Lack of Correlation between in Vitro and in Vivo Replication of Precisely Defined Benz[a]anthracene Adducted DNAs*

Like other polycyclic aromatic hydrocarbons, certain metabolites of benz[a]anthracene have been implicated as potent carcinogens. These effects are thought to be caused by the covalent binding of these species to nucleophilic groups on the bases of DNA. To address the molecular mechanisms by which these molecules induce mutations, this study employed oligonucleotides containing four site-specific N6adenine-benz[a]anthracene diol epoxide adducts. Using a prokaryotic in vivo replication system, we have shown that both non-bay region anti-trans-benz[a]anthracene adducts are essentially nonmutagenic. In contrast, the bay region anti-trans-benz[a]anthracene lesions do induce point mutations at the adduct site. The mutagenic frequency of these bay region lesions is dependent on the stereochemistry about the adduct-forming bond, as well as the strain of Escherichia coli in which they are replicated. The ability of the bacterial replication machinery to bypass the lesions does not correlate with the differences observed in their mutagenesis. While both non-bay region adducts are readily bypassed in vivo, the bay region adducts are both blocking to approximately the same degree. In vitro studies of the interactions of E. coli DNA polymerase III with these adducts have also been undertaken to further dissect the relationship between adduct structure and biological activity.

Under normal growth conditions, Escherichia coli DNA polymerase III and its accessory factors constitute an extremely efficient enzyme system. The holoenzyme complex can incorporate up to 1000 nucleotides/s, and estimates of replication fidelity are in the range of 1 mistake/10 10 bases copied (1). This extremely high fidelity is a result of multiple individual reactions, all of which have intrinsically high fidelity: the polymerization step, 3Ј 3 5Ј proof-reading, and postreplicative mismatch correction. Factors that compromise the fidelity of any of these steps can lead to increased mutation rates and potentially to cell death or transformation/cancer in higher eukaryotes. Polycyclic aromatic hydrocarbons (PAHs) 1 are ubiq-uitous carcinogens that are thought to exert their effects by decreasing polymerase fidelity.
Benz [a]anthracene (BA) is one of the many polycyclic aromatic hydrocarbons which are formed during the incomplete combustion of organic materials, including gasoline, and tobacco. It is found in particularly high concentrations in coal tar and the wood preservative creosote, and it has been shown that levels of BA increase 7-to 8-fold in meat that is grilled over charcoal (2).
Like other PAHs, BA undergoes metabolic activation to form vicinal diol epoxides. The mutagenicity of numerous BA metabolites has been determined (3)(4)(5), and it is widely accepted that the biological effects of BA as well as other PAHs result from the reaction of bay region diol epoxides with the bases of DNA. A bay region diol epoxide is defined as a vicinal diol epoxide in which the epoxide ring encompasses one of the bay region carbon atoms (see Fig. 1). In comparison with other diol epoxides of BA, the bay region diol epoxides are 10 -35 times more mutagenic in certain bacterial strains and 65-125 times more mutagenic in V79 -6 Chinese hamster lung cells (4 -6).
Animal studies implicate BA as a mild carcinogen, especially in comparison with such compounds as benzo[a]pyrene and dibenz[a,h]anthracene (3,7). The low rate of tumorigenicity of BA is thought to be due, at least in part, to the low level of metabolism of BA to the bay region diol epoxide (8 -11).
Exactly what it is about the bay region diol epoxides that leads to the biological effects seen is still unclear. The initial bay region theory of carcinogenesis was based on calculations which showed that the epoxide was particularly reactive at this position and could easily open to form a strongly electrophilic carbocation that would then be attacked by the nucleophilic groups in DNA. It was suggested that this increased reactivity of the bay region diol epoxides was the basis for the increased mutagenicity/carcinogenicity of these compounds (12). As has been pointed out, however, increased reactivity alone cannot account for the increased mutagenicity of these compounds since other (non-bay region) diol epoxides also form covalent DNA adducts with approximately the same efficiency (1). Further, some of the most potent bay region diol epoxides turn out to be some of the least reactive.
An alternative explanation for the observed dichotomy between the bay region and the non-bay region diol epoxides of PAHs is that the adducts which they form adopt conformations that are recognized in different ways by the cellular repair/ replication machinery and that these differences account for the varied mutagenicity of the adducts (13). Clearly, these questions remain to be answered, and we consider site-and stereo-specific BA-DNA lesions to be ideal for determining structure-function relationships of adducts resulting from mutagenic versus non-mutagenic PAH metabolites.
