The Cyclin-dependent Kinase Inhibitor p21 cip1 Mediates the Growth Inhibitory Effect of Phorbol Esters in Human Venous Endothelial Cells*

Long-term application of the phorbol ester phorbol 12,13-dibutyrate (PDBu) inhibits the proliferation of human venous endothelial cells. The cyclin-dependent kinase inhibitor p21 cip1 is a potential candidate mediating the PDBu-induced delayed entry of the cells into S-phase (by ∼10 h when compared with cells stimulated with basic fibroblast growth factor (bFGF)). Levels of p21 cip1 (protein and mRNA) rapidly rise (within ∼2 h) in endothelial cells treated with the active isomer β-PDBu, but not with α-PDBu; this effect is blocked by the mitogen-activated protein kinase kinase-1 (Mek1) inhibitor PD098059 and by the protein kinase C (PKC) antagonists GF109203X and rottlerin (selective for PKC-δ), but not Gö 6976 (selective for Ca2+-dependent PKC isoforms). Rapamycin blocks the PDBu-induced accumulation of p21 cip1 (but not of the cognate mRNA), indicating an action of PKC on p21 cip1 mRNA translation. If endothelial cells are recruited into the cell cycle by bFGF, p21 cip1 mRNA and protein levels rise initially (within 2 h) and decline subsequently such that p21 cip1 drops to a minimum prior to the initiation of DNA synthesis (i.e. after ∼12 h). In bFGF-stimulated cells, changes in p21 cip1 mRNA and protein are strictly linked. In contrast, the levels of p21 cip1 mRNA decline substantially (>10 h) before the protein decreases in PDBu-stimulated cells. Thus, PKC (presumably PKC-δ) regulates the amounts of p21 cip1 in endothelial cells at the level of mRNA accumulation and translation, leading to a rapid and robust induction; following persistent PKC activation, p21 cip1 remains elevated despite reduced mRNA levels, indicating an enhanced stability of the protein. The bFGF-mediated increase in p21 cip1 is blocked by the Mek1 inhibitor, but not by GF109203X; hence, in endothelial cells, induction of p21 cip1 by PKC- and growth factor-dependent signaling is achieved by distinct pathways that converge and require activation of the mitogen-activated protein kinase cascade. The β-PDBu-induced delayed S-phase entry and drop in p21 cip1 are reversed if GF109203X is added 4 h after β-PDBu to prevent persistent PKC activation. These observations indicate a cause and effect relation between sustained p21 cip1 elevations and the delay in S-phase entry induced by β-PDBu.

The addition of phorbol esters to the culture medium inhibits the growth of endothelial cells (1). This observation is difficult to reconcile with the fact that stimulation of the cellular targets of phorbol esters, i.e. the isoforms of protein kinase C, leads to sustained activation of several signaling pathways that are required for the recruitment of quiescent cells into the cell cycle. These include activation of the MAP 1 kinase cascade, which results in stimulation of gene transcription (2,3), as well as the p70 S6 kinase, which enhances the efficiency of mRNA translation (4,5). Previous studies have addressed this paradox and have shown that the PKC activator PDBu exerts a bidirectional effect on growth of human vascular endothelial cells; short-term application of phorbol esters efficiently recruits quiescent endothelial cells into the G 1 -phase of the cell cycle, whereas persistent activation of protein kinase C subsequently slows the progression though G 1 and thereby delays entry of the cells into S-phase (6,7); thus, the net effect of continuous treatment with phorbol esters is a suppression of endothelial cell proliferation (1,6).
Signaling pathways that are linked to inhibition of endothelial cell growth are of potential relevance to cancer therapy. Angiogenesis, i.e. the de novo formation of blood vessels from preexisting capillaries, is a prerequisite for the growth of solid tumors and their metastases (8,9) as well as for the chronic inflammatory response that causes cartilage destruction in rheumatoid arthritis (10). Tumor cells, hypoxic tissue, and leukocytes secrete several endothelial mitogens in a paracrine manner. These growth factors (e.g. fibroblast growth factor, vascular endothelial growth factor, adenosine, interleukin-8, and adhesion molecules) stimulate the growth of endothelial cells by recruiting them from G 0 -phase to G 1 -phase via the stimulation of tyrosine kinase receptors and G protein-coupled receptors (8,9,(11)(12)(13). The ideal antiangiogenic regimen ought to block the action of all endothelial mitogens in a reversible fashion. In theory, this may be achieved by inhibiting a signaling pathway onto which all receptor-generated intracellular stimuli converge to drive the cells through the cell cycle, i.e. the components of the cell cycle machinery that are required for progression through G 1 . The delay in S-phase entry induced by phorbol esters in endothelial cells is thus of interest, as persistent application of phorbol esters results in delayed activation of cyclin-dependent kinases in G 1 (5,6). We have therefore compared the effects of phorbol esters and of a physiological stimulus, i.e. bFGF, on regulatory components that impinge on the cyclin-dependent kinases responsible for G 1 -phase progression. Here, we report that in primary cultures of human endothelial cells, continuous protein kinase C activation leads to a sustained elevation of the cyclin-dependent kinase inhibitor p21 cip1 . This effect is likely to account for the delayed S-phase entry and the resulting growth inhibition. 