A genetic screen identifies FAN1, a Fanconi anemia associated nuclease necessary for DNA interstrand crosslink repair

Genome-wide shRNA screen to identify proteins necessary for resistance to interstrand crosslink (ICL) damage

To identify proteins involved in resistance to crosslinking agents, we screened U2OS cells transduced with a library of 74,905 retroviral shRNAs targeting 32,293 unique human transcripts (Figure 1A). Using competitive hybridization of probes that detect the half-hairpins of the shRNAs, we compared the relative abundance of shRNAs between untreated cells and cells treated with low levels of mitomycin C (MMC). Cells bearing shRNAs that conferred sensitivity were depleted from the treated population. About 2,173 hairpins targeting 2,017 genes conferred sensitivity to MMC using the criteria of an average loss of 2-fold from the treated population (log2>1) (Table S1). Among these were previously known DNA damage response proteins including BRCA1, TOPBP1, RAD18, RAD17, RAD51, RAD54, FANCE, and others. We employed the multicolor competition assay (MCA) (Smogorzewska et al., 2007) with 379 shRNAs against a selected group of genes that made the cutoff in the primary screen to retest for MMC sensitivity. Eighty-four shRNAs tested conferred MMC sensitivity (Table S2). Figure 1B shows the top 38 scoring hairpins in which resistance to MMC was below 80% of control shRNA. To further examine damage sensitivity, pools of siRNAs were also tested (Figure 1C). Based on the results of this assay and domain analysis, two genes, EXDL2 and KIAA1018, were chosen for further study.

EXDL2 - a putative 3′-5′ exonuclease necessary for resistance to MMC

EXDL2 is an uncharacterized 621 amino acid protein with a nuclease domain most similar to the WRN-exo domain in the WRN protein (Figure S1, panels A and B) (Perry et al., 2006). Based on the conservation of the four key negatively-charged residues (DEDD) that serve as ligands for the metal ions, as well as a tyrosine residue which has been shown to be important for the catalysis, EXDL2 is predicted to be a 3′-5′ exonuclease. A mutation in a D. melanogaster ortholog of EXDL2, CG6744, displays a phenotype of hyper-recombination (Cox et al., 2007). In human cells, depletion of EXDL2 using three different siRNAs led to sensitivity to MMC, the Topo1 inhibitor camptothecin (CPT) and the alkylating agent MMS (Figure 1D and S1, panel C). This spectrum of sensitivities is similar to mutants in the Fanconi anemia pathway. Therefore, we tested if FANCD2 ubiquitination was affected in the EXDL2-depleted cells. Based on the normal ubiquitination of FANCD2 before and after damage (Figure S1, panel D) we conclude that EXDL2 is either downstream of FANCD2 in the Fanconi pathway or in a parallel pathway of crosslink repair.

FAN1 (KIAA1018) is required for the resistance to crosslinking agents

A second protein with an interesting domain structure is KIAA1018, which we renamed FAN1 (Fanconi associated nuclease 1) based on the data presented below. FAN1 has a nuclease-like fold called DUF994 (later renamed VRR-NUC) at its C-terminus, potential DNA-binding (SAP) and protein-protein interaction (TPR) motifs in its mid-section, as well as a Rad18–like ZnF domain at the N-terminus (Kinch et al., 2005) (Figure 4A). Depletion of FAN1 with multiple siRNAs leads to sensitivity to crosslinking agents including MMC, chlorambucil, carboplatin, and oxaliplatin, as well as CPT, and MMS (Figure 2A and B).

To examine the evolutionary conservation of FAN1’s role in crosslink repair, we examined the crosslink sensitivity of a C. elegans FAN1 mutant (tm423) that carries a deletion of the SAP domain (Figure S2). fan-1 mutants lay normal numbers of eggs, and show normal larval development as well as no increase in either embryonic lethality or the % of males among their progeny that would suggest a meiotic phenotype (Figure 2D). However, treatment with either MMC or nitrogen mustard (HN2) results in decreased embryonic viability compared to wild type as judged by decreased hatching (p < 0.0001, respectively, chi-square test) (Figure 2E). Thus, FAN1 plays an evolutionarily conserved role in resistance to crosslink damage. Since the tm423 mutant may be a hypomorph, a meiotic phenotype could emerge in null animals.

