Epithelial–mesenchymal transition (EMT) is a process that results in the loss of epithelial properties, such as cell–cell adhesion and baso-apical polarity, and the gain of mesenchymal properties, such as increased ability to migrate on and invade through extracellular matrix proteins (Thiery and Sleeman,2006). EMT is an essential, highly regulated process during early stages of embryonic development in the formation of the primitive streak (Locascio and Nieto,2001). However, inadvertent, unregulated EMT is associated with many types of fibrotic diseases and with oncogenic progression (Thiery and Sleeman,2006). EMT is associated with the loss of expression and/or mis-localization of proteins involved in the formation of cell–cell junctions, such as E-cadherin, zonula occludens (ZO-1) and claudin, and the gain of expression of mesenchymal proteins such as fibroblast specific protein 1 (FSP-1), smooth muscle actin and fibronectin, as well as certain integrins. EMT is also associated with a switch in the expression of intermediate filaments from expression of keratins to expression of vimentin (Thiery and Sleeman,2006).
Loss of E-cadherin expression is a frequent event during cancer progression (Thiery,2002), and many solid cancers such as prostate cancer, pancreatic cancer, renal cancer, gastric cancers, and certain types of breast cancers (e.g., lobular carcinoma) are associated with a loss or down-regulation of E-cadherin expression (Thiery,2002; Thompson et al.,2005). Loss of E-cadherin can be attributed to gene mutations, or to transcriptional down-regulation due to gene methylation, or to inhibition of gene transcription stimulated by activation of certain cellular signaling pathways that result in the activation of repressors of E-cadherin gene transcription, such as Snail 1, Snail 2, and Twist (Batlle et al.,2000; Nieto,2002; Thompson et al.,2005).
EMT has been implicated as being an important event during metastasis (Thiery,2002; Thompson et al.,2005), tumor recurrence (Moody et al.,2005), as well as resistance to growth factor receptor inhibitors, especially epidermal growth receptor (EGFR) inhibitors (Thompson et al.,2005; Yauch et al.,2005). EMT is often considered to be the switch that allows benign, noninvasive and nonmetastatic tumor cells to acquire the capacity to invade and infiltrate surrounding tissue and to ultimately metastasize to distant sites (Thiery,2002). Multiple EMT regulators can enhance tumor progression and metastasis in experimental models. Although EMT can be documented in cell culture as well as in some in vivo experimental models, and proteomic analysis of breast cancer has revealed that circulating mammary tumor cells, or those found in micro-metastases, show evidence of mesenchymal conversion, there is considerable controversy as to whether the conversion of a noninvasive tumor into a metastatic tumor involves EMT (Tarin et al.,2005; Thompson et al.,2005; Lee et al.,2006). This skepticism stems from the apparent rarity of EMT-like morphological changes that are observed in primary tumor sections, and also from the observation that metastases appear histologically similar to the primary tumor from which they are derived. Although elegant transgenic models have shown that EMT is essential for malignant progression and metastasis formation (Perl et al.,1998), the requirement of EMT for spontaneous metastasis of breast cancer has not been assessed.
The 4T1 mouse mammary tumor model is a highly clinically relevant model of spontaneous breast cancer metastasis to multiple sites (Lelekakis et al.,1999). Several syngeneic tumor lines with a spectrum of metastatic phenotypes have been isolated from a spontaneous mammary tumor in a BALB/cfC3H mouse. When injected into the mammary gland of mice, these tumor lines are either nonmetastatic, or produce spontaneous metastases to various sites with varying frequencies (Aslakson et al.,1991; Aslakson and Miller,1992). This model is, therefore, a highly relevant one in which to examine the EMT properties of the various cell lines with differing metastatic properties in vivo (Eckhardt et al.,2005). This model was used to identify genes that influence metastasis (Yang et al.,2004) and resulted in the identification of Twist, a repressor of E-cadherin expression, as a promoter of EMT and metastasis. However, we find little correlation between EMT and the ability to metastasize in this model, and further that there is a large variation in the expression of EMT markers as well as migration and invasive capability between cell lines of differing aggressiveness and propensity to metastasize.
