Metastatic breast cancer remains the chief cause of cancer-related death among women in the Western world. Although loss of cell–cell adhesion is key to breast cancer progression, little is known about the underlying mechanisms that drive tumor invasion and metastasis. Here, we show that somatic loss of p120-catenin (p120) in a conditional mouse model of noninvasive mammary carcinoma results in formation of stromal-dense tumors that resemble human metaplastic breast cancer and metastasize to lungs and lymph nodes. Loss of p120 in anchorage-dependent breast cancer cell lines strongly promoted anoikis resistance through hypersensitization of growth factor receptor (GFR) signaling. Interestingly, p120 deletion also induced secretion of inflammatory cytokines, a feature that likely underlies the formation of the prometastatic microenvironment in p120-negative mammary carcinomas. Our results establish a preclinical platform to develop tailored intervention regimens that target GFR signals to treat p120-negative metastatic breast cancers. Cancer Res; 73(15); 4937–49. ©2013 AACR.
Adherens junctions are required to maintain epithelial tissue integrity. They are established by homotypic interactions between E-cadherin (CDH1) molecules, which in turn control binding of cytosolic catenins that provide linkage to and regulation of the microtubule and actin cytoskeleton (1). Loss or temporal downregulation of E-cadherin is strongly linked to tumor development and progression of several cancer types (2). In breast cancer, timing of adherens junction inactivation has considerable impact on tumor etiology and cellular biochemistry. Although mutational inactivation of E-cadherin is an initiating and causal event in the development of invasive lobular carcinoma (ILC; refs. 3–5), late epigenetic silencing may underlie the progression of invasive ductal carcinoma (IDC; ref. 6) in a process called epithelial-to-mesenchymal transition (EMT; ref. 7). In conjunction with this is the finding that while translocation of p120-catenin (p120; CTNND1) upon E-cadherin inactivation controls Rock-dependent metastasis of ILC, IDC cells do not show dependency on this pathway (8, 9). Thus, p120 may play context-dependent roles in the development and progression of breast cancer.
Under physiologic conditions, p120 binds to the intracellular juxtamembrane domain of E-cadherin (10). Here, p120 controls E-cadherin stability and turnover in a process mediated by Hakai (CBLL1), (11), presenilin-1 (PSEN1; ref. 12), and/or Numb (NUMB; ref. 13). Others and we have shown that genetic ablation of E-cadherin or p120 in mouse mammary epithelial cells results in the induction of apoptosis, indicating that inactivation of adherens junction function is not tolerated in the mammary gland (4, 14, 15). In contrast, genetic inactivation in other organ systems does not induce cell death, but instead induces impaired tissue homeostasis and hyperplasia (16–19). Also, p120 inactivation in mice seems to result in inflammation, which may be caused by a loss of barrier function and the production of inflammatory chemoattractants (17, 18, 20). Interestingly, recent data showed that p120 can function as a bona fide tumor suppressor in the upper gastrointestinal tract. Here, somatic p120 inactivation induced the development of squamous cell carcinoma, which was accompanied by autocrine production of monocyte/macrophage attractants, thus promoting a proinvasive tumor microenvironment (21).
Several studies have indicated that p120 may be lost or inactivated in approximately 10% of IDC breast cancers cases. Loss was defined as absence of expression in more than 10% of the tumor cells, and correlated to absence of progesterone receptor (PGR) expression and poor prognosis (22–24). Here, we have analyzed p120 expression in a comprehensive set of human invasive breast cancer samples and studied the consequences of inactivation of p120 in mammary tumor development and progression in the context of p53 (Trp53) loss using conditional mouse models. Mammary-specific p120 loss in mice resulted in a switch from noninvasive to metastatic mammary carcinoma that phenotypically resembled human metaplastic breast cancer. Furthermore, inactivation of p120 resulted in anoikis resistance, which was exacerbated by a sensitization to growth factor receptor (GFR) signaling due to inactivation of the adherens junction. Finally, we show that loss of p120 results in secretion of inflammatory cytokines, which may promote formation of a prometastatic tumor microenvironment.
Materials and Methods
Additional experimental procedures are described in detail in the Supplementary Experimental Procedures.
Of note, 298 cases of IDC were collected and histologically examined as described in the Supplementary Experimental Procedures.
