Enzymology of base excision repair in the hyperthermophilic archaeon Pyrobaculum aerophilum.

DNA of all living organisms is constantly modified by exogenous and endogenous reagents. The mutagenic threat of modifications such as methylation, oxidation, and hydrolytic deamination of DNA bases is counteracted by base excision repair (BER). This process is initiated by the action of one of several DNA glycosylases, which removes the aberrant base and thus initiates a cascade of events that involves scission of the DNA backbone, removal of the baseless sugar-phosphate residue, filling in of the resulting single nucleotide gap, and ligation of the remaining nick. We were interested to find out how the BER process functions in hyperthermophiles, organisms growing at temperatures around 100 degrees C, where the rates of these spontaneous reactions are greatly accelerated. In our previous studies, we could show that the crenarchaeon Pyrobaculum aerophilum has at least three uracil-DNA glycosylases, Pa-UDGa, Pa-UDGb, and Pa-MIG, that can initiate the BER process by catalyzing the removal of uracil residues arising through the spontaneous deamination of cytosines. We now report that the genome of P. aerophilum encodes also the remaining functions necessary for BER and show that a system consisting of four P. aerophilum encoded enzymes, Pa-UDGb, AP endonuclease IV, DNA polymerase B2, and DNA ligase, can efficiently repair a G.U mispair in an oligonucleotide substrate to a G.C pair. Interestingly, the efficiency of the in vitro repair reaction was stimulated by Pa-PCNA1, the processivity clamp of DNA polymerases.


INTRODUCTION
Most hyperthermophiles, organisms living at temperatures of around 100°C, belong to Archaea (1), the closest prokaryotic relatives of eukaryotes (2). Two key pathways of DNA metabolism, transcription and replication, are highly conserved among Archaea and eukaryotes (3,4), and we were interested to learn whether these similarities extended also to the third domain of DNA metabolism, namely, DNA repair. We chose to study the hyperthermophilic crenarchaeon Pyrobaculum aerophilum, the genomic sequence of which has recently become available (5).
DNA repair processes can be classified into three major categories: damage reversal (DR), recombination repair (RR) and excision repair; the latter can be further subdivided into three biochemical pathways: nucleotide excision repair (NER), mismatch repair (MMR) and base excision repair (BER). Sequence similarity searches for DNA repair genes in P.
aerophilum revealed that the organism apparently lacks key representatives of several of these pathways. Thus, we found only a single representative of the DR class, O 6alkylguanine DNA alkyltransferase, which is present in most Archaea (6), and two copies of the RecA/RAD51 recombinase homologue RadA. It is not known whether the latter genes encode functional polypeptides, as the genome of this organism appears to carry little evidence of recombination events (5). This implies that RR may be rather infrequent, as shown for Sulfolobus acidocaldarius (7), a close relative of P. aerophilum. Only two putative homologues of mammalian NER enzymes, the DNA helicases XPB and XPD (5), could be found, and it is presently unclear whether these proteins are involved in DNA repair or transcription, as components of the transcription factor TFIIH (8). Like other Archaea, the genome of P. aerophilum carries no homologues of the MMR genes mutS and mutL.

