Error-prone Translesion Synthesis by Human DNA Polymerase η on DNA-containing Deoxyadenosine Adducts of 7,8-Dihydroxy-9,10-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene*

When human DNA polymerase η (pol η) encounters N6-deoxyadenosine adducts formed by trans epoxide ring opening of the 7,8-dihydroxy-9,10-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene (BaP DE) isomer with (+)-7R,8S,9S,10R configuration ((+)-BaP DE-2), misincorporation of A or G and incorporation of the correct T are equally likely to occur. On the other hand, the enzyme exhibits a 3-fold preference for correct T incorporation opposite adducts formed by trans ring opening of the (–)-(7S,8R,9R,10S)-DE-2 enantiomer. Adducts at dA formed by cis ring opening of these two BaP DE-2 isomers exhibit a 2–3-fold preference for A over T incorporation, with G intermediate between the two. Extension one nucleotide beyond these adducts is generally weaker than incorporation across from them, but among mismatches the (adducted A*)·A mispair is the most favored for extension. Because mutations can only occur if mispairs are extended, this observation is consistent with the occurrence of A·T to T·A transversions as common mutations in animal cells treated with BaP DE-2 isomers. Adducts with S absolute configuration at the point of attachment of the hydrocarbon to the base inhibit incorporation and extension by pol η to a lesser extent than their R counterparts. Template-primers containing each of the four isomeric dA adducts derived from BaP DE-2 and two adducts derived from 9,10-epoxy-7,8,9,10-tetrahydrobenzo-[a]pyrene in which the 7- and 8-hydroxyl groups of the DEs are replaced with hydrogens exhibit reduced electrophoretic mobilities relative to the unadducted oligonucleotides. This effect is largely independent of DNA sequence. Decreased mobility correlates with an increased rate of incorporation by pol η, suggesting a systematic relationship between the overall DNA structure and efficiency of the enzyme.


,9,10-tetrahydrobenzo[a]pyrene (BaP DE) isomer with (؉)-7R,8S,9S,10R configuration ((؉)-BaP DE-2), misincorporation of A or G and incorporation of the correct T are equally likely to occur. On the other hand, the enzyme exhibits a 3-fold preference for correct T incorporation opposite adducts formed by trans ring opening of the (؊)-(7S,8R,9R,10S)-DE-2 enantiomer. Adducts at dA formed by cis ring opening of these two BaP DE-2 isomers exhibit a 2-3-fold preference for A over T incorporation, with G intermediate between the two. Extension one nucleotide beyond these adducts is generally weaker than incorporation across from them, but among mismatches the (adducted A*)⅐A mispair is the most favored for extension. Because mutations can only occur if mispairs are extended, this observation is consistent with the occurrence of A⅐T to T⅐A transversions as common mutations in animal cells treated with BaP DE-2 isomers.
Adducts with S absolute configuration at the point of attachment of the hydrocarbon to the base inhibit incorporation and extension by pol to a lesser extent than their R counterparts. Template-primers containing each of the four isomeric dA adducts derived from BaP DE-2 and two adducts derived from 9,10-epoxy-7,8,9,10-tetrahydrobenzo-[a]pyrene in which the 7-and 8-hydroxyl groups of the DEs are replaced with hydrogens exhibit reduced electrophoretic mobilities relative to the unadducted oligonucleotides. This effect is largely independent of DNA sequence. Decreased mobility correlates with an increased rate of incorporation by pol , suggesting a systematic relationship between the overall DNA structure and efficiency of the enzyme.
Human pol , although bypassing T-T dimers largely by correct incorporation of A-A with an error rate of 10 Ϫ4 -10 Ϫ2 (12), is highly error-prone on undamaged DNA substrates compared with other nonproofreading DNA polymerases (13). It demonstrates a similarly high error frequency on substrates containing cis platin G-G adducts (14,15) and N-(deoxyguanosin-8-yl)-acetylaminofluorene (14). Moreover, it is significantly error-prone at the (6 -4) T-T photoproduct, inserting a single G opposite the T-T template bases (16). Polymerases that significantly misincorporate are candidates for creating the mutations that initiate carcinogenesis.
Bay-region diol epoxides (DEs) are the ultimate carcinogens derived from mammalian metabolism of polycyclic aromatic hydrocarbons (17)(18)(19). Four stereoisomeric 7,8-dihydroxy-9,10-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene (BaP DE) metabolites are possible, consisting of two diastereomeric pairs of enantiomers: DE-1 in which the benzylic 7-hydroxyl group is cis to the epoxide oxygen and DE-2 in which these groups are trans (Fig. 1A). The DEs react with the exocyclic amino groups of dG and dA residues in DNA to form trans and cis ring-opened adducts (20). Recently, we reported (21) that unlike the blockage seen when replicative polymerases encounter bulky adducts, translesion synthesis by human pol past BaP DE-2-modified dG residues occurs readily with preferred misincorporation of purine nucleotides and dAMP in particular, indicating a potential mechanism for generation of G⅐C to T⅐A mutations.
