. Beer2, Monte M.
Winslow4, Tyler Jacks1,5
1 David H. Koch Institute for Integrative Cancer Research, Department of Biology,
Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
2 Department of Surgery, Thoracic Surgery, University of Michigan Medical School, Ann
Arbor, Michigan 48109, USA
3 Department of Pathology, Tufts University School of Medicine and Veterinary Medicine,
North Grafton, Massachusetts 01536, USA
4 Department of Genetics, Stanford University School of Medicine, Stanford, California
94305, USA
5 Howard Hughes Medical Institute, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, USA
The author performed all the experiments, with some assistance from M.M.W., T.L.D. and
C.K-K. Bioinformatics analysis of microarray data was performed by S.H. and C.W.
Pathology analysis was performed by R.B., while G.C. and D.G.B assisted with clinical data
analysis. All experiments were performed in the laboratory of Tyler Jacks.
79
ABSTRACT
Metastasis accounts for the vast majority of cancer related deaths, yet the molecular
mechanisms that drive metastatic spread remain poorly understood. Here we report that
Tks5, which has been linked to formation of proteolytic cellular protrusions known as
invadopodia, undergoes an isoform switch during metastatic progression in a geneticallyengineered mouse model of lung adenocarcinoma. Non-metastatic primary tumor-derived
cells predominantly expressed a short isoform Tks5short, while metastatic primary tumor- and
metastasis-derived cells acquired increased expression of the full-length isoform Tks5long.
This elevation of Tks5long-to-Tks5short ratio correlated with a commensurate increase in
invadopodia activity in metastatic cells compared to non-metastatic cells. Further
characterization of these isoforms by knockdown and over-expression experiments
demonstrated that Tks5long promoted invadopodia in vitro and increased metastasis in
transplant models and an autochthonous model of lung adenocarcinoma. Conversely,
Tks5short decreased invadopodia stability and proteolysis, acting as a natural dominantnegative inhibitor to Tks5long. Importantly, high Tks5long and low Tks5short expressions in
human lung adenocarcinomas correlated with metastatic disease and predicted worse
survival of early-stage patients. These data indicate that tipping the Tks5 isoform balance to
a high Tks5long-to-Tks5short ratio promotes invadopodia-mediated invasion and metastasis.
80
INTRODUCTION
Despite the high rates of mortality associated with metastatic lung cancer
(Keshamouni et al. 2009; Siegel et al. 2013), the molecular mechanisms underlying disease
progression remain incompletely understood. Metastasis accounts for the vast majority of all
lung cancer fatality, as even 30-50% of early-stage patients who undergo surgical resection
eventually succumb to metastatic relapse, and patients with metastatic disease are almost
always incurable (Keshamouni et al. 2009). The development of more effective therapeutic
interventions for this disease will rely on improving our understanding of metastasis at the
molecular level.
Tks5 (also known as Sh3pxd2a) has been previously implicated in promoting
metastasis because of its role in invasive cellular structures known as invadopodia
(Courtneidge 2012). Invadopodia are actin-rich, proteolytic membrane protrusions that were
initially identified in Src-transformed mouse embryonic fibroblasts (Chen et al. 1984; Tarone
et al. 1985), and later observed in a variety of cultured human cancer cells, including breast
cancer, melanoma, and head and neck squamous cell carcinoma (Seals et al. 2005;
Bowden et al. 2006; Clark et al. 2007). While invadopodia formation is spontaneous in some
cancer cells in culture, it can be further stimulated by activating integrin β1 and receptors for
epidermal growth factor (EGF) and platelet-derived growth factor (PDGF) (Nakahara et al.
1998; Yamaguchi et al. 2005; Philippar et al. 2008; Eckert et al. 2011). These cellular
“invasive feet” have been demonstrated to produce a variety of proteases, including
metalloproteases (MMP2, MMP9, MT1-MMP and the ADAM family of sheddases) and
serine proteases (seprase and uPAR) (Linder 2007), and are capable of digesting various
components of the extracellular matrix in vitro (Kelly et al. 1994). Given their proteolytic
capabilities, invadopodia are thought to facilitate metastasis by enabling tumor cells to
81
breach the basement membrane, degrade the extracellular matrix, and invade into the
stroma during the intravasation and extravasation steps of the metastatic cascade (Murphy
and Courtneidge 2011). In fact, it has been proposed that invadopodia in cancer cells are a
co-opted and dysregulated version of normal cellular structures known as podosomes that
are found in untransformed cells such as osteoclasts, macrophages, dendritic cells,
endothelial cells and smooth muscle cells (Murphy and Courtneidge 2011). Although it is still
unclear whether invadopodia have any physiological function in metastatic invasion during
natural tumor progression (Linder 2009; Sibony-Benyamini and Gil-Henn 2012), animal
transplant studies and intravital imaging have provided some in vivo evidence for a role for
invadopodia in mediating metastasis (Philippar et al. 2008; Gligorijevic et al. 2012).
Tks5 is an important component of invadopodia and mediates invadopodia formation
by acting as a Src-dependent scaffolding protein (Lock et al. 1998). Upon phosphorylation
by Src, the N-terminal phox (PX) homology domain of Tks5 is thought to be released from
intramolecular interactions, and becomes free to bind membrane phosphoinositides
including PI(3,4)P2, thereby localizing Tks5 to the site of invadopodia formation (Abram et
al. 2003; Oikawa et al. 2008). Tks5 also contains five C-terminal Src homology 3 (SH3)
domains, which recruit effector proteins (including AFAP-110, cortactin, and ADAM
metalloproteases) to initiate actin polymerization and matrix degradation (Abram et al. 2003;
Crimaldi et al. 2009). Knockdown of Tks5 (targeting all isoforms simultaneously) abrogates
invadopodia formation and proteolytic function in cultured human breast cancer and
melanoma cells (Seals et al. 2005), and reduces lung metastasis formation by Srctransformed NIH-3T3 mouse embryonic fibroblasts and Ras-transformed human mammary
epithelial cells after intravenous injection or subcutaneous transplantation (Blouw et al.
2008; Eckert et al. 2011).
82
Despite the evidence that Tks5 is important for invadopodia formation in cell lines, its
contribution to the metastatic process in naturally evolving tumors has not been elucidated.
Moreover, previous studies have not accounted for the presence of the two functionally
distinct isoforms (which we have termed Tks5long and Tks5short) that we characterize in this
report. These two Tks5 variants were first detected by immunoblotting in Src-transformed
NIH-3T3 cells when Tks5 was initially identified (Lock et al. 1998); however, no subsequent
functional studies have taken into account the existence of these isoforms. Therefore, the
distinct roles of Tks5long and Tks5short in invadopodia function and cancer invasion are
unknown.
Here we report that Tks5long and Tks5short play distinct and opposing roles in
regulating invadopodia-mediated invasion in lung adenocarcinoma. We show that metastatic
primary tumor- and metastasis-derived cells acquired an elevated ratio of Tks5long-toTks5short expression and a commensurate increase in invadopodia activity compared to nonmetastatic cells. We further demonstrate that the ratio of Tks5long-to-Tks5short expression
regulates invadopodia function in vitro, and influences metastatic potential in transplant
models and a genetically-engineered mouse model of lung cancer. Finally, we provide
evidence that the relative expression of Tks5long and Tks5short represents an important
prognostic factor in human lung cancer. These results highlight the isoform-dependent roles
of Tks5 in invadopodia and metastasis.
83
RESULTS
Metastatic and non-metastatic lung adenocarcinoma cells exhibit differential
expression of Tks5long and Tks5short
To better characterize the cell-state changes and molecular alterations that
accompany tumor progression and metastasis in lung adenocarcinoma, we have recently
developed a KrasLSL-G12D/WT; p53flox/flox mouse model for studying metastatic and nonmetastatic primary tumors (Winslow et al. 2011). This model harbors genetic mutations
frequently found in human lung adenocarcinoma (Rodenhuis et al. 1988; Takahashi et al.
1989), and closely recapitulates the histopathological progression of the human disease
(Jackson et al. 2005). Although these mice develop multiple KrasG12D, p53-/- lung tumors
after inhalation of lentivirus expressing Cre recombinase, only a subset of tumors eventually
acquire full metastatic potential, suggesting that progression to metastasis requires
additional genetic and/or epigenetic events. Importantly, this model allows the identification
of metastatic versus non-metastatic primary tumors, as the metastases that form in these
mice can be matched to their primary tumor of origin based on the common lentiviral
integration site in their genome. Thus, primary tumors (TMet) that have given rise to
metastatic lesions can be distinguished from primary tumors for which no metastasis was
found (TnonMet). Cell lines derived from these TMet and TnonMet tumors were examined for their
gene expression profiles via exon microarrays, which identified expression alterations in
various genes, including Nkx2-1 and Hmga2, that were associated with metastatic
progression (Winslow et al. 2011).
To identify gene isoform expression changes that could be caused by alternative
splicing or differential promoter utilization in our collection of autochthonous tumor-derived
TnonMet and TMet cells, we developed an algorithm that allowed us to query our exon array
84
data for changes in isoform usage. The most striking result from this analysis was a change
in Tks5 isoform expression between TnonMet and TMet cells (Supplementary Figure S2). We
identified two Tks5 isoforms by referencing sequences published on the UCSC genome
browser (http://genome.ucsc.edu/; assembly NCBI37/mm9, gene Sh3pxd2a): Tks5long which
contains exons 1-15, and Tks5short which contains a distinct 5’ sequence from intron 7
followed by exons 8-15 (Figure 1A). Both transcripts encode five SH3 domains in the Cterminus, but only Tks5long contains the N-terminal PX homology domain (Figure 1A).
We confirmed the differential expression of Tks5long and Tks5short in TnonMet and TMet
cells by performing isoform-specific qRT-PCR on a panel of three TnonMet cell lines as well as
five TMet cell lines and their five matching metastasis cell lines (Met). Consistent with the
microarray data, Tks5long transcript levels were on average 4-fold higher in TMet/Met cells
compared to TnonMet cells, while Tks5short was transcribed at a similar level between TMet/Met
and TnonMet cells (Figure 1B). As a result, the ratio of Tks5long-to-Tks5short was on average 6fold higher in TMet/ Met cells than TnonMet cells (Figure 1B). Consistent with the mRNA
expression patterns, the ratio of Tks5long-to-Tks5short proteins (150 kDa and 140 kDa,
respectively) is higher in TMet cells compared with TnonMet cells (Figure 1C).
Interestingly, despite this increase in Tks5long transcripts, because Tks5long only
accounted for a fraction of total Tks5 expression, the levels of total Tks5 transcript did not
vary significantly between TnonMet and TMet/ Met cell lines in our exon microarray or in our
qRT-PCR analysis using primers targeting a common region shared by Tks5long and Tks5short
transcripts (Figure 1B). Thus, the TnonMet and TMet/Met cell lines could be distinguished by
their Tks5long expression or Tks5long-to-Tks5short ratio, but not by total Tks5 level. 5’RACE and
H3K4me3 ChIP-seq analyses suggest that the two isoforms are transcribed from distinct
promoters.
85
Figure 1. TnonMet and TMet lung adenocarcinoma cells exhibit differential expression of
Tks5long and Tks5short.
(A) Tks5long and Tks5short differ in their 5’ coding sequences and hence the presence of the
N-terminal phox homology (PX) domain in the encoded proteins. The 3’ coding sequence
common in the Tks5long and Tks5short transcripts encodes five Src-homology 3 (SH3)
domains in the C terminus of the protein.
(B) Isoform-specific qRT-PCR analysis of Tks5long and Tks5short expression in three TnonMet
(368T1, 394T4, 802T4), five TMet (373T1, 389T2, 393T3, 393T5, 482T1), and five Met
(373N1, 393N1, 393M1, 482N1, 482M1) cell lines. Color dots indicate matching TMet and
Met cell lines in the same lineage (red, 373T1 and 373N1; brown, 393T3 and 393N1; beige,
393T5 and 393M1; orange, 482T1, 482N1 and 482M1). N indicates lymph node metastasis,
while M indicates distant metastasis. Tks5long-specific primers amplify exons 2 and 3
(indicated in orange in panel A), while Tks5short-specific primers amplify a Tks5short-unique 5’
sequence and exons 8 and 9 (indicated in green in panel A). (**) P-value < 0.01; (*) P-value
< 0.05; Student’s T-test.
(C) Immunoblot detection of Tks5 isoforms in a serial dilution of three TnonMet (368T1, 394T4,
802T4) and three TMet (373T1, 393T3, 482T1) cell lysates. Tubulin was used as a loading
control.
86
Expression of Tks5long and Tks5short in TnonMet and TMet cells correlates with
invadopodia formation and function
Given the differences in Tks5 isoform levels between TMet and TnonMet cells, and the
previously reported role of Tks5 in mediating invadopodia activity, we compared three TnonMet
and four TMet cell lines for invadopodia formation and function. To measure invadopodia
formation, we performed immunofluorescence staining to detect the colocalization of two
essential invadopodia components, cortactin and F-actin (Figure 2A). TMet cells, which have
a higher Tks5long-to-Tks5short ratio, displayed a higher frequency of colocalized cortactin and
F-actin than TnonMet cells (Figure 2B). To measure invadopodia function, we examined
invadopodia-mediated proteolysis in TnonMet and TMet cells by culturing them on a thin layer of
FITC-labeled gelatin. The degraded areas can be observed by fluorescence microscopy as
FITC-negative patches that frequently coincide with cortactin/F-actin-stained invadopodia
foci (Figure 2A). The gelatin degradation assay is a more sensitive method to measure
invadopodia activity compared to cortactin/F-actin immunofluorescence staining, because
the effect of degradation is cumulative over time while the presence of invadopodia is
transient. Quantification of the degradation area showed a strong correlation between the
proteolytic capability of these TnonMet and TMet cells and their ratio of Tks5long-to-Tks5short
expression: TMet cells with higher Tks5long-to-Tks5short ratios were more proteolytic on the
gelatin matrix compared with TnonMet cells (Figure 2C). Importantly, we did not observe
significant differences in other invadopodia components at either the total gene expression
or isoform level by exon array analysis (analyzed for Tks4, Src, Cortactin, Afap110, p190
RhoGAP, Arg, N-WASP, Arp2/3 complex subunits, Wave1, Cdc42, Cofilin, Gelsolin, MT1MMP, MMP2, MMP9, ADAM12, ADAM15, and ADAM19) or at the protein phosphorylation
level by western blot (analyzed for Src). Taken together, these observations suggest that an
87
increased ratio of Tks5long-to-Tks5short is associated with the enhanced invadopodia formation
and function that we observe in TMet cells compared to TnonMet cells.
Figure 2. Differential Tks5long and Tks5short expression in TnonMet and TMet lung
adenocarcinoma cells correlates with invadopodia formation and function.
(A) Colocalization of invadopodia components cortactin (green) and F-actin (red), as well as
FITC-negative areas of gelatin degradation, are more readily observed in TMet cells
compared with TnonMet cells. Cells were cultured on a thin layer of FITC-labeled gelatin for 24
hours, and then fixed and processed for immunofluorescence staining. Magnified views of
the regions indicated by the boxed area are shown to the right. Representative images of
TnonMet and TMet cells are shown.
(B) Correlation between invadopodia formation and the ratio of Tks5long-to-Tks5short
expression in three TnonMet cell lines (grey circles; specifically 368T1, 394T4, and 802T4 from
left to right) and four TMet cell lines (orange circles; specifically 393T3, 393T5, 482T1, and
373T1 from left to right). Cells with colocalization of cortactin and F-actin in
immunofluorescence staining were scored as invadopodia-positive. At least 60 cells were
scored for each cell line. Results are representative of three independent experiments.
(C) Correlation between gelatin-matrix degradation and the ratio of Tks5long-to-Tks5short
expression in three TnonMet cell lines (grey circles; specifically 368T1, 394T4, and 802T4 from
left to right) and four TMet cell lines (orange circles; specifically 393T3, 393T5, 482T1, and
373T1 from left to right). Areas of degradation were quantified using ImageJ and normalized
to number of cells per field. At least 50 fields and 1500 cells were analyzed per cell line.
Results are representative of three independent experiments.
88
Knockdown of Tks5long impairs invadopodia activity and metastasis formation.
Since the distinct functions of Tks5long and Tks5short in invadopodia have not been
previously reported, we tested whether Tks5long was specifically required for invadopodia
formation and function in two TMet cell lines. Stable RNAi-mediated depletion of Tks5long
using two isoform-specific short hairpin RNAs reduced Tks5long expression in TMet cells by
55-65% without a significant effect on Tks5short (Figure 3A and 3B). Tks5long knockdown
impaired the ability of TMet cells to form invadopodia as measured by immunofluorescence
staining for cortactin/F-actin foci (Figure 3C and 3D; Supplementary Figure S3A and S3B).
These TMet-shTks5long cells also exhibited significantly reduced extracellular-matrix
proteolysis capability when cultured on FITC-gelatin (Figure 3C and 3E; Supplementary
Figure S3A and S3C). Because these observations were consistent for both shTks5long
shRNAs, it is unlikely that they were the results of off-target effects. Collectively, these data
indicate that Tks5long is necessary for invadopodia activity in metastatic lung cancer cells.
To determine whether the effects of Tks5long knockdown on invadopodia activity in
vitro translate to the inhibition of metastatic ability in vivo, we transplanted TMet-shTks5long
cells subcutaneously into athymic nude mice to assess the metastatic potential of tumor
cells. TMet cells with Tks5long knockdown exhibited a significantly diminished ability to
disseminate from the subcutaneous site and form lung tumor nodules compared to parental
cells 8 weeks after subcutaneous injection (Figure 3F). Of note, the sizes of the
subcutaneous tumors were comparable between the two mouse cohorts, suggesting that
loss of Tks5long expression had no effect on primary tumor growth (Supplementary Figure
S3D). Furthermore, in a separate experiment, we transplanted cells intrasplenically to
assess their ability to extravasate and colonize the liver after draining into the hepatic portal
vein from the spleen. Consistent with data from the subcutaneous transplant experiment,
intrasplenically injected TMet-shTks5long cells showed substantially reduced ability to form
89
liver nodules 3 weeks after transplantation compared to controls (Figure 3G). Importantly,
inefficient liver colonization by TMet-shTks5long cells was not a result of reduced cell
proliferation or increased apoptosis, as the infrequent liver nodules that were formed by TMetshTks5long cells displayed similar mitotic and apoptotic indices compared to parental cells
(Supplementary Figure S3E and S3F). These experiments suggest that Tks5long-induced
invadopodia are not only important for promoting cell invasion during intravasation at the
primary site, but are also required for extravasation and/or colonization at the metastatic
sites, potentially by facilitating metastatic cell exit from blood vessels and/or invasion at
secondary sites.
90
Figure 3. Tks5long is required for invadopodia activity in vitro and metastasis
formation in vivo.
(A) Immunoblot detection of Tks5 isoforms in TMet cells (373T1) expressing control
shLuciferase or shTks5long (sh1 and sh2) hairpins. Tubulin was used as a loading control.
(B) qRT-PCR analysis shows that sh1 and sh2 reduce Tks5long transcripts by 55-65%
without significant effects on Tks5short mRNA levels.
(C) Immunofluorescence staining shows that colocalization of invadopodia components
(cortactin in green, and F-actin in red) and FITC-negative areas of gelatin-matrix
degradation are less frequently observed in TMet-shTks5long cells compared with TMet cells
expressing control shRNA. Magnified views of the regions indicated by the boxed area are
shown to the right. Representative images from TMet cells (373T1) are shown.
(D) Effects of Tks5long knockdown on invadopodia formation in TMet cells (373T1). At least
100 cells were scored for colocalization of cortactin and F-actin in three independent
experiments. All values are mean ± SEM. (*), P-value< 0.05, paired T-test.
(E) Effects of Tks5long knockdown on FITC-gelatin matrix degradation in TMet cells (373T1).
Areas of degradation were quantified using ImageJ and normalized to number of cells per
field. At least 75 fields containing a total of 600 cells were analyzed per condition. All values
are mean ± SEM. (**), P-value< 0.01; (***), P-value< 0.001; Student’s T-test.
(F-G) Tks5long knockdown drastically impairs lung nodule formation 8 weeks after
subcutaneous transplant of GFP-positive TMet cells (393T3) (F), and significantly decreases
liver nodule formation 3 weeks after intrasplenic transplant of TMet cells (373T1). Values are
mean ± SEM, P < 0.05, Student’s T-test.
91
Increased Tks5long expression promotes invadopodia activity and metastasis
formation
To ask whether exogenous expression of Tks5long alone is sufficient to enhance
invadopodia formation and function in TnonMet cells, we generated a lentivirus that allows
doxycycline-inducible expression of Flag-tagged Tks5long (Figure 4A). Increased expression
of Tks5long in two independent TnonMet cell lines promoted invadopodia formation compared
with parental cells as measured by immunofluorescence staining for foci of cortactin/F-actin
colocalization (Figure 4B and 4C; Supplementary Figure S4A). Moreover, these cells also
displayed a dramatic 6- to 14-fold increase in matrix proteolysis in the FITC-gelatin
degradation assay (Figure 4B and 4D; Supplementary Figure S4B). Our data thus
demonstrate that changing the Tks5long-to-Tks5short ratio by increasing Tks5long expression is
sufficient to promote invadopodia formation and function in non-metastatic lung
adenocarcinoma cells.
We then sought to determine whether Tks5long also facilitates tumor metastasis in
vivo. To test this in a physiologically relevant model of tumor progression, we infected
KrasLSL-G12D/WT; p53flox/flox; Rosa26-LSL-TdTomato; CCSP-rtTA mice with a PGK-Cre/TRETks5long lentivirus (Figure 4E). Inhalation of the virus initiates RFP-positive lung tumors
through the concomitant activation of oncogenic Kras and deletion of p53. Infected mice
were then fed with a doxycycline diet starting at 4 weeks post-infection to induce Flagtagged Tks5long expression, allowing us to study the effects of Tks5long specifically on tumor
progression and not initiation. We confirmed inducible expression of our construct through
detection of doxycycline- and rtTA-dependent expression of Flag-tagged Tks5long in these
lung tumors (Supplementary Figure S4C). At 6 months post-infection, although widespread
distant metastases had not yet developed, we observed a significant acceleration in primary
tumor progression in Tks5long mice (n = 9 mice; 221 tumors) compared to the control mice (n
92
= 14 mice; 397 tumors): Tks5long tumors had a smaller proportion of low-grade tumors, and a
commensurate increase in high-grade lesions (Supplementary Figure S4D). These highgrade tumors were invasive into stromal tissues surrounding the blood vessels, and were
thus categorized as grade 4 lesions (Supplementary Figure S4E).