We initiated a project to study the biological effects of sitespecific DNA adducts that form when certain PAH metabolites interact with DNA. A primary goal of this research, in conjunction with the structural studies which are being carried out by Dr. Michael Stone at Vanderbilt University, is to determine the basis for the different mutagenic effects of similar compounds, all of which bind to DNA in the same basic manner.

General Synthetic Procedures
Unless otherwise noted, materials were obtained from commercial suppliers. Methylene chloride (CH 2 Cl 2 ) and pyridine were distilled from calcium hydride after reflux under a nitrogen atmosphere for 4 h. Benzene and tetrahydrofuran (THF) were distilled from sodium/benzophenone after reflux under nitrogen atmosphere for 4 h. All glassware, syringes, needles, and magnetic stirring bars used in moisture-sensitive reactions were oven dried at 140°C and stored in desiccators until used. All moisture-sensitive reactions were conducted under a nitrogen atmosphere. Reactions were monitored by thin-layer chromatography. TLC plates were purchased from EM Science (Kieselgel 60 F254, precoated, 20 ϫ 20 cm, 0.25-mm layer thickness). TLC plates were visualized, after development, under ultraviolet (UV) light (254 cm Ϫ1 ) followed by staining with anisaldehyde-sulfuric acid. TLC of oligonucleotides was carried out using the SureCheck TM system (U. S. Biochemical Corp.). Column chromatography was conducted using silica gel 60 (70 -230 mesh) from EM Science.

Synthesis of Adducted Oligonucleotides
Non-bay Region Benz[a]anthracene-adducted Oligonucleotides-The modified oligonucleotide (2 ϫ 1 mol), 5Ј-CGGA-CA*A-GAAG-3Ј (Ras codon 61), where A* is the 6-fluoropurine nucleoside, was prepared using automated solid-phase synthesis (Applied Biosystems Model 391 synthesizer) with 6-fluoro-5Ј-DMTr-3Ј-phosphoramidite (14,15). The cassettes were emptied into a 1-dram conical vial to which was added non-bay region (Ϯ)-anti-trans-benz[a]anthracene aminotriol (8.3 mg, 28 mol), dry Me 2 SO (600 l), and N,N-diisopropylethylamine (10 l). The mixture was heated for 3 days at 55°C under N 2 . After cooling to room temperature, the supernatant was removed, and the residual beads were washed with CH 3 OH (3 ϫ 1 ml) and diethyl ether (1 ϫ) and air-dried. The beads were transferred to a 3-dram vial, and concentrated NH 4 OH (3 ml) was added. The tightly closed vial was heated for 8 h at 60°C to effect cleavage from the solid support and deblocking. After cooling, the vial was allowed to stand until the excess NH 4 OH had evaporated. The supernatant was transferred to a 50-ml centrifuge tube. The beads were washed with water (2 ϫ 1 ml), and the washings were added to the original supernatant. After lyophilization, the dry materials were dissolved in H 2 O (1 ml), filtered through a 0.45-mm syringe filter, and passed through a 1.5 ϫ 44 cm column of Sephadex G-15 with elution by H 2 O. The major UV-absorbing peak was collected, concentrated, and purified on a reverse-phase column (PRP-1, 7 ϫ 305 mm, Hamilton) at 45°C using a linear gradient of 10 mM ethylenediamine acetate, pH 7.45, containing 6 -12% CH 3 CN over 40 min, flow rate 3.0 ml/min. The two diastereomers were eluted between 22 and 24 min. A second separation on the same column was performed at 25°C using 0.1 M ammonium formate, pH 6.39, containing from 27 to 30% MeOH over 20 min. After removal of solvent by lyophilization, the fractions were desalted on oligonucleotide purification cartridges (Applied Biosystems). After cooling to room temperature, the beads were treated with the procedure described above. Following passage through Sephadex G-15 and lyophilization, the adducted oligonucleotide was dissolved in H 2 O (1 ml), and purified on a reversed-phase column (YMC-ODS-AQ (4.6 ϫ 250 mm)) using a linear gradient of 100 mM ammonium formate, pH 6.5, containing 5-30% CH 3 OH over 30 min, 30% for 5 min, and 30 -100% over 10 min, flow rate 1.25 ml/min. The collected peaks were lyophilized, redissolved, and desalted on OPC cartridges according to the manufacturer directions (Applied Biosystems, Foster City, CA).

