Mechanism of Rad51 filament formation by Rad52 and Rad55-Rad57 in homologous recombination

Abstract

Homologous recombination (HR) repairs double-stranded DNA breaks (DSBs) by generating single-stranded DNA (ssDNA), which is initially coated by Replication Protein A (Rpa). Rad51, a recombinase, catalyzes strand invasion but binds ssDNA with lower affinity than Rpa, necessitating mediator proteins like Rad52 (yeast) or BRCA2 (humans) for Rad51 loading. The mechanisms of this exchange remain unclear. We show that Saccharomyces cerevisiae Rad52 uses its disordered C-terminus to sort polydisperse Rad51 into discrete monomers. Using fluorescent-Rad51 and single-molecule optical tweezers, we visualize Rad52-mediated Rad51 filament formation on Rpa-coated ssDNA, preferentially at ssDNA–dsDNA junctions. Deleting the C-terminus of Rad52 disrupts Rad51 sorting and loading. Addition of the Rad51 paralog Rad55–Rad57 enhances Rad51 binding by ~60%. Despite structural differences, Rad52 and BRCA2 share conserved functional features. We propose a unified “Sort, Stack & Extend” (SSE) mechanism by which mediator proteins and paralogs coordinate Rad51 filament assembly during HR.

The mediator protein Rad52 promotes Rad51 binding onto RPA-coated DNA to initiate homologous recombination. Here, the authors show that Rad52 sorts Rad51 into monomers and stacks the complex on to the ss-dsDNA junction. The Rad55-Rad57 paralog then promotes extension of the Rad51 filament.

Introduction

Accumulation of DNA damage results in genomic instability and is a hallmark of most cancers¹. Double-stranded DNA breaks (DSBs) are particularly harmful as they give rise to chromosomal aberrations and deletions². Homologous recombination (HR) is a key DNA repair pathway that protects cells from the harmful genomic instability effects introduced by DSBs^{3, 4–5}. In HR, a DNA break is preferentially repaired using the sister chromatid as a template. HR is orchestrated through a sequence of DNA break recognition, processing, homology search and recognition, and resolution steps that are catalyzed by over three dozen proteins^{3, 4–5}. In Saccharomyces cerevisiae, the DSB is first recognized and resected to yield 3ʹ ssDNA overhangs, which are subsequently sequestered by the ssDNA binding protein Replication Protein A (Rpa) to serve as a nucleoprotein platform for the recruitment of downstream HR enzymes^6,7. These steps are often called presynapsis. Rad51 is the recombinase that catalyzes homology search and pairs the correct complementary sequences—a process called strand exchange. Rad51 binds to the Rpa-coated ssDNA (nucleation) and forms an ATP-dependent cooperative helical filament (growth). Formation of the pre-synaptic Rad51 filament is a key regulatory event in HR.

One of the impediments to spontaneous Rad51 binding to the resected ssDNA arises from the high affinity of Rpa-ssDNA interactions (K_D < 1 nM) compared to Rad51 binding to ssDNA (K_D ~ 1 µM)^8,9. Thus, several mediator proteins and Rad51-paralogs function to regulate pre-synaptic Rad51 DNA binding/nucleation and filament growth during HR. While pro-HR factors promote the nucleation and growth of the Rad51 filament on Rpa-coated ssDNA, anti-HR factors suppress HR by removing Rad51 molecules from the DNA³. In yeast, Rad52 is the key pro-HR mediator and a functional analog of the human BRCA2 protein¹⁰. However, it should be noted that Rad52 also has other unique roles in HR such as single-strand DNA annealing^11,12. Rad51 filaments are further stabilized by Rad51 paralogs such as Rad55-Rad57 and the Shu complex made of Csm2-Psy3-Shu1-Shu2 subunits¹³. The paralogs are thought to intercalate or cap the ends of the Rad51 filament and protect them from disassembly by anti-HR mediators such as Srs2^{14, 15, 16, 17–18}. Anti-HR mediators are primarily translocases or helicases that utilize ATP-dependent physical movement on ssDNA to remove Rad51^5,19. A key anti-HR mediator in yeast is the Srs2 helicase that physically moves on the ssDNA to strip Rad51^{19, 20, 21, 22–23}. While vast efforts have been undertaken to decipher the functions of mediator proteins and Rad51-paralogs in HR, knowledge of where they bind, how they interact with each other and Rad51, and where they are positioned on the DNA is poorly resolved.

In this study, we focus on the mechanism of Rad52-promoted Rad51 filament formation on Rpa-coated DNA. We recently uncovered that yeast Rad52 is a homo-decameric ring with each subunit possessing an ordered N-terminal and disordered C-terminal half²⁴. The N-terminal region promotes oligomerization of Rad52 and ssDNA interactions, whereas the C-terminal region harbors the Rad51 and Rpa interaction motifs^25,26. Interestingly, we also uncovered additional interactions with Rad51 through the ordered N-terminal ring, but these were asymmetric and captured alongside only one Rad52 subunit²⁴. Thus, we defined two modes of Rad52-Rad51 interactions: Mode-1 interactions occur between Rad51 and the disordered C-terminus of Rad52 and Mode-2 interactions are enacted between Rad51 and the ordered N-terminus of Rad52. But the functional significance of these modes of interaction was not clear.

Several groups have attempted to visualize and capture the dynamics of yeast Rad51 filament formation. Single-molecule DNA curtain assays have been useful in deriving information about binding and dissociation of Rad51, Rad52, Rpa, and Rad51-paralog proteins during pre-synaptic filament formation on long ssDNA substrates^{15,27, 28–29}. However, Rad51 was never directly visualized as attempts to fluorescently label the protein rendered it inactive. Nevertheless, the binding and dissociation signal from fluorescent Rpa served as a reasonable proxy for Rad51 filament dynamics^{29, 30, 31, 32–33}. Liu et. al. recently described a strategy to generate super-folder green fluorescent proteins (sfGFP)-tagged Rad51 (Rad51^GFP) using flexible linkers engineered within the N-terminal non-conserved region. This protein engineering strategy generated a functionally active version of Rad51 in vivo.

Here, using recombinant Rad51^GFP and derivative versions (Rad51^Cy5), we directly visualize Rad51 binding events in single-molecule high-resolution optical trap-based tweezers and fluorescence imaging. Using structural mass spectrometry and biophysical approaches, we define the functions of both modes of Rad52-Rad51 interactions. Mode-1 sorts Rad51 into monomers, whereas mode-2 stacks Rad51 at a single position on Rad52. On Rpa-coated ssDNA carrying junctions, the Rad52-Rad51 complex preferentially binds close to the ss-dsDNA junction, but no extension or growth of the Rad51 nucleoprotein is observed. However, the addition of the Rad55-Rad57 (Rad51-paralog) complex drives further recruitment of Rad51 molecules. Deletion of the C-terminus of Rad52 results in a complete loss of Rad51 binding to the Rpa-coated DNA. The work visualizes and uncovers the mechanism of how Rad52 and Rad55-Rad57 function to catalyze the formation of Rad51 filaments during HR pre-synapsis.

Results

S. cerevisiae Rad51 is polydisperse in solution

Most recombinases including bacterial RecA and human RAD51 are oligomers in solution and cooperatively interact with DNA to form ATP-bound nucleoprotein filaments^{34, 35, 36, 37–38}. Structural studies have revealed the interactions between S. cerevisiae Rad51 protomers within a filament in the absence or presence of DNA (Fig. 1A, B)^37,39. The filament is assembled by docking a conserved FVTA (FxxA) motif into a hydrophobic pocket in the RecA-like domain of the neighboring Rad51 protomer (Fig. 1B)³⁹. To obtain a quantitative measure of the polydispersity of yeast Rad51 under our reaction conditions, we assessed the distributions at both low and high concentrations using mass photometry (MP) and analytical ultracentrifugation sedimentation velocity (AUC^SV), respectively. Both experiments reveal a wide distribution of oligomeric species ranging from monomers to decamers, with the major fraction being dimers (Fig. 1C, D). To further understand the nature of contacts between Rad51 molecules, we performed crosslinking mass spectrometry (XL-MS) analysis using suberic acid bis 3-sulfo-N-hydroxysuccinimide ester (BS3), which crosslinks primary amines within an ~12–24 Å distance⁴⁰. More than 50 crosslinks (XLs) are observed within Rad51 (Fig. 1E), which represents both intra- and inter-Rad51 contacts. The extensive crosslinking observed agrees with our sedimentation and MP results that show heterogeneous states of Rad51 in the absence of DNA and/or ATP. No XLs were identified in the N-terminal disordered region of Rad51 due to the absence of Lys residues in this region. Moreover, this region is unique to S. cerevisiae Rad51 and not conserved in other recombinases.

[See PDF for image]

Fig. 1

S. cerevisiae Rad51 is polydisperse in solution.

A Schematic of Rad51 denoting the positions of the Walker A and B motifs for ATP-binding and hydrolysis. The FVTA motif promotes Rad51 oligomerization. Loops 1 and 2 in site 1 and residues in site 2 are critical for DNA binding. The N-terminal disordered region (NDR), N-terminal lobe domain (NLD), and RecA-like domains are also denoted. B Structure of the Rad51 filament in the absence of DNA (PDB:1SZP). The monomeric units are colored gray and orange. Phe-144 and Ala-147 dock into defined pockets of the adjacent Rad51 protomer to promote oligomerization. C Mass photometry analysis of Rad51 (100 nM) shows a wide range of oligomers ranging from monomers through decamers, with dimers being the predominant species. D Analytical ultracentrifugation sedimentation velocity analysis of Rad51 at a higher concentration (10 µM) also shows a similar distribution of oligomers. E Crosslinking mass spectrometry (XL-MS) analysis of Rad51 using BS3 crosslinker shows extensive intra- and inter-Rad51 crosslinks. Source data are provided as a Source Data file.

Crosslinking mass spectrometry reveals interactions between Rad51 and two distinct binding sites in Rad52

Rad52 physically interacts with Rad51 through two motifs in the disordered C-terminus of Rad52 (Fig. 2A). Using single particle cryogenic electron microscopy (CryoEM), we recently identified another site of interaction between Rad51 and the N-terminal ordered ring of Rad52²⁴. Given the lower resolution of the structural data, the precise motifs that promote this interaction were not identified. Thus, to better understand the details of these two modes of binding, we performed XL-MS analysis of the Rad52-Rad51 complex in the absence of DNA. Rad52 alone shows extensive XLs within the N- and C-terminal halves, and between the two halves (Fig. 2B). These data agree with the structural and hydrogen-deuterium exchange mass spectrometry analysis performed earlier²⁴. To ensure saturation of all Rad51 binding sites on the Rad52 decamer, a 1:10 molar ratio of Rad52 (decamer):Rad51 was used to assemble the Rad52-Rad51 complex. Distinct sets of XLs are captured between Rad51 and the two binding regions in Rad52 (Fig. 2C). For clarity, we define these as two modes of interactions: mode-1 and mode-2. The mode-1 interaction site(s) reside in the C-terminus of Rad52 and crosslink to two regions in Rad51 (residues 100–160 and 300–380). In contrast, the mode-2 interaction site in the N-terminal half of Rad52 predominantly interacts with one region in Rad51 (residues 270–380), although a couple of XLs to a region around residue 120 are also observed (Fig. 2C). It should be noted that BS3 XL-MS only reports on regions that carry primary amines and should be used as a coarse-grain view of the interactions between Rad52 and Rad51. To get a structural perspective of these interactions, we used AlphFold3 (AF) to predict a model of the Rad52-Rad51 complex (Supplementary Fig. 1a, b). In agreement with our results, AF models two regions of contact between Rad52 and Rad51. Mode-1 interactions map to the C-terminal half of Rad52, whereas mode-2 interactions reside in the N-terminal half of Rad52 (Fig. 2D).

[See PDF for image]

Fig. 2

The C-terminus of Rad52 sorts Rad51 into monomers.

A Schematic of Rad52 showing the ordered N-terminal and disordered C-terminal halves of Rad52. The Rpa and Rad51 binding motifs reside in the C-terminal half. B Crosslinking mass spectrometry (XL-MS) analysis of Rad52 reveals extensive crosslinks within the N-terminal (purple) half, the C-terminal (cyan) half, and between the two halves (green). C XL-MS analysis of the Rad52-Rad51 complex reveals intra-Rad52 crosslinks as shown in (B), in addition to two sets of crosslinks between Rad51 and Rad52. Crosslinks between the N-terminal half of Rad52 and Rad51 are shown in blue, while crosslinks between the C-terminal half of Rad52 and Rad51 are shown in red. Interestingly, almost all of the inter- and intra-Rad51 crosslinks observed in the Rad51 alone XL-MS analysis are absent when in complex with Rad52. Only three crosslinks within Rad51 are captured (orange). D AlphaFold3 prediction of the complex between one Rad52 subunit and Rad51. Residues in Rad51 that are proposed to mediate mode-1 interactions with Rad52 are shown in red. Residues that mediate mode-2 interactions are shown in blue and reside in the L1 and L2 loops of Rad51. E Rad52^ΔC, lacking the C-terminal half of Rad52, retains interactions with Rad51. XL-MS analysis of this complex reveals the reappearance of the inter- and intra-Rad51 crosslinks (orange), suggesting that Rad51 forms oligomers when bound to Rad52^ΔC. Thus, the Rad51 sorting properties arise from interactions with the C-terminus of Rad52 (mode-1). Intra-Rad52^ΔC crosslinks are denoted in purple, and those with Rad51 are shown in blue. Source data are provided as a Source Data file.

