Replication acrossO6-Methylguanine by Human DNA Polymerase βin Vitro

Replication in vivo across unrepaired O6-methylguanine (m6dG) lesions by mammalian DNA polymerase β (pol β) during short patch repair may contribute to the cytotoxicity and mutagenesis of m6dG. We have employed in vitro steady state kinetic analysis to investigate the replication of oligonucleotide templates containing site-specific m6dG by human pol β. Our results show that m6dG is a strong but not absolute block to replication by pol β. pol β exhibits mixed kinetic discrimination during overall replication across dG and m6dG. pol β preferentially inserts dTMP rather than dCMP opposite m6dG. However, pol β extends from the dC-m6dG base pair more efficiently than from the dT-m6dG base pair. This is in strong contrast to other polymerases such as the exonuclease-deficient Klenow fragment of Escherichia coli DNA polymerase I (exo−KF) that preferentially extends dT-m6dG by a factor of 10 over dC-m6dG. When both insertion and extension are considered, pol β has a 15-fold overall preference for incorporation of the mutagenic substrate dTTP rather than the nonmutagenic substrate dCTP during replication across m6dG. This suggests that pol β, in concert with the T:G-specific thymine DNA glycosylase, may be intricately involved in the futile cytotoxic repair induced by m6dG. Our results also suggest that replication across m6dG by pol β may contribute to m6dG-induced G → A transition mutations.

Replication in vivo across unrepaired O 6 -methylguanine (m 6 dG) lesions by mammalian DNA polymerase ␤ (pol ␤) during short patch repair may contribute to the cytotoxicity and mutagenesis of m 6 dG. We have employed in vitro steady state kinetic analysis to investigate the replication of oligonucleotide templates containing site-specific m 6 dG by human pol ␤. Our results show that m 6 dG is a strong but not absolute block to replication by pol ␤. pol ␤ exhibits mixed kinetic discrimination during overall replication across dG and m 6 dG. pol ␤ preferentially inserts dTMP rather than dCMP opposite m 6 dG. However, pol ␤ extends from the dC-m 6 dG base pair more efficiently than from the dTm 6 dG base pair. This is in strong contrast to other polymerases such as the exonuclease-deficient Klenow fragment of Escherichia coli DNA polymerase I (exo ؊ KF) that preferentially extends dT-m 6 dG by a factor of 10 over dC-m 6 dG. When both insertion and extension are considered, pol ␤ has a 15-fold overall preference for incorporation of the mutagenic substrate dTTP rather than the nonmutagenic substrate dCTP during replication across m 6 dG. This suggests that pol ␤, in concert with the T:G-specific thymine DNA glycosylase, may be intricately involved in the futile cytotoxic repair induced by m 6 dG. Our results also suggest that replication across m 6 dG by pol ␤ may contribute to m 6

dG-induced G 3 A transition mutations.
Eukaryotic DNA polymerase ␤ (pol ␤) 1 is a nuclear DNA repair polymerase extensively characterized by Wilson (1,2). It mainly carries out gap filling during short patch base excision repair (3)(4)(5)(6)(7). pol ␤ fills short (up to six nucleotides long) DNA gaps (4 -8) and has been found to act in concert with uracil glycosylase (3) and the human G:T-specific thymine glycosylase (9). pol ␤ is also implicated in the repair of DNA damage induced by anticancer agents such as bleomycin, ␥-irradiation (10), cisplatin (11), and alkylating agents (12). Resistance to cisplatin may be partially due to overexpression of pol ␤ (11). Tumors may also develop resistance to radiation therapy and chemotherapy (e.g. cisplatin and alkylating agents) by increased efficiency of DNA repair. pol ␤ has been found to be altered in some human cancers (13,14), suggesting its importance in maintaining genomic stability. However, pol ␤ is the most error prone of all known polymerases tested in vitro (15).
Mechanistic studies of pol ␤ are important for understanding its diverse roles in DNA replication and repair under circumstances leading to mutagenesis, cytotoxicity, tumor progression, and the resistance of tumors to anticancer therapy. Furthermore, the small size, single subunit composition, availability of crystal structure, and absence of confounding associated activities (such as 3Ј-5Ј exonuclease) make pol ␤ a good model for the study of DNA polymerase structure and function.