Antibodies and cytokine array
The following antibodies were used: mouse anti-p120 [Western blot (WB), 1:2,000; immunofluorescence (IF), 1:500; IHC, 1:500; clone 98/pp120; BD Biosciences], tetramethyl-rhodamin-conjugated p120 (immunofluorescence, 1:150; BD Biosciences) mouse anti-E-cadherin (WB, 1:2,000; clone 36/E-cadherin; BD Biosciences), fluorescein isothiocyanate (FITC)-conjugated E-cadherin (IF, 1:150; BD Biosciences), mouse anti-E-cadherin (IHC, 1:200; clone 4A2C7; Zymed, Invitrogen), rat anti-CK 8 (Troma-1; IHC, 1:125; Developmental Studies Hybridoma Bank products), rabbit anti-CK14 (IHC, 1:10,000; BabCo), guinea pig anti-vimentin (IHC, 1:400; RDI), rabbit anti-SMA (IHC, 1:350; Lab Vision), rat anti-mouse F4/80 (IHC, 1:300; Serotec), mouse anti-β-catenin (WB, 1:2,000; BD Biosciences), rabbit anti-αE-catenin (WB, 1:2,000; Sigma) goat anti-AKT1 (C-20; WB, 1:1,000; Santa Cruz Biotechnology), rabbit anti-p-AKT1s473 (WB, 1:1,000; Cell Signaling Technology), rabbit anti-ERK2 (C-14; Santa Cruz Biotechnology), rabbit anti-p-MAPK (p44/42; WB, 1:2,000; Cell Signaling Technology), mouse anti-RhoA (WB, 1:250; Santa Cruz Biotechnology), mouse anti-Rac1 (WB, 1:1,000; Upstate), rabbit anti-Cdc42 (WB, 1:250; Santa Cruz Biotechnology), rabbit anti-EGFR (WB, 1:500; Cell Signaling Technology), rabbit anti-p-EGFRT1068 (WB, 1:1,000; Cell Signaling Technology), and sheep anti-digoxigenin (Roche). Secondary antibodies: horseradish peroxidase (HRP)–conjugated rabbit anti-goat, goat anti-mouse, (1:2,000; Dako), HRP-conjugated goat anti-rabbit (1:2,000; Cell Signaling Technology), rabbit anti-sheep antibodies (Dako), Alexa Fluor 561–conjugated anti-mouse antibodies (Molecular Probes), goat anti-rat Alexa Fluor 488 (Invitrogen), goat anti-rabbit Alexa Fluor 555 (Invitrogen), biotin-conjugated anti-guinea pig (Jackson ImmunoResearch). Biotin label-based mouse antibody array was used according to the manufacturer's recommendations (RayBiotech AAM-BLM-1-4).
Mouse crossings genotyping and generation of cell lines
p120 conditional mice containing loxP sites in intron 2 and 8 of Ctnnd1, (Black-swiss;129SvEvTac; ref. 17) were a kind gift from Al Reynolds (Vanderbilt University Medical Center, Nashville, TN) and crossed onto the Wcre;Trp53F/F mouse model (FVB/N;Ola129/sv; refs. 4, 5). Cohorts of Wcre;Ctnnd1F/+;Trp53F/F and Wcre;Ctnnd1F/F;Trp53F/F mice were bred from the resulting offspring. Wcre;Ctnnd1F/+;Trp53F/F mice were subsequently used to generate Wcre;Trp53F/F mice to control for the introduction of Black-swiss;129SvEvTac genetic material. Mice were bred and maintained on a mixed background of (FVB/N;Ola129/sv). Genotyping was done by PCR as described previously (4, 17). Mice were monitored for the development of mammary tumors by palpation and euthanized by CO2 inhalation when mammary tumor size reached a diameter of 10 mm. Full autopsies were conducted for the analysis of tumor histology and the detection of metastases. Age at the time of euthanasia was used to generate the tumor-free survival curves. For the generation of tumor cell lines, tumors from Wcre;Ctnnd1F/F;Trp53F/F animals were extracted, minced by hand using scalpel blades, and plated onto regular culture dishes in Dulbecco's Modified Eagle Medium (DMEM)-F12 medium as described previously (5).
For stable knockdown of p120, previously described sequences were cloned into a doxycycline-inducible lentiviral expression system (8). For p120 reconstitution experiments, a lentiviral p120-1A cDNA expression construct was generated as described in the Supplementary Experimental Procedures.
Virus production and Rho GTP pulldown
Lentivirus production and transductions were done as described previously (5). In short, 106 Cos-7 cells were seeded onto 10-cm Petri dishes and transiently transfected after 24 hours with third-generation packaging constructs (25) and the indicated viral construct using X-tremeGENE 9 reagent (Roche). Pull-down assays for GTP-loaded RhoGTP family members were conducted as described previously (8, 26).
Mouse Trp53Δ/Δ-7 (WP6) and Trp53Δ/Δ-3 (KP6) cell lines were generated from primary tumors that developed in Wcre;Trp53F/F and K14cre;Trp53F/F female mice, respectively. Cells were cultured as described previously (5). Human breast cancer cell lines T47D and MCF7 were cultured in DMEM-F12 (Invitrogen) containing 6% fetal calf serum, 100 IU/mL penicillin, and 100 μg/mL streptomycin. T47D and MCF7 cell lines were obtained from American Type Culture Collection. Authenticity was tested on May 16, 2012 by means of short-tandem repeat (STR) profiling (LGC Standards), after which large quantities of cells were frozen separately.
