DNA double-strand breaks (DSBs) trigger accumulation of the MRE11–RAD50–Nijmegen breakage syndrome 1 (NBS1 [MRN]) complex, whose retention on the DSB-flanking chromatin facilitates survival. Chromatin retention of MRN requires the MDC1 adaptor protein, but the mechanism behind the MRN–MDC1 interaction is unknown. We show that the NBS1 subunit of MRN interacts with the MDC1 N terminus enriched in Ser-Asp-Thr (SDT) repeats. This interaction was constitutive and mediated by binding between the phosphorylated SDT repeats of MDC1 and the phosphate-binding forkhead-associated domain of NBS1. Phosphorylation of the SDT repeats by casein kinase 2 (CK2) was sufficient to trigger MDC1–NBS1 interaction in vitro, and MDC1 associated with CK2 activity in cells. Inhibition of CK2 reduced SDT phosphorylation in vivo, and disruption of the SDT-associated phosphoacceptor sites prevented the retention of NBS1 at DSBs. Together, these data suggest that phosphorylation of the SDT repeats in the MDC1 N terminus functions to recruit NBS1 and, thereby, increases the local concentration of MRN at the sites of chromosomal breakage.
Double-strand breaks (DSBs) pose a major threat to genomic integrity by promoting mutations, gene deletions, and/or chromosomal translocations (Shiloh, 2003; Kastan and Bartek, 2004). These adverse effects are counteracted by genome surveillance machinery that is designed to rapidly detect DSBs, delay cell cycle progression, initiate DNA repair, or, in case of excessive damage, activate the cell death program (Zhou and Elledge, 2000; Nyberg et al., 2002; Lukas et al., 2004b). Because the maintenance of genome integrity requires extreme accuracy and coordination of elaborate, often intertwined molecular pathways, cells evolved numerous means to increase the efficiency of this process. Among others, virtually every eukaryotic cell responds to DSBs by concentrating signaling, repair, and various adaptor proteins in the vicinity of DNA lesions (Lukas et al., 2005). Although the purpose of this phenomenon (cytologically manifested by the formation of nuclear foci enriched in various DSB regulators) is not yet fully understood, several intriguing scenarios have been considered: for instance, DNA damage–induced protein accumulation may facilitate the formation of active repair holocomplexes, increase the availability of these enzymes at the sites of genetic lesions, generate a local chromatin microenvironment supportive for repair-associated DNA transactions, and/or amplify the DSB-associated signaling (Fernandez-Capetillo et al., 2004; Essers et al., 2006; Stucki and Jackson, 2006; Bartek and Lukas, 2007). Thus, the local concentration of DSB regulators emerges as a vital tool to increase the fidelity of the genome surveillance machinery.
Prominent among the cellular factors that accumulate at DSBs is the MRE11–RAD50–Nijmegen breakage syndrome 1 (NBS1 [MRN]) complex, an essential genome caretaker that regulates crucial steps of the DSB response such as DSB detection, activity of the ataxia telangiectasia mutated (ATM) kinase (the key upstream component of DSB signaling), cell cycle checkpoints (D'Amours and Jackson, 2002; Petrini and Stracker, 2003; Kobayashi et al., 2004; Stracker et al., 2004), and, as the most recent advancements suggest, induction of apoptosis (Difilippantonio et al., 2007; Stracker et al., 2007). In addition, the MRN complex participates in the resection of DNA ends, an essential step required for an error-free DSB repair by homologous recombination (Jazayeri et al., 2006). Such diverse involvement in the DSB response is reflected by the nuclear dynamics of MRN. Most notably, recent studies revealed that the distribution of NBS1 (and other MRN components) at the DSB sites is not uniform but splits into distinct subcompartments (Lukas et al., 2004a; Bekker-Jensen et al., 2006). Although a fraction of MRN interacts with single-stranded DNA formed after enzymatic DSB resection (consistent with the essential role of MRN in DSB resection), the bulk of the DSB-associated MRN accumulates within the vast regions of chromatin marked by ATM-phosphorylated histone H2AX (γ-H2AX). It is indeed this chromatin-associated fraction of MRN that cytologically manifests as the so-called ionizing radiation (IR)–induced foci.
