The FANCJ helicase unfolds DNA-protein crosslinks to promote their repair

Endogenous and exogenous agents generate DNA-protein crosslinks (DPCs), whose replication-dependent degradation by the SPRTN protease suppresses aging and liver cancer. SPRTN is activated after the replicative CMG helicase bypasses a DPC and polymerase extends the nascent strand to the adduct. Here, we identify a role for the 5 0 -to-3 0 helicase FANCJ in DPC repair. In addition to supporting CMG bypass, FANCJ is essential for SPRTN activation. FANCJ binds ssDNA downstream of the DPC and uses its ATPase activity to unfold the protein adduct, which exposes the underlying DNA and enables cleavage of the adduct. FANCJ-dependent DPC unfolding is also essential for translesion DNA synthesis past DPCs that cannot be degraded. In summary, our results show that helicase-mediated protein unfolding enables multiple events in DPC repair.


In brief
When a DNA replication fork encounters a covalent DNA-protein crosslink (DPC), SPRTN cleaves the protein adduct to promote replicative bypass. Yaneva et al.
show that the FANCJ helicase promotes SPRTN activity by unfolding the crosslinked protein. DPC unfolding by FANCJ also allows translesion DNA synthesis past non-degradable DPCs.

INTRODUCTION
To achieve faithful genome duplication, replisomes overcome myriad obstacles, including DNA-protein crosslinks (DPCs). 1,2 DPCs are generated by UV light, chemotherapeutics, and endogenous agents including abasic (AP) sites, formaldehyde, and enzymes such as HMCES and topoisomerases. 3 Experiments in yeast and frog egg extracts identified a pathway of DPC repair that is coupled to DNA replication and involves proteolysis of the protein adduct by a DNA-dependent metalloprotease called Wss1 in yeast and SPRTN in vertebrates. [4][5][6] This pathway, which is conserved in humans, [7][8][9][10][11] does not involve a double-strand break, reducing the risk of gross chromosomal rearrangements. Null mutations in SPRTN cause cell death, whereas hypomorphic germline mutations cause Ruijs-Aalfs syndrome, which involves genome instability, progeria, and a susceptibility to hepatocellular carcinoma. 12,13 Thus, SPRTNdependent DPC repair is critical for cell viability and suppression of human disease.
A model of replication-coupled DPC repair is emerging, primarily from studies in frog egg extracts and in vitro reconstitution. 5,6,[14][15][16] When the replicative CMG helicase collides with a DPC on the leading strand template, the nascent leading strand stalls $30 nucleotides from the adduct due to the footprint of the CMG helicase ( Figure 1A, cartoon), and the E3 ubiquitin ligase TRAIP ubiquitylates CMG. A few minutes after CMG stalls, it resumes translocation and bypasses the intact DPC, which allows extension of the leading strand to within 1 nucleotide of the DPC ( Figure 1A; ''À1''). CMG bypass depends on the 5 0 -to-3 0 helicase RTEL1, which translocates along the undamaged lagging strand template and thereby unwinds DNA beyond the lesion. RTEL1 depletion delays but does not abolish CMG bypass, suggesting that there could be additional backup helicases. In human cells, where RTEL1 mutations do not cause formaldehyde sensitivity, 17 there might be redundancy among such 5 0 -to-3 0 DNA helicases, as seen in worms. 18 How CMG can bypass a large adduct on the translocation strand is enigmatic, but we favor the idea that one of the interfaces in the CMG ring opens and allows the DPC to pass through the resulting gap. 14 The single-stranded DNA (ssDNA) generated downstream of the DPC recruits the E3 ubiquitin ligase RFWD3, which probably further ubiquitylates the DPC and promotes its destruction by the proteasome, which acts redundantly with SPRTN. 6,15 CMG bypass of the DPC is critical to trigger DPC proteolysis by SPRTN, probably because SPRTN activation depends on the leading strand being extended to within a few nucleotides of the DPC, which can only occur after CMG bypass. 6 In direct support of this idea, purified human SPRTN is most active when a DPC resides near DNA structures bearing single-and double-stranded features including ssDNA-double-stranded DNA (dsDNA) junctions. 16 Binding to specific DNA structures likely relieves the autoinhibition of SPRTN's protease domain by its DNA-binding domains. Whether any other events or factors are needed to activate SPRTN is unclear. (B) pDPC Lead was pre-bound with LacR to prevent the leftward replication fork from reaching the DPC and replicated in the indicated egg extracts containing 32 P[a]-dATP and supplemented with buffer, recombinant WT FANCJ, or ATPase mutant FANCJ-K52R, as indicated. At different times, DNA was extracted and digested with AatII and FspI, separated on a denaturing polyacrylamide gel, and visualized by autoradiography. The lower autoradiogram shows nascent leading strands generated by the rightward replication fork, and the upper autoradiogram shows leading and lagging extension products. Light blue bracket, CMG footprint (À30 to À37); orange bracket, products stalled at the adducted base (À1 to +1). The percentage of leading strands that approached from the À30 cluster to the À1 cluster was quantified, and the mean of n = 3 experiments is graphed. Error bars represent the SD. (C) Top: DNA structures generated by XmnI cleavage of pLacO 12 before and after forks progress through the LacR array. pLacO 12 was pre-incubated with LacR and replicated in the indicated egg extracts containing [a-32 P]-dATP. DNA was recovered, digested with XmnI, resolved by native agarose gel electrophoresis, and visualized by autoradiography.
(D) Quantification of the rate of linear product formation in the experiment shown in (C). Error bars represent the SD. See also Figure S1.

Article
After DPC proteolysis, the next step in DPC repair is translesion DNA synthesis (TLS), which extends the leading strand past the peptide adduct ( Figure 1A). 5 TLS is a two-step process in which DNA pol h inserts a nucleotide across from the peptide adduct, followed by strand extension beyond the lesion by a complex of REV1 and DNA pol z. 5,15 Both steps of TLS depend on RFWD3, whose binding to ssDNA ubiquitylates many proteins in the vicinity of the adduct. 15 Surprisingly, although DPC proteolysis normally precedes TLS, TLS still occurs, albeit slowly, when the DPC cannot be degraded. 14 How a TLS polymerase can accommodate a large, intact DPC in its active site and whether this scenario involves special requirements is unknown.
The fact that CMG bypass is not entirely abolished in the absence of RTEL1 14 raises the possibility that other 5 0 -to-3 0 DNA helicases participate in DPC repair. Among the seven vertebrate 5 0 -to-3 0 DNA helicases, FANCJ is of particular interest. Biallelic mutations in FANCJ cause Fanconi anemia, which is characterized by bone marrow failure, cancer predisposition, and sensitivity to bifunctional agents that induce DNA interstrand crosslinks (ICLs) and DPCs. 19 Current models suggest that FANCJ supports ICL repair primarily by promoting homologous recombination (HR), but its specific role in HR is unknown, and it has not been directly implicated in DPC repair. In apparently distinct functions, FANCJ resolves G4 DNA secondary structures to allow nascent strand progression at the replication fork (Sato et al. 20 and references therein), and it suppresses microsatellite instability. 21 Purified FANCJ displaces DNA-binding proteins from DNA, 22 but whether this function contributes to DNA replication is unclear. In summary, FANCJ appears to promote diverse genome maintenance pathways that are tied to DNA replication.
Here, we show that the residual DPC bypass and proteolysis observed in RTEL1-depleted egg extracts is further impaired upon co-depletion of FANCJ, demonstrating that FANCJ backs up this function of RTEL1. Similarly, FANCJ cooperates with RTEL1 to help the replisome overcome non-covalent nucleoprotein complexes. Strikingly, FANCJ depletion alone is sufficient to abolish DPC proteolysis by SPRTN, and this function of FANCJ is independent of its role in promoting CMG bypass of the adduct. Normal SPRTN activity is rescued by wild-type (WT), but not ATPase-deficient, FANCJ. In a reconstituted system, FANCJ's ATPase activity is also required for SPRTN-dependent cleavage of DPCs involving DNA-binding proteins. In this setting, FANCJ unfolds the protein adduct, which exposes the DNA underlying the DPC while also enabling DPC cleavage by SPRTN. In addition, we find that FANCJ's ability to unfold DPCs promotes TLS past non-degradable protein adducts. Together, our results identify FANCJ-dependent protein unfolding as a central event in replication-coupled DPC repair.

