P-cadherin expression in breast carcinomas has been associated with tumors of high histologic grade and lacking estrogen receptor-α, suggesting a link between these proteins. In the MCF-7/AZ breast cancer cell line, blocking estrogen receptor-α signaling with the antiestrogen ICI 182,780 induced an increase of P-cadherin, which coincided with induction of in vitro invasion. Retroviral transduction of MCF-7/AZ cells, as well as HEK 293T cells, showed the proinvasive activity of P-cadherin, which requires the juxtamembrane domain of its cytoplasmic tail. This study establishes a direct link between P-cadherin expression and the lack of estrogen receptor-α signaling in breast cancer cells and suggests a role for P-cadherin in invasion, through its interaction with proteins bound to the juxtamembrane domain.
Classical cadherins are a superfamily of transmembrane glycoproteins responsible for calcium-dependent cell–cell adhesion, mediating homophilic protein interactions (1) . These are modulated by their conserved cytoplasmic juxtamembrane domain and catenin-binding domain, linking them to the actin cytoskeleton. β-, γ-, p120-, and α-Catenins are the best-documented interaction partners (2) . β-Catenin (and perhaps also γ-catenin) is a signaling molecule, implicated in tissue patterning, of which the functions are regulated by binding to the catenin-binding domain of cadherins and by interactions with receptor tyrosine kinases and transcription factors of the lymphocyte enhancer factor/T-cell factor family (2) . P120-catenin was identified as a substrate for Src and several receptor tyrosine kinases and interacts directly with the juxtamembrane domain of cadherins, modulating cadherin clustering and cell motility in a cell-type and phosphorylation state-dependent way (3) . The cadherin/catenin junctional complex is linked to the actin cytoskeleton via α-catenin, thus strengthening its adhesive force (1) .
Reduced expression of E-cadherin is associated with tumor progression in many different cancers, including breast cancer (4) , and may result from mutations, loss of heterozygosity, promoter hypermethylation, or up-regulation of transcriptional repressors, as SIP1, Snail, Slug, or Twist (1) . Moreover, the invasion suppressor function of normally expressed E-cadherin may be overcome by the aberrant expression of N-cadherin (5) or cadherin-11 (6) , which have been associated with progression of breast carcinoma through interference with E-cadherin function (7) .
P-cadherin, another classical cadherin, is expressed in ectodermal tissues, more specifically in the basal layers of stratified epithelia (8 , 9) and in myoepithelial cells of the breast (10) . P-cadherin is implicated in growth and differentiation, as evidenced by knockout mice displaying precocious differentiation of the mammary gland (11) , and is aberrantly expressed in mammary carcinomas of high histologic grade and with a poor prognosis (12, 13, 14, 15, 16) , as well as in other types of carcinomas and proliferative inflammatory lesions (17, 18, 19) . It has been suggested that suppression of the P-cadherin gene is lost during carcinogenesis (9) , but the nature of this mechanism and the biological role of the newly acquired P-cadherin remain to be investigated.
Because aberrant expression of P-cadherin identified a subgroup of estrogen receptor-α-negative breast carcinomas (16) , we raised the hypothesis that the expression of P-cadherin in mammary epithelial cells is hormonally regulated, as described for E-cadherin (20) , N-cadherin (21) , and cadherin-11 (22) .
In mammary epithelial cells, estrogen receptor-α is a key regulator of proliferation and differentiation and a crucial prognostic indicator and therapeutic target in breast cancer. Estrogen receptor-α is a ligand-dependent transcription factor acting through direct transcriptional target activation (23) . Estradiol acts as a potent mitogen for many breast cancer cell lines, and ∼70% of breast carcinomas are estrogen receptor-α positive. This mitogenic effect is blocked by estrogen antagonists. Pure antiestrogens (like ICI 182,780) and selective estrogen receptor modulators (like tamoxifen; ref. 24 ) are used for the treatment of osteoporosis, breast cancer, and other diseases. Continuous exposure of steroid–hormone-responsive breast cancer cell lines to ICI 182,780 leads to resistant sublines, with signaling pathways alternative to estrogen receptor-α (25) . Similarly, in breast cancer, a high number of patients eventually develop antiestrogen resistance for unknown reasons.
Using the antiestrogen ICI 182,780, we investigated a putative molecular and functional link between the absence of estrogen receptor-α signaling and P-cadherin expression in breast cancer cells. To understand the relationship between P-cadherin and the aggressive breast cancer phenotype, we studied the effect of wild-type P-cadherin and several mutants on cell aggregation and invasion. We report that aberrant expression of P-cadherin may result from a lack of estrogen receptor-α signaling and may induce cell invasion in a juxtamembrane domain-dependent manner.
