LRET-derived HADDOCK structural models describe the conformational heterogeneity required for processivity of the Mre11-Rad50 DNA damage repair complex

The Mre11-Rad50-Nbs1 protein complex is one of the first responders to DNA double strand breaks. Studies have shown that the catalytic activities of the evolutionarily conserved Mre11-Rad50 (MR) core complex depend on an ATP-dependent global conformational change that takes the macromolecule from an open, extended structure in the absence of ATP to a closed, globular structure when ATP is bound. We have previously identified an additional ‘partially open’ conformation using Luminescence Resonance Energy Transfer (LRET) experiments. Here, a combination of LRET and the molecular docking program HADDOCK was used to further investigate this partially open state and identify three conformations of ATP-bound MR in solution: closed, partially open, and open, which are in addition to the extended, apo conformation. These models are supported with mutagenesis and SAXS data that corroborate the presence of these three states and suggest a mechanism for the processivity of the MR complex along the DNA.


Introduction
Mre11-Rad50-Nbs1 (MRN) is an essential protein complex required for the repair of DNA double strand breaks (DSBs) (Paull, 2018;Syed and Tainer, 2018). This complex recognizes the broken DNA and begins processing the break via Mre11 exo-and endonuclease activities and Rad50 ATP binding and hydrolysis (Paull, 2018). Nbs1, found only in eukaryotes, further modulates MR activity and signals downstream repair effectors to the site of the break (Deshpande et al., 2016;Oh et al., 2016). If DNA DSBs are not repaired, the cell may undergo cell death via apoptosis, or, if the break is not repaired correctly, a loss of genetic information or gross chromosomal rearrangements can occur potentially resulting in immunodeficiencies and cancer (Ciccia and Elledge, 2010;Oh and Symington, 2018). Many mechanistic and structural studies have been performed on the evolutionarily conserved Mre112-Rad502 (MR) core complex from bacteria, archaebacteria, and eukaryotes and have shown that the complex undergoes a dramatic ATP-induced global conformational change that is required for its various functions. This two-state model, which originated from X-ray crystallographic studies (Lafrance-Vanasse et al., 2015;Lammens et al., 2011;Lim et al., 2011;Möckel et al., 2012), has MR transforming from an extended arms-open-wide conformation, where the two Rad50 nucleotide binding domains (NBDs) are far apart in space, to a 'closed' conformation that sandwiches two ATPs between the NBDs in a more compact, globular structure (Fig. 1A). In the closed conformation, the Mre11 nuclease active sites are occluded and Rad50 can bind DNA (Liu et al., 2016;Möckel et al., 2012;Rojowska et al., 2014;Schiller et al., 2012;Seifert et al., 2016). Once Rad50 hydrolyzes the bound ATPs, the complex returns to the extended conformation where DNA substrates can again access the Mre11 active sites. Critically, ATP binding and hydrolysis, and therefore the cycling between states, appears to be required for DNA unwinding (Cannon et al., 2013), processive Mre11 nuclease activity (Herdendorf et al., 2011;Paull and Gellert, 1998), and downstream signaling events through ATM kinase (Cassani et al., 2019;Lee et al., 2013).
Because the Mre11 active site is occluded in the closed conformation, it was hypothesized that MR nuclease activity originates from either the extended or an otherwise unknown intermediate structure (Deshpande et al., 2014;Lafrance-Vanasse et al., 2015;Lammens et al., 2011;Möckel et al., 2012). A recent cryo-EM structure of E. coli MR (called SbcCD) bound to a double-stranded DNA (dsDNA) substrate revealed a structure where the ADP-bound Rad50s are associated and the Mre11 dimer has moved to one side to interact asymmetrically with the two Rad50s and the dsDNA substrate (Käshammer et al., 2019). In addition, we have previously used Luminescence Resonance Energy Transfer (LRET) experiments to illuminate the presence of a 'partially open' conformation in a truncated construct of hyperthermophilic P. furiosus MR (Pf MR NBD ) in both ATP and ATP-free conditions (Boswell et al., 2020).
