Mutational and DNA binding specificity of the carcinogen 2-amino-3, 8-dimethylimidazo[4,5-f]quinoxaline.

The mutagenic specificity of 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx), a food-borne mutagen and carcinogen, was studied. Plasmid pK19 was modified by photolysis with the 2-azido form of the carcinogen. High pressure liquid chromatography confirmed that the photoactivated azide formed primarily C8 and N2 guanyl adducts. Transformation of modified pK19 into excision repair competent Escherichia coli resulted in dose-dependent increases in genotoxicity and in mutagenesis within the lacZα target sequence. Upon induction of the SOS response, a 20-fold increase in mutation frequency over background was observed. A mutational spectrum for MeIQx, generated by sequencing 125 independent mutants, revealed base substitutions (41%), frameshifts (54%), and complex mutations (5.6%); >90% of the mutations occurred at G-C base pairs. Two hotspots were evident at runs of three or five G-C base pairs; ~60% of the mutations occurred at the hotspot sites. The hotspot at position 2532 produced mainly base substitutions, while that at position 2576 gave exclusively frameshift mutations. A polymerase inhibition assay mapped the sites of MeIQx adducts. Arrest sites were primarily at or one base 3′ to a guanine residue, which correlated well with the distribution of mutations. No direct correlation was seen, however, between intensity of modification and hotspots for mutation.

The mutagenic specificity of 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx), a food-borne mutagen and carcinogen, was studied. Plasmid pK19 was modified by photolysis with the 2-azido form of the carcinogen. High pressure liquid chromatography confirmed that the photoactivated azide formed primarily C8 and N 2 guanyl adducts. Transformation of modified pK19 into excision repair competent Escherichia coli resulted in dose-dependent increases in genotoxicity and in mutagenesis within the lacZ␣ target sequence. Upon induction of the SOS response, a 20-fold increase in mutation frequency over background was observed. A mutational spectrum for MeIQx, generated by sequencing 125 independent mutants, revealed base substitutions (41%), frameshifts (54%), and complex mutations (5.6%); >90% of the mutations occurred at G-C base pairs. Two hotspots were evident at runs of three or five G-C base pairs; ϳ60% of the mutations occurred at the hotspot sites. The hotspot at position 2532 produced mainly base substitutions, while that at position 2576 gave exclusively frameshift mutations. A polymerase inhibition assay mapped the sites of MeIQx adducts. Arrest sites were primarily at or one base 3 to a guanine residue, which correlated well with the distribution of mutations. No direct correlation was seen, however, between intensity of modification and hotspots for mutation.
The covalent binding of an aromatic amine or other carcinogens to DNA is widely regarded as the crucial event in cancer initiation (7). As with most chemical carcinogens, MeIQx and IQ require metabolic activation to exert their biological effects (7,8). Heterocyclic aromatic amines so activated can covalently modify cellular macromolecules, including DNA. The metabolic steps to MeIQx activation are depicted in Fig. 1. Studies in rodents and human tissues reveal that the initial activation step involves oxidation of the exocyclic amino group by cytochrome P4501A2 (9). Esterification of the resulting hydroxylamine species with acetate or sulfate is believed to produce unstable intermediates that decompose to a nitrenium ion, which is the putative DNA-damaging species (9,10). Metabolic studies using human tissue subcellular fractions also indicate the involvement of cytochrome P4501A2 in an initial oxidation event followed by possible O-or N-acetylation to produce highly reactive intermediates that react to form DNA adducts (9).
Studies to deduce the structures of the MeIQx-DNA adducts reveal that its electrophilic derivatives react predominantly at the C8 and N 2 atoms of guanine (11) to form N-(deoxyguanosin-8-ly)-MeIQx (3, dG-C8-MeIQx) and 5-(deoxyguanosin-N 2 -ly)-MeIQx (4, dG-N 2 -MeIQx), respectively ( Fig. 2A). The dG-C8-MeIQx adduct is the major adduct found in vivo in the livers of monkeys fed MeIQx, and also in the livers of rats fed MeIQx where a linear relationship exists between administered dose (doses mimicking human daily exposure) and MeIQx DNA adduct formation (12,13). These studies suggest that MeIQx DNA adducts may be a useful biomarker for exposure to this food-borne toxin.
