Quantitative Analysis of Translesion DNA Synthesis across a Benzo[a]pyrene-Guanine Adduct in Mammalian Cells

Replication across unrepaired DNA lesions in mammalian cells is effected primarily by specialized, low fidelity DNA polymerases. We studied translesion DNA synthesis (TLS) across a benzo[a]pyrene-guanine (BP-G) adduct, a major mutagenic DNA lesion generated by tobacco smoke. This was done using a quantitative assay that measures TLS indirectly, by measuring the recovery of gapped plasmids transfected into cultured mammalian cells. Analysis of PolK+/+ mouse embryo fibroblasts (MEFs) showed that TLS across the BP-G adduct occurred with an efficiency of 48 ± 4%, which is an order of magnitude higher than in Escherichia coli. In PolK–/– MEFs, bypass was 16 ± 1%, suggesting that at least two-thirds of the BP-G adducts in MEFs were bypassed exclusively by polymerase κ (polκ). In contrast, polη was not required for bypass across BP-G in a human XP-V cell line. Analysis of misinsertion specificity across BP-G revealed that bypass was more error-prone in MEFs lacking polκ. Expression of polκ from a plasmid introduced into PolK–/– MEFs restored both the extent and fidelity of bypass across BP-G. Polκ was not required for bypass of a synthetic abasic site. In vitro analysis demonstrated efficient bypass across BP-G by both polκ and polη, suggesting that the biological role of polκ in TLS across BP-G is due to regulation of TLS and not due to an exclusive ability to bypass this lesion. These results indicate that BP-G is bypassed in mammalian cells with relatively high efficiency and that polκ bypasses BP-G in vivo with higher efficiency and higher accuracy than other DNA polymerases.

Genomic DNA is constantly subject to damage caused by both external agents, such as sunlight, and endogenous chemicals, such as reactive oxygen species. Most of this damage is eliminated by error-free DNA repair mechanisms, thereby restoring the DNA to its native sequence (1). However, a significant number of lesions escape repair and might therefore interfere with DNA replication and gene expression. Such in-terference can be mitigated by DNA damage tolerance mechanisms, primarily translesion DNA synthesis (TLS 1 ; also termed translesion replication) (2)(3)(4)(5) and postreplicative recombinational repair (1, 6 -8). The key components in TLS are low fidelity DNA polymerases that specialize in lesion bypass (9 -12). These proteins were conserved in evolution and are present in organisms ranging from Escherichia coli to humans (13). Humans contain at least four specialized DNA polymerases belonging to the Y superfamily (pol, pol, pol, and REV1) as well as several from other polymerase families (e.g. pol (14), pol (15,16), and pol (16,17)). Many of these polymerases have been implicated in TLS in vitro (18 -24). However, there is a paucity of information about the efficiency and fidelity with which they support lesion bypass in living cells.
Pol has a well established biological role in TLS, since it is mutated in all patients examined with the variant form of the hereditary disease xeroderma pigmentosum (10,18). This disease is characterized by sensitivity to sunlight and a marked predisposition to skin cancer. Correspondingly, cells from xeroderma pigmentosum patients are sensitive to and hypermutable by UV light (1). Purified human pol replicates across TT cyclobutyl pyrimidine dimers or normal TT sequences with similar efficiency and accuracy (19,20). These observations suggest that in cells in which pol is inactivated, the bypass of these lesions is effected by other DNA polymerases with reduced efficiency (resulting in increased UV sensitivity) and reduced fidelity (resulting in hypermutability and skin cancer). Evidence for a function in TLS in vivo has also been provided for pol and hREV1. In these cases, decreased expression of the relevant genes with antisense RNA led to reduced UV mutability (14,25).
In vivo experiments showed that fibroblasts in culture, derived from PolK Ϫ/Ϫ mouse embryos, are sensitive to both killing and mutagenesis by BP, relative to otherwise isogenic cells from PolK ϩ/ϩ mouse embryos, suggesting that pol is involved in bypass across BP-G adducts in vivo (26). In vitro experiments have shown that purified pol efficiently bypasses BP-G adducts, preferentially incorporating the correct nucleotide dCMP (21)(22)(23)27). In this study, we used a quantitative indirect TLS assay to specifically analyze bypass across a sitespecific BP-G adduct in mammalian cells. We found that 1) BP-G is bypassed in mammalian cells with a relatively high efficiency of 35-50%, an order of magnitude higher than in E. coli, 2) pol is responsible for at least two-thirds of lesion bypass events, and 3) bypass by pol occurs with lower mutagenicity than bypass by other DNA polymerases.