Thus, a variety of functionally relevant structures of the MR complex may exist in solution. As LRET has been successfully used to characterize the interactions of NBDs in several ABC ATPase membrane proteins (Cooper and Altenberg, 2013;Zoghbi et al., 2017Zoghbi et al., , 2012Zoghbi and Altenberg, 2018), we significantly extended our initial LRET studies on the Pf MR NBD complex to further characterize this partially open conformation. Multiple LRET probes were introduced throughout the Rad50 NBD to determine a network of distances between residues across the Rad50-Rad50 interface (Fig. 1B). These LRET-determined distances were then used as unambiguous distance restraints in the molecular docking program HADDOCK (Dominguez et al., 2003;Van Zundert et al., 2016) to obtain models of the ATPbound MR NBD complex. Here, we present structural models of three distinct conformations of the ATP- where Rad50 contains the coiled-coil domains and an apical zinc hook dimerization motif, confirmed that these conformations are also observed for the complete MR complex. Site-directed mutagenesis was used to disrupt specific conformations, and the effects on Mre11 and Rad50 activities validated our models and put them into a functional context. Small-angle X-ray scattering (SAXS) was also employed to confirm these three conformations of ATP-bound MR NBD in solution and to assign approximate populations to each. In conclusion, we have combined orthogonal biophysical and computational methods to describe three distinct global conformations of ATP-bound Pf MR in solution and demonstrate that they offer valuable insight into how MR functions along the DNA.

Results
Multiple LRET probe positions provide a network of measurements. LRET experiments employ a luminescent lanthanide donor and fluorophore acceptor pair with an appropriate Förster radius (R0) (Zoghbi and Altenberg, 2018). LRET probes are introduced into a protein most easily through a thiolmaleimide reaction with a unique cysteine. For Pf Rad50 NBD , where the coiled-coil domains are truncated, single cysteines were introduced into the naturally cysteine-less construct, and MR NBD activity was tested to ensure an active complex (Supplemental Fig. S1). Mutations were made primarily in loop regions to minimize disruptions to the protein fold. In all, six separate single cysteine mutations were made throughout Rad50 NBD (Fig. 1B). These cysteines were subsequently labeled with thiol-reactive LRET donor or acceptor molecules. To make the MR NBD complex for LRET experiments, equimolar amounts of donor-labeled (Tb 3+ -chelate) Rad50 NBD and acceptor-labeled (Bodipy FL or Cy3) Rad50 NBD were mixed with twice the molar ratio of Mre11 so that 50% of the resulting M2R2 complexes had one donor and one acceptor fluorophore (on separate Rad50 protomers). Not only were identical cysteines mixed within a complex (e.g., Tb 3+ -S13C and Bodipy-S13C), but complexes were also made where cysteine mutants were mixed with other cysteine mutants (e.g., Tb 3+ -S13C and Bodipy-L51C). Additional distance measurements were obtained for a given LRET pair by changing the identity of the acceptor, as the Förster radius for Tb 3+ and Cy3 (61.2 Å) is longer than that of Tb 3+ and Bodipy FL (44.9 Å). Finally, donorand acceptor-labeled cysteines within mixed pairs were swapped (e.g., Tb 3+ -S13C and Bodipy-L51C versus Tb 3+ -L51C and Bodipy-S13C) for added confidence in measurements. In total, 20 different cysteine pairs resulted in 50 unique samples that gave 54 total measured distances between the two Rad50 protomers in the MR NBD complex (Supplemental Table S1).
