Abstract

   Interstrand DNA cross-links (ICLs) represent complex lesions that
   compromise genomic stability. Several pathways have been involved in
   ICL repair, but the extent of factors involved in the resolution of
   ICL-induced DNA double-strand breaks (DSBs) remains poorly defined.
   Using CRISPR-based genomics, we identified FIGNL1 interacting regulator
   of recombination and mitosis (FIRRM) as a sensitizer of the
   ICL-inducing agent mafosfamide. Mechanistically, we showed that FIRRM,
   like its interactor Fidgetin like 1 (FIGNL1), contributes to the
   resolution of RAD51 foci at ICL-induced DSBs. While the stability of
   FIGNL1 and FIRRM is interdependent, expression of a mutant of FIRRM
   (∆WCF), which stabilizes the protein in the absence of FIGNL1, allows
   the resolution of RAD51 foci and cell survival, suggesting that FIRRM
   has FIGNL1-independent function during DNA repair. In line with this
   model, FIRRM binds preferentially single-stranded DNA in vitro, raising
   the possibility that it directly contributes to RAD51 disassembly by
   interacting with DNA. Together, our findings establish FIRRM as a
   promoting factor of ICL repair.
     __________________________________________________________________

   FIRRM plays a role in the maintenance of genome stability.

INTRODUCTION

   Interstrand DNA cross-links (ICLs) are highly cytotoxic DNA lesions
   that produce physical obstacles for vital biological processes,
   including DNA replication, transcription, and recombination. ICLs can
   emerge from endogenous reactive aldehydes but more often arise from
   exposure to either naturally occurring compounds [e.g., mitomycin C
   (MMC)] or chemically synthesized drugs, such as cisplatin or
   cyclophosphamide [reviewed in ([60]1)]. These ICL-inducing drugs are
   widely used as chemotherapeutic agents in the treatment of both solid
   tumors and blood cancers ([61]2). Aside from compromising essential
   biological processes, ICLs, if left unrepaired, result in various
   genomic abnormalities, ranging from point mutations to chromosome
   breakage or missegregation and mitotic catastrophe [reviewed in
   ([62]3)].

   Each type of ICL-inducing agents leads to different types of intra- and
   interstrand cross-links that distort the double helix to different
   levels ([63]1). To detect, signal, and repair these lesions, cells rely
   on a complex response that typically gets triggered when the
   replication machinery encounters the ICL and involves the precise and
   coordinated effort of multiples pathways, including the Fanconi anemia
   (FA), nucleotide excision repair (NER), and the homologous
   recombination (HR) repair pathways, alongside the ATR serine/threonine
   kinase (ATR) checkpoint signaling pathways ([64]1). ICLs are initially
   recognized by the FA complementation group M (FANCM) complex (FANCM,
   FAAP24, MHF1, and MHF2) that serves as a platform for the docking of
   seven additional FANC proteins (FANCA-C, FANCE-G, and FANCL) as well as
   two additional associated proteins (FAAP20 and FAAP100) that altogether
   form the FA core complex. Along with FANCT (UBE2T), the E3-ubiquitin
   ligase FA core complex promotes the monoubiquitylation of the effector
   FANCI/FANCD2 (ID2) heterodimer, a critical step in the processing and
   subsequent repair of the ICLs. Once assembled and activated at the
   lesion, the ID2 heterodimer recruits structure-specific nucleases that
   promote ICL unhooking and acts as a molecular hub for the recruitment
   of translesion synthesis (TLS) DNA polymerases, which bypass the ICL
   adduct, and the subsequent recruitment of HR factors involved in the
   repair of the double-strand breaks (DSBs) generated during this process
   ([65]1). The past few years witnessed the discovery of an important
   decision point during the detection and early processing of ICL, named
   the ICL repair pathway choice. This decision is made between
   replication-independent and replication-dependent pathways and appears
   to be dictated by the structure and the location of the ICL as well as
   the phase of the cell cycle in which it is detected ([66]1). Recent
   dissection of the replication-dependent ICL repair pathway has
   identified a series of recently identified players such as TRAF
   interacting protein (TRAIP) ([67]4, [68]5), nei like DNA glycosylase 3
   (NEIL3) ([69]6), and suppressor of cancer cell invasion (SCAI)-REV3
   like, DNA directed polymerase zeta catalytic subunit (REV3) ([70]7,
   [71]8) that act before the TLS and HR pathways.

   Aside from a better understanding of ICL repair pathway choice, several
   reports have highlighted the complexity of the final stages of ICL
   repair, such as the resolution of RAD51-mediated strand invasion. For
   instance, the AAA^+ adenosine triphosphatase (ATPase) FIGNL1 has been
   shown to bind to RAD51 and promote its unloading from DNA ([72]9), a
   function that is limited by the SWSAP1-SWS1-SPIDR complex ([73]9,
   [74]10). More recently, the helicase HELQ ([75]11) and the
   HROB-MCM8-MCM9 ([76]12) complex have been shown to contribute to
   parallel pathways that promote DNA repair synthesis following D-loop
   formation. However, the spatiotemporal regulation of these functions is
   still unknown.

   Although the first model of ICL repair was drawn two decades ago, a
   complete picture of the factors involved in regulating this repair
   pathway remains largely unclear. Thus, we performed a series of
   CRISPR-based genome-wide screens with the ICL-inducing agent
   mafosfamide (MAF), and we identified FIGNL1 interacting regulator of
   recombination and mitosis (FIRRM) as a player in the response to ICLs.
   In-depth characterization delineated its contribution to regulating the
   later steps of HR pathway. Proximal mapping of FIRRM proximity
   interactors coupled to structure-function analysis identified the AAA^+
   ATPase Fidgetin like 1 (FIGNL1) as a partner of FIRRM, an interaction
   that is critical for the stability of both proteins. Depletion of FIRRM
   resulted in the persistence of RAD51 at DSBs and concomitant cell
   death, two phenotypes that are rescued by a mutant (FIRRM ∆WCF) that
   stabilizes the protein in absence of FIGNL1. Biochemical
   characterization of FIRRM revealed that it directly binds to
   single-stranded DNA (ssDNA) in vitro, raising the possibility that this
   interaction contributes to RAD51 displacement from DNA at DNA damage
   sites. Our model thus defines FIRRM as an integral factor in the
   maintenance of genome stability by promoting RAD51 disassembly at
   ICL-induced DNA damage.

RESULTS

CRISPR screening identifies FIRRM as a modulator of ICL repair in BL cells

   To better understand the factors influencing ICL repair, we undertook
   CRISPR-Cas9 dropout screens in Namalwa and Raji human Burkitt lymphoma
   (BL) cell lines, using the metabolized version of the ICL agent,
   cyclophosphamide as a selective drug (MAF). First, we generated stable
   Namalwa and Raji cell lines expressing Cas9 using lentiviral
   transduction and confirmed genome editing efficiency by targeting the
   FAM83G gene as previously described ([77]13). Next, CRISPR-based
   genome-wide screening was initiated by transducing the Toronto
   knock-out version 1 (TKO v1) single guide RNA (sgRNA) library in both
   human BL Cas9 cell lines ([78]14, [79]15), and transduced cells were
   amplified under puromycin (2 μg/ml) selection for 7 days ([80]Fig. 1A).
   Both Namalwa and Raji transduced cell lines were subsequently treated
   with either preoptimized doses of MAF [25% inhibition concentration
   (IC[25])] or with dimethyl sulfoxide (DMSO) as a vehicle for 14
   doubling times (14 days for both cell lines) before being processed for
   next-generation sequencing (NGS). The DrugZ algorithm was used to
   determine the relative abundance of each sgRNA and identify genes whose
   knockout sensitizes cells to MAF (NormZ score < −2.5, P < 0.01)
   ([81]Fig. 1B, fig. S1A, and table S1) ([82]16). In both screens,
   pathway enrichment analysis identified ICL repair and HR as two
   biological pathways statistically enriched among our identified genes
   of interest (fig. S1B). As expected, most factors of the FA core
   complex and factors implicated in HR repair emerged as top sensitizers
   in both cell lines ([83]Fig. 1B, fig. S1A, and table S1), validating
   our CRISPR-based screening approach to identify genes influencing ICL
   repair. Notably, the recently characterized FIGNL1 ([84]9, [85]17),
   MCM8-MCM9 ([86]12), and SCAI ([87]7) scored significantly in both of
   our screens ([88]Fig. 1B, fig. S1A, and table S1).

Fig. 1. CRISPR-based genome-wide screening identifies FIRRM as a modulator of
MAF sensitivity.

   [89]Fig. 1.
   [90]Open in a new tab

   (A) Schematic representation of the CRISPR-based screening pipeline
   used in BL cells and the screen analysis using the DrugZ algorithm. (B)
   Horizontal scatter plot of DrugZ-generated ranking of the Namalwa MAF
   CRISPR screen. NormZ values are plotted on the y axis, and gene names
   are plotted on the x axis. DDR, DNA damage response; HJ, Holliday
   junction. (C) Venn diagram displaying common sensitizing genes from
   Namalwa and Raji screens (for genes with NormZ ≤ 2). (D) Gene ontology
   biological processes (GO BP) diagram obtained from common sensitizing
   genes from Namalwa and Raji screens. GO term enrichments are ranked by
   statistical significance (P value). The size of the circles indicates
   the number of the 67 common gene hits within a pathway. (E) Network
   analysis displaying protein physical interactions using Cytoscape and
   the GeneMANIA package. The FIRRM-FIGNL1 complex is represented in the
   vicinity of FA and HR repair networks. (F) Schematic representing the
   pipeline used in the two-color competition assay. RPE1-hTERT p53^−/−
   cells were transduced with a sgRNA targeting either the gene of
   interest (GOI) coupled with BFP or LacZ gene coupled with mCherry. BFP-
   and mCherry-expressing cells were mixed at a 1:1 ratio at D[0] and
   treated with IC[25] cisplatin (Cis.). BFP-positive over
   mCherry-positive cells ratio was followed over D[20]. (G) Competitive
   growth results obtained in RPE1-hTERT p53^−/− cells targeted with sgRNA
   against FANCA, MRE11, or FIRRM. Cells were either treated with DMSO as
   a vehicle or cisplatin at its IC[25] (2.4 μM). Mean ± SEM (n = 3
   replicates). Data were analyzed using a one-way Welch’s analysis of
   variance (ANOVA) test and Dunnett’s multiple comparisons test (G). *P <
   0.05, ***P < 0.001, and ****P < 0.0001.

