Transforming Growth Factor β1 Induces Proliferation in Colon Carcinoma Cells by Ras-dependent, smad-independent Down-regulation of p21cip1*

Transforming growth factor β1 (TGFβ1) can act as a tumor suppressor or a tumor promoter depending on the characteristics of the malignant cell. We recently demonstrated that colon carcinoma cells transfected with oncogenic cellularK-rasV12, but not oncogenic cellular H-rasV12, switched from TGFβ1-insensitive to TGFβ1-growth-stimulated and also became more invasive (Yan, Z., Deng, X., and Friedman, E. (2001)J. Biol. Chem. 276, 1555–1563). We now demonstrate that TGFβ1 growth stimulation of colon carcinoma cells is Ras-dependent and smad-independent. In U9 colon carcinoma cells, which are responsive to TGFβ1 by growth stimulation, a truncating mutation at Gln-311 was found in thesmad4 gene. Very little smad4 protein was detected in these cells. Loss of smad4 protein was confirmed by functional studies. In U9 cells co-transfected wild-type smad4, but not mutant smad4, mediated response of the 3TP-lux and pSBE promoter reporter constructs to TGFβ1. Proliferation initiated by TGFβ1 in U9 cells required Ras-mediated down-regulation of p21cip1 protein. Less p21cip1 was associated with cdk2·cyclin complexes in TGFβ1-treated U9 cells, and the cdk2 complexes had increased kinase activity. Elevation of p21cip1 levels diminished proliferative response to TGFβ1. U9 cells expressing DN-N17ras neither proliferated in response to TGFβ1 nor down-regulated the cdk inhibitor p21cip1, and TGFβ1 activation of 3TP-lux in U9 cells was inhibited by DN-N17ras in a dose-dependent manner. TGFβ1 also decreased p21cip1 levels and stimulated proliferation in SW480 cells, which express mutant K-Ras but no smad4 protein. TGFβ1 did not activate or inhibit the p21cip1 promoter construct in U9 cells even in the presence of co-transfected smad4, or alter p21cip1 mRNA levels. Thus the decrease in p21cip1 levels was mediated by a TGFβ-initiated Ras-dependent, but smad-independent post-transcriptional mechanism.

cells, which are responsive to TGF␤1 by growth stimulation, a truncating mutation at Gln-311 was found in the smad4 gene. Very little smad4 protein was detected in these cells. Loss of smad4 protein was confirmed by functional studies. In U9 cells co-transfected wild-type smad4, but not mutant smad4, mediated response of the 3TP-lux and pSBE promoter reporter constructs to TGF␤1. Proliferation initiated by TGF␤1 in U9 cells required Ras-mediated down-regulation of p21cip1 protein. Less p21cip1 was associated with cdk2⅐cyclin complexes in TGF␤1-treated U9 cells, and the cdk2 complexes had increased kinase activity. Elevation of p21cip1 levels diminished proliferative response to TGF␤1. U9 cells expressing DN-N17ras neither proliferated in response to TGF␤1 nor down-regulated the cdk inhibitor p21cip1, and TGF␤1 activation of 3TP-lux in U9 cells was inhibited by DN-N17ras in a dose-dependent manner. TGF␤1 also decreased p21cip1 levels and stimulated proliferation in SW480 cells, which express mutant K-Ras but no smad4 protein. TGF␤1 did not activate or inhibit the p21cip1 promoter construct in U9 cells even in the presence of co-transfected smad4, or alter p21cip1 mRNA levels. Thus the decrease in p21cip1 levels was mediated by a TGF␤-initiated Ras-dependent, but smad-independent post-transcriptional mechanism.
The transforming growth factor ␤ (TGF␤) 1 family of related proteins is a superfamily of secreted factors that have been implicated in diverse phenomena, including growth control, cell adhesion and motility, production of extracellular matrix components, and alteration of cell phenotype (1,2). TGF␤1 induces its varied biological responses through the paired heteromeric complex of type I (T␤RI) and type II (T␤RII) transmembrane serine/threonine kinases. Upon binding of TGF␤1, T␤RII transphosphorylates T␤RI in a highly conserved GS domain, activating T␤RI kinase to its downstream effectors (3). Genetic studies in Drosophila and Caenorhabditis elegans have identified one set of downstream effectors, whose mammalian homologues are termed smads. There are 9 smads in vertebrates: the receptor-regulated smads 1, 2, 3, 5, and 8 with very similar structures, the common interacting smads 4 and 4␤, and the signal-blocking smads 6 and 7, which form a third discrete subgroup. Upon binding TGF␤1 and activation of T␤Rs, the smad2 or smad3 molecules bound to T␤RI are phosphorylated, dissociate to form a heteromeric complex with smad4 trimers, and translocate to the nucleus where they modify transcription (4) following binding to a palindromic smad binding element (SBE), GTCTAGAC (5), or related sequences containing CA-GAC (6). smad2 and smad3 are not interchangeable because they mediate different responses to TGF␤1 (7). TGF␤1 activation of its receptors leads, by as yet unknown pathways, to activation of a series of signaling pathways, in addition to the smads. These pathways include Ras3 ERK (8), Rho3 JNK (9), RhoA3p160ROCK (10), Tak13p38 MAPK (11,12), protein phosphatase 2A3 S6 kinase (13), and possibly others.
