|Home | About | Journals | Submit | Contact Us | Français|
The Nijmegen breakage syndrome 1 (Nbs1) subunit of the Mre11-Rad50-Nbs1 (MRN) complex protects genome integrity by coordinating double-strand break (DSB) repair and checkpoint signaling through undefined interactions with ATM, MDC1, and Sae2/Ctp1/CtIP. Here, fission yeast and human Nbs1 structures defined by X-ray crystallography and small angle X-ray scattering (SAXS) reveal Nbs1 cardinal features: fused, extended, FHA-BRCT1-BRCT2 domains flexibly linked to C-terminal Mre11- and ATM-binding motifs. Genetic, biochemical, and structural analyses of an Nbs1-Ctp1 complex show Nbs1 recruits phosphorylated Ctp1 to DSBs via binding of the Nbs1 FHA domain to a Ctp1 pThr-Asp motif. Nbs1 structures further identify an extensive FHA-BRCT interface, a divalent MDC1-binding scaffold, an extended conformational switch, and the molecular consequences associated with cancer predisposing Nijmegen breakage syndrome mutations. Tethering Ctp1 to a flexible Nbs1 arm suggests a mechanism for restricting DNA end processing and homologous recombination activities of Sae2/Ctp1/CtIP to the immediate vicinity of DSBs.
The Mre11-Rad50-Nbs1 (MRN) complex coordinates detection, signaling, and repair of cytotoxic DNA double-strand breaks (DSBs). Mutation of MRN subunits in Saccharomyces cerevisiae and Schizosaccharomyces pombe causes genotoxin sensitivity and meiotic defects resulting from defective processing of programmed DSBs (reviewed in Williams et al., 2007). In humans, congenitally acquired hypomorphic NBS1 and MRE11 mutations cause radiosensitivity and chromosome instability syndromes, Nijmegen breakage syndrome (NBS) and ataxia-telangiectasia-like disorder (ATLD), respectively (Carney et al., 1998; Stewart et al., 1999).
MRN possesses critical end-bridging and endonucleolytic activities for homologous recombination (HR) repair of DSBs (Williams et al., 2008). MRN also acts in checkpoint signaling by targeting the ATM/Tel1 kinase to DSBs by binding Nbs1 C-terminus (Falck et al., 2005; You et al., 2005). ATM phosphorylates the histone H2AX C-terminal tail near DSBs to form the γH2AX chromatin modification (Stucki and Jackson, 2006). In mammals, checkpoint adaptor MDC1 binds γH2AX, facilitating the intra-S-phase checkpoint (Goldberg et al., 2003; Lou et al., 2006; Stucki et al., 2005). MDC1 then recruits added MRN-ATM complexes via phosphorylation-dependent interactions with Nbs1 FHA-BRCT1-BRCT2 domains (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008) to amplify or sustain checkpoint signals (Lou et al., 2006). However, no MDC1 homologs are known in lower eukaryotes, so other ancestral Nbs1 FHA-BRCT1-BRCT2 domain functions likely exist.
MRN activities in HR involve the diverged protein known as Ctp1 in S. pombe (Limbo et al., 2007), Sae2 in S. cerevisiae, and CtIP in mammals (Sartori et al., 2007). Sae2 cleaves DNA substrates in vitro (Lengsfeld et al., 2007), consistent with 5′ to 3′ single-strand resection of DSBs in vivo to generate 3′ ssDNA tails for HR (Mimitou and Symington, 2008; Zhu et al., 2008). Ctp1Sae2 is required for removal of proteins linked to the 5′-end of DSBs (Neale et al., 2005), including complexes formed by the anticancer drug etoposide and topoisomerase II (Hartsuiker et al., 2009). Ctp1Sae2 activities appear conserved in mammals, and CtIP is a tumor suppressor in mice (Chen et al., 2005). In human cells, CtIP interacts with MRN (Sartori et al., 2007), and co-precipitates with Nbs1 when co-expressed in insect cells, suggesting CtIP binds Nbs1 (Chen et al., 2008), but the mechanism and significance of this interaction are unknown.
Whereas microscopic and crystallographic analyses of Mre11 and Rad50 were integral in defining MRN DNA bridging and nuclease processing activities (Hopfner et al., 2002; Hopfner et al., 2001; Williams et al., 2008), how Nbs1 physically integrates DNA end sensing and processing with cell cycle checkpoint activation remains a mystery. Here we report combined structural and functional dissection of fission yeast and human Nbs1. X-ray structures and small angle X-ray scattering (SAXS) results show an S. pombe Nbs1 folded core is conserved in human Nbs1, and that disease-associated Nbs1 mutations fall into the folded N-terminal FHA-BRCT1-BRCT2 domains defined by our structures. Our results further uncover the basis for phosphorylation-dependent recruitment of Ctp1 to DSBs by Nbs1. Nbs1 emerges as an extended, flexible binding nexus coordinating interactions between the Mre11-Rad50 core, the checkpoint kinase ATM/Tel1, and the DNA end-processing factor Ctp1.