The C-terminus of Rad52 sorts Rad51 into monomers

One striking observation in the XL-MS analysis of the Rad52-Rad51 complex is the loss of almost all intra- and inter-Rad51 crosslinks (Fig. 2C) which were observed in the XL-MS of Rad51 alone (Fig. 1E). Only three XLs within Rad51 were captured in the complex (shown in orange in Fig. 2C) and these mapped to the subunit-subunit interface in the Rad51 filament (Supplementary Fig. 1C). In the AF-model, the Rad51 binding site in the C-terminal half of Rad52 binds across the Rad51-oligomerization interface (Fig. 2D). This interaction is reminiscent of the interactions between the BRC repeats of human BRCA2 and human RAD51 and promotes monomerization (Supplementary Fig. 2)⁴¹. In yeast Rad52, binding mode-1 in the C-terminus is mediated by two known Rad51 interaction patches—316-FVTA-319 and 337-FDPK-340^26,42,43, with Phe-319, Ala-319, and Phe-337 docking into hydrophobic pockets in Rad51 (Supplementary Fig. 2A). Similar Phe residues are conserved in the BRC repeats of BRCA2. However, only the crystal structure of the BRC4-Rad51 complex has been solved⁴¹ (Supplementary Fig. 2B). Thus, we tested whether this region in Rad52 promotes monomerization of Rad51 by deleting the C-terminal half (Rad52^ΔC). In XL-MS analysis of the Rad52^ΔC-Rad51 complex XLs between the two proteins are observed (Fig. 2E). Interestingly, the intra- and inter-Rad51 XLs reappear in the Rad52^ΔC-Rad51 complex, suggesting that the Rad51 monomerization (sorting) property is lost in this mutant. In AUC^SV analysis, Rad52^ΔC binds to Rad51 in agreement with contacts mediated by mode-2 interactions within the N-terminal half of Rad52²⁴ (Supplementary Fig. 3). However, unlike the Rad52-Rad51 complex that produces defined species, the Rad52^ΔC-Rad51 complex is polydisperse. These data agree with the reappearance of the XLs between Rad51 molecules in the Rad52^ΔC-Rad51 complex.

To directly assess the Rad51 sorting function of the C-terminus of Rad52, we generated a C-terminal fragment of Rad52 that encompass the two Rad51-binding motifs (Rad52^ΔN*; Fig. 3A). Based on the XL-MS analysis, we hypothesized that this fragment would shift the equilibrium of Rad51 in solution from oligomers to monomers (Fig. 3B). In MP, Rad52^ΔN* interacts with Rad51 in a concentration-dependent manner and sorts Rad51 into monomeric units (Fig. 3C–E, & Supplementary Fig. 4). These data show that mode-1 interactions are driven by the C-terminal half of Rad52 and function to sort Rad51 into monomers. We had previously shown that Rad51 interacts with the ordered N-terminal ring of Rad52, and this mode-2 interaction occurs in an asymmetric manner within a single subunit in the ring²⁴. A recent study from Ma et al.⁴⁴ further supports these conclusions by showing that an 85-residue segment in the C-terminus of Rad52 is required to chaperone the oligomerization and filament formation properties of Rad51. They report that strains carrying a rad52-F316A or rad52-F337A substitution in the Rad51 binding patches lead to a 33- and 21-fold reduction in Rad51 loading, respectively. Thus, based on our findings, we propose sort and stack features for Rad52 where mode-1 interactions sort Rad51 and mode-2 interactions likely deposit the sorted Rad51 molecules at a defined position on Rad52 towards loading onto the resected ssDNA during the pre-synaptic phase of HR.

[See PDF for image]

Fig. 3

The Rad51-sorting function is mediated by mode-1 interactions and facilitated through the disordered C-terminus of Rad52.

A Schematic of the two Rad51 interaction motifs in the disordered C-terminus of Rad52. Rad52^ΔN* is a construct that possesses both these motifs, and B is able to sort Rad51 from a polydisperse to a monodispersed species in solution. Mass photometry analysis of C Rad51, D Rad52^ΔN*, and E the Rad51-Rad52^ΔN* complex. Rad51 is polydisperse, with monomeric to decameric species observed in solution. Rad52^ΔN* is a single lower molecular weight species in solution. It should be noted that the predicted mass is below the detectable limit of the mass photometry methodology. The Rad51-Rad52^ΔN* complex shows a loss of the higher order Rad51 species and accumulation of a predominantly single species that corresponds to one Rad51 molecule bound per Rad52^ΔN*. Source data are provided as a Source Data file.

Direct visualization of Rad51 binding and filament formation on ssDNA

A method to directly visualize and quantify Rad51 binding onto DNA is required to decipher how mode-1 and mode-2 interactions drive Rad52-promoted Rad51 filament formation. Many attempts by our group and others to generate fluorescently labeled yeast Rad51 using traditional approaches (amino-terminal labeling, maleimide-coupling, or through incorporation of non-canonical amino acids) resulted in loss of activity. Recent work by Liu et al. demonstrated that engineering an sfGFP at position 55 in the disordered and non-conserved N-terminal region of Rad51 through two additional 16 aa linkers (Fig. 4A, B) produced a functional recombinase in vivo⁴⁵. Recombinantly purified sfGFP-Rad51 (Rad51^GFP) shows ssDNA binding, D-loop formation, and ATPase activities similar to Rad51 (Fig. 4C–E, and Supplementary Fig. 5).

[See PDF for image]

Fig. 4

Direct visualization of Rad51 binding events on single-stranded DNA.

A Schematic of the sfGFP-Rad51 (Rad51^GFP) construct. The sfGFP is engineered at position 55 in the non-conserved N-terminal region of S. cerevisiae Rad51. Two 16 aa flexible linkers flank the sfGFP and are required for maintaining Rad51 functionality. B An AF-model of the Rad51^GFP protein. sfGFP and Rad51 are colored green and orange, respectively. The two 16 aa linkers are colored gray. C Rad51 and Rad51-variants support the formation of D-loops. Cy5-labeled ssDNA was incubated with Rad51/Rad51-variant, followed by the addition of Rpa and Rad54+dsDNA, and incubated as depicted. D Image of the gel showing D-loop formation. The smear observed in the lanes corresponding to the Rad51^Cy5 sample arises from the fluorescence signal of the protein (Supplementary Fig. 5c). E Quantification of the percent D-loops formed shows all Rad51 proteins retaining similar levels of D-loop formation activity. F Design of the optical trap assay to visualize Rad51^GFP binding to ssDNA. A ~ 48.5 knt ssDNA was tethered to biotin handles, bound to streptavidin beads, and captured by two optical traps. The trapped DNA is then sequentially moved to channels containing Rad51^GFP in the absence or presence of ATP. No Rad51 binding is observed in the absence of ATP, but robust Rad51^GFP binding to ssDNA is observed in the presence of ATP. G Rad51^GFP binding to ssDNA is dependent on ATP concentration. Data points from multiple Rad51 binding events measured across several tethers (N) for each ATP concentration are shown, and the SEM are plotted as error bars. N = 13, 5, 13, and 16 tethers for experiments with 0.1, 0.34, 1, and 2 mM ATP, respectively. Source data are provided as a Source Data file.

To investigate the mechanism of Rad51 nucleoprotein filament formation, we performed confocal fluorescence imaging of Rad51^GFP on ssDNA under mechanical tension using a LUMICKS C-Trap. This instrument combines dual optical traps with confocal scanning fluorescence microscopy with single-fluorophore sensitivity⁴⁶. Long ssDNA was created by stretching lambda phage dsDNA (~48.5 kbp) that was tethered using streptavidin-biotin linkages on the same strand (Fig. 4F). One strand was mechanically stretched until the complementary strand fell off due to high tension^47,48. Formation of ssDNA was confirmed by fitting to a Freely Jointed Chain Model (FJC)⁴⁹. Rad51^GFP was then allowed to bind to the ssDNA under 5 pN tension for 10–20 s before imaging. A 2D scan of the designated imaging area was performed to locate the confocal plane of the bound protein, followed by a line-scan along the protein-bound region over time, creating a kymograph (Fig. 4F). Rad51 filament formation experiments were carried out either in the channel containing Rad51^GFP, allowing new binding events to occur over time, or by binding in the Rad51^GFP channel followed by imaging in a separate channel to reduce background noise. As expected, Rad51^GFP does not bind to ssDNA in the absence of ATP, but forms filaments in an ATP concentration-dependent manner (Fig. 4F, G). These observations agree with the well-known ATP-dependent DNA binding properties of yeast Rad51^{50, 51–52}. We observed clear binding events of Rad51^GFP with signals resembling multimers. To quantify the multimeric state of each filament, we compared the signal intensity of a monomer with the initial signal intensity of the multimer to calculate the number of molecules in the filament (Supplementary Fig. 6). The signal of the multimer/filament decreased over time, likely due to a combination of photobleaching and Rad51 dissociation. Thus, the initial signal intensity was used as a proxy for establishing the starting size of the filament.

Rad52 sorts Rad51 to enhance binding on non-Rpa-bound ssDNA

To investigate how Rad52 promotes the formation of Rad51 filaments, we measured Rad51^GFP binding to ssDNA when in complex with Rad52. Using the optical trap setup described above, we moved the tethered ssDNA into the channel containing Rad51^GFP in the absence or presence of Rad52, incubated for 20 s, and then imaged and recorded the frequency of Rad51^GFP filament formation per ssDNA tether (Fig. 5A). We found that the presence of Rad52 increased the number of Rad51 multimer binding events (Fig. 5A–C, G). When these experiments were performed on an Rpa-coated ssDNA, no Rad51^GFP binding was observed (Fig. 5D). The Rad52^ΔC protein that lacks the mode-1 interaction motifs does not stimulate Rad51^GFP binding (Fig. 5E). Finally, under these Rad51 concentration regimes (100 nM) we observe poor dsDNA binding activity for Rad51^GFP (Fig. 5F, G). The lack of dsDNA binding activity agrees with the ~7-fold lower affinity of yeast Rad51 to dsDNA (K_D ~ 2.6 µM) compared to ssDNA (K_D ~ 0.36 µM)³⁸. These data show that the sorting function of Rad52 promotes the formation of Rad51 filaments on Rpa-coated ssDNA.

[See PDF for image]

Fig. 5

Rad52 enhances Rad51 binding to ssDNA.

A Schematic of the optical trap experiment showing Rad51^GFP binding to a ~ 48.5 knt ssDNA substrate in the absence or presence of Rad52. Kymographs of optical trap data in the B absence or C presence of Rad52 shows stimulation of Rad51^GFP binding to ssDNA. D Kymograph showing no binding of Rad51^GFP to an Rpa-coated ssDNA in the absence of Rad52. E Rad52^ΔC, which lacks the C-terminal Rad51 and Rpa interaction domains, poorly promotes Rad51^GFP binding to ssDNA. F Under these single-molecule experimental conditions (100 nM Rad51), low levels of Rad51^GFP binding to dsDNA are observed. G Quantitation of Rad51^GFP filaments formed under the denoted conditions. Representative data from a minimum of n = 5 tethers per condition is shown, and the SEM is plotted as error bars. An unpaired t test was performed between the ±Rad52 conditions, and ** denotes p = 0.0098. Source data are provided as a Source Data file.

Rad52 preferentially stacks Rad51 at the ss-dsDNA junction on Rpa-coated ssDNA

The DNA used in the optical trap experiments described above is an ~48 knt ssDNA substrate. However, the resected DNA substrate during HR pre-synapsis possesses both an ss-dsDNA junction and a 3ʹ free end. The technical needs of the optical trap require the DNA to be tethered between two beads, which prevents us from being able to use a substrate with a free 3ʹ end. Therefore, we utilized a modified gapped substrate that, when stretched, produces an ~5 knt ssDNA gap in the middle (Fig. 6A). An Atto-647 fluorophore is positioned within the dsDNA region to infer polarity of the DNA and the two junctions. The Rad51^GFP protein is fully functional in supporting Rad51 filament formation in vivo⁴⁵ and retains ssDNA binding, D-loop formation, and ATPase activities in vitro at levels similar to the unmodified wild-type protein (Fig. 4 & Supplementary Fig. 5). To be certain that our observations do not arise due to some unforeseen artifacts of the linkers and/or the sfGFP, we generated an alternate fluorescent version of Rad51 by engineering a ybbR tag⁵³ at the same position as sfGFP (residue number 55). However, this modification does not require the use of additional linkers. In vitro, both Rad51^GFP and ybbR-tagged Rad51 interacts with Rad52, binds to ssDNA, hydrolyzes ATP, and form D-loops similar to untagged Rad51 (Fig. 4, Supplementary Figs. 5 and 7). A Cy5 fluorophore was attached to the ybbR tag using 4ʹ-phosphopantetheinyl transferase (Sfp synthase) and coenzyme-A conjugated Cy5 (Rad51^Cy5)⁵³.

[See PDF for image]

Fig. 6

Rad51 preferentially localizes to the recessed Rpa-coated ss-dsDNA junctions in the presence of Rad52.

