Activin-A signaling promotes epithelial–mesenchymal transition, invasion, and metastatic growth of breast cancer

Background: Activins belong to the transforming growth factor-β (TGF-β) superfamily of cytokines. Although the role of TGF-β in cancer progression has been highly advocated, the role of activin signaling in cancer is not well known. However, overexpression of activin-A has been observed in several cancers. Aims: The gene expression profile indicated higher expression of Activin-A in breast tumors. Hence the aim of this study was to evaluate the status and role of Activin signaling pathway in these tumors. Methods: Microarray analysis was performed to reveal gene expression changes in breast tumors. The results were validated by quantitative PCR and immunohistochemical analysis in two independent sets of normal and tumor samples. Further, correlation of activin expression with survival and distant metastasis was performed to evaluate its possible role in tumor progression. We used recombinant activin-A, inhibitors, overexpression, and knockdown strategies both in vitro and in vivo, to understand the mechanism underlying the protumorigenic role of this signaling pathway. Results: We report that activin-A signaling is hyperactivated in breast cancers as indicated by higher activin-A, phosphoSMAD2, and phosphoSMAD3 levels in advanced breast cancers. Bone morphogenetic proteins and molecules involved in this signaling pathway were downregulated, suggesting its suppression in breast cancers. Activin-A expression correlates inversely with survival and metastasis in advanced breast cancers. Further, activin-A promotes anchorage-independent growth, epithelial–mesenchymal transition, invasion, angiogenesis, and stemness of breast cancer cells. We show that activin-A-induced phenotype is mediated by SMAD signaling pathway. In addition, activin-A expression affects the tumor-forming ability and metastatic colonization of cancer cells in nude mice. Conclusions: These results suggest that activin-A has a critical role in breast cancer progression and, hence, targeting this pathway can be a valuable strategy in treating breast cancer patients.


Tissue samples and immunohistochemistry
Normal and tumor tissue samples were procured after obtaining informed consent of patients who underwent surgery at the Kidwai Memorial Institute of Oncology, Bangalore. The study has been approved by the Institute Ethical Committee of Kidwai Memorial Institute of Oncology. Details about tissue samples and expression analysis are listed with the Gene Expression Omnibus accession number GSE40206. All the patient details including histopathology data for expression of estrogen receptor, progesterone receptor, Her2/neu, tumor stage/grade, size, and lymph node status were documented for each case. The normal tissue samples were taken from non-tumor-bearing individuals, whereas all the tumor tissues belonged to grade III category. Staining was performed using the standard method of deparaffinization and gradual rehydration, followed by heat-mediated antigen retrieval in tris-EDTA or citrate buffer. The sections were further treated with 3% hydrogen peroxide in methanol to block endogenous peroxidase activity. Primary antibody incubation was done at room temperature for 2 h followed by incubation with an appropriate horseradish peroxidase-conjugated secondary antibody for 30 min. Detection was carried out using supersensitive polymer-horseradish peroxidase immunohistochemistry/diaminobenzidine detection system. The staining for different proteins was performed using the same set of tumor and normal samples. Analysis was performed by the pathologist determining the staining intensity of each section (with minimum of 10% cells positive) and grading from 0 (no staining) to 3+ (highest intensity). The list of primers used in the study is provided in Supplementary Table S1. Proliferation and soft agar assay Proliferation of MCF-7 and MDA-MB-231 cells (from ATCC, Manassas, VA, USA) was assayed using bromodeoxyuridine cell proliferation assay kit (Calbiochem, San Diego, CA, USA) following the manufacturer's protocol (24 h). For soft agar assay, 3,000 and 5,000 cells of MDA-MB-231 and MCF-7 (both from ATCC) were seeded for 12 and 18 days, respectively. The experiments were performed in triplicates for three independent times.

Reporter assay
Matrix metalloproteinase-2 (MMP2) and VEGF reporter constructs were gifts from Etty Benveniste, University of Alabama, and Gail C. Fraizer, Kent State University, respectively. Luciferase reporter assays were carried out using promega kit (E4030, Madison, WI, USA) following the manufacturer's protocol. Renilla luciferase under a TK promoter was used as an internal control. The experiment was performed two independent times in triplicates.

