The Transcriptional Regulators TAZ and YAP Direct Transforming Growth Factor β-induced Tumorigenic Phenotypes in Breast Cancer Cells*♦

Background: The TGFβ and Hippo pathways are dysregulated in metastatic breast cancers. Results: TGFβ-induced cues and nuclear TAZ/YAP converge at the transcriptional level to control gene expression important for tumorigenesis. Conclusion: TAZ/YAP are required to promote TGFβ-induced tumorigenic phenotypes in breast cancer cells. Significance: Our study reveals novel cross-talk between the TGFβ pathway and TAZ/YAP in late-stage breast cancers. Uncontrolled transforming growth factor-β (TGFβ) signaling promotes aggressive metastatic properties in late-stage breast cancers. However, how TGFβ-mediated cues are directed to induce tumorigenic events is poorly understood, particularly given that TGFβ has clear tumor suppressing activity in other contexts. Here, we demonstrate that the transcriptional regulators TAZ and YAP (TAZ/YAP), key effectors of the Hippo pathway, are necessary to promote and maintain TGFβ-induced tumorigenic phenotypes in breast cancer cells. Interactions between TAZ/YAP, TGFβ-activated SMAD2/3, and TEAD transcription factors reveal convergent roles for these factors in the nucleus. Genome-wide expression analyses indicate that TAZ/YAP, TEADs, and TGFβ-induced signals coordinate a specific pro-tumorigenic transcriptional program. Importantly, genes cooperatively regulated by TAZ/YAP, TEAD, and TGFβ, such as the novel targets NEGR1 and UCA1, are necessary for maintaining tumorigenic activity in metastatic breast cancer cells. Nuclear TAZ/YAP also cooperate with TGFβ signaling to promote phenotypic and transcriptional changes in nontumorigenic cells to overcome TGFβ-repressive effects. Our work thus identifies cross-talk between nuclear TAZ/YAP and TGFβ signaling in breast cancer cells, revealing novel insight into late-stage disease-driving mechanisms.

Elevated nuclear levels of the transcriptional regulators TAZ and YAP (TAZ/YAP) are associated with a broad range of aggressive cancers (1). For instance, the extent of nuclear TAZ or YAP levels corresponds with breast cancer tumor grade (2)(3)(4). In breast cancer cells, enhanced nuclear TAZ and YAP levels promote oncogenic transformation and endow cells with tumorigenic properties, including the ability to proliferate, subvert apoptotic cues, migrate, invade, and grow under anchor-age-independent conditions (5)(6)(7)(8)(9). Moreover, high nuclear TAZ levels induce cancer stem cell-like activity (2,10) and promote evasion of certain breast cancer drug therapies (2,11). Thus, understanding the roles of TAZ/YAP is critical for directing efficient breast cancer therapies.
The tumor initiating activity of TAZ/YAP relies on their binding to the TEAD family of transcription factors (TEAD1-4) (10,12,13), indicating that together these factors direct a tumorigenic transcriptional program. Supporting this premise, TAZ/YAP⅐TEAD complexes directly promote the expression of genes encoding oncogenic factors, such as CTGF (also known as CCN2) and CYR61 (also known as CCN1) (12,13), which contribute to human breast cancer progression (14). Nuclear TAZ/YAP activity is highly regulated and governed in large part by the Hippo pathway-regulated LATS1 and LATS2 kinases (15). LATS1/2 kinases phosphorylate TAZ/YAP on conserved serine residues, which promote 14-3-3 binding and subsequent sequestration in the cytoplasm (16,17), and also prime TAZ/YAP for further phosphorylation by CK1⑀/␦ kinases that evoke TAZ/YAP degradation via proteasome-dependent mechanisms (18,19). Additional phosphorylation events destabilize TAZ, including those regulated by Wnt, phosphatidylinositol 3-kinase, and GSK3␤ (20,21). Thus, dysregulation of multiple upstream signals likely contributes to the aberrant nuclear TAZ/YAP activity that is observed in cancers.
TAZ/YAP modify the activity of other transcription factors besides TEADs, including the transforming growth factor-␤ (TGF␤)-activated SMAD complexes (22). TGF␤ is the prototypic member of a family of secreted factors that regulates numerous developmental and homeostatic processes (23). SMAD2 and SMAD3 (SMAD2/3) are the primary mediators of TGF␤-induced transcription. SMAD2/3 are phosphorylated by TGF␤-bound membrane receptors, which induce binding to SMAD4 (24,25), forming active transcriptional complexes that accumulate in the nucleus upon binding to TAZ/YAP (26). In cancer, the role of TGF␤ is complex, as it can suppress early oncogenic events but also promote aggressive late-stage metastatic phenotypes (27,28). What mechanistically distinguishes the different TGF␤-dependent responses is poorly understood.
Several lines of evidence indicate that TGF␤, like TAZ/YAP, promotes aggressive tumorigenic properties in late-stage breast carcinomas (29,30). Given that TAZ/YAP bind to SMAD transcription factors and direct TGF␤ signaling in other contexts (26,31,32), we sought to characterize whether TAZ/YAP define TGF␤-mediated tumorigenic cues in breast cancer cells.
