Heat Shock Factor Hsf1 Cooperates with ErbB2 (Her2/Neu) Protein to Promote Mammary Tumorigenesis and Metastasis*

Background: The role of Hsf1 in mammary tumorigenesis and metastasis remains elusive. Results: Hsf1 deletion inhibits mammary tumorigenesis and metastasis by reducing ERK1/2 activity and epithelial-mesenchymal transition of mammary epithelial cells. Conclusion: Deletion of Hsf1 in mice carrying Her2/Neu significantly reduces breast cancer and metastasis. Significance: These findings indicate a powerful inhibitory effect conferred by Hsf1 in ErbB2-induced breast cancer. ErbB2/Neu oncogene is overexpressed in 25% of invasive/metastatic breast cancers. We have found that deletion of heat shock factor Hsf1 in mice overexpressing ErbB2/Neu significantly reduces mammary tumorigenesis and metastasis. Hsf1+/−ErbB2/Neu+ tumors exhibit reduced cellular proliferative and invasive properties associated with reduced activated ERK1/2 and reduced epithelial-mesenchymal transition (EMT). Hsf1+/+Neu+ mammary epithelial cells exposed to TGFβ show high levels of ERK1/2 activity and EMT; this is associated with reduced expression of E-cadherin and increased expression of Slug and vimentin, a mesenchymal marker. In contrast, Hsf1−/−Neu+ or Hsf1+/+Neu+ cells do not exhibit activated ERK1/2 and show reduced EMT in the presence of TGFβ. The ineffective activation of the RAS/RAF/MEK/ERK1/2 signaling pathway in cells with reduced levels of HSF1 is due to the low levels of HSP90 in complex with RAF1 that are required for RAF1 stability and maturation. These results indicate a powerful inhibitory effect conferred by HSF1 downstream target genes in the inhibition of ErbB2-induced breast cancers in the absence of the Hsf1 gene.

ErbB2/Neu oncogene is overexpressed in 25% of invasive/ metastatic breast cancers. We have found that deletion of heat shock factor Hsf1 in mice overexpressing ErbB2/Neu significantly reduces mammary tumorigenesis and metastasis. Hsf1 ؉/؊ ErbB2/Neu ؉ tumors exhibit reduced cellular proliferative and invasive properties associated with reduced activated ERK1/2 and reduced epithelial-mesenchymal transition (EMT). Hsf1 ؉/؉ Neu ؉ mammary epithelial cells exposed to TGF␤ show high levels of ERK1/2 activity and EMT; this is associated with reduced expression of E-cadherin and increased expression of Slug and vimentin, a mesenchymal marker. In contrast, Hsf1 ؊/؊ Neu ؉ or Hsf1 ؉/؉ Neu ؉ cells do not exhibit activated ERK1/2 and show reduced EMT in the presence of TGF␤. The ineffective activation of the RAS/RAF/MEK/ERK1/2 signaling pathway in cells with reduced levels of HSF1 is due to the low levels of HSP90 in complex with RAF1 that are required for RAF1 stability and maturation. These results indicate a powerful inhibitory effect conferred by HSF1 downstream target genes in the inhibition of ErbB2-induced breast cancers in the absence of the Hsf1 gene.
Amplification or overexpression of oncogenes such as ErbB2/Her2/Neu, a type I transmembrane receptor, occurs in 25% of breast cancer cases and promotes breast cancer invasion and metastasis (1)(2)(3). ErbB2/Her2/Neu is an orphan receptor, but it heterodimerizes with Her3 (ErbB3) and Her4 (ErbB4) family members and, following ligand binding, activates MAPK and PI3K/AKT signaling pathways, leading to enhanced cellular proliferation, increased motility, and reduced apoptosis and eventually cellular transformation and invasion (4,5). During tumorigenesis and metastasis, expression of epithelial cell markers such as E-cadherin, claudin, occludin, and keratins 8,9, and 18 are down-regulated, and expression of mesenchymal markers such as ␣SMA, 2 fibronectin, vitronectin, FSP1, and vimentin are up-regulated (6,7). These alterations include appearance of fibroblast-like morphology, leading to tumor cell invasion, metastasis, recurrence, and chemoresistance (7,8). This epithelial-mesenchymal transition (EMT) is regulated through activation of the transcriptional repressors Slug, Snail, Twist, ZFHX1b, deltaEF1, E12, and E47 or methylation of the E-cadherin promoter (9). Recent studies indicate that activation of ERK/MAPK signaling leads to outgrowth and translocation and proliferation of ErbB2-positive cells using MCF10A acini (10), and in the murine system, mammary epithelial cells undergo EMT when RAS is constitutively activated. Others have suggested that activation of the RAS/MAPK pathway induces EMT, cellular migration, and invasion as well as survival of epithelial cells (11). Data suggest that ErbB2 activation and overexpression alone may not be sufficient to cause tumor cell invasion and metastasis, and there likely are cooperating molecules (e.g. 14-3-3) (12) that drive mammary epithelial cells to acquire invasive properties.
