Specificity for deubiquitination of monoubiquitinated FANCD2 is driven by the N-terminus of USP1

Deubiquitination of FANCD2, FANCI, and PCNA by USP1 is essential for DNA repair signalling. Reconstitution of the system reveals that USP1-mediated specificity towards K561 of FANCD2 is directed by a unique sequence at USP1's N-terminus.


Introduction
Ubiquitination is a reversible post-translational modification that regulates almost every cellular process in eukaryotes. Cycles of ubiquitination and deubiquitination orchestrate the assembly and disassembly of many DNA repair complexes in DNA damage response pathways. These include the Fanconi anemia (FA) pathway, required to repair DNA interstrand crosslinks (ICLs), and the translesion synthesis pathway (TLS), required for DNA damage tolerance (1). FA is a chromosomal instability disorder that results from a dysfunctional ICL repair pathway (2). Central to FA ICL repair is monoubiquitination of two homologous proteins, Fanconi anemia group D2 protein (FANCD2) and Fanconi anemia group I protein (FANCI), at two specific lysines, K561 and K523, in humans, respectively. Monoubiquitination of FANCD2 and FANCI is catalysed by the E3 ubiquitin ligase FANCL and the E2 conjugating enzyme Ube2T (3). Monoubiquitinated FANCD2 (FANCD2-Ub) signals multiple DNA repair proteins to conduct ICL repair (2). A similar specific modification is central to TLS repair, where K164 of proliferating cell nuclear antigen (PCNA) is monoubiquitinated (PCNA-Ub) by the RAD18 E3 ligase and Rad6 E2 (4), which, in turn, recruits TLS polymerases for DNA repair (5). As well as ubiquitination, both ICL and TLS repair require deubiquitination (removal of ubiquitin). Interestingly, although the modifications in each pathway are assembled by distinct enzymes, they are removed by the same deubiquitinase (DUB), the USP1-USP1-associated factor 1 (UAF1) complex (6,7,8). Loss of USP1 function results in an accumulation of FANCD2, FANCI, and PCNA; genomic instability; and a failure to complete the pathways (7,9,10,11,12). In addition to these three substrates, USP1 deubiquitinates a number of other substrates, including the inhibitor of DNA-binding proteins 1-4 (ID1-4), which regulate cell differentiation (13), and TBK1, which is involved in viral infection (14). USP1 belongs to the largest family of DUBs, the ubiquitin-specific proteases (USPs), which contain~50 members. Many USPs show little substrate discrimination between ubiquitin-ubiquitin chains in vitro (15), and can hydrolyse polyubiquitin chains from substrates (16). A few USPs exhibit preference for specific ubiquitin-ubiquitin linkages, such as USP30 for K6-linked Ub chains (17). In contrast, USP1 targets monoubiquitinated substrates and regulates a distinct set of modified proteins. Although molecular mechanisms of ubiquitin removal from ubiquitin are well understood (16), it is less clear how ubiquitin-substrate linkages are specifically targeted. The core catalytic USP domain is~350 amino acids. However, most USPs also contain multiple insertions within the catalytic domain and additional N/C-terminal extensions (18). USP1 has multiple insertions and an extended N-terminus on its USP domain, and their functions are currently unknown.
USP1 has little DUB activity on its own, but is regulated by and forms a stoichiometric complex with UAF1. UAF1 also binds and activates two other DUBs, USP12 and USP46 (19), and studies that reveal how UAF1 binds and activates USP12 and USP46 suggest that 1 UAF1 will bind to USP1 in an analogous manner (20,21). UAF1 acts to stabilise its USP partners and increase catalytic activity (22). UAF1 knockout in mice is embryonic lethal, whereas USP1 knockouts result in a FA-like phenotype, reflecting the additional functions of UAF1 (9,23). In addition to its activation role, UAF1 has a C-terminal SUMOlike domain (SLD) responsible for recruiting USP1 indirectly to FANCD2 and PCNA via a weak interaction with SUMO-like interacting motifs in FANCI and ATAD5, respectively (24). Despite the common activator function of UAF1, loss of either USP12 or USP46 does not result in accumulation of USP1 substrates (19), suggesting that USP1 could target its substrates independent of UAF1. However, it remains unclear how USP1 specifically targets its substrate pool.
Investigating how USPs deubiquitinate their substrates on a molecular level is very challenging because of the difficulty in making physiological and correctly ubiquitinated substrates. To date, most of our understanding of DUB specificity has used ubiquitin-ubiquitin linkages as substrates, likely because of the advances in purifying large quantities of ubiquitin chains. However, there are a few examples of studies that have used monoubiquitinated substrates with a native isopeptide, and these include PCNA-Ub (25) and histones (26). Particularly in the case of histone H2A and H2B, the generation of monoubiquitinated substrates has facilitated the elucidation of the mechanisms of substrate targeting by histone-specific DUBs (26,27). The ability to make physiological substrates allows for a modular approach to understanding the requirements for specificity and how DUBs such as USP1 work at the molecular level.
Here, we report the reconstitution of substrate deubiquitination by USP1-UAF1 that allows for a modular approach to understanding the molecular requirements for deubiquitination of physiological substrates. We define the molecular determinants for substrate deubiquitination via a few residues in a highly conserved and extended N-terminus of USP1. Our analysis indicates that the N-terminus of USP1 harbours a FANCD2-specific binding sequence important for deubiquitinating only one specific lysine on FANCD2: K561-the location of a specific DNA repair signal. Remarkably, we find that the N-terminus of USP1 is important for FANCD2-K561Ub deubiquitination but apparently not for PCNA-K164Ub or FANCI-K523Ub. Finally, we find that the N-terminus of USP1 is sufficient to engineer specificity in a more promiscuous DUB. Our analysis shows that USP1 discriminates between substrates and has direct elements for targeting FANCD2, regardless of whether it is ubiquitinated, providing further insights into how USPs select their substrates.

