Molecular basis of FIGNL1 in dissociating RAD51 from DNA and chromatin

Abstract

Maintaining genome integrity is an essential and challenging process. RAD51 recombinase, the central player of several crucial processes in repairing and protecting genome integrity, forms filaments on DNA. RAD51 filaments are tightly regulated. One of these regulators is FIGNL1, that prevents persistent RAD51 foci post-damage and genotoxic chromatin association in cells. The cryogenic electron microscopy structure of FIGNL1 in complex with RAD51 reveals that the FIGNL1 forms a non-planar hexamer and RAD51 N-terminus is enclosed in the FIGNL1 hexamer pore. Mutations in pore loop or catalytic residues of FIGNL1 render it defective in filament disassembly and are lethal in mouse embryonic stem cells. Our study reveals a unique mechanism for removing RAD51 from DNA and provides the molecular basis for FIGNL1 in maintaining genome stability.

Introduction

RAD51 recombinase is a key protein involved in homologous recombination, the most faithful repair process for a double-stranded DNA break(1–3). RAD51, together with its meiosis specific homologue DMC1, also perform analogous functions in meiotic recombination when parental homologous chromosomes pair and recombine(4). The RAD51-ssDNA filament catalyzes the essential processes of homology search, strand invasion and heteroduplex formation(1, 3, 5). RAD51 needs to be removed from heteroduplex DNA before homology-directed DNA synthesis and repair/recombination can be carried out. Aberrant loading of RAD51 on ssDNA can lead to multiple strand invasions across chromosomes which in turn can lead to the formation of ultrafine bridges and chromosome instability, a hallmark of cancer(6, 7). Furthermore, RAD51 is shown to be essential for DNA replication, especially in dealing with replication stress, and to maintain a stable replication fork(8, 9). Significantly, over-expression and accumulation of RAD51 on chromosome has been associated with increased drug resistance in tumor cells and increased genome instability and apoptosis in normal cells(10). Consequently, RAD51 activity is tightly regulated. For example, the tumor suppressors BRCA2 and RAD51 paralog complex BCDX2 promote RAD51-ssDNA filament formation and stability while several DNA helicases and translocases such as RAD54, the RecQ family, HELQ, RTEL1, and FBH1 can disassemble RAD51-ssDNA or -dsDNA filaments(5, 11–16). FIGNL1 (fidgetin-like 1), an essential gene in mice belonging to the large AAA+ ATPase family, has recently been shown to prevent persistent RAD51 foci formation, prevent replication fork instability and suppress ultra-fine chromosome bridges(17–20). Conditional knockout of FIGNL1 in mouse spermatocytes results in massive overloading of RAD51 and DMC1 in these cells(17, 19). Recently, FIRRM/FLIP has been shown to form a stable complex with FIGNL1 and together they function in several DNA repair pathways including HR and interstrand DNA cross-link (ICL) repair as well as replication fork protection; and FIGNL1-FIRRM/FLIP can disassemble RAD51 filaments(19, 21–25). FIGNL1 has thus been firmly established as an important player in genome maintenance. However, the molecular mechanism of RAD51 modulation by FIGNL1 is currently unknown.

Results

FIGNL1 ATPase is essential for RAD51 filament disassembly

Silencing or deleting FIGNL1 has been shown to be embryonic lethal in mice(17). Indeed Fignl1^−/− mice die during embryogenesis (Fig. S1A) and Fignl1^−/− mouse embryonic stem cells (mESCs) are inviable (Fig. 1A, Fig. S1B–D). The importance of the enzymatic activity of FIGNL1 has been contradictory. Earlier work indicated that ATPase activity was important for RAD51 foci formation and efficient HR(18) but a later study suggested that ATPase activity was not necessary for RAD51 filament disassembly(26). In order to understand if and how enzymatic activity of FIGNL1 affects its functionalities, we developed a system to express various FIGNL1 mutants from the Rosa26 locus in Fignl1^+/− mESCs to ask whether the second endogenous allele could be disrupted (Fig. S1). While Fignl1^−/− cells were viable if they expressed wild-type FIGNL1 from the locus, they were not viable if they expressed the mutant, K456A (K447 in human FIGNL1), in the nucleotide-binding Walker A motif (KA, Fig. 1A, Fig. S1E), indicating that nucleotide binding of FIGNL1 is essential.