   We identified 67 target genes that function as sensitizers in both BL
   cell lines ([91]Fig. 1C and table S1). Gene ontology analysis of these
   hits revealed enrichment for specific biological processes related to
   DNA repair, DNA replication, and DNA damage checkpoint ([92]Fig. 1D and
   table S2), and the network constructed with these common sensitizers
   shows high connectivity between genes of the FA and HR pathways
   ([93]Fig. 1E and fig. S1C). According to this network, we found that
   FIRRM behaves similarly to factors of the cellular response to ICL
   agent in BL cell lines. Previous reports identified FIRRM as a
   modulator of chromosome segregation through the regulation of polo like
   kinase 1 (PLK1) activity([94]18), and its orthologs in Arabidopsis
   thaliana (flip) and Oryza sativa (MEICA: meiotic chromosome association
   1) have been shown to regulate meiotic recombination ([95]19, [96]20),
   thereby emerging as critical player for the maintenance of genome
   stability ([97]21). We therefore focused our attention on FIRRM and
   investigated its role in the cellular response to ICL.

   To validate the contribution of FIRRM in this response, we targeted it
   using a sgRNA in a 20-day CRISPR-based growth competition assay
   ([98]22) that was performed in nontransformed human retinal pigmented
   epithelial cell line retinal pigment epithelial-1 (RPE-1) immortalized
   with telomerase reverse transcriptase (hTERT) (henceforth termed
   RPE1-hTERT p53^−/− cells) ([99]Fig. 1F). Briefly, stable cell lines
   expressing a sgRNA of interest along with a blue fluorescent protein
   (BFP) reporter (sgGOI-BFP) or a control sgRNA against LacZ along with a
   mCherry reporter (sgCtrl-mCherry) were established. At 0 days (D[0]),
   the cell line expressing the sgGOI-BFP was mixed to a 1:1 ratio with
   the sgCtrl-mCherry cell line and treated for D[20] with an IC[25] of
   the ICL-inducing agent cisplatin (2.4 nM) before BFP and mCherry
   measurements at different time points. For comparison, we also tested
   sgRNAs that target representative genes of the FA (sgFANCA) and the HR
   pathways (sgMRE11). In this assay, FIRRM knockout hypersensitizes
   RPE1-hTERT p53^−/− cells to cisplatin ([100]Fig. 1G and fig. S1D),
   demonstrating that the phenotype associated with FIRRM depletion is
   neither cell line nor ICL agent specific. The effect of FIRRM depletion
   on the cellular fitness of cisplatin-treated cells is intermediate of a
   core FA protein and MRE11 homolog, double strand break repair nuclease
   (MRE11) ([101]Fig. 1G and fig. S1D), suggesting that FIRRM is important
   for cellular survival to ICL-inducing agent but may not be a core FA
   gene. Together, this work demonstrates that our CRISPR-based screening
   approach effectively identified genes affecting the response to ICL
   agents, including modulators like FIRRM.

FIRRM depletion induces genomic instability and cellular senescence in a
p53-dependent manner

   To define the role of FIRRM in the maintenance of genome stability, we
   depleted it using a small-interfering RNA (siRNA) in the human bone
   osteosarcoma epithelial U2OS cells (fig. S2A). At 48 hours after
   transfection, FIRRM-depleted U2OS cells accumulated micronuclei (MN)
   ([102]Fig. 2, A and B). MN formation is triggered by unresolved genomic
   instability, such as DSBs and lagging chromosomes ([103]23). In
   FIRRM-depleted cells, we observed that most MN show negative staining
   for centromeres but positive staining for the DSB marker phosphorylated
   form of the histone variant H2AX (γ-H2AX) ([104]Fig. 2, A and B),
   suggesting that they arise from improperly segregated acentric
   chromosome fragments induced by unrepaired DSBs ([105]23). 53BP1
   nuclear bodies (NBs) are another marker of genomic instability,
   representing subnuclear structures assembled around DNA lesions caused
   by replication stress, and transmitted during mitosis to the daughter
   cells ([106]24). Untreated FIRRM-depleted U2OS cells exhibit a twofold
   increase in the number of 53BP1-NBs in the G[1] phase of the cell cycle
   (fig. S2, B and C), suggesting that FIRRM plays a role in preventing
   replication stress-induced DNA breaks. To further explore the impact of
   FIRRM on genomic stability, we quantified the number and the intensity
   of γ-H2AX foci that accumulate in FIRRM-depleted U2OS cells that are in
   interphase. We found that both the number and the intensity of γ-H2AX
   foci were remarkably increased in FIRRM-depleted cells treated with 1
   μM cisplatin for 3 hours compared to the control (cells treated with
   siCtrl) ([107]Fig. 2C and fig. S2D). Depletion of FIRRM is sufficient
   to increase the number and the intensity of γ-H2AX foci in untreated
   cells ([108]Fig. 2C and fig. S2D), indicative of a contribution of
   FIRRM to protect cells from endogenous DNA damage. As a complementary
   approach, we quantified the dynamics of DSBs induced by cisplatin
   treatment (250 nM, 12 hours) in FIRRM-depleted U2OS cells using the
   neutral comet assay. We noted that significant levels of ICL-induced
   DSBs persist in FIRRM-depleted U2OS cells 72 hours after treatment
   compared to control conditions ([109]Fig.2, D and E), indicative of a
   defect in ICL repair.

Fig. 2. Depletion of FIRRM leads to genomic instability and
senescence-induced cell death.

   [110]Fig. 2.
   [111]Open in a new tab

   (A) Representative images of U2OS cells depleted for FIRRM and
   processed for γ-H2AX (red), Lamin-B1 (green), and centromeres (CREST)
   (gray) immunofluorescence. Nuclei were counterstained with DAPI. Scale
   bar, 5 μm. (B) Quantification of cells treated as in (A). Mean ± SD (n
   = 3 independent experiments). (C) Quantification of γ-H2AX foci per
   cell (top) and γ-H2AX foci intensity (bottom) in cells depleted for
   FIRRM and exposed to 1 μM cisplatin or vehicle (DMSO) for 3 hours. Mean
   ± SD (n = 3 independent experiments). (D) Representative images of
   comet neutral assay in U2OS cells depleted or not for FIRRM and treated
   or not with 250 nM cisplatin for 24 hours and collected at the
   indicated time points. (E) Quantification of tail moment as shown in
   (D). Data are represented as box and whiskers (10th to 90th percentile)
   (n = 3, at least 150 cells were measured per experiment in each
   condition). A.I., arbitrary intensity. (F) Representative images of
   RPE1-hTERT WT and p53^−/− cells depleted for FIRRM and stained with SA
   β-gal. Scale bar, 100 μM. (G) Quantification of SA β-gal–positive cells
   as shown in (G). Mean ± SEM (n = 3 independent experiments). (H)
   Karyotype representation of RPE1-hTERT p53^−/− cells depleted
   (sgFIRRM-2) for FIRRM displaying aberrant chromosome copy numbers
   through metaphase analysis. (I) Quantification of apoptotic RPE1-hTERT
   p53^−/− cells depleted for FIRRM and treated with cisplatin. Mean ± SEM
   (n = 4, 30,000 cells measured per experiment in each condition). Data
   were analyzed using an unpaired t test with Welch’s correction (B and
   C), a two-way ANOVA, followed by Tukey multiple comparisons test (E) or
   a one-way Welch’s ANOVA test with Dunnett’s multiple comparisons test
   (G and I). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

   In mammalian cells, DNA damage activates checkpoints that promote cell
   cycle arrest, senescence, or ultimately apoptosis in a p53-dependent
   manner ([112]25–[113]28). Consistent with our previous observations
   that FIRRM depletion results in the accumulation of DNA breaks, we
   noticed that its depletion by two distinct small hairpin RNAs (shRNAs)
   is sufficient to induce senescence in RPE1-hTERT and IMR90 p53
   wild-type (WT) cell lines as monitored by β-galactosidase (β-gal)
   staining and RNA levels of well-established senescent markers
   ([114]Fig. 2, F and G, and fig. S2, E to I). This phenotype was
   significantly reduced in RPE1-hTERT cells where TP53 was inactivated by
   CRISPR technology ([115]Fig. 2, F and G). All our attempts to generate
   p53 WT FIRRM knockout cell lines were unsuccessful. In contrast, we
   could isolate two FIRRM^−/− clones in RPE1-hTERT p53^−/− that we
   confirmed by immunoblotting analysis (fig. S2J). Notably, karyotyping
   of FIRRM knockout RPE1 hTERT p53^−/− clones identified major
   chromosomal rearrangements, particularly duplication of chromosomes
   ([116]Fig. 2H and fig. S2K). While duplication of chromosomes was only
   observed in 4 of 30 karyotypes in sgCtrl cells, this phenomenon is
   exacerbated in FIRRM-depleted clones (18 of 35 for sgFIRRM1 and 21 of
   40 for sgFIRRM2), highlighting the central role of FIRRM in the
   maintenance of genome stability. In this context, increased levels of
   apoptosis detected in FIRRM^−/− cells by annexin V/propidium iodide
   (PI) staining showed that both clones are more prone to cell death upon
   cisplatin treatment ([117]Fig. 2I and fig. S2L), confirming the
   protective role of FIRRM against cell death following ICL induction.

