|Home | About | Journals | Submit | Contact Us | Français|
Oligomeric assembly of Brca1 C-terminal (BRCT) domain-containing mediator proteins occurs at sites of DNA damage. However, the functional significance and regulation of such assemblies are not well understood. In this study, we defined the molecular mechanism of DNA-damage-induced oligomerization of the S. cerevisiae BRCT protein Rad9. Our data suggest that Rad9’s tandem BRCT domain mediates Rad9 oligomerization via its interaction with its own Mec1/Tel1-phosphorylated SQ/TQ cluster domain (SCD). Rad53 activation is unaffected by mutations that impair Rad9 oligomerization, but checkpoint maintenance is lost, indicating that oligomerization is required to sustain checkpoint signaling. Once activated, Rad53 phosphorylates the Rad9 BRCT domain, which attenuates the BRCT-SCD interaction. Failure to phosphorylate the Rad9 BRCT results in cytologically visible Rad9 foci. This suggests a feedback loop wherein Rad53 activity and Rad9 oligomerization are regulated to tune the DNA-damage response.
Various endogenous and exogenous genotoxic insults induce DNA-damage checkpoint signaling. The biological outcomes of checkpoint signaling include the control and coordination of cell-cycle progression, transcription, DNA replication, DNA repair, and apoptosis. These functional endpoints preserve genomic integrity and suppress the evolution of genetically altered cells. These functions are likely to account for the fact that defects in checkpoint signaling are highly correlated with the process of malignancy.
As with other signal transduction pathways, protein phosphorylation is a key molecular event governing checkpoint signaling (Harper and Elledge, 2007; Kastan and Bartek, 2004). A primary biological function of phosphorylation in DNA-damage checkpoint signaling is the promotion of protein-protein interactions. During the response to DNA damage, numerous proteins are phosphorylated by the PI-3-like kinases (PIKKs), ATM, ATR, or DNA-PKcs in mammals and Tel1 and Mec1 in budding yeast, and the effector kinases, Chk1 and Chk2 in mammals and Chk1 and Rad53 in budding yeast (Bartek and Lukas, 2003; Matsuoka et al., 2007; Smolka et al., 2007). These phosphorylation events generate binding sites for forkhead-associated (FHA) and BRCT domains found in numerous DNA-damage response proteins as well as for 14-3-3 proteins. Engagement of the phosphorylated residues by those entities facilitates subsequent phosphorylation events and provides a means to regulate the subcellular localization of checkpoint effectors (Durocher et al., 2000; Fu et al., 2000; Glover et al., 2004).
The tandem BRCT-containing proteins, Brca1, Mdc1, and 53BP1 in mammals, Crb2 in S. pombe, and Rad9 in S. cerevisiae, mediate PIKK phosphorylation of the effector kinases Chk1 or Chk2/Rad53 (Navas et al., 1996; Saka et al., 1997; Stewart et al., 2003; Sweeney et al., 2005; Wang et al., 2002; Yarden et al., 2002). Those BRCT proteins become localized in cytologically observable foci at sites of DNA damage, suggesting that they act in relatively large assemblies that control effector kinase activities (Du et al., 2006; Lisby et al., 2004; Melo et al., 2001; Sanders et al., 2004; Schultz et al., 2000; Scully et al., 1997; Stewart et al., 2003; Toh et al., 2006).
To a large extent, studies of S. cerevisiae Rad9 underlie the canonical view of how mediator proteins function in the DNA-damage response. Upon DNA damage, the Rad9 SQ/TQ cluster domain (SCD) is phosphorylated by PIKKs, Mec1, and Tel1 (Figure 1A). One of these phosphorylation events creates a binding site for the FHA domain of the Rad53 effector kinase (Durocher et al., 1999; Emili, 1998; Schwartz et al., 2002; Sun et al., 1998; Vialard et al., 1998). Mec1 and Tel1 subsequently phosphorylate Rad53 that is associated with Rad9 (Schwartz et al., 2002; Sweeney et al., 2005). This event is followed by Rad53 autophosphorylation, which is required for full activation of the kinase (Chen et al., 2007; Fiorani et al., 2008; Pellicioli et al., 1999; Usui and Petrini, 2007). Oligomeric assembly of phosphorylated Rad9 appears to provide a platform on which high local concentration of Rad53 promotes that autophosporylation step (Gilbert et al., 2001). Consistent with this view, the assembly of Rad9 in chromatin after DNA damage is important for its mediator functions (Javaheri et al., 2006; Naiki et al., 2004; Wysocki et al., 2005).