FANCJ backs up RTEL1 for CMG bypass of DPCs
We previously showed that depletion of RTEL1 from egg extracts delays but does not eliminate CMG bypass of DPCs, suggesting that other 5 0 -to-3 0 helicases might contribute to this process. 14 In addition to RTEL1, vertebrate genomes encode at least six 5 0 -to-3 0 helicases (FANCJ, DDX11, XPD, PIF1, SETX, and DDX3), and all except DDX11 are detectable on chromatin during DNA replication in Xenopus egg extracts. 6,14 To investigate whether these helicases cooperate with RTEL1 to promote CMG bypass, the advance of the leading strand from the À30 position to the À1 position was monitored as a readout of CMG bypass ( Figure 1A). Specifically, we replicated pDPC Lead (STAR Methods), a plasmid containing a site-specific M.HpaII DPC, in Xenopus egg extract depleted of RTEL1 alone or RTEL1 and another 5 0 -to-3 0 helicase. We included [a-32 P]-dATP to label nascent strands. To ensure that the DPC is always encountered on the leading strand template by a single rightward fork, we flanked the DPC with an array of Lac operators bound to Lac repressors (LacRs), which blocks arrival of leftward forks ( Figure 1A). To monitor nascent strand synthesis surrounding the DPC, we digested the DNA with AatII and FspI ( Figure 1A) and visualized the released nascent strands using denaturing urea gels and autoradiography. Release of the rightward leading strand by AatII allowed us to track its approach to the DPC; furthermore, the 3 0 overhangs generated by AatII created fully replicated AatII/FspI lagging strand digestion products that were a few nucleotides longer than the leading strand products, allowing us to distinguish the two ( Figure 1A, green versus red strands). This approach showed that, upon fork collision with the DPC, the nascent leading strands stalled at the À30 position ( Figure 1B, lane 1; Figure 1A, left cartoon). 5,14 CMG then bypassed the intact DPC, allowing extension of the leading strand to the À1 position Figure 2. FANCJ is required for DPC proteolysis (A) pDPC2x Lead was replicated in the indicated egg extracts supplemented with buffer, recombinant WT FANCJ, or ATPase mutant FANCJ-K52R. At the indicated times, plasmid was recovered under stringent conditions, the samples were split and either mock treated (upper panel) or treated with the deubiquitylating enzyme Usp21 (lower panel), followed by DNA digestion and blotting for HpaII. Signal from the Usp21-treated sample was quantified, and peak signal was assigned a value of 100%. The mean of n = 3 independent experiments is graphed. Error bars represent SD. (B) pmeDPC2x Lead was replicated in the indicated egg extracts. At the indicated times, plasmid was recovered under stringent conditions, followed by DNA digestion, and the resulting samples were blotted for HpaII. (C) pmeDPC2x Lead was replicated in the indicated egg extracts supplemented with buffer, recombinant WT FANCJ, or ATPase mutant FANCJ-K52R and HpaII was analyzed as in (A). (D) pmeDPC ssDNA was incubated directly in the indicated nucleoplasmic egg extract (NPE) supplemented with buffer, recombinant WT FANCJ, or ATPase mutant FANCJ-K52R without prior licensing in high-speed supernatant (HSS) to prevent replication initiation. 23 Plasmid was isolated and blotted for HpaII as in (A). (E) pAP ssDNA was incubated in the indicated NPE, which caused HMCES crosslinking to the AP site. Plasmid was isolated by the stringent pull-down procedure as in (A), but the resulting samples were blotted for HMCES instead of HpaII. (F) pAP ssDNA was incubated in the indicated NPE supplemented with MG262 and buffer, recombinant WT FANCJ, or ATPase mutant FANCJ-K52R without prior licensing in HSS. Plasmid was isolated and analyzed as in (E). See also Figure S2. and progression of the nascent lagging strand past the crosslink, as seen from the appearance of the larger AatII/FspI product ( Figure 1B, lanes 2 and 3; Figure 1A, middle cartoon). Arrival of the leading strand at the crosslink triggered DPC proteolysis, and subsequent TLS allowed extension of the nascent leading strand beyond the adduct, as seen from appearance of the smaller AatII/FspI product ( Figure 1B, lanes 3-6; Figure 1A, right cartoon). 5,6,14 Depletion of SETX, PIF1, or DDX3 alone or in combination with RTEL1 had no significant effect on CMG bypass of DPC Lead , as seen from timely extension of the leading strand to the À1 position (data not shown; we did not examine XPD or DDX11). Immunodepletion of FANCJ alone ( Figure S1A, lane 3) also did not impact CMG bypass ( Figure 1B, lanes 13-18; see graph for quantification). However, depletion of both FANCJ and RTEL1 ( Figure S1A, lane 4) led to a substantial further delay in bypass compared to depletion of RTEL1 alone ( Figure 1B, lanes 7-12 versus [19][20][21][22][23][24]. The kinetics of CMG bypass were rescued to the level of RTEL1-only depletion by WT recombinant FANCJ (rFANCJ WT ), but not an ATPase-deficient FANCJ mutant (rFANCJ K52R ) ( Figure 1B, lanes 25-36; Figures S1A and S1B). These results show that FANCJ can partially substitute for RTEL1 in promoting CMG bypass of a DPC.