In the present study we report kinetic analyses of human pol translesion synthesis on DNA oligomers, each of which contained a single, stereospecific N 6 -dA adduct (Fig. 1B) (A*) of BaP DE-2 or a corresponding adduct derived from 9,10-epoxy-7,8,9,10-tetrahydrobenzo-[a]pyrene (BaP H 4 E, Fig. 1A), which has the 7-and 8-hydroxyl groups of BaP DE-2 replaced by hydrogens. Although the predominant DNA adducts formed by BaP DEs are generally at dG (22), a significant role for dA adducts is suggested by observations of dose-dependent differences in mutagenesis by the highly carcinogenic (ϩ)-(7R,8S,9S,10R)-BaP DE-2. Although mammalian cell studies involving high doses of (ϩ)-BaP DE-2 showed mostly G⅐C to T⅐A mutations in cells in culture, lower doses, which may be more reflective of environmental exposures, increased the frequency of mutations at A relative to mutations at G (23)(24)(25). The present study is consistent with a possible link between pol bypass synthesis opposite BaP DE-2-modified dA residues, and the observed A⅐T to T⅐A transversions arising in mammalian cells administered low doses of BaP DE-2.
Assays of base incorporation opposite two BaP DE-2 dA adducts using a single dNTP concentration and a fixed reaction time (26) indicated more efficient base incorporation and extension opposite the S adduct formed by trans ring opening of the highly carcinogenic (27) (ϩ)-(7R,8S,9S,10R)-BaP DE-2 than the R adduct arising from trans opening of its minimally carcinogenic (Ϫ)-(7S,8R,9R,10S)-enantiomer. Translesion synthesis beyond the trans S adduct was also fairly efficient, whereas the trans R adduct completely blocked extension beyond the adduct site. Pol incorporated more T than A opposite the trans S and largely A opposite the trans R adduct in a 5Ј-CA*A-3Ј template sequence. The present, more extensive kinetic analysis using a different sequence (5Ј-GAT-3Ј) also indicated higher incorporation and extension efficiency for pol opposite the trans S relative to the trans R adduct. However, in marked contrast to the previous results, pol showed little discrimination among A, G, and T incorporation opposite the trans S adduct and preferentially incorporated T opposite the trans R adduct.
We explored a possible relationship between the efficiency of base incorporation opposite BaP DE-dA adducts and adduct-induced changes in overall DNA structure and conformation, as reflected by altered gel electrophoretic mobility. With BaP DE-2 dG adducts, a decrease in mobility was observed when a trans S but not a trans R adduct was placed at a template-primer junction (28). Retardation by the trans S BaP DE-2 dG adduct was subsequently shown to be highly sequence-dependent (29). In the present study we measured the effect of an adduct on the electrophoretic mobility of a series of templateprimers containing each of the four isomeric cis-and trans-opened R and S BaP DE-2 dA adducts as well as two structurally related BaP H 4 E adducts. In addition, the sequence dependence of the electrophoretic mobilities of template-primers containing each of the BaP DE-2 adducts was investigated. The stereochemistry of the dA adducts affected gel electrophoretic mobility, and lower mobility was correlated with higher rates of nucleotide incorporation opposite the adduct by pol . However, in contrast to the observations with BaP DE-2 dG adducts, varying the nearest neighboring DNA sequence around a BaP DE-2 dA adduct at a template-primer junction had little impact on gel electrophoretic mobility.

MATERIALS AND METHODS
Oligonucleotides-The template sequence for pol assays was 5Ј-CAGA*TTTAGAGTCTGC-3Ј (Fig. 1C) in which A* represents a BaP DE-2-modified dA residue. Four stereoisomeric BaP DE-2 dA adducts were employed, representing the four possible trans or cis ringopened (ϩ)-and (-)-BaP DE-2 adduct isomers (Fig. 1B). Synthesis and characterization of the template oligonucleotides containing the four stereoisomeric BaP DE-2 dA adducts formed by trans (30) or cis (31) opening of the epoxide as well as the corresponding BaP H 4 E adducts (32) in which the 7-and 8-hydroxyl groups of the DE are replaced by hydrogen have been described. Studies of the effect of sequence on the gel electrophoretic mobility of adducted duplexes used BaP DE-2 adducted oligomers with sequences 5Ј-TTXA*YAGTCTGCTCCC-3Ј, where the neighboring bases X and Y were varied. Characterization of these oligonucleotides is described in the supplemental material. Unmodified templates and primers were synthesized and purified by Oligos Etc., Wilsonville, OR.
