p130Cas/Cyclooxygenase-2 axis in the control of mesenchymal plasticity of breast cancer cells

p130Cas mAbs have been previously described [6]. mAbs to Vinculin were from Millipore (Billerica, MA, USA). Abs to c-Src, p-Tyr PY99, Cyclin D1, Snail, Slug, Twist and Actin were from Santa Cruz Biotechnologies (Palo Alto, CA, USA). pTyr416 c-Src and pJnk (Thr183/Tyr185) Abs were from Cell Signaling (Beverly, MA, USA) and Abs to Cox-2 from Cayman Chemical (Ann Arbor, MI, USA). Secondary antibodies conjugated with peroxidase were from Sigma-Aldrich (St. Louis, MO, USA). Collagen I was from BD Trasduction Laboratories (Franklin Lakes, NY, USA). Doxycycline was purchased from Sigma-Aldrich.

A17 cells were cultured in DMEM-20% FCS and LM2-4175 in DMEM-10% FCS. Doxycycline at a concentration of 1 microgram/ml was directly added to medium and medium was changed every two to three days. The specific inhibitors of c-Src (SU6656) or JNK (SP600125) were used at a final concentration of 10 micromolar and 40 micromolar respectively for 16 hrs. Live images at 10X, 20X, magnification were collected with a Zeiss microscopy (Oberkochen, Germany).

Viral particles of pLVTHM carrying shRNA sequences (Ctr shRNA, p130Cas shRNA, or Cox-2 shRNA) were produced as described in [6]. For Cox-2 downregulation the sequences used were: GCTGTTCCAATCCATGTCAAA (COX-2 shRNA) and TCTCGCTTGGGCGAGAGTAAGCTC (Ctr shRNA).

For p130Cas and Cox-2 expression, human p130Cas cDNA, mouse p130Cas cDNA fused with GFP or human Cox-2 cDNA, respectively, were cloned into pCCL lentiviral vector, and viral particles production was performed as described above.

For silencing p130Cas in LM2-4175 cells, the human shRNA sequence (5'CCTTGCAGTACCCATCGCCTT3') was inserted into pLKO vector purchased from Open Biosystems (Thermo Fisher Scientific, Waltham, MA, USA). Lentiviruses were produced according to manufacturer's instructions.

Total RNA was isolated from cells using TRIzol™ Reagent (Invitrogen Life Technologies, Carlsbad, CA, USA). 1 μg of DNAse-treated RNA (RQ1 RNase-Free DNase kit, Promega, Madison, WI, USA) was retrotranscribed with High Capacity cDNA Reverse Transcription Kit (Invitrogen Life Technologies, Carlsbad, CA, USA). Quantitative PCR was performed on an Applied Biosystems, 7900HT Fast Real-Time PCR System (standard settings) using the Universal Probe Library system (Roche Italia, Monza, Italy) and Platinum™ Quantitative PCR SuperMix-UDG (Invitrogen Life Technologies, Carlsbad, CA, USA). Results were analyzed with the 2^−ΔΔCt method using the 18S rRNA pre-developed TaqMan assay (Invitrogen Life Technologies, Carlsbad, CA, USA) as an internal control. The median expression across samples was used as calibrator. The following primers and probes were used: Cox-2, forward: GATGCTCTTCCGAGCTGTG reverse: GGATTGGAACAGCAAGGATTT, probe number 45; E-cadherin: forward: ATCCTCGCCCTGCTGATT and reverse: ACCACCGTTCTCCTCCGTA, probe number 18.

To generate Cox-2 promoter luciferase reporter plasmids, two different Cox-2-promoter fragments were generated by PCR, using A17 genomic DNA as template, and the following primers: forward (-3195) 5'-CGCGCTCGAGTTTTATTGTTCTGCCCTCATGTGT-3'; forward (-965) 5'-CGCGCTCGAGCAACACAAACACAGTAGGAAGATA-3'; reverse (+39) 5'-CGCGAAGCTTGACTGACTCCTGAAGCTCTTAGCT-3'. The fragments (-3195 bp/-965 bp to +39 bp), respectively were cloned into pGL3-control vector expressing a firefly luciferase (Promega, Madison, WI, USA) using XhoI and HindIII restriction enzymes. The sequences of all constructs were confirmed by sequencing (BMR Genomics, Padova, Italy).