There are no data as yet that have firmly established the DNA binding and mutational specificity of MeIQx. The available genetic data on the activated form of MeIQx come primarily from reversion assays, which reveal it to be a potent frameshift mutagen (14). The most frequent mutations are Ϫ1 and Ϫ2 base deletions involving guanine. In contrast, forward mutational assays of aromatic amines structurally related to MeIQx reveal mainly single base changes. Consistent with that observation, treatment of E. coli with the nitro derivative of 2-amino-3,4-dimethylimidazo [4,5-f]quinoline (MeIQ) induces mainly G-C to T-A transversions in the lacI gene (15). Frameshifts represent a minor fraction of the mutational spectrum. Similarly, G-C to T-A transversions dominate the spectrum of IQ-induced mutations in human fibroblasts (16). Recently, base substitutions were also found to be the prevalent mutation in the activation of ras oncogenes and the inactivation of p53 gene * This work was supported by Grants F32-ES05582 (to M. S. S.) and R32-CA52127 (to J. M. E.) from the National Institutes of Health. 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.
¶ To whom correspondence should be addressed. by heterocyclic aromatic amines (17,18). In contrast, frameshift mutations were seen in the inactivation of the Apc gene by heterocyclic aromatic amines, an event that is believed to be involved in the initiation of human colon carcinogenesis (19).
The first step toward elucidating the mechanisms of MeIQx mutagenicity and carcinogenicity is to establish its unique mutational spectrum in a target gene. One method to achieve that goal is to modify a target gene with the damaging agent and then to determine the pattern of induced mutations upon replication. In addition to genetic studies, analysis of the binding specificity of the DNA-damaging agent can reveal if there is a correlation between sites of mutation and sites of adduction. In the work described here, we investigated the mutational spectrum and DNA binding specificity induced by MeIQx in the ␣-fragment of the lacZ gene of E. coli.

EXPERIMENTAL PROCEDURES
Chemicals and Enzymes-Sequencing primers were synthesized by the Biopolymers Lab, MIT, or on an Applied Biosystems PCRmate and purified by polyacrylamide gel electrophoresis. Sequencing reagents (Sequenase) were obtained from U. S. Biochemical Corp. Deoxyribonuclease I (type IV), nuclease P1 (Penicillium citrium), phosphodiesterase I (type VII), and alkaline phosphatase (type II-S) were purchased from Sigma. Ultrapure dNTPs for polymerase arrest experiments were obtained from Pharmacia Biotech Inc. 5-Bromo-4-chloro-3-indoyl-␤-D-galactopyranoside and isopropyl-␤-D-thiogalactopyranoside were purchased from Gold Biotechnology, Inc. [␣- 35  Preparation of MeIQx-modified DNA-The plasmid pK19 (88 g, 50 pmol) was dissolved in 1 ml of TE buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 7.6) and N 3 -MeIQx ( 14 C-labeled) added in 20 l of dimethylformamide and the solution briefly mixed (final concentrations were 17-430 M for the toxicity study and 43 M for mutational specificity and DNA binding studies). The solution was then irradiated (230 V, 50 Hz, 0.17 amps) at a wavelength of 366 nm for 40 min with a Spectroline ENF-260 UV lamp held 3 cm above the solution. The MeIQx-modified DNA was separated from the unbound reaction products by solvent extraction three times with two volumes of chloroform. The DNA was then precipitated with ethanol (23) and resuspended in TE buffer. Unmodified control DNA was also produced by subjecting the plasmid to the modification conditions in the absence of N 3 -MeIQx. The concentration of recovered DNA was determined by UV absorption at 260 nm, and the extent of covalent binding was determined by liquid scintillation counting on a LKB 1219 Rackbeta scintillation counter.
