Structural mechanisms underlying signaling in the cellular response to DNA double strand breaks

Abstract

DNA double strand breaks (DSBs) constitute one of the most dangerous forms of DNA damage. In actively replicating cells, these breaks are first recognized by specialized proteins that initiate a signal transduction cascade that modulates the cell cycle and results in the repair of the breaks by homologous recombination (HR). Protein signaling in response to double strand breaks involves phosphorylation and ubiquitination of chromatin and a variety of associated proteins. Here we review the emerging structural principles that underlie how post-translational protein modifications control protein signaling that emanates from these DNA lesions.

1. Introduction

Diverse forms of genotoxic agents, such as ionizing radiation, reactive oxygen species, and chemotherapies can lead to the generation of single and double stranded breakages in the DNA backbone. Double stranded breaks (DSBs)¹ can also arise when a replicating polymerase stalls upon encounter with a damaged DNA template, resulting in dissociation of the polymerase from the replicative helicase and rearrangement of the fork structure in a process termed replication fork collapse [1-2]. DSBs are among the most lethal DNA lesions as just a single DSB left unrepaired can cause chromosomal rearrangements and genomic instability.

In actively replicating cells, DSBs can often be repaired in a largely error-free process via homologous recombination (HR), utilizing the undamaged sister chromatid and much of the same cellular machinery that is used during meiotic recombination [1-2]. Efficient HR repair serves as an effective block to tumourigenesis, as mutations in key players in this pathway, such as BRCA1 and BRCA2, can lead to dramatically enhanced cancer risks [3-4]. In addition, since radiotherapy and many DNA-targeted chemotherapies inhibit tumor progression through generation of DSBs, the selective sensitization of tumor cells to these lesions has gained attention as a route to new cancer therapies [5].

DSB repair by HR relies on a cascade of protein-protein interactions that mediates chromosomal structural changes, regulation of the cell cycle, and ultimately recruitment of the HR machinery [6-9]. In this review we focus on the structural principles that have been uncovered that govern how protein phosphorylation and ubiquitination events control protein signaling interactions within this pathway. It should be noted that other pathways, especially non-homologous end joining (NHEJ), are also critical for DSB repair and the reader is referred to recent reviews on this topic [10-11].

2. Overview of protein signaling in the response to DNA double strand breaks

Protein signaling events that emanate from DSBs initiate with specialized protein complexes that recognize the severed ends of the DNA (Fig. 1). The major protein complex thought to recognize DSBs to trigger the HR repair pathway is the MRN (Mre11-Rad50-NBS1) complex, a hetero-hexameric complex that can directly recognize a single DSB or a pair of DSBs to potentiate synapsis (Fig. 1A). A series of elegant structural studies are beginning to reveal how DNA binding and ATPase hydrolysis within the Rad50 subunit induce conformational changes that ultimately drive initiation of DSB signaling [12-19]. Through mechanisms that are still poorly understood, MRN then activates the PI3 kinase ATM, which phosphorylates SQ/TQ sites, to initiate the first steps in DSB signalling. One of the best-characterized and earliest targets of ATM is the histone H2A variant, H2AX, which contains a C-terminal tail not found in H2A [20-21]. Phosphorylation of the H2AX tail (to yield the variant γH2AX) provides a docking site for the next critical protein in DSB signaling, MDC1 (Fig. 1B) [22-24]. MDC1 itself is also phosphorylated by CK2 at a series of SDT sites near its N-terminus that bind the NBS1 component of MRN, thereby enhancing retention of MRN at chromatin near the DSB [25-28]. MDC1 is also phosphorylated at distinct sites by ATM/ATR to provide a binding site for the next protein in the pathway, RNF8 (Fig. 1C) [29-31]. RNF8 is an E3 ubiquitin ligase that partners with the E2 ubiquitin conjugating enzyme Ubc13 to tag H2A/H2B histones near the DSB with polyubiquitin chains. Ubc13 in complex with its partner, Mms2, is a novel E2 that builds polyubiquitin chains that are connected by isopeptide bonds between the ε-amino group of Lys63 of one ubiquitin and the C-terminal carboxylate of the next ubiquitin in the chain [32-37]. RNF8 is not the only E3 involved in generating these chains. A second E3, RNF168, has also been identified in this process (Fig. 1D). Mutation of RNF168 has been implicated in a human immunodeficiency and radiosensitivity disorder termed the RIDDLE syndrome [38]. Initial work indicated that RNF168 acts after RNF8 in the pathway, potentially targeting chromatin regions that were already polyubiquitinated via the RNF168 ubiquitin interacting motifs (UIMs) [39-41]. More recently however, it has been suggested that RNF168 may also play a role in the initial targeting of ubiquitin onto H2A or H2AX in chromatin [40]. The generation of Lys63-linked polyubiquitin is also regulated by a deubiquitinating enzyme, OTUB1, which directly binds and inhibits Ubc13 activity to down-regulate DSB signaling [43-46]. Interestingly, RNF8 also plays a role in the generation of Lys48-linked polyubiquitin near DSBs, which likely is involved in the proteasome-dependent turnover of proteins near the DSB [47].

