Dynamics of nuclear p53 is damage dose-dependent
To obtain data on p53 dynamics and cell fate as a function of DNA-damage strength, we performed dose titration of a DNA-damaging drug, etoposide, on U-2 OS cells expressing a fluorescent p53 reporter. The reporter cell line was generated by infecting U-2 OS cells with lentivirus encoding an established wild-type p53-Venus reporter construct [9]. For the study, we selected an isogenic clone that exhibited behaviors most similar to the parental line (refer to discussion below). We used etoposide as the DNA-damage stimulus, as it is known to trigger DNA double-strand breaks (DSBs) by inhibiting topoisomerase II [13, 14]. To select the appropriate range of etoposide dosage for activating variable amounts of DNA damage, we first measured the level of DNA damage under different etoposide doses. DNA damage, as indicated by the DNA-damage marker, phospho-H2A.X, was present after 12 hours of drug treatment (Figure 1a), and the degree of DNA damage increased with increasing dosage of etoposide from 1 μmol/l to 75 to 100 μmol/l.
To verify that dynamics of the p53-Venus reporter construct in our clonal reporter line mimicked the endogenous p53, we compared the dynamics of the endogenous and fluorescently tagged p53 by western blotting. The dynamics of the p53-Venus construct that we used have been extensively studied previously in MCF7 cells, which all showed that the reporter construct behaved similarly to the endogenous p53 [9, 10]. Our data from the U-2 OS clone also showed similar induction dynamics between p53-Venus and its endogenous counterpart (Figure 1b; note that the level of p53-Venus reporter is about half of the endogenous p53. For quantified results of western blots, see Additional file 1: Figure S1), when cells were treated with both low and high doses of etoposide. Therefore, the fluorescent p53 reporter is probably regulated similarly to the endogenous wild-type p53, and can be used to infer p53 dynamics in response to DNA damage.
From time-lapse microscopy movies, we quantified the real-time induction dynamics of p53 by measuring average Venus fluorescence in the nucleus and cytoplasm, respectively. In general, the fluorescent signal of p53-Venus in the nucleus was significantly higher than that in the cytoplasm. Single-cell p53 fluorescence trajectories showed two distinctive dynamic features: 1) nuclear p53 exhibited periodic pulsing at low drug dose, but monotonic increase at high dose; and 2) the amplitude of monotonically elevated p53 was substantially higher than that of the pulsing mode (Figure 1c; see Additional file 1: Figure S2). These bimodal, damage dose-dependent p53 dynamics are in stark contrast to the dose-independent oscillatory response of p53 to gamma irradiation previously identified in single mammalian cells [8]. The observation of a switch from oscillatory dynamics to a monotonic increase in p53 as DNA damage increased also suggests an intriguing new transitional regulation of p53.
The distinct mode of p53 dynamics strongly correlates with cell fate
Besides p53 dynamics, cumulative survival curves (quantified by analyzing individual cell fate from the time-lapse movies) showed that cell fate was also altered. As the dose of etoposide increased from 1 to 100 μmol/l, the cell-fate profile shifted from primarily cell-cycle arrest (scored by absence of cell division in 72 hours) to primarily cell death, with 100 μmol/l being the saturating dose for the cell-death response (Figure 2a). The dose response of the U-2 OS p53 reporter line was similar to that of the parental line, with the reporter line being slightly more sensitive to etoposide, probably due to the extra copies of p53 (see Additional file 1: Figure S3). Moreover, the single-cell data from the reporter line showed that the phenotypic outcome of arrest versus death correlated strongly with the dynamic mode of p53, as overall 86% of the cells (summed over all drug concentrations) that showed p53 pulsing went into cell-cycle arrest, and more than 96% of those with monotonic p53 induction died. Figure 2b shows the distributions of the four observed response phenotypes at different etoposide doses: 1) pulsing p53 followed by cell-cycle arrest, 2) pulsing p53 followed by cell death, 3) monotonic increase of p53 followed by cell-cycle arrest, and 4) monotonic increase of p53 followed by cell death. At all etoposide doses, phenotypes 1 and 4 accounted for more than 85% of the cells seen. In particular, at low DNA damage (etoposide ≤5 μmol/l), more than 65% of cells showed periodic pulsing and cell-cycle arrest, whereas at high DNA damage (etoposide ≥80 μmol/l) more than 95% cells died after monotonic elevation of p53.