Western blot analysis, immunohistochemistry, and fluorescence
Western blotting was conducted as described previously (27). For immunohistochemistry and fluorescence, tissues and cells were isolated, fixed, and stained as described in the Supplementary Experimental Procedures. In situ hybridization–IHC double staining experiments were carried out using labeled PCR products to identify EGF mRNA as described in the Supplementary Experimental Procedures. Samples were analyzed using a DeltaVision RT system (Applied Precision), equipped with a CoolSnap HQ camera and SoftWorx software. Maximum projections were taken from a stack of deconvolved images.
Anoikis resistance and FACS analysis
Anoikis resistance was analyzed by seeding cells at a density of 20,000 cells per well (in 500 μL) in a 24-well ultra-low cluster polystyrene culture dish (Corning). After 4 days, cells were harvested and resuspended in 75 μL of Annexin V buffer supplemented with Annexin V (IQ Products) and propidium iodide (Sigma-Aldrich). The percentage of anoikis-resistant cells was defined as the Annexin V and propidium iodide–negative population analyzed on a BD FACSCalibur. To determine cellular EGF-binding ability, 400 ng of Alexa Fluor 647–conjugated EGF (Invitrogen) was incubated with 1 × 105 trypsinized ice-cold cells in 100 μL PBS. Cells were washed with PBS to remove unbound EGF-647 and subjected to fluorescence-activated cell sorting (FACS) analysis.
Growth factor stimulation assays
Cells were seeded at 400,000 cells per 6-well in growth factor–free medium for 4 hours, subsequently washed and serum-starved overnight. Next, cells were stimulated with EGF (5 ng/mL; Sigma) or HGF (25 ng/mL; R&D Systems) for 10 minutes. Cells were placed on ice and washed twice with ice-cold PBS containing Ca2+ and Mg2+, and directly lysed in lysis buffer.
Statistical analyses were conducted using GraphPad Prism 5 (GraphPad Software). Analyses on human samples were conducted as described previously (28). For analysis of growth pattern and metastasis formation, Fisher exact test was used. For metastasis-free survival analysis, the log-rank test was used. For anoikis assays, statistical significance was calculated using the Student t test (two-tailed), showing measurements of at least three independent experiments. Error bars in all experiments represent SD or SEM as indicated, of at least triplicate measurements. We considered P values less than 0.05 as statistically significant.
All animal experiments were approved by the University Animal Experimental Committee, University Medical Center Utrecht (Utrecht, the Netherlands). Use of anonymous or coded leftover material for scientific purposes is part of the standard treatment contract with patients in our hospitals.
Loss of p120 expression in invasive breast cancer
To extend previous findings on loss of p120 expression in breast cancer, we conducted immunohistochemistry (IHC) on a panel of 298 invasive ductal breast cancers and determined their clinicopathologic variables (Supplementary Table S1). Membranous p120 was scored (absent/low vs. medium/high) in three independent tissue cores per tumor. Because we hypothesized that loss of p120 may be linked to increased tumor progression, we defined expression as absent/low if more than 10% of the tumor cells were negative for p120 (Supplementary Fig. S1A), as has been used for E-cadherin (29). Using these parameters, we observed that 34% of the IDC samples showed absent/low p120 staining. Upon correlation of p120 expression levels to clinicopathologic variables, a significant association was obtained between p120 loss, high tumor grade (P = 0.007), mitotic index (MAI; P = 0.002), and overall absence of hormone receptor expression (P = 0.039; Supplementary Table S2). Although p120 expression levels did not associate with tumor size, lymph node status, or the other clinicopathologic variables tested (Supplementary Table S2), these data suggest that loss of p120 expression coincides with breast cancer aggressiveness.
Somatic inactivation of p120 results in the formation of metastatic mammary carcinoma
