Modulation of NFκB Activity and E-cadherin by the Type III Transforming Growth Factor β Receptor Regulates Cell Growth and Motility*

Transforming growth factor β is growth-inhibitory in non-transformed epithelial cells but becomes growth-promoting during tumorigenesis. The role of the type I and II receptors in tumorigenesis has been extensively studied, but the role of the ubiquitously expressed type III receptor (TβRIII) remains elusive. We developed short hairpin RNAs directed against TβRIII to investigate the role of this receptor in breast cancer tumorigenesis. Nontumorigenic NMuMG mouse cells stably expressing short hairpin RNA specific to mouse TβRIII (NM-kd) demonstrated increased cell growth, motility, and invasion as compared with control cells expressing shRNA to human TβRIII (NM-con). Reconstitution of TβRIII expression with rat TβRIII abrogated the increased growth and motility seen in the NM-kd cells. In addition, the NM-kd cells exhibited marked reduction in the expression of the adherens junction protein, E-cadherin. This loss of E-cadherin was due to increased NFκB activity that, in turn, resulted in increased expression of the transcriptional repressors of E-cadherin such as Snail, Slug, Twist, and Sip1. Finally, NMuMG cells in which TβRIII had been knocked down formed invasive tumors in athymic nude mice, whereas the control cells did not. These data indicate that TβRIII acts as a tumor suppressor in nontumorigenic mammary epithelial cells at least in part by inhibiting NFκB-mediated repression of E-cadherin.

to transmembrane receptor serine/threonine kinases. TGF␤1 and TGF␤3 can bind with high affinity to the TGF␤ type II receptor, resulting in activation of Smad 2/3 and downstream target genes. In contrast, TGF␤2 binds with low affinity to the type II receptor. The type III receptor (T␤RIII), or betaglycan, binds with high affinity to all three TGF␤ isoforms and is required for presenting TGF␤2 to the type II receptor (2).
T␤RIII has been shown to play an essential role in the formation of the atrioventricular cushion in the development of the heart (3). Consistent with this observation, the T␤RIII null mouse is embryonically lethal because of heart and liver defects (4). The role for T␤RIII in cancer is less clear. Increased expression of TGF␤1 and all three TGF␤ receptors was found in higher grade lymphomas (5). Conversely, reduced expression of T␤RIII was found associated with advanced stage neuroblastomas and ovarian carcinomas (6,7). Similarly, a recent report using cDNA microarrays demonstrated that low T␤RIII expression correlates with higher grade among a cohort of breast cancers (8). Additionally, overexpression of T␤RIII in MDA-231 human breast cancer cells and DU145 prostate cancer cells resulted in decreased tumor invasion in vitro and in vivo (9,10). The reasons for this apparent contradictory role for T␤RIII in these different tumor types have not been elucidated.
Epithelial to mesenchymal transition (EMT) is a process by which TGF␤ can promote tumorigenesis. EMT is characterized by a decrease in epithelial cell markers, such as E-cadherin and ZO-1, and an increase in mesenchymal markers including N-cadherin, vimentin, and fibronectin. This is associated with a decrease in cell-cell adhesion and changes in the actin cytoskeleton. Loss of E-cadherin expression, either through genetic or epigenetic alterations, is the hallmark of EMT in epithelial cells. Several proteins (i.e. Snail, Slug, Twist, and Sip1) have been identified as transcriptional repressors of E-cadherin (11)(12)(13).
NFB is a family of hetero-or homodimeric transcription factors involved in cell survival and regulation of the immune response (14). The NFB signaling pathway appears to be a critical mediator of EMT (15,16). Additionally, it has been reported that NFB is required for EMT during breast cancer progression (17). NFB also appears to be a mediator of Snail expression (15,18). A recent report demonstrated that expression of E-cadherin can down-regulate NFB activity in melanoma cells, suggesting a direct link between these two pathways (19).
The reported conflicting roles of T␤RIII in tumorigenesis lead us to investigate its role in mammary cell transformation.
For this purpose, we used a loss of function approach with short hairpin RNAs (shRNAs) specific to T␤RIII in nontumorigenic NMuMG mammary epithelial cells. In this study we show that knock-down of T␤RIII expression in NMuMG cells results in increased growth, motility, invasion, and tumor formation in vivo using a xenograft mouse model. In addition, we demonstrate that these changes are due to an increase in NFB signaling that, in turn, results in transcriptional repression of E-cadherin. These results were not limited to NMuMG cells because we also observed a similar phenotype in EMT6 mouse mammary tumor cells in which T␤RIII was knocked down by stable RNA interference.