Figure 1. FIGNL1 ATPase activity is required for its functionality in cells.

(A) FIGNL1 is required for the viability of mESCs. Fignl1^+/− cells were targeted at the Rosa26 locus with expression cassettes for wild-type Fignl1 (rWT) or mutants K456A in the ATPase Walker A motif (rKA) or D411C for chemical inhibition (rDC) and then selected for Hyg gene expression from the Rosa26 promoter. After confirmation of correct targeting, the second Fignl1 allele was subjected to CRISPR-Cas9 editing using sgRNAs (c+d) to delete the entire Fignl1 coding region. Genomic DNA was screened by PCR using primers −396 and 409 for the undeleted allele. Successful deletion is indicated by a blue asterisk. The number of Fignl1^−/− colonies is indicated, along with the total number of colonies that were screened. (B) Chemical inhibition of the FIGNL1 ATPase is lethal. Covalent modification of FIGNL1 at D411C in the ATPase domain by ASPIR-1 impairs colony formation. (C) RAD51 accumulates in the chromatin fraction in Fignl1^−/−; rDC but not Fignl1^−/−; rWT cells upon ASPIR-1 treatment. Cells are treated with 0.25 µg/ml ASPIR-1 or DMSO for 24 hr and then subjected to chromatin and cytoplasmic fractionation followed by western blot analysis with antibodies to the indicated proteins. (D) Rad51 heterozygosity rescues survival of the Fignl1 mutant for colony formation in the presence of ASPIR-1 (0.25 µg/ml) for 7 days. Three independent clones were tested (80, 83, 93). (E) Rad51 heterozygosity reduces RAD51 chromatin association in the presence of ASPIR-1 to the level found in untreated Rad51 wild-type cells. Three independent clones were treated with 0.25 µg/ml ASPIR-1 or DMSO for 24 hr and then subjected to chromatin and nuclear fractionation followed by western blot analysis with antibodies to the indicated proteins.

To probe the effects of ATPase activity in cells in a controlled fashion, we used a chemical genetic approach to inhibit the ATPase activity of FIGNL1 by expressing an ATP-analogue sensitive mutant with a cysteine in the active site(27) (Fig. S2A). This modified version of FIGNL1 (D411C, D402 in human FIGNL1) retains ATPase activity, but when exposed to the compound ASPIR-1, a covalent bond is formed at the active site to abrogate ATPase activity(27). We found that while Fignl1^−/− cells ectopically expressing D411C at Rosa26 were viable, treatment of these cells with ASPIR-1 was toxic (Fig. 1A–B, Fig. S2B). Further, ASPIR-1 treatment of D411C-expressing cells resulted in markedly increased association of RAD51 with chromatin (Fig. 1C, Fig. S2C–D), which is known to increase chromosome aberrations/aneuploidy(28). Strikingly, the lethality caused by Fignl1 mutation was rescued by Rad51 heterozygous deletion (Fig. 1D, Fig. S1F). Associated with the rescue, cells displayed reduced RAD51 chromatin accumulation (Fig. 1E), consistent with recent data showing that RAD51 inhibition could rescue cellular defects due to Fignl1 knockout(20). Together, these data support the notion that FIGNL1 ATPase activity is required for its functionality in the prevention of aberrant RAD51-chromatin association and genome instability.

To understand how FIGNL1’s ATPase activity is linked to modulating RAD51’s association with chromatin, we set out to investigate the molecular mechanisms of FIGNL1-mediated RAD51 dissociation from DNA and filament disassembly in vitro. Previous studies have identified three functional domains in FIGNL1: a N-terminal domain containing a largely unstructured region responsible for localisation to DNA damage sites and for interactions with other proteins, followed by the RAD51-binding FRBD (FIGNL1 RAD51-Binding Domain) and AAA+ ATPase domains (Fig. S3A)(18, 19). The N-terminal domain interacts with FIRRM/FLIP (21), and the interactions are important for the mutual stability of FIGNL1 and FIRRM/FLIP(19, 23–25). To ensure protein stability in the absence of interacting partners while maintaining its activities for RAD51 modulation, we purified a N-terminal truncated fragment of human FIGNL1 (residues 287–674), equivalent to those used in earlier studies(18, 26), that exhibits ATPase activity and RAD51 binding, which we term FIGNL1∆N (Fig. S3). As expected and consistent with previous studies(17, 22, 26, 29), FIGNL1∆N can dismantle RAD51 from both ssDNA and dsDNA (Fig. 2, Figs. S3–S4).