Functional inactivation of the adherens junction through somatic inactivation of E-cadherin is a causal event in the development of ILC (5). In E-cadherin–mutant breast cancer, p120 translocates to the cytosol where it exerts a key oncogenic role by regulating anchorage-independent tumor growth and metastasis (8). To study the effect of p120 loss on tumor development and progression of breast cancer, we introduced a conditional p120 allele (Ctnnd1F; ref. 17) onto the Wcre;Trp53F/F (4) noninvasive mammary carcinoma model to produce Wcre;Ctnnd1F/+;Trp53F/F and Wcre;Ctnnd1F/F;Trp53F/F mice. To correct for the differences in genetic background between the Ctnnd1F (17) and the Wcre;Trp53F/F (FVB/N; 129P2/OlaHsd) mice, a control cohort was bred using Wcre;Ctnnd1F/+;Trp53F/F litter mates to produce Wcre;Trp53F/F. Next, females were monitored for spontaneous tumor development. In contrast to conditional inactivation of E-cadherin (4, 5), we observed that the median tumor-free latency (T50) did not significantly change in either Wcre;Ctnnd1F/+;Trp53F/F (T50 = 223 days) or Wcre;Ctnnd1F/F;Trp53F/F female mice (T50 = 214 days) when compared with Wcre;Trp53F/F females (T50 = 213 days; P = 0.5006 and 0.5859, respectively; Fig. 1A; Table 1). Mammary tumors from Wcre;Trp53F/F females were morphologically typed as adenocarcinomas and carcinosarcomas, with expansive growth patterns, dense cellular sheets, and irregular bundles of polygonal to plump spindle-shaped cells (Table 1). Tumors that developed in Wcre;Ctnnd1F/+;Trp53F/F and Wcre;Ctnnd1F/F;Trp53F/F females showed a shift from expansive to invasive growth as compared with Wcre;Trp53F/F animals (P = 0.026 and 0.006, respectively; Table 1). In contrast to somatic E-cadherin inactivation and the development of ILC, both heterozygous and homozygous loss of p120 resulted in metaplastic tumor cells displaying a spindle cell morphology, often expressing vimentin with focal expression of cytokeratin (CK)8 or CK14, reminiscent of an EMT (Fig. 1B, top; Supplementary Table S3; Supplementary Fig. S1B). Tumors developed with high incidence multifocally in different mammary glands and were characterized by a more abundant and dense stromal microenvironment. Mammary tumors from Wcre;Ctnnd1F/+;Trp53F/F females did not show LOH of the Ctnnd1 locus, as tumors expressed membrane-localized p120 and E-cadherin (Fig. 1B, middle and Supplementary Table S3). Mammary tumors from Wcre;Ctnnd1F/F;Trp53F/F females showed large cells with pleomorphic nuclei, coarsely clumped chromatin, and sparse cytoplasm. As expected, all tumors from Wcre;Ctnnd1F/F;Trp53F/F female mice lacked expression of membranous p120 and E-cadherin (Fig. 1B, right and Supplementary Table S3).
|Genotype .||Number of mice .||Median latency, d .||Metastasis .||Local invasion .||AC .||SC/CS .||mILC .|
|Wcre;Trp53F/F||7||213||0 (0%)||0 (0%)||3 (43%)||7 (100%)||0 (0%)|
|Wcre;Ctnnd1F/+;Trp53F/F||20||223||1 (5%)||10 (50%)||1 (5%)||19 (95%)||1 (5%)|
|Wcre;Ctnnd1F/F;Trp53F/F||21||214||9 (43%)||14 (67%)||3 (14%)||18 (86%)||1 (5%)|
|Genotype .||Number of mice .||Median latency, d .||Metastasis .||Local invasion .||AC .||SC/CS .||mILC .|
|Wcre;Trp53F/F||7||213||0 (0%)||0 (0%)||3 (43%)||7 (100%)||0 (0%)|
|Wcre;Ctnnd1F/+;Trp53F/F||20||223||1 (5%)||10 (50%)||1 (5%)||19 (95%)||1 (5%)|
|Wcre;Ctnnd1F/F;Trp53F/F||21||214||9 (43%)||14 (67%)||3 (14%)||18 (86%)||1 (5%)|
NOTE: If tumors were composed of two separate histologic types, both were counted separately.
Abbreviations: AC, adenocarcinoma, glandular-type mammary carcinoma; SC/CS, solid carcinoma/carcinosarcoma, tumor consisting of epithelial and mesenchymal cell types.
When compared with human breast cancer, tumors that developed in Wcre;Trp53F/F mice corresponded to well-differentiated “luminal type” IDC, with low proliferation and expansive growth patterns. They predominantly express luminal CK8, showing little expression of basal CK14 (Supplementary Fig. S1B and Supplementary Table S3). Also, these tumors show a relatively good prognosis with infrequent formation of distant metastases. The highly infiltrative tumors arising in Wcre;Ctnnd1F/F;Trp53F/F females corresponded to poorly differentiated “basal type” tumors. As such, they presented phenotypic features resembling human metaplastic IDC that are characterized by a high proliferation rate, strong nuclear atypia, and expression of CK14 (KRT14) and vimentin (VIM). In contrast to ILC, mouse and human metaplastic tumors displayed decreased p120 levels and a punctate/mislocalized E-cadherin expression pattern (Supplementary Fig. S1C and Supplementary Table S3). Human metaplastic carcinoma corresponds with poor prognosis and rapid onset of distant metastases (30).
Interestingly, and in contrast to dual somatic inactivation of E-cadherin and p53 (4, 5), we observed abundant formation of precursor carcinoma in situ (CIS) lesions in Wcre;Ctnnd1F/F;Trp53F/F females (Fig. 2A). CIS lesions showed loss of p120, but occasionally retained low levels of E-cadherin expression (Fig. 2B and C), indicating that upon p120 inactivation residual E-cadherin remains expressed in a temporal manner, which may underlie the initial noninvasive nature of the CIS-type structures.
