Abstract

   The catalytic cycle of topoisomerase 2 (TOP2) enzymes proceeds via a
   transient DNA double-strand break (DSB) intermediate termed the TOP2
   cleavage complex (TOP2cc), in which the TOP2 protein is covalently
   bound to DNA. Anticancer agents such as etoposide operate by
   stabilizing TOP2ccs, ultimately generating genotoxic TOP2-DNA protein
   cross-links that require processing and repair. Here, we identify RAD54
   like 2 (RAD54L2) as a factor promoting TOP2cc resolution. We
   demonstrate that RAD54L2 acts through a novel mechanism together with
   zinc finger protein associated with tyrosyl-DNA phosphodiesterase 2
   (TDP2) and TOP2 (ZATT/ZNF451) and independent of TDP2. Our work
   suggests a model wherein RAD54L2 recognizes sumoylated TOP2 and, using
   its ATPase activity, promotes TOP2cc resolution and prevents DSB
   exposure. These findings suggest RAD54L2-mediated TOP2cc resolution as
   a potential mechanism for cancer therapy resistance and highlight
   RAD54L2 as an attractive candidate for drug discovery.
     __________________________________________________________________

   RAD54L2 resolves TOP2-DNA adducts and protects cells from anti-cancer
   drugs such as etoposide.

INTRODUCTION

   DNA topoisomerases moderate DNA topology during replication,
   transcription, recombination, and chromatin remodeling ([56]1).
   Topoisomerase 2α (TOP2α) and TOP2β are unique among eukaryotic
   topoisomerases in their ability to form a transient DNA double-strand
   break (DSB) intermediate termed the TOP2 cleavage complex (TOP2cc), in
   which TOP2 is covalently bound to the DSB through a 5′-phosphotyrosyl
   bond ([57]2). Following DNA unwinding or untangling, a religation
   reaction occurs, producing a relaxed DNA product ([58]3). Widely used
   chemotherapeutic drugs, such as etoposide, inhibit DNA religation and
   form abortive TOP2ccs ([59]4–[60]6). TOP2ccs are primarily processed by
   tyrosyl-DNA phosphodiesterase 2 (TDP2), which hydrolyzes the
   5′-phosphotyrosyl bond between TOP2 and DNA ([61]7). This linkage is
   accessible after proteolytic degradation of the TOP2cc by the
   ubiquitin-proteasome system or specialized proteases ([62]8–[63]10).
   Alternatively, attachment of small ubiquitin-like modifier (SUMO)
   proteins to abortive TOP2ccs by zinc finger protein associated with
   TDP2 and TOP2 (ZATT/ZNF451) enables direct recruitment of TDP2 to the
   intact TOP2cc through TDP2’s split SUMO-interacting motifs (SIMs) to
   mediate TOP2 removal ([64]11). Ultimately, processing of abortive
   TOP2ccs converges in the release of a DSB that can be religated through
   canonical nonhomologous end-joining (NHEJ) ([65]12). Distinct from
   TDP2-mediated repair, abortive TOP2ccs can also be directly processed
   by specialized nucleases that “trim” DNA ends and lead to DSB repair by
   NHEJ or homologous recombination (HR) ([66]13, [67]14). The complexity
   of repairing TOP2-mediated DNA damage raises the prospect of as-yet
   uncharacterized proteins being involved in these events.

RESULTS

Focused CRISPR-Cas9 screens identify RAD54L2 as a factor whose loss confers
hypersensitivity to etoposide

   To identify factors involved in TOP2 biology, we performed CRISPR-Cas9
   gene inactivation screens in wild-type (WT) and tumor protein 53 (TP53)
   knockout (TP53^KO) human retinal pigment epithelial (RPE-1) cells using
   a focused dual-guide RNA library targeting 852 DNA damage response
   (DDR)–related genes ([68]Fig. 1A) ([69]15). In addition to identifying
   factors whose loss is known to cause etoposide hypersensitivity, such
   as TDP2, ZNF451, and ataxia telangiectasia mutated (ATM), we identified
   interesting candidates in both WT and TP53^KO RPE-1 cell screens
   ([70]Fig. 1B). These included RAD54 like 2 (RAD54L2) and ERCC excision
   repair 6 like 2 (ERCC6L2), two poorly characterized putative DNA
   helicases.

Fig. 1. Focused CRISPR-Cas9 screens identify factors whose loss imparts
etoposide hypersensitivity.

   [71]Fig. 1.
   [72]Open in a new tab

   (A) Layout of CRISPR-Cas9 screens in WT and TP53^KO RPE-1 cells
   [multiplicity of infection (MOI), drug concentration reducing cell
   viability by 50% (IC[50])]. Figure generated with BioRender. (B) Biplot
   showing gene enrichment scores (NormZ) upon etoposide treatment in WT
   (x axis) and TP53^KO (y axis) RPE-1 cells. Blue or orange circles
   indicate known or previously unknown factors, respectively, to affect
   etoposide sensitivity. (C to G) MTT cell viability assays of two HAP1
   RAD54L2^KO clones upon treatment with etoposide (C), camptothecin (D),
   ionizing radiation (E), ICRF-193 (F), or hydroxyurea (G). n = 3
   independent experiments. Bars represent means ± SEM.

