Research ResourceDNA damage

Comparative Phosphoproteomic Analysis of Checkpoint Recovery Identifies New Regulators of the DNA Damage Response

See allHide authors and affiliations

Science Signaling  23 Apr 2013:
Vol. 6, Issue 272, pp. rs9
DOI: 10.1126/scisignal.2003664


How cells recover from a DNA damage–induced arrest is currently poorly understood. We performed large-scale quantitative phosphoproteomics to identify changes in protein phosphorylation that occurred during recovery from arrest in the G2 phase of the cell cycle caused by DNA damage. We identified 154 proteins that were differentially phosphorylated, and systematic depletion of each of these differentially phosphorylated proteins by small interfering RNA (siRNA) identified at least 10 potential regulators of recovery. Astrin, a protein associated with the mitotic spindle, was among the potential regulators of recovery. We found that astrin controlled the abundance of the cell cycle regulator p53 during DNA damage–induced arrest. Cells in which astrin was depleted had decreased murine double minute 2 (MDM2) abundance and increased p53 at the later stages of the DNA damage response. Astrin was required for continued expression of genes encoding proteins that promote cell cycle progression in arrested cells. Thus, by controlling p53 abundance in cells recovering from DNA damage, astrin maintains the cells in a state competent to resume the cell cycle.


When a G2 cell encounters DNA damage, it activates a checkpoint that prevents entry into mitosis (1). This G2 checkpoint depends on several kinases that initiate and transmit the checkpoint signal to activate DNA repair and arrest the cell cycle (2). To initiate the checkpoint response, ATM (ataxia telangiectasia mutated) and ATR (ATM and Rad3-related) kinases are recruited to the site of damage where they phosphorylate and activate checkpoint kinases 1 and 2 (CHK1 and CHK2) (2). Together, these kinases facilitate a major change in protein phosphorylation that includes inhibitory phosphorylation of cyclin-dependent kinase 1 (CDK1) (1). As a consequence, cyclin B–CDK1 complexes cannot be activated, causing cells to arrest in G2.

Once the DNA damage is repaired, the checkpoint is assumed to be silenced through dephosphorylation or degradation of several key checkpoint components such as ATM or ATR, p53, CHK1, and CHK2 (3). This enables the removal of inhibitory phosphates from CDK subunits and promotes reentry into the cell cycle, but the exact mechanism of this reactivation is currently poorly understood. Previous work shows that a number of kinases and phosphatases play a crucial role during recovery in G2, namely, Polo-like kinase 1 (PLK1), Aurora A, CDK1, CDK2, and Cdc25B (46). Aurora A phosphorylates and activates PLK1 (5, 7), which in turn controls degradation of the kinase WEE1 (6), which phosphorylates, and thus inhibits, CDK subunits during the DNA damage response. Subsequent removal of the inhibitory phosphates is facilitated by the CDC25B phosphatase (6). Activation of PLK1 not only contributes to reactivation of the cell cycle but also promotes silencing of the DNA damage checkpoint at multiple levels (813). This implies that reactivation of the cell cycle is coordinated with silencing of the checkpoint through the action of PLK1.

The DNA damage checkpoint also restricts CDK1 function through the action of p53, which promotes transcriptional activation of the CDK inhibitor p21 (1416). Moreover, p53 can repress the expression of cyclin B and other mitotic regulators (1722). p53 is required for long-term maintenance of the G2 arrest, although it is not required to establish a G2 arrest (23). However, p53-mediated transcriptional repression must be restrained during an ongoing DNA damage response by the wild-type p53-induced phosphatase 1 (WIP1) to maintain cells in a state from which they can recover from the arrest (21). It is currently not known if and how p53 is inactivated when the checkpoint is switched off.

Several groups have performed large-scale phosphoproteomics on cycling cells (24, 25). Similarly, quantitative phosphoproteomics have been applied to identify new substrates of ATM and ATR that contribute to the DNA damage response (26). Here, we have performed large-scale quantitative phosphoproteomic analysis to identify changes in protein phosphorylation that occur during recovery. We identified 154 proteins whose phosphorylation was reproducibly changed during recovery. Systematic depletion of these individual proteins by RNA interference (RNAi) showed that a significant fraction of these proteins were required for recovery from the DNA damage–induced arrest.


