p16INK4A Participates in a G1 Arrest Checkpoint in Response to DNA Damage

ABSTRACT Members of the INK4 protein family specifically inhibit cyclin-dependent kinase 4 (cdk4) and cdk6-mediated phosphorylation of the retinoblastoma susceptibility gene product (Rb). p16 INK4A , a prototypic INK4 protein, has been identified as a tumor suppressor in many human cancers. Inactivation of p16 INK4A in tumors expressing wild-type Rb is thought to be required in order for many malignant cell types to enter S phase efficiently or to escape senescence. Here, we demonstrate another mechanism of tumor suppression by implicating p16 INK4A in a G1 arrest checkpoint in response to DNA damage. Calu-1 non-small cell lung cancer cells, which retain Rb and lack p53, do not arrest in G1 following DNA damage. However, engineered expression of p16 INK4A at levels compatible with cell proliferation restores a G1 arrest checkpoint in response to treatment with γ-irradiation, topoisomerase I and II inhibitors, and cisplatin. A similar checkpoint can be demonstrated inp53−/− fibroblasts that express p16 INK4A . DNA damage-induced G1arrest, which requires the expression of pocket proteins such as Rb, can be abrogated by overexpression of cdk4, kinase-inactive cdk4 variants capable of sequestering p16 INK4A , or a cdk4 variant incapable of binding p16 INK4A . After exposure to DNA-damaging agents, there was no change either in overall levels of p16 INK4A or in amounts of p16 INK4A found in complex with cdks 4 and 6. Nonetheless, p16 INK4A expression is required for the reduction in cdk4- and cdk6-mediated Rb kinase activity observed in response to DNA damage. During tumor progression, loss of p16 INK4A expression may be necessary for cells with wild-type Rb to bypass this G1 arrest checkpoint and attain a fully transformed phenotype.

p16 INK4A is a specific inhibitor of cyclin dependent kinase 4 (cdk4) and cdk6, which participate in the cyclin D-dependent phosphorylation of the retinoblastoma susceptibility gene product, Rb (31). Hyperphosphorylation of Rb inactivates its growth-suppressive properties, allowing cells to enter S phase. Several lines of evidence indicate that p16 INK4A is a tumor suppressor. First, its gene maps to 9p21, a chromosomal locus deranged in many human cancers (15). Second, p16 INK4A is commonly deleted, mutated, or hypermethylated and transcriptionally silenced in tumors that retain wild-type Rb, and ectopic expression of p16 INK4A in these cells at high levels results in G 1 arrest (17,19,22,30,33,35). Furthermore, p16 INK4A -deficient mice are susceptible to several types of malignancies (32), and germ line mutations of p16 INK4A in humans are associated with familial syndromes involving malignant melanoma and pancreatic cancer (8,14,16,40).
The precise mechanism by which p16 INK4A exerts its tumorsuppressive effects is less clear. One straightforward suggestion is that inactivation of p16 INK4A is required for malignant cells to enter S phase efficiently. However, many normal cells express p16 INK4A throughout G 1 and are able to proliferate, suggesting that other mechanisms of tumor suppression must be operating. An alternative mechanism involves the recently identified link between p16 INK4A expression and cellular senescence (1,10,28,32). As fibroblasts or epithelial cells age, p16 INK4A levels increase dramatically, and it has been proposed that loss of p16 INK4A expression is required for cells to escape senescence during their progression to malignancy.
Another possibility is that p16 INK4A plays a role in the maintenance of genome integrity (34). Frequently, following DNA damage normal cells arrest their proliferation at cell cycle checkpoints, the most prominent of which occur at the G 1 -S and G 2 -M boundaries. Arrest allows time for repair prior to continued cell cycle progression. One G 1 arrest checkpoint is controlled by p53 (5,18). In response to DNA damage, p53 levels increase by a posttranscriptional mechanism, resulting in the transcriptional activation of p21 WAF1 , a universal inhibitor of cyclin-dependent kinases, which can mediate G 1 arrest (6,11,42). Inactivation of p53 is the most common genetic event in human cancer, suggesting that loss of a DNA damageinduced G 1 checkpoint is an essential step in tumor progression. This allows damaged DNA to be replicated, which leads to the accumulation of additional mutations and the eventual emergence of a malignant clone.
