Development of a fluorescence reporter system to quantify transcriptional activity of endogenous p53 in living cells

The tumor suppressor protein p53 functions as a hub protein for a variety of signaling pathways involved in cellular stress responses (Bieging and Attardi, 2012; Kastenhuber and Lowe, 2017). The TP53 gene, which encodes p53, is the most frequently mutated gene in many human malignant tumors, suggesting the potential importance of p53 in suppressing tumorigenesis (Olivier et al., 2010). p53 is known to act as a transcriptional regulator, and to be activated by a variety of post-translational modifications induced by various cellular stresses, including hypoxia, nutrient starvation and genotoxic stresses, such as DNA damage. It is well known that the tetramerization of p53 is essential for p53 activation (Kamada et al., 2016). Activated p53 recognizes and binds to a large number of p53 response elements (p53REs) in the genome, leading to the transcriptional regulation of multiple target genes. In this way, p53 inhibits tumorigenesis by inducing various cellular responses, such as apoptosis, cell cycle arrest and DNA repair (Appella and Anderson, 2001; Horn and Vousden, 2007; Meek, 2009). Previous reports have suggested that there are more than 300 target genes of p53 (Fischer, 2017).

In recent years, it has become clear that temporal dynamics plays an important role in intracellular signal transduction (Purvis and Lahav, 2013). In the p53 pathway, the temporal dynamics of p53 is known to play a pivotal role in the regulation of p53-regulated cellular functions. Temporal dynamics of p53 differs depending on the type of cellular stress, inducing distinct cellular responses (Chen et al., 2013; Purvis et al., 2012; Yang et al., 2018). It has also been shown that different doses of the same type of stress induce different temporal dynamics for p53 and consequently lead to different cellular phenotypes (Chen et al., 2013; Paek et al., 2016; Yang et al., 2018). Similar temporal regulation has also been observed in other signaling pathways, including the ERK, NF-κB and nuclear factor of activated T cells (NFAT) pathways (Lee et al., 2009; Marshall, 1995; Nelson et al., 2004; Yissachar et al., 2013). Recently, it has been proposed that cells encode input cellular information into the temporal dynamics of signaling molecules and then decode this information into downstream pathways (Purvis and Lahav, 2013). These mechanisms are thought to play a pivotal role in transmitting cellular information efficiently. However, it has been reported that different cellular responses are induced even when artificially induced temporal dynamics of p53 expression have the same shape as the natural temporal dynamics (Purvis et al., 2012). This suggests that the expression level of p53 does not necessarily reflect its transcriptional activation level (Loewer et al., 2010). In fact, previous studies have suggested that p53 has a very large number of post-translational modifications that regulate its stability, oligomerization status and binding affinity for DNA and transcriptional cofactors in a complex and diverse manner (Murray-Zmijewski et al., 2008). Therefore, to understand the decoding process in the downstream stages of the p53 signaling pathway, it is highly desirable to develop a system that can visualize the activation status of p53 instead of its expression level in single living cells.

In previous attempts to visualize the downstream of p53 signaling pathway in single living cells, transcriptional assay systems using p53RE and endogenous tags based on genome editing technology have been reported (Harton et al., 2019; Stewart-Ornstein and Lahav, 2016). A system using tagged target gene products is able to capture actual changes in protein expression. On the other hand, in terms of monitoring the decoding process of signaling factors, transcription reporters have the advantage of directly observing the transcriptional activity in a manner independent of mRNA stability, protein stability and the translation rate of the protein. Reporter systems that can evaluate p53 tetramer levels, an activated species of p53, have also been reported (Gaglia and Lahav, 2014; Gaglia et al., 2013).

In this study, we report a multiplexed reporter system that enables monitoring of the transcriptional activity of endogenous p53 at the single-cell level with high sensitivity. This reporter system is based on the p53RE-based transcriptional activity assay system previously reported in yeast-based functional assay (Kato et al., 2003; Shimada et al., 1999). Our reporter system is characterized by an internal standard for the transcriptional activity of p53, and by replacing the p53RE, the difference in the p53RE in each target gene can be easily and quantitatively evaluated. In addition, by tracking the fluorescence in the nucleus, it is possible to easily observe the temporal change of transcriptional activity for each single cell. Analysis using our reporter system revealed that the transcriptional activity of p53 was highly heterogeneous even among genetically identical cells. We also found that the transcriptional activation of p53 under etoposide treatment is cell cycle dependent and shows stronger cell cycle dependency than the p53 protein level. Finally, we demonstrate that the transcriptional activity of p53 can be visualized simultaneously with other cellular events when used in combination with other reporter systems.

