RAD51 regulates eukaryotic chromatin motions in the absence of DNA damage

Abstract

In yeasts and higher eukaryotes, chromatin motions may be tuned to genomic functions, with transcriptional activation and the DNA damage response both leading to profound changes in chromatin dynamics. The RAD51 recombinase is a key mediator of chromatin mobility following DNA damage. As functions of RAD51 beyond DNA repair are being discovered, we asked whether RAD51 modulates chromatin dynamics in the absence of DNA damage and found that inhibition or depletion of RAD51 alters chromatin motions in undamaged cells. Inhibition of RAD51 increased nucleosome clustering. Predictions from polymer models are that chromatin clusters reduce chain mobility and, indeed, we measured reduced motion of individual chromatin loci in cells treated with a RAD51 inhibitor. This effect was conserved in mammalian cells, yeasts, and plant cells. In contrast, RAD51 depletion or inhibition increased global chromatin motions at the microscale. The results uncover a role for RAD51 in regulating local and global chromatin dynamics independently from DNA damage and highlight the importance of considering different physical scales when studying chromatin dynamics.

• RAD51 participates in DNA repair by homologous recombination, thereby influencing chromatin mobility. Noncanonical functions of RAD51 are being discovered, prompting the authors to examine if RAD51 regulates chromatin dynamics in the absence of DNA damage.

• Using super-resolution imaging and multiscale particle tracking, they find that RAD51 inhibition alters nucleosome clustering and chromatin motions independently of DNA damage. The effect varies at different spatial scales and is conserved in different model organisms.

• The findings suggest that RAD51 regulates global mobility of chromatin. Defining the mechanisms regulating chromatin motions is essential to understanding the connection between chromatin mobility and genome functions.

INTRODUCTION

Genomic functions including replication, transcription, and repair have an influence on (and may be regulated by) chromatin motions. DNA damage profoundly affects chromatin dynamics (Seeber et al., 2018; Lawrimore et al., 2020; Mine-Hattab and Chiolo, 2020). In yeast, DNA double-strand breaks (DSBs) induced by genotoxic drugs or restriction enzymes lead to increased motion of the chromatin chain flanking the break, as well as globally on other chromosomes (Mine-Hattab and Rothstein, 2012; Neumann et al., 2012; Dion et al., 2013; Seeber et al., 2013; Herbert et al., 2017; Lawrimore et al., 2017). Increased mobility facilitates homology search and strand invasion during homologous recombination repair (HRR) (Mine-Hattab and Rothstein, 2012; Challa et al., 2021; Garcia Fernandez et al., 2022). The impact of DSBs on chromatin dynamics in higher eukaryotes is more nuanced and less understood. DNA repair foci are more mobile than other chromatin domains (Krawczyk et al., 2012) and dysfunctional telomeres (resembling DSBs) diffuse faster than capped, functional telomeres (Dimitrova et al., 2008). Yet, measurements in undamaged nuclear regions have revealed decreased global chromatin motions in response to damage, with higher mobility retained at or near break sites (Liu et al., 2015; Locatelli et al., 2022).

The RAD51 (RADiation sensitive protein 51) recombinase is a key HRR factor. The protein forms helical nucleoprotein filaments on resected single-stranded DNA (ssDNA) and is involved in homology search and pairing with the homologous sequence. The local increase in yeast chromatin motions following DNA damage depends on the RAD51 homologue (scRad51) (Mine-Hattab and Rothstein, 2012; Smith et al., 2018; Smith et al., 2019; Garcia Fernandez et al., 2022). This effect, proximal to the break sites, may be explained, at least in part, by stiffening and increased persistence length of chromatin after the formation of Rad51 nucleofilaments (Mine-Hattab et al., 2017). scRad51 also plays a context-dependent role in global chromatin mobility after DNA damage (Garcia Fernandez and Fabre, 2022). Specifically, scRad51 is necessary for radiation-induced mobility of distal, undamaged loci in diploid cells (Mine-Hattab and Rothstein, 2012; Smith et al., 2018). Whereas scRad51 is generally excluded from stalled replication forks until their relocation to the “safe haven” of the nuclear periphery (Su et al., 2015), peripheral relocation of replication forks arrested by replication fork barriers occurs after scRad51 loading (Kramarz et al., 2020), implicating the protein in directed chromatin motions in the absence of DNA damage. ScRad51 may also participate in the formation of damage-induced nuclear microtubule filaments which mediate DNA break translocation to the periphery of yeast nuclei (Oshidari et al., 2018). Less is known about the contribution of RAD51 to chromatin dynamics in higher eukaryotes. In mammalian cancer cells relying on the alternative lengthening of telomeres (ALT) pathway to maintain their telomeres, damaged telomeres undergo directed motions which depend on RAD51 for homology-directed synthesis of telomeric DNA (Cho et al., 2014). These findings suggest that RAD51 may be an important chromatin mobility factor across species.

Noncanonical roles of yeast and mammalian RAD51 have been described, some of which are independent of the well-known HRR functions of the protein. RAD51 promotes the formation of R-loops of DNA-RNA hybrids (Wahba et al., 2013; Feretzaki et al., 2020; Girasol et al., 2023). While R-loops can lead to genomic instability, they also play physiological roles in transcription activation and termination (Crossley et al., 2019; Niehrs and Luke, 2020). Importantly, accumulation of scRad51 at R-loops occurs before the formation of DNA breaks at these structures (Wahba et al., 2013). In the context of DNA replication stress, RAD51 nucleofilaments protect stalled replication forks, independently from the recombination function of the protein, by promoting fork reversal (Wassing and Esashi, 2021). Considering these HRR-independent functions of RAD51 and the role of the protein in chromatin mobility following DNA damage, we asked whether the protein regulates chromatin dynamics in undamaged cells. We find that RAD51 depletion or a short pharmacological inhibition of RAD51 alters chromatin dynamics, and that this effect is dependent on the physical scale considered.

