Responses to the Major Acrolein-derived Deoxyguanosine Adduct inEscherichia coli *

Acrolein, a reactive α,β-unsaturated aldehyde found ubiquitously in the environment and formed endogenously in mammalian cells, reacts with DNA to form an exocyclic DNA adduct, 3H-8-hydroxy-3-(β-d-2′-deoxyribofuranosyl)-5,6,7,8-tetrahydropyrido[3,2-a]purine-9-one (γ-OH-PdG). The cellular processing and mutagenic potential of γ-OH-PdG have been examined, using a site-specific approach in which a single adduct is embedded in double-strand plasmid DNA. Analysis of progeny plasmid reveals that this adduct is excised by nucleotide excision repair. The apparent level of inhibition of DNA synthesis is ∼70% in Escherichia coli ΔrecA, uvrA. The block to DNA synthesis can be overcome partially byrecA-dependent recombination repair. Targeted G → T transversions were observed at a frequency of 7 × 10−4/translesion synthesis. Inactivation ofpolB, dinB, and umuD,C genes coding for “SOS” DNA polymerases did not affect significantly the efficiency or fidelity of translesion synthesis. In vitroprimer extension experiments revealed that the Klenow fragment of polymerase I catalyzes error-prone synthesis, preferentially incorporating dAMP and dGMP opposite γ-OH-PdG. We conclude from this study that DNA polymerase III catalyzes translesion synthesis across γ-OH-PdG in an error-free manner. Nucleotide excision repair, recombination repair, and highly accurate translesion synthesis combine to protect E. coli from the potential genotoxicity of this DNA adduct.

The ring structure of exocyclic DNA adducts prevents the formation of normal Watson-Crick hydrogen bonds. These adducts miscode in in vitro primer extension studies conducted with purified DNA polymerases (1, 6 -9) and induce point mutations in Escherichia coli and mammalian cells (1, 10 -17). Exocyclic etheno adducts are thought to be responsible for the mutations observed in the tumor suppressor p53 gene of humans and animals exposed to vinyl chloride (18,19).
1,N 2 -Propanodeoxyguanosine adducts appear to be ubiquitous in cellular DNA and may contribute to so-called spontaneous mutagenesis, thereby playing a role in aging and cancer. In this paper, we establish a genotoxic mechanism for ␥-OH-PdG in E. coli and describe the processing of this adduct in bacterial cells.
We have developed a novel experimental approach that allows us to explore cellular responses to DNA adducts at a mechanistic level (23). The plasmid vector used in this study employs strand-specific markers tagged with mismatches. An oligonucleotide containing a single DNA adduct is incorporated into heteroduplex (HD) DNA containing short stretches of mismatches at several locations. The modified DNA is introduced into mismatch repair-deficient hosts, and adduct-related events are measured. Progeny plasmid are analyzed for their marker sequences. Linkage analysis of marker sequences allows us to group progeny according to in vivo processing events, * This research was supported by United States Public Health Services Grants CA76163 (to M. M.) and PO1CA47995 (to A. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Both authors contributed equally to this work. including excision repair, translesion DNA synthesis, and recombination repair.
Oligodeoxyribonucleotides-The synthesis of oligodeoxyribonucleotides containing ␥-OH-PdG or 1,N 2 -(1,3-propano)-2Ј-deoxyguanosine (PdG) (IV in Fig. 1) has been described previously (25,26). Oligonucleotides containing the precursor of ␥-OH-PdG (N 2 -[3,4-dihydroxybutyl] dG) or PdG were purified twice by high pressure liquid chromatography using a Waters Bondapak C18 column (3.9 ϫ 300 mm) and acetonitrile in 100 mM triethylammonium acetate buffer, pH 6.8, at a flow rate of 1 ml/min with the dimethoxytrityl group (DMT) on and off. The acetonitrile gradient was 16 -33% with DMT-on DNA and 0 -20% with DMT-off DNA and applied over 40 min. To generate ␥-OH-PdG, oligonucleotides containing N 2 -[3,4dihydroxybutyl] dG were incubated in 100 mM sodium periodate for Ͼ6 h at room temperature (25). ␥-OH-PdG-containing oligonucleotides were separated from the parental oligonucleotides by high pressure liquid chromatography as described above for DMT-off DNA. All oligonucleotides were subjected to electrophoresis in a denaturing 20% polyacryl-amide gel. Bands were detected by UV shadowing. Oligonucleotides were excised from the gel then purified over a Sep-Pak column (Waters). Purified oligonucleotides were subjected to electrospray mass spectrometry analysis. The results were as follows: 13- Table I; MO strains were constructed by P1 transduction (27) and are mismatch repair-deficient. Plasmids pS and pA have been described (23). These plasmids differ in DNA sequence at three regions but are otherwise identical; hence, HD DNA prepared from these plasmids contains three mismatched regions (see Fig. 3). A mismatched region involving a BamHI site is located 150 nucleotides upstream from the adduct site, the others, containing NheI and SpeI/AatII sites, are located ϳ220 and 3100 nucleotides, respectively, downstream from the adduct (see Fig. 3). As described later, these mismatches serve as strand-specific markers.
