Elucidation of Smad Requirement in Transforming Growth Factor-β Type I Receptor-induced Responses*

Transforming growth factor-β (TGF-β) elicits cellular effects by activating specific Smad proteins that control the transcription of target genes. Whereas there is growing evidence that there are TGF-β type I receptor-initiated intracellular pathways that are distinct from the pivotal Smad pathway, their physiological importance in TGF-β signaling is not well understood. Therefore, we generated TGF-β type I receptors (also termed ALK5s) with mutations in the L45 loop of the kinase domain, termed ALK5(D266A) and ALK5(3A). These mutants showed retained kinase activity but were unable to activate Smads. Characterization of their signaling properties revealed that the two L45 loop mutants did not mediate Smad-dependent transcriptional responses, TGF-β-induced growth inhibition, and fibronectin and plasminogen activator-1 production in R4-2 mink lung epithelial cells lacking functional ALK5 protein. Mutation in the L45 loop region did not affect the binding of inhibitory Smads but did abrogate the weak binding of X-linked inhibitor of apoptosis protein and Disabled-2 to ALK5. This suggests that the L45 loop in the kinase domain is important for docking of other binding proteins. Interestingly, JNK MAP kinase activity was found to be activated by the ALK5(3A) mutant in various cell types. In addition, TGF-β-induced inhibition of cyclin D1 expression and stimulation of PMEPA1 (androgen-regulated prostatic mRNA) expression were found to occur, albeit weakly, in an Smad-independent manner in normal murine mammary gland cells. However, the TGF-β-induced epithelial to mesenchymal transdifferentiation was found to require an intact L45 loop and is likely to be dependent on the Smad pathways.

Transforming growth factor-␤ (TGF-␤) 1 belongs to a family of cytokines that regulate cell proliferation and differentiation of many different cell types (1). TGF-␤ family members, which include TGF-␤s, activins, and bone morphogenetic proteins, were found to possess critical roles during embryogenesis and in maintaining tissue homeostasis during adult life. Deregulated TGF-␤ family signaling has been implicated in multiple developmental disorders and in various human diseases, including cancer, fibrosis, and auto-immune diseases (2).
TGF-␤ family members transduce their signals across the plasma membrane by inducing the formation of heteromeric complexes of specific type I and type II serine/threonine kinase receptors (3). The type I receptor is phosphorylated and activated by the type II receptor (4) and initiates intracellular signaling through activation of downstream signaling components, including the phosphorylation of receptor-regulated (R)-Smad proteins at their extreme C-terminal serine residues. Using chimeric TGF-␤ family receptors, the exposed L45 loop, a nine-amino acid sequence between kinase subdomains IV and V, was found to be important in determining the signaling specificity of type I receptors (5,6) by specifying which Smad isoform is activated (7,8). Accordingly, Smad2 and Smad3 act downstream of TGF-␤ and activin type I receptors, whereas Smad1, Smad5, and Smad8 are phosphorylated by bone morphogenetic protein type I receptors. A number of proteins with anchoring, scaffolding, and/or chaperone activity have been identified that regulate the recruitment of Smads to the type I-type II receptor complex, including Smad anchor for receptor activation (SARA) (9) and Disabled-2 (Dab-2) (10). Whether these components or other identified TGF-␤ receptor-binding proteins may play a role in presentation of other substrates besides Smads remains to be explored. Activated R-Smads form heteromeric complexes with common partner (Co)-Smad, i.e. Smad4, which accumulates in the nucleus, where they control gene expression of genes involved in e.g. growth inhibition, apoptosis, migration, and extracellular matrix production (11). Whereas Smads have been identified as pivotal intracellular effectors for TGF-␤, there is growing evidence that additional pathways are activated downstream of TGF-␤ receptors (11). However, the molecular mechanisms by which these responses are transduced and their physiological significance in TGF-␤ signaling have remained poorly characterized.
Members of the small GTP-binding proteins, including Rac, Rho, and Cdc42, have been reported to become activated by TGF-␤ and to potentiate TGF-␤-induced transcriptional activation (12,13) and induce mobilization of the actin cytoskeleton (14). Three distinct MAP kinases, i.e. extracellular-regulated kinase, c-Jun N-terminal kinase (JNK), and p38, have been shown to be activated by TGF-␤ and other family members (15)(16)(17)(18)(19)(20)(21)(22). Both rapid (within 10 -30 min) as well as delayed (after several hours) activation of MAP kinase have been reported, of which the delayed activation has been shown to depend on Smad pathways (19). The mitogen-activated protein kinase TGF-␤-activated kinase 1 (TAK1) has been shown to be phosphorylated upon TGF-␤ stimulation and can lead to the activation of JNK and p38 MAP kinase pathways (17,23,24). TAK1-binding protein (TAB1) was identified as an activator of TAK1 (25). Hematopoietic progenitor kinase-1 (HPK1) and Xchromosome-linked inhibitor of apoptosis (XIAP) may provide a direct link between TAK1-binding protein (TAB1) and the type I receptor (26 -28). Interestingly, the TGF-␤-induced expression of fibronectin and inhibition of insulin-like growth factorbinding protein-5 or NOV, a secreted glycoprotein, were shown to require JNK activation but to be independent of Smad pathways (29 -31).
