Recognition of Formamidopyrimidine by Escherichia coli and Mammalian Thymine Glycol Glycosylases

The activity of prokaryotic and mammalian thymine glycol (Tg) glycosylases including Escherichia coliendonuclease III (Endo III) and endonuclease VIII (Endo VIII) and mouse Endo III homologue (mNth1) for formamidopyrimidine (Fapy) has been investigated using defined oligonucleotide substrates. 2,6-Diamino-4-hydroxy-5-N-methylformamidopyrimidine, a methylated Fapy derived from guanine, was site specifically incorporated in the oligonucleotide. The substrates containing Fapy:N pairs (N = A, G, C, T) as well as a Tg:A pair, a physiological substrate of Endo III, Endo VIII, and mNth1, were treated by the enzymes and nicked products were quantified by gel electrophoresis. The activity of Endo III and Endo VIII for Fapy varied markedly depending on the paired base, being the highest with G (activity relative to Tg = 0.55 (Endo III) and 0.41 (Endo VIII)) and the lowest with C (0.05 (Endo III) and 0.06 (Endo VIII)). In contrast, mNth1 recognized all Fapy pairs equally well and the activity was comparable to Tg. The results obtained in the nicking assay were further substantiated by the analysis of the Schiff base intermediate using NaBH4trapping assays. These results indicate that Escherichia coli and mammalian Tg glycosylases have a potential activity to recognize Fapy. However, as demonstrated for Fapy:C pairs, their distinctive activities implicate unequal participation in the repair of Fapy lesions in cells.

DNA damage caused by exogenous and endogenous agents is a major threat to genetic integrity of cells, and when unrepaired, it results in lethal and/or mutagenic events of cells (1). The deleterious effects of DNA damage have been also implicated in carcinogenesis and aging (2,3). Among the diverse DNA lesions thus far identified, alterations of the base moiety by reactive oxygen species constitute one of the major classes of DNA damage. In cells, this class of damage is generally restored by the base excision repair (BER) 1 pathway, a highly conserved mechanism across species (4,5). The BER process for oxidized bases is initiated by DNA N-glycosylases that remove damaged bases from DNA. In Escherichia coli, DNA N-glycosylases responsible for this reaction are basically classified into two subgroups depending on the substrate, albeit some overlapping specificities. Oxidized pyrimidine bases such as thymine glycol (Tg) and 5-hydroxycytosine, etc., are excised from DNA by endonuclease III (Endo III) and endonuclease VIII (Endo VIII) (5, 6) encoded by the nth and nei genes, respectively (7)(8)(9). In contrast, oxidized purine bases such as 7,8-dihydro-8-oxoguanine (8-oxoG) and formamidopyrimidine (Fapy) are removed by formamidopyrimidine DNA glycosylase (Fpg) (10,11) encoded by the fpg/mutM gene (12). The nth nei double mutant of E. coli lacking Endo III and Endo VIII exhibits a mutator phenotype and a increased sensitivity to hydrogen peroxide and ionizing radiation (8,9). The mutations observed for the double mutant are mostly C 3 T transitions probably arising from impaired excision of oxidized cytosine lesions (13). The fpg mutant of E. coli deficient in Fpg shows a mutator phenotype though the sensitivity to ionizing radiation or hydrogen peroxide remains unchanged (14,15). The frequent mutations observed for the fpg mutant are G 3 T transversions due to impaired excision of 8-oxoG (15). Thus, the substrate specificity of the two subgroups of DNA N-glycosylases correlates fairly well with the phenotype of their mutants.
