Reduced Expression of the ROCK Inhibitor Rnd3 Is
Associated with Increased Invasiveness and Metastatic
Potential in Mesenchymal Tumor Cells
Cristina Belgiovine1, Roberta Frapolli2, Katiuscia Bonezzi3, Ilaria Chiodi1, Francesco Favero4, Maurizia
Mello-Grand4, Angelo P. Dei Tos5, Elena Giulotto6, Giulia Taraboletti3, Maurizio D’Incalci2, Chiara
1Istituto di Genetica Molecolare, Consiglio Nazionale delle Ricerche (CNR), Pavia, Italy, 2Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy, 3Istituto di Ricerche
Farmacologiche Mario Negri, Bergamo, Italy, 4Cancer Genomics Laboratory, Fondazione Edo ed Elvo Tempia, Biella, Italy, 5Ospedale Generale di Treviso, Treviso, Italy,
6Dipartimento di Genetica e Microbiologia, Universita ` di Pavia, Pavia, Italy
Background: Mesenchymal and amoeboid movements are two important mechanisms adopted by cancer cells to invade
the surrounding environment. Mesenchymal movement depends on extracellular matrix protease activity, amoeboid
movement on the RhoA-dependent kinase ROCK. Cancer cells can switch from one mechanism to the other in response to
different stimuli, limiting the efficacy of antimetastatic therapies.
Methodology and Principal Findings: We investigated the acquisition and molecular regulation of the invasion capacity of
neoplastically transformed human fibroblasts, which were able to induce sarcomas and metastases when injected into
immunocompromised mice. We found that neoplastic transformation was associated with a change in cell morphology
(from fibroblastic to polygonal), a reorganization of the actin cytoskeleton, a decrease in the expression of several matrix
metalloproteases and increases in cell motility and invasiveness. In a three-dimensional environment, sarcomagenic cells
showed a spherical morphology with cortical actin rings, suggesting a switch from mesenchymal to amoeboid movement.
Accordingly, cell invasion decreased after treatment with the ROCK inhibitor Y27632, but not with the matrix protease
inhibitor Ro 28-2653. The increased invasiveness of tumorigenic cells was associated with reduced expression of Rnd3 (also
known as RhoE), a cellular inhibitor of ROCK. Indeed, ectopic Rnd3 expression reduced their invasive ability in vitro and their
metastatic potential in vivo.
Conclusions: These results indicate that, during neoplastic transformation, cells of mesenchymal origin can switch from a
mesenchymal mode of movement to an amoeboid one. In addition, they point to Rnd3 as a possible regulator of
mesenchymal tumor cell invasion and to ROCK as a potential therapeutic target for sarcomas.
Citation: Belgiovine C, Frapolli R, Bonezzi K, Chiodi I, Favero F, et al. (2010) Reduced Expression of the ROCK Inhibitor Rnd3 Is Associated with Increased
Invasiveness and Metastatic Potential in Mesenchymal Tumor Cells. PLoS ONE 5(11): e14154. doi:10.1371/journal.pone.0014154
Editor: Neil A. Hotchin, University of Birmingham, United Kingdom
Received July 23, 2010; Accepted November 8, 2010; Published November 30, 2010
Copyright: ? 2010 Belgiovine et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by Fondazione Cariplo (grant 2006-0734). I.C. post-doctoral fellowship has been supported by Fondazione Cariplo. The
funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: firstname.lastname@example.org
Neoplastic transformation is a gradual process, during which
cells acquire successive mutations, which mainly cause the loss of
proliferation control, the ability to divide indefinitely and invade
other tissues. A critical step in the development of a malignant
cancer is tumor cells’ acquisition of the capacity to migrate and
invade tissues . Different molecular mechanisms are responsible
for the acquisition of a migratory and invasive phenotype, such as
changes in signal transduction pathways involving tyrosine kinases,
changes in cytoskeletal organization and in cell adhesion.
The small GTPases of the Rho family, mainly RhoA, cdc42 and
Rac [2–4], play a pivotal role in regulating the actin cytoskeleton
and cell movement. The RhoA/ROCK and Rac signalling
pathways are required for respectively amoeboid and mesenchy-
mal movements, which are the main types of movements adopted
by tumor cells [5,6]. The mesenchymal movement is typical of
cells displaying an elongated morphology in a 3D environment; it
requires integrin attachment to the extracellular matrix, formation
of focal contacts and pericellular proteolysis. Besides cancer cells of
mesenchymal origin, carcinoma cells can adopt this type of
migration, after undergoing an epithelial-mesenchymal transition
In contrast, some carcinoma cells can move very fast with an
amoeboid shape [8,9]; in vitro studies in 3D environments have
shown that this movement is typical of cells with a rounded
morphology and is associated with the formation of actin cortical
rings and membrane blebbings. The amoeboid movement, which
is potentially faster than the mesenchymal one, does not rely on
integrins, focal contacts and extracellular matrix degrading
PLoS ONE | www.plosone.org1November 2010 | Volume 5 | Issue 11 | e14154
enzymes; it is mainly based on contraction of actomyosin filaments
which enables the cells to squeeze through the extracellular matrix
without the requirement of matrix degradation.
Acto-myosin contractility is strictly dependent on the activity of
the RhoA-dependent kinase ROCK [9–11] and chemical blockade
of ROCK inhibits amoeboid movement . Cells can switch from
one type of invasion mechanism to the other in response to changes
in protein expression, or after treatment with certain compounds
[5,12–18]. For example, inhibition of matrix proteases or integrins
inhibits mesenchymal migration and promotes a transition towards
amoeboid movement [19,20]. Identifying the factors and genes
controlling the different types of motility could help direct
therapeutic strategies aimed at reducing invasion.
Amoeboid migration has been recognized as an important
mechanism of invasion and metastasis of carcinoma cells and, more
recently, of sarcomas . However, little is known about the
possible association between sarcoma malignant progression and
acquisition of the amoeboid phenotype. In this paper, we have
exploited the cen3tel model of isogenic cells at different stages of
transformation, from normal fibroblasts up to metastatic cells, to
study changes in the migratory and invasive potential accompany-
ing human fibroblast neoplastic transformation. The human
fibroblast cell line cen3tel, obtained in our laboratory by telomerase
immortalization, gradually underwent spontaneous neoplastic
transformation [22–24]. Studying cells at different phases of
transformation, we could show that an early event during
transformation was the loss of expression of the CDKN2A locus,
followed by inactivation of p53 and overexpression of c-myc. While
CDKN2A downregulation was not sufficient to make cells tumori-
genic, the ability to induce tumors in nude mice correlated with p53
inactivation and c-myc overexpression. During further culture
propagation, cen3tel cells showed a shorter latency in inducing
tumors, suggesting that they had acquired increased tumorigenicity
and could be a useful tool for obtaining further information on
molecular changes associated with tumor progression .
