Differential modulation of G1 cyclins and the Cdk inhibitor p27kip1 by platelet-derived growth factor and plasma factors in density-arrested fibroblasts.

Stimulation of quiescent Balb/c 3T3 fibroblasts into S phase requires the synergistic action of platelet-derived growth factor (PDGF) and progression factors found in platelet-poor plasma (PPP). Traverse of the G1/S phase boundary and the initiation of DNA replication require functional cyclin E-cyclin-dependent kinase (Cdk) 2 and cyclin A-Cdk2 complexes; however, the mechanisms by which PDGF and PPP regulate Cdk2 activation are not known. Density-arrested fibroblasts contain low levels of cyclins E and A, and high levels of the Cdk inhibitor p27kip1. Exposure of PDGF, which stimulates cell cycle entry but not progression through G1, induces the formation of cyclin D1-Cdk4 complexes that bind p27kip1 and titrate the pool of Kip1 available to inhibit Cdk2. In addition, PDGF stimulates a moderate transient reduction in the abundance of p27kip1 protein. However, limited expression of cyclin E and cyclin A is observed after PDGF treatment, and in the absence of PPP, p27 levels are sufficient to bind and inactivate existing cyclin-Cdk complexes. Although plasma does not significantly increase the proportion of Kip1 bound to cyclin D1-Cdk4, stimulation of PDGF-treated cells with plasma does overcome the threshold inhibition of p27kip1 by further increasing the expression of cyclins E and A and decreasing the amount of Kip1 over a prolonged time period. Our results indicate that the distinct mitogenic activities of PDGF and PPP differentially influence the activation of cyclin E- and cyclin A-associated kinases that ultimately regulate entry into S phase.

The growth of nontransformed eukaryotic cells is controlled by environmental cues which govern the transition from G 1 into S phase. In mammalian cells, growth regulatory signals from serum-derived growth factors are integrated during a late G 1 event called the restriction point (1). Successful execution of this event commits cells to another round of DNA replication, at which time cell cycle progression becomes independent of extracellular mitogens.
Growth factor-stimulated proliferation is achieved, at least in part, by a modulation of the cell cycle machinery consisting of the cylin-dependent kinases (Cdks) 1 and their regulatory cyclin subunits (2,3). While a single kinase, p34 cdc2 , is sufficient to drive progression through the major cell cycle regulatory points at the G 1 /S and G 2 /M phase transitions in yeast (4), multiple distinct Cdc2-related kinases have been identified in higher eukaryotes (5)(6)(7). The involvement of many of these kinases in cell cycle regulation has yet to be established. However, Cdk4 becomes active as a retinoblastoma protein kinase during mid G 1 (8), and overexpression of Cdk4 in epithelial cells reduces the requirement for serum-derived growth factors and confers resistance to TGF-␤-mediated growth inhibition (9). Furthermore, Cdk2 rescues the growth-arrested phenotype of Cdc2-deficient Saccharomyces cerevisiae yeast mutants (5), and ablation of Cdk2 activity prevents the onset of DNA replication in mammalian cells (10,11). Thus, growth stimulatory pathways initiated by extracellular signals must ultimately engage and activate one or more of the G 1 -specific Cdk proteins.
Cdk activation is positively regulated by periodic association with cyclin subunits (2,3). Cdk4 complexes with the D-type cyclins, while Cdk2 primarily associates with cyclin E and cyclin A. Ectopic overexpression of either cyclin D 1 or cyclin E accelerates progression through G 1 and reduces the proliferative requirement for serum-derived growth factors (12)(13)(14). Conversely, abolition of cyclin D 1 or cyclin E activity through the use of neutralizing antibodies or antisense oligonucleotides effectively blocks entry into S phase (15,16). Ablation of cyclin A function also prevents DNA replication (17,18) and disrupts the checkpoint control pathway that couples mitotic initiation to the completion of DNA synthesis (19). However, it appears that the cyclin A-Cdk2 complex may function at a point distal to restriction point traverse. Although the mitogen-dependent expression of cyclin D genes is well defined in mammalian cells (20 -22), growth factor regulation of cyclin E and cyclin A is incompletely understood.
Cyclin interaction with its catalytic partner is necessary but not sufficient for kinase activation. Nonfunctional cyclin-Cdk complexes have been shown to accumulate in serum-stimulated senescent fibroblasts (23) as well as cells that have been growth inhibited by exposure to TGF-␤ (24), ␦ radiation (25), or agents which elevate intracellular levels of cAMP (26). Activation of assembled cyclin/Cdk holoenzymes is negatively regulated by direct interaction with Cdk inhibitory proteins, termed CKIs (27). Several of these inhibitors function as intracellular effectors of antiproliferative environmental signals. In S. cerevisiae, cell cycle arrest in response to mating phermone ␣ is mediated by FAR1, a protein which inactivates the yeast Cdc2 kinase (28). In mammalian cells, DNA damaging agents such as ␦ radiation induce the expression of p53, which transactivates the promoter of the Cdk inhibitor, p21 cip1 (29). Furthermore, TGF-␤-dependent inhibition of Cdk2 activation and S phase entry is mediated by the inhibitor, p27 kip1 (30). Recently it has been shown that p27 expression is down-regulated after interleukin 2 stimulation of T lymphocytes (31), suggesting that Kip1 may serve as a common target of both positive and negative growth regulatory pathways.