EXPERIMENTAL PROCEDURES
Cell Culture, Viral Infections, and shRNA-All of the cells were purchased from American Type Culture Collection (Manassas, VA). NMuMG cells were grown in Dulbecco's modified Eagle's medium (DMEM; Cambrex) supplemented with 10% fetal bovine serum (FBS) and 10 g/ml insulin. EMT-6, Phoenix-Ampho, and 293A cells were grown in DMEM containing 10% FBS in a humidified 5% CO 2 incubator at 37°C. To generate retroviruses expressing short hairpin RNAs specific to mouse T␤RIII, the complimentary oligonucleotides 5Ј-GAAA-UGACAUCCCUUCCAC and 5Ј-GUGGAAGGGAUGUCA-UUUG were annealed and ligated into the BglII/HindIII site of the pSuper vector. Retrovirus expressing shRNA specific to human T␤RIII (5Ј-GAGAUGACAUUCCUUCAAC and 5Ј-GUUGAAGGAAUGUCAUCUC) was used as a control. The resulting pSuper plasmids were transfected into Phoenix-Ampho cells using Superfect transfection reagent (Qiagen) per the manufacturer's directions. Supernatant containing viral particles was collected 72 h after transfection and filtered through a 20-m syringe filter. Retrovirus containing mouse or human short hairpin RNA specific to the type III receptor were used to infect NMuMG cells in the presence of 4 g/ml polybrene (Sigma). Stably expressing cells were selected with 1 g/ml puromycin. The pGEM-4Z rat T␤RIII plasmid was a gift from Dr. Joan Massague (Memorial Sloan Kettering Cancer Center, New York, NY). Rat T␤RIII was digested out of pGEM-4Z using EcoRI and cloned into the EcoRI site of the LZRS-MS-neo retroviral plasmid. NM-kd cells were infected with retrovirus containing rat T␤RIII and selected using 600 g/ml G418 (Invitrogen). Adenovirus containing constitutively active IKK2 (CA-IKK2) or dominant negative IB␣ (dn IB␣) were provided by Dr. Timothy Blackwell (Vanderbilt University, Nashville, TN). Adenovirus containing GFP was used as a vector control. 293A cells were infected with adenovirus to produce a concentrated stock of virions. For adenoviral infection, NMuMG cells were plated to ϳ70% confluency in 100-mm dishes. Adenovirus was added to the cells in 3 ml of serum-free medium for 3 h, at which point 7 ml of medium containing 10% FBS was added. The cells were allowed to grow for 48 h before being subjected to further treatment.
RNA Isolation and Quantitative PCR-Total RNA was extracted using the RNeasy Mini-kit (Qiagen) per the manufacturer's directions. RNA (5 g) was reverse transcribed in a 100-l reaction. Quantitative PCR was carried out on 500 ng of cDNA using the iQ SYBR Green Supermix from Bio-Rad in a Bio-Rad iCycler iQ multicolor real time PCR detection system. Primers were designed using the Universal Probe Library from Roche Applied Science. Primer sequences are listed in Table 1. A standard curve was generated by amplifying known concentration of cDNA using actin primers. All of the reactions were performed in triplicate.
Cell Growth and Motility Assays-The cells (1 ϫ 10 4 cells/ well) were seeded in 12-well plates in medium containing 10% FBS. The cells were harvested every other day for 7 days, and the cell numbers in each well were measured using a Coulter Counter. Three-dimensional growth assays were carried out in growth factor reduced Matrigel (BD Biosciences) as described (21). The cells were dissolved from the Matrigel using Cell Recovery Solution (BD Biosciences), and their numbers were measured in a Coulter Counter. For motility assays, confluent sheets of cells were "wounded" by scraping with a pipette tip at time of treatment. Wound closure in the presence of added ligand was assessed over time as described previously (22). Transwell assays were performed as previously described (23). In brief, the cells (1.5 ϫ 10 5 /well) were plated in serum-free medium in the upper chamber of 8-m transwells (Costar) and incubated with or without 2 ng/ml TGF␤. After 24 h, cells that had migrated to the underside of the transwell filters were fixed and stained utilizing Diff-Quick Stain Set from Dade Behring AG (Dudingen, Switzerland). The cells in five random fields at 200ϫ magnification were counted.