Next, we examined whether the acquisition of invasive behavior upon p120 ablation resulted in an increased metastatic rate. We observed a significant increase in tumor cell dissemination in Wcre;Ctnnd1F/F;Trp53F/F versus Wcre;Ctnnd1F/+;Trp53F/F and Wcre;Trp53F/F female mice (both P < 0.05; Table 1), a phenotype not observed in previously published models where inactivation was targeted to the gastrointestinal tract or skin (17, 18, 20, 21). Metastases phenotypically resembled the primary tumor and localized to regional or distant lymph nodes and lungs (Fig. 1C). Because tumor-free latency was identical in Wcre;Ctnnd1F/+;Trp53F/F versus Wcre;Ctnnd1F/F;Trp53F/F mice, but only Wcre;Ctnnd1F/F;Trp53F/F showed metastatic dissemination, we generated primary cultures from both tumor models and assayed anoikis resistance, a hallmark of metastatic cells (5, 31). In agreement with the in vivo metastatic behavior, we observed that tumor cells derived from Wcre;Ctnnd1F/F;Trp53F/F female mice (Ctnnd1Δ/Δ;Trp53Δ/Δ) were anoikis resistant, whereas neither cells derived from Wcre;Ctnnd1F/+;Trp53F/F nor Wcre;Trp53F/F mammary tumors (Ctnnd1Δ/+;Trp53Δ/Δ and Trp53Δ/Δ, respectively) survived under these conditions (Fig. 2D). These results suggested that homozygous inactivation of p120 is necessary for the acquisition of anchorage independence and subsequent dissemination of mammary tumor cells. In conclusion, we show that conditional loss of p120 induces a transition to highly invasive and metastatic mammary carcinoma.
Loss of p120 results in loss of the adherens junction, transition to a mesenchymal phenotype, and anoikis resistance
Because we observed anoikis resistance in primary tumor cells derived from Wcre;Ctnnd1F/F;Trp53F/F female mice (Fig. 2D), we examined if p120 inactivation was the causal event in this anchorage-independent survival phenotype. To this end, we made use of Trp53Δ/Δ tumor cell lines previously generated from adenocarcinomas that developed in either K14Cre;Trp53F/F (32) or Wcre;Trp53F/F female mice (4). Two independent Trp53Δ/Δ cell lines were transduced using a doxycycline-inducible lentiviral construct targeting p120 (p120-iKD). Doxycycline administration resulted in a strong reduction of p120 protein expression (Fig. 3A), which was accompanied by a decrease in both E-cadherin and β-catenin levels and a slight decrease in α-catenin protein expression (Supplementary Fig. S2A and S2B). Upon p120-iKD, cells lost their typical epithelial appearance and transitioned toward a mesenchymal and motile phenotype (Supplementary Fig. S2F). To confirm the specificity of the RNA interference (RNAi) sequence used, we reconstituted cells with a nontargetable p120 cDNA to near endogenous levels, which completely reverted the doxycycline-induced EMT phenotype (Fig. 3A and Supplementary Fig. S2E and S2F).
We next cultured two independent Trp53Δ/Δ;p120-iKD cell lines in suspension and assayed anoikis resistance in the presence and absence of doxycycline. Although the majority of untreated Trp53Δ/Δ;p120-iKD cells underwent anoikis, administration of doxycycline-induced anoikis resistance (Fig. 3B), which could be fully reverted upon expression of a nontargetable p120 construct (Fig. 3B). To substantiate these findings, we used two E-cadherin–expressing and anchorage-dependent human breast cancer cell lines (T47D and MCF7) in which we conducted knockdown of p120 and assayed anchorage independence. MCF7 was previously reported to acquire anchorage independence in soft agar upon constitutive p120 KD (9). Indeed, as for the mouse mammary carcinoma cells, p120-iKD induced downregulation of the adherens junction members and functionally induced anoikis resistance (Supplementary Fig. S2C and S2D and Fig. 3C and D). In conclusion, our data show that p120 knockdown results in loss of adherens junction function, and suggests that this underlies acquisition of anoikis resistance. Because these features are well-known hallmarks of malignancy (33), our findings suggest that loss of p120 may lead to metastasis through acquisition of anchorage independence.
Loss of p120 potentiates GFR signaling
Upon knockdown of p120 in MCF7, cells acquire Rac-dependent anchorage-independent survival, which is probably due to relieve of E-cadherin–mediated inhibition of Ras (KRAS; ref. 9). Although we could confirm that active Rac1 (RAC1) levels increased in MCF7 upon p120-iKD, we did not detect activation of Rac1 in either Trp53Δ/Δ;p120-iKD cell lines or human T47D;p120-iKD cells after doxycycline administration (Supplementary Fig. S3). Also, no changes were observed in the levels or activity of the other Rho family members RhoA (RHOA) and Cdc42 (CDC42; Supplementary Fig. S3C), indicating that loss of p120 in these cell systems does not result in aberrant Rho GTPase activation.
