A Farnesyl-Protein Transferase Inhibitor Induces p21 Expression and G1 Block in p53 Wild Type Tumor Cells*

Farnesylation is required for the membrane partition and function of several proteins, including Ras. Farnesyl-protein transferase inhibitors (FTIs) were developed to prevent Ras processing and thus to be effective agents for the treatment of cancers harboring mutated ras. However, FTIs inhibit the growth of most tumor cells and several xenograft models, irrespective of whether they possess mutated ras. Furthermore, the antiproliferative effect is not correlated with inhibition of Ki-Ras processing; tumors with wild type rasare inhibited, and FTIs are not particularly toxic. These data suggest that the mechanism of FTI action is complex and may involve other targets besides Ras. To begin to understand how FTIs work, we investigated the mechanism of growth inhibition. FTI causes G1 arrest in a subset of sensitive lines. This is accomplished by transcriptional induction of p21, which mediates the inhibition of cyclin E-associated protein kinase activity, pRb hypophosphorylation and inhibition of DNA replication. Induction of p21 is p53-dependent; it does not occur in cells with mutant p53 or in cells expressing human papillomavirus E6. However, neither p53 nor p21 are required for inhibition of cell proliferation. FTI still blocks the growth of cells deficient in these proteins. In the absence of p21, G1 block is relaxed, DNA replication is not affected, and cells become polyploid and undergo apoptosis. These results suggest that farnesylated protein(s) may be involved in regulating p53 function and in coordinating entrance into S, and that the consequences of FTI treatment are a function of the other mutations found in the tumor cell.

The biologic activity of some proteins is dependent upon their isoprenylation. For example, farnesylation of Ras and geranylgeranylation of small GTP-binding proteins of the Rab and Rho families are necessary for proper subcellular localization and for efficient interaction with regulators and effectors (1,2). This led to the idea that specific inhibitors of farnesyl-protein transferase would be useful anticancer agents by virtue of their ability to prevent Ras processing (1,(3)(4)(5). Several natural products with anti farnesyl-protein transferase activity have been identified, but farnesyl-protein transferase inhibitors (FTIs)1 with little activity against related geranylgeranyl-protein transferase I, were also rationally designed (reviewed in Coleman et al. (3)). These selective agents, including the peptidomimetic FTI L-744,832 we used in our studies (6), inhibit protein farnesylation generally and Ha-Ras processing specifically. Accordingly, FTIs effectively revert the phenotype of Ha-ras-transformed fibroblasts and have remarkable effects on animal tumors that are Ha-ras-dependent (7,8). They prevent the development of breast tumors when administered to murine mammary tumor virus-Ha-ras transgenic mice and when these tumors are allowed to develop, cause their complete remission (7,8).
However, it is not clear that the effects of the drug are due predominantly to inhibition of Ras processing. First, the kinetics for inhibition of cell growth and initiation and maintenance of morphological reversion of v-ras-transformed fibroblasts by FTI does not correlate with the kinetics for inhibition of Ras processing (9). Second, a majority of human tumor cell lines are sensitive to FTI whether or not they contain mutated ras genes (6). Furthermore, although the processing of Ki-and N-Ras is much less sensitive to FTI than is that of Ha-Ras (10 -14), tumor cell lines with the former mutations can be very sensitive to the drug (6). Finally, although FTI inhibits the processing of wild type Ha-Ras, the drug is selective for tumor cells and is remarkably nontoxic in vivo (6 -8).
These observations suggest that the phenotype induced by FTI might be due to effects on other farnesylated target proteins (3)(4)(5). These include lamins, nuclear matrix proteins, Rho B, among others. In fact, Prendergast et al. (9,15) have shown that in transformed fibroblasts the growth and morphological changes induced by FTI correlate more closely with inhibition of RhoB processing than with changes in Ras. This is unlikely to be true in epithelial lineages, however, in which inhibition of growth by FTI is unassociated with characteristic changes in morphology and cytoskeletal architecture.
The mechanism underlying FTI action is of interest for several reasons. The action of the drug is likely to be secondary to specific inhibition of farnesyl-protein transferase. This suggests that one or several farnesylated protein(s) yet to be identified (or to be assigned such a function), play a critical role in the biology of transformation. Second, several FTIs are currently entering phase I trials as anticancer agents in humans.