EXPERIMENTAL PROCEDURES
Materials-Dulbecco's phosphate-buffered saline without calcium chloride and magnesium chloride; minimum essential medium Eagle with Earle's salts and essential and nonessential amino acids; 100 mM glutamine; and fetal calf serum were from Sigma. Dulbecco's modified Eagle's medium and a mixture of penicillin and streptomycin for cell culture were from Invitrogen. JetPEI was from Polyplus-transfection (Illkirch, France).
DNA-DNA oligonucleotides without a lesion were supplied by Sigma-Genosys. Oligonucleotides containing synthetic abasic sites (dSpacer; Sterlin, VA) were synthesized by the Synthesis Unit of the Biological Services Department in our institute or were purchased from Metabion (Martinsried, Germany). Site-specifically modified 12-mer oligonucleotides containing (ϩ)-trans-BPDE-N 2 -dG adduct were generated as previously described (28). The BPDE-modified 54-mer oligonucleotides used to construct the gap-lesion plasmids GP-BPG1 and GP-BPG2 were the products of ligating the BPG1 and BPG2 12-mer oligonucleotides to two 21-mer oligonucleotides, using a 34-mer as a scaffold (Fig. 1A). The resulting 54-mers were separated from the scaffold and excess 21-mers on a 12% denaturing polyacrylamide gel (containing 8 M urea). The construction of the gap-lesion plasmids GP21 (with a synthetic abasic site) and GP20 (without a lesion; Fig. 1B) was previously described (29,30). The gap-lesion plasmids GP-BPG1 and GP-BPG2 were constructed similarly to GP21, except that the insert oligonucleotides contained the BPDE-modified 54-mer templates (Fig.  1B). The plasmid GP20-cm is a chloramphenicol-resistant derivative of GP20. This plasmid was prepared as GP20, except that the vector plasmid used was pSKSL-cm. Plasmid pSKSL-cm is the ligation product of the HindIII-XhoI fragment (2845 bp long) from plasmid pSKSL and a 773-bp fragment containing the chloramphenicol resistance gene from plasmid pACYC184 and carrying PCR-generated termini with XhoI and HindIII sites. The cDNA of the human DINB1 gene, encoding pol, was obtained by reverse transcription-PCR from HeLa cell total mRNA as three overlapping fragments using the following primers: for the 5Ј portion of the DINB1 gene, primer 361 (5Ј-CGGATAAGTTTAT-ACCATGGATAG-3Ј) and primer 362 (5Ј-GGCAATGCCTGCACTGGC-TGTC-3Ј); for the middle portion of DINB1, primer 363 (5Ј-CTCAGTT-GTTTTTGGAACATCAG-3Ј) and primer 364 (5Ј-AGAAACTCTTCTTA-TGAGACA-3Ј); for the 3Ј portion of DINB1, primer 365 (5Ј-CCACTGA-GTGTACATTAGAGAA-3Ј) and primer 366 (5Ј-TTAATGATAAAATGT-TCAATGTTTAC-3Ј). The three amplified fragments were cleaved with restriction nucleases PflMI, XbaI plus PflMI, and XbaI, respectively, and then ligated. Finally, the full cDNA was amplified from the ligation mixture using primers 361 and 366 and cloned into the EcoRV site of plasmid pACYC184 to yield plasmid pACYC184-pol. The sequence of the entire cloned DINB1 gene was then determined. The pol expression vector, pC-pol, was generated by PCR using pACYC184-pol as a template, and the primers 5Ј-TTAGGATCCGGATAGCACAAGGAGAA-GTGT-3Ј and 5Ј-GTTCAATGTTTACTTAAACTCGAGATCAAGGGTA-TGTTTGGG-3Ј. The PCR product was digested with BamHI and XhoI and inserted into BamHI-and XhoI-cleaved pCDNA3 downstream to the cytomegalovirus immediate early promoter. The sequence of pol open reading frame was confirmed by DNA sequencing.
Cell Cultures-Cells from Pol-deficient and wild-type mouse embryonic fibroblasts were previously described (31). The SV40-transformed human fibroblasts, MRC5 (normal) and XP30RO (sv) (XPV; also designated GM3617 (32)) were gifts from A. R. Lehmann (University of Sussex, Brighton, UK). The human fibroblast cells were cultured in Eagle's minimum essential medium. The MEF cells were maintained in Dulbecco's modified Eagle's medium supplemented with 2 mM glutamine. Each medium was supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. The cells were incubated at 37°C in a 5% CO 2 atmosphere.
