Interactions of Human O6-Alkylguanine-DNA Alkyltransferase (AGT) with Short Single-stranded DNAs*

The O6-alkylguanine-DNA alkyltransferase (AGT) repairs O6-alkylguanine and O4-alkylthymine adducts in single-stranded and duplex DNAs. Here we characterize the binding of AGT to single-stranded DNAs ranging in length from 5 to 78 nucleotides (nt). Binding is moderately cooperative (37.9 ± 3.0 ≤ ω ≤ 89.8 ± 8.9), resulting in an all-or-nothing association pattern on short templates. This cooperativity contrasts with the isolated binding seen in recent crystal structures of AGT-DNA complexes. The statistical binding site size S (mean = 5.2 ± 0.1) oscillates with increasing template length. The oscillation period (4.10 ± 0.02 nt/protein) is nearly identical to the binding site size obtained at the highest known binding density (S = 4 nt/protein) and is significantly smaller than the contour length (∼8 bp) occupied in crystalline complexes. A model in which AGT proteins overlap along the DNA contour is proposed to account for these features. Oscillations in intrinsic binding constant Ki and cooperativity factor ω have the same frequency but are of opposite phase to S with the result that the most stable protein-protein and protein-DNA interactions occur at the highest packing densities. We hypothesize that modest binding cooperativity and high binding densities are adaptations that allow AGT to efficiently search for lesions in the context of chromatin remodeling and DNA replication.

DNA repair is crucial for the preservation of cell viability and genetic heredity in the presence of environmental, cellular, and chemical mutagens. The ubiquitous repair protein O 6 -alkylguanine-DNA alkyltransferase 4 (AGT, 5 also called O 6 -methylguanine DNA methyltransferase) plays an essential role in maintaining genomic integrity by repairing O 6 -alkylguanine and O 4 -alkylthymine adducts that form in DNA exposed to alkylating agents (1)(2)(3). Both adducts are mutagenic and carcinogenic (1,4,5), whereas O 6 -alkylguanine adducts are also strongly cytotoxic (6). Ironically AGT also protects cells against chemotherapeutic drugs that methylate or chloroethylate DNA (6,7). Clinical trials of AGT inhibitors are underway in an attempt to increase the efficacy of DNA-alkylating drugs in cancer chemotherapy (8 -10). Despite the interest focused on AGT as a result of its relevance to cancer, relatively little is known about its mechanisms of interaction with DNA and with proteins in its environment.
Human AGT is a monomeric protein (207 amino acids, M r ϭ 21,519) that is expressed constitutively in normal cells (3,7,11). It binds DNA with modest affinity, significant cooperativity, and little sequence specificity (12)(13)(14). In binding it discriminates only weakly between single-stranded and duplex structures (13,14), and it repairs O 6 -alkylguanine lesions in both single-stranded and duplex DNAs (15)(16)(17). In the repair reaction, each protein molecule catalyzes the transfer of a single alkyl group from the O 6 position of guanine or O 4 position of thymine to an active site cysteine (Cys 145 in the human protein). This reaction returns the DNA base to an unmodified state and permanently inhibits the alkyl acceptor activity of the enzyme, 6 which is ultimately degraded (2,18). As a consequence of this life cycle, the number of O 6 -alkylguanine and O 4 -alkylthymine adducts that can be repaired at one time depends on the steadystate cellular concentration of the unalkylated form of AGT (2,3) and on its ability to partition between DNA lesion sites and the vast amount of genomic DNA in which they are embedded.
