|Home | About | Journals | Submit | Contact Us | Français|
The Mre11/Rad50/Nbs1 (MRN) protein complex plays central enzymatic and signaling roles in the DNA-damage response. Nuclease (Mre11) and scaffolding (Rad50) components of MRN have been extensively characterized but the molecular basis of Nbs1 function has remained elusive. Here, we present a 2.3Å crystal structure of the N-terminal region of fission yeast Nbs1, revealing an unusual but conserved architecture in which the FHA and BRCT-repeat domains structurally coalesce. We demonstrate that di-phosphorylated pSer-Asp-pThr-Asp-like motifs, recently identified as multi-copy docking sites within human Mdc1, are evolutionarily conserved Nbs1 binding targets. Furthermore, we identify similar phospho-motifs within Ctp1, the fission yeast orthologue of the human tumor-suppressor, CtIP, and show that their interactions with the Nbs1 FHA domain are necessary for Ctp1-dependent resistance to DNA-damage. Finally, we establish that human Nbs1 interactions with Mdc1 can occur through both its FHA and BRCT-repeat domains, suggesting how their structural and functional inter-dependence may underpin Nbs1 adaptor functions in the DNA-damage response.
Genomic integrity is constantly challenged by the generation of DNA lesions, of which double-stranded breaks (DSBs) are generally considered the most toxic. DSBs can arise through the actions of DNA-damaging chemicals and ionizing radiation, but also occur at sites of stalled DNA replication and as intermediates during programmed genomic rearrangements such as VDJ recombination, immunoglobulin class-switching and meiotic recombination. DNA-damage detection initiates a complex and orchestrated set of intracellular responses involving kinase-signaling cascades, cell-cycle checkpoint activation and the deployment of DNA-repair proteins (Riches et al., 2008). Central to these processes is a diverse group of adaptor and scaffold proteins that function as platforms for the assembly of multi-protein complexes through the combinatorial activities of phosphoserine/threonine binding protein modules, such as forkhead-associated (FHA), Brca1 C-terminal (BRCT)-repeat, Polo-box, WW and WD40-repeat domains (Reviewed in Seet et al., 2006) which interact with specific motifs following their phosphorylation by DNA-damage activated or cell-cycle responsive serine/threonine kinases.
Human Nbs1 (also known as Nibrin or p95) is a 754 amino-acid residue subunit of the Mre11-Rad50-Nbs1 (MRN) complex, which binds to DSBs and acts as a bridge to hold the DNA ends in close proximity and promote their rejoining (Williams et al., 2007). Nbs1 was originally identified as the gene mutated in Nijmegen breakage syndrome (NBS), a rare autosomal-recessive human disease characterized by immune disorders, microcephaly, growth retardation, hypersensitivity to ionizing radiation and predisposition to lymphoid cancers (Carney et al., 1998; Matsuura et al., 1998; Varon et al., 1998). Although homozygous-null Nbs1 mutations are embryonic lethal in mouse models (Zhu et al., 2001), these human diseases are characterized by a variety of hypomorphic mutations that can occur throughout the NBS1 gene.
Many NBS phenotypes overlap with those of another cancer-predisposition syndrome, ataxia-telangiectasia, which is caused by defects in the protein kinase ATM (ataxia-telangiectasia-mutated). It is now evident that Nbs1 functions in ATM activation, while also serving as a substrate for ATM-mediated phosphorylation on Ser-278 and Ser-343 (Difilippantonio and Nussenzweig, 2007). Nbs1 makes direct interactions with ATM and Mre11 through distinct motifs within its C-terminal region (Desai-Mehta et al., 2001; Falck et al., 2005; You et al., 2005) whereas the Nbs1 N-terminal region contains FHA and BRCT motifs that are characteristic of many other DNA-damage response proteins and which often mediate phospho-dependent protein-protein interactions. Functionally, Nbs1 seems to play key roles in most or all of the DNA-damage-checkpoint signaling functions of the MRN complex (D’Amours and Jackson, 2002) through interactions with a number of proteins, the best characterized being Mdc1 (mediator of the DNA-damage checkpoint 1) and ATM itself (Chapman and Jackson, 2008; Falck et al., 2005; Lukas et al., 2004; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008; You et al., 2005; Zhao et al., 2000). However, Nbs1 activities do not appear to be restricted to checkpoint signaling. Indeed, more direct contributions of Nbs1 to DSB repair are suggested by the observation that DNA-damage sequestration of CtIP, a molecule required for efficient DSB resection (Limbo et al., 2007; Sartori et al., 2007), is facilitated through direct interactions with Nbs1 in human cells (Chen et al., 2008).
