Invading Basement Membrane Matrix Is Sufficient for MDA-MB-231 Breast Cancer Cells to Develop a Stable In Vivo Metastatic Phenotype

Introduction The poor efficacy of various anti-cancer treatments against metastatic cells has focused attention on the role of tumor microenvironment in cancer progression. To understand the contribution of the extracellular matrix (ECM) environment to this phenomenon, we isolated ECM surrogate invading cell populations from MDA-MB-231 breast cancer cells and studied their genotype and malignant phenotype. Methods We isolated invasive subpopulations (INV) from non invasive populations (REF) using a 2D-Matrigel assay, a surrogate of basal membrane passage. INV and REF populations were investigated by microarray assay and for their capacities to adhere, invade and transmigrate in vitro, and to form metastases in nude mice. Results REF and INV subpopulations were stable in culture and present different transcriptome profiles. INV cells were characterized by reduced expression of cell adhesion and cell-cell junction genes (44% of down regulated genes) and by a gain in expression of anti-apoptotic and pro-angiogenic gene sets. In line with this observation, in vitro INV cells showed reduced adhesion and increased motility through endothelial monolayers and fibronectin. When injected into the circulation, INV cells induced metastases formation, and reduced injected mice survival by up to 80% as compared to REF cells. In nude mice, INV xenografts grew rapidly inducing vessel formation and displaying resistance to apoptosis. Conclusion Our findings reveal that the in vitro ECM microenvironment per se was sufficient to select for tumor cells with a stable metastatic phenotype in vivo characterized by loss of adhesion molecules expression and induction of pro-angiogenic and survival factors.


Introduction
Metastasis in breast cancer patients accounts for over 90% of the deaths. Preclinical studies reveal that many drugs used in the management of primary tumors are not or less effective against metastasis [1]. Although the mechanism by which metastases develop is still not fully understood, it is generally believed that tumor cells acquire features that affect their metastatic potential during the progression of the tumor; these features include increased survival, invasive and migratory abilities. Metastasis is a complex cascade of sequential steps, none of which being fully understood. Many studies implicated the stroma in the development of metastases. Stroma and cancer cell interactions were found to contribute to cell detachment from primary tumors, intravasation into the blood stream, and extravasation at distant sites where tumor cells can seed and form tumor metastases [2]. Previously, fibroblasts, endothelial cells and macrophages and other stroma cells were reported to be implicated in the occurrence of metastases [2][3][4]. Whereas less is known about the influence of the extracellular matrix (ECM) in the development of metastases [5,2], ECM appears to be involved, particularly through sequestration of soluble factors secreted by stroma cells and acting on tumor cells in a paracrine manner [6][7][8][9]. The ECM can modify the adhesion of cancer cells leading to their invasive behavior. However, it is not clear if the ECM, and particularly the basement membrane, can influence the selection of metastatic cells with a well-defined genotype profile. The basement membrane underlies the endothelium in the vessel wall and is the major barrier to tumor cell extravasation and invasion [10][11]. Matrigel is a basement membrane extract derived from a murine tumor [12]. The components of this tumor basement membrane are similar, both chemically and immunogically, to authentic basement membrane components [13]. Consequently, Matrigel may be used as an experimental model of barrier to identify both inhibitors and activators of invasion [14].
Most of previously reported studies on gene profiling in malignant cells used tumor cells isolated from metastatic foci that were already well established in vivo. Therefore, the contribution of ECM to the selection of metastatic cells has not been rigorously assessed. Here, we report the in vitro selection of two cell populations derived from MDA-MB-231 breast cancer cells [15], on the basis of their high (INV) invasive ability (to migrate through Matrigel). We examined the transcriptomes of these two populations using micro arrays and correlated their gene profiles to their malignant phenotypes.