FIRRM acts downstream of the FA pathway during ICL repair

   The response to ICLs relies on the activation of the FA pathway, which
   is induced when a replication fork encounters covalently linked DNA
   strands ([118]Fig.3A), and subsequent recruitment of specific nucleases
   that release the ICL from one of the two parental strands ([119]29).
   DSBs left behind by nucleolytic processing are then repaired by
   homology-directed recombination pathways such as HR and single-strand
   annealing (SSA) ([120]Fig. 3A). To delineate at which step(s) FIRRM is
   participating in the ICL response, we first assessed its involvement in
   the recruitment and the activation of FANCD2 at ICLs. As expected, both
   FANCD2 accumulation and ubiquitylation at ICLs were abrogated upon
   depletion of FANCA in U2OS cells, serving as a control ([121]Fig. 3, B
   to D). In contrast, FIRRM depletion did not impair the detection of
   FANCD2 at γ-H2AX foci ([122]Fig. 3, B and C) and its
   mono-ubiquitylation status after cisplatin treatment ([123]Fig. 3D).
   These data suggest that FIRRM plays a role downstream of the FA pathway
   during the response to ICLs. We noticed that FANCA-depleted cells were
   hypersensitive to both ICL agents MAF and cisplatin in our viability
   assay, unlike FIRRM-depleted cells, which displayed a much milder
   response to these drugs (fig. S3A). As previously observed with γ-H2AX
   foci ([124]Fig. 2C and fig. S2D), we noted that more FANCD2 foci and
   ubiquitylated form of FANCD2 accumulate in FIRRM-depleted cells at
   steady state and upon cisplatin exposure ([125]Fig. 3, B to D).

Fig. 3. FIRRM is not a core FA gene but is required for the response to ICL
and replicative stress.

   [126]Fig. 3.
   [127]Open in a new tab

   (A) Schematic of ICL repair pathways in mammalian cells. (B)
   Representative images of U2OS cells depleted for FIRRM or FANCA and
   treated with 1 μM cisplatin for 24 hours. Scale bar, 5 μm. (C)
   Quantification of FANCD2 foci per cell (left) and colocalizing with
   γ-H2AX (right) as in (B). Mean ± SD (n = 3, 100 cells quantified per
   experiment in each condition). (D) Whole-cell extracts (WCE) from U2OS
   cells depleted or not for FIRRM or FANCA and treated with cisplatin or
   DMSO was analyzed by immunoblot. (E) Representative images of U2OS
   2-6-5 cells depleted for FIRRM for 48 hours and ER-mCherry-LacR-FOK1-DD
   expression was induced for 4 hours before fixation. Scale bar, 5 μm.
   (F) Quantification of the ratio of indicated protein foci intensity
   over ER-mCherry-LacR-FOK1-DD foci intensity as shown in (E). Mean ± SD
   (n = 4 independent experiments, 50 foci quantified per experiment in
   each condition). (G) Representative images of U2OS cells depleted for
   FIRRM and treated with 1 μM cisplatin for 24 hours and supplemented
   with 10 μM BrdU for 24 hours. Scale bar, 5 μm. (H) Quantification of
   data shown in (G). (I) Similar to (H) except that cells were depleted
   with siCtIP. (J) Representative images of U2OS 2-6-5 cells depleted for
   FIRRM and transfected with an empty mCherry-LacRnls plasmid for 24
   hours. Scale bar, 5 μm (K) Quantification of data shown in (J). Mean ±
   SD (n = 3 independent experiments, 50 foci quantified per experiment in
   each condition). Data were analyzed using a one-way Welch’s ANOVA test
   and Games-Howell’s multiple comparisons test (C, left), an unpaired t
   test (C, right), or with an unpaired t test with Welch’s correction (F,
   H, I, and K). *P < 0.05, ***P < 0.001, and ****P < 0.0001. A.U.,
   arbitrary units; n.s., not significant.

   To further dissect the stage at which FIRRM influences the ICL
   response, we focused our attention on key HR factors and their
   accumulation at DSBs in the absence of FIRRM. We took advantage of a
   DSB reporter system in which the recruitment of these factors can be
   quantified at a single genomic locus. Briefly, DSBs are rapidly induced
   by the recruitment of the ER-mCherry-LacRnls-FOK1-DD fusion protein at
   a LacO array integrated on chromosome 1p3.6 in U2OS cells (U2OS 2-6-5
   cell line) ([128]30, [129]31). In this assay, depletion of FIRRM
   correlated with an exacerbated accumulation of several HR factors at
   FOK1-induced DSBs, including BRCA1 DNA repair associated (BRCA1),
   BRCA2, and RAD51 ([130]Fig. 3, E and F, and fig. S3, B and C). Notably,
   FIRRM depletion did not affect the recruitment of the nonhomologous end
   joining (NHEJ) factor 53BP1 (fig. S3B), suggesting that HR may be
   favored in the absence of FIRRM. Using native bromodeoxyuridine (BrdU)
   labeling ([131]32, [132]33), we noted a significant increase in the
   levels of ssDNA in FIRRM-depleted proliferating cell nuclear
   antigen–positive U2OS cells treated with cisplatin ([133]Fig. 3, G and
   H). This phenotype is abrogated upon depletion of CTBP-interacting
   protein (CtIP) ([134]Fig. 3I), reflective of increased DNA end
   resection and ssDNA generation in the absence of FIRRM.

   To directly assess the role of FIRRM during homology-directed DNA
   repair, we used the well-established HeLa and U2OS direct repeats
   (DR)–green fluorescent protein (GFP) reporter cell lines ([135]34), as
   well as the U2OS single-strand annealing (SA)-GFP reporter cell line
   ([136]35), to measure DNA repair by HR and SSA, respectively. Depletion
   of FIRRM with four different siRNAs resulted in a significant decrease
   in both assays (fig. S3, D to H), particularly with siFIRRM-3 and
   siFIRRM-4, which results in efficient depletion of the protein as
   determined by immunoblotting analysis. Notably, FIRRM depletion had a
   much milder impact than CtIP or RAD51 depletion in both the DR- and
   SA-GFP assays (fig. S3, E to H), suggestive of a regulatory role rather
   than a core contribution of FIRRM in homology-directed DNA repair
   pathways. In these conditions, we did not observe any significant
   impact of FIRRM depletion on the cell cycle distribution of U2OS cells
   (fig. S3I). Beside their contribution to DNA repair, several HR
   factors, including BRCA1, BRCA2, and RAD51, have been shown to protect
   stalled replication forks from nucleolytic processing and promote their
   stabilization ([137]36). To determine whether FIRRM participates in
   this process, we took advantage of the LacO arrays that pose an
   obstacle to DNA polymerase progression when bound by LacR, resulting in
   a stalled replication fork ([138]37). In this assay, depletion of FIRRM
   also correlated with an exacerbated accumulation of several HR factors,
   including BRCA1, BRCA2, and RAD51, upon tethering of a mCherry-LacRnls
   fusion protein at the LacO arrays ([139]Fig. 3, J and K, and fig. S3, J
   and K), supporting a model where FIRRM plays a role at stalled
   replication forks, beyond ICL repair.

   Consistent with these observations, chemogenomic profiling of CRISPR
   screens performed in RPE1-hTERT p53^−/− cells with different genotoxic
   agents showed that FIRRM depletion impairs predominantly cell survival
   (NormZ < −2.3) upon exposure to agents that induce replication
   roadblocks (e.g., cisplatin), base alkylation [methylnitrosoguanidine
   (MNNG)], and oxidative damage (KBrO[3] and H[2]O[2]) (fig. S3L)
   ([140]38). To validate these findings, we exposed U2OS cells to a
   series of DNA-damaging agents and noticed that depletion of FIRRM
   reduced cell viability upon treatment with cisplatin, formaldehyde
   (FormA), MMC, hydroxyurea (HU), and the poly(ADP-ribose) polymerase
   inhibitor BMN673 (fig. S3M). Notably, the lack of FIRRM did not affect
   U2OS viability upon exposure to either ultraviolet (UV) or ionizing
   irradiation (IR) (fig. S3M). Because these types of DNA lesion are
   primarily repaired by NER and NHEJ, respectively, our data suggest that
   FIRRM does not participate in these repair pathways. Together, these
   results support a model where FIRRM is involved in both the repair of
   DSBs by HR and the cellular response to different replication fork
   roadblocks.

FIRRM interacts with FIGNL1 and regulates RAD51 dynamics during ICL repair

   To further define the role of FIRRM in DNA repair, we investigated
   whether its depletion affects RAD51 focus formation and resolution in
   U2OS cells treated with an ICL-inducing agent. As expected, RAD51 foci
   rapidly appeared 3 to 6 hours after the addition of cisplatin to the
   media ([141]Fig. 4, A and B, and fig. S4A). Notably, more than 40% of
   FIRRM-depleted cells still exhibited >5 RAD51 foci 72 hours after
   treatment, a time point that was sufficient for the resolution of these
   foci in the control condition ([142]Fig. 4, A and B). Furthermore, the
   intensity of the RAD51 foci in FIRRM-depleted cells was greater than
   the ones in cells treated with a siRNA control at a late time point
   (fig. S4B), suggesting that RAD51 dynamics at ICL-induced DSBs are
   compromised upon depletion of FIRRM. Partial depletion of RAD51 in
   FIRRM-depleted cells restored cell survival upon cisplatin treatment
   ([143]Fig. 4C and fig. S4C), suggesting that the accumulation of RAD51
   in these cells is toxic and establishing that the regulation of RAD51
   dynamics by FIRRM is critical for cell survival in response to
   cisplatin treatment.

Fig. 4. The FIRRM-FIGNL1 complex regulates RAD51 unloading from DNA.