There are at present 51 members of the TGF␤ superfamily with three TGF␤ isoforms found in humans, types 1, 2 and 3, each with a different spectrum of localization in vivo (14). Use of isoform-specific antisera has shown that only the TGF␤1 isoform is associated with cancer development in a wide series of human neoplasias (colon, breast, prostate, glioma, pancreatic, endometrial, gastric, osteosarcomas, etc.) (15). Elevated expression of TGF␤1, but not TGF␤2 or TGF␤3, was significantly correlated with colon cancer progression to metastases. Patients with elevated levels of TGF␤1 protein in their tumor cells were 18 times more likely to experience recurrence of their disease (16). Furthermore, when metastatic cells were compared with their primary site colon cancer, the level of TGF␤1 protein in the primary site tumor was maintained or increased in the metastatic cells in roughly 75% of cases (17). Elevated TGF␤1 levels could mediate increased tumor aggressiveness in several ways, and autocrine TGF␤1 pathways have been shown to mediate increased invasiveness and metastasis in carcinomas (18). Another autocrine response to TGF␤1 is increased proliferation, which has been observed in resected carcinomas in primary culture (19,20) as well as established cell lines such as U9 colon carcinoma cell line. The colon carcinoma cells, which exhibited increased proliferation in response to TGF␤1, have also become more invasive in vitro and in vivo, suggesting that the two biological responses share at least some components of a common TGF␤1-signaling system (21,22). For example, decreasing TGF␤1 protein levels in the metastatic U9 colon cancer cell line by antisense methodology decreased both U9 cell metastasis to the liver and subcutaneous tumor formation in a nude mouse system, and the tumors that did arise had regained TGF␤1 expression (21). TGF␤1 growth stimulation has been documented in various aggressive carcinoma cells, including a subset of colon carcinomas (20 -25), prostate carcinomas (26), and hepatocellular carcinomas (27). We have previously reported that TGF␤1 growth stimulation in colon carcinoma cells is dependent on mutation in K-Ras (22). In the current study we demonstrate that the TGF␤1-initiated signaling pathways, which mediate increased proliferation are Rasdependent and smad-independent.

EXPERIMENTAL PROCEDURES
Materials-TGF␤1 was purchased from R&D Systems. Antibodies to cyclins E, D 1 , and A, smad2, smad4, and rabbit antibody to cdk2 coupled to agarose were purchased from Santa Cruz Biotechnology, whereas antibodies to p21 and p27 were from Transduction Laboratories. The pan-Ras antibody Y13-259 and c-Ha-Ras Ab1 (clone 235-17.1.1), a mouse monoclonal specific for H-Ras, were purchased from Oncogene Science. Polyvinylidene difluoride transfer paper Immobilon-P was purchased from Millipore and polyethyleneimine cellulose F from EM Separations. All radioactive materials were purchased from PerkinElmer Life Sciences, and ECL reagents were from Amersham Biosciences, Inc. Lovastatin was obtained from Merck, Sharp & Dohme Research Laboratories and dissolved in ethanol (4 mg/0.1 ml), made active by addition of 0.15 ml of 0.1 N NaOH and heating for 2 h at 50°C, neutralized with HCl, then brought to a stock solution of 4 mg/ml with distilled water. All other reagents were from Sigma Chemical Co.
Transient Transfections-For transient transfections, cells were plated in 12-well plates at an approximate density of 3 ϫ 10 5 and cultured for 24 h. LipofectAMINE (Amersham Biosciences, Inc.) (5 /well) was mixed with a total of 720 ng of DNA for 30 min in serum-free medium, then added to the cells in serum-free medium for 16 -18 h with or without 5 ng/ml TGF␤1, and luciferase activity was then determined. Transfected DNA samples were composed of 200 ng of reporter construct, 20 ng of the pCMV-gal ␤-galactosidase expression plasmid and, depending on the experiment, 500 ng of test expression plasmid or vector control DNA. Cells were lysed, and both luciferase and ␤-galactosidase activity were measured using a TD-20/20 luminometer. Luciferase activity was normalized to ␤-galactosidase activity for differences in transfection efficiency. Each experiment was performed in triplicate and performed three to four times.
Cell Culture and Growth Assays-The human colon carcinoma cell lines used in this study were maintained in DME medium containing 7% fetal bovine serum, modified, and supplemented as described (24). Effects on cell growth by TGF␤1 were assayed by direct cell counting using a hemacytometer or by incorporation of [ 3 H]thymidine after culture in serum-free ITS (2.5 g/ml insulin, 1.7 g/ml transferrin, 0.1 M selenous acid, 0.29 M linoleic acid, 1 mg/ml fatty acid free bovine serum albumin)-DME. The N17rasH construct developed by L. Feig (28) had been inserted into a pCNC10 expression plasmid behind a CMV promoter by Dr. Fran Kern (University of Alabama) who provided us with the plasmid. This plasmid also carries a neomycin resistance gene behind a CMV promoter. U9 colon carcinoma cells were transfected with this expression plasmid by calcium phosphate precipitation, and putative N17ras transfectants were selected in G418. Transfectants were isolated after transfection with the empty pCNC10 expression plasmid and used as controls. The p21cip1 coding sequence was inserted C-terminal to EGFP (enhanced green fluorescence protein) in the pUC-derived pEGFP-C2 mammalian expression vector (CLON-TECH) and designated p21-EGFP-C2. Apoptosis was measured by binding of fluorescein isothiocyanate-bound annexin V to phosphatidylserine on the outer layer of the plasma membrane (CLONTECH K2027-1) by fluorescence microscopy.
Immunodetection-Following treatment as indicated and washing with cold phosphate-buffered saline, cells were lysed in a buffer con-taining 25 mM Tris (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 20 g/ml leupeptin, 20 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride, 200 M sodium orthovanadate, and 20 mM sodium fluoride. Lysates were pelleted in a microcentrifuge for 15 min to remove insoluble material. Depending on the experiment, 40 -100 g of the cell lysate was blotted onto polyvinylidene difluoride membranes after separation on SDS-PAGE. The blots were blocked in either TBS or TBST buffer (TBS containing 0.05% Tween 20) and 3-5% nonfat dry milk for 2 h, incubated overnight at 4°C with the primary antibody (1 g/ml Ab to cyclin E, cyclin A, H-Ras, cdc25A, p21cip1, or p15; 0.5 g/ml for cyclin D1; 1:2500 dilution of antibody to p27), and proteins were subsequently detected by enhanced chemiluminescence.
Northern Analysis for p21cip1-20 g of total RNA from each cell line was electrophoresed in a 0.8% agarose-formaldehyde gel, transferred to nylon membranes by downward capillary transfer, and UV crosslinked. The membranes were hybridized to a full-length p21cip1 cDNA. Probes were labeled with 32 P by random priming. The blot was hybridized overnight at 42°C with at least 10 7 cpm of the labeled probe, washed at room temperature three times for 7 min with 1ϫ SSC-1% SDS, then washed for 20 min at 52°C in 0.1ϫ SSC-0.1% SDS and autoradiographed. The blots were stripped and rehybridized to glyceraldehyde-3-phosphate dehydrogenase.
