The 3′ → 5′ Exonuclease of T4 DNA Polymerase Removes Premutagenic Alkyl Mispairs and Contributes to Futile Cycling atO 6-Methylguanine Lesions*

We have studied the processing ofO 6-methylguanine (m6G)-containing oligonucleotides and N-methyl-N-nitrosourea (MNU)-treated DNA templates by the 3′ → 5′ exonuclease of T4 DNA polymerase. In vitro biochemical analyses demonstrate that the exonuclease can remove bases opposite a defined m6G lesion. The efficiency of excision of a terminal m6G·T was similar to that of m6G·C, and both were excised as efficiently as a G·T substrate. Partitioning assays between the polymerase and exonuclease activities, performed in the presence of dNTPs, resulted in repeated incorporation and excision events opposite the m6G lesion. This idling produces dramatically less full-length product, relative to natural substrates, indicating that the 3′ → 5′ exonuclease may contribute to DNA synthesis inhibition by alkylating agents. Genetic data obtained using an in vitro herpes simplex virus-thymidine kinase assay support the inefficiency of the exonuclease as a “proofreading” activity for m6G, since virtually all mutations produced by the native enzyme using MNU-treated templates were G → A transitions. Comparison of MNU dose-response curves for exonuclease-proficient and -deficient forms of T4 polymerase reveals that the exonuclease efficiently removes 50–86% of total premutagenic alkyl mispairs. We propose that idling of exonuclease-proficient polymerases at m6G lesions during repair DNA synthesis provides the biochemical explanation for cellular cytotoxicity of methylating agents.

O 6 -Methylguanine (m6G) 1 is a prototype DNA adduct produced by alkylating agents that is both cytotoxic and mutagenic. Two mechanisms potentially govern the cytotoxic effects of the m6G lesion. O 6 -Methylguanine lesions decrease the efficiency of DNA synthesis by replicative polymerases in vitro (1,2) and inhibit in vitro DNA replication in cell-free extracts (3). Alternatively, the persisting m6G can be cytotoxic through lethal interactions with the mismatch repair pathway in mammalian cells (reviewed in Ref. 4). Loss of O 6 -methylguanine-DNA methyltransferase activity results in the persistence of unrepaired m6G lesions, and such cells are sensitive to killing by alkylating agents. Additional loss of mismatch repair proteins renders O 6 -methylguanine-DNA methyltransferase-deficient cells tolerant of the cytotoxic effects of alkylating agents (5,6). However, the precise biochemical mechanism mediating this toxicity has not been completely defined (4). In one model, mismatch repair activity causes persistent DNA strand breaks which signal cell cycle arrest and apoptosis (4,7).
Replicative and repair-associated DNA polymerases contain 3Ј 3 5Ј proofreading exonucleases, which act in coordination with the polymerase activity to enhance the efficiency of error discrimination. Despite the critical role of polymerases in mediating DNA damage mutagenesis, little is known regarding the biochemistry of this polymerase exonuclease activity at DNA lesions. Eukaryotic DNA polymerases ␦ and ⑀ contain the polymerase domain and the exonuclease domain within the same polypeptide. Thus, during DNA synthesis, the nascent 3Ј-primer terminus must be shuttled between the two active sites in order for excision to occur. In studies using 2-aminopurine fluorescence, the rate-limiting step in excision by the T4 polymerase is translocation of the DNA from the polymerase to the exonuclease site (8,9). Part of the molecular "switch" for this process is thought to be the structure of the DNA (10 -12). The preferred DNA substrate for the polymerase domain is base-paired duplex DNA, whereas that for the exonuclease is melted duplex DNA containing a few single-stranded bases. Any factor that destabilizes duplex DNA is expected to enhance the exonuclease catalytic rate relative to the polymerase catalytic rate. For example, the significant distortion of DNA within the DNA polymerase binding cleft may destabilize duplex DNA and enhance excision by affecting the equilibrium between single-stranded and double-stranded states (13). Alternatively, the melting capacity of the primer terminus determines the rate of editing by the Klenow polymerase (10). The physical structures of m6G⅐T and m6G⅐C base pairs have been determined using crystal (14,15) and solution (16,17) methods. Structurally, the m6G⅐T mispair simulates Watson-Crick alignments, and the m6G⅐T pair is indistinguishable in the minor groove from G⅐C (reviewed in Ref. 18). Thermodynamically, however, the m6G⅐T pair is less stable than both the m6G⅐C and G⅐T pairs in DNA and displays a lower melting temperature (15,19). We report here the efficiency of exonuclease removal of m6G and other alkylation mispairs and the resulting effects on alkylation-induced errors by the T4 DNA polymerase. molecular biology procedures were supplied by Life Technologies, Inc. and used according to the manufacturer's protocol. N-Methyl-N-nitrosourea (MNU) and 5-fluoro-2Ј-deoxyuridine were purchased from Sigma.