Our observations indicate that TGF␤-induced tumorigenic events, such as increased cell migration, invasion, and anchorage-independent growth, require TAZ/YAP. Our data also indicate that, like TAZ/YAP, the TEAD transcription factors interact with TGF␤-induced SMAD2/3 in the nucleus, suggesting that TAZ/YAP⅐TEAD⅐SMAD2/3 complexes coordinate transcriptional events in a concerted manner. Genome-wide microarray analysis of gene expression changes that occur upon knockdown of TAZ/YAP or TEADs, or inhibition of TGF␤ signaling, revealed that TAZ/YAP, TEAD, and TGF␤ regulate overlapping target genes. Interestingly, the direct gene targets NEGR1 and UCA1, which are synergistically regulated by TAZ/ YAP, TEAD, and TGF␤, are necessary for maintaining tumorigenic activity in metastatic breast cancer cells, suggesting that the convergence of TAZ/YAP⅐TEAD-TGF␤ signals is critical for driving late-stage breast cancer phenotypes. Supporting this premise, expression of nuclear-localized TAZ or YAP mutants direct transcriptional events that sensitize untransformed breast cancer cells to adopt tumorigenic phenotypes in response to TGF␤, while also suppressing TGF␤-induced cytostasis. These findings reveal novel cross-talk between TGF␤ and Hippo signaling that we propose is important for late stage tumorigenic events in breast cancer.
Immunofluorescence and Proximity Ligation Assay (PLA)-Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100/PBS, blocked in 2% BSA/PBS, and probed with primary and secondary antibodies outlined in supplemental Table S2. LM2-4 cells were treated with or without TGF␤1 (500 pM, R&D Systems) or SB-431542 (5 M, Sigma) for 24 h before fixing. For the PLA, 2 LM2-4 cells were plated on 96-well microplates (Falcon) and treated with or without TGF␤1 for 24 h. Cells were fixed and permeabilized as described, blocked according to the manufacturer's protocol (Duolink), and probed with the primary antibodies outlined in supplemental Table S2. Anti-mouse MINUS and anti-rabbit PLUS PLA probes (Duolink) were used. Nuclei were stained with Hoechst. All immunofluorescence was visualized by confocal microscopy (LSM 700), and images were processed using Volocity software (PerkinElmer Life Sciences). Images were quantitated using ImageJ software.
Mammospheres-LM2-4 cells were transfected with siRNA, dissociated 24 h later, and resuspended in Mammary Epithelium Growth Medium (MEGM; Lonza) supplemented with B27 (Invitrogen), 20 ng/ml epidermal growth factor (EGF; Pepro-Tech), 20 ng/ml basic fibroblast growth factor (bFGF; Pepro-Tech). Single cells were seeded at 5 ϫ 10 3 cells/ml in 6-well ultra-low attachment plates (Corning Glass) and treated with or without TGF␤1 or SB-431542. Primary spheres were photographed after 7 days and either lysed for RNA (Quick-RNA MiniPrep, Zymo Research) to examine knockdown or dissociated in 0.05% trypsin for 10 min and resuspended as single cells in MEGM for passage. Secondary spheres were photographed after an additional 14 days. Images were analyzed using ImageJ software, and statistics were calculated using Prism software (GraphPad) using a two-tailed unpaired Student's t test.
Immunoprecipitation and Immunoblots-LM2-4 cells examined for endogenous protein expression were treated with or without TGF␤1 or SB-431542 for 2 h and were lysed and examined by immunoblotting. Transfected HEK293T cells expressing the indicated proteins were lysed, subjected to immunoprecipitation using anti-FLAG-conjugated protein-G beads (Sigma), and analyzed by immunoblotting. MCF10A doxycycline-inducible cells were treated with or without doxycycline (0.1 to 100 ng/ml) or TGF␤1 for 24 h and were lysed and examined by immunoblotting. Antibodies are outlined in supplemental Table S2.
Cell Morphology Analysis, Wound Healing, and Transwell Migration-Low density MCF10A doxycycline-inducible cells were pretreated with doxycycline (100 ng/ml, Clontech) for 24 h and then treated with or without TGF␤1 for an additional 24 h. For the wound-healing scratch assays, LM2-4 cells were transfected with siRNA and 24 h later were treated with or without TGF␤1 or SB-431542 for an additional 24 h. MCF10A doxycycline-inducible cells were treated with or without doxycycline or TGF␤1 for 24 h. Monolayers were wounded and photographed after an additional 24 h (LM2-4) or 12 h (MCF10A). Images were analyzed using ImageJ software, and statistics were calculated using Prism software (GraphPad) using a twotailed unpaired Student's t test. Cells used in the transwell assay were transfected with siRNA, trypsinized 24 h later, and resus-pended in low serum media (0.25% FBS). Cells were plated at 10 5 cells/ml on 0.4-M transwell filters (BD Biosciences) pretreated for 24 h with 1 g/ml fibronectin (Millipore). Media ϩ 10% FBS were used in the bottom chamber. Cells were allowed to migrate for 24 h in the presence of TGF␤1 and were subsequently stained with 0.5% crystal violet.