To determine whether other mammalian cells exhibit a bimodal switch of p53 dynamics that correlates with cell fate as a function of DNA-damage strength, we generated another clonal fluorescent p53 reporter cell line in A549 cells. Dose titration results of the A549-p53 reporter line showed that A549 was more resistant than U-2 OS to etoposide, requiring higher concentrations to induce both cell-cycle arrest (5 μmol/l) and cell death (150 μmol/l) to near saturation level (see Additional file 1: Figure S4). In A549 cells, p53 again showed a bimodal switch of dynamics similar to that observed in U-2 OS cells, namely, pulsing at low DNA damage and monotonic increase at high DNA damage (Figure 2c). The strong correlation between cell fate and the two distinct p53 dynamic modes was also seen in A549 cells (Figure 2d). Overall, 97% of A549 cells that displayed p53 pulsing went into cell-cycle arrest, and more than 83% of those with monotonic p53 induction died. Our data thus suggest that the bimodal switch and its correlation with alternative cell fate are common in at least some mammalian cells in response to variable DNA damage.
We noted that although the correlation between cell fate and p53 induction dynamics was evident across the entire cell population, there was strong cell-to-cell variation in terms of the quantitative characteristics. We quantified the single-cell data and population average for the two p53 dynamic modes observed in U-2 OS cells at different etoposide concentrations (Figure 2e). For the pulsing mode, the pulsing period of the initial two to five pulses varied between individual cells from approximately 6 to approximately 40 hours, and the pulsing amplitude spanned a nearly eight-fold range. Furthermore, more than 50% of cells displayed irregular oscillatory behaviors with change in either pulsing period or amplitude over time. The strong single-cell variability was also evident for the monotonic induction mode, as shown by the large standard deviations in the population averages in terms of both the rate of p53 increase and the maximal p53 level (Figure 2e). In addition, our results appeared to show distinct dose dependence of the two p53 dynamic modes. Both the average pulsing period and the pulsing amplitude exhibited a moderate increase as the etoposide concentration increased, whereas the rate of increase and the maximal p53 level under the monotonic induction mode was largely dose-independent.
Mdm2 expression level controls the dynamic mode of p53 induction
p53 level is known to be negatively regulated by the E3 ubiquitin ligase Mdm2, while at the same time Mdm2 is transcriptionally activated by p53. Such a negative feedback loop between p53 and Mdm2 maintains p53 at a low level in unstressed cells. Upon induction of DNA damage, the inhibitory interaction between p53 and Mdm2 is disrupted by phosphorylation of both proteins, resulting in p53 accumulation [15]. To test whether the bimodal switch of p53 induction is a result of differences in Mdm2 activation, we performed immunoblotting to compare the dynamics of ensemble p53 and Mdm2 at 1 and 100 μmol/l etoposide, as under these two extreme concentrations, cells largely exhibited the same p53 dynamics of either pulsing or monotonic increase. The level of p53 increase was substantially higher at high drug dose than low dose for U-2 OS cells (Figure 3), agreeing with the live-cell imaging observations. Intriguingly, whereas expression of Mdm2 steadily increased at 1 μmol/l etoposide, its expression was largely suppressed at 100 μmol/l. Our data thus indicate that attenuation in upregulation of Mdm2 at high DNA damage may enable p53 to accumulate continuously to a high level, exhibiting monotonic elevation that is fundamentally different from the pulsing mode triggered by low damage.
