Translesion synthesis past acrolein-derived DNA adduct, γ-hydroxypropanodeoxyguanosine, by yeast and human DNA polymerase η

γ-Hydroxy-1,N 2-propano-2′deoxyguanosine (γ-HOPdG) is a major deoxyguanosine adduct derived from acrolein, a known mutagen. In vitro, this adduct has previously been shown to pose a severe block to translesion synthesis by a number of polymerases (pol). Here we show that both yeast and human pol η can incorporate a C opposite γ-HOPdG at ∼190- and ∼100-fold lower efficiency relative to the control deoxyguanosine and extend from a C paired with the adduct at ∼8- and ∼19-fold lower efficiency. Although DNA synthesis past γ-HOPdG by yeast pol η was relatively accurate, the human enzyme misincorporated nucleotides opposite the lesion with frequencies of ∼10−1 to 10−2. Because γ-HOPdG can adopt both ring closed and ring opened conformations, comparative replicative bypass studies were also performed with two model adducts, propanodeoxyguanosine and reduced γ-HOPdG. For both yeast and human pol η, the ring open reduced γ-HOPdG adduct was less blocking than γ-HOPdG, whereas the ring closed propanodeoxyguanosine adduct was a very strong block. Replication of DNAs containing γ-HOPdG in wild type and xeroderma pigmentosum variant cells revealed a somewhat decreased mutation frequency in xeroderma pigmentosum variant cells. Collectively, the data suggest that pol η might potentially contribute to both error-free and mutagenic bypass of γ-HOPdG.

␥-Hydroxy-1,N 2 -propano-2deoxyguanosine (␥-HOPdG) is a major deoxyguanosine adduct derived from acrolein, a known mutagen. In vitro, this adduct has previously been shown to pose a severe block to translesion synthesis by a number of polymerases (pol). Here we show that both yeast and human pol can incorporate a C opposite ␥-HOPdG at ϳ190and ϳ100-fold lower efficiency relative to the control deoxyguanosine and extend from a C paired with the adduct at ϳ8and ϳ19-fold lower efficiency. Although DNA synthesis past ␥-HOPdG by yeast pol was relatively accurate, the human enzyme misincorporated nucleotides opposite the lesion with frequencies of ϳ10 ؊1 to 10 ؊2 . Because ␥-HOPdG can adopt both ring closed and ring opened conformations, comparative replicative bypass studies were also performed with two model adducts, propanodeoxyguanosine and reduced ␥-HOPdG. For both yeast and human pol , the ring open reduced ␥-HOPdG adduct was less blocking than ␥-HOPdG, whereas the ring closed propanodeoxyguanosine adduct was a very strong block. Replication of DNAs containing ␥-HOPdG in wild type and xeroderma pigmentosum variant cells revealed a somewhat decreased mutation frequency in xeroderma pigmentosum variant cells. Collectively, the data suggest that pol might potentially contribute to both error-free and mutagenic bypass of ␥-HOPdG. Acrolein ( Fig. 1), the simplest ␣,␤-unsaturated aldehyde, is an environmental contaminant and a product of inborn metabolism. In organisms, acrolein is generated via a number of pathways, such as the oxidation of polyamines and lipid peroxidation (1,2). Like many other bifunctional aldehydes, acrolein reacts with DNA bases to form several DNA adducts, among which the ␥-hydroxy-1,N 2 -propano-2Јdeoxyguanosine (␥-HOPdG) 1 was identified as a major deoxyguanosine (dG) derivative (3,4). Importantly, ␥-HOPdG has been detected in DNA from mammalian tissues (5)(6)(7), suggesting that this ad-duct is generated in vivo. The ␥-HOPdG adduct is formed by conjugate addition of acrolein to N 2 of dG to produce N 2 -(3oxopropyl)dG. Ring closure at N1 leads to the formation of the cyclic adduct (Fig. 1). In the nucleoside and presumably in single-stranded DNA, ␥-HOPdG predominantly exists in the cyclic form, such that at physiological pH, the ring open species cannot be detected spectrophotometrically (8). However, in the presence of a reducing agent, the acyclic form can be trapped as the N 2 -(3-hydroxypropyl) adduct ( Fig. 1).
