Differential specificity of human and Escherichia coli endonuclease III and VIII homologues for oxidative base lesions.

In human cells, oxidative pyrimidine lesions are restored by the base excision repair pathway initiated by homologues of Endo III (hNTH1) and Endo VIII (hNEIL1 and hNEIL2). In this study we have quantitatively analyzed and compared their activity toward nine oxidative base lesions and an apurinic/apyrimidinic (AP) site using defined oligonucleotide substrates. hNTH1 and hNEIL1 but not hNEIL2 excised the two stereoisomers of thymine glycol (5R-Tg and 5S-Tg), but their isomer specificity was markedly different: the relative activity for 5R-Tg:5S-Tg was 13:1 for hNTH1 and 1.5:1 for hNEIL1. This was also the case for their Escherichia coli homologues: the relative activity for 5R-Tg:5S-Tg was 1:2.5 for Endo III and 3.2:1 for Endo VIII. Among other tested lesions for hNTH1, an AP site was a significantly better substrate than urea, 5-hydroxyuracil (hoU), and guanine-derived formamidopyrimidine (mFapyG), whereas for hNEIL1 these base lesions and an AP site were comparable substrates. In contrast, hNEIL2 recognized an AP site exclusively, and the activity for hoU and mFapyG was marginal. hNEIL1, hNEIL2, and Endo VIII but not hNTH1 and Endo III formed cross-links to oxanine, suggesting conservation of the -fold of the active site of the Endo VIII homologues. The profiles of the excision of the Tg isomers with HeLa and E. coli cell extracts closely resembled those of hNTH1 and Endo III, confirming their major contribution to the repair of Tg isomers in cells. However, detailed analysis of the cellular activity suggests that hNEIL1 has a significant role in the repair of 5S-Tg in human cells.

DNA carrying vital genetic information of cells constantly suffers from spontaneous deamination and depurination, alkylation, and oxidation (1)(2)(3). These reactions lead to modifications of the DNA backbone and bases, with the latter predominating. The resulting aberrant bases are potentially genotoxic because of the loss or alteration of base pairing information (4), and hence need to be restored by the cellular repair system (2,3,5). The major repair mechanism for such damage is the base excision repair (BER) 1 pathway (6), which is conserved from bacteria to humans. In the first step of BER, DNA glycosylases with distinct damage specificities detect the aberrant base in the vast sea of normal bases and remove it from the DNA backbone, leaving an apurinic/apyrimidinic (AP) site. The resulting AP site is further processed and repaired by the subsequent action of AP endonuclease (Endo), DNA polymerase, and DNA ligase through the short patch or long patch BER pathway.
The initial search for DNA glycosylases involved in the repair of oxidatively damaged bases in Escherichia coli identified Endo III, Endo VIII, and formamidopyrimidine-DNA glycosylase (7,8). The principal substrates of Endo III and Endo VIII are oxidative pyrimidine lesions. They exhibit redundant damage specificity and catalyze the hydrolysis of the N-glycosidic bond (N-glycosylase activity) and the subsequent incision of an AP site by AP lyase activity via ␤-elimination (Endo III) or ␤,␦-elimination (Endo VIII). The E. coli mutants deficient in both Endo III and Endo VIII are strong spontaneous mutators (9,10) and hypersensitive to the agents that generate reactive oxygen species such as ionizing radiation and hydrogen peroxide (9,11). The principal substrates of formamidopyrimidine-DNA glycosylase are oxidative purine lesions, and it exhibits N-glycosylase and ␤,␦-AP lyase activities. The E. coli mutants deficient in formamidopyrimidine-DNA glycosylase are not sensitive to ionizing radiation but exhibit a mild spontaneous mutator phenotype (12,13). Interestingly, while showing distinct substrate specificity, Endo VIII and formamidopyrimidine-DNA glycosylase belong to the same structural family, the Endo VIII/formamidopyrimidine-DNA glycosylase superfamily (14,15).