Migration/invasion assay
For scratch assay, cells were cultured in a six-well plate till a confluent monolayer was formed. A scratch was made through the center of the well with a tip and images were recorded before and after activin-A treatment. Transwell migration and invasion assays were performed following the manufacturer's protocol, using BD BioCoat Matrigel (NJ, USA) invasion/ control chambers. The experiments were performed in triplicates and repeated two independent times. The images were recorded under a microscope with a fixed camera (×2 magnification) and cells were counted from several random fields.

Zymography
Equal number of cells was seeded and the supernatant was collected after 24 h of serum starvation (stable clones) or activin-A treatment (parental cells). The sample was given a spin and equal volume was loaded into a non-denaturing gelatin containing gel (0.1%). The gel was incubated for 8-10 h followed by Coomassie staining. The images were recorded using a Uvipro platinum gel imaging system (Cambridge, UK).

Mice experiments
For tumor xenograft growth, 5 × 10 5 of MDA-MB-231 and 20 × 10 5 MCF-7 cells were injected into the flank region of 4-to 5-week-old female nude mice. The animals were allocated by the animal facility and 10 animals were placed in each group randomly. Tumor formation was followed for 8 and 10 weeks, respectively. The tumors were excised and the weight of the tumors was plotted. To study metastatic spread of tumor cells, 20 × 10 5 cells were injected into the tail vein of nude mice (5 animals each group). Mice were killed after 10 weeks of injection and various organs were analyzed for nodule formation. The size measurement was carried out on a Gatan Microscopy Suite (Gatan, Pleasanton, CA, USA), using a line scan that gives the number of pixels in a particular direction. The t-test was performed using Graphpad prism software (La Jolla, CA, USA) to evaluate the statistical significance of the data and a P-value above 0.05 was considered significant. All the experiments were performed in accordance