Another dG derivative, 1,N 2 -propanodeoxyguanosine (PdG) (Fig. 1), whose structure is similar to that of the ring closed ␥-HOPdG, has been extensively exploited as a model compound for the ␥-HOPdG and other exocyclic dG adducts in both structural and biological studies. NMR spectroscopy of the PdGadducted oligodeoxynucleotides has revealed that when placed opposite dC, PdG adopts a syn orientation within the duplex and introduces a localized structural perturbation that is pHand sequence-dependent (9,10). The inability of PdG to form normal Watson-Crick hydrogen bonds severely blocks DNA synthesis both in vitro (11,12) and in vivo (13)(14)(15)(16), and the replication that does occur results in mutations (13)(14)(15)(16). Specifically, PdG-induced base substitutions occurred at an overall frequency of 7.8 ϫ 10 Ϫ2 and 7.5 ϫ 10 Ϫ2 /translesion synthesis in the COS-7 (14) and in the nucleotide excision repair-deficient human cells (16), respectively. In both strains, G to T transversions predominated.
Recently, the structure of the ␥-HOPdG-containing oligodeoxynucleotide was solved by NMR spectroscopy (17). These data have indicated that within the duplex, ␥-HOPdG exists primarily in the ring open form. In such a conformation, the modified base participates in standard Watson-Crick base pairing by adopting a regular anti orientation around the glycosidic torsion angle, with the N 2 -propyl chain in the minor groove pointing toward the solvent (17). The structural differences between PdG and ␥-HOPdG within the duplex have led to the hypothesis that the latter lesion would be less blocking for replication and less mutagenic than the former.
Biological studies aimed to test the cytotoxic and mutagenic effects of acrolein-modified DNAs and of site-specific ␥-HOPdG adduct have generated conflicting results. It is known that acrolein itself causes mutations in both bacterial (18) and mammalian (19) systems and has tumor-initiating activity (20). When a DNA vector was treated with acrolein and propagated in human cells, the majority of mutations were single, tandem, and multiple base substitutions that predominantly occurred in G:C base pairs (21). However in bacteria, ␥-HOPdG, the major acrolein-derived dG adduct, is not a strong block for DNA synthesis nor a miscoding lesion (22)(23)(24). Analyses of mutations caused by ␥-HOPdG in wild type Escherichia coli and in polB, dinB, and umuDC deficient strains revealed that in the absence of these "SOS" polymerases, the efficiency and accuracy of the translesion synthesis were not significantly affected (22). In contrast to the prokaryotic data, ␥-HOPdG caused mutations at an overall frequency of 7.4 ϫ 10 Ϫ2 /translesion synthesis when a single-stranded, site-specifically modified vector was propagated in COS-7 cells (24). Interestingly, both the frequencies and types of mutations were remarkably similar to those reported for the PdG adduct (14,16). However, ␥-HOPdG was shown to be only marginally miscoding (Յ1% base substitution) when double-stranded vector was utilized (16). In this investigation, a number of cell lines including HeLa, a nucleotide excision repair-deficient xeroderma pigmentosum group A, and polymerase -deficient xeroderma pigmentosum variant were examined.
Although replication across ␥-HOPdG in vivo was predominantly error-free (from 93 to 100% of the translesional events), the adduct was shown to be a severe block and a miscoding lesion during in vitro DNA synthesis by a number of polymerases. Particularly, replication across ␥-HOPdG by the Klenow exo Ϫ fragment of E. coli polymerase I was significantly inhibited and extremely error-prone (22,23). ␥-HOPdG also strongly blocked DNA synthesis by two major eukaryotic polymerases, pol ␦ and pol ⑀ (24). In the presence of proliferating cell nuclear antigen, little bypass of the adduct by pol ␦ was achieved, and it appeared to be highly mutagenic (24). We hypothesized therefore that in mammalian cells, specialized, translesion DNA synthesis polymerases (25,26) are involved in promoting replication across ␥-HOPdG.
