Efficiency and Accuracy of SOS-induced DNA Polymerases Replicating Benzo[a]pyrene-7,8-diol 9,10-Epoxide A and G Adducts*

Nucleotide incorporation fidelity, mismatch extension, and translesion DNA synthesis efficiencies were determined using SOS-induced Escherichia coli DNA polymerases (pol) II, IV, and V to copy 10R and 10S isomers oftrans-opened benzo[a]pyrene-7,8-diol 9,10-epoxide (BaP DE) A and G adducts. A-BaP DE adducts were bypassed by pol V with moderate accuracy and considerably higher efficiency than by pol II or IV. Error-prone pol V copied G-BaP DE-adducted DNA poorly, forming A·G-BaP DE-S and -R mismatches over C·G-BaP DE-S and -R correct matches by factors of ∼350- and 130-fold, respectively, even favoring G·G-BaP DE mismatches over correct matches by factors of 2–4-fold. In contrast, pol IV bypassed G-BaP DE adducts with the highest efficiency and fidelity, making misincorporations with a frequency of 10−2 to 10−4 depending on sequence context. G-BaP DE-S-adducted M13 DNA yielded 4-fold fewer plaques when transfected into SOS-induced ΔdinB (pol IV-deficient) mutant cells compared with the isogenic wild-typeE. coli strain, consistent with the in vitrodata showing that pol IV was most effective by far at copying the G-BaP DE-S adduct. SOS polymerases are adept at copying a variety of lesions, but the relative contribution of each SOS polymerase to copying damaged DNA appears to be determined by the lesion's identity.

Recent studies suggest that in addition to DNA polymerases (pol) 1 I-III, which are prototypes of the A-, B-, and C-polymerase families, respectively (1,2), Escherichia coli possesses two members of the recently described Y-family of DNA polymerases (3). This new polymerase family is typified by the UmuC, DinB, Rev1, and Rad30 proteins, which represent distinct phylogenetic branches of the Y-family tree (4,5). The Y-family polymerases lack intrinsic exonuclease activity and are distributive in the absence of stimulatory accessory factors (6). Members of this new polymerase family are best characterized by their lesion-bypassing properties; but it is clear that when they replicate undamaged DNA, they do so with fidelities much lower than those of the A-, B-, and C-family polymerases (6).
Both E. coli Y-family pol IV (DinB) and pol V (UmuDЈ 2 C) are expressed at elevated levels as part of the cell's global SOS response to DNA damage (7). It is clear from years of genetic characterization of umuD and umuC mutants that the principal role of pol V is to bypass template bases that effectively block normal pol III-catalyzed DNA replication because deletion of the umu operon or missense mutations in either umuD or umuC effectively render E. coli cells non-mutable, despite being exposed to a variety of mutagens (8 -10). Translesion DNA synthesis catalyzed by pol V results in mutations targeted directly opposite DNA damage sites. However, in addition to its primary role in replicating across from damaged bases, pol V also causes untargeted base substitution mutations in the absence of DNA damage (11).
pol IV is responsible for generating frameshift mutations on undamaged DNA, causing adaptive mutations in non-dividing cells (12,13), mutations that are kept in check by the presence of pol II (14). It is possible, however, that the primary role of pol IV may be to rescue stalled replication forks, which seem to occur at least once per round of replication (15), possibly by extending slipped mispairs that are refractory to proofreading. Genetic experiments also implicate pol IV in translesion synthesis (TLS) of certain types of lesion such as acetylaminofluorene-modified G residues as well as benzo[a]pyrene-7,8-diol 9,10-epoxide (BaP DE) G-adducted DNA (16).
pol II, an SOS-induced DNA polymerase (17)(18)(19) that is the prototype for the B-family of polymerases (18), is a high fidelity enzyme containing an associated 3Ј-exonuclease proofreading activity (20). Although the main cellular role of pol II is believed to be in "error-free" replication restart (21), it also can copy specific DNA lesions in vivo (16,22).