Consistent with these observations, at 8 months post-infection, when mice had
developed both pleural metastases and distant metastases in the liver, kidneys and distant
lymph nodes (Figure 4F), Tks5long-expressing mice developed significantly more distant
metastases (11/15 mice) than the control group (2/8 mice) (P < 0.05 by both Chi-square test
and Fisher’s exact test; Figure 4G). In addition, 15/15 of Tks5long mice developed pleural
metastases and/or distant metastases, while 4/8 mice in the control group remained
metastasis-free (P < 0.003 by Chi-square test, and P < 0.008 by Fisher’s exact test; Figure
4H). Importantly, we did not observe any significant difference in the primary lung tumor
sizes and total lung tumor burden in the Tks5long mice versus control mice (Supplementary
Figure S4F-G), suggesting that the effect of Tks5long on tumor progression and metastasis
was not a consequence of increasing the rate of tumor growth.
Collectively, these data from this mouse model indicate that Tks5long plays an
important role in promoting metastasis in vivo, and are consistent with our in vitro evidence
that Tks5long promotes invadopodia-mediated invasion. Whether tumor progression can be
further stimulated by other mechanisms, including, for example, growth factor shedding
during invadopodia-mediated degradation of the extracellular matrix, remains an open
possibility and will be addressed in future studies.
93
Figure 4. Tks5long is sufficient to promote invadopodia activity in vitro and metastasis
formation in vivo.
(A) Immunoblot detection of Tks5 isoforms in 394T4 TnonMet cells and TnonMet-Flag-Tks5long
cells with or without doxycycline induction of Flag-Tks5long. Tubulin was used as a loading
control.
(B) Immunofluorescence staining shows that colocalization of invadopodia components
(cortactin in green, and F-actin in red) and FITC-negative areas of gelatin-matrix
degradation are more readily observed in TnonMet-Tks5long cells compared with parental
TnonMet cells, both treated with 2 µg/mL doxycycline. Magnified views of the regions indicated
94
by the boxed area are shown to the right. Representative images of TnonMet cells (394T4) are
shown.
(C) Effects of increased Tks5long expression on invadopodia formation in TnonMet cells
(394T4). At least 100 cells were scored for colocalization of cortactin and F-actin in three
independent experiments. All values are mean ± SEM. (*), P-value< 0.05; paired T-test.
(D) Effects of increased Tks5long expression on FITC-gelatin matrix degradation in TnonMet
cells (394T4). Areas of degradation were quantified using ImageJ and normalized to number
of cells per field. At least 40 fields containing a total of 1800 cells were analyzed per
condition. All values are mean ± SEM. (***), P-value< 0.001; Student’s T-test.
(E) Induction of lung adenocarcinomas with doxycycline-dependent overexpression of
Tks5long in an autochthonous mouse model. KrasLSL-G12D/WT; p53flox/flox; CCSP-rtTA mice were
infected with a PGK-Cre/TRE-Tks5long lentivirus. The Cre recombinase initiates lung
adenocarcinomas upon infection. Doxycycline diet later induced Tks5long expression in
these tumors to study the effects of Tks5long on tumor progression without affecting tumor
initiation.
(F) Examples of distant metastases observed in mice with increased Tks5long expression. All
tumors are RFP-positive because of a Rosa26-LSL-TdTomato allele in the mice. BF, bright
field.
(G) Mice with increased Tks5long expression (n = 15) developed more distant metastases
compared to control mice (n = 8). P < 0.03, Chi-square test; P < 0.04, Fisher’s exact test;
two tailed.
(H) Mice with increased Tks5long expression (n = 15) developed more metastases overall
(including distant metastases and pleural metastases) compared to control mice (n = 8). P <
0.003, Chi-square test; P < 0.008, Fisher’s exact test; two tailed.
95
Elevated expression of Tks5short reduces gelatin-matrix degradation and invadopodia
lifetime
In addition to modulating Tks5long expression, the Tks5long-to-Tks5short ratio can also
be altered by increasing the expression of the less well-studied isoform, Tks5short. While
Tks5short levels in TnonMet and TMet/ Met cells were similar, we were interested whether
Tks5short could also participate in regulating invadopodia activity. Although we attempted to
knockdown Tks5short by shRNA, this approach proved to be challenging since the unique
sequence in the Tks5short transcript is very short. Therefore, we chose to infect TMet cells
with a lentivirus that allowed doxycycline-inducible expression of HA-tagged Tks5short (Figure
5A). Interestingly, two independent TMet cell lines with exogenous Tks5short expression both
had drastically reduced gelatin-matrix proteolysis (Figure 5B), suggesting that Tks5short may
act in a dominant-negative manner over Tks5long in regulating invadopodia activity.
We additionally determined the localization of Tks5short by using an antibody specific
for the HA-tag, and found that HA-Tks5short exhibited a diffuse localization in TMet-Tks5short
cells (Figure 5C). This distribution is in contrast to that of endogenous Tks5 in TMet cells
stained with a pan-Tks5 antibody, where the protein (presumably the dominant Tks5long
isoform in TMet cells) is concentrated at foci of cortactin/F-actin colocalization (Figure 5C).
The diffuse distribution of Tks5short suggests that it cannot localize to invadopodia foci on the
cell membrane, conceivably due to lack of a PX homology domain.
Moreover, even though cortactin/F-actin–positive foci of invadopodia could be
observed in TMet-Tks5short cells (Figure 5C), when we measured the lifetime of invadopodia in
these cells using time-lapse fluorescence imaging, they exhibited shortened invadopodia
lifetime compared to parental TMet cells (Figure 5D). Previous studies have shown that
nascent invadopodia remain non-proteolytic for at least one hour before maturing into fully
functional invadopodia that are capable of mediating degradation (Yamaguchi et al. 2005;
96
Oser et al. 2009). Thus, our data suggest that Tks5short may destabilize invadopodia and
interfere with their maturation into a proteolytic state. Hence, the ratio between Tks5short and
Tks5long, rather than the absolute levels of either isoform or total Tks5, appears to be
important for invadopodia-mediated cell invasion.
Figure 5. Tks5short negatively regulates extracellular matrix degradation and reduces
invadopodia lifetime.
(A) Immunoblot detection of Tks5 isoforms in 373T1 TMet and TMet-HA-Tks5short cells with or
without doxycycline induction of HA-Tks5short expression. Tubulin is a loading control.
(B) Increased Tks5short expression in two independent TMet cell lines (373T1, 393T3) impairs
gelatin-matrix proteolysis. Both TMet-Tks5short cells and parental TMet cells were treated with 2
µg/mL doxycycline. Areas of degradation were quantified using ImageJ and normalized to
number of cells per field. At least 40 fields containing a total of 1300 cells were analyzed per
condition. Values are mean ± SEM; (***), P-value< 0.001, Student’s T-test.
(C) Immunofluorescence staining of total Tks5 in 373T1 TMet cells and of HA-Tks5short in
373T1 TMet-HA-Tks5short cells, both treated with 2 µg/mL doxycycline. Invadopodia were
stained by cortactin (green) and F-Actin (red). Magnified views of the regions indicated by
the boxed area are shown below.
(D) Invadopodia lifetime in 373T1 TMet- Tks5short cells are generally shorter than control
373T1 TMet cells as measured by life-cell fluorescence imaging. More than 150 invadopodia
from three independent measurements were analyzed per condition.
97
High expression of Tks5long and low expression of Tks5short correlate with metastatic
progression and poor survival in lung adenocarcinoma patients
We next explored whether a high Tks5long-to-Tks5short ratio correlates with tumor
progression and metastasis in human lung cancer. For this purpose, we analyzed RNA-seq
data from lung adenocarcinoma patients deposited in The Cancer Genome Atlas (TCGA) for
the expression ratio of Tks5long-to-Tks5short. The human Tks5short transcript homologous to
mouse Tks5short was the most abundant isoform alternative to Tks5long in these lung tissues,
although additional transcripts with slight variations may exist in other tissue types
(http://genome.ucsc.edu/; assembly GRCh37/hg19, gene Sh3pxd2a). Importantly, while the
expression patterns of Tks5long and Tks5short in these lung adenocarcinomas were diverse,
there was a trend towards high Tks5long-to-Tks5short expression ratios in patients with stage
III and IV disease (characterized by metastatic invasion in the thoracic cavity and distant
organs, respectively; n = 59) compared to patients with stage IA disease (characterized by a
single, small, localized lesion without detectible metastases; n = 57) (P-value < 0.009, Chisquare test; P < 0.013, Fisher’s exact test; Figure 6A), suggesting that high Tks5long and low
Tks5short expression contributes to promoting metastatic progression.
In addition, we examined Tks5long and Tks5short expression in an independent cohort
of 102 patients with stage I/II lung adenocarcinoma from the University of Michigan.
Interestingly, we observed that higher Tks5long expression and lower Tks5short expression
correlated with worse disease-free survival and overall survival by Kaplan-Meier analysis
(Figure 6B and Supplementary Figure S5), and reflected poor prognosis in a
multivariate analysis by the Cox proportional hazard model after adjustment for gender, age,
stage, and tumor differentiation state (Tables 1 and 2). Importantly, total expression of Tks5
did not demonstrate any survival correlation or prognostic values in these analyses (Figure
6B and Table 1; Supplementary Figure S5 and Table S1), suggesting that the distinction
98
between Tks5 isoforms is critical in analyzing these clinical data. As disease-free survival,
and to a lesser extent overall survival, reflect the rate of post-resection tumor relapse and
thus most likely the magnitude of early micrometastatic spread prior to surgery, our data are
consistent with the conclusion that Tks5long promotes metastasis in human lung cancer,
while Tks5short exerts the opposite effect, and a shift in the balance of the two isoforms may
influence the clinical outcomes in lung adenocarcinoma patients. In addition, Tks5long and
Tks5short may serve as prognostic factors for identifying high-risk patients with early stage
disease who may benefit from adjuvant treatment following tumor resection.
Table 1. Hazard ratios from multivariate Cox-model analysis
of 5-year disease-free survival.
Tks5long
Tks5short
Tks5long:Tks5total
Tks5total
Hazard Ratio (95% CI)
2.1 (1.4-3.2)
0.5 (0.4-0.8)
3.0 (1.6-5.9)
1.1 (0.8-1.7)
P-value
0.0003 ***
0.006 **
0.0009 ***
0.5
(ns)
Analysis was adjusted to age, gender, stage, and differentiation.
(CI, confidence interval; ns, not significant.)
Table 2. Hazard ratios from multivariate Cox-model analysis
of 5-year overall survival.
Tks5long
Tks5short
Tks5long:Tks5total
Tks5total
Hazard Ratio (95% CI)
2.8 (1.7-4.6)
0.5 (0.3-0.9)
2.5 (1.2-4.9)
1.4 (0.8-2.3)
P-value
0.00003 ***
0.01
*
0.01
*
0.2
(ns)
Analysis was adjusted to age, gender, stage, and differentiation.
(CI, confidence interval; ns, not significant.)
99
Figure 6. A high Tks5long-to-Tks5short ratio correlates with metastasis and poor survival
in lung adenocarcinoma patients.
(A) Ratio of Tks5long-to-Tks5short expression in primary lung adenocarcinomas of stage IA
patients (n = 57) and stage III/IV patients (n = 59). P-value < 0.009, Chi-square test; P <
0.013, Fisher’s exact test; two tailed.
(B) Five-year disease-free survival of stage I/II lung adenocarcinoma patients correlates with
Tks5long and Tks5short expression, but not total expression of Tks5. Patients (n = 102) were
divided into two groups based on expression level (high = top two-thirds; low = bottom onethird). All P-values are from Log-rank test.
100
CONCLUSIONS
Metastasis accounts for the vast majority of cancer related deaths, underscoring the
need for a better understanding of the molecular mechanisms that enable tumor cells to
escape from their primary site and spread to other parts of the body. In this study, we report
a shift in the isoform expression of an invadopodia component Tks5 that helps explain the
increased invasiveness of metastatic cells during lung adenocarcinoma progression. Our
data indicate that as primary tumors progress from a non-metastatic state (TnonMet) to a
metastatic state (TMet) and eventually form secondary lesions (Met), tumor cells acquire an
increase in Tks5long-to-Tks5short expression, despite a lack of significant increase in total
Tks5 expression. Using functional experiments in cultured cells and mouse models, we
demonstrate distinct and opposing roles for Tks5long and Tks5short. Tks5long promotes
invadopodia activity and metastasis formation, as knockdown of Tks5long impairs
invadopodia function in vitro and metastasis formation in vivo, while elevated expression of
Tks5long has the opposite effects. Tks5short on the other hand acts as a negative regulator of
invadopodia function, as increased expression of Tks5short interferes with invadopodia
stability and inhibits gelatin proteolysis. Hence, it is the balance of Tks5long and Tks5short
expression, rather than total Tks5 level, that appears to be important for metastatic invasion.
Consistent with these functional analyses, our clinical data demonstrate that high level of
Tks5long expression and low level of Tks5short expression (but not total Tks5 expression)
correlate with metastatic progression in lung adenocarcinoma patients, and predict poor
survival of patients with early-stage disease.
These experiments provide insight into the roles of Tks5 isoforms in metastasis.
Previous studies have demonstrated a role of Tks5 in invadopodia and invasion (Seals et al.
2005; Blouw et al. 2008); however, the specific roles of its isoforms have not been defined.
101
Our data revise the current notion that Tks5 generally promotes invadopodia and
metastasis, and support a model in which Tks5long acts to promote invadopodia formation by
binding to the cellular membrane and recruiting effector proteins for actin polymerization and
protease secretion, while Tks5short acts to regulate invadopodia function by interfering with
their stability and maturation. A shift in the balance of Tks5long and Tks5short expression in
cancer cells may have a profound impact on tumor progression. While our data indicate that
Tks5short interferes with invadopodia stability, the specific mechanism of this regulation
remains to be elucidated. It is conceivable that Tks5short acts by sequestering invadopodia
components away from the cell membrane via its multiple SH3 domains, proline-rich
regions, and phosphorylation sites. Previous biochemical assays have shown that these
functional domains of Tks5 bind to multiple invadopodia components, including N-WASP,
Nck, and ADAM family metalloproteases (Abram et al. 2003; Oikawa et al. 2008; Stylli et al.
2009). Whether these protein interactions mediate the inhibitory function of Tks5short in
invadopodia deserves future investigation. In addition, given the differential isoform
expression of Tks5 in metastatic and non-metastatic tumors, it will be of great interest to
identify the regulatory mechanism of this isoform switch. Data from our 5’RACE and
H3K4me3 ChIP-seq analyses suggest that the two isoforms are transcribed from distinct
promoters. It will be important to dissect the regulatory mechanisms of promoter choice that
lead to the increased expression of Tks5long during tumor progression.
This study also underscores the in vivo role of invadopodia as critical mediators of
metastasis in natural tumor progression. While previous studies have provided important
evidence for a role of invadopodia in mediating metastatic invasion by using cell culturebased invasion assays and transplant models (Seals et al. 2005; Blouw et al. 2008;
Philippar et al. 2008; Eckert et al. 2011; Gligorijevic et al. 2012), these experimental systems
often do not fully recapitulate natural tumor progression and metastatic spread. Here we
102
further establish the role of invadopodia in promoting metastasis in vivo by using an
autochthonous mouse model of metastatic lung cancer. Our data thus help address the
question of whether invadopodia play a physiologically relevant role in metastasis in vivo
(Linder 2009; Sibony-Benyamini and Gil-Henn 2012). The data presented by our study and
others provide a mechanism that explain one of the ways in which tumor cells might
overcome the multiple physical barriers presented by stromal tissues, the extracellular
matrix, and endothelial cells during the intravasation, extravasation and colonization steps of
metastasis. Interestingly, while the gain of Tks5long expression in both TMet and Met cells in
our lung adenocarcinoma model suggests that invadopodia function in both the
invasion/intravasation step at the primary tumor and the extravasation/colonization step at
the metastatic site, the specific contribution of invadopodia to each step of the metastasis
cascade may be context and cell-type dependent. Whereas knockdown of Tks5long in lung
adenocarcinoma cells (this study) and knockdown of total Tks5 in Ras-transformed
mammary epithelial cells (Eckert et al. 2011) diminished the number of metastases formed,
similar total-Tks5 knockdown experiments in Src-transformed mouse embryonic fibroblasts
did not lead to more metastases, but an increase in the volume and vascularization of
metastatic nodules (Blouw et al. 2008). Thus the specific contribution of invadopodia to each
step of the invasion-metastasis cascade in different cancer types remains to be further
dissected in the future.
In addition, this study carries clinical implications for lung cancer patients. Previous
clinical studies of another invadopodia component cortactin indicate that its elevated
expression and dysregulated cellular localization correlate with poor survival in laryngeal
carcinoma and lung adenocarcinoma, respectively, underlining the relevance of invadopodia
activity in predicting clinical outcomes (Gibcus et al. 2008; Hirooka et al. 2011). Consistent
with these studies, our data show that high level of Tks5long expression and low level of
103
Tks5short expression correlate with metastatic progression of lung adenocarcinoma patients,
and predict poor survival of patients with early-stage disease, suggesting that the Tks5longto-Tks5short ratio may serve as a prognostic marker for assessing the metastatic potential of
primary tumors and for identifying early-stage patients who bear higher risks for metastasis
and may benefit from adjuvant therapy after tumor resection. Furthermore, future
development of molecular therapeutic strategies that inhibit Tks5long function or strengthen
Tks5short activity could potentially help inhibit metastatic progression.
Finally, this study demonstrates the value of mouse models and their derivative cell
lines in allowing molecular characterization of the cell state changes that accompany tumor
progression and metastasis, and is representative of recent developments of animal models
for lung adenocarcinoma (Jackson et al. 2005; Politi et al. 2006; Dankort et al. 2007; Ji et al.
2007). Importantly, our approach demonstrates that in order to fully understand the
metastatic process, analysis of gene expression at the isoform level in addition to the total
gene expression level is important. Given the lethal effects of the metastatic phase of
cancer, a deeper insight into the molecular determinants of metastasis will have a significant
impact on cancer mortality and morbidity.
104
MATERIALS AND METHODS
Cell lines
TnonMet, TMet, and Met cell lines were derived from autochthonous tumors in KrasLSL-G12D/WT;
p53flox/flox mice as described previously (Winslow et al. 2011). The MIT Institutional Animal
Care and Use Committee approved all animal studies and procedures. Briefly, lung tumors
were initiated via intratracheal delivery of a lentiviral vector expressing Cre recombinase.
Primary tumors and metastases were harvested at 6-14 months post-infection, and used to
establish cell lines. Each metastasis-derived cell line was matched to its primary tumorderived cell line based on analysis of the unique lentiviral integration site using Southern
blotting or linker-mediated PCR. Thus metastatic primary tumors that had matching
secondary lesions could be distinguished from non-metastatic primary tumors. All cell lines
were cultured in complete media (DMEM with 10% FBS, 50 U/mL penicillin, and 50 mg/mL
streptomycin). Five TnonMet cell lines (368T1, 393T1, 394T4, 802T4, 2557T1), six TMet cell
lines (373T1, 373T2, 389T2, 393T3, 393T5, 482T1) and five Met cell lines (373N1, 393N1,
393M1, 482N1, 482M1) were used for subsequent gene expression analysis and/or
functional experiments in this study.
Exon arrays and differential isoform expression detection
Affymetrix GeneChip Mouse Exon 1.0 ST arrays (Gene Expression Omnibus GSE26874) of
four TnonMet (368T1, 393T1, 802T4, 2557T1) and six TMet (373T1, 373T2, 389T2, 393T3,
393T5, 482T1) cell lines were analyzed for transcriptome-wide isoform switches between
groups TnonMet and TMet using the Partek Genomics Suite software package (v6.4) with a
custom collection of 345,117 probe-sets (~22,000 genes) (Winslow et al. 2011). In
summary, Partek data were post-processed using a custom protocol to rank genes based
105
on a combination of (i) statistical significance in Partek's ANOVA-based test for alternative
isoform expression, (ii) robustly detectible gene expression in both TnonMet and TMet groups,
and (iii) significant deviation of probe intensity difference between TnonMet and TMet groups for
a single probe-set, compared to all probe-sets of a given gene. High-ranking genes were
then further evaluated by manual examination of probe set-based expression profiles in
TnonMet and TMet groups.
qRT-PCR
RNA was purified from cultured cells using RNAqueous kit (Invitrogen), according to the
manufacturer’s instructions. Two micrograms of RNA was reverse-transcribed using a HighCapacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time quantitative PCR
reactions were performed using SYBR Green Jumpstart Taq Ready Mix (Sigma) and an ABI
Prism thermocycler (Applied Biosystem). All gene expression was shown relative to TBP
control. Primers for qPCR were: Tks5long forward: 5’-TTA TCA ACG TGA CCT GGT CTG-3’,
Tks5long reverse: 5’- TTC GGA TCC TTC TGG CCA C -3’; Tks5short forward: 5’-TGG CTC
ACC GCG TGC TTT CTG-3’, Tks5short reverse: 5’- CCT TGC TCT TCA GAT GTG CTC ACA
A-3’; TBP forward: 5’-GGG GAG CTG TGA TGT GAA GT-3’; TBP reverse: 5’- CCA GGA
AAT AAT TCT GGC TCA-3’.
