| Discovery
of Bone Metastasis Genes by Functional Genomics
|
| Dec
1, 2005 |
| By:
Xin
Lu, Yibin
Kang |
| Pharmaceutical
Discovery |
|
Cancer metastasis is a major cause of morbidity and mortality among
cancer patients. Despite its clinical importance and the vast amount of
research on this subject, our understanding of metastasis remains rather
descriptive and fragmented. The genomic complexity of metastatic tumors
makes it exceedingly difficult to associate definite molecular properties
with metastatic phenotypes. The combination of genomic profiling
techniques with animal models of metastasis has provided exciting new
insights into this insidious process. In this article, we will discuss our
research on the functional genomic analysis of bone metastasis in breast
cancer and its potential impact on the clinical management of cancer
metastasis.
New tricks against an old foe
Research on cancer metastasis has been driven by two important clinical
manifestations of the disease. First, although metastasis is the direct
cause of over 90% of cancer deaths, it is in fact a very inefficient
process at the cellular level. Among the millions of cancer cells that
escape from the primary tumor, only a tiny fraction actually result in
clinically significant macrometastases. Secondly, different types of
cancer display different preferred sites of metastasis, which are not
completely explained by the pattern of blood flow alone. For example, bone
metastasis frequently occurs in breast, prostate, and lung cancer, but is
rarely seen in colorectal cancer. An explanation for metastasis
inefficiency and tissue tropism was first proposed by the late British
surgeon, Steven Paget, who published his landmark "seed and
soil" hypothesis over 100 years ago (1). Modern molecular biology has
begun to discover some of the molecular interactions between tumor cells
(seeds) and the stromal environment (soil) that dictate the efficiency and
tissue-specificity of metastasis formation. Molecules with biological
functions in adhesion, migration, proteolysis, chemotaxis, angiogenesis,
and signal transduction are investigated for their roles in cancer
metastasis. Overexpression of candidate metastasis genes in advance stage
tumors or metastases are often followed up with in vitro functional
evaluations or in vivo animal metastasis assays. This traditional
approach of metastasis gene discovery has been instrumental in the
identification of several key metastasis suppressors such as E-cadherin,
and metastasis promoters such as matrix metalloproteases (2, 3). However,
this approach is inherently biased as it often focuses on a pre-selected
list of potentially important genes, and thus is incapable of providing a
comprehensive view of the multigenic nature of metastasis.
The ability of DNA microarrays to simultaneously profile thousands of
human genes in an efficient, cost-effective way has dramatically
revolutionized metastasis research. Gene expression profiling has been
proven to be highly effective in the classification of morphologically
indistinguishable tumor subtypes. In addition, prognosis of
metastasis-free survival can be more accurately predicted by the
expression profiling of primary tumors than by traditional staging and
grading methods (4). However, it is often difficult to decipher biological
processes hidden behind these so-called "prognosis signatures"
that predict metastatic recurrence. The genetic heterogeneity of the human
patient population and the uncertain amount of stromal contamination in
the tumor samples may contribute to the difficulty of deriving a
functionally relevant metastasis signature from the gene expression
profiling of clinical specimens.
An alternative route of metastasis genomic research is to take
advantage of many mouse metastasis models that have been established in
the past half century. Pioneer work done by Fidler and others in the 1970s
suggested that metastasis formation is a selective process (5, 6). A small
subpopulation of highly metastatic cells exists in cancer cell lines
derived from either human patients or experimental animals. When cancer
cells are injected into suitable animal hosts (i.e., syngeneic animals for
animal cancer cells or nude mice for human cancer cells), metastatic
lesions can be formed by these subpopulations of metastatic cells.
Therefore, an animal model can be used effectively as an in vivo
cell sorter to select a subpopulation of cancer cells with enhanced
metastasis ability to a particular organ. Microarray profiling technology
can then be applied to discern important genetic differences between
highly metastatic variants and their closely related, but weakly
metastatic, cousins. Work done by Clark et al. (7) was the first
successful example of such an approach. Currently, this new method of
metastasis gene discovery has been applied to a wide variety of animal
models and several different tumor types. We will focus our discussion
below on how we used a functional genomic approach to understand the
molecular basis of bone metastasis in breast cancer.
