A balanced pyrimidine pool is required for optimal Chk1 activation to prevent ultrafine anaphase bridge formation

Genetic stability is ensured by complete genome duplication and balanced chromosome segregation. DNA replication errors are particularly deleterious because they can affect both DNA duplication and segregation during mitosis (Gelot et al., 2015). Replication stress is defined as any phenomenon affecting completion of the DNA replication program, including alterations to the initiation and elongation steps of DNA replication, conflicts between DNA replication and metabolic pathways, such as transcription and mRNA processing, nucleotide pool disequilibrium, and the overexpression or activation of oncogenes (Hills and Diffley, 2014; Lecona and Fernandez-Capetillo, 2014; Magdalou et al., 2014; Mazouzi et al., 2014). Such stress can jeopardize the completion of DNA replication and chromosome segregation, as illustrated by our previous studies showing that pyrimidine pool disequilibrium resulting from cytidine deaminase (CDA) deficiency is a source of replication stress and subsequent segregation defects (Chabosseau et al., 2011; Gemble et al., 2015). CDA is an enzyme of the pyrimidine salvage pathway that catalyzes the hydrolytic deamination of cytidine and deoxycytidine into uridine and deoxyuridine, respectively (Nygaard, 1986). CDA deficiency leads to intracellular dCTP accumulation, lowering the basal activity of poly(ADP-ribose) polymerase 1 (PARP-1), a multifunctional enzyme involved in many cellular processes, including in particular the response to DNA damage (Tallis et al., 2014). Decreases in PARP-1 activity lead to the under-replication of some ‘difficult-to-replicate’ loci in the genome, resulting in ultrafine anaphase bridge (UFB) formation (Gemble et al., 2015). UFBs reflect a defect in sister chromatid segregation during anaphase. They cannot be stained with conventional DNA dyes or antibodies against histones, but they can be detected with an antibody against Plk1-interacting checkpoint helicase (PICH, also known as ERCC6L), a protein recruited to UFBs (Baumann et al., 2007; Chan et al., 2007). UFBs are present in almost all cell lines and are, therefore, probably physiological structures (Chan et al., 2007). In the absence of exogenous stress, most are of centromeric origin, but some originate from common fragile sites or telomeres (Chan et al., 2007, 2009; Barefield and Karlsedar, 2012). UFB frequency often increases under constitutive or induced replication stress conditions (Chan et al., 2007, 2009).

CDA dysfunction has pathological implications, particularly in cancers, in which CDA activity might be either up- or downregulated, affecting the sensitivity of cancer cells to nucleotide analogs, such as gemcitabine (Ye et al., 2015; Zauri et al., 2015; Serdjebi et al., 2015). The importance of CDA defects and, more broadly, of pyrimidine pool disequilibrium, is illustrated by Bloom syndrome, a rare genetic disease displaying one of the strongest known correlations between chromosomal instability and an increase in cancer risk (German, 1997). In Bloom syndrome(DSA2) cells, the expression of genome instability markers results partly from pyrimidine pool disequilibrium resulting from a CDA defect. In particular, the supernumerary UFBs observed in Bloom syndrome cells result from CDA deficiency, which decreases basal PARP-1 activity (Gemble et al., 2015). Supernumerary UFBs are generated from unreplicated DNA. Thus, these observations raise questions about the mechanism underlying the persistence of unreplicated DNA during mitosis that we will need to resolve to improve our understanding of the pathological consequences of the incomplete genome replication caused by CDA defects.

We explored the potential role of one of the targets and partners of PARP-1, checkpoint kinase 1 (Chk1, also known as Chek1), in preventing the accumulation of unreplicated DNA during mitosis. Chk1, and its kinase, ataxia telangiectasia and Rad3-related (ATR), are key players in the regulation of the intra-S, S–M and G2–M checkpoints (Sancar et al., 2004; Petermann and Caldecott, 2006; Eykelenboom et al., 2013; Smits and Gillespie, 2015). In response to genotoxic stress, the serine 317 and 345 residues of Chk1 are phosphorylated by ATR (Zhao and Piwnica-Worms, 2001; Liu et al., 2000), abolishing Chk1 auto-inhibition and allowing the subsequent activation of downstream checkpoints. Thus, activated Chk1 plays a major role in maintaining replication fork integrity and preventing cells from entering mitosis prematurely (Zuazua-Villar et al., 2014).