Construction of HD DNA Containing a Single ␥-OH-PdG Adduct-The scheme for this construction is shown in Fig. 2. Detailed procedures have been described previously (19). In brief, double-stranded (ds) pA was digested with EcoRV (Step I), and the linearized plasmid DNA was ligated to a blunt-ended duplex 13-mer, 5Ј-d(AGGTACGTAGGAG)/ 3Ј-d(TCCATGCATCCTC), containing a SnaBI site (5Ј-TACGTA) (Step II). Two constructs, each containing a single insert with opposite orientation, were isolated; one of these, pA106, was used in this study.
Single-stranded (ss) pA106 was mixed with EcoRV-digested ds pS (Steps III and IV), treated with NaOH to denature ds pS, then neutralized to form ds DNA. Circular ss pA106 and its complementary strand (derived from ds pS) were annealed to form HD DNA containing a 13-nucleotide gap. An unmodified or modified 13-mer (3Ј-d(TCCATAX-CTCCTC), where X is dG or ␥-OH-PdG) phosphorylated at the 5Јtermini using T4 polynucleotide kinase and ATP, was annealed into the gap and ligated by T4 DNA ligase. The 13-mers are not fully complementary to the gap sequence, because they form mismatches at and adjacent (5Ј and 3Ј) to the adduct site (Fig. 3). These mismatches, located opposite the SnaBI site, also serve as a strand-specific marker. The ligation mixture was treated with SpeI and EcoRV to remove residual ds pS. Closed circular ds DNA was purified by ultracentrifugation in a CsCl/ethidium bromide solution. DNA was concentrated by Centricon 30 (Amicon, Beverly, MA), and the concentration was determined spectrophotometrically.
In this HD construct, six and three base mismatches are formed at the SpeI/AatII and SnaBI sites, respectively. At the BamHI and As MV1932, but mutS201ϻTn5 (23) MO934 As MO933, but uvrA6, malE3ϻTn10 (23) MO936 As MO934, but mal ϩ (23) MO937 As MO936, but ⌬(recA-srl)306ϻTn10 (NheI site), and D/d (AatII/SpeI site). The unmodified complementary strand derived from ss pA106 contains the A-B-C-D linkage (Fig. 3). ␥-OH-PdG was incorporated into the site of b on the strand bearing the a-b-c-d linkage; this strand is the template for leading strand synthesis.
Transformation of E. coli and Analysis of Progeny Plasmid-Unmodified or modified DNA (12 ng) was introduced into mismatch repairdeficient (mutS) MO strains (50 l of electrocompetent cells) by electroporation. The mutS mutation assures that the mismatches will not be repaired. 2ϫ YT medium 16 g of tryptone, 10 g of yeast extract, 5 g/1000 ml NaCl, pH 7) (950 l) was added to the electroporation mixture, then incubated for 20 min at 37°C. A portion (10 -50 l of a 100ϫ dilution) of the mixture was plated onto a 1ϫ YT-ampicillin (100 g/ml) plate to determine the number of transformants in the mixture. The remaining mixture was incubated for additional 40 min and then added to 10 ml of 2ϫ YT containing ampicillin. After culturing overnight, progeny plasmid was prepared by the alkaline lysis method and used to transform E. coli DH5␣ (Life Technologies, Inc.). This second transformation segregates progeny plasmid derived from each strand of the HD DNA. Transformants were inoculated individually in 96-well plates and cultured for several hours. Bacterial cultures were stamped onto filter paper placed on a 1ϫ YT-ampicillin plate and cultured overnight. The filter was treated with 0.5 M NaOH for 11 min, neutralized in 0.5 M Tris-HCl, pH 7.4, for 7 min, washed with 1ϫ SSC (150 mM NaCl, 15 mM Na 3 citrate, pH 7.2) then with ethanol, and baked at 80°C for 2 h. Differential oligonucleotide hybridization (16,23,28) using the 32 P-labeled probes shown in Fig. 3 was employed to detect strand-specific marker sequences and the base located at the position of the adduct. L and R probes were used to confirm the presence of the 13-mer insert.