   To validate these results, we tested the sensitivity of ERCC6L2^KO or
   RAD54L2^KO human HAP1 cells (fig. S1A) to etoposide and various other
   DNA damaging agents. In line with recent reports suggesting a role for
   ERCC6L2 in the DDR ([73]16–[74]19), HAP1 ERCC6L2^KO cells were
   hypersensitive to etoposide (fig. S1B), as well as to other DNA
   damaging agents, including camptothecin and ionizing radiation (fig.
   S1, C and D). In stark contrast, RAD54L2^KO clones were hypersensitive
   to etoposide compared to RAD54L2-proficient parental control (CTRL)
   ([75]Fig. 1C) but were not hypersensitive to the topoisomerase I poison
   camptothecin ([76]Fig. 1D), ionizing radiation ([77]Fig. 1E), or any of
   the other chemotherapeutic agents that we tested (fig. S1, E to H).
   Given that RAD54L2^KO cells were specifically sensitive to etoposide,
   we decided to focus on this factor for the rest of the study.

   We next performed viability assays to test the impact of RAD54L2 on
   cell viability in the presence of the TOP2 catalytic inhibitor ICRF-193
   ([78]6). We observed that, as reported for ZNF451^KO cells ([79]16,
   [80]20), RAD54L2^KO cells were not more sensitive to ICRF-193 than CTRL
   cells, suggesting that, like ZNF451, RAD54L2 does not play a prominent
   role in resolving lesions caused by this compound ([81]Fig. 1F). We
   also observed that RAD54L2^KO clones were not hypersensitive to
   hydroxyurea ([82]Fig. 1G), thus excluding a potential role in promoting
   replication fork reversal, as recently proposed for ZNF451 ([83]21). To
   extend our findings, we generated RAD54L2^KO clones in RPE-1 cells
   (fig. S1I) and confirmed their hypersensitivity toward etoposide
   treatment (fig. S1J). To validate our findings in a third cell line and
   with a different experimental approach, we performed competitive cell
   growth assays in human osteosarcoma (U2OS) cells stably expressing
   Cas9. Briefly, we transduced the cells with two different single guide
   RNAs (sgRNAs) targeting RAD54L2 and monitored the relative growth of
   the edited populations over time in the presence of selected doses of
   etoposide. We observed that sgRNA-mediated depletion of RAD54L2
   hypersensitized U2OS cells to etoposide treatment (fig. S1, K and L).

RAD54L2 binds SUMO and TOP2α/β and requires its SIMs to interact with
SUMO-modified TOP2α/β

   To further explore the link between RAD54L2 and TOP2 biology and
   identify RAD54L2 interactors, we stably expressed mCherry-RAD54L2 in
   RPE-1 cells grown in the presence or absence of etoposide and then
   performed immunoprecipitation experiments followed by mass spectrometry
   (IP-MS; fig. S2A). We observed that RAD54L2 interacts with TOP2α,
   TOP2β, SUMO1, SUMO2, and SUMO3 in untreated conditions ([84]Fig. 2A).
   Pathway enrichment analysis of enriched proteins using PANTHER GOslim
   biological process annotations indicated the most enriched biological
   process as “DNA unwinding involved in DNA replication” (table S4), a
   process that introduces topological stress and is often resolved by
   TOP2 enzymes ([85]22). Moreover, there was a substantial enrichment of
   interactors in the DNA replication pathway (fig. S2B). The interactions
   of RAD54L2 with TOP2α and TOP2β became further accentuated upon
   etoposide treatment ([86]Fig. 2B). In addition, pronounced interaction
   with the SUMO E3 ligase ZNF451 was observed only upon etoposide
   treatment. Collectively, the IP-MS experiments suggested an interplay
   between sumoylation and the ability of RAD54L2 to bind TOP2α, TOP2β,
   and ZNF451. In line with our data, recent studies have identified
   RAD54L2 as a strong SUMO interactor and predicted 12 potential SIMs in
   RAD54L2 ([87]23–[88]25).

Fig. 2. RAD54L2 binds SUMO and TOP2α/β and requires its SIMs to interact with
SUMO-modified TOP2α/β.

   [89]Fig. 2.
   [90]Open in a new tab

   (A and B) Volcano plots showing RAD54L2 interactors identified by IP-MS
   in untreated (UT) cells (A) or etoposide (ETP)–treated cells. (B) Green
   dots bordered by dotted lines indicating P values ≤0.05 and log[2] fold
   changes ≥2 are significant interactors. (C) Schematic of the putative
   domains and relevant properties of RAD54L2. The K311A catalytic site
   mutation is labeled, and the SIMs are highlighted in blue dots. (D)
   Clonogenic survival assays of control (CTRL) cells or RAD54L2^KO cells
   complemented with vectors expressing mCherry, mCherry-RAD54L2^WT,
   mCherry-RAD54L2^K311A mutant, or mCherry-RAD54L2^ΔSIM mutant; n = 3
   independent experiments. Bars represent means ± SEM. (E)
   Coimmunoprecipitation (IP) of mCherry from extracts of untreated or
   etoposide treated RPE-1 cells stably expressing mCherry (EV),
   mCherry-RAD54L2^WT, or mCherry-RAD54L2^ΔSIM mutant. This experiment was
   repeated six times with similar results.

