Human DNA Polymerase κ Bypasses and Extends beyond Thymine Glycols during Translesion Synthesis in Vitro, Preferentially Incorporating Correct Nucleotides*

Human polymerase κ (polκ), the product of the human POLK (DINB1) gene, is a member of the Y superfamily of DNA polymerases that support replicative bypass of chemically modified DNA bases (Ohmori, H., Friedberg, E. C., Fuchs, R. P., Goodman, M. F., Hanaoka, F., Hinkle, D., Kunkel, T. A., Lawrence, C. W., Livneh, Z., Nohmi, T., Prakash, L., Prakash, S., Todo, T., Walker, G. C., Wang, Z., and Woodgate, R. (2001) Mol. Cell 8, 7–8; Gerlach, V. L., Aravind, L., Gotway, G., Schultz, R. A., Koonin, E. V., and Friedberg, E. C. (1999) Proc. Natl. Acad. Sci. U. S. A. 96, 11922–11927). Polκ is shown here to bypass 5,6-dihydro-5,6-dihydroxythymine (thymine glycol) generated in two different DNA substrate preparations. Polκ inserts the correct base adenine opposite thymine glycol in preference to the other three bases. Additionally, the enzyme correctly extends beyond the site of the thymine glycol lesion when presented with adenine opposite thymine glycol at the primer terminus. However, steady state kinetic analysis of nucleotides incorporated opposite thymine glycol demonstrates different misincorporation rates for guanine with each of the two DNA substrates. The two substrates differ only in the relative proportions of thymine glycol stereoisomers, suggesting that polκ distinguishes among stereoisomers and exhibits reduced discrimination between purines when incorporating a base opposite a 5R thymine glycol stereoisomer. When extending beyond the site of the lesion, the misincorporation rate of polκ for each of the three incorrect nucleotides (adenine, guanine, and thymine) is dramatically increased. Our findings suggest a role for polκ in both nonmutagenic and mutagenic bypass of oxidative damage.

Many types of base damage in DNA cause structural modifications that can result in the stalling or complete arrest of high fidelity DNA synthesis during DNA replication (3,4). However, the potential for cell death attendant on arrested DNA replication can be mitigated by a mechanism called translesion DNA synthesis (TLS) (5)(6)(7). This process effects the replicative bypass of sites of base damage, allowing high fidelity semiconservative DNA synthesis to continue. Important new insights into the biochemical mechanism of TLS have recently been gained by the discovery of a number of new DNA polymerases, all of which share the properties of limited fidelity and processivity when copying undamaged DNA, as well as a lack of 3Ј 3 5Ј proofreading exonuclease activity (1,(5)(6)(7)(8)(9). Multiple DNA polymerases of this class have been shown to support TLS of one or more types of base damage in vitro. In some instances, this role is supported by genetic or other biological evidence. Hence, a general theme is beginning to emerge that the redundancy for error-prone DNA polymerases in prokaryotic and especially in eukaryotic cells reflects a requirement for the bypass of multiple types of base damage that can arrest normal DNA replication (5). Recent structural studies on a number of these polymerases suggest that translesion synthesis is effected by a less constrictive, more solvent-accessible active site, which allows for productive interactions with a wider range of template structures, including chemically modified bases (10 -12). The increased error rates observed when copying undamaged DNA in vitro (1,8,9,13) are presumably a direct reflection of this relaxed fidelity for nucleotide incorporation.
