Vol.
8, 61–74, January 2002
Advances in Brief
Contrasting Expression of Thrombospondin-1 and Osteopontin
Correlates with Absence or Presence of Metastatic Phenotype
in an Isogenic Model of Spontaneous Human
Breast Cancer Metastasis
Virginia Urquidi, Derek Sloan, Kanji Kawai,
Dianne Agarwal, Anthony C. Woodman,
David Tarin, and Steve Goodison1
University of California, San Diego Cancer Center and Department of
Pathology, University of California, San Diego, La Jolla, California
92093 [V. U., D. S., K. K., D. A., D. T., S. G.], and Cranfield
BioMedical Centre, Institute of Bioscience and Technology,
Cranfield, University, Bedfordshire, MK43 0AL, United Kingdom
[A. C. W.]
Abstract
Knowledge of the molecular mechanisms involved in met-
astatic spread is needed to facilitate advances in prognostic
evaluation for individual patients and in the design of thera-
peutic interventions to inhibit the process. In an effort to es-
tablish a methodological framework for analysis of molecules
and mechanisms involved in this complex multistep process, we
have developed a well defined experimental system, in which
the role of candidate genes can be screened and tested.
By serial dilution cloning of the MDA-MB-435 breast
tumor cell line and screening by orthotopic implantation
into the mammary fat pad of athymic mice, we have derived
a pair of breast tumor cell lines (M-4A4 and NM-2C5) that
originate from the same breast tumor but have diametrically
opposite metastatic capabilities.
In 74% of inoculated athymic mice, clone M-4A4 metas-
tasized consistently to the lungs, mimicking a major dissemi-
nation route of human breast cancer. Conversely, although
equally tumorigenic, clone NM-2C5 did not metastasize to any
distal site. We have confirmed that the cell lines originate from
a single genetic source by spectral karyotyping and evaluated
the expression of a number of proteins previously implicated in
cellular transformation and metastasis. The ability of M-4A4 to
metastasize was not associated with increased angiogenesis, as
measured by immunohistochemical microvessel density analy-
sis. However, RNA and protein analyses revealed that two
secreted proteins were differentially expressed: osteopontin ex-
pression was increased ⬃30-fold in clone M-4A4 and throm-
bospondin-1 expression was increased ⬃15-fold in clone NM-
Received 7/16/01; revised 10/8/01; accepted 10/8/01.
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
advertisement in accordance with 18 U.S.C. Section 1734 solely to
indicate this fact.
1
To whom requests for reprints should be addressed, at UCSD Cancer
Center, 9500 Gilman Drive, La Jolla, CA, 92093-0912. Phone:
(858) 822-2083; Fax: (858) 822-2084; E-mail: sgoodison@ucsd.edu.
Clinical Cancer Research 61
2C5. These cell lines constitute a stable and accessible model
for the identification of genes involved in the multistep process
of breast tumor metastasis. Manipulation of candidate genes in
these cells will permit evaluation of their functional significance
in the geometric progression of breast cancer.
Introduction
Despite significant advances in the treatment of primary
cancer, the ability to predict the metastatic behavior of a pa-
tient’s cancer, as well as to detect and eradicate such recur-
rences, remains the greatest clinical challenge in oncology. To
make progress in this area it is essential to obtain more detailed
knowledge of the molecular mechanisms involved in metastasis
as a basis for novel approaches to the evaluation of individual
patient prognosis and to rational therapeutic design.
The formation of a secondary tumor colony in a distal site is
the culmination of a complicated series of sequential and highly
selective events (1– 4). To succeed in accomplishing lympho-
genous or hematogenous metastasis, tumor cells must have invaded
the local extracellular matrix and penetrated the vascular endothe-
lium to gain access to the circulation for transport away from the
primary site. In the next phase of the process, cells that survive the
physical stresses of the circulation and surveillance by body de-
fense mechanisms lodge into the capillary bed of a conducive tissue
or organ and exit through the vessel wall. Finally, to thrive and
form secondary deposits, the tumor cells must proliferate and
attract a new vascular supply and other supporting cells from the
host tissue. These consecutive events are dependent upon the co-
ordinated regulation of gene expression.
This multistep nature of metastasis poses difficulties in both
design and interpretation of experiments to unveil the mechanisms
causing the process. Studies on excised fixed human tissues are
complicated by the variance of genetic background between indi-
viduals and by the cellular heterogeneity of a complex tissue mass.
Breast cancer is also a collection of distinct diseases, and it can be
difficult to be certain of the histogenetic classification of the tumor
in advanced cases, thereby causing an inappropriate combination of
data from genetically distinct lesions. However, the major limita-
tion of such studies is the inability to identify those cells in a tumor
mass that are truly capable of metastasis. Even if derived from a
single cell, an advanced carcinoma is a mixture of genotypically
and phenotypically distinct cells, and only a tiny fraction of those
cells may possess the ability to disseminate from the primary
lesion. Furthermore, it is estimated that only 0.1% of tumor cells
that do enter the circulation will form secondary deposits in
a distal organ (5, 6). Analysis of actual metastases may not be
helpful either, because these cells have proliferated in a
non-breast tissue environment and therefore may display a
62 Breast Metastasis Model Characterization
markedly different molecular profile (7) and may not retain
the ability to metastasize again.
Critical to the experimental analysis of metastasis has been
the isolation of human tumor cell lines and the ability to study
their behavior in vivo by inoculation into immune-compromised
mice (8, 9). Several established human breast cancer cell lines
with varying documented abilities of invasiveness and/or mi-
gration in vitro are available, and some are capable of sponta-
neous metastasis in vivo (10), i.e., dissemination from growth in
the mammary gland and proliferation in a distal site. However,
most of these are polyclonal and composed of cell populations
that are heterogeneous in metastatic phenotype, making them
difficult to use as models in studies seeking to define genes
causing metastasis. Several laboratories have obtained cell lines
with increasing metastatic phenotype by recovering and cultur-
ing metastatic deposits derived from primary inoculations and
recycling the cells through several rounds of orthotopic selec-
tion (11, 12). The resulting cell lines represent improved models
for studying metastasis because they are mono- or oligoclonal
and, therefore, of uniform phenotype. The most common diffi-
culty with these models is the lack of a corresponding totally
nonmetastatic, clonally uniform counterpart for comparison,
because the selection process used for the derivation of meta-
static lines cannot be used for the selection of the converse
phenotype. The original cell line, from which the hypermeta-
static cell line is selected, is not an appropriate counterpart for
comparison because it is a heterogeneous polyclonal resource
containing many clones of differing metastatic propensity.
To overcome this experimental conundrum we have devel-
oped an orthotopic model of breast metastasis that enables
comparative molecular screening and functional evaluation of
candidate metastasis-related genes in an isogenic background. A
large panel of monoclonal tumor cell lines were derived by
limiting dilution from the polyclonal breast carcinoma cell line
MDA-MB-435 (13) and systematically tested for metastatic
behavior in athymic mice. Clone M-4A4 was selected from the
many metastatic clonal populations and is highly metastatic to
the lungs. A separate clone, NM-2C5, was found to be totally
nonmetastatic but equally tumorigenic in athymic mice.
In this report we describe characterization of this paired
cell line model, comparing their cellular and histopathological
tumor morphology, in vitro and in vivo growth rates, and mo-
lecular profiling to date. We have demonstrated that the cell
lines originate from the same genetic source by spectral karyo-
typing, and we report data on the expression of a number of gene
products previously implicated in transformation and metastasis
by RNA and immunochemical analyses. The majority of such
targeted comparative analyses revealed equal levels of gene
expression in both clonal cell lines, further supporting the tightly
controlled nature of the experimental system. However, two
major differences were revealed by multiple analyses. The ex-
pression of the secreted proteins TSP-12 and OPN was found to
correlate with the nonmetastatic and the metastatic phenotypes,
2
The abbreviations used are: TSP-1, thrombospondin-1; OPN, os-
teopontin; RT-PCR, reverse transcription-PCR; ER, estrogen receptor;
MMP, metalloproteinase; TIMP, tissue inhibitor of metalloproteinase;
respectively. Verification of the differential expression of these
genes in cultured cells and in xenograft tissues was achieved
using quantitative RNA and protein analyses.
The derivation of this pair of breast tumor cell lines pro-
vides a powerful experimental system for controlled evaluation
of molecular mechanisms involved in metastasis. The informa-
tion already obtained indicates that it will greatly facilitate the
identification of genetic and biochemical differences between
metastatic and nonmetastatic neoplastic breast epithelia. Manip-
ulation of candidate genes and subsequent analysis of upstream
and downstream effects will then permit evaluation of their
functional significance in this clinically important process.
Materials and Methods
Isolation of Clonal Cell Lines and Culture Conditions.
Clonal sublines of the metastatic, polyclonal MDA-MB-435
human breast carcinoma line (10, 13) were isolated by the
limiting dilution technique, with direct microscopic monitoring
of monocellular origin. Monoclonal cell lines were propagated
in RPMI 1640 supplemented with 10% newborn calf serum
(Invitrogen, Carlsbad, CA) at 37°C in a humidified atmosphere
of 5% CO2-95% air. Cells were harvested by washing the
monolayer with PBS and briefly incubating with 0.25% trypsin/
0.02% EDTA. The detached cells were washed by centrifuga-
tion and resuspended in RPMI 1640 media and counted before
passaging or inoculation. Two selected clones of nonmetastatic
and metastatic phenotypes (NM-2C5 and M-4A4, respectively)
were propagated for further investigation. Subsequent analyses
were performed on cultures passaged no more than 10 times
from frozen stock vials designated passage 1 at the time of in
vivo inoculation. Cultures were tested and declared free of
Mycoplasma and common murine pathogens.
Tumorigenicity and Metastasis Formation in Vivo.
