Conformational Dynamics of DNA Repair by Escherichia coli Endonuclease III*

Background: Endonuclease III is responsible for base excision repair of oxidized or reduced pyrimidine bases. Results: Stopped-flow kinetics analysis of endonuclease III interaction with DNA was performed. Conclusion: Endonuclease III uses a multistep mechanism of damage recognition, which likely involves Gln41 and Leu81 as lesion sensors. Significance: The results provide new insight into the mechanism of damage recognition by DNA glycosylases of the helix-hairpin-helix-GPD structural superfamily. Escherichia coli endonuclease III (Endo III or Nth) is a DNA glycosylase with a broad substrate specificity for oxidized or reduced pyrimidine bases. Endo III possesses two types of activities: N-glycosylase (hydrolysis of the N-glycosidic bond) and AP lyase (elimination of the 3′-phosphate of the AP-site). We report a pre-steady-state kinetic analysis of structural rearrangements of the DNA substrates and uncleavable ligands during their interaction with Endo III. Oligonucleotide duplexes containing 5,6-dihydrouracil, a natural abasic site, its tetrahydrofuran analog, and undamaged duplexes carried fluorescent DNA base analogs 2-aminopurine and 1,3-diaza-2-oxophenoxazine as environment-sensitive reporter groups. The results suggest that Endo III induces several fast sequential conformational changes in DNA during binding, lesion recognition, and adjustment to a catalytically competent conformation. A comparison of two fluorophores allowed us to distinguish between the events occurring in the damaged and undamaged DNA strand. Combining our data with the available structures of Endo III, we conclude that this glycosylase uses a multistep mechanism of damage recognition, which likely involves Gln41 and Leu81 as DNA lesion sensors.

Base modifications induced in DNA by a number of endogenous or exogenous factors lead to adverse effects in cells (1)(2)(3). Some damaged DNA bases are potentially mutagenic, and are involved in carcinogenesis and aging (4). Other lesions, such as thymine glycol and 5,6-dihydrouracil (dHU), 3 can block DNA polymerase during the replication (5-7). Non-bulky small base lesions, derived from deamination, alkylation, or oxidation, are excised and replaced in the base excision repair pathway. The first enzymatic step in the base excision repair pathway is typically the excision of the substrate base from duplex DNA by a DNA glycosylase, one of the enzymes that catalyze the cleavage of the N-glycosidic bond between the substrate base and the 2Ј-deoxyribose, creating an abasic (AP) site.
Endo III is a bifunctional DNA glycosylase possessing N-glycosylase and AP lyase activities ( Fig. 1) (12,13). The principal amino acids involved in the catalysis are Lys 120 and Asp 138 . The former is the nucleophile that attacks the C1Ј atom of deoxyribose, resulting in the cleavage of the N-glycosydic bond and subsequent formation of a Schiff base covalent intermediate (Fig. 1, step 1). The following ␤-elimination reaction leads to the departure of the 3Ј-phosphate. The subsequent Schiff base hydrolysis releases the enzyme and leads to formation of a singlestrand break in DNA duplex with an ␣-␤-unsaturated aldehyde at the 3Ј-end and a phosphate at the 5Ј-end (Fig. 1, step 2).
The x-ray structures of free Endo III from E. coli as well as three complexes of Endo III from Geobacillus stearothermophilus with DNA have been published (14,15). The overall structure of enzyme shows two similarly sized globular domains separated by a deep DNA-binding groove that contains the active site cavity (Fig. 2). Endo III binding induces gross conforma-tional changes in DNA, in which the helix axis is bent ϳ55°and the damaged nucleotide is everted to the active site of the enzyme.