Rational clinical development will depend on an understanding of the biochemical basis for their inhibition of cell growth. We undertook to approach these questions by examining the mechanism(s) by which FTI arrested the growth of cancer cells. We found that in a subset of sensitive lines, FTI arrests cells in either G 1 or G 1 and G 2 . In these cell lines, FTI caused a p53-dependent induction of p21waf1/cip1/sdi, which led to in-hibition of cyclin E-and A-associated protein kinase activities, accumulation of hypophosphorylated retinoblastoma protein (pRb), inhibition of DNA synthesis, and cell cycle arrest. However, whereas p21 induction was necessary for G 1 block, it was not necessary for the cytotoxic effects of the drug. In cells in which p53 or p21 were knocked out, FTI caused endoreduplication of DNA and increased apoptosis. These data show that FTI affects cells in a complex manner that is, as it might be expected, dependent upon the complement of mutations in cell cycle regulatory genes in the tumor cell. They also suggest that farnesylated protein(s) may be involved in regulating cellular activity of p53.

EXPERIMENTAL PROCEDURES
Materials-FTI L-744,832 was a generous gift from Drs. N. Kohl  Cell Proliferation and Soft Agar Growth-To investigate the effect of the FTI L-744,832 on the proliferation of MCF7 and HCT116 cells, growth curves and soft agar colony formation assays were performed as described in Sepp-Lorenzino et al. (6). In order to collect cells for cell cycle analysis and apoptosis and to obtain protein samples, cultures were seeded at a density of 2,500 cells/cm 2 and incubated with or without 20 M FTI for increasing time periods. Cells floating in the medium were collected by centrifugation, and attached cells were trypsinized. These populations were combined. The cell number was determined with a Coulter counter, and aliquots were prepared for protein and kinase analysis and determination of cell cycle parameters and apoptosis.
Flow Cytometry and Determination of Apoptosis-Nuclei were prepared according to Giaretti and Nusse (17), stained with ethidium iodide, and analyzed with a Becton Dickinson fluorescence-activated cell sorter. Conversely, samples for whole cell flow analysis were prepared according to Nusse et al. (18). Dual label FACS analysis was employed to determine the effect of the drug on DNA replication. Thirty minutes before harvest, cells were pulsed-labeled with 10 M bromodeoxyuridine (BrdUrd). Samples were fixed and stained with anti-BrdUrd monoclonal antibodies conjugated with fluorescein (Becton Dickinson) and with propidium iodide prior to FACS analysis.
Apoptosis was assessed with quantitative fluorescence microscopy and terminal deoxynucleotidyl transferase assays combined with flow cytometry. Terminal deoxynucleotidyl transferase assays were performed with the apo-BrdUrd kit from Pharmingen according to the manufacturer's instructions.
For quantitative fluorescence microscopy, 10 5 cells were fixed in 3% paraformaldehyde in PBS and stored at 4°C until further analysis. Cells were then washed with PBS and resuspended in 25 l of PBS containing 3 g/ml bisbenzimide. Over 300 cells were counted with a fluorescence microscope, and the percentage of apoptosis was determined.
RNA Isolation and Northern Blots-Cells were extracted with RNAzol (Quality Biotech). Thirty g of total RNA was analyzed by Northern blotting using a HindIII-XbaI 0.7-kilobase pair fragment of p21 cDNA as a probe (kindly provided by Dr. J. Massagué, Memorial Sloan-Kettering). The probe was labeled by random priming. Blots were also probed for a housekeeping gene (glyceraldehyde-3-phosphate dehydrogenase) as a control for RNA loading. p21 and glyceraldehyde-3-phosphate dehydrogenase mRNA levels were quantitated with a Phosphor-Imager and p21 mRNA levels were normalized to glyceraldehyde-3phosphate dehydrogenase mRNA.