   RAD54L2 has been described as a transcriptional coregulator and is a
   member of the RAD54 subfamily of sucrose nonfermentable (SNF2)–type
   chromatin remodeling factors ([91]26, [92]27). Previous biochemical
   characterization unveiled it as a DNA-dependent adenosine
   triphosphatase (ATPase) ([93]28). To dissect which properties of
   RAD54L2 are relevant in relation to TOP2-mediated DNA damage, we
   generated an ATPase catalytic site mutant (K311A), as previously
   characterized ([94]26, [95]29), and a ΔSIM mutant, point-mutated at the
   predicted SIMs identified in previous work ([96]Fig. 2C and fig. S2C)
   ([97]23). We monitored the etoposide sensitivity of RAD54L2^KO cells
   complemented with mCherry alone, mCherry-RAD54L2^WT, or
   mCherry-RAD54L2^ΔSIM by clonogenic survival assays ([98]Fig. 2D).
   Notably, while expression of RAD54L2^WT complemented the etoposide
   hypersensitivity of RAD54L2^KO cells—and actually made them more
   resistant than CTRL cells—ATPase catalytic site or ΔSIM RAD54L2 mutant
   proteins did not complement the etoposide sensitivity ([99]Fig. 2D).
   These data thus implied that the ATPase activity and SIMs of RAD54L2
   are crucial for its role(s) in responding to TOP2-mediated DNA damage.

   To explore these protein variants biochemically and to validate the
   IP-MS results, we coimmunoprecipitated mCherry from extracts of RPE-1
   cells stably expressing mCherry alone [empty vector (EV)],
   mCherry-RAD54L2^WT, or mCherry-RAD54L2^ΔSIM mutant and grown in the
   presence or absence of etoposide ([100]Fig. 2E). We observed that
   RAD54L2^WT interacted with TOP2α and TOP2β, with binding to high
   molecular weight modified forms of these proteins being dramatically
   enhanced upon etoposide treatment. By contrast, while RAD54L2^ΔSIM
   still interacted with lesser/unmodified forms of TOP2α and TOP2β,
   binding to etoposide-induced modified forms of TOP2α and TOP2β was no
   longer evident. Furthermore, we noted that, in line with our IP-MS
   data, binding of RAD54L2^WT to ZNF451 was stimulated upon etoposide
   treatment, although this was not the case for the RAD54L2^ΔSIM mutant.
   Knowing that TOP2α and TOP2β are expressed at different levels in
   different phases of the cell cycle, we tested whether RAD54L2
   interaction with TOP2 proteins is cell cycle regulated by performing
   IPs from extracts of G[1]- or S-phase synchronized cells. As expected,
   TOP2α expression was much lower in G[1] compared to S phase, meaning
   that we could only readily detect its interaction with RAD54L2 in S
   phase (fig. S2D). However, TOP2β expression was essentially equal
   throughout the cell cycle, allowing us to show that RAD54L2 interacted
   similarly with TOP2β in both G[1] and S phase cells (fig. S2D).
   Overall, these results indicated that RAD54L2 likely affects responses
   to TOP2ccs throughout the cell cycle.

RAD54L2 operates in a pathway dependent on ZNF451 and independent of TDP2

   Covalent TOP2-DNA adducts are known to be removed either directly,
   through the concerted actions of specialized nucleases, or in a
   TDP2-dependent manner. To explore where RAD54L2 acts in relation to
   known TOP2-DNA adduct repair processes and to unveil potential novel
   genetic interactions, we performed genome-scale CRISPR-Cas9 gene
   inactivation screens in RPE-1 TP53^KO (CTRL) and RAD54L2/TP53^KO cells
   (fig. S3A) in the presence or absence of etoposide. Ensuing
   bioinformatics analyses indicated that, as expected, RAD54L2 was a
   screen dropout hit in the CTRL background but not in the RAD54L2^KO
   background. Our analyses revealed that RAD54L2 was not epistatic with
   any other factors scoring in the screens and known to be involved in
   promoting repair of TOP2-DNA adducts, including TDP2, ATM, and NHEJ
   components such as NHEJ1, x-ray repair cross complementing 4 (XRCC4),
   and ligase 4 (LIG4) ([101]Fig. 3A; note that in the CRISPR library
   used, sgRNAs targeting ZNF451 were likely noneffective since they were
   not depleted upon etoposide treatment despite this factor’s
   well-established roles in the response to etoposide treatment). To
   validate the screen results, we generated human RPE-1 and HAP1 TDP2^KO
   and TDP2/RAD54L2^KO cells (fig. S3, B and C) and tested them for
   etoposide hypersensitivity via clonogenic survival assays and
   3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide
   (MTT)–based cell viability assays ([102]Fig. 3B and fig. S3D). In
   accord with the screening data, RAD54L2^KO and TDP2^KO cells showed
   mild and moderate etoposide hypersensitivity, respectively, while the
   double TDP2/RAD54L2^KO displayed a notable hypersensitivity phenotype
   indicating synergistic effects of TDP2 and RAD54L2. Similarly to the
   RAD54L2^KO cells, TDP2^KO and TDP2/RAD54L2^KO cells were not
   hypersensitive to ionizing radiation in either RPE-1 or HAP1
   backgrounds (fig. S3, E and F).

Fig. 3. RAD54L2 operates through a mechanism dependent on ZNF451 and
independent of TDP2.

