The structures of cellular oligosaccharides are determined by a series of processing reactions catalyzed by Golgi glycosidases and glycosyltransferases. While there are subtle cell type differences in Golgi enzyme subcompartmentation, in general, glycosylation enzymes are localized within the Golgi cisternae in the same sequence in which they act to modify oligosaccharide substrates. The possibility that this enzyme subcompartmentation may control the types of oligosaccharides expressed by a cell has led to an interest in the signals and mechanisms directing enzyme localization in the Golgi cisternae. All glycosidases and glycosyltransferases characterized thus far have very little sequence homology that might suggest a common Golgi retention signal, but they do share a similar domain structure. They are all type II transmembrane proteins consisting of an amino terminal cytoplasmic tail, a signal anchor transmembrane domain, a stem region, and a large luminal catalytic domain. Their lack of sequence homology suggests that these proteins' Golgi retention signals are not linear amino acid sequences, but most likely involve general characteristics or conformations of larger protein domains. The peptide sequences required for Golgi retention of the N-acetylglucosaminyltransferase I (GlcNAcTI), beta 1,4-galactosyltransferase (GalT) and alpha 2,6-sialytransferase (ST) have been extensively studied. To do this, researchers created mutant and chimeric proteins, expressed these in tissue culture cells, and localized these proteins using immunofluorescence microscopy or immunoelectron microscopy. The cell surface expression of deletion mutants suggested that the deleted sequences were necessary for Golgi retention. Then, if these sequences were fused to a non-Golgi reporter protein and this chimeric or hybrid protein was retained in the Golgi, then these sequences were also sufficient for Golgi retention. Due to differences in reporter proteins used to construct these chimeric proteins, different cell types used for protein expression, different levels of protein expression, and different methods of cell surface protein detection, these experiments have led to somewhat confusing results. However, in general, it appears that the GalT relies primarily on its transmembrane domain for Golgi retention, while the GlcNAcTI and ST have requirements for their transmembrane regions, sequences flanking these regions, and luminal stem sequences. Based on these results, two potential Golgi retention mechanisms have been proposed and are now being tested. The observation that glycosyltransferase transmembrane domains are frequently sufficient for Golgi retention has led to the first of these models, the bilayer thickness model. This model proposes that the shorter transmembrane domains of Golgi proteins prevent them from entering cholesterol-rich transport vesicles destined for the plasma membrane, and that this leads to Golgi retention. The second of these models is supported by the role of multiple protein domains in the Golgi retention of some proteins. This model, the oligomerization/ kin recognition model of Golgi retention, proposes that the formation of insoluble protein homo-oligomers or very large hetero-oligomers prevents protein movement into transport vesicles destined for later compartments. Initial work suggests that the bilayer thickness mechanism may play a role in the retention of some Golgi retained proteins; however, it is not the sole retention mechanism. Other evidence suggests that an oligomerization/kin recognition mechanism may be more common, but definitive proof for its general use in Golgi protein retention is lacking. More research is required to further elucidate the sequences and particularly the mechanisms of Golgi retention. (ABSTRACT TRUNCATED)