A Schematic of the optical trap experiment where the DNA is first preincubated with MB543-labeled-Rpa (Rpa^MB543; 0.25 nM) and then moved to a channel containing fluorescent Rad51 (either Rad51^Cy5 or Rad51^GFP; 50 nM) in the presence of Rad52 (5 nM decamer) and ATP (5 mM). An Atto-647 fluorophore is embedded within one arm of the dsDNA handle (red) to help define polarity. Stacks of frames (kymograph) were recorded by continuous confocal scanning along the DNA axis. The kymographs show binding of either B Rad51^Cy5 (magenta) or C Rad51^GFP (blue) onto the Rpa^MB543-coated ssDNA gap (green) in the presence of Rad52 and ATP. The scale bars for the length of the DNA (y axis) and time (x axis) are denoted for each kymograph. D Photon counts for Rad51^Cy5 (magenta) and Rpa^MB543 (green) were quantified and plotted versus the position of DNA in the kymograph. Rad51^Cy5 preferentially binds to the 5′-recessed junction when in complex with Rad52. In addition, Rad51 and Rpa binding events are not mutually exclusive, suggesting higher-order complex formation between Rpa, Rad52, and Rad51 at this junction. E Similar junction-preferential binding of Rad51^GFP (blue) is observed under these conditions. Compilation of photon count data from experiments using either F Rad51^Cy5 (n = 9 independent tethers) or G Rad51^GFP (n = 13 independent tethers) shows binding events at both junctions and internal regions on the ssDNA gap, but with a strong bias for the 5′-recessed junction. The SEM is plotted as error bars. Source data are provided as a Source Data file.

We initiated the optical trap experiments by first coating the ssDNA with fluorescent Rpa (site-specifically labeled with MB543 on DNA binding domain D; Rpa^MB543)^33,54,55 in one channel and then moving the Rpa^MB543-bound ssDNA substrate to the next channel containing either Rad52-Rad51^Cy5 (Fig. 6B) or the Rad52-Rad51^GFP complex (Fig. 6C). Optical-trap experiments were carried out using 50 nM or 100 nM Rad51^GFP/Cy5 and either stoichiometric or sub stoichiometric amounts of Rad52, relative to the Rad51^GFP/Cy5 concentration used. We make several key observations: The Rad52-Rad51^GFP/Cy5 complex binds close to the ss-dsDNA junctions and at internal positions on the Rpa-coated ssDNA regions with a modest preference for the 5′-recessed position (Fig. 6D–G). The average photon count for Rad51^GFP/Cy5 binding events is brighter close to the junctions relative to the internal binding events (Fig. 6F, G). This behavior is better captured and quantified in experiments with Rad51^Cy5 (Fig. 6B and Supplementary Fig. 8) compared to Rad51^GFP. This is due to bleed through of the GFP and MB543 signals when Rad51^GFP and Rpa^MB543 are used (Fig. 6C). On overlay of the Rad51^Cy5 and Rpa^MB543 intensities as a function of position on the DNA reveals likely interactions of both Rad51 and Rpa with the Rad52 molecule residing close to the junction (Fig. 6D and Supplementary Fig. 8). In addition, we do not observe complete Rpa eviction within regions of Rad51 engagement. However, this interpretation is likely limited by the resolution of the methodology. When binding of Rad51^Cy5 was tested in the presence of Rad52^ΔC, no signal for Rad51^Cy5 binding was observed, suggesting that the C-terminus of Rad52 is required for junction localization (Supplementary Fig. 9A–C). The C-terminus of Rad52 possesses an Rpa-binding motif in addition to the Rad51 mode-1 interaction and sorting functions uncovered in this study (Fig. 2A). Rad52^ΔC does not interact with Rpa in AUC^SV analysis (Supplementary Fig. 9D, E). Thus, the defect in Rad51 binding we observe might arise from a lack of interaction of Rad52^ΔC with Rpa, and/or loss of Rad51 sorting. In addition, we do not observe binding of Rad51^GFP/Cy5 to the dsDNA handles. This is likely due to the lower concentrations of Rad51 used (permissible) for these assays, and the lack of binding is influenced by the higher K_D for dsDNA³⁸.

The recruitment of Rad52-Rad51 onto Rpa-coated ssDNA is independent of ATP

Binding of yeast Rad51 to DNA and filament formation are strictly coupled to ATP^8,38. Interestingly, when the optical trap experiments were carried out in the absence of ATP, we saw binding of Rad52-Rad51^GFP complexes onto Rpa-coated DNA (Supplementary Fig. 10). Binding events were observed at both junction-proximal and at internal sites along the ssDNA. In contrast to experiments carried out in the presence of ATP (Fig. 6), the distribution appears to have a lower bias for junction-proximal regions. In addition, the overall Rad51^GFP photon counts in the absence of ATP (Supplementary Fig. 10E) are nearly half as observed in the presence of ATP (Fig. 6G). In addition, the presence of excess Rad51 does not influence these observations (Supplementary Fig. 10F). These data suggest that the ability of Rad51 to engage DNA or reorganize in the presence of ATP, when still in complex with Rad52, might be an integral feature of Rad51 handoff/nucleation onto DNA.

Higher density Rad51 filaments are formed in the presence of Rad55-Rad57

In the above experiments, the Rad52-Rad51 complex preferentially binds to the junctions, but we do not observe robust growth of the Rad51 filament from the initial event as evidenced by the sudden appearance of the bright Rad51^GFP/Cy5 intensity, but no further increase in the fluorescence track length (Fig. 6 & Supplementary Fig. 8). Even under conditions where we had a two-fold excess of Rad51 over Rad52, no filament growth was observed (Supplementary Fig. 11). Thus, we next tested whether the addition of a Rad51-paralog would stimulate filament growth. Rad51 paralogs possess structural similarity to the Rad51 recombinase, but do not catalyze strand exchange activity and do not form filaments^14,16. However, they are proposed to cap the ends of the Rad51 filament towards inhibiting ATP turnover by Rad51 and thereby helping to catalyze Rad51 filament growth⁵⁶. Genetic studies support this model, as deletion of Rad51 paralogs in yeast leads to loss of recombination and increased sensitivity to DNA-damaging agents^{14,17,57, 58, 59, 60, 61–62}. In yeast, the Rad55-Rad57 complex is the major Rad51-paralog necessary for HR, as both rad55Δ, or rad57Δ cells are defective for recombination^58,63. In DNA curtain single molecule experiments on ssDNA, Roy et. al.^14,16showed that Rad55-Rad57 does not interact with the Rpa-ssDNA complex but assembles with Rad51 (at excess Rad51 concentrations). Rad55-Rad57 enhances the ability of Rad51 to form filaments on ssDNA and promotes faster Rpa displacement¹⁶. To explore this further, we utilized Rad55-Rad57 heterodimer recombinantly produced with GST and MBP tags fused to Rad55 and Rad57, respectively. Basal expression on dextrose media of affinity-tagged versions of Rad55 and Rad57 complements the DNA damage sensitivity of the double deletion mutant rad55Δ rad57Δ; however, galactose overexpression is toxic to cells (Supplementary Fig. 12). Thus, we mixed the Rad52-Rad51^GFP complex with Rad55-Rad57 and visualized events on our Rpa^MB543-coated gapped DNA substrate in the optical trap setup (Fig. 7A).

[See PDF for image]

Fig. 7

Rad55-Rad57, the Rad51-paralog, increases the number of Rad51 molecules bound during nucleation and an increase in filament length.

A Design of optical trap experiments to investigate the effect of Rad55-Rad57 on Rad51 filament formation. Rpa^MB543-coated (green; 0.25 nM) gapped ssDNA were moved to a channel containing Rad51^GFP (blue; 100 nM), Rad52 (5 nM decamer), and Rad55-Rad57 (50 nM) in the presence of ATP (5 mM). B Kymograph from a single tether shows preferential binding of Rad51^GFP close to the 5′-recessed junction. C Photon count of Rad51^GFP binding events along the DNA in the kymograph shown in B and D, across multiple individual tethers (n = 15 tethers) shows broader coverage of the ssDNA gap with Rad51^GFP. E Quantitation of the change in Rad51^GFP count (15 tethers, 46-50 traces), or F the number of Rad51^GFP molecules bound, shows an increase in Rad51 molecules in the presence of Rad55-Rad57 compared to Rad52 alone (15 tethers, 46-50 traces). G Binding of more Rad51 molecules in the presence of Rad55-Rad57 results in an increase in the length of the filament by ~0.1 μm (15 tethers, 28-40 traces). The SEM is plotted as error bars. Two-tailed unpaired t-tests were performed and * and ** denote p-values of 0.0314 (E), 0.0296 (F), and 0.0026 (G), respectively. Source data are provided as a Source Data file.

In the presence of Rad55-Rad57 and Rad52, we observe robust binding of a cluster of Rad51^GFP molecules close to the 5′-recessed junction (Fig. 7B–D & Supplementary Fig. 13), similar to the scenario without the Rad51-paralog. However, there is an additional increase in the number of Rad51 molecules bound, as denoted by the increase in Rad51^GFP photon counts (Fig. 7E, F). The coverage or track length of the Rad51^GFP intensity also appears to increase, which reflects an overall increase in the filament length by ~0.1 µm (Fig. 7G). However, it should be noted that an absolute estimation of filament length cannot be performed due to the lower threshold of the diffraction limit of the confocal microscope associated with the optical trap ( ~ 0.2–0.25 µm).

Discussion

The formation of Rad51 nucleoprotein filaments during pre-synapsis is an essential step to drive HR and is promoted by mediator proteins such as Rad52 in yeast and BRCA2 in humans⁴. Since Rpa binds to ssDNA with higher affinity compared to Rad51, mediator proteins function to overcome this thermodynamic barrier. Elemental to this process are physical interactions of the mediator protein with Rpa, Rad51, and the DNA. However, knowledge of how these binding properties and assembly of the protein complexes enact Rad51 filament formation is poorly resolved. Here, we uncover several functional features of the yeast Rad52 mediator protein that showcase functional similarities to human BRCA2.

Rad52 is a homodecamer and each subunit possesses Rpa and Rad51 binding motifs. Our data show that Rad51 interactions are driven by two distinct binding modes²⁴. Mode-1 is enacted by a Rad51 binding site in the disordered C-terminal region of Rad52, whereas mode-2 is driven through Rad51 interactions with the N-terminal ordered ring structure of Rad52. Here, we show that mode-1 interactions shift Rad51 from a polydisperse complex to monomers. This function resembles the Rad51-monomerization properties of the BRC repeats of BRCA2^41,64,65. In structural studies, mode-2 interactions are enacted in an asymmetric manner with Rad51 molecules stacked alongside one Rad52 subunit in the ring²⁴. Mediator proteins such as BRCA2 have been suggested to sequester the recombinase as a mechanism to keep them inactive and to suppress spurious interactions with DNA and prevent inhibition of other DNA metabolic processes⁶⁶. Similarly, in prokaryotes, stacks of RecA oligomers were observed as dead-end structures, and it was proposed that these must first disassemble into smaller oligomeric forms prior to active filament formation⁶⁷. Whether Rad52 serves such protective roles in yeast remains to be established.

Our data show that the Rad52-Rad51 sorted complexes are preferentially loaded in the vicinity of the ss-dsDNA junction. In the context of pre-synapsis (based on classical definitions of this event), one could interpret this experimental observation as nucleation of Rad51. However, in the optical trap investigations, the Rad52-Rad51 complex binds to the Rpa-coated ssDNA in the absence or presence of ATP. Since Rad51 does not bind to the ssDNA in the absence of ATP, the first step in this mechanism might be the binding of the Rad52-Rad51 complex to Rpa on the ssDNA without the need for Rad51 engagement onto DNA. However, a stronger bias for recognizing the junction is observed in the presence of ATP. This could be interpreted as either a subsequent structural reorganization of the ATP-bound Rad51 molecules complexed to the Rad52 decamer and/or some form of interaction with the ssDNA. The physical interaction between Rad52 and Rpa at this step is essential as Rad52-Rad51 binding is lost when the C-terminus of Rad52 is deleted (Supplementary Fig. 9). This data supports the idea that the initial engagement of the Rad52-Rad51 complex is not dictated by Rad51-ssDNA binding in the context of an Rpa-coated DNA.

When excessive amounts of Rad51 are used in biophysical studies, Rad52 is not required to exchange Rpa on ssDNA substrates^14,16,29,33. When such competition experiments are performed on the gapped DNA substrate used in this study, we also observe that a 20-fold excess of Rad51 will rapidly outcompete Rpa bound to the ssDNA region (t1/2 ~ 200 s⁻¹; Supplementary Fig. 14). In vivo experiments have shown that overexpression of Rad51 in a Δrad52 strain does not restore recombination⁶⁸. In terms of the cellular levels of expression, a 3 to 6-fold higher number of Rad51 molecules/cell is produced compared to Rad52^69,70. Therefore, based on a conservative estimate of the in vivo concentration of Rad52 at ~20 nM^{71, 72–73} a 6-fold excess of Rad51 translates to ~120 nM. Therefore, we reason that the optical trap measurements and conclusions presented here are a better reflection of physiological conditions.

