Role of BRCA2 DNA-binding and C-terminal domain on its mobility and conformation in DNA repair

BRCA2 is an essential protein in genome maintenance, homologous recombination and replication fork protection. Its function includes multiple interaction partners and requires timely localization to relevant sites in the nucleus. We investigated the importance of the highly conserved DNA binding domain (DBD) and C-terminal domain (CTD) of BRCA2. We generated BRCA2 variants missing one or both domains in mouse ES cells and defined their contribution in HR function and dynamic localization in the nucleus, by single particle tracking of BRCA2 mobility. Changes in molecular architecture of BRCA2 induced by binding partners of purified BRCA2 was determined by scanning force microscopy. BRCA2 mobility and DNA damage-induced increase in the immobile fraction was largely unaffected by C- terminal deletions. The purified proteins missing CTD and/or DBD were defective in architectural changes correlating with reduced homologous recombination function in cells. These results emphasize BRCA2 activity at sites of damage beyond promoting RAD51 delivery.


25
Breast cancer-associated protein 2 (BRCA2) is a required component in multi-step genome maintenance 26 processes that are coordinated in time and place. BRCA2 knock-out is lethal in mammalian cells and 27 defective BRCA2 causes increased sensitivity to genotoxic agents, defective DNA repair and reduced 28 homologous recombination activity (Prakash et al., 2015;Sharan et al., 1997;Yu et al., 2000). One role 29 of BRCA2 common to DNA break repair, DNA crosslink repair and replication fork protection, is delivery 30 of RAD51 to sites where it is needed (Holloman, 2011;Sharan et al., 1997;Yuan et al., 1999). RAD51 is 31 also an essential protein whose biochemical function is to form filaments on ssDNA capable of 32 performing strand exchange reactions with homologous partners or otherwise protecting the bound 33 DNA (Baumann & West, 1998;Heyer et al., 2010). We consider essential BRCA2 activity to involve at 34 least (1) spatial relocation in the nucleus resulting in accumulation to sites where RAD51 is needed and 35 (2) molecular rearrangement to release or deposit RAD51 on DNA in an active form. 36 37 Accumulation of the required proteins at the sites of DNA damage is typically defined as the appearance 38 of foci, high local concentration of proteins, in immunofluorescence experiments. During homologous 39 recombination the formation of RAD51 foci is considered a critical step and this has recently also 40 introduced in clinical setting as a test for homologous recombination defects in tumors (Naipal et al., 41 2014). The accumulation of RAD51 into foci depends on functional BRCA2 (Yuan et al., 1999). Several 42 studies have addressed the role of different interactors and domains of BRCA2 on foci formation after 43 DNA damage induction (Shahid et al., 2014). The presence of Partner and Localizer of BRCA2 (PALB2) 44 and its interaction with the N-terminus of BRCA2 is essential for the localization of BRCA2 and RAD51 to 45 foci (Oliver et al., 2009;Xia et al., 2007;Xia et al., 2006), whereas loss of the interaction affects 46 homologous recombination and genome stability in general (Hartford et al., 2016). In chicken DT40 cells 47 both N-terminal interaction with PALB2 and C-terminal DBD have a role in focal accumulation of BRCA2, 48 accumulation is fully eliminated if neither domain is present (Al Abo et al., 2014). Additionally, the 49 interaction of the BRCA2 DNA binding domain with the small DSS1 protein is required for proper 50 localization of BRCA2 to the nucleus and for BRCA2 and RAD51 focus formation (Gudmundsdottir et al.,51 2004; Kojic et al., 2005;Li et al., 2006). 52 53 Accumulation of BRCA2 and RAD51 in DNA damage induced foci necessarily requires a change in their 54 diffusive behavior. Single particle tracking in living mouse embryonic stem (mES) cells revealed that 55 BRCA2 diffuses as multimeric complexes bound to all detectable nuclear RAD51 (Reuter et al., 2014). 56 Individual BRCA2 particles diffuse slowly and are transiently immobile, and this immobility increases in 57 response to DNA damage induction (Reuter et al., 2014). Although they diffuse together, BRCA2 and 58 RAD51 are separated at sites where they accumulate, as determined by super-resolution microscopy 59 (Sanchez et  peptide residing in exon 12 with the C-terminal ARM domain of HSF2BP (Ghouil et al., 2020). Of these 72 recombination activity. We discuss the possible importance of BRCA2 for steps in the homologous 96 recombination beyond its identified role in RAD51 delivery. 97