Fig. 1.

Overview of signaling from DNA double strand breaks (DSBs) to initiate repair by homologous recombination (HR). Phosphorylation events are indicated by red circles, ubiquitylation by purple circles and the domains mediating protein-protein interactions (BRCT, FHA, tUIM, RING) are highlighted.

While the effects of Lys63 polyubiquitylation on chromatin structure have not been determined, it appears that a critical function of these chains is to further recruit other protein complexes to chromatin. The major protein that recognizes these chains in the DNA damage response is RAP80, which contains a novel tandem UIM domain that specifically recognizes at a minimum Lys63-linked diubiquitin (Fig. 1E) [48-50]. More recently, RAP80 has also been shown to bind to ubiquitin chains that also contain the ubiquitin-like protein SUMO via an additional SUMO-interacting motif in RAP80. Critical to the assembly of the ubiquitin-SUMO hybrid chains is another E3 ligase, RNF4, which binds SUMO and links new ubiquitin molecules to pre-existing SUMO chains [51]. RAP80 itself exists as a complex with at least two other proteins, the deubiquinating enzyme BRCC36 and another protein, Abraxas [52-55]. Abraxas does not contain known functional domains but is phosphophorylated at a likely cyclin-dependent kinase site in its C-terminal tail to form a binding site for the breast and ovarian cancer associated protein, BRCA1 (Fig. 1E). BRCA1 can also exist in at least two other complexes; one with the nuclease-associated protein, CtIP [56-57], and the other with the Fanconi anemia-associated helicase, FancJ (also known as BACH1/BRIP1) [58-59]. BRCA1 is thought to be critical for the establishment of HR repair, although the biochemical and structural mechanisms by which this occurs are yet to be elucidated. Part of BRCA1 function likely involves its interaction with PALB2, which could recruit BRCA2 and the HR machinery [60-62]. The N-terminal region of BRCA1 adopts a heterodimeric E3 ubiquitin with another protein partner, BARD1, however the role of this ligase in the HR pathway remains unclear [63-64].

3. Phosphorylation signaling – recognition of phosphate marks by BRCT and FHA domains

The response to DNA double strand breaks initiates with phosphorylation events, largely triggered by PI3 kinases ATM and ATR. In general these phosphorylations create binding sites for specific phospho-peptide recognition protein modules. In the DNA double strand break response, the most common and important modules that mediate these interactions are the BRCT and FHA domains.

3.1. The BRCT domain (Fig. 2)

Fig. 2.

Phospho-peptide recognition by BRCT domains. (A) Structure of the BRCA1 BRCT domain (blue) bound to a pSer-x-x-Phe target peptide (yellow) with the two tandem BRCT domains indicated. (B) Details of pSer recognition by the BRCA1 BRCT. (C) Details of recognition of the Phe +3 residue and C-terminal carboxylate. Note that the MDC1 BRCT uses a similar mechanism to recognize the C-terminus of γH2AX.

