Collective interactions along recombinase-bound D-loops could speed repair of double-strand breaks by destabilizing flanking homoduplex tails Authors: Claudia Danilowicz, Ashini Modi, Afra Sabei, Chantal Prévost, Mara Prentiss Published: 2026 ABSTRACT AbstractDuring repair of double-strand breaks, recombinases form D-loops in which one strand of the broken chromosome base pairs with a strand from an unbroken chromosome. Thus, the double-stranded DNA (dsDNA) in the unbroken chromosome must melt. Rapid repair requires quick melting. In this work, we present experimental results suggesting that both RecA and ScDmc1 form D-loops that distort the structure of long homoduplex tails in regions adjacent to recombinase-bound D-loops. The observed distortion is not simply due to attempted D-loop extension, but it is consistent with local melting of homoduplex tail regions adjacent to recombinase-bound D-loops. Our experiments indicate that the distortion arises from a collective interaction involving torsional stress that extends along RecA-bound D-loops from one homoduplex tail to the other. Furthermore, molecular dynamics simulations show melting in homoduplex tail regions adjacent to RecA-bound D-loops. Finally, we speculate that the destabilization in the base pairing in homoduplex tails flanking D-loops facilitates the extension of D-loops and underlies the previously observed extension of D-loops through 10 contiguous heterologous bases. Introduction Double-strand break (DSB) repair by homologous recombination helps maintain genomic integrity, though it occasionally also promotes genome evolution [1]. To repair a DSB, recombinase proteins form nucleoprotein filaments by binding to single-stranded DNA (ssDNA) regions formed by resection of the ends of the broken double-stranded DNA (dsDNA) [2, 3]. The nucleoprotein filament then probes an unbroken chromosome searching for the region that corresponds to the invading ssDNA sequence in the nucleoprotein filament. Homology is probed by attempting to establish Watson–Crick pairing between the invading strand and one of the strands in the unbroken chromosome [2–5]. If a homologous region is found, strand exchange transfers the base pairing of the complementary strand in the dsDNA from the displaced strand to the invading strand [2–5]. The resulting structure is a recombinase-bound D-loop. The D-loop includes three strands: the displaced strand and a heteroduplex in which the invading ssDNA is base-paired with the complementary strand. In the D-loops, the heteroduplex region is flanked by homoduplex tails comprised of the displaced and complementary strands (Fig. 1). Model of the structure of a 36 bp RecA-bound D-loop flanked by homoduplex tails based on previous modeling and cryoEM structures. No MD evolution is included in the structures. (A) Alternating RecA monomers are shown in blue and white. The invading, complementary, and displaced strands are shown in orange, purple, and light blue, respectively. The 5′ and 3′ labels indicate the ends of the invading strand. (B) Same as A, but the RecA monomers are transparent. (C) Same as B, but the RecA monomers are not shown. The bracket indicates the heteroduplex region. The blue and red circles highlight the boundary regions between the heteroduplex region and the flanking homoduplex tails. (D) Same as C, but the orange, purple, and blue lines are positioned at the outer radius of the invading, complementary, and displaced strands, respectively. Yellow rectangles join strands that are base-paired. (E) Lines corresponding to the complementary strand. The light purple line indicates the heteroduplex region, and the dark purple lines indicate the homoduplex tails. The dsDNA in the unbroken chromosome must melt before one of the strands in the dsDNA can base pair with the invading strand. Since strand exchange occurs in triplets [6], if the dsDNA melting of the homoduplex tails extends well beyond three base pairs, then strand exchange of a triplet will only involve base pairing of two single strands. In contrast, if the homoduplex tail is not melted, then each triplet would have to successively melt before strand exchange of that triplet takes place. Recent experiments with ATP hydrolysis are consistent with RecA-bound D-loops promoting homoduplex tail melting that extends ∼10 bp [7, 8] beyond the homologous region of a D-loop, but in RecA-free D-loops, homoduplex tail melting does not extend one base pair beyond the homologous region of the D-loop [9]. With or without ATP hydrolysis, D-loops are initially bound to RecA [4, 10–12]; however, ATP hydrolysis later drives unbinding of D-loops from RecA, resulting in RecA-free D-loops [10, 11, 13–17]. Thus, it is possible that with or without ATP hydrolysis, RecA-bound D-loops melt flanking homoduplex tails over regions extending up to ∼10 bp, but that the melting reverses once the D-loops become RecA-free. Importantly, such melting could arise from mechanical stress due to distortion in the structure of homoduplex tails in RecA-bound D-loops. In particular, as illustrated in Fig. 1, both experiments [18–20] and modeling [4, 5] suggest that in recombinase-bound D-loops there are strong kinks in the homoduplex tails at the boundary between the heteroduplex region and the homoduplex tails. Those strong kinks represent large deviations from the structure of B-form dsDNA. Furthermore, after the kink, on both the 5′ and 3′ sides of the heteroduplex, more than one helical turn in each homoduplex tail is bound to the nucleoprotein filament. The binding of those homoduplex tails to the nucleoprotein filament explains why protection from digestion by a restriction endonuclease extends 11–14 bases beyond the heteroduplex into the flanking homoduplex tails on both the 5′ and 3′ sides of the heteroduplex [21]. The structural distortion in those kinks and in the nucleoprotein filament-bound regions of the homoduplex could melt homoduplex tails in ∼10 bp regions flanking RecA-bound D-loops. Collective interactions along DNA can influence biological processes [22]. Importantly, twisting or pulling on dsDNA creates stress-induced collective interactions that can locally melt dsDNA [23–27]. Torsional stress in dsDNA can extend over more than 1 kb [28] and influence the function of proteins [29–31], including RecA [32]. Finally, pulling on dsDNA also affects the function of RecA [33] and Rad51 [34]. Importantly, RecA-bound DNA is more extended and less twisted than B-form DNA [35]. Furthermore, both analytical treatments [36] and molecular dynamics [37] suggest that the mechanical stress due to the differential extension of the DNA backbones in RecA-bound D-loops creates collective interactions that make the structure of heteroduplex triplets depend on the number of base pairs in the D-loop and the position of the triplet within the D-loop. In contrast, in the absence of collectivity, each triplet in the heteroduplex would have the same structure regardless of the number of base pairs in the D-loop or the position of the triplet within the D-loop. We performed bulk FRET experiments to investigate structural distortion in homoduplex tails