Biochemical and structural characterization of analogs of MRE11 breast cancer-associated mutant F237C

The MRE11–RAD50–NBS1 (MRN) protein complex plays a vital role in DNA double strand break sensing, signaling, and repair. Mutation in any component of this complex may lead to disease as disrupting DNA double strand break repair has the potential to cause translocations and loss of genomic information. Here, we have investigated an MRE11 mutation, F237C, identified in a breast cancer tumor. We found that the analogous mutant of Pyrococcus furiosus Mre11 diminishes both the exonuclease and endonuclease activities of Mre11 in vitro. Solution state NMR experiments show that this mutant causes structural changes in the DNA-bound Mre11 for both exo- and endonuclease substrates and causes the protein to become generally more rigid. Moreover, by comparing the NMR data for this cancer-associated mutant with two previously described Mre11 separation-of-nuclease function mutants, a potential allosteric network was detected within Mre11 that connects the active site to regions responsible for recognizing the DNA ends and for dimerization. Together, our data further highlight the dynamics required for Mre11 nuclease function and illuminate the presence of allostery within the enzyme.

The Phe237Cys (F237C) mutation in human MRE11 is one such mutation that was found in a breast cancer tumor sample 28 . In the crystal structure of the human MRE11 nuclease and capping domains, the aromatic ring of Phe237 is located in the conserved hydrophobic core of the nuclease domain and is surrounded by the sidechains of Asn212, Ile238, and Asp235 and takes part in a ring stacking interaction with Tyr179 (ref. 31). Thus, it was initially hypothesized that this mutation would disrupt the structure of MRE11 (ref. 31). Here, we describe our efforts to further characterize the effects of this mutant on the structure and function of MRE11. For our in vitro studies, we used the well-studied MR complex from the hyperthermophilic archaea Pyrococcus furiosus (Pf) as a model. Structure-based alignment suggests that an aromatic residue is conserved at this position in eukaryotes ( Fig. 1a; S. pombe Phe242 (F242) and C. thermophilum Phe233 (F233); PDB IDs 4FCX and 4YKE, respectively). In Pf, there is a tyrosine at the corresponding position (Tyr199), and similar packing interactions are observed around this aromatic ring (PDB ID 1II7) 32 . The high level of structural conservation across the domains of life implies that the aromatic residue at this position in Mre11 is important for function. Therefore, Pf Mre11 Tyr199 was mutated to cysteine to produce the corresponding cancer-associated mutant for use in the biochemical and structural studies reported below. We demonstrate that Pf Mre11 Tyr199Cys (Y199C) has decreased exo-and endonuclease activities in vitro, and solution state nuclear magnetic resonance (NMR) spectroscopy data suggest that this decreased catalytic activity originates from altered structure and dynamics within Mre11 that are critical for proper orientation of DNA substrates. Using this and our previously published NMR data on other nuclease deficient mutants, we were able to propose an allosteric network within Mre11. We also made the analogous mutation in S. cerevisiae (F233C) to examine the effect of the mutation in budding yeast and show that this mutation has no effect on DNA DSB repair in S. cerevisiae consistent with other nuclease mutants.

Results
Pf Mre11 Y199C affects both exo-and endonuclease activities. The Mn 2+ -dependent Mre11 exonuclease activity of Pf Mre11-Rad50 complexes (M 2 R 2 or simply MR) was measured using a 2-aminopurine (2-AP) based fluorescence assay. We compared the exonuclease activity of the Y199C MR complex with wild type and His85Ser (H85S) MR complexes (Fig. 1b). Mre11 H85S was used as negative control as it was previously shown to be both exo-and endonuclease inactive 33,34 . As expected, no activity was observed for any of the constructs in the absence of Mn 2+ . Since ATP hydrolysis by Rad50 is required for processive exonuclease activity 21 , we assayed the exonuclease activity in both the presence and absence of ATP. Without ATP, none of the MR complexes showed Mn 2+ -dependent exonuclease activity; however, with ATP/Mg 2+ , we observed differences in the resulting relative 2-AP fluorescence of wild type, H85S, and Y199C MR. The Y199C mutant showed a significant (p < 0.01) reduction of exonuclease activity in comparison to wild type, whereas H85S was completely inactive as expected (Fig. 1b).