Comparative phosphoproteomic analysis of checkpoint recovery

Given the importance of kinases and phosphatases in the DNA damage response and the cell cycle machinery, we reasoned that protein phosphorylation and dephosphorylation must play an important role in recovery. Therefore, a quantitative phosphorylation-centric shotgun proteomics approach was performed to identify alterations in protein phosphorylation during recovery from doxorubicin in different conditions, depicted schematically in Fig. 1A and fig. S1A. U2OS osteosarcoma cells were synchronized in G2 with thymidine and then treated with doxorubicin, an inhibitor of topoisomerase II, to induce double-stranded breaks. Eighteen hours after doxorubicin treatment [a time when all cells are still arrested in G2 (fig. S1B)], we treated cells with caffeine, an inhibitor of ATM and ATR, to induce mitotic recovery [fig. S1C and (6)]. To obtain a cell population in which the activities of ATM and ATR were inhibited, but recovery was prevented, we treated the arrested cells with caffeine in combination with the PLK1 inhibitor BI2536 during the indicated time intervals (Fig. 1A and fig. S1A). Cells treated with caffeine alone will eventually enter mitosis, whereas addition of BI2536 at these time points blocks recovery (5). Mitotic cells were removed by shaking to ensure that our analysis of recovery-associated changes in protein phosphorylation was not contaminated with mitosis-associated changes in protein phosphorylation. This way, the analyzed cell population consisted primarily of late G2 cells, with few mitotic cells (fig. S1D). Peptides from caffeine-treated or caffeine- and BI2536-treated cells were dimethyl-labeled with light or intermediate isotope, respectively, and mixed in equal ratio (27). Phosphopeptides were enriched by strong cation exchange (SCX) chromatography (28) and titanium dioxide (TiO2) (29), and fractions were analyzed by liquid chromatography–tandem mass spectrometry (LC-MS/MS) (LTQ-Orbitrap MS) (Fig. 1A) (30). Using a false discovery rate (FDR) of 1%, we identified and quantified 31,393 unique peptides from the four independent time points. Among these were 6119 unique phosphopeptides, indicative of successful phosphopeptide enrichment. From the 6119 quantified phosphopeptides, 1200 phosphopeptides (from 715 proteins) were altered by at least twofold, indicating that substantial changes in protein phosphorylation occur during recovery from a DNA damage–induced G2 arrest (Fig. 1B and tables S1 to S3). Twenty of these identified proteins (fig. S2A and table S4) interact with the Polo-box domain of PLK1 (31). Consensus motif analysis done by NetworKIN (32) showed that a large fraction of the differentially phosphorylated peptides contained a CDK consensus motif (fig. S2B), consistent with the reactivation of cyclin and CDK activity that accompanies cell cycle reentry. In addition, the casein kinase consensus motif (CK2A1 and CK2A2) was also abundant (fig. S2B), which is partially due to the rather loose consensus of the motif but is also in line with the observation that casein kinase is essential for tolerance to double-stranded breaks in yeast (33). Specific analysis of PLK1 consensus motifs showed that there was an almost threefold enrichment in PLK1 sites identified in the phosphopeptides, which were decreased compared with those that were unchanged in the presence of BI2536 (Fig. 1C). This reduction in the phosphorylation of PLK1 consensus motifs confirmed that BI2536 inhibited PLK1 activity. PLK1 inhibition resulted in the most prominent reduction in phosphorylation of CDK1 consensus motifs at the 6-hour time point (Fig. 1D), in agreement with the kinetics of CDK1 activation during recovery (5). On the other hand, phosphorylation of PLK1 consensus motifs was mostly enriched at the 2-hour time point (Fig. 1D). Comparison of independent biological replicates at the same time point (2 hours of treatment) resulted in a relatively high correlation coefficient of 0.77, indicating reproducibility of the proteomic analysis (fig. S2C). Finally, analysis of protein phosphorylation with commercially available antibodies confirmed decreases in phosphorylated CLASP1 at Thr656, phosphorylated vimentin at Ser56, and phosphorylated STMN1 at Ser25 in the samples treated with caffeine and BI2536, as compared with the samples treated with caffeine alone (fig. S2D), thus confirming the results from our phosphoproteomics.

Fig. 1 Phosphoproteomic analysis of checkpoint recovery.