LRET measurements reveal three distinct sets of distances. Following laser excitation of the Tb 3+chelate moiety and a 200 µsec delay, donor-sensitized Bodipy FL or Cy3 fluorescence emission decay curves were collected for each of the MR NBD LRET samples at 50 °C in the presence of 5 mM Mg 2+ and 2 mM ATP (Fig. 1C). Under these conditions, Rad50 NBD should be >99% bound to ATP as the KD for ATP is ~3 µM and there is no measurable ATP hydrolysis in 1 h at 50 °C. In multi-exponential fits, the emission decays were best described by two or three exponentials depending on the identity of the LRET pair (see Methods). In all cases the first lifetime (<100 µs) is a function of instrument response time and was discarded (Cooper and Altenberg, 2013;Zoghbi et al., 2017Zoghbi et al., , 2012. The Tb 3+ -chelate luminescence decays were also recorded at each probe position in donor-only labeled MR NBD complexes. As expected, the value of the Tb 3+ -chelate lifetime changed with the local environment of each cysteine. Using these Tb 3+ -chelate donor lifetimes, combined with the donor-sensitized acceptor lifetimes and the R0 of the dye pair in the sample, distances were calculated between probes for each LRET pair. For the majority of the 20 cysteine pairs analyzed, combining the data for the Bodipy/Tb 3+ -and Cy3/Tb 3+ -labeled samples gave three distinct distances. For 10 of the pairs (e.g., L51C-L51C), the longer distance (d2 Bodipy ) in the Bodipy/Tb 3+ samples matched the shorter distance (d1 Cy3 ) in the Cy3/Tb 3+ samples, and the Cy3/Tb 3+ samples gave a second, longer distance (d2 Cy3 ) ( Fig. 1D and Supplemental Table S1). This longer distance became 'visible' in samples where Cy3 was the acceptor because the R0 for Tb 3+ and Cy3 is longer. For four of the pairs (e.g., L51C-A66C), the one Bodipy/Tb 3+ distance did not overlap with the two Cy3/Tb 3+ -determined distances, and the combined data resulted in three distances.
For two pairs (A66C-A66C and S93C-S93C), only one distance was seen in the Bodipy/Tb 3+ data, while the Cy3/Tb 3+ data contained two distances. In these samples, the d Bodipy matched d1 Cy3 for a total of two distances. And finally, for four pairs (e.g., S13C-S93C) only Cy3/Tb 3+ samples were made resulting in two distances. Together, these data illuminate the presence of a third solution state in addition to the closed and partially open.
To confirm that the distances observed in MR NBD were the same in full-length MR, all of the cysteine mutations were also introduced into full-length Pf Rad50. We previously reported that the two native cysteines in the zinc hook motif of full-length Rad50 are not efficiently labeled by the LRET fluorophores and do not result in LRET donor-sensitized acceptor signal (Boswell et al., 2020). Tb 3+chelate donor only lifetimes measured in these mutants were identical to those measured for the Rad50 NBD construct, indicating that the local environments of the introduced cysteines do not change between full-length and NBD constructs. Unfortunately, because full-length Rad50 dimerizes at the zinc hook, labeled cysteine mutants could not be mixed and only "identity" LRET pairs (e.g., Bodipy-L51C and Tb 3+ -L51C) could be made. Nonetheless, for all LRET probe positions measured, the distances between full-length MR cysteine pairs were within a few Ångströms of those measured in MR NBD (Fig. 1D).
HADDOCK models of three MR NBD conformations. The measured LRET distances were input as unambiguous restraints in the HADDOCK molecular docking program (Dominguez et al., 2003;Van Zundert et al., 2016), defining the Cb-Cb distance between the LRET-labeled residues. The unambiguous restraints included symmetrical distances for each LRET pair (e.g., both Rad50a protomer L51 to Rad50b protomer S13 and Rad50a S13 to Rad50b L51). These unambiguous restraints were used to dock two Rad50 NBD protomers with one Mre11 dimer in a three-body docking simulation.
The 'closed' HADDOCK model fit very closely to the measured LRET distances ( Fig. 2A, Supplemental Movie S1, and Supplemental Table S2). HADDOCK returned 190 structures in five clusters, with 164 in the top-scored cluster. The Rad50 dimer formed in this model is nearly identical to the AMPPNP-bound dimer structure of Pf Rad50 NBD (PDB: 3QKU, all-atom root-mean-square deviation [RMSD] = 1.11 Å) (Williams et al., 2011). Except in two cases, the Cb-Cb distances between the Rad50 LRET pairs in the HADDOCK model were within ±5 Å of their respective unambiguous LRET restraint.
