The structure-specific endonuclease complex SLX4–XPF regulates Tus–Ter-induced homologous recombination

Vertebrate replication forks arrested at interstrand DNA cross-links (ICLs) engage the Fanconi anemia pathway to incise arrested forks, ‘unhooking’ the ICL and forming a double strand break (DSB) that is repaired by homologous recombination (HR). The FANCP product, SLX4, in complex with the XPF (also known as FANCQ or ERCC4)–ERCC1 endonuclease, mediates ICL unhooking. Whether this mechanism operates at replication fork barriers other than ICLs is unknown. Here, we study the role of mouse SLX4 in HR triggered by a site-specific chromosomal DNA–protein replication fork barrier formed by the Escherichia coli-derived Tus–Ter complex. We show that SLX4–XPF is required for Tus–Ter-induced HR but not for error-free HR induced by a replication-independent DSB. We additionally uncover a role for SLX4–XPF in DSB-induced long-tract gene conversion, an error-prone HR pathway related to break-induced replication. Notably, Slx4 and Xpf mutants that are defective for Tus–Ter-induced HR are hypersensitive to ICLs and also to the DNA–protein cross-linking agents 5-aza-2′-deoxycytidine and zebularine. Collectively, these findings show that SLX4–XPF can process DNA–protein fork barriers for HR and that the Tus–Ter system recapitulates this process. Elango et al. identify a new class of substrates on which the Fanconi anemia SLX4–XPF nuclease operates to mediate homologous recombination at DNA–protein replication fork barriers and promote cellular tolerance of DNA–protein cross-links.

XPF introduces incisions either side of the ICL on one sister chromatid, generating a two-ended DSB opposite an 'unhooked' ICL on the unbroken sister chromatid 25,[28][29][30] . The single-stranded DNA gap opposite the unhooked ICL is filled by translesion synthesis DNA polymerases, and the DSB is repaired by conservative HR. Despite these mechanistic insights, it is unclear whether the full FA mechanism is restricted to ICL repair, or whether FA-mediated incisions can induce HR at other types of fork-stalling lesions such as DNAprotein complexes 31 . A molecularly defined system that recapitulates the FA mechanism in vivo on a mammalian chromosome has been lacking.
To study repair triggered by replication fork stalling at a defined chromosomal locus of cycling mammalian cells, we adapted the E. coli Tus-Ter replication fork barrier (RFB) for use in mammalian cells 32 . An array of six 23-bp (base pair) Ter repeats, when targeted to a specific genomic locus of mouse embryonic stem (mES) cells and bound by the Tus protein, triggers bidirectional fork stalling and HR at the Tus-Ter RFB. Tus-Ter-induced fork stalling also triggers formation of ~10-kb tandem duplications in Brca1-mutant cells, recapitulating the tandem duplication phenotype of BRCA1-mutant breast and ovarian cancers [33][34][35] . Thus, although it is an artificial site-specific DNA-protein fork barrier, the Tus-Ter RFB models the impact of certain endogenous fork-stalling lesions-presumably endogenous DNA-protein barriers-on DNA repair and genomic instability. Conservative HR at Tus-Ter, also called short-tract gene conversion (STGC), is a noncrossover pathway mediated by early components of the FA pathway and the canonical BRCA-Rad51 HR pathway 32,33,36,37 . By contrast, Tus-Ter-induced long-tract gene conversion (LTGC)-a replicative outcome related to break-induced replication-is independent of BRCA genes, Rad51 and the FA pathway 32 . Notably, HR induced at a site-specific DNA-protein fork barrier in Schizosaccharomyces pombe does not proceed via a DSB intermediate 38 . Thus, it is important to determine whether Tus-Ter-induced STGC entails an incision step with formation of a DSB intermediate and, if so, how the incision at Tus-Ter is controlled. To address these questions, we have studied the role of SLX4 in Tus-Ter-induced HR.
In this study, we use CRISPR-Cas9 genome editing to engineer precise, defined mutations in the Slx4 gene of mES cells that carry a reporter of stalled-fork-induced and DSB-induced HR. We systematically examine the role of individual nuclease-interacting domains of SLX4 in stalled fork HR. Our results identify a unique role for the SLX4-XPF complex in Tus-Ter-induced STGC and reveal a role for SLX4-XPF in preserving cell viability in response to chemicals that specifically induce DNA-protein cross-links (DPCs). We define a requirement for the SLX4 UBZ motifs in the recruitment of SLX4 to stalled forks, providing important insight into the mechanisms that connect clinically described SLX4 (FANCP) mutations to genomic instability. We also identify an unexpected role for SLX4 in supporting LTGC in the repair of replication-independent DSB.