The MAP kinase pathways have also been shown to modulate the Smad pathway. Overexpression of dominant negative members of the JNK pathway inhibited TGF-␤/Smad-mediated responses (19,29). Smad3 and AP-1 family members have been shown to cooperate with each other in transcriptional responses by forming direct protein interactions (32)(33)(34)(35)(36)(37). In addition, TGF-␤-induced activation of p38 MAP kinase was shown to induce ATF-2 phosphorylation, which acted synergistically with Smads in transcriptional activation (38). JNK was found to be rapidly activated by TGF-␤ in mink lung epithelial cells (Mv1Lu) and to facilitate TGF-␤-induced Smad activation (19). In contrast, activation of extracellular-regulated kinase MAP kinase has also been shown to induce the phosphorylation in the linker region of R-Smads and, thereby, to inhibit the ligand-induced nuclear accumulation of R-Smads (39 -41).
To investigate the pathways that are activated in the absence of Smad activation we have generated a TGF-␤ type I receptor (ALK5) with mutations in the L45 loop of the kinase domain that are defective in Smad activation and characterized the signaling properties of these mutated type I receptors.
Transcriptional Reporter Assays-One day before transfection Mv1Lu or R4-2 cells were seeded at 4.0 ϫ 10 5 cells/well in 12-well plates. The cells were transfected using FuGENE 6. Where indicated, TGF-␤3 was added into the dishes 24 h after transfection. Subsequently, the cells were cultured in DMEM containing 0.3% FCS for 18 h. In all experiments, ␤-galactosidase (pCH110, Amersham Biosciences) activity was measured to normalize for transfection efficiency. Each transfection was carried out in triplicate and repeated at least twice.
Immunoprecipitation and Western Blotting-To detect Smad2 phosphorylation by constitutively active (ca) ALK5 or its mutant derivatives in COS7 cells, 3 g of pDEF3-FLAG-Smad2 and 3 g of caALK5 or its mutant derivatives were transfected in COS7 cells at 1.5 ϫ 10 6 cells/ 10-cm dish using FuGENE 6. Forty hours after transfection, the cells were lysed in 1 ml of lysis buffer (20 mM Tris (pH 7.4), 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 2.5 g/ml aprotinin, 2 mM sodium vanadate, 40 mM NaF, and 20 mM ␤-glycerophosphate). The cell lysates were precleared with protein G-Sepharose beads (Amersham Biosciences) and incubated with FLAG M5 antibody (Sigma) for 2 h at 4°C. Subsequently, protein G-Sepharose beads were added to the reaction mixture and incubated for 30 min at 4°C. After washing the immunoprecipitates with lysis buffer three times, the proteins in immunoprecipitates and aliquots of total cell lysates were separated by SDS-PAGE and transferred to a Hybond-C Extra membrane (Amersham Biosciences). The membrane was subsequently probed with phosphorylated Smad2-specific antibodies (pS2) (8) or FLAG M5 antibody. Primary antibodies were detected with horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse antibody (Amersham Biosciences) and chemiluminescent substrate. The expression of ALK5 receptors was determined by immunoprecipitation of the [ 35 S]cysteine/methionine-labeled cells (44). The detection of the interactions between caALK5 and I-Smads, between caALK5 and XIAP, or between ALK5 and Dab-2 was performed by immunoprecipitation followed by Western blotting according to the above method, except that cells were lysed in TNE buffer (10 mM Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 2.5 g/ml aprotinin, 2 mM sodium vanadate, 40 mM NaF, and 20 mM ␤-glycerophosphate). The expression of proteins in R4-2 or NMuMG stable transformants was detected by Western blotting of total cell lysates in TNE buffer. pS2 antibody for phosphorylated Smad2, anti-Smad2 antibody (Transduction Laboratories) for total Smad2, anti-HA12CA5 antibody (Roche Molecular Biochemicals) for ALK5 or ␣ALK5 chimera, and anti-GM-CSF ␤R antibody for ␤RII chimera were used as primary antibodies.
Receptor Autophosphorylation Assay-One day before transfection, COS7 cells were seeded at 2.0 ϫ 10 5 cells/3.5-cm dish. The cells were transfected with 1.5 g of pCDNA3-caALK5/HA or its mutant deriva-tives by FuGENE 6. Forty hours later, the cells were lysed with lysis buffer, precleared with protein G-Sepharose beads, and incubated with anti-HA12CA5 antibody for 2 h at 4°C. Protein G-Sepharose beads were then added to the reaction mixture, and incubation was continued another 30 min at 4°C. The immunoprecipitates were washed with lysis buffer three times and with the kinase reaction buffer (10 mM Tris (pH 7.4), 10 mM MgCl 2 , and 2 mM MnCl 2 ) twice and then incubated with kinase reaction buffer containing 14.8 kBq/ml [␥-32 P]ATP (Amersham Biosciences) for 30 min at 25°C. Immunoprecipitates were separated by SDS-PAGE. The expression of receptor proteins was determined by 35 S-metabolic labeling and immunoprecipitation (44). Growth Inhibition Assay-R4-2 transformants were seeded at 1 ϫ 10 4 cells/well in 24-well plates. Before the addition of TGF-␤3, the cells were simultaneously treated (or not treated) with 100 M ZnCl 2 to induce ALK5 expression. Two hours before harvest, the cells were pulsed with 18.5 kBq of [methyl-3 H]thymidine (Amersham Biosciences) in 0.5 ml of culture medium. The cells were fixed with ice-cold 5% trichloroacetic acid for more than 20 min and washed twice with 5% trichloroacetic acid and once with water. Solubilization of the cells was done in 400 l of 0.1 M NaOH for 20 min at room temperature. The 3 H radioactivity incorporated into DNA was determined by liquid scintillation counting.