   [144]Fig. 4.
   [145]Open in a new tab

   (A) Quantification of cisplatin-induced RAD51 foci. Mean ± SD (n = 3,
   100 cells per experiment in each condition). (B) Representative images
   from (A). Scale bar, 5 μm. (C) Viability assay in U2OS cells treated
   with siFIRRM and siRAD51 prior cisplatin. (D) High-confidence proximal
   interactors of FIRRM identified by mTurboID. Bayesian false discovery
   rate, BFDR. (E) Schematic representation of FIRRM and FIGNL1 protein
   domains. DUF, domain of unknown function; WCF, WCF tripeptide sequence;
   FRBD, FIGNL1 RAD51 binding domain; V, vps4 domain. (F) Quantification
   of the colocalization of FIRRM and FIGNL1 at a LacO array. Mean ± SD (n
   = 3, 50 cells per condition). (G) Representative images from (F). Scale
   bar, 5 μm. (H) WCE from U2OS cells depleted for FIRRM and processed for
   FIRRM immunoprecipitation (IP) (n = 3). IgG, immunoglobulin G. (I) Tail
   moment in U2OS cells as in [146]Fig. 2E (n = 3, >150 cells per
   condition). (J) Competitive growth results in FIGNL1 and/or FIRRM
   depleted RPE1-hTERT p53^−/− cells as in [147]Fig. 1G. Mean ± SEM (n =
   3). (K) WCE from U2OS cells were depleted for FIRRM or FIGNL1 and
   analyzed by immunoblot. (L) Turbo-HA-2xStrep-FIRRM U2OS cells were
   treated with siFIRRM and supplemented (black dot) with doxycycline
   (Dox). WCE was analyzed by immunoblot. * Denotes a nonspecific band.
   (M) Quantification of Turbo-HA-2xStrep-FIRRM U2OS cells with ≥5 RAD51
   foci following cisplatin treatment. As in (A). Mean ± SD (n = 3, 100
   cells for each condition). (N) Turbo-HA-2xStrep-FIRRM U2OS cells
   treated with siFIRRM, Dox, and cisplatin before survival were
   monitored. Mean ± SD (n = 4). One-way Welch’s ANOVA test and Dunnett’s
   multiple comparisons test for (C), (J), and (N), two-way ANOVA followed
   by Tukey multiple comparisons test for (I), or unpaired t test with
   Welch’s correction for (M). *P < 0.05, **P < 0.01, ***P < 0.001, and
   ****P < 0.0001.

   To gain insight into the regulation of RAD51 dynamics by FIRRM, we
   mapped its proximal interactome by taking advantage of the miniTurboID
   technology (fig. S4D) ([148]39). Briefly, FIRRM was N-terminally tagged
   with the promiscuous biotin ligase miniTurbo (mTurboID) and stably
   expressed in HEK293 Flp-In cells. This construct was validated for its
   ability to efficiently biotinylate proximal partners by immunoblotting
   analysis (fig. S4E). The proximity interacting network of FIRRM was
   generated at steady state following 3 hours of biotin labeling. A total
   of 155 high-confidence FIRRM interactors found in three technical
   replicates were identified (table S3). Among these interactors, we
   identified three high-confidence candidate proteins that were also
   identified as sensitizers to MAF in our CRISPR screen ([149]Fig. 4D and
   table S1). Two of these interactors, the pericentriolar material 1
   protein and the DNA excision repair protein ERCC6-like (ERCC6L, also
   known as PICH), play a role in mitosis where they respectively
   contribute to the proper assembly of functional centrosomes and the
   resolution of anaphase bridges ([150]39–[151]41). Therefore, we
   reasoned that they most likely play a role in the mitotic function of
   FIRRM ([152]18). The AAA^+ ATPase FIGNL1 was identified as a
   constitutive interactor of FIRRM using mTurboID, an interaction that
   has also been detected previously with two orthogonal approaches [yeast
   two-hybrid ([153]19) and affinity purification coupled to mass
   spectrometry (AP-MS) ([154]17, [155]42)]. Although FIGNL1 has been
   detected at centromeres, two studies support a role for the protein in
   regulating RAD51 filament during HR repair ([156]9, [157]17) raising
   the possibility that FIGNL1 and FIRRM act as a complex in this process.

   To further characterize the interaction between FIGNL1 and FIRRM, we
   took advantage of the U2OS 2-6-5 cell line that contains a LacO array
   integrated into a single locus ([158]30). In this context, transiently
   transfected mCherry-LacRnls-bait fusion proteins accumulate at the
   array in U2OS 2-6-5 cells, and the colocalization of preys is
   quantified at the mCherry locus ([159]Fig. 4, E to G, and fig. S4, F
   and G). Notably, FOK1 nuclease was not expressed in this set of
   experiments. When fused to mCherry-LacRnls, both FIRRM and FIGNL1 were
   able to respectively trigger the accumulation of GFP-tagged FIGNL1 and
   FIRRM at the LacO array ([160]Fig. 4, E to G). In contrast, neither of
   the mCherry-LacRnls-tagged proteins recruited GFP alone, which was used
   as a negative control, at the LacO array. Consistently,
   immunoprecipitation of endogenous FIRRM from U2OS cell extracts treated
   with benzonase confirmed that it can pull down endogenous FIGNL1 in the
   absence of DNA/RNA ([161]Fig. 4H).

   If FIRRM and FIGNL1 cooperate at the same stage of ICL repair, FIGNL1
   depletion should result in phenotypes that are comparable to FIRRM
   depletion. The lack of FIGNL1 correlated with an exacerbated
   accumulation of several HR factors at FOK1-induced DSBs U2OS 2-6-5
   cells (figs. S4, H and I), the persistence of DSBs in U2OS upon
   cisplatin exposure when assessed by the neutral comet assay ([162]Fig.
   4I), and a higher sensitivity to cisplatin in CRISPR-based growth
   competition assay ([163]Fig. 4J). Furthermore, both siRNAs targeting
   FIGNL1 led to a significant defect in HR in the DR-GFP assay (fig.
   S4J). To test whether FIGNL1 depletion is epistatic to the loss of
   FIRRM, we examined the impact of either single or co-depletion of
   FIGNL1 and FIRRM in the neutral comet assay and the competitive growth
   assay upon treatment with cisplatin. In these assays, co-depletion of
   FIRRM and FIGNL1 did not exacerbate the phenotype observed when
   knocking down either of the protein ([164]Fig. 4, I and J). Similarly,
   depleting FIGNL1 and FIRRM led to a similar defect in HR than FIRRM or
   FIGNL1 alone in the DR-GFP assay (fig. S4J). As FIGNL1 promotes RAD51
   unloading at DSBs, our findings support a model where FIGNL1 and FIRRM
   are functionally epistatic in the resolution of DNA breaks.

   During these experiments, we noticed that depletion of either FIRRM or
   FIGNL1 by siRNA drastically reduce the level of its respective partner
   ([165]Fig. 4K), suggesting that FIRRM-FIGNL1 interaction is critical
   for the stabilization of the complex. Consistent with this model, we
   found that reexpression of FIRRM in FIRRM-depleted cells rescued the
   stability of FIGNL1 along with the resolution of cisplatin-induced
   RAD51 foci and cellular resistance to cisplatin treatment in survival
   assay ([166]Fig. 4, L to N, and fig. S4, K to M). The U2OS cell line
   expressing an inducible cassette of TurboID-HA-2xStrep-FIRRM was
   obtained by infecting cells with lentivirus, selecting the transduced
   population with puromycin and isolating clones. We observed that the
   expression of the tagged version of FIRRM induces a concomitant
   reduction of both endogenous FIRRM and FIGNL1 at high doses of
   doxycycline, highlighting the codependency of their stability and
   revealing a tight regulation of their levels in cells (fig. S4N). The
   requirement for such regulation is still unclear as high levels of
   FIRRM by itself is not toxic for the cells (fig. S4O).

A mutant that stabilizes FIRRM in the absence of FIGNL1 efficiently promotes
RAD51 resolution at ICLs

   FIGNL1 is a protein of 674 amino acids with three conserved domains:
   (i) a RAD51 binding domain [FIGNL1 RAD51 binding domain (FRBD)], (ii)
   an AAA^+ ATPase domain, and (iii) a C-terminal Vps4 domain ([167]Fig.
   5A) ([168]9, [169]17). FIGNL1 has been shown to promote RAD51
   dissociation from ssDNA ([170]9, [171]17). To define the role of the
   interplay between FIRRM and FIGNL1 during ICL repair, we first took
   advantage of our ability to detect the interaction of GFP-FIRRM with
   mCherry-LacRnls-FIGNL1 in the LacO/LacR assay ([172]Fig. 4, C and D) to
   rapidly screen a panel of FIGNL1 truncation and mutant proteins
   ([173]Fig. 5A). First, mCherry-LacRnls-FIGNL1 constructs were designed
   to remove or abolish the function of the previously described RAD51
   binding domain (FRBD) ([174]17) and AAA^+ ATPase activity of FIGNL1
   ([175]9) ([176]Fig. 5A and fig. S5A). In this experiment, FIGNL1
   constructs lacking the ability to interact with RAD51 (∆FRBD or F295E
   mutant) still efficiently recruited GFP-FIRRM to the LacO array
   ([177]Fig. 5B and fig. S5, B to D). Similarly, none of the single
   (K447A and D500A) and double (KDm: K447A/D500A) mutations of residues
   that are highly conserved within the AAA^+ ATPase domain affected the
   interaction of FIGNL1 with FIRRM ([178]Fig. 5B and fig. S5, B to D).
   The fact that FIGNL1 constructs encompassing amino acids 1 to 120 or
   121 to 674 are unable to recruit FIRRM to the array suggests that the
   FIRRM binding motif of FIGNL1 spans over the intersection of these
   constructs ([179]Fig. 5B). Consistently, the FIGNL1 fragments 1 to 360
   efficiently promoted the accumulation of FIRRM in our assay.