Immunoblot Analysis-The cells were plated in 60-mm plates and allowed to grow overnight. The cells were then placed in low serum medium (1% FBS) for 16 h, after which the medium was replaced with low serum medium containing either 2 ng/ml TGF␤1 or TGF␤2 for 6 h. The cells were lysed in  NaCl, 1% Nonidet P-40, 20 mM NaF) as previously described (22). Protein concentrations were determined using BCA protein assay reagent (Pierce); 50 g of protein was separated by 9% SDS-PAGE and transferred to nitrocellulose membranes. The membranes were blocked in TBST containing 5% bovine serum albumin for 1 h and then incubated with primary antibody overnight at 4°C. This was followed by incubation with secondary antibody for 1 h at room temperature. The membranes were washed three times in TBST, and the bands were visualized using ECL (Amersham Biosciences).
Transcriptional Reporter Assays-The cells were seeded in 60-mm plates and transfected with 5 g of either 3TP-Lux (provided by Dr. Joan Massague, Memorial Sloan Kettering Cancer Center), NFB-Luc (provided by Dr. Timothy Blackwell, Vanderbilt University), E-cad-Luc (provided by Dr. Amparo Cano, Universidad Autonoma de Madrid, Madrid, Spain), or 0.5 g pCMV-Renilla (Promega) using Superfect transfection reagent (Qiagen) according to the manufacturer's protocol. The next day, the cells were split equally into 48-well plates and incubated overnight in DMEM, 1% FBS, after which the medium was replaced with fresh DMEM, 1% FBS containing either 2 ng/ml TGF␤1 or TGF␤2 for 24 h. Firefly luciferase and Renilla reniformis luciferase activity was measured using a dual luciferase reporter system (Promega) according to the manufacturer's published protocol in a Monolight 3010 luminometer (Analytical Luminescence Laboratory).
ELISAs-Serum-free conditioned medium was removed from growing cells after 72 h and tested using the TGF␤1 or the TGF␤2 Quantikine ELISA kit (R & D Systems) following acid activation as indicated in the manufacturer's protocol. A standard curve using 31.5-2,000 pg/ml human recombinant TGF␤ was generated using the kit reagents and used to calculate the TGF␤ equivalents in the conditioned medium. Each sample was examined in triplicate for a total of three times as described (24). All of the ELISA data are corrected for cell number (pg/ ml/cell number).
Immunofluorescent Microscopy-Immunofluorescent microscopy was performed as previously described (23). Briefly, the cells were grown in 8-well chamber slides for 48 h, fixed in 3% paraformaldehyde, incubated with either phalloidin (1:40 dilution) or E-cadherin (1:500 dilution) primary antibodies for 1 h, and then incubated with fluorescent secondary antibodies for 30 min. Fluorescent images were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss Axiophot upright microscope.
Electrophoretic Mobility Shift Assays-Electrophoretic mobility shift assays were performed using the Promega Gel Shift Assay Kit (Promega) according to the manufacturer's protocol. NM-con and NM-kd cells were grown in DMEM, 1% FBS in the presence or absence of 2 ng/ml TGF␤2 for 1 h. Nuclear extracts were harvested as described previously (25). Nuclear extracts (10 g) were incubated with 32 P-labeled NFB oligonucleotides, separated by 6% SDS-PAGE, and visualized by autoradiography. Unlabeled NFB oligonucleotides (cold NFB) were used as a competitive inhibitor, and unlabeled Oct-1 oligonucleotides were used as a negative control.
Xenograft Assays-The cells (1 ϫ 10 6 cells) were resuspended in 200 l of phosphate-buffered saline and injected with a 22-gauge needle into the right inguinal mammary gland (number 4) of anesthetized 6-week old athymic nude mice (Harlan Sprague-Dawley, Indianapolis, IN) and allowed to grow for 10 weeks. Fat pads, tumors, and lungs were collected, fixed in 10% formalin, and embedded in paraffin.