Three considerations motivate this characterization of AGT binding to single-stranded oligonucleotides. First, the enzyme binds single-stranded DNAs in vitro with affinities close to those observed with duplex DNAs (13,14). Although repair rates have been found to be slower with single-stranded templates than with duplex (19), it remains possible that the repair of single-stranded templates might be a normal function of AGT within the cell. A better understanding of AGT interactions with single-stranded DNAs will contribute to our ability to test the novel possibility. Second, oligonucleotides containing O 6 -alkylguanine offer advantages over the mononucleotide inhibitors currently under clinical trial. Important among these are improved water solubility, greater reactivity with AGT, and efficacy against O 6 -benzylguanine-resistant AGT mutants (20,21). A better understanding of the interactions of the enzyme with single-stranded DNAs may guide the development of therapeutically useful oligonucleotide inhibitors. Finally, recent crystal structures of AGT-DNA complexes depict the isolated binding of enzyme molecules to duplex DNA 7 . This contrasts with the closely packed, cooperative pattern of binding observed in free solution with both single-stranded and duplex templates (Ref. 14 and this work) and suggests that binding in free solution may involve a different ensemble of molecular contacts than is present in the crystal structures. A more complete characterization of cooperative AGT-DNA interactions may help us to account for this difference.

EXPERIMENTAL PROCEDURES
Enzymes and Reagents-T 4 polynucleotide kinase was purchased from New England Biolabs. [␥-32 P]ATP was from ICN Radiochemicals. Electrophoresis grade polyacrylamide was from Fisher. All other chemicals were reagent grade or better.
Human AGT Protein-Recombinant human AGT protein (tagged with His 6 at its C-terminal end) was purified to apparent homogeneity according to published protocols (22). Samples were stored at Ϫ80°C until needed. The purity of the protein was verified by SDS-gel electrophoresis followed by silver staining (23,24). Sedimentation equilibrium data were consistent with a single species with M r ϭ 21,860 Ϯ 400 (result not shown). This value agrees well with previous measurements (13) and is consistent with the value (M r ϭ 21,614) predicted from the sequence of this variant of the protein. The preparations used were Ͼ95% active in transfer of methyl groups from O 6 -[ 3 H]methylguanine-labeled calf thymus DNA to AGT and Ͼ95% active in debenzoylating O 6 -benzylguanine as described previously (25,26). AGT concentrations were measured spectrophotometrically using a molar extinction coefficient, ⑀ 280 ϭ 3.93 ϫ 10 4 M Ϫ1 cm Ϫ1 , calculated from the data of Roy et al. (27). Values of ⑀ 215 /⑀ 280 ϭ 8.2 and ⑀ 260 /⑀ 280 ϭ 0.63 were obtained from UV spectra of the purified protein dissolved in 10 mM Tris buffer, pH 7.6, at 21°C. DNA Substrates-Oligodeoxyribonucleotides of 5,7,9,11,16,22,24,30,41, and 78 residues (sequences shown in Table 1) were synthesized by the Macromolecular Core Facility of the Penn State College of Medicine or were purchased from Operon. These DNA sequences were used previously in studies of DNA binding and/or repair by AGT (see Refs. 13 and 28). In addition, they were selected for a low propensity to form double-stranded secondary structures as predicted by the webbased programs Oligonucleotide Properties Calculator 8 or dnaMATE (29). All DNAs were purified by NENsorb chromatography as directed by the manufacturer. Oligonucleotides of 5, 7, and 9 nt were equilibrated with 10 mM Tris (pH 8.0 at 20°C), 1 mM EDTA buffer by chromatography through short Biogel P2 columns. Longer DNA molecules were dialyzed to equilibrium against the same buffer prior to use. DNA samples for electrophoretic mobility shift analysis were labeled at 5Ј termini with 32 P as described by Maxam and Gilbert (30). Unincorporated [␥-32 P]ATP was removed by buffer exchange using Biogel P2 centrifuge columns pre-equilibrated with 10 mM Tris (pH 8.0 at 21°C), 1 mM EDTA. Single-stranded DNA concentrations were measured spectrophotometrically using extinction coefficients calculated by the nearest neighbor method (31,32). Depending on base composition and sequence, values of ⑀ 260 ranged from 8.45 ϫ 10 3 to 1.04 ϫ 10 4 M Ϫ1 cm Ϫ1 (per base).