Next, we sought to determine the effect of the Y199C mutant on Mre11 ssDNA endonuclease activity. Using a Cy3/BHQ2 fluorescence-based quantitative endonuclease assay, we observed Mn 2+ -dependent activity for wild type MR that is not dependent on ATP/Mg 2+ (Fig. 1c). When comparing the endonuclease activity across the mutants, we observed a significant (p < 0.01), albeit not complete, decrease in function for Y199C MR and no cleavage for H85S MR. To confirm these results, we also monitored the endonuclease activity of the MR complexes against a ssDNA virion plasmid substrate over time (Fig. 1d). In reactions containing Mn 2+ , both wild type and Y199C produced smaller DNA fragments that migrated faster through an agarose gel, indicating successful digestion of the plasmid, whereas H85S gave only intact DNA bands. ssDNA fragments from Y199C cleavage did not migrate as far through the gel as compared to wild type generated fragments revealing that Y199C produces larger endonuclease products than wild type. Thus, the Mre11 Y199C mutant has defects in both exo-and endonuclease activities.
To determine if the effects we observed in exo-and endonuclease activities in the Y199C mutant were due to the inability to bind to substrate DNAs, we determined the binding affinities, via the change in fluorescence anisotropy of labeled DNA, for MR NBD complexes (where the Rad50 nucleotide binding domain construct lacks the majority of the coiled-coil and Zn 2+ hook domains) in the absence of ATP/Mg 2+ . Under these conditions, the observed DNA binding is from Mre11 15 . For both exo-(dsDNA, Fig. 1e) and endonuclease (ssDNA, Fig. 1f) substrates, we observed similar K D s for wild type and Y199C MR NBD complexes: 2.0 and 2.5 μM, respectively for dsDNA and 0.31 and 0.35 μM, respectively for ssDNA. Thus, differences in nuclease function do not arise from the general inability to bind to substrate DNA. Y199C mutant affects Pf Mre11 protein structure and dynamics. To understand the structural effects of the Pf Mre11 Y199C mutant, we turned to solution state NMR spectroscopy. Because of the large size of the Mre11 construct used here (Mre11 nuclease and capping domains dimer, Mre11 ND , ~ 65 kDa), we utilized uniformly deuterated, side-chain methyl group 13 CH 3 -labeled samples and methyl-Transverse Relaxation Optimized SpectroscopY (TROSY) 35,36 . First, we compared the apo form (i.e., Mn 2+ -free and DNA-free) of Mre11 ND Y199C with wild type. The side-chain methyl group resonances in the two-dimensional (2D) methyl-TROSY 13 C, 1 H correlation spectra revealed that the vast majority of the peaks overlay for the two constructs (Fig. 2a), indicating that the global fold of Mre11 ND is unchanged in the mutant. The largest chemical shift perturbations (CSPs) were observed for Leu170 and Leu200 (Fig. 2b), which pack above and below Tyr199. These large peak movements were expected due to the altered chemical environment resulting from the loss of the aromatic residue in Y199C. Other large CSPs were observed for residues in the DNA recognition loop 5 (RL 5), catalytic motif IV, and near the Mre11 dimerization interface (Fig. 2b). There were smaller CSPs observed around RL 3 and 4, which are in a conserved basic patch involved in DNA binding 37 . Other small CSPs were observed in the capping domain which is ~ 46 Å away from the site of the mutation.  www.nature.com/scientificreports/ We have previously biochemically and structurally characterized two Pf Mre11 separation-of-nuclease function mutants: H52S and Y187C 34 . The NMR data collected on those mutants also showed CSPs for side-chain methyl groups around the catalytic motifs and dimerization interface. Since H52S, Y187C, and Y199C have similar biochemical (i.e., loss of exonuclease activity) and structural phenotypes, we questioned whether or not any changes in the structure were common among these mutants. An overlay of the methyl-TROSY spectra for wild type and the three mutants (Fig. 3a) showed that several resonances throughout Pf Mre11 ND move in a linear fashion due to mutations, yet the pattern of the peak movements were not the same for all methyl groups (e.g., 97LeuCD1 and 195LeuCD1 in Fig. 3a). Nevertheless, linear movement as observed here can be indicative of a shift in the population between two states. We therefore performed chemical shift covariance analysis (CHESCA), a method for clustering residues undergoing the same populated weighted average change in peak position because of a perturbation (here, the perturbation is mutation) 38,39 . For CHESCA, the pattern of the four peak positions for a given methyl group (i.e., the peak for the same methyl group in wild type, H52S, Y187C, and Y199C) is compared to the pattern observed for every other methyl group; methyl groups with high correlation (low distance in Fig. 3b) in peak movement patterns are then clustered together. These clusters often uncover allosteric networks