In this study, we have investigated the ability of the five E. coli DNA polymerases to copy bulky adducts derived from trans-ring opening of bay region 7,8-diol 9,10-epoxide metabolites of the polycyclic aromatic hydrocarbon benzo[a]pyrene (BaP DE; see Fig. 1A). Two 7,8-diol 9,10-epoxide diastereomers, each consisting of a pair of enantiomers, are formed metabolically in mammals. These BaP DEs are mutagenic in bacterial and mammalian cells (23), and the predominant isomer formed metabolically, the (ϩ)-enantiomer of the diastereomer (shown in Fig. 1A), whose benzylic hydroxyl group and epoxide oxygen are trans, is highly tumorigenic in mice (24). DNA damage by BaP DEs, which target the exocyclic amino groups of purines in DNA (see Fig. 1A) (25), is most likely an important initiating event in both mutagenesis and cell transformation. Replicative DNA polymerases are generally blocked by bulky adducts (26), and human pol ␣ replicates at best very poorly opposite BaP DE adducts (27), as does the non-SOS-induced E. coli polymerase I Klenow fragment (28,29). Thus, it is attractive to speculate that error-prone TLS by bypass polymerases may be highly significant in the observed induction of mutations by these adducts, especially in SOS-induced cells (30,31). Two central issues are addressed in this study. First, we compare the efficiency and accuracy of TLS by the SOS-induced DNA polymerases when copying the G-and A-BaP DE adducts. Second, data on the efficiency and accuracy of BaP DE adduct bypass by the SOS-induced polymerases in vitro are then compared with in vivo data in ⌬polB (pol II), ⌬dinB (pol IV), and ⌬umuDC (pol V) mutant backgrounds to address the extent to which the model in vitro measurements correlate with in vivo results.

EXPERIMENTAL PROCEDURES
Materials-Ultrapure ATP, dNTPs, E. coli single-stranded DNAbinding protein (SSB), and RecA were purchased from Amersham Biosciences, Inc. ATP␥S was purchased from Roche Molecular Biochemicals. E. coli DNA polymerase I Klenow fragment (KF) and its exo Ϫ derivative were purchased from New England Biolabs Inc. pol II, pol II exo Ϫ , pol III subunits, pol IV (DinB), and pol V (UmuDЈ 2 C) were purified as previously described (32)(33)(34); and all are native enzymes, except for pol IV, which is linked at its N terminus to a maltose-binding protein. BaP DE-containing oligonucleotides were made as previously described (29,30). Three sequence contexts (one for A and two for G adducts) that had previously been used to examine the mutational consequences of these adducts in an E. coli M13 system were used (30). For each of these adducts, two diastereomers are possible, corresponding to an S or R configuration at the point of attachment of the hydrocarbon to the purine base (see Fig. 1A); and both isomers, the BaP DE-S and -R adducts, respectively, were examined. Linear BaP DE-containing DNA templates used in the lesion bypass assay and fidelity study were prepared by ligation of linearized single-stranded M13mp7 DNA with BaP DE-containing oligonucleotides and a synthetic 48-mer as previously described (35). BaP DE-containing circular M13mp7 DNA used in E. coli transfections was prepared as described (30).
In Vitro Translesion Synthesis Analysis-5Ј-32 P-labeled 28-mer primers were annealed to the BaP DE-containing template so that the 3Ј-end of the primer was situated either 2 or 48 nucleotides before the lesion site. The reaction mixtures contained 20 mM Tris-HCl (pH 7.5), 8 mM MgCl 2 , 5 mM dithiothreitol, 0.1 mM EDTA, 25 mM sodium glutamate, 40 g/ml bovine serum albumin, 4% glycerol, 2 nM primer-template, and four dNTPs at 0.5 mM each. pol I KF and pol I KF exo Ϫ were used at 0.05 units/l. pol II and pol II exo Ϫ were present in the reaction at 20 nM. The pol III ␣-subunit, pol IV, and pol V were present at 10, 50, and 200 nM, respectively. For pol II, pol II exo Ϫ , pol III␣, pol IV, and pol V, ␤⅐␥ complex, SSB, and ATP or ATP␥S (for pol V) were included in the reaction at 40 nM, 10 nM, 300 nM, and 1 mM, respectively. Reactions with pol V also included 1 M RecA. Reactions were carried out at 37°C for 10 min and quenched by addition of an equal volume of 40 mM EDTA and 95% formamide. Primer extension products were heat-denatured and resolved on a denaturing polyacrylamide gel. Integrated gel band intensities were measured with a PhosphorImager using ImageQuant software (Molecular Dynamics, Inc.). The bypass efficiencies were measured as the fraction of primers extended beyond the lesion site X: is the integrated gel band intensities of primers extended opposite sites Xϩ1, Xϩ2, etc., and I X and I X-1 are the integrated gel band intensities measured directly opposite the lesion site X and 1 base prior to the lesion site XϪ1, respectively.
