Cdk activity drives senescence from G2 phase

In response to DNA damage a cell can be forced to permanently exit the cell cycle and become senescent. Senescence provides an early barrier against tumor development by preventing proliferation of cells with damaged DNA. By studying single cells, we show that Cdk activity is retained after DNA damage until terminal cell cycle exit. The low level of Cdk activity not only allows cell cycle progression, but also forces cell cycle exit at a decision point in G2 phase. We find that Cdk activity stimulates p21 production, leading to nuclear sequestration of Cyclin B1, subsequent APC/CCdh1-dependent degradation of mitotic inducers and induction of senescence. We suggest that the same activity that triggers mitosis in an unperturbed cell cycle drives senescence in the presence of DNA damage, ensuring a robust response when most needed.


Introduction
In response to DNA damage, the cell cycle is halted to allow DNA repair. This is particularly critical in G2 phase, as entry into mitosis with unrepaired DNA may result in chromosomal aberration and propagation of mutations. Several mechanisms that establish a G2/M arrest by counteracting mitosis-promoting factors have been described [1][2][3]. However, upon DNA damage in S or G2 phase the production of mitosis-inducing factors such as Cyclin A2, Cyclin B1, Aurora A, Aurora B and Plk1 initially continues, albeit at a reduced pace [4][5][6]. As multiple feedback loops ensure a spiraling activation of Cyclin B1-Cdk1, Cyclin A2-Cdk1/2 and Plk1 that ultimately results in mitotic entry [7], maintaining even low levels of mitosis-inducing factors poses the risk to eventually overrun a cell cycle arrest. In fact, suppression of Cdk activity merely by posttranslational modifications is insufficient to sustain a G2/M arrest [8][9][10].
To avoid override of a cell cycle arrest, cells have evolved mechanisms that force terminal cell cycle exit and senescence. We and others have shown that terminal cell cycle exit from G2 phase depends on p53, its transcriptional target p21, and activation of the ubiquitin ligase APC/C Cdh1 that targets a large number of cell cycle regulators for degradation [5,6,[11][12][13]. During this process, cells lose the expression of G2specific proteins, exit the cell cycle, and become senescent, thereby preventing propagation of mutations [11,[13][14][15][16]. What determines whether and when a cell in G2 phase becomes senescent remains unclear.

! %!
There are several indications that induction of senescence is a regulated process.
Terminal cell cycle exit is a sharp transition, whose point-of-no-return is marked by the translocation of Cyclin B1 from the cytoplasm to the nucleus [6,11,17]. Before terminal cell cycle exit is initiated in G2 phase there is a variable delay, whose duration depends on when within S or G2 phase the damage occurred. That is, a cell receiving damage in late G2 exits the cell cycle faster than a cell receiving damage in early G2 phase [6]. These observations suggest that the signaling pathways that mediate senescence and cell cycle progression are interlinked.
Here we show that, despite a severe suppression, Cdk activity is sustained during a cell cycle arrest until terminal cell cycle exit occurs and that this remaining Cdk activity is needed to promote timely induction of senescence. We suggest that the key activity that drives mitosis in the absence of DNA damage drives cell cycle exit in the presence of DNA damage.