Fig. 5. A stable FIRRM is capable of promoting RAD51 unloading.

   [180]Fig. 5.
   [181]Open in a new tab

   (A) Schematic representation of FIGNL1 constructs used in this study.
   (B) Quantification of mCherry-LacRnls constructs colocalizing with
   GFP-FIRRM in U2OS 2-6-5 cells and processed as in [182]Fig.4D. Mean ±
   SD (n = 3). (C) Quantification of RAD51 foci intensity over mCherry
   signal at the LacO array foci. Mean ± SD (n = 3). (D) Representative
   images of U2OS 2-6-5 cells transfected with mCherry-LacR-FIGNL1 and
   GFP-FIRRM. Scale bar, 5 μm. (E) Quantification of U2OS 2-6-5 cells
   transfected with mCherry-LacRnls-FIRRM or -FIGNL1 and processed for
   RAD51 immunofluorescence. (F) Representative images of data shown in
   (E). Scale bar, 5 μm. (G) Schematic representation of FIRRM full length
   (FL) and truncated proteins (top). Quantification of GFP-tagged
   proteins colocalizing with mCherry-LacRnls-FIGNL1 in U2OS 2-6-5 cells
   is presented (bottom). Mean ± SD (n = 3). (H) Representative images of
   data quantified in (D). Scale bar, 5 μm. (I) WCE from U2OS cells
   depleted for FIGNL1 and transfected with mCherry-FIGNL1 and
   3xFLAG-FIRRM FL or ∆WCF for 24 hours was processed for FLAG IP and
   analyzed by immunoblot. (J) WCE from U2OS cells overexpressing stable
   FIRRM FL or ∆WCF constructs depleted (black dot) for endogenous FIRRM
   and treated with Dox for 24 hours before analysis by immunoblot. *
   Denotes a nonspecific band. (K) Viability assay in WT, GFP-FIRRM WT,
   and GFP-FIRRM ∆WCF U2OS cells treated or not with siFIRRM before
   cisplatin treatment. Mean ± SD (n = 3). (L) Quantification of GFP-FIRRM
   WT and GFP-FIRRM ∆WCF U2OS cells with ≥5 RAD51 foci following treatment
   with cisplatin and Dox. Mean ± SD (n = 3). Data were analyzed with an
   unpaired t test with Welch’s correction (C, K, and L), a one-way
   Welch’s ANOVA test, and Dunnett’s multiple comparisons test (G). *P <
   0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

   As mCherry-LacRnls-FIGNL1 efficiently recruits RAD51 to the LacO array,
   we investigated whether this function can be regulated by the levels of
   FIRRM in cells. We observed a significant increase of RAD51 intensity
   at the mCherry locus in FIRRM-depleted U2OS 2-6-5 cells ([183]Fig. 5C),
   suggesting that FIRRM negatively controls or counteracts
   FIGNL1-dependent RAD51 dynamics. Consistent with this hypothesis,
   ectopic expression of GFP-FIRRM had the opposite effect. It resulted in
   a reduced signal of RAD51 at the LacO array when compared to the
   expression of GFP alone ([184]Fig. 5, C and D). While
   mCherry-LacRnls-FIGNL1 efficiently recruits RAD51 to the array, we
   observed that mCherry-LacRnls-FIRRM is unable to do the same ([185]Fig.
   5, E and F), suggesting that FIRRM does not directly interact with
   RAD51.

   FIRRM is a protein of 853 amino acids containing a domain of unknown
   function (DUF4487) ([186]Fig. 5G) ([187]19). FIRRM also harbors a
   conserved WCF tripeptide motif essential for the interaction with
   FIGNL1 as described in the LacO/LacR assay ([188]Fig.5, G and H, and
   fig. S5E) ([189]21). Under these conditions, the C-terminal region of
   FIRRM is dispensable for the interaction with FIGNL1. Using Flag-tagged
   constructs of both WT and ΔWCF FIRRM, we confirmed the importance of
   this motif for FIGNL1 interaction ([190]Fig. 5I). In contrast to FIRRM
   WT, reexpression of a GFP-tagged version of FIRRM ΔWCF mutant in
   FIRRM-depleted U2OS cells failed to restore FIGNL1 protein levels
   ([191]Fig. 5J), confirming that FIRRM-FIGNL1 interaction is critical
   for the stability of this complex. Notably, reexpression of FIRRM ΔWCF
   restored the survival of cells to cisplatin ([192]Fig. 5K and fig.
   S5F), prevented the formation of MN (fig. S5G), and facilitated RAD51
   foci resolution in U2OS cells ([193]Fig. 5L and fig. S5, H to J),
   suggesting that, once stabilized, FIRRM is capable of promoting RAD51
   disassembly in a FIGNL1-independent manner.

FIRRM binds ssDNA where it facilitates the unloading of RAD51 in vitro

   To better understand how FIRRM may regulate RAD51 dynamics, we further
   investigated the biochemical activities and interactions with the RAD51
   recombinase using full-length FIGNL1 and FIRRM either purified from
   bacteria or using Sf9 insect cells (iFIRRM and iFIGNL1) (fig. S6A). As
   reported previously ([194]17), a strong interaction between FIGNL1 and
   RAD51 was detected following immunoprecipitation of RAD51 ([195]Fig.
   6A). In line with our previous observation in the LacO/LacR assay
   ([196]Fig. 5, E and F), no direct interaction was detected between
   recombinant RAD51 and FIRRM in pull-down assays ([197]Fig. 6A). A
   co-complex between RAD51 and FIRRM was detected upon adding FIGNL1 to
   the reaction likely reflecting the formation of the FIRRM-FIGNL1
   complex with RAD51 in vitro.

Fig. 6. FIRRM interacts with ssDNA and SA DNA and forms a complex with FIGNL1
and RAD51.

   [198]Fig. 6.
   [199]Open in a new tab

   (A) Recombinant RAD51, FIGNL1, and FIRRM were processed for
   immunoprecipitation with an anti-RAD51 and analyzed by immunoblot with
   the indicated antibody. (B) Electrophoretic mobility shift assays
   (EMSA) of FIGNL1 and FIRRM on ssDNA. (C) Quantification of the
   percentage of DNA binding shown in (B). (D and E) Quantification of the
   percentage of DNA binding observed in EMSA experiment (fig. S6, C and
   D) on splayed arms and double-strand DNA substrates. (F) Representative
   cryo-EM micrograph of FIGNL1 bound to AMPPNP (1 mM) and MgCl[2] (2 mM).
   Two D classes were obtained from 1287 particles. (G) Model of the
   RAD51-FIGNL1-FIRRM interaction in the presence of DNA.

   During HR, the RAD51 recombinase nucleates on ssDNA to form an active
   nucleoprotein filament, which is highly regulated by several cofactors
   ([200]43). To assess whether FIGNL1 and FIRRM bind ssDNA, we performed
   DNA binding assays. Both full-length FIGNL1 and FIRRM purified from
   bacteria and insect cells bind ssDNA in electrophoretic mobility shift
   assay (EMSA) in the presence or absence of a cross-linking agent
   ([201]Fig. 6B and fig. S6B). In this type of assay, bacterially
   purified FIGNL1 and FIRRM proteins bind to a 60-nucleotide-long ssDNA
   probe and, to a lesser extent, a splayed arms DNA, but not to a
   double-strand DNA probe of the same size ([202]Fig. 6, C to E, and fig.
   S6, B to D). During the purification of FIGNL1, we noticed that the
   protein purified form insect cells eluted from gel filtration column at
   a higher molecular weight than the predicted monomeric form (fig. S6E),
   suggesting that the protein forms multimers like other AAA^+ ATPases.
   Cryo–electron microscopy (cryo-EM) and modeling approaches revealed
   that the protein preferentially associated into hexamers ([203]Fig. 6F
   and fig. S6, E and F), with a central channel that is not large enough
   to accommodate ssDNA. On the basis of these observations, we propose a
   model where the functions of FIGNL1 and FIRRM at ICL repair rely on
   their ability to interact with DNA ([204]Fig. 6G) ([205]44, [206]45).
   In this model, only FIGNL1 interacts with RAD51.

DISCUSSION

FIRRM contributes to HR-mediated ICL repair and beyond

   While the initial steps involved in ICL repair have been extensively
   described, it remains largely unclear how ICL intermediates are
   processed and subsequently resolved. CRISPR-based functional genomics
   have been a powerful strategy to gain insight into several DNA repair
   pathways ([207]38), including the recent characterization of SCAI as a
   modulator of repair pathway choice during ICL processing ([208]7,
   [209]8). Here, we delineated the contribution of FIRRM in the latter
   stage of ICL repair using a similar strategy. FIRRM has been recently
   proposed to play a significant role during the DNA damage response
   ([210]21), and we showed here that lack of FIRRM impairs cell survival
   in response to a series of ICL-inducing agents. Under these conditions,
   FIRRM acts downstream of the FA pathway where it is functionally
   epistatic with the AAA^+ ATPase FIGNL1 to resolve HR-mediated
   intermediates. Specifically, our findings revealed that FIRRM is
   required for the resolution of RAD51 from ssDNA, a phenotype that is
   reminiscent of the activity described for FIGNL1 ([211]9). Consistent
   with a coordinate role of FIRRM and FIGNL1 in regulating HR-mediated
   DNA repair, a mutation in figl1-1, the ortholog of FIGNL1 in A.
   thaliana, is epistatic with flip-1 mutation, the ortholog of FIRRM, in
   limiting meiotic crossover formation ([212]19). Similar findings were
   observed in O. sativa, where targeting of MEICA, the ortholog of FIRRM
   in rice, is critical to prevent aberrant meiotic recombination and
   regulate crossover formation ([213]20). More recently, FIRRM was found
   to genetically interact with factors implicated in the resolution of HR
   intermediates and the promotion of sister chromatid exchange in
   mammalian cells ([214]46). While we could not detect the recruitment of
   endogenous FIRRM at ICLs, the endogenous tag version of FIRRM was
   recently shown to accumulate at MMC-induced damage and at laser UV-A
   micro-irradiation sites after psoralen treatment ([215]46). FIRRM has
   also been detected at stalled replication fork by isolation of proteins
   on nascent DNA (iPOND) ([216]47–[217]49). This observation fits well
   with our findings that FIRRM regulates the recruitment of HR factors at
   stalled replication forks in the LacR assay and suggests that FIRRM
   plays a role beyond the repair of ICL-induced DSBs. The fact that FIRRM
   limits the formation of MN and the accumulation of 53BP1 NBs and
   prevents cellular senescence supports a model where FIRRM participates
   in the response to replication stress.

