Role of p53 and p21waf1/cip1 in senescence-like terminal proliferation arrest induced in human tumor cells by chemotherapeutic drugs

Article metrics

Abstract

Exposure of human tumor cell lines to moderate doses of anticancer agents induces terminal proliferation arrest accompanied by morphologic and enzymatic changes that resemble senescence of normal cells. We have investigated the role of p53 and p21waf1/cip1 in the induction of this response in drug-treated tumor cells. Doxorubicin treatment induced the senescence-like phenotype (SLP) and its associated terminal growth arrest in wild-type HCT116 colon carcinoma cells; this response was strongly decreased but not abolished in HCT116 lines with homozygous knockout of p53 or p21. Transduction of HT1080 fibrosarcoma cells with a genetic inhibitor of p53 also decreased the induction of SLP and increased drug-induced mitotic cell death. To determine if drug-stimulated p21 expression was responsible for senescence-like growth arrest, we have expressed different levels of p21 from an inducible promoter. While high-level overexpression of p21 was sufficient to induce SLP in HT1080 cells, the levels of p21 expressed in doxorubicin-treated cells could account for only a fraction of doxorubicin-induced SLP. Our results indicate that p53 and p21 act as positive regulators of senescence-like terminal proliferation arrest, but their function is neither sufficient nor absolutely required for this treatment response in tumor cells.

Introduction

The proliferative lifespan of normal mammalian cells is limited by replicative senescence (Hayflick and Moorhead, 1961; Campisi, 1997; Duncan and Reddel, 1997), a slow process (50 cell divisions) that appears to be primarily mediated by gradual shortening of telomeres (Shay et al., 1997). A phenotypically similar but much more rapid process is accelerated senescence that can be triggered by DNA-damaging drugs or γ-irradiation (Chen and Ames, 1994; DiLeonardo et al., 1994; Robles and Adami, 1998) or by the introduction of ras or raf oncogenic mutants into normal cells (Serrano et al., 1997; Zhu et al., 1998). Accelerated senescence, like apoptosis, was suggested to be a programmed protective response of the organism to potentially carcinogenic impact (Weinberg, 1997). Both types of senescence are associated with a cellular senescence-like phenotype (SLP), markers of which include enlarged and flattened cell shape, increased granularity and expression of the senescence-associated β-galactosidase (SA-β-gal) active at pH 6.0 (Dimri et al., 1995). Terminal proliferation arrest in both forms of senescence was shown to require the function of p53 and of cyclin-dependent kinase inhibitors, including p21waf1/cip1, the primary mediator of p53-regulated growth arrest, and p16INK4a (Duncan and Reddel, 1997; DiLeonardo et al., 1994; Serrano et al., 1997).

The process of neoplastic transformation involves escape from replicative and accelerated senescence through such mechanisms as the activation of telomerase (an enzyme complex that prevents telomere shortening; Shay, 1997), overexpression of E1A-class oncogenes and mutational inactivation of p53 or p16 (Serrano et al., 1997). Nevertheless, immortal tumor-derived cell lines can undergo terminal growth arrest with features of SLP upon genetic modification, such as somatic cell fusion (Pereira-Smith and Smith, 1988) or ectopic overexpression of p53, RB, p16 or p21 tumor suppressor genes (Sugrue et al., 1997; Xu et al., 1997; Uhrbom et al., 1997; Vogt et al., 1998; Wang et al., 1998b). We have recently demonstrated that the same effect can be achieved in cell lines derived from different types of human solid tumors by treatment with chemotherapeutic drugs, ionizing radiation or differentiating agents (Chang et al., 1999). Induction of SLP, along with mitotic death and apoptosis, is a major response of tumor cells to moderate doses of cytotoxic drugs, and it is the most prominent marker of permanent growth arrest induced by cytostatic differentiating agents. Analysis of the induction of SLP in different tumor-derived cell lines suggested that this response may partly correlate with the cellular p53 status (Chang et al., 1999). In the present study, we have investigated the role of p53 and its downstream mediator of growth arrest, p21, in the induction of SLP in tumor cell lines treated with anticancer drugs. Our results indicate that p53 and p21 act as positive regulators of drug-induced senescence-like terminal proliferation arrest, but the function of these genes is neither necessary nor sufficient to account for this response to anticancer agents.

Results

Effects of p21 and p53 knockouts on doxorubicin response in HCT116 colon carcinoma cells

To investigate the role of p53 and p21 in drug-induced SLP, we have used sublines of HCT116 colon carcinoma cells, where p53 or p21 genes had been inactivated by homozygous knockout (Waldman et al., 1995; Bunz et al., 1998). p53−/−, p21−/− and wild-type HCT116 cell lines were exposed to moderate doses (50 or 100 nM) of doxorubicin. DNA content analysis of cells that remained attached during doxorubicin treatment (Figure 1a) showed a clear difference between the wild-type and both knockout lines in the cell cycle effects of the drug, in agreement with previous studies on p21−/− cells (Waldman et al., 1996). Doxorubicin (which acts primarily at topoisomerase II by stabilizing ‘cleavable complexes’ of this enzyme with DNA; Liu, 1989) induced the accumulation of cells in G2, but a substantial fraction of the wild-type cells remained in G1 throughout drug treatment. The G1 fraction, however, was not retained in p53−/− and p21−/− cell lines (Figure 1a). In addition, doxorubicin treatment led to the appearance of a substantial fraction of apparently polyploid nuclei in p53−/− and p21−/− lines, while few such cells arose in the wild-type line (Figure 1a). These changes in the cell cycle response to doxorubicin have been previously explained by the role of p21 (whose expression is regulated primarily by p53) in damage-induced G1 and G2 growth arrest (Waldman et al., 1996, Bunz et al., 1998). In agreement with this interpretation, the cell cycle changes were similar in both knockout lines but somewhat more pronounced in p21−/− than in p53−/− cells.