MATERIALS AND METHODS
Plasmids and cDNA Constructs.
The hP-cad/pBR322–23-b expression vector, containing the 3.2kb cDNA encoding full-length human P-cadherin (8) , was kindly provided by Prof. Keith R. Johnson (Department of Oral Biology, College of Dentistry and the Eppley Cancer Center, Nebraska Medical Center, Omaha, NE), with the permission from Prof. Yukata Shimoyama (Department of Surgery, International Catholic Hospital, Nakaochiai, Shinjuku, Tokyo, Japan). The cDNA encoding full-length mouse E-cadherin was kindly provided by Jolanda van Hengel (Department of Molecular Biomedical Research, VIB-Ghent University, Ghent, Belgium). Both cDNAs (PC-WT and mEC-WT) were transferred to the expression vector pIRES2-EGFP (Clontech, Palo Alto, CA), allowing easy evaluation of transfection efficiencies due to coexpression of enhanced green fluorescent protein (EGFP). To generate P-cadherin deletion mutants, P-cadherin was EcoRI subcloned into pBluescript (Promega, Madison, WI) and NdeI/SalI digested to remove the region encoding its COOH-terminal tail. PCR fragments corresponding to different lengths of the removed tail, flanked by NdeI/SalI restriction enzyme digest sites at the 5′ and 3′ ends, respectively, were obtained always using the same sense primer (5′-AAGCAGGATACATATGACGTG-3′) and different antisense primers for the following constructs: PC-CT682: 5′-CGTGCCGTCGACCTACTTCCGCTTCTT-3′; PC-CT702: 5′-CGTGCCGTCGACCTAGCCATAGTAGAA-3′; PC-CT711: 5′-CGTGCCGTCGACCTACTGGTCCTCTTC-3′; PC-CT719: 5′-CGTGCCGTCGACCTAGTGGAGCTGGGT-3′; PC-CT727: 5′-CGTGCCGTCGACCTACTCCGGCCTGGC-3′; and PC-CT762: 5′-CGTGCCGTCGACCTACAGGTTCTCAAT-3′. After NdeI/SalI digestion, these products were ligated into NdeI/SalI digested pP-cad-Bluescript, and the resulting construct was EcoRI/SalI transferred to pIRES2-EGFP. Additionally, a mutant with a small deletion in the P120-catenin-binding sequence (lacking the nucleotides coding EEGGG) and retaining the intact catenin-binding domain was created (PC-Δ703–707). Therefore, pP-cad-IRES2-EGFP was XhoI/SmaI digested, and the removed fragment was cut with EarI. After removal of the small fragment between the two EarI restriction sites (encoding EEGGG), the two remaining fragments (XhoI/EarI and EarI/SmaI) were ligated into XhoI/SmaI digested pP-cad-IRES2-EGFP. To create the P-cadherin point mutant (PC-R503H), a PCR product, encompassing the point mutation, was obtained, using the following primers: a sense primer (5′-GGCACCCTCGACCATGAGGATGAG-3′), with the TaqI restriction site in italics and the point mutation in bold; and the antisense primer used for generating PC-CT762. This product was TaqI/NdeI digested and used in a three-point ligation with a BamHI/TaqI and a BamHI/NdeI fragment of pP-cad-Bluescript, followed by EcoRI transfer of the construct to pIRES2-EGFP. Direct sequencing (ABI, Perkin-Elmer, Foster City, CA) was performed for all of the constructs to confirm their integrity.
Restriction Enzymes, Antibodies, and Chemical Reagents.
All of the restriction enzymes were purchased from New England BioLabs (Beverly, MA). Antihuman primary mouse monoclonal antibodies used were against P-cadherin (clone 56) and P120-catenin (clone 98; BD Transduction Laboratories, Lexington, KY), N-cadherin (CH-19 and GC-4), α-tubulin (B-5–1-2; Sigma-Aldrich, Bornem, Belgium), E-cadherin (HECD-1; Takara Biochemicals, Kyoto, Japan), and estrogen receptor-α (NCL-l-ER-6F11; Novocastra, Newcastle, United Kingdom). 17β-Estradiol was purchased from Sigma-Aldrich Química (Sintra, Portugal) and ICI 182,780 was kindly provided by AstraZeneca (Barcarena, Portugal). Both drugs were dissolved in 100% EtOH and added to the culture media. The concentrations used were 10 nmol/L for 17β-estradiol and 100 nmol/L for ICI 182,780, unless mentioned otherwise. Cycloheximide was obtained from Sigma and used at 25 μg/mL. For the control situations, cells were treated only with 100% EtOH.
Cells and Transient Transfection.
