Dynamic recruitment of licensing factor Cdt1 to sites of DNA damage

In eukaryotic cells, dynamic protein complexes regulate the maintenance of genomic stability at multiple levels. Multiprotein complexes assemble onto origins of replication to ensure that replication initiates once per cell cycle, preventing re-replication and safeguarding genomic stability (Arias and Walter, 2007; Blow and Dutta, 2005; Diffley, 2004). Similarly, highly dynamic protein complexes orchestrate the DNA damage response (DDR), which detects aberrant DNA structures and directs cellular responses for repair of a plethora of DNA lesions (Zhou and Elledge, 2000). Coordination of cell cycle regulatory mechanisms with the DDR is essential. DDR pathways must trigger cell cycle arrest in the presence of damage, while cell cycle regulatory mechanisms must signal to the DDR the phase of the cell cycle for the choice of an appropriate repair pathway (Branzei and Foiani, 2008). Factors linking these processes are therefore pivotal for the maintenance of genomic integrity.

The cell cycle regulatory protein Cdt1 has been postulated to link the cell cycle and DDR. Cdt1 regulates replication licensing by controlling the recruitment of mini-chromosome maintenance proteins (MCMs) onto origins of replication (Aparicio et al., 1997; Bell and Dutta, 2002; Blow and Dutta, 2005; Ishimi, 1997; Machida et al., 2005; Nishitani and Lygerou, 2004). Cdt1 is specifically expressed during the G1 phase of the cell cycle (Nishitani et al., 2001; Sakaue-Sawano et al., 2008; Wohlschlegel et al., 2000; Xouri et al., 2007b), ensuring that only after passage through mitosis does chromatin become competent for a new round of DNA replication. Ectopic expression of Cdt1 causes re-replication of DNA in Caenorhabditis elegans (Zhong et al., 2003), Drosophila (Thomer et al., 2004), Xenopus (Arias and Walter, 2005; Davidson et al., 2006; Li and Blow, 2005) and human cells (Vaziri et al., 2003), leading to DNA damage and activation of checkpoint pathways (Liontos et al., 2007; Vaziri et al., 2003). Interestingly, an additional link between Cdt1 and the DDR exists: Cdt1 is targeted for degradation in response to different types of DNA lesions, and this evolutionarily conserved response has been postulated to constitute a novel checkpoint (Higa et al., 2003; Hu et al., 2004; Kondo et al., 2004; Ralph et al., 2006). Cdt1 targeting upon DNA damage occurs independently of the classic DDR pathway mediated by ATM/ATR and CHK1/CHK2 kinases (Higa et al., 2003), but it requires Cdt1 ubiquitylation by the CUL4-DDB1 ubiquitin ligase (Higa et al., 2003; Hu et al., 2004; Nishitani et al., 2006; Senga et al., 2006). Cdt1 ubiquitylation has been shown to require interaction with PCNA in humans, Xenopus, Drosophila and fission yeast (Havens and Walter, 2009; Higa et al., 2006; Hu and Xiong, 2006; Nishitani et al., 2006; Senga et al., 2006) and with the DDB1- and CUL4-associated factor (DCAF) Cdt2 in humans, Xenopus, zebrafish and fission yeast (Higa et al., 2006; Jin et al., 2006; Ralph et al., 2006; Sansam et al., 2006). Cdt2 seems to act as an adaptor bridging the CUL4-DDB1 ubiquitin ligase with its substrates (Lee and Zhou, 2007).

Functional live-cell imaging permits assessment of the dynamics of complex biomolecular interactions within the context of the living cell. Our previous studies showed that Cdt1 exhibits dynamic interactions with chromatin throughout the G1 phase of the cell cycle in live cells, suggesting that Cdt1 constantly scans chromatinduring G1 (Xouri et al., 2007a; Xouri et al., 2007b). In the present study, we have used a micro-irradiation system that induces localized DNA damage in a specific subnuclear volume of the living cell, to investigate the spatio-temporal regulation of Cdt1 following DNA damage. We show that Cdt1 is recruited to sites of DNA damage with fast kinetics. Cdt1 recruitment to damaged DNA takes place during the G1 phase and is a highly dynamic process that requires Cdt1 interaction with PCNA and precedes Cdt1 degradation.