Extracellular Matrix Formation Assay-Cells were seeded in 6-well plates at a density of 2 ϫ 10 6 cells/well in medium containing 100 M ZnCl 2 . After 18 h, the medium was changed to cysteine/methionine-free medium (Invitrogen) with 100 M ZnCl 2 in the presence of 10 ng/ml TGF-␤3, and the incubation was prolonged for another 4 h. During the last 2 h, cells were incubated with 148 kBq/ml 35 S-labeling mixture ProMix (Amersham Biosciences). The cells were removed by washing on ice, once in phosphate-buffered saline (PBS), 3 times in 10 mM Tris (pH 8.0), 0.5% sodium deoxycholate, and 1 mM phenylmethylsulfonyl fluoride, twice in 2 mM Tris (pH 8.0), and once in PBS. Extracellular matrix proteins were scraped off, extracted into SDS sample buffer, and analyzed by SDS-PAGE. Plasminogen activator-1 (PAI-1) was identified as a 45-kDa protein in the extracellular matrix fraction (54). For the detection of fibronectin, R4-2 transformants were seeded at 1.0 ϫ 10 6 cells/6-well plate. After 18 h, the cells were cultured in the presence of 100 M ZnCl 2 for 5 h. Subsequently, the culture medium was changed to DMEM containing 0.3% FCS and 100 M ZnCl 2 with or without 10 ng/ml TGF-␤3. After 16 h, the medium was changed to 1 ml of cysteine/ methionine-free medium containing 100 M ZnCl 2 and 74 kBq/ml 35 Slabeling mixture ProMix with or without 10 ng/ml TGF-␤3. Two hours later, 0.5 ml of culture medium was collected into the tube and incubated with 100 l of gelatin-Sepharose (50% slurry in 50 mM Tris (pH 7.4) and 150 mM NaCl; Amersham Biosciences) and 30 l of 10% Triton X-100 for 16 h at 4°C. Fibronectin is known to bind strongly to gelatin (55). The beads were successively washed once with solution A (50 mM Tris (pH 7.4) and 150 mM NaCl), once with solution B (50 mM Tris (pH7.4) and 500 mM NaCl), and once again with solution A. Gelatinbound proteins were analyzed by SDS-PAGE.
In Vitro JNK Assay-JNK activation was determined by an immune complex kinase assay using GST-c-Jun 1-135 as a substrate (56). 293T cells were seeded at 6.0 ϫ 10 6 /10 cm dish 1 day before transfection. The cells were transfected with 3 g of FLAG-JNK1␣ or 3 g of caALK5/HA or its mutant derivatives by FuGENE 6. After 40 h, cells were lysed in 1 ml of TNE buffer. The cell lysates were precleared with protein G-Sepharose beads and incubated with FLAG M2 (Sigma) antibody for 2 h at 4°C. Subsequently, protein G-Sepharose beads were added to the reaction mixture, and incubation was prolonged another 30 min at 4°C. The immunoprecipitates were washed 3 times with TNE buffer, twice with LiCl buffer (500 mM LiCl, 100 mM Tris (pH 7.6), 0.1% Triton X-100, and 1 mM dithiothreitol), and 3 times with 2ϫ assay buffer (40 mM MOPS (pH 7.2), 4 mM EGTA, 20 mM MgCl 2 , 2 mM dithiothreitol, and 0.2% Triton X-100). Then the immunoprecipitates were incubated in the reaction buffer (20 mM MOPS (pH 7.2), 2 mM EGTA, 15 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Triton X-100, 25 M ATP, and 74 kBq/l [␥-32 P]ATP) containing 0.14 g/l GST-c-Jun-(1-135) at 30°C for 20 min. Simultaneously, the expression of FLAG-JNK1␣ and caALK5/HA or its mutant derivatives was detected by Western blotting using total lysates. To detect the JNK activity in R4-2 cells infected with adenoviral ALK5 or its mutant, R4-2 cells were seeded at 1.0 ϫ 10 6 /10-cm dish 1 day before the infection. After infection with adenoviruses (multiplicity of infection 100), the cells were cultured in DMEM, 10% FCS for 24 h. Subsequently, the cells were starved in DMEM, 0.3% FCS 12 h before TGF-␤3 was added in the dish. After cell lysis with TNE buffer, the kinase assay was carried out as the above except that ␣JNK1 antibody (C-17; Santa Cruz) was used for immunoprecipitations. The anti-HA and pS2 antibodies were used for detection of ALK5 and phosphorylated Smad2, respectively.
Staining of Actin-NMuMG cells or its stable transformants were seeded at 1 ϫ 10 4 cells/well in 8-well glass slides (LAB-TEK) coated with 0.1% gelatin. Eighteen hours later, the cells were starved in DMEM containing 0.3% FCS for 4 h and then stimulated with 10 ng/ml TGF-␤3 or 50 ng/ml human GM-CSF for 20 h. After treatment, the slides were washed once with PBS, fixed for 10 min with 3.7% paraformaldehyde, washed 3 times with PBS, subsequently permeabilized with 0.1% Triton X-100 in PBS for 2 min, and washed again 3 times with PBS. Slides were blocked with 5% normal swine serum (DAKO) in PBS at 37°C for 1 h, and incubated with 5% normal swine serum (in PBS) containing rhodamine-conjugated phalloidin (diluted 1:200) (Molecular Probes). The slides were then washed five times with PBS. To visualize the fluorescence, a confocal laser-scanning microscope (Leica) was used.