Female athymic mice (MF1Nu strain) were housed in an isola-
tion suite for the duration of the experiments. The tumorigenic-
ity and spontaneous metastatic capability of the cell lines were
determined by injection into the mammary fat pad. One million
cells in 0.05 ml of a 1:1 mixture of RPMI 1640 medium and
ECM gel (Sigma Chemical Co., St. Louis, MO) were inoculated
into the anesthetized mouse. Animals were monitored every 2
days for up to 5 months for tumor growth and general health.
The rate of primary tumor growth of the clones was determined
by plotting the means of two orthogonal diameters of the tu-
mors, measured at 7-day intervals. Animals were sacrificed and
autopsied at 3– 6 months postinoculation, unless moribund ear-
lier. Metastasis formation was assessed by macroscopic obser-
vation of all major organs for secondary tumors and confirmed
by histological examination of organs and lymph nodes. Tissue
samples harvested for histological analysis were either fixed and
embedded in paraffin wax or snap-frozen in liquid nitrogen.
Histological Evaluation. Tissues recovered after autop-
sies of inoculated animals were fixed in 4% paraformaldehyde
for 24 h and stored in 70% ethanol. Paraffin embedding, sec-
tioning, and staining were performed using standard protocols.
ECL, enhanced chemiluminescence; MVD, microvessel density; FISH,
fluorescence in situ hybridization; CM, conditioned medium.

To confirm the presence or absence of internal lung and lymph
node metastases, sections were cut at 50-␮m intervals through-
out each of the tissue samples, nine sections per organ. These
sections were stained with H&E, and organs from 10 animals
that were inoculated with each cell line were evaluated for the
presence of microscopic metastasis by three independent pathol-
ogists.
Spectral Karyotyping. Metaphase spreads were pre-
pared according to standard procedures. Briefly, cells in log
phase culture were treated with colcemid (0.3 ␮g/ml final con-
centration; Sigma) for 2 h before harvest by trypsinization.
Centrifuged cells were exposed to hypotonic shock with 75 mM
KCl for 20 min at 37°C and fixed in 3:1 methanol/glacial acetic
acid solution. Cell suspensions were dropped onto slides, air-
dried, and aged at room temperature for several days. Probe
hybridization and image capture/analysis (14, 15) were per-
formed according to the manufacturer’s instructions (Applied
Spectral Imaging, Carlsbad, CA). Probes are created using pu-
rified, single-chromosome templates, and PCR amplified using
degenerate oligo-primers and incorporating three fluorochromes
and two haptens. The specific combination of these five labels
results in a unique spectral signature for each chromosome.
Indirect detection of haptens was performed using Cy5-conju-
gated avidin and Cy5.5-conjugated antimouse IgG (Rockland,
Gilbertsville, PA). Slides were counterstained with 4⬘,6-
diamidino-2-phenylindole and mounted with Prolong antifade
solution (Molecular Probes, Eugene, OR). Spectral imaging was
achieved using a SpectraCube system (Applied Spectral Imag-
ing) mounted on a Zeiss Axiophot microscope, viewed through
a ⫻63 oil-immersion plan/apo objective illuminated by a xenon
lamp (Opti Quip, Highland Mills, NY). Chromosome classifi-
cation was performed with SkyView software (Applied Spectral
Imaging). 4⬘,6-Diamidino-2-phenylindole banding was captured
separately and inverted for alignment with spectral representa-
tions using SkyView software.
RT-PCR. Cultured cells were grown to ⬃75% conflu-
ence before extraction of RNA. Frozen tissues recovered from
nude mice were sectioned on a cryostat, and 100 sections (5 ␮m)
were used for RNA extraction. Total RNA was isolated using an
RNeasy kit (Qiagen, Valencia, CA), and mRNA was purified
with Oligotex (Qiagen), treated with DNase I, and reverse
transcribed using Moloney murine leukemia virus reverse tran-
scriptase with a combination of oligodeoxythymidylic acid and
random decamers (Ambion, Austin, TX). The resulting cDNA
was used as a template for PCR using gene-specific primers.
Hot-start PCR conditions were performed, and cycling condi-
tions were adjusted for primer pair characteristics and estimated
transcript abundance. Amplification products were resolved by
agarose gel electrophoresis. For verification of specificity, prod-
ucts were recovered from gels and sequenced directly using the
amplification primers.
Quantitative PCR Analysis. OPN and TSP-1 mRNA
transcripts in cultured cells and in primary xenograft tumor
tissues recovered from athymic mice were quantified (16) using
the ABI Prism 7700 Sequence Detection System (PE Applied
Biosystems, Foster City, CA). Total RNA was isolated using the
RNeasy kit (Qiagen), and DNA was removed by digestion with
RNase-free DNase A (Ambion). cDNA was synthesized using
Moloney murine leukemia virus reverse transcriptase and an
Clinical Cancer Research 63
oligodeoxythymidylic acid primer (Ambion). PCR was per-
formed using the SYBR Green PCR Master Mix kit containing
SYBR green I dye, AmpliTaq Gold DNA Polymerase, de-
oxynucleotide triphosphates with dUTP, passive reference, and
optimized buffer components (PE Applied Biosystems). PCR
primers were designed against the 3⬘-UTR of the human target
genes using MacVector software (Oxford Molecular, Beaverton,
OR) and checked for the absence of potential binding to mouse
homologue sequences. All primers were used at a final concen-
tration of 100 nM and 1 ␮l of cDNA dilution was added in 25 ␮l
PCR reactions. No-template controls were included for each
target. Thermocycling was initiated with a 10 min, 95°C enzyme
activation step followed by 40 cycles of 95°C for 15 s and 60°C
for 1 min. All reactions were done in triplicate, and each
reaction was gel-verified to contain a single product of the
correct size. Data analysis was performed using the relative
standard curve method as outlined by the manufacturer (PE
Applied Biosystems) and as described previously (17, 18). The
mean glyceraldehyde-3-phosphate dehydrogenase concentration
(primer set supplied by PE Applied Biosystems) was determined
once for each cDNA sample and used to normalize expression
of all other genes tested in the same sample. The relative
difference in expression was recorded as the ratio of normalized
target concentrations for the same cDNA dilution. Primer se-
quences used were: OPN (GenBank accession no. AF052124),
forward primer 5⬘-TGAGAGCAATGAGCATTCCGATG, re-
verse primer 5⬘-CAGGGAGTTTCCATGAAGCCAC; TSP-1
(GenBank accession no. X14787), forward 5⬘-AACAAC-
CCCACACCCCAGTTTG, reverse 5⬘-TTGAAGCAGGCAT-
CAGTCAC.
Immunocytochemistry and Immunohistochemistry.
Cultured cells were grown on glass disks in the appropriate
medium before fixation in cold methanol (10 min, 4°C). Primary
xenograft tumor tissue was either frozen in liquid nitrogen or
fixed in formalin and embedded in paraffin. Cryostat sections (5
␮m) were air dried and subsequently fixed for 20 min in 4%
formaldehyde on Superfrost slides (Fisher Scientific, Pittsburgh,
PA). Paraffin-embedded tissues were dewaxed with xylenes,
followed by a 100%, 95%, 80% ethanol series and rehydrated in
PBS. Standard immunohistochemical procedures were per-
formed using hematoxylin counterstaining. Sections incubated
with secondary antibodies alone served as negative controls.
Antibodies. TSP-1 and focal adhesion kinase antibodies
were obtained from Transduction Laboratories (Lexington,
KY). OPN was from Chemicon (Temecula, CA). CD44 mAb
Hermes-3 (19) against an epitope encoded by CD44 exon 5 was
a kind gift from Dr. E. C. Butcher, Department of Pathology,
Stanford University (Stanford, CA). CD44 mAb 2F10 (R & D
Systems, Inc., Minneapolis, MN) against an epitope encoded by
CD44 exon 11 (CD44v6) was used. ER-␣ F-10, Ras, and p53
were from Santa Cruz Biotechnology (Santa Cruz, CA).
MMP-2, MMP-9, and TIMP-2 were from Oncogene Research
Products (Cambridge, MA). Maspin was obtained from Phar-
Mingen (Franklin Lakes, NJ). Paxillin was from (Sigma Chem-
ical Co.). The phosphatase and tonsin homologue antibody was
a generous gift from DNAX, Palo Alto, CA. The NG2 proteo-
glycan antibody was a generous gift from Dr. W. B. Stallcup,
The Burnham Institute, La Jolla, CA. The Nm23-H1 antibody
64 Breast Metastasis Model Characterization
was a generous gift from Dr. C. Chang, University of California,
San Diego School of Medicine, La Jolla, CA.
Southern Blot Detection of CD44 Transcripts. RNA
extraction and cDNA synthesis are described above. cDNA was
used as a template for CD44-specific PCR using primers which
anneal to CD44 exon 3 and exon 18 respectively (20), yielding
amplification products that include the variant exon sequences
present in CD44 transcripts. Thirty PCR cycles consisting of
94°C for 30 s, 55°C for 1 min, and 72°C for 2 min were
performed after a “hot start” procedure in 50-␮l reactions. PCR
products (5 ␮l) were separated by agarose gel electrophoresis
and transferred to a Hybond N⫹ nylon membrane (Amersham
Biosciences, Piscataway, NJ). The membrane was hybridized
with probes to the standard region (CD44s) and to exon 11 (v6)
of the variant region (20, 21). Exon-specific nucleic acid probes
were synthesized by PCR amplification of a CD44 genomic
clone (22). Probes were gel purified and labeled directly using
the ECL nucleic acid labeling kit (Amersham Biosciences).
Hybridization and detection were performed according to the
ECL manfacturer’s instructions.
