ATAD5 functions as a regulatory platform for Ub–PCNA deubiquitination

Significance

Proliferating cell nuclear antigen (PCNA), a replicative DNA clamp, is mono- or polyubiquitinated upon replication fork stalling to facilitate DNA-lesion bypass. When obstacles for DNA replication are resolved, ubiquitinated PCNA (Ub–PCNA) needs to be deubiquitinated to resume efficient replicative DNA synthesis. It has been unclear how the ubiquitination status of PCNA is dynamically regulated. We revealed how the ATAD5 N-terminal domain recognizes DNA-loaded Ub–PCNA and recruits UAF1–USP1 to coordinate Ub–PCNA deubiquitination with PCNA unloading. Furthermore, we identified the interaction between ATAD5 and USP7–USP11 to efficiently deubiquitinate polyubiquitinated PCNA. Last, we found that specific inhibition of Ub–PCNA deubiquitination by ATAD5 mutation increases sensitivity to DNA damage agents in vitro and in vivo models, indicating importance of timely PCNA deubiquitination through ATAD5.

Abstract

Ubiquitination status of proliferating cell nuclear antigen (PCNA) is crucial for regulating DNA lesion bypass. After the resolution of fork stalling, PCNA is subsequently deubiquitinated, but the underlying mechanism remains undefined. We found that the N-terminal domain of ATAD5 (ATAD5-N), the largest subunit of the PCNA-unloading complex, functions as a scaffold for Ub–PCNA deubiquitination. ATAD5 recognizes DNA-loaded Ub–PCNA through distinct DNA-binding and PCNA-binding motifs. Furthermore, ATAD5 forms a heterotrimeric complex with UAF1–USP1 deubiquitinase, facilitating the deubiquitination of DNA-loaded Ub–PCNA. ATAD5 also enhances the Ub–PCNA deubiquitination by USP7 and USP11 through specific interactions. ATAD5 promotes the distinct deubiquitination process of UAF1–USP1, USP7, and USP11 for poly-Ub–PCNA. Additionally, ATAD5 mutants deficient in UAF1-binding had increased sensitivity to DNA-damaging agents. Our results ultimately reveal that ATAD5 and USPs cooperate to efficiently deubiquitinate Ub–PCNA prior to its release from the DNA in order to safely deactivate the DNA repair process.

Accurate genome duplication is crucial for the transfer of genetic information from one generation to the next (1). A multisubunit replication machinery, known as the replisome, unwinds parental DNA and synthesizes daughter strands (2, 3). During genome duplication, replisomes encounter a variety of DNA lesions that cause DNA helicases or polymerases to be stalled. Eukaryotic cells utilize various mechanisms to bypass these lesions and continue DNA synthesis (4, 5). Proliferating cell nuclear antigen (PCNA) is a key protein that modulates DNA-lesion bypass (6, 7).

PCNA is a homotrimeric ring-shaped DNA clamp that tethers DNA polymerases to replicating DNA (8). PCNA also functions as a molecular hub that coordinates the activities of replication and repair proteins (9–16). Upon replication stalling, PCNA is monoubiquitinated at lysine 164 (17, 18). E2/E3 complex RAD6/RAD18 monoubiquitinates PCNA (19, 20). Mono-Ub–PCNA recruits translesion synthesis (TLS) DNA polymerases, error-prone enzymes that can traverse the damaged DNA template (21, 22). Alternatively, PCNA can be further polyubiquitinated to promote error-free template-switching (TS) (21, 23, 24). In human cells, poly-Ub–PCNA also facilitates fork reversal (FR), resulting in a transient slowing of replication forks (21, 25, 26). K63-linked PCNA polyubiquitination is mediated by the heterotrimeric E2 enzyme UBC13/MMS2 along with the E3 ligase HLTF or SHPRH (27–29).

While PCNA ubiquitination facilitates DNA-lesion bypass, the prolonged presence of Ub–PCNA on the DNA is detrimental to genomic integrity (30–33). TLS polymerases are mutagenic, and Ub–PCNA-mediated fork remodeling can hinder efficient replication fork progression. Furthermore, aberrant Ub–PCNA accumulation on chromatin can impede normal DNA replication by inappropriately recruited TLS/TS/FR factors. Therefore, Ub–PCNA needs to be deubiquitinated when obstacles to replication are resolved. Ubiquitin-specific protease 1 (USP1) is known to deubiquitinate mono-Ub–PCNA. USP1 requires USP1-associated factor 1 (UAF1) for substrate deubiquitination (34, 35).

It was shown that efficient mono-Ub–PCNA deubiquitination by UAF1–USP1 requires an additional factor, ATPase family AAA+ domain-containing protein 5 (ATAD5), the largest subunit of the PCNA-unloading complex (36), while UAF1–USP1 deubiquitinates other substrates such as FANCD2–FANCI on its own (37–39). ATAD5 forms a replication factor C (RFC)-like-complex (ATAD5–RLC) with RFC2-5. ATAD5 is functionally significant in regulating DNA replication and repair (30, 40–45). We have previously reported that ATAD5–RLC possesses potent PCNA-unloading activity (46–48). PCNA-unloading activity of ATAD5 is conferred by a C-terminal AAA+ ATPase domain and RFC2-5 binding domain. In addition to PCNA unloading, ATAD5 also participates in the deubiquitination of Ub–PCNA (36). ATAD5 depletion increases the amount of Ub–PCNA on the chromatin. The ATAD5 N-terminal domain (ATAD5-N) interacts with UAF1–USP1 (36). However, the mechanism by which ATAD5 modulates Ub–PCNA deubiquitination remains unknown.

In addition to UAF1–USP1, a role of USP7 in the deubiquitination of Ub–PCNA was reported (49, 50). USP7 is a deubiquitinase that modulates various cellular processes, including DNA replication and repair (51). USP7 regulates the ubiquitination status of DNMT1 during DNA replication as well as the ubiquitination status of XPC and ERCC6 for nucleotide excision repair (52–54). USP11 is another deubiquitinase that controls DNA repair processes. USP11 deubiquitinates SPRTN for the repair of DNA–protein crosslinks and γH2AX for the double-strand-break repair (55, 56). Moreover, USP7 interacts with USP11 to regulate the function of Polycomb Repressive Complex 1 (PRC1) (57). However, it is currently unknown whether there is a functional interaction between UAF1–USP1, USP7, USP11, and ATAD5 that leads to Ub–PCNA deubiquitination.

We have previously reported that ATAD5–RLC can unload mono-Ub or poly-Ub–PCNA from DNA (46). UAF1–USP1 preferentially deubiquitinates DNA-loaded Ub–PCNA over free Ub–PCNA. Therefore, it is necessary to coordinate PCNA unloading with deubiquitination. Here, we report that ATAD5 functions as a molecular platform for Ub–PCNA deubiquitination. Our biochemical analyses show that ATAD5 recognizes DNA-loaded Ub–PCNA and recruits UAF1–USP1 to facilitate Ub–PCNA deubiquitination. Additionally, ATAD5 specifically binds to USP7 and USP11. ATAD5 promotes poly-Ub–PCNA deubiquitination by UAF1–USP1, USP7, and USP11. Enhancement of Ub–PCNA deubiquitination by ATAD5 is important for maintaining genomic integrity after DNA damage in both cultured cells and mice. Collectively, our findings suggest that ATAD5 enhances Ub–PCNA deubiquitination on DNA to prevent the inappropriate release of Ub–PCNA from DNA, thereby maintaining genomic integrity.

Materials and Methods

Protein Purification.