Results
Slx4 regulates HR at Tus-Ter-stalled replication forks. To study the role of Slx4 in mammalian stalled fork repair, we used dual CRISPR-Cas9-mediated incisions to introduce defined deletions within the endogenous Slx4 gene of mES cells that contained a single copy of a 6xTer-HR reporter targeted to the Rosa26 locus on chromosome VI 32 . In this reporter, HR can be triggered by either Tus-Ter-induced fork stalling or by a site-specific DSB induced by the rare-cutting homing endonuclease I-SceI (Fig. 1a). The reporter enables simultaneous flow cytometric quantitation of error-free HR as STGC (GFP + RFP − ) and aberrant replicative HR responses as LTGC (GFP + RFP + ). We engineered a 125-bp frameshift deletion in exon 2 (Slx4 ∆125 ), thereby disrupting Slx4 early in the open reading frame (Fig. 1b). CRISPR-Cas9 targeting frequently generates a nonsynonymous deletion within the second allele of the target gene 37,53 . In recognition of this fact, we term the second allele in Slx4 ∆125 clones 'Slx4 ∆ ' . We assayed two independent Slx4-mutant clones (Slx4 ∆125/∆ clones 40 and 42) and two isogenic clones that received the same CRISPR-Cas9 treatment but retained wild-type Slx4 (Slx4 +/+ clones 3 and 13) (Fig. 1b). Both Slx4 ∆125/∆ clones expressed normal levels of Slx4 messenger RNA (mRNA) (Extended Data Fig.  1a), suggesting that Slx4 ∆125 may not be a null allele. Colony formation assays showed that Slx4 ∆125/∆ mutants were hypersensitive to the ICL-inducing agent mitomycin C (MMC; Fig. 1c).
To measure the impact of Slx4 disruption on stalled fork HR and DSB-induced HR, we transfected Slx4 ∆125/∆ Ter-HR reporter cells, in parallel, with Tus or I-SceI expression vectors and assayed repair products 72 h posttransfection (Methods). Slx4 ∆125/∆ clones showed an approximately fourfold reduction in Tus-Ter-induced STGC compared with isogenic wild-type clones, whereas Tus-Ter-induced LTGC was unaltered ( Fig. 1d-f). Hence, HR at Tus-Ter was skewed in favor of LTGC (Extended Data Fig. 1b). By contrast, I-SceI-induced STGC was unaltered in Slx4 ∆125/∆ cells, whereas I-SceI-induced LTGC was reduced (Fig. 1d,g,h). Consequently, the ratio of I-SceI-induced LTGC to total HR was reduced in Slx4 ∆125/∆ cells (Extended Data Fig. 1c). These results suggest that mammalian Slx4 has distinct roles in stalled fork HR (where it promotes STGC) and in conventional DSB-induced HR (where it mediates LTGC).
To avoid generating nonsynonymous Slx4 allelic deletions in subsequent experiments, we established an Slx4 haploid reporter cell line (here termed Slx4 +/− ). We used dual CRISPR-Cas9 incisions at exon 2 and exon 15 to delete one copy of the entire 19.3-kb gene (Extended Data Fig. 1d-f). We retrieved no Slx4 −/− clones from this experiment, despite an expected frequency of (60 / 262) 2 = ~13 / 262 (Extended Data Fig. 1d), suggesting that a true Slx4-null is lethal in mES cells. We used direct sequencing of breakpoint PCR products to identify the deletion breakpoint in Slx4 +/− cells and confirmed that the retained allele was wild type at the single guide RNA (sgRNA) target sites (Extended Data Fig. 1f). Slx4 expression in the Slx4 +/− clone was detected at wild-type levels, suggesting that transcription is upregulated in compensation for hemizygosity (Extended Data Fig. 1g). Slx4 +/− cells showed no increased sensitivity to MMC or to the PARP inhibitor olaparib (Extended Data Fig. 1h,i) and were phenotypically wild type for stalled-fork-induced and DSB-induced HR (Extended Data Fig. 1j).