USP1 catalytic domain is sufficient for activity and binding to UAF1
To gain insight into USP1 specificity, we analysed the amino acid sequence and separated the protein into regions termed USP domain, Insert 1, Insert 2, and N-terminus ( Fig 1A). Several elements important for cellular regulation have been previously identified, such as a calpain cleavage site within the N-terminus (28), a degron motif for anaphase-promoting complex/cyclosome Cdh1 targeting within insertion 1 (29), and an auto-cleavage region (G670/G671) within insertion 2 (7). However, whether these regions are important for in vitro DUB activity is unknown. We designed multiple USP1 fragments for expression in Sf21 insect cells, keeping the catalytic domain intact, but systematically removing the insertions and N-terminus ( Fig 1B). Each construct of USP1 is expressed and purified to homogeneity (Fig 1C). To assess catalytic activity and competency, we used ubiquitin-propargylamine (Ub-prg), a "suicide" probe which crosslinks to the active site cysteine residue of DUBs (30). Recombinant USP1 FL , in addition to all of the other fragments, fully reacts with Ub-prg (Fig 1C), thus indicating the catalytic domain is competent and able to bind ubiquitin. USP1 FL has multiple breakdown products, but when insertions 1 and 2 are deleted, the yield and purity are much higher, indicating an increase in protein stability. Indeed, thermal denaturation assays reveal that USP1 ΔNΔ1Δ2 is more thermostable (T m = 44 ± 0.24°C), whereas USP1 FL melts at lower temperatures (T m = 37 ± 0.81°C) ( Fig  S1). Importantly, USP1 requires to be in complex with UAF1 for robust proteolytic activity (22). We find that UAF1 can bind the minimal USP1 catalytic domain (USP1 ΔNΔ1Δ2 ) by size exclusion chromatography (SEC) ( Fig 1D) and stimulate DUB activity on a fluorescent ubiquitinated-dipeptide substrate, Ub-KG TAMRA (Fig 1E). The cleavage rates show that the activity of a USP1 ΔNΔ1Δ2 -UAF1 complex is almost identical to that of USP1 FL -UAF1. Because a minimal pseudosubstrate was used, we further assayed the activity against K63-and K48-linked di-ubiquitin chains ( Fig 1F) and did not detect any differences in activity between USP1 FL and USP1 ΔNΔ1Δ2 . Together, these data show that the catalytic domain of USP1, with stoichiometric amounts of UAF1, is sufficient for DUB activity and that the additional regions of USP1 are not important for in vitro DUB catalysis or UAF1 stimulated activity.

USP1-UAF1 directly deubiquitinates monomeric FANCD2-Ub and FANCI-Ub
The USP1-UAF1 complex can deubiquitinate several monoubiquitinated substrates, including human FANCD2-K561Ub (hsFANCD2-Ub) (6), human FANCI-K523Ub (hsFANCI-Ub) (31), and human PCNA-K164Ub (PCNA-Ub) (25). To understand the requirements for deubiquitination, we optimised a robust method for producing full-length isolated hsFANCD2-Ub and hsFANCI-Ub substrates. The homogeneous preparation of isolated hsFANCD2-Ub and hsFANCI-Ub is very challenging, as previous studies required a non-mammalian FANCD2-FANCI substrate complex, in addition to needing DNA and a six-protein E3 ligase complex (FANCC, FANCE, FANCF, FANCB, FANCL, and FAAP100), to stimulate the modification (32). To enhance the yield and purity of hsFANCD2-Ub and hsFANCI-Ub, we developed a method that requires only a FANCL fragment (FANCL ΔELF ) and a hyperactive mutant form of the E2, Ube2Tv4 (33), to stimulate the reaction (see the Materials and Methods section). Using these enzymes (Fig 2A), we observed robust monoubiquitination of the isolated human FANCD2 and FANCI substrates within 60 min when monitored with either fluorescent ubiquitin (Ub 800 ) (Fig S2A) or GST-Ub ( Fig 2B). To ensure site specificity of ubiquitination, we mutated K561 and K523 to Arg in FANCD2 and FANCI, respectively. In contrast to WT FANCD2 and FANCI, the mutants K561R and K523R are not monoubiquitinated using these reaction conditions, indicating that site specificity is maintained (Fig 2B). We further assessed site specificity by purifying hsFANCD2-Ub and hsFANCI-Ub to homogeneity and analysed both modified and unmodified proteins by mass spectrometry to verify whether the correct lysine was ubiquitinated. We detected ubiquitin on FANCD2-K561 and FANCI-K523 and did not detect any secondary ubiquitination events ( Fig S2B). In addition, these reaction conditions also support monoubiquitination of the isolated Xenopus laevis (xl) FANCD2 and FANCI substrates without the need for any reaction cofactors (Fig S2C). Purification profiles of both hsFANCD2-Ub and hsFANCI-Ub show no apparent changes in their solution states or hydrodynamic radius compared with non-ubiquitinated proteins ( Fig 2C). Finally, we purified hsPCNA-Ub using a previously described method that uses a mutant E2 (UbcH5c S22R (25)) ( Fig 2C). With the substrates purified (Fig 2C), we assessed whether hsFANCD2-Ub, hsFANCI-Ub, or hsPCNA-Ub is directly deubiquitinated by USP1-UAF1 in vitro. The recombinant USP1 FL is able to fully deubiquitinate hsFANCD2-Ub, hsFANCI-Ub, and hsPCNA-Ub at sub-stoichiometric concentrations, in a UAF1dependent manner (Fig 2D). In contrast, a catalytically dead mutant of USP1 (C90S) is unable to deubiquitinate any of the substrates ( Fig 2D). Taken together, these data establish that recombinant USP1-UAF1 is able to directly deubiquitinate diverse monoubiquitinated substrates in vitro without the need for auxiliary factors.

