A monoclonal antibody made against a 135-kD glycoprotein (gp135) on the plasma membrane of Madin-Darby canine kidney (MDCK) cells was used to study the development and maintenance of epithelial cell surface polarity. Immunofluorescence microscopy and immunogold electron microscopy of confluent monolayers demonstrated that gp135 had a polarized cell surface distribution and was only localized on the apical surface. The role of membrane contacts in establishing gp135 polarity was determined by plating cells in low Ca++-medium to prevent the formation of intercellular junctions. Quantitative immunogold electron microscopy demonstrated that gp135 had a polarized distribution on cells lacking membrane contacts and was observed on the apical surface at a density 24 times that of the basal membrane contacting the substratum. The possibility that gp135 was associated with components of the apical cytoskeleton was investigated using cytoskeleton-disrupting drugs. Incubation in cytochalasin D produced a clustering of both actin and gp135, and double-label fluorescence microscopy demonstrated that these proteins were colocalized. Experiments using nocodazole had no effect, suggesting that gp135 could be interacting with actin microfilaments, but not microtubules. Treatment with Triton X-100 extracted approximately 50% of the gp135 and immunofluorescence microscopy indicated that the gp135 which remained associated with the detergent-insoluble cytoskeleton had a distribution identical to that of control cells. Experiments demonstrating that gp23, a nonpolarized glycoprotein, was preferentially extracted from the apical membrane suggested that the improperly sorted apical gp23 did not interact with the cytoskeleton. These results provided evidence that the polarized cell surface distribution of gp135 was maintained through its interaction with actin in the apical cytoskeleton.