To define functional Nbs1 regions, we expressed a series of truncated S. pombe Nbs1 (spNbs1) proteins and mapped structured domains with proteolysis (Figures 1A and 1B). Full-length spNbs1 expresses at low levels in E. coli, but a 60 residue C-terminal deletion (residues 1-553, Nbs1-ΔAT), removing the ATM/Tel1 binding motif (AT) but retaining an Mre11 binding consensus motif (MB) (Falck et al., 2005; You et al., 2005), facilitates robust recombinant over-expression. Trypsin degradation of Nbs1-ΔAT reveals a stable 37kDa folded core (Nbs1-fc, residues 1-329) encompassing the Nbs1 FHA domain and its two BRCT domains (Figures 1A and 1B). N-terminal sequencing of smaller transiently stable trypsin products maps three additional cleavage sites to a divergent sequence linking BRCT2 and the MB motif. Thus, the N-terminal 55% of spNbs1 comprises a composite FHA-BRCT1-BRCT2 folded Nbs1-fc that is linked to the C-terminal Mre11 and ATM binding regions via a protease-sensitive linker peptide. Like human Nbs1 (Difilippantonio et al., 2005), but unlike S. cerevisiae Xrs2Nbs1 (Shima et al., 2005), the integrity of S. pombe Nbs1-fc is critical for MRN complex DSB repair functions (Figure S1).
To examine Nbs1 structure-function, we determined the 2.8 Å resolution X-ray structure of spNbs1-fc with selenomethionine multi-wavelength anomalous dispersion (see Experimental Procedures), (Table S2 and Figure S2). Overall, Nbs1-fc is an elongated, distorted cylinder with dimensions ~35 by ~40 by ~95 Å, composed of joined FHA, BRCT1, and BRCT2 domains (Figure 1C). FHA and BRCT repeat domains mediate phosphorylation-regulated protein interactions, and are ubiquitous amongst DNA repair proteins (Glover et al., 2004; Williams et al., 2004). Like budding yeast Rad53-FHA1 (Durocher et al., 1999) or murine PNK-FHA (Bernstein et al., 2005), the Nbs1 FHA domain folds as a canonical FHA β-sandwich with two opposed 5-stranded anti-parallel β–sheets, and distal variable phosphoprotein-interacting β-strand connector loops (Figure S3).
The tandem Nbs1 BRCT domains adopt BRCT α-β sandwich folds with a central 4-stranded parallel β-sheet surrounded by 3–4 helices, and orient relative to one another with a triple helical hydrophobic inter-BRCT repeat interface. Although similar in structure to hBRCA1 BRCT repeats (Williams et al., 2001), the Nbs1 inter-BRCT repeat angle is offset by ~40 ° relative to BRCA1 tandem BRCTs (Figure S4). Sequence comparisons show that BRCT phosphoprotein interaction motifs (a conserved S/T at the C-terminus of the first β-sheet in BRCT1 (β11-α2 junction in Nbs1), and a TXK motif at the N-terminus of the second helix in the BRCT (α3 in Nbs1) (Clapperton et al., 2004; Williams et al., 2004)) are absent in S. pombe Nbs1, but present in the human, mouse, chicken, and frog Nbs1 homologs (Figure S3).
Despite similarities to known domain structures, key Nbs1 features suggest distinct functions in coordinating DNA repair signaling pathways. Notably, Nbs1 is configured as a divalent binding scaffold to mediate association with phosphorylated binding partners via four possible binding modes (Figure 1E): i) FHA-only binding, ii) BRCT-only binding, iii) simultaneous FHA and BRCT binding for alignment of two distinct binding partners, or iv) recognition of tandem phosphorylation clusters in one target protein via simultaneous FHA/BRCT binding. The phosphoprotein binding sites (spaced ~80 Å apart) could be spanned by an ~22 amino acid spacing (3.8 Å/residue) between phosphorylated residues within a single protein.
A defining Nbs1-fc feature is its extensive FHA-BRCT interface. The FHA N- and C-termini pack at the FHA-BRCT1 interface with N-terminal methionine and the invariant Trp2 buried within the protein core (Figure S5A). Attachment of affinity or epitope tags to the Nbs1 N-terminus (not shown) or a Trp2 to Ala mutation renders spNbs1 non-functional in vivo, suggesting that the FHA-BRCT1 interface is critical for structure and function (Figure S5B). Close FHA and BRCT1 domain juxtaposition results from a short 2-residue linker between the FHA-β10 and BRCT1-β1 strands, and by an Nbs1-specific helical insertion (helix α1) between FHA β-strands β5 and β6 (Figure 1D). At the FHA-BRCT1 boundary, an uninterrupted network of hydrophobic residues from the FHA (Met1, Ile17, Phe19, Cys66, Pro55) and BRCT1 (Met116, Leu136, Pro139, Ala147, Ser145, Arg143), and two flanking salt-bridges (Lys14-Asp194 and Glu57-Arg143) stabilize the FHA-BRCT1 interface. The large buried surface area (~2150 Å2) for the FHA-BRCT1 interface suggests the relative positioning of the FHA-BRCT1-BRCT2 scaffold aligns phosphoprotein interaction sites on opposite faces of Nbs1.