The mammalian Endo III homologue (NTH1) and a functional homologue of formamidopyrimidine-DNA glycosylase (OGG1) have been identified previously, and their functions in BER have been assessed using purified proteins (16 -22), knockout mice (23)(24)(25)(26)(27)(28), and x-ray crystallographic analysis (29,30). It has recently been shown that mammals have Endo VIII homologues (31,32); they are designated NEIL1, NEIL2, and NEIL3 (after Nei-like), demonstrating the conserved organization of DNA glycosylases involved in the repair of oxidatively damaged pyrimidine and purine lesions. Studies into the repair function of the mammalian Endo VIII homologues reveal that like Endo III and Endo VIII, NTH1 and NEIL1/NEIL2 exhibit, albeit not fully, redundant damage specificity and primarily recognize oxidative pyrimidine lesions (14,(31)(32)(33)(34)(35)(36)(37). However, their activities toward oxidized base lesions have been assessed using different substrates (oligonucleotides with different sequence contexts or calf thymus DNA) and assay methods (nicking assays of DNA and release assays of damaged bases), making the quantitative comparison of activity data rather difficult. In light of this fact, we measured and quantitatively compared the activity of human NTH1, NEIL1, and NEIL2 (hNTH1, hNEIL1, and hNEIL2) and that of their E. coli homologues (Endo III and Endo VIII) using common oligonucleotide substrates. We report here that hNTH1, hNEIL1, and hNEIL2 exhibit significantly different activities toward the stereoisomers of thymine glycol (Tg) and other oxidative base lesions, and that this is also the case for Endo III and Endo VIII. These results, together with those obtained from cell extracts, indicate that base lesions generated by reactive oxygen species are removed from DNA at distinct rates in cells, and hence that their genotoxic effects can be differentially attenuated in keeping with their repair kinetics.
DNA Glycosylases-The purification of Endo III, Endo VIII, and hNTH1 were reported previously (19,43). The native form of hNEIL1 and hNEIL2 proteins were purified as follows. Briefly, based on published sequences (31, 32), hNEIL1 and hNEIL2 cDNAs were amplified CATCGATAGCATCCTXCCTTCTCTC C a G, A, and C indicate the base opposite the damage (X) in double-stranded substrates. The length of the complementary strand was the same as the lesion strand for each substrate.
from the human liver cDNA library (Nippon Gene) using the polymerase chain reaction. The amplified DNA fragments were ligated into the NdeI/XhoI site of pET-22b(ϩ) (Novagen). The recombinant plasmids for hNEIL1 and hNEIL2 were designated phNEIL1 and phNEIL2, respectively. E. coli BL21-CodonPlus (DE3)-RIL (Stratagene) was transformed with phNEIL1 or phNEIL2. The original sequence of the inserts was confirmed by sequencing phNEIL1 and phNEIL2 isolated from the host cell. E. coli BL21-CodonPlus (DE3)-RIL harboring phNEIL1 or phNEIL2 was grown in LB media containing chloramphenicol (50 g/ ml) and ampicillin (50 g/ml) at 37°C until A 600 reached 0.6. After the addition of isopropyl-␤-D-thiogalactopyranoside (final concentration, 1 mM), the cell culture was continued at 30°C for 3 h. The following procedures were performed at 4°C or on ice. Harvested cells were disrupted by sonication. The cell lysate was centrifuged, and proteins in the supernatant were collected by ammonium sulfate precipitation (60% saturation). The hNEIL1 protein was purified by SP Sepharose CL-4B, MonoS, and Superdex 75 XK16/50 columns (all from Amersham Biosciences). The hNEIL2 protein was purified by SP Sepharose CL-4B (two cycles) and Superdex 75 XK16/50 (two cycles) columns. The pooled fraction containing hNEIL1 or hNEIL2 was dialyzed against 20 mM Hepes-KOH (pH 7.5), 150 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, and 50% glycerol, and stored at Ϫ80°C (hNEIL1) or Ϫ20°C (hNEIL2). The protein concentration was determined with the BCA protein assay kit (Pierce) using BSA as a standard.
