A single site-specific trans-opened 7,8,9,10-tetrahydrobenzo[a]pyrene 7,8-diol 9,10-epoxide N2-deoxyguanosine adduct induces mutations at multiple sites in DNA.

Site-specific mutagenicity of trans-opened adducts at the exocyclic N(2)-amino group of guanine by the (+)-(7R,8S,9S,10R)- and (-)-(7S,8R,9R,10S)-enantiomers of a benzo[a]pyrene 7,8-diol 9,10-epoxide (7-hydroxyl and epoxide oxygen are trans, BPDE-2) has been determined in Chinese hamster V79 cells and their repair-deficient counterpart, V-H1 cells. Four vectors containing single 10S-BPDE-dG or 10R-BPDE-dG adducts positioned at G(0) or G(-1) in the analyzed 5'-ACTG(0)G(-1)GA sequence of the non-transcribed strand were separately transfected into the cells. Mutations at each of the seven nucleotides were analyzed by a novel primer extension assay using a mixture of one dNTP complementary to the mutated nucleotide and three other ddNTPs and were optimized to quantify levels of a mutation as low as 1%. Only G --> T mutations were detected at the adducted sites; the 10S adduct derived from the highly carcinogenic (+)-diol epoxide was 40-50 and 75-140% more mutagenic than the 10R adduct in V79 and V-H1 cells, respectively. Importantly, the 10S adducts, but not the 10R adducts, induced separate non-targeted mutations at sites 5' to the G(-1) and G(0) lesions (G(0) --> T and C --> T, respectively) in both cell lines. Neither the T 5' to G(0) nor sites 3' to the lesions showed mutations. Non-targeted mutations may enhance overall mutagenicity of the 10S-BPDE-dG lesion and contribute to the much higher carcinogenicity and mutagenicity of (+)-BPDE-2 compared with its (-)-enantiomer. Our study reports a definitive demonstration of mutations distal to a site-specific polycyclic aromatic hydrocarbon adduct.

Mutagenicity of BPDE adducts depends upon efficiency of cellular repair systems to remove the adducts from DNA (mainly nucleotide excision repair (17,18)) and fidelity of DNA polymerases replicating residual adducted sites. Although the normal replicative DNA polymerases pol and pol⑀ are blocked by bulky adducts, members of newly discovered Y superfamily of bypass DNA polymerases such as pol, pol, and pol (19) were found to replicate past sites of various DNA lesions, albeit with low fidelity and low processivity. The frequency of nucleotide misincorporation of pol (20), pol (21), and pol (22) replicating an undamaged template in vitro is in the range of 10 Ϫ3 -10 Ϫ2 . Assuming that the enzymes replicate not only the site of a lesion but possibly also a short stretch of DNA around the lesion with such a low fidelity suggests that besides incorporation of mismatched nucleotides opposite the adducted site, the bypass DNA polymerases could also introduce secondary mutations in the region flanking the adducted site, especially if its DNA structure is perturbed by the presence of the adduct. Multiple mutations in a shuttle vector treated with (Ϯ)-BPDE, believed to be generated by an error-prone polymerase, have been observed (23) in random mutagenesis experiments.
In the present study, we have examined the mutagenicity of 10S-BPDE-dG and 10R-BPDE-dG lesions both at adducted sites and at unmodified flanking sites. We constructed doublestranded plasmid vectors bearing cDNA of the HPRT gene of Chinese hamster V79 cells containing a single adduct positioned at each of the two adjacent sites in the non-transcribed strand of the gene and transfected them separately into V79 cells and also into their nucleotide excision repair-deficient derivative, V-H1 cells (24). These cells are defective in the xeroderma pigmentosum complementation group D/ERCC2 gene encoding for an ATP-dependent DNA helicase (25), an essential subunit of transcription and nucleotide excision repair complex TFIIH (26). They are 9-fold more sensitive to cytotoxic effects of (ϩ)-BPDE-2 than V79 cells and have ϳ50% lower capacity for removal of (ϩ)-BPDE-2-induced adducts from DNA compared with V79 cells (27). We qualitatively and quantitatively evaluated mutagenic effects of both 10R and 10S adducts in a sequence of seven nucleotides by a novel quantitative minisequencing method, and we compared their differences with respect to 10R/10S stereochemistry, position of the adducts at two adjacent sites, and also cellular DNA repair status. The most striking observation was that, in addition to mutations at the adduct sites, significant numbers of mutations were induced 1 or 2 bases 5Ј to these sites by 10S but not 10R adducts.