DNA damage also induces alterations in cyclin D1-cdk4 activity. For example, UV irradiation can lead to decreases in cyclin D1 levels and to inhibitory phosphorylation of cdk4 (24,27,37). Such perturbations may contribute to G 1 arrest following DNA damage. In the present study, we have investigated whether p16 INK4A may also be involved in the response to DNA damage. We have used non-small cell lung cancer (NSCLC) cells, which lack p53 and do not arrest in G 1 following DNA damage. When these cells are engineered to express p16 INK4A at levels compatible with proliferation, the ability to arrest in G 1 in response to DNA damage is restored. Furthermore, we demonstrate that p53 Ϫ/Ϫ fibroblasts maintain a similar G 1 arrest checkpoint in response to DNA damage, which correlates with the level of p16 INK4A they express. Although neither overall p16 INK4A levels nor the amount complexed to cdk4 and cdk6 changes following DNA damage, the presence of p16 INK4A causes a decrease in cdk4-and cdk6-mediated Rb kinase activity and results in G 1 arrest, even in the absence of p53.

MATERIALS AND METHODS
Cell lines. Calcium phosphate precipitation (4) was used to transfect Bing packaging cells (provided by Warren Pear, Massachusetts Institute of Technology, Cambridge, Mass.) with pBPSTR1 (25) or pBPSTR1 into which a cDNA encoding full-length p16 INK4A had been cloned (31). Viral supernatants were used to infect Calu-1 cells, which were then selected in puromycin (0.5 g/ml). A mass population of cells infected with the pBPSTR1 virus was isolated, as were several individual clones arising from pBPSTR1-p16 INK4A infection. Cells were maintained in tetracycline HCl (2 g/ml) and deprived of tetracycline for 24 h to induce higher levels of p16 INK4A expression. These lines were subsequently transfected with expression vectors encoding cdk4, cdk4 variants, and human papillomavirus (HPV) E7, and mass populations were selected for resistance to G418. Calu-1 cells were obtained from the American Type Culture Collection (Rockville, Md.) and primary embryo fibroblasts from p53 Ϫ/Ϫ mice were obtained from Tyler Jacks (Massachusetts Institute of Technology). Early (passage 5)-and late-passage samples of these cells were provided by David Fisher (Dana-Farber Cancer Institute). Normal human bronchial epithelial cells were purchased from Clonetics Corp. (San Diego, Calif.) and maintained in the growth factor-supplemented medium recommended by the supplier.
Construction of cdk4 R24C -HA. A pBlueScript plasmid containing a cDNA encoding cdk4 R24C (41) was obtained from David Beach (Cold Spring Harbor Laboratories, Cold Spring Harbor, N.Y.). An NdeI-BamHI fragment from the 3Ј region of a hemagglutinin (HA)-tagged wild-type cdk4 (in pCMV-neo) (7) was used to replace the analogous region of pBlueScript-cdk4 R24C , and the resulting HA-tagged cdk4 R24C cDNA was cloned into pcDNA3.
DNA damage treatments. Prior to treatment, Calu-1 cells were plated at 10 5 cells/ml and fibroblasts from p53 Ϫ/Ϫ mice were plated at 5 ϫ 10 4 cells/ml. In both cases, this provided cultures that were 50 to 70% confluent at the time of treatment. Twenty-four hours after plating, Calu-1 cells were treated with the following concentrations of chemotherapeutic agents for 48 h: 0.035-g/ml adriamycin (ADR), 2.5 M etoposide, 5 M camptothecin, and 2.6 g/ml cisplatin. In some experiments, cells were cultured for an additional 36 h in the same concentration of drug along with nocodazole (0.4 g/ml). For ␥-irradiation of Calu-1 cells, 10 Gy of irradiation was delivered by a 137 Cs source at 118 cGy/min. This dose was adequate to induce a G 2 arrest in parental Calu-1 cells. Twenty-four hours after plating, fibroblasts from p53 Ϫ/Ϫ mice were treated for 24 h with ADR (0.03 g/ml), etoposide (2 M), and cisplatin (1 g/ml). In some experiments, cells were cultured in the same concentration of drug along with nocodazole (0.4 g/ml) for an additional 16 h. For ␥-irradiation, cells were treated with 20 Gy of irradiation. This dose was required in order to induce a G 2 arrest in late-passage p53 Ϫ/Ϫ embryo fibroblasts.