In order to address the dynamics of Cdt1 regulation in response to DNA damage, we developed a method of inducing spatially restricted DNA damage in a defined subnuclear volume of the living cell. A laser nanosurgery microscope (based on an UVA-pulsed laser), designed to sever biological materials in vivo with high precision (Colombelli et al., 2005), was used to induce local DNA damage. Following exposure to low laser power (supplementary material Fig. S1A), a typical cellular response to DNA double strand breaks (DSBs) was locally activated, verified by the presence at the sites exposed to the laser of proteins involved in DSB recognition and repair. In particular, histone H2AX phosphorylated at serine 139 (γH2AX) and kinase ATM phosphorylated at serine 1981 (Bakkenist and Kastan, 2003; Rogakou et al., 1999) were found at the subnuclear site exposed to the laser (Fig. 1A), as well as proteins involved in DSB-repair pathways through homologous recombination (Mre11, Fig. 1B, and BRCA1, see later) and non-homologous end-joining (Ku70, Fig. 1B). XRCC1, a protein involved in single strand break (SSB) repair (Horton et al., 2008), was also recruited to the site of damage, as previously shown for other types of laser-induced damage (Lan et al., 2004; Mortusewicz and Leonhardt, 2007), implying the presence of SSBs at laser-exposed sites (supplementary material Fig. S1B, upper panel). In addition, TFIIH, a protein involved in nucleotide excision repair (Mone et al., 2001), was also present at the sites of laser-induced damage (supplementary material Fig. S1B, lower panel), suggesting that UV-induced photoproducts are also generated by laser treatment. In contrast to previous protocols for laser-induced DNA damage, pretreatment with DNA sensitizing agents was not necessary for the induction of damage (Lukas et al., 2003; Rogakou et al., 1999; Tashiro et al., 2000), and no nuclear damage was observed either in the DIC images or after staining with Hoechst 33258 (Kim et al., 2005; Kim et al., 2002) (Fig. 1C; supplementary material Fig. S1A). Time-lapse experiments showed that the induced damage did not interfere with cellular viability (supplementary material Fig. S1E). The sliding clamp PCNA was recruited to the site of laser-induced damage in both G1 and S phases of the cell cycle (Fig. 1D; supplementary material Fig. S1C), as previously shown for laser-induced and UV-induced DNA damage (Essers et al., 2005; Lan et al., 2004; Mortusewicz et al., 2005; Perucca et al., 2006; Solomon et al., 2004). PCNA accumulation at the site of damage in G1 reached a maximum 5–10 minutes after laser treatment, and was decreased 40 minutes later (supplementary material Fig. S1C), as previously proposed for RFP-tagged PCNA (Mortusewicz et al., 2005). Protein p53 was not recruited at the site of damage, consistent with previous studies (Bekker-Jensen et al., 2006), but was specifically upregulated in damaged cells (Fig. 1E), suggesting that the localized laser-induced DNA damage was sufficient to activate global checkpoint responses. Three-dimensional analysis of the damaged region showed that the damage was localized only around the laser focus, demonstrating that laser-induced damage could be spatially controlled with high precision, minimizing the volume of the nucleus that was exposed to the damage (supplementary material Fig. S1D).

To investigate the spatial and temporal regulation of Cdt1 upon localized DNA damage, MCF7 cells were exposed to laser-induced damage and subjected to immunofluorescence using antibodies against Cdt1 and the DSB markers γH2AX, Ku70, BRCA1 and pATM. A robust recruitment of Cdt1 to the sites of the laser-induced damage was observed, colocalizing with DSB markers (Fig. 2A,B). Cdt1 recruitment to the site of the laser-induced damage was evident only in G1 cells, as shown using cyclin A expression as a cell-cycle-stage marker in asynchronous cells (Fig. 2A) and by synchronization experiments (supplementary material Fig. S2), consistent with the G1-phase-specific expression of Cdt1 (Nishitani et al., 2001; Sakaue-Sawano et al., 2008; Wohlschlegel et al., 2000; Xouri et al., 2007b). Cdt1 recruitment to the damage site was evident in HeLa cells (Fig. 2C) and in primary fibroblasts (Fig. 2D), suggesting that it is a general cellular response. In order to address whether Cdt1 is recruited to the site of damage following other types of DNA damage, we UV-irradiated a confirmed area of individual cell nuclei using a polycarbonate filter containing pores of defined size (Volker et al., 2001) and analysed the localization of Cdt1 and UV-induced cyclobutane pyrimidine dimers (CPDs) by immunofluorescence. Cdt1 accumulates at sites of UV-induced CPDs (Fig. 2E), showing that Cdt1 is recruited to UV-damaged DNA, similar to recruitment to laser-damaged DNA.

To address the kinetics of Cdt1 accumulation at the site of laser-induced damage in vivo, we performed time-lapse microscopy of cells expressing Cdt1 tagged with the green fluorescent protein (GFP) at its C-terminus (Cdt1GFP). Cdt1GFP adopts the same localization and cell-cycle-specific expression as endogenous Cdt1 (Nishitani et al., 2001; Xouri et al., 2007b) and is degraded upon UV irradiation (supplementary material Fig. S3A), reproducing the behaviour of endogenous Cdt1.

MCF7 cells transiently expressing Cdt1GFP or GFP alone as control were subjected to laser micro-irradiation. As shown in Fig. 3A (see also supplementary material Movie 1), robust accumulation of Cdt1GFP to the sites exposed to the laser was observed within 10 seconds following micro-irradiation and persisted through the timescale of the experiment (10 minutes). To more precisely quantify Cdt1 recruitment to sites of damage, MCF7 cells stably expressing Cdt1GFP at levels similar to endogenous Cdt1 (Xouri et al., 2007b) were micro-irradiated and followed by time-lapse microscopy for 1 hour. The fluorescence intensities at the site of damage and the total nuclear region were calculated through the time-course of the experiment. Cdt1 accumulated at the site of damage seconds after micro-irradiation and persisted throughout the experiment. A decrease in mean intensity at the site of damage was detectable after approximately 45 minutes (Fig. 3B). We then assessed the kinetics of Cdt1 accumulation at sites of damage in HeLa cells, which exhibit fast kinetics of Cdt1 degradation following DNA damage (Higa et al., 2003; Hu et al., 2004; Nishitani et al., 2006; Sansam et al., 2006). A HeLa cell line stably expressing Cdt1GFP at levels similar to endogenous Cdt1 (data not shown) was constructed and subjected to micro-irradiation. Cdt1 was recruited at the site of damage prior to its degradation (Fig. 3C). Similar results were obtained upon UV-induced localized-DNA damage (supplementary material Fig. S3B). Collectively, these results indicate that Cdt1 is recruited within seconds at sites of DNA damage in G1 cells, and recruitment precedes its degradation.