Three-dimensional Invasion-Stable knockdown of TAZ and YAP in LM2-4 cells was accomplished by lentivirus-mediated transduction of shRNA using the pLKO1-puro and pLKO1-neo vectors and subsequent selection with 2 g/ml puromycin and 1.5 mg/ml G418. The shRNA sequences used are listed in supplemental Table S1. Single cells were plated on 100% growth factor-reduced Matrigel (BD Biosciences) using the overlay method (33). Assay media contained 2% Matrigel added to supplemented MEGM, and cells were cultured with puromycin and G418 with medium changes every 3 days. TGF␤1 and SB-431542 were added after 9 days and then cultured for an additional 3 days before being photographed.
Microarrays-LM2-4 cells were transfected with control siRNA or siRNAs targeting TAZ/YAP or all four TEADs (outlined in supplemental Table S1) and were treated 24 h later with TGF␤1 or SB-431542 for an additional 24 h. Total RNA was isolated and purified by Quick-RNA MiniPrep (Zymo Research). Twelve microarrays in total were performed, with each condition carried out three times on separate days. The Boston University Microarray Core generated the data using the Affymetrix Human Gene 1.0 St Array, which covers 27,300 probe sets. The data were filtered using a moderated p value of less than 0.01, and the average fold change in expression of each gene, for each condition, relative to the siCTL ϩ TGF␤ sample was calculated. Fold expression changes relative to siCTL ϩ TGF␤-treated cells were calculated, and statistical significance was assessed using a moderated t test and p values. Hierarchical gene clustering was performed on overlapping genes displaying a p value of Ͻ0.01 with the open source program Cluster 3.0 (34).
Quantitative Real Time PCR (qPCR)-LM2-4 cells were transfected with siRNA and were treated 24 h later with or without TGF␤1 or SB-431542 for an additional 24 h. MCF10A doxycycline-inducible cells were treated with or without doxycycline (0.1 to 100 ng/ml) or TGF␤1 for 24 h. Total RNA was purified using Quick-RNA MiniPrep kit, and cDNA synthesis was performed using 1 g RNA and iScript cDNA synthesis kit (Bio-Rad) according to manufacturer's protocol. qPCR was performed using Fast SYBR Green enzyme (Applied Biosystems) and measured on ViiA 7 real time PCR system (Applied Biosystems). Transcript levels were analyzed using the ⌬⌬C t method and normalized to GAPDH. Primer sequences are indicated in supplemental Table S3.
Cell Proliferation and Cell Cycle Analysis-MCF10A doxycycline-inducible cells were plated (5 ϫ 10 4 cells) and treated with doxycycline with or without TGF␤1 (day 0). Cells were counted each day for 6 consecutive days (day 1-6). For cell cycle analysis, MCF10A doxycycline-inducible cells were treated with doxycycline with or without TGF␤1 for 48 h. 1 ϫ 10 6 cells were fixed overnight in 100% ethanol and stained using 50 g/ml propidium iodide (Sigma) and 100 g/ml RNase A (Sigma). Samples were acquired on the FACScan (BD Biosciences), collecting 1 ϫ 10 4 events, and analyzed using FlowJo software (Tree Star). Statistical analysis was conducted using a twotailed unpaired Student's t test.

Nuclear TAZ/YAP Are Required to Promote TGF␤-induced
Tumorigenic Phenotypes in Breast Cancer Cells-In cancer, the role of TGF␤ is complex, as it can suppress early oncogenic events, such as cell cycle progression, but can also promote late-stage metastatic phenotypes (27,28). What distinguishes these different TGF␤-dependent responses is poorly understood. Several lines of evidence indicate that nuclear TAZ/YAP, like TGF␤, induce tumorigenic properties in late-stage breast carcinomas (29,30). In untransformed mammary epithelium, TAZ/YAP localization is restricted to the cytoplasm by cell compaction/polarity-regulated cues (9,32). Dysregulation of cell polarity cues, which is a hallmark of cancer progression (36), induces nuclear TAZ/YAP localizations. Given our prior work showing that TAZ/YAP bind to and regulate the localization and activity of TGF␤-activated SMAD transcription factors (26,32), we sought to test whether TAZ and/or YAP promote TGF␤-induced tumorigenic events. We began our analysis by examining the relationship between TAZ and YAP localizations and the TGF␤-induced cytostatic response in a panel of mammary epithelial and breast cancer cell lines. Based on published data, we divided the panel into cells that are responsive to TGF␤-induced cytostasis (MCF10A, BT20, HMLE, HS578T, MCF7, and MCF-12) and cells in which TGF␤ induces pro-tumorigenic signals but not growth arrest (MDA-MB-231, MDA-MB-231-LM2-4, SKRB3, and SUM149) (37)(38)(39)(40)(41)(42)(43)(44). Interestingly, we observed that cells displaying high levels of nuclear TAZ/YAP correlate with those in which TGF␤ induces tumorigenic cues (Fig. 1A).