Human cancer cell lines were obtained as described: BT-20 from Peter Coopman (Laboratory of Molecular Biology, Ghent University, Belgium), MCF-7/AZ (MCF7) from Per Briand (The Fibiger Institute, Copenhagen, Denmark), ZR-75.1 and T47D from American Type Culture Collection (Manassas, VA), and HEK 293T (HEK) cells from Veerle De Corte (Department of Biochemistry, Faculty of Medicine and Health Sciences, VIB-Ghent University, Belgium). Cell lines were routinely maintained at 37°C, 10% CO2, in the following media (Invitrogen, Merelbeke, Belgium): 50% DMEM/50% HamF12 (MCF7), DMEM (BT-20, T47D, HEK), or RPMI 1640 (ZR-75.1). All of the media for routine culture contained 10% heat-inactivated fetal bovine serum (Greiner bio-one, Wemmel, Belgium), 100 IU/mL penicillin, 100 μg/mL streptomycin, and 2.5 μg/mL amphotericin B (Invitrogen). To obtain transient transfectants, appropriate expression vectors (2.5 μg) were introduced into HEK cells (2 × 105) with Fugene (Roche Molecular Biochemicals, Mannheim, Germany), and transfection efficiencies were evaluated by measuring EGFP expression by flow cytometry.
The P-cadherin coding sequence was EcoRI digested from pIRES2-EGFP and EcoRI subcloned into the LZRS-IRES-EGFP vector to generate the LZRS-P-cad-IRES-EGFP vector. The LZRS-IRES-EGFP retroviral vector, encoding only EGFP, was used as a control. For the production of retroviral supernatant, the Phoenix-Amphotropic packaging cell line (a kind gift from Dr. Garry P. Nolan, Stanford University School of Medicine, Stanford, CA) was transfected with the LZRS-IRES-EGFP and the LZRS-P-cad-IRES-EGFP plasmids using calcium-phosphate precipitation (Invitrogen) to generate both retroviruses. The viral supernatant was spun (10 minutes at 350 × g), and aliquots were stored at −70°C until use. Transduction of cell lines was performed as described before (26) .
Flow Cytometry Staining and Cell Sorting.
For analysis of E- and N-cadherin surface expression, cells were detached under cadherin saving procedures (27) , and ∼1 × 105 cells were used for staining. Cells were washed with cold PBS containing bovine serum albumen (BSA) and incubated for 30 minutes with the anti-E-cadherin HECD-1 or anti-N-cadherin GC-4 antibodies. This was followed by two washes with PBS/BSA, 30 minutes incubation with biotinylated rabbit antimouse monoclonal antibody, two washes with PBS/BSA, 20 minutes incubation with streptavidin-phycoerythrin, and a final wash with PBS/BSA. For routine analysis of EGFP expression, cells were detached with trypsin/EDTA. Flow cytometric analysis and/or cell sorting were performed as described before (26) .
Biotinylation, Immunoprecipitation, and Immunoblotting.
Immunoprecipitation and immunoblotting experiments and quantification of bands, were performed as published before (28) . For biotinylation, the cells were washed three times with cold PBS and incubated with 0.5 mg/mL of Biotinylation Reagent (EZ-Link Sulfo-NHS-LC-Biotin, Pierce) during 30 minutes at 4°C, followed by four washes with cold PBS before cell lysis. To control for equal loading of total lysates, immunostaining with anti-α-tubulin was performed routinely (not always shown). Each immunoblot was done at least three times, and the ones that were selected to show are representative experiments.
Reverse transcription-PCR (RT-PCR) experiments were done as described previously (28) . Primers specific for P-cadherin cDNA included the following: sense 5′-ACGAAGACACAAGAGAGATTGG-3′ and antisense 5′-CGATGATGGAGATGTTCATGG-3′, to generate a 287-bp product. PCRs were done in a Minicycler (Biozym, Landgraaf, the Netherlands) with an annealing temperature of 55°C.
Slow Aggregation Assays.
For semi-solid substratum, 2 × 104 cells in 200 μL medium were seeded on solidified agar in a 96-well plate (27) . Aggregate formation was evaluated under an inverted microscope after 24, 48, and 72 hours. In suspension, 6 × 105 cells were added to 50 mL-Erlenmeyer flasks in 6 mL of medium. The flasks were incubated on a Gyrotory shaker (New Brunswick Scientific Co., New Brunswick, NJ) at 72 rpm and continuously gassed with humidified 10% CO2 in air. The particle size distribution of the aggregates was measured with a Coulter Particle Size Counter (LS2000, Coulter Company, Miami, FL). The diameter of the particles can be considered as a measure for aggregate formation. Statistical analysis of differences between the particle size distribution curves was done with the Kolmogorov-Smirnov method.
