Cyclin-dependent Kinases Are Inactivated by a Combination of p21 and Thr-14/Tyr-15 Phosphorylation after UV-induced DNA Damage*

Transcription

1 THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 271, No. 22, Issue of May 31, pp , by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A. Cyclin-dependent Kinases Are Inactivated by a Combination of p21 and Thr-14/Tyr-15 Phosphorylation after UV-induced DNA Damage* (Received for publication, January 16, 1996, and in revised form, March 21, 1996) Randy Y. C. Poon, Wei Jiang, Hideo Toyoshima, and Tony Hunter From The Salk Institute for Biological Studies, La Jolla, California The cyclin-dependent kinase (CDK) inhibitor p21 is induced by the tumor suppressor gene product p53 and is thought to be important for the arrest of the cell cycle following DNA damage. Here we have investigated the contribution of p21 in inhibiting different cyclin-cdk complexes that drive different cell cycle transitions following UV irradiation-induced DNA damage in normal human fibroblasts and immortalized rodent fibroblasts. When cells were exposed to a low dose of UV irradiation, both p53 and p21 were induced; the protein kinase activities associated with Cdc2, Cdk2, and Cdk4 were inhibited; and there was a good correlation between their inhibition and binding to p21. p21 alone is likely to be sufficient for the inhibition of Cdk2 because all the cyclin-complexed forms of Cdk2 were associated with p21 after irradiation. In contrast, only a small proportion of Cdk4 and Cdc2 was complexed with p21, although the level of Cdk4 associated with either p21 or p27 was increased after irradiation. Furthermore, recombinant p21 added to an unirradiated cell lysate at the same level as that induced by irradiation damage inhibited only the kinase activity associated with Cdk2. Cdc2 is likely to be inhibited by Thr-14/Tyr-15 phosphorylation after irradiation because Cdc2 was tyrosine-phosphorylated, and recombinant Cdc25 was able to increase its kinase activity significantly. Taken together, these results suggest that different CDKs are inhibited by different mechanisms following UV-induced DNA damage: Cdk2 is inhibited by the elevated level of p21; Cdk4 is inhibited by cooperation of p21 with other CDK inhibitors, like p27, and possibly by phosphorylation; and Cdc2 is inhibited by Thr-14/Tyr-15 phosphorylation. It is likely that these underlying mechanisms that inactivate CDKs are similar for other kinds of DNA damage. Cyclins and cyclin-dependent kinases (CDKs) 1 are key regulators of the eukaryotic cell cycle. Cdc2 is associated with B-type cyclins and regulates M phase (1). Cdk2 is associated with A- and E-type cyclins, and the respective complexes are * This work was supported by United States Public Health Service Grants CA14195 and CA39780 (to T. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Fellow of the International Human Frontier Science Program. Fellow of the American Cancer Society. Fellow of the Leukemia Research Foundation. American Cancer Society Research Professor. To whom correspondence should be addressed: Salk Inst. for Biological Studies, North Torrey Pines Rd., La Jolla, CA Tel.: (ext. 1385); Fax: The abbreviations used are: CDKs, cyclin-dependent kinases; PCNA, proliferating cell nuclear antigen; NRK, normal rat kidney; FACS, fluorescence-activated cell sorter; PAGE, polyacrylamide gel electrophoresis; GST, glutathione S-transferase; Rb, retinoblastoma gene product; Ab, antibody believed to control S phase and the G 1 -S transition, respectively (2, 3). Cdk4 and Cdk6 are associated with the D-type cyclins and are important for G 1 progression (3). The activity of CDKs is tightly regulated by an intricate system of protein-protein interaction and phosphorylation. Association with a cyclin subunit is absolutely required for CDK activation. The activity of the kinase holoenzyme is increased by phosphorylation of Thr-161 and inhibited by phosphorylation of Thr-14 and Tyr-15. Thr-161 is phosphorylated by the CDK-activating kinase (reviewed in Ref. 4) and, at least in Cdk2, can be dephosphorylated by the CDK-interacting phosphatase KAP (5). The Wee1 protein kinase can phosphorylate Tyr-15 of Cdc2 (reviewed in Ref. 4), and the membrane-associated Wee1-related kinase Myt1 can phosphorylate both Thr-14 and Tyr-15 (6). Both Thr-14 and Tyr-15 can be dephosphorylated by the Cdc25 protein phosphatases (reviewed in Ref. 4). The activity of CDKs is also negatively regulated by binding to protein inhibitors; these CDK inhibitors fall into two families based on sequence homology. One CDK inhibitor family includes p15 INK4B (7), p16 INK4A (8), p18 (9), and p19 (10); this family of CDK inhibitors contains ankyrin repeats and is specific for cyclin D-CDK complexes (reviewed in Ref. 11). Recently, p19 ARF, a protein encoded by an alternative reading frame of INK4A, was shown to be able to arrest the cell cycle (12). The other CDK inhibitor family includes p21 (also called CAP20, CIP1, Pic1, Sdi1, and WAF1), p27 (Kip1) (13, 14), and p57 (Kip2) (15, 16). The archetypal member of this family, p21, was isolated as a Cdk2-associated protein and an inhibitor of Cdk2 (17 19), a tumor suppressor gene product p53-inducible gene (20), and a gene whose RNA expression is increased in senescent cells (21). In addition to being induced by p53, p21 is also induced during differentiation by p53-independent pathways (22 31). In contrast to the p16 family, p21 can bind and inhibit a wider spectrum of CDKs, including Cdk2, Cdk4, and, weakly, Cdc2 (32). The exact mechanism through which p21 inhibits CDKs is unknown; there is evidence suggesting that when one molecule of p21 associates with a single cyclin-cdk heterodimer, the complex is still active and that more than one molecule of p21 is required to inhibit a cyclin-cdk complex (32, 33). As well as directly inhibiting the kinase activity of cyclin-cdk complexes, p21 also blocks the phosphorylation of the activating threonine residue in CDKs by CDK-activating kinase (34). Residues of p21, p27, and p57 are related in sequence, and this region has been identified as the cyclin- and CDK-binding region. p21 can also associate with PCNA (35) and inhibit DNA synthesis (21, 36, 37), but not DNA repair in vitro (38, 39). The PCNA-binding region is distinct from the cyclin-cdk-binding region and lies at the very C terminus of p21 (40 44). Either the CDK- or the PCNA-binding region of p21 is sufficient to inhibit DNA replication (40, 42). When cellular DNA is damaged, continued progression through the cell cycle and DNA replication are undesirable