RESULTS
Knock-down of T␤RIII in NMuMG Cells Impairs Response to TGF␤-We developed shRNAs specific to T␤RIII to determine the role of this receptor in mammary epithelial cells. Nontumorigenic NMuMG mouse epithelial cells were infected with a virus containing the shRNA specific to human (NM-con) or mouse (NM-kd) T␤RIII, and individual clones were isolated through serial dilution. The clones were initially screened using semi-quantitative reverse transcription PCR with primers designed to amplify T␤RIII and actin (data not shown). Positive clones were confirmed by receptor cross-linking with 125 I-FIGURE 1. Knock-down of T␤RIII in NMuMG impairs response to TGF␤2. A, NMuMG, NM-con, NM-kd, and NM-kd&RIII cells were affinity labeled with 125 I-TGF␤1 and cross-linked with BS 3 as described under "Experimental Procedures." Labeling was competed with 100 pM cold TGF␤1. Protein lysates were separated by 3-12% gradient SDS-PAGE and visualized by autoradiography. B, subconfluent NMuMG cell monolayers were incubated in DMEM, 1% FBS overnight. The following day, fresh DMEM, 1% FBS containing either 2 ng/ml TGF␤1 or TGF␤2 was added. The cells were lysed in Nonidet P-40 buffer, and protein was harvested 6 h after treatment. Total protein from whole cell lysates was separated by 9% SDS-PAGE and subjected to immunoblot analysis with the indicated antibodies. Actin was used as a control. C, cells were transfected with the 3TP-Lux TGF␤-responsive promoter, serum-starved overnight in 1% FBS, and stimulated with 2 ng/ml TGF␤1 or TGF␤2 for 24 h. The cells were lysed and assayed for luciferase activity as described under "Experimental Procedures." Relative luciferase units (RLU) represents the ratio of firefly to Renilla luciferase activities. Each data point represents the mean Ϯ S.D. of three wells.
TGF␤1. As a control, mouse NMuMG cells were also infected with shRNA to human T␤RIII. Affinity cross-linking with 125 I-TGF␤1 showed loss of the type III receptor protein only in NM-kd cells but not in cells transduced with the control shRNA specific to human T␤RIII (Fig. 1A). To determine the effect of T␤RIII knock-down on TGF␤ signaling, we examined p-Smad2 by Western blot after TGF␤1 and TGF␤2 treatment. NM-kd cells had significantly reduced p-Smad2 levels after treatment with 2 ng/ml TGF␤2 compared with control cells (Fig. 1B). Additionally, the ability of the NM-kd cells to activate the 3TP-Lux TGF␤-responsive promoter after TGF␤ treatment was markedly inhibited, whereas reconstitution of the receptor with a retrovirus encoding rat T␤RIII (NM-kd&RIII) restored ligand-induced 3TP-Lux reporter activity (Fig. 1C). The transduced rat T␤RIII was also detectable by affinity cross-linking with 125 I-TGF␤1 (Fig. 1A).
Knock-down of T␤RIII in NMuMG Cells Results in Increased Growth, Motility, and Invasiveness-TGF␤ is an inhibitor of epithelial cell growth (26). Thus, we examined proliferation and migration of cells in which T␤RIII has been reduced by RNA interference. NM-kd cells grew significantly faster than NMcon cells, and re-expression of rat T␤RIII in the NM-kd cells reduced their growth rate ( Fig. 2A). Additionally, the NM-kd cells were able to migrate and close a wound in a wound closure assay under reduced serum conditions, whereas control cells and NM-kd cells reconstituted with T␤RIII did not close the wound (Fig. 2B). The NM-kd cells were also more invasive than the NM-con cells as determined by their ability to migrate through transwell filters in the presence of low serum. TGF␤1 and TGF␤2 inhibited transwell migration of control cells but not of the NM-kd cells (Fig. 2C). This result is consistent with the dampened transcriptional response to added ligands observed in NM-kd cells (Fig. 1C).
NMuMG cells are nontumorigenic and form small rounded acini when grown in Matrigel. In contrast, the NM-kd cells form invasive branching structures in three-dimensional Matrigel (Fig. 2D). Reconstitution of T␤RIII abrogated the ability of the NM-kd cells to form these structures in Matrigel. To determine whether the changes in observed cell behavior could involve the production of autocrine ligands, we investigated the secretion of TGF␤ ligands by NMuMG cells. ELISAs were used to determine the amount of total TGF␤1 and TGF␤2 secreted from these cells. The NM-kd cells secret significantly less TGF␤1 and TGF␤2 compared with control cells (Fig. 2E).

