Hop2-Mnd1 and Swi5-Sfr1 stimulate Dmc1 filament assembly using distinct mechanisms

Abstract

In meiosis, Dmc1 recombinase and the general recombinase Rad51 are responsible for pairing homologous chromosomes and exchanging strands. Fission yeast (Schizosaccharomyces pombe) Swi5-Sfr1 and Hop2-Mnd1 stimulate Dmc1-driven recombination, but the stimulation mechanism is unclear. Using single-molecule fluorescence resonance energy transfer (smFRET) and tethered particle motion (TPM) experiments, we showed that Hop2-Mnd1 and Swi5-Sfr1 individually enhance Dmc1 filament assembly on single-stranded DNA (ssDNA) and adding both proteins together allows further stimulation. FRET analysis showed that Hop2-Mnd1 enhances the binding rate of Dmc1 while Swi5-Sfr1 specifically reduces the dissociation rate during the nucleation, about 2-fold. In the presence of Hop2-Mnd1, the nucleation time of Dmc1 filaments shortens, and doubling the ss/double-stranded DNA (ss/dsDNA) junctions of DNA substrates reduces the nucleation times in half. Order of addition experiments confirmed that Hop2-Mnd1 binds on DNA to recruit and stimulate Dmc1 nucleation at the ss/dsDNA junction. Our studies directly support the molecular basis of how Hop2-Mnd1 and Swi5-Sfr1 act on different steps during the Dmc1 filament assembly. DNA binding of these accessory proteins and nucleation preferences of recombinases thus dictate how their regulation can take place.

Graphical Abstract

Graphical Abstract.

INTRODUCTION

Homologous recombination (HR) plays multiple critical roles during meiosis. During meiotic prophase, HR promotes the pairing of homologous chromosomes to establish crossovers essential for the accurate segregation between homologous chromosomes during the first meiotic division (1). Most eukaryotes carry two RecA-like recombinases responsible for homologous recombination: Rad51 and Dmc1. While Rad51 is ubiquitous, produced in both vegetative and meiotic cells, Dmc1 is meiosis-specific (2).

Meiotic HR is induced by the programmed formation of DNA double-strand breaks (DSBs) (3). The DSB ends become resected so that 3′-ended single-stranded DNA (ssDNA) is produced (4,5). Both Rad51 and Dmc1 preferentially bind such ssDNA to form a helical nucleoprotein filament structure called the presynaptic filament (2,6,7). In this context, these recombinases actively survey intact double-stranded DNA (dsDNA) to identify homology with the ssDNA they bound. Once homology is identified, ssDNA in the presynaptic filament pairs with its complementary strand of the homologous duplex DNA, leaving the other noncomplementary strand displaced. This early recombination intermediate is called a displacement loop (D-loop). Several meiosis-specific processes exist so that meiotic DSBs induce inter-homolog recombination, eventually producing crossovers (8). While both RecA homologs are essential for successful meiosis, Dmc1 functions predominantly in meiotic HR (9). The mechanism underlying the difference between Rad51 and Dmc1 still remains elusive.

Dmc1, like Rad51, can only function efficiently with several accessory factors (10). The Swi5-Sfr1 complex is an obligate dimer highly conserved from yeast to humans (11–13). Fission yeast and mammalian Swi5-Sfr1 stabilize the presynaptic filaments and stimulate strand exchange of Rad51 and Dmc1 in fission yeast (14–17). Swi5-Sfr1 also facilitates the displacement of replication protein A (RPA) bound to nascent ssDNA, allowing efficient assembly of Rad51 and Dmc1 (16,18,19). Swi5-Sfr1 stimulates ATP-hydrolysis of Rad51 and Dmc1 (16,20,21). Single-molecule analysis revealed that mouse and fission yeast Swi5-Sfr1 stimulate Rad51 filament assembly by preventing Rad51 dissociation (17,22). Interestingly, Mei5-Sae3, the budding yeast counterpart of Swi5-Sfr1, is meiosis-specific and seemingly functions exclusively with Dmc1 (13,23). Meiotic HR becomes defective without Mei5-Sae3, and the phenotypes are indistinguishable from those in the dmc1 null mutant. Meiotic localization of Dmc1 is completely lost without Mei5-Sae3, arguing for its critical role in Dmc1 filament formation and stabilization.

The Hop2-Mnd1 complex is another major accessory factor for Dmc1. The absence of Hop2-Mnd1 disrupts meiosis because meiotic recombination becomes defective (24–32). The budding yeast hop2/mnd1 mutants show aberrant accumulation of both Rad51 and Dmc1, with non-homologous chromosomes associated during prophase I (26,33). Budding and fission yeast Hop2-Mnd1 specifically stimulate Dmc1-driven D-loop formation and strand exchange, while its mammalian counterpart can stimulate Rad51 as well (28,34–38). Mammalian Hop2-Mnd1 stabilizes Rad51 and Dmc1 presynaptic filaments (36–38). Hop2-Mnd1 binds DNA with a much stronger preference for dsDNA over ssDNA (35,39,40), with Hop2-Mnd1 facilitating dsDNA binding of presynaptic filaments (36,37). Structural analysis revealed that Flagellate Giardia lamblia Hop2-Mnd1 takes a curved rod-like structure; one end is associated with an α-helical bundle that likely interacts with the Dmc1 presynaptic filament, while the other winged-helix domains bind dsDNA (39). Taken together, it is plausible that Hop2-Mnd1 facilitates the capture of dsDNA by presynaptic filaments.

Swi5-Sfr1 and Hop2-Mnd1 make unique and synergistic contributions to Dmc1-driven strand exchange, but their molecular details have not been well understood (41,42). In this report, we set out to explore how fission yeast Hop2-Mnd1 and Swi5-Sfr1 modulate Dmc1 filament assembly. Using single-molecule experiments, we showed that Swi5-Sfr1 stimulates Dmc1 assembly by reducing its dissociation rate. Surprisingly, Hop2-Mnd1 stimulates Dmc1 differently by increasing Dmc1’s binding rate on ssDNA. This stimulation arises from the Hop2-Mnd1 binding to DNA first to promote Dmc1 association at the ss/dsDNA junction. Both complexes can act together to establish the stable Dmc1 filament. Our results provide the molecular basis for how these two accessory complexes modulate the Dmc1 nucleoprotein filament assembly using distinct mechanisms.