Figure 2. Disassembly of RAD51 filaments is dependent on FIGNL1 ATPase activity.

(A) Representative negative-stain electron micrographs of RAD51 filaments on both single-stranded and double-stranded DNA in the absence or present of FIGNL1∆N or FIGNL1DN(E501Q), a mutant in the Walker B motif. Red arrows indicate some of the filaments. Scale bars = 100 nm. (B) Quantification of filaments observed in (A) indicates a decrease in the number of filaments per micrograph upon incubation with FIGNL1∆N while FIGNL1DN(E501Q) has little effects. n = 18–20 micrographs per condition (C) Image of gel, showing nuclease protection of dsDNA by RAD51 in the presence of increasing concentrations of FIGNL1∆N or E501Q mutant. (D) Quantification of nuclease protection assays. n = 3, each data point represents the mean ± s.d. WT FIGNL1∆N data quantification from independent repeats and 0.5µM and 0.9µM data points shown in C. Filament disruption assays shown in A-D were carried out using 60nt ssDNA and 60bp dsDNA.

To visualize filament disassembly and to compare the effects of various FIGNL1∆N mutants in order to probe the requirement of ATPase activities, we imaged and quantified RAD51 filaments using negative stain electron microscopy (NS-EM) (Fig. 2A–B, Fig. S4). To assess the activity in solution, we used a nuclease protection assay in which DNA coated by RAD51 is challenged to nuclease treatment(30). A decrease in protection indicates dissociation or remodelling of RAD51 from DNA. In the nuclease protection assays, adding FIGNL1∆N resulted in a decrease in RAD51-mediated protection in a FIGNL1∆N-concentration dependent fashion (Fig. 2C–D, Fig. S3C). Together with the NS-EM results (Fig. 2A–B), these data confirm that FIGNL1∆N disassembles RAD51-DNA filaments. To assess the role of FIGNL1 ATPase activity, we introduced a mutation in the catalytic Walker B motif (E501Q), which is defective in ATP hydrolysis (Fig. S3B–D). The E501Q mutant does not affect RAD51 binding (Fig. S3E) but is severely defective in disassembling filaments, as judged by NS-EM and nuclease protection assays (Fig. 2, Fig. S3C, Fig. S4), suggesting that ATP hydrolysis is required for its activities. Indeed, this mutant resulted in lethality in mESCs (Fig. S5). Not surprisingly, the Walker A mutant (K447A) is also defective in ATPase activity and filament disassembly (Fig. S3B–D).

Taken together, our cellular and in vitro data firmly establish that both nucleotide binding and hydrolysis are required for FIGNL1 to dissociate RAD51 from DNA, filament disassembly and are essential for cell viability.