Loss of T␤RIII Results in Down-regulation of E-cadherin-
The increased motility and invasiveness of the NM-kd cells was suggestive of EMT. Changes in the content and localization of E-cadherin at adherent cell junctions is a hallmark of EMT (27). Western blot analyses, real time quantitative PCR, and immunocytochemistry demonstrated that the NM-kd cells express greatly reduced levels of E-cadherin (Fig. 3, A-C). The ability of the NM-kd cells to transcriptionally activate an E-cadherin promoter luciferase reporter was also diminished as compared with NM-con cells (Fig. 3D). E-cadherin promoter activity is restored in the NM-kd&RIII cells, where the type III receptor had been re-expressed.
The E-cadherin gene can be regulated by a variety of mechanisms including epigenetic changes, such as hypermethylation, as well as transcriptional repression. Therefore, we examined the mRNA levels of the E-cadherin transcriptional repressors Snail, Slug, Sip1, and Twist by quantitative PCR. mRNA levels of Snail, Slug, Sip1, and Twist were significantly increased in NM-kd cells compared with NMcon cells (Fig. 4). This increase correlated with the reduced expression of E-cadherin (Fig. 3). Further, E-cadherin expression was restored with a corresponding decrease in Snail, Slug, Twist, and Sip1 in NM-kd&RIII cells. These data suggest a role for T␤RIII in maintaining the epithelial phenotype through control of the transcriptional repressors of E-cadherin.
EMT-6 Cells, with Reduced Expression of T␤RIII, Demonstrated an Increased Growth Rate and Decreased E-cadherin Expression-To confirm that these results were specific to loss of T␤RIII expression and not specific to NMuMG cells or caused by clonal variation, we knocked down T␤RIII expression in EMT-6 mouse mammary cancer cells. EMT6-con and EMT6-kd cell lines were generated as described for the NMuMG cells (Fig. 1). The clones were screened by semi-quantitative PCR (Fig. 5A), and EMT6-kd clone C5 was used in the remaining experiments. The EMT6-kd cells demonstrated a muted response to TGF␤2, but not TGF␤1, as determined by 3TP-Lux reporter assays (Fig. 5B). Similar to NMuMG cells, EMT6-kd cells grew better in Matrigel compared with EMT6-con cells (Fig. 5C). In addition, these cells showed decreased E-cadherin protein expression and E-cadherin luciferase activity (Fig. 5, D  and E) with a corresponding increase in Snail mRNA (Fig. 5E). These data demonstrate that the effect of T␤RIII loss on growth and E-cadherin expression is not limited to NNuMG cells.
NFB Activity Is Higher in T␤RIII Knock-down Cells-NFB is a known regulator of Snail (28), and NFB activity has been shown to be regulated (positively and negatively) by TGF␤2 (29,

. NM-kd cells have decreased expression of E-cadherin.
A, cells were grown in 8-well chamber slides for 48 h before being fixed in paraformaldehyde and subjected to immunofluorescence as described under "Experimental Procedures." B, subconfluent cell monolayers were incubated in DMEM, 1% FBS overnight. The following day, fresh DMEM, 1% FBS containing either 2 ng/ml TGF␤1 or TGF␤2 was added. The cells were lysed in Nonidet P-40 buffer, and protein was harvested 6 h after treatment. Total protein from whole cell lysates was separated by 9% SDS-PAGE and subjected to immunoblot analysis with antibodies against E-cadherin and actin (control). C, quantitative PCR was used to determine the mRNA levels of E-cadherin. Total RNA was harvested as described under "Experimental Procedures." Reverse-transcribed cDNA (500 ng) was amplified using primers specific for E-cadherin and actin. All of the C t values were equilibrated to the actin control. Each bar indicates the mean of three wells Ϯ S.D. D, cells were transfected with E-cadherin promoter luciferase reporter, serum-starved overnight, and stimulated with 2 ng/ml TGF␤2 for 24 h. The cells were lysed and assayed for luciferase activity as described under "Experimental Procedures." RLU represents the ratio of firefly to Renilla luciferase activities. Each bar indicates the mean of three wells Ϯ S.D.

FIGURE 4. NM-kd cells express higher mRNA levels of several E-cadherin transcriptional repressors.