Although the three-dimensional structure of the Endo III-DNA complex is established, it tells us little about the mechanism of substrate recognition by this enzyme, because the DNA in the structures contains no damaged base. The overall mode of interaction of Endo III with DNA is very similar to that seen with other members of the HhH-GPD structural superfamily, making it possible to compare the kinetic and conformational features of DNA binding and damage recognition with other well studied members of this superfamily (16). Our previous studies (17)(18)(19)(20)(21) of another member of the helix-hairpin-helix (HhH)-GPD superfamily, human 8-oxoguanine-DNA glycosylase (hOGG1), revealed that hOGG1 use multistage mechanism of lesion recognition that includes step-by-step involvement of different amino acids. It was concluded (20) that at the initial step of DNA binding Arg 154 and Arg 204 (corresponding to Arg 78 and Arg 84 of G. stearothermophilus Endo III in their spatial arrangement but not in the alignment, Fig. 3A) "pull out" the cytosine base located opposite the oxoG base and Tyr 203 (spatially corresponding to Leu 82 in G. stearothermophilus Endo III, Fig. 3B) is partially inserted into the DNA duplex. These interactions lead to DNA distortion and eversion of the oxoG base from DNA to the intermediate base-binding site (exo-site) of the enzyme. A thermodynamic analysis (21) argues that the initial step of the DNA substrate binding is mainly governed by a negative enthalpy change due to the formation of favorable contacts between hOGG1 and DNA, as well as due to the partial desolvation of the interface resulting in positive entropy change. The oxoG base is then fully placed into the active site and forms specific interactions within the base recognition pocket detected in the structure of the pre-catalytic hOGG1-DNA complex (22). In addition, Asn 149 (corresponding to Gln 42 in G. stearothermophilus Endo III, Fig. 3B) forms a hydrogen bond between the amide carbonyl of its side chain and the amino group of the opposite cytosine. Interestingly, hOGG1 can also guide the undamaged G base through the exosite into the active site, albeit less efficiently, but cannot cleave its N-glycosidic bond indicating strong discrimination selectivity between damaged and undamaged DNA (23,24). A stepwise thermodynamic analysis (21) reveals that the discrimination of nonspecific G base versus specific oxoG base mostly occurs at the second step of the DNA binding. At the last pre-catalytic step the enzyme-substrate complex is finally adjusted to a catalytically competent conformation. This process is characterized by large endothermicity compensated by a significant increase in entropy likely originating from dehydration of the DNA grooves.
Here we present a kinetic analysis of conformational changes in DNA with another DNA glycosylase from the HhH-GPD structural superfamily, E. coli Endo III. Two fluorescent reporters, 2-aminopurine (aPu) placed 3Ј to the damaged nucleotide and tC O placed opposite to the lesion were chosen for detection of DNA dynamics. Dodecamer oligodeoxyribonucleotide duplexes contained centrally placed specific lesions: dHU, the natural AP site and its uncleavable analogue (3-hydroxytetrahydrofuran-2-yl)methyl phosphate (F) or undamaged nucleotide. The substrates containing dHU are subject to the full enzymatic cycle, which includes DNA binding, N-glycosidic bond cleavage, ␤-elimination, and product release. The substrates containing an AP site are limited to ␤-elimination and product release. The undamaged DNA duplexes and duplexes contain-  ing F reveal the conformational changes in DNA during binding with Endo III uncomplicated by catalytic steps. A comparison of fluorescence kinetic data with available structures for Endo III allowed us to propose the mechanism of Endo III interaction with specific and nonspecific sites.

Experimental Procedures
Purification of Endo III-To purify recombinant E. coli Endo III, 1 liter of E. coli JM105 carrying the pNth 10 (25, 26) plasmid were grown in 2ϫ YT broth supplemented with 50 g/ml of ampicillin at 37°C until A 600 ϭ 0.6 -0.8, then shifted to 30°C and induced overnight with 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside. The cells were harvested by centrifugation, resuspended in 30 ml of the lysis buffer (20 mM HEPES-KOH, pH 7.8, 40 mM NaCl), and lysed using a French press. Insoluble material was pelleted by centrifugation. The supernatant was passed through a Q-Sepharose column (GE Healthcare, Little Chalfont, UK) in the same buffer but containing 250 mM NaCl. The eluate was diluted 2-fold by 20 mM HEPES-KOH (pH 7.8) and loaded into a heparin column (GE Healthcare). Endo III was eluted using a 100 -1500 mM NaCl gradient in 20 mM HEPES-KOH. The fractions containing Endo III were collected, glycerol was added to 50%, and the samples were stored at Ϫ20°C.
Oligodeoxynucleotides-The sequences of oligodeoxyribonucleotides used in this work are listed in Table 1. The oligodeoxyribonucleotides were synthesized by the standard phosphoramidite method using an ASM-700 synthesizer (BIOSSET, Novosibirsk, Russia) in the ICBFM Laboratory of Bionanotechnology. The phosphoramidites were purchased from Glen Research (Sterling, VA). Duplexes were prepared by annealing the modified and the complementary strand at a 1:1 molar ratio.
