Introduction

The Fanconi anaemia (FA) pathway is responsible for repairing DNA interstrand crosslinks (ICLs) and ensuring that genome stability is maintained when replication is stalled (Nalepa & Clapp, 2018). A crucial and central step in this pathway is the mono‐ubiquitination of FANCD2 and FANCI on specific lysines (K561 and K523, respectively, for human proteins) catalysed by a multi‐component ubiquitin ligase (FA‐core complex) and the UBE2T ubiquitin‐conjugating enzyme (Lemonidis et al, 2022). The two ubiquitination events are interdependent, since mutation on either of the two lysines results in greatly impaired in‐cell ubiquitination on the other lysine (Sims et al, 2007; Smogorzewska et al, 2007). Removal of the two ubiquitins, through isopeptide cleavage by the USP1‐UAF1 complex, is also required for ICL repair and maintenance of genomic stability (Oestergaard et al, 2007; Kim et al, 2009; Murai et al, 2011).

Several biochemical data (Sato et al, 2012; Longerich et al, 2014; Rajendra et al, 2014; Chaugule et al, 2020; Rennie et al, 2020) and recent structural evidence (Wang et al, 2021) indicate that FANCD2 is the preferred substrate for ubiquitination and that FANCI ubiquitination likely occurs once FANCD2 has been ubiquitinated. Upon binding to the FA‐core‐UBE2T, the FANCI‐FANCD2 (ID2) complex closes on DNA, and this ID2 remodelling exposes and brings K561 of FANCD2 in proximity to UBE2T's catalytic cysteine for ubiquitination (Wang et al, 2021). Recent structural evidence with chicken ID2 proteins, whereby wild‐type FANCI is replaced with a S558D/S561D/T567D phosphomimetic mutant, suggest that FANCI phosphorylation may also facilitate this ID2 closure on DNA, in the absence of FA‐core‐UBE2T (Sijacki et al, 2022). The ID2 closure on DNA is maintained upon FANCD2 ubiquitination (Alcón et al, 2020; Rennie et al, 2020; Wang et al, 2020). The resulting ID2Ub‐DNA complex can be susceptible to USP1‐UAF1‐mediated deubiquitination. However, further ID2 ubiquitination on FANCI results in enhanced protection of FANCD2's ubiquitin from USP1‐UAF1 action (Rennie et al, 2020). Moreover, FANCI appears to be even more resistant to de‐ubiquitination than FANCD2, in this DNA‐bound di‐monoubiquitinated (IUbD2Ub‐DNA) state (van Twest et al, 2017; Rennie et al, 2020; Wang et al, 2020). Hence, the preferential targeting of FANCD2 for deubiquitination is likely to result in an ID2 complex that is ubiquitinated on FANCI‐only (IUbD2). Currently, we have no information on: (i) what conformation such a complex adopts, (ii) how does it bind to DNA, (iii) how well it supports FANCD2‐ubiquitination and (iv) how efficiently IUbD2 is protected from deubiquitination.

Providing an answer to such questions would greatly enhance our understanding of how the interdependency in FANCI‐FANCD2 in vivo ubiquitination (Smogorzewska et al, 2007) is encoded at the molecular level and elucidate the mechanism by which FANCI and FANCD2 ubiquitination (and deubiquitination) are linked. This is clinically relevant too, since FA‐pathway modulation is associated with both cancer progression and response to cancer treatment agents. Mutations or overexpression of FA genes and/or USP1 are commonly found in cancers (García‐Santisteban et al, 2013; Niraj et al, 2019; Xu et al, 2019; Liu et al, 2020). However, and most importantly, FA‐gene and/or USP1 deregulation is also frequently associated with chemo‐resistance which can be overcome once the expression of the corresponding gene is restored to normal levels (García‐Santisteban et al, 2013; Lim et al, 2018; Xu et al, 2019; Liu et al, 2020). This suggests that FA‐ and/or USP1‐ targeting inhibitors may be beneficial for cancer therapy. USP1, in particular, has been identified as a promising target for cancer‐therapy, for a variety of tumours, including: breast (Lim et al, 2018; Ma et al, 2018; Mussell et al, 2020; Niu et al, 2020), ovarian (Lim et al, 2018; Sonego et al, 2019), colorectal (Xu et al, 2019), Non‐small Cell Lung (Chen et al, 2011), bone (Williams et al, 2011) and glioma (Ma et al, 2019) cancers. Accordingly, there has been growing interest for the development of USP1‐UAF1‐specific inhibitors (Chen et al, 2011; Liang et al, 2014). One such USP1‐UAF1 inhibitor is currently in Phase I clinical trials, for treatment of advanced solid tumours (KSQ Therapeutics Inc, 2021).

