Mechanism, specificity, and function of FANCD2‐FANCI ubiquitination and deubiquitination

Fanconi anemia (FA) is a rare genetic disorder caused by mutations in any of the currently 22 known FA genes. The products of these genes, along with other FA‐associated proteins, participate in a biochemical pathway, known as the FA pathway. This pathway is responsible for the repair of DNA interstrand cross‐links (ICL) and the maintenance of genomic stability in response to replication stress. At the center of the pathway is the monoubiquitination of two FA proteins, FANCD2 and FANCI, on two specific lysine residues. This is achieved by the combined action of the UBE2T ubiquitin‐conjugating enzyme and a large multicomponent E3 ligase, known as the FA‐core complex. This E2‐E3 pair specifically targets the FANCI‐FANCD2 heterodimer (ID2 complex) for ubiquitination on DNA. Deubiquitination of both FANCD2 and FANCI, which is also critical for ICL repair, is achieved by the USP1‐UAF1 complex. Recent work suggests that FANCD2 ubiquitination transforms the ID2 complex into a sliding DNA clamp. Further ubiquitination on FANCI does not alter this closed‐on‐DNA ID2 conformation. However, the resulting dimonoubiquitinated complex is highly resistant to USP1‐UAF1 deubiquitination. This review will provide an update on recent work focusing on how specificity in FANCD2 ubiquitination and deubiquitination is achieved. Recent findings shedding light to the mechanisms, molecular functions, and biological roles of FANCI/FANCD2 ubiquitination and deubiquitination will be also discussed.

Fanconi anemia (FA) is a rare genetic disorder caused by mutations in any of the currently 22 known FA genes. The products of these genes, along with other FA-associated proteins, participate in a biochemical pathway, known as the FA pathway. This pathway is responsible for the repair of DNA interstrand cross-links (ICL) and the maintenance of genomic stability in response to replication stress. At the center of the pathway is the monoubiquitination of two FA proteins, FANCD2 and FANCI, on two specific lysine residues. This is achieved by the combined action of the UBE2T ubiquitin-conjugating enzyme and a large multicomponent E3 ligase, known as the FA-core complex. This E2-E3 pair specifically targets the FANCI-FANCD2 heterodimer (ID2 complex) for ubiquitination on DNA. Deubiquitination of both FANCD2 and FANCI, which is also critical for ICL repair, is achieved by the USP1-UAF1 complex. Recent work suggests that FANCD2 ubiquitination transforms the ID2 complex into a sliding DNA clamp. Further ubiquitination on FANCI does not alter this closed-on-DNA ID2 conformation. However, the resulting dimonoubiquitinated complex is highly resistant to USP1-UAF1 deubiquitination. This review will provide an update on recent work focusing on how specificity in FANCD2 ubiquitination and deubiquitination is achieved. Recent findings shedding light to the mechanisms, molecular functions, and biological roles of FANCI/FANCD2 ubiquitination and deubiquitination will be also discussed. The Fanconi anemia pathway and FANCD2/FANCI ubiquitination Fanconi anemia (FA) is a rare genetic disorder characterized by bone marrow failure, developmental abnormalities, and predisposition to cancer, while at the molecular level, FA cells are highly sensitive to agents (chemotherapeutic or endogenous aldehydes-primarily formaldehyde) that cause interstrand cross-links (ICLs) in DNA. Such symptoms arise due to mutations (usually biallelic) in any of the 22 currently known Fanconi anemia complementation group (FANC) genes (A, B, C, D1, D2, E, F, G, I, J, L, M, N, O, P, Q, R, S, T, U, V, and W). The proteins encoded by these FA genes, along with FA-associated proteins (FAAPs), participate in pathways that promote genomic integrity and/or prevent genomic instability, and the orchestrated action of these proteins is referred to as FA pathway [1][2][3][4].
The FANCI-FANCD2 (ID2) complex and the role of DNA in ID2 ubiquitination and deubiquitination Both FANCD2 and FANCI are ubiquitinated during the S phase of cell cycle [13,14]. However, an increase in their ubiquitination and their colocalization at damage-induced nuclear foci is observed in response to DNA damage [14][15][16], as well as in response to replication stress [16,17]. Both FANCD2 and FANCI are deubiquitinated by the USP1-UAF1 complex [16,[18][19][20], and USP1 deletion experiments in mice and in chicken cell lines have revealed that USP1-UAF1 activity is required for ICL repair and maintenance of genomic stability [21][22][23]. The genetic disruption of FANCI affects the stability, monoubiquitination, and chromatin localization of FANCD2, and of these, only the FANCD2 stability can be restored when an ubiquitin-deficient mutant of FANCI (K523R) is introduced. Conversely, replacement of wild-type FANCD2 with an ubiquitin-deficient mutant form (K561R) also abolishes both the monoubiquitination and chromatin localization of FANCI [14]. The above suggests that FANCD2 is in a constitutive complex with FANCI and that FANCD2 and FANCI monoubiquitination are interdependent. Accordingly, it is now well established that FANCD2 and FANCI assemble to form a complex, designated as the ID2 complex [14,24]. When in complex with FANCI, FANCD2's in vitro ubiquitination has been found to be greatly stimulated by various DNA structures [25][26][27], as well as by ssRNA and synthetic R-loop structures [28]. Furthermore, disruption of ID2 binding to DNA results in impaired FANCD2 ubiquitination [26,29]. Likewise, FANCI ubiquitination (which typically progresses at a much slower rate than FANCD2 ubiquitination) also gets greatly stimulated by DNA, whether this is in complex with FANCD2 or not [26,30]. The above suggests that upon FANCI/ID2 interaction with DNA or other nucleic acids, ubiquitinated FANCI (I Ub ) and two differentially ubiquitinated ID2 complexes (I Ub D2 and ID2 Ub ) may be initially generated for formation of an I Ub D2 Ub complex. The interaction of ubiquitinated ID2 complexes with DNA also seems to affect subsequent removal of ubiquitin from ubiquitinated ID2 complexes. Two studies have shown that dsDNA protects the I Ub D2 Ub complex from excessive USP1-UAF1 deubiquitination [20,27]. However, another study utilizing single-stranded DNA and partially ubiquitinated ID2 (approx. 