To determine whether EMT can predict propensity for spontaneous metastasis, we chose to focus on related mouse breast cancer cell lines (isolated from a spontaneous mammary tumor in a BALB/cfC3H mouse) that have been extensively characterized for their tumorigenic and metastatic behavior in vivo (Aslakson and Miller,1992). The metastatic properties of the various cell lines are summarized in Table 1. Whereas the 4T1 and 66cl4 cell lines readily metastasize from mammary glands with differential organ specificities, the 67NR variant does not metastasize, but does form tumors in the mammary gland. To characterize the EMT status of these various cell lines, we focused on the 67NR (nonmetastatic), 66cl4 (lung metastatic), and 4T1 (metastatic to multiple sites, including lung) derivatives. As shown in Figure 1, the morphology of the three cell types in two-dimensional (2D) and 3D Matrigel cultures differs considerably, with the 4T1 cells having the most epithelial morphology, and the 67NR the most “mesenchymal” morphology. For comparison, we have shown the scp2 normal mouse mammary epithelial cells and NIH 3T3 mouse fibroblasts to illustrate typical epithelial and mesenchymal morphologies, respectively. Both the 4T1 (metastatic) and the scp2 (normal epithelial) cell lines form highly cobblestone monolayers when confluent, and organized spheroids in 3D Matrigel cultures. In contrast, the 67NR (nonmetastatic) cells do not form cobblestone layers and are highly elongated and fusiform in 3D Matrigel culture, resembling the NIH 3T3 fibroblasts.
Table 1. Tumorigenic and Metastatic Properties of Cell Lines Derived From a Single Mammary Tumor That Arose Spontaneously in a Wild-Type BALB/cfC3H Mouse (Aslakson and Miller,1992)
We next examined the expression of epithelial (E-cadherin and γ-catenin/plakoglobin) and mesenchymal (vimentin) markers by these cell lines, and used the epithelial scp2 and the mesenchymal NIH 3T3 cells as controls. As shown in Figure 2, the scp2 mammary epithelial cells express E-cadherin and γ-catenin, but do not express vimentin. As expected, the NIH 3T3 fibroblasts do not express E-cadherin or γ-catenin, but do express robust amounts of vimentin. We also examined the expression of N-cadherin, as the E-cadherin to N-cadherin switch is often a hallmark of tumor cells that have undergone EMT (Maeda et al.,2005). The scp2 epithelial cells do not express N-cadherin, and only minimal amounts are detected in NIH 3T3 fibroblasts. When the three breast tumor cell lines were examined, we found, surprisingly, that the nonmetastatic 67NR cells do not express E-cadherin or γ-catenin, but do express N-cadherin and vimentin (i.e., have undergone EMT), whereas the highly metastatic 4T1 cells express robust amounts of E-cadherin as well as some γ-catenin, suggesting that these cells maintain expression of epithelial proteins. The lung metastatic 66cl4 cells, however, fail to express E-cadherin, express moderate amounts of vimentin and have undergone an E-cadherin to N-cadherin switch, indicating that these cells have undergone EMT. Because the in vivo metastatic propensity of the cell lines was examined using cells that were stably transfected with a firefly luciferase gene (by IVIS bioluminescent imaging; Fig. 6), we confirmed the expression patterns of E-cadherin, vimentin and N-cadherin by Western blotting in the transfected variants. The expression patterns were identical to the untransfected parental cells (data not shown).
We then examined the expression of EMT markers by immunofluorescence microscopy. As shown in Figure 3, the expression patterns of epithelial markers E-cadherin and zonula occludens (ZO-1), and mesenchymal markers N-cadherin and vimentin, mimic those observed by Western blotting. Again, the nonmetastatic 67NR cells fail to express E-cadherin, but express high amounts of vimentin and N-cadherin similar to fibroblasts, whereas the metastatic 4T1 cells express substantial amounts of E-cadherin and ZO-1, similar to the scp2 epithelial cells. Strikingly, E-cadherin and ZO-1 are clearly localized to junctional complexes in the 4T1 cells and normal scp2 epithelial cells (Fig. 3, arrows). In contrast, the 67NR nonmetastatic cells do not form junctional complexes as they fail to express E-cadherin and, although expression of ZO-1 is detected in these cells, it is localized in the cytoplasm and lamellopodia. The 67NR cells do express N-cadherin, which is mostly cytoplasmic, in contrast to 66cl4 metastatic cells, where it appears to localize to cell–cell borders.