Northern Blot Analysis. Total RNA was isolated as
described above and mRNA was purified with Oligotex resin
(Qiagen). RNA was treated with DNase I (Ambion). Northern
blotting was performed using 1.0% agarose/glyoxal gels and
blotting overnight onto NorthernStar nylon membrane (Am-
bion) in diethyl pyrocarbonate-treated 10⫻ SSC. Oligonucleo-
tide primers were designed using sequences obtained from the
National Center for Biotechnology Information GenBank data-
base and gene-specific probes created by PCR using cell line
cDNA as template. Products of expected size were gel purified
and sequenced: TSP-1 (accession no. X14787) forward primer,
5⬘-AACAACCCCACACCCCAGTTTG; reverse primer, 5⬘-TTG-
AAGCAGGCATCAGTCAC; OPN (accession no. AF052124)
forward primer, 5⬘-TGAGAGCAATGAGCATTCCGATG; re-
verse primer, CAGGGAGTTTCCATGAAGCCAC. Probe (25
ng) was labeled using the Decaprime kit (Ambion) and
[␣-32P]dCTP (Amersham Biosciences). Hybridization was per-
formed for 4 –12 h at 68°C in NorthernMax hybridization buffer
(Ambion). Images were obtained by use of a PhosphorImager
(Molecular Dynamics, Sunnyvale, CA), and quantitative values
were obtained using ImageQuant software (Molecular Dynamics).
Western Blot Analysis. Cultured cell lines were grown to
75% confluency, harvested, and lysed for protein extraction ac-
cording to established protocols. Protein concentration was deter-
mined using the Bio-Rad protein assay kit (Bio-Rad, Hercules, CA)
and adjusted to 1 mg/ml by dilution with PBS. For secreted-protein
analysis, serum-free culture medium was removed after 24 h incu-
bation and centrifuged (600 ⫻ g for 10 min) to remove any cellular
material; then, the supernatant was frozen and stored at ⫺70°C.
After 100 times concentration through Ultrafree Centrifugal filters
(Millipore, Bedford, MA), soluble protein was measured as de-
scribed above. Aliquots of standardized samples (5 ␮g of total
protein in 10 ␮l of buffer) and molecular weight standards were
boiled for 5 min before separation by 4 –12% SDS-PAGE under
reducing conditions. The resolved proteins were electroblotted on
to an Immobilon-P membrane (Millipore, Bedford, MA) using
transfer buffer (48 mM Tris, 39 mM glycine, 0.1% SDS, 20%
methanol, pH 9.2). Nonspecific reactions were blocked by incuba-
tion with Tris-buffered saline containing 5% skimmed milk for 1 h,
before the membrane was incubated overnight at 4°C with primary
antibody. Reactivity was determined by the addition of peroxidase-
conjugated secondary antibody at room temperature, and the sig-
nals were visualized using the ECL detection system (Amersham
Biosciences).
Microvessel Density. Three formalin-fixed, paraffin-
embedded specimens of primary tumors were randomly selected
from both NM-2C5 and M-4A4 resources, and several sections
were prepared from each for immunostaining. The CD31 antibody
(murine monoclonal antibody; DAKO, Carpinteria, CA) was used
to identify capillary-sized vessels. Standard avidin-biotin complex
peroxidase immunohistochemical staining was performed on
5-␮m-thick sections from paraffin-embedded, formalin-fixed
tissue. A negative control staining without primary antibody was
performed. The microvessel counting procedure as described by
Weidner (23) was used. The areas with the greatest density of
CD31-positive endothelial cells were designated “hot spots.”
The whole section was scanned at low power (⫻40) to identify
the best fields for counting. Counting was performed on three
separate fields within a hot spot at ⫻200 magnification using a
Nikon E600 microscope. Each stained endothelial cell or cell
cluster was counted as one microvessel, and the presence of a
lumen was not required (see Fig. 5). If two or more CD31-
positive foci appeared to belong to a single continuous vessel,
this was counted as one microvessel. Specimen section edges
and areas of necrosis or sclerosis were ignored. The MVD count
was defined as the sum of the three field counts within the hot
spot. Sections were analyzed by two independent observers
(D. S., S. G.) without knowing prior results or the identity of the
samples. Averages of counts were calculated, and the Wilcox
rank sum test was used to compare medians.
Results
Isolation of Clones and Metastasis Assays. Eighty in-
dependent cell lines were cloned by limiting dilution from the
metastatic polyclonal MDA-MB-435 human breast carcinoma
line. After propagation in culture, clones were tested for tumor-
igenicity and spontaneous metastasis in athymic nude mice by
orthotopic implantation. The goal was to identify clonal cell
lines that had distinct metastatic potential and/or dissemination
behavior. The polyclonal parent cell line MDA-MB-435 was
tumorigenic in all cases and spread to distant sites in 16 of 17
mice examined, with the lungs and the lymph nodes being the
major secondary target organs (Table 1). Clonal cell line M-4A4
was found to be more aggressively metastatic to the lungs than
the parent cell line and to spread only to the lungs and lymph
nodes. Rapidly formed primary tumors (palpable within 2–3
weeks) were evident in all mice examined. Fifteen of 19 mice
(79%) had metastases at either secondary site, with 74% of mice
having detectable lung metastases. Only one mouse had lymph
node metastases in the absence of lung metastases. Conversely,
clonal cell line NM-2C5 was shown to be completely nonmeta-
static from the orthotopic inoculation site. NM-2C5 primary
tumors proliferated at a slower rate than M-4A4 in vivo (median
153 and 98 days, respectively, to the end point of a single
orthogonal 2-cm diameter). However, on examination of 17
mice, no metastases were found in any organ, although the
primary tumors were in situ up to 44 days longer (Table 1). No
 Clinical Cancer Research 65

Table 1 Tumorigenicity and production of metastasis by human breast cancer cells orthotopically implanted in athymic mice
Median days Median size of No. of Animals with Metastasis
No. of postinoculation primary tumor (cm3) Median lung metastatic
Cell line Mice (Range) (Range) Lymph nodes Lungs deposits (Range)
MDA-MB-435 17 98 (63–130) 3.8 (0.56–7.28) 12/17 (71%) 9/17 (53%) 6 (3–20)
M-4A4 19 98 (56–112) 6 (2.46–8) 8/19 (42%) 14/19 (74%) 7 (1–50)
NM-2C5 17 153 (73–174) 4.28 (0.33–7.83) 0/17 0/17 0

Fig. 1 Cell morphology of

nonmetastatic (NM-2C5, left)
and metastatic (M-4A4, right)
cultured cells. Cells were
seeded at passage 5 and im-
aged, without fixing, using dif-
ferential interference micros-
copy (⫻20 magnification).

mice appeared to have any ill-effects from M-4A4 or NM-2C5

tumor growth within the time frame of the experiment. These
two carefully scrutinized clones of opposing metastatic pheno-
type were therefore selected for subsequent analysis of differ-
ential gene expression. These particular clones appear to be
remarkably stable with respect to the metastatic phenotype.
Cells are cultured in vitro anywhere up to passage 10 before
inoculation into the athymic mice, but due to the length of
incubation of the orthotopically implanted primary tumor before
sacrifice (3–5 months; Table 1), the cells must go through many
tens of passages in vivo. However, this extended passage in vivo
does not result in phenotypic drift with regard to metastatic
potential, suggesting a high level of genomic stability.
Cell Morphology and Growth in Culture. A morpho-
logical comparison of the two cell lines of opposite metastatic
potential is shown in Fig. 1. During log phase growth the
NM-2C5 cells grew in tightly clumped aggregates with individ-
ual cells characterized by a spindle-shaped morphology with
cytoplasmic protrusions. The M-4A4 cells displayed a more
flattened morphology with greater cytoplasmic volume and pro-
liferated in a pavement pattern, with less aggregation. As cul-
tures approached confluency, NM-2C5 cells remained attached Fig. 2 Growth rates of cultured NM-2C5 and M-4A4 clonal breast
to the substrate and generated numerous focal aggregates among carcinoma cell lines. Cell numbers were determined by counting live
a disorganized multilayer of cells. Conversely, M-4A4 cells cells harvested from separate culture flasks on a hemocytometer at 24-h
detached from the substrate and divided in suspension above a intervals.
monolayer of confluent cells (data not shown). The log phase
doubling time for both M-4A4 and NM-2C5 cells in vitro was
⬃72 h (Fig. 2). cells routinely form primary tumors in this site; these are ade-
Histological Evaluation of Xenograft Tissue. Fig. 3 nocarcinomas, showing classical features of infiltrating lobular
shows representative histological sections of tumors and metas- carcinomas of the breast including Indian-file arrangement of
tases resulting from inoculation of breast cancer cells into the tumor cells. In primary tumors resulting from M-4A4 inocula-
mammary fat pad of athymic mice. Both M-4A4 and NM-2C5 tion, tumor cells could be seen within blood vessels, but no such
66 Breast Metastasis Model Characterization
blood-vascular invasion was observed in NM-2C5 tumors. Me-
tastases and tumor cell emboli were histologically within both
the lungs and the overlying pleura, and in the cortical sinus of
draining lymph nodes of M-4A4 tumor-bearing mice (Fig. 3).
Such secondary tumor colonies were never detected in mice
hosting mammary tumors formed by NM-2C5 cells.
Spectral Karyotyping. When single-cell subclones are
derived from a polyclonal tumor cell culture, it is usually as-
sumed that they arose from the same original tumor source.