Mammalian TLS Assay in Cultured Cells-The quantitative indirect TLS assay is a modification of the assay previously described (33). The original assay involved transient transfection of mammalian cells with a gap-lesion plasmid (kan R ), along with a normalizing intact plasmid (cm R ) and a carrier plasmid (amp R ). A control transfection is performed in parallel with a gapped plasmid (kan R ) without a lesion (modified base), along with the same normalizing intact plasmid (cm R ) and carrier plasmid (amp R ). After incubation of the transfected cells to allow TLS, the plasmid contents are extracted, and gapped plasmids that were not completely filled and ligated are denatured by alkaline lysis. The plasmid mixture is then transformed into an E. coli recA strain, and transformants are selected in parallel on LB-kan plates (to select for plasmids that underwent TLS) and LB-cm plates (to select for the normalizing plasmid). The extent of bypass is calculated by dividing the ratio of kan R /cm R transformants obtained with the gap-lesion plasmid mixture by the ratio of kan R /cm R transformants obtained with the gapped plasmid (with no lesion) mixture (33). In the modified procedure, we replace the intact normalizing plasmid (cm R ) by a gapped plasmid without a lesion (cm R ), thereby simplifying the TLS assay. Mammalian cells are co-transfected with a plasmid mixture containing the gap-lesion plasmid (kan R ), a control gapped plasmid without a lesion (cm R ), and the carrier plasmid pUC18 (amp R ). After allowing time for gap filling and lesion bypass, the plasmids were extracted from the cells using alkali, such that only filled in plasmids remained intact. To assay the fraction of filled in plasmids, the plasmid mixture was transformed into an indicator E. coli recA strain and plated in parallel, as in the original method, on LB-kan plates (to select for plasmids that underwent TLS) and LB-cm plates (to select for the control filled in plasmid GP20-cm) (Fig. 2). TLS in this case was calculated by the ratio of kan R /cm R E. coli transformants. The two methods yielded similar results. Specifically, the cells were co-transfected with a DNA mixture containing 50 ng of a gap-lesion plasmid (GP21, GP-BPG1, or GP-BPG2; kan R ), 50 ng of a gapped plasmid without lesion (GP20-cm, cm R ), and 10 g of the carrier plasmid pUC18, using jetPEI/DNA complexes (34). The percentage of lesion bypass gap filling was calculated by dividing the number of GP21*, 2 GP-BPG1*, or GP-BPG2* transformants (number of colonies on LB-kan plates) by the number of corresponding GP20 cm* transformants (number of colonies on LB-cm plates). When desired, plasmids were extracted from kan R colonies, and the sequence opposite the lesion was determined by automated DNA sequencing analysis in the Biological Services Department in this Institute.
In Vivo Complementation Assay-For the pol complementation assay, the MEFs were co-transfected with a DNA mixture containing 1.5 g of gap-lesion plasmid (GP-BPG1, kan R ), 1.5 g of gapped plasmid without lesion (GP20-cm, cm R ), and 1 g of expression vector (pC-pol or the control pCDNA3). In this assay, the MEF cells were electroporated with Nucleofector TM (Amaxa GmbH, Köln, Germany) according to the manufacturer's protocol. When desired, plasmids were extracted from kan R colonies, and the sequence opposite the lesion was determined by automated DNA sequencing analysis in the Biological Services Department in this institute.
TLS Assay in E. coli-Gapped plasmids carrying the BP-G adduct (GP-BPG1 and BP-BPG2; kan R ) and the control plasmid without the adduct (GP20; kan R ) were used in parallel to transform UV-irradiated E. coli cells as previously described (8,29,35). The cells were UVirradiated at 20 Jm Ϫ2 , followed by a recovery period of 30 min at 37°C, after which they were transformed in parallel with the gapped plasmids with and without the BP-G adduct using the Ca-MOPS method (8). Survival was calculated by dividing the number of transformants obtained with the gap-lesion plasmid by the number of transformants obtained with the gapped plasmid without the lesion. In this assay, results from two parallel transformations are compared. Although no internal control is included, the same stock of competent cells was used for the gapped plasmids with and without the lesion, and the reliability of the results was assured by performing multiple experiments for each pair of gapped plasmid constructs. The bacterial strains used in this study were E. coli AB1157 (argE3, hisG4, leuB6, proA2, thr1, ara14, galK2, lacY1, mtl1, xyl5, thi1, tsx33, rpsL31, supE44) and E. coli RW118 (leuBϩ, araD139, sulA211, argE3, hisG4, leuB6, proA2, thr1, ara14, galK2, lacY1, mtl1, xyl5, thi1, tsx33, rpsL31, supE44), both proficient in TLS.
