Overexpression of HER2 (erbB2) in Human Breast Epithelial Cells Unmasks Transforming Growth Factor (cid:1) -induced Cell Motility* □ S

We have examined overexpression of the human epidermal growth factor receptor 2 (HER2) to determine if it modifies the anti-proliferative effect of transforming growth factor (TGF)- (cid:1) against MCF-10A human mammary epithelial cells. Exogenous TGF- (cid:1) inhibited cell proliferation and induced Smad-dependent transcriptional reporter activity in both MCF-10A/HER2 and MCF-10A/vec-tor control cells. Ligand-induced reporter activity was 7-fold higher in HER2-overexpressing cells. In wound closure and transwell assays, TGF- (cid:1) induced motility of HER2-transduced, but not control cells. The HER2-block-ing antibody trastuzumab (Herceptin) prevented TGF- (cid:1) induced cell motility. Expression of a constitutively active TGF- (cid:1) type I receptor (ALK5 T204D ) induced motility of MCF-10A/HER2 but not MCF-10A/vector cells. TGF- (cid:1) -in-duced motility was blocked by coincubation with either the phosphatidylinositol 3-kinase inhibitor LY294002, the mitogen-activated protein kinase (MAPK) inhibitor U0126, the p38 MAPK inhibitor

The transforming growth factors (TGF-␤) 1 are polypeptides that are part of the superfamily of structurally related ligands that include the TGF-␤s, activins, and bone morphogenetic proteins. TGF-␤ ligands can modulate cell proliferation, lineage determination, functional differentiation, extracellular matrix production, cell motility, and apoptosis (1). TGF-␤ was originally reported to induce anchorage-independent growth of mouse fibroblasts (2). Later studies indicated that TGF-␤ is a potent inhibitor of epithelial cell proliferation (3,4), suggesting a role in tumor suppression. Transgenic mice expressing active TGF-␤1 S223/225 in the mammary gland exhibit delayed ductal morphogenesis, renewal capacity, and functional differentiation (5)(6)(7). Inhibition of autocrine TGF-␤ signaling in the mouse mammary gland by tissue-specific expression of a truncated dominant-negative TGF-␤ type II receptor (T␤RII) results in accelerated lobulo-alveolar development (8) and enhanced propensity for the development of carcinogen-induced lung, mammary, and skin tumors (9, 10) as well as spontaneous invasive mammary carcinomas (11). Mice with targeted disruption of either the Tgfb1 or the Smad3 genes develop colon adenomas and carcinomas (12,13). Mice heterozygous for deletion of the Tgfb1 gene express 10 -30% of the TGF-␤1 protein, and this partial loss is sufficient to attenuate epithelial growth restraint and inhibit tumor suppression function, resulting in enhanced carcinogen-induced liver and lung tumors (14). MMTV/TGF-␤1 mice are resistant to (7,12-dimethylbenz[a]-anthracene)-induced mammary tumors, and when cross-bred with mice expressing the EGF receptor ligand TGF-␣ in the mammary gland (MMTV/TGF-␣), the bigenic (TGF-␤/TGF-␣) progeny fails to develop TGF-␣-induced mammary hyperplasias and carcinomas while still exhibiting a hypoplastic phenotype (15). Finally, mice overexpressing active TGF-␤1 under the control of the K14 keratinocyte-specific promoter are relatively protected from skin tumors induced by chemical carcinogens (16), further supporting the tumor suppressive role of TGF-␤ signaling.
Most cancer cells lose or attenuate TGF-␤-mediated antiproliferative effects either by mutational inactivation of TGF-␤ receptors or their signal transducers or by less known mechanisms (for review, see Refs. [17][18][19]. There is also abundant experimental evidence to support that excess production and/or activation of TGF-␤ in tumors, including breast cancers, can foster cancer progression by autocrine and paracrine mechanisms (for review, see Ref. 20). These include enhancement of tumor cell motility and survival, increase in tumor neo-angiogenesis and extracellular matrix production, up-regulation of peritumor metalloproteases, and inhibition of host immune surveillance mechanisms (17,21). For example, overexpression of active TGF-␤1 or an activated type I TGF-␤ receptor (T␤RI) in the mammary gland of transgenic mice results in accelera-tion of metastases derived from neu-induced primary mammary tumors (22,23). Furthermore, in transgenic mice expressing the Polyomavirus middle T antigen (PyVmT) in mammary epithelium, blockade of TGF-␤ with a soluble fusion protein consisting of the extracellular domain of T␤RII and the Fc domain of human IgG 1 (sT␤RII:Fc) results in increased apoptosis in primary mammary tumors and a reduction in both circulating tumor cells and lung metastases (24). Taken together, these data suggest that 1) TGF-␤ can suppress some but not all oncogenic signals in epithelial cells, and 2) some oncogenes can engage TGF-␤ signaling and utilize it for tumor maintenance and progression.