MATERIALS AND METHODS

Protein purification

Schizosaccharomyces pombe Dmc1, Hop2-Mnd1, Swi5-Sfr1 and Rad51 were purified as previously described, with full details in the supplemental information (42). In brief, all proteins were induced in E. coli using the T7 promoter system at 18°C for 12 h. In Dmc1, ammonium-sulfate precipitate (35%) was resuspended, passed through SP Sepharose and the flow-through was serially processed through Q Sepharose, HiTrap Heparin, Superdex 200 PG 16/60, and Resource Q. Hop2 and Mnd1 were co-expressed and enriched with nickel-immobilized resin (elution at 200 mM, Roche). After removal of His6 with PreScission (GE Healthcare) protease, the sample was serially processed through HiTrap Heparin and Resource S. Sfr1 and Swi5 were co-expressed and the supernatant of the polyethyleneimine precipitation (0.05%) was precipitated with ammonium sulfate (35%). The resuspended pellet was dialyzed and serially processed through SP Sepharose, HiTrap Heparin, and Superdex 200 PG 16/60. In Rad51, ammonium-sulfate precipitate (35%) was resuspended, passed through SP Sepharose and the flow-through was serially processed through Q Sepharose, HiTrap Heparin, and Resource Q. An SDS gel of purified proteins is shown in Supplemental Figure S1.

smFRET assembly experiment

The dT₁₃₊₁₀ and dT₁₃₊₄₇ DNA substrates used in FRET experiments were prepared by annealing oligos purchased from IDT, with sequence and labeling positions listed in supplemental materials (Supplemental Table S1, Supplemental Figure S2; dT₁₃₊₁₀: F101 and F127; dT₁₃₊₄₇: F101 and F128). The oligos were annealed in an annealing buffer containing 20 mM Tris and 50 mM sodium chloride at pH 8.0. The slides used in smFRET experiments were prepared and assembled as described previously (22). Briefly, slides were PEG-modified containing a small fraction of biotin-labeled PEG and were then incubated with 0.02 mg/ml streptavidin for 5 min. Next, the reaction chamber was washed with buffer containing 0.02 mg/ml BSA and 1× T50 buffer (20 mM Tris, 50 mM NaCl at pH ∼7.7) for 5 min, followed by the addition of 20 pM DNA substrate in buffer containing 0.02 mg/ml BSA and 1× T50 buffer. The typical chamber volume in our setup is ∼20 μl. During the single-molecule experiments, the introduction of components and buffer into the reaction chamber facilitates buffer exchange, effectively removing any previously unbound components from the DNA molecules tethered to the surface. To ensure proper buffer exchange, we performed 2–3 washes of the reaction chamber using the reaction buffer. This is followed by the addition of the final reaction mix containing the specified proteins. Fluorescence images were acquired in the presence of imaging buffer (30 mM Tris–HCl, 120 mM NaCl, 3.5 mM MgCl₂, 2 mM Trolox, 1 mM DTT, 2 mg/ml BSA, 4 mg/ml glucose, 30 U/ml glucose oxidase, 30 U/ml catalase at pH 7.5). Unless specified otherwise, the reaction mixture included 3 μM Dmc1, 2 mM ATP, and given concentrations of accessory proteins, which were pre-incubated at 37°C for 5 min before the addition.

The objective-type total internal reflection fluorescence microscope (Olympus IX71) was coupled with a dual-view system for imaging using an EMCCD camera (iXon Ultra 897, Andor) at 20 Hz. A 532-nm laser (Ventus, 32 mW) was used for the excitation of Cy3, and a 638-nm laser (Omicron, 36 mW) for the excitation of Cy5. Raw fluorescence images were recorded by Glimpse, a generous gift from Jeff Gelles's lab (https://github.com/gelles-brandeis/Glimpse), and the colocalization of dye pairs in Cy3 and Cy5 channels were scored by a program written in IDL. We typically illuminate one field of view using the green laser for 120 seconds, finished with a brief red laser excitation to confirm the presence of Cy5 dyes.

The FRET values of the colocalized dye-pair DNA molecules were analyzed by a program written in MATLAB (MathWorks) (22). For time traces exhibiting alternation among several FRET states, variational Bayesian analysis (vbFRET) was used to get the binding and dissociation dwell times (43). FRET efficiency time traces were calculated from the raw donor and acceptor intensities, and then they were smoothed by a 15-frame moving average filter. A custom script was modified from vbFRET_no_gui.m of the vbFRET source code (https://sourceforge.net/projects/vbfret/) to analyze the FRET time traces. The prior hyperparameters, the optimization threshold, the maximum number of iterations, and the number of restarts were left as the default values. The script concatenated all the traces in a dataset. To serve as spacers, sequences of 300-frame Gaussian random values centered at −0.3 and having a standard deviation of 0.01 were inserted between the traces. vbFRET was performed on the concatenated time trace to infer the underlying hidden Markov model (HMM). To describe the time traces with a 4-state model, the number of states to fit was set to 5 to include the inserted spacers. The posterior was then used to infer the most probable state sequence of the concatenated trace. The original time traces with fitted state sequences were recovered by removing the negative spacers. The dwell times of transitioning from one state to another were collected from the state sequences and analyzed to calculate the transition rates. In these time-trace analyses, more than 55 molecules with >1600 transitions collected from at least three independent experiments were used for each condition.

Transition density plots (TDPs) were produced by analyzing the state sequences. The FRET time series was segmented using the state sequence. The FRET value of each segment was calculated from the median value, excluding the first and the last frames in the segment to avoid getting a data point that was acquired during the transition. Data points used in the TDP were then pairs of FRET values of segments before and after each transition. To filter out self-transitions, transitions with a FRET value change lower than 0.1 were excluded. 2D histograms were then plotted on the range [0, 1], binning into 55 bins on each axis.