Most proteins were purified using the Bac-to-Bac Baculovirus expression system (Thermo Fisher Scientific). Viruses were prepared using Sf9 cells, and proteins were expressed in Hi-5 cells. N-terminal 10x His tagged UAF1 and N-terminal 2x StrepII tagged USP1 were cloned into the pFL multigene expression system and expressed from a single construct. Single polypeptide PCNA was N-terminal 10x His- and C-terminal 3x HSV-3x FLAG tagged. USP7 was N-terminal FLAG-MBP-StrepII tagged and USP11 was N-terminal 10x His-3x FLAG tagged. In the case of ATAD5 N-terminal fragments, the proteins were tagged with N-terminal MBP-StrepII and C-terminal 6x FLAG-S. ATAD5 variants were expressed and purified from yeast strains. Yeast strains used for the purification of ATAD5 N-terminal domains are listed in SI Appendix, Table S2. Single polypeptide PCNA wild-type and mutants that harbor the substitution mutation of lysine by arginine (K164R) were expressed and purified from Escherichia coli BL21 (DE3) (Enzynomics). The proteins were prepared as described previously with few modifications (46). Detailed procedures of protein purification were described in SI Appendix.

DNA Substrates for Deubiquitination Reactions.

130-mer DNA substrate was prepared based on information of previous report (46). Methods for the preparation of bead-attached 130-mer DNA was described in SI Appendix.

Monoubiquitination Reactions.

PCNA was ubiquitinated after loading onto the 130-mer DNA. See SI Appendix for PCNA loading procedure. For the monoubiquitination reaction, a standard 4x ubiquitination reaction buffer was prepared (80 mM HEPES [pH 7.5], 4 mM DTT, 40 mM magnesium chloride, 4 mM ATP). A monoubiquitination reaction mixture was prepared by mixing 7.5 μL of 4x ubiquitination reaction buffer with 22.5 μL of protein mixture (1x mixture: 50 nM UBA1, 100 nM RAD6B-RAD18, and 1.25 μM ubiquitin in water). Next, 30 μL of the monoubiquitination reaction buffer was added to the collected DNA beads where 3xPCNA molecules are loaded. After incubating the reaction mixture at 30 °C in a Thermomixer for 20 min, the reaction was stopped by washing with 0.3 M KCl Buffer H.

Polyubiquitination Reactions.

After the monoubiquitination reaction, monoubiquitination enzymes were removed by washing magnetic beads with 0.3 M KCl Buffer H. Next, polyubiquitination reaction mixture was prepared by mixing 7.5 μL of 4x ubiquitination reaction buffer with 22.5 μL of protein mixture (1x mixture: 100 nM of UBA1, 50 nM of HLTF, 50 nM of UBC13/MMS2, and 1.25 μM ubiquitin in water). Then, 30 μL polyubiquitination reaction buffer was added to the bead and incubated for 30 min at 30 °C in a Thermomixer. The polyubiquitination reaction was stopped by washing with 0.3 M KCl Buffer H before additional treatments.

Deubiquitination Reactions.

Deubiquitination reaction was performed with 30 μL of reaction buffer (25 mM potassium chloride, 50 mM Tris–HCl [pH 7.5], 1 mM DTT, 1 mM magnesium chloride) containing various concentrations of UAF1–USP1, USP7, USP11, and ATAD5 N-terminal fragments. Then, 30 μL deubiquitination reaction buffer was added to the bead and incubated for 20 min at RT in a Thermomixer. The deubiquitination reaction was stopped by washing with 0.3 M KCl Buffer H. Finally, DNA was resuspended in 30 μL of digestion buffer (50 mM Tris–HCl [pH 7.5], 1 mM DTT, 1 mM magnesium chloride) containing 1 unit of DNase I (Promega). Diubiquitin cleavage assay and ubiquitin-AMC cleavage assay were described in SI Appendix.

Electrophoretic Mobility Shift Assay (EMSA).

For EMSA, 130-mer DNA substrate was prepared with FAM-labeled primer instead of biotinylated primer. Various concentrations of UAF1–USP1 complex or ATAD5 N-terminal fragments were incubated with 1 pmole of DNA substrate in reaction buffer (50 mM Tris–HCl [pH8.0], 50 mM potassium chloride, 2.5 mM magnesium chloride, 10% glycerol, 1 mM DTT, 0.1 mg/mL BSA) for 30 min at RT. Reaction products were subjected to nondenaturing acrylamide gel and FAM-labeled DNA substrate was detected via Typhoon RGB (Amersham).

In Vitro Pull-Down Assay.

To check interaction between purified ATAD5 mutants and UAF1, 1 pmol of ATAD5 was mixed with 5 pmol of UAF1 in 100 μL of 0.1 M KCl Buffer H. For checking PCNA interaction, PCNA or Ub–PCNA were isolated from the reaction by DNase I digestion. Protein mixtures were incubated on ice for 30 min and ATAD5 fragments were pulled down by S protein agarose (Millipore). Isolated proteins were analyzed by SDS-PAGE and western blot.

DNA Substrates Digestion for Free-Ub–PCNA Elution.

After the monoubiquitination reaction, DNase I or Hind III (New England biolabs) mixed with deubiquitination reaction buffer were added to the bead and then incubated for 15 min at 37 °C. The eluted Ub–PCNA was mixed with deubiquitinating enzymes.

Antibodies.

Antibodies used in this study were listed in SI Appendix.

Plasmids (or) DNA Constructs and siRNAs.

DNA constructs and siRNAs used in this study were described in SI Appendix.

Cell Culture.

Human embryonic kidney (HEK) 293T cells, U2OS cells (purchased from American Type Culture Collection, Manassas, VA), and their derivative cell lines stably expressing ATAD5 protein were cultured in Dulbecco’s modified Eagle’s medium (Hyclone) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin–streptomycin (Gibco) at 37 °C under 5% CO₂.

Transfections and RNA Interference.

Transfections of plasmid DNA and siRNAs were performed using X-tremeGENE HP DNA transfection Reagent (Roche) and Lipofectamine RNAiMAX (Invitrogen), respectively, according to the manufacturer’s instructions. Transfected cells were harvested 48 h after transfection for further analysis. To deplete only endogenous ATAD5, siRNA targeting 3′UTR of ATAD5 was transfected 4 h after transfecting plasmids expressing wild-type or mutant ATAD5 protein (47).

Chromatin Fractionation.

Cells were lysed with buffer A (100 mM NaCl, 0.3 M sucrose, 3 mM MgCl₂, 10 mM PIPES [pH 6.8], 1 mM EGTA, and 0.2% Triton X100, containing the phosphatase inhibitor PhosSTOP [Roche] and complete protease inhibitor cocktail [Roche]) for 8 min on ice. Crude lysates were centrifuged at 2,300×g at 4 °C for 5 min to separate the chromatin-containing pellet from the soluble fraction. The pellet was digested with 50 units of Benzonase (Enzynomics) for 1 h in RIPA buffer (50 mM Tris–HCl [pH 8.0], 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.5% Na deoxycholate, 1 mM PMSF, 5 mM MgCl₂, containing the phosphatase inhibitor PhosSTOP [Roche] and complete protease inhibitor cocktail [Roche]) to extract chromatin-bound proteins. The chromatin-containing fractions were clarified by centrifugation (15,800×g, 4 °C) for 5 min to remove debris. Protein concentration was determined by the Bradford Assay (Bio-Rad), and the proteins were analyzed by western blotting.

Immunoprecipitation and Western Blot Analysis.