However, it is possible that distinct clones could be derived
from separate tumors present in the individual, which arose via
alternative transformation pathways. It is also possible that a
clonal cell line could be derived from a preexisting contaminant
in the polyclonal source, such as another tumor cell line being
studied in the same laboratory. For subsequent comparative
analysis to be valid it was essential to prove that this metastasis
model was derived from a common genetic source. To confirm
this, a molecular cytogenetic analysis was performed on the
NM-2C5 and M-4A4 clonal cell lines; the comparison of marker
chromosomes present in both lines can distinguish whether a
pair of cells are from the same progenitor. Spectral imaging (14)
is a multicolor FISH procedure capable of classifying all chro-
mosomes in one hybridization experiment (Fig. 4A). Spectral
classification revealed that both cell lines are hyperdiploid
(modal number of 56 –57) and contain multiple chromosomal
aberrations. A number of signature markers were common to
both lines (Fig. 4B). The derivative chromosomes t(1;7), t(1;10),
t(8;15), t(8;11), and t(20;21) were present in 100% of the 20
metaphase spreads analyzed for each cell line. The composition
of these marker chromosomes was confirmed by two-color
FISH (data not shown). The presence of these identical markers
confirms that the two cell lines are derived from a single
Fig. 3 Representative histologi-
cal sections of M-4A4 (upper
left) and NM-2C5 tumors (upper
right) resulting from inoculation
into the mammary fat pad of
athymic mice. The tumors are
poorly differentiated adenocarci-
nomas showing classical features
of an infiltrating lobular carci-
noma of the breast including
Indian-file arrangement of tumor
cells (arrows, upper right; mag-
nification ⫻200). In the upper
left panel, tumor cells are seen
enmeshed within a fibrin embo-
lus in the lumen of a peripheral
blood vessel (arrow), and cross-
sectional profiles of normal
mouse breast ducts are seen in
the adjacent fat pad (ⴱ; magnifi-
cation ⫻200). An example of
pulmonary metastasis (lower
right; magnification ⫻100) and a
tumor cell embolus (lower left;
arrow) are shown in the cortical
sinus of a draining lymph node in
an M-4A4 tumor-bearing mouse.
Magnification ⫻400.
progenitor; however, the spectral karyotype also revealed at
least one chromosomal aberration specific for each cell line. For
example, NM-2C5 contained a derivative chromosome t(12;15)
in 100% of cells examined, a marker chromosome that was
never present in M-4A4. More detailed cytogenetic studies are
in progress to identify genes that may be perturbed by the
cell-specific chromosomal aberrations and to evaluate whether
they play a role in the metastatic phenotype.
Angiogenesis. Angiogenesis is a prerequisite for ad-
vanced tumor growth and is logically believed to be an impor-
tant factor in tumor metastasis. Although the complete process
is very complex, several studies have suggested that a MVD
index of tumors can be of prognostic value. Studies of several
tumor types, including ones of the breast (23, 24), have found
the MVD index to predict independently the clinical outcome in
human subjects. The immunohistochemical detection of CD31,
a member of the immunoglobulin superfamily present on endo-
thelial cells, platelets, and polymorphonuclear leukocytes, was
used to reveal microvessels in three orthotopic xenografts of
clones M-4A4 and NM-2C5 (Fig. 5), and MVD was examined
for association with the metastatic phenotype. MVD counts in
the NM-2C5 series ranged from 17 to 50 per field, with a
Wilcoxon rank sum test value of 92 for three independent fields.
The M-4A4 series counts ranged from 7 to 59 per field with a
three-field rank sum test value of 79. No statistically significant
association with the metastatic phenotype was observed (P 0.1).
Targeted Molecular Analyses. Having evaluated some
characteristics of the two cell lines and their associated tumors,
we then investigated and compared the expression of a number
of genes previously implicated in tumorigenesis and/or metas-
tasis. Expression analyses were performed using immunohisto-
chemistry, Western blotting, Northern blotting, and/or PCR
 Clinical Cancer Research 67

Fig. 4 Spectral karyotyping of the NM-2C5 and M-4A4 cell lines. A, a representative metaphase cell (actually two overlapping cells) of NM-2C5
depicted as a 24-color spectral image of the spectral karyotyping hybridization. Arrows indicate three of the multiple marker chromosomes. B,
representative metaphase cell chromosome complement after spectral karyotyping classification. Numbers alongside the derivative chromosomes
indicate the origin of the translocated material. Arrows indicate marker chromosomes: yellow and green arrows indicate derivative chromosomes
common to both cell lines, and the white arrow indicates a derivative chromosome observed only in nonmetastatic NM-2C5 cells.

Fig. 5 Representative immunohistochemical staining of NM-2C5 and M-4A4 breast cells growing in the mammary fat pad of athymic mice. Tissue
sections were stained with hematoxylin and CD31 immunostaining to reveal endothelium. Left, M-4A4; right, NM-2C5. Original magnification ⫻40.

(Table 2). Targets were selected based on a reported role in
transformation, angiogenesis, extracellular matrix degradation,
or metastasis. A majority of genes evaluated were found to be
either absent or present in both cell lines, and when present, they
were expressed at similar levels. Table 2 lists the genes evaluated
and indicates their presence or absence. Both breast cell lines were
negative for ER-␣ isoform expression. However, the ER-␤ form
did appear to be present at very low levels, detected only by PCR
amplification. ER-␣-positive MCF-7 and ER-␤-positive Du-145
cells were used as positive controls (data not shown).
Expression of the CD44 Gene. The CD44 gene contains
21 exons that can be spliced alternatively (21, 25) to generate a
number of variant protein isoforms. These can be modified
further by post-translational glycosylation to produce a large
family of transmembrane glycoproteins with diverse functions.
Numerous studies have provided evidence that the expression of
68 Breast Metastasis Model Characterization

Table 2 Comparative gene expression analyses of NM-2C5 and v6 antibody detected at least two isoforms in the xenograft
M-4A4 human breast cellsa tissue. The overall CD44 expression and the presence of the
Gene Presence (P) or Analytical Ratio v6-containing isoforms were similar in both M-4A4 and NM-
transcript/protein absence (A) method NM-2C5:M-4A4 2C5. CD44 transcripts were evaluated by exon-specific South-
ER-␣ A/A I, W, R ern blot analysis subsequent to a RT-PCR amplification de-
ER-␤ P/P R equal signed to include all alternatively spliced variants. The nucleic
Progesterone receptor A/A I
acid analyses closely matched the protein profile, with several
EGFRb P/P I equal
Ki67 P/P I equal major bands represented and a less abundant diffuse back-
Cytokeratin A/A I ground. In RT-PCR analyses the range of detectable CD44
EBV A/A I transcripts originates from alternative splicing and possibly the
BCL-2 P/P I equal
detection of unspliced RNA species (22), the latter being un-
CEA A/A I
Maspin P/P W equal translated and detected only because of the exponential ampli-
Paxillin P/P W equal fication of PCR. The range and relative intensity of detected
PTEN P/P W equal CD44s and v6-containing transcripts were comparable in both
NG2 P/P W equal
cultured clonal cell lines and in both M-4A4 and NM-2C5
TSP-1 P/P I, W, R 15:1 recovered xenograft tissues. The observed increased complexity
CD44 P/P I, W, R equal
Nm23-H1 P/P W equal
of CD44 expression in the xenograft tissue is likely caused by
KiSS-1 P/P R equal the heterogeneity of the source material; the xenograft tissue
KAI-I A/A R will contain host stroma, vasculature, and circulatory cells.
E-cadherin A/A R CD44 is ubiquitously expressed in human and mouse tissues,
RhoC P/P R equal
BRMS-1 P/P R equal
and CD44 sequences are highly conserved between the two
OPN P/P I, W, R 1:30 species, so that both the antibodies and the nucleic acid hybrid-
c-myc P/P R&W equal ization probes may cross-react with mouse CD44 molecules.
Ras P/P I equal Extracellular soluble forms of CD44 proteins do occur and have
P53 P/P I equal been evaluated as potential circulating, tumor-associated marker
MMP2 P/P W equal in breast cancer patients (29, 30). Analysis of CM obtained from
MMP9 P/P W equal NM-2C5 and M-4A4 cell line cultures revealed a single soluble
TIMP2 P/P W equal CD44 isoform of equal size and intensity in both samples (data
TIMP-1 P/P W equal
a
not shown).
Cultured tumor cells were analyzed for transcript and/or protein Differential Expression of TSP-1 and OPN. Our first
expression by Western blotting (W), Immunohistochemistry (I), or RNA
analyses (R) using Northern blot and/or PCR amplification. evidence of increased TSP-1 expression in NM-2C5, relative
b
EGFR, epidermal growth factor receptor; CEA, carcinoembry- to its metastatic counterpart M-4A4, came from comparative
onic antigen. total protein analysis of the CM recovered from the cultured
cell lines. Analysis of secreted proteins revealed the presence
of a 150-kDa polypeptide at high concentration in NM-2C5
the CD44 gene is specifically altered in many types of tumors. supernatants, which was virtually absent in supernatants of
Molecular analyses have revealed that there is overproduction of M-4A4 cells grown under identical conditions (Fig. 7A).
a wide array of unusual CD44 mRNA species in a variety of Tryptic digestion and mass spectrometry of the protein ex-
tumor tissues relative to their normal counterparts (26). It has cised from the gel identified the differentially expressed
also been reported that the expression of a particular isoform protein as TSP-1. This identity was verified with specific
that contains the alternatively spliced exon 11 (also termed v6) TSP-1 antibodies using Western analysis (Fig. 7A). The dif-
suffices to confer metastatic behavior upon nonmetastatic rat ferential expression of the TSP-1 gene between the two cell
cells (27, 28). Expression of the CD44 family was evaluated by lines was also confirmed at the transcriptional level by meas-
RNA and protein analyses in the clonal cell lines and in primary uring the relative abundance of steady-state TSP-1 mRNA on
tumor xenografts recovered from athymic mice. Multiple pro- Northern blots (Fig. 7B) and revealed a differential expres-
tein isoforms were detected in the NM-2C5 and M-4A4 cultured sion of approximately 15-fold (Table 2).
cell extracts and in the complex mouse/human xenograft tissue OPN, a secreted calcium-binding phosphoprotein, was tar-
using an antibody (Hermes 3) that recognizes an epitope en- geted specifically for differential expression analysis. Although
coded by standard exon 5, a peptide that is present in all poorly understood functionally, OPN has recently been linked to
expressed isoforms (Fig. 6). An antibody specific for the peptide tumorigenesis and metastasis in experimental animal models
encoded by exon 11 (v6) revealed only one major band in the and human patient studies (31–33). Initial Northern blot analysis
cultured cell lysates. The observed spread of several major of OPN transcript expression in our model revealed approxi-
bands over a diffuse background signal is a consequence of the mately 30-fold more OPN in M-4A4 cells relative to NM-2C5
presence of both alternatively spliced and differentially glyco- (Fig. 7B). The differential expression of secreted OPN protein
sylated isoforms. In the recovered xenograft tissue the Hermes was verified by immunoblotting of CM (Fig. 7A).