The presence of m6G in the final template preparation was verified by matrix-assisted laser desorption-ionization time-of-flight mass spectrometry by the manufacturer. The 15-mer primers were labeled at the 5Ј-end using [␥-32 P]ATP (5000 Ci/mmol) and T4 polynucleotide kinase according to the manufacturer's protocol. The labeled primers were annealed with the template oligonucleotides by combining 1:3 primertemplate molar ratios of each DNA in the presence of SSC. The mixtures were heated at 75°C for 5 min and gradually cooled to room temperature. The primer-template preparations were purified using G25 Sephadex Quickspin columns prior to use.
3Ј 3 5Ј Exonuclease Assay-The exonuclease reaction mixture contained 50 mM Tris-OAc (pH 7.4), 10 mM dithiothreitol, 10 mM Mg(OAc) 2 , 150 mM KOAc, 10 nM DNA primer-template, and 0.5 nM enzyme. Reactions were initiated by the addition of enzyme and were incubated at 37°C. At the indicated times, aliquots were added to an equal volume of stop solution (containing 20 mM EDTA and 90% formamide) to terminate the reactions. Reaction products were separated through a 16% denaturing polyacrylamide gel. Band intensities were quantitated using a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA). The radioactivity at respective time points was normalized to the amount of radioactivity in the control (no polymerase) lane.
Polymerase/Exonuclease Partitioning Assay-T4 polymerase partitioning assays were carried out in the presence of either 200 M dGTP (next correct nucleotide) or 200 M four dNTPs, 5 or 50 nM wild-type enzyme, and 10 nM primer-template using the reaction buffer described above. Aliquots were removed at the indicated time points and quenched, and the products were separated on a 16% denaturing polyacrylamide gel. For reactions containing dGTP only, radioactivity was measured for unmodified substrate (n), and extension (n ϩ 1) or excision (n Ϫ 1, n Ϫ 2) products. For reactions containing all four dNTP substrates, radioactivity was measured for full-length (33-mer) product, intermediate extension products (16 -32-mer), and excision products (n Ϫ 1, n Ϫ 2), relative to the total radioactivity in each sample. The experimental conditions used for the D324A extension assay were the same as above, except the protein concentration was 5 M.
In Vitro Polymerase Reactions-Alkylated DNA substrates were created by random MNU modification (Me 2 SO solvent) of primed, singlestranded DNA synthesis templates and purified prior to use in the polymerase reactions, as previously described (20). The in vitro reactions for mutational analyses contained 2 pmol of ssDNA template at 40 nM concentration. Reaction conditions were 50 mM Tris-OAc (pH 7.4) or 25 mM Hepes-HCl (pH 7.4), 150 mM KOAc, 10 mM Mg(OAc) 2 , 10 mM dithiothreitol, 200 M dNTPs, and 40 pmol of T4 polymerase. Reactions were incubated at 37°C for 60 min and terminated with 15 mM EDTA. The extent of DNA synthesis was determined by parallel reactions (0.2 pmol of DNA; same molar ratios of enzyme to substrate as above) supplemented with 5 Ci of [␣-32 P]dCTP (3000 Ci/mmol). A 64-mer oligonucleotide oligonucleotide was radiolabeled at the 5Ј-end using [␥-32 P]ATP (5000 Ci/mmol) and T4 polynucleotide kinase, and 50 fmol were added to each reaction after termination for use as an internal loading standard. The DNA products were analyzed on an 8% denaturing polyacrylamide gel together with a DNA sequencing ladder generated from the same primer-template, followed by PhosphorImager analysis. The amount of synthesis in each reaction was normalized first to the internal standard and then to the solvent-treated control for each polymerase. No reaction products less than 210 nucleotides in length were observed for any polymerase with either solvent or MNU treatment.