The experiments performed in the absence of ATP also raise an intriguing question about how junction-proximal localization of the Rad52-Rad51 complex is accomplished. Since Rad51 binds to DNA only in the presence of ATP, and neither Rpa nor yeast Rad52 possesses dsDNA binding activity (Supplementary Fig. 15), the engagement of this complex at the junction is likely driven by physical interactions between Rpa and Rad52. But such a model does not explain why Rpa molecules bound at other positions along the ssDNA do not engage the Rad52-Rad51 complex with similar propensity. One possibility is that either Rad52 or Rpa possesses DNA melting activity at the junction, and the melted DNA structure provides a preferred binding location for the Rad52-Rad51 complex. Another possibility is that the configurational arrangement of the DNA binding and protein interaction domains of the Rpa molecule bound at or close to the junction is different compared to the other Rpa molecules bound downstream on the ssDNA regions. However, these possibilities need to be experimentally investigated. These processes are likely further complicated by the presence of chaperone proteins such as Rtt105 that can influence both Rpa and Rad51 interactions with ssDNA^{48,74, 75, 76–77}. For example, in the absence of ssDNA, Rtt105 sequesters Rpa and blocks physical interactions with Rad52⁴⁸. In the presence of ssDNA, Rtt105 is remodeled to release the domains of Rpa, thereby promoting Rad52 binding. In addition, Rad52 engages the ssDNA-Rpa complex by selectively remodeling DNA binding domain-D⁵⁴, which would reside away from the junction due to the defined 5′- > 3′ binding polarity of the domains of Rpa⁷⁸. While the optical trap data presented here gives a bird’s eye view of the binding events at the junction, details of the underlying mechanisms are certainly more complex and need to be deciphered.

Finally, in the presence of ATP, we do not see rapid directional growth of the Rad51 filament on Rpa-coated ssDNA. This property does not change even when excess Rad51 is introduced in the reaction (Supplementary Fig. 11). When Rad55-Rad57 is introduced to the reaction (along with Rad52), we see an ~60% increase in the number of Rad51 molecules bound. However, this does not occur as a gradual increase in the Rad51 signal intensity over time as one would expect from a sequential addition of Rad51 molecules to a growing nucleoprotein filament (Fig. 7). Instead, the initial binding event is brighter, suggesting that Rad55-Rad57 enhances either the number of Rad51 molecules bound to Rad52, or multiple Rad52-Rad51 entities are recruited in the presence of the Rad51-paralog. Consequently, there is also an increase in the length of the filament in the presence of the Rad51-paralog.

Based on these findings, we propose a sort, stack, and extend (SSE) model for mediator and paralog catalyzed formation of Rad51 filaments during HR (Fig. 8). In this model, Rad52 engages a polydisperse population of Rad51 and sorts it into monomers through mode-1 interactions. These interactions are likely not uniform across all the Rad52 subunits in the decamer, as cryoEM studies show selective stacking of Rad51 adjacent to one subunit in the ring through mode-2 interactions. Based on the optical trap data, we suggest that the next step in this model is extension, where Rad51 molecules are handed off from Rad52 onto the DNA in the presence of Rad55-Rad57. However, many mechanistic details in this model need to be parsed. For example, does Rad55-Rad57 form a complex with Rad52 and Rad51 and arrive at the junction? Does Rad55-Rad57 intercalate at positions within the growing filament? If so, how are these binding events differentiated at the junction versus at internal sites? Both Rad55 and Rad57 are ATPases, and how ATP is utilized to promote Rad51 filament growth needs to be established. In addition, how asymmetry is introduced within Rad52 and how the sorted Rad51 molecules are passed on from Rad52 onto the DNA remains to be resolved. Finally, the big question of how the junction is marked, and by whom, remains to be established. We propose that configurational rearrangements of the domains of Rpa likely impart an identity to the junction, which is recognized through physical interactions between Rad52 and Rpa⁵⁴.

[See PDF for image]

Fig. 8

A sort, stack, and extend (SSE) model for mediator and Rad51-paralog-promoted Rad51 filament formation on Rpa-coated ssDNA during HR.

A sort, stack, and extend model for events in presynapsis is depicted. Rad51 is polydisperse in solution and is sorted into defined monomeric units by Rad52. There are two modes of interaction between Rad52 and Rad51. Mode-1 sorts Rad51 into monomers, whereas mode-2 is asymmetric and occurs at one position in the Rad52 ring. The Rad52-Rad51 complex engages the Rpa-coated resected ssDNA during pre-synapsis, and preferential binding to the ss-dsDNA junction is proposed. However, Rad52 is not sufficient to promote Rad51 filament growth. Rad51-paralogs (Rad55-Rad57) promote filament growth/extension. The binding positions of Rad52 and Rad55-Rad57 are speculative and need to be experimentally established.

While yeast Rad52 and human BRCA2 are functional analogs, they share no structural similarity. However, the SSE mechanism that we describe here for Rad52 shares many similarities with BRCA2. BRCA2 interacts with RAD51 through eight BRC repeats (1–8)^64,79. The BRC repeats promote RAD51 binding to ssDNA while suppressing dsDNA interactions^{41,79, 80, 81, 82–83}. BRC1 and BRC4 have been shown to shift RAD51 to monomers⁴¹. Finally, a TR2 region in the C-terminus of BRCA2 interacts with filamentous RAD51 on DNA⁶⁵. Thus, the BRC repeats could be envisioned as the sorter while the TR2 region could be the stacker for pre-synaptic RAD51 filament formation. Recent structural and single-molecule studies on the human RAD51 paralog RAD51B-RAD51C-RAD51D-XRCC2 (BCDX2) support the idea that this complex will cap one end of the RAD51 filament and drive the recruitment of more RAD51 molecules at nucleation^84,85. These findings are similar to our observations for Rad55-Rad57 in this study and highlight an overall conservation in the mechanisms of action between the yeast and human proteins. Interactions between BRCA2 and the BCDX2 complex have also been uncovered⁸⁶. Optical-trap studies of BRCA2 also show preferential localization towards the ss-dsDNA junction on RPA-coated DNA^80,82,87. In addition, no growth of the RAD51 filament is observed in these BRCA2 optical trap studies. Thus, it is reasonable to assume that mediator proteins alone are not sufficient to facilitate filament growth, but function to localize a high concentration of monomeric Rad51 molecules in the vicinity of the ss-dsDNA junction. It is very likely that the addition of the BCDX2 RAD51-paralog complex to BRCA2 and RAD51 will result in growth of the RAD51 filament from the junction. Thus, the SSE mechanism proposed here is likely conserved between lower and higher eukaryotes and possibly even in prokaryotic HR.

Methods

Plasmids

sfGFP-Rad51 and ybbR-Rad51 were synthesized as codon-optimized ORFs and engineered into RSF-Duet1 plasmids (Genscript Inc.). The sfGFP-Rad51 construct was designed as described⁴⁵ and the sfGFP was engineered with flanking 16 aa linkers at the 55th amino acid of the Rad51 coding region. The ybbR tag⁵³ was also engineered at the 55th position in the Rad51 coding region, but no additional linkers were required to generate active protein. Please see the Supplementary Information for amino acid composition of both constructs (Supplementary Fig. 16). The S. cerevisiae Rad52-pTXB1 plasmid (New England Biolabs Inc.) coding for an in-frame C-terminal chitin binding domain (CBD) tag was a kind gift from Dr. Eric Greene (Columbia University). Rad52^ΔC, Rad52^ΔN, and Rad52^ΔN* were generated as described²⁴. The GST-Rad55 MBP-Rad57 plasmid was modified from Bashkirov et al.⁸⁸ to replace the N-terminal His₆ tag of Rad57 with a maltose binding protein (MBP) tag separated by a PreScission Protease cleavage site.

Proteins

S. cerevisiae Rad51, sfGFP-Rad51, and ybbR-Rad51 proteins were recombinantly overproduced using BL21(DE3)pLysS cells and purified using a previously described method⁸. A single transformant was grown in LB broth supplemented with 50 µg/mL kanamycin at 37 °C to an OD₆₀₀ of 0.5–0.7. Protein overexpression was induced by the addition of 0.4 mM IPTG at 18 °C. After overnight induction at 18 °C, the cells were harvested and suspended in resuspension buffer (100 mM Tris-Cl, pH 8.0, 5 mM EDTA pH 8.0, 1 M NaCl, 1 M Urea, 5 mM beta-mercaptoethanol (β-ME), 10% w/v sucrose, 1X protease inhibitor cocktail, and 10% v/v glycerol). All downstream steps were carried out at 4 °C. Cells were first lysed with 0.4 mg/ml lysozyme for 30 min and sonicated for a total of 2 min in batches of 30-second ON-OFF cycles. The lysate was spun at 38,772 × g for 60 min and the clarified lysate was subjected to 40% w/v ammonium sulfate precipitation (0.24 g/mL) on ice. The solution was subsequently spun at 12,298 xg for 60 min. The pellet was resuspended in buffer Q⁰ (20 mM Tris-Cl, pH 7.5, 1 M Urea, 10% v/v glycerol, 0.5 mM EDTA, pH 8.0, 1 mM β-ME, and 1X protease inhibitor cocktail. The resuspended lysate was fractionated over a fast-flow Q-Sepharose column (Cytiva Inc.) and eluted with Q¹⁰⁰-Q⁷⁰⁰ gradient (superscripts denote the NaCl concentration [mM] in the respective buffers). Fractions containing Rad51 were pooled and diluted with H⁰ buffer (30 mM Tris-Cl, pH 7.5, 0.5 mM EDTA pH 8.0, 0.5 mM β-ME, and 10% v/v glycerol) to match the conductivity value of a H¹⁰⁰ buffer and fractionated over a fast-flow Heparin column (Cytiva Inc.) equilibrated with buffer H¹⁰⁰. Rad51 was eluted with a linear gradient of H¹⁰⁰-H¹⁰⁰⁰. Heparin fractions containing Rad51 were concentrated to ~5 mL using an Amicon Ultra spin concentrator (Sigma Inc.). The concentrated Rad51 was loaded onto a Hi-Load 16/600 Superdex 200 pg size exclusion column (Cytiva Inc.) equilibrated with buffer (20 mM Tris-Cl, pH 7.5, 0.5 mM EDTA pH 8.0, 1 mM β-ME, 10% v/v glycerol, and 100 mM NaCl). Fractions containing Rad51 were pooled and dialyzed against storage buffer (20 mM Tris-Cl, pH 7.5, 0.5 mM EDTA, pH 8.0, 1 mM β-ME, 20% v/v glycerol, and 100 mM NaCl), flash-frozen using liquid nitrogen, and stored at −80 °C. Protein concentrations were measured using the following extinction coefficients ε₂₈₀: 11,920 M⁻¹ cm⁻¹ for Rad51 and ybbR-tagged Rad51, and 33,810 M⁻¹  cm⁻¹ for sfGFP-Rad51.

Sfp-synthase was overproduced using a pBAD-MBP-His-SFP plasmid (a kind gift from New England Biolabs - Addgene #141141). The plasmid was transformed into BL21(DE3)pLysS cells, and transformants were grown in LB broth supplemented with 50 µg/mL ampicillin at 37 °C to an OD₆₀₀ of 0.6. Protein overexpression was induced by the addition of 0.2% (v/v) L-arabinose. After overnight induction at 16 °C, the cells were harvested and suspended in resuspension buffer (500 mM NaCl, 5 mM imidazole, and 20 mM Tris-Cl, pH 8.0). All downstream steps were carried out at 4 °C. Cells were lysed with 0.4 mg/ml lysozyme and sonicated for a total of 2 min in batches of 30-second ON-OFF cycles. The lysate was spun at 38,772 × g for 60 min. The clarified lysate was first batch-bound to 5 ml Ni-NTA resin (Gold Biotechnology Inc.) for 3 h. The beads were subsequently washed with resuspension buffer, followed by stepwise elution using resuspension buffer containing 100, 200, or 400 mM imidazole, respectively. The eluate containing Sfp-synthase was then concentrated to ~5 mL using an Amicon Ultra spin concentrator and fractionated over a Hi-Load 16/600 Superdex 200 pg size exclusion column equilibrated with S200 buffer (20 mM Tris-Cl, pH 8.0, 500 mM NaCl, 10% v/v glycerol, and 1 mM β-ME). Please note that Sfp-synthase elutes as two distinct fractions on SEC. While both fractions contain clean protein, the lower molecular mass peak is catalytically active and used for labeling reactions described in this study. Fractions containing Sfp-synthase were flash-frozen using liquid nitrogen, and stored at −80 °C. Sfp-synthase concentration was measured spectroscopically using ε₂₈₀ = 28,880 M⁻¹ cm⁻¹_.

