Netrin-1 regulates invasion and migration of mouse mammary epithelial cells overexpressing Cripto-1 in vitro and in vivo

Human Cripto-1 (CR-1) is a member of the epidermal growth factor (EGF)-CFC family of signaling proteins first cloned from the human teratocarcinoma cell line NTERA2 (Ciccodicola et al., 1989). The glycosylphospatidylinositol (GPI) membrane-linked Cripto (Minchiotti et al., 2000) binds to Nodal, a member of the TGF-β superfamily, which facilitates Nodal interaction with an activin receptor complex composed of the type-I serine-threonine receptor ActRIB (or ALK4) and the type-II activin receptor, either ActRII or ActRIIB. This leads to phosphorylation of Smad2 and/or Smad3, which together with Smad4 can heterodimerize and translocate to the nucleus to mediate transcriptional responses (Bianco et al., 2002; Yeo and Whitman, 2001). Cripto-1 can also function through a Nodal- and ALK4-independent signaling pathway by specifically binding to Glypican-1, a membrane-associated heparan sulfate proteoglycan, which leads to phosphorylation of the tyrosine kinase c-Src and activation of mitogen-activated protein kinase (MAPK) and Akt signaling pathways (Bianco et al., 2003).

Cripto-1 is expressed in a variety of human cancers (Salomon et al., 2000) including breast cancer (Qi et al., 1994). Overexpression of CR-1 has been shown to induce proliferation, migration and invasion of human breast cancer cells (Brandt et al., 1994; Normanno et al., 2004) and of mouse mammary epithelial cells (Bianco et al., 2003; Strizzi et al., 2004; Wechselberger et al., 2001). Furthermore, cells overexpressing CR-1 have been shown to acquire specific biochemical and morphological features suggesting that CR-1 may play a role in promoting epithelial mesenchymal transition (EMT).

Epithelial mesenchymal transition is a normal physiologic process important for embryogenesis, tissue growth, wound healing and tissue repair (Perez-Pomares and Munoz-Chapuli, 2002). During EMT epithelial cells lose their adhesive properties owing to modification in the expression of cellular adhesion molecules like E-cadherin (Boyer et al., 2000; Savagner, 2001). In fact, the overexpression of the genes snail and slug has been shown to play an important role in inducing EMT by negatively affecting the expression of E-cadherin (Cano et al., 2000). Epithelial cells undergoing EMT also show changes in the cytoskeleton. For example, vimentin is a cytoskeleton molecule normally expressed in mesenchymal cells and when expressed in epithelial cells may facilitate their acquisition of a more spindle-shaped morphology during EMT in both normal and disease states (Fuchs and Weber, 1994; Gilles et al., 1996; Lane et al., 1983). Molecules that are involved in growth factor signaling such as Src and phosphoinositide 3-kinase (PI3K) are also activated during EMT (Thiery and Chopin, 1999; Vincent-Salomon and Thiery, 2003). Thus, cellular changes characteristic of EMT facilitate migration and invasion of epithelial tumor cells and have recently been suggested as an index of aggressiveness and increased metastatic potential in different types of malignant tumors (Birchmeier et al., 1996a; Birchmeier et al., 1996b; Gilchrist et al., 2002). Specifically, reports have suggested that EMT may be relevant to the development of human breast cancer, as mutations in E-cadherin expression (Berx et al., 1998; Cano et al., 2000), overexpression of snail and slug (Cano et al., 2000; Hajra et al., 2002), overexpression of vimentin (Hanna et al., 2003) and increased activity of signaling molecules involved in migration and invasion during EMT (Vincent-Salomon and Thiery, 2003) have all been identified in human breast cancer both in vitro and in vivo. In addition, EMT has been suggested to play a role during metastasis and affecting prognosis in human breast cancer (Fuchs et al., 2002; Xue et al., 2003).

Overexpression of mouse Cripto-1 (Cr-1) increases proliferation of mouse mammary epithelial cells and causes them to assume a more mesenchyme-like phenotype. In addition, Cr-1 overexpression increases anchorage-independent growth of mammary epithelial cells in soft agar and migration when cells are grown on plastic or on porous filters coated with extracellular matrix, and during wound healing assays (Wechselberger et al., 2001). Increased expression of Cr-1 also induces the formation of branch-like structures when mouse mammary epithelial cells are grown in a collagen type I matrix (Wechselberger et al., 2001). Overall, these responses are reminiscent of EMT and suggest that Cr-1 overexpression induces this transition in mammary epithelial cells. However, whether Cripto-1-induced migration and proliferation can be influenced by extracellular directional cues is unclear.