Bone metastasis gene discovery by
functional genomics
The human breast cancer cell line
MDA-MB-231 was derived from the pleural effusion of a late stage, ER-,
breast cancer patient in 1974. It has been widely used to study osteolytic
bone metastasis of breast cancer in animal models. Based on microscopic
examination, it is apparent that this cell line contains a heterogeneous
population of individual cells. This cell line, obtained from the American
Type Culture Collection (ATCC), forms x-ray-detectable, bone metastases in
approximately 35% of nude mice within 10-15 weeks of intracardiac
injection. We were also aware of the existence of MDA-MB-231-variant cell
lines from other labs that are more metastatic to bone. Based on these
observations, we hypothesized that there is a subpopulation of cancer
cells in the MDA-MB-231 cell line that is capable of metastasizing to bone
much more efficiently than the majority of the cells in this lineage.
Therefore, we attempted to establish MDA-MB-231 sublines with a particular
metastasis propensity to bone by isolating tumor cells from osteolytic
bone metastases in nude mice.
Figure 1. Schematic representation
of the functional genomic analysis of bone metastasis. Experiments
are carried out in five steps. 1. Cancer cells are injected into
the blood circulation by intracardiac injection. 2. The
development of metastasis is monitored by non-invasive imaging
technologies, such as x-ray imaging (top) or bioluminescent
imaging (bottom). 3. Sublines of cancer cells are established from
tumor cells isolated from bone metastases. 4. The enhanced
metastasis potential of the sublines is tested through a second
round of injection into a new cohort of animals. 5. After a
sufficient number of variant cell lines are established and
tested, microarray profiling and statistical analyses are carried
out to identify candidate metastasis genes.
|
To our surprise, we were able to isolate
several highly metastatic sublines after just one round of in vivo
selection (Figure 1). These sublines metastasize to bone within 6 weeks in
all animals examined. We subjected these sublines to gene expression
profiling using Affymetrix U133A oligo arrays. Group comparison and gene
filtering were carried out using the Genespring software (Silicon
Genetics). A 102-gene expression signature, including 43 overexpressed
genes and 59 underexpressed genes, was found to be associated with the
bone metastasis phenotype. Eleven genes were overexpressed by more than
4-fold in highly metastatic cells and were selected for further functional
validation studies (8).
Validation of candidate metastasis genes
A striking feature of these candidate
bone metastasis genes is that they encode either secreted cytokines or
cell surface receptors. This finding is consistent with the century-old
notion that metastasis formation depends on the pathological interactions
between tumor cells and the stromal environment (1). Biological activities
of these genes suggest their possible roles in promoting metastasis. For
example, the chemokine receptor CXCR4 is a well-known chemotatic receptor
for leukocytes to migrate to their target sites during immune and
inflammatory reactions. CXCR4 has recently been established as a homing
receptor for directional migration of breast cancer cells to metastasis
target sites that express high levels of its ligand (SDF1), including
bone, lymph node, and lung. Identification of CXCR4 in our bone
metastasis gene set, in a completely unbiased fashion, validated the
applicability of our functional genomic approach. Putative angiogenesis
factors, such as CTGF and FGF5, were also found to be highly expressed in
metastatic cells. Several other genes, such as the matrix metalloprotease,
MMP1, and the osteoclast activators, Interleukin-11 (IL-11)
and Osteopontin (OPN), have also been implicated in cancer
metastasis, but have not been rigorously tested in in vivo
metastasis assays.
Figure 2. Functional validation of
candidate metastasis genes. Three approaches can be taken to
functionally validate candidate metastasis genes in mouse
metastasis models. I. Cancer cells that overexpress or
underexpress candidate metastasis genes can be selected from a
heterogeneous parental population and subjected to in vivo
metastasis assays. II. Retroviral vectors can be used to stably
express metastasis gene-targeting RNAi constructs in highly
metastatic cells. III. Retroviral vectors can be used to
ectopically express candidate metastasis genes in weakly
metastatic cells. These genetically modified cells will then be
subjected to in vivo metastasis assays.
|
We first used ectopic expression of
metastasis genes in a weakly metastatic parental cell line to test their
functional importance in promoting bone metastasis (Figure 2). No
significant enhancement of bone metastasis was observed in tumor cells
overexpressing either IL-11 or OPN alone. However, there was
a mild increase in bone metastasis when cells were engineered to
overexpress both genes together. Furthermore, although either CXCR4
or CTGF alone only slightly increased metastasis efficiency, their
overexpression in IL-11/OPN double transfectants led to a dramatic
increase in bone metastasis (8). Therefore, this series of experiments not
only validated the function of these genes in bone metastasis but also
provided direct evidence for the multigenic nature of metastasis.