O 6 -Methylguanine (m 6 dG) is a mutagenic and cytotoxic DNA adduct that can be formed in vivo by such diverse agents as tobacco smoke, methylnitrosourea, and other S n 1 methylating agents (16). In vitro studies (17)(18)(19), as well as in vivo mutagenesis assays (20,21) have shown that m 6 dG preferentially base pairs with dTMP instead of dCMP, thus giving rise to G to A transition mutations. The importance of these lesions has been repeatedly demonstrated. A strong correlation was found between the persistence of m 6 dG lesions and tumors in rodents (22). More recently, the persistence of m 6 dG after treatment with 4-(methylnitrosoamino)-1-(3-pyridyl)-1-butanone (a component of tobacco smoke) has been found to correlate with the activation of the Ki-ras oncogene in lung tumors in mice (23). Introduction of m 6 dG into codon 12 of a synthetic c-Ha-ras gene and transfection into normal NIH-3T3 cells resulted in focus formation and the production of G to A mutations at the position of the m 6 dG (24).
The m 6 dG lesion can be readily repaired in a saturable manner by the suicide enzyme O 6 -methylguanine-DNA methyltransferase (MGMT) (25). This type of repair is error free and noncytotoxic. However, in many human solid tumor cell lines and in some non-tumor tissues the ability to repair m 6 dG is lacking due to the inactivation of MGMT (25)(26)(27)(28)(29), thereby favoring the persistence of unrepaired m 6 dG lesions. Persistence of these lesions is associated with methylation-dependent cytotoxicity. Methylating agents that form m 6 dG in vivo also induce sister chromatid exchanges (29), which are thought to result from persistent gaps (30). The cytotoxicity of unrepaired m 6 dG in eukaryotic cells may be related to its replication blockage and a type of mismatch repair that is independent of MGMT (9,31,32). The initial event in m 6 dG-induced mismatch repair may be the recognition ("tagging") of dT-m 6 dG (or dT-dG after removal of methyl group by MGMT) by the hMSH2-p160 heterodimer (33). Functional deficiency of these mismatch recognition/tagging proteins imparts resistance to alkylation-in-duced cytotoxicity (34). Alternatively, a G:T-specific thymine-DNA glycosylase can also remove thymine from dT-m 6 dG base pairs (31). The abasic site is then acted on by apurinic/apyrimidinic (AP) endonucleases. The short gap can then be filled by pol ␤ in a non-semiconservative manner (9). pol ␤ copurifies with deoxyribonuclease V, an exonuclease that can hydrolyze DNA termini in either the 3Ј-5Ј or the 5Ј-3Ј direction with equal facility (35) and may also catalyze the removal of deoxyribose phosphate residues at the abasic site (36). This suggests that during DNA repair in vivo, pol ␤ acts on a variable sized gap. Inefficient repair of these gaps may contribute to methylationinduced cytotoxicity. Thus pol ␤ and the G:T-specific thymine-DNA glycosylase (and possibly other proteins, such as MutS␣ (37)) may be involved in the futile cycling at m 6 dG lesions. pol ␤ may also be secondarily involved in the related m 6 dG-induced long patch (Ͼ50 nucleotides) mismatch repair process that may also prove to be futile and cytotoxic (32).
We have investigated the kinetics of replication across m 6 dG by pol ␤ in order to dissect the dual roles of this DNA polymerase in the cytotoxic repair and mutagenesis induced by m 6 dG. In this paper we report the kinetic parameters of steady state replication (reviewed in Ref. 38) across dG and m 6 dG by pol ␤. We find that m 6 dG acts as a strong, but not complete, block to replication by pol ␤ and that pol ␤, like other polymerases preferentially inserts dT opposite the lesion. Surprisingly, in striking contrast to other polymerases, we also find that the "correct" base pair, dC:m 6 dG is preferentially extended relative to the "incorrect" base pair, dT:m 6 dG.

EXPERIMENTAL PROCEDURES
Materials-Purified human pol ␤ was very generously supplied by Dr. Samuel Wilson (Sealy Center for Molecular Science, University of Texas Medical Branch, Galveston, TX). Klenow fragment of Escherichia coli DNA polymerase I (KF) was purchased from Boehringer Mannheim Biochemicals. The exonuclease deficient Klenow fragment of E. coli DNA polymerase I (exo Ϫ KF) was purchased from U. S. Biochemical Corp. T7 DNA polymerase and T4 polynucleotide kinase were purchased from New England Biolabs. The activity of the polymerases was standardized by measuring the amount of trichloroacetic acid-precipitable radioactivity from [␣-32 P]dGTP incorporated into activated calf thymus DNA. One unit of polymerase activity is defined as the amount of enzyme required for the incorporation of 10 pmol of [␣-32 P]dGMP into 2 g of activated calf thymus DNA in 30 min at 37°C. DNA templates containing m 6 dG, DNA templates without m 6 dG, and primers were synthesized by Midland Certified Reagent Co., Midland, TX, using phosphoramidite chemistry and purified by anion-exchange high pressure liquid chromatography. The purity and correct site specific incorporation of m 6 dG were confirmed by high pressure liquid chromatography and sequence analysis of the oligonucleotides. Ultrapure deoxyribonucleotide triphosphates (dNTPs) (Sequenase/Polymerase Chain Reaction grade) and Sequenase 2.0 were obtained from U. S. Biochemical Corp. [␥-32 P]ATP (3000 or 6000 Ci/mmol) and [␣-32 P]dGTP (3000 Ci/mmol) were purchased from Dupont NEN.