Base Excision Repair in Pyrobaculum aerophilum
Sartori and Jiricny sites can be accomplished by one of two distinct pathways: short-patch BER, the preferred mechanism that results in the replacement of a single nucleotide residue, and long-patch BER, where the repair tracts are 2-6 nucleotides long (16). In the short-patch pathway, DNA polymerase β (pol-β) extends the upstream fragment by a single nucleotide, and concomitantly removes the dRP moiety by β-elimination (17). The DNA ligase III/XRCC1 complex (18,19) then seals the remaining nick. In long-patch BER, the upstream primer is thought to be extended by pol-β (or pol-δ) by 2-6 nucleotides, which results in the displacement of the dRP residue as part of a 'flap' oligonucleotide (20). This short overhang is then excised by the flap-endonuclease (FEN1), and the resulting nick is sealed by DNA ligase I. Proliferating cell nuclear antigen (PCNA) may also be involved in this process, as it was found to enhance the pol-β-dependent long patch BER by stimulating the activity of FEN1 and to interact with DNA ligase I (21,22). The long-patch pathway may be important for the repair of reduced or oxidized AP-sites, in which the modified dRP residues are resistant to pol-β-catalyzed β-elimination reactions (21).
We set out to explore the BER system of P. aerophilum, the genome of which carries, in addition to the above-mentioned DNA glycosylase genes, also genes encoding orthologues of downstream-acting BER proteins: one putative AP endonuclease, three putative DNA polymerases of the B family, one putative DNA ligase and PCNA1, which was already shown to interact with FEN1, UDGa and PolB3 from P. aerophilum (23). We expressed these proteins in E. coli, and analyzed their ability to catalyze the repair of a G•U mispair in a P. aerophilum Extracts and Purified Proteins-P. aerophilum cultures were grown in the laboratory of Jeffrey Miller (University of California, Los Angeles, CA) and cell-free extracts were prepared by Mahmud Shivji as described previously (24). The cell-free extract was supplemented with 1 mM PMSF and 1 x Complete", EDTA-free (Boehringer Mannheim) protease-inhibitor cocktail and dialyzed overnight at 4°C against 2 liters of buffer containing 25mM sodium phosphate pH 8.0, 50mM NaCl, 10 % glycerol, 0.5mM EDTA and 2mM DTT. The protein concentration was estimated to be 4 mg/ml as determined by the method of Bradford (25). Aliquots were stored at -80°C. The recombinant P.
aerophilum uracil-DNA glycosylases Pa-UDGa and Pa-UDGb were expressed and purified as described previously (10,11 Base excision repair assay using P. aerophilum whole cell-free extract (WCE)-The fluorescently-labeled substrates, containing either a single-nucleotide gap (1nt-gap) or a G•U mismatch at a defined position, were prepared as follows: 1-nt gap, the labeled 23-mer-F and 36-mer oligos were annealed with the complementary 60-mer G oligo; G•U, the labeled 60-mer U-F was annealed with the 60-mer G oligo according to the protocol described previously (10). The oligonucleotide sequences are listed in Table I  Incubation times and temperatures varied as indicated in the figures. After precipitation, the DNA pellets were dissolved in 90% formamide supplemented with 50mM NaBH 4 in order to prevent spontaneous hydrolytic cleavage of the labile AP-sites prior to analyzing the samples as described (10).
DNA polymerase assays-The primer-extension ability of the DNA polymerases was compared using a substrate prepared by annealing the 5'-labeled 23-mer-F primer with the 60-mer G template oligonucleotide (listed in Table I DNA Ligase Assay-The radioactively-labeled substrate, illustrated in Figure 4B, was prepared as follows: 1 µg of a 24-mer oligodeoxythymidilate (dT) was phosphorylated at its The samples were then heated for 5 min at 95 °C and electrophoresed through a 10 % denaturing gel. The gel was dried and the reaction products were detected by autoradiography.
Fluorescently-labeled substrates containing a ligatable nick were used to assay nickjoining activity of the recombinant Pa-DNA-Ligase and of the DNA ligase(s) present in P.

Base Excision Repair in Pyrobaculum aerophilum
Sartori and Jiricny aerophilum WCE. The labeled 24-mer C-F (or 24-mer T-F) and the unlabeled 36-mer oligos were annealed with the complementary 60-mer G oligo (see Table I The repair of AP-sites by purified recombinant P. aerophilum proteins was monitored by using substrates containing either a normal or a reduced AP-site as illustrated in Figure 5B. For normal AP-sites, the G•AP* 60-mer substrate was prepared as described above. The  (10).