FAN1 associates with mismatch repair proteins

To identify FAN1-associated proteins, a HA-tagged FAN1 was purified from 293TREX cells and interacting proteins were identified by LC-MS/MS. Proteomic data was processed using the Comparative Proteomic Analysis Software Suite (CompPASS) (Sowa et al., 2009). The top-scoring interacting proteins were the mismatch repair proteins MLH1, MLH3, PMS1 and PMS2 (Figure 3A). KIAA1018 was previously detected in a proteomic analysis of MLH1 interacting proteins (Cannavo et al., 2007); however, the interaction was not independently confirmed. To confirm these interaction, we performed immunoprecipitations. HA-FAN1 was identified in immunoprecipitates of MLH1, MLH3, and PMS2 (Figure 3B). Reciprocal immunoprecipitations identified all four proteins in precipitation reactions with anti-HA antibodies (Figure 3C). Endogenous FAN1 immunoprecipitations brought down both FAN1 and MLH1 in a FAN1-dependent manner (Figure 3D). The interaction with the mismatch repair machinery raised the possibility that FAN1’s role in ICL resistance is dependent on mismatch repair. To test this, we depleted FAN1 from the HCT116 cells, which carries inactivating mutations in both alleles of MLH1 (Figure 3E). FAN1 depletion with four different siRNAs still increased sensitivity of these cells to crosslinking agents.

FAN1 co-localizes to sites of damage with FANCD2 via its UBZ domain

The domain structure of FAN1 with its well conserved UBZ, SAP, and nuclease domains (Figure 4, panels A and B) led us to test FAN1 localization to sites of DNA damage. We subjected GFP-tagged FAN1 cells to laser microirradiation, which results in localized DNA damage tracks (Bekker-Jensen et al., 2006). Within 15 minutes of microirradiation, GFP-FAN1 localized to sites of DNA damage along with γ-H2AX, a marker of DNA damage (Figure 4C) (Rogakou et al., 1999). The GFP tracks were also seen in the absence of γ-H2AX staining (data not shown). We next asked if FAN1 localized to sites of crosslinked DNA damage caused by MMC. Most MMC-induced foci contained both FAN1 and FANCD2 (Figure 4D). We next tested various FAN1 domain mutants for their ability to localize to sites of laser microirradiation (Figure 5A). Mutants in the nuclease domain behaved like wild type. The N-terminal 373 amino acids lacking SAP or a protein with a SAP domain mutation (L477P) still localized to sites of damage but the strength of the GFP-FAN1 signal was substantially diminished. Only the mutant with two conserved cysteine residues of the UBZ domain substituted by alanines (C44A, C47A) completely failed to localize to microirradiation tracks (Figure 5B). The same UBZ mutant did not form foci in MMC-treated cells (Figure 5C) but was expressed at similar levels as the other alleles (Figure 5C, top panel “no triton”). Several other mutant alleles localized to sites of damage with reduced efficiency (Figure S3). Interestingly, the N-terminal 90 amino-acids containing the UBZ domain was able to co-localize with FANCD2, although with diminished efficiency. Based on these experiments we conclude that the UBZ domain is critical for localizing FAN1 to sites of damage. However, other parts of the protein also assist in localization. Careful kinetic analysis of the behavior of the different mutants will be necessary to resolve this issue.

FANCI and FANCD2 are required for FAN1 localization to sites of DNA damage

To examine possible dependency on the ID complex, we assessed formation of GFP-FAN1 foci after MMC treatment in cells depleted of FANCD2 and FANCI. In FANCI or FANCD2 depleted cells, FAN1 was no longer able to form foci (Figure 6A and S4B). The percentage of cells with foci in cells depleted of either FANCI or FANCD2 was 28% and 24%, respectively while 93% of control depleted cells showed FAN1 foci. FAN1 also fails to form foci in PD20 cells, which lack FANCD2 protein (Figure 6B), but reappear upon complementation with WT FANCD2. PD20 cells expressing the monoubiquitination-defective FANCD2 K561R mutant also fail to form FAN1 foci (Figure 6B). We next asked if FAN1 is required for FANCD2 monoubiquitination. Cells transfected with 4 different siRNAs against FAN1 were treated with MMC. FANCD2 in these cells was ubiquitinated to the same extent as in cells treated with siRNAs against luciferase (Figure 6C). We conclude that FA proteins and specifically FANCD2 monoubiquitination is necessary for the recruitment of FAN1 to sites of damage and that the crosslink sensitivity of FA defective cells is in part due to a failure to recruit FAN1 to sites of DNA damage.