Immunoblotting
The following antibodies were used for immunoblotting: Tks5 (1:1000, Santa Cruz M300;
detects both Tks5long and Tks5short), Tubulin (1:10 000, Cell Signaling 3873), Flag (1:300,
Cell Signaling 2368), and HA (1:500, Cell Signaling 3724).
106
cDNA expression and knockdown
To generate a doxycycline-responsive lentiviral expression vector of N-terminally Flagtagged Tks5long for infecting cultured cells, mouse Tks5long cDNA was PCR amplified from a
pcDNA3-Tks5 plasmid (a generous gift from S. Courtneidge, CA) using forward primer 5'TAC ATC GTT AAC GCC ACC ATG CTC GCC TAC TGC GTG CAA G-3' and reverse
primer 5'-TAC ATC TTA ATT AAT TAC TTG TCG TCG TCG TCC TTG TAG TCG TTC TTC
TTC TCA AGG TAG TTG GAG-3'. The amplicon was subsequently digested using HpaI and
PacI, and cloned into a lentiviral expression vector pCW22tre-optimegaUbcrtTA, which
contains a TRE promoter for doxycycline-induction of the cDNA and a UBC promoter for
constitutive expression of rtTA. The construct was used to infect TnonMet cells.
To generate a doxycycline-responsive lentiviral expression vector of N-terminally HAtagged Tks5short, we performed two-round PCR amplification on TnonMet cDNA using primers
5’-CGG TGC AGA GCT GGC GAC CGA-3’ and 5’-AGT GGC AGC CAA GGC AGC ACG
TT-3’, followed by nested primers 5’- GAG CTG GCG ACC GAG CAG CCT-3’ and 5’-GCA
GCC AAG GCA GCA CGT TGA GT-3’. The Tks5short amplicon was cloned into pCRII-TOPO
(Invitrogen) and sequence verified. The pCRII-TOPO-Tks5short plasmid was then PCR
amplified using primers 5’-TAC ATC GTT AAC GCC ACC ATG GAC AGA GGG CGC CCC
GGC-3’ and 5’-TAC ATC TTA ATT AAT TAG GCG TAG TCA GGC ACG TCG TAA GGA
TAG TTC TTC TTC TCA AGG TAG TTG GAG-3’, followed by HpaI/PacI digestion and
cloning into a lentiviral expression vector pCW22tre-optimegaUbcrtTA. The construct was
used to infect TMet cells.
To knock down Tks5long, shRNAs targeting the 5’ unique region (exons1-7) of
Tks5long were designed using http://gesteland.genetics.utah.edu/siRNA_scales/ and
resources available through G. Hannon’s laboratory at Cold Spring Harbor Laboratories
(http://katahdin.cshl.org/homepage/siRNA). Seven shRNA sequences were cloned into the
107
miR30 sequence and tested for effective knockdown. The oligonucleotides were PCR
amplified using forward primer 5’-CAG AAG GCT CGA GAA GGT ATA TTG CTG TTG ACA
GTG AGC G-3’ and reverse primer 5’-CTA AAG TAG CCC CTT GAA TTC CGA GGC AGT
AGG CA-3’. The amplicon was then digested by XhoI and EcoRI, and cloned into the
retroviral vector MSCV-Hygro. The two oligonucleotides that gave the best Tks5long
knockdown were sh1: 5’-TGC TGT TGA CAG TGA GCG CCT GGA TAA GTT TCC TAT
TGA ATA GTG AAG CCA CAG ATG TAT TCA ATA GGA AAC TTA TCC AGA TGC CTA
CTG CCT CGG A-3’, and sh2: 5’-TGC TGT TGA CAG TGA GCG ACA CAT TTC ACA GTG
TGA CGA ATA GTG AAG CCA CAG ATG TAT TCG TCA CAC TGT GAA ATG TGG TGC
CTA CTG CCT CGG A-3’. These were used to knock down Tks5long in TMet cells. An
oligonucleotide targeting luciferase (5’- TGC TGT TGA CAG TGA GCG CCC GCC TGA
AGT CTC TGA TTA ATA GTG AAG CCA CAG ATG TAT TAA TCA GAG ACT TCA GGC
GGT TGC CTA CTG CCT CGG A-3’) was used as a negative control.
Viral production and infection of TMet and TnonMet cells
To produce lentivirus containing Flag-tagged Tks5long or HA-tagged Tks5short, the lenti-vector
was co-transfected with packaging vectors delta8.2 and VSV-G (gifts from D. Trono) into
293T cells using TransIT-LT1 (Mirus Bio). To produce MSCV containing shTks5long, the
MSCV vector was transfected into phoenix cells using TransIT-LT1 (Mirus Bio). In both
cases, the resultant supernatant was collected at 48 hr and 72 hr post-transfection, and
used to infect TnonMet or TMet cell lines. Infected cells were selected for one week using either
8 µg/mL Blasticidin (in the case of lentiviral infection), or 800 µg/mL Hygromycin (in the case
of MSCV infection).
108
Creation of Lenti-Cre/Tks5long vector
To generate a lentiviral expression vector that allows constitutive expression of Cre
recombinase and doxycycline induction of N-terminally Flag-tagged Tks5long in our mouse
model of lung adenocarcinoma, mouse Tks5long cDNA was PCR amplified from a pcDNATks5 plasmid (a generous gift from S. Courtneidge, CA) using forward primer 5'-TAC ATC
CAA TTG ATC AGC CAC CAT GCT CGC CTA CTG CGT GCA AG-3' and reverse primer
5'-TAC ATC GGC GCG CCT TAC TTG TCG TCG TCG TCC TTG TAG TCG TTC TTC TTC
TCA AGG TAG TTG GAG-3'. The amplicon was subsequently cloned into a lentiviral
expression vector pCW22treoptimegaPgkCre, which contains a TRE promoter for
doxycycline-induction of Tks5long and a PGK promoter for constitutive expression of Cre. The
plasmid was used to initiate tumors with inducible Tks5long expression in mice. An empty
pCW22treoptimegaPgkCre plasmid (lenti-Cre) was used as a negative control.
FITC-gelatin degradation assay
Glass-bottomed 35-mm plates (MatTek) were coated with FITC-gelatin as described in
(Bowden et al. 2001) with some modifications. Briefly, MatTek plates were treated with HCl,
followed by 50 µg/mL poly-L-lysine, and then coated with a thin layer of FITC-labeled 0.2%
gelatin (Sigma) for 1 hr. The gelatin coating was then crosslinked with ice-cold 0.8%
glutaraldehyde (Electron Microscopy Sciences)/PBS for 15 min at 4°C and then for 30 min
at room temperature. Plates were successively washed in PBS (3 × 5 min), 5 mg/mL sodium
borate in PBS (1 × 3 min), and PBS (3 × 5 min), before being incubated for 30 min with
complete tissue culture media. Cells (8x104) were cultured on the gelatin-coated plates for
72 hrs and subsequently processed using standard fluorescence microscopy procedures.
109
Immunofluorescence staining for invadopodia components
Cells (8x104) were grown at 37˚C overnight on 35-mm MatTek plates coated with either
0.2% FITC-labeled gelatin or 0.2% plain gelatin (Sigma). Cells were then fixed in 3.7%
formaldehyde (Electron Microscopy Sciences) in PBS for 20 min, permeabilized with 0.1%
Triton X-100 in PBS for 5 min, and blocked with 1% BSA and 1% FBS in PBS. Subsequently
cells were stained for immunofluorescence microscopy. Primary antibodies include: Tks5
(1:100, Santa Cruz M300; detects both Tks5long and Tks5short), HA-tag (1:100, Cell Signaling
3724), and cortactin (1:100, Millipore 4F11). F-actin was stained with phalloidin (Invitrogen).
Live cell fluorescence microscopy
Cells transfected with a pcDNA3 RFP-β-Actin construct (a generous gift from F. Gertler, MA)
were plated on gelatin-coated MatTek dishes in L-15 Medium (Leibovitz), placed in an
environmental chamber with constant 37˚C temperature, CO2, and humidity, and imaged
every 2.5 minutes for at least 12 hours. The lifetimes of at least 150 invadopodia from three
independent measurements per condition were analyzed using ImageJ.
Subcutaneous transplantation
Nude mice were injected with 5x104 TMet cells (393T3 parental cells, or 393T3 cells
expressing sh1 short hairpin RNA against Tks5long) resuspended in 100 µl PBS under the
skin on their hind flank. Subcutaneously injected mice were analyzed 8 weeks after
injection. To quantify lung tumor nodules all the visible surface tumors were counted under a
dissecting microscope.
110
Intrasplenic transplantation
Nude mice were injected with 5x104 TMet cells (373T1 parental cells, or 373T1 cells
expressing sh1 short hairpin RNA against Tks5long) re-suspended in 200 µl PBS via the
spleen, as described previously (Winslow et al. 2011). Briefly, the animals were given
0.1mg/kg Buprenorphine prior to surgery, and anaesthetized with continuous flow of
isoflurane throughout the procedure. Once the animals were under deep anesthesia, the
abdominal area was disinfected with Betadine and 70% ethanol. The spleen was exposed
through a small incision. Cells were injected into the spleen with a single injection using an
insulin syringe. Cells were given 10 min to travel through the vasculature to the liver, after
which the entire spleen was removed to prevent the formation of a large splenic tumor
mass. To remove the spleen, a dissolvable 4-0 suture was tied snugly around the base of
the spleen including the major splenic vasculature and the spleen was removed. The muscle
wall was closed with 4-0 dissolvable sutures, and the skin incision closed with sterile 7-mm
wound clips (Roboz). Intrasplenically injected mice were analyzed 3 weeks after injection.
Quantification of liver tumor nodules was performed by counting all the visible surface
tumors under a dissecting microscope.
Lentiviral infection of autochthonous mouse model of lung adenocarcinoma
Tumors were initiated by intratracheal infection of mice as described previously (DuPage et
al. 2009). Lentivirus was produced from 293T cell transfection as described above. Virus
was recovered from the supernatant by ultracentrifugation at 25,000 rpm for 90 min, and resuspended in an appropriate volume (200-2000 µl) of PBS. A lentiviral dose of 1000-4000
viral particles induced 25-50 lung tumors per mouse and allowed 6 month survival after
tumor initiation, while a lentiviral dose of 500 viral particles induced ~10 tumors per mouse
and allowed 8 month survival post initiation.
111
Histology and immunohistochemistry
Tissues for histology were fixed in 3.7% formalin in PBS for 24 hours and stored in 70%
ethanol until paraffin embedding. Histological analysis for tumor grade was performed on
formalin-fixed, paraffin-embedded 4-µm sections stained with haematoxylin and eosin.
Immunohistochemistry was performed using the ImmPRESS kit (Vector Lab MP-7401), and
antibodies to phospho-histone H3 (1:200, Cell Signaling) and cleaved caspase 3 (1:1000,
Cell Signaling). Sections were developed with DAB (Vector Lab SK-4100) and
counterstained with haematoxylin. Mitotic index (number of cells per mm2 of tumor area
stained positive for phosphorylated histone H3) and apoptotic index (number of cells per
mm2 of tumor area stained positive for cleaved caspase 3) were quantified.
Clinical analysis
For TCGA dataset, Tks5 isoform expression ratio was obtained from RNAseq alignments of
305 human lung adenocarcinoma samples. SAM files that described the alignments of
RNAseq reads for each sample to the Tks5 locus were obtained and used to calculate the
average depth of coverage for each exon in the Tks5 gene. Tks5long expression was
measured by calculating the average depths of coverage for three consecutive exons near
the 5’ end of the gene (hg19 coordinates chr10:105484028-105484119, 105495490105495566, and 105526852-105526927). Total Tks5 expression (Tks5long + Tks5short) was
measured by calculating the average of three consecutive constitutive exons near the 3’ end
of the gene (hg19 coordinates chr10:105365555-105365674, 105371338-105371387, and
105372610-105372947). An index of Tks5 isoform expression was calculated by [total Tks5/
Tks5long]. A low value of the index represents high expression of Tks5long relative to Tks5short.
Chi-square test and Fisher’s exact test were performed on patients with stage IA disease
(n=57) and stage III/IV disease (n=59) using a cutoff of Tks5 isoform index = 3.6. Patients
112
with Tks5 isoform index of <3.6 were categorized as high Tks5long expression relative to
Tks5short, while patients with Tks5 isoform index of >3.6 were categorized as low Tks5long
expression relative to Tks5short.
For the University of Michigan dataset, Tks5 isoform expressions were measured by
qRT-PCR in 102 primary tumor samples from patients with stage I/II lung adenocarcinoma
without preoperative radiation or chemotherapy. Tissue specimens were obtained with
informed consent after approval from University of Michigan Institutional Review Board and
Ethics Committee. The qPCR primers for human Tks5long are: forward 5’-TGT GAC CTG
GTC TGA CTC CA-3’ and reverse 5’-GTC CTT CTG GCC ACC TTC AA-3’. Primers for total
human Tks5 are: forward 5’-TGC CAA GAA GGA GAT CAG CC -3’ and reverse 5’-TGG
AGG TCT TGT CCG TAG GT-3’. These qRT-PCR data were standardized by β-actin
expression. Levels of human Tks5short expression were calculated by subtracting Tks5long
expression from total Tks5 expression. Based on expression level, patients were divided
into high-expression group (top 2/3) and low-expression group (bottom 1/3) for Tks5long,
Tks5short, total Tks5, or Tks5long-to-total Tks5 ratio. Five-year disease-free survival and
overall survival were analyzed by Kaplan-Meier curves and log-rank test.
Multivariate analysis by the Cox proportional hazard model (adjusted by gender, age, stage,
and tumor differentiation state) was performed using a continuous value of Tks5 mRNA level
to assess survival results. P values (two-tailed) < 0.05 were considered statistically
significant.
Statistical analysis
All statistical analyses were performed using Student’s T-test, unless otherwise specified. Pvalues <0.05 (one-tailed) were considered statistically significant.
113
ACKNOWLEDGEMENTS
This work was supported by a National Institutes of Health grant (5-U01-CA84306) and a
National Cancer Institute grant (P30-CA14051). T.J. is a Howard Hughes Investigator, the
David H. Koch Professor of Biology, and a Daniel K. Ludwig Scholar. M.M.W. is funded by
National Institutes of Health grants (R00-CA151968 and R01-CA175336). C.M.L. is funded
by the Ludwig Center for Molecular Oncology Graduate Fellowship. We thank the Swanson
Biotechnology Center, and especially Denise Crowley and Eliza Vasile, for technical
support. We thank Frank Gertler and Sara Courtneidge for generous sharing of reagents;
Angela Brooks and Matthew Meyerson for assistance with TCGA data; Michele Balsamo
and Russell McConnell for technical support; Nadya Dimitrova, David Feldser, David
McFadden, Thales Papagiannakopoulos, Tuomas Tammela, Wen Xue, Vasilena Gocheva,
Irene Blat, Keara Lane, Kim Mercer, Megan Heimann, and the entire Jacks lab for advice
and experimental assistance.
114
SUPPLEMENTAL FIGURES
Supplementary Figure S1. Generation of TnonMet and TMet cell lines from lung
adenocarcinomas in an autochthonous mouse model.
The genetically engineered mouse model for lung adenocarcinoma used in this study carries
KrasLSL-G12D/WT and p53flox/flox alleles. Lentiviral Cre recombinase mediates activation of
oncogenic Kras and deletion of tumor suppressor p53 in lung epithelial cells, thus initiating
lung adenocarcinomas. Over 6-14 months, a subset of these tumors metastasizes to local
lymph nodes and distant organs. Because each metastasis can be matched to its primary
tumor based on the lentiviral integration site, metastatic primary tumors (indicated in orange
and red) that have seeded secondary lesions can be distinguished from non-metastatic
tumors (indicated in green) that do not have matching metastases. Cell lines have been
derived from these non-metastatic primary tumors as well as from the metastatic primary
tumors and their matching metastases, and are termed TnonMet, TMet, and Met respectively.
115
Supplementary Figure S2. Differential Tks5long and Tks5short expression in TnonMet and
TMet cell lines.
Intensities of Tks5 exon-specific probe-reads from an Affymetrix exon array analysis of four
TnonMet cell lines and six TMet cell lines suggest that Tks5 is expressed as two distinct
isoforms: Tks5long (containing exons 1-15) and Tks5short (containing exons 8-15). Probe-sets
for exons 4 and 7 were not included in the analysis because of low binding intensities
indicative of poorly performing probes.
116
Supplementary Figure S3. Tks5long knockdown impairs invadopodia activity in vitro
and does not affect cell proliferation in vivo.
(A-C) Tks5long knockdown impairs invadopodia activity in vitro.(A) Immunofluorescence
staining shows that colocalization of invadopodia components (cortactin in green, and Factin in red) and FITC-negative areas of gelatin-matrix degradation are less frequently
observed in TMet-shTks5long cells compared with TMet cells expressing control shRNA.
Magnified views of the regions indicated by the boxed area are shown to the right.
Representative images from TMet cells (393T3) are shown. (B) Effects of Tks5long knockdown
on invadopodia formation in TMet cells (393T3). At least 100 cells were scored for
colocalization of cortactin and F-actin in three independent experiments. (C) Effects of
Tks5long knockdown on FITC-gelatin matrix degradation in TMet cells (393T3). Areas of
degradation were quantified using ImageJ and normalized to number of cells per field. At
least 50 fields containing a total of 500 cells were analyzed per condition.
(D-F) Tks5long knockdown does not affect tumor cell proliferation in vivo. (D) Effects of
Tks5long knockdown on size of subcutaneous tumors. Values are mean ± SEM, P-value >
0.28, Student’s T-test. (E) Mitotic index (number of cells per mm2 of tumor area stained
positive for phosphorylated histone H3 by immunohistochemistry) of liver nodules formed 3
weeks after intrasplenic transplant. Values are mean ± SEM, P-value > 0.15, Student’s Ttest. (F) Apoptotic index (number of cells per mm2 of tumor area stained positive for cleaved
caspase 3 by immunohistochemistry) of liver nodules formed 3 weeks after intrasplenic
transplant. Values are mean ± SEM, P-value > 0.11, Student’s T-test.
117
Supplementary Figure S4. Tks5long is sufficient to promote invadopodia activity in
vitro and invasive tumor progression in vivo without affecting primary lung tumor
burden.
(A-B) Tks5long is sufficient to promote invadopodia activity in vitro. (A) Effects of increased
Tks5long expression on invadopodia formation in 368T1 TnonMet cells. At least 100 cells were
scored for colocalization of cortactin and F-actin in three independent experiments. All
values are mean ± SEM. (*), P-value< 0.05; paired T-test. (B) Effects of increased Tks5long
expression on FITC-gelatin matrix degradation in 368T1 TnonMet cells. Areas of degradation
were quantified using ImageJ and normalized to number of cells per field. At least 15 fields
containing a total of 600 cells were analyzed per condition. All values are mean ± SEM. (**),
P-value< 0.01: Student’s T-test.
(C-D) Tks5long is sufficient to promote invasive tumor progression in vivo without affecting
primary lung tumor burden. (C) Immunoblot detection of Tks5 isoforms and Flag-tagged
Tks5long in autochthonous lung adenocarcinomas. KrasLSL-G12D/WT; p53flox/flox mice and KrasLSLG12D/WT
; p53flox/flox; CCSP-rtTA mice were infected with Lenti-Cre or Lenti-Cre/Flag-Tks5long,
and fed with normal rodent diet or doxycycline diet. Flag-Tks5long is expressed only in tumors
induced by Lenti-Cre/Flag-Tks5long in the presence of rtTA and doxycycline. Tubulin was
used as a loading control. (D) KrasLSL-G12D/WT; p53flox/flox; CCSP-rtTA mice infected with LentiCre/Tks5long developed a larger proportion of high-grade tumors (grades 3 and 4) at 6
months post-infection compared with control mice without Tks5long overexpression (including
Lenti-Cre/Tks5long infected mice with no rtTA allele or no doxycycline diet, and Lenti-Cre
118
infected mice). A total of 618 tumors from 23 mice were analyzed. P < 0.0001, chi-squared
test.
(E) Grade 4 tumors are characterized by stromal invasion of tumor cells near the blood
vasculature (marked by an asterisk) as observed in H&E staining. Scale bar = 50 µm.
(F) Areas of primary lung tumors in control mice and Tks5long-expressing mice are not
significantly different. P > 0.23, Student’s T-test.
(G) Lung tumor burdens, as measured by total area of primary lung tumors divided by total
lung area per mouse, in control mice and Tks5long-expressing mice are not significantly
different. P > 0.30, Student’s T-test.
119
Supplementary Figure S5. High Tks5long and low Tks5short expressions correlate with
poor survival of early-stage lung adenocarcinoma patients.
Five-year overall survival of stage I/II lung adenocarcinoma patients correlates with Tks5long
and Tks5short expression, but not total expression of Tks5. Patients (n = 102) were divided
into two groups based on expression level (high = top two-thirds; low = bottom one-third). All
P-values are from Log-rank test.
120
REFERENCES
Abram CL, Seals DF, Pass I, Salinsky D, Maurer L, Roth TM, Courtneidge SA. 2003. The
adaptor protein fish associates with members of the ADAMs family and localizes to
podosomes of Src-transformed cells. Journal of Biological Chemistry 278: 16844-16851.
Blouw B, Seals DF, Pass I, Diaz B, Courtneidge SA. 2008. A role for the
podosome/invadopodia scaffold protein Tks5 in tumor growth in vivo. European journal of
cell biology 87: 555-567.
Bowden ET, Coopman PJ, Mueller SC. 2001. Invadopodia: Unique methods for
measurement of extracellular matrix degradation in vitro. Methods in Cell Biology, Vol 63 63:
613-627.
Bowden ET, Onikoyi E, Slack R, Myoui A, Yoneda T, Yamada KM, Mueller SC. 2006. Colocalization of cortactin and phosphotyrosine identifies active invadopodia in human breast
cancer cells. Experimental Cell Research 312: 1240-1253.
Chen WT, Olden K, Bernard BA, Chu FF. 1984. Expression of transformation-associated
protease(s) that degrade fibronectin at cell contact sites. Journal of Cell Biology 98: 15461555.