Fluorescence-activated cell sorting (FACS) analysis. Cells were collected by trypsinization, washed, and resuspended in 1 ml phosphate-buffered saline. An additional 1 ml of 80% ethanol was added, and cells were fixed overnight at 4°C. Fixed cells were centrifuged and resuspended in 0.5 ml of 500-g/ml RNase A and incubated for 45 min at 37°C. Cells were centrifuged and resuspended in 0.5 ml of 69 M propidium iodide in 38 mM sodium citrate and incubated at room temperature for a minimum of 30 min. Cells were then analyzed for DNA content by flow cytometry (Becton Dickinson, Hialeah, Fla.). Rb kinase assays. Cells were lysed for 1 h at 4°C in 50 mM HEPES (pH 7.2)-150 mM NaCl-1 mM EDTA-2.5 mM EGTA-1 mM dithiothreitol-0.1% Tween 20 supplemented with 10% glycerol-1 mM NaF-0.5 mM sodium orthovanadate-aprotinin (1 g/ml)-leupeptin (1 g/ml)-10 mM ␤-glycerophosphate-phenylmethylsulfonyl fluoride (100 g/ml). Lysates were clarified by centrifugation at 10,000 ϫ g for 10 min, and the supernatants were precleared with rabbit immunoglobulin G prior to immune precipitation using 300 ng each of rabbit antisera against cdk4 and cdk6 (described above). Immune precipitates were collected by using protein A-Sepharose beads equilibrated with lysis buffer containing 4% bovine serum albumin. Beads were washed four times with lysis buffer and then twice with kinase buffer containing 50 mM HEPES (pH 7.2), 10 mM MgCl 2 , 5 mM MnCl 2 , and 1 mM dithiothreitol. After the final wash, 25 l of kinase reaction mix was added, consisting of kinase buffer with 20 M ATP, 10 Ci of [␥-32 P]ATP, and 1 g of GST-Rb (amino acids 792 to 928). Samples were incubated at 37°C for 30 min with occasional mixing, boiled in SDS-PAGE sample buffer, and fractionated by electrophoresis through 10% polyacrylamide gels. Proteins were electrophoretically transferred to nitrocellulose, and phosphorylated GST-Rb was visualized by autoradiography. The nitrocellulose filter was then subjected to Western blotting using cdk4-and cdk6-specific antibodies, to ensure that equivalent amounts had been immune precipitated for each kinase assay.
To prepare the GST-Rb substrate, a culture of Escherichia coli transformed with pGEX-Rb (792-928) was induced with IPTG (isopropyl-␤-D-thiogalactopyranoside) and lysed as previously described (21). Fusion protein was captured on glutathione-Sepharose 4B and released by incubation with reduced glutathione. The concentration of the soluble fusion protein was estimated by Coomassie blue staining of electrophoresed protein in comparison to protein standards of known concentration.

RESULTS
Engineered expression of p16 INK4A in Calu-1 cells. Calu-1 cells are NSCLC cells which have wild-type Rb, deleted p53, and transcriptionally silent and hypermethylated p16 INK4A loci (3,33). Ectopic expression of p16 INK4A at high levels in these and other cell types that retain wild-type Rb produces G 1 arrest (17,19,22,30,33,35). In the present experiments, we generated Calu-1 cells expressing levels of p16 INK4A compatible with proliferation by infection with a tetracycline-suppressible retroviral vector, pBPSTR1 (25), encoding human p16 INK4A (31). Figure 1A shows that several independently isolated clones expressed p16 INK4A in the presence of tetracycline and that expression could be enhanced three-to fourfold by tetracycline withdrawal. Levels of p16 INK4A expression in the absence of tetracycline were comparable to those seen in WI38 diploid human fibroblasts and somewhat higher than those seen in normal human bronchial epithelial (NHBE) cells (Fig. 1A). Levels of p16 INK4A expressed by engineered Calu-1 cells in the presence of tetracycline were similar to those in NHBE cells. Although ectopically expressed p16 INK4A bound to cdk4 and cdk6 (Fig. 1B), the levels of p16 INK4A expressed in these cells did not affect growth rates (Fig. 2). Immune depletion of p16 INK4A from extracts of these cells revealed populations of free cdk4 and cdk6 (see Fig. 10), and depletion of these cdks removed all detectable p16 INK4A (data not shown). Thus, cdk4 and cdk6 are present in excess over p16 INK4A in these cells, resulting in sufficient Rb phosphorylation to allow Sphase entry.