The SLX4 UBZ domain recruits SLX4 to stalled forks for HR.
To study the SLX4 UBZ domain in Tus-Ter-induced HR, we used dual CRISPR-Cas9-mediated incisions to delete in-frame the UBZ1-coding and UBZ2-coding regions in Slx4 +/-Ter-HR reporter cells ( Fig. 2a and Extended Data Fig. 2a). UBZ-encoding mRNA sequences were undetectable in Slx4 ∆UBZ/cells, whereas sequences downstream (encoding the MLR, SAP and SBD domains) were expressed at normal levels (Extended Data Fig. 2b). Slx4 ∆UBZ/clones were hypersensitive to MMC (Extended Data Fig. 2c) but resistant to olaparib, hydroxyurea and camptothecin (Extended Data Fig.  2d,g,h; mES cells lacking Brca1 exon 11 served as positive controls in these colony formation assays). Notably, Slx4 ∆UBZ/cells were hypersensitive to 5-aza-2′deoxycytidine and zebularine, two agents that induce DPCs but not ICLs (Extended Data Fig. 2e,f) 54,55 . A characteristic feature of FA cells is enhanced S/G2 phase accumulation following exposure to MMC 10 . Untreated Slx4 ∆UBZ/cells had cell cycle distribution patterns similar to those of Slx4 +/cells (Fig. 2b). However, in response to titrated doses of MMC, Slx4 ∆UBZ/cells accumulated in G2/M phase more readily than Slx4 +/controls and showed a corresponding reduction in numbers in the S phase fraction (Fig. 2b).
We analyzed repair functions of six independent Slx4 ∆UBZ/clones and six independent isogenic Cas9-sgRNA-exposed Slx4 +/clones. Strikingly, Tus-Ter-induced STGC was reduced approximately sixfold in Slx4 ∆UBZ/clones in comparison with Slx4 +/controls (Fig. 2c). I-SceI-induced STGC was reduced to a lesser extent (Fig. 2d). Transient expression of wild-type human SLX4 restored Tus-Ter-induced STGC to wild-type levels in Slx4 ∆UBZ/clones 2 and 5 but had no effect on Slx4 +/controls (clones 25 and 44) (Fig. 2e). A UBZ 4C>A mutant that specifically disrupts UBZ domain function 24 failed to complement the STGC defect of Slx4 ∆UBZ/cells, despite equivalent levels of expression (Extended Data Fig. 2i,j). In the same experiment, expression of wild-type SLX4 or 4C>A had no statistically significant effect on I-SceI-induced STGC in either Slx4 ∆UBZ/or Slx4 +/clones (Fig. 2f). The ability of wild-type SLX4 but not the 4C>A mutant to complement the defect in Tus-Ter-induced STGC in Slx4 ∆UBZ/cells indicates a specific requirement for the SLX4 UBZ domain in error-free HR (that is, STGC) at stalled forks. By contrast, the failure of wild-type SLX4 to complement the modest defect in I-SceI-induced STGC in Slx4 ∆UBZ/cells provides no clear evidence of a role for the SLX4 UBZ domain in this function. Thus, Slx4 ∆UBZ is a separation-of-function allele that distinguishes control of STGC at stalled forks from STGC at replication-independent DSBs.
The same experiments revealed different effects on LTGC. Tus-Ter-induced LTGC was modestly reduced in the six Slx4 ∆UBZ/clones; hence, the ratio of LTGC to total HR was elevated, reflecting the severe defect in STGC (Extended Data Fig. 3a,b). Tus-Ter-induced LTGC was unaffected by expression of either wild-type SLX4 or the 4C>A mutant (Extended Data Fig. 3c). Thus, there is no clear evidence of a role for the SLX4 UBZ domain in Tus-Ter-induced LTGC. By contrast, I-SceI-induced LTGC was reduced more than twofold in the six Slx4 ∆UBZ/clones in comparison with Slx4 +/controls (Extended Data Fig. 3d); as a result, the ratio of LTGC to total HR was reduced approximately twofold, reflecting the specific reduction in I-SceI-induced LTGC (Extended Data Fig. 3e). Importantly, the LTGC defect was complemented by wild-type SLX4 but not by the 4C>A mutant (Extended Data Fig. 3f). Thus, Slx4 ∆UBZ is a separation-of-function allele that distinguishes control of LTGC at replication-independent DSBs from LTGC at stalled forks.