Various studies have identified different chemotropic factors that regulate the direction of cell migration. Most of these have been identified during neuronal development and include proteins such as, Slits, Ephrins, Semaphorins (Kolodkin et al., 1993), Sonic hedgehog (Charron et al., 2003), bone morphogenic proteins (Butler and Dodd, 2003), Wnts (Yoshikawa et al., 2003) and Netrins (Serafini et al., 1996; Serafini et al., 1994). Sequence and functional analysis have shown that Netrins are a conserved family of secreted proteins that have regional homology to laminins and are capable of regulating axonal outgrowth (Kennedy et al., 1994; Puschel, 1999; Serafini et al., 1996). The direction of Netrin-dependent neuronal outgrowth is determined by the cellular expression of receptors belonging to either the DCC (deleted in colon cancer) or UNC5 families of Netrin-1 receptors (Keino-Masu et al., 1996; Leonardo et al., 1997). These single-pass transmembrane receptors contain immunoglobulin domains with DCC containing fibronectin type-3 domains and with UNC5 containing a thrombospondin type-I domain (Chisholm and Tessier-Lavigne, 1999). The DCC receptors, which include the structurally similar Neogenin receptor, mediate attraction, whereas repulsion is mediated by a complex of DCC and UNC5 receptor families (Hinck, 2004; Hong et al., 1999). The highly conserved family of UNC receptors possess a high level of structural and sequence homology in the ligand binding extracellular domain (Engelkamp, 2002). In humans, UNC5 receptors are composed of UNC5HA, UNC5HB and UNC5HC and correspond to the rodent orthologues UNC5H1, UNC5H2 and UNC5H3, respectively (Arakawa, 2004).

Recent studies have found functioning Netrin molecules outside the nervous system, in the pancreas, intestine (Jiang et al., 2003; Yebra et al., 2003), lung (Liu et al., 2004) kidney, heart and vasculature (Koch et al., 2000; Lu et al., 2004; Park et al., 2004) where they presumably play a role in the development of these organs by regulating the migration of different types of cells. Regulation of the expression of Netrin-1 and its receptors may play a role in tumorigenesis. In fact, Netrin-1 was shown to be reduced in tumors of the prostate and of the nervous system (Latil et al., 2003; Meyerhardt et al., 1999). Low levels of somatic mutations of DCC have been identified in cancers of the brain, stomach, pancreas, colorectum and testicle (Arakawa, 2004) and in a series comparing human colorectal tumors with corresponding normal tissues, of the different UNC5 receptors studied, UNC5A, the orthologue of rodent UNC5H1, showed the highest percentage of altered expression (Thiebault et al., 2003).

Netrin-1 and Neogenin have been shown to be involved in maintaining adhesion between cap cells and luminal cells in the mammary gland terminal end buds (Srinivasan et al., 2003). As Cr-1 is also expressed in the terminal end buds of developing mammary glands (Kenney et al., 1995) and is capable of inducing migration by deregulating cell adhesion and promoting EMT in mammary epithelial cells, we investigated the expression and function of Netrin-1 and its receptors in invasion, migration and colony formation of mouse mammary epithelial cells that overexpress Cripto-1.

Wild-type mouse mammary epithelial cell lines, HC-11 and Eph4, and their counterparts overexpressing Cr-1, respectively HC-11/Cr-1 and EpH4/Cr-1, were grown as previously described (De Santis et al., 1997; Wechselberger et al., 2001). The polyclonal rabbit antibodies against Netrin-1 and UNC5H1 have also been previously described (Srinivasan et al., 2003; Williams et al., 2003). Polyclonal rabbit anti-Neogenin (H-175), polyclonal goat anti-Cripto-1 (F-20) and polyclonal goat anti-vimentin (S-20) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-E-cadherin antibody was purchased from Transduction Laboratories (Lexington, KY). Rabbit polyclonal anti-P-Akt (S473), recombinant mouse Netrin-1 (rmNetrin-1) and blocking antibodies against mouse Neogenin and against UNC5C were all purchased from R&D Systems (Minneapolis, MN). The UNC5C receptor is highly homologous to UNC5H1. In fact, alignment analysis of amino acid sequences shows that UNC5C and UNC5H1 share 65% overall sequence homology and 70% residue identity in the ligand binding extracellular domain. Synthetic inhibitors PP2, LY294002 (LY) and PD98059 (PD) capable of inhibiting the activation of c-Src, PI-3K and MAPK respectively, in mouse mammary epithelial cells overexpressing Cr-1 (Bianco et al., 2003; De Santis et al., 1997; Ebert et al., 1999) were purchased from Calbiochem, San Diego, CA.