In Vitro Lesion Bypass Assay-The DNA substrates used for this assay were prepared by annealing a 5Ј-32 P-end-labeled oligonucleotide primer to the BPDE-modified 54-mer templates, followed by purification on a BioSpin 30 gel filtration column (Bio-Rad). Primer 5Ј-CTG-GTTCAAGTAGCCAGGTAGGACG-3Ј was used for the BP-G1-modified 54-mer, whereas primer 5Ј-CTGGTTCAAGTAGCCAGGTAGGA-3Ј was used for the BP-G2 54-mer, both creating a substrate with the 3Ј terminus located opposite the template base preceding the lesion. Anal-ysis by electrophoresis on native gels revealed that Ͼ95% of the primers were annealed to the template oligonucleotides. All primer extension reactions contained 5 mM dithiothreitol, 5 mM MgCl 2 , 100 M each of dATP, dCTP, dGTP, and dTTP, 50 nM primer/template, and 25 nM DNA polymerase. Primer extension reaction carried out by E. coli polymerase I Klenow fragment also contained 20 mM Tris⅐HCl, pH 7.5, 0.1 mM EDTA, 8 g/ml bovine serum albumin, and 4% glycerol. Primer extension reaction carried out by human DNA polymerase , human DNA polymerase , or human DNA polymerase also contained 25 mM potassium phosphate (pH 7.0), 100 g/ml bovine serum albumin, and 10% glycerol. The primer extension reactions were carried out at 37°C for 3-20 min. Reactions were stopped by adding an equal volume of a mixture of 99.5% formamide, 0.025% bromphenol blue, and 0.025% xylene cyanol. Samples were fractionated by electrophoresis on 15% polyacrylamide gels containing 8 M urea, after which they were dried, visualized, and quantified using a Fuji BAS 2500 phosphor imager. The extent of bypass was calculated by dividing the amount of bypass products by the amount of all extended primers.

TLS across a Benzo[a]pyrene-guanine Adduct Is
More Efficient in Mammalian Cells than in E. coli-The lesion bypass assay system used is a modified version of a previously described assay based on transfection of cells in culture with a gapped DNA plasmid carrying a site-specific lesion at a predetermined site and carrying an intact kan gene. Plasmids are recovered from cells and are introduced into E. coli recA cells for survival and mutational analysis (Fig. 2). Plasmids in which gaps are repaired by TLS in the mammalian cells give rise to colonies in E. coli (33) (also see "Discussion"). To quantify the results, a normalizing control gapped plasmid without the lesion and carrying a different antibiotics marker (cm R ) was co-transfected along with the plasmid carrying the site-specific lesion. In addition, carrier DNA (pUC18, amp R ) is included in the plasmid mixture. The ratio of kan R /cm R transformants represents the extent of gap filling and lesion bypass ( Fig. 2; for details see "Experimental Procedures"). Therefore, the assay measures TLS indirectly via the recovery of gapped plasmids.
We used gapped plasmids carrying site-specific BP-G adducts in order to study their bypass in mammalian cells. Two constructs were utilized, each containing the sequence GCGTCC derived from the p53 gene, carrying the lung cancer mutational hotspot codon 157 (underlined) (36). Plasmid GP-BPG1 (kan R ), carries the sequence 5Ј-CATGCGTCCTAC-3Ј, whereas GP-BPG2 (kan R ) carries the sequence 5Ј-CATGCGTC-CTAC-3Ј (the G is modified with BP). As shown in Table I (top), when the plasmids were alkali-treated and introduced into the E. coli tester strain without prior passage through the murine cells, TLS was very low (0.05-0.07%), representing the background in this system. When plasmid GP-BPG1 (kan R ) was assayed in PolK ϩ/ϩ cells, TLS was nearly 3 orders of magnitude above background, reaching an extent of 48 Ϯ 4%. Similar results were obtained with plasmid GP-BPG2 (kan R ) (52 Ϯ 4%; Table I (top)). This extent of bypass is much higher than values usually reported for bypass across BP-G adducts in E. coli (37)(38)(39). However, since the differences might stem from DNA sequence context effects, which are known to strongly affect bypass across BP-G adducts in vitro and in E. coli in vivo (40 -43), we assayed TLS across the BP-G adduct in E. coli cells, using the same gapped plasmids used in the mammalian cells. The assay, performed as previously described (8,29,35), involved transformation of SOS-induced E. coli cells with plasmid GP-BPG1(kan R ) or GP-BPG2 (kan R ), followed by selection on LB-kan plates. As a control, the cells were transformed in parallel with the gapped plasmid without the lesion. TLS was calculated by the ratio of transformants obtained with the gap-lesion plasmids to the number of transformants obtained with the gapped plasmid without a lesion. As can be seen in Table I (bottom), bypass across the BP-G adduct was poor in E. coli, both with the GP-BPG1 (kan R ) and GP-BPG2 (kan R ) constructs, reaching 1.6% at most. SOS induction caused a FIG. 1. Gap-lesion plasmids used to assay TLS in living cells. A, outline of the construction of 54-mer oligonucleotides carrying site-specific BP-G adducts that were used to build the gap-lesion plasmids. See "Experimental Procedures" for details. B, the structure of gapped plasmids used in this study. The black rectangle represents a damaged nucleotide. The DNA sequence at the gap region is shown. The underlined G represents the BP-G lesion. G1 and G2, the two guanines that were modified with BP in plasmids GP-BPG1 and GP-BPG2, respectively (underlined in A and B); X (in GP21) represents an abasic site. slight increase in bypass, but even under these conditions bypass did not exceed 2%. Thus, bypass across the BP-G adduct was at least an order of magnitude more efficient in mice embryo fibroblasts than in E. coli.