Among the many recently discovered specialized DNA polymerases is one called DNA polymerase (pol) 1 from human cells, a highly conserved structural ortholog of a bacterial polymerase called DNA polymerase IV (2). Pol is encoded by the POLK (DINB1) gene and has a predicted molecular mass of ϳ100 kDa (2). In previous studies, pol was fused to glutathione S-transferase and expressed in insect cells (14). The purified fusion protein was shown to be a template-directed DNA polymerase with limited processivity and fidelity (15). GSTpol protein lacks detectable 3Ј 3 5Ј proofreading exonuclease activity and is not stimulated by recombinant human proliferating cell nuclear antigen (PCNA) in vitro (14). However, in the presence of the three replicative accessory factors, PCNA, replication factor C, and replication factor A, pol exhibits a 50 -200fold stimulation in efficiency but no increase in processivity (16). Additionally, pol interacts physically with PCNA (16). Human pol has optimal activity at 37°C over the pH range 6.5-7.5 and is insensitive to inhibition by aphidicolin or dideoxynucleotides or by NaCl up to 50 mM. Either Mg 2ϩ or Mn 2ϩ can satisfy a metal cofactor requirement for pol activity (14). In vitro pol extends DNA oligonucleotide primers to a position one base short of the end of the DNA template (14).
Full-length purified pol fusion protein is unable to extend a DNA primer past bulky base adducts such as thymine-thymine dimers or -pyrimidine-pyrimidone photoproducts generated by exposure of cells or DNA to UV radiation (17). Similarly, the enzyme does not support TLS past cisplatin intrastrand cross-links in template DNA (14). In contrast, the enzyme can support TLS past acetylaminofluorene-guanine, any of the four stereoisomer adducts resulting from reaction of benzo[a]pyrene-7,8-diol 9,10-epoxide at C-10 with the exocyclic N 2 of guanine (BPDE-G) in template DNA, 1,N 6 -ethenodeoxyadenosine, and sites of base loss (abasic sites) (14,(17)(18)(19). At present, the biological significance of these observations is not clear. Evidence has also been presented that pol can perform efficient extension synthesis following nucleotide incorporation directly opposite DNA lesions (18). Pol additionally supports extension synthesis on primer-template substrates terminating in a 3Ј mispaired base, incorporating nucleotides with a high error rate (20).
POLK mRNA and pol protein are highly expressed in the adrenal cortex of adult mice, beginning in early embryonic life (2). 2 Indeed, at embryonic day 15.5 this is the only tissue in which POLK expression can be detected by in vitro hybridization. 2 Furthermore, this expression pattern appears to be relatively specific, since genes encoding two other recently discovered specialized DNA polymerases, pol and pol are not uniquely or highly expressed in the adrenal cortex. 2 Steroid biosynthesis in the adrenal cortex is known to involve the generation of large amounts of reactive oxygen species, which may result in an unusual burden of oxidative DNA damage in adrenal cortical cells (23,24). 3 Consistent with a possible role in the replicative bypass of oxidative base damage to DNA, the PolK (DinB1) gene is up-regulated in mouse embryo fibroblasts following exposure of cells to either doxorubicin or UV radiation. 2 Both of these agents are known to generate reactive oxygen species that can damage DNA (25)(26)(27)(28). Furthermore, mouse embryo fibroblasts from a mutant mouse strain defective in pol activity manifest increased sensitivity to killing following exposure to UV radiation (29).
In the present studies, we have investigated the ability of human pol to support primer extension in vitro past thymine glycol (Tg) residues in DNA, a biologically important form of oxidative base damage that potently inhibits DNA replication by many high fidelity polymerases (30). 3 Recently, another Y-family DNA polymerase, human pol was reported to bypass Tg lesions in vitro (31). Pol synthesizes DNA past Tg with an efficiency nearly equal to that of undamaged DNA but with an extremely high rate of error (31). Additionally, pol exhibits a stereochemical preference for the R stereoisomer at C-5 of Tg (31). For the present studies, we employed primer-template substrates in which a single Tg residue in the template DNA strand was generated by two different methods (Fig. 1A). Both procedures result in a mixture of the four possible stereoisomers of Tg in different relative proportions (Fig. 1B) (32,33). We show by both qualitative and quantitative steady-state kinetic analysis that pol supports TLS across both of these substrates. During this replicative bypass, the base A is pref-erentially incorporated opposite Tg. Additionally, pol is able to extend the primer template beyond the lesion, preferentially incorporating the correct next base. However, differences are observed both in the efficiency of A incorporation and in the misincorporation rate of G opposite Tg when comparing the two templates. TLS appears to be more efficient and specific opposite the substrate putatively containing a larger proportion of 5S stereoisomers.