To facilitate detection of the endogenous Slx4 gene product in mES cells, we targeted the 3′ end of the residual Slx4 allele in Slx4 +/and Slx4 ∆UBZ/cells with a dual minimal auxin-inducible and small molecule-assisted shutoff (SMASh) degron containing an 8x HA epitope tag positioned N-terminal of the SMASh degron cleavage site 37,56-58 to generate Slx4 deg/and Slx4 ∆UBZdeg/clones (Fig. 2g). The Slx4 deg/and Slx4 ∆UBZdeg/clones expressed comparable levels of Slx4 mRNA (Fig. 2h). We detected nonspecific bands in the whole-cell extract; however, specific HA-tagged gene products were detectable in the chromatin fraction (Extended Data Fig. 2k,i). Therefore, we restricted our western blot analyses to chromatin-bound Slx4. Activation of the dual degron triggered partial degradation of endogenous SLX4, confirming the specificity of the anti-HA western blot signal. Unexpectedly, the HA-tagged Slx4 ∆UBZdeg gene product was more abundant in the chromatin fraction than the wild-type protein (Fig. 2i). We used these tools, in combination with ChIP, to study SLX4 recruitment to Tus-Ter-stalled forks. Strikingly, we noted robust accumulation of wild-type degron-tagged SLX4 (SLX4-deg) at the Tus-Ter RFB but reduced accumulation of degron-tagged SLX4-∆UBZ (SLX4-∆UBZdeg) (Fig. 2j). Thus, the SLX4 UBZ domain mediates Tus-Ter-induced STGC by recruiting SLX4 to the site of fork stalling.

Ter-induced HR.
To study the SLX4-MUS81 interaction in Tus-Ter-induced repair, we used Cas9 dual sgRNA incisions to generate in-frame deletions of Slx4 exons 11-13, encoding the SAP domain, in Slx4 +/-Ter-HR reporter cells ( Fig. 3a and Extended Data Fig. 4a). The mRNA sequence encoding the SAP domain was undetectable in Slx4 ∆SAP/cells, whereas expression of other Slx4 elements was at wild-type levels (Extended Data Fig. 4b). Degron-HA-targeted Slx4 ∆SAP/cells revealed normal levels of HA-tagged gene products (Extended Data Fig. 4c). Slx4 ∆SAP/cells were modestly more sensitive to MMC but not to olaparib, 5-aza-2′-deoxycytidine, zebularine, hydroxyurea or camptothecin (Extended Data Fig. 5a-f), in comparison with isogenic Slx4 +/cells that had been similarly treated with Cas9-sgRNA. We analyzed stalled fork and DSB repair functions in five independent Slx4 ∆SAP/clones and five isogenic Slx4 +/clones and found no effect of SAP domain deletion on any of the repair functions measured (Fig. 3b,c and Extended Data Fig. 4d,e). Thus, the interaction of SLX4 with MUS81 is dispensable for Tus-Ter-induced and I-SceI-induced HR.
To study the SLX4-SLX1 interaction in Tus-Ter-induced repair, we used Cas9 dual sgRNA incisions to generate in-frame deletions of Slx4 exons 13-15, encoding the SBD domain, in Slx4 +/-Ter-HR reporter cells (Fig. 3d and Extended Data Fig. 4f). The mRNA sequence encoding the SBD domain was almost undetectable in the Slx4 ∆SBD/cells, whereas expression of other elements of Slx4 was at normal levels (Extended Data Fig. 4g). Degron-HA-targeted Slx4 ∆SBD/cells revealed normal levels of HA-tagged gene products (Extended Data Fig. 4h). Slx4 ∆SBD/cells were insensitive to MMC, olaparib, 5-aza-2′-deoxycytidine, zebularine, hydroxyurea and camptothecin (Extended Data Fig. 5g-l) in comparison with isogenic Slx4 +/cells that had been similarly treated with Cas9-sgRNA. We analyzed stalled fork and DSB repair functions in six independent Slx4 ∆SBD/clones and six isogenic Slx4 +/clones and found no effect of SBD domain deletion on any of the repair functions measured (Fig. 3e,f and Extended Data Fig. 4i,j). Thus, the interaction of SLX4 with SLX1 is dispensable for Tus-Ter-induced and I-SceI-induced HR.