Clark ES, Whigham AS, Yarbrough WG, Weaver AM. 2007. Cortactin is an essential
regulator of matrix metalloproteinase secretion and extracellular matrix degradation in
invadopodia. Cancer Research 67: 4227-4235.
Courtneidge SA. 2012. Cell migration and invasion in human disease: the Tks adaptor
proteins. Biochemical Society Transactions 40: 129-132.
Crimaldi L, Courtneidge SA, Gimona M. 2009. Tks5 recruits AFAP-110, p190RhoGAP, and
cortactin for podosome formation. Experimental Cell Research 315: 2581-2592.
Dankort D, Filenova E, Collado M, Serrano M, Jones K, McMahon M. 2007. A new mouse
model to explore the initiation, progression, and therapy of BRAFV600E-induced lung
tumors. Genes & development 21: 379-384.
DuPage M, Dooley AL, Jacks T. 2009. Conditional mouse lung cancer models using
adenoviral or lentiviral delivery of Cre recombinase. Nature Protocols 4: 1064-1072.
Eckert MA, Lwin TM, Chang AT, Kim J, Danis E, Ohno-Machado L, Yang J. 2011. Twist1induced invadopodia formation promotes tumor metastasis. Cancer Cell 19: 372-386.
Gibcus JH, Mastik M, Menkema L, de Bock GH, Kluin PM, Schuuring E, van der Wal JE.
2008. Cortactin expression predicts poor survival in laryngeal carcinoma. British Journal of
Cancer 98: 950-955.
Gligorijevic B, Wyckoff J, Yamaguchi H, Wang Y, Roussos ET, Condeelis J. 2012. N-WASPmediated invadopodium formation is involved in intravasation and lung metastasis of
mammary tumors. Journal of cell science 125: 724-734.
121
Hirooka S, Akashi T, Ando N, Suzuki Y, Ishida N, Kurata M, Takizawa T, Kayamori K,
Sakamoto K, Fujiwara N et al. 2011. Localization of the Invadopodia-Related Proteins
Actinin-1 and Cortactin to Matrix-Contact-Side Cytoplasm of Cancer Cells in Surgically
Resected Lung Adenocarcinomas. Pathobiology 78: 10-23.
Jackson EL, Olive KP, Tuveson DA, Bronson R, Crowley D, Brown M, Jacks T. 2005. The
differential effects of mutant p53 alleles on advanced murine lung cancer. Cancer Research
65: 10280-10288.
Ji H, Ramsey MR, Hayes DN, Fan C, McNamara K, Kozlowski P, Torrice C, Wu MC,
Shimamura T, Perera SA et al. 2007. LKB1 modulates lung cancer differentiation and
metastasis. Nature 448: 807-810.
Kelly T, Mueller SC, Yeh YY, Chen WT. 1994. Invadopodia promote proteolysis of a wide
variety of extracellular-matrix proteins. Journal of Cellular Physiology 158: 299-308.
Keshamouni V, Arenberg D, Kalemkerian G. 2009. Lung cancer metastasis: novel biological
mechanisms and impact on clinical practice. Springer, New York.
Linder S. 2007. The matrix corroded: podosomes and invadopodia in extracellular matrix
degradation. Trends in cell biology 17: 107-117.
Linder S. 2009. Invadosomes at a glance. Journal of Cell Science 122: 3009-3013.
Lock P, Abram CL, Gibson T, Courtneidge SA. 1998. A new method for isolating tyrosine
kinase substrates used to identify Fish, an SH3 and PX domain-containing protein, and Src
substrate. Embo Journal 17: 4346-4357.
Murphy DA, Courtneidge SA. 2011. The 'ins' and 'outs' of podosomes and invadopodia:
characteristics, formation and function. Nature reviews Molecular cell biology 12: 413-426.
Nakahara H, Mueller SC, Nomizu M, Yamada Y, Yeh YY, Chen WT. 1998. Activation of beta
1 integrin signaling stimulates tyrosine phosphorylation of p190(RhoGAP) and membraneprotrusive activities at invadopodia. Journal of Biological Chemistry 273: 9-12.
Oikawa T, Itoh T, Takenawa T. 2008. Sequential signals toward podosome formation in
NIH-src cells. The Journal of cell biology 182: 157-169.
Oser M, Yamaguchi H, Mader CC, Bravo-Cordero JJ, Arias M, Chen X, Desmarais V, van
Rheenen J, Koleske AJ, Condeelis J. 2009. Cortactin regulates cofilin and N-WASp
activities to control the stages of invadopodium assembly and maturation. The Journal of
cell biology 186: 571-587.
Philippar U, Roussos ET, Oser M, Yamaguchi H, Kim HD, Giampieri S, Wang YR, Goswami
S, Wyckoff JB, Lauffenburger DA et al. 2008. A Mena invasion isoform potentiates EGFinduced carcinoma cell invasion and metastasis. Developmental Cell 15: 813-828.
Politi K, Zakowski MF, Fan PD, Schonfeld EA, Pao W, Varmus HE. 2006. Lung
adenocarcinomas induced in mice by mutant EGF receptors found in human lung cancers
122
respond to a tyrosine kinase inhibitor or to down-regulation of the receptors. Genes &
development 20: 1496-1510.
Rodenhuis S, Slebos RJC, Boot AJM, Evers SG, Mooi WJ, Wagenaar SS, Vanbodegom
PC, Bos JL. 1988. Incidence and possible clinical-significance of K-ras oncogene activation
in adenocarcinoma of the human-lung. Cancer Research 48: 5738-5741.
Seals DF, Azucena EF, Pass I, Tesfay L, Gordon R, Woodrow M, Resau JH, Courtneidge
SA. 2005. The adaptor protein Tks5/Fish is required for podosome formation and function,
and for the protease-driven invasion of cancer cells. Cancer Cell 7: 155-165.
Sibony-Benyamini H, Gil-Henn H. 2012. Invadopodia: the leading force. European journal of
cell biology 91: 896-901.
Siegel R, Naishadham D, Jemal A. 2013. Cancer statistics, 2013. CA: a cancer journal for
clinicians 63: 11-30.
Stylli SS, Stacey TT, Verhagen AM, Xu SS, Pass I, Courtneidge SA, Lock P. 2009. Nck
adaptor proteins link Tks5 to invadopodia actin regulation and ECM degradation. Journal of
cell science 122: 2727-2740.
Takahashi T, Nau MM, Chiba I, Birrer MJ, Rosenberg RK, Vinocour M, Levitt M, Pass H,
Gazdar AF, Minna JD. 1989. P53 - a frequent target for genetic abnormalities in lungcancer. Science 246: 491-494.
Tarone G, Cirillo D, Giancotti FG, Comoglio PM, Marchisio PC. 1985. Rous-sarcoma virustransformed fibroblasts adhere primarily at discrete protrusions of the ventral membrane
called podosomes. Experimental Cell Research 159: 141-157.
Winslow MM, Dayton TL, Verhaak RG, Kim-Kiselak C, Snyder EL, Feldser DM, Hubbard
DD, DuPage MJ, Whittaker CA, Hoersch S et al. 2011. Suppression of lung
adenocarcinoma progression by Nkx2-1. Nature 473: 101-104.
Yamaguchi H, Lorenz M, Kempiak S, Sarmiento C, Coniglio S, Symons M, Segall J, Eddy R,
Miki H, Takenawa T et al. 2005. Molecular mechanisms of invadopodium formation: the role
of the N-WASP-Arp2/3 complex pathway and cofilin. Journal of Cell Biology 168: 441-452.
123
CHAPTER 3
Foxa2 and Cdx2 cooperate with Nkx2-1
to inhibit lung adenocarcinoma metastasis
Carman Man-Chung Li1, Arjun Bhutkar1, Vasilena Gocheva1, Madeleine J. Oudin1, Shi Yun
Wang1, Saya Date1, Sheng Rong Ng1, Charles A. Whittaker1, Roderick T Bronson2, Eric L.
Snyder3, Frank B. Gertler1, Tyler Jacks1,4
1 David H. Koch Institute for Integrative Cancer Research, Department of Biology,
Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
2 Department of Pathology, Tufts University School of Medicine and Veterinary Medicine,
North Grafton, Massachusetts 01536, USA
3 Departments of Pathology and Anatomy, School of Medicine, University of California, San
Francisco, California 94143, USA
4 Howard Hughes Medical Institute, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, USA
The author performed all the experiments, with some assistance from V.G., M.J.O., S.Y.W.,
S.D. and S.R.N. Bioinformatics analysis was performed by A.J. and C.W. Pathology analysis
was performed by R.B. All experiments were performed in the laboratory of Tyler Jacks.
124
ABSTRACT
The majority of lung cancer mortality is attributed to tumor metastasis. However, little
is known about the molecular mechanisms that drive tumor progression from welldifferentiated, localized lesions to aggressive metastatic cancer. We have previously shown
that downregulation of the pulmonary developmental regulator, Nkx2-1, promotes
metastasis of lung adenocarcinoma. Here, we present evidence that two additional
transcription factors, Foxa2 and Cdx2, synergize with Nkx2-1 to inhibit metastasis of lung
adenocarcinoma. Using transplantation models, we demonstrate that knockdown of Nkx2-1,
Foxa2 and Cdx2 in non-metastatic lung adenocarcinoma cells enhances tumor metastasis
to a level equivalent to metastatic cells. Moreover, intravital imaging and fine-needle
collection assays suggest that Foxa2 and Cdx2 depletion promotes cellular migration, while
Nkx2-1 loss promotes tumor colonization. Furthermore, analysis of tumors from a
genetically-engineered mouse model of lung adenocarcinoma and from human patients
shows that high expression levels of Nkx2-1, Foxa2, and Cdx2 correlate with more
advanced tumor stage and worse survival outcomes. Taken together, our study highlights
the role of these developmental regulators in inhibiting metastasis of lung adenocarcinoma.
125
INTRODUCTION
During tumor progression, cancer cells undergo global gene expression alterations
through which they acquire the traits that allow them to successfully advance through the
multiple steps of the metastatic cascade. These steps include the ability to invade and
migrate through surrounding tissues, intravasate into blood vessels, survive in circulation,
extravasate at secondary sites, and colonize distant organs (Steeg, 2006). A comprehensive
understanding of the upstream regulators that orchestrate this metastasis program is
lacking.
To better understand the molecular mechanisms of tumor progression and
metastasis in a well-defined genetic context, our laboratory has developed a geneticallyengineered mouse model of lung adenocarcinoma, a major subtype of lung cancer that is a
leading cause of cancer death worldwide. Conditional activation of oncogenic Kras and
inactivation of p53 in KrasLSL-G12D/+; p53fl/fl (KP) mice by viral delivery of Cre recombinase to
lung epithelial cells initiates the development of lung adenocarcinomas that closely resemble
the pathophysiological features of the human disease, including the capability to
metastasize to distant organs (Jackson et al., 2005; 2001). Previously, we found that
progression to metastasis in this model was closely associated with decreased expression
of the lung lineage transcription factor Nkx2-1, and knockdown of Nkx2-1 in non-metastatic
tumor cells was sufficient to increase their tumor-seeding ability in transplantation
experiments (Winslow et al., 2011). Nonetheless, two major lines of evidence indicate that
loss of Nkx2-1 alone may not be sufficient for full progression to metastasis. First,
knockdown of Nkx2-1 in non-metastatic lung adenocarcinoma cells does not recapitulate all
of the gene expression changes that occur during the transition from a non-metastatic to
metastatic state (Winslow et al., 2011). Moreover, Nkx2-1 deletion in KP lung
126
adenocarcinomas was not sufficient to induce the metastasis program, but instead
unmasked a latent gastric differentiation state of the tumor cells (Snyder et al., 2013). These
observations indicate that, in addition to Nkx2-1, there likely exist additional regulatory
factors that govern the program necessary for full acquisition of metastatic potential.
To investigate additional regulators of metastasis, we elected to examine the
transcription factors that control the expression of the metastasis mediator Tks5long. A critical
component of the proteolytic cellular protrusions, invadopodia, Tks5long promotes metastasis
in a wide variety of cancer types, including lung adenocarcinoma, and its expression is
consistently upregulated in metastatic cells compared to non-metastatic cells in the KP
model (Li et al., 2013; Murphy and Courtneidge, 2011). We have previously shown that
Tks5long is critical for promoting invadopodia formation and metastatic progression in
transplant and autochthonous mouse models. Moreover, Tks5long expression correlates with
more advanced disease stage and poor survival of lung adenocarcinoma patients.
Importantly, Tks5long is distinct from an invadopodia-inhibiting isoform, Tks5short, by the
presence of the membrane-binding Phox-homology domain and by the use of an
independent promoter for transcription (Li et al., 2013).
Here, we explored the transcriptional regulation of Tks5long to uncover key regulators
of metastasis in lung adenocarcinoma. We identified three transcriptional repressors of
Tks5long: Nkx2-1, Foxa2, and Cdx2, and subsequently showed that they function collectively
as important regulators of a metastasis program in lung adenocarcinoma. While Nkx2-1 and
Foxa2 are known for lineage specification and maintenance of the lungs (among other
organs), Cdx2 expression is limited to the intestines in normal adult tissues, and its role in
lung adenocarcinoma has not been previously explored. Here, we provide evidence that
these three transcription factors function cooperatively as critical regulators in suppressing
lung adenocarcinoma metastasis.
127
RESULTS
Nkx2-1, Foxa2, and Cdx2 synergistically suppress the expression of Tks5long in nonmetastatic lung adenocarcinoma cells
To identify novel mediators of the metastatic program in lung adenocarcinoma, we
focused on the transcriptional regulation of Tks5long. Tks5long has a mechanistically
characterized function in promoting metastasis across a wide variety of cancer types, as it
mediates the formation of invadopodia, which are proteolytic membrane protrusions that
facilitate cellular invasion (Paz et al., 2014). In lung adenocarcinoma, Tks5long is critical for
promoting metastasis, and increased Tks5long expression correlates with poor patient
outcomes (Li et al., 2013). Furthermore, Tks5long is dramatically and consistently upregulated
in our collection of metastatic lung adenocarcinoma cells compared to the non-metastatic
cells derived from the KP model (Li et al., 2013), suggesting that its expression is under tight
regulation. Importantly, H3K4me3 chromatin immunoprecipitation (ChIP)-sequencing
analysis showed that Tks5long is transcribed from its own promoter independent of the other
Tks5 isoform, Tks5short (data not shown), suggesting that its increased expression is likely a
result of transcriptional regulation and not alternative splicing. Given these data, we
hypothesized that the transcriptional regulatory mechanism for Tks5long functions as a key
switch in regulating a broader metastasis program, which includes many more metastasisrelated genes.
To identify potential transcriptional regulators of Tks5long, we utilized an existing gene
expression profile of a panel of cell lines derived from non-metastatic and metastatic primary
lung adenocarcinomas as well as metastases in the KP model (termed TnonMet, TMet, and Met
cells, respectively) (Winslow et al., 2011). From this dataset we generated a list of
transcription factors that were (i) differentially expressed between the collection of TnonMet
128
and TMet/Met cells, and (ii) had predicted binding sites in the Tks5long locus based on
genomic sequence analysis.
This approach neglects potential non-transcriptional regulatory mechanisms of
Tks5long expression, but we wanted to start by first focusing strictly on transcription. Among
the 57 transcription factors that are differentially expressed between TnonMet and TMet/Met
cells, we identified three that meet these criteria: Nkx2-1, Foxa2, and Cdx2. Microarray
gene-expression profiling and qRT-PCR validation confirmed that Nkx2-1, Foxa2, and Cdx2
are highly expressed in TnonMet cells, but are partially or completely lost in TMet/Met cells
(Figures 1A, S1A, and S1B). The expression pattern of each transcription factor is inversely
proportional to Tks5long, suggesting that these three factors may suppress Tks5long
expression (Figure 1A).
In order to determine whether Nkx2-1, Foxa2, and Cdx2 inhibit Tks5long expression,
we knocked down the three transcription factors in TnonMet cells using shRNAs (hereafter
referred to as TnonMet-shNFC cells) and examined the effect on Tks5long expression compared
to control knockdown using shRNAs against firefly and renilla luciferases. Interestingly,
while single or double knockdown of these transcription factors in TnonMet cells only
moderately affected Tks5long expression, triple knockdown of all three transcription factors in
TnonMet-shNFC cells led to a dramatic increase in Tks5long mRNA and protein levels, such that
the levels were comparable to that of TMet cells with the highest Tks5long expression (Figures
1B and 1C). The effect of the triple knockdown on Tks5long expression was synergistic, as
the expression levels of Tks5long far exceeded that predicted by the additive effects of single
knockdown. Importantly, the effect of Nkx2-1, Foxa2, and Cdx2 knockdown was specific to
Tks5long, and did not affect expression of the other Tks5 isoform, Tks5short (Figures 1B and
1C). We validated these results in an independent TnonMet cell line (Figure S1C). These data
129
suggest that Nkx2-1, Foxa2, and Cdx2 may suppress Tks5long expression in a synergistic
manner.
To examine whether these transcription factors are sufficient to suppress Tks5long
expression in metastatic cells, we overexpressed each transcription factor in a doxycyclineinducible manner in TMet cells. Increasing the levels of each transcription factor inhibited
Tks5long expression in a dosage-dependent manner, without affecting Tks5short mRNA levels
(Figures 1D, 1E, and 1F). Furthermore, this suppressive effect on Tks5long expression was
additive, as combined overexpression of Nkx2-1 and Foxa2 (or of Nkx2-1 and Cdx2)
reduced the mRNA levels of Tks5long more significantly than one transcription factor alone
(Figure S1D).
To determine whether these transcription factors suppress Tks5long expression by
directly binding to its genetic locus, we performed chromatin immunoprecipitation (ChIP)qPCR analysis on potential enhancer regions in the Tks5long locus. We observed binding of
Nkx2-1 and Foxa2 to multiple enhancers in the Tks5long locus (Figure 1G), consistent with a
direct role in downregulating gene expression at the genetic locus. In contrast, we could not
detect enrichment for Cdx2 at these Nkx2-1/Foxa2 binding sites or at the predicted Cdx2
binding sites, suggesting that Cdx2 may either bind to the Tks5long locus at a different
location, or suppress Tks5long expression indirectly through other transcription factors.
130
Figure 1. Nkx2-1, Foxa2, and Cdx2 synergistically suppress the expression of Tks5long
in non-metastatic lung adenocarcinoma cells.
(A) mRNA levels of Nkx2-1, Foxa2, and Cdx2 in TnonMet (368T1, 394T4, 802T4) and TMet
(373T1, 393T3, 393T5, 482T1) lung adenocarcinoma cell lines anti-correlate with Tks5long
expression as measured by qRT-PCR. (r and p, Spearman correlation coefficient and pvalue).
(B-C) Knockdown of Nkx2-1, Foxa2, and Cdx2 in TnonMet cells (394T4) derepresses Tks5long
expression, but not Tks5short, as measured by qRT-PCR (B) and immunoblotting (C). Lines
(-) indicate control hairpins against firefly or renilla luciferase. Data are represented as mean
± SD. The p-value was calculated by Student’s t test.
131
Figure 1 (continued). Nkx2-1, Foxa2, and Cdx2 synergistically suppress the
expression of Tks5long in non-metastatic lung adenocarcinoma cells.
(D-F) Overexpression of Nkx2-1, Foxa2, and Cdx2 in TMet cells (393T5) represses Tks5long
expression, but not Tks5short, as measured by qRT-PCR. Data are represented as mean ±
SD. The p-values were calculated by Student’s t test.
(G) ChIP-qPCR analysis of the enrichment of Nkx2-1 (left) and Foxa2 (right) binding at the
Tks5long genomic locus. Data are represented as mean ± SEM of three independent
experiments. SftpA and Hnf4a serve as positive controls for Nkx2-1 and Foxa2 binding,
respectively. GD8: negative control mapping to a gene desert region on murine
chromosome 8. For each enhancer versus GD8, p < 0.03 by Student’s t test.
132
Nkx2-1, Foxa2, and Cdx2 function synergistically to inhibit metastasis in vivo
To test whether Nkx2-1, Foxa2, and Cdx2 suppress metastasis, we transplanted
TnonMet-shNFC cells subcutaneously into nude mice, and examined their ability to
metastasize from the subcutaneous tumor to the lungs over a period of 6 weeks. This assay
tests a full range of metastatic properties, as it requires that tumor cells invade and
intravasate into circulation at the primary site, and then to extravasate and colonize a distant
organ at the metastatic site. Notably, while TnonMet-shNkx2-1 cells formed more lung nodules
than control TnonMet cells, TnonMet-shNFC cells were strikingly more metastatic (Figures 2A
and 2B). The increase in metastatic potential in TnonMet-shNFC cells was more than the
additive effect induced by single or double knockdown. Importantly, the knockdown of these
transcription factors did not significantly affect the size of the primary tumors in the
subcutaneous site (Figure S2). These data suggest that loss of Nkx2-1, Foxa2, and Cdx2
function cooperatively to promote metastasis in lung adenocarcinoma.
133
Figure 2. Foxa2 and Cdx2 synergize with Nkx2-1 in inhibiting metastasis in vivo.
(A) Triple knockdown of Nkx2-1, Foxa2, and Cdx2 in a subcutaneous transplantation assay
increases the metastatic potential of TnonMet cells (394T4) in comparison to control TnonMet
cells and TnonMet cells with single/double knockdown, to a level similar to TMet cells (373T1).
Representative images of lung metastases are shown. Lines (-) indicate control hairpins
against firefly or renilla luciferase.
(B) Quantification of metastasis frequencies. Each circle represents an individual mouse.
Data are represented as mean ± SEM. The p-values were calculated by Student’s t test. *p
< 0.05, **p < 0.01. ****p < 0.0001.