We developed a novel reporter system to quantitatively analyze the transcriptional activity of endogenous p53 in individual living cells (Fig. 1A; Fig. S1A,B). This reporter system is based on the underlying mechanism by which p53 recognizes a specific p53 response element (p53RE) in the p53 target gene and regulates its transcription. This reporter system enables us to quantify the activation status of p53 by using the expression unit p53RE(C3) driving expression of Cerulean–NLS and the p53-independent constitutive expression unit of the simian virus 40 promoter (SV40) driving expression of mCherry–NLS, respectively. The reporter system can also monitor the transcriptional activity of p53 against the endogenous p53RE derived from the CDKN1A gene by the expression unit denoted (p53RE(CDKN1A))-Venus–NLS. Henceforth, we call these reporters (C3)-Cerulean, (SV40)-mCherry and (CDKN1A)-Venus, respectively. Given that a NLS is attached to the C-terminus of the fluorescent protein, the expression level of each reporter can be evaluated by quantifying the fluorescence intensity in the nucleus. These reporter genes were knocked into the AAVS1 site of human lung adenocarcinoma cell line A549 (p53 wild-type) by sequence-specific cleavage with CRISPR/Cas9 and site-specific recombination. The AAVS1 site was selected as a knock-in site because it is an open chromatin region and is less susceptible to repression of gene expression (Sadelain et al., 2011). Another advantage of using the AAVS1 site is that the copy number of the reporter gene is expected to be the same in all cells. These features of the AAVS1 site allowed us to quantitatively capture the transcriptional activity of p53. The fluorescent proteins selected for inclusion in the reporters are suitable for precisely monitoring transcriptional activity because of their high brightness and rapid chromophore formation (Nagai et al., 2002; Rizzo et al., 2004; Shaner et al., 2004). The fluorescence spectra of the three fluorescent proteins do not overlap in the observed wavelength range, so there are no effects from leakage into the other fluorescent channels.

We characterized the properties of the novel reporter system as follows. Etoposide was used as a genotoxic stressor because etoposide induces DNA double-strand breaks (DSBs) by inhibiting topoisomerase II, which activates p53-mediated DNA damage responses (Nitiss, 2009a; Yang et al., 2018). When the reporter cells were treated with etoposide, the fluorescence intensity of (C3)-Cerulean and (CDKN1A)-Venus, both of which represent p53-dependent transcriptional activity, significantly increased, whereas the fluorescence intensity of (SV40)-mCherry, which represents p53-independent transcriptional activity, did not change (Fig. 1B). The fluorescence intensity of (C3)-Cerulean and (CDKN1A)-Venus showed a large distribution after the etoposide treatment, whereas the (SV40)-mCherry intensity was almost unchanged (Fig. 1C; Fig. S1C). This result indicates that the transcriptional activation of p53 in response to genotoxic stress is heterogeneous even among genetically identical cells. Time-lapse imaging of the reporter cells showed that the fluorescence intensity of (C3)-Cerulean and (CDKN1A)-Venus increased from 6 h after stimulation, reaching a peak at 27–30 h, whereas the intensity of (SV40)-mCherry did not change at any time point (Fig. 1D). Analysis with the reporter system showed that the fluorescence intensity of (C3)-Cerulean and (CDKN1A)-Venus increased in a dose-dependent manner with etoposide concentration (Fig. S1D). We found that there were differences in dose dependence between (C3)-Cerulean and (CDKN1A)-Venus. These differences might result from transcriptional regulation of the CDKN1A gene by other factors or differences in affinity between p53 and the p53-binding site (Gartel and Radhakrishnan, 2005; Weinberg et al., 2005). The fluorescence intensity of (SV40)-mCherry did not change under all dose conditions tested. We also verified the p53-dependent response of the reporter system; a p53-knockout reporter cell line, which was confirmed by genomic PCR and western blotting (Fig. S1E,F and Fig. S7), showed only slight changes in the fluorescence intensity of (C3)-Cerulean and (CDKN1A)-Venus following etoposide treatment (Fig. S1G). Taking these results together, we concluded that this reporter system specifically monitors the transcriptional activity of p53. It is unclear why a slight transcriptional activation was observed in the p53-knockout cell line, but this might be due to the other p53-family proteins, p63 and p73; both of these proteins share considerable amino acid sequence identity with p53 and are known to control the expression of p53-regulated genes (Levrero et al., 2000).

A notable feature of our reporter system is that an artificial p53 response element, p53RE(C3) is used separately from the endogenous p53RE to serve as an internal standard for p53 transcriptional activity. Therefore, by replacing the endogenous p53RE with that of another target gene, differences in transcriptional regulation between target genes can be quantitatively analyzed by comparison with this internal standard. To demonstrate this, a total of six reporter cell lines (including CDKN1A) were established using p53RE regions derived from different target genes [CDKN1A, BAX, MDM2, GADD45A, RRM2B, 14-3-3σ (also known as YWHAE)]. After 24 h of etoposide treatment, these reporter cell lines showed no significant change in the distribution of (SV40)-mCherry fluorescence intensity, which exhibits a constitutive expression level (Fig. 2A). On the other hand, the distribution of (C3)-Cerulean fluorescence intensity, an internal standard for p53 activation status, increased in all cell lines, but the distribution did not show any significant differences (Fig. 2B). In contrast, the distribution of Venus fluorescence intensity driven by the p53RE for each gene [hereafter (each gene)-Venus], which reports on transcriptional activity for the endogenous p53RE, showed a large variation among the different cell lines, with particularly large heterogeneity in transcriptional activity against p53RE(CDKN1A) (Fig. 2C). Normalization of the transcriptional activity for each p53RE with the internal standard showed that some genes differed greatly in distribution whereas others did not (Fig. 2D). In particular, the CDKN1A gene exhibited strong transcriptional activation under our experimental condition, followed by the 14-3-3σ gene. The other four genes showed a similar distribution, but with a slight increase in activity for BAX and a slight decrease in activity for RRM2B. Scatter plots of the transcriptional activity of each p53RE relative to the internal standard showed differences in this relationship among the genes (Fig. S2). Specifically, for p53RE(CDKN1A) and p53RE(14-3-3σ), the degree of increase in transcription activity from the endogenous p53RE relative to the internal standard showed a linear relationship, whereas, for the other genes, transcription activity tended to be rapidly increased when the internal standard exceeded a certain level. These differences in activation properties among p53REs might be one means by which cells respond to cellular stress efficiently using specific hub factors.