32 P]ATP and [ 3 H]thymidine were obtained from NEN Life Science Products. Recombinant bovine bFGF and histone H1 were obtained from Boehringer Mannheim (Mannheim, Federal Republic of Germany). Endothelial cell growth supplement was from Technoclon (Vienna, Austria); fetal calf serum was from Life Technologies, Inc.; heparin was from Novo-Nordisk (Bagsvaerd, Denmark); and cell culture dishes were from Costar Corp. (Cambridge, MA). The PKC inhibitor GF109203X (bisindolylmaleimide I) was from Calbiochem, and rottlerin and Gö 6976 (12-(2-cyanoethyl)-6,7,2,13,-terahydro-13-methyl-5-oxo-5H-indolo [2,3-a]pyrrolo [3,4-c]carbazole) were from Alexis Corp. (San Diego, CA). Rapamycin was obtained from Research Biochemicals, Inc. (Natick, MA); PD098059, an inhibitor of MAP kinase kinase-1 (14), was from New England Biolabs Inc. (Beverly, MA). Molecular mass standards (covering the range from 14 to 97 kDa) and reagents for SDS-polyacrylamide gel electrophoresis were from Bio-Rad. The reagents for enhanced chemiluminescence detection were purchased from Amersham International (Buckinghamshire, United Kingdom) or Pierce. The following antibodies and antisera were obtained from commercial sources: mouse monoclonal antibodies against p21 cip1 and p27 kip1 from Transduction Laboratories (Lexington, KY); antisera against cyclin D1 and CDK4 from Santa Cruz Biotechnology Inc. (Santa Cruz, CA); and affinity-purified antiserum or monoclonal antibody against the retinoblastoma gene product (pRb) from Santa Cruz Biotechnology Inc. or Oncogene Science Inc. (Cambridge, MA). The cyclin E antiserum (15) was a generous gift of Dr. J. Roberts (Fred Hutchinson Cancer Research Center, Seattle, WA). The p53 oligonucleotide probe was from American Type Culture Collection (Rockville, MD), and the p21 cip1 cDNA was a generous gift of Drs. O. Pereira-Smith and J. Smith (Baylor College of Medicine, Houston, TX). All other materials were purchased from Sigma if not otherwise indicated.

Materials-[␥-
Cell Culture-Human umbilical venous endothelial cells (HUVECs) were isolated according to Jaffe et al. (16). Cords were rinsed twice with phosphate-buffered saline and then incubated for 20 min with 0.2% collagenase in phosphate-buffered saline. The cell suspension was collected and centrifuged, and the cell pellet was resuspended in medium 199 enriched with endothelial cell growth supplement, 20% fetal calf serum (FCS), 100 mg/ml streptomycin, 100 units/ml penicillin, 0.25 mg/ml amphotericin B, and 5 units/ml heparin and plated in 1% gelatincoated culture dishes. The cells were grown at 37°C in a 5% CO 2 humidified incubator. The following day, the medium was changed to remove erythrocytes and was renewed twice a week. After 4 -6 days, cells formed confluent monolayers and were further subcultured; cell plating efficiency after detachment with 0.02% EDTA was Ͼ98%. The obtained cell population consisted of Ͼ95% endothelial cells as verified by their cobblestone morphology and immunofluorescence staining for von Willebrand factor antigen (Dakopatts, Glostrup, Denmark). Cells were used in the second and third passages.
Immunoblots of p21 cip1 , p27 kip1 , and p53-HUVECs were synchronized in G 0 -phase by contact inhibition of the confluent monolayers. Synchronization was verified by fluorescence-activated cell sorter analysis of propidium iodide-stained cells. Cells were detached as described above and replated in medium 199, 2.5% FCS, 100 mg/ml streptomycin, 100 units/ml penicillin, 0.25 mg/ml amphotericin B, and 5 units/ml heparin for at least 5 h, an interval cells need to adhere. The cells were stimulated with bFGF or ␤-PDBu in the absence and presence of various compounds as indicated in the respective figure legends. At the time points indicated on the respective figures, the cells (ϳ5 ϫ 10 5 ) were harvested in 100 l of lysis buffer A preheated to 80°C and containing 50 mM Tris-HCl (pH 7.4), 120 mM NaCl, 1 mM EDTA, and 0.5% Nonidet P-40 for p21 cip1 and p27 kip1 immunoblots. Lysates were vortexed for 30 s, and DNA and insoluble proteins were removed by centrifugation at 20,000 ϫ g for 20 min. For p53 immunoblots, the cells were harvested in 100 l of lysis buffer B containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 10 g/ml leupeptin, 10 g/ml aprotinin, and 1% Nonidet P-40. After 30 min on ice, lysates were cleared by centrifugation as described above. For immunoblot analysis, lysates (20 g of protein) were resolved by SDS-polyacrylamide gel electrophoresis, and proteins were transferred to nitrocellulose. Mouse monoclonal antibodies against p21 cip1 , p27 kip1 , and p53 were used at dilutions of 1:500, 1:1000, and 1:1000, respectively. The immunoreactive bands were visualized by chemiluminescence using a second antibody that was directed against mouse IgG and conjugated to horseradish peroxidase.
