Cdc2 and Cdk2 Kinase Activated by Transforming Growth Factor-β1 Trigger Apoptosis through the Phosphorylation of Retinoblastoma Protein in FaO Hepatoma Cells*

The signaling pathway leading to TGF-β1-induced apoptosis was investigated using a TGF-β1-sensitive hepatoma cell line, FaO. Cell cycle analysis demonstrated that the accumulation of apoptotic cells was preceded by a progressive decrease of the cell population in the G1 phase concomitant with a slight increase of the cell population in the G2/M phase in response to TGF-β1. TGF-β1 induced a transient increase in the expression of Cdc2, cyclin A, cyclin B, and cyclin D1 at an early phase of apoptosis. During TGF-β1-induced apoptosis, the transient increase in cyclin-dependent kinase (Cdk) activities coincides with a dramatic increase in the hyperphosphorylated forms of RB. Treatment with roscovitine or olomoucine, inhibitors of Cdc2 and Cdk2, blocked TGF-β1-induced apoptosis by inhibiting RB phosphorylation. Overexpression of Bcl-2 or adenovirus E1B 19K suppressed TGF-β1-induced apoptosis by blocking the induction of Cdc2 mRNA and the subsequent activation of Cdc2 kinase, whereas activation of Cdk2 was not affected, suggesting that Cdc2 plays a more critical role in TGF-β1-induced apoptosis. In conclusion, we present the evidence that Cdc2 and Cdk2 kinase activity transiently induced by TGF-β1 phosphorylates RB as a physiological target in FaO cells and that RB hyperphosphorylation may trigger abrupt cell cycle progression, leading to irreversible cell death.

TGF-␤1 1 is a multifunctional cytokine playing critical roles in many cellular processes, including cellular growth, differentiation, and morphogenesis (1,2). Although TGF-␤1 has classically been shown to arrest growth at the G 1 phase of the cell cycle (3), it has more recently been demonstrated to play an important role as an inducer of apoptosis in several cell types, including primary hepatocytes (4), hepatoma cell lines (5,6), prostate epithelial cancer cells (7), ovarian carcinoma (8), and leukemic cells (9). Although the down-regulation of Bcl-2 (8,10), induction of c-fos and c-jun genes (8), and the involvement of caspase family protease(s) (11,12) have each been postulated to play a role in TGF-␤1-mediated apoptosis, the signaling pathways of TGF-␤1-induced apoptosis are still largely unknown.
In eukaryotes, the cell cycle is coordinated by several protein kinases composed of a cyclin-dependent kinase (Cdk) subunit(s) and its corresponding regulatory cyclin subunit(s), and cyclin-dependent kinase inhibitors (13,14). The cell cycle machinery controlling TGF-␤1-mediated growth arrest is relatively well understood. G 1 cyclins, cyclin-dependent kinases, and cyclin-dependent kinase inhibitors are the mediators of the TGF-␤1 effect on the G 1 /S transition (reviewed in Ref. 3). The G 1 cell cycle events that have been shown to be mediated by TGF-␤1 in epithelial cells include an inhibition of Cdc2 synthesis, phosphorylation levels, and kinase activity (15,16); a reduction of Cdk4 synthesis (17), and an inhibition of mRNA expression of Cdk2, Cdk4, CKShs1, cyclin E, and cyclin D (18 -20). Additionally, TGF-␤1 has been reported to induce the expression of the cyclin-dependent kinase inhibitors p27 kip1 , p21 cip1 , and p15 ink4 and to enhance the association of these cyclin-dependent kinase inhibitors with the relevant Cdk complexes (21)(22)(23). Down-regulation of either Cdk2 activity or cyclin A expression by TGF-␤1 has been used as a general indicator of the TGF-␤1-mediated growth inhibitory effect (24 -26). In epithelial cells, TGF-␤1 prevents RB phosphorylation (27,28), retaining RB in a hypophosphorylated state that may suppress progression into the S phase. These effects on RB are postulated to be derived from negative effects of TGF-␤1 on Cdk activity toward RB (17,29). In contrast, cell cycle components controlling TGF-␤1-induced apoptosis have never been explored extensively.
Recently, apoptosis has been hypothesized to be the result of abnormal cell cycle control (30). The overexpression of cyclins (31)(32)(33) and activation of cyclin-dependent kinases have been shown to correlate with the onset of apoptosis in many experimental systems (34 -40). However, the physiological target(s) of these kinases during apoptosis have not been elucidated. The oncogene product Bcl-2 effectively protects cells from apoptotic cell death (41). Recently the inhibition of cell death by Bcl-2 has been linked to the slowdown of cell cycle progression (30,42,43). However, the detailed action mechanism of Bcl-2 for deathsparing activity still remains to be clarified.
We demonstrate here that TGF-␤1 induces apoptosis in FaO rat hepatoma cells through the distinctive controlling mechanism of cell cycle components in FaO cells, which is contradic-tory to the mechanism for TGF-␤1-induced G 1 cell cycle arrest that was previously reported. In addition, we present the evidence for the first time that RB is a functional substrate of Cdc2 and Cdk2 kinase transiently activated by the signal of TGF-␤1 to trigger apoptosis. Moreover, we show that overexpression of Bcl-2 protein protects the cells from TGF-␤1-induced cell death by blocking the induction of Cdc2 mRNA and the subsequent activation of Cdc2 kinase to phosphorylate RB.