Protein extracts and western blots were performed as described in [6]. For tumor protein extraction, tissues were removed, frozen in liquid nitrogen, and homogenized in lysis buffer. Densitometric analysis was performed using the GS 250 Molecular Imager (Bio-Rad, Hercules, CA, USA)

Adhesion assays were performed as described in [11] on dishes coated with 10 microgram/ml Collagen I.

Fvb-Neu mice were challenged subcutaneously in the left inguinal region with 10⁵ A17 Ctr shRNA, A17 p130Cas shRNA or A17 Cox-2 shRNA cells. The incidence and growth of tumors were evaluated twice weekly by measuring with calipers for the two perpendicular diameters. Mice water supplemented with doxycycline (0.1 mg/mL) was protected from light and changed every two to three days. The use of animals was in compliance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health and was approved by the Animal Care and Use Committee of Camerino University

Histology and immunohistochemistry preparations were performed as previously described [27].

For immunohistochemistry, these sections were incubated for 30 min with primary antibodies. After washing, they were overlaid with biotinylated goat anti-rat or anti-rabbit immunoglobulin (Vector Laboratories, Burlingame, CA, USA) for 30 min. Unbound antibodies were removed, and the slides were incubated with avidin-biotin complex/alkaline phosphatase (Dako, Milan, Italy).

Publicly available microarray data from the Netherlands Cancer Institute of 295 early-stage breast cancer biopsies (GSE2034) [28] and from the Koo Foundation Sun Yat-Sen Cancer Center (KFSYSCC) of 327 breast cancer tissues (GSE20685) [29] were used. Before analysis, the dataset was gene mean centered by subtracting the mean value for each gene across all samples of the compendium from all data points, so that in all cases expression values of each data point were reported as positive or negative depending on whether it was higher or lower than the mean value of that gene across the samples.

Statistical analysis was performed using a log-rank (Mantel-Cox) test.

To investigate the role of p130Cas in mesenchymal breast cancer cells, we generated cells expressing doxycycline-inducible control or p130Cas shRNA sequences [6], resulting in p130Cas silencing of about 90% (Figure 1A). Remarkably, upon four days of doxycycline treatment, p130Cas silenced cells underwent a switch from an elongated mesenchymal phenotype to a polygonal epithelial-like shape (Figure 1B, panels a, b) that reverted upon re-expression of p130Cas in silenced cells (Figure 1B, panel c), indicating that p130Cas tuning can control mesenchymal breast cancer cell plasticity.

p130Cas silenced cells revealed decreased expression of the transcriptional factors Snail, Slug and Twist, and of the mesenchymal marker Vimentin, whose levels were restored by re-expression of p130Cas (Figure 1C), or by washing out doxycycline from A17 culture medium (Figure S1 in Additional file 1). Snail, Slug and Twist are known to repress E-cadherin expression during EMT [20, 30]. Quantitative real-time PCR experiments and western blot analysis (Figure 1D) showed that E-cadherin was induced both at mRNA and protein levels upon p130Cas silencing. Consistently, when p130Cas was re-expressed in silenced A17 cells, E-cadherin expression was strongly downregulated, returning to control levels. Immunofluorescence staining clearly showed that upon p130Cas silencing E-cadherin expression becomes detectable in A17 cells with a strong plasma membrane staining (Figure 1E, panels b and c) that is totally missing in control (panel a) and in p130Cas reconstituted cells (panel d). Thus p130Cas can modulate expression of mesenchymal/epithelial markers, resulting in a reversible transition from mesenchymal to epithelial features.

p130Cas has been already shown to play a role in the intrinsic plasticity that allows cells to switch from epithelial to mesenchymal phenotype in pancreatic cancer cells [31], while the second member of the Cas protein family NEDD9 controls EMT in breast [32, 33], and melanoma [34] cancer cells. Remarkably, by mass spectrometry-based profiling, p130Cas tyrosine phosphorylation has been described to be elevated in basal breast cancer cells [35]. Genome-wide transcriptional profiling of a large set of human breast cancer cell lines confirms that EMT features are mostly associated with basal-like tumors [36, 37], suggesting a link between p130Cas expression and basal breast tumors.