To better understand NBS mutations, we analyzed the human Nbs1-fc structure with SAXS and proteolysis. Human Nbs1 (hNbs1) forms a trypsin-stable composite FHA-BRCT1-BRCT2 core (hNbs1-fc, residues 1-334) similar to spNbs1-fc (Figure 2A). As all parts of a structure contribute to the experimental X-ray scattering, SAXS provides an objective, quantitative measure of structural similarity (Putnam et al., 2007). SAXS data of hNbs1-fc closely matches calculated scattering by the spNbs1-fc X-ray structure (χ2= 2.00), revealing conservation of structure despite low sequence conservation of Nbs1-fc (11% identity, 25% similarity). Yet, structure-based sequence alignments show an insertion missing in the spNbs1-fc model corresponds to an α-helix usually found between the third and fourth BRCT domain β-strands (β17 and β18 of BRCT2), that is linked to an hNbs1 Ser278 phosphorylation site loop (human Nbs1 residues 272–295) (Figure S3). To model these missing regions of hNbs1-fc, we generated 100 hNbs1-fc models based on the spNbs1 structure using MODELLER, and evaluated models by fit to the SAXS data. This approach identified 12 models with superior χ2 fits (χ2<1.3, best model χ2=1.25) that all bear extended accessible pSer278 loop conformations (Figures 2B and 2C). Thus, the overall spNbs1-fc fold is maintained in hNbs1-fc, but is decorated with a solvent-accessible phosphorylated loop. Indeed, based on the BRCT1-BRCT2 interface position and Ser278 accessibility on this loop, we hypothesize phosphorylated Ser278 in hNbs1 is suitably placed to regulate BRCT phosphoprotein interactions in cis through an intermolecular auto-inhibitory mechanism (Figure 2C).
Nbs1 SAXS and X-ray structures provide molecular insights into the consequences of patient-derived NBS mutations. The 657del5 frameshift NBS predisposition mutation, accounting for ~90% of patient isolated mutations (Digweed and Sperling, 2004), generates a protein truncation plus a cryptic translational start site from which the Nbs1 C-terminus is produced at reduced levels (Maser et al., 2001). This catastrophic molecular disconnect contributes to radio-sensitivity phenotypes, and ultimately to malignancies in NBS patients (Maser et al., 2001). At the protein level, 657del5 causes molecular scission of hNbs1-p95 into two halves: 1) an N-terminal (p26) fragment that fails to interact with other members of the MRN complex and contains the FHA-BRCT1 with an extended 15 amino acid C-terminus, and 2) a C-terminal (p70) fragment with intact Mre11-ATM binding motifs and an extended N-terminal peptide linked to its BRCT2 domain (Figure 2D). Structurally, 657del5 physically uncouples the Nbs1 FHA-and BRCT-mediated phosphoregulatory interactions of Nbs1 from its Mre11 and ATM/Tel1 interaction regions. Besides split FHA and BRCT domains, we hypothesize protein instability and decreased abundance of Nbs1-p70 underlying cancer incidence (Kruger et al., 2007) is caused by solvent exposure of a large hydrophobic patch on BRCT2 that normally forms the BRCT1-BRCT2 interface (Figure 2D).
Two missense mutations (R215W and L150F) mapping to Nbs1-fc are linked to NBS (Seemanova et al., 2006) and breast cancer (Heikkinen et al., 2006). The R215W (643C>T) mutation occurs with 657del5 in compound heterozygous individuals with NBS presenting unusually severe neurological abnormities (Seemanova et al., 2006). Arg215 lies in the BRCT1-BRCT2 linker region, and salt-bridges with Asp205 and Glu206 in our hNbs1-fc homology models (Figure 2E). Mutation of hNbs1-fc Arg215 to Trp confers trypsin sensitivity, suggesting R215W decreases Nbs1 protein stability (Figure 2G). Conversely, the L150F breast cancer associated mutation displays no measurable stability defect. However, L150F flanks the Nbs1 BRCT phosphoserine interaction cleft (Figure 2F). By analogy to BRCA1 mutations occluding phosphoprotein-binding pockets, L150F may deform the pSer-binding pocket via a relay of hydrophobic distortion (Williams et al., 2004). Whereas patients with Nbs1 mutations decreasing protein stability and abundance present with classical NBS symptoms, a specific linkage of Nbs1 L150F to breast cancers may reflect critical functional roles for the Nbs1 BRCT interactions in tumor suppression. Thus, Nbs1 disease-causing mutations may either directly impact phosphoprotein-binding as seen for BRCA1 (Williams et al., 2004), or global protein stability as seen for BRCA1 (Williams et al., 2001), XPD (Fan et al., 2008), and superoxide dismutase (DiDonato et al., 2003).