We analyzed how the transcriptional activation pattern of p53 differs from cell to cell when the same stress is applied to a homogenous cell population. The transcriptional activity of p53 was quantified simultaneously with detection of cell death by time-lapse live-cell imaging. All subsequent experiments were performed using the reporter cell line with the p53RE region of CDKN1A because transcriptional activation for the endogenous p53RE was most heterogeneous in this cell line. The reporter cells were manually tracked, and cell death was determined by assessing cellular morphological changes (Fig. 3A and Fig. S3A; the images in Fig. 3A were used to make cell trace 3 in Fig. 3B). Interestingly, the pattern of p53 transcriptional activation was heterogeneous even within homogeneous cell populations (Fig. 3B). In addition, even between daughter cells derived from the same parental cell, the p53 activation pattern and the cellular response were often different (Fig. 3B). We next classified all cell traces into whether they came from surviving cells and dead cells, and found that p53 transcriptional activation occurred with a significant difference in cells undergoing cell death (Fig. 3C,D; data for (SV40)-mCherry are shown in Fig. S3B,C). Of note, some of the cell populations that did not undergo cell death also showed strong p53 transcriptional activation. We also confirmed that the p53-dependent transcriptional activity had little correlation with the constitutive expression level as measured by (SV40)-mCherry, indicating that the basal expression level is unlikely to be related to the transcriptional activity of p53 (Fig. S3D). These results strongly suggest that transcriptional activation of p53 is necessary for the induction of cell death. However, there was considerable overlap in the degree of transcriptional activation of p53 between the cells that underwent cell death and those that did not. This overlap again suggests a key role not only for the p53 signaling pathway but also for other signaling pathways involved in cell death by genotoxic stress, as previously reported (Roos et al., 2016).

We classified all cell traces according to cell division time and found that the transcriptional activity of p53 was remarkably reduced in cells that divided 0–3 h before stress addition (presumably early G1 phase) compared to cells that divided 9–12 h before stress addition (presumably late G1–S phase) (Fig. 3E; all data classified by cell division time are shown in Fig. S4A–C). In addition, the proportion of cells undergoing cell death decreased from 43% to 3% (Fig. 3F, gray area), respectively for these two groups. These results suggest that transcriptional activation of p53 upon etoposide treatment is dependent on the phase of the cell cycle, and thereby causes heterogeneous cellular responses. We also compared the fluorescence intensity before and after the stress addition and found no differences between the two groups before stress addition (Fig. 3G). This result indicates that the different patterns of transcriptional activation are not due to the initial activation state of p53. A previous report has shown that the major cytotoxic effects of etoposide are exerted mainly in the S/G2 phase, and our observation is consistent with this report (Hainsworth and Greco, 1995).

We also analyzed the p53 transcriptional activation patterns when cells were treated with UV-C (25 J/m²) irradiation and found that there was considerable cell-to-cell heterogeneity as we showed above for etoposide treatment (Fig. 4A). On the other hand, there was no clear correlation between the intensity of p53 transcriptional activation and cell death as in the case of etoposide treatment. Interestingly, comparison of the time courses of the mean values of surviving and dead cells revealed that p53 activation was stronger in surviving cells at an early stage after the stress addition (Fig. 4A, upper small panel; Fig. 4B). This result differs from that with etoposide treatment, indicating that the p53 transcriptional activation patterns differ depending on the type of stress.

We similarly analyzed the cell cycle dependency of p53 transcriptional activation patterns under UV irradiation conditions. Comparison of p53 transcriptional activation patterns of cells dividing 0–8.5 h before stress addition (presumably G1 phase) or 9–15 h before stress addition (presumably S phase) showed no significant differences until around 20 h after stress addition (Fig. 4C,D). Thereafter, p53 transcriptional activity in the presumably S phase cells plateaued, whereas in the presumably G1 phase cells, transcriptional activity continued to increase until ∼26 h after stress addition and then decreased. Interestingly, the proportion of cells undergoing cell death was 54% (15/28) for presumably G1 phase cells and 69% (25/36) for presumably S phase cells, indicating that the cell groups in which transcriptional activation of p53 occurred more strongly had less cell death. Although details require further study, these results might indicate the importance of the effect of early p53 transcriptional activation on cellular responses under UV stress and the contribution of a p53-independent cell death-inducing pathway.