Formation or functional disruption of adherens junctions can affect EGF receptor (EGFR) activity (34). Whether adherens junction disruption results in activation or inhibition of EGFR seems to be largely cell type–dependent (35, 36). Because p120 knockdown resulted in a loss of adherens junction formation through downregulation of E-cadherin, we wondered whether GFR signaling was affected. To this end, we stimulated doxycycline-treated Trp53Δ/Δ;p120-iKD cells with EGF and assayed EGFR phosphorylation. Although EGF stimulation in control-iKD cells induced a modest EGFR tyrosine phosphorylation, knockdown of p120 resulted in a 1.5- to 2.0-fold increased phosphorylation (Fig. 4A, compare lanes 3 and 4; quantification of three independent experiments in Fig. 4B; nonstimulated in Supplementary Fig. S5A). In line with these findings, we observed that the downstream PI3K/AKT and mitogen-activated protein kinases (MAPK) pathways displayed a 3- to 5-fold increased EGF-induced phosphorylation upon knockdown of p120 (Fig. 4A and B). To substantiate these findings, we made use of the Ctnnd1Δ/Δ;Trp53Δ/Δ cell lines. Similarly to the GFR sensitization after inducible p120 knockdown, Ctnnd1Δ/Δ;Trp53Δ/Δ cells showed a robust increase in EGF-induced phosphorylation of EGFR and MAPK and a modest induction of phosphorylated AKT as compared with Trp53Δ/Δ cells (Supplementary Fig. S6A). Furthermore, EGF stimulation of MCF7;p120-iKD cells also showed a sensitization of EGFR, AKT, and MAPK phosphorylation upon knockdown of p120 (Fig. 4C and D). The increased EGFR signaling following p120 knockdown was not due to increased EGFR levels or EGF binding at the plasma membrane (Fig. 4A and Supplementary Fig. S5C). However, we cannot exclude that p120 knockdown induced autocrine activation of GFR signaling in the presence of anchorage, as serum starvation resulted in low levels of activated EGFR, AKT, and MAPK signaling (Supplementary Fig. S5A and S5B). To determine whether the sensitized GFR signaling was EGFR-specific, we also stimulated the cells with hepatocyte growth factor (HGF) to activate the receptor tyrosine kinase MET. In line with the effects on EGF-dependent signaling, we observed a marked increase of AKT and MAPK phosphorylation upon p120 knockdown and subsequent HGF treatment of mouse and human cells (Supplementary Fig. S4A–S4C), indicating that p120-controlled GFR signaling is a general mechanism. To validate whether the increased GFR sensitivity also occurred in vivo, we investigated the expression of p-AKT and p-MAPK in tumors from Wcre;Trp53F/F;Ctnnd1F/F and Wcre;Trp53F/F female mice. Indeed, loss of p120 induced activation of AKT, and to a lesser extent MAPK signals (Supplementary Fig. S6C). In conclusion, our data imply that loss of p120 leads to an increased sensitization of GFR signaling in breast cancer cells.
GFR sensitization stimulates anchorage-independent viability
Because knockdown of p120 resulted in increased growth factor sensitization of pathways implicated in growth and survival, we used EGF stimulation to examine whether this mechanism could increase anchorage-independent growth and survival of breast cancer cells. We therefore plated Trp53Δ/Δ;p120-iKD and Ctnnd1Δ/Δ;Trp53Δ/Δ cells under anchorage-independent conditions and assayed anoikis resistance. In control cells, EGF stimulation did not mediate survival. However, addition of EGF led to an additional increase in anoikis resistance upon knockdown of p120 (Fig. 4E) and in Ctnnd1Δ/Δ;Trp53Δ/Δ cell lines (Supplementary Fig. S6B). Although less prominent, EGF stimulation also induced a significant increase in anoikis resistance of MCF7;p120-iKD cells (Fig. 4F). Because adherens junction–dependent relieve of GFR inhibition is not specific for a given receptor and MCF7 responds more prominent to HGF than EGF, we also stimulated MCF7;p120-iKD cells with HGF. In line with the EGF-dependent findings in our mouse cell lines, we observed that activation of the MET receptor induced a prominent p120-dependent increase in anoikis resistance of MCF7 (Supplementary Fig. S4E).
In conclusion, we find that loss of p120 enhances anoikis resistance through increased sensitization of GFR signaling pathways, a mechanism that may stimulate metastasis in p120-negative breast cancer.
Loss of p120 leads to a proinvasive microenvironment
On the basis of the prominent presence of stroma in the metaplastic p120 null mammary carcinomas and our finding that loss of p120 induces sensitization of GFR signaling, we examined a potential source of growth factors in metastatic tumors from Wcre;Ctnnd1F/F;Trp53F/F female mice. IHC indeed revealed an abundant presence of vimentin-expressing cells surrounding p120-negative tumor cells (Fig. 5A, middle). We next analyzed p120-negative carcinomas for influx of macrophages, a renowned source of EGF (37). p120-Negative tumors from mouse and human samples contained a large macrophage-dense microenvironment (Fig. 5A and B). Moreover, a combination of IHC and RNA in situ hybridization revealed that the tumor-associated macrophages expressed EGF mRNA (Fig. 5C), indicating that a paracrine source of EGF is present in the tumor microenvironment.