Rad9 chromatin association is promoted by two of its protein domains, both of which appear to bind histone modifications (HMs). Rad9’s tandem BRCT and tudor domains (Figure 1A) bind to phosphorylated histone H2A S129 (Hammet et al., 2007), equivalent to phosphorylated histone H2AX S139 (γ-H2AX) in humans (Downs et al., 2000), and methylated histones, respectively (Grenon et al., 2007; Huyen et al., 2004). These HM-dependent mechanisms of Rad9 binding appear to be conserved in Crb2 and 53BP1, the Rad9 orthologs in S. pombe and mammals (Botuyan et al., 2006; Celeste et al., 2002; Du et al., 2006; Huyen et al., 2004; Nakamura et al., 2004; Sanders et al., 2004). HM-independent chromatin loading is also observed in some cases for Rad9 (Puddu et al., 2008) as well as Crb2 (Du et al., 2006). PIKK phosphorylation of the Rad9 SCD also influences the chromatin retention of Rad9 (Naiki et al., 2004; Toh et al., 2006); however, the molecular basis of this influence is unclear.
In this study, we examined the role of the S. cerevisiae Rad9 BRCT domain in the execution of the DNA-damage response. Previous studies indicate that DNA damage induces Rad9 dimerization via Rad9 BRCT domain interactions (Soulier and Lowndes, 1999). The data presented herein support that idea; however, we show that the Rad9 BRCT engages the Rad9 SCD in response to DNA damage, in contrast to the previous suggestion of DNA-damage-induced BRCT-BRCT interaction. The interaction requires SCD phosphorylation, consistent with the phosphopeptide-binding property ascribed to BRCT domains (Manke et al., 2003; Yu et al., 2003).
Our findings suggest that DNA-damage-induced Rad9 oligomer formation is not required for the initial activation of Rad53. However, BRCT-SCD-mediated interaction is required for maintenance of Rad53 activation and cell-cycle arrest. Rad9 oligomerization appears to occur in chromatin around DNA damage and is inhibited via Rad53-dependent phosphorylation of its BRCT domain. This phosphorylation event attenuates BRCT-SCD interaction. These data suggest a feedback loop in which Rad53 regulates the temporal and spatial properties of checkpoint activation.
To define the mechanism of DNA-damage-induced Rad9 oligomerization, we developed a system to isolate BRCT and SCD Rad9 protein domains that have undergone DNA damage-induced modifications in vivo. Isolation of these domains is effected by the integration of two tobacco etch virus (TEV) protease sites upstream of the BRCT domain (between Rad9 S990 and G991). HA epitope tags were included at the C terminus of RAD9 to create RAD9-TEV-HA (Figure 1A). The RAD9-TEV-HA strain was not sensitive to UV or MMS and was only mildly sensitive to X-irradiation (data not shown). DNA-damage-induced protein interactions were similarly unaffected, as Rad9-TEV-HA was phosphorylated by Mec1/Tel1 and bound to Rad53 in response to MMS (Figure 1B, lane 4, and data not shown). These data indicated that Rad9 functions are preserved in Rad9-TEV-HA.
TEV cleavage of Rad9-TEV-HA efficiently separated N-SCD and BRCT-HA. Rad9-TEV-HA was immunoprecipitated with HA from extracts of MMS-treated cells. Following TEV cleavage, phosphorylated N-SCD was detected in the supernatant of HA IPs by western blot (WB) with N-terminal-specific Rad9 antiserum (raised against Rad9 residues 1–504), whereas BRCT-HA remained bound to HA beads (Figure 1B).
DNA-damage-induced protein interactions of the Rad9 BRCT domain were assessed in pull-down assays. The RAD9-TEV-HA system was used to prepare N-SCD fragments, as described above. To isolate BRCT-HA, immunoprecipitations with N-terminal-specific Rad9 antiserum were carried out in extracts of control and MMS-treated cells, followed by TEV cleavage. In that condition, TEV liberates the BRCT-HA fragment in supernatants while N-SCD is retained in the pellet (Figure 1C, lanes 4 and 7). GST-BRCT pull-down assays were carried out from supernatants containing N-SCD or BRCT-HA. We first asked whether DNA damage induced interaction between GST-BRCT and BRCT-HA. BRCT-HA was not pulled down with GST-BRCT from MMS-treated or control cells (Figure 1C, lanes 5 and 8), suggesting that DNA damage does not induce the Rad9 BRCT-BRCT association.
In contrast, interaction between Rad9 SCD and GST-BRCT was induced by DNA damage. N-SCD supernatants obtained from asynchronous and MMS-treated cells (Figure 1D, lanes 4 and 7) were tested as above for interaction with GST-BRCT. Interaction was seen with N-SCD isolated from MMS-treated, but not untreated, cells. The interaction appeared relatively efficient, as the yield of N-SCD in the pull-down was greater than 10% of the starting material (Figure 1D, lane 8). Notably, the interaction was selective for the slower-migrating (phosphorylated) species of N-SCD (Figure 1D, compare lanes 7 and 8). The majority of Mec1/Tel1 phosphorylation sites within the SCD fall within residues 390–457, distal to T603, which falls with the Rad53 interaction domain (residues 542–620) (Schwartz et al., 2002) (Figure 1A), so that Rad9 BRCT and Rad53 binding need not be mutually exclusive.