EXPERIMENTAL PROCEDURES
Biochemical Reagents-Terminal deoxytransferase and T4 DNA polymerase were obtained from Invitrogen. The Klenow fragment of E. coli DNA polymerase I (exo Ϫ ) was obtained from New England Biolabs. T4 DNA polymerase was obtained from U.S. Biochemical Corp. Deoxynucleoside triphosphates were from Promega. Osmium tetroxide (OsO 4 ) was purchased from Aldrich.
Expression and Purification of GST-Pol-GST-pol fusion protein was purified as previously described (14).
DNA Substrates-The primer used for running start experiments was P4-OX-RS (5Ј-dGAATTCCTGCAGCCCAGGAT); the primer for standing start experiments was P5-OX-SS (5Ј-dGAATTCCTGCAGC-CCAGGATCGACTGGTCC). The primer for steady state extension experiments (k ext ) was P5-OX-SS-A (5Ј-dGAATTCCTGCAGCCCAGGAT-CGACTGGTCCA). The template DNA for these experiments was the sequence 5Ј-dATTCCAGACTGTCAATAACACGGTgGGACCAGTCGA-TCCTGGGCTGCAGGAATTC); the site of thymine glycol modification is underlined. The control template was of the same sequence, except the underlined base was simply thymine. DNA oligonucleotides were purified by denaturing polyacrylamide gel electrophoresis (DPAGE). Five pmol of each primer was 5Ј-end-labeled with T4 polynucleotide kinase (Invitrogen) in the presence of [␥-32 P]ATP, and the unincorporated radiolabel was removed using a Sephadex G-25 spin column equilibrated with STE (100 mM NaCl, 10 mM Tris, pH 8.0, 1 mM Na 2 EDTA). Primers were annealed to template strands in a stoichiometric ratio of 1:1.5 (primer-template) by heating (90°C, 5 min, 1ϫ STE buffer) and cooling on the bench top (15 min). Native polyacrylamide gels showed altered gel mobility for the labeled primers under these conditions when compared with single-stranded primers, indicating duplex DNA character for the template-annealed primers under these conditions.
Thymine Glycol Template 2 (Tg OsO4 ; Higher Proportion of 5R Stereoisomers)-A DNA oligonucleotide of the sequence 5Ј-d(AACACGGTG-GACCAG) (2.4 mg) was incubated in aqueous OsO 4 (3 stoichiometric equivalents, 5% pyridine, 200 l total, 25°C, 30 min). The reaction was extracted with chloroform and ethanol-precipitated before purifying by HPLC and DPAGE. The purified thymine glycol DNA oligonucleotide was characterized by ESI mass spectrometry, which indicated the addition of 34 atomic mass units ((M ϩ H ϩ ) 1ϩ ϭ 4646) to the unreacted DNA oligonucleotide. The purified sample was further characterized by piperidine cleavage analysis of a 5Ј-32 P-end-labeled sample (ϳ1 pmol, 1 M piperidine, 100 l, 30 min, 90°C), which revealed 100% truncation of the DNA oligonucleotide at the thymine position, confirming that the site of modification was exclusively thymine and that 100% of thymines were oxidized. Additionally, the thymine glycol-modified oligonucleotide had distinct gel mobility from the unmodified strand, which aided in both purification and characterization. Background piperidine cleavage at positions corresponding to deoxycytidines, was also observed in incompletely purified samples.