Response of p16 INK4A -expressing Calu-1 cells to DNA-damaging agents. When control Calu-1 cells that do not express p16 INK4A were treated with ADR or etoposide for 48 h, nearly all cells arrested in G 2 (Fig. 3A). On average, 3% of these cells had a G 1 DNA content after ADR treatment, and 10% had a G 1 DNA content after etoposide treatment (Fig. 3B). In contrast, cells expressing p16 INK4A responded to ADR or etoposide treatment by growth arrest with a significantly larger proportion of cells in G 1 (Fig. 3A). After ADR treatment, on average, 12 and 27% of cells had a G 1 DNA content in the presence and absence of tetracycline, respectively (Fig. 3B). After etoposide treatment, the G 1 DNA proportions were 21 and 32% in the presence and absence of tetracycline, respectively. Other chemotherapeutic agents, such as camptothecin and cisplatin (Fig. 4), as well as ␥-irradiation ( Fig. 4 and Table  1) also induced G 1 arrest in Calu-1 cells only when they expressed p16 INK4A .
The response to DNA-damaging agents was a true cell cycle arrest since the proportion of cells in S phase decreased from 30% in untreated cells to 5 to 10% after treatment, and the G 1 arrest was stable during continued culture in the presence of nocodazole (Fig. 5A). In addition, the G 1 arrest was reversible since arrested cells reentered S phase after cessation of DNAdamaging treatments (Fig. 5B). Thus the G 1 arrest had the characteristics of a checkpoint response.
Response of p53 ؊/؊ fibroblasts to DNA damaging agents. To ensure that the results observed with Calu-1 cells were not restricted to an engineered cell line, we performed similar experiments on fibroblasts derived from p53 Ϫ/Ϫ mice. Figure   FIG 14) in the absence of tetracycline were induced to arrest in G 1 with ADR or ␥-irradiation, resulting in less than 10% of these cells having an S-phase DNA content. FACS analysis performed 18 h after withdrawal of the DNA-damaging treatment shows that cells have reentered the cycle, with more than 20% having an S-phase DNA content. a A mass population of Calu-1 cells infected with a control retrovirus (pB-STR1) or a clone (clone 6) of Calu-1 cells infected with a retrovirus expressing p16 INK4A were treated with 10 Gy of irradiation. At the indicated times, cells were collected and DNA content was determined by FACS analysis. Similar results were obtained with two additional p16 INK4A -expressing clones. The same proportions of G 1 DNA were obtained in the presence and absence of nocodazole at 18 h after treatment.
6A shows that early-passage (passage 5) fibroblasts expressed p16 INK4A and that levels of expression decreased with continued passage. Because of the absence of p53 in these cells, the expectation was that a G 1 arrest checkpoint in response to DNA damage would be absent. Surprisingly, however, Fig. 6B shows that these cells have a persistent G 1 arrest response. The proportion of cells arresting in G 1 in response to ␥-irradiation, ADR, etoposide, or cisplatin correlated with the amount of p16 INK4A expressed, so that DNA damage induced a significant proportion of early-but not late-passage cells to arrest in G 1 . The G 1 arrest in p16 INK4A -expressing early-passage fibroblasts was stable during continued culture in the presence of nocodazole (Fig. 6B). Intermediate-passage cells expressed intermediate levels of p16 INK4A , and the proportion of cells arresting in G 1 in response to DNA-damaging agents correlated with these levels of expression (data not shown). In addition, similar to Calu-1 cells, engineered expression of p16 INK4A in late-passage cells at levels compatible with cell proliferation restored a G 1 arrest response to DNA damage (Fig. 7).