Reverse Transcription-PCR Analysis-One day before the infection, MDA-MB-468 cells were plated in 15-cm dishes. The infection with adenoviruses expressing GFP or wild type human Smad4 was carried out as previously described (57). After 24 h, the cells were treated with vehicle or 5 ng/ml TGF-␤1 for 2 or 12 h. At the end of each incubation time total cellular RNA was extracted using RNeasy (Qiagen). Then total RNA (1 g) was digested with RQ1 RNase-free DNase (Promega) and included in cDNA synthesis reaction using Superscript II reverse transcriptase (Invitrogen) with conditions as previously described (58). PCR for fibronectin was performed with the forward primer 5Ј-TG-GAACTTCTACCAGTGCGAC-3Ј and the reverse primer 5Ј-TGTCTTC-CCATCATCGTAACAC-3Ј. PCR conditions were as follows: 95°C for 5 min followed by 33 cycles of 95°C for 30 s, 59°C for 1 min, and 72°C for 1 min, with a final incubation at 72°C for 10 min. PCR for glyceraldehyde-3-phosphate dehydrogenase was carried out with the forward primer 5Ј-ATCACTGCCACCCAGAAGAC-3Ј and the reverse primer 5Ј-ATGAGGTCCACCACCCTGTT-3Ј. PCR conditions were 95°C for 5 min followed by 29 cycles of 95°C for 30 s, 57°C for 1 min, and 72°C for 1 min, with a final incubation at 72°C for 10 min. PCR products were analyzed by 2% agarose electrophoresis and ethidium bromide staining.
Northern Blot Analysis-NMuMG transformants were cultured in DMEM, 0.3% FCS 12 h before stimulation. Total RNAs from NMuMG transformants stimulated with either 10 ng/ml TGF-␤3 or 50 ng/ml GM-CSF were prepared with RNeasy. Fifteen micrograms of total RNA were loaded on the denatured1.5% agarose gel. Blotting and hybridization were performed as previously described (59). The partial cDNAs for mouse PMEPA1 (GenBank TM accession number BG075859) and cyclin D1 (GenBank TM accession number BG083088) were obtained from The Wellcome Trust Sanger Institute (Cambridge, UK), and their sequences were verified.

Generation of ALK5 L45 Mutants Defective in Smad
Activation-To study ALK5-initiated signaling responses that are Smad-independent, we mutated the L45 loop in ALK5 to selectively perturb the ability of ALK5 to bind and activate Smad2 and Smad3. We mutated the aspartic acid residue that is conserved in all type I receptors and the four amino acid residues that we and others previously demonstrated to be important determinants for binding specificity for Smad isoforms between ALK5 and ALK6 (7, 8) (Fig. 1, A and B). We generated single mutants, i.e. ALK5(D266A), ALK5(N267A), ALK5-(D269A), ALK5(N270A), and ALK5(T272A) and a triple mutant, i.e. ALK5(D269A,N270A,T272A), called ALK5(3A).
First we tested the in vitro kinase activity of caALK5 mutants and compared them with caALK5. The constitutively active variants were made by converting threonine residue 204 into aspartic acid (60). caALK5(D266A), caALK5(N267A), and caALK5(3A) but not caALK5(269A), caALK5(N270A), and caALK5(T272A) were found to have a similar enhanced in vitro kinase activity as caALK5 (Fig. 1C, lanes 1-8). We also saw a comparable kinase activity among ALK5, ALK5(D266A) and ALK5(3A) (Fig. 1C, lanes 9, 10, and 12). All ALK5 mutants were also found to form heteromeric complexes with the TGF-␤ type II receptor (T␤R-II) and to bind TGF-␤ with equal efficiency as compared with wild type ALK5 (data not shown). Next, we tested the ability of ALK5 mutants to phosphorylate The amino acid residues that were selected for mutation into alanine residues are indicated with numbers, referring to their position in the ALK5 sequence. The two ␤-strands (␤4 and ␤5) that flank the L45 loop are shown as arrows. C, in vitro autophosphorylation activity of caALK5 and wild type ALK5 and various L45 mutant derivatives. ALK5 constructs were transfected into COS7 cells, immunoprecipitated (IP) with anti-HA antibody, and subjected Smad2 in transfected COS7 cells. Among the three mutants with equivalent kinase activity compared with caALK5, caALK5(D266A) and caALK5(3A) did not phosphorylate Smad2 (Fig. 1D). In addition, we could not see any enhancement of Smad1 phosphorylation by caALK5 or caALK5 mutants (data not shown).
