ATP‐dependent DNA end recognition and nucleolytic processing are central functions of the Mre11/Rad50 (MR) complex in DNA double‐strand break repair. However, it is still unclear how ATP binding and hydrolysis primes the MR function and regulates repair pathway choice in cells. Here, Methanococcus jannaschii MR‐ATPγS‐DNA structure reveals that the partly deformed DNA runs symmetrically across central groove between two ATPγS‐bound Rad50 nucleotide‐binding domains. Duplex DNA cannot access the Mre11 active site in the ATP‐free full‐length MR complex. ATP hydrolysis drives rotation of the nucleotide‐binding domain and induces the DNA melting so that the substrate DNA can access Mre11. Our findings suggest that the ATP hydrolysis‐driven conformational changes in both DNA and the MR complex coordinate the melting and endonuclease activity.
The X‐ray structure of the ATP loaded Mre11/Rad50 DNA repair complex shows that DNA runs along a groove between the Rad50 dimers. ATP hydrolysis causes DNA melting and a conformational change, allowing access to Mre11 active site.
Rad50 binds the DNA minor groove via a groove between the Rad50 dimer subunits.
The Mre11 active site in the full‐length MR complex accesses melted but not duplex DNA.
ATP hydrolysis‐driven lobe‐rotation results in partial DNA unwinding, enabling DNA to access the Mre11 active site.
The structure may indicate that the MR complex recognizes DNA a few bases away from the break end.
DNA double‐strand breaks (DSBs) is one of the most detrimental types of DNA damage, which can be repaired via homologous recombination (HR) or non‐homologous end‐joining (NHEJ) repair pathway (Symington & Gautier, 2011). The Mre11/Rad50/Nbs1 (MRN) complex in mammals (the Mre11/Rad50/Xrs2 (MRX) complex in Saccharomyces cerevisiae) plays a central role in DSB repair by recognizing and resecting the damaged DNA ends, and transducing a signal via activation of ataxia telangiectasia mutated (ATM) kinase (Stracker & Petrini, 2011; Lafrance‐Vanasse et al, 2015). In addition, MRN complex plays a crucial role in telomere maintenance, meiotic recombination, and class switch recombination in B cells (Boulton & Jackson, 1998; Furuse et al, 1998; Moreau et al, 1999; Reis et al, 2012). The importance of MRN function to DNA metabolism is underscored by the fact that deletion of any of its three components leads to embryonic lethality in mice, and hypomorphic mutations in MRN components result in various developmental and neurodegenerative disorders (Varon et al, 1998; Stewart et al, 1999; Buis et al, 2008; Waltes et al, 2009).
In the MR complex, Mre11 dimer is responsible for Mn2+‐ or Mg2+‐dependent nuclease activities, which include 3′–5′ exonuclease, endonuclease, and hairpin opening activities (Connelly et al, 1997; Paull & Gellert, 1998; Hopfner et al, 2000a,b; Trujillo & Sung, 2001; Williams et al, 2008; Park et al, 2011; Schiller et al, 2012). Bacterial homolog of the Mre11/Rad50 complex, SbcD/SbcC, also cleaves the hairpins at inverted repeats and facilitates the replication restarts (Darmon et al, 2010). At the initial stages of DSB repair and meiotic recombination, the yeast MRX complex together with Sae2 nicks the DNA at a set distance from the damaged end, and this process is followed by 3′–5′ resection toward the DSB (Limbo et al, 2007; Garcia et al, 2011; Cannavo & Cejka, 2014; Shibata et al, 2014). Other nucleases, such as Exo1 or BLM/Dna2, perform further resection in the 5′–3′ direction, away from the DSB (Mimitou & Symington, 2008; Zhu et al, 2008; Cejka et al, 2010). These nuclease activities are crucial for removing obstructed DNA ends for HR (Liu et al, 2002; Neale et al, 2005).
Rad50 is an ATP‐binding cassette ATPase that has two lobes, each of which contains Walker A and B motifs. Rad50 shares a similar head domain (nucleotide‐binding domain, NBD) with members of the structural maintenance of chromosomes (SMC) family (Hopfner et al, 2000a,b). ATP‐binding and hydrolysis activities of Rad50 are crucial for binding, unwinding, and tethering of the DNA ends, as well as for ATM kinase activation and regulating the endonuclease activity of the MR complex (Raymond & Kleckner, 1993; Paull & Gellert, 1999; Chen et al, 2005; Lee & Paull, 2005; Deshpande et al, 2014).
Two MR complexes form an elongated shape that can be further divided into ATPase head, zinc‐hook domain, and a long anti‐parallel coiled‐coil arm that connects the head and the hook domains (Hopfner et al, 2001, 2002; de Jager et al, 2001; Moreno‐Herrero et al, 2005). The head region contains two Rad50 ATPase domains and two Mre11 nuclease proteins. The coiled‐coil region of Rad50 at the base of the ATPase head interacts with the C‐terminal three helix‐bundle (or helix‐loop‐helix) of Mre11 (Lammens et al, 2011; Lim et al, 2011; Williams et al, 2011; Möckel et al, 2012).
Structural analyses of prokaryotic MR catalytic domain (MRcd) lacking the coiled‐coil region and zinc‐hook domain have shown that ATP binding and hydrolysis switches the conformation of the MR complex. Head region of MR tetramer complex forms a closed state in the presence of ATP, in which the two Rad50 ATPase domains engage to sandwich two ATP molecules and occlude the active site of Mre11 nuclease dimer (Lim et al, 2011; Möckel et al, 2012). ATP hydrolysis drives the rotation of the two lobes of Rad50 as much as 30° and leads to disengagement of Rad50 dimer, which allows DNA to enter the Mre11 active site for resection (Lammens et al, 2011). This ATP‐dependent conformational change regulates nuclease activity of the MR complex. Based on structural and biochemical data, it has been proposed that the ATP‐bound MR complex with a closed head domain is important for directing the NHEJ pathway, whereas the ATP‐free MR complex with an open head domain directs HR repair (Deshpande et al, 2014).