We next examined the effect of the cell cycle on the transcriptional activation of p53 in more detail using the reporter cell line. The cells were synchronized in a specific phase of the cell cycle; by a double thymidine block (DTB) for G1/S phase synchronization and serum starvation (SS) for G1 phase synchronization. The cells were also synchronized at the G2/M phase by release 8 h after DTB (Fig. 5A). We confirmed that the cell cycle was successfully synchronized under both conditions (Fig. S5A,B). The reporter cell line was synchronized and then treated with etoposide (Fig. 5B). After 24 h of etoposide treatment, the fluorescence intensity of G1-synchronized cells decreased to 60% for (C3)-Cerulean and 50% for (CDKN1A)-Venus compared to that of asynchronous cells. In contrast, the fluorescence intensity of S phase-synchronized cells increased to 170% for (C3)-Cerulean and 220% for (CDKN1A)-Venus. G2-synchronized cells did not show a significant change in transcriptional activity of p53. The level of (SV40)-mCherry was slightly decreased in synchronized cells. Although the details are not clear, it has been reported that CDK1 regulates the balance of the overall cell proliferation rate and global protein synthesis rate (Haneke et al., 2020). Therefore, the decrease in (SV40)-mCherry could be the result of a change in the protein synthesis rate linked to the cell proliferation rate. We also confirmed the effect of cell cycle synchronization on the p53 signaling pathway. The SS treatment did not affect the transcriptional activity of p53 under our conditions (Fig. S5C). The DTB treatment induced slight activation of p53, but there were no differences in the fluorescence intensity of (C3)-Cerulean or (CDKN1A)-Venus at the start of drug treatment (Fig. S5D). Hence, we considered that this difference would not have a significant impact on our experiments.

Several reports have shown that the expression level of p53 does not reflect its transcriptional activity (Loewer et al., 2010; Loffreda et al., 2017). Therefore, we analyzed the expression level of p53 and the level of p53 phosphorylated at Ser15 (p53S15P) after etoposide treatment by immunofluorescence staining using normal A549 cells (Fig. 5C,D). The obtained images were quantified and the distribution of fluorescence intensity compared (Fig. 5E). The expression levels of both p53 and p53S15P were increased under all conditions, but there were some differences among the conditions. Interestingly, the distribution of the p53 expression level in the S/G2-synchronized cells was almost the same as that in asynchronous cells, but the p53 expression level decreased in the G1-synchronized cells. On the other hand, the levels of the p53S15P were substantially increased in S-synchronized cells, but decreased in G1/G2-synchronized cells. The shift in the distribution of p53S15P was more significant in G1-synchronized cells, and was only slight in G2-synchronized cells. The distribution of the (C3)-Cerulean and (CDKN1A)-Venus also showed a shift to a higher expression level in S-synchronized cells and a lower expression level in G1-synchronized cells. The distributions of the (C3)-Cerulean and (CDKN1A)-Venus in G2-synchronized cells indicated an intermediate expression level (Fig. 5E). These results once again indicate that the expression level of p53 does not necessarily reflect the activation state of p53. Furthermore, the coefficient of variation of p53 expression levels and p53S15P levels calculated from the data under cell cycle-unsynchronized conditions revealed that the variation in p53 expression levels did not change significantly before and after the stress addition, whereas the variation in p53S15P levels significantly increased (Fig. S5E). This result might suggest that the heterogeneity in phosphorylated p53 levels is partly responsible for the cell-to-cell heterogeneity in p53 transcriptional activation upon etoposide treatment. Our reporter system can directly quantify the transcriptional activity of p53 and could be a useful tool in studies involving the p53 signaling pathway.