The thymine glycol oligonucleotide was phosphorylated with T4 polynucleotide kinase and ATP, purified using a Sephadex G-25 spin column, and ethanol-precipitated. The resulting phosphorylated, thymine glycol oligonucleotide was ligated to flanker sequences analogously to the preparation of the Tg Br2 substrate by heating in the presence of 5Ј-dAATCCAGACTGTCAAT and 5Ј-phosphoryl-dTCGATC-CTGGGCTGCAGGAATTC and a splinting DNA with the sequence 5Ј-dAGGATCGACTGGTCCACCGTGTTATTGACAGTC. The full-length, DPAGE-purified, ligated thymine glycol substrate (TgOsO 4 ), which has the same sequence as the Tg Br2 substrate (described above), was quantitated by UV and annealed to appropriate primers for primer extension analysis.
DNA Polymerase Assays-Running start and standing start assays shown in Figs. 2-4 were performed as previously described except that gels were phosphorimaged and quantitated using Amersham Biosciences software (14). Steady-state kinetics experiments (Fig. 5) were performed according to previously published methods (35,36). Radiolabeled primer-templates (20 nM) were incubated (10 min) with GSTpol (5 nM) and the indicated concentration of a single deoxynucleotide 5Ј-triphosphate. The resulting primers were resolved by DPAGE, and bands corresponding to unextended primer were quantitated by phosphorimaging. Numerical data corresponding to these gels are shown in Table I. Data were fit to the Michaelis-Menten equation as described by a hyperbolic curve using SigmaPlot 2001. Apparent K m and V max values were calculated from the plots, and k cat values were subsequently calculated from V max . In all cases, data points resulting in greater than 25% primer turnover were not used in plots calculating the V max and K m . Steady state experiments were performed in triplicate, and results were averaged to obtain the values reported in Table I along with the corresponding standard deviations.

Preparation of Two Substrates Containing Different Relative
Amounts of Thymine Glycol Stereoisomers-We generated a single Tg lesion at a defined position in a DNA oligonucleotide template using two different procedures. One substrate (Tg Br2 ) was generated by incorporation of Tg from thymidine glycol-5Јtriphosphate prepared by bromination and oxidation as previously described (33). This procedure has been shown to yield a stereochemical mixture containing 64.3% of the two 5R stereoisomers, and 35.7% of the 5S stereoisomers (33). Following oxidation, the stereochemistry at C-5 is fixed, but epimerization occurs about the C-6 center to yield 87% cis, 13% trans for the 5R stereoisomer and 80% cis, 20% trans for the 5S stereoisomer (33). Epimerization equilibrates at room temperature within a few hours (33), yielding an equilibrium distribution of ϳ55.9% (5R,6S), 8.4% (5R,6R), 28.6% (5S,6S), and 7.1% (5S,6R) (Fig. 1B). When the thymidine glycol triphosphate used to prepare this substrate was digested to the nucleoside with alkaline phosphatase and analyzed by HPLC under conditions previously described (33), the same ratios of the two trans and cis isomers were obtained as reported in the earlier study (data not shown). 4 A second substrate of identical nucleotide se- 1. A, schematic showing the primers and template DNAs prepared for this study. The two methods of thymine oxidation are indicated as well as the position and sequence context of the lesion. B, literature values for the putative stereochemical make-up of each thymine glycol sample. Note that in the Tg OsO4 template, the 5R stereoisomers are in greater relative abundance. quence (Tg OsO4 ) was generated by direct oxidation of the single T residue in the template DNA using osmium tetroxide (OsO 4 ). With this procedure, the 5R and 5S stereoisomers are formed in a ratio of 6:1 (32), and following epimerization the relative percentages of the four stereoisomers are expected to be 74.7% (5R,6S), 11.3% (5R,6R), 11.1% (5S,6R), and 2.9 (5S,6R) (Fig.  1B). Thus, in the Tg OsO4 template, the 5R stereoisomers are expected to be in slightly greater relative abundance than in the Tg Br2 template (Fig. 1B).