The linker region between the tandem BRCT domains appears to stabilize the BRCT domain structure. On this basis, we reasoned that alterations in the linker might diminish phosphopeptide-binding capacity and so impair oligomerization. This region contains Rad9 Ser 1129, which is conserved in Crb2 and 53BP1 (Figure 1E) (Clapperton et al., 2004; Joo et al., 2002; Manke et al., 2003; Shiozaki et al., 2004; Williams et al., 2004; Yu et al., 2003). GST-BRCT carrying rad9-S1129A failed to pull down the PIKK-phosphorylated N-SCD (Figure 1D, lane 9). This suggests that the N-SCD pull-down depends on phosphopeptide binding of Rad9 BRCT. Supporting that interpretation, phosphatase treatment of Rad9-TEV-HA immunoprecipitated from MMS-treated cells abolished pull-down of N-SCD (see Figure S1 available online). Further supporting the requirement for phosphorylation in BRCT-SCD interaction, N-SCD obtained from MMS-treated mec1Δ tel1Δ cells or rad9-6AQ mutant cells where six SQ/TQ sites (T390, T398, T410, T427, S435, and T457) were mutated to AQ (Schwartz et al., 2002) did not interact with GST-BRCT (Figure 1F, lanes 6 and 8). These data demonstrate that the BRCT-SCD interaction depends on DNA-damage-induced phosphorylation of Rad9, and suggest that this interaction is the underlying molecular basis for Rad9 oligomerization in the course of checkpoint signaling.
The consensus phosphopeptide-binding site for the Rad9 BRCT domain is pS/pT-[YILQP]-I-I (Rodriguez et al., 2003). Four residues within the SCD conform to consensus: T390 (TQIV), T427 (TQII), S435 (SQGI), and T457 (TQII). The corresponding residues were restored one at a time in 6AQ mutant to determine the BRCT-SCD-binding specificity. We found that, when prepared from MMS-treated cells, N-SCD containing 5AQ with T427 added back partially recovered the ability to bind to GST-BRCT (Figure 2A, lane 8), whereas the remaining sites had little or no effect (Figure 2A). The apparent adherence of Rad9 BRCT binding to the consensus sequence is consistent with direct binding to the PIKK-phosphorylated Rad9 SCD.
To directly assess direct BRCT-SCD association, we carried out fluorescence polarization (FP) using purified GST-BRCT and a fluorescently labeled T427 phosphopeptide or nonphosphopeptide. We found that GST-BRCT WT imparts polarization of T427 phosphopeptides (Kd = 78 ± 18 μM), but not nonphosphopeptides (Figures 2B and 2C). Consistent with the pull-down assay, the phosphopeptide binding of GST-BRCT S1129A was reduced to a level that precluded reliable detection in this assay (Figure 2C).
To assess the functional significance of the SCD-BRCT interaction, we examined DNA-damage checkpoints in the rad9-S1129A and rad9-6AQ mutants. At the nonpermissive temperature, extensive telomere damage is induced in cdc13 mutants, resulting in a G2/M checkpoint arrest that is entirely dependent upon Rad9 (Gardner et al., 1999; Lydall and Weinert, 1995). Upon release from G1 arrest at nonpermissive temperature (37°), more than 90% of cdc13 cdc15 RAD9+ cells arrested in response to cdc13-induced telomere damage, whereas rad9Δ did not arrest, and nuclear division was evident by 4 hr after release 80% of the cells (Figure 3A). In this assay system, cells that fail to arrest are trapped at telophase due to the inactivation of cdc15, which is required for mitotic exit (Lydall and Weinert, 1995). rad9-S1129A and rad9-6AQ behaved as WT cells until 4 hr at 37°C, but nuclear division was evident in 60% of both mutants at 8 hr (Figure 3A), indicating that maintenance of the checkpoint is impaired in those mutants.
The BRCT-SCD interaction also influenced the response to interstial DNA damage. WT, rad9Δ, yku70Δ, rad9-S1129A, and rad9-6AQ strains carrying an unrepairable HO DSB site (Lee et al., 1998; Shroff et al., 2004) were arrested in G1 and plated on galactose-containing medium to induce HO endonulease expresssion. At 7 hr postinduction, roughly 80% of WT, rad9-S1129A, rad9-6AQ, and yku70Δ were present as microcolonies with two large-budded cells, whereas 84% of rad9Δ microcolonies contained more than three budded cells. At 13.5 hr, 43% of WT colonies remained arrested, whereas few rad9-S1129A and rad9-6AQ did, with 73% and 83% exhibiting microcolonies with more than three large budded cells (Figure 3B). Hence, the BRCT-SCD interaction’s influence on checkpoint maintenance is not specific to cdc13-induced telomere damage.
