The Mre11-Rad50-Nbs1 complex mediates the robust recruitment of Polo to DNA lesions during mitosis in Drosophila

Cells encounter genotoxic stress originating from extrinsic and intrinsic sources (Cannan and Pederson, 2016). DNA double-strand breaks (DSBs) are potentially the most harmful damage as they can cause genomic instability, a hallmark of cancer cells (Jackson and Bartek, 2009). The presence of DSBs in interphase triggers a conserved signaling pathway, also known as the DNA damage response (DDR), that promotes DNA repair concomitantly with cell cycle delay through checkpoint activation (Hartlerode and Scully, 2009; Melo and Toczyski, 2002; Sekelsky, 2017). In brief, DSBs are promptly recognized by DNA damage sensors including the conserved proteins Meiotic recombination 11 homolog 1 (Mre11) and Rad50, which are part of a complex with the less-conserved Nijmegen breakage syndrome protein 1 (Nbs1; also called Nbn) (Lisby et al., 2004; Lukas et al., 2003, 2004; Syed and Tainer, 2018; Williams et al., 2010). Subsequently, the Mre11-Rad50-Nbs1 (MRN) complex recruits the phosphoinositide-3-kinase-related kinase (PIKK) Ataxia telangiectasia mutated (ATM) to the site of DNA damage via direct interactions (Falck et al., 2005). ATM promotes the rapid modification of chromatin flanking the breaks, including extensive phosphorylation of the histone 2A variant H2AX (Rogakou et al., 1998), which serve as docking sites for the C-terminal motif of Breast cancer type 1 susceptibility protein present in Mediator of DNA damage checkpoint protein 1 (MDC1) (Dronamraju and Mason, 2009; Stewart et al., 2003; Stucki et al., 2005). MDC1 directly interacts with Nbs1 and recruits additional MRN complexes to DSBs via an amplification loop (Lukas et al., 2004; Wu et al., 2008). This primary signal is transduced to effectors that direct DNA repair through two main pathways: the error-prone non-homologous end-joining (NHEJ) pathway or the error-free homologous recombination (HR) pathway (Ceccaldi et al., 2016; Syed and Tainer, 2018).

Studies have reported a variety of responses to the presence of DSBs during mitosis, depending on the stage at which the damage occurs (Blackford and Stucki, 2020; Thompson et al., 2019). A single broken chromosome end created by the breakage of a dicentric chromosome during anaphase can produce a chromosomal breakage–fusion–bridge cycle (McClintock, 1938). It can also be repaired by either de novo telomere formation or homolog-dependent restoration of the chromosome terminus (Ahmad and Golic, 1998; Bhandari et al., 2019; Haber and Thorburn, 1984; McClintock, 1939, 1941). In budding yeast, a DSB created during telophase triggers reversion of chromosome segregation and coalescence of sister chromatids that facilitates repair by HR (Ayra-Plasencia and Machín, 2019). In vertebrates, the presence of DSBs in prometaphase activates primary DNA damage signaling, comprising DNA lesion recognition by the MRN complex followed by partial activation of ATM and subsequent H2AX phosphorylation on chromatin proximal to the breaks (Benada et al., 2015; Giunta et al., 2010; Gomez-Godinez et al., 2010; Orthwein et al., 2014; Peterson et al., 2011; Silva et al., 2014; Terasawa et al., 2014). Subsequent downstream signaling pathways that promote DNA repair by NHEJ and HR are inhibited until the next G1 phase (Benada et al., 2015; Giunta et al., 2010; Orthwein et al., 2014; Peterson et al., 2011; Terasawa et al., 2014; van Vugt et al., 2010; Yu et al., 2012). In Drosophila neuroblasts, clustered DSBs on one chromosome are recognized by the mitotic proteins Polo, BubR1 and Bub3. These proteins mediate the proper segregation of the damaged chromatid by tethering the two broken ends (Derive et al., 2015; Royou et al., 2010).

BubR1 and Bub3 act together to stabilize kinetochore–microtubule attachments. They also participate in the spindle assembly checkpoint, which inhibits the anaphase-promoting complex/cyclosome (APC/C), an E3 ubiquitin ligase that triggers anaphase by targeting key substrates for proteolysis, until all chromosomes are properly attached to the spindle (Basu et al., 1999, 1998; Karess et al., 2013; Lopes et al., 2005). In a recent study, we provided evidence that BubR1 and Bub3 depend on each other to localize on DNA lesions during mitosis where they promote local inhibition of the APC/C via sequestration of the APC/C cofactor Cdc20 (Fzy in Drosophila) (Derive et al., 2015).

Polo is a serine/threonine kinase that plays multiple conserved functions during mitosis, including centrosome maturation, bipolar spindle formation, kinetochore function and cytokinesis (Archambault et al., 2015; Llamazares et al., 1991; Sunkel and Glover, 1988). We found that Polo participates in a signaling pathway that tethers chromosome fragments during mitosis, thus preventing genome instability (Royou et al., 2010). However, the molecular pathway governing Polo, BubR1 and Bub3 recruitment to broken chromosomes and their dependency relationship with the DNA damage signaling pathway identified in vertebrate mitotic cells has not been characterized.