We have studied the HER2 proto-oncogene product, the human homolog of erbB2 and neu, to determine if it modulates cellular responses to TGF-␤ in MCF-10A nontumorigenic human mammary epithelial cells. The HER2 protein is a member of the erbB family of transmembrane receptor tyrosine kinases, which also includes the EGF receptor (HER1, erbB1), HER3 (erbB3), and HER4 (erbB4). Binding of ligands to EGFR, HER3, and HER4 induces the formation of homodimeric and heterodimeric, kinase-active complexes to which HER2 is recruited as a preferred partner (25). Even though HER2 is unable to interact directly with receptor ligands, it can amplify signaling pathways activated by erbB co-receptors, which include phospholipase C␥-1, Ras-Raf-MEK-ERK (extracellular signal-regulated kinase), phosphatidylinositol 3-kinase-Akt-p70S6 kinase, Pak-JNKK-JNK (c-Jun NH 2 -terminal kinase), stress-activated protein kinases, and signal transducers and activators of transcription (26). Some of these pathways such as ERK, p38 MAPK, phosphatidylinositol 3-kinase, and c-Jun NH 2 -terminal kinase are also stimulated by TGF-␤ in some cells (18,27), thus providing a biochemical basis for synergy between TGF-␤ and erbB receptor signaling.
As indicated above, TGF-␤ can suppress TGF-␣-induced transformation in the mammary gland (15). Furthermore, TGF-␤ can also suppress EGF-induced mitogenesis in epithelial cells (28). These results imply that TGF-␤ is dominant over the EGF receptor. Because of the ability of HER2 to signal with increased potency than the EGF receptor (26), we speculated that overexpression of HER2 may prevent TGF-␤-mediated growth inhibition and/or modulate other cellular responses of non-transformed mammary epithelial cells to this ligand. In this report, the proliferation of MCF-10A cells overexpressing stably transduced HER2 was still delayed by exogenous TGF-␤. However, MCF-10A/HER2 cells exhibited increased motility and Rac1 activity in response to TGF-␤, suggesting that HER2 and TGF-␤ signals can cooperate in inducing cellular events required for tumor progression.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-MCF-10A cells were purchased from the American Type Culture Collection (Manassas, VA) and maintained in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium supplemented with horse serum, L-glutamine, human recombinant EGF (Calbiochem), cholera toxin (Biomol, Plymouth Meeting, PA), insulin (Calbiochem), and hydrocortisone (Sigma). Experiments were performed in serum-free Dulbecco's modified Eagle's medium /F-12 medium supplemented with 0.2% bovine serum albumin (fraction V). 293T cells were maintained in Dulbecco's modified Eagle's medium, 10% heat-inactivated fetal calf serum. All tissue culture reagents were from Invitrogen unless otherwise specified. Human recombinant TGF-␤1 was from R&D Systems (Minneapolis, MN) and the humanized IgG 1 HER2 antibody trastuzumab (Herceptin) was purchased from the Vanderbilt University Medical Center Pharmacy. ZD1839 (Iressa) was kindly provided by Alan Wakeling (AstraZeneca Pharmaceuticals, Macclesfield, UK). LY294002 and SB202190 were from Calbiochem, U-0126 was from Promega (Madison, WI), and the integrin ␤ 1 blocking antibody was from Chemicon (Temecula, CA).
