The Spemann organizer gene, Goosecoid, promotes tumor metastasis.

Abstract

The process of invasion and metastasis during tumor progression is often reminiscent of cell migration events occurring during embryonic development. We hypothesized that genes controlling cellular changes in the Spemann organizer at gastrulation might be reactivated in tumors. The Goosecoid homeobox transcription factor is a known executer of cell migration from the Spemann organizer. We found that indeed Goosecoid is overexpressed in a majority of human breast tumors. Ectopic expression of Goosecoid in human breast cells generated invasion-associated cellular changes, including an epithelial-mesenchymal transition. TGF-beta signaling, known to promote metastasis, induced Goosecoid expression in human breast cells. Moreover, Goosecoid significantly enhanced the ability of breast cancer cells to form pulmonary metastases in mice. These results demonstrate that Goosecoid promotes tumor cell malignancy and suggest that other conserved organizer genes may function similarly in human cancer.

Full text

Page 1

The Spemann organizer gene, Goosecoid, promotes
tumor metastasis
Kimberly A. Hartwell*†, Beth Muir‡, Ferenc Reinhardt*, Anne E. Carpenter*, Dennis C. Sgroi‡,
and Robert A. Weinberg*†§
*Whitehead Institute for Biomedical Research, Cambridge, MA 02142;†Department of Biology, Massachusetts Institute of Technology,
Cambridge, MA 02139; and‡Department of Pathology, Harvard Medical School, Molecular Pathology Research Unit, Massachusetts
General Hospital, Boston, MA 02129
Contributed by Robert A. Weinberg, September 29, 2006 (sent for review August 16, 2006)
The process of invasion and metastasis during tumor progression
is often reminiscent of cell migration events occurring during
embryonic development. We hypothesized that genes controlling
cellularchangesintheSpemannorganizeratgastrulationmightbe
reactivated in tumors. The Goosecoid homeobox transcription
factor is a known executer of cell migration from the Spemann
organizer. We found that indeed Goosecoid is overexpressed in a
majority of human breast tumors. Ectopic expression of Goosecoid
in human breast cells generated invasion-associated cellular
changes, including an epithelial–mesenchymal transition. TGF-?
signaling, known to promote metastasis, induced Goosecoid ex-
pression in human breast cells. Moreover, Goosecoid significantly
enhanced the ability of breast cancer cells to form pulmonary
metastases in mice. These results demonstrate that Goosecoid
promotes tumor cell malignancy and suggest that other conserved
organizer genes may function similarly in human cancer.
epithelial–mesenchymal transition
T
nation. Such cells undergo changes in cell–cell adhesion, acquire
anchorage independence, gain motility, invade into and out of
the circulation, and colonize distant organs (1). The genetic
bases of these highly complex steps are largely unknown. How-
ever, some analogies exist between metastasizing cells and
migrating subpopulations of cells that mediate tissue reorgani-
zation during embryonic development (2). These analogies
suggest that signaling pathways controlling such embryonic
processes may be reactivated in tumor cells with significant
consequences.
Gastrulation is an embryonic developmental process that
displays some striking similarities to tumor invasion and metas-
tasis. This critical process establishes the basic body plan by way
of highly coordinated cell movements and is initiated by a
conserved group of cells originally characterized in Xenopus
laevis as the Spemann organizer (3). In higher vertebrates, the
organizer equivalents (e.g., the anterior primitive streak in
mouse and Hensen’s node in birds) are regions in which epithe-
lial cells break their cell–cell junctions and ingress into the
interior of the embryo, migrating as individual mesenchymal
cells (4). This shift of cell phenotype is defined as an epithelial–
mesenchymal transition (EMT).
An EMT is marked by the loss of epithelial properties through