Recently, functional homologues of Endo III (Ntg1 and Ntg2 proteins (also called Scr1 and Scr2)) and Fpg (yOgg1 protein) have been identified from Saccharomyces cerevisiae, and the substrate specificities of the expressed proteins have been studied. The yOgg1 protein recognizes 8-oxoG and Fapy, showing a substrate specificity similar to Fpg (16,17). Ntg1 and Ntg2 proteins recognize a variety of oxidized pyrimidines, such as Tg and 5-hydroxycytosine, that are also substrates for Endo III (18 -23). Despite such similar substrate specificities between Endo III and its S. cerevisiae homologues, the latter enzymes also recognize Fapy derivatives (18 -20, 22) that have been reported not to be excised by Endo III (6) and its human homologue (hNTH1) (24). The additional substrate specificity of Ntg1 and Ntg2 for Fapy derivatives is rather unexpected since both proteins show significant amino acid sequence homology to Endo III (24% identity and 46% similarity (Ntg1), 25% identity and 51% similarity (Ntg2)). Moreover, the enzymes have the helix-hairpin-helix (HhH) motif and key amino acids (Lys-243 and Asp-262 (Ntg1), Lys-248 and Asp-267 (Ntg2)) possibly involved in DNA recognition and catalysis. The HhH motif and catalytic amino acids are highly conserved in Endo III homologues (25). Ntg2 (but not Ntg1) also possesses a conserved 4Fe-4S cluster near the C terminus.
The apparently different activities toward a Fapy substrate of Endo III and its S. cerevisiae homologues (Ntg1 and Ntg2) raised a question whether the activity is peculiar to Ntg1 and Ntg2 or has been somehow overlooked in previous studies on the Endo III homologues from other sources. In view of the question above, we have reinvestigated the activity of prokaryotic and mammalian thymine glycol glycosylases for Fapy using defined oligonucleotide substrates. We report here that thymine glycol glycosylases including Endo III, Endo VIII, and the mouse Endo III homologue (mNth1/mNthl1) have a potential activity to excise Fapy from DNA. However, the activity of Endo III and Endo VIII varies dramatically depending on the base opposite Fapy, whereas the activity of mNth1 is essentially independent of the paired base and is comparable to that for Tg.

EXPERIMENTAL PROCEDURES
Enzymes-E. coli DNA polymerase I Klenow fragment and T4 polynucleotide kinase were purchased from Life Technologies, Inc. and New England Biolabs, respectively. Endo III, Endo VIII, and Fpg proteins were overexpressed in E. coli cells harboring plasmids containing the nth, nei, or fpg gene (gifts from S. S. Wallace and Z. Hatahet) and purified as described (8,26). Expression and purification of the mouse Endo III homologue (mNth1/mNthl1) has been reported previously (27). Human 7,8-dihydro-8-oxoguanine glycosylase (hOGG1/hMMH, type 1a isoform) was generously supplied by Nishimura (28). The repair enzymes used in this study were apparently homogenous in SDS-PAGE analysis.
Oligonucleotides-Oligonucleotides used in this study are listed in Table I. The oligonucleotides except 25FP and 19TG were synthesized by the phosphoramidite method and purified by reversed phase HPLC. 25FP containing a single 2,6-diamino-4-hydroxy-5-N-methylformamidopyrimidine (Fapy) was prepared by the reported procedure (29). Briefly, 15PRM was 5Ј-end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase and purified by a C18 Sep-Pak cartridge (Waters). 15PRM was annealed to the template 30COM-C and extended by polymerase I Klenow fragment in the presence of dCTP, dTTP, and 7-methyl-2Јdeoxyguanosine 5Ј-triphosphate (Sigma). The 7-methylguanine residue incorporated opposite C in 30COM-C was converted to Fapy by the alkali treatment at pH 11.4. 25FP was separated from 30COM-C by denaturing polyacrylamide gel electrophoresis (PAGE) and recovered from the gel. Finally, 25FP was annealed to the appropriate complementary strands (25COM-A, -G, -C, -T). 19TG containing a single cis-Tg was synthesized by KMnO 4 oxidation of a 19-mer oligonucleotide that contained thymine at the position of Tg (27,30). The crude 19TG was purified by a C18 Sep-Pak cartridge followed by reversed phase HPLC. 19TG was 5Ј-end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase, purified, and annealed to the complementary strand 19COM-A. The duplex substrate 25OX/25COM-C containing 8-oxoG was con-structed in a similar manner.
