Flipping Duplex DNA Inside Out

The Escherichia coli MutY adenine glycosylase plays a critical role in repairing mismatches in DNA between adenine and the oxidatively damaged guanine base 8-oxoguanine. Crystallographic studies of the catalytic core domain of MutY show that the scissile adenine is extruded from the DNA helix to be bound in the active site of the enzyme (Guan, Y., Manuel, R. C., Arvai, A. S., Parikh, S. S., Mol, C. D., Miller, J. H., Lloyd, S., and Tainer, J. A. (1998) Nat. Struct. Biol. 5, 1058–1064). However, the structural and mechanistic bases for the recognition of the 8-oxoguanine remain poorly understood. In experiments using a single-stranded 8-bromoguanine-containing synthetic oligodeoxyribonucleotide alone and in a duplex construct mismatched to an adenine, we observed UV cross-linking between MutY and the 8-bromoguanine probe. We further observed enhanced cross-linking in the single strand experiments, suggesting that neither the duplex context nor the mismatch with adenine is required for recognition of the 8-oxoguanine moiety. Stopped-flow fluorescence studies using 2-aminopurine-containing oligodeoxyribonucleotides further revealed the sequential extrusion of the 8-oxoguanine at 108 s−1followed by the adenine at 16 s−1. A protein isomerization step following base flipping at 1.9 s−1 was also observed and is postulated to provide additional stabilization of the extruded adenine thereby facilitating its capture by the active site for excision.

The effect of cellular damages by reactive oxygen species in carcinogenesis and aging is well documented (1)(2)(3). Even in the absence of external oxidative stress, normal metabolic processes produce oxidative damages to DNA (4 -6) requiring repair. Oxidative damage of DNA bases alters their base pairing properties (5, 7) thereby interfering with replication and transcription (8). A predominant lesion found in DNA exposed to reactive oxygen species is 8-oxoguanine, which is especially deleterious due to its ability to form a stable Hoogsteen base pair with adenine in addition to the canonical Watson-Crick base pair with cytosine (9,10). The facile by DNA polymerase misincorporation of an adenine across from the 8-oxoguanine (11) results in a mutagenic adenine:8-oxoguanine mismatch, a site where further replication prior to repair would lead to C 3 A or G 3 T transversions. In Escherichia coli, the MutY ade-nine glycosylase (MutY) 1 plays a critical role in preventing mutations stemming from oxidative damages to DNA by excising the adenine from the adenine:8-oxoguanine mismatch.
Like all DNA-nucleotide-modifying enzymes, including DNA methylases, base-excision repair glycosylases, and endonucleases (12)(13)(14)(15), MutY faces the 2-fold task of recognizing and accessing chemical moieties on DNA bases hidden within the double helix of duplex DNA. These enzymes have evolved an elegantly simple strategy for exposing their targets by rotating the phosphodiester bonds surrounding the nucleotide, causing the target base to be flipped out of the DNA helix (16 -21).
Using the E. coli uracil-DNA glycosylase, we have recently reported the first kinetic mechanism of a base-flipping enzyme, defining for the first time the correct temporal sequence of events at the enzyme active site during catalysis (22). Following initial nonspecific binding, the DNA backbone is scanned to locate the site. A protein-induced distortion of the DNA helix at the target site causes the target nucleotide to become mobile, capable of rapid and reversible extrusion from the DNA duplex. The efficient capture of the extruded base, however, requires a protein conformation change, which inserts the side chain of Leu 191 into the DNA duplex at the site vacated by the extruded base. This ratchet-like movement by the protein prevents the flipped-out base from returning into the DNA duplex, thereby driving it into the active site of the enzyme to be excised.
