Purification of RAD51 proteins
Table of Contents
The human RAD51 and RAD51 mutant proteins were purified as previously described23. In brief, His6-tagged human RAD51 and RAD51 mutants (K64A, K70A, K64A/K70A, R27A, S67E, R235D and E59R) were produced in E. coli cells and purified by Ni-NTA agarose chromatography (Qiagen). The His6-tag portion was removed by thrombin protease treatment. The RAD51 proteins were then precipitated with spermidine, and dissolved in potassium phosphate buffer. RAD51 proteins were further purified by MonoQ column chromatography (Cytiva).
Purification of histones
Human histones were purified as previously described38. In brief, human histones (H2A, H2B, H3.1, H4, taillessΔ15 H4 and taillessΔ19 H4) were produced in E. coli cells as His6-tagged peptides. His6-tagged histones were denatured with urea, and purified by Ni-NTA agarose chromatography (Qiagen). The His6-tag portion was removed by thrombin protease treatment, and the histones were further purified by MonoS column chromatography (Cytiva).
Nucleosome preparation
The nucleosomes with the 153-base-pair (with a three-base 3′ ssDNA overhang) and 158-base-pair (with blunt ends) Widom 601 DNA were prepared as previously described38,39. In brief, the histone octamer was reconstituted with histones H2A, H2B, H3.1 and H4, and the resulting histone complex was purified by gel filtration chromatography on a HiLoad16/60 Superdex 200 column (Cytiva). The 158-base-pair Widom 601 DNA fragment with blunt ends was amplified by PCR and purified by native polyacrylamide gel electrophoresis, using a Prep Cell apparatus (Bio-Rad). The sequence of the 158-base-pair DNA fragment is as follows: 5′-CGTGGTGGCCGTTTTCGTTGTTTTTTTCTGTCTCGTGCCTGGTGTCTTGGGTGTAATCCCCTTGGCGGTTAAAACGCGGGGGACAGCGCGTACGTGCGTTTAAGCGGTGCTAGAGCTGTCTACGACCAATTGAGCGGCCTCGGCACCGGGATTCTGAT-3′. Nucleosomes were reconstituted by the salt dialysis method and subsequently purified with a Prep Cell apparatus38.
Preparation of the RAD51–nucleosome complex
RAD51 (2.4 µM) and the nucleosome (0.1 µM) were mixed in reaction buffer (20 mM HEPES-NaOH (pH 7.5), 15 mM NaCl, 1 mM MgCl2, 1 mM dithiothreitol, 0.2 mM 2-mercaptoethanol, 0.03% NP-40 and 1.5% glycerol) in the absence or presence of 1 mM nucleotide (ATP, ADP or AMP-PNP). After an incubation at 37 °C for 30 min, the resulting complexes were fixed by the GraFix method40 in the gradient solution (15–30% sucrose and 0-0.2% glutaraldehyde gradient in 10 mM HEPES-NaOH (pH 7.5), 30 mM NaCl and 1 mM DTT). The samples were applied on top of the gradient solution and centrifuged at 27,000 rpm at 4 °C for 16 h in an SW41Ti rotor (Beckman Coulter). After the ultracentrifugation, 640-μl aliquots were obtained from the top of the gradient solution and analysed by 4% non-denaturing polyacrylamide gel electrophoresis in 0.5× TBE buffer (45 mM Tris-borate and 1 mM EDTA), followed by ethidium bromide or SYBR Gold staining. The fractions containing the RAD51–nucleosome complexes were collected, and the buffer was then exchanged using a PD-10 column (Cytiva) to the final buffer (10 mM Tris-HCl (pH 7.5), 30 mM NaCl and 1 mM dithiothreitol). The resulting sample was concentrated with an Amicon Ultra 30K filter (Merck Millipore).
