Repair of Oxidized Bases in DNA Bubble Structures by Human DNA Glycosylases NEIL1 and NEIL2*

Repair of oxidatively damaged bases in the genome via the base excision repair pathway is initiated with excision of these lesions by DNA glycosylases with broad substrate range. The newly discovered human DNA glycosylases, NEIL1 and NEIL2, are distinct in structural features and reaction mechanism from the previously characterized NTH1 and OGG1 but act on many of the same substrates. However, NEIL2 shows a unique preference for excising lesions from a DNA bubble, whereas NTH1 and OGG1 are only active with duplex DNA. NEIL1 also excises efficiently 5-hydroxyuracil, an oxidation product of cytosine, from the bubble and single-stranded DNA but does not have strong activity toward 8-oxoguanine in the bubble. The dichotomy in the activity of NEILs versus NTH1/OGG1 for bubble versus duplex DNA substrates is consistent with higher affinity of the NEILs for the bubble structures of both damaged and undamaged DNA relative to duplex structure. These observations suggest that the NEILs are functionally distinct from OGG1/NTH1 in vivo. OGG1/NTH1-independent repair of oxidized bases in the transcribed sequences supports the possibility that NEILs are preferentially involved in repair of lesions in DNA bubbles generated during transcription and/or replication.

Repair of oxidatively damaged bases in the genome via the base excision repair pathway is initiated with excision of these lesions by DNA glycosylases with broad substrate range. The newly discovered human DNA glycosylases, NEIL1 and NEIL2, are distinct in structural features and reaction mechanism from the previously characterized NTH1 and OGG1 but act on many of the same substrates. However, NEIL2 shows a unique preference for excising lesions from a DNA bubble, whereas NTH1 and OGG1 are only active with duplex DNA. NEIL1 also excises efficiently 5-hydroxyuracil, an oxidation product of cytosine, from the bubble and singlestranded DNA but does not have strong activity toward 8-oxoguanine in the bubble. The dichotomy in the activity of NEILs versus NTH1/OGG1 for bubble versus duplex DNA substrates is consistent with higher affinity of the NEILs for the bubble structures of both damaged and undamaged DNA relative to duplex structure. These observations suggest that the NEILs are functionally distinct from OGG1/NTH1 in vivo. OGG1/NTH1-independent repair of oxidized bases in the transcribed sequences supports the possibility that NEILs are preferentially involved in repair of lesions in DNA bubbles generated during transcription and/or replication.
Reactive oxygen species are continuously generated as byproducts of respiration and are also induced during the inflammatory response and other pathological processes (1,2). These compounds and radicals are genotoxic and in addition to DNA strand breaks induce a plethora of base lesions, many of which are mutagenic and/or toxic (3). Most of the damaged bases are repaired via the base excision repair pathway that is initiated with their excision by DNA glycosylases with broad substrate range (4). Until recently, only two DNA glycosylases have been identified and characterized in mammalian cells as being responsible for repairing oxidatively damaged bases, 8-oxoguanine-DNA glycosylase (OGG1) 1 and endonuclease three homolog 1 (NTH1). OGG1 and NTH1 preferentially excise purine-and pyrimidine-derived lesions, respectively (5)(6)(7)(8).
Although OGG1 and NTH1 comprise the bulk of DNA glycosylase activity in mammalian cells for repairing oxidized bases, the fact that Ogg1Ϫ/Ϫ and Nth1Ϫ/Ϫ mice have no phenotype suggested the presence of back-up systems for repairing reactive oxygen species-induced lesions in mammalian cells (9,10). Preferential repair of 8-oxoguanine (8-oxoG) from the transcribed strand of a plasmid DNA in Ogg1-null mouse fibroblasts also supports the existence of additional DNA glycosylases (11,12). At the same time, accumulation of 8-oxoG residues in the bulk genome suggests the presence of two distinct processes for repair of 8-oxoG in the transcribed versus nontranscribed sequences (9,11,12). Preferential repair of thymine glycol, observed previously (13) in the transcribed sequences in genomic DNA, also indicates distinct repair systems for transcribed versus bulk DNA sequences. These results are reminiscent of the preferential repair of bulky base adducts in transcribed sequences in DNA via the nucleotide excision repair (NER) pathway, which led to the paradigm of transcription-coupled repair (TCR) as a distinct subpathway of NER (14,15). It is expected that repair of the transcribed strand is critical for preventing synthesis of mutant RNA or sustaining gene expression in order to maintain normal cellular phenotype. In contrast, repair of nontranscribed sequences may have no immediate biological consequences (16).