To examine whether ALK5(D266A) and ALK5(3A) can activate Smad-dependent reporter activity, we first transfected caALK5 or its mutants in Mv1Lu cells. As expected, caALK5 dramatically induced (Ͼ70-fold) Smad-dependent luciferase activity. However, caALK5(D266A) and caALK5(3A) exhibited only 4.1-and 2.3-fold activation, respectively (Fig. 1E). ALK5 is known to be present in cells as a dimer in the absence of ligand (61). It is possible that the exogenous caALK5 L45 mutants can make complexes with endogenous ALK5, which can explain the low residual activity of caALK5 L45 mutants. Thus, introduction of ALK5 L45 mutant in the cells, which possess endogenous functional ALK5, may not lead to a complete Smad-inde-pendent signaling. Therefore, we transfected ALK5 L45 mutants in R4-2 cells that lack functional ALK5. As seen in Fig. 1F, neither ALK5(D266A) nor ALK5(3A) activated (CAGA) 12 -Luc (i.e. a readout for activation of the Smad3/ Smad4 pathway) or 2ϫARE-luc reporter (i.e. a read-out for the Smad2/Smad4 pathway). These results were fully consistent with the inability of ALK5(D266A) and ALK5(3A) to induce Smad2 phosphorylation. We therefore used ALK5(D266A) and ALK5(3A) in the following experiments to examine Smad-(in)dependent signaling.

ALK5 L45 Mutants Do Not Rescue TGF-␤-induced Growth Inhibition, Fibronectin, and PAI-1 Production in Mink Cells
That Are Deficient in ALK5-Mv1Lu cells are potently inhibited in their growth and produce high levels of fibronectin and PAI-1 upon TGF-␤ stimulation. To elucidate the abilities of ALK5(D266A) and ALK5(3A) to mediate these responses, we stably transfected these receptors in R4-2 cells that are deficient in functional ALK5. All receptor constructs were placed

FIG. 2. R4-2 Mink cells expressing ALK5 L45 mutants do not mediate TGF-␤-induced growth inhibition or production of fibronectin and PAI-1.
A, ALK5 or mutant derivatives were subcloned into pMEP4, in which transcription can be induced by ZnCl 2 . Expression of ALK5, ALK5(D266A), and ALK5(3A) is shown by Western blot analysis of total cell lysates with anti-HA antibody. Cells were treated with ZnCl 2 before lysis. B, ectopic expression of wild type ALK5, but not ALK5(D266A) or ALK5(3A) mutants, rescue TGF-␤-induced growth inhibition in R4-2 mink cells. Cells were stimulated with ZnCl 2 where indicated. The relative growth compared with non-treated cells is plotted against the concentration of TGF-␤. All values represent the mean Ϯ S.D. C, ectopic expression of wild type ALK5, but not ALK5 (D266A) or ALK5(3A) mutants, rescue TGF-␤-induced fibronectin and PAI-1 production. Fibronectin and PAI-1 protein levels were analyzed by SDS-PAGE followed by PhosphorImager analysis (Fuji). The asterisk shows a nonspecific band. D, regulation of the fibronectin gene by TGF-␤ is Smad4-dependent. MDA-MB-468 cells were infected with the indicated recombinant adenoviruses and treated with TGF-␤ for the indicated times. Ad-GFP, GFP expressing adenovirus; Ad-Smad4, Smad4 expressing adenovirus. Reverse transcription-PCR reactions with specific primers for fibronectin and glyceraldehyde-3phosphate dehydrogenase (GAPDH) as a internal control were performed, and products were analyzed by ethidium bromide staining.
to an autophosphorylation reaction. Expression controls for ALK5 are shown below by immunoprecipitation of 35 S-labeled cell lysates from parallel transfected COS7 cells with anti-HA antibody. D, effect of caALK5 and L45 mutant derivatives on Smad2 phosphorylation. caALK5 or L45 mutant derivatives were transiently co-transfected with FLAG-Smad2 into COS7 cells. The level of C-terminal Smad2 phosphorylation was determined by Western blotting (WB) with anti-phospho-Smad2 (pS2) antibody. Expression controls for ALK5 and Smad2 are shown below. E, caALK5 L45 mutants activate Smad-dependent luciferase reporter weakly in Mv1Lu cells. caALK5 and its derivatives were transiently transfected with (CAGA) 12  under the transcriptional control of the metallothionein promoter, which can be induced by ZnCl 2 . Expression analysis of type I receptors in R4-2 stable transformants revealed ZnCl 2inducible expression of wild type ALK5, ALK5(D266A), and ALK5(3A) (Fig. 2A). Consistent with the experiment in COS7 cells, Smad2 phosphorylation by TGF-␤ was observed only in wild type ALK5-expressing R4-2 cells after the addition of ZnCl 2 (data not shown). We then investigated whether ALK5(D266A) and ALK5(3A) mutants could be substituted for wild type ALK5 with respect to TGF-␤-induced growth inhibition (Fig. 2B), fibronectin, and PAI-1 protein production (Fig.  2C). However, ALK5(D266A) and ALK5(3A) mutants were not able to mediate these responses. Thus, they appear to depend on an intact L45 loop and are likely to be Smad-dependent responses. Consistent with the results obtained from R4-2 cells that express ALK5 L45 mutants, TGF-␤ did not induce fibronectin mRNA levels in MDA-MB-468 cells that are deficient in Smad4. This TGF-␤-induced response with delayed kinetics (induction after 12 h, but not 2 h) could be rescued after infection of MDA-MB 468 cells with Smad4 adenovirus (Fig.  2D). Thus, our results indicate that the expression of fibronectin is regulated by the TGF-␤/Smad pathway.
