DMC1 and RAD51 bind FxxA and FxPP motifs of BRCA2 via two separate interfaces

Abstract

In vertebrates, the BRCA2 protein is essential for meiotic and somatic homologous recombination due to its interaction with the RAD51 and DMC1 recombinases through FxxA and FxPP motifs (here named A- and P-motifs, respectively). The A-motifs present in the eight BRC repeats of BRCA2 compete with the A-motif of RAD51, which is responsible for its self-oligomerization. BRCs thus disrupt RAD51 nucleoprotein filaments in vitro. The role of the P-motifs is less studied. We recently found that deletion of Brca2 exons 12-14 encoding one of them (the prototypical 'PhePP' motif), disrupts DMC1 but not RAD51 function in mouse meiosis. Here we provide a mechanistic explanation for this phenotype by solving the crystal structure of the complex between a BRCA2 fragment containing the PhePP motif and DMC1. Our structure reveals that, despite sharing a conserved phenylalanine, the A- and P-motifs bind to distinct sites on the ATPase domain of the recombinases. The P-motif interacts with a site that is accessible in DMC1 octamers and nucleoprotein filaments. Moreover, we show that this interaction also involves the adjacent protomer and thus increases the stability of the DMC1 nucleoprotein filaments. We extend our analysis to other P-motifs from RAD51AP1 and FIGNL1.

Graphical Abstract

Figure 1.

Definition of recombinase-binding motifs. The figure depicts schematically the motifs and domains of (A) BRCA2, and (B) the recombinases DMC1 and RAD51. We refer to the conserved FxPP (green) and FxxA (blue) motifs as P- and A-motifs, and regions on the recombinases binding to P- and A-motifs (dots: residues defining the sites) as P- and A-sites, as detailed in the results section. Residue numbers indicated for the domain boundaries and key amino acid positions relate to human proteins. Region deleted in the Brca2^Δ12–14 mouse (Δ) is indicated by a red bracket in panel A; it contains the two HSF2BP-binding sites (light pink) (68,69) and the DMC1-binding PhePP domain (violet). (C) Scheme of the BRCA2 fragments used in this study. The region deleted in the Brca2^Δ12–14 mouse (grey bar) and the fragments defined by Thorslund et al.(14) experimentally (’PhePP domain’, violet bar) and from consensus (‘PhePP motif’, dark blue) are shown at the top. The recombinant and synthetic peptides we used are shown in black. The vertical red bar marks the position of F2406. To avoid coining new terms, we will refer to the BRCA2 region we study, encoded by exon 14 and binding DMC1, as the ‘PhePP domain’, and the set of conserved positions within it as the ‘PhePP motif’, although the region boundaries and position set will be adjusted compared to the original definition by Thorslund et al. (14) based on the new data we present.

Figure 2.

Structure of the DMC1–PhePP complex. (A) AlphaFold models of the DMC1–PhePP interaction. The BRCA2 peptide is colored as a function of the lDDT score per residue. Side chains of the phenylalanine and proline defining the P-motif are displayed in sticks. DMC1 is colored in cyan in its N-terminal region (from M1 to E82), pale green in its linker region (P83 to M97) and dark green in its ATPase domain (V98 to E340). (B) AlphaFold model of the DMC1-BRC4 interaction, displayed as in (A). (C) Crystal structure of the octameric DMC1 (blue, green and wheat) bound to the BRCA2 peptide PhePP (F2s6, red). The blue protomers are observed from Q28 to K339 (one protomer is distinguished by being colored in green), whereas the wheat protomers are observed from I81 to K339. In all protomers, a single loop from A272 to either D283 or K285 is not detected. (D) Focus on the interactions between a wheat DMC1 protomer (surface view) and the adjacent green DMC1 protomer (cartoon view) as well as the red BRCA2 peptide from P2402 to S2414 (cartoon view). The side chains of the BRCA2 peptide are shown in red sticks. BRCA2 residues that are more than 30% buried upon DMC1 binding are marked (red labels correspond to residues more than 80% buried). The side chain of DMC1 F89 (green protomer), located in front of BRCA2 K2413, is displayed in dark green sticks and labeled. In the insets, the positions of the BRCA2 peptides in the X-ray structure (red) and in the AlphaFold models 1 and 2 (blue) are compared by superimposing the interacting DMC1 protomers.