MATERIALS AND METHODS
Reagents-Antibodies against Cdk2, Cdk4, Cdk6, cyclin B, and cyclin D1 were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Antibody against p27 was purchased from Transduction Laboratories (Lexington, KY). Antibodies against p21, Cdc2, and cyclin A were from Oncogene Research Products (Cambridge, MA). Antibody against phosphospecific RB was provided by New England Biolabs, Inc. (Beverly, MA). Histone H1 was purchased from Sigma. Recombinant RB-N terminus fusion protein containing RB residue 181-400 and RB-C terminus fusion protein containing 701-928 were obtained from New England Biolabs (Beverly, MA). Roscovitine, olomoucine, and Pansorbin were purchased from Calbiochem-Novabiochem Corp. (San Diego, CA). [␥-32 P]ATP and enhanced chemiluminescence (ECL) reagents were obtained from Amersham Pharmacia Biotech, and the protein assay reagents were from Bio-Rad. All other reagents and compounds were analytical grades.
Nuclear Staining with Hoechst 33258 -FaO cells were split onto coverslips and incubated for 1 day prior to TGF-␤1 treatment. Cells were treated with TGF-␤1 for the indicated times and fixed with 4% paraformaldehyde (pH 7.4) for 10 min, washed with phosphate-buffered saline (PBS) three times, and then incubated for 10 min in 10 g/ml Hoechst 33258 in PBS. The change of nuclear morphologies was examined by fluorescence microscopy.
DNA Fragmentation Analysis-Total DNA was extracted from the cells at 0, 12, and 24 h after TGF-␤1 treatment. Apoptosis was characterized by DNA fragmentation using TACS TM apoptotic DNA laddering kit (Trevigen, Gaithersburg, MD).
TUNEL Assay-FaO cells were split onto the coverslips and incubated for 1 day prior to TGF-␤1 treatment. Cells were treated with TGF-␤1 for the indicated times and fixed with 4% paraformaldehyde (pH 7.4) for 10 min. TUNEL assay of fragmented DNA was performed as recommended by the manufacturer (Roche Molecular Biochemicals).
Cell Cycle Analysis-DNA content was assessed by staining ethanolfixed cells with propidium iodide and monitoring by FACScan. Cell distribution was determined with a ModFit LT program (Verity Software House, Inc).
Northern Blot Analysis-Total cellular RNA was prepared from FaO cells treated with TGF-␤1 for different incubation times using RNAzol TM B (Tel-Test, Inc., Friendwood, TX), according to the recommendations of the manufacturer. mRNA was prepared from total RNA by use of oligo(dT)-cellulose resin (Life Technologies, Inc.). mRNA (10 g) was loaded onto 1.2% formaldehyde-agarose gel. DNA probes for hybridization were labeled using a random priming kit (Amersham Pharmacia Biotech) and [␣-32 P]dCTP.
Co-immunoprecipitation-FaO cells were treated with TGF-␤1 for different incubation times, and 1 ϫ 10 7 cells were lysed in 0.4 ml of lysis buffer A (1% Nonidet P-40, 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 10 g/ml aprotinin, 10 g/ml leupeptin, 20 mM NaF, 1 mM Na 3 VO 4 , 10 g/ml PMSF) at 4°C for 15 min. Cell lysates were centrifuged at 13,000 rpm for 15 min, and the supernatant was collected. Protein concentration was determined, and 2 mg of protein was used for each co-immunoprecipitation. Immunoprecipitations were performed by incubating lysates with 10 g of the anti-RB, anti-Cdc2, or anti-Cdk2 antibody for 2 h at 4°C. Immune complexes were collected by incubating with 60 l of Pansorbin for 1 h and washed three times with ice-cold buffer A. The pellet was resuspended in 2ϫ SDS-PAGE sample buffer (125 mM Tris-HCl (pH 6.8), 4% w/v sodium dodecyl sulfate, 20% glycerol, 14.4 mM ␤-mercaptoethanol) and boiled for 5 min, and centrifuged at 13,000 rpm for 5 min. The supernatant was collected and subjected to 10% SDS-PAGE. Proteins were transferred to nitrocellulose membranes using standard electroblotting procedures. The presence of co-immunoprecipitated protein was confirmed by Western blotting using specific antibodies.
Immunocomplex Kinase Assay-Cells were collected at different incubation times with TGF-␤1 and washed in PBS. Cells were lysed in buffer A for 15 min on ice. Cell lysates were cleared by centrifugation at 13,000 rpm for 15 min. Protein concentrations were quantitated by Bio-Rad assay. A total of 500 g of protein was used for each immunoprecipitation. Cdc2, Cdk2, Cdk4, and Cdk6 in cell extracts were incubated with a specific antibody (1 g/reaction) for 3 h at 4°C. 15 l of protein A/G-garose (Oncogene Research Products, Cambridge, MA) was added into the mixture, which was then further incubated for 1 h. Immune complexes were centrifuged at 2,500 rpm for 5 min and the precipitates were washed three times with buffer A and twice with kinase buffer (50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM DTT). Cdk kinase assays on histone H1 were performed by mixing the respective immune complexes with 5 g of histone H1 and 10 Ci of [␥-32 P]ATP in 30 l of kinase buffer. Cdk kinase assays on RB were carried out in the same way using 5 g of recombinant RB-N terminus fusion protein or RB-C terminus fusion protein. The kinase reaction was performed at 30°C for 30 min and then terminated with 2ϫ SDS-PAGE sample buffer. The reaction mixtures were resolved by SDS-polyacrylamide gel electrophoresis analysis. Gels were stained with Coomassie Blue staining solution and dried. The extent of phosphorylation was measured by liquid scintillation counting of the gel slices of each substrate.