   [103]Fig. 3.
   [104]Open in a new tab

   (A) Biplot showing gene enrichment scores (NormZ) upon etoposide
   treatment in TP53^KO (x axis) and TP53/RAD54L2^KO (y axis) RPE-1 cells.
   Orange dots relate to well-characterized DDR factors, most discussed in
   the text. (B) Clonogenic survival assays of CTRL, RAD54L2^KO, TDP2^KO,
   and TDP2/RAD54L2^KO RPE-1 cells upon treatment with etoposide. n = 3
   independent experiments. Bars represent means ± SEM. Competitive growth
   assays of RPE-1 CTRL (C), LIG41^KO (D), BRCA1^KO (E), and ZNF451^KO (G)
   cells transduced with virus expressing the indicated sgRNAs in the
   presence of various etoposide concentrations; n = 3 independent
   experiments. Bars represent means ± SEM. (F) Clonogenic survival assays
   of CTRL or ZNF451^KO transduced with a CTRL sgRNA or with an sgRNA
   targeting RAD54L2. n = 3 independent experiments. Bars represent means
   ± SEM. (H) Extracts of etoposide-treated CTRL or ZNF451^KO RPE-1 cells
   stably expressing mCherry [empty vector (EV)] or mCherry-RAD54L2^WT
   were subjected to mCherry IP followed by Western blotting for the
   indicated proteins. This experiment was repeated three times with
   similar results.

   To validate the genetic relationships between RAD54L2 and the HR and
   NHEJ pathways observed in our CRISPR screens, we performed competitive
   cell growth assays in LIG4^KO cells (fig. S3G) or breast cancer 1
   (BRCA1)^KO cells ([105]30), as models of NHEJ and HR deficiency,
   respectively. Thus, we transduced CTRL, LIG4^KO, or BRCA1^KO cells with
   lentiviral vectors containing either of two different sgRNAs targeting
   RAD54L2 and monitored the relative growth of the gene-edited
   populations over time in the presence of selected doses of etoposide.
   As expected from our CRISPR screen data, inactivation of RAD54L2
   sensitized CTRL, LIG4^KO, and BRCA1^KO cells to etoposide ([106]Fig. 3,
   C to E, and fig. S3, H, J, and K). Collectively, these experiments
   suggested that RAD54L2 operates in a hitherto unrecognized pathway for
   TOP2ccs.

   Having found that RAD54L2 likely interacts with sumoylated TOP2 via its
   SIMs, and considering that we observed the SUMO E3 ligase ZNF451 as a
   strong RAD54L2 interactor after etoposide treatment, we tested whether
   ZNF451 and RAD54L2 act in the same pathway. To monitor whether RAD54L2
   loss enhanced the etoposide sensitivity of CTRL or ZNF451^KO cells
   (fig. S3I), we performed clonogenic survival assays in CTRL or
   ZNF451^KO cells transduced with sgRNAs targeting RAD54L2. We observed
   that CTRL cells transduced with sgRNAs targeting RAD54L2 became
   hypersensitive to etoposide ([107]Fig. 3F). However, ZNF451^KO cells
   transduced with sgRNAs targeting RAD54L2 were not further sensitized to
   etoposide, implying an epistatic relationship between ZNF451 and
   RAD54L2. In addition, competitive cell growth assays in these same cell
   lines confirmed that sgRNA targeting RAD54L2 hypersensitized CTRL but
   not ZNF451^KO cells to etoposide ([108]Fig. 3G and fig. S3L).
   Collectively, these data supported a model in which RAD54L2 acts in
   processes controlled by the SUMO E3 ligase ZNF451 in response to
   etoposide.

   In line with the above findings, we observed that interactions between
   RAD54L2 and etoposide-induced posttranslationally modified forms of
   TOP2α and TOP2β were essentially abolished when cells were preincubated
   with the sumoylation inhibitor ML-792 (fig. S3M). Furthermore, by
   performing IPs of RAD54L2 from extracts of CTRL or ZNF451^KO cells, we
   observed that RAD54L2 was not able to effectively interact with
   etoposide-induced post-translationally modified TOP2α and TOP2β in the
   absence of ZNF451, while it still interacted with unmodified TOP2
   ([109]Fig. 3H). Collectively, these data are in accord with there being
   key functional connections between ZNF451 and RAD54L2 in response to
   etoposide treatment.

RAD54L2 counters TOP2ccs formation and prevents DSB generation upon etoposide
treatment

   To investigate cellular readouts of such a mechanism, we tested levels
   of chromosomal DSB formation and resolution in CTRL, RAD54L2^KO,
   TDP2^KO, and TDP2/RAD54L2^KO cells by analyzing formation of γH2AX foci
   (markers of DSBs) after etoposide treatment ([110]Fig. 4A). Notably,
   RAD54L2^KO cells started with significantly higher numbers of γH2AX
   foci than CTRL and TDP2^KO cells. Upon removal of etoposide, however,
   RAD54L2^KO cells resolved damage at a much faster rate than TDP2^KO
   cells, implying that RAD54L2 functions specifically on TOP2cc
   resolution and not DSB repair. The TDP2/RAD54L2^KO cells displayed
   higher levels of γH2AX throughout the experiment, reinforcing our
   conclusion that RAD54L2 and TDP2 act synergistically and in different
   pathways.

Fig. 4. RAD54L2 counters TOP2ccs by promoting their resolution.

   [111]Fig. 4.
   [112]Open in a new tab

   (A) Quantification of γH2AX foci in G[1] cells (cyclin A negative).
   Cells were treated with 30 μM etoposide for 30 min, washed, and left to
   recover for 4 or 8 hours. n > 3 independent experiments. Bars represent
   means ± SEM. One-way analysis of variance (ANOVA) was used for
   statistical testing. (B) DUST assays in CTRL or RAD54L2^KO HAP1 cells
   collected after 1 hour of treatment with 20 μM etoposide or 30 min or 1
   hour after etoposide washout. This experiment was carried out three
   times with similar results. (C) Quantifications of TOP2α or TOP2β
   signal from the DUST assays. Bars represent means ± SEM. (D) Proposed
   mechanism for RAD54L2 in TOP2cc resolution. UBC9, SUMO conjugating
   enzyme 2 ligase. Figure generated with BioRender. ns, not significant.

