Evaluation of the Mutagenic Potential of the Principal DNA Adduct of Acrolein*

Acrolein is produced extensively in the environment by incomplete combustion of organic materials, and it arises endogenously in humans as a metabolic by-product. Acrolein reacts with DNA at guanine residues to form the exocyclic adduct, 8-hydroxypropanodeoxyguanosine (HOPdG). Acrolein is mutagenic, and a correlation exists between HOPdG levels in Salmonella typhimurium treated with acrolein and a resultant increase in mutation frequency. Site-specifically modified oligonucleotides were used to explore the mutagenic potential of HOPdG in Escherichia coli strains that were either wild-type for repair or deficient in nucleotide excision repair or base excision repair. Oligonucleotides modified with HOPdG were inserted into double-stranded bacteriophage vectors using the gapped-duplex method or into single-stranded bacteriophage vectors and transformed into SOS-induced E. coli strains. Progeny phage were analyzed by oligonucleotide hybridization to establish the mutation frequency and the spectrum of mutations produced by HOPdG. The correct base, dCMP, was incorporated opposite HOPdG in all circumstances tested. In contrast, in vitro lesion bypass studies showed that HOPdG causes misincorporation opposite the modified base and is a block to replication. The combination of these studies showed that HOPdG is not miscoding in vivo at the level of sensitivity of these site-specific mutagenesis assays.

The accumulation of DNA damage is believed to play a significant role in genetic diseases, including cancer. An important form of DNA damage results from the addition of electrophiles to form stable adducts to the bases. In the case of bifunctional electrophiles, cyclization can occur to form exocyclic adducts that obstruct Watson-Crick base pairing (1). These adducts are anticipated to be blocks to replication and, therefore, potential premutagenic lesions. This has been demonstrated to be the case with the exocyclic adducts, pyrimido [1,2-␣]purin-10(3H)-one (M 1 G) 1 and its structural analogue, 1,N 2propano-2Ј-deoxyguanosine (PdG) (2)(3)(4).
Acrolein, an ␣,␤-unsaturated aldehyde, is an example of such a bifunctional electrophile. Acrolein is produced exogenously as a product of organic combustion and endogenously as a product of lipid peroxidation (5)(6)(7). Acrolein also arises during cyclophosphamide metabolism (8). It is highly mutagenic to bacterial and mammalian cells and exhibits tumor-initiating activity (9 -11). The most likely mode of action for acrolein-induced mutagenicity is its ability to form adducts to DNA. The major adduct generated by the reaction of acrolein with deoxyguanosine residues in DNA is 8-hydroxypropanodeoxyguanosine (HOPdG), which is structurally related to M 1 G and PdG ( Fig. 1) (12,13). Acrolein has been shown to be mutagenic in Salmonella typhimurium, and a correlation has been observed between HOPdG levels in S. typhimurium treated with acrolein and the resultant increase in mutation frequency (14). The HOPdG adduct has been detected at relatively high levels in DNA from healthy human tissues using 32 P-postlabeling combined with high performance liquid chromatography (15,16).
A major obstacle in attempts to evaluate the mutagenic potential of HOPdG has been the generation of oligonucleotides containing site-specifically incorporated HOPdG. The base lability of HOPdG prevents its direct incorporation into DNA by standard synthetic protocols. Recently, both Khullar et al. (17) and Nechev et al. (18) described strategies for the stable, sitespecific incorporation of HOPdG into oligonucleotides using oxidative generation of the aldehyde function after the base deprotection step. This enables the construction of recombinant M13 genomes containing a single HOPdG adduct for use in mutagenesis assays in Escherichia coli. In the present study, we prepared double-stranded vectors containing HOPdG to evaluate its ability to block replication, to induce mutations, and to be repaired. In addition, single-stranded vectors were constructed to examine the ability of HOPdG to induce mutations in the absence of repair or of replication strand bias. The combination of these approaches revealed that HOPdG does not induce mutations at a level of sensitivity well below that needed to detect mutations caused by other structurally related exocyclic adducts. . Bio-spin 6 columns were purchased from Bio-Rad (Hercules, CA). GELase enzyme was from Epicentre Technologies (Madison, WI). Isopropyl ␤-Dthiogalactoside (IPTG) and 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) were purchased from Gold Biotechnology Inc. (St. Louis, MO).