In this study, we have analyzed ErbB2/Neu-induced mammary tumorigenesis in the absence of the Hsf1 gene. HSF1 is expressed in all tissues and is the main regulator of the heat shock response through control of downstream target genes such as heat shock proteins (HSPs) (13)(14)(15)(16). HSF1 is activated by a variety of environmental stressors and plays a critical function during development as well as during tumorigenesis (13,(17)(18)(19)(20)(21). The Hsf1 Ϫ/Ϫ mice are resistant to skin tumorigenesis induced by 7,1-dimethylbenz(a)anthracene and 12-O-tetradecanoylphorbol-13-acetate, suggesting that the absence of Hsf1 reduces RAS-driven tumor formation; however, the underlying mechanism has not been elucidated (21). The Hsf1 Ϫ/Ϫ mice in the background of TP53 deficiency exhibit a significant reduction in lymphomas (17), and in the background of the p53 mutant, mice exhibit deficiency in the development of sarcomas and carcinomas (21). Hsf1-deficient mice * This work was supported, in whole or in part, by National Institutes of Health are also resistant to diethylnitrosamine-induced liver tumorigenesis and high fat diet-induced obesity, and the underlying mechanism is associated with altered glucose and lipid metabolism (16). In breast cancer, HSF1 levels have been found to be elevated in 80% of breast carcinomas associated with high histologic grade, large tumor size, and increased mortality (22). HSF1 becomes transcriptionally activated following exposure of the cells to heregulin, leading to increased expression of Hsps (23,24). Hsf1 has also been found to interact with metastasisassociated protein 1 (MTA1), leading to changes in gene expression (25). Expression levels of Hsf1 target genes such as HSP90, HSP70, and HSP27 are increased in mammary tumors, leading to a decrease in cellular apoptosis (24). Overexpression of ErbB2/Neu has been shown to activate Hsf1 and up-regulate expression of lactate dehydrogenase-A, which leads to an increase in glycolysis. The study shows that glycolysis promoted by ErbB2/Neu occurs, at least in part, through HSF1 (26). HSF1 therefore plays a complex role in tumorigenesis and may impact the process of tumorigenesis through multiple pathways depending on the tumor origin.
To extend our observation regarding the impact of HSF1 on tumorigenesis in vivo, we have crossed Hsf1 Ϫ/Ϫ mice with MMTV-Her2/Neu ϩ transgenic mice that constitutively express the wild-type form of Her2/Neu in luminal epithelial cells (27). Our results indicate that Hsf1 Ϫ/Ϫ mice exhibit resistance to mammary tumorigenesis, and Hsf1 ϩ/Ϫ -Her2/Neu ϩ mice exhibit significant resistance to lung metastasis. Tumor and metastasis inhibitory effects exerted by complete or partial deletion of the Hsf1 gene appear to be by interference with the activation of ERK1/2 in the mammary tumor tissue and primary mammary epithelial cells, which leads to inhibition of tumor cell proliferation, reduced EMT, and cellular migration.
Kaplan-Meier Tumor-free Survival Curves-Wild-type, Hsf1 ϩ/Ϫ , and Hsf1 Ϫ/Ϫ mice expressing MMTV-Neu were palpated weekly. The date of visible appearance of the tumor was recorded, and mice were euthanized when tumors reached a size of 1-2 cm 3 . Tumors were fixed in 4% paraformaldehyde for histological analysis and frozen in OCT for immunohistochemical analysis or were snap-frozen for Western blot analyses. Tumor-free survival was estimated using Kaplan-Meier method. Statistical analysis was performed using the log-rank (Mantel-Cox) test. Significant differences were considered at p Ͻ 0.05.
Plasmids and Retrovirus Infection-shRNAs targeting sequence 1, GTGGACTCCAACCTGGATA, and sequence 2, gatcccGCTCATTCAGTTCCTGATCTTCAAGAGAGATC-AGGAACTGAATGAGCttttttggaag, of Hsf1 mRNA were synthesized and subcloned into the retroviral vector pLTHR (pLTHR-shRNA-Hsf1). Ha-Ras Val-12 was in the retroviral vector pWZL. For retroviral infection, plasmids were transfected into the packaging cell line, and after 48 h, the supernatant was mixed with 1 g/ml Polybrene for infection of the appropriate cell line. The virus-infected cells were drug-selected and pooled for analyses. Cells infected with empty vector were used for negative control.
Preparation of Primary Breast Epithelial Cell Cultures-For the primary mouse breast epithelial cell culture, mammary glands were dissected from 2-month-old virgin females and digested in DMEM containing 10% FCS plus 1 mg/ml collagenase D overnight. Cells were resuspended in the medium containing 10% FBS, 10 ng/ml EGF, and 5 g/ml bovine insulin. Cells were allowed to attach to the tissue culture dish three times to eliminate the contaminating fibroblasts. The epithelial cells in the supernatant were quantitated and seeded in culture dishes for analyses. MCF10A mammary epithelial cells were maintained in DMEM/F-12 media containing 5% horse serum, 10 ng/ml EGF, 5 g/ml insulin, and 50 g/ml hydrocortisone. To induce EMT, cells were treated with 5 ng/ml of human TGF␤1 (14-8348-62, eBioscience).