Characterization of Adducted Oligonucleotides
Capillary Gel Electrophoresis-Capillary gel electrophoresis was performed on either an Applied Biosystems Model 270, using gel-filled capillaries (Applied Biosystems) and the manufacturer's Micro-Gel buffer, or on a Beckman PACE 2000 system with the manufacturer's e-CAP gel-filled capillaries and Tris borate-urea buffer. Samples were applied at Ϫ5 kV and run at Ϫ15 kV at 30°C.
Each of the four BA-adenine lesions described above was constructed such that it was located in the center of an 11-mer oligonucleotide. The sequence of the 11-mer corresponded to the coding sequence of the human N-Ras gene from position 3 of codon 59 to position 1 of codon 63 (5Ј-C GGA CAA GAA G-3Ј). Thus, the adducted adenine residue was located at position 2 of codon 61, a biologically relevant site since point mutations in this sequence have been detected in numerous tumor types (23)(24)(25).
Adducted M13 Construction-Single-stranded M13mp7L2 DNA was kindly provided by Dr. Christopher Lawrence (University of Rochester). This M13 DNA is unique in that the single-stranded form acquires a hairpin loop structure creating an EcoRI restriction site. Singlestranded M13mp7L2 DNA was digested with EcoRI according to the supplier protocol (New England Biolabs), resulting in the production of linear M13 molecules (cut L2) with only a portion of the multiple cloning site deleted. The linear DNA was purified using Microcon-100 microconcentrators according to the manufacturer instructions (Amicon). The completeness of the digestion was confirmed by separation of the DNAs via electrophoresis through 1.5% agarose gels.
A 51-mer scaffold was employed to position each of the adducted 11-mers, or the corresponding non-adducted 11-mer, within the gap created in the cut L2. A 2.5-fold excess of each 11-mer was annealed to 51-mer scaffold on ice for 30 min. Purified cut L2 was incubated at 90°C for 2 min, cooled on ice, and added to the annealed 11-mer:51-mer mix. The resulting combination of DNA molecules was incubated with T4 DNA ligase according to the supplier instructions (New England Biolabs). The final ratio of cut L2:11-mer:51-mer scaffold was 1:5:2. Ligation resulted in the production of single-stranded circular M13 molecules containing either site-specific BA adducts or the identical nonadducted sequences. To displace the scaffold from the ligation products, a 5-fold molar excess (relative to the concentration of the scaffold) of 51-mer whose sequence is complementary to that of the scaffold was added. The DNAs were heated to 90°C for 2 min and then placed on ice for 10 min. Any portion of the ligation mixture that was not used for transformation immediately was stored at Ϫ20°C (see Latham et al. (26) for a diagrammatic representation of these procedures).
Ligation/Replication Efficiency-An aliquot of each ligation mixture was subjected to electrophoresis through a 1.5% agarose gel. This concentration of agarose efficiently separated ligated, circular M13 molecules from any remaining linear molecules. To determine the relative number of circular molecules that contained the 11-mer insert, the DNAs in these gels were subjected to Southern blot analyses (27) with a radiolabeled 17-mer probe that hybridizes specifically to the adducted 11-mer and flanking M13 sequences.

Replication of Site-specifically Adducted M13 in Vivo
Repair-deficient E. coli, AB2480 (uvrA6, recA13) were made competent according to the calcium chloride method of Sambrook et al. (27) and transformed with aliquots of the ligation mixtures described above. The resultant plaques were counted, and the numbers obtained were normalized with respect to the amount of input circular M13 that contained the 11-mer. The replication efficiency of each species of adducted M13 was expressed as a percentage of that of the unmodified M13.