98
To determine the role of DBD and CTD in DNA damage repair, BRCA2 mobility and structural plasticity 99 we created murine cell lines and purified human BRCA2 protein lacking these domains ( Figure 1A). 100 Mouse ES cell (mES) lines producing tagged variants of BRCA2 (full-length; ΔDBD, containing an internal 101 deletion of amino acids 2401-3143; ΔCTD, truncated at 3143; and ΔDBDΔCTD truncated at 2401) were 102 engineered by homozygous modification of the endogenous Brca2 alleles and addition of a HaloTag at 103 the end of the coding sequence ( Figure 1B, Figure 1 -supplement 1). This allowed us to visualize BRCA2 104 in live or fixed cells (Los et al., 2008) and study the effect of deletions under native expression in the 105 absence of wild type BRCA2 (Figure 2A, Figure 3). To analyze the role of DBD and CTD in vitro, we 106 purified variants of human BRCA2 protein containing the same deletions ( Figure 1A, Figure 4 and 5). 107

Loss of BRCA2 DBD and CTD impairs cell survival and gene targeting 108
To investigate whether the loss of the DBD or CTD affected sensitivity of the cells to DNA damage we 109 performed clonogenic survival assays after treatment with different DNA damaging agents: double 110 strand break induction by ionizing radiation (IR), replication disruption by PARP inhibitor Olaparib and 111 DNA crosslink induction by mitomycin C (MMC) and cisplatin. Loss of CTD caused sensitization to IR (at 5 112 Gy up to 5-fold decrease in surviving fraction), comparable to the effect of deleting non-essential 113 auxiliary homologous recombination protein RAD54 (Essers et al., 1997). In contrast, deletion of DBD led 114 to a higher hypersensitization (18-fold decrease in surviving fraction or 3.6-fold more than in ∆CTD), 115 which was not further exacerbated if CTD was also missing ( Figure 1C). The effects of domain deletion 116 on Olaparib, cisplatin and MMC sensitivities were similar, but notably, loss of CTD did not lead to a 117 significant sensitization to cisplatin nor MMC (Figure 1D-F). Together, these results indicate that the 118 BRCA2 DBD is important for efficient homologous recombination-mediated DNA repair while the CTD is 119 less critical for this cellular activity. 120

129
(G) CRISPR/Cas9 based homologous recombination assay to assess the homologous recombination proficiency of the different 130 BRCA2 mutants. mES cells are transfected with a plasmid encoding Cas9 and the specific gRNA and an repair template with the 131 self cleaving peptide P2A and the mCherry sequence in between two homology arms. Upon proper integration of the donor 132 sequence at the ß-Actin locus the cells will express mCherry. 96 hours after transfection cells are sorted and the frequency of 133 mCherry positive cells is measured (Figure 1 -supplement 2). To correct for difference in transfection efficiency a plasmid 134 expressing BFP is co-transfected. The frequency of positive cells in every experimental replicate is normalized against wildype 135 BRCA2-Halo cells.