(A) Schematic representation of the comparative phosphoproteomic analysis. (B) Table shows the number of identified phosphosites, quantified proteins, and identified and altered phosphopeptides from each sample of U2OS cell lysates. Reproducibility at the 2-hour time point against a biological replicate was assessed in fig. S2C. (C) Percentage of phosphopeptides with CDK1 and PLK1 motifs that showed no change, a decrease, or an increase in the abundance of phosphopeptides. The kinase consensus motifs used in the analysis are indicated. (D) Number of changed phosphopeptides with PLK1 and CDK1 motifs at each time point.

Of the 715 proteins that were differentially phosphorylated, 154 were identified at multiple time points (table S5). A significant fraction of the identified proteins have kinase or phosphatase activity, underlining the importance of phospho-regulation in the recovery process (fig. S3, A and B). Most of the candidates function in binding—either nucleic acid or protein binding. Moreover, a substantial proportion of the identified proteins are known to act in cell cycle control or DNA repair, suggesting that the repair process might be modified once the cells have committed to cell cycle reentry. In addition, 84 proteins in our list of 715 were previously identified in a phosphoproteomic screen for targets of ATM and ATR (26), yet they share only two phosphosites (table S6). This indicates that proteins directly modified by DNA damage–responsive kinases are additionally modified once these kinases are inactivated.

Functional characterization of the recovery-associated phosphoproteome

We next constructed a small interfering RNA (siRNA) library to investigate whether any of the 154 differentially phosphorylated proteins was required for checkpoint recovery. The library, generated from a commercially available genome-wide siRNA library, contained siRNA targeting 147 mRNAs from our list of 154 (siRNAs for seven targets were unavailable in the genome-wide library). The library was plated in pools of four individual siRNA duplexes per target. After protein depletion, cells were analyzed for their ability to recover from a DNA damage–induced G2 arrest. Mitotic index was determined 8 hours after caffeine addition by staining cells with antibodies against phosphorylated histone H3 (pH3) (Fig. 2A). siRNAs against glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or PLK1 were used as negative and positive controls, respectively (Fig. 2A). The screen was performed in triplicate, and proteins whose depletion resulted in greater than 20% reduction in checkpoint recovery in two or more experiments and did not affect mitotic entry in undamaged cultures were selected for further validation. A total of 28 of the 147 proteins fulfilled these criteria and were selected for further analysis (Fig. 2B and table S7). More than two-thirds (20 of 28) displayed altered phosphorylation after 2 and 6 hours of PLK1 inhibition, whereas about one-third was identified as differentially phosphorylated at the 30-min time point in our phosphoproteomic analysis. A minority (11%) of the identified phosphopeptides from these 28 proteins were found to be differentially phosphorylated across all time points.

Fig. 2 siRNA-based screen of the phosphoproteome.

(A) Schematic of the recovery screen in which U2OS cells were transfected with siRNAs and subjected to a recovery assay. (B) Heat plot of mitotic index (blue, reduced; red, increased) from the primary screen using pools of four siRNAs per gene target, in which transfections that reduced recovery by more than 20% over controls in at least two of three independent experiments but did not affect normal mitotic entry were considered hits (arrows). (C) Heat plot of mitotic index from the secondary deconvolution screen of the hits from (B). Genes for which three or more single siRNAs resulted in reduced recovery by more than 20% over controls in at least two of three independent experiments were considered true hits (asterisks).

To exclude possible off-target effects, we performed siRNA deconvolution for these 28 hits and separately tested the checkpoint recovery effects of the individual siRNAs from each pool used in the primary screen. Using stringent criteria in which at least three of four siRNAs of a pool scored positive (reduced cell cycle recovery by at least 20%) in at least two of three experiments, we validated 10 proteins as regulators of the DNA damage checkpoint recovery. These proteins are directly involved in recovery but not in normal G2-M progression (Fig. 2C and table S8). A further 12 proteins on our list showed validation for two of four siRNAs. Thus, our proteomics-directed siRNA screen led to a high hit rate.

Astrin and G2 checkpoint recovery

Depletion of astrin (also known as SPAG5) caused the most prominent recovery defect among all of the hits in our screen. Yet, depletion of astrin did not affect normal G2-M progression (Fig. 2, B and C, and table S7). Astrin is a mitotic spindle–associated protein that is important for normal mitotic spindle architecture and chromosome alignment in mitosis (3436). Astrin is known to interact with Aurora A (37) and can be phosphorylated in vitro by PLK1 (38). Despite its described functions in mitosis (26), nothing is known about a possible function for astrin in G2.