The primers and all unmodified template oligonucleotides for in vitro bypass assays were prepared using an Applied Biosystems automated DNA synthesizer in the Vanderbilt University Molecular Toxicology Molecular Genetics Core. A 12-mer oligonucleotide used as a running start primer had the following sequence: 5Ј-CTGAGCTCGAGT-3Ј. The 13-mer standing start primer sequence was as follows: 5Ј-GAGCTC-GAGTCGA-3Ј. 16-mer oligonucleotides were also synthesized with the following sequence: 5Ј-CTGAGCTCGAGTCGAN-3Ј, where N refers to A, G, C, or T. Unmodified oligonucleotides were purified by electrophoresis on a denaturing 20% polyacrylamide gel. Oligonucleotide sizing markers (8 - (19). The E. coli strain JM105 was used as the host bacterium for the replication of the M13MB102-1 genome on indicator plates.
M13mp7L2 was a generous gift from Christopher Lawrence (University of Rochester, Rochester, NY).
Construction of Adduct-modified M13MB102-1 Genomes-Preparation of gapped-duplex DNA and ligation of adducted or unadducted 8-mers followed an adaptation of the method described by Benamira and Marnett (19). Double-stranded M13MB102-1 was linearized with SacII and BssHII followed by dialysis with a 12-fold excess of singlestranded M13MB102-1 in decreasing concentrations of formamide. The resultant gapped-duplex was resolved by 0.8% low melting point agarose gel electrophoresis. The band corresponding to gapped-duplex DNA was excised, and the DNA was recovered by GELase treatment. HOPdG and dG-containing 8-mers (100 pmol) were phosphorylated using T4 polynucleotide kinase and 1 mM ATP in the buffer supplied by New England BioLabs. For ligation, gapped-duplex DNA was added to each of the phosphorylated 8-mers, along with 400 units of T4 DNA ligase and 1 mM ATP. The reaction proceeded at 16°C overnight in ligation buffer supplied by New England BioLabs. Reaction mixtures were brought up to 100 l with H 2 O, and the DNA was purified by filtration through a modified polyvinylidene fluoride membrane from Millipore. Ligation products were purified on a 0.8% low melting point agarose gel and recovered using GELase enzyme.
Construction of M13mp7L2 Vectors Containing a Single HOPdG Adduct-Single-stranded M13mp7L2 genomes containing a unique site-specific HOPdG were constructed, following a modification of the procedure described by Banerjee et al. (20). Briefly, single-stranded M13mp7L2 DNA was linearized with EcoRI for 2 h at 25°C, utilizing the unique EcoRI site within the hairpin linker region. The linearized DNA was recircularized with a 48-mer oligonucleotide creating a circular single-stranded DNA containing an 8-base gap at the EcoRI site. The ends of the 48-mer oligonucleotide are complementary to the EcoRI site, and its internal sequence of 8 bases is complementary to the dGand HOPdG-modified 8-mers. Recircularization was carried out in 20 mM MOPS/2 mM MgCl 2 /50 mM NaCl buffer (pH 7.4), by heating at 90°C for 3 min and allowing the sample to cool slowly to room temperature overnight. A 100-fold molar excess of modified or unmodified 8-mer was ligated into the gapped vector using 400 units of T4 DNA ligase and 1 mM ATP at 16°C overnight. The 48-mer scaffold was removed by addition of a 10-fold molar excess of anti-48-mer (the complement of the scaffold sequence) to the ligation mixture, followed by heating of the sample at 90°C for 3 min and subsequent dilution with cold 1 mM Tris-HCl (pH 7.5), 0.1 mM EDTA. Samples were used immediately for transformation.