134
The increased number of lung metastases seen with TnonMet-shNFC cells could be
explained by changes in their ability to complete different steps along the metastatic
cascade. We first examined how knockdown of Foxa2, Cdx2 and Nkx2-1 affected cell
morphology and motility in vivo. We observed that TnonMet-shNFC subcutaneous tumors
adopted a mesenchymal morphology similar to that of TMet tumors, in contrast to the
predominantly epithelial morphology in tumors formed by TnonMet and TnonMet-shNkx2-1 cells
(Figures 3A and 3B). This change to mesenchymal morphology for TnonMet-shNFC cells only
occurred in vivo but not in vitro (Figure S3A), strongly suggesting that it is induced by noncell autonomous factors present in the tumor microenvironment. Consistent with this change
to a mesenchymal morphology, qRT-PCR analysis of TnonMet-shNFC tumors showed loss of
the epithelial marker Krt19 and a small increase in expression of mesenchymal markers
Twist, Snail, Zeb1, and N-cadherin compared to TnonMet tumors (Figure S3B). These data
suggest a partial epithelial-to-mesenchymal transition (EMT), an important step in the
metastatic process (Sato et al., 2012), of TnonMet-shNFC cells in vivo.
Because mesenchymal morphology is associated with increased motility (Thiery et
al., 2009; Tsai and Yang, 2013), we performed intravital imaging to monitor migration of
GFP-positive cancer cells within subcutaneous tumors. TnonMet-shNFC tumors contained
significantly more migratory cells than TnonMet and TnonMet-shNkx2-1 cells (Figure 3C).
Furthermore, when we measured the chemotactic ability of the tumor cells by performing in
vivo fine-needle collection assay using 10% fetal bovine serum (FBS) as a chemo-attractant
(Wyckoff et al., 2000), we collected a higher number of GFP-positive cancer cells from
TnonMet-shNFC subcutaneous tumors than TnonMet and TnonMet-shNkx2-1 tumors (Figure 3D).
Taken together, these data suggest that TnonMet-shNFC cells are more motile in vivo than
TnonMet and TnonMet-shNkx2-1 cells.
135
To test whether the enhanced metastatic ability of TnonMet-shNFC cells can also be
explained by differences in the colonization of secondary sites during the metastatic
cascade, we injected tumor cells intravenously into immunocompromised mice to test their
ability to establish tumor nodules upon arriving at the lung capillaries. We observed that
TnonMet-shNFC cells and TnonMet-shNkx2-1 cells had equally high colonization capacity
compared to TnonMet control (Figures 3E and 3F), suggesting that while knockdown of Nkx2-1
enhances metastatic colonization, additional inhibition of Foxa2 and Cdx2 does not further
contribute to this effect.
Taken together, our data support a model in which Nkx2-1, Foxa2, and Cdx2 inhibit
metastasis by acting on different steps of the metastatic cascade. While Nkx2-1 inhibits the
extravasation/colonization step towards the end of the metastatic cascade, Nkx2-1, Foxa2,
and Cdx2 together inhibit migration and invasion in the early steps of metastasis (Figure
3G). The gain of metastatic ability in TnonMet-shNFC cells compared to TnonMet-shNkx2-1 cells
is correlated with an increase in invasion and migration in the primary tumor.
136
Figure 3. Foxa2 and Cdx2 synergize with Nkx2-1 in inhibiting migration in vivo.
(A-B) Triple knockdown of Nkx2-1, Foxa2, and Cdx2 in TnonMet (394T4) subcutaneous tumors
induces a mesenchymal morphology similar to TMet (373T1) tumors, in contrast to the
epithelial TnonMet-shCtrl and TnonMet-shN tumors. (A) Representative H&E staining of
subcutaneous tumors. Scale bar represents 50 µm. (B) Quantification of epithelial and
mesenchymal morphology by a pathologist (R.T.B.). Lines (-) indicate control hairpins
against firefly or renilla luciferase.
(C-D) Knockdown of Nkx2-1, Foxa2, and Cdx2 in TnonMet (394T4) subcutaneous tumors
enhances migration in vivo compared to TnonMet-shCtrl and TnonMet-shN tumors, as measured
by intravital imaging (C), and fine needle collection assay (D). Data are represented as
mean ± SEM. The p-values were calculated by Student’s t test. **p < 0.01, ***p < 0.001,
****p < 0.0001.
(E-F) TnonMet-shNFC and TnonMet-shN (394T4) cells show similar colonization ability in the
lungs after intravenous transplantation. (E) Representative images of lungs with tumor
nodules. (F) Quantification of lung tumor burden. Each circle represents an individual
mouse. Data are represented as mean ± SEM. The p-values were calculated by Student’s t
test. *p < 0.05, ***p < 0.001.
(G) Model for distinct roles of Nkx2-1, Foxa2, and Cdx2 in inhibiting metastatic progression.
137
Nkx2-1, Foxa2, and Cdx2 cooperatively repress a program of metastasis genes
Given the increased metastatic ability of TnonMet-shNFC cells compared to TnonMet and
TnonMet-shNkx2-1 cells, we next investigated whether a network of metastasis-related genes
might be differentially regulated upon knockdown of Nkx2-1, Foxa2, and Cdx2. Transcription
factors generally regulate a broad network of target genes, often genes with similar
functions. Therefore, we hypothesized that in addition to suppressing the expression of
Tks5long – a critical mediator of metastasis – Nkx2-1, Foxa2, and Cdx2 may also regulate the
expression of other metastasis-related genes. To this end, we performed RNA sequencing
(RNA-seq) on TnonMet, TnonMet-shNkx2-1, TnonMet-shNFC, and TMet cells, and employed an
unsupervised blind source separation strategy using Independent Component Analysis
(ICA) (see Supplemental Experimental Procedures and Bhutkar et al.) to elucidate
statistically independent gene expression signatures that characterize the transcriptomes of
these cells. This high-resolution approach allowed us to identify two statistically significant
and biologically relevant signatures that are separate from a “clonal signature” that
embodies the background identity of TMet versus TnonMet-derived cells (Figure 4A): (1) a
signature differentiating Nkx2-1-high samples (TnonMet) from all Nkx2-1-low samples (TnonMetshN, TnonMet-shNFC, and TMet cells), which we termed an “shN signature”; and (2) a signature
clustering TnonMet-shNFC/TMet cells away from TnonMet/TnonMet-shN cells, which we designated
as the “shNFC signature”.
Our analysis provides several lines of evidence that TnonMet-shNFC cells significantly
recapitulated the major metastasis-related gene expression patterns associated with TMet
cells. First, hierarchical clustering based on the top and bottom 2nd percentile of genes
identified in the shNFC signature showed clustering of TnonMet-shNFC cells with TMet cells and
this cluster segregated away from TnonMet and TnonMet-shN cells (Figures 4B and 4C),
indicating that the identified gene expression pattern robustly supports the signature.
138
Second, clustering based on global gene expression profiles after depletion of the
background clonal signature showed that TnonMet-shNFC cells were substantially more
closely related to TMet cells than TnonMet-shN and TnonMet cells (Figure S4A). Finally, the
shNFC signature is highly enriched for the “Tmet/Met” gene set and depleted for the
“TnonMet” gene set (Winslow et al., 2011) when analyzed using Gene Set Enrichment
Analysis (GSEA; Subramanian et al., 2005) (Figure 4D). In contrast, the shN signature does
not show a similar enrichment or depletion pattern of TMet/Met or TnonMet gene sets (Figure
4D).
Of the 388 genes identified by the shNFC signature, 169 genes were upregulated
(among which is Tks5long) and 219 genes were downregulated in the TnonMet-shNFC/TMet cells
compared to the TnonMet/TnonMet-shN cells (Figure 4C). Interestingly, GSEA analyses using
publicly available gene sets in the Molecular Signatures Database (Subramanian et al.,
2005) revealed that the shNFC signature is significantly enriched for gene sets that
represent poor patient prognosis, metastasis/EMT, and TGFβ targets (Figure 4E).
Furthermore, the shNFC signature is significantly depleted for gene sets related to
gastrointestinal/liver-related genes, reflecting the gene expression changes induced upon
knockdown of Foxa2 and Cdx2 (Figure 4E). Many of the gene expression changes identified
in the shNFC signature that were relevant for metastasis or gastrointestinal differentiation
were validated by qRT-PCR and immunoblotting (Figures 4F, 4G, and S4B, and data not
shown).
To directly answer the question of what fraction of the gene expression differences
between our collection of TnonMet and TMet/Met cells was recapitulated by combined
knockdown of Nkx2-1, Foxa2 and Cdx2 in TnonMet 394T4 cells, we performed targeted
pairwise differential analysis. We found that a large fraction (32%) of the genes that were
differentially expressed between TnonMet vs TMet/Met cells also showed significant gene
139
expression alterations by two-fold or more in comparing 394T4-TnonMet vs TnonMet-shNFC cells
(p = 2.22 x 10^-16, hypergeometric test), suggesting that reduced expression of the three
transcription factors can explain about one-third of the gene expression changes in
metastatic progression. In contrast, only 9% of the genes that were differentially expressed
between TnonMet vs TMet/Met cells were found to be altered in comparing 394T4-TnonMet vs
TnonMet-shN cells (p = 2.22 x 10^-16, hypergeometric test).
Taken together, these data argue that the TnonMet-shNFC gene expression program
driven by loss of Nkx2-1, Foxa2, and Cdx2 significantly recapitulated the characteristics of
TMet cells. These findings are consistent with data from our in vivo metastasis assays that
TnonMet-shNFC cells are more metastatic than TnonMet-shN cells, and further support that
Foxa2 and Cdx2 cooperate with Nkx2-1 to regulate a network of metastasis-related genes.
140
Figure 4. Nkx2-1, Foxa2, and Cdx2 together repress a program of metastasis genes.
(A) RNA-seq gene expression analysis of 394T4 TnonMet, TnonMet-shN, TnonMet-shNFC, and
373T1 TMet cells reveals two statistically significant and biologically relevant signatures: an
shN signature, and an shNFC signature, both separate from a clonal signature that
embodies the background identity of TMet versus TnonMet-derived cells.
(B) Dendrogram showing sample relationships via clustering based on top and bottom 2nd
percentile of genes in the shNFC signature.
(C) Differentially expressed genes that drive the shNFC signature distinguishing TnonMetshNFC/TMet cells from TnonMet/TnonMet-shN cells.
(D) GSEA enrichment analysis reveals that the shNFC Signature is highly enriched for
TMet/Met genes and depleted for TnonMet genes, whereas the shN Signature does not show
similar enrichment. (ns, not significant.)
(E) GSEA analysis using the MSigDB curated gene set collection shows that the shNFC
signature is enriched for gene sets associated with metastasis and poor prognosis, and
depleted for gene sets related to Hnf4a-related gastrointestinal/hepatic differentiation.
(F) Examples of pro-metastatic (Tks5long and TGFβ) and anti-metastatic (Mtus1) genes that
were identified in the shNFC signature by RNA-seq. Data are represented as mean ± SEM.
The p-values were calculated by Student’s t test.
(G) Examples of gastrointestinal differentiation genes (Hnf4g, Lgals4, and Lgals6) that were
identified in the shNFC signature by RNA-seq. Data are represented as mean ± SEM. The
p-values were calculated by Student’s t test.
141
The inhibitory effect of Nkx2-1, Foxa2, and Cdx2 on metastasis depends on the
activation of Tks5long, Hmga2, and Snail expression
Given the large number of genes activated in the shNFC signature, we then
examined whether some of these genes are functionally required for promoting the
metastatic capacity of TnonMet-shNFC cells. In addition to Tks5long, we elected to examine the
embryonal proto-oncogene Hmga2 and the epithelial-to-mesenchymal transition (EMT)
transcription factor Snail, as these metastasis-promoting genes are also significantly
upregulated in TMet/Met cells compared to TnonMet cells in microarray-based gene expression
analysis (Figure S5A; Winslow et al., 2011). TnonMet-shNFC cells exhibited a significant
increase in Hmga2 and Snail expression at the levels of both mRNA and protein, exceeding
the levels in TnonMet and TnonMet-single/double knockdown cells (Figures 5A and 5B). The
activation effect was synergistic, as the expression levels of Hmga2 and Snail was much
higher than predicted by the additive effects of single knockdown. To test whether increased
expression of Tks5long, Hmga2, and Snail are required for increased metastatic ability of
TnonMet-shNFC cells, we knocked down these three genes by shRNAs or CRISPR/CassgRNAs (Figure 5C). Decreased expression of Tks5long, Hmga2, or Snail individually
impaired the metastatic ability of TnonMet-shNFC cells, without affecting the size of the
primary subcutaneous tumors (Figures 5D, 5E, and S5D). Importantly, Hmga2 and Snail
knockdown did not affect the expression of Tks5long, or of each other (Figure S5B and S5C),
suggesting that they each contributed individually to increasing metastatic potential. These
data support our hypothesis that Nkx2-1, Foxa2, and Cdx2 may function as key regulators
for a network of metastasis-related genes that include Tks5long, Hmga2, and Snail.
142
Figure 5. The inhibitory effect of Nkx2-1, Foxa2, and Cdx2 on metastasis depends on
the activation of Tks5long, Hmga2, and Snail expression
(A-B) Combined knockdown of Nkx2-1, Foxa2, and Cdx2 in TnonMet cells (394T4)
derepresses the expression of Hmga2 and Snail as analyzed by qRT-PCR (A) and
immunoblotting (B). Lines (-) indicate control hairpins against firefly or renilla luciferase. Data
are represented as mean ± SD. The p-values were calculated by Student’s t test.
(C-E) Knockdown of Tks5long, Hgma2, or Snail dampens the metastatic ability of 394T4
TnonMet-shNFC cells after subcutaneous transplantation. (C) Validation of knockdown by
immunoblotting. (D) Representative images of lungs with tumor nodules. (E) Quantification
of lung tumor nodules. Each circle represents an individual mouse. Control includes TnonMetshNFC cells (n=10 mice) and TnonMet-shNFC-sgRosa cells (n=4 mice); Tks5long knockdown
was generated by hairpins shTks5long#1 (n=5 mice) and shTks5long#2 (n=4 mice); Hmga2
knockdown by hairpin shHmga2#1 (n=5 mice) and sgRNA sgHmga2#2 (n=4 mice); Snail
knockdown by sgRNAs sgSnail#1 (n=5 mice) and sgSnail#2 (n=4 mice). Data are
represented as mean ± SEM. The p-values were calculated by Student’s t test. *p < 0.05,
**p < 0.01.
143
Endogenous expression pattern of Nkx2-1, Foxa2, and Cdx2 correlates with tumor
progression in vivo
To further characterize the suppressive roles of Nkx2-1, Foxa2, and Cdx2 during
tumor progression, we examined their endogenous expression in the KP model of lung
adenocarcinoma. This animal model provides a well-defined genetic context and a
stereotypic temporal pattern of histologic progression, which facilitate the identification of
patterns of gene expression alterations that accompany tumor progression. We analyzed
195 tumor regions from mice ranging from 17 to 33 weeks post-initiation, and scored them
as low grade (grade 1-3) or high grade (grade 4, poorly differentiated) based on nucleocytoplasmic morphology, tumor architecture, and the presence of stromal invasion.
Consistent with previous findings (Winslow et al., 2011), expression of Nkx2-1 and Hmga2
anti-correlated with each other in these lung tumors (Figure 6A), and tumors of Nkx2-1low
and Hmga2high expression were frequently associated with high tumor grade (Figures 6D
and 6E).
The expression pattern of Foxa2 was highly similar to Nkx2-1. Foxa2 expression
strongly correlated with Nkx2-1 and was anti-correlated with Hmga2 (Figures 6B and 6C).
The inverse correlation of Foxa2 expression to tumor grade was more striking than Nkx2-1:
Foxa2high expression was invariably associated with low-grade tumors (117/117; 100%),
while Foxa2low tumors were consistently high grade (21/21; 100%) (Figure 6F). These data
suggest that Foxa2, similar to Nkx2-1, marks an early state of tumor progression, and loss
of Foxa2 expression is a stringent diagnostic marker of high-grade tumors. Interestingly,
loss of Foxa2 expression in high-grade tumors often lagged behind loss of Nkx2-1 (see
examples in Figures 6H third column and S6A). Furthermore, while high-grade tumors were
more often Nkx2-1low (46/67; ~70%) than Nkx2-1mixed (21/69; ~30%), a large fraction of these
high-grade tumors retained mixed Foxa2 expression (46/67; ~70%) instead of being
144
completely Foxa2low (21/67; ~30%). These observations indicate that loss of Foxa2
expression in tumor progression does not occur concurrently with loss of Nkx2-1, but
happens after Nkx2-1 expression is lost (Figure 6I).
Cdx2 staining was detected in a noteworthy fraction of these lung adenocarcinomas
(35/195; ~20%) albeit at a lower frequency than the staining of Nkx2-1 (139/195; ~70%) and
Foxa2 (174/195; ~90%). This lower but significant frequency of Cdx2 staining suggests that
Cdx2 expression may represent a transient state during tumor progression and is only
detectible when a large number of tumors are analyzed. From the Cdx2high tumors, a clear
and consistent pattern emerged, arguing that, unlike Nkx2-1, expression of Cdx2 marks an
intermediate state of tumor progression that is temporally situated after loss of Nkx2-1 but
before loss of Foxa2, and is accompanied by partial activation of Hmga2 expression (Figure
6G). First, the vast majority of Cdx2high tumor areas were Nkx2-1low (8/9, 89%), indicating a
strong anti-correlation between Cdx2 and Nkx2-1 expression. Second, these Cdx2high
regions were frequently Foxa2high (7/9, 87%), suggesting that Cdx2 expression correlates
with Foxa2. In fact, all of the Nkx2-1low Foxa2high tumor regions were invariably Cdx2high (6/6,
100%). We did not see examples of Nkx2-1low, Foxa2low, Cdx2high regions. Finally, the
majority of Cdx2high tumor regions were Hmga2mixed (5/9, 56%), whereas only a small fraction
was Hmga2high (2/9; 22%) or Hmga2low (2/9; 22%). Importantly, even though some of these
Cdx2high regions were found adjacent to Hmga2high, high-grade, and poorly-differentiated
areas, Cdx2high regions themselves were invariably well/moderately differentiated and never
part of the poorly differentiated regions (Figures 6H third column and S6A). Collectively,
these data strongly suggest that Cdx2 expression represents an intermediate state between
loss of Nkx2-1 and full activation of Hmga2 in tumor progression (Figure 6I).
An alternative model that could explain these observations is that Cdx2-expressing
tumors represent a dead-end differentiation state that will never progress to advanced
145
metastatic tumors. However, multiple lines of evidence argue against this model and instead
favor the hypothesis that Cdx2 expression is a transition state. First, within a single tumor,
well/moderately-differentiated tumor regions with strong Cdx2 expression and weak Hmga2
expression were frequently associated with adjacent poorly-differentiated regions that
exhibit a reciprocal expression pattern of low Cdx2 staining and intense Hmga2 staining
(Figures 6H third column and S6A), consistent with the model that tumors lose Cdx2
expression and activate Hmga2 expression during progression to an advanced metastatic
state. Second, cell line-based experiments showed that knockdown of Nkx2-1 in TnonMet cells
derepressed Cdx2 mRNA and protein levels, while knockdown of Foxa2 in TnonMet-shNkx2-1
cells reduced Cdx2 levels (Figures S6B and S6C), suggesting that Cdx2 expression in these
tumor cells is plastic and can be regulated by changes in expression of Nkx2-1 and Foxa2.
Furthermore, ChIP-qPCR analysis detected binding of Nkx2-1 and Foxa2 to an enhancer
downstream of the genomic locus of Cdx2 (Figure S6D). Based on these findings, we
propose a model for the regulation of Cdx2 expression in lung adenocarcinoma, in which the
transcription of Cdx2 is inhibited by binding of Nkx2-1 to a nearby enhancer region. Upon
loss of Nkx2-1, expression of Cdx2 is derepressed in a manner dependent on Foxa2 binding
to the same enhancer (Figure S6E).
Taken together, these observations suggest a model for lung adenocarcinoma
progression (Figure 6I) in which early tumors express Nkx2-1 and Foxa2. As tumors
progress, Nkx2-1 is silenced, leading to activation of Cdx2 and partial activation of
metastasis-promoting genes such as Hmga2 in at least a subset of tumors. Finally,
suppression of Foxa2 leads to reduced Cdx2 expression, and the combined loss of Nkx2-1,
Foxa2 and Cdx2 is required for complete derepression of Hmga2 and other metastasis
promoting genes, resulting in full acquisition of metastatic potential.
146
Figure 6. Endogenous expression pattern of Nkx2-1, Foxa2, Cdx2, and Hmga2
correlates with tumor progression in an autochthonous model of lung
adenocarcinoma.
(A-C) Pairwise correlation of Nkx2-1, Foxa2, and Hmga2 expression in KrasG12D; p53-/- (KP)
lung adenocarcinomas.
(D-F) Correlation of Nkx2-1, Foxa2, and Hmga2 expression in KP lung adenocarcinomas
with tumor grades. Tumor regions were scored as low grade (grade 1-3) or high grade
(grade 4, poorly differentiated) based on nucleo-cytoplasmic morphology, tumor
architecture, and the presence of stromal invasion.
(G) High expression of Cdx2 in KP lung adenocarcinomas is frequently associated with low
expression of Nkx2-1, high expression of Foxa2, low-medium expression of Hmga2, and
low-grade histology.
(H) Representative H&E and IHC images of KP lung adenocarcinomas. Scale bar
represents 150 µm. Inserts show nucleo-cytoplasmic morphology of the tumor cells.
(I) Model summarizing expression changes of Nkx2-1, Foxa2, Cdx2, and Hmga2 in lung
adenocarcinomas.