We confirmed that astrin was also required for spontaneous recovery (without the addition of caffeine) after γ-irradiation or treatment with doxorubicin (Fig. 3A). We observed 50 to 70% fewer mitotic cells in astrin-depleted samples compared to GAPDH-depleted controls. Again, no perturbation of mitotic entry was observed in astrin-depleted cells that were not damaged (Fig. 3A). We subsequently analyzed which proteins interact with astrin in DNA-damaged cells and identified 11 potential astrin interactors in cultures treated with doxorubicin (Fig. 3B), including SKAP, a known partner of astrin (39). Depletion of SKAP inhibited recovery after damage-induced arrest to a similar degree as did depletion of astrin and similarly had no effect on unperturbed G2-M progression (Fig. 3C). This further supports a role for astrin in the cellular response to DNA damage and suggests that it may act in complex with SKAP to promote recovery.

Fig. 3 Validation of astrin as a novel G2 checkpoint recovery component.

(A) U2OS cells were transfected with siRNAs against GAPDH or astrin and subjected to a recovery assay upon 5-Gy γ-irradiation or a 1-hour pulse of 1 μM doxorubicin. Mitotic index was determined by pH3 staining and fluorescence-activated cell sorting (FACS) analysis (n = 3 independent experiments). (B) Potential astrin interactors identified by immunoprecipitation of endogenous astrin from G2-arrested, doxorubicin-treated U2OS cells, assessed by MS (MW, molecular weight; ctrl, control immunoprecipitate). (C) Mitotic index of U2OS cells transfected with siRNAs against GAPDH or SKAP and subjected to a recovery assay as in (A) (n = 3 independent experiments). (D) Quantification of peptides with phosphorylation on Ser401 (pSer401) in control and BI2536-treated U2OS cells, assessed by LC-MS analysis. (E) Mitotic index and astrin protein abundance in U2OS cells expressing RNAi-resistant tetracycline-inducible wild-type (WT) or siRNA (#2)–resistant mutant S401A astrin that were transfected with siRNA against GAPDH or astrin (siRNA#2) and subjected to a recovery assay (n = 3 independent experiments).

Our phosphoproteomic analysis identified Ser401 phosphorylation on astrin to be regulated during G2 checkpoint recovery (table S1). Phosphorylation of astrin at Ser401 was decreased at both 2 and 4 hours after treatment with the PLK1 inhibitor BI2536 (table S5). Consistent with our phosphoproteomics data, we verified phosphorylation of astrin at Ser401 in damaged cultures treated with caffeine, which was substantially reduced with PLK1 inhibition (Fig. 3D). This concurs with previous reports showing that Ser401 is phosphorylated in mitotic cells in a PLK1-dependent manner, and that PLK1 can phosphorylate astrin in vitro (36, 38, 40). Thus, we conclude that PLK1 phosphorylates astrin on Ser401 in vivo.

We next tested the importance of this phosphorylation for G2 checkpoint recovery. To do this, we established stable U2OS cell lines expressing siRNA-resistant wild-type astrin or an S401A mutant under the control of a tetracycline-regulatable promoter. Endogenous astrin was depleted by siRNA before exogenous astrin expression was induced with tetracycline. Overexpression of siRNA-resistant wild-type astrin partially rescued the recovery defect caused by endogenous astrin depletion. In contrast, the S401A mutant was completely unable to restore recovery in astrin-depleted cells, underlining the importance of phosphorylation at Ser401 for the function of astrin in recovery (Fig. 3E).

The role of astrin in the proper activation of p53

Activation of PLK1 and inactivation of p53 are two key events that control the recovery of G2-arrested, DNA-damaged cells (5, 7, 21). Silencing astrin induces p53-dependent apoptosis (41). Therefore, we wondered if the effect of astrin on recovery was dependent on p53 function. Depletion of astrin did not inhibit recovery from a G2 arrest when p53 was co-depleted (Fig. 4A), indicating that the effect of astrin on G2 recovery is entirely mediated by p53. To confirm this, we next compared the effects of astrin depletion in HCT116 cells and HCT116 p53-null (p53−/−) cells (23). Similar to our observations in U2OS cells, depletion of astrin from HCT116 cells resulted in inhibition of checkpoint recovery from a G2 arrest (Fig. 4B). However, astrin depletion had no effect in HCT116 p53−/− cells, further supporting a role for p53 as a mediator of this effect (Fig. 4B). Furthermore, astrin protein abundance was substantially higher in p53-deficient cells compared with wild-type cells, suggesting that astrin itself may be a target of p53 (Fig. 4B).