DNA Polymerase Eta Reactions-His-tagged recombinant human DNA polymerase was produced as described (14). Polymerase reactions contained final concentrations of 40 mM Tris-HCl (pH 8.0), 5 mM MgCl 2 , various concentrations of deoxynucleoside triphosphates, 10 mM dithiothreitol, 250 g/ml bovine serum albumin, 60 mM KCl, 2.5% glycerol, 40 nM 5Ј-32 P-labeled 12-mer primer previously annealed to 60 nM 16-mer template (by heating at 95°C for 5 min, and slowly cooling), and 2 nM pol in a 10-l volume for incorporation assays and one of four 13-mer primers ending in each of the four DNA bases for extension assays. Reactions proceeded at 37°C for 15 min and were stopped by the addition of 10 l of 90% aqueous formamide containing 25 mM EDTA and gel sequencing dye solution followed by heating in a boiling water bath.
Kinetic Analysis of DNA Polymerization-Reaction products were subjected to electrophoresis on 20% polyacrylamide, 7M urea gels run at 60 watts. Nucleotide incorporation across from adducts and extension beyond them was quantitated with a Fujifilm Fluorescent Image Analyzer and Image Gauge V3.12 software. Standing start, steady state kinetic experiments were performed as described previously (21) to quantitate the degree of misincorporation opposite the BaP DE-2-adducted dA. Michaelis-Menten V max /K m values were obtained from hyperbolic fits of V versus S as suggested by Creighton et al. (33). Misincorporation efficiency, f inc , was calculated by dividing the V max /K m for an incorrect nucleotide by the V max /K m for the correct T incorporation. Mispair extension efficiency, f ext , was calculated by taking the V max /K m for dCTP extension of a mispair (X⅐A*, where X ϭ G, A, or C) and dividing by the V max /K m for dCTP extension of the correctly paired T⅐A*.
Gel Electrophoretic Mobility Assays-The 12-mer and 13-mer primers were annealed with adducted or non-adducted template strands and subjected to electrophoresis in nondenaturing 20% polyacrylamide gels in a Hoefer 660 apparatus at 0 -4°C. Bands were positioned with the imaging methods above. Gel electrophoretic mobility was calculated as R f ϭ distance traveled by the adducted duplex/distance traveled by the unadducted duplex.

Patterns of Nucleotide Incorporation Opposite BaP DE-2 dA Adducts-
The patterns of misincorporation by human pol at BaP DE-2 dA adducts in the 16-mer/12-mer sequence described are shown semiquantitatively in Fig. 2. In the absence of adducts, pol preferred to incorporate the correct T (virtually all the primer is extended in the T lane), whereas misincorporation of G or A exceeded misincorporation of C. When all four deoxyribonucleoside triphosphates (dNTPs) were present, pol was able to process to the end of the unadducted template.
All adducts sharply reduced incorporation of the correct T but had much less effect on the various misincorporations. Trans S BaP DE-2 dA adducts gave equivalent incorporations of G, A, and the correct T, whereas trans R adducts gave less misincorporation of G and A than incorporation of T. Cis R or cis S dA adducts also induced approximately similar levels of misincorporation of A and G and correct incorporation of T. The overall levels of (mis)incorporation were highest with the trans S adduct. This adduct gave rise to the greatest enzymatic processivity when all four dNTPs were present; some full-length product was seen. (Mis)incorporation appeared to be the least efficient overall opposite the cis R adducts, and no extension beyond the adducted base was observed when either one or all four dNTPs were present. With all four stereoisomeric BaP DE-2 adducted dA templates, misincorporation of A was slightly greater than misincorporation of G, and misincorporation of C was the least efficient. Adducts with S configuration gave more (mis)incorporation and enzymatic processivity than the corresponding adducts with R configuration.
Kinetic Comparisons of the Rates of Incorporation at Stereoisomeric BaP DE-2 or BaP H 4 E dA Adducts-Steady state kinetic data for incorporation opposite dA in the presence and absence of BaP DE-2 adducts are presented in TABLE ONE. Even in the absence of adducts, pol was error-prone at template dA. Thus, values of f inc , defined as (V max /K m ) incorrect /(V max /K m ) correct , ranged from 2 ϫ 10 Ϫ2 to 5 ϫ 10 Ϫ2 for the unmodified template. In the presence of BaP DE-2 adducts, values of V max /K m for correct incorporation of T decreased by factors of 18 -160, whereas V max /K m values for misincorporation generally decreased less or even increased relative to the control. In consequence, f inc values for base misincorporation opposite the four stereoisomeric BaP DE-2 adducts were large (from 0.1 to 3.0) when compared with those with the unadducted template.
The stereochemistry of the adduct had a substantial effect on both the rate and fidelity of base incorporation. Thus, the four isomeric BaP DE-2 dA adducts exhibited a nearly 9-fold variation in their V max /K m values for correct T incorporation, with a corresponding effect on the f inc values. With a trans R adduct present, the V max /K m for incorporation of T exceeded the rate for A by 3-fold, whereas the other three adduct isomers showed an opposite selectivity such that the rate of incorporation of A exceeded that of T by up to 3-fold (with the cis S adduct). With each of the four isomeric BaP DE-2 dA adducts, V max /K m for G misincorpo-ration was about 70% of the V max /K m for A misincorporation, and misincorporation of C was least efficient.