Concerted Cdk1 and Cdk2 activity regulates Cyclin B1 nuclear translocation and induction of senescence upon DNA damage
We previously employed live-cell microscopy of individual RPE cells encoding a Cyclin B1-eYFP fusion protein at the endogenous CCNB1 locus to study terminal cell cycle exit. Using this system we observed that DNA damage-dependent nuclear translocation and degradation of Cyclin B1 occurred only after S-phase completion [6]. As this is the time when Plk1 and Cdk1 are activated in an unperturbed cell cycle [18], we hypothesized that cell cycle kinases might be involved in regulating cell ! &! cycle exit after DNA damage. To test this idea we monitored the effect of different cell cycle kinase inhibitors on Cyclin B1-eYFP levels and localization after DNA damage using time-lapse microscopy.
To study cell cycle-dependent effects without affecting initiation of the DNA damage response (DDR), we added small molecule inhibitors of cell cycle kinases 1h after addition of the topoisomerase inhibitor Etoposide. Addition of a potent inhibitor of Plk1 did not affect Cyclin B1-eYFP degradation or nuclear translocation upon DNA damage, indicating that Plk1 is not required for these events (Supplementary Fig 1A).  Supplementary Fig 1B and D). Similarly, stimulation of cellular Cdk activity by inhibition of Wee1 led to earlier degradation along with higher rates of DNA damage checkpoint slippage (Supplementary Fig 1C). Taken together, our results suggest that Cdk1/2 activity is needed for timely Cyclin B1 translocation and degradation after DNA damage.
DNA-damage induced nuclear translocation of Cyclin B1 in G2 phase marks a decision point for terminal cell cycle exit and senescence [6,11]. Since inhibition of Cdk1/2 activity substantially delayed nuclear translocation of Cyclin B1, we next sought to test if Cdk activity affects whether cells become senescent. To this end, we assessed senescence-associated markers while perturbing Cdk activity using kinase inhibitors or Cdk RNAi. We quantified the occurrence of !-Galactosidase staining ! '! (Fig 2A and Supplementary Fig 2A), total and foci-associated staining of H3K9Me2 and HP1b as markers of senescence-associated heterochromatin foci (SAHF) [20], as well as expression of IL-6 as marker for the senescence-associated secretory phenotype (SASP) [21,22] (Fig 2B and Supplementary Fig 2B-D). While long-term treatment with RO-3306 and NU6140 was cytotoxic to cells, we found that all these markers were reduced upon combined Cdk1/2 RNAi or addition of Roscovitine, indicating that Cdk activity stimulates senescence. Similarly, the proliferation capacity, measured by total cell numbers and clonogenic growth, increased after temporal Cdk inhibition (Fig 2C, D and Supplementary Fig 2E). Increased Cdk activity may however not necessarily lead to increased senescence, as we find no evidence that Wee1 inhibition changed senescence-associated markers or proliferation capacity during constant exposure to Etoposide for 4 or 5 days (Fig 2A-D and Supplementary Fig 2B-E). Thus, our data suggest that Cdk activity after DNA damage stimulates induction of senescence, but also that cells contain sufficient Cdk activity to promote senescence during prolonged exposure to a DNA-damaging compound.

cycle-dependent manner
Although a DNA damage-mediated checkpoint largely functions by blocking Cdk activity [1,23], our data indicate that Cdk activity is integrated in the DDR. To resolve this apparent paradox, we sought to assess whether Cdk activity persisted after DNA damage. To this end we immunoprecipitated Cyclin A2-eYFP or Cyclin B1-eYFP from gene-targeted RPE cells [6,18] and performed kinase assays on recombinant target proteins that can be phosphorylated by either Cdk2 or Cdk1. Cyclin A2-! (! associated Cdk activity is active through a large part of interphase [24] and was readily detected in a population of unsynchronized cells. Although significantly reduced, Cyclin A2-eYFP associated activity persisted 4h after addition of either Etoposide or the radiomimetic drug Neocarzinostatin (NCS) (Fig 3A). In contrast, Cyclin B1-associated Cdk activity is initially activated at the S/G2 border, and slowly builds up through G2 phase until a dramatic increase initiates mitosis [18,25]. Indeed, we detect a strong Cyclin B1-eYFP associated kinase activity in RPE cells at 10h after release from a thymidine block. However, a markedly reduced Cyclin B1-eYFP associated kinase activity was still present even after 4h Etoposide or NCS treatment, when no mitotic cells were visually detected ( Fig 3B). This indicates that although low compared to the activities that initiate mitosis, both Cyclin A and Cyclin Bassociated activities are present after DNA damage. Similarly, immunoprecipitated Cdk2 from both unsynchronized and G2 synchronized RPE cells showed reduced but persistent activity 4h after Etoposide treatment ( Fig 3C). Thus, Cdk activity can be sustained at a low level after DNA damage in RPE cells.
We next sought to assess phosphorylation of endogenous Cdk targets in damaged and unperturbed RPE and U2OS cells. To detect ongoing Cdk-mediated phosphorylation we added Cdk inhibitors during the last hour of a 4h Etoposide treatment. For both cell lines, we detect Cdk-dependent phosphorylations after Etoposide treatment in whole cell populations (Fig 4A and Supplementary Fig 4A-D) as well as in single G2 cells (Fig 4B and Supplementary Fig 4E). Cdk target phosphorylation is still detectable at the time of Cyclin B1 nuclear translocation, but not after 24h of DNA damage, suggesting that Cdk activity is preserved until terminal cell cycle exit ( Fig   4C and Supplementary Fig 4F). Notably, Cdk target phosphorylation correlated ! )! positively with the levels of the DNA damage marker "H2AX, thus excluding the possibility that only mildly damaged cells retain Cdk activity (Supplementary Fig   4G). Furthermore, "H2AX levels were not affected by RO/NU treatment showing that Cdk inhibition does not result in an overall reduction of DNA damage signaling ( Supplementary Fig 4G). To assess the cell cycle distribution of Cdk activity during an ongoing DDR we performed quantitative immunofluorescence for Cdk-dependent Lamin A/C phosphorylation and sorted the cells according to their relative position in the cell cycle [6,18,26]. To control for cell cycle-dependent differences in background signals and target site-specificity we added Cdk inhibitors 1h before fixation. In accordance with recent data on Cdk2 [27], we detected initial Cdk1/2 target phosphorylation already during G1, from where it slowly rose throughout S phase before it rapidly increased at the S/G2 border ( Fig 4D,  16h and more ( Fig 4E, red tracks). This result is in line with our previous observation that S-phase cells display slower Cyclin B1 accumulation upon DNA damage [6].
Taken together, our data show that Cdk activity is maintained after DNA damage and that the duration and extent of sustained Cdk activity depends on the cell cycle position when DNA damage occurred.