60% FANCD2 ubiquitination and minimum/negligible FANCI ubiquitination) shows that DNA may be required for USP1-UAF1mediated FANCD2 deubiquitination [31]. This finding Fig. 1. The Fanconi anemia core (FA-core) complex and UBE2T. (A) Ubiquitination is a three-step enzymatic process, whereby ubiquitin is first activated by an E1 ubiquitin-activating enzyme (usually UBA1), with which it forms a thiol-ester bond (denoted as:~); then, it is transferred to the catalytic cysteine of an E2 ubiquitin-conjugating enzyme (again as a thiol-ester intermediate); finally, ubiquitin is transferred from the E2 enzyme to a lysine of a target protein with the help of an E3 ubiquitin ligase enzyme. Specificity is conferred by the E2-E3 pair. For FANCD2 and FANCI ubiquitination, UBE2T (also known as FANCT) is the E2 enzyme and the FA-core complex is the E3 enzyme. Whereas an ubiquitin chain can be formed by some E2-E3 pairs on target lysines, UBE2T and FA-core catalyze monoubiquitination reactions. (B) Subunit composition of the FA-core complex. Copies of each protein in the core and respective subcomplexes formed from these proteins are shown. (C) Human FA-core complex surface representation structure (PDB: 6KZP) [7]. The complex is shown at three different orientations with each FA protein labeled and shown in a unique color. Active side proteins are indicated with an apostrophe (') and are shown in different shades of the same color, apart from FANCL and FANCL' which are both colored red. The C terminus of FANCE is contrasts with the observation that ID2 Ub complex can be efficiently deubiquitinated in both the presence and absence of DNA [20]. Thus, further studies are required to elucidate whether the type of DNA, differences in protein preparations including the potential for post-translational modifications, or other factors account for such discrepancies. A prerequisite for efficient in vivo ID2 ubiquitination is ATR-mediated FANCI phosphorylation on S556, S559, and S565 (human numbering) [14,[32][33][34][35][36]. Recent biochemical data support the idea that these phosphorylation events have a dual role: They enhance ID2 ubiquitination by stabilizing the ID2 complex on DNA, while at the same time, they protect the complex from USP1-UAF1-mediated deubiquitination [36]. Interestingly though, USP1-UAF1 activity is not only affected by but is also required for optimal FANCI S559 and S565 phosphorylation [35], which suggests that a negative feedback mechanism may be in place to provide a balance between ubiquitination and deubiquitination. While all evidence suggests that FANCI phosphorylation favors the formation/maintenance of ubiquitinated ID2 complex, FANCD2 phosphorylation appears to disfavor this: in vitro and in-cell experiments with phospho-mimetic FANCD2 mutants, supporting the notion that, in the absence of DNA damage, ID2 recruitment to DNA and subsequent ubiquitination is inhibited via CK2-mediated FANCD2 phosphorylation [37].
The FANCD2-FANCI dimer (ID2 complex) has been shown to preferentially bind branched structures of DNA [38] and specifically ICLs [29,39]. A 7.8 A crystallographic electron density map of FANCI bound to a splayed Y DNA provided insights into how this protein may bind branched DNA structures containing both double-stranded and single-stranded DNA [24]. While all ID2-dsDNA structures available show that only FANCI interacts with dsDNA, recent cryo-EM data of the ID2 complex bound to ICL DNA structure indicate that branching of the DNA facilitates a secondary, albeit less stable, DNA-ID2 interface between the branched dsDNA and FANCD2 ( Fig. 2A), which likely contributes to the increased affinity of ID2 to branched, over nonbranched, DNA structures [39].
The lysines targeted for FANCD2 and FANCI monoubiquitination are within solvent-accessible tunnels of the ID2 interface [24,39] (Fig. 2A). While the lysines are not buried, they are inaccessible for UBE2T ubiquitination without ID2 conformational remodeling. Nevertheless, the presence of FANCI greatly stimulates FANCD2 ubiquitination (instead of restricting this), as shown in FANCL-catalyzed [12,25,40], BL100-CEF-catalyzed [27], and FA-core-catalyzed [6][7][8] FANCD2/ID2 ubiquitination reactions. Moreover, FANCI further restricts the site of FANCD2 ubiquitination to the physiologically modified lysine of the latter [40]. Since DNA can stimulate FANCD2 in vitro monoubiquitination within the ID2 complex [6,12,[25][26][27]29], irrespective of the source of E3 ligase used (FANCL, BL100-CEF, or FA-core complex), it has been suggested that DNA may be responsible for remodeling the ID2 complex to allow access for the E2/E3 machinery [25]. Nevertheless, no major ID2 conformational changes have been identified upon DNA binding, based on the highly similar cryo-EM structures of apo-ID2 complex and the ICL-bound ID2 complex [39]. FANCL, along with various components of the FA-core complex and ubiquitin-charged UBE2T, is hence assumed to promote such conformational change, once ID2 is bound to the DNA. Indeed, that is what recent cryo-EM structures of FA-core-UBE2T-ID2-DNA complexes suggest [7] (Figs 2B and 3B). Moreover, in vitro studies have shown that, while FANCL and BL100 have comparable E3 activities against isolated FANCD2 (whose ubiquitination does not require DNA) [12], BL100 (and/or BL100-CEF) along with DNA is required for efficient ubiquitination of the ID2 complex [8,27]. The FA-core-UBE2T architecture and the contribution of each of the FA-core-UBE2T components for site-specific ID2 ubiquitination will be further discussed below.