within proteins 38,[40][41][42] . CHESCA of Pf Mre11 ND resonances with significant changes in peak position upon mutation resulted in four clusters of residues ( Fig. 3b and 3d). As expected, methyl groups within a cluster have peak movement patterns that are highly correlated with the peak movement patterns from other members of that cluster ( Fig. 3c) but not with those from methyl groups in other clusters. These four clusters map to a line of residues between active site motifs II and III and the dimerization interface (cluster 1), an area of www.nature.com/scientificreports/ Mre11 between Y187 and Y199, which encompasses regions of the protein that may recognize dsDNA ends (cluster 2), various residues throughout the structure (cluster 3), and the dimerization interface (cluster 4) (Fig. 3d). We next used the side-chain methyl group 13 CH 3 -labeled Pf Mre11 ND NMR samples to compare the wild type and mutant side-chain methyl group dynamics on the ps-ns timescale at 50 °C. These motions were characterized by measuring the cross-correlated relaxation rates (η) of methyl 1 H-1 H "forbidden" triple quantum coherences. η describes the ps-ns timescale amplitude of the side-chain methyl group motion and global tumbling 43 . Because there is no evidence that Y199C alters dimerization of Mre11, we interpreted changes in η rates as differences in side-chain motion and not changes in global tumbling. Figure 4a shows representative relaxation curves for individual side-chain methyl groups highlighting the differences observed in the η rates upon Y199C mutation. Figure 4b shows the differences in the η relaxation rates (Δη = η MUT − η WT ) mapped onto the crystal structure of Pf Mre11 ND . The data show that Y199C generally leads to Δη > 0 (average Δη rate was 27.2 s −1 ), which indicates decreased side-chain methyl group flexibility throughout the entire structure. Specifically, we observed the largest Δη values for methyl groups adjacent to the site of the mutation as well as in the capping domain and the basic patch, which mirror the changes observed above in the CSPs. Collectively, the NMR data show that the immediate area around the site of the mutation was predominantly affected by Y199C, as expected, and that perturbations to the conserved basic patch, a region suggested to be important for proper positioning of the exonuclease substrate, could be in part what is causing the lower exonuclease activity of the mutant.  1 H methyl TROSY NMR spectra for wild type (blue), H52S (red), Y187C (green), and Y199C (orange) Pf Mre11 ND . Peak assignments are given for several peaks that were considered for CHESCA. The colored bars correspond to the cluster where that peak is located (see (b) and (d)). (b) Dendrogram of methyl groups determined by CHESCA. A distance cut-off of 0.5 (horizontal black line) was used to define the clusters. Clusters 1, 2, 3, and 4 are colored red, green, blue, and purple, respectively. (c) Representative pairwise inter-residue correlation plots of the combined chemical shifts (δ methyl ) for each of the four clusters. The Pearson's correlation coefficient (R P ) is given in the lower right corner. (d) Structure of Pf Mre11 ND highlighting the side-chain methyl groups whose CSPs from mutation cluster according to CHESCA. Methyl groups not clustered are colored grey. DNA recognition loops (RL) and conserved catalytic motifs (I-V) are indicated. The location of the conserved basic patch is outlined by the red circle.  We also compared η rates across wild type, Y199C, and the two separation-of-nuclease function Pf Mre11 ND mutants (Fig. 4c). Here, we noted that several side-chain methyl groups in the Y199C mutant had a significant deviation from the η rates observed for wild type, H52S, and Y187C. When the side-chain methyl groups with deviant η rates were mapped onto the structure of Pf Mre11 ND (Fig. 4c, right), we observed a clustering of residues in the capping domain and linkers connecting the nuclease and capping domains-regions consistent with the proposed ssDNA binding site 33 . Therefore, differences between the endonuclease activities of the Y199C, which has impaired activity (Fig. 1), and H52S and Y187C, which have wild type-like activity, could be due to differences in the dynamics in this region. Moreover, comparing the pattern of methyl groups with deviant η rates to methyl groups that clustered in CHESCA (Fig. 3d), we noticed very few methyl groups are observed in both sets. This observation supports the notion, exemplified by separation-of-function mutations 33,34 and inhibitors 44 , that different regions within Mre11 are responsible for exo-and endonuclease functions. Furthermore, the observed structural and dynamic perturbations to side-chain methyl groups distal to the site of mutation (in the nuclease motif and dimerization interface) combined with the CHESCA clusters further substantiate a role for allostery within Mre11.