Determination of Ras GTP/GDPϩGTP Ratio-Cells were labeled overnight with 200 Ci of 32 P-labeled orthophosphate in serum-free, phosphate-free ITS-DME, then lysed and immunoprecipitated with affinity-purified monoclonal antibody Y13-259, exactly as described (30). After washing the immunoprecipitates, GTP and GDP were eluted with 20 l of 1 M KH 2 PO 4 , pH 3.4, with heating for 3 min at 90°C, then separated by thin layer chromatography on polyethyleneimine cellulose F developed with 1 M KH 2 PO 4 , pH 3.4, followed by autoradiography.
Band Analysis-Immunoblots were scanned using a Lacie Silverscanner DTP and a Power Macintosh 7500, and densitometry was performed using the IP Lab Gel program (Scanalytics).

TGF␤1 Decreases Levels of p21cip1 When It Stimulates
Proliferation-About half of all colon cancers exhibit Ras mutations, with the vast majority in K-Ras, and very rarely, if ever, in H-Ras. Recent results from our laboratory (22) have shown that colon cancer cells can respond to TGF␤1 by growth if they express transfected oncogenic K-Ras proteins, but not transfected oncogenic H-Ras proteins or an equal abundance of transfected wild-type K-Ras proteins. In our earlier studies about 40% of resected colon cancers placed into primary culture, including most metastatic tumors, responded to TGF␤1 by growth stimulation; therefore, the oncogenic K-Ras-expressing cell lines serve as models for this biological response (19,20). The HT29 colon carcinoma sublines U9 and HP1 are not mutated in Ras, however, and yet respond to TGF␤1 by growth and enhanced invasion and metastasis (21,23,31,32). We undertook these studies to determine their molecular basis for growth stimulation.
In our earlier study, TGF␤1 down-regulated the abundance of the cdk inhibitor p21cip1 when it stimulated cell proliferation in colon carcinoma cells expressing transfected oncogenic K-Ras proteins (22), but the functional significance of this reduction was not determined. We now tested the effect of TGF␤1 on p21cip1 abundance in HT29 sublines. The HT29 parental line is not clonal, so the isolated sublines represent cells within the tumor with differing malignant potential. These include the weakly invasive HD3 and HD4 sublines, which respond to TGF␤1 by growth inhibition through a block in phosphorylation of the retinoblastoma protein, and the highly invasive U9 and HP1 sublines, which respond to TGF␤1 by increased proliferation through increased retinoblastoma protein phosphorylation (25). Each of the HT29 sublines displays the same mutations in the colon cancer genes APC and p53 (31).
Two HT29 sublines, one growth-inhibited and the other growth-stimulated by TGF␤1, were treated with a range of TGF␤1 concentrations. TGF␤1 increased p21cip1 protein levels 9-fold when added to HD3 cells at 4 ng/ml, a concentration that inhibits cell proliferation (Fig. 1A). In contrast, TGF␤1 decreased the abundance of p21cip1 to one-third of control levels in U9 cells (Fig. 1A, loading and blotting controls shown below the p21cip1 lanes), similar to its down-regulation of p21cip1 in HD6-K-Ras V12 cells (22). Our initial observation, that the retinoblastoma protein was more highly phosphorylated in U9 and HP1 cells induced to proliferate by TGF␤1 (25), implied that TGF␤1 increased the kinase activity of the cdk2⅐cyclin complexes in these cells. This hypothesis was confirmed when the histone kinase activity of cdk2⅐cyclin complexes was determined by immunoprecipitation of cdk2. Equal amounts of cdk2 were immunoprecipitated (Fig. 1C, lower lanes), but TGF␤1 increased the cdk2 kinase activity of U9 cells while inhibiting the cdk2 kinase activity of HD3 cells (Fig. 1C, one of four experiments with similar results). Analysis of the composition of such complexes revealed that after TGF␤1 treatment there was less p21 associated with cdk2 complexes in U9 cells, whereas more p21cip1 associated with cdk2 complexes in TGF␤1-treated HD3 cells ( Fig. 1B; 150% more lysate was used for the cdk2 immunoprecipitates from U9 cells, so the decrease in p21cip1 levels would be more evident). Addition of a cdk2 peptide during the immunoprecipitation blocked the association of p21cip1 with cdk2 (Fig. 1B). These cdk2⅐cyclin⅐cdk inhibitor compositions are consistent with modulations in histone H1 kinase activity caused by TGF␤1 treatment (Fig. 1C). Thus TGF␤1 stimulated the growth of certain colon carcinoma cells by decreasing the total amount of the cdk inhibitor p21, which in turn led to a decrease in the amount of p21cip1 capable of associating with cdk2⅐cyclin complexes and an increase in the kinase activity of the cdk2⅐cyclin complexes. Such an increase in cdk2⅐cyclin activity by TGF␤1 was the likely cause of the increase in retinoblastoma protein phosphorylation seen in TGF␤1-treated U9 and HP1 cells in our earlier studies (25). The cell growth induced by TGF␤1 was not due to a block in apoptosis. U9 cells were treated with 5 ng/ml TGF␤1 for 2 days, and the percentage of apoptotic cells was determined by binding of annexin V. Both untreated and TGF␤1-treated cultures contained 0.9% apoptotic cells (a total of 3996 cells assayed).