BRCT domains are 90-100 amino acid domains first identified at the C-terminus of BRCA1 [65-66], where they are essential for the tumour suppressor function of the protein [67-68]. BRCT domains often occur as a pair of tandem repeats that pack together in a head to tail arrangement (Fig. 2A) [69]. Shortly after the demonstration that these tandem domains function as phospho-peptide modules [59,70], a series of studies demonstrated that the tandem BRCT domains of both BRCA1 and MDC1 bind their phospho-peptide targets though a similar mechanism [24,71-79]. In general, these proteins recognize phospho-serine/phospho-threonine residues via a conserved phosphate binding pocket in the N-terminal repeat that provides ligands for the phosphate oxygen atoms (Fig. 2B) [68,80]. The BRCA1 and MDC1 BRCTs bind the phosphorylated residue in an orientation that favors pSer binding over pThr, due to steric effects of the pThr methyl group [81]. In addition, these proteins also specifically recognize a Phe/Tyr at the +3 position with respect to the pSer (Fig. 2C) [59,70]. This recognition is via a hydrophobic cleft formed at the interface of the two BRCTs. Interestingly, this pSer-x-x-Phe/Tyr peptide motif is positioned at the C-terminus of both the MDC1 target, γH2AX, as well as the BRCA1 target, Abraxas. Both BRCA1 and MDC1 directly recognize the peptide C-terminal carboxylate via a salt bridging interaction from a conserved arginine residue, which significantly enhances binding affinity [72].

3.2. FHA domain – phosphopeptide interactions (Fig. 3)

Fig. 3.

Phospho-peptide recognition by FHA domains. (A) Structure of the FHA domain of RNF8 (green) bound to its pThr-containing target peptide (orange). (B) Detailed view of the recognition of the pThr and neighboring residues by the RNF8 FHA. (C) Phospho-peptide binding by the NBS1 FHA domain drives a conformational change that leads to a rearrangement of BRCT-BRCT packing. Structure of the NBS1 FHA-BRCT-BRCT domain bound to a CtIP peptide is in pink, while the apo structure is in grey. The structures were aligned on their FHA domains to reveal the resulting rotation of the C-terminal BRCT with respect to the rest of the structure.

FHA domains, initially identified as a signaling domain in certain Forkhead transcription factors [82], are phospho-peptide binding modules that fold into a β-sandwich structure in which the phospho-threonine containing peptide is bound by loops protruding from one edge of the β-sandwich (Fig. 3A) [83]. FHA domains are found in MDC1 [84-86], Nbs1 [87-89], and RNF8 [29] where they are essential for critical phospho-peptide interactions. In contrast to BRCT domains that recognize both pSer- and pThr-containing sequences, FHA domains recognize only pThr-containing targets.

The structure of the RNF8 FHA bound to an optimized pThr peptide provides a general model for FHA target interactions (Fig. 3A) [29]. FHA domains recognize the pThr through a conserved pocket that provides hydrogen-bonding and salt-bridging ligands for the phosphate (Fig. 3B) [88]. Specificity for the pThr is provided by a small hydrophobic pocket that specifically recognizes the γ-methyl of the pThr [90]. Like BRCT peptide recognition, the RNF8 FHA also shows strong selectivity for a Phe/Tyr residue at the +3 position, and the structure reveals a small hydrophobic pocket that contacts this residue [29]. The MDC1 FHA domain appears to recognize an internal, N-terminal ATM phosphorylation site, stabilizing the self-association of MDC1 FHA in response to phosphorylation [84,85].

Intriguingly, the NBS1 FHA domain does not exist as an isolated domain but instead is found as part of a larger fusion of the N-terminal FHA with a tandem BRCT repeat [87,89,91]. Structures of the NBS1 FHA-BRCT-BRCT with target peptides from CtIP [87,89] reveal the conserved recognition of the pThr residue by the FHA, as well as recognition of the neighboring aspartate. However, peptide binding appears to trigger a conformational change that initiates from binding-induced alterations in the FHA peptide binding surface that result in a rotation of the C-terminal BRCT domain with respect to the rest of the structure (Fig. 3C) [89]. This conformational change could present a mechanism by which binding at the FHA ultimately controls subsequent phospho-peptide recognition at the BRCT. The pThr-Asp FHA target sequence revealed in this structural work is likely recognized in a similar way as the pSer-Asp-pThr-Asp MDC1 sequence that is also recognized by this protein [26,84-85,87,89].

4. Generation and recognition of Lys-63 – linked polyubiquitin in DSB signaling

4.1. Generation of Lys63-linked polyubiquitin by Mms2-Ubc13

Early structures of human Mms2-Ubc13 [36] and the closely related S. cerevisiae Uev1a-Ubc13 complex [37] revealed that Mms2 and Uev1a adopt an E2-like fold, although lacking the active site cysteine that characterizes true E2 enzymes. Mms2 binds to a hydrophobic patch on Ubc13 distant from the catalytic cysteine and not conserved in other E2s, explaining why only Ubc13 participates with Mms2 to build Lys63 polyubiquitin. Further NMR and crystallographic work [32,34-35] revealed that Mms2 provides a binding site for an acceptor ubiquitin in an orientation such that its Lys63 is positioned to attack the thioester of the acceptor ubiquitin at the Ubc13 active site (Fig. 4A).