Quantitative PCR was used to determine the mRNA levels of Snail, Slug, Twist, and Sip1 as described in the legend to Fig. 3 and under "Experimental Procedures." Each bar indicates the mean of three wells Ϯ S.D. 30). Therefore, we examined NFB activity using a NFB-responsive luciferase reporter in transiently transfected NMuMG cells. The NM-kd cells showed high basal levels of NFB reporter activity as compared with NM-con cells. Re-expression of the type III receptor in NM-kd cells reduced NFB transcription to the same levels as those seen in NM-con cells (Fig.  6A). The ability of NFB to bind to DNA was confirmed by electrophoretic mobility shift assays using 32 P-radiolabeled oligonucleotides containing the putative NFB-binding site (Fig. 6B). Interestingly, TGF␤2 blocked DNA binding in both cell lines. To examine the role of NFB in the invasive phenotype seen in the NM-kd cells, we used an adenovirus containing either dominant negative IB␣ (dn IB␣ S32A/S36A) to abrogate NFB function or a constitutively active IKK2 (CA-IKK2 D177E/D181E) to induce NFB signaling. NM-kd and NM-con cells were transduced with adenoviruses, plated in growth factor reduced Matrigel 24 h later, and allowed to grow for 8 days. NM-kd cells infected with dn IB␣ containing adenovirus showed a significant reduction of growth in Matrigel (Fig. 6, C and D). Transduction of adenoviruses encoding dn IB␣ into NM-kd cells significantly diminished NFB promoter activity. On the other hand, NM-kd cells infected with CA-IKK2 showed a significant increase in NFB reporter activity that was not seen in NM-control cells (Fig. 6, E and F). This result implies that the endogenous type III TGF␤ receptor may counteract the NFB-inducing effect of CA-IKK2. It is significant to note that the reduction in NFB activity did not reduce the ability of the NM-kd cells to grow in Matrigel down to that of the NM-con cells and that these cells still retain their invasive phenotype, suggesting the involvement of additional signaling pathways on the cells invasive behavior.
The decreased ability to grow in Matrigel corresponded to a significant increase in base-line E-cadherin promoter activity in FIGURE 5. EMT-6 cells with reduced T␤RIII exhibit increased growth rate and decreased E-cadherin expression. A, semi-quantitative PCR was performed to determine the mRNA levels of T␤RIII. Total RNA was harvested as described under "Experimental Procedures." Reverse-transcribed cDNA (500 ng) was amplified using primers specific for T␤RIII and actin. B, cells were transfected with the 3TP-Lux TGF␤-responsive promoter, serum-starved overnight in 1% FBS, and stimulated with 2 ng/ml TGF␤1 or TGF␤2 for 24 h. The cells were lysed and assayed for luciferase activity as described under "Experimental Procedures." RLU represents the ratio of firefly to Renilla luciferase activities. Each data point represents the mean Ϯ S.D. of three wells. C, equal numbers of cells were plated in Matrigel and grown for 8 days. The colonies were photographed and then dissolved from Matrigel as described under "Experimental Procedures" and counted using a Coulter Counter. D, subconfluent cell monolayers were incubated in DMEM, 1% FBS overnight. The following day, fresh DMEM, 1% FBS containing either 2 ng/ml TGF␤1 or TGF␤2 was added. The cells were lysed in Nonidet P-40 buffer and protein harvested 6 h after treatment. Total protein from whole cell lysates was separated by 9% SDS-PAGE and subjected to immunoblot analysis with the indicated antibodies. Actin was used as a control. E, quantitative PCR was used to determine the mRNA levels of E-cadherin and Snail as described under "Experimental Procedures." Each bar indicates the mean of three wells Ϯ S.D. NM-con versus NM-kd cells. Infection with an adenovirus containing dn IB␣ up-regulated the low basal E-cadherin promoter activity in NM-kd but not in NM-con cells (Fig. 7, A and  C). Additionally, NM-con cells infected with adenovirus containing constitutively active IKK2 showed a decrease in E-cadherin promoter reporter activity (Fig. 7, A and B), consistent with the role of NFB signaling on EMT. In accordance with these results, E-cadherin mRNA, as measured by quantitative PCR, is significantly increased in NM-kd cells transduced with dn IB␣ (Fig. 7C) with a corresponding decrease in Snail (Fig. 7D). These changes were not observed in the control cells (Fig. 7, C and D).