DNA Digestion Conditions and Analysis by HPLC-MeIQx-modified DNA (50 g) was dissolved in 1 ml of reaction buffer (5 mM Tris-HCl, 10 mM MgCl 2 , pH 7.5), and deoxyribonuclease I was added to a final concentration of 0.05 mg of enzyme/ml. After incubation for 5 h at 37°C, nuclease P1, dissolved to a final concentration of 1 mg of enzyme/ml, was added and incubation was continued for 3 h at 37°C. Phosphodiesterase I (0.04 U) was then added, followed by 1 unit of alkaline phosphatase, and incubation carried out for another 18 h at 37°C. The reaction mixtures were then added to 3 ml of ethanol and the proteins removed by precipitation on ice, followed by centrifugation. The supernatant, which contained greater than 95% of the radioactivity, was lyophilized to dryness.
HPLC analysis of the DNA digest was performed on a Hewlett Packard 1090 M HPLC system with an online Berthold LB C-1 radioactivity monitor. A TosoHaas TSKgel ODS-80TM HPLC column (4.6 mm ϫ 25 cm, 5-m particle size, Stuttgart, Germany) was used for all separations. A linear gradient that started at 90% solvent A (25 mM KH 2 PO 4 , pH 2.5, 5% CH 3 CN) and 10% solvent B (100% CH 3 CN) and proceeded to 40% solvent B after 40 min was employed to separate adducts. The flow rate was 1 ml/min. Fractions were taken every 30 s and measured for radioactivity by liquid scintillation counting.
SOS Induction and Transformation of E. coli-E. coli DL7/pGW16 cells were grown in 100-ml batches in Luria-Bertani (LB) broth (23) to a cell density of approximately 1 ϫ 10 8 cells/ml. A portion of the culture was then irradiated in 10 mM MgSO 4 with UV fluences of 30 -45 J/m 2 to induce the SOS response (20,24). After a recovery period of 40 min at 37°C, the cells were harvested by centrifugation, washed with water, and resuspended in water as described by Yarema et al. (24). Unmodified and MeIQx-modified DNAs were introduced into E. coli DL7/ pGW16 cells by electroporation. The prepared cells (ϳ10 9 viable cells in 90 l) and 10 -100 ng of unmodified or MeIQx-adducted DNA were mixed and aliquoted into prechilled Bio-Rad Gene-Pulser cuvettes (0.2-cm electrode gap). The cells were electroporated in a BTX Electro Cell Manipulator 600 at 50 microfarads, 129 ohms, at 12.5 kV/cm for both SOS non-induced and induced cells. Immediately after electroporation, the cells were diluted with 1 ml of SOC medium (23) and allowed to recover for 1 h at 37°C. An aliquot of the transformed bacteria was then plated on LB plates containing kanamycin (50 g/ml) to determine the number of transformants. Transformation efficiencies were 1-4 ϫ 10 6 /g and 1-3 ϫ 10 4 /g for unmodified DNA and MeIQx-modified DNA, respectively. A portion of the remaining cells (0.5 ml) was diluted in LB (9.5 ml) containing kanamycin (50 g/ml) and grown for 4 h at 37°C. The cells were then pelleted by centrifugation and plasmid DNA isolated from the preparations using Wizard minipreps (Promega). The isolated plasmid DNA was then retransformed into BZ234 cells (made competent by CaCl 2 ; Ref. 23), and after heat shock allowed to incubate at 37°C for 1 h. The transformation mixture was then plated on minimal medium containing 5-bromo-4-chloro-3-indoyl-␤-D-thiogalactopyranoside and isopropyl-␤-D-thiogalactopyranoside. Mutant colonies in this system appear as either light blue or white after 48 h of incubation at 37°C. Mutation frequencies were determined from 14 separate sets of experiments as the number of light blue or white colonies divided by the total number of colonies.