To confirm that Rad53 activation was intact in rad9-S1129A and rad9-6AQ, we inactivated CHK1 in those mutants. Rad9 promotes the activation of both Rad53 and Chk1 in response to DNA damage (Blankley and Lydall, 2004; Gardner et al., 1999; Sanchez et al., 1999); hence it was important to address the possibility that the residual checkpoint functions in rad9-S1129A and rad9-6AQ are entirely dependent on Chk1. This scenario predicts that Chk1-deficient rad9-S1129A and rad9-6AQ would phenocopy rad9Δ or rad53KD chk1Δ double mutants in cdc13 cdc15. Neither rad9-S1129A chk1Δ nor rad9-6AQ chk1Δ double mutants did so (Figure 3C), confirming that Rad53 is activated in rad9-S1129A and rad9-6AQ cells.
Having established genetic evidence that Rad53 was activated in the context of impaired BRCT-SCD interaction, we examined kinase activity in vitro. Auto- and transphosphorylation activity of Flag-Rad53 in rad9-S1129A and rad9-6AQ was assessed by IP kinase assays from extracts of cdc13 cells, prepared 2 hr after shifting to nonpermissive temperature. For both auto- and transphosphorylation activity, activation of Rad53 in rad9-S1129A was comparable to WT, whereas activation in rad9-6AQ was reduced by approximately 50% (Figures 4A and 4B).
Rad53 activation is correlated with hyperphosphorylation, and exit from arrest is correlated with disappearance of the hyperphosphorylated Rad53 species (Pellicioli et al., 1999; Sun et al., 1996). In rad9-6AQ and rad9-S1129A, wild-type levels of hyperphosphorylated Rad53 were detected at 2 hr after the shift to nonpermissive temperature (Figures 4A and 4C), consistent with the observation that checkpoint activation is unaffected by those mutations. However, hyperphosphorylated Rad53 began to wane at 4 hr in rad9-S1129A and rad9-6AQ and was undetectable by 8 hr. The dynamics of Rad53 hyperphosphorylation corresponded temporally with the checkpoint maintenance defect in rad9-S1129A and rad9-6AQ. Collectively, these genetic and biochemical data support the interpretation that BRCT-SCD-mediated Rad9 oligomerization is dispensable for Rad53 activation but required for the maintenance of Rad53 activation and the checkpoint-dependent cell-cycle arrest.
H2AX phosphorylation appeared to be dispensable for checkpoint maintenance, whereas histone methylation was required. We examined Rad53 activation and checkpoint activation in a cdc13 cdc15 hta1/2-S129* mutant that deleted the four residues (SQEL) of the Hta1 and Hta2 C termini including the PIKK phosphorylation site, S129 (Downs et al., 2000), and a deletion mutant of the histone methylase Dot1 that is important for Rad9 chromatin loading (Toh et al., 2006; Wysocki et al., 2005). Eight hours after shifting to nonpermissive temperature, a time point at which it was absent from rad9-S1129A and rad9-6AQ, hyperphosphorylated Rad53 persisted in hta1/2-S129* and wild-type, but not dot1Δ (Figure 4D). Checkpoint behavior mirrored Rad53 phosphorylation, as the cdc13-induced arrest was maintained in hta1/2-S129* and wild-type, but not in rad9-S1129A and dot1Δ mutants (Figure 4D), indicating that the functional impact of the BRCT-γ-H2AX interaction is distinct from that of BRCT-SCD and tudor-methylated histone interactions. Further, rad9-S1129A dot1Δ mutant showed an additive defect in checkpoint function (Figure S2), suggesting that BRCT-SCD interaction and DOT1-dependent Rad9 chromatin loading may independently influence Rad53 activity.
Crb2, the S. pombe Rad9 ortholog, exists in a constitutively dimeric or multimeric structure (Du et al., 2004; Kilkenny et al., 2008). We found that Rad9 was similar in this regard. HA-RAD9 and MYC-RAD9 were coexpressed, and recipriocal immunopreciptations from extracts of untreated as well as MMS-treated cells revealed constitutive association of the two species (Figure 4E and data not shown). However, unlike Crb2, which self-associates via its BRCT domain (Kilkenny et al., 2008), Rad9 self-association appears to be BRCT domain independent. This interpretation is suggested by the fact that GST-BRCT did not pull down BRCT-HA (Figure 1C), that the constitutive dimerization of Rad9 was not disrupted by a rad9-S1136A mutation that corresponds to dimerization-defective crb2-S666A (Kilkenny et al., 2008) (Figure 4E), and that neither rad9-S1129A nor rad9-6AQ affected constitutive interaction (Figure 4E). Therefore, the BRCT-SCD-mediated Rad9 oligomerization induced by DNA damage appears to reflect the oligomerization of dimeric (or multimeric) assemblies of Rad9.