Whole-cell lysates were prepared by lysing cells with buffer X (100 mM Tris–HCl [pH 8.5], 250 mM NaCl, 1 mM EDTA, and 1% Nonidet P-40, 5 mM MgCl₂) supplemented with the phosphatase inhibitor PhosSTOP (Roche), complete protease inhibitor cocktail (Roche), and 50 units of Benzonase for 1 h at 4 °C. Lysates were cleared by centrifugation (15,800×g, 4 °C, 5 min). FLAG-tagged proteins were incubated with anti-FLAG M2 agarose affinity beads (Sigma) for 1 h at 4 °C with constant rotation. The beads were washed three times using buffer X, and the bead-bound proteins were eluted with buffer X containing 0.15 mg/mL FLAG peptide. Coimmunoprecipitated proteins were loaded onto SDS-PAGE and analyzed by immunoblotting. To check phospho-shift of ATAD5 fragments, FLAG immunoprecipitated ATAD5 variants were treated with λ-phosphatase (NEB). The proteins were detected by means of a ChemiDoc MP imaging system (Bio-Rad). The signal intensity of the bands was quantified by ImageLab software version 5.2.1 (Bio-Rad).

EdU Incorporation Analysis.

Cell lines expressing ATAD5 Wild-type and UB1m mutant in ATAD5 knock-out 293AD cells were labeled with 10 mM EdU for 1 h. For flow cytometry analysis, samples were prepared using the Click-iTTM Plus EdU Alexa FluorTM 647 Flow Cytometry Assay Kit (Invitrogen) according to the manufacturer’s instructions and subjected to FACS analysis (FACSVerse™ flow cytometer, BD Biosciences). Data were analyzed using FlowJo software (Tree Star).

Lentivirus-Based Stable Cell Line Generation.

Cells expressing wild-type or UB1m, UB1m+UB4m, UB1m+UB4m+DBm (UB1m: UAF1-binding defective, UB4m: PCNA-binding defective, DBm: DNA-binding defective) ATAD5 were generated by infecting lentiviral expression cassette to ATAD5 knock-out HEK293 AD cells. Detailed procedures of cell line generation were described in SI Appendix.

Cell Survival Assay.

Cells were seeded in 96-well culture plates. After 24 h, cells were treated with indicated dose of UV (254 nm) or H₂O₂ (2 h) and incubated for 24 h (UV) or 48 h (H₂O₂) in fresh media. Cell survival was measured by the Cell Proliferation Reagent WST-1 (Roche) according to the manufacturer’s instructions. In brief, WST-1 assay reagent was added to the cell culture media and incubated for 2 h and the amount of produced formazan dye was analyzed by measuring the absorbance at 440 nm. Percent cell survival was normalized to that of control cells.

AP-MS Analysis.

Detailed procedures of AP-MS analysis for the identification of ATAD5 (1 to 500) interacting USPs and phosphorylation sites in ATAD5 (1 to 400) were described in SI Appendix.

Animal Care.

All animal procedures were approved and performed according to the guidelines provided by the Ulsan National Institute of Science and Technology’s Institutional Animal Care and Use Committee (IACUC-UNIST19-05).

UV Irradiation.

Six-week-old hair-removed +/+ and ΔUB3 mice were irradiated by UV ramp with 365 nm 1.5 kJ/m² daily twice and after 24-h recovery, skin was collected. Control non-UV area of skin of each mouse was protected from UV using aluminum foil.

Paraffin Sections.

First, 1 cm × 1 cm skin pieces were fixed with 10% neutral buffered formalin for 7 d and 3 mm × 10 mm piece were embedded in paraffin. Then, skin embedded block were sectioned in 5 μm with a microtome.

TUNEL Staining.

The TUNEL assay was performed on paraffin section using “In Situ Cell Death Detection Kit, Fluorescein” (Cat-11684795910) from Roche according to the manufacturer’s instructions. Briefly, paraffin sections on slides were first permeabilized with PBS/1% BSA/0.1% Triton X-100, then incubated with the TUNEL reaction Mixture for 30 min at 37 °C, washed several times with PBS, which was then followed by immunofluorescent staining. Image analysis procedure of TUNEL staining was described in SI Appendix.

Histology Analysis.

Dissected testes were fixed in Bouin’s fixative overnight at 4 °C, progressively dehydrated in a graded ethanol series, and embedded in paraffin wax. Sections (6 μm) were deparaffinized, rehydrated, and stained with hematoxylin and eosin.

Tissue Immunofluorescent Staining.

Slides with paraffin sections were washed in PBT (0.1% Tween 20 in PBS) and autoclaved in 10 mM citric acid (pH 6.0) to retrieve antigenicity. Slides were blocked in 5% serum (matched to the species of the secondary antibody) in PBS for 1 h at room temperature and incubated with primary antibodies overnight at 4 °C prior to detection with secondary antibodies. Image analysis procedure of tissue immunofluorescent staining was described in SI Appendix.

Quantification and Statistical Analysis.

Quantification of immunoblots was performed using Bio-Rad Image Lab. Statistical analyses were completed using GraphPad Prism. Statistical details of individual experiments are described in the figure legends and Results. Unless otherwise stated, all experiments were performed three times and representative experiments were shown. Bar graphs with mean and SD were presented. Individual data points were overlaid with graphs as scatter dot plot. In all cases, *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.

Results

ATAD5-N Facilitates Deubiquitination of Ub–PCNA by UAF1–USP1.

As previously reported, ATAD5 depletion led to Ub–PCNA accumulation (Fig. 1A). Ub–PCNA accumulation was reduced by the expression of ATAD5 (1 to 692) (Fig. 1A). These findings prove that ATAD5-N controls PCNA ubiquitination status. It has been proposed that ATAD5-N is implicated in Ub–PCNA deubiquitination through interactions with UAF1–USP1 (36). However, the role of ATAD5 in the deubiquitination process is still unclear and ATAD5-N modulated Ub–PCNA deubiquitination has not been reconstituted. First, we identified the minimum ATAD5 region that lowers Ub–PCNA levels in ATAD5-depleted cells (Fig. 1A). Although a previous study claimed that ATAD5 (1 to 400) was sufficient for lowering Ub–PCNA levels (36), our current analysis unambiguously showed that ATAD5 (1 to 500) is necessary (Fig. 1A). It is noteworthy that the majority of ATAD5 (1 to 500), but not ATAD5 (1 to 400), was chromatin associated (Fig. 1A and SI Appendix, Fig. S1 A and B).

Fig. 1.

ATAD5-N facilitates the Ub–PCNA deubiquitination by UAF1–USP1. (A) ATAD5 (1 to 500) is required for deubiquitination of Ub–PCNA. Chromatin fractionation was performed after expression of the indicated ATAD5 variants in ATAD5-depleted 293T cells. Endogenous ATAD5 was depleted by siRNA targeting 3′ UTR. *: Ubiquitinated ATAD5 (1 to 692). (B) Schematic diagram of Ub–PCNA deubiquitination assay. (C) Autocleavage inactive wild-type UAF1–USP1 deubiquitinates DNA-loaded Ub–PCNA. In vitro deubiquitination assays were performed with DNA-loaded ubiquitinated 3xPCNA. USP1 (C90S) is a catalytically inactive mutation. (D) ATAD5 (1 to 500) facilitates deubiquitination of DNA-loaded Ub–PCNA in reconstituted system. DNA-loaded Ub–PCNA was deubiquitinated by UAF1–USP1 in the presence of purified ATAD5 (1 to 500).