MATERIALS AND METHODS
Oligonucleotides-All non-adducted oligonucleotides were prepared at IDT (Coralville, IA). The 19-mer oligonucleotide and primers for quantitative minisequencing (Fig. 2) were gel-purified. Adducted oligonucleotide 18-mers (5Ј-AAACTG 0 G Ϫ1 GAAAGCCAAAT) containing a single trans-opened BPDE adduct at one or the other of the numbered dG residues were prepared by solid-phase synthesis using the mixed 10R/10S diastereomers of the appropriately protected adducted phosphoramidite as described (28,29), followed by high pressure liquid chromatography separation of the resultant pair of diastereomeric oligonucleotides. Synthesis and characterization of the diastereomeric pair of oligonucleotides bearing trans-opened BPDE adducts at G 0 with 10R and 10S configuration at the point of attachment of the base to the hydrocarbon have been reported (28). A second R/S pair of 18-mers with trans-opened BPDE adducts at G Ϫ1 was prepared (3-mol scale) by the same methodology and purified by high pressure liquid chromatography as described for the adduct at G 0 (28) (Hamilton PRP-1 column (7 ϫ 305 mm), eluted with a gradient from 0 to 35% solvent B in solvent A over 20 min, where solvent A is 0.1 M (NH 4 ) 2 CO 3 , pH 7.5, and solvent B is a 1:1 mixture of solvent A with CH 3 CN adjusted to the same pH). The late eluting oligonucleotide (t R 19.0 min) was assigned as contain-ing the trans-opened 10S-BPDE dG adduct and the early eluting oligonucleotide (t R 17.6 min) as containing the 10R adduct, on the basis of the CD spectrum (14,30) of the known nucleoside adducts obtained upon enzymatic hydrolysis (12) of the late eluting adduct. The CD spectra of the 18-mer oligonucleotides themselves exhibited bands at 320 -350 nm that were positive for the early eluting (10R) isomer and weakly negative for the late eluting (10S) isomer, consistent with previous observations (31) of other oligonucleotides containing transopened BPDE dG adducts of known absolute configuration as follows: 18-mer modified with the 10S or 10R adduct at the position G 0 , 5Ј-AAACTGGGAAAGCCAAAT; 18-mer modified with the 10S or 10R adduct at the position G Ϫ1 , 5Ј-AAACTGGGAAAGCCAAAT; 19-mer, 5Ј-CATATTTGTGTCATTAGTG; 59-mer scaffold, 5Ј-GAATTCTCATCTT-AGGCTTTGTATTTGGCTTTCCCAGTTTCAGTAATGACACAAATA-TG; primer A, 5Ј-TGCGGGATCCCTCCTCACACCGCT; primer B, 5Ј-C-TGCTTTCCCTGGTCAAGCGG; primer C, 5Ј-GAAATTAATACGACTC-ACTATAGGG; primer D, 5Ј-GCAGATTCAACTTGAATTCTCATC. Sequence of primers used for mutation analysis by quantitative minisequencing is shown in Fig. 2.
Scaffold-directed Extension of the Adducted 18-Mers to the Adducted 37-Mers-The adducted oligonucleotides and the non-adducted control were extended on their 5Ј termini with the 19-mer to the final 37-mers by scaffold-directed ligation. Adducted 18-mers or the corresponding non-adducted 18-mer (10 pmol), 5Ј-labeled with 32 P, were incubated with the 19-mer and 5Ј-32 P-labeled 59-mer scaffold in a molar ratio 1:2:1.5 at 65°C for 5 min and cooled to room temperature over 10 min. Ligase reactions (30 l) containing 66 mM Tris-HCl, 5 mM MgCl 2 , 1 mM dithioerythritol, 1 mM ATP, and 0.5 units of T4 DNA ligase were incubated at 25°C for 2 h at pH 7.5 and terminated by adding 15 l of stop solution containing 90% formamide, 10 mM NaOH, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF. Reaction products were separated on 10% PAGE and detected by autoradiography. The 37-mer bands were excised from the gel and briefly washed with water. The gel was repeatedly treated with water by three rounds of heating and cooling (65°C for 5 min and 0°C for 5 min) to elute the DNA, and the DNA was precipitated from the eluate with ethanol. The total yield of the ligation product was ϳ40% for the non-adducted oligonucleotide and ϳ25% for the adducted oligonucleotides.
Preparation of Plasmid Vectors Containing the Full-length Hamster HPRT cDNA-The HPRT cDNA was prepared by RT-PCR from total mRNA of V79 cells using primer A and primer D as described before (32). Primer A and primer D create BamHI and EcoRI sites, respectively. Double-stranded pCR3 vector (5.1 kb, Invitrogen) containing the HPRT cDNA was prepared by TA cloning (Invitrogen) of the PCR product. The orientation of the HPRT cDNA insert was screened by HindIII digestion of the plasmid DNA. The construct with the antisense orientation of the insert with respect to the cytomegalovirus promoter (pCR3/HPRT antisense ), used for preparation of adducted vectors, yields a 289-bp fragment, whereas the construct with the sense orientation (pCR3/HPRT sense ) yields a 596-bp fragment. Sequence of the insert in pCR3/HPRT antisense was further verified by dideoxy sequencing using ThermoSequenase Cycle Sequencing Kit (U. S. Biochemical Corp.).