In this study, we examined the kinetics of Drosophila Polo, BubR1 and Bub3 at DNA lesions created with a pulsed-UV laser in a precise spatiotemporal manner during the different phases of mitosis to assess the coordination of their recruitment to DSBs and investigated their interplay with the DSB sensor (the MRN complex).

The kinetics of Polo recruitment to DSBs in mitosis were first examined by inducing surgical damage to one chromosome in live Drosophila neuroblasts at precise times during mitosis with a pulsed-UV laser. The signal of GFP::Polo on the resulting DNA lesions was subsequently monitored by time-lapse imaging (within 1-5 s of irradiation). This method of micro-irradiation (IR) creates complex DNA lesions including DSBs (Aleksandrov et al., 2018). We first calibrated laser power to determine the optimal IR dose for our experiments. We identified 20% laser power as the lowest dose that produced a 100% response (i.e. the presence of a detectable Polo signal at the site of damage) (Fig. S1A-D). In contrast to 20% laser power, lower IR doses resulted in a dramatic decrease in the frequency of cells exhibiting a detectable GFP::Polo signal at the lesions and a significant delay in the appearance of GFP::Polo, accompanied by reduced levels of GFP-Polo at the site of damage (Fig. S1A-D). Because no noticeable changes in GFP::Polo kinetics were observed between 20 and 30% power, we chose to induce DSBs using 20% laser power in all subsequent experiments. This response was specific to damage on the chromosomes as no GFP::Polo foci were detected in the cytoplasm after IR (Fig. S1E).

Because Polo is targeted to different subcellular structures depending on the stage of mitosis, we reasoned that its recruitment to DSBs could be dependent upon the stage of mitosis during which DSBs were created. We therefore measured the kinetics of GFP::Polo on DSBs induced at different phases of mitosis. Cells were irradiated in prometaphase (after nuclear envelope breakdown and at least 2 min before anaphase onset), metaphase (less than 2 min before anaphase onset), anaphase (within 2 min of anaphase onset) and telophase/G1 (>6 min after anaphase onset, during chromatin decondensation and nuclear expansion). Following IR, Polo signal was consistently detected at the site of DNA damage when the damage was induced during prometaphase, metaphase and anaphase (Fig. 1A). By contrast, no GFP::Polo signal or only traces were detected at the site of IR-induced damage at telophase/G1 phase (T/G1) (Fig. 1A,C,F). In most cases, Polo exhibited biphasic kinetics at DSBs, with a period of accumulation followed by a period of dissociation. In some instances, a short plateau was observed between the phases of association and dissociation (Fig. 1B, PM curve). Polo signal was detected on DSBs 20-40 s following IR during all stages except T/G1 (Fig. 1C). Its half-time recruitment was variable (ranging from 1 to >6 min), particularly when IR was applied during prometaphase (Fig. 1D), and was correlated with the time elapsed between IR and anaphase onset (R²=0.5, Fig. S2A). Importantly, Polo onset of dissociation from the damaged chromatin consistently occurred near or after anaphase onset, in concert with Cyclin-dependent kinase 1 (Cdk1) inhibition (Fig. 1B,E). The maximum levels of Polo at DNA lesions varied between cells, but did not correlate with the time elapsed between IR and anaphase onset (Fig. 1F and Fig. S2B). However, Polo never reached high levels on lesions induced after anaphase onset (Fig. 1F). We noted that a pool of Polo remained associated with DSBs more than 6 min following anaphase onset, which corresponds to the average time of completion of nuclear envelope re-formation in Drosophila neuroblasts (Fig. 1A, last images in each panel) (Montembault et al., 2017). We conclude that Polo is rapidly recruited to DNA lesions encountered at any point during mitosis and that the efficiency of its recruitment to DSBs correlates with Cdk1 activity.

Because Polo dissociation from DNA lesions consistently occurred after sister chromatid separation, at a time of Cdk1 inhibition, we determined whether Polo maintenance at DSBs correlated with the level of Cdk1 activity. To do so, we analyzed Polo signal in cells arrested in prometaphase with high Cdk1 activity as a result of colchicine treatment, which rapidly depolymerizes microtubules in mitotic neuroblasts (Fig. S3). We observed that the first phase of Polo association at lesions was similar in untreated and colchicine-treated cells (Fig. 1G,H). No significant difference in the half-time of Polo recruitment or its maximum levels at DNA lesions was observed between control and prometaphase-arrested cells (Fig. 1I,J). These results indicate that a functional spindle is not required for the efficient recruitment of Polo to DNA lesions. Interestingly, although Polo underwent a phase of dissociation from the lesions when untreated cells entered anaphase, Polo levels remained high and constant on the lesions in prometaphase-arrested cells (Fig. 1H). These results are consistent with a model in which Cdk1 activity is required for the maintenance of Polo at mitotic DSBs. Collectively, these results demonstrate the highly dynamic nature of Polo localization during mitosis, when it can be readily recruited to DNA lesions upon demand.