Kip1 activity may also be regulated through association with cellular proteins such as Cdk4. p27 kip1 associates with the cyclin D-Cdk4 complex in a cell cycle-dependent manner (32), and treatment of human keratinocytes with the antiproliferative agent TGF-␤ leads to a redistribution of p27 kip1 from Cdk4 to Cdk2, correlating with an inhibition of Cdk2 activity and cell cycle arrest (33). Baculovirus-produced cyclin D 2 -Cdk4 complexes can facilitate activation of Cdk2 in vitro (34), suggesting that a sequestering of Kip1 by Cdk4 may also be an important component of the mitogenic response to environmental proliferative signals. However, this hypothesis has not been rigorously tested in vivo.
Balb/c 3T3 fibroblasts are a nontransformed mouse cell line that has been extensively characterized with regard to the proliferative requirements for specific serum-derived growth factors (34). Platelet-derived growth factor (PDGF) acts early in the cell cycle to stimulate the G 0 to G 1 transition and render cells competent to respond to progression factors contained in platelet-poor plasma (PPP). Sequential exposure of quiescent fibroblasts to PDGF and PPP is sufficient to stimulate traverse of the restriction point and initiate commitment to DNA synthesis. Using the Balb/c 3T3 fibroblast system, we have examined the molecular mechanisms by which growth regulatory signals such as PDGF and plasma factors cooperatively activate the Cdk2 kinase during late G 1 . Our results suggest that both a PDGF-dependent association of Kip1 with Cdk4 and a plasma-dependent reduction in Kip1 levels are essential for the activation of cyclin E-and cyclin A-dependent kinases.

MATERIALS AND METHODS
Cell Culture-Balb/c 3T3 mouse embryo fibroblasts (clone A31) were cultured in a water-jacketed incubator with a humidified atmosphere (5% CO 2 , 95% air) maintained at 37°C. Experimental cultures were grown to confluency in 100-mm Petri dishes using Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated bovine calf serum (Colorado Serum Company), 4 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. Cells were used 3-4 days after density arrest. PDGF (BB) was purchased from Biosource International, Camarillo, CA, and plasma was obtained fropm the Southwest Florida Blood Bank, Tampa, FL.
Immunoprecipitations and Kinase Assays-Balb/c 3T3 cell lysates were prepared exactly as described for the preparation of immunoblots. Kinase activities were immunoprecipitated from 100 g of extract using rabbit polyclonal antisera (approximately 4 h at 4°C). Immunoprecipitates were brought down with protein A-agarose (30 min at 4°C), and pellets were washed twice in immunoblot lysis buffer and twice in kinase buffer (50 mM Tris, pH 7.4, 10 mM MgCl 2 , 1 mM dithiothreitol). Pellets were incubated in 50 l of kinase buffer containing 5 M ATP, 100 g/ml histone H1, and 0.1 C1/ml [␥-32 P]ATP for 5 min at 30°C. Reactions were stopped by boiling 3 min in 50 l of 2 ϫ Laemmli buffer, and separated on a 10% polyacrylamide gel. Histone phosphorylation was visualized by autoradiography (0.5-3 h at Ϫ70°C with intensifying screens) and quantitated using a PhosphorImager and ImageQuant software (Molecular Dynamics).
Inhibition experiments were performed by mixing extracts of proliferating cells containing active cyclin A-kinase complexes with inhibitorcontaining extracts for 1 h at 37°C prior to immunoprecipitation of cyclin A. Density-arrested Balb/c 3T3 cells were treated with 20 ng/ml PDGF with 10% calf serum in Dulbecco's modified Eagle's medium for 18 h then harvested for proliferating cell extracts. Unless noted otherwise, extracts were mixed in a 1:1 ratio of protein (100 g:100 g).

Induction of Cyclin E-and Cyclin
A-associated Kinase Activities in Balb/c 3T3 Fibroblasts-To examine the expression of cyclin E and cyclin A after mitogenic stimulation, densityarrested Balb/c 3T3 fibroblasts were treated with PDGF, PPP, or both. Whole cell extracts were harvested after 16 h (for cyclin E) or 21 h (for cyclin A) and analyzed by immunoblotting ( Fig.  1). Quiescent Balb/c 3T3 cells contained undetectable amounts of cyclin A protein, while low levels of cyclin E were typically observed in unstimulated fibroblasts. A weak induction of both cyclin A and cyclin E expression was observed after exposure to PDGF, a factor which stimulates cell cycle entry from quiescence but does not support traverse of the late G 1 restriction point or subsequent entry into S phase. Although plasma by itself did not affect cyclin expression, the abundance of both cyclin A and cyclin E was increased to maximal levels in fibroblasts stimulated with a combination of PDGF and PPP. In contrast to the cyclins, Cdk2, the catalytic subunit of cyclins A and E, was relatively abundant in quiescent cells and underwent only a modest increase after growth factor stimulation (data not shown).