The function of the chromatin-bound MRN has been subjected to intensive investigation and yielded important insights. Thus, it was found that the forkhead-associated (FHA) domain of NBS1 is necessary for its retention at the DSB-flanking chromatin and that its disruption impairs IR-induced foci formation of the entire MRN complex (Zhao et al., 2002; Cerosaletti and Concannon, 2003). Reconstitution experiments in human cells derived from the NBS patients suggested that the NBS1 N terminus, where the FHA domain resides, is important for survival, optimal ATM activity, and the intra–S-phase checkpoint after IR (Tauchi et al., 2001; Zhao et al., 2002; Cerosaletti and Concannon, 2003, 2004; Lee et al., 2003; Horejsi et al., 2004; Cerosaletti et al., 2006). However, because these assays were performed on the background of low levels of hypomorphic NBS1 alleles, the exact role of the chromatin-bound MRN remained a matter of debate. An important breakthrough in this discussion has been recently provided by Difilippantonio et al. (2005, 2007), who generated a mouse strain in which the endogenous NBS1 gene was replaced by a mutant with the disrupted FHA domain. Remarkably, although these mice were viable, they displayed defects in the DNA damage–induced G2/M- and S-phase checkpoints, an increased incidence of chromosomal aberrations, attenuated ATM activity after low doses of IR, and a decreased radiation survival. Although it is possible that the NBS1 N terminus supports other functions than the chromatin tethering of MRN, these results provide the most compelling evidence so far that the chromatin-bound MRN contributes to reach the threshold of the DNA damage signaling (especially after a low dose of IR), thereby guarding against chromosomal instability.
Although the requirement of the NBS1-FHA domain for the MRN focus formation has been established, the way this domain communicates with the DSB-flanking chromatin and/or the associated proteins is less clear. An earlier study suggested that the NBS1-FHA domain directly binds to γ-H2AX (Kobayashi et al., 2002). However, more recent results showed that the interaction of NBS1 with the DSBs requires MDC1, a large adaptor protein and an important upstream coordinator of the DSB-induced chromatin response (Goldberg et al., 2003; Stewart et al., 2003; Lukas et al., 2004a; Lou et al., 2006). MDC1 contains tandem BRCA1 C-terminal (BRCT) domains, which bind with high specificity to phosphorylated Ser139 of γ-H2AX (Stucki et al., 2005). As a result, MDC1 is among the first proteins to accumulate at the DSB sites, and its productive assembly in this compartment is a prerequisite for retention of most of the known chromatin-binding DSB regulators, including the MRN complex (Stucki and Jackson, 2006). The key evidence for the causative role of MDC1 for MRN focus formation was provided by several independent studies, which all showed that the siRNA-mediated knockdown of MDC1 by RNA interference in human cells (Goldberg et al., 2003; Stewart et al., 2003; Lukas et al., 2004a) or a complete MDC1 knockout in mouse cells (Lou et al., 2006) prevented MRN retention at the DSB-flanking chromatin. Importantly, kinetic measurements in living cells extended these results by showing that the MDC1 knockdown abolished MRN retention at DSBs from the earliest stages after DNA damage and that in the absence of MDC1, NBS1 could not concentrate at the DSB-flanking chromatin despite the fact that H2AX was efficiently phosphorylated in these compartments (Lukas et al., 2004a).
At lest three additional pieces of evidence suggested that the retention of MRN at the DSB-flanking (and γ-H2AX decorated) chromatin is not direct but is mediated by interaction with MDC1. First, peptides derived from the phosphorylated γ-H2AX C terminus interacted with MRN only in the presence of MDC1 (Lukas et al., 2004a). Second, MDC1 can be efficiently copurified with MRN from cell extracts (Goldberg et al., 2003; Stewart et al., 2003; Lukas et al., 2004a). Importantly, this interaction was dependent on the phosphate-binding FHA domain of NBS1, suggesting that the formation of a productive MDC1–MRN complex is phosphorylation dependent (Lukas et al., 2004a). Third, real-time imaging of protein assembly in live cells revealed that NBS1 accumulates at the DSB-flanking chromatin with a dynamics that is indistinguishable from that of MDC1 (but significantly faster than other chromatin-binding proteins such as 53BP1 or BRCA1), an observation that further supported the emerging model of the intimate relationship between MDC1 and the MRN complex (Lukas et al., 2004a; Bekker-Jensen et al., 2005; Mailand et al., 2007).
Although intriguing, these results left several mechanistic questions unanswered: most notably, it remains unclear whether the phosphosubstrate of the NBS1-FHA domain is indeed MDC1 itself or whether the MDC1–NBS1 interaction requires yet another, hitherto unrecognized (phospho)intermediate. Furthermore, the identity of the protein kinases and the phosphorylation sites that generate the recognition signal for the NBS1-FHA domain (and thus promote retention of MRN on the DSB-flanking chromatin) have not been determined. In an attempt to resolve these issues, we performed a detailed analysis of the MDC1–NBS1 interaction in vitro and in cells. We show that the NBS1-FHA domain binds to the N-terminal part of MDC1 enriched in acidic Ser-Asp-Thr (SDT) repeats. Furthermore, we provide evidence that the SDT repeats are constitutively phosphorylated, at least in part, by casein kinase 2 (CK2). These SDT phosphorylations are functionally significant because they trigger productive interaction between MDC1 and NBS1 and determine the retention of MRN at the DSB-modified chromatin.