Nucleotide Misincorporation and Mismatch Extension Analysis-A standard gel kinetic assay was used to measure the misincorporation and mismatch extension fidelity as previously described (36,37). The concentration of primer-template DNA in the reactions was 2 nM. The reaction conditions were the same as in the translesion synthesis assay, except that different sets of primers were annealed to the template, and incorporation of individual dNTP substrates was measured as a function of concentration to determine apparent V max and K m values for each substrate. Standing-start primer extension reactions (38) were performed with incubation times chosen to satisfy single-completed-hit conditions, i.e. Ͻ20% of the input primer molecules were extended (36,38). Apparent V max and K m values were determined by plotting primer extension velocity versus dNTP concentration, fitting the data to a rectangular hyperbola with nonlinear least-squares regression using Sigmaplot. The misincorporation efficiency (f inc ) was determined as the ratio of apparent V max /K m for incorrect incorporation to apparent V max /K m for correct incorporation. The intrinsic mismatch extension efficiency (f 0 ext ) was determined as the ratio of apparent V max /K m for mismatch extension to apparent V max /K m for correct match extension (36,37).
BaP DE-containing M13 Transfection and Analysis of Progeny Bacteriophage-The transfection procedure was the same as previously described (30) with the following modifications. AB1157 (F Ϫ ) was used as the wild-type strain, with CSH50 (FЈ) added as the indicator prior to plating. STL1336 (AB1157 with ⌬polB::⍀) and SR1157U (AB1157 with ⌬umuDC::cat) were used as pol II-deficient and pol V-deficient strains, respectively. The pol IV-deficient strain was constructed by P1 transduction into AB1157 using a lysate from RW626 (AB1157 with dnaE486 zae502::Tn10 ⌬polB::⍀ ⌬dinB::Kan ⌬umuDC::cat), selecting for the associated kanamycin resistance of the ⌬dinB allele. M13 DNAs containing A-BaP DE-S and G-BaP DE-S adducts in contexts I and IV (see Fig.  1) and unadducted control DNAs were used in parallel for comparison. pol II, IV, and V were induced by UV irradiation as noted in Table IV. Analysis of progeny bacteriophage was carried out by a combination of probe hybridization and sequencing following the protocol referred to as "Experiment I" in Ref. 30.

BaP DE Lesion Bypass in Vitro
Using the Five E. coli DNA Polymerases-All five E. coli DNA polymerases were used to copy each of the two diastereomeric (10S and 10R) transopened A-and G-BaP DE adducts. The A adducts were examined in context I, and the G adducts in contexts III and IV (Fig.  1). The notation for the oligonucleotide sequence contexts was adopted from a previous in vivo study (30). Two running-start primer-template DNAs were used, one with a primer 3Ј-end located 2 nucleotides from the lesion (Fig. 2, upper panels) and the other with the primer end 48 nucleotides upstream from the lesion (Fig. 2, lower panels). TLS efficiencies for the three SOS polymerases (pol II, IV, and V) are calculated in Table I.
Each of the five polymerases was able to catalyze nucleotide incorporation, at least to some extent, directly opposite A-BaP DE adducts with both 10S and 10R configurations (Fig. 2). However, only two of the polymerases, pol I KF and pol V, were able to replicate beyond the lesion (Fig. 2, lanes 1 and 6). Bypass of either A adduct isomer by KF was weak (ϳ3-5% efficiency) even in the presence of a large enough excess of KF to extend almost all of the input primer ( Fig. 2, lower left panel, lane 1). The pol V-catalyzed TLS required RecA ϩ SSB and was stimulated in the presence of ␤⅐␥ processivity factors (data not shown). Lesion bypass by pol V was processive and much more robust when poorly hydrolyzable ATP␥S was used in place of ATP for assembling the RecA nucleoprotein filament (Fig. 2, compare lanes 5 and 6; and Table I). Contact between pol V and the 3Ј-tip of the RecA nucleoprotein filament is required both for TLS to occur and for pol V to remain bound at a primer 3Ј-end (39 -41). Disassembly of the filament in the 5Ј to 3Ј direction occurred in the presence of ATP, but not ATP␥S (42,43), causing termination of pol V synthesis roughly 1-4 bases downstream from the lesion (Fig. 2, upper left panel, lane 5 for both the S and R isomers). pol V showed a somewhat greater TLS efficiency when replicating beyond the R rather than the S isomer (Table I, ATP␥S column).