136
To assay homology search and DNA strand exchange functions of homologous recombination, we 137 performed a FACS-based gene targeting assay (Yao et al., 2017), in which Cas9 is used to induce a double 138 strand break in the β-actin locus that is repaired by a donor plasmid including a mCherry coding 139 sequence ( Figure 1G). The absolute gene targeting frequency of about 5% (Figure 1 -supplement 2 to nuclear sites where it is needed, typically observed as foci in cell imaging. We focused on the 147 response of BRCA2 and RAD51 to IR-induced DNA damage because timing of this response in wild type 148 mES cell lines is well defined, and in contrast to genotoxic chemicals, damage induction is instantaneous 149 and synchronous. We visualized BRCA2 protein with a bright photostable fluorophore via the HaloTag 150 using JF646 HaloTag-ligand (Grimm et al., 2015) combined with RAD51 immunofluorescence (Figure 2A). 151 As the DBD of BRCA2 binds DNA in vitro (Yang et al., 2002), it might contribute to BRCA2 localization 152 and/or retention at the sites of damage. However, we observed formation of both spontaneous and IR-153 induced nuclear BRCA2 and RAD51 foci in all three BRCA2 deletion variants, where BRCA2 and RAD51 154 foci appear to overlap to a large extent (Figure 2A).   Figure 2B). Upon irradiation, the number of BRCA2 foci increased in all deletion variants, and the 172 difference in fold increase compared to cells producing full-length BRCA2 was either small or absent. 173 The effect of DBD and CTD deletion on the intensity of background BRCA2 foci was much more 174 pronounced (1.5 fold reduction) ( Figure 2C). Interestingly, in all cell lines IR-induced increase in the 175 number of foci was accompanied by a decrease in focus intensity, suggesting that BRCA2 re-localises 176 from the background to the IR-induced foci, but this effect was suppressed in the deletion variants (only 177 7% reduction in ∆DBD compared to 30% in cells expressing full-length BRCA2). 178 We further analyzed RAD51 focus formation and resolution over 24 hours after IR treatment ( Figure  179  suppressed to a lesser extent in ∆CTD and the double mutant. The effect of DBD and CTD deletion on 185 RAD51 focus intensity dynamics was also pronounced. In the control cells, changes in RAD51 foci 186 intensity paralleled changes in their number: peaking at 2 hours, decreasing gradually thereafter ( Figure  187 2F). In all three deletion variants, focus intensity increase was absent (ΔDBD); reduced (1.2-fold increase 188 compared to 2-fold in control); or delayed (peak at 8 hours compared to 2 hours in control) ( Figure  189 2F,G). Taken together, these results show that the deletion mutants of BRCA2 do accumulate RAD51 190 proteins to IR-induced lesions, however, less RAD51 accumultes and its turnover is supressed. 191 DBD and CTD are not essential for BRCA2 mobility response to DNA damage 192 Previously we used single particle tracking (SPT) to determine diffusive behavior of BRCA2-GFP and 193 observed a mobile fraction, which diffused slower than expected due to frequent transient interactions, 194 and an immobile fraction, which increased upon induction of DNA damage (Reuter et al., 2014).

225
The BRCA2 deletion variants all had a similar apparent diffusion coefficient, mobile molecules diffuse 226 with a rate like the full-length protein ( Figure 3D, Figure 3 -supplement 1). The increase in immobile 227 tracklets after IR for the variants BRCA2 ΔDBD and ΔCTD was similar to full-length, indicating that these 228 domains separately are not essential for this change in mobility ( Figure 3E). However, the immobile 229 fraction for BRCA2 ΔDBDΔCTD, missing both regions, did not increase after IR as much as the others (3% 230 increase compared to 7-10%, Figure 3D). Thus, either the DBD or CTD is sufficient for BRCA2 mobility 231 changes in response to IR but protein missing both of these domains reduces this response. As deletion 232 of single domains, ΔDBD or ΔCTD, did cause increased sensitivity to DNA damaging agents (Figure 1) we 233 can conclude that diffusion changes in response to DNA damage were not sufficient to assure cell 234 survival or proper homologous recombination activity. 235