147
Nkx2-1, Foxa2, and Cdx2 gene expression signatures predict clinical outcomes of
lung adenocarcinoma patients
We next asked if our observations could provide prognostic information relevant to
human lung adenocarcinomas. To this end, we analyzed RNA-seq expression data for 488
lung adenocarcinoma primary tumors from patients with stage I-IV disease from The Cancer
Genome Atlas (TCGA, http://cancergenome.nih.gov/). Unsupervised signature analysis (see
Supplemental Experimental Procedures and Bhutkar et al.) of the expression patterns of
NKX2-1, FOXA2, CDX2, and HMGA2 in these tumors revealed three gene expression
signatures (Figures 7A and 7B). Interestingly, these signatures closely correlated with the
expression patterns of Nkx2-1, Foxa2, Cdx2, and Hmga2 in the KP mouse model of lung
adenocarcinoma. The first signature is driven by high expression of NKX2-1 and FOXA2, as
well as low expression of HMGA2, similar to the early tumors in mice. The second signature
is characterized by high expression of CDX2, similar to tumors in the intermediate state.
Finally the third signature is characterized by high expression of HMGA2, similar to
advanced tumors in our mouse model.
Importantly, these signatures also strongly correlated with clinical outcomes (Figures
7C and 7D). The NKX2-1/FOXA2-driven signature and the CDX2-driven signature were
associated with favorable and intermediate disease stage and patient survival, respectively.
In contrast, the HMGA2-driven signature was associated with the worst stage and survival.
Furthermore, the prognostic values of these four gene pattern-derived signatures were more
powerful than analysis using a single gene alone (data not shown). Finally, a multivariate
Cox proportional hazard regression analysis showed that these signatures of gene
expression pattern were significantly prognostic factors independent of gender, age, and
disease stage (Figure 7E). These results provide further support to our proposed model that
early-stage tumors are marked by NKX2-1 and FOXA2 expression, while intermediate
148
tumors acquire expression of CDX2, and most advanced tumors acquired expression of
HMGA2. Taken together, these data argue that our proposed model for lung
adenocarcinoma progression is highly relevant for human cancer. Our findings provide
important information for the prognosis of lung adenocarcinoma patients, and may inform
future development of therapeutic strategies for this highly metastatic disease.
149
Figure 7. NKX2-1, FOXA2, and CDX2 gene expression signatures predict clinical
outcomes of lung adenocarcinoma patients.
(A) Heatmap showing the expression patterns of NKX2-1, FOXA2, CDX2, and HMGA2 in
signatures identified in TCGA human primary lung adenocarcinoma (n = 488) using ICA.
(B) Box-and-whisker plots of standardized expression levels of NKX2-1, FOXA2, CDX2, and
HMGA2 in each signature. Horizontal dashed line reflects mean expression level across all
tumors. The p-values were calculated by Student’s t test. ****p < 0.0001.
(C-E) Analysis of the top 10th percentile of the 488 lung adenocarcinoma patients in each
signature identified by ICA shows that the three signatures correlate with disease stage (C)
and overall survival (p = 0.0006 by Log-rank test) (D). Multivariate Cox proportional hazard
regression analysis of overall survival after adjustment for gender, age, and stage (E).
150
CONCLUSION
Our study shows that Nkx2-1, Foxa2, and Cdx2 function collectively to suppress
metastatic progression of lung adenocarcinoma. Evidence from the autochthonous KP
model of lung adenocarcinoma, derivative cell lines, and human patients supports a model
of tumor progression in which loss of Nkx2-1 is accompanied by derepression of Cdx2 in the
transition from early to intermediate tumor state. Silencing of Foxa2 and Cdx2 in addition to
Nkx2-1 allows progression to advanced tumors with metastatic potential (Figure 6I). This
tumor suppressive effect is at least in part explained mechanistically by the observation that
Nkx2-1, Foxa2 and Cdx2 collectively inhibit cell migration in vivo, while Nkx2-1 also
independently inhibits colonization at distant sites. In human lung adenocarcinoma, the
expression profile of NKX2-1, FOXA2, CDX2, and HMGA2 predicts tumor differentiation
states that correlate with disease stage and survival outcome. Thus, our results are highly
relevant for the human disease, and provide prognostic information for lung
adenocarcinoma patients.
In terms of gene expression, combined loss of Nkx2-1, Foxa2, and Cdx2 is sufficient
to activate a network of transcriptional targets that account for a significant fraction of the
dysregulated genes in TMet cells, including Tks5long, Hmga2, and Snail. This finding strongly
suggests that these three factors function as key regulators in restraining the metastasis
program in lung adenocarcinoma. The fact that suppression of Nkx2-1, Foxa2, and Cdx2 is
sufficient to derepress this network of metastasis-related genes also suggests that the
transcriptional activators for at least a subset of these target genes may be readily available
to act upon loss of these three transcriptional suppressors. As such, Nkx2-1, Foxa2 and
Cdx2 may function as important regulatory nodes in governing transcriptional programs that
together restrain tumor metastasis.
151
The role of Nkx2-1, Foxa2, and Cdx2 in regulating metastasis can be considered
from a developmental perspective, as tumor progression is often associated with
dysregulation of normal development and differentiation. Nkx2-1 and Foxa2 are important
developmental regulators of the lungs as well as other endoderm-derived organs (Kimura et
al., 1996; Minoo et al., 1999; Wan et al., 2005; Zhou et al., 1997). Nkx2-1 is a
homeodomain-containing transcription factor essential for differentiation of the lungs during
early embryogenesis (Kimura et al., 1996). In the lungs, Nkx2-1 is expressed in all epithelial
cells in early pulmonary development, but becomes progressively restricted to alveolar type
II and Clara cells in adults (Minoo et al., 1999), where Nkx2-1 activates expression of
pulmonary-specific genes, including SftpA, B, C and the Clara cell secretory protein gene,
CCSP (Minoo et al., 1999). Foxa2 is a forkhead transcription factor that is expressed in the
endoderm and cooperates with its paralog Foxa1 in mediating organogenesis of the lungs,
as well as the stomach, intestine, liver, pancreas, midbrain, and prostate (Kaestner, 2010).
Foxa2 is important for alveolarization of the lungs during development (Wan et al., 2004).
Adult lungs express Foxa2 in the bronchiolar epithelium and alveolar type II cells (Besnard
et al., 2004). Interestingly, while Foxa2 expression and function overlap largely with Foxa1,
we did not see differential expression of Foxa1 between TnonMet and TMet/Met cells
(Supplemental Fig. S7). In contrast to Nkx2-1 and Foxa2, the expression of Cdx2 is not
appreciably expressed in normal embryonic or adult lungs. Instead, as a member of the
Caudal-type homeobox protein, Cdx2 is required for intestine morphogenesis during
embryonic development, and is expressed in the small and large intestines in adults (Beck
and Stringer, 2010). Cdx2 also functions in an earlier stage in development for trophoblast
formation and axial patterning (Beck and Stringer, 2010). While detection of Cdx2
expression in the KP lung adenocarcinoma is perhaps surprising, it is consistent with other
152
reports that CDX2 is aberrantly expressed in human lung adenocarcinomas (Grimminger et
al., 2009; Yatabe et al., 2004).
Our findings suggest that during progression of lung adenocarcinoma, partial loss of
the lung differentiation program by silencing of Nkx2-1 can lead to aberrant activation of an
alternative differentiation program of the intestine driven by Cdx2 in at least a subset of
these tumors. As such, the activation of the latent intestinal program may serve as a
redundant mechanism in the cells to restrain tumor dedifferentiation and metastatic
progression. Because expression of Cdx2 is dependent on Foxa2, further loss of Foxa2
expression in more advanced tumors leads to loss of the Cdx2-driven intestinal program,
leading to a cellular state of more primitive differentiation and higher metastatic potential,
reflected by the expression of the embryonal proto-oncogene Hmga2, EMT factor Snail, and
invadopodia mediator Tks5long. The aberrant activation of Cdx2 has also been identified to
induce intestinal differentiation in other tumor types, including gastric cancer, esophageal
cancer, nasal adenocarcinoma, pancreatic cancer, and ovarian cancer, and in some cases
has been shown to associate with favorable prognosis (Guo et al., 2004; Matsumoto et al.,
2004; Mizoshita et al., 2003; Yuasa, 2003). These studies, together with our findings,
strongly suggest that the activation of Cdx2-driven alternative differentiation program in
tumors may be a common phenomenon in the evolution of cancer development, and may
serve as a mechanism to restrain malignant progression.
Taken together, our study has shown that the developmental transcription factors
Foxa2 and Cdx2 function synergistically with Nkx2-1 as important regulators in inhibiting
metastasis of lung adenocarcinoma. These data provide strong evidence for the important
roles of active and latent developmental regulators in restraining the programs of tumor
dedifferentiation and metastatic progression. Our findings also shed light on the complexity
of the interplays between different differentiation programs in the course of tumor evolution.
153
MATERIALS AND METHODS
Autochthonous K-rasG12D/WT; p53-/- lung tumors and derivative cell lines
Lung tumors were initiated via intratracheal delivery of Lenti-Cre or Adeno-Cre in KrasLSL-G12D/WT; p53flox/flox mice as described previously (DuPage et al., 2009). The MIT
Institutional Animal Care and Use Committee approved all animal studies and procedures.
TnonMet, TMet, and Met cell lines were generated previously using Lenti-Cre-initiated
primary tumors and metastases harvested at 6-14 months post-infection (Winslow et al.
2011). All cell lines were cultured in complete media (DMEM with 10% FBS, 50 U/mL
penicillin, and 50 mg/mL streptomycin). Five TnonMet cell lines (368T1, 393T1, 394T4, 802T4,
2557T1), six TMet cell lines (373T1, 373T2, 389T2, 393T3, 393T5, 482T1) and five Met cell
lines (373N1, 393N1, 393M1, 482N1, 482M1) were used for subsequent gene expression
analysis and/or functional experiments in this study.
cDNA expression and knockdown
To generate TMet-TRE-Nkx2-1, -Foxa2, or -Cdx2 cell lines, the respective coding
sequences were cloned into the lentiviral expression vector pCW22tre-optimegaUbcrtTA,
which contains a TRE promoter for doxycycline-induction of the cDNA and a UBC promoter
for constitutive expression of rtTA. In addition, to allow constitutive overexpression of Nkx2-1
in TMet cells, the Nkx2-1 coding sequence was cloned into the retroviral expression vector
MSCV. These vectors were used to infect TMet cells.
To generate TnonMet-shNFC cells, shRNAs targeting Nkx2-1, Foxa2, Cdx2, or
Luciferase were cloned into lentiviral vectors and used to infect TnonMet cells. These cells
were subsequently infected with lentiviral vectors expressing shRNAs to knockdown Tks5long
or Hmga2, or FLAG-Cas9 and sgRNAs targeting Hmga2 or Snail.
154
List of shRNAs
Gene
shRNA ID
miR30 shRNA sequence
Nkx2-1
shNkx2-1
TGCTGTTGACAGTGAGCGACGCCATGTCTTGTTCTACCTTTAGTGAAGC
CACAGATGTAAAGGTAGAACAAGACATGGCGCTGCCTACTGCCTCGGA
Foxa2
shFoxa2
TGCTGTTGACAGTGAGCGACCACAGTGATCTGTCATTCTATAGTGAAGC
CACAGATGTATAGAATGACAGATCACTGTGGCTGCCTACTGCCTCGGA
Cdx2
shCdx2
TGCTGTTGACAGTGAGCGCCTGGGCTTTCTTCTCCACAAATAGTGAAGC
CACAGATGTATTTGTGGAGAAGAAAGCCCAGATGCCTACTGCCTCGGA
Tks5long
shTks5long#1
TGCTGTTGACAGTGAGCGCCTGGATAAGTTTCCTATTGAATAGTGAAGC
CACAGATGTATTCAATAGGAAACTTATCCAGATGCCTACTGCCTCGGA
Tks5long
shTks5long#2
TGCTGTTGACAGTGAGCGACACATTTCACAGTGTGACGAATAGTGAAGC
CACAGATGTATTCGTCACACTGTGAAATGTGGTGCCTACTGCCTCGGA
Hmga2
shHmga2#1
TGCTGTTGACAGTGAGCGAAAGGACTATATTAATCACTTTTAGTGAAGC
CACAGATGTAAAAGTGATTAATATAGTCCTTCTGCCTACTGCCTCGGA
Luciferase
shLuciferase
TGCTGTTGACAGTGAGCGCCCGCCTGAAGTCTCTGATTAATAGTGAAGC
CACAGATGTATTAATCAGAGACTTCAGGCGGTTGCCTACTGCCTCGGA
shRenilla
TGCTGTTGACAGTGAGCGCAGGAATTATAATGCTTATCTATAGTGAAGC
CACAGATGTATAGATAAGCATTATAATTCCTATGCCTACTGCCTCGGA
sgRNA ID
sgRNA target sequence (5’ to 3’)
(PAM sequence in bold)
sgHmga2#2
GTCCTCGCTTCTGTGGCACCTGG
Snail
sgSnail#1
GTCCTGCAGCTCGCTATAGTTGG
Snail
sgSnail#2
CGCTATAGTTGGGCTTCCGGCGG
sgRosa
GAAGATGGGCGGGAGTCTTCTGG
Renilla
List of sgRNAs
Gene
Hmga2
mRosa26
Immunoblotting
Immunoblotting was performed on whole-cell lysates, using antibodies against Nkx21 (Epitomics 2044), Foxa2 (Cell Signaling 8186), Cdx2 (Cell Signaling 12306), Tks5 (Santa
Cruz sc-30122), Hmga2 (Cell Signaling 5269), Snail (Cell Signaling 3879), HSP90 (Cell
Signaling 4877), and β-tubulin (Cell Signaling 2128).
155
qRT-PCR
RNA was purified from cultured cells or tissues using the RNAqueous kit (Invitrogen),
and was reverse-transcribed using a High-Capacity cDNA Reverse Transcription Kit
(Applied Biosystems). Quantitative RT-PCR was performed using SYBR Green Jumpstart
Taq Ready Mix (Sigma), or TaqMan Fast Universal Master Mix and probes (Applied
Biosystems).
List of qRT-PCR primers and Taqman probes
qRT-PCR primers
Target
gene
Forward
Reverse
TBP
GGGGAGCTGTGATGTGAAGT
CCAGGAAATAATTCTGGCTCA
Tks5long
TTATCAACGTGACCTGGTCTG
TTCGGATCCTTCTGGCCAC
Tks5short
TGGCTCACCGCGTGCTTTCTG
CCTTGCTCTTCAGATGTGCTCACAA
Nkx2-1
GCTGTCCTGCTGCAGTTGTTG
AGCTCGAGCGACGTTTCAAG
Foxa1
Taqman
probe
Mm00446971_m1
Mm00484713_m1
Foxa2
GACTGGAGCAGCTACTACGC
TCATTCCAGCGCCCACATAG
Cdx2
CAGCAGTCCCTAGGAAGCCA
GCAGCCAGCTCACTTTTCCT
Hnf4a
AGAGGTTCTGTCCCAGCAGATC
CGTCTGTGATGTTGGCAATC
Hmga2
GGGCAGCCGTCCACATCAGC
TCACAGGTTGGCTCTTGCTGC
Snail
Mm00441533_g1
Slug
Mm00441531_m1
Twist1
Mm00442036_m1
Vimentin
Mm01333430_m1
Zeb1
Mm00495564_m1
Cdh1
Mm01247357_m1
Cdh2
Mm01162497_m1
Krt17
Mm00495207_m1
Krt19
Mm00492980_m1
Angptl2
Mm00507897_m1
Dock8
Mm00613802_m1
Dapk2
Mm00802402_m1
Mtus1
Mm00628662_m1
TGFβ
Mm01178820_m1
156
ChIP-qPCR
Adherent TMet cells (393T5) overexpressing Nkx2-1, Foxa2, or Cdx2 were washed
once in PBS, and cross-linked in 1% formaldehyde diluted in PBS for 10 min at room
temperature. The reaction was stopped by 100 mM glycine, followed by 5mg/ml BSA in
PBS, and subsequently washed twice in cold PBS. Cells were harvested and re-suspended
in lysis buffer [50mM Tris-HCl, pH 8.1, 10mM EDTA, 1% SDS, 1X complete protease
inhibitors (Roche)], and sonicated with a Diagenode Bioruptor to obtain 300-500 bp
fragment size. Fragmented chromatin was diluted in IP buffer [20mM Tris-HCl pH 8.1,
150mM NaCl, 2mM EDTA, 1% Triton X-100] and incubated overnight at 4°C with Protein G
magnetic beads (Dynabeads: Invitrogen) that had been pre-incubated with antibodies
against Nkx2-1 (Bethyl A300-BL4000)), Foxa2 (Santa Cruz sc-6554), or isotype controls
(rabbit IgG and goat IgG, Abcam). Immunoprecipitates were washed six times with wash
buffer [50mM HEPES pH 7.6, 0.5M LiCl, 1mM EDTA, 0.7% Na deoxycholate, 1% NP-40]
and twice with TE buffer. Immunoprecipitated (or no IP input) DNA recovered in 100ul 1x
Elution Buffer [1% SDS, 0.1M NaHCO3] over 6 hours at 65°C, and column purified with
QiaQuick columns (Qiagen). qRT-PCR was performed using Fast Sybr Green Master Mix
(Applied Biosystems) on a StepOne Plus Real Time PCR system.
List of primers for ChIP-qPCR:
Genetic locus
Forward primer
Reverse primer
GD8
GGGCACTGCTAAACTCTTGC
GATGTGGGAGACTGGAGGAA
SftpA
TTGCCTTCTGTGGTTCTGTG
TACACAGCTCAGGTTCCTTCAG
Hnf4a
TGCCATGACAAAAGCATGAC
GGTGGGTGGATACGTTAAACAG
Tks5long-enhancer1
AGCGTGTACTTCATAGGGCT
TTGTCAGAGTCTCGCCAACA
Tks5long-enhancer2
CAGTGTCTCTTCCGGAGTCA
GGTTAGCAAAGCCCTTGTCC
Tks5long-enhancer3
TTTGACTGATTCGCGGCTCT
ACTGGCAGTAGTGTGCATGG
Tks5long-enhancer4
TCTCCTCCATGACCTAGCCC
GGCCAACAGTACATGAGCCA
Cdx2
CGGTTGCTATGCGTTGCC
157
GCCGACTTTTGAACCTCTAACC
Transplantation assays for metastasis
For subcutaneous transplantation, 5x104 cells resuspended in 100 µl PBS were
injected under the skin on the hind flank of nude mice. Mice were analyzed 6.5 weeks after
injection. For intravenous transplantation, 5x104 cells resuspended in 100 µl PBS were
injected into the lateral tail vein. Animals were analyzed 2.5 weeks post-injection.
Intravital imaging
Multiphoton imaging of GFP-labeled tumors was performed as described previously
(Wyckoff et al., 2011). Briefly, subcutaneous tumors at 5-6 weeks post-injection were
exposed by skin flap surgery performed on anesthetized animals. Tumors were imaged with
an Olympus FV1000 multiphoton microscope using a 25x 1.05 NA water immersion
objective with correction lens. Thirty-minute time-lapse movies were analyzed for frequency
of migratory GFP-positive tumor cells using ImageJ. Three mice were used per condition,
with 4-7 fields imaged per mouse.
In vivo fine-needle collection assay
The in vivo invasion assay was performed as previously described (Wyckoff et al.,
2000). In brief, 4-6 catheterized micro-needles held in place by micromanipulators were
inserted into the primary tumor of an anesthetized mouse. Needles contained a mixture of
10% Matrigel, 0.01 mM EDTA with L-15 media +/- 10% FBS. After 4 hr, the contents of the
needle were extruded and the total number of tumor cells that migrated into each needle
was quantified using DAPI. Three mice were used per condition.
158
Histology and immunohistochemistry
Tissues for histology were fixed in 10% formalin for 24 hours and stored in 70%
ethanol until paraffin embedding. Histological analysis for tumor grade was performed by a
pathologist (R.T.B.) on 4-µm sections stained with haematoxylin and eosin (H&E).
Immunohistochemistry (IHC) was performed on a Thermo Scientific Autostainer 360
machine followed by a hematoxylin counterstain, using antibodies against Nkx2-1
(Epitomics), Foxa2 (Cell Signaling), Cdx2 (Cell Signaling), or Hmga2 (Biocheck).
RNA-sequencing (RNA-seq) analysis
RNA-seq analysis was performed on 394T4 TnonMet, TnonMet-shNkx2-1, TnonMet-shNFC,
and TMet 373T1 cells in biological duplicates. RNA was isolated with the RNAqueous Total
RNA Isolation Kit (Life Technologies), and cDNA libraries were prepared with the TruSeq
RNA Sample Preparation Kit (Illumina). Sequencing was performed on an Illumina HiSeq
2000 instrument to obtain single-end 40-nt reads. All reads that passed quality metrics were
mapped to the UCSC mm9 mouse genome build (http://genome.ucsc.edu/) using RSEM (Li
and Dewey, 2011). Raw estimated expression counts were upper-quartile normalized to a
count of 1000 (Bullard et al., 2010). Independent Component Analysis (ICA) was performed
as described below to identify biologically relevant signatures that characterize the global
gene expression profiles of these samples. Targeted differential analysis for overlaps with
TnonMet/TMet/Met dataset was performed using EBSeq v1.4.0. All RNA-seq analyses were
conducted in the R Statistical Programming language (http://www.r-project.org/), including
signature analysis, hierarchical clustering, and multidimensional scaling (MDS). Gene set
enrichment analysis (GSEA) was carried out using the pre-ranked mode with default
settings (Subramanian et al., 2005). Heatmaps were generated using the Heatplus package
in R.
159
Clinical analysis
RNA-seq gene expression profiles of the primary tumors and the relevant clinical
data of 488 lung adenocarcinoma patients were obtained from the Cancer Genome Atlas
(TCGA; http://cancergenome.nih.gov/). Independent Component Analysis (ICA) was
performed as described below to identify biologically relevant signatures that characterize
the gene expression patterns of NKX2-1, FOXA2, CDX2, and HMGA2 in these human
tumors. Kaplan-Meier survival analysis was conducted with patients in the top 10th percentile
of each signature, and significance was assessed using log-rank test. Multivariate Cox
proportional hazard regression analysis with adjustment for gender, age, and stage was
performed on the overall survival of patients in the top 10th percentile of each signature.