Despite an elegant model how ATP binding and hydrolysis induces the conformation change of the MR complex, it does not fully explain biochemical properties of ATP‐free MR. For example, based on the open conformation of ATP‐free MRcd complex, this MR complex should still be able to cleave its substrate DNA even in the absence of ATP because there is no restriction in accessing dsDNA to Mre11. Surprisingly, the full‐length prokaryotic MR complex or eukaryotic MRN/X complex does not exhibit endonuclease activity in the absence of ATP (Connelly et al, 1997; Hopfner et al, 2000a,b; Lee et al, 2003; Chen et al, 2005; Deshpande et al, 2014). These studies thus indicate that the ATP‐free intact MR complex may adopt a conformation, which could be different from the open conformation of the ATP‐free MRcd complex. In this case, it is unclear how the substrate DNA can access the Mre11 active site in the full‐length MR complex, and how ATP promotes the DNA unwinding and endonuclease activities of the MR complex.
Here, by using structural, biochemical, and genetic approaches, we examined a binding model for the ATPγS‐bound MR and DNA complex. Crystal structure of the ATPγS–MR–DNA complex revealed that double‐stranded (ds) DNA binds to the central groove at the dimeric interface of Rad50 dimer in a partly deformed conformation. DNA binding analyses suggest that the full‐length nucleotide‐free MR complex forms an ATPase head arrangement, different from that of the coiled‐coil deleted MRcd complex, in such a way that dsDNA cannot access Mre11. ATP hydrolysis triggers the conformational change and the domain rotation of Rad50, which subsequently melts the dsDNA. Biochemical analyses showed that a bubble but not intact duplex DNA can be cleaved by Mre11 in the ATP‐free state, suggesting that ATP‐dependent DNA binding, DNA melting, and Rad50 dimer disengagement are all essential for the efficient resection of DNA ends by the MR complex. Based on the integrative approach, we propose a new model by which the ATP hydrolysis‐driven concerted motion of Rad50 and Mre11 couples the DNA binding and melting by Rad50 to endonucleolytic cleavage by Mre11 in DSB repair process.
Overall structure of the ATPγS–MR–DNA complex
To understand how ATP promotes the MR complex to recognize the DNA, we determined the structure of the ATPγS–MR–DNA complex. We crystallized the complex containing Methanococcus jannaschii (Mj) Mre11 (full‐length) and Rad50 ATPase domain (residues 1–190 and 825–1,005) in the presence of ATPγS and a 25‐base pair (bp) dsDNA containing 2‐nt 5′ overhangs (Fig 1A and B). The crystal contains two molecules of the MR complex in the asymmetric unit, which are arranged in an ATPγS‐bound MR tetramer similar to that described previously (Lim et al, 2011; Möckel et al, 2012). The 3.1 Å resolution electron density map revealed a 22‐bp dsDNA with a 3‐nt overhang at one 5′ end and a 1‐nt overhang at the other 5′ end, which runs across the central groove at the interface of Rad50 dimer, with its central axis tilted approximately 15° relative to the dimer interface (Fig 1C and D, Table 1, Appendix Fig S1, Movie EV1).
The 38‐Å‐long and 25‐Å‐wide central groove at Rad50 dimeric interface is formed only in the presence of ATP or an analogue (Figs 2A and EV1, Hopfner et al, 2000a,b; Lim et al, 2011; Williams et al, 2011; Möckel et al, 2012). The DNA duplex interacts with the Rad50 ATPase dimer in a near‐symmetrical manner. Both Rad50 monomers bind to each end of the DNA strand via three motifs, namely loops α1‐α2, β7‐α3, and β5‐β6 at the N‐terminal lobe (or lobe I, residues 1–137 and 939–1,005) (Fig 2A–C). We refer to each Rad50 ATPase domain as Rad50 A (green) and Rad50 B (blue) and each DNA strand as “template” (orange) and “non‐template” (yellow) (Fig 2A and E). The structure of the DNA bound at the central groove of Rad50 dimer is consistent with previous modeling‐based prediction, as well as functional studies showing that ATP binding‐induced dimerization is important for DNA binding (Raymond & Kleckner, 1993; Hopfner et al, 2000a,b, 2001; Lim et al, 2011; Rojowska et al, 2014).
ATP‐dependent symmetric DNA recognition by the MR complex
Rad50‐DNA contacts occur primarily between phosphate groups and protein side‐chain groups in a sequence‐independent manner (Movie EV2). Interactions between one Rad50 molecule (Rad50 A) and one end of the DNA strand are similar to those between another Rad50 molecule (Rad50 B) and the opposite end of the DNA strand (Fig 2B–E). The most noticeable features are loop α1‐α2 (residues 51–58), which wedges the minor groove and loop β5‐β6, which protrudes into the major groove (Fig 2B and C). Three successive phosphate groups in the minor groove interact with lobe I. Asn57 and the main chain of Tyr58 bind to the first phosphate group (Cyt8′), Ala52 (main‐chain) and Tyr58 interact with the second phosphate group (Cyt7′), and Thr107 and Ser109 recognize the Ade6′ phosphate group. Arg92 side chain interacts with the Ade5′ and Gua4 bases in the major groove (Fig 2B and E). A conserved Arg86 is located near the Cyt7′ phosphate (4.2 Å), and its guanidinium group stacks against the backbone of loop α1‐α2 and interacts with Asn112, thereby stabilizing the two DNA‐interacting loops, α1‐α2 and β7‐α3. Interaction between Rad50 and DNA in the complex structure is reinforced by the cluster of positively charged residues along the groove, as supported by mutational analyses of Rad50 (Fig EV1B; Lim et al, 2011; Rojowska et al, 2014).
Comparison of two different DNA binding modes by the nucleotide‐bound MR complex
Recently determined structure of the Thermotoga maritima (Tm) Rad50 nucleotide‐binding domain (NBD) complexed with the C‐terminal motif of TmMre11 and a 15‐bp dsDNA (4W9M) revealed a mechanism by which β‐γ‐imidoadenosine 5′‐phosphate (AMP‐PNP)‐bound Rad50 dimer recognizes the DNA (Rojowska et al, 2014). In TmRad50‐DNA structure, DNA binds to only one of the two Rad50 NBDs, mainly to a strand‐loop‐helix at the N‐terminal lobe (equivalent to β7‐loop‐α3 of MjMR) with additional contact with the coiled‐coil region. The overall location and orientation of DNA within TmRad50 markedly differ from those within MjMR (Fig EV2). The presence of two different binding sites for DNA from the two structures can be interpreted that Rad50 may interact with DNA through multiple sites in the head domain.