The FHA domain of NBS1 interacts with the N-terminal part of MDC1
Endogenous MDC1 and NBS1 proteins efficiently interact both before and after DNA damage (Fig. 1 A) and contain numerous structural motifs and potential targets for posttranslational modifications (Fig. 1 B) that might be involved in this interaction. To elucidate the requirements for the MDC1–MRN interaction, we first constructed a series of HA-tagged fragments progressively spanning the N-terminal part of MDC1 (Fig. 1 C). We transiently expressed each of these fragments in a U2OS cell line in which the endogenous MDC1 was down-regulated by stably integrated short hairpin RNA (shRNA; Fig. 1 D; Bekker-Jensen et al., 2006) and tested their ability to interact with endogenous NBS1. The MDC1 fragments were constructed so that they were not affected by the shRNA to endogenous MDC1 (see Materials and methods), and the robust down-regulation of MDC1 in this cell line (designated U2OS/shMDC1) allowed us to assess the binding efficiency of each fragment unbiased by higher-order complexes containing the endogenous protein. Under these conditions, a fragment spanning the first 500 amino acids of MDC1 appeared to be necessary and sufficient to bind cellular NBS1 (Fig. 1 C, lane 6), and this binding did not further increase after including the more C-terminal sequences (Fig. 1 C, lanes 7 and 8). Interestingly, although MDC1 also contains an FHA domain at the very N terminus, this domain (either isolated or in the context of a larger fragment spanning up to the first 200 amino acids of MDC1) appeared to be entirely dispensable for NBS1 interaction (Fig. 1 C, lanes 1 and 2).
We then performed a reciprocal analysis and tested the structural requirements of NBS1 to interact with the MDC1 N terminus. To this end, we coexpressed the HA-tagged N-terminal fragment of MDC1 spanning the first 1,100 amino acids (p1100) with various Myc-tagged variants of NBS1 and assessed their interaction by an immunoprecipitation assay. Consistent with a previous study that measured the interaction of NBS1 mutants with full-length MDC1 (Lukas et al., 2004a), we found that a point mutation within the NBS1-FHA domain (NBS1R28) completely uncoupled NBS1 from the MDC1 N terminus (Fig. 1 E, lane 2). In contrast, neither disruption of the NBS1 C terminus responsible for interaction with MRE11 and ATM (NBS1ΔC) nor mutation of the three major phosphorylation sites targeted by the ATM kinase (NBS13A) had a measurable effect, and these mutants interacted with the MDC1 fragment as efficiently as the wild-type (WT) NBS1 (Fig. 1 E). Together, these data suggest that the MDC1 N terminus contains the necessary structural signature to allow recognition of MDC1 by the FHA domain of NBS1.
The SDT-rich region in the MDC1 N terminus mediates interaction with NBS1
Next, we subjected the N-terminal fragment of MDC1 to a bioinformatic analysis and searched for conserved motifs that may potentially mediate the interaction with NBS1. Interestingly, the region between amino acids 210 and 460 (that is, the part of MDC1 identified in the previous section as necessary and sufficient to bind NBS1) contained up to six acidic SDT motifs, some of which are conserved among different vertebrate species (Fig. 2 A and Fig. S1). To test whether the SDT clusters influence the MDC1–NBS1 interaction, we generated another set of MDC1 fragments either spanning different parts of the SDT-containing region or harboring an in-frame deletion of the entire SDT region (Fig. 2 B). The fragment containing all six SDT repeats efficiently bound to endogenous NBS1 when expressed in the U2OS/shRNA cell line (Fig. 2 C, lane 4). Strikingly, even shorter fragments containing either three N-terminal (SDT1) or four C-terminal (SDT2) repeats still interacted with NBS1, albeit with a lower efficiency than the full-length SDT region (Fig. 2 C, lanes 5 and 6). Conversely, deletion of the entire SDT region completely abolished the ability of the MDC1 N terminus to bind NBS1 under these conditions (Fig. 2 C, lane 3). Together, these data indicate that the SDT repeats within the MDC1 N terminus might be involved in binding to NBS1, and we set out to further explore this possibility.