Cell Extracts-Cell extracts were prepared on ice or at 4°C. The HeLa cell extract was prepared from confluent cells. The cell pellet was suspended in three volumes of 50 mM Tris-HCl (pH 7.5), 3 mM EDTA, 5 mM Mg(CH 3 COO) 2 , 3 mM 2-mercaptoethanol, 300 mM KCl, 1 mM phe-FIG. 2. PAGE analysis of the reaction products formed by incubation of oligonucleotide substrates containing 5R-Tg and 5S-Tg with Endo III and Endo VIII homologues. 30TG5R/A and 30TG5S/A (both 5 nM, the Tg strand was 5Ј-end 32 P-labeled) containing 5R-Tg and 5S-Tg, respectively, were incubated with different amounts of the indicated enzymes at 37°C for 30 min. Products were separated by 16% denaturing PAGE. The autoradiograms of the gels are shown. The amounts of enzyme used in the reactions were as follows (from left to right lanes for both 5R-Tg and 5S-Tg): 0, 0.25, 0.5, 1, and 2 ng for hNTH1; 0, 1, 2, 5, and 10 ng for hNEIL1; 0, 2.5, 5, 10, and 20 ng for hNEIL2; 0, 0.025, 0.05, 0.1, and 0.2 ng for Endo III; and 0, 0.5, 1, 2, and 4 ng for Endo VIII.
FIG. 3. Differential activities of Endo III and Endo VIII homologues for the 5R-Tg and 5S-Tg isomers. The percentage of nicked products was determined by the PAGE analysis as shown in Fig. 2, and is plotted against the amount of enzyme used for the assay (average of two experiments). Symbols: q, 5R-Tg; OE, 5S-Tg. The enzyme used is indicated above each panel.
nylmethanesulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml pepstatin. The cells were disrupted with 30 strokes of a tight-fitting Dounce homogenizer and centrifuged at 100,000 ϫ g for 30 min. The proteins in the supernatant were collected by ammonium sulfate precipitation (60% saturation). The proteins were resuspended in 20 mM Tris-HCl (pH 7.5), 20 mM NaCl, 1 mM dithiothreitol, and 1 mM EDTA, dialyzed against the same buffer, and used for activity assays. The E. coli cell extract was prepared from exponentially growing E. coli AB1157. The cell pellet was suspended in 10 volumes of 10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM phenylmethanesulfonyl fluoride, and 0.1 mg/ml lysozyme, and allowed to stand for 30 min. The cells were disrupted by sonication. The cell lysate was centrifuged, and the supernatant was used for activity assays. The protein concentration was determined with the BCA protein assay kit.
Activity Assays-The substrates (all 50 fmol, Table I) were incubated with DNA glycosylases in appropriate buffer (10 l) at 37°C for 30 min. The amount of proteins was varied depending on the activity of enzymes, and is indicated in the figures. The buffer used for assays with Endo III, Endo VIII, and hNEIL1 was 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 100 mM NaCl, and 0.1 mg/ml BSA. The buffer for hNTH1 was 20 mM Hepes-KOH (pH 8.0), 0.25 mM dithiothreitol, 0.25 mM EDTA, 50 mM KCl, and 0.1 mg/ml BSA, and that for hNEIL2 was 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 50 mM NaCl, and 0.1 mg/ml BSA. The reaction was terminated by the addition of gel loading buffer (0.05% xylene cyanol, 0.05% bromphenol blue, 20 mM EDTA, and 98% formamide). After heating at 70°C for 5 min, products were separated by 16% denaturing PAGE and the radioactivity in the gel was analyzed on a phosphorimaging analyzer BAS2000 (Fuji). For assays with the cell extracts, the substrates (30TG5R and 30TG5S, both 50 fmol) were incubated with the E. coli or HeLa cell extracts (0.5-4 g) in 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 100 mM NaCl, and 0.1 mg/ml BSA (10 l) at 37°C for 30 (HeLa extracts) or 10 min (E. coli extracts). Products were analyzed as described for purified DNA glycosylases.