In this study we found that, upon neoplastic transformation,
cen3tel cells increased their migratory and invasive capacity by
adopting a protease-independent/ROCK-dependent mechanism
of invasion. We show here that Rnd3 (also known as RhoE), a
cellular inhibitor of ROCK-I , plays a relevant role in
regulating invasion and metastasis formation of sarcoma cells.
Materials and Methods
Procedures involving animals and their care were conducted in
conformity with the institutional guidelines that are in compliance
with national (Decreto Legge No. 116, Gazzetta Ufficiale, Suppl.
40, Feb. 18, 1992; Circolare No. 8, Gazzetta Ufficiale, July, 1994)
and international laws and policies (European Economic Com-
munity Council Directive 86/609, Official Journal Legislation
358. 1, Dec. 12, 1987; Guide for the Care and Use of Laboratory
Animals, United States National Research Council, 1996). The
study was reviewed and approved by the IRFMN Animal Care
and Use Committee (IACUC), which includes ‘‘ad hoc’’ members
for ethical issues. Aproval ID Frap1.
The cen3tel cellular system
The cen3tel telomerase immortalized cell line was obtained
from primary cen3 fibroblasts, derived from a centenarian, by
infection with an hTERT-containing retrovirus . Cen3tel cells
were used at different steps of propagation reflecting different
phases of transformation . In particular, we used cen3tel cells
at five phases of propagation (up to around population doubling
(PD) 1000): early cen3tel cells, these cells are at the initial passages
after infection with the hTERT containing retrovirus (between
PDs 34 and 45) and show a behaviour similar to that of primary
cen3 fibroblasts; mid cen3tel cells (around PD 100), these cells are
at an early phase of transformation, they are able to grow in the
absence of solid support, but are not tumorigenic in nude mice;
tumorigenic cen3tel cells, which induce tumors when inoculated
subcutaneously into nude mice, were further subdivided in three
groups, according to the time required for tumor formation, which
decreases at increasing PDs (see first section of the Results),
tumorigenic cells of phase I (around PD 160), phase II (around PD
600) and phase III (around PD 1000).
Cell culture, transfection and plasmid
Primary and immortalized cells were grown in Dulbecco’s
modified Eagle’s Medium (DMEM, Celbio) supplemented with
10% fetal bovine serum (Lonza), 2 mM glutammine, and 1% non-
essential amino acids (Euroclone), 0.1 mg/ml penicillin (Euro-
clone), 100 U/ml streptomycin (Euroclone) at 37uC in an
atmosphere containing 5% CO2. To analyze cellular morphology
and organization of the actin cytoskeleton (see section ‘‘Immuno-
fluorescence’’) in a 3D environment, cells were plated on top of
Matrigel (8 mg/ml) (BD Biosciences) in 8 chamber polystyrene
vessels (BD Transduction Laboratories Biosciences Falcon) and
incubated at 37uC with 5% CO2in complete DMEM to allow
Matrigel invasion. Cells in Matrigel were treated with 10 mM
Y27632 (Calbiochem) for 6 hours. Cell morphology was observed
using an optical microscope Olympus IX71 equipped with a 4x
objective (NA 0.13). Images were taken with a digital camera Cool
SNAPES(Photometrics) using the MetaMorph software.
Cells were transfected with the linearized plasmids EGFP-C1
(Clontech) or EGFP-C1-Rnd3 (kindly provided by Dr. Pierre
Roux, CRBM-CNRS FRE2593, France) using Lipofectamine
2000 (Invitrogen) according to the manufacturer’s instructions.
Transfected cells were selected and expanded in complete medium
containing 0.5 mg/ml G418 (Invitrogen).
Motility and Invasion Assays
Cell motility and invasiveness were assayed using modified
Boyden chambers with polycarbonate PVP-free Nucleopore filters
(8 mm pore size) . Supernatant of NIH-3T3 cells was used as a
reference attractant and was added to the lower compartment of
the Boyden chamber. For motility, filters were coated with 0.1%
gelatin. For invasion, filters were coated with a thick layer of the
reconstituted basement membrane Matrigel (0.5 mg/ml; Becton
Dickinson). Cells were detached, washed in DMEM containing
0.1% BSA, resuspended in the same medium at a concentration of
56105/ml, and added to the upper compartment of the chamber.
When indicated, the ROCK inhibitor Y27632 (Calbiochem) or
the MMP inhibitor Ro 28–2653 (kindly provided by H.W. Krell,
Roche Diagnostics Gmbh, Penzberg, Germany) was added to the
cells 1 hour before the assay and left throughout the assay. After
4 hours (motility) or 6 hours (invasion), filters were stained with
Diff-Quik (Baxter), and migrated cells in 10 high-power fields were
counted. We choose these times of analysis to avoid that possible
proliferation differences between cell lines could affect the results.
Results are expressed as the number of migrated or invaded cells
or as the percentage of control invasion (inhibitor treated or
transfected cells). Statistical differences between groups were
evaluated by ANOVA followed by Bonferroni post hoc test.
Cells were seeded at different density in 12 well plates (Corning)
containing 19 mm diameter coverslips and incubated at 37uC for
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org2November 2010 | Volume 5 | Issue 11 | e14154
24 hours. Filamentous actin (F-actin) was stained with phalloidin-
TRITC (P1951, Sigma-Aldrich) diluted 1:300 and nuclei with
10 min. The phosphorylated form of myosin II regulatory light
chain (p-MLC) was stained using the anti-pS19-MLC primary
antibody (#3671, Cell Signaling) diluted 1:50 and an anti-rabbit
secondary antibody conjugated with FITC (Jackson Immunor-
esearch), after an overnight block in 0.3% BSA in PBS at 4uC. To
detect F-actin in cells in Matrigel matrix, cells were fixed for 30
minutes with 4% paraformaldehyde-0.25% glutaraldheyde in PBS
on ice and then treated for 30 minutes with 0.2% Triton-X 100 in
PBS. Cells were then incubated with phalloidin-TRITC (P1951,
Sigma-Aldrich) diluted 1:1000 and with 0.2 mg/ml DAPI for
3 hours. Slides were analyzed with the Leica TCS SP2 confocal
laser microscope using a 40x objective (NA 1.32). Figures were
assembled using Adobe Photoshop and Adobe Illustrator.
