Glucosylceramide (GlcCer) is synthesized at the cytosolic surface of the Golgi complex while enzymes acting in late steps of glycosphingolipid biosynthesis have their active centers in the Golgi lumen. However, the topology of the "early" galactose-transferring enzymes is largely unknown. We used short-chain ceramides with either an 2-hydroxy fatty acid (HFA) or a normal fatty acid (NFA) to determine the topology of the galactosyltransferases involved in the formation of HFA- and NFA-galactosylceramide (GalCer), lactosylceramide (LacCer), and galabiosylceramide (Ga2Cer). Although the HFA-GalCer synthesizing activity colocalized with an ER marker, the other enzyme activities fractionated at the Golgi density of a sucrose gradient. In cell homogenates and permeabilized cells, newly synthesized short-chain GlcCer and GalCer were accessible to serum albumin, whereas LacCer and Ga2Cer were protected. From this and from the results obtained after protease treatment, and after interfering with UDP-Gal import into the Golgi, we conclude that (a) GlcCer and NFA-GalCer are synthesized in the cytosolic leaflet, while LacCer and Ga2Cer are synthesized in the lumenal leaflet of the Golgi. (b) HFA-GalCer is synthesized in the lumenal leaflet of the ER, but has rapid access to the cytosolic leaflet. (c) GlcCer, NFA-GalCer, and HFA-GalCer translocate from the cytosolic to the lumenal leaflet of the Golgi membrane. The transbilayer movement of GlcCer and NFA-GalCer in the Golgi complex is an absolute requirement for higher glycosphingolipid biosynthesis and for the cell surface expression of these monohexosyl sphingolipids.