Figure 3.

Validation of the DMC1–PhePP binding interface. (A) Crystal structure of the interface between DMC1 and the BRCA2 PhePP (F2s6) peptide, colored as in Figure 2D. In BRCA2, only side chains buried upon binding are displayed, and their label is colored as a function of their contribution to binding, as defined in panels B. In DMC1, buried sidechains are dark-red and labeled, DMC1-F89 forming cation-π interaction with BRCA2-K2413 is green, residues conserved in RAD51 are boxed. (B) Alignment of the central region of human BRCA2 F2s3 (2379–2433) sequence with homologous vertebrate BRCA2 sequences, conservation score and consensus sequence. Conserved positions tested by alanine substitution are displayed above the conservation plot. Circles above indicate the importance of the position for the interaction. (C) GST pull-down between BRCA2-F2s3 substitution variants and purified recombinant DMC1. Replicates are shown in Supplementary Figure S4B. Bound DMC1 band intensities normalized to the wild-type GST-F2s3 lane is shown in pink. (D) GST pull-down of DMC1 variants with GST-F2s3 BRCA2 fragment. A replicate is shown in Supplementary Figure S4C. Bound DMC1 band intensities normalized to the wild-type lane is shown in pink. (E) Interaction between BRCA2–PhePP (fragments F2s3 and F2s6, GST-tagged) and DMC1 (purified or in crude bacterial lysate, with or without his-tag) and its truncation variants, determined by GST pull-downs. A replicate is shown in Supplementary Figure S4D. Bound DMC1 band intensities normalized to the full-length his-DMC1, is shown in pink.

Figure 4.

A- and P- recombinase-binding motifs and sites. (A) Alignment of the sequences of human RAD51 and DMC1 recombinases. Green and blue dots correspond to residues buried at >30% upon binding of DMC1 to PhePP (P-site) and RAD51 to BRC4 (A-site), respectively. (B) Representation of the surfaces binding to PhePP (P-site; green) and BRC4 (A-site; blue), as defined in (A), onto the 3D structure of RAD51 bound to BRC4 (PDB 1N0W). To display the PhePP peptide bound to the RAD51 P-site, the crystal structures of DMC1-F2s6 and RAD51-BRC4 were superimposed, and the structure of DMC1 was hidden. (C, D) AlphaFold models of the DMC1 and RAD51 complexes with P- and A-motifs from different proteins (see also Supplementary Figure S5). The ATPase domain surface of the recombinase is displayed, with the P- and A-sites colored in green and blue, respectively. Peptides containing P- or A-motifs are shown as cartoons, colored based on the AlphaFold IDDT score (red - accurate to white - poorly defined). The phenylalanine side chain of the motifs is shown as sticks. (E) Sequences of P- and A-motif peptides. (F) BioLayer Interferometry experiment performed using an immobilized biotinylated PhePP (F2s6) peptide and increasing concentrations of DMC1. The measured apparent K_d is about 1 nM, which is unexpectedly higher than the affinity measured in solution by ITC (Supplementary Figure S2E). (G) BioLayer Interferometry experiments revealing that P- and A-motifs bind with different affinities to DMC1 (see also Supplementary Figures S6A-B). Using a DMC1 concentration of 5 μM, an interaction was detected only in the case of PhePP, RAD51AP1, BRC4 and FIGNL1-P. Apparent affinities for RAD51AP1, BRC4 and FIGNL1 were about 1.6, 4.3 and 16.2 μM, respectively. No binding was detected for PhePP_mut, TR2 and BRC1.

Figure 5.