3 antibody revealed a similar pattern of protein expression, Having determined the differential expression of TSP-1
including the predominant standard isoform (molecular weight and OPN in M-4A4 and NM-2C5 cultured cells, we examined
of Mr ⬃110,000) and a range of isoforms of higher mass. The the expression of TSP-1 and OPN transcripts in primary xe-

Fig. 6 Comparative CD44 expression analyses.
CD44 protein (left, upper and lower) expression
was evaluated by Western blot analysis. Lysates
(5 ␮g of protein/well) from cultured cells (upper)
and from recovered primary xenograft tissue
(lower) of NM-2C5 (2) and M-4A4 (4) were
incubated with antibodies to epitopes encoded by
CD44 variant exon 11 (Exon 11) and exon 3
(Standard). Molecular mass (kDa) of prestained
protein markers is indicated to the left of each
panel. CD44 transcripts were evaluated by
Southern hybridization of RT-PCR products ob-
tained from amplification of NM-2C5 (2) and
M-4A4 (4) mRNA, recovered from cultured cells
(upper right) and primary xenograft tissue (lower
right). The blot was hybridized with probes spe-
cifically annealing to exon 11 (Exon 11) and
exon 5 (Standard) of the CD44 gene. Molecular
mass (bp) of nucleic acid markers is indicated to
the right of each panel.
nograft tissue recovered from nude mice. Real-time quantitative
PCR analysis of three primary tumors originating from each
clone revealed that OPN and TSP-1 transcripts were also dif-
ferentially expressed in vivo. Real-time quantitative PCR results
(normalized against average values for the housekeeping gene
GAPDH) revealed that TSP-1 mRNA was 22-fold higher in
NM-2C5 tissue relative to M-4A4, and OPN mRNA was 21-fold
higher in M-4A4 tissue relative to NM-2C5. Real-time PCR
evaluation of TSP-1 and OPN expression in cultured cells
revealed 33-fold and 88-fold differential expression, respec-
tively (lower fold differences revealed using Northern and
Western analyses were presumably underestimated by solid-
phase filter analyses). Immunochemical analysis of xenograft
tissue revealed that both OPN and TSP-1 were expressed, pri-
marily accumulating in the extracellular matrix, but truly quan-
titative protein analysis was unattainable because both are se-
creted proteins. We used the cytogenetic evaluation of the cell
lines (Fig. 4) to estimate gene dosage and to assess the gross
integrity of the chromosomal structure at the OPN and TSP-1
loci. OPN is a single-copy gene located on chromosome 4, and
both NM-2C5 and M-4A4 cell lines were found to contain two
copies of this chromosome of normal constitution. TSP-1 is also
a single-copy gene, located on chromosome 15 at position
15q15. Chromosome 15 does show some rearrangement in our
cell lines (Fig. 4), but this does not appear to affect the TSP-1
gene. Two chromosomes 15⬘ with structural integrity at 15q15
were always present in both cell lines. Furthermore, the chro-
mosome 15 translocation prevalent in NM-2C5 cells has a
breakpoint mapped as t(12;15)(q22;q26.1), and the t(8;15) trans-
location common to both cell lines is a t(8;15)(q24;q21), as
determined by G-banding and FISH (data not shown).
Clinical Cancer Research 69
Discussion
The experimental approach adopted in this work is based
on the knowledge that advanced tumors are highly heterogene-
ous (34, 35) and that clones that are capable of completing the
multistep process of metastasis from the primary lesion may be
only a tiny fraction of the total tumor cell population. The
evolution of a neoplastic cell to an invasive and then a meta-
static cell depends upon a transient period of genomic instability
and/or the ability to rapidly adjust to changing local microen-
vironments during its dissemination, with appropriate adaptive
shifts in gene expression patterns. Earlier studies on the conti-
nuity of metastatic capabilities of clonal tumor cell populations
cultured over several generations in vitro (34, 36) repeatedly
showed that this pattern of behavior is retained and heritably
transferred within a cell population, but that it can drift in
degree, or even be extinguished, over time. It is, therefore,
probably programmed by genetic mechanisms that remain stable
over sufficient periods of time to make them amenable to
analysis but are not necessarily constant or predictable. In this
work we have taken special precautions to avoid the tendency
for clonal drift in metastatic capability over many cell genera-
tions described by Poste, Nicholson, and others (36 –38). This
trend was restricted in our experimental system by limiting
clonal expansion to less than 10 passages in vitro for all cell line
analyses. However, because of the orthotopic inoculation used
in our model, the cell lines experienced extended passage in vivo
of many tens of passages, yet this did not result in phenotypic
drift with regard to metastatic potential, suggesting a high level
of genomic stability.
Because of the method of derivation of the majority of cell
line models capable of spontaneous metastasis, no completely
70 Breast Metastasis Model Characterization
nonmetastatic counterpart has been isolated for direct compar-
ative evaluation. In contrast, our goal was to isolate a matched
pair of clonal cell lines with diametrically opposite metastatic
capabilities, derived from the same tumor source and, therefore,
originating from a common genetic progenitor. It was reasoned
that this experimental design would help to narrow the search
for differentially expressed genes relevant to metastasis. An-
other feature of our approach is the use of the immune-compro-
mised mouse to inform us of the metastatic behavior of in-
oculated human breast cells. The inoculum was already
monoclonal; therefore, no recovery or reculturing of cells was
necessary or desired. The human polyclonal parent MDA-MB-
435 cell line, the source for the derivation of this metastasis
model, was known to be metastatic; thus, the probability of
isolating a metastatic clone was high. However, the isolation of
a nonmetastatic clone was much less certain because of the
difficulty of confirming a negative result. Hence, the method
used to derive a cell line metastasis model such as the M-4A4
and NM-2C5 pair requires a major investment of effort. How-
ever, this effort can yield a valuable investigative resource with
regard to biological relevance, because the cell lines of opposing
metastatic phenotype have emerged through natural selection
within the original human tumor and have not been genetically
manipulated by the investigator or selected by passage through
a murine host.
Our approach to the investigation of the molecular basis of
metastasis uses these paired cell lines as a starting point, focus-
Fig. 7 Differential expression of
TSP-1 and OPN in M-4A4 and NM-
2C5 cells. A, protein levels in serum-
free CM from cultured cells. a: the
higher expression level of TSP-1 in
NM-2C5 was clearly noticeable by
Coomassie Blue staining of an SDS-
PAGE gel. Note that TSP-1 expres-
sion in M-4A4 decreases as cell den-
sity increases. b: confirmation of the
identity of TSP-1 and of its differen-
tial expression was obtained by West-
ern blot using anti-TSP-1 antibodies.
c: Western blot with OPN specific
antibody, showing a substantially
higher expression in M-4A4. B,
Northern analysis reveals that the
level of TSP-1 mRNA is higher in
NM-2C5 (a), and the level of OPN
mRNA is higher in M-4A4 (b).
ing directly on the differences between defined metastatic and
nonmetastatic monoclonal cell populations. We considered
commencing with analysis of recovered human primary and
metastatic tumor xenograft tissues resulting from inoculation of
these lines but decided that this could lead to interpretive diffi-
culties, because of the obligatory mix of human tumor and
murine host cells. We reasoned that an initial differential obser-
vation in the pure cell populations followed by human-specific
verification of any promising candidates in the primary xe-
nograft material constitutes a rational approach. Initially, we
targeted specific molecules for study in this model, selected on
the basis of reported roles in transformation, angiogenesis, ex-
tracellular matrix degradation, or metastasis. The oncogene
products c-myc and Ras were equally present in both trans-
formed cell lines. The levels of these proteins were far greater
than those observed in nontransformed cells, indicative of over-
expression and/or alteration of protein half-life, most often
caused by genetic mutation of these genes (39, 40). The cell
adhesion molecule E-cadherin and the cytokeratins are specifi-
cally expressed in epithelial tissue, and in tumor cells, the loss
of E-cadherin expression parallels tumor progression toward a
malignant invasive state and is correlated with a loss of the
epithelial phenotype (41). These molecules were absent from
both of the clonal breast cell lines, suggesting some degree of
loss of differentiation. Their absence may also be linked to the
negative ER status of both cell lines, inasmuch as there are
reports that the expression of these genes in breast epithelia is

regulated by estrogens (42). In any case, the data indicate that
these classes of molecules are not actively involved in generat-
ing the differences between the metastatic phenotypes of the two
clones.
Angiogenesis, the process by which new vasculature is
formed from preexisting vessels, is an essential component to
primary tumor growth and distant metastasis. The degree of
tumor vascularization is a result of pro- and antiangiogenic
factors originating from both the tumor cells and those of the
host (43). Analysis of the recovered xenografts did not reveal
any significant difference in MVD between the M-4A4 and
NM-2C5 cell line-derived tumors, suggesting that the ability of
M-4A4 to metastasize is not associated with increased angio-
genesis. However, further work is required to evaluate the
constitution of the neovasculature as recent reports have indi-
cated that it is not only microvessel density but the size and the
composition of the vasculature that may influence the passage of
cells into the circulation (44).
The invasion of cancer cells into the surrounding host
stroma is also a crucial step in the process of metastasis (45).
The MMP family is a major component of extracellular matrix
degradation and remodeling and could, therefore, have a role in
enhancing the dissemination of cells from a primary tumor (46).
