Translesion DNA Synthesis Catalyzed by Human Pol h and Pol k across 1, N 6 -Ethenodeoxyadenosine*

1, N 6 -Ethenodeoxyadenosine, a DNA adduct generated by exogenous and endogenous sources, severely blocks DNA synthesis and induces miscoding events in human cells. To probe the mechanism for in vivo translesion DNA synthesis across this adduct, in vitro primer extension studies were conducted using newly identified human DNA polymerases (pol) h and k , which have been shown to catalyze translesion DNA synthesis past several DNA lesions. Steady-state kinetic analyses and analysis of translesion products have revealed that the synthesis is > 100-fold more efficient with pol h than with pol k and that both error-free and error-prone syn-theses are observed with these enzymes. The miscoding events include both base substitution and frameshift mutations. These results suggest that both polymerases, particularly pol h , may contribute to the translesion DNA synthesis events observed for 1, N 6 -ethenodeoxya-denosine in human cells. cartridge (Waters). Purified oligonucleotide primers were labeled at the 5 9 end with [ g - 32 P]ATP and T4 polynucleotide kinase. Primers were annealed to templates by mixing at a 1:1.2 molar ratio in 10 m M Tris-HCl (pH 7.5), 1 m M EDTA, and 100 m M NaCl by heating to 80 °C followed by slow cooling. For primer extension and standing start kinetic studies (25) of nucleotide insertion and extension, the 32 P-labeled primers (5 9 -GTTCTAGCGTG-TAGGT, 5 9 -GTTCTAGCGTGTAGGTAT, and 5 9 -GTTCTAGCGTGTAG- GTATN (where N 5 A, C, G, or T)) were annealed to a 28-mer template (5 9 -CTGCTCCTCXATACCTACACGCTAGAAC (where X 5 dA or e dA)), generating substrates 1, 2, and 3, respectively. sequence context used in the template of substrates 1, 2, 3 identical to used in miscoding studies in human cells (22). It not possible to separate various TLS products by the method described above when this sequence context was employed. Therefore, we used the sequence (substrate 4) that has been shown to permit separation of various TLS products by gel electrophoresis (27).

In the last few years, several new human DNA polymerases (pols), 1 which are likely to be involved in translesion DNA synthesis (TLS), were discovered. This list includes pol (1), pol (2,3), pol (4), and pol (5). Pol , pol , and pol are encoded by the hRAD30A, hRAD30B, and DINB1 genes, respectively. These new pols form a Rad30/UmuC/DinB/REV1 superfamily (2,3,6). Pol consists of two gene products: hREV3 containing pol activity and hREV7 with an unknown function (5). In general, these pols catalyze TLS more efficiently than previously known pols. They synthesize DNA in a distributive manner and tend to show lower replication fidelity than other pols such as pol ␣, pol ␤, and pol ␦ when unmodified DNA is used as a template (7)(8)(9)(10).
There are several pieces of evidence for the involvement of human pol and pol in TLS in vivo (5,(11)(12)(13). Although the involvement of human pol or pol has not yet been established in vivo, the Escherichia coli homologue of pol , pol IV, has been shown to play a role in TLS in vivo (14). Pol , which is missing in xeroderma pigmentosum variant cells (13,15), is shown to catalyze efficient TLS across the cis-syn cyclobutane thymine-thymine dimer by inserting two dAMPs opposite the lesion (16). Therefore, the cancer proneness of xeroderma pigmentosum variant patients is thought to be caused by the lack of accurate TLS across this and/or other UV photo products. Pol also catalyzes TLS across other DNA lesions such as cisplatin G-G intrastrand cross-link (16), acetylaminofluorene-dG (16), and 8-oxodeoxyguanosine (17) with relatively high fidelity. On the other hand, TLS across (ϩ)-trans-anti-benzo[a]pyrene-N 2 -dG is reported to be error-prone (18). Pol is also shown to conduct TLS across an abasic site (19,20), acetylaminofluorene-dG (19,20), (Ϫ)-trans-anti-benzo[a]pyrene-N 2 -dG (20), and 8-oxodeoxyguanosine (20).
Based on these findings, we are motivated to study the efficiency and fidelity of TLS catalyzed by these novel pols across 1,N 6 -ethenodeoxyadenosine (⑀dA) to probe the in vivo TLS mechanism. We have shown that ⑀dA is miscoding in simian and human cells by inducing ⑀dA3 T, ⑀dA3 G, and ⑀dA3 C (21,22). This adduct is produced in animals exposed to vinyl compounds such as the human carcinogen vinyl chloride. Surprisingly, this adduct is also found in unexposed animals and humans with lipid peroxidation products being the suspected source of this adduct (23). Our in vitro primer extension studies indicate that pol catalyzes TLS more efficiently than pol and that both pols catalyze error-free and error-prone TLS.