TPM dT gapped DNA substrate preparation

DNA substrates for TPM experiments containing ssDNA gaps were prepared as previously described (22). The duplex handles were first prepared by auto-sticky PCR reactions using primers containing abasic sites to create a 351/331 or 151/131 ssDNA/dsDNA hybrid handle (44). polydT single-stranded oligos, purchased from IDT, were treated with T4 polynucleotide kinase (NEB) to generate a 5′-phosphate-modified end. Primers and DNA substrates are shown in supplemental materials (Supplemental Table S1, Supplemental Figure S2). The handles and dTn oligos were then ligated using T4 DNA ligase (NEB) at 37°C for 48 hours, followed by gel purification (QIAGEN). Before the experiments, excess of the biotin-labeled oligos (P77, Supplemental Table S1) for streptavidin-labeled bead attachment and junction oligos (for the 45 + 45 substrate, P265, Supplemental Table S1) was annealed to the DNA substrates at room temperature for 30 min to get the final substrates.

TPM filament assembly experiment

The preparation of 220 nm streptavidin-coated polystyrene beads (Bangs Lab.) was described previously (45). In filament assembly experiments, a glass slide was first incubated with 10 μg/ml anti-digoxigenin for 10 min and then with 2 mg/ml BSA for 30 min to prevent non-specific binding. 4 nM dT gapped DNA substrates were incubated on the slide for 30 min to attach DNA substrates on the slide through the anti-digoxigenin/digoxigenin linkage (46). Extensive buffer washing was used to remove any unbound DNA and oligos, followed by the introduction of streptavidin-coated polystyrene beads for visualization. TPM reactions were performed with 1.5 μM SpDmc1 in TPM reaction buffer (1 mM DTT, 2 mg/ml BSA, 2 mM ATP, 2 mM phosphoenolpyruvate, 0.02 U/μL pyruvate kinase, 30 mM Tris–HCl, 120 mM NaCl, 3.5 mM MgCl₂ at pH 7.5). Images of the polystyrene bead tethers were acquired by an inverted optical microscope (Olympus IX71) as previously described (47). Bead Brownian motion (BM) amplitude was calculated by the standard deviation of bead centroid positions over 1000 frames and was analyzed using MATLAB. The solver function of Excel fitted the time-course change of BM, as described previously (22,48).

Dwell time analysis

The dwell times were modeled by exponential distributions. The exponential time constants or, equivalently, rate parameters were determined using a maximum likelihood estimator. The standard errors of the estimates were estimated using bootstrapping. For a dwell time dataset with a size of N, 5000 bootstrap samples were created by resampling N points from the data with replacement. An exponential parameter was estimated from each bootstrapping sample, forming a set of bootstrap estimates. The standard deviation of the bootstrap estimates was reported as the standard error of the parameter estimate.

RESULTS

Swi5-Sfr1 and Hop2-Mnd1 stimulate the nucleation of Dmc1

Nucleation of recombinases on DNA to form stable recombinase-DNA nucleoprotein filaments is the first committed step for DNA strand exchange reaction central to homologous recombination. We had previously studied the dynamics of recombinase binding and dissociation on DNA during the nucleating cluster formation (22,49). A previous study has suggested the structural differences in Dmc1 and Rad51 nucleoprotein filaments, which might allow accessory proteins to modulate their functions differentially (50). We now aim to investigate how the Swi5-Sfr1 (SS) and Hop2-Mnd1 (HM) complexes affect the nucleating cluster formation of Dmc1 using single-molecule FRET (smFRET) experiments. We used a DNA substrate containing an 18-bp double-strand handle and a 23-nt secondary-structure-free polydT with the donor and acceptor dye separated by 13-nt (dT₁₃₊₁₀, Figure 1A). Every single recombinase monomer is known to bind to 3-nt ssDNA, so at most 4 Dmc1 molecules occupy over the 13-nt between the dye pair in this substrate. Without Dmc1, the ssDNA strand is flexible, and the distance between dye pairs is short, resulting in a high FRET value of 0.64 ± 0.05. When Dmc1 binds to DNA between the dye pair, Dmc1 extends the DNA, leading to a lower FRET value. This FRET substrate thus allows us to monitor the Dmc1 binding dynamics within the 13-nt segment near the ss/dsDNA junction in the context of overall 23-nt long ssDNA. Even though both Swi5-Sfr1 and Hop2-Mnd1 harbor DNA binding properties, DNA binding of these individual proteins did not cause changes in FRET histograms without Dmc1 (Figure 1B, left, and Supplemental Figure S3 for dT₁₃₊₄₇ substrate). Therefore, all FRET changes result from the Dmc1 binding.

Figure 1.

Swi5-Sfr1 and Hop2-Mnd1 stimulate Dmc1 binding on DNA. (A) Scheme of single-molecule FRET DNA binding assay. The dT₁₃₊₁₀ DNA substrate contains an 18-bp double-strand handle, a 23-nt poly(dT) ssDNA with the Cy3 and the Cy5 dye pair separated by 13-nt. When Dmc1 binds to DNA, the separation of the dye pair increases and leads to decreased FRET. (B) FRET histograms of Dmc1 binding to DNA based on the more than ninety time-courses. The DNA-only state has the FRET value of 0.64 ± 0.05 (orange), and including HM, SS or HM + SS do not result in FRET change (left). The percentage of the DNA-only state decreases in the presence of 3 μM Dmc1 and 0.5 μM accessory proteins as specified (Hop2-Mnd1, HM; Swi5-Sfr1, SS).

When Dmc1 alone (3 μM) was added to the dT₁₃₊₁₀ DNA substrate, the FRET histogram (compiled from >90 individual DNA time-courses of 100-sec recordings) contained a major DNA-only peak (∼95% population) and a very small population with FRET value near 0.2 (Figure 1B). This indicates that Dmc1 alone cannot form stable nucleoprotein filaments on 23-nt ssDNA in the presence of ATP. When either Hop2-Mnd1 or Swi5-Sfr1 was added with Dmc1, the DNA-only population dropped to 91% or 81%, respectively, suggesting that these accessory proteins stimulate Dmc1 binding on ssDNA. Interestingly, when adding two accessory proteins together with Dmc1, lower FRET states became apparent, and the DNA-only population was further reduced to 59% shown in the histogram (Figure 1B). These results indicate that both accessory proteins stimulate the Dmc1 nucleating cluster formation.