The C terminus of Polo contains two Polo box domains (PBDs) that target Polo to different subcellular structures, such as centrosomes and unattached kinetochores, via recognition of phosphoepitopes that serve as docking sites to the PBDs (Elia et al., 2003a,b) (for review see Schmucker and Sumara, 2014). These phosphoepitopes on Polo targets are generated by priming kinases such as Cdk1, or by Polo itself (Kang et al., 2006; Lénárt et al., 2007). This prompted us to determine whether Polo kinase activity is required for its own recruitment to DNA lesions. To do so, we treated cells with BI2536, an inhibitor of Polo kinase activity (Lénárt et al., 2007; Riparbelli et al., 2014; Steegmaier et al., 2007). Polo inhibition is known to result in prolonged prometaphase (Conde et al., 2013; Lénárt et al., 2007; Llamazares et al., 1991; Sumara et al., 2004; Sunkel and Glover, 1988). Therefore, we compared GFP::Polo kinetics at DNA lesions in BI2536-treated cells with its kinetics in cells arrested in prometaphase after colchicine treatment, as colchicine did not affect the association phase of Polo at DSBs (as shown in Fig. 1H,I). We found that Polo inhibition using BI2536 severely altered GFP::Polo kinetics at DSBs (Fig. 2A,B). The initiation and the half-time of GFP::Polo recruitment to DNA lesions were delayed more than threefold upon BI2536 treatment (Fig. 2C,D). Similarly, BI2536-treated cells did not display the high levels of GFP::Polo on lesions observed in untreated cells or after colchicine treatment (Fig. 1J, Fig. 2E). Cells treated simultaneously with colchicine and BI2536 exhibited similar defects in Polo kinetics at DSBs to cells treated with BI2536 alone (Fig. 2A-D). Our results indicate that the rapid and robust recruitment of Polo to DNA lesions depends on its kinase activity, as reported for its localization to centrosomes and kinetochores in human cells (Lénárt et al., 2007).

Previously, we reported that BubR1 and Bub3 localized as a complex to damaged chromatids throughout mitosis, where they are required to sequester Fzy via its direct interaction with the KEN motif in BubR1 (Derive et al., 2015). We therefore analyzed the kinetics of GFP-tagged BubR1, Bub3 and Fzy after IR, as described for GFP::Polo. We observed that GFP::BubR1, GFP::Bub3 and GFP::Fzy were efficiently recruited to DNA breaks if the damage occurred before anaphase (Fig. 3A). Their kinetics on DNA lesions was similar to GFP::Polo. All three proteins underwent a phase of association, culminating a few minutes after anaphase onset, followed by a phase of dissociation during anaphase/telophase (Fig. 3B,E). BubR1, Bub3 and Fzy shared similar timing of recruitment, which is consistent with their association to DSBs as a complex. However, their kinetics differed from Polo in three ways: First, their initiation and half-time of recruitment at the site of DNA damage were four and three times slower, respectively, compared with Polo (Fig. 3C,D). Second, unlike Polo, all three proteins dissociated completely from DBSs at telophase (Fig. 3B). Third, these proteins failed to localize to DNA lesions in more than half of the cells if the damage was induced after anaphase onset (Fig. 3A). This latter observation might be related to their slower rate of recruitment to DNA lesions, combined with their rapid and complete dissociation at telophase.

The kinetics of Polo and the BubR1/Bub3/Fzy complex relative to each other at DSBs were monitored using simultaneous live-cell imaging of GFP::BubR1 and mCherry::Polo after IR (Fig. 4). In all instances, we observed that GFP::BubR1 foci appeared after mCherry::Polo at the site of damage (Fig. 4A-D). Consistently, the half-time of BubR1 recruitment was delayed compared with Polo (Fig. 4E). For both proteins, the phase of dissociation was initiated after anaphase onset (except in one cell for mCherry::Polo) (Fig. 4F). However, whereas BubR1 dissociation was complete at telophase, a pool of Polo remained on the lesions and disassembled at a later stage during the next interphase (Fig. 4G). Consequently, the half-time of Polo dissociation from damaged chromatids was slower than that of BubR1 (Fig. 4H). These simultaneous dual-color imaging data confirmed our previous observations on the rapid association and slower dissociation kinetics of Polo at DNA breaks relative to BubR1.

The observation that Polo recruitment to DSBs precedes BubR1 and Bub3 prompted us to assess the requirement of Polo activity for the localization of BubR1 and Bub3 to sites of DNA damage. To do so, we monitored BubR1 and Bub3 dynamics in cells carrying a strong hypomorphic allele of polo (polo¹⁰) (Donaldson et al., 2001). Because attenuation of Polo activity induces a prolonged prometaphase (Conde et al., 2013; Llamazares et al., 1991; Sunkel and Glover, 1988), we compared the kinetics of BubR1 and Bub3 in polo¹⁰ mutant cells with wild-type cells prolonged in prometaphase by colchicine treatment. Attenuation of Polo function severely altered the kinetics of BubR1 and Bub3 recruitment to DNA lesions (Fig. 5A-C). The initiation and half-time of recruitment of both BubR1 and Bub3 to DNA breaks were greatly delayed compared with their kinetics in colchicine-induced prometaphase-arrested cells (Fig. 5D,E). To confirm this result, we monitored the kinetics of GFP::BubR1 in wild-type cells treated with the Polo inhibitor BI2536. Upon pharmacological inhibition of Polo activity, BubR1 signal was not detected at the site of DNA damage in the majority of cells (Fig. 5F). Collectively, these data indicate that Polo promotes the efficient recruitment of BubR1 and Bub3 to DNA breaks.