TLS across Thymine Glycol by Pol-We previously reported that the GST-pol fusion protein used in the present studies does not support TLS across cisplatin adducts in DNA (14). In contrast, incubation of a primer-Tg Br2 template with purified GST-pol fusion protein revealed progressively more efficient TLS past Tg as a function of increasing enzyme concentration (Fig. 2). The high fidelity replicative enzyme DNA polymerase ␦ from calf thymus did not support bypass across the Tg lesion at the enzyme concentrations tested, although it exhibited primer extension of the undamaged control template DNA that was characteristically stimulated by the accessory factor PCNA (Fig. 2, lanes 10 -14). Like all of the replicative polymerases, pol␦ exhibits a 3Ј-5Ј exonuclease activity, which yielded faint primer degradation bands in lanes 10 -14 below the portion of the gel shown in Fig. 2 (data not shown). Consistent with the results of previous studies (37), comparable levels of the Klenow exo Ϫ form (devoid of 3Ј 3 5Ј exonuclease activity) of E. coli DNA polymerase I also bypassed Tg in this substrate (Fig. 2). Essentially identical results were observed with the primer-Tg OsO4 template (Fig. 3A). Once again, bypass was observed in the presence of Escherichia coli Klenow exo Ϫ fragment (Fig.   3B). However, comparable amounts of the high fidelity replicative DNA polymerases from phage T4 (Fig. 3B) and phage T7 (data not shown) characteristically did not support bypass of Tg. Additionally, when we performed side-by-side standing start experiments in which the two templates were compared directly, we observed essentially identical levels of bypass by GST-pol (Fig. 3C). In all reactions with or without the presence of Tg primer, extension terminated one nucleotide short of the end, a previously described intrinsic property of pol (14).
Fidelity of Nucleotide Incorporation Opposite Tg by GST-Pol-To determine the fidelity of TLS across Tg by the GSTpol fusion protein, we performed standing start primer extension reactions in the presence of each of the four individual deoxynucleoside triphosphates. As shown in Fig. 4, the correct complementary base A is qualitatively preferentially incorpo-  3. A, GST-pol exhibits identical bypass of an alternatively prepared thymine glycol template, Tg OsO4 , which is constitutionally identical to Tg Br2 (Fig. 2) experiments, but stereochemically contains a greater proportion of 5R thymine glycol stereoisomers. Radiolabeled primer-templates (5 nM) were incubated with all four deoxynucleotide triphosphates (100 M total) at the indicated quantities of each polymerase, and the resulting primer extension products were resolved by DPAGE. Lanes 1 and 4, control experiments with no enzyme added. Lanes 2 and 3, undamaged template with 1 and 5 nM GST-pol. Lanes 5 and 6, Tg OsO4 template with 1 and 5 nM GST-pol. B, the TLS pattern for Klenow fragment of E. coli Pol I (exo Ϫ ) with Tg OsO4 , is identical to that shown for Tg Br2 in Fig. 2, and T4 DNA polymerase shows characteristic replicative arrest at Tg in Tg OsO4 . Experiments were performed analogously to those in A. Lanes  rated opposite Tg. However, the incorrect bases C, T, and especially G are also misincorporated. To compare the efficiency and fidelity of nucleotide incorporation in undamaged template DNA and that of DNA containing a single Tg residue at the identical position but prepared using two different protocols, we performed experiments under steady state conditions ( Fig. 5 and Table I).
A comparison of the parameter k cat /K m for the incorporation of A opposite T and Tg shows a 20-fold reduced efficiency for the Tg Br2 template and a 50-fold reduced efficiency for the Tg OsO4 template relative to T (Table I and Fig. 6A). When copying the undamaged template, the GST-pol fusion protein misincorporates nucleotides opposite the T residue with frequencies of 2.7 ϫ 10 Ϫ3 (G), 9.6 ϫ 10 Ϫ4 (C), and 2.7 ϫ 10 Ϫ4 (T) relative to the correct nucleotide A ( Table I). The same comparison for nucleotide incorporation opposite Tg in either substrate (i.e. the frequency of misincorporation of G, C, and T relative to the correct incorporation of A) reveals ϳ1.2-3-fold reduced discrimination between the correct and incorrect nucleotide in all cases except G misincorporation opposite the Tg OsO4 template lesion (Table I and Fig. 6B). Nonetheless, the preference for A opposite Tg is 2-3 orders of magnitude greater than for any other nucleotide. After A, the base most frequently incorporated opposite either T and Tg is G.