A phenotypic hallmark of mesenchymal cells is the ability to migrate on and invade through extracellular matrices and basement membranes. These properties are also frequently attributed to metastatic ability. We, therefore, assessed the ability of the three cell lines to migrate and invade. Cell migration was determined using two distinct assays, a monolayer “wounding” assay and a transwell chamber cell migration assay. As shown in Figures 4 and 5, the metastatic 4T1 cells, which have a largely epithelial morphology and express E-cadherin, are highly migratory, whereas the nonmetastatic 67NR cells, which express mesenchymal markers, are less migratory than the 4T1 cells but are more migratory than the largely mesenchymal and metastatic 66cl4 cells. These cells are the least migratory in both assays. Invasion through matrigel (Fig. 5) mimics this pattern, in which the E-cadherin positive 4T1 cells are highly invasive compared with the largely mesenchymal 67NR cells and the 66cl4 cells. These data show that the expression of EMT markers, per se, does not predict invasive and migratory ability, and also suggest different forms of migration and invasion (e.g., sheet movement versus single cells) are adopted by these cell types.
We next assessed the metastatic properties of these different cell lines in vivo. To noninvasively track metastasis formation, we engineered the cell lines to stably express the firefly luciferase gene. Stable cell lines expressing this gene were then inoculated into the mammary glands of syngeneic BALB/c mice, and primary tumor and metastasis formation were followed by bioluminescent imaging using the IVIS 200 imaging system as described in the Experimental Procedures section. As shown in Figure 6, all three cell lines formed tumors in the mammary glands within 2 weeks. Lung metastases were detected in all mice inoculated with the 66cl4 cells within 8 weeks. Formation of metastases in mice inoculated with the 4T1 cells was more aggressive, but sporadic, and metastases were detected in multiple sites, including the lungs (Fig. 6). In some cases, metastases were only detected after resection of the primary tumor. However, although primary tumors readily formed in the mammary glands of mice inoculated with the 67NR cells (Fig. 6), no metastases were detected in any organ sites at 100 days after inoculation of the cells. These data agree with previous findings using these cell lines (Aslakson and Miller,1992) and suggest suppression of metastatic ability in the 67NR cells compared with the 4T1 and 66cl4 cell lines.
Gene expression can often be altered between in vitro and in vivo conditions. We therefore examined the expression of the EMT markers, E-cadherin and vimentin, in the primary and metastatic tumors formed by the three cell lines. Normal and tumor tissues were fixed, paraffin-embedded, and stained using immunohistochemistry. As shown in Figure 7A, the expression patterns of E-cadherin and vimentin in the primary tumors reflected the pattern observed in culture. Specifically, the nonmetastatic 67NR tumor cells do not express E-cadherin, but do express vimentin in vivo, whereas the highly metastatic 4T1 cells express E-cadherin and no vimentin in vivo. Importantly, the expression of E-cadherin in primary tumors formed by 4T1 cells is clearly localized to the plasma membrane at cell–cell borders as well as within the cytoplasm (Fig. 7A, inset; arrows). This pattern of staining mirrors that observed in cultured 4T1 cells (see Fig. 3). As observed in cell culture, the 66cl4 cells fail to express E-cadherin and express moderate amounts of vimentin in vivo. This pattern of expression of these two EMT markers is maintained in the lung metastases derived from the 4T1 and the 66cl4 cell lines. In particular, high levels of E-cadherin are expressed in the 4T1 lung metastatic tumor tissue (Fig. 7B), whereas the 66cl4 lung metastatic tumors express more vimentin. It should be noted that metastases were completely absent from mice bearing 67NR cell-derived primary tumors (Fig. 7B). These data show that the pattern of expression of the EMT markers is maintained in the primary and metastatic tumors, and that epithelial to mesenchymal transition does not strictly predict the metastatic ability of the tumors in vivo.