SLX4-XPF binding is required for Tus-Ter-induced STGC.
To study the SLX4-XPF interaction in Tus-Ter stalled fork repair, we used dual sgRNAs to generate in-frame deletion of the MLR-encoding exons 5 and 6 in Slx4 +/-Ter-HR reporter cells ( Fig. 4a and Extended Data Fig. 6a). Slx4 mRNA transcripts in Slx4 ∆MLR/clones lacked the deleted region but were detected at normal levels (Extended Data Fig. 6b). Degron-HA-targeted Slx4 ∆MLR/cells showed a modest reduction in levels of HA-tagged gene products (Extended Data Fig. 6c). Slx4 ∆MLR/clones were slow-growing compared with isogenic Slx4 +/controls that had been similarly exposed to Cas9-sgRNA but retained wild-type Slx4. In untreated cells, the proportion of cells in late S/G2/M phase was increased approximately twofold in Slx4 ∆MLR/cells in comparison with Slx4 +/cells (Fig. 4b). The S phase fraction was correspondingly reduced. Similarly, following exposure of cells to titrated doses of MMC, Slx4 ∆MLR/cells showed more dramatic late S/G2/M phase accumulation and S phase depletion than Slx4 +/controls (Fig. 4b). Slx4 ∆MLR/cells were hypersensitive to MMC, 5-aza-2′-deoxycytidine and zebularine but not to olaparib, hydroxyurea or camptothecin in comparison with Slx4 +/controls (Extended Data Fig. 6d-i). The slow growth and late S/ G2/M accumulation suggest that Slx4 ∆MLR/cells do not resolve replication stress efficiently. To explore this effect further, we analyzed metaphase spreads from Slx4 +/-, Slx4 ∆UBZ/and Slx4 ∆MLR/cultures that were either untreated or exposed to 20 ng ml −1 of MMC for 12 h. Both Slx4 ∆UBZ/and Slx4 ∆MLR/cells showed elevated levels of breaks and radial chromosomes in comparison with Slx4 +/controls, but Slx4 ∆MLR/cells had a more severe defect than Slx4 ∆UBZ/cells (Fig. 4ce). Thus, both Slx4 ∆UBZ/and Slx4 ∆MLR/cells displayed a classical FA chromosome breakage and/or aberration phenotype. In a second experiment, we analyzed Slx4 ∆SAP/and Slx4 ∆SBD/cells (Extended  Fig. 6j-l). Slx4 ∆SAP/cells were indistinguishable from Slx4 +/cells. Slx4 ∆SBD/cells had modestly increased frequencies of chromosome breaks and radial chromosomes compared with Slx4 +/cells.
We analyzed stalled-fork-induced and DSB-induced HR in six independent Slx4 ∆MLR/− clones and six independent Slx4 +/isogenic clones. Tus-Ter-induced STGC was almost undetectable in Slx4 ∆MLR/cells, whereas LTGC was detected at wild-type levels ( Fig.  5a and Extended Data Fig. 7a). I-SceI-induced STGC was reduced in Slx4 ∆MLR/clones, whereas I-SceI-induced LTGC was marginally reduced (Fig. 5b and Extended Data Fig. 7b). A caveat when interpreting these results is the possibility that the slow growth of Slx4 ∆MLR/cells might have artifactually altered the measured HR frequencies 59 . To address this concern, we performed complementation experiments in Slx4 ∆UBZ/-Ter-HR reporter cells, which exhibited no proliferation defect. In addition to testing full-length hSLX4 expression vectors, we generated 'mini-SLX4' constructs that encoded hSLX4 polypeptides truncated carboxy-terminal to the BTB domain and therefore lacked the MUS81-binding SAP and SLX1-binding SBD domains 28 . These mini-SLX4 constructs included wild type, 4C>A and ∆MLR variants (Fig. 5c). Following transient transfection, hSLX4 was expressed equivalently from each construct, although the wild-type mini-SLX4 protein was more abundant than other products (Fig. 5d,e). Expression of mini-SLX4 complemented the defect in Tus-Ter-induced STGC in Slx4 ∆UBZ/cells as efficiently as wild-type SLX4 (Fig. 5f). Importantly, both full-length SLX4∆MLR and mini-SLX4∆MLR failed to complement the defect in Tus-Ter-induced STGC in Slx4 ∆UBZ/cells, behaving similarly to 4C>A mini-SLX4. No significant effect of any of these constructs on Tus-Ter-induced STGC was observed in Slx4 +/cells, and I-SceI-induced STGC was unaffected by any SLX4 construct in either cell type (Fig. 5f,g). These findings show that intact SLX4 UBZ and MLR domains are independently required for Tus-Ter-induced STGC. The absence of interallelic complementation between endogenous Slx4 ∆UBZ and exogenous SLX4∆MLR alleles suggests that intact UBZ and MLR domains must each be present on the same SLX4 molecule for SLX4 to mediate Tus-Ter-induced STGC.
