Nearest neighbor affects G:C to A:T transitions induced by alkylating agents.

The influence of local DNA sequence on the distribution of G:C to A:T transitions induced in the lacI gene of E. coli by a series of alkylating agents has been analyzed. In the case of nitrosoguanidine, two nitrosoureas and a nitrosamine, a strong preference for mutation at sites proceeded 5' by a purine base was noted. This preference was observed with both methyl and ethyl donors where the predicted common ultimate alkylating species is the alkyl diazonium ion. In contrast, this preference was not seen following treatment with ethylmethanesulfonate. The observed preference for 5'PuG-3' site over 5'-PyG-3' sites corresponds well with alterations observed in the Ha-ras oncogene recovered after treatment with NMU. This indicates that the mutations recovered in the oncogenes are likely the direct consequence of the alkylation treatment and that the local sequence effects seen in E. coli also appear to occur in mammalian cells.


Introduction
Alkylating agents represent a major class of potential human mutagens and carcinogens. These agents give rise to a large number of DNA adducts, both in vivo and in vitro (1,2). The major adducts suspected of being responsible for the mutational and carcinogenic effects of these agents are the 06-alkylguanine and the O0alkylthymine adduct. Evidence indicating the potency of these adducts is derived from a number of sources. These include the observation that persistence and/or existence of adducts generally correlates with carcinogenesis in the target organs (3)(4)(5). Furthermore, these adducts have been shown to mispair in in vitro systems (6)(7)(8). In addition, 06-methylguanine and 04-methylthymine have been shown to mispair with thymine and guanine, respectively, in semiin vivo site-specific mutational systems (9)(10)(11). Hence, there would appear to be more than adequate evidence indicating a role for these adducts in alkylation-induced mutagenesis.
Our approach to these questions has concentrated on the analysis of mutational specificity. These studies also indicate major roles in mutagenesis for the 06-guanine and 04-thymine adducts. For example, agents producing almost exclusively the 06 adduct such as ethylmethanesulphonate (EMS), N-methyl-N-nitro-N-nitrosoguanidine (MNNG), and N-nitroso-N-methylurea (NMU) yield almost exclusively G:C to A:T transitions (12)(13)(14). In contrast, N-nitroso-ethylurea (ENU), which produces both the O6 and O0 lesions, induces both G:C *Biology Department, York University, 4700 Keele Street, Toronto, Ontario, Canada M3J 1P3. to A:T and A:T to G:C transitions in a ratio similar to that of the adducts (15,16). Finally, the analysis of mutation in strains defective in DNA repair are also consistent with these models (12,15).
While these mutagenic lesions thus appear capable of accounting for the specificity of the base substitutions encountered following treatment with a wide variety of alkylating agents, there are multiple steps between the exposure to the chemical and the final observation of the mutation. Among the major factors requiring consideration are the deposition of damage and the influence of DNA repair. Others include the nature and quantities of other molecules in the cell that can act as traps and thereby protect the DNA by essentially altering the effective dose. Further considerations need to include the nature of the mutation being selected, as a large fraction of DNA sequence changes are unlikely to lead to selectable phenotypic alterations.
The analysis of mutation at the DNA sequence level is also capable of shedding light on some of these questions as well. For example, a knowledge of mutational specificty at the DNA sequence level can also provide information on the influence of local DNA sequence. We have recently reported a nearest neighbor effect on G:C to A:T transition mutations induced by MNNG (13). The nature of this effect was that mutations were found to be much more likely to occur at guanine residues proceeded 5' by a purine base; thus developed the "PuG rule" for mutation induced by MNNG. In this communication we examine the validity and possible cause of the PuG rule following mutagenesis with other alkylating agents.
The approach used in this study involves the analysis of mutation in the lacI gene of Escherichia coli. Mutations in the first 180 base pairs of the gene were genetically identified and then cloned onto a specially constructed M13 vector, mRS81, by in vivo recombination. The methodologies in obtaining mutants are identical to those described elsewhere (13). Those methods for the cloning and sequencing of mutations are as described by Schaaper et al. (17). All the lacI mutations reported here were induced in a repair proficient strain of E. coli known as NR3835.

Results and Discussion
Mutation Induction by Alkylating Agents The survival and LacI-mutation induction data for a number of alkylating agents is given in Table 1. In addition to agents already mentioned, the table includes data on the nitrosamine N-nitroso-N-methyl-N-ax-acetoxybenzylamine (NMAB). This is a direct-acting analog of the esophogeal carcinogen N-nitroso-N-methyl-N-benzylamine (NMB). As can be seen in Table 1, the treatments cover a range of doses and some of the agents are clearly more toxic and/or mutagenic than others. From the table, however, it can be concluded that only a very small fraction of mutations sequenced following treatment are likely to be spontaneous in origin.

