Pathway- and Expression Level-Dependent Effects of Oncogenic N-Ras: p27^Kip1 Mislocalization by the Ral-GEF Pathway and Erk-Mediated Interference with Smad Signaling

Abstract

Overactivation of Ras pathways contributes to oncogenesis and metastasis of epithelial cells in several ways, including interference with cell cycle regulation via the CDK inhibitor p27^Kip1 (p27) and disruption of transforming growth factor β (TGF-β) anti-proliferative activity. Here, we show that at high expression levels, constitutively active N-Ras induces cytoplasmic mislocalization of murine and human p27 via the Ral-GEF pathway and disrupts TGF-β-mediated Smad nuclear translocation by activation of the Mek/Erk pathway. While human p27 could also be mislocalized via the phosphatidylinositol 3-kinase/Akt pathway, only Ral-GEF activation was effective for murine p27, which lacks the Thr157 Akt phosphorylation site of human p27. This establishes a novel role for the Ral-GEF pathway in regulating p27 localization. Interference with either Smad translocation or p27 nuclear localization was sufficient to disrupt TGF-β growth inhibition. Moreover, expression of activated N-Ras or specific effector loop mutants at lower levels using retroviral vectors induced p27 mislocalization but did not inhibit Smad2/3 translocation, indicating that the effects on p27 localization occur at lower levels of activated Ras. These findings have important implications for the contribution of activated Ras to oncogenesis and for the conversion of TGF-β from an inhibitory to a metastatic factor in some epithelial tumors.

The function and regulation of cyclin-dependent kinases (CDK), cyclins, and CDK inhibitors are an important crossroad for the cellular readout of both transforming growth factor β (TGF-β) and Ras signaling, which have opposite effects on the proliferation of epithelial cells; while Ras signaling induces proliferation and contributes to neoplastic processes (4, 7, 8, 14, 66), TGF-β mediates growth arrest (1, 43, 65). Constitutive activation of Ras signaling pathways can interfere with TGF-β-mediated cell cycle arrest and contribute to tumor promotion and invasiveness (13, 65). We therefore set out in the current work to identify Ras effector pathways that can interfere with the effect of TGF-β on the cell cycle in epithelial cells.

TGF-βs inhibit cell proliferation and suppress the formation of a variety of epithelial tumors (23, 24, 41, 59). TGF-β signals via two receptors, type I (TβRI) and type II (TβRII). These receptors and the co-Smad (Smad4) act as tumor suppressor genes in various tumor types, including colorectal cancers, carcinoma, and T-cell lymphoma (9, 23, 33, 41, 42, 53). TGF-β mediates phosphorylation of TβRI by TβRII, followed by phosphorylation of the R-Smads (Smad2 and -3) by TβRI (2, 43, 65). Smad2/3 associate with Smad4 and accumulate in the nucleus to regulate gene transcription. Growth arrest by TGF-β occurs via interference with cell cycle progression. Depending on the cell type, TGF-β inhibits proliferation by suppressing the expression of c-Myc, cyclin A, Cdc25A, and CDK4/6 (13, 15, 17, 29, 55, 73) and by inducing the CDK inhibitors p15^Ink4B (p15) and p21^Waf1/Cip1 (12, 26, 28). In Mv1Lu mink lung epithelial cells, p15 induction is prominent; p15 releases p27^Kip1 (p27) from CDK4/6, enabling it to inhibit CDK2 (56, 58, 64); the cyclin E-CDK2 activity and the ensuing hyperphosphorylation of the retinoblastoma protein are necessary for transition from G₁ to S phase (27, 64). p27 is a key intermediate in cell cycle arrest and is regulated by degradation (10, 40, 45, 52, 68, 75) and by translocation to the cytoplasm (19, 20, 30, 36, 38, 60, 67, 74). Shifting p27 to the cytoplasm partitions it away from the nuclear CDK2 and correlates with impairment of p27 function (3, 32, 38, 51, 63).

Expression of oncogenic Ras or overactivation of Ras signaling pathways has been linked to loss of TGF-β growth inhibition (13, 34, 35, 38, 50, 61, 62). This may convert the TGF-β response to promotion of metastasis (13, 35, 49). However, the mechanisms by which Ras overactivation disrupts growth arrest by TGF-β are not fully understood, and multiple pathways may be involved. Thus, it was shown that activated protein kinase B/Akt can induce cytoplasmic mislocalization of human p27 by phosphorylating it on Thr157 (36, 67, 74). However, we found (38) that murine p27, which lacks Thr at position 157, is also mislocalized to the cytoplasm by activated Ras. This mislocalization, accompanied by loss of TGF-β growth inhibition, suggests that another Ras-activated pathway may be involved. Moreover, some studies of cells expressing high levels of constitutively active Ras reported that Erk activation can directly inhibit TGF-β-mediated Smad nuclear translocation (34, 61). However, other studies, including ours, showed impairment of TGF-β growth arrest upon low-level expression of activated Ras in the absence of effects on the Smad response (31, 35, 38).

We therefore hypothesized that the differences in these reports may reflect distinct effects of oncogenic Ras on p27 localization and on Smad translocation via different Ras effector pathways, as well as different dependences of these effects on its expression levels. Our results point to a novel mechanism, besides the Akt-mediated phosphorylation of human p27, whereby oncogenic N-Ras mislocalizes both murine and human p27 via the Ral-guanine nucleotide exchange factor (Ral-GEF) pathway. This effect, which is sufficient to abrogate TGF-β antiproliferative activity, is manifested even at relatively low expression levels of activated N-Ras. At high expression levels, oncogenic N-Ras also hampers the nuclear translocation of Smad2/3 in response to TGF-β, an effect which is mediated by the Mek/Erk pathway. Thus, constitutively active N-Ras can impair the cytostatic function of TGF-β via distinct pathways exhibiting different sensitivities to the Ras expression level.

MATERIALS AND METHODS

Reagents.

