PCNA is involved in the EndoQ-mediated DNA repair process in Thermococcales

To maintain genome integrity for transfer to their offspring, and to maintain order in cellular processes, all living organisms have DNA repair systems. Besides the well-conserved DNA repair machineries, organisms thriving in extreme environments are expected to have developed efficient repair systems. We recently discovered a novel endonuclease, which cleaves the 5′ side of deoxyinosine, from the hyperthermophilic archaeon, Pyrococcus furiosus. The novel endonuclease, designated as Endonulcease Q (EndoQ), recognizes uracil, abasic site and xanthine, as well as hypoxanthine, and cuts the phosphodiester bond at their 5′ sides. To understand the functional process involving EndoQ, we searched for interacting partners of EndoQ and identified Proliferating Cell Nuclear Angigen (PCNA). The EndoQ activity was clearly enhanced by addition of PCNA in vitro. The physical interaction between the two proteins through a PIP-motif of EndoQ and the toroidal structure of PCNA are critical for the stimulation of the endonuclease activity. These findings provide us a clue to elucidate a unique DNA repair system in Archaea.

DNA is always under threat of change or loss of genetic information by endogenous or exogenous influences. To maintain genome integrity for their offspring, and to prevent disorder of a cell system, all living organisms have evolved DNA repair mechanisms. One of the predominant DNA damages encountered by cells is base deamination 1 . Deamination of cytosine, adenine, and guanine gives rise to uracil, hypoxanthine, and xanthine, respectively. Uracil and hypoxanthine are also mis-incorporated into the nascent DNA strand by DNA polymerase during replication. If these bases remain in DNA, they lead to point mutations during replication due to wrong base pairing. Generally, the deaminated bases are released from the DNA strand by a lesion-specific DNA glycosylase. The resultant apurinic/apyrimidinic (AP) site is recognized and the DNA backbone is cut by AP endonuclease (APE). DNA polymerase synthesizes the new strand and DNA ligase fills the nick. This repair pathway is called base excision repair (BER) 2,3 . Uracil-DNA glycosylase (UDG), which removes uracil, is the most conserved DNA glycosylase in all domains of life, Bacteria, Archaea and Eukarya. The UDGs are now classified into five families, based on their substrate specificity and amino acid sequence motifs in the active site, although the UDGs form a single protein superfamily with a common structural fold 4,5 , suggesting that the repair of damaged bases have been divergently evolved.
Besides the fundamental DNA repair systems like BER, organisms thriving in extreme environments are thought to have developed efficient DNA repair systems, since harsh conditions such as high temperature, ionizing radiation, and acidic/basic pH promote DNA damage. Endonuclease Q (EndoQ) is an enzyme recently isolated from the hyperthermophilic archaeon, Pyrococcus furiosus 6 . This enzyme (PfuEndoQ) recognizes uracil, hypoxanthine, AP site and xanthine, and cleaves the phosphodiester bond at the 5′ side of the damaged base, leaving 5′ phosphate and 3′ hydroxyl groups. EndoQ is conserved in the Thermococcales (the genus Pyrococcus and Thermococcus) and some methanogenic archaea, but it does not belong to any of the previously described groups of DNA repair proteins. The homolog from Thermococcus kodakarensis (TkoEndoQ) also exhibited the same biochemical properties 6 . Furthermore, it is of note that a homolog is found in a few bacteria, but so far not in any eukaryotic organism.
Biochemical characterization of EndoQ showed that it is involved in the damaged DNA base repair system. However, there is no evidence for how EndoQ functions in this process. Although another hypoxanthine specific endonuclease, Endonuclease V (EndoV), is considered to function in removing deaminated adenine in P. furiosus, as well as in E. coli and other prokaryotes 7 , our in vitro analyses predicted that EndoQ and EndoV are not involved in the same repair pathway, but rather work independently 8 . Furthermore, PfuEndoQ is expected to act more effectively on hypoxanthine-containing DNA than EndoV from P. furiosus cells 8 .
To address the question of how EndoQ works in the repair of damaged bases in DNA in the Thermococcales, we have been searching for its interaction partners. Proliferating cell nuclear antigen (PCNA) plays an essential role in DNA transactions, including replication, repair, recombination, and cell cycle control 9 . PCNA is a ring-shaped trimeric complex. The central hole of the PCNA ring encircles double-stranded DNA to provide a scaffold to many proteins that acts on DNA, and it is called the clamp molecule. The β -clamp (identified as the β subunit of DNA polymerase III) in Bacteria has same functions as PCNA 10 . Proteins interacting with PCNA possess a consensus sequence motif called PIP (PCNA-interacting protein) box (Qxxhxxaa: x, any amino acid; h, hydrophobic residues; a, aromatic residues) 11,12 . A similar motif to PIP box is also conserved as a β -clamp binding sequence in Bacteria 10 . In this study, we found a PIP box-like motif at the C-terminal region of EndoQ. With respect to proteins that are involved in the early steps of the BER pathway from Archaea, previous studies showed that PCNA interacts with UDG and APE and enhances the glycosylase activity of UDG and the 3′ -5′ exonuclease activity of APE in P. furiosus 13,14 . It has also been shown that UDG from Sulfolobus solfataricus 15 and from Pyrobaculum aerophilum 16 interact with their PCNA. Hence, the PIP box-like motif in the EndoQ protein implies the possibility that PCNA is involved in EndoQ function. Here we report the physical and functional association of PCNA with EndoQ in vitro and propose a repair pathway in the Thermococcales.