Fig. 4.

Structure and regulation of Mms2-Ubc13. (A) Mms2 (orange) binds an acceptor ubiquitin (purple) such that its Lys63 is positioned to attack the thioester of the donor ubiquitin bound to the active site Cys87 of Ubc13 (blue). (B) RNF8 RING domain exists as a coiled-coil – stabilized dimer (green) and binds to Ubc13. Highlighted in red is Asp443, which regulates the ability of the RING domain to target nucleosomes for ubiquitination. (C) OTUB1 (green) binds Ubc13~Ub, blocking RNF8 access and interfering with Mms2 binding to Ubc13. In this figure, Mms2 (orange surface) is docked onto Ubc13 to illustrate the predicted clash with the N-terminal OTUB1 helix (αN). OTUB1 binding is allosterically regulated through the binding of free ubiquitin.

4.2. RNF8 and RNF168 ubiquitin ligases in DSB signaling

Mms2-Ubc13 alone only slowly makes polyubiquitin chains but the rate of chain formation is greatly enhanced by RNF8 [30-31,92-93]. Structural studies have revealed that the RNF8 RING domain adopts the zinc finger fold found in other RING E3 ligases [92,94] (Fig. 4B). However, a striking novel structural feature of RNF8 is its long coiled-coil immediately N-terminal to the RING domain, which is critical for dimerization, as well as binding and activation of Mms2-Ubc13 [92]. The RNF8 RING binds a surface on Ubc13 that appears to be a largely conserved E3 binding site in other E2 enzymes [94-96]. In contrast, RNF168 RING domain does not appear to have the same coiled-coil structure, and the crystallized RNF168 RING construct is in fact a monomer [42,92,97]. The RING domain of RNF168 does not interact tightly with Ubc13 and is much less efficient in the stimulation of Ubc13 activity [92,97]. Recent data have suggested that RNF168 may play an important role in linking the initial ubiquitin molecule to H2A within nucleosomes [42]. This work suggests that single positively charged residue (Arg57) exposed on the surface of the RNF168 RING domain mediates this specificity [42]. The analogous residue in RNF8 is negatively charged (Asp443) providing an explanation for the finding that only RNF168 and not RNF8 can monoubiquitinate nucleosomes in vitro.

4.3. Regulation of Ubc13 by OTUB1

Lys63-linked chain building is also subject to repression via a mechanism that appears to be sensitive to feedback inhibition from the free ubiquitin pool in the cellular milieu of the DSB. Central to this mechanism is OTUB1, a deubiquitinase (DUB) enzyme initially isolated via its association with ovarian cancer [98-99]. OTUB1 specifically binds and hydrolyzes Lys48-linked polyubiquitin using its active site cysteine, which attacks and hydrolyzes the isopeptide ubiquitin-ubiquitin linkage [100]. Intriguingly, the functional active site of OTUB1 is not required to inhibit Ubc13. Instead, OTUB1 inhibits Lys63-linked chain formation through direct interactions with Ubc13 [44-45] (Fig. 4C). These interactions are stabilized by interactions with free ubiquitin as well as donor ubiquitin linked to the Ubc13 active site [43,46]. Structural studies of these complexes [43,46] reveal that the donor ubiquitin binds an otherwise disordered N-terminal segment of OTUB1 in a manner that is facilitated by binding of a second, free ubiquitin. Complex formation likely represses Lys63-linked chain formation in at least three ways. First, binding of OTUB1 occludes the RNF8 docking site on Ubc13, which would reduce efficiency and could facilitate dissociation of Ubc13 from the DSB. Second, the N-terminal helical segment and non-covalent ubiquitin could partially sterically block Uev1/Mms2 binding to Ubc13. Finally, the orientation of the covalently linked, donor ubiquitin may be such that it is less reactive to attack from the incoming lysine of the acceptor ubiquitin. Support for this comes from a recent structure of an E2 enzyme-ubiquitin conjugate with a bound E3 ligase [101-102]. This work indicates that E3 ligases may interact with the donor ubiquitin to hold the ubiquitin in a conformation that enhances the reactivity of the thioester. The OTUB1-Ubc13~Ub structures reveals an alternative conformation that may instead stabilize the thioester to inhibit chain formation.