Establishment of Bcl-2-or E1B 19K-overexpressing FaO Stable Cell Line-FaO cells were transfected with a mammalian expression vector containing full-length bcl-2 cDNA and with a vector containing fulllength E1B 19K cDNA (Ref. 44; kindly provided by Dr. Zarchuck, National Institutes of Health, Bethesda, MD) and selected with changes of fresh media containing G418 (500 g/ml) for 3 weeks. FaO sublines stably transfected with an empty vector were used as a control. The expression of the transfected gene products in the selected cell lines was confirmed by Northern blot analysis.

TGF-␤1 Induces Apoptosis in the Rat Hepatoma Cell FaO-
TGF-␤1 is a potent apoptotic inducer in hepatocytes (4 -6). FaO is a well differentiated hepatoma cell line that is very sensitive to the apoptotic effect of TGF-␤1, even in complete media containing 10% fetal bovine serum, as we previously reported (12,45). TGF-␤1 induces cell death in FaO cells with morphological changes such as cell shrinkage and cytoplasmic blebbing, which are characteristic of progression of apoptosis ( Fig.  1A). In addition, chromatin has a characteristic condensed and fragmented appearance following Hoechst 33258 staining of cells treated with TGF-␤1 for 12 h. Moreover, TGF-␤ induced a clear 180 -200-base pair internucleosomal DNA cleavage after TGF-␤1 treatment (Fig. 1B).

Regulatory Mechanism of Cell Cycle Components during TGF-␤1-induced Apoptosis Does Not Overlap with TGF-␤1induced Growth
Arrest Pathway-Next, we explored the possible differences between the TGF-␤1-mediated apoptotic pathway and the TGF-␤1-induced G 1 cell cycle arrest pathway. First, we examined whether TGF-␤1-induced G 1 cell cycle arrest is a prerequisite for apoptosis. Interestingly, cell cycle analysis demonstrated that the induction of apoptosis by TGF-␤1 was not preceded by G 1 cell cycle arrest. Rather, a progressive decrease of the G 1 phase population and an increase in the cells in G 2 /M phase were observed prior to TGF-␤1-induced apoptosis (Fig. 2). These results suggest that loss of G 1 cell cycle arrest may be implicated in the TGF-␤1-induced apoptotic mechanism.
We then attempted to study whether these changes of the cell cycle profile during TGF-␤1-induced apoptosis can be explained in terms of the changes in expression of the cell cycle regulatory components. We were interested in whether some cell cycle regulatory components are shared between the TGF-␤1-induced apoptotic pathway and the TGF-␤1 growth inhibitory pathway, or if the regulatory mechanisms controlling cell cycle progression are completely separate for the two pathways. We explored the expression of cyclins, Cdks, and Cdk inhibitors, which have previously been implicated in TGF-␤1mediated cell cycle arrest during TGF-␤1-induced apoptosis (Fig. 3A). The protein levels of Cdk2, Cdk4, Cdk6, cyclin E, and p27 were not significantly altered. The levels of proteins such as cyclin A, cyclin B, and cyclin D1, which are involved in the progression of the cell cycle, showed a gradual rise with a sustained peak at around 8 h after TGF-␤1 treatment, and then a subsequent decline. Upon TGF-␤1 treatment, the Cdc2 protein was induced at 4 h, reached its peak at 8 h, and then decreased to its previous level by 24 h. Several recent papers suggest that tyrosine phosphorylation of Cdc2 (46,47) is associated with the inactivation of Cdc2 kinase activity. However, no significant change in the tyrosine phosphorylation level of Cdc2 was observed during TGF-␤1-induced apoptosis (data not shown).
To investigate whether the changes in protein levels of some cell cycle components are caused by the changes in mRNA levels, Northern blot analysis was performed after TGF -␤1 treatment (Fig. 3B). Levels of Cdc2, cyclin A, and cyclin B mRNAs were dramatically induced at 2 h after TGF-␤1 treatment and then showed a gradual decline. The level of cyclin D1 mRNA was induced at 2 h after TGF-␤1 treatment and was sustained for at least 12 h. The mRNA level of p21 was quite abundant, but was slightly induced at 2 h after TGF-␤1 treatment, and then gradually decreased. In contrast, Cdk2, cyclin E, and p27 did not demonstrate any significant changes in mRNA levels during the progression of TGF-␤1-induced apoptosis. These results suggest that the expressions of cyclins, Cdks, or Cdk inhibitors involved in TGF-␤1-induced growth inhibition are regulated differently during TGF-␤1-induced apoptosis, except for a slight induction in p21.
