ATP Hydrolysis by RAD50 Protein Switches MRE11 Enzyme from Endonuclease to Exonuclease*

Background: RAD50-MRE11-Nbs1 complex is essential for DNA repair. Results: ATP binding by RAD50 closes the complex; MRE11 is an endonuclease. ATP hydrolysis opens the complex; MRE11 is an exonuclease. Conclusion: ATP hydrolysis is a switch converting MRE11 from an endonuclease to an exonuclease. Significance: ATP-dependent nuclease switch provides a mechanism of how RAD50-MRE11 complex is able to coordinate DNA repair. MRE11-RAD50 is a key early response protein for processing DNA ends of broken chromosomes for repair, yet how RAD50 nucleotide dynamics regulate MRE11 nuclease activity is poorly understood. We report here that ATP binding and ATP hydrolysis cause a striking butterfly-like opening and closing of the RAD50 subunits, and each structural state has a dramatic functional effect on MRE11. RAD50-MRE11 has an extended conformation in solution when MRE11 is an active nuclease. However, ATP binding to RAD50 induces a closed conformation, and in this state MRE11 is an endonuclease. ATP hydrolysis opens the RAD50-MRE11 complex, and MRE11 maintains exonuclease activity. Thus, ATP hydrolysis is a molecular switch that converts MRE11 from an endonuclease to an exonuclease. We propose a testable model in which the open-closed transitions are used by RAD50-MRE11 to discriminate among DNA ends and drive the choice of recombination pathways.

MRE11-RAD50 is a key early response protein for processing DNA ends of broken chromosomes for repair, yet how RAD50 nucleotide dynamics regulate MRE11 nuclease activity is poorly understood. We report here that ATP binding and ATP hydrolysis cause a striking butterfly-like opening and closing of the RAD50 subunits, and each structural state has a dramatic functional effect on MRE11. RAD50-MRE11 has an extended conformation in solution when MRE11 is an active nuclease. However, ATP binding to RAD50 induces a closed conformation, and in this state MRE11 is an endonuclease. ATP hydrolysis opens the RAD50-MRE11 complex, and MRE11 maintains exonuclease activity. Thus, ATP hydrolysis is a molecular switch that converts MRE11 from an endonuclease to an exonuclease. We propose a testable model in which the open-closed transitions are used by RAD50-MRE11 to discriminate among DNA ends and drive the choice of recombination pathways.
Despite its importance in recombination, the multifunctional roles of the RAD50-MRE11-Nbs1 complex in carrying out repair are poorly understood (1-3, 5, 6). During non-homologous end joining, binding of RAD50-MRE11-Nbs1 to chromosome ends prevents their degradation by processive nucleases (7)(8)(9)(10)(11), whereas in the HR pathway RAD50-MRE11-Nbs1 promotes DNA resection of the broken DNA ends (12)(13)(14). RAD50-MRE11 is also a component of the broad DNA replication machinery, where it plays key roles in replication fork stability and replication restart after lesion-induced fork blockage (14 -18). Additionally, by interacting with either ATM or ATR kinase, the RAD50-MRE11-Nbs1 complex can activate the cell-cycle checkpoint in response to DSB or fork stalling, respectively (19 -24). Thus, the RA50-MRE11-Nbs1 complex has been implicated in every aspect of DSB processing from the initial detection to the triggering signaling pathways and to the choice among pathways for repair. How this single protein complex is able to orchestrate these diverse and complex functions is unknown.
At overhangs, MRE11 binds to ssDNA-dsDNA junctions in the same groove (14). The shape and extensive contacts with the DNA backbone enable two subunits of MRE11 to simultaneously contact both dsDNA and ssDNA substrates (14,28). Thus, there is a physical basis for how MRE11 might act as a multifunctional DNA end-processing enzyme. Yet, how the MRE11 nuclease discriminates among DNA ends is not clear. In vivo, during DSB repair, MRE11 is needed to process DNA ends, but in vitro DNA recognition by RAD50-MRE11-Nbs1 has no clear biochemical preference for DNA ends, hairpin loops, or internal sites in DNA (27)(28)(29)34). Furthermore, the mechanisms by which RAD50 regulates MRE11 activity is poorly understood. RAD50 belongs to the structural maintenance of chromosomes family of proteins (35)(36)(37)(38)(39) and functions as an ABC-ATPase (40) in which RAD50 forms a composite ATP binding site comprising the N-terminal Walker A and C-terminal Walker B motifs (26,37,38,41,42) (Fig. 1A). The N-and C-terminal subdomains come together by intramolecular collapse of the coiled-coil domains, and ATPinduced dimerization results in sandwiching of ATP between the Walker A and B site of one monomer with the signature motif of the other monomer (37,38).
In both P. furiosus and Thermotoga maritime, binding of the non-hydrolysable ATP analogues (AMP-PNP) to RAD50 induces a dramatic conformational rotation of the C-terminal ATPase subdomains relative to the N-terminal half in the RAD50 subunits (37,38,41,42). AMP-PNP binding positions the RAD50 arms together and poises the MRE11 nuclease for wider access to DNA (14,26,37,38,41,42), yet neither the effects of the conformational change nor the ATP dependence of the MRE11 nucleolytic functions is clear. Mutation in the ATP binding domains of RAD50 impairs HR, implying that ATP is essential to carry out MRE11 nucleolytic processing and repair (43,44). However, the ssDNA exonuclease activity and the ssDNA endonuclease activity of MRE11 do not require ATP (27,28), whereas the endonucleolytic cleavage of blocked DNA, dsDNA exonuclease, and 5Ј dsDNA endonuclease do require ATP (27)(28)(29). The role of RAD50 ATP hydrolysis in regulating MRE11 activity remains enigmatic.