Activity Assays for Substrates Containing Fapy and Tg-Duplex substrates (5 nM), 25FP/25COM-N (N ϭ A, G, C, T) and 19TG/ 19COM-A, were incubated with Endo III, Endo VIII, or mNth1 in appropriate buffers (10 l) at 37°C for 5 min. The composition of the buffer for Endo III and Endo VIII was 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 100 mM NaCl (buffer A), and that for mNth1 was 20 mM Hepes-KOH (pH 8.0), 50 mM KCl, 0.25 mM EDTA, and 0.25 mM dithiothreitol (buffer B). The amount of the repair enzymes used in the experiments was indicated in the table or figures. To ensure the linear response of the product formation, appropriate amounts of the enzymes and incubation time for the assay were determined by varying these parameters in preliminary kinetic experiments (data not shown). These reaction conditions were used for the activity assays. After incubation, the sample was mixed with gel loading buffer (0.05% xylene cyanol, 0.05% bromphenol blue, 20 mM EDTA, and 98% formamide), heated at 50°C for 5 min, and separated by 16% denaturing PAGE. The gel was autoradiographed at Ϫ80°C overnight. Alternatively, the radioactivity of the separated bands was analyzed by Fuji BAS 2000.
NaBH 4 Trapping of Schiff Base Intermediates-An aqueous solution of 500 mM NaBH 4 (1 l) was added to trapping reaction buffers (8 l) containing substrates (25FP/25COM-N (N ϭ A, G, C, T) or 19TG/ 19COM-A, final concentration 5 nM). Immediately after, the solution was mixed with repair enzymes (Endo III (5 ng), Endo VIII (20 ng), mNth1 (5 ng), all in 1 l) and incubated at 37°C for 5 min. The final compositions of the trapping reaction buffers were the same as those for the activity assay buffers A and B described above, except that NaCl (Endo III and Endo VIII) and KCl (mNth1) were omitted and 50 mM NaBH 4 was present. After incubation, the solution was mixed with SDS-loading buffer (100 mM Tris, 8% SDS, 24% (v/v) glycerol, 4% 2-mercaptoethanol, 0.02% SERVA Blue G) and heat-denatured. The sample was separated by 10% SDS-PAGE. Autoradiography and quantitation of the radioactivity were performed as described above.

RESULTS
Activity of Endo III, Endo VIII, and mNth1 for Fapy Paired with C-To examine the incision activity for Fapy, Endo III, Endo VIII, and mNth1 were incubated with a duplex substrate 25FP/25COM-C containing a Fapy:C pair. Cytosine was chosen as a paired base since Fapy used in this study was originally derived from guanine and a Fapy:C pair could be a naturally occurring base pair. PAGE analysis of the reaction products revealed that treatments with Endo III (Fig. 1A) and Endo VIII ( Fig. 1B) resulted in weak bands corresponding to ␤and ␦-elimination products, respectively, which migrated somewhat slower (␤) and faster (␦) than the size marker (M). Although the amount of incision products increased with that of incubated Endo III and Endo VIII (Fig. 1, A and B, lanes 1-3), only minor proportions of the substrate were converted to the incision product even with an excess of the enzymes. These results suggest that both Endo III and Endo VIII recognize Fapy paired with C but the activity is extremely low. In contrast, incubation with mNth1 resulted in strong bands corresponding to the ␤-elimination product (Fig. 1C), showing highly efficient recognition of Fapy by mNth1.
To compare the activity of Endo III, Endo VIII, and mNth1 for Fapy on the biologically relevant basis, two substrates containing Fapy (25FP/25COM-C) and Tg (19TG/19COM-A) were treated with each enzyme under the same conditions. Prior to the quantitative activity assay, the reaction conditions (the amount of the enzyme and incubation time) were properly adjusted based on the preliminary experiments. Under these conditions, the formation of the product was essentially within a linear (or not saturating) range, thereby allowing direct comparison the activities for Fapy and Tg (This was also the case for the data in Fig. 3). Table II summarizes the amount of   TABLE I List of oligonucleotides used in this study products (i.e., percentage of nicked substrate) for both substrates. Comparison of the yield of products for Fapy and Tg showed that the activity of Endo III and Endo VIII for the Fapy:C pair was approximately 20-fold lower than that for the Tg:A pair, an intrinsic substrate of these enzymes. However, mNth1 recognized both Fapy and Tg equally well.