Recombinant TGF-β1, recombinant platelet-derived growth factor-BB (PDGF) and epidermal growth factor (EGF) were from R&D Systems. The phosphatidylinositol 3-kinase (PI3K) inhibitor LY294002 and the Mek inhibitor U0126 were purchased from Calbiochem. The bromodeoxyuridine (BrdU) labeling kit with anti-BrdU antibodies (catalog no. 93-3943) was from Zymed Laboratories. Rabbit immunoglobulin Gs (IgGs) against Smad3 (reactive with Smad3 and Smad2; sc-8332), human p27 (sc-528), Erk2 (sc-154), and Akt1/2 (sc-8312) were obtained from Santa Cruz Biotechnology. Mouse monoclonal antibody against phospho-Erk1/2 (M8159) was from Sigma, anti-phospho-Akt(Ser473) (catalog no. 9271) was from Cell Signaling, and affinity-purified rabbit IgG against the hemagglutinin (HA) epitope tag (anti-HA; A190-108A) was from Bethyl Laboratories. Mouse monoclonal pan-Ras antibody (Ab-3) and anti-N-Ras (Ab-1) were obtained from Calbiochem. Affinity-purified biotinylated goat anti-rabbit (GαR) IgG, biotinylated goat anti-mouse (GαM) IgG, and Cy3-streptavidin were from Jackson ImmunoResearch. Peroxidase-coupled GαM and GαR IgGs were from Dianova.

Plasmids.

Constitutively active human N-Ras^61K in the pMX-IRES-GFP1.1 retroviral vector and GFP-K-Ras^12V in pEGFP-C3 were described by us earlier (38, 48). Human N-Ras^61K and its effector loop mutants N-Ras^61K/35S, N-Ras^61K/40C, and N-Ras^61K/37G in pcDNA3 and in pZIP-NeoSV(X)1 (described in reference 76), as well as pBABE-puro with a cDNA insert encoding HA-tagged Rlf fused to the plasma membrane-targeting sequence of K-Ras (Rlf-CAAX) (77), were a generous gift from C. J. Der and A. D. Cox. Murine green fluorescent protein (GFP)-p27 was constructed from N-terminally HA-tagged p27 in pCG-N-BL (a gift from J. Kato) (72) by subcloning the BamHI-HindIII fragment into BglII-HindIII-digested pEGFP-C1 (Clontech). Human HA-p27 in pcDNA3 (60) was a gift from M. Pagano, and dominant-negative RalA (Ral^28N) in pCAG(SB) (22) was a gift from L. A. Feig. The pGex-2TH vector for bacterial expression of the glutathione S-transferase (GST)-fused Ras binding domain (RBD) of Raf-1, mutated (A85K) to enhance Ras binding (18), was kindly supplied by A. Burgess.

Tissue culture and transfection.

Mv1Lu cells and HEK-293T cells (American Type Culture Collection) were grown as described earlier (38, 74). For immunofluorescence and BrdU incorporation assays, subconfluent cells plated on glass coverslips placed in 35-mm dishes were transfected with 3 μg DNA (total of cotransfected plasmids, supplemented to 3 μg with empty vector where needed) using jetPEI (Polyplus Transfection). For Western blotting and RBD pull-down assays, cells were grown in 10-cm plates and transfected with a total of 7 to 10 μg DNA.

Immunofluorescence microscopy.

Cells grown on glass coverslips were transfected as described above, and after the appropriate time and treatments (see the figure legends), were fixed by 4% paraformaldehyde in phosphate-buffered saline (PBS) (30 min; 22°C) and permeabilized with Triton X-100 (0.2%; 5 min). After being blocked (30 min) with 200 μg/ml goat γ-globulin in Hanks' balanced salt solution without sodium bicarbonate and phenol red supplemented with 20 mM HEPES (pH 7.2) and 2% bovine serum albumin (BSA), they were labeled successively with various antibodies (see the figure legends) in the same buffer. All incubations with antibodies were for 45 min at 22°C, with three extensive washes after each step. Cells were mounted with Prolong Antifade (Molecular Probes), and fluorescence digital images were recorded with a charge-coupled device camera as previously described (21).

BrdU incorporation.

Mv1Lu cells were seeded for 1 day on glass coverslips (65,000 cells/35-mm dish) and cotransfected with transfection marker (pEGFP) and a six-fold excess of N-Ras^61K, one of its effector loop mutants, or empty vector. After 24 h, they were incubated with or without TGF-β1 (10 pM; 24 h; 37°C). The cells were incubated with BrdU (1:100 dilution from the Zymed BrdU labeling kit) for another 24 h, fixed with 4% paraformaldehyde to preserve GFP fluorescence under the disruptive treatments required for BrdU antibody labeling, and permeabilized with 0.2% Triton X-100. The cells were washed with PBS, blocked (10 min) with TBST (50 mM Tris, pH 7.4, 100 mM NaCl, 0.1% Tween 20) containing 1% BSA, and treated sequentially with 2 N HCl and 0.1 M sodium borate, pH 8.5 (10 min; 22°C), to enable BrdU immunostaining. After being blocked with goat γ-globulin, they were labeled successively in TBST-BSA as described above with the following antibodies: (i) mouse anti-BrdU (Zymed kit; 1:50 dilution), (ii) biotinylated RαM IgG (5 μg/ml), and (iii) Cy3-streptavidin (1.2 μg/ml). Transfected cells were identified by GFP fluorescence and scored for nuclear BrdU labeling.

Immunoblotting.

Cells seeded at 10⁶/10-cm dish were grown for 24 h and serum-starved for another 24 h. They were then incubated (12 h; 37°C) with LY294002 (20 μM) or U0126 (50 μM) and stimulated (5 min) with 100 ng/ml PDGF or EGF (see Fig. 3). They were washed with cold PBS and lysed on ice (30 min) with lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM EDTA, 1 mM EGTA, 1.5 mM MgCl₂, 5 μg/ml leupeptin, 1 mM benzamidin, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml pepstatin, 5 units/ml aprotinin, 0.1 mM sodium orthovanadate). After centrifugation to remove cell debris, samples were boiled (5 min) in sodium dodecyl sulfate (SDS) sample buffer containing dithiothreitol and analyzed by SDS-polyacrylamide gel electrophoresis (PAGE). After electrotransfer, the blots were blocked with TBST containing 5% milk powder (1 h at 22°C for anti-phospho-Akt, 12 h at 4°C for all other primary antibodies) and incubated with anti-phospho-Akt (1:1,000; 12 h at 4°C in TBST-5% BSA) or anti-phospho-Erk (1:10,000; 1 h at 22°C in TBST). After three 10-min washes (22°C) with TBST, the blots were incubated (1 h; 22°C) with peroxidase-GαM or peroxidase-GαR IgG (1:10,000 in TBST) and washed extensively, and adsorbed antibodies were detected by enhanced chemiluminescence (ECL) (Amersham). To probe for total Akt and Erk, the blots were stripped with boiling 0.1 M glycine, pH 2.5 (two 10-min washes), and reprobed with anti-Akt (1:1,000) or anti-Erk (1:1,500), followed by peroxidase-coupled secondary antibodies and ECL as described above.