Effect of cdk4 variants and HPV E7 expression on the G 1 arrest response.
To confirm that the G 1 arrest in Calu-1 cells depended on p16 INK4A , we expressed a cdk4 variant in which lysine-35 is replaced by methionine (20) (Fig. 8A). This variant has no kinase activity but binds p16 INK4A efficiently (30), and when overexpressed, it sequesters p16 INK4A , thereby preventing its interaction with endogenous cdk4. Figure 9 shows that expression of cdk4 K35M reversed the G 1 arrest response in p16 INK4A -expressing Calu-1 cells. A similar result was obtained by expressing cdk4 D185N , another kinase-defective variant that may also sequester p16 INK4A (39) (Fig. 8A and 9).
These results suggest that kinase-defective cdk4 variants prevent p16 INK4A from inhibiting endogenous cdk4, thereby permitting Rb phosphorylation and reversal of the G 1 arrest response to DNA damage. Consistent with this model, overexpression of wild-type cdk4 overcame the G 1 arrest response in p16 INK4A -expressing cells (Fig. 8A and 9). This was further confirmed by the ability of another cdk4 variant, cdk4 R24C to reverse the G 1 arrest response ( Fig. 8A and 9). This variant has FIG. 6. Response of p53 Ϫ/Ϫ fibroblasts to DNA-damaging agents. (A) Immunoblot analysis of p16 INK4A expression in late-and early-passage primary embryo fibroblasts from p53 Ϫ/Ϫ mice. The blot was stripped and reprobed for cdk4 and proliferating-cell nuclear antigen to demonstrate equal loading. (B) Early-and late-passage fibroblasts were treated with ␥-irradiation, ADR, etoposide, and cisplatin and then analyzed for DNA content by propidium iodide staining and FACS analysis. Cells treated with ␥-irradiation were analyzed 12 h after treatment, and cells treated with chemotherapeutic agents were analyzed at the end of 24 h of continuous exposure. Early-passage cells were also cultured for an additional 16 h in the presence of nocodazole to demonstrate the stability of the G 1 arrest. full kinase activity but does not bind and is therefore not inhibited by p16 INK4A (41). Finally, the G 1 arrest response was abrogated by expressing the HPV E7 protein, which inactivates pocket proteins (Fig. 8B and 9). Thus, the p16 INK4A -mediated G 1 arrest in response to ADR is dependent on p16 INK4A itself and pocket proteins such as Rb and further depends on inhibition of cdk4.
Effect of DNA-damaging agents on cell cycle protein expression and cdk activity. G 1 arrest following DNA damage involves p53-dependent and -independent events. In cells with wild-type p53, DNA damage results in increased levels of p53 (5,18), which induce increased expression of the non-INK4 cdk inhibitor, p21 WAF1 (6), leading to G 1 arrest (6,11,42). Figure 10A shows that levels of p21 WAF1 did not rise following ADR or etoposide treatment, indicating that the G 1 arrest response was not due to p53-independent recruitment of the p21 WAF1 pathway. Similarly, there was no change in levels of ectopic p16 INK4A in Calu-1 cells (Fig. 10A), in levels of endogenous p16 INK4A in mouse embryo fibroblasts (Fig. 10B), or in levels of cdk4 in both cell types (Fig. 10A and 11B) following DNA damage. In addition, DNA damage did not increase the amount of p16 INK4A complexed to cdk4 and cdk6 (Fig. 10C).
In contrast, ADR treatment and ␥-irradiation routinely produced lower levels of cyclin D1 in engineered Calu-1 cells and in mouse embryo fibroblasts whether or not they expressed p16 INK4A (Fig. 11). In the absence of p16 INK4A , this decrease was not associated with a measurable reduction in cdk4-or cdk6-mediated Rb kinase activity. However, in the presence of p16 INK4A , kinase activity was reduced (Fig. 12).