FANCJ backs up RTEL1 in progression through a LacR array
In addition to promoting CMG bypass of DPCs, RTEL1 is required for efficient replisome progression through non-covalent LacR-DNA complexes. 14 To address whether this process also involves FANCJ, we replicated a plasmid containing an array of 12 lacO repeats bound by LacR. DNA was recovered at various time points and digested with XmnI, followed by native gel electrophoresis. Replication forks initially converged at the outer edges of the LacR array, generating a discrete X-shaped replication intermediate that was subsequently converted to a linear DNA species when converging forks met ( Figure 1C, cartoon and lanes 1-6). As we showed previously, RTEL1 depletion from egg extract slowed the appearance of linear species, indicating its requirement for efficient replisome progression through the LacR array ( Figure 1C, lanes 1-6 versus 7-12; see Figure 1D for quantification) 14 . Immunodepletion of FANCJ from egg extract had no significant effect on accumulation of linear molecules ( Figure 1C, lanes 13-18; Figure 1D), but it enhanced the defect seen in RTEL1-depleted extract ( Figure 1C, lanes 19-24; Figure 1D). This delay was fully rescued to the level seen in RTEL1-depleted extract by rFANCJ WT , but not rFANCJ K52R ( Figure 1C, lanes 25-36; Figure 1D). The same result was observed when fork progression was examined at higher resolution using urea PAGE gels ( Figure S1C). We conclude that FANCJ backs up RTEL1 to promote efficient helicase progression past covalent and non-covalent proteinaceous barriers.

FANCJ promotes DPC proteolysis
We previously showed that CMG bypass is a prerequisite for efficient DPC proteolysis. 14 Given that FANCJ depletion impairs DPC bypass in RTEL1-depleted extract ( Figure 1B), we expected that FANCJ depletion would further compromise DPC proteolysis in the absence of RTEL1. To test this idea, we replicated a plasmid containing two closely spaced leading strand DPCs (DPC Lead ) in extract that was depleted of FANCJ, RTEL1, or both. To monitor degradation of the DPC during DNA replication, we isolated the plasmid, digested DNA, and blotted for HpaII. We also treated the digested chromatin with the deubiquitylating enzyme Usp21, which collapses ubiquitylated M.HpaII into a single band for easier quantification. As expected, 14 DPC degradation was delayed in extracts depleted of RTEL1 ( Figure 2A, lanes 1-4 versus 5-8). Importantly, DPC proteolysis was more severely inhibited in extract co-depleted of RTEL1 and FANCJ (Figure 2A, lanes [13][14][15][16]. This defect was rescued to the level seen in RTEL1-only depletion by the addition of rFANCJ WT , but not rFANCJ K52R ( (legend continued on next page) ll OPEN ACCESS Article proteasome, 6 FANCJ depletion enhanced the proteolysis defect observed in RTEL1-depleted extract, suggesting that FANCJ contributes to a fully functional proteasome pathway (Figure S2A). The same additive effect of combined depletion was observed in extracts supplemented with proteasome inhibitor, implicating FANCJ in SPRTN-mediated DPC destruction (Figure S2B). Thus, in the absence of RTEL1, FANCJ appears to stimulate both sub-pathways of DPC proteolysis.

FANCJ is required for SPRTN activity independently of DPC bypass
To further investigate the involvement of FANCJ in the SPRTN pathway, we examined replication-coupled degradation of a DPC whose lysine residues have been chemically methylated (meDPC). The meDPC cannot undergo ubiquitylation and thus cannot be degraded by the proteasome, but it is still susceptible to SPRTN-mediated degradation, which yields a discrete HpaII fragment ( Figure 2B, lanes 1-5). 6,14 As we reported before, RTEL1 depletion only slightly delayed the appearance of the SPRTN-dependent HpaII fragment, consistent with its partial effect on CMG bypass ( Figure 2B, lanes 6-10). 14 In contrast, FANCJ depletion alone or in combination with RTEL1 depletion abolished SPRTN-dependent meDPC proteolysis ( Figure 2B, lanes [11][12][13][14][15][16][17][18][19][20], and this defect was reversed by rFANCJ WT , but not rFANCJ K52R ( Figure 2C). Thus, even in the presence of RTEL1, FANCJ is essential for SPRTN activity. FANCJ depletion abolished SPRTN activity ( Figure 2B) but had no effect on CMG bypass of DPCs ( Figure 1B), suggesting that FANCJ promotes SPRTN activity independently of DPC bypass. To test this idea, we exploited the fact that SPRTN can be activated independently of DNA replication or CMG bypass if the DPC is positioned within a ssDNA gap. In this setting, the 3 0 end flanking the gap was extended toward the DPC, triggering SPRTN activity ( Figure 2D, lanes 1-6). 6 Importantly, FANCJ depletion abolished ssDNA gap-induced SPRTN activity, and the defect was rescued by rFANCJ WT , but not rFANCJ K52R (Figure 2D, lanes 7-24). These findings show that FANCJ helicase activity supports SPRTN-dependent DPC proteolysis independently of the replication fork or CMG bypass.
One possible explanation for FANCJ's role in DPC proteolysis is that it recruits SPRTN to the DPC. However, we saw no difference in SPRTN recruitment to the HpaII pDPC in the presence and absence of FANCJ ( Figure S2C). Another explanation is that after CMG bypass, FANCJ unwinds DNA secondary structures surrounding the DPC. To test this hypothesis, we flanked the DPC with tracts of thymidines, which should not form any secondary structure. As shown in Figure S2D, FANCJ was still required for SPRTN activity in this context. A third possibility is that RPA binding to the ssDNA surrounding the DPC inhibits SPRTN and that FANCJ removes RPA to relieve this inhibition. However, in the context of a ssDNA gap substrate, depletion of RPA did not restore meDPC proteolysis in FANCJ-depleted extract (data not shown). We conclude that FANCJ is required for HpaII-DPC proteolysis by SPRTN, independent of FANCJ's role in promoting CMG bypass of the DPC, and not related to SPRTN recruitment, DNA secondary structure disruption, or RPA displacement.
FANCJ is required to promote SPRTN proteolysis of a native DPC Before exploring further FANCJ's mechanism of action, we addressed whether it promotes proteolysis of an endogenous DPC containing HMCES. HMCES forms DPCs with AP sites in ssDNA, which prevents AP site cleavage and formation of double-strand breaks. 24 We recently found that in egg extracts, endogenous HMCES crosslinks to the AP site generated when DNA replication triggers ICL unhooking by the NEIL3 DNA glycosylase. In this setting, HMCES is subsequently degraded by SPRTN. 25 A simpler approach to generate an HMCES-DPC involves supplementing egg extract with a plasmid carrying an AP site that resides in a ssDNA gap ( Figure 2E). 15 As seen in the context of AP-ICL repair, 25 HMCES-DPC proteolysis in this setting was delayed after SPRTN depletion ( Figure 2E, lanes 1-12). Importantly, FANCJ depletion delayed HMCES degradation to a similar extent as SPRTN depletion ( Figure 2E, lanes [13][14][15][16][17][18], and the combined depletion of FANCJ and SPRTN did not further stabilize HMCES relative to the single depletions (Figure 2E, lanes [19][20][21][22][23][24], consistent with FANCJ functioning in the SPRTN pathway. As seen for SPRTN-dependent HpaII destruction, the effect of FANCJ depletion on HMCES proteolysis was rescued by rFANCJ WT , but not rFANCJ K52R ( Figure 2F). We conclude that FANCJ is essential for efficient proteolysis of an endogenous HMCES-DPC.
FANCJ is sufficient to promote SPRTN proteolysis of a native DPC In biochemical reconstitutions, we previously showed that human SPRTN cleaves a protein G-based DPC in the absence of FANCJ or other proteins, as long as the DPC resides near a ssDNA-dsDNA junction. 16 However, most native DPCs involve DNA-binding proteins such as histones. 26 To investigate how FANCJ affects proteolysis of a native DPC, we used human proteins to reconstitute proteolytic repair of a DPC formed by the catalytic SRAP domain of HMCES, which interacts tightly with the underlying DNA. 27,28 We incubated HMCES SRAP with an AP site-containing oligonucleotide to form a HMCES SRAP -DPC ( Figure S3C for quantification). While we were unable to test a human FANCJ ATPase mutant in our assays because of aggregation of the recombinant protein (FANCJ-K52R, data not shown), a requirement for FANCJ's ATPase activity was indicated by the inability of frog (B) Limited proteolysis of HMCES SRAP -DPCs. Free DNA or the human HMCES SRAP -DPC (WT or R98E) were incubated alone or in the presence of recombinant human FANCJ and trypsin as indicated for 5, 10, or 15 min at 30 C prior to analysis by denaturing SDS-PAGE. Green arrow, tryptic cleavage site accessible in natively folded HMCES SRAP -DPC; orange arrows, cleavage sites exposed upon unfolding of the HMCES SRAP adduct. See also Figure S4. rFANCJ K52R to support DPC cleavage ( Figure S3D). In addition, we tested a Fanconi anemia-causing FANCJ patient variant (FANCJ-A349P), which hydrolyzes ATP and translocates on ssDNA but fails to produce enough force to unwind DNA structures, such as G4 quadruplexes. 29 FANCJ-A349P did not support SPRTN activity ( Figure 3A, lanes 8-10), suggesting that force generation by FANCJ's ATPase motor is required for DPC cleavage. While a prior study showed that RPA stimulates FANCJ's ability to displace proteins from DNA, 22 FANCJ's stimulation of SPRTN activity was not affected by low concentrations of RPA, whereas high concentrations were inhibitory (Figure S3E). These results suggest that the requirement for FANCJ in DPC proteolysis is conserved in humans and involves a direct collaboration between the motor activity of FANCJ and SPRTN.