Cryo-EM grid preparation and data collection
Aliquots (2.5 μl) of the purified RAD51–nucleosome complexes in the absence or presence of nucleotide (ATP, ADP or AMP-PNP) were applied to glow-discharged Quantifoil R1.2/1.3 200-mesh Cu grids. The grids were blotted at 4 °C for 4 or 6 s under 100% humidity using a Vitrobot Mark IV (Thermo Fisher Scientific), and then rapidly frozen in liquid ethane. Cryo-EM data of the RAD51–nucleosome complexes were collected on a Krios G4 microscope (Thermo Fisher Scientific) operating at 300 kV and a magnification of 81,000× (pixel size of 1.06 Å). The data acquisition was performed using the EPU automation software (Thermo Fisher Scientific). The defocus range varied from −1.0 to −2.5 μm. A K3 BioQuantum direct electron detector (Gatan) was used, and a stack of 40 frames was obtained for each dataset. The detailed conditions that were used for obtaining the cryo-EM data are shown in Extended Data Tables 1 and 2.
Image processing
The detailed process is shown in Extended Data Figs. 2–5 and 7. All frames in the movies of each dataset were aligned using MotionCor241 with dose weighting, and the contrast transfer function (CTF) estimation was then performed using CTFFIND442 on digital micrographs. Micrographs were selected on the basis of the strong correlation of the CTF. The following image-processing steps were performed using Relion 4 beta216. Picked particles by Laplacian-of-Gaussian (LoG)-based auto-picking were subjected to two-dimensional (2D) classification, and 2D class averages with nucleosome and additional densities were used as references for the following particle picking. Picked particles were extracted from micrographs with 2× binning. Further 2D classification was performed to discard junk particles. An initial model was then generated de novo, and several rounds of 3D classification were performed using a reasonable model as a reference. After removing the 2× binning, Bayesian polishing and CTF refinement were conducted. A mask was created around the RAD51 ring, and further 3D classification was performed using the created mask. The final map was generated by using high-quality classes for sharpening in each class with various numbers of RAD51 molecules.
In the dataset of RAD51–nucleosome complexes containing the 153-base-pair DNA obtained in the presence of ATP, two classes were identified: one with the RAD51 ring bound to the linker DNA and nucleosome, and another with an additional RAD51 ring bound to the nucleosome without linker DNA binding. In the first round of 3D classification, these classes were separated. For the image processing of the RAD51 ring bound to the nucleosome without linker DNA binding, focused refinement on the RAD51-ring structure was performed. To analyse the binding of the RAD51 L1 loop to the sticky DNA end, focused refinement was performed after the CTF refinement by masking the sticky DNA end and RAD51. We conducted this focused refinement with the RAD51 protomers bound to the linker DNA without selecting specific ring structures. To analyse RAD51 binding to the histone H4 tail, focused refinement was performed by masking the region around the histone H4 tail. We conducted this focused refinement with the H4 tail without selecting specific ring structures.
In the analysis of the dataset of the RAD51–nucleosome complex containing the 158-base-pair DNA with blunt ends obtained in the presence of ATP, only the RAD51 ring bound to the linker DNA was analysed. The structures were refined separately, on the basis of the number of protomers in each RAD51 ring. To analyse the binding of the linker DNA to the RAD51 L1 loop, focused refinement was performed by masking the DNA blunt end and RAD51. We conducted this focused refinement with the RAD51 protomers bound to the linker DNA without selecting specific ring structures.
In the analysis of the samples obtained in the presence of ADP, fractions separated by GraFix were obtained: one containing complexes with RAD51 rings bound to linker DNA, and the other containing two rings of RAD51 bound to the nucleosome. For each dataset, the RAD51 ring was focused and refined, resulting in the final maps.
For the samples obtained in the presence of AMP-PNP, two datasets of the F1 and F2 fractions were collected individually (Extended Data Fig. 5). For the image processing of F1, the processes were performed as described above. For the image processing of F2, the 2D class averages of nucleosomes with additional densities were obtained, and used as the reference for particle picking. The filament structure of RAD51 bound to the nucleosome was obtained by 2D classification and two rounds of 3D classification, and used as the reference for Topaz particle picking43. After 2D and 3D classifications, Bayesian polishing and CTF refinement were performed without 2× binning, and the dimer and monomer structures of the nucleosome–RAD51 filament complex were obtained. In addition, 2D class averages of the naked DNA–RAD51 filament structure were obtained, and used as the reference for particle picking. By 2D classification and two rounds of 3D classification, the cryo-EM map of the naked DNA–RAD51 complex was obtained from the reference-based particle picking of the filament structure.