As these data suggest that alterations in expression of genes related to EMT do not correlate with metastatic potential, we performed global gene expression analysis on primary tumors derived from nonmetastatic (67NR) and metastatic (66cl4, 4T1) cell lines. For this purpose, primary tumor tissue from three individual mice per cell line was subjected to laser microdissection to prevent contamination from the surrounding stroma. Subsequently, total RNA was isolated and hybridized to the Affymetrix Mouse Genome GeneChip 430 2.0. We analyzed RNA from three independent 67NR tumors and three independent tumors each from the 4T1 and 66cl4 cell lines. We compared the gene expression profiles of the nonmetastatic 67NR tumors to the metastatic 4T1 and 66cl4 tumors. Therefore, three biological replicates were used for the 67NR nonmetastatic tumors and six biological replicates were used for the metastatic tumors. Using the SAM (significance analysis of microarrays) algorithm, we identified 127 differentially regulated genes at a false discovery rate of 2% with 53 being up-regulated and 74 down-regulated in primary tumors from metastatic compared with nonmetastatic cell lines (Fig. 8). Some of the relevant genes whose expression are up-regulated in both of the metastatic cell types relative to the nonmetastatic 67NR cells are carbonic anhydrase 9 (CA9), Ephrin receptor B2 (EphB2), and CD24 antigen. CA9 and EphB2 are both associated with poor breast cancer patient prognosis (Brennan et al.,2006; Hussain et al.,2007; Trastour et al.,2007), while CD24 is a marker of mammary stem cells (Shackleton et al.,2006). To validate the differential expression of some of these genes, we stained the 67NR, 4T1, and 66cl4-derived primary tumor sections with an anti-CA9 antibody. As shown in Figure 9, the expression levels of CA9 are highly elevated in the metastatic 66cl4 and 4T1 tumors relative to the nonmetastatic 67NR tumors, which are largely negative for expression of CA9. We also immunostained these tumors with an anti-Twist antibody. As shown in Figure 9A, there is very little difference in staining intensity of Twist between the three tumor types. In addition, we interrogated the DNA microarray gene expression data for Twist mRNA expression, as shown in Figure 9B. There is no significant difference in the expression of Twist mRNA between the 67NR, 4T1 and 66cl4 primary tumors. These data further suggest that EMT, per se, does not correlate with in vivo spontaneous metastatic ability.
Epithelial–mesenchymal transition (EMT) is a process that has been implicated in cancer progression (Thiery,2002; Thompson et al.,2005). EMT has also recently been implicated in the sensitivity of cancer cells to inhibitors of epidermal growth factor receptor (EGFR), with cells that have undergone EMT being resistant to such inhibitors (Yauch et al.,2005). The involvement of EMT in metastasis of breast and other cancers has been controversial and continues to be debated and analyzed (Tarin et al.,2005; Thompson et al.,2005; Lee et al.,2006). In this study, we have analyzed the clinically relevant mouse 4T1 breast cancer model to determine whether EMT can be linked to metastatic ability. The availability of metastatic and nonmetastatic variants derived from a spontaneous mammary tumor from a BALB/c mouse provides a powerful model for the identification of genes involved in the metastatic process. EMT is considered to promote the early stages of metastasis (i.e., invasion and migration) that allow primary tumor cells to invade the basement membrane and disseminate into the circulation.
