The ADP-ribosyltransferase PARP10/ARTD10 Interacts with Proliferating Cell Nuclear Antigen (PCNA) and Is Required for DNA Damage Tolerance*

Background: PCNA mono-ubiquitination at stalled replication forks recruits translesion synthesis polymerases for fork restart. Results: The mono-ADP-ribosyltransferase PARP10 interacts with PCNA through a PIP-box. PARP10 knockdown results in DNA damage hypersensitivity and defective translesion synthesis. Conclusion: PARP10 participates in PCNA-dependent DNA damage tolerance. Significance: This is the first time that post-translational modification by mono-ADP-ribosylation is implicated in DNA repair. All cells rely on genomic stability mechanisms to protect against DNA alterations. PCNA is a master regulator of DNA replication and S-phase-coupled repair. PCNA post-translational modifications by ubiquitination and SUMOylation dictate how cells stabilize and re-start replication forks stalled at sites of damaged DNA. PCNA mono-ubiquitination recruits low fidelity DNA polymerases to promote error-prone replication across DNA lesions. Here, we identify the mono-ADP-ribosyltransferase PARP10/ARTD10 as a novel PCNA binding partner. PARP10 knockdown results in genomic instability and DNA damage hypersensitivity. Importantly, we show that PARP10 binding to PCNA is required for translesion DNA synthesis. Our work identifies a novel PCNA-linked mechanism for genome protection, centered on post-translational modification by mono-ADP-ribosylation.

The ability of cells to repair damaged DNA and faithfully replicate and transfer the genome to their progeny is essential for suppressing cellular transformation (1). To avoid genomic instability, cells employ numerous mechanisms that detect, signal, and repair DNA lesions, stabilize and protect complex structures such as replication forks and DNA fragile sites, and ensure the correct and timely replication and segregation of chromosomes. These mechanisms are subjected to complex cellular regulation and are frequently inactivated in cancer.
DNA replication in S-phase is a particularly dangerous cellular process, since complex chromosomal structures need to be unfolded and reassembled, mutations can occur spontaneously due to nucleotide misincorporation, and unrepaired DNA lesions can block polymerase progression threatening the stability of replication forks and leading to DNA breaks (2). Many processes occurring at replication forks are controlled by the protein proliferating cell nuclear antigen (PCNA), 2 a homotrimeric DNA clamp that provides processivity to DNA polymerases and acts as a docking platform for numerous repair and replication-associated factors (3). PCNA participates in an astoundingly large number of diverse genomic stability mechanisms, including chromatin assembly and remodeling (4), cell cycle control (5), prevention of re-replication (6), and suppression of toxic recombination (7,8).
PCNA is also essentially required for efficient DNA repair. Several DNA repair pathways including mismatch repair, base excision repair, and nucleotide excision repair employ PCNA to help DNA scanning, lesion recognition, and DNA repair synthesis (3). Moreover, PCNA post-translational modifications by ubiquitin and SUMO further expand the repertoire of PCNA functions. In response to replication-blocking DNA damage, ubiquitinated PCNA coordinates the filling of single-stranded gaps behind replication forks through the recruitment of specialized effector complexes (9 -13). Mono-ubiquitinated PCNA specifically interacts with translesion synthesis (TLS) polymerases of the Y family, including Rev1, Poln, Polk, and Pol. These polymerases contain a specific PCNA-interacting region termed PIP-box, which is common to many PCNA interactors (14), but also ubiquitin-binding domains, which ensure the specificity of the interaction to ubiquitinated PCNA. Thus recruited to stalled replication forks, TLS polymerases replace the normal replicative polymerases and can catalyze DNA synthesis across and beyond the lesion, because of their more relaxed active site (12,(15)(16)(17)(18). Frequently, this results in mutations caused by incorrect nucleotide insertion across the lesion, low fidelity, and lack of proofreading activity. Typically, replicative polymerases switch back on the template shortly after the lesion is bypassed to ensure high fidelity replication. This is achieved through PCNA de-ubiquitination by the enzyme USP1, which is essential for suppressing mutagenesis (19).