The N-terminus of USP1 is critical for FANCD2-Ub deubiquitination
To understand the minimum requirements for deubiquitination, we assessed whether the catalytic domain of USP1 is sufficient for deubiquitinating its physiological substrates. To test this, we assayed multiple USP1 fragments. As shown, USP1 FL -UAF1 readily deubiquitinates hsFANCD2-Ub, hsFANCI-Ub, and hsPCNA-Ub ( Fig 3A). In addition, the minimal catalytic module (USP1 ΔNΔ1Δ2 -UAF1) is sufficient for deubiquitinating hsFANCI-Ub ( Fig 3A). Surprisingly, USP1 ΔNΔ1Δ2 -UAF1 is unable to deubiquitinate hsFANCD2-Ub and there is slower activity on hsPCNA-Ub at early time points (Fig 3A), indicating the catalytic domain is not sufficient despite the catalytic domain being active on other substrates including hsFANCI-Ub. In light of this unexpected result, we wanted to determine which deleted region of USP1 is responsible for the loss in deubiquitinating activity on hsFANCD2-Ub. We, therefore, assayed each deletion: USP1 ΔN , USP1 Δ1 , and USP1 Δ2 . Both deletions of inserts 1 and 2 cleave at the same rate as USP1 FL ; however, deletion of the N-terminus results in a large loss of activity on hsFANCD2-Ub ( Fig 3B). In contrast to hsFANCD2-Ub, none of the deletions have any apparent effect on hsFANCI-Ub deubiquitination or hsPCNA-Ub deubiquitination ( Fig  3B). Because the minimal USP1 ΔNΔ1Δ2 is still an active DUB, we wanted to determine whether at higher concentrations hsFANCD2-Ub could be deubiquitinated by this fragment. Full deubiquitination of hsFANCD2-Ub can only be achieved by increasing the concentration of USP1 ΔNΔ1Δ2 to a 1:1 ratio (Fig 3C). This is in contrast to USP1 Δ1Δ2 , which is able to deubiquitinate hsFANCD2-Ub at sub-stoichiometric concentrations ( Fig 3C). These data indicate that although the N-terminus is not required for catalytic activity per se, it indeed contributes to a more productive hsFANCD2-Ub DUB activity.
A recent report suggests that xlFANCI is required for human USP1-UAF1 to remove ubiquitin from xlFANCD2-Ub (32). In contrast, we found that xlFANCD2 does not require the presence of xlFANCI to be deubiquitinated ( Fig 3A). Therefore, we speculated whether our observation was because of the use of human FANCD2 substrate. To test this, we purified xlFANCD2-Ub to homogeneity (Fig S2C-E) and assayed with human USP1-UAF1. In contrast to previous reports, we found that USP1 FL and USP1 Δ1Δ2 are able to fully deubiquitinate xlFANCD2-Ub ( Fig 3D). Furthermore, consistent with our observations using hsFANCD2-Ub as substrate, deletion of the N-terminus reduces USP1 activity for xlFANCD2-Ub (Fig 3D). Because deletion of the N-terminus of USP1 does not result in an apparent defect for Ub-KG TAMRA (Fig 1E), K63/K48 diUb (Fig 1F), hsFANCI-Ub (Fig 3B), and hsPCNA-Ub (Fig 3B), it is likely that the USP1 N-terminus is a specific FANCD2-Ub requirement that is extended from the catalytic domain. Taken together, these data suggest that the N-terminus of USP1 harbours a substrate targeting sequence specific for FANCD2.

The N-terminus drives specific FANCD2-K561-Ub deubiquitination
The ubiquitination site of FANCD2 is at a specific residue, K561. We speculated whether the FANCD2 targeting by the N-terminus of USP1 also extends to the site of ubiquitination, that is, for K561-Ub. To test this, we generated FANCD2 and modified lysines distinct from K561 ( Fig 4A). We decided to use a mutant E2 (UbcH5c S22R ) that primarily monoubiquitinates proteins (25). As the monoubiquitination activity was weak, we used the E3 ligase RNF4 fragment (RNF4-RING fusion, RNF4 RR ) to increase the activity (34). The UbcH5c S22R /RNF4 RR pair resulted in robust and multiple monoubiquitination events on hsFANCD2 ( Fig 4B). To ensure K561 is not being monoubiquitinated by UbcH5c S22R /RNF4 RR , we used K561R hsFANCD2 as the substrate. We performed an E3 assay to monoubiquitinate hsFANCD2 and hsFANCD2 K561R, which we term hsFANCD2 K561-Ub and hsFANCD2 KX-Ub, respectively ( Fig 4A), and treated the reaction products with USP1 FL -UAF1 or USP1 ΔN -UAF1. Interestingly, whereas USP1 FL -UAF1 deubiquitinates both substrates K561-Ub and KX-Ub at similar rates, USP1 ΔN -UAF1 shows clear activity on KX-Ub but little detectable activity on K561-Ub ( Fig 4C). These data show that the USP1 N-terminus targeting sequence also specifies the site for DUB activity on FANCD2-Ub, that is, K561-Ub.