PolK Ϫ/Ϫ Mouse Embryo Fibroblasts Are Deficient in TLS across a Benzo[a]pyrene-guanine Adduct-We examined the role of pol in bypass across BP-G adducts by performing the TLS assay in parallel in PolK ϩ/ϩ and PolK Ϫ/Ϫ cells. As can be seen in Fig. 3A and Table II, bypass across the BP-G adduct in gapped plasmid GP-BPG1 (kan R ) in PolK Ϫ/Ϫ MEFs was 16 Ϯ 1%, 3-fold lower than in PolK ϩ/ϩ MEFs. Similar results were obtained with plasmid GP-BPG2 (kan R ). In this case, TLS was 52 Ϯ 4 and 20 Ϯ 3% in PolK ϩ/ϩ and PolK Ϫ/Ϫ cells, respectively (Table II).
In order to examine whether the decreased TLS in PolK Ϫ/Ϫ is indeed due to the lack of pol, we used a plasmid expressing human pol from the cytomegalovirus immediate early promoter to complement the deficiency in PolK Ϫ/Ϫ cells. Fig. 3A and Table II show the results of such complementation experiments performed with plasmid GP-BPG1 (kan R ). TLS in the control PolK Ϫ/Ϫ cells that were transfected with the empty expression vector was 20 Ϯ 2.5%. However, when co-transfected with the plasmid expressing human pol, bypass increased to 49 Ϯ 6%, similar to the levels observed in PolK ϩ/ϩ cells (Table II; 2. Outline of the quantitative TLS assay. Cultured mammalian cells are transfected with a mixture of a gap-lesion plasmid (kan R ), a control gapped plasmid without a lesion (cm R ), and a carrier plasmid (amp R ). After gap filling by TLS in the mammalian cells, the plasmids were extracted and introduced into an indicator E. coli cell for analysis. Only gapped plasmids that were completely filled in the mammalian cells gave rise to bacterial colonies on selective media. TLS is calculated by the ratio of kan R /cm R colonies. See "Results" for details.

TABLE I Extent of bypass across BP-G adducts in mouse embryofibroblasts and in E. coli cells
Plasmid mixtures containing the indicated gap-lesion plasmid (kan R ), the gapped plasmid GP20-cm (cm R ), and the carrier plasmid pUC18 (amp R ) were introduced into PolK ϩ/ϩ mouse embryo fibroblasts by transfection, after which the DNA was extracted and introduced into the E. coli indicator strain. Top, bypass levels were calculated by dividing the number of kan R colonies by the number of cm R colonies. The table shows the average of at least six transfections. Bottom, E. coli cells, either unirradiated or UV-irradiated, were transformed in parallel with the indicated gap-lesion plasmid and with a control gapped plasmid without the lesion. Bypass levels were obtained by dividing the number of transformants obtained with the gap-lesion plasmid by the number of transformants obtained with the control gapped plasmid. a The number of transformants obtained in a typical assay with 100 l of the transformation mixture. For plates carrying over 300 transformants, appropriate dilutions were made in order to obtain an accurate colony count. b A DNA mixture containing the indicated gap-lesion plasmid was introduced into E. coli JM109recA, without prior passage through the mouse cells.
c The E. coli strains used in this study are proficient in TLS. Genotypes are given under "Experimental Procedures." d Number of colonies obtained with the gap-lesion plasmid (number on the left) and the control gapped plasmid (number on the right) in a typical assay with 100 l of the transformation mixture. In order to obtain an accurate colony count, appropriate dilutions were made to get 100 -300 colonies/plate. two-thirds of the BP-G adducts in our system.
