The mammalian Delta6-desaturase coded by fatty acid desaturase 2 (FADS2; HSA11q12-q13.1) catalyzes the first and rate-limiting step for the biosynthesis of long-chain polyunsaturated fatty acids. FADS2 is known to act on at least five substrates, and we hypothesized that the FADS2 gene product would have Delta8-desaturase activity. Saccharomyces cerevisiae transformed with a FADS2 construct from baboon neonate liver cDNA gained the function to desaturate 11,14-eicosadienoic acid (20:2n-6) and 11,14,17-eicosatrienoic acid (20:3n-3) to yield 20:3n-6 and 20:4n-3, respectively. Competition experiments indicate that Delta8-desaturation favors activity toward 20:3n-3 over 20:2n-6 by 3-fold. Similar experiments show that Delta6-desaturase activity is favored over Delta8-desaturase activity by 7-fold and 23-fold for n-6 (18:2n-6 vs 20:2n-6) and n-3 (18:3n-3 vs 20:3n-3), respectively. In mammals, 20:3n-6 is the immediate precursor of prostaglandin E1 and thromboxane B1. 20:3n-6 and 20:4n-3 are also immediate precursors of long-chain polyunsaturated fatty acids arachidonic acid and eicosapentaenoic acid, respectively. These findings provide unequivocal molecular evidence for a novel alternative biosynthetic route to long-chain polyunsaturated fatty acids in mammals from substrates previously considered to be dead-end products.