Replication of O6-methylguanine-containing DNA by repair and replicative DNA polymerases.

The biological consequences of O6-methylguanine (m6G) in DNA are well recognized. When template m6G is encountered by DNA polymerases, replication is hindered and trans-lesion replication results in the preferential incorporation of dTMP opposite template m6G. Thus, unrepaired m6G in DNA is both cytotoxic and mutagenic. Yet, cell lines tolerant to m6G in DNA have been isolated, which indicates that some cellular DNA polymerases may replicate m6G-containing DNA with reasonable efficiency. Previous reports suggested that mammalian pol β could not replicate m6G-containing DNA, but we find that pol β can catalyze trans-lesion replication; however, the lesion must reside in the optimal context for pol β activity, single- or short nucleotide gapped substrates. Primed single-stranded DNA templates, with or without template m6G, were poor substrates for pol β as reported in earlier studies. In contrast, trans-lesion replication by bacteriophage T4 DNA polymerase was observed for primed single-stranded DNA templates. Replication of m6G-containing DNA by T4 DNA polymerase required the gp45 accessory protein that clamps the polymerase to the DNA template. The rate-limiting step in replicating m6G-containing DNAs by both DNA polymerases tested was incorporation of dTMP across from the lesion.

O 6 -Methylguanine (m6G) 1 -DNA methyltransferase repairs m6G residues in DNA; however, 20 -30% of human solid tumor cell lines do not express this repair activity (1). Exposure of cells lacking m6G-DNA methyltransferase to alkylating agents such as MNNG results in high levels of mutations, sister chromatid exchanges, and cell death (reviewed in Ref. 2). Yet, cells unable to repair the m6G damage but tolerant to the killing activities of MNNG have been isolated (reviewed in Refs. 2, 3). As these tolerant cells remain sensitive to the mutagenic effects of alkylating agents, questions about the replication of m6G-containing DNA arise.
DNA polymerases are predicted to encounter m6G in three DNA environments: 1) in single-or short nucleotide gaps formed during short patch DNA repair, 2) in lengthy single-stranded regions formed by long patch repair, and 3) at replication forks. Single-nucleotide gaps may be produced by short patch mismatch repair activity that is normally directed to the repair of G:T mispairs that are the result of 5-methylcytosine (5mC) deamination (4). The enzyme that initiates this repair process is a DNA G:T mismatch-specific thymine-DNA glycosylase that removes the mispaired thymine from G:T DNA to generate an apyrimidinic site (5). A G:T thymine-DNA glycosylase also initiates the removal of thymine from m6G:T base pairs (6). The abasic site is further processed to generate a single-nucleotide gap across from guanine for G:T mismatches and across from m6G for m6G:T mismatches. While pol ␤ can efficiently fill in single-nucleotide and small gaps across from undamaged DNA templates (7), pol ␤ activity on single-nucleotide gaps across from m6G has not been reported. Pol ␤ replication is blocked, however, by template m6G in long singlestranded DNA templates (8).
Long patch mismatch repair also appears to act on m6G:T base pairs. The model proposed by Karran and Marinus (9), Scudiero et al. (10), Goldmacher et al. (3) and others as reviewed by Karran and Bignami (2) is that mismatch repair enzymes recognize m6G base pairs in DNA as mismatches and produce long excision tracks in the non-m6G strand. Replication of the single-stranded region can result in incorporation of dTMP across from m6G to again produce the m6G:T base pair. Several studies demonstrate that m6G templates the preferential incorporation of dTMP in vivo (11)(12)(13) and in vitro (14 -16). The resulting m6G:T mispair is again recognized by long patch repair enzymes, and the process of excision followed by replication is repeated. These futile cycles of excision and resynthesis are predicted to produce persistent discontinuities in the DNA that are thought to contribute to the cytotoxic effects of m6G in DNA. This model is supported by the observation that at least some cell lines that lack methyltransferase, but are tolerant to MNNG, also lack long patch mismatch repair (17,18).
In addition to replication associated with the repair of m6Gcontaining DNA, tolerant cells must be able to replicate chromosomes in preparation for cell division. Although several DNA polymerases have been shown to have some bypass activity, trans-lesion replication is poor (14).