Electrophoretic Mobility Shift Assays (EMSAs)-Binding reactions were carried out at 20 Ϯ 1°C in 10 mM Tris (pH 7.6), 100 mM NaCl, 1 mM dithiothreitol, and 10 g/ml bovine serum albumin. Protein-DNA complexes were formed by adding appropriate amounts of AGT to solutions containing 32 P-labeled oligodeoxyribonucleotides. Mixtures were equilibrated at 20 Ϯ 1°C for 30 min. Duplicate samples incubated for longer periods gave identical results, indicating that equilibrium had been attained (result not shown). Electrophoresis was carried out in 10% polyacrylamide gels (acrylamide:N,NЈ-methylenebisacrylamide ϭ 75) cast and run at 8 V/cm in buffer consisting of 10 mM Tris acetate, pH 7.6, supplemented with NaCl to match the conductivity of the protein-DNA samples (33). Autoradiographs were obtained with Eastman Kodak Co. X-Omat Blue XB-1 film exposed at 4°C. Gel segments containing individual electrophoretic species were excised using the developed film 7 The structures obtained by Daniels et al. (22)   as a guide and counted in a scintillation counter by the Cerenkov method. Similar results were obtained by densitometry of appropriately exposed autoradiographic films. The serial dilution method (34,35,47) was used to obtain self-consistent estimates of binding stoichiometry (n) and the association constant (K n ). For the cooperative binding mechanism nP ϩ D^P n D, the association constant is K n ϭ [P n D]/ [D][P] free n . Here P represents AGT protein, D represents DNA, and P n D represents the AGT-DNA complex of protein-DNA ratio n. Separating variables and taking logarithms gives Equation 1. In many cases stoichiometries, association constants, and cooperativity parameters were evaluated by direct titration. Solutions of AGT protein (typically 5 ϫ 10 Ϫ7 M Յ [AGT] Յ 3 ϫ 10 Ϫ4 M) were added directly to [ 32 P]DNA solutions (typically ϳ5 ϫ 10 Ϫ8 M), and samples were analyzed by native gel electrophoresis. Because the total concentration of protein binding sites on DNA was always significantly less than that of the protein, the approximation [P] ϭ [P] free was used. The dependence of binding density on the free protein concentration [P] was given by the McGhee-von Hippel isotherm (36) as modified by Record and co-workers (37) to account for finite lattice size (Equations 2 and 3). ͓P͔ Here is the binding density (protein molecules/nucleotide), K is the equilibrium association constant for binding a single site, is the cooperativity parameter, N is the length of the DNA in nucleotides, and s is the size of the site (in nucleotides) that a protein molecule occupies to the exclusion of others. In this context, the cooperativity parameter is equal to the equilibrium constant for the process of moving a protein from an isolated site to a singly contiguous one or from a singly contiguous site to a doubly contiguous one. Thus, it is a measure of the population-averaged equilibrium constant for interaction between proteins occupying adjacent sites on the DNA.
Analytical Ultracentrifugation-Human AGT protein and oligodeoxyribonucleotides were dialyzed against 10 mM Tris (pH 7.6), 1 mM dithiothreitol, 1 mM EDTA, 100 mM NaCl. Analytical ultracentrifugation was performed at 20 Ϯ 0.1°C in a Beckman XL-A centrifuge using an AN60Ti rotor. Most scans were obtained at 252 nm to minimize protein contribution to the DNA signal (38), although some data sets were obtained at 260 nm. Equilibrium was held to be attained when scans taken 6 h apart were indistinguishable. Typically equilibration times of 24 h met this criterion for AGT-DNA mixtures. Five scans were averaged for each sample at each wavelength and rotor speed.
For DNAs with small numbers of protein interaction sites, cooperative binding can be described by the simple mechanism nP ϩ D^P n D in which free DNA (D) is in equilibrium with saturated complex (P n D), but intermediates with subsaturating protein stoichiometries are not present at significant concentrations. At sedimentation equilibrium, the radial distribution of absorbance for such a system is given by Equation 4.