Although most of the residues interacting with DNA via their side chains are polar or charged, they are not conserved in three available prokaryotic Rad50 structures (Fig 2D and E). Weak conservation of the DNA‐binding residues at the central groove suggests that these residues in Rad50 molecules from various species may contribute differently in DNA recognition. Upon DNA binding, a 1,945 Å2 surface area of MjRad50 becomes buried, whereas a 500 Å2 area is buried in TmRad50, suggesting that symmetric DNA recognition by MjRad50 involves more extensive contacts than asymmetric binding. Alternatively, difference in DNA binding between two structures may be attributed to the differences in the DNA length used in the crystallization. The distance between each α1‐α2 loops in the MjRad50 dimer is about 18 bp (Fig 2A).
Full‐length and truncated MR complexes exhibit different DNA‐binding patterns
To validate the DNA binding by the ATPγS–MjMR complex observed in the crystal structure, we introduced four mutations (Δ53–55 (deletion of 53–55), R86E, R92E, and T107E, Figs 2B and 3A) into the residues of MjRad50 that likely contribute to DNA contact or their immediate vicinity, and examined the DNA binding activities of the MjMRcd (residues 1–190 and 825–1,005) mutants using a closed circular dsDNA (φx174RFII). Interestingly, all MjMRcd wild‐type and mutants exhibited similar robust DNA binding activities in the nucleotide‐free condition (Fig 3B). Because Pyrococcus furiosus (Pf) Rad50cd interacted with DNA only in the presence of a nucleotide (Hopfner et al, 2000a,b), these results suggest that DNA binding in AMP‐PNP‐free condition is primarily contributed by MjMre11 and therefore is not dependent on the binding activity by Rad50. This interpretation is consistent with the reported structure, in which DNA‐binding surface of Mre11 is open in the nucleotide‐free MRcd complex (Lammens et al, 2011). In the presence of AMP‐PNP and Mg2+, all MjMRcd mutants (R86E, R92E, and T107E) except for MjMRcd (Δ53–55) exhibited significantly reduced binding to the DNA relative to wild‐type (WT) MjMRcd (Fig 3C). We used AMP‐PNP instead of ATPγS, because weak but residual hydrolyzing property of ATPγS might activate the MR nuclease activity, and interfere for an accurate detection of DNA binding to the mutants. The results validate our structure and the proposed DNA‐binding interface as critical for binding of Rad50 to DNA (Hopfner et al, 2001; Lim et al, 2011).
MRcd lacks the essential coiled‐coil and zinc‐hook domains, which could restrict the movement of the Rad50 ATPase head as proposed by Lee et al (2013). It is therefore of a significant interest if the full‐length MR complex forms a similar conformation as that of the MRcd and exhibits the nucleotide dependent and independent DNA binding pattern as shown in Fig 3B and C. We thus examined the DNA binding activity of the full‐length MR complex. Full‐length MjMR cannot be produced in soluble form; therefore, we used full‐length TmMR proteins. Unlike MjMRcd wherein ATP‐free form binds to DNA independently of Rad50 mutations, both a nucleotide‐free and nucleotide‐bound full‐length TmMR cannot efficiently bind to DNA if Rad50 harbors mutations at Arg87, Lys95, and Lys115 (Fig 3D and E). Lys115 of TmRad50 is equivalent to Thr107 of MjRad50 in aligned structures. The K115E TmMR mutant binds to DNA at high protein to DNA ratio. Both ATP‐free and ATP‐bound forms of wild‐type or each mutant TmMR proteins exhibit similar DNA binding activities and patterns. We observed similar DNA binding patterns between MRcd (or full‐length MR) and a linear DNA (pCDFDuet‐1 cleaved by EcoRI, Fig EV3A–D). All of Rad50 mutants used here retained full ATPase activities (Appendix Fig S2). The results raise the possibility that the structure of full‐length MR is different from that of MRcd and the coiled‐coil and zinc hook restrict the NBD movement of the nucleotide‐free MR complex. We propose that the NBDs of nucleotide‐free full‐length MR are arranged in such a way the access of dsDNA to Mre11 is impeded (see Discussion for further details).
To test this idea further, we compared the nuclease activities of free Mre11, ATP‐unbound and ATP‐bound MRcd and the full‐length MR using hairpin DNA as a substrate (Fig 3F and G). We surmised that nucleotide‐free MRcd retains the ability to cleave hairpin DNA to the level identical or similar to free Mre11 because the active site of Mre11 in the MRcd complex is not blocked and open for the cleavage. By contrast, the nucleotide‐free full‐length MR complex exhibits no or weak nuclease activity because Rad50 prevents the access of DNA to Mre11 in nucleotide‐free condition. Indeed, the nucleotide‐free MjMRcd complex exhibits a hairpin‐cleavage activity comparable to that of the Mre11 alone. Most importantly, the nucleotide‐free full‐length MR exhibits virtually null (or very weak) hairpin‐cleavage activity. Addition of ATP significantly increases endonuclease activity of the full‐length TmMR protein. ATP did not stimulate endonuclease activity of the MjMRcd complex (Fig 3F; Lim et al, 2011). We concluded that coiled‐coil and zinc‐hook domain of Rad50 are critical in regulating the nuclease activity of Mre11 under ATP‐dependent manner, consistent with the reported studies (see Discussion, Lee et al, 2013; Hohl et al, 2015; Barfoot et al, 2015).
Conformation of a DNA bound to the ATPγS–MR complex
In the MjMR‐DNA complex, the α1‐α2 and β5‐β6 loops sandwich the 3rd to 6th phosphates of the template strand, which results in a partly deviated strand from its ideal B‐form (Figs 2B and EV4A and B). The width of the minor groove (distance between the 7'th and 12th phosphate groups), which is wedged by loop α1‐α2, increases from 13.2 Å (for B‐form DNA) to 15.3 Å. A major groove at the center flanked by the minor grooves at both sides is constricted to 17.3 Å (distance between the 12th and 17'th phosphate groups; 20.6 Å for B‐form DNA), with a small kink (approximately 7°, Fig EV4). The width between the two strands also increases. Distance between the 11'th and 12th phosphate groups changes to 18.4 Å from 17.3 Å upon binding to Rad50.