Saccharomyces cerevisiae Rad52, Rad52^ΔC, Rad52^ΔN, and Rad52^ΔN* proteins were purified as described²⁴. Unlabeled and MB543-labeled S. cerevisiae Rpa were purified as described^33,54,55,89. For this study, Rpa labeled in the DNA binding domain D⁵⁵ was used in the optical trap experiments. S. cerevisiae Rad54 was purified as a functional GST-fusion protein as described (ref. ⁸⁷). S. cerevisiae GST-Rad55 MBP-Rad57 was overproduced and purified as a heterodimeric complex similar to a previously described method⁸⁸. The plasmid was transformed into a protease deficient yeast strain with rad55 and rad57 deletions WDHY 6008 (MATα ura3-52 trp1 leu2-∆1 his3-∆200 pep4::HIS3 prb1-∆1.6 R can1 rad55::hphMX rad57::LEU2). Transformants were grown in minimal media (6.7 g/l yeast nitrogen base without amino acids, 0.87 g/L dropout mixture without uracil, 2% sodium lactate, and 3% glycerol) at 30 °C for 24 h from OD₆₀₀ 0.2 to 1.5. Protein overexpression was induced by the addition of galactose to 2% (w/v) at 18 °C for 24 h. Collected cells were frozen in liquid nitrogen and stored at −80 °C. All subsequent steps were carried out at 4 °C in the presence of protease inhibitors (1 mM PMSF, 2 μM leupeptin, 4 μM pepstatin A, and 4 mM benzamidine) except where noted. 80 g of cells were thawed in lysis buffer (25 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 50 mM NaCl, 10% glycerol, 1 mM DTT, 2% isoamyl alcohol, 2 mM ATP, and 4.5 mM MgCl₂) and lysed by glass beads (0.5 mm, Biospec) using a Biospec Bead Beater using six cycles of 30 s of bead beating followed by 2 min in an ice bath. The cell debris was separated from the lysate by centrifugation at 95,834 × g for 60 min. The clarified lysate was subjected to 45% w/v ammonium sulfate precipitation (0.258 g/mL) on ice for 2 h and centrifuged at 13,870 × g for 20 min. The pellet was resuspended in Buffer A (25 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 50 mM NaCl, 10% glycerol, 5 mM DTT, 2 mM ATP, and 4.5 mM MgCl₂) for 30 min and centrifuged at 13,870 × g for 20 min to remove any insoluble protein. The resuspended lysate was batch-bound to 5 mL of glutathione agarose (ThermoFisher Scientific USA Inc.) overnight. In a column the resin was washed with Buffer A followed by Buffer B (25 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 10% glycerol, 0.01% IGEPAL CA-630, 1 mM DTT, 2 mM ATP, and 4.5 mM MgCl₂) with 2 M NaCl and Buffer B with 50 mM NaCl. Fractions were eluted by Buffer B with 500 mM NaCl and 20 mM glutathione. Fractions containing Rad55-Rad57 were applied to a 1 mL amylose resin column (New England Biolabs, Inc.) and washed with Buffer B containing 2 M NaCl, Buffer B containing 50 mM NaCl, and Storage Buffer (25 mM Tris-HCl pH 7.5, 0.1 mM EDTA, 50 mM NaCl, 10% glycerol, and 1 mM DTT without protease inhibitors). Fractions were eluted by Storage Buffer with 20 mM maltose. Fractions containing Rad55-Rad57 were pooled, and buffer exchanged using an Amicon Ultra filter unit (Sigma Inc.) with Storage Buffer without maltose. Rad55-Rad57 was concentrated to 0.5 mL, and aliquots were flash-frozen using liquid nitrogen and stored at −80 °C. The protein concentration was calculated using the extinction coefficient ε₂₈₀ = 168,000 M⁻¹ cm⁻¹. Rad55-Rad57 was purified free of detectable ssDNA and/or dsDNA nucleases, assayed as previously described⁹⁰. No DNA contamination was detected with an absorbance ratio A_{280 nm}/A_{260 nm} of 1.3.

Serial dilution assay

The Rad55-Rad57 plasmid and its empty vector were transformed into a wild-type W303 strain WDHY 2220 (MATα ade2-1 can1-100 his3-11,15 leu2-3,112 trp1-1 RAD5 ura3::loxP) and rad55 rad57 deletion strain WDHY 4251 (MATα ade2-1 can1-100 his3-11,15 leu2-3,112 trp1-1 RAD5 ura3::loxP rad55::hphMX rad57:LEU2). The transformants were grown for 6 h to log phase, and the OD₆₀₀ was adjusted to 0.4. 9 μL of seven 5-fold serial dilutions were plated on synthetic deficient media lacking uracil, containing either 2% dextrose or 2% galactose and 0% or 0.01% methyl methanesulfonate (MMS, Sigma Inc.). Plates were grown at 30 °C and imaged after 3 days.

Rad51 labeling with CoA-Cy5

The CoA-Cy5 conjugate was synthesized as described⁵³. CoA-Cy5 was generated by preparing a 1000 µL reaction by adding 1.6 mg CoA-trilithium salt (Sigma Inc.) in 750 µL 100 mM Na₃PO₄ pH 7.0 to a solution containing 1 mg Cy5-maleimide dye in 250 µL DMSO. This reaction mixture was stirred at room temperature in the dark for 60 minutes. The mixture was then passed through a C18 reversed-phase HPLC column (Avantor-VWR Inc.). Peak fractions carrying the CoA-Cy5 conjugate were collected and stored at −20 °C. For the labeling reaction, 5 µM CoA-Cy5, 5 µM ybbR-Rad51, and 0.1 µM SFP synthase were mixed in a 500 µl reaction in labeling buffer (50 mM HEPES, pH 7.5, and 10 mM MgCl₂). The reaction was incubated at room temperature for 30 min in the dark and then passed through a Biogel P-4 column (Bio-Rad Inc.) equilibrated with buffer (20 mM Tris-Cl, pH 8.0, 500 mM NaCl, and 10% v/v glycerol). The final protein concentration of Rad51-Cy5 and the labeling efficiency were calculated with the ε₂₈₀ = 11,920 M⁻¹ cm⁻¹ for the protein and ε₆₅₀ = 250,000 M⁻¹ cm⁻¹ for Cy5. In addition, the value at 280 was adjusted using a correction factor of 0.03 to account for the contribution of the Cy5 to the absorption signal. The Rad51-Cy5 protein was flash-frozen using liquid nitrogen and stored at −80 °C. During experiments, the Rad51-Cy5 protein was not subject to repeated freeze-thaw cycles.

Crosslinking mass spectrometry

XL-MS experiments and analyses were carried out as described²⁴. Briefly, stock solutions of Rad52 and Rad51 were diluted to 1.8 mg/mL and 2.2 mg/mL, respectively, in buffer (30 mM HEPES, 200 mM KCl, pH 7.8) and incubated together for 30 min. Crosslinking reaction and sample processing were conducted in a similar fashion as described before²⁴. Briefly, stock solutions of Rad52^ΔC and Rad51 were diluted to 0.9 mg/mL and 1.5 mg/mL, respectively, in buffer (30 mM HEPES, 300 mM NaCl, pH 7.8) and incubated together (Rad52^ΔC to Rad51 molar ratio as 1:10) for 5 min on ice. The diluted proteins were incubated with primary amine reactive 3 mM (final) bis(sulphosuccinimidyl)suberate (Sigma), and 20 µL of the sample was taken at various time points (0, 15, and 30 min) and immediately quenched with 2 µL of 1 M ammonium acetate (Sigma). Quenched samples were diluted with 4× Laemmli gel loading buffer to a final volume of 30 µL, boiled for 5 min and then resolved on 4–20% (w/v) gradient SDS-PAGE gels (Bio-Rad) with Tris-glycine buffer. To visualize protein, gels were stained with Gelcode blue safe protein stain (ThermoFisher Inc.). The major protein bands from 15 min time point were excised, briefly destained, and proteins were reduced with 10 mM dithiothreitol (Sigma) for 30 min at 56 °C and then alkylated with 50 mM iodoacetamide (Sigma) for 25 min at room temperature in the dark. Next, proteins were digested with 100 ng porcine sequencing grade modified trypsin (Promega) overnight at 37 °C. LCMS experiments were performed as described before⁹¹. Briefly, tryptic peptides were separated by reverse phase XSelect CSH C18 2.5µm resin (Waters) on an in-line 150  × 0.075 mm column using an UltiMate 3000 RSLCnano system (ThermoFisher Inc.). Peptides were eluted using a 60 min gradient from 98:2 to 40:60 solvent A:B ratio (solvent A = 0.1% formic acid, 0.5% acetonitrile; solvent B = 0.1% formic acid, 99.9% acetonitrile). Eluted peptides were ionized by electrospray (2.4 kV) followed by mass spectrometric analysis on an Orbitrap Eclipse Tribrid mass spectrometer (Thermo). MS data were acquired using the FTMS analyzer in profile mode at a resolution of 120,000 over a range of 375 to 1400 m/z with advanced peak determination. Following HCD activation, MS/MS data were acquired using the FTMS analyzer in profile mode at a resolution of 15,000 over a range of 150 to 2000 m/z with a stepped collision energy of 27–33%. Raw data files were converted to mgf format using ProteoWizard 3.0⁹² and then uploaded to the Spectrum Identification Machine 1.5.6⁹³ for crosslink identification. Crosslinking patterns between bound and unbound proteins were compared across the same migration range in the gel, monomers to monomers, dimers to dimers, etc.

Mass photometry

MP measurements were carried out on a TwoMP instrument (Refeyn Inc.), which was allowed to warm up for an hour prior to the experiments. Glass coverslips (No. 1.5H thickness, 24 × 50 mm, VWR Inc.) were cleaned by sequential sonication in isopropanol and deionized water for 15 min, respectively. Cleaned coverslips were dried under a stream of nitrogen. The molecular weight standard (β-amylase, with three species of 56, 112, and 224 kDa, respectively), Rad51, Rad52^∆N, and Rad52^∆C were diluted in the same buffer comprised of 50 mM Tris-acetate, pH 7.5, 50 mM KCl, and 5 mM MgCl₂. β-amylase was diluted to a final concentration of 200 nM, while Rad51, Rad52^∆N, and Rad52^∆C were diluted to a final concentration of 100 nM. Three different stoichiometries of both Rad52^∆N and Rad52^∆C were tested separately (1:1, 3:1, and 5:1) with a fixed concentration of Rad51. For these experiments, 500 µL reactions were prepared in the same buffer with the relevant ratios of Rad52^∆N/Rad52^∆C and Rad51 concentrations. For our experiments, we used a constant concentration of 100 nM for Rad51 against 100, 300, and 500 nM Rad52^∆N/Rad52^∆C. The reactions were spun down on a table-top centrifuge for one minute and then allowed to incubate at room temperature for precisely 20 min. The buffer was also maintained at room temperature throughout the analysis. For each measurement, a clean coverslip with a 6-well silicone gasket was placed onto the oil-immersion objective and samples were added into each well as described below. After aligning the laser approximately to the center of the first well, 15 µL of the buffer was added, and focus was obtained. Then, 1 µL of the 200 nM β-amylase was added to the same well (final concentration of 12.5 nM) to obtain high-resolution data with a significant number of particles (~2500) over a 60 s recording. This data was used to perform mass calibration. The experimental samples were similarly added to their respective wells after finding focus with 15 µL of the buffer. For data analysis, single-particle landing events were identified and converted to mass units using the above-mentioned standard calibration, extracted from videos, and non-linear least squares fitted with Gaussian mixture model to quantify the underlying populations using the Refeyn DiscoverMP software.

Analytical ultracentrifugation

Sedimentation velocity experiments were performed with an Optima analytical ultracentrifuge (Beckman-Coulter Inc.) using An-50Ti rotor at a speed of 40,000 rpm at 20 °C. Samples were prepared by dialyzing against 30 mM HEPES, pH 7.8, 100 mM KCl, 10 % glycerol, 1 mM TCEP and concentrations of each component of samples are mentioned in the figure. Sample (380 µL) and buffer (400 µL) were filled in each sector of 2-sector charcoal quartz cells. Absorbance was monitored at 280 nm for samples using absorbance optics, and scans were obtained at 3 min intervals. The density and viscosity of the buffer at 20 °C were calculated using SEDNTERP⁹⁴. The continuous distribution (c(s)) model was used to fit the AUC data by SEDFIT program^95,96.

Co-relative optical-trap experiment and analysis

A 48.5 kilobase (Kbp) Lambda DNA was prepared with three biotins on the 3′ and 5′ ends of the same strand. Lambda DNA was purchased from New England Biolabs (N3011). This linear DNA comes with 12 nucleotides (nt) 5′ overhangs. Three short oligos to anneal to the sticky ends were purchased from Integrated DNA Technologies Inc. Oligo 1: 5′-GGG CGG CGA CCT GGA CAA-3′; oligo 2: 5′-AGG TCG CCG CCC TTT TTT T/iBiodT/T/iBiodT/T/iBiodT/-3′; oligo 3: 5′-/iBiodT/T/iBiodT/ T/iBiodT/T TTT TTT AGA GTA CTG TAC GAT CTA GCA TCA ATC TTG TCC-3′. The DNA was stored in TE buffer (10 mM Tris-HCl, pH 8.0, and 0.1 mM EDTA). The final ssDNA construct was prepared as described⁴⁷. Briefly, a Lumicks BV Inc. commercial optical trap with combined confocal microscopy was used. The system has a microfluidic setup with five experimental channels, each with a height of 100 µm. Re-usable glass slides provided by the company are extensively passivated before the experiment to create a uniform surface. Passivation was performed sequentially with 0.5 mL of each component in separate syringes: 0.1% BSA (Sigma Inc.) flowing at 1 bar for 5 min, then at 0.4 bar pressure for 25 min, followed by a 10-minute rinse with RNase-free water at 1 bar pressure for 5 min, and 0.4 bar pressure for 5 min. This was followed by 0.5% Pluronic F-127, which was flowed at 1 bar pressure for 5 min, then at 0.4 bar pressure for 25 min, followed by 1X Phosphate Saline Buffer (PBS) at 1 bar pressure for 5 min and 0.4 bar pressure for 5 min. Streptavidin-coated polystyrene particle beads of average size 4.8 µm [0.5% w/v] (Spherotech Inc.) were diluted 1:250 in 1X PBS containing 137 mM NaCl, 2.7 mM KCl, 8 mM Na₂HPO₄, and 2 mM KH₂PO₄ (10× PBS was purchased from ThermoFisher Scientific USA Inc.). 100 ng of DNA was made in 0.5 mL of 1× PBS. DNA was captured between two streptavidin beads and mechanically denatured by moving one bead to the overstretched region to create ssDNA. ssDNA was confirmed by fitting the force-distance (FD) curve to the FJC (contour length 48.5 kbp/27.160 µm; persistence length 0.9 nm; stretch modulus 1000 pN) in real-time. DNA was held for 5 s in the fully ssDNA state, then returned to a required tension on the ssDNA position for the fluorescence experiments.