In the gastrointestinal tract and skin, conditional p120 loss resulted in cytokine secretion and subsequent attraction of immune cells (18, 20, 21). To determine whether p120 loss would also control production of inflammatory cytokines in breast cancer, we assayed culture supernatant from Trp53Δ/Δ;p120-iKD, Ctnnd1Δ/Δ;Trp53Δ/Δ, and control Trp53Δ/Δ cells using an antibody-based cytokine array. Using these tools, we observed that loss of p120 indeed increased secretion of numerous cytokines. Several of these factors are responsible for macrophage and lymphocyte attraction/maturation and stimulation of cancer-associated fibroblasts such as IL-1α (IL1A; ref. 38), IL-11 (IL11; ref. 39), and IL-12 (IL12A; ref. 40; Supplementary Table S4). These results suggest that p120-negative tumor cells are a paracrine source for the influx of inflammatory cells and subsequent formation of a prometastatic microenvironment.
In conclusion, we show that loss of p120 results in the excretion of cytokines that may control influx of macrophages and other stromal cells that can promote progression of p120-negative tumor cells through the induction of sensitized GFRs signaling pathways.
p120 Loss leads to anoikis resistance and metastasis
In breast cancer, disruption of adherens junction complex formation through early mutational inactivation of E-cadherin leads to the development of ILC (3–5). Upon inactivation of E-cadherin, p120 translocates to the cytosol, where it functions as an oncogene controlling Rock-mediated anoikis resistance of ILC cells (8). In contrast to ILC, most IDC cases retain expression of a functional adherens junction complex that may be inactivated through epigenetic mechanisms at later stages of breast cancer progression (6). Interestingly, and in contrast to the salivary gland (17), gastrointestinal tract (20, 21), dental enamel (16), ocular tissue (19), and skin (18), somatic inactivation of p120 in the mammary gland is not tolerated (15). These tissue-specific differences are further exemplified by the fact that conditional p120 ablation in the upper gastrointestinal tract resulted in the formation of invasive squamous esophageal carcinomas (21).
To determine a possible role for p120 during breast cancer progression, we crossed the conditional p120 allele (41) onto the Wcre;Trp53F/F mouse model of noninvasive breast cancer. Using this compound mouse model, we observed that—whereas onset and incidence of primary tumor development were unaffected—homozygous p120 loss resulted in metastatic disease. Interestingly, metastatic spread was not reported in other conditional p120 tumor models (17, 18, 20). Metastatic spread of tumors from Wcre;Trp53F/F and Wcre;Ctnnd1F/+;Trp53F/F female mice was a rare event. Because E-cadherin remained expressed and derivate cell lines functionally displayed anchorage dependence, it indicated that E-cadherin expression is the rate-limiting factor preventing initial tumor cell invasion and metastasis upon p120 loss. Supporting this assumption is our observation that E-cadherin is temporally retained on the cell surface upon p120 inactivation in preinvasive CIS-type lesions from Wcre;Ctnnd1F/F;Trp53F/F mice. Also, we have shown previously that low E-cadherin expression levels are sufficient to induce anchorage dependency in anoikis-resistant mouse ILC (mILC) cells (4, 5).
Interestingly, and in contrast to mutational E-cadherin loss, we observed that somatic p120 inactivation did not result in the formation of ILC, but instead resulted in metaplastic carcinoma. Differences may be explained by the fact that p120 plays a key oncogenic role in ILC progression through myosin phosphatase Rho interacting protein-dependent regulation of the Rho, Rock, and the cytoskeleton (8, 42). Moreover, in contrast to E-cadherin inactivation, we did not observe metastases in typical ILC dissemination sites such as bone marrow and gastrointestinal tract in the current p120 conditional knockout model. Finally, dual inactivation of p120 and p53 did not accelerate tumor development as seen for combined E-cadherin and p53 loss (4, 5). These phenotypic and functional differences between p120 and E-cadherin suggest that although the overall consequence is adherens junction inactivation and metastasis, p120 loss leads to a biochemically and functionally different breast cancer phenotype.
p120 Loss sensitizes cells to GFR signaling
How does p120 loss regulate anoikis resistance? Our data indicate that p120 loss results in a decrease of E-cadherin–dependent proapoptotic signals in the absence of anchorage, resulting in anoikis. In this scenario, heterozygous p120 loss will not result in metastasis, because residual membranous E-cadherin expression will lead to anoikis in cells that escape from the primary tumor. An additional scenario may be provided by GFR signaling. Membrane receptors can directly influence the activity of unrelated neighboring receptors (43). Depending on cellular context, E-cadherin expression may influence EGFR activation (35, 36). Indeed, studies in gastric cancer showed that germline and somatic (in-frame deleterious) mutations in the E-cadherin extracellular domain lead to increased activation of EGFR signaling (44). We show here that inactivation of the adherens junction through loss of p120 leads to increased sensitization of GFR signaling, a phenomenon that is likely dependent on E-cadherin turnover. Although it is still controversial whether this mechanism is ligand dependent, our data indicate that the GFR hypersensitivity upon p120 loss is not caused by upregulation of EGFR expression levels or increased EGF binding at the cell surface (Fig. 4A and Supplementary Fig. S5C). Given our observation that E-cadherin–dependent inhibition of GFR signaling is not specific for a given receptor, it seems unlikely that specific posttranslational modifications or a specific phosphatase/kinase regulates this process.