Architectural rearrangement of BRCA2 variants 236
Our observations so far indicated that BRCA2 function needed for DNA damage survival and 237 homologous recombination includes activities beyond immobility. We considered that homologous 238 recombination DNA damage response requires dynamic interaction between BRCA2 and RAD51 at a 239 scale not evident in our (live) cell imaging. Although BRCA2 and RAD51 diffuse together in the nucleus,  (Figure 5A and B). Incubation with 277 ssDNA shifts the solidity distribution to a single peak at 0.7 (Figure 5: FL BRCA2 + ssDNA). The BRCA2 C-278 terminal deletion variants had fewer extended molecules to begin with, as solidity shows a peak 279 distribution around 0.9 for all the 3 variants; ΔDBD, ΔCTD and ΔDBDΔCTD (Figure 5 and Table S3). In 280 striking contrast to the full-length BRCA2, interaction with ssDNA did not change the distribution of 281 oligomers and shape (solidity) for any of the deletion variants (Figure 5, Figure 5 -supplement 1). Both 282 the DBD and CTD have to be present for BRCA2 to undergo the conformational change associated with 283 ssDNA interactions. 284 Upon incubation with RAD51, full-length BRCA2 assemblies become largely monomeric (74%) and adopt 295 a more regular compact conformation, with 33% having a rod-like shape (major to minor axis ratio >1.5) 296 ( Figures 4A, B, Figure 4 -Supplement 1,2 and Table S4 All deletion variants also become largely monomeric upon interaction with RAD51, but to a lesser extent 301 than the full-length BRCA2 (40-55% for variants vs 74% for full-length). However, all the variants 302 included about one-third of the complexes as dimers: BRCA2 ΔDBD (28%), BRCA2 ΔCTD (32%) and 303 BRCA2 ΔDBDΔCTD (33%), which was more than the full-length BRCA2 (18%) (Figure 4B, C: right panel). 304 Only BRCA2 ΔCTD-RAD51 formed rod-shaped assemblies similar to full-length BRCA2 (Figures 4B and  305 Table S4). Removing either the DBD, CTD or both, reduced BRCA2 oligomerization and to a lesser extent 306 reduced RAD51 induced changes in oligomerization and architecture (Figure 4 supplement 1 and 2). The 307 effect of the BRCA2 DBD and CTD domain deletions on cellular response to DNA damage (Figure 1) and 308 their effect on the architecture of BRCA2 and its complexes with RAD51 and ssDNA did correlate. The 309 architectural changes we defined here may report on important BRCA2 cellular functions. 310

311
Here we investigated the role of DBD and CTD on the diffusive behavior and ligand-induced structural 312 plasticity of BRCA2. We correlated these observations with functional consequences of deleting these 313 domains, individually and in combination, in living cells. Our panel of isogenic precision-engineered cell 314 lines allowed us to label endogenously expressed BRCA2 directly via the HaloTag. We found that 315 substantial reduction in DNA damage resistance, especially in DBD-deficient cells, was accompanied by 316 only subtle changes in dynamics and localization of BRCA2. In contrast, the ability of purified 317 recombinant BRCA2 to undergo structural rearrangements was strongly affected by DBD or CTD 318