DNA binding induces rotation of the coiled‐coil arm of Rad50
ATPγS‐bound MjMR also undergoes conformational changes upon binding to DNA. Superposition of apo‐MjMR (3AV0, magenta in Fig 4A) and ATPγS–MjMR–DNA (green) by aligning an NBD reveals notable differences in the orientation of the coiled‐coils, which are rotated inward by 27° and become more parallel in the DNA‐bound MjMR structure (Fig 4A). This is consistent with the time‐resolved atomic force microscope analysis of the human MRN–DNA complex (Moreno‐Herrero et al, 2005). In a close‐up view, helix α6 slides toward the DNA (by 1.5 Å for 14 Cα atoms), which shifts Lys144 to interact with phosphate backbone (Fig 4B and C). Translation of helix α6 moves Leu155 and Leu156 to pack tightly against the hydrophobic pocket formed between lobe I and lobe II (Fig EV5A and B). Also, a new ion‐pair network (Asp159, Lys163, Arg939) is formed, which interacts with main chain of Gly120 (Fig EV5C and D). These structural transitions would stabilize the rotated coiled‐coils in the DNA‐bound MjMR structure. In addition, loop α1‐α2 is rearranged into a more compact form to fit into the minor groove of DNA (Fig EV5E).
Basis for ATP‐dependent DNA melting by Rad50
Because DNA is positioned at the interface of Rad50 dimer, ATP hydrolysis could rotate each lobe to deform the DNA strand further. Two lobes of Rad50 can be rotated up to 30° upon ATP hydrolysis (Hopfner et al, 2000a,b; Lim et al, 2011; Williams et al, 2011). To understand how ATP hydrolysis affects the DNA conformation, we superimposed ADP‐MjRad50 (3AUX) and the ATPγS–MjMR–DNA complex structures by aligning their lobe II domains (1.1 Å rmsd in 66 Cα positions). The superimposed structures reveal the gross movement (as much as 26 Å) of the α1‐α2 loop, in such a way that this loop could collide with a few DNA residues during the conformational change (Fig 5A, Movie EV3). In addition, rotation of lobe I shifts the positions of the β5–β7 strands as much as 27 Å and opens the gate for Mre11 (Fig 5B).
To determine whether domain rotation of Rad50 melts DNA, we examined DNA unwinding activity of a TmMR complex containing a nuclease‐inactive (H94S) Mre11 variant on 50‐bp dsDNA. The experiment was not performed at the optimal temperature for the enzyme activity (55°C) as significant portions of the control DNA would be melted at this temperature. Thus, we performed the unwinding analysis at 25°C to compare relative dsDNA unwinding activities of various TmMR complexes. Prokaryotic MR complex failed to melt DNA in the absence of ATP or in the presence of ATPγS (Fig 5C). By contrast, the addition of ATP stimulated the melting activity of TmMR. WT TmMR exhibited a limited DNA unwinding activity (~8%), which is similar to the unwinding activity of the human MRN complex (Paull & Gellert, 1999). Unlike the human complex, which can only melt short (17 bp) dsDNA molecule, the prokaryotic MR complex lacking Nbs1 can unwind much longer DNA strands (Paull & Gellert, 1999). The TmMR (Δ54–56) mutant exhibited slightly reduced DNA unwinding activity, whereas the other mutants showed almost no DNA melting activity (Figs 3A and 5D). This finding suggests that ATP‐dependent partial DNA unwinding activity is conserved in both prokaryotic MR and eukaryotic MRN complexes (Paull & Gellert, 1999; Chen et al, 2005; Cannon et al, 2013).
Coupling of DNA melting activity of Rad50 and in vitro endonuclease activity of Mre11
Next, we investigated whether ATPase‐driven DNA melting by the MR complex stimulates endonuclease activity of Mre11. As predicted, WT TmMR exhibited ATP‐dependent nuclease activity toward an 81‐bp DNA strand at both 25 and 55°C (Fig 6A, Appendix Figs S3 and S4). Four TmMR mutants (Rad50 Δ54–56, R87E, R94E/R95E, and R95E, see Fig 3A for corresponding residues in other species) exhibited no or weak endonuclease activities in the absence or presence of ATP, suggesting that the DNA binding activities of Rad50 are important for the resection by Mre11 (Figs 3B–E and 6A). The TmMR complex containing the Rad50 Δ54–56 mutant did not cleave dsDNA, even if it retains limited DNA unwinding activities. Possibly, relatively weak DNA unwinding activities of the TmMR (Δ54–56) complex may not be sufficient for robust endonucleolytic cleavage.
One intriguing possibility is that Rad50‐ and ATP‐dependent DNA melting facilitates the access of Mre11 for DNA cleavage. The model explains as to how cells couple ATP hydrolysis to nuclease activity of Mre11 in full‐length TmMR. In the superimposed ADP‐MjRad50 and ATPγS–MjMR–DNA complex structures, the lobe I of the ADP‐MjRad50 shifts as much as 27 Å, which could allow the access of ssDNA to Mre11 (Fig 5B). To test this idea further, we examined the binding and nuclease activities of WT and mutant TmMR complexes toward a substrate containing a 15‐nt bubble at the middle of the 81‐bp dsDNA. All TmMR proteins interacted with the bubble DNA, although mutants exhibited reduced bubble DNA binding activities compared to that of WT TmMR (Figs 6B and EV3E and F). All TmMR mutants exhibited significantly higher endonuclease activities toward the bubble substrate than toward the intact 81‐bp dsDNA in the absence of ATP (Fig 6A and C, left panel). The mutants generated approximately 23‐nt and 33‐nt fragments due to the cleavage at or near the bubble site, which differed from the cleavage product of the duplex substrate produced by WT TmMR (Fig 6C, left panel). The addition of ATP strongly stimulated WT TmMR to produce 13‐nt, 20‐nt, 23‐nt, and 33‐nt fragments as major products (Fig 6C, right panel). By contrast, ATP did not stimulate the nuclease activities of the TmMR mutants. These findings suggest that unwound DNA but not intact dsDNA could access the Mre11 active site for DNA cleavage, which is stimulated by ATP binding and hydrolysis of Rad50.