Immunoprecipitation of Cyclin D1 and pRb-HUVECs were synchronized in G 0 -phase and stimulated as described above. At each time point, 5 ϫ 10 6 cells were harvested in 0.8 ml of ice-cold lysis buffer C (pH adjusted to 7.0 with NaOH) containing 50 mM Hepes, 50 mM ␤-glycerophosphate, 250 mM NaCl, 5 mM EDTA, 5 mM NaF, 0.2 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 0.5% Nonidet P-40, 10 g/ml leupeptin, and 10 g/ml aprotinin. Lysates were vortexed for 10 min; DNA and insoluble proteins were removed by centrifugation at 20,000 ϫ g for 15 min. Thereafter, the lysates were precleared with protein G-agarose (Oncogene Science Inc.) for 2 h at 4°C; subsequently, 2 g of affinity-purified cyclin D1 antiserum or pRb antibodies were added for 2 h, followed by a 2-h incubation with 20 l of protein G-agarose. After six washes in lysis buffer C, the immunoprecipitated material was released by boiling in Laemmli sample buffer and resolved on an 8% SDS-polyacrylamide gel. Proteins were transferred to nitrocellulose, and antibodies against p21 cip1 and p27 kip1 (see above) or antisera against cyclin D1 (1:300), CDK4 (1:300), and pRb (1:1000) were used for immunostaining. The antigen-antibody complexes were visualized as described above.
Cyclin E-dependent Kinase Activity and Cyclin E Immunoblots-HUVECs were synchronized in G 0 -phase and stimulated as described above. At each time point, 5 ϫ 10 6 cells were harvested by scraping them off in 0.8 ml of lysis buffer D (pH adjusted to 7.5 with NaOH) containing 20 mM Hepes, 50 mM ␤-glycerophosphate, 100 mM NaCl, 2.5 mM EDTA, 1 mM NaF, 0.5 mM Na 3 VO 4 , 0.5 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 10% glycerol, 10 g/ml leupeptin, and 10 g/ml aprotinin. Lysates were vortexed for 10 min; DNA and insoluble proteins were removed by centrifugation at 20,000 ϫ g for 15 min. Samples were incubated with 2 g of cyclin E antiserum at 4°C for 2 h and subsequently with 20 l of protein G-agarose for 2 h. After three washes with lysis buffer D and three washes with a buffer containing 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 2 mM dithiothreitol, 0.01 mM ATP, 100 mM NaCl, 2.5 mM EDTA, 1 mM NaF, 0.5 mM Na 3 VO 4 , 0.5 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 10% glycerol, 10 g/ml leupeptin, 10 g/ml aprotinin, and 5 g/ml histone H1, the samples were incubated at 37°C for 20 min in the same buffer with the addition of 2 g of histone H1 and 10 Ci of [␥-32 P]ATP per sample. Thereafter, the samples were boiled in Laemmli sample buffer and resolved on a 10% SDS-polyacrylamide gel. The gel was dried and exposed to Kodak X-Omat film. Aliquots of the immunoprecipitates were also withdrawn for immunoblots with the cyclin E antiserum. The affinity-purified antiserum against cyclin E was used at a dilution of 1:2000 for immunostaining. The antigen-antibody complexes were visualized as described above.
[ 3 H]Thymidine Incorporation-The effect of bFGF, FCS, ␤-PDBu, and GF109203X on [ 3 H]thymidine incorporation into DNA was determined as described previously (17) with minor modifications. Briefly, confluent growth-arrested endothelial cells were detached with EDTA and seeded in 96-well plates (1-2 ϫ 10 4 cells/well) in the presence of medium containing 2.5% FCS, 100 mg/ml streptomycin, 100 units/ml penicillin, 0.25 mg/ml amphotericin B, and 5 units/ml heparin. After 5 h, cells were treated with various compounds as indicated in the respective figure legends. At each time point, 0.1 ml of medium 199 containing [ 3 H]thymidine (0.3-0.5 Ci) was added for 3 or 4 h. At the end of each incubation, the medium was removed, and the cells were detached by trypsinization and lysed by a freeze-thaw cycle. The particulate material was trapped onto glass-fiber filters using a Skatron cell harvester, and the radioactivity retained was measured by liquid scintillation counting. The blank level of [ 3 H]thymidine retained on the filters in the absence of DNA synthesis (determined in the presence of aphidicolin) was Ͻ50 cpm. Assays were done in sextuplicate.
Northern Blots-Total cellular RNA was prepared by homogenizing the cells (4 -6 ϫ 10 6 cells) in guanidinium isothiocyanate, followed by extraction of the homogenate with phenol/chloroform/3-methylbutan-1ol. Aliquots (20 -25 g) of each RNA sample were applied to a 1% agarose gel containing formaldehyde and transferred to Hybond N ® nylon membranes by vacuum blotting. The coding sequence of p21 cip1 was excised from the plasmid pCMVexSdi1 by digestion with BamHI (18); the p21 cip1 and p53 probes were labeled using the Amersham Megaprime labeling kit. Hybridization was performed overnight at 65°C in a solution containing 6 ϫ SSC (SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate), 5 ϫ Denhardt's solution, and 100 mg/ml salmon sperm DNA and the labeled cDNA probe. Blots were washed with 2 ϫ SSC, twice with 2 ϫ SSC and 0.1% SDS, and twice with 0.1 ϫ SSC for 15 min each at 65°C. The membranes were exposed to Kodak X-Omat AR5 films at Ϫ80°C. Each experiment was carried out at least three times on different primary cultures of human endothelial cells.