Bypass across the BP-G Adduct Is More Mutagenic in the Absence of Pol-Next we analyzed the specificity of nucleotide insertion opposite BP-G adducts. As shown in Table III, correct incorporation of C opposite BP-G adducts was more frequent than the incorrect incorporation of other nucleotides in both PolK ϩ/ϩ and PolK Ϫ/Ϫ cells. However, incorporation of the incorrect nucleotides (A, G, or T) was significantly greater in PolK Ϫ/Ϫ than in PolK ϩ/ϩ cells (Table III). With plasmid GP-BPG1 (kan R ), 29% of the bypass events in PolK ϩ/ϩ cells resulted in mutations, compared with 50% in PolK Ϫ/Ϫ cells (Fig.  3B, Table III). With plasmid GP-BPG2 (kan R ), bypass was only slightly more mutagenic in PolK Ϫ/Ϫ cells (71%) compared with PolK ϩ/ϩ (53%). With both plasmids, the main mutagenic event was misinsertion of A opposite the BP-G adduct. Complementation experiments revealed that while in the PolK Ϫ/Ϫ cells transfected with the empty vector, 44% of the bypass events were mutagenic, in such cells co-transfected with the polexpressing plasmid, only 17% of the bypass events were mutagenic (Fig. 3B, Table III). Thus, expressing pol in PolK Ϫ/Ϫ cells renders TLS across the BP-G adduct less mutagenic than in cells lacking pol.
Pol Is Not Required for TLS across a Synthetic Abasic Site in Mouse Embryo Fibroblasts-To examine the substrate specificity of TLS by pol, we performed experiments with gapped plasmids carrying a site-specific synthetic abasic site. Bypass across the abasic site reached levels of 36 Ϯ 4% and 38 Ϯ 4% in PolK ϩ/ϩ and PolK Ϫ/Ϫ cells, respectively (Table IV, top). Analysis of the specificity of base insertion opposite the abasic site revealed that purines, primarily A, were preferentially inserted opposite the abasic site (Table IV, bottom), consistent with our results in human cells in culture (33). Similar spectra of base insertions were observed in PolK ϩ/ϩ and PolK Ϫ/Ϫ cells (Table  IV, bottom). These results indicate that pol is not required for in vivo bypass across an abasic site.
Pol Is Not Required for Bypass across the BP-G Adduct-To explore the polymerase specificity for TLS, we examined the ability of pol encoded by the human XPV gene to bypass BP-G lesions in DNA. As shown in Table V (top), bypass in the human cell line MRC5 (XPV ϩ/ϩ ) was 35 Ϯ 4 and 39 Ϯ 4% for plasmids GP-BPG1 (kan R ) and GP-BPG2 (kan R ), respectively. Similarly, bypass in the human XPV-defective cell line XP30R was 40 Ϯ 3 and 44 Ϯ 5% for plasmids GP-BPG1 (kan R ) and GP-BPG2 (kan R ), respectively (Table V, top). We also examined the specificity of nucleotide insertion opposite the BP-G adduct and found that in both cell lines bypass was largely accurate: 90% C insertion in XPV ϩ/ϩ cells and Ͼ95% C insertion in XPV Ϫ/Ϫ cells. This accuracy of bypass is higher than in the PolK ϩ/ϩ MEFs (71%; Table III). The reason for the difference in accuracy is not clear. It may stem from differences between human and mice cells and/or transformed cell lines versus embryo cells. In conclusion, pol is not required for bypass across the BP-G adduct in living cells.
In Vitro Bypass across the BP-G Adduct by Specialized DNA Polymerases-We examined the possibility that the primary role of pol in TLS across the BP-G adduct stems from the failure of other DNA polymerases to bypass this lesion. It was previously shown that pol can bypass BP-G adducts (44 -46). However, since TLS across BP-G is strongly influenced by DNA sequence context (40 -43), we determined the ability of purified human pol, pol, and pol to bypass the BP-G adduct in the same DNA sequence context that was used in our in vivo studies. Fig. 4 shows the kinetics of TLS across the BP-G lesion by purified recombinant human DNA polymerases , , and .

TABLE II Extent of bypass across a BP-G adduct in PolK ϩ/ϩ and PolK Ϫ/Ϫ MEFs in the presence or absence of pol expressed from a plasmid
The plasmid mixtures contained the indicated gap-lesion plasmid (kan R ) and the control gapped plasmid GP20-cm (cm R ) and, when indicated, also the pol expression vector pC-pol, or the empty vector pCDNA3. The plasmids were transfected into PolK ϩ/ϩ or PolK Ϫ/Ϫ MEFs, after which the DNA was extracted and introduced into the E. coli indicator strain. Bypass levels were calculated by dividing the kan R /cm R transformants obtained. The table shows the average bypass results of at least three transfections.

Mouse cells (MEFs)
Gap-lesion plasmid Expression vector The number of transformants obtained in a typical assay with 100 l of transformation mixture except for the complementation experiments, where 500 l transformation mixture was used to compensate for a lower efficiency of the electroporation transfection method. This did not affect the results of lesion bypass, which are calculated by the ratio of Kan R /cm R colonies.