We also used a dT₁₃₊₄₇ substrate of a total of 60-nt secondary-structure-free polydT ssDNA, but with the same 13-nt dye pair separation near the junction, to monitor Dmc1 binding (Supplement Figure S4A). The overall longer ssDNA length in the dT₁₃₊₄₇ DNA substrate is expected to allow the formation of longer and more stable Dmc1 nucleoprotein filaments. Adding 3 μM Dmc1 to dT₁₃₊₄₇ substrates leads to a higher Dmc1-bound fraction (∼ 62% of DNA-only population, Supplement Figure S4B), as expected. Adding Hop2-Mnd1 or Swi5-Sfr1 with Dmc1 leads to a reduced DNA-only population (40% or 55%). Adding both accessory proteins further reduced the DNA-only population to ∼20%. The results from both dT₁₃₊₁₀ and dT₁₃₊₄₇ experiments confirm that both Hop2-Mnd1 and Swi5-Sfr1 stimulate Dmc1 nucleation on DNA.

Hop2-Mnd1 increases the binding rate of Dmc1, and Swi5-Sfr1 decreases the dissociation rate of Dmc1

Time-courses of the dT₁₃₊₁₀ FRET substrates offer a resolution to monitor individual Dmc1 binding and dissociation (Figure 2A). We used the variational Bayesian FRET (vbFRET) to analyze the observed dynamic alternations in FRET states in the presence of Dmc1 and accessory proteins, as we previously used in studying Rad51 recombinases (22). The 13-nt dye pair separation allows up to 4 Dmc1 bindings. Our analysis returned at most 4 FRET states, corresponding to 0, 1, 2, and 3 Dmc1 bindings. The absence of the fourth Dmc1 binding-induced FRET state could result from the unstable nucleoprotein filaments due to the short total ssDNA length used. Exemplary time-courses shown in Figure 2B-D are consistent with the FRET histograms (Figure 1B) that Dmc1 excursion to the lower FRET states is more frequently seen in the presence of either Hop2-Mnd1 or Swi5-Sfr1 complex. To confirm the state assignment made by vbFRET, we compiled FRET values before and after individual transitions, as shown in the transition density plot (TDP, Figure 2E–G). Symmetric FRET transitions are seen in these transition density plots. Also, the FRET states align well within these three sets of experiments. Thus, these TDPs offer confidence in the validity of FRET state assignments. It is also apparent that transitions between lower FRET states (bottom left of the TDP) are more frequent in the presence of Swi5-Sfr1 or Hop2-Mnd1, reflecting that Swi5-Sfr1 and Hop2-Mnd1 stimulate to populate Dmc1 in those lower FRET states (more than 1 Dmc1 bound state).

Figure 2.

Hop2-Mnd1 increases Dmc1 binding on DNA, and Swi5-Sfr1 decreases the dissociation. (A) Single-molecule FRET binding assay determines Dmc1 binding and dissociation rates in the dT₁₃₊₁₀ substrate. (B–D) Exemplary FRET time-courses of 3 μM Dmc1 (B) binding to DNA, and in the presence of 1 μM Hop2-Mnd1 (C, HM) or Swi5-Sfr1 (D, SS). Raw data (light grey), average (dark grey), and the fitting by vbFRET (blue line) of the FRET states. (E–G) Transition density plot of the Dmc1 binding and dissociation transitions. The x-axis is the FRET before the transition, and the y-axis is the one after the transition. The heat maps indicate the transition density among the four FRET states corresponding to Dmc1 binding and dissociation. (H, I) Apparent binding rates of Dmc1 nucleating cluster dynamics at different concentrations of Hop2-Mnd1 (H, orange) and Swi5-Sfr1 (I, green). 0, 1, 2, 3 refer to the numbers of bound Dmc1 molecules within the Cy3–Cy5 pair. (J, K) Apparent dissociation rates of Dmc1 nucleating cluster dynamics in the presence of Hop2-Mnd1 (J, orange) and Swi5-Sfr1 (K, green). Error bars are the standard error by bootstrapping 5000 times.

Dwell time analysis in vbFRET allows us to determine the apparent reaction rates of individual binding and dissociation transitions. In the fixed Dmc1 concentration of 3 μM, we titrated concentrations of Hop2-Mnd1 and Swi5-Sfr1 from 0.25 μM to 3 μM and analyzed these FRET time-courses to obtain the rates of each Dmc1 binding/dissociation transitions. The apparent binding rates are summarized in Figure 2H and I (values in Supplemental Table S2), with open bars designated for the Dmc1-only experiments and displayed in both panels for easy comparison. The 0→1 refers to the apparent binding rate from the no-Dmc1 (0) state to the 1-Dmc1-bound (1) state. At increasing Hop2-Mnd1 concentrations, binding rates of Dmc1 increase for 0→1 and 1→2 transitions, but not for 2→3 transitions (Figure 2H), surprisingly. Hop2-Mnd1 stimulation on Dmc1’s binding rates depends on its concentrations up to 2 μM of Hop2-Mnd1, with no additional stimulation at 3 μM. A maximum of ∼2-fold increase in Dmc1 binding rate was seen in the presence of Hop2-Mnd1. In contrast, adding Swi5-Sfr1 does not result in a statistically significant change in Dmc1 binding rates, even though some stimulation is seen for the 2→3 transition at high Swi5-Sfr1 concentrations (>1 μM, Figure 2I). Thus, two accessory proteins modulate Dmc1’s binding step differently.

Interestingly, the apparent dissociation rates show the opposite trend (Figure 2J, K). Hop2-Mnd1 does not significantly affect the dissociation rate of Dmc1, even though there might be some stabilization effect at 1→0 transition at high concentrations (3 μM). Swi5-Sfr1, on the other hand, reduces the dissociation rate of Dmc1, most significantly for 3→2 and 1→0 transitions. For the 3→2 transition, 0.5 μM SS is sufficient for the apparent stabilization effect. For the 1→0 transition, 2 μM SS can significantly stabilize Dmc1 from dissociation about 3-fold (∼ 0.6 s^-1 for Dmc1-only and ∼0.2 s^-1 for Dmc1+SS). The reduction in Dmc1 dissociation by Swi5-Sfr1 is consistent with the previous studies using mouse SWI5-SFR1 and RAD51 (22). Again, Hop2-Mnd1 and Swi5-Sfr1 act differently on the dissociation step of Dmc1 from DNA. Thus, adding both Swi5-Sfr1 and Hop2-Mnd1 shows additive effect (Supplemental Table S4).