To investigate requirements for Cdt1 recruitment to sites of DNA damage, we tested the ability of mutated forms of Cdt1 to accumulate to micro-irradiated subnuclear regions. A mutant with six alanine substitutions at the N-terminal PIP box motif of Cdt1 (A6Cdt1) (see Nishitani et al., 2006), required for interaction with PCNA and Cul4/DDB1-dependent degradation, was compared to a mutant with alanine substitutions of the Cy 68-RRL-70 motif (CyCdt1), required for interactions with cyclin A and/or cyclin E and SCF/Skp2-dependent Cdt1 degradation after S-phase onset (Liu et al., 2004; Nishitani et al., 2006; Sugimoto et al., 2004). As CyCdt1 is mutated in the nuclear localization sequence (nls) of Cdt1, SV40 nls sequences were added before the GFP in all constructs. Wild-type Cdt1, A6Cdt1, CyCdt1 and the double mutant A6CyCdt1, all tagged with nlsGFP, as well as nlsGFP itself as control, were transfected in MCF7 cells and their abilities to accumulate to sites of damage were assessed by time-lapse microscopy following micro-irradiation. As shown in Fig. 4A, A6Cdt1nlsGFP and the double mutant A6CyCdt1nlsGFP lost the ability to accumulate to sites of micro-irradiation, suggesting that interactions with PCNA are required for Cdt1 recruitment to the damaged site. By contrast, CyCdt1nlsGFP accumulated at the site of DNA damage similarly to wild-type Cdt1, suggesting that Cy-motif-mediated interactions with cyclins A/E are dispensable for Cdt1 recruitment to DNA damaged sites. Consistently, coexpression of Cdt1GFP and PCNA–RFP in the same cell revealed that these proteins were rapidly recruited in parallel to the site of damage. By contrast, an A6Cdt1GFP mutated form was not recruited to the DNA damaged site (Fig. 4B; supplementary material Movies 2 and 3).

To confirm that Cdt1 recruitment to the sites of damage requires PCNA, we knocked down PCNA expression in MCF7 cells by 60% using short interfering RNA (siRNA) (Fig. 4C, lower panel), and tested whether the ability of transiently expressed Cdt1GFP to be recruited to the site of the laser-induced damage depended on the presence of PCNA. As shown in Fig. 4C (upper panel), both Cdt1 and PCNA accumulated at the site of damage in non-target siRNA-treated cells. By contrast, in cells treated with siRNA for PCNA, Cdt1 accumulated at the site of damage only in cells where PCNA was present at the site of damage, confirming that Cdt1 recruitment is PCNA-dependent. Taking these data together, we conclude that Cdt1 rapidly accumulates to the site of DNA damage through interactions with PCNA.

Cdt1 targeting for degradation upon DNA damage requires the Cul4A-Ddb1 ubiquitin ligase and the DCAF subunit Cdt2 (Higa et al., 2006; Jin et al., 2006; Ralph et al., 2006; Sansam et al., 2006). We next tested the behaviour of Cdt2 in response to laser-induced localized DNA damage. MCF7 cells coexpressing Cdt1 tagged with mCherry and Cdt2 tagged with GFP were micro-irradiated. As shown in Fig. 5A and in supplementary material Movie 4, Cdt2 rapidly accumulated at the site of damage, in parallel with Cdt1. Similarly, the endogenous protein Cdt2 was recruited at the site of damage, colocalizing with the DSB marker γH2AX (supplementary material Fig. S4A). We then constructed GFP-tagged forms of Cul4A and Ddb1 and proceeded to time-lapse microscopy following micro-irradiation. GFP–Cul4A was detected at the site of damage in living cells (supplementary material Fig. S4C, upper panel), as was the endogenous Cul4A protein (supplementary material Fig. S4B). Similarly to Cul4A, DDB1 was also present at the site of laser-induced damage in living cells expressing GFP–DDB1 (supplementary material Fig. S4C, lower panel), suggesting that the subunits of Cul4A-Ddb1^Cdt2 ubiquitin ligase that targets Cdt1 for degradation in response to DNA damage are recruited to the site of damage.

We next addressed whether PCNA and Cdt2 can be recruited at the site of damage in the absence of Cdt1. MCF7 cells were treated with siRNA for Cdt1, and the ability of PCNA (Fig. 5B) and Cdt2–GFP (Fig. 5C) to be recruited at the site of the laser-induced DNA damage was assessed. Although the treatment of cells with siRNA for Cdt1 downregulated Cdt1 expression effectively (supplementary material Fig. S5A,B), both PCNA and Cdt2–GFP could be recruited to the site of damage in the absence of Cdt1 (Fig. 5B,C). These data suggest that both PCNA and Cdt2 recruitment to the site of damage is Cdt1-independent. The DNA repair factors Ku80 and XRCC1 appeared similarly unaffected by Cdt1 depletion (supplementary material Fig. S5C). Likewise, Cdt1 was still detected at sites of damage in cells depleted of Cdt2 by RNAi (supplementary material Fig. S6), indicating that Cdt2 is not essential for Cdt1 recruitment.