The SDT repeats are phosphorylated in vivo
We reasoned that if the SDT repeats mediate the MDC1–NBS1 interaction, the Ser/Thr residues within the repeats should be phosphorylated in vivo. To test this important prediction, we analyzed phosphorylation patterns of various forms of MDC1 (Fig. 3 A) immunopurified from U2OS cells metabolically labeled with [32P]orthophosphate. As expected, the tryptic phosphopeptide map of endogenous MDC1 revealed considerable complexity, indicating that numerous phosphorylation events converge on these large adaptor proteins (Fig. 3 B). Strikingly, the majority of the phosphopeptide spots observed in the endogenous protein were preserved in the N-terminal fragment of MDC1 (Fig. 3 C), and some of the most prominent spots were clearly also resolved in the isolated SDT region (Fig. 3 E). Indeed, a systematic analysis of various deletion fragments and/or the SDT mutants in which the Ser/Thr residues were substituted to Ala confirmed that at least four of the prominent SDT clusters are subjected to phosphorylation in vivo (Table I). A typical example of such an assay is provided in Fig. 3 (F and G), where mutation of the second (amino acids 299/301) or the sixth (amino acids 453/455) SDT repeat eliminated the respective prominently labeled spot. Consistently, an N-terminal fragment of MDC1 terminating within the SDT region (at amino acid 400) lacked the spot corresponding to the sixth SDT (amino acids 453/455; Fig. 3 D). In addition, assessment of the spot charges after mutating single phosphoacceptor sites (Fig. S2 A) and a direct phospho–amino acid analysis of the prominently labeled SDT repeats (Ser299/Thr301 and Ser453/Thr455, respectively; Fig. S2 B) confirmed that both Ser and Thr residues were simultaneously phosphorylated. Together with recent results from the large-scale phosphoproteomic projects that independently confirmed phosphorylation of the SDT clusters (Table I), these data show that the SDT repeats within the MDC1 N terminus are indeed phosphorylated in vivo and suggest that the SDT-associated phosphates may serve as docking sites for the FHA domain of NBS1.
CK2 phosphorylates the SDT repeats and promotes interaction between the MDC1 N terminus and NBS1
To test this hypothesis, we first tried to identify the kinase that can phosphorylate the SDT repeats. Sequence analysis by the recently introduced NetworKIN approach (Linding et al., 2007) revealed that the SDT clusters strongly resemble consensus sites for the acidophilic protein kinase CK2. Therefore, from Escherichia coli, we purified a GST-tagged SDT fragment (spanning amino acids 181–480) and its variant in which the Ser/Thr residues in all six SDT repeats were changed to Ala (designated as 12A) and subjected these fragments to phosphorylation by CK2 in an in vitro kinase assay. Indeed, CK2 efficiently phosphorylated the WT SDT fragment in vitro (Fig. 4 A, lane 5), and this phosphorylation was inhibited by adding the CK2 inhibitor into the kinase reaction (Fig. 4 A, lane 8). Significantly, phosphorylation of the SDT-12A mutant was very inefficient (Fig. 4 A, lane 6), suggesting that the bulk of the CK2-mediated phosphorylations within the MDC1 N terminus was indeed targeted to the SDT repeats.
To test the impact of CK2-mediated phosphorylation of the MDC1 N terminus on its interaction with NBS1, we phosphorylated (or mock treated) the GST-tagged WT or the phosphorylation-deficient (12A) SDT fragments (using nonradioactive ATP in the kinase reaction) and assessed their ability to interact with an in vitro–translated 35S-labeled NBS1. Efficient phosphorylation of the WT (but not the 12A) SDT fragment was confirmed by a marked mobility shift on the SDS gel (Fig. 4 B, bottom). Strikingly, only the CK2-phosphorylated but not the unmodified SDT region interacted with NBS1 in this in vitro binding assay (Fig. 4 B, lane 3). Importantly, this interaction required the intact FHA domain of NBS1, as the R28 mutant of NBS1 failed to bind to the MDC1 N terminus even though this was efficiently phosphorylated (Fig. 4 B, lane 8). The SDT-12A mutant did not bind to any form of NBS1 under these conditions (Fig. 4 B, lanes 4, 5, 9, and 10).
To explore the emerging cross talk between CK2 and the N terminus of MDC1, we tested the impact of CK2 inhibition on SDT phosphorylation in cells. Indeed, preincubation of U2OS cells with a CK2 inhibitor significantly reduced the overall incorporation of radioactive phosphate to the ectopically expressed SDT fragment (Fig. 5 A). More importantly in this context, the ensuing phosphopeptide analysis revealed that this was accompanied by an ∼50% reduction of phosphate incorporation into two prominent SDT repeats (Ser299/Thr301 and Ser453/Ser455; Fig. 5 B). Thus, CK2 can at least partially phosphorylate the SDT repeats in vivo. To further support this conclusion, we immunopurified HA-tagged MDC1 N terminus (Fig. 6 A) or endogenous MDC1 (Fig. 6 B) from cells and subjected these immunocomplexes to an in vitro kinase assay (without any additional substrate). In both cases, MDC1 copurified with an autocatalytic kinase activity that was strongly attenuated by a specific CK2 inhibitor (Fig. 6, A and B). Interestingly, this MDC1-associated CK2 activity was clearly present in undamaged cells (Fig. 6, A and B; lane 1) and did not appreciably increase after exposing the cells to IR (Fig. 6 A, lane 2). Collectively, the data in Figs. 4–6,56 show that CK2 can phosphorylate the SDT repeats and suggest that this may contribute to trigger a productive interaction between the MDC1 N terminus and the FHA domain of NBS1.
