The MRN complex and topoisomerase IIIa–RMI1/2 synchronize DNA resection motor proteins

DNA resection—the nucleolytic processing of broken DNA ends—is the first step of homologous recombination. Resection is catalyzed by the resectosome, a multienzyme complex that includes bloom syndrome helicase (BLM), DNA2 or exonuclease 1 nucleases, and additional DNA-binding proteins. Although the molecular players have been known for over a decade, how the individual proteins work together to regulate DNA resection remains unknown. Using single-molecule imaging, we characterized the roles of the MRE11–RAD50–NBS1 complex (MRN) and topoisomerase IIIa (TOP3A)–RMI1/2 during long-range DNA resection. BLM partners with TOP3A–RMI1/2 to form the BTRR (BLM–TOP3A–RMI1/2) complex (or BLM dissolvasome). We determined that TOP3A–RMI1/2 aids BLM in initiating DNA unwinding, and along with MRN, stimulates DNA2-mediated resection. Furthermore, we found that MRN promotes the association between BTRR and DNA and synchronizes BLM and DNA2 translocation to prevent BLM from pausing during resection. Together, this work provides direct observation of how MRN and DNA2 harness the BTRR complex to resect DNA efficiently and how TOP3A–RMI1/2 regulates the helicase activity of BLM to promote efficient DNA repair.

Here, we use single-molecule fluorescence imaging to decipher the functions of individual resectosome components during DNA resection. Both MRN and TOP3A-RMI1/2 help BLM to initiate DNA unwinding. MRN and TOP3A-RMI1/2 also stimulate DNA2-mediated resection. Finally, MRN synchronizes the translocation speeds of BLM and DNA2 to prevent BLM pausing. We reveal that MRN and TOP3A-RMI1/2 are regulatory resectosome components that initiate DNA resection and synchronize the individual motors during kilobase-long DNA processing.

MRN prevents BTRR dissociation from DNA
Having established that TOP3A-RMI1/2 helps initiate DNA unwinding, we tested whether MRN further stimulates the BTRR complex. MRN is important for BLM recruitment to DSB and has been shown to stimulate unwinding activity (30,60,61). As expected, MRN colocalizes with BTRR and moves together with BTRR during DNA unwinding (Fig. 4A).
To better understand the effect of MRN on BLM activity, we repeated the helicase assays with MR and MRE11 subunits (Fig. 4, B-E). MR was sufficient to initiate the helicase activity of BLM but decreased processivity. MRE11 alone did not stimulate BLM initiation, suggesting the RAD50 subunit is critical for regulating the activity of BLM. To test whether the ATPase of MRN is also required to stimulate BLM, we repeated helicase experiments with the ATPase-deficient MR(S1202R)N (62,63). Interestingly, MR(S1202R)N decreased both BLM helicase initiation and processivity. In vitro pulldown experiments showed that BLM interacts with both NBS1 and RAD50; MRE11 did not interact with BLM (Fig. S3A). MRN and BLM may also interact indirectly via RPA and/or TRR (64). We conclude that the ATPase of MRN activity is important for stimulating BLM, possibly by initial DNA unwinding and/or promoting DNA tethering at the ssDNA/dsDNA junction.

MRN synchronizes the BLM and DNA2 motors
The unwinding rate of DNA2 is approximately three-fold slower than that of BLM in the presence of RPA (Fig. 2F). The yeast homolog of BLM, Sgs1, and Dna2 also show approximately two-fold difference in unwinding rates (65). This difference in unwinding rates of BLM and DNA2 can lead to discoordination between the two motors. Consistent with this notion, we observed that 45% of the BTRR-DNA2 resectosomes paused for >30 s during DNA resection (n = 42/94) with 12% (n = 5/42) of these complexes pausing two or more times during their resection trajectories (Fig. 4, F and G). The change in resection velocity after the pause was heterogeneous and did not correlate with the prepause velocity (Fig. S3E). Adding MRN suppressed these pauses (83% of resectosomes did not pause; n = 68/82). MRN also suppresses pauses with the minimal MRN-BLM-DNA2 assembly (83% did not pause; n = 25/30). MR and MRE11 did not stimulate resection and suppress pausing (Fig. 4, G-I). Interestingly, MR(S1202R)N decreased processivity, velocity, and could not suppress pausing (Fig. 4, G-I). Thus, the ATP-dependent activities of MRN, along with BLM and DNA2, are also required to promote efficient DNA resection, possibly by stimulating the engagement of BLM with the ss/ds junction. We also tested whether pausing was sequence or GC content specific. The GC content of λ-DNA is greater on the cosL side than on the cosR side of our DNA substrate (Fig. S3B). Therefore, we assayed resection from the GC-rich (cosL) end or GC-poor (cosR) ends. The pausing frequency was similar on both substrates, suggesting that pausing is not strongly sequence or GC content dependent but instead may be caused by the accumulation of ssDNA (Fig. S3, C and D). We conclude that MRN coordinates BLM and DNA2 to stimulate efficient DNA resection.