Next, we purify human ATAD5 (1 to 500) and assess its function in Ub–PCNA deubiquitination (Fig. 1 B–D and SI Appendix, Fig. S1). Following the binding of the TALE protein to DNA to inhibit PCNA from sliding off, PCNA was loaded onto bead-coupled primer-template DNA using RFC and was subsequently ubiquitinated on DNA by RAD6B-RAD18 (Fig. 1B) (46). To facilitate the analysis of PCNA ubiquitination status, we created a subunit fused PCNA trimer (3xPCNA), in which individual protomers were connected by flexible linkers (SI Appendix, Fig. S1C). RFC efficiently loaded the 3xPCNA to DNA, and the loaded 3xPCNA remained stably bound to DNA (SI Appendix, Fig. S1D). We were able to track the K164 ubiquitination status of individual protomers using the 3xPCNA (SI Appendix, Fig. S1 E–H). All three of the PCNA protomers were monoubiquitinated, according to the mobility shift of 3xPCNA in SDS-PAGE. The efficient DNA loading and ubiquitination of 3xPCNA indicate that the subunit fused PCNA is functional. DNA-loaded Ub–PCNA was then deubiquitinated by UAF1–USP1 (Fig. 1C and SI Appendix, Fig. S1I). Wild-type USP1 that is autocleavage-inactive (G670A, G671A) deubiquitinated Ub–PCNA subunits, whereas catalytically inactive USP1 (C90S) did not. Multiple bands appear between triple-monoubiquitinated and unmodified 3xPCNA, in addition to the two predicted bands. When the subunit-fused PCNA is randomly ubiquitinated, it produces a combination of dual or single mono-Ub–3xPCNA that is ubiquitinated at various positions. Migration of 3xPCNA in SDS-PAGE varies depending on the locations of ubiquitinated protomers (SI Appendix, Fig. S1 G and H).

Importantly, ATAD5 (1 to 500) significantly enhanced the deubiquitination of Ub–PCNA (Fig. 1D and SI Appendix, Fig. S1J). ATAD5 (1 to 500) did not exhibit any deubiquitination activity by itself (SI Appendix, Fig. S1K). In contrast, ATAD5 (1 to 500) had no impact on cleavage of K63-linked diubiquitin or Ub–AMC by UAF1–USP1 (SI Appendix, Fig. S1 L–N). These findings demonstrate that ATAD5-N modulates Ub–PCNA deubiquitination process directly and is crucial for UAF1–USP1 to efficiently deubiquitinate Ub–PCNA. Our results also show that the deubiquitination activity of ATAD5 can be functionally separated from its PCNA-unloading activity, because ATAD5 C-terminal domain mediates PCNA-unloading.

Ub–PCNA Deubiquitination Requires the Stable Interaction of ATAD5-N with UAF1–USP1.

It is known that interaction between ATAD5 and UAF1–USP1 plays a crucial role in controlling the level of Ub–PCNA (36). However, it has not yet been determined how ATAD5 binding to UAF1–USP1 affects Ub–PCNA deubiquitination at the molecular level. The results of in vitro pull-down experiments demonstrated that ATAD5-N binds to UAF1 directly (SI Appendix, Fig. S2 A and B). In addition, USP1 depletion had no effect on the coprecipitation of UAF1 and ATAD5 (SI Appendix, Fig. S2C). These findings imply that ATAD5 interacts with UAF1–USP1 through UAF1. To further understand how the ATAD5–UAF1 interaction regulates Ub–PCNA deubiquitination, we aimed to generate ATAD5 mutants that are specifically defective in UAF1 binding. First, the minimal UAF1-binding domain was determined. Although it has been suggested that deleting ATAD5 residues 368 to 384 weakens UAF1 binding (36), the ATAD5 region that interacts with UAF1 has not been precisely characterized. Immunoprecipitation experiments with ATAD5 variants demonstrated that ATAD5 (282 to 400) is required for stable UAF1 binding (Fig. 2A and SI Appendix, Fig. S2D). We designate this region as the ATAD5 UAF1-Binding domain (ATAD5 UB). To identify residues essential for UAF1 binding, we mutated multiple conserved amino acid patches in ATAD5 UB and perform immuno-precipitation (SI Appendix, Fig. S2 E–G). Through this approach, the essential UB1 and UB3 motifs for UAF1 interaction were identified. The UB3 motif overlaps with the previously claimed UAF1-interacting region (36), but UB1 is identified in this study. Mutation in either of the two motifs prevented UAF1 binding (Fig. 2B and SI Appendix, Fig. S2 F–H).

Fig. 2.

ATAD5–UAF1 interaction is essential for Ub–PCNA deubiquitination. (A) ATAD5 (282 to 400) is required for stable binding to UAF1–USP1. The indicated ATAD5 variants were immuno-precipitated using anti-FLAG beads to monitor the coisolation of UAF1 and USP1. (B) UB1 and UB3 motifs of ATAD5 are crucial for UAF1 binding. After expression of full-length ATAD5, FLAG pull-down was performed. See SI Appendix, Fig. S2E for the information of mutations. BRD4A is a binding control protein. (C) The β-propeller domain and the SLD2 of UAF1 interact with ATAD5. Indicated UAF1 variants were expressed and then isolated with S-protein beads. (D) ATAD5–UAF1 interaction is crucial for Ub–PCNA deubiquitination. Chromatin fractionation was performed following the expression of ATAD5 variants in 293T cells depleted of ATAD5. Numbers below Ub–PCNA blots indicate relative amount of Ub–PCNA. (E) ATAD5 UB1 motif is essential for association with UAF1. After mixing purified ATAD5 (1 to 500) and UAF1, S-protein pull-down was performed. (F) Quantification of the UAF1 coisolated with ATAD5 (1 to 500) in E. Error bars indicate the SD (n = 3). (G) ATAD5–UAF1 interaction is essential for facilitating the deubiquitination of Ub–PCNA. DNA-loaded Ub–PCNA was deubiquitinated by UAF1–USP1 in the presence of wild-type or UB1m ATAD5 (1 to 500). (H) Quantification of the level of Ub–PCNA deubiquitination shown in G. The Y axis indicates the relative amount of the unmodified PCNA compared to the fully ubiquitinated PCNA in the mock reaction. The error bars indicate SD (n = 3).

Additionally, we find that UB-containing ATAD5 fragments were phospho-shifted in SDS-PAGE (SI Appendix, Fig. S2I). We questioned whether phosphorylation status of UB affects UAF1 binding. Several phospho-residues were identified within or upstream of UB by mass spectrometric analysis (SI Appendix, Fig. S2J). Mutation of putative phospho-sites and SDS-PAGE mobility analysis were used to identify phosphorylated residues (SI Appendix, Fig. S2K). Among the other variants, only the S306A mutation eliminated the phospho-shift. It is noteworthy that the majority of exogenously expressed ATAD5 (158 to 400) was phospho-shifted. Anti-phospho-S306 antibody confirmed the phosphorylation of S306 (SI Appendix, Fig. S2L). S306 is a potential target of CKII, and CKII depletion decreased S306 phosphorylation (SI Appendix, Fig. S2M). CKII depletion or phospho-dead S306A mutation decreased UAF1 binding (SI Appendix, Fig. S2 M and N). On the contrary, the phospho-mimetic S306E mutant restored the mobility shift and interacted with UAF1 similarly to the wild type (SI Appendix, Fig. S2O). These findings indicate that ATAD5 S306 is constitutively phosphorylated by CKII, and this phosphorylation is not essential but enhances UAF1 binding. Even though phosphorylated S306 increased UAF1 binding, the S306A mutation had no significant effect on Ub–PCNA deubiquitination in cells (SI Appendix, Fig. S2P). Because ATAD5 (1 to 500) was purified from budding yeast, S306 is not phosphorylated. The phospho-mimetic S306E mutation had no significant effect on UAF1 binding or Ub–PCNA deubiquitination in vitro (SI Appendix, Fig. S2 Q–T). In spite of this, we introduced the S306E mutation to recombinant ATAD5 (1 to 500) for the in vitro studies of ATAD5–UAF1 and ATAD5–PCNA interactions.