Retroviral and Adenoviral Vectors-c-Myc-tagged HER2 cDNA (provided by Dr. James Staros, Vanderbilt University) was excised using XbaI/HindIII enzymes, purified by agarose-gel electrophoresis, and converted to blunt ends with Klenow DNA polymerase. The retroviral expression vector pBMN-IRES-EGFP (29) (provided by Dr. Gary Nolan, Stanford University) was digested with SnaBI and dephosphorylated at the 5Ј-ends with alkaline phosphatase (Promega). The blunt-ended vector and the HER2 cDNA were ligated and then replicated in DH5␣ (Invitrogen). To confirm that the HER2 insert had no truncations, the vector was digested with StuI and also sequenced using the primer 5Ј-GACCTTACACAGTCCTGC-3Ј. Restriction enzymes, T4-DNA ligase, and DNA polymerases were all from New England BioLabs (Beverly, MA). 293T retrovirus-packaging cells (2.5 ϫ 10 6 /60-mm dish; ATCC) were transfected with 5 g of pBMN-HER2-IRES-EGFP or 5 g of pBMN-IRES-EGFP (control) for 24 h using FuGENE 6 (Roche Applied Science). For each case, cells were cotransfected with 3 g of pHCMV-G (30) and 3 g of pSV-⌿-env-MLV (pSV-pol/gag) (31) (provided by Dr. Jane Burns, University of California, San Diego, CA). Virus-containing medium was collected 48 -72 h later and passed through a 45-m filter. MCF-10A cells (10 5 /60-mm dish) were transduced with control or HER2-encoding retroviral vectors for 2 h. After 5 passages, cells stably expressing EGFP were sorted by flow cytometry and expanded. EGFP expression was monitored with an inverted fluorescent microscope and maintained at 100% throughout all experiments.
The hemagglutinin (HA)-tagged ALK5 T204D adenoviral construct was provided by Dr. Kohei Miyazono (Japanese Foundation for Cancer Research, Tokyo, Japan). Stock of recombinant viruses was replicated in 293 cells and titered using the adenovirus expression vector kit (Takara Biomedicals, Tokyo). MCF-10A cells were infected with ALK5 T204D or a control ␤-galactosidase adenovirus at a multiplicity of infection of 10 plaque-forming units per cell for 1 h, unless otherwise specified. After 48 h, ␤-galactosidase expression was assessed using the ␤-galactosidase enzyme system (Promega). Expression of HA-tagged mutant ALK5 was confirmed by HA and T␤RI immunoblot analyses.
Cell Cycle Analysis-Cells were harvested by trypsinization and labeled with 50 g/ml propidium iodide (Sigma) containing 125 units/ml protease-free RNase (Calbiochem). A total of 10,000 stained nuclei were analyzed in a FACSCalibur flow cytometer (BD Biosciences) as described previously (33).
Cell Motility Assays and BrdUrd Incorporation-Cell motility was assessed in wound closure assays as described previously (32). For wound closure experiments, cells were allowed to reach confluence in serum-containing complete growth medium and then incubated for 16 h in serum-free medium before the addition of TGF-␤ or other inhibitors. Wound closure was monitored by microscopy at various times. In some cases wounded monolayers were incubated with BrdUrd (Zymed Laboratories Inc., San Francisco, CA) for 3 h and then stained with a biotinylated BrdUrd antibody according to the BrdUrd staining kit protocol (Zymed Laboratories Inc.). Transwell migration was performed utilizing 5-m pore, polycarbonate filters (Corning Costar Corp., Cambridge, MA) as described (32). Cells migrated to the underside of transwell filters were fixed, stained with Diff-Quick (Dü dingen, Switzerland), and counted by bright field microscopy at 200ϫ in five random fields.
Transcription Reporter Assays-Cells were seeded in 24-well plates and transfected with 3 g/35-mm dish of the Smad-dependent reporter construct p(CAGA) 12 -luciferase (provided by J.-M. Gauthier, Laboratoire GlaxoSmithKline, Les Ulis Cedex, France) along with 0.02 g/well of pCMV-Renilla using FuGENE 6 according to the manufacturer's protocol. Firefly and Renilla reniformis luciferase activities were measured using the dual luciferase assay system (Promega), and the data were normalized utilizing the ratio of firefly to R. reniformis luciferase as described previously (32).