This interpretation requires that rad9-S1129 and rad9-6AQ would exhibit intragenic complementation, as dimers containing the two gene products would contain WT BRCT domains (from Rad9-6AQ) and WT SCDs (from Rad9-S1129A). To test this, RAD9 and rad9-S1129A were expressed in rad9-6AQ cdc13 cdc15 cells. In rad9-S1129A-expressing rad9-6AQ cells, checkpoint maintenance was indistinguishable from cells expressing WT RAD9, indicating that rad9-S1129A/rad9-6AQ heterodimers can maintain checkpoint activation as well as WT (Figure 4F). These data demonstrate that DNA-damage-induced BRCT-SCD interactions promote oligomerization of mulitmeric Rad9 in vivo and exclude the possibility that a folded structure of Rad9 (via intramolecular interaction of BRCT and SCD) is functional.
Many DNA-damage response proteins form DNA-damage-induced nuclear foci (Lisby and Rothstein, 2004). Available evidence suggests that a minimum of seven protein molecules must be closely spaced in order for these proteins to be visible in the light microscope (Joglekar et al., 2006). We reasoned that the BRCT-SCD-mediated Rad9 oligomerization might affect the formation or persistance of DNA-damage-induced Rad9 foci.
RAD9-YFP (yellow fluorescent protein)- and rad9-6AQ-YFP-expressing cdc13 cdc15 cells were released at nonpermissive temperature from G1 arrest to induce DNA damage. We observed that a small fraction of WT as well as rad9-6AQ cells (less than 10%) formed nuclear Rad9-YFP foci, as previously reported (Melo et al., 2001). In both cases, Rad9 signal was too dim to reliably identify focus-positive cells (Figures 5A and 5B). Nevertheless, rad9-S1129A and rad9-6AQ did not interfere with Rad9 chromatin binding in response to DSBs induced by HO endonuclease when examined by chromatin immunoprecipitation (ChIP) (Figure S3 and data not shown). Hence, BRCT-SCD-mediated Rad9 oligomerization does not appear to exert a strong influence on the extent of Rad9 DNA-damage association.
However, Rad53 activity exerted a strong negative influence on Rad9’s accumulation in DNA-damage foci. In contrast to WT (RAD53) in which Rad9 foci were not scoreable, we found that rad53-KD cells formed easily detected Rad9 foci at 4 hr after shift to nonpermissive temperature (Figures 5C and 5D). We observed similar genetic dependencies in profiles of X-irradiation-induced Rad9 focus formation (Figure S4). PIKK phosphorylation of Rad9 SCD, and thus BRCT-SCD association, is responsible for the effect, as rad9-6AQ rad53-KD reduced the number of foci-positive cells (17.6%) at 4 hr (p = 0.03, Figure 5D and Figure S5).
The influence of Rad53 was also evident by ChIP. Quantitative PCR of anti-HA chip showed that HA-Rad9-bound DNA at 0.05 kb from the HO-induced DSB sites increased 11-fold in WT 3 hr after DSB induction, compared to a 25-fold increase of HA-Rad9-bound DNA in rad53-KD (p = 0.006) (Figure 5E and Figure S3). The effect of Rad53 on chromatin association appeared to be dependent on Rad9 BRCT function, as no enhancement of Rad9 ChIP in rad53-KD rad9-S1129A relative to rad53-KD was detected (p = 0.004) (Figure 5E). These data suggest Rad53 activity suppresses the accretion of Rad9 oligomers at sites of DNA damage via an effect on the BRCT-SCD association.
We found that the mechanism of Rad53’s influence on Rad9 DNA-damage association was effected via phosphorylation of the Rad9 BRCT domain. Rad53 immunoprecipitates from extracts of MMS-treated cells were incubated in kinase assay conditions. Consistent with the previous reports (Jia-Lin Ma and Stern, 2008; Lee et al., 2003), Rad9 was phosphorylated in a Rad53-dependent manner (Figure 6A, lanes 2, 4, and 6). To localize the phosphorylated domain of Rad9, we constructed GST fusions expressing five segments of Rad9 (Rad9-A to Rad9-E) as substrates for in vitro kinase assays using bacterially produced Rad53 (His-Rad53; Figure 6B). The GST-Rad9-D fragment (residues 991–1309) comprising the Rad9 BRCT was most efficiently phosphorylated by His-Rad53 (Figure 6B).