FIG. 3.

Ral^28N blocks p27 cytoplasmic mislocalization by N-Ras^61K, while Mek inhibition restores TGF-β-mediated Smad2/3 nuclear translocation. (A and B) Pharmacological inhibitors are effective in blocking growth factor stimulation in Mv1Lu cells. After 24 h in serum-free medium, cells were incubated with LY294002 (20 μM) or U0126 (50 μM) for 12 h. They were then stimulated (100 ng/ml; 5 min) with PDGF (A) or EGF (B), lysed, and subjected to 12.5% SDS-PAGE, loading 100 μg protein/lane for Akt and 60 μg/lane for Erk analysis. Electrotransfer was followed by immunoblotting with anti-Akt or anti-Erk antibodies to determine the total levels of these proteins or with antibodies specific to phospho-Akt (P-Akt) and phospho-Erk (P-Erk) to determine the activated forms. Visualization was by ECL. PDGF induced a 3.5- ± 0.8-fold increase in P-Akt, and EGF induced a 7.0- ± 1.2-fold increase in P-Erk (n = 3 in both cases); both effects were totally blocked by the respective inhibitors LY294002 and U0126. (C) Interference of inhibitors with N-Ras^61K effects on p27 and Smad2/3 localization. Mv1Lu cells were cotransfected with murine GFP-p27 together with an excess (sixfold) of empty pcDNA3 vector (Ctrl) or N-Ras^61K in pcDNA3. To assess the effects of dominant-negative Ral^28N (DN Ral), cells were triply transfected with GFP-p27 together with both N-Ras^61K and Ral^28N at DNA ratios of 1:6:6, respectively. After 24 h, the cells were incubated for 16 h with LY294002 (20 μM) or U0126 (50 μM); control samples and cells transfected with Ral^28N were left untreated. The cells were then incubated with TGF-β1 (100 pM; 20 min; 37°C), fixed/permeabilized, and labeled for Smad2/3 as in Fig. 2. Typical images of GFP-p27 (green) and Smad2/3 (Cy3; red) are shown. In the Smad2/3 images (bottom row), arrows indicate the transfected cells, identified by GFP-p27 fluorescence (top row). Bar, 20 μm. (D and E) Quantification of p27 and Smad2/3 localization. Cells were scored for nuclear versus cytoplasmic localization of GFP-p27 as in Fig. 1. The Smad2/3 localization results shown were obtained by counting cells exhibiting cytoplasmic GFP-p27, as described in the legend to Fig. 2D. Qualitatively similar, albeit somewhat smaller, effects of N-Ras^61K on Smad2/3 nuclear translocation in response to TGF-β were observed in the entire population of transfected cells (not shown). The bars are means plus standard errors of the mean of three or four samples in each case, scoring 100 cells per sample.The asterisks indicate a significant difference (Student's t test) from the control (*, P < 10⁻⁴; **, P < 2 × 10⁻⁵).

Determination of the levels of Ras and of activated Ras.

The levels of total Ras proteins were determined by immunoblotting. Cells seeded in 10-cm dishes were transfected with N-Ras^61K expression vectors and a smaller amount (one-sixth of total DNA) of pEGFP (see Fig. 6). After 24 h, they were lysed in buffer containing 0.5% NP-40 (48). Lysates (30 μg protein) were subjected to 12.5% SDS-PAGE, followed by immunoblotting using mouse pan-Ras antibody (1:2,500; 1 h at 22°C) or anti-N-Ras (1:20; 1 h), followed by peroxidase-GαM IgG (1:10,000) and ECL. To determine the levels of activated Ras (Ras-GTP), cell lysates (500 μg protein) were precipitated by glutathione-Sepharose beads coupled to GST-RBD, prepared as described previously (18). The precipitated Ras-GTP was dissolved in SDS-PAGE sample buffer and subjected to SDS-PAGE and immunoblotting as described above.

FIG. 6.

Different dependences of p27 mislocalization and Smad2/3 nuclear translocation on the expression level of N-Ras^61K. (A) Ras expression levels in Mv1Lu cells transfected with N-Ras^61K in pcDNA3 and in retroviral vectors. Mv1Lu cells were transfected with empty vector [pcDNA3 or one of the retroviral vectors pMX-IRES-GFP1.1 or pZIP-NeoSV(X)1; Ctrl], with N-Ras^61K in pcDNA3, or with N-Ras^61K in one of the retroviral vectors (designated pMX or pZIP); a small amount of pEGFP (one-sixth) was included. The cells were lysed 24 h later. Transfection efficiencies were similar (∼40% expressing cells), as evaluated from the percent cells expressing GFP (by fluorescence microscopy) prior to lysis. Aliquots of the lysates (30 μg protein) were subjected to SDS-PAGE, followed by immunoblotting with pan-anti-Ras antibody (left), to determine total Ras. To determine the level of GTP-bound activated Ras, aliquots (500 μg protein) were subjected to the GST-RBD pull-down assay (see Materials and Methods), followed by SDS-PAGE and immunoblotting with pan-anti-Ras. The immunoblots were visualized by ECL. The data shown are representative of three independent experiments. Transfection with N-Ras^61K in pcDNA3 or in a retroviral vector resulted in 5.0- ± 0.2-fold and 2.1- ± 0.2-fold increases in activated Ras, respectively. Similar results (right) were obtained when blotting was performed with anti-N-Ras antibodies. (B) Typical images of p27 (GFP; green) and Smad2/3 (Cy3; red) localization. Mv1Lu cells were cotransfected with GFP-p27 in pEGFP, together with a sixfold excess of either empty vector (pcDNA3 or pMX; Ctrl), N-Ras^61K in pcDNA3, or N-Ras^61K in the pMX retroviral vector. N-Ras^61K in the pZIP retroviral vector yielded results similar to those obtained with the pMX vector (Fig. 7A). After 24 h, the cells were incubated with TGF-β1 (100 pM; 20 min; 37°C), fixed/permeabilized, and subjected to labeling of Smad2/3 as in Fig. 2. In the Smad2/3 images, arrows indicate the transfected cells, identified by GFP-p27 fluorescence. Bar, 20 μm. (C and D) Quantification of p27 and Smad2/3 localization. The percentage of cells with nuclear Smad2/3 was scored in cells with cytoplasmic p27 localization, as described in Fig. 2D. Similar, although somewhat smaller, effects were found for Smad2/3 in the entire population of transfected cells (data not shown). The bars are means plus standard errors of the mean of three or four samples (100 cells per sample) in each case. The asterisks indicate a significant difference from the control (P < 1.5 × 10⁻⁶ in panel C; P < 3 × 10⁻⁴ in panel D).