The role of UBE2T and the FA-core complex in ID2 ubiquitination As mentioned above, UBE2T coupled with BL100 can support in vitro ID2 ubiquitination, in the absence of other FA-core subunits, which indicates that AG20 and CEF play an auxiliary to BL100 function. A FANCA asymmetric dimer is presumably required for disordered in the structure, and hence, only the N-terminal part of FANCE (E NTD ) is shown. (D) FANCL RING binding to UBE2T's core/UBC (ubiquitin-conjugating) domain allosterically activates UBE2T, by reconfiguring the positions of three basic residues specifically targeting an acid patch on FANCD2 around its acceptor lysine (K561). Human UBE2T Core (UBC domain only) apo structure (PDB: 1YH2) [82] was superimposed to human UBE2T Core bound to the RING domain (residues 289-375) of FANCL (PDB: 4CCG) [11], using PyMOL. UBE2T's catalytic cysteine (C86), and the three basic residues (R84, K91, and K95) forming the basic triad are shown and illustrated as sticks.

4814
The the assembly of an active asymmetric FA-core complex. In support for this, a large proportion of FA pathological missense mutations are located within FANCA dimerization domains; moreover, lack of FANCA dimerization is associated with assembly of a twofold symmetric FA-core complex, which is predicted to be inactive [7]. FANCA dimerization occurs via the C termini of two FANCA molecules [7,41], whereas the two FANCG molecules bridge the FANCA dimer (on which FAAP20 also binds) to the dimeric BL100 module [7] (Fig. 1C). BL100, which is a dimer of trimers, composed of two copies of FANCB, FANCL, and FAAP100 [6,7,27,42], and bridges the dimeric AG20 module to a single copy of CEF [7]. The above interactions result in an asymmetric FAcore complex, with an active side primed for UBE2T and ID2 binding [7] (Figs 1C and 3A-C). The structural configuration of the asymmetric FA-core complex enables the active side FANCL (FANCL') and a C-terminal domain (CTD) of FANCE (FANCE CTD , which is flexible in the absence of ID2) to bind and remodel ID2-DNA for FANCD2 ubiquitination, whereas the inactive (within FA-core) FANCL binds the N-terminal segment of FANCE, along with FANCC and FANCF [7] (Figs 1C and 3A-C). The asymmetry might be crucial to ensure that only one copy of CEF complex will be bound to the BL100 module, thus allowing (the active side) FANCL' to engage with FANCE CTD and UBE2T for ID2 remodeling and ubiquitination [7] (Fig. 3A,B). Nevertheless, addition of the AG20 module in BL100-CEF-catalyzed ID2 ubiquitination reactions has not been found able to further improve FANCD2/FANCI in vitro ubiquitination [27]. Furthermore, the reaction rate of ID2 ubiquitination with chicken proteins is comparable between BL100 and FA-core-catalyzed reactions [6,8].
The above suggests that AG20 is not directly involved in catalysis. On the other hand, human CEF has been reported to significantly enhance in vitro BL100catalyzed ubiquitination of FANCD2 and FANCI, but only when these proteins exist in the ID2 heterodimer [27]. However, this function may not be very well conserved among vertebrates, since CEF appears to be dispensable for maximal ID2 ubiquitination in in vitro reactions with chicken proteins [6,8]. Earlier reports have shown that CEF is required for recruiting the BL100 unit to sites of DNA damage [43] and for bridging the ID2 complex to BL100 [27]. In a nonredundant fashion, the AG20 module also works to recruit BL100 to sites of DNA damage, but independently of CEF: Simultaneous knockdown of both FANCG and FANCF results in a complete abolishment of FA-core complex chromatin recruitment and FANCD2 monoubiquitination; however, single deletion of FANCG or FANCF does not [43]. As mentioned above, FANCE CTD (residues 273-536) interacts with the N terminus (residues 91-311) of FANCD2, acting thus as a substrate adaptor protein [7,44,45] ( Fig. 3C). This interaction has been additionally proposed to result in a reshaped, more stable ID2 interface [42]. Upon FANCE CTD -FANCD2 binding, a rotation of a~160 amino acid segment of FANCD2 occurs, contributing to partial exposure of FANCD2's K561 [7]. Recently, it has been shown that CEF can also bind to FANCI, in the absence of FANCD2 [27]. However, it remains to be established what function this interaction may serve, since no FANCE-FANCI interaction has been observed upon FANCD2 ubiquitination [7]. For site-specific FANCD2/FANCI ubiquitination, FANCL is the RING E3 ligase, which binds and activates UBE2T for site-specific FANCD2/FANCI ubiquitination. The crystal structure of the drosophila fulllength FANCL revealed that this protein contains, apart from a RING domain, also an ELF (E2-Like Fold), as well as a DRWD (Double RWD) domain (Fig. 3D); the latter was found capable of binding both FANCI and FANCD2 [46]. The equivalent DRWD domain in human FANCL was later crystallized and designated as URD (UBC-RWD Domain). This has a similar structure to the drosophila FANCL DRWD (despite differing significantly in amino acid conservation) and likewise binds both FANCD2 and FANCI substrates in vitro [47]. Indeed, FANCL's URD, also referred to as RWD2-RWD3 domain, has been shown to bind FANCI, in recent cryo-EM