Some Features of Base Pair Mismatch and Heterology Escherichia coli Repair in

We have used artificially constructed heteroallelic heteroduplex molecules of bacteriophage lambda DNA to transfect Escherichia coli, and E. coli mutants deficient in various functions involved in the adenine methylation-directed mismatch repair system, MutL, MutS, MutH, and lJvrD (MutU). Analysis of the allele content of single infective centers shows that this repair system often acts on several mismatches, separated by as many as 2000 bp, on one of the strands of a heteroduplex molecule. When the methyl-directed mismatch repair system is disabled by mutH or uurD mutations, localized mismatch repair becomes prominent. This prominent localized repair that can result in separation of very closely linked markers requires the functions MutL and MutS, is independent of adenine methylation, and appears to reflect another mechanism of mismatch repair. Heterologycontaining heteroduplex molecules with a deletion in one strand often escape processing. However, when the heterology includes the stem and loop structure of a transposon, TnlO, the transposon is lost. HE demonstration of in vivo processing of misT matched base pairs in the duplex structure of DNA has provided insight into features of the fidelity of replication as well as the clustering of genetic exchanges. In the case of replication fidelity, it has been suggested (WAGNER and MESELSON 1976) that mismatched bases introduced as replication errors on a newly synthesized DNA strand are preferentially excised and the gaps filled by repair synthesis. The incorrect nucleotide is thus removed. In the case of clustering of genetic exchanges, it has been shown that several closely linked parental markers can be included in a region of heteroduplex DNA created by recombination (WHITE and Fox 1974). Localized excision of mismatches and repair synthesis from the complementary strand template could result in products that display clustered coincident exchanges. Evidence for a role for mismatch repair in replication fidelity comes from the characterization of mutations that simultaneously result in striking elevation of mutation frequency (GLICKMAN and RADMAN 1980) and defects in the repair process (Lu, CLARK, and MODRICH 1983). The functions, so far recognized, that play roles in this mismatch repair system are MutL, MutS, MutH, and UvrD (MutU). The repair has been shown to be directed by the absence of adenine methylation at GATC dum methylation sites (PUKKILA et al. 1983) and can involve replacement of a thousand or more nucleotides. The efficiency with which repair occurs is dictated by the nature of the base pair mismatch (DOHET, WAGNER and RADMAN 1985). Since adenine methylation occurs after repliGenetics 117: 381-390 (November 1987) cation, errors in a newly synthesized strand of DNA would be subject to correction, thus contributing to replication fidelity. This subject has recently been reviewed by RADMAN and WAGNER (1986), CLAVERYS and LACKS (1 986) and MFSELSON (1 987). The DNA products of genetic recombination include regions in which the two strands of the duplex are contributed by different parental DNA molecules (FOX 1978). The presence of mutational sites distinguishing the two parents, within such heteroduplex regions, could result in either base pair mismatches (WHITE and Fox 1974) or regions of nonhomology (LICHTEN and Fox 1984) within the DNA duplex. Repair processes operating on such structures play a role in determining the final outcome of the recombination process. In an effort to elaborate the rules that dictate the processing of base pair mismatches and regions of sequence nonhomology, we have examined the products of transfection of E. coli with bacteriophage lambda DNA molecules harboring such structures. The hosts for transfection included mut+ bacteria as well as various mutants defective in methyl directed mismatch repair: mutL, mutS, mutH, and uvrD. The phage genotypes present among the products emerging from bacteria transfected with artificially constructed heteroduplex molecules reveal some of the features of the processing events that the DNA molecules have experienced. Substantial differences are evident in the products of processing that are observed when various bacterial mutator functions are disabled. When the heterodu382 S. Raposa and M. S. Fox plex molecules harbor a set of three base pair mismatches covering a span of about 2000 bp, the pattern of allele loss evident in the products of single bursts from mu2 + bacteria seems to reflect frequent coincident loss of all three alleles from one or the other strand. With strains defective in mutL or mutS, little processing is evident; the phage products in most of the infective centers include all of the alleles present in the parental molecule. In contrast, with strains defective in the MutH or UvrD functions, much of the substantial processing that is evident appears to reflect the repair of mismatches one at a time so that most bursts are mixed for at least one of the allelic pairs present in the parental molecule. The localized processing that is prominent in mutH and uvrD strains appears to be independent of dam-directed adenine methylation and requires the activities of both the MutL and MutS functions. The residual repair that is evident in mutL and mutS strains remains to be accounted for. There appear, therefore, to be at least two distinct mechanisms that result in repair of mismatched base pairs (FOX and RAPOSA 1983; LIEB 1983). For the case of heteroduplex molecules harboring a non-homology