Preparation of Whole-mount Mammary Glands-Numbers four and nine mammary glands were extracted from virgin females and flattened and stretched out on a glass slide. The slides were air-dried for 5 min and fixed in 4% paraformaldehyde. Tissues were transferred to 100% acetone (three times for 1 h) to defat. Tissues were then transferred to 100% and then 90% EtOH (1 h each). Glands were stained in hematoxylin (3 h), rinsed and incubated in 50% EtOH (acidified with 25 ml HCl), and then transferred into 70, 95, and then 100% EtOH (1 h each) and xylene (twice for 1 h). Glands were stored in xylene.
BrdU Labeling-For in situ labeling of mammary epithelial cells, 8-week-old mice were injected with 100 g/g body weight BrdU three times within a period of 24 h. Mammary glands were fixed, processed, sectioned, and analyzed for number of cells labeled with BrdU. For cultured cells, coverslips containing cells were labeled for 2 h with 10 mM BrdU, and following immunostaining, the number of BrdU-positive cells was quantified.
Transwell Migration Assays-Cell migration was performed as described previously (28). Briefly, the surface under the transwell (8 m pore size; Costar) was coated with 10 g/ml collagen I for 16 h at 4 ºC. Mammary epithelial cells cultured from 2-month-old mice were dispersed using trypsin and then resuspended in serum-free medium at a density of 2 ϫ 10 5 cells/ml. One hundred microliters of cell suspension was added to the upper chamber of the transwell. Cells were allowed to migrate for 48 h in presence or absence of 5 ng/ml TGF␤1. After this time, cells remaining in the upper chamber were removed, and the cells on the undersurface of the transwells were fixed and stained with crystal violet. The number of cells that had migrated to the undersurface was quantified using a phase-contrast microscope.
Immunoblotting-Immunoblotting experiments were performed according to standard protocols (29). Cells were lysed in lysis buffer, and 30 g of protein from each sample was resolved by SDS-PAGE and processed for immunoblotting analyses.
Histology and Immunohistochemical and Immunofluorescence Staining-For histology, tissues were fixed in 4% paraformaldehyde and embedded in paraffin, and tissue sections were stained with H&E and analyzed using the Elston and Ellis technique (30) for the histological grading system. Tumors were analyzed according to tube formation, nuclear pleomorphism, and number of mitotic cells (30). Analyses were performed to estimate whether tumors were well, moderately, or poorly differentiated. For immunocytochemistry, cells were cultured on glass coverslips, fixed in 4% paraformaldehyde, permeabilized (0.1% Triton X-100 in PBS), and blocked (5% milk in PBS) for 30 min at 25 ºC. Fixed cells were incubated with primary antibodies overnight at 4 ºC, rinsed, and incubated for 1 h with fluorescent-conjugated secondary antibody (Molecular Probes). Cells were further rinsed and mounted with Vectashield containing DAPI (Vector Laboratories) and analyzed using a confocal or Zeiss fluorescence microscope. For immunohistochemistry, 7-m tissue sections were deparaffinized in xylene and rehydrated in a series of alcohol/water mixtures. Antigen retrieval was performed by placing tissue sections in 10 mM sodium citrate, pH 6.0, and steamed for 30 min. After rinsing in PBS, tissue sections were blocked (3% BSA in PBS for 2 h at 4 ºC) and then incubated in primary antibody (in 3% BSA in PBS plus 1% Tween 20) overnight at 4 ºC. Antibody/antigen was detected using Cy3 fluorescent-conjugated anti-mouse (or rabbit) IgG secondary antibody. Nuclei were stained with DAPI, and sections were processed, mounted, and analyzed as above (31,32).
Statistical Analyses-All experiments were performed with at least 3-40 mice. Data are presented as means Ϯ S.D. Statistical significance between experimental groups was assessed using Student's t test and p Ͻ 0.05 was considered significant.

Hsf1 Promotes Development of Her2/Neu-induced Mammary
Tumors and Lung Metastasis-To determine the function of Hsf1 in mammary tumorigenesis, we selected the MMTV-Her2/Neu ϩ transgenic mouse model crossed with Hsf1 Ϫ/Ϫ mice (27). In this model, Her2/Neu expression in mammary epithelial cells induces mammary tumors and lung metastasis in high percentages of mice. For Her2/Neu ϩ mice (thereafter named Neu ϩ ), a cohort of virgin female mice expressing Neu in the presence or absence of the Hsf1 gene was palpated weekly to detect the presence of breast tumors. In contrast to Hsf1 Ϫ/Ϫ Neu ϩ mice where development of mammary tumors was rare (7%), high percentages of Hsf1 ϩ/ϩ Neu ϩ mice developed tumors (74%) (Fig. 1A). At 499 days post-observation, 22% of the Hsf1 ϩ/ϩ Neu ϩ mice developed tumors, whereas none of the Hsf1 ϩ/Ϫ Neu ϩ mice did. At 599 days, 14% of the Hsf1 ϩ/ϩ Neu ϩ mice and 16% of Hsf1 ϩ/Ϫ Neu ϩ mice developed tumors. These data suggest a period of delay in the onset of tumorigenesis in the absence of one copy of the Hsf1 gene. There were, however, no significant overall differences between these two groups in terms of tumor-free survival. During the period of observation, only one Hsf1 Ϫ/Ϫ Neu ϩ mouse developed a small mammary tumor and that occurred at 707 days.