Elevation of p21cip1 Protein Levels Blocks TGF␤1 Growth Stimulation-A longer exposure of the Western blots from U9 cells was needed to detect p21cip1 (Fig. 1A), suggesting that steady-state levels of p21cip1 were lower in these cells. Both U9 and HD3 cells exhibit autocrine regulation by TGF␤1 (32), so it is possible that their endogenous TGF␤1 regulates their steady-state p21cip1 levels. Steady-state levels of cell cycle modulators were compared between HT29 sublines growthstimulated by TGF␤1 (U9, HP1) and sublines growth-inhibited by TGF␤1 (HD3, HD4). Western blotting demonstrated that levels of the cdk inhibitor p21cip1 were markedly lower in both growth-stimulated lines ( Fig. 2A, lower panel, protein loading controls) whereas levels of other cell cycle modulators were not substantially altered (Fig. 2B). The extent of the difference in p21cip1 abundance between these lines varied in different experiments (results from two experiments shown in Fig. 2, A and B) and may reflect differential response to the insulin and insulin-like growth factor 1 found in different batches of fetal calf serum. However, in growth media the level of p21cip1 was always lower in the U9 and HP1 cell lines, which were capable of growth stimulation by TGF␤1. In contrast to the differences in abundance of p21cip1, all four lines exhibit similar steadystate levels of the cdk inhibitors p15 and p27, the tyrosine phosphatase cdc25A, which dephosphorylates cdc2 on regulatory tyrosine residues, and the cyclins D1 and E. A small increase in the steady-state level of cyclin A was seen in both TGF␤1-stimulated cell lines (Fig. 2B), which may simply reflect their increased growth rate.
Elevated expression of p21cip1 blocked the proliferation of U9 cells. An expression plasmid for p21cip1 driven by a CMV promoter and tagged with EGFP was transfected into U9 cells, and G418 selection was attempted. The U9 cells expressing FIG. 1. TGF␤1 modulation of the abundance of the cdk inhibitor p21cip1 in two HT29 sublines: TGF␤1 growth-inhibited HD3 cells and TGF␤1 growth-stimulated U9 cells; association of p21cip1 with cdk2⅐cyclin complexes, and assay of kinase activity. A, log phase cells were treated with 0 -32 ng/ml TGF␤1 in serum-free ITS medium for 54 h before Western blotting for p21cip1. Blotting controls are shown in the lower panels. The decrease in p21cip1 levels by TGF␤1 in U9 cells was seen in each of five independent experiments. B, log phase cells were treated with 5 ng/ml TGF␤1 in serum-free ITS medium for 54 h before immunoprecipitation of cdk2 by anti-cdk2 coupled to agarose, then analysis of cdk2 immunoprecipitates by Western blotting for p21. 1.5-fold as much cell lysate was used for the analysis of cdk2-bound p21cip1 in U9 cells because of the lower expression levels of p21cip1. pep, addition of the cdk2 peptide before immunoprecipitation. Data shown is representative of three separate experiments. C, log phase cells were treated with 5 ng/ml TGF␤1 in serum-free ITS medium for 54 h before immunoprecipitation of cdk2 by anti-cdk2 coupled to agarose, then the kinase activity of the cdk2⅐cyclin complexes was determined by an in vitro kinase reaction on histone H1. The kinase reaction products were separated by SDS-PAGE, and the histone H1 band was detected by autoradiography. The amount of cdk2 in each immunoprecipitate was detected by Western blotting (lowest band). Data shown are one of four experiments with similar results. exogenous p21cip were identified by the co-expressed EGFP protein by fluorescence microscopy. However, after 4 -5 days the vast majority of U9 cells with overexpressed p21cip1 detached from the substratum and then died within the next few days, whereas the control transfectants expressing only the EGFP protein were still viable, attached cells. We concluded that U9 colon carcinoma cells responded to highly elevated levels of p21cip1 by growth arrest and apoptosis as has been reported for other cell types. Therefore, there was no inherent resistance to p21cip1 expression in U9 cells. We then tested whether a more modest elevation of endogenous p21cip1 levels would block TGF␤1-induced cell proliferation. Increasing the level of insulin 16-fold in the serum-free ITS-DME medium caused a 10-fold increase in p21cip1 levels after 24 h (Fig. 2C). TGF␤1 was added for an additional 2 days to half of the cultures, leading to a 4-fold decrease in p21cip1 levels. Parallel cultures were examined for proliferation. TGF␤1 induced a 2-fold increase in proliferation in U9 cells cultured in standard ITS-DME but only a modest 30% increase in proliferation in U9 cells with elevated p21cip1 levels due to insulin treatment (Fig.  2D). Increasing insulin levels did not increase U9 cell proliferation (Fig. 2D, compare lanes 1 with 3), probably because the increase in p21cip1 levels was balanced by increases in other regulatory molecules such as cyclins. However, increasing p21cip1 levels did diminish the growth stimulation by TGF␤1.
Functional, Nonmutated TGF␤ Receptors in TGF␤1 Growthstimulated Cells-We next investigated the TGF␤1 signaling pathways in TGF␤1 growth-stimulated U9 cells and TGF␤1 growth-inhibited HD3 cells. Possibly a mutation in either T␤RI or T␤RII might explain growth stimulation by TGF␤1. This hypothesized mutation could not be in the extracellular TGF␤binding domain because TGF␤ binding was normal (25). Functional TGF␤ receptors had been detected in both TGF␤1 growth-inhibited cells (HD3, HD4) and TGF␤1 growth-stimulated cells (HP1, U9) by [ 125 I]TGF␤1 binding studies (25). All of these cells express equal levels of T␤RI, T␤RII, and T␤RIII mRNA and protein, and these receptors are transported to the cell surface (33). T␤RII and T␤RI were sequenced in three HT29 subclones: TGF␤1-resistant HD6 cells (25), TGF␤1 growth-inhibited HD3 cells, and TGF␤1 growth-stimulated U9 cells, and compared with the published sequences. DNA sequencing was performed on reverse transcriptase-PCR-generated overlapping fragments from bp 220 through 767 of the coding region of T␤RII, which covers the poly(A) tracts mutated in repair-deficient syndromes (34)  Blue staining in the same blot previously probed for p21cip1. B, the same cell lysates were analyzed by Western blotting for the cdk inhibitors p21, p15, and p27, the phosphatase cdc25A, and the cyclins D1, E, and A. C, Western blot for p21cip1. The upper portion of the blot is shown to demonstrate equal loading. U9 cells were cultured in serum-free ITS-DME supplemented with 40 g/ml insulin for 1 day or left unsupplemented (Ϫ lane), then 5 ng/ml TGF␤1 was added to half of the insulin-supplemented cultures for an additional 2 days. D, U9 cells were cultured for 3 days in serum-free ITS-DME supplemented with 40 g/ml insulin or 5 ng/ml TGF␤1 as indicated, then proliferation measured by incorporation of were seen in T␤RI in any of the three lines. In both U9 and HD3 cells TGF␤ receptors were functional, because they phosphorylated smad2 following addition of TGF␤1. A time-course study demonstrated that smad2 was phosphorylated with the same kinetics in HD3 and U9 cells, as shown by Western blotting using phospho-specific antisera, and controlled by blotting for both phosphorylated and unphosphorylated forms (Fig.  3). We concluded that proliferative response to TGF␤1 in U9 colon carcinoma cells was not caused by TGF␤ receptor mutation.