To further investigate this relationship, we sought to determine the roles of nuclear TAZ/YAP in the human MDA-MB-231-LM2-4 (herein referred to as LM2-4) metastatic breast cancer cell line (45), a highly aggressive derivative of triplenegative basal subtype MDA-MB-231 cells (46). A fraction of LM2-4 cells in culture are capable of generating clonal mam-mospheres under anchorage-independent conditions (Fig. 1B), which is often used as a measure of the self-renewing potential of tumorigenic cells in vitro (47). TGF␤ treatment of LM2-4 cells led to dramatic increases in the number and size of mammospheres observed (Fig. 1B), similar to that observed with TGF␤ treatment of other mammary cells (30). The self-renew- , YAP (siYAP), or TAZ and YAP (siTAZ/YAP). Cells were left untreated, treated with TGF␤ or SB-431542 ϩ TGF␤, and grown in anchorage-independent conditions. Primary mammospheres were examined for knockdown or were passaged into secondary spheres. Secondary mammospheres following SB-431542 (SB) treatment, or transfection with siTAZ, siYAP, or siTAZ/YAP, were unable to be determined due to low numbers. Representative images are shown, and three independent experiments from each condition were quantitated, measuring the number of colonies formed and the size of each colony. Black error bars represent the average ϩ S.E., and red error bars represent the average Ϯ S.E., *, p Ͻ 0.025; **, p Ͻ 0.005; ***, p Ͻ 0.0001 (t test). C, LM2-4 cell lysates were immunoblotted to examine endogenous levels of the indicated proteins upon TGF␤ or SB-431542 treatment compared with GAPDH (loading control). D, LM2-4 cells were transiently transfected with siCTL or siTAZ/YAP. Cells were left untreated, treated with TGF␤, or SB-431542 ϩ TGF␤. Monolayers were wounded and analyzed for cell migration. E, LM2-4 cells stably expressing control shRNA (shCTL) or shRNA targeting TAZ and YAP (shTAZ/YAP) were treated with TGF␤ or SB-431542 ϩ TGF␤ and incubated in three-dimensional Matrigel culture conditions. Representative images from three independent experiments are shown.
ing properties of the cells within the mammospheres were assessed by their ability to form secondary clonal spheres (47), and we found that TGF␤ also promoted secondary mammosphere formation. Co-treatment of the cells with the TGF␤ receptor antagonist SB-431542 abolished the formation of primary mammospheres, validating that the observed effects are indeed generated via canonical TGF␤ receptor-mediated signals (Fig. 1B) (48,49). As expected, SB-431542 treatment eliminated the TGF␤-induced phosphorylation of SMAD2 and SMAD3 in these cells (Fig. 1C). Individual TAZ or YAP knockdown also repressed the number and size of TGF␤-induced mammospheres (Fig. 1B). However, simultaneous knockdown of both TAZ and YAP dramatically reduced mammosphere formation (Fig. 1B), indicating redundant roles for TAZ and YAP in transducing TGF␤-mediated cues required for anchorageindependent growth.
We further investigated other hallmark tumorigenic properties that may be mediated by TGF␤ and TAZ/YAP in metastatic breast cancers, including cell migration and invasion (36). We found that treatment of LM2-4 cells with TGF␤ led to increases in cell migration in an in vitro wound-healing scratch assay (Fig.  1D), similar to previous work (50). As expected, co-treatment with the TGF␤ receptor antagonist SB-431542 blocked TGF␤induced cell migration (Fig. 1D). Simultaneous knockdown of TAZ/YAP using siRNA also abolished TGF␤-induced LM2-4 cell migration (Fig. 1D). Similarly, SB-431542 treatment or shRNA-mediated TAZ/YAP knockdown abolished the ability of three-dimensional colonies of LM2-4 cells to invade into the surrounding Matrigel matrix in the presence of TGF␤ (Fig. 1E). Taken together, our observations indicate that TAZ/YAP are critical mediators of TGF␤-induced tumorigenic events, including mammosphere formation, cell migration, and invasion.
TAZ/YAP, TEADs, and SMADs Converge to Regulate a TGF␤-induced Transcriptional Program in Breast Cancer Cells-Studies indicate that TAZ/YAP-induced cell transformation relies on the recruitment of TAZ/YAP to DNA by the TEAD family of transcription factors (TEAD1-4) (12,13). TAZ and YAP also bind TGF␤-activated SMAD complexes to control SMAD localization and activity in a variety of cell types, including mammary epithelial cells (26,32). Recent work has shown that TAZ/YAP⅐TEAD⅐SMAD2/3 complexes control transcriptional events important for maintaining human embryonic stem cell pluripotency (35). Thus, we hypothesized that similar complexes may also be present in late stage breast cancers such that TEAD and SMAD transcription factors can cooperatively facilitate TAZ/YAP-mediated tumorigenic activity. We found that TEAD2 and TEAD4 associate with SMAD3, as well as YAP ( Fig. 2A), and these interactions were unaffected by stimulation with a constitutively active TGF␤ receptor (TGF␤R1-T240D (51)). Given that TAZ/YAP exhibit a predominantly nuclear localization in LM2-4 cells (Fig. 2B), we speculated that TAZ/ YAP⅐TEAD might be interacting with TGF␤-activated SMAD2/3 to specify pro-tumorigenic transcriptional events. To acquire both protein interaction and localization information, we performed in situ PLA. PLA is a sensitive technique used to visualize the localization and association of endogenous protein complexes (proteins localized within 40 nm of each other) by microscopy (52). Using PLA, we observed TAZ/ YAP⅐SMAD2/3 interactions in both the nucleus and cytoplasm of untreated LM2-4 cells (Fig. 2C). Upon TGF␤ treatment, nuclear TAZ/YAP-SMAD2/3 binding became much more apparent in the nucleus (Fig. 2C), consistent with nuclear TAZ/ YAP⅐SMAD2/3 complexes directing transcriptional events (26,32). We also detected endogenous TAZ/TEAD1 interactions in the nucleus of LM2-4 cells with or without TGF␤ stimulation (Fig. 2D), which were increased slightly upon TGF␤ treatment (Fig. 2D). TEAD1⅐SMAD2/3 interactions were readily detected in the nucleus of LM2-4 cells, particularly after TGF␤ treatment (Fig. 2E), suggesting these complexes stabilize upon nuclear accumulation of SMADs. Taken together, our observations indicate that TAZ/YAP, TEAD, and SMAD interact in TGF␤-stimulated metastatic breast cancer cells and suggest that they may form transcriptional complexes that function together in the nucleus.