Although exposure to PDGF stimulated limited expression of cyclin A protein in nonproliferating cells, it was not sufficient to induce cyclin A-associated kinase activity (Fig. 1). Similarly, cyclin E-Cdk complexes immunoprecipitated from PDGF-or plasma-treated cells failed to phosphorylate histone H1 above basal levels. Identical results were also observed when Cdk2 was immunoprecipitated from fibroblasts treated with PDGF or plasma alone (data not shown). However, cyclin A-and FIG. 1. Induction of cyclin A and cyclin E expression and associated kinase activities. Density-arrested Balb/c 3T3 cells were stimulated with 25 ng/ml PDGF-BB or 10% PPP, or both. Whole cell extracts were prepared after a 16-h stimulation for analysis of cyclin E, and after a 21-h stimulation for cyclin A. Top panel, extracts from quiescent and stimulated cells were resolved on a 10% SDS-polyacrylamide gel, immunoblotted, and probed with polyclonal antibodies to cyclin A or cyclin E. Bottom panel, cyclin E and cyclin A immunoprecipitates from nontreated and stimulated cell extracts were analyzed for in vitro histone H1 kinase activity. Phosphorylated proteins were separated on a 10% SDS-polyacrylamide gel and visualized by autoradiography.
cyclin E-associated kinase activities were dramatically increased in cells receiving both PDGF and PPP. Induction of kinase activity under these conditions was greater than that predicted for an additive response, indicating that plasmaderived progression factors can act synergistically with PDGF to regulate the activation of cyclins A and E associated Cdk2.
Growth Factor Modulation of the Cdk Inhibitor Kip1-Mixing of extracts from quiescent Balb/c 3T3 cells with an equal amount of lysate from growth-stimulated fibroblasts containing active cyclin-Cdk complexes caused an inhibition of kinase activities associated with cyclin A, cyclin E, Cdk2, and to a lesser extent Cdc2 ( Fig. 2A). Thus, failure to stimulate cyclin Aand cyclin E-dependent kinases in PDGF-treated cells could in principle reflect an inability to down-regulate one or more of the Cdk inhibitory factors present in quiescent fibroblasts. As the predominant cyclin-dependent kinase activity is associated with cyclin A in Balb/c 3T3 fibroblasts, we have used cyclin A-Cdk complexes to measure inhibition in the following experiments.
Although exposure of quiescent fibroblasts to plasma alone had no effect on inhibitor levels, stimulation of density-arrested Balb/c 3T3 cells with either PDGF or a combination of PDGF and PPP resulted in a nearly identical reduction of free cyclin A/Cdk inhibitory activity that was biphasic in nature ( Fig. 2B). Activity of the Cdk inhibitor(s) was rapidly and dramatically reduced by approximately 75% between 2 and 6 h after treatment with PDGF or PDGF/PPP (Fig. 2C). After this time, free inhibitor activity continued to decline at a more gradual rate until it was completely abolished by 12-15 h poststimulation, a point coincident with S phase entry and normal cyclin A activation in those cells exposed to a full complement of growth factors. Down-regulation of free inhibitory activity persisted for at least 24 h after initial mitogen stimulation.
The majority of Cdk inhibitory activity could be restored to growth factor-treated cell lysates when samples were heattreated prior to incubation with proliferating cell extracts (Fig.  2B). These results suggest that a Cdk inhibitor present in both quiescent and stimulated Balb/c 3T3 fibroblasts was reversibly masked by interaction with a heat-labile factor after exposure to PDGF: p27 kip1 had previously been shown by others to be heat-stable in other cells (30). However, boiled extracts of stimulated cells were less effective in inhibiting the cyclin A-Cdk complex than identically treated extracts of quiescent cells. Therefore, a decrease in the abundance of inhibitory factors may also contribute to the apparent down-regulation of inhibitory activity after growth factor treatment. Inhibition from boiled extracts decreased slowly starting 6 -9 h after stimulation (Fig. 2C). However, in boiled extracts from cells treated with PDGF alone, decline in activity was transient, and the ability to inhibit cyclin A-Cdk complexes returned to basal levels at later time points. In contrast, total cellular inhibitory activity continued to decline over the time course of the experiment when cells were exposed to both PDGF and PPP.
In order to identify the factor responsible for cyclin A/Cdk inactivation, boiled lysates from quiescent and stimulated Balb/c 3T3 cells were precleared with antibodies to p27 kip1 prior to use in mixing experiments. Immunodepletion of Kip1 from extracts of both nonstimulated cells and cells treated with PDGF and PPP for 24 h eliminated essentially all inhibitory activity toward the cyclin A-Cdk complex (Fig. 3A). Immunoprecipitation of [S 35 ]methionine-labeled cells and Western blot analysis using the anti-p27 antibody indicated that this antibody did not recognize p21 cip1 or p57 kip2 (data not shown). These data suggest that Kip1 is the primary negative regulator of cyclin A-dependent kinase activity in Balb/c 3T3 fibroblasts.