Cell culture
Human breast adenocarcinoma MDA-MB-231 cells were obtained from the American Type Culture Collection (Manassas, VA, USA). The INV and REF cells isolated as described underneath were maintained in 10% FBS-DMEM, 1% sodium pyruvate with 1% penicillin and 1% streptomycin at 37uC in a humidified atmosphere containing 5% carbon dioxide. Human microcapillary brain endothelial hCMEC/D3 cells [16,17] were grown in collagen type I-coated dishes in EBM-2 media supplemented with 2.5% FBS, hydrocortisone and growth factors (VEGF, IGF, EGF and bFGF EGM-2 Bullet kit, Cambrex-Biowhittaker).

Isolation of MDA-MB-231 derived INV and REF variants
MDA-MB-231 cells (2.5610 5 ) were seeded in 8 mm-pore size Boyden chambers (Becton Dickinson) coated with Matrigel (Falcon, MA, USA) and incubated for 16 h at 37uC in a 5% CO2 atmosphere. Invading cells, referred to as INV cells, were collected from the bottom of the membrane in 0.5 mM EDTA and cultured in a 24-well plate. The selection was repeated eight times sequentially in order to obtain highly invasive cells. To select non-invasive cells (REF), the same procedure was used and the cells that remained on top of the membrane after the first round of selection were collected.

Cell growth, migration and invasion assay
For cell growth assays, INV and REF cells were seeded at 2610 5 cells per well in 6-well plates and cultured for 72 h. Cell migration experiments were performed using the Boyden chambers as described above: the cells (2.5610 5 ) were put into the upper chamber and 10% FCS-DMEM was used as a chemoattractant. For invasion assays, chambers were coated with Matrigel as described above. Then, cells on the lower face of the filter were fixed, stained with crystal violet and counted under the optical microscope.

Adhesion and transmigration assays
Tumor cells were harvested in 0.5 mM EDTA and seeded at 5610 5 cells per well in 96-well plates pre-coated with 1 mg/ml of fibronectin (Sigma, France) and incubated at 37uC for 1 h. The plates were washed with PBS, and adherent cells were stained with crystal violet/methanol (0.5%/20%) for 1 h at room temperature. The plates were again washed, and were incubated in ethanol: 0.1 M citrate (1:1) pH 4.2 for 30 min at room temperature and absorbance was measured at 540 nm using an ELISA reader (Labsystem, GMI Inc, Minnesota, USA). To study tumor cell adhesion to endothelial cell monolayers, hCMEC/D3 cells were seeded at 10 4 cells per well on collagen type I-coated 96-wells plates and left to grow until confluence. The cells were then starved for 48 h and incubated for an additional 24 h with the proinflammatory cytokines IFNc (100 U/ml-Preprotech Inc, NJ, USA.) and TNFá (200 U/ml-Preprotech Inc.). Next, the INV and REF cells at 80% confluence were harvested using PBS-EDTA and labeled for 30 min at 37uC with CMFDA, a green fluorescent cell tracker (Invitrogen, Cergy Pontoise, France). Then, 10 5 INV or REV cells were added to the endothelial cell monolayer and incubated for 30 min at 37uC. The unattached cells were removed and adherent cells were lysed with water. Fluorescence was measured at 485/530 nm using a microplate reader (Labsystem, France). For transmigration assays, Boyden chambers were coated with hCMEC/D3 endothelial cells and CMFDA-labeled tumor cells were counted by fluorescent microscopy.

Gelatine zymography
INV and REF cells were cultured in serum-free media for 24 h in 6-well plates. Conditioned media were then lyophilized and normalized to the number of cells. They were mixed with nonreducing Laemmli buffer and run on a 10% SDS-polyacrylamide gel containing 1 mg/ml gelatin. The gel was washed three times at room temperature in a solution containing 2.5% (v/v) Triton X-100 and incubated at 37uC for 24 h in 50 mM Tris/HCL, pH 7.5, 0.2 M NaCl, 5 mM CaCl 2 and 0.05% Brij 35. The gel was then stained for 60 min with 0.25% (w/v) R-250 Coomassie blue in 45% (v/v) methanol/10% (v/v) acetic acid and destained in 25% methanol (v/v)/10% acetic acid (v/v).