With each of the two diastereomeric G adducts in both sequence contexts, all five polymerases could incorporate a base opposite the adduct. Incorporation at the lesion site was most easily seen when the enzymes were given a long running start (primer 2) (Fig. 2, lower panels). The efficiencies of nucleotide incorporation at the lesion site varied, as did the efficiencies with which further primer extension (TLS) occurred beyond this site. Both pol IV and V carried out TLS past G-BaP DE-R moieties, with pol IV being much more efficient. Thus, pol IV is by far the best at primer extension past the G-BaP DE adducts and is clearly better than pol V, which was most efficient for TLS at A adducts (Table I). Of the four G lesions (two isomers in two sequence contexts), three were bypassed by pol IV with similar high efficiencies (ϳ70%), and the fourth (G-BaP DE-R in context III) was bypassed about half as well. pol V exhibited a modest diastereomer preference for bypassing the G-BaP DE-R isomer in both sequence contexts, and the only apparent effect of sequence context on pol V-catalyzed TLS was that the "favored" G-BaP DE-R isomer in context IV was bypassed to a significant extent (7%) even when the RecA filament was assembled in the presence of ATP (Table I).
In summary, pol IV bypassed only G-BaP DE adducts, favoring the R over the S isomer in one context but not in the other, whereas pol V bypassed both A-and G-BaP DE lesions to about the same extent, favoring the R over the S isomer in a sequence context-independent manner. The replicative pol III carried out limited synthesis (5.1%) past the G-BaP DE-R isomer in context IV, but no significant synthesis past G-BaP DE-S (Fig. 2, lower panels). No detectable synthesis beyond either G-BaP DE isomer was observed for pol III in context III.  ured the relative nucleotide incorporation and primer extension efficiencies (Tables II and III) for the three SOS polymerases (pol II, IV, and V) at the A-BaP DE-S and -R and G-BaP DE-S and -R adducts using the sequence contexts shown in Fig. 1. The nucleotide incorporation specificities for the three SOS-induced polymerases are unusual, with major differences in pol V, IV, and pol II incorporation and extension fidelities depending on sequence context, adduct configuration at C-10 (R or S), and whether the adduct is at A or G.

Fidelity of Nucleotide Incorporation and Extension at BaP
pol V Nucleotide Incorporation and Extension Fidelity at BaP DE Template Adducts-pol V catalyzed misincorporation of A opposite G-BaP DE-S in context III by a factor of 350-fold greater than the "correct" incorporation of C (Table II). An anomalously high misincorporation specificity was also observed with G-BaP DE-R, where formation of an A⅐G-BaP DE-R mispair was favored by a factor of 132-fold compared with a C⅐G-BaP DE-R correct pair (Table II). Thus, pol V strongly favors making G⅐C to T⅐A transversions in this sequence context. Both A⅐G-BaP DE mispairs were also extended with higher efficiency than their correctly paired counterparts: 23fold greater for the A⅐G-BaP DE-S isomer compared with C⅐G-BaP DE-S and 3.9-fold greater for the corresponding matched and mismatched G-BaP DE-R isomers in context III (Table III).
An important point is that the A⅐G-BaP DE misincorporation errors in context III are unlikely to result from a simple template slippage event (44,45), with the lesion out of the helical plane, because the template base immediately downstream from the adduct is not a T, but is instead a C. The favored misincorporation of A opposite G-BaP DE also cannot be caused by backward slippage of the primer relative to the template strand. This type of misalignment event is precluded in context III because an incoming dATP substrate is unlikely to be incorporated readily opposite the As on the template strand immediately upstream from the lesion (Fig. 1). In contrast, backward slippage events should be more likely to result in misincorporation of A opposite the G adduct in context IV because in this sequence, two Ts are just upstream from the lesion. Remarkably, however, A⅐G-BaP DE-S and -R mispairs have f inc values of "only" 0.78 and 0.18, respectively, in context IV, compared with 350 and 132 in context III, a result clearly inconsistent with backward slippage of the primer. The lower frequency of A⅐G-BaP DE mispairs relative to correct pairs in sequence context IV, might, however, result from a looping out of the adducted G that could enable pol V to use an unmodified G downstream in this sequence as a template for incorporating a correct C.