Model building
The atomic models of the RAD51–nucleosome complexes were built using the atomic coordinates of the histone octamer from the human nucleosome (Protein Data Bank (PDB) ID: 5Y0C)44 and the atomic coordinates of a 145-bp Widom 601 sequence from the Xenopus laevis nucleosome (PDB ID: 7OHC)45. The atomic model of RAD51 was built using the crystal structure of human RAD51 (PDB ID: 5NWL)46, and refined using the cryo-EM map of the highest-resolution RAD51 single molecule with phenix.real_space_refine47. The atomic coordinates of the NLD and RecA domains were adjusted and fitted to each cryo-EM map. The sequences of the nucleosomal DNA and linker DNA were modified using Chimera48. The atomic coordinates of the DNA were refined by manual editing with ISOLDE49 and Coot50. The resulting atomic coordinates of RAD51 and nucleosome were fitted to the cryo-EM map by rigid body fitting, using the ‘Fit in Map’ mode of ChimeraX51. The major clashes were modified with phenix.real_space_refine and Coot.
For model building of the histone H4 tail bound to the RAD51, the atomic coordinates were refined by manual editing with ISOLDE and Coot.
For model building of the RAD51 filament–nucleosome complex, the DNA was built by connecting the nucleosomal DNA (PDB ID: 7OHC), the kinked DNA (PDB ID: 1WD1)52 and the extended DNA from the human RAD51 post-synaptic complex (PDB ID: 5H1C)28. The atomic coordinates of the nucleosomal DNA were refined by manual editing with ISOLDE.
Assay for RAD51–nucleosome or DNA binding
The nucleosomes (0.1 µM) or the 153-base-pair DNA (0.01 µM) and RAD51 or RAD51 mutants (0.24, 0.48 and 0.72 μM for DNA-binding assay, and 1.2, 2.4 and 3.6 μM for nucleosome-binding assay) were incubated at 37 °C for 30 min in the reaction buffer (20 mM HEPES-NaOH (pH 7.5), 15 mM NaCl, 1 mM MgCl2, 1 mM dithiothreitol, 0.2 mM 2-mercaptoethanol, 0.03% NP-40 and 1.5% glycerol) in the absence or presence of 1 mM ATP, ADP or AMP-PNP. The samples were analysed by 4% non-denaturing polyacrylamide gel electrophoresis in 0.5× TBE buffer (45 mM Tris-borate and 1 mM EDTA), followed by ethidium bromide staining. Band intensities were quantitated by an Amersham Imager 680 with ImageQuant TL (Cytiva).
Visualization of RAD51 in the absence of nucleosomes and DNA
Wild-type (WT) RAD51 (92.5 µM) and the K64A/K70A (73.3 µM) and R235D (55.2 µM) mutants were incubated at 37 °C for 30 min in reaction buffer (34 mM HEPES-NaOH (pH 7.5), 135 mM NaCl, 1 mM MgCl2, 0.9 mM dithiothreitol, 1.8 mM 2-mercaptoethanol, 0.03% NP-40 and 9% glycerol) in the presence of 1 mM AMP-PNP. Aliquots (2.5 μl) were applied to glow-discharged Quantifoil R1.2/1.3 200-mesh Cu grids. The grids were blotted at 4 °C for 4 or 6 s at 100% humidity, and then rapidly frozen in liquid ethane. Micrographs of RAD51 were collected on a Krios G4 microscope operated at 300 kV and a magnification of 81,000× (pixel size of 1.06 Å).
