Main

Cell cycle progression from G1 to S phase is governed by CDK4 and the homologous CDK6 as well as CDK2 (ref. 4). CDK4 and CDK6 are activated by D-type cyclins at early to mid-G1 phase, whereas CDK2 is activated by E- and A-type cyclins during late G1 and S phase, respectively4. The activities of CDK4 and CDK6, and of CDK2, are constrained by the p16 INK4 family and the p21 Cip/Kip family inhibitors, respectively4.

TGF-β potently inhibits cell proliferation by causing cell cycle arrest at the G1 phase1,2. Smad proteins can mediate TGF-β growth inhibitory responses1,2. In the basal state, Smad2 and Smad3 are distributed throughout cells. In response to TGF-β, Smad2 and Smad3 are phosphorylated at the carboxyl terminus by TGF-β receptor and form complexes with Smad4 (refs 1, 2). These complexes then accumulate in the nucleus and regulate the transcription of target genes that include cell cycle regulators1,2, such as the CDK inhibitors p15 and p21 and the protooncogene c-myc (refs 7–16). Smad3 is important in antiproliferative responses. For instance, hyperproliferation is a component of the carcinogenic process that leads to the development of metastatic colon cancer in Smad3-/- mice17. The loss of Smad3 expression increases susceptibility to tumorigenicity in human gastric cancer18. Mice lacking Smad3 display squamous hyperplasia in the stomach (C.-X. Deng, personal communication). Smad3-/- mice also show accelerated wound healing, partly owing to an increased rate of re-epithelialization19. A variety of primary cells from Smad3-/- mice are resistant to the growth inhibitory effects of TGF-β (refs 20–22), indicating that Smad3 has a key function in responsiveness to TGF-β.

Because Smads contain potential CDK phosphorylation sites, we analysed whether they are substrates for CDKs. To investigate whether cyclin D–CDK4 can phosphorylate Smads, we performed an in vitro kinase assay with endogenous CDK4 immunoprecipitated from Mv1Lu mink lung epithelial cells by an affinity-purified CDK4-specific antibody. Full-length Smad proteins fused to glutathione S-transferase (GST) were used as substrates. Two frequently used fusion proteins of retinoblastoma protein (Rb) with GST were included as positive controls. As shown in Fig. 1a, the highly homologous Smad2 and Smad3 were phosphorylated. Smad4, which is homologous to Smad2 and Smad3 in the amino-terminal and C-terminal domains but divergent from Smad2 and Smad3 in the middle proline-rich region, was not phosphorylated (Fig. 1a), nor was GST alone (data not shown). The phosphorylation was specific, because preincubation of the antibody with the CDK4 antigen peptide before immunoprecipitation significantly inhibited phosphorylation (Fig. 1a). Because CDK4 has a very strict substrate specificity, indicated by its inability to phosphorylate the canonical CDK substrate histone H1 (ref. 4), we further verified that CDK4 can indeed phosphorylate Smads by using bacterially expressed and in vitro reconstituted CDK4 (Fig. 1b). Notably, Smad3 was phosphorylated to a greater extent than Rb by immunoprecipitated CDK4 from Mv1Lu cells (Fig. 1a) and many other cell lines (data not shown) as well as by reconstituted CDK4 (Fig. 1b). Further substrate titration experiments comparing Smad3 and Rb phosphorylation by using either immunoprecipitated CDK4 or reconstituted CDK4 confirmed that Smad3 is an excellent substrate for CDK4 (Supplementary Fig. S1 and Supplementary Table S1). Similar experiments indicated that Smad3 and Smad2, but not Smad4, can also be specifically phosphorylated by immunoprecipitated CDK2 (Fig. 1c) and bacterially expressed and in vitro reconstituted cyclin E–CDK2 or cyclin A–CDK2 complexes (Fig. 1d).