NM-kd Cells Form Invasive Tumors in Vivo-NMuMG cells are not able to grow and form tumors in athymic nude mice. Because of the apparent gain-of-function effects as a result of knock-down of T␤RIII, we hypothesized that these cells would be able to form tumors in vivo. The cells were injected subcutaneously into the number 4 right inguinal mammary fat pad of nude mice. The mice were sacrificed 12 weeks later, and fat pads were harvested for further pathology. None of the mice injected with NM-con cells developed tumors. However, five of five mice injected with NM-kd cells formed high grade, poorly differentiated carcinomas with mesenchymal and fibroblastoid features (Fig. 8). Macroscopic and microscopic examination of mouse lungs did not reveal any metastases.

DISCUSSION
It has recently been reported that the loss of T␤RIII correlates with higher grade breast cancer, suggesting that this receptor acts as a tumor suppressor (8). Additionally, the loss of T␤RIII has been associated with the development of renal cell carcinoma (31). Two recent reports have also suggested a tumor-suppressor role for T␤RIII in prostate cancer (9,32). The mechanisms by which T␤RIII may exert an antitumor effect are not known. In contrast, TGF␤1, as well as all three TGF␤ receptors, were found to be expressed at higher levels in high grade lymphomas compared with patients diagnosed with low grade lymphomas (5), suggesting a potential oncogenic role for T␤RIII.
TGF␤ signaling can act as a tumor suppressor or a tumor promoter depending on the cellular context. The molecular actions that result in the switch from suppressor to promoter are under intense investigation. The pro-oncogenic properties of TGF␤ include increased motility and invasiveness and enhancement of EMT (33). Excess production of TGF␤ in this FIGURE 6. T␤RIII knock-down results in an increase in NFB activity. A, cells were transfected with NFB-Luc, an NFB-responsive promoter luciferase plasmid and allowed to grow for 24 h before being harvested and tested for luciferase activity. RLU represents the ratio of firefly to Renilla luciferase activities. Each bar indicates the mean of three wells Ϯ S.D. B, subconfluent cell monolayers were incubated in DMEM, 1% FBS overnight. The following day, the cells were treated with fresh DMEM, 1% FBS or fresh DMEM, 1% FBS containing 2 ng/ml TGF␤2 for 1 h. The nuclear extracts were harvested from cells, and electrophoretic mobility shift assays were performed as described under "Experimental Procedures." C-F, cells were infected with adenovirus containing either GFP alone (ad-GFP), dominant negative IB␣ (ad-dn IB␣), or constitutively active IKK2 (ad-CA-IKK2) and either plated in Matrigel (C and D) or transfected with NFB-Luc (E). The cells transfected with NFB-Luc were left to grow for 1 week before being harvested and tested for luciferase activity. RLU represents the ratio of firefly to Renilla luciferase activities. Each data point represents the mean Ϯ S.D. of three wells. The cells grown in Matrigel were photographed and then dissolved from Matrigel as described under "Experimental Procedures" and counted using a Coulter Counter (D). Each bar indicates the mean of three wells Ϯ S.D. Western blot analysis was used to demonstrate infection efficiency of the adenovirus (F).
setting results in an increase in extracellular matrix production, angiogenesis, and an increased production of matrix metalloproteinases (34).
The ability of TGF␤ to induce EMT has been well documented, although the molecular mechanisms have still to be elucidated. It has been clearly demonstrated that TGF␤ can induce the expression of Snail and Slug (35,36). TGF␤1 treatment of epithelial cells results in a mitogen-activated protein kinase-dependent induction of Snail (35). A recent report demonstrates that Snail is required for TGF␤-induced EMT through activation of the phosphatidylinositol 3-kinase and Akt pathway (37). TGF␤2, specifically, has been shown to activate Slug during EMT required for the development of the chick heart (36). TGF␤2and Slug-mediated EMT are dependent on the expression of the TGF␤ type III receptor because blocking this receptor inhibits the initial stages of EMT (36). TGF␤ can also directly modulate NFB signaling. TGF␤ inhibits NFB activity in B cells (38), and NFB has been reported to inhibit TGF␤ signaling via up-regulation of the inhibitory Smad 7 (39), suggesting that there is cross-talk between these two pathways. Furthermore, secretion of FGF5 and TGF␤2 can induce NFB activity, and this autocrine induction results in the constitutive activation of the NFB pathway observed in many tumors (29).