Finally, we demonstrated the simultaneous imaging of p53-dependent transcription and cell cycle progression at the single-cell level. We developed a novel reporter stable cell line using the Fucci reporter system for visualization of the cell cycle (Fig. 6A; Fig. S6A). Owing to the limitation of available fluorescence channels, we used only the Cdt1 (30–120) fragment, which is stabilized in G1 phase and degraded in S/G2/M phase (Sakaue-Sawano et al., 2008). Hence, this reporter system monitors G1 phase-specific fluorescence through the Fucci-G1-Cerulean cassette, transcriptional activity against p53RE (CDKN1A) through Venus, and the constant expression level through mCherry. Time-lapse imaging of the reporter cells showed that the fluorescence intensity of Fucci-G1-Cerulean increases after cell division (Fig. 6B, white arrowhead; cell trace 4 in Fig. 6D shows the cells in Fig. 6B). The fluorescence intensity of (CDKN1A)-Venus increased after etoposide treatment and decreased after removing the drug. To automatically classify these cell traces, we determined a classification rule (Fig. 6C; see the Materials and Methods for detail). Using this rule, we classified all cell traces as early G1, late G1, S, and G2 by the cell cycle phase at the start of drug treatment (Fig. 6C). We refer to these groups of cell traces as early G1, late G1, S and G2 population, respectively. Applying this classification rule to the cell traces of this reporter system, we can simultaneously observe p53-dependent transcriptional activity, the cell cycle phase and cellular response (Fig. 6D). All cell traces were then classified according to the classification rule, and we found that the transcriptional activation pattern of p53 had different characteristics that varied in a cell cycle-dependent manner [Fig. 6E; data for (SV40)-mCherry are shown in Fig. S6D]. On average, the transcriptional activity of p53 in the G1 population was lower than that in S/G2 population. The proportion of cells undergoing cell death in the G1 population was also lower than that in S population. In late G1 population, some cells showed high transcriptional activity of p53 and cell death. This is probably because the late G1 population contains the boundary population of G1/S phase cells. These cells showed on average high p53 transcriptional activity as well as a high proportion of cell death. Interestingly, the proportion of cell death in S-population was remarkably different from that in G2-population even though these two populations showed similar p53-activation levels. Transcriptional activation of p53 in the G2 population occurred more rapidly than that in S population and stopped at 12 h after drug treatment. These differences in activation patterns might have affected the final cellular outcome. We note that the pattern of Fucci-G1-Cerulean after drug treatment differed between the G1 and S/G2 population. It is well known that etoposide induces mainly G2 arrest, and therefore, the pattern of Fucci-G1-Cerulean indicates that the G1 population enters S phase, and then all populations are arrested at the G2 phase after drug treatment. Moreover, we confirmed that after release from etoposide, cells failed to show a clear cell cycle re-entry over the next 12 h. Overall, our system proved to be useful for visualizing the transcriptional activity of p53 along with other biological phenomena.

In this study, we have developed a reporter system that allows the sensitive observation of p53 transcriptional activity in single living cells. Our reporter system is unique in that it uses p53RE(C3), which consists only of p53-binding sequences, as an internal standard for p53 transcriptional activity. By comparing the responses of endogenous p53REs from several target genes with the responses of p53RE(C3), it is possible to quantitatively analyze differences in transcriptional regulation among target genes. In fact, transcriptional activity for p53RE(C3) showed a similar distribution among all cell lines, whereas transcriptional activity for endogenous p53REs showed a different distribution among cell lines (Fig. 2B,C). This result suggests that the responsiveness of each p53RE is affected by other factors, such as p53 affinity, other regulators that bind to p53RE and post-translational modifications of p53. We also calculated the ratio of Venus to Cerulean fluorescence intensity per cell to more clearly demonstrate differences in the responsiveness of endogenous p53REs (Fig. 2D; Fig. S2). Our reporter system design, integrating internal standards, provides a useful framework for quantitative analysis of differences in responsiveness of multiple target genes regulated by hub transcription factors, not just p53.

We found that the transcriptional activation pattern of p53 under etoposide treatment differs among genetically identical cells (Fig. 3). It has been previously reported that the temporal dynamics of p53 in A549 cells under etoposide treatment showed either ‘periodic pulsing’ or ‘monotonic induction’, with the proportion of cells displaying ‘monotonic induction’ increasing in a dose-dependent manner (Chen et al., 2013; Yang et al., 2018). It has also been reported that the majority of ‘periodically pulsing’ cells exhibit cell cycle arrest, whereas the majority of ‘monotonic induction’ cells undergo cell death. In our study, although we did not observe the p53 expression and p53 transcriptional activity at the same time, the characteristics of the transcriptional activation patterns in the cells that underwent cell death suggest that cells that showed strong p53 transcriptional activation correspond with ‘monotonic induction’ cells and cells that did not show strong p53 transcriptional activation correspond with ‘periodic pulsing’ cells. Previous studies have shown that p53REs found in different p53 target genes have various affinities for p53 and thus different thresholds for p53 expression levels (Harton et al., 2019; Wu et al., 2017). These differences might result in the separate decoding of a single temporal dynamics of p53 into transcriptional dynamics for each target gene. In addition, it is generally known that the affinity of p53 for p53REs from genes associated with cell death is lower than that for p53REs of genes associated with cell cycle arrest (Weinberg et al., 2005). These properties might be a part of the mechanisms that amplify differences in the temporal dynamics of p53 among cells and confer robustness to the signaling pathway.