In the same experiments, Tus-Ter-induced LTGC was unaffected by any of the above-noted hSLX4 constructs in either Slx4 ∆UBZ/or Slx4 +/cells (Extended Data Fig. 7c). By contrast, I-SceI-induced LTGC, which was specifically impaired in Slx4 ∆UBZ/cells, was complemented by expression of full-length wild-type SLX4 (consistent with the results of our previous experiments) but not by the ∆MLR mutant (Extended Data Fig. 7d). Similarly, wild-type mini-SLX4 complemented the defect in I-SceI-induced LTGC, whereas mini-SLX4 mutants 4C>A and ∆MLR failed to do so. These results suggest that I-SceI-induced LTGC requires the same elements of SLX4 that are required for Tus-Ter-induced STGC.

The XPF nuclease domain in Tus-Ter-induced and I-SceI-induced STGC.
The above findings implicate the SLX4-XPF interaction in Tus-Ter-induced STGC. We therefore established tools to study XPF in stalled fork repair. We used CRISPR-Cas9 with dual sgRNA targeting to delete an entire ~42.3-kb Xpf allele in Ter-HR reporter cells (Extended Data Fig. 8a,b). Xpf expression was reduced by ~50% in Xpf +/cells compared with an isogenic Xpf +/+ clone, implying an absence of transcriptional compensation for hemizygosity (Extended Data Fig. 8c). Nonetheless, Xpf +/-Ter-HR reporter cells had wild-type levels of the XPF binding partner ERCC1, MMC and olaparib resistance, and repair frequencies (Extended Data Fig.  8d-f,g). (We were unable to detect the endogenous XPF protein in mES cells.) To determine whether the XPF nuclease domain contributes to stalled fork HR, we used Cas9 with dual sgRNAs to delete in-frame the nuclease-domain-encoding region of Xpf in Xpf +/cells (Fig. 6a,b). We obtained one Xpf ∆Nuc/clone containing an in-frame deletion (clone 1) and three that contained a frameshift (clones 2, 11 and 13). Analysis of Xpf expression in each mutant clone revealed the  absence of nuclease-domain-encoding sequences but normal levels of expression of regions encoding the upstream helicase domain and the downstream HhH2 domain (Extended Data Fig. 8h). Western blot analysis of the HhH2-binding XPF partner ERCC1 in whole-cell extracts revealed variably reduced abundance of ERCC1 in Xpf ∆Nuc/clones (Extended Data Fig. 8i). Chromatin association of ERCC1 was detected in Xpf ∆Nuc/clone 1 but not in the three Xpf ∆Nuc/frameshift clones, suggesting that interaction of XPF with ERCC1 is required for chromatin association of ERCC1 (Extended Data Fig.  8j). All four Xpf ∆Nuc/clones displayed a proliferative defect similar to that of Slx4 ∆MLR/cells, associated with an increase in the G2/M fraction and a reduction in the S phase fraction of unperturbed cells (Extended Data Fig. 8k). Similarly, following exposure of cells to titrated doses of MMC, Xpf ∆Nuc/cells showed more dramatic G2/M arrest and S phase depletion than Xpf +/controls. Xpf ∆Nuc/cells were hypersensitive to MMC, 5-aza-2′-deoxycytidine and zebularine and modestly sensitive to olaparib and camptothecin in comparison with Xpf +/controls ( Fig. 6c-f,h). Xpf ∆Nuc/cells were resistant to hydroxyurea treatment (Fig. 6g). We prepared metaphase spreads from Xpf ∆Nuc/or Xpf +/cells that were either untreated or exposed to 20 ng ml −1 MMC. Xpf ∆Nuc/cells showed dramatically increased frequencies of MMC-induced chromatid breaks and radials, consistent with the status of XPF/FANCQ as an FA gene (Fig. 6i-k). We analyzed stalled fork and DSB repair functions in each of the four Xpf ∆Nuc/-Ter-HR reporter clones and four isogenic Xpf +/clones. Similar to Slx4 ∆MLR/cells, Xpf ∆Nuc/cells showed near-complete loss of Tus-Ter-induced STGC (Fig. 6l) but no alteration in Tus-Ter-induced LTGC (Extended Data Fig. 8l). We also noted a significant