NaBH 4 Trapping Reactions-NaBH 4 trapping reactions were per-formed under conditions similar to those for the activity assay using 19AP/A as a substrate and Endo VIII, hNEIL1, and hNEIL2 (all 200 ng) as enzymes. In the reaction buffer for Endo VIII and hNEIL1, 100 mM NaCl was replaced by 50 mM NaCl plus 50 mM NaBH 4 , and for hNEIL2, 50 mM NaCl was replaced by 50 mM NaBH 4 . The sample was incubated at 37°C for 30 min. After incubation, the sample was mixed with SDS loading buffer (100 mM Tris, 8% SDS, 24% (v/v) glycerol, 4% 2-mercaptoethanol, and 0.02% SERVA Blue G), heated, and separated by 10% SDS-PAGE. Autoradiography and quantitation of the radioactivity were performed as described above.
Cross-link Reactions with 25OXA-The duplex of 25OXA/C (10 fmol) was incubated with hNEIL1 or hNEIL2 (200 ng) in the activity assay buffer described above (10 l) at 37°C for up to 1 h. The sample was mixed with SDS-loading buffer, heated, and separated by 10% SDS-PAGE.
FIG. 4. Differential activities of Endo III and Endo VIII homologues for oxidative base lesions and AP sites. The substrates (all 5 nM; the strand with a lesion was 5Ј-end 32 P-labeled) containing an AP site (19AP/A), hoU (25HOU/G), hoC (25HOC/G), and mFa-pyG (34FP/C) were incubated with different amounts of the indicated enzymes at 37°C for 30 min, and products were quantified by PAGE analysis. The percentage of nicked products is plotted against the amount of enzyme used for the assay (average of two experiments). Symbols: f, AP site; q, hoU; ࡗ, hoC; OE, mFapyG. The enzyme used is indicated above each panel.
trapped species, which migrated slower than that formed with Endo VIII (29.6 kDa). This result further confirmed that the hNEIL1 and hNEIL2 preparations were free from Endo VIII and Endo III (23.4 kDa).
Activity of Purified Enzymes for Tg Isomers-hNTH1, hNEIL1, hNEIL2, and their E. coli homologues (Endo III and Endo VIII) were incubated with 30TG5R/A and 30TG5S/A containing 5R-Tg and 5S-Tg, respectively, and the products were analyzed by denaturing PAGE (Fig. 2). Fig. 3 shows plots of the amount of nicked products (average of two experiments) against that of enzyme used for the assay, which was varied depending on the activity. hNTH1 and hNEIL1 recognized both 5R-Tg and 5S-Tg isomers, but their specificity for the isomers differed significantly. hNTH1 excised 5R-Tg much more preferentially compared with 5S-Tg (Fig. 3A), whereas hNEIL1 excised 5R-Tg only slightly better than 5S-Tg (Fig. 3B). When 5R-Tg and 5S-Tg in 30TG5R and 30TG5S, respectively, were converted to urea residues by mild alkaline treatment, hNTH1 and hNEIL1 exhibited the same activity toward urea residues derived from the two Tg isomers (data not shown). These results confirmed that the differential specificities of hNTH1 and hNEIL1 toward the Tg isomers originate from the distinct configurations at C-5 and C-6 of the pyrimidine ring. The activity of hNEIL2 for the Tg isomers was below the detection limit (Fig. 3C). From the slope of the essentially linear part of the plot in Fig. 3, the activity for the two Tg isomers was calculated as [nicked substrate]/[enzyme]/min, where square brackets denote the molar concentration, and is summarized