To date, we have examined the expression of the collagenases
MMP-2 and MMP-9 and their inhibitors TIMP-1 and -2. Anal-
ysis of these secreted proteins revealed no differences in protein
levels in the CM recovered from the cells in culture (Table 2).
The enzymatic activity of the MMP-2 and MMP-9 secreted
proteins was also equal, as tested by gelatin zymography (data
not shown). We are currently examining the expression of other
members of the MMP family in this experimental system.
Changes in the expression profile of CD44 variant isoforms
have been observed in many carcinomas, including those of the
breast (26, 47), and much interest and investigative effort has
centered on the possible involvement of CD44 isoforms in
tumor metastasis (48). Gunthert et al. (27) demonstrated that an
antibody that inhibited metastasis of a rat pancreatic adenocar-
cinoma cell line specifically recognized an epitope encoded by
CD44 exon v6 (exon 11) of this gene. Transfection of a non-
metastatic cell clone from the same rat pancreatic cancer, with
a construct containing this exon, induced it to become metastatic
(49, 50). Furthermore, two other groups provided evidence
indicating that, in experiments with cultured human melanoma
(51) and lymphoma cell lines (52), there appeared to be a
correlation between raised CD44 expression and metastatic ca-
pability. However, there is also evidence that CD44 function is
not always necessary for metastatic capability. A murine lym-
phoma cell line with a CD44 gene double knockout was re-
ported to retain its local invasiveness and metastatic potential
despite losing all hyaluronan-binding ability (53). Analysis of
our human breast metastasis model revealed multiple isoforms,
including the metastasis-implicated CD44v6 isoforms, but at
equivalent levels in both clones. In this breast metastasis model,
specific CD44 isoform expression does not appear to be suffi-
cient to confer metastatic capability.
The two genes that have been identified thus far in our
current work as being significantly differentially expressed in
our experimental model have each been implicated previously in
tumorigenesis and metastasis. TSP-1 is a homotrimeric multido-
Clinical Cancer Research 71
main glycoprotein with disulfide-linked subunits of molecular
weight Mr 150,000 and is encoded by a gene mapped to human
chromosome 15. The protein is synthesized and secreted by
numerous cell types including platelets, endothelial cells,
macrophages, fibroblasts, vascular smooth muscle cells, type II
pneumocytes, and keratinocytes (reviewed in Refs. 54 and 55).
Because of its interaction with a wide variety of proteins, TSP-1
has been implicated in a number of biological processes includ-
ing coagulation, cell adhesion, modulation of cell-cell and cell-
matrix interactions, control of tumor growth and metastasis, and
angiogenesis (55). The effect of TSP-1 expression on tumor
growth and metastasis has been examined in vivo by orthotopic
inoculation of TSP-1-overexpressing MDA-MB-435 cells. A
dose-dependent inhibition of spontaneous pulmonary metastases
was reported, but this was paralleled by an inhibition of primary
tumor growth (56). TSP-1 has also been implicated in the
phenomenon of concomitant tumor resistance. This refers to the
ability of some large primary tumors to hold smaller, distant
tumors in check, preventing their progressive growth. Volpert et
al. (57) demonstrated this phenomenon using tumors formed by
the human fibrosarcoma line HT1080 in nude mice and obtained
data indicating that it was caused by secretion of TSP-1 by the
primary tumor. However, some of the properties attributed to
TSP-1 are conflicting: TSP-1 can be both adhesive and antiad-
hesive in vitro, can stimulate or inhibit angiogenesis, and can
inhibit or enhance proteolytic enzyme activity. The composition
of the matrix, the availability of cytokines and proteases, and the
expression of various receptors in a given cellular environment
might explain these opposing functions. Our findings are in
agreement with TSP-1 playing a role in the reduction of meta-
static potential, but it does not appear to be through inhibition of
angiogenesis per se, as judged by microvessel density evalua-
tion, in the NM-2C5 xenografts.
A wide variety of transformed cells in culture express
much higher levels of OPN than do their normal counterparts
(58, 59), and a series of functional studies links OPN causally to
tumor progression. Transfection of antisense OPN constructs or
OPN-targeting ribozymes into transformed tumorigenic cells
can result in impaired anchorage-independent growth, as well as
diminished tumor-forming ability in vivo (31, 32, 60). Further-
more, transfection of OPN expression constructs into cells pro-
ducing noninvasive mammary epithelial tumors has been re-
ported to convert them to a malignant phenotype (33, 58). OPN
is also typically overexpressed in human tumors (61, 62), in-
cluding breast cancer (63); and among cell lines derived from
human breast tumors, the highest levels of expression are found
in the most metastatic (64, 65). Consistent with overexpression
of OPN in tumors, the concentration of OPN in the blood of
patients with metastatic cancer is substantially elevated in com-
parison with the low levels that normally circulate. Higher OPN
levels in patients with metastatic breast cancer may be associ-
ated with an increased number of involved sites and decreased
survival (66). There may also be a link between OPN function
and CD44 expression. OPN binds to naturally expressed and
stably transfected variant CD44 isoforms in a manner that is
specific, dose-dependent, and inhibitable by anti-CD44 antibod-
ies. The CD44-OPN interaction can mediate chemotaxis or
attachment, depending on presentation of OPN in soluble or
immobilized form (67). We have shown that multiple CD44
72 Breast Metastasis Model Characterization
isoforms exist in both the NM-2C5 and M-4A4 cell lines, and
although CD44 does not itself appear to be sufficient for me-
tastasis, it may be necessary for OPN function.
Relatively few studies have produced functional data dem-
onstrating the role of candidate genes in the regulation of human
breast tumor cell metastasis, but some of the most interesting
candidates have been identified using the MDA-MB-435 breast
tumor cell line (68 –70). Accordingly, we tested the expression
of KAI1 and KISS-1, molecules that have both been reported to
suppress metastasis in MDA-MB-435 cells (68, 71) and in other
experimental systems (70, 72). By RNA analysis, KAI1 expres-
sion was not detected in either the NM-2C5 or M-4A4 clonal
cell lines. Conversely, KISS-1 mRNA was found to be ex-
pressed in both cell lines at equal and relatively high levels. The
metastasis suppressor gene Nm-23-H1 (73) has also been found
to quantitatively inhibit MDA-MB-435 metastasis (74), but the
Nm23 protein product was found to be equally expressed in both
NM-2C5 and M-4A4 cell lines, as evidenced by Western blot
analysis.
The most recent gene reported to be implicated function-
ally in MDA-MB-435 metastasis is a breast metastasis suppres-
sor gene, termed BRMS-1. Based upon cytogenetic data that
identified chromosome 11 as containing multiple genetic aber-
rations associated with breast cancer progression, Phillips et al.
(75) showed that introduction of a normal chromosome 11 into
MDA-MB-435 cells significantly reduced the metastatic poten-
tial of the cell line without affecting tumorigenicity. Subsequent
analyses identified the location of one of the genes responsible
for this effect (69), and it was named breast metastasis suppres-
sor-1 (BRMS-1). Stable transfection of BRMS-1 cDNA into
MDA-MB-435 and MDA-MB-231 cell lines correlated with
locally invasive tumor growth in athymic mice and with signif-
icant reduction of metastases to the lungs and lymph nodes (69).
However, quantitative RNA evaluation of BRMS-1 expression
in NM-2C5 and M-4A4 cells revealed equal levels of expression
in both cell lines; and, therefore, BRMS-1 does not appear to
suppress metastasis in M-4A4 cells. Furthermore, our initial
cytogenetic analysis of NM-2C5 and M-4A4 cells has revealed
no genetic aberrations on the chromosome 11q arm where
BRMS-1 resides. The small GTPase RhoC molecule has also
recently been identified as being influential in the metastasis of
the A375 series of human melanoma cell lines (76). By cDNA
array-based comparative analysis of an increasingly metastatic
series of cell lines derived by reculturing of metastatic deposits,
RhoC expression was identified as correlating with metastatic
capability. Subsequent overexpression of RhoC or a dominant-
negative RhoC in recipient A375 cells confirmed this correla-
tion. RhoC was found to be expressed equally in both the
metastatic M-4A4 and the nonmetastatic NM-2C5 cells; thus,
RhoC expression was not limiting for either cell line. As is the
case with many of the previously implicated metastasis-related
genes, in our model RhoC expression may be permissive, or
even essential for metastasis. However, our data suggest that
expression of this molecule is not, in itself, enough to cause
metastasis, as evidenced by the data from NM-2C5.
Perhaps the most significant experimental difficulty with
cell line analyses, reemphasized by our current study, is the
heterogeneity of the cultured populations. Most widely available
cell lines have been isolated from advanced tumors or ascites
containing millions of tumor cells, and these lines have often
been excessively passaged and have passed through many lab-
oratories, making interlaboratory comparisons difficult. Further-
more, many previous studies of metastasis-implicated genes
have logically investigated cell clones that have been transfected
to achieve high level expression of the exogenous gene. How-
ever, comparative metastasis analyses then have often been
made with the phenotype of the polyclonal parent cell line,
without determining the metastatic capability of the specific
selected clone before transfection. In such circumstances, it
remains possible that the phenotype displayed by the transfected
clone was already present before transfection and that a differ-
ence observed between this selected clone and its polyclonal
parent cell line merely reflects heterogeneity in the original
population (39, 42).
Despite decades of research, no single consistent marker or
effector of metastatic behavior has yet been identified. Given the
complexity of the metastatic process during which a dissemi-
nating tumor cell must accomplish several sequential tasks and
survive in many different environments, it is far more likely to
be a set of genes that is responsible for manifesting the overall
phenotype. Modern technological advances now permit the ap-
plication of high throughput gene expression analysis, by meth-
ods such as gene-chips and spotted microarrays, to identify
whether coordinated patterns of expression of clusters of genes
are involved in this complex process. Current work in this
laboratory is focused on applying these approaches to this tested
and well characterized experimental system.