Immunofluorescent Microscopy-Cells were treated with TGF-␤1 on glass coverslips and then fixed and permeabilized as described previously (34) followed by incubation with phalloidin-Texas Red (2 units/ml in phosphate-buffered saline for 30 min; Molecular Probes). Fluorescent images of coverslips mounted in AquaPolyMount (Polysciences) were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss Axiophot upright microscope.
Rac1 Activity Assay-The pGEX plasmid containing GST-PBD (glutathione S-transferase fused with the Pak family binding domain for Cdc42/Rac1) (35) was a gift from R. Cerione (Cornell University, New York, NY). GST-PBD was expressed in Escherichia coli and immobilized by glutathione-Sepharose 4B beads after the GST gene fusion system protocol (Amersham Biosciences). Fifty g of GST-PBD were incubated with 800 g of protein from precleared cell lysates for 3 h at 4°C. GST-PBD was affinity-precipitated using glutathione-Sepharose 4B beads (Amersham Biosciences), and the adsorbed proteins were eluted from the beads by suspending them in Laemmli sample buffer followed by SDS-PAGE and Rac1 immunoblot analysis as described previously (34).
cDNA Array and Northern Analysis-Cells were incubated for 16 h in serum-free medium and then treated with TGF-␤1 for 24 h. Total RNAs were isolated with Trizol reagent (Invitrogen), digested with RNasefree DNase (0.1 unit/1 g RNA; Promega), and purified using a silica gel column (RNeasy Kit, Qiagen). Purified RNAs (50 g) were reversetranscribed to cDNAs, and these were labeled with Cy3-conjugated dCTP (MCF-10A/HER2) or Cy5-conjugated dCTP (MCF-10A/vector). The procedures for reverse transcription, RNA template digestion, and cDNA purification and hybridization to a human 5000 cDNA array were performed following the protocols listed in www.array.vanderbilt.edu. Cy3-and Cy5-labeled cDNAs were scanned at 532 and 635 nm, respectively, in GenePix Pro (Axon Instruments, Inc., Foster City, CA). DNA polymerase, primers, and fluorescent dye-conjugated nucleotides were from Invitrogen unless otherwise specified. For Northern analysis, RNAs were resolved by agarose gel electrophoresis, transferred onto nylon membranes (Schleicher & Schü ll), and hybridized with cDNA probes using Express hybridization solution (Clontech, Palo Alto, CA) following the manufacturer's protocol. The cDNA probes were made from 3Ј-ESTs excised from cloning vectors, 1) T␤RII 3Ј-EST (accession number AA487034, restriction enzymes EcoR-I/Xho-1) and 2) metallothionein I-H (H77597, EcoR-I/Pac-1). The excised 3Ј-ESTs were isolated by agarose gel electrophoresis, labeled with [␣-32 P]dCTP (PerkinElmer Life Sciences) using the Random Primer labeling system (Invitrogen), and purified with a silica gel column (Qiagen).

TGF-␤ Induces
Motility of MCF-10A/HER2 Cells-We have examined the effects of HER2 overexpression on the response of human mammary epithelial cells to TGF-␤. MCF-10A nontumorigenic mammary epithelial cells were stably transduced with a retroviral vector encoding Myc-tagged HER2 (Fig. 1A). By Western analysis with HER2 and c-Myc antibodies, protooncogene expression was confirmed in the transduced cells. Both HER3 and a 185-kDa phosphotyrosine band were identified in HER2 immunoprecipitates from the HER2-transfected cells, indicating that HER2 was constitutively phosphorylated and associated with HER3 (erbB3) in the absence of exogenous ligands (Fig. 1B). Treatment with recombinant TGF-␤ (for 24 h) increased the fraction of cells in G 1 phase of the cell cycle in MCF-10A/HER2 (34 3 47%) and control cells (32 3 49%) with a simultaneous reduction in the S-phase fraction and a 50% reduction in cell number after 6 days (Figs. 1, C and D).