Bypass was assayed using primed oligonucleotides, whose DNA sequence is the same as in the gap lesion plasmids used in the in vivo experiments. Pol bypassed the BP-G adduct in either of the two sequence contexts examined with similarly high efficiency, reaching ϳ40% in 10 min (Fig. 4). In contrast, pol exhibited different bypass extents at the two sequence contexts. On template G1, bypass was slower than by pol. However, on template G2, pol and pol showed similar bypass extents, although the product distribution was different. Pol was unable to bypass the adducts under these conditions. Thus, consistent with previous results (44 -46), at least one other DNA polymerase, pol, is capable of bypassing a BP-G adduct in vitro, although it is not essential for this bypass reaction in vivo.

DISCUSSION
The multiplicity of DNA polymerases, particularly from the Y superfamily, suggests that lesion bypass is an important housekeeping process in protecting mammalian cells from genotoxic agents. The study of the intricate network of TLS polymerases and their mode of action in the cell requires quantitative tools. We have used a plasmid-based quantitative TLS assay, developed in our laboratory, to study the involvement of pol in TLS in cultured cells. The major advantages of this system are its specificity for TLS and its quantitative results, which are essential for the in depth understanding of in vivo TLS. The main disadvantages of this system are that it is indirect (based on plasmid recovery) and that it utilizes a nonreplicating episomal substrate rather than a chromosomal substrate. Nonetheless, the system can be viewed as a model for gap filling during lagging strand DNA replication. Indeed, the system is responsive to the type of DNA damage and to the cellular composition of DNA polymerases, as shown in this and in previous studies (33,47). The specificity of this assay for TLS is dictated by the structure of the plasmid used, namely a gap-lesion plasmid. Indeed, nucleotide excision repair is known not to act on single-stranded DNA regions (1), and in the absence of an intact complementary strand, recombinational repair cannot act either, as demonstrated previously (33). Base excision repair, on the other hand, can potentially interfere with the assay, since some DNA damage-specific DNA glycosylases, which initiate this pathway, can act on single-stranded DNA, leading to the formation of abasic sites (1). This means that if a DNA glycosylase were to excise the BP-G adduct, the TABLE III Insertion specificity opposite a BP-G adduct during TLS in PolK ϩ/ϩ and PolK Ϫ/Ϫ MEFs in the presence or absence of pol expressed from a plasmid The plasmid mixtures contained the indicated gap-lesion plasmid (kan R ) and the control gapped plasmid GP20-cm (cm R ) and, when indicated, also the pol expression vector pC-pol or the empty vector pCDNA3. The plasmids were transfected into PolK ϩ/ϩ or PolK Ϫ/Ϫ MEFs, after which the DNA was extracted and introduced into the E. coli indicator strain. Plasmids were extracted from kan R colonies and subjected to DNA sequence analysis. The table shows the DNA sequence opposite the lesion obtained for individual clones.

TABLE IV Extent and specificity of bypass across an abasic site in
PolK ϩ/ϩ and PolK Ϫ/Ϫ MEFs The plasmid mixtures containing the indicated gapped plasmid (GP21; kan R or GP20-cm; cm R ) and the carrier plasmid pUC18 were introduced into the indicated mammalian cells, after which the DNA was extracted and introduced into the E. coli indicator strain. Top, bypass levels were calculated by dividing the number of kan R colonies obtained for GP21 (containing the abasic site) by that obtained for GP20-cm (with no lesion; cm R ). The average of three experiments is presented. Bottom, plasmids were extracted from kan R colonies containing GP21 descendants obtained in the top of the  19 19 a The number of transformants obtained in a typical assay, with 100 l of transformation mixture.  20 20 a The number of transformants obtained in a typical assay, with 100 l of transformation mixture. observed effects would have been due to abasic sites rather than the BP-G adduct. This possibility is ruled out by two arguments: (a) there was no difference in bypass across an abasic site in PolK ϩ/ϩ and PolK Ϫ/Ϫ cells, whereas there were clear differences in the bypass across BP-G in these two cell types, as described above; (b) there is no known DNA glycosylase that acts on BP-G. Thus, the results obtained with the gapped plasmid carrying the BP-G adduct in the singlestranded DNA region can be attributed solely to TLS in the mammalian cells.
Our results indicate that the BP-G adduct is bypassed in mammalian cells with considerable efficiency, amounting to 35-50%. This is more than 20-fold higher than the extent of bypass in SOS-induced E. coli across the same BP-G lesion, using the same gapped plasmid constructs. These results, together with previous results demonstrating much higher bypass across an abasic site in mammalian cells compared with E. coli (33), indicate that TLS in general is much more efficient in mammalian cells than in E. coli and is therefore likely to be a more significant repair function in mammals than in E. coli.