Metastasis requires the concerted actions of multiple genes that belong to
different functional classes. Lacking any one of them may cripple the
entire metastasis cascade. The inherent inefficiency of metastasis and the
rarity of highly metastatic cells could therefore be explained by the
multigenic nature of metastasis since tumor cells need to acquire multiple
metastasis genes in order to become highly metastatic.
Our second approach took advantage of
the natural variation of the parental MDA-MB-231 cells. We cloned 46
sublines (Single Cell Progenies, SCPs) from 46 individual cells of the
parental cell line by dilution cloning. We then used Northern blot
analysis to detect the expression of five metastasis genes (IL11, OPN,
CXCR4, MMP1, and CTGF) in these sublines. Only one subline
(SCP2) expressed all five genes at high levels; three (SCP20, 25, and 46)
overexpressed four genes; and one (SCP28) expressed three genes (8). Most
of the individual cells from the parental cell line did not express any of
the five metastasis genes. The metastatic behavior of SCPs in mice matched
very well with predictions, based on the number of metastasis genes
expressed. Thus, cell lines that express 4 or 5 metastasis genes are
extremely efficient in bone metastasis, similar to the cell lines selected
from bone metastases in vivo. Cell lines that do not express any
metastasis genes did not form a bone lesion even after 100 days of
incubation in mice (8). This set of experiments not only confirmed the
functional importance of bone metastasis genes, but also provided evidence
for the selective nature of the metastasis process.
The
applications of RNA interference (RNAi) technology in mammalian genetics
provide an alternative, and perhaps more, efficient way of validating
metastasis genes. Because of the multigenic nature of metastasis, it is
often difficult to observe significant increases in metastasis efficiency
by ectopic expression of a single metastasis gene. Therefore,
overexpression experiments often lead to false-negative results. In
comparison, suppressing a key mediator of metastasis may crumble the
entire metastasis cascade and result in a significant attenuation of
metastasis phenotype. Therefore, RNAi-mediated gene knockdown in highly
metastatic cells may more readily generate significant results in
metastasis assays. However, caution must be taken when designing and
interpreting RNAi knockdown experiments because of the uncertain,
off-target effects of RNAi. Multiple RNAi-targeting constructs need to be
tested and ideally, a conditional RNAi or a "put-back"
experiment is required to firmly establish the functional importance of a
candidate metastasis gene. An RNAi approach was used by Yang et al.
(9) to validate twist as a metastasis gene that promotes tumor cell
intravasation through a process called epithelial-mesenchymal transition (EMT).
We recently used this technology to analyze the role of the TGFβ/Smad
signal transduction pathway in bone metastasis. TGFβ is a cytokine
that is abundantly stored in the bone matrix and can be released into the
stromal environment upon bone destruction by osteolytic metastasis. We
found that two of the bone metastasis genes, IL-11
and CTGF, were potently activated by TGFβ. Therefore, a
positive-feedback loop may exist between tumor cells and the bone matrix
to fuel the destruction of bone and promote the growth of cancer cells. We
used retroviral vectors to stably express Smad4 RNAi constructs in highly
metastatic cells. These cells were rendered irresponsive to TGFβtreatment
and became weaker in bone metastasis formation. Restoration of Smad4
expression by ectopic expression of an RNAi-insensitive, Smad4 construct
led to increased bone metastasis (Kang et al., unpublished data).
Therefore, the TGFβ pathway may play a pro-metastasis role during
tumor progression.