The in vitro DNA replication system ( Fig. 1) consists of short oligo-nucleotide primers and complementary templates containing either m 6 dG (abbreviated as X) or dG at a defined site in the template strand. The 44-mer template is a biologically relevant M13 DNA sequence (nucleotides 118 to 152 of the minus strand).
Primer-Template Annealing-The 32 P-labeled primers were annealed to the normal and m 6 dG containing templates (100 -200 pmol/ reaction) in a hybridization buffer containing 20 mM Tris-HCl (pH 7.8), 10 mM MgCl 2 , 50 mM NaCl, and 1 mM dithiothreitol (DTT). The primer to template ratio was 2:1 (1:2 in some standing start extension studies). The reaction mixture was incubated for 5 min at 90°C and then cooled slowly to room temperature over a period of about 2 h.
Primer Extension Assays-10 -100 pmol of 32 P-labeled primer-template were replicated with different DNA polymerases for different times at 37°C in 10-l reactions containing variable concentrations of dNTPs. Primer extension with pol ␤ was carried out in the presence of 50 mM Tris-HCl (pH 7.5), 2 mM DTT, and 10 mM MgCl 2 . Primer extension with KF or exo Ϫ KF was carried out in the presence of 10 mM Tris-HCl (pH 7.8), 7 mM MgCl 2 , and 12 mM DTT. Replication with T7 DNA polymerase was carried out in the presence of 40 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM DTT, and 0.1 mg/ml bovine serum albumin.
Primer extension reactions were stopped by the addition of stop solution containing 20 mM EDTA in 90% formamide and kept on ice. Before loading onto the gel, the reactions were denatured by heating at 100°C for 5 min. The replication products were resolved on 16% polyacrylamide, 8 M urea denaturing gels at 2000 V for 4 h followed by autoradiography. Autoradiograms were scanned to quantitate the bands using the BioImage Application system (Millipore Inc., Ann Arbor, MI). Film exposures were such that the intensities of the bands to be quantitated were well within the linear range of the apparatus. Relative velocity was determined by the ratio of integrated optical density (IOD) for the target band (t n ) and beyond (t n ϩ t nϩ1 ϩ t nϩ2 ϩ . . . . .) divided by the target band minus one (t n-1 ) normalized per minute of reaction time (39). The units of velocity are thus relative IOD (rIOD) min Ϫ1 . The following equation was used to derive the velocity, rIOD min Ϫ1 ϭ IOD 03ϱ ͑lnϩ1͒ IOD ͑tnϪ1͒ ϫ min reaction time (Eq. 1) The K m(app) (concentration of substrate dNTP at which the observed reaction velocity is half-maximal) and V max(rel) (maximum observed reaction velocity) as well as the associated standard errors were determined by using the K cat ® enzyme and ligand binding program (Biometallics Inc., Princeton, NJ). This program fits the velocity versus substrate concentration data directly to the Michaelis-Menten equation using a robust weighted non-linear regression method (40). The efficiency (f) was calculated as (V max(rel) /K m(app) ). Time course studies (data not shown) indicated that under our standard conditions using pol ␤, the reaction velocity is linear (i.e. obeys steady state kinetics) for up 5 min.
Sequencing-Primed templates were sequenced using Sanger's dideoxynucleotide sequencing method (41). In short, 3 units of Sequenase 2.0, 40 M of all four dNTP's, 40 M each dideoxynucleotide triphosphate, 50 mM NaCl, 3.6 mM MnCl 2 , and 16.7 mM DTT were mixed and primer extension was carried out for 5 min before termination of the reaction by the addition of stop solution. The sequencing lanes were run side-by-side with the primer extension reactions on denaturing polyacrylamide gels, as described above.

RESULTS
Replication of Normal and m 6 dG-modified Templates by pol ␤-We have performed in vitro DNA replication studies of primed normal and m 6 dG modified templates by pol ␤ and three other DNA polymerases in the presence of all 4 dNTPs. The polymerases encounter m 6 dG after a four-base running start. Fig. 2 shows the replication of normal (TGG) and m 6 dGmodified (TXG) 44-mer templates by pol ␤, KF, exo Ϫ KF, and T7 DNA polymerases. During replication of the TXG template, each of the polymerases exhibited pausing immediately before and at m 6 dG, as well as bypass replication. pol ␤ was blocked by m 6 dG to a greater extent than KF and exo Ϫ KF, while T7 polymerase was blocked the most. In this experiment, since the  Fig. 3B also illustrates the prelesion block at dT, the 3Ј neighboring base to m 6 dG. Fig. 4, A and B, show, respectively, the Hanes-Woolf plots of insertion of dTMP and dCMP opposite m 6 dG by pol ␤ in a running a start mode. The kinetic parameters of nucleotide insertion opposite dG and m 6 dG are presented in Table I.
