The mutD (dnaQ) gene of Escherichia coli codes for the proofreading activity of DNA polymerase III. The very strong mutator phenotype of mutD5 strains seems to indicate that their postreplicational mismatch repair activity is also impaired. We show that the mismatch repair system of mutDS strains is functional but saturated, presumably by the excess of DNA replication errors, since it is recovered by inhibiting chromosomal DNA replication. This recovery depends on de novo protein synthesis.The mechanisms ensuring accuracy of DNA replication have been elucidated by study of mutator mutants affecting specific accuracy functions (for a review, see reference 4). In Escherichia coli, mutator mutations mapping in the dnaE gene presumably affect nucleotide selection by DNA polymerase III (28). Mutations mapping in the dnaQ gene (also called mutD) affect the proofreading 3'->5' exonuclease function of the DNA polymerase III holoenzyme (6, 8), whereas mutH, mutL, mutS, and mutU mutations cause defects in methyl-directed postreplicative mismatch correction (for reviews, see references 21 and 24). mutD5 and dnaQ49 mutants are the most potent mutator strains known. All classes of transition, transversion, and frameshift mutations are increased up to 105-fold when mutD5 strains are grown in rich medium (5, 10, 11). The mutD5 gene of E. coli encodes the epsilon subunit, which carries out the 3'-*5' proofreading exonuclease function of the DNA polymerase III holoenzyme (6,8). However, the high-mutation-rate effect of mutD5 strains cannot be explained only by the defect in exonuclease activity, as this mutation rate is comparable to the error rate of in vitro replication with a polymerase devoid of proofreading activity (9, 16). The strong mutator phenotype of mutD5 suggested that mismatch repair functions may be impaired as well. This prediction was supported by results from DNA heteroduplex transfection experiments (23,25,26). One hypothesis is that the high error rate in DNA replication saturates mismatch repair in mutD5 mutants (23). Therefore, we have tested the mismatch repair capacity of mutD5 strains under conditions in which chromosomal DNA replication was or was not allowed. Mismatch repair activity was determined by the extent of pure infective centers derived from infection with packaged hemimethylated heteroduplexes of lambda DNA. Under these conditions, bacteria proficient in mismatch repair produced essentially pure infective centers, whereas bacteria deficient in mismatch repair produced mostly mixed infective centers.To evaluate mismatch repair activity, pure hemimethy-* Corresponding author. t Present address: Institute of Molecular Biology, University of Oregon, Eugene, OR 97403. lated heteroduplexes of lambda DNA containing an A. C or a G T mismatch were artificially constructed and introduced into mutD5 and other appropriate strains. Hemimethylated heteroduplexes of lambda DNA with defined mismatches were prepared as previously described (19) by reannealing the separated DNA strands of a lambda mutant carry...