In several experiments, mitomycin C was used to induce SOS functions in E. coli. Overnight cultures of MO strains were diluted 20-fold with prewarmed 2ϫ YT and cultured for 2 h. Mitomycin C (1-10 g/ml) was added to the cultures and incubated at 37°C for 30 min with shaking. Cells were prepared for electroporation by repeated washings with H 2 O.
Interpretation of Results-When a HD construct bearing a single ␥-OH-PdG residue is introduced into a host cell, various events may occur (Fig. 4). If the adduct is removed by excision repair before being replicated, the immediate 5Ј-and 3Ј-flanking mismatches also are removed. Gap-filling DNA synthesis converts 5Ј-CXA (b) to 5Ј-ACG (B); therefore, progeny derived from the repaired strand contain the linkage of a-B-c-d (progeny III) (Fig. 4, step 1). When the construct is replicated in the absence of DNA repair, progeny produced from the unmodified strand contain the A-B-C-D linkage (step 2), whereas those from the modified strand have a-b-c-d (step 4) following translesion synthesis (TLS). When DNA synthesis is blocked by the adduct, the block may be overcome by UmuDЈ 2 C/RecA-assisted TLS or recombination repair (daughter strand gap repair) (29). In recombination repair, the ss gap is filled by strand transfer from the unmodified parental strand (step 5). The 3Ј-end of the blocked nascent strand is used to replicate the transferred region (step 6). These processes create a Holliday junction. When the Holliday junction is migrated by the RuvA⅐RuvB complex, a Holliday junction-specific helicase, and resolved by RuvC, a Holliday junction-specific endonuclease (step 7), progeny with the linkage of A-B-C-D (progeny I), a-B-C-d (progeny IV), a-B-C-D (progeny V), and A-B-C-d (progeny VI) are created.
In Vitro Primer Extension Studies-Oligonucleotides used for primer extension studies are as follows: 28-mer templates, 5Ј-CTGCTCCTCX-ATACCTACACGCTAGAAC, where X is dG, ␥-OH-PdG, or PdG; an 18-mer primer for incorporation experiments, 5Ј-GTTCTAGCGTGTAG-GTAT; 19-mer primers for extension experiments, 5Ј-GTTCTAGCGT-GTAGGTATN, where N is dA, dG, dC, or dT; and a 16-mer primer for read-through experiments, 5Ј-GTTCTAGCGTGTAGGT. All oligonucleotides were purified by electrophoresis in a denaturing 20% polyacrylamide gel. The 5Ј-32 P-end-labeled primers were hybridized to templates at a molar ratio of 1:1.2 in a buffer containing 50 mM Tris-HCl, pH 7.5, 5 mM EDTA, and 500 mM NaCl. Annealing reactions were conducted by heating at 70°C for 5 min followed by slow cooling.
Reaction mixtures ( mg/ml bovine serum albumin, and 10% glycerol. Reactions were conducted at 25°C for 30 min and stopped by adding 10 l of formamide dye mixture (90% formamide, 0.001% xylene cyanol, 0.001% bromphenol blue, 20 mM EDTA). Samples were heated at 95°C for 5 min, and aliquots (1.5 l) were subjected to electrophoresis in a denaturing 20% polyacrylamide gel (0.4 mm thick, 40 cm long) at 2600 V for 3 h. Gels were analyzed by a PhosphorImager using ImageQuaNT software (Molecular Dynamics). Fig. 4 predicts that progeny III derived from plasmids subjected to excision repair will contain the a-B-c-d linkage. The unmodified control construct, which contained three base mismatches and no adduct, yielded Ͻ1% of progeny III in the uvrA (MO937) and uvr ϩ (MO939) strains (Table II), indicating that the mismatches were not subjected to nucleotide excision repair. When the adducted construct was introduced into the uvrA strains (MO937 and MO934), progeny III accounted for Ͻ3% of plasmids analyzed. When the same construct was introduced into uvrϩ strains (MO939 and MO933), fractions of progeny III ranged from 32 to 35%. Thus, ␥-OH-PdG is a substrate for UvrABC-catalyzed nucleotide excision repair in the presence of the 5Ј-and 3Ј-flanking mismatches.