Whereas the relative preference for incorporating G compared with the other bases opposite T in the undamaged template or Tg in Tg Br2 template is similar (Table I and Fig. 6B), this preference is increased ϳ5-20-fold in the Tg OsO4 template (Table I and Fig. 6B). This may relate to the fact that the Tg OsO4 template is expected to contain a greater proportion of the 5R stereoisomers. This stereochemically based difference presumably interferes with discrimination of the purine versus pyrimidine character of the incoming nucleotide. A myriad of models are tenable to explain this phenomenon, especially since the site of hydrogen bonding is on the opposite end of the template base. It is important to note that the relative percentage of each of the stereoisomers is not anticipated to be greatly different between the two templates tested. The 5R stereoisomers are expected to be only ϳ25% more abundant in the Tg OsO4 template. Hence, one might predict even greater stereochemical effects on TLS by pol in a stereochemically pure 5R or 5S sample. Accordingly, in vivo the 5S stereoisomers may be the preferred substrate for this polymerase.
Having established that GST-pol inserts a base opposite Tg in a largely correct fashion, we examined the fidelity of extension beyond the lesion. Recent reports have documented the ability of pol to extend mispaired primer-termini promiscuously (20). Additionally, extension of the Tg-A base pair is believed to be the arresting substrate for most of the polymerases arrested by Tg, since a number of them insert A correctly opposite Tg but are unable to incorporate the next base (38 -43). We determined steady state values for the extension of a primer terminating with correctly base-paired deoxyadenosine (P5-ox-ss-A) opposite the Tg lesion (in Tg OsO4 ), using each of the four deoxyribonucleoside triphosphates. The next correct base, C, was incorporated in clear preference to the other bases (Table I).
We also measured k ext for C incorporation using the Tg Br2 template. This returned k cat and K m values very similar to those observed with Tg OsO4 , suggesting little or no stereochem- The nucleotide incorporated is shown immediately above each gel, and the micromolar dNTP concentration incubated in each reaction is shown immediately below the corresponding lane of each gel. Experiments were performed using the undamaged template (A), the Tg Br2 template (B), or the Tg OsO4 template (C). Raw data were analyzed as described under "Experimental Procedures," and the resulting steady state kinetic parameters are reported in Table I. Note that most of the panels shown represent initial conditions with a broad dNTP range. dCTP and TTP gels in B show representative results from more narrowly focused working ranges used in subsequent experiments.  Table I. A, k cat /K m values for nucleotide insertion opposite thymine glycol in the Tg OsO4 template (black bars), the Tg Br2 template (blue bars), and the undamaged template (yellow bars). The left graph shows directly plotted k cat /K m values obtained for incorporation of the bases A, C, G, or T, emphasizing the difference in incorporation efficiency of A between the control and Tg templates. The right graph is a y axis blown up version of the same graph to emphasize that A is incorporated in great preference to the other bases in all three of the templates and that GST-pol may exhibit stereochemical preference during insertion of A opposite of thymine glycols. B, comparison of the f inc values obtained for each template, measuring the degree of preference for incorporation of the correct base. f inc is defined as (k cat /K m ) incorrect /(k cat /K m ) correct . The left graph is a direct plot of the f inc values for nucleotide insertion opposite thymine glycol in the Tg OsO4 template (black bars), the Tg Br2 template (blue bars), and the undamaged template (yellow bars), emphasizing the high degree of preference for A incorporation in each template. The graph on the right is a y axis blown up version of the same graph to emphasize that the misincorporation frequency of G increases in template Tg OsO4 , suggesting that thymine glycol stereochemistry influences the ability of GST-pol to discriminate between the purines during incorporation opposite the lesion. ical influence in the ability of pol to extend beyond the Tg lesion (Table I). Measurement of k ext revealed that the other three bases, A, G, and T were misincorporated with about the same relative efficiency, approximating the high level of G misincorporation opposite Tg in the Tg OsO4 template. However, the levels of base misincorporation observed were considerably lower than those reported for extension performed with undamaged but mispaired substrates in previous studies (20). In summary, GST-pol clearly prefers the correct base both opposite and at least 1-2 bases beyond the Tg residue in the DNA sequence context tested here.