Hop2-Mnd1 stimulates Dmc1 nucleation at the ss/dsDNA junction

FRET results suggest that Hop2-Mnd1 stimulates Dmc1 nucleoprotein filament assembly by increasing Dmc1’s binding on DNA without significantly affecting its dissociation. To understand how Hop2-Mnd1 proteins stimulate Dmc1 assembly, we used the tethered particle motion (TPM) experiments to monitor nucleation events associated with stable nucleoprotein filament assembly. In TPM experiments, recombinase binding to DNA extends the DNA length, leading to the increase in bead Brownian motion (BM) in the DNA tethers (22,46,51) (Figure 3A). Snapshots of BM histograms 15 min after adding the 1.0 μM Dmc1 showed that ∼ 63% of the DNA tethers are assembled (dT₁₀₀ substrate, Supplemental Figure S5). In the presence of Hop2-Mnd1 or Swi5-Sfr1, assembled fraction increases (74% and 70%). When both accessory proteins were added together, assembled fraction further increased (>89%), consistent with our FRET measurements.

Figure 3.

Hop2-Mnd1 binding to DNA junction stimulates Dmc1 nucleation. (A) Scheme of tethered particle motion filament assembly assay. As Dmc1 initiates nucleation on ssDNA, Dmc1 recombinases extend DNA, increasing bead Brownian motion (BM). (B) Exemplary time-courses and the BM histogram of filament assembly on dT₉₀ DNA substrates in the presence of 1.5 μM Dmc1-only. The dark grey region is the dead time of observation during the introduction of proteins. The nucleation time is the duration between the protein introduction and the onset of continuous BM increase. (C–E) Histograms of nucleation times of filament assembly on dT₉₀ DNA substrates in the presence of 1.5 μM Dmc1 (C), with 50 nM Swi5-Sfr1 (D, SS) or Hop2-Mnd1 (E, HM). n is the number of nucleation events included. (F–H) Histograms of nucleation times of filament assembly on dT₄₅₊₄₅ DNA substrates in the presence of 1.5 μM Dmc1 (F) with 50 nM Swi5-Sfr1 (G) or Hop2-Mnd1 (H). The dT₄₅₊₄₅ contains two 45-nt dT ssDNA sandwiched by a 24-bp dsDNA, with two 5′-end dsDNA junctions. (I) Summary of nucleation times of Dmc1 in the presence of Hop2-Mnd1 or Swi5-Sfr1 on the dT₉₀ and dT₄₅₊₄₅ substrates shown in C-H. Experiments of introducing Dmc1-only on the pre-incubation of Swi5-Sfr1 or Hop2-Mnd1 complex on DNA substrates are listed as (SS→Dmc1) or (HM→Dmc1), respectively. Error bars are the standard error by bootstrapping 5000 times. (J) Determination of the apparent nucleation rate constants of Dmc1 for ssDNA and ss/dsDNA junctions using dT₉₀ (one 5′-ss/dsDNA junction) and dT₄₅₊₄₅ (2 such junctions). The y-intercept refers to the apparent rate constants for nucleation by 90-nt ssDNA, and the slope refers to that for ss/ds junctions. In the presence of Hop2-Mnd1 (orange), the slope increases ∼4-fold compared to the Dmc1-only (blue), while the intercept is similar. In the case of Swi5-Sfr1 (green), the slope is unchanged, while the intercept increases by about ∼35%.

Real-time measurement of bead BM amplitude offers the kinetics of stable recombinase filament assembly with details. We used a dT₉₀ DNA substrate to monitor how Dmc1 assembles. The dwell time between the addition of Dmc1 and the onset of the continuous BM increase is the nucleation time (Figure 3B, Supplemental Figures S6 & S7). The average nucleation time for the Dmc1-only (1.5 μM) reaction is 104.8 ± 9.8 s (Figure 3C). Adding Swi5-Sfr1 together with Dmc1 only slightly reduces the nucleation time to 83.7 ± 8.0 s. In the presence of Hop2-Mnd1, nucleation time is reduced two-fold to 52.8 ± 7.7 s (Figure 3C–E, I, Supplemental Table S3). Apparently, Hop2-Mnd1 stimulates the Dmc1 nucleation step in filament assembly.

Previous studies suggested that Dmc1 from budding yeast and mouse were found to preferentially nucleate on the 5′-ss/dsDNA junctions (46). Using the same previously developed dT₄₅₊₄₅ DNA substrate containing the same total ssDNA lengths of 90-nt but with two ss/dsDNA junctions (46), we set out to test if SpDmc1 behaves similarly. The nucleation times of dT₄₅₊₄₅ for the SpDmc1-only reaction is 77.6 ± 8.2 s (Figure 3F–I, Supplemental Table S3), about 1.3-fold stimulation compared to dT₉₀ substrate (104.8 ± 9.8 s). It thus confirms that SpDmc1 also possesses similar ss/dsDNA junction nucleation preference. Nucleation times of dT₄₅₊₄₅ experiments in the presence of Swi5-Sfr1 or Hop2-Mnd1 are both shorter than the corresponding ones using dT₉₀ substrates (Figure 3C-E). Swi5-Sfr1 showed ∼1.3-fold stimulation (64.7 ± 7.6 s for dT₄₅₊₄₅; 83.7 ± 8.0 s for dT₉₀), but Hop2-Mnd1 showed a higher ∼1.7-fold stimulation (30.9 ± 3.4 s for dT₄₅₊₄₅; 52.8 ± 7.7 s for dT₉₀). As FRET experiments showed that Hop2-Mnd1 preferentially increases Dmc1’s binding rate (Figure 2D), we wonder how Hop2-Mnd1 stimulates Dmc1’s association with DNA during the nucleation step. One model is that Hop2-Mnd1 forms complexes with Dmc1 in solution, and the complex binds to DNA junction with a higher affinity. The other model is that Hop2-Mnd1 binds to dsDNA first and then recruits Dmc1 to the DNA junction. To distinguish these models, we changed the order of the addition of Hop2-Mnd1 by first adding Hop2-Mnd1 to the reaction chamber containing dT₉₀ DNA substrates for five min, washing extensively to remove unbound Hop2-Mnd1 and then introducing Dmc1 (HM→Dmc1, Figure 3I, Supplemental Figure S8). Adding Hop2-Mnd1 and Dmc1 sequentially results in a nucleation time of 47.2 ± 6.1 s, similar to adding both together (52.8 ± 7.7 s). During single-molecule experiments, when components and buffer are introduced into the reaction chamber, a buffer exchange occurs, effectively removing any previously unbound components from the surface-tethered DNA molecules. As a result, the sequential addition of HM first and then Dmc1 ensures the removal of any HM that is not bound to DNA substrates. Therefore, it is highly likely that HM stably binds to DNA within the timescale of our experiments. This result clearly indicates that Hop2-Mnd1 stimulates Dmc1 association by binding to DNA first to form the Hop2-Mnd1/DNA complex. It also suggests that Hop2-Mnd1 binding to DNA is kinetically fast, compared to Dmc1 nucleation, and that Hop2-Mnd1-DNA binding is sufficiently stable to resist buffer exchange. In contrast, the experiment of sequential adding Swi5-Sfr1 and Dmc1 (SS→Dmc1, Figure 3I, Supplemental Figure S8) resulted in a nucleation time of 96.5 ± 13.5 s, similar to adding Dmc1 only.

