Two auxiliary factors promote Dmc1-driven DNA strand exchange via stepwise mechanisms

The fission yeast Hop2-Mnd1 complex was purified, along with other necessary proteins (SI Appendix, Fig. S1A), and its effect on Dmc1-driven strand exchange was tested in an assay that employs 5.4-kb circular ssDNA (css) and homologous linear dsDNA (lds; Fig. 1A). Pairing between css and the complementary strand of lds produces an intermediate (joint molecule, JM), and once the reaction is complete, a nicked circular DNA (nc) as the final product. Hop2-Mnd1 promoted Dmc1 activity at concentrations as low as 0.13 µM (1/40th the concentration of Dmc1), as demonstrated by the production of JM and nc (Fig. 1B). Interestingly, at higher concentrations, JMs of higher molecular weight (referred to as hyper JMs hereafter, denoted with asterisks in Fig. 1B) appeared with some signal observed in the well. The amount of nc declined as hyper JMs increased, suggesting that most DNA substrates are incorporated into strand exchange reactions that are never completed. The inclusion of ATP, Dmc1, and Hop2-Mnd1 is indispensable for formation of hyper JMs and nc (SI Appendix, Fig. S1B). We confirmed that the smear corresponding to hyper JMs is not an artifact caused by insufficient proteolysis before agarose gel electrophoresis, which could lead to a similar retardation in the mobility of DNA (SI Appendix, Fig. S1C). Hyper JMs are likely to be comprised of css and lds molecules that have undergone incomplete strand exchange because heat treatment resolved most of the JM species, leading to an increase in nc and lds (SI Appendix, Fig. S1D). Hyper JM structures might be produced via strand invasion of a single css into multiple lds molecules (42), thus forming a high-molecular weight DNA network (43).

Mechanistically, mouse Hop2-Mnd1 was proposed to stimulate human Rad51/Dmc1 by enhancing the ability of presynaptic filaments to capture dsDNA (37, 41, 44). This property was tested using fission yeast Hop2-Mnd1. The assay employed css annealed to a biotin-conjugated oligonucleotide, which allows precipitation of the css using streptavidin-coated magnetic beads (SI Appendix, Fig. S2A). css was initially incubated with Dmc1 in the presence of AMP-PNP, a nonhydrolyzable ATP analog, to form stable presynaptic filaments. The reaction was then supplemented with Hop2-Mnd1. Either homologous (PhiX174) or nonhomologous (pBluescript) linear dsDNA was added to the reaction, and the fraction of dsDNA associated with css (pellet, P) was separated from the rest (supernatant, S) using magnetic beads. dsDNA efficiently precipitated with css only when the reaction contained both Dmc1 and Hop2-Mnd1 (SI Appendix, Fig. S2B). PhiX174 and pBluescript were pulled down to a similar extent, indicating that homology between css and dsDNA is not necessary for dsDNA to be pulled down. These results strongly argue that, much like its mouse counterpart, fission yeast Hop2-Mnd1 facilitates dsDNA capture by Dmc1 presynaptic filaments.

Next, we decided to closely examine the biochemical similarities/differences between Hop2-Mnd1 and Swi5-Sfr1, another major auxiliary factor of Dmc1. Time-course analysis of Dmc1-driven strand exchange revealed that, in the presence of Hop2-Mnd1, the majority of lds is converted into a wide range of JMs, accounting for ∼70% of total DNA by 20 min, with the sum of nc and JMs plateauing at 60 min (Fig. 1C). By contrast, the disappearance of lds was more gradual in the reaction containing Swi5-Sfr1, and the sum of nc and JMs progressively increased up until 120 min. Thus, strand exchange stimulated by Hop2-Mnd1 is much quicker than that stimulated by Swi5-Sfr1. The apparent difference in the stimulation of Dmc1 by these two auxiliary factors prompted us to examine if they act synergistically in this reaction. Indeed, when Hop2-Mnd1 and Swi5-Sfr1 were both incubated with Dmc1 presynaptic filaments, a more robust production of hyper JMs was seen (Fig. 1D).

We next examined the physical interactions between Hop2-Mnd1 and the other proteins employed here (SI Appendix, Fig. S3A). Hexahistidine-tagged Hop2-Mnd1 (6xHis-Hop2-Mnd1) was incubated with either Dmc1 or Rad51, then 6xHis-Hop2-Mnd1 was precipitated with nickel-conjugated beads. Dmc1 and, to a lesser extent, Rad51 was pulled down with 6xHis-Hop2-Mnd1, which is consistent with a previous report (34). Interaction of 6xHis-Hop2-Mnd1 with Swi5-Sfr1 or RPA was barely detectable.