RESULTS

Oncogenic N-Ras induces cytoplasmic mislocalization of p27 via the Ral-GEF pathway.

In our previous work (38), we demonstrated that stable expression of low levels of constitutively active N-Ras^61K in mink lung epithelial cells (Mv1Lu) induces mislocalization of p27 to the cytoplasm, sequestering p27 and CDK2 in different compartments and disrupting TGF-β-mediated growth arrest. The interference with the TGF-β response occurred at the level of p27 localization and interactions, as the upstream TGF-β signaling (including Smad nuclear translocation and transcriptional activation) were unaffected (38). Interestingly, the Ras-induced mislocalization was observed with murine p27 (38), which lacks the Thr157 Akt phosphorylation site reported to induce cytoplasmic mislocalization of human p27 (36, 63, 67, 74). This prompted us to study the effects of oncogenic N-Ras^61K and its effector domain mutants (66, 76) on the localization of murine and human p27 in transiently transfected Mv1Lu cells. In these effector loop mutants, the effector-specific mutation is introduced into the background of constitutively active N-Ras^61K, leading to selective activation of one downstream signaling pathway: the Mek/Erk (N-Ras^61K/35S), the PI3K (N-Ras^61K/40C), or the Ral-GEF (N-Ras^61K/37G) pathway. In accord with our previous results (38), transient expression of N-Ras^61K (which activates the full repertoire of Ras effectors) impaired the nuclear localization of transfected murine and human p27 constructs (Fig. 1A to C), as well as of endogenous human p27 (Fig. 1D). This is in line with our former demonstration that N-Ras^61K also mislocalized endogenous p27 in Mv1Lu cells (38). Interestingly, selective activation of the Ral-GEF pathway by Ras^61K/37G produced the same effect. In contrast, the effector loop mutant activating Mek/Erk (N-Ras^61K/35) failed to mislocalize either murine or human p27 (Fig. 1), suggesting that this pathway is not involved. Activation of PI3K (by N-Ras^61K/40C) was as potent as N-Ras^61K in mislocalizing human p27 but had only a small effect on murine p27 (Fig. 1); this partial effect may be mediated indirectly through Ral, since PI3K activation was shown to enhance the catalytic activity of Ral-GEF by promoting its association with PDK1 (70). Importantly, endogenous human p27 exhibited the same pattern of response to the different N-Ras^61K effector loop mutants as the overexpressed tagged p27 constructs (Fig. 1D). These findings demonstrate a major role for the Ral-GEF pathway in the cytoplasmic mislocalization of both human and murine p27 by constitutively active N-Ras. The absence of the Akt target site on murine p27, rendering it sensitive to Ral-GEF activation while eliminating Akt-dependent effects that may complicate the analysis, makes the Mv1Lu/murine p27 system highly suitable to study the effects of the Ral-GEF pathway on p27 mislocalization and its consequences.

FIG. 1.

Activation of the Ral-GEF pathway by N-Ras^61K/37G induces mislocalization of both murine and human p27. Mv1Lu cells (A to C) or HEK-293T cells (D) were cotransfected (see Materials and Methods) with (A and B) murine GFP-p27 in pEGFP and an excess (sixfold) of empty pcDNA3 vector (Ctrl), N-Ras^61K, or one of its effector domain mutants in pcDNA3. (C) Murine GFP-p27 was replaced by human HA-p27 in pcDNA3. (D) pEGFP (transfection marker) replaced murine GFP-p27. After 24 h, the cells were fixed with 4% paraformaldehyde. For immunofluorescent labeling (C and D), the paraformaldehyde-fixed cells were permeabilized with 0.2% Triton X-100 prior to being immunostained. Fluorescence images wererecorded by a charge-coupled device camera. Bar, 20 μm. (A) Typical images of murine GFP-p27 localization. (B) Quantification of murine GFP-p27 localization. (C) Quantification of human HA-p27 localization. To label HA-p27, the cells were incubated successively with (i) rabbit anti-HA (2 μg/ml), (ii) biotinylated GαR IgG (5 μg/ml), and (iii) Cy3-streptavidin (1.2 μg/ml). (D) Quantification of the localization of endogenous human p27. Endogenous p27 in HEK-293T cells was labeled by successive incubations with (i) rabbit anti-p27 (1.25 μg/ml), (ii) biotinylated GαR IgG (5 μg/ml), and (iii) Cy3-streptavidin (1.2 μg/ml). The bars are means plus standard errors of the mean of five to seven samples in each case, scoring 100 transfected cells per sample for nuclear and cytoplasmic localization of p27. The asterisks indicate significant differences from the control (*, P < 0.003; **, P < 10⁻⁶; Student's t test). Mainly nuclear localization is evident for the control and for the Mek/Erk-activating mutant N-Ras^61K/35S. N-Ras^61K/37G, which activates Ral-GEF but not the Mek/Erk or the PI3K pathway, was as effective as constitutively active N-Ras^61K in mislocalizing both murine and human p27. In contrast, the PI3K-activating mutant N-Ras^61K/40C was highly effective in mislocalizing human HA-p27 but had a much weaker effect on the localization of murine GFP-p27. Control experiments (not shown) on Mv1Lu cells transfected with dominant-negative H-Ras^17N in pRSV (54) or N-Ras^17N in pEGFP showed no mislocalization of murine p27 and even a slight increase (to 86 to 92%) in the percentage of cells with nuclear p27.

Impairment of Smad2/3 nuclear translocation in response to TGF-β and of p27 localization is mediated through distinct Ras-activated pathways.