HSV-tk Forward Mutational Assay-Mutational analyses of the reaction products were performed as reported previously (20). Briefly, the ssDNA synthesis products were digested with EcoRV and MluI restriction enzymes, and 203-base pair double-stranded products were purified. An equivalent yield of product (50 -100 ng; 0.4 -0.8 pmol), as estimated by gel electrophoresis, was obtained for all solvent and MNU-treated polymerase reactions. These fragments were hybridized to gapped duplex molecules (0.05 pmol) containing one chloramphenicolresistant strand and one chloramphenicol-sensitive strand. To select for HSV-tk mutations, an aliquot of the final hybridization was used to transform recA13, upp, tdk E. coli strain FT334 by electroporation and plated on VBA-selective media. Progeny of the DNA strand produced during in vitro synthesis were selected with 50 g/ml chloramphenicol, and HSV-tk mutant plasmids were selected with 40 M 5-fluoro-2Јdeoxyuridine. The HSV-tk mutant frequency is defined as the number of 5-fluoro-2Ј-deoxyuridine ϩ chloramphenicol-resistant colonies divided by the total number of chloramphenicol-resistant (Cm R ) colonies.
In the experiments presented here, a total of 0.14 -3 ϫ 10 5 Cm R transformants (solvent treatment) and 0.06 -6 ϫ 10 5 Cm R transformants (MNU treatment) were analyzed for mutants for each polymerase. Mutational spectra were derived from independent mutants, as described (21). Differences in proportions of specific types of errors were analyzed statistically using Fisher's exact test (two-tailed).

RESULTS
O 6 -Methylguanine-containing Templates Are Substrates for the 3Ј 3 5Ј Exonuclease Activity of T4 Polymerase-The intrinsic exonuclease activity of T4 polymerase on various natural and m6G-containing templates was examined by incubating the enzyme and DNA in the absence of dNTP substrates. As expected for control templates, the T4 exonuclease activity exhibited "proofreading" activity by degrading terminal G⅐T mispair-containing substrates with a greater efficiency (t1 ⁄2 ϳ6 min) than correctly paired G⅐C-containing substrates (t1 ⁄2 ϳ60 min), under enzyme-limiting conditions (Fig. 1). In contrast, the exonuclease exhibited little or no discrimination between the two m6G-containing substrates. Both m6G⅐C and m6G⅐T substrates were degraded with similar efficiencies (t1 ⁄2 ϳ4.5 min and 4 min, respectively), near that observed for the G⅐T mispaired substrate (Fig. 1). More rapid excision was observed for all four substrates during the first minute of the reaction, compared with later times. This observation may reflect the high processivity of the T4 exonuclease, since excision of the 15-mer primers was complete under the substrate excess conditions utilized. These results show that although m6G-containing templates are substrates for the T4 exonuclease activity, this activity may not contribute a significant "proofreading" function to DNA polymerase errors, because both m6G⅐C and m6G⅐T substrates are excised with similar efficiencies.
T4 Polymerase Partitioning between Polymerase and Exonuclease Active Sites at m6G Lesions-Partitioning of the T4 enzyme between the polymerase and exonuclease activities was assayed in the presence of the next correct nucleotide, dGTP (Fig. 2). The correctly paired natural substrate (G⅐C) was efficiently extended by the enzyme with minimal exonuclease degradation. In contrast, polymerase activity was highly inhibited using a mispaired G⅐T substrate (Ͻ1% extension), relative to the paired substrate (90% extension), and exonucleolytic degradation products were predominant over the time course analyzed. However, inclusion of m6G in the substrate eliminated this discrimination, and the exonuclease activity predominated in reactions containing either the m6G⅐C or the m6G⅐T substrate (Fig. 2, C and D).