Lysates were obtained from the mouse mammary epithelial cells and western blotting was performed as previously described (Bianco et al., 2002). A 1:1000 dilution was used for all primary antibodies unless otherwise stated. HC-11/Cr-1 cells were transfected with either an anti-Cr-1 small interfering RNA (siRNA) or irrelevant scrambled siRNA, used as a control, both custom made by Qiagen (Valencia, CA). The anti-Cr-1 siRNA sequence is designed to target the Cr-1 mRNA sequence at AAACAGCTAAATTATCTTTAA (GenBank accession number NM_011562). The transfection experiments with siRNA were performed using Qiagen transfection reagents for siRNAs and following the manufacturer's instructions for transfection of cells with siRNAs. Western blotting for analysis of Cr-1 expression was performed on lysates collected from the transfected cells after 72 hours. Reverse transcriptase-PCR for analysis of Cr-1 expression in cells treated with the anti-Cr-1 siRNA was performed as previously described (Kenney et al., 1995).

For treatment with synthetic inhibitors, approximately 2×10⁵ EpH4 or EpH4/Cr-1 cells were seeded and grown until 70-80% confluent. The cells were then serum starved for 24 hours and subsequently treated for 8 hours with 10 μM PP2 or LY, or 20 μM PD before being harvested for western blotting. To quantify the expression of the different proteins analyzed, western blots were scanned by densitometric analysis, which was performed using the public domain NIH Image analyzer developed at the US National Institutes of Health (http://rsb.info.nih.gov/nih-image/). Final densitometric readings were normalized against actin for equal loading and expressed as optical densitometric (OD) units.

For immunofluorescence, approximately 1×10⁵ cells were cultured overnight in Lab-Tek dual chamber slides (Nalge Nunc, Naperville, IL). Culture medium was removed and cells were washed twice with PBS, fixed with ice-cold 100% methanol for 10 minutes and air-dried. Slides were then washed three times with PBS, blocked for 30 minutes with 5% normal goat serum and incubated for 1 hour with primary rabbit antibodies against Netrin-1, Neogenin or UNC5H1 (1:100). Cells were again washed three times with PBS and incubated for 30 minutes with goat anti-rabbit Alexa Fluor-conjugated secondary antibody (1:600) (Molecular Probes, Eugene, OR). Slides were finally mounted with Vectashield (Vector Labs, Burlington, CA), a mounting medium containing DAPI for identification of cell nuclei.

For immunohistochemistry, 5-μm-thick sections of paraffin-embedded, formalin-fixed mammary tumors from MMTV-CR-1 transgenic mice (Wechselberger et al., 2005) were deparaffinized in xylene, rehydrated in a series of graded ethanols, and predigested with ready-to-use pepsin solution (Digest-All3; Zymed, San Francisco, CA) for 6 minutes at 37°C. Endogenous peroxidase activity was blocked for 10 minutes with 0.3% H₂O₂ in methanol. The sections were then incubated for 30 minutes at room temperature with anti-Netrin-1 or anti-Neogenin primary antibodies (1:100). Immunostaining was carried out using the Vectastain ABC kit (Vector, Burlingame, CA) and following the manufacturer's instructions. Color was developed with DAB peroxidase substrate (Vector) and sections counterstained with haematoxylin. When appropriate, immunostaining intensity was quantified by using the public domain NIH digital image analyzer described.

Cell invasion and migration across a basement membrane matrix was evaluated using a commercially available 12- or 24-well plate cell invasion/migration assay kit (Chemicon, Temecula, CA) and following the manufacturer's instructions. Briefly, ∼3.5×10⁵ cells were seeded into individual invasion chambers, which in turn were placed in 12-well plates containing low serum (2% FBS) culture medium with or without 25 or 50 ng/ml rmNetrin-1 in the lower chamber and incubated for 36 hours. Non-invading cells were carefully wiped off the upper surface of the invasion filters with a swab. Cells that invaded and migrated through the matrix-containing membrane and reached the lower surface of the invasion chamber were stained with crystal violet and counted in at least four different high power fields (hpf) using a light microscope. In parallel experiments ∼3.5×10⁵ EpH4/Cr-1 or HC-11/Cr-1 cells were seeded in normal 12-well plates, grown until ∼70-80% confluent and then treated with 50 ng/ml rmNetrin-1. After 48 hours, lysates were collected from these cells and western blotting performed for analysis of the expressions of Neogenin and UNC5H1.