Results
EndoQ homologs have a PIP-box motif at the C-terminus. An alignment of the amino acid sequence showed that most EndoQ homologs from Archaea, except for the Methanomicrobiales, have PIP box-like motifs at their C-terminal region (Fig. 1). Thus we assumed that EndoQ proteins would interact with PCNA through the motifs. It is also of note that the endoQ gene is present in Bacteria, such as Bacillus subtilis and Disulfovivrio sp., although EndoQ is mainly conserved in Archaea 6 . It is yet to be determined if these endoQ genes are expressed in the bacterial cells and have a function to cleave the DNA at the damaged site. However, the consensus sequences of the β -clamp binding motif 10 were found in the C-terminal region of the putative sequences of the bacterial EndoQ homologs. It will, therefore, be interesting to investigate if the physical and functional interactions between EndoQ and the clamp molecules from Bacteria, even though PCNA and β -clamp are thought to have evolved independently (see Supplementary Fig. S1).

Preparation of TkoEndoQ and TkoPCNA1 proteins. To investigate the interaction between EndoQ
and PCNA from T. kodakarensis, we prepared the mutant EndoQ with truncation of the PIP-box-like sequence and mutant PCNA with point mutations at the interface of the protomers for disruption of the ring structure. We deleted the amino acids from position 409 to 421 for TkoEndoQ, and designated it TkoEndoQ ΔPIP . It is known that the D143A/D147A mutant of PfuPCNA cannot form a stable ring structure in solution 17 , and therefore, the corresponding E143A/D147A mutations were made in TkoPCNA1. T. kodakarensis has two PCNAs, and PCNA1, but not PCNA2, is essential for cell viability 18,19 . Recombinant proteins expressed in E. coli, i.e., TkoEndoQ WT (MW: 48080.3), TkoEndoQ ΔPIP (MW: 46491.5), TkoPCNA1 WT (MW: 28239.4) and TkoPCNA1 E143A/D147A (MW: 28137.4) were purified to near homogeneity (Fig. 2). To confirm the disruption of the ring structure of TkoPCNA1 E143A/D147A in solution, purified TkoPCNAs were subjected to gel filtration analysis (see Supplementary  Fig. S2). Each protein eluted as a single peak, but the elution positions were different. The molecular weight estimation of TkoPCNA1 E143A/D147A was 37.3 k, while TkoPCNA1 WT was 99.1 k from the elution profiles. It is already known that PCNA molecules are eluted slightly earlier than the calculated molecular weights 17 . This result suggests that TkoPCNA1 E143A/D147A exists as a monomer in solution even at a high concentration (160 μ M). Maintenance of the structural conformation of TkoEndoQ after deletion of the C-terminal PIP region was supported by the comparison of the CD spectra from TkoEndoQ WT and TkoEndoQ ΔPIP . Two spectra that were almost superimposed were obtained from the two proteins (see Supplementary Fig. S3). Further experiments were performed using these purified proteins.