Nucleotide incorporation opposite the trans R adduct was the least error-prone, with all values of f inc for incorrect bases Ͻ1.0. Although the highest V max /K m for correct T incorporation was observed opposite the trans S adduct, this was accompanied by parallel increases in V max /K m for incorporation of the incorrect bases, such that this adduct exhibited the highest rate of incorporation of each nucleotide, correct or incorrect, but fidelity opposite this adduct was low, with f inc values close to 1 for misincorporation of G or A. Cis R and cis S BaP DE-2 dA adducts exhibited similar poor fidelity (high f inc ). As suggested qualitatively by Fig. 2, the S adducts generally exhibited about a 5-fold greater rate of G, A, and C misincorporation compared with their R counterparts. This is true whether the comparison is between trans R versus trans S adducts and cis R versus cis S adducts, derived from enantiomeric BaP DE-2 isomers, or trans R versus cis S and cis R versus trans S adducts, derived from the same BaP DE-2 isomer. The presence of an R adduct lowered the V max /K m for G, A, or C misincorporation 3-5-fold compared with the unadducted template, whereas in most cases an S adduct either caused no change or a slight increase (up to 2-fold) in V max /K m .
Nucleotide incorporation by pol opposite two BaP H 4 E adducts with R configuration, which lack the 7-and 8-hydroxyl groups present in the DE adducts, is generally less efficient compared with BaP DE-2 adducts with the same absolute configuration. The reduction in rate was less for misincorporation of A than for any other nucleotide, with two consequences. The f inc values for A misincorporation were higher with BaP H 4 E adducts than with the corresponding BaP DE-2 adducts, and the rates for A misincorporation were ϳ3 times that for G with the BaP H 4 E adducts, in contrast to the less than 2-fold rate difference between these nucleotides observed with the DE adducts.

Kinetic Comparisons of the Rates of Extension and Overall Translesion Synthesis Beyond (Mis)pairs Containing BaP DE-2 or BaP H 4 E dA
Adducts-Kinetic data for extension of 13-mer primers with a correct or incorrect base opposite the adduct are also presented in TABLE ONE. V max /K m for extension beyond a correctly or incorrectly paired dA DE adduct was generally smaller than V max /K m for either correct or incorrect base incorporation opposite the adduct. V max /K m for extension beyond adduct (A*)-containing mispairs was well below V max /K m for mispair extension in the absence of adduct. Although for unadducted DNA mispair extension was more than 10 times less efficient than extension beyond a correct A-T base pair, extension of the adducted template-primers showed less discrimination between mispairs and correctly paired A*⅐T. In particular A*⅐A mispairs were extended with nearly equal or greater efficiency than correct A*⅐T pairs. A*⅐C mispairs were extended somewhat less efficiently, whereas A*⅐G mispairs were extended the least well.
Extension is more efficient past BaP DE-2 adducts with S configuration compared with R adducts, in keeping with the greater processivity seen in the four dNTP lanes with S adducts compared with R adducts (Fig. 2). The V max /K m for extension when an S adduct was present was 2-10-fold greater than the value for the comparable R adduct, with the greatest difference in the rate of extension at R versus S adducts seen with extension of A*⅐G mispairs. The differences in rate between trans and cis adducts with the same absolute configuration at C-10 were small, and extension past the cis R adduct was the least efficient.
With BaP H 4 E adducts, V max /K m values for extension ranged from equal to severalfold greater than the corresponding V max /K m values for incorporation, the opposite of what was seen with BaP DE-2 adducts. Whereas rates for either correct or incorrect nucleotide incorporation opposite BaP H 4 E adducts were lower than for the comparable reactions opposite BaP DE-2 adducts, 13-mer primers with A and T across from BaP H 4 E trans R adducts gave extension rates equal to the highest rates of extension seen with BaP DE-2 adducts. With the trans R BaP H 4 E adduct, V max /K m values for extension were equal to or higher than those for incorporation regardless of which nucleotide was extended, yielding f ext values similar to f inc . With the cis R BaP H 4 E adduct, V max /K m for extension was highest when a pyrimidine (C or T) was opposite the adduct, in contrast to the greater efficiency of A incorporation opposite this adduct.