Fig. 4 Astrin controls recovery through p53.

(A) Mitotic index, assessed by pH3 staining and FACS analysis, of U2OS cells transfected with the indicated siRNAs and subjected to standard recovery assay. The graph shows the means and SDs of three independent experiments. Astrin knockdown efficiency and p53 protein abundance were monitored by Western blot. (B) Mitotic index, astrin knockdown efficiency, and p53 protein abundance were determined in wild-type and p53−/− HCT116 cells treated as in (A). The graph shows the means and SDs of four independent experiments.

Unattenuated activation of p53 causes cells to lose the ability to resume cell division because of a progressive loss of essential mitotic activators, such as PLK1 and cyclins A and B, as a result of p53-dependent transcriptional repression (4, 21). Moreover, the oscillatory pattern of p53 abundance after DNA damage is required to maintain the arrest and ensure proper cell cycle reentry (42, 43). We therefore investigated whether astrin also acts to attenuate p53 activity after activation of the DNA damage checkpoint in G2 and whether it affects the oscillatory behavior of p53. We observed an increase in p53 abundance after doxorubicin treatment in siGAPDH-transfected cells, with a first peak at 6 to 9 hours after treatment and a second peak 14 to 16 hours afterward (Fig. 5A). However, in astrin-depleted cells, the protein abundance of p53 increased to its maximum point as early as 3 hours after doxorubicin treatment and remained high for the duration of the experiment (Fig. 5A). The increased abundance of p53 in astrin-deficient cells also correlated with a significant increase in the abundance of p21, a cell cycle arrest protein whose gene is transcriptionally activated by p53, and with reduced abundance of cyclin A and PLK1, cell cycle progression proteins whose genes are transcriptionally repressed by p53 (Fig. 5B). These data suggest that depletion of astrin leads to excessive transcriptional repression by p53, resulting in a reduced capacity to recover from the G2 arrest.

Fig. 5 Astrin regulates p53 signaling.

(A to C) The protein abundance of p53 (A), its targets (B), and MDM2 (C) in G2-synchronized, doxorubicin-treated (1 hour, 1 μM) U2OS cells transfected with the pool of siRNAs targeting astrin or GAPDH was assessed by Western blot. Protein abundance was normalized to their loading control, and the means and SDs of three independent experiments at 4 and 18 hours (A), 18 hours (B), or 10 and 18 hours (C) after doxorubicin treatment are shown.

The mRNA abundance of p53 was not affected by astrin depletion, indicating that the regulation of p53 by astrin was most likely at the posttranscriptional level (fig. S4A). The E3 ubiquitin ligase murine double minute 2 (MDM2) is a major regulator of p53 protein abundance, which has been shown to be essential for the dynamic behavior of p53 in response to double-strand DNA breaks (44). MDM2 is also a transcriptional target of p53, which functions as a part of a feedback loop in response to damage (44). In control cells, MDM2 protein abundance was increased in response to DNA damage and reached its maximum abundance at 10 hours after doxorubicin treatment. In contrast, astrin-depleted cells showed a lower basal abundance of MDM2, which did not increase in response to DNA damage, even in the presence of higher abundance of p53 (Fig. 5C). MDM2 mRNA abundance was similarly induced upon doxorubicin treatment in both control and astrin-depleted cells, with a peak at 8 and 12 hours after treatment, respectively (fig. S4, A and B), suggesting that astrin could regulate MDM2 protein abundance independently of p53. Only at later time points did the amount of MDM2 mRNA begin to return to endogenous abundance, although this return was much less pronounced in astrin-depleted cells compared with control cells (fig. S4B), probably as a consequence of the high abundance of p53 maintained in the absence of astrin (Fig. 5A). Therefore, the overall reduction in MDM2 protein abundance together with the increase in p53 protein abundance in astrin-depleted cells suggests that MDM2-dependent proteasomal degradation of p53 is impaired in cells lacking astrin. To confirm that the increase in p53 was indeed due to impaired proteasomal degradation of p53, we treated doxorubicin-treated cultures with the proteasomal inhibitor N-carbobenzyloxy-l-leucyl-l-leucyl-l-leucinal (MG132). Whereas MG132 treatment increased p53 abundance in control cells, it failed to further increase it in astrin-depleted cells (fig. S4C). This is consistent with the notion that proteasomal degradation of p53 is impaired in astrin-depleted cells, and indicates that astrin may suppress the protein abundance of p53 by promoting MDM2-dependent degradation.