2 13284 Mechanism of UV-induced Cell Cycle Arrest because this would lead to increased DNA mutagenesis and contribute to tumorigenesis. Two options adopted by damaged cells are either to stop the cell cycle, thus allowing time for the DNA to be repaired, or to eliminate the damaged cells by apoptosis. Both cell cycle arrest (45) and apoptosis (46 48) after ionizing irradiation are dependent on p53 function. The level and activity of p53 are increased following DNA damage, but exactly how the cell detects the damaged DNA and sends a signal to regulate p53 is unclear. Once activated, p53 then induces the expression of p21, which in turn may be responsible for the cell cycle arrest. About 50% of all tumors contain p53 mutations (49); hence, it is likely that the failure to induce p21 in cells with mutated p53 is one of the key mechanisms of tumorigenesis. It has long been known that UV irradiation stimulates the accumulation of p53 (50) and that the p53 response induced by UV irradiation differs from that induced by ionizing irradiation (51). There is a good correlation between radiation-induced DNA damage and induction of p21 (52 55). Whether p21 is sufficient to account for the inhibition of cyclin-cdk activities after DNA damage is clearly a key question. Mice that lack p21 develop normally without observable defects up to 7 months (56), which is in marked contrast to mice that lack p53, which develop tumors spontaneously (57). Nonetheless, p21 / embryonic fibroblasts are significantly impaired in their ability to arrest in G 1 following DNA damage by ionizing irradiation (56, 58, 59), indicating that p21 is important, but not totally sufficient for the DNA damage checkpoint. In contrast, the lack of p21 does not affect p53-dependent apoptosis in thymocytes (56, 58). Here we set out to investigate the contribution of p21 to the inhibition of different cyclin-cdk complexes after UV-induced DNA damage in fibroblasts. We found that p21 is probably only sufficient for the inhibition of Cdk2, whereas Thr-14/Tyr-15 phosphorylation is likely to be critical for Cdc2 inhibition. In this study, we have used normal human diploid fibroblast cell strains (AG1523 and HSF8), mouse Swiss 3T3 fibroblasts, and normal rat kidney (NRK) fibroblasts, which are all relatively normal cell lines containing functional p53. Essentially the same results were obtained with these different cells, and most of the data presented here were obtained with AG1523 cells. MATERIALS AND METHODS Cell Culture Normal human foreskin AG1523 diploid fibroblasts were obtained from the National Institute on Aging Cell Repository, Institute for Medical Research (Camden, NJ). Normal human HSF8 fibroblasts were cultured from human neonatal foreskin. 2 Mouse Swiss 3T3 fibroblasts and NRK cells were obtained from the American Type Culture Collection (Rockville, MD). Cells were grown in Dulbecco s modified Eagle s medium supplemented with 10% (v/v) fetal calf serum in a humidified incubator at 37 C in 10% CO 2. AG1523 and HSF8 cells were used between passages 8 and 25. Cell-free extracts were prepared as described (60). The protein concentration of cell lysates was measured with a bicinchoninic acid protein assay system (Pierce) using bovine serum albumin as a standard. For fluorescence-activated cell sorter (FACS) analysis, cells were harvested, fixed in 70% ethanol, and stained with propidium iodide prior to FACS analysis using a FACScan machine (Becton Dickinson Advanced Cellular Biology). Metabolic Labeling of Cells Nearly confluent cells were washed twice with phosphate-buffered saline and incubated in methionine- and cysteine-free Dulbecco s modified Eagle s medium containing 10% (v/v) dialyzed fetal calf serum and 0.1 mci/ml [ 35 S]methionine/cysteine (Du- Pont NEN) for 12 h in a humidified incubator at 37 C in 10% CO 2.To measure the half-life of p21, the labeled cells were washed twice with phosphate-buffered saline, irradiated with the indicated dose of UV (see below), and incubated with Dulbecco s modified Eagle s medium containing 10% (v/v) dialyzed fetal calf serum. Cell-free extracts were 2 W. Jiang, unpublished results. prepared at the indicated times as described above and immunoprecipitated with anti-p21 antibodies. The samples were analyzed by SDS- PAGE, and the amount of 35 S-labeled p21 was quantitated with a PhosphorImager (Molecular Dynamics, Inc.). UV Irradiation of Cells Cells were irradiated with UV essentially as described by Lu and Lane (51). Cells were grown to near confluence in Dulbecco s modified Eagle s medium supplemented with 10% (v/v) fetal calf serum in a humidified incubator at 37 C in 10% CO 2. The medium was removed, and the cells were irradiated with the indicated dose of UV (UV-C) in a Stratalinker UV cross-linker 2400 apparatus (Stratagene). Fresh medium was then added, and cells were returned to the incubator. Unless stated otherwise, cells were harvested, and cell extracts were prepared as described above 12 h after irradiation. Constructs and Expression and Purification of Recombinant Proteins GST-Cdc25B and GST-Wee1 were as described (5). The GST-Rb construct containing Rb residues was a gift from J. Wang (University of California at San Diego). Human p21 with a hexahistidine tag at the C terminus (p21-h6 in pet21d) was a gift from M. Howell (Imperial Cancer Research Fund, London). Expression of GSTand histidine-tagged proteins in bacteria and purification by GSHagarose and Ni 2 -nitrilotriacetic acid-agarose chromatography, respectively, were as described (60). Histone H1 Kinase and Rb Kinase Assays Immunoprecipitates equilibrated with kinase buffer (80 mm sodium -glycerophosphate, ph 7.4, 20 mm EGTA, 15 mm Mg(OAc) 2,and1mM dithiothreitol) were mixed with 10 l of kinase buffer containing 50 M ATP, 1.25 Ci of [ - 32 P]ATP, and 1 g of histone H1 or 2 g of GST-Rb. No unlabeled ATP was added in assays of histone H1 kinase activity associated with p21. The samples were incubated at 23 C for 30 min, and the reactions were terminated by addition of 30 l of SDS sample buffer. The samples were subjected to SDS-PAGE, and phosphorylation was detected and quantitated with a PhosphorImager. Antibodies Rabbit anti-cdk2 and rabbit anti-cyclin B1 antisera (61), the A17 anti-cdc2 monoclonal antibody (62), and the E72 anticyclin A monoclonal antibody (60) were as described previously. Anti- PCNA monoclonal antibody (PC10) (63) was a gift from J. Gannon and T. Hunt (Imperial Cancer Research Fund, South Mimms, Great Britain). These antibodies recognized the respective antigens from human, mouse, and rat