RAD50-MRE11 is a key and essential DNA processing enzyme for DSB repair, and regulation is nucleotide-driven. Despite elegant structural and molecular analyses, there is not yet a sufficiently clear picture of how nucleotide dynamics in the RAD50 subunits regulate the MRE11 nuclease activities. To address this issue, we have combined biochemical analysis, fluorescence energy transfer, and small angle x-ray scattering to evaluate nucleotide regulation and the dynamic interplay between RAD50 and MRE11. We find that ATP binding and ATP hydrolysis drive dramatic and opposing structural rearrangement in the RAD50-MRE11 complex and specify the nucleolytic activity of MRE11. ATP binding to RAD50 shifts the equilibrium to the closed conformation, conditions under which MRE11 is an endonuclease, whereas ATP hydrolysis opens the complex, conditions under which MRE11 is an exonuclease. ATP hydrolysis is a molecular switch that converts MRE11 from an endonuclease to an exonuclease.

EXPERIMENTAL PROCEDURES
Oligonucleotides-Oligonucleotides were purchased from Integrated DNA Technologies, Coralville, Iowa. All oligonu-cleotides were gel-purified. In anisotropy DNA binding experiments, fluorescein-labeled (6-carboxyfluorescein) DNA was used. The sequences of the oligonucleotides are presented in supplemental material.
Proteins-P. furiosus proteins and protein complexes were overexpressed and purified following a previously published procedure (25,38).
Nucleotide and DNA Binding Assay-All binding experiments were performed at room temperature in buffer A (50 mM Tris, pH 7.5, 100 mM NaCl, 0.1% PEG-6000, 2.5% glycerol, and the indicated concentration of MgCl 2 and MnCl 2 ). Reactions were incubated for 10 min before measurement. Anisotropy was measured using an Infinite M1000 plate reader (Tecan Group Ltd.) at 25°C. In nucleotide competition binding experiments, 2 nM BODIPY-labeled nucleotide was incubated with 2 M protein (4 M binding sites) and the indicated concentrations of "cold" nucleotide. The K i values were calculated according to the mass-action binding model using K D for BODIPY-labeled nucleotide obtained from a direct binding assay. All anisotropy measurements were performed in triplicate.
Cross-linking of Nucleotide and DNA Binding Assay-Experiments were performed at a concentration of 100 nM protein and 500 nM DNA. UV cross-linking of labeled nucleotides to pfRAD50 was performed in a Stratalinker 1800 (Stratagene) for 7.5 min. All binding experiments were performed at 25°C in buffer A and the indicated concentrations of MgCl 2 and MnCl 2 . Cross-linked products were resolved by SDS-PAGE, and the signals were quantified with phosphorimaging.
Fluorescence Resonance Energy Transfer (FRET)-Nucleotide binding experiments were performed at 25°C in buffer A and the indicated concentrations of MgCl 2 and MnCl 2 . Twenty-microliter reactions contained 1 M FRET donor (BODIPY-FITC) nucleotide, 1 M FRET acceptor (BODIPY-TR) (Texas Red (TR)) nucleotide, 1.5 M protein (3 M binding sites), and when applicable, 3 M DNA. After all components were added, the reactions were incubated for 10 min before measurement. Fluorescence intensities of the FRET donor (excitation/emission 470/522 nm), FRET acceptor (excitation/emission 590/ 615), and FRET signal (excitation/emission 470/615) were measured using an Infinite M1000 plate reader (Tecan Group Ltd.). FRET efficiency was calculated as a ratio of the FRET signal divided by the fluorescence intensity of the donor. Data were normalized with FRET efficiency values obtained in reactions containing no protein.
Small Angle X-ray Scattering (SAXS)-SAXS data were collected at the ALS beamline 12.3.1 (Lawrence Berkeley National Laboratory, Berkeley, CA). The scattering vector is defined as q ϭ 4sin /, where 2 is the scattering angle. All experiments were performed at 20°C, and data were processed as described (45,46). The data acquired at short and long time exposures (1 and 10 s) were merged for calculations using the entire scattering spectrum. The experimental SAXS data for different protein concentrations were measured for aggregation using Guinier plots (46). Aggregated samples were excluded from analysis. The program GNOM (46) was used to compute the pair distance distribution functions, P(r). This approach also provided the maximum dimension of the macromolecule, Dmax.
Sedimentation Velocity-Samples were dialyzed in 25 mM HEPES, pH 8.1, 100 mM NaCl, 5 mM MgCl 2 , and 1 mM MnCl 2 . All sedimentation velocity experiments were performed with a Beckman Optima XL-A (Brea, CA) at 20°C. Samples were centrifuged at a rotor speed of 40,000 rpm using a two-channel centerpiece in an An-55 rotor using dialysis buffer as a reference. The absorbance of the samples was measured at 280 nm for L4-MRE11 Ϯ ATP. The 40-base pair double-stranded DNA was tagged with fluorescein at the 5Ј end. Thus, the absorbance of the DNA-containing samples was detected at 497 nm for L4-MRE11, DNA Ϯ ATP. The sedimentation velocity data were analyzed using the enhanced van Holde-Weischet module of Ultrascan 9.9 with hydrodynamic corrections for buffers used (48).
