Specificity of Mutagenesis by 4-Aminobiphenyl A POSSIBLE ROLE FOR N-(DEOXYADENOSIN-8-YL)-4-AMINOBIPHENYL AS A PREMUTATIONAL LESION*

Mutagenesis by N-acetoxy-N-trifluoroacetyl-4-ami- nobiphenyl, a reactive form of the human bladder carcinogen 4-aminobiphenyl (ABP), was studied in Esch- erichia coli virus M13mplO. N-acetoxy-N-trifluoroac-etyl-4-ABP-treated DNA containing 140 lesions/ duplex genome, when introduced into excision repair-competent cells induced for SOS mutagenic processing, resulted in a 40-fold increase in mutation frequency over background in the lacZa gene fragment. DNA sequence changes were determined for 20 independent mutants. G-C base pairs were the major targets for base pair substitution mutations, although significant mutagenic activity was also observed at certain A-T base pairs. Deletion and frameshift mutations also were found in this sample. The salient feature of this partial “mutational spectrum” was a hotspot that oc- curred at position 6357 (amino acid 30 of the j3-galac-tosidase fragment encoded by M13mplO); this A-T to T-A transversion appeared in 6 of the 20 mutants. The property of ABP to mutate A-T base pairs was consist- ent with the result that N-hydroxy-ABP reverted Salmonella typhimurium

always contains a set of structurally diverse carcinogen-nucleotide adducts (Basu and Essigmann, 1988). It is suspected that misreplication or misrepair of a subset of these adducts gives rise to mutations, which in turn may be the genetic precursors of the cancer phenotype. Given the wide range of DNA adduct structures that usually forms, it is a challenging task to determine which lesion(s) pose the most significant mutagenic risks to the cell. One strategy for establishing such relationships is to use the mutational spectrum of the chemical carcinogen to formulate hypotheses about which lesions are premutagenic. These hypotheses are subsequently tested in a model system in which a single carcinogen-DNA adduct is situated at a unique site in a phage or plasmid genome and then replicated in vivo. The determination of the type and amount of mutation induced usually enables an assessment of the extent to which the adduct under examination could have contributed to the mutational spectrum induced by the collection of lesions formed by the DNA-damaging agent.
The goal of the study described here was to identify the premutagenic lesions of 4-aminobiphenyl (ABP),' an aromatic amine used in industry until it was discovered to cause bladder cancer in exposed workers (Melick et al., 1971). Human exposure to ABP continues today owing to its presence in cigarette smoke (Patrianakos and Hoffmann, 1979;Bryant et al., 1987). ABP is mutagenic in Salmonella typhimurium (McCann et al., 1975;Kadlubar et al., 1982;Beland et al., 1983), in Escherichia coli (Pai et al., 1985), and in mouse lymphoma cells in culture (Oberly et al., 1984). When metabolically activated, ABP reverts S. typhirnurium strains TA1538, TA98, and TAlOO but not TA1535; in the presence of pKM101, ABP appears to cause either base pair or frameshift mutations. All of these strains are believed to detect mutagens that act primarily at G-C base pairs (Levin et al., 1982).
The ability of ABP to induce mutations at G-C targets led us to investigate the mutagenic activity of the dGa-ABP adduct as the initial step toward defining the mutagenic consequences of all ABP-DNA adducts. Although two guanine lesions have been identified, we chose to focus our investigation on the more abundant of the two, dGR-ABP, because it is directly accessible to chemical synthesis (Lasko et al., 1987) and because its effect on DNA structure is better understood. In the work reported here, we have undertaken an analysis of the potential of this adduct to cause base pair substitution mutations by using an Ml3mplO genome containing one dG8-ABP at a defined site. We also have defined a partial mutational spectrum for the full set of ABP lesions in the CY fragment of the lac2 gene of M13mplO.