origins. Anti-human cyclin E monoclonal antibodies (HE12 for immunoblotting and HE172 for immunoprecipitation) were generous gifts from E. Lees (DNAX, Palo Alto, CA), and anti-mouse cyclin E polyclonal antiserum was a generous gift from J. Roberts (Fred Hutchinson Cancer Research Center, Seattle). Anti-cyclin D1 polyclonal antiserum was raised in rabbits against the C-terminal peptide (CTPTDVRDVDI) and recognized human, mouse, and rat cyclin D1. Anti-human Cdk4 antiserum was raised in rabbits against the C-terminal peptide (CHSYLHKDEGNPE); on immunoblots, but not by immunoprecipitation, this antibody also recognized a cross-reactive band migrating just below Cdk4. Anti-phosphotyrosine antibodies were a mixture of monoclonal antibodies PY20 and PY72. Anti-mouse and anti-rat p53 monoclonal antibody (Ab122) and anti-human p53 monoclonal antibody (DO-1; Santa Cruz Biotechnology, sc-126) were gifts from A. Tsou and W. Eckhart (The Salk Institute). Anti-human p16 antibody was raised in rabbits against a C-terminal peptide of human p16 (CNHARIDAAEGPSDI). Rabbit anti-mouse p27 antisera were raised either against the C-terminal peptide of mouse p27 or against full-length GST-p27 (14). Anti-human p27 antibodies were a monoclonal antibody for immunoblotting (Pharmingen, 13231A) (which also recognized mouse and rat p27) and a 1:1 mixture of two polyclonal antibodies for immunoprecipitation (Santa Cruz Biotechnology, sc-527 and sc-528) (a gift from G. Orend, La Jolla Cancer Research Foundation, La Jolla, CA). Anti-p21 antibodies were as follows: AbI, a polyclonal antibody raised against human GST-p21 (Pharmingen, 15431); AbII, a monoclonal antibody raised against full-length human p21 (Santa Cruz Biotechnology, sc-817) (a gift from A. Tsou and W. Eckhart); or AbIII, a purified polyclonal antibody raised against the C- terminal peptide of p21 (Santa Cruz Biotechnology, sc-397). The AbI and AbII anti-p21 antibodies only recognized human p21, whereas the AbIII anti-p21 antibody recognized p21 from human, mouse, and rat. Immunological Procedures Immunoprecipitation was performed as described (60); immunoblotting was also performed as described (64), except that a SuperSignal CL-HRP substrate system (Pierce) was used to detect the horseradish peroxidase-conjugated secondary antibodies. Immunodepletions were performed as described (60), except that immunodepletion experiments were performed with two rounds instead of three rounds of immunoprecipitation.

3 Mechanism of UV-induced Cell Cycle Arrest FIG. 2.Time course of UV response. Normal human AG1523 fibroblasts were untreated (lane 1) or irradiated with a UV dose of 30 J/m 2 (lanes 2 5) or60j/m 2 (lanes 6 9). Cells were harvested at the indicated times after irradiation, and cell extracts were prepared. Equal amounts of protein (10 g) were resolved by SDS-PAGE and immunoblotted with antibodies against p53, p21 (with AbI), cyclin A, or cyclin D1 as indicated. FIG. 1. UV dose response of p53, cyclins, CDKs, and CDK inhibitors in human fibroblasts, mouse Swiss 3T3 fibroblasts, and rat NRK cells. Normal human AG1523 fibroblasts (A), mouse Swiss 3T3 fibroblasts (B), and rat NRK cells (C) were treated with the indicated dose of UV irradiation. After culture for 12 h, cell extracts were prepared as described under Materials and Methods. Equal amounts of protein (10 g) were resolved by SDS-PAGE and immunoblotted with antibodies against p53, p21 (with AbII), p27, p16, cyclins, or CDKs as indicated. For the Cdk4 immunoblots, the arrow indicates Cdk4; the faster migrating band cross-reacted with the antibody on immunoblots, but not by immunoprecipitation. RESULTS Accumulation of p21 Correlates with p53 at a Low Dose, but Not a High Dose, of UV Irradiation To determine whether there is a good correlation between the levels of p53 and p21 after UV-induced DNA damage, we irradiated cells with different doses of UV (UV-C) and examined the amounts of p53, p21, and other cell cycle regulatory proteins in the cells by immunoblotting. UV-C was used instead of solar UV-A and UV-B because we primarily intended to use UV as a source of DNAdamaging agent instead of generating active oxygen intermediates (65). Fig. 1A shows that an increase in p53 level was detected after normal human fibroblasts were irradiated with a UV dose of 30 J/m 2 and remained elevated up to the highest dose used (60 J/m 2 ). In contrast to p53, however, although p21 was induced at low doses of irradiation as expected (increased at 30 J/m 2 and peaked at 40 J/m 2 ), p21 was present only at the basal level at UV doses of 50 J/m 2 or higher. Similar results were observed with NRK (Fig. 1B) and Swiss 3T3 (Fig. 1C) cells. Staining with anti-p21 antibodies followed by immunofluorescence microscopy indicated that 90% of the cells expressed an elevated level of p21 after irradiation with a UV dose of 30 J/m 2 (data not shown). The failure to see p21 induction at higher doses of UV is striking since p53 was present at a very high level in these cells. Previously, we found that when Swiss 3T3 cells are irradiated with UV, both p53 and p21 are induced, whereas the level of the p21-related inhibitor p27 is reduced (60). Under these circumstances (which is equivalent to 40 J/m 2 for AG1523 cells shown in Fig. 1A), cyclin levels also decrease; most notably, cyclin D1 is destroyed much faster than the rest, which causes the p27 that is normally associated with cyclin D1 to become transiently redistributed to cyclin A/E-Cdk2. This redistribution of p27 may contribute to the rapid inactivation of Cdk2 following DNA damage. Similar results were observed with normal human fibroblasts (Fig. 1A); cyclins (A, B1, D1, and E) and p27 were significantly reduced by 12 h after irradiation at 40 J/m 2, when the p21 level was at a maximum. We observed that at lower UV doses (20 30 J/m 2 ), the levels of cyclins B1 and D1 were unchanged, but the levels of cyclins A and E and p27 were slightly elevated. CDK levels (Cdc2, Cdk2, and Cdk4) did not change significantly at UV doses up to 60 J/m 2, although at higher UV doses (40 J/m 2 or higher), the phosphorylation-modified forms of Cdc2 (upper band, representing Thr- 14/Tyr-15-phosphorylated forms) and Cdk2 (lower band, representing the Thr-160-phosphorylated form) were significantly reduced, as expected based on the destruction of their cyclin partners. The level of another CDK inhibitor, p16, did not vary significantly after UV irradiation. We have not examined the levels of other p16 family inhibitors. In this paper, we focus on the role of p21 in cells arrested at two different levels