Cox-2 is frequently associated with aggressive breast cancer [38]. Cox-2 was found significantly overexpressed in A17 cells, where it correlates with their mesenchymal signature [22, 23]. Interestingly, in p130Cas silenced cells the expression of Cox-2 markedly decreased, and was restored by re-expressing p130Cas (Figure 2A, left panel). qRT-PCR showed that in p130Cas silenced cells Cox-2 mRNA was reduced by 80% compared to control cells, and restored to control levels after p130Cas re-expression in silenced cells (Figure 2A, middle panel), suggesting that p130Cas exerts a transcriptional control on Cox-2 expression. Luciferase assays on two DNA fragments corresponding to a short (-965, +39) and a long (-3195, +39) Cox-2 promoter indicated that p130Cas silencing significantly decreased Cox-2 promoter activity (Figure 2A, right panel). Adhesion-dependent Cox-2 induction has been previously described [39‐41]. Consistently, plating control and p130Cas silenced cells on Collagen I-coated dishes for different times, showed that Cox-2 induction both at mRNA and protein levels and was markedly delayed and decreased in p130Cas silenced cells (Figure 2B). Taken together, these results show that p130Cas is a key upstream element in the regulation of Cox-2 expression in breast cancer cells. As Cox-2 has been proposed as a mediator of breast tumor epithelial-stroma interactions, which promote growth and progression of in situ tumors [42], these results suggest that p130Cas can behave as a master regulator of tumor/microenvironment interactions.

Interestingly, the p130Cas-dependent expression of Cox-2 is instrumental for the regulation of breast cancer cells plasticity. Indeed, re-expression of Cox-2 in p130Cas silenced cells reverted cells to a mesenchymal morphology (Figure 2C, panel c) and restored Snail, Slug and Twist expression (Figure 2D). Accordingly, cells expressing doxycycline-inducible Cox-2 shRNAs in which Cox-2 was knocked down by about 90% (Figure 2E, right panels), exhibited a clear switch from an elongated to a polygonal epithelial shape (Figure 2E, panel b). Moreover, these cells showed marked downregulation of Slug and Twist transcriptional factors, while p130Cas expression was not affected (Figure 2E, right panels). These results indicate that p130Cas controls Cox-2 expression and that Cox-2 is involved in p130Cas-dependent maintenance of mesenchymal phenotype, thus establishing a p130Cas/Cox-2 axis that sustains the mesenchymal features of breast cancer cells.

To investigate the role of p130Cas/Cox-2 axis on tumor growth, syngeneic mice were subcutaneously injected with 10⁵ control or p130Cas silenced cells and treated with doxycycline in drinking water. Within three weeks, all the 31 mice injected with control cells gave rise to tumors with a mean diameter of 8 mm. In contrast, 38% (28 out of 73) of mice injected with p130Cas silenced cells did not give rise to detectable tumors and the remaining 45 mice developed small tumors, with a mean diameter of 2 mm (Figure 3A, left panel). Interestingly, p130Cas silencing was sufficient to halt tumor growth in mice that have already developed tumors with a diameter of 3 to 4 mm. Indeed, by adding doxycycline to drinking water two weeks after cell injection, p130Cas silenced tumors regressed, becoming undetectable by palpation within two to three weeks, while control tumors continued to grow (Figure 3A, right panels). Consistently, after doxycycline withdrawal p130Cas silenced tumors resumed growing. These data strengthen the in vivo relevance of p130Cas as a major regulator of the tumorigenic properties of mesenchymal breast cancer cells. We have previously shown that intranipple injection of p130Cas siRNAs in the mammary gland of Balb/c-NeuT mice significantly decreases the number of cancer lesions compared to glands injected with control siRNAs [6], with a significant downregulation of proliferative and survival pathways. Overall these data indicate that tight modulation of p130Cas levels can affect in vivo tumor properties of distinct breast cancer subtypes, implying the compelling need of studying its transcriptional regulation in normal mammary epithelial cells and in tumors in the near future.

Hematoxylin and eosin staining of tumor sections showed that tumors derived from p130Cas silenced cells consisted of cells with an epithelial-like shape, while the control tumors presented elongated, mesenchymal cells (Figure 3B, panels a, b). Moreover, immunohistochemistry analysis indicated that tumors from p130Cas silenced cells were characterized by decreased vascularization and proliferation (CD31 and pCNA staining), and increased apoptosis (de novo expression of Caspase-3) (Figure 3B, panels c-h).