Cdk activity during DNA damage promotes p21 production
Cell cycle exit and senescence from G2 phase after DNA damage depends on p53 and its transcriptional target p21 [5,6,28,29]. We therefore sought to investigate whether Cdk activity could enhance p53 and p21 expression. In contrast to this hypothesis, we found p53 expression to be elevated when Cdk1 and Cdk2 were inhibited or knocked down, suggesting that Cdk activity does not enhance p53 levels ( Fig 5A, B and Supplementary Fig 5A). Cdk inhibition also led to a slight increase in ATM-target phosphorylation, which might contribute to the observed elevated p53 levels To further investigate Cdk-dependent p21 induction, we next focused on the determinants of p21 expression. Although p21 was induced by Etoposide addition, ! "+! p21 levels in single cells correlated poorly with the levels of "H2AX staining, indicating that p21 expression is not solely regulated by the amount of damaged DNA present in a cell ( Supplementary Fig 5C). In contrast, p21 levels after Etoposide addition showed a strong cell cycle-dependency: p21 was expressed in all cells in G1 phase, virtually absent in cells in S phase [32], and dramatically increased in cells that had crossed the S/G2 border ( Fig 5D). Inhibition of Cdk1 and Cdk2 decreased p21 levels both in cells in G1 and G2 phase, indicating that Cdk activity affects p21 expression throughout the cell cycle ( Fig 5D, top panel). Analyzing p53 levels, we found stronger induction in S and G2-compared to G1-phase cells. When Cdk activity was inhibited, p53 levels were elevated in all cell cycle phases, again reaching the highest expression in S and G2 phase cells (Fig 5D bottom panel). Thus, our data indicate that the expression levels of p53 and p21 are regulated through the cell cycle and that Cdk activity decreases p53 levels but increases p21 levels in all cell cycle phases.
We next sought to test if Cdk activity affects production or degradation of p21.