For unknown reasons, WT TmMR produced fragments of approximately 13 nt and 20 nt as major products in the presence of ATP. The dsDNA cleavage by the MR complex occurs in a symmetrical manner (Appendix Fig S4E). Based on the ATPγS–MjMR–DNA structure, each wedge interacts with minor grooves, which are ~9–13 nt and 20–24 nt away from one end of the DNA, and this region of DNA is likely to be melted by the ATP hydrolysis‐driven NBD rotation.
Puzzlingly, in six independent experiments using a fresh prepared protein sample, the TmRad50 K115E mutant repeatedly produced a 3‐nt fragment as a major product. We do not know the biochemical basis of such activity but wondered if the altered binding of DNA to the MR complex by the mutation affects the hand‐off process of a substrate, which may have incorrectly guided the phosphodiester bonds at the Mre11 active site. This unusual activity of the K115E mutant may lead to a reduced in vivo hairpin activity (Fig 6E, R131E variant) and increased sensitivity of rad50Δ sgs1Δ cells expressing the mutant toward damage‐inducing drugs (see below, Appendix Fig S5).
DNA binding to Rad50 is important for DSB repair
We then determined whether DNA binding by Rad50 contributes to in vivo functions of the MRX complex in budding yeast. Centromeric plasmids expressing yeast rad50 variants with mutations at or near the DNA‐binding surface (Δ55–57, R96E, K103E/K104E, R131E, see Fig 3A) were introduced into yeast cells deleted for rad50 or rad50 sgs1. Yeasts were then subjected to DNA damage by treating cells with the indicated dose of camptothecin (CPT), methyl methanesulfonate (MMS), and phleomycin (PHL) whose repair depends on binding and resection of DNA ends by MRX complex. Considering severe deficiency of these mutants except Δ55–57 in DNA binding, we are surprised to find that most single rad50 mutants except R96E and R131E showed near wild‐type levels of resistance to genotoxic agent treatment (Figs 3A–E, 6D left panel, and EV3A–D, Appendix Fig S5 top panel). The results may reflect the limited conservation of the DNA‐contacting residues of Rad50 proteins between prokaryotic MR and yeast MRX complexes. Alternatively, the results could be due to the limitations of in vitro DNA binding and endonuclease assay and/or additional factors related to in vivo environment. In this scenario, a single‐site mutation might not be sufficient to disrupt in vivo DNA binding by Rad50 due to clustering of weak binding contacts. We thus combined each mutation to generate double or triple rad50 mutants and assessed their sensitivities from rad50Δ or rad50Δ sgs1Δ cells that express the rad50 mutants upon genotoxic agent treatment. We then found that the rad50 bearing Δ55‐57/K103E/K104E or K103E/K104E/R131E mutation conferred a severe sensitivity to all three drugs in rad50 or both rad50 sgs1 deleted cells (Fig 6D right panel, Appendix Fig S5 bottom panel). Notably, all rad50 variants were expressed at levels almost indistinguishable from that of WT Rad50, with exception of R96E, which showed the significant reduction in rad50 level (Appendix Fig S6). All the mutant proteins except R96E mutant also sustain the integrity of MRX complex, suggesting that multi‐site mutations did not compromise the stability and gross structural properties of the complex (Appendix Fig S7). Thus, unlike the prokaryotic R86E variant that forms a stable Mre11/Rad50 complex, the severe drug sensitivity of rad50‐R96E in yeast likely stems from an inability to produce stable complex, which is known to be essential for all functions associated with MRX complex.
To further investigate the effect of the rad50 mutations on endonuclease activity of MRX in vivo, we measured the rate of recombination between Alu repeats from S. cerevisiae harboring these mutations (Lobachev et al, 2002). To this end, we engineered rad50 mutations into its own genomic locus in yeast strains carrying inverted Alu repeats at lys2 locus and monitored the frequency of recombination by scoring LYS+ revertants. Three rad50 mutants, Δ55–57, R131E, and K103E/K104, exhibited moderately decreased lys2 reversion frequency, whereas rad50 R96E mutant exhibited a near null level of lys2 revertant formation (Fig 6E, Table EV1). When single mutations were combined, the effects became more pronounced such that two multi‐site mutants (Δ55–57/K103E/K104E and R131E/K103E/K104E) exhibited severe recombination defects. By contrast, the K147E/K103E/K104E mutant that is resistant to damage‐inducing drugs exhibited recombination efficiency similar to that of wild‐type RAD50. These results mirror that of the cell survival assay toward CPT, which suggests that the sensitivity toward CPT is tightly correlated with in vivo endonuclease activity of the MRX complex in the yeast strain. Collectively, these results support the premise that DNA binding activity of Rad50 is critical for endonuclease activity of MRX in vivo even though yeast can readily tolerate a single‐site mutation at the interface of Rad50 and the multiple‐site mutations is needed to express its effect.
Damaged DNA recognition by the MRN/X complex is the central feature of the early DSB repair events. The MRN complex can bind to DNA in either the absence or the presence of ATP (Alani et al, 1990; Hopfner et al, 2000a,b; Bhaskara et al, 2007; Lee et al, 2013). However, a number of studies showed that the ATP‐induced NBD dimer formation is important for the functional DNA binding by Rad50. Yet, the underlying biochemistry and molecular biology of ATP‐dependent DNA binding and how it plays critical roles in DNA unwinding, stimulation of the endonuclease activities, end joining, and ATM kinase activation by the MRN (or MRX) complex are still elusive (Alani et al, 1990; Hopfner et al, 2000a,b; Lee et al, 2003; Cannon et al, 2013; Deshpande et al, 2014). Recently, Rojowska et al reported crystal structure of the DNA‐bound MR complex in the presence of AMP‐PNP (2014). In the structure, the DNA molecule binds to only one of the two NBDs of the MR complex despite ATP‐induced NBD dimer formation is essential for DNA binding. Thus, it remains unclear how ATP binding to the MR complex promotes diverse ATP‐dependent functions of the complex. Mutational studies revealed that functionally critical DNA‐contacting residues are located at the central groove, which is distant from the DNA‐binding site in the reported structure, suggesting that additional DNA binding site is likely present in the NBDs of Rad50 dimer (Lim et al, 2011; Rojowska et al, 2014).