Here we report an unusual activity of NEIL1 and NEIL2 in excising lesions from DNA bubble structures. In contrast, neither OGG1 nor NTH1 shows any base excision activity for lesions present in a bubble or single-stranded (ss) DNA. This novel structural preference of the NEILs raises the possibility that these enzymes are involved in a distinct function in vivo for the repair of oxidized bases.

EXPERIMENTAL PROCEDURES
Oligonucleotide Substrates-Two 51-mer oligos containing either 5-OHU or 8-oxoG at position 26 from the 5Ј-end (shown in Table I) were purchased from Midland Certified Reagent. The undamaged control oligo contained C at position 26. The sequences of complementary oligos, purchased from Invitrogen, containing G or C opposite the lesion or containing sequences for producing bubble structures containing 5, 11, and 19 unpaired bases, and named B5, B11 and B19 respectively, are also shown in Table I. Two hundred fifty pmol each of the lesion strand and a complementary strand were heated at 94°C for 2 min in 50 l of phosphate-buffered saline and then slowly cooled to room temperature. To produce 32 P-labeled substrates, the single-stranded oligos were labeled at the 5Ј terminus with [␥-32 P]ATP and polynucleotide kinase before annealing.
Enzymes-Recombinant, wild type (WT) NEIL1, NEIL2, and OGG1 were purified to homogeneity from E. coli (17,18,26). The K53L mutant of NEIL1, encoding a C-terminal His tag fusion polypeptide, was generated in the pET22b vector using the Quick Change Site-directed Mutagenesis kit (Stratagene). The K49R mutant of NEIL2, without a fusion tag, was similarly generated in the pRSET(B) expression vector. The NEIL1 mutant was purified by affinity chromatography on a Ni 2ϩ column and subsequent chromatography in a HiTrap-SP column (26). The untagged WT and mutant NEIL2 were purified as before (18). NTH1 was a gift of R. Roy (27).
Incision Assay of DNA Glycosylases-DNA strand cleavage, due to intrinsic AP lyase activity of DNA glycosylases after lesion excision, was assayed by using duplex oligo containing 5-OHU or 8-oxoG in the 32 P-labeled strand, after incubation with the DNA glycosylases at 37°C for the indicated times in 10-l reactions containing 40 mM Hepes-KOH (pH 7.5), 50 mM KCl, and 100 g/ml bovine serum albumin. The reactions were stopped with 70% formamide, 30 mM NaOH, and the cleaved oligo products were then separated by denaturing gel electrophoresis in 20% polyacrylamide gel containing 7 M urea, 90 mM Tris borate (pH 8.3), and 2 mM EDTA. For analysis of enzyme kinetics, we incubated 5-OHU⅐G and 5-OHU⅐B11 oligo substrates (7.8 -125 nM) with either NEIL1 (at 17.5 nM for 3.5 min) or NEIL2 (at 20 nM for 5 min) at 37°C. The rate of product formation was linear for both enzymes under these conditions. The radioactivity in the DNA bands was quantitated by analysis in a PhosphorImager (Amersham Biosciences) using Image-Quant software.
Electrophoretic Gel Mobility Shift Assay-The 51-mer 5-OHU-containing oligo and a control oligo of identical sequence except for substitution of 5-OHU with C in position 26 (Table I) were labeled with 32 P at the 5Ј terminus by T4 polynucleotide kinase and then annealed with appropriate complementary strands to generate either perfect duplexes or 11-nt bubble structures (B11) flanked by duplex regions. The DNAs (1.5 nM) were incubated with WT and mutant NEIL1 (5-20 nM) for 10 min or WT and mutant NEIL2 (15-60 nM) for 15 min at 22°C in a buffer containing 40 mM Hepes (pH 7.3), 50 mM KCl, 12% glycerol, and bovine serum albumin (100 g/ml), before electrophoresis in nondenaturing polyacrylamide gels (10%) in Tris-glycine buffer (pH 8.4). The protein-bound DNA was quantitated in the linear range by Phospho-rImager analysis as before. The dissociation constants of binding were calculated according to Taylor et al. (28).