Stopped-flow Experiments-The experiments were conducted essentially as described previously (27)(28)(29). All experiments were carried out at 25°C in buffer containing 50 mM Tris-HCl (pH 7.5), 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, and 7% glycerol (v/v). An SX.18MV stopped-flow spec-trometer (Applied Photophysics, Leatherhead, UK) fitted with a 150-W Xe arc lamp was used. The dead time of the instrument was 1.4 ms. The excitation wavelengths were 310 and 360 nm for the aPu and tC O fluorescent dyes, respectively. The emission was monitored using long pass wavelength filters (Corion) at 370 nm for aPu and 395 nm for tC O . Endo III was placed in one syringe of instrument and rapidly mixed with the substrate in another syringe. The concentration of substrates in all experiments was 1.0 M, and the concentration of Endo III was varied in the 0.5-4.0 M range. The reported concentrations of reactants are those in the reaction chamber after mixing. Typically, each trace shown is the average of four or more individual experiments; the reported rate constants represent the mean Ϯ S.D. of such data sets. In the figures, the individual traces are manually offset for clarity.
Kinetic Data Analysis-The kinetic parameters were obtained by global non-linear fitting using the DynaFit software (BioKin, Pullman, WA) (30) as described previously (21,(31)(32)(33). The approach is based on the fluorescence intensity variation in the course of the reaction due to sequential formation and further transformation of the DNA-enzyme complex and its conformers. The software performs numerical integration of a system of differential equations with subsequent nonlinear least-squares regression analysis. The response factors of the intermediates (that are essentially the products of the fluorescence quantum yields) are treated as fitting parameters in the data processing. In the evaluated mechanisms, except for the first bimolecular step, all other reactions are first order reactions. In the data processing, the kinetic information is obtained from the "temporal behavior" of the fluorescence intensity, not from the "amplitudes" of the specific signal contributions. The "response factors" for different conformers resulting from the fits were not used in the determination of the equilibrium constants, but rather provided additional information on the fluorescence intensity variations in different complexes and conformers. Processing of individual kinetic curves does not unambiguously provide the kinetic parameters; therefore, global fits of sets of kinetic curves obtained at different concentrations of the reactants was used. The fits of all relevant rate constants for the forward and reverse reactions, as well as the specific molar responses for all intermediate complexes, were optimized. After that rate constants for the forward and reverse reactions were fixed and additional optimization of the specific molar responses of intermediate complexes for individual kinetic traces was done.
Product Analysis-To analyze the products formed by Endo III, the substrates were 5Ј-32 P labeled using phage T4 polynucleotide kinase and [␥-32 P]ATP and the reaction was performed under the conditions described above. The products were precipitated by adding 10 volumes of 2% LiClO 4 in acetone. The precipitates were washed three times with 100 l of acetone, dried, dissolved in 4 l of water and 3 l of loading buffer (80% formamide, 20 mM EDTA, 0.1% xylene cyanol, 0.1% bromphenol blue), and analyzed by 20% denaturing PAGE. To observe formation of the AP site after base excision, the reaction mixture was treated with 0.3 M NaOH and heated for 15 min at 75°C to cleave the DNA backbone at AP sites. The gels were visualized using Agfa CP-BU x-ray film (Agfa-Geavert, Mortsel, Belgium), and the autoradiograms were scanned and quantified using Gel-Pro Analyzer software (Media Cybernetics, Rockville, MD). The time course curves of product accumulation were fitted by a single exponential function using Origin software (OriginLab Corporation, Northampton, MA) (Equation 1), where A is the amplitude, k obs is the rate constant, and t is the reaction time.

Conformational Transitions in the Damaged DNA Strand-
The most informative approach to follow conformational transitions occurring in protein-DNA interactions through fluorescence changes involves a combination of the intrinsic protein fluorescence from tryptophan residues and DNA fluorescence from incorporated reporters. However, we were unable to detect changes in Endo III fluorescence in the reactions with any ligand or substrate, which is most likely due to suboptimal location of Trp residues or lack of significant conformation mobility in the protein molecule. Thus we have resorted to a set of fluorescent reporters to follow DNA dynamics in the complex.
We have first examined the activity of Endo III on DNA substrates containing aPu or other fluorescent reporters either 5Ј or 3Ј to dHU. The presence of aPu did not affect the reaction rate in comparison with the dHU substrate carrying no reporter (data not shown). The aPu residue located immediately 5Ј to lesions was less sensitive to the conformational changes of DNA than when placed at the 3Ј-side (data not shown). In contrast, analysis of the product accumulation by PAGE showed that other fluorescent base analogs (C py , tC O , and 3HC) placed at the 5Ј-side of dHU make the substrate totally resistant to cleavage by Endo III (data not shown), similar to recent findings for DNA glycosylase Nei, which also excises oxidized pyrimidines but belongs to another structural superfamily (33). Therefore, we constructed 12-base pair oligodeoxyribonucleotide duplexes with an aPu reporter on the 3Ј-side of the site of unmodified G, or modified F, AP, and dHU residues as specific ligands and substrates for Endo III ( Table 1). The substrates containing dHU are subject to the full enzymatic cycle, which includes DNA binding, N-glycosidic bond cleavage, ␤-elimination, and product release. The substrates containing an AP site are limited to ␤-elimination and product release. The undamaged DNA duplexes and duplexes containing F reveal the conformational changes in DNA during binding with Endo III uncomplicated by catalytic steps.