Figure 1: CDK4 and CDK2 can phosphorylate Smad3 and Smad2 in vitro.
figure 1

a, Top: immunoprecipitated (IP) CDK4 kinase assay. CDK4 immunoprecipitated from 240 µg Mv1Lu cell lysate with 1.2 µg CDK4 antibody in the absence or presence of 9 µg of the antigen peptide (mouse CDK4 amino acids 282–303) was used in a kinase assay with 1 µM substrates. Bottom: protein amounts indicated by Coomassie blue staining. b, Top: reconstituted CDK4 kinase assay. Reconstituted cyclin D–CDK4 (500 ng) was used to phosphorylate 0.4 µM substrates. Bottom: protein level indicated by Coomassie blue staining. c, Immunoprecipitated CDK2 kinase assay. CDK2 immunoprecipitated from 240 µg Mv1Lu cell lysate with 1.2 µg CDK2 antibody in the absence or presence of 6 µg of the antigen peptide (human CDK2 amino acids 283–298) was used in a kinase assay with 1 µM substrates. d, Reconstituted CDK2 kinase assay. Reconstituted cyclin E–CDK2 (13 ng) and reconstituted cyclin A–CDK2 (96 ng) were used to phosphorylate 1 µM substrates.

Full size image

To determine whether endogenous Smad3 is phosphorylated by G1 CDKs, we synchronized Mv1Lu cells at the G0/G1 phase by contact inhibition as described previously23. Cells were then released from growth arrest by plating into fresh medium and, at different time points, were labelled with 32P-orthophosphate and immunoprecipitated by a Smad3-specific antibody to analyse the endogenous Smad3 phosphorylation status. Unlabelled cells prepared in parallel were analysed by flow cytometry to determine the cell cycle distribution at each time point. As shown in Fig. 2a, Smad3 phosphorylation oscillated in a cell-cycle-dependent manner. The maximal phosphorylation of Smad3 occurred at the G1/S junction, indicating that Smad3 was phosphorylated by G1 CDKs. Subsequent immunoprecipitation of the 32P-labelled cell lysates with an antibody against Rb showed that the peak of Rb phosphorylation slightly lagged behind that of Smad3 phosphorylation (Fig. 2a), suggesting that Smad3 is a good substrate for CDK4 in vivo.

Figure 2: Smad3 is phosphorylated by endogenous G1 CDKs in vivo.
figure 2

a, Endogenous Smad3 is phosphorylated in a cell cycle-dependent manner. Mv1Lu cells were synchronized by contact inhibition, then plated into fresh medium, labelled with 32P-orthophosphate, immunoprecipitated (IP) first with a Smad3 antibody and subsequently with an anti-Rb antibody. The ERK activity profile, analysed by a phosphotyrosine antibody that recognizes only activated ERK (pERK1 and pERK2) in unlabelled lysates, indicates that ERK might contribute to Smad3 phosphorylation at very early, but not at the peak, phosphorylation time points. b, Schematic diagram of Smad3 structure. c, Mutation of the Thr 8 and four phosphorylation sites in the Smad3 linker region markedly decreases CDK4 and CDK2 phosphorylation in vitro. WT, wild-type. d, Left, p16 or p21 transfected into Mv1Lu/L17 cells can inhibit endogenous CDK to phosphorylate Smad3. Middle and right, cyclins transfected into COS cells can activate endogenous CDKs to phosphorylate Smad3, and p16 and p21 can inhibit this activity.

Full size image

Smad3 contains nine potential CDK phosphorylation sites, four of which are located in the proline-rich linker region: Thr 178, Ser 203, Ser 207 and Ser 212 (Fig. 2b and Supplementary Fig. S2). The threonine at amino acid position 8 is potentially a good site, particularly for CDK4 (ref. 24). Through a number of mutational studies, we found that simultaneous mutation of Thr 8 and the four sites in the linker, designated Smad3 (T8V/Linker Mut), markedly decreased phosphorylation by CDK4 or CDK2 (Fig. 2c), indicating that CDK4 and CDK2 phosphorylation might occur within these five sites in vitro.