Insertion of dCMP opposite dG was chosen as the standard determination with a relative efficiency value of 1.0 to allow for comparison between insertion of dCMP and dTMP opposite both template dG and m 6 dG. For the correct (dCMP) versus incorrect (dTMP) insertion opposite the normal base, dG, in the running start mode the V max(rel) discrimination is only 5-fold (39.4 rIOD min Ϫ1 for dCMP versus 8.5 rIOD min Ϫ1 for dTMP) while there is a very strong, 8000-fold, K m(app) bias (1.6 M for dCMP versus 12.9 mM for dTMP) ( Table I). The predominantly K m(app) -mediated discrimination suggests that binding of the incoming dTTP to pol ␤-primer-template complex and formation of the T-G base pair is highly inefficient because T-G is an unstable, non-Watson-Crick (wobble) type of base pair (42).
The V max(rel) for nucleotide insertion opposite m 6 dG by pol ␤ is drastically reduced compared to insertion opposite template dG (i.e. V max(rel) discrimination) ( Table I). There is an almost 1000-fold decrease in the V max(rel) for insertion of dCMP opposite m 6 dG (0.04 rIOD min Ϫ1 ) versus dG (39.4 rIOD min Ϫ1 ), and a 71-fold decrease in the V max(rel) for insertion of dTMP opposite m 6 dG (0.12 rIOD min Ϫ1 ) versus dG (8.5 rIOD min Ϫ1 ) ( Table I). This indicates that the new phosphodiester bond 5Ј to dC or dT paired with m 6 dG forms with difficulty, probably because of sterochemical hindrances due to the base pairing of the incoming nucleotide with m 6 dG (42) or because of interference with hydrogen bonding between arginine 283 of the polymerase and the template base (43). Meanwhile, the K m(app) for insertion of dCMP opposite m 6 dG (83.5 M) is 52-fold greater than the K m(app) for insertion of dCMP opposite dG (1.6 M) (Table I). This indicates that pol ␤ employs a mixed (K m(app) and V max(rel) based) discrimination (with V max(rel ) bias playing a predominant role) to reduce the efficiency (an index of the ease of the given step in DNA replication) of insertion of dCMP opposite m 6 dG by 10 5 -fold relative to its insertion opposite normal dG. It is striking that when the template base is m 6 dG instead of dG, the discrimination between insertion of dCMP versus dTMP is mixed, but milder and reversed (i.e. both V max(rel) and K m(app) differences are seen during the insertion of dCMP opposite dG and during the insertion of dTMP opposite m 6 dG) (Table I). However, there is a drastic 1700-fold increase in the K m(app) for insertion of dTMP opposite dG (12.9 mM) relative to the insertion of dTMP opposite m 6 dG (7.6 M) (Table I). This is consistent with the formation of a relatively unstable T-G wobble base pair compared to the formation of a relatively stable Watson-Crick alignment for the incoming dTTP base pair with m 6 dG. This inference from kinetic replication studies is supported by structural NMR studies (42). It is notable that the overall efficiency of insertion of dTMP opposite m 6 dG is 24-fold greater than the insertion of dTMP opposite dG (Table I).
In summary, pol ␤ has a 10-fold lower K m(app) and a 3-fold higher V max(rel) for insertion of the mutagenic nucleotide dTMP opposite m 6 dG as compared to insertion the nonmutagenic nucleotide dCMP. With this mixed K m(app) and V max(rel) bias, pol ␤ has greater than a 33-fold greater efficiency for the insertion of dTMP relative to the insertion of dCMP opposite  (Table I). Thus, pol ␤ suffers a decrease in both efficiency and fidelity during nucleotide insertion opposite m 6 dG relative to insertion opposite dG.
Standing Start Insertion Opposite m 6 dG-Although running start insertion is the more physiological mode of replication, we also investigated the effect of m 6 dG on the kinetics of standing start nucleotide insertion by pol ␤ using 24 nucleotide primers ending immediately prior (5Ј) to the m 6 dG in the template strand (Table II). Using these primer-templates, we again found that discrimination between dCTP and dTTP for insertion opposite the normal dG is mediated by differences in K m(app) while the insertion of dC or dT opposite m 6 dG is determined primarily by a large decrease in the V max . This may indicate that the rate of insertion of a normal base opposite the m 6 dG is governed in part by interactions between the template strand and the active site of the DNA polymerase. It is interesting to note that misinsertion is more efficient in the standing start relative to the running start mode. This may be the result of the slightly processive nature of pol ␤ under these conditions (44).