RESULTS
Characterization of DNA Bubble Structures-We annealed equimolar amounts of lesion-containing and complementary oligos to generate either completely base-paired duplexes or bubble structures of various unpaired regions flanked by duplex sequences as indicated in Table I. The melting temperatures (T m ) of the annealed oligo duplexes, calculated according to Oligo Analyzer 3.0 (Integrated DNA Technologies, Inc.), ranged from 50 to 74°C, depending on the bubble size. Thus the duplexes and bubble structures should be stable in our studies.
We then carried out detailed analysis of the structure of the duplex and B11 oligos prepared by annealing a lesion-containing or a normal strand with various complementary strands. Both strands in annealed DNAs were labeled with 32 P. In all cases, we observed a single band of the expected mobility after electrophoresis in nondenaturing gels for both 5-OHU-and normal base-containing oligos (Fig. 1). The mobility of the duplex was in between that of ssDNA and the bubble structure. These results indicated that the duplex and bubble structures were stable and homogeneous without contamination by singlestranded oligos (Fig. 1A). We confirmed the presence of a single-stranded region in the control B11 oligo containing C in position 26 by examining its susceptibility to the single strandspecific mung bean endonuclease (Fig. 1B). Duplex sequences, 20-mer in length on either side of the bubble in B11 oligo, should be resistant to the ssDNA-specific endonuclease, whereas the single-stranded sequences in these oligos would be degraded by the enzyme. As shown in Fig. 1B, with a limited amount of the nuclease, the duplex oligo remained unchanged (lanes 6 and 7), whereas a significant fraction of the B11, DNA in which both strands were radiolabeled (lanes 2 and 4), was converted into 20 -22-mer fragments from either strand (lanes 3 and 5) after denaturing gel electrophoresis. As expected, the ssDNA was extensively degraded by nuclease treatment (lane 9). Thus, the bubble oligo contained stable 20-mer duplex sequences flanking an ssDNA region. It should also be noted that no significant hairpin duplex region in the oligos was predicted from the sequence analysis under our assay condition (Oligo Analyzer 3.0, Integrated DNA Technologies, Inc.). These results also predict that B5 and B19 bubble oligos have similar base unpaired regions flanked by duplex segments. Preparation of Inactive NEIL1 and NEIL2 Mutants-To investigate the affinity of NEIL1 and NEIL2 for DNA bubbles containing substrate lesions by electrophoretic mobility shift assay (EMSA), as described later, inactive mutants of NEIL1 and NEIL2 were generated. Previous studies (29,30) have shown that E. coli Nei and Fpg contain a conserved Lys residue essential for base excision but not for the AP lyase activity. We therefore mutated the corresponding conserved Lys residues (Lys-53 in NEIL1 and Lys-49 in NEIL2) and purified both K53L NEIL1 and K49R NEIL2 mutant proteins to homogeneity ( Fig. 2A). Analysis of activity of the WT and mutant enzymes for excision of 5-OHU and strand cleavage of 32 P-labeled 5-OHU-containing duplex oligos confirmed our prediction that the NEIL1 and NEIL2 mutants lack DNA glycosylase activity (Fig. 2B).