In a previous study, we reported that none of the core kinetochore proteins (including CenpA/Cid, CenpC, Spc105, Spc25 and Nuf2) nor members of the spindle assembly checkpoint (Mps1, Mad1, Mad2) localized on I-CreI- or laser-induced DSBs (Derive et al., 2015; Royou et al., 2010). These data suggest that the formation of a neo-kinetochore on broken chromosomes is highly unlikely. As Cid is required for the localization of Polo and BubR1 on kinetochores (Blower and Karpen, 2001), we reasoned that Polo and BubR1/Bub3 localization on DNA breaks is likely to be mediated via interactions with components that do not depend on kinetochore proteins. Recent studies in vertebrates reported that components of the early steps of the DDR, including the MRN complex, ATM and γH2AX, form foci on damaged chromosomes during mitosis (Giunta et al., 2010; Gomez-Godinez et al., 2010; Orthwein et al., 2014). Similarly, γH2A.Z foci were detected on damaged chromatids in Drosophila neuroblasts (Royou et al., 2010), suggesting that activation of the initial steps of the DDR in mitosis is conserved in Drosophila. It therefore seemed possible that activation of the first steps of the DDR during mitosis promotes the recruitment of Polo and, consequently, BubR1 and Bub3 to DNA lesions.

Thus, we examined the dynamics of the DSB sensors Mre11 and Rad50 on sites of laser-induced DNA damage in mitotic neuroblasts. One dose of mRFP-Rad50 rescued the lethality associated with a rad50^EP strong hypomorph mutation, demonstrating that mRFP did not ablate Rad50 function. Next, the dynamics of GFP::Mre11 and mRFP::Rad50 were monitored on DNA lesions at different stages of mitosis. Mre11 and Rad50 signals appeared within 5-10 s of IR, which was even more rapid than the Polo signal (Fig. 6A). As was the case for Polo, BubR1 and Bub3, both Mre11 and Rad50 exhibited biphasic kinetics at DNA lesions, with a phase of association promptly followed by a phase of dissociation (Fig. 6A,B). We also observed the re-association of Mre11 and Rad50 at sites of DNA damage, coupled with their nuclear localization upon entry into the next interphase (Fig. 6A, last insets). Remarkably, the biphasic kinetics of Mre11 and Rad50 did not correlate with the timing of anaphase onset (Fig. 6B), in contrast to the kinetics of Polo, BubR1, Bub3 and Fzy. Consequently, the half-time of Mre11 recruitment was less than 1 min, regardless of the stage of mitosis at which DSBs were induced (Fig. 6C,D). These observations suggest that the level of Cdk1 activity has minimal effect on MRN kinetics at DSBs. However, when IR was applied during anaphase, Mre11 did not reach the same levels on DSBs as observed during other points in mitosis (Fig. 6E). This puzzling observation led us to test the influence of Cdk1 activity on Mre11 dynamics at lesions by monitoring GFP::Mre11 in colchicine-treated, prometaphase-arrested cells after IR. No difference in Mre11 kinetics was detected between cells going through mitosis and prometaphase-arrested cells (Fig. 6C-E). This result suggests that, unlike Polo, high Cdk1 activity does not prevent Mre11 dissociation from DSBs.

Next, we examined the differences in kinetics between the MRN complex and Polo on mitotic DNA lesions by dual-color imaging of neuroblasts expressing both GFP::Mre11 and mCherry::Polo (Fig. 6F,G). As previously observed, GFP::Mre11 signal promptly appeared at the site of IR, preceding mCherry::Polo (Fig. 6F-H). GFP::Mre11 recruitment to DSBs was brief and invariably initiated its dissociation in less than 3 min after IR, regardless of the stage of mitosis (Fig. 6F,G). In contrast, mCherry::Polo consistently remained associated at the damaged chromosome until anaphase onset (Fig. 6E,G,I). This point is illustrated by the observation that its half-time of recruitment has a greater variance than that of GFP::Mre11 (Fig. 6H).

Because Mre11 precedes Polo at the site of IR, we hypothesized that the recognition of DSBs by the MRN complex during mitosis facilitates the recruitment of Polo and, subsequently, BubR1 and Bub3 to DNA breaks. To test this idea, we monitored the dynamics of Polo in rad50 mutant cells. Rad50 is an essential protein in Drosophila, and null or strong hypomorphic alleles result in lethality at the pupal stage (Ciapponi et al., 2004; Gorski et al., 2004). Furthermore, Rad50 as well as Mre11 and Nbs1 have conserved functions in DNA repair signaling and telomere protection (Ciapponi et al., 2004, 2006; Gorski et al., 2004; Sekelsky, 2017).

We first confirmed that Mre11 was no longer detected on DSBs in mitosis in rad50^EP strong-hypomorph mutant cells (Fig. 7A). Then, we monitored the kinetics of recruitment of GFP::Polo in rad50^EP mutant cells after IR (Fig. 7B,C). We observed that attenuation of Rad50 severely altered Polo dynamics at DSBs. The time of Polo appearance at DNA lesions and its half-time of recruitment were both greatly delayed (Fig. 7D,E). Similarly, Polo did not reach the high levels at DNA lesions observed in wild-type cells in rad50^EP mutant cells (Fig. 7F). These results indicate that the MRN complex acts upstream of Polo at mitotic DSBs.