Protein Analysis-Extracts for protein analysis were prepared from PBS-washed cell pellets that had been stored at Ϫ20°C. Pellets were lysed with 2% SDS, for pRb detection or with 1% Nonidet P-40 for Cdk activity assays. For total cell lysates, pellets were dissolved in 2% SDS in 50 mM Tris, pH 8, boiled 10 min, and sonicated three times for 30 s each time on ice. Nonidet P-40 lysates were obtained by incubating the cell pellets on ice for 15 min in 1% Nonidet P-40, 50 mM Tris, pH 7.4, 150 mM NaCl, 40 mM NaF, 1 mM V0 4 Na 3 , 10 g/ml each of soybean trysin inhibitor, aprotinin, and leupeptin, and 1 mM phenylmethylsulfonyl fluoride. Insoluble material was removed by centrifugation at 15,000 ϫ g for 5 min. Protein concentration was measured with the bicinchoninic acid protein assay (Pierce). Ten percent polyacrylamide gels were used to detect cyclins, Cdks, and p21, whereas 7% gels were used for analysis of pRb. Gels were transferred onto nitrocellulose and probed with specific antibodies. All the antibodies for immunoblotting were from Santa Cruz. Protein detection was by chemiluminescence (ECL, Amersham Corp.).
In Vitro Kinase Reactions-Cyclin-associated Cdk 2 and Cdc 2 protein kinase activity were assayed in vitro with histone H1 as the substrate. One-hundred g of Nonidet P-40 lysates were immunoprecipitated with 2 g of anti-cyclin antibodies (sc-751 for cyclin A, sc-198 for cyclin E, and sc-181 for cyclin B, Santa Cruz Biotechnology). Immune complexes were washed four times with Nonidet P-40 buffer and twice with 20 mM Tris-HCl, pH 7.4, 7.5 mM MgCl 2 , and 1 mM dithiothreitol. Kinase reactions were carried out for 15 min at 37°C in 40 l of kinase buffer containing 20 mM Tris HCl pH 7.4, 7.5 mM MgCl 2 , 1 mM dithiothreitol, 30 M ATP, 10 Ci of [␥-32 P]ATP, and 2 g of histone H1. Cdk 4 activity was analyzed in anti-Cdk 4 immunoprecipitates with a glutathione S-transferase-pRb fragment (Promega) as a substrate. Reactions were terminated by addition of 5ϫ SDS-polyacrylamide gel electrophoreiss sample buffer and then boiled for 5 min. Proteins were separated on 10% gels and transferred onto nitrocellulose, and blots were exposed to x-ray film or to a PhosphorImager screen. Kinase activity was quantitated with a FUJIX PhosphorImager. Blots were also probed for cyclins, kinases, and p21 using specific antibodies and detection by chemiluminescence (ECL, Amersham).
Transfections-Transient and stable transfections were carried out using Lipofectin (Life Technologies, Inc.) according to the manufacturers' instructions. Transient transfections were employed to study the effect of FTI on the induction of the p21 promoter. Luciferase reporter constructs under control of the p21 promoter, either inact (WWP) or lacking the major p53 regulatory site (DM) (19), were cotransfected with SV40-␤-galactosidase into MCF7 cells. After 16 h, cultures were split into 6-well plates and treated for either 24 or 48 h with or without 20 M FTI in triplicate. Cells were harvested, and aliquots were employed for determination of ␤ -galactosidase and luciferase activity. Luciferase activity was normalized to ␤-galactosidase, which was unaffected by the FTI.
HCT116 E6 transfectants were obtained upon co-transfection of the HPV-E6-AU1 epitope tagged under the control of the SV40 promoter (pSG5-HPV-E6-AU1) (kindly provided by Dr. M. Stöppler, Georgetown University Medical Center) with a neomycin resistance-conferring plasmid (pSV40-neo). Single cell clones were obtained after 2 weeks of growth in the presence of 500 g/ml G418. Clones were characterized by E6 expression (upon immunoprecipitation and immunoblotting with anti-AU1 antibody (BabCo)), by the lower basal levels of p53 and by the decreased induction of p53 by doxorubicin (350 nM for 24 h).