Architecture of the FIGNL1-RAD51 complex

Other known ATPases that disassemble RAD51 filaments (collectively termed anti-recombinases) function as DNA translocases or helicases via their ATPase domains(14, 31), whereas the AAA+ domain of FIGNL1∆N does not bind DNA (Fig. S3F), suggesting that FIGNL1 acts on RAD51 instead of on DNA. To gain insights into the molecular mechanism of FIGNL1-dependent regulation of RAD51, we utilised FIGNL1∆N^E501Q, which forms a stable complex with RAD51 but is defective in filament disassembly, to obtain a structure of FIGNL1 in complex with RAD51, which likely represents an initial engaged state of FIGNL1. Using cryogenic electron microscopy (cryoEM), we determined the structure of the FIGNL1∆N^E501Q -RAD51 complex in the presence of ATP and Mg²⁺ (Fig. S6). Single particle analysis produced a 3D reconstruction that revealed a hexameric ring shape, which we attributed to FIGNL1, a AAA+ protein that forms hexamers in the presence of nucleotide (ATPγs) (Fig. S6E)(24), and additional density above the ring, tethered to the hexameric density, to be RAD51 (Fig. 3A). The resolution of this complex is limited, with a global resolution of ~ 8 Å, likely due to the innate flexibility of the FIGNL1-RAD51 complex (Fig. S6). Indeed the identified RAD51-binding FRBD is predicted to be in a largely unstructured region (Fig. S3A). Using AlphaFold2(32, 33), we generated a model of the FIGNL1 FRBD in complex with a RAD51 trimer, which resembles a filament (Fig. S7A). This model predicts that the AAA+ domain is tethered to the FRBD-RAD51 sub-complex via a flexible linker (Fig. S7A), consistent with our cryoEM reconstruction (Fig. 3A). In this model, the FRBD can bind to a RAD51 trimer using two sites, separated by ~ 40 amino acids (Fig. S7A). The first site (FKTA), which has been previously identified(18), is predicted to bind to RAD51 analogous to that of the BRCA2 BRC4 motif (FxxA), that binds at the RAD51 protomer interface in the filament(34). A second site (FVPP), which is also present in BRCA2 and RAD51AP1(35, 36), binds the adjacent RAD51 protomer via a different location on RAD51 and does not overlap with BRC4 binding site (Fig. S7A). Mutating either of the two sites is only moderately defective in our nuclease protection assays while mutating both sites severely reduced its ability in filament disassembly (Fig. S7B). Murine ESCs are viable even when both sites are mutated (Fig. S7C), suggesting there are other interaction sites between RAD51 and FIGNL1, either directly or via other interacting partners, to partially compensate for the lost interaction site. A recent study has shown that mutating site 1 (FxxA to ExxE) failed to suppress ultra-fine bridge (UFB) formation in U2OS cells(20), suggesting a crucial role of this interaction site in UFB suppression.

Figure 3. CryoEM structures of the FIGNL1-RAD51 complex and FIGNL1 AAA+ hexamer.

(A) Top and side views of the cryoEM map and model of the FIGNL1DN^E501Q-RAD51 complex in the presence of ATP.Mg²⁺. (B) Top and side views of the cryoEM map (2.9Å) of the FIGNL1 AAA+ hexamer in the presence of ATP.Mg²⁺. (C) Top and side views of the atomic model of the FIGNL1 AAA+ hexamer modelled from the map shown in B.

To improve the resolution of the reconstruction, we focused on the FIGNL1∆N hexamer for further processing, which was resolved to 2.9Å (Fig. 3B, Fig. S8). The six FIGNL1 AAA+ domains form a spiral hexamer (Fig. 3B), with chain F at the base of the spiral and chain A at the top. The quality of the electron density map is sufficient to allow a structural model of the AAA+ domain of FIGNL1 to be built (Fig. 3C, Table 1, Fig. S8). We identify clear density for Mg²⁺-ATP, bound in the nucleotide binding pocket of chains A, B, C, D and E, in between adjacent protomers (Fig. 3C, Fig. S8E). The electron density of chain F is of poorer quality, presumably due to increased flexibility of this subunit (Fig. S8B, S8E). The structural model allowed us to map mutations found in cancer samples and variants of unknown significance in patients with genetic disorders (Fig. S9A–B)(37, 38). Interestingly, using AlphaMissense (Fig. S9C, Table 2) (39), E501Q has a high pathogenicity score (0.97) and in human disease mutation database ClinVar, it occurred in 7 samples, consistent with defects we observe in vitro and in cells. Several mutations are located at the protomer-protomer interface, and these mutations could interfere with proper assembly of the functional hexamer (Fig. S9B–C). Further investigations are required to confirm their potential consequences.

Table 1.