Comparing deubiquitination activity of E. coliand Sf21-expressed USP1
To determine whether the N-terminus dependence of hsFANCD2-Ub deubiquitination by USP1 was due to possible post-translational modifications in eukaryotic cells, we also expressed and purified USP1 Δ1Δ2 and USP1 ΔNΔ1Δ2 from E. coli and tested their activity ( Fig S3). A comparison of Sf21-and E. coli-expressed USP1 shows only minor differences in activity on hsFANCD2-Ub, and confirms that the activity depends on the N-terminus of USP1 ( Fig S3A). Interestingly, a small difference in activity is apparent for hsPCNA-Ub deubiquitination when comparing USP1 Δ1Δ2 and USP1 ΔNΔ1Δ2 , and this difference is more obvious in insect cell-expressed USP1 (Fig S3B), perhaps indicating some form of regulation within the N-terminus for targeting hsPCNA.
However, as deubiquitination of hsFANCD2-Ub still depends on the N-terminus regardless of how USP1 is expressed, we used the E. coli USP1 to assess point mutations within the N-terminus.
The N-terminus of USP1 contains critical residues for FANCD2-Ub deubiquitination To identify residues within the N-terminus of USP1 required for hsFANCD2-Ub specificity, we looked at regions that are well conserved within the N-terminus of USP1 ( Fig 5A). The N-terminus of USP1 is almost completely conserved from humans to zebrafish between residues 19 and 40 ( Fig 5A). Similar to deleting residues 1-54 of USP1, deletion of residues 21-29 also results in a specific loss in activity for hsFANCD2-Ub but not hsPCNA-Ub (Fig 5B). We next determined whether any of the side chains are important, so we mutated each residue to alanine and assayed for DUB activity using Ub-prg, hsPCNA-Ub, hsFANCI-Ub, and hsFANCD2-Ub ( Fig S4). All point mutants are active with Ub-prg (Fig S4A), and deubiquitinate hsPCNA-Ub and hsFANCI-Ub (Fig S4B). In contrast, there is a loss in activity for several point mutants, such as N21A, R22A, L23A, S24A, and K26A, on hsFANCD2-Ub deubiquitination (Fig S4B). To quantify the relative differences between USP1 mutants, we purified hsFANCD2-Ub with a fluorescently labelled Ub 800 that allows us to detect only the ubiquitinated FANCD2 (hsFANCD2-Ub 800 ) (Fig S5). Using this approach, we monitored the deubiquitination of hsFANCD2-Ub 800 by each mutant and quantified the remaining FANCD2-Ub as a percentage of the input (Fig 5C). We found that whereas USP1 Δ1Δ2 -UAF1 cleaves~80% of the hsFANCD2-Ub 800 , USP1 ΔNΔ1Δ2 -UAF1 shows no apparent cleavage (Fig 5C). Interestingly, R22A and L23A are able to only deubiquitinate~10% of the hsFANCD2-Ub 800 (Fig 5C). We did not further characterise L23A as it co-purifies with a large contaminant chaperone, which may suggest some misfolding (Fig S4). We next assessed whether the charge at R22 is critical for deubiquitination as it is conserved as a lysine in zebrafish (Fig 5A). We generated R22K, R22A, and R22E variants (Fig S4C), and as expected, these had no apparent effect on hsPCNA-Ub or hsFANCI-Ub deubiquitination (Figs 5D and S4D). In contrast, we observed that R22A has an intermediary affect and R22E is comparatively more compromised on hsFANCD2-Ub deubiquitination (Fig 5D). R22K retains WT activity on hsFANCD2-Ub (Fig 5D), suggesting a positive charge is required. Finally, we found that in the full-length USP1-UAF1 context, a single point mutation in its N-terminus (R22E) is sufficient for specifically disrupting hsFANCD2-Ub deubiquitination but not hsFANCI-Ub or hsPCNA-Ub (Fig S6A-C). Taken together, these data indicate that R22 plays an important role in deubiquitination of hsFANCD2-Ub. However, it is clear that multiple conserved residues (N21, R22, L23, S24, and K26) that are within a short N-terminal region may play a role in driving USP1-mediated FANCD2-Ub deubiquitination.
Mutation or deletion of the USP1 N-terminus specifically and negatively impacts FANCD2-Ub deubiquitination. We hypothesised that this region in the N-terminus of USP1 encodes a FANCD2binding site. To test this, we purified UAF1 with an N-terminal Strep tag and formed stable complexes with USP1 Δ1Δ2 , R22A, R22K, or USP1 ΔNΔ1Δ2 . We assayed interaction with hsFANCD2 and found that a USP1 Δ1Δ2 -UAF1 complex can capture the non-ubiquitinated substrate ( Fig S7A). In contrast, when using USP1 ΔNΔ1Δ2 -UAF1 as bait, hsFANCD2 is not enriched above the levels seen in the beads control ( Fig S7A). These data indicate the USP1 binding to hsFANCD2, which is almost completely dependent on the N-terminus of USP1. In addition, USP1 Δ1Δ2 -R22A only partially interacts with hsFANCD2, and USP1 Δ1Δ2 -R22K does not weaken the substrate interaction-both results are consistent with the hsFANCD2-Ub deubiquitination assays (Figs S7A and 5A). As USP1's interaction with FANCD2 is dependent on the N-terminus, we wanted to determine whether this is specific for hsFANCD2, so we similarly assayed interactions of hsFANCI and hsPCNA. Interestingly, we can detect a weak interaction with both hsFANCI and hsFANCI-Ub that is unaffected by deletion of the N-terminus ( Fig S7B). We could not detect an interaction between USP1 and hsPCNA, but we observed an interaction with hsPCNA-Ub (Fig S7C), which is, however, not dependent on the N-terminus of USP1. Together, these data indicate that the N-terminus of USP1 directly interacts with FANCD2.