ATPase Assay-Forty-microliter assays were performed at 55°C in buffer A containing the indicated concentrations of Exonuclease Assays-Steady state 2-aminopurine (2-AP) fluorescence 3Ј-5Ј exonuclease assays were performed at 55°C in buffer A containing the indicated concentrations of MgCl 2 and MnCl 2 . Sixty-microliter assays contained 2 M protein (4 M active sites), 0.2 mM nucleotide, and the indicated concentrations of DNA substrate (1, 2.5, 5, 10, 15, and 25 M). After 2, 4, 8, 12, and 20 min, 10-l aliquots were removed from the reaction and quenched with 15 l of 30 mM EDTA. Release of 2-AP from the duplex DNA was measured as an increase in fluorescence at 375 nm (excitation at 310 nm) in a Tecan Infinite M1000 plate reader. The rate of hydrolysis was calculated using free 2-AP as a standard and fitted to the Michaelis-Menten equation.
Endonuclease Assays-The endonuclease assay using oligonucleotide templates contained 2 M MRE11 or L4-MRE11 protein, 0.2 mM nucleotide, 5 mM concentrations of both MgCl 2 and MnCl 2 , and ϳ1 nM DNA substrate (10,000 cpm). Reactions were incubated at 55°C for 2 h, stopped with gel loading buffer, and resolved by 12% urea-PAGE followed by phosphorimaging. For intact plasmid DNA templates, the assays contained 2 g of ss plasmid, 1 M full-length (FL) or L4-MRE11 protein, and 1 mM ATP incubated with 5 mM concentrations of the indicated bivalent cation. Both fulllength RAD50-MRE11 and L4-MRE11 were used as indicated. Samples were taken at 0, 30, and 60 min. Disappearance of bands indicated incision.

Nucleotide Binding Affinity of RAD50 Is Not Affected by Its
Association with MRE11-To probe the relationship between the ATP (Mg 2ϩ ) binding, ATP hydrolysis, and MRE11 nuclease activity, we purified three forms of RAD50 from P. furiosus: the full-length FL pfRAD50 (Fig. 1, A, left, and B) and two truncated forms of pfRAD50 ( Fig. 1, A, right, B, and C). For the larger truncated forms, pfRAD50-L4 (L4) contains 25 residues of the coiled-coil domains (Fig. 1C, right), whereas in pfRAD50-L1 (L1) (Fig. 1C, left) only 15 residues of each coiled-coil domain are preserved (14,37,42). Both ends of coiled-coil helices in pfRAD50-L1 and pfRAD50-L4 are connected by a GGSGG linkers and associate with MRE11 as dimers (Fig. 1B). Both full-length and truncated forms of pfRAD50 form a tetrameric complex with MRE11 comprising two subunits of the MRE11 nuclease and two subunits of RAD50 (Fig. 1D). For our experiments, the FL protein was purified in a complex with full-length MRE11 protein. L1 and L4 were either purified alone or in a complex with a C-terminal-truncated form of the MRE11 nuclease containing the amino acids 1-381 (MRE11 1-381 ), which retains its nuclease activities (37). The MRE11 1-381 is insoluble in the absence of RAD50 protein, and in experiments with MRE11 nuclease alone, we used a slightly smaller, more soluble form of the protein, MRE11 1-342 (Fig. 1B, lane 1). MRE11 1-342 retains its nuclease activity; however, it is unable to bind RAD50. The purified proteins migrated on an SDS-PAGE gel according to their expected molecular weights, with no significant degradation products (Fig. 1B). In subsequent text, pfRAD50, pfRAD50-L1, pfRAD50-L4, and pfMRE11-RAD50-FL will be referred to as RAD50, L1, L4, and FL, respectively. MRE11 1-381 , will be referred to as MRE11.
We tested whether association with MRE11 affected nucleotide affinity for wild type and mutant forms of RAD50. Nucleotide affinity was measured by fluorescence anisotropy using either labeled ATP-BODIPY-fluorescein (ATP-FITC), or ADP-BODIPY-Texas Red (ADP-TR), or ADP-FITC and ADP-TR ( Fig. 1, E and F). P. furiosus is a widely used extremophile that optimally grows at 55°C, and RAD50-MRE11 in this organism is inactive at 25°C (25). Thus, in these experiments binding was performed at 25°C to prevent ATP hydrolysis in the RAD50 subunits during the measurements. ATP and ADP binding to RAD50 required Mg 2ϩ or Mn 2ϩ as structural co-factors (Table  1) (14,27,28), and in their absence nucleotide binding was not measurable (Table 1). However, nucleotide affinity for FL, L1, or L4 or their complexes with MRE11 was not significantly different whether Mg 2ϩ or Mn 2ϩ was added independently or in a stoichiometric mixture (Table 1). Nucleotide affinity was also unaffected when L4 or L4-MRE11 was bound to either ssDNA or dsDNA (Fig. 1E, supplemental Table SI). Under all conditions, affinity was modest (around 1 M) ( Table 1).
As a second approach, we measured the affinity of nucleotide binding for FL, L1, L4, or their complexes with MRE11 using UV cross-linking. However, ATP and ADP dissociated rapidly from RAD50 ( To ensure that the fluorescence analogues did not reduce affinity or promote rapid nucleotide dynamics, we measured the K i for competition of ATP-FITC and ATP-TR with other nucleotides. We found that K D for binding of ATP-FITC (Mg 2ϩ ) and the K i for its inhibition by unlabeled ATP (Mg 2ϩ ), non-hydrolysable ATP analogues ATP␥S (Mg 2ϩ ), or the transition state analog ADP-BeF n (Mg 2ϩ ) were similar (supplemental Fig. S2, Table 1). Thus, the nucleotide binding to L4 and L4-MRE11 in solution was reversible (Table 1). Collectively, these findings revealed that nucleotide affinity for FL, L1, or L4 was remarkably unaffected by either the structure or function of the RAD50 subunits. Nucleotide affinity for RAD50 was the same whether or not it was associated with MRE11 whether the RAD50 ATPase (Mg 2ϩ ) or whether the MRE11 nuclease (Mn 2ϩ ) was active, or RAD50 was bound to DNA (Fig. 1E, supplemental Table SI). Thus, RAD50 did not significantly contribute to the regulation of MRE11 at the level of nucleotide or DNA affinity.