flanking recombinase-bound D-loops. To determine the extent of homoduplex tail distortion, we measured the emission from fluorescein in homoduplex tails. When the dsDNA is in B-form, the emission is quenched; therefore, the emission increase we observed indicated that there is structural distortion in the homoduplex tails that extends up to 15 bp beyond the homologous region of the dsDNA. The increase in emission was eliminated by a nick situated between the fluorophores and the homologous region of the dsDNA. Interestingly, both bulk FRET and molecular dynamics simulations showed that the structural disruption in the homoduplex tails depends on N, the number of homologous bp in the D-loop. Finally, we found that the distortion in one homoduplex tail depends on the presence of a homoduplex tail on the other side of the homologous region. In sum, our results suggest that a collective interaction extending along recombinase-bound D-loops from one homoduplex tail to the other destabilizes the base pairing in the homoduplex tail regions adjacent to the D-loop. Materials and methods Preparation of dsDNAs The oligonucleotide sequences used to prepare the dsDNAs, as well as the ssDNA-RecA filaments, are listed in Supplementary Tables S1-S13. Preparations of 180 bp dsDNA constructs with fluorophores were achieved by annealing complementary oligonucleotides and subsequently ligating corresponding fragments overnight at 16°C with T4 DNA ligase in ligase reaction buffer (50 mM Tris, 10 mM MgCl2, 1 mM ATP, and 10 mM dithiothreitol, pH 7.5 (NEB)), yielding ∼180 bp dsDNA products. Purification was performed using gel electrophoresis on 3% agarose in TBE (Tris/Borate/EDTA) buffer for 2 h (6 V/cm). The bands were visualized with a midrange UV trans-illuminator. Finally, by comparison with a dsDNA ladder, the dsDNA band of the correct size was cut out and extracted from agarose using a Nucleospin kit (Machery and Nagel, Bethlehem, PA). The dsDNA products were finally concentrated using 100 kDa YM100 Amicon filters (Millipore). After purification and concentration of the products, the concentration was measured at 260 nm. The oligonucleotides used for 180 bp dsDNA with rhodamine and fluorescein labelled dsDNA positioned, respectively, 58 and 57 bases from the 5´-end (taking the rhodamine strand as reference) and the unlabelled oligonucleotides (O1-4) are shown in Supplementary Table S1. The corresponding oligonucleotides for the 180 bp rhodamine and fluorescein labelled dsDNA with labels in the middle of the construct, as well as the unlabelled oligonucleotides (O26-30), are shown in Supplementary Table S5. The short 75 or 90 bp labeled dsDNAs were obtained by heating a mixture of the corresponding oligonucleotides to 90°C and then cooling back to 40°C at 1°C per step, equilibrating for 1 min at each step shown in Supplementary Tables S2, S6, and S12. Similarly, the dsDNA constructs with an internal bubble and nicked constructs were prepared by annealing the corresponding oligonucleotides shown in Supplementary Table S8. The 306 bp dsDNA was prepared by PCR using primers shown in Supplementary Table S11, following a typical protocol for EconoTaq DNA polymerase (Lucigen). The PCR product was purified using a Nucleospin kit (Machery and Nagel, Bethlehem, PA). Strand exchange reactions DNA strand exchange in vitro was achieved by mixing 0.06 μM ssDNA-RecA filaments with 0.06 μM labeled dsDNA at 37°C. We used several dsDNAs and ssDNAs of varying lengths of homology, but with a total length of 98 nt. The sequences of the oligonucleotides purchased from Integrated DNA Technologies (Coralville, IA) are listed in Supplementary Tables S1-13. The ssDNA-RecA filaments were prepared by mixing the corresponding ssDNA with RecA (NEB) at a 3:1 ratio (bases: protein; (typically 5.3 μM (in bases) = 53 nM 98-nt ssDNA and 1.8 μM RecA), 1 mM ATPγS, and 0.4 μM single-stranded binding protein, SSB (Abcam) in RecA buffer (70 mM Tris–HCl, 10 mM MgCl2, and 5 mM dithiothreitol, pH 7.6 (NEB)). Filament formation proceeded at 37°C for 10 min. When strand exchange reactions were combined with restriction enzyme digestion, the buffer was replaced with rCutSmart buffer (50 mM KAc, 20 mM Tris-acetate, 10 mM MgAc2, and 100 mg/ml recombinant BSA, pH 7.9 (NEB)) for filament preparation and subsequent strand exchange reaction. Reactions in ATP were performed similarly; the ssDNA-RecA filament was prepared by mixing the ssDNA with RecA (NEB) at a 3:1 ratio (bases: protein), 1 mM ATP, a regeneration system containing 10 U/ml of pyruvate kinase and 3 mM phosphoenolpyruvate in RecA buffer (70 mM Tris–HCl, 10 mM MgCl2, and 5 mM dithiothreitol, pH 7.6 (NEB)). Filament formation proceeded at 37°C for 10 min. Finally, the filament aliquot was mixed with the D-loop-like construct (oligonucleotides shown in Supplementary Fig. S2). ScDmc1 protein [38] was a gift from Professor Hung Yuan (Peter) Chi (National Taiwan University). Reactions with ScDmc1 were performed in 35 mM Tris–HCl, pH 7.5, 1 mM DTT, 1 mM ATP, 2.5 mM MgCl2, 0.5 mM CaCl2, 75 mM KCl, and 0.1 mg/mL BSA. The filaments were prepared with ScDmc1 and ssDNA at a 3:1 ratio (bases: protein; (typically 5.3 μM (in bases) = 53 nM 98-nt ssDNA and 1.8 μM ScDmc1) at 37°C for 10 min. DNA strand exchange in vitro was achieved by mixing 0.06 μM ssDNA-ScDmc1 filaments with 0.06 μM labeled dsDNA at 37°C. Fluorescence measurements To measure the fluorescent signal of these strand exchange reactions, the ssDNA/RecA filament and dsDNA mixture was placed in a quartz cuvette. Using FluorEssence spectroscopy software and the automated FluoroMax spectrofluorometer (Horiba, Edison, NJ), the emission of the fluorescein label was read using a 493 nm excitation wavelength and a 2 nm slit. The fluorescent emission in counts per second (cps) was detected at 518 nm with a 2 nm slit. The reaction was run for 30 min, and emission measurements were collected every 1 second with an integration time of 0.5 s. The temperature was kept constant at 37°C. When strand exchange reactions were combined with restriction enzyme digestion, the strand exchange measurement was paused after 30 min for the addition of the restriction enzyme and finally resumed for another 30 min. Gel electrophoresis experiments Filaments of different lengths of homology and total length 98 nt (Supplementary Table S11) were prepared by incubating corresponding oligonucleotides with RecA (New England Biolabs (NEB)) at 3:1 ratio (bases: protein), 1 mM ATPγS, and single-stranded binding protein, SSB (Abcam) as described above in CutSmart buffer (NEB) (50 mM KAc, 20 mM Tris-acetate, 10 mM MgAc2, and 100 mg/ml BSA, pH 7.9) at 37°C for 10 min. After filament formation, 306-bp dsDNA with one fluorescein label at one of the 5′ ends was added and incubated for 30 min. After the strand exchange took place, the reaction products were incubated with NciI (NEB) restriction enzyme for 1 h at 37°C. Each aliquot was treated with 40 mM EDTA (final concentration 2 mM), 10% SDS (final concentration 0.5%) and 20 mg/ml proteinase K (final concentration 0.5 mg.ml) at 37°C for 10 min and finally mixed with Blue Gel loading buffer (NEB) and run on 3% agarose gel for 2 h at 60 V using Tris-acetate EDTA (TAE) buffer. The bands were visualized at 488 nm on a Sapphire Biomolecular Imager (Azure Biosystems). The bands were integrated using ImageJ [39]. Molecular dynamics (MD) simulations Models of nucleoprotein filaments with increasing D-loop lengths were constructed starting from the CryoEM structure with PDB code 7jy9 [19]. That structure is composed of 9 RecA monomers bound to a 27-nucleotide single-stranded oligonucleotide (ssDNA) and a 42-base pair (bp) dsDNA. In the PDB structure, the 42-bp dsDNA presents a 10 bp-long heterologous bubble designed to artificially stabilize a 10 bp-long D-loop resulting from local annealing of the ssDNA and the dsDNA complementary strand into a heteroduplex. We modified the sequence of the displaced strand so that the homology with the complementary strand was restored over the whole dsDNA length, so reverse base pairing exchange is possible in the D-loop. We subsequently defined a subunit formed by the central monomer and the three DNA triplets it binds to, and we used that subunit as a building block to simultaneously increase the length of the D-loop (by increments of three base levels) and the filament (by one monomer increment), using construction tools from the Heligeom module of the PTools library [40, 41]. The detailed procedure is described in [37] and follows a step-by-step building protocol similar to that described in reference [42]. This procedure produced a site I-bound heteroduplex where the two backbones were almost perfectly continuous, but minimization using the NAMD2.12 software [43, 44] was required to close the backbone of the displaced strand in site II. During minimization, the phosphate atoms on each side of the strand disruption locations, every three bases, were restrained to their initial position. All systems were prepared using the VMD auto PSF builder (VMD 1.9.2) [45]. The structures were first solvated using a TIP3P water model in boxes submitted to periodic boundary conditions in a physiological ionic concentration of 0.15 mol/l. The simulated systems contained between 435 000 and 900 000 atoms. Molecular dynamics (MD) simulations were performed on all systems using the NAMD software [44] versions 2.12 and 2.13, and the CHARMM36m force field including the CMAP correction [46] and using the Particle Mesh Ewald (PME) method to account for long-range electrostatic interactions [47]. Time steps were set to 2 fs using the SHAKE algorithm [48]. Temperature and pressure were regulated via a Langevin dynamics approach, employing a Nosé-Hoover-Langevin piston for control. Throughout the equilibration phase, the protein Cα carbon and DNA P positions were harmonically restrained to their initial coordinates, with the force constant reduced incrementally from 0.5 to 0.05 kcal·mol − 1·Å−2 during 60 ns. During the 200 ns production phase, no restraints were applied. To find the spacing between base pairs in the homoduplex tails, trajectory files were processed using the MDAnalysis Python library (version 2.9.0) [49, 50]. Distances between complementary and displaced nucleobases were computed based on atom groups defined for each residue pair listed in the triplet table. For both the complementary and displaced strands, the selected atom groups included only heavy atoms belonging to the nucleobase ring and exocyclic substituents (N1, C2, O2, N3, C4, N4, C5, C6, C5M, N6, O6, N7, C8, N9). To identify a consistent point of comparison for each pair of bases, a pairwise distance search between all base heavy atoms was performed at the first frame of the trajectory, and the atom pair with the minimum separation was recorded. For each subsequent frame, the instantaneous inter-base distance was computed as the Euclidean distance between the coordinates of this same atom pair. This yielded a time evolution of the base proximities across the full trajectory for every residue pair in the dataset from which the root mean square deviation (RMSD), representing the degree of distance fluctuation, was calculated. Results Probing disruption in homoduplex tails flanking D-loops ATP hydrolysis promotes unbinding of RecA from D-loops [10, 15], resulting in RecA-free D-loops; however, even with ATP hydrolysis in vivo, most of the homology testing occurs before the D-loops become RecA-free [10, 51]. D-loops formed in ATPγS, a slowly hydrolysable analog of ATP, do not unbind from RecA. Thus, we studied homoduplex tail disruption in RecA-bound D-loops in ATPγS [52]; however, we studied ScDmc1-bound D-loops in ATP since ATP hydrolysis does not drive unbinding of ScDmc1 from D-loops [53]. We used bulk FRET to probe the structure of the homoduplex tails in recombinase-bound D-loops by placing fluorescein and rhodamine fluorophores on the displaced and complementary strands, respectively (Fig. 2). We followed ΔF vs. time, the change in fluorescein emission as a function of time relative to a negative control. The negative control was the average fluorescein emission value for a heterologous filament after 1800 s in the presence of labeled dsDNA and ATPγS. Initially, the rhodamine label quenched the fluorescein emission; however, if the structure of the dsDNA changed in the dsDNA region containing the fluorophores, the fluorescein emission could increase. The two fluorophores were positioned d = 2 bp beyond the homologous region of the invading strand. The fluorophores monitored either the structure of the homoduplex tail on the 3′ side of the D-loop (3′ homoduplex tail) (Fig. 2A) or the structure of the homoduplex tail on the 5′ side of the D-loop (5′ homoduplex tail) (Fig. 2B). Structure distortion in the homoduplex tails flanking a D-loop with N = 35. (A). Schematic of the experiment to probe distortion in the homoduplex tail on the 3′ side of the homologous region of the invading strand. The light blue and purple strands represent the displaced and complementary strands, respectively. The blue ovals represent the recombinase monomers. The orange line indicates the region of the invading strand that is homologous to the dsDNA, and the magenta lines indicate the regions of the invading strand that are not homologous to the dsDNA. The filament is aligned so the orange region is adjacent to the homologous region in the dsDNA. The green star and red circle indicate the positions of the fluorescein and rhodamine fluorophores. The orange dashed lines extend the locations of the ends of the homologous region of the invading strand into the dsDNA. The magenta bracket indicates that the fluorescein is positioned 2 bp beyond the 3′ end of the invading strand. Initially, the fluorescein is quenched by the nearby rhodamine, as indicated by the dark green star. If D-loop formation disrupts the structure in the homoduplex tail, fluorescein emission can increase, as indicated by the light green star. (B). Same as A, but for testing the distortion on the 5′ side of the homologous region of the invading strand. (C). Change in emission relative to a negative control (heterologous filament), ΔF, vs. time for RecA-mediated homologous recombination in ATPγS. The red and