in Table II. According to the data in Table II, the specificity ratio toward 5R-Tg versus 5S-Tg is 13:1 for hNTH1 (i.e. 1.4 ϫ 10 Ϫ2 min Ϫ1 versus 1.1 ϫ 10 Ϫ3 min Ϫ1 ) and 1.5:1 for hNEIL1 (i.e. 5.1 ϫ 10 Ϫ3 min Ϫ1 versus 3.3 ϫ 10 Ϫ3 min Ϫ1 ), demonstrating marked differences in the isomer specificity between hNTH1 and hNEIL1. It can also be deduced from the activity data (Table II) that for 5R-Tg, hNTH1 exhibits a higher turnover rate than hNEIL1 (hNTH1:hNEIL1 ϭ 2.7:1), whereas for 5S-Tg, hNEIL1 exhibits a higher turnover rate than hNTH1 (hNTH1: hNEIL1 ϭ 1:3). These results are in contrast to those reported recently for mouse NTH1 (mNTH1) and NEIL1 (mNEIL1) (36), although the isomer specificities of mNTH1 and mNEIL1 for the two Tg isomers are similar to those observed for hNTH1 and hNEIL1 in this study. The ratio of the turnover rates for 5R-Tg (estimated from the reported data) is mNTH1: mNEIL1 ϭ 1:29, and that for 5S-Tg is 1:570, indicating that mNEIL1 is an extremely efficient enzyme as compared with mNTH1 for both Tg isomers, which was not the case for hNEIL1 (see above).
The activity of E. coli Endo III and Endo VIII for the two Tg isomers was determined in a similar manner (Figs. 2 and 3). Like hNTH1 and hNEIL1, Endo III and Endo VIII exhibited  Table II were standardized to that for an AP site, and are plotted against the lesions. The enzyme used is indicated above each panel.
FIG. 6. Differential excision capacities of HeLa and E. coli cell extracts for the 5R-Tg and 5S-Tg isomers. 30TG5R/A and 30TG5S/A (both 5 nM) containing 5R-Tg and 5S-Tg, respectively, were incubated with different amounts of extracts from HeLa and E. coli cells at 37°C for 30 (HeLa) or 10 min (E. coli). Products formed with the HeLa (panel A) and E. coli (panel B) extracts were separated by 16% denaturing PAGE. The bands of ␤-elimination and ␤,␦-elimination products are indicated with ␤ and ␦, respectively. The amounts of HeLa and E. coli cell extracts used in the reaction were 0, 0.5, 1, 2, and 4 g (from left to right lanes for both 5R-Tg and 5S-Tg). The percentage of nicked products was determined by the PAGE analysis as described above, and is plotted against the amount of the cell extract used for the assay. Panels C and D show the results with the HeLa and E. coli cell extracts, respectively. Symbols: q, 5R-Tg; OE, 5S-Tg. significantly different specificity for the two isomers. However, Endo III excised 5S-Tg better than 5R-Tg (Fig. 3D), and the specificity ratio toward 5R-Tg versus 5S-Tg was 1:2.5 (i.e. 1.9 ϫ 10 Ϫ1 min Ϫ1 versus 4.7 ϫ 10 Ϫ1 min Ϫ1 , Table II). Thus, despite being homologues, hNTH1 and Endo III have an opposite preference for the Tg isomers. Endo VIII preferentially excised 5R-Tg as compared with 5S-Tg (Fig. 3E), and the specificity ratio toward 5R-Tg versus 5S-Tg was 3.2:1 (i.e. 3.0 ϫ 10 Ϫ2 min Ϫ1 versus 9.3 ϫ 10 Ϫ3 min Ϫ1 , Table II). Although hNEIL1 has a slight preference of 5R-Tg over 5S-Tg, the difference in the isomer specificity of Endo VIII (3.2-fold) is greater than that of hNEIL1 (1.5-fold). It is likely from the activity data (Table II) that Endo III exhibits higher turnover rates than Endo VIII for both 5R-Tg (6.3-fold) and 5S-Tg (51-fold).