Isolation and Sequencing of Mutants-Light blue or white colonies were picked and purified by three rounds of single colony purification. Double-stranded DNA isolated from mutant colonies (Wizard minipreps, Promega) was denatured and then sequenced according to the method of Sanger (25).
Polymerase Arrest Assay-These assays were carried out as de- scribed in Ross et al. (26) except that 33 P-labeled primers were used. Primers were end-labeled with T4 polynucleotide kinase (New England Biolabs) and [␥-33 P]ATP as described (23). DNA modified with MeIQx (2 g) was denatured with alkali and precipitated. The pellet was then resuspended in 2 l of Sequenase reaction buffer (200 mM Tris-HCl, 100 mM MgCl 2 , 250 mM NaCl, pH 7.5) and 6.5 l of water, prior to the addition of an equimolar amount of 33 P-end-labeled primer. Annealing of the primer to the template was accomplished by heating the mixture to 65°C in a water bath for 2 min and then allowing the mixture to cool for 30 min. To the annealed template mixture was added 1 l of 0.1 M dithiothreitol and 2 l of Sequenase (T7 DNA polymerase) version 2.0, diluted 1:8 in enzyme dilution buffer (10 mM Tris-HCl, 5 mM dithiothreitol, 0.5 mg/ml BSA, pH 7.5). A 3-l aliquot of this mixture was added to a prewarmed tube containing 2.5 l of 250 mM deoxynucleotide triphosphates and incubated for 5 min at 37°C. The reactions were terminated with 4 l of stop solution containing 95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF. The samples were heated to 75°C for 2 min prior to loading 3 l on a 6% polyacryamide gel. The gel was dried before exposing it to a Molecular Dynamics PhosphorImager screen for analysis.
PhosphorImager Analysis-Polymerase arrest data were quantitated by using ImageQuant software on a Molecular Dynamics PhosphorImager 400S. Densitometry scans were recorded for each lane on the gel. Peaks were integrated and areas compared to unmodified pK19 run under identical conditions. The intensity of the termination sites diminished with increasing distance from the primer in an exponential manner. The scans were normalized by generating a best fit exponential curve determined by Cricket Graph (Computer Associates) to the densitometry scan and then subtracting these values from each of the densitometry points. The relative intensities of the peaks (or stop sites) were then classified using five levels of intensity, where 1 was a weak arrest site and 5 was a strong arrest site.

RESULTS
The mutagenic consequences of MeIQx damage were evaluated in a forward mutational assay based on ␣-complementation between the lacZ␣ fragment on the kanamycin-resistant plasmid pK19 and the host, E. coli strain BZ234, M15 protein.
The lacZ␣ gene fragment, which encodes a portion of ␤-galactosidase, was the genetic target used to measure MeIQx-induced mutations. The azide of MeIQx (N 3 -MeIQx) was used to modify pK19 in vitro by photolyzing the compound under long wavelength UV light in the presence of the plasmid. Arylazides have been used recently for mutagenesis studies, because they generate short-lived arylnitrenium ions nonenzymatically. They have been shown to form the same DNA adducts as those of the putative in vivo activated species, N-acetoxy-MeIQx (27). The mutagenic potency of N 3 -MeIQx also correlates with Nacetoxy-MeIQx in reversion assays (27), and N 3 -MeIQx has been shown to form a similar ratio of N 2 and C8 guanine adducts as the N-acetoxy compound. 2 The modified plasmids were enzymatically digested and the resulting nucleosides analyzed by HPLC, as shown in Fig. 2B. The major peak observed, which represented 52% of the radioactivity incorporated into DNA, was identified as the dG-C8-MeIQx adduct by comparing its retention time to a synthetic standard. The dG-N 2 -MeIQx adduct was also present at 8.6%. This distribution of C8:N 2 adducts was similar to that seen in previous experiments in which calf thymus DNA was modified with N-acetoxy-MeIQx (11). 3 At a concentration of 43 M N 3 -MeIQx, a total of 33 MeIQx adducts/pK19 plasmid was introduced using this method.