blue curves show results for fluorophores on the 3′ and 5′ sides of the homologous region of the invading strand, respectively. (D). Same as C, but for ScDmc1 in ATP. When N = 35, for either RecA (Fig. 2C) or ScDmc1 (Fig. 2D), as strand exchange progressed, the fluorescein emission from each homoduplex tail increased with time, suggesting that D-loop formation distorted each of the homoduplex tails flanking the recombinase-bound D-loops. Furthermore, the distortion extended at least 2 bp beyond the homologous region of the invading strand. The increase in emission versus time indicates a change in the structure of the homoduplex tails due to D-loop formation. Importantly, we also compared FRET experiments that measured emission for fluorescein in the homologous region to the emission for fluorescein in the 3′ and 5′ homoduplex tails (Supplementary Fig. S1A). Emission from fluorophores in the homologous region measures the D-loop population. Interestingly, the emission for fluorophores in the 3′ homoduplex tail was approximately half the emission for fluorophores in the homologous region, suggesting that the distortion in the 3′ homoduplex tails is substantial. Furthermore, the results suggest that adding a 3′ heterologous region to the invading strand increases the emission for the homoduplex tails dominantly because the 3′ heterologous region increases the D-loop population (Supplementary Fig. S1A-C); In sum, the presence of a 3′ heterologous region on the invading strand does not significantly affect disruption in the homoduplex tails; therefore, not all figures in this work include both results with a heterologous 3′ region on the invading strand and results without a heterologous 3′ region on the invading strands. Distortion of the homoduplex tails flanking D-loops: dependence on RecA presence, and attempted strand exchange Supplementary Fig. S1D shows control experiments related to the results shown in Fig. 2. Those controls consider FRET experiments for the following (Supplementary Fig. S1D): 1. Heterologous filament. 2. Homologous ssDNA with no RecA. 3. RecA only, no ssDNA. 4. Experiments in ATP with fluorophores in the 3′ homoduplex tail using the invading strand with two flanking heterologous regions. None of those controls showed any significant increase in emission. The first control showed that the observed distortion depends on the formation of stable D-loops, which do not occur if the invading strand is heterologous. The second and third controls showed that neither RecA alone nor homologous ssDNA alone produced significant tail distortion, suggesting that the distortion is produced by RecA- bound D-loops formed by homologous recombination. The experiment in ATP was the same as the experiments that produced the dark red curve in Supplementary Fig. S1B, except the latter was done in ATPγS. The signal in ATPγS was significant, but the signal in ATP was negligible. Importantly, previous work suggests that with ATP hydrolysis, homoduplex tails transiently melt while D-loops are RecA-bound [8]. Since ATP hydrolysis unbinds RecA from D-loops [10, 15], the absence of distortion in ATP (Supplementary Fig. S1D) combined with previous results suggests that melting is not maintained in RecA-free D-loops [7]. In sum, only experiments in ATPγS with homologous ssDNA-RecA filaments showed sustained distortion in the structure of homoduplex tails, suggesting that the observed distortion was caused by stable RecA-bound D-loops formed by homologous recombination. Probing whether homoduplex tails flanking D-loops are available for restriction enzyme digestion Previous work had already shown that restriction enzyme digestion was inhibited outside the homologous region of the invading strand [21]. That inhibition could result from deformation of the homoduplex tail, steric hindrance due to binding of the tail to RecA, or both. Furthermore, either effect could be influenced by attempts to extend the D-loop into the heterologous region. Thus, to further probe the structure of the homoduplex tails, we followed strand exchange and subsequent restriction enzyme digestion using FRET (Fig. 3A–D), and we considered both filaments that can attempt to extend the D-loop and filaments that cannot attempt to extend the D-loop. Probing the disruption of homoduplex tails using restriction enzyme digestion and FRET. (A) Schematic of a control for restriction enzyme digestion of dsDNA, yielding the maximum possible emission. The fluorescein and rhodamine FRET pair is separated by 11 bp, and the restriction digestion site is positioned between them. (B) Same as A, but a homologous ssDNA-RecA filament interacts with the dsDNA. The light blue and purple strands represent the displaced and complementary strands, respectively. The blue ovals represent the recombinase monomers. The orange line indicates the region of the invading strand that is homologous to the dsDNA, and the magenta lines indicate the regions of the invading strand that are not homologous to the dsDNA. The filament is aligned so the orange region is adjacent to the homologous region in the dsDNA. The green star and red circle indicate the positions of the fluorescein and rhodamine fluorophores. The 98-nt invading strand has N = 82, and the fluorophores are positioned in the fully homologous region, so they should be well protected from restriction enzyme digestion. (C) Same as B, but the invading strand has N = 35 and a 3′ heterologous region as well as a 5′ heterologous region; the homologous region of the invading strand terminates 3 bp before the restriction site. (D) Same as C, but the invading strand terminates at the end of the homologous region, so there is no 3′ heterologous region. (E) ΔF vs. time curves. The magenta curve shows the results for dsDNA only (A). The green curve shows the result for the invading strand with N = 82 and the fluorophores positioned in the homologous region (B). The light blue curve shows the results for an invading strand with N = 35 for filaments with heterologous regions at each end (C). The dark blue curve shows the results for an invading strand with N = 35 for filaments with a heterologous region on the 5′ side only (D). The black curve shows results for a heterologous filament. (F) The fraction of dsDNA that was cut was calculated from the curves shown in E, considering the pink curve as complete digestion. For the experiments shown in Fig. 3, the restriction digestion site for the NciI enzyme was between rhodamine and fluorescein fluorophores in 90-bp dsDNA. A control for the efficiency of the digestion reaction is performed with 90- bp dsDNA and NciI (Fig. 3A and E, pink curve). To form stable D-loops, strand exchange was performed with a fully homologous filament that included 82 homologous bp that fully covered the restriction site and 16-nt additional heterologous bases to provide filament stability [54]. After strand exchange proceeded for 30 min, subsequent addition of NciI showed a small increase in emission (Fig. 3E, light green curve), suggesting that the restriction site within the stable D-loop was not accessible to the enzyme. Interestingly, our results