(DAPI) in PBS for
Western blot analysis
Whole-cell lysates were prepared using the RIPA buffer (50 mM
Tris-HCl pH 8, 150 mM NaCl, 1% Nonidet P40, 0.1% SDS,
0.1% DOC, 1x protease inhibitor cocktail (Roche) and 0.2%
Na3VO4) or the Laemmli buffer (60 mM Tris-HCl, pH 6.8, 10%
glycerol, 5% b-mercaptoethanol, 2% SDS, 0.02% bromophenol
blue). The anti-Rnd3 antibody [R 6153, clone 4 (Sigma-Aldrich),
dilution 1:500] was used on extracts prepared with the RIPA
buffer. The anti-pS19-MLC (1:500) was used on extracts prepared
with the Laemli buffer and blotted onto PVDF membranes
(Biorads). The anti-c-tubulin antibody (T6557, clone GTU-88,
Sigma-Aldrich) was used as control for protein loading. Secondary
Horseradish Peroxidase conjugated antibodies were from Jackson
ImmunoResearch (anti-mouse IgG 115-035-146, anti-rabbit IgG
111-035-003). Chemiluminescent assays (Pierce) were used to
detect the secondary antibody signal.
RNA extraction and microarray analysis
Total RNA was extracted using the Trizol reagent (Invitrogen)
from actively dividing primary fibroblasts (PD 15) and from cen3tel
cells representing the five phases of propagation (PD 37, 97, 167,
618, 1032 and 1042;cells at PDs 1032 and 1042 both represent cells
of tumorigenic phase III). Microarray probe preparation, hybrid-
ization on Agilent Whole Genome 44k oligo microarrays and
scanning werecarried out as previously described . Images were
analyzed using Feature Extraction software (Agilent Technologies)
version 8.1. Output files were then treated with the Limma (linear
models for microarray data) package  available within
Bioconductor (http://www.bioconductor.org/). Linear models
were fitted to the normalised data to compare cen3tel cells at each
population doubling with control cen3 fibroblasts. The empirical
Bayes method was used to compute the moderated t-statistic and F-
statistic . P-values were adjusted for multiple testing by the
Benjamini-Hochberg approach . To select differently expressed
transcripts, a cut-off of 0.01 was applied to the adjusted p-value.
Microarray data have been deposited in the NCBI database GEO
(Gene Expression Omnibus), accession number GSE157442.
The in vivo tumorigenic potential of cen3tel cells was evaluated by
injecting 107cells in the right flank of female athymic nude mice or
female SCID mice (Harlan Italy, Bresso, Italy). Mice were monitored
two/three times a week to assess tumor appearance and growth. To
investigate the metastatic ability of cen3tel cells, 26106cells were
injected i.v. in nude or SCID mice. Mice were monitored daily and
sacrificed at the appearance of distress symptoms. Animals were
autopsied to evaluate metastases. To establish the capacity of primary
tumors to give spontaneous metastases, tumor masses obtained after
s.c. injection of cen3tel cells were surgically removed under isoflorane
anesthesia and the mice were kept alive until the appearance of distress
symptoms. After sacrifice, animals were autopsied to evaluate
metastases. For counting metastatic foci, tissues were collected and
stored in BOUIN.Forhistopathologicexamination, tumor masses and
animal parenchymal organs were fixed in 10% buffered formalin, and
paraffin embedded. 4-mm sections were cut and stained with
hematoxylin and eosin.
In vivo tumorigenicity and metastatic potential of cen3tel
cells at different stages of propagation
Cen3tel cells at different PDs were inoculated subcutaneously in
immunocompromised mice. Analysis of cen3tel cells at tumori-
genic phase I and phase II (respectively around PD 160 and PD
600, see Material and Methods for the description of the cellular
system) confirmed our previous results , with tumors becoming
detectable with latencies of about one month and eight days,
respectively (Fig. 1A). Conversely, phase III tumorigenic cen3tel
cells (PD 900–1020) generated tumors already evident two days
after inoculation (Fig. 1A). Histological analysis revealed that the
tumors developed by cen3tel cells at the first and second
tumorigenic phases were pleomorphic sarcomas (Fig. 1B and C),
those developed by phase III cen3tel cells showed a hemangio-
pericytoma-like vascular pattern, similar to human poorly
differentiated, round-cell synovial sarcoma (Fig. 1D). Cen3tel cells
of tumorigenic phase III recapitulated the histological features of
synovial sarcomas even in the absence of the SYT-SSX fusion
transcript (data not shown), which is diagnostic of this sarcoma
subtype , suggesting the involvement of other genetic or
epigenetic mechanisms, possibly associated with the still poorly
understood SYT-SSX downstream targets.
Cen3tel cells at different PDs were also injected intravenously to
test whether they were able to give metastases. Both cen3tel cells at
tumorigenic phases I and II did not induce metastases (in 3 mice
each, Fig. 1E and F). In contrast, in the lung of 10 out of 10 mice
injected with cen3tel cells around PD 1000, a high number of
metastases (always more than 50 metastatic foci per lung) were
observed 4 weeks after injection (Figure 1G).
Tumorigenic cen3tel cells are more motile and invasive
than non-tumorigenic cells
To study motility and invasiveness in cen3tel cells at different
PDs, we used Boyden chambers with filters coated either with
gelatin or with a thick layer of the reconstituted basement
membrane Matrigel. The number of cells able to migrate through
the porous membrane coated by gelatin was considered a measure
of the migratory capacity, and the number of cells passing through
Matrigel a measure of the invasive capacity. As illustrated in
Figure 2A, which shows both spontaneous motility (black columns)
and motility in response to a chemoattractant (grey columns),
parental cen3 fibroblasts, as well as early and mid cen3tel cells,
had a very limited migratory potential; in contrast, tumorigenic
cen3tel cells were clearly able to migrate through the porous
membrane. The same holds true for the ability to invade Matrigel
(Fig. 2B). Interestingly, invasiveness was also high in the absence of
the chemoattractant. Thus, cen3tel neoplastic transformation is
associated with increased migratory and invasive potential. In the
in vitro experiments, we could see no clear difference in the
migratory or invasive potential between the cells in the three
phases of tumorigenicity. In particular, in cen3tel cells around PD
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org3November 2010 | Volume 5 | Issue 11 | e14154
1000, there was no clear increase in in vitro invasion, that could be
related to the acquisition of the metastatic ability.