Y199C mutant affects Pf MRE11 DNA binding. We next added dsDNA and ssDNA to side-chain methyl group 13 CH 3 -labeled NMR samples of Y199C Mre11 ND to determine if the mutant has an effect on the structure of DNA-bound Mre11 ( Fig. 5a and b). To identify differences in the DNA-bound structures, we calculated the deviation of the CSPs between the DNA-bound forms of wild type and Y199C Mre11 ND (ΔCSP = |CSP WT-DNA − CSP Y199C-DNA |). Significant ΔCSPs (i.e., those greater than one standard deviation of the median ΔCSPs) were observed for both dsDNA-and ssDNA-bound forms of Mre11 ND and are mapped onto the crystal structure of Pf Mre11 ND in Fig. 5c (green and cyan spheres, respectively). In general, the number of ΔCSPs was similar for both dsDNA-and ssDNA-bound Mre11. We have previously shown that perturbations in methyl dynamics around a conserved basic patch were not correlated with a decrease in affinity for dsDNA, as measured by fluorescence polarization, but were instead associated with altered stability of the dsDNA substrate and in turn, a decrease in exonuclease activity 34 . A similar situation may be occurring in Pf Mre11 Y199C.
Since the Y199C mutant disrupts endonuclease activity, we compared the significant ΔCSPs observed here upon ssDNA binding with those previously observed for H52S and Y187C 34 . We reasoned that side-chain methyl groups that uniquely experienced a difference in ssDNA binding for the Y199C mutant could be related to the disruption of endonuclease activity. Interestingly, ~ 1/2 of the ssDNA-bound ΔCSPs observed for Y199C were unique (i.e., they were not observed in ssDNA-bound H52S or Y187C), and these side-chain methyl groups clustered within the structure to two main regions: above the site of the mutation where the largest CSPs are observed (e.g., Leu170 and Leu172; Fig. 2) and along the base of the central β-sheet. These residues may therefore offer the first clue into which regions of Mre11 ND are important for endonuclease activity.  www.nature.com/scientificreports/ stably integrated into the endogenous mre11 locus of S. cerevisiae. Western blots confirmed that mutant Mre11 expression was equal to wild type (Fig. 6a). First, we examined whether or not the analogous cancer-associated mutant was sensitive to the genotoxic agents CPT and MMS, a topoisomerase I inhibitor and a nucleobase alkylating agent, respectively. Figure 6b, top, shows that mre11-F233C repairs dsDNA breaks generated by CPT or MMS, as this yeast strain grew as efficiently as wild type and mre11-H125S on both the non-treated and drugtreated plates. We also performed survivability assays with phleomycin, which generates 'clean' DNA DSBs (i.e., without protein adducts), and as shown in Fig. 6b, bottom mre11-F233C repairs this form of DNA damage. To determine if other nucleases (e.g., Dna2 or Exo1) [45][46][47] were compensating for the nuclease deficiency of mre11-F233C in the DDR, we monitored the sensitivity to phleomycin in a sgs1Δ background (Fig. 6b, bottom). As previously shown, the nuclease inactive mre11-H125N demonstrated extreme sensitivity to phleomycin when Sgs1, a helicase associated with Dna2, is knocked out 47 , removing a 'back up' nuclease from the DDR. On the other hand, mre11-F233C again demonstrated wild type ability to survive on phleomycin even in the sgs1Δ background (Fig. 6b, bottom). Thus, the remaining nuclease activities of this mutant may be sufficient to maintain the DDR in the absence of Sgs1/Dna2. Though mre11-F233C was not sensitive to genotoxic agents, we were still interested in determining its effectiveness in each of the two major DNA DSB repair pathways. The ability to perform NHEJ was assessed by transforming yeast cells with a linearized plasmid containing a yeast origin of replication and complementary nutritional marker (HIS3) 48,49 . Restriction enzyme digestion was performed at a site lacking homology to the yeast genome; therefore, NHEJ is required to repair the DNA DSB and for the yeast to grow on histidine dropout media. mre11-F233C showed wild-type levels of NHEJ repair efficiency (Fig. 6c), as did mre11-H125S. Lastly, to probe for a defect in HR repair, the DNA end resection efficiency of mre11-F233C was assessed via a real time www.nature.com/scientificreports/ PCR-based assay 50,51 . Kinetic analysis showed mre11-F233C exhibited similar levels of end resection as both wild type and mre11-H125S at an HO endonuclease induced DNA DSB near the MAT locus (Fig. 6d). Thus, mre11-F233C mutated yeast are fully capable of performing NHEJ and HR, and these data are consistent with previous reports of Mre11 nuclease activity being dispensable for NHEJ and HR in yeast 52 .