Western blot analysis of p130Cas silenced tumors showed a significant in vivo p130Cas silencing together with Cox-2 downregulation, compromised activation of c-Src and JNK kinases and decreased expression of Cyclin D1 (Figure 3C, left panels). A parallel downregulation of Snail, Slug and Twist expression was also detected (Figure 3C, right panels), indicating that p130Cas silencing compromises tumor growth through inhibition of cell signaling controlling cell cycle progression and the acquirement of epithelial-like features. In parallel, syngeneic mice were subcutaneously injected with 10⁵ Cox-2 silenced or control A17 cells and treated with doxycycline in drinking water. As shown in Figure 3D, while mice injected with control cells gave rise to tumors with a mean diameter of 10 mm within six weeks, mice injected with Cox-2 silenced cells give rise to barely detectable tumors. Taken together these data show that p130Cas/Cox2 axis controls in vivo survival and proliferative pathways of mesenchymal breast cancer cells and silencing of either p130Cas or Cox-2 is sufficient for switching cells to an epithelial state leading to impaired tumor growth.

To assess whether the p130Cas/Cox-2 axis is effective also in the human setting, we chose the human lung metastatic MDA-MB-231 subpopulation LM2-4175 [25, 26] as they recapitulate A17 cell features with high levels of Cox-2 expression and a mesenchymal phenotype. Upon infection with lentiviral particles carrying human p130Cas shRNA, the marked downregulation of p130Cas was associated with a concomitant decrease in Cox-2, Snail, Slug and Twist (Figure 4A, left panels). Accordingly, p130Cas silenced cells reorganized in colonies that lost their elongated protrusions, acquiring a more polygonal shape (Figure 4A, panels b and e), as quantified by a marked decreased in length/width ratio. Re-expression of a mouse full-length p130Cas-GFP fused protein (mFLCas) in LM2-4175 p130Cas silenced cells, re-established Cox-2 and mesenchymal markers expression at the same level of control cells (Figure 4A, left panels), and consistently p130Cas reconstituted cells reacquired elongated protrusions (Figure 4A, panels c and f). Moreover, p130Cas silencing led to a strong reduction of c-Src and JNK activities (Figure 4A, left panels), similar to those observed in in vivo tumor grafts derived from p130Cas silenced A17 cells. Interestingly, cell treatment with specific inhibitors of c-Src (SU6656) or JNK (SP600125) activities for 16 hrs, caused a switch to an epithelial morphology similar to that observed upon p130Cas downregulation (Figure 4B, left panels). Consistent with the fact that Src and JNK controls Cox-2 expression (for review, [43]), both inhibitors caused downregulation of Cox-2, and a reduction in Snail, Slug and Twist expression (Figure 4B, right panels), without grossly affecting p130Cas levels. In addition, cells treated with the c-Src inhibitor SU6656 showed a decrease in JNK activity, while the JNK inhibitor SP600125 did not affect c-Src phosphorylation, suggesting that Src activity is upstream to JNK activation. Moreover, in A17 cells, luciferase assays revealed that the reporter expression driven by Cox-2 promoter was decreased by the use of Src inhibitor and practically abrogated with JNK inhibitor (Figure S2 in Additional file 1). Overall these data show that the p130Cas/Cox-2 axis is effective both in the mouse and in the human setting. c-Src and JNK kinases appear as sequential players in this axis and their pharmacological inhibition was sufficient to downregulate Cox-2 and to induce an epithelial phenotype. These results also suggest the potential clinical application of targeting c-Src through pharmacological inhibitors in breast tumors expressing high levels of p130Cas and Cox-2, the same strategy already proposed in HER2-positive trastuzumab-resistant tumors to overcome trastuzumab resistance [44].

Finally, in order to evaluate whether the p130Cas/Cox-2 axis has clinical relevance in human breast cancer, publicly available microarray data from the Netherlands Cancer Institute of 295 early-stage breast cancer biopsies [28] and from the Koo Foundation Sun Yat-Sen Cancer Center (KFSYSCC) of 327 breast cancer tissues [29] were analyzed. Kaplan-Meier curves showed that p130Cas and Cox-2 double positivity was associated with the lowest time survival (log-rank Van de Vijver P = 0.0025 and Kao P = 0.0071) (Figure 4C, D), and the highest frequency of recurrence (log-rank P = 0.0013) (Figure 4C), indicating that high levels of p130Cas/Cox-2 co-expression relates to the worst prognosis in breast cancer. Previous data have already shown that high levels of p130Cas correlate with intrinsic resistance to tamoxifen treatment in a large subset of estrogen receptor (ER)-positive human breast tumors (reviewed in [3]). Moreover, in human breast cancers overexpression of both HER2 and p130Cas is associated with poor prognosis [10].