Translocation of Smad to the nucleus is a key early step in TGF-β signaling. In view of the controversy regarding the ability of oncogenic Ras to interfere with the translocation of Smads to the nucleus in response to TGF-β (31, 34, 35, 38, 61), we studied the effects of N-Ras^61K and its effector loop mutants on this parameter. Typical results are shown in Fig. 2A and B, and data derived from several experiments integrating data from a few hundred cells are depicted in Fig. 2C and D. These studies demonstrate that TGF-β1 induces nuclear accumulation of Smad2/3 in the vast majority of the control (mock-transfected) cells; this accumulation is significantly reduced by the fully active N-Ras^61K and by the N-Ras^61K/35S mutant, which selectively activates the Mek/Erk pathway, but not by the other effector loop mutants (Fig. 2A to C). Unlike Smad translocation, the cellular localization of murine GFP-p27 followed in the same cells was not affected by TGF-β, and cytoplasmic mislocalization of p27 was mediated only by the Ral-GEF-activating mutant N-Ras^61K/37G (Fig. 2A and B), reinforcing the notion that the effects of constitutively active N-Ras on p27 localization and on Smad translocation in response to TGF-β are mediated independently by separate Ras-activated pathways.

FIG. 2.

Impairment of TGF-β-induced Smad2/3 nuclear translocation by N-Ras^61K is mediated via the Mek/Erk pathway. Mv1Lu cells were cotransfected with murine GFP-p27 and N-Ras^61K mutants in pcDNA3 as in Fig. 1. In the control (Ctrl), empty pcDNA3 replaced the N-Ras construct. After 24 h, the cells were incubated without (A) or with (B) TGF-β1 (100 pM; 20 min; 37°C), fixed/permeabilized, and processed for immunofluorescence (see Materials and Methods). They were labeled successively with (i) rabbit anti-Smad3, reactive with both Smad3 and Smad2 (5 μg/ml); (ii) biotinylated goat anti-rabbit IgG (5 μg/ml); and (iii) Cy3-streptavidin (1.2 μg/ml). Bar, 20 μm. The white arrows in the Smad2/3 images indicate the transfected cells (which also show GFP-p27 fluorescence). As shown in the arrow-marked cells in panel B, in the absence of TGF-β, Smad2/3 was mainly cytoplasmic. TGF-β failed to induce Smad2/3 nuclear translocation in cells transfected with constitutively active N-Ras^61K and the N-Ras^61K/35S mutant but not with the other mutants. (C) Quantification of Smad2/3 localization in the entire transfected-cell population. Transfected cells were identified by GFP-p27 fluorescence, and Smad2/3 nuclear localization was scored in the GFP-expressing cells. (D) More pronounced effects are observedin cells displaying cytoplasmic p27 localization. Only transfected cells exhibiting cytoplasmic GFP-p27 were scored. The results shown are with TGF-β1; in the absence of ligand, the results were similar to those in panel C (not shown). In both panels C and D, the percentages of cells with predominantly nuclear Smad2/3 were derived by scoring 100 cells per sample in four or five samples. The bars are means plus standard errors of the mean. The asterisks indicate a significant difference from the control in the presence of TGF-β (P < 0.0003, except for N-Ras^61K/35S in panel D, which yielded a P value of <0.0008).

Since the expression level of N-Ras mutants among the transfected cells is subject to significant variations, we examined the effects of activated N-Ras mutants on Smad translocation in the subpopulation of cells exhibiting cytoplasmic p27. This selects for cells expressing activated N-Ras at levels that are sufficient at least for induction of p27 mislocalization. Indeed, in this subpopulation, the effects of both N-Ras^61K and N-Ras^61K/35S on Smad2/3 localization after exposure to TGF-β1 were more pronounced (Fig. 2D). We therefore employed this protocol for further studies of TGFβ-mediated Smad2/3 nuclear translocation. The enhanced effect on Smad translocation in this subpopulation raises the possibility that high expression levels of activated Ras are required for its interference with Smad translocation in response to TGF-β, a notion supported by our studies of cells expressing different levels of N-Ras^61K or the loop mutants (see Fig. 5 and 6). Differences in Ras expression levels therefore provide a plausible explanation for the different reports on the ability of activated Ras to impair TGF-β-mediated Smad translocation (31, 34, 35, 38, 61).

FIG. 5.

Growth arrest by TGF-β is disrupted by N-Ras mutants that mediate either p27 mislocalization or impairment of Smad2/3 nuclear translocation. Mv1Lu cells were cotransfected with GFP (pEGFP vector, as a marker for transfected cells) and an excess (sixfold) of empty pcDNA3 vector (Ctrl), N-Ras^61K, or one of the effector domain mutants in pcDNA3. After 24 h, the cells were incubated without (A) or with (B) TGF-β1 (10 pM; 24 h; 37°C). They were then subjected to BrdU incorporation and immunostaining (resulting in Cy3-labeled BrdU), as detailed in Materials and Methods. The arrows in the BrdU images (red) indicate transfected cells, identified by GFP fluorescence (green). Bar, 20 μm. (A and B) Typical images of BrdU incorporation. (C) Quantification of BrdU incorporation in Mv1Lu cells transiently expressing Ras effector loop mutants. Transfected cells (identified by GFP expression) were scored for nuclear BrdU labeling. The bars are means plus standard errors of the mean of four samples in each case, scoring 100 cells per sample. Inhibition of BrdU incorporation by TGF-β1 was highly significant (marked by asterisks) in the control cells (P < 2 × 10⁻⁶) and in cells expressing the N-Ras^61K/40C mutant (P < 0.005) but was completely lost in cells expressing N-Ras^61K, N-Ras^61K/37G, or N-Ras^61K/35S (P > 0.1; comparing pairs of samples without and with TGF-β).

To further validate the involvement of the specific pathways identified by the N-Ras effector loop mutants in mediating p27 mislocalization and in disrupting Smad2/3 nuclear translocation, we examined the ability of specific inhibitors of Ras-activated pathways to impair the effects exerted by constitutively active N-Ras^61K. The PI3K/Akt and the Mek/Erk pathways were inhibited by LY294002 and U0126, respectively, and the Ral-GEF pathway was inhibited by overexpression of Ral^28N, a dominant-negative Ral mutant (22). The effectiveness of the pharmacological inhibitors in the Mv1Lu cell system is shown in Fig. 3, where incubation with LY294002 blocked the elevation in phosphorylated Akt following stimulation with PDGF (Fig. 3A), while U0126 blocked the EGF-induced increase in Erk phosphorylation (Fig. 3B).