FANCJ unfolds the protein adduct
We speculated that FANCJ promotes SPRTN activity by translocating into the DPC, which remodels the protein adduct. To test this idea, we asked whether disrupting the native conformation of the protein adduct would bypass the requirement for FANCJ. We first heat-denatured the DPC before adding SPRTN, but this led to only a low level of DPC cleavage in the absence of FANCJ (Figure 4A, lane 12, red arrow). As an independent approach to destabilize the DPC, we generated a HMCES SRAP -DPC with reduced DNA-binding activity. We utilized a previously described HMCES SRAP R98E variant, which shows almost no activity in DNA gel shifts but forms DPCs, implying significant residual DNA binding ( Figures S4A and S4B). 24 As seen for HMCES SRAP WT ( Figure 4A, lanes 7-8), HMCES SRAP R98E was only cleaved in the presence of active SPRTN and FANCJ (Figure 4A, compare lanes 20 and 23-24). However, upon heat denaturation, the mutant DPC was cleaved efficiently in the absence of FANCJ ( Figure 4A, lanes 28 versus 31). We speculated that the mutant adduct remained denatured following heat treatment while the WT adduct refolded. To test this possibility, we analyzed WT and mutant HMCES SRAP -DPCs by native PAGE before and after heat treatment. Prior to denaturation, both DPCs entered the gel ( Figure S4C, lanes 3 and 4, and Figure S4D for quantification). In the case of WT HMCES SRAP , a non-covalent complex between the DPC and free HMCES SRAP was also observed ( Figure S4C, lane 3). Following heat treatment, the majority of WT DPCs still entered the gel and migrated at the original position, consistent with a native conformation (although the noncovalent complexes disappeared). In contrast, HMCES SRAP R98E-DPCs remained in the well, indicating a non-native, misfolded state ( Figure S4C, compare lanes 7 and 8). Our data demonstrate that cleavage of a DPC formed between a native DNA-binding protein and DNA requires FANCJ, whereas when such a DPC is unfolded, FANCJ is dispensable.
To test directly whether FANCJ unfolds the DPC, we probed the conformation of the protein adduct using limited proteolysis. The Remarkably, the addition of FANCJ exposed additional tryptic cleavage sites in native WT and R98E HMCES SRAP -DPCs very close to the DNA ( Figure 4B, lanes 7-9 and 20-22, orange arrows). No effect was observed in the absence of ATP or upon addition of the patient variant FANCJ-A349P ( Figure S4E). The same tryptic cleavage sites were exposed upon heat denaturation of the R98E HMCES SRAP -DPC ( Figure 4B, lanes 24-26, orange arrows), but not the WT HMCES SRAP -DPC ( Figure 4B, lanes [11][12][13], which confirms that the WT adduct retains a native conformation upon heat denaturation. We conclude that FANCJ partially or completely unfolds the protein adduct. FANCJ exposes DNA underlying the DPC We next asked whether unfolding of the protein adduct exposes the underlying DNA. To this end, we placed a HaeIII-restriction enzyme site in the dsDNA 1 nucleotide from the HMCES SRAP -protein adduct ( Figure 5A). HaeIII cleaved the free DNA but failed to do so upon DPC formation, suggesting that the protein adduct blocked access of the restriction enzyme ( Figure 5A, compare lanes 1-2 with 5-6). In this setting, FANCJ restored HaeIII cleavage in an ATP-dependent manner ( Figure 5A, lanes 7-9), and, as seen for SPRTN activity, the requirement for FANCJ was bypassed by heat denaturation of the HMCES SRAP R98E-DPC ( Figure S5A). In contrast, a HaeIII site placed 8 nucleotides away from the DPC was efficiently cleaved independently of FANCJ ( Figure 5A, lanes 14-17). Together, these experiments suggest that FANCJ-dependent DPC unfolding exposes the DNA beneath the DPC, which might allow SPRTN to bind the ssDNA-dsDNA junction and undergo activation. To test whether providing DNA access is sufficient for DPC cleavage, we placed the ssDNA-dsDNA junction adjacent to the DPC footprint, 8 nucleotides from the crosslinking site (SPRTN cleaves protein G adducts up to 10 nucleotides away from an activating structure 16 ). At this position, the ssDNA-dsDNA junction was fully accessible, as indicated by efficient HaeIII cleavage ( Figure 5B, lanes 7-9). However, cleavage of the HMCES SRAP -DPC by SPRTN was only observed upon addition of FANCJ or heat denaturation of the R98E mutant variant ( Figure 5B, lanes 10-12, and Figure S5B). This result indicates that DPC unfolding is required even when the ssDNA-dsDNA junction is accessible and raises the question why the motor activity of FANCJ is not required for protein G-DPC proteolysis.
To address this, we tested the conformation of the protein G adduct using limited proteolysis. Strikingly, we observed a major tryptic cleavage site very close to the DNA that was independent of FANCJ (Figure S5C), suggesting that a flexible, unstructured part of protein G near the attachment site is available for SPRTN cleavage. In summary, our data show that FANCJ-dependent DPC unfolding exposes the underlying DNA, which might allow SPRTN to bind and undergo de-repression of its protease domain. However, in the context of structured DPCs, FANCJ is still required even when an activating DNA structure is accessible, probably to unfold the DPC and allow its entry to the narrow SPRTN active site.
FANCJ is required to promote translesion synthesis past stable DPCs Once a DPC has undergone proteolysis, the remaining peptide adduct is bypassed by TLS, which leads to recoupling of the leading strand with CMG ( Figure 1A). However, we previously showed that even when a DPC fails to be degraded (e.g., due to SPRTN depletion and DPC methylation), TLS can still extend the nascent leading strand past the intact adduct, albeit more slowly than usual, as seen from the large accumulation of À1 species (Figure 6A, lanes 1-6 versus 13-18). 14 Strikingly, FANCJ depletion abolished TLS past the meDPC in SPRTN-depleted extract, permanently arresting the leading strand at the À1 position and preventing accumulation of the mature leading strand product ( Figure 6A, lanes [19][20][21][22][23][24]. By contrast, TLS was independent of FANCJ when the DPC was unmethylated and could therefore be ubiquitylated and degraded by the proteasome (Figure 6A, lanes 7-12, and Figure 1A). Furthermore, when the DPC was pre-digested to a short peptide using proteinase K, TLS still depended on Rev1 6 but was independent of FANCJ ( Figure 6B, lanes [11][12][13][14][15][16][17][18][19][20]. These results indicate that FANCJ is not a general TLS factor but rather is crucial at large protein adducts that cannot be degraded. Our results support the idea that DPC proteolysis normally precedes TLS, but if DPC proteolysis fails, FANCJ promotes TLS past the intact DPC. We next addressed whether FANCJ is also required for TLS when a stable DPC is located in a ssDNA gap. 6 In this setting, FANCJ depletion also greatly inhibited TLS, and we observed that during extension, the 3 0 end flanking the gap initially stalled at the À3 position, followed by slow progression to À1 ( Figure 6C, lanes [8][9][10][11][12][13][14]. TLS and the efficient approach to À1 were restored by rFANCJ WT , but not rFANCJ K52R ( Figure 6C, lanes 15-28). Similarly, loss of FANCJ also caused nascent leading strands to arrest further away from the DPC in the context of full replisome collision with the adduct ( Figure 1B, lanes 13-18, pink  arrowhead). These stalling products disappeared upon addition of rFANCJ WT , but not rFANCJ K52R ( Figure 1B, lanes 25-36).
To address whether these effects stem from a direct role of FANCJ in promoting TLS at a DPC, we crosslinked the HMCES SRAP domain downstream of a primer template junction and added human Pol h or yeast Pol z-Rev1 polymerase with and without FANCJ. In the absence of FANCJ, Pol h failed to bypass the HMCES SRAP -DPC and was unable to advance (B) pmeDPC Lead or pPeptide Lead (generated via proteinase K digestion of pmeDPC Lead ) was pre-incubated with LacR, replicated in the indicated egg extracts containing [a-32 P]-dATP, and analyzed as in Figure 1B primer past the intact DPC ( Figure 6E, lanes 6-9); combining Pol z-Rev1 and Pol h did not result in synergistic effects (data not shown). We propose that FANCJdependent unfolding of the DPC allows leading strands to advance toward and eventually bypass the large protein adduct.