Mutated, Nonfunctional smad4 in TGF␤1 Growth-stimulated Colon Carcinoma Cells-We next investigated whether TGF␤1 activated the remaining elements of the smad signaling pathway in U9 cells, resulting in transcriptional activation. Transient transfection experiments were performed in U9 colon carcinoma cells with transcriptional reporters, which contain TGF␤responsive elements. TGF␤1 activated the p2XSBS/tkCAT construct in melanoma cells whether their response to TGF␤1 was either increased or decreased proliferation (35). TGF␤1 did not activate p2XSBS/tkCAT in U9 cells while inducing a 3-fold activation of this reporter in Mv1Lu cells in parallel experiments (Fig. 4A). This construct includes SBE sequences within elements of the smad-responsive collagen type VII promoter, and its lack of activation in U9 cells indicated that their smad signaling pathway might be aberrant. Similar data was observed with the pCAL2 reporter, which is composed of cyclin A promoter elements (36) (negative data not shown). TGF␤1 did activate the p3TP-lux reporter construct about 5-fold in U9 cells, but activation by TGF␤1 was increased over 200-fold when wild-type smad4 was co-transfected (Fig. 4B). Mutant smad4 expression plasmids encoding either the C-terminal MH2 domain of smad4 (37) or a mutant smad4 unable to bind DNA (100T) (38) could not substitute for smad4 in enhancing activation of 3TP-lux in U9 cells. These smad4 mutants also did not inhibit TGF␤1-induced activation. 3TP-lux is composed of three TGF␤ response elements and a fragment of the plasminogen activator inhibitor-1 promoter and can respond to smad-independent signaling pathways activated by TGF␤1 (39), including JNK (40). These data strongly suggested that smad4 was nonfunctional in U9 cells and that 3TP-lux was activated by TGF␤1 in U9 cells in a smadindependent manner. To resolve this issue we co-transfected pSBE4-BV/luc, a reporter containing four repeats of an 8-bp palindromic SBE, which is known to be activated only by a smad3⅐smad4 complex (5) into U9 cells together with smad4 expression plasmids. The pSBE4-BV/luc reporter was only activated by TGF␤1 when wild-type smad4, but not a dominant negative smad4, was co-transfected (Fig. 4C). Parallel experiments demonstrated that TGF␤1 could not activate a mutant form of this reporter construct, p6MBE, when co-transfected with neither wild-type or mutant smad4 (Fig. 4D). Therefore, endogenous smad4 was nonfunctional in TGF␤1 growth-stimulated cells.
smad4 has been found to be infrequently mutated in benign colon tumors and noninvasive carcinomas, whereas smad4 mu-tations occur in over 30% of metastatic colon carcinomas (38,(41)(42)(43), so we tested the hypothesis that smad4 might be functionally inactivated in metastatic U9 cells by mutation. The U9 cell smad4 gene and the HD3 cell smad4 gene were sequenced from cDNA. Both smad4 genes contained the same mutation, which converted Gln-311 to a stop codon, which would yield a protein of 310 amino acids, with the MH2 domain deleted. Such a mutated smad4 protein would not be able to interact with other smads and probably would not be able to trimerize. A smad4 deletion mutant (⌬274 -321), which encompasses some of the amino acids deleted because of the Q311stop mutation, exhibited virtually no capacity to mediate TGF␤1 activation of the p3TP-lux reporter (44). However, no shorter smad4 form was detected when lysates from U9, HD3, and a third HT29 subline, HD6, cells were examined by Western blotting with an NIH3T3 cell lysate as control (Fig. 5A). Thus the C-terminaldeleted smad4Q311st form appeared to be unstable. Lysates from a colon carcinoma cell line, SW480, which contains an oncogenic mutated K-ras gene, also exhibited no smad4 protein, as has been shown previously (45). smad2 served as the loading and blotting control. Therefore, all TGF␤1 signaling in U9 and HD3 cells must be smad-independent because of the absence of smad4 protein.
We then determined whether TGF␤1 would induce proliferation and decrease p21cip1 levels in other carcinoma cells. Two colon carcinoma cell lines were compared: SW480, which expresses no smad4 protein, and SKCO-1, which does (Fig. 5B). Treatment with TGF␤1 decreased p21cip1 levels in SW480 cells and increased proliferation a modest 43% while affecting neither p21cip1 levels nor proliferation in SKCO-1 cells (Fig.  5C). Induction of proliferation by TGF␤1 in SW480 cells but not in SKCO-1 cells had been shown previously (46).
The Decrease in p21cip1 Levels Requires Activated, Mature Ras Proteins-Both SKCO-1 and SW480 cells express mutated K-ras oncogenes. In the absence of smad4 protein, TGF␤1 was able to down-regulate p21cip1 and stimulate proliferation in SW480 cells. In our earlier study, expression of K-Ras oncoproteins was essential to convert the TGF␤1-insensitive HD6 colon carcinoma cell line to a line capable of growth stimulation in response to TGF␤1 (22). However, one of the novel functions of the K-Ras oncoproteins first demonstrated in this study was to mediate post-translational maturation of T␤RIII, which was incomplete in HD6 cells. U9 cells, in contrast, have functional TGF␤ receptors capable of binding TGF␤1 (25) and transmitting the signal to smad2 (Fig. 3). Perhaps oncogenic K-Ras is needed to mediate T␤RIII maturation, but only wild-type Ras proteins are needed to mediate cell proliferation. Oncogenic Ras proteins are highly activated, with about 50% of Ras proteins binding GTP in untreated SKCO1 colon carcinoma cells, which express oncogenic K-Ras, whereas only about 1% of the Ras proteins in untreated HD3 cells bind GTP (30). Steadystate levels of activated Ras proteins were compared in untreated U9 and HD3 cells by measuring the Ras-bound GTP/ GDP ratio. Total Ras proteins were immunoprecipitated from H 3 [ 32 P]O 4 -prelabeled HD3 and U9 cells, and the Ras-bound GDP and GTP was separated by thin layer chromatography. The percentage of Ras-bound GTP was 5.4-fold more in U9 cells than in HD3 cells (Fig. 6A). Therefore, roughly about 5% of wild-type Ras proteins were activated in untreated U9 colon carcinoma cells and 1% in HD3 cells, similar to data we reported earlier (30). The significance of this Ras activation in U9 cells is unknown, but it may contribute to response to TGF␤1.