We next examined the effect of TGF-␤ on the motility of HER2-overexpressing cells in wound closure and transwell motility assays. After 18 h, MCF-10A/HER2 cells migrated into the wounded area but did not close the wound, whereas in MCF-10A/vector cells the wound remained completely open. Similar results were observed in cells plated in transwells (Fig.  2, A and B, first column). The addition of TGF-␤ resulted in complete wound closure by MCF-10A/HER2 cells but had no effect on motility of control cells (Fig. 2, A and B, second  column). TGF-␤-induced motility was blocked by the addition of a saturating concentration of the HER2 blocking antibody Herceptin (Fig. 2, A and B), thus implying that the effect of TGF-␤ was HER2-specific. The EGF receptor tyrosine kinase inhibitor ZD1839 (Iressa), which has been shown to block phosphorylation of HER2 (36), also inhibited TGF-␤-induced wound closure completely (data not shown).
Expression of ALK5 T204D Enhances Motility in MCF-10A/ HER2 Cells-To determine whether the motility effect was specific to activated TGF-␤ receptors, we induced TGF-␤ signaling in both cell types by expressing an HA-tagged adenovirus encoding a constitutively active mutant of T␤RI (ALK5 T204D ). Mutation of threonine 204 in ALK5 to aspartic acid leads to constitutive activation of the type I receptor serine/threonine kinase, allowing it to signal in the absence of added ligand or T␤RII (37). Cells infected with a ␤-galactosidase adenovirus at the same multiplicity of infection were used as controls. The efficiency of transduction was Ͼ90% as evaluated by in situ staining for ␤-galactosidase activity 48 h after infection (data not shown). Expression of ALK5 T204D was confirmed by HA immunoblot analysis (Fig. 3A). The ␤-galactosidase-transduced MCF-10A/HER2 cells exhibited slightly higher basal motility than the non-motile MCF-10A/vector cells. However, ALK5 T204D induced wound closure and transwell migration in HER2-overexpressing but not in control cells (Fig. 3, B and C).
TGF-␤-induced Motility of HER2-overexpressing MCF-10A Cells Requires Multiple Signaling Pathways-TGF-␤-induced cell motility in HER2 cells was blocked by the addition of mitomycin C, a compound that inhibits DNA synthesis (data not shown). Therefore, we examined if TGF-␤ stimulated DNA synthesis in the wounded monolayers as measured by BrdUrd incorporation. Interestingly, the addition of TGF-␤ clearly stimulated BrdUrd incorporation in MCF-10A/HER2 but not MCF-10A/vector cells, but this was limited to cells in the edge of the wound (Fig. 4A). Because TGF-␤ can activate non-Smad signaling pathways or induce molecules associated with cell motility (32,34), we examined for evidence of activated (phosphorylated) Akt, ERK and p38 MAPK as well as increased integrin ␤ 1 by immunocytochemistry focusing on cells in the leading edge of the wound. Using commercially available (phospho-specific) antibodies, we were unable to detect the abovementioned molecules. However, TGF-␤-induced wound closure was completely blocked in the presence of the phosphatidylinositol 3-kinase inhibitor LY294002, the MEK1/2 (mitogenactivated protein kinase/extracellular signal-regulated kinase kinase) inhibitor U0126, the p38 MAPK inhibitor SB202190, and the CD29 integrin ␤ 1 blocking antibody (Fig. 4B).
TGF-␤ Induces Actin Reorganization and Lamellipodia in MCF-10A/HER2 Cells-We next investigated if reorganization of the actin cytoskeleton, which is required for cell migration, was being induced by exogenous TGF-␤. Wounded monolayers of HER2-overexpressing and vector control cells were stimulated with TGF-␤. Within 15 min, actin stress fibers, as measured by phalloidin-Texas Red, formed along the largest axis in MCF-10A/HER2 cells within the leading edge of the wound. In addition, the TGF-␤-stimulated advancing cells exhibited membrane ruffling and the formation of lamellipodia. These changes were not seen in MCF-10A/vector cells (Fig. 5A). These changes suggested activation of the Rac1 and Cdc42 GTPases (38). Rac1 and Cdc42 activity was measured by precipitating active GTP-bound forms of these proteins using a GST fusion protein containing the p21-activated kinase (Pak) binding domain (GST-PBD). The Pak family of 62-68-kDa serine/threonine kinases are the best characterized effector proteins of Rac1 and Cdc42 and have been shown to bind preferentially to the GTP-bound forms of these proteins (39). Total Rac1 levels were unchanged in HER2-transduced compared with control cells (Fig. 5B, top). However, higher levels of Rac1 were present as the bound form with GST-PBD in MCF-10A/HER2 compared with control cells, suggesting that MCF-10A/HER2 has higher levels of activated Rac1. Preincubation with Herceptin inhibited both basal and TGF-␤-stimulated Rac1 activity in the HER2-transduced but not control cells (Fig. 5B, bottom). Cdc42 was undetectable by immunoblot analysis in the GST-PBD eluates (data not shown).
