Mismatch - stimulated killing ( bacteriophage A / heteroduplex DNA / DNA methylation / mlsmatch repair / double - strand break )

DNA duplexes with or without mismatches and with or without adenine-methylated GATC sequences were prepared from separated strands of bacteriophage A DNA and used to transfect Escherichia cofi. Unmethylated heteroduplexes containing one or more repairable mismatches transfect cells with a functioning mismatch repair system less efficiently than they transfect cells deficient in mismatch repair. No difference is observed when the duplexes contain no mismatch or a poorly repaired mismatch or when the heteroduplexes are fully or hemimethylated. These results and the phenotypes of E. coil dam mutants suggest that the E. colh mismatch repair system may introduce double-strand breaks in unmethylated DNA at or near repairable mismatches. The Escherichia coli mismatch repair system is able to recognize noncomplementary base pairs in DNA and acts, apparently via localized excision and resynthesis, to replace mispaired bases (see ref. 1 for review). Regions of DNA in which GATC sequences are fully adenine-methylated appear to be refractory to mismatch repair (2, 3), and it appears to be the transient undermethylation of newly synthesized GATC sequences in the region immediately following the replication fork that allows mismatch repair to operate only on newly synthesized strands and, thereby, to remove replication errors (1-5). E. coli deficient in adenine methylation (dam) have been found to have a mutator phenotype (6), as would be expected if mismatch repair occurs on either strand of unmethylated DNA-i.e., is undirected-or does not operate at all on unmethylated DNA. The results of experiments utilizing artificially constructed heteroduplexes of bacteriophage X DNA introduced into E. coli cells by means of CaCl2mediated transfection have shown that mismatch repair can operate on either strand of unmethylated DNA (3, 5, 7). Although, theoretically, undirected mismatch repair should have the same effect on mutation frequencies as no mismatch repair, the spontaneous mutation frequency ofdam mutants, which have undirected mismatch repair, is lower than that of mutH, mutL, mutS, or mutU single mutants or dam mut double mutants, all of which are deficient in mismatch repair (1-5, 7-9). Since polymerase errors presumably occur with roughly equal frequency in dam, mut, and dam mut cells and since the mismatch repair system does not appear to be able to distinguish mutant and wild-type strands when both are unmethylated, it may be that the action of the mismatch repair system on some fraction of the replication errors occurring in dam cells causes cell death, such that presumptive mutants are selectively lost from the population (1, 8). Such mismatch-stimulated killing would account for the findings that dam mutants are more sensitive than wild-type bacteria to base analogs, ethyl methanesulfonate, N-methyl-N'-nitro-N-nitrosoguanidine, methyl methanesulfonate, and UV irradiation, and that these sensitivities can be alleviated by the addition of a mut mutation, which renders the dam cells deficient in mismatch repair (1). The experiments reported here were designed to allow a direct demonstration of mismatch repair-dependent loss of the ability to form infective centers after transfection-i.e., inactivation-of mismatch-containing DNA heteroduplexes. Separated strands of unmethylated bacteriophage X DNA were annealed to form duplexes with and without base-pair mismatches. These DNA molecules were used to transfect wild-type and mismatch repair-deficient (mut) bacteria. The relative transfection efficiencies of X DNA duplexes with and without mismatches in wild-type and mut bacteria indicate that mismatch repair-dependent inactivation of mismatchcontaining DNA does occur. MATERIALS AND METHODS X phages with sequenced mutations in the C1 gene were obtained from Franklin Hutchinson (Yale University). Procedures for strand preparation, annealing, and transfection have been described (7). Unmethylated DNA is prepared from phages grown in dam (deficient in adenine methylation) bacteria, GM 33 (6). Fully methylated DNA is prepared from phages grown in a methylase-overproducer strain (10). [The GATC sequences in X DNA prepared from phages grown in wild-type E. coli are only -75% methylated (2).] The assay for inactivation of mismatch-containing DNA heteroduplexes-i.e., loss of the ability to form infective centers in transfection assays-involves mixing annealed DNA, with or without mismatches, with phenol-extracted DNA of X imm434. Aliquots of the mixtures are used in transfections ofwild-type and mut bacteria. Transfected cells are plated, before lysis, to form infective centers. Individual infective centers are transferred to plates seeded with C600, which plates all phages used in these experiments, and C600 (X imm434), which does not plate X imm4". Inactivation is detected as a decrease in the fraction of infective centers derived from heteroduplex DNA in transfections ofwild-type bacteria relative to transfections of mut bacteria. RESULTS AND DISCUSSION It has been suggested that undirected mismatch repair might cause cell death by making double-strand breaks in DNA (1, 8). This idea is supported by the findings that dam recA double mutants are inviable, whereas dam recA mut triple mutants are viable (8, 11). [Double-strand break repair in E. coli requires recA gene product (see ref. 12).] If double-strand breaks are the mechanism of mismatchstimulated killing, and if the mismatch repair system acts in a comparable way on X DNA heteroduplexes, the effect should be detectable as a decrease in the transfection efficiency ofunmethylated mismatch-containing heteroduplexes in transfections of wild-type cells relative to transfections of mismatch repair-deficient cells. It is conceivable that doublestrand breaks could be created by simultaneous, or nearly 2576 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Proc. Natl. Acad. Sci. USA 83 (1986) 2577 simultaneous, nuclease attack on both strands of a single mismatch or, alternatively, by overlapping excision tracts from two relatively close mismatches. To allow the detection ofheteroduplex inactivation by either mechanism, unmethylated heteroduplexes were prepared with three closely spaced repairable mismatches. The experiments reported here involve measuring the difference between transfection efficiency of mismatch-containing DNA heteroduplexes in wild-type and mismatch repair-deficient (mut) bacteria. To provide an internal standard against which efficiency can be measured, the heteroduplex DNA is mixed with phenol-extracted DNA from phages (X imm434), which can be distinguished genetically from the phages from which heteroduplex DNA is prepared. The multiplicity of transfection is low, such that infective centers contain phages derived from only one DNA molecule (7). Relative transfection efficiency is defined as the ratio of the number of heteroduplex-derived infective centers to the number of X imm434-derived infective centers. (Relative transfection efficiency depends on the particular mix of DNAs used and is therefore useful only for purposes of comparison between transfections using the same DNA mixture.) Survival is defined as the ratio of the relative transfection efficiency in wild-type cells to that in mut cells. Inactivation is defined as (1 survival). The data are presented in Table 1. Each point has been confirmed in at least one separate experiment, most often with at least one different strand preparation. The data in Table 1 (line 1) indicate that unmethylated heteroduplexes containing three mismatches are inactivated in wild-type cells relative to cells deficient in mismatch repair. To eliminate the possibility that the inactivation is due to some artifact introduced by the strand separation and annealing procedures, unmethylated DNA duplexes without mismatches were prepared from separated strands under conditions identical to those used to prepare heteroduplexes. No inactivation can be detected when these duplexes are used in the transfection mix (line 10). If the heteroduplexes are either fully adenine-methylated, which inhibits mismatch repair (2, 3), or hemimethylated, which restricts mismatch repair to the unmethylated strand (1-5), inactivation is virtually eliminated (lines 2 and 3). Thus, it appears that the mismatch repair system acts to inactivate unmethylated mismatch-containing DNA. Table 1. Inactivation of mismatch-containing DNA heteroduplexes by the E. coli mismatch repair system Transfected bacteria