Two different mechanisms for mutagenesis following treatment with methyl methanesulfonate (MMS) are suggested from the dose-response curve that is best fit by the linear quadratic model where m = 0.130D + 0.038D2 (D = dose measured as alkylations per nucleotide X 10(3), APdN; m = percent sex-linked recessive lethals, SLRL). A predominant component of the dose-response curve at moderate to high dose is the quadratic component which is interpreted as the result of two single-strand breaks. The distribution of methyl adducts in vivo is consistent with the previously determined in vitro distribution of methyl adducts on DNA following treatment with MMS. With the use of HPLC, 82% of the 3H-labeled adducts are found on the N-7 of guanine. It has previously been shown by both in vitro studies and in vivo correlation with mutagenesis that the N-7 alkyl guanine is not itself a predominately genotoxic lesion; however, N-7 alkyl guanine destabilizes guanine resulting in an increased rate of hydrolysis producing apurinic sites. In data presented in this paper, the loss of labeled adducts is shown to be at a rate consistent with hydrolysis of the destabilized alkyl guanine. The apurinic site thus produced should be converted to single-strand breaks by AP endonucleases once sperm has fertilized the egg. Single-strand breaks are repaired by excision repair which is not error-prone; however, multiple breaks producing a proximity effect should lead to double-strand breaks that are repaired by an error-prone process. Mutations that are induced by a proximity effect would account for the quadratic term. It is hypothesized that a proximity effect is produced when two breaks are sufficiently close together to prevent using the complementary strand as a template. The linear component of the dose-response curve is probably due to alkylation of oxygens in the purine or pyrimidine ring leading to mispairing. However, due to the low frequency of ring-oxygen alkylation following treatment with MMS, this important genotoxic site is not the predominant one observed at experimental levels normally used in the laboratory. From the dose-response curve, it is calculated that at mutation frequencies of 10 times the spontaneous frequency or higher, the predominant mechanism is the multi-hit component; however, at mutation induced frequencies of 0.1 of the spontaneous frequency, which are levels more likely to be encountered in man's exposure to environmental mutagens, the dominant mechanism is the linear component.(ABSTRACT TRUNCATED AT 400 WORDS)