For collagen type I invasion assay (29) , six-well plates were filled with 1.3 mL of neutralized collagen type I (0.09% w/v, Upstate Biotechnology, Inc., Lake Placid, NY) and incubated for at least 1 hour at 37°C to allow gelification. 1 × 105 cells of a single-cell suspension were seeded on top of the gel, and cultures were incubated at 37°C for 24 hours. Using an inverted microscope controlled by a computer program, invasive and superficial cells were scored blind-coded in 12 fields of 0.157 mm2. The invasion index expresses the percentage of cells invading into the gel over the total number of cells counted. For Matrigel invasion assay, transwell chambers with polycarbonate membrane filters (6.5 mm diameter, 8 μm pore size, Costar, Corning, NY) were coated with 20 μL of a Matrigel solution (Becton Dickinson). 1 × 105 cells were added to the upper compartment of the chamber. In the lower compartment, conditioned cell culture medium of the MRC-5 human embryonic lung fibroblast cell line was added as a chemoattractant. After 24 hours of incubation at 37°C, the upper surface of the filter was cleared from nonmigratory cells with a cotton swab and washes with serum-free DMEM. The remaining (invasive) cells at the lower surface of the filter were fixed with cold methanol and stained with 4′, 6-diamidino-2-phenylindole (Sigma, 0.4 mg/mL). Invasive cells were scored by counting 50 fields per filter with a fluorescence microscope, at ×250 of magnification. Rat myofibroblast DHD-FIB cells were routinely included as a positive control for invasion in both assays. Each experiment was repeated at least three times. For collagen invasion assay, data are expressed as mean ±SD; for Matrigel invasion assay, a representative experiment is shown, with the SD for the number of cells scored on the 50 microscopic fields. Statistical significance was determined by t test, and differences between groups were analyzed using the ANOVA; P < 0.05 was considered significant.
The Antiestrogen ICI 182,780 Up-Regulates P-Cadherin in Estrogen Receptor-α–Positive Breast Cancer Cell Lines.
To test the hypothesis that estrogen receptor-α negatively regulates P-cadherin, we examined the expression of estrogen receptor-α and cadherins in breast cancer cell lines by Western blot (Fig. 1A) ⇓ . Interestingly, higher levels of P-cadherin were found in estrogen receptor-α–negative BT-20 cells.
A 24-hour treatment with the antiestrogen ICI 182,780 (10−7 mol/L) increased P-cadherin protein levels in MCF7 and ZR-75.1 cells but not in BT-20 cells (Fig. 1B) ⇓ . There were no significant changes in P-cadherin levels observed in T47D cells, bearing already higher pretreatment levels of P-cadherin and lower levels of estrogen receptor-α than the responsive cell types. ICI 182,780-induced increase of P-cadherin was associated with a decline of estrogen receptor-α levels (Fig. 1B) ⇓ .
For additional investigation, we chose the MCF7 cell line, because it is estrogen receptor-α positive, highly responsive to estrogen, and extensively investigated as a model of breast cancer. In these cells, ICI 182,780 induced, respectively, up- and down-regulation of P-cadherin and estrogen receptor-α in a time- and dose-dependent way (Fig. 2, A and B) ⇓ . A decrease of estrogen receptor-α levels was already observed after 6 hours of treatment, whereas P-cadherin levels nearly doubled after 12 hours. After 24 hours of exposure to ICI 182,780, higher P-cadherin and lower estrogen receptor-α levels persisted for several days, with normalization 96 hours after ICI 182,780 withdrawal (Fig. 2C) ⇓ . To examine whether or not the effect of ICI 182,780 on P-cadherin expression was mediated via estrogen receptor-α, we did a competition experiment. As already described (30) , 17β-estradiol readily decreased estrogen receptor-α levels, although to a lesser extent than ICI 182,780 (Fig. 2D) ⇓ . Importantly, 17β-estradiol counteracted the ICI 182,780-induced up-regulation of P-cadherin (Fig. 2D) ⇓ and accelerated normalization of P-cadherin levels in cells treated for 24 hours with ICI 182,780 (Fig. 2E) ⇓ . Together, these results suggest that not the decrease in estrogen receptor-α, but the lack of estrogen receptor-α signaling is responsible for the increase of P-cadherin by ICI 182,780.