Mutation Assay-The frequency and spectrum of point mutations that resulted at the position of the adduct were determined using a differential hybridization strategy described previously (26,28). SOS Induction-A saturated culture of AB1157 E. coli was diluted 1:100 with Luria-Bertani medium and grown at 37°C to an A 600 of approximately 0.2. The cells were pelleted by centrifugation at 3000 ϫ g for 5 min and resuspended in 1/6 volume of 1 ϫ M9 salts (48 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , 8 mM NaCl, 18 mM NH 4 Cl). They were then transferred to a sterile 35 ϫ 10 mm tissue culture dish. While stirring, the cells were irradiated with 254-nm light at a fluence of 40 W cm Ϫ2 for 50 s. A mock control was not exposed to 254-nm light. Two volumes of 2 ϫ YT medium were then added, and the cells were grown in the dark at 37°C for 45 min (final A 600 was approximately 0.7). Finally, cells were made competent and transformed as described above. To allow bacteriophage propagation, transformed cells were plated on Luria-Bertani medium in the presence of a 5-fold excess of UT481 E. coli (met thy ⌬(lac-pro) hsdR BamHI hsdM ϩ supDTn10/FЈ traD36 proAB lacI q Z ⌬M15). The resultant plaques were treated as those described above.

Replication of Adducted DNA in Vitro
M13 Templates-All proteins used in the in vitro replication studies (except bovine serum albumin (BSA) and proteinase K, which were purchased from Sigma and U. S. Biochemical Corp., respectively) were provided by Dr. Mike O'Donnell (Cornell University Medical College).
M13 molecules containing either site-specifically modified adenine residues, or the corresponding unadducted sequence were produced as described above. After displacing the 51-mer scaffold, ligation products were separated from any remaining linear DNAs on a 1.4% low melting point agarose gel. The gel was stained with 0.5 g/ml ethidium bromide, and the DNA was visualized under long wavelength (300 nm) ultraviolet light. The bands containing circular M13 were excised, and the agarose was digested with ␤-agarase according to the supplier instructions (New England Biolabs). Purification of the circular M13 molecules was again achieved using Microcon-100 microconcentrators. An aliquot of each sample was subjected to electrophoresis through a 1.5% agarose gel, and DNA concentrations were estimated via ethidium bromide staining. All templates were diluted to approximately 40 fmol/l.
Annealing of template to primer was carried out in a 12-l volume containing 44 mM Tris-OAc (pH 7.8); 88 mM KOAc; and 13 mM MgOAc. Templates (200 fmol each) were annealed to a 5-fold molar excess of a 21-mer primer that was designed to initiate replication 1.8 kilobases upstream of the adducted adenine. A total of 1.6 g of single-stranded binding protein (SSB) was added to coat the remaining single-stranded regions of the circular M13 templates. Samples were incubated at 80°C for 2 min and slowly cooled to 37°C. Preinitiation solution (3 l of 300 M each dTTP, dGTP, and dCTP; 110 M dATP; 110 M [␣ 32 P]dATP; 0.2 g/l beta sliding clamp; 6 mM dithiothreitol; 3 mM ATP; 6 g/l BSA; 33 mM Tris-OAc (pH 7.8); 66 mM KOAc; and 10 mM MgOAc) was then added. Finally, 0.1 g of purified E. coli DNA polymerase III* (Pol III*) was added to initiate replication. Pol III* is the complete replication complex except for the ␤ sliding clamp. The final reaction volume was 17 l. After 3 and 10 min, 8 l of each reaction were removed and added to 1.7 l of 6 ϫ stop buffer containing 3% SDS and 120 mM EDTA. Proteinase K (1 l of a solution containing 2 g in 10 mM CaCl 2 ) was then added, and incubation at 37°C was continued for 30 min. Alkaline agarose gel electrophoresis loading buffer (1.5 l of a 7 ϫ solution consisting of 21% Ficol 400 (Sigma); 350 mM NaOH; 7 mM EDTA; 0.2% bromcresol green; and 0.3% xylene cyanol) was added, and the samples were stored at Ϫ70°C until analyzed.
Products of in vitro replication were separated via electrophoresis through 0.8% alkaline agarose gels according to Sambrook et al. (27). After neutralization, the gels were dried under vacuum and placed under a storage phosphor screen for subsequent quantitation (Molecular Dynamics).

In Vitro Mutagenesis
Template Construction-Both adducted and non-adducted 11-mers were ligated at their 3Ј ends to 22-mers as described in Latham et al. (26) and Chary et al. (28). The 33-mer templates that resulted had the following sequence: 5Ј-C GGA CAA GAA GAA TTC GTC GTG ACT GGG AAA AC-3Ј. The adducted adenine was thus located at position 28 (counting from the 3Ј end). A 17-mer oligonucleotide primer with the sequence 5Ј-TCA CGA CGA ATT CTT CT-3Ј was utilized to determine the relative propensity of exonuclease-deficient E. coli DNA polymerase III (Pol III ␥ Ϫ ⑀ Ϫ Ϫ ) to insert each of the four common nucleosides opposite the adducted adenine residue in vitro.