RAD50 Is Double-occupied with ATP Both On and Off DNA-As judged by the crystal structure, binding of two AMP-PNP molecules induces dimerization of the head domains and a dramatic structural change that closes the RAD50-MRE11 complex (38,42). However, in solution we found that AMP-PNP was a poor competitor for ATP-FITC bound in the L4 subunits (supplemental Fig. S2). Thus, we tested whether nucleotide occupation altered the structures of L1, L4, or their complexes with MRE11 when reversible ATP analogues were added in solution, and whether the alterations depended on DNA binding.
In the first experiments we tested whether ATP occupied one or two of the ATP sites in L4. RAD50 was incubated with an equal concentration of ATP-FITC (as a donor) and ATP-TR (as an acceptor), and we tested whether nucleotide addition resulted in FRET at 25°C, conditions under which ATP hydrolysis in L4 is inactive (Fig. 2, A and B). If two composite Walker sites were occupied with ATP in solution, then we anticipated a FRET signal. Indeed, a strong FRET signal was observed for L4 and its complexes with MRE11 when ATP-FITC and ATP-TR were added together (Fig. 2B). The FRET signal increased as the ATP binding sites were filled and diminished as the excess pro-tein competed for the donor and acceptor nucleotides ( Fig. 2A). However, in all cases, the maximum FRET signal occurred at a 1:1 ratio of nucleotide per RAD50 binding site, indicating that both sites were occupied equally (starred bars, in Fig. 2, A  and B).
In our nucleotide binding experiments (Fig. 1, E and F), the best fit of experimental data to the Langmuir isotherm did not deviate from a theoretical curve, implying that ATP binding to the RAD50 subunits was not cooperative. Thus, in solution the RAD50 dimer bound two ATP nucleotides in close proximity in the presence or absence of MRE11 (Fig. 2, C and D), and nucleotide binding to one subunit did not influence nucleotide binding to the other subunit. The strong FRET signal induced by ATP binding to L4 and L4-MRE11 did not depend on the presence of DNA, indicating that both complexes existed in solution primarily in a double-occupied state (Fig. 2, E-G).
Binding of Two ATPs Promotes "Open" and "Closed" Conformations of RAD50 and RAD50-MRE11 Complexes-The FRET signal observed after binding of ATP (Mg 2ϩ ) to L4 and L4-MRE11 provided evidence that the RAD50 subunits of L4-MRE11 moved closer together. Indeed, we found that ATPbound L4-MRE11 formed a closed structure which exceeded that expected of simple dimerization (Fig. 2D). Both L1 and L4 ( Fig. 1) dimerize (37,38), yet we were unable to observe a FRET signal between ATP-FITC and ATP-TR when bound to L1, and the FRET signal for L1-MRE11 was weak (Fig. 2B, supplemental  Fig. S3). The lack of a FRET signal for the nucleotide-bound L1 complex was not due to differences in nucleotide affinity (Table  1). Nucleotides bound equally well to L1 and L4 or their complexes with MRE11, and the K i for competition with unlabeled ATP was similar to the K D for ATP-FITC and ATP-TR binding to L1 (Table 1). The observations implied that in solution as well as in the crystal structure (41,42), binding of ATP to L4 and L4-MRE11 induced a closed complex that required the coiledcoil domain of RAD50.
In solution, however, ATP (Mg 2ϩ ) binding to RAD50 was necessary but not sufficient for closure. In the absence of ATP  ATP Hydrolysis Opens Closed Complex-We tested the consequences of ATP hydrolysis on the structural conformation of the L4-MRE11. At 25°C, the FRET signal was consistently lost when ATP-FITC and ADP-TR (supplemental Fig. S3) or ADP-FITC and ADP-TR (Fig. 2F) were used as FRET pairs, suggesting that ATP hydrolysis might drive the opening of ATP-bound RAD50-MRE11. To test this hypothesis, we measured the hydrolytic activity of RAD50-MRE11 at 55°C (the temperature optimum for the catalytic activity of pfRAD50) side by side with the FRET intensity of the complex. The rate of ATP hydrolysis was derived from the level of conversion of ATP to ADP after their resolution using thin layer chromatography (TLC) (Fig. 3,  A and D). As expected, at 55°C, ATP hydrolysis in the L4-MRE11 complex was active in the presence of Mg 2ϩ (Fig. 3,  A and D), whereas little activity was observed when the added ion was Mn 2ϩ (required for MRE11 nuclease activity) (Fig. 3, C  and D). Mn 2ϩ competed with Mg 2ϩ for binding to RAD50 but failed to support significant ATP hydrolytic activity. Consequently, the rate of ATP hydrolysis was intermediate between Mg 2ϩ and Mn 2ϩ when both ions were present (Fig. 3, B and D).