In the present study, we showed that an ATPγS‐bound tetrameric MR complex recognizes the duplex DNA via a positively charged surface at the central groove between two Rad50 NBDs that forms only in the presence of ATP or ATPγS. This structural feature explains why ATP is required for the functional DNA binding by Rad50 (Alani et al, 1990; Raymond & Kleckner, 1993; Hopfner et al, 2000a,b). DNA binding by ATPγS‐bound Rad50 NBDs is clearly distinguishable from that by SMC proteins such as cohesin and condensin that interact with DNA via their hinge domain despite the fact that all members of this family share similar ATPase head domain structures (Hirano, 2006). Presumably, ATP‐dependent DNA binding property of the NBD of Rad50 has evolved to perform unique function in tethering, signaling, and resecting of DNA ends.
Previous structural studies using the coiled‐coil deleted prokaryotic MRcd complex showed that the nucleotide‐free MR complex forms an open head conformation, whereas the nucleotide‐bound MR complex forms a closed head structure, where the active site of Mre11 is blocked by Rad50 dimer (Lammens et al, 2011; Lim et al, 2011; Möckel et al, 2012). Indeed, our EMSA and nuclease activity analyses (Fig 3B and C) reconfirm these structural findings. Surprisingly, the nucleotide‐free full‐length MR complex exhibited DNA binding and cleavage patterns different from that of the nucleotide‐free coiled‐coil deleted MRcd (Figs 3D–G and EV3C and D). Although we do not exclude a possibility that the observed DNA binding and cleavage differences between the full‐length and the coiled‐coil deleted MR complexes might reflect different temperature optima for the archaeal versus eubacterial thermophilic enzymes or the limited conservation between these two complexes, we interpret the data that the presence of the coiled‐coils and a zinc hook in full‐length MR might impose restraints on the ATP‐free Rad50 head domains. Recent studies showed that zinc‐hook and head domains interdependently regulate each other via coiled‐coil arms in yeast MRX (Hohl et al, 2015). The coiled‐coil and hook domains are important for a myriad of the globular domain function of MRX including end joining, resection, telomere maintenance, Tel1 activation, and assembly of the MRX complex (Hopfner et al, 2002; Hohl et al, 2011, 2015; Barfoot et al, 2015). We propose that the two NBDs are located closely together in the nucleotide‐free full‐length MR complex, which could prevent the access of DNA to Mre11. Consistent with this model, Nbs1, which binds to the closed state of human MR, but not the open state, interacts efficiently with the ATP‐free intact human MR complex (Lee et al, 2013). Furthermore, cross‐linking analyses of the archaeal full‐length Rad50 head domain showed that two NBDs can be linked by hydrogen peroxide in the ATP‐free state, suggesting that two head domains are arranged in close proximity (Deshpande et al, 2014). ATP binding and hydrolysis by the archaeal full‐length MR and MRcd complexes were shown significantly different (Deshpande et al, 2014). Nelson and co‐workers suggested that an ATP‐free open state is unlikely present in the full‐length phage T4 MR complex (Barfoot et al, 2015).
Symmetrical binding of DNA to the central groove of an ATP‐bound Rad50 dimer provides several insights into understanding the functions of the MR complex. First, positioning of the DNA molecule on top of the two ATP‐bound NBDs could allow Rad50 to melt the DNA at least partially via ATP hydrolysis‐driven lobe rotation (Figs 5A and C, and 7, Paull & Gellert, 1999). Upon ATP hydrolysis, the α1‐α2 loop of Rad50 undergoes positional alteration, and during this process, the α1‐α2 loop from each Rad50 could collide with phosphate backbone residues at minor groove and disrupt its structure, which may explain the ATP‐dependent DNA melting activity by the MR complex (Fig 5A, Movie EV3). Second, the rotated lobes of Rad50 can generate a space, through which ssDNA but not dsDNA passes to access Mre11 in the MR complex (Figs 5B and 7). This model might explain the similar dsDNA‐binding patterns by the nucleotide‐free and nucleotide‐bound intact MR complexes, and impaired endonuclease activities toward dsDNA by the nucleotide‐free intact MR complex (Fig 3B–G). Consistently, the bubble DNA, but not intact dsDNA, was cleaved by the ATP‐free MR mutants, and the bubble DNA bound more efficiently to the ATP‐free MR complex than intact dsDNA (Fig 6A–C). The model we propose may provide a basis for the observation that the presence of ATP and Rad50 stimulates the endonuclease activity of Mre11, although Mre11 dimer alone can cleave hairpin DNA (Fig 3F and G; Williams et al, 2008; Sung et al, 2014).
Previous TmMR‐DNA and current MjMR‐DNA structures demonstrated that the MR complex recognizes DNA region distant from the end. It thus remains unclear how the MR complex recognizes DNA end and catalyzes ATM‐dependent signaling. In the end blocked DNA, the MRX complex is preferably bound to a site which is ~15–25 nt away from the 5′ blocked end and makes an endonucleolytic incision (Cannavo & Cejka, 2014; Shibata et al, 2014). Symmetrical DNA binding by the M2R2 tetramer requires two minor grooves, in which each minor groove is recognized by three motifs of each Rad50 molecule (Fig 2A, B and E). Such structure‐specific recognition by Rad50 may explain DNA binding at the internal site by the MR complex. We do not rule out the possibility that the MR complex binds to the internal region and slides along the DNA to sense the minor groove near the terminal end. The MR complex might recognize DNA ends by inducing Rad50‐dependent DNA deformation from the ideal B‐form DNA conformation. DNA deformation then could trigger ATM activation signaling and assembly of end‐joining complex at the DNA ends (Lee & Paull, 2005; Lee et al, 2013). Given that the conformations of the ATP‐free and ATP‐bound MR complexes play important roles in determining the DSB repair pathway (Deshpande et al, 2014), the conformational change of MR and the associated DNA could dictate the repair pathway choice. Further studies are required to understand whether the DNA conformation would play a crucial role in sensing, end joining, and resection by the MR complex.
Materials and Methods
Protein expression and purification for the MjMRcd complex
The MjMR complex was produced by co‐expressing the two proteins from Escherichia coli Rosetta (DE3), as described previously (Lim et al, 2011). Briefly, cDNAs encoding residues 1–190 and residues 825–1,005 of MjRad50 were inserted into pCDFDuet‐1 and pETDuet‐1, respectively, and MjMre11 (residues 1–366) was inserted into the pET28a vector. The complex was first purified by Ni‐NTA affinity chromatography. The MjMR complex was subsequently eluted with 300 mM imidazole in the same buffer. Fractions containing the MjMR complex were purified using cation exchange (Resource S) and gel‐filtration chromatography (Superdex 200), and concentrated by ultrafiltration. All MjMR mutant proteins were purified by the same procedure.