Activity of Purified Enzymes for Other Oxidative Base Lesions-hNTH1, hNEIL1, hNEIL2, Endo III, and Endo VIII were incubated with the substrates containing hoU (25HOU/ G), hoC (25HOC/G), mFapyG (34FP/C), an AP site (19AP/A), a urea residue (19UR/A), fU (25FU/A), hmU (25HMU/A), and 8-oxoG (25OG/C), and products were analyzed by denaturing PAGE (data not shown). Fig. 4 shows typical plots of the amount of nicked product (average of two experiments) against that of enzyme used for the assay. From these plots, the activity of the enzymes toward the individual lesions was calculated as described for the Tg isomers (Table II). The activity for the selected lesions standardized to that for an AP site is plotted in Fig. 5, together with that for the Tg isomers. hNEIL1 exhibited comparable or better activities for 5R-Tg, 5S-Tg, urea, hoU, and mFapyG than for an AP site ( Fig. 5B and Table II). hoC was a relatively poor substrate of hNEIL1, with the activity for this lesion being one-ninth of that for 5R-Tg, the best substrate of hNEIL1. fU, hmU, and 8-oxoG were worse substrates than hoC (Table II), and hence the activities for these lesions were essentially negligible. Consistent with a previous report (33), hNEIL2 showed significant activity for AP sites (Fig. 5C), which was comparable with that of hNEIL1 (hNEIL1 ϭ 3.5 ϫ 10 Ϫ3 min Ϫ1 and hNEIL2 ϭ 2.7 ϫ 10 Ϫ3 min Ϫ1 , Table II). The incision product of an AP site by hNEIL2 was a mixture of ␤and ␤,␦-elimination products, whereas that by hNEIL1 was exclusively a ␤,␦-elimination product (data not shown). Although hNEIL2 recognized hoU and mFapyG, the activities for these lesions were marginal, at only 1/50 -1/100 of that for an AP site (Fig. 5C and Table II). We checked the effect of trimethylamine N-oxide that was reported to enhance the activity of hNEIL2 (33), but no significant enhancement was observed under the present assay conditions. The damage specificity of hNTH1 was notably different from that of hNEIL1 (Fig. 5A). Unlike hNEIL1, hNTH1 showed a preference toward an AP site over the base lesions. The relative activity for the tested lesions (shown in parentheses) was AP (1.0) Ͼ 5R-Tg (0.61) Ͼ urea (0.48) Ͼ mFapyG (0.30) Ͼ hoU (0.17) Ͼ hoC (0.074) Ͼ 5S-Tg (0.048).
With respect to the preference of base lesions relative to an AP site, Endo VIII exhibited specificity similar to that of its human homologue hNEIL1 (Fig. 5E), and as was the case for Endo III and hNTH1 (Fig. 5D). In addition, hoC was a consistently poor substrate not only for the human enzymes (hNTH1, hNEIL1, and hNEIL2) but also for the E. coli enzymes (Endo III and Endo VIII). A notable difference between the activity of human (hNTH1 and hNEIL1) and E. coli homologues (Endo III and Endo VIII) was the activity for mFapyG, which was a fair to good substrate for hNTH1 and hNEIL1 but not for Endo III and Endo VIII. This result is consistent with our previous observations for mouse NTH1, Endo III, and Endo VIII (19).
Activity of Cell Extracts for Tg Isomers-The substrates containing 5R-Tg (30TG5R/A) and 5S-Tg (30TG5S/A) were incubated with the extracts from HeLa and E. coli cells, and products were analyzed by denaturing PAGE (Figs. 6, A and B). Fig.  6, C and D, are plots of the amount of nicked products against that of the cell extract used for the assay. The HeLa cell extract excised 5R-Tg much more preferentially compared with 5S-Tg. The profiles of the excision of the two Tg isomers with the HeLa cell extract (Fig. 6C) closely resembled that with hNTH1 (Fig.  3A). In contrast, the E. coli cell extract preferentially excised 5S-Tg as compared with 5R-Tg. The profiles of the excision of the two Tg isomers with the E. coli cell extract (Fig. 6D) were similar to that with Endo III (Fig. 3D). The "Discussion" provides more detailed analysis of the cellular activity for the Tg isomers.