The differential effects of TGF␤ on epithelial cells may be dependent on the activation or inhibition of other oncogenes or tumor suppressor genes. Oncogenic Ras has been shown to overcome the growth-inhibitory effects of TGF␤ (40), and high expression of Ras and active Smad2 are required for metastasis (41). Additionally, the cells may be refractory to TGF␤ growth inhibition but may still rely on TGF␤ signaling for invasion. For example, it has been reported that the tumorigenic 4T1 cell line, which is not growth inhibited by treatment with TGF␤, requires functional TGF␤ signaling for its ability to metastasize (42). Another report demonstrated that treatment of tumor cells with a truncated soluble TGF␤ type II receptor had no effect on tumor cell growth but was able to block tumor metastases (43). In endothelial cells, TGF␤ can signal through two distinct type I receptors, ALK1 and ALK5. ALK1 signaling through Smads 1/5 results in cell proliferation and migration, whereas ALK5 signaling through Smads 2/3 is growth-inhibitory (44). Goumans et al. (45) demonstrate that ALK5 signaling is required for ALK1 activation and that ALK1 signaling is in turn inhibitory of ALK5. In addition, endoglin, an endothelial cell-specific type III TGF␤ receptor, is required to mediate the activation of ALK1 in these cells (46). In this report we present data to support a tumor suppressor role of T␤RIII in NMuMG  were resuspended in phosphate-buffered saline and injected into the number 4 right inguinal fat pad of athymic nude mice (n ϭ 5/group) and allowed to grow for 10 weeks. Tumors were harvested, fixed in formalin, and embedded in paraffin followed by staining with hematoxylin and eosin. A, number of mice that developed tumors. B, representative low and high power hematoxylin and eosin-stained, 5-m tissue sections depicting tumor-free mammary gland from mice injected with control cells and a high grade carcinoma in a mouse injected with NM-kd cells.
nontumorigenic mammary cells. Decreased expression of T␤RIII resulted in an increased growth rate both in monolayer and in three-dimensional Matrigel and increased motility and invasion through transwell filters. All of these effects on growth and motility were abrogated by expression of exogenous rat T␤RIII in the NM-kd cells. Additionally, NM-kd cells were able to form invasive tumors in immunodeficient mice.
Our data are consistent with previous reports suggesting a tumor suppressor role for T␤RIII in breast cancer. In this study, we report for the first time a mechanism by which this receptor inhibits cell growth and motility. We demonstrate that T␤RIII inhibits growth and motility of NMuMG cells by repressing NFB. The loss of T␤RIII results in increased NFB signaling and, in turn, transcriptional repression of E-cadherin resulting enhanced invasiveness in vitro and tumorigenesis in vivo. How T␤RIII is modulating NFB activity is less clear. Lu et al. (29) demonstrated that secreted TGF␤2 could activate NFB. The loss of T␤RIII results in diminished TGF␤2 expression (Fig. 2) and may therefore inhibit further the ability of TGF␤2 to activate NFB in autocrine fashion.
Other signaling pathways may also be involved. It has recently been reported that T␤RIII can activate p38 signaling in the absence of TGF␤ ligand (47), suggesting that this receptor may interact with signaling pathways independent of TGF␤. This is an intriguing speculation because T␤RIII has no known signaling motif, although its C terminus has been shown to be necessary for TGF␤2-induced signaling (48). T␤RIII may also modulate NFB indirectly through interactions with other proteins. The cytoplasmic tail of T␤RIII has been shown to bind the GTPase-activated protein G␣ I subunits interacting protein, C terminus, which stabilizes the expression of T␤RIII at the cell surface (48), and ␤-arrestin 2, which is involved in the degradation of T␤RIII (49), resulting in another layer of regulation of TGF␤ signaling. There may be additional, yet to be identified, protein-binding partners for this receptor that may regulate signaling. An interaction of T␤RIII with signaling pathways independent of TGF␤ would also explain the incomplete reversal of the ability of NM-kd cells to grow in Matrigel when NFB activity is diminished (Fig. 5).
In summary, the data presented herein suggest a role for T␤RIII in regulating cell growth and motility. In nontumorigenic NMuMG mouse mammary epithelial cells, the loss of T␤RIII resulted in an increased ability to grow and invade both in vitro and in vivo because of up-regulated NFB activity and the loss of E-cadherin expression.