DISCUSSION
Our data implicate p16 INK4A in a G 1 arrest checkpoint in response to DNA damage. We demonstrated this first in Calu-1 NSCLC cells, which express wild-type Rb but do not express p53 or p16 INK4A and undergo cell cycle arrest in G 2 in response to DNA damage. Although high levels of ectopic p16 INK4A expression in these cells produce G 1 arrest (33), the levels engineered in the present experiments were compatible with cell proliferation, presumably because cdk4 and cdk6 were in stoichiometric excess compared to p16 INK4A , as demonstrated by depletion experiments. While these levels of p16 INK4A expression did not inhibit cell proliferation, they did result in G 1 arrest in response to a wide variety of DNAdamaging agents that exert their effects in different ways. In addition, the proportion of cells arrested in G 1 correlated with the amount of p16 INK4A expressed by the cell population.
The G 1 arrest response in these cells depended on expression of p16 INK4A since it was abrogated by kinase-defective cdk4 variants that sequester p16 INK4A . These variants prevent p16 INK4A from inhibiting endogenous cdk4, thereby allowing cells to proceed through G 1 into S phase. Overexpression of wild-type cdk4 achieved a similar result, again most likely due to p16 INK4A sequestration. Furthermore, G 1 arrest was reversed by a cdk4 variant incapable of binding p16 INK4A and by the HPV E7 protein, indicating that the G 1 arrest response was dependent on the ability of p16 INK4A to inhibit cdk4 activity in a cell with active pocket proteins, such as Rb.
The G 1 arrest response induced by DNA damage was accompanied by decreases in cyclin D1 levels. Although Calu-1 cells engineered to express low levels of p16 INK4A proliferated like their parental cells, it is possible that this amount of p16 INK4A artificially sensitized them to decreases in cyclin D1 and that p16 INK4A is not normally involved in this DNA damage checkpoint. This is unlikely to be the case for several reasons. First, the amount of p16 INK4A expressed by engineered Calu-1 cells was similar to that expressed by WI38 diploid fibroblasts and NHBE cells, indicating that the levels of were transfected with the indicated expression vectors and controls, and pooled populations of stable transfectants were isolated. Cell lysates were analyzed for expression of transfected proteins by immunoblotting with anti-cdk4 or anti-HA antibodies. All transfected cDNAs encoded HA-tagged proteins except for cdk4 K35M (A), and ectopic expression of this transfected cDNA was inferred by the increased amount of total immunoreactive cdk4 compared to pRc/CMV control transfectants. Also shown is the lack of effect on p16 INK4A levels by expression of cdk4, cdk4 variants, or E7. (B) The control for E7 expression was the same expression vector into which a mutant E7 (⌬DLYC) unable to bind Rb was cloned (23). However, expression of E7⌬DLYC could not be documented, and this control should be considered a vector control rather than an inactive E7 control. expression approximated physiological levels seen in nontransformed cells in culture.
In addition, our results were not confined to cell lines in which p16 INK4A was ectopically expressed but were also observed in primary embryo fibroblasts from p53 Ϫ/Ϫ mice. We found that early-passage cells expressed high levels of p16 INK4A and underwent G 1 arrest in response to several DNA-damaging agents. This was a surprising result and indicates that DNA damage-induced G 1 arrest does not solely depend on the presence of p53 or on events immediately downstream from p53 such as induction of p21 WAF1 . Late-passage cells expressed much lower levels of p16 INK4A and no longer demonstrated a G 1 arrest response after DNA damage.
Other reports on studies using human diploid fibroblasts or mouse embryo fibroblasts from wild-type mice have described progressive increases in p16 INK4A levels with continued passage (1,10,32,43). The behavior of the p53 Ϫ/Ϫ mouse embryo fibroblasts was different in our experiments, although we did not analyze p16 INK4A levels at every passage, so increases in p16 INK4A prior to a subsequent decrease may have been missed. Nonetheless, as p16 INK4A expression diminished, the proportion of cells arresting in G 1 in response to DNA damage progressively decreased. Admittedly, a variety of other genetic (or epigenetic) alterations which might have influenced the DNA damage-induced G 1 arrest response could have occurred in these cells during passage. However, there was a clear correlation between the level of endogenous p16 INK4A in these p53 Ϫ/Ϫ fibroblasts and their degree of G 1 arrest in response to DNA-damaging agents.