DISCUSSION
Our results suggest a central role for FANCJ in replication-coupled DPC repair ( Figure 7). After CMG stalls, RTEL1 unwinding past the DPC creates a stretch of ssDNA downstream of the adduct. FANCJ binds to this ssDNA and translocates back toward the adduct, which it unfolds, thereby facilitating DPC proteolysis by SPRTN, TLS past non-degradable DPCs, and possibly CMG bypass. To our knowledge, protein unfolding by DNAdependent ATP motors has not been described.

The role of FANCJ in SPRTN activity
In reconstitution experiments, SPRTN is sufficient to cleave a protein G-DPC, 16 but cleavage of a HMCES SRAP -DPC requires the ATPase activity of FANCJ, as does destruction of HpaII in egg extracts. Critically, HpaII and the SRAP domain are tightly folded and bind DNA intimately, 27,30 whereas the protein G construct we used has no DNA-binding activity and contains an unstructured His-tag adjacent to the attachment site. These findings suggested that the motor activity of FANCJ might unfold crosslinked DNA-binding proteins and thereby promote SPRTN activity. Consistent with this model, FANCJ's ATPase activity enhanced HMCES SRAP -DPC proteolysis by trypsin, and when the DPC was irreversibly unfolded (through a combination of a point mutation and heat denaturation), the requirement for FANCJ in SPRTN activity was abrogated. Moreover, a clinically relevant FANCJ mutation (A349P) that specifically disrupts FANCJ's ability to convert ATP hydrolysis to force generation 29 inhibited SPRTN activation by FANCJ. We envision at least two mechanisms by which DPC unfolding promotes SPRTN activity. First, unfolding allows SPRTN binding to the ssDNA-dsDNA junction, promoting SPRTN de-repression. Consistent with this idea, unfolding exposes DNA in the immediate vicinity of the DPC. Second, unfolding allows the DPC to access SPRTN's narrow active site cleft. 31 Consistent with this idea, DPC unfolding is important even when the activating DNA structure is placed adjacent to the DPC. Future work will be required to test these models more directly.

FANCJ promotes translesion DNA synthesis past stable DPCs
In unperturbed DPC repair reactions, DPC proteolysis precedes TLS, 5 and when DPC proteolysis is blocked, TLS is delayed 14 but now depends absolutely on FANCJ. These data show that DPC proteolysis normally facilitates TLS but that in the absence of proteolysis, TLS can still proceed with the assistance of FANCJ. When we prevented DPC proteolysis in the context of the gapped substrate, leading strands stalled at the À1 position before undergoing TLS, whereas in the absence of FANCJ, there was a prolonged arrest at the À2 and À3 positions ( Figure 6C). This result shows that in the context of an intact DPC, FANCJ is critical for approach to the À1 position. Similarly, in reconstituted reactions, FANCJ enabled Pol z-Rev1 to not only approach a DPC but to also synthesize across the lesion. We propose that FANCJ-dependent unfolding of the protein adduct enables polymerase approach by reducing the footprint of the DPC and that it enables extension by allowing the bulky protein adduct to access the polymerase active site.