Tumorigenic cen3tel cells show actin cortical rings and a
spherical morphology in a 3D environment
Cen3tel cells routinely grown on plastic showed a clear change
in morphology during culture propagation. Cells switched from a
typical elongated fibroblastic shape, distinctive of cells up to
around PD 100, to a polygonal shape characteristic of cells that
had become neoplastically transformed . To test whether
neoplastic transformation and the change in cell morphology were
associated with changes in the actin cytoskeleton, we analyzed the
organization of the actin filaments in cells at different stages of
propagation. We stained F-actin using its interacting molecule
phalloidin, conjugated to a fluorochrome. A clear difference in F-
actin organization was observed between tumorigenic and non-
tumorigenic cen3tel cells (Fig. 3A–F). As expected, F-actin was
organized in stress fibers in primary cen3 fibroblasts (Fig. 3A) and
the same organization was also observed in early and mid cen3tel
cells (Fig. 3B, C). In tumorigenic cen3tel cells (Fig. 3D–F),
phalloidin staining clearly highlighted the change in morphology
with cortical rings formed along the inner cell periphery.
Together with the change in actin organization, there was a
parallel re-distribution of p-MLC (Fig. 3G–L). In fact, while in
non-tumorigenic cells, the phosphorylated form of myosin II was
distributed in fibers along the cells (Fig. 4G–I), in tumorigenic cells
it mainly formed a ring around the inner cell periphery (Fig. 4J–L).
The change in cytoskeletal organization observed in tumori-
genic cen3tel cells suggests that the transformed cells could have
switched to an amoeboid-type movement. Cells migrating through
this movement have a characteristic rounded shape when grown in
a deformable gel like Matrigel and show cortical actin rings with
membrane blebbings . To analyze cen3tel tumorigenic cell
morphology and actin organization in 3D, we plated them on the
top of Matrigel and allowed them to invade the gel. After
24 hours, cells were first analyzed by phase contrast microscope
(Fig. 4A, A’–C’) and then fixed and stained with phalloidin
conjugated with TRITC (Fig. 4A, G’–I’, M’–O’). Tumorigenic
cells had a rounded morphology (A’–C’), with F–actin cortical
rings (G’–I’, M’–O’) (see also Movie S1). It is worth noticing that
the rounded morphology was also observed in cells of the sarcoma-
like tumor masses generated by cen3tel cells in animal models.
ROCK inhibition leads to an elongated morphology and
reduces cen3tel cell invasion
Rounded cell morphology and cortical actin rings depend on
the activity of the Rho-effector kinase ROCK, which acts by
increasing myosin-II-mediated actin filament stabilization and
contraction . Inhibition of ROCK causes the loss of the
spherical cellular morphology . To test whether this was the
case in tumorigenic cen3tel cells, we exposed cells that had
invaded Matrigel to the ROCK chemical inhibitor Y27632.
Treatment with the inhibitor produced a more elongated
morphology (Fig. 4A, D’–F’) and altered the actin cortical rings
Figure 1. Tumor induction by cen3tel cells. A) Growth curves of tumors induced by cen3tel cells at PD 165 (empty circles), PD 616 (empty
squares) and PD 902 (black triangles). B–D) Histological analysis (hematoxylin and eosin staining) of tumors induced by cen3tel cells at PD 165 (B), PD
616 (C) and PD 977 (D). E–G). Lungs from mice intravenously injected with tumorigenic cen3tel cells of phase I (E), phase II (F) or phase III (G), only in
(G) are metastases visible.
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org4November 2010 | Volume 5 | Issue 11 | e14154
(Fig. 4A, J’–L’, P’–R’), indicating that ROCK activity is involved
in the change of shape of tumorigenic cen3tel cells.
It is known that the invasiveness of rounded tumor cells depends
on ROCK activity, while it is independent of the activity of matrix
proteases [10,19]. Global gene expression profiling of cen3tel cells
at different PDs showed a decrease in the expression of different
types of extra-cellular matrix proteases during cen3tel propagation
in vitro, particularly of several matrix-metalloproteases (MMPs)
(Table 1), and zymographic analysis showed a decrease in the
activity of MMP2 and MMP9 in tumorigenic cen3tel cells (data
not shown), suggesting that increased motility in late cen3tel cells
was not strictly dependent on pericellular matrix proteolysis.
To test the roles of ROCK and metalloproteases on cen3tel cell
invasiveness, we analyzed the invasive ability of cen3tel cells at the
three stages of tumorigenesis in the presence of the ROCK
inhibitor Y27632 or in the presence of the metalloprotease
inhibitor Ro 28–2653. As shown in Fig. 4B, Y27632 reduced the
invasiveness of cen3tel cells at all three different stages (black
columns), while the metalloprotease inhibitor did not affect it (grey
columns). The opposite result was observed in HT1080 cells,
which use mesenchymal invasion (Fig. 4B). These results indicate
that tumorigenic cen3tel cells acquired a protease-independent
type of invasion, which requires the activity of ROCK kinase.
Rnd3 expression inversely correlates with the invasive
and metastatic properties of tumorigenic cen3tel cell
By microarray analysis and western blotting we did not observe
any evident change in ROCK or RhoA expression in tumorigenic
cen3tel cells (data not shown); however, microarray and western
blot analyses revealed that the cellular inhibitor of ROCK-I, Rnd3
(Riento et al. 2003), was expressed at lower levels in tumorigenic
cen3tel cells compared to non-tumorigenic ones (Fig. 5A, B),
suggesting that Rnd3 downregulation might play a role in the
transition to the ROCK dependent mode of invasion.