Analysis of nucleation times on the substrates containing different amounts of 5′-ss/dsDNA junctions allows the determination of the apparent nucleation rate constants of Dmc1 for ssDNA and ss/dsDNA junctions (Figure 3J). At the same Dmc1 concentration and the fixed total length of ssDNA (90-nt), the nucleation rates can be expressed as: rate = (constant)*(k_ss^apparent+ k_junction*numbers-of-junction). Thus, the y-intercept of Figure 3J refers to the apparent rate constants for nucleation on the 90-nt ssDNA, and the slope refers to that for ss/ds junctions. In the presence of Hop2-Mnd1, a significant 4-fold increase in the Dmc1’s nucleation on the ss/dsDNA junctions is seen (k_junction), while no change is seen in Dmc1’s nucleation on ssDNA (k_ss^apparent). However, in the presence of Swi5-Sfr1, no effect on the ss/dsDNA junctions is shown. Therefore, Hop2-Mnd1 stimulates Dmc1 nucleation by increasing Dmc1’s association with DNA through the ss/dsDNA junctions, while Swi5-Sfr1 does not.

A previous study reported that SpDmc1 can bind to dsDNA (52). To check whether Dmc1 initiates filament assembly by binding to the dsDNA handle in the substrate, we used a duplex-only DNA substrate (439 bp dsDNA) as a control experiment. The time course of ds439 did not show a BM increase in the 500 s time scale (Supplemental Figure S9A), and the histogram of BM indicates that the BM before and 10-min after the Dmc1 introduction is the same (Supplemental Figure S9B, C). Therefore, Dmc1 cannot initiate nucleation by binding to dsDNA, and our observation of Dmc1 assembly resulted from the nucleation on the ssDNA and the junction part.

Swi5-Sfr1 stabilizes Dmc1 by reducing the onset of filament dissociation

Our earlier studies suggested that Swi5-Sfr1 stabilizes the Rad51 filament by preventing Rad51 dissociation (22). We then tested whether Swi5-Sfr1 also stabilizes Dmc1 filaments. In this TPM filament disassembly experiment, we first prepared extended Dmc1 filaments, followed by extensive buffer wash to remove any free Dmc1 to initiate disassembly (Figure 4A). There are two kinetic parameters obtained from this filament disassembly experiment: dissociation time and disassembly rate. We determined how stable the Dmc1 nucleoprotein filaments were by measuring the dissociation time, the dwell time between the removal of free Dmc1 molecules by buffer wash and the initiation of the continuous BM decrease (as marked by the double-headed arrow in Figure 4B-C). It reflects the rate-limited step for the dissociation of the first Dmc1 molecule. The BM of the Dmc1-extended filaments for the dT₉₀ DNA substrate was ∼70 nm. After the buffer wash, the BM of the Dmc1 filament persisted for a while before continuous dissociation took place. Eventually, all Dmc1 dissociated, resulting in the DNA-only substrate (BM of ∼39 nm, Figure 4B-C). Our results showed that the dissociation time of Dmc1-only is 115.1 ± 7.5 s (Figure 4D, Supplemental Table S3). In the presence of 50 nM Swi5-Sfr1, the Dmc1 filament is stabilized, increasing the dissociation time to 172.9 ± 12.7 s. This likely reflects that Swi5-Sfr1 could bind to the disassembling end of the Dmc1 filaments to reduce the onset of filament dissociation. On the other hand, in the presence of 50 nM Hop2-Mnd1, the dissociation time is similar to those with Dmc1-only (107.1 ± 6.4 s, Figure 4C).

Figure 4.

Swi5-Sfr1 reduces the onset of the disassembly of the Dmc1 nucleoprotein filament. (A) Scheme of tethered particle motion filament disassembly assay. Dmc1 is incubated with dT₉₀ DNA substrate in the reaction chamber for 15 min to become a fully extended filament in the presence of Hop2-Mnd1 (HM) or Swi5-Sfr1 (SS) when indicated. The filament is extensively buffer-washed to remove free proteins. When Dmc1 disassembles from the filament, bead BM decreases. The dissociation time is the duration between the beginning of the buffer wash and the onset of continuous BM decrease. (B, C) Exemplary time-courses of Dmc1 filament disassembly on dT₉₀ DNA substrate in the presence of 1.5 μM Dmc1 (B) or with an additional 50 nM Swi5-Sfr1 (C). The dark grey region was the dead time of observation when buffer washes occurred. (D) Dissociation times of Dmc1 filament in the presence of Hop2-Mnd1 or Swi5-Sfr1 on dT₉₀ substrate. Error bars are the standard error by bootstrapping 5000 times. (E–G) The histograms of disassembly rates represented the slopes in the Brownian motion decrease of Dmc1 filament in the presence of Hop2-Mnd1 or Swi5-Sfr1 on the dT₉₀ substrate.