Previous studies demonstrated that prolonged binding of the MRN complex to chromatin was sufficient to trigger the recruitment and activation of downstream components of the DDR independently of DNA damage (Bonilla et al., 2008; Soutoglou and Misteli, 2008). To determine whether ectopic localization of the MRN complex to chromatin was sufficient to recruit Polo independently of DNA lesions in mitosis, we fused Mre11 to the Escherichia coli lactose repressor (LacI), tagged with GFP. The construct was introduced into flies carrying an X chromosome with 256 repeats of the lac operator sequence (LacO) located near the telomere (Fig. 8A) (Belmont and Straight, 1998). We first checked that the fusion protein was efficiently recruited to laser-induced DNA damage during mitosis, and that it co-localized with mRFP::Rad50 and mCherry::Polo at the site of damage (Fig. S4A,B). Next, we monitored the localization of the fusion proteins in flies carrying one X chromosome with 256 LacO repeats. As expected, GFP::LacI::Mre11 signal was observed as distinct foci on LacO arrays in 73% of mitotic cells (Fig. 8B). We did not observe any obvious defects in chromatid segregation, suggesting that the persistent tethering of Mre11 to chromatin did not affect genome integrity. In the seven mitotic cells where GFP::LacI::Mre11 was detected on LacO arrays, three cells exhibited mRFP::Rad50 foci colocalizing transiently with GFP::LacI::Mre11. This result indicates that the ectopic localization of Mre11 on chromatin resulted in the recruitment of Rad50, as demonstrated in human cells (Fig. 8B) (Bonilla et al., 2008; Soutoglou and Misteli, 2008). Remarkably, mCherry::Polo signal was also detected with GFP::LacI::Mre11 foci in 25% of cells (Fig. 8B). This result indicates that the tethering of Mre11 to chromatin is sufficient to recruit Polo independently of DNA damage.

In vertebrates, the pathway downstream of MRN and ATM involved in DNA repair by NHEJ or HR is inhibited during mitosis (Giunta et al., 2010; Orthwein et al., 2014; Peterson et al., 2011; Terasawa et al., 2014). To determine whether this inhibition is conserved in Drosophila after IR, we monitored the dynamics of Okra (Drosophila homolog of human RAD54), a helicase involved in DNA repair by HR (Kooistra et al., 1999; Schupbach and Wieschaus, 1991; Sekelsky, 2017). As expected, a prominent Okra::GFP focus was detected at the site of laser-induced damage in interphase nuclei, indicating that DNA repair by HR was initiated (Fig. S5, cell#2). In contrast, no Okra::GFP signal was observed on damaged chromosomes in mitotic cells (Fig. S5, cell#1). Collectively, theses results provide evidence that, similar to vertebrates, the early steps of the DDR are activated upon DNA damage in Drosophila mitotic cells. This is followed by inhibition of the Okra-dependent downstream repair pathway.

Using live single-cell analysis combined with surgical damage of one chromosome at a defined time during mitosis, we have performed detailed analysis of the kinetics of the mitotic proteins Polo, BubR1, Bub3 and Fzy and the DSB sensors Mre11 and Rad50 on mitotic DSBs. Our studies show that Mre11 and Rad50 transiently associate with DSBs and facilitate the robust recruitment of Polo and, consequently, the BubR1/Bub3 complex to DSBs. This, combined with our previous observation that Polo, BubR1 and Bub3 are required for tethering of the two broken chromosome ends, can be integrated within the following model: DSBs in mitosis are promptly marked by the MRN complex, promoting the recruitment of Polo kinase (Fig. 8C). Subsequently, the MRN complex dissociates from DSBs, possibly to prevent activation of downstream repair components, while Polo kinase activity promotes the subsequent accumulation of the BubR1/Bub3 complex at DSBs. This complex sequesters Fzy, thereby inhibiting the APC/C locally. Polo and BubR1/Bub3 levels are maintained at DSBs until mid-anaphase, thus facilitating the persistent tethering of broken fragments at a crucial time during poleward chromosome movement. Upon segregation of the broken DNA fragment, Polo and the BubR1/Bub3/Fzy complex dissociate from the DNA lesions. However, a pool of Polo persists on DSBs through the next interphase, while MRN re-accumulates at the site of damage to promote downstream repair signaling (Fig. 8C) (Derive et al., 2015; Royou et al., 2010).

We observed that Polo promptly accumulates at DSBs at any point during mitosis, including anaphase, albeit with lower levels. The kinetics of its recruitment are sigmoidal, suggesting the cooperative binding of Polo at the site of damage. Given that Polo localization to subcellular structures is mediated through its interaction with primed phosphorylated substrates (Elia et al., 2003a,b; Schmucker and Sumara, 2014), our findings that pharmacological inhibition of Polo kinase activity severely alters Polo recruitment to DSBs, but does not completely prevent it, supports the conclusion that Polo kinase activity is not required for the initial steps of DSB association but promotes its cooperative binding to DSBs through a self-priming mechanism. Interestingly, recent studies have reported that MRE11 is a substrate of polo-like kinase 1 (PLK1) in human cells (Li et al., 2017; Xu et al., 2018). It is therefore conceivable that Polo promotes its recruitment to DSBs via phosphorylation-dependent priming of Mre11.