As mentioned before, tumors arising in Neu ϩ mice metastasize to the lungs with about 70% frequency (27). To determine whether there were differences in metastasis ability between tumors that developed in Hsf1 ϩ/ϩ Neu ϩ or Hsf1 ϩ/Ϫ Neu ϩ mice, lungs were collected from tumor-bearing mice and examined for the presence of tumor nodules. Histological analyses indicate that 26% of Hsf1 ϩ/ϩ Neu ϩ mice exhibited lung metastasis, in contrast to 0% of the Hsf1 ϩ/Ϫ Neu ϩ mice (Fig. 1B).
Interestingly, there were significant differences in multiple parameters in the tumors that developed in Hsf1 ϩ/Ϫ Neu ϩ mice compared with Hsf1 ϩ/ϩ Neu ϩ mice. For example, tumors in the Hsf1 ϩ/Ϫ Neu ϩ mice exhibited a 40% reduction in Ki-67 immunostaining compared with tumors in the Hsf1 ϩ/ϩ Neu ϩ mice, indicating a lower proliferating capacity of tumor cells in this genetic background (Fig. 1C). Furthermore, Hsf1 ϩ/Ϫ Neu ϩ tumors exhibited a significant increase in TUNEL-positive cells indicating cells were undergoing apoptosis (Fig. 1D). The tumors arising in Hsf1 Ϫ/Ϫ Neu ϩ and Hsf1 ϩ/Ϫ Neu ϩ mice also exhibited a significant reduction in vascularization compared with tumors arising in Hsf1 ϩ/ϩ Neu ϩ mice. This was confirmed by CD31 immunostaining (Fig. 1E).
Activation of ERK1/2 through ErbB2 and RAS has been shown to induce EMT, which plays an important role during development and in tumorigenesis, resulting in tumor cell spreading and increased cellular migration and invasiveness to the surrounding tissue (11,34). As we observed significantly less p-ERK1/2 and a lower number of tumors that metastasized to the lungs in Hsf1 ϩ/Ϫ Neu ϩ compared with Hsf1 ϩ/ϩ Neu ϩ tumors, we examined the expression of epithelial and mesenchymal markers in tumors that evolved in Hsf1 ϩ/ϩ Neu ϩ and Hsf1 ϩ/Ϫ Neu ϩ mice. IHC staining of epithelial and mesenchymal markers showed lower expression of E-cadherin in Hsf1 ϩ/ϩ Neu ϩ tumors, whereas expression of E-cadherin remained elevated in Hsf1 ϩ/Ϫ Neu ϩ tumors (Fig. 2C). Loss of ␤-catenin, which destabilizes the epithelial cell-cell junction, was also examined, and Hsf1 ϩ/ϩ Neu ϩ tumors exhibited significant loss of ␤-catenin expression compared with Hsf1 ϩ/Ϫ Neu ϩ tumors (Fig. 2C). Loss of epithelial cell markers correlated with increased expression of mesenchymal markers vimentin and ␣SMA in Hsf1 ϩ/ϩ Neu ϩ tumors as expected, whereas the expression of these mesenchymal markers was sig-nificantly lower in Hsf1 ϩ/Ϫ Neu ϩ tumors. IHC data were confirmed using immunoblotting (Fig. 2, D and E). The data indicated a reduction in the expression of the Slug transcription factor in Hsf1 ϩ/Ϫ Neu ϩ tumors as a potential regulator of E-cadherin expression, leading to a reduction in EMT.

JOURNAL OF BIOLOGICAL CHEMISTRY 35649
There was no significant reduction in the levels of CDC37, HOP, HSP40, or p23, which are known to be Hsp90a cochaperones ( Fig. 2D and data not presented).
Taken together, these data indicate that absence of the Hsf1 gene inhibits mammary tumorigenesis. Furthermore, a reduction in HSF1 expression level in Hsf1 ϩ/Ϫ Neu ϩ mice reduces lung metastasis and is associated with reduced cellular proliferation and a reduction in p-ERK1/2 and EMT.