RT-PCR revealed an increase of P-cadherin mRNA after ICI 182,780 treatment, suggesting that the higher P-cadherin protein expression results from an up-regulation of P-cadherin transcripts (Fig. 2F) ⇓ . This was confirmed by a micro-array study performed on 17β-estradiol– or ICI 182,780-treated MCF7 cells, in which 17β-estradiol did not alter P-cadherin mRNA levels, whereas ICI 182,780 induced an 8-fold increase. Finally, it remained to be determined whether induction of the P-cadherin gene (CDH3) was a direct effect of ICI 182,780 or required prior induction of other genes. We addressed this question by blocking protein synthesis in cells, because the induction of primary target proteins or immediate early genes should not be sensitive, whereas secondary targets should be blocked. The treatment of MCF7 cells with cycloheximide, a de novo protein synthesis inhibitor, largely blocked P-cadherin up-regulation by ICI 182,780 (Fig. 2G) ⇓ , which is consistent with a requirement for newly synthesized proteins, probably induced by ICI 182,780, before CDH3 activation. In contrast, as expected, this drug did not block estrogen receptor-α down-regulation mediated by ICI 182,780 (Fig. 2G) ⇓ .
ICI 182,780 Decreases Cell–Cell Adhesion and Increases Invasiveness of MCF-7/AZ Cells.
MCF7 cells formed compact aggregates on top of soft agar or when incubated in Erlenmeyer flasks under continuous shaking (Fig. 3A ⇓ , panel i, and Fig. 3B ⇓ ). In presence of ICI 182,780, this effect was counteracted (Fig. 3A ⇓ , panel ii, and 3B ⇓ ). Even a 24-hour pretreatment with ICI 182,780, followed by testing these cells in the absence of ICI 182,780, was sufficient to prevent the formation of large aggregates (Fig. 3A ⇓ , panel iii). On plastic substratum, no changes in morphology or migrating behavior (as measured by a wound healing assay) could be observed upon treatment with ICI 182,780 (data not shown).
Whereas MCF7 cells failed to invade in collagen type I and Matrigel invasion assays, a 24-hour pretreatment with ICI 182,780 was sufficient to induce invasion of these cells in both assays (Fig. 3, C and D) ⇓ . These proinvasive effects of ICI 182,780 were counteracted by 17β-estradiol (Fig. 3, C and D) ⇓ , indicating that they are mediated by interference with estrogen receptor-α signaling.
Aggregation and invasion of MCF7 cells, in the presence of ICI 182,780, mimics the behavior of the poorly aggregating and invasive estrogen receptor-α–negative and P-cadherin–positive BT-20 cells (Fig. 1A) ⇓ , which remained unchanged upon treatment with ICI 182,780 (Fig. 3, E and F) ⇓ .
P-Cadherin Expression Increases Invasiveness but Does Not Alter Cell–Cell Adhesion of MCF-7/AZ cells.
Cells, retrovirally transduced to encode only EGFP (MCF7.LIE) or both P-cadherin and EGFP (MCF7.P-cad), were sorted to >90% EGFP positivity (Fig. 4A) ⇓ . P-cadherin levels were higher at the cell surface in P-cadherin–transduced cells (Fig. 4B) ⇓ . The levels of cell-surface E-cadherin were the same in P-cadherin–transduced cells, as in vector-transduced cells (Fig. 4, A and B) ⇓ , excluding an effect of the exogenous cadherin on the levels of the major endogenous cadherin.
On plastic substratum, P-cadherin–transduced MCF7 cells, like their parental or vector-transduced cells, formed epithelioid islands, showing no morphotype differences (data not shown). Transduction with P-cadherin did not interfere with E-cadherin–mediated cell–cell adhesion (Fig. 4, C and D) ⇓ . However, in a wound healing migration assay, P-cadherin–transduced cells migrated faster (data not shown) and, in contrast to parental or vector-transduced (LZRS-IRES-EGFP) controls, invaded into collagen type I and Matrigel (Fig. 4, E and F) ⇓ .
P-Cadherin-Induced Invasion Is Not Breast Cancer Cell or Endogenous Cadherin-Specific.
P-cadherin retroviral transduction was also done on HEK cells, expressing at their surface low and high levels of E- and N-cadherin, respectively (Fig. 1A ⇓ and Fig. 5A ⇓ ), and being invasive neither into collagen type I nor into Matrigel. Sorting of vector- or P-cadherin–transduced cells resulted in populations having either moderate or high EGFP expression (HEK.LIE.Med, HEK.LIE.High, HEK.P-cad.Med, and HEK.P-cad.High; Fig. 5, A and B ⇓ ). As for MCF7 cells, no differences in morphotype or aggregation were observed between parental and transduced cells (Fig. 5, C and D) ⇓ . Although there was a down-regulation of superficial N-cadherin in the highest P-cadherin–expressing cells (Fig. 5B) ⇓ , this did not result in a significant decrease in total levels of N-cadherin (Fig. 5A) ⇓ . P-cadherin–transduced cells were significantly more invasive into collagen type I or Matrigel than vector-transduced cells, with higher invasiveness of the cells expressing more P-cadherin (Fig. 5, E and F) ⇓ . In both assays, the control cells with higher LZRS-IRES-EGFP expression levels showed an increased invasion index when compared with the ones with moderate levels of expression. This may be due to the insertion of viral promoters into the host genome, leading to the aberrant activation of host genes. However, although this observation highlights the care that should be taken when using these systems, it does not influence the interpretation of our results as such: the values of the P-cadherin–transduced cells remain significantly different from those of the respective vector-transduced cells.