MATERIALS AND METHODS
Materials-[␥-32 P]ATP was purchased from Amersham Pharmacia Biotech. Human pol and pol were purified as described (13,19). Pol ␦ was purified to apparent homogeneity from calf thymus (24). Human proliferating cell nuclear antigen (PCNA) was a generous gift from Paul Fisher (State University of New York, Stony Brook, NY). T4 polynucleotide kinase and EcoRI were purchased from New England Biolabs. Ultrapure deoxyribonucleic acid triphosphates were purchased from Roche Molecular Biochemicals.
DNA Substrates-Oligonucleotides were purchased from Oligos Etc. (Wilsonville, OR) or synthesized in the laboratory of Francis Johnson (State University of New York, Stony Brook, NY). Oligomers were purified by electrophoresis on a 20% polyacrylamide gel containing 7 M urea, detected by UV shadowing, excised from the gel, eluted from gel slices, and desalted using a SEP-PAK C18 cartridge (Waters). Purified oligonucleotide primers were labeled at the 5Ј end with [␥-32 P]ATP and T4 polynucleotide kinase. Primers were annealed to templates by mixing at a 1:1.2 molar ratio in 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 100 mM NaCl by heating to 80°C followed by slow cooling. For primer extension and standing start kinetic studies (25) of nucleotide insertion and extension, the 32 P-labeled primers (5Ј-GTTCTAGCGTG-TAGGT, 5Ј-GTTCTAGCGTGTAGGTAT, and 5Ј-GTTCTAGCGTGTAG-GTATN (where N ϭ A, C, G, or T)) were annealed to a 28-mer template (5Ј-CTGCTCCTCXATACCTACACGCTAGAAC (where X ϭ dA or ⑀dA)), generating substrates 1, 2, and 3, respectively.
Kinetic Studies of Nucleotide Insertion and Extension-Standingstart reactions (25) (10 l) contained 40 nM substrate 2 or 3 (substrate 2 for insertion analysis and substrate 3 for extension analysis), 0 -2 mM dNTP(s), and a reaction buffer (see above). Initiation and termination of reactions were conducted as described above. Aliquots (1 l) were subjected to electrophoresis in a denaturing 20% polyacrylamide gel.
Data Analysis-Integrated gel band intensities were measured using a PhosphorImager and ImageQuant software (Molecular Dynamics). Nucleotide incorporation parameters were determined (25). Less than 20% of the primers were extended in these steady-state kinetic analyses, ensuring single-hit kinetics (26). Values for the Michaelis-Menten constant (K m ) and V max for incorporation opposite dA and ⑀dA were obtained by least squares nonlinear regression to a rectangular hyperbola. k cat was calculated by dividing V max by the enzyme concentration. The frequency of insertion (F ins ) and extension (F ext ) were calculated using the equation F ins or ext ϭ (k cat /K m ) adduct /(k cat /K m ) control (25). Standard errors derived from the curve-fitting are included.
Analysis of TLS Products-DNA synthesis reaction mixtures (10 l) contained 50 nM substrate 4, 100 M dNTPs, the appropriate buffer (see above), and enzyme (1.5 units of pol ␦, 36 fmol of pol , or 56 fmol pol ) were incubated at 23 Ϯ 1°C for 15 min and then 37 Ϯ 1°C for 45 min (27). Reactions were stopped by adding 10 l of a formamide dye mixture and heating to 95°C for 5 min. Samples were subjected to electrophoresis in a denaturing 20% polyacrylamide gel (35 ϫ 42 ϫ 0.04 cm). Full-length products were extracted from the gel and annealed to a complementary 38-mer. The annealed products were digested with EcoRI (100 units) for 1 h at 30°C and then 1 h at 15°C. This digestion generates 32 P-labeled 18-mers from the fully extended products. The products were separated in a two-phase polyacrylamide gel (15 ϫ 72 ϫ 0.04 cm) (27). This method allows the separation of four base substitution products and frameshift products.
The DNA template of substrate 4 is different from the template of substrates 1, 2, and 3 in the DNA sequence surrounding ⑀dA. The  1 and 7) containing dA, dC, dG, or dT opposite the lesion and with those of standards containing one-(⌬ 1 ) and two-base (⌬ 2 ) deletions. sequence context used in the template of substrates 1, 2, and 3 is identical to that used in miscoding studies in human cells (22). It was not possible to separate various TLS products by the method described above when this sequence context was employed. Therefore, we used the sequence (substrate 4) that has been shown to permit separation of various TLS products by gel electrophoresis (27).