Here A(r) is the absorbance at radial position r and ␣ P , ␣ D , and ␣ P n D are absorbances of protein, DNA, and protein-DNA complex at the reference position r 0 , and ⑀ is a base-line offset that accounts for radial position-independent differences in the absorbances of different cell assemblies. The reduced molecular weights of AGT protein, DNA, and protein-DNA complexes are given by Here M P and M D are the molecular weights of protein and DNA, n is the protein:DNA ratio of the complex, r is the solvent density, is the rotor angular velocity, R is the gas constant, and T the temperature (kelvin). The partial specific volume of AGT (v P ϭ 0.744 ml/g) was calculated by the method of Cohn and Edsall (39) using partial specific volumes of amino acids tabulated by Laue et al. (40). The partial specific volume of single-stranded NaDNA at 0.1 M NaCl (0.502 ml/g) was estimated by interpolation of the data of Cohen and Eisenberg (41). Partial specific volumes of each protein-DNA complex were estimated using Equation 5.
This relationship is based on the assumption that there is no significant change in partial specific volumes of the components upon association. This approach has been used successfully for other protein-protein and protein-DNA interactions (42)(43)(44), although there are notable exceptions (45). Estimated with Equation 5, v P n D is ϳ0.73 for the complexes described here. Accordingly a 1% error in our estimate of v P n D results in an error in M P n D of nearly 3%. However, a comparison of stoichiometries estimated by sedimentation equilibrium and EMSA methods ( Table 2) suggests that errors in v P n D are unlikely to be large. Equation 4 was used in global analysis of multiple data sets obtained at different macromolecular concentrations and rotor speeds (46). In this method, the values of ␣ P , ␣ D , ␣ P n D , and ⑀ are unique to each sample, but the value of n must be common to all data sets. Terms accounting for non-ideal effects were not included because there was no evidence of non-ideality (see Table 1).

RESULTS
Determination of Stoichiometries-Human AGT binds single-stranded and duplex DNAs with significant cooperativity (13,14). When short single-stranded DNAs are titrated with AGT, the major DNA species present at equilibrium are free DNA and the saturated protein-DNA complex (Fig. 1). The electrophoretic mobilities of these AGT complexes decrease with increasing DNA length (not shown), but protein-DNA stoichiometries cannot be reliably estimated from this effect (33, 35). Accordingly we used sedimentation equilibrium analyses to establish the stoichiometries of the protein-DNA complexes formed with single-stranded DNAs of 5,7,9,11,16,30, and 78 nucleotides (representative data are shown in Fig. 2). In each case the data are fit by the sedimentation equation corresponding to the concerted binding model (nP ϩ D^P n D; Equation 4). The small, symmetrical residuals (upper panels) indicate that this model is consistent with the mass distributions present in these samples. Molecular weights of protein-DNA complexes and free DNAs were obtained as parameters of these fits. 9 The stoichiometry of each complex was inferred from the known molecular weights of the DNA and AGT protein, and the results are summarized in Table 2. The masses and stoichiometries of complexes formed with single-stranded 16-, 30-, and 78-mers were within error the same as values reported previously for analogous complexes formed with full-length human AGT protein prepared without the C-terminal His 6 tag (13). These results confirm our previous conclusion that the presence of the C-terminal His 6 tag has negligible effect on affinity and stoichiometry of AGT-DNA complexes (14). Under comparable solution conditions, increasing DNA length resulted in complexes of higher stoichiometry, indicating that for the DNA concentration range explored here and under conditions of AGT excess stoichiometry is limited by the number of binding sites available on each DNA.
Additional stoichiometry values were determined by native gel electrophoresis (EMSA) using the serial dilution procedure 9 The excellent correspondence of the molecular weights observed for the free DNA species with those predicted from their sequences (summarized in Table 1) establishes that each DNA is single-stranded and demonstrates the absence of significant steric and electrostatic non-idealities under these solution conditions.    Table 2, and values of K n are in Table 3. The agreement of stoichiometry values determined by EMSA with those obtained by sedimentation equilibrium indicates that the same complexes are detected by both methods and demonstrates that dissociation of complexes during electrophoresis is not a significant factor under the conditions that were used here.