To further confirm the role of Mdm2 in regulating the bimodal p53 dynamics, we attenuated Mdm2 activity in U-2 OS cells by using either RNA interference (RNAi) knockdown or nutlin-3, a small-molecule inhibitor of Mdm2 [16], and compared the resulting p53 dynamics and cell fate with those after treatment of 1 μmol/l etoposide alone (low DNA damage). Mdm2 knockdown efficiencies were generally greater than 90%, and Mdm2 knockdown alone (without etoposide) increased p53 to a level similar to that seen with 1 μmol/l etoposide alone (Figure 3b). Approximately 70% of U-2 OS cells under Mdm2 knockdown alone continued to divide, and less than 10% cell death occurred in the 72 hours of the imaging experiment. Figure 3c compares the distributions of the four different phenotypes, as discussed above, after treatment with 1 μmol/l etoposide plus one of the following four conditions: control (no siRNA), control siRNA (control for the pro-apoptotic effect of transfection), Mdm2 knockdown, and 10 μmol/l nutlin-3. We chose to use 10 μmol/l nutlin-3 because at this concentration nutlin-3 alone induced little cell death (<4% of U-2 OS cells died after 72 hours). Control siRNA transfection slightly sensitized cells to etoposide-induced cell death (approximately 5%), but did not alter distribution of the phenotypic responses, with about 90% of cells exhibiting pulsing p53 followed by cell-cycle arrest, similar to that of control cells. By contrast, Mdm2 knockdown not only significantly upregulated p53 level (Figure 3b) and changed the mode of p53 dynamics from being approximately 90% oscillatory to approximately 95% monotonic increase, but also shifted cell fate from approximately 90% cell-cycle arrest to approximately 80% cell death (Figure 3c). Similar changes in p53 dynamics and cell fate were also seen with etoposide plus nutlin-3. These data strongly suggest that the bimodal p53 dynamics are generated by differential Mdm2 upregulation modulated by DNA-damage strength, and that the bimodal switch functions as a crucial control over arrest versus death, as cell fate can be altered by changing p53 dynamics.
Attenuation of Mdm2 expression at high DNA damage could be due to reduced gene transcription or translation, or increased protein degradation. Real-time quantitative PCR (qPCR) results of Mdm2 showed that the gene-expression level of Mdm2 was largely similar under low and high etoposide concentrations (Figure 3d). Therefore, the distinct Mdm2 expression that we observed at the protein level in response to variable DNA-damage strength is not due to differential gene expression, but possibly due to differences in translation or Mdm2 protein degradation, such as by HAUSP [17] and ATM [18].
Bimodal switch of p53 dynamics modulates the overall transcriptional activity of p53
p53 is known to modulate cell fate through both its transcriptional activity in the nucleus, which upregulates downstream target genes, and its transcription-independent pro-apoptotic activity in the cytoplasm [4, 19]. To determine whether the bimodal switch of p53 dynamics exerts control over alternative cell fates by modulating the transcriptional activity of p53 in upregulating distinct pro-survival and pro-death genes, we measured, by real-time qPCR, the gene-expression dynamics of a set of well-characterized p53 target genes that are known to regulate cellular response to DNA damage [20]. These target genes include those involved in cell-cycle arrest (CDKN1A (p21) and GADD45A), DNA repair (XPC), senescence (PML and YPEL3) [21, 22] and apoptosis (APAF1, TP53AIP1, BAX, PUMA and NOXA). All target genes that we profiled, except BAX (largely not induced), showed much higher upregulation in response to high DNA damage (Figure 4a), when p53 was strongly elevated in the monotonic mode, irrespective of their particular functions. Of the three most highly induced genes under monotonic p53 in response to high DNA damage were two cell-cycle arrest genes, that is, CDKN1A (approximately 30-fold increase) and GADD45A (approximately 60-fold), and one pro-death gene, NOXA (approximately 35-fold).