LRET pairs A66-A66 and A66-S93 deviated more significantly with differences of 7.4 and 11.0 Å, respectively. This deviation could arise from slight differences in the loop structures between the solution (LRET) and crystal states (i.e., input PDB; note that HADDOCK does not move backbone atom positions during model refinement). Specifically, we observed the expected relative positions within the associated Rad50s for residues in the Walker A motif (N32) of one protomer and the D-loop (D829) and signature motif (S793) of the other (Fig. 2D, top and movie S1). In this closed model, Rad50 interacts with Mre11 via the capping domain and along the top of the nuclease domain as observed in the T. maritima (PDB: 3THO) , M. jannaschii (PDB: 3AV0) (Lim et al., 2011), and E. coli (PDB: 6S6V) (Käshammer et al., 2019) nucleotide-bound structures (with Ca RMSDs of 4.9 Å, 1.9 Å, and 4.7 Å, respectively) (Supplemental Fig. S2A). Like the M. jannaschii structure, each Rad50 protomer makes contact with only one of the Mre11 capping domains mainly through interactions between capping domain b18 and Rad50 Lobe II aE and b8-10. In total, eight and four ionic or hydrogen bond interactions are made between Rad50 and the Mre11 capping and nuclease domains, respectively. In particular, unique contacts not seen in the other two conformations are made between Rad50 E831 and Mre11 H17 in the nuclease domain and Rad50 E758/E761 and Mre11 Y325/K327 in the capping domain. The combination of these interactions occludes the Mre11 nuclease active site for dsDNA, as previously described (Lim et al., 2011;Möckel et al., 2012).
The 'partially open' HADDOCK simulation returned nine clusters of models, and analysis of the top four clusters concluded that the distances between LRET pairs for each cluster had an average standard deviation of ~0.3 Å when compared among the top three clusters and increased to ~0.7 Å when adding the fourth (Fig. 2B, Supplemental Movie S1, and Supplemental Table S2). 13 out of 20 of the Cb-Cb distances in the HADDOCK model are within ±5.7 Å of the measured LRET distances (Supplemental Table S2). Interestingly, all of the pairs with larger (>5.7 Å) deviations included either A66 or S93, again  Table S2).
The Rad50 protomers have moved apart considerably (~41 Å D829-D829 Ca-Ca distance). Moreover, the orientation of the Rad50 protomers with respect to Mre11 has changed significantly (Supplemental To ensure that the unambiguous distance restraints obtained from one probe position were not dominating the structure calculations, HADDOCK runs were performed with systematic dropouts of all restraints calculated from a specific cysteine position. For example, for the L51C probe position, L51C-S13C, L51C-L51C, L51C-A66C, L51C-S93C, L51C-N774C, and L51C-V866C distances were all removed from the unambiguous restraints, and HADDOCK runs were repeated for each of the three sets of distances (closed, partially open, and open). In general, none of the dropouts appreciably changed the overall conformation of any of the states (Supplemental Fig. S3).
Destabilizing solution state conformers alters MR activity. Next, to decrease the stability of one or two conformations over the others, charge reversal mutations were made to several Mre11 residues that directly interact with Rad50. As there are a handful of shared interactions between Mre11 and Rad50 in the various conformation combinations, it was impossible to completely disrupt a given state. These Mre11 mutants were combined with full-length Rad50 to make MR complex, and then both Mre11 and Rad50 activities were tested (Fig. 3). To assay Mn 2+ -dependent Mre11 3'-to-5' exonuclease activity, two dsDNA substrates were interrogated: Exo2 and Exo11, which have a fluorescent 2-aminopurine incorporated as the second or eleventh nucleotide from the 3'-end of one strand, respectively. Although MR does not require ATP to cleave the Exo2 substrate, ATP is required for Exo11 as the enzyme needs to progress eleven base pairs along the DNA duplex to reach the position of the fluorescent nucleotide (Boswell et al., 2020;Herdendorf et al., 2011).