Figure 1
figure1

Effects of p53 and p21 inhibition on cell cycle response to doxorubicin. (a) FACS analysis of DNA content of attached HCT116 wild-type, p21−/− and p53−/− lines. Top to bottom: untreated cells, cells treated with 50 nM doxorubicin for 1 day (1d) or 3 days (3d). (b) FACS analysis of DNA content of control and L56SN-transduced HT1080 3′SS6 cells. Untreated cells (top), cells treated with 20 nM doxorubicin for 1 day (middle) or for 3 days (bottom)

Analysis of doxorubicin-induced cell death, however, revealed a substantial difference between the p21−/− and p53−/− lines. As illustrated in Figure 2a, morphological types of cell death in doxorubicin-treated HCT116 line included cells with apoptotic morphology (Figure 2a), as well as large cells containing many separate micronuclei with evenly stained chromatin (mn), characteristic of mitotic catastrophe (Lock and Stribinskiene, 1996; Hendry and West, 1997; Torres and Horwitz, 1998; Chang et al., 1999). In agreement with previous reports (Waldman et al., 1996), drug-induced cell death was drastically increased in the p21−/− line. While the wild-type cells treated for 4 days with 50 nM doxorubicin showed apoptosis or mitotic death in a total of 38% of the attached cells (Figure 3a), this fraction was increased to 90% in p21−/− cells (Figure 3a). In contrast to the p21 knockout line, doxorubicin-treated p53−/− cells showed only a minor increase relative to the wild-type line in the percentage of the attached apoptotic and micronucleated cells (50% total, Figure 3a). Long-term cell survival assays also indicated a strong increase in drug sensitivity for p21−/− and a modest increase for p53−/− cell lines relative to the wild-type cells (data not shown).

Figure 2
figure2

Microscopic analysis of cell death and SA-β-gal expression in HCT116 cells. Cells were treated with 50 nM doxorubicin for 4 days (a, c – f) or 100 nM doxorubicin for 1 day and grown without drug for 10 days to allow colony formation (b), fixed with 2% formaldehyde/0.2% glutaraldehyde and stained with X-gal for SA-β-gal activity. In (a), cells were counterstained with hematoxilin and eosin and photographed at phase contrast with 400-fold magnification (20 μM scale bar). The photographs were taken at 400-fold magnification (50 μM scale bar) for (b) and 200-fold (20 μM scale bar) for (c – f). (a) Cells with multiple micronuclei (mn) and apoptotic cells (a) in doxorubicin-treated wild-type HCT116 line. (b) SA-β-gal expression in colonies and single cells of doxorubicin-treated wild-type HCT116 cells. (c) SA-β-gal expression in untreated wild-type HCT116 cells. (d) SA-β-gal expression in doxorubicin-treated wild-type HCT116 cells. (e) SA-β-gal expression in doxorubicin-treated p21−/− HCT116 cells. (f) SA-β-gal expression in doxorubicin-treated p53−/− HCT116 cells

Figure 3
figure3

Effects of p53 and p21 inhibition on the induction of cell death and SLP in doxorubicin-treated HCT116 cells. (a) Percentages of attached HCT116 wild-type, p21−/− and p53−/− cells with apoptotic morphology (A), cells with multiple micronuclei (MN), or SA-β-gal+ cells, and the percentage of SA-β-gal+ cells in the dying (A+MN) population after 4-day exposure to 50 nM doxorubicin. The percentages of dying or SA-β-gal+ cells in untreated cell populations were 2%. All the experiments were repeated with similar results. Error bars represent the Poisson standard deviation calculated as the square root of counted events and expressed as per cent abundance. (b) FACS analysis of cell division in HCT116 wild-type, p21−/− and p53−/− lines treated for 1 day with 100 nM doxorubicin and grown without drug for 6 days. Prior to drug treatment, cells were labeled with PKH2 and 10 000 cells were analysed in each sample. D: dividing cells. GA: growth-arrested cells

The induction of SLP in doxorubicin-treated HCT116 cells was analysed by staining cells for SA-β-gal activity using 5-bromo-4-chloro-3-indolyl β-D-galactoside (X-gal) at pH 6.0. As illustrated in Figure 2c,d, doxorubicin treatment efficiently induced SA-β-gal expression and morphologic features of SLP (enlarged and flattened shape and increased granularity) in wild-type HCT116 cells; the SA-β-gal+ fraction increased during continuous treatment with 50 nM doxorubicin reaching 60% level after 4 days of drug exposure (Figure 3a). To analyse the association of SLP with proliferative failure, wild-type HCT116 cells were treated for 1 day with 100 nM doxorubicin and grown in drug-free media for 10 days. By this time, the cell death process was essentially completed, and the surviving cells either formed colonies or remained on the plate as single cells or small clusters. As illustrated in Figure 2b, the growing colonies were SA-β-gal−, but most of the non-proliferating cells were SA-β-gal+, indicating that SLP was associated with HCT116 cells that undergo long-term growth arrest after release from the drug.

The p21−/− and p53−/− lines also showed SLP induction after exposure to doxorubicin (Figure 2e,f). The frequencies of SA-β-gal+ cells after 4 day treatment with 50 nM doxorubicin reached close to 20% in the knockout lines (as compared to 1 – 2% in the untreated subconfluent cells) (Figure 3a), indicating that p53 and p21 are not absolutely required for this response. On the other hand, the SA-β-gal fraction in both knockout lines was significantly (three times) lower than in the wild-type HCT116 cells treated in parallel (Figure 3a). In all three cell lines, the SA-β-gal+ fraction of the dying (apoptotic+micronucleated) cells was similar to the SA-β-gal+ fraction in the total cell population (Figure 3a), indicating that SLP and cell death were independent responses to doxorubicin treatment and that decreased SLP induction in the knockout lines was independent from the effect of the knockouts on drug-induced cell death.