of DNA damage: at a relatively low dose of UV (30 J/m 2 ), when p21 was induced and the level of cyclins and CDKs remained high (Fig. 1), whereas their activities were inhibited (see below); and at a relatively high dose of UV (60 J/m 2 ), when the level of p21 did not increase. Fig. 2 shows the time course of accumulation or disappearance of p53, p21, and cyclins A and D1 after the cells were irradiated with a low dose (30 J/m 2 ) or a high dose (60 J/m 2 )of UV. Interestingly, a smaller band around 18 kda was detected by the anti-p21 polyclonal antiserum used here, but was not detected by the anti-p21 monoclonal antibody (compare with Fig. 1) or an anti-c-terminal peptide antibody (data not shown). This smaller polypeptide (presumably a degradation product of p21) was more prominent and accumulated with time after treatment with the high dose of UV in comparison with the low dose. Similar results were obtained with Swiss 3T3 cells using another polyclonal antibody (data not shown). Hence, an intriguing possibility is that at the high dose of UV, p21 was still being synthesized, but was rapidly converted into this smaller product. p21 Has a Half-life of 30 min To determine whether p21 is specifically degraded after irradiation (especially at a high

4 13286 Mechanism of UV-induced Cell Cycle Arrest FIG. 3.Turnover of p21 in presence or absence of UV irradiation. Growing normal human AG1523 fibroblasts were metabolically labeled with [ 35 S]methionine. Cells were then UV-irradiated at 30 J/m 2 ( ) or60j/m 2 (Ç) or not irradiated (E). A pulse-chase experiment was then performed as described under Materials and Methods. Cell extracts were prepared at the indicated time points and immunoprecipitated with anti-p21 antibodies. The samples were analyzed by SDS- PAGE, and the amount of 35 S-labeled p21 was quantitated with a PhosphorImager. UV dose), we measured the turnover rate of p21 in untreated and irradiated cells. Cells were first metabolically labeled with [ 35 S]methionine and then UV-irradiated, followed by culture in complete medium without [ 35 S]methionine. Cell-free extracts were prepared at different times after irradiation; p21 was then immunoprecipitated; and the relative amount of 35 S radioactivity in p21 was quantitated (Fig. 3). The turnover rate of p21 was relatively high, with a half-life of 30 min. Moreover, there was no significant change in the half-life of p21 after UV irradiation at either 30 or 60 J/m 2. Since there was no increase in the stability of p21 after irradiation at 30 J/m 2, we conclude that the greater accumulation of p21 at 30 J/m 2 was mainly due to increased synthesis. With such a short half-life, a slight acceleration in p21 turnover at a UV dose of 60 J/m 2 would be hard to detect, but a small change in half-life is unlikely to account for the difference in p21 levels between 30 and 60 J/m 2. Northern blot analysis showed that transcription of p21 increased after UV irradiation at 30 J/m 2, but not at 60 J/m 2 (data not shown), indicating that the p53-dependent transcription of p21 is impaired after a high dose of irradiation (see Discussion ). Inhibition of Cyclin-CDK- and p21-associated Kinase Activity after UV Irradiation AG1523, HSF8, Swiss 3T3, and NRK cells stopped proliferating when irradiated with UV. FACS analysis showed that the S phase cells disappeared after UV irradiation at either 30 or 60 J/m 2, and cells were arrested in both G 1 and G 2 phases of the cell cycle (Fig. 4A). When assayed using histone H1 or GST-Rb substrates, the kinase activities associated with the mitotic cyclin (cyclin B1), S phase cyclin (cyclin A), and G 1 phase cyclins (cyclins D1 and E) were all diminished after UV irradiation at 30 J/m 2 even though the levels of cyclins and CDKs were unchanged (Fig. 4B). To show that cyclins were not dissociated from their CDK partners, cyclins were immunoprecipitated, and the immunoprecipitates were then immunoblotted for the respective CDK partners. Fig. 4C shows that the cyclin B1-Cdc2, cyclin A-Cdk2, and cyclin E-Cdk2 complexes were all present after irradiation at 30 J/m 2. Cyclin D1-Cdk4 appeared to be less stable compared with other cyclin-cdk complexes. At 60 J/m 2, these cyclin-cdk complexes were not present as expected since the cyclins were destroyed (Fig. 1). Therefore, the arrest in cell cycle progression after UV irradiation is likely to be due to the inhibition of the kinase FIG. 4. Inhibition of kinase activity of cyclin-cdk after UV irradiation. A, FACS analysis. Growing normal human AG1523 fibroblasts were either untreated or irradiated at 30 or 60 J/m 2. After 12 h, the cell cycle distribution of the cells was analyzed by FACS analysis. B, inactivation of kinase activity association with cyclins after UV irradiation. AG1523 cells were either untreated (odd-numbered lanes) or irradiated at 30 J/m 2 (even-numbered lanes) as described above, and cell extracts were prepared after 12 h. The cell extracts (200 g) were immunoprecipitated with antibodies against cyclin A (lanes 1 and 2), cyclin B1 (lanes 3 and 4), cyclin D1 (lanes 5 and 6), or cyclin E (lanes 7 and 8). The histone H1 kinase activity (lanes 1 4, 7, and 8) or the Rb kinase activity (lanes 5 and 6) was assayed and quantitated with a PhosphorImager. C, cyclin-cdk association after UV irradiation. AG1523 cells were irradiated as described above, and cell extracts were prepared after 12 h. The cell extracts (200 g) were immunoprecipitated with antibodies against cyclin A (top panel), cyclin B1 (second panel), cyclin D1 (third panel), or cyclin E (bottom panel). The immunoprecipitates (IP) were immunoblotted with antibodies against the respective partners of the cyclins: Cdc2 (for cyclin B1), Cdk2 (for cyclins A and E), or Cdk4 (for cyclin D1). activity of various cyclin-cdk complexes at the low dose of UV and the destruction of cyclins at the high dose of UV. To determine whether p21 is responsible for the inhibition of cyclin-cdk complexes at the low dose of UV, we next investigated whether there was any change in the level of cyclins associated with p21 after irradiation. To show that the anti-p21 antibodies we used could immunoprecipitate p21 and associated proteins, cell extracts derived from 35 S-labeled extracts were immunoprecipitated with anti-p21 antibodies (Fig. 5A). Note that no 27-kDa band potentially representing the p21- related p27 was detected; moreover, the anti-p21 antibodies did not recognize either recombinant p27 or p57 on immunoblots (data not shown), indicating that the antibodies are relatively specific for p21. Similarly, we did not observe any p21 coimmunoprecipitated with anti-p27 antibodies (Fig. 5B). In this connection, we note that we have never detected p21 and p27 in the same complex, a theoretical possibility if two molecules of p21 or p27 are involved in the inhibition of one molecule of cyclin-cdk complex. p21 immunoprecipitates from extracts of untreated cells or cells treated with a UV dose of 30 J/m 2 were immunoblotted with antibodies against cyclins and PCNA. We found that there