In the closed conformation, Mre11 nuclease domain residue H17 interacts with Rad50 E831 (Supplemental Fig. S4A). MR H17E had ~30% of wild type exonuclease activity on Exo2 DNA but does not cleave Exo11 DNA. Williams et al. identified H17 as a 'wedge residue' that helps to unwind the dsDNA helix (Williams et al., 2008). The observation that processive MR exonuclease activity is hindered in the H17E mutant supports that function. MR H17E decreased the Vmax of ATP hydrolysis to ~60% of wild type MR, indicating that this residue also assists in stabilizing the closed conformation from which hydrolysis proceeds. Mre11 Y325 and K327 are located in the capping domain on b18 and interact with E761 and E758 of Rad50, respectively, in the closed conformation (Supplemental Fig. S4A Fig. S4A and S4B). D313 and its neighboring residues in a loop at the top of the capping domain might be acting as a pivot point for Rad50 to rotate between the two states. Surprisingly, MR D313K had significantly increased exonuclease activity (Fig. 3). In fact, a ~2-fold increase in activity was observed for Exo2 in the absence of ATP; thus, we hypothesize that destabilizing both the closed and partially open conformations increases the population of the open conformation that accommodates dsDNA substrate. Activity against the Exo11 substrate increased ~2.5-fold in this mutant. In contrast to its high exonuclease activity, MR D313K reduced the Vmax for ATP hydrolysis by more than 50%, which was expected since the closed state is destabilized.
In the partially open complex, Mre11 K277 in b16 interacts with Rad50 E750, which is in aE at the base of a coiled-coil, whereas in the open complex, Mre11 K279 also in b16 interacts with Rad50 E783 in the short loop between b9 and b10 (Supplemental Fig. S4B and S4C). As these two residues are close in sequence space, a double mutant was constructed to destabilize both partially open and open conformations. The most striking feature of MR K277E/K279E was that it had no nuclease activity on Exo2 without ATP but increased to ~50% of wild type levels when ATP was added (Fig. 3). No other mutant tested required ATP for the Exo2 substrate, implying that it requires hydrolysis to open the complex before dsDNA can bind. Moreover, this mutant displayed the lowest activity in the presence of AMPPNP suggesting that the combination of the mutations and non-hydrolyzable analog effectively stabilized the closed state, fully occluding the dsDNA substrate. On the Exo11 substrate, the activity was also ~50% of wild type. This double mutant has no effect on ATP hydrolysis activity, since the closed conformation can readily form.
Finally, mutants were made for Mre11 R177, E181, and E152, which all make unique contacts in the open conformation (Supplemental Fig. S4C). These residues are along the "top edge" of the Mre11 nuclease domain in aD and aE, and R177/E181 and E152 are structurally homologous to part of the socalled "latching loop" and "fastener," respectively, of the E. coli SbcCD cryo-EM "cutting state" structure (Käshammer et al., 2019). The MR R177E/E181R double mutant had only ~60% of wild type exonuclease activity but wild type ATP hydrolysis activity, which is consistent with there being less open and more closed conformation. On the other hand, MR E152K, which interacts with Rad50 K860 in aE, showed more impaired exonuclease activity than R177/E181, but surprisingly decreased the Vmax for ATP hydrolysis by more than 50% (Fig. 3). As this mutant should readily form the closed conformation, this result was unexpected. Like the K227E/K279E mutant, both MR R177E/E181R and MR E152K had very little Exo2 activity in the presence of AMPPNP, again suggesting a complete stabilization of the closed conformation.  (Fig. 4). The open model did not fit well to the experimental SAXS data (c 2 = 13.6).

SAXS corroborates HADDOCK MR
Conversely, for the ATP-free MR NBD SAXS profile, a reasonable fit was only obtained with the open model (c 2 = 1.26). Next, MultiFoXS was used to fit the ATP-bound MR NBD SAXS profile to either two or three of the LRET-derived models simultaneously. The fit to two models improved the c 2 significantly and combined populations of open (19%) and closed (81%) conformations (Fig. 4B). The fit using all three models did not improve the c 2 but did assign populations of 15% open, 18% partially open, and 67% closed. The two-state fit to the apo MR NBD SAXS profile only slightly improved the c 2 (1.26 vs 1.05) and gave 90% open and 10% closed, while the 3-state fit gave 89% open, 5% partially open, and 6% closed again without improving c 2 further. Thus, the SAXS data supports the three states observed in LRET data and demonstrates the expected shift in population to the closed form in the presence of ATP.