Among DNA polymerases proficient in translesion synthesis, yeast polymerase (a product of the RAD30 gene) (27) and its human counterpart (a product of the RAD30A (XPV, POLH) gene) (28,29) both possess a unique ability to replicate efficiently and accurately past a cis-syn cyclobutane pyrimidine dimer (30,31), the predominant DNA lesion caused by ultraviolet irradiation. In the yeast Saccharomyces cerevisiae, deletion of RAD30 confers moderate sensitivity to UV irradiation and leads to increased UV-induced mutagenesis (32). Mutations in the human RAD30A gene cause the variant form of xeroderma pigmentosum (XPV), suggesting that predisposition of XPV individuals to sunlight-induced skin cancer is due to the lack of accurate translesion DNA synthesis across UV-induced DNA lesions (28,29,33). Yeast and human pol also efficiently bypass a product of oxidative DNA damage, the 7,8-dihydro-8oxoguanine, and do so in a predominantly error-free manner (34). In addition, several other DNA lesions were reported to be substrates for human (35)(36)(37)(38)(39) and yeast (35,40) pol .
In the present study, the ability of yeast and human pol to perform translesion DNA synthesis across ␥-HOPdG has been examined, and the efficiency and fidelity of synthesis have been tested using steady-state kinetic analyses. To further explore the bypass mechanism, comparative studies were also performed with two model DNA adducts: PdG, which mimics the cyclic form of ␥-HOPdG, and N 2 -(3-hydroxypropyl)dG, which is similar to ␥-HOPdG in its ring open form. In addition, the mutagenic potential of ␥-HOPdG was tested in vivo in both human fibroblasts and pol -deficient XPV cells utilizing a site-specifically modified single-stranded pMS2 vector.

EXPERIMENTAL PROCEDURES
Materials-T4 DNA ligase, T4 polynucleotide kinase, and EcoRV were obtained from New England BioLabs (Beverly, MA). S1 nuclease and proteinase K were purchased from Invitrogen. [␥- 32  Oligodeoxynucleotides-12-mer oligodeoxynucleotide modified with ␥-HOPdG, 5Ј-GCTAGC(␥-HOPdG)AGTCC-3Ј, was kindly provided by Dr. T. M. Harris and Dr. C. M. Harris (Vanderbilt University, Nashville, TN), and it was prepared by a previously described procedure (8). The 24-mer oligodeoxynucleotide, 5Ј-GCAGTATCGCGC-(PdG)CGGCATGAGCT-3Ј, adducted with PdG was synthesized as described (41) and was a generous gift from Dr. L. J. Marnett (Vanderbilt University, Nashville, TN). Nondamaged 12-and 24-mer with a dG in place of ␥-HOPdG or PdG, respectively, were purchased from Midland Certified Reagent Co. (Midland, TX). All of the other oligodeoxynucleotides were synthesized by the Molecular Biology Core Laboratory of the National Institute of Environmental Health Sciences Toxicology Center at the University of Texas Medical Branch (Galveston, TX) and purified by electrophoresis through a 15% denaturing PAGE (in the presence of 7 M urea).