For characterization of Rad51 binding events on the long ssDNA experiments, 100 nM of sfGFP-Rad51 was used for each experiment. The buffer used was 100 mM NaCl, 50 mM HEPES pH 7.5, and 10 mM MgCl₂. The ATP concentration was varied as denoted. The imaging buffer contained 0.8% (w/v) dextrose, 165 U/mL glucose oxidase, 2170 U/mL catalase, and 2–3 mM Trolox to increase the fluorescence lifetime of the fluorophores. Imaging settings were 0.1–0.5 ms exposure time (per pixel), with a constant pixel size of 100 nm. Excitation wavelengths were Green (575–625 nm) and Red (670–730 nm). Laser power was selected to maintain 5 microwatts at the objective. The data was analyzed by custom-written Python scripts. Pylake package of Lumicks was used in python to process the data. Scripts used can be accessed through the following Github link (https://github.com/spangeni/Rad51_eGFP-Paper). Further raw data can be made available upon request.

For characterization of Rad51-binding events on the gapped DNA substrates, data were collected on a different optical trap instrument with the following modifications to the procedure and analysis. Prior to use, the protein channels of the microfluidics chip were passivated according to Lumicks protocol with minor modifications. Briefly, 0.1% BSA resuspended in Rad51 buffer (50 mM Tris-Cl, pH 7.5, 50 mM NaCl, 10 mM MgCl₂ and 0.2 mg/mL BSA) was flowed in for 20 min at 0.35 bar, followed by 0.5% Pluronic F-127 in the same buffer for 20 min at 0.35 bar. Finally, a minimum of 300 µL of Rad51 buffer flowed into equilibrate the microfluidics chip. Biotinylated ds/ss DNA hybrid precursor (17.852 kbp, Cat # 00027Lumicks Inc.) and 4.35 µm Streptavidin-coated polystyrene beads from Lumicks were diluted in 1X running buffer (137 mM NaCl, 2.7 mM KCl, 1.12 mM phosphates, 5 mM sodium azide, and 0.5 mM EDTA, pH 7.4) according to manufacturer’s instructions. ds/ss DNA hybrid precursor was captured between two beads at 0.29 pN/nm trap stiffness, and ss DNA gap was generated by pulling the DNA in a low-salt buffer (0.1× running buffer). Formation of ssDNA gap was verified by comparison to the worm-like chain model in the instrument-associated Bluelake software. In addition, ssDNA gap was confirmed by visualizing DNA using MB543-labeled yeast Rpa protein, which binds to only the ssDNA region with high affinity. Prior to visualization of the interaction of yeast Rad51/Rad52 with Rpa-coated ssDNA, the three proteins were diluted in Rad51 buffer. The ss/dsDNA hybrid was stretched between two beads with a force of 10 pN at a flow pressure of 0.1 bar. The DNA was moved to a microfluidics flow channel containing 0.25 nM Rpa^MB543 in imaging buffer (Rad51 buffer supplemented with 2–3 mM Trolox, 0.165 U/µL glucose oxidase, 2.17 U/ µL catalase and 0.8% (w/v) D-glucose monohydrate). After confirming the coating of the ssDNA gap with fluorescent Rpa, the DNA was then moved to another microfluidics channel containing 50 nM or 100 nM Cy5-labeled Rad51 and 50 nM Rad52 or Rad52^ΔC in imaging buffer with or without 5 mM ATP. Confocal microscopy was carried out using green (561 nm) and red (638 nm) lasers, with pixel size of 100 nm and 2 ms per pixel. Stacks of frames (kymograph) recorded by continuous confocal scanning along the DNA axis over time were visualized using the Lumicks-developed Lakeview Pro software.

Optical-trap data analysis

All kymographs (and image scans, when necessary) were exported as.h5 files and processed using specific Python scripts on a Jupyter notebook. Photobleaching rates were determined by acquiring multiple kymographs at the bottom of the flow-cell and collecting the total photon count until it dropped to at least 50%, before fitting the data points to a single-exponential decay fit model (https://github.com/schaichm/Photostability-calculator.git). Due to the highly unstable photo-physical nature of sfGFP, and the lack of distinct, quantifiable photobleaching steps, the lowest possible intensity of Rad51^GFP was acquired in our optical-trap system by screening for apparent monomer binding events and determining the initial photon count of multiple such traces after background subtraction. The frequency of multiple initial photon counts was subsequently plotted and fitted to a Gaussian model to determine the apparent monomer photon count (Supplementary Fig. 17). Change in photon count upon binding was derived by first converting the.h5 kymograph to a pixel (y axis) and frame index (x axis) image, and then summing the photon count through each frame to derive a total photon count versus frame index plot for our channel(s) of interest. The regions of interest were selected carefully to maximize the signal contributed by the binding event. Any changes in the photon count were fitted using the AutoStepFinder algorithm to objectively derive the average photon count upon binding⁹⁷. An ~10-second window was selected at the start of a new binding event, and the photon counts averaged. Consequently, the average photon count upon binding was divided by the average photon count at baseline (the window before the first binding event in each kymograph) to derive the change in photon count upon binding. The number of Rad51 molecules detected upon binding was derived by dividing the average photon count upon binding by the photon count of an apparent Rad51^sfGFP monomer. The length of each binding event along the y axis (in microns and nucleotides) was derived by applying a Sobel image filter with a directional filter emphasizing horizontal edges to the pixel and frame index image to enhance sudden changes in contrast. For Edge detection on the Kymograph, we used the script ‘Analysis/Edge Detection on a Kymograph’ from https://github.com/lumicks/harbor (2025), developed by Aafke van den Berg. These high-contrast regions were tracked using the KymoTracker function, and then the tracks were overlaid on the original kymograph to verify accurate edge tracking. Positional binding analysis was performed on pixel and frame-index images, where the average photon count was calculated for each pixel through all frames within the window of analysis and was subsequently matched to the distance in microns based on the known pixel size (0.1 μm).

We do note some technical limitations that should be accounted for when interpreting the apparent photon counts for monomeric Rad51^GFP. sfGFP displayed relatively rapid photobleaching and presented as poorly defined steps that could not be distinctly tracked. Ideally, we would have aimed to derive a consistent step-size, or a single-step photobleaching scenario that lets us determine the photon count of a GFP-labeled Rad51 monomer. However, despite optimizing buffer components to enhance photostability and employing a range of laser power and dwell-time parameters, we were unable to establish an ideal scenario. In such a case, we screened for apparent monomer binding events at low Rad51 concentrations, and the initial photon counts of such events were assessed via frequency distribution to determine the most consistent apparent monomer photon count value (Supplementary Fig. 17). The raw measurements, on the other hand, such as the photon count in a window of interest, can be interpreted with confidence. A similar caveat applies to ssDNA-Rad51 filament length analysis in this study. Since the confocal modality in the optical-trap setup is diffraction limited (~200—250 nm), absolute measurements of such measurements (on the gapped DNA used) fall around that limit and thus curtail the extent to which we can interpret the data.

Biolayer interferometry (BLI) analysis of Rad51 ssDNA binding

BLI experiments were performed using an 8-channel Octet R8 instrument (Sartorius) using the Octet BLI Discovery v13.0.1.19 software for data collection and analysis. Measurements were done at 30 °C while shaking at 400 rpm. 30 nM 5′-biotinylated (dT)₆₆ was immobilized to streptavidin (SA)-coated biosensor tips (Sartorius) that were pre-hydrated in the reaction buffer (50 mM Tris-Cl, pH 7.5, 50 mM NaCl, 2.5 mM ATP and 10 mM MgCl₂) for at least 15 min. The baseline signal was measured in the reaction buffer for 120 s before the sensors were moved to the wells containing varying concentrations of Rad51. Rad51 was allowed to bind ssDNA on the tips with an association time of ~450–500 s for unlabeled Rad51 and Rad51^Cy5, and ~1000 s for Rad51^GFP. The association time was optimized such that binding events reached equilibrium. This was followed by dissociation in the respective reaction buffer for ~300 s to eliminate any non-specific binding. All conditions were run in triplicates (n = 3 independent experiments). Appropriate concentrations to reach the steady-state plateau for Rad51-ssDNA binding were pre-determined via a BLI-based titration measurement, and this concentration range was subsequently used for all the experiments. A reference channel with association steps using the reaction buffer was maintained and used to normalize the experimental data. BLI data was analyzed using the Octet Analysis Studio v13.0.1.35 software using reference channel subtraction. The raw reads were processed, normalized to baseline, and response units (RU) were plotted as a function of Rad51 concentration. Given the complexity of Rad51 oligomerization on and off the DNA, we are unable to ascertain (with confidence) quantitative binding parameters from the BLI measurements.

D-Loop assay

Rad51-catalyzed D-loop assays were performed with slight modifications to the previously described methods⁹⁸. A 5′ Cy5-labeled ssDNA oligonucleotide, consisting of 90 nucleotides (10 nM concentration, equivalent to 0.9 µM nucleotides), was incubated with 0.3 µM of wild-type Rad51, Rad51^ybbR, Rad51^GFP, or Rad51^Cy5 in a buffer containing 35 mM Tris-acetate (pH 7.5), 100 mM NaCl, 7 mM magnesium acetate, 2 mM ATP, 1 mM DTT, 0.25 mg/mL BSA, 20 mM phosphocreatine, and 100 µg/mL phosphocreatine kinase. After a 10-minute incubation at 30 °C, 50 nM Rpa was added, and the mixture was incubated for an additional 10 min. Rad54 (120 nM) and supercoiled plasmid dsDNA (10 nM) were added, with samples taken at 0, 5, and 10 min as indicated. Samples were then deproteinized, separated on 1% agarose gels, and documented using a GE Amersham Imager 600. All bands were quantified through densitometry using Image Lab (Bio-Rad).

ATPase assay

The ATPase assay was carried out as previously described using PEI-F cellulose thin layer chromatography to separate ATP-[γ-³²P] and hydrolyzed inorganic phosphate⁹⁰. In 10 μL reactions, 2 μM Rad51 was mixed with 10 μM nt phiX174 virion DNA (NEB, N3023S) in Rad51 buffer supplemented with 1 mM DTT and 200 μM ATP. Reactions were initiated by the addition of 0.1 μCi of ATP-[γ-³²P] and incubated at 30 °C.

Structural models

A model of the Rad51-Rad52 complex was generated by using the respective amino acid sequences as input for the web-service AlphaFold server (https://alphafoldserver.com/) powered by AlphaFold3⁹⁹. Default settings were used to generate predicted structure models for Rad52-Rad51. Molecular graphics and analyses were performed using UCSF ChimeraX¹⁰⁰. The top-ranked predictions are displayed in the figures (Supplementary Fig. 1a, b).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Acknowledgements

This work was supported by grants from the National Institutes of Health, NIGMS R35GM149320 to E.A., R35GM122569 to T.H., and R01GM58015 and R01GM137751 to W.D.H. We also acknowledge instrumentation support from the NIH Office of the Director S10OD030343 to E.A., and S10OD28650 to B.B. Funding for the Montana State Mass Spectrometry Facility was made possible in part by the MJ Murdock Charitable Trust, NIGMS P20GM103474, and the MSU Office of Research and Economic Development. We also acknowledge generous financial support from the Doisy Research Fund of the Edward A. Doisy Department of Biochemistry and Molecular Biology at Saint Louis University School of Medicine. We also acknowledge generous technical support for C-Trap data analysis from the team at Lumicks Inc. and Dr. Matthew Schaich (Univ. of Pittsburgh).

Author contributions

J.D., A.M., and S.K.—performed AUC and MP experiments, generated Rad52 constructs, performed experiments, and generated figures. Assisted with manuscript preparation. A.M.—performed MP and optical trap experiments, generated Rad51^GFP and Rad51^Cy5 constructs and proteins, performed experiments, generates figures. Assisted with manuscript preparation. S.P. and T.H.—performed optical trap experiments and analysis of Rad51 interactions on ssDNA. Manuscript edits and revisions. M.G. and A.M.—performed optical trap experiments on the gapped DNA substrate. Manuscript preparation. V.K.—generated fluorescent Rpa proteins for the study. S.G., J.L., and W.D.H.—designed, purified, and characterized Rad55-Rad57, and performed D-loop and ATPase assays. Manuscript edits and revisions. A.T.—provided information and advice on design of the sfGFP-Rad51 construct. Manuscript revisions. B.B. and M.T.—performed XL-MS experiments and analysis. Manuscript edits and revisions. E.A.—Study design and directed research. Manuscript generation.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The crosslinking mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD065573. All remaining data supporting the findings of this study including large kymograph metadata are available from the corresponding author upon request. Raw data for the XL-MS are included in the Source data file.  are provided with this paper.