To gain insight into the mechanisms controlling recovery from a DNA damage–induced arrest, we performed a phosphoproteomic screen and uncovered more than 700 proteins whose phosphorylation status was altered during recovery, 154 of which showed reproducibly altered phosphorylation at multiple time points. This indicates that substantial changes in protein phosphorylation occur as cells silence the DNA damage checkpoint and reenter the cell cycle. This is in line with the vast changes in protein phosphorylation that occur when the checkpoint is activated; more than 700 proteins are differentially phosphorylated in response to damage (26).

We found substantial enrichment for PLK1 consensus motifs in the peptides that were differentially phosphorylated during recovery, and most of the regulated phosphopeptides contain a CDK consensus, in agreement with our previous observations that PLK1 controls recovery through reactivation of cyclin and CDK (6). Many of the proteins phosphorylated or dephosphorylated during recovery are also known ATM and ATR substrates (26). However, during recovery, these proteins were phosphorylated or dephosphorylated at sites that are distinct from the known ATM and ATR sites. This suggests that inactivation of these proteins is not only regulated through dephosphorylation but might also involve phosphorylation-dependent inactivation or degradation. This concurs with the current knowledge that PLK1 plays an essential role in both the silencing of the DNA damage checkpoint and the activation of the cell cycle reentry machinery, and that both aspects of recovery are connected (6, 10).

Depletion of a substantial proportion of the differentially phosphorylated proteins exclusively affected mitotic entry after a DNA damage–induced arrest, similar to PLK1, indicating that these proteins actively promote recovery. The fact that such a high percentage of these proteins is involved in recovery indicates that our phosphoproteomic screen successfully identified physiologically relevant proteins. Moreover, the high hit rate shows that the combination of shotgun phosphoproteomics and siRNA-based screens is an excellent approach to uncover new regulators of a given cellular process. Astrin plays important roles in different aspects of mitosis (39, 45, 46), but no clear functions for astrin in interphase have been reported. Through its interaction with the nuclease Apollo (also known as SNM1B), astrin may function in the prophase checkpoint in response to spindle stress (47). Our results clearly demonstrate that astrin is required for mitotic entry after a DNA damage–induced G2 arrest, and that its phosphorylation by PLK1 at Ser401 is important for this function. Phosphorylation of this residue by PLK1 is also reported in several large screens for PLK1 substrates (36, 38, 40), both during entry into mitosis and in mitotic arrest, although its functional relevance to the mitotic functions of astrin was not reported. We demonstrate that the small kinetochore-associated protein SKAP, a known mitotic partner of astrin (39), was also required for checkpoint recovery. Together, these findings indicate that astrin and SKAP may function in complex as an effector of PLK1 signaling that, similar to PLK1 itself, plays distinct roles in the cell cycle: (i) regulating mitotic entry after a DNA damage–induced arrest in G2, and (ii) controlling different aspects of mitosis. We found that astrin promoted G2 checkpoint recovery through negative regulation of p53. Indeed, astrin is not the only PLK1 substrate that negatively controls p53. PLK1 can directly phosphorylate GTSE1 (G2 and S phase–expressed protein 1), promoting its translocation to the nucleus where it binds p53 and subsequently shuttles p53 out of the nucleus to promote its proteolytic degradation (48). Consistent with this, GTSE1 was differentially phosphorylated during recovery (table S1). Negative regulation of p53 involves several additional antagonists, such as MDM2 and WIP1. In this respect, it is important to note that in response to γ-irradiation, p53 activation occurs in an oscillatory pattern, controlled by MDM2 and WIP1 (42, 49). Thus, during an ongoing DNA damage response, a cell switches between phases of high or low p53 activity. This oscillation may enable continued expression of cell cycle regulatory genes that are essential for the G2-M transition, such as cyclin B and PLK1 (21). Our results show that the absence of astrin increased the abundance of p53 and perturbed its oscillatory behavior. This correlated with a reduced abundance of cell cycle regulatory proteins essential for the G2-M transition, probably due to excessive p53-dependent repression of the associated genes. The suppressive effect of astrin on p53 abundance appears to be mediated by MDM2 because MDM2 protein abundance is severely reduced in astrin-deficient cells, even in the presence of high amounts of mRNA. Further studies are needed to elucidate how exactly astrin regulates the protein abundance of MDM2.