To determine whether blocking of Netrin-1 receptors would affect invasion and migration, Cr-1-overexpressing cells were pre-incubated with 10 μg/ml of either anti-UNC5C or anti-Neogenin blocking antibody for 30 minutes prior to seeding in the invasion chambers that contained 50 ng/ml rmNetrin-1. Approximately 2×10⁵ cells were seeded in invasion chambers that were contained in the 24-well plate cell invasion kit and incubated as described above. For the quantification of invading cells in these sets of experiments, direct cell counts were not obtained in order to reduce possible underestimation artifact owing to selection of non-representative areas on the invasion membrane as a consequence of the relatively low number of cells seeded. Instead, stain from the invading cells that reached the lower surface of the invasion membrane was extracted with 10% acetic acid and optical density quantified at 560 nm.

For colony formation, a 2 ml layer of ready-to-use Matrigel solution (Collaborative Biomedical Products, Bedford, MA) was pipetted in six-well cell culture plates. Sterile disks of blot paper preabsorbed with PBS or with PBS containing approximately 100, 200 or 400 ng/ml of rmNetrin-1 were placed in the center of the Matrigel and covered with a second layer (2 ml) of Matrigel. Approximately 5×10⁵ Eph4/CR-1 cells in complete culture media were seeded in each well. Formation of spherical colonies was evaluated after 48 hours. To determine the number of colonies, an automated colony counter (Artek Systems Corp, Farmingdale, NY) was adjusted to count colonies having a diameter greater than 500 μm. The colonies were counted in three evenly spaced circumferences defined as proximal (P), medial (M) or distal (D) to the disks of preabsorbed blot paper that served as the source of Netrin-1.

Five-week-old FVB/N or MMTV-CR-1 female mice were implanted with cholesterol pellets. Different cholesterol pellets were formulated to continuously release various doses (25 or 50 ng/day) of rmNetrin-1 for 2 weeks and were prepared as previously described (Vonderhaar, 1987). The pellets were then surgically implanted into the right inguinal mammary gland at approximately the same distance from the mammary lymph node for each animal. A total of 14 FVB/N mice were used. Four FVB/N mice were implanted with cholesterol-only pellets, five were implanted with pellets releasing 25 ng/d of rmNetrin-1 and five with pellets releasing 50 ng/day of rmNetrin-1. A total of 14 MMTV-CR-1 transgenic mice were also used and distributed in each experimental group as described above for the FVB/N mice. Two weeks after implantation, mammary glands were surgically removed and analyzed by whole mount morphology at 10× magnification. For whole mount preparation, mammary glands were spread and fixed on glass slides with Carnoy's solution (glacial acetic acid:ethanol; 1:3) for 60 minutes at room temperature. The glands were rehydrated and stained overnight in aluminum carmine solution. The glands were subsequently dehydrated, cleared in xylene and mounted for microscopic observation. Digital microphotographs were taken using a Polaroid DMC-1 digital camera (Polaroid, Cambridge, MA) mounted on a Leica MZ125 microscope (Leica, Wetzlar, Germany). For each mammary gland, ductal elongation was represented as the distance measured on the microphotographs in millimeters from the center of the mammary gland lymph node to the tip of the farthest growing duct in direction of the inserted pellet. Mammary gland tissue from the MMTV-CR-1 transgenic mice treated with control or Netrin-1-releasing pellets were also processed for immunohistochemistry as described above and analyzed for expression of UNC5H1, E-cadherin, vimentin and P-Akt. Care and use of the experimental animals for this study was in compliance with the relevant animal welfare laws, guidelines and policies at NIH.

Quantitative values are represented as the mean of quadruplicate results. All in vitro experiments were repeated at least three times. The statistical significance of the difference between groups analysed was determined by the Wilcoxon rank sum test. Comparisons resulting in a P value of less than 0.05 were considered statistically significant and identified in the figures with an asterisk (*).

Western blot analysis of cell lysates showed that overexpression of Cr-1 in mouse HC-11 or EpH4 mammary epithelial cells was associated with reduced expression of Netrin-1 and increased expression of Neogenin when compared to wild-type (WT) cells (Fig. 1A). No significant difference in the expression of UNC5H1 between the wild type and Cr-1-overexpressing HC-11 or EpH4 cells was detected by western blot analysis (Fig. 1A). The decreased expression of Netrin-1 and the increased expression of Neogenin in Cr-1-overexpressing mouse mammary epithelial cells compared to the wild type were confirmed by immunofluorescence analysis of HC-11/WT and HC-11/Cr-1 cells (Fig. 1B). Immunohistochemistry showed decreased staining intensity for Netrin-1 compared to the intensity of staining obtained for Neogenin in areas of the MMTV-CR-1 transgenic mice mammary tumors containing large anaplastic, mesenchyme-like tumor cells (Fig. 1C). Similar to the mammary epithelial cells overexpressing Cr-1, the anaplastic, mesenchyme-like tumor cells in the MMTV-CR-1 mammary tumors also expressed higher levels of CR-1 and of biochemical markers suggestive of EMT when compared to the more differentiated areas of the tumor (Strizzi et al., 2004).