P-Cadherin Mediates Invasion of HEK 293T Cells via Its Juxtamembrane Domain.
To identify the P-cadherin domain(s) necessary for its proinvasive effects, we used several P-cadherin constructs (Fig. 6A) ⇓ for transient transfection of the HEK cell line. Biotinylation and immunoblotting confirmed expression of all of the constructs at the plasma membrane (Fig. 6B) ⇓ . Transient transfection with P-cadherin induced invasion into collagen type I, as observed with stably transduced HEK cells (Fig. 6C) ⇓ .
The P-cadherin point mutant, PC-R503H (Fig. 6A) ⇓ , representing the missense mutation in CDH3, found in hypotrichosis with juvenile macular dystrophy (31) , failed to support strong cell–cell adhesion unlike wild-type P-cadherin. 6 Most likely, the reason for this failure is the disruption of the strongly conserved LDRE Ca2+-binding motif in the fourth extracellular domain of P-cadherin. Nevertheless, PC-R503H still induced invasion (Fig. 6C) ⇓ .
Mutants of the P-cadherin cytoplasmic tail were also generated (Fig. 6A) ⇓ . Transfection into HEK cells showed that PC-CT762, retaining the intact P-cadherin juxtamembrane domain, induced invasion-like wild-type P-cadherin (Fig. 6C) ⇓ . Because this mutant is truncated just before the catenin-binding domain, we assume that β-catenin, γ-catenin, or any other protein that binds to this region are not needed for P-cadherin–mediated invasion.
With mutants within the juxtamembrane domain (Fig. 6A) ⇓ , statistically significant invasion into collagen was seen only with the truncation mutants that still retained the intact juxtamembrane domain (PC-CT719 and PC-CT727; Fig. 6C ⇓ ). The somewhat decreased ability of PC-CT719 to induce invasion (Fig. 6C) ⇓ might be due to its lower expression levels (Fig. 6B) ⇓ . In line with the results obtained with the truncation mutants and confirming that the catenin-binding domain is not involved in the proinvasive effects, the PC-Δ703–707 mutant (lacking EEGGG in the P120-catenin binding site), with impaired P120-catenin binding (Fig. 6D) ⇓ , was not able to induce invasion of HEK cells into collagen type I (Fig. 6C) ⇓ . In conclusion, P-cadherin needs its intact juxtamembrane domain to induce invasion of HEK cells into collagen type I.
To exclude that the gain of any exogenous cadherin, retaining its juxtamembrane domain, would be sufficient for a proinvasive effect, we demonstrated that HEK cells transfected with mouse wild-type E-cadherin cDNA (Fig. 6A) ⇓ failed to invade into collagen type I (Fig. 6C) ⇓ . In conclusion, the juxtamembrane domain of P-cadherin confers to this molecule the specific ability to induce invasion of HEK cells, in the presence of the endogenously expressed cadherin.
We demonstrated that the antiestrogen ICI 182,780 increased time- and dose-dependently P-cadherin expression in estrogen receptor-α–positive breast cancer cells. This increase could be completely reverted by 17β-estradiol, categorizing CDH3 as an estrogen-repressed gene and pointing to 17β-estradiol as a key regulator of this cadherin. In addition to competing for binding to estrogen receptor-α, ICI 182,780 also increases its breakdown (24) . As a result, ICI 182,780 abrogates estrogen receptor-α signaling and the subsequent regulation of 17β-estradiol responsive genes. Because the human P-cadherin promoter (GI: 2950171) does not contain the consensus sequence 5′-GGTCAnnnTGACC-3′ of the estrogen-responsive elements (32) , 17β-estradiol is unlikely to have a direct inhibitory effect on transcription of the CDH3 gene. Instead, the increase of P-cadherin by ICI 182,780, some hours after the decrease of estrogen receptor-α, and its inhibition by cycloheximide, pleads for the existence of a CDH3-regulating transcription factor. In the absence of estrogen receptor-α signaling (as in estrogen receptor-α–positive cells treated with ICI 182,780 or in estrogen receptor-α–negative cells), this 17β-estradiol–regulated factor might account for the high P-cadherin levels in some breast cancer cell lines and for the inverse correlation between estrogen receptor-α and P-cadherin expression in mammary tumors.