We next examined the abilities of the specific inhibitors to restore p27 nuclear localization or TGF-β-mediated Smad2/3 nuclear translocation in Mv1Lu cells overexpressing N-Ras^61K. Cells were either doubly transfected to coexpress murine GFP-p27 and N-Ras^61K (for studies with the pharmacological inhibitors) or triply transfected to also express Ral^28N (for studies of the effects of dominant-negative Ral). At 24 h posttransfection, the doubly transfected cells were incubated (16 h) with LY294002 or U0126, while cells transfected with Ral^28N and control samples were incubated only with medium. All samples were then incubated with TGF-β1 (100 pM) and assayed for the cellular localization of GFP-p27 and Smad2/3. As shown in Fig. 3C and D, only Ral^28N (but not the PI3K/Akt or Mek/Erk inhibitors) restored the nuclear localization of GFP-p27 to a level similar to that of the control (60% of the triply transfected cells) but had no effect on TGF-β-mediated Smad2/3 translocation. In contrast, only the Mek inhibitor U0126 was capable of restoring TGF-β-mediated Smad2/3 accumulation in the nucleus to the levels observed in control cells, while either LY294002 or expression of Ral^28N was completely ineffective (Fig. 3C and E). Mislocalization of p27 is also induced by constitutively active K-Ras, since coexpression of GFP-K-Ras^12V with murine HA-p27 (using the same conditions as in Fig. 3C) resulted in cytoplasmic mislocalization of the latter (the percentage of cells with nuclear HA-p27 went down from 80% ± 5% to 37% ± 6%; n = 4). This is in accord with our earlier report on the induction of p27 mislocalization by constitutively active H- and K-Ras (38). Importantly, the K-Ras^12V effect is mediated through the Ral-GEF pathway, as demonstrated by its reversal upon inclusion of dominant-negative Ral^28N, along with GFP-K-Ras12V, in the transfection (nuclear p27 = 68% ± 6%; n = 4).

To confirm the hypothesis that the effects of activated Ras on p27 localization involve the Ral-GEF pathway and are distinct from its effects on Smad translocation, we examined the ability of Rlf-CAAX, a constitutively activated Ral-GEF generated by fusing Rlf to the CAAX membrane-targeting sequence of K-Ras (77), to mislocalize p27 or to interfere with the induction of Smad translocation by TGF-β. Mv1Lu cells were cotransfected with murine GFP-p27, together with Rlf-CAAX (or pBABE-puro with no insert as a control), incubated with TGF-β1 to promote Smad nuclear translocation, and assayed for the cellular localization of GFP-p27 and endogenous Smad2/3. The results (Fig. 4) demonstrate that Rlf-CAAX induces p27 mislocalization as effectively as N-Ras^61K or N-Ras^61K/37G but has no detectable effect on Smad nuclear translocation in response to TGF-β. Taken together with the results of the experiments with the N-Ras^61K effector loop mutants and the pathway-specific inhibitors, these findings demonstrate that p27 cytoplasmic localization and impaired TGF-β-mediated nuclear translocation of Smad2/3 are two independent effects of constitutively active Ras, exerted via activation of distinct downstream pathways.

FIG. 4.

Rlf-CAAX induces p27 mislocalization but does not affect Smad2/3 nuclear translocation. Mv1Lu cells were cotransfected with murine GFP-p27 (in pEGFP) together with a sixfold excess of HA-Rlf-CAAX (in pBABE-puro) or empty pBABE-puro vector (Ctrl). After 24 h, the cells were incubated with TGF-β1 (100 pM; 20 min; 37°C), fixed/permeabilized, and labeled for Smad2/3 as in Fig. 2. (A) Typical images of GFP-p27 and Smad2/3. Smad2/3 is endogenously expressed in all the cells, while Rlf-CAAX and GFP-p27 are expressed only in the transfected cells, identified by GFP-p27 fluorescence (marked by arrows in the Smad2/3 images). Bar, 20 μm. (B and C) Quantification of p27 and Smad2/3 localization. The percentage of cells with nuclear Smad2/3 was scored in cells with cytoplasmic p27 localization, as in Fig. 2D. Similar effects (not shown) were obtained for Smad2/3 in the entire population of transfected cells. In the absence of TGF-β, only 4 to 6% of the cells exhibited nuclear Smad2/3 labeling, both in mock-transfected and in Rlf-CAAX-transfected cells. The bars are means plus standard errors of the mean of three or four samples (100 cells per sample). The asterisks indicate a significant difference from the control (P < 1.5 × 10⁻⁶).

Interference with either p27 nuclear localization or Smad translocation disrupts TGF-β-mediated growth arrest.

In view of reports that Ras overactivation can interfere with TGF-β-mediated growth arrest (13, 34-36, 38, 50, 61, 62, 67, 74), we examined the abilities of the different N-Ras^61K effector loop mutants (constitutive activation of different Ras effector pathways) to abrogate TGF-β-induced growth arrest, measured by inhibition of BrdU nuclear incorporation. Mv1Lu cells were cotransfected with GFP (as a transfection marker), together with N-Ras^61K or one of its effector domain mutants. At 24 h posttransfection, the cells were incubated with TGF-β1 (10 pM; 24 h), followed by incubation with BrdU for an additional 24 h. The cells were then fixed and permeabilized under conditions that preserve GFP fluorescence and immunostained for BrdU (see Materials and Methods). Typical images are shown in Fig. 5A and B, and the averaged data derived from several experiments are depicted in Fig. 5C. As shown in the figure, TGF-β strongly attenuated BrdU nuclear incorporation in the control cells. N-Ras^61K and each of the effector loop mutants, N-Ras^61K/37G (activating Ral-GEF) or N-Ras^61K/35S (activating the Mek/Erk pathway), completely abolished the TGF-β-induced inhibition of BrdU incorporation. In contrast, N-Ras^61K/40C (activating PI3K/Akt) was ineffective in reversing the effect of TGF-β on BrdU incorporation (Fig. 5). Since N-Ras^61K/37G induces p27 mislocalization (Fig. 1) but does not interfere with TGF-β-mediated Smad2/3 translocation (Fig. 1 and 2D) and N-Ras^61K/35S disrupts Smad translocation but not p27 localization (Fig. 1 and 2D), we conclude that the effects of the Mek/Erk and the Ral-GEF Ras effector pathways are independent of each other and that overactivation of one of these pathways is sufficient to interfere with TGF-β-induced growth arrest in Mv1Lu epithelial cells.