To test whether Rnd3 levels affected tumorigenic cen3tel cell
invasiveness, we transfected phase II and phase III tumorigenic
cells with an EGFP-Rnd3-expression vector and isolated clones
with exogenous Rnd3 expression, three from phase II tumorigenic
cen3tel cells and two from phase III cells (Fig. 5C). Compared to
mock-transfected cells, in which the EGFP signal was, as expected,
both nuclear and cytoplasmic (Fig. 5D), in the cells transfected
with the recombinant vector, the EGFP-Rnd3 fusion protein
showed the same subcellular localization described for the
endogenous protein : it was present in the cytoplasm, at the
plasma membrane and showed a perinuclear accumulation
suggestive of a localization in the Golgi apparatus (Fig. 5E).
Rnd3 expressing clones showed proliferation rates similar to that
of parental and mock transfected cells (data not shown).
Rnd3 overexpression induces actin fiber disassembly in several
cell types . In the clones stably expressing Rnd3, we found no
major changes in cellular morphology compared to parental cells,
or actin depolymerization (not shown). This might be related to
the level of expression of the protein in the clones, which may not
be high enough and, above all, to its stable expression. In fact, it
has been shown that in mouse 3T3 fibroblasts, the strong effects
induced by acute Rnd3 expression on the cytoskeleton are only
transient and disappear within hours, even though the levels of the
protein are still high . In the Rnd3 expressing clones we did
find lower level of p-MLC than in parental and mock-transfected
cells, which could be due to an Rnd3-induced reduction in ROCK
kinase activity (Fig. 5F).
We analyzed invasion in the five Rnd3 expressing clones and, in
all of them, it was lower than in mock transfected cells (Fig. 5G),
indicating that Rnd3 expression can modulate invasion of
neoplastically transformed human fibroblasts.
We then investigated whether exogenous Rnd3 expression
reduced the in vivo metastatic potential of phase III tumorigenic
cen3tel cells. We analyzed the metastatic potential of Rnd3 clone 16,
because of the greater stability of Rnd3 expression in this clone
compared to that in clone 8 (not shown). We first inoculated mock-
transfected control cells and Rnd3 clone16 cells subcutaneously in
nude mice, to verify their tumorigenic potential. Cells from both
transfected populations developed tumors; however, after about 10
days, when tumorswerearound 250 mm3,the tumorswererejected,
probably because of an immune response to EGFP expressing cells
. To avoid rejection, we inoculated the cells into SCID (Severe
Combined Immuno-Deficiency) mice, which are deficient in T and
B cells. In these mice, both cell lines developed tumors, which
showed similar growth rates, indicating that Rnd3 does not affect
subcutaneous tumor growth (Fig. 5H). To analyse metastasis
formation, cells were injected intravenously in SCID mice. Mice
were sacrificed after about six weeks; metastatic masses were
detected in the lungs (from 4 to 43 metastases/mouse with a median
number of 9) and in the adrenal glands of the six mice inoculated
with mock-transfected cells (Fig. 5I), while they were not observed in
any of the eight mice inoculated with Rnd3 clone 16 cells.
Taken together, these results indicate a role for Rnd3
downregulation in promoting invasion and metastasis formation
by mesenchymal tumor cells.
The cen3tel cellular system [22,23], derived from human
telomerase immortalized fibroblasts, has allowed us to follow the
Figure 2. Motility and invasiveness of cen3tel cells at different
PDs. A) Motility and (B) invasiveness were assessed in Boyden chambers,
in the absence (spontaneous migration, black columns) or in the
presence (grey columns) of NIH-3T3 supernatant, used as chemoattrac-
tant. To assess invasion, the membrane of the Boyden chambers was
covered with a thick layer of Matrigel. The numbers of migrated cells,
mean and SE from 2–6 independent experiments are shown. Stimulated
migration and invasion of tumorigenic cen3tel cells were significantly
higher (respectively p,0.01 and p,0.05 for migration and invasion) than
that of cen3 fibroblasts or non-tumorigenic cen3tel cells.
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org5 November 2010 | Volume 5 | Issue 11 | e14154
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org6November 2010 | Volume 5 | Issue 11 | e14154
stepwise process of cellular transformation, from normal fibro-
blasts up to tumorigenic cells with metastatic potential. In this
paper, we focused on the study of molecular and biological
features associated with changes in invasiveness during the
progression from normal to tumorigenic and metastatic fibroblasts.
During transformation, cen3tel cells increased their migratory and
invasive potential and gradually acquired the ability to induce
metastases in immunocompromised mice. We show here that the
increased invasiveness does not depend on metalloprotease
activity, as expected for cells of mesenchymal origin, but on the
activity of the ROCK kinase; in addition, we show that Rnd3
negatively regulates invasion and metastasis formation.
Although ROCK-dependent invasion was first described in
carcinoma and melanoma cell lines , evidence has been
recently reported that cells of mesenchymal origin can adopt a
protease-independent amoeboid type of movement, making
ROCK a possible therapeutic target for sarcomas . In mouse
3T3 fibroblasts a switch to amoeboid movement was observed
upon p53 inactivation , while up-regulation of the Rho/
ROCK signalling was found in highly metastatic rat sarcoma cells,
together with the loss of MMP2 activity and an increased
generation of protrusive forces, typical of the amoeboid movement
. Moreover, it has been shown that microtubule destabilization
through stathmin overexpression can also lead to acquisition of
Figure 3. Actin and pMLC cellular distribution in cen3 primary fibroblasts and cen3tel cells. To detect actin, cells were seeded on a
coverslip and incubated with TRITC-labelled phalloidin (A–F, red signal), which binds to F-actin. Tumorigenic cells show a polygonal shape compared
to non-tumorigenic ones and show actin cortical rings. To detect pMLC, indirect immunofluorescence was done with an anti-pMLC primary antibody
and a FITC conjugated secondary antibody (G–L, green signal). In tumorigenic cells, PML is mainly distributed along the inner membrane. (Images
were taken with a confocal microscope, 40x objective, bars 25 mm; single confocal sections are shown).