␥-OH-PdG Is a Substrate for Nucleotide Excision Repair-Pathway 1 depicted in
␥-OH-PdG Adduct Inhibits DNA Synthesis in E. coli-The HD constructs were introduced into MO937 (⌬recA, uvrA, alkA1, tag1). With the unmodified control construct, more than 96% of progeny were derived from replication of both strands (progeny I and II, Table II). The ratio of progeny I to II, however, was not 1:1 but twice that of progeny I. This result is consistent with that of a previous study (23). Using the modified construct, the number of progeny I derived from the unmodified strand (78%) markedly exceeded that of progeny II derived from TLS (17%). This result indicates that ␥-OH-PdG is a strong but not complete block to DNA synthesis. The apparent efficiency of TLS is estimated to be 27% (17/64). The induction of SOS functions did not increase significantly the fraction of progeny II in MO934. The fractions of progeny II obtained in the uvrϩ strains (MO939 and MO933) were lower than those in the uvrA strains (MO937 and MO934) due to efficient repair of the adduct by nucleotide excision repair.
TLS across ␥-OH-PdG Is Essentially Error-free-The adducted HD construct was introduced into MO933 ("wild"), MO934 (uvrA), MO220 (mutD5), and MO221 (mutD5, ⌬umuDC) in the presence or absence of induced SOS functions. Initial numbers of transformants in the transformation mixtures were determined by plating a portion of the transformation mixture. This analysis revealed Ͼ1 ϫ 10 6 and Ͼ1 ϫ 10 4 transformants per transformation in the absence and presence of induced SOS functions, respectively. The lower transformation efficiency of mitomycin Ctreated cells is thought to be due to DNA damage. Progeny plasmid DNA was purified following overnight culture of the transformation mixture in the presence of ampicillin and then digested with SnaBI, the site for which is located in marker B (Fig. 3). This digestion removes progeny derived from the unmodified strand and plasmids subjected to excision repair or recombination repair (see Fig. 4), facilitating analysis of TLS events. Targeted events were analyzed by differential oligonucleotide hybridization and DNA sequencing. This analysis revealed that almost all targeted events were ␥-OH-PdG 3 dG, indicating accurate TLS. Only one ␥-OH-PdG 3 dT transversion was observed among 282 transformants of SOS-uninduced MO933, yielding a miscoding frequency of 0.35%. The numbers of transformants analyzed were 144 for SOS-induced MO933; 190 each for SOS-induced and -uninduced MO934 and MO220; and 144 each for SOS-induced and -uninduced MO221. Targeted mutations were not observed in these strains. When data for the four strains are combined, the frequencies of targeted mutations are 0.12% and Ͻ0.15% in the absence and presence of induced SOS functions, respectively. Thus, TLS across ␥-OH-PdG is highly accurate in E. coli.
DNA Polymerase III and/or pol I Catalyze Accurate TLS-To address the question of which DNA polymerase(s) is (are) responsible for the TLS, we constructed MO233, a strain that lacks all "SOS DNA polymerases" such as pol II (polB), pol IV (dinB), and pol V (umuDC), in addition to the mutS and uvrA genes. MO234 is a control strain that expresses all SOS DNA polymerases. The degree of TLS (number of progeny II) is not significantly different in MO233 and MO234 in the presence or absence of induced SOS functions (Table III). The analysis of targeted events revealed a single ␥-OH-PdG 3 dT transversion in SOS-induced MO233, suggesting that pol III or pol I is responsible for this mutation. These results suggest that non-SOS DNA polymerase(s) conduct(s) TLS across ␥-OH-PdG with high fidelity.