We show here that a lowering of basal PARP-1 activity levels in CDA-deficient cells compromises optimal Chk1 activation and decreases the efficiency of downstream checkpoints, leading to the accumulation of unreplicated DNA during mitosis and, ultimately, UFB formation. These results establish an unexpected link between the nucleotide pool and Chk1 activation and reveal a new role for Chk1 in preventing chromosome segregation defects.

PARP-1 catalyzes poly(ADP-ribosyl)ation (PARylation) by transferring ADP-ribose units from nicotinamide (NAD⁺) onto diverse acceptor proteins (Luo and Kraus, 2012). The interaction of Chk1 with PAR is required for its efficient retention on DNA, and, thus, for its optimal activation (Min et al., 2013). We hypothesized that the lower levels of PARP-1 activity in CDA-deficient cells might decrease the stability of Chk1 on DNA and its activation. We analyzed the relative abundance of Chk1 in the chromatin-bound fractions from CDA-deficient or -proficient cells. GAPDH, BLM and histone 3 (H3) were used as loading controls for the cytoplasm, soluble nuclear and chromatin-bound fractions, respectively (Fig. 1A; Fig. S1A). Chk1 was found to be less abundant in the chromatin-bound fraction of PARP-1-deficient cells than in that of control cells. We thus confirmed that PARP-1 defects decrease the stability of Chk1 on DNA (Min et al., 2013) (Fig. S1A). Chk1 was also less abundant in the chromatin-bound fraction of CDA-deficient cells than in that of control cells (Fig. 1A). Additionally, PARP-1 was less abundant in the chromatin-bound fraction in the absence of CDA (Fig. 1A). Thus, CDA-deficient cells have small amounts of Chk1 bound to their DNA, suggesting that the activation of this protein might be weak in these cells.

We then investigated whether Chk1 activity was affected in CDA-deficient cells. Phosphorylation of the serine 317 residue of Chk1 is a classical marker of the activation of this protein (Zhao and Piwnica-Worms, 2001; Katsuragi and Sagata, 2004; Ng et al., 2004). However, this phosphorylation cannot be detected by standard immunoblotting in basal conditions without exogenous stress. We thus activated Chk1 by exposing CDA-deficient cells, or PARP-1-depleted cells as a control, to UVC (8 J/m²) and analyzed Chk1 phosphorylation by immunoblotting. For CDA deficiency, we used two isogenic cell models, CDA-depleted HeLa cells and the corresponding control cells [HeLa-shCDA and HeLa-Ctrl_(CDA), respectively] and Bloom syndrome cells expressing neither BLM nor CDA [BS-Ctrl_(BLM)] and the corresponding control cells expressing both BLM and CDA (BS-BLM) (Gemble et al., 2015). Chk1 phosphorylation levels were lower in PARP-1-deficient cells than in control cells (Fig. S1B), confirming that PARP-1 contributes to optimal Chk1 activation (Min et al., 2013). In CDA-deficient cells, Chk1 phosphorylation levels were 34% lower than those in CDA-expressing cells after exposure to UVC (Fig. 1B). These observations were reproduced by treating both HeLa-shCDA and BS cells with 25 nM CPT for 2 or 8 h (Fig. 1C; Fig. S1C).

CDA deficiency leads to an intracellular accumulation of dCTP (Chabosseau et al., 2011; Gemble et al., 2015). The addition of dC to the culture medium of CDA-proficient cells reproduces the increase in dCTP levels observed in CDA-deficient cells, mimicking CDA deficiency (Chabosseau et al., 2011; Gemble et al., 2015). We treated CDA-proficient and CDA-deficient cells with dC before CPT treatment. Lower levels of PARP-1 activity, resulting in lower levels of Chk1 activation, were observed in both cell types (Fig. 1D; Fig. S1E). Thus, increasing the size of the intracellular dCTP pool is sufficient to weaken the activation of PARP-1 and Chk1 in response to genotoxic stress and to decrease the activation of these proteins further in CDA-depleted cells. These data reveal an unexpected link between nucleotide pool disequilibrium and Chk1 activation.