Construction of site-specifically modified linear templates for in vitro replication assays was done according to the previously described procedure (24). Sequences of the resulting oligodeoxynucleotides were identical: 5Ј-GCTAGCGAGTCCGCGCCAAGCTTGGGCTGCAGCAGG-TC-3Ј, where the underlined G is either ␥-HOPdG or nonadducted dG and 5Ј-GCAGTATCGCGCGCGGCATGAGCTGCGCCAAGCTTGGGC-TGCAGCAGGTC-3Ј, where the underlined G is either PdG or nonadducted dG. To obtain the N 2 -(3-hydroxypropyl)dG-containing DNA substrate, 10 l of 1 M NaBH 4 dissolved in 1 M Hepes buffer (pH 7.4) were added twice to 200 l of the ␥-HOPdG-adducted 38-mer oligodeoxynucleotide (1-2 M). Each addition of the reducing agent was followed by incubation at room temperature for 4 h. DNA was then dialyzed against 10 mM Tris-HCl (pH 7.0), 1 mM EDTA overnight using Slide-A-Lyzer Dialysis Cassette (3,500 molecular weight cut off). To confirm the completeness of reduction, the polypeptide trapping technique was utilized (42) modified by A. J. Kurtz for ␥-HOPdG-containing DNAs. Briefly, probes of both ␥-HOPdGand reduced ␥-HOPdG-adducted oligodeoxynucleotides (50 nM) were incubated with 50 mM lysine-tryptophan-lysine-lysine in the presence of 25 mM NaCNBH 3 and 100 mM Hepes (pH 7.4) for 5 h. The reactions were terminated by the addition of an equal volume of 95% (v/v) formamide, 20 mM EDTA, 0.02% (w/v) xylene cyanol, and 0.02% (w/v) bromphenol blue and heating at 90°C for 2 min. Next, DNAs were resolved through a 15% denaturing PAGE and visualized with PhosphorImager Screen. Under these conditions, no trapping was detected in reactions with ␥-HOPdGcontaining oligodeoxynucleotide, whereas the ␥-HOPdG-containing DNA was completely complexed with the polypeptide.
Pol Purification-Purifications of yeast pol and human pol were done as described in Refs. 27 and 31, respectively.
Primer extension and single-nucleotide incorporation experiments with yeast pol were carried out as described (27) and with human pol as in Ref. 31. Briefly, the reaction mixture (10 l) contained 5 nM primer annealed to a template, 25 mM Tris-HCl buffer (pH 7.5), 10 mM NaCl, 5 mM MgCl 2 , 10% glycerol, 100 g/ml of bovine serum albumin, 5 mM dithiothreitol, 100 M of each of the four dNTPs (primer extension experiments), or 10 M individually (single-nucleotide incorporation experiments), and yeast or human pol at the concentrations as indicated in the figure legends. The reactions were incubated at 22°C and terminated by the addition of 4ϫ excess of stop solution consisting of 95% (v/v) formamide, 20 mM EDTA, 0.02% (w/v) xylene cyanol, and 0.02% (w/v) bromphenol blue. The reaction products were resolved through a 20% denaturing PAGE and visualized by a PhosphorImager screen.
Steady-state Kinetic Analysis-Steady-state kinetic assays were carried out under the same conditions as the DNA polymerase assays except that 1 nM yeast or human pol and 20 nM DNA substrates were used with various concentrations of one of the four nucleotides. The reactions were quenched after 5 min. Quantitative analyses were performed using a PhosphorImager screen and Image-Quant 5.0 software (Molecular Dynamics, Sunnyvale, CA). Calculations of rates of nucleotide incorporation were done as described in Ref. 43. The rates of nucleotide incorporation were graphed as a function of nucleotide concentration, and the k cat and K m parameters were obtained from the best fit of the data to the Michaelis-Menten equation.
Construction of Circular Single-stranded pMS2 DNA Modified with ␥-HOPdG-The 12-mer oligodeoxynucleotides containing either ␥-HOPdG or a nondamaged dG were phosphorylated at the 5Ј end with ATP and inserted into single-stranded pMS2 shuttle vector as described earlier (24). The two ligated samples were designated pMS2(dG) and pMS2(␥-HOPdG).