Hsf1 Promotes Mammary Gland Morphogenesis-To determine the underlying mechanism for the significant reduction in tumor development in Hsf1 Ϫ/Ϫ Neu ϩ mice, we analyzed mammary glands of adult Hsf1 ϩ/ϩ , Hsf1 ϩ/Ϫ , and Hsf1 Ϫ/Ϫ mice in the presence or absence of the Neu transgene. Maturation of the mammary gland occurs mainly postnatally (36). Cellular differentiation and migration are limited to the cap layer at the terminal end buds that migrate to the fat pad and stroma. Terminal end bud growth is regulated by a variety of growth factors, leading to ordered ductal development (36). Therefore, mammary gland numbers 4 and 9 of virgin female mice were examined for ductal morphogenesis. At 8 weeks of age, the ductal tree of Hsf1 ϩ/ϩ mice filled the entire fat pad with substantial alveolar development (Fig. 3A) (36). The ductal tree of Hsf1 ϩ/Ϫ but specifically Hsf1 Ϫ/Ϫ mice appeared to possess fewer lateral side branches. Interestingly, expression of Neu ϩ did not increase the number of ductal branches observed in Hsf1 Ϫ/Ϫ mice, but it did increase slightly the number of ductal branches in Hsf1 ϩ/ϩ Neu ϩ and Hsf1 ϩ/Ϫ Neu ϩ mice (Fig. 3B). Quantification of lateral side branches of the indicated genotypes is presented in Fig. 3C.
Mammary glands from all genotypes were analyzed for histological changes (Fig. 3D). A decrease in the number of ductal side branches is evident following histological analyses of the cross-sections of the mammary glands in mice deficient in Hsf1, but this is less evident in Hsf1 ϩ/Ϫ mice. To investigate if the expressions of ␣SMA, a myoepithelial marker, and keratin 18, a luminal epithelial marker, were comparable between the genotypes, IHC staining of cross-sections of the mammary glands was performed (Fig. 3, E and F). The results show that both myoepithelial cells and luminal epithelial cells were present, and the relative expression levels of ␣SMA and keratin 18 were comparable between the genotypes.
Hsf1 Protects Mammary Epithelial Cells from Apoptosis and Increases Their Proliferative Capacity-Hsf1 and its downstream target genes protect cells against apoptotic cell death and reverse the senescence phenotype (15,(37)(38)(39). To determine whether the reduction in ductal branching observed in Hsf1 Ϫ/Ϫ mice correlated with increased epithelial cell apoptosis, we performed TUNEL assays on mammary gland tissue sections prepared from Hsf1 ϩ/ϩ , Hsf1 ϩ/Ϫ , and Hsf1 Ϫ/Ϫ mice in the presence or absence of the Neu ϩ transgene. Data indicate that the number of apoptotic cells is significantly increased in the absence of the Hsf1 gene (Fig. 4, A and B). The number of The increase in apoptosis in mammary epithelial cells in Hsf1 Ϫ/Ϫ mice also coincided with lower numbers of proliferating cells that were detected following in situ BrdU labeling of ductal epithelial cells (Fig. 4, C and D). The reduction in cellular proliferation using BrdU labeling also was present in cultured primary mammary epithelial cells prepared from Hsf1 ϩ/ϩ , Hsf1 ϩ/Ϫ or Hsf1 Ϫ/Ϫ mice with or without expression of the Neu transgene (Fig. 4, E and F). Taken together, these data indicate that ductal epithelial cells undergo increased apoptotic cell death and reduced cellular proliferation in the absence of one or both copies of the Hsf1 gene.
HSF1 Promotes EMT and Increases Epithelial Cell Migration in Response to TGF␤-TGF␤ cooperates with signaling pathways such as RAS and WNT to induce EMT; this is a critical process during tumorigenesis (34). To further examine whether the EMT program is affected in Hsf1 Ϫ/Ϫ mammary epithelial cells and whether this affects cellular migration, Neu ϩ primary cultured epithelial cells from Hsf1 ϩ/ϩ Neu ϩ , Hsf1 ϩ/Ϫ Neu ϩ , and Hsf1 Ϫ/Ϫ Neu ϩ cells were exposed to TGF␤ and immunostained for the presence of E-cadherin, ␤-catenin, and the mesenchymal marker ␣SMA. The expression levels of E-cadherin and ␤-catenin are reduced in response to TGF␤ in Hsf1 ϩ/ϩ Neu ϩ cells as expected (Fig. 5, A and B). This was associated with an increase in the expression of ␣SMA 24 -48 h following exposure of cells to TGF␤. However, reduced levels of HSF1 led to a reduction in E-cadherin and ␤-catenin degradation as they remain elevated in Hsf1 ϩ/Ϫ and Hsf1 Ϫ/Ϫ cells compared with Hsf1 ϩ/ϩ cells (Fig. 5A). The level of mesenchymal marker ␣SMA is increased during the 48-h incubation in the presence of TGF␤ in Hsf1 ϩ/ϩ , but this was much less pronounced in

Hsf1 Modulation of Mammary Tumorigenesis
Hsf1 ϩ/Ϫ Neu ϩ and Hsf1 Ϫ/Ϫ Neu ϩ cells. This also was associated with an increase in fibroblast cell morphology in Hsf1 ϩ/ϩ Neu ϩ cells. Immunoblot analyses confirmed activation of TGF␤ signaling through activation of p-Smad 3, increase in Slug, reduction in E-cadherin (2-5-fold at 24 -48 h) and ␤-catenin (5-fold at 24 -48 h) levels, and an increase in the mesenchymal markers N-cadherin (1.5-fold at 24 -48 h) and vimentin (3-and 2-fold at 24-48 h) in Hsf1 ϩ/ϩ Neu ϩ cells but not in Hsf1 ϩ/Ϫ Neu ϩ cells. We found that exposure of cells to TGF␤ increased the activities of ERK1/2 and AKT (Ser(P)-473), which are downstream of TGF␤ signaling and are known inducers of EMT. The levels of the phosphorylated forms of ERK1/2 and AKT were also reduced by 3-5-fold in response to TGF␤ in Hsf1 ϩ/Ϫ Neu ϩ cells compared with Hsf1 ϩ/ϩ Neu ϩ cells, which increased by 3-5fold relative to untreated control cells (Fig. 5D). The expression of total ERK1/2 and AKT was not significantly different between the genotypes.