We then addressed the recruitment kinetics of Cdt1, Cdt2 and PCNA accumulation in the damaged region in living cells. In parallel, we explored the recruitment kinetics of the cyclin kinase inhibitor p21^Cip1, which was previously shown to accumulate to the site of damage through interaction with PCNA (Perucca et al., 2006) and to be proteolysed through Cul4-Ddb1^Cdt2, similarly to Cdt1 (Abbas et al., 2008; Kim et al., 2008; Nishitani et al., 2008). To this end, MCF7 cells expressing fused forms of the proteins tagged with GFP were micro-irradiated, and time-lapse microscopy used to record their recruitment over 15 minutes. Quantification of protein accumulation at the site of damage with respect to time revealed that Cdt1, PCNA, Cdt2 and p21^Cip1 started to accumulate at the site of damage only some seconds (approx. 5–10 seconds) after the induction of damage (Fig. 5D). PCNA and p21^Cip1 showed parallel accumulation profiles and reached maximum accumulation approximately 400 seconds after damage induction. Cdt1 accumulation kinetics, however, differed from PCNA and p21^Cip1, with Cdt1 reaching a maximum accumulation at the site of damage much earlier, approximately 100 seconds after the laser treatment. Cdt2 continued to accumulate after Cdt1 had reached its maximum levels, up to 200 seconds after damage induction. These findings suggest that the binding sites of PCNA/p21^Cip1, Cdt2 and Cdt1 at the site of damage differ with respect to saturation time. They further indicate that increased PCNA levels at the site of damage at later time-points are not sufficient to recruit additional Cdt1.

Time-lapse analysis allowed us to compare the extent and timing of accumulation of Cdt1, PCNA, Cdt2 and p21^Cip1 at the site of damage. It is, however, not suitable for addressing the mobility of the molecules once bound to the site of damage. In order to address the dynamics of Cdt1, PCNA, Cdt2 and p21^Cip1 at the site of damage, we employed fluorescence recovery after photobleaching (FRAP). MCF7 cells transfected with GFP-tagged proteins were micro-irradiated to induce local DNA damage in a stripe pattern. After binding equilibrium was reached at the site of damage, FRAP was performed by bleaching a rectangular area placed over the DSB-containing nuclear stripe, or in a neighbouring cell's nucleoplasm that had not been damaged (Fig. 6A), and then monitoring fluorescence recovery over time (Fig. 6B).

Qualitative analysis of the acquired data revealed a highly dynamic behaviour of these proteins in the absence of damage but distinct dynamics at the site of damage, with Cdt1 and p21^Cip1 showing fast recoveries in contrast to markedly reduced total recoveries of PCNA and Cdt2 (Fig. 6C). To obtain quantitative binding parameters from the photobleaching experiments, we analysed fluorescence recovery curves using an exactly solvable model that assumed a simplified rectangular geometry (with one relevant spatial dimension). The model accounted for uniform binding sites scattered throughout the nucleus, and explicitly specified additional binding sites introduced by the damage (for detailed description see Materials and Methods and supplementary information available at http://edoc.mpg.de/display.epl?mode=doc&id=521078&col=18&grp=125). The model reproduced accurately the observed recoveries and allowed quantitative discrimination of the contributions arising from diffusion versus binding interactions (supplementary material Fig. S7). In non-damaged cells, all recoveries were very rapid (within seconds) and consistent with a ‘diffusive’ recovery, which was specified in the model by an effective diffusion constant, D_eff (Fig. 6C).

In laser-micro-irradiated cells expressing Cdt1GFP, a large pool (approximately 80%) of the Cdt1 molecules that accumulated to the site of damage was highly dynamic, with a robust upper limit to the residence half-life of 1 second. A smaller immobile part (roughly 20%) at the damage site had an estimated residence time in the order of minutes (Fig. 6C). The rapid recovery of Cdt1, along with its observed accumulation at the site of damage, implied that 90% of Cdt1 molecules in the nucleus will bind at the site of damage within a time period of less than 2 minutes (see supplementary information at http://edoc.mpg.de/display.epl?mode=doc&id=521078&col=18&grp=125 for description). Similarly to Cdt1, over 90% of p21^Cip1 molecules were able to recover with an upper limit to the residence half-life of 3 seconds. By contrast, our analysis revealed that Cdt2 molecules exist in two subpopulations, with only approximately 40% able to rapidly recover and 60% of Cdt2 molecules highly immobile at the damaged region. Strikingly, PCNA molecules were stably bound at the damaged site with a robust lower limit to the residence half-life of PCNA molecules at the damaged site of approximately 20 minutes, consistent with previous findings (Essers et al., 2005; Mortusewicz and Leonhardt, 2007; Solomon et al., 2004). The long residence half-life of PCNA at the site of damage was further supported by following FRAP kinetics of PCNA for over 30 minutes (supplementary material Fig. S8). Taking these data together, we conclude that PCNA provides a stable binding interface at the site of damage for dynamic interactions of Cdt1 and p21^Cip1. The stable binding of a large fraction of Cdt2 molecules might facilitate robust modification of multiple substrates and further suggests that Cdt2 does not move as a complex with Cdt1 at the site of damage.

Here, we have investigated the spatial and temporal regulation of the replication protein Cdt1 in response to DNA damage. We combined spatially restricted DNA damage by laser micro-irradiation with functional imaging to acquire quantitative information on Cdt1 behaviour in the presence of DNA damage. We show that Cdt1 is rapidly recruited at the site of DNA damage within the living cell. Our data suggest that dynamic interactions with a stably bound PCNA scaffold might facilitate recruitment and coordination of the cell cycle and DDR.