Binding Density Depends on DNA Length-The number of nucleotide residues per protein molecule (S) is a measure of binding density that reflects the extent to which the DNA is occupied in a protein-DNA complex. When proteins are packed efficiently on templates of optimal length, S is minimized; when gaps occur in packing or partial length sites are present at DNA ends, S Ͼ S min , and the value of S overestimates the separation between the start points of successive binding sites. Such end effects are particularly severe when DNA templates are short (36,37). We exploited this effect to estimate the size of the site occupied by AGT on single-stranded templates. As shown in Fig. 4A, S oscillates with increasing DNA length (L) for short templates (Յ30 nt), whereas above this length, the separation between values of L that we tested is too great for oscillation to be evident. To determine the underlying period of the oscillation, the relation 10 S ϭ A cos (BL) ϩ C was used to model the dependence of S on L (Fig. 4B). In this equation, A is the amplitude of the oscillation, B is its angular frequency in degrees/nt, and C is an offset equal to the mean value of S. A fit of this relation to the data returned an angular frequency of B ϭ 87.4 Ϯ 0.4°/nt, consistent with models in which successive binding sites are separated by 360/ (87.4 Ϯ 0.4) ϭ 4.1 Ϯ 0.02 nt along the DNA contour. No other angles in the range 20°Յ B Յ 160°yield similarly small values of 2 (Fig. 4C), indicating that there is only one plausible binding site frequency for this system.
The value of S is significantly greater for the 78-nt DNA than for any of the smaller single-stranded DNAs that we tested. Because incomplete occupancy of DNA ends should be equally possible for all DNAs, the less efficient packing of AGT on the 78-nt DNA may reflect the presence of gaps between groups of tightly packed protein molecules or a uniform but less tightly packed structure. Despite less efficient packing, the inclusion of complexes containing 41-and 78-nt DNAs in the frequency analysis returned an angle of 86.9 Ϯ 0.5°in good agreement with the value obtained with shorter templates (result not shown). Thus, although the uniformity of packing may decay with increasing template length, the dominant picture remains one of a protein array in which the average separation between the start of two adjacent binding sites is ϳ4 nt. This spacing is significantly shorter than the ϳ8 bp/protein occupied in crystalline AGT complexes formed with duplex DNA (22,48) and the opti-   FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5 mum site size found for the rate of methyl group transfer from oligonucleotides (7 nt; Ref. 15). As discussed below, our working hypothesis is that binding sites for AGT molecules overlap along the contour of single-stranded DNA.

AGT Interactions with DNA
DNA Length Dependence of K and -Changes in binding density may be accompanied by changes in the intrinsic association constant of AGT with DNA (K i ) or in the cooperativity with which AGT forms multiprotein complexes (). To explore these possibilities we recast our binding data according to the short lattice variant of the McGhee-von Hippel relation (Equations 2 and 3). This relates binding density to free protein concentration [P] in terms of K i , , and the statistical binding site size S. Shown in Fig. 5A are Scatchard plots of data obtained from serial dilution and direct titration experiments. The smooth curves represent fits of Equations 2 and 3 to the data in which values of S were calculated from stoichiometries determined independently (by sedimentation equilibrium or serial dilution as described above). The good correspondence of fits to data indicates that binding models that include neighbor exclusion and positive cooperativity can account for the binding activities in these systems.