Activin-A signaling is active in breast cancer
Microarray study done in our laboratory revealed INHBA and various other genes involved in the activin signaling pathway to be differentially expressed in breast cancers (invasive ductal carcinoma) compared with normal tissue samples, suggesting activation of this signaling pathway (Supplementary Figure S1).
This differential expression of the INHBA was not dependent on the category of breast tumors such as estrogen receptor, progesterone receptor, or Her2 status. We used an independent set of 15 normal (from non-cancer individuals) and 30 breast cancer samples (grade 3), to analyze the expression of various components of activin-A signaling pathway. As summarized in Table 1 and shown in Figure 1a, we observed upregulation of various components of activin-A signaling pathway including INHBA, SMAD2, and activin type II receptor. Interestingly, some of the negative regulators of activin signaling pathway such as follistatin, β-glycan, IGSF1, and IGSF10 showed downregulation in   breast tumors compared with normal samples. Furthermore, although TGF-β1 expression was upregulated, TGFβRII showed a very significant downregulation in tumors, compared with normal tissues (Supplementary Figure S2). We also analyzed various available breast cancer gene expression data sets (oncomine.org).
In accordance with our study, analysis of these data sets shows that activin-A signaling components are frequently deregulated in breast cancers (Supplementary Table S2). TGF-β/activin signaling has been shown to be opposed by BMP signaling pathways in development and disease. 1,25,26 In congruent with this, BMP isoforms 2, 3, and 6, and various other genes involved in BMP signaling such as SMAD1, ZNF521, RGMA, and Gremlin1 were found to be downregulated. To confirm our results, we performed immunohistochemistry with another set of 13 normal and 29 tumor samples. As shown in Figure 1b (i), most of the tumors have higher levels of activin-A compared with normals tissues. Activation of activin signalling results in phosphorylation of SMAD2 and SMAD3. In good correlation, breast tumors showed increased phosphoSMAD2 and phosphoSMAD3 levels compared with normal tissues (Figure 1b, ii and iii). In addition, in the same set of tumors, BMP2 staining showed a reduced expression pattern (Figure 1b, iv), compared with normal tissues. We also analyzed the expression of inhibin in some normal and breast tumor samples (data not shown) and found that most of the tumor samples have very low levels of inhibin compared with normal tissues. This suggests that overexpression of INHBA results in reduced inhibin expression, possibly due to INHβA homo-dimerization. In conclusion, our data shows that breast tumors have higher levels of activin-A and low levels of various BMPs, suggesting activation of activin signaling pathway in these tumors. Further, we also analyzed various publicly available breast cancer clinical data sets (co.bmc.lu.se/gobo/gsa.pl). As represented in Figure 1c (ii), the analysis shows that INHBA expression is negatively correlated with the overall survival of grade 3 breast cancer patients. However, there was no significant correlation between INHBA expression and overall patient survival, when all    Activin-A induces epithelial-mesenchymal transition in breast cancer cells Epithelial-mesenchymal transition (EMT), a process in which epithelial cells lose some of their characteristics and acquire properties of mesenchymal cells, has been proposed to have a pivotal role in invasion/metastasis of cancer cells. Unlike TGF-β, the role of activin-A in this regard is unknown. Western blot analysis shows that treatment or overexpression of activin-A leads to a decrease in the expression of E-cadherin and an increase in the expression of various mesenchymal markers in MCF-7 cells (Figure 3a, i and ii). Interestingly, we could also observe a mesenchymal phenotype in activin-A-overexpressing MCF-7 cells (Supplementary Figure S9). EMT is also marked by stress fiber formation associated with changes in the cytoskeleton. Confocal microscopy indicated that treatment with activin-A leads to downregulation of E-cadherin, induction of α-smooth muscle actin, and stress fibre formation in MCF-7 cells (Figure 3b). Similar results were observed on treatment or knockdown of activin-A in MDA-MB-231 cells (Figure 3a, iii and iv). In conclusion, similar to TGF-β, activin-A also is an inducer of EMT in cancer cells.
Activin-A promotes migration and invasion of MDA-MB-231 breast cancer cells EMT has been associated with migratory and invasive behavior of cancer cells. 27 Hence, we studied whether activin-A has any effect on migratory and invasive behavior of MDA-MB-231 cells. Woundhealing (scratch) assay showed that activin-A treatment of MDA-MB-231 cells promotes their migration, whereas knockdown of activin-A results in decreased migration of these cells (Figure 4a, i and ii). We also performed transwell migration assay and observed that treatment of activin-A increases, whereas stable knockdown of activin-A decreases migration of MDA-MB-231 cells (Figure 4b, i and ii). To assess whether activin-A affects the invasive behavior of MDA-MB-231 cells, we performed Matrigel invasion assay and observed that treatment with activin-A increases, whereas knockdown of activin-A reduces invasive potential of MDA-MB-231 cells (Figure 4c, i and ii). Next, we performed zymography to analyze MMP2 and MMP9 activity. The results show that activin-A treatment increases, whereas its knockdown decreases MMP2 activity in MDA-MB-231 cells (Figure 4d, i and ii). Further, we used Luciferase-conjugated MMP2 promoter and assayed its inducibility by activin-A. Activin-A treatment of HEK-293T cells transfected with the reporter construct showed a significant increase in the promoter activity (Figure 4e). These results suggest that activin-A can activate transcription of MMP2 and promote migration and invasion of breast cancer cells.  Many recent studies have shown that SMAD3 has an important role in TGF-β-induced EMT, migration, and invasion. 28,29 Hence, we used a specific inhibitor (SIS3) and small hairpin RNA-mediated stable knockdown of SMAD3 in MDA-MB-231 cells. The results show that ablation of SMAD3 activity leads to blockade of activin-A-induced EMT and invasion in these cells (Figure 4f-i). Together, these results suggest that activin-Ainduced phenotype is dependent on SMAD3 signaling in MDA-MB-231 cells.