G⅐G mispairs in context III, which are favored over C⅐G base pairs by factors of 2-and 3.9-fold for S and R isomers, respectively, could have resulted from a similar primer-template misalignment with the modified base looped out of the helical plane, allowing incorporation of G opposite the downstream template C. However, we think it is much more likely that G is incorporated directly opposite the G-BaP DE adduct because the very high apparent K m values (0.74 mM for the R isomer and Ͼ1 mM for the S isomer) (data not shown) are characteristic of direct misincorporations (46,47). Much smaller apparent K m values intermediate in magnitude between Watson-Crick base pair formation and non-Watson-Crick nucleotide misincorporation are a hallmark of correct incorporations on misaligned primer-template DNA (46,47). Once formed, both G⅐G-BaP DE-S and -R mispairs were extended more efficiently than the corresponding correctly matched C⅐G-BaP DE pairs: 2.2-fold for the R isomer and 1.8-fold for the S isomer (Table  III).
pol V is the only SOS polymerase that could bypass the A-BaP DE adducts efficiently (Fig. 2). The major errors observed when pol V copied A-BaP DE were A⅐A-BaP DE-R mispairs (f inc ϭ 0.23) and A⅐A-BaP DE-S mispairs (f inc ϭ 0.08). Thus, the primary mutational event when pol V copies an A-BaP DE moiety is expected to be an A⅐T to T⅐A transversion.  G⅐A-BaP DE-R and -S errors occurred less frequently, with f inc values of 0.026 and 0.05, respectively (Table II) (Table III). The same was true for the A-BaP DE-R isomer, except for C⅐A-BaP DE-R, which was extended with an ϳ20-fold lower efficiency (Table III). pol IV Nucleotide Incorporation and Extension Fidelity at BaP DE Template Adducts-pol IV-catalyzed TLS was favored for G-BaP DE adducts, with pol IV able to copy and bypass these lesions more effectively and much more accurately than pol V (Fig. 2 and Tables II and III). Thus, pol IV correctly incorporates C opposite both R and S isomers of G-BaP DE, with error frequencies typically in the range of 10 Ϫ1 to 10 Ϫ4 , in contrast to pol V, which favors misincorporation of A compared with the correct incorporation of C opposite the same two G-BaP DE isomers in context III by factors of 350 and 132 (see above). The accurate and efficient incorporation opposite both G-BaP DE isomers in both sequence contexts is mirrored by the ability of pol IV to extend DNA containing C⅐G-BaP DE base pairs in a similar manner. With C⅐G-BaP DE correct base pairs, extension occurred with efficiencies (apparent V max /K m values) and specificities (f 0 ext ) on the order of 10 2 -to 10 4 -fold higher than for any of the mispairs (Table III).
Both A-BaP DE-S and -R adducts were also copied relatively accurately by pol IV, with f inc values ranging from a high of 0.12 for C⅐A-BaP DE-S to a low of 2 ϫ 10 Ϫ3 for A⅐A-BaP DE-S (Table II). A curious property of pol IV is its ability to extend A⅐A-and G⅐A-BaP DE mispairs 10 -100-fold more efficiently than T⅐A-BaP DE correct pairs for the R isomer and to extend correct and incorrect base pairs with roughly equal efficiency for the S isomer (Tables II and III). However, their ease of extension is probably of little consequence because these mispairs are made with such low efficiency by pol IV (Table II). These data are consistent with the observation that pol IV effectively carried out TLS beyond G-but not A-BaP DE lesions (cf. Fig. 2).
pol II exo Ϫ Nucleotide Incorporation and Extension Fidelity at BaP DE Template Adducts-The in vitro behavior of pol II is perhaps less "interesting" compared with pol V and IV because pol II shows considerably less ability to bypass any of the BaP DE adducts. Even so, there are several kinetic points of interest. For example, pol II favored, by 380-and 36-fold, the misincorporation of G opposite G-BaP DE in sequence context III for the S and R isomers, respectively (Table II). However, the opposite was true in sequence context IV, where the correct incorporation of C opposite G-BaP DE was favored by ϳ25and 17-fold for the S and R isomers, respectively (Table II).
The G⅐G-BaP DE misincorporation catalyzed by proofreading-defective pol II is clearly attributable to template slippage, with correct incorporation of G occurring opposite the template C immediately downstream from the G-BaP DE adduct in context III (Fig. 1). In accordance with the "dNTP-stabilized" primer-template misalignment mechanism (46,47), the apparent K m for misincorporating G opposite G-BaP DE (ϳ270 M for both S and R isomers) (data not shown) is much lower than for the correct incorporation of C (ϳ1400 M for both isomers) (data not shown). In context IV, however, where both direct incorporation opposite the adduct and/or incorporation opposite a misaligned template G downstream from the lesion can occur, the apparent K m values are ϳ2-fold lower for C⅐G-BaP DE base pairs compared with G⅐G-BaP DE mispairs for both S and R isomers (data not shown).