Independent Component Analysis (ICA)
For the analysis of RNA-Seq samples sequenced in this study and the comparative
analysis with TCGA dataset, an unsupervised blind source separation strategy using
Independent Component Analysis (ICA) was applied to elucidate statistically independent
gene expression signatures within RNA-Seq expression data (Bhutkar et al.; Hyvärinen and
Oja, 2000; Rutledge and Jouan-Rimbaud Bouveresse, 2013). ICA is a general-purpose
signal processing and multivariate data analysis technique in the category of unsupervised
matrix factorization methods. Based on input data consisting of a genes-samples matrix,
ICA uses higher order moments to characterize the dataset as a linear combination of
statistically independent latent variables. These latent variables represent independent
components based on maximizing non-gaussianity, and can be interpreted as independent
source signals that have been mixed together to form the dataset under consideration. Each
component includes a weight assignment to each gene that quantifies its contribution to that
component. Additionally, ICA derives a mixing matrix that describes the contribution of each
160
sample towards the signal embodied in each component. This mixing matrix can be used to
select signatures among components with distinct gene expression profiles across the set of
samples. All computations were done in the R Statistical Programming Language. The R
implementation of the core JADE algorithm (Joint Approximate Diagonalization of
Eigenmatrices) (Biton et al., 2013; Nordhausen et al., 2012; Rutledge and Jouan-Rimbaud
Bouveresse, 2013) was used along with custom R utilities.
Other statistical analyses
All other statistical analyses were performed using Student’s T-test, unless otherwise
specified. P-values < 0.05 (two-tailed) were considered statistically significant.
161
ACKNOWLEDGEMENTS
We thank the Swanson Biotechnology Center, and especially Jeff Wyckoff,
Eliza Vasile, Denise Crowley, Kathleen Cormier, Michael Brown, and Michele Griffin for
technical support. We also thank Hideo Watanabe for assistance with ChIP-PCR; Monte
Winslow, Nadya Dimitrova, Mandar Muzumdar, Nikhil Joshi, Wen Xue, Thales
Papagiannakopoulos, Francisco Sanchez-Rivera, Tuomas Tammela, Kim Mercer, Kim
Dorans, and the entire Jacks lab for advice and experimental assistance. This work was
partially supported by the Cancer Center Support Grant (CCSG) P30-CA14051 from the
National Cancer Institute, grants from the Howard Hughes Medical Institute and the National
Institutes of Health (5-U01-CA84306) to T.J., DoD Breast Cancer Research Program grant
(W81XWH-12-1-0031) to M.J.O, and funds from the Ludwig Center at MIT to F.B.G. T.J. is a
Howard Hughes Investigator and a Daniel K. Ludwig Scholar.
162
SUPPLEMENTAL FIGURES
Supplemental Figure S1. Related to Figure 1
(A) Normalized expression values of Nkx2-1, Foxa2, and Cdx2 in TnonMet, TMet, and Distant
Met cells from microarray data in Winslow et al. (2011). Data are presented as Box-whisker
plot (5%–95%). The p-values were calculated by Student’s t test.
(B) Protein level of Nkx2-1, Foxa2, and Cdx2 in representative TnonMet (802T4, 394T4,
368T1), TMet (482T1, 373T1, 393T5, 393T3), and Met (482M1, 393M1) cells.
(C) Knockdown of Nkx2-1, Foxa2, and Cdx2 in an independent TnonMet cell line (368T1)
increases Tks5long expression, but not Tks5short, as measured by qRT-PCR. Data are
represented as mean ± SD. The p-value was calculated by Student’s t test.
(D) Combined overexpression of Foxa2 and Nkx2-1 (left) or Cdx2 and Nkx2-1 (right) in an
independent TMet cell line (393T3) further represses Tks5long, but not Tks5short, compared to
single overexpression, as measured qRT-PCR. Foxa2 and Cdx2 are expressed in a
doxycycline-inducible manner, while Nkx2-1 is expressed constitutively. Data are
represented as mean ± SD. The p-value was calculated by Student’s t test. **p < 0.01, ***p
< 0.001.
163
Supplemental Figure S2. Related to Figure 2
Size of subcutaneous tumors after transplantation of 394T4 TnonMet cells, TnonMet with
single/double knockdown, TnonMet-shNFC cells, and 373T1 TMet cells. Each circle represents
an individual mouse. Lines (-) indicate control hairpins against firefly or renilla luciferase.
Data are represented as mean ± SEM.
164
Supplemental Figure S3. Related to Figure 3.
(A) TnonMet and TnonMet-shNFC cells (394T4) in culture condition have similar morphology,
unlike their distinct epithelial and mesenchymal morphology in vivo.
(B) TnonMet-shNFC (394T4) subcutaneous tumors lose epithelial marker Krt19, and partially
gain mesenchymal markers Twist, Snail, Zeb1, and Cdh2 compared to TnonMet-shCtrl tumors.
Each circle represents an individual mouse. Data are represented as mean ± SEM. The pvalue was calculated by Student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001.
165
Supplemental Figure S4. Related to Figure 4
(A) Multidimensional Scaling (MDS) clustering showing sample relationships before and
after subtraction of clonal background signature. Removal of cell line-dependent clonal
signature unveils underlying clustering of TnonMet-shNFC/TMet cells away from TnonMet/TnonMetshN cells.
(B) qRT-PCR validation of expression changes of pro-metastatic gene TGFβ and antimetastatic gene Mtus1 that were identified in shNFC signature. Data are represented as
mean ± SD. The p-value was calculated by Student’s t test.
166
Supplemental Figure S5. Related to Figure 5
(A) Normalized expression values of Hmga2 and Snail in TnonMet, TMet, and Distant Met cells
from microarray data in Winslow et al. (2011). Data are presented as Box-whisker plot (5%–
95%). The p values were calculated by Student’s t test.
(B) qRT-PCR detection of knockdown of Tks5long (shTks5long#1), Hmga2 (shHmga2#1), and
Snail (sgSnail#1) in 394T4 TnonMet-shNFC cells. Data are represented as mean ± SD.
(C) Hmga2 and Snail knockdown did not affect the expression of each other, or the
expression of Tks5long, as measured by qRT-PCR. Data are represented as mean ± SD.
(D) Size of subcutaneous tumors after transplantation of 394T4 TnonMet-shNFC cells, and
TnonMet-shNFC cells with knockdown of Tks5long, Hmga2 or Snail. Each circle represents an
individual mouse. Data are represented as mean ± SEM.
167
Supplemental Figure S6. Related to Figure 6
(A) Representative tumor where moderately-differentiated Cdx2-positive area is adjacent to
and not overlapping with poorly-differentiated Hmga2-positive area. Scale bar represents
150 µm.
(B-C) Expression of Cdx2 detected by qRT-PCR (B) and immunoblotting (C) in two
independent TnonMet cell lines upon knockdown of Foxa2, Nkx2-1, or both factors. (B) 394T4
and 368T1 cells. Data are represented as mean ± SD. (C) 394T4 cells.
(D) ChIP-qPCR detects binding of Nkx2-1 and Foxa2 to an enhancer of the Cdx2 genomic
locus. Data are represented as mean ± SEM of two independent experiments (for Nkx2-1
ChIP) and five independent experiments (for Foxa2 ChIP). SftpA and Hnf4a serve as
positive controls for Nkx2-1 and Foxa2 binding, respectively. GD8: negative control mapping
to a gene desert region on murine chromosome 8. For each enhancer versus GD8, p < 0.05
by Student’s t test.
(E) A model for regulation of Cdx2 expression by Nkx2-1 and Foxa2.
168
Supplemental Figure S7. Expression level of Foxa1 in TnonMet, TMet, and Met cell lines.
Foxa1 was not differentially expressed between TnonMet (368T1, 394T4, 802T4), TMet (373T1,
393T3, 393T5, 482T1), and Met (373N1, 393M1, 393N1, 482M1, 482N1) cells as measured
by qRT-PCR. Data are represented as mean ± SEM. The p-value was calculated by
Student’s t test.
169
REFERENCES
Beck, F., and Stringer, E.J. (2010). The role of Cdx genes in the gut and in axial
development. Biochem Soc Trans 38, 353–357.
Besnard, V., Wert, S.E., Hull, W.M., and Whitsett, J.A. (2004). Immunohistochemical
localization of Foxa1 and Foxa2 in mouse embryos and adult tissues. Gene Expr Patterns 5,
193–208.
Bhutkar, A., Blat, I., Boutz, P.L., Cameron, E.R., Chen, P.Y., Chen, S., Ferretti, R., Gurtan,
A.M., Ianari, A., Muzumdar, M.D., et al. High-resolution signature discovery in NGS
expression datasets using Blind Source Separation (In Submission).
Biton, A., Zinovyev, A., Barillot, E., and Radvanyi, F. (2013). MineICA: Independent
component analysis of transcriptomic data.
Bullard, J.H., Purdom, E., Hansen, K.D., and Dudoit, S. (2010). Evaluation of statistical
methods for normalization and differential expression in mRNA-Seq experiments. BMC
Bioinformatics 11, 94.
DuPage, M., Dooley, A.L., and Jacks, T. (2009). Conditional mouse lung cancer models
using adenoviral or lentiviral delivery of Cre recombinase. Nat Protoc 4, 1064–1072.
Grimminger, P., Ling, F.C., Neiss, S., Vallböhmer, D., Lurje, G., Schneider, P.M., Hölscher,
A.H., Metzger, R., and Brabender, J. (2009). The role of the homeobox genes BFT and
CDX2 in the pathogenesis of non-small cell lung cancer. Anticancer Res 29, 1281–1286.
Guo, R.J., Suh, E.R., and Lynch, J.P. (2004). The role of Cdx proteins in intestinal
development and cancer. Cancer Biol Ther 3, 593–601.
Hyvärinen, A., and Oja, E. (2000). Independent component analysis: algorithms and
applications. Neural Netw 13, 411–430.
Jackson, E., Olive, K., Tuveson, D., Bronson, R., Crowley, D., Brown, M., and Jacks, T.
(2005). The differential effects of mutant p53 alleles on advanced murine lung cancer.
Cancer Res 65, 10280–10288.
Jackson, E., Willis, N., Mercer, K., Bronson, R., Crowley, D., Montoya, R., Jacks, T., and
Tuveson, D. (2001). Analysis of lung tumor initiation and progression using conditional
expression of oncogenic K-ras. Gene Dev 15, 3243–3248.
Kaestner, K.H. (2010). The FoxA factors in organogenesis and differentiation. Curr Opin
Genet Dev 20, 527–532.
Kimura, S., Hara, Y., Pineau, T., Fernandez-Salguero, P., Fox, C.H., Ward, J.M., and
Gonzalez, F.J. (1996). The T/ebp null mouse: thyroid-specific enhancer-binding protein is
essential for the organogenesis of the thyroid, lung, ventral forebrain, and pituitary. Gene
Dev 10, 60–69.
Li, B., and Dewey, C.N. (2011). RSEM: accurate transcript quantification from RNA-Seq
170
data with or without a reference genome. BMC Bioinformatics 12, 323.
Li, C.M.-C., Chen, G., Dayton, T.L., Kim-Kiselak, C., Hoersch, S., Whittaker, C.A., Bronson,
R.T., Beer, D.G., Winslow, M.M., and Jacks, T. (2013). Differential Tks5 isoform expression
contributes to metastatic invasion of lung adenocarcinoma. Gene Dev 27, 1557–1567.
Matsumoto, K., Mizoshita, T., Tsukamoto, T., Ogasawara, N., Hirata, A., Shimizu, Y.,
Haneda, M., Yamao, K., and Tatematsu, M. (2004). Cdx2 expression in pancreatic tumors:
Relationship with prognosis of invasive ductal carcinomas. Oncol Rep 12, 1239–1243.
Minoo, P., Su, G., Drum, H., Bringas, P., and Kimura, S. (1999). Defects in
tracheoesophageal and lung morphogenesis in Nkx2.1(-/-) mouse embryos. Dev Biol 209,
60–71.
Mizoshita, T., Tsukamoto, T., Nakanishi, H., Inada, K.-I., Ogasawara, N., Joh, T., Itoh, M.,
Yamamura, Y., and Tatematsu, M. (2003). Expression of Cdx2 and the phenotype of
advanced gastric cancers: relationship with prognosis. J Cancer Res Clin 129, 727–734.
Murphy, D.A., and Courtneidge, S.A. (2011). The 'ins' and “outs” of podosomes and
invadopodia: characteristics, formation and function. Nat Rev Mol Cell Bio 12, 413–426.
Nordhausen, K., Cardoso, J.F., Miettinen, J., Oja, H., and Ollila, E. (2012). JADE: JADE and
other BSS methods as well as some BSS performance criteria (R package version).
Paz, H., Pathak, N., and Yang, J. (2014). Invading one step at a time: the role of
invadopodia in tumor metastasis. Oncogene 33, 4193–4202.
Rutledge, D.N., and Jouan-Rimbaud Bouveresse, D. (2013). Independent Components
Analysis with the JADE algorithm. TrAC Trends in Analytical Chemistry 50, 22–32.
Sato, M., Shames, D.S., and Hasegawa, Y. (2012). Emerging evidence of epithelial-tomesenchymal transition in lung carcinogenesis. Respirology 17, 1048–1059.
Snyder, E.L., Watanabe, H., Magendantz, M., Hoersch, S., Chen, T.A., Wang, D.G.,
Crowley, D., Whittaker, C.A., Meyerson, M., Kimura, S., et al. (2013). Nkx2-1 represses a
latent gastric differentiation program in lung adenocarcinoma. Mol Cell 50, 185–199.
Steeg, P.S. (2006). Tumor metastasis: mechanistic insights and clinical challenges. Nat Med
12, 895–904.
Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette, M.A.,
Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., et al. (2005). Gene set enrichment
analysis: a knowledge-based approach for interpreting genome-wide expression profiles. P
Natl Acad Sci Usa 102, 15545–15550.
Thiery, J.P., Acloque, H., Huang, R.Y.J., and Nieto, M.A. (2009). Epithelial-mesenchymal
transitions in development and disease. Cell 139, 871–890.
Tsai, J.H., and Yang, J. (2013). Epithelial-mesenchymal plasticity in carcinoma metastasis.
Gene Dev 27, 2192–2206.
171
Wan, H., Dingle, S., Xu, Y., Besnard, V., Kaestner, K.H., Ang, S.-L., Wert, S., Stahlman,
M.T., and Whitsett, J.A. (2005). Compensatory roles of Foxa1 and Foxa2 during lung
morphogenesis. Jbc 280, 13809–13816.
Wan, H., Kaestner, K.H., Ang, S.-L., Ikegami, M., Finkelman, F.D., Stahlman, M.T.,
Fulkerson, P.C., Rothenberg, M.E., and Whitsett, J.A. (2004). Foxa2 regulates
alveolarization and goblet cell hyperplasia. Development 131, 953–964.
Winslow, M.M., Dayton, T.L., Verhaak, R.G.W., Kim-Kiselak, C., Snyder, E.L., Feldser, D.M.,
Hubbard, D.D., Dupage, M.J., Whittaker, C.A., Hoersch, S., et al. (2011). Suppression of
lung adenocarcinoma progression by Nkx2-1. Nature 473, 101–104.
Wyckoff, J.B., Segall, J.E., and Condeelis, J.S. (2000). The collection of the motile
population of cells from a living tumor. Cancer Res 60, 5401–5404.
Wyckoff, J., Gligorijevic, B., Entenberg, D., Segall, J., and Condeelis, J. (2011). HighResolution Multiphoton Imaging of Tumors In Vivo. Cold Spring Harb Protoc 2011,
pdb.top065904–pdb.top065904.
Yatabe, Y., Koga, T., Mitsudomi, T., and Takahashi, T. (2004). CK20 expression, CDX2
expression, K-ras mutation, and goblet cell morphology in a subset of lung
adenocarcinomas. J Pathol 203, 645–652.
Yuasa, Y. (2003). Control of gut differentiation and intestinal-type gastric carcinogenesis.
Nat Rev Cancer 3, 592–600.
Zhou, L., Dey, C.R., Wert, S.E., Yan, C., Costa, R.H., and Whitsett, J.A. (1997). Hepatocyte
nuclear factor-3beta limits cellular diversity in the developing respiratory epithelium and
alters lung morphogenesis in vivo. Dev. Dyn. 210, 305–314.
172
CHAPTER 4
DISCUSSION AND FUTURE DIRECTIONS
173
In this thesis, I have utilized tools from a mouse model of lung adenocarcinoma,
primary tumor cell lines, and gene expression analysis of human lung adenocarcinoma to
investigate the molecular regulatory mechanisms of metastasis. Using these approaches, I
have shown that the metastatic invasion of tumor cells is regulated by the isoform
expression of the invadopodia component Tks5. In addition, I have established that
expression of the pro-metastatic isoform of Tks5, Tks5long, is synergistically suppressed by
the transcription factors Nkx2-1, Foxa2, and Cdx2. Furthermore, I have demonstrated that
combined loss of Nkx2-1, Foxa2, and Cdx2 leads to the activation of a number of
metastasis-related genes besides Tks5long, and these alterations in gene expression
promote metastasis of lung adenocarcinoma. Finally, I have provided evidence from mouse
and human tumors that supports a model for dynamic expression of Nkx2-1, Foxa2, and
Cdx2 that correlates with lung adenocarcinomas progression. Altogether these findings
contribute to our understanding of the complex mechanisms that regulate cancer
metastasis.
Tks5 isoforms regulate invadopodia and metastatic spread of lung adenocarcinoma
In an effort to identify the gene expression alterations that occur during metastasis of
lung adenocarcinoma, we examined the transcriptome profiles of non-metastatic and
metastatic KrasG12D/+; p53-/- (KP) tumor cell lines using microarray-based data (Winslow et
al., 2011). Our analysis led us to identify two previously uncharacterized isoforms of Tks5,
as they exhibited the most striking pattern of differential isoform expression between nonmetastatic and metastatic cells out of all the isoform expression changes that we have
analyzed in the microarray dataset. These two variants, which we named Tks5long and
Tks5short, share the same 3’ coding sequences at the Tks5 genetic locus, but differ in the
inclusion and exclusion of the 5’ exons that encode the phox homology (PX) domain. Each
174
isoform is transcribed from an independent promoter, with Tks5short being the dominant
isoform in non-metastatic cells, and Tks5long as the dominant isoform in metastatic cells. Our
functional studies have demonstrated that Tks5long promotes invadopodia-mediated matrix
degradation, and is required for metastasis in vivo. In contrast, Tks5short inhibits invadopodia
function by destabilizing invadopodia foci and preventing their maturation. Furthermore, our
analysis of clinical data has indicated that a high ratio of Tks5long-to-Tks5short expression
correlates with more advanced tumor stage and worse survival of patients.
Taken together, our findings have revised the previous notion in the field that Tks5
generally promotes invadopodia, and instead established that distinct isoforms of Tks5
regulate invadopodia activity in antagonizing manner. As such, a balance between Tks5long
and Tks5short expression is critical for invadopodia activity and metastasis. Importantly, while
many positive regulators of invadopodia have been previously identified, our understanding
of the negative regulatory mechanisms of invadopodia remains relatively limited (see
Introduction for more details). Our study of Tks5short has revealed a novel mechanism for
restraining invadopodia activity. Given that the normal-cell counterparts of invadopodia,
podosomes, play significant roles in developmental processes as well as normal physiology
and therefore are likely to be tightly regulated, it will be important to investigate the extent to
which Tks5short is required to constrain podosome activity in these contexts. To this end,
isoform-specific loss of function experiments by RNAi may be challenging given that
Tks5short only has 585 nucleotides that are unique from Tks5long. However, silencing of
Tks5short expression may be achieved by applying the emerging CRISPR genome-editing
technology to specifically disrupt the promoter of Tks5short, without affecting the transcription
of Tks5long. Additionally, the exact mechanisms by which Tks5short destabilize invadopodia
remains to be further characterized. Because Tks5short lacks the PX domain necessary for
membrane localization but retains the SH3 domains that are capable of binding other
175
invadopodia components, we propose that Tks5short may function by sequestering
invadopodia components away from the cell membrane, thereby preventing proper
assembly and maturation of invadopodia. This model is supported by our observation that
Tks5short is diffusely localized in the cytoplasm, and that the invadopodia foci are short-lived
and lack proteolytic function. Similar observation has been reported in studies using a ΔPXmutant form of Tks5long (Santiago-Medina et al., 2015). Furthermore, isolated SH3 domains
of Tks5 have been shown to be capable of interacting with various invadopodia components
in co-immunoprecipitation experiments (Abram and Courtneidge, 2003; Rufer et al., 2009).
This hypothesis can be further tested in future studies by examining the interactions
between Tks5short and invadopodia components.
Another contribution from our findings is providing in vivo evidence to support a role
of invadopodia in metastasis. Because most studies correlating the role of invadopodia to
metastasis have historically relied on in vitro assays or transplantation experiments, there
has been skepticism in the field regarding whether invadopodia have a functional role in
clinically relevant settings of metastasis. Here, by providing evidence from an
autochthonous mouse model of lung cancer and clinical data from lung adenocarcinoma
patients, our findings have argued that invadopodia do have a relevant function in
metastasis. Interestingly, the role of invadopodia in metastasis has primarily been attributed
to their ability to degrade the extracellular matrix (ECM). A complementing mechanism for
invadopodia to promote metastasis can be the release of growth factors from the ECM as a
result of invadopodia-mediated proteolysis. This mechanism is corroborated by our
observation that KP lung tumors that overexpress Tks5long exhibit an increase of tumor
grades. Furthermore, we have also found that TnonMet cells with triple knockdown of Nkx2-1,
Foxa2 and Cdx2 adopt a mesenchymal morphology only in vivo but not in vitro, suggesting
that this morphological change is induced by factors present in the tumor microenvironment.
176
However, these are indirect evidence, and more comprehensive proof will come from further
studies in vivo.