We find that contrary to expectations, the “EMT phenotype” cannot be used to predict whether a primary breast tumor will metastasize or not, at least in this murine model of spontaneous breast cancer metastasis. The results presented in this study show the following: (1) The nonmetastatic 67NR variants clearly show hallmarks of EMT. These cells do not express E-cadherin, but do express vimentin, characteristics associated with a mesenchymal phenotype. In addition, these cells express the “mesenchymal” cadherin, N-cadherin, and are moderately invasive in vitro. (2) The highly metastatic 4T1 cells, in contrast, appear very epithelial, and express E-cadherin and γ-catenin. E-cadherin and ZO-1 are localized to cell–cell junctions, and these cells do not express N-cadherin and express very little vimentin. Thus, these cells would be defined as being largely “epithelial.” However, the 4T1 cells are highly migratory and invasive in vitro, clearly showing that EMT is not a prerequisite for invasiveness. (3) The lung metastatic 66cl4 cells have undergone an EMT and they have undergone a E-cadherin to N-cadherin switch (Maeda et al.,2005). However, these cells are the least migratory of all three cell lines and they do not invade through matrigel over the time course examined in this study. These results suggest that tumor cells can spontaneously metastasize in vivo, without exhibiting invasive and migratory properties in vitro, and cells that exhibit hallmarks of EMT are unable to metastasize. Finally, The EMT markers observed in culture are maintained in vivo, indicating that the nonmetastatic or metastatic properties of the cell lines in vivo are not due to alterations in expression of the epithelial and mesenchymal protein markers in vivo.
EMT is a dynamic process and “metastable” cells that exhibit both epithelial and mesenchymal markers are known to exist (Lee et al.,2006). It is possible that some 4T1 cells undergo EMT in vivo, and that these are the cells that form the lung metastases. However, the EMT markers observed in cell culture in this model are maintained in vivo, indicating that the metastases develop from cells with epithelial markers, or that any mesenchymal cells present have undergone a “mesenchymal-to-epithelial” transition (MET), which may be required for the establishment of the metastases. These possibilities reflect the dynamic nature of EMT.
The findings reported here are in contrast with those of Yang et al. (2004), in which differential gene expression between 4T1 cells and 67NR cells resulted in the identification of Twist, a transcriptional repressor of E-cadherin (Yang et al.,2004). Twist was found to induce EMT and promote a metastatic phenotype. We have failed to detect differences in Twist gene and protein expression in the tumors derived from the 4T1, 66cl4 and 67NR cells. The reasons for the differences in our results as compared to those reported by Yang et al. (2004) are not clear, but could be attributed to clonal differences between cell lines. In addition, it should be pointed out that Twist protein expression was not verified in 4T1 and 67NR tumors in vivo by Yang et al. (2004).
Collectively, the data presented in this study bring into question the prevailing paradigms that EMT always facilitates metastasis, and that the migration and invasive properties of cancer cells in vitro correlate with in vivo metastasis. Our findings suggest that other processes and mechanisms must dictate metastatic ability, including the manner in which cells disseminate from the primary tumor. Our findings also suggest caution in extrapolating certain phenotypic properties of cancer cells to their ability to spontaneously metastasize, a process that involves many steps such as tumor cell dissemination, circulation, extravasation, and growth at the distant site, resulting in the survival of only a minority of tumor cells leading to metastasis formation (Talmadge,2007).
The 4T1 model provides a powerful model for the identification of genes and processes that promote or inhibit spontaneous metastasis in vivo (Eckhardt et al.,2005). The differential global gene expression analysis carried out here between primary tumors formed by the nonmetastatic 67NR cells compared with the metastatic 4T1 and 66cl4 cells have identified several candidate genes that may influence metastatic ability. These include, amongst others, CA9, CD24 and EphB2. Carbonic anhydrase gene expression is highly up-regulated in the primary tumors of the metastatic 4T1 and 66cl4 cells relative to the primary tumors generated from the nonmetastatic 67NR cells. CA9 is a hypoxia-regulated gene and several reports have shown that its expression is associated with poor survival of breast cancer patients (Brennan et al.,2006; Hussain et al.,2007; Trastour et al.,2007). We have confirmed that CA9 expression is indeed elevated in tumors derived from metastatic 66cl4 and 4T1 cells relative to the nonmetastatic 67NR cells. The expression of EphB2 is also associated with shorter disease-free survival of breast cancer patients (Wu et al.,2004), and this protein is also present at higher levels in the primary tumors generated from the metastatic cells relative to the nonmetastatic cells. The expression of the CD24 antigen, a mammary stem cell marker (Shackleton et al.,2006), is also expressed at higher levels in the primary tumors generated from the metastatic cell lines relative to the 67NR nonmetastatic cells. In depth analysis of these (and other) genes in this breast cancer model should provide further insights into breast cancer metastasis.