Several lines of evidence suggest that Cdk1 activity modulates Polo kinetics at DSBs. First, Polo initiates its dissociation from DSBs invariably during anaphase, in concert with the decline of Cdk1 activity. Second, Polo levels are constantly maintained at DSBs in prolonged prometaphase when Cdk1 activity is high. Finally, Polo never reaches the high levels on DSBs created during anaphase. Although the molecular mechanism by which Cdk1 contributes to Polo kinetics at DSBs remains to be elucidated, we can propose plausible possibilities. Numerous Polo substrates are primed by Cdk1 (Lowery et al., 2005); thus, Cdk1 may facilitate the enrichment of Polo at DNA lesions by priming Polo targets associated with DSBs. Alternatively, Cdk1 may sustain Polo activation at DSBs directly through phosphorylation of the T-loop in the kinase domain (Mortensen et al., 2005), or indirectly by inhibiting counteracting phosphatases. The observation that Polo is successfully recruited to DSBs induced at anaphase, albeit with less efficiency, suggests that Cdk1 activity is not essential for the initial steps of Polo recruitment to DSBs but is important for its cooperative binding and maintenance at the lesions.

In this study, we demonstrate that one of the functions of Polo at mitotic DSBs is to target the BubR1/Bub3 complex. Polo precedes BubR1/Bub3 on DSBs and attenuation of its kinase activity dramatically alters the kinetics of BubR1/Bub3 at DSBs. This raises the question of how Polo mediates the recruitment of BubR1/Bub3 to DSBs. One possibility is via direct phosphorylation of BubR1 by Polo. In a previous study, we found that a 432 amino acid fragment of BubR1 (from amino acid 330 to amino acid 762) was sufficient for its localization to mitotic DSBs. This sequence encompasses a conserved motif called KARD, which is phosphorylated by PLK1 in human cells, resulting in the interaction of BubR1 with the phosphatase PP2A-B56α at kinetochores (Espert et al., 2014; Kruse et al., 2013; Suijkerbuijk et al., 2012). However, it is unlikely that Polo mediates the localization of BubR1/Bub3 to DSBs via direct phosphorylation of the BubR1 KARD motif, because mutation of the putative Polo phosphosites to alanine or aspartate had no obvious effect on the localization of BubR1 to mitotic DSBs (Derive et al., 2015). However, this 432 amino acid sequence includes seven other Polo putative phosphosites, including serine at position 581, which is phosphorylated in vivo in Drosophila embryos (Zhai et al., 2008). It would be relevant to determine whether any of these putative Polo phosphosites are important for BubR1 targeting to DSBs. Alternatively, Polo might target BubR1/Bub3 to DSBs indirectly by promoting interactions with another substrate. Studies have shown that Bub3 is targeted to the kinetochore via direct interaction with the kinetochore scaffold KNL1. This interaction involves the recognition of phosphorylated MELT repeats in KNL1 by two residues in Bub3 (Primorac et al., 2013). Importantly, although the MELT repeats are phosphorylated primarily by the spindle checkpoint component Mps1, studies have shown that Plk1 also contributes to this phosphorylation (Espeut et al., 2015). Interestingly, sequence analysis reveals that Drosophila Rad50 contains two MELT repeats. In addition, the residues identified in human BUB3 that are crucial for its interaction with the phosphorylated MELTs are conserved in Drosophila Bub3. Although speculative, it is possible that Polo phosphorylates Rad50 on its MELT repeats at mitotic DSBs, and these phospho-MELTs, in turn, target the BubR1/Bub3 complex via direct interaction with Bub3. It would be relevant to determine the importance of the conserved Bub3 residues and the Rad50 MELT repeats in the recruitment of BubR1/Bub3 to DSBs.

Although the detection of DSBs in interphase can be facilitated by confinement of the DSBs sensors in the nucleus and their association to chromatin, the recognition of DSBs during mitosis poses a greater challenge, as the DSBs sensors are dispersed throughout the cytoplasm (Polo and Jackson, 2011). Here, we show that cytoplasmic Mre11 and Rad50 readily accumulate on DSBs in mitosis with a half-time of recruitment estimated at around 30-40 s, modestly faster than that observed for human RAD50 on interphase DSBs using the same method of IR (t_1/2∼60 s) (Aleksandrov et al., 2018). This result is consistent with the detection of mitotic and interphase DSBs by MRN via a conserved mechanism. Experiments using single molecule tracking of MRN on in vitro reconstituted DNA homoduplexes have provided evidence that chromatin-associated MRN detects free DNA ends by scanning the DNA molecule through facilitated diffusion (Myler et al., 2017). However, as MRN is excluded from the condensed chromosomes, this chromatin-dependent diffusion-based mechanism is unlikely to occur on mitotic DSBs. It is possible that the initial detection of free DNA ends occurs through a different pathway, possibly through cytoplasmic diffusion but, once associated, a minimal amount of MRN triggers the rapid and robust accumulation of additional MRN molecules via cooperative binding, as suggested by the sigmoidal shape of Mre11 recruitment to mitotic DSBs. We cannot rule out that an undetected pool of MRN complex remains associated to condensed chromatin during mitosis and senses DSBs through active diffusion, as demonstrated in vitro (Myler et al., 2017).