We have characterized a method of inducing DNA damage in locally predefined regions of the living cell with minimal invasion to cell physiology. Several experimental results support this notion. Laser treatment did not produce detectable nuclear damage, in contrast to previously described methods (Kim et al., 2005; Kim et al., 2002). Damage was localized only around the laser focus, minimizing the volume of the cell exposed to the damage. Consistently, cell viability was not affected in long time-lapse experiments. In addition, laser-induced damage was generated without pretreatment of cells with DNA-sensitizing agents that could affect the DDR (Lukas et al., 2005). We showed that our system induces a typical response to DSB formation because factors involved in recognition and repair of DSBs were rapidly recruited to the damage site. In addition, single strand breaks and UV-induced photo-products were present at laser-exposed sites, as evident by recruitment of repair factors for these lesions. The stabilization of p53 protein levels only in laser-treated cells suggests that the laser-induced damage is sufficient to elicit global checkpoint responses. Thus, the system used in this study induces spatially localized DNA damage that activates a typical response to DSB formation with minimal perturbation to cellular physiology, and therefore can be used to investigate regulation of the DDR within the context of the living cell.

We have shown that Cdt1 rapidly accumulates at the site of DNA damage. Cdt1 recruitment occurred in both MCF7 and HeLa tumor cell lines upon laser-induced and UV-induced localized DNA damage as well as in laser-treated primary cells, suggesting that Cdt1 accumulation at the damage site is a general feature of human cells upon DNA damage. We report that Cdt1 recruitment to the site of damage requires interaction with PCNA. Mutated forms of Cdt1 that have previously been shown not to interact with PCNA (Nishitani et al., 2006) lost the ability to accumulate at the site of damage. Knocking down the expression of PCNA using siRNA showed that Cdt1 recruitment to the site of damage was dependent on the presence of PCNA.

We employed FRAP to show that PCNA is stably bound at the site of damage, consistent with previous studies (Essers et al., 2005; Mortusewicz and Leonhardt, 2007; Solomon et al., 2004). By sharp contrast, the majority of Cdt1 molecules at the site of damage exhibited dynamic binding with residence times of less than 1 second. We observed a similar recruitment with dynamic binding for another PCNA-interacting protein, p21^Cip1. Several studies report a striking ability of PCNA to interact with multiple binding partners involved in DNA replication and DNA repair, and to control cellular functions through the coordination of these interactions (Maga and Hubscher, 2003; Moldovan et al., 2007). Similarly to Cdt1, many PCNA-interacting proteins contain a common PCNA-interacting protein motif (PIP box), are believed to share common interfaces for interaction on the PCNA homotrimeric structure (Maga and Hubscher, 2003; Moldovan et al., 2007; Warbrick, 2000), and are recruited to sites of DNA damage through interaction with PCNA (Moggs et al., 2000; Mortusewicz et al., 2006; Mortusewicz et al., 2005; Perucca et al., 2006). The recruitment of multiple proteins through the same binding interface on PCNA might be facilitated by rapid interactions to a stably bound PCNA trimer. In addition, co-regulation of their access to PCNA and competition for the same binding regions might contribute to fine-tuning their recruitment and function. Cdt1 recruitment at the site of damage might regulate or be regulated by the recruitment of other PCNA-interacting proteins. In support of this idea, a synthetic peptide derived from p21^Cip1 containing the PIP box motif affects the direct binding of Cdt1 to PCNA in a Xenopus in vitro system (Arias and Walter, 2006), and overexpression of the p21^Cip1 or p57^Kip2, but not their non-PCNA-interacting mutants, blocks Cdt1 degradation in mammalian cells upon UV-irradiation (Hu and Xiong, 2006).

We showed that both Cdt1 and PCNA are recruited to the site of damage in parallel, within a few seconds of the induction of damage. However, quantification of recruitment kinetics revealed that Cdt1 accumulation reached a plateau much earlier than PCNA, suggesting that additional molecules of PCNA accumulating a few minutes after the damage are not sufficient to recruit additional Cdt1 molecules. It is unlikely that rapid degradation of Cdt1 at the site of damage balances additional Cdt1 recruitment because total levels of Cdt1 in the nucleus were constant over the time course of the experiment (data not shown) and Cdt1 levels at the site of damage did not decrease after PCNA reached a plateau. Temporally or spatially regulated changes in Cdt1–PCNA binding affinity could be a possible reason for the observed differences. Changes in the post-translational modification status of PCNA or its interacting proteins have previously been reported to alter binding affinities with its multiple partners (Henneke et al., 2003). Alternatively, a factor required for Cdt1 accumulation in addition to PCNA might become limiting, or a factor inhibiting Cdt1 recruitment might accumulate at the site of damage.

Cdt2 belongs to the recently discovered DCAF protein family comprising members that, as parts of the CUL4-Ddb1 ubiquitin ligase complex, provide specificity to the ligase for different substrates (Angers et al., 2006; He et al., 2006; Higa et al., 2006; Jin et al., 2006). Cdt2 is required for the ubiquitylation and subsequent degradation of Cdt1 (Higa et al., 2006; Jin et al., 2006; Ralph et al., 2006; Sansam et al., 2006) and p21^Cip1 (Abbas et al., 2008; Kim et al., 2008; Nishitani et al., 2008), both after onset of S phase and upon DNA damage.

We showed that upon localized DNA damage, Cdt2 is rapidly recruited to the site of damage, predicting a possible involvement of Cdt2 in DDR pathways. In support of this idea, in cells treated with ionizing radiation, Cdt2 is required for the activation of the G2–M checkpoint of the cell cycle in human and zebrafish (Sansam et al., 2006). We also detected Cul4 and DDB1 at the site of laser-induced damage, suggesting that an active CUL4-Ddb1^Cdt2 ubiquitin ligase complex might assemble.