hNbs1 binding to hMDC1 depends on both the FHA and BRCT domains (Chapman and Jackson, 2008), probably via bipartite Nbs1 FHA-BRCT interactions (Figure 1E - type IV binding). Consensus phosphoprotein interaction motifs (Williams et al., 2004) are missing in spNbs1 BRCT domains, but present in hNbs1 and higher eukaryotes (Figure S3). As S. pombe also lacks an MDC1 ortholog, the Nbs1 FHA domain may mediate key phosphorylation-dependent interactions with other DSB repair factors through an FHA-only (Figure 1G - class I) interaction.
We addressed the functional importance of the FHA phosphoprotein-binding pocket in spNbs1. Arg27 and Lys45 are conserved residues predicted to contact phosphothreonine (pThr) in protein binding targets (Durocher et al., 1999), so we constructed strains replacing one or both residues with alanine. While neither single mutant is sensitive to DNA damage (except at high doses for nbs1-R27A), the double mutant (nbs1-RKAA) is sensitive to ionizing radiation (IR), the topoisomerase I poison camptothecin (CPT), and the DNA replication inhibitor hydroxyrurea (HU) (Figure 3A). These mutations had no effect on Nbs1 abundance (Figure 3B).
Interestingly, nbs1-RKAA phenotypes matched those of nbs1-s10 (Figure 3A), a Gly103Asp mutant (Akamatsu et al., 2008) that destabilizes the phosphoprotein-binding loop encompassing Lys45 (Figure S6). Besides causing genotoxin sensitivity, nbs1-s10 has a synthetic lethal interaction with rad2Δ (Akamatsu et al., 2008). Rad2 (FEN-1/RAD27) is a structure-specific endonuclease that processes Okazaki fragments during DNA replication. Unprocessed Okazaki fragments lead to replication fork collapse, causing damage that is repaired by an HR pathway requiring MRN (Akamatsu et al., 2008). Further revealing the importance of the Nbs1 FHA domain in mediating repair of collapsed replication forks, nbs1-RKAA rad2Δ double mutants are inviable (Figure 3C).
The nbs1-RKAA and nbs1-s10 phenotypes suggest that Nbs1 FHA domain facilitates a phosphorylation-dependent interaction with a key HR protein. As a high-copy suppressor of nbs1-s10 (Akamatsu et al., 2008), Ctp1 may interact with the Nbs1-FHA, consistent with our finding that ctp1+ overexpression rescues nbs1-RKAA (Figure 3D).
To test if Nbs1 is required for Ctp1 localization to DSBs, we used chromatin immunoprecipitation (ChIP) to measure enrichment of Mre11, Nbs1 and Ctp1 at four sites 0.2–16 kb away from a defined DSB created with HO endonuclease (Limbo et al., 2007). We assayed γH2A as a marker enriched at regions flanking a DSB, and we performed these assays in a mre11-H134S mutant background (Williams et al., 2008), as ablated Mre11 endonuclease activity increases ChIP signals for Mre11, Nbs1 and Ctp1 at DSBs, presumably because defective endonuclease processing of DNA ends increases MRN retention at a DSB. Mre11-H134S, Nbs1, Ctp1 and γH2A were strongly enriched around the DSB in nbs1+ cells (Figure 3E). The γH2A signal was weakest nearest the DSB (0.2 kb), while Ctp1 and Nbs1 signals were strongest at this position. No Ctp1 enrichment at the DSB was detected in nbs1Δ cells (Figure 3E), even though Ctp1 abundance is unaffected by loss of Nbs1 (Akamatsu et al., 2008). Mre11-H134S enrichment was unaffected in nbs1Δ cells (Figure 3E). Truncated Nbs1 in nbs1-FHAΔ cells had a weak or transient interaction with the DSB, but reduced abundance of Nbs1-FHAΔ complicates interpretation of this result. Importantly, Ctp1 enrichment at the DSB was strongly impaired in nbs1-ΔFHA cells. The same defect was seen in nbs1-RKAA and nbs1-s10 mutants, yet the mutant Nbs1 proteins exhibit no defect in associating with the DSB (Figure 3E). Quantitative real-time PCR confirmed diminished recruitment of Ctp1 to the DSB in the nbs1 mutants (Figure 3F).