FAcore-UBE2T-ID2 structures, captured prior to FANCD2 ubiquitination [7]. The main function of the ELF domain, designated also as RWD1 in human FANCL, appears to be in binding FANCB-FAAP100 for formation of a BL100 complex [7,27,42]. However, Fig. 3. Structural basis of FANCD2 ubiquitination by the FA-core and UBE2T. During FA-core-catalyzed FANCD2 ubiquitination, a C-terminal segment of FANCE (E CTD ) binds to FANCD2, while FANCL inserts into the ID2 interface in proximity to FANCD2' K561. The above interactions, along with FANCI-binding to dsDNA, result in ID2 transitioning form an open-to a closed-state conformation. In the closedstate ID2 structure, UBE2T is able to engage with FANCD2. As a result, the cysteine of UBE2T carrying the ubiquitin (C86) is positioned within proximity to the target lysine of FANCD2 (K561) for ubiquitin transfer. (A) Surface representation of FA-core-UBE2T bound to ID2-dsDNA (open-state ID2; PDB: 6KZS) [7]. The overall complex is shown at two orientations with each FA protein labeled and shown in a unique color. Active side proteins are indicated with an apostrophe (') and are shown in different shades of the same color, apart from FANCL and FANCL' that are both colored red.  (Fig. 3A,B,D). Accordingly, an intact ELF is essential for in vivo FANCD2 ubiquitination [48]. The ELF domain of FANCL is also capable of binding ubiquitin's Ile44 patch [48], but it is currently unknown whether there is a FANCLubiquitin-binding step during ID2 ubiquitination. FANCL's RING domain plays role in ID2 remodeling prior to FANCD2 ubiquitination; a hairpin of FANCL' RING domain inserts into the ID2 interface in proximity to FANCD2's K561, interacting with both FANCD2 and FANCI (Fig. 3E), as demonstrated in a recent cryo-EM structure of ID2 bound to FA-core complex [7]. Crucially though, the RING domain of FANCL binds and recruits UBE2T with high specificity (Figs 1D and 3F). This is achieved by a combination of electrostatic and hydrophobic interactions, which are unique for the UBE2T-FANCL RING pair [11].
FANCL's RING interaction with UBE2T optimizes several uniquely positioned positively charged residues within UBE2T (Fig. 1D) to drive specific lysine selection on FANCD2 (K561) via recognition of a conserved negatively charged patch adjacent K561 [7]. Indeed, the importance of these interactions for FANCD2 ubiquitination has been demonstrated with charge-reversal mutations on UBE2T's catalytic basic triad. Such mutations impair FANCD2 ubiquitination, but FANCD2 ubiquitination can be restored if reverse-charge mutations are also applied on FANCD2's acidic patch [12]. Moreover, these acidic residues are aligned within proximity to UBE2T basic residues prior to ubiquitin transfer, in a recent cryo-EM FA-core-UBE2T-ID2-DNA structure [7] (Fig. 3F). Lastly, targeted mutagenesis on UBE2T (E54R, P93G, and P94G mutations), based on the allosteric conduit identified, can also result in optimization of UBE2T's charged residues for enhanced FANCD2 ubiquitination (in both the presence and the absence of FANCL), without compromising site specificity [12]. UBE2T contains, apart from its conserved core (UBC) domain, also a~40 amino acid Cterminal extension appended to this. Although it is unclear whether this region has any role in FANCI ubiquitination, all available evidence suggests that this is dispensable for FANCD2 ubiquitination. Previous studies have shown that UBE2T is ubiquitinated on both K91 and that C-terminal extension [10,40], but a UBE2T mutant lacking these sites (UBE2T-ΔC-K91R) efficiently supports FANCD2 ubiquitination, both in vitro and in vivo [40]. Moreover, no density for this region has been identified in recent cryo-EM FA-core-UBE2T-ID2-dsDNA structures [7].
As mentioned above, BL100 results in more efficient ID2 ubiquitination than FANCL alone. Although it is currently unclear why this occurs, it is likely that FANCL is more stable within a BL100 complex. FANCB is responsible for dimerization of the BL100 complex, while FAAP100, which adopts a similar to FANCB fold, most likely stabilizes FANCB (and hence the dimeric BL100 complex) and is additionally required for bridging BL100 to AG20 [6,7,27,42]. However, the recruitment of two FANCL molecules by the FANCB-FAAP100 tetramer does not result in simultaneous ubiquitination of FANCD2 and FANCI within an ID2 complex, as previously hypothesized [42]. The BL100 dimeric architecture within the FAcore is not fully symmetric, but instead adopts a pseudo twofold symmetry [6,7]. Furthermore, several studies point to the fact that that FANCD2 is the preferred substrate of ubiquitination within the ID2 complex [7,14,[25][26][27]29]. However, in the absence of FANCI, FANCD2 is a poor substrate for FA-corecatalyzed ubiquitination [6][7][8]. This may be due to FANCD2's ability to form homodimers, in which the target lysine is buried in the FANCD2-FANCD2 interface, as indicated in a recent cryo-EM structure of recombinant chicken FANCD2, coupled with sizeexclusion FANCD2/FANCI/ID2 experiments [49]. Alternatively (or additionally), FANCI's interaction with FANCL and/or UBE2T [7] may be required for efficient targeting of FANCD2's K561 by the FA-core complex.