In transient transfection assays, Flag–Smad3 was phosphorylated, and p16 and p21 each inhibited the phosphorylation (Fig. 2d, left panel). Smad3 (T8V/Linker Mut) was phosphorylated to a much lower level than the wild-type Smad3. To determine whether the introduction of exogenous cyclins can activate endogenous CDK to phosphorylate Smad3, the wild-type or mutant version of Flag–Smad3 was cotransfected either individually or together with cyclins D, E or A. As shown in Fig. 2d (middle panel), the phosphorylation of wild-type Flag–Smad3 was significantly increased by cotransfected cyclins. The effects of cyclins D or E can be inhibited by cotransfected p16 or p21, respectively (Fig. 2d, right panel). In contrast, phosphorylation of Smad3 (T8V/Linker Mut) was only slightly increased by cyclins D or E. These observations provide additional evidence that Smad3 is phosphorylated by CDK4 and CDK2 in vivo and that the phosphorylation occurs within these five sites.

To determine the exact CDK phosphorylation sites in Smad3, we generated phosphopeptide antibodies against each of the five potential phosphorylation sites: Thr 8 and the four sites in the linker (Thr 178, Ser 203, Ser 207 and Ser 212). Each of the five sites was phosphorylated by both CDK4 and CDK2 in vitro, and only Thr 8, Thr 178 and Ser 212 were phosphorylated by CDK4 and CDK2 in vivo (Supplementary Figures S3–S5). To confirm that the other four potential sites (Thr 131, Ser 316, Ser 391 and Ser 415) cannot be phosphorylated by CDK, we also generated phosphopeptide antibodies against each of these sites and found that these four sites indeed cannot be phosphorylated by CDK4 or CDK2 (data not shown). Mitogen-activated protein (MAP) kinase was shown to phosphorylate Smad3 in the linker region25. We found that Ser 203 and Ser 207 were phosphorylated by MAP kinase and that Thr 178 was phosphorylated mostly by CDK and to a lesser extent by MAP kinase (Supplementary Fig. S5).

To analyse the role of CDK phosphorylation of Smad3, we examined the effect of mutation of the Smad3 CDK phosphorylation sites on the p15 reporter gene. As shown in Fig. 3a, each of T8V, T178V and S212A has a higher activity than the wild-type Smad3 to stimulate the p15 promoter, and the triple mutant (T8V/T178V/S212A) has the highest activity. Moreover, the GAL4–Smad3 triple mutant (T8V/T178V/S212A) has a higher activity than the wild-type GAL4–Smad3 on a GAL4–luciferase reporter gene in the absence as well as in the presence of different concentrations of TGF-β (Fig. 3b). We also found that cotransfection of CDK4 RNA-mediated interference (RNAi) or CDK2 RNAi constructs increases the basal and TGF-β-induced p15 reporter gene activity (Fig. 3c). Taken together, these results indicate that CDK phosphorylation of Smad3 can inhibit its transcriptional activity.

Figure 3: Mutation of CDK phosphorylation sites in Smad3 leads to an increased p15 level and reduced c-myc expression in reporter gene assays.
figure 3

a, Smad3-/- MEF were transfected with a p15 reporter gene together with the wild-type or various CDK phosphorylation mutants of Smad3. Similar results were obtained in HepG2 cells, and HepG2 cells were used to examine the Smad3 protein expression levels, as shown in the gel at the bottom. Black bars, without TGF-β; hatched bars, with TGF-β. b, Mv1Lu/L17 cells were transfected with a GAL4 reporter gene and TGF-β receptor along with either GAL4–Smad3 (open circles; WT) or GAL4–Smad3 (filled circles; Triple Mut), which contains the T8V, T178V and S212A mutations. c, CDK4 RNAi or CDK2 RNAi increases p15 reporter gene activity as analysed in HepG2 cells. Black bars, without TGF-β; hatched bars, with TGF-β. d, Smad3-/- MEF were transfected with a c-myc reporter gene along with the wild-type or various CDK phosphorylation mutants of Smad3 and analysed as indicated. Black bars, without TGF-β; hatched bars, with TGF-β. Error bars show standard deviation (s.d.) of at least three independent transfection results.

