Lutein, a dihydroxy derivative of alpha-carotene (beta,epsilon-carotene), is the most abundant carotenoid in photosynthetic plant tissues where it plays important roles in light-harvesting complex-II structure and function. The synthesis of lutein from lycopene requires at least four distinct enzymatic reactions: beta- and epsilon-ring cyclizations and hydroxylation of each ring at the C-3 position. Three carotenoid hydroxylases have already been identified in Arabidopsis, two nonheme diiron beta-ring monooxygenases (the B1 and B2 loci) that primarily catalyze hydroxylation of the beta-ring of beta,beta-carotenoids and one heme-containing monooxygenase (CYP97C1, the LUT1 locus) that catalyzes hydroxylation of the epsilon-ring of beta,epsilon-carotenoids. In this study, we demonstrate that Arabidopsis CYP97A3 (the LUT5 locus) encodes a fourth carotenoid hydroxylase with major in vivo activity toward the beta-ring of alpha-carotene (beta,epsilon-carotene) and minor activity on the beta-rings of beta-carotene (beta,beta-carotene). A cyp97a3-null allele, lut5-1, causes an accumulation of alpha-carotene at a level equivalent to beta-carotene in wild type, which is stably incorporated into photosystems, and a 35% reduction in beta-carotene-derived xanthophylls. That lut5-1 still produces 80% of wild-type lutein levels, indicating at least one of the other carotene hydroxylases, can partially compensate for the loss of CYP97A3 activity. From these data, we propose a model for the preferred pathway for lutein synthesis in plants: ring cyclizations to form alpha-carotene, beta-ring hydroxylation of alpha-carotene by CYP97A3 to produce zeinoxanthin, followed by epsilon-ring hydroxylation of zeinoxanthin by CYP97C1 to produce lutein.