While the FA-core complex with UBE2T seems to specifically open the ID2 dimer for FANCD2 ubiquitination [7], FANCD2 ubiquitination is required for efficient FANCI's ubiquitination. Indeed, substitution of FANCD2 with a nonubiquitinatable lysine-to-arginine mutant results in complete loss of FANCI ubiquitination in in vitro BL100-or BL100-CEF-catalyzed ID2 ubiquitination reactions [27]. Furthermore, in an in vitro setting with genetically engineered UBE2T, FANCI, and FANCD2 proteins, 'switching off' only FANCD2 ubiquitination results in decreased FANCI ubiquitination, while 'switching back on' FANCD2 ubiquitination significantly restores FANCI ubiquitination [12]. Recent cryo-EM structures of FANCD2ubiquitinated ID2 complex have shed light into how FANCD2 ubiquitination can enhance FANCI ubiquitination: ubiquitin conjugation on FANCD2's lysine 561 is associated with major FANCD2 conformational changes [30,39,49], resulting in exposing FANCI's lysine 523 for ubiquitination [39] (Fig. 4). This lysine is likely targeted by UBE2T by a similar allosteric mechanism of FANCL-mediated UBE2T activation [12]. However, it remains to be determined how FA- core and UBE2T specifically engage with ID2 Ub (or ID2) for FANCI ubiquitination. Interestingly, and contrary to human FANCD2, human FANCI in isolation can be efficiently ubiquitinated by the FA-core complex [7]. The nearly 10-fold excess protein abundance of FANCI over FANCD2, reported in U2OS cells [50], indicates that FANCI may be ubiquitinated in isolation to some extent in cells too. However, a substantial fraction of cellular FANCI may not be available for ID2 formation and ubiquitination. Apart from participating in the FA pathway, FANCI also functions in ribosome biogenesis in the nucleolus and such function requires neither FANCD2 nor ubiquitination [51]. Furthermore, all available in vivo data suggest that the presence of ubiquitinatable FANCD2 is required for efficient FANCI ubiquitination: Decreased FANCI protein levels and undetectable FANCI ubiquitination have been reported in the absence of cellular FANCD2, whereas introduction of a K561 FANCD2 mutant to these cells could restore only the FANCI cellular levels, but not FANCI ubiquitination [14]. Hence, FANCD2 ubiquitination is required for initiation and/or maintenance of FANCI ubiquitination.

The molecular function of FANCD2 and FANCI ubiquitination
Although it has been known since 2001 that FANCD2 becomes ubiquitinated in response to DNA damage, the molecular function of this modification had been elusive for nearly two decades. Ubiquitination of the ID2 complex has been long hypothesized to result in recruitment of ubiquitin-binding factors, essential for ICL repair or restoration of replication. Ubiquitin-binding proteins typically recognize a hydrophobic patch on ubiquitin that includes Leu8, Ile44, and Val70 [52]. However, the recent cryo-EM structure of human dimonoubiquitinated ID2 complex bound to DNA (I Ub D2 Ub -DNA) revealed that both conjugated ubiquitins have this hydrophobic patch buried within the I Ub D2 Ub interface [39] and are thus not easily accessible to ubiquitinbinding proteins. Indeed, another study utilizing pulldown binding experiments showed that ID2 ubiquitination had no effect on ID2 interactions with several DNA-repair proteins that contain ubiquitin-binding domains [53]. Three independent studies have now shown that upon FANCD2 ubiquitination, major FANCD2 conformational changes occur, resulting in a newly formed ID2 interface, formed by the two Cterminal 'arms' of FANCI and FANCD2 (Arm interface), and a monoubiquitinated ID2 complex (ID2 Ub ) encircling double-stranded DNA (dsDNA) [30,39,49] (Fig. 5). The resulting ID2 Ub complex has significantly increased affinity for dsDNA, compared with the nonubiquitinated ID2 complex [30,49]. Further ubiquitination on FANCI does not seem to impact on the overall ID2 Ub structure and coupled DNA affinity, since the dimonoubiquitinated (I Ub D2 Ub ) complex also displays a comparable dsDNA affinity and overall ID2 conformation to the ID2 Ub complex [30,39] (Fig. 5). No specificity for a particular DNA sequence has been reported for ubiquitinated/nonubiquitinated ID2. Accordingly, in all available ubiquitinated ID2-dsDNA cryo-EM maps, the dsDNA is less well resolved, which indicates that dsDNA populates multiple positions/orientations upon ID2 binding. Furthermore, structural evidence suggests that the I Ub D2 Ub complex is able to bind only to a dsDNA stretch (of about 30 kb) of various DNA structures, due to electrostatic interactions occurring between a semicircular I Ub D2 Ub basic groove and the acidic DNA backbone [39]. The above suggests that ubiquitinated ID2 may act as a sliding DNA clamp. Indeed, evidence indicates that the I Ub D2 Ub complex is able to slide on dsDNA, since it can come off a linear stretch of dsDNA, but cannot disengage from a circular dsDNA of the same length [39]. DNA sliding has been suggested to enable the ID2 complex to travel away Fig. 4. FANCD2 ubiquitination leads to structural rearrangements in the ID2 complex, resulting in K523 of FANCI being more accessible for ubiquitination. Structural changes upon FANCD2 ubiquitination occurring in the ID2 complex surrounding K523 of FANCI. The FANCI-FANCD2-DNA complex structure (PDB: 6VAA [39]; displayed as cartoon and surface) was superimposed to FANCI-FANCD2 Ub -DNA complex (PDB: 6VAF [39]; shown as cartoon only), and centered on K523 of FANCI (colored as pink in 6VAA and dark red in 6VAF) using PyMOL. The superimposition of the two structures shows that FANCD2 helices move away from FANCI's K523, while the FANCI helix containing K523 is also shifted away from FANCD2.   [39]. Each complex is shown at three different angles. FANCD2 ubiquitination stabilizes a secondary FANCD2-FANCI interface formed between the C-terminal 'arms' of each protein (Arm ID2 interface), resulting in the ID2 complex encircling dsDNA. The FANCD2 and FANCI lysines that get modified (K561 and K523, respectively) are indicated in red. FANCD2 (green), FANCI (cyan), DNA backbone (orange), ubiquitin conjugated to FANCD2 (black), and ubiquitin conjugated to FANCI (dark blue). The dissociation constants (Kd values) of each complex for dsDNA in solution, as calculated from protein-induced fluorescence enhancement (PIFE) data [30], are shown. The relative resistance (low/high) of FANCD2's ubiquitin to USP1-UAF1 deubiquitination activity is also indicated for mono (ID2 Ub -DNA) and dimonoubiquitinated (I Ub D2 Ub -DNA) ID2 complexes. from ICLs, replication forks, or other related DNA structures, on which ID2 ubiquitination has occurred [39]. In vitro studies have shown that, upon ubiquitination, multiple ID2 complexes can accumulate on long stretches of dsDNA (150 bp or longer), forming filament-like structures; the latter have been suggested to act as a DNA-shield, protecting the free dsDNA from the unwanted action of nucleases, helicases, or other factors [53]. Indeed, FANCD2 ubiquitination has previously been shown to protect stalled replication forks from degradation [54]. However, it is not yet known whether ID2 complexes form such filament-like structures in vivo.
The fact that FANCI ubiquitination does not further change the overall structure and associated dsDNA affinity of the ID2 Ub complex suggests that FANCI ubiquitination may have a complementary function to FANCD2 ubiquitination. Indeed, in vitro deubiquitination experiments with USP1-UAF1 and either ID2 Ub -DNA or I Ub D2 Ub -DNA substrates have shown that FANCI ubiquitination largely protects FANCD2's ubiquitin from USP1-UAF1 activity [30]. This may also explain why introducing a nonubiquitinatable FANCI mutant (K523R) in FANCD2 ubiquitination-proficient cells results in greatly reduced FANCD2 ubiquitination [14]: The ubiquitin conjugated to FANCD2 would not be sufficiently protected against USP1-UAF1 activity, when the FANCD2 forms a complex with FAN-CI K523R . Key for this protection appears to be FANCI's ubiquitin interaction with FANCD2, which involves the Ile44 hydrophobic patch of ubiquitin [39]. Indeed, I44 alanine substitution of FANCI's conjugated ubiquitin makes FANCD2's ubiquitin more susceptible to USP1-UAF1 activity [30]. In the I Ub D2 Ub -DNA complex, FANCI's ubiquitin is also partially protected from USP1-UAF1 activity, since FANCI's deubiquitination appears to be even slower than FANCD2's [27,30]. Although the precise reason for this is currently unknown, the I Ub D2 Ub -DNA complex structure suggests that the ubiquitin conjugated to FANCD2 is less buried than the ubiquitin conjugated to FANCI [39,55] and hence may be more accessible for USP1-UAF1 deubiquitination. The following section will focus on how USP1-UAF1 interacts with the FANCD2-ubiquitinated ID2 complex for FANCD2 deubiquitination.

Mechanism of FANCD2 deubiquitination
As previously mentioned, FANCI/FANCD2 deubiquitination is specifically catalyzed by the USP1-UAF1 complex. USP1, which belongs to the ubiquitin-specific protease (USP) class of deubiquitinating enzymes, is the subunit binding to ubiquitin and catalyzing its removal from (the conjugated with that) proteins. As revealed in recent crystal and cryo-EM structures of USP1 [55], this has a typical USP-fold, consisting of Fingers, Palm, and Thumb subdomains [56]. Apart from its USP-core domain, USP1 also has two insertions and an Nterminal extension appended to this domain. On its own, USP1 has little activity and hence requires activation by UAF1 (USP1 Associated Factor 1; also known as WDR48) for efficient catalytic turnover [19]. UAF1, which can also form a complex with and activate USP12 and USP46 [57], binds and activates all USP partners via its WD40-repeat b-propeller domain [19,57]. This domain specifically interacts with the finger domain of USPs, as demonstrated in recent UAF1-USP12 [58,59], UAF1-USP46 [60], and UAF1-USP1 [55] structures. While USP12 and USP46 also require a second activator (WDR20) for maximal activity, USP1 does not [61,62]. However, recent work has indicated that part of USP1's low intrinsic activity is due to autoinhibition conferred by the two insertions to its USP domain, which is alleviated upon binding to UAF1 [63]. A 54 amino acid segment within USP1's N-terminal extension is required for efficient cleavage of ubiquitin from FANCD2's K561, but is dispensable for FANCI's K523 ubiquitin removal [20].
Although limited information exists on the role of UAF1's Ancillary domain, the SUMO-like domain (SLD) of UAF1 has been proposed to bind a putative SUMO-interacting motif (SIM) on FANCI, and a SIM peptide sequence on the PCNA-binding protein hELG1. Accordingly, replacement of wild-type UAF1 with a mutant UAF1 version lacking part of its SLD, in chicken DT40 cells, results in impaired FANCD2 and PCNA deubiquitination [64].