EMT is critical during the progression of carcinomas to an invasive state (34). We therefore used a transwell migration assay to examine whether the reduced EMT observed in Hsf1 ϩ/Ϫ and Hsf1 Ϫ/Ϫ cells reduces the migratory ability of Neu ϩ primary epithelial cells harboring reduced HSF1 compared with Hsf1 ϩ/ϩ Neu ϩ cells. The data presented in Fig. 5E indicate that a significant reduction in the migratory ability of Neu ϩ primary mammary epithelial cells that are deficient in or have lower levels of HSF1. The reduced cellular migration of cells with lower levels of HSF1 and expressing Neu was evident in the presence or absence of TGF␤. Treatment of cells with TGF␤ increased the ability of Hsf1 ϩ/ϩ Neu ϩ cells to migrate, but it was less effective when Hsf1 level was reduced (Fig. 5E,  lower panel).
The results obtained using mouse mammary tumors expressing lower HSF1 levels or using primary epithelial cells prepared from Hsf1 ϩ/Ϫ or Hsf1 Ϫ/Ϫ mammary glands suggest FIGURE 5. Reduction in HSF1 levels inhibits EMT and cellular migration. A-C, mammary epithelial cells were cultured from number 4 mammary glands of 2-month-old virgin female mice. Cells were exposed to TGF␤ for 0, 24, or 48 h. Cells were then fixed and stained with antibody to E-cadherin (green), ␤-catenin (red), or ␣SMA (green). Propidium iodide (red) (A) and DAPI (blue) (B and C) show nuclear staining. Bar, 20 m. D, immunoblot analyses of mammary epithelial cells prepared from Hsf1 ϩ/ϩ Neu ϩ or Hsf1 ϩ/Ϫ Neu ϩ mice (as in A) showing the levels of the indicated proteins following exposure to TGF␤. ␤-Actin is loading control. E, transwell assay. Mammary epithelial cells were cultured from number 4 mammary glands of 2-month-old virgin female mice. Cells were placed in the transwell as described under "Experimental Procedures" and allowed to migrate to the undersurface. Cells were then fixed, stained, and photographed. Quantification of the cells migrated in the undersurface is presented in the lower panel. Bars are mean Ϯ S.D. Experiments were repeated three times. Statistical significance is as follows: **, p Ͻ 0.01; ***, p Ͻ 0.001. OCTOBER 12, 2012 • VOLUME 287 • NUMBER 42 that they respond poorly to activate ERK1/2 even in the presence of TGF␤ stimulation compared with wild-type cells. To recapitulate some of the effects observed in mouse mammary epithelial cells in the absence of the Hsf1 gene, we knocked down HSF1 in MCF10A cells using shRNA (sequence 1, see "Experimental Procedures") (Fig. 6A). We then determined whether MCF10A cells can activate ERK1/2 using constitutively active Ha-Ras Val-12 as an efficient inducer of ERK1/2 signaling pathway downstream of ErbB2, and TGF␤ receptors when Hsf1 levels are reduced (33). Immunoblotting experiments show that MCF10A cells expressing Hsf1 shRNA and c-Ha-Ras Val-12 exhibit significantly lower levels of p-ERK1/2 (10-fold reduction compared with MCF10A cells expressing scrambled shRNA). We also observed a 2-fold reduction in the levels of p-RAF1 and p-MEK in cells expressing Hsf1 shRNA that also expressed Ha-Ras Val-12 , compared with cells that expressed Ha-Ras Val-12 alone. These data indicate that lowering the Hsf1 levels leads to alterations in MCF10A cells, reducing their ability to fully activate ERK1/2. No reductions in the total levels of RAF1, MEK, ERK1/2, AKT, or p-AKT were observed by immunoblotting (Fig. 6B and data not shown). We also tested another Hsf1 shRNA (sequence 2, see "Experimental Procedures") to ensure that P-ERK1/2 was reduced in Ha-RASexpressing cells when Hsf1 level was reduced (Fig. 6C).