RESULTS
Previous studies have shown that ABP and its metabolite N-hydroxy-ABP are mutagenic, but there are few data avail-

~~ ~
The "Experimental Procedures" are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. Schematic outline of the experimental strategy allowing assessment of mutagenic potential of all ABP adducts, or of a single lesion. A , mutation assays using Ml3mplO DNA randomly adducted with ABP lesions. DNA was treated with Nacetoxy-N-trifluoroacetyl-ABP as described under "Experimental Procedures." The adducted DNA was transfected into excision repairproficient cells that were competent for SOS mutagenic processing due to the presence of the mutagenesis-enhancing plasmid, pGW16. Progeny phage were isolated and scored for the presence of colorless or pale blue mutant plaques, which indicated a defect in j3-galactosidase activity. DNA sequencing of mutants allowed analysis of the mutational spectrum caused by ABP lesions. B, in parallel to mutation assays with randomly adducted DNA, the Ml3mplO cloning vector was used to assess the mutagenic potential of a single carcinogen residue situated at a unique site within the Ml3mplO genome. M13mplO-ABP1 DNA containing dG8-ABP at position 6270 in the minus strand was constructed (Lasko et al., 1987). This DNA was transfected into host cells, either excision repair-proficient or -deficient, containing the mutagenesis-enhancing plasmid pGW16. Progeny phage were screened for base substitution mutations at the PstI site by iterative cleavage of replicative form DNA with PstI or by a plaque color assay that specifically detected G to T transversion mutations at position 6270. able on the specificity of the mutational process. The mutational spectrum of ABP in E. coli was defined by introducing ABP adducts randomly into an M13 phage genome containing a fragment of the lac2 gene. In the same vector, the ability of the major dG8-ABP adduct to induce base substitution rnutations at the PstI site was assessed. An overview of the experimental strategy appears in Fig. 2.
Mutational Specificity of N-Acetoxy-N-trifluoroacetyl-ABP, a Reactive Form of ABP-Reaction of double stranded DNA with N-acetoxy-N-trifluoroacetyl-4-aminobiphenyl (Fig. 2 A ) produced 140 ABP residues/genome (M13mp10-ABP14,) or about 1 adduct/26 guanine residues. The accompanying number of abasic sites in the modified replicative form DNA was estimated to be 0.8 (+.O.l)/genome. Under the same reaction conditions, single stranded DNA yielded over 400 ABP adducts/genome. When calf thymus DNA modified in parallel to the same binding level was analyzed by reversed-phase high performance liquid chromatography, the adduction pattern was found to be similar to DNA reacted with N-hydroxy-ABP in vitro ; dGa-ABP, dG2-ABP , and d A 8 -A B P (Fig. 1) accounted for 66, 6, and 8% of the radioactivity, respectively. Other peaks in the profile each contained no more than 3% of the total radioactivity. A mutation in the lac& fragment of M13mplO resulting in loss of complementation (thus little or no p-galactosidase activity) produces an easily distinguishable colorless or pale blue plaque, compared to a wild-type blue plaque, on an

Mutation frequency of M13mpl0-ABP~0 DNA
Data from at least five separate transfection tubes were pooled for each experiment. Numbers in parentheses are Poisson 95% confidence intervals for the mutation frequencies. ND, not determined. A nonconcurrent buffer and solvent control experiment revealed 0 mutant plaques/8,268 total plaques, a Poisson upper 95% confidence limit of 4.3 x 10" for mutation frequency induced by reaction conditions. Transfection efficiency of control DNA was at least 10' transfectants/ng. ND ND indicator plate. Detectable mutations include base pair substitutions both in the regulatory and structural regions of the target gene, frameshift mutations, and insertions and deletions (LeClerc and Istock, 1982;Kunkel, 1984;LeClerc et al., 1984). The rnucAB operon of plasmid pGWl6, a point mutant derived from plasmid pKMlOl (Walker, 1978), encodes enhanced SOS mutagenic processing compared to its progenitor.
In the presence of pGW16, mutagenesis of UV-treated M13mplO single stranded DNA is detectable without prior irradiation of E. coli used for transfection assays.3 This property makes pGW16 a convenient tool for the analysis of mutagenesis by lesions that require SOS mutagenic processing (Walker, 1985).