A recent cryo-EM structure of a human USP1-UAF1-ID2 Ub -DNA complex has provided insights into how the USP1-UAF1 complex deubiquitinates FANCD2 [55]. This structure has revealed that USP1 extracts the ubiquitin from the FANCD2 Ub -FANCI interface, to enclose it into USP1's Palm and Thumb USP subdomains. As a result, the ubiquitin is 160°r otated relative to its position in the ID2 Ub -DNA complex (without USP1-UAF1), and the isopeptide bond between K561 of FANCD2 and G76 of ubiquitin is in proximity to USP1's active site C90 for cleavage (Fig. 6A). The structure additionally revealed two important UAF1-FANCI interfaces: UAF1's SUMOlike domain (SLD) interacting with FANCI's 249-260 amino acid loop (Fig. 6A,B) and UAF1's b-propeller interacting with FANCI's 547-576 amino acid loop (Fig. 6A,C). The importance of these interactions is supported by associated biochemical data, whereby deletion of UAF1's SLD or phospho-mimetic mutations on FANCI at FANCI-b-propeller interface impairs FANCD2 deubiquitination [55]. The SIM-like sequence of FANCI that interacts with UAF1's SLD (DELLDVV) is different and far away from the one previously postulated (VIPLQQGEEEEEEEE) [64]. It From those, the ones that likely interact using their side chain are labeled (cyan for FANCI and salmon for UAF1). A part of FANCI's 547-576 amino acid segment (in proximity to UAF1's b-propeller) not modeled in the structure (residues 555-562) due to poor density, is shown as dotted line. S565 (also displayed, labeled and highlighted with an orange asterisk) has been previously shown to be phosphorylated in an ATR-dependent way. (D) USP1-UAF1 complex likely engages the I Ub D2 Ub complex in the same way as the ID2 Ub complex (for FANCD2 deubiquitination), since superimposition of I Ub D2 Ub -DNA structure (PDB: 6VAE) [39] to the USP1-UAF1-ID2 Ub -DNA structure (PDB: 6AY1) [55] does not reveal any clashes of FANCI's ubiquitin (blue spheres) with UAF1. is thus unclear how deleting part of the latter sequence results in increased levels of ubiquitinated FANCD2 in cells [64], since this has shown to have no effect on in vitro FANCD2 deubiquitination [55]. Upon USP1-UAF1 binding, the ID2 Ub -DNA complex is still in the closed-state conformation [55], whereby the Arm ID2 interface is formed around dsDNA by the FANCI and FANCD2 C termini [7,39,49,55] (Figs 2A, 5 and 6 (Fig. 7). Because FANCI ubiquitination is not predicted to occlude binding of UAF1 to FANCI (Fig. 6D), it has been suggested that USP1-UAF1 targets the ubiquitin of FANCD2 in dimonoubiquitinated ID2 complex (I Ub-D2 Ub ), in the same manner as in the ID2 Ub complex [55]. However, some subtle conformational changes upon FANCI ubiquitination may make this binding less effective. Indeed, pull-down experiments using a catalytically inactive GST-USP1-UAF1 as bait have revealed that USP1-UAF1 binds preferentially to ID2 complexes containing nonubiquitinated (rather than ubiquitinated) FANCI [27]. This is consistent with FANCD2 being deubiquitinated less efficiently in the I Ub D2 Ub -DNA complex than in the ID2 Ub -DNA complex [20,30]. Further, biochemical experiments coupled with a USP1-UAF1-I Ub D2 Ub structure are required to explain how FANCI's ubiquitination interferes with USP1-UAF1catalyzed FANCD2 deubiquitination. More work is also required to answer how the N terminus of USP1 facilitates removal of ubiquitin from FANCD2, since no clear density for this USP1 segment was observed in the USP1-UAF1-ID2 Ub -DNA structure. Lastly, it is still not understood how interaction of ubiquitinated ID2 with DNA can influence USP1-UAF1-mediated deubiquitination. Despite the reported ability of USP1-UAF1 complex, or individual USP1/UAF1 proteins, to bind various DNA structures [31,65], no such interactions were observed in the USP1-UAF1-ID2 Ub -DNA structure [55] (Fig. 6A).

Molecular and cellular functions of FANCI/FANCD2 deubiquitination
The role of USP1-UAF1 might be more complicated than simply removing the ubiquitins in a timely manner. At least one of the roles of USP1-UAF1 action (on ubiquitinated ID2 complexes) appears to be in enriching a dimonoubiquitinated ID2 (I Ub D2 Ub ) population [20,27,30]. This is achieved through USP1-UAF1 preferentially targeting the ID2 Ub -DNA complexes, which are more susceptible to cleavage than the more resistant I Ub D2 Ub -DNA complexes [20,30]. Indeed, it has been shown that any imbalance in the ratio of ubiquitinated FANCD2 to ubiquitinated FANCI, caused by a faster rate of FANCD2 ubiquitination, can be corrected by USP1-UAF1 in vitro; FANCD2 has been shown to be deubiquitinated faster than FANCI, until an equilibrium can be reached whereby ubiquitinated FANCD2 and FANCI become equimolar [27]. USP1-UAF1 activity may also regulate the levels of ID2 in chromatin, by effectively reducing its affinity for dsDNA (since ID2 binds weaker to dsDNA than ID2 Ub and I Ub D2 Ub ). Indeed, loss of USP1 activity results in accumulation of FANCD2 on chromatin [21,22]. However, USP1-UAF1 may also act on ubiquitinated ID2 complex that has been removed from chromatin, since DNA-free dimonoubiquitinated ID2 complexes have been reported to be more susceptible to USP1-UAF1 activity [20,27]. The DVC1-p97 ubiquitin segregase complex has been proposed to remove from chromatin, ubiquitinated ID2 complexes that are additionally SUMOylated and subsequently polyubiquitinated, as a way of controlling ID2 dosage at DNA damage sites [66].