ALK5 L45 Mutant Interacts with I-Smads but Not XIAP or Dab-2-I-Smads bind to activated type I receptors and can compete with R-Smads for receptor binding (62)(63)(64). However, the domain(s) in the type I receptor responsible for I-Smad binding have not been characterized. To test the interaction of Smad6 or Smad7 with ALK5(3A) mutant, I-Smads and receptors were transfected into COS7 cells and subjected to immunoprecipitation followed by Western blotting. The ALK5(3A) mutant was found to interact with Smad6 or Smad7 as efficiently as wild type ALK5 (Fig. 3A). This suggests that the L45 loop is not important for interaction with I-Smads. However, when we analyzed the interaction of two other components known to bind type I receptors, i.e. XIAP (47) (Fig. 3B) and Dab-2 (10) (Fig. 3C), we found that none of them interacted with the ALK5(3A) mutant using a similar strategy as above. The L45 loop region may, thus, not only be important for Smad binding but also for interaction with other signaling components.
293T cells were chosen because they have very little endogenous ALK5 and can be efficiently transfected. Expression constructs for HA-tagged caALK5 or caALK5(3A) was co-transfected with an expression construct for JNK in 293T cells, and cell lysates were subjected to immunoprecipitation with anti-FLAG M5 antibody followed by in vitro kinase reaction using GST-c-Jun 1-135 as a substrate. Like caALK5, caALK5(3A) was found to potently activate JNK (Fig. 4A). In contrast, the phosphorylation of Smad2 in 293T cells was not enhanced by caALK5(3A) but was dramatically induced by caALK5 (data not shown). Consistent with these results, TGF-␤ induced a significant JNK activation in R4-2 cells that were infected with wild type ALK5 or ALK5(3A) mutant adenoviruses. Peak levels of JNK activation were reached 10 min after TGF-␤ challenge (Fig. 4B). TGF-␤ induced Smad2 phosphorylation in R4-2 cells expressing wild type ALK5 but not in cells expressing ALK5(3A) mutant (Fig. 4B). Thus, TGF-␤-induced JNK activation occurs in a Smad-independent manner. Further support for ALK5(3A)-induced activation of JNK was obtained by analyzing the activation of Gal4-c-Jun in R4-2 cells. caALK5(3A) significantly induced the Gal4-cJun transcriptional reporter, albeit weaker than caALK5 (Fig. 4C, upper panel). Consistent with the Gal4-cJun assay, the activation of pAP-1-Luc, which can be activated by JNK, was also weakly induced by caALK5(3A) compared with that of caALK5 (Fig. 4C, middle  panel). Transfection of ALK5(3A) in R4-2 cells mediated a 2-fold activation of pAP-1-Luc reporter in response to TGF-␤ (Fig. 4D). In contrast, the Gal4-CHOP assay, which exhibits a readout of the activated p38 pathway (65), was not significantly affected by caALK5(3A) but was influenced by caALK5 (Fig.  4C, lower panel). Characterization of Chimeric GM-CSF/TGF-␤ Receptors-GM-CSF is known to be a species-specific ligand. Chimeric receptors between GM-CSF and TGF-␤ receptors have shown that TGF-␤ signaling can be reconstituted in a system independent of TGF-␤ ligand (42,66). We generated a chimeric receptor between the extracellular domain of GM-CSF ␣R and the intracellular domain of ALK5(3A) or ALK5(D266A), which FIG. 3. ALK5(3A) mutant interacts with I-Smads but not with XIAP nor Dab-2. A, caALK5 or caALK5(3A) mutant was cotransfected with 6ϫMyc-Smad6 or 6ϫMyc-Smad7. To show interaction between ALK5 and I-Smads, the cell lysates were subjected to immunoprecipitation (IP) with anti-HA antibody followed by Western blotting (WB) with anti-Myc antibody. Expression controls for ALK5 or I-Smads are shown below by Western blotting on total cell lysates. B, caALK5 without or with 3A mutation of the L45 loop was cotransfected with expression constructs for GST-XIAP. To show interaction between components, cell lysates were first subjected to immunoprecipitation with anti-HA antibody followed by Western blotting with anti-GST antibody. Expression controls are shown below by Western blot analysis of total cell lysates. C, ALK5 or ALK5(3A) was cotransfected with FLAG-Dab-2 in COS7 cells. The cell lysates were immunoprecipitated with anti-HA antibody followed by Western blotting with FLAG M5 antibody. The expression of ALK5 and Dab-2 is shown below by Western blotting of total lysates. allowed us to study TGF-␤ signaling in a cell upon stimulation with GM-CSF when co-transfected with GM-CSF ␤R/T␤R-II chimera (Fig. 5A) . Both chimeric receptor chains were subcloned in retroviral expression vectors. After infection of NMuMG cells, chimeric receptor-expressing clones were sorted by fluorescenceactivated cell sorter analysis. We analyzed TGF-␤-or GM-CSFinduced Smad2 phosphorylation in wild type NMuMG cells and in cells expressing GM-CSF ␤R/T␤R-II chimera alone (termed ␤II) or together with GM-CSF ␣R/ALK5 (termed ␣wt/␤II), GM-CSF ␣R/ALK5(D266A) (termed ␣(D266A)/␤II), or GM-CSF ␣R/ALK5(3A) (termed ␣(3A)/␤II) (Fig. 5A). As expected, TGF-␤ induced Smad2 phosphorylation in all cells, whereas GM-CSF only induced Smad2 phosphorylation in cells expressing ␣wt/␤II but not in other cell clones (Fig. 5B). Thus, consistent with previous results, the intracellular domain of neither ALK5(D266A) nor ALK5(3A) induced Smad2 phosphorylation when the chimeric receptors were activated. In addition, GM-CSF stimulated a significant increase in AP1-Luc activity in ␣(3A)/␤II cells. In- terestingly, the luciferase activity was higher after 6 h compared with 18 h of GM-CSF stimulation (Fig. 5C).