Cell death detection
Cells (2610 5 ) were cultured for 96 h in a serum-free medium or with 2.5 mM Doxorubicin for 24 h in a serum-containing medium. Then, cells were harvested and apoptotic cells were labeled using the Annexin V-FITC Apoptosis Detection kit (Beckman coulter, Fullerton CA, USA). The samples were then analyzed by flow cytometry (Becton Dickinson, Heidelberg, Germany). For detection of cell death in vivo, tumor sections (5 mm) were deparaffinized and rehydrated, then analyzed for DNA fragmentation using a TumorTACS kit (R&D Systems, Abington, UK). Various fields (10 per section) were selected for analysis and the intratumor aponecrotic cells were counted as previously described [18]. For AKT phosphorylation assay, cells were treated for 1 h or 72 h with serum free medium and stimulation with IGF-1 at 10 ng/ml for 30 min was used as positive control.

Western blots
INV and REF cell lysates were clarified by centrifugation, and protein concentrations were measured by bicinchoninic acid protein assay (Bio-Rad, Marne-La-Coquette, France). Equal amounts of total proteins were resolved by 6% SDS-PAGE and transferred to nitrocellulose membrane. The membrane was blocked in 5% nonfat dried milk in Tris-buffered saline containing 1% Tween 20 and incubated with primary antibodies against NRP-1 (Santa-Cruz Biotechnologies inc., CA) or ß-actin (Sigma, Saint quentin Fallavier, France), AKT or pAKT (Cell signaling technology, Boston, USA) for 1 h, and then with an appropriate conjugated secondary antibody for 30 min at room temperature. Immunoreactive bands were detected by chemiluminescence according to the manufacturer's instructions.

VEGF quantification
INV or REF cells seeded in 24-well culture plates (2610 5 /well) in DMEM-10% FCS were allowed to adhere for 24 h. Conditioned media containing growth factors secreted by the cells were collected after 48 h of serum starvation. The VEGF content in these media was measured by quantitative enzymelinked immunoabsorbent sandwich assay (ELISA) using the Quantikine kit (R&D Systems, Abington, UK) according to the manufacturer's instructions. VEGF contents were estimated from a standard curve established from serial dilutions of VEGF protein standards (R&D Systems). The values reported for VEGF secretion are normalized to cell number. Reported results are normalized to the expression of the PPIA, GAPDH, TBP, beta 2 Microglobin, Actin, RPL0 and RPL19 genes that exhibit little variation in this data set. Similar results were obtained for all reference genes, the representative data is normalized to PPIA. Fold induction relative to internal controls was calculated by the delta Ct evaluation method [19].

Gene Expression Profiling
Total RNA was isolated from eight cell cultures (four biological replicates corresponding to four different passages of both INV and REF cells) using the RNeasy mini kit (Qiagen, Valencia, CA, Germany) by direct cell lysis in 10 cm culture dishes containing 600 ml RLT Buffer following the manufacturer's instructions. Purified RNA was quantified with a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Hudson, NH, USA) and its quality was evaluated by running the Eukaryotic Total RNA Nano Assay in the Bioanalyser 2100 (Agilent Technologies, Chandler AZ, USA) using the Agilent RNA 6000 Nano Kit. Probes (300 ng of the total RNAs) were prepared for hybridization to Affymetrix GeneChipH Human Gene 1.0 ST arrays according to the manufacturer's instructions. After scanning, eight CEL files generated by GCOS (Affymetrix GeneChipH Operating Software) were uploaded into Expression Console Software (Affymetrix, Santa Clara, CA, USA) for normalization and probe set summarization using the RMA (Robust Multichip Analysis) algorithm [20]. The resulting data in terms of probe set intensities were exported into a .txt file for further statistical analysis. The microarray data has been deposited in the Gene Expression Omnibus (GEO) database at http://www. ncbi.nlm.nih.gov/geo/ with the accession number GSE16838.