T4 Polymerase Exhibits Idling on m6G Substrates-The 3Ј 3 5Ј exonuclease assays (Figs. 1 and 2) demonstrate that this activity could excise both nonmutagenic (C) and mutagenic (T) bases incorporated opposite m6G lesions. We next examined whether the polymerase or the exonuclease activity predominates in the presence of all four dNTP substrates (Fig. 3). Experiments performed using natural G templates with or without a mispaired primer terminus resulted in the accumulation of full-length product (Fig. 3, A and B). As expected, the 3Ј 3 5Ј exonuclease activity was not detectable in the presence of the G⅐C substrate and 200 M dNTPs. However, using the mispaired G⅐T substrate, excision products accumulated during the initial time points with little extension (Fig. 3B). With increased time of incubation, full-length product accumulated concomitant with disappearance of excision bands, suggesting that efficient extension occurred only after complete excision of the terminal mispaired thymine.
The enzyme polymerization activity was found to be highly inhibited using lesion-containing templates. For both m6G⅐C and m6G⅐T substrates, excision products persisted throughout the course of the reaction and predominated over extension products (Fig. 3, C and D). Increasing the enzyme concentration 10-fold (enzyme excess conditions) did not affect the polymerase/exonuclease balance, since the ratio of extension/excision products remained ϳ1:1 ( Table I). Accumulation of the 15-mer products probably results from the repeated cycles of incorporation and excision of bases at the lesion-containing primer terminus. In order to examine the m6G substrate preference of the polymerase activity, extension of m6G-containing substrates was examined using the exonuclease-compromised D324A T4 polymerase (Fig. 3, E and F). The enzyme was found to extend the m6G⅐T substrate somewhat more efficiently than the m6G⅐C substrate (Table I). However, even under these large enzyme excess conditions, the m6G lesion is inhibitory to the T4 polymerase, since 20 -50% of the substrates remained unextended. The appearance of excision products (Fig. 3, E and F) may represent either residual exonuclease activity or nonenzymatic cleavage in the presence of Tris buffer, as has been reported by others (22).
Genetic Analyses Reveal That the T4 Exonuclease Activity Can Proofread Alkyl-directed DNA Mispairs-In the biochemical assays, we observed a similar yield of full-length reaction products for both the nonmutagenic (m6G⅐C) and mutagenic (m6G⅐T) substrates (Fig. 3). In order to quantitate the extent to which the exonuclease contributes to error avoidance, we performed the in vitro HSV-tk forward mutation assay (20). The wild-type, proofreading-proficient enzyme was compared with two proofreading-deficient forms: D324A, which exhibits a 10 4 to 10 5 -fold reduction in exonuclease activity, and D219A, which exhibits a 10 3 -fold reduction in exonuclease activity (22,23). At 200 M dNTPs, using unmodified DNA templates, the mutation frequency for wild-type T4 polymerase was 2.4 Ϯ 0.72 ϫ 10 Ϫ4 , approximately 5-fold lower than either the D324A (21 ϫ 10 Ϫ4 ) or the D219A (10 Ϯ 2.5 ϫ 10 Ϫ4 ) exonuclease-deficient enzymes and near the background for this assay (20).
In the HSV-tk forward mutation assay, chemically treated, oligonucleotide-primed ssDNA is used as a template for DNA polymerase reactions. The DNA reaction products are digested with restriction enzymes, and a 203-base pair DNA synthesis product is purified. To recover and analyze these DNA fragments for the presence of mutations, a gapped duplex molecule is used that is formed by hybridization of a linear chloramphenicol-resistant DNA fragment to a chloramphenicol-sensitive ssDNA. DNA synthesis fragments containing potential mutations within the thymidine kinase gene are rescued by hybridization to the gapped duplex, forming heteroduplex plasmid molecules that are used to transform E. coli. Incubation of the transformed bacteria in the presence of chloramphenicol selects progeny of the DNA strand synthesized in vitro (20). HSV-tk mutant plasmids are selected by plating the bacteria in the presence of 5-fluoro-2Ј-deoxyuridine. The resulting HSV-tk mutant frequency is a measure of the proportion of DNA fragments containing mutations.