RESULTS
We have previously shown that FTI has a profound inhibitory effect on the anchorage-dependent and -independent growth of a majority of human tumor cell lines (6). In order to begin to identify the target(s) responsible for the antiproliferative effect of the FTI, we investigated the mechanism of growth inhibition in several of the sensitive lines identified in our previous screen (6). FTI inhibited the growth of various cell lines without apparent changes in cell cycle parameters. On the other hand, in another subset of sensitive lines, inhibition of growth was accompanied by cell cycle arrest. For example, the human breast carcinoma cell line MCF7 is sensitive to the FTI (6). These cells are estrogen-dependent human breast cancer cells, harboring wild type p53 and ras, and functional pRb; they posses low levels of epidermal growth factor receptor and ErbB2 and abundant levels of insulin-like growth factor I receptor (20). As shown in Fig. 1A, FTI inhibited MCF7 cell proliferation and resulted in an increase in the proportion of G 1 cells at the expense of cells in S phase, whereas the proportion of cells in G 2 and mitosis remained largely unchanged (Fig. 1B). G 1 arrest was accompanied by hypophosphorylation of pRb as seen in Fig. 1C. Control cells maintained hyperphosphorylated pRb while in a logarithmic growth state; hypophosphorylation of pRb was detected at days 5 and 6 when cultures began to reach confluence.
The mechanism by which FTI induced the hypophosphorylation of pRb and G 1 block was investigated. FTI treatment led to a time-dependent inhibition of cyclin E-and A-associated kinase activities, whereas Cdk 4 kinase activity remained unchanged ( Fig. 2A and not shown). The kinase activity of cyclin E⅐Cdk 2 complexes declined earlier than that associated with cyclin A (24 h versus 48 -72 h). FTI induced a decline in the steady-state levels of cyclin A (2-3-fold) and a slight decrease in Cdk 2, which might explain the effect on kinase activity of this complex (Fig. 2B). On the other hand, the levels of Cdk 4 and cyclins E and D3 remained largely unchanged, whereas FTI increased cyclin D1 levels by 2-3-fold (Fig. 2B).
As cyclin E kinase levels fell in the absence of major changes in cyclin E and Cdk 2 expression, the effects of FTI on the expression of Cdk inhibitors (Cdi) were determined. No changes were observed in the levels of p27, p15, p16, p18, and p19 (not shown); however, p21waf1/cip1/sdi was induced by treatment of MCF7 cells with 20 M FTI (Fig. 3A). We also found that p21 levels rose with increased cell density in control cultures, yet the levels induced by FTI treatment are 2.5-5-fold higher than those of controls. It is also important to point out that the cell density in FTI-treated cultures remained very low compared with control cultures (Fig. 1A), thus if a direct com-parison could be made under equal density conditions the induction of p21 by FTI would be of a greater magnitude.
The kinetics of induction of p21 by FTI were compared with those of doxorubicin, an agent known to induce DNA damage and p21. Cells were treated for increasing time periods with either 20 M FTI or 350 nM doxorubicin and p21 levels were compared with those of untreated controls (Fig. 3B). Doxorubin induced a rapid and marked increase in the steady-state levels of p21, detected after 3 h of drug treatment. In contrast, FTI induction took at least 16 h to become evident. A maximun p21 level was attained after 24 h in FTI when p21 was 5-fold higher than that in untreated controls. The induction of p21 by the FTI was accompanied by an increase in the steady-state levels of p21 mRNA (Fig. 3C). Increased p21 mRNA was also detected in control cultures as cell density increased; however, FTI induction was 4-fold greater (484 versus 2012 arbitrary units).
We observed that the cells in which FTI induced p21 and caused G 1 arrest had wild type p53. MCF7 and human colon carcinoma HCT116 are both p53 wild type cells, and in both, FTI treatment leads to increased p21, Cdk inhibition, Rb hypophosphorylation, and cell cycle block (Figs. 1, 2, and 6). On the contrary, FTI did not induce p21 in cell lines with mutated p53 (such as Colo 205, DU145, PC3, and MDA MB-468) that are sensitive to its effects on cell growth (Fig. 4A). In view of these results we sought to determine whether the induction of p21 in response to the FTI was dependent on p53.
The effect of the FTI on the transcriptional activation of the p21 promoter and its p53 dependence were investigated using a set of reporter constructs in which the luciferase gene was Cells were harvested by trypsinization, and nuclei were prepared and stained with ethidium bromide and analyzed by flow cytometry. C, MCF7 cells grown as described for B were harvested, and total cell extracts were made in SDS buffer. pRb was detected by immunoblotting on 7% polyacrylamide gels.