We also showed that the pattern of transcriptional activation of p53 following treatment with etoposide is highly dependent on phase of the cell cycle. It is well known that the cellular response to etoposide is cell cycle dependent, and that etoposide is most cytotoxic during the S/G2 phase (Hainsworth and Greco, 1995). Our results are consistent with these conclusions, as the transcriptional activity of p53 tended to be higher in S/G2 cells than in G1 cells (Figs 5B and 6E). The main reason for this cell cycle dependency might be the difference in the total amount of DNA damage. Topoisomerase IIα, the primary cellular target of etoposide, is mainly responsible for resolving the topological problems associated with replication and chromosome segregation in the S/G2 phase (Nitiss, 2009b). In addition, topoisomerase IIα is known to be upregulated during the S/G2 phase (Goswami et al., 1996; Hsiang et al., 1988). Therefore, the amount of DNA damage caused by the same dose is expected to be higher in S/G2 cells. In fact, several reports have shown that there is a significant increase of DSBs in S/G2 cells (Muslimović et al., 2009; Tammaro et al., 2013). The cell cycle dependency of p53S15P levels is probably closely related to these changes upstream of the p53 signaling pathway. On the other hand, transcriptional activation of p53 following UV irradiation showed a heterogenous response, similar to that of etoposide treatment even though it was largely independent of the cell cycle. Interestingly, there was also little correlation between transcriptional activity of p53 and cell death events. It has been reported that under UV irradiation, cross-inhibition of c-Jun N-terminal kinases 1 and 2 (JNK1/2; also known as MAPK8 and MAPK9, respectively) via p38 MAPKs generates cell-to-cell variability in JNK activity, resulting in stochastic cell death (Miura et al., 2018). It is also known that JNK has multiple phosphorylation sites on the transactivation domain of p53 and regulates p53 stability and transcriptional activity (Buschmann et al., 2001; Oleinik et al., 2007; Saha et al., 2012). Therefore, under UV irradiation, heterogeneity in upstream JNK activity could be one of the factors causing heterogeneity in transcriptional activity of p53 and might contribute to stochastic cell death. Although the detailed upstream mechanisms of the p53 signaling pathway under various stress conditions requires further investigation, it is likely that heterogeneity in the p53 signaling pathway occurs at different layers due to various factors, such as those mentioned so far.

In our experiments, we did not find marked cell cycle dependency in the expression level of p53 (Fig. 5C,E). On the other hand, the level of p53S15P showed a remarkable cell cycle dependency, which is consistent with the results observed for transcriptional activity (Fig. 5D,E). A possible mechanism to explain this observation is the contribution of direct transcriptional repression of p53 by MDM2. MDM2 functions as an E3 ubiquitin ligase and regulates p53 expression; MDM2 also directly represses transcription by concealing the transcriptional activation domain of p53 (Karni-Schmidt et al., 2016; Oliner et al., 1993). Therefore, the phosphorylation of Ser15 might lead to disassociation of p53 from MDM2 and increase the proportion of p53 that is free of MDM2, resulting in cell cycle-dependent transcriptional activation. MDM4, a homolog of MDM2 that does not have E3 ubiquitin ligase activity but has direct transcriptional repressive activity, might also be involved in this mechanism (Karni-Schmidt et al., 2016). Additionally, inhibition of p53 tetramer formation by an apoptosis repressor with a caspase recruitment domain (ARC; also known as NOL3) could also be involved in the cell cycle-dependent response. ARC is known to be repressed by transcription through p53 and degradation by MDM2 (Foo et al., 2007; Li et al., 2008). Therefore, the formation of p53 tetramers is inhibited in a cell cycle-dependent manner. These factors would allow cells to regulate the transcriptional activity of p53 in a cell cycle-dependent manner without any marked changes in p53 expression levels.

Although p53 was strongly transcriptionally activated in both the S and G2 phases compared to in the G1 phase, almost none of the cells in G2 phase underwent cell death (Fig. 6E). It has been reported that activation of p53 in the G2 phase is sufficient to induce cellular senescence (Johmura et al., 2014; Krenning et al., 2014). Therefore, most of the cells stressed in the G2 phase are thought to undergo cellular senescence. It is known that p21 (also known as CDKN1A) plays an important role in the induction of cellular senescence. p21 is rapidly degraded in S-phase cells, which might induce a different response in the G2 phase by predominantly inducing the expression of cell death-related gene products (Bornstein et al., 2003). These findings can only be obtained by comparing global regulation at the transcriptional level with the expression of individual target gene products.

In this study, we developed an experimental system that enables us to monitor the activation status of p53 in single living cells. We have also demonstrated that the cell cycle is strongly involved in the heterogeneous response to genotoxic stress in genetically identical cells by single-cell tracking using this system. Recently, it has become clear that crosstalk between the p53 signaling pathway and other signaling pathways involved in the cell cycle and cell proliferation plays a very important role (Hanson and Batchelor, 2022; Yang et al., 2017). Therefore, in the future, simultaneous visualization of the activation states of p53 and these factors in single cells is expected to provide a detailed understanding of how these signaling pathways contribute to cell cycle progression and the regulation of various cellular responses. In conclusion, this reporter system might be an important tool to provide new insights into the transduction mechanisms regulated by the p53 signaling pathway.