MeIQx-modified pK19 was introduced into SOS-induced or uninduced E. coli strain DL7 containing the mutagenesis-enhancing mucAB genes (SOS functions) on plasmid pGW16 by electroporation. Mutagenesis enhancing plasmids have been used previously to determine the mutations from bulky adducts in excision repair competent cells where they increase SOSassociated mutagenesis (14,15,20). After a growth period, the cells were plated on LB or allowed to grow for subsequent DNA isolation. The toxicity of MeIQx was assessed at this point by counting the number of colonies produced. The isolated DNA was retransformed into the strain BZ234, which contains the complementary M15 protein (DL7/pGW16 does not provide ␣-complementation). Mutant colonies resulting from loss of ␣-complementation appeared as either light blue or white. A dose-dependent increase in toxicity and mutagenicity was seen over the range 17- The mutational frequency (MF) was calculated by dividing the number of light blue plus white colonies by the total number of transformants. When MeIQx-modified pK19 was transformed into uninduced DL7/pGW16 cells, a MF of 5.3 ϫ 10 Ϫ4 was found. 4 In comparison, the background MF assessed using pK19 subjected to the same buffer and photolyzing conditions as the modified plasmid was 1.5 ϫ 10 Ϫ4 . Upon transformation of the MeIQx-adducted plasmid into SOS-induced cells, the MF increased to 7.5 ϫ 10 Ϫ4 for the plasmid containing 7.3 MeIQx adducts and to 2.9 ϫ 10 Ϫ3 for the plasmid containing 33 MeIQx adducts. The background MF did not increase significantly with the induction of SOS functions. Since the mutation frequency in the modified DNA was on average 20-fold higher than the control DNA, it was concluded that 95% of the mutations were caused by the MeIQx modification.
MeIQx-induced Mutational Spectrum-Plasmids modified with 43 M N 3 -MeIQx were used for subsequent evaluation of mutational specificity. MeIQx-induced lacZ mutant colonies 2 R. Turesky, unpublished results. 3 It is possible that several of the other peaks observed in the chromatogram are oligonucleotides containing C8 or N 2 dG adducts. At high levels of DNA modification, enzymatic hydrolysis may have been incomplete. 4 Transformations with uninduced cells were only preformed on the plasmid containing 33 MeIQx adducts. It is assumed that the less modified plasmid would result in an even lower MF. Subsequent evaluation of mutational specificity was accomplished using the plasmid containing 33 MeIQx adducts, so this matter was not pursued.

FIG. 2. Structures of dG-C8-MeIQx and dG-N 2 -MeIQx (A) and HPLC analysis of an enzymatic digest of plasmid pK19 modified with 43 M N 3 -MeIQx (B).
were purified by three rounds of single colony purification, and their DNA was isolated for sequencing. A 170-bp portion of the lacZ gene including the polylinker region and 93-bp host M15 deletion was analyzed for mutations. A total of 203 individual mutants were sequenced, and 125 (61%) contained mutations, which was consistent with the fact that only a portion of the ␤-galactosidase insert was sequenced. The mutational spectrum for MeIQx in SOS-induced DL7/pGW16 is shown in Fig. 3. The mutations induced by MeIQx are clearly not random; they are divided between frameshifts (54%) and single base substitutions (41%) with a small contribution from complex mutations (5.6%). With the exception of three A-T to T-A transversions, all of the mutations occurred at G-C base pairs. Ninetythree spontaneous mutants were also sequenced; however, only five spontaneous mutations were found in the region of interest. The spontaneous mutations were a T-A to A-T transversion at 2531, a C-G to T-A transition at 2523, a T-A to G-C transversion at 2531, C-G to A-T transversion at 2512, and loss of a G-C base pair at 2576. Four of the spontaneous mutations did not correlate with any of the MeIQx-derived mutations, but the frameshift seen at base pair 2576 is situated in one of the hotspots for mutation and represents one of the 40 mutants found in this region.