The intracellular accumulation of dCTP resulting from CDA deficiency leads to a 30–50% decrease in basal PARP-1 activity (Gemble et al., 2015). We therefore investigated whether PARP-1 activity was also compromised in response to acute exposure to genotoxic stress, reflecting suboptimal Chk1 activation. We quantified PAR levels in CDA-deficient cells following treatment with either CPT or UVC, by immunofluorescence analysis with customized software for the automatic counting of PAR foci (Gemble et al., 2015). Basal PARP-1 activity in untreated CDA-deficient cells was weak, as previously reported (Gemble et al., 2015), but we also observed significantly fewer PAR foci in CDA-deficient cells treated with 25 nM CPT or exposed to 8 J/m² of UVC than in control cells exposed to the same stimuli (Fig. 1E,F; Fig. S1D). Thus, the PARP-1 activity of CDA-deficient cells is weak not only in basal conditions, but also in response to genotoxic stress.

These data indicate that, in CDA-deficient cells subjected to genotoxic stress, the intracellular accumulation of dCTP decreases PARP-1 activity, thereby compromising optimal Chk1 activation.

Our data showing that Chk1 activation was compromised in CDA-deficient cells subjected to genotoxic stress led us to explore the possible functional consequences for downstream cell-cycle delays in these cells. Chk1 plays a key role in the S- and G2–M-phase checkpoints in response to DNA damage (Smits and Gillespie, 2015; Segurado and Tercro, 2009). S-phase checkpoint activation suppresses replication origin firing and stabilizes stalled replication forks, leading to a decrease in DNA synthesis and an accumulation of cells in S-phase (Smits and Gillespie, 2015; Segurado and Tercero, 2009). Thus, to quantify S-phase checkpoint efficiency, we first measured DNA synthesis in CDA-deficient cells and control cells by exposing them to UVC (8 J/m²) and adding EdU to the cell culture medium for one hour, either immediately after UVC treatment or 16 h later. Then we analyzed EdU incorporation into DNA by flow cytometry, to determine the relative proportions of DNA synthesis after UVC treatment, with decreases considered to reflect S-phase checkpoint efficiency. DNA synthesis had decreased one hour after UVC treatment in all cells analyzed (Fig. 2A; Fig. S2A), but PARP-1-depleted cells displayed higher levels of DNA synthesis than PARP-1-expressing cells, as previously reported (Min et al., 2013) (Fig. S2A). Similarly, CDA-depleted cells displayed higher levels of DNA synthesis one hour after UVC exposure than CDA-expressing cells (Fig. 2A). DNA synthesis had returned to basal levels 16 h after UVC exposure in both cell types, confirming that the decrease in DNA synthesis resulted from transient checkpoint activation rather than cell death (Fig. 2A; Fig. S2A). The inhibition of Chk1 by CHIR-124, its specific inhibitor (Tse et al., 2007), abolished the decrease in DNA synthesis one hour after UVC treatment, in both CDA-deficient and CDA-proficient cells (Fig. S2B). The efficiency of Chk1 inhibition was confirmed by the accumulation of Cdc25A protein, the degradation of which is dependent on Chk1 activity (Boutros and Ducommun, 2008) (Fig. S2C). These data confirmed that the decrease in DNA synthesis in response to UVC resulted from Chk1-dependent activation of the S-phase checkpoint, and revealed a lower efficiency of this checkpoint in CDA-deficient cells.

Further support for these conclusions was provided by an analysis of the cell cycle distribution of cells with and without CDA expression, at various times after exposure to genotoxic stress. The increase in the percentage of cells in S-phase (red arrows in left panels, Fig. 2B,C) in response to CPT or UVC treatment was significantly smaller in CDA-deficient cells (Fig. 2B,C, right panels) than in control cells, and entry into S-phase was inhibited by CHIR-124 treatment (Fig. S2D,E).

Thus, Chk1-dependent S-phase checkpoint activation was less efficient in the absence of CDA, reducing the inhibition of DNA synthesis and resulting in the accumulation of smaller number of cells in S-phase, in response to a genotoxic stress.

We also observed that more CDA-deficient cells than control cells accumulated in the G2–M phase (Fig. 2B,C, left panels). We therefore investigated the efficiency, in CDA-deficient cells, of the G2–M checkpoint, which prevents entry into mitosis in the presence of DNA damage (Dai and Grant, 2010). We quantified histone 3 with a phosphorylated serine 10 residue (H3S10), a specific marker of mitosis, after CPT treatment. The H3S10 signal was still visible eight hours after treatment with 25 nM CPT in CDA-deficient cells, whereas it was not detectable in control cells (Fig. 2D). Thus, a higher percentage of cells enter mitosis, despite DNA damage, in the absence of CDA. Consistent with these results, analysis of the mitotic index revealed a significantly higher percentage of CDA-deficient cells than of control cells in mitosis after CPT treatment (Fig. 2E). These results were confirmed by flow cytometry analysis, which showed the percentage of H3S10-positive cells after CPT treatment to be significantly higher for CDA-deficient cells than for CDA-proficient cells (Fig. S2F). Thus, G2–M checkpoint activation was less efficient in CDA-deficient cells.