Preference for Mutation at 5'-PuG-3'Sites
The reported experiments with MNNG (13) indicate an influence of the 5' base on mutation induction. This effect is documented in Table 2, where it can be seen that 125 MNNG-induced mutations occurred at the 8 available 5'-PuG-3' sites, compared with 39 mutations occurring at the 17 potential 5'-PyG-3' sites. From these data it appears that mutation is almost seven times more likely to occur at a G residue proceeded by a purine than a pyrimidine. A similar analysis of the data for MNU, ENU, and MNAB also indicate a strong preference for mutation occurring at 5'-PuG-3' over that seen for 5'-PyG-3' sites.
In the case of the EMS treatment, the preference for 'Ratio of induced over spontaneous mutation frequency.
Abbreviations: MNNG, N-methyl-N-nitrosoguanidine; NMAB, Nnitroso-N-methyl-N-a-acetoxybenzylamine; MNU, methylnitrosourea; ENU, N-nitroso-N-ethylurea; EMS, ethylmethanesulfonate. 5'-PuG-3' sites is much less obvious. 5'-PuG-3' sites are preferred over 5'-PyG-3' sites by a factor of only 2.0. Moreover, mutations are most commonly recovered at sites proceeded by a C. This observation, however, is further complicated by the fact that DNA repair influences the distribution ofmutation (12). As we previously demonstrated, the excision of the 06-ethylguanine residues by the E. coli uvrABC repair pathway did not occur with equal efficiency at all sites; sites flanked by A:T base pairs are more efficiently repaired than sites flanked by G:C base pairs. In the absence of excision repair, i.e., in a UvrBstrain, the slight bias favoring 5'-PuG-3' sites disappears (12), and we concluded that the PuG rule does not apply to EMS-induced mutation.
The mutational spectra for the above treatments are summarized in Figure 1.

Origin of the Nearest Neighbor Effect
The observation that the mutability of a site might be influenced by local DNA sequence was first proposed by Benzer (18) as an explanation for the occurrence of mutational hotspots in the bacteriophage T4 rII gene. Since then, a number of examples of sequence effects have been observed (16,19). These influences have been shown to reflect reactivity (20), secondary structure of DNA (21,22), and repairability (12,23).
If we assume that the primary source of G:C to A:T transitions induced by these agents is the 06-alkylguanine adduct, then the 5' flanking base pair may influence mutation distribution by affecting the initial distribution of these lesions; the efficiency ofrepair ofthese lesions; and the fidelity of DNA polymerase at those positions in the sequence.
The 5' flanking sequence may exert its effect by affecting the distribution of 06-methylguanine lesions by modifying the reactivity of the 06 position of guanine.
Pullman and Pullman (24) have shown that the molecular electrostatic potential, an important component of a site's reactivity, can be strongly influenced by neighboring base pairs. Another important component of reactivity is steric accessibility, clearly a factor also subject to influence by local DNA sequence.
that the efficiency of repair of 06-methylguanine could vary several-fold with respect to the position of the lesion. However, in this case the influence of the 5' flanking base was modest, less that twofold. We note that differential repairability is unlikely to play a major role in the observed preference, since under the treatment conditions employed here, the repair capacity of the methyltransferase system is most likely fully saturated (25). Finally, the distribution of mutations may reflect the influence of neighboring base sequence on the fidelity of DNA synthesis past miscoding lesions. On the whole we tend to favor the hypothesis that it is the initial DNA damage distribution that is responsible for the observed 5'-PuG-3' preference. The argument is that this preference is observed under a number of conditions and at different doses (results not all shown). It is seen when the alkyltransferase repair pathway is quite probably saturated and when ethyl adducts, which are very much less efficiently repaired, are involved. We prefer the suggestion that the reactivity of the O6 position of guanine is substantially enhanced when proceeded by a purine.

Relevance of This Effect to Carcinogenesis
The 5' flanking base influence seen in these studies has an important counterpart in the results of studies on the activation of Ha-ras oncogenes by methylnitrosourea (MNU) (26). MNU was found to preferentially induce the activation of the protooncogene by a G:C to A:T transition at the second guanine in codon 12 (GGC), even though a transition at the first guanine can activate this gene. The finding that all the alkylating agents predicted to have a common intermediate, the alkyl dia-zonium ion, show this preference can be taken as a strong indication that the oncogene alteration was a direct consequence of the mutagenic treatment. Hence, this reactive intermediate may demonstrate the same mutational specificity in bacteria as in mammalian cells.
We are grateful to F. Allen for her technical assistance in carrying out so many cloning and sequencing reactions. This research was supported by the Natural Science and Engineering Council of Canada through grants A2814 and G1598.