The types of mutations observed in the mutational spectrum are summarized in Table I. Among the frameshift mutations, both single base additions and deletions were observed. Loss of a G-C base pair was the major frameshift mutation seen, accounting for 51% of the total mutations. The most frequent type of single base substitution was a transversion (25%), with G-C to T-A mutations dominating the spectrum. The few A-T to T-A transversions that were seen may have been caused by an MeIQx-adduct on the adjacent guanine interfering with proper replication at this adenine or may be due to modification of the adenine itself. The transitions seen were exclusively G-C to A-T and were present in 12% of the total mutations. Complex mutations consisted of double mutants, addition of AC, a TG to A mutant, and a GGTA to TT mutant; collectively these represented 5.6% of the mutants. Two hotspots were seen in the MeIQx-induced mutational spectrum. The first was at the 5Ј-GGG sequence starting at position 2532 and consisted mainly Sites of Polymerase Arrest-Most bulky DNA adducts inhibit replication in vivo (24,28,29). To determine whether MeIQx was also a replication blocking lesion and to examine the sequence specificity of the modified sites, a polymerase arrest assay was performed. Plasmid pK19 modified to 33 adducts/ plasmid (the same level used for mutant isolation; approximately 2 MeIQx adducts/170-bp LacZ fragment) was used for this analysis. A 5Ј-end-labeled primer was annealed to denatured MeIQx-modified plasmid template and extended with T7 DNA polymerase (Sequenase version 2.0, which contains no 3Ј to 5Ј exonuclease activity). Exact polymerase stop sites resulting from the presence of MeIQx adducts were visualized by electrophoresing the extension products and comparing their positions to dideoxy sequencing products of unmodified pK19 plasmid. In this manner the position of the stop sites was determined to the nearest base. As a control, the replication mapping assay was also conducted with unmodified pK19 and analyzed on the same gel as the modified DNA. Quantitation of the observed bands was accomplished by PhosphorImager analysis. Stop sites found in the lacZ region are shown in Fig.  3. The stop sites varied in intensity and were predominantly at guanine residues. The intensity of the bands was ranked from 1 to 5 in increasing order of intensity. Greater than 85% of these bands were at guanine or the adjacent 3Ј base. This percentage increased to 93% when arrest sites two bases 3Ј to guanine were included. These results are in agreement with the idea that the polymerase stopped opposite or 3Ј to the MeIQx lesion. Sites of polymerase arrest that were one base before potentially modified guanines were the strongest sites, with intensities of 4 and 5. Sequences with hotspots for stoppage were runs of guanines, 5Ј-AGC, 5Ј-CGT, and 5Ј-TGC. When arrest sites occurred adjacent to or at a single guanine surrounded by pyrimidines, the location of the adduct could be deduced as MeIQx forms no known pyrimidine adducts. This was also the case for arrest sites at the end of a run of purines. When the arrest site was in the middle of a run of purines, however, the exact location of the adduct could not be deduced. Therefore, precise linkage of adduct site to arrest band was not attempted.
When polymerase arrest sites were compared with the mutational spectrum, several trends were observed. A polymerase arrest band was present at or 3Ј to the putative guanine adduct for all mutations found except for those at positions 2603, 2554, and two A-T to T-A transversions at positions 2574 and 2505. A reasonable correlation between sites of modification and sites of mutation can thus be assumed, although many more arrest sites were seen than mutations. This, of course, was expected to a certain degree owing to the redundancy in the genetic code and the fact that point mutations at the third positions of codons generally do not score. We also observed that the intensities of the sites of modification did not necessarily correlate with the mutational intensities. For example, comparison of the mutational hotspots at positions 2532 and 2576, reveals that the hotspot for (Ϫ1) frameshift mutations at position 2576 was only moderately blocking to the polymerase, with the highest intensity being two for this sequence. In contrast, the mutational hotspot for base substitutions at position 2532 was also a hot spot for polymerase arrest with intensities of 4 and 5, the strongest sites in this assay. DISCUSSION The heterocyclic aromatic amines have attracted much recent interest because of suspicion that they may be responsible for the induction of genetic diseases, especially cancer, in humans. A representative of this class of carcinogens, MeIQx, is one of the most abundant dietary heterocyclic aromatic amines, with dietary intakes estimated at 0.2-2.6 g/person/day (12). As a first step toward understanding the mode of action of MeIQx, we have investigated its effects on DNA polymerases in vitro and in vivo.