Transfection of HC-11/Cr-1 cells with an anti-Cr-1 siRNA, capable of reducing Cr-1 mRNA and protein expression in these cells (Fig. 2A), caused Netrin-1 and Neogenin expression to return to levels that were comparable to those detected in wild-type HC-11 cells by western blotting (Fig. 2B). EpH4/Cr-1 cells do not express endogenous Nodal (Bianco et al., 2002). When Nodal-independent signaling in EpH4/Cr-1 cells was blocked with the synthetic c-Src inhibitor PP2 (Bianco et al., 2003), there was an increase in Netrin-1 expression and a decrease in Neogenin expression compared to that in untreated cells (Fig. 2C). Treatment of EpH4/Cr-1 cells with the PI-3K inhibitor LY was also associated with an increase in Netrin-1 expression and a decrease in Neogenin expression (Fig. 2C,D). No significant effect on the expression of Netrin-1 or Neogenin was detected when EpH4/Cr-1 cells were treated with the MAPK inhibitor, PD (Fig. 2C,D). These results suggest that blocking Cripto-1 signaling through a Nodal-independent pathway, which is dependent on c-Src and PI-3K, rescues Cripto-1-dependent loss of Netrin-1 expression.

Cripto-1 overexpression in mouse mammary cells is associated with reduced expression of the intracellular adhesion molecule E-cadherin and increased expression of vimentin, which are characteristic of cells undergoing EMT (Strizzi et al., 2004). Akt activation has also been shown to occur during Cr-1-dependent migration of mouse mammary epithelial cells and in human breast cancer cells overexpressing Cr-1 (Bianco et al., 2003; Normanno et al., 2004). EpH4/Cr-1 cells grown in medium containing exogenous rmNetrin-1 (50 ng/ml) showed an increase in the expression of E-cadherin and a decrease in the expression of vimentin (Fig. 3A). In addition, reduced phosphorylation of Akt was observed in Netrin-1-treated EpH4/Cr-1 cells compared to levels in EpH4/Cr-1 cells grown in culture medium without exogenous rmNetrin-1 (Fig. 3A).

Mammary epithelial cells overexpressing Cripto-1 have been shown to possess an increased capacity for invasion and migration (Bianco et al., 2003; Normanno et al., 2004; Strizzi et al., 2004; Wechselberger et al., 2001). In fact, when medium containing 2% FBS was added to the bottom of the wells in the invasion assay, EpH4/Cr-1 or HC-11/Cr-1 cells showed a significant increase in the number of cells that invaded across the matrix-covered membrane (Fig. 3B). However, when 25 or 50 ng/ml soluble rmNetrin-1 was added to the 2% FBS-containing medium in the bottom of the wells, the number of invading EpH4/Cr-1 or HC-11/Cr-1 cells was significantly decreased (Fig. 3B). Furthermore, EpH4/Cr-1 or HC-11/Cr-1 cells that were treated with 50 ng/ml exogenous rmNetrin-1 showed a decrease in Neogenin expression and an increase in UNC5H1 expression as determined by western blotting (Fig. 3C). This suggested that Netrin-1 might bind to UNC5H1 and may therefore be involved in producing the anti-invasive effect that was observed in the Cr-1-overexpressing cells. To address this possibility, we used a neutralizing antibody generated against the extracellular domain of UNC5C with the supposition that this antibody would also block UNC5H1 function, as the ligand binding extracellular domain of UNC5 proteins are highly conserved (Engelkamp, 2002). Indeed, pretreatment of EpH4/Cr-1 cells with a neutralizing antibody against UNC5C restored the invasive properties of EpH4/Cr-1 cells in the presence of rmNetrin-1 (Fig. 3D). In contrast, pretreatment of EpH4/Cr-1 cells with a neutralizing antibody against Neogenin did not restore the invasiveness of the cells, but further decreased the number of invading EpH4/Cr-1 cells across the matrix coated membranes in the presence of rmNetrin-1 (Fig. 3D). Thus treating Cr-1-overexpressing mouse mammary epithelial cells, which have downregulated Netrin-1 expression, with exogenous soluble rmNetrin-1 reverses the biochemical characteristics of Cr-1-dependent EMT, suppresses activation of Akt and reduces cell invasiveness. In accord with reduced cell motility, there is an upregulation of the UNC5 class of Netrin-1 receptors that act to inhibit or to redirect the attractive responses to Netrin-1.