Next, we investigate how UAF1 interacts with ATAD5. UAF1 is composed of a β-propeller domain with seven WD40 (WD) repeats, an ancillary domain, and two SUMO-like domains (SLDs) (SI Appendix, Fig. S2U). To investigate ATAD5 binding, domain deletion mutants of UAF1 were prepared (SI Appendix, Fig. S2U). SV40 nuclear localization signal (NLS) was added to the C terminus of UAF1 to prevent mislocalization. Interestingly, deletion of the β-propeller domain abolished the ATAD5–UAF1 interaction (SI Appendix, Fig. S2U). Ablation of SLDs reduced ATAD5 binding, but did not eliminate it. Previous research claimed that ATAD5 primarily interacts with UAF1 SLD2 (38), but our results indicate that ATAD5 binds to both the β-propeller domain and the SLDs of UAF1. UAF1 binds to FANCI for the deubiquitination of FANCI–FANCD2 (39). Similar to the ATAD5–UAF1 interaction, both the β-propeller domain and the SLDs of UAF1 participate in FANCI binding. We mutated FANCI-interacting residues in UAF1 and examined whether ATAD5 binds to UAF1 in a manner similar to FANCI (Fig. 2C and SI Appendix, Fig. S2V). ATAD5 binding was eliminated by mutations in the β-propeller domain, while mutations in SLD2 reduced ATAD5 binding. The binding of USP1 was unaffected by those mutations. Collectively, our findings demonstrate that ATAD5–UAF1 interaction is more complex than was previously anticipated, with multiple interaction motifs ensuring the association between ATAD5 and UAF1.

Mutations in UB1 or UB3 that disrupt the ATAD5–UAF1 interaction significantly increased the level of Ub–PCNA in cells, highlighting the significance of ATAD5 binding to UAF1 in regulating Ub–PCNA level (Fig. 2D). In vitro, as in cells, the UB1m mutant ATAD5 (1 to 500) did not interact with UAF1 (Fig. 2 E and F and SI Appendix, Fig. S2W). Importantly, ATAD5 (1 to 500, UB1m) and ATAD5 (1 to 500, UB1m+UB3m) were severely deficient in promoting deubiquitination of Ub–PCNA in reconstituted systems (Fig. 2 G and H and SI Appendix, Fig. S2 W–Y). These biochemical results demonstrate that the interaction between ATAD5 and UAF1 is essential for UAF1–USP1 to efficiently deubiquitinate Ub–PCNA.

ATAD5 Bridges Ub–PCNA and UAF1–USP1.

ATAD5 (1 to 400) is sufficient for UAF1 binding (SI Appendix, Fig. S2B), but ATAD5 (401 to 500) is required to enhance Ub–PCNA deubiquitination (Fig. 1A). These results indicate that ATAD5–UAF1 interaction is not sufficient for the facilitation of Ub–PCNA deubiquitination. Interestingly, we found that ATAD5 (1 to 500) bound to PCNA and Ub–PCNA (Fig. 3 A and B). The highly conserved UB4 motif of ATAD5, residues 399 to 410, locates just downstream of the UB3 motif (SI Appendix, Fig. S2 E and H). The mutation of the UB4 motif substantially reduced PCNA and Ub–PCNA binding (Fig. 3 A and B and SI Appendix, Figs. S2E and S3A). However, the UB4m mutation had no effect on UAF1 binding (Fig. 2 E and F). The UB1m mutation did not affect PCNA binding (SI Appendix, Fig. S3B). These results suggest that UB4 (ERQQFMKAFRQP) is a noncanonical PCNA-binding motif (ATAD5 PB) that is distinct from the conventional PIP box (Q-X-X-(V/L/M/I)-X-X-(F/Y)-(F/Y)). Importantly, ATAD5 (1 to 500, UB4m) failed to fully promote Ub–PCNA deubiquitination in reconstituted assays (Fig. 3 C and D). Efficient Ub–PCNA deubiquitination in cells also required ATAD5–PCNA interaction (SI Appendix, Fig. S3C). It has been unclear how ATAD5 links UAF1–USP1 and Ub–PCNA for deubiquitination. Our results indicate that ATAD5-N, not the ATAD5 C-terminal PCNA-unloading domain, recognizes Ub–PCNA for deubiquitination. Therefore, ATAD5-N serves as a molecular platform that bridges UAF1–USP1 and Ub–PCNA to facilitate the deubiquitination (SI Appendix, Fig. S3D).

Fig. 3.

ATAD5 recognizes Ub–PCNA for deubiquitination. (A) ATAD5 (1 to 500) binds to Ub–PCNA through UB4 motif. PCNA and Ub–PCNA were obtained by DNase I digestion after PCNA-loading and PCNA-ubiquitination reactions, respectively. Purified wild-type or UB4m ATAD5 (1 to 500) was incubated with PCNA or Ub–PCNA and then isolated with S-protein beads. (B) Quantification of PCNA or Ub–PCNA that is coisolated with ATAD5 (1 to 500) in A (n = 2). (C) ATAD5–PCNA interaction is important for the facilitation of Ub–PCNA deubiquitination. The indicated amount of wild-type or UB4m ATAD5 (1 to 500) is included in the deubiquitination reaction. (D) Quantification of the level of Ub–PCNA deubiquitination shown in C. The error bars indicate SD (n = 3).

ATAD5 Facilitates the Deubiquitination of Ub–PCNA Prior to PCNA Unloading.

As shown in Figs. 1–3, ATAD5 is an essential factor for the efficient deubiquitination of Ub–PCNA by UAF1–USP1. We questioned why ATAD5 participates in Ub–PCNA deubiquitination. Intriguingly, UAF1–USP1 deubiquitinated DNA-loaded Ub–PCNA more efficiently than free Ub–PCNA which were released from DNA by DNase I treatment (Fig. 4A and SI Appendix, Fig. S4 A and B). UAF1 interacts with DNA, and DNA enhanced FANCI–FANCD2 deubiquitination by UAF1–USP1 (37). The DNA binding-deficient UAF1 3A mutation (37) substantially reduced Ub–PCNA deubiquitination (Fig. 4B and SI Appendix, Fig. S1I), suggesting that DNA is required for the efficient action of UAF1–USP1 on Ub–PCNA. These findings imply that UAF1–USP1 deubiquitinates DNA-loaded Ub–PCNA preferentially. On the other hand, ATAD5–RLC can unload Ub–PCNA from DNA (46). The level of Ub–PCNA in the soluble fraction increases in UAF1-binding defective ATAD5 cells (SI Appendix, Fig. S4E). Importantly, ATAD5 (1 to 500) enhanced the deubiquitination of DNA-loaded Ub–PCNA but not of free Ub–PCNA which were released from DNA by DNase I or Hind III treatment (Fig. 4C and SI Appendix, Fig. S4 C and D).

Fig. 4.

ATAD5 facilitates Ub–PCNA deubiquitination on DNA. (A) UAF1–USP1 deubiquitinates DNA-loaded Ub–PCNA preferentially. DNA-loaded Ub–PCNA is produced as described in Fig. 1B. Free Ub–PCNA was obtained by digestion of DNA by DNase I after the ubiquitination reaction. (B) UAF1 (3A)–USP1 is defective in deubiquitination of DNA-loaded Ub–PCNA. UAF1 (3A) is a DNA-binding defective mutant (37). (C) ATAD5 (1 to 500) facilitates deubiquitination of DNA-loaded Ub–PCNA. After PCNA ubiquitination, DNA was cleaved by Hind III to release Ub–PCNA from DNA. (D) ATAD5 (439 to 500) is required for efficient DNA binding. EMSA was performed with the DNA probe shown in the upper panel. The DNA probe was fluorescently labeled with FAM at the indicated position. (E) Quantification of DNA binding shown in D. The relative amounts of the shifted DNA compared to the unbound DNA in the mock reaction are presented. The error bars indicate SD (n = 3). (F) ATAD5 (439 to 500) is required for efficient Ub–PCNA deubiquitination. The indicated amounts of ATAD5 variants are included in the deubiquitination reaction. (G) Quantification of the level of Ub–PCNA deubiquitination shown in F. The error bars indicate SD (n = 3).