Crystallographic studies on the catalytic core domain of MutY reveals an active site binding pocket for an extruded adenine (23), suggesting that MutY also uses a similar baseflipping reaction mechanism. However, MutY is unique among base-excision repair enzymes in recognizing a mismatch between a damaged 8-oxoguanine and a normal adenine while exclusively catalyzing the removal of the undamaged base (24 -26). Unlike other base-flipping enzymes, the recognition target of MutY must extend beyond the adenine being excised; thus, the characteristic base-flipping reaction mechanism where the extruded base is also the site of the chemical reactions cannot fully explain the mode of target recognition by MutY.
Biochemical evidence shows that the C-terminal domain missing from the crystal structure of the catalytic core domain is required for 8-oxoguanine recognition (27)(28)(29). NMR studies of this missing domain reveal significant structural homologies with the known structure of MutT (29 -31), an enzyme that hydrolyzes deoxy-8-oxoguanosine 5Ј-triphosphate. As MutT uses a binding pocket to recognize the 8-oxoguanine portion of its substrate (32), the presence of an analogous pocket has been postulated for the C-terminal domain of MutY (29,30). If such a cleft were to exist, then the recognition of an 8-oxoguanine by MutY would likely require the flipping of the non-scissile 8-ox-oguanine out of the DNA helix in addition to the expected extrusion of the adenine for excision. However, no direct biochemical or structural evidence currently exists for such a double base-flipping mechanism.
Here we report the results of UV cross-linking experiments using 8-bromoguanine-containing oligodeoxyribonucleotides to probe the topological nature of the structural contact between MutY and 8-oxoguanine. In addition, we also show results of real time stopped-flow fluorimetric detection of both 8-oxoguanine and adenine base flipping. These results will be discussed in the context of the existing structural and mechanistic models of base-flipping enzymes to provide direct insight into the mechanism of target recognition and selectivity by MutY.
Oligodeoxyribonucleotides and Enzymes-Synthetic oligodeoxyribonucleotides were either synthesized by Gene Link (Hawthorne, NY) or by the Central Laboratory Services of the Center for Gene Research, Oregon State University and further purified by urea-polyacrylamide gel electrophoresis and electroelution as described in Bao et al. (33). Concentrations were determined spectrophotometrically using extinction coefficients calculated according to Cantor et al. (34). The 21nucleotide-long single-stranded oligodeoxyribonucleotides are listed in Table I with their nucleotide sequences. Oligodeoxyribonucleotides were 5Ј-32 P-labeled using T4 polynucleotide kinase (USB Corp.) and [␥-32 P]ATP (3000 Ci/mmol, Amersham Biosciences) as described previously in Bao et al. (33).
The E. coli MutY was purified from E. coli BL21(DE3) harboring the overproducing plasmid pET24a/MutY-8, which contained a copy of the mutY gene obtained by Pfu-catalyzed PCR of E. coli BL21(DE3) inserted between the NdeI and the BamHI restriction sites of pET24a. Overproduced MutY was purified by a protocol adapted from published methods (35, 36) using a combination of cation exchange on a Bio-Rad Hi-S column and affinity chromatography on a double-stranded DNAcellulose column. The MutY concentration was determined spectrophotometrically using an extinction coefficient of ⑀ 280 ϭ 77,510 M Ϫ1 cm Ϫ1 . Purified MutY had an A 400 to A 280 ratio between 0.15 and 0.19 and was stored in MutY Storage Buffer at Ϫ80°C.
UV Cross-linking with 8-Bromoguanine-Samples (100 l) containing 1 M MutY and 0.5 M DNA in Buffer Y were irradiated at 20°C for 40 min in a 96-well microtiter plate 5 mm below a UVP Model UV-LMS-38 lamp (Upland, CA) set to 302 nm and quenched directly into SDS gel loading buffer. Radiolabeled cross-linked products were separated from uncross-linked DNA by SDS-PAGE on a 12% acrylamide gel and visualized using a Molecular Dynamics PhosphorImager (Amersham Biosciences). Handling of 8-bromoguanine-containing DNA prior to UV irradiation was carried out under reduced illumination provided by a 25-watt red incandescent light bulb using amber-colored microfuge tubes.