Immunoblotting of whole cell extracts demonstrated that Kip1 expression was influenced by both PDGF and plasma factors (Fig. 3B). The p27 protein was relatively abundant in quiescent Balb/c 3T3 cells but was moderately reduced in response to treatment with PDGF. The PDGF-mediated reduction in Kip levels reached a nadir 12-18 h after stimulation; however, Kip1 expression increased by 24 h, correlating temporally with a return of total cellular inhibitory activity to the maximal basal level observed in nonstimulated cells (Fig. 2B). In contrast, fibroblasts stimulated with PDGF in the presence of plasma displayed a more dramatic and prolonged reduction in Kip1 expression. By 24 h after stimulation, cells treated with PDGF and PPP contained less than 50% of the p27 expressed in cells receiving PDGF alone (Fig. 3C). Kinetics of the plasmadependent decline in Kip1 levels at later time points closely paralleled the reduction of inhibitory activity detected in boiled cell lysates. These data show that p27 expression is differen- Growth factor-stimulated down-regulation of Cdk inhibitory activity. A, density-arrested Balb/c 3T3 cells were stimulated with 10% serum and 10 ng/ml PDGF-BB. After 18 h, extracts containing active cyclin-Cdk complexes were isolated and mixed with inhibitorcontaining extracts of quiescent cells. 100 g of stimulated cell extract were incubated with 100 g of quiescent cell lysate for 1 h at 37°C. Cyclin A, cyclin E, Cdk2, and Cdc2 were immunoprecipitated with polyclonal antibodies, and histone H1 kinase activity was determined. Phosphorylated proteins were separated on a 10% SDS-polyacrylamide gel, and kinase activity was quantitated on a PhosphorImager. 100% activity is the activity without inhibitor-containing extracts added. B, density-arrested Balb/c 3T3 cells were stimulated with 25 ng/ml PDGF-BB in the presence and absence of PPP. At the times (hrs) indicated, cells were harvested and whole cell extracts were prepared. Nontreated extracts and extracts heated 5 min at 100°C were mixed with proliferating cell lysates for 1 h at 37°C. Cyclin A was immunoprecipitated with a polyclonal antibody, and histone H1 kinase activity was determined. The (ϩ) is without inhibitory extracts added. C, cyclin A-dependent kinase activity shown in B was quantitated using a Phos-phorImager, and the percentage of inhibition was graphed. Data points represent the average of three separate experiments. 100% activity is the inhibition produced by quiescent extracts. tially modulated by PDGF and PPP, and indicate that plasmaderived growth factors are required for a full down-regulation of Kip1 protein.
Although Kip1 levels were dramatically reduced after stimulation with PDGF and PPP, under no condition was Kip1 expression completely abolished. The data presented in Fig. 2 indicate that the low level of Kip1 protein present in cells exposed to PDGF and PPP is sufficient to inhibit the majority of cyclin A-dependent kinase activity when released from a masking factor by heat treatment. These results imply that the PDGF-mediated association of p27 with a heat labile silencing factor is likely to be essential for the removal of Kip1 inhibitory activity.
Requirement for Protein Synthesis during Down-regulation of Free Inhibitory Activity-To further examine the mechanism by which growth factors decreased the level of free inhibitory activity in Balb/c 3T3 cells, density-arrested fibroblasts were stimulated with PDGF and PPP in the presence of cycloheximide (Fig. 4A). Whole cell extracts of cells exposed to growth factors or cycloheximide or a combination of both were harvested 12 h after stimulation and used without heat pretreatment in mixing experiments. Addition of cycloheximide at 10 g/ml blocked greater than 95% of serum-stimulated protein synthesis, but did not by itself affect levels of inhibitory activity in unstimulated cells. Thus, termination of new Kip1 synthesis is not sufficient to rapidly eliminate the Cdk inhibitory activity present in quiescent Balb/c 3T3 fibroblasts. Furthermore, the normal removal of free inhibitory activity after treatment with PDGF or PDGF/PPP was effectively blocked when cells were stimulated in the presence of cycloheximide. Treatment of fibroblasts with the RNA synthesis inhibitor 5,6-dichlorobenzimidazole riboside also antagonized the growth factor-dependent reduction in Cdk inhibition, although not to the extent observed after exposure to cycloheximide.
Identical effects on inhibitory activity were observed in cells that received cycloheximide simultaneously with exposure to PDGF/PPP and those in which cycloheximide was added 2 h after mitogenic stimulation (Fig. 4B). However, down-regulation of free inhibitory activity was markedly less sensitive to protein synthesis inhibition by 4 h after growth factor treatment. Moreover, addition of cycloheximide 4 h prior to harvest only weakly influenced the amount of inhibitory activity. These results, taken together with the data presented in Fig. 4A, indicate that the elimination of free Cdk inhibitory activity is absolutely dependent on protein synthesis 2-4 h after exposure to PDGF.