In spite of the array of Nbs1 activities, little is currently known about how it mediates its diverse protein interactions at the molecular level. Here, we describe an integrated structural, biochemical, genetic and cell-biological analysis of the Nbs1 N-terminal FHA/BRCT-repeat region. This work reveals how the conserved architecture and phospho-dependent protein binding functions of the Nbs1 N-terminus play crucial roles in mediating Nbs1 adaptor functions in the DNA-damage response to promote DSB signaling and repair.
Following extensive screening of Nbs1 proteins from several species, we were able to express and purify 382 and 323 residue N-terminal fragments of human Nbs1 (hNbs1) and yeast Schizosaccharomyces pombe Nbs1 (spNbs1), respectively. The structure of the S. pombe protein was determined by single-wavelength anomalous diffraction methods at an initial resolution of 2.6Å, and then further refined against data extending to 2.3Å spacing collected from crystals of reductively methylated protein (Table S1). Representative electron density for the SAD-phased selenomethionine structure and the final refined methylated protein are shown in Figure S1.
Overall, the spNbs1 N-terminus is extended, with approximate dimensions of 95 × 30 × 25 Å (Figure 1A). The N-terminal 114 residues adopt a β-sandwich fold characteristic of FHA domains: small phospho-dependent binding modules that function through recognition of phosphothreonine (pThr)-containing motifs in target proteins (Mahajan et al., 2008). Many functional studies of Nbs1 have been predicated on the assumption that only a single BRCT motif exists C-terminal to the Nbs1 FHA domain. Notably, the spNbs1 structure reveals the presence of tandem BRCT repeats, despite the lack of any significant homology between the second motif and other known BRCT domains. In the light of previous sequence analyses of various Nbs1 orthologues (Becker et al., 2006) and comparison of our spNbs1 structure with the NMR analysis of the second BRCT repeat from Xenopus laevis Nbs1 (Xu et al., 2008) (Figure 1B), it seems that the FHA/BRCT2 composition is universally conserved in eukaryotic Nbs1 proteins. Most significantly, our structure reveals an unprecedented structural fusion of the FHA domain and first BRCT repeat. Rather than being flexibly linked in a ‘beads-on-a-string’ arrangement as might be expected in a classical modular domain system, they are associated through a substantial interface within which 2150 Å2 of solvent-accessible surface is buried (Figure 1C). The bulk of the interfacial interactions occur between the base of the FHA domain and α1, α3, αL1 and α1/β2 loop of BRCT1. Additional stabilizing contacts are provided by the helical insertion in the FHA domain (α0) and the α1-β2/β2-β3 loops from BRCT1. Overall, the relative orientation of the FHA, BRCT1 and BRCT2 domains for three independent molecules (two methylated, one unmethylated) in the two crystal forms are almost identical, suggesting that the extensive FHA-BRCT1 interactions impose a rather rigid sub-structure (Figure 1D).
Several lines of evidence indicate that the overall architecture observed for spNbs1 is common to all eukaryotic Nbs1 orthologues. Thus, although limited sequence identity for interfacial residues is evident between spNbs1 and hNbs1, the non-polar character of amino-acid residues that contribute core FHA-BRCT1 interactions is conserved. (Figure 2A). In addition, a strip of conserved hydrophobic surface follows the path of the linker extending from the BRCT1-FHA interface to BRCT2. Notably, the N-terminal Met-Trp pair of spNbs1 is buried in the structure and forms an intrinsic part of the non-polar FHA-BRCT1 interface (Figure 2B). Trp-2 is highly conserved in animals and fungi and because of its bulk, it presumably insulates the initiator methionine from co-translational excision (Giglione et al., 2004). Plants, however appear to have followed a different evolutionary path to the same structural end-point whereby an extra valine residue is positioned to take the place of Met-1 and maintain the integrity of the FHA-BRCT1 interface following more efficient methionine-amino-peptidase cleavage expected when small hydrophobic residues are present in the second position (Martinez et al., 2008). Overall, the importance of the structure around Met-1/Trp-2 is exemplified by the insolubility/aggregation of recombinant Nbs1 expressed with either N-terminal GST or hexa-histidine affinity tags (data not shown). These observations, together with the absence of any significant ‘linker’ region between the C-terminus of the FHA domain and N-terminus of BRCT1 in all Nbs1 sequences (Becker et al., 2006) (Figure 2C), support our contention that the structural coupling and precise spatial apposition of the FHA and BRCT domains is likely to be crucial for Nbs1 functions in all eukaryotes.