Template ( This brought the reaction volume to 10 l. Extension was carried out at 37°C, and reactions were terminated by the addition of 2 l of 6 ϫ stop buffer (3% SDS, 120 mM EDTA). Reaction products were separated by electrophoresis through 15% polyacrylamide sequencing gels. Visualization and quantitation of reaction products was via a Molecular Dynamics PhosphorImager.

Purity of Stereochemically defined BPDE Adducts-
The usefulness of regio-and stereo-specifically adducted oligonucleotides for structural and biological studies has been well documented (29,30). For such studies, evaluation of the purity of the oligonucleotides is essential. The oligonucleotides used in the studies reported herein were purified initially by HPLC. The purity at this stage was evaluated by TLC and capillary gel electrophoresis (Fig. 2). Mass spectroscopy showed that the oligonucleotides were of the expected molecular weight. In addition, the oligonucleotides were enzymatically hydrolyzed to their component nucleosides, which were present in the predicted ratios. Inasmuch as the syntheses were carried out with racemic aminotriols, it was important to examine the stereochemical identity and purity of the PAH adduct in the oligonucleotides. This could be evaluated from the enzyme hydrolysate by chromatographic comparison with the synthetic nucleosides. In the case of the non-bay region BA-adducted oligonucleotides, cross-contamination of the 11R and the 11S isomers was seen to the extent of 15-20%; for the well resolved bay region oligonucleotides, the cross-contamination was less (ϳ10%).
Further purification of the adducted oligonucleotides was carried out by polyacrylamide gel electrophoresis, and the purity of the resulting 11-mers was determined by analysis of 32 P-end-labeled products (Fig. 3). All oligonucleotides were de-termined to be Ͼ99.4% adducted 11-mers.
Replication of Adducted M13 within E. coli-A scaffolding strategy was employed to introduce site-and stereo-specifically modified 11-mers, or their non-adducted counterparts, into the multiple cloning site of single-stranded M13 (see Ref. 28 and "Experimental Procedures"). In vivo replication of these DNAs was carried out in E. coli both with and without prior SOS induction. Plaques that contained site-specific mutations were analyzed by differential hybridization strategies (Fig. 4). Neither of the non-bay region adducts inhibited replication of the M13 (Fig. 4B) or induced a significant number of point mutations at the site of the adduct (Fig. 4C). SOS induction of the cells prior to transformation did not affect the replication of these adducts (Fig. 4B). The M13 molecules that carried the non-bay region 11R-BA lesions were replicated at least as efficiently as their unadducted counterparts (replication efficiency ranged from 92 to 144% that of the unadducted M13 ( Fig. 4B)). Replication of non-bay region 11S-BA adducted DNA was also uninhibited within bacteria (replication efficiency range of 94 -325% (Fig. 4B)). Of the Ͼ18,000 plaques that were generated from these non-bay region BA adducted M13 molecules (in AB2480, approximately 10,000 and 6000 plaques were screened for the non-bay region 11R-BA and the non-bay region 11S-BA adducted DNAs respectively, and approximately 2000 (11R-BA) and 1200 (11S-BA) plaques were screened under SOSϩ conditions in AB1157 cells), only four were positive for point mutations at the site of the lesion (see Fig. 4C). Therefore, we conclude that these lesions are non-mutagenic when processed within E. coli.