Microarray data analysis
Analyses were carried out in R, a free software environment available at http://www.r.-project.org/. To identify differentially expressed genes, differences between the two conditions were calculated on each day of measurement and then analyzed using a double-sided paired t-test implemented in the anapuce R package (http://www.agroparistech.fr/mia/outil.html). Pairing was carried out to remove any source of variation due to the day of measurement from the comparison of the two conditions. The variance was split between subgroups of genes with homogeneous variance [21]. Statistical raw p-values were adjusted for multiple comparisons using the Benjamini-Hochberg procedure [22] which controls the False Discovery Rate (FDR) after the estimation of the proportion of genes non-differentially expressed using the smoother method [23]. The level of statistical significance was set at 0.05 for all comparisons. Gene annotations and functions were assessed using NetAffx TM Analysis Center (Affymetrix, Santa Clara, CA, USA), Ingenuity pathway Analysis (IPA) software (Ingenuity SystemsInC.) and data mining.

Xenografts in nude mice
All in vivo experiments were carried out with the approval of Département d'Expérimentation Animale, Institut Universitaire d'Hématologie, Hôpital Saint-Louis ethical committee (permit number 75516) and according to the UKCCCR guidelines. The animals were kept in a temperature-controlled room on a 12: 12 light-dark schedule with food and water ad libidum. Aliquots of 2610 6 of INV (n = 7) or REF cells (n = 7) were inoculated subcutaneously (s.c) near the mammary gland of 4-week-old athymic nude mice (nu/nu, Janvier, France, n = 14). Tumor volume was measured once a week along two major axes using calipers. Tumor volume (mm 3 ) was calculated as follows: V = 4/3pR 1 2 R 2 where R 1 ,R 2

Immunohistochemical analysis
Tumor specimens were fixed with 4% paraformaldehyde and embedded in paraffin using standard procedures. For immunohistochemical studies, 5 mm-thick sections were cut, deparaffinized and rehydrated. Routinely, sections were stained in haematoxylin and eosin. Endogenous peroxidase was inactivated with 3% H 2 O 2 and washed in TBS (0.05 M Tris, 1.5 M NaCl, pH 7.6) followed by preincubation in 10% normal goat serum for 1 h at room temperature. Endothelial cells were specifically labeled with GSL-1 isolectin B4 (Vector Laboratories, Burlingame, CA, U.S.A.) as described previously [24]. For each tumor, 10 randomly selected sections were analyzed.

Intracardiac experimental metastasis assay
We performed mice intracardiac injection of bioluminescent INV and REF cells as also described previously in order to observe large dissemination behavior of each subpopulation [25]. For this study, luciferase-expressing INV and REF cell pools were obtained by stable transfection with a plasmid carrying a G418 resistance gene and the luciferase gene (pEGFPLuc, BD Biosciences, Rungis, France) using Fugen-6 (Roche Diagnostics, Meylan, France). Transfected cells were cultured in 10% FCS-DMEM supplemented with G418 at 1 mg/ml. For experimental metastasis, female nude mice (8-10 weeks) were anesthetized by intraperitoneal injection with 120 mg/kg ketamine and 6 mg/kg xylazine on the day of cell injection or by exposure to 1-3% isoflurane on subsequent imaging days. pEGFPLuc-transfected INV or REF cells (10 5 ) in 100 ml of sterile DPBS were injected into the left ventricle of the heart by nonsurgical means.
Anesthetized mice were given intraperitoneal injections of Dluciferin (Caliper Life Science) and placed in the IVIS TM Imaging System (Xenogen, Caliper Life Science, MA, USA) and imaged from both dorsal and ventral views. Successful intracardiac injection was revealed on day 0 by systemic bioluminescence distributed throughout the whole animal. Only mice with satisfactory injection were used for subsequent experiments. Metastasis analysis was monitored by in vivo imaging once a week for up to eight weeks.

