We have attempted to isolate samples of apical and basal-lateral plasma membranes from cultured fetal human RPE. Cells from confluent, dome-forming cultures were disrupted with a Dounce apparatus. Nuclei and melanin granules were sedimented by centrifugation at 2600 g for 10 min. The supernates were layered over gradients of 17.5-65% sorbitol and centrifuged at 122,000 g for 5 hr. Fractions were grouped into "density windows" on the basis of their biochemical marker contents. Na,K-ATPase and alkaline phosphatase overlapped but did not precisely parallel one another, suggesting associations with two partially separated membrane populations; in density window I, alkaline phosphatase was enriched 4.3-fold, and Na,K-ATPase was enriched 1.7-fold, whereas in window II the corresponding enrichment factors were 7.7 and 6.7. These markers were well resolved from a mitochondrial marker, but they were overlapped by endoplasmic reticulum and Golgi markers. Additional density gradient centrifugations, performed after samples had been suspended in 55% sorbitol, further separated alkaline phosphatase- and Na,K-ATPase-containing membranes from endoplasmic reticulum and Golgi membranes, yielding alkaline phosphatase and Na,K-ATPase cumulative enrichment factors of 6.8 and 2.5 for the sample from window I and 9.3 and 10.9 for the sample from window II. Subsequent phase partitioning analysis of the sample from window I further enriched an alkaline-phosphatase-rich membrane population, which is believed to represent the RPE basal-lateral membranes. The sample from density window II contained two membrane populations, both enriched in Na,K-ATPase, alkaline phosphatase, and galactosyltransferase, and both of which appear to be derived from the apical plasma membrane. SDS-PAGE and Western blotting confirmed a correlation between Na,K-ATPase catalytic activity and Na,K-ATPase alpha subunit immunoreactivity.