The precise mechanism by which p16 INK4A functions in a G 1 arrest checkpoint remains to be elucidated. We have demonstrated that DNA damage does not induce an increase in p16 INK4A levels. Instead, we and others have observed significant decreases in levels of cyclin D1 expression following DNA damage (24,27). While this response also occurred in transformed cells that do not express p16 INK4A , it did not result in a diminution of total cdk4 and cdk6 kinase activities as determined by an in vitro kinase assay. In contrast, when transformed cells were engineered to express p16 INK4A , DNA damage-mediated decreases in cyclin D1 levels were associated with a significant decrease in cdk4-and cdk6-mediated Rb kinase activity. Thus, although DNA damage does not produce a change in the levels of p16 INK4A in cells that express it, its presence is required for a decrease in Rb kinase activity and resultant G 1 arrest following DNA damage.
One simple explanation of these findings would be that lower levels of cyclin D1 after DNA damage permitted increased amounts of p16 INK4A to bind to cdk4 and cdk6. However, our data indicate that this is not the case. The depletion experiments (Fig. 10) demonstrated that cdks are in excess and that there are roughly equivalent amounts of p16 INK4A -free cdk4 and cdk6 both before and after exposure to DNA damage. Although we cannot exclude the possibility that subtle quantitative changes in the amount of p16 INK4A complexed to target kinases could result in a biological effect, it is clear that dramatic accumulation of p16 INK4A in cdk4 or cdk6 complexes does not occur following DNA damage.
Rather, it is likely that p16 INK4A plays a passive but essential role in the context of other perturbations that occur in the cyclin D-cdk4-Rb axis after DNA damage. For example, tyrosine phosphorylation of cdk4 is required in order for cells to arrest in G 1 after UV light treatment (37). Therefore, it is possible that a proportion of cdk4 molecules in the cells we examined becomes tyrosine phosphorylated in response to DNA damage. In the absence of p16 INK4A , there must be sufficient amounts of active cyclin D-cdk4 complexes to allow phosphorylation of Rb and bypass of the checkpoint. However, in the presence of p16 INK4A , the available cdk4 following DNA damage that is not inhibited by either tyrosine phosphorylation or p16 INK4A association may become limiting, leaving insufficient active cyclin D-cdk4 to allow cell cycle progression. In this scenario, steady-state levels of p16 INK4A or the amount of p16 INK4A complexed to cdk4 need not change following DNA damage, but the presence of p16 INK4A is nonetheless essential for G 1 arrest to occur.
Another p53-independent contribution to G 1 arrest following DNA damage may come from the redistribution of p27 Kip1 . Unlike INK4 family members, which appear to displace D-type cyclins from cdks, p27 Kip1 can exist in complex with the cyclincdk holoenzyme (26). In growing cells, most p27 Kip1 is associated with cyclin D1-cdk4 (27). UV irradiation has been reported to reduce the levels of both cyclin D1 and p27 Kip1 , but since the reduction in cyclin D1 is greater, p27 Kip1 redistributes to cyclin A-cdk2, causing a reduction in cdk2 kinase activity and G 1 arrest (27). p27 Kip1 is readily detectable in the cells we examined, and its levels do not change after DNA damage (data not shown). We have not yet investigated the association of p27 Kip1 with cyclin-cdk complexes in these cells, but it is tempting to speculate that a more significant redistribution of p27 Kip1 to cdk2 complexes might occur in the presence of p16 INK4A after DNA damage. If so, this mechanism would again be consistent with a necessary but passive role for p16 INK4A in the G 1 arrest response.