To determine whether Rad9 BRCT is phosphorylated in vivo, we assessed mobility shift of the BRCT-HA fragment released from Rad9-TEV-HA by TEV protease. A lower-mobility BRCT-HA species became evident when prepared from MMS-treated wild-type cells (Figure 6C, lane 8, and Figure 1B, lane 8). Phosphatase treatment abolished the lower-mobility form (Figure 6D, lane 5), indicating that Rad9 BRCT is phosphorylated in response to MMS treatment. We did not detect phosphorylated BRCT-HA in rad53-KD cells treated with MMS (Figure 6C, lane 10), indicating that phosphorylation of Rad9 BRCT depends on Rad53 kinase activity.
Rad9 BRCT phosphorylation was not detected in mec1Δtel1Δ cells, a setting in which both Rad53-Rad9 interaction and Rad53 activation are blocked (Vialard et al., 1998) (Figure 6C, lane 12). Phosphorylation was also undetectable in rad53-T354A T358A (rad53-TA) and rad53-T354A cells expressing hypomorphic rad53 alleles (Chen et al., 2007; Fiorani et al., 2008; Usui and Petrini, 2007) (Figure 6E, lanes 10 and 11), indicating that phosphorylation of Rad9 BRCT requires fully activated Rad53. In contrast, deficiency of Dun1 kinase, which is downstream of Rad53 (Allen et al., 1994; Chen et al., 2007), did not compromise phosphorylation of BRCT-HA (Figure 6E, lane 12). These data suggest that Rad9 BRCT is a direct target of activated Rad53 kinase after DNA damage. Mass spectrometry of Rad9 BRCT domains prepared from extracts of damaged cells to identify the residue(s) modified by Rad53 in vivo was unsuccessful. We find that Rad53-phosphorylated Rad9 BRCT becomes relatively resistant to protease cleavage, suggesting that phosphorylation imparts a structural change in the BRCT domain (Figure S6).
To define the functional significance of BRCT domain phosphorylation by Rad53, we examined its effect on BRCT-SCD interaction. We phosphorylated GST-BRCT by His-Rad53-WT and carried out GST pull-down assays from supernatants containing PIKK-phosphorylated N-SCD as above. The interaction with Mec1/Tel1-phosphorylated N-SCD was strongly inhibited (Figure 7A, reaction Z, lane 3), while treatment with His-rad53-KD did not affect the ability of GST-BRCT to pull down the N-SCD (reaction X, lane 1). Phosphatase treatment reversed both mobility shift and inhibitory effect of the reaction Z-treated GST-BRCT (Figure 7A, lane 5). Consistent with this result, phosphorylated GST-BRCT exhibited minimal binding to T427 phosphopeptide in the fluorescence polarization assay (Figure 7B). These data indicate that binding to PIKK-phosphorylated Rad9 SCD is inhibited by Rad53-mediated phosphorylation of the Rad9 BRCT domain and support the interpretation that Rad53 regulates BRCT-SCD-mediated Rad9 oligomerization induced by DNA damage.
Here we describe a molecular mechanism of DNA damage-induced Rad9 oligomerization and describe its role in the regulation of the DNA-damage checkpoint. Our data suggest that DNA damage, and the ensuing Mec1/Tel1 phosphorylation of the Rad9 SCD, induces interaction between the Rad9 BRCT domain and the phosphorylated SCD. This mode of Rad9 oligomer formation is dispensable for initial Rad53 activation but required for maintenance of the checkpoint. This mode of intermolecular association contrasts the BRCT-BRCT interaction previously proposed to account for DNA-damage-induced Rad9 self-association (Soulier and Lowndes, 1999), although the data supporting the previous model are also consistent with the model proposed here. Unlike BRCT-BRCT interaction, intermolecular BRCT-SCD interaction could in principle accommodate a broad range of oligomerization states of Rad9 (Figure 7C) and permit a commensurately broad range in the amplitude of Rad9-dependent checkpoint signaling. Finally, we show that Rad53, once fully activated, inhibits further oligomerization by phosphorylation of the Rad9 BRCT domain, suggesting a feedback mechanism of regulation. The details of this model are described below.