RAD50-MRE11-Nbs1 Exo-endonuclease Switch
We correlated the ATPase activity of RAD50 with changes in FRET intensity. Remarkably, when L4 or L4-MRE11 was incubated with ATP (Mg 2ϩ ) for 10 min at 55°C, the FRET signal P]ATP. D, the percentage of ADP formed was quantified by phosphorimaging and used to calculate ATP hydrolysis rates (expressed in turnover number per minute) in the presence or absence of DNA. E, shown is FRET between ATP-FITC and ATP-TR bound to L4-MRE11 at 55°C in Mg 2ϩ , Mn 2ϩ , or a stoichiometric ratio of both Mg 2ϩ /Mn 2ϩ , or no ion was measured as described in Fig. 2. ATP hydrolysis and/or the addition of Mn 2ϩ is sufficient to lose FRET and open the closed complex. F, shown is the nucleotide binding affinity measured at 55°C for reactions in E using fluorescence anisotropy for ATP-FITC and ATP-TR. The excitation and emission of wavelengths were 470 and 522 nm for ATP-FITC and 590 and 615 nm for ATP-TR; gray is no protein, the green bar is ATP-FITC, and the red bar is ATP-TR. The presence (ϩ) or absence (Ϫ) of DNA is indicated. Samples in the presence (ϩ) or absence (Ϫ) of heating to 55°C are indicated.
diminished to the base value, indicating that the L4 dimer was in close proximity before but not after ATP hydrolysis (Fig. 3E). Thus, in the absence of nuclease activity, ATP hydrolysis alone was capable of opening the closed L4-MRE11 complex. Loss of the FRET intensity could arise if either the donor or acceptor nucleotides dissociated upon heating. Thus, in parallel reactions we measured nucleotide binding of both the TR-and the FITC-labeled nucleotides by anisotropy at their distinct max (Fig. 3F). Binding of the ATP analogues was not significantly different from L4-MRE11 before and after heating under physiological ionic strength. ATP binding to L4-MRE11 (supplemental Fig. S3) was specific because, as expected, little to no nucleotide bound to L4 (supplemental Fig. S3) or to L4-MRE11 (Fig. 3F, Table 1) in the absence of added ions. Thus, the loss of FRET was due to hydrolysis-driven separation of the RAD50 subunits.
The FRET intensity of L4-MRE11 was low before or after heating when both the RAD50 ATPase and the MRE11 nuclease were active simultaneously (Mn 2ϩ and Mg 2ϩ together). However, at 55°C the FRET intensity of L4-MRE11 was consistently higher relative to either ATP (Mn 2ϩ ) or to ATP (Mg 2ϩ ) added alone (Fig. 3E). Because the rate of ATP hydrolysis in Mn 2ϩ and Mg 2ϩ is slower relative to Mg 2ϩ alone, the results implied that ATP hydrolysis modulated the extent of opening of the nuclease active L4-MRE11, which was in an intermediate conformation between a fully closed and a fully open state when both L4 and MRE11 were active enzymes.
The Open Configuration of RAD50-MRE11 Binds Best to DNA-We asked whether the open and closed conformations differed when L4-MRE11 was bound to DNA. However, for all measurements tested, we were unable to observe a conformational difference in the DNA-bound and DNA-free RAD50-MRE11 complexes. Using SAXS (45), the addition of ATP (Mg 2ϩ ) resulted in a more "bumpy" scattering curve and a sharper P(r) function maximum, consistent with a more compact, spherical conformation of the closed complex (Fig. 4A). However, the ATP-dependent conformational changes of the DNA-free and DNA-bound complexes were indistinguishable (Fig. 4B). Sedimentation velocity analysis (51,52) confirmed that the addition of ATP (Mg 2ϩ ) to DNA-bound L4-MRE11 was characterized by an increase in the s 20,w value from 8.0 to 8.5 (Fig. 4D), but the analogous alteration was observed upon ATP binding to L4-MRE11 in the absence of DNA (Fig. 4C). Additionally, the rate of ATP hydrolysis in the RAD50 subunits of L4-MRE11 was equivalent in the presence or absence of DNA (Fig. 3), implying that opening of the closed complex occurred equally well under both conditions. No differences in the FRET signal of the DNA-bound (Fig. 2, E and F) and DNA-free L4-MRE11 (supplemental Fig. S3) were observed under any of the conditions measured. Thus, RAD50 acted as an autonomous unit in which opening and closing resulted from continuous cycles of ATP binding, ATP hydrolysis, and ATP exchange whether or not L4-MRE11 was bound to DNA.
Because the presence of DNA did not influence the structural transition, we tested whether the structural transition altered the affinity for DNA for the RAD50-MRE11 complex (Fig. 4, E  and F). We added purified L4 or L4-MRE11 to FITC-labeled ssDNA or dsDNA and measured the DNA binding affinity under conditions of Mg 2ϩ (closed configuration) or Mn 2ϩ (open configuration). These experiments were performed at 25°C to silence enzymatic activity of RAD50-MRE11 (49). In agreement with previous results (14,25,27,28,38), we found that both L4 and MRE11 were able to bind DNA (supplemental Table SII). However, the open and closed L4-MRE11 complexes had significantly different DNA affinities. For L4, at 25°C DNA binding affinity was modest and did not change whether L4 was bound with ATP or whether Mg 2ϩ or Mn 2ϩ were the added ions (supplemental Table SII). In contrast, DNA binding affinity of MRE11 and L4-MRE11 substantially increased (10 -20-fold) in the presence of Mn 2ϩ (Fig. 4, E and F, supplemental Table SII). The increase in DNA binding affinity of L4-MRE11 was observed whether Mn 2ϩ was added independently or together with Mg 2ϩ (supplemental Table SII). Thus, RAD50-MRE11 bound best to DNA in the presence of Mn 2ϩ , conditions under which the complex was open (Fig. 2, E  and F).