Low expression levels of N-Ras^61K induce p27 mislocalization and block TGF-β growth arrest without interfering with TGF-β-induced Smad translocation.

The impairment of TGF-β-mediated Smad2/3 nuclear translocation and growth arrest by transient expression of N-Ras^61K or N-Ras^61K/35S in pcDNA3 (Fig. 2, 3, and 5) differs from our earlier demonstration that low-level expression of N-Ras^61K by a retroviral vector mislocalizes p27 and impairs TGF-β growth inhibition without affecting TGF-β-mediated Smad2/3 translocation and transcriptional responses (38). Since protein expression is controlled primarily by the promoter region of the vector, this difference could be due to different expression levels of the Ras proteins, depending on the expression vector (16). We therefore compared the expression levels of N-Ras^61K in Mv1Lu cells, using either pcDNA3 (expression driven by the cytomegalovirus promoter) or retroviral vectors [pMX-IRES-GFP1.1 or pZIP-NeoSV(X)1, with the Moloney murine leukemia virus long terminal repeat promoter]. As shown in Fig. 6A, transfection with N-Ras^61K in pcDNA3 resulted in a large increase (fivefold) in the level of activated (GTP-bound) Ras, while transfection under similar conditions with N-Ras^61K in a retroviral vector led only to a modest increase (about twofold) in activated Ras. In both cases, the increase in the level of total Ras compared to mock-transfected cells was significantly less (1.2-fold). The effects of the different expression levels of N-Ras^61K on p27 localization and on Smad2/3 nuclear translocation are shown in Fig. 6. Cytoplasmic mislocalization of p27 was effectively mediated by either high (transfection with pcDNA3) or low (transfection with a retroviral vector) expression of N-Ras^61K (Fig. 6B and C). In contrast, disruption of TGF-β-mediated Smad2/3 nuclear translocation required high expression levels of N-Ras^61K and was not observed following transfection with N-Ras^61K in a retroviral vector (Fig. 6B and D).

To further investigate the dependence on the expression level of activated N-Ras, we studied the effects of low expression levels of N-Ras^61K effector loop mutants [in pZIP-NeoSV(X)1] on p27 localization, Smad 2/3 nuclear translocation in response to TGF-β, and TGF-β-induced growth arrest (Fig. 7). Under these conditions, the only effector domain mutant that was effective was N-Ras^61K/37G (activating Ral-GEF). This mutant induced p27 mislocalization to the same extent as fully active N-Ras^61K (Fig. 7A) and completely blocked TGF-β-mediated growth arrest as measured by BrdU incorporation (Fig. 7C). On the other hand, none of the effector loop mutants (as well as N-Ras^61K) had any effect on the ability of TGF-β to induce Smad2/3 nuclear translocation (Fig. 7B). These results demonstrate that relatively low expression levels of activated N-Ras are sufficient to induce p27 mislocalization via the Ral-GEF pathway independent of the Mek/Erk pathway, while disruption of Smad2/3 translocation requires higher expression levels. This is in accord with our earlier demonstration (38) that this p27 mislocalization interferes with the ability of p27 to bind to and inhibit nuclear cyclin E-CDK2 and is sufficient to disrupt growth inhibition by TGF-β, although the earlier steps of TGF-β signaling, including Smad nuclear translocation and Smad-mediated transcriptional activation, are intact.

FIG. 7.

Selective impairment of p27 localization and TGF-β-mediated growth arrest, but not Smad nuclear translocation, by low levels of N-Ras^61K effector loop mutants. Mv1Lu cells were cotransfected with GFP-p27 (A and B) or GFP (C) in pEGFP, together with an excess of one of the N-Ras^61K effector loop mutants [in pZIP-NeoSV(X)1] as in Fig. 6B to D. The bars are means plus standard errors of the mean of three or four samples in each case, scoring 100 cells per sample. (A and B) Quantification of p27 and Smad2/3 localization. Twenty-four hours posttransfection, cells were incubated with 100 pM TGF-β1 (20 min) and assayed for p27 and Smad2/3 localization as in Fig. 6B to D. The asterisks indicate significant differences from the control (P < 2 × 10⁻⁴). (C) TGF-β-mediated inhibition of BrdU incorporation. Twenty-four hours posttransfection, cells were incubated with TGF-β1 (10 pM; 24 h), followed by a BrdU incorporation assay as in Fig. 5. Inhibition of BrdU nuclear incorporation by TGF-β1 was significant (asterisks) in the control (cotransfection with an empty pZIP vector) and in cells expressing N-Ras^61K/40C or N-Ras^61K/35S (P < 0.005), but was completely lost (P > 0.2) upon expression of N-Ras^61K or N-Ras^61K/37G.

DISCUSSION

Expression of constitutively active Ras in epithelial tumor cells contributes to their oncogenesis and metastatic potential (7, 14, 49, 50, 65, 66). The oncogenic effect of constitutively active Ras is enhanced by its ability to alter the cellular readout to growth-inhibitory extracellular stimuli, interfering with their ability to induce cell cycle arrest (13, 36, 38, 67, 74). In this context, expression of oncogenic Ras or overactivation of Ras signaling pathways interferes in many cases with the inhibition of cell cycle progression by TGF-β, compromising its function as a tumor suppressor (13, 34, 35, 38, 50, 61, 62). Moreover, since the ability of TGF-β to induce epithelial-to-mesenchymal transition and to promote extracellular-matrix production is not reduced by expression of activated Ras, it can lead to enhanced metastasis in response to TGF-β (13, 35, 38, 49, 50, 62). In spite of its relevance to tumorigenesis and metastatic processes, the interference of Ras overactivation with TGF-β-induced growth arrest is still not fully understood, and numerous different mechanisms have been proposed to explain these phenomena (11, 31, 34, 35, 38, 39, 61). In the present work, we show that expression of activated (oncogenic) N-Ras can disrupt the cytostatic effects of TGF-β via distinct pathways that display highly different dependences on the Ras expression level. We demonstrate that activation of the Ral-GEF pathway, which occurs even at low expression levels of activated N-Ras, mediates cytoplasmic mislocalization of p27 and is sufficient to disrupt TGF-β antiproliferative activity (Fig. 1 and 3 to 7). The ability of this pathway to induce such effects is novel and may be relevant to the emerging importance of the Ral-GEF pathway in oncogenesis (25, 37). When expressed at higher levels, oncogenic N-Ras also interferes with the nuclear translocation of Smad2/3 in response to TGF-β, an effect that is mediated independently via activation of the Raf/Mek/Erk pathway (Fig. 2 to 7).