Code availability

The Python scripts used to process kymographs, and calculate photon counts, filament length, and measure positional binding have been deposited at GitHub https://github.com/AyushMistry25¹⁰¹.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-025-61811-0.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Negrini, S; Gorgoulis, VG; Halazonetis, TD. Genomic instability-an evolving hallmark of cancer. Nat. Rev. Mol. Cell Biol.; 2010; 11, pp. 220-228.1:CAS:528:DC%2BC3cXitlaisb4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20177397]

2. Pfeiffer, P; Goedecke, W; Obe, G. Mechanisms of DNA double-strand break repair and their potential to induce chromosomal aberrations. Mutagenesis; 2000; 15, pp. 289-302.1:CAS:528:DC%2BD3cXlslyit74%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10887207]

3. Wright, WD; Shah, SS; Heyer, WD. Homologous recombination and the repair of DNA double-strand breaks. J. Biol. Chem.; 2018; 293, pp. 10524-10535.1:CAS:528:DC%2BC1cXht1yns7zF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29599286][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6036207]

4. Heyer, WD; Ehmsen, KT; Liu, J. Regulation of homologous recombination in eukaryotes. Annu. Rev. Genet.; 2010; 44, pp. 113-139.1:CAS:528:DC%2BC3cXhs1SmurnM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20690856][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4114321]

5. San Filippo, J; Sung, P; Klein, H. Mechanism of eukaryotic homologous recombination. Annu. Rev. Biochem.; 2008; 77, pp. 229-257.1:CAS:528:DC%2BD1cXos1ektbY%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18275380]

6. Wold, MS. Replication protein A: a heterotrimeric, single-stranded DNA-binding protein required for eukaryotic DNA metabolism. Annu. Rev. Biochem.; 1997; 66, pp. 61-92.1:CAS:528:DyaK2sXktFOjur4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9242902]

7. Marechal, A; Zou, L. RPA-coated single-stranded DNA as a platform for post-translational modifications in the DNA damage response. Cell Res.; 2015; 25, pp. 9-23.1:CAS:528:DC%2BC2MXitVCnug%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25403473]

8. Antony, E et al. Srs2 disassembles Rad51 filaments by a protein-protein interaction triggering ATP turnover and dissociation of Rad51 from DNA. Mol. Cell; 2009; 35, pp. 105-115.1:CAS:528:DC%2BD1MXps1ejsLc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19595720][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2711036]

9. Caldwell, CC; Spies, M. Dynamic elements of replication protein A at the crossroads of DNA replication, recombination, and repair. Crit. Rev. Biochem. Mol. Biol.; 2020; 55, pp. 482-507.1:CAS:528:DC%2BB3cXhslWgtr%2FF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32856505][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7821911]

10. Mortensen, UH; Lisby, M; Rothstein, R. Rad52. Curr. Biol.; 2009; 19, pp. R676-677.1:CAS:528:DC%2BD1MXhtVensL7L [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19706272]

11. Bai, Y; Davis, AP; Symington, LS. A novel allele of RAD52 that causes severe DNA repair and recombination deficiencies only in the absence of RAD51 or RAD59. Genetics; 1999; 153, pp. 1117-1130.1:CAS:528:DyaK1MXnslSjs7Y%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10545446][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1460819]

12. Shi, I et al. Role of the Rad52 amino-terminal DNA binding activity in DNA strand capture in homologous recombination. J. Biol. Chem.; 2009; 284, pp. 33275-33284.1:CAS:528:DC%2BD1MXhsVCjtb7N [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19812039][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2785170]

13. Bonilla, B; Hengel, SR; Grundy, MK; Bernstein, KA. RAD51 gene family structure and function. Annu. Rev. Genet.; 2020; 54, pp. 25-46.1:CAS:528:DC%2BB3cXhtlyitLvE [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32663049][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7703940]

14. Roy, U. & Greene, E. C. The role of the Rad55-Rad57 complex in DNA repair. Genes12, 1390 (2021).

15. Roy, U; Kwon, Y; Sung, P; Greene, EC. Single-molecule studies of yeast Rad51 paralogs. Methods Enzymol.; 2021; 661, pp. 343-362.1:CAS:528:DC%2BB3MXis1yltLnM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34776219]

16. Roy, U et al. The Rad51 paralog complex Rad55-Rad57 acts as a molecular chaperone during homologous recombination. Mol. Cell; 2021; 81, pp. 1043-1057 e1048.1:CAS:528:DC%2BB3MXhtF2lsLc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33421364][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8262405]

17. Liu, J et al. Rad51 paralogues Rad55-Rad57 balance the antirecombinase Srs2 in Rad51 filament formation. Nature; 2011; 479, pp. 245-248.2011Natur.479.245L1:CAS:528:DC%2BC3MXhtlGksL7K [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/22020281][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3213327]

18. Veaute, X et al. The Srs2 helicase prevents recombination by disrupting Rad51 nucleoprotein filaments. Nature; 2003; 423, pp. 309-312.2003Natur.423.309V1:CAS:528:DC%2BD3sXjs1yntbc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12748645]

19. Krejci, L et al. DNA helicase Srs2 disrupts the Rad51 presynaptic filament. Nature; 2003; 423, pp. 305-309.2003Natur.423.305K1:CAS:528:DC%2BD3sXjs1yntbk%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12748644]

20. Marini, V; Krejci, L. Srs2: the “Odd-Job Man” in DNA repair. DNA Repair.; 2010; 9, pp. 268-275.1:CAS:528:DC%2BC3cXitlSktbc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20096651][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2845805]

21. Seong, C; Colavito, S; Kwon, Y; Sung, P; Krejci, L. Regulation of Rad51 recombinase presynaptic filament assembly via interactions with the Rad52 mediator and the Srs2 anti-recombinase. J. Biol. Chem.; 2009; 284, pp. 24363-24371.1:CAS:528:DC%2BD1MXhtVGgsr3M [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19605344][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2782029]

22. Qiu, Y et al. Srs2 prevents Rad51 filament formation by repetitive motion on DNA. Nat. Commun.; 2013; 4, 2013NatCo..4.2281Q [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23939144]2281.

23. Dhingra, N. et al. The Srs2 helicase dampens DNA damage checkpoint by recycling RPA from chromatin. Proc. Natl. Acad. Sci. USA118, e2020185118 (2021).

24. Deveryshetty, J et al. Yeast Rad52 is a homodecamer and possesses BRCA2-like bipartite Rad51 binding modes. Nat. Commun.; 2023; 14, 2023NatCo.14.6215D1:CAS:528:DC%2BB3sXitVyns7%2FM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37798272][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10556141]6215.

25. Krejci, L; Damborsky, J; Thomsen, B; Duno, M; Bendixen, C. Molecular dissection of interactions between Rad51 and members of the recombination-repair group. Mol. Cell Biol.; 2001; 21, pp. 966-976.1:CAS:528:DC%2BD3MXmtVOrug%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11154282][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC86686]

26. Krejci, L et al. Interaction with Rad51 is indispensable for recombination mediator function of Rad52. J. Biol. Chem.; 2002; 277, pp. 40132-40141.1:CAS:528:DC%2BD38XnvVeht7c%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12171935]

27. Kong, M; Greene, EC. Mechanistic insights from single-molecule studies of repair of double strand breaks. Front Cell Dev. Biol.; 2021; 9, [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34869333][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8636147]745311.

28. Kaniecki, K; De Tullio, L; Greene, EC. A change of view: homologous recombination at single-molecule resolution. Nat. Rev. Genet.; 2018; 19, pp. 191-207.1:CAS:528:DC%2BC2sXhvFGmtbnK [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29225334]

29. Gibb, B et al. Protein dynamics during presynaptic-complex assembly on individual single-stranded DNA molecules. Nat. Struct. Mol. Biol.; 2014; 21, pp. 893-900.1:CAS:528:DC%2BC2cXhsFKms7jP [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25195049][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4190069]

30. Greene, EC; Wind, S; Fazio, T; Gorman, J; Visnapuu, ML. DNA curtains for high-throughput single-molecule optical imaging. Methods Enzymol.; 2010; 472, pp. 293-315.1:CAS:528:DC%2BC3cXhtV2rt7vE [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20580969][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7233429]

31. Deng, SK; Gibb, B; de Almeida, MJ; Greene, EC; Symington, LS. RPA antagonizes microhomology-mediated repair of DNA double-strand breaks. Nat. Struct. Mol. Biol.; 2014; 21, pp. 405-412.1:CAS:528:DC%2BC2cXjvVyqs7s%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24608368][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3980576]

32. Ma, CJ; Gibb, B; Kwon, Y; Sung, P; Greene, EC. Protein dynamics of human RPA and RAD51 on ssDNA during assembly and disassembly of the RAD51 filament. Nucleic Acids Res.; 2017; 45, pp. 749-761.1:CAS:528:DC%2BC1cXivFajtbc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27903895]

33. Pokhrel, N et al. Monitoring Replication Protein A (RPA) dynamics in homologous recombination through site-specific incorporation of non-canonical amino acids. Nucleic Acids Res.; 2017; 45, pp. 9413-9426.1:CAS:528:DC%2BC1cXmtV2gtLs%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28934470][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5766198]

34. Masui, R; Mikawa, T; Kato, R; Kuramitsu, S. Characterization of the oligomeric states of RecA protein: monomeric RecA protein can form a nucleoprotein filament. Biochemistry; 1998; 37, pp. 14788-14797.1:CAS:528:DyaK1cXmt1aiur0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9778353]

35. Forget, AL et al. RecA dimers serve as a functional unit for assembly of active nucleoprotein filaments. Biochemistry; 2006; 45, pp. 13537-13542.1:CAS:528:DC%2BD28XhtV2gtLrO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17087507]

36. Gasior, SL et al. Assembly of RecA-like recombinases: distinct roles for mediator proteins in mitosis and meiosis. Proc. Natl. Acad. Sci. USA; 2001; 98, pp. 8411-8418.2001PNAS..98.8411G1:CAS:528:DC%2BD3MXls1Wis7s%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11459983][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC37451]

37. Liu, J., Gore, S. K. & Heyer, W. D. Local structural dynamics of Rad51 protomers revealed by cryo-electron microscopy of Rad51-ssDNA filaments. Nucleic Acids Res.53, https://doi.org/10.1093/nar/gkaf052 (2025)

38. Ito, K et al. Real-time tracking reveals catalytic roles for the two DNA binding sites of Rad51. Nat. Commun.; 2020; 11, 2020NatCo.11.2950I1:CAS:528:DC%2BB3cXhtFKjurzL [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32528002][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7289862]2950.

39. Conway, AB et al. Crystal structure of a Rad51 filament. Nat. Struct. Mol. Biol.; 2004; 11, pp. 791-796.1:CAS:528:DC%2BD2cXmtVamtrw%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15235592]

40. Merkley, ED et al. Distance restraints from crosslinking mass spectrometry: mining a molecular dynamics simulation database to evaluate lysine-lysine distances. Protein Sci.; 2014; 23, pp. 747-759.1:CAS:528:DC%2BC2cXot1ynsLc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24639379][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4093951]

41. Pellegrini, L et al. Insights into DNA recombination from the structure of a RAD51-BRCA2 complex. Nature; 2002; 420, pp. 287-293.2002Natur.420.287P1:CAS:528:DC%2BD38XovVektLY%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12442171]

42. Kagawa, W et al. Functional analyses of the C-terminal half of the Saccharomyces cerevisiae Rad52 protein. Nucleic Acids Res.; 2014; 42, pp. 941-951.1:CAS:528:DC%2BC2cXhtlShsr8%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24163251]

43. Ma, E. et al. Rad52-Rad51 association is essential to protect Rad51 filaments against Srs2, but facultative for filament formation. Elife7, e32744 (2018)

44. Ma, E. et al. A large C-terminal Rad52 segment acts as a chaperone to Form and Stabilize Rad51 Filaments. Nat. Commun.16, 5589 (2025).

45. Liu, S et al. In vivo tracking of functionally tagged Rad51 unveils a robust strategy of homology search. Nat. Struct. Mol. Biol.; 2023; 30, pp. 1582-1591.1:CAS:528:DC%2BB3sXhslektrvO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37605042]

46. Comstock, MJ; Ha, T; Chemla, YR. Ultrahigh-resolution optical trap with single-fluorophore sensitivity. Nat. Methods; 2011; 8, pp. 335-340.1:CAS:528:DC%2BC3MXit1aitrk%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21336286][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3732480]

47. Pangeni, S et al. Rapid long-distance migration of RPA on single stranded DNA occurs through intersegmental transfer utilizing multivalent interactions. J. Mol. Biol.; 2024; 436, 1:CAS:528:DC%2BB2cXkt1Wlsr4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38360091][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10949852]168491.

48. Kuppa, S et al. Rtt105 regulates RPA function by configurationally stapling the flexible domains. Nat. Commun.; 2022; 13, 2022NatCo.13.5152K1:CAS:528:DC%2BB38Xitlahtb3L [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36056028][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9440123]5152.