At sites of DNA lesions, PCNA can also be multiubiquitinated by K63-linked ubiquitin chains. This modification is believed to promote a so far poorly characterized error-free mechanism of lesion bypass, involving a template switch to the newly replicated strand of the sister chromatid (3,20). Finally, PCNA can also be modified by covalent addition of the ubiquitin-related protein SUMO. This S-phase specific modification was shown to repress toxic recombination events of replicating chromatids by recruiting factors that can remove the recombination protein RAD51 from DNA (7,8,21,22).
ADP-ribosylation is a unique post-translational modification that has been shown to participate in a large number of cellular processes including transcriptional regulation, cell signaling, cell death, energy metabolism, and DNA repair (23)(24)(25). This modification is catalyzed by poly(ADP-ribose) polymerases (PARPs), also termed ADP-ribosyltransferases diphteria toxinlike (ARTDs). These enzymes use NADϩ as substrate to transfer ADP-ribosyl to an amino acid receptor (glutamic acid, aspartic acid, or lysine). The moiety can also be transferred to another ADP-ribosyl molecule, resulting in formation of poly-(ADP-ribose) (PAR) chains, which can be linear or branched. The PARP family contains at least seventeen members (26). The best characterized are the founding members PARP1and PARP2, which account for well over 90% of cellular ADP-ribosyl transferase activity and are known to participate in many cellular processes, including repair of DNA strand breaks (23,24,27).
It has been recently shown that a number of PARP family members are unable to catalyze poly(ADP-ribose) chain formation. The members of this subset of PARPs lack a critical active site glutamate residue and thus are only able to transfer a single ADP-ribosyl moiety (25, 28 -30). PARP10 (also termed ARTD10) is a mono-ADP-ribosyl (MAR) transferase, originally identified as a Myc-interacting protein (31). Subsequently it has been proposed to be important for the G1/S cell cycle transition (32) as well as for caspase-dependent apoptosis (33). Similar to other PARPs, PARP10 has auto-catalytic activity (30). Recent work has uncovered a number of other substrates targeted for MARylation by PARP10: GSK3␤ MARylation was shown to inhibit its kinase activity (34), while MARylation of NEMO was shown to block the activation of NF-B pathway by extracellular signals (35). A protein microarray screen identified over 70 substrates for PARP10-catalyzed MARylation (34), but the extent to which most of these substrates are modified in vivo is unknown. Recently, the macrodomain-containing protein PARP14/ARTD8 was shown to recognize MARylated PARP10 and NEMO, thus acting as a reader for protein MARylation (35,36).
To date, it is not known if PARP10, or protein MARylation, directly participate in DNA repair (29). Here we show that PARP10 specifically binds to ubiquitinated PCNA and participates in PCNA-mediated translesion synthesis.

EXPERIMENTAL PROCEDURES
Cell Culture and Protein Techniques-Human HeLa, 293T, and U2OS cells were grown in DMEM (Lonza) supplemented with 15% fetal calf serum. Native whole cell extracts for co-immunoprecipitation and GST-pulldown studies, were obtained by incubating cells for 30 min on ice with HEPES lysis buffer (50 mM HEPES, 1% Triton, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 10 mM MgCl 2 containing protease inhibitors). For co-immunoprecipitation experiments, lysates were prepared in the presence of 20 mM crosslinking agent DSP (Thermo Scientific). Denatured whole cell extracts were prepared by boiling cells in 100 mM Tris, 4% SDS, 0.5 M ␤mercaptoethanol. Antibodies used for Western blotting and immunoprecipitation were: PCNA (Abcam), PARP10 (Novus), PARP1 (Abnova), Myc, and actin (both from Santa Cruz Biotechnology).
LUMIER with Bait Control Assay-Reporter 293T cells stably expressing a codon-optimized Renilla luciferase fused to the C terminus of human PCNA were reverse transfected with 3ϫ Flag-tagged PCNA binding partners or GFP control. LUMIER with BACON was done as previously described (37). The interaction strength between PCNA and its binding partners was calculated as [PCNA/partner].