Cdks Are Transiently Activated during TGF-␤1-induced Apoptosis-Cyclin-dependent kinases (Cdks) are regulated by their association with cyclins; cyclin A (Cdc2, Cdk2), B (Cdc2), D (Cdk2, Cdk4, and Cdk6), and E (Cdk2) (13,14). Since the expression of Cdc2 and cyclins including cyclin A, B, and D1 was changed during TGF-␤1-induced apoptosis, we investigated whether the activities of cyclin-dependent kinases are affected by the changes in the expression of Cdc2 itself or the associated cyclins. The kinase activities associated with Cdc2 and Cdk2 were measured by immunoprecipitation with respective antibodies, followed by a kinase assay on histone H1 as an exogenous conventional substrate (Fig. 4A). Cdc2 kinase activity showed an approximately 4.4-fold rise after 8 h of treatment with TGF-␤1, and a gradual decline after a peak in parallel with the change in the Cdc2 protein level as shown in Fig. 3A. Interestingly, Cdk2-associated histone H1 kinase activity was also markedly increased, despite the fact that the Cdk2 protein level was not changed. Western blot analysis of immunoprecipitated Cdk2 revealed the enhanced association of cyclin D1 with Cdk2 at the peak time of Cdk2 kinase activity (Fig. 4B), suggesting that increased cyclin D1 expression contributes to the increase of Cdk2 kinase activity. Recent studies have reported that histone H1 is a poor substrate for Cdk4 and Cdk6 (48). In consistent with their results, not only Cdk4 but also Cdk6 kinase activity on histone H1 was barely detected before or after TGF-␤1 treatment in our study (data not shown). Therefore, the changes in Cdk4 and Cdk6 kinase activity were analyzed again using the RB-C terminus fusion protein containing residues 701-928, which is a commonly used substrate for Cdk4 or Cdk6 kinase assay. However, activation of Cdk4 and Cdk6 kinase on RB-C-terminal fusion protein was not significant compared with Cdc2 and Cdk2 (Fig. 4A).
Both Cdc2 and Cdk2 Kinase Activity to Phosphorylate RB Are Dramatically Enhanced at the Early Phase of TGF-␤1-induced Apoptosis-The most studied G 1 cyclin/Cdk substrate is the product of retinoblastoma tumor suppressor protein (RB). Since the kinase activities of Cdks were transiently activated in this study, we investigated the phosphorylation status of endogenous RB in response to TGF-␤1. Western blot analysis using anti-phosphospecific RB antibody demonstrated a dramatic increase in RB phosphorylation, which showed a peak at 8 h following TGF-␤1 treatment (Fig. 5A), coinciding with the activation of Cdks observed in Fig. 4. The kinases responsible for RB phosphorylation in vivo have been reported to be members of the Cdk family, including Cdk2 in association with cyclin E and A, as well as Cdk4 and Cdk6 in association with D-type cyclins (reviewed in Ref. 49). Massive RB hyperphosphorylation in response to TGF-␤1 as shown in Fig. 5A may not be explained by the slight increase in Cdk4 and Cdk6 kinase activity on RB. At least seven of potential Cdk phosphorylation sites (Ser-249, -807, and -811, and Thr-252, -373, -821, and -826) have been shown to be phosphorylated in vivo (50,51). Therefore, our data suggest that N terminus of RB is also involved in RB hyperphosphorylation, although C-terminal fragment (amino acids 791-928) of RB has been conventionally used as an exogenous substrate for RB kinase such as Cdk2 and Cdk4. To examine which Cdk(s) play a major role for RB hyperphosphorylation, immune complex kinase assay of Cdk was performed using two types of RB substrate, including a recombinant RB-N terminus fusion protein containing RB residues 181-400 and an RB-C terminus fusion protein containing RB residues 701-928 (Fig. 5B). Surprisingly, not only Cdk2 but also Cdc2 demonstrated a significant increase in RB kinase activity on both the N and C termini of RB protein, peaking at 8 h after TGF-␤1 treatment. Interestingly, N-terminal RB protein phosphorylated by activated Cdc2 revealed two discrete bands with an equal intensity, whereas the RB protein phosphorylated by activated Cdk2 demonstrated a higher intensity in the upper shifted band. This result suggests that Cdc2 and Cdk2 kinase have a differential substrate specificity or preference for specific sites at RB-N terminus. In contrast, Cdk4 revealed a very trifling increase in kinase activity on RB-N terminus fusion protein. Cdk6 showed a slight activation on both RB-N terminus and RB-C terminus. These results indicate that activated Cdc2 as well as Cdk2 are responsible for a massive RB hyperphosphorylation following TGF-␤1 treatment. We further examined whether the Cdk2 or Cdc2 complex binds stably to RB to facilitate RB phosphorylation in vivo. We immunoprecipitated Cdk2-and Cdc2-containing complexes, respectively, and tested for coprecipitation of RB. As shown in Fig. 5C, Cdk2-or Cdc2-specific antibody precipitated Cdk2 or Cdc2, respectively, and coprecipitated more RB at the cell extracts prepared from cells treated with TGF-␤1 for 8 h. Therefore, these results suggest that the association of RB with Cdc2 or Cdk2 complex increases at the time of activation of these kinases. In conclusion, our results strongly suggest the existence of Cdc2-RB and Cdk2-RB complexes in vivo, and RB is a physiological target of inappropriately activated Cdc2 and Cdk2 by TGF-␤1.