Cross-link Formation with Oxanine-hNEIL1 and hNEIL2 were incubated with 25OXA/C containing Oxa at 37°C for up to 1 h, and products were analyzed by SDS-PAGE (Fig. 7). Although hNEIL1 and hNEIL2 did not excise Oxa from DNA at appreciable rates (data not shown), both enzymes were crosslinked to Oxa, resulting in slow-migrating bands in the SDS-PAGE analysis. Endo VIII but not Endo III and hNTH1 was similarly cross-linked with Oxa under these conditions (45) (and data not shown), implying a common architecture of the active site pocket in human and E. coli Endo VIII homologues (hNEIL1, hNEIL2, and Endo VIII).  7. Cross-link formation of hNEIL1 and hNEIL2 with Oxa. 25OXA/C (the Oxa strand was 5Ј-end 32 P-labeled) was incubated with hNEIL1 or hNEIL2 at 37°C for 0, 10, 30, 45, and 60 min. The resulting cross-link complexes containing 25OXA and hNEIL1 or hNEIL2 were separated from free 25OXA by 10% SDS-PAGE.

DISCUSSION
In this study we have shown that hNTH1, hNEIL1, and hNEIL2 exhibit quantitatively different specificities toward oxidatively damaged bases and AP sites. hNTH1 and hNEIL1 recognized both oxidatively damaged bases and AP sites, showing a redundant spectrum of damage recognition. However, the specificity toward the individual lesions differs significantly between hNTH1 and hNEIL1 (Fig. 5, A and B). This was also the case for their E. coli homologues (Endo III and Endo VIII, Fig. 5, D and E). In contrast, hNEIL2 recognized AP sites exclusively, and the excision efficiency for the oxidatively damaged bases (hoU and mFapyG) was low (Fig. 5C). The present results, together with those reported for mammalian NTH1 (16 -19), NEIL1 (14,(31)(32)(33)(34)(35)(36)(37), and SMUG1 (43,44), show an elaborate backup system of mammalian DNA glycosylases that work in the first step of BER for oxidatively damaged DNA. For example, at least two or three enzymes can participate in the repair of Tg and urea residues (NTH1 and NEIL1), hoU (SMUG1, NTH1, and NEIL1), and mFapyG and its unmethylated analogue (OGG1, NTH1, and NEIL1). This may partly explain the lack of overt phenotypes of NTH1-knockout mice (23,24). The repair function of hNEIL2 in BER needs to be further assessed because it excises hoU and mFapyG with only limited efficiency (Fig. 5C). In addition, hNEIL1 and hNEIL2 excise oxidative base lesions in single-stranded and bubble DNA structures, suggesting their repair role during DNA replication and/or transcription (34,37).
Interestingly, the two Tg isomers were excised from DNA with differential efficiencies by both HeLa and E. coli cell extracts, and the specificity for the isomers was opposite between HeLa and E. coli cells (Fig. 6, C and D). It is known that ionizing radiation generates the two Tg isomers at comparable rates (46). Thus, the present results indicate that the 5R-Tg isomer is preferentially removed from chromosomal DNA in irradiated human cells, whereas the 5S-Tg isomer is preferentially removed in irradiated E. coli cells. The 5R-Tg and 5S-Tg isomers exert a similar destabilizing effect on duplex DNA (39) and equally constitute strong blocks to DNA synthesis catalyzed by polymerase ␣ (47). However, with polymerase , translesion synthesis past 5R-Tg is more efficient than that past 5S-Tg (47). Thus, 5S-Tg is a more persistent as well as intense blocking lesion than 5R-Tg because of slow removal and translesion synthesis (at least for polymerase ), and possibly constitutes a burden for mammalian cells. It remains to be seen whether the two isomers have distinct effects on other biological processes such as DNA replication catalyzed by other replicative and TLS polymerases or transcription, and also whether the species-specific excision of a particular Tg isomer results in any distinct biological consequences.