G-aPu Ligand-The binding of undamaged G-aPu ligand by Endo III did not lead to appreciable changes in the aPu fluorescence intensity (data not shown) at any enzyme:substrate ratio indicating that this strand may be only slightly distorted in the complex of Endo III with nonspecific DNA.
F-aPu Ligand-When the concentration of Endo III varied in the 0.5-4 M range with the constant 1 M concentration of the uncleavable analog of AP site, F-aPu ligand, the fluorescent traces indicated that the binding was essentially complete within 10 s (Fig. 4A). The amplitude of the fluorescence change was dependent on the Endo III concentration. The decrease in the aPu fluorescence intensity is suggestive of the transition of the aPu base into a more hydrophobic environment, most reasonably explained by an insertion of amino acid side chains of Endo III (such as Gln 41 ) into the vacant space present in the DNA double helix due to the abasic void. As expected, the amplitude of aPu fluorescence intensity change increases with the increase in the Endo III concentration, reflecting a shift of binding equilibrium toward the complex. The kinetic data are satisfactorily described by Scheme 1 with a single equilibrium step. Rate and equilibrium constants derived from Scheme 1 are shown in Table 2.
AP-aPu Substrate-AP site is a cleavable substrate for Endo III, undergoing ␤-elimination after binding to the enzyme (see Fig. 1). Fig. 4B shows the time course fluorescence traces upon Endo III interaction with the AP-aPu substrate. The concentration of Endo III varied in the 1-4 M range at the fixed 1 M concentration of the substrate. The shapes of the fluorescence curves reveal four phases in the process: (i) a slight increase in fluorescence up to 0.2 s, (ii) a decrease phase up to 3 s, (iii) a final increase in the fluorescence intensity, and (iv) a plateau. The minimal kinetic mechanism corresponding to these fluores-cence changes contains two reversible binding steps, one irreversible step of catalysis, and an equilibrium stage of the product release from the complex with the enzyme (Scheme 2). The rate and equilibrium constants obtained by fluorescence trace fitting are shown in Table 2. The affinity of Endo III for the natural AP site was ϳ12-fold higher than for its synthetic analog, F; the association constant K bind were 5 ϫ 10 6 M Ϫ1 and 0.4 ϫ 10 6 M Ϫ1 , respectively. Moreover, the first binding step was characterized with a higher rate constant of complex formation (k 1 ) for the AP-aPu substrate than for the F-aPu ligand. Despite this, similar association constants K 1 were observed for both DNA duplexes. The observed difference in the affinity of Endo III for F-and AP-containing DNA is provided by the second binding step, absent in the F-aPu ligand. It can be inferred that the adjustment of the enzyme-substrate complex into a catalytically competent conformation requires the second step of the productive binding process. The third step in Scheme 2 is irreversible, most likely attributed to the chemical step of catalysis, i.e. AP lyase reaction, characterized with k 3 aPu ϭ 0.38 Ϯ 0.02 s Ϫ1 . This rate constant is an order of magnitude higher than the rate constant for the same reaction characterizing hOGG1 with its notably low turnover rate (18).
dHU-aPu Substrate-Although no structure of Endo III in a complex with base-containing DNA is available, the structures of G. stearothermophilus Endo III bound to abasic DNA (15) suggests that the damaged base should be flipped out from the stack into the active site pocket. The eversion of dHU base from the DNA helix should lower the hydrophobicity of the environment of the aPu residue and therefore induce an increase in the aPu fluorescence signal. The observed fast increase in the aPu fluorescence intensity up to 5 ms (Fig. 4C) shows that Endo III induces the damaged base eversion at an early stage of DNA binding. After the fast eversion of the dHU base from DNA the void is plugged by amino acid side chains. The void-filling process, observed also with abasic DNA, is characterized by a decrease in the fluorescence intensity and extends up to 20 s. A further increase in fluorescence intensity may be associated with the catalytic step and release of the product from the active site of the enzyme. The minimal four-step kinetic scheme describing the observed changes of aPu fluorescence intensity is identical to that proposed for the AP-aPu substrate (Scheme 2) and contains two equilibrium steps of substrate binding followed by one irreversible chemical step and then by an equilibrium step of product release. The kinetic constants that satisfy Scheme 2 are presented in Table 2. As can be seen from their values, Endo III possesses ϳ4-fold lower affinity for dHU base than for AP site at the first binding step, whereas the second binding step, which corresponds to the void-filling process, does not discriminate between AP and dHU substrates. The products of conversion of both specific substrate conversions have lower affinities for Endo III than AP and dHU substrates (compare K P aPu ). Although Endo III processes dHU substrates in two consecutive reactions, DNA glycosylase and AP lyase, only one irreversible step characterized by the rate constant k 3 aPu was identified. The value of k 3 aPu , 0.054 s Ϫ1 , was 6.7-fold lower than for the AP-aPu substrate, indicating that under our experimental conditions (close to single turnover) the rate is limited by the N-glycosylase reaction, in agreement with the earlier reports (10,48).