The G⅐G-BaP DE-R mismatch in context III was extended ϳ2-fold better than the correct C⅐G-BaP DE-R (Table III). Unexpectedly, an A⅐G-BaP DE-R mismatch, which was formed 19-fold more efficiently than C⅐G-BaP DE-R by direct misincorporation of A (K m Ͼ 1 mM) (data not shown), was extended 27-fold better than either G⅐G-or C⅐G-BaP DE-R (Table I). Yet, the incorporation of A opposite the G-BaP DE-S isomer was not detectable, nor was a preformed A⅐G-BaP DE-S mispair extendable in sequence context III (Tables II and III). Proofreadingdefective pol II copied A-BaP DE adducts somewhat accurately, with the highest error rates being 8% for incorporation of A opposite A-Ba PDE-S and 5.7 and 4.1% for incorporation of G opposite the A-BaP DE-S and -R isomers, respectively (Tables  II and III). Extension of the correct pair was favored over extension of the three mispairs for both isomers (Tables II and  III). BaP DE Lesion Bypass in Vivo in pol II-, pol IV-, and pol V-null Mutant Backgrounds-We measured the ability of E. coli to produce progeny virus after transfection with BaP DE-S-adducted M13 DNA using sequence context IV for the G adduct and sequence context I for the A adduct (Table IV). The G-BaP DE-S adduct appeared to be bypassed ϳ12-fold more efficiently than the A-BaP DE-S adduct in non-SOS-induced wild-type cells, where ϳ9% of the input G-BaP DE-adducted DNA gave rise to M13 plaques compared with ϳ0.7% for the A-BaP DE-adducted DNA (Table IV). SOS induction in wildtype cells increased the yield of plaques from the A-BaP DE adduct by 8-fold (to 5.7%) and the yield from the G-BaP DE adduct by an additional 2-fold (to 18.9%) There were no discernible differences in A-BaP DE yield in the absence of any individual SOS polymerase.
In contrast to the lack of a specific polymerase effect on A-BaP DE-S survival, the absence of pol IV in SOS-induced cells reduced the G-BaP DE-S survival from ϳ20 to 5.4% (Table  IV). These in vivo data are consistent with the in vitro data showing that pol IV was by far the best at bypassing the G-BaP DE-S adduct in context IV (Fig. 2). In the absence of SOS induction, deletion of pol II had a pronounced effect on G-BaP DE DNA survival, reducing the yield from 8.7 to ϳ0.4% (Table  IV). pol II is present at roughly 50 molecules/cell (48), suggesting that this constitutive level of pol II may be sufficient to bypass the G-BaP DE-S adduct in the absence of SOS induction in vivo. Consistent with the in vivo data are in vitro data showing that proofreading-proficient pol II was able to bypass this adduct (Fig. 2). In contrast, when SOS was turned on, the absence of pol II had no discernible effect on G-BaP DE DNA progeny plating efficiency (Table IV), suggesting that surely pol IV and perhaps pol V then bypass the lesion. A clearly surprising result, however, is that in the absence of SOS induction, the G-BaP DE-S adduct appeared to be bypassed equally poorly in the absence of either pol V (0.3%) or pol II (0.4%) (Table IV). One would presume that the presence of pol II in the pol V-null background should result in at least a 1.4% survival, as observed in the pol IV-null background (Table IV). Perhaps these data hint at the interesting possibility that interactions between these polymerases may be needed for their ability to function in vivo. A three-dimensional illustration of the in vitro BaP DE adduct bypass efficiencies of the SOS polymerases is shown alongside the in vivo survival of transfected adductcontaining DNA in polymerase mutant backgrounds in Fig. 3. The most striking results are the 4-fold reduction in survival when replicating a G-BaP DE-S adduct in SOS-induced cells lacking pol IV and the loss of ability to copy this same adduct in SOS-non-induced cells when any one of the SOS polymerases was missing ( Fig. 3B and Table IV). DISCUSSION The related UmuC/DinB/Rev1/Rad30 polymerases, recently renamed the Y-family of DNA polymerases (3), appear to be earmarked for use in specialized DNA synthesis reactions, including copying damaged DNA, rescuing stalled replication forks, meiosis-associated DNA repair, and somatic hypermutation (6,49,50). The relationship between the in vivo function and biochemical properties of pol V has been investigated extensively with respect to its role during SOS mutagenesis in E. coli (33,35,39,41). pol V, acting in conjunction with RecA, SSB, and ␤⅐␥ processivity proteins, is primarily responsible for