Although we were previously unable to co-precipitate Ctp1 and Nbs1 (Limbo et al., 2007), their mammalian counterparts associate in vivo (Chen et al., 2008; Sartori et al., 2007). Indeed, by adjusting cell extraction conditions and utilizing an mre11-H134S background, we could detect Nbs1 in a Ctp1 immunoprecipitation when both proteins were expressed at endogenous levels (Figure 3G). Importantly, nbs1-RKAA disrupted this interaction (Figure 3G). Collectively, these results reveal a critical role for the Nbs1 FHA domain in mediating Ctp1 localization to DSBs.
To test if Ctp1 phosphorylation is required for Nbs1 binding, we focused on potential casein kinase 2 (CK2) phosphorylation sites in Ctp1, as CK2 phosphorylates SDT sites in hMDC1 that interact with the Nbs1 FHA in mammalian cells (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008). NetPhos 2.0 (http://www.cbs.dtu.dk/services/NetPhos/) identified two clusters of probable CK2 sites in the Ctp1 N-terminus: the first spanning Ser77, Thr78 and Thr79, and the second containing Ser87 and Thr89 (Figure 4A). Based on their similarity to MDC1 SDT repeats, we designated these clusters as SXT repeats. While mutation of either SXT repeat had no noted effect on DNA damage sensitivity, a mutant removing all Ser/Thr in the repeats (ctp1-5A) was extremely damage sensitive suggesting the SXT repeats are functionally redundant, not cooperative (Figure 4B). The triple threonine mutant (ctp1-T78A T79A T89A) was as DNA damage sensitive as ctp1-5A, while strains with paired Ctp1 Thr-Ala substitutions (ctp1-T78A T89A or ctp1-T79A T89A) are fully functional (Figure S7), so a single Thr within the SXT repeats is sufficient for function. The ctp1-5A mutant grew slowly, as is typical for Ctp1-defective strains (Limbo et al., 2007), but did not reduce Ctp1 abundance or nuclear localization (Figures 4C and 4D), indicating that Ctp1-5A is not grossly destabilized or misfolded. ChIP analysis revealed a striking reduction in Ctp1 enrichment at a DSB in the ctp1-5A mutant, as seen in the nbs1-RKAA background (Figure 4E).
To test if Ctp1 is phosphorylated on its SXT repeats, we performed SDS-PAGE analyses of FLAG-tagged Ctp1 immunoprecipitated from ctp1+ or ctp1-5A cells. In the absence of DNA damage, Ctp1-5A migrates faster than wild-type Ctp1 (Figure 4F, lanes 1–2), indicating that basal Ctp1 phosphorylation is altered in this mutant. Consistent with this observation, phosphatase treatment causes both wild-type and mutant Ctp1 to migrate faster than either untreated sample, indicating that Ctp1 is basally phosphorylated on SXT repeats and other sites (Figure 4F, lanes 5–6). Exposure to the radiomimetic drug bleomycin induces further Ctp1 phosphorylation that results in a shift to slower mobility, but this shift is ablated in Ctp1-5A (Figure 4F, lane 3–4). This DNA damage-induced hyper-phosphorylation of Ctp1 requires Mre11, Nbs1 and either of the ATM/ATR-like kinases (Tel1/Rad3) in fission yeast (Akamatsu et al., 2008).
To test if the SXT repeats in Ctp1 mediate direct interactions with the FHA domain of Nbs1, we monitored in vitro association of Nbs1-fc with phosphorylated Ctp1 peptides by size exclusion chromatography (Figure 5). Nbs1-fc directly binds and co-elutes with phosphorylated fluorescein (FITC) conjugated Ctp1 peptides encompassing either of the functionally redundant SXT repeats. These interactions were phosphorylation-dependent and reduced by the nbs1-RKAA mutation.
To define Nbs1-Ctp1 interactions, we co-crystallized and determined the 2.15 Å resolution X-ray structure of the first Ctp1 SXT repeat bound to Nbs1-fc (Table S2, Figures 6 and S8). FITC-Ctp1 phosphopeptides enabled us to visually identify yellow Nbs1-fc-Ctp1 complex crystals in a background of peptide-free Nbs1-fc crystals (Figure 6B, Experimental Procedures). Phosphorylated Ctp1 binds in an electropositive groove formed by FHA residues Arg27, Lys45, Phe77, Lys76, Lys104, Ser44, and Lys41 (Figure 6C), and buries 644 Å2 of solvent accessible surface at the Ctp1-Nbs1 interface. Ctp1 Ile72 - pSer77 and residues C-terminal to Glu81 (Asp82 and Glu83) are disordered. In contrast, 2Fo-Fc and Fo-Fc difference maps show electron density for Ctp1 pThr79, positions pThr(−1) (Thr78) through pThr(+2) (Glu81), and pSer77 main-chain, but not the pSer phosphate (Figure S8), underscoring the importance of these residues for Nbs1 binding.