The mutation frequency among progeny phage increased when M13mp10-ABP140 DNA was introduced into excisionrepair competent cells containing plasmid pGW16 ( Table I).
The damaged DNA yielded a 25-50-fold increase in mutation frequency compared to control, untreated replicative form DNA. A separate mutation assay on DNA subjected to identical buffer and extraction conditions yielded no mutant plaques among 8268 examined, placing an upper 95% Poisson confidence limit for mutation frequency induced by buffer conditions of 4.3 x In cells lacking pGW16, no mutant plaques were detected in 3421 examined; the upper 95% Poisson confidence limit for mutation frequency under these conditions was 1.1 x Thus, SOS mutagenic processing was necessary for mutagenesis to be observed in M13mplO-ABPllo DNA. Survival of phage immediately after heat shock, measured as infective centers, was 0.2% in DL7 (uur+) cells, 0.3% in DL7/pGW16 cells, but less than 0.005% (no plaques detected) for both DL6 (uurA) and DL4 (uurC) cells. Survival in these experiments was the number of infective centers formed by adducted DNA compared to the number resulting from the same amount of unmodified DNA. In a separate experiment, buffer and solvent control DNA showed no reduction in survival compared to untreated DNA. These data show that, at 140 adductslphage genome, excision repair proficiency markedly increased the survival of N-acetoxy-Ntrifluoroacetyl-ABP-treated DNA. Mutagenesis was not investigated further in excision repair-deficient bacteria because of the low viability of the adducted DNA in these strains.
Partial Mutational Spectrum of ABP Lesions-Twenty independent lac2 mutants were collected from excision repairproficient cells containing pGW16. Subsequently, their DNA was sequenced to provide a partial mutational spectrum for ABP lesions (Table I1 and   CGT (Arg) to GGT ( G l d neous mutations; however, because the mutation frequency in adducted DNA was on average 40-fold higher than control DNA, it was reasonable to assume that the mutants were mainly due to the ABP lesions. The noteworthy features of the ABP-induced spectrum were a hotspot (6 occurrences/20 independent mutants sequenced) for an A-T to T-A transversion at position 6357, other transversion mutations (mostly occurring at G-C base pairs), transition mutations, two (-1) and one (+I) frameshift mutations, and deletion mutations (Table 11). Because adenine and guanine bases are known targets for modification by ABP, we assume that the transversions and -1 deletion frameshifts due to N-acetoxy-N-trifluoroacetyl-ABP treatment originated at these sites. We have not as yet determined whether or not the A-T to T-A hotspot exists in the background mutational spectrum. Confirmation of the 4-base deletion mutations by ABP is required since we cannot conclusively rule out the possibility of contamination of DNA or phage from site-specific mutagenesis experiments (see below).
The finding of a hotspot for mutagenesis by ABP lesions at an A-T base pair prompted an evaluation of mutagenesis by N-hydroxy-ABP in S. typhimurium strain TA104, a strain designed to detect mutation at A-T targets (Levin et al., 1982(Levin et al., , 1984. Treatment of TA104 cells with N-hydroxy-ABP in a liquid suspension assay (Kado et al., 1983) at a dose of 50 ~L M resulted in a doubling of the number of revertant colonies at >60% survival; the dose-response curve was nearly identical to that of TAlOO assayed in parallel. Although the interpretation of TA104 mutagenesis in terms of exact DNA sequence alterations is complicated by the occurrence of extragenic ochre suppressor revertants, these data generally support the conclusion that in an SOS-processing environment, ABP lesions cause base pair substitution mutations at both G-C and A-T targets.