RESULTS

RAD51 modulates mammalian chromatin microdomains motions

RAD51 is a key influencer of chromatin mobility in response to DNA damage (Mine-Hattab and Rothstein, 2012; Cho et al., 2014; Smith et al., 2018; Smith et al., 2019; Garcia Fernandez et al., 2022). Because a growing number of functions independent from DNA repair are being discovered for the protein, we set out to determine whether RAD51 affects chromatin dynamics in the absence of DNA damage. We recently developed a method based on structured illumination to map the motions of chromatin microdomains in mammalian cells (Bonin et al., 2018; Locatelli et al., 2022). The method relies on a diffractive optical element producing an array of beamlets that activate photoactivatable GFP fused to histone H2A in the nucleus of cells (PAGFP-H2A; Figure 1A), which results in a 7 × 7 lattice of fluorescent spots. The measurements reflect diffusion (D) of chromatin microdomains.

FIGURE 1:

RAD51 regulates chromatin motions at the microscale. (A) Schematic illustration of the structured illumination system used to track chromatin microdomain motions in cells expressing PAGFP-H2A. KD of RAD51 expression in U2OS cells verified by Western blot (B) and immunostaining (C). Immunoblot signal intensities were quantified by densitometry. Normalized intensity values are shown below the blots (mean ± SD; N = 2). Cells used for immunostaining were treated with MMC to induce DNA damage. The proportion of nuclei that accumulated RAD51 foci is shown in the graph (N = 2). Scale bar, 10 µm. (D) Representative images of the photoactivated chromatin lattices (PAGFP-H2A) and corresponding diffusion (D) values (bubble plots) in control and RAD51 KD cells. Measurements with fixed cells show a modest contribution of noise. Scale bar, 5 µm. D values are quantified in the graph. ***, P = 0.0005 (Mann–Whitney test). N = 50–90 cells from three independent experiments. (E) Chromatin microdomain diffusion in cells treated with RAD51 inhibitors B02 (100 µM) and RI-1 (25 µM). ****, P < 0.0001 (t-test). N = 30–100 cells from two biological replicates.

To determine whether RAD51 modulates mammalian chromatin motions, we generated a U2OS cell line with reduced RAD51 levels by stably expressing Cas9 and a RAD51-specific guide RNA. We achieved a ∼60% reduction in RAD51 expression, but not complete depletion, as determined by Western blot and immunostaining (Figure 1, B and C). This was expected since RAD51 is an essential gene in mammalian cells (Tsuzuki et al., 1996). RAD51 knockdown (KD) cells had increased sensitivity to mitomycin C, a DNA cross-linker inducing DNA breaks predominantly repaired by homologous recombination (Supplemental Figure S1). These results confirmed reduced RAD51 function in the KD cell line.

Chromatin microdomains were significantly more mobile in RAD51 KD cells compared with the parental cell line (Figure 1D). To further assess the influence of RAD51 on chromatin motions, we used the B02 (Huang et al., 2011; Huang et al., 2012) and RI-1 (Budke et al., 2012) inhibitors, which prevent the interaction of the protein with ssDNA and filament formation. As expected, these treatments prevented the formation of RAD51 repair foci in cells with DNA damage (Supplemental Figure S2). B02 and RI-1 treatments significantly increased chromatin microdomain motions, (Figure 1E), suggesting that RAD51 limits chromatin motions at the microscale.

Increased chromatin microdomain mobility upon RAD51 inhibition is independent from DNA damage

RAD51 is a key factor for DNA DSB repair by HRR. Accordingly, prolonged inhibition of RAD51 is expected to increase the DNA damage burden, complicating data interpretation. We and others have indeed shown a complex effect of DNA damage on chromatin dynamics (Seeber et al., 2018; Locatelli et al., 2022). We therefore identified treatment conditions with the B02 RAD51 inhibitor, namely a dose of 10 µM applied for 1 h, which did not lead to reduced cell viability (Figure 2A), nor to the induction of DNA breaks, quantified by neutral comet assays (Figure 2B), by immunostaining for γH2AX (Figure 2C), and visualized with mCherry fused to the c-terminus of 53BP1 (Dimitrova et al., 2008) (Figure 2D). With these conditions (10 µM B02 for 1 h), acceleration of chromatin microdomain motion was conserved (Figure 2E), despite the absence of DNA damage induction. Within a population, U2OS cells have varying levels of endogenous DNA breaks. We therefore also guided chromatin mobility analyses to cells with no or very few DNA damage foci and still observed B02’s effect on chromatin mobility (Figure 2F). Finally, HRR is predominantly active in S- and G2-phases of the cell cycle when sister chromatids are available for recombination repair. In experiments with enriched G1 cell populations, we measured the same B02-induced increased chromatin microdomain motions, suggesting that this effect is not restricted to S/G2-phases (Figure 2G). As expected, the short (1 h) exposure to B02 did not alter cell cycle distribution (Supplemental Figure S3), ruling out potential effects associated with proliferation. We found previously that manipulation of chromatin compaction with histone deacetylase or demethylase inhibitors alters chromatin microdomain motions (Locatelli et al., 2022). However, in our short treatment conditions, no appreciable change in methylated nor acetylated histone H3 (H3K9me3 and H3K9ac) was detected (Supplemental Figure S4). Collectively, the results suggest that inhibition of RAD51 affects chromatin mobility independently from RAD51's function in DNA repair.