Base Excision Repair in Pyrobaculum aerophilum
Sartori and Jiricny

Base Excision Repair activities supported by P. aerophilum extracts-In our earlier
studies, we demonstrated that whole cell-free extracts (WCE) of P. aerophilum could catalyze the removal of uracil from a G•U mismatch and the subsequent cleavage of the resulting AP-site (10). As this results in the generation of a single nucleotide gap through the removal of the abasic sugar-phosphate, we wanted to test the capacity of P. aerophilum WCE to repair this latter lesion ( Figure 1A). Upon addition of deoxyribonucleotide triphosphates and magnesium, we observed preferential incorporation of a single nucleotide into the one-  12). As the WCE of P. aerophilum was apparently able to carry out all the steps of BER, we set out to characterize the enzymes involved.
Identification of BER genes in P. aerophilum-We showed earlier that P. aerophilum possessed at least three different DNA glycosylases able to remove uracil from a G•U mismatch in vitro (9)(10)(11). The action of a monofunctional uracil-DNA glycosylase (e.g. and glutamate (E) side chains to form a trinuclear zinc cluster (29). As the 275-amino acid ORF identified contained all the latter metal-binding residues, we named its protein product Pa-EndoIV (Fig. 1B).
The removal of uracil from a G•U mismatch and the cleavage of the resulting AP-site on its 5'-side gives rise to a single-strand break where the upstream fragment is terminated Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny with a 3'-hydroxyl group and the downstream fragment has a deoxyribose-phosphate (dRP) group at its 5'-terminus. In mammalian short-patch BER, DNA polymerase-β (pol-β) removes the dRP by its associated AP-lyase activity and simultaneously extends the 3'terminus by one nucleotide (17,30). The dRP-lyase step, which is rate-limiting in BER (31), was shown to be carried out by the 8 kDa N-teminal domain of pol-β, a member of the Xfamily of DNA polymerases. The sequence of pol-β was used to search for putative homologues in the P. aerophilum sequence database, albeit without success. Indeed, only family-B DNA polymerases, thought to be involved primarily in DNA replication, have been identified in crenarchaeal organisms to date (32). As shown in Figure 1C, the P. aerophilum genome encodes three family-B DNA polymerases, one each of the B1, B2, and B3 subfamilies, denoted as Pa-PolB1, Pa-PolB2, and Pa-PolB3, respectively (5, 33).
Interestingly, while Pa-PolB1 and Pa-PolB3 contain all known signatures of replicative DNA polymerases, the Pa-PolB2 sequence is substantially shorter and lacks several conserved motifs, one of which encodes a putative proofreading exonuclease (Fig. 1C). P. aerophilum enzyme is also ATP-rather than NAD-dependent (see (35,36) for reviews).
As shown in Figure 1D  Pa-UDGa and Pa-UDGb are both monofunctional DNA glycosylases that generate AP-sites in the 60-mer oligonucleotide duplex G•U, but are unable to convert these into strand breaks, because they lack an associated DNA lyase activity (Fig. 2C, lanes 2 and 3).
The 60-mer substrate could be cleaved at these sites by β-δ-elimination with hot alkali (Fig.   2C, lanes 4 and 5), which gave rise to a labeled 23-mer fragment terminated with a 3'phosphate group. Pa-EndoIV also cleaved the AP-sites created by both these enzymes.
However, because the 60-mer substrate was cleaved on the 5' side of the abasic residue by In addition to their sensitivity to aphidicolin, the three major mammalian polymerases involved in DNA replication (α, δ and ε) do not incorporate dideoxyribonucleotide monophosphates (ddNMPs) into DNA. Pfu pol, a representative member of the archaeal Bfamily of DNA polymerases, behaved similarly (Fig. 3C, lanes 8,9). In contrast, the human pol-β is not inhibited by aphidicolin (Fig. 3C, lane 4), but incorporates ddNMPs into DNA Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny quite efficiently (20,41); the primer extension reaction is thus inhibited by these substances (Fig. 3C, lane5). In all these assays, Pa-PolB2 resembled pol-β. It remained unaffected by 2mM aphidicolin (lane 12), but was inhibited by dideoxynucleoside triphosphates at a ddCTP/dCTP ratio of 10, albeit not as efficiently as pol-β (cf. Lanes 5 and 13).
Next, we wanted to test whether DNA synthesis by Pa-PolB2 is affected by the recombinant P. aerophilum polymerase processivity factor Pa-PCNA1. As shown in Fig. 3D, Finally, we wanted to test whether these archaeal polymerases displayed binding affinity towards a double-stranded 60-mer oligo containing a single nucleotide gap. This substrate, which is generated during the initial steps of BER by the cleavage of an AP-site and the subsequent removal of the baseless sugar-phosphate (dRP), was bound by Pa-UDGb (Fig. 3E, lane 6). Pa-UDGb interacted weakly also with the G•C homoduplex DNA (Fig. 3E, lane 2) and more strongly with AP-sites (10). Pa-PolB2 also bound to the gapped-substrate in an EMSA assay (Fig. 3E lanes 8 and 9), albeit with an affinity lower than Pa-UDGb (lane 6). Interestingly, the mobility of the retarded band in the polyacrylamide matrix was slightly lower in the presence of Pa-PCNA1 (lanes 10 and 11), suggesting that the DNA, Pa-PolB2 and Pa-PCNA1 formed a ternary complex. This finding was rather unexpected, in view of the fact that Pa-PolB2 lacks a PCNA binding motif and that its polymerase activity was not stimulated by the sliding clamp, in contrast to Pa-PolB3, which interacts and is stimulated by

Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny
Pa-PCNA, yet could be seen to form no stable complex on this DNA substrate (lanes 12, 13). However, in the light of data presented below, Pa-PCNA1 may have a role in BER that is distinct from its role as a processivity factor in DNA synthesis. Procedures for details). After endonucleolytic cleavage, the RAP substrate is resistant to βelimination, and its "repair" can be accomplished only by displacement of the entire downstream 36-mer (see above), or by excision mediated by a 5→3 exonuclease (FEN1), followed by gap-filling (pol-δ) and ligation (long-patch repair). In contrast, the unreduced Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny dRP moiety can be removed also by a β-elimination reaction, as described for the mammalian system (30). As shown in Figure 5B, the AP-site was sensitive to this reaction even in the absence of added enzymes, especially at higher temperatures (upper panel, lanes 1

Catalytic Properties of Pa
and 2), and this reaction was accelerated in the presence of Pa-PolB2 and Pa-DNA-Ligase [lane 3, see also ref. (43)]. As expected, the RAP-site was stable under these experimental conditions (lower panel, lanes 1-3). In the presence of Pa-EndoIV (lanes 4), both sites were efficiently cleaved. Note that the product band generated by Pa-EndoIV migrated slightly faster through the polyacrylamide gel than the product resulting from the spontaneous βelimination reaction. This is because the former product has a 3'-OH terminus, whereas the latter has an α,β-unsaturated aldehyde at its 3'-terminus (41).
In the presence of Pa-PolB2 and dCTP, the 3'-termini of both substrates were extended by a single nucleotide (lanes 5), but only the AP substrate was repaired to the 60- However, when this experiment was carried out with Pa-PolB3, an identical result was obtained (data not shown), implying that the dRP moiety could be removed under our reaction conditions in the presence of either polymerase. We therefore tested both polymerases in the subsequent experiment, in which we wanted to test whether the recombinant P. aerophilum BER proteins described above could catalyze the repair of a G•U mismatch in the 50-mer/60-mer oligonucleotide substrate used above (Fig. 1A). As shown in that Pa-PolB2 is more likely to be involved in DNA repair, rather than in DNA replication.