Multiplying f inc ϫ f ext for all misincorporation/extension combinations (TABLE ONE) allows comparison of the overall efficiency of pol translesion synthesis at BaP DE-2 or BaP H 4 E dA adducts. Bypass via A*⅐A mispairs is the most efficient, with values of f inc ϫ f ext generally close to or greater than 1 in the presence of adducts compared with 4.9 ϫ 10 Ϫ3 in their absence. All cis ring-opened dA adducts showed greater overall efficiency for bypass of an A*⅐A mispair than for bypass of a correct A*⅐T. Trans S BaP DE-2 dA adducts, derived from (ϩ)-(R,S,S,R)-BaP DE-2, allowed extension of a primer containing a mispaired A nearly as often as one containing a correctly paired T. However, translesion synthesis using the correct A*⅐T pair was favored over translesion synthesis with A*⅐A by severalfold with the trans R adduct derived from (Ϫ)-(S,R,R,S)-BaP DE-2. Bypass with A*⅐G or A*⅐C mispairs was generally less efficient, often by a considerable margin.
Correlations between Rates of Incorporation and Reduced Gel Electrophoretic Mobility of Template-primers Containing BaP DE-2 or BaP H 4 E dA Adducts-Placement of a BaP DE-2 adduct at the template-primer junction in the 16-mer template/12-mer primer used in our kinetic studies results in mobility shifts on a non-denaturing gel that are stereoisomer-dependent (Fig. 3). Relative mobility was quantified by use of the equation R f ϭ d add /d u , where d add represents the distance migrated by the adducted duplex, and d u represents the distance migrated by the unadducted (control) duplex. The changes in gel electrophoretic mobility with changing adduct stereoisomer (range in R f from 0.93 for trans S BaP DE-2 adducts to 0.99 for BaP H 4 E adducts), although small compared with the effect seen with G adducts (range in R f of 0.92 for cis S BaP DE-2 adducts to 0.76 for trans S BaP DE-2 adducts) (data not shown), are highly reproducible.
The availability of kinetic rates of incorporation permits the plotting of V max /K m for incorporation by pol of each dNTP at each type of adduct against gel mobility, R f (Fig. 4). The individual V max /K m values were added together to make a composite V max /K m for all four dNTPs. It is clear that the greater the decrease in gel mobility induced by a given BaP DE-2 dA adduct, the larger the V max /K m value for (mis)incorporation. Trans R and cis R BaP H 4 E adducts induced the least reduction in gel electrophoretic mobility and showed the lowest V max /K m values for incorporation by pol . The best correlations with gel mobility (R 2 ϳ 0.8) are seen with purine nucleotide misincorporation and the sum of the V max /K m values for all four nucleotides.
The effect of BaP adducts on gel electrophoretic mobility of duplexes formed by annealing the adducted templates with the 13-mer primers used to determine rates of mismatch extension was assessed as well. In general trans S and cis S BaP DE-2 adducts induced the greatest decrease in gel mobility, whereas cis R BaP H 4 E adducts induced the least. How-  ever, unlike what was observed with incorporation on the 12-mer primers, moderate correlation between gel mobility and rate of (mis)pair extension was only seen with extension of A and T and only if the BaP H 4 E adducts were excluded (data not shown).

Lack of Effect of DNA Sequence Changes on the Adduct-determined Rank Order of Gel Electrophoretic Mobility of BaP DE-2 dA Adducted
Template-primers-Because the electrophoretic mobility of templateprimers containing BaP DE-2 dG adducts had been shown to be highly sequence-dependent (29), we examined the effect of sequence on the electrophoretic mobility of BaP DE-2 dA-adducted template-primers. For these studies template sequences 5Ј-TTXA*YAGTCTGCTCCC-3Ј, where X and Y are the nearest neighbors to a BaP DE-2-adducted A*, were annealed to a complementary 12-mer primer. The R f values seen with the four BaP DE-2 isomers in these eight sequences followed the same pattern (cf. Fig. 3) seen with the different 16-mer/12-mer (local sequence GA*T) employed in our kinetic studies, with the trans S adduct resulting in the greatest retardation and the cis R the least. There is little effect of nearest neighbor sequence on the variation in gel mobility with adduct isomer (Fig. 5). The lines connecting R f values for a given adduct isomer are mostly parallel, with only two crossovers seen. Thus, it appears that the nearest neighbor sequence surrounding a BaP dA adduct at a template-primer junction has relatively little influence on the rank order of gel electrophoretic mobility induced by adduct stereochemistry.

DISCUSSION
Overview-The present studies on bypass synthesis by human pol at BaP DE-2 adducts aim to determine the structural bases for why stereoisomeric adducts induce different specificities and rates of nucleotide incorporation and extension. X-ray crystal structures for a number of Y-family polymerases have been described (34 -36), including the catalytic fragment of yeast pol (34) and a Dpo4 structure with a BaP DE-2-adducted DNA and an incoming dNTP present (35).  Lower values of R f correspond to slower migration of the adducted template-primers relative to the unadducted control (cf. Fig. 3).
These show the Y family polymerases to have open active sites that accommodate adducts. However, because no such structure for human pol exists, the analysis will first consider the structures of adducted oligonucleotides.