Discussion
DNA resection is catalyzed by either the EXO1 or DNA2 nucleases. Although EXO1 may be the predominant resection nuclease in human cells (48,(66)(67)(68), DNA2 is better at processing apurinic/apyrimidinic sites and 8-oxoguanines (69). Furthermore, a super-resolution imaging study found comparable recruitment of both DNA2 and EXO1 at induced DSBs, suggesting that both nucleases are required during DNA resection (61). We had previously shown that MRN and BLM act as processivity factors for EXO1 in the presence of RPA (28,50). We expand on the earlier study to show that TOP3A-RMI1/2 does not stimulate BLM-EXO1 resection. Instead, TOP3A-RMI1/2, in combination with MRN, stimulate DNA2-mediated resection. In addition, MRN plays a scaffolding role by assembling the resectosome at a DSB and suppressing the dissociation of BTRR from DNA. MRN also prevents pausing by coordinating BLM and DNA2 during DNA resection (Fig. 5).
BLM can either processively unwind DNA or strand switch between the Watson or Crick ssDNA strands (50, 70-72). To promote processive movement and suppress strand-switching, BLM must engage the ss/dsDNA junction rather than the ssDNA formed during unwinding (73,74). MRN recognizes both dsDNA and ssDNA/dsDNA junctions and makes protein-protein interactions with BLM through both the RAD50 and NBS1 subunits (Fig. S3A). We propose that these physical interactions with BLM increase the processivity of BLM and suppress pausing by anchoring BLM to the ss/ds DNA junction. Similarly, the TRR complex also binds free DNA ends and interacts with BLM (34). Thus, TRR may also prevent BLM from engaging partially unwound DNA to promote processive helicase activity.
Here, we show that DNA resection requires the concerted activity of motors of both BLM and DNA2. However, the helicase activity of DNA2 is only detectable when its nuclease is disabled in vitro (54,56). How is DNA2 then able to use its helicase activity within the resectosome? One possibility is that a physical interaction between BLM and DNA2 stimulates the helicase of DNA2 and/or suppresses its nuclease activity. BLM and DNA2 physically interact and colocalize at DNA ends in vitro and in cells (30,50,61,75). WRN also interacts with DNA2 via its helicase domain, which is conserved in BLM (45). Future studies will be required to further map this interaction and its significance for regulating both motors. BLM may also promote the helicase of DNA2 by providing a long 5 0 -ssDNA overhang that engages that domain. The structure of mouse DNA2 reveals that ssDNA threads through a narrow protein channel that requires >10 nt to interact with both the nuclease and helicase domains. In the absence of such ssDNA, the nuclease activity may degrade ssDNA that cannot thread into the helicase domain (55). However, the unwinding of the long ssDNA flap by BLM may overcome the inhibition by the nuclease domain to engage the helicase and provide stimulation of both motors. This is consistent with our results and previous in vitro and in vivo data that showed that hd BLM and its yeast homolog, Sgs1, downregulate DNA resection (19,20,30,32,50).
Pausing during DNA resection has been observed with the bacterial RecBCD nuclease-helicase complex following recognition of the recombination hot spot sequence χ (crossover hot spot instigator-Chi) (76)(77)(78). RecBCD is functionally reminiscent of BTRR-DNA2 because it also encodes a fast (RecD) and slow (RecB) motor of opposite polarity. Two motors that move along opposite DNA strands with different speeds will generate a long ssDNA loop between them. Such loops are generated by the Escherichia coli RecBCD helicasenuclease because of differences in the translocation rates between the RecD and RecB motors (79). The molecular origin of RecBCD pausing stems from the slower motor "catching up" with the faster motor because of a conformational switch after χ recognition.
The underlying reason for pausing by BTRR-DNA2 is unknown, but it is unlikely to depend on a χ-like DNA sequence as sequence-dependent resection regulation has not been observed in yeast and human resectosomes. We conjecture that BTRR may unwind DNA in front of DNA2, leading to a growing ssDNA loop that ultimately pauses the entire complex. In this model, MRN prevents the accumulation of such ssDNA loops by synchronizing BLM and DNA2 helicase velocities. In support of this hypothesis, a recent singlemolecule study showed that BLM retains contact with ssDNA as it unwinds dsDNA (75). Additional high-resolution electron microscopy and other biochemical studies will be required to directly image an ssDNA loop between BLM and DNA2.
MRE11 and BLM colocalize early at DNA breaks immediately following damage in cells (61). This is consistent with our results showing that MRN assembles the resectosome at a DNA break. However, MRE11 and BLM do not associate as closely in the later stages of DNA resection (61). The role of MRN may thus be critical in the initiation and early coordination of BLM and DNA2. In the later stages of DNA resection, pausing may act as a negative DNA resection signal. Such pauses slow resection, possibly limiting over-resection and giving RAD51 sufficient time to complete the homology search. Together, this work shows that its conserved accessory factors regulate the helicase activity of BLM and that coordination with MRN and DNA2 stimulates DNA resection and, ultimately, efficient HR.