The ability of p16 INK4A to participate in G 1 arrest following DNA damage depends on its expression during G 1 . In this respect, p16 INK4A differs from other closely related members of the INK4 family, all of which are potent inhibitors of cdk4 and cdk6. For example, in many cell types, p15 INK4B expression depends on transforming growth factor ␤ treatment (9,29), while the expression of p18 INK4C and p19 INK4D is cell cycle regulated and restricted to S phase (13). p16 INK4A 's expression during G 1 in a wider variety of normal cell types than other INK4 family members (34,36) may explain why it is a more frequent target for inactivation. One would predict that any INK4 family member that is expressed during G 1 in a given cell type might be a tumor suppressor in that cell. Consistent with  Fig. 3. Cell lysates were analyzed by immunoblotting for p16 INK4A , p21 WAF1 , and cdk4. (For p16 INK4A and cdk4, 100 g of total cell lysate protein was analyzed, but because of the low abundance of p21 WAF1 in these cells, 150 g of protein was analyzed and exposure times to X-ray film were prolonged.) (B) Late-and early-passage primary fibroblasts from p53 Ϫ/Ϫ mice were untreated or were treated with ␥-irradiation (␥), ADR, or etoposide as described in the legend to Fig. 6. Cell lysates were analyzed by immunoblotting for p16 INK4A . (C) Calu-1 cells expressing p16 INK4A (clone 14) were left untreated or were treated with ADR, and 44 to 48 h later cells were radiolabeled with [ 35 S]methionine. Some cell lysates were depleted of p16 INK4A by several rounds of immune precipitation. Control lysates and those depleted of p16 INK4A were then analyzed by immune precipitation for p16 INK4A , cdk4, and cdk6. The persistence of cdk4 and cdk6 in extracts depleted of p16 INK4A indicates that the cdks were present in stoichiometric excess over p16 INK4A . Furthermore, the amounts of free cdk4 and cdk6 after ADR treatment were approximately the same as those before treatment (as determined by phosphorescence image analysis), indicating that ADR treatment did not result in an increased proportion of p16 INK4A bound to cdk4 and cdk6. Calu-1 cells infected with pBPSTR1 or with pBPSTR1-p16 INK4A (clone 6) were left untreated (NT) or were treated with ADR (A) or ␥-irradiation (␥) as described in the legend to Fig. 3, and cell lysates were analyzed by immunoblot for cyclin D1 expression. The blots were stripped and reprobed for cdk4 to demonstrate equal loading. (B) Early-and late-passage fibroblasts from p53 Ϫ/Ϫ mice were untreated or treated with ␥-irradiation, ADR, or etoposide (E) as described in the legend to Fig. 6, and cell lysates were examined by immunoblot for cyclin D1 expression. In addition to recognizing the primary cyclin D1 protein, the anti-cyclin D1 antibody used in this analysis also recognizes a less abundant protein of higher apparent molecular weight in murine cells which has been shown by protease mapping to be cyclin D1 as well (2). Similar results were observed with a second anti-cyclin D1 antibody. The blots were stripped and reprobed for cdk4 to demonstrate equal loading.
FIG. 12. Dependence of reduced Rb kinase activity on p16 INK4A expression after DNA damage. Calu-1 cells infected with pBPSTR1 (C) or pBPSTR1-p16 INK4A (clone 14) were untreated (NT) or treated with ADR (A) as described in the legend to Fig. 3. Lysates were subjected to immune precipitation using anti-cdk4 and anti-cdk6 antibodies, and the precipitates were used to phosphorylate GST-Rb in vitro. Kinase assays were analyzed by SDS-PAGE followed by electrophoretic transfer to nitrocellulose. Phosphorylated GST-Rb was detected by exposing the nitrocellulose filter to X-ray film. The presence of equal amounts of cdk4 and cdk6 in the precipitates was then determined by staining the nitrocellulose filter with anti-cdk4 and anti-cdk6 antibodies. this idea, p15 INK4B is expressed during G 1 in T lymphocytes and is a tumor suppressor in T-cell leukemias (12,36).
During tumor progression, loss of the G 1 arrest checkpoint in response to DNA damage is essential for the eventual emergence of a malignant clone. Our data indicate that unless p16 INK4A activity is lost, the cell types we examined can still arrest in G 1 after DNA damage, even in the absence of p53. This may explain why most tumors with wild-type Rb inactivate both p16 INK4A and p53 and why the frequency of p16 INK4A inactivation in human cancers rivals that of p53. However, while inactivation of p16 INK4A in tumors with wild-type Rb is nearly universal, some malignant cells do express wild-type p53. Our data predict that these cells may not have a fully competent G 1 arrest checkpoint in response to DNA damage, despite the presence of p53, and that p16 INK4A expression may augment this response. We are currently testing this hypothesis in tumor cells that express wild-type p53 and in fibroblasts from p16 Ϫ/Ϫ mice.