Immunofluorescence Assays-Cells were fixed with 4% paraformaldehyde for 10 min at room temperature, followed by extraction with 0.3% Triton X-100 for 10 more minutes on ice. Slides were blocked with 5% BSA, 0.1% Triton in PBS for 30 min, followed by incubation with the primary antibody diluted in 3% BSA in PBS, for 1 h at 37°C. After three PBS washes, the secondary antibody (Alexa Fluor 488 from Invitrogen) was added for 1 h. Slides were mounted with DAPI-containing Vectashield mounting medium (Vector Labs). Primary antibodies used include RAD51, GFP (both from Santa Cruz Biotechnology), phospho-RPA32 S33 (Bethyl Laboratories), and ␥H2AX (Abcam).
Plasmids and Small Interfering RNA (siRNA)-The cDNA for PARP10 and PCNA were purchased from PlasmID, and Origene, respectively. For transient transfection, cDNA fragments were cloned into pCMV-Myc (Clonetech). For bacteria expression, the pGEX plasmid (GE Healthcare) was used. Plasmid transfections were performed using Lipofectamine LTX (Invitrogen). Lipofectamine RNAiMAX (Invitrogen) was used for siRNA transfection. siRNA oligonucleotides were purchased from Invitrogen and Qiagen. The siRNA sequences used are shown below (if the oligo number is not shown in the figure, the sequence no. 1 was employed). As controls, AllStars Negative Control (Qiagen) and Stealth RNAi Negative Controls (Invitrogen) were employed. siPARP10 1: GCCTGGTGGAGATGGTGCTATTGAT; siPARP10 2: TGAAGGACCGGATATGACTGGCTTT; siBRCA2: TTGAAGAATGCAGGTTTAATA; siPARP1: AAA-CATGGGCGACTGCACCATGATG.
BrdU Incorporation-Quantification of DNA replication by BrdU incorporation was done as previously described (38). Briefly, 5 ϫ 10 5 cells were incubated with 20 mM BrdU for 30 min, washed, and fixed overnight at 4°C in 70% ethanol. Cells were then incubated with 2 N HCl/0.5% Triton X-100 for 30 min and 0.1 mM sodium tetraborate pH 8.5 for 1 min. Next, cells were incubated with anti-BrdUrd antibodies (Pierce) and Alexa-Fluor 488-conjugated anti-mouse secondary antibodies (Invitrogen) for 30 min each and analyzed using a FACSCalibur (BD Biosciences) instrument.
Functional Assays-Clonogenic assays were perfomed as previously described (39). Two rounds of siRNA were performed, 24 h apart. The day after the second siRNA treatment 400 cells were plated in each dish, and 8 h later the DNA-damaging drugs were added. Colony formation was scored after 2 weeks using Crystal Violet solution (ScholAR Chemistry). Translesion synthesis SupF assay was previously described (40,41). Following two rounds of siRNA, 293T cells were transfected with UVC-irradiated (1000 J/m 2 ) pSP189 (SupF) plasmid. Three days later, the plasmid was recovered using a miniprep kit (Promega), DpnI digested and transformed into MBM7070 indicator bacteria. Transformants were selected on plates containing 1 mM IPTG and 100 g/ml X-gal. The ratio of white (mutant) to total (blue ϩ white) colonies was scored as mutation frequency.