Standing Start Extension from Terminal Base Pairs-Time course and kinetic studies were performed to compare extension by pol ␤ and exo Ϫ KF from the four different terminal base pairs dC-dG, dT-dG, dC-m 6 dG, and dT-m 6 dG. In the time course studies, extension from dC-m 6 dG and dT-m 6 dGT terminal base pairs (in a T-m 6 dG-G context) was evaluated using pol ␤ in comparison with exo Ϫ KF. As shown in Fig. 5A, we find that exo Ϫ KF extends from the dT-m 6 dG base pair at a faster rate than from dC-m 6 dG. The initial rates of extension (picomoles of primer extended per minute averaged over the first 2.5 min of reaction time (picomoles min Ϫ1 )) from dC-m 6 dG and dT-m 6 dG by exo Ϫ KF are 0.29 and 0.74 pmol min Ϫ1 , respectively, a 2.6fold preference for extension from dT-m 6 dG. Fig. 5B shows that pol ␤, in contrast to exo Ϫ KF, preferentially extends from dCm 6 dG rather than dT-m 6 dG terminal base pairs. The initial rates of extension from dC-m 6 dG and dT-m 6 dG by pol ␤ during the first minute of the reaction are 4.12 and 0.87 pmol min Ϫ1 , respectively, a 4.7-fold preference for extension of dC-m 6 dG.
Kinetic studies were also performed to further characterize the standing start extension from all the four terminal base pairs by pol ␤. Fig. 6, A and B, show representative Hanes-Woolf plots of standing start extension from the dT-m 6 dG and dC-m 6 dG base pairs by pol ␤. The kinetic parameters of base pair extension are tabulated in Table III. In the standing start mode, there is no significant difference in the K m(app) and only a 2-fold difference in the V max(rel) for extension from dC-dG and dC-m 6 dG base pairs by pol ␤ (Table  II). Thus the efficiency varies only by 2-fold. However, the K m(app) for extension from dT-m 6 dG (21.6 M) is 7-fold less than that from dT-dG (142 M) (Table II). Concomitantly, the V max(rel) for extension from dT-m 6 dG (0.11 rIOD min Ϫ1 ) is 3-fold higher than that from dT-dG (0.03 rIOD min Ϫ1 ) (Table  II). This mixed discrimination leads to a 24-fold preference for extension from dT-m 6 dG rather than dT-dG. The K m(app) values for extension from dC-m 6 dG versus dT-m 6 dG are similar. However, the V max(rel) for extension from dC-m 6 dG (0.21 rIOD min Ϫ1 ) is 2-fold higher than that from dT-m 6 dG (0.11 rIOD min Ϫ1 ) (Table II). This leads to an overall 2-fold preference for standing start extension from dC-m 6 dG versus dT-m 6 dG. The relative efficiency of extension from the four terminal base pairs varies in the order: dC-dG Ͼ dC-m 6 dG Ͼ dT-m 6 dG Ͼ Ͼ dT-dG.
Running Start Extension-Running start extensions from dC-dG, dC-m 6 dG, and dT-m 6 dG base pairs with dCTP as the next nucleotide substrate were also evaluated (results not shown). A comparison of the kinetics of running start extension from dC-dG and dC-m 6 dG base pairs also shows 2-fold overall preference for extension from dC-dG versus dC-m 6 dG. Running start extension from dC-m 6 dG (K m(app) ϭ 1.1 M) and dT-m 6 dG (K m(app) ϭ 13.5 M) shows an exclusively K m(app) -based 13-fold preference of extension from dC-m 6 dG versus dT-m 6 dG. This bias is much more pronounced than, but in agreement with, the 2-fold preference for standing start extension from dC-m 6 dG versus dT-m 6 dG. For all three base pairs: dC-dG, dC-m 6 dG, and FIG. 4. Typical Hanes-Woolf plots of running start insertion opposite m 6 dG by pol ␤. A, Hanes-Woolf plot of running start insertion of dTMP opposite m 6 dG by pol ␤ opposite m 6 dG is presented. Reaction conditions for dTMP incorporation were as described in the legend for Fig. 3. B, shows dCMP incorporation opposite m 6 dG. Incorporation was measured using 0.3 unit of pol ␤ and 10 pmol of 23-mer-TXG primer-template. The concentration of dATP was held fixed at 10 M. Reaction time was 5 min.