Specific Inhibitors of Cdc2 and Cdk2 Kinase Block TGF-␤1induced Apoptosis-Recently, activation of cyclin-dependent kinases, either Cdc2 or Cdk2, has been shown to correlate with the onset of apoptosis (34 -40). However, the significance of activation of specific cyclin-dependent kinases in apoptosis has been controversial in several cell culture systems induced by diverse stimuli (52)(53)(54). In order to determine whether the activation of these kinases is a required signal for TGF-␤1induced apoptosis, we examined the effect of roscovitine and olomoucine (55,56), specific inhibitors for both Cdc2 and Cdk2, on TGF-␤1-induced apoptosis. FaO cells were pretreated with roscovitine or olomoucine at the various concentrations for 1 h and then treated with TGF-␤1. The extent of apoptosis was assessed by the TUNEL assay at 24 h after TGF-␤1 treatment (Fig. 6A). Treatment of these inhibitors blocked the DNA fragmentation of FaO cells that is normally induced by TGF-␤1 in a dose-dependent manner. The death-blocking effect of roscovitine was more potent than that of olomoucine. Cells treated with 30 M roscovitine or 100 M olomoucine prior to TGF-␤1 treatment were almost resistant to apoptosis (Fig. 6B). Next, we investigated whether RB hyperphosphorylation in response to TGF-␤1 is affected by the treatment of 30 M roscovitine. Western blot analysis using anti-phosphospecific RB antibody demonstrated no significant changes in the RB phosphorylation levels (Fig. 6C). To confirm whether this specific inhibitor blocks TGF-␤1-induced apoptosis by the inhibition of Cdc2 and Cdk2 kinase activity to phosphorylate RB, immune complex kinase assays of Cdks were performed using the control cell extracts and the cell extracts treated with 30 M roscovitine and TGF-␤1 for 8 h (Fig. 6D). The activities of Cdc2 and Cdk2 kinase on RB induced by TGF-␤1 were almost suppressed with the treatment of this inhibitor. This result confirms that increased Cdc2 and/or Cdk2 kinase activity to phosphorylate RB is required for TGF-␤1-induced apoptosis.

Overexpression of Bcl-2 Suppresses TGF-␤1-induced Apoptosis by Specifically Blocking Cdc2
Kinase Activation-Overexpression of the bcl-2 gene has been shown to block the cell death caused by diverse death stimuli in many cell types (41), although its mechanism of action still remains obscure. In addition, adenovirus E1B 19K protein, another Bcl-2 family member, is presumed to prevent the cell death, although its role in cell death has been studied less intensively than the role of Bcl-2 (57,58,60). First, we investigated whether overexpression of Bcl-2 or E1B 19K protein can block TGF-␤1-induced cell death. Stable cell lines expressing Bcl-2 or E1B 19K were generated (Fig. 7A). Most of the Bcl-2-overexpressing sublines were completely resistant to TGF-␤1-induced apoptosis as estimated by the TUNEL assay, despite slight differences in the expression of bcl-2 transcripts between them, whereas TGF-␤1-induced apoptosis was partially blocked in all E1B 19K protein-expressing cell lines (Fig. 7, B and C).
In order to test whether TGF-␤1-induced activation of Cdc2 or Cdk2 kinase directly correlates with the occurrence of apoptosis, the changes in Cdc2 and Cdk2 kinase activities were examined using a Bcl-2-expressing FaO subline (Bcl-2/FaO) and an E1B 19K protein-expressing FaO subline (E1B 19K/ FaO), respectively (Fig. 8A). Cell extracts were isolated from the two cell lines at the indicated times after TGF-␤1 treatment and subjected to a Cdc2-or Cdk2-associated histone H1 kinase assay. As shown in Fig. 8A, no significant activation of Cdc2 kinase activity was observed following TGF-␤1 treatment in Bcl-2/FaO cells, whereas Cdk2 kinase activity was markedly increased in these cells, similar to the increase shown in TGF-␤1-treated FaO cells (Fig. 4A). In E1B 19K/FaO cells, which showed a very slow progression of apoptosis in response to TGF-␤1, there was a gradual increase in Cdc2 kinase activity up to 24 h following TGF-␤1 treatment. However, Cdk2 kinase activity was still significantly activated with a peak at 8 h after TGF-␤1 treatment as observed in FaO cells. The similar results were obtained in Cdc2-or Cdk2-associated kinase assay on RB (data not shown). These results demonstrate that activation of Cdc2 kinase is closely linked to TGF-␤1-induced apoptosis in FaO cells. On the other hand, Cdk2 kinase can be activated in response to TGF-␤1 without concomitant apoptosis. Therefore, we can conclude again that activation of Cdc2 is essential for induction of apoptosis by TGF-␤1 in FaO cells, even though TGF-␤1 can activate both Cdc2 and Cdk2 kinases. Also these results suggest that the anti-apoptotic function of Bcl-2 and E1B 19K protein may be mediated through inhibition of Cdc2 kinase activation. Next, we investigated whether overexpression of anti-apoptotic genes affects the changes in the phosphorylation levels of endogenous RB following TGF-␤1 treatment (Fig. 8B). Whereas a very slight increase in RB phosphorylation with a peak at 8 h after TGF-␤1 treatment was observed in Bcl-2-overexpressing cells, phosphorylation of RB increased gradually up to 12 h in 19K protein-overexpressing cells. These results suggest again that the apoptotic occurrence may be associated with a massive hyperphosphorylation of RB.