suggest that even a fully heterologous filament reduced digestion by ∼40%, possibly because the filament is blocking access to the restriction site due to constant and futile attempts to form a stable D-loop (Fig. 3E, black curve). Thus, we compared the protection for restriction sites in homoduplex tails to the protection for heterologous filaments (Fig. 3E, black curve) and the protection for sites in the homologous region of stable D-loops (Fig. 3E, light green curve). Since Fig. 2 shows results for filaments with N = 35, we examined protection for restriction sites in homoduplex tails when N = 35. We compared results for two different filaments. For both filaments, the 3′ homologous end was three bases from the restriction site. One filament could attempt strand exchange into the 3′ homoduplex tail because the invading strand had a 3′ heterologous region that extended beyond the restriction site (Fig. 3C and E, light blue curve). In contrast, the other filament could not attempt strand exchange into the 3′ homoduplex tail because the invading strand did not have a 3′ heterologous region that extended beyond the restriction site (Fig. 3D and E, dark blue curve). Both filaments showed more protection than a heterologous filament (Fig. 3E and F), but the filament without a heterologous 3′ region showed less protection than the filament with a heterologous 3′ region. Thus, D-loops protect 3′ homoduplex tails even when attempted strand exchange into the tail is not possible; however, protection is enhanced if the invading strand has a heterologous 3′ region. This result is consistent with results in Supplementary Fig. S1. Importantly, steric hindrance would not affect the results shown in Supplementary Fig. S1; therefore, together Fig. 3 and Supplementary Fig. S1 suggest that D-loops produce distortions in flanking heteroduplex tails, and that though a heterologous 3′ region in the filament is not required for distortion of the 3′ homoduplex tails, the heterologous 3′ region in the filament increases distortion. Distortion into the homoduplex tails flanking D-loop-like structures not formed by recombinases Additionally, we tested dsDNA that contained a heterologous bubble to evaluate whether the intersection between three-strand DNA structures and homoduplex tails produces a significant tail distortion in the absence of RecA (Supplementary Fig. S2). When the dsDNA with the heterologous bubble interacted with ssDNA that was homologous to one of the strands in the heterologous bubble, some increase in emission was observed one bp into the dsDNA tail. In contrast, homologous ssDNA-RecA filaments produced a significantly larger increase in emission in ATPγS, but there was no signal in ATP. Furthermore, in ATPγS, homologous ssDNA-RecA filaments increased emission for fluorescein positioned 15 bp beyond the heterologous bubble in the dsDNA, but homologous ssDNA alone did not (Supplementary Fig. S2). In sum, the results in Supplementary Fig. S2 indicate that without RecA, in D-loop like structures, junctions between dsDNA and three-stranded DNA structures can distort the dsDNA tail within a few base pairs of the heterologous bubble; however, even in D-loop like structures, RecA is required for the distortion on the 3′ side of the ssDNA to extend 15 bp beyond the heterologous bubble in the dsDNA. Extent of the distortion into the homoduplex tails flanking RecA-bound D-loops Figs 2 and 3 do not indicate whether the distortion in RecA-bound D-loops created by homologous recombination extends more than a few base pairs into the homoduplex tails. Thus, we did additional experiments with fluorophores positioned d = 1, 4, 6, 8, 15, and 30 bp beyond the 3′ end of the invading strand, so strand exchange could not be attempted in that region (Fig. 4A). The dsDNA is the same for all experiments. We controlled it by changing the sequence of the invading strand. For d ≤ 15, D-loop formation increased fluorescein emission in the 3′ tail, so the effect caused by D-loop formation extended beyond a single helical turn in the homoduplex tail (Fig. 4B) even when the invading strand did not have a 3′ heterologous region that could attempt strand exchange. Fig. 4C shows graphs of ΔF values at 1800 seconds, ΔF1800, as a function of d. The ΔF1800 values in Fig. 4C are normalized to the value for d = 4. The bright red curve represents the results shown in Fig. 4B, and the pink curve shows results from additional experiments. We also probed distortion in the homoduplex tail on the 5′ side of the homologous region for various values of d (Fig. 4D). For d = 1, results for the 5′ and 3′ tails were approximately the same (Fig. 4E); however, for d ≥ 4, results for the 5′ tail (blue curve) were lower than results for the 3′ tail (bright red curve) (Fig. 4F). In sum, the results in Figs 2–4 and Supplementary Fig. S1 suggest that RecA-bound D-loops formed by homologous recombination can produce significant distortion in flanking homoduplex tails even when strand exchange cannot be attempted; however, that distortion is not maintained in RecA-free D-loops. FRET measurements of the distortion into the homoduplex tails flanking RecA-bound D-loops for different separations d between the fluorophores and the homologous region of the invading strand. (A) Schematic of the 50-nt filament and 180 bp dsDNA interaction. The light blue and purple strands represent the displaced and complementary strands, respectively. The blue ovals represent the recombinase monomers. The orange line indicates the region of the invading strand that is homologous to the dsDNA, and the magenta line indicates the region of the invading strand that is not homologous to the dsDNA. The filament is aligned so the orange region is adjacent to the homologous region in the dsDNA. The orange dashed lines extend the locations of the ends of the homologous region of the invading strand into the dsDNA. The green star and red circle indicate the positions of the fluorescein and rhodamine fluorophores. The magenta bracket indicates that the fluorescein is positioned d bp beyond the homologous region of the invading strand, where d varied from 1 to 30 bp. (B) ΔF vs. time for d = 1 (red), 4 (orange), 6 (green), 8 (blue), 15 (purple), and 30 bp (magenta). (C) Normalized ΔF1800 vs. d, including a result for d = −4, which corresponds to the emission value when fluorophores are in the homologous region. The results are normalized so that the value at d = 4 is 1. The pink and red curves represent results from different days. Error bars show rms deviation. (D) Analogous to A except showing the assay that probes distortion in the 5′ homoduplex tail as a function of the separation between the fluorophores and the 5′ end of the homologous region of the invading strand. (E) Same as B except for the assay shown in D. (F) The blue curve shows the ΔF1800 values as a function of d for the curves shown in D. The red curve repeats the d > 0 results shown in C; in F, the results are not normalized. Experiments probing the influence of the number of homologous bases in the filaments Evidence for collective interactions is provided by experiments that