PhePP binds DMC1 oligomers. (A) Superposition of the crystal structure of a dimer of DMC1–PhePP onto the crystal structure of RAD51 ATPase domain bound to BRC4. The two DMC1 protomers are in wheat and green, as in Figure 2C. Only the best resolved PhePP (F2s6) peptide is shown (in red). RAD51 is not displayed and the BRC4 peptide is in blue. In DMC1 green protomer, the side chains of F85 at the interface with the wheat protomer and F89 at the interface with PhePP are shown in sticks. In PhePP, the side chains of two residues important for binding to DMC1, i.e. F2406 and K2413 (interacting with F89), are also represented. In BRC4, only the side chain of F1524, essential for binding RAD51, is displayed. (B) GST pull-down assay with GST-tagged BRCA2 peptides and either wild type DMC1 or its F85A variant (replicated in Supplementary Figure S7A). (C) Docking of three PhePP peptides onto the cryo-EM structure of the DMC1–ssDNA filament (PDB 7C9C). Each peptide was positioned by superimposing the DMC1 filament monomer (either A, B or C) onto the crystal structure of DMC1 (81–339) bound to PhePP, as shown in Supplementary Figure S8. DMC1 (81–339) structures were subsequently hidden. The ssDNA binds on the opposite side of the filament, as compared to PhePP, it is thus not visible in this view. (D) Zoom-in from (C). The side chains of PhePP important for binding to DMC1 are displayed in sticks. The side chains of the DMC1 residues in front of the two lysines are displayed. The same side chains were identified in front of the two lysines in the crystal structure of DMC1–PhePP. (E) EMSA analysis of the filaments formed under Ca-ATP conditions by wild-type or substitution variant (F85A, F89A) DMC1 on a 90 nt ssDNA in the absence or presence of F2s3 or its F2406A variant. (F, G) BLI curves corresponding to the interaction between PhePP (F2s6) or different mutated PhePP peptides F2s6-F2406A, F2s6-K2404E, F2s6-K2413E, F2s6-K2411A-K2413A, F2s6-R2401A-K2404A and ssDNA-DMC1 filaments. On panel F, a schematic representation of the BLI experiment is displayed: the biotinylated ssDNA is coupled to a streptavidin sensor, DMC1 is assembled onto ssDNA, and different concentrations of F2s6 peptides are added. On panel G, the binding curves show that F2s6 binds to the ssDNA-DMC1 filaments with a micromolar affinity (K_d value obtained in the steady-state mode), whereas for its variants, no significant binding was detected in our conditions. A replicate is shown in Supplementary Figure S7D.

Figure 6.

PhePP stabilizes DMC1 filaments. All experiments were performed under Ca-ATP conditions. (A) Representative images of the SFM scans of the complexes formed by DMC1 on a DNA molecule corresponding to resected DSB end, in the presence of the PhePP (F2s3) peptide, its F2406A variant or storage buffer as negative control. (B) Quantification of the wild-type and F89A DMC1 filament lengths in the presence of BRCA2 F2s3 (PhePP wild-type and F2406A variant), as analyzed by SFM. Symbol shapes and colors correspond to independent experiments (n = 2–4). Red lines indicate mean and 95% confidence intervals. Statistically significant differences (p< 0.05) are indicated, as determined by ANOVA with Kruskal–Wallis test. (C–E) TEM statistical analysis of DMC1 filaments. (C_1–3) Representative views of DMC1–ssDNA (100 nt) filaments imaged using positive (C_1-2) or negative staining (C₃). (D_1–3) Same as (C) but in presence of 2 μM F2s3. (E_1–3) Same as (C) but in presence of 8 μM F2s6. (F) Statistical analysis of the filament lengths. (G) EMSA analysis of DMC1 filaments. Filaments were assembled in the same conditions as for TEM analysis in presence of various concentrations of F2s3 (1, 2 and 4 μM) and F2s3 F2406A mutant (4 μM) on the ssDNA (100 nt) (G₁), or on the dsDNA (200 bp) (G₂), in presence of various concentrations of F2s6 (2, 4, 8 and 15 μM) and F2s6 variants (8 μM) on the ssDNA (100 nt) (G₃). The effect of peptides was also tested in absence of DMC1 on ssDNA (G₄).