The breast cancer cell lines 4T1, 66cl4, and 67NR were kindly provided by Dr. Fred Miller (Karmanos Cancer Institutes, Detroit, MI). These lines are derived from a single mammary tumor that arose spontaneously in a wild-type BALB/cfC3H mouse and form primary mammary tumors with equivalent kinetics that differ dramatically in their metastatic potential (Aslakson and Miller,1992; Eckhardt et al.,2005). The scp2 mouse mammary epithelial cell line and the NIH 3T3 mouse embryonic fibroblast cell line were previously described (Todaro and Green,1963; Somasiri et al.,2001). All breast cancer cell lines, stable cell lines expressing luciferase, and the NIH 3T3 cell line were grown in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mmol/L L-glutamine, 1 mM mixed nonessential amino acids, penicillin (100 units/ml), and streptomycin (100 μg/ml; Invitrogen, Rockville, MD). Scp2 cells were grown in a 1:1 mixture of DMEM and F12 (DMEM-F12) containing 10% FBS, insulin (5 μg/ml, Sigma, Mississauga, ON, Canada), 2 mmol/L L-glutamine, penicillin (100 units/ml), and streptomycin (100 μg/ml). All cells were incubated at 37°C with 5% CO2.
Establishing Stable Cell Lines
The CMV promoter was amplified from pcDNA 3.1/V5-His A (Invitrogen Life Technologies, Carlsbad, CA) with the following primers: 5′-TTC ACT CGA GCG ATG TAC GGG CCA GAT ATA C-3′, which introduced a XhoI site, and 5′-GCG CAA GCT TAA TTT CGA TAA GCC AG -3′, which introduced a Hind III site. The CMV promoter PCR fragment was digested with restriction enzymes Xho I and Hind III, and ligated with T4 DNA ligase to the Xho I and Hind III sites of pGL 4.20 (Promega, Madison, WI), a plasmid shuttle vector encoding the luciferase reporter gene luc 2, to make pGL4CMV. pGL4CMV was isolated and sequenced. pGL4CMV was transfected into 80% confluent 4T1, 66cl4, and 67NR cells using a modified LipofectAMINE protocol (Invitrogen Life Technologies). After transfection, the cells were grown for 2 days in a nonselective growth medium that was then replaced with a selection medium containing puromycin (Sigma, Mississauga, Ontario, Canada). After 2 weeks, individual colonies were isolated using cloning cylinders (Corning Incorporation, Corning, NY) and selected for luciferase activity using a Luciferase Assay System (Promega, Madison, WI) following the protocol recommended by the manufacturer. Luciferase activity was quantified in a Lumat LB 9507 tube luminometer (Berthold Technologies USA LLC, Oak Ridge, TN) and single cell-derived clones were subsequently expanded in selection medium for use in experiments.
Spheroid formation was performed as described previously (Lee et al.,2007). Briefly, eight-well chamber slides (Lab Tek II Chamber Slides, Nalge Nunc International, Rochester, NY) were coated with 50 μl/well of ice-cold Extracellular Matrix Extract (Matrigel, BD Bioscience, Bedford, MA). Suspended cells (1.2 × 104 cells in 100 μl of growth medium) were placed on top of the Matrigel and allowed to attach for 30 min at 37°C. 150 μl of growth medium, containing 10% v/v of Matrigel, was added on top of cell layer and cells were cultured at 37°C for 4 days. Brightfield images of 3D aggregates were subsequently acquired and analyzed.