Mutagenesis Experiments-Transformation of cells and determination of mutation frequency were performed as described with slight modifications (21). Briefly, bacteria were grown to logarithmic phase and irradiated with UV light to induce the SOS response. The UV dose was determined by irradiating cells at increasing times from 0 to 3 min and then plating dilutions of the irradiated cells on Luria-Bertani plates. The optimal UV dose corresponded to roughly a 10% survival rate of the cells compared with no exposure. After irradiation, the cells were made competent for transformation as described previously (21). Competent cells (20 l) were transformed with 100 ng of singlestranded M13mp7L2 or Ϸ30 ng of double-stranded M13MB102-1 using a BTX TransPorter Plus electroporation system. Immediately following transformation, 1 ml of SOC medium (20 g/liter bacto-tryptone/5 g/liter bacto-yeast extract/20 mM glucose/2.5 mM KCl/10 mM MgCl 2 /9 mM NaCl) was mixed with the cells, and the bacteria were plated on Luria-Bertani plates and incubated at 37°C overnight (22). To determine mutation frequencies for M13MB102-1, the entire phage population was eluted from primary transformation plates, diluted, and replated with JM105 on X-gal/IPTG indicator plates to give ϳ500 plaques per plate (23). In M13mp7L2 experiments, the primary transformation plates were used for mutation analysis. The plaques on the plates were then lifted onto nylon membranes and probed for base pair substitutions by differential hybridization with 13-mer probes for the M13MB102-1 experiments (4) or 14-mer probes (ATTCGGAXACCCAC, where X ϭ A, C, G, or T) for the M13mp7L2 experiments. Membranes from modified and unmodified phage plates were split into four dishes, with each dish containing one of the four hybridization probes.
Frameshift mutations induced by HOPdG in M13MB102-1 were detected by phenotypic screening with X-gal/IPTG during the secondary plating. M13MB102-1 harbors a single deletion at position 6284 of the lacZ gene, so frameshift mutations that restore the reading frame are detected as blue plaques against a background of colorless plaques.
In Vitro DNA Lesion Bypass Assays-DNA polymerase assays were performed essentially as described (24). Primers were 5Ј-radiolabeled using T4 polynucleotide kinase and [␥-32 P]ATP and purified with Biospin 6 columns (Bio-Rad). Substrates were generated by mixing modi- fied or unmodified 31-mer template with 32 P-labeled primer in a 5:1 template:primer molar ratio in the presence of buffer (20 mM MOPS, pH 7.4, 2 mM MgCl 2 , and 50 mM NaCl) and incubating at 90°C for 2 min, followed by slow cooling to room temperature. Polymerase reaction mixtures (10 l) contained 5 nM of the 32 P-labeled DNA template:primer substrate in 50 mM MOPS (pH 7.4), 10 mM MgCl 2 , 2 mg/ml bovine serum albumin, 5 mM dithiothreitol, with 100 M dNTPs (dATP, dCTP, dGTP, and dTTP individually or together as indicated) and the indicated amounts of Klenow fragment (exo Ϫ ). The reactions were carried out for 30 min at 30°C for the running-start primer and 37°C for the standing-start primer and were then quenched by the addition of 10 mM EDTA in 90% formamide. The products of the reaction were resolved on a 20% polyacrylamide gel using the Ultrapure Sequagel sequencing system (National Diagnostics, Atlanta, GA). The positions of the bands were visualized by autoradiography and imaged with PhosphorImager analysis (Molecular Dynamics, Sunnyvale CA).

Mutagenicity of HOPdG-modified Double-stranded Vectors-
Construction of vectors containing HOPdG followed the basic procedure described by Benamira and Marnett (19). The M13MB102-1 vector was used during this preparation. M13MB102-1 contains a single deletion at position 6284 in the lacZ␣ gene of the M13MB102 genome. Frameshift mutations that restore the reading frame of lacZ can be detected by phenotypic screening in the presence of IPTG and X-gal. Double-stranded M13MB102-1 DNA was linearized with BssHII and SacII, then dialyzed with an excess of (ϩ)-single-stranded phage DNA to create a duplex with an 8-nucleotide gap in the (Ϫ) strand. Phosphorylated 8-mer oligonucleotides containing HOPdG or dG were ligated into the gap at a position corresponding to position 6256 in the M13 genome. The resultant closed circular duplexes were separated from incompletely ligated duplexes in ethidium bromide-containing agarose gels. The modified genomes were then transformed into E. coli strains that were either wild-type or deficient in DNA repair. The transformed cells were plated to produce a lawn of progeny phage. The plaques were eluted, and an aliquot of the stock was replated to yield ϳ500 plaques per plate. Plaque DNA from the secondary plating was lifted onto nylon membranes and probed by differential hybridization with radiolabeled probes specific for each type of base pair substitution. Frameshift mutations were detected by phenotypic screening using IPTG/X-gal during the secondary plating. HOPdG did not increase the frequency of frameshift mutations compared with unadducted genomes in any of the strains tested. Table I lists the outcomes of in vivo replication when HOPdG-containing genomes were replicated in three separate E. coli strains. All strains were UV-irradiated before transformation to increase sensitivity to mutation by inducing the SOS response. No base pair substitutions were detected following replication of HOPdG-containing genomes in any of the strains with or without SOS induction. LM102 is wild-type for repair, whereas LM103 contains a uvrA6 mutation that eliminates nucleotide excision repair. No increase in mutation frequency was detected when the HOPdG-modified vectors were replicated in the LM103 strain. LM119 contains deletions in two major apurinic/apyrimidinic endonucleases, exonuclease III and endonuclease IV. Apurinic/apyrimidinic endonucleases function to cleave the phosphodiester backbone at abasic sites, and their action is associated with the base excision repair pathway. If base excision repair is involved in the removal of   HOPdG, a deficiency in this repair pathway should result in an increase in the overall mutation frequency compared with repair-proficient cells. Replication of HOPdG-modified vectors in LM119 did not result in the occurrence of base pair substitutions.