FIG. 2. FTI induces the inhibition of cyclin E-and A-associated kinases in MCF7.
Cells grown as described for Fig. 1B were harvested and washed with PBS, and protein extracts were made in 1% Nonidet P-40 lysis buffer. A, the kinase activity of cyclin E-and A-containing complexes was assayed in vitro with histone H1 as a substrate. B, steady-state levels of cyclins and Cdks were analyzed by electrophoresis on 10% SDS-polyacrylamide gels and immunoblotting with specific antibodies. driven either by the intact p21 promoter (WWP-luc) or by a truncated p21 promoter, in which the major p53-binding element at Ϫ2.4 kilobase pairs is missing (DM-luc) (19). Transcription from the wild type p21 promoter was induced by the FTI in p53 wild type cells, but not in cells with mutant p53 (Fig.  4B). In addition, the p53 binding site was necessary for transcriptional activation from the p21 promoter by FTI, as no significant changes in transcription were detected from the truncated promoter in p53 wild type cells treated with FTI (Fig. 4B).
The requirement for p53 was further investigated by assessing the effects of FTI on cells engineered to express the E6 gene of the human papilloma virus 16 (HPV16). HPV16 E6 specifically targets p53 for degradation via the ubiquitin-dependent pathway (21). E6 was introduced into HCT116 cells and stable transfectants were isolated 2 weeks later as G418-resistant clones. Expression of E6 was confirmed by immunodetection of the AU1-tagged protein. These clones contained decreased basal levels of p53, and the up-regulation of p53 following DNA damage was attenuated (Fig. 5A). The induction of p21 by doxorubicin was decreased in the E6-expressing clones compared with neomycin-resistant controls, and no increase in p21 occurred in response to the FTI (Fig. 5A). Moreover, the small increase in p53 levels detected in the neomycin controls treated with FTI was absent from the p53 deficient cells. These results support the conclusion that the FTI induces the transcriptional activation of the p21 gene by a p53-dependent mechanism.
Are p21 and p53 required for the antiproliferative effect of the FTI? For this purpose we tested the effect of the FTI on the anchorage-dependent and -independent growth of HCT116 cells expressing HPV16 E6 or lacking the p21 gene (16). As seen in Fig. 5B, inability of the FTI to induce p21 due to absence of the gene or to loss of wild type p53 function did not alter the capacity of the drug to inhibit cell growth. This result is in agreement with our previous finding that FTI effectively inhibited the proliferation of cells with mutated p53 (6). Thus, it seems that p21 induction constitutes only one of the pathways through which FTI inhibits cell growth. If that is the case, what is the mechanism of growth inhibition in the absence of p21? We compared the effects of the FTI on HCT116 parental cells and on their counterparts lacking p21. As expected, only the parental cells could up-regulate p21 in response to FTI treatment and induce the hypophosphorylation of Rb (Fig. 6A). Increased cellular levels of p21 led to enhanced association of the Cdi with cyclin-Cdk complexes and consequential inhibition of cyclin E-and A-associated kinase activity (Fig. 6B). In contrast, in p21Ϫ/Ϫ cells the FTI neither caused decreased Cdk activity or pRb hypophosphorylation (Fig. 6, A and B). These results suggest that the inhibition of cyclin E-associated protein kinase activity and induction of Rb hypophosphorylation by FTI require the p53-dependent induction of p21.