ATP Hydrolysis Prevents Loss of MRE11 3Ј to 5Ј Exonuclease Activity-MRE11 is a 3Ј to 5Ј exonuclease on dsDNA and an endonuclease on ssDNA at protruding 3Ј and 5Ј ends and 3Ј branches (27)(28)(29). Because L4-MRE11 could form open and closed states when bound to DNA, we tested whether the different structural states had consequences on the nuclease functions of the MRE11.
To directly measure the exonuclease activity of MRE11, we synthesized a DNA duplex in which a 2-AP base was incorporated at the second position from the 3Ј end of the DNA duplex ( Fig. 5A) (14). The 2-AP is intrinsically fluorescent, but its fluorescence is quenched when incorporated in dsDNA (Fig. 5A). If the MRE11 exonuclease was active, then we anticipated a gain in fluorescence intensity when 2-aminopurine is released during 3Ј to 5Ј nucleolytic activity. At 55°C in the presence of Mn 2ϩ added alone (Fig. 5B) or together with Mg 2ϩ (Fig. 5D), MRE11 1-342 excision resulted in efficient release of 2-AP. MRE11 1-342 does not bind ATP, and as expected, catalytic activity of MRE11 1-342 was not significantly affected by the addition of ATP, ADP, or the non-hydrolyzable analog ATP␥S (Fig. 5, B and D).
We repeated the MRE11 exonuclease assay for the L4-MRE11 complex. Indeed, when associated with RAD50, MRE11 was an active exonuclease in the absence of nucleotides as long as Mn 2ϩ was added to the reaction (Fig. 5, C and E). Surprisingly, the presence of RAD50 markedly suppressed the specific activity of MRE11 exonuclease activity (k cat /K m ) (supplemental Table SIII) around 2-fold relative to MRE11 alone (compare Fig. 5, B and C, supplemental Table SIII). Furthermore, the exonuclease activity of L4-MRE11 in the presence of ATP (Mn 2ϩ ), ATP␥S (Mn 2ϩ ), or ADP (Mn 2ϩ ) was not substantially different from that observed in the absence of nucleotide (Fig. 5C). These findings implied that RAD50 intrinsically inhibited the exonuclease activity of L4-MRE11, and in the absence of ATP hydrolysis, exonuclease suppression was independent of bound nucleotide.
We tested the effects on the MRE11 exonuclease activity if RAD50 was also an active ATPase (ATP (Mg 2ϩ and Mn 2ϩ )) (Fig. 5E). However, we observed only a modest rise in the exo-nuclease activity of MRE11 under these conditions (compare Fig. C and Fig. 5E). To test whether RAD50 ATPase activity promoted the MRE11 exonuclease activity, we blocked it using ATP␥S (Fig. 5E). Unexpectedly, the specific activity (k cat /K m ) of the L4-MRE11 exonuclease activity was suppressed around 2-fold by the addition of ATP␥S (Mg 2ϩ and Mn 2ϩ ) relative to the addition of ATP (Mg 2ϩ and Mn 2ϩ ) (Fig. 5E, supplemental Table SIII) and about 5-fold relative to MRE11 alone (compare Fig. 5, D and E, supplemental Table SIII). Thus, the level of exonuclease activity of RAD50-MRE11 depended on ATP hydrolysis but not by a canonical stimulation mechanism. Rather, ATP hydrolysis prevented shutdown of MRE11 exonuclease activity. Because it was suppressed by ATP␥S only under conditions of active hydrolysis (compare Fig. 5, C and E), these findings implied that maintenance of exonuclease activity depended on the open conformation. Consistent with a regulatory role for ATP hydrolysis in maintaining the open complex, the exonuclease activity was highest when L4-MRE11 was ADP-bound (Fig. 5E).
To determine the extent to which ATP hydrolysis regulated MRE11 exonuclease activity, we measured the catalytic rate of ATP hydrolysis relative to the rate of exonuclease activity. RAD50 was a weak ATPase (Fig. 3D), with a maximum rate no greater than 15 molecules of ATP per minute (per ATP binding site) under optimum conditions. ATP hydrolysis did not change when RAD50-MRE11 was bound to DNA. However, the rate of ATP turnover in the L4 subunits of L4-MRE11 was Ͼ100-fold higher than the k cat of the MRE11 exonuclease (compare Fig. 3D and supplemental Table SIII). Thus, ATP hydrolysis did not limit the velocity of nucleotide release from the 3Ј end.
ATP Hydrolysis in RAD50 Switches MRE11 from Exonuclease to Endonuclease-We next evaluated the effect of nucleotidedependent structural dynamics of the endonuclease activity of MRE11 in the presence of Mg 2ϩ and Mn 2ϩ . To measure the endonuclease activity, we developed a second DNA template in which the endo-and exonucleotide could be visualized individually or simultaneously by adjusting the solution conditions (Fig. 6A). The 59-nucleotide template folded back to form a duplex region of 22 base pairs, a 3-nucleotide loop, and 6 nucleotide ssDNA overhangs on both the 3Ј and the 5Ј end. The MRE11 endonuclease activity is highest on ssDNA at protruding 3Ј and 5Ј ends and 3Ј branches (27)(28)(29). Thus, the hairpin template was 32 P-labeled on the 5Ј terminus, and the endonuclease activity was measured by the reduction in the length of the incision product after resolution on PAGE gels (Fig. 6B).