FIGURE 2:

Increased chromatin microdomain mobility upon RAD51 inhibition is independent from DNA damage. (A) Cell viability, assessed by MTT assay 24 h after B02 treatment. Values are normalized to untreated cells. *, P < 0.05; ns, not significant (one-sample t test; N = 3 biological replicates). (B) Detection of DNA DSBs with neutral comet assays in cells treated with B02 (10 or 100 µM, 1 h). Bleomycin treatment (Bleo; 20 mU/ml for 1 h) was used as positive control. Averaged values from two biological replicates (>300 cells/condition) are shown in the graph. (C) γH2AX immunostaining of cells treated as in B. Scale bar, 10 µm. γH2AX foci quantification is shown in the graph. **, P < 0.01; ns, not significant (Kruskal–Wallis and Dunn's multiple comparisons test; N > 300 cells from two biological replicates). (D) DNA break foci (labeled with mCherry-53BP1_ct) in cells treated with B02 (10 µM, 1 h). Gamma adjustment (0.5) was performed to visualize both nucleoplasmic signals and DSB foci. Scale bar, 10 µm. DSB foci are quantified in the bar graph (mean ± SEM). ns, not significant; N = 900–1000 cells from three independent experiments. (E) Chromatin diffusion (D) in untreated U2OS cells and in cells treated with B02 (10 µM, 1 h). ****, P < 0.0001; N = 53 (control) and N = 100 (B02) cells from three independent experiments. (F) Chromatin D values in cells treated as in E, but only including cells with ≤ 3 mCh-53BP1_ct DSB foci, to avoid cells with endogenous DNA breaks. *, P = 0.003; N = 12 (control) and N = 16 (B02) cells from three experiments. (G) Effect of B02 on chromatin diffusion in G1-phase. Enrichment of the G1 cell population after synchronization with a double thymidine (TT) block was confirmed by flow cytometry (left). Chromatin D values are shown in the graph. *, P = 0.013; N = 39 (control) and 41 (B02) cells from two experiments.

RAD51 inhibition increases nucleosome clustering

To measure the effect of RAD51 inhibition on the dynamics and nanoscale organization of single nucleosomes, we used a single-molecule imaging approach with cells expressing histone H2B fused to the HaloTag, labeled with a fluorescent HaloTag ligand (Kefer et al., 2021) (Figure 3A). First, we tracked single nucleosomes in time-lapse movies from cells with sparsely labeled H2B. As shown in Figure 3B, there was no significant difference in nucleosome diffusion speed in cells treated with B02 compared with controls. Therefore, our results show different dependencies for chromatin motions at different physical scales.

FIGURE 3:

RAD51 inhibition does not impact single-nucleosome motions but increases nucleosome clustering. (A) Schematic of oblique light sheet imaging of single nucleosomes and micrograph of a U2OS cell expressing histone 2B fused to HaloTag (H2B-HT) labeled with JF646 HaloTag ligand (HTL). (B) Diffusion of single nucleosomes in untreated cells and in cells treated with B02 (10 µM, 1 h); ns, not significant (P = 0.07; N = 3391 [Control] and N = 3661 [B02] nucleosome traces from two independent experiments). (C) Representative dSTORM super-resolution (SR) images of cells treated as in B. Scale bars, 5 µm. (D) Illustration of spot cluster analysis using Besag's L-function. A random particle distribution along the radius of the nucleus (r) yields a flat line whereas a peak in the function indicates clustering. The amplitude and position of the peak describes the degree of clustering at a given distance band. (E) Nucleosome cluster analysis of untreated and B02-treated cells. Means ± SD are shown. N = 12 (control) and N = 18 (B02) cells.

Next, we analyzed nanoscale chromatin organization with super-resolution imaging, using direct STochastic Optical Reconstruction Microscopy (dSTORM) (Van de Linde et al., 2011). We used the same H2B-HT cell system, but with dense labeling of nucleosomes with the HaloTag ligand (Figure 3C; Supplemental Figure S5A). Qualitatively, nucleosomes were more dispersed in the control than in cells treated with B02, although the number of nucleosomes detected in the reconstituted dSTORM images were similar in both conditions. Quantification of spatial dependencies using Besag's L-function (Figure 3D; Supplemental Figure S5, A and B) confirmed a clustered distribution of nucleosomes at the 100 nm distance band (Bancaud et al., 2009; Ricci et al., 2015; Nozaki et al., 2017; Miron et al., 2020). In cells treated with B02, the L-function had a higher peak, shifted to the left (Figure 3E), interpreted as increased nucleosome clustering and smaller cluster sizes. An alternative explanation would be that B02 massively decreases histone contents. However, this was not seen by Western blot analysis (Supplemental Figure S4A). Moreover, quantification from dSTORM images revealed no difference in the total number of H2B-HT molecules between controls and RAD51-inhibited cells (Supplemental Figure S5C). Hence, there was rearrangement rather than loss of nucleosomes in B02-treated cells.