To determine if the observed decrease in drug-induced SLP in p21 and p53 knockout lines was associated with lower proliferative arrest after drug treatment, we have analysed the division of drug-treated cells using a lipophilic fluorescent compound PKH2. PKH2 incorporates stably into the plasma membrane and is evenly divided between daughter cells, allowing one to distinguish cells that underwent a different number of divisions by their decreasing PKH2 fluorescence (Horan and Slezak, 1989). The wild-type, p53−/− and p21−/− cells were labeled with PKH2, exposed to 100 nM doxorubicin for 24 h and plated in drug-free media; changes in PKH2 fluorescence were then analysed by FACS. In agreement with previous reports of strong damage-induced growth arrest in the wild-type HCT116 line (Waldman et al., 1996, 1997), only a few of these cells divided by day 3 after release from the drug. This growth arrest, however, was partly reversible, and a substantial fraction of dividing cells emerged by day 6 after release from the drug (fraction D in Figure 3b), although 43% of the wild-type cells remained growth-arrested (GA). The GA fraction, however, was greatly diminished (down to 11%) in p21−/− cells treated and analysed in parallel (Figure 3b). The GA fraction of p53−/− cells was also diminished relative to the wild-type cell line, albeit to a lesser extent than in p21 knockout cells (29%) (Figure 3b). These results demonstrate that p21 and p53 knockouts in HCT116 cells decrease the induction of SLP and its associated long-term growth arrest in drug-treated cells.

p53 inhibition increases mitotic death and decreases drug-induced SLP in HT1080 fibrosarcoma cells

The above observations with p53−/− and p21−/− HCT116 lines strongly suggested that p53 and p21 play a role in drug-induced SLP, but the clonal nature of the wild-type and knockout HCT116 cell lines did not allow us to rule out the possibility that the observed differences could be due to clonal variation. We have therefore investigated changes in cellular drug response resulting from the inhibition of p53 function in a mass population of cells transduced with a p53-inhibiting genetic suppressor element (GSE). As the recipient cell line, we have used subline 3′SS6, derived from HT1080 human fibrosarcoma cells after transfection with the murine ecotropic retrovirus receptor (which renders cells highly susceptible to infection with ecotropic retroviruses) and with the modified bacterial LacI repressor (which allows for isopropyl-β-thio-galactoside (IPTG)-regulated gene expression) (Chang and Roninson, 1996). The p53 status of HT1080 3′SS6 cells was determined by PCR amplification and sequencing of all the protein-coding exons of the p53 gene; no p53 mutations were detected by this analysis. To inhibit the wild-type p53 function in HT1080 3′SS6 cells, we used GSE56, a p53-derived peptide that inhibits all the tested biological functions of wild-type p53 (Ossovskaya et al., 1996). 3′SS6 cells were infected with the ecotropic retrovirus L56SN that carries GSE56 and the neo (G418 resistance) gene, or with a control insert-free retrovirus. Cell populations transduced with nearly 100% efficiency (as determined by comparing cell growth in the presence or absence of G418) were analysed without further selection. This experimental design assures that the differences between the experimental and control cells would not be due to clonal variability or artifacts of selection.

The cellular levels of p21 protein in GSE56-transduced and control cells were measured by a quantitative ELISA assay, before and after treatment with doxorubicin. Doxorubicin treatment of the control 3′SS6 cells resulted in time- and dose-dependent induction of p21, with the highest levels observed after 48 h of drug exposure (data not shown). Transduction with L56SN resulted in almost complete inhibition of basal and drug-induced levels of p21 (Figure 4a), indicating that p21 expression in HT1080 cells was strongly dependent on p53. Similar results were obtained by Western blotting analysis of p21 and another p53-regulated protein, Mdm2, and by RNA analysis of p21 expression (data not shown). To determine if GSE56 had the expected effect on cell cycle response to doxorubicin, L56SN-transduced and control cells were incubated for 3 days with 20 nM doxorubicin and then released into drug-free medium. FACS analysis of DNA content (Figure 1b) revealed the expected differences between the two populations. In particular, GSE56 strongly decreased the fraction of cells that remained in G1 during doxorubicin treatment and slightly increased the fraction of polyploid cells (Figure 1b and data not shown). While these changes resembled the effects of p21 or p53 knockout in HCT116 cells (Figure 1a), the effect of p53 inhibition was somewhat smaller in HT1080 cells, reflecting a lower stringency of drug-induced G1 and G2 arrest in wild-type HT1080 relative to HCT116 cells.

Figure 4
figure4

Effects of p53 inhibition on the induction of cell death and SLP in drug-treated HT1080 cells. (a) p21 levels in control and L56SN-transduced HT1080 3′SS6 cells, determined by ELISA assay. Cells were untreated or treated with 50 nM doxorubicin for 48 h. (b) FACS analysis of GFP fluorescence in LmXSE-transduced control (left) and Lm56SE-transduced (right) HT1080 3′SS6 cells treated with 30 nM doxorubicin. Top to bottom: untreated cells, cells treated with doxorubicin for 2 days and released for 1 day (2+1d rel) or for 5 days (2+5d rel). (c) Percentages of apoptotic (A) or micronucleated (MN) cells attached after 3 day treatment of control and L56SN-transduced HT1080 3′SS6 cells with 20 nM doxorubicin. The percentages of dying cells in untreated cell populations were 2%. (d) Percentages of SA-β-gal+ cells in control and L56SN-transduced HT1080 3′SS6 populations after 3 day treatment with 20 nM doxorubicin (shown for total cells and for the dying (A+MN) cell population), 1 μg/ml aphidicolin, 2 μM cisplatin, 2 μM cytarabine, 1000 RAD γ-irradiation, and 8 ng/ml taxol. The percentages of dying or SA-β-gal+ cells in untreated cell populations were 3%