For the full‐length TmMR complex, the cDNAs encoding full‐length TmMre11 and TmRad50 were inserted into pET28a and pCDFDuet‐1, respectively, and co‐expressed in Escherichia coli Rosetta (DE3). The full‐length WT or mutant TmMR complex was purified by Ni‐NTA affinity chromatography, anion exchange (Hi‐trap Q), and gel‐filtration chromatography (Superdex 200).
Crystallization and data collection
Crystals of the ATPγS–MjMR complex bound to DNA were grown at 18°C by the hanging‐drop vapor diffusion method. The crystallization buffer contained 24% PEG 400, 0.1 M Tris–HCl (pH 8.0), 1 mM ATPγS, 0.1 mM MgCl2, and 0.1 M trimethylamine hydrochloride. A 25‐bp duplex DNA with a 2‐nt 5′ overhang at each end was used for crystallization. The ATPγS–MjMR‐DNA crystals formed in the space group P212121 (a = 84.3 Å, b = 130.1 Å, and c = 166.6 Å) and contained two ATPγS–MjMR complexes and one DNA molecule in the asymmetric unit. Diffraction data were collected at −170°C using crystals flash‐frozen in crystallization buffer containing 30% (v/v) glycerol. Diffraction data were collected at 0.9795 Å on Beamline 5C at the Pohang Accelerator Laboratory. Diffraction data were integrated and scaled using the HKL2000 package (Otwinowski & Minor, 1997).
To improve the usable resolution and quality of the resulting electron density maps, we used the Karplus CC* (Pearson correlation coefficient)‐based data cutoff approach (Table 1; Karplus & Diederichs, 2012, 2015). We took into account following criteria to improve the resolution limit: (i) CC1/2 > ~40% based on data merging statistics, (ii) a Karplus CC* analysis against unmerged intensities in the Phenix package (higher CC* relative to CCwork and CCfree) to avoid overfitting, and (iii) Rfree of the highest resolution shell against the refined structure being < ~40% (Appendix Fig S1A).
Structure determination and refinement
The structure of the ATPγS–MjMR–DNA complex was determined by the molecular replacement method. The MjRad50 molecules were located with the PHASER program, followed by a search for MjMre11 (McCoy et al, 2007; Lim et al, 2011; Sung et al, 2011). After density modification, an electron density map was generated at a resolution of 3.1 Å using the PHENIX program (Adams et al, 2010); the map was of good quality and revealed most DNA chains. Successive rounds of model building using COOT (Emsley & Cowtan, 2004) and refinement using PHENIX (using rigid body, xyz coordinates, individual B‐factor, NCS, and TLS refinements) were performed to build the complete model. The final refined model of the ATPγS–MjMR–DNA complex at 3.1 Å (Rwork/Rfree = 22.2/26.9%) contains 92.9% of residues in the most favored region and no residue in the disallowed region (Table 1).
DNA binding assays
For the DNA binding analysis, MjMRcd or TmMR proteins were incubated with a closed circular dsDNA molecule (φX174 RFII; NEB), a linear dsDNA (EcoRI‐treated pCDFDuet‐1), or a bubble DNA in 20 μl reactions containing reaction buffer (25 mM Tris–HCl, pH 7.4, 50 mM NaCl, 5 mM β‐mercaptoethanol, and 5% glycerol) in the absence or presence of 1 mM AMP‐PNP and 5 mM MgCl2. For DNA binding analysis, each protein sample was incubated with φX174 RFII or a linear dsDNA (3.5 nM; the molar ratio of protein: DNA was 50:1, 100:1, 200:1, 500:1, 1,000:1, or 2,000:1) for 30 min at 4°C. After mixing with 6× loading dye [0.4% orange G, 0.03% bromophenol blue, 0.03% xylene cyanol FF, 15% Ficoll 400, 10 mM Tris–HCl (pH 7.5)], the samples were resolved by electrophoresis (100 V for 20 min) through a 0.5% agarose gel in 0.5× TB buffer and visualized by ethidium bromide staining. For the bubble DNA binding analysis, a 15‐nt bubble DNA (2.5 nM) was incubated with various amount of MjMRcd or TmMR proteins at 4°C for 30 min. The molar ratio of protein: DNA was 100:1, 200:1, 500:1, 1,000:1, or 2,000:1. The reaction mixture was resolved by 6% native polyacrylamide gel electrophoresis.
DNA unwinding assays
DNA unwinding reaction mixtures (20 μl) contained 2.5 nM of 32P‐labeled DNA substrate and TmMR mutant proteins (5 or 12.5 nM) in reaction buffer (25 mM HEPES, pH 7.4, 25 mM NaCl, 5 mM β‐mercaptoethanol, 4% glycerol, 50 nM MgCl2, and 0.1 mM ATP). The reactions were incubated at 25°C for 15 min and then stopped by the addition of EDTA and SDS to final concentrations of 5 mM and 0.3%, respectively. The helicase reaction products were analyzed by polyacrylamide gel electrophoresis using non‐denaturing 15% polyacrylamide gels and were separated for approximately 90 min at 100 V. The intensity of the unwound substrate bands was analyzed by ImageQuant TL (Amersham Biosciences).
The reaction mixture (20 μl) contained 10 nM DNA substrate and 100 or 200 nM TmMR (or MjMRcd) in reaction buffer (25 mM HEPES, pH 7.4, 5 mM β‐mercaptoethanol, 50 mM NaCl, and 5% glycerol) containing 5 mM MnCl2, 1 mM MgCl2, and 1 mM ATP (or no ATP). The reaction mixture was incubated for 30 min at 55°C (or for 15 min at 25°C) and stopped by adding the same volume of 2× reaction stop buffer (95% formamide, 18 mM EDTA, 0.025% SDS, and 0.01% bromophenol blue), followed by 5 min of boiling at 100°C. The products were resolved on 15% denaturing polyacrylamide gels containing 7 M urea in 1× Tris–borate–EDTA for 120 min at 13 Vcm−1. The intensities of the uncleaved substrate bands were quantitated using a phosphorimager. All assays were repeated at least three times.