The least squares "best" values of K i and obtained as fitting parameters are summarized in Table 3. Tests using fixed values of K i show that K i and are weakly anticorrelated (an increase in one parameter produces a slight decrease in the other). On the other hand, the quality of the fits measured by the 2 parameter decreased significantly if either parameter varied from its most probable value by more than the 95% confidence limits (results not shown). Thus, although the values of these parameters cannot be considered unique, the data quality is sufficient to allow us to parse binding affinity into reasonably well defined estimates of K i and . Values of the product K i ⅐ obtained from analysis of direct titration data are similar to monomer-equivalent association constants, K mono ϭ 1/[P] free, midpoint , calculated from serial dilution data 11 (14). This agreement indicates that both assays are monitoring the same binding processes and argues against a significant systematic bias in either assay. 12 In addition, the values of K n and K mono shown here for binding the single-stranded 16-mer are closely similar to values reported previously for the binding of wild-type and C145A mutant human AGT proteins to the same DNA under the same buffer conditions (14). This supports our conclusion that the presence of the C-terminal His 6 tag on the protein used in the current studies has a negligible effect on the affinity of human AGT for this standard DNA template.
Values of K i and oscillate in phase with each other as DNA length increases (Fig. 5B). The fact that these oscillations are significantly larger than the confidence limits associated with individual parameters and the fact that they are in phase argue that this oscillation is not a consequence of fitting correlation between K i and (as discussed above, K i and are weakly anticorrelated). Thus, the orientations of AGT proteins and DNA that maximize K i also position each protein to interact optimally with neighboring AGT molecules. Intriguingly, oscillations in K i and have similar periods but are of opposite phase to the oscillation of S with DNA length. As a result, the most compact complexes (with smallest S) have relatively stronger protein-DNA and protein-protein interactions than do less compact assemblies with similar stoichiometries. This suggests that AGT complexes formed with single-stranded DNA are somewhat malleable so that when packing density is not constrained at the maximum value (giving ϳ4 nt/protein) rearrangements may occur that increase configurational entropy at the expense of slightly suboptimal protein-DNA and/or protein-protein contacts. This interpretation is supported by the decay of the amplitudes of oscillations in K i and with increasing template length (Fig. 5B). Such decay might be expected if packing interactions become increasingly degenerate with template length.

DISCUSSION
The data presented here are consistent with a cooperative binding model in which AGT molecules bind single-stranded DNA at ϳ4-nt intervals. Because a single molecule of AGT occupies ϳ8 bp of DNA duplex (22,48), it seems likely that AGT molecules overlap along the DNA contour 13 as shown schematically in Fig. 6. Overlapping binding provides opportu-  a Estimated using the relation Ϫn log͓P͔ free ϭ log K n obtained From Equation 1 when log͓P n D͔/͓D͔ ϭ 0 (61). The range reflects 95% confidence limits in stoichiometry (n) returned by these analyses. b The range reflects 95% confidence limits. c Obtained by propagating 95% confidence limits of K i and .
nities for protein-protein interactions that may contribute to cooperativity and accounts for the efficient DNA-dependent cross-linking of AGT molecules that we have recently observed. 14 Modestly cooperative binding like that described here (37.9 Ϯ 3.0 Յ Յ 89.8 Ϯ 8.9) is sufficient to account for the range of binding site sizes (4 nt Յ S Յ 9 nt) previously reported for saturated complexes formed with single-stranded DNAs (13,14,17). On short DNAs (n Յ 40 nt), values in this range are sufficient to produce tight packing that results in oscillation in the binding site size with minima when the number of nucleotides is an integral multiple of 4. On longer DNAs, this cooperativity is not sufficient to suppress binding degeneracy so the statistical binding site size gradually increases and oscillations in K i and decay toward average values (Fig. 5).
AGT is not unique in its reactivity toward single-stranded DNA. Other DNA repair proteins and factors that bind and/or repair single-stranded DNA include the human and vaccinia virus uracil-DNA glycosylases (49,50), human DNA glycosylases NEIL1 and NEIL2 (51), human apurinic/apyrimidinic endonuclease (APE1) (52), and Xeroderma pigmentosum group A correcting protein (XP-A) (53,54). In addition, XP-A is associated with the replication protein A single strand-binding protein (55,56), whereas components of the transcription-coupled nucleotide excision repair complex are associated with XP-B and XP-D DNA helicases (57). It seems reasonable to expect that these interaction partners increase the availability of single-stranded templates for repair. Although no such association has been found to date for AGT, this may be because the interactions are too weak for detection by current methods. Alternatively single-stranded DNA may be sufficiently available for AGT binding, making association with helicases or single strand-binding proteins unnecessary for its function.