To determine whether protein synthesis inhibition affected the growth factor-mediated decrease in p27 expression, lysates from quiescent Balb/c 3T3 cells stimulated in the presence and absence of cycloheximide were immunoblotted and probed with  2). B, density-arrested cells were treated with PDGF and PPP as above. At the indicated times after stimulation, cycloheximide was added to culture medium. Cells represented in lane 8 received no cycloheximide. 12 h after initial exposure to mitogens, extracts were prepared and mixed with proliferating cell lysates. Control extracts of proliferating cells were mixed with lysis buffer (lane 1) or nontreated quiescent cell extract (Q, lane 2). Cyclin A was immunoprecipitated from mixed extracts and histone H1 kinase activity was determined. C, cycloheximide was added to quiescent Balb/c 3T3 cells stimulated with PDGF in the presence or absence of PPP as described in A. After 12 h, extracts were isolated, immunoblotted, and probed with polyclonal antibodies Kip1 and cyclin D 1 .
antibodies to Kip1 (Fig. 4C). After a 12 h stimulation, no detectable difference in Kip1 levels was observed in cells exposed to cycloheximide and PDGF or PDGF/PPP compared with cells treated with growth factors alone. In contrast, the addition of cycloheximide completely abolished PDGF-induced expression of cyclin D 1 in the same cells. Thus, the effect of cycloheximide on mitogen-dependent reduction in Cdk inhibitory activity is not mediated at the level of Kip1 expression. These results demonstrate that the decline in p27 levels achieved after PDGF stimulation is not sufficient to effect a decrease in Cdk inhibition.
Interaction of Kip1 with Cdk4 -The moderate reduction in p27 expression stimulated by PDGF during early time points was not accompanied by a corresponding decrease in total cellular inhibitory activity observed in boiled lysates. Failure to detect modulation of inhibitory activity could be due to the presence of saturating amounts of Kip1 in our mixing experiments. Although titration of nontreated extracts indicated that experiments using these lysates were performed within a linear range of inhibition, boiling of extracts could, in principle, release additional stores of inhibitor if a pool of Kip1 was sequestered in density-arrested cells. Dose response studies indicated that 100 g of extract from nontreated quiescent Balb/c 3T3 cells were sufficient to maximally inhibit cyclin A-dependent kinase activity immunoprecipitated from 100 g of proliferating cell extract (Fig. 5, A and B). However, an identical inhibition of the cyclin A-Cdk complex was obtained using only 50 g of heat-treated cell extracts. These results suggest an excess of Kip1 exists in quiescent Balb/c 3T3 fibroblasts; however, approximately half of the functional p27 pro-tein was sequestered in an inactive state.
Since many Cdk inhibitors were originally isolated by virtue of their ability to bind cyclin-Cdk complexes, potential candidate molecules which might sequester p27 in quiescent fibroblasts include the cyclin and Cdk proteins themselves. To test this possibility, extracts of nonstimulated Balb/c 3T3 cells were incubated with either normal rabbit serum or antibodies to cyclin E, cyclin A, or Cdk4 (Fig. 5C). The antibodies were then immobilized on protein A-agarose beads and removed from the lysate by centrifugation. Both the boiled supernatant and the boiled eluate of the immunoprecipitated pellet were then assayed for inhibition toward a cyclin A-Cdk complex. Of the antibodies used, the Cdk4 immunoprecipitate was found to contain the highest level of Cdk inhibitory activity, while a smaller portion of inhibitor was associated with cyclin E. The amount of inhibitory activity released from the Cdk4 complex after heat treatment was comparable to the activity remaining in the supernatant after immunodepletion of the Cdk4 kinase. These results suggest that the majority of sequestered inhibitor in density-arrested Balb/c 3T3 fibroblasts is associated with Cdk4.
In quiescent Balb/c 3T3 fibroblasts, approximately 50% of the total cellular Kip1 protein is sequestered, while the remainder exists in a free active state. However, within 12 h after stimulation with PDGF, all remaining p27 is sequestered. Previously we have demonstrated that expression of the growth regulatory Cdk4 subunit, cyclin D 1 , is induced by PDGF during this time period (21). As cyclin D 1 has been reported to associate with the Kip1 inhibitor both in vitro (8) and in vivo (34), we examined the role of D 1 during PDGF-dependent down-regulation of free inhibitory activity. While Cdk4 levels did not fluctuate during cell cycle progression, cyclin D 1 was undetectable in quiescent cells and increased dramatically upon growth factor stimulation (Fig. 6A). D 1 expression was elevated within 3 h after exposure to PDGF/PPP, coincident with the onset of a decline in free Cdk inhibitory activity. Further increases in cyclin D 1 levels were inversely proportional to the reduction of Cdk inhibition until after 12 h when D 1 levels decreased. Immunoprecipitation of D 1 after a 12-h exposure to PDGF coprecipitated both the Cdk4 kinase and p27 (Fig. 6B). These results suggest that the cyclin D 1 -Cdk4 complex may regulate the availability of functional Kip1 protein in growth-stimulated fibroblasts.
To examine the effect of a cyclin D/Cdk4/Kip1 interaction on free inhibitory activity, cyclin D 1 and Cdk4 were immunoprecipitated from extracts of cells stimulated with PDGF or PDGF/ PPP for various times. Both the supernatant and the pellet were then boiled and assayed for inhibition of cyclin A-depend-

FIG. 5. Inhibitor activity is sequestered by Cdk4 in quiescent Balb/c 3T3 cells.