In MCF7 breast cancer cells, ICI 182,780 treatment led to a decreased cell–cell adhesion and promotion of invasion in vitro. This is in line with the finding that 17β-estradiol (33) and even the unliganded receptor (34) may decrease in vitro invasiveness and motility of breast cancer cells, suggesting that some estrogen-regulated genes negatively control invasion. Because this control is lost in cells treated with high concentrations of ICI 182,780, which up-regulate P-cadherin, the effect of the latter was additionally investigated on in vitro aggregation and invasion of cells retrovirally transduced with P-cadherin. Surprisingly, retroviral transduction of MCF7 and HEK cells with P-cadherin had no detectable influence on cell–cell adhesion. This result suggests that P-cadherin does not shift the aggregation balance established by the other cadherins in these systems. By contrast, such balance may well be changed for invasion, as demonstrated with P-cadherin–transduced cells. It should be noted that this does not allow us to draw conclusions about the necessity of P-cadherin up-regulation for ICI 182,780-induced invasion of MCF7 cells. In contrast to P-cadherin–transduced cells, which migrated faster than controls in a wound healing migration assay, ICI 182,780-treated cells did not. This might be due to the fact that the extent by which P-cadherin is up-regulated by ICI 182,780 may not be sufficient to promote motility as such or, alternatively, the growth-inhibitory effect of ICI 182,780 nullified the promigratory effect of P-cadherin in this assay. Furthermore, ICI 182,780 up-regulated additional proinvasive genes in MCF7 cells, such as MMP-2 and -9, of which the expression was not influenced by P-cadherin (data not shown). Hence, whereas high levels of P-cadherin may be sufficient for induction of invasion, ICI 182,780-induced invasion might require the synergistic action of multiple genes. This hypothesis, in which a critical level of P-cadherin seems to be needed for its proinvasive activity, is supported by the comparison between the invasive and highly P-cadherin–positive BT-20 and T47D cells and the noninvasive and weakly P-cadherin–positive MCF7 and ZR-75.1 cells (Fig. 1A) ⇓ . In contrast to its proinvasive activity in our cells, transfection of other cell lines with P-cadherin inhibited invasion (35 , 36) , 6 suggesting that P-cadherin may act both as an invasion promoter and suppressor, depending on the cell type and its invasive status. Transgenic mice expressing high levels of P-cadherin in the normal mammary epithelium (37) contributed little to this issue, because they did not produce tumors, and because neu oncogene-induced mammary tumors in P-cadherin transgenic mice were always P-cadherin negative.
In the present study, the proinvasive action of P-cadherin is unlikely to be the result of alterations in cell–cell adhesion, because the assays score invasion of single cells into or through a matrix, the retroviral transduction of MCF7 and HEK cells with P-cadherin did not change aggregation, and the point mutant PC-R503H, incapable of supporting strong P-cadherin mediated adhesion, still induced invasion. We presume that the proinvasive activity of P-cadherin is due to changes in signaling pathways.
Recently, Wong and Gumbiner (38) attributed the anti-invasive activity of wild-type E-cadherin to its interaction with β-catenin. An E-cadherin mutant, retaining the catenin-binding domain but with a point mutation that abolishes P120-catenin binding, was still able to suppress invasion. By contrast, in P-cadherin, maintenance of the juxtamembrane domain is crucial for the induction of invasion, irrespective of the catenin-binding domain. Although the juxtamembrane domain is highly conserved between cadherins, its function is very context-dependent, being implicated in both positive and negative regulation of cadherin activity. Cells expressing mutated E-cadherin juxtamembrane domain are weakly adherent (39) , more motile, but still epithelioid. Upon formation of adhesive contacts, the juxtamembrane domain recruits and activates Rac, regulating the actin cytoskeleton (40) . In another context, the juxtamembrane domain may inhibit aggregation mediated by classical cadherins and induce cell motility (41 , 42) or, alternatively, exclude another cadherin from junctions and regulate cell proliferation (43) . Via its binding to P120-catenin, this domain has been implicated recently in maintenance of the stability of endogenous cadherins (44 , 45) . Thus, a possible mechanism for the induction of invasion by P-cadherin might be its competition with the endogenous cadherin for the available P120-catenin, leading to the destabilization of pre-existing anti-invasive cadherin/catenin complexes. Yet, we consider this possibility less likely. Although the down-regulation of N-cadherin in HEK cells by high levels of the several P-cadherin constructs coincided with stimulation of invasion (Fig. 6C ⇓ and Supplementary Data), moderate P-cadherin expression levels, leaving the endogenous cadherin unchanged, were sufficient to induce invasion. Furthermore, transfection of HEK cells with E-cadherin did not induce invasion (Fig. 6C) ⇓ , despite decreased endogenous cadherin in highly expressing cells (Supplementary Data) and expected competition for cadherin-binding proteins.