The apparent structural conservation between Nbs1 orthologues allows us to place in a molecular context various human disease-associated Nbs1 mutations that are located within the FHA/BRCT region. To date, around 90% of patients suffering from NBS are homozygous or heterozygous for a founder mutation, 657del5, a 5 bp deletion in exon 6 within the coding region for the BRCT1-BRCT2 linker that results in production of C- and N-terminally truncated proteins termed p26 and p70, respectively (Maser et al., 2001) (Figure S2A). In addition, a number of germ-line and somatic missense mutations in the Nbs1 N-terminal region have been noted in patients suffering from NBS and in a variety of cancers (http://www.nijmegenbreakagesyndrome.net) (Huang et al., 2008) (Figure S2B). Although the degree to which these mutations might contribute to disease is mostly unclear, we note that four (T148I, L150F, I171V and V210F) lie within a significant hydrophobic cluster in the BRCT1-domain core and the BRCT1-linker interface, suggesting that these lesions may result in structural perturbations, potentially contributing to Nbs1 protein instability in affected individuals (Figure S2C). Nonetheless, it remains possible that some mutations (such as V28I located immediately C-terminal to the highly conserved FHA phosphopeptide binding residue Arg-27) directly affect phospho-dependent and/or phospho-independent interaction surfaces on Nbs1.
The functional importance of the Nbs1 FHA/BRCT region is well established but the molecular basis for this remains obscure. Recruitment of the MRN complex to sub-nuclear foci at DSB sites has been proposed to involve a direct interaction between human Nbs1 (hNbs1) and γH2AX (Kobayashi et al., 2002), the ATM-phosphorylated form of variant histone H2A that serves as a major marker for damaged chromatin. However, by isothermal titration calorimetry (ITC), we were unable to detect any interactions of recombinant hNbs1 or spNbs1 proteins with peptides representing human or S. pombe γH2AX (Figure S3). This confirms the lack of binding of hNbs1 to γH2AX peptides observed previously by peptide pull-down assays (Chapman and Jackson, 2008) and strongly suggests that direct γH2AX-binding by Nbs1 is unlikely to contribute significantly to MRN localization to damaged chromatin.
More recently, the mammalian MRN has been shown to be indirectly sequestered to DNA-damage sites through Nbs1 interactions with a cluster of six di-phosphorylated casein-kinase 2 (CK2) target sites within Mdc1, each with a core consensus sequence phospho-Ser, Asp, phospho-Thr, Asp (pSDpTD; (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008). This Mdc1/MRN/ATM complex is, in turn, tethered to damaged chromatin via interactions between the Mdc1 BRCT-repeat domain and γH2AX (Lee et al., 2005; Stucki et al., 2005). Consistent with the dependence of Mdc1-Nbs1 interactions on the Nbs1 FHA domain (Chapman and Jackson, 2008; Lukas et al., 2004; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008), we found that the hNbs1 fragment bound a synthetic Mdc1 pSDpTD motif peptide (Mdc1 residues 325-336) with an affinity of ~1 μM (Figure 3A, upper panel; Table S2) and a stoichiometry significantly greater than 1:1 (see below). In addition, and in spite of the absence of any clear Mdc1 orthologues in yeast, the purified spNbs1 protein bound the same Mdc1 phosphopeptide with a slightly higher affinity of 0.5 μM (Figure 3A, lower panel). By contrast, neither spNbs1 nor hNbs1 bound to a peptide of the same sequence in which the threonine was not phosphorylated, reflecting the acute pThr-selectivity of FHA domains.
