A short-chain analogue of galactosylceramide (6-NBD-amino-hexanoyl-galactosylceramide, C6-NBD-GalCer) was inserted into the apical or the basolateral surface of MDCK cells and transcytosis was monitored by depleting the opposite cell surface of the analogue with serum albumin. In MDCK I cells 32% of the analogue from the apical surface and 9% of the analogue from the basolateral surface transcytosed to the opposite surface per hour. These numbers were very similar to the flow of membrane as calculated from published data on the rate of fluid-phase transcytosis in these cells, demonstrating that C6-NBD-GalCer acted as a marker of bulk membrane flow. It was calculated that in MDCK I cells 155 microns membrane transcytosed per cell per hour in each direction. The fourfold higher percentage transported from the apical surface is explained by the apical to basolateral surface area ratio of 1:4. In MDCK II cells, with an apical to basolateral surface ratio of 1:1, transcytosis of C6-NBD-GalCer was 25% per hour in both directions. Similar numbers were obtained from measuring the fraction of endocytosed C6-NBD-GalCer that subsequently transcytosed. Under these conditions lipid leakage across the tight junction could be excluded, and the vesicular nature of lipid transcytosis was confirmed by the observation that the process was blocked at 17 degrees C. After insertion into one surface of MDCK II cells, the glucosylceramide analogue C6-NBD-GlcCer randomly equilibrated over the two surfaces in 8 h. C6-NBD-GalCer and -GlcCer transcytosed with identical kinetics. Thus no lipid selectivity in transcytosis was observed. Whereas the mechanism by which MDCK cells maintain the different lipid compositions of the two surface domains in the absence of lipid sorting along the transcytotic pathway is unclear, newly synthesized C6-NBD-GlcCer was preferentially delivered to the apical surface of MDCK II cells as compared with C6-NBD-GalCer.