Two reporter plasmids, pAAVS1-(p53RE(C3))-Cerulean-NLS-(SV40)-mCherry-NLS-2A-Puromycin^R and pAAVS1-(p53RE(CDKN1A))-Venus-NLS-Neomycin^R, were constructed as shown in Fig. S1A,B. The former plasmid carried the artificial p53RE(C3), which consists of three tandem repeats of the p53-binding motif (5′-AGACATGTCCGGACATGTCT-3′) with a 16 bp spacer (5′-ACTAGCGGCTGTCACT-3′). Using this plasmid, we can estimate the activation status of p53 from the expression level of Cerulean-NLS [comprising cyan fluorescent protein with three tandem repeats of the nuclear localization signal (PKKKRKV×3) from the simian virus large T-antigen] in the nucleus. This plasmid also contains the expression unit (SV40)-mCherry-NLS-2A-Puromycin^R. The 2A peptide sequence (GSGEGRGSLLTCGDVEENPG/P) derived from the Thosea asigna virus interrupts the normal peptide bond and leads to ribosomal skipping (Szymczak et al., 2004). Thus, the red fluorescent protein (mCherry-NLS) and puromycin-resistance gene product can be separately expressed in a p53-independent manner. The latter plasmid carries the p53RE derived from the CDKN1A gene, which was originally reported in Shimada et al. (1999) [GenBank accession no. U24170, nucleotides 2241–3258]. Using this plasmid, a yellow fluorescent protein with an NLS sequence (Venus-NLS) is p53-dependently expressed in the nucleus. Both plasmids contain the left and right homology arm to the AAVS1 genomic integration site derived from the AAVS1 SA-2A-puro-pA donor plasmid (Addgene number #22075) (Kotin et al., 1992). The mRNA degradation signal, an AU-rich element (5′-ATTTATTTATTTATTTATTTA-3′), is inserted in the 3′-noncoding region of Venus-NLS and Cerulean-NLS to monitor the quick response in p53-dependent transcriptional activity (Chen and Shyu, 1995).

Five reporter plasmids were also constructed by replacing p53RE(CDKN1A) of pAAVS1-(p53RE(CDKN1A))-Venus-NLS-Neomycin^R to p53RE(each gene); each gene is BAX, MDM2, GADD45A, RRM2B or 14-3-3σ. The following p53RE regions were used for each gene, based on the transcription start point: BAX; −896 to −332, MDM2; 594 to 801, GADD45A; 1388 to 1721, RRM2B; 1953 to 2584, 14-3-3σ; −2553 to −1768 (Kato et al., 2003; Shimada et al., 1999).

Using pAAVS1-(p53RE(CDKN1A))-Venus-NLS-Neomycin^R, we also constructed a cell cycle reporter plasmid, pAAVS1-(p53RE(CDKN1A))-Venus-NLS-(SV40)-mCherry-NLS-2A-Puromycin^R-2A-Fucci-G1-Cerulean (Fig. S6A). This plasmid can separately express the mCherry-NLS, the puromycin-resistance gene product and Fucci-G1-Cerulean, which is Cerulean fused to a portion of human Cdt1 from pFucci-G1 Orange (Amalgaam, MBL), which enable us to distinguish the G1/S boundary by quantification of the increase or decrease of fluorescence just in the nucleus (Sakaue-Sawano et al., 2008).

The plasmid pBS-p53Nn-Zeo^R-p53Cc for making the p53-knockout cell line carried a 974 bp DNA fragment from the 5′-noncoding region of the human TP53 gene (nucleotide number −971 to 4) and the 869 bp DNA fragment of the 3′-noncoding region in TP53 starting from the stop codon. Using this plasmid, we can replace the p53 coding region with the Zeocin-resistance gene from pcDNA3.1/Zeo (Invitrogen).

The guide RNA expression plasmids for the CRISPR/Cas9 system were constructed using pX330 from Addgene (#42230). pX330-AAVS1-1, -p53N, or -p53C carried the DNA fragments 5′-GTCCCCTCCACCCCACAGTG-3′, 5′-GCGGGTCACTGCCATGGAGG-3′ or 5′-GGAGAATGTCAGTCTGA GTC-3′ between the two BbsI sites in pX330, respectively.

A549 cells (p53 wild type) were obtained from American Type Tissue Culture Collection (ATCC; Rockville, MD, USA). Cells were cultured on a 35 mm dish in Dulbecco's modified Eagle's medium (DMEM), high glucose (Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS; Japan Bioserum), 100 units/ml penicillin and 100 µg/ml streptomycin (Thermo Fisher Scientific). For stress addition, cells were treated with 10 µg/ml etoposide (Wako) or irradiated with 25 J/m² UV-C (254 nm) with a FUNA UV Crosslinker, FS-800 (Funakoshi). All cells were maintained in a humidified atmosphere of 5% CO₂ at 37°C.

For the establishment of the reporter cell line, A549 cells were plated on a 35 mm dish for 24 h and transfected with 0.2 µg of pX330-AAVS1-1 vector and 0.4 µg knock-in donors by using Lipofectamine 2000 (Invitrogen). At 24 h after transfection, the cells were seeded onto a 35 mm dish and treated with 0.5 µg/ml puromycin (Invivogen) and 1 mg/ml G418 (Roche) for more than 14 days. The culture medium was changed every 3–5 days. The selected cell lines were subcloned in a 96-well plate. For the establishment of the p53-knockout cell line, the A549 cells with both [AAVS1](p53RE(C3))-Cerulean-NLS-(SV40)-mCherry-NLS and (p53RE(CDKN1A))-Venus-NLS were plated on a 35 mm dish for 24 h and transfected with 0.6 µg of pBS-p53Nn-Zeo^R-p53Cc, 0.5 µg of pX330-p53N and 0.5 µg of pX330-p53C using 4 µl of Lipofectamine 2000, selected with 100 µg/ml Zeocin (Invivogen) and subcloned with a 96-well plate.