TABLE I
Kinetic parameters of running start nucleotide insertion opposite dG and m 6 dG by pol ␤ All experiments were performed using 10 -100 pmol of the 20-mer primer (the 23-mer primer was used for the insertion of dTMP opposite dG) hybridized to the 44TXG template (where X denotes either dG or m 6 dG) as described under "Experimental Procedures" using limiting enzyme concentration (0.3 unit/reaction) and varying the appropriate dNTP. dT-m 6 dG, extension in the running start mode is severalfold more efficient than in the standing start mode, the opposite of the trends seen for insertion. Furthermore, in the standing start mode of extension, differences in V max(rel) , K m(app) , and efficiency are less pronounced than during running start extension. These observations may be related to pol ␤ functioning semi-processively rather than distributively in our system, as has been previously noted by Werneburg et al. (44).
Overall Replication across m 6 dG by pol ␤-It is notable that while there are drastic K m(app) and V max(rel) differences during nucleotide insertion opposite m 6 dG versus dG (Tables I and II), the K m(app) and V max(rel) differences are not very pronounced during extension from terminally base paired m 6 dG versus dG (Table III). Thus pol ␤ seems to be more discriminating during the initial nucleotide insertion step rather than during the later base pair extension step. Furthermore, the preferential insertion of dTMP is partially opposed by preferential extension from dC-m 6 dG.
In Table IV the overall efficiency of insertion and extension from the four base pairs dC-dG, dT-dG, dC-m 6 dG, and dTm 6 dG has been calculated as a product of the efficiency of insertion and the efficiency of standing start extension. The overall efficiency is an index of the relative ease of replication by pol ␤ across template dG or m 6 dG using dCTP or dTTP as the nucleotide substrates. Replication across template dG with dCTP as the nucleotide substrate is chosen as the standard with a relative overall efficiency value of 1.0. During replication across template dG, dCTP is favored over dTTP as the nucleotide substrate with greater than a 10 6 -fold preference (with a 10 4 -fold preference at the insertion step and a 10 2 -fold preference at the extension step). When dCTP is the substrate nucleotide, the replication inhibiting nature of template m 6 dG becomes evident in light of the 10 5 -fold decreased overall efficiency as compared to dG. This large bias occurs almost exclusively at the initial insertion step. On the other hand when dTTP is the nucleotide substrate, m 6 dG is a significantly better template base than dG as indicated by an almost 600-fold greater overall efficiency, with equal 24-fold contributions at both the insertion and extension steps. A comparison of the overall relative efficiencies shows the following trend in the ease of overall replication (insertion and extension): dC-dG Ͼ Ͼ dT-m 6 dG Ͼ dC-m 6 dG Ͼ dT-dG.
Kinetics of exo -KF Insertion and Extension in the Same Sequence Context-For comparison to these results obtained with pol ␤, Table V shows the kinetics of standing start extension by exonuclease-deficient pol I-KF. Note that the K m(app) for extension from a dT-m 6 dG base pair is 10-fold lower than for extension from a dC-m 6 dG base pair while the V max(rel) is the same, thus the relative efficiency of extension from dT-m 6 dG by exo Ϫ KF is 10-fold higher than the efficiency of extension of the dC-m 6 dG base pair, in substantial agreement with the kinetic data shown in Fig. 5. Table VI presents the overall efficiency of running start nucleotide insertion and standing start extension by E. coli exo Ϫ KF in this sequence context. It is clear that the dT-m 6 dG base pair is greatly favored over the dC-m 6 dG base pair, as has been demonstrated previously by other approaches (19). These results are also similar to those obtained for insertion and extension of dTMP and dCMP opposite m 6 dG by E. coli KF in a different sequence context (45). Table VII allows easy analysis of the type (i.e. K m(app) and V max(rel) ) and relative extent of discrimination exhibited by pol ␤ during replication across template dG and m 6 dG using dCTP and dTTP as the nucleotide substrates, during both insertion and extension. Attention is brought to the pronounced discrimination seen during nucleotide insertion opposite both dG and m 6 dG. This contrasts with the less pronounced discrimination during extension from the four base pairs. This theme is exemplified by the severe K m(app) and V max(rel) differences seen during insertion of dCMP opposite dG versus m 6 dG on one hand and the virtual lack of bias during extension from dC-dG versus dC-m 6 dG. Table VII also shows the relative contributions of insertion and extension efficiencies to the overall efficiencies of favored versus unfavored replication events. DISCUSSION m 6 dG Lesions Inhibit Repair Replication by pol ␤-pol ␤ is likely to replicate across persistent m 6 dG lesions in vivo when MGMT-mediated repair is saturated or lacking. We have performed in vitro kinetic studies to characterize the mechanisms of replication across site-specific m 6 dG adducts by pol ␤. Our results suggest that pol ␤, in concert with the G:T-specific thymine glycosylase, may contribute to the cytotoxic repair induced by m 6 dG. Our results also indicate that the infrequent bypass replication across m 6 dG by pol ␤ may lead to insertion of dTMP opposite m 6 dG and thus contribute toward G 3 A mutagenesis by methylating agents in vivo.