Paired Base Effects on the Recognition of Fapy-It is known that the activity of certain BER enzymes varies depending on the base opposite a lesion. For example, 8-oxoG paired with C and A is excised by Fpg and hOGG1 with different efficiencies (Ref. 29, and references cited therein). The repair activity of Endo III for 5-hydroxypyrimidines (31) and a ring fragmentation product of thymine C 5 -hydrate (32) also varies depending on the paired base. In view of these facts, substrates containing all four possible Fapy pairs (25FP/25COM-N (N ϭ A, G, C, T)) were constructed and tested for Endo III, Endo VIII, and mNth1. Fig. 2 shows the results of product analysis by PAGE for Fapy pairs together with Tg. The yield of the ␤-elimination product by Endo III (Fig. 2A, lanes 3-6) and the ␦-elimination product by Endo VIII (Fig. 2B, lanes 3-6) varied remarkably depending on the paired base, whereas that by mNth1 (␤elimination product) did not (Fig. 2C, lanes 3-6). Based on the repeated assays as shown in Fig. 2, the incision activity for Fapy:N pairs (N ϭ A, G, C, T) were quantified and compared (Fig. 3). The results for Tg obtained under the same reaction conditions were also included in Fig. 3. The activity of Endo III for Fapy decreased in the following order of paired bases: G (activity relative to Tg ϭ 0.55) ϭ A (0.55) Ͼ T (0.3) Ͼ C (0.05). The variation of the activity was 11-fold between the most (G, A) and least (C) preferred pairs. The corresponding order for Endo VIII was G (0.41) Ͼ A (0.14) Ͼϭ T (0.09) Ͼϭ C (0.06), with the activity variation of 7-fold between the most (G) and least (C) preferred pairs. For mNth1, all Fapy pairs were equally good substrates and the activity was comparable to Tg, i.e., Fapy:G (activity relative to Tg ϭ 1.   tional BER enzymes that release a damaged base (DNA Nglycosylase activity) and incise a phosphodiester bond (AP lyase activity) form a covalent Schiff base intermediate with DNA during the reaction (25). Formation of the intermediate has been substantiated by the NaBH 4 or NaCNBH 3 trapping assay for several Endo III homologues including Endo III itself (33), Ntg1/Ntg2 (21,22,34), and hNTH1 (35,36). Thus, formation of a Schiff base intermediate between the substrate containing Fapy and thymine glycol glycosylases (Endo III, Endo VIII, mNth1) provides additional evidence for their activity to Fapy. Duplex substrates containing Fapy (25FP/25COM-N, where N ϭ A, G, C, T) and Tg (19TG/19COM-A) were incubated with Endo III, Endo VIII, and mNth1 in the presence of NaBH 4 and the trapped Schiff intermediate was analyzed by SDS-PAGE (Fig. 4). With all tested enzymes, bands migrating more slowly than free substrates were observed, showing the Schiff base formation. Based on these assays, the amount of the trapped intermediate was quantified and the percentage of the substrate cross-linked to the enzymes was compared. With Endo III (Fig. 5A) (Fig. 5B). The paired base effect on the Schiff base formation between the Fapy substrates and Endo III/Endo VIII correlates fairly well with that on the incision activity of these enzymes shown in Fig. 3 (A and B). With mNth1, the paired base-dependent variation of the Schiff base formation was much less obvious than with Endo III and Endo VIII. Moreover, the amounts of trapped intermediates for the Fapy substrates were virtually comparable to that for Tg (G (0.99), A (0.88), T (0.79), C (0.57)). These results were also consistent with the activity assay of mNth1 shown in Fig. 3C.