Since Dmc1 can physically bind to both Swi5-Sfr1 (14) and Hop2-Mnd1 (SI Appendix, Fig. S3A), we sought to test whether these interactions involve the same site on Dmc1. A competition binding assay was devised where 6xHis-Hop2-Mnd1 and Dmc1 were incubated with increasing amounts of Swi5-Sfr1. As a control, 6xHis-Hop2-Mnd1 and Dmc1 were incubated with various concentrations of untagged Hop2-Mnd1, which should compete with 6xHis-Hop2-Mnd1 for Dmc1 binding. As expected, the presence of untagged Hop2-Mnd1 reduced the pull-down efficiency of Dmc1 by 6xHis-Hop2-Mnd1 in a concentration-dependent manner (SI Appendix, Fig. S3 B and C). By contrast, the amount of pulled-down Dmc1 was not affected by the presence of Swi5-Sfr1. Furthermore, the amount of Swi5-Sfr1 that was pulled down by 6xHis-Hop2-Mnd1—in the presence of Dmc1—was increased at higher concentrations of Swi5-Sfr1, strongly arguing that these two auxiliary factors bind to different interfaces on Dmc1. This noncompetitive binding is consistent with the synergistic stimulation of Dmc1 by Hop2-Mnd1 and Swi5-Sfr1.

The ability of Hop2-Mnd1 to interact with Rad51 prompted us to examine if it stimulates Rad51-driven strand exchange. While Swi5-Sfr1 stimulated both Dmc1 and Rad51 (14), Hop2-Mnd1 had little effect on the activity of Rad51 (SI Appendix, Fig. S4A). However, mild stimulation of Rad51-driven strand exchange was seen when reactions containing Swi5-Sfr1 were supplemented with Hop2-Mnd1 (SI Appendix, Fig. S4B). This suggests that Hop2-Mnd1 can also stimulate Rad51, albeit to a far lesser extent than Dmc1.

How can Dmc1, when stimulated by Hop2-Mnd1, produce JMs more robustly and quickly than Swi5-Sfr1 (Fig. 1C)? One possibility is that Hop2-Mnd1 allows Dmc1 presynaptic filaments to access lds with more freedom. We hypothesized that Swi5-Sfr1–stimulated Dmc1 can only access lds at its ends, while Hop2-Mnd1 also permits internal entries. The lds substrate used so far has ends created by ApaLI, which generates 5′ overhangs. To examine the relationship between strand exchange and dsDNA ends, lds with various ends were created (Fig. 2A). lds prepared with XhoI, like ApaLI, was utilized efficiently by Dmc1 in the presence of Swi5-Sfr1 (Fig. 2B). When lds with 3′ overhangs was prepared by PstI or AatII, strand exchange was substantially reduced, more so for AatII than PstI. Virtually no strand exchange was seen when lds was prepared with StuI, suggesting that blunt ends hindered strand exchange. However, with SspI-digested lds, which also has blunt ends, some products were seen. Since the SspI-created ends are A/T-rich, how easily the duplex around an end is separated could also be an important parameter. Nonetheless, these results indicate that Swi5-Sfr1–stimulated strand exchange by Dmc1 relies heavily on the ends of lds. In sharp contrast, robust JM and nc formation was seen in Hop2-Mnd1–stimulated reactions containing lds substrates prepared with any of the six different restriction enzymes (Fig. 2). Thus, Hop2-Mnd1 nullifies the end dependency in Dmc1-driven DNA strand exchange.

Swi5-Sfr1 is an auxiliary factor with “mediator” activity for Dmc1 (15), meaning that it can facilitate the displacement of ssDNA-bound RPA by Dmc1 (45). We examined if Hop2-Mnd1 also has mediator activity. The strand exchange assay used above was modified so that ssDNA was coated with RPA first, followed by the addition of other proteins and lds. Swi5-Sfr1 stimulated Dmc1 in a concentration-dependent manner (Fig. 3A). Although some products were obtained with Hop2-Mnd1 in a concentration-independent manner, the yield was substantially reduced, with a ∼5-fold reduction in the final products. This indicates that Hop2-Mnd1 does not possess robust mediator activity for Dmc1.