A double thymidine block (DTB) was used for S phase synchronization and serum starvation (SS) for G1 phase synchronization (Bostock et al., 1971; Johmura et al., 2014). For the DTB, A549 reporter cells were treated with 2 mM thymidine for 12 h, released for 12 h, and then treated with 2 mM thymidine for 12 h again before stress addition. For SS, A549 cells were treated with DMEM without FBS for 48 h before stress addition.

Live-cell imaging was performed with a BIOREVO BZ-X710 (Keyence) machine for the data in Figs. 3, 4 and a BIOREVO BZ-9000 (Keyence) machine for other data. The following filter sets were used for observation: for mCherry, the Texas Red filter (excitation filter: 560/40 nm; emission filter: 630/60 nm; dichroic mirror: 595 nm; for BIOREVO BZ-X710) or the TRITC filter (excitation filter: 540/25 nm; emission filter: 605/55 nm; dichroic mirror: 565 nm, for BIOREVO BZ-9000); for Venus, the YFP filter (excitation filter: 500/24 nm; emission filter: 542/27 nm; dichroic mirror: 520 nm); and for Cerulean, the CFP filter (excitation filter: 448/20 nm; emission filter: 482/25 nm; dichroic mirror: 458 nm). All live-cell imaging was performed with a Plan Fluor D 10×/NA030 objective lens (Nikon).

The transcriptional activity from different promoters was monitored by three fluorescent proteins fused with NLS, so that the transcriptional activity was detected as the fluorescence intensity present in the nucleus. Quantitative analyses of fluorescence intensity were performed as previously described (Imagawa et al., 2009). The detailed procedure used for detection and quantification of fluorescent signals with image analysis was as follows: (1) the original image is smoothed by applying a Gaussian filter, and then the background signal is subtracted; (2) based on the fluorescence image of mCherry, areas with values above a set threshold are detected as nuclei; and (3) the fluorescence intensity of each reporter is calculated based on the detected nuclear area. Single cells were tracked manually via the phase and fluorescence images. Cell death was determined manually from the changes of cell morphology seen in phase images (Fig. S3A). Experimental data under UV irradiation conditions only (Fig. 4) were analyzed using Fiji and LIMTracker (Aragaki et al., 2022; Schindelin et al., 2012).

A549 cells were plated on a 35 mm dish and their cell cycles were synchronized (the detailed schedules are shown in Fig. 5A), and then etoposide stimulation was performed for 24 h. Cells were fixed with 10% neutral buffered formalin solution (Wako) for 15 min, washed in phosphate-buffered saline (PBS), permeabilized with 0.2% Triton X-100 in PBS for 15 min, and blocked with 10% FBS in PBS for 1 h. After that, cells were incubated with the primary antibody in 1% FBS in PBS for 16 h at 4°C, and incubated with the secondary antibody and DAPI (Wako) in 1% FBS in PBS for 30 min at room temperature. The primary antibodies used were mouse monoclonal anti-p53 (DO-1; 1:500, Santa Cruz Biotechnology, SC-126) for determination of the p53 expression level, and mouse monoclonal anti-p53(Ser15) (16G8; 1:250, Cell Signaling, 9286) for determination of the p53 Ser15 phosphorylation level. The secondary antibody was Alexa-Fluor-488 anti-mouse-IgG (1:1000, Invitrogen, A-11001). Cells were imaged on a BIOREVO BZ-9000 fluorescent microscope (Keyence). The following filter sets were used for observation: for Alexa Fluor 488, the GFP filter (excitation filter: 470/40 nm; emission filter: 535/50 nm; dichroic mirror: 495 nm); and for DAPI, the DAPI filter (excitation filter: 360/40 nm; emission filter: 460/50 nm; dichroic mirror: 400 nm). Nuclei were detected by using DAPI images and the relative expression levels of proteins in the nuclear region were quantified.

Cell traces were classified automatically by the cell cycle phase at the start of drug treatment based on the expression pattern of Fucci-G1-Cerulean and the time of cell division (Fig. 6C). For cells classified as G1 phase, those that were treated within the first 6 h after cell division and those that were treated after the first 6 h following cell division were further classified as early G1 and late G1 cells, respectively. For cells classified as S/G2 phase, those that were treated within the first 6 h after the peak and those that were treated after the first 6 h following the peak were further classified as S and G2 cells, respectively. The time window of each cell cycle phase was decided by the measured parameters of cell cycle length and cell cycle population of A549 cells [Fig. S6B,C; cell cycle length: 21.9±0.2 h; cell cycle population: G1 (51.4±2.4%), S (35.2±1.7%), G2 (12.3±0.3%); mean±s.d.]. Specifically, we calculated the length of the S phase and obtained a value of 7.4 h, and therefore we separated cells that were treated first 6 h after the peak or not were classified S or G2 cells, respectively.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260918.reviewer-comments.pdf.