Preparation of Plasmid Vectors Containing Site-specific 10S-or 10R-BPDE Adducts in the HPRT cDNA-Four plasmid vectors adducted with BPDE and a non-adducted control were prepared by the enzymatic extension of the BPDE-adducted and control 37-mer oligonucleotides (see above for preparation) and annealed to the single-stranded circular pCR3/HPRT antisense DNA as shown in Fig. 3. The single-stranded DNA was isolated from the supernatant of an Escherichia coli culture transformed with pCR3/HPRT antisense plasmid after co-infection with M13K07 helper phage (33). Double-stranded DNA was prepared as follows: a mixture (72 l) containing 0.5 pmol of single-stranded pCR3/ HPRT antisense , 0.5 pmol of 37-mer oligonucleotide, 20 mM Tris-HCl, pH  7.4, 5 mM MgCl 2 , and 50 mM NaCl was incubated at 95°C for 3 min and then at 65°C for 3 min and cooled slowly (15 min) to room temperature. After adding 1 mM DTT, 1 mM ATP, 500 M dNTPs, T4 gene 32 protein (5 g), T4 DNA polymerase (5 units), and T4 DNA ligase (2.5 units), the reaction (100 l) was incubated at 25°C for 2 h. Following extraction with phenol/chloroform/isoamyl alcohol (25:24:1), DNA was precipitated with 2 volumes of cold ethanol after adding 0.3 M sodium acetate, pH 5.2, and 2 g/ml tRNA. After centrifugation, the DNA pellet was dissolved in 50 mM Tris-HCl, pH 7.6, 5 mM MgCl 2 , 5 mM DTT, and 50 g/ml bovine serum albumin (100 l) and digested with 20 units of exonuclease III for 1 h at 37°C to remove partially extended reaction products. Extraction (phenol/chloroform/isoamyl alcohol) and ethanol precipitation was repeated. The pellet was dissolved in 10 mM Tris-HCl, 150 mM NaCl, 10 mM MgCl 2 , 1 mM DTT, 100 g/ml bovine serum albumin, pH 7.9 (25 l). The DNA was cut with BamHI ϩ EcoRI (5 units each) for 1 h at 37°C. Reaction products, including two BamHI-EcoRI fragments (15-and 31-mer) and a 24-mer EcoRI-EcoRI fragment, were separated on a 0.9% preparative agarose gel. Bands corresponding to 713 bp (HPRT cDNA) and ϳ5 kb were eluted from the gel (QIAquick Gel Extraction kit, Qiagen, Valencia, CA) and were re-ligated at 16°C for 16 h in a 50-l mixture containing 66 mM Tris-HCl, 6.6 mM MgCl 2 , 10 mM DTT, 66 M ATP, and 1 unit of T4 DNA ligase. This step changes the orientation of the insert from the antisense to sense and also separates residual single-stranded DNA. The reaction was stopped by 500 mM EDTA (1 l), and products were applied on the Qiagen QIAquick column (Nucleotide Removal kit, Qiagen). DNA was eluted in a sterile 10 mM Tris-HCl buffer, pH 8.5. The final concentration of pCR3/ HPRT antisense -BPDE and the non-adducted control used for cell transfections was 20 ng/l. All four plasmid vectors were prepared at least three times.
Transfection of V79 and V-H1 Cells with Adducted Plasmid Vectors-The Chinese hamster V79 cells were obtained from the ATCC and propagated in minimum Eagle's medium containing 10% dialyzed and heat-inactivated fetal bovine serum in an incubator with controlled humidified atmosphere containing 5% CO 2 . Before transfection, the cells were seeded at a density 2 ϫ 10 5 per 60-mm dish. Following 24 h of incubation, cells were transfected with 0.3 g of one of the modified plasmids (10S or 10R adduct at the G 0 or G Ϫ1 position) or the control plasmid using Effectene (Qiagen). The cells were incubated for 24 h, washed with phosphate-buffered saline, and supplied with fresh medium. After 48 h following the transfection, the cells were harvested by trypsinization and seeded in the same medium supplemented with 500 g/ml G418 (Invitrogen) at a density 3 ϫ 10 4 per 60-mm dish to select permanently transfected cells. Ten dishes were prepared from each transfection. Cellular colonies developed after an 11-day incubation with G418 were harvested by trypsinization from each dish and collected by centrifugation. To ensure analysis of at least 100 independent cellular clones from each plasmid transfection, 10 plates containing at least 60 cellular colonies in a random mixture were harvested. Thus, each transfection experiment yielded 5 ϫ 10 pooled samples, which were separately processed and analyzed.
The repair-efficient Chinese hamster V-H1 cells (kindly provided by Dr. M. Zdzienicka, Leiden University Medical Center, Leiden, The Netherlands) were grown and transfected in the same way with several changes. The cells were seeded at a density 2.6 ϫ 10 5 per a 60-mm dish, transfected with 0.5 g of the plasmid, and seeded after the transfection at a density 1 ϫ 10 5 per 60-mm dish.
Preparation of DNA Samples for Mutation Analysis-Chromosomal DNA from harvested cells was isolated using High Pure PCR Template Preparation kit (Roche Molecular Biochemicals). DNA fragment (337 bp) encompassing the examined region was prepared by PCR amplification using primer B corresponding to region 437-457 of the HPRT gene (exon 6) (32) and primer C corresponding to the T7 promoter region of the plasmid construct; the endogenous cellular HPRT gene is not amplified using this set of primers. The reaction was carried out in a 50-l mixture containing 200 ng of DNA, 200 nM primers, 200 M dNTP, and Taq DNA polymerase (0.025 unit/l, Qiagen). Following 30 cycles (95°C for 30 s, 55°C for 10 s, and 72°C for 10 s) with the initial 3 min at 95°C and the final 7 min at 72°C, the residual dNTPs were removed by a 30-min alkaline phosphatase treatment (1 unit per each 20 l of the reaction). The product was purified using QIAquick PCR Purification kit (Qiagen). Concentration of the product was estimated based on absorbance at 260 nm, and the purity was checked on 1.5% agarose gel containing 0.5 g/ml ethidium bromide. Standard 337-bp fragments used as positive and negative controls for quantitative minisequencing were amplified from plasmids pCR3/HPRT antisense containing all four nucleotides at each of the seven examined positions (Fig.  2). These plasmids were prepared by site-directed mutagenesis using a QuikChange Kit (Stratagene, La Jolla, CA).