A, various amounts of boiled or nonboiled (no pretreat) quiescent cell extracts were incubated with 100 g of proliferating cell extract (stimulated 21 h with 25 ng/ml PDGF-BB and 10% PPP). Cyclin A was immunoprecipitated from mixed extracts, and histone H1 kinase activity was determined. Control extracts of proliferating cells (Ϫ, lane 1) were incubated with lysis buffer. B, cyclin A-dependent kinase activity measured in A was quantitated using a PhosphorImager and graphed as a percentage of maximum activity. Data points represent the average of three separate experiments. C, normal rabbit serum (NRS) or antibodies to cyclin E, cyclin A, or Cdk4 immobilized on protein A-agarose beads were used in immunoprecipitation of quiescent cell extracts. The beads and the supernatant were separated, boiled, and assayed for inhibition of cyclin A-associated kinase activity.
FIG. 6. Association of Kip1 and Cdk4 with cyclin D 1 . A, densityarrested Balb/c 3T3 cells were stimulated with 25 ng/ml PDGF-BB and 10% PPP. At the times indicated, whole cell extracts were isolated, immunoblotted, and probed with polyclonal antibodies to Cdk4 and cyclin D 1 . B, quiescent Balb/c 3T3 cells were stimulated with PDGF-BB (25 ng/ml) in the presence and absence of 10% PPP. 12 h after stimulation, extracts were prepared and cyclin D 1 was immunoprecipitated with a monoclonal antibody covalently linked to agarose. Immunoprecipitates were resolved on a 10% SDS-polyacrylamide gel, immunoblotted, and probed sequentially with polyclonal antibodies to cyclin D 1 , Cdk4, and Kip1. ent kinase activity. As expected, heat treatment of either the Cdk4 or cyclin D 1 immunoprecipitates released an activity that efficiently inactivated the cyclin A-Cdk complex (Fig. 7A). Cdk inhibitory activity was complexed with the Cdk4 kinase for at least 21 h after exposure to PDGF. However, a marked reduction in the amount of inhibitor bound to Cdk4 was observed in cells stimulated in the presence of plasma, particularly at the later time points of 18 and 21 h. The identical phenomenon was observed when cyclin D 1 was immunoprecipitated from cells treated with PDGF and PPP. We next measured the inhibitory activity remaining in the supernatant after immunodepletion of Cdk4 or cyclin D 1 . Although a considerable amount of inhibitory activity associated with Cdk4 in quiescent fibroblasts, sufficient inhibitory activity remained in the lysate after the removal of either Cdk4 or D 1 to maximally inhibit the cyclin A/Cdk enzyme (Figs. 5C and 7B). In contrast, Cdk inhibition was decreased when cyclin D 1 or Cdk4 was depleted from extracts of PDGF-treated cells. However, the most striking reduction of inhibitory activity was observed after D 1 or Cdk4 was cleared from lysates of fibroblasts that were stimulated with PDGF in the presence of plasma (Fig. 7C). These data demonstrate that, while Cdk4 interacts with p27 in both quiescent and stimulated fibroblasts, the cyclin D 1 -Cdk4 complex has a greater effect on Kip1 availability in PPP-treated cells due to the lower abundance of the inhibitor under this condition.
Interaction of Kip1 with Cyclin E in Growth-stimulated Cells-Our results suggest that, in the absence of plasma, p27 is sufficiently abundant to bind and inactivate cyclin-Cdk complexes assembled in response to PDGF stimulation. To determine whether Kip1 associated with cyclin E-dependent kinases in growth-stimulated fibroblasts, we immunoprecipitated cyclin E from extracts of cells treated with PDGF or PDGF/PPP. Immunoprecipitates were then immunoblotted and probed for the presence of p27 (Fig. 8). While Kip1 coprecipitated with inactive cyclin E-Cdk complexes in both quiescent and PDGFtreated cells, nearly undetectable levels of Kip1 complexed with the active cyclin E/Cdk holoenzyme in cells exposed to both PDGF and plasma. Consistent with this observation, a high level of Cdk inhibitory activity could be dissociated from cyclin E protein immunoprecipitated from lysates of PDGF-treated cells, while considerably less inhibitory activity interacted with cyclin E after exposure to plasma. These results support the hypothesis that most of the Kip1 protein remaining in PDGF/ PPP-stimulated cells is titrated away from Cdk2 by the cyclin D 1 -Cdk4 complex.