Rad9 DNA-damage association precedes or is coincident with its initial phosphorylation by Mec1/Tel1 (Hammet et al., 2007; Javaheri et al., 2006; Toh et al., 2006; Wysocki et al., 2005) (Figure 7Ci). Subsequently, “naive” (i.e., hypophosphorylated) Rad9 species could engage via BRCT-SCD interaction, an event that would in turn potentiate SCD phosphorylation as well as Rad53 recruitment (Figures 7Cii and 7Ciii). This aspect of the model is consistent with the observation that BRCT-SCD-mediated Rad9 oligomerization is not required for Rad9 DNA-damage association, and with the observation that SCD phosphorylation is less robust in rad9-S1129A mutants, a setting in which the oligomerization of Rad9 would be reduced (Figure S7). Thus, DNA-damage-induced Rad9 oligomerization allows amplification and maintenance of phosphorylated (activated) Rad9 pool and thereby sustained activation of Rad53. This interpretation is supported by the fact that mutations that impair BRCT-SCD interaction result in precocious release from checkpoint arrest and failure to maintain Rad53 activation.
We propose a feedback loop in which activated Rad53 phosphorylates the Rad9 BRCT domain and contributes to the turnover of Rad9 oligomers by suppressing BRCT-SCD-mediated Rad9 oligomerization (Figure 7D), thereby promoting release of PIKK-phosphorylated Rad9. This regulatory step may account for the observation that cdc13- or ionizing radiation-induced foci were difficult to detect in WT cells (Figure 5A and Figure S4) (Melo et al., 2001; Toh et al., 2006), whereas Rad9 foci were readily apparent in rad53-KD cells (Figure 5C). We favor the view that Rad9 DNA-damage association is transient and highly dynamic, accounting for the fact that cytologically visible Rad9 assemblies are rare and that this behavior reflects Rad53’s inhibition of oligomerization. Our data suggest that impairment of HM-dependent chromatin association does not markedly affect the activation of Rad9-dependent checkpoints (Hammet et al., 2007; Javaheri et al., 2006; Lazzaro et al., 2008; Toh et al., 2006; Wysocki et al., 2005). This raises the possibility that non-chromatin-associated Rad9 may exert a substantial influence on checkpoint signaling.
The inhibition of BRCT-SCD-mediated Rad9 oligomerization by Rad53 may be relevant in two nonexclusive steps, both of which could account for its effect on checkpoint maintenance. First, Rad53 may phosphorylate pre-existing BRCT-SCD-associated Rad9 species to promote their disassembly (Figure 7Di). Second, phosphorylation of non-SCD-engaged Rad9 BRCT may inhibit its recruitment into oligomer assemblies (Figure 7Dii). In either case, blocking Rad9 oligomerization may increase accessibility of phosphorylated SCD to promote the Rad9-Rad53 complex formation and sustain the pool of activated Rad53. It is also likely that the regulation of the Rad9 oligomer formation in chromatin may prevent interference with other chromosome metabolism processes after DNA damage (e.g., DNA repair, chromatin remodeling). Finally, an appealing possibility is that Rad53’s liberation of phosphorylated Rad9 may facilitate its interaction with other components of the DNA-damage response.
Both S. cerevisiae Rad9 and S. pombe Crb2 are constitutively dimeric although the data do not exclude higher order constitutive association; however the molecular determinants of dimerization appear to differ. Rad9 S1136, localized in the BRCT linker corresponds to Crb2 S666 which is required for constitutive dimerization (Kilkenny et al., 2008). Whereas crb2-S666A is checkpoint deficient (Kilkenny et al., 2008), we found that rad9-S1136A had no effect on dimerization, phosphopeptide binding, or checkpoint functions (Figure 4E and Figure S8). This may suggest that Rad9 self-associates in a BRCT-independent manner, as reported for 53BP1 (Ward et al., 2006). Although Rad9 S1136 is also identified as a DNA-damage-induced phosphorylation site (Albuquerque et al., 2008), mass spectrometry analysis of BRCT-HA shows that the phosphorylation of S1136 is Rad53 independent (data not shown).
The functional relationship of Rad9 and Rad53 is likely to be analogous in their human orthologs, 53BP1 and Chk2. In human cells, 53BP1 and Chk2 appear to be less important for the regulation of cell-cycle progression after DNA damage than for DNA repair and apoptotic induction. Nevertheless, 53BP1 focus formation and Chk2 activation are observed in damaged human cells as well as in preneoplastic lesions, leading to the hypothesis that the DNA-damage response is an inducible barrier to malignant progression (Bartkova et al., 2005; Gorgoulis et al., 2005). In this regard, the molecular mechanisms described here may also be relevant to chromosome dynamics and repair, as well as apoptotic regulation and tumor suppression by the mammalian DNA-damage response network.