FAN1 interacts with FANCD2

The dependence of FAN1 foci formation on the presence of monoubiquitinated FANCD2 raised a possibility that FANCD2 and FAN1 interact. Cells expressing GFP-FAN1 were subjected to immunoprecipitation with anti-FANCD2 antibodies. GFP-FAN1 strongly co-immunoprecipitated with FANCD2 (Figure 6D). Therefore, FAN1 and FANCD2 interact in vivo.

FAN1 has an endonuclease and 5′exonuclease activity

FAN1 contains a conserved nuclease domain. To assess its activity, we examined whether HA-FAN1 complexes immunoprecipitated from 293TREX cells could act on radiolabled substrates including 3′ and 5′ flaps, replication forks, and nicked substrates. Two nuclease activities were observed. One was a 5′ to 3′ exonuclease especially active on 3′FLAP substrate (Figure 7A, lane 8) as well as a nicked substrate (Figure 7A, lane 17). The second activity was an endonuclease activity on a 5′FLAP substrate (Figure 7A, lane 2), replication fork substrate (Figure 7A, lane 11) and nicked substrate (Figure 7A, lane 20). The endonuclease activity on the 5′FLAP substrate was seen on the top strand at the junction of the two DNA duplexes, not on the single strand flap itself. Neither the exonuclease nor the endonuclease activities were seen in the immunoprecipitates from cells expressing control HA protein or a mutant FAN1 with the key residues predicted to be necessary for the catalysis substituted to alanines (Q864A, D960A, E975A, K977A) (Figure 7A, lanes 3, 9, 12, 18, and 21), although the WT and mutant proteins were immunoprecipitated to the same extent (Figure S5A). To further show that this activity is intrinsic to FAN1 as opposed to a factor like MLH1/PMS2 associated with FAN1 in mammalian cells, we expressed the last 644 amino acids of human FAN1 (aa373-1017) as a His₆ fusion protein in bacteria (Figure 7B, Figure S5B and C). Indeed, the purified protein had the activities seen with the mammalian FAN1. More robust activity of FAN1 purified from bacteria gave us an opportunity to define the substrate specificity. Using a 3′-labeled substrate, we have confirmed that the exonuclease activity was a 5′- to 3′ activity (Figure 7B, lane 26). The major endonuclease activity of FAN1 was observed on the top strand of the 5′FLAP substrate at the junction of the DNA duplexes (Figure 7B, lane 8) and across from a nick (Figure 7B, lane 35). A weak 5′FLAP endonuclease activity was seen on the bottom strand (Figure 7B, lane 11). On a replication fork, the major activity was on a top strand (Figure 7B, lane 29) with some activity seen on the bottom strand (Figure 7B, lane 32). FAN1 activity was stimulated by ATP (Figure S5E). We noted a well-conserved motif in FAN1 (GFDQGIHGEGST, amino acids 826-837 of human protein) that may be able to bind to ATP. The significance of this motif remains to be determined. A mutant FAN1 with the key residues predicted to be necessary for the catalysis substituted to alanines (Q864A, D960A, E975A, K977A) (Figure 7B) or a mutant with just two (E975A, K977A) (Figure S5F) lacked the 5′-3′ exonuclease and the endonuclease activities. The bacterially-purified proteins did have some contaminating 3′ exonuclease activity (Figure 7B, lane 27). Based on these in vitro experiments, we conclude that FAN1 possesses an intrinsic nuclease activity that participates in DNA repair.

An shRNA screen identifies many putative players in crosslink resistance

Interstrand crosslinks are among the most lethal lesions to cells and their repair pathway(s) remain poorly understood. Given the difficulty that ICLs create during replication and transcription, it is critical to identify all the participants in the repair process. Therefore we performed an shRNA screen to identify new components of crosslink repair. Despite the limitations of RNAi, we were able to confirm a number of proteins involved in resistance to crosslinking agents using shRNAs and siRNAs. Among these were C4orf21, which has both a ZF-GRF, a presumed DNA binding domain, and a domain with similarities to helicases and FLJ25006, which has a kinase domain with predicted serine/threonine activity. Among the known DNA repair proteins was POLQ, an error prone translesion DNA polymerase (Arana et al., 2008; Seki et al., 2004). POLQ in chicken cells has been shown to be involved in repair of oxidative damage but not in crosslink repair (Yoshimura et al., 2006). However, in C. elegans, POLQ has been implicated in ICL repair, in a pathway distinct from the Fanconi anemia pathway but in the same pathway as the C. elegans BRCA1 ortholog (Muzzini et al., 2008).