   We next used the detection of ubiquitylated and sumoylated
   TOP-DNA-protein cross-links (DPCs) (DUST) assays to monitor levels and
   SUMO posttranslational modifications of trapped TOP2α and TOP2β
   ([113]8). In this assay, only proteins covalently bound to DNA copurify
   with nucleic acids and are readily detected by Western blotting using
   antibodies targeting SUMO1, SUMO2/3, TOP2α, or TOP2β. We observed that
   in the presence of etoposide, levels of SUMO1- or SUMO2/3-conjugated
   DPCs and trapped TOP2β were higher in RAD54L2^KO cells compared to CTRL
   cells ([114]Fig. 4, B and C). Overall, these data thus supported a role
   for RAD54L2 in affecting TOP2cc levels.

DISCUSSION

   Collectively, our findings support a model in which RAD54L2 interacts
   with TOP2α and TOP2β, recognizes SUMO-modified TOP2ccs through its
   SIMs, and uses its ATPase activity to help resolve TOP2ccs and prevent
   DSB exposure. Furthermore, we have demonstrated that RAD54L2 acts with
   ZNF451 but in processes distinct from those mediated by TDP2, whose
   role in removing TOP2ccs is also promoted by ZNF451 activity. Our work
   thus identifies a key function for RAD54L2 and adds yet another layer
   of complexity to cellular mechanisms that safeguard the genome from
   TOP2-mediated DNA damage ([115]Fig. 4D), which occur not only in
   response to etoposide but also under normal physiological conditions.
   Since TDP2 gene mutations have been linked to hereditary neurological
   disease in individuals with seizures, ataxia, and intellectual
   disability ([116]31, [117]32), we speculate that RAD54L2 may protect
   against neurological disorders. In addition, in line with our finding
   that overexpression of RAD54L2 in WT cells generates etoposide
   resistance ([118]Fig. 2D), high expression of RAD54L2 was recently
   associated with inferior event-free and overall survival of acute
   myeloid leukaemia (AML) patients in the AML02 cohort ([119]33–[120]35).
   As RAD54L2-mediated TOP2cc resolution could represent a mechanism for
   intrinsic or acquired tumor resistance to chemotherapies including TOP2
   poisons, it represents a potential biomarker and a candidate target for
   anticancer drug discovery.

MATERIALS AND METHODS

Cell culture and cell line generation

   RPE-1 WT cells were originally obtained from J. Pines. RPE-1 TP53^KO
   cells were generated as described previously ([121]36). HAP1-Cas9 and
   RPE-Cas9 RAD54L2 knockout cells were generated by transduction of a
   RAD54L2 dual-guide RNA plasmid (guide A: ATTGGACAGCTGGTAGTAGG, guide B:
   GTACCCTCACCTCATCACCA). For genome-wide screens, RAD54L2 knockout cells
   were generated by transient transfection of a synthetic guide RNA
   (gRNA) targeting RAD54L2 (CTCTTCTTGCTGTGCTGCTT) in RPE-1 TP53^KO cells.
   TDP2 knockout cells were generated by transient transfection of a
   synthetic gRNA targeting TDP2 (GTAGAAATATCACATCT). LIG4 knockout cells
   were generated by transient transfection of a synthetic gRNA targeting
   LIG4 (GGAAGAATTAGTCTCATTGC) in RPE-1 TP53^KO cells. ERCC6L2^KO cells
   were purchased from Horizon (ID: HZGHC006416c004). RPE-1 cells were
   cultured at 37°C and 5% CO[2] in Dulbecco’s modified Eagle’s
   medium:nutrient mixture Ham’s F-12 (DMEM/F-12, Sigma-Aldrich)
   supplemented with 17 ml of 7.5% NaHCO[3] per 500 ml (Sigma-Aldrich),
   10% (v/v) fetal bovine serum (FBS; BioSera), penicillin (100 U/ml),
   streptomycin (100 μg/ml) (Sigma-Aldrich), and 2 mM l-glutamine.
   Blasticidin (10 μg/ml; Sigma-Aldrich) was used to select for
   Cas9-expressing cells. Cells were additionally cultured with puromycin
   (1.5 μg/ml) during selection of the transductants in the CRISPR-Cas9
   screens. HAP1 cells were cultured at 37°C and 5% CO[2] in Iscove’s
   modified Dulbecco’s medium (Thermo Fisher Scientific) supplemented with
   10% (v/v) FBS (BioSera), penicillin (100 U/ml), streptomycin (100
   μg/ml) (Sigma-Aldrich), and 2 mM l-glutamine. LentiX 293T and U2OS
   cells were cultured at 37°C and 5% CO[2] in DMEM (Sigma-Aldrich). RPE-1
   cells were synchronized in G[1] phase by treating with 5 μM palbociclib
   (S1116-SEL, Stratech Scientific Ltd.) for 20 hours. S-phase
   synchronization was obtained by treating with 2 mM thymidine (T1895,
   SLS Limited) for 20 hours and then releasing in fresh medium for 2
   hours. When indicated, 10 μM ML-792 (HY-108702, MedChemExpress) was
   added for 30 min before etoposide treatment.