5 Mechanism of UV-induced Cell Cycle Arrest FIG. 5. Association of p21 with cyclin-cdk and PCNA. A, immunoprecipitation of p21 from 35 S-labeled cells. Normal human AG1523 fibroblasts were metabolically labeled with [ 35 S]methionine, and cell extracts were prepared. The cell extracts (200 g) were immunoprecipitated with normal rabbit serum (NRS) (lane 1) or anti-p21 antibodies (AbI) (lane 2), and the samples were analyzed by SDS-PAGE (17.5%) followed by phosphoimaging. The positions of molecular mass markers (in kilodaltons) are shown on the left. The position of p21 is indicated; the other proteins that associated with p21 were (indicated by arrows, in ascending order) CDKs, cyclin D1, PCNA, and cyclin A according to Ref. 33. B, immunoprecipitation of p27 from 35 S-labeled cells. Swiss 3T3 cells were metabolically labeled with [ 35 S]methionine, and cell extracts were prepared. The cell extracts (200 g) were immunoprecipitated with anti-p27 C-terminal peptide antibodies in the absence (lane 2) or presence (lane 3) of competing peptide. In lane 1, an anti-p27 immunoprecipitate was boiled in 1% SDS, and the supernatant was diluted 10-fold, followed by re-immunoprecipitation with antip27 antibody. The samples were analyzed by SDS-PAGE followed by phosphoimaging. The positions of molecular mass markers (in kilodaltons) and the position of p27 are indicated. C, irradiation-induced binding of cyclins and PCNA to p21. Normal human AG1523 fibroblasts were either untreated (lanes 1 and 2) or irradiated at 30 J/m 2 (lanes 3 and 4), and cell extracts were prepared. The cell extracts (400 g) were immunoprecipitated with normal rabbit serum (lanes 1 and 3) or antip21 antibodies (AbI) (lanes 2 and 4), and the immunoprecipitates were immunoblotted with antibodies against cyclin A, B1, D1, or E or PCNA as indicated. D, inactivation of p21-associated kinase activity after UV irradiation. AG1523 cells were either untreated (lane 1) or irradiated at 30 J/m 2 (lane 2) as described above, and cell extracts were prepared after 12 h. The cell extracts (200 g) were immunoprecipitated with anti-p21 antibodies, and the histone H1 kinase activity was assayed and quantitated with a PhosphorImager. Gel analyses of the phosphorylation reactions are shown in the lower panel, and their quantitations are shown in the upper panel. was a small amount of cyclins A, B1, and D1 and PCNA associated with p21 in untreated cells, but the relative amount of these proteins that associated with p21 increased significantly after irradiation (Fig. 5C). A similar increase in association with p21 was also seen with CDKs (data not shown). Intriguingly, the level of cyclin E associated with p21 was similar with FIG. 6. Immunodepletion of p21. Normal human AG1523 fibroblasts were either untreated (lanes 1 and 2) or irradiated at 30 J/m 2 (lanes 3 and 4), and cell extracts were prepared after 12 h. The cell extracts were immunodepleted with normal rabbit serum (NRS) (lanes 1 and 3) or anti-p21 antiserum (AbI) (lanes 2 and 4) as described under Materials and Methods. The supernatant after immunodepletion (10 g) was then resolved by SDS-PAGE and immunoblotted with antibodies against p21 (AbI), p27, Cdc2, Cdk2, Cdk4, cyclin A, cyclin B1, cyclin D1, cyclin E, and PCNA as indicated. or without irradiation. Hence, the increased association between cyclin-cdk and p21 may explain the inhibition of the kinase activity of cyclin-cdk complexes. Furthermore, the kinase activity associated with p21 itself also diminished after irradiation (Fig. 5D), suggesting that the cyclin-cdk complexes that associated with p21 after irradiation were inactivated. p21 Binds to All of the Cdk2, but Only to a Small Population of Cdc2 and Cdk4, after UV Irradiation Although there is a correlation between increased binding of p21 to cyclin-cdk and the inhibition of cyclin-cdk kinase activity, this does not prove that p21 is responsible for the inactivation of cyclin-cdk complexes. A prerequisite condition for p21 to be responsible for cyclin-cdk inhibition is that p21 has to be associated with all the cyclin-cdk complexes. Although the fact that the cyclin- CDK complexes associated with p21 were inactive after UV irradiation (see above) suggests that p21 was present in excess, we set out to establish whether this was the case. For this purpose, we determined what proportion of the different cyclin and CDK populations was bound to p21 in cell extracts derived from untreated cells or cells that received a UV dose of 30 J/m 2 by immunodepletion (Fig. 6). Cell extracts were immunodepleted for two cycles with either anti-p21 antiserum or normal rabbit serum as a control. What remained in the supernatant after immunodepletion was examined by immunoblotting with specific antibodies. Fig. 6 shows that p21 was induced by UV irradiation as expected (lanes 1 and 3), and it was immunodepleted with anti-p21 antiserum (lanes 2 and 4). No p27 was depleted by this procedure. We found that cyclin A was partly depleted with anti-p21 antiserum without UV treatment, but, significantly, all of the cyclin A was depleted with anti-p21 antiserum after UV treatment. In contrast, cy-