Next, the possible action mechanism by which Bcl-2 and 19K regulates the activation of Cdc2 and Cdk2 in response to FIG. 4. Cdc2 as well as Cdk2 kinase is dramatically activated during TGF-␤1-induced apoptosis. A, Cdk immune complex kinase assays. Whole cell extracts were prepared from FaO cells treated with TGF-␤1 for the indicated time points, and equal amounts of the lysates (500 g) were used for the specific immune complex kinase assay. Cdc2 and Cdk2 immune complex kinase assay was performed using histone H1 as a substrate as described under "Materials and Methods." Cdk4 and Cdk6 immune complex kinase assay was carried out using RB-C terminus fusion protein containing residues 701-928 as a substrate. Samples were analyzed by 10% SDS-polyacrylamide gel electrophoresis and autoradiography. Representative data from three independent experiments are shown. -Fold change in the respective kinase activity was calculated from liquid scintillation counting of each gel slice and denoted as numbers. B, association of cyclins with Cdk2. Cdk2 was immunoprecipitated (IP) from the control cells and cells treated with TGF-␤1 for 8 h. Amounts of immunoprecipitated kinases were determined by incubating blots with anti-Cdk2 antibodies. Anti-cyclin D1 antibody showed increased coprecipitation with Cdk2 at 8 h after TGF-␤1 treatment.
TGF-␤1 was investigated by the expression analysis of the related cell cycle components in the stable cell lines overexpressing Bcl-2 and 19K. No significant difference in the expression pattern of Cdk2 and cyclin D1 was observed among Bcl-2-, 19K-overexpressing cells, and control cells (pcDNA3/FaO), explaining the reason that the consequent activation of Cdk2 in response to TGF-␤1 is not affected by the overexpression of these anti-apoptotic gene products (Fig. 9A). Interestingly, TGF-␤1 failed to induce Cdc2 mRNA and protein in Bcl-2expressing cells. In contrast, in 19K-expressing cells, the delayed induction of Cdc2 mRNA and protein was observed (Fig.  9, A and B). Therefore, these results indicate that the blocking effect of TGF-␤1-induced apoptosis by Bcl-2 or 19K may be closely associated with the suppression of induction in Cdc2 mRNA and the subsequent Cdc2 kinase activity.
In conclusion, TGF-␤1-induced increases in Cdc2 and Cdk2 kinase activities significantly contribute to the transient hyperphosphorylation of RB in FaO cells. The consequent inactivation of RB may promote the failure of controlled cell cycle progression, leading to irreversible cell death. DISCUSSION Apoptosis is a complex cellular response with multiple signaling pathways and regulatory proteins. We have previously reported that activation of caspase-2 may have a crucial role at the execution stage of TGF-␤1-induced apoptosis in FaO rat FIG. 5. Both Cdc2 and Cdk2 kinase activated during TGF-␤1induced apoptosis significantly phosphorylate RB. A, hyperphosphorylation of RB during TGF-␤1-induced apoptosis. Cell extracts were prepared from the respective cells treated with TGF-␤1 for 0, 2, 4, 8, 12, and 24 h and subjected to Western blotting with anti-phosphospecific RB antibody. B, Cdc2-, Cdk2-, Cdk-4, and Cdk6-associated immune complex kinase assay using recombinant RB fusion protein. Cell extracts were prepared from the cells not treated with TGF-␤1 and from cells treated with TGF-␤1 for 8 h. RB-N terminus fusion protein containing RB residues 181-400 or RB-C terminus fusion protein containing residues 701-928 was used as a substrate for the respective immune complex kinase assay. Two phosphorylated RB-N terminus fusion proteins by activated Cdc2 were marked by asterisks. Representative data from three independent experiments with similar results are shown. -Fold change in the respective kinase activity was calculated from liquid scintillation counting of each gel slice and denoted as numbers. C, association of RB with active Cdc2 and Cdk2 immune complex. Cell extracts were prepared from cells not treated with TGF-␤1 and from cells treated with TGF-␤1 for 8 h. RB, Cdk2, and Cdc2 complex were immunoprecipitated (IP) from each cell extract using specific antibodies, and immune complexes were resolved on 10% SDS-polyacrylamide gel, transferred to nitrocellulose. Coprecipitation of RB was examined by immunoblotting with RB-specific antibody.