These results show that the low levels of Chk1 activation in CDA-deficient cells decrease the efficiency of the DNA damage-induced S and G2–M checkpoints, leading to a premature passage of cells from S-phase to mitosis despite the presence of DNA damage and, potentially, under-replicated DNA, particularly at late-replicating sequences. As CDA-deficient cells contain smaller amounts of DNA-bound Chk1, we hypothesized that the low levels of basal Chk1 activity contributed to the formation of UFB-containing unreplicated DNA in the absence of exogenous genotoxic treatment.

The excess UFB formation in CDA-deficient cells is a marker of unreplicated DNA accumulation during mitosis, but the origin of the basal UFBs remains unclear (Gemble et al., 2015). Chk1 inhibition triggers the premature entry of S-phase cells into mitosis before the completion of DNA replication (Zuazua-Villar et al., 2014), and we observed low levels of Chk1 activation in CDA-deficient cells (Fig. 1), which are known to present supernumerary UFBs (Gemble et al., 2015). We therefore hypothesized that Chk1 inhibition might favor the accumulation of unreplicated DNA during mitosis, and, consequently, UFB formation. We treated CDA-deficient or CDA-proficient cells with Chk1 inhibitor CHIR-124 or with VE-821, an inhibitor of ATR, the kinase that activates Chk1 in response to DNA damage (Zhao and Piwnica-Worms, 2001), and analyzed UFB frequency in anaphase cells. Increases in H3S10 expression in both CDA-proficient and CDA-deficient cells treated with VE-821 revealed premature entry into mitosis, confirming ATR inhibition (Fig. 3A). We found that the inhibition of Chk1 or ATR led to an increase in UFB frequency in CDA-proficient cells, to levels similar to those in CDA-deficient cells, whereas it had no effect on UFB frequency in CDA-deficient cells (Fig. 3B–D). These results demonstrate the importance of the ATR–Chk1 pathway for preventing supernumerary UFB formation. As the inhibition of ATR or Chk1 had no effect on UFB frequency in CDA-deficient cells, these results suggest that CDA and ATR–Chk1 prevent UFB formation through the same pathway. Our previous results showing that CDA-deficient cells display a constitutive activation of both γH2AX and Chk2 (Gemble et al., 2015) led us to investigate the possible involvement of the ATM–Chk2 pathway in supernumerary UFB formation in these cells. We analyzed UFB frequency in CDA-deficient cells and in CDA-proficient cells left untreated or treated with KU55933, a specific inhibitor of ATM (Hickson et al., 2004). As expected, the level of phosphorylation of the threonine 68 residue of Chk2 by ATM was much lower in the KU55933-treated cells, confirming the inhibition of ATM (Fig. S3A). ATM inhibition had no effect on UFB frequency in CDA-expressing cells or CDA-deficient cells (Fig. S3B), demonstrating an absence of ATM–Chk2 pathway involvement in UFB formation.

We have shown that the treatment of CDA-deficient cells with very low doses of CPT (2 pM) is sufficient to increase basal levels of PARP-1 activity to those observed in untreated CDA-proficient cells. Such a restoration of PARP-1 activity is sufficient to prevent the accumulation of unreplicated DNA and, thus, supernumerary UFBs (Gemble et al., 2015). We investigated whether the ATR–Chk1 pathway acted downstream from PARP-1 to prevent UFB formation in CDA-deficient cells, by determining whether the suppression of UFB formation by low-doses of CPT required a functional ATR–Chk1 pathway. In CDA-proficient cells, the inhibition of Chk1 or ATR stimulated UFB formation, independently of treatment with low doses of CPT (Fig. 3E,F, black bars). In CDA-deficient cells, low doses of CPT abolished the formation of supernumerary UFBs, unless Chk1 or ATR was inhibited (Fig. 3E,F, gray bars). Thus, PARP-1 activation prevents the formation of supernumerary UFBs in CDA-deficient cells in an ATR–Chk1-dependent manner.

These findings demonstrate that the accumulation of unreplicated DNA and, thus, of supernumerary UFBs in the absence of CDA is caused by low levels of PARP-1 activity, resulting in a weakening of ATR–Chk1-dependent checkpoints.