Control
Activin-A promotes tumorigenicity of breast cancer cells The process of tumor formation is highly complex. TGF-β signaling has been shown to have an important role in the establishment of tumors in vivo. [30][31][32] Hence, we wanted to investigate how activin-A would affect the tumor-forming ability of cancer cells in vivo. We injected activin-A-overexpressing MCF7 cells in the flank of immunocompromised mice and followed till the tumors reached to a prominent size. Our results show that activin-Aoverexpressing MCF-7 cells have better tumor-forming ability in comparison with control cells (Figure 5a, i). We also injected only half a million activin-A knockdown MDA-MB-231 and control cells (optimal cell number generally used is two million) subcutaneously in nude mice. Although 7 out of 10 animals formed tumors in control mice, only 3 out of 10 animals injected with activin-A knockdown cells could lead to tumor formation (Figure 5a, ii). Significant differences in the weight of the tumors were also observed. We also performed immunohistochemistry on the tumors formed by MCF7 cells (Figure 5b). We observed that MCF7 overexpressing activin-A tumors have higher Ki-67 percentage (~80%) as compared with control tumors (~50%). In addition, staining for various EMT markers confirmed the mesenchymal state of tumors formed by activin-A-overexpressing MCF-7 cells. Recruitment of blood vessels or de-novo formation of blood vessels can influence the tumor growth in vivo. Hence, we investigated the expression of VEGF, which is known to promote angiogenesis. Activin-A treatment or its stable overexpression in MCF-7 cells induced VEGF expression (Figure 5c, i and ii). Further, we performed Luciferase reporter assay to study the regulation of VEGF promoter by activin-A. Activin-A treatment of HEK-293T transfected with Luciferase-conjugated VEGF promoter showed  that activin-A induces activity of VEGF promoter significantly (Figure 5d). We also performed tail vein injections with activin-Aoverexpressing MCF-7 cells. Even though we did not find a significant difference in the number of nodules formed, we observed that activin-A-overexpressing cells formed much bigger nodules as compared with the control cells (Figure 5e, iii and iv). This suggests that growth of colonized MCF-7 cells is promoted by activin-A expression. Tumor-forming ability of cancer cells and aggressiveness of various cancers has been associated with the presence of cells having stem-like phenotype. 33,34 Activin-A has been shown to regulate expression of various stemness markers such as Nanog, Sox2, and Oct4 in various cells. 35,36 Hence, to investigate whether activin-A expression affects stemness of breast cancer cells, we analyzed population of CD44 high and CD24 low cells. Fluorescence-activated cell sorting analysis of activin-A-overexpressing MCF-7 cells and activin-A knockdown MDA-MB-231 cells shows that activin-A expression leads to enrichment of breast cancer stem-like cells (Figure 5f, i). In addition, treatment of MCF-7 and MDA-MB-231 cells with activin-A leads to an increase in expression of several stemness markers (Figure 5f, ii). Taken together, these results suggest that activin-A has multiple effects on tumor establishment and progression in vivo.