Next, the stabilizing effect of Hop2-Mnd1 on the Dmc1 presynaptic filament was examined using the RPA chase assay (Fig. 3B). A biotin-conjugated oligonucleotide was first annealed to css, and this was then incubated with Dmc1 to form the presynaptic filament. Subsequently, reactions were supplemented with either Swi5-Sfr1 or Hop2-Mnd1, followed by the addition of RPA to challenge the presynaptic filaments by competing for css binding. css was then precipitated by streptavidin-coated magnetic beads and the amount of css-bound Dmc1 was examined. Stable Dmc1 filaments remain associated with css, whereas unstable filaments allow RPA to bind the css, leading to a reduction in the amount of css-bound Dmc1. Inclusion of these auxiliary factors had little effect on the binding of Dmc1 to css in the absence of RPA (SI Appendix, Fig. S5A). When preincubated with Swi5-Sfr1, Dmc1 effectively resisted displacement by RPA (Fig. 3C, compare lane 1 with lanes 4 and 5). Without Swi5-Sfr1, the amount of Dmc1 bound to css was drastically reduced, even in the presence of Hop2-Mnd1, and this was accompanied by an increase in css-bound RPA (Fig. 3C, lanes 1–3). Thus, unlike Swi5-Sfr1, Hop2-Mnd1 does not stabilize the Dmc1 presynaptic filament against displacement by RPA.

In order to more directly assess Dmc1 filament stability, the fluorescence anisotropy assay was employed. In this assay, fluorescently labeled ssDNA (72-mer oligo dT) was preincubated with Dmc1, then either Swi5-Sfr1 or Hop2-Mnd1 was added. Next, the reaction was diluted and the change in fluorescence anisotropy was monitored in real time. Upon dilution, anisotropy values showed a gradual decline (Fig. 3D), indicating that the mass associated with ssDNA, primarily contributed by Dmc1, was decreasing. This decline was less steep when Swi5-Sfr1 was included: at 0.6 µM Swi5-Sfr1 (one-fifth the concentration of Dmc1), k_off was reduced by 8.9-fold (SI Appendix, Fig. S5B). In sharp contrast, the inclusion of Hop2-Mnd1 had no obvious effect on anisotropy, indicating that Swi5-Sfr1 is an efficient stabilizer of the Dmc1 presynaptic filament while Hop2-Mnd1 is not.

The results thus far uncovered the attributes unique to each auxiliary factor. Swi5-Sfr1 is an efficient mediator/stabilizer of Dmc1 filaments, whereas Hop2-Mnd1 allows Dmc1 presynaptic filaments to engage dsDNA in the strand exchange reaction irrespective of the donor’s ends. To examine if these unique attributes function in a collaborative manner, a condition requiring both activities was employed by precoating css with RPA and utilizing lds with StuI-digested ends. Hyper JMs were only produced when both Swi5-Sfr1 and Hop2-Mnd1 were included in the reaction, demonstrating that these factors stimulate Dmc1-driven strand exchange in a complementary fashion (Fig. 3E). We note that this synergism was only observed for hyper JM formation, while the levels of nc was actually reduced (also seen in Fig. 1 B and D). This may be due to unproductive strand invasion events wherein one css invades into multiple lds molecules, some of which may already be engaged in reactions with other css molecules. The formation of such DNA networks would be expected to hinder the strand transfer reactions—which specifically drive the conversion of JMs into nc—for all DNA molecules involved.

HR in vivo is initiated at a DSB that is resected to form 3′-ended ssDNA at its ends, which searches for homology embedded somewhere within intact dsDNA. This homology is unlikely to be exposed at or near a DSB. Thus, we introduced a set of substrates that better mimic the strand exchange that occurs in vivo. A reasonably long (1.1 kb) linear ssDNA (lss) was employed as the recipient molecule (the molecule which incurs a DSB is termed recipient because it receives information from a donor dsDNA once homology is identified). As a donor substrate, one assay employed linear dsDNA (3.7 kb) with the region homologous to the ssDNA embedded near the middle of the molecule (Fig. 4 A, Left). Another assay employed nicked circular dsDNA (Fig. 4 A, Right), which lacks DNA ends, as a donor. With these two sets of substrates, ssDNA was preincubated with Dmc1, followed by further incubation with RPA and dsDNA with or without auxiliary factor(s). In both cases, Hop2-Mnd1 stimulated product formation, which was further promoted by the presence of Swi5-Sfr1, while Swi5-Sfr1 on its own barely stimulated Dmc1 (Fig. 4B). When ssDNA was precoated with RPA, efficient product formation was seen only when the reaction was supplemented with both Hop2-Mnd1 and Swi5-Sfr1 (Fig. 4C), further highlighting the distinct but complementary roles played by these two auxiliary factors in stimulating Dmc1-driven strand exchange.