Physical interaction between TkoEndoQ and TkoPCNA1.
To investigate whether TkoEndoQ physically binds TkoPCNA1, surface plasmon resonance (SPR) analysis was performed using the purified proteins. As shown in Fig. 3, TkoEndoQ showed the positive sensorgram against the immobilized TkoPCNA1, and the responses increased in a protein concentration-dependent manner. The K D value for the interaction between the two proteins was 55 nM, which was calculated from the sensorgrams of seven different concentrations of TkoEndoQ. On the other hand, TkoEndoQ ΔPIP did not show any response with TkoPCNA1 even at a high concentration up to 800 nM. These results clearly indicated that the PIP-box located in the C-terminus of TkoEndoQ is essential for its interactions with TkoPCNA1. In this experiment, TkoPCNA1 was fixed on a sensorchip at less than 2 μ M, in which TkoPCNA1 WT exists as a monomer in solution as we showed previously 18 . Therefore, TkoEndoQ should binds to the monomeric form of TkoPCNA1 as observed in many other PCNA binding proteins.
Stimulation of endonuclease activity of TkoEndoQ by TkoPCNA1. To gain an infromation of how the physical interaction between EndoQ and PCNA contribute to DNA repair and the genome integrity, a cleavage assay using TkoEndoQ and TkoPCNA1 was conducted. Using an assay condition, in which TkoEndoQ WT exhibited 9% cleavage on one deoxyinosine (dI)-containing DNA, the rate of the cleavage was increased in a TkoPCNA1 concentration-dependent manner (Fig. 4a, lanes 2 to 5). When TkoPCNA1 WT was added at 180, 600 and 1800 nM (60, 200 and 600 nM; as a trimer) to the reaction, the rate of the cleavage was increased to 11%, 24% and 41%, respectively. Conversely, when TkoEndoQ ΔPIP or the monomeric mutant of TkoPCNA1 E143A/D147A was used, this stimulation was not detected. Notably, the TkoEndoQ ΔPIP mutant showed 6-7% cleavage either with or without TkoPCNA1 WT (Fig. 4a, lanes 6 to 9). The TkoPCNA1 E143A/D147A mutations did not affect the cleavage activity of the TkoEndoQ WT (Fig. 4a, lanes 11 to 1413). These results support our observation that the TkoPCNA1 stimulated endonuclease activity of EndoQ depending on the presence of the PIP box-like motif, and the ring structure of the PCNA is important for this function. The SPR experiment shown above supports that TkoEndoQ specifically binds to the monomeric form of TkoPCNA1. However, the ring structure TkoPCNA1 is necessary to stimulate the endonuclease activity of TkoEndoQ as shown here, although one EndoQ molecule on one PCNA

Interaction of EndoQ and PCNA is conserved in the Thermococcales. To confirm that the
EndoQ-PCNA interaction is conserved in the Thermococcales, purified PfuEndoQ and PfuPCNA were used for the interaction/stimulation analyses (see Supplementary Fig. S4). PfuPCNA clearly stimulated the endonuclease activity of PfuEndoQ on the dI-containing DNA by 6-7 fold (see Supplementary Fig. S5). Because the purified TkoEndoQ protein has more non-specific binding property to DNA and proteins as compared with PfuEndoQ, a higher salt concentration (0.4 M NaCl) was required for its manipulation in vitro. In addition, the endonuclease activity of TkoEndoQ showed more salt-resistance than PfuEndoQ. From these differences, the cleavage assay for PfuEndoQ was performed under reduced concentration of NaCl (0.18 M). To confirm that the EndoQ and PCNA were in the same complex in the cells, an immunoprecipitation experiment was performed using extracts from exponentially growing P. furiosus cells and antibodies raised against TkoEndoQ and PfuPCNA (a cross-reactivity of PfuEndoQ against the anti-TkoEndoQ antibody was confirmed before this IP experiment). PfuEndoQ and PfuPCNA co-precipitated with anti-TkoEndoQ or anti-PfuPCNA antibody, respectively (see Supplementary Fig. S6).