Nanoscale clusters lower the mobility of chromatin loci in a biophysical chromatin model

To predict the effect of nucleosome clustering on the mobility of chromatin loci, we used a bead-and-spring polymer model with different levels of looping (cross-links) to simulate clusters. Cross-links within or between chains in the model represent chromatin interactions with proteins such as condensin or cohesin. Introduction of cross-linking within or between polymer chains recapitulates key experimental observations of nucleoli in budding yeast (Hult et al., 2017; Walker et al., 2019). For example, changes in cross-linking account for the heterogeneity in the density of ribosomal DNA within the nucleolus, the crescent shape of the nucleolus, and the position of two chains in cells with the ribosomal DNA split between two chromosomes (Hult et al., 2017). To determine the consequences of nucleosome clustering on chromatin mobility, we examined chromatin motion in the absence or presence of cross-linking (Figure 4A). The cross-links are dynamic through regulation of the mean On and Off times of binding. Two cross-linking regimes were implemented: fast crosslinking (0.19 s On / 0.019 s Off), which results in rigid chromatin clusters, and slow crosslinking (10 s On/1 s Off), resulting in more flexible clusters. Our model predicts a decrease in chromatin mobility in the presence of either rigid or flexible clusters compared with unconstrained chains (no looping) (Figure 4B).

FIGURE 4:

Relationship between chromatin cross-linking and mobility in a polymer model. (A) Schematic illustration of the bead-and-spring chromatin model at 5 kilobase (kb) resolution. (B) MSD plots of beads in the polymer model corresponding to the left arm of chromosome III (Chain 5, 24 beads total) or to the right arm of chromosome XII (Chain 24, 546 beads total). The plots represent mean ± SEM (N = 10 simulations) of different beads. Different looping behaviors were compared: no crosslinking (green), long cross-link lifetimes (slow crosslinking, pink), and short cross-link lifetimes (fast cross-linking, blue).

RAD51 inhibition reduces motions of individual chromatin loci

To assess B02’s effect on the dynamics of individual chromatin loci, matching the physical scale represented by our biophysical model, we expressed the Lac-repressor fused to GFP and a nuclear localization signal (LacI-GFP) in U2OS cells with a stable genomic integration of lacO arrays (Vidi et al., 2012) (Figure 5A). As expected, LacI-GFP diffusion values measured in live cells were more than an order of magnitude higher than in fixed cells, indicating a negligible contribution of noise and drift to the chromatin diffusion measurements (Figure 5B). Using the same B02 treatment condition as for the chromatin microdomain and single-nucleosome analyses (1 h, 10 µM), we found that the drug significantly decreased LacI-GFP motions (Figure 5B), as predicted in silico.

FIGURE 5:

RAD51 inhibition reduces motions of individual chromatin loci across species. (A) Schematic and micrograph of a U2OS cell with a genomic integration of lacO DNA arrays and expressing the lac repressor fused to GFP (LacI-GFP). Gamma adjustment (0.5) was performed for visualization of both the LacI-GFP focus and pan-nuclear signals. Scale bar, 10 µm. (B) Diffusion (D) of LacI-GFP foci in untreated cell and in cells treated with B02 (10 µM, 1 h). Fixed cells were used as controls. *, P = 0.042; N = 37 (fixed cells), N = 57 (control), and N = 81 (B02) LacI-GFP foci from three independent experiments. (C) Yeasts with a genomic integration of lacO visualized with LacI-GFP. A spindle pole body marker (SPB; Spc29-RFP) is used for image registration. Scale bar, 2 µm. (D) Diffusion of LacI-GFP foci in yeasts treated with B02 (10 µM, 1 h). The control was untreated. The data are presented as a MSD plot (mean ± SEM; N = 81 [control] and N = 77 [B02]). (E) LacO/LacI-GFP in Arabidopsis roots. A file of cell nuclei is shown enlarged in the inset. Scale bar, 20 µm. (F) Displacement calculated from the motions of spot pairs, corresponding to the fluctuation of spot-spot distances. (G) Quantification of average LacI-GFP spot displacements. Roots were placed on regular MS medium (control) or on medium supplemented with B02 (10 µM) for 18 h before the measurements. The symbols represent individual cells from two different lines (#26 and #112). **, P = 0.0011; ****, P < 0.0001

RAD51 is a highly conserved protein. Human RAD51 notably shares 67% with yeast scRad51 and 69% identity with the Arabidopsis thaliana RAD51 orthologue (AtRAD51). Molecular docking studies indicate that B02 binds within the RAD51 dimerization interface and/or RAD51’s Walker ATP binding site of RAD51 (Vydyam et al., 2019; Shkundina et al., 2021). Both domains are conserved across species (Supplemental Figure S6). To determine whether the effect of RAD51 inhibition on chromatin loci motions is a general phenomenon, we used other model organism in which Lac array systems have been developed to track chromatin motions: budding yeast and the plant Arabidopsis thaliana. As in mammalian cells, we measured slower chromatin loci motions after B02 treatment in yeasts (Figure 5, C and D). In Saccharomyces cerevisiae, DSBs cause the accumulation of Rad52 into prominent repair foci (Lisby et al., 2001; Lawrimore et al., 2017). Using cells expressing Rad52-GFP, we found that, as in mammalian cells, the B02 treatment did not cause DNA breaks, nor prevent the DNA damage response elicited by bleomycin treatment (Supplemental Figure S7, A and B). The effect of B02 therefore appears to be unrelated to HRR functions of scRad51. At a higher concentration (100 µM), B02 did block Rad52 foci formation following bleomycin treatment, indicating that the inhibitor is also active against scRad51 (Supplemental Figure S7C).