deletions. 319
Despite their adjacent location in the C-terminal part of BRCA2 (and frequent simultaneous loss due to 320 human cancer-predisposing mutations), DBD and CTD are functionally distinct. CTD, although much 321 shorter that the DBD, performs several distinct functions: cell cycle-controlled phosphorylation-322 dependent stabilization of RAD51 filament in vitro; replication fork protection from excessive nucleolytic 323 processing; and nuclear import. In mouse BRCA2 an additional nuclear localization signal is present at 324 the N-terminus, but in the human protein there is no such redundancy, which exaggerates the 325 consequence of even short C-terminal truncations, because these produce human BRCA2 that cannot 326 localize to the nucleus (Sarkisian et al., 2001;Spain et al., 1999). Controlled deletions, including the 327 internal DBD deletion, allowed us to avoid some of the confounding effects complicating previously used 328 mutant or patient cell models. 329 The DBD is the evolutionarily defining part of BRCA2, conserved from fungi to humans, but its function is 330 less defined than that of other, "younger" BRCA2 regions. Information on DBD function focuses on its 331 interaction with an intrinsically disordered acidic protein DSS1. Our findings reinforce the notion that 332 despite its deep phylogenetic roots, the DBD is not what makes BRCA2 essential for general viability of 333 animal cells. Absence of the DBD leads to significant sensitization to DNA interstrand crosslinks, PARP 334 inhibitor and radiation, not further exacerbated by additional CTD deletion (Figure 1). The role of DSS1 335 interaction remains puzzling: On the one hand, it is as conserved the DBD itself, was shown to be 336 required for BRCA2 stability and intracellular localization (Li et al., 2006); mutations in DSS1 binding 337 phenocopy BRCA2 deficiency, as does DSS1 depletion (Zhao et al., 2015). But, on the other hand, in fungi 338 DSS1 is only required for DNA repair when DBD is present (Kojic et al., 2005) (Figure 1E, F) where CTD 357 deletion has no effect. This finding is also suggesting that the described fork protection and RAD51 358 filament stabilization functions of the CTD are not essential for DNA crosslink repair. This is at odds with 359 studies that describe the role of fork protection in crosslink repair, and in particular with previous 360 findings on ES cells with different CTD-disrupting Brca2 alleles (Atanassov et  We observed RAD51 focus formation in some cell lines that were however deficient in homologous 368 recombination at two-ended double-strand breaks. Despite strong sensitization to radio-and 369 chemotherapeutic agents, only careful quantification of the numbers and intensity of RAD51 foci at 370 multiple timepoints after radiation revealed subtle differences in the deletion variants. The reduced 371 intensity of RAD51 foci in cells lacking DBD and CTD indicates that the repair process is delayed or 372 reduced at some point beyond delivery of RAD51 by BRCA2 to the sites of damage. This suggests that 373 RAD51 foci quantification, although useful to identify more homologous recombination deficient 374 samples than BRCA1/2 mutations, could occasionally return false positive results for homologous 375 recombination function at two-ended breaks. In and of itself this observation is not surprising as there 376 will be steps important for homologous recombination function downstream of RAD51 focus formation. 377 However, it does raise a note of caution in the context of employing the RAD51 focus formation assay in 378 pre-clinical and clinical settings with the aim to find BRCAness phenotypes. Additional homologous 379 recombination markers may need to be explored to overcome this limitation. 380 The DBD and CTD of BRCA2 did markedly affect protein architecture and conformational changes in 381 response to binding partners. These domains contribute to oligomerization, when we remove them, in 382 DBD and CTD deletion variants, the BRCA2 population was less oligomeric. This is an agreement with a 383 recent study where interaction of N and C-terminal fragments of BRCA2 is indicated to contribute to 384 oligomerization of BRCA2 (Le et al., 2020). However, in our study the oligomeric forms induced by 385 RAD51 binding remain unchanged, which is likely mediated by interaction with the intact BRC repeats. 386 The characteristic conformational change of irregular compact particles-to-extended architecture of full-387 length BRCA2 in response to ssDNA was severely impaired in all the investigated deletion variants. 388 Together, the inability of the deletion variants to rearrange in vitro in the presence of ssDNA coupled 389 with the impaired homologous recombination in vivo suggests DBD and CTD interactions of BRCA2 are 390 important for optimal BRCA2 activity at the sites of damage. Similar regulatory function is reported for 391 other proteins in that interact with BRCA2 such as DSS1, which also affects the conformation of BRCA2 392 (Le et al., 2020). Suggesting that regulation of RAD51 by BRCA2, is affected by conformational 393 rearrangement of BRCA2 and is mediated at different levels by self-interaction of BRCA2 and its 394 interaction partners. 395 Comparing all the molecular endpoints we analyzed (diffusion, foci, architecture) we conclude that 396 although none correlated perfectly with the functional outcomes (survival and recombination assay), the 397 magnitude of the effect on architectural plasticity was the closest reflection. We are only starting to 398 tease apart the relationship between structural plasticity and cellular function. BRCA2 function may 399 depend not so much on the existence of one structural form or another but on the lifetime of specific 400 conformations affected by its interactors and local chromatin organization, parameters that will need to 401 be quantified. restriction fragment) were separately purified from gel. The resulting restriction fragment was used as 409 template for two PCRs with overhanging primers containing the required gRNA sequence (see Table S1). 410 Using Gibson assembly, the two fragments and the digested vector backbone were assembled and 411 transformed in E. coli (DH5 alpha). The correct integration of the gRNA sequence in the isolated plasmid 412 was validated by Sanger sequencing. For incorporation of two gRNA in a single plasmid, px459 was 413 modified to contain two U6 promotors and gRNA sequences separated by a short spacer. 414 The donor template for C-terminal tagging of BRCA2 with HaloTag was derived from the plasmid that 415 was used to make BRCA2-GFP knock-in cell lines (Reuter et al., 2014). This plasmid contains 3' and 5' 416 homology arms (6.6 and 5.4 kb homology) for integration of the construct at the BRCA2 locus. The GFP 417 sequence was removed by restriction digest and replaced with the HaloTag sequence by Gibson 418 Assembly. The HaloTag sequence was obtained by PCR from pENTR4-HaloTag (gift from Eric Campeau, 419 Addgene #29644). The donor plasmids for the ΔCTD, ΔDBDΔCTD-HaloTag contain a 6kb homology arm 420 upstream of the deletion, while the downstream homology arm was identical to the full-length 421 construct (See Figure 1 -supplement 1). The ΔDBD donor construct was made by introducing the coding 422 sequence from exon 27 of mouse BRCA2 excluding the stop codon in the ΔDBDΔCTD-HaloTag donor 423 construct. 424 The PiggyBac iRFP720-PCNA construct was generated using Gibson assembly, by inserting the iRFP720 425 sequence (Shcherbakova & Verkhusha, 2013)  immunofluorescence was performed as described above and DNA was stained using DAPI. 491