Preferential Excision of 5-OHU from a DNA Bubble Structure-We compared the DNA glycosylase activity of NEIL1, NEIL2, and NTH1 in excising 5-OHU from duplex, singlestranded, and bubble oligos B5, B11, and B19 with 5, 11, and 19 unpaired bases, respectively. The lesion base was positioned in the middle of the bubble in all cases. We have confirmed the observation by Takao et al. (20), as shown in Fig. 3, that NEIL1 is active with ssDNA (Fig. 3A, lane 6), and we further showed that NEIL2 (Fig. 3A, lane 11) but not NTH1 (Fig. 3B, lane 5)  was similarly active with ssDNA. More remarkably, both NEIL1 and NEIL2 were highly active in excising 5-OHU when it was placed inside a bubble in B5, B11, and B19, in an otherwise duplex DNA (Fig. 3A, lanes 3-5 and 8 -10). In con-trast, NTH1 was completely inactive with the same substrate (Fig. 3B, lanes 3 and 4). This was expected, because NTH1 was also inactive with the ssDNA substrate (Fig. 3B, lane 5), as was shown before. NEIL1 is generally more active in base excision from duplex DNA than NEIL2 (17,18). However, although both NEIL1 and NEIL2 had comparable activity with the bubble substrates, NEIL2 showed almost 4-fold higher excision activity when 5-OHU was present in an 11-or 19-nt bubble than in a duplex. We determined the kinetic parameters of NEIL1 and NEIL2 with oligos containing 5-OHU either paired with G or present in the B11 bubble. The G⅐5-OHU pair should be generated in the genome in situ after oxidative deamination of C (31). Both NEIL1 and NEIL2 have higher catalytic specificity in excising 5-OHU from an 11-nt bubble than from a G⅐5-OHU pair in the duplex (Table II). Although both NEIL1 and NEIL2 prefer 5-OHU located in the bubble, NEIL1 has higher turnover and catalytic specificity than NEIL2. On the other hand, the preference of NEIL2 for the 11-nt bubble substrate relative to duplex DNA is higher (about 7-fold) than that of NEIL1 (about 3-fold).
Preferential Excision of 8-OxoG in a Bubble Structure-We showed earlier that with the duplex DNA NEIL1 but not NEIL2 was active in excising 8-oxoG from an 8-oxoG⅐C pair (17,18). However, we have now observed that NEIL2 could excise 8-oxoG when it was present inside a bubble (Fig. 4, lane 6). Like NTH1, OGG1, the major 8-oxoG-excising enzyme in eukaryotic cells, was completely inactive in excising 8-oxoG from both a bubble structure or ssDNA (Fig. 4, lanes 9 and 10). Interestingly, NEIL1 had higher 8-oxoG excision activity when the lesion was present in the 8-oxoG⅐C duplex (lane 2) than in the bubble (lane 3).
Affinity of NEIL1 and NEIL2 for Bubble Conformation-The lower K m values of both NEIL1 and NEIL2 for the 5-OHU substrate in bubble relative to duplex DNA suggested that these enzymes have intrinsic affinity for the bubble structure. FIG. 1. Characterization of single-stranded, bubble-containing, and duplex oligos. A, the undamaged (lanes 1-3) and 5-OHU-containing oligos (lanes 4 -6) were annealed to generate duplex and bubble structures as described under "Experimental Procedures" and analyzed by native PAGE. Their secondary structures are depicted on the left. Asterisks indicate 32 P label. B, sensitivity of nondamaged oligos of various structures to mung bean nuclease. Lane 1, markers, 51-mer substrate and predicted 20-mer nuclease product. Even-numbered lanes, control; odd-numbered lanes, after mung bean nuclease treatment. The drawings below the lanes indicate the structure of the oligos as in A. We therefore analyzed their relative affinity for duplex versus bubble DNA by EMSA. Because the activity of NEIL1 and NEIL2 does not require a cofactor, we could not carry out EMSA with WT NEILs and substrate oligos. We therefore used inactive mutants of the enzymes with 5-OHU-containing oligos. We also examined the affinity of WT NEIL1 and NEIL2 with control oligos that are identical to the 5-OHU-containing oligos except for substitution of 5-OHU with C. Fig. 5 shows a representative analysis of WT NEIL1 and NEIL2 complexes with duplex and bubble-containing DNA with normal bases. The slower moving complex with bubble DNA appears to be due to the binding of two enzyme molecules which became more pronounced with increasing protein concentration (Fig. 6A). It should be noted that NEIL1 and NEIL2 have no excision activity with normal bases, as expected (data not shown).