Tks5long expression is suppressed by cooperation between Nkx2-1, Foxa2, and Cdx2
The consistent upregulation of Tks5long in metastatic KP lung adenocarcinoma cells
compared to non-metastatic cells led us to hypothesize that there is a common regulatory
mechanism that activates Tks5long transcription in metastasis. We have found that Nkx2-1,
Foxa2, and Cdx2 function cooperatively to inhibit Tks5long level in non-metastatic cells, and
their loss in metastatic cells leads to activation of Tks5long expression. Our data have shown
that combined knockdown of the three transcription factors is required to fully induce Tks5long
expression. Furthermore, overexpression of each of the three factors is sufficient to
suppress Tks5long transcription.
While we have demonstrated that Tks5long expression can be regulated by these
transcription factors, other regulatory mechanisms for Tks5long expression exist. Recently,
Sara Courtneidge and colleagues have reported that the abundance of the Tks5long protein
in NIH3T3 mouse fibroblasts is increased in the presence of active Src kinase, while the
mRNA level of Tks5long remains unchanged (Cejudo-Martin et al., 2014). However, the
molecular mechanisms of how Src activity regulates Tks5long protein levels remain to be
elucidated. It is possible that the regulation by Src at the post-translational level allows for
more rapid adjustment of Tks5long expression than the regulation by transcription factors. Of
note, we saw that in our KP tumor cells, differences in active Src levels (as measured by
immunoblotting of Y418 phospho-Src) does not correlate with Tks5long protein levels. These
data suggest that Tks5long expression can be regulated in a context-dependent manner.
From the perspective of development, it is intriguing to consider why the
dedifferentiation process induced by loss of the lineage regulators Nkx2-1, Foxa2, and Cdx2
177
can lead to expression of Tks5long in lung cancer. This raises the possibility that Tks5long may
play a role in organogenesis of the lungs. Matrix metalloproteinases have been shown to
mediate branching and alveolarization during lung development, potentially by remodeling
the basement membrane and cleaving ligands to regulate growth factor-mediated signaling
(Kheradmand et al., 2002; Wang et al., 2010). Therefore, it is possible that formation of
invadopodia, driven by increased Tks5long expression, may occur during the branching
process of lung development. However, this hypothesis would assume that additional
regulatory mechanisms exist to allow Tks5long expression in the presence of Nkx2-1 and
Foxa2 in the developing lungs. This will be an interesting model to test in future studies.
Nkx2-1, Foxa2, and Cdx2 synergistically suppress lung adenocarcinoma metastasis
We have presented evidence that Nkx2-1, Foxa2, and Cdx2 collectively function to
inhibit metastasis, as loss of these three transcription factors enhanced metastasis in a
subcutaneous transplantation model. In particular, loss of Nkx2-1 alone promoted the
colonization step of metastasis, whereas combined loss of Nkx2-1, Foxa2 and Cdx2
promoted tumor cell migration. The increased metastatic ability upon loss of Nkx2-1, Foxa2,
and Cdx2 can be explained by activation of multiple metastasis-promoting genes in addition
to Tks5long, such as the embryonal proto-oncogene Hmga2 and the EMT mediator Snail. It is
very likely that only a subset of these activated genes are directly regulated by Nkx2-1,
Foxa2 and Cdx2, while other targets may be indirectly controlled via secondary transcription
factors downstream of Nkx2-1, Foxa2 and Cdx2. The direct targets of these three
transcription factors can be identified in the future by ChIP-seq analysis. Combining these
data with findings from expression patterns of the three transcription factors over the course
of tumor development in the KP mouse model and in human patients, we have proposed a
model for lung adenocarcinoma progression. This model starts with high Nkx2-1 and Foxa2
178
expressions that restrain tumors in a well-differentiated non-metastatic state. Subsequent
loss of Nkx2-1 leads to activation of Cdx2, shifting the cells to an aberrant, albeit still
differentiated, state. Finally, suppression of Foxa2 leads to loss of Cdx2, and these
alterations synergize with silencing of Nkx2-1 to induce a dedifferentiated stem-like state in
the tumor cells, leading to activation of a metastasis program.
Taken together, our observations lead to two important inferences. First, our findings
may reflect the redundant nature of the cellular program in impeding tumor progression. It is
noteworthy that combined loss of the three transcription factors is required to increase
metastatic potential of TnonMet cells to a level comparable to TMet cell, while loss of a single
transcription factor is insufficient to recapitulate the same effect. This hypothesis about
redundancy is corroborated by the results from a previous study in our laboratory, which has
shown that deletion of Nkx2-1 in the KP model of lung adenocarcinoma is not sufficient to
promote metastasis (Snyder et al., 2013). In fact, loss of Nkx2-1 expression at tumor
initiation leads to activation of a gastric differentiation program driven by Hnf4α, Foxa1 and
Foxa2 in these cancer cells. These observations further highlight the redundancy of
regulatory programs that need to be altered in the progression to metastasis.
Second, our observations suggest that a relatively small number of factors could be
responsible for the vast amount of gene expression alterations in metastasis. This
hypothesis is based on our finding that combined knockdown of the three transcription
factors in TnonMet cells can recapitulate a significant fraction of the gene expression changes
observed between TnonMet and TMet cells. Among these gene expression changes are both
driver events that directly promote metastasis, as well as passenger events that reflect the
difference in cellular states. Therefore, we propose that a small number of transcription
factors can function as regulatory nodes to govern a network of downstream targets either
directly or indirectly to mediate metastasis. In the future, it will be important to identify
179
additional transcription factors that are responsible for the other gene expression differences
between TnonMet and TMet cells.
Finally, it is also important to emphasize that transcriptional regulation is not the only
mechanism for activating the cellular changes that are required for metastasis. The
metastatic potential can also arise from non-transcriptional mechanisms, such as chromatin
modifications, alternative pre-mRNA splicing, protein modifications, protein localization, as
well as non-coding RNAs mediated functions. As such, some of the players in metastasis
may not be captured in our transcriptome analysis of non-metastatic and metastatic cells.
Therefore, a more holistic approach that integrates these global alterations between nonmetastatic and metastatic cells will be important to fully understand the metastasis program.
Tumor dedifferentiation versus tumor stem cell expansion
Some studies argue that the resemblance of advanced tumors to a stem-like state
can be explained by the expansion of cancer stem cells instead of a dedifferentiation
process in tumors. One example is the study by Zena Werb and colleagues on the
expression of the luminal cell fate regulator Gata3 in breast cancer (Kouros-Mehr et al.,
2008). Using an orthotropic transplantation model of MMTV-PyMT driven breast cancer, the
authors of the study observed that Gata3 expression was reduced in advanced,
dedifferentiated primary tumors and metastases, whereas forced overexpression of Gata3 in
advanced mammary carcinomas was sufficient to induce differentiation and inhibit
metastasis. Surprisingly, conditional deletion of Gata3 in early, well-differentiated tumor cells
did not lead to increased progression, but instead caused apoptosis of tumor cells. Based
on these observations, the authors concluded that mammary cancer progression and
metastasis are driven by the expansion of a small population of Gata3-negative tumor stem
cells. However, we propose that tumor dedifferentiation is an equally likely alternative
180
explanation for these observations, and this model will postulate that additional gene
expression alterations are required to synergize with loss of Gata3 to allow progression to
metastasis.
In the KP model of lung adenocarcinoma, several lines of evidence argue that
progression to the metastatic state is attained by dedifferentiation rather than expansion of
tumor stem cells. First, high-grade tumor regions with low Nkx2-1, Foxa2 and Cdx2 and high
Hmga2 expression are found to be located within larger low-grade regions with high Nkx2-1,
Foxa2 and Cdx2 and low Hmga2 expression, suggesting the high-grade region emerge from
the low-grade tumor. Second, the Hmga2-positive high-grade regions have been shown to
harbor the same lentiviral integrate sites in the genome as the surrounding Hmga2-negative
low-grade regions, thus unequivocally arguing that these associated regions originate from
the same tumor-initiating cells (Winslow et al., 2011). Third, our gene expression
experiments have demonstrated that non-metastatic cells can be altered to activate at least
part of the metastatsis program by silencing of Nkx2-1, Foxa2 and Cdx2. Together, these
findings have convincingly demonstrate that the metastatic program in lung adenocarcinoma
depends on a dedifferentiation process.
More generally, the models of tumor dedifferentiation and stem cell expansion are
likely to be non-mutually exclusive, and may occur in a manner dependent on the specific
tumor type. As such, the interplays between differentiation states and cancer progression
reflect the complexity of the regulatory mechanisms in metastasis.
Dedifferentiation and activation of alternative state in metastasis
Our observation that loss of Nkx2-1, Foxa2 and Cdx2 expression promotes
metastasis underscores the intricate link between metastasis and differentiation states,
particularly in terms of dedifferentiation and activation of alternative differentiation states.
181
First, the role of Nkx2-1 and Foxa2 in inhibiting lung adenocarcinoma progression
can be considered from a perspective of tumor dedifferentiation. Nkx2-1 and Foxa2 are both
required for specifying the lung lineage in embryos and adults. Therefore, it is possible that
loss of these two transcription factors favors tumor progression because of the loss of the
terminal lung differentiation state and reversal to a more stem-like cellular program. Our
results are corroborated by the data from Don Nguyen and colleagues (Cheung et al.,
2013), which have shown that two other lung transcription factors Gata6 and Hopx can
restrain metastasis of lung adenocarcinoma by maintaining the alveolar differentiation state.
Furthermore, numerous additional examples for a role of dedifferentiation in promoting
metastasis can be found in other cancer types (see Introduction for more details). In this
regard, our findings, along with the aforementioned studies, support a model in which
inactivation of cell lineage transcription factors contribute to tumor metastasis.
Second, the role of Cdx2 in inhibiting metastasis of lung adenocarcinoma can be
considered in the context of alternative differentiation state. We have observed that Cdx2
expression is upregulated upon loss of Nkx2-1, and is subsequently downregulated upon
additional silencing of Foxa2. The detection of Cdx2 in lung tumors is unexpected, because
Cdx2 is a lineage transcription factor for the small and large intestines, and its expression is
not observed in developing and adult lungs. This observation is not likely to be an artifact of
the mouse model, as previous studies on human lung adenocarcinomas have reported
expression of Cdx2 in a subset of patients (Grimminger et al., 2009; Yatabe et al., 2004).
While the role of Cdx2 in lung tumor progression has not been well characterized previously,
our findings strongly suggest that the activation of Cdx2 upon loss of Nkx2-1 in lung tumors
reflects the activation of an alternative differentiation program of the intestine that serves as
a redundant mechanism to restrain dedifferentiation and metastatic progression. The lungs
and the intestines are developmentally related, as they are both derived from the developing
182
gut tube. The foregut gives rise to the lungs, stomach, and part of the small intestine (part of
the duodenum), while the midgut and hindgut give rise to the remainder of the small
intestine (part of the duodenum, the jejunum, the ileum) and the large intestine. Importantly,
while we cannot formally exclude the possibility that these Cdx2-expressing cells represent
a dead end that will never further progress to metastasis, our data argue that this Cdx2positive state is more likely to be a transition state between loss of Nkx2-1 and upregulation
of Hmga2 during progression to metastasis. This hypothesis is based on our observation
that the increased Cdx2 expression in TnonMet-shNkx2-1 cells can be repressed upon further
knockdown of Foxa2. Future studies can distinguish between the two possible models by
using a reporter for Cdx2 expression in these lung tumors, for example by using a KP;
Cdx2-FlpO-ER; Rosa26-Frt-Stop-Frt-GFP mouse line that are infected with a lenti-Cre virus.
It is important to note that while the intestinal marker Cdx2 is activated in at least a subset of
KP tumors upon loss of Nkx2-1 expression during tumor progression, engineered deletion of
Nkx2-1 at tumor initiation in contrast leads to activation of a gastric program driven by Hnf4a
(Snyder et al., 2013). This difference may reflect the different cellular states of early and late
tumors, and the existence of multiple possible alternative differentiation states in the
progression of the same tumor type. Finally, activation of alternative differentiation programs
have been observed in other tumor types. For example, aberrant activation of intestinal
program (driven by Cdx2 and Cdx1) has been observed in gastric cancer, esophageal
cancer, nasal adenocarcinoma, pancreatic cancer, and ovarian cancer (Guo et al., 2004;
Matsumoto et al., 2004; Mizoshita et al., 2003; Yuasa, 2003), while foregut genes (driven at
least partly by Hedgehog/Gli) have been found to be upregulated in pancreatic intraepithelial
neoplasia (Prasad et al., 2005). Collectively, these observations reflect the dynamic nature
of differentiation states in tumor progression, and the redundancy of regulatory pathways
that can potentially restrain the progression to metastasis.
183
Potential mechanisms for loss of lineage transcription factors in metastasis
One outstanding question raised by findings from our study and others is what
caused the eventual downregulation of the lineage transcription factors during tumor
progression. This loss of expression is likely not due to genetic deletion or mutation, as
these sequence alterations are rarely detected (Basseres et al., 2012; Kouros-Mehr et al.,
2008). There is evidence that the silencing of gene expression is associated with chromatin
modifications such as methylation of the promoter and histones (Basseres et al., 2012).
However, it is not clear whether these epigenetic modifications are the upstream drivers of
the silencing event, or simply function as markers that reflect the transcriptional states of the
genes.
Here, we propose two models that might explain the loss of transcription factors in
tumor progression. One model is the induction of cellular changes by environmental stimuli,
such as changes in the oxygen level, metabolite concentrations, growth factors, hormones,
cytokines, or other signaling ligands present in the ECM or secreted by stromal cells. These
external signals can activate cellular pathways that lead to silencing of the lineage
transcription factors directly or indirectly. This model is partially supported by the finding that
TGFβ can inhibit Nkx2-1 expression in cultured human lung adenocarcinoma cells (Saito et
al., 2009). Furthermore, the shNFC signature we have identified is enriched for gene sets of
TGFβ targets. Another possible model is that the loss of lineage transcription factors is a
stochastic event. As a result of the noise and fluctuations of cellular processes, there may
be intrinsic heterogeneity in the expression levels of these transcription factors within the
tumor population. In either situation, those cells with lower expression of lineage
transcription factors may gain a selective advantage, and therefore eventually expand to
dominate the entire population of tumor cells.
184
These two models are not necessarily mutually exclusive. Furthermore, for these
models to be valid, the silencing effects induced by external stimuli or stochastic fluctuations
would need to remain relatively stable even after the stimuli or the selective pressure is
removed, perhaps through maintenance of the epigenetic state of the genes. This postulate
is required to explain the prevailing observation that the lineage transcription factors that
have been suppressed in the primary tumors remain inactive in the metastases in vivo or the
tumor-derived cell lines in vitro. In the future, it will be important to examine these
possibilities in order to obtain a more comprehensive understanding of the regulation of
metastasis.
Implications on therapeutic strategies
Traditionally, transcription factors are not perceived as ideal therapeutic targets for
treating cancer because they lack domains that can be inhibited by small molecules.
However, with the recent advancements of CRISPR/Cas based tools for gene activation and
inactivation, as well as the emerging technologies for tumor-specific deliveries of therapeutic
molecules, it may not be implausible to consider therapeutic strategies that reactivate the
silenced lineage-specific transcription factors in order to constrain the metastatic potential of
malignant cells. Given the widely observed suppression of lineage transcription factors in
various tumor types, such a treatment approach may provide great impact. However, the
findings that some of these transcription factors can also promote tumor progression by
acting as lineage-survival oncogenes call for a note of caution. While activation of these
transcription factors may induce differentiation and restrain metastasis, over-activation may
lead to expansion of tumor subpopulations that benefit from the proliferation advantage
conferred by these lineage transcription factors. Therefore, the expression levels of these
185
lineage transcription factors need to be precisely controlled in order for such therapeutic
strategies to be beneficial.
A unifying theme for differentiation and metastasis?
It is paradoxical that lineage transcription factors play dual roles in tumor
progression. On one hand, there is overwhelming evidence that suppression of these
transcription factors can promote tumor progression, including the findings in this thesis that
loss of Nkx2-1, Foxa2 and Cdx2 can hasten metastasis of lung adenocarcinoma. On the
other hand, amplifications and overexpression of these three transcription factors and others
have been found to favor tumorigenesis in various contexts, as we have discussed in the
Introduction.
Perhaps one unifying theme that can reconcile these observations is that tumor
progression is a process of evolutionary selection. While the end result of tumor metastasis
is invariably detrimental, the paths for tumors to reach a metastatic state are diverse. Just as
populations of organisms can evolve over time to adapt to environmental challenges, cancer
cells within a tumor are also capable of undergoing dynamic evolution and selection in
response to pressure or stress. As such, our understanding of the complexity of cancer
metastasis is an important first step to cure this disease.
186
REFERENCES
Abram, C., and Courtneidge, S. (2003). The adaptor protein fish associates with members of
the ADAMs family and localizes to podosomes of Src-transformed cells. Jbc 278, 16844–
16851.
Basseres, D.S., D'Alò, F., Yeap, B.Y., Loewenberg, E.C., Gonzalez, D.A., Yasuda, H.,
Dayaram, T., Kocher, O.N., Godleski, J.J., Richards, W.G., et al. (2012). Frequent
downregulation of the transcription factor Foxa2 in lung cancer through epigenetic silencing.
Lung Cancer 77, 31–37.
Cejudo-Martin, P., Yuen, A., Vlahovich, N., Lock, P., Courtneidge, S.A., and Diaz, B. (2014).
Genetic Disruption of the Sh3pxd2a Gene Reveals an Essential Role in Mouse
Development and the Existence of a Novel Isoform of Tks5. PLoS ONE 9, e107674.
Cheung, W.K.C., Zhao, M., Liu, Z., Stevens, L.E., Cao, P.D., Fang, J.E., Westbrook, T.F.,
and Nguyen, D.X. (2013). Control of alveolar differentiation by the lineage transcription
factors GATA6 and HOPX inhibits lung adenocarcinoma metastasis. Cancer Cell 23, 725–
738.
Grimminger, P., Ling, F.C., Neiss, S., Vallböhmer, D., Lurje, G., Schneider, P.M., Hölscher,
A.H., Metzger, R., and Brabender, J. (2009). The role of the homeobox genes BFT and
CDX2 in the pathogenesis of non-small cell lung cancer. Anticancer Res 29, 1281–1286.
Guo, R.J., Suh, E.R., and Lynch, J.P. (2004). The role of Cdx proteins in intestinal
development and cancer. Cancer Biol Ther 3, 593–601.
Kheradmand, F., Rishi, K., and Werb, Z. (2002). Signaling through the EGF receptor
controls lung morphogenesis in part by regulating MT1-MMP-mediated activation of
gelatinase A/MMP2. J Cell Sci 115, 839–848.
Kouros-Mehr, H., Bechis, S.K., Slorach, E.M., Littlepage, L.E., Egeblad, M., Ewald, A.J., Pai,
S.-Y., Ho, I.-C., and Werb, Z. (2008). GATA-3 links tumor differentiation and dissemination
in a luminal breast cancer model. Cancer Cell 13, 141–152.
Matsumoto, K., Mizoshita, T., Tsukamoto, T., Ogasawara, N., Hirata, A., Shimizu, Y.,
Haneda, M., Yamao, K., and Tatematsu, M. (2004). Cdx2 expression in pancreatic tumors:
Relationship with prognosis of invasive ductal carcinomas. Oncol Rep 12, 1239–1243.
Mizoshita, T., Tsukamoto, T., Nakanishi, H., Inada, K.-I., Ogasawara, N., Joh, T., Itoh, M.,
Yamamura, Y., and Tatematsu, M. (2003). Expression of Cdx2 and the phenotype of
advanced gastric cancers: relationship with prognosis. J Cancer Res Clin 129, 727–734.
Prasad, N.B., Biankin, A.V., Fukushima, N., Maitra, A., Dhara, S., Elkahloun, A.G., Hruban,
R.H., Goggins, M., and Leach, S.D. (2005). Gene expression profiles in pancreatic
intraepithelial neoplasia reflect the effects of Hedgehog signaling on pancreatic ductal
epithelial cells. Cancer Res 65, 1619–1626.
Rufer, A.C., Rumpf, J., Holleben, von, M., Beer, S., Rittinger, K., and Groemping, Y. (2009).
Isoform-Selective Interaction of the Adaptor Protein Tks5/FISH with Sos1 and Dynamins. J
187
Mol Biol 390, 939–950.
Saito, R.-A., Watabe, T., Horiguchi, K., Kohyama, T., Saitoh, M., Nagase, T., and Miyazono,
K. (2009). Thyroid transcription factor-1 inhibits transforming growth factor-beta-mediated
epithelial-to-mesenchymal transition in lung adenocarcinoma cells. Cancer Res 69, 2783–
2791.
Santiago-Medina, M., Gregus, K.A., Nichol, R.H., O'Toole, S.M., and Gomez, T.M. (2015).
Regulation of ECM degradation and axon guidance by growth cone invadosomes.
Development 142, 486–496.
Snyder, E.L., Watanabe, H., Magendantz, M., Hoersch, S., Chen, T.A., Wang, D.G.,
Crowley, D., Whittaker, C.A., Meyerson, M., Kimura, S., et al. (2013). Nkx2-1 represses a
latent gastric differentiation program in lung adenocarcinoma. Mol Cell 50, 185–199.
Wang, Q., Uhlirova, M., and Bohmann, D. (2010). Spatial restriction of FGF signaling by a
matrix metalloprotease controls branching morphogenesis. Dev Cell 18, 157–164.
Winslow, M.M., Dayton, T.L., Verhaak, R.G.W., Kim-Kiselak, C., Snyder, E.L., Feldser, D.M.,
Hubbard, D.D., Dupage, M.J., Whittaker, C.A., Hoersch, S., et al. (2011). Suppression of
lung adenocarcinoma progression by Nkx2-1. Nature 473, 101–104.
Yatabe, Y., Koga, T., Mitsudomi, T., and Takahashi, T. (2004). CK20 expression, CDX2
expression, K-ras mutation, and goblet cell morphology in a subset of lung
adenocarcinomas. J Pathol 203, 645–652.
Yuasa, Y. (2003). Control of gut differentiation and intestinal-type gastric carcinogenesis.
Nat Rev Cancer 3, 592–600.
188