Although the spNbs1 FHA domain is closely related to previously determined FHA-domain structures, superposition with a canonical Chk2 FHA-phosphopeptide complex (Li et al., 2002) predicts severe steric clashes of the spNbs1 FHA domain with the phosphopeptide. This results from phenylalanine/tyrosine substitutions in Nbs1 at a conserved position (Phe-77 in spNbs1) normally occupied by an asparagine in other FHA domains (Asn-166 in Chk2) (Figure S4A; Figure 3B). These observations suggest that bound phosphopeptide ligands follow a different path across the spNbs1 FHA binding surface, interacting with a surface groove formed by K76-F77 and S42-I-S-K45 (Figure 3B). These motifs form the β6-β7/β7-β8 loops that are more distant from one another in Nbs1 than in other FHA structures and are instead connected through a bridging water molecule. This generates a somewhat polar floor for the peptide-binding site and potentially provides for some structural flexibility in the relative juxtaposition of the peptide-binding loops. Lys-76 and Lys-45 each lie close to the pThr-binding site and potentially contact the phosphoryl moiety directly (Figure S4B), while Gly-103 and Ser-42 structure the rear of the pocket.
The above ideas were tested by site-directed mutagenesis. Significantly, of the three conserved basic residues located around the canonical FHA-domain binding site, the largest effect on binding to the Mdc1 pSDpTD peptide was observed for alanine substitution of the highly conserved Arg-27 residue (corresponding to Chk2 Arg-117; Figure S4) which reduced apparent binding affinity ~100-fold (Figure 3C; Table S2). More modest effects were seen for the K76A and K45A mutations, which reduced apparent affinity 7- and 5-fold, respectively. A G103D mutation has been implicated in the slow growth and DNA-damage hypersensitivity of the S. pombe nbs-s10 strain identified in a mutagenesis screen (Akamatsu et al., 2008). Given its structural context, we speculated that this substitution might directly affect phosphopeptide binding and, as shown in Figure 3C, this is clearly the case.
To explore the functional relevance of the somewhat unusual FHA-mediated binding mode implied by the above data, we introduced the same mutations into S. pombe cells and determined their effects on cell growth and sensitivity to various DNA-damaging agents. Consistent with previous findings (Akamatsu et al., 2008), a strain bearing the spNbs1 G103D mutation exhibited slow growth and hyper-sensitivity to a range of agents that cause DNA-damage and/or perturb DNA replication (Figure 3D). Moreover, a strain bearing the R27A mutation that severely reduced phosphopeptide binding in vitro, closely mimicked the slow-growth and DNA-damage hypersensitivity phenotypes of the nbs-s10 strain, implying that the G103D/s-10 phenotypes are not merely caused by improper Nbs1 folding. By contrast, strains with alanine substitutions of Lys-45 or Lys-76, which resulted in more modest phosphopeptide binding defects, were not slow-growing and exhibited DNA-damage sensitivity profiles close to those of the wild-type strain (Figure 3D). Nevertheless, these two strains showed significant hyper-sensitivity when exposed to increased concentrations of methyl-methane sulphonate, presumably reflecting the modest impairment of phospho-binding activity observed in ITC experiments. Collectively, these structural, biochemical and functional analyses thus provided strong evidence that FHA-dependent phosphopeptide interactions by spNbs1 are required for its roles in promoting normal growth and resistance to DNA-damaging agents.
The above findings suggested that binding to pSDpTD-like motifs is an important, evolutionarily conserved function for Nbs1. In this regard, we noted that phospho-dependent interactions of budding yeast Nbs1/Xrs2 with the XRCC4 homologue, Lif1 (Matsuzaki et al., 2008) have previously been shown to require the Xrs2 FHA domain and two putative phosphorylation sites (Thr-387 and Thr-417) within Lif1 (Palmbos et al., 2008), both of which are clearly related to the Mdc1 SDTD motifs (Figure 4A). While S. pombe has no clear Mdc1 or Lif1 orthologues, we were intrigued by previous observations that the growth and DNA-repair defects of the nbs-s10 strain (harboring the functionally disruptive G103D FHA-domain mutation) are rescued by Ctp1 over-expression. We confirmed this and further showed that Ctp1 over-expression can suppress the impaired growth and DNA-damage hyper-sensitivity of a strain carrying the spNbs1 R27A FHA mutation that reduces phospho-binding function by ~100-fold in vitro (Figure 4B).