The biological role of USP1-UAF1 action on ID2 is less straightforward, since deleting or impairing expression of USP1 or UAF1 has consequences that are not restricted to ID2 ubiquitination. Apart from acting on monoubiquitinated FANCD2 and FANCI, the USP1-UAF1 complex is also known to remove ubiquitin from monoubiquitinated proliferating cell nuclear antigen (PCNA) [20,67], and recently, many more USP1-UAF1 targets have been discovered [68][69][70][71][72][73][74][75][76][77]. Unlike UAF1 deletion, which is embryonically lethal in mice [78], USP1 deletion is viable, but results in various defects, including increased perinatal lethality, male infertility, cross-linker hypersensitivity, and an FA-like phenotype [22]. At the cellular level, blocking USP1 activity leads to increased cytotoxicity in response to UV and cisplatin, as well as impaired homologous recombination and sister-chromatid exchange [79]. Many of these effects, though, may be attributed to functions of USP1-UAF1 on other proteins than ID2, and indeed, USP1 inhibition has been shown to affect both the FA pathway and DNA translesion synthesis, by inhibiting both FANCD2 and PCNA deubiquitination [79]. Hence, introduction of a USP1 mutant that is specifically defective against ubiquitinated FANCD2  (and/or FANCI), in cells, may be particularly useful for determining the biological role of USP1-UAF1 in the FA pathway. To this effect, deletion of targeted mutagenesis on USP1 N terminus has been shown to specifically impact on FANCD2 deubiquitination, without having much of an effect on FANCI or PCNA deubiquitination [20]. Similarly, a cellular UAF1 mutant that specifically blocks UAF1-FANCI interaction [55] (Fig. 6A-C) may be also useful in assessing USP1-UAF1's biological role in deubiquitinating FANCD2. Despite any wide-reaching effects of USP1 depletion, USP1 activity seems to be specifically required for formation of FANCD2 damage foci. Indeed, a defect of both basal (S phase-specific) and DNA damage-induced FANCD2 foci has been observed upon USP1 deletion in mouse embryonic fibroblasts [22] and HeLa cells [80]. Moreover, treatment of U2OS cells with a USP1-specific inhibitor is also associated with a reduction in DNA damage-induced FANCD2 foci [79]. These effects are not due to impaired cH2AX focus formation (which are required for formation of FANCD2 foci [81]), since cH2AX focus formation was found to be unaffected by the loss of USP1 [22,80]. Moreover, USP1 deletion does not impair FANCD2 chromatin localization. In fact, loss of USP1 is associated with increased ubiquitination and enrichment of FANCD2 in chromatin [21,22]. Hence, USP1 action is likely required for correct localization of ID2 within the chromatin. One possibility of how this may occur is by USP1-UAF1 effectively limiting the population of ID2 that is ubiquitinated. This may be necessary to restrict the I Ub D2 Ub population to sites of DNA damage or replication arrest (where ubiquitination occurred), as opposed to a widespread distribution across the chromatin. Clearly, more work is required to determine whether USP1-UAF1 action contributes to the formation or maintenance of ID2 foci, and how this achieved.

Perspective
Our knowledge of the requirements and mechanism of site-specific FANCD2 ubiquitination and deubiquitination has greatly expanded over the last few years. Very recently, structural and biochemical studies have also revealed that ubiquitin conjugation on FANCD2 facilitates clamping of ID2 complex on DNA. Furthermore, cryo-EM structures have provided insights into how the FA-core complex engages ID2 for FANCD2 ubiquitination, and how USP1-UAF1 engages the ID2 Ub complex for ubiquitin removal. What is less clear is how the FA-core complex targets the ID2 complex for FANCI ubiquitination, and how this ubiquitin is subsequently removed by USP1-UAF1. FANCI ubiquitination has been found to largely protect the ubiquitin of FANCD2 from USP1-UAF1 activity. However, this may not be the only function of FANCI ubiquitination. That is because an ID2 complex where only FANCI is ubiquitinated (I Ub D2) may be formed upon FANCI/ID2 ubiquitination and/ or I Ub D2 Ub deubiquitination. Our knowledge on such a complex is only that it also has a higher affinity for dsDNA than the ID2 complex, albeit not as high as that of ID2 Ub /I Ub D2 Ub complexes [30]. What is also less clear is the mechanism by which DNA can influence ID2 deubiquitination. Lastly, despite the fact that (at least part of) the molecular functions of FANCD2/ FANCI ubiquitination have now been uncovered, it is still poorly understood how FANCD2/FANCI ubiquitination and deubiquitination actually contribute to ID2 foci, ICL repair, and maintenance of genomic stability. Clearly, more work is required to answer these questions and improve further our understanding of the mechanisms and roles of ID2 ubiquitination and deubiquitination. Instrumental for these are expected to be the recent advances in preparation of ubiquitinated FANCI and FANCD2, coupled with the ability to now obtain high-resolution cryo-EM structures of protein complexes.