RESULTS
PARP10 Interacts with PCNA via Its PIP-box-To identify genomic stability mechanisms operating at replication forks, we searched for novel PCNA-interacting proteins. We observed that the ADP-ribosyltransferase PARP10/ARTD10 contains a PCNA-interacting peptide (PIP) box at position 834 (Fig. 1A). The sequence QEVVRAFY matches the PIP-box consensus sequence QXXhXXaa (where h stands for hydrophobic, a for aromatic, and X for any amino acid). This sequence is located at the very beginning of the ADP-ribosyltransferase domain. While this domain, and especially the active site region (amino acids 884 -934 in PARP10) are highly conserved among PARPs, the PIP-box sequence is present exclusively in PARP10 (Fig. 1B), suggesting that PARP10 might specifically interact with PCNA using this PIP-box. To address this, we employed the LUMIER co-immunoprecipitation assay, which allows for quantification of protein interactions in vivo (37). We co-expressed in 293T cells luciferase-tagged PCNA and Flag-tagged PARP10 (and other PCNA-interacting proteins as positive controls) and performed Flag immunoprecipitation. Co-purified luciferase-tagged PCNA was detected by luminescence measurement. We found that PARP10 interacts with PCNA; the strength of this interaction was similar to that of other known PCNA-interacting proteins (Fig. 1C). To confirm that PARP10 is a novel PCNA interaction partner, we bacterially expressed and purified recombinant GST-tagged PARP10, either fulllength or a fragment (800-end) spanning the ARTD domain and thus including the PIP box. Both PARP10 species were able to interact with PCNA from native whole cell extracts of 293T cells. In contrast, control empty GST could not pull-down PCNA (Fig. 1D). Further confirming the interaction between PCNA and PARP10, an antibody against endogenous PARP10 co-immunoprecipitated endogenous PCNA from HeLa cells (Fig. 1E). The interaction between endogenous PARP10 and PCNA was increased following UV irradiation, suggesting that this interaction participates in the response to DNA damage.
We next investigated if PCNA interaction occurs through the PIP-box domain of PARP10. We first analyzed two PARP10 mutants: one bearing the QEVVRAFY to QEVARAAA mutation (termed "Mut"), and one with a PIP-box (QEVVRAFY) deletion (termed "Del" or "⌬PIP"). The variants were expressed in bacteria with GST-tags, and purified to identical yields as wild-type versions. Both PIP mutants failed to interact with PCNA ( Fig. 2A). Similar results were obtained when we introduced the mutations in the 800-end fragment (data not shown). We also expressed Myc-tagged wild type PARP10, or the ⌬PIP variant in 293T cells. Reciprocal co-immunoprecipitation experiments showed that wild type, but not the PIP-box mutant PARP10 interacts with endogenous PCNA (Fig. 2, B and C). Next, we thought to introduce mutations in the PIP-box binding site of PCNA. Structural studies found that PIP-boxes plug into a hydrophobic pocket below the interdomain connecting loop of PCNA (3). A recent study showed that a point mutation in this hydrophobic pocket, M40R reduces PCNA binding to a peptide derived from FEN-1 (42). Consistent with this, when we performed GST-PARP10 pulldowns with Myc-tagged PCNA variants, we observed that the M40R mutant shows reduced interaction to PARP10 (Fig. 2D). Thus, we conclude that PARP10 specifically interacts with PCNA, using its PIP-box motif.
PARP10 also contains two ubiquitin-interacting motifs (UIMs) positioned in front of the ARTD domain. These domains were recently shown to bind K63-multiubiquitin chains such as those present on activated TRAF6 upon NFB pathway activation by extracellular ligands (35). Since PCNA is known to be ubiquitinated following replication fork stalling at sites of DNA damage, we investigated if the UIMs are involved in binding to ubiquitinated PCNA. First, we confirmed that PARP10 binds ubiquitin. We expressed Myc-tagged PARP10 or a PARP10 variant lacking the UIMs (termed "⌬UIM)" in 293T cells and used extracts of these cells for interaction studies with recombinant GST-ubiquitin. Wild type PARP10, but not the ⌬UIM version was able to interact with ubiquitin (Fig. 2E), confirming that PARP10 binds ubiquitin using the UIM domains. Next, we investigated binding to GST-tagged recombinant PCNA, or to a PCNA variant fused to ubiquitin. Such PCNAubiquitin fusions were previously successfully used to mimic ubiquitinated PCNA, which is present at very low cellular amounts incompatible with binding studies. For example, PCNA-ubiquitin fusions were employed to investigate interactions with TLS polymerases (43) and the crosslink repair factor SNM1 (44). Similar to PARP10, these factors contain PIP-boxes and ubiquitin-interacting domains, and are thought to be recruited to stalled replication forks by ubiquitinated PCNA. When we incubated recombinant GST-PCNA with extracts of 293T cells expressing Myc-tagged PARP10, the PCNA-PARP10 interaction was clearly observed also in this scenario. Importantly, PARP10 interacted significantly stronger (about 2-fold) with the PCNA-ubiquitin fusion (Fig. 2E). When we employed the PARP10 ⌬UIM variant, the interaction with PCNA was still present, but the binding to PCNA-ubiquitin was not strengthened any longer. Altogether, these results suggest that PARP10 interacts with ubiquitinated PCNA, using its PIPbox and UIM domains to bind to the PCNA and ubiquitin moieties, respectively.