The recruitment of Mre11 and Rad50 to mitotic DSBs is strikingly transient. Maximum levels are observed ∼2 min following laser IR and they subsequently dissociate rapidly from DSBs (t_1/2 dissociation ∼2 min), until the next interphase when they re-accumulate on the lesions. Although the kinetics of Drosophila MRN at DSBs during interphase is not known, our result contrasts with observations of human RAD50 in the interphase nucleus, where its removal from DSBs is slow (t_1/2>30 min) (Aleksandrov et al., 2018). We currently do not know the biological significance of the prompt dissociation of Rad50 and Mre11 from mitotic DSBs and how it is controlled. One possibility is that the transient association of MRN with mitotic DSBs is sufficient to initiate the recruitment of Polo and, consequently, BubR1/Bub3, but its ensuing rapid dissociation prevents the unscheduled recruitment of downstream DNA repair components to damage chromatin, which could otherwise compromise sister chromatid segregation. This idea is consistent with the observations that DNA repair signaling downstream of MRN is inhibited during mitosis (Giunta et al., 2010; Orthwein et al., 2014; Peterson et al., 2011; Terasawa et al., 2014) (this study) and that its reactivation provokes telomere fusion (Orthwein et al., 2014). If this model is correct, one might expect to observe an alteration of chromosome structure that would lead to segregation defects upon prolonged association of MRN with mitotic DSBs. However, no noticeable defects in chromosome segregation were observed in cells where Mre11 was persistently tethered to X chromosome-associated LacO arrays. These results should be interpreted cautiously because mRFP::Rad50 colocalized with LacO-associated GFP::LacI-Mre11 foci in less than half of the cells observed and most of the GFP::LacI::Mre11 and mRFP::Rad50 signal was associated with large cytoplasmic aggregates. Therefore, the amount of functional MRN complex generated using this approach might be insufficient to induce a dominant negative effect. The future challenge is to decipher the underlying mechanisms that control MRN kinetics on mitotic DSBs.

We found that depletion of MRN severely alters Polo kinetics at mitotic DSBs. Conversely, ectopic targeting of Mre11 to chromosomes is sufficient to recruit Polo, albeit with low efficiency. Collectively, these data strongly argue that the MRN complex plays a positive role in targeting Polo to DSBs. As discussed above, MRN might promote Polo association with DSBs via direct interaction with Mre11. Alternatively, MRN might target Polo to DSBs indirectly through the recruitment of additional DNA repair components that interact with Polo. In interphase, MRN recruits several effectors including MDC1 and C-terminal binding protein interacting protein (CtIP) to DSBs. Both MDC1 and CtIP are also found on mitotic DSBs (Giunta et al., 2010; Leimbacher et al., 2019; Wang et al., 2018). In addition, MDC1 interacts with TopBP1 on damaged chromosomes where they form filaments reminiscent of the Polo-coated tethers observed in Drosophila (Leimbacher et al., 2019; Royou et al., 2010). Importantly, systematic human interactome studies revealed an interaction between Plk1 and TopBP1 (Huttlin et al., 2015). Moreover, recent studies reported that the presence of DSBs in nocodazole-treated human cells triggers the Plk1-dependent phosphorylation of CtIP (Li et al., 2017). Therefore, it is possible that the MRN complex mediates the recruitment of Polo to DSBs indirectly by promoting the accumulation of one of its substrates such as TopBP1 or CtIP. The fact that Polo is maintained at DSBs until mid-anaphase or in prolonged prometaphase, at a time when MRN has already dissociated from DSBs, supports the notion that MRN is required for the initiation of Polo recruitment to mitotic DSBs but, unlike Cdk1, is dispensable for its maintenance.

Future studies will determine the molecular mechanisms by which MRN and Cdk1 control Polo association with mitotic DSBs and how Polo, in turn, facilitates the recruitment of BubR1/Bub3 to broken chromosomes. Current anti-cancer therapies rely heavily on the induction of extensive DNA damage, either by IR or genotoxic agents. Thus, deciphering the molecular pathways that process DNA damage in healthy mitotic cells will improve our understanding of tumor resistance to these therapies and may help identify novel targets that will enhance the efficacy of these therapies.

Flies were grown on corn flour and yeast extract medium under standard conditions at 25°C. The h2az>H2A.Z::mRFP stock was obtained from the Bloomington Drosophila Stock Center (Indiana University, USA). bub3>EGFP::Bub3 stock was provided by Christian Lehner (Institute of Molecular Sciences, University of Zurich, Switzerland). The bubR1>GFP::BubR1 stock was provided by R. Karess (Institute Jacques Monod, Paris, France). The polo>GFP::Polo stock was provided by Claudio Sunkel (Instituto de Biologia Molecular e Cellular, Porto, Portugal). The ubi>GFP::Fzy stock was provided by Jordan Raff (Sir William Dunn School of Pathology, University of Oxford, UK). The GFP::Jupiter is described by Karpova et al. (2006). The bubR1[1] (Basu et al., 1999), polo[10] (Donaldson et al., 2001) and rad50[EP1] (Ciapponi et al., 2004; Gorski et al., 2004) mutations were previously described. The transgenic stocks ubi>GFP::Mre11, ubi>GFP::LacI::Mre11, ubi>mRFP::Rad50, ubi>Okra::GFP and polo>mCherry::Polo were produced in our laboratory. The transgenic stock ubi>GFP::Mre11 and ubi>mRFP::Rad50 are described by Murcia et al. (2019).