Together, our results show that astrin plays an important role in the DNA damage response by keeping p53 protein abundance in balance during the checkpoint arrest, keeping the cells competent to recover. This concurs with the observation that an oscillating p53 response is compatible with recovery from damage, whereas sustained p53 signaling leads to irreversible arrest and senescence (44). As a consequence, it is expected that astrin, through its effect on MDM2 and p53 dynamics, will influence cell fate decisions in the context of the DNA damage response.

With the validation of astrin as a regulator of recovery, our comprehensive phosphoproteomic analysis in combination with the siRNA-based functional screen was a successful approach for the identification of new components of the recovery pathway. Our complete data set should serve as a valuable resource for future work investigating the roles that other proteins identified here may have in cell cycle recovery. Furthermore, the strategy developed here could easily be applied to the analysis of other signaling networks involved in different cellular processes.

Materials and Methods

Cell culture, transfection, and drugs

U2OS and HCT116 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 6% fetal bovine serum (FBS) and antibiotics. Astrin-inducible cell lines were generated by transfection of U2TR cells (U2OS cells expressing the tetracycline repressor) with pcDNA4/TO plasmids (Invitrogen) with the standard calcium phosphate transfection protocol, and stable clones were selected in DMEM with 6% tetracycline system–approved FBS (Clontech), antibiotics, and zeocin (400 μg/ml). Expression was induced by the addition of tetracycline (1 μg/ml). siRNAs were reverse-transfected with HiPerFect (Qiagen) according to the manufacturer’s guidelines. Thymidine, puromycin, nocodazole, doxorubicin, and caffeine were purchased from Sigma and used at 2.5 mM, 2 μg/ml, 250 ng/ml, 0.5 to 1 μM, and 5 mM, respectively. BI2536 (Boehringer Ingelheim Pharma) was used at 100 nM. Cell synchronization, DNA damage application, and recovery assays were performed as previously described (5) and outlined schematically in Fig. 1 and fig. S1.

Plasmids and oligonucleotides

Human SPAG5 (astrin) complementary DNA (cDNA) was purchased from OriGene (SC116082), converted into an RNAi-resistant mutant (S401A) by site-directed mutagenesis, and subcloned into pCDNA4/TO plasmid (Invitrogen). pRetrosuper (pRS) and pRS-p53 were previously described (21). The siRNA library was composed of four On-Target duplexes per gene (Dharmacon). siRNAs targeting p53, SKAP, and GAPDH mRNAs were On-Target SMART pools from Dharmacon. siRNA against PLK1 was previously described (5).


Antibodies against p53 (DO-1), cyclin A2 (H-432), FoxM1 (C-20), p21 (C-19), and actin (I-19) were from Santa Cruz Biotechnology and were used at 1:1000 dilution. Antibody against MDM2 was from BD Biosciences (1:500 dilution), and phosphorylated (Ser10) histone H3 was from Millipore (1:1500 dilution). Phosphorylated (Thr656) CLASP1 was from Abgent, phosphorylated (Ser56) vimentin was from Cell Signaling Technology, and phosphorylated (Ser25) STMN1 was from Abnova (all at 1:500 dilution). Antibody against astrin (MAP126) was from Bethyl Laboratories (1:500 dilution), and antibody against PLK1 (1:1000 dilution) was produced as previously described (5).

Protein extraction, proteolytic digestion, and peptide labeling

Protein extraction, digestion, and peptide labeling were performed as previously described (27, 28).

Phosphopeptide enrichment and MS analysis

SCX was performed as described (28), and 35 fractions were collected. Offline TiO2 chromatography (50) was used to enrich further selected SCX fractions. Analyses were performed with a nano-LC-LTQ-Orbitrap (Thermo Scientific), as described previously (30).