Discussion
The MRE11 F237C mutation was found in an analysis of ~ 13,000 genes in 11 breast cancers. Additionally, analysis of MRE11 mutations cataloged in cBioportal 53,54 revealed another mutation at this position (F237L) found in breast invasive ductal carcinoma 55 . In both cases, the mutation at F237 would have been one of many other mutations in various other genes detected in these tumors 28,55 . Therefore, we obviously do not know if these mutations were necessary and/or sufficient to cause the disease. Our in vitro data does show that the analogous mutation in a model Mre11 (Pf Mre11 Y199C) has significantly decreased exo-and endonuclease activities and changes the chemical environment of the protein side-chains in both ss-and dsDNA binding regions. Yet, we and others have shown that this phenotype is not sufficient to alter the DNA DSB response in budding yeast 34,52 . Thus, if this spontaneous mutation was an underlying cause for carcinogenesis, it is likely due to the disruption of some other aspect of MRN function, protein-protein interactions, and/or regulation, which occur in humans but not lower organisms.
The data presented here are comparable to our recent study of two other Pf Mre11 separation-of-nuclease function mutations: Y187C and H52S 33,34 . Y187C and H52S were exonuclease inactive but endonuclease active, whereas Y199C showed significant reduction in both nuclease activities. NMR data suggests that, like Pf Mre11 Y187C and H52S, the Y199C mutant not only changed the structure of Mre11 in the immediate vicinity of the mutated residue but also in functional regions far from the site of the mutation and led to a dramatic overall decrease in side-chain methyl group motions within Mre11. In the case of Y187C and H52S, we determined that the separation-of-nuclease function phenotype is derived from an inability of Mre11 to stabilize the dsDNA exonuclease substrate, and since Mre11 binds to ssDNA and dsDNA differently, exonuclease activity was affected, whereas endonuclease activity was not 34 . Because of the similarity in the NMR data, we propose that a similar phenomenon is occurring in Y199C, though perhaps to a lesser extent as some exonuclease activity is retained in this mutant.
Combining our data here with our previous NMR results for Y187C and H52S allowed us to perform CHESCA (Fig. 3) and uncovered a network of residues in Mre11 ND that couples the regions important for DNA binding, and possibly DNA end recognition, with the active site and dimerization interface. X-ray crystal structures and methyl group CSPs from our previous NMR experiments on various Mre11 ND -DNA complexes have both shown that DNA binding alters the environment of the dimerization interface 34,37 . CHESCA identified a specific set of residues through the core of the Mre11 nuclease domain that might relay DNA binding information to the dimerization motif. We have also examined the effect of these mutants on dsDNA binding (Fig. 5). Interestingly, all of the residues in cluster 1 (red spheres in Fig. 3d) displayed significant deviations from wild type in dsDNA binding (i.e., significant ΔCSPs) in at least two of the three exonuclease deficient mutants we have characterized. This result strengthens the CHESCA output and implies that this network of residues is indeed important for exonuclease activity. Thus, we predict that the mutation of residues within this cluster may also lead to an exonuclease inactive phenotype.