DNA Polymerase I-catalyzed TLS Is Diminished and Errorprone-Incorporation studies using pol I (KF exo Ϫ ), in which a FIG. 3. Nucleotide sequence of regions containing strand-specific markers and probes used for oligonucleotide hybridization. The top strand is ss pA106, and the bottom strand is derived from EcoRV-treated ds pS. Marker sequences in the sequence-specific probes A-D (overscored) and a-d (underlined) are highlighted. L and R probes (underlined) were used to identify progeny containing the 13-mer insert. Probes SG, ST, SA, SC, and SD were used to determine which base replaced ␥-OH-PdG. Ϫ, direct connection of bases: e.g. G-G is GG; ϳ, sequence interruption. Approximate distances between markers are shown in parentheses.

TABLE III
Linkage analysis in SOS DNA polymerase-deficient E. coli primer terminus was located one base 3Ј from the adduct site and a single dNTP was present, showed that this polymerase inserts dAMP and dGMP and, to a much lesser extent, dCMP and dTTP opposite ␥-OH-PdG: Frequencies of incorporation were 93, 88, 7, and 5% for dATP, dGTP, dCTP, and dTTP, respectively, in a reaction using 85 nM KF exo Ϫ (Fig. 5A). Primer extension studies, in which a primer terminus was located opposite ␥-OH-PdG and all four dNTPs were present, showed that dAMP, dCMP, and dGMP termini were extended similarly and the dTMP terminus was extended poorly: Yields of extended products were 63, 57, 44, and 18% with the dA, dC, dG, and dT termini, respectively, in the reaction containing 85 nM enzyme (Fig. 5B). These results suggest that pol I promotes error-prone DNA synthesis across ␥-OH-PdG. When the readthrough experiments were conducted using a 16-mer primer in the presence of all four dNTPs, doublet bands, caused by differing migration rates due to the difference in the nucleotide inserted, were observed opposite the adduct (X) and the next base (C) (Fig. 5C). This result indicates that different nucleotides are inserted opposite this adduct and then extended. Effects of the nucleotide difference is obscured as the length increases. This result is consistent with those of incorporation and extension studies described above. Because dCMP is not the preferred nucleotide inserted, TLS catalyzed by this polymerase is likely to be error-prone.
Extension of primers beyond ␥-OH-PdG or PdG was observed at 85 and 170 nM KF exo Ϫ (Fig. 5C). The fractions of primers extended beyond these modified bases are 36% at 85 nM and 69% at 170 nM for ␥-OH-PdG and 11% at 85 nM and 23% at 170 nM for PdG. These results indicate that PdG is more inhibitory than ␥-OH-PdG.
Blockage of DNA Synthesis Is Rescued by Recombination Repair-When the ␥-OH-PdG-containing HD construct was introduced into MO933, MO934, MO233, or MO234, the relative fractions of progenies IV, V, and VI increased. The sum of these three types of progeny ranged from 8 to 19% in the uvrA strains (MO934, MO233, and MO234). In the uvr ϩ strain (MO933), percentages were 7 and 5% in the absence and presence of induced SOS functions, respectively. The lower values may reflect the substantial repair of the adduct in this uvr ϩ strain. When the modified HD construct was introduced into the ⌬recA strains (MO937 and MO939), fractions of the three types of progeny were low: 2.3% in MO937 and 1% in MO939. In a previous report (23), mismatches were shown not to increase the number of recombinants. Therefore, the increases observed can be ascribed to a RecA-dependent recombination repair mechanism (daughter strand gap repair) in response to the DNA synthesis block caused by the adduct (Fig. 4). These increases are consistent with the general finding that ␥-OH-PdG inhibits DNA synthesis. DISCUSSION In this report, we studied the genotoxicity of ␥-OH-PdG in E. coli and the response of this organism to the presence of the adduct. Our results indicate that ␥-OH-PdG is a good substrate for nucleotide excision repair; in this respect, it is similar to the structurally related PdG and malondialdehyde-derived propeno-dG adduct, pyrimido[1,2-a]purin-10(3H)-one (M 1 G) (III in  Tables II and  III. Excision repair of the adduct (step 1) converts the sequence of the adducted region to B; replication of a repaired molecule produces progeny I and III. When modified DNA replicates (steps 2-4), the unmodified strand produces progeny I and the modified strand yields progeny II following TLS (step 4). When the adduct inhibits DNA synthesis, daughter-strand gap repair (steps 5-7) operates to overcome the inhibition. This mechanism, involving strand transfer (step 5), formation of a Holliday junction, gap-filling synthesis (step 6), branch migration, and resolution of a Holliday junction by RuvC resolvase (step 7), produces progeny I, IV, V, and VI.   (30). ␥-OH-PdG inhibits DNA synthesis; the apparent efficiency of TLS is 27% in the absence of induced SOS functions. The inhibition of DNA synthesis created by ␥-OH-PdG is partly overcome by daughter strand gap repair (recombination repair) as determined by the increase in the number of recombinants of progenies IV, V, and VI (Tables II and III). Because the parental unmodified strand is used to fill in a gap generated as a result of the synthesis block (refer to Fig. 4), this repair is mechanistically accurate and therefore contributes to an error-free recovery from DNA synthesis block.