down-regulation of epithelial components (e.g., E-cadherin and
cytokeratins), and the acquisition of mesenchymal proteins (e.g.,
N-cadherin and vimentin) in their stead (5). This transition can
impart additional mesenchymal properties to embryonic epithelial
cells, such as motility and invasiveness, which enable various cell
movements during gastrulation and other subsequent developmen-
tal processes requiring tissue remodeling (6). During cancer patho-
genesis, EMTs are similarly thought to confer these phenotypes
upon carcinoma cells, enabling them to complete some of the steps
requiredforsuccessfulinvasionandmetastasis(2).Indeed,multiple
umorcellsacquiretheabilitytometastasizebyevolvingtraits
that allow them to overcome multiple barriers to dissemi-
experimental studies provide support for such a functional link
betweenEMTandtumormetastasis.Forexample,certainproteins
that can induce an EMT in mammalian mammary epithelial cells,
such as Twist and TGF-?, were found to be necessary for the
metastatic behavior of tumor cells in vivo (7, 8).
The parallels between gastrula organizer biology and tumor
malignancy suggest that common signals may drive gastrulation
and metastasis. We therefore directed our attention to the
Goosecoid (Gsc) gene, which encodes a well conserved tran-
scription factor that was first identified as the most highly
expressed homeobox gene in the Spemann organizer (9–11). Gsc
can recapitulate many of the properties of the organizer when
ectopicallyexpressedintheamphibianembryo(12)andisknown
to promote cell migration in X. laevis (13). Moreover, elements
of the TGF-? superfamily and Wnt??-catenin signaling path-
ways, which are known to be involved in tumor invasion and
metastasis (2), can induce Gsc expression in embryonic cells and
are required for Spemann organizer formation (14, 15). For
these reasons we sought to ascertain whether Gsc also plays a
role in neoplastic disease. Goosecoid and its encoded protein
have not been previously studied in the context of human cancer
pathogenesis. The results described here strongly support the
notion that this embryonic transcription factor can indeed be
appropriated opportunistically by human cancer cells, allowing
such cells to acquire certain characteristics needed to overcome
key barriers to tumor metastasis.
Results
Elevated Goosecoid Expression in Human Breast Tumors. The expres-
sion patterns of the Goosecoid gene have not been well charac-
terized in human or murine adult tissues. To determine whether
a role for the GSC developmental gene in cancer was plausible,
we undertook to examine human tumor specimens for evidence
of GSC mRNA. Because probes for this gene were not included
in published microarray expression studies to the best of our
knowledge, we were unable to assess GSC expression patterns
through database mining. We therefore measured GSC levels in
a cohort of microdissected human breast tumors of three prev-
alent pathological subtypes: atypical ductal hyperplasia (ADH),
ductal carcinoma in situ (DCIS), and invasive ductal carcinoma
(IDC) (16). The 72 tumor samples examined were each accom-
panied by a patient-matched sample of normal breast epithe-
lium. The normal samples were presumably proliferative per
Author contributions: K.A.H., D.C.S., and R.A.W. designed research; K.A.H., B.M., and F.R.
performedresearch;K.A.H.,A.E.C.,D.C.S.,andR.A.W.analyzeddata;andK.A.H.wrotethe
paper.
The authors declare no conflict of interest.
Abbreviations: EMT, epithelial–mesenchymal transition; HMEC, human mammary epithe-
lial cell; MDCK, Madin–Darby canine kidney; IDC, invasive ductal carcinoma; ADH, atypical
ductal hyperplasia; DCIS, ductal carcinoma in situ.
§To whom correspondence should be addressed at: Whitehead Institute for Biomedical
Research, 9 Cambridge Center, Cambridge, MA 02142. E-mail: weinberg@wi.mit.edu.
© 2006 by The National Academy of Sciences of the USA
www.pnas.org?cgi?doi?10.1073?pnas.0608636103PNAS ?
December 12, 2006 ?
vol. 103 ?
no. 50 ?
18969–18974
CELL BIOLOGY