Induction of cell cycle block by FTI also required p21. FTI blocked DNA replication in parental cells; there were 4-fold fewer BrdUrd-positive cells in FTI-treated cultures compared FIG. 3. Up-regulation of p21waf1/ cip1/sdi by FTI treatment. A, MCF7 cell lysates were prepared as described for Fig. 2A, and p21 was detected on total cell extracts by immunoblotting with anti-p21 antibodies. Quantitation was performed with a Bio-Rad image analyzer. B, time course for p21 induction by FTI and doxorubicin. C, p21 mRNA levels were analyzed by Northern blotting and normalized to the levels of glyceraldehyde-3phosphate dehydrogenase. FIG. 4. p21 up-regulation by FTI is dependent on p53 function. A, a set of p53 positive and negative cell lines was tested for the ability of 20 M FTI to induce p21 as described for Fig. 3. B, activation of the p21 promoter by FTI requires an intact p53-binding site and wild type p53 activity. MCF7 (p53ϩ/ϩ) and MDA MB 468 (p53 mut) cells were transiently transfected with reporter constructs driven by the full-length p21 promoter (WWP) or a truncated promoter lacking the major p53 regulatory site (DM). Twenty-four h after transfection, cells were treated for an additional 24 or 48 h with or without 20 M FTI. Luciferase activity was normalized to the activity of cotransfected SV40-␤-galactosidase. Each point was done in triplicate. The graph represents the average of three independent experiments for the 24-h time point. At 48 h of treatment proportional results were obtained.

p21 Mediates Cell Cycle Block by FTI
with untreated controls (Fig. 6C). Conversely, p21Ϫ/Ϫ cells in which the FTI did not inhibit Cdk activity, continued to synthesize DNA in the presence of the drug (Fig. 6C). FACS analysis corroborated these results and revealed that the lack of Cdk inhibition and consequential unrestricted DNA replication was associated with the formation of polyploid cells (Fig.  7A). While the FTI caused parental cells to arrest in G 1 and G 2 at the expense of cells in S, the response of p21 knockout cells was different; S phase remained prominent, the G 1 block was relaxed, and cells accumulated with greater than 4n DNA content. Ungated FACS analysis revealed the appearance of 8n and 16n peaks in the p21Ϫ/Ϫ cells treated with the drug (Fig.  7A). FTI also caused the development of polyploid cells in cells expressing HPV16 E6 (Fig. 7B).
Bisbenzimide staining showed that parental cells exposed to the FTI were predominantly in interphase, whereas p21 knockout cultures showed an increased proportion of cells in mitosis (3.7% versus 9.25%) (Fig. 8, A-D). Double-staining with antitubulin antibodies provided evidence for endoreduplication in the FTI-treated p21Ϫ/Ϫ cells, as mitotic forms with four centrioles were common (Fig. 8E). However, some of these mitosis were abnormal and seemed to lead into apoptosis (Fig. 8D). Cell cycle analysis in combination with terminal deoxynucleotide transferase assays corroborated these results and suggested that inhibition of protein farnesylation in the absence of a p21-mediated cell cycle block, translates into increased vulnerability for cell death. DISCUSSION FTIs were developed as drugs that would cause the specific inhibition of Ras processing and would thus be effective agents for the treatment of tumors with mutations in the ras protooncogene (1,3,4). Potent and highly specific FTIs are now available, and they have proven in preclinical studies to be promising drugs. However, contrary to expectations, FTIs inhibit the growth of tumor cells with wild type ras and inhibit the processing of this protein while causing only minor toxicity to normal tissue (6 -8). Moreover, FTIs inhibit the growth of tumor cells containing mutations in the Ki-ras gene (6) at doses that inhibit the processing of Ha-Ras but not Ki-Ras (10 -14). 2 These data suggest both that FTI may exert its effects on cancer cells by inhibiting other targets (for example, Rho B (9,15,22)) and that the role played by activated Ras in human epithelial tumors is poorly understood. We undertook studies to determine the mechanism of induction of growth inhibition by FTI in order to develop a context in which to investigate these issues.
Most tumor cell lines, which were sensitive to the FTI in terms of inhibition of growth, did not undergo a cell cycle specific growth arrest in response to the drug. However, a subset of sensitive lines did block in G 1 or in G 1 and G 2 . These results are in agreement with the recent observations of Sebti and collaborators (23,24). They compared the cell cycle effects of the farnesyl-protein transferase inhibitor FTI-277 and of a geranylgeranyl-protein transferase I inhibitor in human tumor cell lines. Unlike inhibition of protein geranylgeranylation, which caused a G 0 /G 1 arrest in all cell lines examined, inhibition of farnesylation resulted in either no effect, or in G 1 or G 2 /M blockade, in a cell line-specific fashion.