FANCJ helps CMG overcome obstacles
Our data show that FANCJ is partially redundant with RTEL1 in allowing CMG to bypass DPCs and in promoting replisome progression through a LacR array. In the case of DPC bypass, we propose that FANCJ partially substitutes for RTEL1 in unwinding DNA beyond the DPC via translocation on the lagging strand template. The fact that FANCJ depletion alone has no effect on CMG bypass suggests that DPC unfolding is not essential for CMG bypass or that FANCJ functions redundantly with other factors in this step of DPC repair. In the case of replisome progression through LacR arrays, we propose that like RTEL1, FANCJ cooperates with CMG in the disruption of these non-covalent nucleoprotein complexes by translocating on the lagging strand template. Alternatively, or in addition, FANCJ might disrupt LacR complexes by functioning on the leading strand template behind CMG. Thus, after passage of CMG beyond a LacR-lacO complex and extension of the leading strand to the lacO site, pol ε might dissociate, allowing LacR to re-bind. The re-formed LacR-lacO complex would prevent further progression of the leading strand and cause limited CMG uncoupling. Analogous to its function in DPC repair, FANCJ might then bind to the ssDNA downstream of lacO on the leading strand template and motor back to displace LacR. Translocation of FANCJ into the tightly bound protein would either displace LacR directly or unfold LacR, disrupting its interaction with DNA.
Limitations of the study How a DNA-dependent translocase such as FANCJ unfolds protein adducts and whether DPCs are unfolded partially (as depicted in Figure 7) or completely remain exciting open questions. Another important question is whether FANCJ regulates DPC repair in cells. Notably, FANCJ was identified as the secondstrongest hit in genome-wide screens for formaldehyde sensitivity in RPE1 cells. 17 In agreement, we found that knocking out the FANCJ gene in U2OS cells resulted in formaldehyde sensitivity, which was complemented by re-expression of WT enzyme, but not by FANCJ-K52R or FANCJ-A349P ( Figure S6). However, unlike SPRTN knockouts, FANCJ knockouts are viable. Although this could indicate that FANCJ does not support SPRTN activity in mammals, our biochemical reconstitutions with human SPRTN and FANCJ suggest otherwise. More likely, another 5 0 -to-3 0 helicase is redundant with FANCJ. Consistent with a dedicated role for FANCJ in DPC repair, FANCJ knockouts are more sensitive to formaldehyde than knockouts in other FANC genes, 17 and double knockouts of FANCJ and FANCD2 display added sensitivity toward crosslinking agents. 21 Interestingly, mutations in SPRTN, FANCJ, and RTEL1, all of which function in DPC repair in egg extracts, have different phenotypes in humans. 12,19,32 This could be due to the fact that these proteins, especially the helicases, have numerous functions and that hypomorphic human mutations probably cause a partial loss in a subset of these functions. Understanding the precise role of FANCJ and other proteins in human DPC repair and how their dysfunction causes disease are important future goals.

Conclusion
Together with recent work on FANCJ-dependent resolution of G4 quadruplexes, 20,33 our results suggest a general model for the action of FANCJ in overcoming replicative obstacles. When CMG stalls at a major barrier on the leading strand template (such as a DPC or G4), an accessory helicase unwinds DNA beyond the barrier. This can involve RTEL1 translocating 5 0 to 3 0 along the lagging strand template to facilitate bypass of DPCs or DHX36 translocating 3 0 to 5 0 along the leading strand template to facilitate bypass of G4s. FANCJ then loads onto the unwound leading strand template and translocates back toward the obstacle, on which it exerts force. FANCJ thereby unfolds the barrier, allowing protease activity and/or leading strand extension past the obstacle. We identify FANCJ-mediated unfolding of DPCs and other obstacles as a versatile new function in the DNA repair toolbox.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank Daniel Semlow and Alex Wu for critical feedback on the manuscript, Kim Remans and Julia Flock (EMBL Protein Expression and Purification Core Facility) for help with protein production, and Puck Knipscheer and F. Ulrich Hartl for providing antibodies and/or recombinant proteins. We thank Alex Wu for performing one of the repeats of Figure   Oligonucleotide sequences used in this study are provided in Table S1 N

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Johannes C. Walter (johannes_walter@hms.harvard.edu).

Materials availability
All plasmids are available on request.
Data and code availability d Original western blot and gel images reported in this paper have been deposited at Mendeley and are publicly available as of the date of publication. The DOI is listed in the key resources table. d This study did not generate original code. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Xenopus laevis
Egg extracts were prepared using Xenopus laevis (Nasco Cat #LM0053MX). All experiments involving animals were approved by the Harvard Medical Area Institutional Animal Care and Used Committee and conform to relevant regulatory standards.