Page 2

published studies of normal human breast tissue (17). Because
all samples in this cohort were obtained by laser capture
microdissection, the samples do not contain significant numbers
of stromal cells.
Quantitative real-time RT-PCR was used to compare levels of
GSC mRNA represented in individual samples. The abundance
of GSC mRNA in the normal tissue samples was found to be low,
because signals were not detected until a high cycle number
during PCR amplification. Strikingly, GSC expression was ele-
vated in 56 of 72 tumors (78%) compared with corresponding
patient-matched normal tissue samples (Fig. 1). By subtype, 71%
of ADH samples, 79% of DCIS samples, and 78% of IDC
samples contained a level of GSC mRNA above that of patient-
matched normal tissue, and this pattern of GSC up-regulation
was found to be significant in each case (P ? 0.02 for ADH, P ?
0.01 for DCIS, and P ? 0.01 for IDC samples). The average
extent of elevation of GSC mRNA across all samples per subtype
was 5.9-, 9.6-, and 6.9-fold in the ADH, DCIS, and IDC samples,
respectively, compared with corresponding normal samples. A
more detailed view of the GSC expression data set can be found
in Table 1, which is published as supporting information on the
PNAS web site. Together, these results show that, in a majority
of human ductal-type breast tumors, GSC expression is signif-
icantly elevated above normal levels, consistent with a role for
this developmental gene in human cancer, as hypothesized.
Goosecoid Elicits an EMT and Enhances Cell Motility. To identify the
functional consequences of Gsc expression in adult epithelial
cells, we stably expressed this protein in immortalized human
mammary epithelial cells (HMECs) and in Madin–Darby canine
kidney (MDCK) epithelial cells using retroviral transduction
(Fig. 2A). Neither of these parental cell lines expressed substan-
tial levels of Gsc protein by Western blotting (Fig. 2A). In both
cell types, we observed that the population of cells expressing
ectopic Gsc lost cell–cell contacts and displayed a scattered
distribution in culture (Fig. 2B), whereas control cells retained
their typical epithelial morphology, continuing to grow as groups
of cobblestone-like cells. The morphological changes evident in
the Gsc-expressing cells were suggestive of an EMT. We there-
fore examined the status of known EMT markers in these cells.
The Gsc-expressing cells demonstrated marked down-regulation
of E-cadherin, ?-catenin and ?-catenin proteins, concordant
with the apparent loss of adherens junctions (Fig. 2 C and D).
These cells had replaced their cytokeratin-based intermediate
filament network with one based on vimentin and stained
positively for the mesenchymal protein N-cadherin (Fig. 2 C and
D). Moreover, the Gsc-expressing HMECs were found to be
substantially more migratory in transwell migration assays than
control cells (Fig. 2E). Our results demonstrate that Gsc induces
the central hallmarks of an EMT and cell motility in adult
mammalian epithelial cells, recapitulating cellular changes driv-
ing gastrulation in higher vertebrates.
TGF-? Signaling Induces Goosecoid Expression in Adult Breast Epithe-
lial Cells. The Wnt??-catenin and TGF-? superfamily signaling
cascades are required for Spemann organizer formation and Gsc
gene expression (18), and these same pathways have been impli-
catedintumormetastasis(2).BecauseGscrecapitulatedaspectsof
its embryonic organizer function in adult mammalian epithelial
cells, we tested whether these two organizer-associated signaling
cascades induce GSC expression in these cells. We found that the
enhancement of Wnt??-catenin signaling by two approaches failed
to activate GSC expression. Specifically, GSC mRNA expression
was not increased in HMECs either by expression of a nondegrad-
able form of ?-catenin (?N90 ?-catenin) (19) or by a constitutively
active form of Lef-1 (Lef-vp16) (20), a DNA-binding protein that
associates with ?-catenin to induce transcription of target genes
(Fig. 3A). We confirmed these constructs were transcriptionally
functional using the Topflash?Fopflash reporter system (data not
shown) (21).
In contrast, expression of constitutively active TGF-? type 1
receptor (22) in these cells using retroviral transduction did
induce GSC mRNA expression (Fig. 3B). GSC mRNA was also
induced in nontransduced HMECs in a dose-dependent manner
Fig. 1.
sample is shown with the lowest value of each pair in foreground. Pairs are grouped by tumor pathological subtype and sorted within groups according to the
level of GSC mRNA in the tumor samples. All values displayed were normalized to the average of the GSC mRNA levels in the normal samples, which is set as the
y value 1 in the graph. Values outside the scale of the y axis are marked by an asterisk.
Quantification of Goosecoid expression in human tumors. The relative level of GSC mRNA in each tumor (blue) and corresponding normal (red) tissue
18970 ?
www.pnas.org?cgi?doi?10.1073?pnas.0608636103Hartwell et al.