To explore the possible overlap in transcriptional activity by TAZ/YAP, TEAD, and SMAD complexes, we used microarrays to compare the global expression profiles of LM2-4 cells treated as follows (complete data available in supplemental Table S4): 1) transfected with control siRNA (siCTL) and treated with TGF␤; 2) transfected with siRNA targeting both TAZ/YAP (siTAZ/YAP) and treated with TGF␤; 3) transfected with siRNA targeting all four TEAD (TEAD1-4) family members (siTEAD) and treated with TGF␤; and 4) transfected with control siRNA (siCTL) and treated simultaneously with TGF␤ and SB-431542. In terms of significant gene expression differences (p value Ͻ 0.01) relative to siCTL ϩ TGF␤ treatment, 461 genes overlapped between siTAZ/YAP and siTEAD conditions (Fig. 3A). This gene set displayed a high degree of correlation in expression (R ϭ 0.86). The expression of 594 genes changed following SB-431542 treatment, and of these, 176 genes overlapped with siTAZ/YAP conditions. Of these 176 genes, 80 were also altered following TEAD knockdown (Fig. 3A).
Interestingly, genes for which expression was altered among all three experimental conditions exhibited distinct expression correlations. Unbiased clustering segregated TAZ/YAP⅐TEAD-TGF␤-regulated genes into four different groups as follows: group 1, repressed following siTAZ/YAP, siTEADs, or TGF␤ inhibition (therefore normally induced by the presence of these factors); group 2, repressed following siTAZ/YAP or siTEAD treatment but induced by TGF␤ inhibition; group 3, induced following siTAZ/YAP, siTEADs, or TGF␤ inhibition (therefore normally repressed by the presence of these factors); and group 4, induced by siTAZ/YAP and siTEADs but repressed by TGF␤ inhibition. The top five genes with altered expression in each group are listed in Fig. 3A. Quantitative PCR analysis confirmed the respective knockdown of TAZ/YAP and TEADs knockdown in each sample (Fig. 3B), as well as the expression changes observed from our microarray results for each group (Fig. 3, C-F). Notable genes for group 1 included the following: neuronal growth regulator 1 (NEGR1), urothelial cancer associated 1 (UCA1), and CTGF. Elevated expression of the group 1 genes NEGR1, UCA1, and CTGF relied on the presence of TAZ/YAP, TEADs, and active TGF␤ signaling (Fig. 3C), suggesting that TAZ/YAP⅐TEAD-TGF␤ synergize to promote the expression of these genes. In agreement with our observations, CTGF has recently been confirmed as an important transcriptional target of YAP⅐TEAD⅐SMAD complexes that promotes tumorigenesis in human malignant mesothelioma (31). NEGR1, UCA1, and CTGF expression was abolished following TAZ/YAP or TEAD knockdown in the absence of TGF␤ (Fig. 3C), suggesting that although specific TGF␤ signals rely on TAZ/YAP⅐TEAD, the basal level of TAZ/YAP⅐TEAD activity does not require TGF␤, and therefore TAZ/YAP⅐TEAD complexes may function dominantly to TGF␤ signals.

NEGR1 and UCA1 Are Direct Targets of TEADs and Are Necessary to Maintain Tumorigenic Breast Cancer Phenotypes-
Our analysis of LM2-4 cells indicates that TAZ/YAP, TEAD, and TGF␤ co-regulate the expression of a distinct subset of genes. To examine the importance of these genes in tumorigenesis, we focused our attention on group 1 genes, as these are synergistically induced by TAZ/YAP, TEAD, and TGF␤ and include CTGF, a defined mediator of TAZ/YAP-induced tumorigenesis and cancer stem cell-like phenotypes (2, 31). The top two genes synergistically induced by TAZ/YAP⅐TEAD and TGF␤ identified in our analysis were NEGR1 and UCA1. NEGR1 encodes a cell adhesion molecule that plays a role in neuronal growth and development (53)(54)(55)(56)(57)(58)(59). UCA1 encodes a long noncoding RNA that is expressed in development, is turned off in homeostatic tissues, and has been found to be highly expressed in bladder carcinomas (60). To determine whether these are direct transcriptional targets of TAZ/YAP, TEAD, and SMAD2/3, we performed chromatin immunoprecipitation (ChIP). Examination of the promoter regions of NEGR1, UCA1, and CTGF revealed consensus TEAD binding (61) and SMAD-binding motifs (62). ChIP of TAZ/YAP, TEAD, and SMAD2/3 from LM2-4 cell lysates revealed enrichment of these factors at the NEGR1, UCA1, and CTGF promoters, with SMAD2/3 recruitment only apparent after TGF␤ treatment (Fig. 4, A-C).