The BM time-courses also disclosed the disassembly rate of recombinases once the disassembly commences, as shown by the slope of decreasing BM (evident between 100–130 s in Figure 4B). The Dmc1 disassembly rate remains unchanged in the presence of Hop2-Mnd1 (0.356 ± 0.045 nm/s for Dmc1 + HM, 0.392 ± 0.046 nm/s for Dmc1-only, Figure 4E), consistent with the FRET result and the model that Hop2-Mnd1 does not affect the Dmc1 dissociation significantly. Surprisingly, the disassembly rate increases ∼1.6-fold in the presence of Swi5-Sfr1 (0.642 ± 0.110 nm/s, Figure 4E and Supplemental Table S3). As Swi5-Sfr1 stimulate ATP hydrolysis of Dmc1 and Rad51 (16,21), it is possible that Swi5-Sfr1 facilitates ATP hydrolysis of Dmc1 and thus its dissociation upon the onset of the filament disassembly. Also note that once the disassembly is initiated, no pause was seen during BM decrease, suggesting a continuous Dmc1 dissociation. Thus, it is possible that Swi5-Sfr1 binding to the disassembling end and other parts of the Dmc1 filament contribute differently. Reducing the onset of Dmc1 filament dissociation (longer dissociation time) must have resulted from the Swi5-Sfr1 capping the Dmc1 at the disassembling end. This interesting observation might reflect efficient Dmc1 usage in the filamentous stage.

DISCUSSION

Homologous recombination is tightly regulated spatiotemporally in meiosis, so regulating Dmc1 during filament assembly would serve as an effective regulatory strategy. Here, we used two types of single-molecule experiments to monitor how Dmc1 assembles on DNA to form nucleoprotein filaments in the presence of two accessory protein complexes, Hop2-Mnd1 and Swi5-Sfr1. Surprisingly, these two complexes regulate Dmc1 filament assembly at different steps. Hop2-Mnd1 binds to DNA first to promote Dmc1 assembly at the ss/dsDNA junction. Swi5-Sfr1, on the other hand, suppresses the dissociation of Dmc1 molecules from the assembled filament. Together, these complexes contribute to the efficient assembly of the Dmc1 filament (Figure 5).

Figure 5.

Hop2-Mnd1 and Swi5-Sfr1 stimulate Dmc1 through different mechanisms. Dmc1 filament assembly is stimulated by both Hop2-Mnd1 (HM) and Swi5-Sfr1 (SS) but with different mechanisms. Hop2-Mnd1 enhances the association of Dmc1 by binding on DNA and recruiting Dmc1 to the DNA junction. Swi5-Sfr1 reduces the dissociation of Dmc1.

Adding Hop2-Mnd1 and Swi5-Sfr1 together leads to a significant stimulation for Dmc1 assembly. To best illustrate this effect, we monitored the filament assembly of the dT₁₃₊₁₀ substrate, a substrate not long enough for Dmc1 to form stable nucleoprotein filaments. Given the fixed Dmc1 concentration, Dmc1-alone renders ∼5% of DNA in filament form, Hop2-Mnd1 or Swi5-Sfr1 returned increases additional 4% and 14% in filament form (Figure 1B). If stimulation by Hop2-Mnd1 and Swi5-Sfr1 is independent, a combined 23% filament form will be expected (5%, 4% and 14%). Surprisingly, adding all three proteins together showed 41% in filament form, a nearly two-fold increase in assembly stimulation. Given that Hop2-Mnd1 and Swi5-Sfr1 bind to different interfaces on Dmc1 (42) and act on different steps of the Dmc1 assembly, Hop2-Mnd1 and Swi5-Sfr1 could synergistically promote the Dmc1 filament assembly in kinetics, allowing enhanced recombination efficiency.

We showed that Hop2-Mnd1 stimulates Dmc1 binding during the nucleating cluster formation and thus reduces nucleation time during the assembly of stable filaments. Careful examination of the Hop2-Mnd1 effect on Dmc1 showed that it preferentially affects the first two binding rates (0→1 and 1→2) (Figure 2H), seen in FERT analysis. As there is only 13-nt long between Cy3 and Cy5 dyes, Hop2-Mnd1 effect on the third Dmc1 might not be apparent in the current setup. Also, TPM experiments showed that Hop2-Mnd1 stimulation depends on the amounts of ss/dsDNA junctions (Figure 3I). Analysis of apparent nucleation rates on the substrates containing different 5′-ss/dsDNA junction numbers showed a significant 4-fold increase in the Dmc1’s nucleation on the ss/dsDNA junctions in the presence of Hop2-Mnd1. At the same time, no change is seen in Dmc1’s nucleation on ssDNA segments (Figure 3J). Surprisingly, Swi5-Sfr1 showed no such junction effect (Figure 3J). Even though Hop2-Mnd1 binds to both ssDNA and dsDNA substrates, it binds to duplex DNA with higher affinity (36,38–40,53). The sequential addition of Hop2-Mnd1 and Dmc1 leads to the same stimulation on Dmc1 nucleation (Figure 3I). Given that Dmc1 preferentially initiates nucleation at ss/dsDNA junctions (46), it is likely that Hop2-Mnd1 binds to the duplex or near the ss/ds junction in the DNA substrates and recruits Dmc1 to the junction.

While the most significant stimulation effect of Hop2-Mnd1 is seen in the Dmc1’s binding rate, Hop2-Mnd1 also shows some stabilizing effect on Dmc1 dissociation, most apparent for 1→ 0 transition (Figure 3J). vbFRET analysis showed that at the high Hop2-Mnd1 concentration, Dmc1’s 1→0 dissociation rate constant is reduced by 50% (Supplemental Table S2, 0.32 s^-1, compared to 0.62 s^-1 for Dmc1-only). This stabilization effect is consistent with the model that Hop2-Mnd1 likely binds near the ss/dsDNA junction and interacts with the Dmc1 molecule in proximity.