Cryo-EM data collection, refinement and validation statistics

	FIGNL1-RAD51 complex (EMDB-18946) (PDB 8R64)
Data collection and processing
Magnification	81000 X
Voltage (kV)	300 kV
Electron exposure (e–/Å2)	50
Defocus range (μm)	−0.7 to −2.1
Pixel size (Å)	1.1
Symmetry imposed	C1
Initial particle images (no.)	1.8 million
Final particle images (no.)	57,484
Map resolution (Å)	3.2
FSC threshold	0.143
Map resolution range (Å)	2.8 to 5.6
Refinement
Initial model used (PDB code)	3D8B, Alphafold2
Model resolution (Å)	3.4 / 3.1
FSC threshold	0.5 / 0.143
Model resolution range (Å)	2.9 to 3.7
Map sharpening B factor (Å2)	−65
Model composition
Non-hydrogen atoms	13636
Protein residues	1774
Ligands	ATP: 5
	ADP: 1
	Mg2+: 6
B factors (Å2)
Protein	69.2
Ligand	60.1
R.m.s. deviations
Bond lengths (Å)	0.003
Bond angles (°)	0.53
Validation
MolProbity score	1.33
Clashscore	5.38
Poor rotamers (%)	0
Ramachandran plot
Favored (%)	97.8
Allowed (%)	2.2
Disallowed (%)	0

FIGNL1 coordinates the RAD51 N-terminal peptide in its hexameric pore using pore loops

Unexpectedly, we found additional density consistent with a polypeptide in the central pore of the FIGNL1 AAA+ hexamer (Fig. 4A). The polypeptide is tightly coordinated by two highly conserved pore loops of FIGNL1 which form a helical staircase (Fig. 4B–D). Many AAA+ proteins utilise pore loops to interact with their respective substrates, either protein polypeptides or nucleic acids(40–43). The N-terminal pore loop (residues 470 to 476, pore loop 1, PL1) contains a lysine-tryptophan dipeptide (KW) at its tip which intercalates the side chains of the peptide (Fig. 4D). The C-terminal pore loop (residues 506–514, pore loop 2, PL2) is less tightly associated with the peptide, but habours histidine 514 that tracks along the peptide backbone (Fig. 4C–D). The extra density in the pore suggests that in addition to the two identified FRBD sites, there are additional interaction sites between RAD51 and FIGNL1. We could fit residues 2–13 of the RAD51 N-terminus, which belong to an unstructured region not observed in previous crystal and cryoEM structures (Fig. S10A). The peptide is well resolved, with a local resolution of 2.7 Å (Fig. S10A–B). High Q-scores(44) not only indicate a good fit of these residues into the density, but the density has a high resolvability, providing further confidence in the residue assignment (Fig. S10A). In this structural model, the highly conserved pore loops encircle the physiochemically conserved hydrophobic/polar amphipathic residue pattern found in the N-terminus of RAD51 (Fig. 4D–E, Fig. S10B–E), via a close network of interactions, which may provide RAD51 sequence preference (Fig. 4D–E; Fig. S10D–E).

Figure 4. FIGNL1 coordinates the N-terminus of RAD51 through its central hexameric pore.

(A) Extra density observed in the central pore of the FIGNL1 AAA+ hexamer. (B) Sequence conservation plots of the pore loops of FIGNL1 in vertebrates (n=576). Residues 473, 474 and 514 are labelled. (C) The pore loops of FIGNL1 form two helical staircases enclosing the RAD51 N-terminus (D) The pore loop residues intercalate with the RAD51 N-terminal residues. (E) Conservation plot of the N-termini of RAD51 in 548 vertebrates. Numbering refers to the human RAD51 sequence. (F) Nuclease protection of ssDNA by RAD51 or RAD51 with N-terminal 20 a.a. deleted (RAD51DN), in the presence of increasing concentrations of FIGNL1DN (G) Quantification of ATPase activity of FIGNL1 alone or incubated with RAD51 or RAD51∆N. (H) Dose-response ATPase activity of FIGNL1∆N in the presence of increasing concentrations of the RAD51 N-terminal peptide or with the first 5 amino acids replaced by alanine (n = 4).