In view of these numerous reports that demonstrate that m6G in DNA inhibits replication by many DNA polymerases but that m6G-containing DNA is replicated, nevertheless, in tolerant cells, experiments were designed to measure the ability of repair and replicative DNA polymerases to replicate m6G-containing DNA. The two DNA polymerases studied were 1) a mammalian repair DNA polymerase, pol ␤, and 2) a replicative DNA polymerase, bacteriophage T4 DNA polymerase with its accessory protein, the product of gene 45 (gp45) which clamps the DNA polymerase to the DNA template. The T4 DNA polymerase-gp45 complex is functionally analogous to the hu-* This work was supported by grants from the Alberta Cancer Board and the National Cancer Institute of Canada with funds from the Canadian Cancer Society. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Scientist of the Alberta Heritage Foundation for Medical Research. To whom correspondence should be addressed. Tel.: 403-492-5383; Fax: 403-492-9234; E-mail: lreha@gpu.srv.ualberta.ca. 1 The abbreviations used are: m6G, O 6 -methylguanine; pol ␤, DNA polymerase ␤; MNNG, N-methyl-NЈ-nitro-N-nitrosoguanidine; DTT, dithiothreitol; dNTP, deoxynucleotide triphosphate; gp45, T4 DNA polymerase accessory protein, the product of bacteriophage T4 gene 45. man pol ␦-PCNA complex (19). Although replication of m6Gcontaining substrates by several DNA polymerases including pol ␤ and the T4 DNA polymerase have been examined previously, studies presented here attempted to approximate in vivo conditions more closely. Thus, while pol ␤ was reported to have no trans-lesion replication activity on a single-stranded m6Gcontaining DNA template (8), we observed pol ␤ m6G translesion activity on short gapped templates, the optimal substrates for pol ␤ replication activity (7). In contrast, T4 DNA polymerase trans-lesion replication was dependent upon association with the gp45 accessory protein and on DNA templates that resemble DNA replication forks or long, single-stranded repair tracks.
In addition to the highly purified recombinant pol ␤ and T4 DNA polymerases, preparations of pol ␤ from m6G-sensitive and m6G-tolerant cell lines were tested on the m6G-containing substrates in order to determine if tolerant cells encoded a mutant pol ␤ with increased ability to replicate m6G-containing DNA. Mutant T4 DNA polymerases with reduced 3Ј 3 5Ј-exonuclease activity or increased intrinsic processivity were also examined for trans-lesion replication.

EXPERIMENTAL PROCEDURES
DNA Polymerases and Accessory Proteins-Recombinant rat pol ␤ was provided by S. Wilson and purified as described (20). Human pol ␤ was partially purified from human glioma cell lines A1235 and A1235MR4 (1,21). Both A1235 cell lines lack O 6 -methylguanine-DNA methyltransferase activity, but the A1235MR4 cells are tolerant to the cytotoxic effects of m6G in DNA, whereas the A1235 cells are hypersensitive. Extracts from 10 confluent 100-mm tissue culture plates were prepared as described (22,23). The extracts were chromatographed on a Mono Q HR5/5 column (Pharmacia Biotech Inc.) in buffer containing 40 mM Tris-HCl (pH 7.4), 0.1 mM EDTA, 10 mM ␤-mercaptoethanol, and a salt gradient from 25 to 500 mM NaCl. Fractions eluting at 240 mM NaCl contained DNA pol ␤ activity as determined by several criteria: high activity on short gapped DNA substrates, but low activity on primed single-stranded DNA templates; and inhibition by dideoxy-NTPs, but resistance to aphidicolin. A second set of fractions containing DNA polymerase activity eluted at a higher salt concentration (Ͼ0.4 M NaCl). This polymerase activity was different from pol ␤ activity in the following ways: high activity on long, single-stranded DNA templates compared to short gapped DNA substrates, inhibition by ddNTPs, and sensitivity to aphidicolin. These properties are characteristic of eukaryotic replicative DNA polymerase, pol ␣, ␦, and ⑀. The fractions containing apparent pol ␤ activity were pooled and dialyzed against 20 mM sodium phosphate buffer (pH 7.0) containing 50% glycerol, 5 mM DTT, and 0.1 mM EDTA. The dialyzed fractions were stored at Ϫ80°C. This partially purified pol ␤ fraction was used in the assays described below.
Purification and characterization of wild type and mutant recombinant bacteriophage T4 DNA polymerases have been described (24 -26). The T4 DNA polymerase accessory protein, the gene 45 protein (gp45), was purified from the expression vector constructed by Lin et al. (27) and purified by a modification of the procedure described by Morris et al. (28). A crude extract was prepared, and the extract was applied to a 75-ml, 5-cm diameter Q-Sepharose column. The gp45 bound to the column, and fractions containing gp45 eluted at 210 mM NaCl from a gradient of 50 -500 mM NaCl in Ao buffer (40 mM Tris-HCl (pH 7.4), 0.1 mM EDTA, 10 mM ␤ mercaptoethanol, and 10% glycerol). These fractions were pooled and further purified by chromatography through a 55-ml hydroxylapatite (Bio-Rad, HTP) 5-cm diameter column in buffer containing 50 -200 mM potassium phosphate (pH 7.0), 0.1 mM EDTA, 10 mM ␤ mercaptoethanol, and 10% glycerol. Fractions containing gp45 eluted at 135 mM potassium phosphate. The single major contaminating band was then removed by chromatography on a 1-ml Mono-Q column (Pharmacia) with a 20-ml linear gradient of 50 -350 mM NaCl in Ao buffer. The gp45 eluted at 200 mM NaCl as an apparently homogeneous protein.
DNA Substrates-Gapped and single-stranded DNA templates were made using combinations of synthetic oligonucleotides (Fig. 1). The 45-mer template strand had either a G or m6G residue as indicated. A three-nucleotide gapped substrate was formed by annealing a 24-mer that was complementary to the template DNA sequence upstream from the m6G residue plus an 18-mer primer that was complementary to template sequence downstream from the m6G residue (Fig. 1, panel A).