TPM disassembly experiments showed that Swi5-Sfr1 can reduce recombinase dissociation from the stable nucleoprotein filaments of Rad51 (22) and Dmc1 (Figure 4). Interestingly, the stabilization effect of Swi5-Sfr1 for Rad51 filaments seems more apparent than for Dmc1 in S. pombe. In the presence of 50 nM Swi5-Sfr1, the dissociation time of Rad51 increased by 2.6-fold (22), but only a 1.5-fold increase for Dmc1 (Figure 4D). The difference might be caused by the DNA-binding properties of recombinases, as 0.3 μM Rad51 is sufficient to observe efficient filament assembly in TPM (22), but 1.5 μM Dmc1 is required. smFRET experiments of Rad51 and Swi5-Sfr1 on short DNA substrates also confirmed that Swi5-Sfr1 prevents Rad51 dissociation during its nucleating cluster formation (Supplemental Figure S10, and Supplemental Table S5). As Swi5-Sfr1 can fit into the groove of Rad51 recombinase filaments for stabilization (54), it is possible that Swi5-Sfr1 can stabilize the Dmc1 filament using the same mechanism.

In contrast, Hop2-Mnd1 does not affect Rad51 binding nor dissociation rates in the same dT₁₃₊₁₀ substrate (Supplemental Figure S10). Differential stimulation of Hop2-Mnd1 on Dmc1 and Rad51 filament assembly can result from the biochemical properties of Hop2-Mnd1, Dmc1, and Rad51 proteins. In yeast and mouse, distinct nucleation preference is seen between Dmc1 and Rad51. Dmc1 preferentially initiates nucleation at the ss/dsDNA junctions, while Rad51 initiates at ssDNA segments (46). Given that Hop2-Mnd1 binds dsDNA with higher affinity, it is reasonable for Hop2-Mnd1 to execute its stimulation specifically on Dmc1 nucleation but not Rad51.

Dmc1-mediated recombination is initiated from the assembly of the Dmc1 filament, followed by the strand exchange process. Both Swi5-Sfr1 and Hop2-Mnd1 activate the Dmc1-driven strand exchange (42). Previous work has closely examined their impact on the stability of the Dmc1 filaments and found that Swi5-Sfr1 greatly stabilizes the Dmc1 filaments while Hop2-Mnd1 does not (42). Specifically, when Dmc1 filaments were preincubated with Hop2-Mnd1 and then challenged by the addition of RPA, the Dmc1 assembled on ssDNA became easily replaced by RPA. On the other hand, upon the challenge of RPA, the Dmc1 persisted much longer on ssDNA when preincubated with Swi5-Sfr1. This observation is in line with the fluorescence anisotropy measurements in which Dmc1 filaments preincubated with Swi5-Sfr1 have a significantly longer binding time to ssDNA compared to those with Dmc1 alone, while Hop2-Mnd1 has little effect on stabilizing the Dmc1 filaments. In this work, we focused on the Dmc1 filament assembly kinetics and showed that both Swi5-Sfr1 and Hop2-Mnd1 stimulate Dmc1 filament assembly but through distinct mechanisms. Swi5-Sfr1 reduces the dissociation of Dmc1 from the Dmc1 filament while Hop2-Mnd1 promotes the association of Dmc1. This finding agrees with the observation that Hop2-Mnd1 promotes the formation of the Rad51/Dmc1 filament (36,37). Taken together, the findings above point towards the importance of Swi5-Sfr1 in stabilizing Dmc1 oligomers while Hop2-Mnd1 promotes Dmc1 oligomerization; such Dmc1 oligomers will remain dispersive unless solidified by other factors such as Swi5-Sfr1. Hop2-Mnd1 is thought to actively participate in assessing the similarity between two DNA molecules, which likely involves dynamic assembly and disassembly of Dmc1 (42). Hop2-Mnd1-mediated Dmc1 oligomerization might play a key role in this process.

What is the implication of having two distinct mechanisms for promoting Dmc1 assembly? Dynamic assembly and disassembly of RecA recombinase nucleoprotein filaments have been shown to facilitate the finding of homologous DNA in cells by reducing dimensionality (55). Hop2-Mnd1-directed Dmc1 assembly could be advantageous for such homology search. Dmc1 assembly is nucleated at the ss/dsDNA junction (46), which happens to be where nascent ssDNA is generated by the action of various 5′-to-3′ exonucleases (56). The Hop2-Mnd1-directed Dmc1 assembly might be actively involved in the homology assessment of newly exposed ssDNA. Cytological observations using budding yeast showed a massive accumulation of Dmc1 and Mei5-Sae3 (the budding yeast counterpart of Swi5-Sfr1) in the absence of Hop2-Mnd1 (13,23). Even though the Hop2-Mnd1-directed mechanism is not essential for Dmc1 assembly, chromosomes, unfortunately, are associated with non-homologous partners under such conditions. Thus, it suggests that homology searching is inaccurate or defective without Hop2-Mnd1 (26,33). Apparently, the initial homology search likely relies on the ssDNA segment adjacent to its end. In contrast, DSB ends would keep being resected with the continuous generation of nascent ssDNA at the ss/dsDNA junction. DNA sharing short sequence homology with break-associated ssDNA can exist at multiple locations in the genome. Therefore, the initial homology assessment likely would need further validation by the continuity of homology towards its 5′ direction before the full commitment to HR. It is possible that Swi5-Sfr1-stabilized Dmc1 filaments represent locations where the initial round of homology searching is being conducted, awaiting further validation conducted by the Hop2-Mnd1-directed mechanism to move on to the later stages of HR.

Differences in DNA binding properties of accessory proteins and nucleation preferences of Dmc1 and Rad51 dictate the differential needs for efficient recombination progression. Our systematic studies of the accessory proteins on these two recombinases form a molecular basis to elucidate how these accessory proteins regulate and orchestrate these recombination events. Our studies further provide the molecular basis for why recombinases are regulated by different sets of accessory protein complexes based on their biochemical characteristics. Considering that Rad51 has been suggested to function as an accessory factor in meiotic recombination (9), it would be interesting to explore how these two accessory complexes act together in the presence of Rad51 and Dmc1.

DATA AVAILABILITY

The data underlying this article are available in the article and in its online supplementary material.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

National Science and Technology Council of Taiwan [107-2113-M-002-010, 110-2113-M-002-020 to H.-W.L.]; Grants-in-Aid for Scientific Research from Japan Society for the Promotion of Science (JSPS) [A-JP22H00404 to H.I., B-JP18H02371 to H.T.]. Funding for open access charge: National Science and Technology Council of Taiwan.

Conflict of interest statement. None declared.