Modification of a plasmid with the azide of MeIQx generated a genome that contained both C8 and N 2 guanine adducts. Subsequent transformation of this modified DNA into E. coli produced a dose-dependent increase in toxicity. Increased mutagenicity was also observed and, as with many bulky adducts including 4-aminobiphenyl (ABP) and 2-(acetylamino)fluorene (AAF), MeIQx required SOS induction for its mutagenesis (20,30). The SOS dependence of the MeIQx-induced mutations (MF of 2.9 ϫ 10 Ϫ3 at 33 adducts/plasmid) was detected as a 5-fold increase in MF over that of uninduced cells and a 20-fold increase in MF over that of unmodified DNA. These results are comparable to those of King and co-workers who found a mutation frequency of 1.3 ϫ 10 Ϫ3 for AAF, a 6.2-fold increase over uninduced cells, in the lacZ gene of M13mp9 with 63 adducts/ molecule under similar host conditions (31). Lasko et al. (20) found a MF of 8.0 ϫ 10 Ϫ4 for ABP corresponding to a 25-fold increase in MF over uninduced cells. If the MFs are taken on a per adduct basis within the lacZ mutational target, a single MeIQx adduct had a MF of 0.10%. By comparison, AAF and ABP had mutation frequencies of 0.019% and 0.14%/adduct/ lacZ mutational target, respectively, indicating that the mutation frequency of MeIQx is comparable to other carcinogenic aromatic amines.
While the mutation frequencies are similar, the lacZ mutational spectrum of MeIQx is unique when compared with other bulky adducts. MeIQx was able to induce frameshifts and base substitutions at an almost equal frequency, with 91% of the mutations occurring at G-C base pairs. The most abundant mutations seen were single base deletions of G-C, G-C to T-A transversions, and G-C to A-T transitions. Of the mutations that did not occur at G-C base pairs, all occurred at a base pair adjacent to a G-C base pair. The mutational spectrum generated here differs from previous studies in bacteria on the structurally related compound, MeIQ, where base substitution mutations are the major mutagenic event. In addition, in mammalian cells, IQ induces primarily base substitution mutations with very little contribution from frameshift mutations (15,16). It would appear that MeIQx produces a very distinct mutational spectrum, although neither of these previous studies employed the lacZ gene as a reporter of mutations so a strict comparison cannot be made.
A substantial portion of the MeIQx-induced mutations (60%) was clustered at positions 2576 -2581 and 2532-2534. Both of these sites are in G-C clusters adjacent to runs of adenines, which may influence the reactivity to the activated MeIQx at these sites, since runs of adenines are known to bend DNA and thus structurally influence adjoining sequences (32). Interestingly, different types of mutations occurred at these two sites. The sequence at position 2532 produced mainly base substitutions (although some frameshifts were also detected), whereas the sequence at position 2576 produced exclusively single base frameshift mutations. Of the five spontaneous mutations found, only one, a frameshift at 2576, occurred at a site in the MeIQx-induced spectrum.