It was previously shown that PCNA interacts with another member of the PARP family, namely PARP1, to inhibit its ADPribose polymerase activity (45). When we depleted PARP1 using siRNA, we observed that PARP10 co-immunoprecipitated with PCNA to the same extent as in control cells (Fig. 2F). Thus, we conclude that the PARP10-PCNA interaction occurs independently of PARP1.
PARP10 Is Required for Genomic Stability and DNA Damage Tolerance-PCNA is essential for genomic stability by acting as a master regulator of DNA replication and replication-linked processes. We wondered if PARP10 is involved in maintenance of genomic stability. To investigate PARP10 functions, we employed two different siRNA oligonucleotides targeting PARP10. Both siRNA oligonucleotides were able to knockdown PARP10 protein levels efficiently (Fig. 3A). PARP10 knockeddown HeLa cells were hyper-sensitive to the DNA-damaging agent mytomycin C (MMC), known to induce alkylating DNA damage and DNA crosslinks (Fig. 3, B and C). HeLa cells transfected with siRNA oligonucleotides targeting PARP10 were also significantly more sensitive to the replication fork-stalling agent hydroxyurea (HU), which acts by depleting nucleotide pools (Fig. 3D). We obtained similar results when using other DNA-damaging agents such as UV irradiation (Fig. 3E). In contrast, PARP10-depleted cells were not sensitive to ionizing radiation (IR)-mimeticum drug bleomycin, known to induce double-stranded breaks (Fig. 3F). These results show that PARP10 is a novel genomic stability factor required for cellular resistance to DNA-damaging agents that induce replication fork stalling.
Because PARP10 interacts with PCNA and is required for HU and UV resistance, we investigated if PARP10 participates in the response to stalled replication forks. We monitored recovery from replication fork stalling, by BrdU incorporation following exposure of HeLa cells to UV irradiation. Control cells slightly reduced BrdU incorporation under the conditions investigated; in contrast, PARP10-depleted cells showed a much more drastic and prolonged reduction in BrdU incorporation (Fig. 3G), suggesting that replication or repair of damaged chromatin is slower in PARP10-deleted cells. We conclude that PARP10 is required for removal or replicational bypass of DNA lesions in S-phase.
Accumulation and clearance of DNA damage can be monitored by the chromatin recruitment of DNA damage-associated proteins. Following PARP10 knockdown, we observed a marked increase in UV-induced ␥H2AX foci, a marker of damaged DNA (Fig. 4A). Moreover, we observed that phospho-RPA foci are also increased following PARP10 depletion from HeLa cells, suggesting increased or prolonged replication fork stalling (Fig. 4B). Finally, we observed increased spontaneous and CPTinduced RAD51 foci (Fig. 4, C and D). Altogether, these results argue that DNA damage is not efficiently cleared in PARP10depleted cells. These results show that PARP10 knockdown results in accumulation of DNA damage. We conclude that human PARP10 is required for DNA repair and genomic stability. Recombinant ubiquitin, PCNA, or a PCNA-ubiquitin in-frame fusion (lower panel) were incubated with extracts of 293T cells overexpressing Myc-tagged wild type PARP10, or a PARP10 variant harboring a deletion of the UIMs. F, PARP1 is not necessary for the PCNA-PARP10 interaction. 293T cells expressing myc-PARP10 were transfected with control or PARP1-targeting siRNA. Cells were lysed, and lysates were subjected to anti-PCNA immunoprecipitation. PARP10 co-purified with PCNA regardless of PARP1 levels. PARP1 also co-immunoprecipitated with PCNA in control cells, as previously shown (45).