Full size image

Smad3 also plays a critical role in the downregulation of c-myc expression by TGF-β (refs 13–16), which is necessary for subsequent p15 and p21 induction1. Mutation of Smad3 CDK phosphorylation sites also increased its ability to downregulate c-myc (Fig. 3d). Downregulation of c-myc might involve an active repression mechanism, a possibility that requires further investigation.

The above observations prompted us to ask whether Smad3 can inhibit cell cycle progression from G1 to S phase, and whether mutation of the CDK phosphorylation sites renders it more effective in executing this function. Previous studies have shown that Smad3 together with Smad2 and Smad4 activates p15 expression, and introduction of Smad3 into Smad3-/- mouse embryonic fibroblasts (MEF) restores the p15 reporter gene induction by TGF-β (ref. 11). Smad3-/- primary MEF proliferated faster than the wild-type littermate MEF in the 3H-thymidine incorporation assay, and the TGF-β growth-inhibitory effects were largely lost in the Smad3-/- primary MEF (Fig. 4a and ref. 21). To determine whether mutation of the CDK phosphorylation sites enables Smad3 to be more effective at inhibiting G1 cell cycle progression, we generated retroviral vectors that express either the wild-type Smad3 or the various CDK phosphorylation mutants of Smad3, and then used them to infect Smad3-/- primary MEF. As shown in Fig. 4b, wild-type Smad3 increased the cell population in G0/G1 phase and decreased the cell population in S phase. The various CDK phosphorylation mutants of Smad3 augmented these effects. Accordingly, these mutants were more effective than the wild-type Smad3 in inhibiting cell proliferation as measured by the 3H-thymidine incorporation assay (Fig. 4c) and cell number (Fig. 4d), accompanied by increased p15 expression and lower c-Myc levels (Fig. 4e). In addition to the Smad3-/- primary MEF, we have also observed the same trend in other cell types including the Mv1Lu epithelial cells, which contain relatively low levels of wild-type Smad3 (data not shown). Taken together, these observations indicate that phosphorylation of Smad3 by CDK facilitates cell cycle progression from G1 to S phase.

Figure 4: Mutation of the CDK phosphorylation sites in Smad3 leads to increased antiproliferative activities.
figure 4

a, Smad3-/- primary MEF and the wild-type littermate control MEF (both at passage 3) were compared in a 3H-thymidine incorporation assay in the absence (black bars) and in the presence (hatched bars) of TGF-β. The average of four experiments is shown. Error bars represent standard deviation (s.d.). b–e, Smad3-/- primary MEF (passage 3) were infected with retrovirus vector, retroviral wild-type Smad3 or various Smad3 CDK phosphorylation mutants. Infected cells were then split and seeded for fluorescence-activated cell sorting analysis 2 days later (b; black bars, G0/G1 phase; hatched bars, S phase; grey bars, G2/M phase), 3H-thymidine incorporation assay (c; black bars, without TGF-β; hatched bars, with TGF-β), measurement of cell number about 5 days later (d) and analysis of p15 and c-Myc levels (e). The error bars in b–d represent s.d. of four experiments. The infected MEF were treated with TGF-β for 24 h in e. Total RNA (10 µg) and protein (15 µg) were analysed by northern blot (top) and immunoblotting (bottom), respectively.

Full size image

Thus, we have shown that Smad3 is phosphorylated by CDK4 and CDK2. Mutation of its CDK phosphorylation sites increases its transcriptional activity and antiproliferative function. We propose that under physiological conditions, phosphorylation of Smad3 by CDK inhibits its transcriptional activity, contributing to a decreased level of p15 and an increased level of c-Myc, thus facilitating cell cycle progression from G1 to S phase. In the presence of physiological concentrations of TGF-β, normal cells are inhibited by TGF-β. However, cancer cells often contain high levels of CDK activities because of frequent amplification, translocation or overexpression of the cyclin D1 gene or inactivation of the tumour suppressor p16 (refs 5, 6). In addition, overexpression of cyclin E and decreases in CDK inhibitor p27 levels also occur in cancer cells26. Inactivation of Smad3 and presumably the homologous Smad2 by extensive CDK phosphorylation may provide an important mechanism for resistance to the TGF-β growth-inhibitory effects in cancers.