Western blotting had demonstrated that TGF␤1 down-regulated p21cip1 levels in U9 cells in a dose-dependent manner (Fig. 1A). The relationship between the presence of mature Ras proteins and p21 abundance was examined further by treating FIG. 3. smad2 is phosphorylated in response to TGF␤1 with the same kinetics in TGF␤1-growth-stimulated U9 colon carcinoma cells and TGF␤1-growth-inhibited HD3 cells. Parallel cultures of HD3 and U9 HT29 colon carcinoma sublines were treated with 5 ng/ml TGF␤1 for 0, 5, 15, and 30 min, then phospho-smad2 and total smad2 proteins were detected by Western blots.
U9 cells with lovastatin, a compound that blocks the posttranslational modification of Ras proteins necessary for their function. Lovastatin inhibits hydroxymethylglutaryl-CoA reductase and thus depletes cells of farnesyl pyrophosphate. It does not diminish total levels of Ras but inhibits both Ras farnesylation and geranylgeranylation (47,48). Both H-Ras  and K-Ras4B are post-translationally farnesylated, whereas K-Ras proteins have also been shown to be geranylgeranylated when cells are treated with a farnesyltransferase inhibitor (reviewed in Ref. 49). Thus lovastatin is predicted to block prenylation of both of these Ras proteins (50). The abundance of p21cip1 was increased in a dose-dependent manner by treatment of U9 cells with increasing concentrations of lovastatin (Fig. 6B), suggesting that inhibition of Ras maturation prevents Ras proteins from mediating p21cip1 down-regulation. Furthermore, lovastatin treatment prevented TGF␤1 from reducing p21cip1 levels in U9 cells (Fig. 6C). Blocking the prenylation of Ras proteins by lovastatin blocked the ability of Ras proteins to mediate the signal from TGF␤1 to down-regulate p21cip1. These studies, taken together, demonstrate that a cell must maintain a threshold level of activated, mature Ras proteins for it to respond to TGF␤1 by reduction of p21cip1 levels, which in turn leads to enhanced cell proliferation.
Dominant Negative Ras Blocks Proliferative Response to TGF␤1-To confirm that Ras proteins were essential for both p21cip1 down-regulation and growth stimulation by TGF␤1, a construct encoding both the H-ras gene bearing a dominant negative S17N mutation and a neomycin resistance gene was transfected into the U9 colon carcinoma cell line, and putative stable transfectants were isolated in G418 by standard methods (51). Overexpression of a dominant negative H-Ras protein inhibits the activity of all Ras proteins. U9 cells were also transfected with the empty pCNC10 expression plasmid, and stable transfectants were isolated. Expression of the N17ras construct was monitored by performing Western blots with a monoclonal antibody specific for H-Ras. The U9-N17ras transfectant exhibited increased expression of H-Ras proteins compared with the control transfectant, U9-N17EV (empty vector) (Fig. 7A), confirming expression of the dominant-negative H-Ras construct. A nonspecific band in the immunoblot serves as an internal control. This construct had earlier been shown to block the TGF␤1 signaling pathways in HD3 colon carcinoma cells that mediate the maturation of ␤1 integrin (51). Therefore, the N17ras proteins were expected to be functional in U9 cells and possibly block some TGF␤1 signaling pathways that require active Ras proteins.
Proliferative response to TGF␤1 was then compared in U9-N17ras transfectant cells, the U9-N17EV empty vector control transfectant, and the U9 parental line. After 2 days of exposure to 4 ng/ml TGF␤1 in serum-free ITS-medium, the U9 parental cells and the U9-N17EV empty vector control transfectant cells exhibited a 2-to 3-fold increase in cell number (Fig. 7B). However, the U9-N17ras transfectant cells exhibited no modulation in cell growth. Therefore, proliferative response to TGF␤1 was dependent on activated, functional Ras proteins. In parallel Western blots, expression of DN17ras was shown to block the down-regulation of p21cip1 caused by TGF␤1, whereas p21cip1 down-regulation was seen in vector controls and in the parental line (Fig. 7C). To confirm that Ras proteins participate in TGF␤1 signaling in U9 cells, increasing concentrations of a DN-N17ras expression plasmid were co-transfected with p3TPlux in U9 cells. DN-N17ras induced a dose-dependent inhibition of activation of p3TP by TGF␤1 (Fig. 7D). Therefore, inhibiting the TGF␤1 to activated Ras pathway with DN-N17ras blocked proliferative response, p21cip1 down-regulation, and activation of the p3TP reporter in U9 colon carcinoma cells. p21cip1 Levels Down-regulated by TGF␤1 by a Post-transcriptional Mechanism-TGF␤1 did not decrease p21cip1 abundance in U9 cells by down-regulating transcription of p21 mRNA. The p21cip1 promoter reporter construct WWP-luc was transfected into U9 cells and as a control into Mv1Lu cells. TGF␤1 activated the WWP-luc reporter about 2-fold in Mv1Lu cells while not activating or inhibiting basal activation levels of this reporter in U9 cells (Fig. 8A). As an internal control, phorbol 12-myristate 13-acetate was shown to activate the p21cip1 promoter reporter in U9 cells showing that the construct was efficiently expressed. These transient transfections studies were confirmed by Northern analysis. A 24-or 48-h exposure to a range of TGF␤1 levels, from 2 to 16 ng/ml, was ineffective in modulating p21 mRNA levels in U9 (Fig. 8B). Therefore, TGF␤1 down-regulation of p21cip1 levels in these colon carcinoma cells must occur by a post-transcriptional mechanism.