Clinical applications of the bone
metastasis signature
The ultimate goal of the functional
genomic analysis of metastasis in animal models is to provide therapeutic
targets for drug development and diagnostic markers for the detection of
emerging metastasis. Since most metastasis genes that we found encoded
secreted proteins and cell surface receptors, there was a reasonable
chance that the overexpression of these genes in metastatic tumor cells
could be detected in body fluids such as blood and urine. This possibility
has several important implications. First, if the detection method was
sensitive enough, a protein profile that is associated with a particular
kind of metastasis (e.g., bone or lung metastases) could be detected when
the metastasis was sufficiently small enough to allow for surgical
intervention or effective pharmacological suppression. In addition,
knowing the identities of the metastasis mediators would facilitate the
design of "personalized" therapeutics to treat each individual
cancer patient more effectively. We used ELISA assays to detect the serum
levels of the metastasis protein MMP1 in breast cancer patients with or
without bone metastasis. High levels of serum MMP1 was associated with
shorter bone metastasis-free survival (P < 0.001) (Kang et al.,
unpublished data). Importantly, only about 20% of breast cancer patients
with bone metastasis had high levels of serum MMP1, suggesting that there
are alternative metastasis genes that might have a functional overlap with
MMP1 in other patients. These results further underscore the importance of
knowing the individual genetic basis of metastasis in order to optimize
treatment.
We further tested the possibility of
detecting bone metastasis signatures in primary tumors. Although highly
metastatic cells may represent a small subpopulation of cells in the
primary tumor, microarray profiling may still be sensitive enough to
detect the overexpression of bone metastasis genes in primary tumors.
Indeed, hierarchical clustering of 63 primary human breast carcinomas by
bone metastasis signature allowed segregation of tumor samples into two
major groups: one with bone metastasis as the first metastatic site and
another group with other sites of primary metastases (10). However, the
segregation of these two tumor subgroups was not robust, perhaps
reflecting the rarity of metastatic cells in the primary tumor and the
unclear nature of the bone metastasis signature.
Future directions
Our work on combining gene expression
profiling with mouse models of metastasis to identify functionally
important bone metastasis genes has established a working model for future
studies. Although several metastasis genes were successfully identified
and validated using this approach, major hurdles remain before we can
obtain a comprehensive and coherent understanding of the biology of bone
metastasis. We used only one cell line from one cancer patient in this
study. Although the results were validated and applicable to a large
population of human cancer patients, the idiosyncratic, single cell line
may still limit the interpretation of experimental results in more general
terms. Additional cell lines or fresh tumor samples may be needed to
extend metastasis gene research. Furthermore, while gene expression
profiling represents a technological quantum leap in terms of information
density compared with traditional approaches, multidimensional analyses of
metastasis requires the integration of other global assay tools, such as
proteomics.
Identification and validation of
metastasis genes is the first step toward understanding the biology of
metastasis and providing rational therapeutic designs. Functional
mechanisms of metastasis proteins need to be studied in greater detail in
both in vitro assays and in vivo animal models. The
interplay between tumor cells and the stromal microenvironment will become
a focal point of future research efforts. Therapeutics will be designed to
target the functions of metastasis genes or the stromal components that
facilitate the function of these genes. More sensitive and systematic
metastasis protein detection assays, such as protein arrays or mass
spectrometric techniques, will also be required to facilitate the
detection of metastasis in cancer patients.
Yibin Kang is an Assistant
Professor in the Department of Molecular Biology at Princeton University
(Princeton, NJ). Xin Lu is a graduate student in the Kang lab.
Yibin Kang can be reached at 609-258-8834; ykang@molbio.princeton.edu
.
Reference
1. S.
Paget, Lancet 1, 571–3 (1889).
2. W. Birchmeier, K. M. Weidner, J.
Hulsken, J. Behrens, Semin Cancer Biol. 4, 231-9 (1993).
3. S. McDonnell, L. M. Matrisian, Cancer
Metastasis Rev. 9, 305-19 (1990).
4. L. J. van 't Veer et al., Nature
415, 530-6 (2002).
5. I. J. Fidler, D. M. Gersten, I. R.
Hart, Adv Cancer Res 28, 149-250 (1978).
6. I. R. Hart, Am J Pathol 97,
587-600 (1979).
7. E. A. Clark, T. R. Golub, E. S.
Lander, R. O. Hynes, Nature 406, 532-5 (2000).
8. Y. Kang et al., Cancer Cell
3, 537-49 (2003).
9. J. Yang et al., Cell 117,
927-39 (2004).
10. A. J. Minn et al., J Clin
Invest 115, 44-55 (2005).
|