Site-specific Mutagenesis of dGR-AH' in M13mplO"Seven of 14 base substitution mutations in the mutational spectrum of ABP occurred at G-C base pairs, suggesting that one or both of the two known guanine adducts (Fig. I)  "___""""""~-~"""""""""""""""""""""~  Table 11. The parentheses open above the first or only base involved in the mutation. The PstI site, base pairs 6267-6272, was the location chosen for adduct-directed, site-specific mutagenesis studies. bp, base pairs. known ABP adducts to mutagenesis, the ability of dG8.ABp to induce base pair substitution mutations at the PstI site in Ml3mplO DNA was investigated (Fig. 2B).
In excision repair-proficient cells, there was no evidence that dG8-ABP caused base pair substitution mutations at genome position 6270 (Fig. 4). This was found to be the case even when a sufficient number of infective centers was examined so that the limit of detection with 95% confidence was below a mutation frequency of 0.02%. By contrast, single stranded DNA treated with 72 J/m2 UV light as a positive control produced a mutation frequency of 9 X lo-', demonstrating that the cells were indeed capable of the SOS processing required for UV-and ABP-induced mutagenesis. In excision repair-deficient backgrounds (uurA and uurC), induced base pair substitution mutagenesis was below 0.2%. In the two cases where mutation appeared to rise above background levels (DL4 cells + UV and DL6 cells -UV), DNA sequencing of six PstI-resistant mutants from these tubes revealed that five of the resistant phage were in-frame deletion mutants of -3 base pairs and one was a misclassified wild

FIG. 4. Mutation frequencies for base substitution (other than G-C to T-A) mutagenesis induced at the PstI site in
M13mplO-ABP1. Four rounds of selection with PstI followed by exonuclease I11 treatment were performed in order to remove wildtype molecules containing an intact PstI site from the population. The details of PstI selection and calculation of mutation frequency are described under "Experimental Procedures." Confidence intervals were derived from Poisson 95% confidence limits propagated through the calculation. Two transfection tubes were pooled for each data point. Each column in the histogram represents data from lo6 progeny phage. Top panel illustrates the results when excision repair-proficient or -deficient cells were not pretreated with UV light; the bottom panel shows results when UV was used to induce SOS mutagenic processing beyond the level normally observed in cells containing pGW16. The mutation frequency for M13 replicative form ( R F ) control DNA in uurA cells is not shown because only two rounds of cleavage were performed on this data point. type that had evaded PstI selection. Two PstI-resistant phage from control tubes were found to be A-T to T-A mutants at position 6271. It is conceivable that the in-frame deletion mutations were induced by the adduct but, as explained below, complete analysis of frameshift and deletion mutagenesis was beyond the scope of this study.
The G-C to T-A transversions at position 6270 in the minus strand were screened for directly. This transversion resulted in an opal codon (TGA) in the transcribed strand. Therefore, phage that exhibited a colorless plaque phenotype on the usual plating bacteria (GW5100, an amber suppressor strain) but were pale blue when plated on NR8044 (an opal suppressor plating strain) were likely to have possessed a G-C to T-A transversion at the PstI site. The frequency of G to T transversions was not elevated above the background (less than 0.03%). A complication of this screen resulted from the presence among the progeny phage of deletion mutants missing the central four bases (5'-TGCA-3') of the PstI site; these mutants were shown earlier to result from the genetic engineering procedures used to prepare the singly adducted genome (Loechler et al., 1984). Because these small deletions were relatively frequent (0.3-4.0%) and they yielded colorless plaques, it was impossible to screen easily for frameshift or deletion mutations induced by the adduct.

DISCUSSION
ABP Is a Versatile Mutagen-The most striking feature of the mutational spectrum is that ABP is a very versatile mutagen. An evaluation of a sample of 20 changes induced by ABP lesions in a genetic background that included the mutagenesis-enhancing plasmid pGW16 showed that base substitution mutations occurred at both G-C and A-T base pair targets (Table 11). This was consistent with earlier studies and with our observations on mutagenesis by ABP and its metabolites in S. typhimurium his tester strains. In addition, the mutational spectrum revealed frameshift mutations, again consistent with results from S. typhimurium reversion assays.