DNA Polymerase Activity on Control and ⑀dA-Modified
Templates-Pol , pol , and pol ␦/PCNA were assayed for polymerase activity on both unmodified and ⑀dA-modified templates. The primer (substrate 1; Fig. 1A) allowed the addition of two nucleotides before encountering the adduct. Although all three pols were capable of synthesizing across ⑀dA (Fig. 1, B and C), this lesion posed a much stronger block to pol ␦ than to pol and pol when compared with the control templates. A very small amount of the full-length product was generated by pol ␦ only when PCNA was added to the reaction mixture (compare lanes 10 and 11 with lane 12 in Fig. 1B), revealing the enhancing role for PCNA in TLS. Pol seems to catalyze TLS more efficiently than pol .
Kinetic Studies of Nucleotide Incorporation and Extension-To determine the efficiency and fidelity of TLS catalyzed by pol and pol , we first determined steady-state kinetic parameters (K m and k cat ) for nucleotide incorporation opposite dA and ⑀dA using substrate 2. The internal 13 nucleotides (5Ј-CTCCTCXATACCT) of this template are identical to those used in the miscoding studies in human cells (22). The kinetic data (F ins ) indicate that pol incorporates a nucleotide opposite ⑀dA more efficiently than pol . Pol inserts the correct dTMP opposite ⑀dA twice as efficiently as dAMP and dGMP and 13 times more efficiently than dCMP. This dTMP insertion is ϳ68 times less efficient than that opposite dA. Similarly, pol also inserts dTMP most efficiently opposite ⑀dA, followed by dGMP and then dAMP, but its efficiency is ϳ1000 times less than the incorporation opposite dA. These results indicate that dTMP, the correct nucleotide, is preferentially inserted opposite ⑀dA by both pols.
We then determined steady-state kinetic parameters for nucleotide extension from four different 3Ј termini located opposite dA or ⑀dA using substrate 3. The kinetic data (F ext ) indicate that pol extends from all the termini more efficiently than pol . Pol extends the primer with the correct dTMP terminus more efficiently than the other three termini when the modified template was used. This extension from the dTMP terminus is ϳ55 times less efficient than that from the dTMP terminus located opposite dA. In experiments using pol , the efficiency of extension from the 3Ј terminus followed the order of dAMP Ͼ dGMP Ͼ dTMP Ͼ dCMP, indicating that unlike pol , the incorrect pairings are extended better than the correct ⑀dA:T pairing.
Based on these insertion and extension kinetic parameters,

FIG. 3. Model for the induction of base substitutions and onebase deletions after the incorporation of dGMP. dGMP is inserted
opposite ⑀dA (X) (step1). This terminal dG is extended (step 6) or misaligned with dC located 5Ј to X, rendering X extrahelical (step3). Extension of the misaligned primer generates one-base deletion (step 5). Realignment (step 7) and extension (step 8) yields a base substitution mutation. Misalignment is also generated by the dNTP-stabilized misalignment mechanism (29) in which a slippage event occurs first, causing X to be extrahelical (step 2), and the incoming dGTP stabilizes this misalignment (step 4).  the relative efficiency of TLS was determined by multiplying F ins and F ext . The results indicate that pol catalyzes TLS across ⑀dA more efficiently than pol . With pol , TLS with ⑀dA:T is dominant, and its efficiency is 3.0, 4.4, and 45 times greater than TLS with ⑀dA:A, ⑀dA:G, and ⑀dA:C, respectively.
The same analysis for pol shows that the efficiency of TLS with ⑀dA:T is 2.1 and 2.7 times greater than TLS with ⑀dA:A and ⑀dA:G, respectively. These results indicate that accurate TLS is dominant but not exclusive with both pols.
Miscoding Specificity of ⑀dA-Because steady-state kinetic analysis includes only one of four dNTPs in the reaction mixture, and frameshift mutations are not detected, we determined the miscoding specificity of ⑀dA in the presence of four dNTPs. We analyzed polymerization products using substrate 4 (Fig. 2). Although fully extended products were observed with the unmodified template for all three pols, only pol and pol produced full-length products with the modified template (data not shown). One possible explanation is that the 10-mer primer is not long enough to accommodate both pol ␦ and PCNA (28), and pol ␦ alone cannot catalyze TLS as shown in Fig. 1B.