The results presented above suggest that the low levels of ATR–Chk1 activity observed in the absence of CDA lead to the premature entry of S-phase cells with partially replicated chromosomes into G2–M, resulting in an excess of UFB-containing unreplicated DNA.

These data led us to investigate whether delaying entry into mitosis might allow cells to complete DNA replication before entering mitosis, thereby preventing UFB formation. The treatment of cells with low doses of RO-3306, a cyclin-dependent kinase 1 (Cdk1) inhibitor, or with volasertib, a polo-like kinase 1 (Plk1) inhibitor, delays entry into mitosis (Eykelenboom et al., 2013; Rudolph et al., 2009; Raab et al., 2015).

Cells were treated with either RO-3306 or volasertib, leading to an increase in the percentage of cells in the S and G2 phases (Fig. 4A,B), and an increase in cyclin B1 levels (Fig. 4C,D). This delaying of entry into mitosis was associated with a significant decrease in the frequencies of unreplicated centromeres and UFBs during mitosis, not only in CDA-deficient cells, but also in CDA-proficient cells (Fig. 4E–G; Fig. S4A,B). We checked that the cells in anaphase had previously experienced extended S and G2 phases by adding EdU to cell cultures one hour before RO-3306 treatment for the labeling of S-phase cells (Fig. S4C). All the mitotic cells analyzed were EdU-positive, confirming that cells in anaphase with low UFB frequencies had previously gone through extended S and G2 phases (Fig. S4D). Finally, given that PICH is regulated by Plk1 and Cdk1 (Baumann et al., 2007), we checked that low doses of RO-3306 had no effect on the location of PICH at the centromeres of mitotic cells (Fig. S4E), confirming that the decrease in UFB frequency did not result from a change in PICH distribution.

These data indicate that delaying entry into mitosis is sufficient to promote the completion of DNA replication, thereby preventing the formation of basal and supernumerary UFBs. These findings suggest that all UFBs result from the accumulation of unreplicated DNA during mitosis.

We recently reported a new mechanism by which excess cellular dCTP, resulting from CDA deficiency, partially inhibits the basal activity of PARP-1, leading to the under-replication of some ‘difficult-to-replicate’ loci, resulting in UFB formation during mitosis (Gemble et al., 2015). Here, we identify a downstream target of PARP-1, the ATR–Chk1 pathway, dysfunctions of which lead to an accumulation of UFB-containing under-replicated DNA. Our findings reveal that pyrimidine pool disequilibrium jeopardizes Chk1 activation. We demonstrate that the intracellular accumulation of dCTP resulting from CDA deficiency reduces PARP-1 activity and, thus, Chk1 activation, thereby weakening the S and G2–M checkpoints. We suggest that this decrease in checkpoint efficiency promotes the accumulation of unreplicated DNA during mitosis, leading to excess UFB formation (Fig. 5).

We found that PARP-1 activity levels were low in CDA-deficient cells not only in basal conditions, but also in response to genotoxic stress. The intracellular accumulation of dCTP thus also impairs the enhancement of PARP-1 activation. PARP-1 interaction with DNA is necessary for its activation (Langelier and Pascal, 2013); the low levels of PARP-1 in the chromatin-bound fraction from CDA-deficient cells studied here suggest that increases in dCTP concentration disturb the interaction of PARP-1 with DNA, thereby compromising its subsequent activation.

We also found that Chk1 levels were low in the chromatin-bound fraction from CDA-deficient cells. These results are consistent with those of a previous study showing that PAR, supplied by PARP-1, interacts with Chk1, promoting its efficient retention on DNA and its optimal activation (Min et al., 2013). Thus, the small amounts of DNA-bound Chk1 probably result from the low level of PARP-1 recruitment to DNA, and, thus, from insufficient Chk1–PAR interaction. As expected, the low levels of DNA-bound Chk1 affected the activation of this protein in response to genotoxic stress, confirming that optimal PARP-1 activity is required for the correct enhancement of Chk1 activation. Culturing cells with deoxycytidine, which is known to increase intracellular levels of dCTP and to decrease PARP-1 activity (Gemble et al., 2015), was sufficient to reproduce the cellular phenotype associated with CDA-deficiency; poor Chk1 activation in response to genotoxic stress. These data reveal an unexpected requirement for a balanced nucleotide pool to ensure optimal Chk1 activation.