A single-nucleotide gapped substrate was prepared by annealing a 20-mer primer (Fig. 2, panel B) in place of the 18-mer oligonucleotide. A two-nucleotide gapped substrate that allowed measurement of incorporation across from m6G or G as well as extension was prepared by annealing the 20-mer primer upstream and a 23-mer downstream of the template m6G ( Fig. 1 panel C). "Gaps" were flanked by a 3Ј-OH group for extension of the 18-or 20-mer primers and a 5Ј-phosphate (P) in the upstream 23-or 24-mers. Primer extension DNA templates did not have a downstream oligomer (Fig. 1B).
The DNAs used were prepared by automated DNA synthesis. The m6G-containing DNA was synthesized by the Regional DNA Synthesis Laboratory at the University of Calgary as described by Sibghat-Ullah and Day (6). Approximately 65-70% of the m6G-containing template could be replicated and templated the preferential incorporation of dTMP in the complementary strand; the remainder of this template was refractory as a DNA template in any of the in vitro replication assays. We inferred from these observations that 65-70% of the template was authentic m6G-containing DNA.
The 18-nucleotide primer (Fig. 1, A and B) and the 20-nucleotide primer (Fig. 1, B and C) were labeled with 32 P at the 5Ј-end using T4 polynucleotide kinase (Pharmacia) and standard reaction conditions (29). The labeled primers were annealed with the other oligonucleotides illustrated in Fig. 1 by combining equal concentrations of each oligonucleotide (45 nM) in 50 l of 50 mM Tris-HCl (pH 8.0) buffer containing 10 mM dithiothreitol (DTT). The mixtures were heated at 70°C for 5 min and gradually cooled in the water bath to room temperature over 2 h.
Bacteriophage T4 DNA Polymerase Assay Conditions-DNA polymerase reactions were done in two steps. The first step was a 5-min preincubation at 37°C with all reaction components except for Mg 2ϩ . The final concentrations of reaction components were 16.7 mM (NH 4 ) 2 SO 4 , 72 mM Tris-HCl (pH 8.8), 200 g/ml bovine serum albumin, 100 M dNTPs, 1.5 mM DTT, 8.3% glycerol, 0.15 mM EDTA, 3.75 nM DNA substrate, and 6.25 nM T4 DNA polymerase. For reactions with gp45, the same reaction conditions were used with the addition of 7.5% polyethylene glycol (Carbowax PEG-8000, Fisher) and 8 M gp45 monomer. The ratio of DNA substrate to DNA polymerase to gp45 hexamer was 0.6:1.0:213. The high concentrations of gp45 and polyethylene glycol allowed formation of the DNA polymerase-gp45 complex without the additional accessory proteins (30).
The second step was to start the reactions by the addition of Mg 2ϩ , or Mg 2ϩ and heparin, to give a final concentration of 6.7 mM MgCl 2 . Heparin, when added, was at a final concentration of 0.1 mg/ml and was sufficient to trap all free DNA polymerase. The total reaction volume was 12 l. The conditions for producing an effective heparin trap for T4 DNA polymerase were described previously (26). The reactions were incubated at 37°C for the indicated times. The reactions were stopped with an equal volume (12 l) of gel loading solution (92% deionized formamide, bromphenol blue, and xylene cyanol blue markers in TBE (100 mM Tris-HCl (pH 8.3), 100 mM boric acid, 2 mM EDTA)) and placed on ice.
Reaction products were separated by electrophoresis through 15% polyacrylamide gels with 8 M urea in TBE buffer using 45 watts constant power for 80 min. The 32 P-labeled products were visualized by autoradiography with Kodak X-Omat AR film. Reaction products were quantitated by densitometry using a Bio-Rad model GS-670 imaging densitometer. Several film exposures were used to keep the signal within the linear range of the film. Analysis of image data was done using the Molecular Analyst software program (version 1.4) provided by Bio-Rad. The rate constant was determined as the slope of the intensity versus time.
Pol ␤ Assay Conditions-DNA pol ␤ reactions were also done in two steps. Reaction components, 35 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1.5 mM DTT, 200 g/ml bovine serum albumin, 100 M dNTPs, 6.3% glycerol, 3.75 nM DNA substrate, and 6.25 nM pol ␤, were preincubated for 5 min at 37°C. The reactions were then started with either Mg 2ϩ or a mixture of Mg 2ϩ and heparin at the same concentrations used for T4 DNA polymerase. The conditions for producing an effective heparin trap for pol ␤ were found to be identical to the conditions used for the T4 DNA polymerase. Reactions products were processed as described for the T4 DNA polymerase reactions. The concentration of pol ␤ activity in preparations from human glioma cells was determined by comparing activity with the homogeneously purified rat pol ␤ and the nonmodified gapped DNA substrate.