6 13288 Mechanism of UV-induced Cell Cycle Arrest FIG. 7.Immunodepletion of CDKs. Normal human AG1523 fibroblasts were either untreated (lanes 1 3) or irradiated at 30 J/m 2 (lanes 4 6), and cell extracts were prepared as described under Materials and Methods. The cell extracts were immunodepleted with normal rabbit serum (NRS) (lanes 1 and 4), anti-cdk2 antibodies (lanes 2 and 5), or anti-cdk4 antibodies (lanes 3 and 6). The supernatant after immunodepletion (10 g) was then resolved by SDS-PAGE and immunoblotted with antibodies against Cdk2, Cdk4, and p21 (AbIII) as indicated. The arrow indicates the position of Cdk4 as described for Fig. 1. clin B1 was not detectably depleted with anti-p21 antiserum either with or without UV treatment. Cyclin D1 was also only slightly depleted with anti-p21 antiserum. Interestingly, cyclin E was depleted with anti-p21 antiserum both with and without UV treatment, consistent with the earlier results that the amount of cyclin E co-immunoprecipitated with anti-p21 antiserum was the same under both circumstances (Fig. 5B). No depletion of Cdc2 or Cdk4 was observed, consistent with their cyclin partners not being depleted. For Cdk2, a small proportion of the protein was depleted with anti-p21 antiserum in the absence of UV treatment, but a higher proportion of Cdk2 was depleted with anti-p21 antiserum after irradiation, consistent with the observed depletion of its partners, cyclins A and E. Significantly, although not all of the Cdk2 was depleted from lysates of either control or UV-treated cells, all of the faster migrating form of Cdk2 (the Thr-160-phosphorylated and active form of Cdk2) was depleted from UV-treated, but not control, cell lysates. This is explicable by the fact that p21 binds to cyclin-cdk2 complexes with much higher affinity than to free Cdk2, and only cyclin-associated Cdk2 is phosphorylated on Thr-160; thus, free unphosphorylated Cdk2 (which is in excess over cyclins A and E in the cell) was not depleted with p21. Depletion of p21 did not decrease the total level of PCNA significantly, suggesting that p21 bound to only a small proportion of the total PCNA after UV-induced DNA damage. Essentially the same results as above was obtained with nondiploid mouse Swiss 3T3 fibroblasts (data not shown). To examine what proportion of the total p21 was associated with Cdk2 and Cdk4, we did the converse experiment of immunodepleting Cdk2 or Cdk4. Fig. 7 shows that in untreated cells, a small proportion of the p21 was associated with Cdk2, and most of the p21 was associated with Cdk4. After irradiation, however, immunodepletion of either Cdk2 or Cdk4 only partially depleted p21. These results suggest that p21 is mainly associated with cyclin D-Cdk4 in normal growing cells, which is similar to what has been observed for p27 (60). Furthermore, the increased level of p21 after UV-induced DNA is likely to be sufficient to saturate the available cyclin-associated Cdk2 and Cdk4. However, since not all of the cyclin D-Cdk4 complex is associated with p21 after irradiation, the rest of the cyclin D-Cdk4 complex must somehow be unable to bind p21, perhaps because it is already occupied by another inhibitor such as p27. In conclusion, these results indicate that p21 is associated with only a portion of Cdk2 in growing cells (all of the cyclin E-Cdk2 complex and some of the cyclin A-Cdk2 complex), but p21 binds to all of the active Cdk2 after UV irradiation. Although there is an increase in the binding of cyclin B1-Cdc2, cyclin D1-Cdk4, and PCNA to p21 following irradiation, only a minor portion of these proteins is associated with p21 either FIG. 8. Inhibition of CDKs by recombinant p21-h6. A, cell extracts from untreated AG1523 cells (10 g) were mixed with different amounts of purified p21-h6 protein (5 l of 27-fold (lane 1), 81-fold (lane 2), and 243-fold (lane 3) diluted) and compared with extracts from untreated AG1523 cells (lane 4) and AG1523 cells UV-irradiated at 30 J/m 2 (lane 5) by immunoblotting with anti-p21 antibodies (AbII). B, cell extracts from normal growing AG1523 cells (100 g) were mixed with purified p21-h6 (5 l of 8-fold diluted, i.e. equivalent to the final concentration of p21-h6 in A, lane 2) and incubated at 25 C for 30 min. The extracts were then immunoprecipitated with antibodies against Cdc2 (lanes 1 and 2), Cdk2 (lanes 3 and 4), or Cdk4 (lanes 5 and 6). The kinase activity of the immunoprecipitates was measured using histone H1 (lanes 1 4) or GST-Rb (lanes 5 and 6) as substrate. Gel analyses of the phosphorylation reactions are shown in the lower panels, and the quantitations are shown in the upper panels. before or after irradiation. p21 Is Sufficient to Inhibit all of the Cdk2, but Not Cdc2 and Cdk4, after UV Irradiation We next asked whether the amount of p21 induced after UV-induced damage was sufficient to inhibit the kinase activities of CDKs. The amount of p21 in extracts derived from irradiated cells was compared with the amount of p21 in extracts from normal cells mixed with serial dilutions of purified recombinant p21-h6 by immunoblotting (Fig. 8A). We then used the amount of recombinant p21 that provided the cell extracts with a total level of p21 that was similar to that after irradiation (Fig. 8A, lane 2) and determined whether various cyclin-cdk complexes were inhibited under these circumstances. Fig. 8B shows that adding recombinant p21-h6 to the level observed after irradiation completely abolished Cdk2-associated kinase activity, but only partially inhibited Cdk4-associated kinase activity and had no effect on Cdc2-associated kinase activity. This supports the earlier conclusion that p21 is sufficient to inhibit all Cdk2- associated kinase activity, but only partially affects Cdk4 and has no effect on Cdc2. p21 and p27 Cooperate in the Inhibition of Cdk4 If p21 is only sufficient to inhibit Cdk2 after UV-induced DNA damage, how are other CDKs like Cdk4 and Cdc2 inhibited? One possibility is that more than one CDK inhibitor is involved. To see whether the p21-related p27 also plays a role in the inhibition of cyclin D1-Cdk4 in conjunction with p21, either p27 alone or p21 and p27 together were immunodepleted from cell extracts derived from normal cells or irradiated cells (Fig. 9). As seen in Fig. 5, no significant cross-reaction between the antibodies against p21 and p27 was detected. Immunodepleting p21 and p27 together depleted a considerable portion of the cyclin D1 before irradiation and nearly all of the cyclin D1 after irradiation, and in both cases, the combined depletion removed more cyclin D1 than either p21 (Fig. 6) or p27 alone. Therefore, after irradiation, a significant portion of the cyclin D1-Cdk4 complexes can be accounted for in association with either p21 or p27, which may be responsible for the inhibition of their kinase