RESULTS
Expression Pattern of p21 cip1 Protein and mRNA during G 1phase Progression-When applied to quiescent (G 0 ) endothelial cells in a pulsatile manner (1.5 h), the sole addition of ␤-PDBu suffices to promote the entry of 40 -60% of the total cell population into the cell cycle as assessed by propidium iodide staining and fluorescence-activated cell sorter analysis (data not shown). However, if ␤-PDBu is applied continuously, the subsequent onset of [ 3 H]thymidine incorporation into DNA, i.e. S-phase entry, is delayed by ϳ10 h compared with cells that have been treated with physiological stimuli such as bFGF or FCS (Fig. 1A). This phenomenon has been attributed to a phorbol ester-dependent delay in the activation of cyclin-dependent kinases in G 1 -phase. To identify the relevant upstream signaling events that may regulate cell cycle progres-sion, we evaluated the expression of the cyclin-dependent kinase inhibitors p21 cip1 and p27 kip1 . Quiescent endothelial cells were stimulated with either bFGF or PDBu; the expression pattern of the cyclin-dependent kinase inhibitors was assessed at the level of the protein in the subsequent cell cycle (Fig. 1B) and compared with the onset of DNA synthesis (Fig.  1A). In G 0 , endothelial cells contain abundant amounts of p27 kip1 , but only low levels of p21 cip1 (cf. Fig. 2A). At the earliest time point that was assessed after mitogenic stimulation (4 h), the cellular levels of p21 cip1 were elevated in cells treated with bFGF or ␤-PDBu, whereas the levels of p27 kip1 remained essentially unaltered ( Fig. 1B; see also Fig. 2A); the increase in p21 cip1 was maintained throughout G 1 -phase and dropped as cells incorporated [ 3 H]thymidine. In bFGF-stimulated cells, the drop in p21 cip1 was observed after 12 h; as the incorporation of [ 3 H]thymidine started to decline, p21 cip1 reaccumulated (Fig. 1B, lanes labeled bFGF 20ϩ and 24ϩ) and dropped again as DNA synthesis resumed (lane labeled bFGF 28ϩ). In ␤-PDBu-treated cells, p21 cip1 remained elevated up to H]thymidine incorporated was determined as outlined under "Experimental Procedures." Assays were done in sextuplicate. B, in a parallel experiment, cells from the same primary culture (5 ϫ 10 5 cells/dish) were stimulated with 20 ng/ml bFGF ϩ 10% FCS or 1 M ␤-PDBu and were harvested in 100 l of lysis buffer A at the indicated time points. After removal of insoluble proteins, lysates were resolved by SDS gel electrophoresis, and the blots were analyzed using monoclonal antibodies against p27 kip1 (upper row) and p21 cip1 (lower row) as described under "Experimental Procedures." Lane Std, a cellular lysate (20 g) prepared from UV-irradiated human fibroblasts that was used as a standard for p21 cip1 and p27 kip1 . C, growth-arrested synchronized endothelial cells (5 ϫ 10 6 cells/dish) were stimulated with 20 ng/ml bFGF ϩ 10% FCS or 1 M ␤-PDBu. At the indicated time points, cells were harvested, and total RNA was prepared as outlined under "Experimental Procedures"; an aliquot thereof (20 g) was separated on a 1% agarose gel and blotted onto a nylon membrane, which was probed with p21 cip1 . The RNA blots were also reprobed with a p53-and a 28 S RNA-specific probe to verify that the differences observed cannot be accounted for by unequal loading (not shown).
FIG. 2. Induction of p21 cip1 and p53 protein expression and mRNA by bFGF and ␤-PDBu. For the analysis of p21 cip1 protein levels, quiescent endothelial cells (5 ϫ 10 5 cells/dish) were stimulated with 1 M ␤-PDBu or vehicle (A and C) or 20 ng/ml bFGF ϩ 10% FCS (B) and were harvested in 100 l of lysis buffer A for p21 cip1 and p27 kip1 or lysis buffer B for p53 at the indicated time points. After removal of insoluble proteins, lysates were resolved by SDS gel electrophoresis, and the blots were analyzed using a mixture of monoclonal antibodies against p21 cip1 and p27 kip1 (A and B) or against p53 (C). Lane Std, a cellular lysate (10 g) prepared from UV-irradiated human fibroblasts that was used as a standard for p21 cip1 and p27 kip1 . For p21 cip1 mRNA analysis, growth-arrested synchronized endothelial cells (5 ϫ 10 6 cells/ dish) were stimulated with 1 M ␤-PDBu (A and C) or 20 ng/ml bFGF ϩ 10% FCS (B). At the indicated time points, cells were harvested, and total RNA was prepared as outlined under "Experimental Procedures"; an aliquot thereof (20 g) was separated on a 1% agarose gel and blotted onto a nylon membrane, which was probed with p21 cip1 (A and B) or p53 (C). 24 h, and the subsequent decline coincided again with the entry into S-phase (Fig. 1B, lane labeled PDBu 28ϩ). In contrast, stimulation of endothelial cells with ␤-PDBu did not defer but rather accelerated the decline in p27 kip1 (Fig. 1B). Although we have attempted to optimize the synchronization of the cells in G 0 , there was some variability in individual cell batches with respect to the onset (i.e. 11-14 and 25-30 h for bFGF and ␤-PDBu, respectively) and the duration of [ 3 H]thymidine incorporation (cf. also Figs. 1 and 6, A and B). This variation is not unexpected in experiments with primary cell cultures. However, we stress that irrespective of this variation, the decline in p21 cip1 always occurred prior to the onset of DNA synthesis. The changes in the p21 cip1 protein levels of bFGF-stimulated cells were paralleled by appropriate changes in the p21 cip1 mRNA (Fig. 1C); this finding is consistent with a previous report that showed that in a human fibroblast cell line, serum stimulation resulted in an oscillation of the p21 cip1 mRNA along the cell cycle (19). In bFGF-stimulated cells, the decline in the protein levels of p21 cip1 and of its cognate mRNA coincided, indicating that the protein turned over rapidly (Fig. 1,  compare B and C). In contrast, if the cells had been stimulated with ␤-PDBu, p21 cip1 mRNA declined substantially (Ͼ12 h) before the drop in the protein (Fig. 1, compare B and C). This discrepancy between the p21 cip1 mRNA and protein levels can only be accounted for by a ␤-PDBu-induced stabilization of the protein.
