Two human cell lines (A549 and U937) with cytosolic thymidine phosphorylase (TP) activity were used to evaluate the potential of 5'-deoxy-5'-[F-18]fluorothymidine ([F-18]DFT) as a tracer of intracellular TP expression. Cellular metabolism of DFT led to the production of 5-[F-18]fluoro-2,5-dideoxy-D-ribose-1alpha-phosphate ([F-18]FddR-1P), in analogy to the metabolism of thymidine, which produces 2-deoxy-D-ribose-1alpha-phosphate (dR-1P). A549 cells showed the highest production rate of FddR-1P. After A549 cells were exposed to [F-18]DFT for 40 min, the relative intracellular concentration of [F-18]FddR-1P was more than sevenfold higher in cells than its precursor in the incubating medium. For the same amount of time, a twofold concentration was seen in U937 cells. However, uptake ratios did not rank with the corresponding TP activities found in cell extracts [TP activity ratio (U937:A549)=1.6] that were independently determined with a labeled thymidine/thymine cleavage assay. The discrepancy of TP activity ratios was traced to the instability of FddR-1P in cells. This was evident from the fact that cells accumulated radioactivity by producing FddR-1P, but activity also effluxed from cells over 1 h when the medium was subsequently made tracer free. The dominant labeled molecule released by cells was characterized as a neutral and lipophilic molecule, which was presumed to be a deoxynucleoside. Our results indicate that [F-18]DFT would not be effective for imaging TP expression because its initial metabolite undergoes further conversion to a diffusible secondary metabolite, allowing activity loss from cells.