The increased ability of Eph4/Cr-1 cells compared to the wild type, to form colonies in 3D extracellular matrix has been previously determined (Wechselberger et al., 2001). To determine whether exogenous Netrin-1 was capable of affecting colony formation by EpH4/Cr-1 cells, these cells were seeded in 3D Matrigel containing disks of filter paper that had been preabsorbed with different concentrations of Netrin-1. Spherical colonies having diameters ≥500 μm were scored after 48 hours in equally divided areas situated proximal, medial or distal to the disks of blot paper preabsorbed with either PBS or rmNetrin-1. EpH4/Cr-1 colonies were more homogenously distributed throughout the Matrigel containing the PBS source whereas there were significantly fewer colonies formed in the areas proximal to the Netrin-1 source (Fig. 4A). In some of the 3D cultures, EpH4/Cr-1 cells actually formed a noticeable ring of growth inhibition around the source of Netrin-1 (Fig. 4B). This effect was more evident with blot paper preabsorbed with 100 or 200 ng/ml of rmNetrin-1. Blot paper pre-absorbed with 400 ng/ml rmNetrin-1 did not inhibit colony formation of EpH4/Cr-1 cells (data not shown). Quantification of the colonies formed in Matrigel with diameters greater than 500 μm revealed a significant reduction in the number of colonies formed by Eph4/Cr-1 in proximity to the blot paper preabsorbed with rmNetrin-1 compared to blot papers preabsorbed with PBS (Fig. 4C).

When control pellets containing cholesterol were introduced into the virgin mammary gland of 5-week-old FVB/N or MMTV/CR-1 mice, normal ductal elongation was observed in whole mounts of mammary gland from these mice after 2 weeks (Fig. 4D). Ductal elongation in FVB/N mammary glands containing pellets releasing 25 ng/day was similar to that observed in control FVB/N mammary glands (Fig. 4D). However, there was a significant reduction in ductal elongation in the mammary glands of MMTV-CR-1 mice containing pellets releasing 25 ng/day of Netrin-1 (Fig. 4D,E). The inhibitory effect of Netrin-1 on mammary gland ductal elongation was not observed in mice containing pellets releasing greater amounts (50 ng/day) of rmNetrin-1 (data not shown).

Immunohistochemical analysis of the terminal end buds in the mammary glands treated with the Netrin-1 pellets showed an overall increase in expression of UNC5H1 throughout the terminal end bud structures compared to MMTV-CR-1 transgenic mammary glands treated with control pellets where UNC5H1 appeared to stain only the peripheral area of these structures (Fig. 5). Increased staining for E-cadherin was also observed in the terminal end buds of MMTV-CR-1 mammary glands containing the Netrin-1 pellet compared to that in the control (Fig. 5). Finally, positive staining for P-Akt was more intense in the epithelial and stromal cells in mammary glands from MMTV-CR-1 mice containing control pellets compared to mammary glands from MMTV-CR-1 transgenic mice implanted with the Netrin-1-releasing pellets. Moreover, the average intensity of staining for P-Akt in the terminal end buds as evaluated by digital image analysis was significantly reduced by almost one-third (32%) (Range 24-42% reduction; n=5; P=0.008) in MMTV-CR-1 mammary glands containing Netrin-1 pellets compared to terminal end buds in MMTV-CR-1 glands containing control pellets (Fig. 5). Vimentin stained poorly in MMTV-CR-1 mammary glands containing either control or Netrin-1 pellets and showed no difference in the levels of expression between the two different mammary glands (data not shown).

The present study demonstrates that mouse mammary epithelial cells overexpressing Cr-1 in vitro have reduced expression of Netrin-1 and increased expression of Neogenin. Analysis of immunostained histological sections of mammary tumors from MMTV-CR-1 transgenic mice revealed that in the areas where mammary tumors are composed of large anaplastic, mesenchyme-like tumor cells, immunostaining for Netrin-1 was less intense compared to that for Neogenin. These results suggest that overexpression of Cripto-1 is capable of modulating the expression of Netrin-1 and Neogenin.

Cripto-1 is expressed in a greater number of infiltrating ductal carcinomas or intralobular carcinomas than in ductal carcinomas in situ (Panico et al., 1996). In addition, expression of CR-1 is higher in colon tumors than in the adjacent noninvolved colon epithelium and correlates with tumor stage, increased regional lymph node metastases and a higher rate of colorectal cancer recurrence (Gagliardi, 1994; Kuniyasu et al., 1991). In gastric carcinoma, the incidence of CR-1-positive cases was more frequent in late stage, locally invasive tumors than in early stage, noninvasive cancers (Kuniyasu, 1994). From these studies, the association between CR-1 expression and characteristics of tumors undergoing EMT such as local tissue invasion, lymph node metastasis and cancer recurrence (Gotzmann et al., 2004; Thiery, 2003b; Thiery and Chopin, 1999; Vincent-Salomon and Thiery, 2003), suggest that CR-1 may be capable of promoting a more aggressive phenotype in human tumor cells by inducing EMT.

