The beta-galactoside alpha 2,6-sialyltransferase has been localized to the trans cisternae of the Golgi apparatus and the trans Golgi network where it transfers sialic acid residues to terminal positions on N-linked oligosaccharides. It is a type II transmembrane protein possessing a 9-amino acid amino-terminal cytoplasmic tail, a 17-amino acid signal anchor domain, and a 35-amino acid stem region which tethers the large luminal catalytic domain to the membrane anchor. Previous work has demonstrated that the soluble sialytransferase catalytic domain is rapidly secreted from Chinese hamster ovary cells. These results suggest that the signals for Golgi apparatus localization do not reside in the catalytic domain of the enzyme but must reside in the cytoplasmic tail, signal anchor domain, and/or stem region. To determine which amino-terminal regions are required for Golgi apparatus localization, mutant sialyltransferase proteins were constructed by in vitro oligonucleotide-directed mutagenesis, expressed in Cos-1 cells, and localized by indirect immunofluorescence microscopy. Signal cleavage-sialyltransferase mutants which consist of only the stem and catalytic domain of the enzyme are not rapidly secreted but are retained intracellularly and predominantly localized to the Golgi apparatus. However, deletion of either the stem region or the cytoplasmic tail of the membrane-bound sialyltransferase does not alter its Golgi apparatus localization. In addition, sequential replacement of the amino acids of the sialyltransferase signal anchor domain with amino acids from the signal anchor domain of a plasma membrane protein, the influenza virus neuraminidase does not alter the Golgi apparatus localization of the sialyltransferase. These observations suggest that sequences in the signal anchor region and stem region allow the Golgi apparatus localization of the membrane-bound and soluble forms of the sialytransferase, respectively, and that both regions may contain Golgi apparatus localization signals.