For this discussion it is hypothesized that polymerase specificity depends on the relative populations of DNA template-primer junctions containing adducts and the preference exhibited by the polymerase for those structures that best fit the enzyme active site. Previously we postulated that the high degree of purine misincorporation by human pol at BaP DE-2 dG adducts depended on  intermediates with the adducted dG in the unusual syn glycosidic conformation (21); a similar structural explanation also applies to the lesser but still high degree of purine misincorporation by human pol at BaP DE-2 dA adducts observed in the present study.
Fidelity of Translesion Synthesis by pol at dA Adducts Versus dG Adducts-The strong preference for purine nucleotide misincorporation at BaP DE-2 dG adducts (21) is partly due to enhancement of V max /K m for misincorporation. In the presence of dG adducts, values of V max /K m for A misincorporation were increased relative to misincorporation opposite the normal base by factors ranging from 1.6-fold for cis S dG adducts to nearly 8-fold for trans S dG adducts in a 5Ј-CG*A-3Ј sequence. On the other hand, TABLE ONE of the present study shows that with dA adducts there is no generalized increase in the rate of purine misincorporation, with V max /K m in the presence of adduct ranging from 0.2 to 2 times the V max /K m observed with undamaged DNA. For A misincorporation, the presence of adducts yields V max /K m values ranging from 0.19 to 1.15 times what is seen in the absence of adducts. S dA adducts showed rates of incorrect base incorporation that were close to those observed with the unadducted control, but only the trans S adduct gave a rate enhancement (ϳ2-fold) for misincorporation of G.
In addition to increasing the rates of purine misincorporation, dG adducts cause a greater decrease in the rate of incorporation of the correct nucleotide (C) than do dA adducts (T incorporation). Thus, lower f inc values for purine misincorporation are seen with BaP DE-2 dA versus dG adducts. Decreased fidelity at BaP DE-2 dA adducts fits the model for polymerase fidelity put forth by Beard et al. (38), where the decrease in fidelity is caused mostly by a decrease in the rate of incorporation of the correct nucleotide, whereas rates of misincorporation change little, but it leaves the enhancement of the rate of purine misincorporation at dG adducts without an explanation.
The effects of dG and dA adducts on extension differ as well. With dG adducts, V max /K m for extension (on the order of 10 Ϫ7 M primer extended/M dNTP/min) is generally, with the exception of cis R adducts, 10-fold lower than V max /K m for misincorporation, and at the extreme, V max /K m for extension of an A⅐G* mispair at a trans S adduct is only 1.3% of the V max /K m for A misincorporation across from G*. Thus, with dG adducts, the combination of adduct stereoisomer and incoming purine nucleotide producing the highest rate of misincorporation also exhibits the slowest rate of extension. With dA adducts, rates of extension are generally only a few-fold lower than the rates for incorporation, except in the case of G mispairs, whose rates of extension are an order of magnitude lower. Thus, the overall rate of bypass of dA adducts is higher than dG adducts. The efficiency of bypass by misincorporating purines (f inc ϫ f ext ), however, is low at dA adducts compared with dG adducts because C is poorly incorporated and extended at dG adducts, whereas T is well incorporated and extended at dA adducts.
Adducted DNA Structure and Nucleotide Incorporation Specificity-NMR structures of DNA template-primer junctions containing either no base (39) or a correct C (40) opposite a trans S BaP DE-2 dG adduct have been determined. In these structures the hydrocarbon lies stacked with the previous (3Ј on the adducted strand) base pair when the primer strand terminates one base before the adduct (39), but when C is added across from the adducted G, the adduct moves into the minor groove, oriented toward the 5Ј end of the adducted strand (40). This places the adduct in the direction of extension. Thus, it is not surprising that when C is placed across from trans S BaP DE-2 adducted G, extension by pol becomes undetectable (21). No NMR structures exist for template-primers containing BaP DE-2 adducts on dA, although structures have been determined for duplex DNA with a centrally located adduct. These structures indicate that duplexes containing trans-opened BaP DE-2 adducts with R configuration and a correctly paired T opposite the adduct have the hydrocarbon intercalated toward the 5Ј side of the modified A (41)(42)(43). The single mispaired R dA adduct that has been examined (44) had a G opposite a trans R dA adduct. In this structure the hydrocarbon is intercalated to the 5Ј side of the adducted A, the mispaired G is turned out into the major groove, and the dA* residue is in the normal anti conformation. Duplex DNA containing a trans S BaP DE-2 dA adduct correctly paired with T has to date been intractable to NMR structure determination because the adduct is disordered on the NMR time scale and/or adopts multiple conformations; however, an analogous trans S BaP DE-1 dA adduct (45) paired with T intercalates toward the 3Ј side of the modified A, and the adducted dA predominantly adopts a syn glycosidic conformation. This preference for the syn conformation is also seen in a 9-mer duplex containing a trans S BaP DE-2 dA adduct in a Topal-Fresco (37) mispair with G (46). Such a mispair stabilizes the S adducted duplex, such that T m values for 9-mer duplexes containing A and G mispairs with the trans S BaP DE-2 dA adduct are higher than those for the correctly paired duplex containing complementary T. In contrast, the T m of the duplex containing the correct A*⅐T pair with the trans R adduct was comparable or higher than either purine mismatch (44). Available data suggest that the cis BaP DE-2 dA adducts follow the same structural trends as the trans adducts. As in the case of the trans adducts, the cis R BaP DE-2 dA adduct has the hydrocarbon intercalated toward the 5Ј side of the modified the dA residue, which retains a normal anti conformation about the glycosidic bond (47). In contrast, the cis S adduct, like its trans S counterparts, has to date not permitted NMR characterization.