To further investigate the role of NEGR1 and UCA1 in TGF␤-mediated tumorigenesis, we examined the consequences of reducing their expression following siRNAmediated knockdown. Knockdown of NEGR1 or UCA1 repressed the migration of LM2-4 cells treated with TGF␤ in wound-healing scratch assays (Fig. 5A) and in transwell migration assays (Fig. 5B). Knockdown of either NEGR1 or UCA1 also suppressed the ability of LM2-4 cells to form large mammosphere colonies in the presence of TGF␤ (Fig. 5C), consistent with pro-tumorigenic roles for NEGR1 and UCA1. The results of these experiments reflect our observations with TGF␤ inhibition (SB-431542 treatment) or TAZ/YAP knockdown, suggesting that cooperative regulation of NEGR1 and UCA1 expression by TAZ/YAP⅐TEAD⅐SMAD complexes is necessary to promote tumorigenic phenotypes.

Nuclear TAZ and YAP Cooperate with TGF␤ to Promote Phenotypic and Transcriptional Changes in Nontumorigenic Cells-
Based on the results uncovered from our gene expression studies, we decided to test whether ectopic expression of nuclear TAZ/YAP in nontumorigenic human mammary MCF10A cells would lead to the induction of TGF␤-dependent transcriptional events similar to those we characterized in the malignant LM2-4 cells. Stable expression of nuclear TAZ or YAP mutants can transform epithelial cells (2,5,7,8), but this occurs following weeks of stable selection. Similarly, treatment of cells with TGF␤ for several days to weeks is required to observe tumorigenic events in mammary epithelial cells (30,63). To prevent confounding issues with long term culture conditions, we generated MCF10A cells that express a nuclear-localized and sta-  (siTEAD1-4), and then treated with TGF␤ or SB-431542 ϩ TGF␤. RNA from cell lysates was harvested, and global gene expression profiles were examined using Affymetrix microarrays. The Venn diagram highlights the number of genes with significant expression changes occurring for the indicated condition relative to the siCTL ϩ TGF␤ sample. Hierarchical clustering was performed on the significantly changing genes, which revealed four major clusters as indicated. Top significantly changing genes of interest are highlighted in each of the four clustered groups. B-F, LM2-4 cells were transiently transfected with siCTL, siTAZ, siYAP, siTAZ/YAP, or siTEADs and treated with or without TGF␤ or SB-431542 ϩ TGF␤. Relative expression of genes indicated in the microarray analysis was determined by qPCR. All data are shown as the average of three independent experiments ϩ S.E. B, confirmation of knockdown. C, group 1, genes repressed by siTAZ/YAP, siTEAD1-4, and SB-431542 treatment. D, group 2, genes repressed by siTAZ/YAP and siTEAD1-4 but induced by SB-431542 treatment. E, group 3, genes induced by siTAZ/YAP, siTEAD1-4, and SB-431542 treatment. F, group 4, genes induced by siTAZ/YAP and siTEAD1-4 but repressed by SB-431542 treatment.
ble TAZ mutant (TAZ(4SA)) (7) or YAP mutant (YAP(5SA)) (9) in a doxycycline-inducible manner. These TAZ/YAP mutants have the LATS kinase-induced phosphorylation sites substituted to alanines, preventing their cytoplasmic sequestration and proteasomal degradation (7,9). Titration of increasing amounts of doxycycline evoked subtle to high expression of TAZ(4SA) or YAP(5SA) in these cells (Fig. 6A). High levels of TAZ(4SA) or YAP(5SA) expression for short time frames (24 h) had minimal effects on the morphology of these cells (Fig. 6B). Short treatments of TGF␤ led to flattening of cells (Fig. 6B), a morphology indicative of cells undergoing cell cycle arrest, as has been described for MCF10A cells post-TGF␤ treatment (64). Strikingly, simultaneous doxycycline and TGF␤ treatment led to rapid cell morphology changes that differed from either condition alone, with the cells becoming more spindle-like and elongated (Fig. 6B). Furthermore, TAZ(4SA)-or YAP(5SA)expressing cells treated with TGF␤ displayed much more rapid cell migration in a wound-healing scratch assay, as compared with either condition alone (Fig. 6C), indicating that nuclear TAZ/YAP synergize with TGF␤ to promote cell morphology and cell migration changes.
In accordance with our expression analysis of LM2-4 cells, we found that nuclear TAZ or YAP function in concert with TGF␤ to control transcriptional events in MCF10A cells. For example, TAZ or YAP synergized with TGF␤ to promote the transcription of group 1 genes in an inducible fashion, including the expression of NEGR1, UCA1, and CTGF (Fig. 6D). Increased TAZ(4SA) or YAP(5SA) levels also induced the expression of group 2 genes (e.g. OCLN and CYFIP2), whereas TGF␤ repressed this group of genes (Fig. 6E). Conversely, group 4 genes, such as LBH and PMEPA1, were induced by TGF␤ but repressed in an inducible fashion by nuclear TAZ or YAP (Fig. 6F). Intriguingly, group 3 genes were undetectable in MCF10A cells, which may reflect the more differentiated state of these cells compared with LM2-4 cells. Together, our data indicate that the relationship between TAZ/YAP and TGF␤ is conserved in mammary-derived cells, and our observations support the idea that dysregulated TAZ/YAP and TGF␤ work in concert to control transcriptional events.