Because constitutively active T␤RI (ALK5 T204D ) mimicked the ligand-induced effect on cell motility (Fig. 3), we next determined whether active ALK5 interacted with Rac and Pak. Transduction of mutant ALK5 resulted in Smad2 phosphorylation but did not alter Rac1 and Pak1 protein levels in MCF-10A/HER2 cells (Fig. 5C, left). Interestingly, HER2 was detected by immunoblot analysis in Rac1 and Pak1 precipitates from MCF-10A/HER2 cell lysates under basal conditions. This association was increased by both the addition of TGF-␤ (within 15 min) or transduction with ALK5 T204D but not with the ␤-galactosidase control adenovirus. HA immunoblots of Rac1 pull-downs revealed the presence of ALK5 T204D (Fig. 5C,  right), implying that in HER2-overexpressing cells activation of TGF-␤ results in the association of HER2, Rac, Pak, and the type I TGF-␤ receptor.
TGF-␤ Enhances Smad-dependent Transcription and Differential Gene Expression in MCF-10A/HER2 Cells-The ability of TGF-␤ to inhibit proliferation of MCF-10A/HER2 monolayers (Fig. 1, C and D) implied that Smad function was retained in cells overexpressing HER2. On the other hand, the different effect on motility as a function of HER2 overexpression suggested the possibility of differential ligand-induced gene expression. To test TGF-␤-induced transcription, a reporter construct containing 12 Smad binding elements repeated in tandem, p(CAGA) 12 -luciferase, was transiently transfected into MCF-10A/HER2 and MCF-10A/vector cells along with pCMV-Renilla. Normalized luciferase activity indicated that TGF-␤1 induced Smad-dependent transcription in both cell types, but this induction was 7-fold higher in MCF-10A/HER2 cells. Herceptin minimally reduced TGF-␤-induced reporter activity in MCF-10A/HER2 but not in vector control cells, suggesting that its ability to block ligand-induced motility was independent of an inhibitory effect on gene transcription (Fig. 6A).
Because of the difference in magnitude of TGF-␤-induced reporter activity between both cell types, we next examined differential ligand-induced gene expression (at 24 h) in both cells by cDNA array analysis. The raw data for the differen- tially expressed genes in TGF-␤-treated cells is available in Supplemental Table I. There were 161 genes up-regulated and 215 genes down-regulated at least 3-fold in TGF-␤-treated MCF-10A/HER2 relative to TGF-␤-treated MCF-10A/vector cells. Fig. 6B contains 12 genes that were up-regulated and 10 genes that were down-regulated, all with previously reported functions, in HER2-overexpressing cells relative to vector controls. Among this list, the T␤RII and the metallothionein 1H genes were up-regulated (3.5-and 8.1-fold, respectively) by added ligand in HER2 cells relative to control cells. Up-regulation of these two RNAs was confirmed by Northern analysis, which indicated differences in steady state levels of these transcripts in both cell types (Fig. 6C), consistent with the magnitude of difference reported in the cDNA array analysis. DISCUSSION We have attempted to determine if the HER2 proto-oncogene is dominant over TGF-␤ and whether it can synergize with TGF-␤ on cellular responses associated with epithelial transformation. Overexpression of HER2 in MCF-10A nontumor mammary cells did not abrogate TGF-␤-induced anti-mitogenic effects. However, it was clearly permissive for TGF-␤-induced motility. This motility was blocked by preincubation with Herceptin, a humanized IgG 1 that binds to the ectodomain of the HER2 receptor and blocks its function (40), suggesting that the effect of TGF-␤ was HER2-specific. Similar to exogenous TGF-␤, expression of activated T␤RI without added ligand also induced wound closure and transwell migration in HER2-overexpressing cells but not in control cells with undetectable HER2 protein, further indicating that high levels of HER2 are required for TGF-␤-stimulated cell migration. The retention of the antiproliferative effect of TGF-␤ with the simultaneous induction of cell motility, which is associated with a more transformed phenotype, is consistent with a recent report of bi-transgenic tumors expressing neu, the rat homolog of HER2, and active TGF-␤1. Mammary tumors arising in MMTV/neu x TGF-␤1 S223/225 bi-transgenic mice were smaller and less proliferative than cancers occurring in MMTV/neu mice. However, the bigenic tumor cells were more motile, locally invasive, and metastatic than tumors expressing neu alone (23). In addition, transgenic mice expressing activated forms of neu and activated ALK5 in the mammary gland exhibit longer mammary tumor latency compared with mice expressing MMTV/neu alone. However, the bigenic mice show a higher frequency of extravascular lung metastases (22).