Hsf1 Modulation of Mammary Tumorigenesis
Previous studies showed that the maturation of RAF1 is regulated by HSP90 and its co-chaperones (35,40). HSF1 regulates the expression of HSP90a and HSP70, as well as their cochaperones, which contain heat shock elements in their promoters (13,14). MCF10A cells expressing Hsf1 shRNA showed 20 -40% reduction in the levels of Hsp90a cochaperones CDC37, p23, and HOP (Fig. 6D). To determine whether HSP90a-RAF1 complexes were affected in cells expressing Hsf1 shRNA, we ectopically expressed myc-Raf1 together with Hsf1 shRNA with or without Ha-Ras Val-12 . Myc-Raf1 was immunoprecipitated, and immunoblotting was performed to detect HSP90a in the complex. Fig. 6E shows that the amount of RAF1 binding to HSP90a was significantly reduced in cells expressing Hsf1 shRNA compared with cells that expressed scrambled shRNA. These data indicate a mechanism by which ERK1/2 activity may be reduced in Hsf1 shRNA-containing cells. This suggests that a small reduction in the levels of HSPs and their cochaperones affects complexes of HSP90a with their client proteins more significantly than their total expression in cells.
As noted before, the process of tumorigenesis involves increased cellular migration (34). To examine whether knockdown of Hsf1 leads to changes in MCF10A cell migration, we performed wound healing and transwell assays. Although cellular migration was comparable between mock-treated cells and cells expressing Hsf1 shRNA, cells expressing c-Ha-Ras Val-12 were able to migrate significantly more distances when they expressed wild-type levels of HSF1 compared with cells expressing Hsf1 shRNA (Fig. 6, F and G). We also evaluated the ability of Hsf1 shRNA-containing cells to migrate using a transwell assay. Fig. 6G shows that MCF10A cells containing Hsf1 shRNA showed a significant reduction in migration ability compared with mock-transduced cells using the transwell assay. Taken together, the above results indicate that Hsf1 deficiency leads to reduced p-ERK1/2 activity, inefficient EMT, and reduced cellular migration in response to TGF␤ in mammary epithelial cells.

DISCUSSION
We sought to determine the role of Hsf1 in breast cancer in a cohort of mice deficient in the Hsf1 gene that overexpressed the Neu oncogene in luminal epithelial cells of the mammary gland. The results show that Hsf1 Ϫ/Ϫ Neu ϩ mice are resistant in developing breast cancer, whereas Hsf1 ϩ/Ϫ Neu ϩ mice exhibit some delay in the onset of breast cancer development. However, the overall tumor-free survival was not significantly delayed when compared with the Hsf1 ϩ/ϩ Neu ϩ mice. Interestingly, however, Hsf1 ϩ/Ϫ Neu ϩ mice exhibit a significant reduction in lung metastasis compared with Hsf1 ϩ/ϩ Neu ϩ mice. To elucidate the mechanism underlying the inhibition of tumorigenesis and metastasis observed in Hsf1 Ϫ/Ϫ or Hsf1 ϩ/Ϫ Neu ϩ mice, respectively, we analyzed multiple parameters such as proliferative capacity, apoptotic rate, and EMT in the tumors and in mammary epithelial cells expressing reduced levels of the Hsf1 gene. We found alterations of multiple parameters that could reduce tumor growth and metastasis in the absence of the Hsf1 gene. We observed a significant reduction in proliferative capacity of cells deficient in HSF1 following both in situ as well as in vitro BrdU labeling of cells, and a significant increase in the in situ detection of apoptotic cell death of mammary epithelial cells in the absence of the Hsf1 gene. Other alterations that appear to significantly inhibit tumor growth and metastasis is the inability of Hsf1 ϩ/Ϫ and Hsf1 Ϫ/Ϫ cells to effectively show ERK1/2 activity. ERK1/2 are downstream of ErbB2 receptor signaling, and ErbB2-positive mammary epithelial cells have been shown to require activity of MEK/ERK1/2, and not AKT, to drive the translocation of cells to the ductal lumen and their subsequent proliferation (10). Other studies have shown that expression of the active form of RAF in MCF10A acini leads to an increase in cell movement (41). ErbB2-driven AKT activation has been shown to activate NF-B, suppressing apoptosis, and to activate mammalian target of rapamycin, leading to an increase in cellular proliferation (33). Both activation of ERK1/2 and AKT led to an increase in the expression of a number of genes, including VegfA and CCNDI (cyclin D1), that led to an increase in angiogenesis and cellular proliferation (33). We therefore anticipate that lower levels of p-ERK1/2 and p-AKT lead to increased apoptotic cell death (Fig. 1D) and a decrease in cellular proliferation (Fig. 1C) and in angiogenesis (Fig. 1E), leading to reduced tumor cell proliferation and metastasis when the level of HSF1 is reduced in Hsf1 ϩ/Ϫ Neu ϩ mice. There is increasing evidence that ERK1/2 (and more recently ERK2 (42)) induce EMT in mammary epithelial cells (11,34). Therefore, reduction in ERK1/2 activity could lead to the reduction in EMT that we observe in tumors formed in Hsf1 ϩ/Ϫ Neu ϩ mice compared with Hsf1 ϩ/ϩ Neu ϩ mice. A lower level of EMT also was observed in an in vitro response to TGF␤ stimulation, which is a strong inducer of EMT. The results showed reduced activated p-ERK1/2 in Hsf1 ϩ/Ϫ Neu ϩ cells compared with Hsf1 ϩ/ϩ Neu ϩ cells, and a reduction in EMT was associated with reduced levels of Slug transcriptional repressor in Hsf1 ϩ/Ϫ Neu ϩ cells, which leads to increased expression of E-cadherin and reduced expression of N-cadherin, a mesenchymal marker, as expected (11,34). Downstream effects of reduction in EMT led to reduced cellular migration in Hsf1 ϩ/Ϫ Neu ϩ cells compared with Hsf1 ϩ/ϩ Neu ϩ as expected.