Finally, Y199C disrupts endonuclease activity, whereas the Y187C and H52S mutants were wild type-like. NMR data for Y199C showed some changes within the structure as a result of ssDNA binding which not seen in the other mutants. The residues involved in these changes are unique from those observed upon dsDNA binding and in the CHESCA clusters implying that different networks might exist within Mre11 for relaying ss-versus dsDNA binding information. In support of this, we note that Y199C displayed significant rigidification in the capping domain that was not observed in Y187C or H52S. In conclusion, we have shown through the characterization of nuclease deficient mutants that discrete changes in Mre11 structure and dynamics are critical for exo-and endonuclease activity and likely underlie the separation of the nuclease functions.

Materials and methods
Protein expression and purification. Mre11 point mutations were introduced into the expression plasmid using a modified Quikchange approach (Stratagene). Full-length Pf Mre11 and truncated Pf Mre11 ND (amino acids 1-342 comprising the nuclease and capping domains) were expressed by transforming the appropriate plasmid into E. coli BL21(DE3) C41 cells (Sigma). For unlabeled protein, cells were grown in LB media, and expression was induced with 1 mM IPTG at 37 °C for 4 h. To prepare U-2 H, Ileδ1-[ 13 CH 3 ], Leuδ/Valγ-[ 13 CH 3 , 12 CD 3 ], Metε-[ 13 CH 3 ]-labeled Pf Mre11 ND samples for NMR spectroscopy, cells were grown in deuterated 2 × M9 minimal media 56 with 1 g/L 14 NH 4 Cl and 3 g/L 2 H, 12 C-glucose as the sole nitrogen and carbon sources. 100 mg of each 13 CH 3 -labeled Ile, Leu, Val, and Met metabolic precursors (Cambridge Isotope Laboratories, Inc) 57 were added to the media 45 min prior to induction with 1 mM IPTG. Isotopically labeled Mre11 ND was then overexpressed for ~ 16 h at 37 °C. In both cases, purification proceeded as previously described 34 .
Full-length and nucleotide binding domain (NBD) constructs of Pf Rad50 were expressed and purified as described in Boswell et al. 58 . Mre11-Rad50 complexes were made by mixing equimolar ratios of full-length Mre11 and Rad50, heating at 60 °C for 15 min, and cooling to room temperature. www.nature.com/scientificreports/ 30 µL exonuclease reaction was set up in a 1.5 mL microcentrifuge tube and contained 0.5 μM MR and 1 μM Exo11 dsDNA substrate in Reaction Buffer (50 mM HEPES, 150 mM NaCl, 0.1% PEG-6000, 2.5% glycerol, 1 mM TCEP, pH 7.0). Samples were made without co-factor, with 1 mM MnCl 2 , or with 1 mM MnCl 2 , 5 mM MgCl 2 , 1 mM ATP. After a 30 min incubation at 60 °C, the reactions were removed from the heat block, which stops the nuclease reaction, and 25 μL were transferred to a black 385-well plate. The 2-AP fluorescence resulting from Mre11 exonuclease activity was measured using a BioTek Synergy Neo2 multimode plate reader as previously described 34 .