The SDT repeats are required for MDC1–NBS1 interaction in vivo
To directly test whether the SDT repeats determine MDC1–NBS1 interaction in vivo, we expressed in U2OS cells the HA-tagged MDC1 N termini with progressively increasing mutations within the SDT repeats and assessed their ability to bind endogenous NBS1. The isolated mutations of the Ser/Thr residues within the second (amino acids 299/301) or the sixth (amino acids 453/455) SDT cluster had only marginal effects, and these mutants bound NBS1 almost as efficiently as the WT MDC1 N terminus (Fig. 7 A, lanes 1–3). Strikingly, a combination of these mutations (designated as 4A) markedly impaired the ability to bind NBS1 (Fig. 7 A, lane 4), and mutation of the Ser/Thr residues in all six SDT clusters (12A) abolished the MDC1–NBS1 interaction completely (Fig. 7 A, lane 5). Thus, consistent with the in vitro binding experiments described in Fig. 4, the integrity of the SDT repeats appears to be critical to support the interaction between MDC1 and NBS1 in cells. In addition, the progressive attenuation of this interaction achieved by increasing mutations of the SDT-associated Ser/Thr residues suggested that the SDT repeats (especially the dominant ones such as Ser299/Thr301 and Ser453/Thr455) bind NBS1 in a redundant fashion.
The SDT repeats determine the retention of NBS1 at the DSB-modified chromatin
Having obtained biochemical evidence for a link between the phosphorylated N terminus of MDC1 and the FHA domain of NBS1, we set out to investigate the functional relevance of these observations for chromatin retention of NBS1 at the DSB-modified chromatin. To this end, we again used the U2OS-derived cell line with stably depleted endogenous MDC1 (MDC1/shRNA; Fig. 1 D). To test the suitability of this cellular model for the in vivo reconstitution experiments with heterologous MDC1 alleles, we first confirmed that the endogenous MDC1 in these cells was indeed down-regulated to the degree that prevented any cytologically discernible accumulation of NBS1 in the laser-generated DSB tracks (Fig. S3). We then reconstituted this cell line with WT MDC1 or the phosphorylation-deficient variants in which the Ser and Thr residues were mutated in two (4A) or in all six (12A) SDT repeats. All of these MDC1 variants were generated in the context of full-length proteins, tagged with the HA epitope on the N terminus, and rendered insensitive to shRNA. Strikingly, although the WT and both phosphorylation-deficient forms of MDC1 avidly accumulated at the DSB-flanking chromatin, only the WT version was able to fully restore the retention of NBS1 in this compartment (Fig. 7 B, left). In contrast, cells expressing the MDC1-4A mutant retained only residual amounts of NBS1 at the DSB-modified chromatin (Fig. 7 B, middle), and MDC1-12A completely failed to support any cytologically discernible accumulation of NBS1 in these regions (Fig. 7 B, right). Importantly, these cell-based data closely matched the progressive loss of MDC1–NBS1 interactions in the pull-down assay using the analogous MDC1 phosphodeficient mutants (Fig. 7 A, compare lanes 1, 4, and 5). Together, these data suggest that the SDT-associated phosphorylations trigger productive interaction between MDC1 and NBS1 and facilitate retention of the latter protein at the DSB-modified chromatin.
Our results elucidate the nature of one of the most proximal steps in the spatiotemporal evolution of DSB-modified chromatin. Specifically, although availability of the MDC1 adaptor protein and the NBS1-FHA domain were known to regulate MRN chromatin retention, the way these two factors communicate with each other remained elusive. Our findings mechanistically explain this conundrum by identifying the conserved SDT repeats in the MDC1 N terminus and by showing that their phosphorylation generates recognition sites for the NBS1-FHA domain. In addition, we identify CK2 as a candidate kinase that can phosphorylate the SDT repeats and show that the integrity of the SDT-associated phosphoacceptor residues is required for MRN retention at the sites of DNA damage.