FIG. 6. Treatment with roscovitine or olomoucine blocks TGF-␤1-induced apoptosis inhibiting activities of Cdc2 and Cdk2 kinase on RB. A, dose dependence of roscovitine and olomoucine on TGF-␤1-induced apoptosis. FaO cells were pretreated with roscovitine or olomoucine at the indicated concentrations (Conc.) for 1 h prior to TGF-␤1 treatment for 24 h. The TUNEL procedure was carried out following the instructions for the in situ cell death kit (Roche Molecular Biochemicals). Apoptotic cells were counted, and the percentage of apoptotic cells was denoted as graphs. B, representative pictures to demonstrate the effect of roscovitine and olomoucine on TGF-␤1-induced apoptosis. TUNEL assay was performed with the cells treated as follows, and pictures were taken under a light microscope (magnification, ϫ200): a, untreated FaO cells; b, FaO cells treated only with TGF-␤1 for 24 h; c, FaO cells pretreated with roscovitine (30 M) for 1 h and further treated with TGF-␤1 for 24 h; d, FaO cells pretreated with olomoucine (100 M) for 1 h and then treated with TGF-␤1 for 24 h. C, phosphorylation levels of RB in the cells treated with TGF-␤1 after pretreatment of 30 M roscovitine. Cell extracts were prepared from the cells at the denoted time points after treatment of TGF-␤1 together with roscovitine and subjected to Western blotting with anti-phosphospecific RB antibody. D, inhibition of activities of Cdc2 and Cdk2 kinase on RB by roscovitine. Cell extracts were prepared from the control cells and from the cells treated with TGF-␤1 for 8 h after pretreatment of 30 M roscovitine. Cdc2-, Cdk2-, Cdk4-, and Cdk6-associated immunocomplex kinase assay using recombinant RB-N terminus fusion protein or RB-C terminus fusion protein was performed as described in Fig. 5B. hepatoma cells (12). In this study, we explored the signaling pathways at the early stages of TGF-␤1-induced apoptosis. The regulatory pattern of cell cycle components after TGF-␤1 treatment showed the signal for a rapid cell cycle progression, including the transient transcriptional induction of cyclin A, B, D1, and Cdc2, together with hyperphosphorylation of RB. Furthermore, the progressive decrease in the G 1 cell cycle population occurred before entry into apoptosis. Therefore, TGF-␤1induced apoptosis takes a separate signaling pathway from the TGF-␤1-induced cell cycle arrest pathway, which is one of the most studied aspects of TGF-␤ function. Lack of cellular capacity to accommodate these excessive proliferating signals induced by TGF-␤1 may contribute to a failure in coordinated control at the checkpoint of cell cycle transition and a resultant irreversible cell death in FaO cells.
Expression of various cyclins (30 -33) and inappropriate ac-tivation of cyclin-dependent kinases (34, 35, 37-39, 61, 62) have been implicated in apoptotic triggering of several experimental systems. However, it has been a matter of controversy whether activation of cyclin-dependent kinases is a secondary event of apoptosis, depending on the systems. De Luca et al. (52) suggested that Fas-induced changes in Cdc2 and Cdk2 kinase activities are not sufficient enough to trigger apoptosis in HUT-78 cells, and some models of apoptosis have demonstrated that enhanced activation of Cdc2 kinase is not required for apoptosis to occur (53,54). Furthermore, the inhibitor of Cdk2 has been reported to induce apoptosis in some cells (63,64). In contrast, evidence that Cdc2 activity is required for apoptosis is provided by the observation that overexpression of a dominant negative Cdc2 mutant blocks Fas/APO-1-induced apoptosis in Jurkats cells (40), and that cells with a temperature-sensitive Cdc2 mutant are unable to undergo apoptosis in A, Cdc2-and Cdk2-associated immune complex kinase assay using histone H1. The same procedure was performed as described in Fig. 4. B, changes in the phosphorylation levels of RB in Bcl-2-and 19 kDa-overexpressing cells in response to TGF-␤1. Each cell line overexpressing Bcl-2 or E1B 19K was treated with TGF-␤1 for the indicated times, and cell extracts were prepared for the Western blotting to detect the phosphorylated forms of RB. response to a variety of stimuli when cultured at the restrictive temperature (38). Inhibition of cyclin B1-Cdc2 kinase activity has been proposed as one mechanism by which some transforming oncogenes protect against apoptosis (65).
Our studies on the changes in Cdk activities during TGF-␤1induced apoptosis demonstrated that all Cdks analyzed in our study were activated with a peak at 8 h after TGF-␤1 treatment. In our study, the significant role of Cdc2 and/or Cdk2 kinase in TGF-␤1-induced apoptotic signaling was demonstrated by the result that roscovitine or olomoucine, potent Cdc2 and Cdk2 inhibitors, effectively blocked TGF-␤1-induced apoptosis. In particular, a direct correlation was observed between the activation of Cdc2 and the progression of apoptosis in response to TGF-␤1. Not only the apoptotic occurrence induced by TGF-␤1 but also the blocking of apoptosis by roscovitine or by the overexpression of anti-apoptotic gene products such as Bcl-2 and E1B 19K were closely associated with activation of Cdc2 kinase. Therefore, these results suggest that activation of Cdc2 is critical for TGF-␤1-induced apoptosis. Interestingly, Cdk2 kinase activity was activated in response to TGF-␤1 regardless of apoptotic occurrence, suggesting that only the increase in this Cdk2 activity may not be sufficient to trigger apoptosis.
