To determine whether base-to-base variations in the rate of excision repair influence the distribution of mutations, we have developed a method to measure UV photoproducts at individual nucleotides in the Escherichia coli chromosome. Specific gene fragments are 3' end-labeled using a sequence-specific oligonucleotide to direct the site of labeling, and photoproducts are identified by enzymatic incision. On the nontranscribed strand of the E. coli lacI gene, the cyclobutane pyrimidine dimer frequency was 2- to 8-fold higher in chromosomal DNA than in a cloned DNA fragment. The chromosomal lesion frequency corresponded to the frequency of UV-induced mutations at mutation hot spots reported in the literature. Only 0-30% of cyclobutane dimers at various sites on this strand were excised in 20 min. In contrast, repair on the transcribed strand was 80-90% complete in 20 min. However, the transcribed strand contained an excision repair "slow spot" at the site of its single mutation hot spot: At this site, no repair occurred for the first 10 min, after which repair proceeded more slowly than typical of that strand. In an mfd strain, deficient in a factor that couples repair to transcription in cell extracts, the excision rate at individual nucleotides on the transcribed strand was minimal at most sites for at least 30 min. Wild-type E. coli's bias for producing mutations at photoproducts on the nontranscribed strand, reported to require the mfd gene, therefore appears to be due to an excision repair system specific for the transcribed strand of chromosomal DNA.