RESULTS
Pol ␤ Activity on m6G-containing DNA Substrates-In studies of short patch repair of m6G-T base pairs, we observed efficient replication of single-nucleotide gaps across from template m6G by partially purified pol ␤ from human glioma cell lines (discussed below). This result was unexpected from earlier reports of pol ␤ activity on an m6G-containing singlestranded DNA substrate that demonstrated that pol ␤ could not incorporate nucleotides across from m6G (8). These earlier studies, however, did not use what is now known to be the optimal DNA substrate for pol ␤ activity, substrates with small gaps (7). In order to verify that pol ␤ can incorporate nucleotides across from m6G in gapped substrates but not on substrates with long single-stranded regions, a homogeneous preparation of recombinant rat pol ␤ was tested with the DNA substrates shown in Fig. 1.
One of the gapped substrates used was the three-nucleotide gapped substrate illustrated in Fig. 1, panel A. With no m6G in the template, 44% of the primer was fully extended in 15 s to the ϩ3 position, and significant displacement synthesis was observed to the ϩ4 position (Fig. 2, panel A, lane 1). Heparin was used to measure the processivity of pol ␤ DNA replication. When pol ␤ dissociates from the DNA template, the enzyme is effectively trapped by heparin. A burst of replication was observed in the presence of the heparin trap; 12% of the primer was fully extended without dissociation (Fig. 2, panel A, lane 2). When these experiments were repeated with the m6G-substrate (lanes 3 and 4), most of the primer extended terminated one nucleotide before the template m6G residue, with or without the heparin trap. Thus, the burst of processive DNA replication by pol ␤ was limited primarily to a two-nucleotide extension in the presence of template m6G. Without the heparin trap, the enzyme could cycle back onto the ϩ2 extended primer for repeated attempts at trans-lesion replication (Fig. 2, panel  A, lane 3).
Similar results were observed with single-and double-nucleotide gapped substrates (data presented below). For all short gapped substrates, with or without m6G, efficient replication by pol ␤ required a 5Ј-terminal phosphate on the downstream oligomer (data not shown). Requirement for the downstream 5Ј-phosphate for optimal pol ␤ activity was reported previously (7).
Pol ␤ activity was reduced considerably with single-stranded DNA templates (Fig. 1B). Reaction conditions were the same as those used for the gapped DNA substrates. With no m6G in the primer extension template, only 7% of the primer was extended (Fig. 2B, lane 1). Pol ␤ replication on the single-stranded substrate was also dissociative as indicated by the "ladder" pattern of products (Fig. 2B, lane 1) and the absence of replication with the heparin trap (Fig. 2B, lane 2). With the m6G substrate (Fig.  2B, lane 3), 7% of the primer was again extended, but all of the products terminated one position before the template m6G residue. No trans-lesion replication was observed with the heparin trap (Fig. 2B, lane 4).
Because pol ␤ first converted the three-nucleotide gapped substrate into a single-nucleotide gapped substrate before replication of template m6G (Fig. 2A, lane 4), the specificity of nucleotide incorporation across from m6G was investigated with a preformed single-nucleotide gap positioned directly across from either G or m6G. This substrate was formed by annealing the 24-mer plus the 20-mer primer to the template 45-mer (Fig. 1). The reactions contained either 100 M dCTP, 100 M dTTP, or both nucleotides each at 100 M. The rate of incorporation of dTMP was 20-fold higher for the m6G template than for the nonmodified G template that verifies the mutagenic mistemplating of m6G (Table I and Fig. 3). Incorporation of dCMP across from m6G was also detected, but the rate was 5-fold less than that of dTMP.
In the cell, however, both dCTP and dTTP as well as dATP and dGTP are present. The rate of nucleotide incorporation in reactions with both 100 M dCTP and dTTP and the singlenucleotide gapped substrate was comparable with the rate of dTMP incorporation in reactions that contained only dTTP (Table I and Fig. 3). Whether dCMP or dTMP was incorporated can be determined by the mobility of the ϩ1 product in DNA   FIG. 1. DNA substrates. Synthetic oligonucleotides were prepared as described under "Experimental Procedures." The 45-mer template strand contained either m6G or guanine at the position indicated by the *. The 18-or 20-mer primers were labeled at the 5Ј-end with 32 P. For gapped substrates, the downstream 23-or 24-mer was phosphorylated at the 5Ј-end. The 5Ј phosphorylation is indicated by an underline.
sequencing gels. The 3Ј-terminal nucleotide has a significant effect on mobility (31). The primer extended with dCMP has a greater mobility than the primer extended with dTMP (Fig. 3). In reactions with both 100 M dCTP and dTTP, primarily the slower moving 3Ј-dTMP-terminated primer was detected (Fig. 3).