CRISPR-Cas9 screens

   For DDR-focused CRISPR-Cas9 screens, biological duplicates of RPE-1 WT
   and TP53^KO Cas9-expressing cells were transduced at a multiplicity of
   infection (MOI) of 0.3- and 1000-fold coverage of the custom dual-gRNA
   DDR library as described previously ([122]15). Afterward, transductants
   were selected with puromycin for 12 days before treatment with dimethyl
   sulfoxide (DMSO) or etoposide [drug concentration reducing cell
   viability by 50% (IC[50])] for 7 days with three additional days
   without any drug. Library preparation and next-generation sequencing of
   the samples were performed as described previously ([123]15). Guide
   enrichment analysis was performed using DrugZ to compare DMSO-treated
   to etoposide-treated samples ([124]37).

   For the genome-wide CRISPR-Cas9 screens, RPE-1 TP53^KO or RPE-1
   TP53/RAD54L2^KO Cas9-expressing cells were transduced at an MOI of 0.2-
   and 500-fold coverage of the Gattinara library ([125]38). Afterward,
   transductants were selected with puromycin for 8 days and either left
   untreated or treated with etoposide (IC[30]) for 15 days. Genomic DNA
   from cell pellets was isolated using the QIAamp Blood Maxi Kit
   (Qiagen), and genome-integrated sgRNA sequences were amplified by
   polymerase chain reaction (PCR) using the Q5 Mastermix (New England
   Biolabs Ultra II) and i7 multiplexing barcoded primers. The final
   gel-purified products were sequenced on Illumina NovaSeq 6000 systems.
   Guide enrichment analysis was performed using DrugZ to compare
   DMSO-treated to etoposide-treated samples ([126]37).

MTT viability and clonogenic survival assays

   For MTT cell viability assay, 1000 HAP1 cells were plated in 96-well
   plates in triplicate and treated with the indicated drugs 24 hours
   later. Four days later, the medium was removed, and cells were
   incubated with MTT (0.5 mg/ml). Four hours later, 10% SDS was added,
   and absorbance was read the day after at 595 nm. For clonogenic assays,
   200 RPE-1 cells were plated in six-well plates, treated with drugs 24
   hours later, and stained and counted 10 to 12 days later, when visible
   colonies were formed. Cells were treated with etoposide (VP-16, Cayman
   Chemicals), carboplatin (C2538, Sigma-Aldrich), camptothecin (C9911,
   Sigma-Aldrich), 5-aza-2′-deoxycytidine (Merck, A3656), formaldehyde
   (Sigma-Aldrich, F1635), olaparib/AZD2281 (Stratech Scientific, A4154),
   ICRF-193 (GR332, BIOMOL International), or hydroxyurea (H8627,
   Sigma-Aldrich).

Competitive cell growth assay

   CTRL or ZNF451^KO human RPE-1 cells were transduced (MOI >1) with
   mCherry-sgLacZ or green fluorescent protein (GFP)–sgRAD54L2 (#1 or #2).
   Transduced cells were selected with puromycin 2 days after transduction
   and cultured for 2 more days until non-transduced cells were dead. In
   six-well plates, 50,000 cells expressing mCherry-sgLacZ were mixed with
   50,000 cells expressing GFP-sgRAD54L2 (#1 or #2) and plated with DMSO
   or etoposide. Cell mixtures were collected at the indicated days and
   subcultured (to prevent confluency) into fresh medium containing DMSO
   or etoposide. Cell mixtures were analyzed by fluorescence-activated
   cell sorting (Beckman Coulter, Cytoflex S) for cells expressing GFP or
   mCherry. The ratio of GFP^+/mCherry^+ cells was analyzed with FlowJo
   v9. To calculate the efficiency of indel formation after sgRNA
   transduction, DNA was extracted from cells 4 days after transduction
   and regions near the cut site were PCR-amplified, sequenced, and
   analyzed by TIDE (Netherlands Cancer Institute,
   [127]http://shinyapps.datacurators.nl/tide/).

Plasmid, siRNA transfection, and viral transduction

   The mCherry plasmid was obtained from VectorBuilder (VB200726-1045daz).
   Cloning of RAD54L2 cDNA into the mCherry vector was done by Gibson
   assembly (New England Biolabs) following PCR of RAD54L2 using the
   forward primer 5′-AGGATGACGATGACAAGAGCTCAGACGAATCTGCCTCAGG and reverse
   primer 5′-TCGAGGTCGACACGCGTGTTTTTCCCAGTGACCTCTATCAC and vector
   digestion with Hpa I/Afe I (New England Biolabs) restriction enzymes.
   RAD54L2 K311A mutant was generated by mutagenesis using primers
   5′-GAGATCACTTGCAAAGTTGCCCCCAGACCCATGCTGTG and
   5′-CACAGCATGGGTCTGGGGGCAACTTTGCAAGTGATCTC. RAD54L2 ΔSIM mutant (mutant
   details in table S2) was generated by gene synthesis (GeneWiz). The
   synthesized open reading frame was then cloned into the mCherry vector
   by Gibson assembly using the same conditions as above with the reverse
   primer 5′-CGAGGTCGACACGCGTGTTTTTCCCAGTGGCCTCTGCCGC. To generate stable
   cells, virus was first produced in LentiX 293T cells by cotransfecting
   the packaging constructs psPAX2 (Addgene, #12260) and pMD2.G (Addgene,
   #12259) with the plasmid of interest using TransIT-LT1 (Mirus Bio)
   according to the manufacturer’s protocol. Viral particles were then
   incubated with cells in the presence of polybrene (10 μg/ml) (Merck),
   followed by positive selection with geneticin (Gibco).