The Nbs1 peptide-binding groove interacts with Ctp1 on three sides (Figures 6C and 6D) and significant rearrangements of the interacting loops occur as they “grasp” Ctp1 (Supplemental Movies S1 and S2). The pThr79 phosphate directly contacts the side chains of Arg27, Ser44, and Lys45. Residues Arg27, Lys45, Ser42, Ile43, Lys76, Ser44, and a tightly bound water molecule mold the pThr interaction site surface (Figure 6D). This pocket appears to specifically bind pThr preferentially over pSer or pTyr: it is too small for pTyr binding and tightly contours around the pThr79 side chain methyl. Redundant contacts by Arg27 and Lys45 with the pThr phosphate explain the mild phenotypes observed for nbs1-R27A and nbs1-K45A single mutants as compared to the nbs1-RKAA double mutant (Figure 3A).
Our structure of the Nbs1-Ctp1 complex, conservation of the Nbs1 FHA phosphopeptide-binding cleft, and homology of the Ctp1 SXT repeats to Nbs1-binding motifs in human MDC1 (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008) and S. cerevisiae Lif1 (Palmbos et al., 2008) together define an Nbs1 FHA-binding consensus sequence motif as a pThr with an Asp at position pThr(+1) (Figures 6D and 6E). Conserved residues Lys76 and Phe77, along with Lys41 and Lys104 recognize Ctp1 Asp80 (pThr (+1)). Conversely, recognition of Ctp1 Glu81 (pThr(+2)) by Nbs1 Lys104 appears specific to the S. pombe Nbs1-Ctp1 interaction. While Phe77 makes van der Waals contacts to the peptide backbone, Lys76 hydrogen bonds to the pThr79 main chain carbonyl and combines with Lys41 in salt-bridges that sandwich the acidic pThr(+1) specificity position (Asp80). Consistent with structural observations, an Nbs1 mutant (Ala76-Ala77) impairs Ctp1 peptide binding, showing the conserved Lys76-Phe77 loop is crucial for the interaction (Figure 5).
Whereas most FHA domains recognize pThr(+3) position, specific interactions with Nbs1-FHA in cells are probably achieved by unique recognition of pThr-Asp motifs. Added specificity for the CtIP-MRN complex interaction in complex eukaryotes may also occur via avidity, as CtIP also binds Mre11 and Rad50 (Sartori et al 2008). Conservation of acidic residues C-terminal to pThr(+2) and phosphorylation at pSer(−2) could reflect kinase recognition-specificity for efficient in vivo phosphorylation, rather than a preferred Nbs1-FHA binding consensus (Figure 6E), as a single or double phosphorylated Ctp1 peptide binds Nbs1-fc with similar affinity (Figure 5B).
Use of a minimalist pThr-Asp Nbs1-FHA interaction consensus binding motif may direct MRN binding to multiple cellular targets. We predict the Xrs2Nbs1-Lif1 interaction facilitating MRN(X) DSB repair by non-homologous end joining in budding yeast (Palmbos et al., 2008) assembles via interactions between the Xrs2Nbs1-FHA domain and Lif1 CK2 phosphoacceptor sites analogous to the Ctp1 SXT repeats (Figures 6E and 6F). By contrast, mammalian Nbs1 binding to clusters of MDC1 SDTD repeats depends on both FHA and BRCT phosphoprotein interaction sites, and phosphorylation of the serine and threonine in the SDTD motif (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008). As BRCT domains typically bind pSer targets with specificity for the pSer(+3) position (Glover et al., 2004; Manke et al., 2003), our data predict that Nbs1-MDC1 complexes assemble through a bipartite interaction between the Nbs1 FHA domain binding the pTD of an MDC1 SDTD motif and the Nbs1 BRCT domains recognizing a neighboring MDC1 pSXXD consensus of a second SDTD repeat (Figures 6E and 6F).
Nbs1 might provide rigid architectures to align its multiple interaction partners. Yet, Nbs1 proteolytic lability suggests its C-terminus is disordered (Figure 1B). To test these possibilities, we examined spNbs1-fc and spNbs1-ΔAT structures in solution with SAXS. While spNbs1-fc SAXS closely matches the X-ray structure, the maximum particle dimension (Dmax) for spNbs1-ΔAT is much larger (~175 Å), indicating a very elongated structure (Figures 7A and 7B). SAXS objectively measures folding: Kratky plots are parabolic for folded structures, asymptotic for unfolded structures, and intermediate for partially folded structures (Putnam et al., 2007). Consistent with an unstructured Nbs1 C-terminus, the spNbs1-ΔAT Kratky plot is intermediate between parabolic and asymptotic curves, while spNbs1-fc has a parabolic plot showing spNbs1-fc is folded consistent with proteolysis and crystal structures (Figure 7C).