For L4-MRE11, we anticipated that if the MRE11 endonuclease were active, then it would remove either of the two arms of the template, one of which would retain the 32 P-label (Fig. 6,  A and B, red arrows). Indeed, when we incubated L4-MRE11 with the DNA substrate in the presence of both ATP (Mn 2ϩ ) and ATP (Mg 2ϩ ), we observed a doublet band corresponding to endonucleolytic removal of the overhang at the junction between the dsDNA and ssDNA and 1-2 nucleotides from the end (lane 3, Fig. 6B, red arrows). The endonucleolytic function of MRE11 was specific to MRE11 and required ATP binding. No incision product was observed after the addition of ATP in the absence of L4-MRE11 (lane 1, Fig. 6B) or for the L4-MRE11 complex without ATP (lane 2, Fig. 6B). Furthermore, no incision activity was observed for MRE11 1-342 and L4-MRE11 in the absence of Mn 2ϩ under any conditions tested (lanes 10 and 11, Fig. 6B).
In contrast to the exonuclease activity of MRE11, however, ATP hydrolysis in the subunits of L4 was not required for endonuclease activity of MRE11. When ATP␥S was the added nucleotide, L4-MRE11 generated an incision product that was fivesix nucleotides shorter than the original template and more specific for ssDNA/dsDNA junction (lane 4, Fig. 6B). No incision product was detected when ADP was the added nucleotide (lane 5, Fig. 6B), conditions under which the exonuclease activity was highest.
We repeated the analysis with MRE11 alone in the absence of L4. Indeed, we observed promiscuous exonuclease and endonuclease products of MRE11   (lanes 6 -9, Fig. 6B), consistent with loss of its inhibition by RAD50 (Fig. 5C). Thus, regulation of the endo-and exonuclease activities of MRE11 depended on the presence of L4. Moreover, the exonuclease and endonuclease functions depended differently on the open and closed complexes of L4-MRE11. ATP hydrolysis shifted equilibrium toward the open complex, a condition under which the exonuclease activity of MRE11 was maintained. In contrast, the endonuclease activity of MRE11 depended only on ATP binding, which tended to close the RAD50-MRE11 complex.
To further test the dependence of MRE11 endonuclease activity on the ATP dynamics of RAD50, we repeated the assay in the presence of Mn 2ϩ alone or together with Mg 2ϩ using a series of templates comprising 3Ј overhangs, 5Ј overhangs, or blunt ends (supplemental Fig. S4). Each of these templates was labeled with 32 P on the 5Ј end (supplemental Fig. S4A). Within 60 min the single-stranded tails were removed by MRE11 endonuclease activity in templates containing the 3Ј overhang whether or not Mg 2ϩ was present in the reaction. Each reaction product was shorter by five-six nucleotides, consistent with incision at the ssDNA-dsDNA junction. As a control, we followed the reaction product of a 5Ј overhang, which cannot be processed by the MRE11 exonuclease function. Indeed, consistent with an endonuclease activity, incision resulted in loss of the 5Ј end ssDNA tail, and the 5Ј 32 P-labeled product was no longer observed. No activity was observed in the absence of ATP. Thus, the endonuclease activity of MRE11 did not require ATP hydrolysis in the RAD50 subunits.
In a final set of experiments, we measured the endonuclease activity on ss plasmid DNA using both the FL and L4-MRE11 proteins (Fig. 6C). Due to its circular nature, any incision event will arise from endonucleolytic activity on the plasmid DNA. The linearized template was detected by ethidium staining after resolution on agarose gels. Indeed, ATP (Mn 2ϩ ) addition to either L4-MRE11 or FL resulted in loss of the DNA template within 30 min (Fig. 6D, ATP-Mn 2ϩ ). In the absence of Mn 2ϩ , no degradation was observed when ATP (Mg 2ϩ ) was the added nucleotide (Fig. 6D, ATP-Mg 2ϩ ). Thus, the endonuclease activity of RAD50-MRE11 depended on ATP binding but not on its hydrolysis. Collectively, these results implied that opening and

RAD50-MRE11-Nbs1 Exo-endonuclease Switch
closing of RAD50-MRE11 complexes imparted distinct physiological roles that fundamentally converted the nucleolytic function of MRE11 between exonuclease and endonuclease functions. ATP hydrolysis served as a switch between these two activities.

DISCUSSION
RAD50 forms a tight complex with MRE11, yet the molecular mechanism by which RAD50 and MRE11 communicate to orchestrate damage detection is poorly understood. Moreover, ATP regulates MRE11 functions, but the role of ATP binding in discriminating DNA end processing and the role of ATP hydrolysis are unknown. We report here that RAD50 coordinates these two activities of MRE11 by ATP-dependent structural rearrangements and a remarkable ability to switch MRE11 nucleolytic activity. Continuous cycles of ATP binding and ATP hydrolysis impart a "butterfly" movement that opens and closes the RAD50-MRE11 complex (Fig. 6E). ATP binding shifts the equilibrium to closed conformation, in which MRE11 is an endonuclease. ATP hydrolysis shifts the equilibrium to the open state in which MRE11 is an exonuclease (Fig. 6E). We find that ATP hydrolysis acts as a biological switch that converts the MRE11 endonuclease into an exonuclease (Fig. 6E). Although conformational changes in proteins commonly play regulatory roles, the findings presented here are among the first examples where structural alterations are used to entirely change the function of the protein.
ATP-dependent nuclease switching provides a plausible mechanism and critical insight into how a single RAD50-MRE11 complex is able to coordinate DNA repair, DNA end recognition, and decision points for directing repair to HR, non-homologous end joining, or microhomology-mediated end joining repair pathways.