In Xenopus egg extracts, Cdt1 ubiquitylation by the CUL4-Ddb1^Cdt2 ubiquitin ligase takes place on chromatin and requires interactions with chromatin-bound PCNA through the PIP box motif of Cdt1 (Arias and Walter, 2006; Havens and Walter, 2009; Jin et al., 2006). Cdt1 binding to PCNA seems to be required for the subsequent recruitment of Cdt2 and Ddb1 onto chromatin (Jin et al., 2006). Consistently, our data suggest that upon laser-induced damage, PCNA acts as a scaffold for Cdt1 recruitment at the site of damage, Cdt1 is not required for PCNA recruitment, and Cdt2 is not required for Cdt1 recruitment. However, in contrast to in vitro experiments, our in-cell analysis suggests that Cdt2 recruitment at the site of damage is independent of Cdt1 because siRNA-mediated depletion of Cdt1 did not prevent Cdt2 recruitment to the sites of damage. In addition, quantitation of recruitment kinetics showed that Cdt2 continues to accumulate on chromatin after Cdt1 reaches saturation. FRAP analysis showed highly dynamic binding of the majority of Cdt1 molecules at the site of damage, in contrast to the slower exchange observed for the majority of Cdt2 molecules. These data suggest that Cdt1 and Cdt2 can bind to damaged chromatin independently of each other. Our analysis does not exclude that a fraction of Cdt1 will contact Cdt2 on chromatin. The use of Cdt1 mutants able to bind PCNA but not Cdt2 (Havens and Walter, 2009) will hopefully help unravel Cdt1–Cdt2 interactions at the site of damage. In living cells, Cdt2 might be recruited to sites of damage by multiple pathways. Indeed, at least one more CUL4-Ddb1^Cdt2 substrate, p21^Cip1, accumulates at the site of damage in live cells. However, as p21^Cip1 exhibits highly dynamic behaviour at the site of damage, similar to that of Cdt1, interactions in addition to substrate binding might be required to stabilize Cdt2 at sites of damage.

Our experiments in living cells clearly demonstrated that, in response to laser- or UV-induced DNA damage, Cdt1 is recruited to sites of damage followed by its degradation (Fig. 7). Why is Cdt1 targeted to the site of damage? One possibility is that Cdt1 recruitment to the site of DNA damage facilitates its ubiquitylation by the CUL4A-Ddb1^Cdt2 ubiquitin ligase and its subsequent proteasomal degradation. We showed that Cdt1 recruitment coincides with the accumulation of the CUL4A-Ddb1 ubiquitin ligase and the adaptor protein Cdt2 to the site of damage. Moreover, we demonstrated that Cdt1 recruitment to the damaged site necessitates interaction with PCNA stably bound at the site of damage, which is also indispensable for its targeting for degradation. Earlier studies reported accumulation of conjugated ubiquitin in ionizing-radiation-induced nuclear foci (Morris and Solomon, 2004; Polanowska et al., 2006) and in laser-micro-irradiated tracks (Mailand et al., 2007), suggesting that ubiquitin-related events take place at repair centres. The observed recruitment of CUL4A-Ddb1^Cdt2 ubiquitin ligase, PCNA and Cdt1 at the damaged site leads to an increase in their local concentration that might enhance the efficiency of Cdt1 ubiquitylation. The highly dynamic behaviour of Cdt1 molecules at the site of damage predicts that a large pool of Cdt1 molecules is able to interact with PCNA over a short time period, which could facilitate rapid and processive poly-ubiquitylation by CUL4A-Ddb1^Cdt2. Although there are no available data showing specific ubiquitylation of Cdt1 molecules at the site of damage, the ubiquitylation of Cdt1 molecules by CUL4A-Ddb1^Cdt2 onto chromatin in Xenopus extracts (Arias and Walter, 2006; Havens and Walter, 2009; Jin et al., 2006) and the multiple targets of CUL4A-Ddb1 ubiquitin ligase onto chromatin in response to DNA damage (Kapetanaki et al., 2006; Sugasawa et al., 2005; Wang et al., 2006), are consistent with this notion. Local modification of CUL4A-Ddb1^Cdt2 substrates at the sites of damage could facilitate specificity for the presence of damage and co-regulation of multiple substrates.

However, the observed fast recruitment of Cdt1 to the site of damage before detectable degradation might indicate a more direct role of Cdt1 in the DDR that would precede its degradation. FRAP data analysis and modelling revealed that almost 90% of the Cdt1 molecules will bind at the site of damage within a time period of less than 2 minutes. If Cdt1 binding to the damaged sites was the primary trigger for Cdt1 breakdown, Cdt1 degradation on the same timescale would be expected. By contrast, we observed that Cdt1 is proteolytically degraded with much slower kinetics, suggesting that Cdt1 recruitment to damaged sites is not immediately followed by its degradation. In a similar way, recruitment of another CUL4A-Ddb1^Cdt2 target, p21^Cip1 to damaged sites has been suggested to serve direct functions in the DDR prior to its degradation (Prives and Gottifredi, 2008). Similarly, DDB2 recruitment to UV-induced damaged sites is required for XPC recruitment and repair, before its ubiquitylation by CUL4A-Ddb1 and its subsequent degradation (El-Mahdy et al., 2006; Luijsterburg et al., 2007; Sugasawa et al., 2005). We showed that Ku80 and XRCC1 (required for the repair of DSBs and SSBs, respectively, during the G1 phase) can still be recruited to sites of damage in the absence of Cdt1, making it unlikely that Cdt1 is required for initial damage recognition of these lesions. In addition, no defects in the phosphorylation of γH2AX and ATM were detected in Cdt1-depleted cells (data not shown). Cdt1 interacts with multiple proteins, including histone-modifying activities (Iizuka et al., 2006; Miotto and Struhl, 2008), which could be targeted to the site of damage through interactions with Cdt1. In addition, Cdt1 is specifically expressed during the G1 phase of the cell cycle and its recruitment at the site of damage could signal to the DDR the phase of the cell cycle for the choice of the appropriate repair pathway. For example, the choice between homologous recombination and non-homologous end-joining in repairing DSBs has been shown to be dependent on the cell cycle phase (Branzei and Foiani, 2008; Huertas and Jackson, 2009; Limbo et al., 2007; Sartori et al., 2007; Shrivastav et al., 2008). Future experiments will address whether the Cdt1 recruitment to sites of damage shown here could be important for linking the cell cycle to the DDR pathways.