Although its C-terminus may undergo disorder-to-order transitions upon binding Mre11 and ATM, we do not see stable domains (>10 kDa) when Nbs1 is trypsin digested while in complex with Mre11 (Figure S9). Thus, C-terminal Mre11 and ATM interacting regions may be peptide interaction motifs, rather than globular domains. Sequence-based order predictions of Nbs1 homologs (Figures 7D and S10) suggest flexibility is a conserved Nbs1 feature, so inherent pliability of the Nbs1 C-terminus may be needed for function. Furthermore, flexibility extends into Nbs1-fc: the closed, phosphopeptide grasped FHA conformation flips an Arg switch at the FHA-BRCT interface and promotes conformational changes in the BRCT linker region, a ~20° rotation about the BRCT1-BRCT2 interface, and a ~10 Å shift at the BRCT2 C-terminal end in our experimentally determined X-ray structures (Figure 7E).
Nbs1 provides the key phosphoprotein sensor and recruitment scaffold for the Mre11-Rad50 complex. Nbs1 forms a folded N-terminal FHA-BRCT1-BRCT2 core, conserved from S. pombe to humans based upon the SAXS results, that uniquely connects an FHA phosphoprotein-binding site with BRCT domains via an extensive interface. The direct fusion of two phosphoprotein-binding domains implies functional crosstalk between binding sites may occur. Interestingly, structural overlays of the apo and peptide-bound crystal structures reveal an Nbs1-fc conformational change that may facilitate FHA-BRCT phosphoprotein binding site communication and pull on the flexible tether to Mre11 (Figure 7E, Supplemental Movies S3 and S4).
Integrating our results with known data supports a testable architectural hypothesis for MRN-Ctp1 (Figure 7F). MRN complex components and Ctp1 co-localize to within ~200 bp (<700 Å), of a defined endonuclease cut DSB in vivo (Figure 3E). Ctp1 is retained at the break site via direct phosphorylation-dependent binding to the N-terminal Nbs1 FHA domain. SAXS and X-ray structures define the reach from the Nbs1 C-terminal Mre11 interaction region to the N-terminal FHA domain as ~175 Å. Prior results also help delineate MRN-Ctp1 complex architecture at a DSB: 1) the Mre11-Rad50 DNA binding head can directly bind and bridge DNA ends (Chen et al., 2001; Williams et al., 2008), 2) the Mre11/Rad50 heterotetrameric core forms a four-lobed DNA binding head composed of dimeric Mre11 and the Rad50 ATPase domains, with Rad50 coiled-coils emanating from the head (Hopfner et al., 2001, Hopfner et al. 2002), and 3) the Mre11-Nbs1 binding interface maps to Nbs1 C-terminal motifs (Falck et al., 2005; You et al., 2005) and an Mre11 dimer surface patch away from the DNA binding cleft (Williams et. al, 2008). Rad50 coiled-coils are positioned to define the reach of the MRN complex to bridge homologous sister chromatids during HR repair (Hopfner et al., 2002; Moreno-Herrero et. al., 2005). Direct synapsis of DNA ends by Mre11 (Williams et al., 2008), and the tethering length of the Nbs1 connection between Ctp1 and Mre11 defined here support a mechanism for localizing both the MRN complex and Nbs1-linked Ctp1 to regions directly flanking a DSB (Figure 7F). Consistent with this composite structural model, archaeal Mre11/Rad50 core complex (Hopkins and Paull, 2008) or eukaryotic MRN/Ctp1Sae2 5′-3′ resection activity is limited to regions proximal to a DSB (Mimitou and Symington, 2008; Zhu et al., 2008).
Flexible Nbs1 molecular associations may accommodate large-scale MRN complex conformational changes, proposed to regulate ATM activation and dimer-to-monomer transitions (Lee and Paull, 2005). Like the flexible linkage of PCNA to ligase for DNA replication (Pascal et al., 2006), the malleable nature of Nbs1 and the Rad50 coils (Moreno-Herrero et al., 2005), provides a molecular basis for concerted MRN DNA break sensing and repair activities while preserving the integrity of a structurally dynamic chromatin template. With its apex FHA binding motif and conformational switch encompassing the FHA-BRCT interface, the mobile Nbs1 hub appears suitable to promote integration of repair sensing and effector activities of MRN by interface exchange and handoff interactions with multiple partners, as proposed for PCNA (Chapados et al., 2004) and BRCA2-Rad51 partners (Shin et al., 2003).
Recombinant S. pombe Nbs1 variants were expressed from pET29b as C-terminal 6-His tagged proteins in E. coli Rosetta2 (DE3) cells (Novagen) grown at 16 °C in Terrific Broth. After Ni-NTA affinity chromatography, proteins were purified by Superdex 75 (GE Healthcare) (spNbs1-fc, residues 1-329) or Superdex 200 (Nbs1-ΔAT, residues 1-553) gel filtration, and anion exchange chromatography. G103D, R27A/K45A and K76A/F77A mutations were made with Quickchange (Stratagene).
spNbs1-fc crystals were grown at 8 °C by hanging drop vapor diffusion. Nbs1-fc crystals were grown by mixing 1 μL protein solution (10 mg/mL spNbs1-fc in 500 mM NaCl, 20 mM Tris-HCl pH 7.5, 0.1% β–mercaptoethanol) and 1uL of crystallization solution A (2 M NaCl, 100 mM sodium acetate pH 5.5). Plate morphology was improved with crush dilution seeding into drops with 1 μL crystallization solution B (1.2–1.4 M potassium formate, 100 mM sodium acetate pH 4.6, 5% w/v polyethylene glycol 8000) and 1 μL protein solution. For data collection, crystals were transferred to cryo-protectant (crystallization solution B plus 26% ethylene glycol) and flash cooled in liquid nitrogen.