4.4. Recognition of Lys63 polyubiquitin by RAP80

An array of helical motifs and domain structures have been identified that specifically recognize ubiquitin and participate in ubiquitin-mediated signaling [103]. One of the simplest and best studied is the UIM (ubiquitin interacting motif) that adopts a single helix that binds to the hydrophobic face of the ubiquitin β-sheet centered on Ile44. In general, single UIM domains bind weakly to single ubiquitin units with K_D’s >100 μM [104]. In DSB signaling, RAP80 specifically recognizes Lys63 polyubiquitin through a novel tandem UIM motif [48-50] (Fig. 5). Biophysical and structural studies have revealed that both UIMs engage individual ubiquitins in model Lys63-linked diubiquitin [48-50]. Specificity for Lys63-linked diubiquitin is achieved through the linkage of the two UIMs [48-50]. Upon binding diubiquitin, this region adopts a helical structure so that the entire UIM is a single contiguous helix. This helical structure positions the two ubiquitin binding surfaces on the same side of the helix and at the correct distance to allow cooperative binding of the diubiquitin. Binding affinities of model UIMs oscillate as linker size is increased in phase with the helical repeat, providing support for this model [48-50]. Lys48-linked chains cannot adopt the extended structure of the Lys63-linked chains, and this likely explains the fact that these chains are bound with much lower affinity [49]. The picture may be even more complex has it has been recently shown that RAP80 also contains a binding motif for the small, ubiquitin-like protein, SUMO, and that RAP80 can bind with <0.2μM affinity to hybrid Lys63-linked ubiquitin/SUMO chains [51,105].

Fig. 5.

Recognition of Lys63-linked di-ubiquitin by the RAP80 tandem UIM (tUIM). The tUIM is shown in green and grey and the sequence of the human RAP80 tUIM is displayed below. Conserved acidic residues in the tUIM and the positively charged residues in ubiquitin with which they interact are shown as sticks. The conserved Ala-Ser residues in the tUIM and Ile44 in ubiquitin that make hydrophobic contact are displayed as spheres.

5. Future challenges in the structural study of DSB signalling

While significant advances have been made in understanding the structures and biochemical functions of isolated components of DSB signalling, we lack structural data that integrates these components into a larger functional framework. Part of the challenge is that many of the signalling proteins involved are enormous (2000 residues or more) and are thought to be highly flexible and post-translationally modified. There have been some recent successes, most notably advances in our understanding of MRN structure and function (section 2), however how MRN interacts with and activates PI3 kinases is still poorly understood. The recent successes in the crystallization and low resolution structure determination of the PI3 kinase family member, DNA-PKcs gives some hope that this question will ultimately be amenable to structural analyses [106]. Another major challenge is to visualize these signalling interactions in the context of nucleosomal structures, to understand the larger impact of these interactions on chromatin structure and dynamics [107]. Many of these challenges will involve the use of hybrid structural methods, incorporating X-ray crystallography with other methods such as electron microscopy or small angle X-ray scattering (SAXS) that can be more amenable to the study of large flexible complexes.

Finally, the structural analyses of DSB signalling interactions could provide key information for the rational design of small molecule inhibitors of this system. While the interactions that have been detailed here are not classically “druggable”, there have been some successes. For example, small molecule inhibitors of Ubc13 have recently been isolated that repress NF-κB signalling (which depends on Ubc13-dependent Lys63-linked polyubiquitylation) and inhibit the proliferation of lymphoma cell lines that depend on NF-κB activity [108]. In addition, there have also been reported small molecule and peptidic inhibitors of interactions between the BRCA1 BRCT domain and its phospho-peptide targets [109-110] and these inhibitors can sensitize human cells to PARP inhibition [111]. This provides support for the notion that selective inhibition of the DSB signalling pathway could provide routes for new cancer therapies.

Highlights.

• Double strand break (DSB) signaling is initiated with lesion recognition by MRN

• DSB recognition leads to protein kinase activation and H2AX phosphorylation

• DSB-linked phosphorylation is specifically recognized by BRCT and FHA domain proteins

• phosphorylation leads to modification of chromatin by Lys63-linked polyubiquitin

• Lys63-linked polyubiquitin is recognized by RAP80 tUIM and recruitment of BRCA1