Saccharomyces cerevisiae strains and DNA damage sensitivity assays
The S. cerevisiae strains used in this study are listed in Extended Data Table 3. The rad51 deletion strain (Δrad51 strain) was generated by replacing the endogenous Rad51 gene with the kanamycin resistance gene (kanMX6). To construct rad51Δ + rad51 WT and mutant stains, the rad51 deletion (rad51Δ::kanMX6) strain was transformed with DNA fragments containing Rad51 (WT, Δrad51, rad51K122A, rad51K128A or rad51K122A/K128A) -Ura3 genes, which were amplified by PCR or purchased (Integrated DNA Technologies). Strains were selected on synthetic complete medium without uracil (SC-Uracil: 6.7 g l−1 Difco yeast nitrogen base without amino acids (BD Biosciences, 291940), 1.92 g l−1 yeast Synthetic Drop-out medium supplements (Merck, Y1501-20G), 2% glucose and 2% Difco Bacto Agar).
S. cerevisiae cells were grown at 30 °C in yeast complete medium (YPD: 1% yeast extract, 2% peptone and 2% glucose) overnight. The pre-cultures were twofold diluted in YPD medium and incubated at 30 °C for 2 h. Afterwards, 8.0 × 107 cells grown in YPD medium were collected and suspended in 1 ml of sterile water, and tenfold serial dilutions were prepared. For all spots, 5-μl aliquots of serial dilution samples were spotted on YPD plates in the absence or presence of 0.02% (v/v) MMS, 30 µM CPT or 150 mM HU. To assess the X-ray sensitivity, yeast cultures spotted onto YPD plates were irradiated with a CellRad X-ray irradiator (Faxitron Bioptics). The plates were incubated at 30 °C for several days. The quantification was performed using the third spot (1:100 dilution) of X-ray irradiation, according to a previously described method53.
Protein extraction from S. cerevisiae and western blots
Cells (1.0 × 108) grown in YPD medium were collected and suspended in 500 µl of ice-cold sterile water, and 75 µl of lysis buffer (2 M NaOH, 7.5% 2-mercaptoethanol) was added. After an incubation on ice for 10 min, 75 µl of 50% (v/v) trichloroacetic acid was added. After another 10 min incubation on ice, pellets obtained by centrifugation were resuspended in 60 µl Laemmli Sample Buffer (Bio-Rad, 1610737) with 5% (v/v) 2-mercaptoethanol, and the pH of the suspension was adjusted to alkaline using 1 M Tris (pH 8.8). Samples were then incubated at 65 °C for 10 min and the supernatant was used as the extracted proteins.
To detect endogenous S. cerevisiae Rad51, the extracted proteins were separated by SDS 10%-polyacrylamide gel electrophoresis. The gels were transferred onto membranes using an iBlot 2 Gel Transfer Device (Thermo Fisher Scientific), and the membranes were blocked with Blocking One-P (Nacalai Tesque). The membranes were then probed with the rabbit anti-S. cerevisiae Rad51 (1:5,000; BioAcademia, 62-101) antibody, with HRP-conjugated anti-rabbit IgG (1:5,000; Merck; NA9340) as the secondary antibody. As a loading control, α-tubulin was detected by HRP-conjugated anti-tubulin α (1:5,000; Bio-Rad, MCA77P). Can Get Signal (TOYOBO) was used for antibody dilution. Signals were enhanced by ECL Prime (Cytiva) and detected using an Amersham Imager 680 (Cytiva).
Statistical analysis
Statistical analyses were performed using R and Python. For the electrophoretic mobility shift assays, differences in band intensities were assessed between the canonical nucleosome and each nucleosome containing a histone H4 deletion mutant, as well as between RAD51 and each RAD51 mutant at each RAD51 concentration. In the spot assay, differences in spot intensities were estimated between the + WT strains and each mutant strain. Welch’s t-test was used to assess the differences in the means of the two datasets without conducting any pre-tests, as recommended54.
Use of large language models
ChatGPT was used for grammatical correction of the text and supplied the basis of the Python programs. The programs were used to generate the quantification graphs and to process PDB files.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.