We concentrated our efforts in trying to understand whether there was a pattern that would explain the cell cycle-specific effects of the FTI. We found that cell lines that arrested in G 1 (or in G 1 and G 2 /M), were wild type for p53. The three p53-wild type human tumor cell lines we studied, MCF7 (breast), HCT116 (colon), and LNCaP (prostate), arrested in G 1 in response to FTI treatment (Figs. 1B, 6B, and 7). 3 The only other cell line in our inventory that is p53 wild type is the immortalized human breast epithelial line MCF10A. However, we have previously shown (6) that untransformed cells are severalfold more resistant for growth inhibition than cancer lines, thus, at 20 M FTI no cell cycle effects were observed (not shown).
As shown in this study for MCF7 and HCT116, growth arrest was associated with activation of p53, inhibition of cyclin E-dependent protein kinase, and hypophosphorylation of the Rb protein. Inhibition of cyclin E⅐Cdk 2 occurred prior to cell cycle arrest and in the absence of significant changes in cyclin E⅐Cdk 2 or increase in p27 expression (Figs. 2 and 6, and not shown). We investigated the mechanism of cyclin E-dependent protein kinase inhibition and found that it was mediated by a p53-dependent induction of the Cdi p21waf1/cip1/sdi. p21 plays a crucial role in controlling G 1 /S progression and in coordinating the G 1 and G 2 checkpoints (16, 19, 26 -28). Besides inhibiting Cdk activity, p21 also affects DNA replication (but not DNA repair) directly, by binding to and interfering

p21 Mediates Cell Cycle Block by FTI
with the role of proliferating cell nuclear antigen in replication (29 -33). Moreover, p21 was also shown to regulate the activity of stress-activated protein kinases (34). We found that increased p21 was responsible for the FTI-induced inhibition of cyclin E⅐Cdk 2 kinase activity, which in turn resulted in decreased Rb phosphorylation, inhibition of G 1 /S progression and cell cycle arrest. HCT116 cells in which p21 had been knocked out no longer inhibited cyclin E-dependent protein kinase activity in response to the FTI, nor did they arrest in G 1 (Figs. 6 and 7). The results strongly suggest that the primary basis for induction of G 1 arrest by FTI is induction of p21.
The increase in p21 levels by the FTI was detected after 16 h of drug treatment and reached a maximum after 48 h (Fig. 3, A and B) when nuclear accumulation was evident (data not shown). Nuclear translocation of p21 following up-regulation was also detected in other cell systems (35,36). The increase in p21 protein levels was explained, in part, by a concomitant elevation in the levels of p21 mRNA (Fig. 3C). We cannot FIG. 6. Up-regulation of p21 is necessary for the FTI-induced hypophosphorylation of Rb, for the inhibition of cyclin E-and A-associated kinase activities, and for inhibition of DNA replication. Parental and p21 knockout cells were seeded at 2,500 cells/cm 2 and treated with 20 M FTI for increasing time periods. A, p21, p53, and Rb were detected by immunoblotting with specific antibodies and B, the activity of Cdk⅐cyclin complexes was assayed as in Fig. 2. Immunoprecipitates were immunoblotted for p21. C, cultures treated or not with 20 M FTI for 72 h were pulsed-labeled with BrdUrd for 30 min prior to harvesting. Dual-stain immunofluorescence was conducted by detecting incorporated BrdUrd with monoclonal anti-BrdUrd antibodies and secondary antibodies conjugated to fluorescein isothiocyanate and by staining total DNA with bisbenzimide. The photographs depict representative frames. Quantitation was performed by counting over 300 cells per condition.
p21 Mediates Cell Cycle Block by FTI exclude at the moment the participation of other types of posttranslational regulation as it was reported that p21 levels can be controlled by changes in protein and mRNA stability (37)(38)(39).
We also detected elevated levels of p21 in untreated cells as their growth rates decreased as a function of cell density (Fig.  3, A and B). Given that the cell density of the FTI-treated cultures is severalfold lower that that of untreated controls (Fig. 1A), the mechanism for p21 up-regulation due to FTI treatment is density-independent. Another observation from our kinetic studies is that the acumulation of p21 due to treatment with FTI is delayed compared with the increase caused by direct DNA damage inflicted with doxorubicin (Fig. 3B). It is possible that this variation reflects the two different mechanism of actions for these drugs. While the effect of doxorubicin is direct and p21 is rapidly induced, the lag time observed in FTI-treated cultures might reflect the half-life of a farnesylated protein.
