A non-enzymatic method that was previously shown to create single-strand DNA breaks at the location of (6-4) photoproducts in the overall genome was adapted to measure (6-4) photoproducts at the level of the gene. This approach employs a photoisomerization step that converts the (6-4) photoproduct into the Dewar valence photoisomer, which is more alkaline labile and allows for the creation of a single-strand DNA break at this site. These strand breaks were quantified via Southern analysis and the DNA repair of (6-4) photoproducts was measured over 2, 4 and 8 h after a UV dose of 40 J/m2. A comparison of repair efficiency in the actively transcribed dihydrofolate reductase (DHFR) gene, a transcriptionally inactive genomic region and the overall genome (as measured by radioimmunoassay) showed preferential repair of the active gene. The active DHFR gene showed 59% repair by 8 h compared to 33% repair in the inactive downstream region. Analysis of (6-4) photoproduct repair in the transcribed and non-transcribed strands of the DHFR gene indicate some strand specificity with 62% repair in the transcribed strand at 8 h compared with 43% repair in the non-transcribed strand. However, this strand bias is much less distinct than has been reported for the major UV photoproduct, the cyclobutane pyrimidine dimer.