Bioluminescent Imaging
Bioluminescence images were acquired with the IVIS imaging system (Xenogen) 5 minutes after intraperitoneal injection of Dluciferin into the anesthetized animals. Acquisition time at the beginning of experiment was 5 min and was reduced in keeping with signal strength to avoid saturation. Analysis was performed using LivinImage software (Xenogen) by measurement of photon flux (photon/s/cm 2 ) within a region of interest (ROI) drawn around the bioluminescence signal to detect the metastatic sites.

Statistical analysis
Student's t-test was used for statistical analysis. P,0.05 was considered significant.

Selection of INV subpopulation from MDA-MB-231
Using Matrigel coated Boyden chambers, we established two different sub-populations of MDA-MB-231 cells. These included the invasive MDA-MB-231 subpopulation (INV) that was isolated from a less invasive subpopulation (REF) following eight rounds of invasion assays in the Boyden Chamber. Their invasive phenotype was confirmed by migration and invasion assays (Fig. 1). REF cells were expanded and frozen after the first round. As expected, INV cells were up to 4-fold more invasive than REF cells (Fig. 1A and right panel), and up to 3-fold more motile (Fig. 1B and right  panel). These properties were stable and not lost after maintenance in culture or freezing. To assess the invasive capacity of INV and REF cells in more detail, the corresponding conditioned media were collected and analyzed for gelatinase enzymatic activity (Fig. 1C). MMP-9 activity was higher in media conditioned by INV cells as compared to REF cells. In contrast, MMP-2 activity remained unchanged (data not shown). Similarly, the mRNA expression of uPA, uPAR and PAI-1 were found significantly up-regulated in INV cells as evaluated by real time PCR (Fig. 1D).

Transcriptomic analysis of INV and REF cells
We used the Affymetrix Human Gene 1.0ST arrays that contain 28000 annotated genes (Table 1 and (Table 1), 44% are involved in cell adhesion and cell-cell junctions, and 22% in induction of apoptosis and G-protein coupled receptor signaling (Fig. S1). Of the upregulated genes in INV ( Table 2) with known function, 25% are involved in survival via negative regulation of apoptosis, 17.5% in cytoskeleton reorganization, cell proliferation and chemotaxis, 12.5% in cell adhesion and cell-cell junctions and 10% in Gprotein coupled receptor signaling (Fig. S1).   We validated the array results by real-time PCR for selected genes (indicated with asterisks in tables 1 and 2).

Adhesion and transmigration abilities of INV and REF cells
The main differences between the transcriptome profiles of the two cell populations concerned molecules involved in cell-ECM and cell-cell junctions. Therefore, we tested INV and REF cells for adhesion and transmigration through endothelial cells. As observed in Fig. 2A,

the morphology of INV cells in monolayer cultures differed from that of REF cells. INV but not REF monolayers cells seemed to present gaps between cells. In addition, REF cells were round whereas INV cells were fusiform.
We next tested adhesion of INV and REF cells to fibronectin and hCMEC/D3 brain microcapillary endothelial cells (Fig. 2 B and C, respectively). Consistent with findings for the transcriptomes, REF cells adhered to fibronectin whereas INV cells did not (the optical density was the same as for negative controls, equivalent to fold induction of 1) (Fig. 2B). Although both cell populations adhered to the endothelium formed by hCMEC/D3 human brain microcapillary endothelial cells, INV cells again showed a lower adhesion capacity than REF cells both in the presence or absence of cytokines (Fig. 2C) (proinflammatory cytokines have been reported to activate endothelial cells and induce tumor cell attachment to the endothelium [26]).
Next, we compared the ability of INV and REF cells to transmigrate through hCMEC/D3 endothelial cells (Fig. 2D)