In this work, we show that a transient IUbD2‐DNA complex is most likely formed due to significantly faster rate of FANCD2 over FANCI deubiquitination. We further demonstrate that FANCI ubiquitination maintains the closed‐on‐DNA ID2 conformation when FANCD2 ubiquitination is lost. We lastly show that, in this conformation, FANCD2 ubiquitination is favoured, while FANCI deubiquitination is restricted. Similar to IUbD2‐DNA complex having a propensity to transform into an IUbD2Ub‐DNA complex, the ID2Ub‐DNA complex also has the propensity to give rise to a di‐mono‐ubiquitinated complex: this is achieved due to ID2 displaying significantly faster kinetics of FANCI ubiquitination upon FANCD2 ubiquitination. Hence, our results indicate that ubiquitination of either ID2 subunit results in an ID2‐DNA clamp that promotes ubiquitination of the other subunit.

Results

To assess the difference between FANCD2 and FANCI deubiquitination, we assayed IUbD2Ub complex deubiquitination by USP1‐UAF1 in a time‐course. As illustrated in Fig 1, FANCI deubiquitination progresses at a much slower rate than FANCD2 deubiquitination. This suggests that an ID2 complex which is ubiquitinated only on FANCI (IUbD2), may derive from USP1‐UAF1‐mediated IUbD2Ub deubiquitination. Previous protein‐induced fluorescence enhancement (PIFE) assays in our lab showed that fully ubiquitinated or FANCD2‐only‐ubiquitinated ID2 complexes display a 10‐fold increase in affinity for double‐stranded DNA (dsDNA), relative to non‐ubiquitinated ID2; however, FANCI‐only‐ubiquitination results in only a 3‐fold enhancement in ID2‐DNA affinity (Rennie et al, 2020). A possible interpretation for this would be that an IUbD2 complex has a different conformation from ID2 and ID2Ub/IUbD2Ub, which would allow a different mode of binding to dsDNA. We thus sought to determine the structure of such complex bound to dsDNA, to elucidate how this may differ from ID2Ub and IUbD2Ub, and understand how IUbD2 exactly interacts with dsDNA.

Figure 1
Figure 1
Full size image

FANCD2 deubiquitination progresses at much faster rate than FANCI deubiquitination

FANCIUb‐FANCD2Ub‐DNA complexes were assembled in vitro, and FANCD2Ub and V5‐FANCIUb deubiquitination by USP1‐UAF1 (50 nM final) was monitored at room temperature in a time course: at indicative time‐points, aliquots of each reaction were removed and analysed by western blotting using FANCD2 and V5 antibodies. Experiment was repeated six times (two technical replicates with three different preparations of ubiquitinated FANCI) and FANCI/FANCD2 ubiquitination levels were calculated following quantification of ubiquitinated and non‐ubiquitinated FANCI/FANCD2 bands from the blots. For each protein, all calculated values for all time‐points were used for fitting to a one‐phase decay model.