smad4 Is Not Sufficient to Restore Regulation of p21cip1 Expression by TGF␤1-Transient overexpression of smad4/ dpc4 has been reported to induce p21cip1 expression in the presence or absence of TGF␤1 (52). Expression of smad4 enabled TGF␤1 to activate the 3TP-lux and pSBE4 promoter FIG. 7. TGF␤1 down-regulates the cdk inhibitor p21cip1 in a rasdependent manner. A, Western blot of total Ras proteins in dominant-negative N17ras transfectant of U9 cells and empty vector control transfectants U9-N17EV. The nonspecific band in the lower panel is loading and transfer control. B, growth assay: parallel cultures of U9 parental cells, DN-N17ras U9 cell transfectants, and U9-N17EV empty vector control transfectants were cultured for 2.5 days with 4 ng/ml TGF␤1 or left untreated. Cell numbers were assayed by direct cell counting in a hemacytometer. C, Western blot showing that 2.5 days of treatment with 5 ng/ml TGF␤1 decreased p21cip1 levels in U9 cells, in empty vector control transfectants U9-N17EV, but not in dominant-negative N17ras U9 cell transfectants. The lower panel is a nonspecific band demonstrating loading and transfer control. D, DN-N17ras blocks TGF␤1 activation of 3TP-lux in a dose-dependent manner. Transient co-transfection of a series of concentrations of the DN-N17ras expression plasmid together with the 3TP-lux promoter reporter for 24 h followed by treatment with 5 ng/ml TGF␤1 or no treatment for 24 h before assay.
reporter constructs in U9 cells (Fig. 4, B and C). However, expression of smad4 did not enable TGF␤1 to activate the p21cip1 promoter reporter in U9 cells (Fig. 8A, last 4

bars).
Thus U9 cells may have additional genetic alterations that limit their ability to activate p21cip1 transcription via the smad pathway. Similar results had been found in SW480.7 cells in which restoration of smad4 expression was not sufficient either to rescue TGF␤1-antiproliferative responses or to allow TGF␤1 induction of p21cip1 (45). DISCUSSION Expression of TGF␤1 protein increases in various neoplasias and is maintained or even increased in many metastatic cancers compared with the primary site tumor (15,17). TGF␤1 is a very potent natural growth inhibitor for epithelial cells, but the vast majority of carcinomas either are not growth-inhibited by TGF␤1 or show modest response. Because of this diminished effect on growth, the selection for increased levels of TGF␤1 protein in tumors in vivo has often been explained by its paracrine effects, such as the induction of angiogenesis through production and secretion of platelet-derived growth factor (53) and vascular endothelial growth factor, and TGF␤'s immunosuppressive functions. Tumor cells themselves were thought to be insensitive to TGF␤, because inactivating mutations in TGF␤ receptors had been found in various tumor types. TGF␤ receptor type II is mutated in a subset of colon cancers (about 15%), which display microsatellite instability (34). The human colorectal cancers with microsatellite instability are weakly aggressive, with a decreased rate of metastasis, and are not likely to be fatal. Therefore, tumors displaying microsatellite instability and mutation in T␤RII and thus not responding to TGF␤1 are not representative of the aggressive tumors in which elevated TGF␤1 protein levels are significantly correlated with disease progression (15). In addition, elevated TGF␤1 levels can modulate tumor aggressiveness through autocrine mechanisms. Expression of DN-T␤RII in EpH4rastransformed mammary epithelial cells and highly metastatic CT26 mesenchymal cells blocked invasion and metastasis, proving the necessity for autocrine TGF␤ signaling in these functions (18). In addition, decreasing TGF␤1 protein levels in the metastatic U9 colon cancer cell line by antisense method-ology decreased both U9 cell metastasis to the liver and subcutaneous tumor formation in a nude mouse system, and the tumors that did arise had regained TGF␤1 expression (21). Thus in 80 -85% of colon cancers and in other cancers that have functional TGF␤ receptors such as the cells in this report, an aggressive cancer phenotype can result from mutations in smads and other signaling pathways that mediate growth arrest, with retention or even activation of other TGF␤ signaling pathways, such as Ras.
Mutation of smad4/dpc4 occurs in over 30% of aggressive, metastatic colon cancers, and in about 50% of pancreatic carcinomas (41,43). Mutations in other smad genes have also been found, but infrequently, less than 5% (54). The smad4 Gln-311 mutation to a stop codon, which we report in U9 and HD3 colon carcinoma cells, has not been found in other tumor types to our knowledge. Other mutations to stop codons have also been found in the MH2 domain of smad4, however, and include 343stop, 358stop, 412stop, and 515stop, each of which blocks the ability of dpc4/smad4 to mediate transcription (55). The 311stop mutation is in a codon conserved in smads 1-4, near the N terminus of the MH2 domain. Because smad4 was mutated and functionally inactive in U9 cells and in SW480 cells that responded to TGF␤1 by proliferation, smad4 is not necessary for this response to TGF␤1. No smad4 protein was detected in another HT29 subline, HD6. Because its derivative line, HD6-K-Ras V12 , was stimulated to proliferate by TGF␤1 (22), we can conclude that this proliferation also occurred in a smad-independent manner. Mitogenic response to TGF␤1 has been seen in several types of aggressive tumor cells (56,57) and has been reported to be caused by oncogenic Ras in human prostate TSU-Pr1 carcinoma cells by a smad-independent pathway (26). We believe that TGF␤1's action as a mitogen is not a cell line phenomenon; our original studies demonstrated that TGF␤1 could stimulate proliferation of primary cultured colon carcinomas, as assayed by direct cell counting as well as [ 3 H]thymidine incorporation (20). TGF␤1 induces both mitogenesis and increased invasion in vivo and in vitro in U9 colon carcinoma cells (32), so it is likely that both TGF␤1 responses are linked and may be mediated by similar pathways.