MCF7, HeLa cells and primary fibroblasts were grown in DMEM with 10% fetal bovine serum at 37°C and 5% CO₂. HeLa cells stably expressing Cdt1GFP were prepared using Cdt1–GFP expression plasmid as described previously for the stable MCF7 cells (Xouri et al., 2007b). For transient transfection, cells plated in 35-mm dishes were transfected with 1 μg of plasmid DNA for 22 hours using FuGENE 6 (Roche).

For live-cell experiments, cells were plated on MatTek dishes (MatTek Corporation) in phenol red-free, CO₂-independent medium (Invitrogen). For synchronization experiments, MCF7 cells were treated with 90 ng/mL nocodazole for 11 hours and mitotic cells harvested by shake off. Mitotic cells were plated and left to proceed to G1 as indicated.

siRNA oligos against Cdt1, PCNA, Cdt2 or non-targeted siRNA (Dharmacon smart pools; Dharmacon, Lafayette, CO) were transfected using Dharmafect 2 siRNA transfection reagent. After 48 hours, cells were transfected with the appropriate DNA plasmids and analysed 22 hours following the second transfection.

UV-irradiation using isopore filters of 5 μm diameter (Millipore) was performed as previously described (Volker et al., 2001).

DNA damage was induced using a frequency tripled Nd:YAG-pulsed laser (JDS Uniphase, Grenoble, France), at a wavelength of 355 nm and a theoretical pulse duration of less than 500 picoseconds. The beam was coupled through the epifluorescence port of an Axiovert 200M Zeiss inverted microscope (Carl Zeiss, Göttingen, Germany) to allow simultaneous irradiation and imaging, and was focused through a Zeiss C-Apochromat 63× 1.2 NA water immersion objective lens. The micro-irradiations were performed along a laser line target, graphically defined with the mouse on the live window and along which 70–80 pulses were distributed at up to 500 Hz. Reproducible results were obtained by repeating this irradiation five times, the total time for irradiation did not exceed 1.5 seconds; irradiation happened during acquisition. Optimal energy per pulse was measured at approximately 100 nJ before the objective lens across the whole study (corresponding to 220 a.u. in supplementary material Fig. S1). Laser parameters were: 470 picoseconds theoretical pulse width, 200 nJ at 500 Hz corresponds to about 0.2 mW average power, i.e. 2% of the total laser power (HighPowerChip, JDS Uniphase, Grenoble, France). Arbitrary units for laser power in supplementary material Fig. S1 do not correspond linearly to laser power values due to a non-linear acousto-optical device response curve. Damage is most probably caused by material ionization below the threshold of plasma formation (Vogel and Venugopalan, 2003) due to the short pulse duration (470 picoseconds) (Colombelli et al., 2005). Typically, 50–100 cells were micro-irradiated within less than 10 minutes and returned for incubation for the time indicated before fixation.

Correlative fluorescence live imaging and immunofluorescence was performed as previously reported (Colombelli et al., 2008). Briefly, laser ablation at very high energies (>2 μJ per pulse) is used to inscribe cell location by plasma formation within the glass volume of the coverslip, approximately 20 μm below the studied sample. This technique permits a field of view to be subsequently located on any microscope under transmission contrast.

Simultaneous irradiation and live imaging of cells was performed on the same microscope equipped with a temperature-controlled incubator (37°C) and a Hamamatsu ORCA CCD camera (Hamamatsu Photonics KK, Hamamatsu City, Japan).

Cells were fixed in 4% paraformaldehyde and permeabilized with 0.3% Triton X-100 or alternatively using −20°C methanol. Primary antibodies used were: anti-Cdt2 antibody raised in rabbits against the bacterially expressed and purified 6×His-tagged C-terminal 150 amino acids of human Cdt2, α-Cdt1 was described previously (Nishitani et al., 2001), α-CPDs (Kamiya Biomedicals), α-γH2AX (Upstate), α-ATM (phospho-ATM at serine 1981; Rockland), α-MRE11 (Gene Tex), α-BRCA1 (Santa Cruz Biotechnology), a-Ku70 (Lab Vision), α-PCNA (Santa Cruz Biotechnology), α-CUL4A (Rockland), α-XRCC1 (Abcam), α-TFIIH (Santa Cruz Biotechnology). DNA was stained with Hoechst 33258 or DAPI.

Cdt1GFP, Cdt1nlsGFP, CyCdt1nlsGFP, A6Cdt1GFP, A6Cdt1nlsGFP A6CyCdt1nlsGFP constructs were as described previously (Xouri et al., 2007b). Constructs containing three copies of the SV40 nuclear localization sequences prior to the GFP (nlsGFP) were used when studying CyCdt1 and A6CyCdt1 mutants (Fig. 4A), which bear mutations in the nuclear localization sequence of Cdt1 and fail to correctly localize to the nucleus. Constructs without exogenous nls sequences were used for all other experiments.