To address these questions, we used our in vitro reconstitution approach (Arkinson et al, 2018; Rennie et al, 2020, 2021) to assemble a 1:1:1 IUbD2‐DNA complex, from purified IUb, D2 and dsDNA (61 bp long), and subsequently determined its structure by cryo‐EM. Such reconstitution approach has been successfully applied to produce ID2Ub/IUbD2Ub‐DNA structures (Rennie et al, 2020) and USP1‐UAF1‐bound ID2Ub‐DNA structures (Rennie et al, 2021). These complexes have been shown to adopt the same closed ID2 conformation as the one observed in ubiquitinated ID2‐DNA complexes produced following FA‐core‐catalysed ID2 ubiquitination (Alcón et al, 2020; Wang et al, 2020). This suggests that, despite the requirement of FA‐core for opening up the ID2 complex for subsequent ubiquitination (Wang et al, 2021), ID2 ubiquitination is actually required for both producing and maintaining the final closed ID2 conformation. Hence, we reasoned that our in vitro assembled complex would also be structurally indistinguishable from a complex produced through removal of FANCD2's ubiquitin from IUbD2Ub. Our final IUbD2‐DNA map, made of 139,601 image particles, was at 4.1 Å global resolution and had a local resolution ranging from 2.8 to 13.9 Å (Fig EV1A–E; Table 1). By 2D classification we also obtained few smaller‐sized‐particle 2D class averages (four classes; 62,961 particle images in total), likely corresponding to dissociated monomeric proteins (Fig EV1B). The presence of such smaller particles, combined with IUbD2‐DNA‐specific particles adopting a preferred orientation (Fig EV1D), restricted to some extent the overall resolution of our final map. Similarly, chicken ID2 and ID2Ub complexes (containing either wild‐type or phosphomimetic FANCI) have been reported to exhibit too, a preferred orientation of cryo‐EM particles images (Alcón et al, 2020; Sijacki et al, 2022). Hence, a strikingly uneven angular distribution of cryo‐EM particle images may be a common feature shared by ID2 complexes, irrespective of ubiquitination. Using the structure of IUbD2Ub‐DNA (Wang et al, 2020; PDB: 6VAE) as initial model (but with the ubiquitin conjugated to FANCD2 removed) for refinement, we obtained an atomic model of the IUbD2‐DNA structure at 4.1 Å resolution (Table 1; Figs 2A and EV1E). Despite the limited overall map resolution, our maps had several well‐resolved regions for modelling, like the one surrounding and including FANCI's K523 isopeptide linkage with G76 of ubiquitin (Fig EV1F) and the region of FANCI‐FANCD2 C‐termini interaction (Fig EV1G). Many FANCI and FANCD2 loops, as well as the FANCI N‐terminus (region corresponding to the first 170 aa) had very poor density and were thus unmodelled (Fig 2A). Modelled regions of relatively poor density included the dsDNA (Fig EV1H), the central region of FANCD2 and an N‐terminal part of FANCI (Fig EV1E).

Figure EV1
Figure EV1
Full size image

Figure

A. Example micrograph with scale bar.

B. Example 2D classes. Circular mask is 170 Å in diameter. 2D classes surrounded by a green box correspond to IUbD2‐DNA complex particles, while smaller‐sized 2D classes surrounded by a red box likely correspond to monomeric IUb/D2 proteins.

C. Fourier Shell Correlation (FSC) curves: between the two half maps produced in the final local non‐uniform refinement (shown in blue) and between the refined model and final map (shown in orange).

D. Particle orientation (viewing direction distribution) in the final map. Total number of particles: 139,601.

E. Top: Locally filtered map coloured by local resolution, viewed from three different angles. Bottom: Corresponding structural model viewed under same orientations.

F. IUbD2‐DNA structure with corresponding map density (locally filtered map), centred on the isopeptide bond between K523 of FANCI and G76 of ubiquitin. Some well‐resolved side‐chains are illustrated as sticks and indicated.

G. Interaction between FANCI and FANCD2 C‐termini with corresponding map density (locally filtered map). A beta‐sheet consisting of a FANCI and a FANCD2 strand is formed (residues 1,285–1,289 of FANCI and residues 1,384–1,388 of FANCD2). This is held in place through hydrophobic and electrostatic interactions with a FANCD2 helix (1,351–1,377 aa). Residues predicted to participate in such interactions are shown as sticks and indicated. Selected side chains, for which there is good density are also shown as sticks. For clarity, adjacent to that region elements of the IUbD2‐DNA structure and map are not shown.

H. IUbD2‐DNA structure centred on DNA. Density corresponding to the 27 bp modelled DNA is shown as orange mesh. Colouring of structure is as in (E–G).

Table 1 Cryo‐EM data collection and processing, and subsequent model building and refinement.
Full size table
Figure 2
Figure 2
Full size image

FANCIUb‐FANCD2 complex is a DNA clamp

A. FANCIUb‐FANCD2 (IUbD2) structure bound to double‐stranded DNA. The structure was determined by cryo‐EM, using a 4.1 Å global resolution map. Two different views of the structure are shown. Unmodelled regions (due to poor density) extending 20 amino‐acid stretches are indicated at the bottom.