Activity of Endo III, Endo VIII, and mNth1 for 8-OxoG-The repair enzymes (Endo III, Endo VIII, and mNth1) used in the present study were overexpressed and purified from E. coli cells carrying the wild type fpg gene that codes for Fpg, the major repair enzyme for both Fapy and 8-oxoG in E. coli cells. To demonstrate that the activity of Endo III, Endo VIII, and mNth1 for Fapy was not due to the contaminating Fpg protein, the enzymes were incubated with 25OX/25COM-C containing an 8-oxoG:C pair and products were analyzed by PAGE. None of these enzymes exhibited a detectable activity to 8-oxoG (Fig.   6, lanes [3][4][5], although the same substrate was efficiently incised by Fpg (Fig. 6, lane 6) and hOGG1 (Fig. 6, lane 7), resulting in ␦and ␤-elimination products, respectively. These results indicate that the activity for Fapy is not due to contaminating Fpg but is indeed associated with Endo III, Endo VIII, and mNth1. This conclusion was further supported by the mode of strand cleavage characteristic to Endo III (␤-elimination), Endo VIII (␦-elimination), and mNth1 (␤-elimination) (Fig. 2) as well as the presence of a single species of the cross-linked intermediate for each enzyme (Fig. 4). DISCUSSION Biological Implications of the Fapy Repair Activity of Endo III, Endo VIII, and mNth1-In the present study, it has been shown that thymine glycol glycosylases from E. coli (Endo III and Endo VIII) and mouse (mNth1) have a potential activity to remove the Fapy lesion derived from guanine. Combining the present results and those reported for the S. cerevisiae (Ntg1 and Ntg2) (18 -20, 22), the repair activity of Fapy is potentially conserved among the thymine glycol glycosylases across species. However, the activity of the E. coli enzymes, but not the mouse enzyme, was dramatically influenced by the base opposite this lesion. Endo III and Endo VIII efficiently recognized Fapy when it paired with purines (particularly G in the case of Endo VIII) and very poorly when paired with C (Fig. 3). The very weak activity of Endo III for a Fapy:C pair relative to Tg may explain the reason why Endo III released Tg but not G-derived Fapy from ␥-irradiated DNA substrates in the previous study (6) (for A-derived Fapy, see the next section). According to the reported data (19,22), Ntg1 and Ntg2 recognize Fapy:C pairs in ␥-irradiated DNA and methylated/alkalitreated poly(dG-dC) as efficiently as Tg. These results suggest that Ntg1 and Ntg2 resemble the mammalian homologue (mNth1) rather than the E. coli enzymes (Endo III and Endo VIII) with respect to the activity for Fapy:C pairs. The substrate specificity of Endo III homologues of Schizosaccharomyces pombe (Nth-Spo) (37) and human (hNTH1) (24) has been also examined previously using ␥-irradiated or H 2 O 2 /Fetreated DNA. Nth-Spo and hNTH1 released several pyrimidine damages, but neither enzyme released Fapy derivatives from the damaged DNA. Therefore, as far as eukaryotic enzymes are concerned, yeast Ntg1 and Ntg2 but not Nth-Spo recognize Fapy:C pairs, and the mouse (mNth1) but not human (hNTH1) enzyme recognize Fapy:C pairs. The negligible activity of hNTH1 and Nth-Spo toward Fapy:C pairs was rather surprising since the amino acid sequence of mNth1 shows 81% identity to hNTH1 in 300 overlapping residues and 58.2% identity to Nth-Spo in 226 residues (27). Additionally, all activity assays of mNth1 (this study), hNTH1 (24), and Nth-Spo (37) were performed with recombinant histidine-tagged proteins. An apparent difference in the assay conditions was the enzyme substrate, which contained a single Fapy lesion and multiple types of lesions in the present (mNth1) and previous (hNTH1 and Nth-Spo) studies, respectively. Thus, the influences of co-existing damage in the substrate together with other possible factors need to be assessed to solve the apparent activity difference toward Fapy lesions.
It has been shown previously by transfection studies that Fapy derived from G is lethal but not mutagenic (38,39), suggesting that G-derived Fapy does not form mispairs with A, G, and T. Thus, as far as G damage is concerned, a Fapy:C pair is the predominant form that repair enzymes encounter in cells. In view of the very weak activity of Endo III and Endo VIII for a Fapy:C pair (20-fold lower than Tg), the activity of these enzymes for a Fapy:C pair may not be physiologically important in E. coli cells relative to Fpg. In contrast, the situation will be different in eukaryotic cells whose Endo III homologues (at least mNth1, Ntg1, Ntg2) have a strong activity for Fapy:C pairs. Very recently, knockout mice deficient in OGG1 protein, a functional homologue of Fpg, have been generated (40). Although the activity for 8-oxoG was completely diminished in the tissue extracts from ogg1 Ϫ/Ϫ null mice, that for Fapy (paired with C) was only partially reduced (ϳ1/3 of the ogg1 ϩ/ϩ mice), indicating that in vivo repair activity of Fapy resided on not only OGG1 but also another protein with a redundant activity. This observation is consistent with the activity of mNth1 for Fapy:C pairs demonstrated in this study. Accordingly, unlike in prokaryotic cells, Fapy lesions are likely repaired by the action of both OGG1 and NTH1 (or Ntg1/Ntg2) proteins in eukaryotic cells.