DISCUSSION
Tg has been associated with potent blocking of DNA replication by high fidelity polymerases, in particular full-length E. coli DNA polymerase I or the Klenow exo ϩ fragment (38 -43). Our results (and those reported elsewhere (37)) indicate that at DNA/enzyme stoichiometric equivalents approximating those generally employed for in vitro TLS by Y family polymerases, the Klenow exo Ϫ fragment of E. coli polymerase I does indeed support limited bypass of Tg. This presumably derives from the absence of the 3Ј 3 5Ј proofreading exonuclease activity and from our use of a highly purified Tg template DNA devoid of background oxidation at neighboring deoxycytidine residues. Previous studies utilizing Tg templates were prepared with a vast stoichiometric excess of OsO 4 without removal of secondarily oxidized DNA (see "Experimental Procedures"). We have observed reduced Tg bypass by both Klenow exo Ϫ and GST-pol using DNA prepared in this manner (data not shown). Hence, we suggest that the replicative arrest observed in previous studies reflects, at least in part, the presence of multiple oxidative lesions rather than a single Tg residue. More relevant to the present studies, Tg is unambiguously an arresting lesion in experiments using various eukaryotic DNA polymerases (18,31).
As previously mentioned, pol is highly expressed in cells in the mouse adrenal cortex. The adrenal cortex is a site of active steroidogenesis, and the covalent adduction of estrogen metabolites with DNA has been demonstrated experimentally (44). Additionally, steroidogenesis is associated with the potential for generating reactive oxygen species (45)(46)(47). This tissuespecific expression pattern of pol suggests a role in the response to DNA damage from reactive oxygen species and/or aromatic hydrocarbons. Support for the former role derives from the observation that the mouse Dinb1 gene is up-regulated after exposure of cells to agents known to promote oxidative DNA damage and from the present studies demonstrating that GST-pol can support bypass of a well characterized form of oxidative DNA damage, Tg, incorporating A opposite the lesion in preference to the other bases. Support for a role in TLS across sites of aromatic hydrocarbon base damage derives from two observations. First, levels of POLK expression have been correlated with an upstream promoter element that is stimulated by aromatic hydrocarbons (48). Second, pol has been shown to support TLS past two different deoxyguanosine adducts with aromatic hydrocarbons, acetylaminofluoreneguanine and BPDE-G (14,17,18).
The propensity for accurate TLS of Tg by GST-pol is manifested both with respect to the nucleotide inserted directly opposite the lesion and extension for at least 1-2 bases. GSTpol inserts the correct nucleotide with a specificity of Ͼ98% and extends correctly with ϳ97% accuracy. Taken together (i.e. the multiples of these yields), GST-pol would be expected to bypass a stereochemically mixed Tg environment with greater than 92% accuracy, inserting the correct nucleotide opposite and (at least) two nucleotides beyond the lesion. This contrasts with the reported kinetic parameters for human pol, which inserts A opposite Tg almost as efficiently as it does opposite T in undamaged DNA but with a level of overall misincorporation of ϳ7% (31). The authors of this study did not report misincorporation rates for extension, but even in the unlikely case that those were modest, the overall yield would decrease with each extension event, yielding an overall rate much lower than that exhibited by GST-pol.