Fully extended products were digested with EcoRI and electrophoresed in a two-phase polyacrylamide gel. When the unmodified template was used, pol ␦-catalyzed products showed only one band (Fig. 2, lane 2) that co-migrated with the dT marker, indicating accurate DNA synthesis. Pol also catalyzed faithful synthesis (Fig. 2, lane 5). Pol mainly catalyzed error-free DNA synthesis, but two additional bands were also observed, co-migrating with the dG and two-base deletion standards (Fig. 2, lane 3), indicating that pol produced errors on the unmodified template. When the modified template was used, at least five and four products were observed for pol and pol , respectively (Fig. 2, lanes 4 and 6), co-migrating with the dG, dA, dT, dC, or one-base deletion markers (Fig. 2, lanes 1  and 7). These products were quantified based on the amount of radioactivity in the bands (Table II). Consistent with the results of the steady-state kinetic analysis, pol dominantly catalyzed accurate TLS with dT on the modified template. However, substantial amounts of products containing dA, dG, dC, or one-base deletion were also observed. On the other hand, pol dominantly catalyzed TLS with one-base deletion, followed by dT, dA, and dC incorporation. The results shown in Table II indicate that error-prone TLS is dominant for both pols when frameshift mutations are included and that pol catalyzes accurate TLS more frequently than pol , but still more than 50% of the TLS is error-prone. DISCUSSION Pol has been reported to catalyze TLS across several DNA lesions in a relatively error-free manner (1,16,17). Our steadystate kinetic analyses and the analysis of TLS products have revealed that TLS catalyzed by pol across ⑀dA, like the benzo[a]pyrene dG adduct (18), is significantly erroneous. Although accurate TLS with dTMP insertion opposite ⑀dA is predominant, pol also frequently catalyzed erroneous TLS, causing base substitutions and frameshift mutations (Tables I  and II and Fig. 2). Pol catalyzes TLS across several DNA lesions in relatively error-free and error-prone manners (19,20). Our steady-state kinetic analysis shows that pol preferentially incorporates dTMP, followed by dAMP, opposite ⑀dA (Table I). The product analysis experiment (Table II and Fig.  2), however, has revealed that one-base deletion events were dominant followed by dTMP insertion products, indicating that pol -catalyzed TLS is also erroneous. Neither pol nor pol can be characterized as simply error-free or error-prone polymerases.
The overall efficiency of TLS, determined by F inc ϫ F ext , for A:⑀dA and G:⑀dA is similar for both pols: 9.0 ϫ 10 Ϫ5 versus 6.1 ϫ 10 Ϫ5 for pol and 1.5 ϫ 10 Ϫ5 versus 1.2 ϫ 10 Ϫ5 for pol (Table I). However, analysis of TLS products shows that dA incorporation is preferred to dG by both pols: 19.8% dA versus 9.1% dG for pol and 18.8% dA versus 2.9% dG for pol (Table  II). It is likely that some dGMP incorporated opposite ⑀dA misaligned (Fig. 3, step 3) to generate a one-base deletion, whereas dAMP incorporation does not cause this misalignment. Accordingly, dAMP incorporation opposite ⑀dA leads to a base substitution, whereas dGMP incorporation results in both a base substitution and a one-base deletion. Another mechanism envisioned is "dNTP-stabilized misalignment," which was observed for pol ␤ in TLS across abasic sites (29). According to this mechanism, the slippage event occurs first, causing ⑀dA to be extrahelical (step 2), and the incoming dGTP stabilizes this misalignment (step 4). Continuous extension from this terminus results in a one-base deletion (step 5), whereas realignment (step 7) and extension (step 8) result in a base substitution. The hallmark of the dNTP-stabilized misalignment mechanism is the relatively low K m for dNMP insertion at the terminus opposite a DNA lesion, which suggests that the incorporation is actually opposite the base 5Ј to the lesion (29). In the insertion kinetic studies with pol , the K m value for dGMP insertion opposite ⑀dA (42.7 M) is not much different from that for dTMP insertion opposite dA (21.4 M) (Table I), which suggests that dGMP is inserted opposite dC, 5Ј to ⑀dA, and extension from this terminus results in a one-base deletion (steps 2 to 4 to 5). On the other hand, with pol the K m value for dGMP insertion opposite ⑀dA (126 M) is very different from that for dTMP insertion opposite dA (9.4 M) (Table I). This suggests that dGMP is incorporated opposite ⑀dA, followed by misalignment and subsequent extension of the primer. This is the likely mechanism for the induction of one-base deletions (the dominant event) by pol .
Our mutagenesis experiments (22) have shown that miscoding events account for 10 -20% of TLS in human cells. Although it is not possible to speculate as to what extent these pols contribute to TLS in vivo, our results suggest that if these pols are involved in TLS in vivo, then it is likely to be error-prone.
In vivo experiments using human cells lacking these pols are necessary to clarify this point.