Conformational Transitions in the Complementary DNA Strand-To register the conformational transitions in the DNA strand opposite to the lesion, three fluorescent base analogs (C py , aPu, and tC O ) were tested. In previous work, we have used C py and tC O to study conformational dynamics induced in the substrate by DNA glycosylases Fpg (31) and Nei (33), respectively. In the case of Endo III interaction with the dHU-containing duplexes the initial screening experiments have shown that DNA conformational changes were more pronounced if the tricyclic cytosine analog tC O was placed opposite to the damaged base compared with C py or aPu reporters (data not shown). Interestingly, earlier reports characterized tC O as a base analog that is highly fluorescent but has a lower sensitivity to the environment in double-stranded DNA in comparison with aPu (49,50). An aPu residue placed opposite dHU was insensitive to the binding by Endo III. Therefore, tC O located opposite to the lesion was used for the study of conformational transitions in the complementary DNA strand.
G/tC O Ligand-Nonspecific binding of the G/tC O ligand by Endo III led to an increase in tC O fluorescence intensity up to 10 s (Fig. 5A). As was suggested for another member of the HhH-GPD structural superfamily, hOGG1 (20), the initial step of DNA binding includes hydrogen bond formation between two helix-invading amino acids, Arg 154 and Arg 204 (Arg 78 and Arg 84 in G. stearothermophilus Endo III, Fig. 3A), and the cyto-

TABLE 2
The rate and equilibrium constants for interactions of Endo III with X-aPu-substrates* SCHEME 2

Conformational Dynamics of DNA Repair by E. coli Endo III
sine base located opposite to the damaged base. Furthermore, as was suggested for Tyr 203 and Asn 149 of hOGG1 (20), the side chain of Leu 81 of Endo III (an equivalent of Tyr 203 , Fig. 3B) can be partially inserted into the DNA duplex during nonspecific complex formation, and the side chain carbonyl of Gln 41 of Endo III (an equivalent of Asn 149 , Fig. 3B) contacts the base opposite to the lesion site. Therefore, similarly to hOGG1, the tC O fluorescence intensity increase may be due to destabilization and local unwinding of the DNA duplex. Fitting the experimental data to a one-site binding model (Scheme 1), we calculated the values of the forward and reverse rate constants (Table  3). Although a direct comparison between G/tC O and G-aPu nonspecific ligands was precluded by the lack of fluorescent changes in the latter, the rate constant values show that Endo III interacts with G/tC O very slowly, ϳ10 3 -fold slower than with the specific AP-aPu or dHU-aPu substrates (see Table 2).

F/tC O Ligand-
The interaction of Endo III with F/tC O ligand produced a two-phase change in the fluorescence intensity in the stopped-flow experiments (Fig. 5B). The fast decrease phase was complete by 10 ms, followed by an increase phase up to 5 s. The stopped-flow data were fitted to Scheme 3; the forward and reverse rate constants are presented in Table 3.