Mre11 endonuclease assay. A Cy3/BHQ2 fluorescence-based quantitative MR endonuclease assay was performed using a 17-nucleotide ssDNA substrate containing a Cy3 fluorophore at the 5′-end and a BHQ2 quencher at the 3′-end (5′-Cy3-TCT CTA GCA GTG GCGCC-BHQ2-3′; IDT). Note, Cy3 and BHQ2 are close enough on the intact ssDNA substrate that resonance energy transfer between Cy3 and BHQ2 results in low fluorescence signal. The experimental conditions and protocol were the same as described above for the 2-AP exonuclease assay except that 0.2 μM Cy3/BHQ-2 ssDNA was used as the substrate. Cy3 fluorescence resulting from endonuclease-cleaved substrate was measured as described previously 34 . Qualitative endonuclease reactions on agarose gels were performed using 1 μg of ΦX174 single stranded virion circular DNA (New England Biolabs) as the substrate. 30 μL reactions containing 0.5 μM MR in Reaction Buffer were set up in 1.5 mL microcentrifuge tubes. Reactions were incubated at 60 °C without co-factor, with 1 mM MnCl 2 , or with 1 mM MnCl 2 , 5 mM MgCl 2 , 1 mM ATP. 6 µL time points were removed from the tube after 10, 20, and 30 min and quenched with 1% w/v SDS, 10 mM EDTA, and 0.5 mg/mL Proteinase K. Cleaved products were resolved on an 0.8% agarose gel, and visualized as previously described 34  To analyze the effect of mutation or DNA binding, a weighted chemical shift perturbation (CSP) was calculated from the 1 H (δ H ) and 13 C (δ C ) chemical shifts according to where δ i is the chemical shift value for wild type or DNA-free, δ j is the chemical shift value for mutant or DNAbound, and σ H and σ C are the standard deviations for methyl group specific δ H and δ C from the Biological Magnetic Resonance Data Bank.
Methyl group cross-correlated relaxation rates (η) were measured as described 43 . Pairs of triple ("forbidden") and single ("allowed") quantum data sets were collected in an interleaved manner for relaxation delay time (T) points of 2, 4, 6, 8, 10, 15, 20 and 25 ms. Ratio of peak intensities (Int forbid /Int allow ) were fit to where C is 0.75 and δ accounts for relaxation from external protons 43 . Errors in peak intensities were set to the noise of the spectra, and the reported errors in η are from the covariance of the fit.

Methyl chemical shift covariance analysis.
To perform the chemical shift covariance analysis (CHESCA) 38 , a weighted combined methyl chemical shift value (δ methyl ) was calculated from the 1 H (δ H ) and 13 C (δ C ) chemical shifts as Int forbid Int allow = C(ηtanh( η 2 + δ 2 T)) η 2 + δ 2 − δtanh( η 2 + δ 2 T) www.nature.com/scientificreports/ where σ H and σ C are the standard deviations for methyl group specific δ H and δ C from the Biological Magnetic Resonance Data Bank. Side-chain methyl groups were selected for CHESCA using a median-absolute-deviation (MAD) approach 61 , which is used to find outliers from a distribution, with a cut-off of modified Z-score > 0.5. CHESCA was then performed on methyl groups with significant CSPs as a function of mutation with an inhouse written python script (available upon request) that used the scientific python (scipy) library as previously described 42 . Briefly, unit-less inter-residue correlation distances 62,63 were calculated between each methyl group using the set of δ methyl values for the four Mre11 ND constructs. Correlation distances were then clustered using a complete linkage algorithm 64 . Finally, the four clusters were derived from a distance cut-off of 0.5 which was empirically determined based on the dendrogram in Fig. 3b.
Yeast assays. C-terminal 13 × myc-tagged mre11 mutants were constructed in the integrated GAL10-HO cassette W303 S. cerevisiae strain (MATa leu2::GAL-HO-LEU2 hml ∆hmr ∆RAD5). The kanMX gene was cloned from pFA6a-kanMX vector (Addgene plasmid number 39296) 65 and juxtaposed to the 3′ end of the mre11 gene for geneticin (G418 di-sulfate salt from Sigma) antibiotic selectivity. The mre11∆ strain was prepared by replacing the mre11 open reading frame with the kanMX gene. The sgs1Δ strain was prepared by replacing the sgs1 open reading frame with the trp1 gene, which was cloned from the pRS414 vector (ATCC). Verification of the integrated genes was accomplished by PCR (mre11 and sgs1) and western blot (mre11). The genotoxin sensitivity assay, NHEJ repair assay 48,49 , and real time PCR-based end resection assay 50,51 were all performed as previously described 34 .
Statistical analysis. Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software Inc., San Diego, CA, USA). All data are presented as the mean and standard deviation (SD) of at least three replicates.
As the exonuclease assay has only one variable, we used a one-way ANOVA with Dunnett correction. For all statistical tests, p < 0.01 was considered statistically significant.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.