reduction in I-SceI-induced STGC (Fig. 6m) and a modest reduction in I-SceI-induced LTGC (Extended Data Fig. 8m). Thus, Xpf ∆Nuc/cells phenocopy Slx4 ∆MLR/cells in stalled fork repair. The slow growth phenotype of Xpf ∆Nuc/cells suggests that these repair frequencies should be interpreted with caution. However, all of our experiments point to SLX4-XPF as a critical and specific mediator of Tus-Ter-induced STGC. Further, the Slx4 and Xpf mutants with the most severe FA phenotype exhibited the most profound impairment of Tus-Ter-induced STGC and were hypersensitive to DPC-inducing agents.

Discussion
We show here that the structure-specific endonuclease scaffold SLX4 has a critical and specific function in conservative HR (that is, STGC) at Tus-Ter-stalled forks. The SLX4 ubiquitin-binding UBZ and XPF-binding MLR domains are required for Tus-Ter-induced STGC, whereas the SAP and SBD domains are dispensable, excluding a requirement for the mitosis-associated SLX4 trinuclease complex. Taken together, our findings suggest that a major function of SLX4 in Tus-Ter-induced STGC is to position XPF at the stalled fork, promoting XPF-mediated incision of the bidirectionally Tus-Ter-arrested fork (Fig. 7). Such a function would recapitulate the critical role of SLX4-XPF in the 'unhooking' step of ICL repair 9,25,30 , with the exception that only one XPF-mediated incision would be required to generate a two-ended DSB at Tus-Ter because of the absence of a cross-link. This model explains why SLX4 plays a major part in Tus-Ter-induced STGC but not in I-SceI-induced STGC. In the former case, if SLX4-XPF coordination at the stalled fork were defective, production of a DSB at the stall site would be suppressed, resulting in low levels of HR. In the latter case, the DSB is induced by the I-SceI endonuclease, bypassing any requirement for SLX4-XPF in DSB formation. SLX4-XPF may have additional generic roles in HR, such as in flap-processing. Key components of the FA pathway required for efficient Tus-Ter-induced STGC include FANCM 37 , the FA core complex 33,37 , the BRCA-Rad51 pathway 32 and SLX4-XPF (this study). In light of these findings, it will be important in future to determine whether Tus-Ter-induced STGC requires TRAIP-mediated CMG ubiquitination 17 , monoubiquitination of FANCD2-FANCI 25 and/or a fork reversal step 23 . If, indeed, the full FA mechanism is required for Tus-Ter-induced STGC, this would extend the scope of the FA pathway beyond ICL repair and indicate that the FA pathway can also process some DNA-protein RFBs for HR 31 . In support of this idea, we found that cells defective for SLX4-XPF were hypersensitive to 5-aza-2′-deoxycytidine and zebularine. These cytidine analogs are incorporated into DNA and trap de novo DNA methyltransferases (DNMTs) in chromatin as covalent complexes (that is, DPCs) but do not form ICLs 54,55 . Importantly, the cellular toxicity of 5-aza-2′-deoxycytidine is caused by its capacity to cross-link DNA methyltransferases to DNA, not by downstream effects on DNA methylation 54 . Although it has been shown that the FA pathway is not necessary for repair of one class of DPCs, some DPCs encountered during replication might escape proteolytic mechanisms of DPC bypass and be processed by the FA pathway for HR 31,[60][61][62][63] . Experiments are underway to determine whether FA pathway components in addition to SLX4 and XPF are involved in 5-aza-2′-deoxycytidine resistance. In this regard, aldehydes, a class of chemicals to which FA mutant cells are hypersensitive, are known to form both ICLs and DPCs 6,12-15 . The Tus-Ter system may prove to be a useful model of FA-mediated repair of endogenous DPCs.