In general, Hsf1-deficient cells are specifically deficient in inducing a heat shock response when exposed to elevated temperature or other forms of stressors (13). However, in different cell types, and dependent on whether cells are primary or express different oncogenes, Hsf1-deficient cells exhibit reduced levels of different HSPs (13,16,43,44). Hsf1 is known to drive the expression of HSP110, HSP90a, HSP70 (inducible), and HSP40, as well as other heat shock element-containing genes such as p23, CDC37, and HOP, which are cochaperones for HSP90. The HSP90 complexes have been shown to be involved in ErbB2, RAF, and AKT maturation, and they are degraded when cells are exposed to HSP90 inhibitors (35,40,44). In addition, tumorigenesis process is known to increase the levels of molecular chaperones, and it places a higher demand on the function of molecular chaperones (45,46). As noted before, microarray and immunoblotting analyses of Neu ϩ tumors, Neu ϩ primary mammary epithelial cells expressing lower levels of HSF1, or human mammary epithelial MCF10A cells expressing Hsf1 shRNA did not show significantly lower levels of all the components of the HSP90 machinery. However, we noted lower levels of HSP70 and HSP90a in Hsf1 ϩ/Ϫ Neu ϩ tumors by 20 -40% and no significant reduction in HOP, CDC37, or p23. In MCF10A cells expressing Hsf1 shRNA, we observed 20 -40% reduction in the level of HSP90 cochaperones CDC37, HOP, and p23. Interestingly, however, immunoprecipitation experiments show a significant reduction in RAF1 association with HSP90a in cells expressing Hsf1 shRNA if there is some perturbation in the expression of HSPs or their cochaperones. We therefore anticipate that lower HSF1 levels Hsf1 Modulation of Mammary Tumorigenesis OCTOBER 12, 2012 • VOLUME 287 • NUMBER 42 leads to a base-line reduction in multiple chaperones, causing improper formation of HSP90 with its client proteins, such as RAF1, leading to inefficient activation of RAS/RAF/MEK/ ERK1/2 signaling. Although AKT is also a client protein of Hsp90, a comparable reduction in pAKT levels was not observed in MCF10A. We therefore envision where reduction in HSF1 levels could impact ErbB2 and TGF␤ signaling pathways, affecting tumor growth. Inhibition of HSP90 leads to degradation of its many client proteins such as RAF1, ErbB2 (and other receptor tyrosine kinases), and AKT as has been reported previously (35). Hsf1 deficiency reduces the levels of HSP90a and other chaperones such as HSP70 and their client proteins. Although the client proteins may not be degraded entirely, there is a reduction in their levels (such as RAF1 presented here) in chaperone complexes. This leads to reduced signaling through receptor tyrosine kinases and reduced tumor growth or metastasis. Recent studies show that inhibition of HSP90 using geldanamycin reduced levels of TGF␤ receptor and signaling leading to reduction in TGF␤-induced p-Smad3 (47) and EMT (48). Although in both studies induction of Hsf1-induced HSP70 was associated with this effect and this would not be relevant in our studies where Hsf1 has been deleted, nevertheless, a reduction in the levels of HSP90 and/or its cochaperones as presented in our studies could potentially reduce TGF␤-induced signaling through the TGF␤ receptor leading to a reduction in EMT. Significant reduction in mammary tumorigenesis and metastasis in the absence of the Hsf1 gene in Hsf1 Ϫ/Ϫ Neu ϩ or Hsf1 ϩ/Ϫ Neu ϩ mice could be due to the reduced level of expression and signaling through the ErbB2 receptor. Previous reports indicate that overexpression of Her2 in MCF10A cells carrying Hsf1 shRNA leads to reduced ability of cells to grow on soft agar or in nude mice (49). The mechanism proposed was higher levels of p21 and down-regulation of HSP70 and HSP27. In addition, using the constitutively active form of Her2, it has recently been shown that Hsf1 Ϫ/Ϫ mice exhibit resistance to tumorigenesis (50). The underlying mechanism is thought to be suppression of angiogenesis through Hsf1 regulation of the HuR-HIF1 pathway. The above data are consistent with our results suggesting that Hsf1 deficiency leads to reduced levels of HSPs, leading to inhibition of critical signaling molecules that are known clients of Hsf1 downstream targets leading to suppression of tumorigenesis.