To probe the importance of the RAD51 N-terminus, which has not previously been shown to have specific functions in filament regulation, we investigated if FIGNL1 requires the RAD51 N-terminus for filament disassembly. RAD51 lacking the N-terminal 20 residues (RAD51∆N) can form filaments and protect DNA from nuclease digestion (Fig. S11A–C). Strikingly, RAD51∆N filaments could not be efficiently disrupted by FIGNL1∆N in the nuclease protection assay (Fig. 4F, Fig. S11D). This is not due to the lack of interactions between FIGNL1∆N and RAD51∆N, as the binding, although reduced, is still sufficiently high to ensure interactions under the experimental conditions (Fig. S11E). This is consistent with the FRBD being the main recruitment site between FIGNL1 and RAD51. Given that ATPase activities of some AAA+ proteins are stimulated by their respective substrates(41, 45), we tested whether RAD51 stimulates the ATPase activity of FIGNL1. The ATPase activity of FIGNL1∆N is enhanced in the presence of RAD51 (Fig. 4G) but this enhancement is abolished when the N-terminus of RAD51 is deleted (Fig. 4G). Adding FIGNL1∆N defective in ATPase activity (K447R or E501Q) does not stimulate total ATP hydrolysis, while adding ATPase defective RAD51 (K133R, RAD51-WA) still showed significant enhancement, suggesting the majority of stimulation is due to increased ATPase activity of FIGNL1 in the presence of the N-terminus of RAD51 (Fig. S12A–B). A synthetic peptide containing the complete unstructured 22 amino acid RAD51 N-terminus, or the N-terminal 13 or 18 amino acids, can stimulate FIGNL1 ATPase equivalently (Fig. 4H, Fig. S12C). This stimulation relies on the very N-terminal amphipathic region of RAD51 observed in our cryoEM structure as replacing the first 5 residues with alanine in this peptide diminishes the stimulation (Fig. 4H). Furthermore, FIGNL1 ATPase stimulation is specific to the RAD51 N-terminus, as high concentrations of a peptide of poly-glutamate, tubulin C-terminal tails(46), ssDNA, and dsDNA all fail to enhance its ATPase activities (Fig. S12C–D). Together these data reveal that FIGNL1∆N utilises the RAD51 N-terminal region to stimulate its ATPase activity, and that the RAD51 N-terminus is essential for its ability to disassemble RAD51 filaments.

Pore loop integrity is required for filament disassembly and mESC viability

To corroborate the structural data and to further investigate the importance of the FIGNL1 pore loops, we substituted the KW of pore loop 1 to a glutamic acid and alanine respectively (K473E,W474A, herein referred to as the PL mutant) (Fig. S13A). This mutant retains near wild-type ATPase activity and RAD51 binding (Fig. S13B–C). This mutant ATPase activity can be stimulated by the RAD51 N-terminus, but to a lesser extent compared to wild-type (Fig. S13D). Critically, this mutant is severely defective in disassembling RAD51 filaments on both dsDNA and ssDNA as can be visualised using NS-EM (Fig. 5A–B) and nuclease protection assays (Fig. 5C–D). The equivalent mutation in mESCs is incompatible with cell viability (Fig. 5E). Together these data support a crucial role of the pore loops for FIGNL1 activity.

Figure 5. Mutation of pore loop 1 confers loss of RAD51 filament disassembly and cell lethality.

(A) Representative micrographs of RAD51-DNA filaments treated by FIGNL1∆N bearing mutations in pore loop 1 (PL mutant) with ss or ds-DNA. Scale bars = 100 nm (B) Quantification of experiments shown in (A), highlighting the loss of filament disassembly by FIGNL1∆N bearing PL mutation. n = 20, data are shown as mean ± s.d. WT FIGNL1∆N data is for comparison and is as shown in Fig. 2. (C) Nuclease protection of dsDNA coated by RAD51 upon treatment with wildtype or PL mutant of FIGNL1∆N. (D) Quantification of RAD51 + FIGNL1∆N^PL experiments shown in c, n = 3, data is shown mean ± s.d. WT FIGNL1∆N data quantification from independent repeats shown in Fig. 3 and 0.5µM and 0.9µM data points shown in C. (E) FIGNL1 pore loop mutant K483E/W484A (PL) is not compatible with cell survival. An expression cassette for the mutant (rPL^Myc) was targeted to Rosa26 locus as in Fig. 1A. No Fignl1^−/− colonies were obtained after CRISPR-Cas9 gene editing in Fignl1^+/− cells with gRNAs c and d as in Fig. 1A. (F) A proposed model of RAD51 filament disassembly by FIGNL1. FIGNL1 is recruited to the RAD51 filament via its FRBD domain and forms a hexamer, enclosing the N-terminus of RAD51, which stimulates the ATPase activity of FIGNL1 and promotes translocation of the N-terminus in the hexamer pore, leading to the unfolding/removing of the RAD51 from the filament, promoting disassembly. (G) SDS-PAGE gels showing RAD51 degradation by proteases in the presence of FIGNL1DN and its mutants with ATP. (H) quantification of RAD51 band intensity over time.