Effect of cAMP on Growth Factor-dependent Regulation of Kip1-Previously it has been shown that growth-stimulated macrophages treated with cAMP analogs arrest during mid G 1 due, at least in part, to a failure to down-regulate Kip1 expression (26). Balb/c 3T3 fibroblasts are also growth-inhibited by chronic elevation of intracellular cAMP levels (35). To examine the effect of cAMP on cyclin A/Cdk2 inhibitor activity in 3T3 cells, density-arrested fibroblasts were treated with PDGF or PDGF/PPP in the presence and absence of the cAMP-inducing agent cholera toxin and the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine. Whole cell extracts of stimulated fibroblasts were immunoblotted and probed with antibody to cyclin D 1 (Fig. 9). PDGF-mediated expression of cyclin D 1 was severely reduced by the addition of cAMP inducing agents to the stimulation medium. In contrast, elevation of cAMP levels did not prevent the normal decline in Kip1 protein levels after a 12-h exposure to PDGF (data not shown). In addition, cAMP antagonized the growth factor-dependent elimination of free Cdk inhibitory activity. These results suggest that cAMP does not interfere with the early component of Kip1 down-regulation in Balb/c 3T3 fibroblasts. However, because these cells eventu-

FIG. 7. Association of Cdk inhibitory activity with Cdk4 and D 1 in mitogen-stimulated cells.
A, density-arrested Balb/c 3T3 cells were stimulated with PDGF (25 ng/ml) in the presence and absence of 10% PPP. At the times indicated, whole cell extracts were isolated, and cyclin D 1 and Cdk4 were immunoprecipitated. Antibodies were immobilized on protein A-agarose beads and removed from the lysate by centrifugation. Pelleted beads were washed extensively and heated to 100°C. Eluate of the boiled pellet was assayed for inhibition of cyclin A-dependent kinase activity. B, supernatants of cyclin D 1 and Cdk4 immunoprecipitaions described in A were boiled and assayed for inhibition of cyclin A/kinase activity. Immunoprecipitations of cyclin D 1 and Cdk4 were performed on extracts of quiescent cells and cells that were stimulated for 21 h. C, kinase activity measured in A and B at 0 and 21-h time points was quantitated using a PhosphorImager. Data are expressed as a percentage of the decrease from control activity immunoprecipitated from proliferating cell extracts mixed with lysis buffer .   FIG. 8. Cyclin E associated with p27 kip1 in quiescent and PDGF treated cells. Density-arrested cells were treated for 15 h with PDGF (25 ng/ml) with and without 10% PPP. Equal amounts of protein (10 g) from extracts of quiescent cells, cells stimulated with PDGF, and cells treated PDGF with plasma were analyzed for the amount of cyclin E, and 20 g of extract were used to determine cyclin E-Cdk2 histone H1-associated activity as indicated. The total p27 kip1 level was also determine by direct Western analysis of the extracts (10 g) along with the p27 kip1 protein found to be associated with cyclin E by immunoprecipitation of cyclin E (in 40 g of protein) followed by Western analysis using anti p27 kip1 antibody. The bottom panel shows the amount of cyclin A-associated histone kinase inhibitory activity that was associated with the immunoprecipitated cyclin E (in 40 g of protein) as determined after boiling the cyclin E immunoprecipitate.
ally leak through the cAMP-mediated block after prolonged periods of growth arrest, the effect of cAMP on the critical plasma-dependent reduction in Kip1 at later time points could not be accurately assessed. DISCUSSION Progression through the Balb/c 3T3 cell cycle is regulated by the sequential and synergistic action of PDGF and plasmaderived progression factors (34). While the downstream targets of these mitogens and their cognate receptors are incompletely defined, it is clear that the growth regulatory pathways activated by these factors must ultimately impinge upon the cyclindependent kinases and their cyclin subunits. Previously we have demonstrated that PDGF and other competence agents which govern the G 0 /G 1 transition directly engage the cell cycle machinery via modulation of cyclin D 1 expression (21). Here it is shown that PDGF and plasma factors cooperatively induce the cyclin E-and cyclin A-dependent kinase activities required for traverse of late G 1 and the initiation of DNA replication. Activation of the PDGF receptor resulted in a limited induction of cyclin E and cyclin A expression that was not sufficient to overcome the threshold of inhibition by p27 kip1 and allow kinase activation. However, addition of plasma to PDGF-treated cells stimulated maximal cyclin expression and an overall reduction of Kip1 levels, thereby promoting the activation of cyclin associated Cdks. Thus, distinct proliferative signals from both PDGF and PPP converge upon common targets which regulate cell cycle progression from G 1 into S phase in 3T3 fibroblasts. These results provide a molecular basis of how competence and progression factors might synergistically stimulate cell growth through unique modulation of the activities of specific Cdk kinases during the traverse of G 1 .