Mutagenesis Experiments-Transfection of pMS2(dG) and pMS2(␥-HOPdG) into cTAG and SV80 cells, isolation of DNA, amplification in E. coli DH10B cells, and differential hybridization analysis were done as previously described (24). Hybridization with the progeny plasmid DNA was performed using [␥-32 P]ATP-labeled 18-mer oligodeoxynucleotide probes (5Ј-GATGCTAGCNAGTCCATC-3Ј, where N refers to A, T, G, or C). Whatman 541 filters containing hybridized colonies were exposed to X-Omat AR film overnight, and autoradiographs were developed to identify mutation frequency and types of mutations. Representative colonies were subjected to dideoxy sequencing (44) to confirm the presence of the mutations. A 20-mer primer (5Ј-CCATCTTGT-TCAATCATGCG-3Ј) sequence around 100 nucleotides downstream of the adduct was used for sequencing the region containing the 12-mer oligodeoxynucleotide in progeny plasmid DNA.

RESULTS
In Vitro Lesion Bypass with Yeast DNA Polymerase -To examine whether yeast pol was able to replicate past a ␥-HOPdG adduct, running start primer extension experiments were performed ( Fig. 2A). A 21-mer primer was annealed to the template DNA so that it allowed the addition of 9 nucleotides before encountering the adduct (Ϫ9 primer). On the nondamaged DNA substrate, primers were efficiently extended by yeast pol (Fig. 2A, lanes 1-4). On the ␥-HOPdG-containing substrate ( Fig. 2A, lanes 5-8), yeast pol appeared to be capable of bypassing the lesion and forming full-length products. However, DNA synthesis was partially inhibited right before the DNA lesion and opposite from it.
To understand better the importance of ring opening during replication, primer extension experiments were carried out using two model DNA substrates: the PdG adduct, which is an analogue of the ring closed form of the ␥-HOPdG, and the reduced ␥-HOPdG, which is similar to the ring open form of the natural adduct. In the case of the 50-mer PdG-containing substrate, 21-mer primer was positioned on the template so that the incorporation of 15 nucleotides was needed before reaching the lesion (Ϫ15 primer). Because both efficiency and accuracy of the DNA synthesis are known to be sequence-dependent (43,45), an additional nondamaged control 50-mer DNA template was utilized that had the same sequence as the PdG-adducted template. These data revealed that the PdG adduct was a much stronger block for replication by yeast pol than ␥-HOPdG. Under conditions that allowed an efficient replication of the nondamaged DNA template ( Fig. 2A, lanes 13-16), DNA synthesis on the PdG-adducted template was greatly inhibited one nucleotide before the lesion, and synthesis was completely aborted after incorporating a nucleotide opposite the lesion ( Fig. 2A, lanes 17-20). However, replication by yeast pol beyond the PdG can be achieved but at much higher concentrations of enzyme (data not shown). With the reduced ␥-HOPdG-adducted template ( Fig. 2A, lanes 9 -12), the bypass efficiency by yeast pol seemed to be comparable with that on the ␥-HOPdG-adducted template.
The specificity of nucleotide incorporation by yeast pol opposite and beyond the lesions was also tested. To identify the nucleotide that is incorporated by this polymerase opposite the adducted base, single-nucleotide incorporation experiments were carried out using standing start DNA substrates in which 3Ј terminus of the primer was located one nucleotide before the lesions (Ϫ1 primers) (Fig. 2B). On both nondamaged substrates, yeast pol preferentially incorporated a C opposite G (Fig. 2B, lanes 3 and 18). Incorporation of a T and to a lesser extent an A and a G was also observed, especially on the 38-mer template. Interestingly, incorporation of a correct nucleotide (C) was predominant opposite each of the modified bases, namely the ␥-HOPdG (Fig. 2B, lane 8), the reduced ␥-HOPdG (Fig. 2 B, lane 13), and the PdG (Fig. 2B, lane 23) adducts.
To test whether any misincorporation occurred past the le- sion site, single-nucleotide incorporation experiments were carried out using DNA substrates in which the correct nucleotide (C) was primed with the adducted base (0 primers). No nucleotide misincorporation was observed on any of the adducted templates examined (data not shown).
Thus, yeast pol is capable of bypassing the ␥-HOPdG adduct, and in contrast to all other polymerases tested so far (22)(23)(24), it predominantly incorporates the correct nucleotide opposite and downstream of the lesion. In addition, these data show that a cyclic PdG is a much stronger block for replication by yeast pol than an acyclic reduced ␥-HOPdG, but neither of the model adducts seem to be particularly miscoding for this polymerase.