B. Ubiquitination of either subunit of the ID2 complex results in increased affinity to double‐stranded DNA (dsDNA). Left: Normalised fluorescent changes of IRDye700‐labelled 32 bp DNA (125 nM) when incubated with increasing concentrations (ranging from 1 nM to 2.5 μM) of FANCI (I) or ubiquitinated FANCI (IUb) in the presence of a constant excess concentration (equal to the maximum concentration of I/IUb used) of FANCD2 (D2) or ubiquitinated FANCD2 (D2Ub). As a control, normalised fluorescent changes of IRDye700‐labelled DNA (125 nM) when incubated with increasing concentrations of FANCD2 (ranging from 40 nM to 5 μM) were monitored as well. For each protein/complex, the experiment was conducted 2–4 times (technical replicates) and all data points from the replicate experiments were used for fitting of a one‐site binding model. Right: Apparent ID2, IUbD2 and D2 Kd values (and associated uncertainties, all in nM) for dsDNA measured from model fitting (n: number of binding experiments per protein/complex).

Interestingly, IUbD2 has the same closed‐on‐DNA conformation as IUbD2Ub and ID2Ub (Fig 2A). We hypothesised that the apparent lower than expected enhancement of ID2‐DNA affinity upon FANCI ubiquitination measured before, may be due to IUb dissociating from D2 at low concentrations. Indeed, previous PIFE assays showed that at lower concentrations of IUbD2 (< 100 nM) there had been negligible protein‐binding induced fluorescence enhancement of labelled DNA, while at higher protein concentrations (> 500 nM), the IUbD2‐binding induced fluorescence enhancement of labelled DNA, had been comparable to that achieved with IUbD2Ub and ID2Ub complexes (Rennie et al, 2020). Hence, to ensure complex formation at low FANCI concentrations, we performed dsDNA‐binding PIFE assays for ID2, IUbD2, ID2Ub and IUbD2Ub as before (Rennie et al, 2020), but this time we titrated only FANCI (ubiquitinated or not), while having a constant high concentration of FANCD2 (ubiquitinated or not) – equal to the maximum concentration of FANCI used. With such set‐up, our PIFE assays reveal a significant (> 10‐fold) increase in ID2 affinity for dsDNA when either FANCI, FANCD2 or both FANCI and FANCD2 were ubiquitinated (Fig 2B). FANCD2 on its own had negligible binding to dsDNA (Fig 2B), while ubiquitinated FANCD2 had been previously shown to have no affinity for dsDNA on its own either (Rennie et al, 2020). The above indicate that FANCI‐ubiquitination is responsible for maintaining the clamping of the ID2 complex on DNA, when FANCD2 ubiquitination is lost.

The overall conformation of the IUbD2 complex is very similar to that of IUbD2Ub (Fig EV2A). The most noticeable differences are (i) slight movements of FANCD2 and FANCI helices in the region where FANCD2's ubiquitin interacts with FANCI and (ii) the high level of disorder in the FANCI N‐terminus proximal to that region (residues 1–170), upon loss of FANCD2 ubiquitination (Fig EV2B and C).

Figure EV2
Figure EV2
Full size image

Figure

A. The absence of a FANCD2‐conjugated ubiquitin in IUbD2‐DNA structure is associated with a disorder in the N‐terminus of FANCI (residues 1–170), when compared with IUbD2Ub‐DNA structure. The two structures were aligned in Pymol and visualised from the same angle, either on their own (left and centre), or together (right).

B. Helices of FANCI involved in interaction with FANCD2's ubiquitin (IUbD2Ub‐DNA structure; FANCI: blue, Ubiquitin: magenta), are positioned differently when the ubiquitin is removed (IUbD2‐DNA structure; FANCI: orange). IUbD2Ub‐DNA and IUbD2‐DNA structures were fitted to IUbD2‐DNA sharpened map (yellow) and centred on FANCI N‐terminus.

C. Removal of ubiquitin (magenta) from FANCD2 results in slight movements affecting several FANCD2 helices, from α31 (helix where ubiquitin is conjugated) up to α18. FANCD2 helices of IUbD2‐DNA and IUbD2Ub‐DNA are better aligned towards the N‐terminus of FANCD2 (N‐terminally to, and including α18 helix of FANCD2), whereby FANCD2 interacts with the ubiquitin (green) conjugated to FANCI. The structures shown in (A), were centred on the central part of FANCD2.