The amount of Kip1 available to bind and inactivate cyclin Eand cyclin A-associated kinases was regulated in a mitogen-dependent fashion by at least two distinct mechanisms: 1) active Kip1 protein was sequestered by Cdk4 which repressed inhibition toward cyclin A-Cdk complexes in lysate mixing experiments, and 2) total protein levels of Kip1 were decreased. Reduction of Kip1 expression occurred in two phases that were differentially regulated by PDGF and PPP. Exposure of quiescent fibroblasts to PDGF stimulated a moderate decline in p27 levels that began within 6 h of mitogen stimulation. Although in several cell types Kip1 expression is elevated in response to antiproliferative signals such as TGF-␤ (33), this early component of Kip1 elimination was not affected by inhibitors of Balb/c 3T3 cell growth such as cAMP and cycloheximide. However, the PDGF-mediated removal of free Cdk inhibitory activity was prevented under conditions of protein synthesis inhibition. Therefore, the down-regulation of Kip1 levels achieved after treatment with PDGF alone was not sufficient to ablate inhibition of the cyclin A/Cdk enzyme as determined in the in vitro assays. In contrast, stimulation of PDGF-treated cells with PPP resulted in a more pronounced decline in p27 expression, particularly at later time points when cyclin A-associated kinase activity was maximal. Thus, the greatest decrease in p27 was observed under conditions that stimulated DNA synthesis. This plasma-dependent reduction of Kip1 levels, together with the PDGF-mediated inactivation of Kip1 by Cdk4, critically limited the interaction of Cdk inhibitors with cyclin E and cyclin A kinase partners.
Dissociation of Kip1 from labile proteins after heat treatment revealed that enough inhibitor was present to reduce cyclin A-dependent kinase activity by 70 -80%, even after maximal down-regulation of Kip1 expression in plasma-treated cells. Therefore, sequestering of free Kip1 protein after growth factor stimulation is likely to be essential for the activation of cyclin E-and cyclin A-associated kinases. Previously it has been shown that cyclin D-Cdk4 complexes compete with cyclin E-Cdk2 and cyclin A-Cdk2 for binding of the Kip1 inhibitor in vitro (30). These data suggest that growth factor-mediated assembly of cyclin D/Cdk4 holoenzymes and consequent association with p27 may facilitate activation of Cdk2 later in the cell cycle. Consistent with this hypothesis, treatment of epithelial cells with TGF-␤ elevates the synthesis of p15 which displaces Kip1 from Cdk4. p27 kip1 is then available to bind and inhibit Cdk2 (33). In Balb/c 3T3 fibroblasts stimulated with PDGF either in the presence or absence of plasma, a substantial proportion of Kip1 was associated with Cdk4. As a consequence of this Kip1/Cdk4 interaction, the cellular pool of inhibitory activity was depleted by greater than 50%. Our results suggest that Cdk4 is an integral component of a mitogenstimulated feed-forward mechanism which promotes activation of Cdk2 in Balb/c 3T3 cells.
The affinity of Cdk4 for Kip1 in vitro is increased by association with a cyclin subunit (36). Cdk4 assembles combinatorially with three-dimensional-type cyclins which are differentially induced in various cell types. Transcripts for all three of the D cyclins are expressed during the G 1 phase of the Balb/c 3T3 cell cycle; however, only cyclin D 1 is up-regulated in response to PDGF (21). The kinetics of cyclin D 1 increase temporally correlated with a reduction in free inhibitor levels after PDGF stimulation. Furthermore, D 1 expression was first detected during a window of time when down-regulation of inhibitory activity is absolutely dependent on new protein synthesis. As cyclin D 1 levels increased, removal of Cdk inhibitory activity became less sensitive to protein synthesis inhibition. Immunoprecipitation of cyclin D 1 during peak expression coprecipitated Kip1 protein, and boiling of the immunoprecipitate released a considerable amount of Cdk inhibitory activity. Comparison of D 1 and Cdk4 immunoprecipitates revealed a nearly identical pattern of association with the inhibitor, suggesting that Cdk4 modulation of Kip1 availability in growth-stimulated cells was primarily effected by complexes containing cyclin D 1 .
However, down-regulation of Cdk inhibition may not be strictly dependent on cyclin D 1 . Treatment of Balb/c 3T3 cells with cAMP-inducing agents inhibited D 1 expression, but only weakly antagonized the removal of free Cdk inhibitory activity FIG. 9. Agents that elevate cAMP inhibit the induction of cyclin D 1 . Cultures of Balb/c 3T3 cells were treated with PDGF (25 ng/ml) with or without 10% PPP in the presence or absence of cholera toxin (CT, 0.5 g/ml) and 3-isobutyl-1-methylxanthine (IBMX, 10 M). After 12 h of treatment cells were harvested, extracts were prepared, and Western analyses were performed with antibodies to cyclin D 1 (panel A). Panel B shows the amount of inhibitor activity present in the extracts prepared for panel A.
after PDGF stimulation. Presently, it is not known whether Cdk4 expressed under these conditions sequesters Kip1 in association with another cyclin partner, or whether the small amount of D 1 induced in the presence of cAMP is sufficient to modulate Kip1 availability. However, Cdk4 was found to bind a large amount of inhibitory activity in quiescent cells despite the absence of cyclin D 1 protein. Transcripts for both cyclin D 2 and cyclin D 3 are relatively abundant in density-arrested Balb/c 3T3 fibroblasts (21), and cyclin D 2 -Cdk4 complexes compete more effectively for Kip1 binding in vitro than do cyclin D 1 -Cdk4 complexes (36). Thus, the Cdk4/p27 interaction may be directed by various cyclin partners during different stages of the cell cycle. One consequence of Kip1 association with Cdk4 in quiescent cells may be to maintain Cdk4 in an inactive state until normal cell cycle progression is initiated in response to growth factor stimulation.