Little if any smad4 protein was detected in U9, HD3, and FIG. 8. TGF␤1 down-regulates p21cip1 post-transcriptionally. A, TGF␤1 does not inhibit the WWP-luc p21 promoter reporter in U9 colon carcinoma cells in the presence or absence of co-transfected smad4. WWP-luc was transiently transfected without or with (last four bars) co-transfection of a wild-type smad4 expression plasmid, and the plasmids were allowed to express for 16 h. The cells were then divided into two groups in serum-free DME, one of which was treated with 5 ng/ml TGF␤1 for an additional 24 h. Additional parallel U9 cell cultures were not treated with TGF␤1, but with 10 ng/ml phorbol 12-myristate 13-acetate diluted in 0.2% Me 2 SO or the Me 2 SO diluent alone. All experiments were performed four times with triplicate assay points. Luciferase values were normalized to 100% for controls Ϯ S.E. bars. The Mv1Lu data is shown ϫ 0.01 to allow data to be shown on the same scale. B, Northern blots for p21cip1 were performed on parallel U9 cell cultures treated with 0 -16 ng/ml TGF␤1 for 24 and for 48 h. Blots were stripped and reprobed with glyceraldehyde-3-phosphate dehydrogenase.
HD6 colon carcinoma cells, and this may be due to rapid degradation. The L440R mutation in the MH2 domain of smad2 results in a dramatic reduction in steady-state levels, which is due to rapid ubiquitination and proteolysis (58). Recently, oncogenic Ras was shown to induce ubiquitin-proteasome-mediated degradation of smad4 (59). HD3 colon carcinoma cells were also shown to express very little smad4 protein. In prior studies HD3 cells exhibited growth inhibition by TGF␤1 (23,25).Thus, growth inhibition induced by TGF␤1 in HD3 cells must also be smad-independent. TGF␤1 can induce several physiological responses without smad4, including growth inhibition and the induction of fibronectin and several collagens in smad4 knock-out cells, either embryo fibroblasts (39) or MDA-MB 468 cells, which have a homozygous deletion in smad4 (60,61). smad4 is also dispensable for induction of the endogenous PAI-I gene (39), so the 5-fold activation of the p3TP-lux reporter construct by TGF␤1 in smad4-mutant U9 cells (Fig. 4B) is probably due to TGF␤1-mediated activation of TPA-responsive elements added to the 3TP-lux construct to enhance responsiveness (62). Studies are currently underway to determine whether TGF␤1 can initiate colon carcinoma cell proliferation in the presence of wild-type functional smad4.
TGF␤1 has been shown to induce several signaling pathways in addition to smads, one of them the Ras3 ERK pathway (63). We have shown in this study that Ras is essential for TGF␤1 signaling, which results in proliferation by analysis of a DN-N17ras transfectant. Multiple Ras effector pathways are known to contribute to cell cycle progression (64). Ras contributes to cell cycle progression through the Raf-MEK-ERK induction of cyclin D1 (65). However, there is little ERK activation in the latter stages of G 1 when cyclin D1 expression is maximal. Phosphatidylinositol 3-kinase activity is necessary for this induction and for entry into S phase, through activation of Akt/ protein kinase B and the Rho family GTPase Rac (64). Phosphatidylinositol 3-kinase may play a significant role in TGF␤1 signaling through Ras to mediate growth. This report documents that wild-type Ras proteins were at least 5-fold more activated in cells responsive to TGF␤1 by growth stimulation than in cells that did not exhibit this response. Consistent activation of wild-type Ras proteins may be due to the elevated expression of growth factors in many aggressive tumors. U9 cells with wild-type, activated Ras proteins have been shown to express many growth factors, including members of the EGF and fibroblast growth factor families and to respond to these factors by proliferation (53,66). Therefore, any or all of these growth factors may mediate activation of wild-type Ras proteins by autocrine mechanisms in U9 cells. We can conclude that colon carcinoma cells with inactive smad4 may respond to TGF␤1 by proliferation if they exhibit functional TGF␤ receptors and Ras proteins activated either by mutation (HD6-K-Ras V12 cells or SW480 cells) or wild-type Ras proteins activated by autocrine growth factors. In our prior study (22) K-Ras oncoproteins, but not H-Ras oncoproteins, mediated proliferative response to TGF␤1, but the relative roles of wild-type K-Ras and H-Ras proteins have not been evaluated in this response. Oncogenic H-Ras V12 has been shown to block TGF␤1mediated inhibition of RIE-1 (rat/intestinal epithelial) cells through degradation of smad4 but not to induce proliferation (59), so the Ras isoform may be critical.
Down-regulation of certain cdk inhibitors by TGF␤1 has been found by several investigators, including our group, and is likely to contribute to autocrine growth stimulation. Autocrine TGF␤1 accelerated the growth of two hepatocellular carcinoma cell lines, constitutively activated smad2, and suppressed the transcription of endogenous p15ink4B and a p15ink4B promoter construct, which was blocked with neutralizing antibody to TGF␤1 (27). Lower levels of p15ink4B mRNA were seen in lysates of human hepatocellular cancers compared with normal hepatic tissue and were correlated with primarily nuclear expression of smad2, possibly reflecting smad2 constitutive activation by autocrine TGF␤1 in vivo (27). The switch to a spindle cell morphology in epidermal carcinogenesis has also been correlated with the loss of p15ink4B (67). This is functionally similar to our observations that, following stable transfection with an oncogenic cellular K-Ras V12 construct, colon carcinoma cells with wild-type Ras lose cell polarity and become multilayered, invasive, and more aggressive in vivo and down-regulate p21cip1 and the tumor suppressor PTEN (22,68). U9, U9H, and HP1 colon carcinoma cells, all of which were spindleshaped and fibroblastoid and grew in response to TGF␤1 (23,24), exhibited low levels of p21cip1 and down-regulated p21cip1 even further in response to TGF␤1 (this report). Thus downregulation of cdk inhibitors may be a common mechanism by which TGF␤1 stimulates growth, with the selection of the cdk inhibitor being cell type-specific: p21cip1 in colon carcinomas, p15ink4B in hepatocellular carcinomas and squamous cell carcinomas, and p57kip1 in osteoblastic cells (69).