Among the validated genes with strong phenotypes were two putative nucleases. We validated involvement of one, EXDL2, in crosslink sensitivity using multiple siRNAs. EXDL2 has a WRN-like exonuclease domain and promises to shed light on the mechanism of crosslink resistance. The other protein FAN1, with an intriguing domain structure, has now been placed as a bona fide nuclease necessary for crosslink repair in the Fanconi anemia pathway.

Evolutionary conservation of FAN1 function

FAN1 is easily identifiable in Dictyostelium discoideum, which has orthologs of several Fanconi anemia proteins (Zhang et al., 2009) but no obvious orthologs are apparent in Drosophila or Xenopus. Interestingly, an S. pombe ortholog of FAN1 has a SAP domain, TRP domain and a nuclease domain, thus it is predicted to also function in DNA repair transactions. Fission yeast does not possess the classical FA proteins except for an ortholog of FANCM, Fml1, which promotes Rad51-dependent gene conversion at stalled replication forks and limits crossing over during mitotic double-strand break repair (Sun et al., 2008). It will be interesting to test if the S. pombe FAN1 mutant is sensitive to crosslinking agents and if so, how it functions without the other proteins present in human cells. This could shed light on alternative pathways for crosslink repair in human cells.

Interaction of FAN1 with mismatch repair (MMR) proteins

We identified all human MutL proteins in nearly stoichiometric complexes with FAN1, suggesting a highly conserved function. Despite the strong interaction with MutLs, we have yet to identify a mismatch repair defect in extracts derived from cells depleted of FAN1 using shRNAs. This could be due to the limitations of RNAi to create the equivalent of a null mutation. Alternatively, the MutL complexes may play roles outside of mismatch repair.

During mismatch repair in bacteria, MutL complexes are recruited to sites of mismatches by MutS complexes that sense the mismatch lesions. MutL complexes then recruit the UvrD helicase and the MutH endonuclease, which it also activates, to initiate repair. MutL complexes have also been shown to have an endogenous endonuclease activity that could participate in mismatch repair (Kadyrov et al., 2006). It is possible that MutL complexes could play a similar role in ICL repair. In place of MutS complexes, the monoubiquitinated ID complex would play the analogous role as an ICL lesion sensor to recruit the FAN1-MutL complex to introduce one or more of the four incisions needed for ICL repair. In addition, this complex could recruit additional factors to aid in repair. The MutLα (MLH1-PMS2 heterodimer) complex has been identified as an interactor of FANCJ (BRIP1), a helicase involved in crosslink repair (Peng et al., 2007). A direct MLH1-FANCJ interaction was required for FANCJ to complement the crosslink sensitivity of a patient cell line with a mutation in the FANCJ gene. Thus, it is possible that FAN1 and MutLα might act as a bridge between the ID complex and FANCJ allowing the nuclease activity of FAN1 to pair with FANCJ’s helicase activity during the crosslink repair process. It will be important to determine if the interaction of MLH1 with FAN1 and FANCJ is mutually exclusive and whether FAN1 or MLH1 are required to localize FANCJ to the sites of crosslinks.

FAN1 localization to sites of crosslink damage and interaction with FANCD2

FAN1 localizes to laser microirradiation sites as well as to crosslink-induced damage foci where it co-localized with FANCD2. Localization of FAN1 to foci was dependent on the UBZ domain of FAN1, the presence of FANCI and FANCD2, and the monoubiquitination of FANCD2. Based on these findings, we hypothesize that FAN1 is recruited to sites of damage by the ID complex using monoubiquitinated FANCD2 and FANCI as an interaction platform for FAN1’s UBZ domain.

Since several mutant alleles of FAN1 examined, including the truncation allele lacking the SAP domain and the TPR domain, showed decreased GFP-FAN1 staining at the sites of DNA damage, it is very likely that in addition to the interaction between the ubiquitinated ID complex and the UBZ domain, FAN1 is stabilized in damage foci by interactions with the DNA or other proteins.

FAN1 was also identified by our group as a phospho-protein in a proteomic analysis of DNA damage response (Zhou and Elledge, in preparation). The identified phosphorylated SQ site, S210, in the human protein is evolutionarily conserved in mouse, chicken and fish and is the only SQ/TQ site in FAN1 with such high extent of conservation. Phosphorylation plays a pivotal role in activating FA pathway and it is likely that the activity of enzymes participating in repair will be tightly regulated. Therefore, it will be important to establish the functional consequences of abolishing FAN1 phosphorylation.