Mutation Detection Assay Using Quantitative Minisequencing-The method is based on a single nucleotide primer extension assay (34,35) modified to quantify mutations in a known sequence context at levels lower than 10%. A primer annealed to the DNA template (337-bp fragments), one nucleotide before the analyzed site (Fig. 2), is extended with a mixture of one dNTP and three other ddNTPs. When annealed to a fragment carrying a specific analyzed nucleotide (mutation), the primer can be extended with the complementary dNTP before being terminated by ddNTP incorporation at the subsequent nucleotide, whereas elongation of the primer annealed to fragments not containing the analyzed nucleotide is terminated immediately with ddNTP. The level of mutation is determined by calculating the ratio of differently extended primers following their separation on denaturing polyacrylamide gel and analysis with PhosphorImager (see for 30 s, 50°C for 10 s, and 72°C for 10 s) with the initial 1 min at 95°C were used. The reaction was stopped with 1 volume of 95% formamide, 20 mM EDTA, 10 mM NaOH, 0.05% bromphenol blue, and 0.05% xylene cyanol FF. Reaction products were separated on 15% PAGE in 1ϫ TBE buffer and visualized by autoradiography. To increase the throughput of the assay, up to four sets of samples were subsequently loaded into the same gel. For quantitative evaluation (Fig. 5), the signal from the gel was transferred onto a screen and scanned by Cyclone Phos-phorScreening System (Packard Instrument Co.), and the amount of radioactivity associated with each spot was quantified using software provided with the system.

Construction of Site-and Stereospecifically Adducted
Plasmid Vectors-BPDE-adducted plasmid vectors were constructed by the in vitro enzymatic extension of 37-mer oligonucleotides after their annealing to a single-stranded plasmid template as shown in Fig. 3 and described under "Materials and Methods." The plasmid contains the HPRT gene inserted in the antisense orientation toward the plasmid promoter. The 37-mers were prepared from 18-mer oligonucleotides modified with 10S or 10R adducts at the G 0 or G Ϫ1 sites (see "Materials and Methods" and Fig. 4A). The sequence of the 18-mers corresponds to region 629 -646 of the non-transcribed strand of the HPRT gene in Chinese hamster cells. The use of 37-mers instead of 18-mers in the reaction substantially increases the yield of the covalently closed circular reaction products (data not shown). After the reaction, covalently closed circular DNA was digested with restriction enzymes EcoRI and BamHI and re-ligated yielding a vector with the HPRT gene in the sense orientation and a single BPDE adduct in the non-transcribed strand of the gene (Fig. 3). Purity of the final products was checked by digestion with restriction enzymes EcoRI and MslI and separation of the 32 P-labeled restriction fragments on denaturing PAGE (Fig. 4B). The presence of BPDE adducts retards mobility of the adducted fragments through the gel, which translates into the shift of the corresponding bands (Fig.  4B). Scanning of the gel using the Cyclon system detected at least 99.3% of a band corresponding to the adducted fragment in adducted samples. This indicates that the adducts were very stable during the preparation of the plasmid vectors and that the adducted vectors contain less than 0.7% of contaminants.
Transfection of Cells with Adducted Plasmid Vectors-Four adducted plasmids and a non-adducted control were separately transfected into repair-proficient V79 cells and repair-deficient V-H1 cells. Each experiment was repeated at least twice with a different plasmid preparation. Doubling times of exponentially growing V79 and V-H1 cells were ϳ14 and ϳ18 h, respectively. After transfection, doubling times of V79 and V-H1 cells increased to ϳ19 and ϳ34 h, respectively. There was no difference in proliferation of the cells transfected with different plasmid constructs. For the selection with antibiotics, cells were seeded in a density ensuring approximately the same yield of cellular colonies per plate from both cell lines. Because the cells kept proliferating at a time between the transfection and seeding (48 h), and their proliferative rate was different, it is reasonable to expect that 5.8 of the V79 cellular colonies and 2.7 of the V-H1 cellular colonies on average originated from a single cell. Thus, samples harvested for mutation analysis contained a random mixture of cellular colonies (pools) that were not necessarily independent. A sufficient total number of colonies (ϳ600) in random pools was harvested from each plasmid transfection to ensure that at least 100 of these were independent clones. Assuming that permanent transfection decreases the proliferative rate of the affected cells more than the nonaffected cells, which are not selected, this number of analyzed independent colonies is the lowest possible estimate.