A possible scenario may therefore be that E-cadherin elicits a general inhibitory effect on the recruitment of proximal activating molecules and docking adaptors such as Gab1 (GAB1), Grb2 (GRB2), or Sos (SOS1) to sites of tyrosine kinase phosphorylation (45, 46), resulting in a dampening of pathway activation downstream of the GFR. Alternatively, the extracellular domain of E-cadherin could play a role in preventing activation of EGFR, because shedding of E-cadherin has been implicated in EGF-dependent activation of the PI3K/AKT and MAPK pathways and subsequent cell survival (47). Thus, although further research is needed to delineate the exact mechanism that controls the sensitization of GFR signaling upon p120 loss, we propose that increased sensitization of GFR signaling may be caused by a relieve of direct or indirect mechanical restriction, or inhibition of ligand binding as a result of steric hindrance between the adherens junction and GFRs.
Loss of p120 results in a prometastatic microenvironment
Although the precise nature of the influx of inflammatory cells may depend on the model system studied, several groups reported that p120 ablation may lead to inflammation (18, 20, 21). In line with this, we found that loss of p120 results in the development of stromal dense and macrophage-rich mammary tumors. Furthermore, we show that loss of p120 leads to secretion of several cytokines, which may control the recruitment of inflammatory cells (e.g., macrophages) that have been implicated in inflammation-associated cancer initiation and promotion (48) and are correlated with poor prognosis in breast cancer (49, 50). Moreover, it has been well established that mammary tumor cells may instigate a paracrine loop that involves the production of chemoattractants and subsequent recruitment of EGF-producing macrophages, resulting in activation of prometastatic pathways in tumor cells (51, 52).
In closing, we propose that p120 inactivation and subsequent E-cadherin loss is a late event in IDC tumor progression. To our knowledge, there have been no reports on mutational p120 inactivation in breast cancer. Although such mutations could have easily been missed using past sequencing techniques and the fact that p120 loss is often observed in only a small percentage of the tumor, CTNND1 mutations are most probably a rare event. Conversely, this could also indicate that p120 inactivation in IDC may be caused by epigenetic or other indirect mechanisms, as has been shown for E-cadherin (53–55). This, and because of the fact that homozygous p120 inactivation in mice does not induce the formation of mILC, leads us to hypothesize that inactivation of p120 probably occurs during late progression of invasive breast cancer. Although the exact mechanism that regulates p120 inactivation remains to be resolved, our data show that E-cadherin–dependent inhibition of GFR signaling is relieved upon inactivation of p120. As a consequence, anoikis-resistant cells are sensitized to prometastatic growth factors. Because others and we show that p120 loss also facilitates concomitant formation of a prometastatic microenvironment, inactivation of p120 induces multiple hallmarks of metastatic cancer. A simplified model of our findings is presented in Fig. 6.
We think that our findings may have substantial clinical ramifications. In our model, GFR signaling can be enhanced independent of GFR expression levels. Thus, our data imply that the mere presence of certain GFRs is of clinical importance, provided that breast cancer cells are p120 negative. Accordingly and depending on the GFRs expressed, patients suffering from p120-negative IDC may be eligible for treatment with GFR inhibitors targeting the expressed receptors. We have thus uncovered a tumor-promoting role for p120 in breast cancer progression that provides an alternate rationale for therapeutic intervention of p120-negative metastatic breast cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: R.C.J. Schackmann, J. Jonkers, P.W.B. Derksen
Development of methodology: R.C.J. Schackmann, T. Peeters, P.J. van Diest, P.W.B. Derksen
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.C.J. Schackmann, S. Klarenbeek, E.J. Vlug, S. Stelloo, M. Tenhagen, J.F. Vermeulen, P.J. van Diest, J. Jonkers
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.C.J. Schackmann, S. Klarenbeek, S. Stelloo, J.F. Vermeulen, P.J. van Diest, J. Jonkers, P.W.B. Derksen
Writing, review, and/or revision of the manuscript: R.C.J. Schackmann, S. Klarenbeek, J.F. Vermeulen, E. van der Wall, P.J. van Diest, J. Jonkers, P.W.B. Derksen
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Klarenbeek, M. van Amersfoort, M. Tenhagen, T.M. Braumuller, P. van der Groep, P.W.B. Derksen
Study supervision: J. Jonkers, P.W.B. Derksen
The authors thank Eva Schut-Kregel for excellent technical support, Juliet Daniel for providing the p120-1A cDNA and Al Reynolds for the p120 conditional mice. Participants of the MMM meeting and members of the Bos, Burgering, and Pathology labs are acknowledged for help and fruitful discussions. The authors also thank the University Medical Center (UMC) Biobank.
This work was supported by the UMC Cancer Center and grants from the Netherlands Organization for Scientific Research (NWO-VIDI 917.96.318) and the Dutch Cancer Society (UU-KWF 2011 5230).
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.