Modification of oligonucleotide-primed single-stranded DNA templates with MNU resulted in a dose-dependent inhibition of DNA synthesis for the three polymerases. However, at low levels of modification, the inhibition of synthesis observed for the wild-type polymerase was greater than that observed for either exonuclease-deficient enzyme (Fig. 4A). Genetic analyses of the reaction products from these modification reactions are shown in Fig. 4B. For the wild-type polymerase, a linear mutation versus dose-response curve was observed, with a 30fold increased frequency at 5 mM MNU (110 Ϯ 39 ϫ 10 Ϫ4 ), relative to solvent control (3.6 Ϯ 3.9 ϫ 10 Ϫ4 ). We repeated the wild-type dose-response curve in Hepes buffer to eliminate potential nonenzymatic exonuclease activity, and again observed a 37-fold increased mutation frequency at 5 mM MNU (270 ϫ 10 Ϫ4 ), relative to solvent-treated templates (7.3 ϫ 10 Ϫ4 ). The mutation rate (defined by the slope of the doseresponse curves) of the D324A polymerase was 7-fold greater than that of wild-type T4, while that of the D219A was 2-fold greater than wild-type (Fig. 4B). At the highest dose tested (5 FIG. 1. T4 polymerase 3 3 5 exonuclease activity with m6Gcontaining substrates. Priming oligonucleotides were hybridized to either natural (G) (top panels) or alkylated (m6G) (bottom panels) template oligonucleotides to create the 3Ј-terminal base pairs indicated (template-primer). The 15-mer primers were labeled at the 5Ј-end with 32 P prior to hybridization. The exonuclease reactions contained 0.5 nM T4 polymerase and 10 nM DNA substrate in a Tris-OAc (pH 7.4) reaction buffer. Reactions were terminated, and the products were separated on a 16% denaturating polyacrylamide gel. A, representative PhosphorImager scans of reaction products. Lane 1, control reaction containing no polymerase; lanes 2-5, products after 0.5-, 1-, 5-, and 15-min incubation with enzyme at 37°C. B, quantitation of reaction products. The radioactivity was quantitated for the starting substrate (15-mer) and degradation products at each time point and normalized to the control reaction. Each data point represents the average of two independent determinations. Solid lines and symbols, natural templates; dotted lines and open symbols, m6G templates. Circles, 3Ј C; triangles, 3Ј T. mM MNU), the absolute mutation frequencies measured for the D324A and D219A polymerases were 740 ϫ 10 Ϫ4 and 250 ϫ 10 Ϫ4 , respectively.
Comparison of the wild-type and exonuclease-deficient doseresponse curves demonstrates that the exonuclease of T4 polymerase is able to remove 50 -86% of total alkylation mispairs. Random modification of ssDNA by MNU results in alkylation at template bases in the order: G Ͼ Ͼ A Ͼ C (24). DNA sequence analyses revealed that the wild-type enzyme produced exclusively G 3 A transition mutations using alkylated templates (Fig. 5). The frequency of G 3 A transition mutations at 5 mM MNU treatment was increased 2000-fold, relative to solvent treatment (Table II). At the same dose, the G 3 A mutation frequency for the D324A polymerase was 2.6-fold higher than that for the wild-type polymerase (Table II), indicating that the exonuclease activity removed only ϳ38% of m6G⅐T mispairs. Inspection of the mutational spectra derived from all three forms of the T4 polymerase reveals that the majority of errors removed by the exonuclease involve methylated C and A lesions (Table II). In the D324A spectrum, eight mutational events (26%) induced by MNU were C 3 T transitions. This corresponds to a frequency of 2.3 ϫ 10 Ϫ2 , nearly equivalent to the G 3 A frequency and 190-fold greater than the solvent-treated control frequency. Other prevalent mutations in the D219A and D324A MNU-induced spectra include C 3 A transversions (18 and 10%, respectively), A 3 T transversions (15 and 10%, respectively), and one-base deletions at template A (7 and 22%, respectively).  2-4, respectively).