The details are mentioned in the Supplemental Data.
cdc13 cdc15 assay was performed as mentioned (Lydall and Weinert, 1995). Rad9-YFP was observed as described (Burgess et al., 2007). The exposure time was 3.5 s. We used ROI stamp tool of the Volocity software (Improvision Ltd.) to determine focus-positive cells. All cells were marked with the tool (5 × 5 pixels) at possible focus regions or nonfocus regions, if cells did not have the apparent regions, and quantified for fluorescent intensity. An arbitrary threshold number, 1070 (AU/pixel), was set based on the samples at time 0. We considered the one above the threshold a focus-positive cell. P value was calculated using the two-tailed Wilcoxon rank sum test. Micorocolony formation assay was carried out as follows. Cells were grown in YP-lactate media and arrested at G1 by α factor. After washing out α factor, 4 × 105 cells were plated on a 10 cm Petri dish of YP+2% galactose media, followed by microscopic observation at the indicated time after plating. More than 95% of cells were observed as isolated single G1 cells on the 10 cm dish at time 0.
Yeast cell extracts were prepared, and immunoprecipitation (IP) and Flag-Rad53 IP kinase assay were performed as described (Usui and Petrini, 2007). At least 6 mg of cell extract including Rad9-TEV-HA was incubated with 5 μg of anti-HA mouse monoclonal antibody (mAb) (12CA5, MSKCC mAb core facility) bound to ProteinG beads (Calbiochem). After washed, anti-HA IPs were incubated in 150 μl of TEV buffer (10 mM Tris-HCl [pH 8.0], 0.5 mM EDTA, 1 mM DTT, 0.1% NP-40, 150 mM NaCl, 2 mM NaF, 5 mM β-glycerophosphate) with AcTEV protease (Invitrogen, 20 units/IP) at 30°C for 1.5 hr. Supernatants that contain N-SCD were collected and incubated with 0.1 μg GST-BRCT proteins bound to 10 μl of glutathione Sepharose 4B (GE Healthcare) at 4°C for 2 hr. To obtain BRCT-HA in supernatants, IP was carried out with anti-Rad9 rabbit polyclonal antibody (UWM60). The GST-BRCT-bound beads were washed with TEV buffer and analyzed by WB with anti-Rad9 (UWM60), anti-GST (GE Healthcare), or anti-HA (Bethyl Laboratories, Inc) antibodies. To test the effect of Rad53 phosphorylation of Rad9 BRCT in the GST pull-down assay, 0.1 μg GST-BRCT was incubated with 0.05 μg His-Rad53-WT or -KD in kinase buffer (50 mM HEPES-NaOH [pH 7.9], 50 mM KCl, 0.1% NP-40, 2% glycerol, 10 mM MgCl2, 5 mM ATP) at 30°C for 1 hr and purified by 10 μl of glutathione beads. For phosphatase treatment, the GST-BRCT beads were incubated with 10 units of CIP (NEB) at 37°C for 20 min with or without 10 mM NaV. The GST-BRCT beads were extensively washed prior to pull-down assay.
5-FAM-labeled phospho- or nonphospho-T427 peptides (200 nM) (KAELET QIIAK) were incubated with GST fusion proteins at the various concentrations in binding buffer (50 mM HEPES-NaOH [pH 7.5], 150 mM NaCl, 4 mM DTT). Fluorescent polarization (FP) was measured on SpectraMax M5 (Molecular Devices) using 495 nm excitation and 525 nm emission, and the data were analyzed by Prism (GraphPad Software Inc).
Cells were arrested at G2/M by nocodazole (15 μg/ml) in YP-lactate. Galactose was added to induce HO-DSBs. After cells were fixed with 1% formaldehyde for 15 min, chromatin solution was prepared from 5 × 108 cells as essentially described (Aparicio et al., 2004), followed by IP with 0.5 μg anti-HA mAb (12CA5). Immunoprecipitated DNA was purified as described (Shroff et al., 2004) and quantified by real-time PCR as described (Kim et al., 2008). The data were presented as fold increase after HO-DSB formation, normalized to time 0. P value was calculated using two-tailed Wilcoxon rank sum test.
This work was supported by GM56888, GM59413, and the Joel and Joan Smilow Initiative (J.H.J.P.). T.U. was a special fellow of the Leukemia and Lymphoma Society. The authors thank members of their laboratories for insights and A. Koff, X. Zhao, and C. Lima for critical reading of the manuscript. They are also grateful to H. Erdjument-Bromage and P. Tempst for mass spectrometry analysis, G. Bryant and D. Spagna for real-time PCR analysis, J. Chen and V. Yong-Gonzalez for YFP observation, and J. Gareau for FP. The authors would like to thank San San Yi at Memorial Sloan-Kettering Cancer Center (MSKCC) Microchemistry and Proteomics Core Laboratory (supported by National Cancer Institute [NCI] Cancer Center Support Grant P30 CA08748) for the synthesis of the fluorescent peptides.
The Supplemental Data include eight figures and Supplemental Experimental Procedures and can be found with this article online at http://www.cell.com/molecular-cell/supplemental/S1097-2765(08)00892-7.