Two lines of evidence implicate pol in cellular responses to BP; PolK Ϫ/Ϫ mouse embryo fibroblasts were found to be sensitive to killing and mutagenesis by BP (26), and the PolK gene is subject to arylhydrocarbon receptor-dependent inducible transcription (48). These results clearly indicate that pol plays an important role in protecting mouse cells from the killing and mutagenic effects of BP, consistent with a role of pol in TLS in vivo. However, the same results can be explained by the involvement of pol in responses to DNA damage other than TLS, such as error-free repair and checkpoint activation. Indeed, a link between pol and checkpoint activation, including the physical interaction between DinB and Hus1 and Rad1, was recently reported in Schizosaccharomyces pombe (49). Similarly, pol⑀ was reported to be a checkpoint protein (50). The experiments presented in this study provide evidence for the direct involvement of pol in TLS across BP-G adducts in living cells, since cells lacking pol exhibited a 3-fold reduction in TLS across this lesion, and expression of human pol in these cells resulted in complementation of the bypass defect. Based on these results, pol is responsible for the bypass of at least two-thirds of the BP-G adducts. In fact, since in the absence of pol another polymerase may carry out bypass cross BP-G to some extent, it is possible that in PolK ϩ/ϩ cells pol is responsible for the bypass of more than two-thirds of the BP-G lesions. This is quite remarkable, given the multiplicity of DNA polymerases present in mammalian cells (at least 14 additional DNA polymerases) (5, 51), underscoring the principle of DNA damage specificity in the bypass activity of at least certain TLS DNA polymerases. FIG. 4. In vitro bypass across BP-G by purified Y family human DNA polymerases. Primer extension assayed were performed with the indicated purified recombinant human DNA polymerases, using primed oligonucleotides carrying a site-specific BP-G adduct in the template strand, as described under "Experimental Procedures." A and B, experiments performed with BP-G at two different sequence contexts, which are shown at the top of the gel images (the modified G is underlined). The upper panels show phosphor images of reaction products, whereas the lower panels show the quantification of lesion bypass based on these images. Lane 1 in each panel shows a control incubation of the reaction mixture without any DNA polymerase. Other lanes contain the products of reactions performed with the following DNA polymerases: human pol (lanes 2-4); human pol (lanes 5-7); human pol (lanes 8 -10); E. coli polymerase I (Klenow fragment) (lanes [11][12][13]. For each DNA polymerase, three time points of the reaction are shown. The bands of oligonucleotides shorter than 26 nucleotides (A, lanes [11][12][13] represent excision products of the 3Ј-5Ј proofreading activity of polymerase I KF. The important role of pol in bypass across the BP-G adduct does not stem from an inherent inability of other DNA polymerases to bypass this lesion, since purified pol was found to bypass the BP-G adduct in one of the constructs tested with an efficiency similar to that of pol. This is consistent with the idea that regulation of TLS DNA polymerases plays an important role in the final TLS outcome. It is clear, however, that pol is not the only DNA polymerase that can bypass BP-G in vivo, since bypass extents of 16 -20% were obtained in cells lacking pol. At this point, the identity of these polymerases is still unknown; however, a possible candidate is pol, based on its ability to bypass the BP-G adduct in vitro, as shown in this (Fig. 4) and in previous studies (44 -46). It is currently unknown whether pol functions in the bypass of DNA lesions other than BP-G. However, it is likely that additional modified bases, perhaps adducts with other aromatic compounds, will be bypassed by pol.
It was previously reported that purified pol is a promiscuous extender of mispaired termini in vitro (52,53). At this stage, we do not know whether pol performs both the misinsertion and extension steps in vivo or whether an additional polymerase is involved in the bypass reaction. Analysis of the specificity of nucleotide incorporation opposite the BP-G site indicates that the major mutagenic event involves the misinsertion of A, consistent with previous results (54,55). Interestingly, bypass was more mutagenic in the absence of pol, resulting in an ϳ2-fold higher frequency of incorporation of incorrect nucleotides. This is not a big effect on mutation frequency, but it may translate to a much bigger effect on the mutagenic outcome of BP adducts in carcinogenesis, since the cumulative mutational outcome of DNA lesions grows exponentially with the number of hits. Overall, the results presented here, together with previous results, indicate that pol bypasses BP-G adducts with higher efficiency and higher accuracy than other DNA polymerases. This resembles the action of pol on cyclobutyl pyrimidine dimers, making pol the second demonstrated case of a TLS polymerase with the ability to bypass a specific lesion with higher accuracy than other DNA polymerases. Collectively, these results suggest that at least some DNA polymerases function to bypass particular DNA lesions with higher efficiency and higher accuracy than others. In this sense, pol and pol represent DNA damage tolerance enzymes, which are dedicated to minimizing the hazards of certain unrepaired lesions in DNA by restoring DNA replication in the face of arrest, with a reduced probability of mutations compared with other polymerases.