The kinetic profiles of the excision of the Tg isomers with the HeLa (Fig. 6C) and E. coli (Fig. 6D) cell extracts closely resembled those of hNTH1 (Fig. 3A) and Endo III (Fig. 3D), respectively, suggesting their major contribution to the cellular activity to Tg. This is also consistent with the observation that the residual activity for 5R-Tg (prepared by chemical oxidation of T with KMnO 4 or OsO 4 ) in cells deficient in NTH1 (NTH1knockout mouse) or Endo III (E. coli nth mutant) is 5-10% of the wild type cells (9,23,48) (and data not shown). Given that the residual activity for 5R-Tg is attributable to NEIL1 and Endo VIII in mammalian and E. coli cells, respectively, the contribution of the individual enzymes to the excision of 5S-Tg in cells (P 5S ) can be estimated using their activity ratio for the 5S-Tg and 5R-Tg isomers (F 5S/5R ) and observed cellular activity for the 5R-Tg isomer (P 5R ). The estimated contribution of hNTH1 to the excision of 5S-Tg (P 5S ) is 52-70% and that of hNEIL1 is 30 -48% (Table III), indicating fairly active involve-ment of hNEIL1 in the repair of 5S-Tg in human cells. This is mainly because of the higher turnover rate of hNEIL1 for 5S-Tg (3-fold) than that of hNTH1. Similar calculations for the E. coli enzymes (P 5S in Table III) show that Endo III contributes exclusively to the repair of 5S-Tg (99%) relative to Endo VIII (1%) in E. coli cells. The distinct contributions of hNEIL1 and Endo VIII to the repair of the 5S-Tg isomer agree semiquantitatively with the presence of the ␤,␦-elimination product (the hallmark of hNEIL1) with the HeLa cell extract and its absence (the hallmark of Endo VIII) with the E. coli cell extract (lanes for 5S-Tg in Fig. 6, A and B).
We have previously shown that Oxa, a major guanine lesion produced by the reaction with nitric oxide or nitrous acid (49,50), forms cross-links with DNA-binding proteins such as histone, HMG protein, and DNA glycosylases (45). The reaction with histone and HMG proteins is very slow, occurring on a time scale of days, but that with DNA glycosylases is very rapid, occurring in less than an hour, and possibly involves the direct interaction of Oxa in DNA with the active site residue of DNA glycosylases (tentatively assigned to Lys or Arg). Interestingly, despite sharing activities for oxidative pyrimidine lesions, Endo VIII but not Endo III and hNTH1 are crosslinked to Oxa (45). Accordingly, Oxa is a simple but useful lesion to probe the architecture of the active site of DNA glycosylases. The present study has shown that hNEIL1 and hNEIL2 form cross-links to Oxa at rates comparable with that of Endo VIII (Fig. 7), although they exhibit no appreciable N-glycosylase activity toward Oxa. This result suggests that hNEIL1, hNEIL2, and Endo VIII share in common the -fold of the active site, which is distinct from that of Endo III and hNTH1. Similar to Endo VIII (51,52), hNEIL1 and hNEIL2 are likely to have a capacity to accommodate certain types of purine lesions, but not all of them are excised. The determination of the three-dimensional structures of hNEIL1 and hNEIL2 and their comparison to those of Endo VIII (53), Endo III (54,55), and hNTH1 will provide further insight into the mechanisms underlying the differential recognition of oxidatively damaged bases by these enzymes. Toward the end of the preparation of this manuscript, two papers showing the differential specificity of Endo III and Endo VIII homologues toward the Tg isomers became in press (56,57).