In Vitro Lesion Bypass with Human DNA Polymerase -Primer extension reactions and single-nucleotide incorporation experiments were carried out with human pol (Fig. 3) using the same set of the primer/templates as with the yeast enzyme. Similar to the yeast pol , human polymerase was able to replicate past the ␥-HOPdG (Fig. 3A, lanes 5-8) and the reduced ␥-HOPdG lesions (Fig. 3A, lanes 9-12). However, unlike yeast pol , at higher enzyme concentrations human pol appeared to bypass the PdG adduct (Fig. 3A, lanes 17-20).
Single-nucleotide incorporation experiments with human pol revealed significant differences between the human and yeast enzymes in their discrimination abilities during nucleotide insertion opposite the ␥-HOPdG adduct. Whereas yeast pol preferentially incorporated the correct nucleotide (C) opposite the lesion, human polymerase extended the Ϫ1 primer almost equally well in the presence of A, C, and G (Fig. 3B,  lanes 6 -10). On the PdG-adducted template, the difference between these two polymerases was even more striking. In contrast to the yeast pol that incorporated a C opposite PdG, human polymerase inserted A, G, and T better than the correct nucleotide (Fig. 3B, lanes 21-25). Interestingly, incorporation by human pol is much more accurate opposite the reduced ␥-HOPdG adduct (Fig. 3B, lanes 11-15) than opposite the nonreduced adduct (Fig. 3B, lanes 6 -10).
Single-nucleotide incorporation experiments were carried also out using 0 primers with the C primed with the adducted base. Yielding data similar to that of the yeast pol , human polymerase preferentially incorporated the correct nucleotide on all five substrates tested (data not shown).
Efficiency of Nucleotide Incorporation and Extension-To compare the efficiency of translesion synthesis by yeast and human pol , steady-state kinetic parameters k cat and K m were first determined for the correct nucleotide (C) incorporation opposite dG in two different sequence contexts and also opposite ␥-HOPdG, reduced ␥-HOPdG, and PdG adducts. The reactions were performed using the same 21-mer Ϫ1 primers as in the single-nucleotide incorporation experiments. For yeast pol , C is incorporated opposite the ring closed PdG adduct with a 1600-fold lower efficiency (k cat /K m ) than C is incorporated opposite the unadducted dG (Table I)  a For nucleotide incorporation opposite a given adduct, the fold reduction in efficiency was calculated as (k cat /K m ) G /(k cat /K m ) adduct . Similarly, the fold reduction in the efficiencies for extention were calculated as (k cat /K m ) normal primer terminal base paired /(k cat /K m ) adducted primer terminal base paired .
b Fidelity of incorporation was calculated as (k cat /K m ) incorrect /(k cat /K m ) correct . c ND, not determined.
incorporates a C opposite the ring open reduced ␥-HOPdG with only a 12-fold lower efficiency than opposite the unadducted dG. The efficiency of incorporation opposite the ␥-HOPdG adduct is in between these two extremes with a 190-fold reduction relative to the unadducted dG. The same trends were also observed with human pol (Table II). Next, the steady-state kinetic parameters were determined for the extension from a C residue paired with the modified bases and were used to determine the efficiency of extending from each adduct relative to the extension from an unadducted dG (Tables I and II). For both yeast and human pol , the efficiencies of extensions from the ␥-HOPdG and the reduced ␥-HOPdG were reduced ϳ5-30-fold relative to the unadducted dG. In contrast, the extension from the PdG was blocked to a much greater extent, especially in the case of the yeast enzyme (6800-fold; Table I).

