The routes by which solutes traverse a perturbed cerebral endothelium are, arguably, transcellular or intercellular. Rapid freezing rather than aldehyde fixation, during which cytoplasmic vesicles continue to fuse, may capture the momentary formation of either route. The only cerebral vessels thin enough and accessible for optimal freezing are pial capillaries, present in frogs but not mammals. Accordingly, the pial surface of frogs that were (a) normothermic (22 degrees C), (b) hypothermic (5 degrees C for 3-5 days), (c) normo- or hypothermic and exposed topically to hyperosmotic urea (2.5-3M) and (d) injured by a cold probe (-70 degrees C) applied to the skull were rapidly frozen. About 50 mg ferritin was infused intravenously in groups (a), (b) and (c). Thin plastic sections were examined morphometrically. The endothelium of all groups was indented by large, luminal and abluminal pits 0.08-0.32 um wide, which communicated with some cytoplasmic vesicles but did not form transcellular channels. The number of pits in the normal frogs (5/um2 of membrane) was similar to the hyperosmotic group (4/um2 of membrane) but half of that in the cold lesion groups (10/u2 of membrane). In the endothelium of the hyperosmotic groups, both normothermic and hypothermic, the intercellular clefts were patent instead of being sealed by tight junctions. Ferritin traversed such clefts to enter the perivascular basement membrane. The pits and open clefts were not autolytic features of the excised pieces of brain because they also typified hypothermic tissue. The route of passive permeation across cerebral endothelium during hyperosmotic treatment is, therefore, paracellular rather than transcellular.