DISCUSSION
In the experiments described above we showed that the genome of P. aerophilum (5) encodes a full set of BER proteins. In addition to several DNA glycosylases (9)(10)(11), ORFs encoding one AP endonuclease, three DNA polymerases of the B family and one DNA ligase could be identified by sequence homology searches. In addition, the genome encodes also two homologues of the polymerase processivity factor PCNA (23). We expressed Pa- Taking the repair of uracil arising through spontaneous hydrolytic deamination of cytosine as an example, earlier work showed that the in vitro BER process could in this case be initiated by any one of at least three uracil-DNA glycosylases: Pa-UDGa (10), Pa-UDGb (11) or Pa-MIG (9). However, we would argue that the enzyme most likely to initiate this repair process in vivo would be Pa-UDGb. As Pa-UDGa could be shown to interact with Pa-PCNA1, this glycosylase might be more likely to act in the repair of uracil residues incorporated into the newly-synthesized strand in the form of dUMP during DNA replication (45). Pa-MIG is a DNA glycosylase that can address both G•U and G•T mispairs, the latter thought to arise through the deamination of 5-methylcytosine residues in DNA (9). Its Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny participation in short patch BER of G•U mispairs cannot be ruled out, but its low activity and abundance in cell-free extracts of P. aerophilum (10) might imply that this enzyme plays only a secondary role to the more active and more abundant Pa-UDGb in vivo. It was for this reason that we selected the latter glycosylase as the initiating enzyme in our BER reconstitution experiments.
The AP-site arising through the removal of uracil by Pa-UDGb has to be incised at its 5'-end by an AP endonuclease. Pa-EndoIV appears to be the only enzyme of this kind encoded in the genome of P. aerophilum, and it is thus likely that this protein indeed participates in BER in this organism. Pa-EndoIV is required also for the processing of APsites generated by the action of glycosylases/lyases, such as enzymes of the Nth family (46) that cleave the DNA backbone by β-elimination concomitantly with base removal. The α,βunsaturated aldehyde generated by these enzymes blocks the 3-terminus of the upstream fragment such that the repair polymerase cannot initiate the gap-filling reaction until this moiety is cleaved off. Pa-EndoIV can indeed catalyze the removal of these residues in vitro (data not shown).
Following the action of Pa-UDGb and Pa-EndoIV, the 3'-terminus of the incised strand has a free hydroxyl group, and can thus serve as a primer for the repair polymerase in a gapfilling reaction. However, the end of the extended primer cannot be ligated to the 5'-terminus of the incised strand until the baseless sugar-phosphate (dRP) that is blocking this site is removed. This reaction can take place spontaneously to some extent, especially at elevated temperatures (Fig. 2C), but it is unlikely that any organism would rely on this process. In mammals, the β-elimination reaction that removes the dRP residue is catalyzed by the Nterminus of pol-β (47). We did not directly test the dRP lyase activity of Pa-PolB2 and Pa-Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny PolB3 in an in vitro assay, because the heat-labile dRP residues are readily cleaved at the high incubation temperature required for their optimal activity; we could therefore not distinguish between non-enzymatic and enzymatic cleavage. Nevertheless, both enzymes yielded ligatable substrates in our in vitro assays (Fig. 5C), which attests to their ability to catalyze the removal of this blocking lesion. Moreover, the AP-site containing substrate was readily cleaved by Pa-PolB2 and/or Pa-DNA-Ligase by β-elimination (Fig. 5B, lane 3; data not shown), reflecting an intrinsic AP-lyase activity in these enzymes. It is therefore possible that the 5'-dRP group is removed by these enzymes in a BER reaction. A similar BER pathway was proposed to take place in mitochondria by Bogenhagen et al., where both DNA polymerase γ and mtDNA ligase appear to possess a dRPlyase activity (43).
Based on our results, it is difficult to implicate either polymerase in P. aerophilum BER in vivo. Both Pa-PolB2 and Pa-PolB3 carried out the gap-filling reactions with similar efficiencies, and the yields of the repaired 60-mer products were comparable in assays using either enzyme. However, the enzymatic properties of the former enzyme resembled pol-β in many aspects: Pa-PolB2 contains several basic amino acid residues at its N-terminus that might impart it with a dRPase activity, it has limited processivity that is not stimulated by Pa-PCNA1 (Fig. 3D), and it is thus less prone to carry out strand displacement (Fig. 5C, cf. Lanes 10 and 11). In contrast, the primary structure and biochemial properties of crenarchaeal B3-DNA polymerases resemble more enzymes of the B1 family (40,(48)(49)(50); the B1 and B3 polymerases are therefore most likely involved in DNA replication in Crenarchaea (34).
Thus, by analogy with the mammalian BER systems, where pol-β appears to be the major BER polymerase, but where the involvement of pol-δ in long patch BER could not be excluded (45), Pa-PolB2 would appear to be a better candidate for carrying out the gap-Base Excision Repair in Pyrobaculum aerophilum Sartori and Jiricny filling function in short-patch BER in P. aerophilum, even though Pa-PolB3 may also participate in this process.
The role of Pa-PCNA1 in the P. aerophilum BER process in vitro is puzzling. This homotrimeric sliding clamp, the primary function of which is to increase the processivity of replicative DNA polymerases [reviewed in (51)], failed to stimulate Pa-PolB2. On the contrary, its presence in the reaction appeared to restrict the activity of Pa-PolB2 in the gapfilling reactions to the addition of a single nucleotide, while suppressing its strand displacement activity (Fig. 5). Yet, although Pa-PolB2 has no consensus PCNA binding motif, it appeared to interact with the sliding clamp in the EMSA assay on a substrate containing a single nucleotide gap, and the presence of Pa-PCNA1 in the reaction improved efficiency of the ligation reaction, especially at mispaired termini (Fig. 4C). It is possible that the association of the polymerase and the ligase with Pa-PCNA1 leads to the stabilization of DNA termini, which might otherwise fray at the elevated temperatures employed in these assays. Pa-PCNA1 might thus be fulfilling a role of a molecular matchmaker or scaffold protein similar to that postulated for XRCC1 in mammalian systems (19). It is interesting to note in this regard that the P. aerophilum genome does not appear to encode an XRCC1 homologue.
In conclusion, we have identified homologues of the mammalian BER proteins in the hyperthermophilic crenarchaeon Pyrobaculum aerophilum and could show that these proteins can carry out efficient G•U → G•C repair in an oligonucleotide substrate. This is to our knowledge the first report describing the reconstitution of the archaeal BER process from