PARP10 Is Required for Mutagenesis-Ubiquitinated PCNA recruits TLS polymerases to bypass DNA lesions thereby restarting stalled replication forks. TLS is a potentially mutagenic process, and conversely PCNA ubiquitination is required for maintaining normal mutation levels (16,19,46). Since PARP10 interacts with PCNA, we investigated if PARP10 is involved in mutagenesis. Using the SupF shuttle plasmid mutagenesis assay (40,41), we observed that PARP10 knockdown results in a 2-fold reduction in UV-induced mutation levels, compared with control (Fig. 5A). This result suggests that PARP10 contributes to maintaining normal mutation levels. Indeed, overexpression of wild type Myc-tagged PARP10 resulted in increased mutation rates (Fig. 5A). To test the role of PCNA interaction with PARP10 in TLS, we constructed siRNA-resistant variants of PARP10, and employed them to complement the TLS phenotype of PARP10-silenced cells in the SupF assay. Wild type PARP10 could correct this phenotype; in contrast, the PARP10 PIP-box mutant did not (Fig. 5, B and C). Moreover, expression of the PARP10 mutant G888W, lacking catalytic activity (30), also did not result in correction of the TLS phenotype. These results show that both PCNA interaction and the MARylation activity are essential for PARP10 function in TLS.
Because PARP10 knockdown resulted in reduced mutation rates, and PCNA ubiquitination is required for mutagenic TLS, we investigated if PCNA ubiquitination levels are affected by PARP10. We observed that PARP10 knockdown results in a significant decrease in PCNA ubiquitination (Fig. 5D). Ubiquitinated PCNA promotes mutagenesis by recruiting mutagenic TLS polymerases to replication forks. The formation of Rev1 foci following DNA damage exposure was previously found to be dependent on PCNA ubiquitination (43). We observed that PARP10 knockdown resulted in reduced accumulation of the TLS polymerase Rev1 to UV-induced nuclear foci (Fig. 5E). Altogether, our results strongly argue that PARP10 controls mutagenesis by regulating PCNA ubiquitination and subsequently recruitment of TLS polymerases to replication forks.

PARP10 Is a Novel Player in the DNA Damage Response
Program-DNA damage repair is a fundamental process for all living organisms. In metazoans, DNA repair is particularly important to ensure genomic stability and thus suppress cellular transformation and tumor formation. Many DNA repair pathways cooperate to detect, process, and remove DNA lesions in a timely and efficient manner. DNA damage signaling and repair factors are frequently found to be inactivated in cancers, highlighting the importance of genomic instability for cellular transformation.
The risk of accumulating mutations and structural damage of the chromosomes is particularly high during DNA replication, a process requiring complex gymnastics involving protein-DNA complexes. DNA lesions can block the progress of replication forks; prolonged replication fork stalling leads to disas- sembly of the replication machinery and double strand break formation (47). Cells have developed DNA damage response mechanisms that protect and stabilize replication forks, promote lesion bypass, and integrate cell cycle signals (3,47,48).
The ADP-ribosyltransferase family of proteins has long been implicated in protecting against DNA damage. The poly(ADP-ribose) polymerase PARP1 is important for activation of the DNA damage response and promotes recruitment of repair enzymes to chromatin (23,25,49,50). Much less is known about other ADP-ribosyltransferases. In particular, the substrates and functions of mono-ADP-ribosyltransferases remain mostly mysterious. The MARylating enzyme PARP10/ARTD10 was shown to participate in caspase-dependent apoptosis and NF-B pathway suppression (33,35). Here we show for the first time that PARP10 is required for DNA damage resistance. We found that PARP10depleted cells are hypersensitive to DNA-damaging agents, including MMC and HU (Fig. 3) and show increased chromatin binding of DNA damage response and repair proteins ␥H2AX, RAD51, and phospho-RPA (Fig. 4). These results suggest that PARP10-depleted cells accumulate DNA damage, and rule out that PARP10 DNA damage hypersensitivity is indirectly caused by its previously proposed role in apoptosis. Instead, our results suggest that PARP10 is actively participating in S-phase repair. Indeed, PARP10-depleted cells show an inability to restart DNA replication following exposure to S-phase-specific DNA damage (Fig. 3G). We propose that PARP10 is important for promoting replication completion by activating TLS at DNA damage sites.