Quantitative Minisequencing and Calculation of the Mutation Level at a Specific Site-Chromosomal DNA from pooled cellular samples was used for amplification of the 337-bp DNA template encompassing a region originally containing the adducted nucleotide. The amplified template is a mixture of fragments with a DNA sequence reflecting the mutagenicity of the adduct at the region. Quantitative minisequencing was used to assess both the type and the level of mutations at the region as described under "Materials and Methods" and Fig. 5. To validate accuracy and sensitivity of the assay model, binary mixtures of 337-bp fragments differing in a single nucleotide at each site of the region (pseudo-pools) were analyzed. These fragments were prepared by PCR amplification of 337-bp region from the HPRT cDNA manipulated in plasmid vectors by site-directed mutagenesis. The relative content of one fragment in the other was 50, 25, 12.5, 6.25, 3.13, 1.56, 0.78, 0.39, 0.19, and 0%. Composition of these binary mixtures represented all potential single nucleotide mutations in the analyzed region (5Ј-A ϩ3 C ϩ2 T ϩ1 G 0 G Ϫ1 G Ϫ2 A Ϫ3 ). The ratio between a level of the fragment detected by the assay (f measured ) and a predicted level of the fragment in the mixture (f predicted ) represents a correction factor (k). The means Ϯ S.D. of the correction factor were calculated at each analyzed point from at least three independent experiments and plotted. The relationship between k and the level of the analyzed fragment was found linear between 0.78 -50%. Fig. 8 is included as Supplemental Material and shows the results from analyses of model binary fragment mixtures differing by a nucleotide in positions found mutated in the study (C ϩ2 , G 0 , and G Ϫ1 ). The average k values FIG. 5. Quantitative minisequencing of binary mixtures of DNA fragments with a single nucleotide difference at the position G 0 (see Fig. 2). A DNA fragment containing A, C, or T instead of G at the G 0 position (A, C, or T fragment) was mixed with the fragment containing G (G fragment) at different ratios as indicated. A 32 P-labeled primer annealed one nucleotide before the analyzed site was extended by a cycling reaction with ThermoSequenase DNA polymerase in a mixture containing one dNTP complementary to the nucleotide at the analyzed site and the three other ddNTPs. Reaction products were separated on denaturing 15% PAGE, and the signal was transferred from the gel onto a screen and scanned by Cyclone PhosphorScreening System (Packard Instrument Co.). The primer extended more than once (with dNTP ϩ ddNTP) generates the upper spot, whereas the primer extended only once (with ddNTP) generates the lower spot. Radioactivity of the upper spot corresponds to the amount of the analyzed fragment, whereas radioactivity of the lower spot corresponds to the amount of the second fragment in the mixture. The amount of radioactivity associated with each spot was quantified using software provided with the system. Plotted numbers were calculated according to the formula: (S 2 Ϫ(C n2 S 1 )/(C n1 ))/((S 1 ϩ S 2 ) Ϫ(C n2 S 1 )/(C n1 )) ϫ 100 ϭ mutation (%). S 1 and S 2 indicate the amount of radioactivity associated with the lower and upper spots of the analyzed samples, respectively; C n1 and C n2 indicate the same for the negative control containing 0% of the analyzed fragment (only pure G fragment).
FIG. 6. Determination of G 3 T mutations at the G 0 position by quantitative minisequencing. PCR fragments (frgmt.) were amplified from DNA of V79 cells transfected with plasmid vectors adducted with 10S-or 10R-BPDE or nonadducted at G 0 position (A) or G Ϫ1 position (B). 32 P-Labeled primer annealed one nucleotide before the analyzed site (G 0 primer, see Fig. 2) was extended by a cycling reaction with ThermoSequenase DNA polymerase in a mixture containing dATP, ddGTP, ddTTP, and ddCTP. Reaction products were separated on denaturing 1.5% PAGE; the signal was transferred from the gel onto a screen and scanned by Cyclone PhosphorScreening System (Packard Instrument Co.).
(0.85-1.05) calculated for each mutation in the region from model binary mixtures (the rest of the results not shown) were used to re-calculate levels of mutations found in unknown samples. The results demonstrate accuracy and linearity of the assay in tested binary mixtures containing more than 1% of the analyzed fragment. Accuracy of the assay decreases sharply when levels of the analyzed fragment are lower than 1% and limits sensitivity of the assay to 0.3% (see Supplemental Material Fig. 8). Thus, mutagenic changes at levels higher than 0.3% can be qualitatively detected and at levels higher than 1% can be accurately quantified by the assay.
Mutagenicity of BPDE Adducts in Repair-proficient and Repair-deficient Cells-Mutagenicity of 10S-BPDE-dG and 10R-BPDE-dG lesions in the sequence 5Ј-A ϩ3 C ϩ2 T ϩ1 G 0 -G Ϫ1 G Ϫ2 A Ϫ3 was examined at the adducted sites (G 0 and G Ϫ1 ) and also at their 5Ј-and 3Ј-flanking sites (from ϩ3 to Ϫ3, Fig.  2 and also Fig. 6). G 0 corresponds to the hotspot (G-634) of the non-transcribed strand of the gene found in random mutagenesis studies of (ϩ)-BPDE-2 in V79 cells (10,32).