Replication of the two-nucleotide gapped substrate can be considered in two steps: 1) extension to the ϩ1 position required incorporation of dCMP or dTMP opposite template m6G, and 2) replication to the ϩ2 position required extension of the m6G:C or m6G:T primer template and incorporation of dGMP opposite template C (Fig. 1, panel C). In reactions with either dCTP or dTTP but not dGTP, the rates of either dCMP or dTMP incorporation across from m6G were slower with the two-nucleotide gapped substrate than for the single-nucleotide gapped substrate which suggests that the single-nucleotide gapped substrate may be the optimal substrate for pol ␤ activity (  Fig. 1, m6G is indicated by a * over the G residue. All four dNTPs, each at 100 M, were present. Reactions were incubated for 15 s, and the products were resolved by gel electrophoresis. The gels were dried and exposed to x-ray film. The amount of gap-filling or primer extension was quantitated by densitometer scanning. The values for extension are relative to the amount of labeled DNA. Reactions without heparin are indicated by a minus sign, whereas reactions with heparin are indicated by the plus sign. Detailed reaction conditions are described under "Experimental Procedures."

TABLE I
Replication of m6G DNA substrates by DNA pol ␤ Rates of nucleotide extension were determined as described under "Experimental Procedures." Reactions were incubated for 15 s, 1, 10, and 20 min. Rates were determined from densitometric scanning of exposed x-ray films (note Fig. 3 a The DNA substrates were the single nucleotide gapped substrate with either G or m6G in the template position across from the gap or the two nucleotide gapped substrate illustrated in Fig. 1, panel C. b Full extension was observed by the earliest time point in reactions with the nonmodified template and the correct nucleotide, dCTP. The extension rate of 1 s Ϫ1 on nonmodified templates is from published reports and is included as a reference point to compare reactions with the m6G template (8).
c For the two-nucleotide gapped substrate, dCTP and dGTP are the correct nucleotides for incorporation (Fig. 1, panel C).
across from m6G was slowed from 0.004 s Ϫ1 for the singlenucleotide gapped substrate to 0.001 s Ϫ1 for the two-nucleotide gapped substrate (Table I). Adding the next correct nucleotide, dGTP, in reactions with dTTP did not appreciably affect the rate of trans-lesion replication (0.004 s Ϫ1 compared with 0.006 s Ϫ1 ), but the addition of dGTP to reactions with dCTP slowed the rate of trans-lesion replication by about 3-fold (0.0003 s Ϫ1 compared with 0.001 s Ϫ1 ).
The rate-limiting step in filling the two-nucleotide gapped m6G substrate was incorporation of dTMP or dCMP across from m6G and not extension of either the T:m6G or C:m6G base pair since the rates of ϩ1 and ϩ2 extension were similar, 0.0003 and 0.0002 s Ϫ1 for the dCTP/dGTP reaction and 0.004 and 0.003 s Ϫ1 for the dTTP/dGTP reaction. Thus, although incorporation of a nucleotide opposite m6G was slow, most of this product was extended relatively rapidly. This result was in contrast to replication of the nonmodified G template with dTTP and dGTP. Both misincorporation of a dTMP opposite template G and extension of the T:G mispair were kinetically slow steps.
Pol ␤ from Human Cancer Cell Lines-Having confirmed that recombinant rat pol ␤ can replicate m6G residues located in short gapped DNA substrates, albeit slowly and requiring recycling of the enzyme on and off the primer template, we re-examined pol ␤ purified from human glioma cell lines. Both cell lines were deficient in m6G-methyltransferase activity, but while the parental cell line was highly sensitive to the cytotoxic effects of alkylating agents (cell line A1235), a cell line isolated from the parent (A1235MR4) was tolerant to alkylating agents. Two hypotheses to explain the tolerance were tested: 1) to determine if long patch mismatch repair was defective in the tolerant cell line as observed by Branch et al. (17) and Kat et al. (18) in studies of other m6G-tolerant cell lines, and (2) to determine if some aspect of short patch repair was altered in the tolerant cell line. We report here on studies designed to test the second hypothesis.
Both the A1235 m6G-sensitive cell line and the A1235MR4 m6G-tolerant cell line catalyze removal of thymine from m6G:T base pairs (6). Thus, single-nucleotide gaps are expected to occur in both sensitive and tolerant cell lines in response to attempted correction of m6G:T base pairs. Tolerance could arise if gap-filling activity by pol ␤ and subsequent ligation were more effective in tolerant cells than in sensitive cells. Partially purified preparations of pol ␤ from sensitive and tolerant cell lines were tested on the three-nucleotide gapped substrate (Fig. 4). When similar amounts of the partially purified human pol ␤ (lanes 1 and 2) and recombinant rat pol ␤ (lanes 3 and 4) were tested on the m6G-containing three-nucleotide gapped substrate, good gap-filling activity was de-tected for both pol ␤ preparations. In comparison to the 15-s reactions shown in Fig. 2, panel A, more gap-filling activity was observed in the longer 1-and 10-min reactions. Similar results were observed with the single-nucleotide gapped substrate. No differences were detected between pol ␤ preparations from the m6G-sensitive or -tolerant cell lines, but less displacement synthesis to the ϩ4 position was observed for the partially purified pol ␤ preparations than for the highly purified rat pol ␤ (compare Fig. 4, lanes 2 and 4).