copying UV-damaged DNA such as TT cis-syn photodimers and TT (6-4) photoproducts (33) and abasic template lesions (33,39,41,51), with fidelities corresponding to in vivo mutational spectra for these three lesions (35). There are two additional SOS-induced DNA polymerases, pol IV, another Y-family member encoded by dinB (52), and pol II (17)(18)(19). pol IV causes frameshift errors in phage (53) and is responsible for generating "adaptive" frameshift mutations in non-dividing E. coli (12,13), whereas pol II plays a pivotal role in error-free replication restart in UV-irradiated cells (21) and also acts to reduce adaptive mutation levels (14,54). Thus, pol IV and II are believed to carry out specialized reactions principally involving undamaged DNA, whereas the role of pol V seems restricted to copying damaged DNA. It is crucial, however, to avoid "pigeonholing" the SOS polymerases into categories of those that copy either damaged or undamaged DNA, especially because so little is known concerning their actual cellular occupations. Although the genetic (55)(56)(57) and biochemical (35) data support a primary, perhaps exclusive role for pol V in copying UV-damaged DNA and abasic lesions, different polymerases might well be used to copy other types of damaged or modified template bases. Previous exam-ples that support this point in E. coli include the use of pol II to copy abasic sites in the absence of heat shock proteins (22) and the involvement of both pol II and V (but not pol IV) in copying an N-acetyl-2-aminofluorene G adduct (16). In a reversal of logic, it turns out that the high fidelity pol II generates Ϫ2 frameshifts during TLS of acetylaminofluorene G, whereas the low fidelity pol V is responsible for error-free lesion bypass (16). Most closely related to our experiments are data showing that  Table I. The column heights represent the in vitro BaP DE TLS percentage catalyzed by pol II, IV, and V (with ATP or ATP␥S). B, plating efficiencies of M13 DNA containing a site-specific BaP DE adduct in wild-type E. coli compared with pol II-, pol IV-, and pol V-null mutant strains with or without SOS induction (from Table IV). pol IV and V carry out error-free and Ϫ1 frameshift TLS, respectively, when copying a G-BaP DE-adducted template in vivo (16).
SOS Polymerase-catalyzed BaP DE Adduct Bypass Efficiencies-G-BaP DE adducts were bypassed most efficiently by pol IV in vitro, irrespective of isomer and sequence context, with TLS efficiencies ϳ3-10-fold higher than those for pol V (Fig. 3A and Table I). pol II was able to bypass G-BaP DE, albeit with poor efficiency, in just one of the two sequence contexts (context IV) (Fig. 3A and Table I). The TLS efficiency reflects the fraction of primer molecules that are extended beyond the lesion. The biological relevance of these in vitro observations is supported by the observation that when transfected into pol IV (⌬dinB)-null mutant cells induced for SOS, G-BaP DE-S-adducted M13 DNA gave rise to 4-fold fewer plaques relative to wild-type cells or compared with null mutants in either pol II (⌬polB) or pol V (⌬umuDC) (Fig. 3B and Table IV).
pol IV, which bypassed G-BaP DE adducts with efficiencies between 36 and 71%, did not significantly bypass A-BaP DE adducts, nor did pol II (Fig. 3A and Table I). pol V, however, bypassed both A-and G-BaP DE adducts with roughly similar efficiencies, which were lower than that of pol IV for the G adducts ( Fig. 3A and Table I). The A-BaP DE-S adduct was bypassed ϳ3-fold less efficiently than G-BaP DE-S in SOSinduced cells in vivo ( Fig. 3B and Table IV). However, no difference was observed in the "survival" of A-BaP DE-adducted M13 DNA in any of the wild-type or polymerase-deficient mutant backgrounds ( Fig. 3B and Table IV). Consequently, no clear conclusion can be drawn concerning which polymerase or polymerases might be involved in replicating past A-BaP DE-S. pol I showed a weak ability to bypass the A-BaP DE-S adduct, so perhaps it might play a role in vivo, even though it was not as active in bypassing this lesion compared with pol V in vitro (Fig. 2, lanes 1 and 6). It has previously been shown that proofreading-proficient and -deficient forms of pol I KF exhibit different patterns of TLS and misincorporations when copying A-BaP DE adducts depending on the stereochemistry at C-10 (58). pol III holoenzyme was essentially unable to bypass any of the BaP DE adducts (Fig. 2,  lane 3).