studied the dependence of the distortion on N, since an N-dependent distortion is a signature of collectivity. The N dependence of the disruption in the 3′ homoduplex tail was probed with filaments of several lengths of homology and 180 bp labeled dsDNA (Fig. 5, Fig. 6, and Supplementary Fig. S3). Importantly, a variation in emission with N could be either due to an increase in distortion, or an increase in D-loop population, but the D-loop populations for N = 50 and N = 82 are comparable or slightly larger for N = 50 (Supplementary Fig. S4). In contrast, for N = 82, emission from fluorescein in the homoduplex tail is significantly larger than for N = 50 (Fig. 5B and D and Supplementary Fig. S3). In sum, despite comparable D-loop populations, emission for fluorescein positioned in the homoduplex tails is significantly larger for N = 82 than for N = 50, suggesting that the distortion does depend on N, which is a signature of a collective interaction along the filament. Extent of the distortion into the homoduplex tails as a function of N for N = 50 and N = 82. (A) Schematic of the interaction between nucleoprotein filaments with N = 50 or N = 82 and 180 bp dsDNA. The light blue and purple strands represent the displaced and complementary strands, respectively. The green star and red circle indicate the positions of the fluorescein and rhodamine fluorophores. The blue ovals indicate recombinase monomers. The orange line indicates the region of the invading strand that is homologous to the dsDNA, and the magenta lines indicate the regions of the invading strand that are not homologous to the dsDNA. The filaments are aligned so the orange region is adjacent to the homologous region in the dsDNA. The green arrows highlight the boundaries of the homologous region, and the orange dashed line projects the end of the homologous region in the filament to the dsDNA. The distance between the 3′ end of the invading strand and the nearest fluorophore is d = 1 bp (magenta bracket). (B) ΔF vs. time for N = 50 (purple) and N = 82 (magenta) for fluorophores in the 3′ homoduplex tail. The result for a heterologous filament is shown in black. (C) Same as A, but for fluorophores in the 5′ homoduplex tail. (D) Same as B but for fluorophores in the 5′ homoduplex tail. Extent of the distortion into the homoduplex tails as a function of N for N = 30, 35, and 50. (A) Schematic of the interaction between 180 bp dsDNA and several different filaments. The light blue and purple strands represent the displaced and complementary strands, respectively. The green star and red circle indicate the positions of the fluorescein and rhodamine fluorophores. The blue ovals indicate recombinase monomers. The orange line indicates the region of the invading strand that is homologous to the dsDNA, and the magenta lines indicate the regions of the invading strand that are not homologous to the dsDNA. The filaments are aligned so the orange region is adjacent to the homologous region in the dsDNA. The boundaries of the orange regions are indicated by the green arrows. The orange dashed line extends the end of the homologous region of the invading strand to the dsDNA, and the magenta bracket represents the distance d between the 3′ end of the invading strand and the nearest fluorophore. (B) ΔF vs. time curves for N = 30 (orange), N = 35 (green) and N = 50 (purple) for fluorophores in the 3′ homoduplex tail. The result for a heterologous filament is shown in black. The error bars show rms deviation. (C) Same as A, but for fluorophores in the 5′ homoduplex tail. (D) Same as B but for fluorophores in the 5′ homoduplex tail. (E) Bar graph for the ratio between homoduplex tail disruption and population average calculated from restriction enzyme digestion and FRET experiments; disruption in the 3′ homoduplex tails (red) and in the 5′ homoduplex tails (blue) where the values for each measurement were normalized to the value for N = 50. For N < 50, emission from fluorescein in homoduplex tails decreases with decreasing N (Fig. 6); however, for N < 50, the D-loop population also decreases with N. Thus, to separate the N dependence of the homoduplex tail distortion from the N dependence of the D-loop populations, we performed several assays that measured the N dependence of the D-loop populations. First, we measured D-loop population using fluorophores in the homologous region (Supplementary Fig. S5A). When strand exchange progressed for 30 min, there was almost no signal at N = 20, but there was a significant signal for N ≥ 30 (Supplementary Fig. S5B). The signal increased with N until N = 50, but the result for N = 82 was the same as the result for N = 50. We then added a restriction enzyme and measured protection against restriction enzyme digestion using FRET (Supplementary Fig. S5A and C). For shorter filaments, the larger increase in emission due to restriction enzyme digestion (red and yellow curves) almost overlaps the curve for a heterologous filament (black curve). In contrast, N = 50 and 82 (green and blue curves, respectively), show more protection against digestion. Finally, we probed protection from restriction enzyme digestion using a gel electrophoresis assay to quantify the fraction of uncut dsDNA (Supplementary Fig. S5D-G). The assay was repeated 3 times. The gray line in Supplementary Fig. S6A indicates the average values for the gel electrophoresis results. The gel assay showed no significant protection until N = 20. Protection increased with N until N = 50, but protection for N = 82 was slightly smaller than for N = 50. Experiments showed that the gel electrophoresis assay result was not sensitive to the position of the fluorophores along the dsDNA (not shown), so the N dependence was not an artifact resulting from a changing distance between the ends of the homologous region and the fluorophores. Thus, all three assays suggest that the D-loop population was small for N < 20, increased with N until N = 50, and saturated for N ≥ 50 because D-loops with N ≥ 50 are irreversible [55]. In contrast, Supplementary Fig. S6B shows that fluorescein emission due to distortions in the homoduplex tails was small for N < 30 and increased strongly from N = 50 to N = 82. The bar graph in Fig. 6E summarizes the results for homoduplex disruptions normalized by the average population values calculated from restriction enzyme digestion and FRET experiments. The graphs show that the ratio of the emission in the homoduplex tails to the D-loop population increases with increasing N for N ≤ 50, so for N ≤ 50, the N dependence of emission for fluorophores in the homoduplex tails is not simply a consequence of N-dependent changes in the D-loop populations. In sum, for N = 30 to N = 82, homoduplex tail distortion increases with N, suggesting that there is a collective interaction along the D-loop. Probing effects of torsion If a collective torsional interaction between the D-loop and the homoduplex tail contributes to the structural distortion in the homoduplex tail, D-loop formation would not distort the homoduplex tail if a nick were inserted between the homologous region of the dsDNA and the homoduplex