TABLE I Quantitation of partitioning of polymerase and 3Ј 3 5Ј exonuclease activities of T4 polymerase
The assay was performed as described under "Experimental Procedures" in the presence of 200 M dNTP substrates, and the corresponding gel analysis of products is shown in Fig. 3. The amount of radioactivity in each indicated product band was quantitated after a 60-s incubation, relative to the total DNA product present in that sample. Reactions contained either 5 nM or 50 nM wild-type T4 polymerase or 5 M D324A polymerase and 10 nM m6G-containing template DNA. ND, none detected.  2. T4 polymerase/exonuclease partitioning activity with m6G-containing substrates. The DNA substrates and reaction conditions described in the legend to Fig. 1 were repeated, except that 200 M dGTP was added to each reaction; enzyme concentration was 5 nM, and DNA substrate concentration was 10 nM. Aliquots were removed at the indicated time points and quenched, and the products were separated on a 16% denaturing polyacrylamide gel. Radioactivity was measured by PhosphorImager analyses for unmodified substrate (n) and extension (n ϩ 1) or excision (n Ϫ 1, n Ϫ 2) products. Lane 1, DNA substrate without enzyme; lanes 2-5, contain products after 15-, 30-, 60-, and 120-s incubation times. The values below each lane were quantitated from the 60-s reaction and are expressed as a percentage of the control (no polymerase). The DNA substrates used were G⅐C (A), G⅐T (B), m6G⅐C (C), and m6G⅐T (D).

DISCUSSION
Our results demonstrate that the 3Ј 3 5Ј exonuclease activity of T4 DNA polymerase recognizes and removes both dCMP and dTMP incorporated opposite a defined m6G lesion (Figs. 1  and 2). This activity results in severe DNA synthesis inhibition at the lesion in the presence of dNTP substrates, due to repeated cycles of incorporation and excision (Fig. 3). Consistent with this, we observed less DNA synthesis inhibition for exonuclease-deficient forms of the T4 polymerase, relative to the wild-type enzyme, using MNU-treated DNA templates (Fig.  4A). Genetic analyses demonstrate that the exonuclease removes ϳ38% of the mutagenic m6G⅐T mispairs but virtually 100% of other types of alkyl-directed mispairs (Fig. 5 and Table II).
Our observations can be explained by taking into account the differential basis of substrate discrimination by the polymerase and exonuclease enzymatic activities. As has been observed previously by others (2), exonuclease-deficient T4 polymerase displays a preference for extension of m6G⅐T over m6G⅐C substrates (Fig. 3), consistent with a polymerase activity that discriminates on the basis of structural configuration of bases (18). Physical determinations have shown that the m6G⅐C base pair is in altered wobble geometry, while the mutagenic m6G⅐T base pair simulates Watson-Crick geometry. In contrast, we observed that the T4 3Ј 3 5Ј exonuclease activity displayed little discrimination between nonmutagenic and mutagenic base pairs opposite the m6G lesion, and both were excised as efficiently as the natural G⅐T mispair (Fig. 2). This result is consistent with a thermodynamic mechanism of substrate recognition by the exonuclease. Both m6G⅐C and m6G⅐T base pairs are less stable than correct G⅐C (⌬⌬G ϭ 46 and 46.8 kJ/mol, respectively) (15). Moreover, the m6G⅐T base pair is of equivalent stability as an m6G⅐C base pair (⌬⌬G ϭ 0.8 kJ/mol), whereas the natural G⅐T base pair is substantially less stable than a G⅐C base pair (⌬⌬G ϭ 30.9 kJ/mol). Thus, the significant thermodynamic difference for correctly paired and mispaired natural bases is eliminated by formation of the m6G lesion. Interestingly, we observed that the exonuclease activity was highly effective in removing alkylated mispairs at template A and C residues, resulting in a 7-fold reduction in the MNUinduced T4 polymerase mutation frequency ( Fig. 5 and Table  II). The premutagenic lesions for these mutational events have not been defined by site-specific lesion studies. However, the nearly 200-and 40-fold increases in the frequency of C 3 T transitions and C 3 A transversions, respectively, observed for the D324A enzyme upon MNU treatment are consistent with earlier studies demonstrating misincorporation of dAMP Ͼ TMP on methylated poly(dC) templates (25). We presume that the premutagenic lesion for these events is 3meC, which is formed in significant amounts by MNU modification of singlestranded DNA (24). The mutational events at methylated A residues may result from 1meA adducts (25) and/or apurinic sites resulting from spontaneous decay of 7meA lesions. Alkylation of purine and pyrimidine base pairing positions may destabilize the primer/terminus, allowing for efficient exonuclease removal.