Due to its pivotal role in cell cycle progression, p27 is subject to multiple layers of regulation. In proliferating cells, p27 is found mainly in association with cytoplasmic CDK4 and/or CDK6 (27, 64). Anti-proliferative signals, such as cell-cell contact and TGF-β, mediate p27 partner switching, i.e., the release of p27 from CDK4/6, binding to CDK2, and induction of growth inhibition (64, 65, 73). In order to mediate cell cycle arrest by inhibition of CDK2, p27 must localize to the nucleus, and cytoplasmic mislocalization of p27 can have profound effects on cell cycle regulation (3, 32, 38, 51). Several reports point to the importance of the phosphorylation status of p27 as a determinant of its subcellular localization. Phosphorylation of Thr187 by CDK2 and of Ser10 by human interacting kinase stathmin are thought to induce nuclear export of p27 (6, 45, 46), mediated by Jab1 and CRM1 (71) and/or by interaction with mNPAP60 (47). Phosphorylation of Thr198 by Akt was also reported to induce cytoplasmic mislocalization of p27 (19). Recently, Thr157, which is located at the center of the putative bipartite nuclear localization signal motif of human p27, was proposed as a target for Akt in human breast cancer cells (36, 67, 74). Phosphorylated Thr157 was subsequently shown to impair the binding of p27 to α3 and α5 importins, in addition to promoting the association of p27 with cytoplasmic 14-3-3 (63). However, our earlier (38) and current (Fig. 1 and 3 to 6) studies demonstrate that Ras-activated pathways other than the PI3K/Akt pathway can effectively mislocalize p27 to the cytoplasm. This is suggested by the finding that not only human p27, but also its murine counterpart, which has Ala instead of Thr at position 157, are mislocalized to the cytoplasm by expression of N-Ras^61K (38) (Fig. 1 and 3), its Ral-GEF-activating effector loop mutant N-Ras^61K/37G, or the constitutively active Ral-GEF Rlf-CAAX (Fig. 1, 3, and 4). These results establish a novel role for the Ral-GEF pathway in the regulation of the subcellular localization of p27 and its effects on the cell cycle. The ability of the Ral-GEF pathway to regulate p27 and the cell cycle gains additional support from several recent reports. (i) Ral activation by N-Ras^61K was shown to promote anchorage-independent growth of a human fibrosarcoma cell line, HT1080, by enhancing the degradation of p27 (78); we have formerly shown (38) that p27 is mislocalized in this N-Ras^61K-expressing human cell line and that its nuclear localization is restored upon treatment with a Ras inhibitor. (ii) Oncogenic Ras was shown to enhance the downregulation of p27 in RIE-1 intestinal epithelial cells via a pathway independent of Raf or PI3K activation (57). (iii) It was recently demonstrated that the Ral-GEF pathway, and not the Raf or PI3K pathway, is critical for Ras transformation in human cells (25, 37). It should be noted that cytoplasmic mislocalization of p27 may have additional effects aside from disrupting the inhibitory interactions of p27 with CDK2. Thus, interactions of p27 with cytoplasmic Grb2 can inhibit Grb2-SOS complex formation (44), and Grb2 can also downregulate p27 (69). A shift of p27 to the cytoplasm can also affect cell migration by regulating RhoA activation (5).

Importantly, the cytoplasmic mislocalization of p27 by oncogenic Ras is sufficient to impair TGF-β-mediated growth inhibition (38) (Fig. 5 and 7). This is demonstrated by the disruption of TGF-β-induced growth arrest upon expression of an effector loop mutant (N-Ras^61K/37G) activating only the Ral-GEF pathway, which does not affect Smad nuclear translocation in response to TGF-β (Fig. 5). It is further reinforced by the loss of the TGF-β effect on the cell cycle in cells expressing low levels of either N-Ras^61K or the Ral-GEF-activating mutant N-Ras^61K/37G, since these low expression levels induce p27 mislocalization without affecting Smad translocation (Fig. 6 and 7). As we have shown previously (38), the interference of N-Ras^61K with the TGF-β response in Mv1Lu cells expressing low levels of N-Ras^61K occurs at the level of p27 localization and interactions, since the upstream signaling of TGF-β (including Smad nuclear translocation and transcriptional activation) is not affected (38).

At higher expression levels, activated N-Ras also interferes with the TGF-β-induced Smad2/3 nuclear translocation (Fig. 2 and 3). This effect is mediated by another pathway (the Raf/Mek/Erk pathway), as indicated by the use of N-Ras^61K effector loop mutants, activated Rlf-CAAX, and inhibitors of specific Ras-activated pathways (U0126 to inhibit the Mek/Erk pathway, LY294002 to inhibit the PI3K pathway, and dominant-negative Ral^28N to block Ral-GEF signaling) (Fig. 2 to 4). Thus, constitutively active N-Ras can disrupt TGF-β-mediated growth arrest by two independent mechanisms: induction of cytoplasmic mislocalization of p27 (via the Ral-GEF pathway for both murine and human p27 and via the PI3K/Akt pathway for human p27) and interference with Smad nuclear translocation via the Raf/Mek/Erk pathway. The latter effect is in line with former studies of cells overexpressing high levels of constitutively active H-Ras, where Erk activation was reported to directly inhibit Smad translocation and response (34, 61). Our finding that this effect is not detected at relatively low expression levels of activated Ras explains other observations, where low-level expression of constitutively active Ras was found to abrogate TGF-β signaling in spite of normal Smad translocation and transcriptional response (31, 35, 38).

In summary, the findings reported here demonstrate that constitutively active N-Ras can affect p27 localization and TGF-β-mediated growth inhibition via several distinct pathways and that these effects depend differentially on the expression level of activated Ras. We suggest that the amount of oncogenic Ras expressed in the cell and the extent of constitutive activation of specific Ras-dependent pathways are important determinants of the cellular response to TGF-β. Thus, in tumor cells exhibiting relatively low levels of oncogenic Ras, the major effect of Ras on TGF-β growth inhibition is exerted via p27 mislocalization. At higher expression levels, both p27 localization and the translocation of Smad proteins in response to TGF-β are compromised.