To further confirm the similar induction dynamics of pro-survival and pro-death genes at the protein level, we compared, by western blotting, expression levels of p21 (CDKN1A), PUMA, and BAX from U-2 OS cells treated with 1 μmol/l (low damage) and 100 μmol/l (high damage) etoposide. Expression of both the pro-survival gene p21 and the pro-death gene PUMA were significantly induced under both low and high DNA-damage levels, whereas expression of another important pro-death gene, BAX, was not upregulated (Figure 4b, e). These largely agreed with the qPCR data. Hence, our results suggest that the bimodal switch of p53 dynamics modulates the transcriptional activation of pro-survival and pro-death genes similarly, and do not differentially activate these distinct genes in a damage dose-dependent manner. The observed bimodal switch did not appear to regulate cell fate through the previously reported mechanisms of differential transactivation of pro-survival and pro-apoptotic promoters [23], or through the p53-mediated transrepression of pro-survival genes [24]. Instead, our data point to a mechanism in which pulsing p53 retains a low level of p53 with target gene expression sufficient only for inhibiting cell-cycle progression, whereas the monotonic p53 results in strong induction of p53 that activates sufficiently high levels of pro-death genes to trigger cell death, superseding the concomitant cell-cycle arrest response.
Bimodal switch of p53 dynamics modulates the p53 cytoplasmic localization, and this correlates with the pro-apoptotic activity
The single-cell imaging data showed that although under low etoposide dosage the cytoplasmic p53 signal was close to that of the fluorescence background, there was an observable increase in cytoplasmic p53 fluorescence in response to high drug dosage, although the majority of the activated p53 was localized in the nucleus under both conditions. Shown in Figure 5a are representative single-cell trajectories of p53 fluorescence in the cytoplasm, relative to that in the nucleus from U-2 OS cells treated with 100 μmol/l etoposide. The average p53 fluorescence in the cytoplasm was found to be about 10% of that in the nucleus. Cell-to-cell variability was again evident, with the abundance of cytoplasmic p53 ranging from about 5% to 25% of that in the nucleus. The single-cell data indicated that the cytoplasmic accumulation of p53, which may be proportional to the cytoplasmic pro-apoptotic activity of p53, was modulated by the bimodal switch of p53 dynamics as a function of DNA-damage strength.
To determine the level of cytoplasmic p53 at the ensemble level, we performed subcellular fractionation. Using western blotting, we compared levels of nuclear and cytoplasmic p53 in U-2 OS cells treated with no drug, 1 μmol/l and 100 μmol/l etoposide, respectively (Figure 5b). Tubulin and transcriptional factor II B (TFIIB) were used as cytoplasmic and nuclear markers to confirm the quality of fractionation. In contrast to cells treated with 1 μmol/l etoposide, in which little cytoplasmic p53 was detected, treatment with 100 μmol/l etoposide induced substantial accumulation of cytoplasmic p53 (approximately 15% of that in the nucleus). The level of cytoplasmic p53 correlated with the differential cell fate, that is, more than 90% cell-cycle arrest at low damage, where cytoplasmic p53 was low, and more than 95% cell death at high damage, where cytoplasmic p53 was significant. These data thus suggest that in addition to the transcriptional activation of pro-death genes, bimodal switching of p53 dynamics may also regulate cell fate by modulating cytoplasmic accumulation of p53, which possibly correlates with its pro-apoptotic activity in the cytoplasm, as DNA damage increases.
Besides the induction level, activities of p53 are also known to be regulated extensively by post-translational modifications, such as phosphorylation and acetylation [25, 26]. Previous studies have identified several post-translational modifications on p53 that play crucial roles in regulating differential cell fate, such as Ser-20 phosphorylation by Chk1/Chk2 [27], Ser-46 phosphorylation by HIPK2 [28], Lys-120 acetylation by Tip60 [29], and Lys-382 acetylation by CBP/p300 [30]. To determine whether these modifications affect the cytoplasmic accumulation of p53, we measured, by western blotting, the above four post-translational modifications in the cytoplasmic and nuclear fractions of p53, respectively (Figure 5b). Whereas the extent of p53 phosphorylation at serine-20 and serine-46 was largely proportional to the expression level of p53, irrespective of the sub-celluar localization, the cytoplasmic fraction of p53 appeared to be preferentially modified by acetylation at lysine-120 and lysine-382, indicating that these two post-translational modifications may potentially have important roles in regulating p53 translocation to cytoplasm and/or its cytoplasmic pro-apoptotic activity.