Comparative Mutagenesis of A B P Lesions throughout the Genome and at a Specific Site-The ability of the major DNA lesion caused by ABP in vivo or by reaction of activated forms of the chemical with DNA in uitro, dG8-ABP, to induce base substitution mutations was evaluated at a specific site in the M13mplO genome. The singly adducted genome yielded no increase in base pair substitution mutagenesis over background (less than 2 X mutants/progeny phage in excision repair-proficient cells, and less than 2 X mutants/progeny phage in excision repair-deficient cells). This negative result raises the question as to what frequency of mutation was expected for the "average" ABP lesion, based on our results with M13mp10-ABP14,. When a series of assumptions is made, including random modification, equal probability of repair at any target site, and equal probability of mutation at any given target, it is possible to calculate the "average" mutagenic efficiency of an ABP lesion (Schaaper and Loeb, 1981;Koffel-Schwartz et al., 1984). Given the mean induction of mutagenesis above background of 80 x and the mean number of adducts per target of 6, the average adduct results in mutation 1.4 times out of every 1000; i.e. the mutation frequency in vivo is -0.14%. Thus, the results of the sitespecific mutagenesis study indicate that the mutagenic efficiency of dG8-ABP built into the PstI site was at least 7-fold lower than this value under the conditions of the experiment, and below the limit of detection of our mutation assay. This result could indicate that the dG8-ABP adduct is not highly mutagenic. However, the partial mutational spectrum in the lacZa fragment, in combination with the results of many other workers on other mutagens (notably Benzer, 1961;Miller, 1983; and, with other aromatic amine lesions, Koffel-Schwartz et al., 1984) indicate that the effect of neighboring DNA sequence is important. A crucial future experiment will be to evaluate mutagenicity by an ABP adduct at a known "hotspot" for mutagenesis, as deduced from Table 11. An additional explanation for the inability to detect induced mutagenesis by dG8-ABP at position 6270 concerns the kinetics of DNA repair. The large number of lesions per genome in M13mplO-ABP,,, may well have saturated the repair system(s) for this type of damage, while the comparatively low dose of one ABP adduct per genome in M13mplO-ABP1 may have resulted in rapid repair of the single lesion and thus no observed mutagenesis. Experiments in uvrA and uvrC cells revealed no consistent increase in dG8-ABP-induced base pair substitution mutagenesis compared to excision repair-proficient cells (Fig. 4). This may have resulted from leakiness of the repair-deficient phenotype or from the ability of another cellular repair system to act on the single ABP lesion. In the case of 06-methylguanine (Loechler et al., 1984) for which there is a well characterized, suicidal repair enzyme (Lindahl, 1982), mutagenesis in single stranded DNA modified specifically at the PstI site was elevated from 0.4 to -20% when the repair system for this lesion was compromised. Further experimentation under conditions of saturated repair and SOS mutagenic processing may resolve this question.
Another plausible reason for lack of mutagenesis by the single dG8-ABP lesion in the duplex M13 genome is that there may have been a bias in favor of recovering progeny phage replicated from the unadducted plus strand of M13mplO-ABP,. Since double stranded DNA was used for transfections in the site-specific mutagenesis assays, progeny could come either from the intact plus strand or the adducted minus strand. Penetrance of the minus strand phenotype is at least -30% (Hill-Perkins et al., 1986;Kunkel and Alexander, 1986) and possibly higher . Because no direct measurement of the penetrance of the plus versus minus strands was made, a correction has not been applied to the raw mutation frequencies reported here. This correction would also alter the control mutation frequency estimates by the same factor. An additional and more severe bias in favor of progeny of the unadducted plus strand could have resulted from selective blockage of DNA polymerase I11 by the adduct in the minus strand. Consequently, few if any progeny phage may have been produced with the genotype of the adductcontaining strand. Such a model is strongly suggested by recent results of Koffel-Schwartz et al. (1987).

Specificity of Mutagen
quence-dependent and had a component that was independent of an umuDC processing environment; N-hydroxy-AFinduced mutagenesis was less governed by sequence context but was found to be umuDC-dependent.