pol ␤ was previously reported to be unable to bypass m 6 dG (46,47). We have demonstrated that pol ␤ is able to bypass m 6 dG, but at a much reduced efficiency. In the study conducted by Abbotts et al. (47), m 6 dG was located in a run of dT's and the activating metal ion was manganese. In our experiments m 6 dG is located in a heterogeneous sequence context (Fig. 1) and the activating metal ion is magnesium. It is interesting to note that  in our sequence context the insertion of dTMP opposite dG is very disfavored with an overall relative efficiency of 2 ϫ 10 Ϫ7 . This is in distinct contrast to other sequence contexts where the relative efficiency of formation of the dT-dG base pair is on the order of 10 Ϫ4 (44). In both the studies by Abbotts et al. (47) and Voigt and Topal (46), pol ␤ was blocked by m 6 dG during replication in the presence of all four dNTP substrates. Our studies with individual dNTP substrates show that the mechanism of this block is the preferential incorporation of dTMP opposite m 6 dG but poor extension from the dT-m 6 dG base pair. pol ␤ (and other DNA polymerases KF, exo Ϫ KF, and T7 DNA polymerase) all pause at the prelesion and lesion site when replicating across m 6 dG. In agreement with our observations, a previous study by Menichini et al. (48) showed that KF exhibits prelesion pausing when replicating across O 4 -methylthymine. Abbotts et al. (47) also showed that pol ␤ was blocked mostly at the prelesion site before m 6 dG. Pausing before or at m 6 dG may be due to polymerase stalling and/or dissociation from the primer-template before or after nucleotide incorporation opposite m 6 dG, respectively. In agreement with the lesion bypass capability of pol ␤ reported here, there have been previous studies reporting that pol ␤ can replicate past lesions such as apurinic sites (49), cis-syn thymine dimers (50) and (GpG)cisplatin lesions (51).
Preferential Insertion of dTMP Opposite m 6 dG and Extension from dC-m 6 dG-pol ␤ has a 33-fold preference for insertion of the mutagenic nucleotide substrate dTMP, rather than the nonmutagenic dCMP, opposite m 6 dG. In this respect pol ␤ is similar to KF, T4 and T5 polymerases which also preferentially insert dTMP (17)(18)(19). On the other hand, pol ␤ has an opposing 2-fold (13-fold from a running start) preference for extension from dC-m 6 dG rather than dT-m 6 dG. This contrasts with the 10-fold preferential extension from dT-m 6 dG by exo Ϫ KF, whereas previously published experiments using KF in difference sequence contexts have observed up to a 3-fold preference for extension from dT-m 6 dG (45). The efficiency of incorporation of the next correct nucleotide to extend a terminal base pair depends on the degree of base pairing stability of the terminal base pair and the ensuing local helical distortion. Phosphodiester bonds 3Ј and 5Ј to dC paired with m 6 dG have been shown to be distorted (42). The observation that dC-m 6 dG is extended almost as well as dC-dG suggests that, in contrast to KF and exo Ϫ KF, pol ␤ is able to extend from poorly bonded terminal base pairs and is relatively insensitive to local helical perturbations.
The finding that pol ␤ extends from dC-m 6 dG better than from dT-m 6 dG is difficult to rationalize in view of NMR data which shows that the dT-m 6 dG base pair more stable, has Watson-Crick alignment, and the alkyl group of m 6 dG has an anti orientation, causing less distortion of local DNA structure, while dC-m 6 dG is a less stable wobble base pair, and the alkyl group of m 6 dG has an syn orientation which causes more distortion of local DNA structure (42,52). The lack of 3Ј-5Ј exonuclease function, lower processivity, fidelity, and other less understood features of pol ␤ may be involved in imparting such reduced and reversed discrimination during extension from this unusual base pair. It has been previously reported that pol ␤ efficiently extends from primer templates with one or more terminal mismatches (53).