Signaling Pathways Involved in the ␤-PDBu-mediated Induction and Regulation of p21 cip1 -To characterize the initial induction of p21 cip1 , endothelial cells were subjected to shortterm incubations in the presence of 1 M PDBu. The addition of the phorbol ester to quiescent cells led to a rapid accumulation of p21 cip1 (Fig. 2A). In contrast, the levels of p27 kip1 , which were readily detectable in quiescent cells, remained unaffected by the short-term incubation with ␤-PDBu. However, an appropriate stimulus did regulate the levels of p27 kip1 in endothelial cells. Incubation of endothelial cells with transforming growth factor-␤, which is known to redistribute cellular levels of p27 kip1 (20), led to the accumulation of p27 kip1 (data not shown). We stress that a virtually identical increase in p21 cip1 was also observed if asynchronously growing cells were stimulated with ␤-PDBu and irrespective of whether total cellular protein or a detergent extract was used for analysis (data not shown). The accumulation of p21 cip1 was paralleled by a rise in the cognate mRNA, which reached a maximum within 2-4 h ( Fig. 2A). The levels of p21 cip1 protein and mRNA also increased if quiescent endothelial cells were stimulated with bFGF ( Fig. 2A), albeit to a lesser extent than following the addition of ␤-PDBu (cf. Fig. 1B). In contrast, we did not detect any ␤-PDBu-induced change in the levels of p53 irrespective of whether the determination was done by immunoblotting cellular lysates or by RNA blots (Fig. 2C).
The experiments depicted in Figs. 1 and 2 suggested that direct activation of protein kinase C by ␤-PDBu resulted in an initial rapid increase in p21 cip1 . Growth factor receptors with tyrosine kinase activity are capable of recruiting phospholipase C␥ via the interaction of the SH2 domain with autophosphorylated tyrosine residues; the resulting activation of phospholipase C is expected to trigger PKC activation (21). We have therefore assessed whether the bFGF-induced increase in p21 cip1 resulted from the activation of a PKC isoform. This was clearly not the case as GF109203X did not block the response to bFGF while blunting the response to ␤-PDBu (Figs. 3A and 4, A and C). In contrast, incubation of endothelial cells with PD09805, an inhibitor of MAP kinase kinase-1 (Mek1; see Ref. 14), blocked the induction of p21 cip1 by both ␤-PDBu and bFGF (Fig. 3A). Hence, the bFGF receptor-and PKC-dependent stim-ulation converge upstream of MAP kinase and depend on the activation of Mek1 to raise p21 cip1 levels in endothelial cells. This response is not restricted to endothelial cells and can also be observed in other cell types, e.g. A431 cells. 2 Previous experiments have verified that PD098059 effectively inhibits MAP kinase activation in endothelial cells at the concentration employed (17). A discrepancy is evident if the amounts of p21 cip1 mRNA and protein are compared in quiescent cells; while the mRNA is readily detectable, the level of the protein is very low. This discrepancy has been noted earlier and has been interpreted as evidence that the cellular concentration of p21 cip1 is presumably regulated not only at the level of mRNA transcription, but also at the level of mRNA translation (22). We have tested this hypothesis by employing the immunosuppressant rapamycin. The p70 S6 kinase regulates the translation of mRNA transcripts, and rapamycin blocks the activation of the p70 S6 kinase irrespective of the upstream signaling pathway employed by the stimulus (5). At concentrations Ն1 nM, rapamycin attenuated the ␤-PDBu-induced increase in p21 cip1 protein levels (Fig. 3B, lanes labeled R), whereas the accumulation of p21 cip1 mRNA was virtually unaffected (Fig.  3C, lanes labeled R); in contrast, the Mek1 inhibitor PD098059 and the PKC inhibitor GF109203X blocked the ␤-PDBu-induced increase in p21 cip1 protein and mRNA levels (Fig. 3, B  and C, lanes labeled PD and GF). Hence, ␤-PDBu raises p21 cip1 levels in endothelial cells not only by inducing the accumulation of the cognate mRNA, but also by regulation at the translational level.
Protein  R0.1, R1, and R10, respectively); 20 M PD098059 (lanes labeled PD); 1 M GF109203X (lanes labeled GF); and vehicle (lane labeled Co) for 2 h and analyzed as described for A. C, in a parallel experiment, cells (5 ϫ 10 6 cells/dish) were incubated as described for B. After 2 h, the cells were harvested, and total RNA was prepared as outlined under "Experimental Procedures"; an aliquot thereof (20 g) was separated on a 1% agarose gel and blotted onto a nylon membrane, which was probed with p21 cip1 . The mRNA blot was also reprobed with a p53-specific probe as a loading control (not shown).