Interestingly, like GST-pol, pol misincorporates G with the highest frequency relative to the other bases (31). Pol synthesizes DNA more rapidly than pol, with k cat /K m values approximately an order magnitude higher, but is more errorprone when synthesizing past Tg. Most provocatively, however, pol apparently prefers the 5R Tg stereoisomers over the 5S forms. Steady state kinetics results were not reported for 5S; however, running start synthesis clearly demonstrated less robust bypass. In the present study, we observed the opposite behavior for GST-pol, which exhibits small, but reproducible differences in the efficiency of incorporation of A opposite Tg in templates differing only in stereochemical composition. The template that exhibits more efficient turnover and greater accuracy of incorporation by GST-pol may contain a greater relative abundance of the 5S stereoisomer. Whereas pol is approximately an order of magnitude more efficient than pol in vitro, the role of accessory proteins may influence the k cat /K m values reported here and elsewhere. Additionally, up-regulation of the POLK gene in tissues with DNA oxidative damageprone environments may promote higher levels of pol.
Our observation that the misincorporation rate for G by GST-pol is increased in the template with a putatively greater proportion of 5R stereoisomers may reflect important features of the pol active site. The conformation of the Tg base is calculated to be the "half-chair" in which the most abundant 5R stereoisomer, (5R,6S) has been calculated to reside largely in a conformation that places the methyl group of C-5 pseudoequatorial and the hydroxyl group of C-6 pseudoaxial (49). It has been proposed that such a structure could generate a G-T wobble base pair (49). Perhaps the Tg base preference for the half-chair "down" conformation in 5R,6S adjusts the plane of the opposite edge of the base (the hydrogen bonding surface), allowing for a G-T wobble base pair, whereas in the opposite stereoisomer (5S,6R) a half-chair "up" conformer would predominate, which may leave the hydrogen bonding edge of the FIG. 7. Schematic drawing of the most important conformers expected from the most abundant stereoisomers of thymine glycol. The 5R,6S form (left drawings, putatively in greater abundance in Tg OsO4 ) may give rise to G-T wobble base-pairing in the GST-pol active site, resulting in higher levels of misincorporation of G. The 5S,6R isomer (right drawings) may adopt a much different conformation. dR, deoxyribose. As a result of different half-chair conformations, the stereoisomers may each present the hydrogen bonding surface of the base at different angles, giving rise to the observed differences for misincorporation of G between the Tg Br2 and Tg OsO4 templates. base more closely aligned with that of a normal thymine template base (Fig. 7). Interestingly, human Rev1 polymerase, another Y family member, has been shown to be a dCMP nucleotidyl transferase that preferentially incorporates C opposite a number of DNA lesions but incorporates T next most frequently (50). Thus, while pol and pol apparently prefer incorporating purines opposite template lesions, Rev1 inserts pyrimidines preferentially. This contrast suggests that some Y family polymerases may be more specialized for purine-based lesions and others for pyrimidine-based lesions.
Importantly, GST-pol is able to extend beyond sites of Tg damage. The enzyme exhibits an increase in overall misincorporation beyond the Tg lesion, echoing previous reports on extension beyond sites of base damage. However, GST-pol manifests a significant preference for the correct base, incorporating C greater than 97% of the time immediately beyond the site of the Tg lesion. This observation is significant, because Tg lesions are typically arresting to polymerases following the point of nucleotide incorporation opposite a lesion (38 -43).
Pol has also been shown to efficiently extend mispaired bases, BPDE-G adducts, acetylaminofluorene-guanine adducts, and AP sites (18,29). These observations have prompted the notion that pol has a primary role in extension beyond sites of damage rather than in bypass per se during TLS, as suggested for pol (21). Our results with Tg suggest that pol supports both less efficient but more accurate incorporation opposite, and extension beyond, lesions that occur spontaneously in tissues with high levels of oxidative damage. Similar conclusions have been drawn for the bypass of BPDE-G adducts (18). In the final analysis, it is important to bear in mind that interpretations of the experiments reported here (and elsewhere) require independent biological and genetic confirmation.