Recent studies have shown FIGNL1 is important in meiotic DMC1-focus formation and resolution(17, 19). We thus set out to test the ability of FIGNL1∆N in disassembling DMC1 filaments in vitro. DMC1 does not efficiently protect dsDNA from nuclease(47); we therefore used ssDNA in our assays. FIGNL1∆N could also reduce nuclease protection by DMC1 (Fig. S14A–B), indicating disruption of nucleoprotein complexes, and both DMC1 and DMC1 N-terminal peptide stimulate ATPase activity of FIGNL1∆N (Fig. S14C–D). However the DMC1 N-terminal peptide is ~3-fold less effective at stimulating FIGNL1 ATPase activity compared to that of RAD51 under the same experimental conditions (Fig. S14D). DMC1 and RAD51 N-terminal regions share some degree of conservation (Fig. S14E), suggesting that FIGNL1 substrate selection is somewhat plastic. Importantly, the pore loop mutant is severely defective in disassembling DMC1 filaments (Fig. S14A). Together these data suggest that FIGNL1 can also disrupt DMC1-DNA association using a similar mechanism to that of RAD51, despite being less efficient in our assays. The precise mechanism and sequence specificity of FIGNL1 requires further investigation.

Discussions

FIGNL1 acts as a peptide translocase to disociate RAD51 from DNA

Our data presented here suggest a molecular mechanism of how FIGNL1 acts on RAD51 during dissociation from DNA and filament disassembly. We showed that FIGNL1 AAA+ domains do not bind DNA and DNA does not stimulate its ATPase activities (Fig. S3F, S12C), and thus are unlikely to act as an ATP-dependent DNA translocase. Instead, we show that the ATPase activity is stimulated by RAD51, specifically via the very N-terminus, which we observe is enclosed by the FIGNL1 AAA+ hexamer. Therefore FIGNL1 likely acts directly on RAD51 via interacting with RAD51 N-terminus. The FRBD is responsible for the direct recruitment of FIGNL1 to RAD51. Site 1 (FxxA) binds to RAD51 at a location that overlaps with the protomer interface in a RAD51 filament, suggesting that site 1 would prefer to bind to the 3’-end of a RAD51 filament or a gap in the filament. Site 2 (FVPP), however, could bind to a RAD51 molecule inside the filament, suggesting that FIGNL1 can be recruited to anywhere along the filament. Mutating one of the two sites has mild defects in our nuclease protection assays (Fig. S7B). Interestingly, in vivo, mESCs with both sites mutated are still viable while those with pore loop mutations are not (Fig. S6C), confirming that FRBD mainly acts in recruitment and that among the three interaction sites between FIGNL1 and RAD51, pore loops enclosing the RAD51 N-terminus are the most important. It is likely that in vivo, FIGNL1 could also be recruited to RAD51 via other interacting partners, such as FIRRM/FLIP, which is shown to bind to RAD51(25). Our data thus support a model that FIGNL1 is recruited to the filament through its FRBD (and other interacting partners), on the side of the filament. FIGNL1 hexamer assembles around and encloses the RAD51 N-terminus via an intricate network of interactions (Fig. 5F). Indeed 3D variability analysis of the FIGNL1-RAD51 structure reveals a mixture of tetrameric, pentameric, and hexameric configurations of the FIGNL1 AAA+ hexamer on the RAD51 N-terminus (Fig. S15). This model is also consistent with the observed minimal level of ATP hydrolysis by FIGNL1 at submicromolar concentrations, which is stimulated by RAD51 N-terminus since full ATPase activity requires hexamer formation (Fig. 4H). A similar mechanism has been proposed for Katanin, which oligomerizes as a hexamer on microtubules(48).