In contrast, the bay region BA adducts did inhibit M13 replication as evidenced by decreased plaque forming abilities (Fig.  4B). They also induced a relatively large proportion of point mutations at the site of the adduct (Fig. 4C). Specifically, when allowed to propagate within repair-deficient AB2480, those M13 DNAs bearing bay region 1R-BA and bay region 1S-BA adducts produced 17 and 11%, respectively, of the number of plaques as the same quantity of unadducted M13. Under these conditions, the template DNAs carrying bay region 1R-BA lesions produced 2.6% A 3 G transition mutations, whereas the bay region 1S-BA adducted M13 molecules produced 0.3% transitions at the lesion site. In total, replication of these DNAs within AB2480 cells produced 665 (1R-BA adducted) and 345 (1S-BA adducted) plaques. In wild-type AB1157 cells, in the absence of SOS induction, bay region 1R-BA lesions decreased replication efficiency to 7% of that of unadducted M13 and produced 0.4% A 3 T transversions and 7.3% A 3 G transitions. In the same cell line, bay region 1S-BA adducted DNAs were replicated 16% as efficiently as unadducted and induced 4.2% A 3 G transitions. Under these conditions, bay region 1R-BA adducted M13 generated 275 plaques, and 570 plaques were generated from the bay region 1S-BA adducted DNAs. When SOS conditions were induced in this bacterial strain, bay region 1R-BA containing DNAs were replicated 21% as efficiently as unadducted and induced A 3 T transversions (1.4%) and A 3 G transitions (10.7%). Those M13 molecules carrying bay region 1S-BA decreased replication efficiency to 19% of that of unadducted M13 when SOS functions were induced and produced 0.14% A 3 T transversions and 2.1% A 3 G transitions (Fig. 4, B and C). In total, ϳ800 plaques were produced from bay region 1R-BA adducted M13, and ϳ700 plaques were produced from the replication of bay region 1S-BA adducted DNAs when SOS functions were induced.
All data described above were the result of multiple experiments in which each species of adducted DNA was introduced into each type of E. coli, under a specified set of conditions, in a minimum of three separate experiments. In the case of the bay region BA adducted M13, as many as 12 different transformations were carried out in an attempt to generate large numbers of plaques. The decrease in the plaque-forming efficiencies of these bay region adducted M13 molecules made it impossible to screen an equal number of plaques, but the concomitant increase in mutagenesis opposite these lesions was so dramatic that it was quite obvious that these lesions were indeed more mutagenic than their non-bay region complements. Further, during the course of the investigations described, multiple batches of each species of adducted oligonucleotide had to be prepared, and in no case did the results across these experiments vary greatly.
Replication of Adducted M13 in Vitro-Purified circular M13 molecules, either adducted or unmodified, were used as templates for polymerization by isolated E. coli Pol III holoenzyme (Pol III) in vitro. Replication was initiated 1.8 kilobases upstream of the adducted adenine, and polymerization was carried out in the presence of [␣ 32 P]dATP. Reaction products were separated by electrophoresis through alkaline agarose gels and analyzed via autoradiography. As indicated in Fig. 5, when M13 molecules which were linearized in the vicinity of the adduct site (Cut-L2 Control) were utilized as templates, 92-95% of the products accumulated in a single band corresponding to 1.8 kilobases in length. When a non-adducted 11-mer was substituted for the adducted 11-mer during ligation, those templates supported a 10-fold increase in the amount of primer that was extended to lengths greater than 1.8 kilobases. A circular M13 molecule, which had not been subjected to manipulation (linearization, ligation, and isolation) prior to reaction (Circular-L2 Control) coded for the production of more than 73% of primers that were extended beyond the 1.8-kilobase length (Fig. 5).
Initially, we had hypothesized that those lesions which inhibited replication and were mutagenic in vivo (the bay region lesions) would block Pol III in vitro, whereas the non-mutagenic, non-blocking adducts (non-bay region) would be readily bypassed. This, however, proved not to be the case as replication of each of the adducted templates resulted in essentially complete blockage of Pol III at or near the site of the lesion (Fig. 5).
Nucleotide Incorporation Opposite the Site of the Lesion-Due to the lack of correlation between adduct-induced blockage in vitro and in vivo, we chose to examine the propensity of the purified polymerase complex to incorporate each of the common nucleotides opposite each of the adducted adenine residues. The strategy employed for these studies also afforded an increased degree of precision over the system that utilized adducted M13 molecules as templates since the products of synthesis could be resolved to the level of a single nucleotide as opposed to an approximate size in kilobases (as was the case in the experiments utilizing adducted M13 templates). Synthetic oligonucleotide template-primer complexes in which the 3Ј terminus of the primer was one nucleoside upstream of the adducted adenine were used to determine the efficiency with which purified Pol III ␥ Ϫ ⑀ Ϫ Ϫ would insert a given dNTP opposite the site of the lesion (Fig. 6A). The sequence context in which the adducted adenine was located was kept the same as in the other assays described. The most prevalent mutation that was seen when these lesions were processed intracellularly, A 3 G transition, was not recapitulated in vitro. Even after 120 min, there was essentially no incorporation of dCTP opposite any of the adducted adenine residues (Fig. 6B). It is interesting to note, however, that after 120 min, the adduct that induced the most mutations in vivo, bay region 1R-BA, was the lesion opposite which Pol III ␥ Ϫ ⑀ Ϫ Ϫ most efficiently incorporated dT. This differential is not seen at the 10-min time point (see Fig. 6).