RAD50 Provides Check and Balance System for MRE11-dependent DNA Repair-RAD50 is exquisitely designed to act as a critical "check and balance" system to coordinate the DNA repair function of MRE11. Arguably, the most important modular function of RAD50 is to avoid unregulated activity of MRE11. In the absence of RAD50, MRE11 behaves like a "runaway train" and operates indiscriminately and simultaneously on both ssDNA and dsDNA (Fig. 6B). Unregulated MRE11mediated activity predicts an increase in the demand for double strand break repair, but recruitment of unregulated MRE11 itself would cause more breaks. Thus, the interaction of RAD50 with MRE11 is poised to minimize toxic and futile cycles of nuclease activity. Importantly, suppression of the MRE11 nuclease by RAD50 is its only nucleotide-independent function (Fig. 5). Thus, the repressive action of RAD50 provides an intrinsic gate-keeping function and a global check of MRE11 activity. Interestingly, a negative regulatory function of RAD50 protein was recently observed in another archaeal species Methanococcus jannaschii (26). Thus, our results imply that repression of MRE11 is likely to be a common function of RAD50.
The "balance" for the DNA repair activity of MRE11 is provided by the nucleotide-dependent properties of RAD50. In the ADP-bound state, the endonuclease activity of RAD50 is lost, and the exonuclease activity is at its maximum (Figs. 5 and 6, A   and B). Because ATP and ADP bound in RAD50 subunits are in rapid exchange with solution nucleotides (Fig. 1G), ATP replaces ADP, and the endonucleolytic form of MRE11 is active again. Thus, opening and closing of RAD50-MRE11 are reminiscent of a DNA clamp, which characterizes other DNA binding members of the ABC-ATPase superfamily (40,53,54).
We find that RAD50-MRE11 most frequently exists as an ATP-bound endonuclease in the closed conformation. However, as judged by its biochemical properties, limiting the exonuclease activity of MRE11 appears to be the most crucial role of RAD50 nucleotide dynamics. Indeed, in the absence of ATP hydrolysis, we find that the exonuclease plummets nearly 5-fold relative to the MRE11 alone (Fig. 5). Under these conditions, MRE11 would be unable to generate microhomology domains or to remove blocked ends for successful recombination or end joining (supplemental Fig. S5). In this regard, ATP hydrolysis is not a canonical "activator" of MRE11 but is designed to avoid loss of its exonuclease function and maintain it at a useful level. The activity of the MRE11 as an exonuclease is maximal when RAD50 is an ADP-bound complex and the endonuclease function is suppressed (Figs. 5 and 6). However, the rate of ATP hydrolysis exceeds the rate of MRE11 exonuclease activity, and rapid ADP-ATP exchange quickly returns RAD50-MRE11 to a closed ATP-bound state. Collectively, the nucleotide dynamics in the RAD50 subunits provide a brief window within which MRE11 can function as the exonuclease and strictly limits the number of nucleotides released from the 3Ј end.
Model for Switching-directed Pathway Choice-The RAD50-MRE11-Nbs1 complex has been implicated in every aspect of DSB processing from the initial detection to the choice among pathways for repair. Based on the nucleotide dynamics of RAD50, we propose a testable model in which the structural changes integrate the nucleolytic processing of MRE11 and the tethering functions of RAD50-MRE11 to direct pathway choice. MRE11 has both endo-and an exonuclease activity in vitro, but whether these functions are coordinated remains enigmatic. However, the continuous MRE11 exo-endo cycle poises MRE11 for ordered and sequential nuclease functions in rapid succession. Opening and closing of RAD50-MRE11 is independent of DNA binding. However, RAD50-MRE11 binds best to DNA in an open exonuclease-active form, and ADPbound RAD50-MRE11 has a preference for ssDNA overhang in that state (Fig. 6E). The rapid exchange of ADP for ATP exchange results in closing of ATP-bound RAD50-MRE11 around the ssDNA or dsDNA junction, yielding endonucleolytic incision of the overhang and production of a blunt end (Fig. 6E).
ATP hydrolysis promotes a second opening event, and in the open state, RAD50-MRE11 adopts extended conformation and either dissociates or is poised to span between DNA ends as a tethering bridge. At flush ends, the open RAD50-MRE11 is competent for 3Ј to 5Ј MRE11 exonuclease activity and, after nibbling into the 3Ј end, likely hands the ends to Sae2/CtIp for 5Ј to 3Ј resection and successful HR (6,31). RAD50-MRE11 has a dsDNA melting activity (29), which presumably aids in 3Ј to 5Ј exonuclease activity.
Because both the endo-and exonuclease activities of MRE11 occur in the presence of ATP, RAD50-MRE11 is unlikely to efficiently discriminate among hairpin opening, ssDNA endonuclease, dsDNA exonuclease, and 5Ј dsDNA endonuclease in vitro. However, RAD50-MRE11 affinity for particular DNA ends provides a plausible means for the precise roles for RAD50-MRE11 in double strand break repair in vivo. RAD50-MRE11 does not efficiently process 3Ј overhangs or hairpin DNA, but it binds to those substrates with significant affinity (27). In these cases we envision a mechanism by which MRE11-RAD50 acts as a scaffold and recruits DNA-PK (55) (or another suitable component of the repair machinery) to complete end processing. The ability of RAD50-MRE11 to process a particular type of end provides a plausible and simple driver to direct pathway choice. The rapid opening and closing and nuclease switching ensures that the complex broadly recognizes most DNA broken ends but uses end recognition to distinguish between direct nucleolytic processing versus scaffolding and recruitment. Whatever the detailed mechanisms, RAD50-MRE11 is a "first responder" for DSBs. The structural rearrangements of RAD50-MRE11 complex and its ability to switch nucleolytic functions poise it for integration of ATP dynamics, nucleolytic processing, and pathway selection.