Overexpression of Bcl-2 family proteins protects against the cell death induced by diverse death stimuli, although their action mechanisms are still obscure (41). Previous reports have demonstrated that the inhibition of death by Bcl-2 is associated with alterations in the expression and localization of Cdk proteins or cyclin A (35,37). In addition, the level of Bcl-2 protein in T cells has been connected with the retardation of the G 1 /S transition through the sustained level of p27 (66) or through dephosphorylation of RB (43). Therefore, the protective effect of Bcl-2 against cell death may be accomplished by modulating the cell cycle progression, i.e. by increasing the length of the G 1 phase. In this study, we have shown that the overexpression of Bcl-2 inhibited the induction of Cdc2 mRNA in response to TGF-␤1. In particular, the complete blocking of TGF-␤1-induced apoptosis by Bcl-2 may be accomplished by inhibiting the inappropriate activation of Cdc2 kinase through the suppression of TGF-␤1-induced Cdc2 expression.
RB has been regarded as a major effector for G 1 cell cycle arrest induced by TGF-␤1 in many cell types, including epithelial cells. We present the several lines of evidence that RB serves as a target of activated Cdks in TGF-␤1-induced apoptosis of FaO cells. An increase in the hyperphosphorylated forms of RB following TGF-␤1 treatment coincides with the activation of all Cdks. In addition, when TGF-␤1-induced apoptosis is blocked, either by the pretreatment of specific inhibitors of Cdc2/Cdk2 or by the overexpression of anti-apoptotic gene products, RB hyperphosphorylation was not observed. RB hyperphosphorylation by both Cdc2 and Cdk2 markedly increased in response to TGF-␤1, whereas Cdk4 and Cdk6 demonstrated only slight increases in their kinase activities on RB. Furthermore, the association of RB with Cdc2 or Cdk2 was detected in active Cdc2 or Cdk2 immune complex at the peak of their RB kinase activities.
RB contains 16 Ser/Thr-Pro motifs, which are potential Cdk phosphorylation sites (50,51,67). The kinases responsible for RB phosphorylation in vivo are known to include members of the Cdk family, including Cdk2 in association with cyclin E and A as well as Cdk4 and Cdk6 in association with D-type cyclins (56). Cdc2 kinase activity to phosphorylate RB has been documented from in vitro experiments using purified Cdc2 complex (50,68). In this study, we demonstrate the evidence for the first time that Cdc2 activated during apoptosis phosphorylates RB not only at N terminus but also at C terminus. The appearance of two discrete phosphorylated bands as shown in Cdc2 immune complex kinase assay on RB-N-terminal fusion protein (Fig. 5B) suggests that there may be more than one phosphorylation site between residues 181 and 400 of RB protein by Cdc2 kinase. Supporting our idea, Ser-289, Thr-252, and Thr-373 of RB protein have been reported to correspond closely to the consensus sequence for phosphorylation by Cdc2 kinase (50).
Evidence is emerging that various Cdks differentially phosphorylate RB at distinctive residues in vitro (68,69). Phosphorylation at particular residues of RB may affect the binding of RB to only particular subsets of its interacting partners (70). This suggests that different arrays of RB phosphorylation by distinct Cdks may result in differential regulation of downstream effector pathways. Therefore, it will be very interesting to examine how the decision making for TGF-␤1-induced apoptosis is controlled by RB hyperphosphorylation. An abrupt RB hyperphosphorylation by activated Cdks may result in its functional loss as a coordinator at G 1 /S cell cycle transition and may trigger the subsequent irreversible cell death. However, we cannot exclude the possibility that the phosphorylation event at particular sites of RB by activated Cdc2 may be sufficient for apoptotic triggering through downstream signaling, such as a release of E2F.
Involvement of RB in apoptosis has already been suggested by several groups. Massive apoptotic cell death as well as inappropriate cell proliferation was observed in RB deficient mice (71). Ectopic overexpression of adenovirus E1A, human papilloma virus E7, E2F, cyclin D1, or c-Myc, all of which associate with and/or inactivate RB, has been demonstrated to be associated with cell death (72)(73)(74)(75)(76). RB was phosphorylated during apoptosis of Balb/c-3T3 fibroblasts induced by serum FIG. 9. Detailed action mechanism of Bcl-2 and 19K to block TGF-␤1-induced apoptosis. PcDNA3/FaO, Bcl-2/ FaO, and 19K/FaO cells were treated with TGF-␤1 for the indicated time points, respectively. A, Western blot analysis. Equal amounts of cell extracts (40 g) were resolved by SDS-PAGE and analyzed by Western blotting using antibodies specific for the indicated proteins. B, Northern blot analysis of Cdc2 mRNA. mRNA was isolated from the respective cell lines at the indicated time points. Equal amounts of mRNA (10 g) were subjected to Northern blotting for the denoted gene products. The changes in the levels of Cdc2 mRNA were compared among three cell lines. deprivation (77). Furthermore, the underphosphorylated active form of RB plays a crucial role in protecting cells from radiation-induced apoptosis in cell line SAOS-2 (78). However, the signaling pathways leading to activation of RB kinases and the action mechanism of specific Cdks modulating RB function to trigger apoptosis need to be investigated further.
In conclusion, we demonstrated clearly that TGF-␤1 induces transient up-regulation of Cdc2 and several cyclins in FaO cells. The subsequent activation of Cdc2 and Cdk2 is responsible for the marked hyperphosphorylation of RB. This RB phosphorylation event may be linked to the resultant failure in G 1 cell cycle arrest, inappropriate S phase entry, and ultimately to irreversible cell death.