The weakening of downstream checkpoint efficiency is a direct consequence of the low levels of Chk1 activity in the absence of CDA. Indeed, CDA-deficient cells present incomplete S and G2–M checkpoint activation in response to genotoxic stress. These data confirm our previous findings that, in response to UVC, BLM-deficient cells display partial escape from the G2–M cell cycle checkpoint (Ababou et al., 2002). The failure of CDA-deficient cells to activate the S and G2–M checkpoints optimally in response to genotoxic stress might account for the accumulation of DNA damage observed directly by electron microscopy in these cells and indirectly detected through the constitutive activation of the γH2AX–Chk2 pathway (Gemble et al., 2015).

Our findings suggest that insufficient Chk1 activity allows cells with partially replicated chromosomes to enter mitosis, leading to the formation of UFB-containing under-replicated DNA in CDA-deficient cells. We would thus expect to reproduce this phenotype by inhibiting Chk1 or its sensor kinase, ATR. We found that ATR–Chk1 pathway inhibition resulted in excess UFB formation in CDA-proficient cells but not in CDA-deficient cells. These results not only support our hypothesis, but also indicate that CDA, PARP-1, Chk1 and ATR act through the same pathway to prevent the accumulation of unreplicated DNA, and, thus, the formation of supernumerary UFBs, during mitosis. Consistent with this finding, we demonstrated that restoring PARP-1 activity by treating CDA-deficient cells with very low doses of CPT did not prevent UFB formation if either Chk1 or ATR was inhibited.

The ATR–Chk1 pathway prevents S-phase cells from entering mitosis prematurely (Eykelenboom et al., 2013; Zuazua-Villar et al., 2014). Thus, the low levels of Chk1 activation in CDA-deficient cells might promote premature entry into mitosis, before the completion of DNA replication, thereby accounting for the excess of UFB-containing unreplicated DNA formation. Consistent with this hypothesis, we showed that delaying entry into mitosis led to a decrease in the frequencies of unreplicated centromeres and UFBs during mitosis, not only in CDA-deficient cells, but also in CDA-proficient cells. These results indicate that all UFBs are derived from unreplicated DNA sequences, regardless of CDA status. We suggest that UFBs can be considered markers of the accumulation of under-replicated DNA during mitosis. This raises the question of the significance of UFBs, as these structures have been detected in all cells tested, whether primary or transformed, under physiological conditions (Chan et al., 2007). The MUS81–EME1 and ERCC1 nucleases and mitotic DNA synthesis have been reported to process unreplicated DNA sequences in mitosis (Gemble et al., 2015; Ying et al., 2013; Naim et al., 2013; Bergoglio et al., 2013), preventing excess UFB formation. All the experimental strategies used to date have succeeded only in reducing UFB numbers to the threshold UFB frequency in control cells, a physiological level. Our results are the first to show that delaying entry into mitosis is sufficient to reduce physiological UFB frequency in control cells to below this threshold. We therefore suggest that UFB formation is an additional mechanism for processing unreplicated DNA during mitosis, in situations in which processing by nucleases or mitotic DNA synthesis has failed.

Cell lines were cultured in DMEM supplemented with 10% FCS. BS-Ctrl_(BLM) and BS-BLM cells were obtained and cultured as previously described (Chabosseau et al., 2011). HeLa-Ctrl_(CDA) and HeLa-shCDA cells were obtained and cultured as previously described (Gemble et al., 2015). HeLa-Ctrl_(PARP-1) and HeLa-shPARP-1 cells were cultured as previously described (Gemble et al., 2015).

Deoxycytidine (dC) was provided by Sigma Aldrich (D0779); Camptothecin (CPT) was provided by Sigma Aldrich (C9911); CHIR-124 was provided by Selleckchem (S2683); MK-1775 was provided by Selleckchem (S1525); VE-821 was provided by Selleckchem (S8007); volasertib was provided by Selleckchem (S2235); KU55933 was provided by Selleckchem (S1092) and RO-3306 was provided by Calbiochem (217699). Drugs were added to the cell culture medium at the following concentrations: dC, 1 mM; camptothecin, 2 pM or 25 nM; CHIR-124, 400 nM; VE-821, 10 µM; volasertib, 5 nM; KU55933, 10 µM; RO-3306, 2.5 µM.