Bacteriophage T4 DNA Polymerase Activity on m6G-containing DNA Substrates-Strong gap-filling activity on the threenucleotide nonmodified gapped substrate was observed for wild type T4 DNA polymerase, but T4 DNA polymerase was less processive than pol ␤. In the presence of the heparin trap, just 3% of the primer was extended to the ϩ3 and ϩ4 positions by T4 DNA polymerase (Fig. 5, panel A), whereas 13% of the primer was extended by pol ␤ (Fig. 2A, lane 2). High processivity was observed, however, in reactions with the T4 DNA polymerase-gp45 complex (Fig. 5, panel A). Products longer than three nucleotides were produced, particularly in reactions with the T4 DNA polymerase-gp45 complex. In separate reactions, no degradation of the downstream oligonucleotide was observed. Thus, the T4 DNA polymerase-gp45 complex can catalyze significant displacement replication. Only limited gap filling was observed with the m6G-gapped substrate, and this activity required gp45 (Fig. 5, panel B). As observed for pol ␤ (Fig. 2), most replication terminated one nucleotide before template m6G at the ϩ2 position.
Two-mutant T4 DNA polymerases were also tested with the three-nucleotide gapped substrates, the D112A ϩ E114A-T4 Reaction products were resolved by gel electrophoresis and quantitated by densitometry (see Table I). In lane 1 for each set of reactions, the incubation time was 15 s at 37°C; in lane 2 the incubation time was 1 min; in lane 3 the incubation time was 10 min; and in lane 4 the incubation time was 20 min. Labeled primer alone was loaded into the two lanes between the dTTP reaction set and the reaction set with dTTP plus dCTP. Note that the ϩ1 products terminating with dCMP ( lanes  3-4) in the dCTP reaction set migrate slightly faster than ϩ1 products terminating with dTMP (lanes 2-4) in the dTTP reaction set. The ϩ1 products in the reaction set with both dCTP and dTTP have the mobility of the slower migrating products with 3Ј-terminal dTMP.  lanes 1 and 2) and rat pol ␤ (lanes 3 and 4) were incubated with the three-nucleotide gapped m6G substrate used in Fig. 2. Equal units of human and rat DNA polymerase activity were added to reactions as determined by assays with the nonmodified gapped substrate. All four dNTPs were present at 100 M. Reactions in lanes 1 and 3 were incubated for 1 min, and reactions in lanes 2 and 4 were incubated for 10 min. Preparations of human pol ␤ isolated from A1235 or A1235MR4 cell lines were identical in all assay conditions tested. The exonuclease activity detected in the pol ␤ purified from the glioma cells is a contaminating activity that does not seem to affect replication of m6G-containing DNA but may affect the amount of ϩ4 product that is longer than the three-nucleotide gap and, thus, must arise from some type of displacement synthesis.
DNA polymerase which lacks 3Ј 3 5Ј-exonuclease activity (25) and the L412M-T4 DNA polymerase which has increased intrinsic processivity (26). The mutants were no more active under these conditions than the wild type enzyme, even in the presence of gp45 (data not shown). The wild type and mutant T4 DNA polymerases were also inactive on single-nucleotide gapped substrates with template m6G.
In contrast to the poor primer extension activity observed for pol ␤ on the single-stranded DNA template (Fig. 2, panel B), T4 DNA polymerase was highly active and replicated the fulllength (ϩ27) product even in the presence of the heparin trap (Fig. 5, panel C). As expected, the T4 DNA polymerase-gp45 complex was more processive than T4 DNA polymerase alone (Fig. 5, panel C), and this processivity was required for bypass replication of template m6G (Fig. 5, panel D). In the absence of gp45 most primer extension stopped at position ϩ2, one position before template m6G.
The exonuclease-deficient D112A ϩ E114A-DNA polymerase and the L412M-DNA polymerase which has increased intrinsic processivity were also tested with the single-stranded m6G DNA substrate. Neither of the mutant DNA polymerases were able to produce any more full-length product than detected for the wild type enzyme during the 15-s incubation or with the heparin trap (Fig. 6). As observed for wild type T4 DNA polymerase, the addition of gp45 was required by the mutant DNA polymerases for trans-lesion replication. Although the twomutant DNA polymerases proved to be no better than wild type T4 DNA polymerase in incorporating a nucleotide across from template m6G, the mutant DNA polymerases did not degrade the primer. Note that about 40 -50% of the primers were degraded to smaller products by the wild type T4 DNA polymerase, with or without gp45, during the 15-s incubation (Fig. 5,  panel D).
As observed for the wild type T4 DNA polymerase-gp45 complex, the exonuclease-deficient D112A ϩ E114A-DNA polymerase complex with gp45 also did not produce any ϩ3 product, with or without the heparin trap (Fig. 6). Similarly, no ϩ3 product was detected for the L412M-DNA polymerase (data not shown). These observations indicate that most replication stopped at ϩ2 before the template m6G, and, if a nucleotide was incorporated opposite m6G, the DNA polymerase did not dissociate from the DNA template which would produce ϩ3 product. For wild type T4 DNA polymerase, the T:m6G primer terminus may be extended to produce full-length product or degraded to produce the ϩ2 product. For the exonucleasedeficient D112A ϩ E114A-DNA polymerase, degradation was not a possibility, but the same amounts of ϩ2 and full-length products were produced as observed for the wild type T4 DNA polymerase-gp45 complex (Fig. 6). Thus, incorporation of a nucleotide opposite template m6G is coupled to the subsequent extension reaction, at least under these reactions conditions with 100 M dNTPs.