Next, we tested the effect of B02 on plant chromatin motions, using lacO/LacI-GFP Arabidopsis thaliana lines (Meschichi et al., 2022). We focused on primary root tips, which have clear LacI-GFP signals (Figure 5E). To assess the effect of RAD51 inhibition on chromatin motions in this system, seedlings were transferred onto B02-containing growth substrate (or growth substrate without B02 as control) and imaged 16–18 h later. We extended the treatment time (compared with mammalian cells and yeasts) to ensure penetration of the molecule throughout the root tissue. The diffusion speed of individual LacI-GFP foci was reduced in roots treated with B02 (Supplemental Figure S8). In the division zone, the vast majority of cell nuclei have two fluorescent LacI-GFP spots. We could therefore use paired particle tracking to assess B02’s effect on chromatin motions. With this approach, fluctuation of the distance between the fluorescent spot is measured as a function of time (Liu et al., 2015) (Figure 5F). Comparing the average change in spot separation between frames (Figure 5G) further indicates that the effect of B02 on chromatin dynamics (at least at the level of individual chromatin loci) is conserved in plant cells. We made the same observation in two independent Arabidopsis lacO/LacI-GFP lines. In conclusion, using the same molecular tool in different species, we find that the effect of RAD51 inhibition on chromatin loci dynamics is conserved.

DISCUSSION

Our results show that inhibition of RAD51 alters chromatin dynamics, in the absence of DNA damage, and that this effect is dependent on the physical scale considered. Previous work revealed that nuclear processes influence chromatin motions differentially at different scales. For example, the cell cycle status influences motions of individual chromatin loci (1–2 kb) but not of single nucleosomes or chromatin microdomains (Heun et al., 2001; Ma et al., 2019; Iida et al., 2022; Locatelli et al., 2022). Also, the transcription machinery “stirs” actively transcribed regions, and motions of enhancers and promoters increase upon activation (Gu et al., 2018). Yet, globally, active RNA polymerase II constrains nucleosome motions, likely by stabilizing chromatin domains around transcription hubs (Nagashima et al., 2019). Chromatin motion regimes are distinct at different physical scales, and we indeed find that RAD51 inhibition has an impact on chromatin motions at the microscale, without affecting single-nucleosome motions. At the nanoscale, thermal effects are likely major contributors to the molecular dynamics.

At the scale of chromatin microdomains, RAD51 inhibition or depletion leads to a global increase in chromatin dynamics. A possible explanation is that nucleic acid structures stabilized by RAD51 (such as R-loops) may be inherently less mobile or less prone to collective microscale motions. For example, R-loops regulate DNA loop extrusion by cohesin (Zhang et al., 2023), a key driver of chromatin dynamics. Interestingly, a scRad51 mutant defective in strand exchange (and hence HRR) but able to bind ssDNA retains the capacity to increase diffusion of chromatin loci distal from DNA breaks, indicating that ssDNA binding is an important regulator of global chromatin mobility (Smith et al., 2018). We note that rad51Δ yeast strains have similar chromatin dynamics compared with wild-type cells in the absence of DNA damage (Seeber et al., 2013; Smith et al., 2018), suggesting that in yeast, depleting versus inhibiting Rad51 leads to different outcomes.

It is also important to consider that RAD51 binds double-stranded DNA (dsDNA) (Benson et al., 1994; Zaitseva et al., 1999) and that B02 may inhibit dsDNA binding by RAD51 (Huang et al., 2012). In contrast to ssDNA binding, necessary for the recombination process, dsDNA binding by RAD51 protects DNA and inhibits pairing with homologous sequences (Zaitseva et al., 1999). The noncanonical function of RAD51 in the protection of reversed replication forks (distinct from its recombination function) relies on dsDNA binding (Halder et al., 2022). It remains to be established if RAD51 binds dsDNA in other contexts, which may globally affect chromatin dynamics. RAD51 ChIPseq data do indicate an enrichment of the protein at active chromatin regions (which are preferentially repaired by HRR), where it colocalizes with transcription factors (Aymard et al., 2014; Hazan et al., 2019). A reanalysis of these datasets indicates preferential binding of RAD51 to enhancer box elements (Kang et al., 2021), suggesting a role of RAD51 in transcription regulation, with the caveat that it is difficult to control for the presence/absence of DNA breaks in cell population–based assays.

Homology between RAD51 orthologues in eukaryotes is high. Moreover, B02 inhibits RAD51 foci formation and HRR both in human cells (Huang et al., 2012; Urban et al., 2016) and in Plasmodium falciparum (Vydyam et al., 2019), indicating that B02’s inhibitory mechanism is conserved across species. It was therefore not surprising to measure similar effects of the B02 RAD51 inhibitor in mammalian cells, yeasts, and plants. As in mammalian cells, the effect in yeasts was highly likely independent from scRad51’s role in HRR since B02 did not cause DNA damage in yeast cells. As documented in yeast, DNA damage in plants increases the mobility of LacO chromatin arrays (Meschichi et al., 2022). Yet (and in contrast to the increased motions of chromatin microdomains), B02 reduced mobility of lacO genomic loci across species, further arguing for an effect independent from the canonical role of RAD51 in HRR. Based on super-resolution imaging, the drug also increases chromatin nanodomain clustering, predicted by our model to reduce mobility at the 5 kb scale (5 kb DNA/spring). Future work is needed to determine how RAD51 inhibition affects nanoscale chromatin organization.