Thus far all the known AAA+ proteins that enclose substrate peptides in their hexamer central pores utilise ATP hydrolysis to translocate or remodel the substrate(49). Given the similarities of the structure of FIGNL1∆N in complex with RAD51 presented here with that of Vps4, Spastin, Katanin and TRIP13 in complex with their substrate peptides(40, 41, 50–52) (Fig. S16), we propose that FIGNL1 acts as a peptide translocase that leads to RAD51 remodelling. Using limited proteolysis, we showed that RAD51 was sensitive to Proteinase K digestion in the presence of FIGNL1∆N and ATP, but markedly less so in the presence of PL or EQ mutants (Fig. 5G–H). Proteolytic patterns of RAD51 also differ between FIGNL1 mutants, suggesting that RAD51 conformation is altered by FIGNL1, resulting in the differential proteolytic sensitivities (Fig. S17). Indeed, FIGNL1-RAD51 N-terminal interactions strongly resemble that of TRIP13 in complex with the with MAD2 N-terminus (Fig. S16). TRIP13 uses its ATPase activity to translocate the unstructured N-terminal region of MAD2 via its pore loops, partially unfolding MAD2 and converting an active closed conformation to an inactivate open conformation (53, 54).

Similarly to other AAA+ translocases, FIGNL1 translocates by ATP hydrolysis along the RAD51 N-terminus via the pore loops(55–57), leading to pulling or partial unfolding/remodelling of the RAD51 molecule within the filament, destabilising and disassembling the filament as the N-terminal domain links adjacent ATPase domains of RAD51 in the filaments(58). Indeed the FIGNL1-PL mutant is defective in filament disassembly, despite stimulated ATPase activities, and does not increase RAD51 sensitivity to proteolysis, supporting the notion that the PL residues are important for its translocating activities analogously to other AAA+ translocases(41, 52). Our model suggests that in addition to acting on RAD51 nucleoprotein filaments, FIGNL1 could also act on RAD51 bound to other proteins including nucleosomes, as shown recently(59) as the action does not require the acccess of FIGNL1 to DNA.

FIGNL1’s unique mechanism in dissociating RAD51 from chromatin defines its critical roles in cell viability

The unique mechanism of FIGNL1 in disassembling RAD51 from DNA filaments, acting anywhere along the RAD51 filaments, the potential ability of dissociating RAD51 from other binding partners, including nucleosome-bound RAD51, and the ability to remodel RAD51, suggest FIGNL1 act as a general RAD51 regulator. Our DMC1 data suggests a conserved mechanism in FIGNL1-mediated recombinase filament disassembly. This is distinct from other anti-recombinases which act on DNA from the end of filaments, and may explain the critical role of FIGNL1 in cell viability that cannot be substituted by other anti-recombinases. Excess RAD51 accumulation on ssDNA is predicted to cause abberant HR and interfere with ongoing replication. Furthermore accumulation of RAD51 on dsDNA and chromatin could interfere with other DNA transactions such as chromatin remodelling and transcription. In normal cells, FIGNL1 and other anti-recombinases prevent RAD51 accumulation on chromatin and abberant loading on DNA. Absence of FIGNL1 will result in RAD51 chromatin accumulation, some of which cannot be removed by other recombinases, leading to cell death. Indeed, heterozygous Rad51 deletion results in reduced RAD51 chromatin association, and rescues viability associated with Fignl1 mutation. Our work here thus explains the critical roles of FIGNL1 in cell viability and reveals the molecular mechanism of RAD51 filament disassembly and the maintainence of genome stability.

Funding Statement

This work was funded by Breast Cancer Now and a Wellcome Trust Investigator Award (210658/Z/18/Z) and a Wellcome Trust Discovery Award (227769/Z/23/Z) to X.Z.; and Starr Cancer Consortium award I13-0055 and NIH R35 CA253174 to M.J.

Data Availability

All the data have been deposited in wwPDB with access codes EMDB-18946, PDB code 8R64