These data suggested to us that the spNbs1 FHA domain might promote interactions with Ctp1, and that spNbs1 FHA-domain defects can thus be circumvented by Ctp1 over-expression. In line with this idea, like its human counterpart CtIP (Chen et al., 2008; Sartori et al., 2007), S. pombe Ctp1 is a phospho-protein that is recruited to DSBs in a MRN-dependent fashion to promote DSB resection and DNA repair by homologous recombination (Akamatsu et al., 2008; Limbo et al., 2007). Moreover, while inspecting the Ctp1 sequence, we observed that it contains two motifs (Site 1 and Site 2) closely resembling the CK2-consensus, Nbs1-binding sites of Mdc1 and Lif1 (Figure 4A).
To investigate whether Ctp1 is phosphorylated on the Site 1 and 2 motifs in vivo, we generated a phospho-specific antibody directed against a di-phosphorylated peptide comprising Ctp1 Ser-87 and Thr-89 (pS87pT89). We then used this antibody to probe Ctp1-immunoprecipitates from extracts of S. pombe strains expressing epitope-tagged wild-type Ctp1 or Ctp1 mutants (‘2TA’ and ‘3TA’) in which the Thr residues in Site 1 or Site 2 were converted to non-phosphorylatable Ala residues (Figure 4C). Importantly, while the Ctp1 pS87pT89 antibody recognized wild-type Ctp1, its immunoreactivity was lost in the case of the mutant proteins, thus revealing that Ctp1 Site 2 (Ser-87/Thr-89) is phosphorylated in vivo. Consistent with this, we noted that the wild-type Nbs1 derivative migrated more slowly on SDS-polyacrylamide gels than the mutant proteins, while λ phosphatase treatment converted all proteins to a faster migrating form. Collectively, these findings indicated that Ctp1 is subject to multiple phosphorylations in vivo, some of which reside in the Site1/2 motifs. In accord with the apparent constitutive phosphorylation of CK2 motifs in mammalian Mdc1 (Chapman and Jackson, 2008; Melander et al., 2008; Spycher et al., 2008; Wu et al., 2008), we detected Ctp1 Ser-87/Thr-89 phosphorylation irrespective of whether the S. pombe cells had been treated with a DNA-damaging agent (Figure 4C). Significantly, we further noted that the DNA damage-dependent reduction in Ctp1 mobility previously attributed to phosphorylation by S. pombe orthologues of ATR and ATM, Tel1 and Rad3 (Akamatsu et al., 2008) was not observed in the Nbs1-deletion and FHA mutant strains, and was reduced or abolished in the 2TA and 3TA strains (Figure 4D). Together, these data suggested a mechanistic link between Nbs1 FHA-binding activity and priming phosphorylation of Ctp1 by CK2 as a pre-requisite for additional DNA-damage dependent Ctp1 phosphorylation by ATM/ATR-family kinases.
The foregoing data prompted us to investigate whether Ctp1 phosphosites bind spNbs1 in vitro. This revealed tight spNbs1 binding to phosphorylated forms of both the Ctp1 Site1 and Site 2 motifs (Figure 5A; Table S2). Indeed, the Site 2 peptide containing both pSer-87 and pThr-89 bound some 30-fold tighter than a di-phosphorylated Site 1 peptide (pSer-77/pThr-79), with an apparent Kd of ~50 nM that is among the highest affinity interactions for any FHA domain reported to date. Significantly, interactions of spNbs1 with Site 1 or Site 2 peptides phosphorylated only on pThr-79 or pThr-89 were 8-fold and 10-fold weaker than those for the respective di-phosphorylated versions. This effect recapitulates the 17-fold reduced affinity observed for hNbs1 binding to the mono- versus di-phosphorylated Mdc1 peptide (Figure 5B), demonstrating a remarkably similar pattern of specificity for the fission yeast and human Nbs1 orthologs.
To further investigate the potential functional significance of Ctp1 Sites 1 and 2 as binding epitopes for spNbs1, we generated mutant strains bearing alanine substitutions of Thr-79, Thr-89 or both. Perhaps surprisingly, both the individual and double site Ctp1 mutants largely corrected the slow-growth and DNA-damage hypersensitivity phenotypes of Ctp1-defective cells (Figure 5C). Nevertheless, while displaying little or no hypersensitivity towards low doses of the radio-mimetic drug phleomycin (data not shown), the double mutant did display some hypersensitivity to higher phleomycin levels. We thus speculated that phosphorylation of a third threonine, Thr-78 within Ctp1 Site 1, might also contribute to Ctp1 function and, accordingly, we found that the spNbs1 FHA/BRCT2 region bound tightly to a Ctp1 peptide containing singly phosphorylated Thr-78 with an apparent Kd of ~0.2 μM (Table S2). Moreover, while a mutant strain bearing the single T78A mutation displayed wild-type growth and drug-sensitivity profiles, combination of the T78A, T79A and T89A mutations yielded a strain that was hypersensitive to camptothecin, phleomycin, and methyl-methane sulphonate, albeit not to the same degree as the Ctp1-null strain (Figure 5A).