At first glance, it appears surprising that CK2, a ubiquitous and constitutively active kinase, regulates MDC1, a protein that functions almost exclusively in stochastic events triggered by genotoxic stress in which it operates strictly locally at the sites of chromosomal lesions. To comprehend this seemingly unusual cross talk, it is important to realize that MDC1 and NBS1 interact already in unstressed cells (Goldberg et al., 2003; Stewart et al., 2003), and, as we show in this study, even this predamage interaction is at least partly regulated by CK2. Based on this observation, the first obvious scenario that comes to mind is that as a result of its constitutive phosphorylation by CK2, MDC1 forms a stable holocomplex with MRN. In this model, the MRN focus formation simply reflects the propensity of MDC1 (as a fourth subunit of the MRN complex) to avidly interact with γ-H2AX. Although simple and attractive, several pieces of evidence suggest a more subtle way of regulation. First, MRN clearly has functions that do not require MDC1 (Petrini and Stracker, 2003; Falck et al., 2005; Jazayeri et al., 2006), and a strong affinity trap on SDT-phosphorylated MDC1 would likely reduce MRN's flexibility and limit its accessibility to nuclear compartments where it operates independently of MDC1. Indeed, kinetic analysis of NBS1 and MDC1 mobility in living cells revealed marked differences between these two proteins: although both NBS1 and MDC1 become retained at the DSB sites (a feature that determines formation of the cytologically discernible nuclear foci), the residence time of MDC1 at DSBs was nearly one order of magnitude longer compared with NBS1. As a result, although MDC1 becomes immobilized at the DSB-flanking chromatin, NBS1 remains highly mobile, a feature compatible with its flexible exchange among distinct DSB-generated subcompartments (Lukas et al., 2003, 2004a). Thus, the interaction between MDC1 and NBS1 is dynamic, and these two factors do not appear to form a rigid holocomplex.
Given these findings, the new data presented in this study, and similar results reached by Spycher et al. (see p. 227 of this issue), our current model describing the mechanism of MRN chromatin retention is as follows. In undamaged nuclei (Fig. 8, top), CK2 phosphorylates the SDT repeats of MDC1. Combined with the ability of the NBS1-FHA domain to recognize phosphorylated SDTs, this allows constitutive yet highly dynamic MDC1–MRN interaction, which remains amenable to further regulation and prevents sequestration of any of these factors in rigid aggregates. The latter feature could be crucial to allow the rapid recognition of DNA breaks by the MRN complex, a process that is independent of MDC1 (see Introduction). After DSB generation (Fig. 8, bottom), the most proximal chromatin modification includes phosphorylation of the H2AX C terminus (γ-H2AX). Although several DSB regulators contain phosphate interaction motifs and domains, the MDC1-BRCT domain appears to interact with γ-H2AX with the highest affinity (Stucki et al., 2005), resulting in a rapid coating of the γ-H2AX–primed chromatin by MDC1. At this point, the MDC1 N terminus becomes important for the ensuing events on the DSB-flanking chromatin (Fig. 8, bottom). We propose that constitutive phosphorylation of the SDT repeats (translated to a continuous generation of binding sites for the NBS1-FHA domain) is likely the most efficient means to avoid any delay in recruiting MRN to the modified chromatin rapidly evolving around the incipient DBS lesions. Indeed, integration of a constitutive signal is consistent with the simultaneous recruitment of MDC1 and NBS1 to DSBs (Lukas et al., 2004a; Mailand et al., 2007) and suggests that in vertebrates, the SDT repeats of MDC1 coevolved with MRN to prime the latter for the fastest possible arrival at the sites of harmful chromosomal lesions.
Although our data and the complementary results obtained by Spycher et al. (2008) strongly implicate CK2 in phosphorylating the MDC1 N terminus, we cannot exclude the possibility that other acidophilic kinases contribute to regulate the MDC1–NBS1 interaction. Most notably in this regard, we show in Fig. 5 that chemical inhibition of CK2 reduced but did not abolish phosphorylation of the prominently phosphorylated SDT repeats (the maximum reduction of phosphate incorporation to these repeats was ∼50% compared with CK2-proficient cells). Unlike complete genetic disruption of the dominant SDT repeats (Fig. 7 B), this degree of the SDT phosphorylation impairment was not sufficient to prevent accumulation of NBS1 at the DSB sites (our unpublished data).