Interestingly, tC O placed opposite to the lesion site allowed detection of the events occurring earlier during Endo III binding to an F-site than those revealed by aPu reporter placed on the 3Ј-side of this lesion (compare Figs. 4A and 5B). This observation supports the hypothesis that formation of the initial specific enzyme-DNA recognition complex directly involves both damaged and complementary DNA strand (20). Besides, the rate constants presented in Tables 2 and 3 (Fig. 5C). The second phase of decrease in the tC O fluorescence proceeded between 10 ms and 1 s. This two-step drop in the tC O fluorescence presumably reflects productive AP site recognition by Endo III. The ␤-elimination reaction resulting in the damaged strand cleavage, as well as the product release, proceeded at times Ͼ1 s and induced an increase in the   (Fig. 5C). The minimal kinetic scheme describing the observed changes in tC O fluorescence intensity was identical to Scheme 2. The rate and equilibrium constants of the steps estimated from this kinetic scheme are listed in Table 3. In the case of AP substrate the first binding step detected by the tC O fluorescence was ϳ23-fold faster than the step detected by the aPu fluorescence. Nevertheless, the total association constants K bind are similar for both substrates. The rate constant of the irreversible catalytic step k 3 tCo was 0.13 s Ϫ1 , somewhat lower than k 3 aPu ϭ 0.38 s Ϫ1 obtained for the AP-aPu substrate. Therefore, tC O moderately affects the rate of the catalytic step, presumably due to an increased size of this residue in comparison with the natural cytosine base.
dHU/tC O Substrate-Binding of the dHU/tC O substrate by Endo III induced changes in the tC O fluorescence similar to those observed with the AP/tC O substrate (Fig. 5D). When the base is present, two chemical steps occur after the formation of the catalytically competent complex: cleavage of the N-glycosidic bond and ␤-elimination reaction. The additional chemical step of N-glycosidic bond cleavage slightly shifts the rising phase of the curves into longer times compared with the AP/tC O substrate. The traces can be fit to the minimal kinetic Scheme 2 containing two equilibrium substrate binding steps followed by an irreversible chemical step and then with an equilibrium product release step. The rate constants of the elementary steps estimated from this kinetic scheme are listed in Table  3. It can be seen that k 3 tCo ϭ 0.077 s Ϫ1 is very close to the k 3 value determined by aPu fluorescence (k 3 aPu ϭ 0.054 s Ϫ1 ). The rate constants corresponding to the binding step are also in a general agreement for both reporters. Therefore, in the case of the dHU substrate the conformational perturbation induced in DNA by Endo III binding occur nearly simultaneously in both strands.
Opposite Base Specificity of Endo III-It has been reported that the base opposite the lesion may influence the activity of Endo III but the opposite base specificity strongly depends on the lesion: whereas thymine glycol-, 5,6-dihydrothymine-, formamidopyrimidine-, and N-(2-deoxy-␤-D-erythro-pentofuranosyl)-N-3-[(2R)-hydroxyisobutyric acid]-urea-containing substrates show a moderate to pronounced preference for G opposite the lesion, urea, and 5-hydroxycytosine are excised equally well from all pairs (48,51,52). We have examined the opposite base specificity of dHU excision by Endo III in a single turnover mode (Fig. 6). The formation of the nicked products was directly detected by PAGE. Endo III was found to moderately discriminate the bases placed opposite to the lesion; the order of preference G Ͼ A Ͼ T Ͼ C was observed. This order is in agreement with the data obtained for other lesions showing at most a severalfold preference for G over A and purines over pyrimidines (48,51,52). No significant difference was observed when the reaction products were treated or not by alkali, indicating that the AP lyase activity does not limit the reaction and proceeds with a rate similar to the rate of the glycosylase reaction; this is consistent with the observation of only a single fluorescence change in both AP substrate and dHU substrate cleavage.

Discussion
Endo III from E. coli is a prototypical member of the endonuclease III structural superfamily of DNA glycosylases, defined according to the presence of two structural elements, a HhH motif and an extended glycine/proline-rich loop containing a catalytic Asp residue (G/P…D loop) (16). Crystal structures have been reported for free Endo III from E. coli (14,53) and for Endo III from G. stearothermophilus cross-linked to DNA as a reduced Schiff base intermediate or non-covalently bound to F-containing DNA (15). All structures show that Endo III is a two-domain ␣-helical protein, with one domain organized into a six-helix barrel, and another domain containing an iron-sulfur [4Fe-4S] 2ϩ cluster (FeS).
Because the conformational changes in the protein globule of OGG1, another Endo III superfamily member, could be followed by Trp fluorescence with a reasonably good time and amplitude resolution (17,18,20), we expected to be able to obtain Trp fluorescence traces for Endo III-DNA interactions as well. E. coli Endo III contains two Trp residues, Trp 132 and Trp 178 . The latter is buried inside the FeS domain, tightly packed against the ␣A helix, and is unlikely to be considerably mobile in the course of DNA substrate recognition and processing. Trp 132 , however, is located in an inter-domain loop on the surface of the protein opposing the DNA-binding groove, and we anticipated that DNA binding would induce significant conformational changes to be observed by Trp fluorescence. Our failure to detect Trp fluorescence changes indicates that either Endo III binds DNA with minimal domain movement, or the solvent exposure of Trp 132 changes little upon DNA binding. Our preliminary structure of E. coli Endo III cross-link to DNA 4 shows no great deviations from the structure of the wildtype protein (C␣ root mean square deviation: 2.12-Å overall, 2.09-Å six-helix barrel domain, 1.27-Å FeS domain) and shifted but still widely exposed side chain of Trp 132 , so both factors probably contribute to the observed lack of Trp fluorescence change.