Because ATAD5-N promoted Ub–PCNA deubiquitination on DNA, we investigated whether ATAD5 recognizes DNA for the deubiquitination. We performed electrophoretic mobility shift assay (EMSA) with gapped DNA and ATAD5 N-terminal fragments (Fig. 4 D and E and SI Appendix, Fig. S4F). ATAD5 (1 to 500) bound to DNA with similar affinity as UAF1–USP1. DNA binding to ATAD5 (1 to 400) and ATAD5 (1 to 438) was two- to threefold lower than that of ATAD5 (1 to 500), indicating that ATAD5 (439 to 500) is important for efficient DNA binding. On the other hand, mutations that impair UAF1- or PCNA-binding had no effect on DNA binding (SI Appendix, Fig. S4 G and H). Importantly, ATAD5 (1 to 438) was deficient in Ub–PCNA deubiquitination despite containing UAF1- and PCNA-binding motifs (Fig. 4 F and G). ATAD5 (1 to 400) was additionally deficient in Ub–PCNA deubiquitination due to its lack of PCNA-binding motif. ATAD5 (472 to 484) contains conserved positively charged amino acids (SI Appendix, Figs. S2H and S4I). Alanine mutations of those positively charged amino acids, ATAD5 DBm, reduced DNA binding (SI Appendix, Fig. S4 J–L). In addition, ATAD5 DBm promoted Ub–PCNA deubiquitination less efficiently than the wild type (SI Appendix, Fig. S4 M–O). These results suggest that ATAD5–DNA interaction is important for promoting Ub–PCNA deubiquitination. It is notable that binding motifs for UAF1, PCNA, and DNA are clustered. Combination of mutations in the ATAD5 deubiquitination domain, such as UB1m+UB4m and UB1m+UB4m+DBm, exhibited additive defects in Ub–PCNA deubiquitination compared to UB1m alone, suggesting that each domain contributes to Ub–PCNA deubiquitination (SI Appendix, Fig. S4 P–T).

Collectively, our findings imply that ATAD5–UAF1 interaction is not sufficient for Ub–PCNA deubiquitination, whereas ATAD5 mediated enzyme–substrate complex formation on DNA is necessary for efficient Ub–PCNA deubiquitination prior to PCNA unloading.

The ATAD5 Function in Ub–PCNA Deubiquitination Is Important for Genomic Integrity.

Accumulation of free Ub–PCNA may lead to improper chromatin loading of Ub–PCNA, as RFC could reload Ub–PCNA to primer-template DNA (SI Appendix, Fig. S5A). If Ub–PCNA is loaded to primer-template DNA without DNA damage, the false lesion-bypass signal may impede unchallenged DNA synthesis. We investigate how improper regulation of Ub–PCNA deubiquitination affects genome integrity. ATAD5 is an essential component of the PCNA-unloading complex. Therefore, ATAD5 deficiency results in a variety of cellular abnormalities that are primarily caused by an abnormal accumulation of chromatin-bound PCNA. To precisely inhibit Ub–PCNA deubiquitination, the ATAD5 deubiquitination mutation was implemented. DNA synthesis is not impaired in ATAD5 UB1m cells in the absence of replication stress (SI Appendix, Fig. S5B). However, ATAD5 deubiquitination mutant cells were more sensitive to UV or H₂O₂ (Fig. 5 A and B). As anticipated, the UB1m mutation substantially increased the Ub–PCNA level, particularly after DNA damage (SI Appendix, Fig. S5 C-D). Intriguingly, UV or H₂O₂ treatment resulted in a greater accumulation of γH2AX in ATAD5 UB1m cells than in wild-type cells, suggesting that the failure of timely Ub–PCNA deubiquitination causes the accumulation of DNA damage. Additionally, Ub–PCNA level was increased in ATAD5 deubiquitination mutant cells even without DNA damage agents (SI Appendix, Fig. S5E). These findings suggest that timely deubiquitination is important for the clearance of Ub–PCNA generated by spontaneous replication obstacles, and its improper regulation causes the accumulation of DNA damage.

Fig. 5.

The ATAD5-regulated Ub–PCNA deubiquitination is important for maintaining genomic integrity. (A and B) Misregulated Ub–PCNA deubiquitination increases sensitivity to DNA damage agents. The cell viability of wild-type, UB1m, UB1m+UB4m, or UB1m+UB4m+DBm ATAD5 expressing cells treated with the indicated dose of UV (A) or H₂O₂ (B), was determined. (C) Genotyping of wild type allele (Left) and UB3Δ mutant allele (Right). 12 bp deletion is shown in the mutant allele by genome analyzer (SI Appendix, Fig. S5F). (D and E) ATAD5–UAF1 interaction is important for genomic integrity in the mouse. The skin of three wild-type or three UB3Δ mutant mice (6 wk old) was irradiated with 1.5 kJ/m² of 365 nm UV daily twice. After 24-h recovery, mice skins were sectioned, and TUNEL assay was performed to measure DNA breaks (D), and number of TUNEL-positive cells were quantified (E). A higher number of TUNEL positive cells were observed in UB3Δ mice skin in the defined area (One-way ANOVA *P < 0.05, ***P < 0.001).

Next, we evaluated the in vivo function of ATAD5 in Ub–PCNA deubiquitination. To investigate this, ATAD5 mutant mice containing a four-amino acid deletion in the UB3 motif (UB3Δ) of murine ATAD5 (mATAD5) were produced (Fig. 5C and SI Appendix, Fig. S5F). We first examined the effect of the UB3Δ mutation on mouse embryonic stem (ES) cells derived from mutant embryos. UB3Δ mATAD5 was defective in UAF1 binding (SI Appendix, Fig. S5G) and accumulated chromatin-bound Ub–PCNA, particularly after UV irradiation (SI Appendix, Fig. S5H). We further analyzed the UB3Δ mutant mice (Fig. 5 D and E and SI Appendix, Fig. S5 I–M) to determine the effects of the mutation on murine genomic integrity. The homozygous mutant mice were born in the expected Mendelian ratios, developed normally, and did not show significant postnatal growth defects (SI Appendix, Fig. S5I). The adult homozygous mutant males and females were fertile, and histological analysis of the testis showed proper spermatogenesis (SI Appendix, Fig. S5 J and K). Because Ub–PCNA was accumulated under replication stress, we UV-irradiated the skin of wild-type or UB3Δ mutant mice and examined their sensitivity to DNA damage (Fig. 5 D and E). Surprisingly, the TUNEL assay confirmed that more DNA damage was accumulated in UB3Δ mouse skin compared to the wild-type mouse skin. Furthermore, γH2AX and cleaved caspase3 (c-CASP3)–positive cells were increased in the skin of UB3Δ mutant mice compared to the skin of wild-type mice (SI Appendix, Fig. S5 L and M). These results imply that proper Ub–PCNA deubiquitination is crucial for mitigating replication stress at the organismic level.

ATAD5 Facilitates Ub–PCNA Deubiquitination by USP7 and USP11.