As alluded to in the preceding paragraphs of this section, this model poses an important conceptual question: how is constitutive and ubiquitous phosphorylation of the MDC1 N terminus transformed to a signal that locally and transiently enhances chromatin affinity for NBS1? The answer to this question is not trivial, but several scenarios come to mind (Fig. 8, bottom). First, it is possible that just the sheer increase in the local density of MDC1 (and the corresponding phosphates on the SDT repeats) may be sufficient to delay MRN at the DSBs to an extent that still allows its dynamic exchange between distinct DSB-generated subcompartments yet prevents its dilution in the undamaged parts of the nucleus. Second, although CK2 activity does not increase after DNA damage, one can envisage that DSBs may locally regulate (inhibit) chromatin-associated phosphatases and thereby decrease the rate of SDT-phosphate turnover. In principle, this may also increase local density of phosphorylated SDT repeats and thereby stabilize the interaction between the MDC1 N terminus and the NBS1-FHA domain. Third, an alternative (but not mutually exclusive) model is based on our observation that the MDC1 N terminus harbors a cluster of six SDT sites and that at least some of them contribute to the MDC1–NBS1 interaction in a cooperative fashion (Fig. 7). Thus, it is possible that although only a subset of the SDT repeats generates the basal level of MDC1–MRN interaction in undamaged cells, the DSB-associated MDC1 undergoes additional conformational changes in its N terminus that may allow phosphorylation of the entire SDT region and, thereby, locally increase the interaction platform for NBS1 and indeed the entire MRN complex. Further studies will be required to discriminate among these scenarios.
Finally, we note that the role of CK2 in the DNA damage response is not unprecedented. The most striking parallel with our findings has been provided by Loizou et al. (2004), who reported that CK2 phosphorylates XRCC1, an adaptor protein involved in single-strand DNA break repair. This and the subsequent studies (Clements et al., 2004; Bekker-Jensen et al., 2007; Iles et al., 2007; Kanno et al., 2007) showed that CK2-mediated phosphorylation was required for the interaction of XRCC1 with other repair proteins such as polynucleotide kinase, aprataxin, and XIP1 (the latter protein is also known as APLF and PALF, respectively). It is noteworthy that all of these proteins interacted with XRCC1 by their FHA domains, which is in striking analogy to the MDC1–NBS1 interaction described in this study. In addition, an exciting conceptual parallel between the CK2-XRCC1 and CK2-MDC1 interplay is provided by the fact that in both cases, CK2 targets large scaffold proteins that by themselves do not posses enzymatic activity but rather appear to integrate upstream signaling to organize other proteins at the vicinity of DNA lesions. The fact that CK2 targets different adaptor proteins engaged in distinct DNA repair processes (single-strand breaks and DSBs) suggests that phosphorylation of these molecular adaptors coevolved to increase the efficiency of the genome surveillance machinery. Although the recent genetic data strongly support the role of MRN chromatin retention for survival and genomic stability after DSB-generating insults (see Introduction), the CK2-mediated phosphorylation of XRCC1 facilitates the single-stranded break repair by promoting the recruitment of polynucleotide kinase and aprataxin and by protecting XIP1 from proteolytic degradation (Clements et al., 2004; Loizou et al., 2004; Bekker-Jensen et al., 2007; Iles et al., 2007; Kanno et al., 2007). When combined together, the previous findings and our new data support the emerging model that one important function of CK2 is to increase the efficiency of genome surveillance by targeting interaction matchmakers and thereby promoting the local concentration and/or stability of DNA damage regulators at sites of genetic lesions.
Materials And Methods
Cell culture and gene transfer
The human osteosarcoma cell line U2OS (American Type Culture Collection) was cultured in DME (Invitrogen) supplemented with 10% FCS, 50 U/ml penicillin, and 50 mg/ml streptomycin. The U2OS stable derivative expressing the MDC1-targetting shRNA (U2OS/shRNA) was described previously (Bekker-Jensen et al., 2006). All transfections were performed with FuGENE6 (Roche) according to the manufacturer's instructions, and cells were harvested after 24 h.
HA-tagged human MDC1 (pcDNA3-neo) was a gift from D. Stern (Yale University, New Haven, CT). HA-MDC1 fragments were generated by PCR and subcloned into a pcDNA4/TO (Invitrogen) plasmid. To the shorter N-terminal MDC1 fragments, a triple nuclear localization signal was added by cloning the fragments into pCMV-myc/nuc (Invitrogen) vector and thereafter subcloning back into pcDNA4/TO by PCR. Myc-NBS1-pCMV plasmids were described previously (Horejsi et al., 2004). The QuikChange Site-Directed Mutagenesis kit (Stratagene) was used to generate point mutations; deletions were made using the Phusion Site-Directed Mutagenesis kit (New England Biolabs, Inc.) and performed according to the manufacturer's instructions. Plasmid constructs were verified by sequence analysis.
Antibodies, chemicals, and immunochemical techniques
HA antibodies F7 and Y11 were purchased from Santa Cruz Biotechnology, Inc. The rabbit Nbs1 and Myc antibodies originated from Abcam. Rabbit and sheep polyclonal MDC1 antibodies were provided by S. Jackson (Gurdon Institute, Cambridge, UK). The specific CK2 inhibitor InSolution DMAT was obtained from EMD and was used in concentrations specified in the figure legends. For immunoprecipitation, cells were lysed, and proteins were extracted in a high salt buffer (20 mM Tris, pH 7.5, 400 mM NaCl, 0.5% NP-40, 1 mM EDTA, and protease and phosphatase inhibitor mixture). After centrifugation of the lysates, an equal volume of the same buffer without NaCl was added to produce a final NaCl concentration of 200 mM. Kinase assays and immunoblotting were performed as previously described (Falck et al., 2005; Mailand et al., 2006) except for the MDC1-associated kinase assay, in which no exogenous substrate was added.