Discussion
We presented here that EndoQs from T. kodakarensis and P. furiosus interact with PCNA, and therefore, EndoQ may be involved in the replication-associated repair pathway at the replication fork, as proposed previously for P. furiosus UDGs 13 . It was also reported that APE of P. furiosus interacts with its cognate PCNA both in vivo and in vitro 14 . Furthermore, an efficient BER process, in which UDG and APE are bound simultaneously to the same PCNA trimer, and an efficient progress of the repair process including the sequential cleavages of the glycosyl bond of uracil and the diester bond has been proposed 14,20 . The multiprotein complex, including UNG2, APE1 (AP endonuclease), XRCC1, Polα , β , δ , ε , DNA ligase 1, and DNA dependent protein kinase, was also isolated from the nuclei of human cycling cells 21 . These reports indicate that archaea, possessing EndoQ may have more efficient repair systems for the damaged bases during replication fork progression.
It is now well known that many of the family B DNA polymerases from the hyperthermophilic archaea, including P. furiosus PolB, specifically recognize uracil bases in the template strand and stall complementary strand synthesis. This property of the archaeal DNA polymerases has been implicated as an intrinsic activity for the removal of uracil bases [22][23][24] . It is also possible that PolB and UDG bind to the same PCNA ring to switch at the uracil site. In addition to UDG and PolB, dUTPase, which probably contributes to precise DNA replication by preventing dUTP incorporation in the cells, is also found in P. furiosus 25 . Functional associations of PolB, UDG and dUTPase were proposed as a complex named 'uracilosome' for the efficient escape from uracil under hyperthermophilic conditions 26 , although the complex has not been isolated from any hyperthermophilic archaea. In addition to these molecules, we propose here that EndoQ is a member of the "uracilosome" in the Thermococcales and likely in other archaea harboring its homologs. Uracil is produced by the frequently occurring deamination of cytosine, especially at high temperatures, and therefore, it is possible that the hyperthermophilic archaea acquired the efficient prevention system to alleviate mutations by cytosine deamination.
We showed here that EndoQs also interact with PCNA likely through the PIP-box-like motif in their C-terminal regions. The predicted PIP-boxes are QRSITEFL in T. kodakarensis and QRTLLQYI in P. furiosus, respectively, and these are typical consensus sequences of the PIP-box. The location of these sequences at the very C-terminus is also typical among the PCNA-binding proteins. In the case of UDG and APE in P. furiosus, PCNA binding sites are not in the terminus, but the internal part of the proteins, and shorter versions of the PIP-box, AKTLF in UDG 13 and TIAGI 14 in APE, were proposed, as well as for DNA ligase, which also has a shorter version of the PIP box, QKSFF, in its internal site 27 . The apparent K D values, calculated from the SPR analysis were 55 nM for TkoEndoQ and TkoPCNA1. These results suggest that EndoQ has stronger affinity to PCNA as compared with UDG and APE. The apparent K D values for PfuUDG and PfuAPE with PfuPCNA are 220 nM and 1 μ M, respectively 13,14 . In consideration with a very close relationship between P. furiosus and T. kodakarensis, EndoQ may mainly work for removal of uracil and also other damaged bases in the Thermococcal cells.
The Thermococcales have one family B DNA polymerase (PolB) and one family D DNA polymerase (PolD), which are supposed to be replicative DNA polymerases 28 . However, genetic analyses showed that the polB gene can be disrupted in T. kodakarensis genome and it may mainly work for repair processes 29 . Our previous in vitro study showing that PolB prefers gap-filling type substrates to primer-extension type substrates, while the substrate preference of PolD is the opposite, supports this prediction 30 . We have also confirmed that the PolB of P. furiosus has strand displacement activity in vitro (Kimizu et al., unpublished result). Taken together with these results, strand displacement DNA synthesis by PolB, cleavage of the resultant flapped DNA by Fen1 endonuclease, and nick-sealing by DNA ligase will occur after incision by EndoQ, as in the case of the BER pathway. PCNA will have an important role to provide a scaffold for EndoQ, PolB, Fen1 and Lig to work on DNA efficiently for their sequential tasks (Fig. 5). Further analyses will elucidate this prediction of the damaged base repair process in the Thermococcales. It is also of evolutionary interest that the endoQ gene is not found in the hyperthermophilic archaeal subdomain of Crenarchaeota, which includes organisms such as Sulfolobus solfataricus, Sulfolobus islandicus, Aeropyrum pernix and Pyrodictium occultum. However, the gene is conserved in the methanogenic archaea, suggesting that this gene was likely acquired or invented in the archaeal subdomain Euryarchaeota, which includes the methanogens (hyperthermophilic, thermophilic and mesophilic), the halophiles, and the Thermococcales. The presence of the gene in some bacteria is not surprising, as the methanogens tend to grow in association with bacteria in many environments including the soil and mammalian guts, and this important gene can be acquired through horizontal gene transfer. We are currently investigating the function of the EndoQ homologs in both the mesophilic and hyperthermophilic methanogens to help shed more light on the evolution and distribution of this very fascinating DNA repair enzyme.
In conclusion, we presented here the physical and functional interactions between EndoQ and PCNA. EndoQ is probably acquired for the efficient repair of damaged bases in hyperthermophilic archaea and evolved in the archaeal and bacterial domains to form a repairsome with PCNA.