We have shown that a key function of the SLX4 UBZ domain is to support efficient retention of SLX4 at the stall site. As noted above, specific inactivation of UBZ domain function is a common defect associated with clinically described FANCP mutations 47,50 . Our data suggest that a key defect in FANCP-mutant cells of patients with FA cells is reduced retention of SLX4 at stalled forks. Unlike the true SLX4-null, which may be cell lethal when homozygous according to the present study and previous work 51,52 , germline homozygous SLX4 ∆UBZ mutations may be compatible with human development because the mutation is hypomorphic, reducing but not abolishing SLX4 action at stalled forks.
Unexpectedly, we identified SLX4 as a mediator of DSB-induced LTGC. This finding might appear to connect DSB-induced LTGC to mitotic DNA synthesis, which is mediated by SLX4 (ref. 64 ), MUS81 (ref. 65 ) and SLX1 (ref. 66 ). However, interaction of SLX4 with MUS81 or SLX1 was dispensable for DSB-induced LTGC; that is, there was no indication of a role for the mitotic SLX4 trinuclease complex. Instead, DSB-induced LTGC required the SLX4 UBZ and MLR domains, implicating ubiquitin chain binding and XPF in the mechanism. Further studies will be needed to decipher how SLX4-XPF-mediated incisions contribute specifically to DSB-induced LTGC.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41594-022-00812-9. Fig. 2 | The SLX4 uBZ domain promotes resistance to DPc-inducing drugs. a. Strategy for in-frame deletion of UBZ domain-encoding regions of Slx4. Red half-arrow heads: genotyping PCR primers. Gel shows PCR products using gDNA from Slx4 +/and Slx4 ∆UBZ/clones. Sequencing chromatogram shows in-frame breakpoint of Slx4 ∆UBZ allele. b. RT-qPCR analysis of mRNA encoding UBZ, MLR, SAP and SBD domains in Slx4 +/and Slx4 ∆UBZ/clones. Data normalized to Gapdh mRNA using the 2 -ΔCT method, mRNA expression in Slx4 ∆UBZ/samples were normalized to Slx4 +/of the same experiment. Each data point is an average of three technical replicates. Data shows mean of three independent biological replicates (n = 3). Analysis by unpaired Student's t-test (n = 3). Error bars: standard deviation. P = 0.0125 for mRNA expression of the UBZ region in Slx4 +/compared to Slx4 ∆UBZ/-. No significant differences were observed in expression levels of all other domains. c-h. Quantification of colony formation of Slx4 +/-, Slx4 ∆UBZ/and Brca1-∆exon11 clones in the presence of MMC (c), Olaparib (d), 5-aza-2′-deoxycytidine (e) Zebularine (f), Hydroxyurea (g) and Camptothecin (h). Data shows mean of values of three biologically independent replicates, n = 3. Analysis using Student's t-test. P-value *p < 0.05 and ***p < 0.001. Red asterisks refer to comparison between Slx4 +/and Slx4 ∆UBZ/clones; blue asterisks denote comparison between Slx4 +/and Brca1-∆exon11 clones and blue bracket with red asterisks denotes comparison between Slx4 ∆UBZ/and Brca1-∆exon11 clones. Error bars: s.d. i. RT-qPCR analysis of N-terminal Flag-tagged wild-type fulllength human SLX4 and UBZ 4 C > A expression plasmids 48 h after transfection. Data normalized to Gapdh mRNA using the 2 -ΔCT method and compared to empty vector (EV) control. Each data point is an average of three technical replicates. Data shows mean of three independent biological replicates (n = 3). Error bar: s.d. j. Western blot analysis of the chromatin-bound fraction of various human 3xFLAG-SLX4 full-length and UBZ 4 C > A transiently expressed for 48 hours and blotted with anti-FLAG antibody. k. Western blot analysis of whole cell extract (WCE) and chromatin-bound fraction of clones of Slx4 +/and Slx4 ∆UBZ/tagged with a dual degron containing a C-terminal 8xHA tag and untagged hemizygote cells (+/-). Asterisk (*) denotes nonspecific bands.