All plasmids were verified by sequencing before being injected into fly embryos to generate transgenic stocks (Bestgene Inc.).

The LacI sequence was amplified from the pGEX6p2 vector using primers that added the AttB1 and AttB2 Gateway sequences and the sequence of a GAGAGAGA flexible peptide linker to the N and C termini of LacI, respectively. The Mre11 coding sequence was amplified from pENTR-N-Mre11 vector with primers that added the GAGAGAGA linker and the AttB1 and AttB2 Gateway sequences on N and C termini of Mre11, respectively. The two PCR products were combined and amplified to get the final PCR product that was inserted into the entry Gateway vector. This product was recombined with a pUbi-GFP::GAT destination vector (a gift from Jordan Raff) to obtain the plasmid ubi>GFP::LacI::Mre11.

The Okra coding sequence was amplified from the pOT2-Okr vector (DGRC, plasmid n°LD35220) using primers that added the AttB1 and AttB2 Gateway sequences on the N and C termini of Okra, respectively. The PCR product was inserted into the pDONR/Zeo vector (Life technologies). This resulting entry vector was recombined into the pUbi-GAT::GFP vector containing the ubiquitin promoter and a C-terminal GFP tag (a gift from Jordan Raff).

The polo>mCherry::Polo construct was obtained by multistep cloning. First, a region of 881 nucleotides upstream of the Polo initiation codon was amplified from genomic DNA using primers introducing BamHI and EcoRI restriction sites. mCherry was amplified with a forward primer introducing an EcoRI restriction site and a reverse primer introducing a NotI site plus a GAGAGAGA linker. The cDNA Polo sequence was amplified with forward and reverse primers introducing NotI and KpnI sites, respectively. A region of 704 nucleotides after the STOP codon of PoloCDS was amplified from genomic DNA using forward and reverse primers introducing XhoI and KpnI sites, respectively. Each PCR product was first inserted into a pGEM-T easy vector. After digestion, the products were ligated into a pAttB vector. Ligation of this product resulted in the final pAttB polo>mCherry::Polo vector.

For live imaging of neuroblasts, late third-instar larvae were dissected in PBS (phosphate buffer saline, 1×) and their central nervous system transferred in a drop of PBS on a coverslip. The brain was gently compressed by capillary action and the coverslip sealed with Halocarbon oil 700 (Sigma). Each preparation was monitored for less than 1 h. For colchicine and BI2536 treatments, whole brains were incubated in PBS containing 10 µM colchicine (Sigma) and/or 1 µM BI2536 (Selleckchem) for 30 min before preparation for live imaging as described above. These conditions for colchicine and BI2536 treatment were optimal to produce >30 min delay in prometaphase. Live analysis was performed at room temperature with a 100× oil Plan-Apochromat objective lens (NA 1.4) and an Axio-Observed.Z1 microscope (Carl Zeiss) equipped with a spinning disk confocal (Yokogawa), an EMCCD Evolve camera (Photometrics and Roper Scientific) and 491 nm (100 mW; Cobolt calypso) and 561 nm (100 mW; Cobolt Jive) lasers. Images were acquired with Metamorph software (Roper Scientific). Every 5 and/or 20 s, 12 Z-series of 0.5 µm steps were acquired. All images are maximum-intensity projections. A 355 nm microchip laser (Teem Photonics) (passively Q-switched SNV-20F-000) with a 21 kHz repetition rate, 0.8 µJ energy/pulse, 2 kW of peak power and 400 ps pulse width, powered with an iLas2 PULSE system (Roper Scientific) was used at 10, 15, 20 or 30% power with one pulse of a spot length of 100 points to induce surgical damage to one chromosome. For all experiments, 20% laser power was used unless specified otherwise.

ImageJ (Fiji, National Institutes of Health) was used for image analysis and fluorescence quantification. The level of our fluorescence-tagged proteins at DNA lesions was measured as follows: The average fluorescence intensity of a defined region of interest at the site of IR (ROI A) and the average fluorescence intensity of the cytoplasm near the site of IR (ROI B) were measured over time from maximum projection images (12z of 0.5 µm depth). The background of the camera was removed for each value. For each time point, the level of signal at the site of damage was equal to the difference between the average signal of ROIs A and B divided by the average signal in ROI B {[(ROI A-background)−(ROI B-background)]/(ROIB-background)}. The initiation of recruitment at DNA lesions corresponds to the time where the difference between ROI A and B signals is positive. To determine the half-time of recruitment (t_1/2) and maximum level parameters, each data set (from 0 min to the time of peak level) was fitted to the following sigmoid function: Y=Max/{1+10^{[((t1/2)−X)×Hillslope]}} using Prism (Graphpad), where Max is the maximum level. To estimate the half-time of dissociation, each data set was normalized to Max=1 and the data from Y=1 to the bottom plateau was fitted to the following sigmoid function: Y=Bottom+(1−Bottom)/{1+10^{[((t1/2)−X)×Hillslope]}}. In Fig. 1H, Fig. 4B,C and Fig. 6G, the data were normalized to Max=1.

We thank Derek McCusker and Cameron Mackereth, as well as Royou team members for fruitful discussion and critical reading of the manuscript.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.244442.reviewer-comments.pdf