The frequency of progeny II derived from the TLS pathway accounts for Ͼ30% among all progeny in the strain (MO233) lacking all the SOS DNA polymerases (pol II, pol IV, and pol V), and the SOS induction did not significantly increase the number of progeny II in MO233 or MO234, a control strain containing the SOS DNA polymerases (Table III). These results suggest that one or more constitutive DNA polymerases, pol III and/or pol I, play the major role in this TLS, which is highly accurate. Only two ␥-OH-PdG 3 T transversions were observed among the total of 2687 transformants of various E. coli strains obtained in the presence or absence of SOS induction. The overall targeted mutation frequency is 0.07%. Marnett and his colleagues also have observed the lack of mutagenicity of this adduct (accompanying article (39)). Because pol I is necessary for the replication of ColE1 origin-based plasmids (31), inactivation of this gene is not possible. Our in vitro primer extension studies show that this polymerase appears to catalyze error-prone TLS across ␥-OH-PdG (Fig. 5). pol I is believed to catalyze the filling of small gaps and the formation of DNA primers in E. coli. Therefore, we speculate that pol III, but not pol I, is responsible for this highly accurate TLS. The accurate TLS was also observed in the strain (MO220) carrying a mutation in the mutD (dnaQ) gene that codes for the pol IIIassociated 3Ј35Ј-exonuclease (⑀ subunit) (31). This exonuclease removes nucleotides from the 3Ј terminus when incorrect nucleotides are inserted during pol III-catalyzed DNA synthesis. Therefore, if this TLS is catalyzed by pol III, the result of the accurate TLS in this strain suggests that correct dCMP is inserted almost exclusively opposite ␥-OH-PdG and extended: The accurate TLS can be ascribed to the selection of the correct deoxyribonucleoside triphosphate, dCTP. Another possibility is that only ␥-OH-PdG:dC pair, but not other incorrect pairs, can be extended by pol III.
Results of this study are very different from those reported for the model acrolein adduct PdG, which strongly blocks DNA synthesis and efficiently induces targeted PdG 3 T and PdG 3 A mutations in E. coli and simian kidney cells (12,15). PdG presents a stronger block than ␥-OH-PdG when tested by primer extension experiments in vitro. When a 25-mer containing ␥-OH-PdG was incubated at room temperature with its complementary strand, two species migrating more slowly than the starting material appeared, accounting eventually for 5% of the total oligonucleotide. 2 This observation suggests that interstrand cross-linking has occurred, an event that would take place only if ␥-OH-PdG exists in a ring-open form similar to that reported for M 1 G paired with dC in duplex DNA (32). An NMR structural study (see accompanying article (40)) shows unequivocally that ring opening of ␥-OH-PdG occurs when the adduct is located opposite dC in duplex DNA, in a conformation allowing formation of Watson-Crick bonds. Because ␥-OH-PdG is located in a mismatched region in our HD DNA, the structure of the adduct cannot be stated with certainty; however, Watson Crick hydrogen bonding generated by ring opening of ␥-OH-PdG would ac-count fully for the enhanced translesion synthesis and relative lack of miscoding of the natural acrolein adduct.
In summary, our results show that the combination of nucleotide excision repair, accurate TLS, and daughter strand gap repair protects well E. coli from the genotoxicity of ␥-OH-PdG.