Taken together, these data imply that the antimitogenic and prometastatic effects of TGF-␤ can exist simultaneously. Similar to blockade of HER2 with Herceptin, blockade of phosphatidylinositol 3-kinase, extracellular signal-regulated kinase, p38 MAPK, and integrin ␤ 1 with specific inhibitors completely prevented TGF-␤-stimulated migration of MCF-10A/HER2 cells in wound closure assay (Fig. 4B). These molecules have all been shown to be associated with epithelial to mesenchymal transition, and cell migration was stimulated by TGF-␤ (32,34,(41)(42)(43)(44). Whether these signaling/adhesion molecules are specific to the synergistic effects of TGF-␤ and HER2 on cell motility or whether these results suggest the requirement of these pathways for any type of cell movement not necessarily associated with TGF-␤ and HER2 require further investigation.
One obvious rapid response to exogenous TGF-␤ in HER2- overexpressing cells, but not observed in control cells, was the formation of actin stress fibers as well as membrane ruffling and lamellipodia. This result suggested the involvement of the Rac1 and Cdc42 GTPases. Indeed, activation of HER2 with heregulin has been shown to reorganize the actin cytoskeleton and activate Pak1, a direct target of Rac1 (45), thus enhancing cell migration. Consistent with this previous report, HER2 was constitutively phosphorylated in MCF-10A/HER2 cells, and these cells exhibited constitutive association of HER2 with Pak1 and Rac1 and higher basal Rac activity than control cells were subjected to immunoblot analysis with antibodies for HER2, HA, P-Smad2, total Smad 2/3, Rac1, Pak1, and actin. Right panel, cell lysates (200 g) were precipitated with normal rabbit IgG, Rac1, or Pak1 antibodies. Immunoprecipitates (IP) were subjected to HER2, Rac1, Pak1, and HA immunoblot analyses. An HA band clearly distinct from the IgG band is apparent in the Rac1 precipitate from ALK5 T204D but not ␤-galactosidase-transduced cells.
with low HER2 levels. Both the addition of TGF-␤ and transduction with constitutively active ALK5 T204D markedly enhanced this association. Interestingly, ALK5 was detectable in Rac1 pull downs, and these pull-downs also contained HER2 and Pak1 (Fig. 5C). Whether the endogenous type I TGF-␤ receptor is part of such ligand-activated complex in non-transduced cells requires further investigation. Data consistent with this result were reported recently; Rac1 activity in situ was higher in transgenic tumors expressing neu and active TGF-␤1 compared with tumors expressing neu alone (23). In normal mouse mammary epithelial cells, TGF-␤ activates Rac1, which in turn activates p38 MAPK, a serine-threonine kinase involved in TGF-␤-mediated epithelial to mesenchymal transition and cell motility (34,43,46).
In the report herein, Herceptin inhibited basal and TGF-␤stimulated Rac1 activity in MCF-10A/HER2 cells, suggesting that the latter was required for ligand-induced cell motility. Based on these data, we speculate that activated HER2 and TGF-␤ signaling converge at Rac1, thus providing a threshold of sustained activation that leads to an enhanced migratory phenotype. Rac1 has also been implicated in integrin-specific control of cyclin D1 expression and cell cycle entry (47). The possible association between Rac1 activity and the enhanced DNA synthesis observed in TGF-␤-stimulated cells at the edge of closing wounds (Fig. 4A) remains to be determined. Our data do not necessarily conflict with other possible mechanisms by which TGF-␤ and HER2 can cooperate in the induction of cell migration and invasion. For example, MCF-10A cells stably expressing activated human erbB2 (HER2) and TGF-␤ but not erbB2 alone secrete autocrine factors that were sufficient to induce cell migration (48).