FAN1 is a nuclease in the Fanconi anemia pathway

Biochemical analysis of FAN1 isolated from human cells revealed that FAN1 acts as both an endonuclease and as an exonuclease. The major endonucleolytic activity of FAN1 appears to act opposite a nick and at branched structures with a 3′ end at the branch point (Figure 7C). FAN1’s minor activities are towards a 3′ FLAP and towards a replication forks across a 5′ end at the branch point. The 5′ to 3′ exonuclease activity is active on most 5′ ends although its strength varies depending on the specific substrate, with those substrates processed efficiently by the endonuclease activity being poorer substrates for the exonuclease than those inefficiently processed by the endonuclease. The ultimate test of FAN1’s activity will come from examining its function on crosslinked substrates in the setting of DNA replication, since this is where ATR is activated and where the interaction with the monoubiquitinated ID complex positions FAN1 at the sites of damage. However, the in vitro activities of FAN1 seen so far fit well with the repair activities hypothesized to be present at the crosslink and other lesions that necessitate replication restart (Figure 7D). FAN1 may be involved in the unhooking of the interstrand crosslink in the setting of replication. The ID complex is required for this process and since it is also necessary for FAN1 localization at the site of DNA damage, FAN1 is a good candidate for this activity. The exonuclease activity of FAN1 might also be important for the processing of the DNA strands to generate 3′ overhangs for the homologous recombinational repair that is necessary to restart the replication fork after the crosslink is removed or to gap repair unreplicated regions when both forks reach the ICL. Lastly, the final steps of removal of the fully unhooked crosslink might also rely on FAN1 activity. This is among the first instances for which we have an understanding of how a nuclease important for crosslink resistance localizes to sites of damage. Although other nucleases including XPF-ERCC1, MUS81-EME1 and SLX1 have been implicated in crosslink repair (Ciccia et al., 2008; Munoz et al., 2009; Svendsen et al., 2009) we still do not understand how they are recruited to the sites of DNA damage and the nature of their relationship to Fanconi anemia proteins.

FAN1 is a candidate tumor suppressor

Based on FAN1’s function in ICL resistance, co-localization, dependence on FANCI and FANCD2 function, as well as interaction with FANCD2, FAN1 is a candidate Fanconi anemia gene. Since three of the FA genes, FANCD1 (BRCA2), FANCN (PALB2), and FANCJ (BRIP1) are also mutated in familial breast cancer predisposition syndrome (Rahman et al., 2007; Wooster et al., 1995) FAN1 should be sequenced in the appropriate cohort of patients who display familiar predisposition to breast cancer but lack identified predisposing mutations.

The identification of FAN1 brings us closer to an understanding of the biochemical pathway involved in DNA crosslink repair and sets the stage for more precise examination of the repair process in reconstituted crosslink repair systems.

Cell Culture, plasmids, antibodies, RNAi, RT-qPCR

U2OS, DR-U2OS, HeLa, 293T, 293TREX, and HCT116 cells were grown in Dulbecco Modified Eagle medium (DMEM) supplemented with 10% (v/v) FBS (Invitrogen), 100 units of penicillin per ml, and 0.1 mg streptomycin per ml. PD20 cells were grown as above but with 15% FBS. Plasmids were constructed using recombinational cloning via the Gateway system (Invitrogen). KIAA1018 clone was obtained from Origene and the WT or truncation mutants were amplified and recombined into pDONR223 (Lamesch et al., 2007). pDONR223 derivatives were recombined into appropriate recipient vectors using LR clonase (Invitrogen). Mutagenesis was performed using multisite mutagenesis kit (Agilent) (primers are listed in Table S4). Antibodies against FAN1 were raised in rabbits against peptide CGQSDSAKREVKQKIS (YenZym) and affinity purified using the antigenic peptide. The other antibodies were FANCD2 (Novus NB100-182), GFP (Roche 11814460001), HA (Covance MMS-101R), MLH1 (Santa Cruz sc-582), MLH3 (Bethyl A301-849A & A301-850A), PMS1 (Santa Cruz sc-615), and PMS2 (BD Pharmigen 556415). siRNA transfections were performed using Lipofectamine RNAiMAX as suggested by the manufacturer with the final siRNA concentration of 50 nM. siRNA sequences are listed in Table S5. For RT-qPCR, Supperscript III reverse transcriptase followed by Platinum cybergreen super mix (Invitrogen) were used according to the instructions. GAPDH or Actin were used as controls.

Whole genome shRNA screen

The pool-based shRNA screen using half-hairpin (HH) barcode deconvolution was performed as described before (Schlabach et al., 2008).