The cooperative binding mechanism described here for single-stranded DNA and reported previously for duplex DNA (13,14) contrasts with the independent binding shown in currently available crystal structures 7 of AGT-DNA complexes (22,48). At present we cannot account for this difference. To date, we have tested the binding of wild-type human AGT over a range of temperature (4°C Յ T Յ 30°C), salt concentrations (0 M Յ [KCl] Յ 0.4 M), magnesium ion concentrations (0 mM Յ [Mg 2ϩ ] free Յ 10 mM), and zinc availability (Zn 2ϩ -depleted and Zn 2ϩ -saturated proteins and Zn 2ϩ -saturated protein with 0 mM Յ [Zn 2ϩ ] free Յ 10 mM). We have tested Cys 145 -methylated and -benzoylated proteins, active site (Cys 145 ) mutations, both C-terminal and N-terminal His 6 -tagged proteins, and a range of single-stranded and duplex templates. Under all conditions, significant binding cooperativity was observed (14,26,58). We conclude that cooperative binding is a bona fide activity of human AGT.
What do these results tell us about the substrate interactions of AGT? Within our sample set, variations in K i ⅐ with changes in substrate identity are small, indicating that binding affinity is not strongly dependent on base composition or sequence. These results are consistent with previous observations based on binding competition assays (13) and the relative sequence independence of DNA repair efficiency (16). The absence of a dependence on base composition or sequence is likely to ensure that repair activity is uniformly available to all sequences. 14 M. Fried, unpublished result. Cooperative binding may be an important part of the DNA repair mechanism of AGT. In the current results, the most compact complexes have the strongest protein-DNA and protein-protein interactions. Overlapping binding concentrates repair activity at a higher density on the DNA contour than would be available in non-overlapping binding modes. With a new binding site starting every 4 nt, only modest protein displacement is needed for surveillance of every nucleotide in the complex. Together these results suggest that compact, cooperative binding may be part of an efficient lesion search and repair mechanism. Finally although the moderate cooperativity detected here (37.9 Ϯ 3.0 Յ Յ 89.8 Ϯ 8.9) is insufficient to extend high density binding over hundreds of residues of DNA, it can generate densities approaching one protein every 4 residues for sections of template Յ50 residues long. Duplex regions of this size are likely to become available between nucleosomes during chromatin remodeling, and single-stranded segments may be transiently available within transcription or DNA replication complexes. The processive movements of these complexes may allow systematic surveillance of large parts of the genome by AGT. In addition, repair just upstream of a replication fork is likely to represent the last opportunity to prevent the conversion of a promutagenic lesion into an actual mutation. Because of its proximity to the replication fork, such last ditch repair may be limited to single-stranded templates.
The results presented here raise a number of questions for future investigation. First, is the mechanism of cooperative binding the same on singlestranded and duplex DNAs? In vitro measurements indicate that both secondary structures are bound with positive cooperativity and remarkably similar affinities and binding densities 15 (13,14). These results are intriguing because single-stranded and duplex DNAs differ greatly in torsional rigidity. The relative flexibility of single-stranded DNA may allow cooperatively bound AGT proteins to form identical protein-DNA contacts at a binding density of one protein every 4 nucleotides. A comparable density of one protein every 4 base pairs will not result in identical protein contacts with the more rigid B-form duplex unless each protein is rotated approximately (4 bp⅐protein Ϫ1 / 10.4 bp⅐turn Ϫ1 ) ϫ 360°⅐turn Ϫ1 ϭ 138.4°⅐protein Ϫ1 with respect to its immediate neighbor (similar rotations are shown schematically in Fig.  6). If protein-protein contacts constrain the geometry of the cooperative complex to values other than ϳ138°/ binding step, the cooperative assembly might cause local DNA unwinding (at rotation angles Ͻ138°/binding step) or overwinding (at rotation angles Ͼ138°/binding step).