DNA alkylation lesions inhibit forward DNA synthesis by 3Ј 3 5Ј exonuclease-proficient DNA polymerases ( Fig. 4A; Ref. 26). Our biochemical observations (Figs. 1-3) indicate that the exonuclease activity acts as a kinetic barrier to DNA synthesis on m6G lesion-containing templates by preventing accumulation of 3Ј-terminal m6G substrates. Furthermore, the m6G⅐C and m6G⅐T intermediates are extended inefficiently by T4 polymerase. However, the presence of m6G is not an absolute barrier to DNA synthesis, since we observed 3% full-length product with wild-type T4 polymerase in the presence of 200 M dNTPs at approximately equimolar concentrations of enzyme and DNA (Table I). Moreover, the presence of accessory proteins may enhance the ability of T4 polymerase to synthesize past lesions, as has been observed for E. coli polIII (27) and pol ␦ (28) in the presence of their cognate processivity factors. Recent evidence in S. cerevisiae has led to the suggestion that specific translesion synthesis polymerases may assist in overcoming m6G inhibition (29). Together, these biochemical observations are consistent with in vivo data demonstrating that m6G-mediated cytotoxicity is not observed in the first S phase after alkylation treatment (30,31). Nonetheless, the idling we observed at the site of m6G, which we interpret as repeated cycles of base incorporation and exonuclease excision, could lead to disruption of local dCTP and TTP pools during DNA replication, resulting in a biochemical signal for cell cycle arrest or apoptosis. The most profound implication of the idling by proofreadingproficient polymerases at sites of m6G lesions is during repair DNA synthesis. In eukaryotic cells, m6G has been suggested to be cytotoxic when used as a substrate by mismatch repair enzymes, in what has been termed a "futile cycle." In this model, DNA repair of the strand opposite the lesion will produce an m6G-containing gapped DNA substrate for either polymerase ␦ or polymerase ⑀ (32). The presence of the lesion in this gap will severely inhibit the efficiency of DNA repair synthesis, since we have observed that less than 10% of m6Gcontaining templates give rise to full-length product by wildtype T4 polymerase. Our observations provide a biochemical explanation for how persistent gaps arise in genomic DNA from methylated substrates. Moreover, human cell lines have been reported that are O 6 -methylguanine-DNA methyltransferasedeficient and mismatch repair-proficient yet exhibit full resistance to the cytotoxic action of MNU (5). Extrapolating our data, these cell lines may harbor a mutation in the exonuclease domain of polymerase ␦ or polymerase ⑀ that alleviates the idling and thus the cytotoxicity of MNU.
Our study demonstrates the biochemical significance of the polymerase-associated 3Ј 3 5Ј exonuclease activity in processing alkylation adducts, particularly the m6G lesion. The idling we observed opposite m6G could have implications for improv-

TABLE II
Mutational specificity of wild-type and 3Ј 3 5Ј exonuclease-deficient T4 polymerases Independent FUdR-resistant transformants were obtained by electroporation of E. coli with DNA synthesis products from the HSV-tk in vitro forward mutational assays (Fig. 4) and selective plating as described under "Experimental Procedures." DNA treatment refers to modification of the primed ssDNA template prior to the polymerase reactions (solvent ϭ Me 2 SO; MNU ϭ 5 mM dose). DNA sequence analyses within the MluI-EcoRV target region were performed to determine the precise polymerase error causing the mutant phenotype. The frequency of each mutational event was calculated by multiplying the proportion of total mutants by the overall mutation frequency for the corresponding polymerase reaction. a Other mutations were one A 3 G, two G 3 T base substitutions, two complex mutations, and three insertions (23 total mutational events; f ϭ 1.4 ϫ 10 Ϫ4 ). b 20 total mutational events; f ϭ 120 ϫ 10 Ϫ4 . c Other mutation: one T 3 A base substitution (10 total mutational events; f ϭ 12 ϫ 10 Ϫ4 ). d Other mutations were one A 3 C, one G 3 T base substitutions, one deletion (31 total mutational events; f ϭ 890 ϫ 10 Ϫ4 ).
ing the therapeutic potential of chemotherapeutic agents. Nucleoside analogs or DNA adducts that thermodynamically destabilize the nascent DNA are expected to be substrates for the exonuclease activity; thus, the most effective cytotoxic agents will be those that produce substrates favored by the exonuclease domain over the polymerase domain, as has been observed for the prototype m6G lesion.