Based on the available NMR evidence, it seems likely that the conformational heterogeneity exhibited by the trans S dA adducts in DNA duplexes may include a significant syn-rotated component, and it is possible that a similar involvement of syn conformations may also pertain to cis S adducts. In our in vitro assays both the cis and trans S adducted templates are significantly better substrates for incorporation by pol than templates containing the corresponding cis and trans R adducts. In particular, the V max /K m for purine misincorporation is 4 -6fold higher at both cis and trans S BaP DE-2 dA adducts compared with the corresponding R adducts. Thus, the general observation of relatively efficient purine misincorporation opposite S adducts is consistent with involvement of a Topal-Fresco intermediate. Such an intermediate is particularly attractive since pol from yeast requires hydrogen bonding between the incoming nucleotide and the template base (48,49). Although Hoogsteen base pairing is probably not involved with a T template base (49), the analogous Topal-Fresco hydrogen bonding is more likely when two purines are mispaired. The small preference for A over G misincorporation may be in part a reflection of the "A rule" for incorporation opposite non-instructional lesions (50). However, it is also consistent with Topal-Fresco hydrogen bonding since creating hydrogen bonds in a Topal-Fresco mispair between a syn-oriented purine and G requires a double tautomer of G (enol, imino), whereas in a purine-A mispair the A is tautomeric at a single site (imino). The relative preferences for forming the requisite G and A tautomers could differ. With BaP DE-2 dA adducts, V max /K m values for G incorporation clustered around 74% of the V max /K m for A incorporation regardless of adduct configuration (range 63-87%). With the two BaP H 4 E dA adducts, the relative V max /K m also favors A incorporation such that G incorporation is only ϳ35% that of A incorporation.
Adducted DNA Structure and V max /K m Values for Extension-As noted above, BaP DE-dA adducts with R configuration in duplex DNA (42)(43)(44) intercalate to the 5Ј side of the modified A, in the direction of further extension, whereas adducts with S configuration intercalate to the 3Ј side (45,46), away from further extension. If this is a reflection of what is encountered during extension, it could explain the higher V max /K m for extension exhibited by S adducts compared with their R counterparts (TABLE ONE), and the overall greater processivity that pol shows with S adducts compared with R adducts (Fig. 2). Although relative efficiencies of A and G misincorporation opposite an adduct differ by less than a factor of 2, extension of A*⅐A mispairs was considerably more efficient (ϳ3-8-fold) than extension of A*⅐G mispairs in all cases except the cis H 4 E adduct. This preference for extension of A*⅐A mispairs could result from differences in the structure or stability of the two types of hydrogen-bonded Topal-Fresco mispairs (see the previous paragraph) which the enzyme encounters on extension. Because pol has been selected by evolution to bypass T-T UV dimers by inserting A-A (9, 10), it is possible that structural feature(s) of the enzyme have evolved that favor insertion of A opposite other lesions in addition to T-T dimers as well as recognition of an A already present opposite a lesion as the "correct" choice for further extension.
T-T dimers in duplex DNA create a bend in the DNA structure (51). Therefore, it may be hypothesized that the preferred substrate for pol is one that is bent, and it would not be surprising that within the series of BaP DE-2 and BaP H 4 E dA adducts, the greater the bend induced by the adduct, the higher the V max /K m for incorporation will be. Although it is not clear what structural factors lead to the observed stereochemistrydependent reduction in gel mobility of the template-primers (Fig. 3), these mobility differences could reflect structural changes in the overall shape of the DNA imposed by different dA adducts. In the case of duplex DNA, gel mobility retardation by BaP DE adducts has been ascribed to induction of bends or flexible hinge joints at the site of the DNA adducts (Refs. 52 and 53 and references therein). Retardation of oligonucleotides containing template-primer junctions by BaP DE-dG adducts has also been ascribed to DNA bending (29). Thus, it is notable that increasing V max /K m for incorporation correlates with a decrease in gel mobility, especially for purine misincorporation. The two BaP H 4 E dA adducts induce the least reduction in gel mobility. An NMR structure for the cis R BaP H 4 E adduct opposite T in the center of a DNA duplex (54) shows far less local distortion induced by the adduct than a corresponding NMR structure for a cis R BaP DE-2 adduct (47). It must be borne in mind, however, that the effects of adducts in fully duplexed DNA may be quite different from their effects at a template-primer junction.