Figure 4. Morphology in 3D and control of invasion in tumorigenic cen3tel cells. A) Morphology and F-actin organization of tumorigenic
cen3tel cells embedded in Matrigel matrix and effect of the ROCK inhibitor Y27632. Left panels A’–F’: phase contrast images (10x objective, bar
50 mm). Right panels: direct immunofluorescence with TRITC-labelled phalloidin. Iimages were taken using a confocal microscope (40x objective, bar
10 mm); G’–L’: single confocal sections, M’–R’: average of multiple confocal sections. 3D reconstruction of the image in panel O’ is showed in
Supplementary Video 1. B) Effect of the ROCK inhibitor Y27632 and of the MMP inhibitor Ro 28–2653 on invasion of cen3tel cells at different PDs and
of the fibrosarcoma cell line HT1080. Invasion was measured using Boyden chambers. Black columns: Y27632 (10 mM); grey columns: Ro 28–2653
(0.1 mM). The percentages of invasion of control cells (in the absence of inhibitors, white columns) are shown, mean and SE from 2–4 independent
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org7November 2010 | Volume 5 | Issue 11 | e14154
amoeboid movement in sarcoma cells and that the tumor
suppressor protein p27kip1is an important factor for the control
of cellular morphology and motility of transformed fibroblasts by
regulating microtubule stability [38,39].
In cen3tel cells, neoplastic transformation was associated with
an increasing invasive capacity, linked to a change in cell
morphology from elongated to rounded, in the organization of
the actin cytoskeleton from stress fibers to cortical rings, and in the
subcellular distribution of p-MLC. Chemical inhibition of ROCK
led to a reversal of the rounded morphology and to a decrease in
the invasive capacity. Unlike in HT1080 fibrosarcoma cells, we
found that invasion of tumorigenic cen3tel cells was not reduced
upon treatment with a matrix protease inhibitor, indicating that
cells of mesenchymal origin can spontaneously undergo a
transition towards amoeboid movement during transformation.
At the molecular level, this switch could be linked to two main
changes: the decrease in the expression of matrix protease genes
and the reduced expression of the Rho GTP-binding protein
Rnd3. Experimental inhibition of pericellular proteolysis can
induce a transition from mesenchymal to amoeboid movement
; in our cellular system, spontaneous downregulation of matrix
protease genes, through a mechanism still to be clarified, might
have contributed to the switch in the type of movement.
Rnd3, together with Rnd1 and Rnd2, belongs to the Rnd family
of small GTP-binding proteins always bound to GTP, which is
involved in the control of cytoskeletal organization . Micro-
array analysis did not show any significant change in the
expression of Rnd1 and Rnd2 during malignant transformation of
cen3tel cells. Rnd3 is an inhibitor of ROCK-I [32,40] and
evidence has been reported that it can play a role in controlling
morphology and invasion of rounded tumor cells. Pinner and
Sahai  showed that a fine balance between PDK1 and Rnd3
expression is required for the amoeboid movement of melanoma
cells, because the two proteins have opposing roles in controlling
ROCK-I activity. PDK1 drives ROCK-I to the plasma membrane
where it stimulates actomyosin phosphorylation and contraction,
while Rnd3 can bind to ROCK-I and inhibit its activity, reducing
motility. Gadea et al.  showed that overexpression of Rnd3 in
mouse embryonic fibroblasts in which amoeboid migration had
been induced by p53 knockdown led to reduced invasion. Our
observation that ectopic Rnd3 expression in phase II and III
tumorigenic cen3tel cells reduced their invasive capacity points to
Rnd3 as a possible regulator of invasion of cells of mesenchymal
origin. In addition, the low levels of myosin phosphorylation in
Rnd3 transfected clones is in agreement with the hypothesis that
Rnd3 hampers amoeboid movement by inhibiting ROCK kinase
activity. Finally, the low metastatic potential observed in phase III
tumorigenic cen3tel cells that ectopically express Rnd3 suggests
that Rnd3 exerts its effect also in vivo and that its downregulation
can have a pro-metastatic effect.
Rnd3 shows reduced expression in prostate cancer and seems to
have a protective role against breast cancer ; however,
increased Rnd3 expression has been found in other epithelial
tumors and in metastatic melanoma cells , suggesting that the
genetic background of cancer cells can influence the role of Rnd3
in trasformation. Very little is known about Rnd3 expression in
human sarcomas. Our results indicate that reduced Rnd3
expression in tumor cells of mesenchymal origin can be linked
to the acquisition of a ROCK-dependent mode of invasion and
can increase the metastatic potential. Thus, analysis of Rnd3
expression in human sarcomas might give information on the
mechanism of invasion adopted by the tumor cells and help
orientate the therapeutic strategy .
It is worth noticing that the increase in the invasive potential
and Rnd3 downregulation occur when cen3tel cells are tumori-
genic but not yet metastatic. This suggests that the acquisition of
invasive ability by tumor cells, although necessary, is not sufficient
to give them full metastatic ability. Recent studies [43–47] suggest
that even normal cells can disseminate and dissemination from
primary tumors can actually precede the ability to form
metastases; additional genetic or epigenetic changes are then
required for the subsequent production of metastases. It is possible
that cen3tel cells progressively acquired different and complemen-
tary malignant properties besides invasiveness that ultimately led
to a fully metastatic cell population. In agreement with this,
preliminary findings indicate a progressive increase in the
Table 1. Expression of different matrix metalloproteinase (MMP) genes in cen3tel cells at different PDs, relative to parental cen3
*The values are the log2 of the ratio.
1Cen3tel 1034 and cen3tel 1042 represent cells of tumorigenic phase III.
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org8November 2010 | Volume 5 | Issue 11 | e14154
expression of pro-angiogenic factors (such as VEGF and FGF2)
and a concomitant decrease of angiogenesis-inhibitory factors
(particularly TSP-1 and TSP-2) during cen3tel malignant
progression, pointing to angiogenesis as a candidate pro-metastatic
feature acquired by late-phase metastatic cen3tel cells.
In conclusion, the results presented here do suggest that
sarcoma cells can adopt a protease-independent/ROCK-depen-
dent mechanism of invasion and indicate that the Rnd3 gene can
play a role in regulating the invasiveness and metastatic capacity of
mesenchymal tumor cells.
3D reconstruction of a tumorigenic phase III cen3tel cell in Matrigel
in which F-actin was highlighted using TRITC-labelled phalloidin
Tumorigenic cen3tel cells have a rounded morphology.
