Data on DNA repair rates of specific types of DNA lesions are very limited in humans in situ. Rate of repair of UV-induced DNA damage was followed in the skin of 17 volunteers up to 3 weeks of UV exposure, using a (32)P-postlabelling technique for the determination of specific photoproducts. The subjects of skin phototypes I and IV were exposed to 40 mJ/cm(2) of solar simulating radiation on buttock skin, and biopsies were taken at 0 h, 48 h and 3 weeks of exposure for the analysis of two cyclobutane pyrimidine dimers, TT=C and TT=T, and two 6-4 photoproducts, TT-C and TT-T, as trinucleotides. Repair rates were heterogeneous for different photoproducts. T=T dimers were repaired slower than C=T dimers, and 2.3-9.0% of the initial T=T damage remained unrepaired after 3 weeks, and was detectable in 16/17 subjects. The identity of the identified photoproducts was confirmed by a photochemical reversion assay. Damage level correlated with skin types, type I being more sensitive than type IV in an age-matched comparison. This is the first time the persistence of defined human DNA damage is demonstrated up to 3 weeks. Long-lasting DNA damage increases the likelihood of mutations.