Page 3

by the addition of soluble, activated TGF-?1 to the cell culture
medium (Fig. 3C). When TGF-?1 was applied to HMECs
expressing nondegradable ?-catenin or the constitutively active
form of Lef-1 or GFP control, GSC expression was not induced
to a level greater than that achieved without activation of the
Wnt??-catenin pathway (data not shown). Together, these ex-
periments demonstrate that TGF-? signaling induces GSC in
adult breast epithelial cells as do related mesoderm-inducing
signals in gastrulating embryos and other cells (14, 23, 24). We
did not observe ?-catenin acting synergistically with these signals
in our system, contrary to observations in Xenopus embryos (14).
This may reflect distinct roles for Goosecoid in Xenopus and
humans, as well as distinct mechanisms of regulation.
Goosecoid Enhances the Metastatic Ability of Cancer Cells. Because
Goosecoid triggered an EMT and enhanced cell motility in adult
epithelial cells (both known correlates of invasive and metastatic
ability) we tested whether this gene could also promote tumor
metastasis. Gsc was ectopically expressed in GFP-labeled MDA-
MB-231 human breast cancer cells (Fig. 4A). The cells of this line
are weakly metastatic and quasi-mesenchymal, in that they do
not express E-cadherin and do express vimentin, yet they display
an epithelial-like morphology in culture (25). We observed that
upon the introduction of Gsc, the MDA-MB-231 cells acquired
a spindle-like morphology more typical of mesenchymal cells
(Fig. 4B) as well as an increased degree of motility (Fig. 4C).
ControlorGsc-expressingMDA-MB-231cellswereinjectedinto
the tail veins of mice and lungs were examined for metastases 8–10
weeks after injection (Fig. 4D). At both time points, a greater
Fig. 2.
caninekidneyepithelialcells.(A)EctopicexpressionofGscinHMECsandinMDCK
epithelialcellsbyWesternblotting.(B)Phase-contrastmicrographsofHMECsand
MDCKcellsexpressingeitherGscorGFPcontrol.(C)Expressionlevelsofepithelial
proteins E-cadherin, ?-catenin, and ?-catenin and mesenchymal proteins N-
cadherin and vimentin in HMECs and MDCK cells expressing either Gsc or GFP
control by Western blotting. ?-Actin protein is shown as a loading control. (D)
ImmunofluorescencestainingforepithelialproteinsE-cadherinandcytokeratins
and mesenchymal protein vimentin in MDCK cells expressing either Gsc or GFP
control.Antibodystainingisshowninred,andHoechstnuclearstainingisshown
in blue. (E) Quantification of the migratory abilities of HMECs expressing Gsc or
GFP control by transwell migration assay. Movement toward medium with or
withoutgrowthfactorsupplements(EGF,insulin,andhydrocortisone)isgraphed
asthepercentageoftotalcellsassayedthatmigratedafter48h.Assaysweredone
in triplicate, and the averages with SEM are shown.
Effects of Goosecoid expression in immortalized human breast and
Fig. 3.
levels in HMECs containing empty vector, nondegradable ?-catenin (?N90
?-cat), or constitutively active Lef-1 (Lef-vp16). Each bar represents the aver-
age with SEM of triplicate assays. (B) Relative GSC mRNA expression levels in
HMECs expressing either empty vector or constitutively active TGF-? type 1
receptor. Each bar represents the average with SEM of triplicate assays. (C)
Relative GSC mRNA expression levels in HMECs treated with activated TGF-?1
ligand at various concentrations for 3 or 6 days. Each bar represents the
average with SEM of triplicate assays.
Induction of Goosecoid in HMECs. (A) Relative GSC mRNA expression
Hartwell et al. PNAS ?
December 12, 2006 ?
vol. 103 ?
no. 50 ?
18971
CELL BIOLOGY