Discussion
The disulfide bonds and non-hydrolyzable ATP analogs used in all of the closed MR NBD crystal structures and the selection for closed particles during the E. coli SbcCD cryo-EM model refinement imply that there is more than one conformation of ATP-bound MR. Nevertheless, the ATP-bound closed form has been referred to as the "resting" state given the cellular concentration of ATP and the KD for ATP-binding to Rad50 (Käshammer et al., 2019). Even so, both closed and partially open conformations were also identified in apo MR NBD (Boswell et al., 2020). In fact, the existing structural data (Deshpande et al., 2014;Williams et al., 2011) and the LRET-derived HADDOCK models presented here provide evidence for at least three conformations for Pf MR NBD in solution: closed, partially open, and open. As LRET is a distance-dependent phenomenon and distances could not be measured for probes more than ~85 Å apart, we did not obtain data for the extended conformation. Likewise, there was no evidence of MR NBD complexes containing dissociated Mre11 dimers (Käshammer et al., 2019;Saathoff et al., 2018), since the LRET probes on the Rad50 NBDs would also be too far apart in that scenario. Interestingly, this is not the first time that conformations with these names have been suggested for MR. Williams and coworkers (Williams et al., 2011) found that a combination of molecular dynamics (MD)-simulated conformational models best described their ATP-free and ATP-bound Pf MR NBD SAXS curves. Although those MD models do not resemble the models presented here, they nevertheless opened the door for conformational heterogeneity in MR, a possibility also proposed by others (Deshpande et al., 2014;Lafrance-Vanasse et al., 2015).
Protein expression and purification were performed as previously described for full-length Rad50 (Boswell et al., 2020), Rad50 NBD (Boswell et al., 2018), and Mre11 (Rahman et al., 2020). ATP hydrolysis assay. Rad50 ATP hydrolysis assays were performed essentially as described by Boswell et al. (Boswell et al., 2018). 0-300 µM ATP was titrated into microfuge tubes containing either 2.5 µM MR NBD complex or 2 µM MR complex and 50 mM Tris, 80 mM NaCl, 1% glycerol, 5 mM MgCl2, pH 7. Reactions without protein were included for each ATP concentration to control for ATP degradation and PO4 contamination. 60 µL reactions were incubated at 65 °C for 60 min after which the tubes were placed on ice. 50 µL of each reaction was then transferred to the wells of clear, flat-bottom 96-well plates and 100 µL of cold BIOMOL Green (Enzo Lifesciences) colorimetric reagent was added. After a 30 min incubation at room temperature to allow the color to develop, the amount of inorganic phosphate released by hydrolysis was quantified using the absorbance mode on a Synergy Neo2 multi-mode plate reader.  autofluorescence, emission due to direct excitation of the acceptor, and scattering of the excitation pulse), donor emission intensity was collected for Tb 3+ at 50 Hz through a 490/10 nm band-pass filter (Omega Optical) and donor-sensitized acceptor emission intensities were collected at 100 Hz through 520/10 nm (Bodipy FL) or 570/10 nm (Cy3) band-pass filters. PTI FeliX32 software was used to fit the Bodipy FL and Cy3 emission decay curves to either a two-or three-exponential function depending on the identity of the LRET pair. For samples where the distance between the LRET pair in the nucleotide-bound crystal structure was £42 Å, the data fit well to three exponentials for both Bodipy-and Cy3-labeled samples.