ATP hydrolysis assay
ATP hydrolysis activity was quantified by the BIOMOL GREEN assay (BIOMOL Research Labs, Inc.) using 1 mM ATP and various concentrations of MR proteins (200, 400, 800 nM). The mixtures were incubated at 37 or 55°C (MjMRcd or TmMR, respectively) for 20 min in buffer containing 25 mM Tris–HCl (pH 7.5), 50 mM NaCl, 5 mM MgCl2, and 1 mM DTT and were subsequently incubated on ice. A suitable amount of BIOMOL GREEN reagent was added to the mixtures, which were incubated for 20 min at 25°C and then measured at OD620. The amounts of released free phosphate in the mixtures were determined by subtracting the background phosphate.
All of the mutants used in this study were constructed by PCR‐based methods (Stratagene). For the MjRad50 mutants, residues 53–55 were deleted, or Arg86, Arg92, or Thr107 was mutated to glutamic acid. For the TmRad50 mutants, residues 54–56 were deleted, or Arg87, Lys95, or Lys115 was mutated to glutamic acid; Arg94 and Lys95 were simultaneously mutated to glutamic acid. For the TmMre11 nuclease‐inactive mutant, His94 was mutated to serine. For the ScRad50 mutants, residues 55–57 were deleted; Arg96, Lys103, Lys104, or Arg131 was mutated to glutamic acid; or Lys103 and Lys104 were simultaneously mutated to glutamic acid.
Recombination rate determination
This assay follows typical fluctuation test described in Lobachev et al (2002) with minor modification. Briefly, 500–1,000 yeast cells were inoculated in YEPD and cultured for 3 days. Cells were diluted and plated onto Lysine dropout plate, to monitor the LYS+ recombination event, and YEPD plate to monitor the input cell number, respectively. After 3 days culture in 30°C incubator, colonies were counted. At least 12 independent colonies from two isolates were scored to calculate recombination rate and 95% confidence intervals as described (Spell & Jinks‐Robertson, 2004).
Drug sensitivity assay
Yeast cells expressing WT Rad50, Rad50 variants with different mutations, or rad50 gene deleted were cultured in YEPD media for 32–72 h until they reached a density of approximately 4 × 107 cells/ml. Subsequently, approximately 2 × 106 cells were transferred into 200 μl of sterilized distilled water and subjected to 1:5 serial dilutions. Cells at different densities were spotted onto YEPD agar plates containing the indicated concentrations of camptothecin, phleomycin, or methyl methanesulfonate and incubated at 30°C for 48–72 h.
Yeast genetic analyses
The yeast strains used in this study are derivatives of SLY1A (hoΔ hml::ADE1 MATα hml::ADE1 ade1 leu2‐3,112 lys5 trp1::hisG ura3‐52 ade3::GAL10::HO) or ALE94. All single gene deletions or replacements were performed using a PCR‐derived selection gene product flanked by short terminal sequences homologous to the ends of each gene's open reading frame, or to the sequence flanking the region to be replaced (Li et al, 2013). Saccharomyces cerevisiae strain ALE94 (MATα, ade5‐1, his7‐2, leu2‐3,112:: p305L3 (LEU2), trp1‐289, ura3‐Δ, lys2::AluIR) was utilized to monitor the recombination event stimulated by inverted Alu repeats. rad50 mutations were introduced into genomic locus and confirmed by sequencing. More than two independent isolates with each mutation were utilized for subsequent genetic analysis.
Western blot analysis
Trichloroacetate (TCA)‐precipitated S. cerevisiae lysates were prepared as described before (Yaffe & Schatz, 1984) with modification: S. cerevisiae were grown to A600 of 1.0 in plasmid‐maintaining liquid media at 30°C. Cells were harvested and lysed with 1 ml 0.2 M NaOH, 1% β‐mercaptoethanol. Total protein was precipitated by additional 10% TCA. Protein pellet was re‐suspended in 50 μl SDS–PAGE sample buffer. For analysis, 20 μl of each sample was separated by 8% SDS–polyacrylamide gel and transferred onto nitrocellulose membrane, followed by immunoblotting with anti‐rad50 and anti‐tubulin antibodies. Antibodies against S. cerevisiae Rad50 were a kind gift of Dr. John Petrini.
Yeast two‐hybrid assay
Yeast two‐hybrid system analysis was performed with a GAL4 DNA‐binding domain (BD)‐fused Mre11 and a GAL4 activation domain (AD)‐fused Rad50. The bait (BD‐Mre11) and the prey (AD‐Rad50) proteins were co‐expressed in the yeast strain PBN204 (MATa, pGAL1‐lacZ, pGAL1‐URA3, pGAL2‐ADE2, trp1‐901, leu2‐3, gal4Δ, gal80Δ). PBN204 contains URA3, ADE2, and beta‐galactosidase as reporter genes and trp1 and leu2 as selection marker genes (Panbionet Inc.). Co‐transformation of BD‐polypyrimidine tract‐binding protein (PTB) and AD‐PTB served as a positive control for the protein–protein interaction. PTB is a homo‐dimeric protein. Plasmids pGBKT7 and pGADT7 were used as a negative control.
YL carried out crystallization and structure determination with the help of YK and GHG; YL, FL, SS, AJ, TK, and GHG participated in biochemical experimental design and data analysis; FL and S‐EL designed and performed yeast genetics experiment; A‐KK and O‐KS performed yeast two‐hybrid experiment; YC and S‐EL conceived of the project and wrote the paper.
Conflict of interest
The authors declare that they have no conflict of interest.
Expanded View Figures PDF
Source Data for Expanded View
We thank K. Lobachev, R. Deshpanda, T. Paull, Y.C. Kim, K.T. Kim, and P.A. Karplus for reagents and helpful discussions. We thank G.O. Ahn, for proofreading. We also thank J. Petrini for providing us antibody against yeast Rad50. This work was supported by grants from the National Research Foundation of Korea (NRF) funded by the Korea Government (MEST, No. 2015R1A2A1A05001694, No. 2012‐054226, and No. 20120008833), a rising Star Program (POSTECH) and BK21 Program (Ministry of Education) to Y.C. and from NIH (GM071011) to S.E.L. Coordinates and structure factors have been deposited to RCSB (5DNY and 5F3W).
FundingNational Research Foundation of Korea (NRF)
- © 2015 The Authors