Mutagenicity of Modified Single-stranded Vectors-The use of duplex vectors may mask the mutagenicity of lesions because of DNA repair or strand bias during replication. Therefore, we constructed single-stranded vectors containing HOPdG at a defined site to eliminate these possibilities. Vector molecules were constructed by the method described by Banerjee et al. (20). Single-stranded M13mp7L2 was digested with EcoRI, utilizing a unique site within a short hairpin region formed in the single-stranded DNA. The vector was then recircularized by annealing with a scaffold oligonucleotide. The ends of the scaffold are complementary to the ends of the cut vector, but the scaffold leaves an eight-nucleotide gap within the duplex region. 8-mer oligonucleotides containing HOPdG or dG, in the same sequence context used during the construction of M13MB102-1 vectors, were ligated into the gap. The scaffold was removed by denaturation in the presence of an excess of its complement. The adduct-modified vector was transformed by electroporation into SOS-induced E. coli, and the transformed cells were plated to produce a lawn of progeny plaques. The plaques were lifted onto nylon membranes and assayed for mutations by differential hybridization with oligonucleotide probes specific for base pair substitutions. No mutations were observed in LM102 cells, complementing the results observed in the duplex vector studies (Table II).
In Vitro Lesion Bypass with Purified Polymerase-The results of the in vivo experiments suggest that HOPdG is not highly mutagenic and is not a strong block to replication. Therefore, we tested the ability of HOPdG to block DNA replication, using an in vitro lesion bypass assay. Extension of a 12-mer running-start primer annealed with a 31-mer template containing HOPdG at the sixteenth position from the 3Ј-end was compared with an unmodified template. The template: primer substrate was extended by the Klenow fragment of DNA polymerase I (exo Ϫ ). The newly synthesized products of this reaction were separated on a 20% polyacrylamide gel and imaged using PhosphorImager analysis. Extension of the primer opposite an unadducted strand by Klenow fragment in the presence of all four dNTPs completely extended to the full-length product. However, full-length extension of template: primers containing HOPdG was substantially inhibited. Blockage of extension by Klenow fragment occurred opposite to the thymidine immediately preceding HOPdG (Fig. 2). At increased concentrations of polymerase (molar ratio of enzyme to substrate greater than 1:1), a second product was observed, corresponding to an arrest of replication opposite HOPdG. Bypass of HOPdG to produce full-length extension products was observed at all enzyme concentrations tested for Klenow fragment (exo Ϫ ), although the extent of bypass was modest.