Western blotting was carried out on breast cancer, NIH 3T3, and scp2 cell lysates as indicated. Cells were lysed in RIPA lysis buffer (20 mM Tris-HCl, pH 7.5; 150 mM NaCl; 0.5% deoxycholate, 1% NP-40; 0.1% sodium dodecyl sulfate [SDS]; 2 mM EDTA; 2 mM Na3VO4; 1 mM NaF; 2 mM β-glycerophosphate; “Complete” protease inhibitor cocktail [Roshe Diagnostics, Laval, QC, Canada]). A total of 20 μg of total cell lysates were separated by 6% or 10% SDS-polyacrylamide gel electrophoresis, transferred to PVDF membrane (Millipore Corporation, Billerica, MA) and probed with primary antibodies as described previously (Somasiri et al.,2001). Mouse monoclonal antibodies directed against E-cadherin, N-cadherin, and Vimentin were from BD Pharmingen (Mississauga, ON, Canada); anti–β-actin antibodies were from Sigma and rabbit polyclonal anti–γ-catenin antibodies were from Santa Cruz (Santa Cruz Biotechnology Inc., Santa Cruz, CA). Secondary HRP-linked antibodies used were anti-mouse IgG (Cell Signalling, Danver, MA) and goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories Inc. West Grove, PA). Antibody reactions were visualized by an enhanced chemiluminescence (ECL) detection system (Amersham Biosciences UK limited, Buckinghamshire, England).
Light and Immunofluorescent Microscopy
Both brightfield and fluorescent images were obtained using a Zeiss microscope (Axioplan2 Imaging, Photonitech, Singapore) equipped with a CCD camera. For immunofluorescence, cells were plated on Lab-Tek Chamber Slides (Nalge Nunc International), fixed with 4% paraformaldehyde in phosphate buffered saline (PBS), permeabilzed with 0.2% Triton X-100 in PBS, and stained with mouse monoclonal antibodies directed against E-cadherin, N-cadherin, vimentin (BD Pharmingen), and rat monoclonal anti–ZO-1 antibodies (Chemicon, Temecula, CA). Samples were mounted in Vectashield Mounting Medium containing 4′,6-diamidine-2-phenylidole-dihydrochloride (DAPI; Vector Laboratories Inc., Burlingame, CA).
Migration and Invasion Assay
For the wound healing assay, confluent cells were scraped with a 10-μl Gilson pipette tip. After wounding, the culture medium was changed to remove detached and damaged cells, and wound closure was monitored microscopically every 12 hr. Five different wounds for each cell line were made and migration was determined as the average distance from the wound origin covered by cells 12 hr after wounding.
Chemotactic migration of tumor cells was analyzed on Transwell filters (6.5 mm diameter, 8 μM pore size) from Corning Corporation. Trypsinized cells (5 × 104) were washed with PBS, resuspended in DMEM supplemented with 0.2% fatty-acid free bovine serum albumin, and placed into the insert with a filter. The lower chamber was filled with medium containing 10% FBS as a chemoattractant. After 24 hr, cells in the upper chamber were removed with a cotton swab and filters were washed twice with PBS. Cells present on the lower surface of the filters were fixed in methanol for 10 min. Filters were then cut from the inserts and immersed in mounting medium containing DAPI (Vector Laboratories Inc, Burlingame, CA). The nuclei from five different fields of each Transwell were counted and averaged to represent the number of migrated cells per field at ×100 magnification. In a similar manner, the invasiveness of tumor cells was evaluated using Transwell filters coated with Matrigel (BD Matrigel Invasion Chamber, BD Biosciences). Each experiment was repeated three times, and results were averaged.
Primary and Metastatic Tumor Model and In Vivo Bioluminescent Imaging
Female BALB/c mice (7–9 weeks old, Taconic Laboratories, Oxnard, CA) were purchased, housed under pathogen-free conditions and kept on a 12 hr light/12 hr dark cycle in the Animal Research Centre, BC Cancer Research Centre (Vancouver, BC, Canada), according to the guidelines of the Canadian Council on Animal Care and the University of British Columbia Animal Care Committee. The mice were injected with 1 × 106 viable cells (4T1/CMVLUC, 66cl4/CMVLUC, and 67NR/CMVLUC) into the right fourth mammary gland (50 μl in PBS per mouse). At 15 and 56 days after injection, in vivo bioluminescent imaging was conducted with a cryogenically cooled IVIS 200 system (Xenogen Corporation, Alameda, CA) using Living Image acquisition and Living Image 2.50.1 analysis software (Xenogen). Briefly, the mice were injected intraperitoneally with 200 μl of D-luciferin (15 mg/ml in PBS), placed in a light-tight chamber and imaged with a CCD camera (Xenogen). Images were acquired 10 min after D-luciferin administration.