In order to investigate the specificity of spNbs1 interactions with the Site 1 and Site 2 Ctp1 motifs, we constructed a strain containing alanine mutations of Thr-78, 79, 89 and Thr-71 that is located immediately N-terminal to the Site 1 region, but not within a SDTD-like sequence. In this case, the additional mutation did not further increase damage sensitivity (Figure S5A), a lack of effect that correlated well with the absence of significant binding of spNbs1 FHA/BRCT2 to a pThr-71 phosphopeptide (Figure S5B; Table S2). These and additional data (Figure S6; Table S2) revealed an ‘extended’ specificity of the spNbs1 FHA domain beyond that normally observed for such modules. Binding absolutely depends on the presence of at least one pThr, and a further preference for glutamate/aspartate is evident at the pThr +3 position that is often a major determinant of FHA specificity (Durocher et al., 2000). For Thr-79 and Thr-89, additional serine phosphorylation in the -2 position increases affinity by ~10-fold. However, other positions, notably those in the C-terminal acidic stretch of the motif, contribute significantly to binding affinity reflecting a plasticity of the Nbs1 FHA domain that appears to have evolved to exploit and accommodate additional sequence requirements for Ctp1 phosphorylation.
Taken together, the above findings strongly suggested that S. pombe MRN, through the Nbs1 FHA domain, interacts directly with Ctp1 through redundant CK2 phosphorylation motifs. To test this idea, we first created an S. pombe strain containing Nbs1 in which the FHA-BRCT1 region had been deleted. As expected, these cells exhibited a severe hypersensitivity to low levels of DNA-damaging agents closely resembling the phenotypes of the G103D FHA mutant and Nbs1Δ strains (Figure 5D). Strikingly, however, hypersensitivity was completely rescued when the Nbs1 protein lacking the FHA/BRCT1 region was expressed as a fusion to full-length Ctp1 (Figure 5D). The ability of this fusion to circumvent a requirement for the FHA/BRCT1 domain thus provided strong support for Ctp1 binding as a primary function of the spNbs1 FHA/BRCT region.
Finally, we turned our attention to the role of the Nbs1 BRCT-repeat domain. Our spNbs1 structure revealed that the relative arrangement and packing of the tandem BRCT motifs resembles that previously observed in structures of other BRCT-repeat domains, many of which display phosphopeptide-binding functions (Figure 6A). SpNbs1 does not contain lysine and serine/threonine residues known to make direct phosphate contacts in other BRCT-phosphoprotein complexes (Figure 6B). They are, however, present in hNbs1 (Lys-160 and Ser-118) and although other positions involved in accessory main-chain peptide interactions and sequence-specific binding are not (Figure 6B), we nonetheless speculated that the Nbs1 N-terminus might possess dual phospho-recognition surfaces. In this regard, and in contrast to spNbs1, we noticed that titrations of wild-type hNbs1 with a variety of Mdc1-derived pSDpTD peptides consistently showed evidence of two interaction sites, manifested as binding stoichiometries significantly and reproducibly greater than 1:1 (Figure 6C; Table S2). Moreover, although mutation of the conserved phospho-binding residues in the hNbs1 FHA domain (Arg-28 or Ser-42), or in BRCT1 (Lys-160) had surprisingly small apparent effects on phosphopeptide-binding affinity (Figure 6C; Table S2), each of these reduced the stoichiometry by ~2-fold. These findings suggested that the effects of mutations in each individual domain were being partially obscured by interactions mediated by the other. Indeed, while a single FHA mutation (R27A) is sufficient to dramatically reduce phosphopeptide binding in spNbs1, we found that binding to hNbs1 was only abolished when both FHA and BRCT1 sites were disrupted in a S42A/K160M double mutant.