MATERIALS AND METHODS

Mammalian cell culture and treatments

U2OS osteosarcoma cells were cultured in DMEM supplemented with 10% FBS (Sigma) in a humidified incubator at 37°C, 5% CO₂. Cultures were tested routinely for Mycoplasma contamination; all tests were negative. KD of RAD51 was achieved by transducing cells with lentiviral particles combining Cas9 and sgRNA under the hEF1a promoter (Horizon discovery, Edit-R All-in-one lentiviral system [lentiviral particles; 10⁷ TU/ml], catalogue no. VSGH11936-247505860). The target sequence was TTGGCCCACAACCCATTTCA, corresponding to exon 2 of RAD51 (genomic location: chr15:40698817-40698839). A 1:50 dilution of the lentiviral particles was used for transduction. Transduced cells were selected with puromycin (2 µg/ml). For imaging, cells were seeded in 35 mm glass-bottom dishes (MatTek) at 100,000 cells per dish. For fixation, cells were incubated 20 min in 4% paraformaldehyde (Sigma # HT5011; 20 min). Treatments with B02 (3-(Phenylmethyl)-2-[(1E)-2-(3-pyridinyl)ethenyl]-4(3H)-quinazolinone; Merck, catalogue no. SML0364) were for 1 h, at the concentrations indicated in the Results section. RI-1 (GlpBio, catalogue no. GC18140) was diluted to 25 µM and added to the cells for 24 h. Bleomycin (Cayman Chemical; 20 mU/ml, 2 h) or mitomycin C (Accord; 2.5 µM, 18 h) was applied to the cells to induce DNA damage. The double thymidine block was used for cell cycle synchronization as described (Harper, 2005).

Tracking chromatin microdomains in mammalian cells

Photoactivated lattices of chromatin microdomains were generated in U2OS PAGFP-H2A cells (Locatelli et al., 2022) with a custom diffractive optical element module inserted into the condenser arm of an inverted Olympus IX83 microscope (Bonin et al., 2018). Cells were kept at 37°C in the custom-built enclosure of the microscope. Images were taken with a 60x oil lens (N.A. = 1.35) and an sCMOS camera (Hamamatsu ORCA-Flash 4.0), using the CellSens software. Movies were recorded over one minute at 3.3 frames per second (fps). The image sequences were registered using the StackReg plugin in FIJI (Thevenaz et al., 1998). Tracking of chromatin microdomains was done in MATLAB, as described previously (Bonin et al., 2018; Locatelli et al., 2022). To avoid potential confounding effects of large nuclear deformation, cells with distortions greater than an arbitrary threshold (ΔL ≥ 400 pixels, with ΔL, the cumulative change in the perimeter of the largest quadrilateral defined by the grid of spots, see below) were excluded from the analyses.

Tracking genomic loci in mammalian cells

U2OS cells with a stable genomic integration of Lac repressor binding arrays (lacO; 256 copies; Vidi et al., 2012) were used to measure the mobility of genomic loci. LacI-GFP was transiently expressed using Lipofectamine 3000 (Invitrogen)-mediated transfection, performed 48 h before imaging. Image sequences were collected with an IX83 Olympus microscope equipped with an ORCA-Quest qCMOS camera (Hamamatsu) and a 40x air objective (N.A. = 0.95). A TokaiHit stage incubator was used to maintain environmental conditions (37°C, 5% CO₂). Excitation light was from the 475 nm line of a LED illumination system (X-Cite TURBO, Excelitas). Emitted light was filtered using a 525/50 nm filter (Chroma). The frame rate was 3 fps. For analysis, individual cells were cropped from the time lapses and registered with StackReg (rigid body translations/rotations) (Thevenaz et al., 1998). The MOSAIC Particle Tracker (Sbalzarini and Koumoutsakos, 2005) was used to detect LacI-GFP spots, track trajectories, and compute chromatin diffusion from mean squared displacement (MSD) plots. Some diffusion values were aberrant (supraphysiological), likely due to tracking errors, and were excluded using an outlier analysis (ROUT method, Q = 0.1).

Tracking genomic loci in yeasts

Budding yeast strain KS406 (MATalpha, ade5-1, trp1-289, ura3del, leu2-3, 112, lys2::insI-Sce1 [lacO array next to RAD16 promoter, tetO array next to LYS2 promoter (pRS305tetO)], arg4::hisG Gal1/10 I-Sce1, thr1::HISpLacI-GFP:Nat, ade1::URAptetR-CFP:Hyg, Spc29-RFP:Bsd) and KBY3584 (YEF 473A Rad52-GFP::Kan Spc29-CFP::HygB) were grown to midlog phase at 24°C in liquid YEPD and imaged in YC complete medium with 2% filter-sterilized glucose added. Image sequences were acquired on an Eclipse Ti wide-field inverted microscope (Nikon) with a 100x Apo TIRF 1.49 N.A. objective (Nikon) and a Clara charge-coupled device camera (Andor) using the Nikon NIS Elements imaging software. Images were taken over 20 min with 30 s intervals. Each interval contained a seven-step Z-stack of 400-nm step size, acquired for KS406 in the GFP (lacO array), RFP (SPB) and Trans channels, or for KBY3584 in the GFP (Rad52), CFP (SPB) and transmission channels. After manual cropping, time lapses were tracked using TrackMate (FIJI plugin) and a custom MATLAB program to compute MSD, or Rad52 positive cells counted.