Page 4

number of pulmonary metastases were visible in the mice injected
with Gsc-expressing cells. Quantification of the observed lung
nodules at 8 weeks using image analysis indicated a 4-fold increase
in the average number of metastases in the mice injected with
Gsc-expressing cells compared with control animals (Fig. 4E).
This demonstrated enhancement of metastasis might have
arisen as a consequence of a Gsc-induced stimulation of prolif-
eration in vivo. To address this possibility, we directly compared
the in vivo proliferation rates of these two cell populations by
injecting them either into the s.c. space or into the mammary
glands of mice. In fact, the resulting primary tumors generated
by the Gsc-expressing cells grew more slowly than did control
tumors at both sites (data not shown). The demonstration that
Gsc-expressing breast cancer cells formed significantly greater
numbers of metastases in murine lungs despite proliferating
more slowly in vivo provide strong indication that Goosecoid
expression enhances the metastatic ability of MDA-MB-231
human breast cancer cells.
Discussion
Our results demonstrate that the Goosecoid homeobox tran-
scription factor, a major orchestrator of Spemann organizer
biology during gastrulation, plays an important role in activating
cell properties associated with tumor progression to malignancy.
To date, a limited number of developmental transcription fac-
tors, including SNAI1 (Snail), SNAI2 (Slug), and Twist, have
also been linked to metastasis (7, 26–28). During embryogenesis
these transcription factors are required for mesoderm formation
in Drosophila and neural crest development in vertebrates (5),
two processes in which an EMT and cell migration are critical
(6). Although the concept of the EMT as a driving force behind
human cancer metastasis is well described, there are still very
limited in vivo data demonstrating that genes inducing the
mesenchymal state contribute functionally to tumor metastasis
(5). Here we have found that Gsc is sufficient to enhance
metastatic behavior in an in vivo model of experimental metas-
tasis. Gsc may augment metastatic colonization by promoting
extravasation, cell survival in the environment of the lung, or
migration to hospitable microenvironments within the lung.
Our finding that GSC expression is up-regulated in the vast
majority of clinical ductal-type tumors supports a role for this
embryonic transcription factor in human breast cancer. The
up-regulation of GSC occurs quite early in multistep cancer
progressionratherthanconcurrentlywiththeovertdisplayofthe
invasive phenotype. This result is not unusual for breast carci-
noma progression; for example, the HER2?neu gene, known to
promote invasive cell behavior (29) and routinely used to inform
both patient treatment and prognosis, is likewise already over-
expressed in human tumors before the overt onset of invasive-
ness (30). Our observations are in concordance with other gene
expression studies examining different stages of ductal-type
breast cancer progression, which have shown that most expres-
sion changes associated with invasiveness are already present in
preinvasive tissue (16, 31). Moreover, our observations are
consistent with other published results demonstrating that sev-
eral genes shown to promote the metastatic behaviors and poor
prognosis of aggressive cancers, such as Slug and HOXB13, are
expressed in clinical specimens before the appearance of the
malignant tumor phenotype (32, 33). Thus, it possible that, in
human ductal-type breast tumors, Goosecoid primes cells for the
expression of aggressive phenotypes, which manifest themselves
in the context of subsequent alterations.
Taken together, the present results implicate the Spemann
organizer gene, Goosecoid, in tumor metastasis. Moreover, they
Fig. 4.
expressing either Gsc or GFP control by Western blotting. (B) Phase-contrast micrographs of MDA-MB-231 cells expressing either Gsc or GFP control. (C)
Quantification of the migratory abilities of MDA-MB-231 cells expressing Gsc or GFP control by transwell assay, graphed as the percentage of total cells assayed
thatmigratedafter16h.Assaysweredoneintriplicate,andtheaverageswithSEMareshown.(D)Representativebrightfieldandfluorescenceimagesofmouse
lung lobes 10 or 8 weeks after tail vein injection of MDA-MB-231 cells expressing either Gsc or GFP control. (E) Quantification of the number of metastatic foci
in the lungs of mice 8 weeks after tail vein injection of MDA-MB-231 cells expressing either Gsc or GFP control (n ? 6; trend was confirmed by four independent
experiments). Quartiles, medians, and the P value of the mean are shown.
Goosecoid expression changes the behavior of MDA-MB-231 human breast cancer cells in vitro and in mice. (A) Gsc expression in MDA-MB-231 cells
18972 ?
www.pnas.org?cgi?doi?10.1073?pnas.0608636103 Hartwell et al.