In repair-proficient V79 cells, the guanine adduct derived from (ϩ)-BPDE-2, 10S-BPDE-dG, induced 2.5 Ϯ 1.3 and 2.8 Ϯ 1.1% of G 3 T mutations of the adducted nucleotides at the G 0 and G Ϫ1 sites, respectively. The guanine adduct derived from (Ϫ)-BPDE-2, 10R-BPDE-dG, induced 1.8 Ϯ 1.0 and 1.9 Ϯ 0.7% of G 3 T mutations of the adducted nucleotides at the G 0 and G Ϫ1 sites, respectively (Fig. 7A). Analysis of secondary mutations of nucleotides at sites flanking the lesion on the 5Ј-end revealed that the 10S adduct, but not the 10R adduct, induced C 3 T mutations (0.7 Ϯ 0.4%) at the (ϩ2) position when the G 0 site was adducted, and G 3 T mutation (1.3 Ϯ 0.5%) at the G 0 position when the G Ϫ1 site was adducted. Interestingly, adducts at the G 0 or G Ϫ1 site did not induce mutations of the T nucleotide located at the (ϩ1) position. Also, no mutations were found in the 3Ј-vicinity of the adducted sites (up to 3 nucleotides) or further upstream in the 5Ј-direction (3 and 4 positions analyzed overall from the adducted G 0 and G Ϫ1 sites, respectively).
The effect of repair deficiency in the V-H1 cells on the observed frequency of mutations is relatively small (a factor of 2 or less) and is only significant for the constructs containing 10S adducts at G 0 (Fig. 7B). For example, in the VH-1 cells, the 10S adduct induced 4.6 Ϯ 1.5 and 2.1 Ϯ 0.6% of G 3 T mutations of the adducted nucleotides at G 0 and G Ϫ1 sites, as compared with 2.5 and 2.8%, respectively, in the V79 cells (see above). The 10R adduct induced 1.9 Ϯ 0.8 and 1.2 Ϯ 0.6% of G 3 T mutations at the adducted nucleotides when at the G 0 and G Ϫ1 sites, respectively. Analysis of secondary mutations of nucleotides at sites flanking the lesion on the 5Ј-end showed the same pattern of mutations found in V79 cells; the 10S adduct, but not the 10R adduct, induced C 3 T mutations (1.7 Ϯ 0.6%) at the (ϩ2) position when the G 0 site was adducted and G 3 T mutations (1.5 Ϯ 0.7%) at the G 0 position when G Ϫ1 was the adducted site. As with the V79 cells, no other mutation type or other mutated nucleotide was found in the analyzed region. The formation of secondary mutations depends on the presence of the 10S adduct but probably not on the formation of primary mutations. If so, tandem mutations (two mutations on the same analyzed DNA fragment) would be detected. In the primer extension assay tandem T mutations would give rise to multiple A incorporations in the primers used. No such multiple extensions of the primer were observed. Although we cannot exclude the possibility of formation of tandem mutations in levels below the detection limit of the assay (0.3%), random formation of secondary mutations independent of the primary mutations is a more probable scenario. Large standard deviations of the means of the presented data originate in large differences in the individual data from each of the 10 samples containing pools of cellular colonies, not from irreproducibility of the assay. Repeated analyses of the same samples showed remarkable reproducibility (maximum scatter did not exceed 10% of the calculated values, data not shown).

FIG. 7. Mutagenic effects of 10S-and 10R-BPDE adducts in the vicinity of the lesion in V79 (A) and V-H1 cells (B).
Note the difference in vertical axis scales for the two cell lines. Types of mutations and their levels were determined by quantitative minisequencing at each indicated site as described under "Materials and Methods" (detection limit 0.3%). Data shown are means Ϯ S.D. of values acquired from at least two experiments. In each experiment, 10 samples were analyzed from transfection of each plasmid construct. Adduct position is indicated by an asterisk.

DISCUSSION
At the target sites (G 0 and G Ϫ1 ), the 10S adduct and the 10R adduct induced only G 3 T mutations. This mutation was found to be dominant in previous studies examining mutagenicity of site-specific adducts both in prokaryotic (31,36,37) and eukaryotic cells (37), although G 3 C and G 3 A mutations have also been observed in some sequences. The sequence context of our adducted G 0 site is identical to the ϳTGGϳ sequence examined in simian kidney cells (37). There was no significant difference between mutagenicity at the adducted site for the 10S-BPDE-dG lesion relative to the 10R-BPDE-dG lesion, and a preponderance of G 3 T mutations was observed, as in the present study. However, the level of mutation induced by both adducts was substantially higher than in our study (13 versus 1.8 -2.5%), and low levels (Ͻ1%) of G 3 C and G 3 A mutations were also detected. These differences may stem mainly from differences in the experimental systems; the single-stranded vector system used by Page et al. (31) and Moriya et al. (37) eliminates the involvement of DNA repair, whereas the double-stranded vector system used in our study is sensitive to DNA repair (both in V79 and in V-H1 cells, which have ϳ50% remaining capacity to remove adducts derived from the (ϩ)-(7R,8S,9S,10R)-enantiomer of BPDE from DNA compared with V79 cells (27)). Consequently, the results of our study show substantially lower mutagenicity of the examined adducts and reflect both DNA repair of the adducts and fidelity of their bypass.