DNA in complex with both G. stearothermophilus Endo III (15) and E. coli Endo III 4 is grossly distorted at the site of the 4 D. O. Zharkov, unpublished data. lesion, with the DNA axis kinked ϳ55°, the minor groove widely open, and the damaged nucleotide everted from the helix into the active site pocket of the enzyme. Four distinct structural elements of the protein are responsible for this distortion, discussed below for the published G. stearothermophilus structure (refer to the alignment in Fig. 7 for the corresponding E. coli residues). First, a short turn connecting ␣-helices ␣B and ␣C, together with the adjacent helix parts (Ser 40 -Leu 47 ), is inserted into the minor groove of the DNA double helix to push the damaged nucleotide out. The void left after the eversion is filled by Gln 42 , and the adjacent bases are stacked against its side chain amide group (the 3Ј-base) and the planar peptide bond Gln 42 -Cys 43 (the 5Ј-base). Second, ␣-helices ␣F and the beginning of ␣G (Arg 78 -Asn 85 ) buttress the orphaned nucleotide opposite the lesion and two flanking nucleotides, with Leu 82 wedging between the orphaned base and the base 5Ј to it and introducing a kink into this strand. Third, the damaged strand 3Ј to the lesion presses against the hairpin and the beginning of the second helix (␣I) of the HhH motif (Leu 115 -Lys 121 ); these interactions kink the damaged strand, and Lys 121 forms the Schiff base during the reaction (4). Finally, Asp 139 at the very end of the G/P…D loop, the first half of the ␣J helix (Thr 140 -Arg 144 ), and Arg 186 , pinch the phosphates p 0 and p Ϫ2 (the distance between the phosphorus atoms of these groups is 7.4 Å, cf. ϳ13 Å mean distance between every other phosphate in the same strand of B-DNA), thus helping to evert the damaged nucleotide and kink the DNA axis.
This arrangement of DNA and protein residues provides insights into the conformational transition dynamics associated with the changes in fluorescence of the reporters incorporated into DNA (Fig. 8). Fluorescence of synthetic base analog aPu is quenched in a less polar environment, so the decrease in the fluorescence of this reporter at the initial phases of reaction with the F ligand and AP and dHU substrates is obviously associated with the insertion of Gln 41 into the space vacated by the everted nucleotide and stacking of its amide group against aPu. Multiple solution NMR studies of DNA duplexes show that F and AP tend to be intrahelical or slightly displaced toward the major or minor groove when the opposite base is well stacked (reviewed in Ref. 54). In our case, when the opposite guanine base is in the …GGC… context, the abasic nucleotides are likely intrahelical and therefore the adjacent bases are solvent accessible in free DNA, so Endo III binding and Gln 41 stacking would decrease the fluorescence from the baseline. No structure of dHU-containing DNA has been reported but, as other 5,6-saturated pyrimidine lesions (54), it is probably displaced or partially extrahelical, also exposing the adjacent bases to the solvent in free DNA. Fast increase in the aPu fluorescence in the case of the dHU substrate indicates fixation of the damaged base in the extrahelical state. However, longer times of fluorescence decrease with dHU indicate that the insertion of Gln 41 in the DNA void after eversion of this lesion in the active site is energetically less favorable compared with AP substrate.
When another fluorescent reporter, tC O , was incorporated opposite to the lesion, a single decrease in its fluorescence for the F ligand (Fig. 5A) and a two-phase decrease with the AP and dHU substrates (Fig. 5, B and C) were recorded. On the contrary, binding of the G ligand as well as the second step of F ligand binding lead to an increase in tC O fluorescence. The ␣F-␣G motif of Endo III binds the undamaged strand exclusively through the backbone carbonyls and is flexible enough to accommodate G or A in two different conformations (15), so tC O will be presumably bound without significantly disrupting functionally important interactions with the undamaged strand. In the same vein, although Endo III showed some specificity for a base opposite the lesion, it was capable of efficiently excising dHU paired with all nucleobases (Fig. 6), unlike the Endo III superfamily member hOGG1, which strongly discriminates against adenine opposite the lesion and forms several specific contacts with the orphaned base (22).