FIRRM forms a complex with FIGNL1 to regulate RAD51 accumulation at DSBs

   Previous reports identified FIGNL1 as a key regulator of the resolution
   of RAD51 foci in different biological contexts ([218]9, [219]17,
   [220]19). Our systematic proteomics analysis of FIRRM identified FIGNL1
   as a constitutive proximal interactor that we validated in cellula and
   by coimmunoprecipitation. Our data clearly show that the FIRRM-FIGNL1
   interaction is necessary for the stability of the complex, reminiscent
   of what has been previously observed with the BRCA1/BARD1 heterodimer,
   where each member of the heterodimer controls the abundance, stability,
   and function of the other ([221]50, [222]51).

   Our in vitro biochemical analysis provided better insight into the role
   of FIGNL1 and FIRRM during RAD51 dynamics. First, our characterization
   showed that FIGNL1, unlike FIRRM, strongly interacts with RAD51.
   Second, we noticed that both FIGNL1 and FIRRM preferentially interact
   with ssDNA, which is produced during the DNA resection step of HR. The
   integration of these data, along with previous findings demonstrating
   the ability of FIGNL1 to promote RAD51 filament disassembly in vitro
   ([223]9), supports a model where FIRRM acts as a scaffolding platform
   for FIGNL1 hexameric ring structure to promote the release of ssDNA
   from RAD51 ([224]Fig. 6G), thereby regulating RAD51 unloading at DSBs.
   While this model is attractive, it does not fully explain why FIRRM
   alone, when stabilized by the deletion of its WCF motif, promotes the
   resolution of RAD51. Additional biochemical investigation will be
   required to fully address this function. Similarly, further studies
   will be required to understand FIGNL1 and FIRRM cooperate with numerous
   other players, such as partner and localizer of BRCA2 (PALB2)/BRCA2
   ([225]52) and RPA1 related single stranded DNA binding protein,
   X-linked (RADX) ([226]53, [227]54), which respectively promote RAD51
   nucleoprotein assembly and displace RAD51 from ssDNA, to regulate RAD51
   filament dynamics in mammalian cells.

FIRRM has independent function of FIGNL1 with potential relevance for cancer
therapy

   In this study, we found that the stability of FIRRM and FIGNL1 is
   interdependent. Whether both proteins are always in complex with one
   another remains however unknown. In species such as fungi and other
   model organisms such as Caenorhabditis elegans and Drosophila
   melanogaster ([228]19), only FIGNL1 is conserved raising the
   possibility that it has conserved functions that are independent of
   FIRRM. Consistently, the predicted FIRRM interacting region is located
   in a less conserved region of FIGNL1. Depending on the cellular
   context, it is also possible that FIRRM forms different molecular
   complexes to preserve genomic stability, similar to other scaffolding
   proteins such as REV7 (MAD2L2: mitotic arrest deficient 2 like 2)
   [reviewed in ([229]55)]. In line with this hypothesis, our
   structure-function analysis identified a mutant of FIRRM that
   stabilizes the protein in the absence of FIGNL1 and is functionally
   capable of resolving RAD51 foci, preventing MN formation and promoting
   survival to cisplatin. These findings suggest that FIRRM has functions
   that are independent of FIGNL1 in the regulation of RAD51 dynamics and
   ICL repair. Furthermore, our proteomic analysis identified ERCC6L as
   one of the top constitutive interactors of FIRRM, which may likely
   reflect a role during mitotic progression ([230]18). These previously
   unidentified functions of FIRRM may be relevant in cancers where high
   levels of FIRRM is associated with poor outcome, including gliomas
   ([231]56), and therefore may be a potential prognostic marker in
   multiple tumor types ([232]57). Whether FIGNL1 is also expressed in
   these cancers is unknown. Together, our study identified FIRRM as a
   regulator of ICL-induced genomic instability and concomitant survival
   by modulating RAD51 dynamics. In line with this finding, high
   expression of FIRRM in cancer cells may provide an evolutionary
   advantage to resist stringent cellular conditions such as replication
   stress.

METHODS

Cell culture and plasmids transfections

   All cell lines were maintained at 37°C and 5% CO[2]. Namalwa and Raji
   BL cell lines (a gift of J. Pelletier, McGill University) were cultured
   in RPMI 1640 medium (Wisent) and supplemented with 20% fetal bovine
   serum (FBS; Sigma-Aldrich) and 1% penicillin-streptomycin (Wisent).
   RPE1-hTERT p53 WT and p53^−/− cells (a gift of D. Durocher, University
   of Toronto) and IMR90 cells [National Institute of General Medical
   Sciences (NIGMS) Human Genetic Cell Repository I90-10] were cultured in
   Dulbecco’s modified Eagle’s medium (DMEM) medium (Wisent) and
   supplemented with 10% FBS and 1% penicillin-streptomycin. U2OS
   (American Type Culture Collection, number HTB-96) and U2OS 2-6-5 cell
   lines [a gift from R. Greenberg (University of Pennsylvania,
   Philadelphia, PA) ([233]30)] were cultured in McCoy’s medium (Life
   Technologies) and supplemented with 10% FBS. Hela DR-GFP cells, U2OS
   DR-GFP, and U2OS SA-GFP cells (gift of J. Stark, City of Hope) were
   cultured in DMEM and McCoy’s medium (Wisent), respectively. These media
   were supplemented with 10% FBS and 1% penicillin-streptomycin.

   RPE1-hTERT Cas9 p53^−/− FIRRM^−/− cells were generated by transduction
   of the parental cell line with lentiviral vectors harboring
   FIRRM-sgRNAs targeting exons 5 and 13 (table S4). Cells were selected
   with puromycin for 4 days, and clones were isolated by single-cell
   sorting. Knockout cells were validated using immunoblotting. Stable
   Cas9-expressing BL cell lines (Namalwa and Raji Burkitt cells) were
   generated using the LentiCas9-Blast vector as previously described
   ([234]58) and validated for Cas9 expression by immunoblotting analysis.
   Cell lines used for all mTurboID-MS experiments were generated in
   HEK293 Flp-In T-REx cells as previously described ([235]14) and a pool
   of stable transfectants selected with hygromycin (200 μg/ml; Multicell)
   and blasticidin (5 μg/ml;InvivoGen). Expression of bait proteins were
   induced with tetracycline (1 μg/ml) for 24 hours. U2OS Turbo-HA-2xStrep
   FIRRM WT/∆WCF or GFP-FIRRM WT/∆WCF cells were generated by transduction
   of the parental cell line with pCW57.1-Turbo-HA-2xStrep or pCW57.1-GFP
   expressing lentiviral vectors. Cells were selected with puromycin (2
   μg/ml), and clones were isolated using a limited dilution approach.
   Expression of the constructs was induced with the addition of
   doxycycline (0.1, 1.0, or 5 μg/ml) and validated by immunoblotting
   analysis.

   Plasmid transfections were done using Lipofectamine 2000 (Invitrogen)
   and TransIT-LT1 (Mirus) transfection agents according to the
   manufacturer’s instructions. Lentiviruses were produced as previously
   described ([236]59). All cell lines were validated using short tandem
   repeat markers and tested negative for mycoplasma contamination.

RNA interference

   The pLKO shRNA plasmids against FIRRM were obtained from the McGill
   Platform for Cellular Perturbation as part of the TRC/RNAi Consortium
   from the Broad Institute. Nontargeting shRNA control was purchased from
   Addgene. Single siRNA duplex sequences targeting FIRRM, FIGNL1, FANCA,
   a custom nontargeting sequence, and SMARTPool siRNAs targeting RAD51 or
   CtIP were purchased from Horizon Discovery. Unless stated otherwise,
   all siRNAs were transfected at a concentration of 25 nM in a forward
   transfection for 48 hours using RNAiMAX (Invitrogen) according to the
   manufacturer’s instructions. Knockdowns were confirmed by
   immunoblotting or reverse transcription quantitative polymerase chain
   reaction (RT-qPCR) analyses. All siRNA, shRNA, and RT-qPCR primers
   sequences are detailed in table S4.

Chemicals and sources of DNA damage

   In the FOK1 system, DSBs were induced at the LacO array by inducing the
   nuclear localization [4-hydroxytamoxifen (100 nM, #3412, Tocris)] and
   stabilization (Shield-1 ligand, 0.5 μM; CIP-S1-0001, CheminPharma) of
   the ER-mCherry-LacR-FOK1-DD nuclease for 4 hours before
   immunofluorescence sample preparation. Replication fork stall at the
   LacO array was achieved by transfecting the empty mCherry-LacR fusion
   protein 24 hours before fixation. DNA damages were induced by exposing
   cells to either ionizing irradiation (IR 1 or 10 Gy) with a CellRad
   (Precision X-Ray Inc.), cisplatin treatments (Tocris; 1 μM for 3 or 24
   hours at 250 nM in time course experiment and 2 μM in survival assay),
   MAF (1.30 and 1.57 μM in Namalwa and Raji screens, respectively;
   Toronto Research Chemicals), HU (4 and 1 mM in survival assay;
   Sigma-Aldrich), formaldehyde (150 μM; Sigma-Aldrich), MMC (150 nM;
   Sigma-Aldrich), talazoparib (3 μM; BMN 673, Selleck Chemicals), or UV
   (10 J/m^2). UV irradiations were carried on with a germicidal lamp (243
   nm).