Fidelity of Nucleotide Incorporation by Yeast and Human Pol
Opposite ␥-HOPdG-In the single-nucleotide incorporation experiments, yeast and human pol displayed different accuracies of replication across the ␥-HOPdG adduct. To further evaluate the accuracy of nucleotide incorporation opposite the lesion, kinetic analyses were carried out using Ϫ1 primer, and the frequencies of misincorporation were calculated as the ratio of k cat /K m of the incorrect nucleotide to the correct nucleotide (43). These data showed that yeast pol synthesizes past ␥-HOPdG relatively accurately with efficiency of incorporation of a C ϳ75 times higher than that of the next most preferred nucleotide (G) ( Table I). In contrast, human pol discriminated poorly between the correct and wrong nucleotides incorporating opposite ␥-HOPdG. Particularly, high misincorporation fre-quencies were observed for A and G (Table II).
Mutagenicity of ␥-HOPdG-modified Single-stranded pMS2 Vectors in Normal Human Fibroblasts and XPV Cells- Table  III shows the outcomes of in vivo replication of pMS2 (dG) and pMS2 (␥-HOPdG) in SV80 and XPV cells. The data presented for XPV cells were obtained from five independent experiments. All of the 1104 E. coli transformants resulting from replication of modified pMS2 (␥-HOPdG) in XPV cells were picked and grown in 96-well plates. Hybridization analysis revealed that 767 colonies hybridized with either one of the four probes, whereas 337 colonies did not hybridize with any of the four probes. Of those transformants that did not hybridize with any sequence-specific probe, none of those hybridized to sequences immediately upstream of the oligodeoxynucleotide ligation site, suggesting that this deletion was not caused by the adduct. Although 96% of the hybridized transformants did not contain any targeted mutations (Table III), 1.3% (10/767) were G to A transitions, 0.5% (4/767) were G to C transversions, and 2.1% (16/767) were G to T transversions. Sequencing of plasmid DNA prepared from these colonies confirmed the presence of T, C, or A, respectively, opposite the site of the adducted guanine. No mutations were observed when 192 colonies were screened out of transformants obtained from nonadducted pMS2(dG).
When these experiments were repeated in SV80 normal human fibroblasts, all of the 288 transformed colonies subsequently obtained from two experiments were analyzed for mutations by differential hybridization strategy. Although only 92 colonies hybridized with either one of the four probes, 89% (82/92) contained the correct base opposite the adducted guanine,  Bypass of ␥-HOPdG by Polymerase whereas 8.6% (8/92) were G to T transversions, and 1.1% (1/92) were G to C and G to A mutations. None of the colonies from the control pMS2(dG) transformants showed any mutation. Thus, XPV cells appeared to have a lower mutation frequency (3.9%) when compared with normal human fibroblast cells (11%).

DISCUSSION
The ␥-HOPdG adduct was not a significant block for replication when site-specifically modified vectors were propagated in E. coli (22)(23)(24) or in mammalian cells (16,24). In E. coli, the adduct appeared not to be miscoding (22)(23)(24). Depending on the cell type and vector used, 93-100% of the translesion events were nonmutagenic during in vivo replication in mammalian cells (16,24). Thus, in both prokaryotic and eukaryotic systems, DNA polymerases exist that are able to synthesize past ␥-HOPdG efficiently and in a predominantly error-free manner. On the other hand, none of the polymerases examined in vitro so far, namely, Klenow exo Ϫ fragment of E. coli pol I (22,23), calf thymus pol ␦ (24), and human pol ⑀ (24), were able to incorporate the correct nucleotide opposite this adduct. In the present study, yeast pol has been identified as the first polymerase that possesses an ability to replicate across the ␥-HOPdG adduct relatively accurately. Comparable efficiency of DNA syntheses past ␥-HOPdG was also observed for human pol , but this polymerase displayed a much higher propensity for misincorporation. Single-nucleotide experiments as well as steady-state analyses showed that human pol frequently incorporates an A or a G opposite ␥-HOPdG and therefore is likely to introduce G to T and G to C transversions.