A chimera fusion of USP1 N-terminus to the USP2 catalytic domain provides FANCD2 specificity Deletion of the N-terminus of USP1 does not appear to affect the catalytic activity of the USP1 catalytic module (USP1 ΔNΔ1Δ2 ), except in case of FANCD2-Ub deubiquitination. We, therefore, speculated whether the N-terminus can influence the efficiency of a distinct DUB. To test this, we chose the catalytic domain of USP2, generally considered a versatile/promiscuous DUB because of its lack of substrate specificity (35). We created and purified a USP2 chimera (USP1 1-60 -USP2) where residues 1-60 of USP1 are present in the N-terminal to the USP2 catalytic domain. From the in vitro assays, we found that whereas 100 nM of USP2 is able to completely deubiquitinate 1 μM of hsFANCD2-Ub by 40 min (Fig 6A), the chimeric USP1 1-60 -USP2 achieves this within 10 min (Fig 6A), showing a clear gain of activity. Because the N-terminus increases USP2 activity against FANCD2-Ub, we tested whether the N-terminus augments its activity towards other ubiquitinated substrates. In contrast to hsFANCD2-Ub, there are no obvious differences in the activity of USP2 and the USP2 chimera using hsFANCI-Ub as the substrate (Fig 6B). In addition, only minor differences in the deubiquitination activity of hsPCNA-Ub are observed at early time points (Fig 6C). These data show that the addition of USP1 N-terminus in cis confers a gain of activity for USP2 on hsFANCD2-Ub.

FANCD2-FANCI-DNA complex deubiquitination depends on the N-terminus of USP1
The FANCD2 and FANCI substrates can form a heterodimer on DNA, and this ensemble has been shown to be important for in vitro monoubiquitination (32,36,37). However, it is not yet known how the monoubiquitinated FANCD2 and FANCI proteins assemble or interact once modified. It is, therefore, unclear whether hsFANCD2-Ub is deubiquitinated as part of a complex or on its own. A recent report has showed that the di-monoubiquitinated frog ID2 complex (both FANCD2 and FANCI monoubiquitinated), generated in the presence of plasmid DNA, remains resistant to deubiquitination by a GST-tagged USP1-UAF1 complex and only the removal of DNA permits efficient DUB activity (32). Because under physiological conditions the monoubiquitinated ID2 complex is normally localized on DNA, we therefore wanted to test whether deubiquitination of the hsID2-DNA complex is also dependent on the N-terminus of USP1. To test this, we first monoubiquitinated the hsID2 heterodimer in the presence of double stranded DNA (dsDNA) (36) and arrested the ubiquitination reactions with apyrase. We subsequently treated half of the substrate sample with benzonase nuclease to eliminate the DNA before subjecting both the nucleasetreated and untreated substrate samples to USP1-UAF1 activity. Whereas USP1 FL is able to deubiquitinate some of the hsFANCD2-Ub in the presence of both DNA and hsFANCI, a portion of hsFANCD2-Ub and hsFANCI-Ub remains resistant to deubiquitination (Fig 7A), consistent with the recent findings using frog substrates (32). Interestingly, when DNA is removed, all of the hsFANCD2-Ub and hsFANCI-Ub is deubiquitinated by USP1 FL -UAF1; in contrast, USP1 ΔN -UAF1 and USP1 R22E -UAF1 do not deubiquitinate hsFANCD2-Ub (Fig 7A). To allow a clear assessment of whether DNA, FANCI-Ub, or both factors inhibit deubiquitination, we monoubiquitinated FANCD2-FANCI K523R -DNA complexes. Using this set-up, we observed that USP1 FL -UAF1 is able to deubiquitinate hsFANCD2-Ub-FANCI with and without benzonase treatment (Fig 7B). These data suggest that DNA does not inhibit FANCD2-Ub deubiquitination except in the context of the di-monoubiquitinated ID2-DNA complex. Interestingly, and despite the presence of FANCI, which may also recruit the DUB complex (24), USP1 ΔN -UAF1 and USP1 R22E -UAF1 are not able to fully deubiquitinate hsFANCD2-Ub, suggesting that any indirect recruitment of USP1-UAF1 by FANCI does not completely compensate for the loss of the USP1 N-terminus. Thus, the N-terminus of USP1 contains a FANCD2 determinant sequence that functions at the core of FANCD2  deubiquitination by specifically binding and increasing deubiquitination efficiency for FANCD2-Ub even in the presence of both dsDNA and FANCI (Fig 7C).