Increased metastatic sites colonization and reduced survival in INV cell-inoculated mice
In order to study the subclones metastasis behaviour in the whole body, Luc-transfected subclones of INV and REF (1610 5 ) were injected in the circulation (intracardiac route as described previously, [25]) of nude mice (n = 9) and their abilities to colonize secondary sites were evaluated weekly during eight weeks (Fig. 3A). Metastases were detected after two and three weeks in mice injected with INV and REF cells, respectively. Five weeks after injection, multiple metastatic sites were detected in mice inoculated with INV cells (Fig. 3A and C). In general, in each mouse inoculated with REF cells, one metastatic site was detected, mainly in bone rather if only one mouse developed lung metastasis which lead to rapid death at day 37 (data not shown. In addition, ex vivo imaging of various tissues after the final in vivo imaging mainly revealed the presence of lesions in bones for REF cells injections (legs and skull) and brain, lymph nodes and ovary for INV ones (Fig. 3B). Ovarian lesions were only found in INV cellinjected mice (4 of 9 mice) and most smooth tissue lesions were also in mice injected with INV cells. The number of metastatic sites was 2 to 3 fold higher in the INV group than in the REF group of mice (Fig. 3C). INV cell-injected mice died earlier than REF cell-injected mice; only 20% of INV as compared to 80% of the REF group survived to the end of the experiment (Fig. 3D).  near the mammary gland of nude mice (Fig. 4). Both cell populations caused the development of 30 mm 3 tumors after one week. After 21 days, the volumes of INV tumors were 2.6 fold higher than those of REF tumors (172 mm 3 vs 67 mm 3 , respectively, Fig. 4A). Surprisingly, in vitro proliferation of INV cells was half that of REF cells after 72 h of cell culture in FCS10%/DMEM (data not shown). Thus, the faster growth of INV tumors was unlikely to have been due to a higher proliferation rate. We therefore evaluated cell survival within REF and INV tumors by in situ DNA fragmentation labeling with biotinylated nucleotides. This technique results in brown staining of apoptotic cell nuclei and diffuse brown staining in the cytoplasm and nuclei of necrotic cells (Fig. 4B). The numbers of dead cells were lower in INV tumors than REF tumors (15% versus 22%, P = 0.046; Fig. 4C). In vitro after 24 h of treatment with 2.5 mM doxorubicin (Fig. 5), 90% of REF and 59% of INV cells had undergone apoptosis as revealed by Annexin-V labeling (Fig. 5A). Similarly, following 96 h of serum deprivation, 40% of REF cells and only 9.5% of INV cells were found to be apoptotic (Fig. 5B).

Increased angiogenesis in INV xenografts
We next assessed angiogenesis in both INV and REF tumors using GSL-1 lectin to labelmouse endothelial cells (Fig. 6). Both tumor types exhibited vascular areas (Fig. 6A). The number of endothelial cells per area was significantly higher in INV than REF tumors (132 versus 94, respectively, p,0.05, Fig. 6B). Analysis of genes involved in angiogenesis including those encoding VEGF and their receptors [31][32][33][34][35] by qRT-PCR revealed stronger VEGF, VEGFR2 and NRP-1 expression in INV compared to REF cells (Fig. 6C). The over-production of VEGF and NRP-1 proteins was confirmed by ELISA tests for VEGF (Fig. 6D) and western blot analysis for NRP-1 (Fig. 6E).