Page 5

suggest that the reactivation of conserved organizer genes may
be a recurrent theme in human cancer metastasis. Our findings
therefore warrant a comprehensive examination of these genes
in multiple types of human malignancies.
Materials and Methods
RNA Preparation and RT-PCR. The clinical cohort examined was
previously described (16). Seventy-two tumor samples were
obtained from 40 patients, 28 of whom had two or more
pathological subtypes of breast cancer detectable at diagnosis,
and each was accompanied by a patient-matched normal breast
tissue sample. The Massachusetts Institute of Technology Com-
mittee on the Use of Humans as Experimental Subjects and the
Massachusetts General Hospital Human Research Committee
approved this study of deidentified samples. cDNAs from the
previous study were additionally analyzed for GSC by real-time
quantitative PCR analysis by using the ABI 7900HT system as
previously described (16). The sequences of the GSC-specific
fluorogenic MGB probe (5? to 3?) and the PCR primer pair,
respectively, were as follows: VIC, CCCACCGTAGTATTTAT,
GCCGCCCGCGACTAG, and CACTTTATTGTACTGT-
CACCCTTAATTTAAC. Statistical significance was calculated
for this clinical data set by using the paired Student t test, and
relative expression was calculated as described (7).
For cell line analyses, total RNA was purified by using RNA
STAT-60 (Tel-Test, Friendswood, TX) and RNase-free DNase set
(Qiagen, Valencia, CA) according to the manufacturer’s instruc-
tions. Hexanucleotide mix (Roche, Indianapolis, IN) was used for
reverse transcription. Quantitative real-time RT-PCR was per-
formed in triplicate by using the iCycler apparatus (Bio-Rad,
Cambridge, MA) and SYBR-Green detection reagent, either from
stock (Molecular Probes, Eugene, OR) or in commercial master
mix (PerkinElmer Applied Biosystems, Foster City, CA). The
sequences of the GSC-specific primer pairs were (5? to 3?) TCT-
CAACCAGCTGCACTGTC (left) and GGCGGTTCTTAAAC-
CAGACC (right), and those of the GAPDH-specific pairs were
AGCCACATCGCTCAGACAC (left) and AATGAAGGGGT-
CATTGATGG (right). Experimental data were normalized to
GAPDH, and relative expression was calculated as described (7).
Expression Constructs and Virus Generation. Full-length mouse
Goosecoid cDNA (34) (provided by Martin Blum, Universitât
Hohenheim,Stuttgart,Germany)wassubclonedwithorwithoutan
HA-antigen tag at the amino terminus into the pWZL-Blasticidin
vector. A corresponding vector containing the GFP gene was used
as control. ?N90 ?-catenin consisting of mouse ?-catenin contain-
ing amino-terminal deletions of 90 aa (19) and Lef-vp16 consisting
ofmouseLef-1fusedtothetransactivationdomainfromtheherpes
simplex virus VP16 protein (20) (provided by Masahiro Aoki, The
Scripps Research Institute, La Jolla, CA) were expressed by using
thepBabe-Puromycinvector.Activated,myc-tagged,humanTGF?
type I receptor cDNA (22) (provided by Joan Massague ´, Sloan–
Kettering Cancer Center, New York, NY) was expressed by using
the pWZL-Blasticidin vector. pWZL and pBabe amphotropic
viruses and lentiviruses were generated and used for target cell
infection as previously described (35).
CellCulture.TheMDCKcelllinewasobtainedfromAmericanType
Culture Collection (Manassas, VA) and cultured in DMEM sup-
plemented with 10% heat-inactivated FCS. The immortalized,
nontransformed HMEC line, expressing the SV40 early region and
hTERT, was previously described (36) and cultured in DMEM and
F12 medium (1:1) containing the supplements EGF (10 ng?ml),
insulin (10 ?g?ml), and hydrocortisone (0.5 ?g?ml), with noted
exceptions. The Gsc-expressing HMEC cells were generated by
using differential trypsinization of the polyclonal population of
Gsc-transduced cells to separate out the scattered, less adherent
cells from those not expressing substantial amounts of Gsc, as
confirmed by Western blotting. Soluble, activated TGF-?1 ligand