reflecting the presence of a deletion in one of the parental strands, bursts often include alleles from both parental phage. This is true in all of the bacterial strains examined. However, heteroduplex molecules harboring a heterology resulting from the presence of a transposon, TnlO, in one of the strands, experience a different fate. The transposon is lost in most or all of the products of heteroduplex transfection. MATERIALS AND METHODS T h e E. coli and bacteriophage lambda strains that were used are described in Table 1. P I transductions followed the procedures described by MILLER (1972). T h e mut double mutants were constructed by P1 transduction, selecting for the drug resistant markers of the appropriate mutatorinserted transposon. T h e presence of the second mutator mutation was confirmed by P1 back-crosses using the candidate double mutants as donors. Mutator activity was monitored by the frequency of trimethoprim-resistant (MILLER 1972) o r lambda-resistant bacteria. Standard methods for growing lambda in supplemented lambda broth and plating on trypticase agar have been described elsewhere (WHITE and FOX 1974). Phage stocks were grown in BNN45 lacA using the NZC broth method (BLATTNER et aE. 1977). Phage concentration with polyethylene glycol and subsequent CsCl gradient purification have been described previously by LICHTEN and Fox (1983). In this case, however, the densities of the CsCl steps in the step gradient were 1.6, 1.4 and 1.3 g/ml. DNA denaturation and strand separation with poly U, G were carried out by the method of DAVIS, BOTSTEIN and ROTH (1 980). T h e CsCl gradients of denatured DNA were collected by puncturing the bottoms of polyallomer tubes and collecting 2-drop fractions into Eppendorf tubes. Five microliters of each fraction and an equal volume of ethidium bromide (1 mg/ml) were spotted on a piece of parafilm and photographed under UV light to determine the positions of the heavy and light peaks. Each fraction from the two peaks was diluted 50-fold in T E buffer (0.01 M Tris, 0.001 M EDTA) and the optical density at 260 nm measured. T o reduce contamination of material in the light peak by heavy strands, the light fractions were pooled, renatured and rerun on a second CsCl gradient according to the procedure of NEVERS and SPATZ (1 975). T h e purity of each fraction was assessed by measuring biological activity in a transfection experiment before and after self-annealing. T h e yield of infective centers when complementary strands were annealed was 300 times greater than with unannealed or self-annealed DNA. Fractions from each peak were pooled and stored at 4" . Before transfection, equal volumes of the separated strands were hybridized in 50% formamide at 26" for 1 hr. Undermethylated and hypermethylated DNAs were isolated from phage grown on a dam13::Tn9 derivative of BNN45 and a BNN45 derivative harboring the Dam overproducing plasmid p T P l 6 6 (MARINUS, POTEETE and ARRAJ 1984), respectively. T h e level of methylation was confirmed by susceptibility to restriction by MboI. For those experiments involving the transposon-harboring heteroduplex molecules and those examining the impact of adenine methylation, mixed random heteroduplex molecules were used. They were prepared by mixing purified phage in the appropriate ratios to give a total DNA concentration of 100 pg/ml in SM (50 mM Tris, pH 8, 100 mM NaCI, 10 mM MgS04). T h e mixture was made 20 mM with EDTA (pH 8.0) and heated to 65" for 15 min. NaOH was added to a final concentration of 0.1 M and the mixture allowed to stand at room temperature for 10 min. Tris-HC1 (pH 7.2) was added to a final concentration of 0.2 M and the solution was mixed with an equal volume of formamide. T h e mixtures were incubated at 45" for 1 hr. Sham preparations were prepared by the same protocol, with the NaOH and Tris mixed before addition to the DNA. Transfection was carried out using a modification of the method of MANDEL and HICA (1970). Overnight cultures of bacteria in LB were diluted 100-fold in LB and incubated at 37". When the cells reached a density of 3 X 10s/ml, they were pelleted in a refrigerated centrifuge, suspended in an equal volume of cold 0.1 M MgC12, repelleted and suspended in % volume of cold 0.1 M CaCI2, and held on ice for 20 min. T h e cells were pelleted again, resuspended in %o volume of cold 0.1 M CaClp, and used immediately for transfection. An aliquot of 0.2 ml competent cells and 0.1 ml lambda heteroduplex DNA (1 pg/ml) was mixed and placed in ice water for 15 min, then at 37" for 20 min. Appropriate dilutions were plated as infective centers. Isolated plaques were picked with a capillary tube and resuspended in 1 ml SM buffer. The contents of individual plaques were replated and the genotypes of phage present were determined in the following manner: the nonpermissive indicator, MI82 (suf+), was used to detect the presence of wild type (P') phage, since neither Pam3 nor Pam80 mutants can grow on this host. T h e Pam80 allele was detected on a plate with a lawn of Q L (sul l l ) that is permissive for Pam80 mutants but not for Pam3 phage. Indicator plates with a lawn of D6431 (sul) were used to discriminate between Pam3 and Pam80 phage, since Pam80 mutants grow very poorly while Pam3 phage grow very well on this indicator. T h e double mutant, Pam3 PamBO, was not detectable by these assays. A permissive indicator plate with D6432 (sul l ) was also spotted with the resuspended phage to score for the c l allele ( i . e . , clear, turbid and mottled spots). As a confirming check, Mismatch and Heterology Repair