GSE56 transduction of HT1080 3′SS6 cells resulted in pronounced sensitization of these cells to cytotoxic drugs. This sensitizing effect is best demonstrated by an experiment where GSE56 was cloned into a retroviral vector LmXSE, which carries Green Fluorescent Protein (GFP) as a selectable marker (Kandel et al., 1997). HT1080 3′SS6 cells were infected with ecotropic retroviruses derived from an insert-free LmXSE vector or from Lm56SE vector that carries GSE56. As shown by FACS analysis of GFP fluorescence (Figure 4b), the transduction gave rise to mixed populations of infected cells, with 82% GFP positivity in LmXSE-infected and 71% in Lm56SE-infected cell populations (the lower intensity of GFP fluorescence observed with the Lm56SE vector reflects an apparently non-sequence-specific effect of inserts in LmXSE on GFP expression; Kandel et al., 1997). These infected populations were treated for 2 days with 30 nM doxorubicin and then transferred into drug-free media; the fraction of GFP+ cells was analysed on consequent days. In the control LmXSE-transduced cells, the size of the GFP+ fraction was not significantly altered after doxorubicin treatment and release from the drug (Figure 4b). In contrast, Lm56SE-transduced cells showed a striking decrease in the GFP+ fraction, which became nearly undetectable by day 5 after release, indicating a strong selection against GSE56 in doxorubicin-treated cells (Figure 4b). Similar selection against Lm56SE-transduced cells was also observed upon treatment with aphidicolin and taxol (data not shown). The sensitizing effect of GSE56 was also indicated by the analysis of L56SN-transduced cells, which showed a fivefold decrease in clonogenicity after treatment with 20 nM doxorubicin relative to the control.

As previously described (Pellegata et al., 1996; Chang et al., 1999), drug-treated or irradiated HT1080 cells, unlike HCT116, show little evidence of apoptosis. The most prominent morphological type of cell death in this line is mitotic catastrophe, evidenced by the formation of enlarged micronucleated cells with uneven distribution of chromosomes among the micronuclei (Chang et al., 1999). The increased cell death in the L56SN-transduced population of HT1080 3′SS6 cells treated for 3 days with 20 nM doxorubicin was associated with >twofold increase in the percentage of micronucleated cells relative to the control and the appearance of a detectable minority of cells with apoptotic morphology (Figures 4c and 5a,b). These results indicate that p53 inhibition sensitizes HT1080 cells to drug-induced mitotic cell death.

Figure 5
figure5

Microscopic analysis of cell death and SA-β-gal expression in HT1080 3′SS6 cells. Cells were treated with 20 nM doxorubicin for 3 days (a – d), untreated (e) or treated with 5 mM IPTG for 4 days (f), fixed and stained with X-gal for SA-β-gal activity. The photographs were taken at 400-fold magnification with phase contrast for (a) and (b), and at 200-fold magnification for (c – f) (20 μM scale bars (a and c)). (a) Cells with multiple micronuclei (mn) and apoptotic morphology (a) in doxorubicin-treated control HT1080 3′SS6 population. (b) Cells with multiple micronuclei (mn) and apoptotic morphology (a) in doxorubicin-treated L56SN-transduced HT1080 3′SS6 population. The inset shows bright-field image of the apoptotic cell. (c) SA-β-gal expression in doxorubicin-treated control HT1080 3′SS6 cells. (d) SA-β-gal expression in doxorubicin-treated L56SN-transduced HT1080 3′SS6 cells. (e) SA-β-gal expression in untreated p21-9 cells. (f) SA-β-gal expression in IPTG-treated p21-9 cells

As described in detail elsewhere (Chang et al., 1999), doxorubicin treatment of HT1080 3′SS6 cells induced SA-β-gal expression and morphologic changes associated with senescence; these SLP markers were associated with cells that underwent terminal proliferation arrest within a small number of cell divisions after release from the drug. SLP induction by doxorubicin was strongly (about threefold) decreased in the L56SN-transduced population relative to the control cells (Figures 4d and 5c,d). GSE56 also decreased the SA-β-gal+ fraction among the micronucleated and apoptotic cells that arise after doxorubicin treatment (Figure 4d), indicating that this effect of GSE56 is independent from the promotion of mitotic death. Similar or stronger effects were observed when SLP was induced by ionizing radiation, aphidicolin, cytarabine, cisplatin or taxol (Figure 3c), and in a cell population transduced with GSE56 in Lm56SE vector and isolated by FACS on the basis of GFP fluorescence (data not shown). Although diminished, SLP induction in doxorubicin-treated L56SN-transduced cells was still readily apparent (up to 20% SA-β-gal+, as compared to 1 – 3% among untreated cells).