verified that the PDBu-induced increase in p21 cip1 was mediated via activation of a protein kinase C isoform (presumably the ␦-isoform) based on the following criteria. (i) The halfmaximum response to ␤-PDBu was observed in the nanomolar concentration range (Fig. 4A). (ii) The effect of phorbol esters was stereospecific, as the inactive isomer ␣-PDBu failed to affect the levels of p21 cip1 (Fig. 4B). (iii) The addition of GF109203X (bisindolylmaleimide I), a inhibitor of typical (i.e. Ca 2ϩ -dependent) and novel (i.e. Ca 2ϩ -independent) protein kinase C isoforms (23), blunted the effect of ␤-PDBu (Figs. 3A and 4A). We noted that the sole addition of GF109203X caused a modest increase in p21 cip1 . This effect was even more pronounced for other PKC inhibitors such as staurosporine and H-7, which actually enhanced the effect of ␤-PDBu (Fig. 4B). These compounds, however, are less specific as protein kinase C inhibitors, as they also act as potent inhibitors of several other serine/threonine kinases (23). In addition, prolonged in-cubation of endothelial cells with staurosporine (but not with GF109203X) caused the cells to detach from the matrix and promoted cell death (data not shown). (iv) The induction of p21 cip1 was blocked by 10 M rottlerin; this inhibitor is selective for PKC-␦ at this concentration (24). In contrast, Gö 6976, a compound structurally related to GF109203X and staurosporine, which inhibits the ␣and ␤-isoforms of PKC with IC 50 values in the nanomolar range (25), was ineffective up to 1 M (Fig. 4C).
In several human cell lines, p21 cip1 is induced by oxidative stress caused by intracellular glutathione depletion with diethyl maleate, an effect that is prevented by the glutathione precursor N-acetylcysteine (26). Similarly, N-acetylcysteine blocks several effects of phorbol esters (27)(28)(29). We have therefore determined the level of p21 cip1 in the presence of these compounds or, alternatively, stimulated the endothelial cells to generate an endogenous free radical, namely NO. Preincubation with N-acetylcysteine or diethyl maleate neither inhibited nor potentiated, respectively, the increase in p21 cip1 induced by ␤-PDBu (Fig. 4D). Finally, treatment of the cells with the Ca 2ϩ ionophore A23187 (calcimycin), under conditions that lead to a very pronounced activation of NO synthase (30) and subsequently to cell death (data not shown), had no effect on p21 cip1 levels (Fig. 4D).
Functional Relevance of the Phorbol Ester-induced Increase in p21 cip1 -To be functionally relevant for the delay in G 1phase progression, p21 cip1 must form a complex with a cyclindependent kinase that is active in early G 1 -phase. We therefore immunoprecipitated cyclin D1 from lysates of cells that were stimulated with either bFGF or ␤-PDBu. Equivalent amounts of cyclin D1 and of the associated cyclin-dependent kinase CDK4 were recovered (Fig. 5A). However, even at the earliest time point investigated (i.e. 3 h after stimulation; Fig. 5A), these complexes contained higher amounts of p21 cip1 when isolated from lysates of ␤-PDBu-treated cells (Fig. 5A, lanes  labeled IP). While a significant portion of p21 cip1 was coimmunoprecipitated with the cyclin D1 antiserum, only low levels of p27 kip1 were recovered in the complex, and the majority of p27 kip1 was detected in the supernatant from the immunoprecipitation (Fig. 5A, compare lanes labeled IP and SN). At all time points investigated (up to 12 h after stimulation), complexes of cyclin D1 and CDK4 that were immunoprecipitated from ␤-PDBu-treated cells contained higher levels of p21 cip1 than those isolated from bFGF-stimulated cells (data not shown).
These findings predict a reduced activity of the cyclin D1-CDK4 complex. The retinoblastoma protein pRb represents one of the key endogenous substrates of cyclin D1-CDK4, and hyperphosphorylation of pRb is an essential step in G 1 -phase progression prior to S-phase entry (31). This modification is associated with a reduced mobility of pRb on SDS-polyacrylamide gels. We have therefore enriched pRb by immunoprecipitation from cellular lysates and analyzed the electrophoretic mobility of the protein after stimulation of the cells with bFGF and ␤-PDBu. As can be seen from Fig. 5B, the time point at which pRb is phosphorylated to a significant extent is delayed by several hours in ␤-PDBu-stimulated cells. The induction of cyclin E synthesis and the activation of cyclin E-CDK2 are downstream of cyclin D1-CDK4 activation. Accordingly, cyclin E synthesis (data not shown) and cyclin E-dependent kinase activity (as assessed by histone H1 phosphorylation after immunoprecipitation of the complexes) were delayed in PDBustimulated cells when compared with cells that had been recruited into the cell cycle with bFGF (Fig. 5C).
Reversal of the PDBu-induced Delay in S-Phase Entry-Taken together, the findings presented so far indicate that activation of PKC by phorbol esters raises the levels of p21 cip1 to inhibit activation of G 1 -phase cyclins and that the sustained increase in p21 cip1 results in delayed S-phase entry. This interpretation predicts that the effect of ␤-PDBu should override the physiological stimulation by bFGF when both stimuli are presented in combination. In addition, if the sustained PKC activation is prevented, both the delay in S-phase entry and in the drop in p21 cip1 ought to be prevented. Both predictions have been verified. (i) ␤-PDBu, even when added 4 h after stimulation with bFGF, suppressed [ 3 H]thymidine incorporation in endothelial cells (Fig. 6A, Ç), and this effect of ␤-PDBu was reversed by the PKC inhibitor GF109203X (Fig. 6A, Ⅺ). (ii) If GF109203X was added to the endothelial cells 4 h after stimulation by ␤-PDBu, essentially no delay in the onset of DNAsynthesis was observed (Fig. 6B, Ç). If GF109203X was added 12 h after ␤-PDBu, the delay was only partially prevented (Fig.  6B, É); conversely, if the order of addition was reversed, i.e. GF109203X was added before ␤-PDBu, the cells were obviously not recruited into the cell cycle, and no [ 3 H]thymidine incorporation was observed (data not shown). In parallel experiments,

FIG. 5. Immunoprecipitates of cyclin D1 complexes (A), phosphorylation of the retinoblastoma protein (B), and cyclin E-CDK2 activity (C) in bFGF-and ␤-PDBu-treated endothelial cells.