Second, is the mechanism by which damaged bases enter the active site similar for single-stranded and duplex templates? Torsional stress exerted on duplex DNA may influence the extrusion of alkylguanines from the base stack to form the extrahelical conformation seen in crystalline repair complexes (22,48). Such effects would not be available on single-stranded DNA and may account for the more rapid repair of duplex DNAs than single-stranded DNAs in vitro (19).
Third, how do cooperative interactions influence the rates of AGT binding and dissociation from target DNAs? We have shown previously that AGT is monomeric in the absence of DNA over a wide range of solution conditions (13,14). This result argues against the maintenance of cooperative proteinprotein contacts once AGT proteins have dissociated from DNA and against a role for multimeric protein complexes in DNA binding. Independently of whether AGT binds and dissociates from DNA as a monomer, the portion of a cooperative array of bound proteins that is active in protein binding and/or dissociation is likely to influence the kinetics of these processes. Our current thinking is influenced by the notion that the addition or removal of a protein unit from the middle of a cooperative assembly is likely to be slow compared with the addition or removal of a unit from one end. If this is the case, the rate of 15 For example, S ϭ 4 nt/protein on single-stranded 16-mer DNA and S ϭ 4 bp/protein for the corresponding 16-mer duplex (14). Forward titrations were carried out as described under "Experimental Procedures," and binding was detected and quantitated by electrophoretic mobility shift assay. Each data set is a composite derived from two or three independent titrations. The smooth curves correspond to non-linear least squares fits of Equations 2 and 3 to the data using fixed values of S determined by equilibrium sedimentation or serial dilution EMSA. The values of K and obtained as parameters of these fits are compiled in Table 3. B, dependence of S, , and K on DNA length (L). The data for the dependence of S on L are a subset of that shown in Fig. 4A. The data for the dependence of and K on DNA length are from the experiments shown in A. The error bars correspond to 95% confidence limits estimated for each parameter. The oscillations of and K with increasing L appear to have the same period but opposite phase as those of S.
transfer of AGT molecules between DNA segments may depend more strongly on the concentration of cooperative assemblies (and hence the concentration of end monomers) than on the concentration of DNA-bound AGT monomers themselves. Experiments designed to test these ideas are underway. Finally, do alkyl transfer rates depend on the length of the cooperative complex? Repair requires the correct juxtaposition of an active AGT monomer and the damaged base. If the correct juxtaposition of AGT and lesion is not achieved during formation of the protein-DNA complex, achieving it may require repositioning of the entire protein array. If this is the case, productive binding and repair might be more rapidly achieved with short protein arrays than with long protein arrays. Similarly do alkyl transfer rates depend on substrate length as seen for the binding parameters K i and ? If so, oligonucleotides with lengths that are multiples of 4 nt will be more efficient substrates than oligonucleotides of intermediate lengths. These factors may contribute to the development of clinically useful oligonucleotide inhibitors of AGT. Proteins are shown schematically as filled ovals. The DNA is represented by a black rod. A, non-overlapping, adjacent binding. The minimum value of the statistical binding site size S is the number of nucleotides occupied by each AGT monomer to the exclusion of other AGT monomers. B, overlapping binding viewed from perpendicular to the DNA axis. Two proteins bind different surfaces of the same DNA segment, and the minimum value of S is the number of nucleotides between the first residue in one binding site and the first residue in the next site. Our frequency data do not specify the rotational sense of the assembly. A right-handed complex is shown for purposes of illustration. C, overlapping binding viewed along the DNA axis. The DNA is represented by the black circle in the center. Three proteins, represented by gray, magenta, and blue ovals, are bound with a rotational offset angle of ␣ ϳ120°. A right-handed complex is shown for purposes of illustration.