Plasmids

   DNA sequences of sgRNAs were cloned into a modified form of
   LentiGuide-puro with BFP or mCherry marker as described in ([237]22,
   [238]60). The cDNA for FIRRM (MHS1010-202739778) and FIGNL1
   (MHS6278-202759761) were obtained from Horizon Discovery, and coding
   sequences with respective attB1/B2 adapters were PCR-amplified and
   subcloned into a pDONR221 vector using a BP clonase II reaction
   according to the manufacturer’s instructions (Invitrogen), to generate
   ENTRY vectors. ENTRY clones containing FIRRM constructs (amino acids 1
   to 175 and 1 to 735) and FIGNL1 constructs (amino acids 1 to 120, 121
   to 674, 1 to 360, 361 to 674, and 295 to 344) were also generated using
   this approach. FIRRM deletion (∆WCF), FIGNL1 mutations and depletion
   constructs (∆295-344, F295E, K447A, D500A, and K477A/D500A) were
   directly introduced in ENTRY vectors using the QuickChange approach.
   QuickChange Site-Directed Mutagenesis was conducted using primers
   listed in table S4 using the QuickChange (Agilent) or Q5-Site-directed
   mutagenesis kit (New England Biolabs) and following the manufacturer’s
   protocol. LR recombining reactions were then conducted to transfer
   ENTRY constructs into pDEST-mTurboID (a gift from A.-C. Gingras,
   University of Toronto), pDEST pCW57.1-TurboID-HA-2xStrep ([239]61),
   pDEST pCW57.1 GFP ([240]59), pDEST-mCherry-LacRnls (a gift from D.
   Durocher, University of Toronto), or pDEST-FRT-TO-eGFP-nls. To produce
   recombinant proteins, full-length codon-optimized human FIGNL1
   (UniProt, entry [241]Q6PIW4) and FIRRM (UniProt, entry [242]Q9NSG2)
   were purchased from GenScript and subcloned into a pGex6p2–glutathione
   S-transferase (GST)–His plasmid or a modified pFastBac dual vector
   (pfbdm) to be expressed in Escherichia coli and baculovirus,
   respectively. FIRRM and FIGNL1 were subcloned into pfbdm using
   NEBuilder HiFi DNA Assembly (New England Biolabs). The modified pFBDM
   expression vector includes an N-terminal GFP tag removable with tobacco
   etch virus protease. All sequences were validated through Sanger
   sequencing. A list of plasmids used in this study is provided in table
   S6.

CRISPR- based genome-wide screening

   Namalwa and Raji genome-wide screens were executed as previously
   described ([243]14). Briefly, 245 million Namalwa and Raji cells
   expressing Cas9 were transduced with TKO v1 sgRNA lentiviral library at
   a multiplicity of infection of 0.2, ensuring coverage of at least
   500-fold for each individual sgRNA represented in the cell population.
   Two days later, the selection of fully edited transduced cells was
   achieved by adding puromycin to the media at a final concentration of 2
   μg/ml for 7 days. Following a 2-day recovery in fresh media without
   selection, cells were split into two pools, in triplicate, at a cell
   density of 45 million cells (D[0]). The first pools were treated with
   MAF at its IC[25] (1.30 and 1.57 μM in Namalwa and Raji screens,
   respectively) and the second pool with DMSO as a vehicle. Cells were
   then cultured for 14 doubling times with puromycin at a concentration
   of 1 μg/ml. Cells were counted and passaged every 3 days at a cell
   density of 0.37 million cells/ml to maintain fold coverage of 500 cells
   per sgRNA, until D14. At each time point, cell pellets were frozen for
   subsequent genomic DNA purification. Genomic DNA isolation was
   performed as described in ([244]14). For NGS library preparation,
   sgRNAs were amplified from genomic DNA using two rounds of nested PCR
   with inner oligo and outer oligo annealing. The initial outer PCR was
   performed with TaKaRa ExTaq DNA Polymerase Hot-Start Version polymerase
   (Takara) using forward and reverse outer primers (table S4). PCR
   products were pooled, and ~ 2% of the input was amplified using Hot
   Start Q5 polymerase to add Illumina HiSeq adapter sequences (table S4).
   The resulting product from each pooled sample was further purified
   following separation on an 8% 0.5× Tris/borate/EDTA (TBE)
   polyacrylamide gel. The library NGS was quantified using qPCR and
   analyzed by deep sequencing on the HiSeq 2500 Illumina platform. Reads
   were trimmed of NGS adapter sequences using the Cutadapt tool. Reads
   were aligned to the sgRNA library index file using Bowtie2. BAM files
   were generated using samtools, and total read count tables were
   subsequently generated using the MAGeCK count command. DrugZ algorithm
   was used to identify gene knockouts which were depleted or enriched
   from D14 populations in comparison to D0 ([245]16).

Acquisition of immunoblotting images

   Immunoblot images were either acquired from film exposure or with an
   Azure Biosystems c300 Imaging System instrument. Briefly, membranes
   were exposed to Azure Radiance ECL (VWR) for 1 min before acquisition
   on the Azure apparatus with the sensitivity parameter set to “Normal”
   (1108 × 834). Whole pictures were adjusted for luminosity and contrast
   with the Adobe Photoshop software.

CRISPR-based competition growth assay

   Growth competition assays were conducted as described in ([246]22) and
   ([247]60). Briefly, RPE1-hTERT Cas9 p53^−/− were transduced with
   lentiviral particles of Lenti-mCherry-sgRNA-LacZ or Lenti-BFP-sgRNA-GOI
   (FIRRM/FIGNL1/FANCA/MRE11). Twenty-four hours after transduction, cells
   were selected with puromycin (15 μg/ml) for 4 days. BFP and
   mCherry-labeled cells were mixed at a 1:1 ratio and seeded in a
   six-well plate. At the initial time point (T0), cells were treated with
   cisplatin at 6 μM (IC[25]) or vehicle (DMSO), and the levels of each
   fluorophore were measured via flow cytometry. Cells were maintained
   under these conditions and subcultured for 20 days. The ratio of BFP to
   mCherry fluorescent cell population was assessed via flow cytometry
   every 5 days.

Apoptosis

   RPE1-hTERT Cas9 p53^−/− FIRRM^−/− clones 1 and 2 were seeded and
   treated with an overnight treatment of 4 μM cisplatin. The next day,
   cells were washed with Dulbeccos’s phosphate-buffered saline (D-PBS),
   and fresh medium was added to the plate. Forty-eight hours after
   cisplatin treatment, cells were processed for annexin V (BioLegend) and
   PI staining, following the manufacturer’s instructions. Analyses of
   annexin V and PI signals were done on at least 30,000 events acquired
   on a BD Fortessa (Becton Dickinson). Data were analyzed using the
   FlowJo software as previously described ([248]62).

Senescence-associated β-gal assay

   Senescence-associated β-gal (SA β-gal) assays were performed as
   previously described ([249]63). Briefly, shRNA expressing REP1 hTERT
   p53WT or p53^−/− or IMR90 cells were fixed with 0.5% glutaraldehyde in
   PBS, washed, and kept at 4°C in PBS supplemented with 1 mM MgCl[2] (pH
   6). Cells were stained with X-Gal solution containing potassium
   ferricyanide in PBS supplemented with 1 mM MgCl[2] (pH 6). Images were
   acquired, and the percentage of SA β-gal–positive cells was quantified.

Neutral comet assay

   Neutral comet assay was performed as previously described in ([250]14,
   [251]64, [252]65) and following the manufacturer’s specifications
   (Trevigen). Briefly, siRNA-treated U2OS cells were treated or not with
   250 nM cisplatin 24 hours after tranfection. The next day, cells were
   washed off twice with D-PBS before the addition of fresh media. Cells
   were then harvested at the indicated time point, resuspended at 1 ×
   10^5cells/ml in D-PBS, combined with prewarmed (37°C) molten LMAgarose
   at a ratio of 1:10, and evenly spread in CometSlides (Trevigen, catalog
   no. 4250-200-03). Comet slides were immersed into neutral lysis buffer
   overnight at 4°C and processed for electrophoresis and SYBR Gold
   (Invitrogen, catalog no. S11494) staining to reveal the comets as
   previously described ([253]14). Images were acquired with the EVOS FL
   Cell Imaging System microscope, and the tail moments were quantified
   using ImageJ with the OpenComet plug-in ([254]66). For each condition,
   at least 150 cells were analyzed.

Cell viability and cell growth assays

   siRNA-treated U2OS cells were exposed to different DNA-damaging agents
   as indicated. Forty-eight hours after the addition of treatment, cells
   were harvested and counted using a hemocytometer.
   Turbo-HA-2xStrep-FIRRM WT/∆WCF and GFP-FIRRM FL/∆WCF cell lines were
   supplemented with doxycycline either 6 hours after siRNAs transfection
   or 24 hours after seeding, and cell numbers were counted at the
   indicated time point following induction.