PCNA tagged to mRFP (mRFP–PCNA) and to GFP (GFP–PCNA) were kindly provided by Cristina Cardoso (Leonhardt et al., 2000; Mortusewicz et al., 2005). Cdt1–mCherry was constructed by replacing GFP in the Cdt1GFP construct by mCherry (Nishitani et al., 2001). To construct GFP–Cdt2, full-length Cdt2 was cloned into pEGFP–C1 (Clontech) at the BglII-HindIII sites. Similarly, full-length CUL4A was subcloned from pcDNA3.1–HA–Cul4A (Nishitani et al., 2008) to pcDNA3.1–GFP (Invitrogen) at sites EcoRI-XhoI. GFP–DDB1 was constructed by subcloning full-length FLAG–DDB1 to pEGFP–C2 (Clontech). In addition, full-length p21^Cip1 was subcloned from 3×FLAG–p21^Cip1 (Nishitani et al., 2008) to pEGFP–C1 vector (Clontech) between EcoRI-BamHI restriction sites.

Mean fluorescence intensities were quantified using ImageJ 1.37g (Wayne Rasband, National Institute of Mental Health, Bethesda, MD). Where required, cell motility was corrected using ImageJ plug-in MultiStackReg (Thevenaz et al., 1998). For quantifications described in Fig. 3B,C, mean intensity per pixel at the site of damage or for total nucleus in each damaged cell were normalized at each time point as follows: background mean intensity per pixel measured in a region devoid of cells was subtracted; the resulting values were divided with the mean intensity per pixel in the nucleus of a neighbouring undamaged cell, to correct for loss of fluorescence due to photobleaching; the resulting values were divided by the average of all pre-irradiation values and plotted with respect to time. This normalization allows changes in intensities over time at the site of damage and the total nucleus to be directly compared. For quantification described in Fig. 5D, mean intensity per pixel at the site of damage or for total nucleus in irradiated cells were background-subtracted and corrected for photobleaching using neighbouring non-damaged cells as above; obtained values were then divided by the total nuclear mean intensity at each time point to normalize for differences in expression levels between constructs and loss of fluorescence due to laser-induced photobleaching. The resulting values were divided by the average of all pre-irradiation values and plotted with respect to time. With this normalization, accumulation at the site of damage is shown as fold increase over 1, facilitating direct comparison of different proteins. For Fig. 4A, total fluorescence accumulating at the site of damage was calculated by subtracting from the integrated intensity at the site of damage the intensity contributed by nucleoplasmic fluorescence in the same area. Values were normalised for total integrated intensity in each cell and plotted with respect to time. Similar results were obtained when data in Fig. 4A were normalized as in Fig. 3 or Fig. 5.

Imaging on fixed samples was performed on SP2 AOBS Sirius or on SP5 Leica confocal microscopes (Leica Microsystems, Wetzlar, Germany) equipped with 63× 1.4 NA oil-immersion lenses. The images in supplementary material Fig. S3A were acquired on a Nikon Eclipse TE 2000-U epifluorescence microscope equipped with 60× 1.4 NA oil-immersion lens.

Photobleaching experiments were perfomed on a Leica SP2 AOBS Sirius microscope equipped with a 63× 1.4 NA oil-immersion lens. A defined stripe of 1.29 μm width was placed to the nuclear midpoint of non-irradiated cells or over the DSB-containing nuclear stripe where the GFP-tagged proteins were accumulated in the laser-treated cells. GFP was excited using the 488 nm laser line. Fifty prebleach images were recorded with 4% laser power of the 488 nm line, followed by a double bleach pulse of 0.200 seconds using the 456, 476, 488 and 496 nm laser lines combined at maximum power. After bleaching, 300 images were recorded at 0.200-second intervals with 4% laser power of the 488 nm line. For analysis of Cdt2 and PCNA binding, only cells with diffuse nuclear staining (which are not in the S phase of the cell cycle) were analysed.

Mean fluorescence intensities in the FRAP region and across the entire nucleus (including the FRAP region) were calculated and background subtracted. The bleached region and whole nucleus profiles were then individually normalized to their prebleach values (obtained from averaging the ten images immediately preceding the bleaching). The final recovery profile was then obtained by dividing these two normalized profiles (FRAP region divided by whole nucleus). This final division removes the global decrease in fluorescence upon bleaching as well as any later gradual bleaching that occurred during the recovery (the latter was always less than a few percent). An incomplete recovery (to a value less than 1) indicates the presence of an immobile fraction.

Recoveries in undamaged and damaged nuclei were interpreted using an analytic Laplace-transform-based solution of the full reaction-diffusion equations in a rectangular geometry. For the undamaged nuclei, all recoveries were complete and consistent with ‘effective diffusion’. The recoveries were therefore governed by two parameters, the asymptotic recovery value and the effective diffusion constant, which were fitted by minimizing χ² to individual nuclear recoveries. The effective diffusion constants were used, along with other information from the pre-images and the asymptotic recovery value, to predict and/or fit the recoveries in the damaged nuclei (see supplementary information at http://edoc.mpg.de/display.epl?mode=doc&id=521078&col=18&grp=125 for more details on data analysis).

We thank Stefan Terjung (Advanced Light Microcopy Facility, EMBL) for expert assistance with microscopy, Cristina Cardoso (Technische Universität Darmstadt, Germany) for kindly providing FP-PCNA constructs, Caido Sirinian for expert assistance with plasmid constructions and the Light Microscope Facility of Patras Medical School. This work was supported by grants from EU (NEST-004995), University of Patras Karatheodori Program and the Association for International Cancer Research to Z.L. and an EMBO short-term fellowship to V.R.