GM-CSF/ALK5 L45 Mutants Can Modulate Gene Expression but Fail to Induce Stress Fibers in NMuMG Cells-Transcrip-
tional analysis on ␣wt/␤IIand ␣(3A)/␤II-expressing NMuMG cells upon stimulation with GM-CSF indicated that the genomic response is much stronger in the presence than the absence of Smad signaling. 2 Among the limited group of genes that were found to be up-regulated and down-regulated by GM-CSF (and TGF-␤) in ␣wt/␤IIand ␣(3A)/␤II-expressing cells are PMEPA1 and cyclin D1, respectively (Fig. 6, A and B). Wild type ALK5 was found to mediate a stronger signal than ALK5(3A) mutant; this suggests that both Smad-dependent and Smad-independent signaling are needed to efficiently regulate these genes by TGF-␤. PMEPA1 was initially identified as an androgen-regulated prostatic mRNA (67) without functional annotation. Cyclin D1 was previously shown to be a target of TGF-␤ (68) and has been implicated in TGF-␤-induced growth arrest.
NMuMG cells transdifferentiate from an epithelial phenotype to a spindle-shaped morphology in response to TGF-␤ as can be demonstrated by a reorganization of the actin cytoskeleton (69,70). We therefore analyzed ␤II, ␣wt/␤II, and ␣(3A)/␤II for TGF-␤-and GM-CSF-induced stress fiber formation. As expected, we found that TGF-␤ induced stress fiber formation in all cell lines (Fig. 6C), whereas GM-CSF induced stress-fiber formation in ␣wt/␤II cells (Fig. 6C) but not in ␣(3A)/␤II cells (Fig. 6C), ␣(D266A)/␤II cells (data not shown), and ␤II cells (Fig. 6C). In addition, we also observed stress fiber formation by GM-CSF in an ␣wt/␤II-expressing Swiss3T3 transformant but not in ␣(D266A)/␤IIor ␣(3A)/␤II-expressing Swiss3T3 transformants (data not shown). Thus, an intact L45 loop is critical for ALK5-mediated induction of stress fibers. receptor to the nucleus, where they affect the transcription of target genes. TGF-␤-induced responses that are independent of Smads have been reported (11). However, their physiological importance is not well understood. Therefore, we investigated the pathways that are activated in mutant TGF-␤ type I recep-tors (ALK5s) defective in Smad activation but with retained kinase activity. Most of the TGF-␤-induced responses in a variety of cell lines that we examined, including growth inhibition, fibronectin and PAI-1 protein production, and stress fiber formation, were dependent on having an ALK5 with in-FIG. 6. GM-CSF/ALK5 L45 mutants can mediate effects on gene expression of specific genes but fail to induce stress fibers in NMuMG cells. The expression of PMEPA1 and cyclin D1 were enhanced and decreased in the chimeric GM-CSF/ALK5 L45 mutant stimulated with GM-CSF, respectively. A, Northern blot using total RNA from each transformant stimulated with 10 ng/ml TGF-␤3 (T) or 50 ng/ml GM-CSF (G) for 1 and 6 h was performed using PMEPA1 as a probe (upper panel). Relative expression levels (normalized using 28 S) compared with non-stimulated cells are indicated. Equal loading of RNA samples is shown by ethidium bromide stain of gel before Northern blotting (lower panel). B, cyclin D1 mRNA in NMuMG transformants stimulated with 10 ng/ml TGF-␤3 (T) or 50 ng/ml GM-CSF (G) for the indicated times was detected by Northern blot analysis (upper panel). Relative expression levels (normalized using 28 S) compared with non-stimulated cells are indicated. Equal loading of RNA samples is shown by ethidium bromide stain of gel before Northern blotting (lower panel). C, cells stably expressing ␤II in the absence or presence of ␣wt and ␣(3A) or nontransfected cells treated without or with TGF-␤3 or GM-CSF were stained with phalloidin for polymeric actin. tact L45 loop and are likely to be Smad-dependent. ALK5mediated activation of JNK was found to be independent of Smads. In addition, T␤R-I was found to weakly regulate the expression of specific genes in a Smad-independent manner. Taken together these results indicate that different independent signaling pathways are initiated from the activated type I receptor. Although our results show that Smad-independent signaling through TGF-␤ receptor is not sufficient to mediate many of the regulatory effects of TGF-␤ on cell proliferation and differentiation, further studies are under way to investigate the effects of the mutants in more short term responses, like mobilization of the actin cytoskeleton. In addition, our data certainly do not rule out an important role for TGF-␤-induced Smad-independent signaling in either promoting, inhibiting, or redirecting the Smad pathway or an important role for Smadindependent signaling in TGF-␤-induced responses that are induced via Smad-mediated changes in expression of genes, e.g. growth factors, their receptors, and AP1 family members.