Immunofluorescence and microscopy imaging

   For imaging purposes, cells were plated in 24-well imaging plates,
   treated with etoposide (30 μM, 30 min), washed, and released for the
   indicated period of time. After treatments, cells were fixed in 4%
   paraformaldehyde for 10 min, permeabilized in phosphate-buffered saline
   (PBS)–0.2% Tween for 10 min, and blocked in PBS–5% bovine serum albumin
   (BSA) for at least 30 min. After 1 hour of incubation at room
   temperature with primary antibodies (γH2AX, Cell Signaling Technology,
   2577 and cyclin A, BD Biosciences, 611268), cells were washed in
   PBS–0.2% Tween three times and incubated with secondary antibodies for
   45 min. After three more washes in PBS–0.2% Tween, nuclei were stained
   with 4′,6-diamidino-2-phenylindole for 10 min. Images were acquired and
   analyzed using the Opera Phoenix microscope.

Immunoblotting

   Cells were lysed in Laemmli buffer [2% SDS, 10% glycerol, and 60 mM
   tris-HCl (pH 6.8)], incubated for 5 min at 95°C, and resolved by
   SDS–polyacrylamide gel electrophoresis (SDS-PAGE). After transferring
   onto a nitrocellulose membrane, the membranes were blocked with 5% milk
   or BSA in TBS-T buffer (tris-buffered saline with Tween 20, 0.1%) and
   incubated with primary antibodies diluted in 5% milk or BSA in TBS-T.
   The membranes were washed in TBS-T and probed with secondary
   antibodies. After secondary antibody incubation, membranes were washed
   in TBS-T, incubated with enhanced chemiluminescence (ECL) mixture for 5
   min in the dark, and developed using films or a ChemiDoc imaging system
   (Bio-Rad). Antibodies used were as follows: RAD54L2 (Abcam, ab86063),
   TOP2α (Bethyl Laboratories, A300-054A), TOP2β (BD Biosciences, 611493),
   mCherry (Abcam, ab167453), tubulin (Sigma-Aldrich, T9026), TDP2
   (Bethyl, A302-737A), ZNF451 (Sigma-Aldrich, SAB2108741), LIG4
   (ab193353, Abcam), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH)
   (ab8245, Abcam).

Immunoprecipitation

   For IP, around 2 million RPE-1 cells were plated in 15-cm dishes (2 per
   condition). Two days later, the cells were treated with 100 μM
   etoposide for 1 hour. After treatment, the cells were washed twice with
   cold PBS and lysed in 1.5 ml of IP buffer [20 mM tris-HCl (pH 7.5), 150
   mM NaCl, 2 mM MgCl[2], 10% glycerol, 0.5% NP-40, 20 mM N-ethylmaleimide
   (NEM), and EDTA-free protease and phosphatase inhibitors] and 15 μl of
   benzonase (Millipore) for 45 min. Lysates were centrifuged at 15,000
   rpm for 10 min, and supernatants were incubated with 20 μl of RFP-Trap
   magnetic beads (ChromoTek) for 2 hours at 4°C. The samples were washed
   4× with IP buffer and finally eluted in 40 μl of lithium dodecyl
   sulfate (LDS) buffer 2× + 1 mM dithiotheitol (DTT).

MS-based proteomics

   Bound proteins were eluted in NuPAGE LDS Sample Buffer (Life
   Technologies) supplemented with 1 mM DTT and boiled at 75°C for 15 min.
   The eluates were alkylated with 5.5 mM chloroacetamide for 30 min in
   the dark before being loaded onto 4 to 12% gradient SDS-PAGE gels.
   Proteins were stained with the Life Technologies Colloidal Blue
   Staining Kit and digested in-gel using trypsin. Peptides were extracted
   from the gel and desalted on reversed-phase C18 StageTips. Peptide
   fractions were analyzed on a quadrupole Orbitrap mass spectrometer (Q
   Exactive or Q Exactive Plus, Thermo Fisher Scientific) equipped with an
   ultrahigh-performance liquid chromatography system (EASY-nLC 1000,
   Thermo Fisher Scientific) as described ([128]39, [129]40). Peptide
   samples were loaded onto C18 reversed-phase columns (15 cm length, 75
   μm inner diameter, 1.9 μm bead size) and eluted with a linear gradient
   from 8 to 40% acetonitrile containing 0.1% formic acid over 2 hours.
   The mass spectrometer was operated in data-dependent mode,
   automatically switching between MS and MS^2 acquisition. Survey
   full-scan MS spectra [mass/charge ratio (m/z) 300 to 1700] were
   acquired in the Orbitrap. The 10 most intense ions were sequentially
   isolated and fragmented by higher-energy C-trap dissociation (HCD)
   ([130]41). An ion selection threshold of 5000 was used. Peptides with
   unassigned charge states, as well as with charge state less than +2,
   were excluded from fragmentation. Fragment spectra were acquired in the
   Orbitrap mass analyzer.