SOS Polymerase Misincorporation and Extension Fidelity at BaP DE Adducts-Not only did pol IV copy G-BaP DE adducts more efficiently compared with pol V, but it did so much more accurately. Thus, pol IV favors incorporation of C opposite R and S isomers of BaP DE by between 2 and 4 orders of magnitude, whereas pol V favors formation of A⅐G-BaP DE mispairs by factors of 130 -350 (Table II). The extension efficiencies for the two enzymes mirror the incorporation efficiencies. pol IV favored extension of C⅐G-BaP DE correct base pairs by 1000 -10,000-fold over any of the mispairs (Table III). True to form, pol V favored extension of the favored A⅐G-BaP DE mispairs over the C⅐G-BaP DE correct pairs (Table III).
The bottom line is that the action of pol V on G-BaP DE adducts would be expected to generate G⅐C to T⅐A transversions, whereas few base substitution mutations would be expected to occur when G-BaP DE adducts are copied by pol IV. We made a preliminary estimate of the types and frequency of mutations by sequencing a subset of M13 progeny containing G-BaP DE-S adducts in context IV (Fig. 1) in the different polymerase-null mutant backgrounds that were SOS-induced. The in vitro kinetic data and in vivo mutational data agree with respect to the "strongest" prediction: the frequency of G⅐C to T⅐A mutations increases by ϳ7-fold in a pol IV-null mutant background. With the G-BaP DE-S adducts, G⅐C to T⅐A mutations occur with a frequency of ϳ1% in the SOS-induced wildtype strain (two mutants/176 DNA molecules sequenced) com-pared with 6.8% in the absence of pol IV (three mutants/44 DNA molecules sequenced).
Although these numbers are admittedly small, the conclusions are supported by data in the pol II-null background, where the G to T transversion frequency is again 1% (one mutant/104 DNA molecules sequenced). G to T error frequencies have an intermediate value of ϳ3.1% in the pol V-null background (3/98). These errors are perhaps made by pol II, which forms A⅐G-BaP DE-S mispaired intermediates at an elevated frequency of ϳ7% in context IV (although the favored pol II mispair is with T at a frequency of ϳ14%) ( Table II).
Because of the small numbers and the obvious complexities of attempting to make a direct comparison of the in vitro and in vivo data, e.g. the inability to take into account the effects of proofreading and post-replication mismatch repair (acting on double-stranded M13 replication intermediates), no strong conclusions can be drawn concerning which of the polymerases is responsible for causing specific types of mutations in vivo. Despite this caveat, however, several conclusions can be drawn from the data. The most important conclusion, from a biological perspective, is that the three SOS polymerases (pol II, IV, and V) are each likely to play some role in bypassing BaP DE adducts. Based on both the in vivo and in vitro kinetic data, pol IV copies and bypasses G-BaP DE adducts with highest efficiency and accuracy. In contrast, the errant pol V copies G-BaP DE adducts with remarkable inaccuracy and furthermore favors extension of the resultant mispairs over correct pairs. The most exceptional errors are an A⅐G-BaP DE-S mispair, favored by 350-fold over a C⅐G-BaP DE-S correct pair, and an A⅐G-BaP DE-R mispair, favored by 132-fold over its correctly paired counterpart. Both of these mispairs are selectively extended by pol V with efficiencies of 23-and 3.9-fold over the corresponding correct pairs. In contrast to its ability to catalyze G-BaP DE TLS efficiently, pol IV is essentially unable to bypass R and S isomers of A-BaP DE. On the other hand, pol V copies A-BaP DE adducts efficiently and relatively "accurately," making A⅐A-BaP DE-R mispairs with a frequency of only 23% and A⅐A-BaP DE-S mispairs with a frequency of 8.4%. Extension of both of these mispairs by pol V occurs with efficiencies similar to those of extension of T⅐A-BaP DE correct pairs. Thus, the kinetic data support a growing body of evidence that although each of the SOS polymerases appears to have a primary role to play in the cell (pol V for error-prone TLS on heavily damaged DNA, pol IV for relieving blocked replication forks on undamaged DNA, and pol II for catalyzing error-free replication restart), each enzyme can also play a role in copying various types of DNA template damage.