Single-turnover Excision Assays-Single-turnover experiments were performed using a KinTek RFQ-3 Rapid Chemical Quench (KinTek Instruments, State College, PA) instrument maintained at a constant 37°C with a Neslab RTE-111 refrigerated water bath. Reactions were initiated by rapidly mixing 15 l of MutY with 15 l of 5Ј-32 P-labeled duplex DNA substrate and were chemically quenched with 90 l of 0.2 M NaOH. The abasic site-containing excision products were then heated to 90°C for 8 min to cleave the phosphoribose backbone of the DNA at the abasic site to generate 10-nucleotide products that were separated from the 21-nucleotide substrates by electrophoresis on a 20% acrylamide, 8 M urea gel. Following visualization with a Molecular Dynamics PhosphorImager (Amersham Biosciences), the intensities of DNA substrate and product bands were quantified using ImageQuant software (Amersham Biosciences) as described in Bao et al. (33) and Wong et al. (22). Single-turnover time courses with enzyme present in excess over substrate were fitted to a single exponential function, Stopped-flow Assays-Real time fluorescence changes were measured using a pneumatically driven KinTek SF 2001 stopped-flow spectrophotometer (KinTek Instruments) fitted with a 100-watt mercury arc lamp (OSRAM HBO 103 W/2). Reactions were maintained at a constant 37°C with a Neslab RTE-111 refrigerated bath. The 2-aminopurine was excited at ex ϭ 313 nm, and its fluorescence emission was monitored at Ͼ350 nm using a Thermo Corion LG-350-F filter (Franklin, MA). Time courses were fitted by nonlinear least-square regression to a sum of exponential terms, In the case of the tPAT:bAOT, where the time course required a lag period, k 1 of a triple exponential function, n ϭ 3, was constrained at 110 s Ϫ1 , whereupon the regression analysis would then return a small negative best-fit value of A 1 to account for the lag.

RESULTS AND DISCUSSION
UV Cross-linking via 8-Bromoguanine-To directly detect structural contacts between MutY and 8-oxoguanine, we synthesized the 21-nucleotide bABT with a photoreactive 8-bromoguanine (B) substituted for the 8-oxoguanine (see Table I). Cross-linking experiments were carried out with either 5Ј-32 Plabeled single-stranded *bABT or the duplex tAAT:*bABT containing an adenine:8-bromoguanine mismatch. Single-turnover adenine excision assays performed with tAAT:bABT showed only a ϳ2-fold reduction in the rate of excision (data not shown) relative to a "normal" adenine:8-oxoguanine mismatched duplex, tAAT:bAOT, validating the suitability of the 8-bromoguanine substitution in these cross-linking studies.
Cross-linked products formed after irradiation at 302 nm for 40 min were resolved by SDS-polyacrylamide gel electrophoresis. Fig. 1 shows the results of experiments carried out at 1 M MutY and 0.5 M DNA, conditions under which active site titration experiments indicated near maximal productive binding of substrate DNA (data not shown). A distinct radiolabeled band consistent with DNA-cross-linked MutY appeared with either single-stranded *bABT (Lane 4) or duplex tAAT:*bABT (Lane 8) with the former showing moderately higher efficiency of cross-linking. The overall extent of cross-linking (Ͻ10%) was poor partly due to the non-optimal wavelength of irradiation. Significantly better cross-linking was observed with irradiation at 254 nm; however, we chose to cross-link at the longer wavelength to avoid unnecessary damage to the DNA and protein as well as to avoid non-8-bromoguanine-induced, nonspecific DNA cross-linking. Control experiments with 1 M bovine serum albumin (Lanes 1 and 5) or no MutY (Lanes 2 and 6) did not  show cross-linking nor did samples quenched prior to irradiation 2 (Lanes 3 and 7).