Acknowledgments
We are grateful for the expertise of the University of California,
San Diego Histology Core facility for help with histological prepara-
tions. We also thank Carrie Viars for technical and analytical advice
during cytogenetic evaluations.
References
1. Tarin, D. Clinical and experimental studies on the biology of metas-
tasis. Biochim. Biophys. Acta, 780: 227–235, 1985.
2. Tarin, D., and Matsumura, Y. Recent advances in the study of tumour
invasion and metastasis. J. Clin. Pathol., 47: 385–390, 1994.
3. Fidler, I. J. Host and tumour factors in cancer metastasis. Eur. J. Clin.
Invest., 20: 481– 486, 1990.
4. Fidler, I. J., Kumar, R., Bielenberg, D. R., and Ellis, L. M. Molecular
determinants of angiogenesis in cancer metastasis. Cancer J. Sci. Am.,
4 (Suppl 1): S58 –S66, 1998.
5. Tarin, D., Price, J. E., Kettlewell, M. G., Souter, R. G., Vass, A. C.,
and Crossley, B. Mechanisms of human tumor metastasis studied in
patients with peritoneovenous shunts. Cancer Res., 44: 3584 –3592,
1984.
6. Fidler, I. J. Metastasis: quantitative analysis of distribution and fate
of tumor emboli labeled with 125 I-5-iodo-2⬘-deoxyuridine. J. Natl.
Cancer Inst., 45: 773–782, 1970.
7. Radinsky, R. Molecular mechanisms for organ-specific colon carci-
noma metastasis. Eur. J. Cancer, 31A: 1091–1095, 1995.
8. Fidler, I. J. Rationale and methods for the use of nude mice to study
the biology and therapy of human cancer metastasis. Cancer Metastasis
Rev., 5: 29 – 49, 1986.
9. Fidler, I. J. Orthotopic implantation of human colon carcinomas into
nude mice provides a valuable model for the biology and therapy of
metastasis. Cancer Metastasis Rev., 10: 229 –243, 1991.

10. Price, J. E., Polyzos, A., Zhang, R. D., and Daniels, L. M. Tumor-
igenicity and metastasis of human breast carcinoma cell lines in nude
mice. Cancer Res., 50: 717–721, 1990.
11. Bruns, C. J., Harbison, M. T., Kuniyasu, H., Eue, I., and Fidler, I. J.
In vivo selection and characterization of metastatic variants from human
pancreatic adenocarcinoma by using orthotopic implantation in nude
mice. Neoplasia, 1: 50 – 62, 1999.
12. Gehlsen, K. R., Davis, G. E., and Sriramarao, P. Integrin expression
in human melanoma cells with differing invasive and metastatic prop-
erties. Clin. Exp. Metastasis, 10: 111–120, 1992.
13. Cailleau, R., Olive, M., and Cruciger, Q. V. Long-term human
breast carcinoma cell lines of metastatic origin: preliminary character-
ization. In Vitro, 14: 911–915, 1978.
14. Schrock, E., du Manoir, S., Veldman, T., Schoell, B., Wienberg, J.,
Ferguson-Smith, M. A., Ning, Y., Ledbetter, D. H., Bar-Am, I., Soen-
ksen, D., Garini, Y., and Ried, T. Multicolor spectral karyotyping of
human chromosomes. Science (Wash. DC), 273: 494 – 497, 1996.
15. Macville, M., Veldman, T., Padilla-Nash, H., Wangsa, D., O’Brien,
P., Schrock, E., and Ried, T. Spectral karyotyping, a 24-colour FISH
technique for the identification of chromosomal rearrangements. Histo-
chem. Cell Biol., 108: 299 –305, 1997.
16. Heid, C. A., Stevens, J., Livak, K. J., and Williams, P. M. Real time
quantitative PCR. Genome Res., 6: 986 –994, 1996.
17. Schmittgen, T. D., Zakrajsek, B. A., Mills, A. G., Gorn, V., Singer,
M. J., and Reed, M. W. Quantitative reverse transcription-polymerase
chain reaction to study mRNA decay: comparison of endpoint and
real-time methods. Anal. Biochem., 285: 194 –204, 2000.
18. Rajeevan, M. S., Vernon, S. D., Taysavang, N., and Unger, E. R.
Validation of array-based gene expression profiles by real-time (kinetic)
RT-PCR. J. Mol. Diagn., 3: 26 –31, 2001.
19. Jalkanen, S., Jalkanen, M., Bargatze, R., Tammi, M., and Butcher,
E. C. Biochemical properties of glycoproteins involved in lymphocyte
recognition of high endothelial venules in man. J. Immunol., 141:
1615–1623, 1988.
20. Aogi, K., Kitahara, K., Urquidi, V., Tarin, D., and Goodison, S.
Comparison of telomerase and CD44 expression as diagnostic tumor
markers in lesions of the thyroid. Clin. Cancer Res., 5: 2790 –2797,
1999.
21. Woodman, A. C., Goodison, S., Drake, M., Noble, J., and Tarin, D.
Noninvasive diagnosis of bladder carcinoma by enzyme-linked immu-
nosorbent assay detection of CD44 isoforms in exfoliated urothelia.
Clin. Cancer Res., 6: 2381–2392, 2000.
22. Goodison, S., Yoshida, K., Churchman, M., and Tarin, D. Multiple
intron retention occurs in tumor cell CD44 mRNA processing. Am. J.
Pathol., 153: 1221–1228, 1998.
23. Weidner, N. Current pathologic methods for measuring intratu-
moral microvessel density within breast carcinoma and other solid
tumors. Breast Cancer Res. Treat., 36: 169 –180, 1995.
24. Ogawa, Y., Chung, Y. S., Nakata, B., Takatsuka, S., Maeda, K.,
Sawada, T., Kato, Y., Yoshikawa, K., Sakurai, M., and Sowa, M.
Microvessel quantitation in invasive breast cancer by staining for factor
VIII-related antigen. Br. J. Cancer., 71: 1297–1301, 1995.
25. Goodison, S., Yoshida, K., Sugino, T., Woodman, A., Gorham, H.,
Bolodeoku, J., Kaufmann, M., and Tarin, D. Rapid analysis of distinc-
tive CD44 RNA splicing preferences that characterize colonic tumors.
Cancer Res., 57: 3140 –3144, 1997.
26. Goodison, S., and Tarin, D. Clinical implications of anomalous
CD44 gene expression in neoplasia. Front. Biosci., 3: E89 –E109, 1998.
27. Gunthert, U., Hofmann, M., Rudy, W., Reber, S., Zoller, M.,
Haussmann, I., Matzku, S., Wenzel, A., Ponta, H., and Herrlich, P. A
new variant of glycoprotein CD44 confers metastatic potential to rat
carcinoma cells. Cell, 65: 13–24, 1991.
28. Rudy, W., Hofmann, M., Schwartz-Albiez, R., Zoller, M., Heider,
K. H., Ponta, H., and Herrlich, P. The two major CD44 proteins
expressed on a metastatic rat tumor cell line are derived from different
splice variants: each one individually suffices to confer metastatic
behavior. Cancer Res., 53: 1262–1268, 1993.
Clinical Cancer Research 73
29. Lackner, C., Moser, R., Bauernhofer, T., Wilders-Truschnig, M.,
Samonigg, H., Berghold, A., and Zatloukal, K. Soluble CD44 v5 and v6
in serum of patients with breast cancer. Correlation with expression of
CD44 v5 and v6 variants in primary tumors and location of distant
metastasis. Breast Cancer Res. Treat., 47: 29 – 40, 1998.
30. Martin, S., Jansen, F., Bokelmann, J., and Kolb, H. Soluble CD44
splice variants in metastasizing human breast cancer. Int. J. Cancer, 74:
443– 445, 1997.
31. Gardner, H. A., Berse, B., and Senger, D. R. Specific reduction in
osteopontin synthesis by antisense RNA inhibits the tumorigenicity of
transformed Rat1 fibroblasts. Oncogene, 9: 2321–2326, 1994.
32. Su, L., Mukherjee, A. B., and Mukherjee, B. B. Expression of
antisense osteopontin RNA inhibits tumor promoter-induced neoplastic
transformation of mouse JB6 epidermal cells. Oncogene, 10: 2163–
2169, 1995.
33. Oates, A. J., Barraclough, R., and Rudland, P. S. The identification
of osteopontin as a metastasis-related gene product in a rodent mam-
mary tumour model. Oncogene, 13: 97–104, 1996.
34. Fidler, I. J., and Kripke, M. L. Metastasis results from preexisting
variant cells within a malignant tumor. Science (Wash. DC)., 197:
893– 895, 1977.
35. Fidler, I. J. Tumor heterogeneity and the biology of cancer invasion
and metastasis. Cancer Res., 38: 2651–2660, 1978.
36. Poste, G., and Fidler, I. J. The pathogenesis of cancer metastasis.
Nature (Lond.), 283: 139 –146, 1980.
37. Nicolson, G. L., and Poste, G. Tumor cell diversity and host
responses in cancer metastasis—part I—properties of metastatic cells.
Curr. Probl. Cancer, 7: 1– 83, 1982.
38. Nicolson, G. L., Winkelhake, J. L., and Nussey, A. C. An approach
to studying the cellular properties associated with metastasis: some in
vitro properties of tumor variants selected in vivo for enhanced metas-
tasis. In: Weiss L. (ed.), Fundamental Aspects of Metastasis, pp. 291–
303. Amsterdam, North-Holland, 1976.
39. Bardeesy, N., Bastian, B. C., Hezel, A., Pinkel, D., DePinho, R. A.,
and Chin, L. Dual inactivation of RB and p53 pathways in RAS-induced
melanomas. Mol. Cell Biol., 21: 2144 –2153, 2001.