The affinity of WT NEIL1 and NEIL2 for normal DNA was calculated from the two-step binding Reaction 1, where D is free duplex or bubble oligo, and P, NEIL1 or NEIL2, is present in significant excess over DNA. K a1 and K a2 are the equilibrium constants for primary and secondary binding. Assuming the free duplex as the reference state, the partition function, Q, for the binding reaction is given by Reaction 2 (28,32).
With this definition, F, the fraction of duplex bound in the primary mode (the first shifted band) is given by Reaction 3. REACTION 3 With two complexes and knowing the relative concentrations of these complexes and free DNA, the fraction of bound DNA was measured at various enzyme concentrations. K a1 , i.e. 1/K d , could then be calculated from the binding isotherms by data fitting using the Sigma plot (representative plots are given in Fig. 6, B and C).
We calculated the apparent K d values of interaction of WT NEIL1 and NEIL2 with normal duplex and B11 bubble oligos (Table III, top). We could not calculate the K d values of these enzymes for single-stranded oligos because the complexes were not stable during EMSA. Interestingly, our initial EMSA results using the Tris borate buffer system were not reproducible. On the other hand, the results obtained from gels containing Tris-glycine buffer were reliable and used in calculating the above constants. We carried out similar binding studies of K53L NEIL1 and K49R NEIL2 mutants with 5-OHU-containing oligos. Table III, bottom, summarizes these results. It is evident from these EMSA results that both WT and mutant NEILs have an inherent affinity for the bubble structures. Similar EMSA studies with NTH1 and OGG1 did not indicate the formation of stable complexes of these enzymes with singlestranded and 11-nt bubble oligos containing normal bases (data not shown).

DISCUSSION
Mammalian NEIL1 and NEIL2 share many oxidized DNA base lesions as substrates with OGG1 and NTH1, the other two previously characterized DNA glycosylases involved in repair of these lesions. However, unlike the latter, which belong to the Nth family, NEIL1 and NEIL2 belong to the distinct Fpg/Nei family of enzymes based on reaction chemistry. It is interesting that there is no significant homology between NEIL1 and NEIL2 (17,18). Nevertheless, we show here that NEIL1 and NEIL2 share a common preference for an unusual DNA structure. Most DNA glycosylases of both bacterial and eukaryotic origin are active only with duplex DNA substrates. This is  4. Relative activity of NEIL1, NEIL2, and OGG1 in excision of 8-oxoG from duplex and bubble-containing oligos. An 8-oxoG-containing oligo (500 nM) of the same sequence as the 5-OHUoligo (Table I)  expected because the complementary strand serves as a template for the repair of gaps generated in the lesion-containing strand after damage removal. Thus, neither NTH1 nor OGG1 showed detectable activity for excising 5-OHU or 8-oxoG from ssDNA. In contrast, NEIL1 and NEIL2 have significantly higher activity and catalytic specificity for 5-OHU located in a single-stranded sequence in a bubble than for those in duplex DNA. Furthermore, the activity of NEIL2 was much higher with the bubble substrate than with ssDNA. We used 5-OHU and 8-oxoG in oligos of the same sequence as common substrates for the DNA glycosylases in order to avoid any potential impact of the sequence context. We confirmed an earlier observation that NEIL1 is active with ssDNA (20), and we showed for the first time that NEIL2 has similar activity.
Our results indicate that NEIL1 had comparable activity with ssDNA and duplex DNA and with DNA containing 5-19-nt bubbles. In contrast, NEIL2 had much higher activity with the bubble DNA than with duplex or ssDNA. The preference of NEIL2 for the bubble substrate extends to repair of 8-oxoG (Fig. 3). Compared with a robust activity with the 8-oxoG⅐C pair in duplex DNA, OGG1 showed no activity when 8-oxoG was present in ssDNA or in an 11-nt bubble.