Results from DNA repair-proficient (V79) and DNA repairdeficient (V-H1) cells show only minor differences. It is possible that the defect of V-H1 cells in the xeroderma pigmentosum complementation group D/ERCC2 gene (25) influences mainly transcriptional coupled repair (18) and causes a decreased ability of the cells to preferentially repair BPDE adducts from the transcribed strand of active genes (27), whereas the efficiency of global genomic repair (18), which is responsible for the repair of the non-transcribed strand (location of the adducts in this study), may be affected only marginally. Both lesions in both adducted G sites generated the same kind of mutation (G 3 T), and the 10S adduct induced identical secondary mutations in these two cell lines. Quantitatively, no significant difference in mutagenicity of the 10R adduct was observed in corresponding sites in V79 and V-H1 cells. However, mutagenicity of the 10S adduct was significantly higher at the G 0 site in the V-H1 cells than in the V79 cells (p Ͻ 0.006), although similar at the G Ϫ1 site in both cell lines (p Ͼ 0.1). The data suggest that repair of the 10S adduct is more efficient than the 10R adduct in the non-transcribed strand but also that repair efficiency of the 10S adduct may differ from site to site. These results are in accordance with a study by Custer et al. (29) who demonstrated a remarkable resistance of the 10R adduct to DNA repair in vitro compared with 10S adduct using a whole cell extract. However, in a different sequence context, no difference in repair efficiency of these two adducts in vitro was observed (38).
Mutations remote from a specifically modified base in DNA were first described by Lambert et al. (39) for frameshifts induced by an acetylaminofluorene adduct. The present study reports a definitive demonstration of substitution mutations remote from the target site induced by a single, site-specific polycyclic aromatic hydrocarbon DE adduct. These non-targeted mutations were observed only when the adduct has the 10S configuration and were found within two nucleotides on the 5Ј-side of the adducted base, namely at C 2 when G 0 was modified and at G 0 when G Ϫ1 was modified (for sequence see Fig. 2). Notably, the absence of non-targeted mutations with 10R adducts at either position provides strong internal evidence that these mutations with the 10S adducts do not result from any artifact of the oligonucleotide synthesis, since for each sequence the 10R and 10S adducted oligonucleotides were prepared together from the mixed 10R/10S diastereomers of the phosphoramidite (see "Materials and Methods"), and thus the two diastereomeric oligonucleotides underwent identical treatment prior to final chromatographic purification. The nontargeted mutations at both sites represented mutagenic changes of the original nucleotides to T, suggesting their bypass with a mismatched A. Furthermore, the T that is immediately 5Ј to the G 0 lesion or separated by one base from the G Ϫ1 lesion was not found to be mutated in the presence of either lesion. An attractive explanation for these data is that translesion synthesis results in A being inserted (either correctly or incorrectly) as the preferential nucleotide at the adducted site and at the two additional 5Ј-flanking nucleotides. In vitro studies with isolated bypass DNA polymerases (pol , , , and ) on templates containing a single 10S-BPDE-dG or 10R-BPDE-dG lesion identified pol as involved in low fidelity insertions at these lesions in vitro (40,41). The enzyme incorporates mostly A and G opposite the lesion site (40,41) and thus is likely to be responsible at least in part for mutations induced by BPDE-dG lesions in mammalian cells (mostly G 3 T with minor G 3 C and G 3 A). Although the enzyme extends the primer beyond the lesion site rather inefficiently, its preference to extend mispaired primers containing a purine opposite the adduct and its high misincorporation frequency on non-damaged templates (10 Ϫ3 -10 Ϫ2 ) (41) are features that may contribute to the formation of non-targeted mutations and would be worthy of further exploration by studies in vitro and in cells lacking functional pol.
The 10S BPDE adducts at the G 0 and G Ϫ1 sites induced identical non-targeted mutations in both repair-proficient (V79) and repair-deficient (V-H1) cell lines, and the frequency of the non-targeted mutations correlated with the frequency of targeted mutations. The increased level of the targeted mutation at the G 0 site (G 3 T) in V-H1 cells was accompanied by a similar increase in the level of the non-targeted mutations (C 3 T) in these cells compared with V79 cells, whereas no difference in the frequency of non-targeted mutation (G 3 T at the G 0 site) between V-H1 and V79 cells was observed when the levels of the targeted mutation (G 3 T at the G Ϫ1 site) were similar in both cell lines. Assuming that the differences in the targeted mutation frequency between these two cell lines reflect only the efficiency of the 10S adduct removal from a particular site, it seems likely that the adduct level determines the frequency of not only targeted but also non-targeted mutations.