We also examined the effect of heterology (70 bp), located at either the 5′-, 3′-, or both ends of lss (Fig. 5A). Products were practically undetectable without any auxiliary factors, and the inclusion of just Swi5-Sfr1 had a marginal effect on product formation. The inclusion of Hop2-Mnd1 resulted in reasonably efficient product formation when lss had at least one end free of heterology, achieving a yield of 40–50%, while the presence of heterology at both ends substantially reduced the reaction efficiency (∼10%). When both Hop2-Mnd1 and Swi5-Sfr1 were added to the reaction, the lss species with heterology at either end and those without any heterology achieved even higher yields (70–80%). Strand exchange was much less efficient when both ends of lss contained heterology (<40%). Taken together, we conclude that homology exposed at ssDNA ends is important, but not essential, for strand exchange.

All of the reactions were conducted in the scale of 10 µL at 30 °C. The following conditions were employed with PhiX174 substrates unless otherwise stated. Strand exchange buffer (30 mM Tris-Cl [pH 7.5], 120 mM NaCl, 30 mM KCl, 3.5 mM MgCl₂, 2 mM adenosine triphosphate [ATP], 1 mM dithiothreitol, 5% glycerol, 8 mM phosphocreatine, 8 units/mL creatine phosphokinase) containing 10 µM in nucleotides (µMnt) PhiX174 virion DNA (i.e., css [New England BioLabs, NEB]) was supplemented with 5 µM Dmc1 or Rad51 and incubated for 5 min. The indicated concentration of auxiliary factor(s), typically 0.25 µM for Hop2-Mnd1 and 0.5 µM for Swi5-Sfr1, was added to the reaction, which was further incubated for 5 min. One micromolar RPA was then added and incubated for 10 min. The strand exchange reaction was started by adding 10 µMnt of linearized dsDNA (lds, PhiX174 RF I DNA [NEB], ApaLI digested unless otherwise stated), which lasted for either 1 h or 2 h (as specified). Strand exchange between css and lds initially produces JMs as intermediates, which are then converted to nc and a linear ssDNA upon completion of strand exchange. lss is not visible on the gel because the band overlaps with css (Fig. 1A). In the case of the D-loop system, 5 µMnt linear ssDNA was supplemented with 5 µM Dmc1 and incubated for 10 min. The indicated concentration of auxiliary factor (typically 0.25 µM for Hop2-Mnd1 and 0.5 µM for Swi5-Sfr1), 1 µM RPA and 5 µMnt nc or linear dsDNA, were then mixed to start the reaction, which lasted for 1 h (or as indicated). In the case of mediator assays using PhiX174 substrates, 10 µMnt of PhiX174 virion DNA (css) was incubated with RPA at 2 µM for 10 min, then mixed with Dmc1, auxiliary factor(s) at 0.25 µM for Hop2-Mnd1 and/or 4 µM for Swi5-Sfr1, and linear PhiX174 RF I DNA (ApaLI digested unless otherwise stated) at 10 µMnt, which lasted for 1 h. In the case of D-loop mediator assays, 5 µMnt linear ssDNA was supplemented with 2 µM RPA and incubated for 10 min. Then, 5 µM Dmc1, indicated concentrations of auxiliary factor (typically 0.25 µM for Hop2-Mnd1 and 0.5 µM for Swi5-Sfr1), and 5 µMnt nc or linear dsDNA, were then mixed to start the reaction, which lasted for 1 h. At the end of all of the reactions described above, 10-µL reactions were subjected to psoralen-ultraviolet cross-linking to stabilize labile DNA structures and 2 µL of stop solution was added (30 mM Tris·Cl [pH 7.5], 44 mM EDTA, 3% SDS, 15 mg/mL proteinase K). Following a 5-min incubation, DNA was resolved in 1% agarose gels and stained with SYBR Gold (Thermo Fisher Scientific). Gels were then imaged using LAS4000 mini (GE Healthcare). Data acquisition and analysis were done as before except FIJI (65) was employed for signal quantification (14, 23). In brief, in the case of PhiX174-based substrates, background was subtracted and the amount of signal for areas corresponding to lds, nc, and JM bands was measured with the JM values divided by 1.5 to compensate for the extra signal generated by these three-stranded DNA molecules. The sum of the values was set to 100% and the percentage of total DNA corresponding to nc or JM was calculated. For the total yield, the percentage of DNA corresponding to nc and JM was combined. In the case of the D-loop assay, the sum of the value for dsDNA (nc or lds) and that for the product was set to 100%, and the yield was expressed as the percentage of the product per the sum. The values from three independent experiments were averaged and plotted along with error bars (SD). In SI Appendix, Fig. S4B, statistical significance was assessed by a paired t test.

All relevant data are presented in the manuscript. Requests for reagents or further information should be directed to H.T. Additional materials and methods are in SI Appendix.