The variations in the quantum yield of tC O are not straightforward to interpret; it seems to be stacking-dependent in single-stranded DNA but relatively invariable in double-stranded DNA, and, if tC O is located between two G bases, the fluorescence in single-stranded DNA is quenched relative to doublestranded DNA (50). Therefore, we hypothesize that the first decrease in fluorescence in cases of F ligand and AP and dHU substrates may be associated with DNA kinking upon Endo III  (15)) and are alternatively marked in black and red to show the borders. Asterisks indicate identical residues, colons, conserved residues; dots, residues with at least some physicochemical properties conserved. The alignment was produced with Clustal Omega (65).
binding, or with Leu 81 wedging and eversion of the damaged nucleotide from DNA (Fig. 8). Notably, these events seem to be also displayed in the aPu fluorescence traces (Fig. 4, A-C) before the insertion of Gln 41 , indicating that the initial stages of DNA distortion do not require active helix penetration by the ␣B-␣C turn. The shapes of tC O fluorescence traces (Fig. 5, A-D) suggest that the increase in tC O fluorescence upon nonspecific binding may also be associated with partial DNA melting. As can be seen from Tables 2 and 3, for F-and AP-DNA binding k 1 aPu Ͻ k 1 tCo , indicating that this disruption of DNA base pairing proceeds after the formation of initial contacts with Endo III. Such disturbance of DNA conformation in both specific and nonspecific complexes may result from the attempts of the enzyme to flip out the sampled base independently of whether it is damaged or not. Active eversion of undamaged bases from DNA in the process of lesion search was also shown for hOGG1 (23,24). Also, in all available structures of DNA glycosylases with undamaged DNA (23, 55-60) the wedging residue of the enzyme is already inserted in DNA, so Leu 81 of Endo III might be expected to behave in the same way as it was recently shown by single-molecules studies in Refs. 61 and 62. Therefore, the other possible molecular event causing a tC O fluorescence increase in the first phase of nonspecific DNA binding might be insertion of Leu 81 as a part of lesion search, affecting the environment of tC O in fully stacked undamaged DNA in another way than in the case of already destabilized damaged DNA.
The second phase of F ligand binding induced an increase in the tC O fluorescence signal (Fig. 5B). On the other hand, the second phase observed with cleavable AP and dHU substrates (Fig. 5, C and D) led to a decrease in the tC O fluorescence signal. Therefore, this stage is not likely to arise from interactions with the ␣F-␣G motif, because they are identical in the Endo III complex with the F ligand and with the reduced Schiff base intermediate (15). Rather, we suggest that the extended size of the tC O base allows it to partially overlap with the bases in the damaged strand and respond to the structural changes occurring when DNA around the lesion adopts a catalytically competent conformation. A comparison of DNA structure in close vicinity of the lesions in the cases of F and reduced Schiff base (15) indicates that F is pulled deeper into the active site pocket of the enzyme, and the space between the adjacent bases is narrower with F. The relaxation of this DNA distortion upon pre-catalytic conformational adjustment may ease stacking with the protruding part of tC O thus causing a further decrease in its fluorescence.
A comparison of kinetic data obtained by aPu and tC O fluorescence measurements (compare k 3 aPu and k 3 tCo values, Table  2 and 3) together with the lack of additional product accumulation after an alkaline treatment support the conclusion that the ratelimiting step of Endo III catalysis is base excision rather than substrate binding or DNA backbone cleavage. This may be related with a need to adjust the active site conformation to a wide spectrum of oxidatively damaged bases recognized and removed by Endo III (9,11,26,52), precluding the existence of a stiff preformed active site highly specific for one particular substrate.

Conclusion
The kinetic analysis of conformational transitions in DNA ligands and substrates for E. coli Endo III DNA glycosylase, based on stopped-flow fluorescence traces of reporter base analogs aPu and tC O , located in the damaged strain and in the opposite chain, respectively, reveals the dynamics of different regions of DNA around the damaged site during the interaction with Endo III and opens a window into the mechanism of enzyme catalysis (Fig. 8). As in the case of previously studied DNA glycosylases, E. coli Fpg (28,29,31,63) and human OGG1 (20,21), binding of Endo III to DNA begins most likely with an insertion of the wedge amino acids, Gln 41 and Leu 81 , into the helix (61,62). Insertion of other amino acids and eversion of the damaged nucleotide into the active site proceed at the next stage of specific lesion recognition, resulting in "stapling" of the enzyme on DNA. After that the active site is adjusted to the catalytically competent conformation, and the catalytic steps end up in the cleavage of the N-glycosidic bond and ␤-elimination.