On the day of removal of the primary tumor, tumor-bearing mice were given an intraperitoneal injection of ketamine (75 mg/kg body weight) and xylazine (10 mg/kg). Fifteen minutes later, the animals were checked for depth of anesthesia and the primary tumors were removed, weighed and prepared for endpoint assays. After surgery, the animals were subcutaneously injected with buprenorphine (0.1 mg/kg). The animals were allowed to recover and returned to the environmental chamber. Animals were monitored daily and killed by CO2 inhalation at 57 days after inoculation. Upon killing, lung tissue was harvested and prepared for the following assays.
Histopathology and Immunohistochemistry
For histologic examination, the primary tumor and lungs from tumor-bearing mice (three individual mouse samples/tumor type) were fixed in 4% paraformaldehyde in PBS and then 5-μm paraffin sections were either hematoxylin and eosin (H&E) stained or processed for immunohistochemical analysis and counterstained following standard procedures. Sections were incubated with mouse E-cadherin and vimentin monoclonal antibodies (BD Pharmingen) or rabbit anti CA9 (M-100) and twist (H-81) polyclonal antibodies (Santa Cruz Biotechnology Inc). Mouse IgG2a, mouse IgG1 (BD Pharmingen) and rabbit IgG (Santa Cruz Biotechnology Inc) were used as controls. Antibody binding was detected using the DakoCytomation Envision System (DakoCytomation, Carpintenia, CA) and Vector NovaRED substrate kit (Vector Laboratories Inc., Burlingame, CA) following the protocol of the manufacturer.
Laser Microdissection (LM)
Altogether nine primary breast cancer samples (3 each of 4T1/CMVLUC, 66cl4/CMVLUC, and 67NR/CMVLUC) were investigated. Tissue samples were embedded in TissueTek OCT (Finetek U.S.A. Inc., Torrance, CA), snap-frozen in liquid nitrogen, and stored at −80°C for further analysis. Frozen specimens were cryosectioned in 8-μm sections onto membrane slides (Molecular Machines and Industries Inc., Knoxville, TN), fixed in 70% ethanol, stained using hematoxylin for a few seconds, and dehydrated in ethanol and xylene. After dehydration, the slides were air-dried for 5 min and subjected to LM. All solutions were prepared using diethyl pyrocarbonate-treated water. RNase free instruments and RNaseZap (Ambion, Foster City, CA) were used throughout the procedure.
All tissues included in this study were re-examined by a clinical pathologist dedicated to the project. Ten to thirty sections of each primary tumor were prepared. Primary tumor cells were microdissected using a MMI CellCut LM system (Molecular Machines & Industries Inc.). Necrotic areas, when present, were avoided. Microdissected tissues were placed in MMI Eppendorf tubes (Molecular Machines & Industries Inc.) and processed for RNA extraction. LM of tissue sections on each membrane slide was performed within 1.5 hr after tissue staining.
RNA Isolation, Labeling, and Hybridization of Microarray Targets
Total cellular RNA was extracted from each population of microdissected tissues using a commercially available system (RNAeasy micro kit, Qiagen, Mississauga, ON). Subsequently, samples were re-extracted as described above and resuspended in 10 μl of diethylpyrocarbonate-treated water. RNA was quantified and qualified using an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA). RNA probes were labeled according to the supplier's instructions using 2 μg of total RNA (Affymetrix, Santa Clara, CA). Analysis of gene expression was carried out using the Mouse Genome 430 2.0 microarray (Affymetrix) which contains 45,000 probe sets. Hybridization and washing of gene chips were carried out according to the supplier's instructions. A Genechip scanner 3000 7G with autoloader (Affymetrix) was used for reading out the microarrays.
Microarray Hybridization and Analysis
Data analysis was performed using the Bioconductor based CARMAweb software package (Rainer et al.,2006). Raw data were preprocessed by the RMA (robust multiarray analysis) method, and differentially expressed genes were determined using the SAM (significance analysis of microarrays) algorithm (Tusher et al.,2001).
This work was supported by a program project grant from the Canadian Breast Cancer Research Alliance (CBCRA) with funds from the Canadian Breast Cancer Foundation and the Cancer Research Society.