For the UVC experiment, we used the cells to seed six-well plates, which were incubated for 24 h before UVC exposure. Exponentially growing cells were irradiated in 1 ml PBS. Plates were placed uncovered under a UVC lamp (G25T8 germicidal lamp, Sankyo Denki, Japan) operating at 0.02 mW/cm². Dosimetry was performed with a VLX 3W radiometer (Vilber Lourmat, France) equipped with a 254 nm probe.

All cells were routinely checked for mycoplasma infection.

Cells were lysed in 8 M urea, 50 mM Tris HCl, pH 7.5 and 150 mM β-mercaptoethanol, sonicated and heated at 75°C for 10 min. Samples (equivalent of 2×10⁵ cells) were subjected to electrophoresis in NuPAGE Novex 4–12% Bis-Tris pre-cast gels (Life Technologies). The procedures used for gel electrophoresis and immunoblotting have been described elsewhere (Gemble et al., 2015), except that a Bio-Rad ChemiDoc XRS+ camera was used for detection. Primary and secondary antibodies used were as follows: rabbit anti-BLM antibody (1:5000; ab2179 from Abcam); rabbit anti-CDA antibody (1:500; ab56053 from Abcam); rabbit anti-β-actin antibody (1:10,000; Cat no. A2066 from Sigma); rabbit anti-PARP-1 antibody (1:4000; ALX-210-302 from Enzo Life Sciences); rabbit anti-Chk1 (1:500; sc-8408 from Santa Cruz Biotechnology); rabbit anti-Chk1-S317 (1:500; 2344 from Cell Signaling); mouse anti-H3S10 (1:500; 14955 from Abcam); mouse anti-Cdc25A (1:500; sc-7389 from Santa Cruz Biotechnology); rabbit anti-H3 (1:5000; ab1791 from Abcam); rabbit anti-cyclin-B1 (1:1000; sc-245 from Santa Cruz Biotechnology); mouse anti-GAPDH (1:500; G8795 from Sigma); rabbit anti-Chk2 (1:500; 2662 from Cell Signaling); rabbit anti-Chk2-T68 (1:500; 2661 from Cell Signaling); rabbit anti-H2AX (1:500; 2595 from Cell Signaling); rabbit anti-H2AX-S139 (1:500; 2577 from Cell Signaling) and horseradish-peroxidase-conjugated goat anti-rabbit-IgG or goat anti-mouse-IgG (1:5000; Cat nos. sc-2054 and sc-2055, respectively, from Santa Cruz Biotechnology).

We used 4×10⁵ HeLa cells to seed the wells of a six-well plate. The cells were left untreated or were treated as mentioned in the figures. They were then washed in PBS and fixed by incubation in 4% paraformaldehyde for 20 min. Nuclear DNA was detected by mounting slides in ProlongH Gold Antifade reagent supplemented with DAPI (Invitrogen), making it possible to distinguish between mitotic and interphase cells. Images were acquired with a Leica DM RXA microscope equipped with a motorized xy stage, using a 40× PlanApo N.A. 1.25 objective and a CoolSNAP HQ interline CCD camera (Photometrics). For each slide, a mosaic of 10×10 partly overlapping images was acquired with a Metamorph software (Molecular Devices) routine developed in-house. Image collections were assembled into a mosaic with the ‘Stitching 2D/3D’ plugin32 (available from http://fly.mpi-cbg.de/~preibisch/software.html) for ImageJ software (National Institutes of Health, Bethesda, Maryland). Mitotic index was calculated by measuring the percentage of mitotic cells in each mosaic.

Immunofluorescence staining and analysis were performed as previously described (Gemble et al., 2015). Primary and secondary antibodies used were as follows: rabbit anti-PICH (1:150; H00054821-D01P from Abnova); mouse anti-PICH (1:400; H00054821-M01 from Abnova); human anti-CREST (1:100; 15-234-0001 from Antibodies Inc); goat anti-human Alexa Fluor 633 (1:500; A21091 from Life Technologies); goat anti-rabbit Alexa Fluor 555 (1:500; A21429 from Life Technologies); goat anti-mouse Alexa Fluor 555 (1:500; A21050 from Life Technologies). Cell images were acquired with a 3D deconvolution imaging system consisting of a Leica DM RXA microscope equipped with a piezoelectric translator (PIFOC; PI) placed at the base of a 63× PlanApo N.A. 1.4 objective, and a CoolSNAP HQ interline CCD camera (Photometrics). Stacks of conventional fluorescence images were collected automatically at a z-distance of 0.2 mm (Metamorph software, Molecular Devices). Images are presented as maximum intensity projections, generated with ImageJ software, from stacks deconvolved with an extension of Metamorph software (Savino et al., 2001). EdU incorporation into DNA was visualized with the Click-it EdU imaging kit (C10338 from Life Technologies), according to the manufacturer's instructions. EdU was incubated with cells at a concentration of 10 µM for 1 h.