Addition of the proteasome inhibitor MG-132 and the protein translation inhibitor
Cycloheximide affected p21 levels in Etoposide treated G2 cells, suggesting that p21 is continuously produced and degraded. Combined Cdk inhibition and proteasome inhibition resulted in lower p21 levels than in control Etoposide treated cells, suggesting that Cdk-mediated p21 expression cannot be explained solely by differences in p21 degradation ( Fig 5E). In line with this finding, Cdk activity did not significantly affect p21 stability after DNA damage ( Fig 5F). In contrast, we detect a 20 -30%-decrease in p21 mRNA levels after Cdk inhibition in Etoposide treated cells (Control) (Fig 5G and Supplementary Fig 5E). Thus, Cdk activity stimulates p21 production at least partly by increasing the amount of p21 mRNA. While the remaining Cdk activity sustains important cellular functions during a DDR, it also poses a risk for genome stability. If cells with damaged DNA progress into G2 phase they need to be prevented from entering mitosis, which otherwise would result in chromosome missegregation and propagation of mutations. Indeed, in the absence of p53 or p21, a cell cycle arrest in G2 phase is eventually overrun [8]. We show that Cdk activity is coupled to negative feedback by inducing p21 expression, leading to subsequent Cyclin B1 nuclear translocation and terminal cell cycle exit. As increasing Cdk activity drives mitotic entry, the incorporation of Cdk activity as a positive regulator of p21 expression provides an elegant mechanism to ensure cell cycle exit and senescence when most needed. Thus, our data highlights the overall importance

Plasmids, cloning, purification and transfection
The use of the live-cell sensor for Cdk2 activity (kindly provided by Tobias Meyer and Sabrina Spencer) has been described previously [27].
To clone substrates for the kinase assays, DNA fragments corresponding to the optimal Cdk2 substrate peptide HHASPRK or STPLSPTRIT peptide derived from Lamin A were ligated in frame into pGEX6P plasmid [51]. GST, GST-CDK2 or GST-LAMS22 substrates were purified from BL21 bacteria induced by 0.5 mM IPTG for 5h using glutathione beads.

Inhibitors, and RNAi
The inhibitors used in this study were employed at the following concentrations: were purchased from Dharmacon and employed at a concentration of 20 nM using HiPerFect (Qiagen) transfection at 48h and 24h before analysis of the phenotype.

Live-cell microscopy and quantitative immunofluorescence
Live-cell imaging experiments were done as previously described [6].
For quantitative immunofluorescence cells were fixed and immunostained as previously described [6]. Images were acquired on an ImageXpress system (Molecular Devices) using a 40x NA 0.6 or a 60x NA 0.7 objective. Images were manually screened for aberrant staining or illumination, and processed and analyzed To determine cell proliferation potential, cells were seeded in quadruplicates in 6-well plates 1 day before treatment with Etoposide with and without different inhibitors.
Cells were counted after 4 days of treatment, reseeded into fresh medium, and counted again after an additional 2 days.
To determine clonogenic capacity, cells were treated with Etoposide with and without different inhibitors. After 4 days 5000 cells were seeded into fresh medium in quadruplicates in 6-well plates. After an additional 7 days the samples were fixed ! ")! with 10% formaline and stained with 0.5% (w/v) crystal violet before the number of colonies was assessed.

Kinase assay
Asynchronously growing RPE Cyclin A2-eYFP cells or RPE cells released for 6h from thymidine block (2.5 mM, 24h) were treated with DMSO or with indicated      Cells were treated with Etoposide or mock treated with DMSO and fixed after 4h.
Cdk inhibitors were added 1h before fixation. More than 250 cells were analyzed for each condition. G2 cells were identified from DAPI staining. Statistical hypothesis testing was performed using two-sided t-test.  (E) Immunofluorescence quantification of Cyclin B1 pS126, Lamin A/C pS22 and Cdc6 pS54 nuclear fluorescence intensity of interphase 4n DNA content U2OS cells treated with Etoposide or mock treated with DMSO and fixed after 4h. Cdk inhibitors were added 1h before fixation. G2 cells were identified from DAPI staining. More than 150 cells were analyzed for each condition. Statistical hypothesis testing was performed using two-sided t-test.
(F) Immunofluorescence quantification of Cyclin B1 pS126, Lamin A/C pS22 and Cdc6 pS54 nuclear fluorescence intensity in interphase RPE cells with 4n DNA content cells treated as in (E) but fixed at 24h. More than 150 cells were analyzed per condition. Statistical hypothesis testing was performed using two-sided t-test.     Figure 3 -Cyclin A and Cyclin B associated activity is preserved during a DDR Figure 4 -Low levels of Cdk activity are preserved for several hours during a DDR in a cell cycle-dependent manner ! !