DISCUSSION
Since BA is known to be metabolized to both bay region and non-bay region diol epoxides, and there is a clear distinction between the mutagenicity of these two classes of metabolites (4, 31-33), we consider site-specific benz[a]anthracene-DNA adducts to represent ideal models to study the mechanism by which PAHs induce mutation. This study, therefore, was designed to examine the processing of two bay region BA-DNA adducts and two non-bay region BA-DNA adducts by E. coli DNA polymerases both in vivo and in vitro. M13 molecules carrying these lesions were allowed to replicate within E. coli either with or without prior induction of the cellular SOS response. The same lesions were also incorporated into templates that could be replicated in vitro.

Mutagenesis of BA-DNA Lesions in Vivo
The results of these experiments supported our hypothesis that the cellular replication of templates carrying bay region lesions would result in a higher frequency of point mutations at the site of the adduct than would replication of phage that contained non-bay region adducts. To our knowledge, this is the first study which has demonstrated that even in the same sequence context, site-specific bay region lesions are mutagenic, and the corresponding non-bay region lesions are not. These data suggest that indeed, there is something inherently different about the structure of the different adducted DNA molecules that determines their mutagenicity. While it is not our intent to overemphasize the quantitative aspects of these data, clearly some generalizations can be drawn. First, M13 molecules carrying bay region BA lesions produce fewer plaques than either those carrying non-bay region BA adducts or the corresponding non-adducted M13. Second, the bay region BA adducts induce point mutations at the site of the adduct, whereas the non-bay region lesions are essentially non-mutagenic. Third, the bay region 1R-BA lesion is the most mutagenic of all the adducts examined in this study. Fourth, there is a distinct difference in the number of mutations that were induced by the bay region lesions when they were propagated in different types of E. coli. And fifth, the primary type of FIG. 5. Blockage of Pol III in vitro. Circular M13 molecules containing either site-specifically modified adenine residues or the corresponding unadducted sequences were utilized as templates for polymerization by E. coli DNA polymerase III holoenzyme in vitro. The percent of products in which synthesis was blocked in the vicinity of the lesion is given. substitution introduced is a transition mutation (A 3 G), consistent with the mutations induced by other site-specific adenine N 6 adducts in the same sequence context (26,28).
Regarding differential mutagenesis in different cell types, since AB2480 E. coli are deficient in recA and the binding of RecA protein to single-stranded DNA is thought to be the event that initiates the SOS response, AB1157 cells were chosen to examine the mutagenesis of BA lesions when SOS conditions were induced. AB1157 and AB2480 are essentially isogenic strains of E. coli that vary in their repair capacity. AB1157 cells are wild type with respect to repair, whereas AB2480 cells are deficient in both RecA and UvrA activities. As we had hypothesized, the bay region BA lesions induced the highest frequency of point mutations when replicated in AB1157 cells that had been exposed to UV irradiation prior to transformation. Somewhat surprising, however, was the dramatic difference in mutation frequency that was solely dependent on the cell type in which the adducted DNAs were processed. We propose that the presence of RecA, along with basal levels of UmuC and UmuD, in the AB1157 cells can explain this differential processing of BA lesions in different cell types.
Normally, SOS functions are activated in E. coli in response to massive amounts of DNA damage. Cellular repair activities produce regions of single-stranded DNA to which RecA protein binds, and this is thought to initiate the SOS response (35)(36)(37). Under normal conditions, the extensive nature of DNA damage requires that the proteins involved in translesion synthesis (RecA, UmuC, and the activated form of UmuD) be expressed in high quantity. In the experiments described above, however, there is a single DNA lesion per M13 molecule being introduced into the cell. Further, this lesion is contained on a singlestranded DNA template. It seems reasonable, therefore, that the normal complement of RecA protein (estimated to be 7200 copies per cell (38)) would be sufficient to bind to the singlestranded DNA template and activate some of the constitutive UmuD protein of the cell to its active form (UmuDЈ), which in conjunction with UmuC would promote error prone bypass of the BA lesion even in the absence of a full SOS response. Estimates of the cellular complement of UmuD and UmuC are 180 and 15 copies, respectively, per uninduced cell (39). This hypothesis is supported by observations that the SOS response is activated simply by the presence of single-stranded DNA when E. coli are infected with mutant M13 that replicate much more slowly than wild-type M13 (40).