Discussion
A cycle of monoubiquitination and deubiquitination of FANCD2 is critical for completion of FA ICL repair (8). Removal of the ubiquitin signal from FANCD2 is conducted by the USP1-UAF1 DUB, and disruption of USP1 catalytic activity results in an accumulation of FANCD2-Ub and FA-like phenotypes (6,9,11,23). Despite multiple DUBs existing at DNA replication and repair events, USP1 appears to be the only DUB whose loss results in an accumulation of FANCD2-Ub (6). The molecular determinants and mechanisms of how DUBs, such as USP1, target and regulate distinct pools of substrates have remained elusive, because of a lack of molecular insights. Here, we report a unique feature and a potential short linear motif at the N-terminus of USP1, which shows that USP1 specifically and directly targets K561 of FANCD2, providing a molecular foundation for how USP1 targets one of its substrates with high specificity. USP1 activity is subject to layers of regulation during cell-cycle progression. For example, USP1 is turned over throughout the cell cycle, is activated and stabilised by binding partners such as UAF1, and, in addition, undergoes autocleavage to allow targeting by both the N-end rule and C-end rule pathways (7,11,22,38). Interestingly, UAF1 is an abundant protein with multiple functions, including activating USP12 and USP46 (19). It has been suggested that UAF1 may indirectly target USP1 to specific substrates such as FANCD2 or PCNA via a C-terminal SLD that binds to SUMO-like interacting motifs on protein partners of FANCD2 and PCNA (24) (Fig 7C). However, disruption of either USP12 or USP46 does not affect FANCD2-Ub levels, showing that FANCD2 is not an overlapping substrate (19). Because the SLD of UAF1 would target multiple DUBs to the same substrates, further layers of regulation may exist to ensure the correct DUB is targeted to FANCD2-Ub or PCNA-Ub. By reconstituting FANCD2-Ub, FANCI-Ub, and PCNA-Ub deubiquitination by USP1-UAF1, we identified the minimal components required for deubiquitination and discovered a unique sequence within the N-terminus of USP1 critical for removing ubiquitin from FANCD2 ( Fig  7C). We noticed a relatively small, but reproducible difference in activity on PCNA-Ub, which is apparent when deleting the N-terminus. Dependence on the N-terminus for PCNA-Ub deubiquitination may become more apparent when PCNA-Ub is within physiological contexts, such as being DNA bound and interacting with proteins such as polη (5). Further studies may reveal additional layers of specificity for substrates such as PCNA-Ub, which could also be mediated by the N-terminus of USP1-UAF1. Moreover, it remains unclear whether the insertions of USP1 play a more direct role in substrate specificity and deubiquitination.
The N-terminus of USP1 binds specifically to FANCD2, and not FANCI or PCNA. Thus, USP1 appears to display a higher level of substrate targeting for FANCD2. It is interesting that the N-terminus of USP1 is not required for FANCI-Ub deubiquitination in vitro because FANCI shares a similar ubiquitination site with FANCD2 both structurally (39) and on a sequence level, allowing us to speculate that perhaps FANCD2 is the major substrate of USP1. Although we showed that FANCD2-Ub with FANCI and DNA does not obviously affect USP1-UAF1 deubiquitination in vitro, FANCI has been shown to regulate FANCD2 deubiquitination in cells via phosphomimetic mutations (40) and SUMO-like interacting motif mutations (24). Therefore, under certain conditions, FANCI may also play a role in the direct regulation of FANCD2-Ub deubiquitination. For example, we and others have shown that di-monoubiquitinated DNA-bound ID2 complexes remain resistant to deubiquitination (32), suggesting that FANCD2-Ub and FANCI-Ub in the presence of DNA reciprocally regulate their deubiquitination by USP1-UAF1. How di-monoubiquitination of ID2-DNA regulates the ability of USP1-UAF1 to remove ubiquitin from FANCD2 will be an interesting mechanism to investigate. For example, does the ID2-DNA complex affect the USP1 N-terminus binding or directly occlude the ubiquitin tail from proteolytic cleavage?
The N-terminus of USP1 is an extension of the USP domain and is~75 amino acids in length. It is unclear how flexible, structured, or compact the N-terminus will be in a relative position to the USP domain and catalytic site. Interestingly, the N-terminus is modular, as it can be added to other DUB catalytic domains such as USP2 and enhance FANCD2-Ub-specific deubiquitination. Therefore, it is difficult to predict whether the N-terminus of USP1 directly contributes to catalysis when FANCD2 is bound, or simply acts to recruit FANCD2-Ub and stabilise the isopeptide bond within the active site. Because USP1 ΔN -UAF1 loses activity for ubiquitinated K561, but not for other lysines of FANCD2, this allows us to speculate that the N-terminus binds and precisely positions FANCD2 to increase activity for a specific lysine position. Another possibility is that FANCD2-K561-Ub is not in a fully accessible conformation for deubiquitination and that the N-terminus modulates this conformation to allow the formation of a more productive complex. This is reminiscent of a recently identified monoubiquitination site of SETDB1, which protects the ubiquitin from deubiquitination via multiple SETDB1-ubiquitin interactions (41). Perhaps FANCD2 also partially protects and occludes ubiquitin at K561, and the N-terminus of USP1 is required to relieve this effect. It will be interesting to determine how the N-terminus binds to FANCD2, how it modulates USP1 function towards FANCD2, and whether other PTMs or processes within the N-terminus can modulate its function.
It is not understood how the majority of USPs target their substrates specifically. Some USPs target the signal itself; for example, USP18 targets ISG15 (42) and USP30 targets K6-linked ubiquitin chains (17,43). However, here we have an example: USP1, which we have shown targets a specific monoubiquitinated substrate and, therefore, recognises both ubiquitin and the substrate itself. Another specific monoubiquitinated substrate is histone 2B, which is monoubiquitinated on K120 and targeted by the Spt-Ada-Gcn5 acetyltransferase (SAGA) DUB module, which contains USP22 (UBP8 homologue). Whereas UBP8 has contributions for targeting nucleosomes, other members of the complex such as sgf11 make more significant contributions via the acidic patch on H2A/ H2B, and therefore likely govern most of the substrate specificity (27). Whereas SAGA DUB achieves specificity via a protein complex, other DUBs such as USP7 contain substrate-targeting regions within the same polypeptide. USP7 displays substrate-targeting regions via its N-terminal tumor necrosis factor-receptor associated factor-like domain for p53 (44). However, in contrast to the effect we see with the disruption of USP1's N-terminus on FANCD2-Ub deubiquitination, disruption of the N-terminus of USP7 has only a minor effect on p53 deubiquitination in vitro (45). We show that USP1 specifically targets one ubiquitinated lysine made by a specific E3 ligase, FANCL. Another DUB, USP48, was recently published to also specifically oppose the monoubiquitination made by another lysine-specific E3 ligase, BRCA1 (26). This suggests that some DUBs and E3 activities may have coevolved to oppose each other on specific cellular signals and maintain an appropriate equilibrium of non-ubiquitinated and ubiquitinated substrates.