Alternatively, P-cadherin may generate a specific proinvasive signal via its juxtamembrane domain. In this hypothesis, the binding of proteins to the P-cadherin juxtamembrane domain may differ from their binding to E- or N-cadherin by strength, conformation, or recruitment of other members of the complex. This, in turn, may result in the activation of pathways that overcome the suppressive signals mediated by the endogenous cadherins.
Although binding of proteins to the juxtamembrane domain of P-cadherin has just been documented for P120-catenin (46) , other molecules, like Hakai and presenilin-1 (PS-1), have been reported to bind to the juxtamembrane domain of classical cadherins as well, to a sequence adjacent to or overlapping the P120-catenin–binding domain, thereby competing with P120-catenin for binding (47 , 48) . Although the significance of these interactions is not well known, we cannot exclude the possibility that disruption of the P120-catenin–binding sequence introduces conformational changes and/or uncouples the interaction of these or other proteins, which could be responsible for our observations. Striking examples of this were shown for E-cadherin, where functional differences have been noted between larger and minimal deletions of the juxtamembrane domain, with even the minimal changes disrupting binding of multiple molecules (47) .
Data about the role of P120-catenin in normal and cancer cells are conflicting. Positive and negative regulation of cell–cell adhesion and motility possibly reflect differences in cell type, cadherins, P120-catenin isoforms, and shuttling between cadherin-bound and cytoplasmic pools (3) . When overexpressed in the cytoplasm, P120-catenin may regulate the actin cytoskeleton and cell motility, through Rho GTPases (49) . Similar to the differences seen between E-and N-cadherin in terms of strength (3) and preference (50) of binding to distinct P120-catenin isoforms, P120-catenin binding to P-cadherin may be unique. This unique interaction may influence its impact on the activity of the Rho GTPases, possibly making the cells more prone to invade. Alternatively, the panel of molecules recruited by P120-catenin may differ depending on the isoform or on the cadherin it is bound to.
In conclusion, our study establishes an as yet unknown role for P-cadherin in cancer: (1) P-cadherin expression is regulated through estrogen receptor-α signaling, suggesting that the inverse in vivo correlation between these molecules stems from a causal relationship; (2) P-cadherin induces invasion, in the context of endogenous E- or N-cadherin expression; because P-cadherin expression in breast cancer is far more frequent than aberrant expression of N-cadherin, its physiologic relevance is more likely to be higher (15) ; (3) except from the presently demonstrated induction of invasion, no regulatory functions have been described for the P-cadherin juxtamembrane domain. This establishes a novel role for this domain and distinguishes P-cadherin–mediated invasion from invasion induced by N-cadherin, which depends on a physical interaction of its extracellular domain with the fibroblast growth factor receptor (51) . Remarkably, although the P-cadherin juxtamembrane domain differs in only few amino acids from the corresponding E-cadherin domain, it exerts an opposite function: whereas the E-cadherin juxtamembrane domain suppresses motility (52) , the P-cadherin juxtamembrane domain is necessary for induction of invasion. To understand why such related domains can have opposite functions, it will be crucial to identify new interaction partners and/or to study if the interaction of known partner molecules differs between cadherins.
The authors would like to acknowledge Profs. Keith R. Johnson, Yukata Shimoyama, and G.P. Nolan for providing reagents, and Astra Zeneca (Portugal) for providing ICI 182,780 to use in this study.
Grant support: Ph.D. grants from the Portuguese Science and Technology Foundation (BD/1450/2000, J. Paredes) and from the Belgian Fund for Scientific Research-Flanders (C. Stove, V. Stove, and V. Van Marck).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: J. Paredes and C. Stove contributed equally to this article and should be considered as first authors. Supplementary data for this article can be found at Cancer Research Online (http://cancerres.aacrjournals.org).
Requests for reprints: Fernando Schmitt, Institute of Pathology and Molecular Immunology of Porto University, Rua Roberto Frias s/n, 4200-465 Porto, Portugal. Phone: 351225570700; Fax: 351225570799; E-mail:
↵6 V. Van Marck, C. Stove, V. Stove, J. Paredes, M. Bracke. P-cadherin promotes cell-cell adhesion and counteracts invasion in human melanoma, manuscript in preparation.
- Received March 4, 2004.
- Revision received August 31, 2004.
- Accepted September 15, 2004.
- ©2004 American Association for Cancer Research.