Preparation of DNA constructs
To generate pDPC plasmids, either pJLS2 or pJLS3 were nicked with Nt.BbvcI (DPC Lead ) and ligated with an oligonucleotide containing a fluorinated cytosine (dFdC_lead; sequences provided in Table S1) and subsequently cross-linked to M.HpaII-His 6 or methylated M.HpaII-His 6 to generate pDPC Lead and pDPC 2xLead or pmeDPC Lead and pmeDPC 2xLead , respectively, as previously described. 5 Creation of pDPC ssDNA and pmeDPC ssDNA was previously described. 6 Briefly, pJLS2 was nicked with Nb.BbvcI and ligated with an oligonucleotide containing a fluorinated cytosine (dFdC_bottom). The dFdC-containing plasmid was then nicked with Nt.BbvcI and the resulting 31 bp fragment was melted and captured by annealing with an excess of the complementary oligo (Top_capture). Remaining oligos were then degraded by Exonuclease I (New England BioLabs) treatment. The gapped plasmid was subsequently cross-linked to M.HpaII-His 6 or methylated M.HpaII-His 6 to generate pDPC ssDNA or pmeDPC ssDNA , respectively, as previously described. 5 The C5-Fluor dC modified plasmids were mixed with either methylated M.HpaII or nonmethylated M.HpaII in M.HpaII reaction buffer (50 mM Tris-HCl, pH 7.5, 5 mM 2-mercaptoethanol, 10 mM EDTA) and supplemented with 100 mM S-adenosylmethionine (NEB, Ipswich, MA) for 12-18 h at 37 C. Creation of pAP ssDNA plasmid was previously described. 15 Briefly, pJLS2 was nicked with Nb.BbvcI and ligated with an oligonucleotide containing a uracil (dUdC_bottom). The dU-containing plasmid was then nicked with Nt.BbvcI and the resulting 31 bp fragment was melted and captured by annealing with an excess of the complementary oligo (Top_capture). Remaining oligos were then degraded by Exonuclease I (New England BioLabs) treatment. The gapped plasmid was subsequently in experiments.
Xenopus egg extracts and DNA replication Xenopus egg extracts were prepared as described. 39 Briefly, licensing was carried out by supplementing a high-speed supernatant (HSS) of egg cytoplasm with plasmid DNA at a final concentration of 7.5-15 ng/mL. For radiolabeling DNA replication products, [a-32 P]-dATP was added to HSS prior to the DNA. For replication in the presence of LacI, 1 volume of plasmid (75 ng/mL) was incubated with an equal volume of 12 mM LacI for 30 min prior to transfer into HSS so that the final concentration of plasmid was 7.5 ng/mL. 5 Licensing mixes were incubated for 30 min at room temperature to assemble pre-replicative complexes (pre-RCs). To prevent licensing, Geminin was added to HSS at a final concentration of 10 mM and incubated for 10 min at room temperature prior to addition of plasmid DNA. To initiate replication, 1 volume of licensing reaction was mixed with 2 volumes of nucleoplasmic extract (NPE) that had been diluted 2-fold with 1xELB-sucrose (10 mM Hepes-KOH pH 7.7, 2.5 mM MgCl 2 , 50 mM KCl, 250 mM sucrose). 0.5 mL aliquots of replication reaction were typically stopped with 5-10 volumes of replication stop buffer (8 mM EDTA, 0.13% phosphoric acid, 10% ficoll, 5% SDS, 0.2% bromophenol blue, 80 mM Tris-HCl at pH 8), treated with 1 mg/mL Proteinase K. For nascent strand analysis, 2.5 mL aliquots of replication reaction were stopped in 10 volumes of sequencing stop buffer (0.5% SDS, 25 mM EDTA, 50 mM Tris-HCl pH 8.0) followed by addition of 1.25 mL of 190 ng/mL RNase A and incubated for 30 min at 37 C. After RNase digestion, 1.25 mL of 900 ng/mL Proteinase K was added to the DNA samples and incubated overnight at room temperature. Following the Proteinase K treatment, samples were diluted to 150 mL with 10 mM Tris-HCl pH 8.0. The samples were extracted once with an equal volume of phenol/chloroform followed by one extraction with an equal volume of chloroform. The DNA was then precipitated with the addition of 0.1 vol 3M sodium acetate pH 5.2 and 1 mL glycogen (20 mg/mL stock) and resuspended in 7.5 mL of 10 mM Tris-pH 7.5. For RTEL1 immunodepletion and rescue experiments, NPE was supplemented with $200 nM recombinant wild type or mutant Xenopus RTEL1 and incubated for 15 min prior to replication initiation. For MG262 (stock 20 mM; Boston Biochem. Cat# I-120) treatment, NPE was supplement with 200 mM MG262 and incubated for 15 min prior to mixing with HSS (133.33 mM final concentration in replication mix). Samples were analyzed by native 0.8% agarose gel electrophoresis. Gels were exposed to phosphorscreens and imaged on a Typhoon FLA 7000 phosphorimager (GE Healthcare). Band or total lane intensities were quantified using Multi-Gauge software (Fujifilm) with subtraction of appropriate background.

Nascent strand analysis
To nick radiolabeled nascent leading-strands, 3-4 mL of extracted and ethanol precipitated DNA at 1-2 ng mL À1 was incubated in 1x cutsmart buffer (New England BioLabs) with 0.45 units ml À1 Nt.BspQI (New England BioLabs) in a 5 mL reaction at 37 C for 2 h. Nicked DNA (3.5-4 mL samples) was separated on 4% polyacrylamide sequencing gels. To digest radiolabeled nascent leading-strand 3-4 mL of extracted and ethanol precipitated DNA a 1-2 ng mL À1 was incubated in 1x cutsmart buffer (New England BioLabs) with 1 unit ml À1 AatII (New England BioLabs) and FspI (New England BioLabs) in a 5 mL reaction at 37 C for 2 h. Digestion reactions were stopped with 0.5 volumes of Sequencing Stop solution (95% formamide, 20 mM EDTA, 0.05% bromophenol blue, 0.05% xylene cyanol FF). Digested DNA (3.5-4 mL samples) was separated on 7% polyacrylamide sequencing gels. Gels were dried and subjected to phosphorimaging using a Typhoon FLA 7000 phosphoimager. Gels were quantified using Multi Gauge software (Fujifilm).
To quantify the percentage of CMG that underwent bypass, the radioactive signal of all leading strands located between positions +1 and À29 on the gel (reflecting CMGs that have bypassed) was divided by the radioactive signal for leading strands between positions +1 and À44 (reflecting CMGs that have stalled at the lesion or undergone bypass). To help visualize bands, brightness and contrast of some scanned gels were adjusted globally using ImageJ. Quantification of radioactive gels was performed using Typhoon imaging software.

Antibodies and immunodepletion
The xlFANCJ-N antibody was raised against amino acids 69-249 of Xenopus laevis FANCJ. 34 FANCJ antibody was affinity purified from serum using the FANCJ antigen according to standard protocols. In Western blotting of NPE, the affinity-purified xlFANCJ-N antibody recognized $160 and $140 kD bands (data not shown). Both bands were immunoprecipitated from NPE by affinity-purified xlFANCJ-N antibody, but this antibody partially co-depleted FANCM and FANCA (data not shown). The xlFANCJ-C antibody was raised against a C-terminal peptide of FANCJ (CNRENRLSRSRNKGVSSFFLD) by Bethyl laboratories, and it specifically recognized the 160 kD FANCJ band without appreciably co-depleting the 140 kD FANCJ band, which we infer is a C-terminal truncation. It also did not co-precipitate FANCA or FANCM (data not shown). The following antibodies were described previously: RTEL1-N, 14 CDC45, 35 M.HpaII, 6 PSMA3, 6 SPRTN-N, 6 Histone H3 (Cell Signaling Cat #9715S), Mcm6, 14 and HMCES. 25 For FANCJ immunodepletion, 4 volumes of purified xlFANCJ-N, xlFANCJ-C antibody (1 mg mL À1 ), or an equivalent amount of rabbit IgG purified from non-immunized rabbit serum (Sigma) were incubated with 1 volume of Protein A Sepharose Fast Flow (PAS) (GE Healthcare) overnight at 4 C. For RTEL1 immunodepletion, 3.5 volumes of purified RTEL1 antibody (1 mg mL À1 ) or an equivalent amount of rabbit IgG purified from non-immunized rabbit serum (Sigma) were incubated with 1 volume of Protein A Sepharose Fast Flow (PAS) (GE Healthcare) overnight at 4 C. For SPRTN immunodepletion, 4 volumes of SPRTN serum was incubated with 1 volume of Protein A Sepharose Fast Flow (PAS) (GE Healthcare) overnight at 4 C. For mock depletion, 4 volumes of preimmune serum from matched rabbit, was used. In each case, one volume of antibody-conjugated Sepharose was added to 5 volumes of precleared HSS or NPE and incubated for 1 h at 4 C. The HSS or NPE was collected and incubated two more times with antibody-conjugated Sepharose for a total of three rounds of depletion. The depleted HSS or NPE was collected and used immediately for DNA replication, as described above. For FANCJ immunodepletions, xlFANCJ-C antibody was used for the first round of depletion, and the xlFANCJ-N antibody was used for the second and third rounds of depletion. This procedure avoided significant co-depletion of FANCM and FANCA. We speculate that these proteins interact with the C-terminus of the 160 kD form of FANCJ, and that they are displaced from FANCJ by the C-terminal antibody during the first round of depletion. deubiquitylating enzyme Usp21 for 1 h at 37 C. Subsequently, all residual buffer was removed, and the beads were resuspended in 7.5 mL of Benzonase buffer containing 1 mL of Benzonase (Sigma) at 37 C for 1 h to allow for DNA digestion and DPC elution, after which the beads were pelleted and the supernatant M.HpaII eluate was mixed with 2X Laemmli sample buffer for subsequent western blotting analysis. Equal volumes of the protein samples were blotted with the indicated antibodies. To help visualize bands in plasmid and DPC pull-down experiments, brightness and contrast of some Western blots were adjusted globally using ImageJ. Quantification was performed using ImageJ.