UAF1–USP1 is a major deubiquitinase for Ub–PCNA. However, it is still possible that additional deubiquitination enzymes regulate Ub–PCNA deubiquitination. We questioned whether ATAD5-N cooperates with other deubiquitination enzymes for Ub–PCNA deubiquitination. To identify ATAD5-N interactors, FLAG-ATAD5 (1 to 500) is transiently expressed in 293T cells and affinity purified. Coisolated proteins with ATAD5 (1 to 500) were then analyzed by mass spectrometry (SI Appendix, Fig. S6A). As expected, UAF1 and USP1 were the most enriched proteins. Interestingly, several other USPs were coisolated with ATAD5 (1 to 500). We focused on USP7 and USP11 among ATAD5-N bound USPs, because it was reported that USP7 deubiquitinates Ub–PCNA and that USP11 binds to USP7. We thought that USP12 and USP46 were isolated through their interaction with UAF1, because these enzymes require UAF1 as a cofactor (34). To validate the role of USP7 and USP11 in Ub–PCNA deubiquitination, we measured the level of Ub–PCNA after depleting them with RNAi (Fig. 6A). USP7 or USP11 depletion increased the amount of chromatin-bound Ub–PCNA particularly after UV treatment, but to a lesser extent than USP1 or ATAD5 depletion. This indicates that Ub–PCNA is a substrate for USP7 and USP11 in cells. Next, we validated the interaction between ATAD5 and USP7 or USP11 using immuno-precipitation (Fig. 6B and SI Appendix, Fig. S6 B and C). USP7 and USP11 were both coimmunoprecipitated with ATAD5 (Fig. 6B). In addition, ATAD5 was coimmunoprecipitated with FLAG-tagged USP7 and FLAG-tagged USP11 (SI Appendix, Fig. S6 B and C). The low quantity of copurified USP7 and USP11 with ATAD5 suggested that the interaction between ATAD5 and USP7–USP11 is more transient than ATAD5–UAF1 interaction. Additionally, USP7 was coprecipitated with USP11, confirming previously published findings (57). Immunoprecipitation experiments with ATAD5 deletion variants showed that USP11 binds to the N terminus of ATAD5, ATAD5 (1 to 240) (Fig. 6C). On the other hand, USP7 was primarily bound to ATAD5 (384 to 600). USP7 contains three domains: the TRAF domain (1 to 204), the catalytic domain (205 to 537), and UBL domains (538 to 1,102) (SI Appendix, Fig. S6D). ATAD5 was associated with the UBL domains of USP7 (SI Appendix, Fig. S6E). In the case of USP11, the catalytic domain of USP11 (295 to 963) interacted with ATAD5 (SI Appendix, Fig. S6 D and F). USP11 catalytic domain also bound to the TRAF domain of USP7 (SI Appendix, Fig. S6 G and H). We identified the ATAD5-N motifs which participate in USP7 or USP11 binding through mutational analysis (SI Appendix, Figs. S2H and S6 I and J). These findings suggest that USP7–USP11 binds to ATAD5 through motifs distinct from the UAF1-binding motif.

Fig. 6.

ATAD5 enhances Ub–PCNA deubiquitination by USP7 and USP11 and facilitates poly-Ub–PCNA deubiquitination. (A) USP1, USP7, and USP11 regulate Ub–PCNA level. ATAD5 or indicated USP was depleted by siRNA treatment. The level of chromatin-bound Ub–PCNA was monitored after chromatin fractionation. Numbers below PCNA and Ub–PCNA blots indicate relative amount of PCNA and Ub–PCNA. (B) ATAD5 binds to USP7 and USP11. FLAG-tagged ATAD5 was pulled down from ATAD5 wild-type add-back ATAD5 knock-out cells. (C) USP7- and USP11-binding domains of ATAD5 are distinct from UAF1-binding motif. Indicated ATAD5 variants were expressed and then FLAG immuno-precipitated. (D) USP7 deubiquitinates DNA-loaded mono-Ub–PCNA more efficiently than free mono-Ub–PCNA which were released from DNA by Hind III treatment. (E) USP11 demonstrates higher deubiquitination activity for DNA-released mono-Ub–PCNA compared to DNA-loaded mono-Ub–PCNA. (F and G) ATAD5 enhances the deubiquitination activity of USP7 and USP11 for DNA-loaded mono-Ub–PCNA. In vitro Ub–PCNA deubiquitination assay was performed with USP7 (F), or USP11 (G), in the presence of ATAD5 (1 to 603). (H) ATAD5 (1 to 500) enhances cleavage of the polyubiquitin chain from DNA-loaded poly-Ub–PCNA by UAF1–USP1. (I and J) ATAD5 (1 to 603) facilitates degradation of the polyubiquitin chain on DNA-loaded poly-Ub–PCNA by USP7 and USP11.

We next investigated whether ATAD5 enhances the deubiquitination of Ub–PCNA by USP7 and USP11 (Fig. 6 D–G and SI Appendix, Fig. S6 K–P). Purified USP7 and USP11 deubiquitinated Ub–PCNA, but higher enzyme concentrations of USP7 or USP11 were required to fully deubiquitinate Ub–PCNA compared to UAF1–USP1 (Fig. 6 D and E and SI Appendix, Fig. S6 K–M). Intriguingly, USP7 and USP11 exhibited different specificity for free Ub–PCNA compared to UAF1–USP1 (Figs. 4A and 6 D and E). USP7 could deubiquitinate DNA-released PCNA, but deubiquitinated DNA-loaded Ub–PCNA more efficiently. In the case of USP11, DNA-released Ub–PCNA was deubiquitinated more efficiently compared to DNA-loaded Ub–PCNA. Because USP7 is bound to ATAD5 (380 to 600), we performed the Ub–PCNA deubiquitination assay in the presence of ATAD5 (1 to 603) (Fig. 6 F and G and SI Appendix, Fig. S6 N–P). ATAD5 (1 to 603) enhanced USP7- and USP11-mediated Ub–PCNA deubiquitination on DNA, despite the relative weak interaction between ATAD5 and USP7–USP11. 11m+7m mutation in ATAD5 partially reduced the USP7 and USP11 binding (SI Appendix, Fig. S6J). The mutation did not significantly reduce the deubiquitination of Ub–PCNA in vitro, possibly due to incomplete inactivation of USP7 or USP11 binding (SI Appendix, Fig. S6 N and Q–T). These results indicate that ATAD5-N associates with USP7 and USP11 to facilitate Ub–PCNA deubiquitination.

ATAD5 Cooperates with USPs to Deubiquitinate poly-Ub–PCNA.

Under replication stress conditions, HLTF or SHPRH polyubiquitinates PCNA (28, 29). The poly-Ub–PCNA drives replication fork remodeling and needs to be deubiquitinated to resume unchallenged DNA synthesis. Next, we investigated whether ATAD5 facilitates poly-Ub–PCNA deubiquitination. In addition to mono-Ub–PCNA, poly-Ub–PCNA accumulated on the chromatin following ATAD5 depletion, suggesting a role for ATAD5 in the deubiquitination of poly-Ub–PCNA (Fig. 6A). ATAD5 and USPs may work together to deubiquitinate poly-Ub–PCNA. Therefore, we evaluated this hypothesis in vitro (Fig. 6 H–J and SI Appendix, Fig. S6U). After monoubiquitination by RAD6B-RAD18, K63-linked poly-Ub–PCNA was generated by incubating mono-Ub–PCNA with purified UBC13-MMS2 and HLTF (SI Appendix, Fig. S6V and W). The addition of UAF1–USP1 resulted in the cleavage of the polyubiquitin chain from PCNA (Fig. 6H and SI Appendix, Fig. S6X). Intriguingly, addition of ATAD5 (1 to 500) significantly enhanced the poly-Ub–PCNA deubiquitination. The addition of ATAD5 (1 to 500) did not significantly affect the chain length of the cleaved polyubiquitin. These findings indicate that ATAD5 stimulates UAF1–USP1 to cleave the isopeptide bond between the polyubiquitin chain and PCNA. Similar to the deubiquitination of mono-Ub–PCNA, both the UAF1-binding mutant and PCNA-binding mutant forms of ATAD5 (1 to 500) were severely defective in poly-Ub–PCNA deubiquitination. The amount of cleaved polyubiquitin chain did not perfectly correlate with poly-Ub–PCNA deubiquitination, possibly due to partial cleavage of polyubiquitin chain in solution by UAF1–USP1. In the case of USP7, ATAD5 (1 to 603) enhanced the USP7-mediated poly-Ub–PCNA deubiquitination (Fig. 6I and SI Appendix, Fig. S6Y). USP7 differed from UAF1–USP1 in poly-Ub–PCNA deubiquitination in that it promoted polyubiquitin chain shortening. The average length of the polyubiquitin chain cleaved by USP7 is shorter than when poly-Ub–PCNA was incubated with UAF1–USP1, but the chain was not completely digested to monoubiquitin. ATAD5 (1 to 603) appeared to enhance the activity of USP7 to cleave ubiquitin links within polyubiquitin chains. Additionally, ATAD5 (1 to 603) promoted the USP11-mediated polyubiquitin chain cleavage (Fig. 6J and SI Appendix, Fig. S6Z). The majority of the USP11-cleaved ubiquitin was monoubiquitin, indicating that USP11 cleaved the polyubiquitin chain from the free chain end. Collectively, USP1, USP7, and USP11 have distinct cleavage activities on poly-Ub–PCNA, and ATAD5 enhances the ability of each enzyme to deubiquitinate poly-Ub–PCNA cooperatively.