We note that the observed k cat for C incorporation opposite the undamaged G template (ϳ5 min Ϫ1 ; Table I) is slower than the rate of nucleotide incorporation measured during processive synthesis (ϳ80 min Ϫ1 ; Ref. 46) for yeast pol . This suggests that k cat reflects the rate of DNA release and thus is an underestimate of the actual rate of nucleotide incorporation. Nevertheless, because the observed K m is expected to be decreased with the k cat in a compensatory manner, the efficiencies of nucleotide incorporation (k cat /K m ) determined under steady-state conditions provide a measure of catalytic efficiencies of the enzyme. More detailed kinetic studies are needed, however, to more accurately define the mechanisms controlling the fidelity of pol opposite these DNA adducts.
The nucleotide incorporation data for pol are in agreement with results of the in vivo replication assays when site-specifically modified single-stranded pMS2 vector was propagated in XPV cells. Overall mutagenic frequency determined in the XPV cells (3.9 ϫ 10 Ϫ2 /translesion synthesis) was about two and three times less than that in COS-7 (24) and normal human cells, respectively. Importantly, lower frequencies of transversions (particularly G to T) in XPV cells, but not G to A transitions, accounted for the observed differential between two types of cells. Thus, pol might potentially contribute to the bypass of the ␥-HOPdG adduct in mammalian cells being responsible for both error-free and error-prone replicative events.
Based on the NMR spectroscopy data, a model of error-free bypass of ␥-HOPdG has been proposed in which the incoming dCTP triggers a structural rearrangement of the adduct from the ring closed to the ring open form. This change allows the formation of the standard Watson-Crick hydrogen bonds, stabilizes the structure, and facilitates the subsequent extension reaction (17). To examine the role of ring opening during replication by pol , we compared the efficiency of incorporation opposite ␥-HOPdG to the incorporation opposite the two model adducts: PdG and reduced ␥-HOPdG. For both yeast and human pol , cyclic PdG was a very strong block for the incorporation of a C relative to the acyclic reduced ␥-HOPdG. For incorporation opposite ␥-HOPdG, both polymerases had an in-termediate incorporation efficiency. Ring opening was also important for the extension from a C paired with the adduct. For both yeast and human pol , relative efficiencies of extension were similar when ␥-HOPdGand reduced ␥-HOPdG-modified DNA substrates were used. By contrast, the cyclic PdG adduct is a very strong block for extension by these polymerases, especially for the yeast enzyme. Overall, these data are consistent with the proposed model of de los Santos (17), such that ring opening of ␥-HOPdG is essential not only for efficient incorporation opposite the lesion by yeast and human pol but also for efficient extension. However, from these data it cannot be concluded whether the incoming nucleotide causes the transformation of the adduct from the ring closed to the ring open form or whether the equilibrium is shifted toward ring open conformation by protein-DNA interactions in the polymerase active site.
The steady-state kinetic analyses and single-nucleotide incorporation experiments have revealed significant differences between yeast and human pol with respect to their accuracies of replication across modified bases. For the human enzyme, frequencies of misincorporation opposite ␥-HOPdG were on average, 1 order of magnitude higher than for the yeast enzyme. In addition, the incorporation by human pol opposite PdG was extremely error-prone, whereas yeast pol inserted the correct nucleotide preferentially.
The proficient ability of yeast and human pol to replicate across the ring open form of ␥-HOPdG strongly indicates that in spite of the fact that it is located in the minor groove, the presence of this adduct on the templating residue poses no significant hindrance to these polymerases. This suggests the lack of any specific contact of these enzymes with the minor groove of the templating residue, which would permit pol to replicate across DNA adducts, which protrude into the minor groove.
Although DNA synthesis past ␥-HOPdG by pol is very efficient when the adduct exists in its ring open form, in vivo replication data (16, this report) clearly show that pol is not solely responsible for bypass of this lesion in humans. Thus, another polymerase is likely involved in translesion synthesis across ␥-HOPdG. The yet unidentified polymerase may be able to efficiently bypass the ring closed form of ␥-HOPdG and perhaps other exocyclic dG adducts (1,2), in which N1 modification prevents Watson-Crick pairing.