Arguably, the most important known function of USP1 is to regulate FANCD2, as the monoubiquitinated form is important for both ICL repair and origin of replication firing in latent conditions (46). In addition, non-ubiquitinated FANCD2 is important in DNA surveillance mechanisms (47), meaning the presence of both FANCD2-Ub and FANCD2 is important for DNA integrity and cell proliferation. Moreover, USP1 depletion increases oncogeneinduced senescence and plays a pivotal role in protecting genomic instability by preventing FANCD2-Ub aberrant aggregation (12). USP1-UAF1, therefore, works to recycle and maintain appropriate levels of both monoubiquitinated and non-ubiquitinated FANCD2. Inhibition of USP1-UAF1 activity by the specific small molecule ML323 leads to an increased sensitivity to platinum-based compounds in resistant cells (11). Furthermore, the USP1-UAF1 complex has been shown to play critical roles in homologous recombination (10). This collection of studies reveals that USP1 would be an excellent anticancer target by disrupting its ability to deubiquitinate FANCD2-Ub, as this would lead to oncogene-induced senescence and cellular sensitivity to crosslinking agents (8,9). However, USP1-UAF1 targets multiple substrates, including the inhibitor of DNA-binding proteins and TBK1, meaning a catalytic activity inhibitor may affect multiple processes. It is possible that a more specific inhibitor could block FANCD2-USP1 interaction, such as the N-terminus, and may provide a more specific therapy.
All steps of purifications were performed at 4°C and completed within 24-36 h of lysis. Cell pellets were routinely suspended in lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5% glycerol, 10 mM imidazole, 10 mM 2-mercaptoethanol with freshly added MgCl 2 [2 mM], protease inhibitor [EDTA-free] tablets, and benzonase). For FANCD2, FANCI, and FANCL ΔELF , the lysis buffer contained 400 mM NaCl. Sf21 cells were lysed by a homogeniser, followed by sonication at 40% amplitude, 10 s on/10 s off for 12 cycles. Suspended E. coli cells were sonicated at 80% amp, 20 s on/40 s off for 12 cycles. All lysates were clarified at 40,000 g for 45 min and filtered (0.45 μM). Proteins were bound to respective resins (either Ni-NTA for His tags or glutathione [GSH] for GST tags) and washed extensively with lysis buffer with 500 mM NaCl. GST-tagged proteins were eluted by incubation with GST-precision protease or lysis buffer with 10 mM GSH. His-Smt3-tagged proteins were eluted by on-column cleavage using His-ULP1 protease. His-TEV-or His-3C-tagged proteins were eluted by lysis buffer with 250 mM imidazole and lower NaCl for ion exchange chromatography (typically 100 mM NaCl). To remove tags, proteins were incubated with respective proteases (His-TEV protease or GST-precision protease) overnight and tagged proteases were removed by binding to respective resins. Tags were not removed from FANCD2, FANCI, USP2, or GST-ubiquitin. Anion exchange for USP1, UAF1, FANCD2, FANCI, GST-USP2, GST-Ub, and PCNA was performed using a high-performance Q (1 or 5 ml) column and the column eluted with a linear gradient (50 mM Tris, pH 8.0, 100-1,000 mM NaCl, 5% glycerol, and 10 mM 2-mercaptoethanol). E. coli-expressed USP1 was not purified by anion exchange.
Ub-prg BL21 E. coli carrying Ub-intein-CBD was grown to OD 600 : 0.6, induced with 0.5 μM IPTG and expressed for 24 h at 16°C. Cells were harvested and suspended in 50 mM Na 2 HPO 4 , pH 7.2, 200 mM NaCl, and 1 mM EDTA before lysis and clarification. Lysates were bound to chitin resin and washed extensively in 50 mM Na 2 HPO 4 , pH 7.2, and 200 mM NaCl. The resin was washed with wash buffer (20 mM Na 2 HPO 4 , pH 6.0, 200 mM NaCl, and 0.1 mM EDTA) before incubation overnight at 4°C with wash buffer containing 100 mM MESNa. The eluent was collected and a second elution was performed in wash buffer plus 100 mM MESNa. The ubiquitin-MESNa was concentrated to~5 ml and buffer exchanged into 50 mM Hepes, pH 8.0. Ubiquitin-MESNA was either stored at −80°C or directly reacted with 0.25 M propargylamine (prg) for 4 h at 20°C with mild shaking in the dark. Ub-prg was finally purified using SD75 (16/60) in 50 mM Hepes, pH 8.0, and 150 mM NaCl, and flash frozen for storage at −80°C.

Reacting Ub-prg with DUBs
To crosslink DUBs with Ub-prg, DUBs (2 μM) in 20 μl were incubated in DUB buffer (50 mM Tris, pH 7.5, 120 mM NaCl, and 10 mM DTT) with and without Ub-prg (6 μM) for 10 min at room temperature. Reactions were stopped by the addition of SDS-PAGE loading buffer (Invitrogen) and visualised using 4-12% Bis-Tris SDS-PAGE (Invitrogen) run with NuPAGE MOPs SDS running buffer (Thermo Fisher Scientific) and Coomassie staining. A slower migrating band indicates reaction with Ub-prg.

Thermofluor assays
Thermofluor experiments were carried out using a CFX96 real-time PCR detection system (Bio-Rad). Samples containing USP1 (40 μl) at 0.05-0.2 mg/ml in 50 mM Tris, pH 8.0, 120 mM NaCl, 10 mM DTT, and 5× SYPRO orange were dispensed in 96-well plates. The samples were heated from 25°C to 95°C with increments of 1°C/min, and fluorescence was measured at each interval. Data were analysed as previously described (49), and a mean T m°C was calculated from three different USP1 concentrations performed in triplicate.