Methods

Phosphorylation in vitro by immunoprecipitated CDK

Mv1Lu cells were lysed in a buffer containing 50 mM Tris pH 7.5, 150 mM NaCl, 0.5% Nonidet P40, 1 mM dithiothreitol (DTT) and protease and phosphatase inhibitors. Antibodies against CDK2 (M2) and CDK4 (C-22) from Santa Cruz Biotechnology were used for immunoprecipitations. The kinase reaction was carried out for 1 h in 30 µl containing 50 mM HEPES pH 7.4, 15 mM MgCl2, 1 mM EGTA, 0.1% Tween 20, 1 mM DTT, 50 µM ATP, 5 µCi [γ-32P]ATP (3000 Ci mmol-1) and substrates at 30 °C. GST–Rb (773–928) contains the proline-rich region, and GST–Rb (379–928) contains in addition the Rb pocket domain.

Phosphorylation in vitro by reconstituted CDK

GST–CDK4, GST–CDK2 and His-tagged cyclins D, E and A were expressed and purified from Escherichia coli as described previously27. GST–CDK and His–cyclin were mixed and incubated with HeLa extracts in the presence of ATP and Mg2+ to activate CDK4 or CDK2. Cyclin D–CDK4 was reconstituted as described previously28 except that no MnCl2 was included. The cyclin D–CDK4 complex was then purified with glutathione agarose beads. Cyclin E–CDK2 and cyclin A–CDK2 were reconstituted as described previously27 and purified by successive Ni-nitrilotriacetate–agarose and glutathione–agarose chromatography. The purified cyclin–CDK complexes were used to phosphorylate substrates for 40 min in 20 µl reactions containing 35 mM HEPES pH 7.4, 10 mM MgCl2, 1 mM EGTA, 0.1% Tween 20, 1 mM DTT, 15 µM ATP and 5 µCi [γ-32P]ATP for CDK4 or 100 µM ATP and 2 µCi [γ-32P]ATP for CDK2 at 30 °C.

Phosphorylation in vivo

For analysis of endogenous Smad3 phosphorylation, Mv1Lu mink lung epithelial cells were synchronized at G0/G1 phase by contact inhibition in complete medium as previously described23. In brief, Mv1Lu cells were grown to full confluence in complete medium. Cells were then split and plated into fresh medium. At different time points, cells were phosphate-starved for 0.5 h and then labelled for 1.5 h with 1 mCi ml-1 32P-orthophosphate. Cells were lysed in a buffer containing 10 mM Tris pH 7.8, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P40, and protease and phosphatase inhibitors, and immunoprecipitated by a Smad3-specific antibody. For analysis of transfected Flag-tagged Smad3 phosphorylation, Mv1Lu/L17 or COS cells were transfected by DEAE-dextran. At 30–36 h after transfection, cells were phosphate-starved for 45 min and labelled with 32P-orthophosphate at 1 mCi ml-1 for 2.5 h followed by immunoprecipitation by a Flag antibody.

Retrovirus production

Wild-type or CDK phosphorylation mutant Smad3 were subcloned into the retroviral vector pLZRSΔ-IRES–GFP29 and transfected into ecotropic phoenix packaging cells to produce retroviruses as described previously30. Smad3-/- MEF were infected at greater than 90% efficiency.

Generation of MEF and 3H-thymidine incorporation assay

Smad3+/- mice were crossed and MEF were generated as described previously21; 2 × 105 Smad3-/- primary MEF and the wild-type littermate control MEF (both at passage 3) were seeded in six-well plates for 24 h, then treated for 24 h with or without 500 pM TGF-β. During the last 4 h, 5 µCi of 3H-thymidine was added to the culture, and 3H-thymidine incorporation was assayed as described previously21.