Figure 5. Rnd3 expression in cen3tel cells and its effect on their invasive and metastatic capacity. A) Results of microarray analysis. Rnd3
expression in cen3tel cells is indicated relative to cen3 primary fibroblasts. The values are the average of the results for ten spots corresponding to the
same probe for Rnd3 (bar: standard error); cen3tel at PD 1034 and at PD 1042 both represent cells of tumorigenic phase III. B) Western blotting
analysis of Rnd3 expression in primary cen3 fibroblasts and cen3tel at different stages of transformation. c-tubulin was used as loading control. C)
Western blotting analysis of recombinant EGFP-Rnd3 expression in transfected clones. C1 and C2 are mock-transfected clones from phase II and III
tumorigenic cen3tel cells, respectively (the endogenous protein and the recombinant one were detected on the same membrane with the anti-Rnd3
antibody, but part of the space between the two signals has been eliminated to make the figure smaller). c-tubulin was used as loading control. D)
Subcellular localization of EGFP (yellow signal) or E) of the EGFP-Rnd3 fusion protein in cells stably transfected with either the empty vector or the
Rnd3 expression vector. The exogenous Rnd3 protein is localized around the plasma membrane and in the cytoplasm where there is a perinuclear
accumulation. Images were taken using an optical microscope (60x objective). F) Western blot analysis of the levels of pMLC in Rnd3-expressing and
mock-transfected clones (C1 and C2). c-tubulin was used as loading control. G) Invasion of clones from phase II (grey bars) and III (black bars)
tumorigenic cen3tel cells transfected with the Rnd3expression vector. Invasion is expressed as the percentage of invasion in cells transfected with the
empty vector (white bars). The average of the results from at least two independent experiments is shown (*: p,0.05; **: p,0.001). H) Growth curves
of tumors obtained in SCID mice after subcutaneous inoculation of phase III tumorigenic cen3tel cells, either mock-transfected (continuous line) or
stably transfected with the Rnd3 expression vector (dashed line). I) Lungs (left) and adrenal gland (right) from SCID mice six weeks after intravenous
inoculation of the mock-transfected phase III tumorigenic cen3tel cells; metastases are visible on both organs. In the box, the median number of
metastases per animal and the range are reported.
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org9 November 2010 | Volume 5 | Issue 11 | e14154
and the nucleus by staining with DAPI. Images of different focal
planes were taken with a confocal microscope (objective 40x).
Found at: doi:10.1371/journal.pone.0014154.s001 (0.05 MB
We are very grateful to Dr. Giovanna Chiorino (Fondazione Edo ed Elvo
Tempia, Biella, Italy) for supervision of microarray analysis, to Dr. Patrizia
Vaghi (University of Pavia, Italy) for confocal microscope analysis and to Dr.
Pierre Roux (CNRS FRE2593, France) for the EGFP-C1-Rnd3 plasmid.
We thank Dr. Valentina Di Felice (University of Palermo, Italy) for
suggestions about immunofluorescence on cells in Matrigel and Ms Giudith
Baggott for help in editing the English style. CB is a PhD student of the
University of Pavia (Dottorato in Scienze Genetiche e Biomolecolari).
Conceived and designed the experiments: CM. Performed the experi-
ments: CB RF KB IC MMG APDT. Analyzed the data: FF. Wrote the
paper: CM. Performed the bulk of the experiments: CB. Helped to draft
the manuscript: CB IC EG GT MD. Carried out the in vivo experiments:
RF. Performed the in vitro migration and invasion assays: KB. Conceived
the transfection experiments: IC. Participated in the experimental work:
IC. Analyzed microarray data: FF MMG. Performed the histological
analysis: APDT. Discussed the results: EG. Participated in designing the
study: GT MD.
1. Weinberg RA (2008) Mechanisms of malignant progression. Carcinogenesis 29:
2. Frame MC, Brunton VG (2002) Advances in Rho-dependent actin regulation
and oncogenic transformation. Curr Opin Genet Dev 12: 36–43.
3. Raftopoulou M, Hall A (2004) Cell migration: Rho GTPases lead the way. Dev
Biol 265: 23–32.
4. Heasman SJ, Ridley AJ (2008) Mammalian Rho GTPases: new insights into
their functions from in vivo studies. Nat Rev Mol Cell Biol 9: 690–701.
5. Friedl P, Wolf K (2003) Tumour-cell invasion and migration: diversity and
escape mechanisms. Nat Rev Cancer 3: 362–374.
6. Sahai E (2005) Mechanisms of cancer cell invasion. Curr Opin Genet Dev 15:
7. Polyak K, Weinberg RA (2009) Transitions between epithelial and mesenchymal
states: acquisition of malignant and stem cell traits. Nat Rev Cancer 9: 265–273.
8. Farina KL, Wyckoff JB, Rivera J, Lee H, Segall JE, et al. (1998) Cell motility of
tumor cells visualized in living intact primary tumors using green fluorescent
protein. Cancer Res 58: 2528–2532.
9. Wyckoff JB, Jones JG, Condeelis JS, Segall JE (2000) A critical step in metastasis:
in vivo analysis of intravasation at the primary tumor. Cancer Res 60:
10. Sahai E, Marshall CJ (2003) Differing modes of tumour cell invasion have
distinct requirements for Rho/ROCK signalling and extracellular proteolysis.
Nat Cell Biol 5: 711–719.
11. Wyckoff JB, Pinner SE, Gschmeissner S, Condeelis JS, Sahai E (2006) ROCK-
and myosin-dependent matrix deformation enables protease-independent
tumor-cell invasion in vivo. Curr Biol 16: 1515–1523.
12. Pankova K, Rosel D, Novotny M, Brabek J. The molecular mechanisms of
transition between mesenchymal and amoeboid invasiveness in tumor cells. Cell
Mol Life Sci 67: 63–71.
13. Friedl P (2004) Prespecification and plasticity: shifting mechanisms of cell
migration. Curr Opin Cell Biol 16: 14–23.
14. Gadea G, de Toledo M, Anguille C, Roux P (2007) Loss of p53 promotes RhoA-
ROCK-dependent cell migration and invasion in 3D matrices. J Cell Biol 178:
15. Sahai E, Garcia-Medina R, Pouyssegur J, Vial E (2007) Smurf1 regulates tumor
cell plasticity and motility through degradation of RhoA leading to localized
inhibition of contractility. J Cell Biol 176: 35–42.
16. Gadea G, Sanz-Moreno V, Self A, Godi A, Marshall CJ (2008) DOCK10-
mediated Cdc42 activation is necessary for amoeboid invasion of melanoma
cells. Curr Biol 18: 1456–1465.
17. Sanz-Moreno V, Gadea G, Ahn J, Paterson H, Marra P, et al. (2008) Rac
activation and inactivation control plasticity of tumor cell movement. Cell 135:
18. Parri M, Taddei ML, Bianchini F, Calorini L, Chiarugi P (2009) EphA2
reexpression prompts invasion of melanoma cells shifting from mesenchymal to
amoeboid-like motility style. Cancer Res 69: 2072–2081.