Fluorescent polarisation (FP) assays with ubiquitin-TAMRA assays USP1-UAF1 samples were prepared by diluting USP1 variants (10 μM) and UAF1 (10 μM) in DUB buffer for 10 min at room temperature. For FP assays, USP1-UAF1 samples were diluted and incubated in DUB buffer with 0.1 mg/ml ovalbumin at 2× concentration indicated in reaction. Typically, reactions with Ub-KG TAMRA (UbiQ) were started by adding and mixing 10 μl of 2× DUB (20 nM) to 10 μl 2× Ub-KG TAMRA (600 nM) to make a final concentration of 300 nM substrate with 10 nM DUB with a volume of 20 μl. Reactions were monitored at 25°C for 1 h by measuring FP at 2-min intervals in 384-well round-bottom corning black plates with a PHERAstar FSX. FP values for each well were fitted using a "one-phase decay" model in Prism (GraphPad). pMax was monitored using Ub-KG TAMRA without DUB and pMin using KG TAMRA .

Purifying monoubiquitinated proteins
Purification of PCNA monoubiquitinated at K164 (PCNA-Ub) was performed as previously described (25). Monoubiquitinated FANCD2 at K561 and FANCI at K523 were purified using GST-Ub, which leaves a "GPLGS" over hang at the N-terminus of ubiquitin following precision protease cleavage. Typically, purified FANCD2 or FANCI (after anion exchange) at 4 μM was monoubiquitinated by equimolar Ube2Tv4, FANCL ΔELF , 50 nM E1, and 8 μM GST-3C-ubiquitin in E3 buffer (50 mM Tris, pH 8.5, 150 mM NaCl, 5% glycerol, 1 mM TCEP, 2.5 mM MgCl 2 , and 2.5 mM ATP) for 30 min at room temperature. xlFANCD2 ubiquitination was performed for 10 min rather than 30 min. Reactions were arrested using apyrase followed by anion exchange chromatography (high-performance Q 1 ml). Importantly, prolonged incubation of FANCD2 in low NaCl concentrations (below~200 mM NaCl) will lead to large losses in protein. GST-Ubsubstrates were bound to GSH resin at 4°C before extensive washes using 50 mM Tris, pH 8.0, 400 mM NaCl, 5% glycerol, and 2 mM DTT. Ubiquitinated proteins were eluted from GSH resin using GSTprecision protease. Finally, ubiquitinated substrates were purified via gel filtration using an S6 Increase (10/300) column in 20 mM Tris, pH 8.0, 400 mM NaCl, 5% glycerol, and 5 mM DTT before concentrating to~5 mg/ml and flash freezing in liquid nitrogen for storage at −80°C. Ubiquitinated FANCD2 or FANCI proteins were visualised on Bis-Tris 4-12% gradient SDS-PAGE gels run in MOPs SDS running buffer, the gels were "overrun" so the 20-kD reference band in the Precision Plus prestained protein ladder (Bio-Rad) was at the bottom of the SDS-PAGE gel. To purify FANCD2-Ub 800 , FANCD2 was monoubiquitinated with ubiquitin 800 instead of GST-Ub, and subsequently purified by anion exchange and gel filtration as above.
Deubiquitination reactions using recombinant substrates USP1-UAF1 complexes were assembled and used to make 2× DUB stocks at 200 nM, or as indicated in Fig 3C. Ubiquitinated substrates were made as 2× stocks: purified hsFANCD2-Ub (2 μM), hsFANCI-Ub (2 μM), hsPCNA-Ub (6 μM), xlFANCD2-Ub (2 μM), and diubiquitin chains (20 μM) were incubated in DUB buffer on ice for 10 min. To initiate hydrolysis, DUBs (200 nM) and substrates (2 or 6 μM) were incubated in a 1:1 ratio (typically 5 μl: 5 μl) for 30 min at room temperature or an indicated time. Reactions were terminated using SDS-PAGE loading buffer (Invitrogen). To analyse deubiquitination, 300 ng of FANCD2-Ub or FANCI-Ub,~600 ng of PCNA-Ub, or 800 ng of diubiquitin was separated on 4-12% Bis-Tris SDS-PAGE gels and Coomassie stained. Deubiquitination was determined by the loss of ubiquitinated proteins or the emergence of non-ubiquitinated substrates and monoubiquitin. For USP2 DUB assays, the intensity of ubiquitinated substrates was measured using a 700-nm channel after Coomassie staining, calculated as mean % of the input and plotted against time, with error bars showing standard deviation from the mean.

DUB-step reactions
DUB-step reactions followed three steps: (1) ubiquitination (see the Ubiquitination reactions section), (2) arrest with apyrase to stop ubiquitination, and (3) deubiquitination. Briefly, FANCD2 (2 μM) ubiquitinated on K561 or KX was immediately diluted 5 μl: 5 μl with DUBs (200 nM) diluted in DUB buffer, giving a final concentration of DUB at 100 nM and FANCD2 at 1 μM. DUB-steps were left at room temperature for an indicated duration (usually 30 min), and 300 ng of FANCD2-Ub 800 was visualised on SDS-PAGE (4-12% Bis-Tris) with an 800-nM channel followed by Coomassie staining. The percentage of residual FANCD2-Ub 800 was calculated as a percentage of the input and plotted for each DUB. For non-fluorescent FANCD2-Ub-FANCI-DNA complexes, 300 ng of FANCD2 was loaded on 4-12% Bis-Tris SDS-PAGE for Coomassie staining and 50 ng for Western blotting.