Using a heterologous promoter containing 12 Smad binding elements as a reporter of TGF-␤-induced gene transcription, we found that HER2-overexpressing cells exhibited significantly higher ligand-induced reporter activity. This result suggested differences in ligand-induced gene expression between both cell types that were investigated by cDNA array analysis. A list of differentially up-regulated or down-regulated genes in MCF-10A/HER2 versus MCF-10A/vector cells is included in Fig. 6B. Some of these are worth highlighting. For example, steadystate levels of T␤RII and metallothionein 1H mRNAs were up-regulated ϳ3and 8-fold, respectively, in ligand-stimulated HER2 compared with control cells. Interestingly, studies in fibroblasts have shown up-regulation of TGF-␤ binding in response to injury (49). Metallothioneins are cysteine-rich, low M r proteins that have been associated with more virulent phenotypes of breast cancer (50,51). Three genes induced by TGF-␤ in HER2-overexpressing cells have been associated with metastatic behavior of cancer cells: laminin (52), S110 calciumbinding protein A2 (53), and metastasis-associated factor (mta1) (54). Vimentin, a marker of fibroblasts or epithelial cells that have undergone epithelial to mesenchymal transition, and serum-and glucocorticoid-regulated kinase, a serine-threonine kinase that has been shown to protect mammary cells from apoptosis (55,56), were also up-regulated. Contrary to vimentin, keratin 8, a marker of epithelial cells reported to be reduced in breast cancers relative to benign breast tissues (57), was preferentially reduced by added ligand in MCF-10A/HER2 cells. The quiescin Q6 gene, whose expression is induced as fibroblasts leave the cell cycle (58), was also down-regulated in HER2 cells. Interestingly, the interleukin 1 (IL-1) receptor antagonist (IL-1 ra) gene was predominantly down-regulated in HER2 cells. IL-1 ra levels correlate directly with the presence of estrogen receptors in breast cancers. Because IL-1, which is induced by TGF-␤ in wounded keratinocytes (59), is also associated with estrogen receptor-negative, poor prognosis cancers (60), the down-regulation of IL-1 ra might be permissive for IL-1 actions.
In summary, overexpression of the HER2 tyrosine kinase in mammary epithelial cells alters cellular responses to TGF-␤. Although this overexpression does not change the overall anti-FIG. 6. TGF-␤ enhances Smad-dependent transcription and differential gene expression in MCF-10A/HER2 cells. A, cells were transiently transfected with p(CAGA) 12 -luciferase and pCMV-Renilla in complete growth medium. The following day cells were incubated in serum-free medium for 16 h and treated or not with 1 ng/ml TGF-␤ for an additional 24 h. Normalized p(CAGA) 12 -lux reporter activity was measured as indicated under "Experimental Procedures." Each bar represents the mean Ϯ S.D. relative luciferase activity calculated from 4 wells. B, differentially expressed genes in TGF-␤-treated (24 h) MCF-10A/HER2 relative to MCF-10A/vector cells. TM, tumor marker; CC, cell cycle regulatory molecule; CS, cytoskeleton; CK, cytokine or growth factor; TF, transcription factor; ST, signal transducer; EM, extracellular matrix component. C, Northern analysis of up-regulated T␤RII transcript (4.2 kilobases (kb)) and metallothionein 1H transcript (0.37 kilobases) in HER2-overexpressing cells. The 4.7-and 1.9-ribosomal RNA units are shown in an ethidium bromide-stained formaldehydeagarose gel immediately above. mitogenic effect of TGF-␤, it is permissive for TGF-␤-induced motility and gene expression. These data suggest that HER2 and TGF-␤ signaling synergize in the transformation of mammary epithelial cells. If operative in human breast cancers such cooperation would provide a basis for the combined use of HER2 and TGF-␤ inhibitors in patients with breast carcinoma.