Peptides were synthesized at the Scripps Research Institute core facility. Colorless peptide-free and yellow peptide-bound (FITC-conjugated) crystals grew in the same drop. 10 mg/mL spNbs1-fc in 40 mM NaCl, 20 mM Tris 7.5, 0.1% β-mercaptoethanol was mixed with FITC-Ctp1-1 (FITC-KKIQELD(pS)T(pT)DEDEI-nh2) at a 4:1 Ctp1-1/spNbs1-fc ratio. Crystals were grown at 8 °C by hanging drop vapor diffusion by mixing 1 μL of Nbs1-Ctp1 complex with 1 μL crystallization solution C (18% Polyethylene glycol 3350, 100 mM MES pH 6.1–6.5, 260 mM potassium thiocyanate). Crystals were washed in cryoprotectant (crystallization solution C with 26% glycerol) and yellow peptide-containing crystals were flash cooled in liquid nitrogen for data collection.
Data processing, phasing, model building and refinement are detailed in Supplemental data. Final refined models of apo Nbs1 at 2.8 Å resolution (R= 23.6, Rfree=28.5) and Nbs1/Ctp1 complex at 2.2 Å resolution (R=21.6, Rfree=26.4) exhibit excellent geometry (Table S2).
Coordinates and structure factors for Nbs1-fc (3HUE) and Nbs1-fc/Ctp1 (3HUF) complex are in the Protein Data Bank.
FITC-conjugated peptide binding to Nbs1-fc variants was monitored by co-elution in gel filtration or fluorescence polarization for sample concentrations detailed in the Supplemental Material.
SAXS data were collected at SIBYLS beamline 12.3.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory. SAXS scattering data was collected for 2.5, 5.0, and 10.0 mg/mL samples for Nbs1-fc, 2 mg/mL samples of spNbs1-ΔAT, and 10 mg/mL hNbs1-fc. SAXS analysis was done as described in the Supplemental Data.
Strains used are listed in Table S1.
Exponentially growing cells were exposed to bleomycin as described (Akamatsu et al, 2008). Whole-cell extracts were prepared by disrupting in lysis buffer (50 mM Tris pH 8.0, 150 mM NaCl, 2.5 mM EDTA, 0.002% NP-40, 50 mM NaF, protease inhibitor tablet (Roche, complete Mini)) with a bead beater.
Soluble whole-cell extracts were made from exponentially growing cells by disrupting in low-salt lysis buffer B1: (50 mM Tris pH 8.0, 50 mM NaCl, 2.5 mM EDTA, 0.002% NP-40, 50 mM NaF, protease inhibitor tablet (Roche, complete Mini)) with a bead beater. Ctp1-TAP and associated proteins were precipitated from extracts with a 50% suspension of IgG Sepharose (GE Healthcare), washed 3 times with B1 buffer, resolved on 8% SDS-PAGE, and examined by Western blotting.
Ctp1 cDNA was cloned as N-terminal GFP fusion proteins under control of the thiamine-repressible nmt1 promoter (pREP41). Cells were then grown in the absence of thiamine for 20 hours to induce GFP-Ctp1 expression. Images were collected as described (Limbo et al., 2007).
Ctp1-FLAG immunoprecipitation from whole-cell extracts and subsequent λPPase treatment of precipitates were done as described (Akamatsu et al., 2008).
Fellowship support was provided to R.S.W. (Canadian Institutes of Health Research, the Alberta Heritage Foundation for Medical Research, and the Skaggs Institute for Chemical Biology), G.E.D. (National Institutes of Health NRSA), and Y.Y. (The Uehara Memorial Foundation). Work on Mre11 complex and Ctp1 in the authors’ laboratories is supported by National Cancer Institute grants CA117638, CA92584 and CA77325. SIBYLS beamline (BL12.3.1) efforts at the Advanced Light Source of Lawrence Berkeley National Laboratory are supported by U.S. Department of Energy program IDAT (for Integrating crystallography and X-ray scattering) under Contract Number DE-AC02-05CH11231. We thank SIBYLS staff G. Hura and M. Hammel for expert SAXS data collection assistance, B. Chapados, C. Chahwan for discussions, and E. Getzoff and G. Williams for comments.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.