Generation of HMCES SRAP -DPCs
DPCs were generated between HMCES SRAP WT and mutant variants and a 30mer Cy5-fluorescently-labelled forward oligonucleotide oDY_54. HMCES SRAP was prediluted to 40 mM in purification buffer D and forward oligonucleotide was prediluted to 1 mM in DNA dilution buffer (50 mM HEPES/KOH pH 7.5, 100 mM KCl, 10% Glycerol, 0.4 mg/mL BSA). Cross-linking was carried out in 10 mL final volume containing 1 mL forward oligonucleotide, 0.5 mL HMCES SRAP , 0.48 mL UDG (New England BioLabs) and 8.02 mL reaction buffer (20 mM HEPES/KOH pH 7.5, 50 mM KCl, 10 mM MgCl 2 , 2 mM TCEP, 0.1 mg/mL BSA), resulting in final concentrations of 0.1 mM DNA, 2 mM HMCES SRAP , and 0.1 U/mL UDG. The reaction was incubated for 1 h at 37 C. Next, 1 mL of 15mer complementary reverse oligonucleotide oHR_127 (12 mM in nuclease-free H 2 O) was added to the cross-linking reaction. Annealing was performed by incubating the reaction for 2 min at 37 C followed by a decrease in temperature of 1 C/min until 20 C was reached. In experiments using ssDNA DPCs, the reverse oligonucleotide was replaced by H 2 O. In experiments using heat-denatured DPCs, the reactions were incubated for 5 min at 60 C prior to reverse oligonucleotide annealing. All DPCs were prepared immediately prior to cleavage assays.
For analysis of DPCs by native PAGE, HMCES SRAP was prediluted to 40 mM in purification buffer D and forward oligonucleotide was prediluted to 1 mM in DNA dilution buffer. The assay was carried out in 10 mL final volume with 1 mL forward oligonucleotide, 0.5 mL HMCES SRAP , 0.48 mL UDG (1 U) and 8.02 mL reaction buffer (20 mM HEPES/KOH pH 7.5, 50 mM KCl, 10 mM MgCl 2 , 2 mM TCEP, 0.1 mg/mL BSA). The reactions were incubated for 1 h at 37 C to allow DPC formation. For heat-denaturation, DPCs were then incubated for 5 min at 60 C. 4 mL of 6x Orange G loading dye was added and the samples were separated on 6% native PAGE gels with 0.5x TBE as running buffer at room temperature. Gels were photographed using a BioRad Chemidoc MP system using appropriate filter settings for Cy5 fluorescence. To help visualize bands, brightness and contrast of some gel images were adjusted globally using ImageJ. Quantification was performed using ImageJ; the fraction of DPC in the well was determined by dividing the amount of DPC retained in the well by the total amount of DPCs (retained in well plus DPC in gel).

Generation of protein G-DPCs
Recombinant N-terminally His-tagged protein G (BioVision, #6510) was conjugated to a fluorescently-labelled oligonucleotide 30_FAM_X15 using proFIRE Amine Coupling Kit (Dynamic Biosensors, #PF-NH2-1) as described previously. 42 DPC cleavage assays For the experiment shown in Figure 3A, FANCJ dependent-cleavage of HMCES SRAP -DPCs by SPRTN was assessed in a reaction volume of 10 mL containing 1 mL of the HMCES SRAP DNA cross-linking reaction described above, 80 nM FANCJ, 100 nM SPRTN in a final reaction buffer of 9.5 mM HEPES/KOH pH 7.5, 70 mM KCl, 2 mM MgCl 2 , 2 mM ATP, 2% Glycerol, 5.5 mM TCEP and 0.1 mg/mL BSA. For the experiments shown in Figures 4A, S3B, and S3E, the reaction was performed identically, but using 100 nM FANCJ in a final reaction buffer of 17 mM HEPES/KOH pH 7.5, 85 mM KCl, 2 mM MgCl 2 , 2 mM ATP, 3% Glycerol, 5.5 mM TCEP and 0.1 mg/mL BSA. For the experiment shown in Figure S3D, the reaction was performed identical, but using 20 nM X.l. rFANCJ in a final reaction buffer of 12 mM HEPES/KOH pH 7.5, 2 mM Tris/HCl, 20 mM NaCl, 70 mM KCl, 2 mM MgCl 2 , 2 mM ATP, 2% Glycerol, 5.5 mM TCEP and 0.1 mg/mL BSA. Reactions were incubated for 1 h at 30 C and stopped by the addition of 5.5 mL SDS loading buffer. The reactions were then boiled for 1 min at 95 C and resolved on 4-12% or 12% SDS-PAGE gels. Gels were photographed using a BioRad Chemidoc MP system using appropriate filter settings for Cy5 fluorescence. ImageJ was used for quantification; the fraction of cleaved DPCs was determined by dividing the amount of cleaved DPCs by the total amount of DPCs (cleaved plus uncleaved).

DNA binding assays
Electrophoretic mobility shift assays (EMSAs) were used to analyze binding of recombinant HMCES SRAP WT and mutant variants to a Cy5-fluorescently labeled 30-mer oligonucleotide oDY_54. HMCES SRAP was prediluted to 40, 10 and 2.5 mM in purification buffer D and forward oligonucleotide was prediluted to 1 mM in DNA dilution buffer. Binding was carried out in 10 mL final volume with 1 mL forward oligonucleotide, 0.5 mL HMCES SRAP and 8.5 mL reaction buffer (20 mM HEPES/KOH pH 7.5, 50 mM KCl, 10 mM MgCl 2 , 2 mM TCEP, 0.1 mg/mL BSA). The reactions were incubated for 20 min on ice. 4 mL of 6x Orange G loading dye was added and the samples were separated on 6% native PAGE gels with 0.5x TBE as running buffer at 4 C. Gels were photographed using a BioRad Chemidoc MP system using appropriate filter settings for Cy5 fluorescence.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical details of each experiment (including the exact value of n, what n represents and precision measures) can be found in the figure legends.

OPEN ACCESS
Article