In Vitro Replication
Blockage-We interpreted the inhibition of cellular replication of M13 containing bay region BA adducts as being potentially due to physical blockage of cellular polymerase molecules at or near the site of the adduct. In an attempt to reproduce this blockage in vitro, we used site-specifically adducted, covalently closed circular M13 molecules as templates for replication by FIG. 6. Single nucleotide incorporation opposite adducted adenine. Site-specifically modified 33-mer oligonucleotides were hybridized to 17-mer primers whose 3Ј termini were positioned 1 base upstream of the adducted adenine (the sequences of these DNAs are depicted in the illustration) (A). The exonuclease-deficient form of E. coli DNA polymerase III (Pol III ␥ Ϫ ⑀ Ϫ Ϫ ) was examined for its propensity to incorporate each of the four common nucleosides opposite the site of the lesion. Percent incorporation is given (B). purified E. coli DNA polymerase III holoenzyme. Synthesis was initiated 1.8 kilobases upstream of the adduct site. Thus, if Pol III were blocked at the site of the lesion, one would expect a 1.8-kilobase product, whereas bypass synthesis would result in a product DNA of approximately 7.2 kilobases (the size of M13).
In these assays, all four adducts served as efficient blocks to replication (see Fig. 5). The observation that the non-bay region lesions were preferentially bypassed in vivo, suggests that the adducts are processed differently in the intracellular environment. Of multiple possible explanations, our favored interpretation is that the polymerase machinery that encounters the adducts in vitro is somehow fundamentally different from the polymerase that allows translesion synthesis in vivo. It may be that a different subassembly of polymerase III is responsible for the replication of adducted M13 molecules in vivo, and this subassembly is more likely to bypass the lesion than the holoenzyme. This idea is supported by Kornberg (1) who argues that the ϳ10 Pol III holoenzyme molecules present in a rapidly dividing cell (our cells are in log phase when they are transformed) are likely to be involved in replication of cellular DNA and are not available for replication of exogenous M13 molecules. Maki et al. (41) offer further support for this hypothesis in a study demonstrating that Pol I may substitute for the activity of a defective Pol III core. Bonner et al. (42) also show that Pol II can become processive when in the presence of Pol III accessory factors. Perhaps the most convincing argument for a different polymerase, however, can be made in light of a study that was published from the laboratory of Dr. Myron Goodman (43) subsequent to the submission of this manuscript. In this study, they demonstrate that Pol II is indeed involved in the replication of episomal DNAs in E. coli. It seems probable, therefore, that something other than the replicative form of Pol III is responsible for the replication of the exogenous DNAs and that this polymerase complex more efficiently bypasses non-bay region BA adducts. Indeed, we have observed a marked difference in the ability of various polymerases to conduct translesion synthesis on oligonucleotides carrying these same non-bay region BA lesions (data not shown).
Mutagenesis-When replicated within E. coli, there is a clear distinction in the mutagenic frequency of the various BA adducts examined in this study. By far, the most mutagenic lesion examined is the bay region 1R-BA adduct, and the most abundant mutation is a transition (A 3 G). This would correspond to the incorporation of dC opposite the adducted adenine residue. In vitro, there was essentially no difference in the amount of dC that was incorporated opposite the bay region 1R-BA adducted adenine and any of the other lesions. Again, these data suggest that there is a fundamental difference in either the polymerase machinery that replicates the adducted DNAs in vitro and in vivo or in the structure that the adduct assumes in the different assay systems. Work from the Grollman laboratory (34) has demonstrated a marked difference in the propensity of various polymerases to insert a given nucleotide opposite an 8-oxoguanine residue. This further supports the idea that there is a different enzyme system acting on these DNAs in vivo and in vitro.
Whatever the reason, we conclude that the in vitro assays described herein are not predictive of the processing of BA adducted DNAs intracellularly. These findings emphasize the importance of exercising extreme caution when attempting to make generalizations about the cellular processing of DNA lesions based on data obtained in vitro.