Analysis of SLP induced by p21 overexpression from an inducible promoter indicates that p21 induction is insufficient to account for doxorubicin-induced SLP

p53-mediated upregulation of p21 expression in drug-treated cells provides the primary mechanism of damage-induced transient growth arrest (el-Deiry et al., 1993; Harper et al., 1993). The observed decrease in drug-induced SLP in p21−/− cells and in cells where p21 expression was inhibited by GSE56 suggested that p21 induction may also be responsible for senescence-like terminal growth arrest in drug-treated cells. To test this hypothesis, we have expressed p21 from an inducible promoter to determine whether p21 overexpression is sufficient to induce SLP. HT1080 3′SS6 cells carry the LacI repressor, which allows one to use physiologically neutral doses of β-galactosides (such as IPTG) for regulated expression of promoters coupled with lac operator sequences. A clonal cell line p21-9 was derived from 3′SS6 after transduction with retroviral vector LNXCO3 (Chang and Roninson, 1996) carrying the coding sequence of the human p21 gene under the IPTG-inducible promoter. As will be described elsewhere, the addition of IPTG to p21-9 cells induced stable growth arrest, which became largely irreversible after 3 or more days of exposure to IPTG (Chang, Broude, Watanabe, Fang, Abdryashitov and Roninson, manuscript in preparation). This growth arrest was associated with SLP, as indicated by SA-β-gal expression and senescent-like morphology observed in IPTG-treated p21-9 cells (Figure 5e,f) (IPTG had no effect on SA-β-gal expression in HT1080 3′SS6 cells transduced with insert-free LNXCO3 vector; data not shown). SLP induction by IPTG was dose-dependent and showed an excellent correlation with the levels of p21 induced by different concentrations of IPTG (Figure 6c). To determine if SLP induction by p21 required other p53-dependent functions, p21-9 cells were transduced with GSE56 in Lm56SE retroviral vector, and the infected cells were isolated by FACS on the basis of GFP fluorescence. Lm56SE-transduced cells were treated for 3 days with 50 μM IPTG or with 50 nM doxorubicin and stained for SA-β-gal. The percentage of SA-β-gal+ cells induced by IPTG was decreased only 1.5-fold in the GSE56-transduced population (Figure 6a); this modest decrease was associated with an apparently non-specific reduction in the IPTG-induced p21 level (data not shown). In contrast, GSE56 decreased doxorubicin-induced SA-β-gal expression in p21-9 cells fivefold (Figure 6a). These results indicate that overexpression of p21 is sufficient to induce SLP and irreversible growth arrest, and that this effect is largely independent of other p53-regulated functions.

Figure 6
figure6

Effects of p21 induction on SLP in HT1080 3′SS6 cells. (a) Induction of SA-β-gal expression by 3 day treatment with 50 μM IPTG (left) or 50 nM doxorubicin (right) in the control and Lm56SE-transduced p21-9 cells. (b) Western blot of p21 induction in HT1080 3′SS6 cells by 48 h treatment with different drugs and percentages of SA-β-gal+ cells arising under the same treatment conditions. (c) Dose-dependent induction of p21 protein and SA-β-gal in p21-9 cells by 48 h exposure to IPTG. p21 levels were determined by ELISA. (d) Dose-dependent induction of p21 protein and SA-β-gal in p21-9 cells by 48 h exposure to doxorubicin. (e) Comparison of the percentages of SA-β-gal+ cells corresponding to different doses of doxorubicin, as calculated for p21 values induced by the corresponding doses of doxorubicin (d) using the IPTG induction curves of (c) (solid bars) and as actually observed in (d) (hatched bars)

Since overexpression of the exogenous p21 gene from an inducible promoter induced SLP, we asked whether the induction of the endogenous p21 expression by cytotoxic drugs could be responsible for SLP in drug-treated cells. The levels of p21 induced by 2-day treatment of HT1080 3′SS6 cells with different drugs showed, however, no apparent correlation with the percentages of SA-β-gal+ cells arising under the same treatment conditions (Figure 6b). Since the time course of p21 induction may vary among different drugs, this result could only be viewed as preliminary. We have therefore compared the induction of p21 and SLP in p21-9 cells treated with different doses of IPTG or doxorubicin for 48 h, the period corresponding to the highest p21 levels in doxorubicin-treated cells. Both agents induced p21 and SA-β-gal in a dose-dependent manner (Figure 6c,d); immunocytochemical analysis (not shown) indicated that p21 expression was relatively homogeneous among cells treated with the same dose of IPTG or doxorubicin. The relationship between p21 levels and the percentages of SA-β-gal+ cells, however, differed greatly for the two inducers. Figure 6e shows a comparison of the actual percentages of SA-β-gal+ cells induced by three different concentrations of doxorubicin with calculated values that would have resulted from the induction of the same p21 levels by IPTG. The calculated p21-based values were four times lower than the actual percentages of SA-β-gal+ cells induced by the corresponding doses of doxorubicin, indicating that p21 expression in doxorubicin-treated cells could account for only a fraction of drug-induced SLP.

Discussion

We have previously shown that senescence-like terminal proliferation arrest is one of the major responses of human tumor cells to moderate doses of different anticancer agents (Chang et al., 1999). In the present study, we have investigated the role of p53 and p21 genes in this effect of chemotherapeutic drugs. We have found that the senescence-like response was strongly decreased but not abolished in HCT116 colon carcinoma cell lines where p53 or p21 were inactivated by homozygous knockout, and in HT1080 fibrosarcoma cells where p53 function and its associated p21 expression were repressed by retroviral transduction of a p53-inhibiting GSE. We have also found that p21 overexpression from an inducible promoter was sufficient to induce SLP, but p21 levels expressed in doxorubicin-treated cells could account for only a fraction of drug-induced SLP. Taken together, these results indicate that p53 and p21 act as positive regulators of senescence-like terminal proliferation arrest in drug-treated tumor cells, but their function is neither sufficient nor even absolutely required for this response to anticancer drugs.