Peptide identification and quantitative analysis

Raw data were converted to single DTA files using DTA SuperCharge and merged into Mascot Generic Format (MGF) files, which were searched with an in-house licensed Mascot v2.2 search engine against Human SwissProt 56.2 database concatenated with reversed sequences as decoy [containing 40,656 sequences (20,328 forward sequences)]. Carbamidomethyl cysteine and oxidized methionines were set as fixed modifications; serine, threonine, and tyrosine phosphorylations were set as variable modifications; quantification was set with dimethyl double mode. The mass tolerance of the precursor ion was 10 parts per million (ppm) and 0.9 dalton for fragment ions. The FDR was determined as <1% (Mascot score threshold of 31) with the decoy database approach, and the MGF files were trimmed at a Mascot score threshold of 31 (1% FDR) by RockerBox (51). MSQuant v1.5 was used to quantitate the amounts of the identified phosphopeptides and determine the exact phosphorylation site within the peptide (52). Every phosphopeptide quantitation was manually validated; peptides with low signal-to-noise ratios, low number of MS scans, or overlapping peaks were not included for quantitative purposes. Histogram plots of the ratio of whole quantified peptides in each experiment were used to normalize the ratios.

Data analysis

A program to extract information from MSQuant output was developed in-house to identify the position of each phosphorylation site and its status in current available databases (Phospho ELM and SwissProt). The Panther Classification System was used to classify proteins with regulated phosphopeptides. Gene-E was used to generate a heat map presenting the results of the siRNA analysis. NetworKIN was used to predict in vivo kinases for the phosphosites.

Automated image analysis

Cells were grown in 96-well plates (Viewplate-96, Perkin Elmer) in 100 μl of culture medium. Cells were fixed by the addition of 100 μl of 8% formaldehyde solution to the medium to prevent loss of mitotic cells and subsequently permeabilized with cold methanol. Cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) and the indicated antibodies. Image acquisition was performed using a Cellomics ArrayScan VTI (Thermo Scientific) with a 10× (0.50 numerical aperture) objective. Image analysis was performed with a Cellomics ArrayScan HCS Reader (Thermo Scientific). In short, cells were identified on the basis of DAPI staining, and they were scored as “mitotic” if the pH3 staining reached a preset threshold (Target Activation Bioapplication).

FACS, immunofluorescence, immunoprecipitation, and Western blot

FACS analysis was performed as described previously (6). Immunofluorescence was performed as previously described (5). For immunoprecipitations, cells were lysed in buffer containing 50 mM Hepes-NaOH, 150 mM NaCl, 0.5 mM EDTA, and 0.2% NP-40, supplemented with protease and phosphatase inhibitors. Immunoprecipitation and Western blots were performed as described previously (53).

Supplementary Materials

Fig. S1. Schematic of the checkpoint recovery assay.

Fig. S2. In-depth proteomics data analysis.

Fig. S3. Enrichment of candidates with particular functions.

Fig. S4. Astrin affects the p53/MDM2 feedback loop.

Table S1. All quantified phosphopeptides.

Table S2. All identified peptides (unphosphorylated).

Table S3. All identified proteins.

Table S4. Comparing phosphosites that are phosphorylated by PLK1 and phosphosites of Polo box interactors.

Table S5. Protein with changes on phosphopeptide abundance (at least in two experiments).

Table S6. Comparing phosphosites that are phosphorylated by PLK1 during recovery and also phosphorylated with pSQ/TQ motif upon DNA damage.

Table S7. Primary siRNA screen to identify proteins involved in recovery and unperturbed mitotic entry control.

Table S8. Secondary screen for siRNA pool deconvolution.

References and Notes

Acknowledgments: We thank W. Bruinsma, A. Janssen, L. Krenning, L. Macurek, A. Lindqvist, D. Warmerdam, T. Y. Low, J. Munoz, and J. Gouw for their input and valuable discussions. We thank D. A. Egan for the assistance in siRNA screening. Funding: This work was supported by a TOP-GO grant from the Netherlands Organization for Scientific Research (NWO ZonMW 912100651 to R.H.M., S.M., and V.A.H.), VIDI Grant 700.10.429 (to S.M.), and the Gravitation Grant of the Cancer Genomics Netherlands Consortium. Author contributions: V.A.H., M.A.-F., A.J.R.H., S.M., and R.H.M. designed the experiments; V.A.H., M.A.-F., Y.J.X., and M.A. performed the experiments; V.A.H. and H.W.P.v.d.T. performed computational work; V.A.H., M.A.-F., A.J.R.H., S.M., and R.H.M. analyzed the data; and V.A.H., M.A.-F., and R.H.M. wrote the paper. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article