The present observation that mutations are induced 1 or more bases away from the target site by a polyaromatic hydrocarbon DE lesion is consistent with results of our recent study (42) of mutations in an E. coli-M13 system induced by several cis-opened adducts derived from both BPDE-2 (the diastereomer used in the present study, whose benzylic hydroxyl group and epoxide oxygen are trans) and BPDE-1 (benzylic hydroxyl group and epoxide oxygen cis). More limited data in E. coli by Jelinsky et al. (36) using a single trans-opened 10S adduct of BPDE-2 had also led to the tentative suggestion of non-targeted mutations at a base immediately adjacent to this adduct. In our previous study (42) using the sequence ϳG 6 C 5 G 4 G 3 G 2 G 1 G 0 ϳ with cis-opened adducts at G 0 , non-targeted substitution mutations 4 and 6 bases remote from the target site were described, but their significance was not fully recognized. No significant non-targeted substitutions had been detected in this DNA sequence containing trans-opened BPDE-1 or BPDE-2 dG adducts in the same experimental system (31). The most prevalent non-targeted mutations (fre-quency 2-4%) induced by the cis-opened BPDE-1 and BPDE-2 dG adducts were T substitutions at G 6 and G 4 , and they occurred both in the absence of mutations at the target site (G 0 ) and in combination with G 0 3 T mutations at this site. These non-targeted mutations were not observed with the control (non-adducted) sequence, and most significantly they were observed only when the adducts at G 0 had 10S but not when they had 10R configuration, in analogy to our present observations. Because no significant non-targeted mutations were observed in a different (ϳG 6 C 5 G 4 T 3 T 2 C 1 G 0 ϳ) sequence containing BPDE-1 or -2 adducts at G 0 (42), the above mutations were most likely related to the run of 5 guanines in the ϳG 6 C 5 G 4 G 3 G 2 G 1 G 0 ϳ sequence. Despite the specific sequence effect, as well as the differences in experimental systems, the similarity between these results in E. coli and the present observations in mammalian cells is intriguing and suggests that mutations remote from the lesion site induced by polyaromatic hydrocarbon DE adducts may constitute a not uncommon mechanism of polyaromatic hydrocarbon mutagenesis whose significance has not been appreciated previously.
A limited number of site-specific mutagenesis studies with other types of bulky DNA adducts has demonstrated induction of non-targeted substitution mutations at non-adducted sites in the vicinity of the lesion. They include studies of the N7-guanyl adduct of aflatoxin B 1 8,9-epoxide and the adduct's ring opened formamidopyrimidine form in SOS-induced E. coli (43,44) as well as N-deoxyguanosin-8-yl)-2-acetylaminofluorene and N-(deoxyguanosin-8-yl)-2-aminofluorene in COS-7 cells (45). Although it is difficult to make any generalization about the mechanism of these non-targeted mutations by comparing the data of these studies to those presented here (different experimental models, structure of the adducts, and sequence context), it is clear at this point that the formation of various non-targeted mutations is likely related to the bulky character of the adducts and that the orientation of the adducts may play a significant role. The hydrocarbon of the 10S BPDE-dG adduct orients toward the 5Ј-side of the modified base in the minor groove of duplex DNA in the sequence 5Ј-CGC (15) as well as 5Ј-TGC (46) and could thus cause structural perturbations 5Ј to the lesion in our study (5Ј-TGG and 5Ј-GGG), resulting in the observed non-targeted mutations 5Ј to the adduct. In contrast, the 10R BPDE-dG adduct, which orients in the opposite direction toward the 3Ј-end of the modified strand in duplex DNA (16), was not found to cause any mutations at non-target bases on either side of the adduct. Interestingly, about 13% of the total mutations in SOS-induced E. coli caused by a dG adduct of aflatoxin B 1 8,9-epoxide (43), which intercalates into the helix on the 5Ј-side of the modified G base (47,48), were primarily C 3 T mutations at the dC immediately 5Ј to the lesion (G) in a 5ЈϳCGAϳ sequence. In contrast, an N-(deoxyguanosin-8-yl)-2-acetylaminofluorene lesion, whose hydrocarbon moiety displaces the modified G and intercalates in its place opposite the complementary C (49), gave rise to base misincorporation on the 3Ј-side of the lesion site (50).
The dependence of non-targeted mutations on adduct configuration is of particular interest in light of the marked differences in mutagenicity and carcinogenicity between (ϩ)-and (Ϫ)-BPDE-2 enantiomers. Higher tumorigenic activity of (ϩ)-BPDE-2 compared with (Ϫ)-BPDE-2 has been demonstrated. (ϩ)-BPDE-2 induced lung tumors and skin tumors when injected intraperitoneally into newborn mice and applied topically to the skin of adult mice, respectively; but (Ϫ)-BPDE had little or no carcinogenic activity (7,8). Although the present mutagenesis study showed little difference in the mutagenicity of (ϩ)-and (Ϫ)-BPDE-2 dG adducts at specific sites, (ϩ)-BPDE-2 was ϳ11 times more mutagenic than (Ϫ)-BPDE-2 in the HPRT gene of V79 cells on a per dose basis in a random mutagenesis study (11). Many factors can contribute to these different results, including more efficient adduct formation from (ϩ)-relative to (Ϫ)-BPDE-2 (12), differences in relative proportions of dG and dA adducts formed from the two enantiomers (12,13), as well as sequence effects on adduct formation, repair (51), and bypass (41). Furthermore, the higher mutagenicity of the (ϩ)-BPDE-2 adducts observed in random mutagenesis studies may result in part from their mutagenic effects on the whole region (i.e. non-targeted mutations in the vicinity of the adduct) rather than only at a particular adducted site. Because not all DNA mutations lead to amino acid substitutions, and some substitutions may be silent in terms of function, it is obvious that induction of a mutation at more than one site by a single adduct increases the probability of an amino acid change that could generate a protein with a compromised function. Consequently, the mutagenic and tumorigenic potential of a BPDE adduct leading to "multiposition" mutations would be significantly higher. In summary, results of this study contradict the intuitive notion that point mutations arise exclusively by erroneous replication of the modified base and suggest that the higher mutagenic and carcinogenic activity of (ϩ)-BPDE-2 compared with (Ϫ)-BPDE-2 may partly stem from the capability of its major dG adduct to induce DNA mutations at multiple sites.