Preferential Incorporation of dTMP during Overall Replication across m 6 dG-With regard to the overall replication across m 6 dG, pol ␤ has a 16-fold preference for incorporating dTTP rather than dCTP. This mutagenic bypass replication of m 6 dG FIG. 6. Representative Hanes-Woolf plots of standing start extension from dT-m 6 dG and dC-m 6 dG by pol ␤. A presents standing start extension from dT-m 6 dG by pol ␤ (0.4 unit) using the 25-mer(T)-44TXG primer-template (100 pmol/reaction). Two determinations were made at each concentration of the next nucleotide substrate, dCTP. The concentration of dATP was fixed at 10 M. B presents standing start extension of dCm 6 dG from the 25-mer(C)-44TXG primertemplate. The reaction time was 2 min.  is 10 4 -fold less efficient than replication across dG. This indicates that pol ␤ is strongly, but not completely, blocked at m 6 dG and that bypass replication is predominantly mutagenic and would contribute to production of G to A mutations in vivo. pol ␤ exhibits mixed kinetic discrimination (i.e. both K m(app) and V max(rel) differences) and decreased fidelity during replication across template m 6 dG. In agreement with our observations, Boosalis et al. (54) have also reported a mixed discrimination by pol ␤. pol ␤ employs a more rigorous kinetic discrimination (i.e. severalfold differences in K m(app) and V max(rel) ) during the initial nucleotide insertion step that contrasts with the much decreased discrimination during subsequent base pair extension. Our results show that enhanced kinetic discrimination may not always be concordant with fidelity (in terms of mutagenesis). pol ␤ exhibits reduced fidelity but greater kinetic discrimination during nucleotide insertion opposite m 6 dG, but has greater fidelity and reduced kinetic discrimination during subsequent base pair extension from m 6 dG.
The effect of Processivity-Depending on the replication system, pol ␤ may function in a distributive mode (i.e. recycles, dissociates, and reassociates, with each nucleotide incorporation event) or a slightly processive mode (i.e. incorporates a few nucleotides before recycling). In a distributive mode pol ␤ may have relatively weak, inefficient binding interactions with the primer-template for easier recycling. Binding and interaction may be better in the slightly processive mode. Furthermore, a distributive mode would predict that the kinetic parameters of running start and standing start extensions of different base pairs should be equivalent. Due to complete recycling after each terminal base pair extension event pol ␤ would always function in a standing start mode even in a running start assay. On the other hand, in a slightly processive mode the initial binding step is not repeated. Thus running start extension may be more efficient when pol ␤ is in a slightly processive versus a distributive mode. Our kinetic analysis shows significantly greater efficiencies for running start versus standing start extensions (not shown). Furthermore, the relatively weak, incomplete intermediate pause sites of pol ␤ (seen in Fig. 2) indicate a slightly processive mode. In agreement with our observations, the slightly processive mode of pol ␤ during gap filling as well as during replication of synthetic oligonucleotide primertemplates has been previously reported (8,44).
The Role of pol ␤ in the Futile Repair of m 6 dG-During genomic replication by replicative polymerases, dTMP be may preferentially inserted opposite persisting m 6 dG lesions. The dT-m 6 dG base pair is recognized by a G:T-specific thymine glycosylase which preferentially excises the thymine (31). The subsequent action of AP endonucleases creates a short gap. When filling the short gap containing m 6 dG, pol ␤ preferentially incorporates dTMP opposite m 6 dG with a low efficiency. The dT-m 6 dG base pair is not efficiently extended, thus leaving a long-lived gap or nick. Once the nick is sealed, the G:Tspecific thymine glycosylase may again excise the thymine opposite the m 6 dG, followed by another round of action of AP endonucleases (and/or pol ␤/deoxyribonuclease V), gap filling by pol ␤, etc. This futile cycling of the repair process (32) may sequester the repair machinery at m 6 dG lesions. The resulting long lived gaps may also induce sister chromatid exchanges and other types of cytotoxic responses. Incision by thymine glycosylase has also been shown in different DNA environments to be either sensitive or insensitive to the nature of the 5Ј neighboring base to dG or m 6 dG base paired with T (55,56). Thus the local DNA sequence may modulate the cytotoxicity of m 6 dG during the futile cyclic repair mechanism.
Summary-This is the first report to dissect the kinetic parameters of replication across m 6 dG by pol ␤ in vitro. Our results are physiologically significant since they provide mechanistic insight into the involvement of pol ␤ in the futile repair of m 6 dG that may lead to cytotoxicity. This study also provides a kinetic basis to explain how pol ␤ may be partially blocked by m 6 dG and contribute to the mutagenesis of m 6 dG in vivo. Moreover, this study provides important functional and mechanistic information regarding the kinetic discrimination, fidelity, and processivity of pol ␤ during in vitro replication across normal and miscoding template bases.   replication across dG and m 6 dG using dCTP and dTTP as the nucleotide substrates The following symbols were used: 0, no difference; Ϫ, negative discrimination, i.e. increased K m(app) , decreased V max(rel) , or decreased f (Ϫ ϭ 3-10ϫ, ----ϭ 10 -10 2 ϫ); ϩ, positive discrimination, i.e. decreased K m(app) , increased V max(rel) , or increased f (ϩ ϭ 3-10ϫ; ϩϩ ϭ 10 -10 2 ϫ; ϩϩϩ ϭ 10 2 -10 3 ϫ; ϩϩϩϩ ϭ 10 3 -10 4 ϫ; ϩϩϩϩϩ ϭ Ͼ10 4 ϫ).