Tracking genomic loci in plant cells

The Arabidopsis thaliana lacO/LacI-GFP lines used in this study are in the Col-0 background. Seeds of the lines #26 and #112 were sterilized for 5 min in 70% ethanol and dried before sowing on 100 × 15 mm Petri dishes containing 1x MS medium supplemented with 1% sucrose, vitamins (1 µg/ml thiamine, 0.5 µg/ml pyridoxine, and 0.5 µg/ml nicotinic acid), 0.05% MES and 0.8% agar. The pH of the medium was adjusted to 5.5. Seeds were stratified in the dark at 4°C for 2–3 days to synchronize germination and grown vertically under continuous light (150 µmol photons m⁻² s⁻¹). Five-day-old seedlings were transferred onto new MS plates containing 0 (control) or 10 µM B02 for 16–18 h. For measurements, the seedlings were placed in 35 mm glass-bottom dishes (MatTek) and covered with medium “slabs” with the same composition as described above. Image sequences were recorded from the root division zone at 40x magnification, with the same IX83 microscope used for mammalian Lac arrays (see above). The temperature was maintained at 22°C by cooling the room to ∼18°C and heating the custom microscope enclosure to the desired temperature. The acquisition time was 15 min, with 6 s intervals. Z-stacks were collected over a 3 µm range (seven frames, 0.5 µm z-spacing). Focus was maintained with a laser-based z-drift compensation system (TruFocus; IX3-ZDC). Image preprocessing was done in FIJI. Maximal intensity projections were generated from the z-stacks and intensity was normalized throughout time series. Images (corresponding to the entire root division zone) were registered using StackReg and LacI-GFP spots tracked with the MOSAIC two-dimensional (2D) tracker. For paired particle tracking, individual cells were cropped from the movies before tracking. The distance between the two spots, retrieved from the x/y spot coordinates was used to generate MSD using custom code in MATLAB.

Single-nucleosome tracking

We used U2OS cells stably expressing H2B fused to the HaloTag and a custom-built oblique light sheet microscope for single-nucleosome tracking, as described previously (Kefer et al., 2021). Before imaging, the cells were incubated with 10 pM fluorescent JF 646 HaloTag ligand (Grimm et al., 2015) for 1 h, washed with PBS, and incubated in DMEM without phenol red for at least 30 min. Cells were kept at 37°C in a live cell imaging chamber (Tokai Hit). Images were taken with a 60x oil immersion objective (N.A. = 1.2, Olympus, also used to generate the oblique light sheet), an ORCA-Flash 4.0 sCMOS camera, and a solid-state 640 nm laser. μManager (Edelstein et al., 2014) was used to capture images at a frame rate of 20 fps. Single-nucleosome motions were quantified with custom software written in Python. Briefly, the localization of each nucleosome spot was detected and fitted with a 2D Gaussian function. Nucleosomes trajectories were established with a multitemporal association tracking algorithm and used to compute MSD from which the diffusion coefficient D was calculated.

Super-resolution imaging of single nucleosomes

The same imaging platform as for single-nucleosome tracking was used for dSTORM. Cells were labeled (1 pM, 16 h), washed, and fixed in 3.7% formaldehyde solution in PBS buffer (pH = 7.5).

Prior to imaging, dSTORM buffer was added to the cells. This buffer is based on glucose oxidase/catalase/cysteamine, as described previously (Van de Linde et al., 2011). Images were acquired at a rate of 100 fps using μManager with a laser power of 20 mW at the objective. For single-nucleosome localization, we first selected nuclear regions and reduced background noise using FIJI's rolling ball method with a radius of 10 pixels. To minimize the localization error caused by pixel dependent readout noise from sCMOS cameras, we used an iterative algorithm (Huang et al., 2013) that uses a maximum likelihood estimation of the position and intensity of each nucleosome (Smith et al., 2010). After obtaining precise x,y positions of each nucleosome, we analyzed the clustering using Besag's L-function L(r) (Besag, 1977). The L-function (Equation 1) is obtained by transforming Ripley's K-function K(r) in Equation 2 where a is the area of the window, r is the distance separation, and n is the number of data points (Ripley, 1977). The indicator function I(d_ij ≤ r) gives 1 if the distance is less than or equal to r. Degree of clustering is measured by L(r) − r.

1

2

Immunofluorescence

Immunofluorescence analyses of fixed U2OS cells was done as described previously (Locatelli et al., 2022), with antibodies against RAD51 (AbCam, catalogue no. Ab63801; 1:300), γH2AX (Millipore, clone JBW301, 2 µg/ml); H3K9me3 (Cell Signaling Technology, catalogue no. 13969S; 1:800), and H2K9ac (AbCam, catalogue no. Ab12179; 1:500). Secondary antibodies were coupled to AlexaFluor dyes (Life Technologies; 1:500). Nuclei were stained with 0.5 µg/ml 4′,6-diamidino-2-phenylindole (DAPI). Samples were mounted using ProLong Gold antifade (Molecular probes) and imaged with an IX83 Olympus microscope, using a 60x oil immersion objective (N.A. = 1.35) and the appropriate filter cubes (Chroma).

Imaging cytometry

To assess cell cycle distribution, DAPI-stained nuclei were imaged with a 4x (N.A. = 0.13) lens and a qCMOS camera (ORCA-Quest, Hamamatsu) on an IX83 fluorescence microscope, applying shading correction. Background was subtracted using FIJI's built-in rolling ball method (using a radius of 100 pixels). Nuclei were segmented with StarDist (Schmidt et al., 2018) to retrieve integrated densities of DAPI signals. Because signal intensity varied slightly between images, intensity values for each nucleus of an image were scaled by subtracting the mode value (corresponding to the G1 peak position), then adding an arbitrary value (400) to avoid negative intensities. Intensity histograms were plotted in GraphPad Prism for visualization.