19. Wolf K, Mazo I, Leung H, Engelke K, von Andrian UH, et al. (2003)
Compensation mechanism in tumor cell migration: mesenchymal-amoeboid
transition after blocking of pericellular proteolysis. J Cell Biol 160: 267–277.
20. Carragher NO, Walker SM, Scott Carragher LA, Harris F, Sawyer TK, et al.
(2006) Calpain 2 and Src dependence distinguishes mesenchymal and amoeboid
modes of tumour cell invasion: a link to integrin function. Oncogene 25:
21. Rosel D, Brabek J, Tolde O, Mierke CT, Zitterbart DP, et al. (2008) Up-
regulation of Rho/ROCK signaling in sarcoma cells drives invasion and
increased generation of protrusive forces. Mol Cancer Res 6: 1410–1420.
22. Mondello C, Chiesa M, Rebuzzini P, Zongaro S, Verri A, et al. (2003)
Karyotype instability and anchorage-independent growth in telomerase-
immortalized fibroblasts from two centenarian individuals. Biochem Biophys
Res Commun 308: 914–921.
23. Zongaro S, de Stanchina E, Colombo T, D’Incalci M, Giulotto E, et al. (2005)
Stepwise neoplastic transformation of a telomerase immortalized fibroblast cell
line. Cancer Res 65: 11411–11418.
24. Maga G, Crespan E, Wimmer U, van Loon B, Amoroso A, et al. (2008)
Replication protein A and proliferating cell nuclear antigen coordinate DNA
polymerase selection in 8-oxo-guanine repair. Proc Natl Acad Sci U S A 105:
25. Chardin P (2006) Function and regulation of Rnd proteins. Nat Rev Mol Cell
Biol 7: 54–62.
26. Taraboletti G, Giavazzi R (2004) Modelling approaches for angiogenesis.
Eur J Cancer 40: 881–889.
27. Deaglio S, Vaisitti T, Aydin S, Bergui L, D’Arena G, et al. (2007) CD38 and
ZAP-70 are functionally linked and mark CLL cells with high migratory
potential. Blood 110: 4012–4021.
28. Smyth G (2005) Limma: linear models for microarray data. In: Gentleman R,
Carey CV, Dudoit S, Irizarry R, Huber W, eds. Bio-informatics and
Computational Biology Solutions using R and Bioconductor. New York:
Springer. pp 397–420.
29. Smyth GK (2004) Linear models and empirical bayes methods for assessing
differential expression in microarray experiments. Stat Appl Genet Mol Biol 3:
30. Benjamini Y, Hochberg,Y (1995) Controlling the False Discovery Rate: A
Practical and Powerful Approach to Multiple Testing. Journal of the Royal
Statistical Society Series B (Methodological) 57: 289–300.
31. Haldar M, Randall RL, Capecchi MR (2008) Synovial sarcoma: from genetics
to genetic-based animal modeling. Clin Orthop Relat Res 466: 2156–2167.
32. Riento K, Ridley AJ (2003) Rocks: multifunctional kinases in cell behaviour. Nat
Rev Mol Cell Biol 4: 446–456.
33. Riento K, Totty N, Villalonga P, Garg R, Guasch R, et al. (2005) RhoE function
is regulated by ROCK I-mediated phosphorylation. Embo J 24: 1170–1180.
34. Guasch RM, Scambler P, Jones GE, Ridley AJ (1998) RhoE regulates actin
cytoskeleton organization and cell migration. Mol Cell Biol 18: 4761–4771.
35. Villalonga P, Guasch RM, Riento K, Ridley AJ (2004) RhoE inhibits cell cycle
progression and Ras-induced transformation. Mol Cell Biol 24: 7829–7840.
36. Steinbauer M, Guba M, Cernaianu G, Kohl G, Cetto M, et al. (2003) GFP-
transfected tumor cells are useful in examining early metastasis in vivo, but
immune reaction precludes long-term tumor development studies in immuno-
competent mice. Clin Exp Metastasis 20: 135–141.
37. Micuda S, Rosel D, Ryska A, Brabek J. ROCK inhibitors as emerging
therapeutic candidates for sarcomas. Curr Cancer Drug Targets 10: 127–134.
38. Belletti B, Pellizzari I, Berton S, Fabris L, Wolf K, et al. p27kip1 controls cell
morphology and motility by regulating microtubule-dependent lipid raft
recycling. Mol Cell Biol 30: 2229–2240.
39. Belletti B, Nicoloso MS, Schiappacassi M, Berton S, Lovat F, et al. (2008)
Stathmin activity influences sarcoma cell shape, motility, and metastatic
potential. Mol Biol Cell 19: 2003–2013.
40. Riento K, Villalonga P, Garg R, Ridley A (2005) Function and regulation of
RhoE. Biochem Soc Trans 33: 649–651.
41. Pinner S, Sahai E (2008) PDK1 regulates cancer cell motility by antagonising
inhibition of ROCK1 by RhoE. Nat Cell Biol 10: 127–137.
42. Klein RM, Aplin AE (2009) Rnd3 regulation of the actin cytoskeleton promotes
melanoma migration and invasive outgrowth in three dimensions. Cancer Res
43. Husemann Y, Geigl JB, Schubert F, Musiani P, Meyer M, et al. (2008) Systemic
spread is an early step in breast cancer. Cancer Cell 13: 58–68.
44. Molloy TJ, Bosma AJ, van’t Veer LJ (2008) Towards an optimized platform for
the detection, enrichment, and semi-quantitation circulating tumor cells. Breast
Cancer Res Treat 112: 297–307.
45. Podsypanina K, Du YC, Jechlinger M, Beverly LJ, Hambardzumyan D, et al.
(2008) Seeding and propagation of untransformed mouse mammary cells in the
lung. Science 321: 1841–1844.
46. Weinberg RA (2008) Leaving home early: reexamination of the canonical
models of tumor progression. Cancer Cell 14: 283–284.
47. Weinberg RA (2008) The many faces of tumor dormancy. Apmis 116: 548–551.
Rnd3 and Tumor Cell Invasion
PLoS ONE | www.plosone.org 10November 2010 | Volume 5 | Issue 11 | e14154