tail. Fig. 7 shows that without the nick, D-loop formation increased emission from fluorescein in the homoduplex tail; however, if a nick separated the fluorophores from the homologous region of the invading strand, then D-loop formation did not increase the emission of fluorescein in the homoduplex tail (Fig. 7A-H). Importantly, the nick did not significantly affect D-loop population (Fig. 7I and J). Thus, the observed emission reduction in emission represents a nick-induced reduction in the distortion in the homoduplex tail rather than a nick-induced reduction in the D-loop population. In sum, previous work [8] and Fig. 7 are consistent with the observed homoduplex tail distortion resulting from torsional stress due to a collective interaction that extends along a recombinase-bound D-loop from the homologous region into the homoduplex tails. Effect of a nick on the distortion in the 3′ and 5′ homoduplex tails. (A) Schematic of the experiments without a nick with a 35-nt homologous filament (total length 98 nt) and 75 bp dsDNA. (B) ΔF vs. time curves for heterologous ssDNA-RecA filament (black) and homologous ssDNA-RecA filament (blue). The error bars represent the variation between two runs. (C) Same as A but for a dsDNA with a nick in the displaced strand separating the fluorophores in the 3′ homoduplex tail from the homologous region of the invading strand. (D) ΔF vs. time curves for the experiments illustrated in C. The red and gray curves represent results for homologous and heterologous ssDNAs, respectively. (E) Same as A, but for fluorophores in the 5′ homoduplex tail. (F) ΔF vs. time curves for the experiments illustrated in E. (G) Same as E, but for dsDNA with a nick in the displaced strand between the fluorophores and the homologous region of the dsDNA. (H) ΔF vs. time curves for the experiment illustrated in G. (I) Schematic of experiments probing the effect of the nick on D-loop population for an 82-nt filament (98 nt total length) and 180 bp dsDNA; only the construct with the nick is shown. (J) ΔF vs. time curves for the experiment illustrated in I. Probing the influence of one homologous tail on the distortion observed in the other homoduplex tail If the collective interaction extends along the D-loop from one homoduplex tail to the other, then eliminating one homoduplex tail would change the observed distortion in the other homoduplex tail. Thus, we tested emission from fluorescein in one homoduplex tail with or without the other homoduplex tail. We then compared that result to the D-loop populations with and without the other homoduplex tail measured by emission from fluorescein in the homologous region. Supplementary Fig. S7A shows schematics of the assays that measure the distortion in the 3′ homoduplex tail. Supplementary Fig. 7A illustrates the case in which N = 35, with the homologous region flanked by homoduplex tails extending over at least 18 bp and an invading strand that contained 50 homologous bases. The homoduplex tails on the 3′ sides were the same for both N = 35 and N = 50, but for N = 50 at 37°C, there was no homoduplex tail on the 5′ side of the D-loop that was formed after strand exchange. Supplementary Fig. S7A shows ΔF vs. time curves for N = 35 (orange) and N = 50 (purple). Fluorescein emission due to distortion in the structure of the 3′ tail was much larger for N = 35 than for N = 50. In contrast, fluorescein emission due to fluorophores in the homologous region of the invading strand was significantly larger for N = 50 than for N = 35, indicating that the D-loop population for N = 50 is larger than the D-loop population for N = 35 (Supplementary Fig. S7B). Thus, though the D-loop population was larger for N = 50, the fluorescein emission from fluorophores in the 3′ homoduplex tail was much smaller, suggesting that the 5′ homoduplex tail increased the distortion in the 3′ homoduplex tail. Supplementary Fig. 7C and D show analogous results for the distortion in the 5′ homoduplex tail. Similarly, when N = 70 or 88, decreasing the length of one homoduplex tail suppressed the distortion in the other homoduplex tail (Supplementary Fig. S8). In sum, the distortion of the structure of one homoduplex tail was strongly enhanced by the presence of the other homoduplex tail. This result suggests that there is a collective interaction that extends along the length of the D-loop from one homoduplex tail to the other. Molecular dynamics simulations probing the influence of the number of homologous bases Finally, additional evidence for a collective interaction that extends along D-loops is shown in Supplementary Fig. S9, which shows results of 200 ns molecular dynamics (MD) simulations of D-loops containing different N values. MD simulations indicate that as a result of mechanical stress, the structures of triplets in RecA-bound D-loops depend both on the position of the triplet and the length of the D-loop [37]. The x-axis represents the number of backbone phosphates separating each base pair from the homologous region of the dsDNA. The y-axis shows the rms deviation between the bases in each pair, averaged over the 200-ns simulation time. For all N, the free ends of the dsDNA showed thermal melting. For N = 9 base pair separations, the homoduplex tails were very stable except for the thermal melting at the free ends of the dsDNA. For N = 18, there was some small instability in the homoduplex tails flanking the D-loop since base pairs with m = 4 to m = 13 show the same rms deviation of slightly less than 1 Angstrom; however, no rms deviation significantly exceeded 1 Angstrom, so there was no melting in the homoduplex tails flanking the D-loops. Importantly, when N = 54, melting extended at least 1 bp beyond the heteroduplex into the 3′ homoduplex tail, and the base pair destabilization extended at least 2 bp into the 3′ homoduplex tail. In contrast, no homoduplex tail melting was observed close to the 5′ end for any N value. The 200-ns simulation times were quite short, so longer times might have revealed more melting into the homoduplex tails. In sum, consistent with experimental results in this work, the simulations suggest that homoduplex tails in RecA-bound D-loops are destabilized by a collective interaction, and that the destabilization increases as the length of the D-loop increases. Discussion Most of the results in this work were done in ATPγS, and the experiments that were done in ATP did not observe distortion extending even 1 bp into the homoduplex tails; however, even with ATP hydrolysis, D-loops are initially RecA-bound [10], and homoduplex tails drive rapid reversal of RecA-free D-loops shorter than 160 bp [54]. Thus, we propose that with ATP hydrolysis, RecA-bound D-loops distort ∼10 bp regions in flanking homoduplex tails; however, any homoduplex tail distortion flanking a RecA-bound D-loop quickly relaxes after ATP hydrolysis creates RecA-free D-loops. Support for this suggestion is provided by previous work with ATP hydrolysis that showed homoduplex tails shorter than ≤ 10 bp melted when they flanked a 73-bp D-loop [8]; however, if the tails were extended to 15 bp, almost no melting was observed [8]. Furthermore, Fig. 4 shows that in RecA-bound D-loops, there is little emission from fluorophores positioned more than ∼15 bp beyond the homologous region. In sum, previous work and this work indicate that with or without ATP hydrolysis, Rec-bound D-loops distort ∼10 bp regions in flanking homoduplex tails; however, if unmelted regions remain in the homoduplex tails, then the D-loop distortions rapidly relax after RecA unbinds. Evidence that kinks in the homoduplex tails flanking RecA-bound D-loops could lead to dsDNA melting was also provided by earlier MD simulations [4] that considered formation of the 8-bp D-loops, which are the first metastable structures formed during homologous recombination [4, 56–58]. Those MD simulations showed that binding of the dsDNA to a homologous ssDNA-RecA filament leads to a structure in which two base pair triplets from the dsDNA are bound to the secondary binding site (site II) in RecA [4, 5]. After those two base pair triplets bind, a base duplet from each site II-bound triplet melts, relocates, and forms a base pair duplet with two bases in the invading strand. Subsequently, the kink in the homoduplex tail on the 3′ side of the invading strand promotes melting of a base duplet beyond the two base triplets that were bound to site II. That melted duplet then transfers its base pairing to a duplet in the invading strand [4, 5]. In sum, MD simulations showed that during the formation of a RecA-bound 8-bp D-loop, the kink in the 3′ homoduplex tail promotes dsDNA melting and strand exchange; however, it was not clear whether homoduplex tail melting played a role in the extension of a D-loop beyond 8 bp. Progress of D-loops through short heterologous regions due to base substitution is important in vivo [15], and in vitro experiments showed that D-loops do incorporate ≤ 4-bp substitutions without ATP hydrolysis [13] and ≤ 10-bp substitutions with ATP hydrolysis [7]. We propose that after formation of the metastable 8-bp product, RecA-bound D-loops can incorporate short substitutions because RecA extends the complementary strand so that it is in registration with the RecA-bound invading strand in the homologous region beyond the substitution [13]. In contrast, insertions or deletions in the sequence of one strand can block RecA-bound D-loop extension because the RecA-bound complementary and invading strands are not in registration in the homologous region beyond the insertion or substitution [13]. In the following, we illustrate how the extension of the complementary strand allows recombinase-bound D-loops to incorporate short substitutions within D-loops without forming a heteroduplex within the short substitution. To highlight the importance of the extension of the complementary strand, first consider the hypothetical case in which the complementary strand is not extended. After formation of an 8-bp metastable D-loop, if the complementary strand is not extended, then the 1.5 x extension of the invading strand in recombinase-bound D-loops [4, 5, 35–37] implies that the invading and complementary strands will not be in registration in the homologous region beyond the substitution. Thus, if the complementary strand is not extended, the D-loop cannot progress into a homologous region beyond a short substitution (Fig. 8A-C). Illustration of the proposed mechanism for the extension of D-loops promoted by homoduplex tail melting in the kinks, allowing 6-bp heterologous regions to be easily bypassed if they are flanked by sufficient homology. (A) The blue, purple, and orange colors correspond to the displaced, complementary and invading strands. Initially, the dsDNA is in B-form, but the invading ssDNA is extended to 1.5x the B-form length due to binding to RecA. (B) The dsDNA has bound to RecA, and a short D-loop has formed, resulting in homoduplex kinks on both sides of the heteroduplex. The red dots indicate site II residues that are bound to the displaced strand backbone. The black bases indicate the heterologous insert, and the green bases indicate the beginning of the flanking homology. The green arrow indicates the bases in the homologous triplet in the invading strand. The difference in extension between the RecA-bound ssDNA and the dsDNA requires that the dsDNA extends in order to base pair with the invading strand. The 3′ side of the D-loop is flanked by a 6-bp heterologous region that cannot form a stable heteroduplex, so the difference in extension between the invading strand and the 3′ homoduplex tail prevents the D-loop from extending beyond the heterologous insert into the flanking homologous region. (C) The D-loop reverses. (D) Same as A. (E) Same as B. (F) The kink melts the homoduplex tail. (G) The displaced strand can bind to site II and extend to 1.5 times the B-form length. Since the complementary strand is still bound to the displaced strand in the homoduplex tail, the complementary strand also extends enough to allow the flanking homologous region to form a heteroduplex product. Progress through the flanking homologous region then stabilizes the D-loop despite the heterologous insert. In contrast, if the complementary strand does extend, then the complementary and invading strands can be in registration in a homologous region beyond the substitution. We propose that the complementary strand extends because site II binding extends the displaced strand backbone. In one scenario, the dsDNA melts first, and then site II binding extends the displaced strand (Fig. 8D-F) [4, 5]. That extension will then stretch the complementary strand if the bases remain paired in a homoduplex region beyond the melted region. Alternatively, homoduplex dsDNA could extend and untwist due to site II binding of the displaced strand backbone. In that case, the differential extension of the displaced and complementary strands would promote rapid melting of the site II-bound dsDNA [4]. In either scenario, the result is a displaced strand that is extended by site II binding and a complementary strand that is not base-paired and is in registration with the invading strand [4] (Fig. 8G). If the complementary strand extension extends through the substitution, the D-loop can form a stable heteroduplex in the homologous region even though there is no base pairing in the substitution. Finally, we speculate that the D-loop region in which no heteroduplex is formed is limited by mechanical stress ∼10 bp [4, 5]. In sum, we speculate that in vivo, collective interactions that extend through recombinase-bound D-loops from one homoduplex tail to the other allow strand exchange to exploit melting in flanking homoduplex tails to facilitate D-loop extension and allow D-loops to progress through regions of contiguous heterology extending over up to 10 bp [7]. Importantly, in vivo homology testing involves ∼75 bp [59, 60], and previous theoretical work has shown that homology testing ∼75 bp allows highly mismatch-tolerant recombinases to achieve the same DSB repair accuracy as homology testing that rejects all mismatches [51], so extending D-loops through short regions of heterology does not decrease the accuracy of DSB repair. Supplementary Material