Western blot

Cells were lysed in PBS containing 2% SDS, then sonicated. Protein concentration was quantified using the DC protein assay (Bio-Rad, catalogue no. 500-0122), and 30 µg of each sample were resolved by SDS–PAGE. Proteins were transferred onto nitrocellulose membranes for immunoblotting with antibodies against H3K9me3 (Cell Signaling Technology, catalogue no. 13969S; 1:1000), H3K9ac (AbCam, catalogue no. Ab12179; 1:1000), H2A (Cell Signaling Technology, catalogue no. 12349S; 1:1000), RAD51 (AbCam, catalogue no. Ab63801; 1:1000), and HSC70 (Santa Cruz Biotechnology, catalogue no. sc7298; 1:1000). Secondary antibodies were anti-rabbit or anti-mouse horseradish peroxidase–coupled IgGs (GE Healthcare, catalogue nos. NA934 and NA931; 1:10000). Enhanced chemiluminescence signals were captured with a Fusion Solo imaging system (Vilber Lourmat) and quantified using FIJI.

Cell viability assay

Cells were seeded in 96-well plates at a density of 10,000 cells per well and treated as described above. MTT reagent (3-4,5-dimethylthiazol 2,5-diphenyltetrazolium bromide; 10 µl,  5 mg/ml) was added to each well and incubated in the cell culture incubator for 2 h. The medium was discarded, and formazan crystals were resuspended in 100 µl DMSO. Absorbance was measured using a microplate reader (Tecan Group Ltd.) at 650 nm. For clonogenic survival assays, cells were trypsinized and seeded into 12-well plates (10,000 cells/well). Colony formation was visualized after 7 d by fixing cells in formalin and staining 20 min with crystal violet (0.5% in 20% ethanol). Plates were air-dried before imaging.

Chromatin polymer model

Chromosomes were modeled as bead-spring chains, where each chain is analogous to one of the 32 chromosome arms in budding yeast (Hult et al., 2017). In this model, individual beads represent nonoverlapping chromatin segments 5 kb in length. The entire yeast genome is represented by 2803 beads organized across 32 chains (16 chromosomes). Each bead is described by its x/y/z coordinates and a set of global parameters dictating bead volume, Brownian motion, tensile forces between adjacent beads, viscosity, drag, and interactions with the nuclear membrane (Hult et al., 2017). The three-dimensional (3D) coordinates of individuals beads were recomputed every 0.001 s of simulation time based on the sum of these forces, generating sequential snapshots of chromatin spatial conformations. The initial configuration of chromosomes was with centromeres tethered to the inner surface of the nucleus, at the same point, according to the Rabl configuration (Berger et al., 2008; Taddei and Gasser, 2012). The telomeres on each of the arms were also tethered to the inner nuclear membrane across six distinct positions. In the starting configuration (time 0), the beads in each arm are arranged linearly between the two tether sites. Crosslinks were coded as the property of select beads to form crosslinks with every other bead that shares this property independent of their chain of origin. Cross-links are realized as additional springs connecting the centers of two beads and influencing bead positions as additional sources of force. Analogous to springs of the bead-spring model, the cross-links do not have volume or shape in themselves. When two beads are cross-linked, there is an additional tension force between them of magnitude 50x relative to springs linking the polymer chain. Beads form cross-links with a bead that is not currently cross-linked to another bead; every pair of cross-linked beads will remain bound for a set amount of time or until the distance between them exceeds 90 nm. Once a cross-linker detaches, both beads are unavailable for cross-linking for a set amount of time. The ratio of bound-to-unbound time is defined in initial parameters of the simulation as the Mean On/Off time (Walker et al., 2019). To allow the system to normalize, data collection began after 2400 s of simulation time. The 3D coordinates of individual beads were saved every dT = 0.1 s, for a total of 1300 s of data per simulation. The simulation code, code to generate MSD, and the exponential function fit were run in ImageTank (Visual Data Tools).

Statistical analyses

Statistical analyses were done using GraphPad Prism 9. The D'Agostino–Pearson omnibus normality test was used to verify normal distributions. Nonparametric tests were used if the data did not pass the normality test (at alpha = 0.05). The t test (or Mann–Whitney test) was used for comparisons between two conditions, whereas ANOVA and Tukey post-hoc test (or Kruskal–Wallis and Dunn's multiple comparisons test) was used for datasets with more than two conditions. P-values < 0.05 were considered significant. All statistical tests were two sided. The box and whisker plots represent medians, 25th–75th percentiles (boxes), and 10th–90th percentiles (whiskers).

Abbreviations used:

B02

3-(Phenylmethyl)-2-[(1E)-2-(3-pyridinyl)ethenyl]-4(3H)-quinazolinone, RAD51 inhibitor

dSTORM

direct stochastic optical reconstruction microscopy

DSB

DNA double-strand break

FPS

 frames per second

HRR

homologous recombination repair

HT

HaloTag

KD

 knock down

LacI

lac repressor

LacO

lactose operon

MSD

mean-squared displacement

PAGFP

photoactivatable GFP

RI-1

3-chloro-1-(3,4-dichlorophenyl)-4-morpholino-1H-pyrrole-2,5-dione, RAD51 inhibitor.