PARP10 Couples Protein MARylation to the DNA Damage Response-Our data argue that upon replication fork stalling, PARP10 is recruited to stalled replication forks to promote genomic stability. However, the exact nature of PARP10 participation in replication fork stability is still unclear. Since PARP10 has been characterized as a mono-ADP-ribosylransferase, it is likely that this activity is required for PARP10 function in DNA repair. Indeed, we found that a catalytic inert PARP10 mutant is unable to promote TLS (Fig. 5). The substrates of PARP10 relevant for TLS are still unknown. A recent protein microarray in vitro screen identified over seventy potential PARP10 substrates (34), but there is little indication of the DNA damage-relevant substrate(s).
DNA damage signaling and repair are highly regulated and dynamic processes. To efficiently control these processes, cells employ a staggering array of post-translational modifications that can affect protein localization, function, and stability. Our data strongly suggest that protein MARylation is a novel posttranslational modification involved in DNA repair.
PARP10 and PCNA Ubiquitination-Our study identifies PARP10 as a novel PIP-box containing, PCNA-interacting protein. While the PIP-box is present at the very beginning of the conserved PARP domain, the motif is found uniquely in PARP10. In the crystal structure of PARP1, the region is present in a ␣-helix (termed ␣A) fold exposed on the surface of the domain (30). This argues that even though the overall fold of the PARP domain is conserved throughout the family members, PARP10 has gained the special ability to bind to PCNA on the surface of this domain. However, since PCNA binding occurs in close proximity to the PARP domain, it is possible that this interaction also modulates PARP10 catalytic activity. Indeed, preliminary studies showed that PCNA enhances the in vitro auto-MARylation activity of PARP10. 3 Thus, PCNA might contribute to both localization and catalytic activity of PARP10.
Our study adds PARP10 to the long list of PIP-box-containing PCNA-interacting partners. How competition of PCNA binding partners is regulated is still mysterious, but differences in interaction strength, as well as post-translational modifica-tions may play important roles (3,5,51). Indeed, our results suggest that PARP10 binding to PCNA is enhanced by PCNA ubiquitination (Fig. 2), which provides an additional binding surface for PARP10 and its UIM domains. Not only is PCNA ubiquitination important for efficient PARP10 binding, but it also requires normal PARP10 levels. Indeed, PARP10 knockdown results in significant reduction in PCNA ubiquitination, as well as TLS polymerase recruitment and translesion synthesis rates (Fig. 5). PARP10 binding to ubiquitinated PCNA might protect and stabilize this modified form of PCNA, similar to the effect of Srs2 binding to yeast SUMO-modified PCNA (8). Perhaps PARP10 binding blocks the binding of enzyme USP1, which de-ubiquitinates PCNA (19). Alternatively, PARP10 might actively inhibit USP1. PARP10 might also act on the RAD6-RAD18 machinery for PCNA ubiquitination, either by directly activating it, or by creating chromatin structures that recruit and activate it. In conclusion, PARP10 seem to be part of a positive feedback loop that recognizes and amplifies the PCNA ubiquitination signal. This is strikingly similar to the situation previously described for Spartan/C1orf124, another reader of PCNA ubiquitination involved in translesion synthesis (52)(53)(54). It will be important to investigate if Spartan and PARP10 work together. Interestingly, PARP10 contains two tandem UIM domains, found to bind to K63-linked multi-ubiquitin chains (35). PCNA is itself subjected to K63-linked multiubiquitination, and was recently shown to recruit the ZRANB3 helicase (55). This raises the possibility that PARP10 might also interact with multi-ubiquitinated PCNA.
Our results showing that PARP10 is required for cellular resistance to DNA-damaging drugs suggests that PARP10 activity might represent a barrier against efficient genotoxic cancer therapy. Thus, PARP10 expression or activity screening might represent a potent biomarker to predict therapeutic responses. Inhibition of DNA repair mechanisms has been shown to potentiate the effect of genotoxic cancer therapy (56,57). Several PARP10 inhibitors have already been identified using in vitro enzymatic assays. Identification of efficient and selective inhibitors of cellular PARP10 might provide tools for better cancer therapy, in combination with camptothecin or replication-inhibiting cancer drugs.