(R & D Systems, Minneapolis, MN) was used at a working
concentration of 100 pM, or 2.5 ng?ml, in the presence of 5% calf
serum. The MDA-MB-231 cell line was maintained in DMEM
supplemented with 10% FBS.
Antibodies, Immunoblotting, and Immunofluorescence. A rabbit
polyclonal antibody against Gsc was generated by using a
KLH-conjugated peptide of the sequence CSENAEK-
WNKTSSSKA,andresultingantiserawereaffinity-purified(Co-
vance, Philadelphia, PA). The specificity of the antibody was
confirmed by Western immunoblotting using whole-cell lysates
expressing either tagged or untagged ectopic Gsc. Other primary
antibodies used were vimentin (V9, catalog no. MS129P from
Neomarkers, Fremont, CA), N-cadherin [catalog no. 180224
from Zymed (San Francisco, CA) and catalog no. 610920 from
BD Transduction Labs, San Jose, CA], ?-actin (catalog no. 8226
from Abcam, Cambridge, MA), pan-cytokeratin (catalog no.
071M from Biogenex, San Ramon, CA), ?-catenin, ?-catenin,
and E-cadherin (catalog nos. C21620, 610254, and 610182 from
BD Transduction Labs). Standard procedures were used for
immunoblotting and immunofluorescence.
Transwell Migration Assays. Cells were plated on cell culture
inserts (Falcon, West Chester, PA) containing a filter with
8.0-?m pores. Total cells and migrated cells were quantified by
using crystal violet staining after time indicated and compared
with control for differences in cell number as described (37).
Mice and Injection of Tumor Cells. Female NOD-SCID mice (prop-
agated on site) and nude mice (NCR nude; Taconic, Hudson,
NY) were used in these studies, and all protocols were approved
by the Massachusetts Institute of Technology Committee on
Animal Care. Nude mice received 400 rad of ?-radiation using
a dual137Cesium source 1 day before tumor cell injection. Mice
were anesthetized with either avertin (i.p.) or isoflurane (inha-
lation). For orthotopic injections, 1 million cells in 30 ?l of
Matrigel (Becton Dickinson, San Jose, CA) diluted 1:2 in
medium were injected into each of two mammary glands per
NOD-SCID mouse. For s.c. injections, 2 ? 106cells in 160 ?l of
Matrigel diluted 1:2 in medium were injected at each of three
sites per nude mouse. For tail vein injections, 2 ? 106cells in 200
?l PBS were injected per mouse. Tumor diameters were mea-
sured multiple times per week by using precision calipers.
Visualization and Quantification of GFP-Labeled Lung Metastases.
Upon necropsy, lungs of injected mice were removed, separated
into individual lobes, and examined under a Leica MZ 12
fluorescence dissection microscope. Images of both faces of all
lobes were captured at identical settings, and the fluorescent
metastatic nodules in each image were analyzed by using Cell-
Profiler image analysis software developed in the laboratory of
David Sabatini (www.cellprofiler.org) (38). The unpaired Stu-
dent t test was used for statistical comparisons of these data.
We thank I. Ben-Porath, J. Yang, S. Mani, C. Kuperwasser, B. Elenbaas,
R. Hynes, J. Lees, T. Ince, S. McCallister, L. Xu, R. Lee, A. Orimo, L.
Spirio, C. Scheel, A. Karnoub, S. Stewart, and other members of the
R.A.W. laboratory for invaluable input during the course of this work.
We also thank M. Blum, M. Aoki, J. Massague ´, M. van de Wetering, B.
Elenbaas, C. Kuperwasser, and L. Spirio for reagents; M. Brooks and M.
Rockas for technical assistance; and G. Bell for help with statistical
analysis. We acknowledge the support of the W. M. Keck Biological
Imaging Facility at Whitehead Institute. K.A.H. is the recipient of a U.S.
Army Predoctoral Breast Cancer Fellowship. R.A.W. is a Daniel K.
Ludwig Cancer Research Professor and an American Cancer Society
Research professor. This research was supported by National Institutes
of Health Grant R01-CA078461 and by a grant from the Breast Cancer
Research Foundation.
Hartwell et al.PNAS ?
December 12, 2006 ?
vol. 103 ?
no. 50 ?
18973
CELL BIOLOGY