To further explore the apparent dual phospho-binding mode of hNbs1, we attempted to ‘uncouple’ FHA and BRCT domain contributions by using a peptide synthesized with a variant pSDpSD motif that lacks a phosphothreonine and would not, therefore, be expected to interact with hNbs1 FHA domain. In line with the lack of ITC evidence for an additional binding site in spNbs1 and the absence of any of the characteristic BRCT phospho-binding residues in this protein, no interaction of spNbs1 with the pSDpSD peptide was detectable (Figure 6D). In marked contrast, significant binding of the pSDpSD peptide to both wild-type hNbs1 and the R28A hNbs1 FHA mutant was maintained but, crucially, no interaction was seen with the K160M BRCT1 mutant. Consistent with these and previous biochemical analyses (Chapman and Jackson, 2008), tagged wild-type hNbs1 expressed ectopically in human cells efficiently co-immunoprecipitated Mdc1, while Mdc1 was undetectable in immunoprecipitates of Nbs1 bearing either the R28A or the K160M point-mutation (Figure 6E). Thus, both FHA-dependent and BRCT-dependent phospho-specific interactions by Nbs1 are required for its stable interactions with Mdc1 in cell extracts.
Collectively, these data establish a clear in vitro phospho-dependent binding activity for the hNbs1 BRCT-repeat domain. Furthermore, the lack of hNbs1 interactions with the mono-phosphorylated pSDTD or γH2AX peptides containing a single phosphoserine (Figure 4A and Figure S3) implies a preference of the hNbs1 BRCT-domain for doubly-phosphorylated motifs that differs considerably from previously reported pS/TxxF/Y phospho-dependent BRCT specificities (Manke et al., 2003; Yu et al., 2003). Consistent with this model, hNbs1 lacks the highly conserved arginine corresponding to residue 1699 in Brca1 and residue 1933 in Mdc1 that mediates crucial contacts with main-chain atoms of bound phosphopeptides. Taken together, these observations point to a rather distinct overall phospho-motif binding mode for the hNbs1 BRCT-repeats, although further structural analysis of hNbs1 phosphopeptide complexes will clearly be necessary to ascertain the precise molecular basis for such interactions.
Here we have presented a combined structural, biochemical and cell-biological investigation of the N-terminal region of the key DSB repair and DNA damage-signaling protein, Nbs1. Our crystallographic analysis has revealed a conserved super-modular FHA-BRCT2 architecture that, along with our supporting biochemical data, explains the apparent functional inter-dependence of the Nbs1 FHA and BRCT motifs first noted some time ago (Cerosaletti and Concannon, 2003).
Through comparison of the in vitro binding activities of fission yeast and human Nbs1 fragments, we have identified a similar FHA domain specificity for pSDpTD-like CK2 motifs originally identified as hNbs1 binding sites in Mdc1. On this basis, we have identified related sequences in S. pombe Ctp1 that are phosphorylated in cells, bind in a phospho-dependent manner and with high affinity to the spNbs1 FHA domain in vitro and which, when mutated, impair Ctp1 function in promoting resistance to DNA-damaging agents in vivo. The resemblance of these sites to others located in budding yeast Lif1 further suggests that the reported interactions between Lif1 and Xrs2 (Palmbos et al., 2008) are likely to employ a closely related binding mode. Significantly, an increase in affinity for di-phosphorylated sequences as compared with singly phosphorylated motifs appears to be a characteristic of both human and yeast Nbs1. Taken together with the existence of multiple SDTD motifs in Mdc1, Ctp1 and Lif1 and the combinatorial effects of Site 1 and Site 2 mutations on Ctp1 function and its interactions with spNbs1, our data suggest that Nbs1 has evolved an ability to respond to varying levels of overall phosphorylation of its various interacting partners.