DISCUSSION
Activin-A, a member of the TGF-β superfamily, binds to its cognate receptor and activates SMAD2/3 signaling pathway. However, unlike TGF-β, the role of activin in cancer is not well known. Here we show that high-grade breast tumors have activation of activin-A signaling pathway. Interestingly, we did not observe much significant change in the expression of TGF-β ligands. We also demonstrate that many components and regulators of activinsignaling pathway are deregulated, favoring the activation of this   tumors revealed significant downregulation of SMAD1, which acts as a downstream mediator of the BMP signaling pathway. Downregulation of SMAD1 in breast tumor cells will abrogate any BMP signaling (locally or at a distant site) in these cells. BMP signaling pathway has been shown to have an antagonistic role to TGF-β/activin signaling in various physiological and pathological conditions. 37,38 In addition, recent studies have shown that BMP signaling has a tumor-suppressive function in cancers. 39 Hence, hyperactivation of activin-A signaling and loss of BMP signaling may have a critical role in the clinical outcome of breast cancers. We show that INHBA expression correlates negatively with the overall survival of high-grade breast tumors and metastasis-free survival of breast cancer patients. In agreement with these results, FST and TGFBR3 expression confers a good prognosis for breast cancers. It suggests that activin-A signaling has an important role in the dissemination of breast cancer cells and, hence, may determine the outcome of the disease. We show that overexpression or knockdown of activin-A affects anchorageindependent growth of breast cancer cells. We also show that activin-A induces EMT and promotes migration and invasion in breast cancer cells. Primarily, the SMAD2/3 signaling pathway inhibits proliferation of normal epithelial cells and, hence, is considered to have a tumor-suppressor function. Similar to TGF-β, proliferation of normal or tumor cells in the earlier stages of tumor development may be inhibited by activin-A. However, it is interesting to note that this signaling pathway still remains intact in majority (~98%) of the breast cancers. 40 With the disease progression, these cells may become refractory to the growth inhibitory effect of activin-A signaling, as exemplified in prostate cancer. Activin has been shown to inhibit proliferation of LNCaP and DU145 (low-and moderate-grade PCa) but not PC3 cell line (high grade). Interestingly, circulating levels of activin-A were demonstrated to increase significantly in metastatic prostate and breast cancers. 41,42 In addition, SMAD deletions are observed only in 1-2% of breast cancers. 40 In other words, this signaling pathway remains intact in most of the breast cancers. Given the abundant expression of activin-A in tumor cells and the presence of an intact signaling pathway, it is reasonable to believe that these cells have acquired resistance to the growth inhibitory effect of activin-A. Moreover, it is quite comprehensible that SMADs may then have an important role in tumor progression. Although TGF-β is well known to induce EMT in various cell lines, not much is known about activin-A in this regard. EMT is considered to be a prerequisite for tumor cells to migrate and invade into neighboring tissues, and hence lead to metastasis. In recent times, even nodal, another member of the TGF-β superfamily that signals through the activin receptors has been reported to promote cancer progression. Our results demonstrate that activin-Ainduced expression of EMT markers and invasion is SMAD3 dependent. SMAD3 has been shown to promote growth of breast cancer cells in nude mice. 43 However, activin-A may also activate non-SMAD signaling pathways, which may contribute to its pro- Activin-A promotes breast cancer M Bashir et al tumorigenic actions. We observed that activin affects tumor formation and colonization of tumor cells in nude mice. The data suggest that activin-A expression affects the establishment of tumors, which is a very complex process. Establishment of tumors or metastasis is also influenced by interaction between tumor cells and the microenvironment. Hence, factors that can modulate the tumor microenvironment can have a key role in cancer progression. In this context, activin may also be important in the establishment of metastases from disseminated cells by modifying the tumor niche. 22 One of the factors known to be important for tumor growth is VEGF, which leads to recruitment of blood vessels, hence providing nourishment to the tumor cells. As we have demonstrated here, activin-A induces VEGF expression and may influence the proliferation of the cancer cells in vivo. It may, to some extent, also explain how activin overexpression promotes proliferation of tumor cells in vivo. It will be interesting to evaluate how activin-A expression would affect the response of breast cancer patients to various conventional treatments. Activin-A is known to have an important role in maintenance of stem cell phenotype 11 and stemness. 36 In addition, EMT has been shown to induce drug resistance and stem cell-like phenotype in cancer cells. 44 It has also been shown that tumors that are more aggressive are less differentiated and vice versa. We show that activin-A expression affects the stemness of breast cancer cells and hence may contribute to the aggressiveness of the disease. Our study highlights the importance of activin signaling in the progression of breast tumors. Administration of a circulating dominant-negative type II TGF-β receptor in mice has been shown to prevent metastasis of breast tumors. 45 However, our study emphasizes the role of activins and the expression pattern of these ligands should be considered from a clinical perspective. It is possible that different tumors may use either TGF-β or activin in a context-dependent manner. Hence, it is important to carefully examine the expression of these ligands while designing strategies to block their actions. Although most studies in the past have focussed on the role of TGF-β, this study emphasizes activin-A's role in the progression of breast tumors. Interestingly, many other cancers have been reported to have increased expression of activins, suggesting that activation of the activin signaling pathway may be widely involved in carcinogenesis. As activins have multiple roles in physiological context, their role in cancer may be equally diverse. Importantly, our study emphasizes the role of SMAD pathway in the progression of breast tumors and targeting this pathway may be a useful strategy in the treatment of breast cancers. In conclusion, we show that activin-A induces EMT, promotes invasion, and enhances the tumor-forming ability and metastatic growth of breast cancers.