Analysis of MS data

   Raw data files were analyzed using MaxQuant (development version
   1.6.14.0) ([131]42). Parent ion and MS2 spectra were searched against a
   database, and peptide lists were searched against the human UniProt
   FASTA database released in February 2021 using the Andromeda search
   engine ([132]43). Spectra were searched with a mass tolerance of 6
   parts per million (ppm) in MS mode, 20 ppm in HCD MS^2 mode, strict
   trypsin specificity, and allowing up to three miscleavages. Cysteine
   carbamidomethylation was searched as a fixed modification, whereas
   protein N-terminal acetylation and methionine oxidation were searched
   as variable modifications. The dataset was filtered on the basis of
   posterior error probability to arrive at a false discovery rate (FDR)
   of below 1% estimated using a target-decoy approach ([133]44). The
   “match between run algorithm” in the MaxQuant quantification was
   enabled ([134]45). The MaxLFQ protein groups data calculated by
   MaxQuant were further analyzed with the Perseus (v1.6.13.0) of the
   MaxQuant computational platform ([135]46). Identified peptides were
   filtered for potential contaminants and reverse reads localization
   probability (>75%). Missing values were replaced by random numbers
   drawn from a normal distribution with a width of 0.3 and down shift of
   1.8. Protein groups with no proteotypic and less than two peptide
   identifications were excluded. Only protein groups identified in three
   of five replicates in at least one treatment group were included.
   Label-free protein quantitation (LFQ) was performed with a minimum
   ratio of 2 ([136]47). LFQ intensities normalized by the MaxLFQ
   algorithm were log[2]-transformed and calculated for the average
   log[2](FC) (fold change). P values were calculated by Student’s t test
   ([137]48). For RAD54L2 interactors, protein groups with fold change 2
   and P value ≤0.05 were considered significant interactors. Volcano
   plots were made in RStudio (v1.3.1093) with customized scripts. For
   pathway enrichment analysis, significant interactors (FDR < 0.05) from
   the untreated IP-MS experiments (RAD54L2-UT/RFP-UT) were used for a
   statistical overrepresentation test with the whole human genome as
   reference using PANTHER GOslim biological process annotations
   ([138]http://pantherdb.org/). Significant interactors linked to DNA
   replication in the untreated samples were identified as significantly
   enriched using the Reactome (v80) pathway browser
   ([139]https://reactome.org). Full results and enrichment analyses are
   available (table S3 and S4).

DUST assay

   The DUST assay ([140]8) is an extension of the RADAR assay developed by
   Maizels and colleagues for detecting proteins covalently bound to DNA
   ([141]49, [142]50). The DUST assay is an extension of the assay that
   also allows detection of posttranslational modifications on proteins
   covalently bound to DNA by using antibodies directed against the
   modifications. The RADAR purification scheme eliminates proteins that
   are not covalently bound to nucleic acids. For the experiments
   described here, cells were grown in 60 mM dishes to around 50 to 70%
   confluence. After specific treatment conditions as described, the
   medium was removed by aspiration and cells were resuspended in 600 μl
   of DNAzol (Invitrogen), which also contained 1× protease inhibitor
   cocktail (Sigma-Aldrich), 1 mM DTT, and 20 mM NEM (Sigma-Aldrich).
   After 10 min of incubation at 4°C with gentle rocking, 300 μl of
   ice-cold 100% ethanol was added to each plate and the plate was
   agitated until a white precipitate formed. The lysate was transferred
   to a microcentrifuge tube and centrifuged at maximum speed for 15 min
   at 4°C. The supernatant was removed, and the pellet was washed with 75%
   ethanol and spun for 2 min. All the supernatant was removed, and the
   pellet was air-dried briefly. The pellets were resuspended in 200 μl of
   tris-ethylenediaminetetraacetic acid (EDTA) (TE) buffer overnight at
   4°C. The following day, the samples were sonicated for 10 s two times
   using the Branson 450 Sonifier with a small probe and set at 30% power.
   The DNAs were quantitated using the Qubit dsDNA BR Assay. Then, 5 μg of
   DNA was digested by adding CaCl[2] to a final concentration of 5 mM and
   1 μl of micrococcal nuclease (Thermo Fisher Scientific) and incubating
   the samples for 30 min at 37°C digest DNA; RNA is also digested by
   micrococcal nuclease, eliminating the need for a ribonuclease step. The
   reactions were stopped by adding Laemmli buffer and boiled for 5 min.
   The samples were loaded on 4 different 4 to 15% TGX Mini-protean
   precast gels (Bio-Rad) and each indicated protein was probed on a
   separate blot. After wet transfer to polyvinylidene difluoride, the
   membranes were blocked and incubated with the indicated antibodies
   [TOP2α (Bethyl Laboratories, A300-054A), TOP2β (BD Biosciences,
   611493), SUMO2/3 (ab3742, Abcam), SUMO1 (Abcam ab11672)] overnight at
   4°C. The following day, the membranes were washed and incubated with
   secondary antibodies [GE Healthcare, NAV931V, ECL–anti-mouse
   horseradish peroxidase (HRP)–linked antibody; GE Healthcare, NAV9340V,
   ECL–anti-rabbit HRP-linked antibody] and visualized using Femto or Atto
   ECL reagents (Thermo Fisher Scientific).

Statistics and reproducibility

   Graphs and statistical tests were made using Prism v9 (GraphPad
   Software). One-way analysis of variance (ANOVA) was used to analyze the
   data in [143]Fig. 4A.

Figures and schematics

   Schematics were made using BioRender ([144]https://biorender.com/), and
   figures were made with Adobe Illustrator.

Acknowledgments