Confocal microscopy 492
Confocal images were acquired at a Zeiss Elyra PS1 system with additional confocal scan unit coupled to 493 an Argon laser for 488nm excitation (Alexa 488) and additional 30mW 405nm (DAPI), 10mW 561nm 494 (CF568, JF549) and 633nm (Alexa 647, JF646) lasers. A 63x (NA 1.4, Plan Apochromat DIC) objective was 495 used for imaging. At least 3 positions per condition were selected based on DAPI signal, subsequently 496 automatic multi position imaging was performed and for every position. Fluorescence based autofocus 497 was used to find the center of the nuclei. A z-stack of 11 slices with 500 nm axial spacing from the center 498 was acquired, while the lateral pixel size was 132*132 nm. 499 Foci quantification 500 BRCA2 and RAD51 foci were automatically quantified using CellProfiler (Carpenter, 2006). The analysis 501 script can be found at https://github.com/maartenpaul/DBD_foci. In short, from maximum projections 502 of the confocal images, nuclei were segmented using a global threshold (minimum cross-entropy) based 503 on the DAPI signal. Subsequently within the masked image based on segmented nuclei RAD51 foci were 504 identified using global threshold (Robust background) method with 2 standard deviations above 505 background. The integrated intensity of EdU signal per nucleus was also measured and used to 506 determine the EdU positive cells. Based on the distribution of the integrated intensity of EdU signal per 507 nucleus a fixed threshold was set at 500 a.u., cells above this threshold were defined EdU positive. Also 508 the integrated intensity per focus for BRCA2 and RAD51 were obtained from CellProfiler. The data was 509 exported as CSV files from CellProfiler. R and Rstudio was used to plot the data (example script can be 510 found at the Github repository mentioned above). 511