Paired Base Effects on the Incision Activity of Fapy-The present study has shown that Endo III and Endo VIII, but not mNth1, recognize Fapy in a paired base-dependent manner (Fig. 3). Endo III recognized Fapy:A and Fapy:G pairs most efficiently and a Fapy:C pair least efficiently (Fig. 3A). These paired base effects are quite different from those observed for Fpg and hOGG1, which recognize all Fapy base pairs with comparable efficiencies (29). Endo III excises a variety of pyrimidine lesions, and a common feature of these substrates is the loss of aromatic character due to saturation of the C5-C6 double bond (e.g. pyrimidine glycols and photohydrates), ring fragmentation (e.g. urea and ␤-ureidoisobutyric acid residues), and ring contraction (e.g. hydantoin derivatives). The structural alterations of these substrates occur along the C4-C5-C6-N1 bonds of the pyrimidine ring. When the structure of Fapy is superimposed on the pyrimidine lesions recognized by Endo III, the ruptured imidazole ring of Fapy overlaps the C4-C5-C6-N1 region. An example for Tg is shown in Fig. 7A, where the deoxyribose moieties of Fapy and Tg were primarily superimposed to reflect the DNA structure. Thus, the ruptured imidazole ring mimics a defective pyrimidine. It is possible that Endo III potentially senses this feature of Fapy for the initial stage of damage recognition. Concerning the paired base effects, a Fapy (derived from G):C pair is likely to form hydrogen bonds similar to a G:C pair, albeit weaker, since the functional groups involved in hydrogen bonding remain intact in Fapy (Fig. 7, B and C). This notion is also supported by the theoretical calculation of the stabilization energy of the Fapy:C pair (41). Since Endo III uses a flip-out mechanism (42), the stabilization of the Fapy:C pair by hydrogen bonds may retard the extrusion of the Fapy residue into the active site pocket that is suggested to accommodate a flipped out base. Considering that A-derived Fapy was not excised by Endo III from ␥-irradiated DNA (6), similar stabilization may occur for a base pair between A-derived Fapy and thymine. Conversely, if a bulky purine is placed opposite Fapy, the steric crash between Fapy and the purine in a helix extrudes the Fapy residue into a partially extrahelical position, a geometry like in a pretransition state of base flipping. A presumable disposition of a Fapy:G pair is shown in Fig. 7 (D and E). Accordingly, the lack of stabilizing hydrogen bonds and partial extrusion of the Fapy residue facilitate excision of Fapy paired with purines.
Endo VIII showed a paired base specificity similar to Endo III though activity to a Fapy:A pair was lower than a Fapy:G pair (Fig. 3B). Endo VIII is a functional homologue of Endo III, but its amino acid sequence shows no homology to Endo III (8,9). The overlapping substrate specificity including Fapy pairs between Endo III and Endo VIII implies that Endo VIII employs a damage sensing mechanism similar to Endo III, though the final products formed by the AP lyase activity of Endo III and Endo VIII are different (␤-and ␦-elimination products, respectively). In contrast to Endo III and Endo VIII, mNth1 exhibited a consistently high activity for all Fapy base pairs (Fig. 3C). The distinctive activities of Endo III and mNth1 for Fapy pairs are rather surprising since the fundamental architecture and catalytic mechanism are predicted to be conserved between Endo III and mNth1 based on the amino acid sequence homology and the characteristic HhH and 4Fe-4S motifs (27), as well as the formation of common Schiff base intermediates (Fig. 4C). The present results suggest that, upon binding to DNA, mNth1 may induce extra structural perturbations in DNA such as kinks or bends over Endo III to extrude the Fapy residue from a stable Fapy:C pair. Currently we do not know the exact origin of such extra perturbations conferred by mNth1. The answer to this question must await for determination of three-dimensional structures of DNA-protein binary complexes of mNth1 and Endo III.