49. Candelli, A et al. A toolbox for generating single-stranded DNA in optical tweezers experiments. Biopolymers; 2013; 99, pp. 611-620.1:CAS:528:DC%2BC3sXpvVykurw%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23444293]

50. Sung, P; Robberson, DL. DNA strand exchange mediated by a RAD51-ssDNA nucleoprotein filament with polarity opposite to that of RecA. Cell; 1995; 82, pp. 453-461.1:CAS:528:DyaK2MXnsFSgtr0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/7634335]

51. Ogawa, T; Yu, X; Shinohara, A; Egelman, EH. Similarity of the yeast RAD51 filament to the bacterial RecA filament. Science; 1993; 259, pp. 1896-1899.1993Sci..259.1896O1:CAS:528:DyaK3sXitVyjs7s%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/8456314]

52. Sung, P. Catalysis of ATP-dependent homologous DNA pairing and strand exchange by yeast RAD51 protein. Science; 1994; 265, pp. 1241-1243.1994Sci..265.1241S1:CAS:528:DyaK2cXmt1emsr8%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/8066464]

53. Yin, J; Lin, AJ; Golan, DE; Walsh, CT. Site-specific protein labeling by Sfp phosphopantetheinyl transferase. Nat. Protoc.; 2006; 1, pp. 280-285.1:CAS:528:DC%2BD28XhtFOitbvF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/17406245]

54. Pokhrel, N et al. Dynamics and selective remodeling of the DNA-binding domains of RPA. Nat. Struct. Mol. Biol.; 2019; 26, pp. 129-136.1:CAS:528:DC%2BC1MXnt1Gjtrg%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30723327][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6368398]

55. Kuppa, S; Pokhrel, N; Corless, E; Origanti, S; Antony, E. Generation of fluorescent versions of saccharomyces cerevisiae RPA to study the conformational dynamics of its ssDNA-binding domains. Methods Mol. Biol.; 2021; 2281, pp. 151-168.1:CAS:528:DC%2BB3MXisFals7zN [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33847957][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8326916]

56. Rein, HL; Bernstein, KA; Baldock, RA. RAD51 paralog function in replicative DNA damage and tolerance. Curr. Opin. Genet. Dev.; 2021; 71, pp. 86-91.1:CAS:528:DC%2BB3MXhsFCgtLrL [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34311385][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8671181]

57. Godin, S et al. The Shu complex interacts with Rad51 through the Rad51 paralogues Rad55-Rad57 to mediate error-free recombination. Nucleic Acids Res.; 2013; 41, pp. 4525-4534.1:CAS:528:DC%2BC3sXmt1OgtLc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23460207][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3632125]

58. Hays, SL; Firmenich, AA; Berg, P. Complex formation in yeast double-strand break repair: participation of Rad51, Rad52, Rad55, and Rad57 proteins. Proc. Natl. Acad. Sci. USA; 1995; 92, pp. 6925-6929.1995PNAS..92.6925H1:CAS:528:DyaK2MXntVSktrg%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/7624345][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC41443]

59. Lovett, ST; Mortimer, RK. Characterization of null mutants of the RAD55 gene of Saccharomyces cerevisiae: effects of temperature, osmotic strength and mating type. Genetics; 1987; 116, pp. 547-553.1:CAS:528:DyaL2sXlsFWjsbw%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/3305159][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1203167]

60. Sugawara, N; Wang, X; Haber, JE. In vivo roles of Rad52, Rad54, and Rad55 proteins in Rad51-mediated recombination. Mol. Cell; 2003; 12, pp. 209-219.1:CAS:528:DC%2BD3sXmt1Cqu70%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12887906]

61. Jasin, M; Rothstein, R. Repair of strand breaks by homologous recombination. Cold Spring Harb. Perspect. Biol.; 2013; 5, a012740. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24097900][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3809576]

62. Lisby, M; Rothstein, R. DNA repair: keeping it together. Curr. Biol.; 2004; 14, pp. R994-996.1:CAS:528:DC%2BD2cXhtVKrtb3K [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15589147]

63. Fortin, GS; Symington, LS. Mutations in yeast Rad51 that partially bypass the requirement for Rad55 and Rad57 in DNA repair by increasing the stability of Rad51-DNA complexes. EMBO J.; 2002; 21, pp. 3160-3170.1:CAS:528:DC%2BD38XlsVSlurs%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12065428][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC126052]

64. Carreira, A; Kowalczykowski, SC. Two classes of BRC repeats in BRCA2 promote RAD51 nucleoprotein filament function by distinct mechanisms. Proc. Natl. Acad. Sci. USA; 2011; 108, pp. 10448-10453.2011PNAS.10810448C [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21670257][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3127913]

65. Appleby, R; Joudeh, L; Cobbett, K; Pellegrini, L. Structural basis for stabilisation of the RAD51 nucleoprotein filament by BRCA2. Nat. Commun.; 2023; 14, 2023NatCo.14.7003A1:CAS:528:DC%2BB3sXit1yrtr3P [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37919288][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10622577]7003.

66. Tarsounas, M; Davies, AA; West, SC. RAD51 localization and activation following DNA damage. Philos. Trans. R. Soc. Lond. B Biol. Sci.; 2004; 359, pp. 87-93.1:CAS:528:DC%2BD2cXitVCrtL4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15065660][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1693300]

67. Logan, KM; Skiba, MC; Eldin, S; Knight, KL. Mutant RecA proteins which form hexamer-sized oligomers. J. Mol. Biol.; 1997; 266, pp. 306-316.1:CAS:528:DyaK2sXhs1Gmtrc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9047365]

68. Asleson, EN; Okagaki, RJ; Livingston, DM. A core activity associated with the N terminus of the yeast RAD52 protein is revealed by RAD51 overexpression suppression of C-terminal rad52 truncation alleles. Genetics; 1999; 153, pp. 681-692.1:CAS:528:DyaK1MXmvFGmsL8%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10511548][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1460769]

69. Cherry, JM et al. SGD: saccharomyces genome database. Nucleic Acids Res.; 1998; 26, pp. 73-79.1:CAS:528:DyaK1cXovVWqsw%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9399804][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC147204]

70. Ghaemmaghami, S et al. Global analysis of protein expression in yeast. Nature; 2003; 425, pp. 737-741.2003Natur.425.737G1:CAS:528:DC%2BD3sXotV2iu7c%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/14562106]

71. Kharlamova, MA et al. Monomers and short oligomers of human RAD52 promote single-strand annealing. Proc. Natl. Acad. Sci. USA; 2025; 122, 1:CAS:528:DC%2BB2MXhtVWlt7jM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/40184180][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12002259]e2420771122.

72. Chan, YH; Marshall, WF. Organelle size scaling of the budding yeast vacuole is tuned by membrane trafficking rates. Biophys. J.; 2014; 106, pp. 1986-1996.2014BpJ..106.1986C1:CAS:528:DC%2BC2cXotVaqur0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24806931][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4017318]

73. Cho, NH et al. OpenCell: endogenous tagging for the cartography of human cellular organization. Science; 2022; 375, 1:CAS:528:DC%2BB38XntlChtrs%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35271311][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9119736]eabi6983.

74. Corda, Y., Maestroni, L., Luciano, P., Najem, M. Y. & Geli, V. Genome stability is guarded by yeast Rtt105 through multiple mechanisms. Genetics217, iyaa035 (2021).

75. Li, S. et al. Rtt105 functions as a chaperone for replication protein A to preserve genome stability. EMBO J.37, e99154 (2018).

76. Wang, X. et al. Rtt105 promotes high-fidelity DNA replication and repair by regulating the single-stranded DNA-binding factor RPA. Proc. Natl. Acad. Sci. USA118, e2106393118 (2021)

77. Wang, X et al. Rtt105 stimulates Rad51-ssDNA assembly and orchestrates Rad51 and RPA actions to promote homologous recombination repair. Proc. Natl. Acad. Sci. USA; 2024; 121, 1:CAS:528:DC%2BB2cXitVCmtLfK [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39145931][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11348298]e2402262121.

78. Kolpashchikov, DM et al. Polarity of human replication protein A binding to DNA. Nucleic Acids Res.; 2001; 29, pp. 373-379.1:CAS:528:DC%2BD3MXnt1OmtQ%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11139606][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC29666]

79. Jensen, RB; Carreira, A; Kowalczykowski, SC. Purified human BRCA2 stimulates RAD51-mediated recombination. Nature; 2010; 467, pp. 678-683.2010Natur.467.678J1:CAS:528:DC%2BC3cXht1GqurbL [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20729832][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2952063]

80. Bell, JC; Dombrowski, CC; Plank, JL; Jensen, RB; Kowalczykowski, SC. BRCA2 chaperones RAD51 to single molecules of RPA-coated ssDNA. Proc. Natl. Acad. Sci. USA; 2023; 120, 1:CAS:528:DC%2BB3sXot1Kgur8%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36976771][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10083600]e2221971120.

81. Bignell, G; Micklem, G; Stratton, MR; Ashworth, A; Wooster, R. The BRC repeats are conserved in mammalian BRCA2 proteins. Hum. Mol. Genet.; 1997; 6, pp. 53-58.1:CAS:528:DyaK2sXlsleqsw%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9002670]

82. Davies, AA et al. Role of BRCA2 in control of the RAD51 recombination and DNA repair protein. Mol. Cell; 2001; 7, pp. 273-282.1:CAS:528:DC%2BD3MXis1Kitrc%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11239456]

83. Miron, S et al. DMC1 and RAD51 bind FxxA and FxPP motifs of BRCA2 via two separate interfaces. Nucleic Acids Res.; 2024; 52, pp. 7337-7353.1:CAS:528:DC%2BB2cXislWjt7jM [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38828772][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11229353]

84. Greenhough, LA et al. Structure and function of the RAD51B-RAD51C-RAD51D-XRCC2 tumour suppressor. Nature; 2023; 619, pp. 650-657.2023Natur.619.650G1:CAS:528:DC%2BB3sXht1OntLbJ [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37344587][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7614784]

85. Rawal, Y et al. Structural insights into BCDX2 complex function in homologous recombination. Nature; 2023; 619, pp. 640-649.2023Natur.619.640R1:CAS:528:DC%2BB3sXht1Ontb3I [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37344589][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10712684]

86. Thrasher, J. G., Fagunloye, A. A. G., Justiniano, F. S., Bernstein, K. A. & Jensen, R. B. RAD51 Paralogs and RAD51 Paralog Complexes BCDX2 and CX3 Interact with BRCA2. bioRxiv. https://doi.org/10.1101/2024.10.10.617680 (2024).

87. Belan, O et al. Visualization of direct and diffusion-assisted RAD51 nucleation by full-length human BRCA2 protein. Mol. Cell; 2023; 83, pp. 2925-2940.e2928.1:CAS:528:DC%2BB3sXhsFCrt7%2FJ [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37499663][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7615647]

88. Bashkirov, VI; Herzberg, K; Haghnazari, E; Vlasenko, AS; Heyer, WD. DNA damage-induced phosphorylation of Rad55 protein as a sentinel for DNA damage checkpoint activation in S. cerevisiae. Methods Enzymol.; 2006; 409, pp. 166-182.1:CAS:528:DC%2BD2sXmt1Giu78%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16793401]

89. Kaushik, V et al. Fluorescent human RPA to track assembly dynamics on DNA. Methods; 2024; 223, pp. 95-105.1:CAS:528:DC%2BB2cXis1Wgs7c%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38301751][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10923064]

90. Zhang, XP; Heyer, WD. Quality control of purified proteins involved in homologous recombination. Methods Mol. Biol.; 2011; 745, pp. 329-343.1:CAS:528:DC%2BC38XktV2ksLg%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21660703][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4064682]

91. Boyer, NR et al. Structural insights into redox signal transduction mechanisms in the control of nitrogen fixation by the NifLA system. Proc. Natl. Acad. Sci. USA; 2023; 120, 1:CAS:528:DC%2BB3sXhs1ylt7zO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37459513][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10372690]e2302732120.

92. Chambers, MC et al. A cross-platform toolkit for mass spectrometry and proteomics. Nat. Biotechnol.; 2012; 30, pp. 918-920.1:CAS:528:DC%2BC38XhsVyjs7fO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23051804][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3471674]

93. Lima, DB et al. SIM-XL: A powerful and user-friendly tool for peptide cross-linking analysis. J. Proteom.; 2015; 129, pp. 51-55.1:CAS:528:DC%2BC2MXitVykur4%3D

94. Philo, JS. SEDNTERP: a calculation and database utility to aid interpretation of analytical ultracentrifugation and light scattering data. Eur. Biophys. J.; 2023; 52, pp. 233-266.1:CAS:528:DC%2BB3sXjtlCrtrk%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36792822]

95. Schuck, P; To, SC; Zhao, H. An automated interface for sedimentation velocity analysis in SEDFIT. PLoS Comput. Biol.; 2023; 19, e1011454.2023PLSCB.19E1454S1:CAS:528:DC%2BB3sXhvFKgtbrN [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37669309][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10503714]

96. Schuck, P. Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and lamm equation modeling. Biophys. J.; 2000; 78, pp. 1606-1619.2000BpJ..78.1606S1:CAS:528:DC%2BD3cXhsF2jtbg%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10692345][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1300758]

97. Loeff, L; Kerssemakers, JWJ; Joo, C; Dekker, C. AutoStepfinder: a fast and automated step detection method for single-molecule analysis. Patterns; 2021; 2, 1:CAS:528:DC%2BB38XhvF2nurrO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34036291][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8134948]100256.

98. Liu, J. et al. Srs2 promotes synthesis-dependent strand annealing by disrupting DNA polymerase delta-extending D-loops. Elife6, e22195 (2017)

99. Abramson, J et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature; 2024; 630, pp. 493-500.2024Natur.630.493A1:CAS:528:DC%2BB2cXhtlSntbjJ [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38718835][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11168924]

100. Meng, EC et al. UCSF ChimeraX: Tools for structure building and analysis. Protein Sci.; 2023; 32, 1:CAS:528:DC%2BB3sXit12ntLzK [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37774136][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10588335]e4792.

101. Ayush Mistry.AyushMistry25/PhotonCountCode_2024: CTFM Script for Deveryshetty et al. Nat Comms 2025 (Version NatComms2025). Zenodo. https://doi.org/10.5281/zenodo.15678263 (2025).