While both IUbD2Ub‐DNA (EMD‐21138) and ID2Ub‐DNA (EMD‐21139) maps display relatively poor density for FANCI N‐terminus (Wang et al, 2020), there is virtually no density for that part in both our locally‐filtered and Phenix‐auto‐sharpened map (Figs EV1E and 3A, respectively). Lack of density in the N‐terminus of FANCI has been also observed before, upon extraction of FANCD2's ubiquitin by USP1 (Rennie et al, 2021). We thus propose that the high level of disorder in the FANCI N‐terminus is a direct consequence of the loss of binding between the ubiquitin of FANCD2 and the N‐terminus of FANCI. Similarly, reduced density for the N‐terminus of FANCD2 has been previously observed in closed state ID2 conformations in which there is no ubiquitin conjugated to FANCI, like in the ID2Ub‐DNA (EMD‐21139) map (Wang et al, 2020; Fig 3A; right) and USP1‐UAF1‐ID2Ub‐DNA (EMD‐11934) map (Rennie et al, 2021).

Figure 3
Figure 3
Full size image

Ubiquitination of either of the two ID2 subunits enhances ubiquitination of the other

A Comparison of cryo‐EM density distribution among IUbD2‐DNA (Phenix‐auto‐sharpened map), IUbD2Ub‐DNA (EMD‐21138) and ID2Ub‐DNA (EMD‐21138) maps. IUbD2‐DNA and ID2Ub‐DNA maps, as well as IUbD2Ub‐DNA model (PDB: 6VAE) were aligned to IUbD2Ub‐DNA in ChimeraX. A different colour was applied for each of the protein chains of IUbD2Ub‐DNA model (FANsCI: slate blue, Ubiquitin‐on‐FANCI: green, FANCD2: cyan, Ubiquitin‐on‐FANCD2: magenta), while DNA was coloured red. Then each map was coloured according to nearby (within 6 Å) residue colours. Contour levels were adjusted (IUbD2‐DNA: 6.21, IUbD2Ub‐DNA: 0.0194 and ID2Ub‐DNA: 0.0162) to achieve comparable volumes among all displayed maps (ranging from 8.6 to 9.4 × 104 Å3). Arrows indicate regions of poorer density (in IUbD2‐DNA and ID2Ub‐DNA maps) relative to other regions of the map, as well as to equivalent positions in the other two maps.

B Both K561 of FANCD2 and K523 of FANCI become more accessible upon ubiquitination of the other ID2 subunit. Structural comparison of relative accessibility of FANCD2‐K561, upon FANCI ubiquitination (left panel), and of FANCI‐K523, upon FANCD2 ubiquitination (right panel). The relative positions of these lysines upon conjugation with ubiquitin, are also shown for comparison. Residues of the other ID2 subunit within 8 Å distance from the epsilon‐amino‐group of the corresponding lysine are indicated as sticks. The distance to the nearest residue is shown prior and upon ubiquitination of the other ID2 subunit. In either case this increases, upon ubiquitination of the other subunit, further than 10 Å.

C, D ID2 ubiquitination on FANCI results in increased rate of FANCD2 ubiquitination (B), whereas ID2 ubiquitination on FANCD2 results in increased rate of FANCI ubiquitination (C). Protein complexes were assembled in vitro on ice in the presence of dsDNA (32 bp) and their in vitro ubiquitination at 30°C was subsequently monitored in a time‐course: at indicative time‐points, aliquots of the reaction were removed and analysed by western blotting using FANCD2 and V5 antibodies (Top). For each protein complex, data‐points from three replicate experiments (three technical replicates) were used in fitting to a one‐phase association model (Bottom).

We reasoned that the relative disorder in the N‐terminal regions of FANCI or FANCD2, in IUbD2‐DNA or ID2Ub‐DNA complexes, respectively, might be crucial for ubiquitination of FANCD2 or FANCI, correspondingly. Despite the slight movements of IUbD2's FANCI and FANCD2 towards the region where FANCD2‐conjugated ubiquitin would be (Fig EV2B and C), FANCD2's K561 is fully accessible for ubiquitination, when compared to its position in the ID2‐DNA complex (~80 Å2 buried; Figs 3B and EV3). Similarly, several FANCD2 residues in proximity to FANCI's K523 (in ID2‐DNA complex) are positioned further away upon FANCD2 ubiquitination, and hence FANCI's K523 becomes more accessible for ubiquitination in the ID2Ub‐DNA complex than in the ID2‐DNA complex (Figs 3B and EV3). These observations led us to the hypothesis that ubiquitination of either of the two subunits of ID2 (FANCI or FANCD2) actually favours ubiquitination of the other subunit. Indeed, time‐course ubiquitination assays reveal that: FANCD2 ubiquitination is stimulated when FANCI is already ubiquitinated; and similarly, FANCI ubiquitination is stimulated when FANCD2 is already ubiquitinated (Fig