For samples where the LRET pair was separated by more than 45 Å, the Bodipy emission decays fit to only two exponentials while the Cy3 decays fit to three. Lifetime distributions shorter than ∼100 µs were discarded as these are largely a function of the instrument response time (Cooper and Altenberg, 2013;Zoghbi et al., 2017Zoghbi et al., , 2012Zoghbi and Altenberg, 2018). Donor Tb 3+ -chelate fluorescence decays were recorded at each probe position in donor-only labeled MR NBD complexes. Each Tb 3+ -chelate fluorescence decay fit well to two exponentials, with the longer lifetime comprising >85% of the signal. This longer lifetime was used in the distance analysis. From the Tb 3+ -chelate, Bodipy FL, and Cy3 lifetimes, the distances between donor and acceptor molecules (R) were calculated with and where E is the efficiency of energy transfer, tDA is the donor-sensitized lifetime of the acceptor (Bodipy FL or Cy3), tD is the lifetime of the donor (Tb 3+ ), and R0 is the Förster distance between Tb 3+ and Bodipy FL (44.9 Å) or Tb 3+ and Cy3 (61.2 Å). Errors are the standard deviations from the mean of at least three measurements.
HADDOCK. Molecular docking of the MR NBD complex was done using the GURU interface on the HADDOCK 2.4 webserver (Dominguez et al., 2003;Van Zundert et al., 2016). For the three-body docking protocol, the PDB inputs were 3DSC (Pf Mre11 dimer lacking the helix-loop-helix (HLH) motifs and Ctermini, DNAs deleted) (Williams et al., 2008) and two monomers of 3QKU (AMPPNP-bound Pf Rad50 NBD in complex with the Mre11 HLH motif) (Williams et al., 2011). Except for increasing the number of structures for rigid body docking (it0) to 3000, all settings used were default. The three linkers attaching the Pf Mre11 capping domain to the nuclease domain were allowed to be fully flexible (Y222-V236, Y249-G254, and V266-F273). To allow HADDOCK to move the two Rad50 monomers relative to each other, we input Rad50 as two identical monomers. C2 symmetry was enforced between the two Rad50 NBD monomers and between the two monomers of the Mre11 dimer. distances were input as unambiguous restraints and defined as the Cb-Cb distance between the LRETlabeled residues, ±5 Å. For distance restraints greater than 75 Å, the bounds were increased to ±7 Å as the lifetime fits giving these distances were in a more error-prone region (i.e., the flat section) of the Tb 3+ -Cy3 LRET efficiency curve. The model with the lowest HADDOCK score in each run was considered as the best structure. To ensure that the unambiguous distance restraints obtained from one probe position were not dominating the structure calculations, HADDOCK runs were performed with systematic dropouts of all restraints calculated from a specific cysteine position (Supplemental Fig. S3 (Classen et al., 2013;Dyer et al., 2014;Hura et al., 2009). For each sample, a total of 33 frames were collected at 0.3 sec intervals at 10 °C with a cell thickness of 1.5 mm. The sample to detector distance was 2 meters and the beam wavelength was 11 keV/1.27 Å. Buffer profiles were subtracted from sample profiles and the buffer subtracted frames for each sample were averaged using SAXS FrameSlice (https://sibyls.als.lbl.gov/ran).
Compared to the 3QKU PDB used as the input for HADDOCK runs, the Rad50 NBD protein construct used in the LRET and SAXS samples has an additional 7 amino acids on one side of the truncated coiled-coils and 8 amino acids on the other and the linker (GGAGGAGG) connecting them is slightly longer. To ensure that the structural models were as close to the protein construct used for the SAXS experiments as possible, the Rad50 coiled-coil and linker along with the 14 amino acid loop connecting the Mre11 helix-loop-helix and nuclease domain were modeled with loop modeling via Rosetta (version 3.11) (Mandell et al., 2009;Wang et al., 2007). This resulted in models only missing the 47 C-terminal amino acids of Mre11 (Supplemental Fig. S6            Movie S1. The movie depicts the transitions between the closed, partially open, and open conformations first shown from the 'side' view and then from the 'top' view. Next, Mre11 is hidden to highlight the position of the Rad50 Walker A (N32, magenta), signature motif (S793, yellow), and D-loop (D829) in these three conformations. Mre11 is colored orange, whereas Rad50 is colored blue and teal. The morph between the conformations was generated in Chimera (version 1.15) and rendered in PyMOL (version 2.4).