FTI caused a p21-dependent G 1 arrest in only a few epithelial tumor cell lines, and these all contained wild type p53. Reduction of p53 expression by introduction of E6 protein prevented FTI induction of p21 (Fig. 5A); cells expressing mutated p53 did not up-regulate p21 in response to the FTI, whereas wild type cells did (Fig. 4A), and translocated p21 into the nucleus (data not shown), an indicator of wild type p53 function (36). Second, FTI induced transcription from the p21 promoter ( Fig. 4B) only in p53 wild type cells, and this required the major p53 binding site in the promoter (Fig. 4B). Deletion of the p53 binding sequence at Ϫ1.3 kilobase pairs (19) decreased the magnitude of the FTI induction (Fig. 4B). The truncated promoter might still contain a weak p53-responsive site at Ϫ75 which could account for the residual activity seen by us and others (19,39). Third, we investigated whether FTI altered p53 levels. In MCF7 cells, p53 levels were unchanged by FTI (data not shown). However, MCF7 cells express high basal levels of wild type p53 and p53 activation is not associated with further increases in its steady-state levels (40,41). On the contrary, HCT116 cells have low basal levels of p53, which become elevated upon DNA damage. In these cells, the FTI led to a modest induction of p53 (1.5-2-fold increase) as detected by immunoblotting (Figs. 5A and 6A). Finally, it was reported that activation of p53 is correlated with up-regulation of cyclin D1 (42)(43)(44). We did observe an increase in the steady-state levels of cyclin D1 (2-3-fold) in response to FTI treatment (Fig. 6B). The weight of the evidence is, then, that FTI induces p21 via a p53-dependent pathway. The mechanism whereby FTI activates p53 transcriptional activity is unknown. The p53 effect could result from induction of low level DNA damage, but we have no evidence for this. It is also possible that a farnesylated protein plays a role in maintaining DNA stability or functions in the p53 sensor pathway, or that the FTI suppresses a pathway that down-regulates both p53 function and DNA replication.
The majority of tumor cell lines with mutated p53 were growth inhibited by FTI, but did not arrest in G 1 (6) (data not shown). It was not surprising, then, to find that when p21 was knocked out from or E6 was expressed in tumor cells with wild type p53, FTI still caused cell growth arrest and death, although it no longer caused G 1 block or inhibition of DNA synthesis (Figs. 6 and 7). Instead, unscheduled endoreduplication of DNA occurred and polyploid cells underwent apoptosis (Figs. 6 -8). Polyploidy was also induced in HCT116 p21Ϫ/Ϫ cells by doxorubicin and other chemotherapeutic agents (28). Although many parallels can be drawn between the response of p21Ϫ/Ϫ cells to DNA damaging agents and to the FTI, there are also differences, as FTI-treated cells entered mitosis (Fig. 8,  D and E), whereas other treatments led to arrest in a "G 2 -like state" (16,28). It is possible that the mismatch repair deficiency of HCT116 cells (45) may contribute to polyploidy; however, rereplication was also observed in human and mouse cells lacking p53, Rb, or p21 when treated with drugs that interfere with mitotic spindle assembly (25,46). In addition to allowing cells to continue to cycle, loss of checkpoint control under these conditions is associated with bulk replication of DNA and apoptosis (25). The mechanism underlying this phenomenon is obscure. In MCF7 and HCT116 cells, the effects of the FTI vary as a function of p53 and p21 status. This reinforces the point that the effects of perturbing particular regulatory pathways are dependent upon the cellular context. HCT116 contains mutated Ki-ras; MCF7 contains wild type ras. The effects of FTI may vary as a function of ras genotype, but in these cells, the FTI effect on cell cycle was primarily dependent on p53 status. These data suggest that the consequences of administering FTI to patients with tumors with or without mutated p53 will be different. Clinical protocols will have to stratify patients for p53 as well as ras status. Tumors with mutant p53 and wild type ras, such as some breast, prostate, and small cell lung cancers, may be especially sensitive to treatment with DNAdamaging agents in combination with FTI.