Nuclear TAZ and YAP Overcome TGF␤-induced Cytostasis in Nontumorigenic Cells-A hallmark trait of TGF␤ is its ability to suppress tumorigenesis in normal epithelium and early stage cancers, particularly through cell cycle inhibition. However, TGF␤ signals lose their ability to induce cytostasis in late stage cancers via poorly understood mechanisms (27,28). TGF␤-induced cell cycle arrest has been previously described in MCF10A cells (64), so we sought to explore the relationship between TGF␤, nuclear TAZ/YAP, and cell cycle progression. We performed proliferation assays using control MCF10A cells or cells with doxycycline-inducible nuclear TAZ(4SA) or YAP(5SA) expression. TGF␤-induced cytostasis was evident in control MCF10A cells (Fig. 7A). Strikingly, we found that expression of TAZ(4SA) or YAP(5SA) overcomes TGF␤ growth arrest, as cells treated simultaneously with doxycycline and TGF␤ proliferated similarly to control cells (Fig. 7A). To investigate whether the proliferative differences were due to cell cycle alterations, we used fluorescence-activated cell sorting analysis (FACS) to examine the DNA content of these cells. We found that TGF␤ treatment arrests cells in the G 1 phase of the cell cycle, and that TAZ(4SA) or YAP(5SA) expression reverses the G 1 phase arrest (Fig. 7, B and C). Our data therefore suggest that nuclear TAZ/YAP are responsible for the switch in TGF␤ activity from tumor-suppressive to tumorigenic in later stage breast cancers by converging to direct a distinct transcriptional program (see model in Fig. 8).

DISCUSSION
We have found TAZ/YAP to be necessary for transduction of TGF␤-induced tumorigenic phenotypes in metastatic breast cancer cells, such as clonal anchorage-independent growth, cell migration, and invasion. Interactions between endogenous TAZ/YAP, TEAD, and SMAD2/3 in the nucleus suggest that these complexes coordinate their activities at the transcriptional level. Through genome-wide expression analysis, we show that TAZ/YAP, TEAD, and TGF␤ regulate individual and common gene targets both positively and negatively, implying a complex level of transcriptional regulation and cross-talk between these factors. Of those gene targets we identified, many have yet to be characterized in breast cancer, and therefore our work may highlight previously unrecognized factors contributing to tumorigenesis. Of note, epithelial-mesenchymal transition-related genes were not enriched among the overlapping TAZ/YAP⅐TEAD-TGF␤-regulated subset, indicating that the TAZ/YAP⅐TEAD⅐SMAD2/3 complex drives aggressive behaviors of metastatic breast cancer cells downstream from the loss of epithelial cell polarity. Our transcriptional signature may thus reveal insight into the TAZ/YAPmediated tumorigenic program occurring in late-stage cancers, as MDA-MB-231 cells, and their LM2-4 derivatives possess mesenchymal properties. Indeed, the two genes that we characterized, NEGR1 and UCA1, proved to be necessary for the anchorage-independent growth and migratory properties of LM2-4 cells. TAZ/YAP and TGF␤ synergistically induce the expression of NEGR1 and UCA1 (group 1 genes), and given that TAZ/YAP, TEADs, and SMAD2/3 are enriched at the promot- ers of these genes, direct transcriptional synergy between TAZ/ YAP⅐TEAD⅐SMAD complexes likely promotes their expression in breast cancer.
Out of the 80 genes co-regulated by TAZ/YAP, TEAD, and TGF␤, 21 of them encode membrane proteins, several of which function as cell surface receptors, and 13 of them encode secreted proteins. The enrichment of such genes may reflect important non-cell autonomous alterations that are regulated by TAZ/YAP⅐TEAD and TGF␤ signals. Such signals are important for the pro-tumorigenic activity of TAZ and YAP (65,66), and thus we propose that cross-talk between TAZ/YAP⅐TEAD and TGF␤ signals demarcate a distinct local cellular environment that may promote a tumor-initiating niche. The well documented TAZ/YAP⅐TEAD target CTGF best highlights a secreted factor that is cooperatively induced by TGF␤. CTGF is a well established target of TGF␤-activated SMAD2/3 transcription factors (67) but is also an important driver of TAZ/ YAP-induced tumorigenic events (2,13). We observe that CTGF expression relies on the presence of TAZ/YAP, TEADs, and TGF␤ signaling, and nuclear TAZ or YAP mutants syner- gize with TGF␤ to strongly induce CTGF expression. Therefore, as in malignant mesotheliomas (31), the synergistic regulation of the CTGF promoter likely promotes aggressive breast cancer phenotypes.