Live cell imaging 512
For tracking experiments cells were labelled with 5 nM JF549-HaloTag ligand for 15-30 minutes at 37 o C 513 in mouse ES imaging medium (FluoroBrite DMEM (ThermoFisher, 10 % FCS supplemented with non-514 essential amino acids, 0.1 mM β-mercaptoethanol, pen/strap and 1,000 U/ml leukemia inhibitory 515 factor). Subsequently cells were incubated twice for 15 minutes with fresh imaging medium, while 516 washing the cells once with PBS in between. Microscopy experiments were performed at a Zeiss Elyra PS 517 complemented with a temperature-controlled stage and objective heating (TokaiHit). Samples were 518 kept at 37 o C and 5% CO 2 while imaging. For excitation of JF549 a 100mW 561 nm laser was used. The 519 samples were illuminated with HiLo illumination by using a 100x 1.57NA Korr αPlan Apochromat (Zeiss) 520 TIRF objective. Andor iXon DU897 was used for detection of the fluorescence signal, from the chip a 521 region of 256 by 256 pixels (with an effective pixel size of 100*100 nm) was recorded at 31. linking with a maximum displacement of 1.2 µm and a gap size of maximum 1. Tracks had to be at least 531 5 frames long to be processed further. 532 Subsequently, the track data was imported in R for analysis using a home-build script 533 (https://github.com/maartenpaul/DBD_tracking). Tracks were segmented in tracklets using the ML-MSS 534 software described in (Arts et al., 2019) (https://github.com/ismal/DL-MSS), using a 3-state deep-535 learning prediction model. Apparent diffusion constants for the tracklets were estimated by determining 536 the slope of the MSD(t) curve. From all tracklets that were at least 10 frames in length. 537

SFM sample preparation, imaging and analyses 564
For BRCA2-RAD51 reactions, aliquots of BRCA2 stored at -80 °C were thawed and diluted four-fold in 10 565 mM HEPES pH 8.0 buffer, to subsequently prepare a reaction of 2.5 nM BRCA2 construct in 22 mM 566 HEPES pH 8.2, 112 mM NaCl, 0.125 mM EDTA, 2.5 % glycerol, 0.25 mM DTT. Samples were incubated at 567 37 °C in the absence or presence of 250 nM RAD51 for 30 minutes without shaking. 568 For BRCA2-ssDNA reactions, after dilution as mentioned above, the protein was incubated at 37 °C for 569 30 minutes with linear 90 nt ssDNA oligo (3.4 µM in nt) (5'-570 AF647/AATTCTCATTTTACTTACCGGACGCTATTAGCAGTGGCAGATTGTACTGAGAGTGCACCATATGCGGTGTG 571 AAATACCGCACAGATGCGT-3'). After incubation 50 µM spermidine was added to the sample. 572 Samples for SFM imaging were prepared by depositing 20 µl reaction volume on a freshly cleaved mica 573 (Muscovite mica, V5 quality, EMS) for 2 minutes, followed by a 2 ml wash using 18 MΩ water and drying 574 in filtered (0.22 µm) air. SFM images were obtained with a Nanoscope IV (Bruker), using tapping mode in 575 air with a silicon probe, NHC-W, with tip radius <10 nm and resonance frequency range of 310-372 kHz 576 (Nanosensor, Veeco Instruments, Europe). All images were acquired with a scan size of 2 × 2 µm at 512 × 577 512 pixels per image at 0.5 Hz. Images were processed using Nanoscope analysis (Bruker) for 578 background flattening. Quantitative analysis of the images was performed as described using SFMetric 579 software (Sanchez et al., 2017, Sidhu et al. 2020). In volumetric analyses, a comparison of the oligomeric 580 volume of the different regions with RAD51 (56 nm 3 ) showed that the monomer volume of RAD51 is 581 much lower than the threshold volume and thus free RAD51 is removed from analysis ( Figure S2). 582 The conformation of the molecules was quantified by the parameters of solidity. Solidity measures the 583 irregular shape of the selected molecule by using the ratio of the area of the selected molecule to the 584 area of a convex hull, that completely encloses the molecule. Solidity is presented in a scale of 1 to 0, 585 where a value of ~1 signifies a globular molecule while a value ~0 represents a highly irregular molecular 586 shape. 587

588
We thank the Optical Imaging Centre for use and technical assistance of the optical microscopes; Ihor 589 Smal (Erasmus MC) for assistance in single-molecule tracking analysis; Luke Lavis (HHMI Janelia) for 590 providing HaloTag ligands; Niklas Bachmann for assistance in making the BRCA2-Halo ΔCTD cell line. We 591 thank Joyce Lebbink (Erasmus MC) and Nick van der Zon (Erasmus MC) for critically reading the 592 manuscript. 593