Primer extension studies were performed in the presence of individual dNTPs using a 13-mer standing-start primer. Klenow fragment (exo Ϫ ) preferentially incorporated dAMP and dGMP residues opposite HOPdG (dGMP Ͼ dAMP), although some incorporation of dCTP was also observed (Fig. 3). Thus, in contrast to in vivo replication, in vitro replication with Klenow fragment reveals that HOPdG is able to strongly block replication and to miscode for insertion of purines opposite to it. DISCUSSION In this report, we investigated both the mutagenic potential and repair of an HOPdG adduct site-specifically positioned in double-stranded and single-stranded M13 genomes and replicated in E. coli. Our results show that replication of doublestranded vectors containing HOPdG does not induce base pair substitutions or frameshift mutations at the level of sensitivity of the assay (ϳ10 Ϫ3 ). The same results were obtained in bacterial strains that are either wild-type for repair or deficient in nucleotide excision repair or base excision repair. This observation does not result from a masking effect due to the doublestranded nature of the vector, because HOPdG also was not mutagenic when incorporated into single-stranded vectors. Thus, even though the entire base pairing region of guanine is blocked, DNA polymerases still appear to insert C opposite HOPdG in vivo.
The low mutagenicity of HOPdG is quite different from that of other six-membered exocyclic guanine derivatives when compared in an identical sequence context. PdG and M 1 G induce base pair substitutions and frameshift mutations in vitro and in vivo (2,4). HOPdG, M 1 G, and PdG do not induce frameshift mutations in the sequence used in the present study, but PdG and M 1 G do induce frameshifts in a sequence containing reiterated CGs. The structural difference between PdG and HOPdG is the presence or absence of a hydroxyl group, whereas M 1 G and HOPdG also differ by the pi character of the exocyclic ring of M 1 G (Fig. 1). An analogy may be drawn to the study of Langouet and coworkers (25), which demonstrated that ethano-dG and etheno-dG, the five-membered homologues of PdG and M 1 G, are mutagenic in E. coli, whereas HO-ethano-dG, the homologue of HOPdG, is only weakly mutagenic.
The lack of mutations induced by HOPdG is not the result of it being a strong block to in vivo replication or to its removal by repair systems, as demonstrated by the use of single-stranded vectors in mutagenesis assays. One possible explanation for HOPdG's nonmutagenic nature is that the structure of HOPdG may exist in multiple forms, where one form is miscoding but the other not (Fig. 4). A related interconversion has been observed for M 1 G, which exists in a hydrolytically ring-opened form (N 2 -(3-oxopropenyl)-dG) when it is opposite dC residues but in the ring-closed form when it is opposite dT residues (Fig.  4). The ring-opened form appears to be substantially less mutagenic than the ring-closed form. One can postulate the existence of an analogous ring-opened form for HOPdG in which the ring-opened form, N 2 -(3-oxopropanyl)-dG, would be expected to be substantially less mutagenic. In fact, a very recent study by de los Santos et al. (26) indicates that HOPdG exists entirely in the ring-opened form in duplex DNA and that the ring-opened form hydrogen bonds effectively to dC residues. This type of interconversion of opened and closed forms of exocyclic adducts may account for the low mutagenicity of HO-ethano-dG as speculated by Langouet et al. (25).
The relevance of our results to the mutagenicity and tumorinitiating activity of acrolein is supported by recent studies by Yang and coworkers (27), who reported that HOPdG was not mutagenic in E. coli. It seems unlikely that HOPdG is the principal adduct responsible for acrolein mutagenicity. This suggests that the mutagenic effects are mediated by other acrolein-DNA adducts. Isomeric 6-HOPdG isomers are formed on reaction of acrolein with DNA in vitro and have been detected in human tissue (12,13,15). Even if 6-HOPdG ringopens in duplex DNA, it will still contain a 3-oxopropanyl group on N1 of dG, which should alter translesion synthesis. In addition, there are other adducts that may contribute to the genotoxicity of acrolein. Recently, Wang et al. (28) reported that the hydrolysis product of crotonaldehyde-␤-hydroxybutyraldehyde adds to the exocyclic amino group of deoxyguanosine to form unstable imine adducts. The imine adducts are present in concentrations 50-fold higher than those of the 6-methyl-8hydroxypropanodeoxyguanosine adduct generated by reaction of deoxyguanosine with crotonaldehyde. It is likely that related N 2 -imino adducts are formed by reaction of deoxyguanosine residues with hydrated acrolein (i.e. 3-hydroxypropionaldehyde). It will be interesting to evaluate the mutagenicity of these adducts using site-specific procedures analogous to those described in the present work. However, the chemical instability of these novel exocyclic and N 2 adducts may make this a formidable experimental challenge.