Discussion

DNA replisomes frequently encounter various lesions during DNA synthesis. Lesions are bypassed in order to complete DNA replication, preventing the prolonged replication fork stalling and the formation of single-stranded DNA gaps. Bypassed lesions are repaired in a postreplicative manner (21). PCNA ubiquitination is a key activator of DNA-lesion bypass. Because lesion bypass mechanisms are inaccurate and slow, Ub–PCNA must be deubiquitinated in a timely fashion. In this study, we reveal how ATAD5 modulates the deubiquitination of mono- or poly-Ub–PCNA.

We showed that UAF1–USP1 preferentially deubiquitinates DNA-loaded Ub–PCNA (Fig. 4A). Recent research suggested that insert L1 of USP1 is important for the deubiquitination of DNA-loaded Ub–PCNA (58). In addition, our results show that DNA binding property of UAF1 is important for Ub–PCNA deubiquitination on DNA (Fig. 4B). These findings suggest that both UAF1 and USP1 interact with DNA to efficiently deubiquitinate DNA-loaded Ub–PCNA. Because DNA-lesion induced K164-ubiquitination of PCNA occurs only on DNA, the requirement of DNA for efficient deubiquitination of Ub–PCNA may ensure specific turn-off of lesion bypass signal. However, this property of UAF1–USP1 can be challenged by the release of Ub–PCNA from DNA.

Eukaryotic cells have ELG1-RLC or ATAD5–RLC that is specialized for PCNA unloading. We previously showed that ATAD5–RLC can unload Ub–PCNA from DNA (46). Therefore, ATAD5 needs to participate Ub–PCNA deubiquitination process to coordinate Ub–PCNA deubiquitination and PCNA unloading. PCNA unloading should not precede Ub–PCNA deubiquitination. If Ub–PCNA is unloaded before deubiquitination, the population of Ub–PCNA that escapes deubiquitination may increase.

If Ub–PCNA persists after legion bypass, it may inhibit proper polymerase exchange and hinder the resumption of normal DNA synthesis. On the other hand, if Ub–PCNA is unloaded without deubiquitination, accumulated Ub–PCNA in nucleosol can be reloaded to primer-template DNA by RFC. The inappropriately loaded Ub–PCNA may interfere with unchallenged DNA synthesis thereby decreasing genome stability (Fig. 5). The relatively mild sensitivity of deubiquitination defective ATAD5 mutant cells against DNA damage agents suggests that there is another layer of regulatory mechanism that limits inappropriate usage of free Ub–PCNA.

The coordination of Ub–PCNA deubiquitination and PCNA unloading is achieved by the tight association of UAF1–USP1 with ATAD5. ATAD5 (1-500)-UAF1–USP1 can be purified as a stoichiometric heterotrimer. ATAD5 and UAF1–USP1 appear to form a stable complex in cells and operate as a functional unit. It is interesting that UAF1 utilizes the same residues for binding ATAD5 and FANCI. These findings suggest that UAF1 has a conserved recognition surface for substrates. ATAD5 is not a substrate for UAF1–USP1, but ATAD5-N bridges Ub–PCNA and UAF1–USP1. A PCNA-binding motif is located just downstream of the UAF1-binding motif in ATAD5-N (Fig. 3). ATAD5-N appears to keep UAF1–USP1 in position so that it can efficiently access the isopeptide bond between ubiquitin and PCNA K164, analogous to what FANCI does for the deubiquitination of FANCD2. Because of this structural organization, ATAD5 can specifically enhance deubiquitination of Ub–PCNA by USP1. Interestingly, each component of the Ub–PCNA deubiquitinating complex, UAF1, USP1, and ATAD5-N, binds to DNA. Furthermore, DNA binding to each component of the deubiquitination enzyme complex is important for the efficient deubiquitination of Ub–PCNA (Fig. 4). These findings suggest that intensive interactions between DNA and the deubiquitination enzyme complex may ensure the deubiquitination of Ub–PCNA on DNA.

PCNA has two faces. Ubiquitin moiety is attached to the back face of PCNA. UAF1–USP1 and ATAD5-N should be positioned on the back face of PCNA. On the other hand, PCNA-unloading machinery, which contains ATAD5-C, approaches the front face of PCNA, so ubiquitination does not impede unloading. Therefore, ATAD5–RLC encompasses both sides of PCNA. If either side of PCNA is bound by other proteins, ATAD5–RLC may not be able to access PCNA. In addition, DNA binding by UAF1, USP1, and ATAD5 may require the presence of free DNA surrounding Ub–PCNA. This characteristic of ATAD5–RLC can provide a mechanism for the selective unloading and deubiquitination of unoccupied PCNA or Ub–PCNA.

In addition to UAF1–USP1, ATAD5-N interacts with USP7 and USP11 to promote Ub–PCNA deubiquitination by these enzymes (Fig. 6). ATAD5-mediated enhancement of Ub–PCNA deubiquitination by various USPs is consistent with the idea that ATAD5 functions as a bridge between USPs and Ub–PCNA. UAF1, USP7, and USP11 bind to distinct ATAD5-N motifs. ATAD5-N forms a stable complex with UAF1–USP1, but its interactions with USP7 and USP11 are weak. USP7 and USP11 may associate transiently with the ATAD5–UAF1–USP1 complex. ATAD5 enhanced Ub–PCNA deubiquitination by USP7 and USP11 in vitro (Fig. 6). USP7 and USP11 may function as a backup deubiquitinase for Ub–PCNA. Otherwise, USP7 and USP11 may aid in the concerted deubiquitination of Ub–PCNA in which multiple subunits are ubiquitinated.

In addition to mono-Ub–PCNA, ATAD5 facilitates poly-Ub–PCNA deubiquitination (Fig. 6). ATAD5-N enhances the activity of USP1 to cleave the link between PCNA and polyubiquitin chain. This result supports an idea that ATAD5–UAF1 interaction enhances positioning of USP1 active site adjacent to the isopeptide bond between ubiquitin and PCNA. USP7 cut the polyubiquitin chain into varying lengths, whereas USP11 primarily produces monoubiquitin by cleaving ubiquitin from the free end of the polyubiquitin chain. Due to the variable length of polyubiquitin chain, the loose association of USP7 and USP11 to ATAD5-N can provide flexibility for the efficient degradation of polyubiquitin chains. Through the cooperation of ATAD5, UAF1–USP1, USP7, and USP11, poly-Ub–PCNA can be efficiently recycled to unmodified PCNA and ubiquitin monomers.

Our findings elucidate that ATAD5 serves as a molecular platform for Ub–PCNA deubiquitination and coordinates Ub–PCNA deubiquitination with PCNA unloading. ATAD5 function is essential for the timely deubiquitination of DNA-loaded Ub–PCNA. This report provides a molecular mechanism by which deubiquitination function of ATAD5 contributes to the maintenance of genomic integrity.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.