Discussion
We report that breast cancer cell invasion through Matrigel resulted in the generation of an aggressive subpopulation (INV) with a stable malignant genotypic/phenotypic profile. The main characteristic of these cells was loss of adhesion to ECM, and in particular fibronectin and endothelial cells. Accordingly, in nude mice, INV cells were more competent than the control cells (REF) for extravasation through the endothelium to colonise tissues and form metastases. Metastases formed by INV cells were mainly  found in various soft tissue (brain, ovary, lymph node and bone) as compared to REF ones (Exclusively in bones). This result was in accordance with previous work of Jenkins et al [36] which demonstrated that dissemination in other sites revealed a hallmark of increased metastatic potential. This indicates that the ECM is sufficient to influence the enrichment of a stable metastatic subpopulation. We also found that INV cells failed to induce significant platelet adhesion and/or aggregation (data not shown), a process previously reported to be involved in metastasis processes [37]. Previously, cell migration was reported to result from a balance between cell adhesion and de-adhesion, both acquired for motility. Indeed, strong adhesion would tightly attach cancer cells to the substrate and prevent their movement. Reduced cell adhesion to both ECM and endothelial cells, observed for our INV subpopulation, is consistent with their enhanced migration ability, and their capacity to cross the endothelial barrier to invade secondary organs and form metastases. Thus, we found that the cells obtained by transmigration through an extracellular matrix displayed a loss of adhesion behavior. Studies on breast cancer progression indicate that loss of cadherin or cell polarity and the acquisition of a mesenchymal phenotype are associated with an increase in invasiveness [38]. Accordingly, our INV cells, the invasive subpopulation, had a fusiform shape with large spaces between cells and thus resembled mesenchymal cells. Also, we found weak expression of OCLN and strong expression of TGFb2, an association which was previously demonstrated to be linked to the breast cancer epithelial mesenchymal transition (EMT) [39]. Thus, our results support the notion that the basement membrane may also contribute to the EMT leading to plasticity of metastatic cells.
Using GSEA MsigDB [40] data bases, we found that the functions of 38% of down-regulated and 47% of up-regulated genes in INV are known (see also  apoptosis. INV cells also over-expressed the proangiogenic factor (VEGF) and its receptor NRP-1, both involved in the autocrine survival of cells and microvessel permeability, processes required for cell extravasation [41]. Other genes including IRAK1, NKAP, FLNA, XIAP and TGFß2 were also up-regulated: XIAP and TGFb2 are associated with the NFKB pathways possibly being pivotal potential signaling targets [42,43]. To note, also the implication of UPA, UPAR and PAR-I in the NFKB pathway and that acted in concert with MMP-9 to enhance INV cell invasion. Thus, further studies on the NFKB signaling in these cell would seem interesting to undergo. Some pro-apoptotic genes were down-regulated in INV cells. However, important pro-apoptotic gene as for example, BCL2 was not found to be upregulated, (Fig. S2). These include rather CASP4 and cyclin D2, previously reported to be involved in genistein and doxorubicin sensitization, respectively [44,30]. The down-regulation of cyclin D2 in INV cells suggests neosis, an important process that mediates tumor cell resistance to apoptosis, heterogeneity and continuity [45]. These findings are in agreement with reports of the critical role of survival during the metastatic process of tumor cells [46]. Further, AKT analysis showed no differences between the INV and REF cells, suggesting that several survival genes mentioned above acted in concert to increase resistance to death. Interestingly, in vivo, we only observed small differences in cell death between INV cell-derived tumors and REF tumors, suggesting that the observed difference in tumor volumes might result from a dormancy exit. Indeed, several reports reveal the participation of angiogenesis in the exit from dormancy by inhibiting apoptosis [47]. The in vitro identification of anti-apoptotic or survival targets in this work constitutes an interesting and original finding that could lead to better treat aggressive sub populations. In particular, it could be a good strategy to prevent cancer relapse and to better investigate tumor dormancy in term of therapy as well as diagnosis. In basis of these first results, further investigation should address preclinal models to indentify targets in order to treat and also survey long-term cancer survivors minimizing the risk of cancer recurrence.

Conclusions
We demonstrated that using only basement membrane extracellular matrix it is possible to obtain a stable metastatic population without using expensive and longer in vivo procedures. Our results suggest that future strategies for anti-metastatic therapies should target molecules involved in angiogenesis and survival rather than in cell adhesion. We report here the genetic, phenotypic and functional characterization of invasive (INV) cancer cell population. We believe that such cells will provide unique cellular models for in vitro screening of anti-metastatic drugs with clinical potential.