UV cross-linking demonstrated direct contact between MutY and the 8-bromoguanine during catalysis. In a duplex DNA substrate where this base is hidden in the DNA helix, such contact could occur either from the insertion of a part of the enzyme into the helical space or by the extraction of the base from the helix into a binding pocket on the enzyme surface. Therefore, the observation of cross-linking with the duplex construct alone does not constitute compelling evidence of flipping of the 8-substituted guanine. However, the observation of cross-linking with the single-stranded DNA, where the rigid structural determinants of the duplex and the scissile adenine are absent, implies that the recognition of the 8-substituted guanine requires neither the prior binding of duplex DNA nor the extrusion of adenine. In addition, the increased cross-linking observed with single-stranded DNA suggests that the site of cross-linking on MutY is more readily accessible to a "free" 8-bromoguanine residue than one hidden within the helical space of duplex DNA. Conversely if contact were to occur with a protein side chain inserted into the DNA duplex to probe for the 8-substituted-guanine, then less cross-linking would be expected with the single-stranded construct where the "binding site" provided by the helical cage of the duplex is absent. 3 Consequently these results support the hypothesis of a preexisting 8-oxoguanine binding pocket on the enzyme surface as proposed based on the structural homology to MutT (29 -31).
Stopped-flow Detection of Double Base Flipping-To directly monitor the extrusion of 8-oxoguanine from the DNA duplex, we synthesized an 8-oxoguanine:adenine mismatched substrate with the fluorescent base analog 2-aminopurine (P), positioned 5Ј-adjacent to the 8-oxoguanine (O), tAAT:bPOT. The reduction of quantum yield of 2-aminopurine from basestacking interactions (37) has been widely exploited in studies of DNA-metabolizing enzymes (22, 38 -41). Therefore, we expected to observe an enhancement of the fluorescence intensity of the 2-aminopurine probe in response to the loss of basestacking interaction upon extrahelical extrusion of its neighboring 8-oxoguanine. Similarly a homologous duplex substrate, tPAT:bAOT, with the 2-aminopurine probe adjacent to the scissile adenine was also synthesized to monitor adenine flipping. Single-turnover adenine excision assays at 500 nM MutY and 250 nM DNA showed identical time courses for the fluorophore-containing duplex substrates *tAAT:bPOT and *tPAT: bAOT and the non-fluorescent, normal substrate *tAAT:bAOT (Fig. 2). Nonlinear regression best fit of the data yielded a rate constant for excision for all three substrates at the active site of 0.25 Ϯ 0.03 s Ϫ1 in agreement with reported values (26,29), validating the suitability of these fluorescent duplexes as substrate analogs in quantitative kinetic measurements. Fig. 3a shows the real time fluorescence emission of the 2-aminopurine-containing tAAT:bPOT (250 nM) within 40 ms of mixing with 500 nM MutY in a stopped-flow fluorimeter. On this time scale, we observed a rapid, single-exponential increase in fluorescence with a rate constant of 108 Ϯ 13 s Ϫ1 . With the 2-aminopurine probe positioned next to the 8-oxoguanine, this experiment directly demonstrates 8-oxoguanine base flipping at 108 s Ϫ1 under these reaction conditions. In a similar experiment performed under identical conditions using a substrate where the fluorophore was positioned to monitor adenine flipping (tPAT:bAOT), the observed fluorescence enhancement appeared later, slower, and with multiple exponential phases. Fig. 3b shows split time-based stopped-flow traces for adenine flipping (top trace), 8-oxoguanine flipping (middle trace), and a negative control (bottom trace). The time course for tPAT:bAOT corresponding to adenine flipping occurred with an initial 30-ms lag followed by two distinct exponential increases. Comparison with the initial 30 ms of the tAAT:bPOT time course shows that the initial lag observed with adenine flipping coincided with 8-oxoguanine flipping, indicating that the extrusion of the adenine follows 8-oxoguanine flipping. The subsequent biphasic increase in fluorescence further reflects a two-step process for adenine extrusion with an initial extrusion of the adenine with an apparent rate constant of 16 Ϯ 1.2 s Ϫ1 followed by a subsequent step at 1.9 s Ϫ1 . In a negative control experiment using a duplex containing a non-cognate 8-oxoguanine: cytosine base pair, tACT:bPOT, no fluorescence change was detected (Fig. 3b, bottom trace).