We used 4×10⁵ HeLa cells to seed the wells of a six-well plate. The cells were left untreated or were treated as mentioned in the figures. They were then washed in cold PBS on ice and fixed by incubation in a 1:1 (vol/vol) mixture of methanol and acetone for 10 min on ice. After three washes in PBS–Tween (0.05%), cells were incubated overnight at 4°C with a mouse anti-PAR antibody [1:500; generously provided by Valérie Schreiber (Illuzzi et al., 2014)]. The cells were washed three times with PBS–Tween (0.05%) and incubated with a goat anti-mouse Alexa-Fluor-555-conjugated antibody (1:500; A21050 from Life Technologies) for 2 h. After two washes in PBS–Tween (0.05%), cells were mounted on slides with Prolong Gold with DAPI (P36931 from Life Technologies). Cell images were acquired with a 3D deconvolution imaging system consisting of a Leica DM RXA microscope equipped with a piezoelectric translator (PIFOC; PI) placed at the base of a 63× PlanApo N.A. 1.4 objective, and a CoolSNAP HQ interline CCD camera (Photometrics). Stacks of conventional fluorescence images were collected automatically at a z-distance of 0.2 mm (Metamorph software, Molecular Devices). Images are presented as maximum intensity projections, generated with ImageJ software, from stacks deconvolved with an extension of Metamorph software (Savino et al., 2001). PAR foci per nucleus were counted by a customized macro using a semi-automated procedure, as follows: the nucleus stack was first smoothed using a median filter (radius 5), the user defined an intensity value as a threshold (one value for all experiments); a mask was then generated and transferred onto the stack of foci so that only foci in nuclei were analyzed. A top-hat filter was applied to the result to eliminate local background and facilitate the segmentation process, based on a simple threshold (user-defined value). Finally, the macro counted and characterized foci. At least 500 nuclei were analyzed for each condition.

Cells were detached by treatment with Accutase (Sigma), immediately washed in 1× PBS, fixed in 70% ethanol and stored at −20°C overnight. They were then washed in 1× PBS and 1× staining buffer (554656 from BD Pharmingen). DNA content was visualized by incubating the cells with 7-AAD (559925 from BD Pharmingen) for 15 min at room temperature.

DNA synthesis was visualized with the Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (C-10419 from ThermoFisher Scientific), according to the manufacturer's instructions. EdU was used at a concentration of 10 µM for 1 h.

Cell cycle analysis or EdU incorporation was visualized with a FACSCalibur (Becton-Dickinson), by analyzing 50,000 cells per condition. Data were then analyzed with FlowJo software (Tree Star Inc.).

Cells were detached by treatment with Accutase (Sigma), immediately washed in 1× PBS, fixed in 70% ethanol and stored at −20°C overnight. They were then washed two times in 1× PBS and incubated for 1 h with an anti-histone 3 (phospho S10) antibody (1:200; 14955 from Abcam) in Stain Buffer (554656 from BD Pharmingen). Cells were washed in 1× PBS and incubated for 30 min with a goat anti-mouse Alexa-Fluor-633-conjugated antibody (1:500; A21052 from Life Technologies), propidium iodide (P4864 from Sigma) and RNAse A (R6513 from Sigma) in Stain Buffer at 37°C.

P-Histone 3 staining and DNA content were visualized with a FACSCalibur (Becton-Dickinson), by analyzing 10,000 cells per condition. Data were then analyzed with FlowJo software (Tree Star Inc.).

Subcellular analysis was performed with the Subcellular Protein Fractionation Kit for Cultured Cells (78840 from ThermoFisher Scientific), according to the manufacturer's instructions.

At least three independent experiments were carried out to generate each dataset and the statistical significance of differences was calculated with Student's t-test, as indicated in the figure legends.

We acknowledge the assistance of Marie-Noëlle Soler (PICT-IBiSA, Institut Curie, Orsay) and Charlène Lasgi (Flow Cytometry platform, Institut Curie, Orsay). We thank V. Schreiber and Jean-Christophe Amé (Institut de recherche de l'Ecole de biotechnologie de Strasbourg, Strasbourg) for providing us with anti-PAR antibodies.