The overexpression of CR-1 in mammary glands of aged multiparous mice leads to the formation of mammary papillary adenocarcinomas (Wechselberger et al., 2005). These tumors show evidence of EMT such as reduced expression of E-cadherin and increased expression of vimentin and snail, especially in areas containing anaplastic, mesenchyme-like tumor cells (Strizzi et al., 2004). In the present study, Netrin-1 expression was reduced relative to Neogenin in these same regions. Moreover, immunostaining for CR-1 was increased in areas of the tumors exhibiting characteristics of EMT compared to the areas of the CR-1 transgenic mouse mammary tumors that possess a more differentiated papillary phenotype (Strizzi et al., 2004). In this regard, the reduction of Netrin-1 expression observed in mouse mammary epithelial cells overexpressing Cr-1 is similar to the reduced expression of Netrin-1 observed in the mesenchyme-like tumor cells in the mammary tumors from the CR-1 transgenic mice. Therefore, the loss of an epithelial phenotype, which is induced by Cripto-1 overexpression in mammary epithelial cells, is associated with a reduction in Netrin-1 expression as mammary epithelial cells exhibit a more motile and mesenchyme-like phenotype.

Targeted inhibition of Cr-1 expression with a specific siRNA resulted in the reversion of Netrin-1 and Neogenin expression to levels that were detected in wild-type mammary epithelial cells, suggesting that Cr-1 may play a role in regulating Netrin-1 and Neogenin expression. As EpH4 cells do not express Nodal (Bianco et al., 2002), and as treatment of EpH4/Cr-1 cells with PP2 or LY was followed by an increase in Netrin-1 expression and a decrease in Neogenin expression, this indicates that Cripto-1 may be affecting Netrin-1 and Neogenin expression by signaling through a pathway that involves Nodal-independent activation of c-Src and/or PI-3K/Akt in EpH4/Cr-1 cells. The addition of exogenous rmNetrin-1 to the culture medium of mouse mammary epithelial cells overexpressing Cr-1 resulted in a reduction in the expression of Neogenin. This suggests that exogenous Netrin-1 is capable of either reverting effects of Cr-1 on Neogenin expression or that there may be a potential negative feedback mechanism by which Netrin-1 regulates the expression of Neogenin. The latter possibility has been described in other systems whereby cells compensate for high concentrations of ligand expression by adjusting the expression of cell surface receptors (Frank et al., 1996).

Local release of Netrin-1 from pellets implanted in vivo in the mammary glands of virgin MMTV-CR-1 transgenic mice but not in control mammary glands of virgin FVB/N mice resulted in an inhibition in the elongation of the developing mammary ducts through the fat pad. This is interesting, as CR-1 overexpression in the mammary gland of virgin MMTV-CR-1 transgenic mice is associated with enhanced ductal side branching and elongation (Wechselberger et al., 2005). The family of UNC5 receptors can mediate, alone or as a coreceptor with Neogenin, the repulsive effects of Netrin-1 (Dickson and Keleman, 2002; Hinck, 2004; Hong et al., 1999). Netrin-1 in the presence of an UNC5 receptor is capable of reducing migration and filipodial extension of endothelial cells in vitro and in vivo (Lu et al., 2004). Likewise, Netrin-1 can cause epithelial cells of the developing lung to migrate away from a Netrin-1 source (Liu et al., 2004). Our in vitro data suggest that UNC5 may be involved in mediating the negative effects of Netrin-1 on ductal elongation. Although overexpression of Cripto-1 in the mouse mammary epithelial cells did not affect the expression of UNC5H1, when Cr-1 overexpressing cells are cultured in the presence of exogenous rmNetrin-1, these cells are found to overexpress UNC5H1 receptors. Immunohistochemical analysis of the mammary terminal end buds of MMTV/CR-1 transgenic mice treated with Netrin-1 pellets showed increased expression of UNC5H1 throughout the terminal end bud compared to mammary terminal end buds in MMTV/CR-1 transgenic mice treated with control pellets, which showed reduced UNC5H1 staining limited to the periphery of the terminal end bud. Thus the UNC5 receptors may also be involved in mediating the reduction in the number of colonies formed in Matrigel and in impairing the migration and invasion of EpH4/Cr-1 cells across matrix-coated membranes in response to exogenous rmNetrin-1. Interestingly, the inhibitory effects of Netrin-1 on colony formation by Eph4/Cr-1 cells and on ductal elongation in the mammary glands of MMTV-CR-1 transgenic mice were not observed at higher doses (data not shown). A similar dose-response effect for Netrin-1 activity on axon outgrowth was also observed and has been attributed to the requirement for Netrin-1 to induce receptor clustering that is impaired at higher concentrations of Netrin-1 (Serafini et al., 1994).