In vivo labeling and phosphopeptide mapping
U2OS cells transfected with the indicated constructs were labeled for 4 h in phosphate-free DME with 20 mM Hepes, pH 7.2, and 5% dialyzed phosphate-free DME with 1 mCi/ml 32P (PBS43; GE Healthcare). Two-dimensional phosphopeptide and phospho–amino acid mapping were performed as described previously (Blume-Jensen et al., 1995; Hansen et al., 1996). Tryptic phosphopeptides were separated on an electrophoresis system (HTLE-7000; CBS Scientific Company, Inc.) at 2,000 V for 30 min at 14°C. The second dimension chromatography was performed in isobutyric acid buffer for 14–16 h. Separated peptides were subjected to phosphorimager analysis (FLA-3000; Fuji).
CK2 kinase assay
1 μg of purified GST or GST-MDC1 fragments was phosphorylated with 100 U of recombinant CK2 kinase in the presence of 10 μCi γ-[32P]ATP for 15 min at 37°C according to the manufacturer's instructions. The reaction was stopped by the addition of Laemmli sample buffer, and phosphorylated proteins were visualized by resolving the samples on SDS-PAGE followed by autoradiography.
In vitro binding assay
2 μg of purified GST-MDC1(181–480) fragments were left untreated or phosphorylated with recombinant CK2 (New England Biolabs, Inc.) for 30 min at 37°C in the presence of 1 mM ATP. The GST proteins were then bound to glutathione–Sepharose beads (GE Healthcare), and the beads were washed three times in EBC buffer (Mailand and Diffley, 2005). The beads were then incubated with in vitro–translated 35S-labeled Nbs1 for 2 h at 4°C, washed four times with EBC buffer, resuspended in Laemmli sample buffer, and resolved by SDS-PAGE and autoradiography.
Laser microirradiation and microscopy
Induction of localized DSBs by microirradiation was performed as described previously (Lukas et al., 2003, 2004a; Bekker-Jensen et al., 2005, 2006). In brief, U2OS cells and their derivatives were grown on glass coverslips in the presence of 10 μM BrdU for 24 h and subsequently exposed to the pulsed UVA laser (λ = 337 nm) along a narrow track spanning the entire nuclear diameter. After an additional hour, the cells were fixed in 4% formaldehyde and subjected to immunostaining. Images were acquired with a confocal microscope (LSM 510; Carl Zeiss, Inc.) through a plan-Neofluar 40× NA 1.3 oil immersion objective (Carl Zeiss, Inc.). For dual color imaging, secondary antibodies coupled to AlexaFluor dyes with excitation wavelengths of 488 and 568 nm were used. Image acquisition and basic image processing were performed with LSM software (Carl Zeiss, Inc.).
Online supplemental material
Fig. S1 shows sequence alignment of the SDT-rich MDC1 regions in several vertebrate species. Fig. S2 provides in vivo evidence for the simultaneous phosphorylation of Ser and Thr residues within the individual SDT repeats. Fig. S3 shows that constitutive down-regulation of MDC1 by shRNA impairs retention of NBS1 at the sites of DNA damage.
We thank Manuel Stucki for stimulating discussions and communicating unpublished results, Claudia Lukas for comments and critical reading of the manuscript, Lars Ronnstrand for help with phosphopeptide analysis, and Stephen Jackson, Manuel Stucki, and David Stern for reagents.
This work was supported by grants from the Danish Cancer Society, Danish National Research Foundation, European Union (integrated project DNA Repair and Active p53), European Science Foundation (EuroDYNA), Danish Research Council, Swedish Society for Medical Research Stockholm, and the John and Birthe Meyer Foundation.
F. Melander and S. Bekker-Jensen contributed equally to this paper.
J. Falck's present address is Novo Nordisk A/S, DK-2880 Bagsverd, Denmark.
Abbreviations used in this paper: ATM, ataxia telangiectasia mutated; BRCT, BRCA1 C terminal; CK2, casein kinase 2; DSB, double-strand break; FHA, forkhead-associated; IR, ionizing radiation; MRN, MRE11–RAD50–NBS1; NBS, Nijmegen breakage syndrome; SDT, Ser-Asp-Thr; shRNA, short hairpin RNA; WT, wild type.