Previous studies have demonstrated that the sequence 5Ј-CCCCC, within the same context of the lacZ in M13, is prone to frameshift mutations when bulky adducts are present. Gupta et al. found a high frequency of single base deletions at this sequence in their work on AAF (31). Misincorporation of nucleotides at homopolymeric sequences, resulting in frameshift mutations, is thought to be a common occurrence in DNA synthesis, potentially arising from slippage of the two DNA strands during replication (33,34). Fuchs and co-workers have studied the frameshift mutations induced by AAF in site-specifically modified DNA duplexes (34). They speculate that destabilization of the helix by AAF at the sequence 5Ј-CGGGA is relieved by formation of a bulged intermediate; rotation of the AAF moiety into the helix stabilizes the mutagenic intermediate (34). Interestingly, both the hotspots for MeIQx-induced mutations are at 5Ј-GGGA sequences and frameshift mutations were found at both positions. The C8 adduct of MeIQx is known to rotate into the syn orientation on a deoxyguanosine in an analogous manner to the AAF C8 adduct. It is plausible that MeIQx may also induce destablization in a similar manner to AAF, producing a bulged intermediate that causes misinsertion of cytosine opposite the lesion.
Single base substitutions were seen at a frequency of 41% in the mutational spectrum of MeIQx with the most dominant being G-C to T-A transversions followed by G-C to A-T transitions. The high frequency of G-C to T-A transversions may be explained by the known propensity of DNA polymerase to insert adenine opposite a noninstructional lesion such as a bulky adduct or an apurinic site (35). If the premutagenic bulky adducts dG-C8-MeIQx or dG-N 2 -MeIQx evade repair, they may induce a G-C to T-A transversion by forcing a replication error during DNA synthesis leading to preferential insertion of adenine. Alternatively, formation of an apurinic site may have occurred. Apurinic sites are thought to be formed from alkylation of the guanine N-7 position of heterocyclic aromatic amines and subsequent loss of the base (36). In fact, if one compares the distribution of base substitution mutations seen by Kunkel for AP site mutagenesis (59% A, 28% T, and 11% G incorporated opposite a template G) and those observed for MeIQx (64% A, 31% T, and 4% G incorporated opposite a template G), the distribution is quite comparable (37). The similarity supports the idea that the base substitutions observed by MeIQx may be due to AP site mutagenesis, an observation that has also been made for the induction of base mutations by MeIQ (15).
A polymerase arrest assay was used to determine the binding specificity of MeIQx and the ability of MeIQx-modified DNA to inhibit polymerization. MeIQx proved to be a replication blocking lesion with arrest sites at or 3Ј to guanine residues. While there was overall convention between the positions of mutation and the sites of adduct formation, there was little correlation between intensity of modification and mutational hotspots. This is especially evident when observing the hotspot for mutations at base pairs 2576 -2581. That site was only a weak arrest site in the polymerase arrest assay, yet 32% of the MeIQx-induced mutations occurred at this position. Similar observations have been made by others (29,30).
One important feature of MeIQx is its mutagenic versatility, inducing both frameshift and base substitutions almost equally. The conformation of an MeIQx adduct is likely to influence its persistence and its mutagenic impact. NMR studies on the 2Ј-deoxyguanosine MeIQx-adduct at C8 have shown that the preferred glycosidic conformation is syn (11). Similarly, dG-C8-AAF, which induces mainly frameshift mutations, displays the syn conformation both at the nucleoside (38) and duplex DNA (39) levels. It is conceivable that the dG-C8-MeIQx adduct in the syn conformation is responsible for the observed frameshift mutations. It is reasonable to speculate that the base substitutions may be derived from the corresponding dG-N 2 -MeIQx adduct. In contrast to the C8 adduct, the conformational preference was shown to be anti for the N 2 adduct of MeIQx on 2Ј-deoxyguanosine (11). Although, there have been no NMR investigations on the structural effects of bulky heterocyclic aromatic amine N 2 adducts in DNA, modeling shows that the N 2 adduct should be accommodated within the minor groove and be available for base pairing. Induction of base substitution mutations by N 2 adducts would result when polymerases traverse the lesion but fail to replicate it faithfully. Deciphering the conformations of the adducts of MeIQx will require synthesis of site-specifically modified DNAs followed by NMR structural investigations. This information in conjunction with mutagenesis studies involving singly modified genomes will be an important next step toward understanding the details underlying the mutagenic activity of the adducts of MeIQx.