7 Mechanism of UV-induced Cell Cycle Arrest FIG. 9.Immunodepletion of p21 and p27. Normal human AG1523 fibroblasts were either untreated (lanes 1, 2, 5, and 6) or irradiated at 30 J/m 2 (lanes 3, 4, 7, and 8), and cell extracts were prepared. The cell extracts were immunodepleted with normal rabbit serum (NRS) (lanes 1, 3, 5, and 7), anti-p27 antibodies (lanes 2 and 4), or a mixture of anti-p21 and anti-p27 antibodies (lanes 6 and 8). The supernatant after immunodepletion (10 g) was then resolved by SDS-PAGE and immunoblotted with antibodies against p21 (AbIII), p27, and cyclin D1 as indicated. activity. Cdc2 Is Inhibited by Thr-14/Tyr-15 Phosphorylation after UV Irradiation Since no significant amount of cyclin B-Cdc2 binding to CDK inhibitors was detected after UV irradiation, we next investigated whether the inhibition of Cdc2 kinase activity might be due to negative regulatory phosphorylation of Thr-14 and Tyr-15. We found that there was a slight increase in tyrosine phosphorylation in the Cdc2 immunoprecipitates after cells were irradiated with a UV dose of 30 J/m 2 (data not shown). A more appreciable increase in tyrosine phosphorylation was seen in Cdc2 that associated with cyclin B1 (Fig. 10A), suggesting that Tyr-15 phosphorylation may be a mechanism that inhibits the activity of cyclin B1-Cdc2 after irradiation. Furthermore, when purified GST-Cdc25B fusion protein was incubated with the Cdc2 immunoprecipitates derived from cells receiving a UV dose of 30 J/m 2 or no irradiation (Fig. 10B), the histone H1 kinase activities of both active Cdc2 (from untreated cells) and inactivated Cdc2 (from irradiated cells) were increased, although the -fold increase was much greater with Cdc2 from irradiated cells. Moreover, the maximal level of Cdc2 activation achieved with GST-Cdc25B was the same with Cdc2 immunoprecipitates from control and irradiated cells. These results suggest that the inhibition of Cdc2 after irradiation may be due to Tyr-15 and/or Thr-14 phosphorylation (Cdc25 can dephosphorylate both residues). The same result was obtained with GST-Cdc25C (data not shown). Under the same conditions, the histone H1 kinase activity of the control Cdc2 immunoprecipitate was inhibited by phosphorylation with GST-Wee1, a Tyr-15 kinase, whereas there was very little further inhibitory effect of GST-Wee1 on Cdc2 activity from UV-irradiated cells, consistent with the idea that Tyr-15 was already nearly fully phosphorylated. In contrast to Cdc2, the Cdk2 immunoprecipitates could be activated by GST-Cdc25B only before, and not after, UV irradiation (Fig. 10B). This is consistent with the fact that all cyclin-cdk2 complexes were inhibited by p21 after UV irradiation. No effect on the kinase activity of Cdk4 was detected in a similar experiment (data not shown); it is possible that Cdk4 is not regulated by Tyr-17 phosphorylation or is not a substrate for Cdc25B and Wee1. DISCUSSION Information on how different cyclin-cdk complexes are inhibited following DNA damage in normal cells is important for our understanding of how cancer cells escape this cell cycle checkpoint and may have implications for the design of cancer therapies. Here we show that the kinase activities associated with Cdc2, Cdk4, and Cdk2, which represent engines for G 2 -M, G 1, and G 1 -S progression, respectively, are all inhibited after FIG. 10. Thr-14/Tyr-15 phosphorylation of Cdc2. A, normal human AG1523 fibroblasts were either untreated (lane 1) or irradiated at 30 J/m 2 (lane 2), and cell extracts were prepared. The cell extracts (400 g) were immunoprecipitated with anti-cyclin B1 antibodies, and the immunoprecipitates (IP) were immunoblotted with monoclonal antibodies against phosphotyrosine (P-Tyr). B, normal human AG1523 fibroblasts were either untreated (lanes 1, 3, and 5) or irradiated at 30 J/m 2 (lanes 2, 4, and 6), and cell extracts were prepared. The cell extracts (200 g) were immunoprecipitated with either anti-cdc2 antibody (upper panel) or anti-cdk2 antibody (lower panel), and the immunoprecipitates were incubated with 1 g of purified GST (lanes 1 and 2), GST- Cdc25B (lanes 3 and 4), or GST-Wee1 (lanes 5 and 6) in the presence of 15 mm Mg 2 and1mmatp at 25 C for 30 min. The immunoprecipitates were then washed, and the histone H1 kinase was assayed. Histone H1 phosphorylation was quantitated with a PhosphorImager and is expressed as the percentage of that of untreated cells. UV irradiation of fibroblasts. It is likely that the cell uses different mechanisms for the inhibition of each class of CDKs. The p53-dependent induction of p21 is likely to be sufficient for the inhibition of Cdk2-associated kinase activity, whereas an increase in Thr-14/Tyr-15 phosphorylation appears to be responsible for the inhibition of Cdc2-associated kinase activity. Contrary to current belief, p21 alone is unlikely to be sufficient for the inhibition of cyclin D1-Cdk4; at least in the cells studied here, a combination of p21 and p27 appears to be responsible for the inhibition of cyclin D1-Cdk4 activity. Thus, cyclin E- Cdk2 and cyclin A-Cdk2 are more important targets for p21 than cyclin D1-Cdk4. p21 is able to inhibit a relatively wide variety of cyclin-cdk complexes, although the affinity for Cdk2 and Cdk4 is much higher than for Cdc2 (32), which may explain the relative minor role played by p21 in the inhibition of Cdc2. It is surprising that all of the cyclin E-Cdk2 complex is already complexed with p21 even before DNA damage in growing fibroblasts. Nevertheless, cyclin E-Cdk2 is active in growing cells, which is consistent with the suggestion that the first molecule of p21 that associates with a cyclin-cdk complex does not inhibit its activity and that the binding of a second p21 molecule is required to inhibit its kinase activity (32, 33). The fact that the kinase activity associated with p21 is diminished after UV irradiation, coupled with the fact that a significant fraction of p21 is left behind after depletion of UV-treated cell lysates with anti-cdk2 and anti-cdk4 antibodies, implies that the

8 13290 Mechanism of UV-induced Cell Cycle Arrest p21-associated cyclin-cdk complexes are saturated with p21 after DNA damage (i.e. with more than one p21 molecule/ cyclin-cdk complex). Since it is difficult to directly compare the kinase activity associated with CDKs and that of p21, we do not know the relative activity of cyclin-cdk complexes that are free or associated with p21 before and after DNA damage. Nevertheless, our in vitro titration experiments with recombinant p21 show that there is enough p21 present in UV-irradiated cells to inhibit all of the cyclin-cdk2 activity in an untreated cell. Here we used a dose of UV (30 J/m 2 ) that caused CDK inhibition and cell cycle arrest, but did not diminish the level of cyclins. We have shown before that at a slightly higher dose of UV (40 50 J/m 2 ), where p21 is induced and cyclins are destroyed, the inhibitor p27 may provide a rapid Cdk2 inhibitory response before p21 is synthesized (60). Cyclin D1-Cdk4 acts as a reservoir for p27 in normal growing cells, and after UVinduced damage, cyclin D1 is destroyed rapidly; hence, the net effect is that the p27 that usually binds to cyclin D1-Cdk4 is redistributed rapidly to cyclin A/E-Cdk2. At 30 J/m 2, cyclins are not destroyed, and there is no redistribution