Consistent with previous data that implicate Smads in TGF-␤-induced growth arrest (71,72), we found that ALK5 L45 mutants defective in Smad activation were unable to mediate growth arrest despite being able to down-regulate cyclin D1 (Fig. 6B). Cyclin D1 has been implicated in TGF-␤-induced growth arrest. However, either TGF-␤-induced down-regulation of cyclin D1 in Smad-independent manner is not efficient enough or Smad-dependent regulation of other cell cycle regulators, e.g. up-regulation of CDK inhibitors p15 and p21 and down-regulation of c-Myc, are needed for anti-proliferative action of TGF-␤. In EpH4 polarized mammary epithelial cells ALK5-initiated signaling to growth arrest has been shown to occur in part independently of the Smad pathway via phosphatase 2A-mediated inactivation of p70S6K (73). Whether our ALK5 L45 mutants can mediate (partial) growth arrest in EpH4 cells remains to be investigated. The dependence for intact L45 loop for the observed ALK5-induced PAI-1 and fibronectin protein production in R4-2 cells (Fig. 2C) is consistent with the inability of ALK5 L45 mutants to activate the PAI-1 promoter based reporter (CAGA) 12 -Luc assay (Fig. 1F) and the lack of TGF-␤-induced fibronectin mRNA levels in Smad4 deficient MDA-MB-468 cells (Fig. 2D). The observed TGF-␤-induced fibronectin has been previously shown to be dependent on JNK activation (29). However, the ability of the ALK5 L45 mutant to activate JNK without inducing fibronectin levels indicates that JNK activation alone by TGF-␤ is not sufficient for the induction of fibronectin production. TGF-␤ has been reported to induce fibronectin protein levels independent of Smad4 (29). However, we found that TGF-␤-induced fibronectin mRNA levels require Smad4 (Fig. 2D). The reason for this discrepancy remains to be elucidated.
The ALK5(3A) mutant was found to interact with I-Smads (Fig. 3). I-Smads have been shown to compete with R-Smads for interaction with activated type I receptors (62)(63)(64). Taken together, these results suggest that R-Smads and I-Smads interact also with a region in the receptor, such as glycine-serinerich domain (74). The weak binding of XIAP and Dab-2 was found to be dependent on the intact L45 loop in ALK5. Thus, the L45 loop region may be important for interaction not only R-Smads but also with other signaling components. Therefore, to complement our studies on Smad-independent signaling using ALK5 L45 mutant-expressing cells, we are currently examining the responses on TGF-␤ in cells that lack certain R-Smads or Smad4.
Activation of JNK was shown to be enhanced by the caALK5(3A) mutant (Fig. 4). Thus, this response is independent of the Smad pathway. The observation that ALK5(3A) is weaker than wild type ALK5 in activating c-Jun-based tran-scriptional reporters suggests that Smad signaling contributes to the activation of this reporter, as previously shown (32)(33)(34)(35)(36)(37). XIAP has been implicated as a link between the type I receptor and the JNK pathway (26). However, ALK5(3A), which is not capable of binding to XIAP, can induce moderate JNK-mediated reporter activities (Fig. 4B). This suggests that XIAP is dispensable for JNK activation mediated by TGF-␤ in R4-2 cells. Although the TGF-␤ receptor-interacting protein Daxx has been known to mediate JNK activity (21), we were unable to show enhancement of JNK activity in the presence of caALK5 or caALK5(3A) in 293T cells (data not shown). Receptor-mediated activation of Smads may occur at early endosomes as SARA (Smad anchor for receptor activation), the molecule presenting R-Smads to the type I receptor, is exclusively located in this organelle (75)(76)(77). It will be of interest to examine whether TGF-␤-induced JNK activation is initiated at the plasma membrane or at early endosomes.
The TGF-␤-induced change of epithelial cells into fibroblastoid-shaped cells and formation of actin stress fibers were found to require an intact L45 loop. This result is consistent with previous data that show that Smads alone can weakly induce stress fiber formation and cooperate with activated ALK5 for full epithelial-to-mesenchymal transition (70). While our manuscript was in preparation, Yu and co-workers reported on the characterization of a similar ALK5 L45 mutant capable of inducing apoptosis but not epithelial-to-mesenchymal transition via a p38 pathway in NMuMG cells (48). We observed a weak but not consistent p38 phosphorylation in response to activation of mutant chimeric GM-CSF⅐TGF-␤R complex. The differences between our observations and those of Yu et al. (48) may be because of the use of constitutively active ALK5 L45 mutant by Yu et al. (48) compared with ligandmediated activation of chimeric GM-CSF/TGF-␤ receptor complex by us. An advantage of the use of constitutively active receptor is that the signal is built-up in the cell, allowing easier detection. However, we found that caALK5 L45 mutants, when expressed in Mv1Lu cells expressing endogenous ALK5, very weakly activated a Smad-dependent reporter (Fig. 1E). In addition, caALK5 L45 mutant receptors will induce a sustained response, which does not occur when ALK5 L45 mutant receptors are activated by ligand.
In conclusion, using Smad-activation-defective L45 loop mutants of ALK5, we have shown that ALK5-mediated JNK activation, and certain gene responses can occur independent of the Smad pathway. The ALK5 L45 mutants will be important tools to examine the requirement of Smads, and other signaling components that bind to the L45 loop for the various effects by TGF-␤.