We have additionally identified genes that are activated by both TAZ and YAP but repressed by TGF␤ signaling (group 2 genes) and, reciprocally, genes repressed by TAZ/YAP but induced by TGF␤ (group 4 genes). These groups of genes were somewhat surprising as they indicate that TAZ/YAP and TGF␤ direct opposing transcriptional events, and therefore suggest that a subset of TGF␤-activated SMAD activity does not rely on TAZ/YAP and vice versa. Based on the products encoded by several of these genes, we speculate that nuclear TAZ/YAP may override tumor-suppressive or negative feedback mechanisms initiated by TGF␤. For example, PMEPA1, which we found is induced by TGF␤ and inhibited by TAZ/YAP (group 4 gene), encodes a transmembrane protein that sequesters SMAD complexes in the cytoplasm (68). Thus, nuclear TAZ/YAP may function to overcome the induced expression of this gene to sustain pro-tumorigenic TGF␤ signals.
Historically, TAZ and YAP have been considered to be activators of gene transcription. However, our data indicate that TAZ/YAP play repressive roles as well (group 3 and 4 genes). We hypothesize TAZ/YAP⅐TEAD complexes execute this repressive function by various means. Recent work has shown that TAZ/YAP recruit the nucleosome remodeling and deacetylation (NuRD) complex to repress gene expression (35). Yorkie (Yki), the homolog of TAZ/YAP in Drosophila melanogaster, is also known to associate with chromatin-modifying proteins (69,70). Thus, TAZ/YAP⅐TEAD complexes likely function directly to inhibit transcription in breast cancers through similar recruitment of repressive factors to control local chromatin remodeling at promoters. However, TAZ/ YAP⅐TEAD complexes may also function in an indirect manner, particularly in conjunction with TGF␤ signaling, by binding, and re-localizing SMAD complexes (26,32). SMAD redistribution by TAZ/YAP may explain why nuclear TAZ or YAP affects the expression of certain target genes (group 2 and 4) more dramatically in MCF10A cells in the presence of TGF␤. Moreover, TAZ/YAP binding to SMADs is evident in the nucleus and in the cytoplasm (Fig. 2C), suggesting that interactions between these proteins in different localizations may direct distinct events.
Of interest, nuclear TAZ or YAP is capable of overcoming TGF␤-induced cytostasis (Fig. 7), which is a major mechanism by which TGF␤ functions as a tumor suppressor in early stage cancers (27). Consistent with this, we find that constitutively nuclear TAZ/YAP is evident in breast cancer cell lines where TGF␤ has lost its ability to induce cytostatic signals (Fig. 1A). TAZ/YAP drive the expression of cell cycle regulators (6), which may account for the ability of these factors to overcome cell cycle arrest. Indeed, our gene expression analysis in LM2-4 cells identified several cell cycle regulators as TAZ/YAP-regulated genes (e.g. CDKL1, CCNA1, CCNB1, and CCND3). However, given that TAZ/YAP bind SMAD complexes, we also speculate that TAZ/YAP may be capable of redirecting TGF␤-  We propose that increased nuclear TAZ/YAP, resulting from defects in upstream Hippo pathway signals, overcome TGF␤-mediated tumor suppressive functions (e.g. cytostasis) and concomitantly drive tumorigenic transcriptional events by promoting the activity of TEAD⅐SMAD complexes.
induced SMADs away from their cell cycle-repressive transcriptional roles toward those that promote tumorigenesis.
Our phenotypic and transcriptional analysis revealed redundant functions for TAZ and YAP. For example, TAZ and YAP have redundant roles in mediating TGF␤-induced mammosphere formation. Additionally, TAZ and YAP redundantly regulate the expression of group 1 genes NEGR1 and UCA1 (Fig. 3C). Interestingly, TAZ knockdown alone led to increases in UCA1 expression, which may reflect compensatory YAP hyperactivity in this context. A redundant role for these factors is further implied on account of similar effects resulting from nuclear TAZ or YAP mutant expression in MCF10A cells. Such redundancy is consistent with the overlapping roles of TAZ/ YAP in early development (71). However, we also present evidence for divergent transcriptional activity, based on specific gene expression reliance on either TAZ or YAP exclusively. For example, the expression of CTGF was repressed by TAZ or TAZ/YAP knockdown in LM2-4 cells but not by YAP knockdown alone (Fig. 3C). Thus, TAZ appears to have a dominant role in regulating CTGF expression in LM2-4 cells. Interestingly, recent work has revealed that YAP, in cooperation with TGF␤, has critical roles in controlling the expression of CTGF in malignant mesotheliomas (31). Thus, it appears that context defines dominance of TAZ or YAP.
Effective treatments of late-stage breast cancers are lacking, and our current understanding of the important signals driving and maintaining proliferation and metastasis is unclear. Our work has revealed critical intersections between TAZ/YAP, TEAD, and TGF␤ signaling in directing pro-tumorigenic phenotypes in breast cancer, and provides novel mechanisms by which the TGF␤ program may be directed toward aggressive tumorigenic phenotypes. Given the well documented roles of TGF␤ in late-stage cancers, recent efforts have been focused on optimizing new TGF␤ signaling inhibitors, which are currently in pre-clinical and clinical trials (72). Although advancement with such treatments is logical, our work suggests that enhanced efficacy may be achieved by treatment or co-treatment with current (73) or future TAZ/YAP⅐TEAD inhibitors.