A, quiescent endothelial cells (5 ϫ 10 6 cells) were incubated with 1 M ␤-PDBu (lanes labeled ␤-P) and 20 ng/ml bFGF (lanes labeled bFGF) for 3 h and subsequently harvested in 0.8 ml of ice-cold lysis buffer C. The lysates were incubated with 2 g of cyclin D1 antiserum for 2 h and subsequently immunoprecipitated with protein G-agarose as described under "Experimental Procedures." Immunoprecipitates (lanes labeled IP) and aliquots of supernatants (lanes labeled SN) were resolved by SDS gel electrophoresis; the blots were analyzed using a mixture of monoclonal antibodies against p21 cip1 and p27 kip1 as well as antisera against cyclin D1 and CDK4; IgG indicates the position of the immunoglobulin light chains. B, quiescent endothelial cells (5 ϫ 10 6 cells) were incubated with 1 M ␤-PDBu (lanes labeled ␤-P) or 20 ng/ml bFGF (lanes labeled FGF) for the indicated time periods and subsequently harvested in 0.8 ml of ice-cold lysis buffer C. Lysates were incubated with pRb antiserum (2 g) for 2 h and subsequently immunoprecipitated with protein G-agarose as described under "Experimental Procedures." Immunoprecipitates were resolved by SDS gel electrophoresis; the blots were analyzed using a monoclonal antibody against pRb. ppRb and pRb indicate the migration of hyperphosphorylated and hypophosphorylated pRb, respectively. C, quiescent endothelial cells (5 ϫ 10 6 cells) were incubated with 1 M ␤-PDBu (lanes labeled ␤PDBu) or 20 ng/ml bFGF (lanes labeled bFGF) for the indicated time periods and harvested in 0.8 ml of lysis buffer D. Samples were incubated with cyclin E antiserum (2 g) for 2 h and subsequently immunoprecipitated with protein G-agarose. The kinase reaction with histone H1 as substrate was carried out as described under "Experimental Procedures." Samples were resolved on an SDS-polyacrylamide gel, and the dried gel was exposed to Kodak X-Omat film. Thereafter, the medium was removed, and the cells were detached and lysed. The particulate material was trapped onto glass-fiber filters, and the radioactivity retained was measured by liquid scintillation counting. Assays were done in sextuplicate. B, quiescent endothelial cells (2 ϫ 10 4 cells/well) were stimulated with 20 ng/ml bFGF ϩ 10% FCS (E) or 100 nM ␤-PDBu (Ⅺ, Ç, and É); 3 M GF109203X was added after 4 h (Ç) and 12 h (É). [ 3 H]Thymidine incorporation was measured as described for A. C, in parallel to the experiment depicted in B, endothelial cells (5 ϫ 10 5 cells/dish) were stimulated with 20 ng/ml bFGF ϩ 10% FCS (lanes labeled FGF), 100 nM ␤-PDBu (lanes labeled ␤-P), or 100 nM ␤-PDBu followed by the addition of 1 M GF109203X after 4 h (lanes labeled GF). At the indicated time points, the cells were harvested in 100 l of lysis buffer A. The lysates (30 g) were resolved by SDS gel electrophoresis, and the blots were analyzed using a monoclonal antibody against p21 cip1 . Lane Std, a cellular lysate (30 g) prepared from UV-irradiated human fibroblasts that was used as a standard for p21 cip1 .
we have assessed the pattern of p21 cip1 expression in cells that were first stimulated with ␤-PDBu and subsequently received GF109203X (Fig. 6C). This regimen reversed the delay in the drop in p21 cip1 such that the pattern of expression was comparable to that observed in bFGF-treated cells. The decline in p21 cip1 levels again preceded the onset of DNA synthesis. DISCUSSION Previous studies have demonstrated that induction of p21 cip1 is an important element in the coordinated response that leads to cell cycle arrest following exposure to DNA damage (32)(33)(34). In addition, up-regulation of p21 cip1 has been proposed to participate in cell senescence (18) and in cell cycle withdrawal upon terminal differentiation (20,(35)(36)(37). In this study, we show that in primary cultures of human endothelial cells, the levels of p21 cip1 vary along the cell cycle; p21 cip1 is low in quiescent cells, increases as cells are recruited into G 1 -phase by a mitogenic stimulus, drops prior to the onset of S-phase, and subsequently reaccumulates. This pattern is observed regardless of whether the cells are stimulated with bFGF, a physiological growth factor, or subjected to a long-term incubation with ␤-PDBu. However, bFGF leads to a transient induction of p21 cip1 , the protein turns over rapidly, and the decline in mRNA and protein levels is tightly linked. While the accumulation of p21 cip1 in response to cell damage or growth inhibitors can be rationalized from a teleological point of view, the induction of an inhibitor of G 1 -phase progression by mitogens seems counterintuitive. However, expression of a cyclin-dependent kinase inhibitor may be required for efficient complex formation between CDK4 and cyclin D1 such that low concentrations of p21 cip1 act as an assembly factor, whereas high concentrations inhibit the kinase activity (38). This dual action of p21 cip1 is evident in endothelial cells; in contrast to bFGF, activation of protein kinase C causes a prolonged elevation of p21 cip1 in G 1 -phase. ␤-PDBu elicits this effect by regulating at least three distinct processes, namely (i) the accumulation and (ii) translation of p21 cip1 mRNA; and (iii) upon long-term incubation of the endothelial cells in the presence of the phorbol ester, p21 cip1 is stabilized such that the protein levels remain elevated despite a reduction in the cognate mRNA. The resulting robust and sustained elevation of p21 cip1 is functionally relevant as it accumulates in complexes with cyclin D1-CDK4. Accordingly, the phosphorylation of pRb and the induction of cyclin E-dependent kinase activity are delayed in ␤-PDBu-treated endothelial cells as shown here and described previously (6,7). In contrast, the fluctuations in the levels of p27 kip1 , a cyclin-dependent kinase inhibitor related to p21 cip1 (22,31), do not coincide with the delayed onset of DNA synthesis. The ␤-PDBuinduced delayed drop in p21 cip1 and S-phase entry can be appropriately reversed by the PKC inhibitor GF109203X. This substantiates our conclusion that the two effects are causally related. Although we obviously cannot rule out the participation of additional signaling pathways, the elevation of p21 cip1 and the resulting inhibition of cyclin D1-dependent kinase activity are per se sufficient to account for the delayed progression through G 1 induced by ␤-PDBu.