Immunofluorescence microscopy

   U2OS and U2OS 2-6-5 cells were grown on glass coverslips and fixed with
   2% (w/v) paraformaldehyde (PFA) in PBS for 20 min at room temperature
   or with 100% methanol for 20 min at −20°C (RAD51 time-course
   experiments). When indicated, PFA-fixed cells were permeabilized cells
   with 0.3% (v/v) Triton X-100 in PBS for 20 min at room temperature.
   Cells were then incubated with blocking buffer (10% goat serum, 0.5%
   NP-40, and 0.5% saponin in 1× PBS) for 30 min at room temperature and
   then incubated with the indicated primary antibodies (table S5) for 2
   hours at room temperature. After three washes with PBS, cells were
   incubated for 1 hour at room temperature with the indicated secondary
   antibodies (table S5) and 4′,6-diamidino-2-phenylindole (DAPI; 0.4
   μg/ml). Coverslips were mounted onto glass slides with ProLong Diamond
   Antifade Mounting agent (Invitrogen). Images were acquired using a
   Zeiss LSM900 laser-scanning microscope equipped with a 40× and 63× oil
   lens. In all micrographs, dashed lines indicate the nucleus outline,
   and insets represent a 10-fold magnification of the indicated fields.
   Each quantification was done on at least three biological replicates,
   and at least 100 cells or 50 foci were counted in each experiment.
   Unless stated otherwise, significance was assessed by performing an
   unpaired t test with Welch’s correction.

BrdU ssDNA detection assay

   BrdU ssDNA detection assay was done as previously described ([255]32).
   Briefly, siRNA-treated U2OS cells were incubated with 1 μM cisplatin
   and 10 μM BrdU (catalog no. [256]B23151, Thermo Fisher Scientific) for
   24 hours. Cell were then sequentially subjected to a nuclear
   preextraction buffer [10 mM Pipes (pH 7.0), 100 mM NaCl, 300 mM
   sucrose, 3 mM MgCl[2], 1 mM EGTA, and 0.5% Triton X-100] and a
   cytoskeleton stripping buffer [10 mM tris (pH 7.4), 10 mM NaCl, 3 mM
   MgCl[2], 1% Tween-20, and 0.5% sodium deoxycholate]. Following three
   washes in PBS, cells were fixed with 2% (w/v) PFA for 20 min on ice and
   then with 100% methanol for 5 min at −20°C. Cells were further
   permeabilized with 0.5% Triton X-100 for 10 min on ice before blocking
   with 3% bovine serum albumin (BSA) in PBS on ice for 1 hour. Cells were
   incubated with an anti-BrdU antibody overnight at 4°C. After three
   washes in PBS, cells were incubated with an anti-mouse Alexa Fluor 488
   for 1 hour at room temperature followed by DAPI (0.4 μg/ml) for 5 min
   at room temperature. Coverslips were mounted and imaged as described
   above.

mTurboID sample preparation for MS

   HEK293 Flp-In cells expressing mTurboID-tagged protein or mTurbo tag
   alone were seeded in 150 mm plates in technical duplicates. Induction
   of fusion protein and proximity biotinylation was conducted as
   previously described ([257]14, [258]62). Cells were harvested, washed,
   and processed for streptavidin pull-down as previously described
   ([259]14, [260]62). Beads were washed with radioimmunoprecipitation
   assay buffer and 50 mM ammonium bicarbonate (ABC; pH 8.0) before being
   resuspended in 100 μl of ABC. Trypsin digestion was performed on beads
   overnight at 37°C. The next day, supernatant-containing peptides were
   collected and pooled before lyophilization by vacuum centrifugation.

   Analysis of MS data was conducted on an Orbitrap Fusion (Thermo Fisher
   Scientific) as previously described with minor modifications ([261]62).
   Raw files were analyzed with the Comet, XTandem! and Mascot search
   engines using the human RefSeq database (version 20170518). To estimate
   the interaction statistics, we used SAINTexpress [([262]67) (version
   3.6.1)] on proteins with iProphet protein probability ≥ 0.9 and unique
   peptides ≥ 2. Each bait was compared against its respective negative
   control, which comprised of pulldowns from human embryonic kidney cells
   expressing the empty vector in triple technical replicates.
   Interactions displaying an averaged probability (AvgP) ≥ 0.7 were
   considered as statistically significant (table S3). Unfiltered
   contaminants, such as keratin, BirA*, and β-gal, were discarded in
   every bioinformatics analyses. Using the ProHits-Viz online tool
   ([263]prohits-viz.org), we generated a dot plot representing the
   relative AvgSpec of identified proteins for both baits.

Immunoprecipitation

   U2OS cells were transfected with the indicated siRNA duplexes for 24
   hours and then cotransfected with the indicated mCherry-LacR and/or 3×
   FLAG expression vectors. After 24 hours, cells were lysed in high-salt
   (HS) buffer (300 mM NaCl, 50 mM tris-HCl, 1 mM EDTA, and 1% Triton
   X-100) complemented with 1× cOmplete, EDTA-free Protease Inhibitor
   Cocktail (Roche), 20 mM N-ethylmethylamine, 1 mM NaF, and 0.2 mM
   Na[3]VO[4]. Cleared cell lysates were immunoprecipitated using either 1
   μg of the indicated antibody coupled to 40 μl of packed protein G
   Sepharose beads (catalog no. GE17-0618-01, Sigma-Aldrich) for 3 hours
   at 4°C. Beads were washed four times with HS buffer + 0.1% BSA and
   eluted in 2× Laemmli buffer for immunoblotting.

Protein expression and purification

   E. coli BL21RP were transformed with pGex6p2-GST-FIGNL1-His plasmid or
   pGex6p2-GST-FIRRM-His plasmids. Colonies were resuspended in 1 liter of
   LB with antibiotics and grown at 37°C until an optical density between
   0.4 and 0.6. The expression was induced using 0.1 mM
   isopropyl-β-d-thiogalactopyranoside and incubated overnight at 16°C.
   Bacteria were harvested by centrifugation, and the pellet was frozen on
   dry ice. Cells were lysed in PBS300 (PBS containing 300 mM NaCl), 1 mM
   dithiothreitol (DTT) and protease inhibitors, 0.05% Triton X-100,
   followed by four times 30-s on/off cycles of sonication on ice. The
   cell lysate was incubated with 1 mM MgCl[2] and benzonase (15 U/ml) at
   4°C for 45 min followed by centrifugation at 35,000 rpm for 1 hour. The
   soluble cell lysate was incubated with GST-Sepharose beads for 90 min
   at 4°C with gentle rotation. The beads were washed twice with PBS300
   followed by incubation with PBS300 containing 5 mM adenosine
   triphosphate and 15 mM MgCl[2], for 45 min at 4°C. GST-Sepharose beads
   were washed twice with PBS300 and once with P5 buffer [50 mM NaHPO[4]
   (pH 7.0), 500 mM NaCl, 10% glycerol, 0.05% Triton-X-100, and 5 mM
   imidazole]. The beads were incubated with PreScission protease (60
   U/ml) for 5 hours at 4°C in P5 buffer. The supernatant was collected
   and incubated with 400 μl of Talon beads for 1 hour at 4°C on gentle
   rotation. The beads were washed two times with P5 followed by two times
   with P30 (P5 buffer with a final concentration of 30 mM imidazole).
   Proteins were eluted twice with 200 μl of P500 (P5 buffer containing
   500 mM imidazole), dialyzed for 1 hour at 4°C against the storage
   buffer [20 mM tris-HCl (pH 7.4), 200 mM NaCl, 10% glycerol, and 1 mM
   DTT], and stored in aliquots at −80°C.

DNA-binding assays (EMSA) with fixation

   γ-^32P-labeled DNA oligonucleotides (100 nM; ssDNA–JYM 925, dsDNA–JYM
   925 + 945, SA–JYM 925 + 926, and 5′biotin–JYM 1086) [sequence of
   oligonucleotides (company) are indicated in table S4] were added to
   tris-acetate binding buffer [40 mM tris-acetate (pH 7.5), 20 mM NaCl, 2
   mM CaCl[2,] 0.02% Tween, and 1 mM DTT]. An increasing amount of
   purified FIGNL1 or FIRRM (0 to 25 nM) was added to the DNA-buffer mix
   and was incubated at 37°C for 15 min. The total reaction volume was 10
   μl. After incubation, the protein-DNA complexes were fixed with 0.2%
   glutaraldehyde for 15 min at 37°C. The reactions were subjected to
   electrophoresis at 150 V for 2.5 hours at room temperature on an 8% TBE
   acrylamide gel, and γ-^32P-labeled DNA was visualized by
   autoradiography.

Immunoprecipitation with recombinant protein

   Protein A/G Sepharose beads (Pierce) were blocked in
   immunoprecipitation buffer (IPB) buffer [20 mM KPO[4] (pH 7.4), 175 mM
   KCl, 0.5 mM EDTA, 0.5% NP-40, 10% glycerol, and 1 mM DTT] with BSA (1
   mg/ml) for 20 min at room temperature. Anti-RAD51 14B4 (Novus) was
   added for 20 min at room temperature followed by the addition of the
   purified proteins for 15 min. Immunoprecipitates were washed four times
   in 1 ml of IPB buffer and visualized by Western blotting using the
   indicated antibodies.

Cryo-EM sample preparation and data collection

   FIGNL1 samples (10 μM) in 1× PBS were incubated with 2 mM MgCl[2] and 1
   mM adenylyl-imidodiphosphate (AMP-PNP), immediately applied to graphene
   oxide lacey carbon grids (300 mesh), and vitrified using an FEI
   Vitrobot Mark IV with 100% humidity and 4°C settings. Samples were
   blotted for 1.5 to 2 s before being plunged into the liquid ethane.
   Initial screening and optimization of conditions were performed on an
   FEI Tecnai G2 F20 operated at 200 kV at facility for electron
   microscopy research (FEMR) (McGill University) using a Gatan 626 single
   tilt cryo-holder. Screening images were recorded in a Gatan Ultrascan
   4000, 4000 by 4000 charge-coupled device camera. Images taken at a
   nominal magnification of ×62,000 were imported into CryoSPARC (v4.1.1)
   ([264]68). Contrast transfer function (CTF) parameter estimation was
   performed using the Patch CTF estimation. Blob picker was used to pick
   particles out of 15 micrographs to generate two-dimensional averages.
   The FIGNL1 model was generated using the AlphaFold prediction for the
   FIGNL1 monomer and applying the hexameric symmetry seen in other AAA^+
   ATPases.

Acknowledgments