Single-molecule fluorescence microscopy
All single-molecule data were collected on a Nikon Ti-E microscope in a prism-total internal reflection fluorescence configuration equipped with a prior H117 motorized stage. Flow cells were loaded into a custom-designed stage insert incorporating a chip mount, fluidic interface, and heating element (49). All experiments were maintained at 37 C by a combination of an objective heater (Bioptechs) and a custombuilt stage-mounted heating block. The flow cell was illuminated with a 488 nm laser (Coherent) through a quartz prism (Tower Optical Co). Data were collected with a 200 ms exposure, 2 s shutter (Vincent Associates) resulting in 1800 frames in 1 h, through a 60× water-immersion objective (1.2 numerical aperture; Nikon), a 500 nm long pass (Chroma), and a 638 nm dichroic beam splitter (Chroma), which allowed twochannel detection through two EMCCD cameras (Andor iXon DU897; cooled to −80 C). Images were collected using Nikon NIS-Elements software and saved in an uncompressed TIFF file format for later analysis (see later).
In our imaging buffer (40 mM Tris [pH 8.0], 60 mM NaCl, 200 μg/ml bovine serum albumin [BSA], 2 mM DTT, 2 mM MgCl 2 , and 1 mM ATP), we typically observe intermittent fluorescent emission (blinking), which is an intrinsic property of single QDs (84). These blinking events indicate that our fluorescent trajectories are from an individual QD because of the unlikely situations of two QDs blinking simultaneously.

Particle tracking analysis
The image stacks collected from the EMCCD cameras were exported as full-resolution TIFF stacks. To correct for XYstage sample drift, a stationary particle on the flow cell surface was picked, and its position was tracked by fitting the point-spread function to a 2D-Gaussian using a customwritten ImageJ script (available at: https://github.com/ finkelsteinlab/single-particle-tracking-scripts). XY drift was then subtracted from all resectosome complexes during postprocessing. For each frame, the point spread function of DNA-bound proteins was fit to a 2D Gaussian to obtain (x, y) coordinates with subpixel resolution. We ensured that resectosome components were DNA bound by briefly stopping buffer flow at the beginning of each experiment. Stopping buffer flow recoils the DNA and all associated proteins to the diffusion barrier, providing a useful control that these are not surface-tethered particles. Only DNA-bound particles were included in all subsequent analyses. We did not attempt to analyze the trajectories of particles that moved less than 1 kb, which is approaching the resolution of the DNA curtain assay under the buffer flows used here (500 bp).
Particle trajectories were analyzed in MATLAB R2018aversion (MathWorks). For individual moving particles, the processivity was determined by measuring the distance traveled along DNA, and velocity was determined by fitting the time-dependent position along DNA to a line. To determine DNA binding lifetimes, we measured the time each molecule was bound to DNA. The survival probabilities were fit to a single exponential decay in MATLAB. Particles from at least two flow cells were pooled for the final analysis. Statistical significance was determined via the Student's t test.
Fluorescent protein labeling 3xHA-BLM (40 nM) were conjugated to QDs preincubated with a rabbit anti-HA antibody (ICL Lab) on ice for 10 min in 20 μl. Next, BLM was incubated with the anti-HA QDs at a ratio of 1:2 for an additional 10 min on ice, diluted with imaging buffer (40 mM Tris [pH 8.0], 60 mM NaCl, 200 μg/ml BSA, 2 mM DTT, 2 mM MgCl 2 , and 1 mM ATP) to 200 μl and injected into the flow cell. FLAG-TOP3A-RMI1/2 (80 nM) or FLAG-DNA2 (40 nM) were labeled with QDs preincubated with a mouse anti-FLAG antibody (Sigma-Aldrich) on ice for 10 min prior to injection. In addition, biotin-MRN (2 nM) was labeled via streptavidin QDs. Saturating biotin was added to the protein-QD conjugates to bind free streptavidin sites prior to injection.

Quantification and statistical analysis
For Figures 1-4, n represents the number of molecules. Quantification and statistical analyses were done using MAT-LAB (version: R2018a). Fluorescent particles were tracked using an in-house ImageJ script (available at https://github.com/ finkelsteinlab/single-particle-tracking-scripts) where the positions of individual molecules on DNA were determined by fitting the point spread function to a 2D Gaussian. Trajectories were used to calculate the velocity and processivity for BLM and DNA resection complexes. Statistical details of experiments can be found in the Results section and figure legends where indicated.

Contact for reagent and resource sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ilya Finkelstein (ilya@finkelsteinlab.org).
Supporting information-This article contains supporting information.