This study has investigated the metabolic modification of [3-14C]docosatetraenoate (22:4(n-6)) and [3-14C]docosapentaenoate (22:5(n-3)) by human cells in culture. Fetal skin fibroblasts converted as much as 20% of the incorporated [14C]22:4(n-6) to [14C]20:4(n-6) within 6 h and 41% within 48 h. Retroconversion of incorporated [14C]22:5(n-3) was less than 13% at all time points. Chain shortening of [14C]22:4(n-6) was also 2-6-fold greater than that of [14C]22:5(n-3) in retinoblastoma and vascular endothelial cells. Fibroblasts, vascular endothelial cells and retinoblastoma cells all elongated substantially more [14C]eicosapentaenoate than [14C]arachidonate to the respective C22 fatty acids. Within 3-4 days, fibroblasts incubated with either [14C]20:5(n-3) or [14C]22:5(n-3) had the same ratio of radiolabeled C22:C20 fatty acids in cellular glycerolipids. By contrast, the cells incubated with [14C]22:4(n-6) or [14C]20:4(n-6) did not reach a common C22/C20 equilibrium by 5 days. Although fibroblasts were found to desaturate [14C]22:5(n-3), a substantial lag time was observed; [14C]22:6(n-3) was 2% at 48 h and 20% at 96 h. By contrast, synthesis of [14C]22:6(n-3) by retinoblastoma cells was 51% within 6 h and greater than 90% at 96 h. Desaturation of [14C]22:4(n-6) was observed in retinoblastoma cells, but not in fibroblasts. These results thus suggest that the ratio of C22C20 polyunsaturated fatty acids in cells is regulated by the relative rates of retroconversion and chain elongation, with the net effect of the two processes favoring C20 for n-6 and C22 for the n-3 fatty acids. Furthermore, although fibroblasts desaturate [14C]22:5(n-3), the process appears to be qualitatively different from that of retinoblastoma cells.