Our experiments identify PKC-␦ as the candidate isoform that mediates induction of p21 cip1 in response to ␤-PDBu. This interpretation is supported by the observation that GF109203X and rottlerin antagonized the action of ␤-PDBu. GF109203X inhibits classical (Ca 2ϩ -dependent) and novel (Ca 2ϩ -independent) PKC isoforms with nanomolar affinities (23,25); at the concentration at which rottlerin inhibited the effect of ␤-PDBu on the expression of p21 cip1 , the compound is selective for PKC-␦ (IC 50 ϭ 3-6 M), whereas 10 -30-fold higher concentrations are required to inhibit the other Ca 2ϩ -independent as well as classical PKC isoforms (24). In contrast, Gö 6976, which selectively inhibits Ca 2ϩ -dependent PKC isoforms with IC 50 values in the nanomolar range (25), did not affect the ␤-PDBudependent induction of p21 cip1 . H-7 and staurosporine raised the basal levels of p21 cip1 . This was also seen, albeit to a lesser extent, in response to GF109203X. H-7 and staurosporine, which inhibit several protein kinase C isoforms as well as other serine/threonine kinase isoforms (23), have previously been reported to increase the expression of p21 cip1 (39). Contrary to GF109203X, H-7 and staurosporine potentiated the effect of ␤-PDBu. It is at present not clear whether this phenomenon reflects a balance between protein kinase C isoforms, which exert opposite effects on the expression level of p21 cip1 and which have different sensitivities for inhibitors, or the involvement of other kinases.
DNA damage and inhibition of DNA and RNA synthesis cause the accumulation of the tumor suppressor protein p53, which induces the expression of p21 cip1 ; this signaling pathway is understood in considerable detail (see Ref. 40 for review). Several findings, however, argue against a role of p53 in the ␤-PDBu-dependent induction of p21 cip1 ; while p21 cip1 and its mRNA rapidly increased in endothelial cells stimulated with the phorbol ester, the level of p53 remained unaffected. Phorbol esters also raise p21 cip1 levels in human cells that are deficient in p53 (41) or that contain a mutated form of p53 (e.g. A431 cells). 2 Additional p53-independent mechanisms that result in the induction of p21 cip1 are less well characterized; these include oxidative stress (26). However, our experiments rule out that induction of p21 cip1 by phorbol esters can simply be accounted for by a stress-induced response of endothelial cells. Previous experiments indicate that stimulation of quiescent cells with serum-derived growth factors (20,36) or epidermal growth factor (42) leads to a transient accumulation of p21 cip1 . Transient induction of p21 cip1 was also seen in endothelial cells stimulated with bFGF. However, our work clearly shows that bFGF and ␤-PDBu use distinct but convergent signaling pathways to raise p21 cip1 levels; contrary to ␤-PDBu, the effect of bFGF does not depend on the activation of GF109203X-sensitive protein kinase C isoforms. Nevertheless, induction of p21 cip1 by both ␤-PDBu and bFGF is abolished by the Mek1 inhibitor PD098059. This compound has also recently been shown to suppress the nerve growth factor-stimulated accumulation of p21 cip1 in NIH3T3 cells expressing the nerve growth factor receptor TRKA (43). Induction of p21 cip1 is also seen after transfection with inducible forms of Raf-1, the kinase that is upstream of Mek1 (43,44). Hence, we conclude that in endothelial cells, the signals elicited by physiological mitogens and protein kinase C activation converge at the level of the MAP kinase cascade and require its activation to induce the expression of p21 cip1 . However, this pathway is clearly not universally utilized in all cell types; for instance, in the monocytic cell line U937, stimulation of MAP kinase fails to raise p21 cip1 levels, and the induction by phorbol esters depends on the transcription factor Sp1 (45). Furthermore, contrary to human endothelial cells, where PKC-␦ mediates the phorbol ester-dependent accumulation of p21 cip1 , it is the ␣-isoform of PKC that raises p21 cip1 levels and that induces a growth arrest in a human epithelial cell line (46). Taken together, these data show that the mechanisms by which induction of p21 cip1 is achieved are distinct in different cells, and this diversity may thus also be exploited to selectively arrest a given cell type in an intact organism. Cellular immortalization is associated with abrogation of one or several regulatory pathways that are required for the normal control of the cell cycle. We stress that our experiments were carried out on primary cultures of human endothelial cells. Thus, we believe that the observations in this study are relevant to the normal endothelial cell cycle in vivo.