The large amplitude of the 1.9 s Ϫ1 exponential phase implicated a large favorable forward equilibrium for this step. However, an 8-fold smaller excision rate constant of 0.25 s Ϫ1 would rule out this step as being excision. In the paradigmatic baseflipping kinetic mechanism of the E. coli uracil-DNA glycosylase, a conformational change step following base flipping corresponds structurally to the insertion of a leucine side chain of the enzyme into the DNA helix to occupy the space vacated by the extruded base (22). The leucine insertion isomerization conferred additional stabilization for the "flipped-out" conformation of the extruded base by preventing its return into the DNA double helix. By analogy, the 1.9 s Ϫ1 exponential phase likely reflects a similar isomerization step in the reaction mechanism of MutY to facilitate the capture of the extruded adenine in preparation for excision.
Double Base-flipping Model for Target Selection-Identical rate constants were observed irrespective of the placement of the 2-aminopurine to detect 8-oxoguanine or adenine flipping consistent with the sequential three-step mechanism shown in  Ϫ1 . An isomerization step is proposed to account for the third exponential term observed of 1.9 s Ϫ1 based on analogy with the reaction mechanism of the E. coli uracil-DNA glycosylase. This isomerization would provide additional stabilization of the flipped-out adenine by preventing its return to the DNA helix. In the case of the E. coli uracil-DNA glycosylase, this is accomplished via the insertion of a leucine side chain into the DNA helix to occupy the space vacated by the flipped-out base (19,22). requiring the 8-oxoguanine to be flipped before the adenine, the enzyme ensures that only those adenines mispaired to it are targeted for excision. Similarly the stabilization provided by the isomerization step following the extrusion of both bases of the mismatch ensures that only base pairs containing both 8-oxoguanine and adenine bases become captured with commitment toward catalysis. Accuracy in target selection is thereby achieved as each step of the mechanism functions to trigger the forward progress toward catalysis while coordinately stabilizing the previous steps.
In addition, increased efficiency is achieved in searching for a scissile adenine. By sequentially coordinating the selection of the 8-oxoguanine and the adenine bases, the double base-flipping mechanism establishes a hierarchical order for the search process where the much rarer 8-oxoguanine constituent is first targeted. Interestingly the 8-oxoguanine of this mismatched base pair is readily discernable by the syn conformation of its glycosidic bond as illustrated in structural studies (42), and biochemical evidence further suggests that MutY can recognize the syn conformation of the 8-oxoguanine base in this context (43). MutY therefore likely takes advantage of this conformational feature in its scan along the DNA backbone to trigger the double base-flipping mechanism upon location of an 8-oxoguanine:adenine lesion. Such a hierarchically ordered search model would greatly enhance the efficacy of the search by mitigating the need to extrude and examine each individual base in the duplex DNA.
The reaction rate constants reported here represent macroscopic rate constants observed at 250 nM DNA and 500 nM MutY. The resolution of the microscopic forward and reverse rate constants for each elementary step must await the completion of concentration dependence studies currently underway, although preliminary results suggest that the reaction conditions used were near saturation. However, the design and use of the 2-aminopurine-containing base-flipping-sensitive substrate pair tAAT:bPOT and tPAT:bAOT unambiguously demonstrates 8-oxoguanine flipping. In addition, by establishing the temporal sequence of the substrate recognition steps leading up to catalysis, we were able to observe the coordination of the mechanistic interplay between steps along the catalytic pathway responsible for substrate recognition. These stopped-flow studies not only provide direct kinetic detection of key structural intermediates, but more importantly, by placing these structures in their proper temporal order, they allow us to deduce the dynamic mechanistic basis by which this class of diverse DNA-metabolizing enzymes (13) recognizes and gains access to a naturally hidden target.