The lower levels of drug-induced SLP and long-term growth arrest in p53−/− and p21−/− relative to wild-type HCT116 lines could be due, in theory, to clonal variability rather than the effect of the knockouts. This concern, however, is not applicable to the interpretation of the experiments where p53 function and p21 expression were inhibited in a mass population of HT1080 cells by transduction with a GSE56-carrying retrovirus, without further selection or subcloning. Since both the knockouts and GSE56 transduction had a similar inhibitory effect on the induction of SLP by chemotherapeutic drugs, we conclude that p53 and p21 act as positive regulators of this response. On the other hand, the observation that SLP was still inducible (albeit to a reduced extent) in HCT116 cells with homozygous knockout of p53 or p21 provides definitive evidence that these genes are not absolutely required for the senescence-like response. In agreement with this conclusion, we have previously found that all three (out of 14) tumor-derived cell lines that failed to induce SLP after 2-day treatment with moderate doses of doxorubicin were p53-mutant, but SLP was inducible in three other p53-mutant cell lines and in two lines where p53 function was attenuated by the E6 protein of papilloma virus (Chang et al., 1999). Our present finding that SLP can be induced even in p53−/− cells suggests that this response may contribute to treatment outcome in p53-negative human tumors.

It has been clearly established that p53 and p21 mediate G1 and G2 cell cycle arrest induced by DNA damage (el-Deiry et al., 1993; Waldman et al., 1995; Bunz et al., 1998). Several lines of evidence indicate, however, that drug-induced SLP is not a component of this protective growth arrest: (i) SA-β-gal expression and other markers of SLP develop in most HT1080 cells prior to terminal growth arrest (Chang et al., 1999), and p53 inhibition by GSE56 inhibits this early response to drug treatment; (ii) As shown by Waldman et al. (1996) and observed in the present study, almost all of the wild-type HCT116 cells undergo prolonged p21-mediated growth arrest until about 3 days after release from the drug. Many of these cells, however, eventually resume cell division, and such cells do not show the markers of SLP and (iii) SA-β-gal+ cells undergo cell death with a similar probability to SA-β-gal- cells, indicating that SLP induction does not correlate with the protective p21-mediated G1 or G2 arrest. Thus, the role of p53 and p21 in drug-induced SLP is not merely an extension of growth arrest regulated by these genes.

Decreased SLP induction in p53- or p21-inhibited cells also is not a consequence of an increase in drug-induced cell death. Regardless of the cellular p53 and p21 status, the percentage of SA-β-gal+ cells was always found to be the same among the dying (micronucleated and apoptotic) cells and in the total cell population, indicating that the induction of SLP and of cell death were independent events. Furthermore, the effects of p53 and p21 inhibition on SLP were more uniform in different cell lines than the effects on drug-induced cell death. Inhibition of wild-type p53 by GSE56 strongly increased drug-induced cell death in HT1080 cells, as evidenced by an increase in the percentage of cells containing multiple micronuclei, the appearance of an apoptotic cell fraction and selection against GSE56-containing cells. These effects of GSE56 in HT1080 cells agree with previously reported hypersensitivity of a p53-mutant HT1080 subline to radiation (Pellegata et al., 1996) and closely resemble the effects of p21 knockout in HCT116 line. An increase in drug-induced cell death in p21−/− relative to the wild-type HCT116 cells has been previously described (Waldman et al., 1996) and explained by the role of p21 in drug-induced protective G2 growth arrest (Bunz et al., 1998). It seems likely that the inhibition of p21 expression in GSE56-transduced HT1080 cells is responsible for increased sensitivity of these cells to drug-induced mitotic catastrophe. In contrast to GSE56-transduced HT1080 and p21−/− HCT116 cells, the p53−/− line of HCT116 showed only a small increase in cell death relative to wild-type HCT116. While the difference between the p21−/− and p53−/− lines could be due to clonal variability, it is conceivable that the relatively modest effect of p53 knockout on drug-induced cell death in HCT116 cells could reflect the balance between the p21-mediated protective effect of p53 and the known positive effect of p53 on drug-induced apoptosis (Lowe et al., 1993). In contrast to HCT116, the latter effect of p53 would not have a significant effect on the death of HT1080 cells that are largely refractory to apoptosis, thus explaining a major sensitizing effect of GSE56 in this cell line.

The ability of drugs to induce SLP even in p53−/− and p21−/− cells indicates that this response can also be mediated by some p53 and p21-independent pathways. The observation that p21 induction accounts for only a fraction of drug-induced SLP also indicates that some genes other than p53 or p21 play an essential role in this response. p16 was shown to be involved in accelerated senescence of normal cells (Serrano et al., 1997), but HT1080 and HCT116 cells are p16-deficient (Whitaker et al., 1995; Myohanen et al., 1998). The telomerase activity, which plays a key role in the escape of neoplastic cells from replicative senescence (Shay, 1997), appears to be undiminished in drug-treated HT1080 cells (Holt et al., 1997) and in cisplatin-treated nasopharyngeal carcinoma cells that show markers of SLP (Wang et al., 1998a). The nature of the genes mediating p53- and p21-dependent and independent components of drug-induced senescence-like growth arrest is currently under investigation.

The relative importance of different pathways of cell death and senescence-like growth arrest in tumor cell response to anticancer drugs may vary depending on the cell type, the nature of the drug and the conditions of drug exposure. While the cytotoxic responses (mitotic death and apoptosis) appear to be the primary determinants of survival in cells treated with high doses of conventional chemotherapeutic drugs, SLP induction is relatively more pronounced at moderate drug concentrations and may be an important determinant of treatment outcome in low-dose continuous infusion protocols. Furthermore, SLP appears to be the major correlate of permanent growth arrest in cells treated with cytostatic differentiating agents (Chang et al., 1999). Understanding the mechanism of senescence-like terminal proliferation arrest and manipulating this process may therefore help to improve the efficacy of cancer therapeutic regimens with a low systemic toxicity.