The complex substrate preferences of NEIL1 and NEIL2 raise the question about their affinity for various DNA structures. Our EMSA results with undamaged DNA showed that both NEIL1 and NEIL2 have higher affinity for bubble structures than for duplex DNA. Furthermore, the binding of NEILs to ssDNA was less specific than to the B11 bubble. These data strongly suggest that whereas NEIL1 and NEIL2 stably bind to single-stranded sequences even in the absence of substrate lesions, their initial interaction with DNA requires a duplex structure. The similar trend in binding of NEIL1 and NEIL2 with lesion-containing and nondamaged DNA in singlestranded, duplex, and bubble conformation supports our conclusion that these enzymes have intrinsic affinity for the bubble structure.
We then assessed the contribution of the base lesion to the binding by using inactive mutants of NEIL1 and NEIL2 with 5-OHU-containing oligos in various DNA structures. We observed a lower affinity of the mutants for the substrate oligo compared with that of the WT enzymes for the undamaged oligo (Table III). It thus appears likely that significant changes in the tertiary structures of these enzymes were induced by the mutations. Nevertheless, the affinity of both NEIL1 and NEIL2 for the bubble DNA was consistently higher than for the duplex DNA. This is in strong contrast to the situation with NTH1 and OGG1 which did not produce stable complexes with either ssDNA or B11 bubble.
Tainer and co-workers (33) have proposed a push-pinch-pull model for lesion detection by DNA glycosylases while scanning by NEILs along the DNA backbone. It is unclear how such a  model could apply to initial scanning when the duplex DNA changes abruptly to ssDNA before reaching the lesion. This issue may be resolved once the structures of DNA-bound NEILs are elucidated. Given the absence of requirement for ATP whether NEIL1 and NEIL2 move unidirectionally on the DNA strand is not known. Furthermore, it is highly likely that these enzymes, like other DNA glycosylases, flip out the damaged nucleotide before catalysis (34,35). It should thus be easier for them to insert the base lesion into the catalytic pocket from ssDNA or a loop than from a duplex DNA. The inability of OGG1 and NTH1 to similarly utilize ssDNA or bubble DNA substrates may be due to distinct mechanisms involved in stabilizing the enzyme-substrate complex for the NEILs versus other DNA glycosylases.
Finally, the physiological implications of the unusual substrate structure preference of NEILs deserve to be discussed. The bubble DNA containing noncomplementary bases used in these studies was designed to represent unwound duplex regions in the genome. Similar bubbles are transiently formed in vivo during both transcription and DNA replication. Although preferential repair of DNA template strands during DNA replication has not been established, we had earlier hypothesized about replication-associated repair of mutagenic bases in mammalian cells (36). During DNA replication, several distinct DNA helicases function in unwinding the DNA duplex ahead of the replication fork to generate a bubble structure (37). Sphase-specific activation of NEIL1 raises the possibility that it is involved in repair of the damage present in the replication bubble.
In contrast, NEIL2, independent of cell cycle expression, could be involved in TCR. TCR of bulky base adducts (e.g. UV photoproducts), which prevent RNA chain elongation, was identified as a distinct subpathway of NER. The blockage of RNA polymerase II at the damage site triggers repair of the transcribed strand, which is required for survival and maintenance of genomic integrity (14, 15, 38 -40). Preferential repair of oxidative damage in transcribed sequences via base excision repair was also observed (11)(12)(13). Nevertheless, because oxidatively damaged bases, unlike bulky adducts, do not block transcription, at least in vitro, how TCR is activated during in vivo repair of these bases is not clear (41). We propose the following model of how NEIL2 initiates repair of oxidized bases in a transcription bubble. After base excision by NEIL2 in the transient bubble, generated ahead of the growing RNA chain, the DNA strand is cleaved due to the AP lyase activity of the enzyme. In contrast, OGG1 and NTH1 are unable to carry out base excision in the bubble. This could explain their dispensability in repairing damage from the transcribed sequences. The single-strand break generated by NEIL2 prevents forward movement of the transcription complex and induces either its retrograde movement or dissociation from the template (38). The bubble then collapses to reform the duplex structure, allowing completion of the strand break repair and resumption of transcription.
Whether NEILs interact with transcription and replication complexes in order to carry out targeted repair of DNA templates in a coordinated fashion remains to be investigated.