The exonuclease-deficient D112A ϩ E114A-DNA polymerase was used to measure specificity of nucleotide incorporation FIG. 5. Bacteriophage T4 DNA polymerase activity on gapped and single-stranded DNA substrates. Bacteriophage T4 DNA polymerase, with or without the accessory protein gp45, was incubated with the three-nucleotide gapped substrate, with template G (panel A) or with template m6G (panel B). Analogous experiments were performed with the single-stranded DNA substrate with template G (panel C) or with template m6G (panel D). All four dNTPs, each at 100 M, were present. Reactions were incubated for 15 s, and the products were resolved by gel electrophoresis. The gels were dried and exposed to x-ray film. The amount of gap-filling or primer extension was quantitated by densitometer scanning. Primers degraded by the 3Ј 3 5Ј-exonuclease activity are summed together and listed as exo. The values for primer extension and degradation are relative to the amount of labeled DNA. Reactions without heparin are indicated by a minus sign, whereas reactions with heparin are indicated by the plus sign. Detailed reaction conditions are described under "Experimental Procedures." across from m6G (Table II). In reactions with either dCTP or dTTP, dTMP was incorporated at a faster rate opposite m6G than dCMP in the absence of gp45 (0.01 s Ϫ1 compared to 0.004 s Ϫ1 ); however, in the presence of gp45, dCMP and dTMP were incorporated at comparable rates. When both dCTP and dTTP were present at equal concentrations (100 M), however, dTMP was incorporated preferentially as demonstrated for pol ␤. Thus, although in reactions with either dCTP or dTTP either nucleotide was incorporated with similar effectiveness by the T4 DNA polymerase-gp45 complex, dTMP was incorporated preferentially when a mixture of the two nucleotides was present. Thus, the addition of the gp45 processivity subunit does not diminish the ability of the DNA polymerase to distinguish between dTTP or dCTP if both nucleotides are present. In experiments not shown here, dTMP was still incorporated preferentially even when dCTP was present in 5-fold excess (100 M dCTP and 20 M dTTP). DISCUSSION Previous experiments indicated that DNA polymerases replicated m6G-containing substrates with difficulty (Escherichia coli DNA pol I, T4 DNA polymerase) or not at all (mammalian pol ␤). Since certain cells are tolerant to considerable amounts of m6G modified DNA, some cellular DNA polymerases must be able to replicate m6G-containing DNA reasonably well, or tolerant cells have one or more mutant DNA polymerases with enhanced m6G trans-lesion replication activity.
Experiments presented here confirmed that pol ␤ cannot replicate past an m6G residue located in a single-stranded region of DNA (Fig. 2, panel B) as would occur at a replication fork or in a long patch mismatch repair track (8). The optimal DNA substrates for pol ␤, however, are gapped substrates with single-stranded regions from one to six nucleotides (7). With small gapped DNA substrates, pol ␤ incorporated dTMP across from m6G if the DNA on the 5Ј-side of the gap was phosphorylated (Fig. 2, panel A). With prolonged incubation, significant gap-filling replication was observed (Fig. 4). T4 DNA polymerase also catalyzed an efficient gap-filling reaction on the threenucleotide gapped substrate, but gap filling was generally not processive (Fig. 5A). The T4 DNA polymerase-gp45 complex was highly processive, but on the gapped substrate only a low level of nucleotide incorporation across from m6G was detected (Fig. 5, panel B). The T4 DNA polymerase-gp45 complex was even less active in m6G-trans-lesion replication on single-nucleotide gapped substrates which indicates that pol ␤ is better suited than T4 DNA polymerase for gap-filling replication with template m6G.
Unlike pol ␤, the phage T4 DNA polymerase alone, and particularly when complexed with gp45, catalyzed efficient replication of the primed single-stranded DNA substrate (Fig. 5, panel C). The gp45 protein clamps the T4 DNA polymerase to DNA and increases the processivity of DNA replication as demonstrated in reactions with the heparin trap (Fig. 5, panel  C). Trans-lesion replication was observed on the singlestranded DNA substrate and was dependent on the gp45 accessory protein (Fig. 5, panel D).
In reactions with exonuclease-proficient and -deficient T4 DNA polymerase-gp45 complexes, there was no accumulation of product that terminated at the lesion (ϩ3) (Fig. 6). Primers terminated one position before the template m6G (ϩ2) or fulllength (ϩ27) product was observed (Fig. 5, panel D). This observation indicates that once a nucleotide was incorporated opposite m6G, the DNA polymerase was committed to extension of the primer terminus. Thus, the rate-limiting step in trans-lesion replication was incorporation of a nucleotide across from m6G. Incorporation of a nucleotide opposite template m6G was also the rate-limiting step for pol ␤, but pol ␤ produced a small amount of product that terminated opposite template m6G (Table I). Both pol ␤ and T4 DNA polymerase incorporated dTMP preferentially across from template m6G (Tables I and II and Fig. 3).
