We are attempting to understand the processes required to accurately r
eplicate the repetitive DNA sequences whose instability is associated
with several human diseases. Here we test the hypothesis that the cont
ribution of exonucleolytic proofreading to frameshift fidelity during
replication of repetitive DNA sequences diminishes as the number of re
peats in the sequence increases. The error rates of proofreading-profi
cient T7, T4, and Pyrococcus furiosis DNA polymerases are compared to
their exonuclease-deficient derivatives, for +1 and -1 base errors in
homopolymeric repeat sequences of three to eight base pairs. All three
exonuclease-deficient polymerases produce frameshift errors during sy
nthesis at rates that increase as a function of run length, suggesting
the involvement of misaligned intermediates. Their wild-type counterp
arts are all much more accurate, suggesting that the majority of the i
ntermediates are corrected by proofreading. However, the contribution
of the exonuclease to fidelity decreases substantially as the length o
f the homopolymeric run increases. For example, the exonuclease enhanc
es the frameshift fidelity of T7 DNA polymerase in a run of three A .
T base pairs by 160-fold, similar to its contribution to base substitu
tion fidelity. However, in a run of eight consecutive A . T base pairs
, the exonuclease only enhances frameshift fidelity by 7-fold. A simil
ar pattern was observed with T4 and Pfu DNA polymerases. Thus, both po
lymerase selectivity and exonucleolytic proofreading efficiency are di
minished during replication of repetitive sequences. This may place an
increased relative burden on post-replication repair processes to red
uce rates of addition and deletion mutations in organisms whose genome
contains abundant simple repeat DNA sequences.