Several human hereditary neuromuscular disease genes are associated with the expansion of CTG or CGG triplet repeats. The DNA syntheses of CTG triplets ranging from 17 to 180 and CGG repeats from 9 to 160 repeats in length were studied in vitro. Primer extensions using the Klenow fragment of DNA polymerase I, the modified T7 DNA polymerase (Sequenase), or the human DNA polymerase β paused strongly at specific loci in the CTG repeats. The pausings were abolished by heating at 70°C. As the length of the triplet repeats in duplex DNA, but not in single-stranded DNA, was increased, the magnitude of pausings increased. The location of the pause sites was determined by the distance between the site of primer hybridization and the beginning of the triplet repeats. CGG triplet repeats also showed similar, but not identical, patterns of pausings. These results indicate that appropriate lengths of the triplets adopt a non-B conformation(s) that blocks DNA polymerase progression; the resultant idling polymerase may catalyze slippages to give expanded sequences and hence provide the molecular basis for this non-Mendelian genetic process. These mechanisms, if present in human cells, may be related to the etiology of certain neuromuscular diseases such as myotonic dystrophy and Fragile X syndrome.

Long CTG triplet repeats which are associated with several human hereditary neuromuscular disease genes are stabilized in ColE1-derived plasmids in Escherichia coli containing mutations in the methyl-directed mismatch repair genes (mutS, mutL, or mutH). When plasmids containing (CTG)180 were grown for about 100 generations in mutS, mutL, or mutH strains, 60-85% of the plasmids contained a full-length repeat, whereas in the parent strain only about 20% of the plasmids contained the full-length repeat. The deletions occur only in the (CTG)180 insert, not in DNA flanking the repeat. While many products of the deletions are heterogeneous in length, preferential deletion products of about 140, 100, 60, and 20 repeats were observed. We propose that the E. coli mismatch repair proteins recognize three-base loops formed during replication and then generate long single-stranded gaps where stable hairpin structures may form which can be bypassed by DNA polymerase during the resynthesis of duplex DNA. Similar studies were conducted with plasmids containing CGG repeats; no stabilization of these triplets was found in the mismatch repair mutants. Since prokaryotic and human mismatch repair proteins are similar, and since several carcinoma cell lines which are defective in mismatch repair show instability of simple DNA microsatellites, these mechanistic investigations in a bacterial cell may provide insights into the molecular basis for some human genetic diseases.