of p27 from cyclin D1-Cdk4 to cyclin A/E-Cdk2. Here we show that, like p27, most of the basal p21 is also associated with cyclin D- Cdk4, so it is likely that p21 is also redistributed rapidly from Cdk4 to Cdk2 at an intermediate level of DNA damage. We consistently observed a small increase in the levels of cyclins E and A after irradiation at 30 J/m 2, which intriguingly are the two cyclins that are inhibited by p21. One possibility is that increased binding of p21 to cyclin-cdk complexes may inhibit the degradation of the cyclins. The level of cyclin A also increases after HeLa cells are irradiated with ionizing radiation (66). The exact contribution of p21, p27, or other inhibitors and cyclin degradation to the inhibition of Cdk2 depends on the extent of DNA damage. There are also subtle variations depending on cell type, but some principles emerge. At a low level of DNA damage, the main long-term contributor to Cdk2 inactivation is p21 induction; at a high level of DNA damage, cyclin degradation becomes the main factor in the inactivation of Cdk2; and at an intermediate level of DNA damage, both p21 induction and cyclin degradation may be involved, with the possibility of a more rapid p21/p27 redistribution response. We found that the increase in p21 level alone is probably insufficient to account for the inhibition of cyclin D1-Cdk4. Although p21 and p27 together bind to the majority of the cyclin D1-Cdk4 complexes, a significant portion of the cyclin D1-Cdk4 complexes are not associated with p21 or p27. It is possible that other CDK inhibitors, not investigated here, also contributed to the inhibition of cyclin D1-Cdk4 after DNA damage. In this connection, we found that there was a reduction in the level of cyclin D1-Cdk4 complexes after irradiation at 30 J/m 2 (Fig. 4C), and therefore, it is possible that the p16 family CDK inhibitors may be involved since they bind Cdk4 and block rebinding of cyclin D1. We did not detect an increase in p16 following UV irradiation, but p15, p18, or p19 may be upregulated. It has also been reported that phosphorylation of Cdk4 at Tyr-17 is required for the G 1 arrest of NRK cells following UV irradiation (67). Unlike Cdc2 and Cdk2, we were unable to activate cyclin D1-Cdk4 with GST-Cdc25B, but we have no direct evidence that phosphorylation of Tyr-17 does not play a role in inhibiting Cdk4 activity in UV-irradiated human fibroblasts, nor do we know whether Cdk4 Tyr-17 is a substrate for Cdc25B. On the contrary, it appears that Cdc2 is mainly inhibited by Tyr-15 phosphorylation after UV irradiation. We do not know whether Thr-14 phosphorylation is also increased after UV irradiation because Cdc25 can dephosphorylate both Thr-14 and Tyr-15; but it is clear that Tyr-15 phosphorylation is involved, as suggested by the increase in Cdc2 phosphotyrosine content. An increase in Cdc2 tyrosine phosphorylation has also been observed in the human HaCaT keratinocyte cell line (with mutated p53) after UV irradiation (68). The increase in Tyr-15 (and Thr-14) phosphorylation in Cdc2 could be due to an increase in the kinase activity of Wee1 and/or Myt1 (or other unidentified Wee1-related kinases) or could be due to a decrease in Cdc25 phosphatase activity. We are currently exploring the different possibilities. In Xenopus egg extracts, an increase in Cdc2 tyrosine phosphorylation in the presence of incompletely replicated DNA has been observed (69). However, when cyclin B-Cdc2 is inhibited in Xenopus egg extracts arrested at the G 2 -M checkpoint with aphidicolin, there is no change in the activities of the kinases and phosphatases that regulate Thr-14, Tyr-15, and Thr-161, and an inhibitor of Cdc2 has been proposed to be responsible for the decreased activity in this situation (70). At a high dose of UV irradiation (60 J/m 2 ), we found that the levels of p21 and cyclins were diminished, despite the fact that p53 was induced and was present at high levels. The cell cycle is probably arrested under these circumstances not by an increase in CDK inhibitors or tyrosine phosphorylation, but by the loss of cyclins. It is possible that the DNAs encoding cyclins and p21 received so much damage that they are no longer transcribed. In support of this, Northern blot analysis indicated that the amount of p21 mrna remained at basal levels after UV irradiation at 60 J/m 2. However, p53 levels rise under these conditions, and since the p53 gene should also have been affected, it is likely that an increase in p53 stability explains its accumulation. The increase in the 18-kDa p21 fragment observed after UV irradiation at 60 J/m 2 is intriguing. The 18- kda fragment does not contain the C-terminal region of p21 since an anti-c-terminal peptide antibody did not recognize the fragment. DNA fragmentation typical of apoptosis was not detected by FACS analysis or DNA ladder analysis, and the cells still excluded trypan blue at 12 h after cells were irradiated at 60 J/m 2. However, smeared DNA degradation was observed after treating cells with a high dose of UV, and most of the cells failed to recover after such a high dose of UV irradiation. At a low dose of UV, cultured fibroblasts arrest and then appear to recover and continue growing. Thus, the mechanisms of cell cycle arrest we have defined may have physiological relevance in allowing the cell to repair UV-induced DNA damage before progressing into S phase or mitosis. The rapid turnover rate of p21 described here may be important for the cell to re-enter the cycle. Indeed, the fact that p21 has such a short half-life means that there is no need to invoke a specific p21 destruction mechanism for the cell to escape from the cell cycle arrest. Consistent with this idea, we find that the levels of p21 begin to fall at 24 h after irradiation. This is in contrast to human diploid fibroblasts exposed to ionizing radiation, where the cells never recover from the arrest and maintain high levels of p21 for several days (53). This implies that fibroblasts respond to UV-induced DNA damage (thymine dimers) and -ray-induced damage (double-strand breaks) in different ways. The experiments reported here have mostly used human diploid fibroblasts, but we have obtained very similar results with p53 rodent fibroblasts cell lines. However, we do not know whether diploid epithelial cells will respond in the same way, which would have obvious relevance to the exposure of skin to sunlight. But recent studies have shown that p53 and p21 are induced in human skin epithelium following UV irradiation (71). Furthermore, UV-C was used here mainly as a