Replication bypass of template m6G by either pol ␤ or the T4 DNA polymerase-gp45 complex was reduced under reaction conditions containing the heparin trap, which indicates that the DNA polymerases frequently dissociated from the template when template m6G residues were encountered (Figs. 2, 5, and 6). Dissociation was observed even when the T4 DNA polymerase was clamped to the DNA by the gp45 protein (Fig. 5,  panel D, and Fig. 6). Together, these observations suggest that m6G residues in various DNA contexts are not efficiently replicated. The reduced need to replicate m6G DNA, as would occur in long patch mismatch repair defective cells, may then contribute to tolerance to this type of DNA damage (17,18). FIG. 6. Replication of m6G-containing DNA by an exonucleasedeficient T4 DNA polymerase. Wild type and the exonuclease-deficient D112A ϩ E114A-T4 DNA polymerases were assayed under reaction conditions described for Fig. 5D. The DNA polymerases were complexed with gp45 and preincubated with the m6G single-stranded DNA template and dNTPs. Reactions were initiated with Mg 2ϩ and heparin. Lanes 2 and 4 contain the non-extended primer. The reaction with wild type T4 DNA polymerase ϩ gp45 is in lane 1; the reaction with the exonuclease-deficient D112A ϩ E114A-T4 DNA polymerase ϩ gp45 is in lane 3. The reactions with the wild type and exonucleasedeficient T4 DNA polymerases were nearly identical. The L412M-DNA polymerase activity was also the same (data not shown). Approximately 30% of the primer was extended to the ϩ2 position and 10% of the primer was extended fully (lanes 1 and 3). DNA polymerase Rates of nucleotide extension were determined with the 3Ј 3 5Ј exonuclease-deficient D112A ϩ E114A-DNA polymerase. The DNA template was the 20-mer primed single-stranded DNA illustrated in Fig. 1, panel B. Reaction rates for the correct incorporation of dCMP across from template G were not measured but are presumed to be approximately 100 s Ϫ1 (34). Extension rates were determined as described for Table I There is still a question, however, as to how chromosomes containing large quantities of m6G are replicated even if cells lack long patch mismatch repair. Although we report here replication of template m6G residing in short gaps by pol ␤ and in single-stranded DNAs by the T4 DNA polymerase-gp45 complex, replication is not robust. Mutant T4 DNA polymerases with reduced exonuclease activity or increased processivity may assist trans-lesion replication by allowing repeated attempts at nucleotide incorporation across from m6G without concurrent primer degradation. When the wild type T4 DNA polymerase-gp45 complex encountered template m6G, bypass was observed about 25% of the time and dissociation about 75%. (Compare 32% of ϩ2 product formed with 10% full-length product in the reaction with heparin in Fig. 5, panel D.) If the T4 DNA polymerase was allowed to reassociate with the primer terminus, which occurred in reactions without the heparin trap, primer degradation was observed. About 44% of the primers were degraded to lengths shorter than the starting 18-mer (Fig. 5, panel D). In reactions with the mutant T4 DNA polymerases complexed with gp45, again most primer extension stopped at ϩ2, one position before template m6G; however, the primers were not degraded. The absence of primer degradation allows more opportunity by the mutant DNA polymerases to attempt trans-lesion replication.
Although pol ␤ from sensitive and tolerant human glioma cell lines were equally proficient in gap replication with template m6G, we did not measure another activity of pol ␤, the ability to excise 5Ј-terminal deoxyribose phosphate residues (32). The association of 5Ј-DNA-deoxyribophosphodiesterase activity with DNA polymerase activity makes pol ␤ an ideal enzyme for short patch DNA repair, and recent studies demonstrate that pol ␤ functions specifically in base-excision repair in vivo (33). It would be informative to test if 5Ј-DNA-deoxyribophosphodiesterase activity is affected by m6G.
There may be other factors in the cell besides DNA polymerases and mismatch repair that could contribute to the tolerance of m6G in DNA. Because the "tolerant" cells studied are cells adapted for growth in culture and are also frequently derived from malignant or tumor cells, several differences compared to primary "normal" cells are expected. Extension of the S phase of the cell cycle, for example, could contribute to tolerance by allowing a longer time for chromosome replication; hence, a longer time for trans-lesion replication. Along with an extended S phase, apoptosis would also need to be prevented or the onset delayed. Another possibility is that standard methods of chromosome replication in tolerant cells may be supplemented by other mechanisms such as replication restart and recombinational repair that may assist replication of damaged chromosomes. Because of the use of alkylating agents in che-motherapeutic regimes, it is necessary to continue to pursue the molecular basis of tolerance to m6G. These studies, in turn, will provide important insights into differences between normal and cancer cells.