Relationships between dA Adduct Bypass Activity and Crystal Structures of Y-family Polymerases-To date crystal structures are unavailable for human pol . The use of existing crystal structures for analogous Y-family polymerases to interpret the efficiency and fidelity of human pol at BaP DE-2 dA adducts in template-primer junctions is subject to various caveats, as is discussed below. A crystallographic structure has been determined for the catalytic core of yeast pol (34). However, the crystals did not contain DNA, whose orientation relative to the protein had to be modeled, and prediction of the location of bulky hydrocarbon adducts would not be possible. Furthermore, even though the yeast and human enzymes have both evolved to operate on cis-syn T-T dimers, yeast pol incorporates nucleotides by a different mechanism compared with the human enzyme (55).
Dpo4 and Dbh are archaeal polymerases for which crystal structures have been reported (35,56). The Dbh crystals did not include DNA, whereas one Dpo4 structure is based on crystals of Dpo4 bound to a template-primer containing a cis ring-opened BaP DE-2 dA adduct in the template strand across from its correct primer partner T at the template-primer junction. An incoming nucleotide, dATP, is paired with the next template base, T. In this structure two conformations are observed per unit cell (35). One has the hydrocarbon moiety intercalated between the end of the primer and the incoming nucleotide. The other has the hydrocarbon moiety in the minor groove. The intercalated structure parallels the NMR structure observed, with the adduct placed in the middle of a DNA duplex, lending credence to the preceding discussion based on NMR structures. The adduct in the minor groove structure faces away from the bulk of the enzyme, so that in both conformations the bulky adduct can be easily accommodated in the active site along with an incoming nucleotide. The open, adduct-accommodating active site of Dpo4, observed with both conformations of the adducted template-primer, is characteristic of most Y-family DNA polymerases, including the catalytic core of yeast pol (34) and Dbh (56).
The Dpo4 crystal complex does not turn over. Failure to do so could reflect either the unusual presence of 0.1 M calcium acetate or incorrect spacing of catalytic amino acid residues and metal ions. In the Dpo4 crystals, the normal Mg(II) was replaced by the larger Ca(II). It has been argued (57) that Ca(II) is an adequate model for Mg(II) since the positions of the two Ca(II) ions in a Dpo4 structure are superimposable with the two Mg(II) ions in the structure of a ternary complex of T7 DNA polymerase. However, in addition to potentially changing the geometry of the complex with substrate, the much larger Ca(II) should also be less effective in polarizing a P-O bond to catalyze nucleotidyl transfer.
The adducted template-primer in the Dpo4 structure represents extension beyond a correct T in the primer across from the adducted A in the template. Thus, this crystal structure cannot shed light on whether anti to syn rotation of the adducted A or the incoming nucleotide occurs in the intermediate for purine misincorporation across from such an adduct.
Crystal structures for the human Y-family DNA polymerases (58) and (59) have also been published. However, the structure for pol has been criticized; " . . . the pol crystals appeared to be partially disordered, and the reported structure contains a large number of stereochemical abnormalities" (36). The crystal structure for the catalytic core of human pol (59) in the absence of DNA shows that pol has a more restricted active site than other Y-family enzymes, which could account for its greater fidelity on undamaged DNA relative to pol . Another critical difference between pol and pol is in their processing of BaP DE-2 N 2 dG adducts. Human pol inserts the correct dCMP opposite a trans 10S BaP DE-2 dG adduct (60), in sharp contrast to the preferential incorporation of purine nucleotides across from the same adducted dG exhibited by human pol (21).

Mutational Specificities of dA Adduct Bypass by Pol Compared with Mammalian Cells Treated with BaP DE-2-
The values of f inc ϫ f ext (TABLE ONE) show that when compared with G⅐A* or C⅐A* mispairs, pol is biased in favor of bypass synthesis using A⅐A* mispairs, which in the absence of repair would result in A⅐T to T⅐A transversions. This result is consistent with mutations observed in mammalian cells containing BaP DE-dA adducts (23)(24)(25)61).
In particular, mutational spectra in repair-deficient Chinese hamster VH-1 cells (61) treated with (ϩ)-BaP DE-2 provide the simplest comparison with our enzymatic data in vitro, since potential repair processes are not superimposed on the mutational results. In these studies A⅐T to T⅐A transversions predominated for mutations at A with both high and low doses of (ϩ)-BaP DE-2. Similar results were also obtained with repair-proficient V79 cells (23,24), indicative of incomplete repair of the dA adducts from which the observed mutations arise. Although it is unknown what enzyme(s) are responsible for error-prone bypass of BaP DE-dA adducts leading to mutations in mammalian cells, these results are consistent with a role for pol in this process.