The anti-invasive effects of Netrin-1 on EpH4/Cr-1 cells was significantly attenuated when these cells were preincubated with blocking antibodies against UNC5C, which is highly homologous in the extracellular domain to mouse UNC5H1 (Engelkamp, 2002). However, preincubation of EpH4/Cr-1 cells with an anti-Neogenin blocking antibody did not have the same effect but actually further inhibited the invasion of Eph4/Cr-1 cells across the matrix-coated membranes. This effect may possibly be due to a decrease in attractive cues that the Neogenin receptors may have been capable of inducing in EpH4/Cr-1 cells in response to Netrin-1. This result also suggests that it is unlikely that Neogenin is acting alone (Rajagopalan et al., 2004) or as a coreceptor with UNC5 (Hong et al., 1999) in mediating repulsion in EpH4/Cr-1 cells. However, it cannot be excluded that, alternatively or in combination with the UNC5 receptor, Netrin-1 may have affected Cr-1-dependent invasion and colony formation of mammary epithelial cells and ductal elongation in mammary glands by blocking Cripto-1 signaling. In fact, Akt activity was reduced in EpH4/Cr-1 cells when cultured in the presence of exogenous Netrin-1. The reduction in P-Akt expression was also detected by immunohistochemistry in epithelial cells of the terminal end buds and surrounding stromal cells of MMTV/CR-1 mammary glands containing Netrin-1 pellets compared to MMTV/CR-1 mammary glands containing control pellets.

Mechanisms that regulate cell adhesion and migration play a fundamental role not only during normal tissue development and differentiation but also in the survival and spread of tumor cells (Cavallaro and Christofori, 2004; Lee and Juliano, 2004; Thiery, 2003a; Van Roy and Mareel, 1992). In this respect, factors that may affect Netrin-1-dependent adhesion in mammary epithelial cells should provide some insight into the mechanisms involved in the spread of potential tumor cells. The association between overexpression of Cripto-1 and induction of biochemical changes important for EMT, such as reduction of E-cadherin levels and increased expression of vimentin both in vitro and in vivo, has been previously described (Ebert et al., 2000; Strizzi et al., 2004). Furthermore, Cripto-1 induces morphologic changes and activation of signaling molecules known to enhance cell migration and invasion (Bianco et al., 2003; Normanno et al., 2004; Wechselberger et al., 2001). These findings support a potential role for Cripto-1 during tumorigenesis.

From our data it appears that a possible mechanism by which Cripto-1 may induce a more aggressive phenotype in mammary epithelial cells is by reducing Netrin-1 expression and affecting the profile of Netrin-1 receptor expression. Exogenous Netrin-1 was capable of increasing E-cadherin and decreasing vimentin expression in EpH4/Cr-1 thus reversing Cr-1-dependent biochemical changes in vitro, which are important for EMT. Immunohistochemical analysis of the mammary terminal end buds from MMTV/CR-1 transgenic mice treated with Netrin-1 pellets also showed an increase in E-cadherin expression compared to mammary terminal end buds from MMTV/CR-1 mammary glands treated with control pellets. However, immunohistochemical analysis showed poor expression for vimentin in the mammary glands analyzed from both the Netrin-1-treated and control MMTV/CR-1 transgenic mice. This latter observation may be due to the fact that hyperplastic lesions and mammary tumors in which EMT markers, including vimentin, are overexpressed were identified in aged, multiparous mice (Strizzi et al., 2004). In the present study, the Netrin-1 pellets were implanted in virgin 5-week-old MMTV/CR-1 mice and mammary glands were harvested shortly after, suggesting that prolonged exposure of mammary epithelial cells to CR-1 effect is probably needed in order to induce changes in vimentin expression. The increased invasion and migration in epithelial cells is also a major characteristic of Cripto-1-induced EMT. Netrin-1 was capable of reducing Cr-1-dependent invasion and migration of mammary epithelial cells both in vitro and in vivo. As increased expression of Cripto-1 has been detected in a number of human cancers, including breast cancer, it will be informative to investigate the exact relationship between Netrin-1 and Cripto-1 expression in these tumors.

The Authors thank Brenda Wallace-Jones for her excellent technical assistance. L.H. was supported by funds from the American Cancer Society Research Scholar Grant #RSG0218001MGO. N.N. was supported by funds from the Italian Association for Cancer Research (AIRC).