40. Chen, C. R., Kang, Y., and Massague, J. Defective repression of
c-myc in breast cancer cells: a loss at the core of the transforming growth
factor ␤ growth arrest program. Proc. Natl. Acad. Sci. USA, 98: 992–
999, 2001.
41. Potter, E., Bergwitz, C., and Brabant, G. The cadherin-catenin
system: implications for growth and differentiation of endocrine tissues.
Endocr. Rev., 20: 207–239, 1999.
42. MacCalman, C. D., Farookhi, R., and Blaschuk, O. W. Estradiol
regulates E-cadherin mRNA levels in the surface epithelium of the
mouse ovary. Clin. Exp. Metastasis, 12: 276 –282, 1994.
43. Folkman, J. Endothelial cells and angiogenic growth factors in
cancer growth and metastasis. Introduction. Cancer Metastasis Rev., 9:
171–174, 1990.
44. Streit, M., Velasco, P., Riccardi, L., Spencer, L., Brown, L. F.,
Janes, L., Lange-Asschenfeldt, B., Yano, K., Hawighorst, T., Iruela-
Arispe, L., and Detmar, M. Thrombospondin-1 suppresses wound heal-
ing and granulation tissue formation in the skin of transgenic mice.
EMBO J., 19: 3272–3282, 2000.
45. Aznavoorian, S., Murphy, A. N., Stetler-Stevenson, W. G., and
Liotta, L. A. Molecular aspects of tumor cell invasion and metastasis.
Cancer (Phila.)., 71: 1368 –1383, 1993.
46. McCawley, L. J., and Matrisian, L. M. Matrix metalloproteinases:
multifunctional contributors to tumor progression. Mol. Med. Today, 6:
149 –156, 2000.
47. Woodman, A. C., Sugiyama, M., Yoshida, K., Sugino, T., Borgya,
A., Goodison, S., Matsumura, Y., and Tarin, D. Analysis of anomalous
CD44 gene expression in human breast, bladder, and colon cancer and
correlation of observed mRNA and protein isoforms. Am. J. Pathol.,
149: 1519 –1530, 1996.
48. Stamenkovic, I., Aruffo, A., Amiot, M., and Seed, B. The hemato-
poietic and epithelial forms of CD44 are distinct polypeptides with
74 Breast Metastasis Model Characterization
different adhesion potentials for hyaluronate-bearing cells. EMBO J.,
10: 343–348, 1991.
49. Hofmann, M., Rudy, W., Zoller, M., Tolg, C., Ponta, H., Herrlich,
P., and Gunthert, U. CD44 splice variants confer metastatic behavior in
rats: homologous sequences are expressed in human tumor cell lines.
Cancer Res., 51: 5292–5297, 1991.
50. Hofmann, M., Rudy, W., Gunthert, U., Zimmer, S. G., Zawadzki,
V., Zoller, M., Lichtner, R. B., Herrlich, P., and Ponta, H. A link
between ras and metastatic behavior of tumor cells: ras induces CD44
promoter activity and leads to low-level expression of metastasis-spe-
cific variants of CD44 in CREF cells. Cancer Res., 53: 1516 –1521,
1993.
51. Birch, M., Mitchell, S., and Hart, I. R. Isolation and characterization
of human melanoma cell variants expressing high and low levels of
CD44. Cancer Res., 51: 6660 – 6667, 1991.
52. Sy, M. S., Guo, Y. J., and Stamenkovic, I. Distinct effects of two
CD44 isoforms on tumor growth in vivo. J. Exp. Med., 174: 859 – 866,
1991.
53. Driessens, M. H., Stroeken, P. J., Rodriguez Erena, N. F., van der
Valk, M. A., van Rijthoven, E. A., and Roos, E. Targeted disruption of
CD44 in MDAY-D2 lymphosarcoma cells has no effect on subcutane-
ous growth or metastatic capacity. J. Cell Biol., 131: 1849 –1855, 1995.
54. Bornstein, P. Diversity of function is inherent in matricellular
proteins: an appraisal of thrombospondin 1. J. Cell Biol., 130: 503–506,
1995.
55. Lawler, J. The functions of thrombospondin-1 and -2. Curr. Opin.
Cell Biol., 12: 634 – 640, 2000.
56. Weinstat-Saslow, D. L., Zabrenetzky, V. S., VanHoutte, K., Frazier,
W. A., Roberts, D. D., and Steeg, P. S. Transfection of thrombospondin
1 complementary DNA into a human breast carcinoma cell line reduces
primary tumor growth, metastatic potential, and angiogenesis. Cancer
Res., 54: 6504 – 6511, 1994.
57. Volpert, O. V., Lawler, J., and Bouck, N. P. A human fibrosarcoma
inhibits systemic angiogenesis and the growth of experimental metas-
tases via thrombospondin-1. Proc. Natl. Acad. Sci. USA, 95: 6343–
6348, 1998.
58. Senger, D. R., Asch, B. B., Smith, B. D., Perruzzi, C. A., and
Dvorak, H. F. A secreted phosphoprotein marker for neoplastic trans-
formation of both epithelial and fibroblastic cells. Nature (Lond.)., 302:
714 –715, 1983.
59. Senger, D. R., Perruzzi, C. A., Gracey, C. F., Papadopoulos, A., and
Tenen, D. G. Secreted phosphoproteins associated with neoplastic trans-
formation: close homology with plasma proteins cleaved during blood
coagulation. Cancer Res., 48: 5770 –5774, 1988.
60. Behrend, E. I., Craig, A. M., Wilson, S. M., Denhardt, D. T., and
Chambers, A. F. Reduced malignancy of ras-transformed NIH 3T3 cells
expressing antisense osteopontin RNA. Cancer Res., 54: 832– 837,
1994.
61. Brown, L. F., Papadopoulos-Sergiou, A., Berse, B., Manseau, E. J.,
Tognazzi, K., Perruzzi, C. A., Dvorak, H. F., and Senger, D. R. Os-
teopontin expression and distribution in human carcinomas. Am. J.
Pathol., 145: 610 – 623, 1994.
62. Tuck, A. B., O’Malley, F. P., Singhal, H., Harris, J. F., Tonkin,
K. S., Kerkvliet, N., Saad, Z., Doig, G. S., and Chambers, A. F.
Osteopontin expression in a group of lymph node negative breast cancer
patients. Int. J. Cancer, 79: 502–508, 1998.
63. Sharp, J. A., Sung, V., Slavin, J., Thompson, E. W., and Henderson,
M. A. Tumor cells are the source of osteopontin and bone sialoprotein
expression in human breast cancer. Lab. Invest., 79: 869 – 877, 1999.
64. Sung, V., Gilles, C., Murray, A., Clarke, R., Aaron, A. D., Azumi,
N., and Thompson, E. W. The LCC15-MB human breast cancer cell line
expresses osteopontin and exhibits an invasive and metastatic pheno-
type. Exp. Cell Res., 241: 273–284, 1998.
65. Tuck, A. B., Arsenault, D. M., O’Malley, F. P., Hota, C., Ling,
M. C., Wilson, S. M., and Chambers, A. F. Osteopontin induces in-
creased invasiveness and plasminogen activator expression of human
mammary epithelial cells. Oncogene, 18: 4237– 4246, 1999.
66. Singhal, R. L., Prajda, N., and Weber, G. Intracellular signalling:
phosphatidylinositol lipid metabolism in cancer cells. Indian J. Bio-
chem. Biophys., 34: 25–28, 1997.
67. Weber, G. F., Ashkar, S., Glimcher, M. J., and Cantor, H. Receptor-
ligand interaction between CD44 and osteopontin (Eta-1). Science
(Wash. DC)., 271: 509 –512, 1996.
68. Lee, J. H., and Welch, D. R. Suppression of metastasis in human
breast carcinoma MDA-MB-435 cells after transfection with the metas-
tasis suppressor gene, KiSS-1. Cancer Res., 57: 2384 –2387, 1997.
69. Seraj, M. J., Samant, R. S., Verderame, M. F., and Welch, D. R.
Functional evidence for a novel human breast carcinoma metastasis
suppressor, BRMS1, encoded at chromosome 11q13. Cancer Res., 60:
2764 –2769, 2000.
70. Yang, X., Welch, D. R., Phillips, K. K., Weissman, B. E., and Wei,
L. L. KAI1, a putative marker for metastatic potential in human breast
cancer. Cancer Lett., 119: 149 –155, 1997.
71. Phillips, K. K., White, A. E., Hicks, D. J., Welch, D. R., Barrett,
J. C., Wei, L. L., and Weissman, B. E. Correlation between reduction of
metastasis in the MDA-MB-435 model system and increased expression
of the Kai-1 protein. Mol. Carcinog., 21: 111–120, 1998.
72. Dong, J. T., Isaacs, W. B., Barrett, J. C., and Isaacs, J. T. Genomic
organization of the human KAI1 metastasis-suppressor gene. Genomics,
41: 25–32, 1997.
73. Steeg, P. S., Bevilacqua, G., Sobel, M. E., and Liotta, L. A.
Identification and characterization of differentially expressed genes in
tumor metastasis: the nm23 gene. Basic Life Sci., 57: 355–360, 1991.
74. Leone, A., Flatow, U., VanHoutte, K., and Steeg, P. S. Transfection
of human nm23–H1 into the human MDA-MB-435 breast carcinoma
cell line: effects on tumor metastatic potential, colonization and enzy-
matic activity. Oncogene, 8: 2325–2333, 1993.
75. Phillips, K. K., Welch, D. R., Miele, M. E., Lee, J. H., Wei, L. L.,
and Weissman, B. E. Suppression of MDA-MB-435 breast carcinoma
cell metastasis following the introduction of human chromosome 11.
Cancer Res., 56: 1222–1227, 1996.
76. Clark, E. A., Golub, T. R., Lander, E. S., and Hynes, R. O. Genomic
analysis of metastasis reveals an essential role for RhoC. Nature
(Lond.)., 406: 532–535, 2000.