Overall, the intimate association of the Nbs1 FHA and BRCT-repeat domains observed in our structure suggests that the fixed spatial relationship between the two regions may be important in allowing a concerted mode of FHA and BRCT domain activity. Our cell-biological and biochemical approaches have shown a phosphorylation-dependent binding capability for the human Nbs1 BRCT-repeat domain that is exercised through interactions distinct from those previously observed for the BRCT-repeats of Brca1 and Mdc1. The apparent necessity for binding of the human Nbs1 BRCT-repeat domain to di-phosphorylated motifs is intriguing and contrasts the rather more accessory effect of the additional phosphoserine on interactions of the Nbs1 FHA domain. It is thus tempting to speculate that these binding properties confer a degree of flexibility to hNbs1 responses through the use of separate phospho-binding domains with distinct, but overlapping specificities. At the molecular level, this arrangement might, for example, accommodate binding and juxtaposition of different protein interaction partners, each primed through phosphorylation by upstream kinases, in a manner reminiscent of 14-3-3 function (Gardino et al., 2006). Alternatively, the FHA and BRCT-repeat domains might recognize dual phospho-dependent interaction sites within a single molecule such as Mdc1. In support of this, we have shown that individual mutation of phospho-binding residues in either the FHA or BRCT-repeat domains of hNbs1 is sufficient to severely compromise its interactions with Mdc1 in human cells, consistent with previously reports that intact FHA and BRCT-repeat domains of Nbs1 (Cerosaletti and Concannon, 2003; Wu et al., 2008) and more than a single Mdc1 pSDpTD motif (Spycher et al., 2008) are necessary for MRN localization to DNA-damage foci in human cells.
In contrast to the human protein, the role of the spNbs1 BRCT-repeat domain remains unclear and, consistent with the absence of residues analogous to those imparting direct phosphate interactions in Mdc1, Brca1 and also hNbs1, we have been unable to demonstrate a phospho-binding function for the spNbs1 BRCT-repeat domain. The presence of alternative interaction surfaces in this domain that might mediate phospho-binding cannot be discounted but it may also be that the spNbs1 BRCT-repeats function through as yet uncharacterized phospho-independent interactions. Indeed, the phospho-independent binding capability reported for FHA domains (Li et al., 2002; Nott et al., 2009) and previously seen in several BRCT-repeat domain complexes (Derbyshire et al., 2002; Doré et al., 2006; Joo et al., 2002; Sibanda et al., 2001) could potentially specify additional, evolutionarily conserved binding functions for the Nbs1 N-terminus, thus expanding the versatility of this region as a platform for promoting DNA repair and DNA-damage checkpoint signaling.
Synthetic genes encoding spNbs1 (1-324) and hNbs1 (1-382) were produced by GENEART AG (www.geneart.com) and expressed in pET22b as C-terminal (His)6-tag fusion proteins for purification utilizing a Ni-Sepharose affinity chromatography followed by Source 15Q ion-exchange and Superdex 75 size exclusion chromatography (GE Healthcare). Mutations were introduced using the QuikChange protocol (Strategene) and all constructs were verified by DNA sequencing. For preparation of seleno-methionine (SeMet) labelled spNbs1, plasmid was transformed into B834 (DE3) cells (Novagen). Protein expression was performed in supplemented M9 minimal media (Molecular Dimensions) and purified as for the native protein. SeMet incorporation was estimated to be > 98% by mass spectrometry.
Growth of diffraction quality crystals of native and SeMet incorporated spNbs1 was aided by carboxypeptidase A digestion. Crystallization was carried out by sitting-drop vapour diffusion using an Oryx8 nano-dispensing robot (Douglas Instruments). Crystals grew at a protein concentration of 8 mg/ml against a reservoir containing 0.1 M MES pH 6.0, 0.1 M MgCl2 and 8% w/v PEG 6000. Diffraction data were collected at 100 K at a wavelength corresponding to the peak of the anomalous (f”) signal on beamline I04 at the Diamond Light Source, UK from crystals cryoprotected with 30% glycerol. Methods for crystal improvement by reductive methylation along with details of the structure solution and refinement and biochemical/cell biology procedures are provided in Supplemental Material.
We thank Hiroshi Iwasaki for the kind gift of reagents, P. Walker for invaluable assistance with X-ray data collection, K. Rittinger for advice on ITC measurements and support from the NIMR large-scale laboratory. EH is supported by a CRUK grant C20600/A6620 and TC acknowledges funding by an MRC Program Grant G0600233. Research the SPJ laboratory is supported by grants from Cancer Research UK, the European Union (Integrated Project DNA repair, LSHG-CT-2005-512113, and Genomic Instability in Cancer and Precancer, GENICA, HEALTH-F2-2007-201630), and core infrastructure funding from Cancer Research UK and the Wellcome Trust. SJS is grateful to the Medical Research Council, UK for continuing support and Diamond Light Source, UK for synchrotron access and beamline support. Atomic coordinates have been deposited with the Protein Data Bank - accession numbers 1XXX and 1YYY.
The authors state that they have no competing financial interests.