Protein purification and mass spectrometry

293 TREX cells expressing FAN1 were lysed and immunoprecipitated using anti-HA antibodies (Sowa et al., 2009; Svendsen et al., 2009). Complexes were either used for DNA cleavage assays, subjected to immunoblotting, or eluted with HA peptide and trypsinized prior to mass spectrometry (Sowa et al., 2009; Svendsen et al., 2009). Processing of samples for mass spectrometry as well as analysis of proteomic data using CompPASS was as described (Sowa et al., 2009). For immunoprecipitations with FANCD2, chromatin fraction was prepared as described (Moldovan et al., 2009). For DNA cleavage assays, the immune complexes were washed 3 times in buffer containing 20 mM Tris HCl pH 8.0, 5 mM MgCl2, 1 mM DTT. Bacterial proteins (FAN1 373-1017) were expressed using pDEST17 (Invitrogen), induced with L-arabinose and purified on a Ni-NTA column. The proteins were dialyzed against a buffer containing 50 mM Tris HCl pH7.5, 100 mM NaCl, 0.01 %NP40, 10% glycerol, 1 mM DTT, 0.5 mM EDTA and were stored at −80°C.

Multicolor competition assay

Experiments were done as described (Smogorzewska et al., 2007).

Laser-induced damage and immunofluorescence

Microirradiation was performed as described previously (Bekker-Jensen et al., 2006). Immunofluorescence experiments were done as described (Smogorzewska et al., 2007).

In vitro cleavage assays

In vitro cleavage of DNA substrates was performed using the FAN1 immune complexes or bacterially-purified FAN1 in conjunction with previously described DNA substrates (Ciccia et al., 2003; Ip et al., 2008; Rass and West, 2006; Svendsen et al., 2009). DNA cleavage assays were performed using 5′-³²P or 3′ -³²P-end labeled DNA substrates. Substrates were generated by annealing oligonucleotides and were purified by polyacrylamide gel electrophoresis as described previously (Ciccia et al., 2003; Ip et al., 2008; Rass and West, 2006; Svendsen et al., 2009). The sequences of substrates are provided in Table S4. Radiolabeled substrates were incubated with the indicated immune complexes or bacterially-purified FAN1. DNA cleavage assays were performed in 20 mM Tris HCl pH 8.0, 5 mM MgCl2, 1 mM ATP, and 1 mM DTT. For bacterially-purified FAN1, 20 ng of WT protein or 40 ng of mutant protein was used with each substrate. After 30 minutes (for the bacterial protein) or 2 hours (for the immunoprecipitated proteins) at 37°C, reaction mixtures were treated with 1% proteinase K in SDS prior to electrophoresis on either 12% polyacrylamide gels (native) or 16% polyacrylamide-urea gels (denaturing). Reaction products were visualized by autoradiography.

C. elegans Genetics

C. elegans strains were cultured at 20°C under standard conditions (Brenner, 1974). The N2 Bristol strain was used as the wild-type background. The following mutations and chromosome rearrangements were used in this study: LGIII: him-18(tm2181) (Saito et al., 2009), qC1[dpy-19(e1259) glp-1(q339) qIs26] (III); LGIV: fan-1(tm423), nT1[unc-?(n754) let-? qIs50] (IV; V), nT1[qIs51] (IV; V).

The fan-1(tm423) mutant, obtained from the Japanese National Bioresource Project, carries a 411 bp in-frame deletion encompassing parts of exons 5 to 8. This deletion results in the loss of the predicted SAP motif.

DNA interstrand crosslink sensitivity assay in C. elegans

Young adult worms were treated with 0, 250 or 500 μM of mitomycin C (MMC; Sigma) in M9 buffer containing E. coli OP50 with slow shaking in the dark for 19 hours. Treatment with nitrogen mustard (mechlorethamine hydrochloride; Sigma) was similar, but with doses of 0, 50 or 100 μM. Following treatment with MMC or HN2, animals were plated to allow recovery for 3 hours. 20 animals were plated 5 per plate and hatching was assessed for the time period 22–26 hours from the start of treatment. Each damage condition was replicated at least three times in independent experiments. him-18/slx-4 mutants, shown previously to be extremely sensitive to ICL inducing agents (Saito et al. 2009), were used as a control. Since untreated him-18(tm2181) mutants have reduced hatching, embryonic viability after DNA damage treatment was plotted as a percentage of the hatching after DNA damage normalized by that in untreated animals (relative hatching) (Saito et al., 2009).