Materials and methods

Cell lines and retroviral transduction

HCT116 wild-type, p21−/− (clone 80S4) and p53−/− (clone 379.2) lines (Waldman et al., 1995; Bunz et al., 1998) were a gift of Dr B Vogelstein (Johns Hopkins University). HT1080 subline 3′SS6 was derived after transfection with the murine ecotropic retrovirus receptor and LacI repressor genes as previously described (Chang and Roninson, 1996). The p53 status of 3′SS6 cells was determined by amplifying exons 2 – 11 of p53 from genomic DNA, followed by automated DNA sequencing of both strands. All the cells were grown in DMEM with 10% FC2 serum (Hyclone).

Retroviral vector L56SN carrying GSE56 and the neo gene (Ossovskaya et al., 1996) was a gift of Dr AV Gudkov. The vector Lm56SE was constructed by ES Kandel in our laboratory by inserting GSE56 into retroviral vector LmXSE that carries Enhanced GFP as a selectable marker (Kandel et al., 1997). The LNp21CO3 vector was constructed by cloning the 492-bp coding sequence of the human p21 cDNA from pZL-WAF1 (el-Deiry et al., 1993) in the NotI and BglII sites of IPTG-inducible retroviral vector LNXCO3 (Chang and Roninson, 1996). Insert-free vectors LNCX (Miller and Rosman, 1989), LmXSE and LNXCO3 were used for control infections with L56SN, Lm56SE and LNp21CO3, respectively. Retroviral transduction of HT1080 3′SS6 cells was carried out using BOSC23 ecotropic packaging cell line (Pear et al., 1993) as previously described (Schott et al., 1996). Cell populations transduced with L56SN and LNCX were used without further selection, since plating of infected cells in the presence of G418 showed that the infection rates in these cases were close to 100%. LNp21CO3 and LNXCO3-transduced cell populations were selected with G418. Clonal line p21-9 was derived from the LNp21CO3-transduced population of HT1080 3′SS6 cells by end-point dilution. Lm56SE and LmXSE-transduced cell populations were isolated on the basis of GFP fluorescence using FACS (EPICS Elite-ESP, Coulter) as described (Kandel et al., 1997).

Drug and IPTG treatment and microscopic analyses

All the drugs were purchased from Sigma; IPTG was from Gibco – BRL. JL Shepherd Model 143 irradiator was used for γ-irradiation. Growth inhibition assays were carried out by plating 4-10×104 cells per 3.5 cm plate prior to drug exposure. For IPTG treatment, cells were plated at 2×104 per 3.5 cm plate prior to the addition of IPTG. At the end of IPTG treatment, cells were rinsed twice with PBS, incubated with regular media and refreshed with regular media 24 h later. Cell growth was measured by methylene blue staining as described (Perry et al., 1992). Colony formation assays were done by plating 2000 cells per 10 cm plate and allowing cells to form colonies for 8 – 10 days after release from the drug.

Cells on plates were fixed and stained for SA-β-gal activity using X-gal at pH 6.0 as described (Dimri et al., 1995); SA-β-gal+ cells were detected by bright-field microscopy. To detect micronuclei-containing and apoptotic cells, cells after SA-β-gal staining were counterstained with hematoxilin and eosin and analysed by phase-contrast microscopy. The percentages of cells in different categories were determined after scoring 100 – 400 cells for each sample.

FACS assays

For DNA content analysis, 3×105 cells per 6 cm plate were treated with different drugs. After treatment, attached cells were trypsinized and analysed by PI staining and FACS analysis using Becton Dickinson FACSort as described (Jordan et al., 1996). In some experiments, floating cells were collected by centrifugation, combined with attached cells and analysed by PI staining.

For PKH2 analysis of cell division, 107 cells were trypsinized and labeled with PKH2 (Sigma) according to the manufacturer's protocol. Cells were plated at 5×105 per 6 cm plate and PKH2 fluorescence was monitored on consecutive days by FACS analysis, after PI staining to exclude dead cells. As a control for the stability of PKH2 labeling, no changes in cell fluorescence were observed in p21-9 cells that were growth-arrested for 5 days by IPTG induction of p21. ModFit cell cycle analysis program (Verity Software) was used to determine the percentages of dividing and non-dividing cells.

Protein assays

ELISA measurement of p21 protein was carried out using WAF1 ELISA kit (Oncogene Research). Each sample contained approximately 103 cells; p21 levels were normalized for the protein content determined with Bio-Rad Protein Assay reagent. Western blotting was carried out using Western View (Transduction Laboratories), with anti-CIP1 antibody for p21WAF1/CIP1 (Transduction Laboratories) and MDM2 (Ab-1) antibody for Mdm2 (Oncogene Research). Immunocytochemical staining for p21 was carried out using WAF-1 (Ab-1) antibody (Oncogene Research) and LSAB 2 peroxidase kit (DAKO).

References

  1. Bunz F, Dutriaux A, Lengauer C, Waldman T, Zhou S, Brown JP, Sedivy JM, Kinzler KW and Volgelstein B. . 1998 Science 282: 1497–1501.

  2. Campisi J. . 1997 Eur. J. Cancer 33: 703–709.

  3. Chang B-D and Roninson IB. . 1996 Gene 183: 137–142.

  4. Chang B-D, Broude EV, Dokmanovic M, Zhu H, Ruth A, Xuan Y, Kandel ES, Lausch E, Christov K and Roninson IB. . 1999 Cancer Res. 59: 3761–3767.