In view of the structural similarity between the major AF and ABP adducts, our results for ABP and those reported for AF provide an interesting comparison. The mutational efficiency that we found for ABP was similar to those reported for other aromatic amine lesions (Koffel-Schwartz et al., 1984;Bichara and Fuchs, 1985). Moreover, in our hands ABPinduced mutagenesis was similar to that induced by AF, in that it was dependent on SOS mutagenic processing. With ABP, however, we observed a lower proportion of the G-C to T-A transversion and the presence of a hotspot at an A-T base pair target. The appearance in the mutational spectrum of ABP (Table 11) of a hotspot at an A-T base pair, coupled with the knowledge that a dA8-ABP adduct is among the lesions produced by ABP, suggests that site-specific mutagenesis in this DNA sequence context would be informative.
One plausible cause of the transversion mutations is an apurinic site formed either before or after entry of the adducted DNA into the cell (Schaaper and Loeb, 1981;Kunkel, 1984). The former possibility is unlikely, however, in view of the low number of apurinic/apyrimidinic sites per molecule of M13mp10-ABP140 prior to transfection (determined empirically to be 0.8) and the low mutation frequency in the solvent control. The average apurinic/apyrimidinic site formed prior to transfection would need to have a mutagenic efficiency of 20% in double stranded DNA in order to have been responsible for this mutation; this is much higher than has been reported thus far in any forward mutation assay. To the extent that apurinic/apyrimidinic sites may have acted as intermediates in the mutagenic activity of ABP, it is more likely that they formed inside the cell, probably via the action of repair enzymes or, less likely, by spontaneous depurination. In support of the involvement of apurinic/apyrimidinic sites in mutagenesis by ABP is the observation that ABP induced transversion mutations (Table 11) that presumably originated from purines (ABP forms no known pyrimidine adducts). These data are in accord with the established mutagenic specificity of apurinic/apyrimidinic sites (Loeb and Preston, 1986).
A second possible mechanism of mutagenesis by ABP would require that the premutagenic adduct(s) evade repair and induce mutation by forcing a replication error during DNA synthesis. In this regard it is noteworthy that the hotspot for the A-T to T-A mutation is adjacent to a run of 5 G-C base pairs, and it is tempting to speculate that the presence of the G-C run may act as a "clamp" that offers this sequence decreased susceptibility to the action of DNA repair enzymes version mutations for ABP, as well as for AF and other aromatic amines, is consistent with conformational changes in DNA that can arise from carcinogen modification at C-8 of guanine and adenine (Swenson and Kadlubar, 1981). From theoretical as well as spectroscopic studies, it appears that nine-deoxyribose linkage (Broyde et al., 1985;Shapiro et al., 1986). This has been suggested to occur preferentially with a destabilized or unwound DNA helix such as that obtained during replication. The syn conformation allows carcinogen stacking with a neighboring base and provides the potential for frameshift or missense mutations. However, such a conformational change (Fig. 5 ) also places the O6 and N-7 atoms of the modified guanine in a position to mispair with N-1 and N2 of a guanine or with an N6 and N-1 of an adenine (imino on U 8 -A B P . It also is noteworthy that the prevalence of trans-dGR-ABP can readily adopt a syn conformation about the gua-:esis by 4-Aminobiphenyl tautomer) in the complementary strand (Drake and Baltz, 1976;Topal and Fresco, 1976), resulting in G-C to C-G or G-C to T-A transversions, respectively. Similarly, dA8~ABP might be expected to be converted to its syn conformation in which a mispair could occur between N 6 and N-7 of the modified adenine and N 6 and N-1 of a complementary adenine, resulting in an A-T to T-A transversion. The existence of such base pairs is supported by ab initio self-consistent field and dispersion energy calculations that predict the relative stability of hydrogen bonds at O6 and N-7 of guanine and at N 6 and N-7 of adenine (Hobza and Sandorfy, 1987).