  5. Chen Q and Ames BN. . 1994 Proc. Natl. Acad. Sci. USA 91: 4130–4134.

  6. Duncan EL and Reddel RR. . 1997 Biochemistry (Mosc) 62: 1263–1274.

  7. DiLeonardo A, Linke SP, Clarkin K and Wahl GM. . 1994 Genes Dev. 8: 2540–2551.

  8. Dimri GP, Xinhua L, Basile G, Acosta M, Scott G, Roskelley C, Medrano EE, Linskens M, Rubej I, Pereira-Smith O, Peacocke M and Campisi J. . 1995 Proc. Natl. Acad. Sci. USA 92: 9363–9367.

  9. el-Deiry WS, Tokino T, Velculescu VE, Levy DB, Parsons R, Trent JM, Lin D, Mercer WE, Kinzler KW and Vogelstein B. . 1993 Cell 75: 817–825.

  10. Harper JW, Adami G, Wei N, Keyomarsi K and Elledge S. . 1993 Cell 75: 805–816.

  11. Hayflick L and Moorhead PS. . 1961 Exp. Cell Res. 37: 585–621.

  12. Hendry JH and West CML. . 1997 Int. J. Radiat. Biol. 71: 709–719.

  13. Holt SE, Aisner DL, Shay JW and Wright WE. . 1997 Proc. Natl. Acad. Sci. USA 94: 10687–10692.

  14. Horan PK and Slezak SE. . 1989 Nature 340: 167–168.

  15. Jordan MA, Wendell K, Gardiner S, Derry WB, Copp H and Wilson L. . 1996 Cancer Res. 56: 816–825.

  16. Kandel ES, Chang BD, Schott B, Shtil AA, Gudkov AV and Roninson IB. . 1997 Somatic Cell Mol. Genet. 23: 325–340.

  17. Liu LF. . 1989 Annu. Rev. Biochem. 58: 351–375.

  18. Lock RB and Stribinskiene L. . 1996 Cancer Res. 56: 4006–4012.

  19. Lowe SW, Ruley HE, Jacks T and Housman DE. . 1993 Cell 74: 957–968.

  20. Miller DA and Rosman GJ. . 1989 Biotechniques 7: 980–986.

  21. Myohanen SK, Baylin SB and Herman JG. . 1998 Cancer Res. 58: 591–593.

  22. Ossovskaya VS, Mazo IA, Chernov MV, Chernova OB, Strezoska Z, Kondratov R, Stark GR, Chumakov PM and Gudkov AV. . 1996 Proc. Natl. Acad. Sci. USA 93: 10309–10314.

  23. Pear WS, Nolan GP, Scott ML and Baltimore D. . 1993 Proc. Natl. Acad. Sci. USA 90: 8392–8396.

  24. Pellegata NS, Antoniono RJ, Redpath JL and Stanbridge EJ. . 1996 Proc. Natl. Acad. Sci. USA 93: 15209–15214.

  25. Pereira-Smith OM and Smith JR. . 1988 Proc. Natl. Acad. Sci. USA 85: 6042–6046.

  26. Perry ME, Rolfe M, McIntyre P, Commane M and Stark GR. . 1992 Mutat. Res. 276: 189–197.

  27. Robles SJ and Adami GR. . 1998 Oncogene 16: 1113–1123.

  28. Schott B, Iraj ES and Roninson IB. . 1996 Somatic Cell Mol. Genet. 22: 292–309.

  29. Serrano M, Lin AW, McCurrach ME, Beach D and Lowe SW. . 1997 Cell 88: 593–602.

  30. Shay JW. . 1997 J. Cell Physiol. 173: 266–270.

  31. Sugrue MM, Shin DY, Lee SW and Aaronson SA. . 1997 Proc. Natl. Acad. Sci. USA 94: 9648–9653.

  32. Torres K and Horwitz SB. . 1998 Cancer Res. 58: 3620–3626.

  33. Uhrbom L, Nister M and Westermark B. . 1997 Oncogene 15: 505–514.

  34. Vogt M, Haggblom C, Yeargin J, Christiansen-Weber T and Haas M. . 1998 Cell. Growth Differ. 9: 139–146.

  35. Waldman T, Kinzler KW and Vogelstein B. . 1995 Cancer Res. 55: 5187–5190.

  36. Waldman T, Lengauer C, Kinzler KW and Vogelstein B. . 1996 Nature 381: 713–716.

  37. Waldman T, Zhang Y, Dillehay L, Yu J, Kinzler K, Vogelstein B and Williams J. . 1997 Nature Medicine 3: 1034–1036.

  38. Wang X, Wong SC, Pan J, Tsao SW, Fung KH, Kwong DL, Sham JS and Nicholls JM. . 1998a Cancer Res. 58: 5019–5022.

  39. Wang Y, Blandino G, Oren M and Givol D. . 1998b Oncogene 17: 1923–1930.

  40. Weinberg RA. . 1997 Cell 88: 573–575.

  41. Whitaker NJ, Bryan TM, Bonnefin P, Chang A C-M, Musgrove EA, Breithwaite AW and Reddel RR. . 1995 Oncogene 11: 971–976.

  42. Xu H-J, Zhou Y, Ji W, Perng G-S, Kruzelock R, Kong C-T, Bast RC, Mills GB, Li J and Hu S-X. . 1997 Oncogene 15: 2589–2596.

  43. Zhu J, Woods D, McMahon M and Bishop JM. . 1998 Genes Dev. 12: 2997–3007.

Download references

Acknowledgements

We thank Dr B Vogelstein for the wild-type, p21−/− and p53−/− HCT116 cell lines, Dr AV Gudkov for the L56SN vector, Dr ES Kandel for constructing the Lm56SE vector, and Dr K Hagen for assistance with flow sorting. This work was supported by grants R01CA62099 and R37CA40333 from the National Cancer Institute.

Author information

Correspondence to Igor B Roninson.

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • drug effects
  • p53
  • p21
  • senescence
  • cell death

Further reading