Breakdown of the blood-retinal barrier (BRB) is an early event in diabetic and galactosemic rats, but the location and nature of the specific defect(s) are controversial. Using an electron microscopic immunocytochemical technique, the retinas of normal, diabetic, and galactosemic rats were immunostained for endogenous albumin. Normal rats showed little evidence of BRB breakdown at either the inner barrier (retinal vasculature) or the outer barrier (retinal pigment epithelium) (RPE). In diabetic and galactosemic rats, as was true in human diabetics, BRB breakdown occurred predominantly at the inner BRB, but in some cases at the outer barrier as well. Treatment with the aldose reductase inhibitor sorbinil largely prevented BRB failure in galactosemic rats. In the inner retina of diabetic and galactosemic rats, albumin was frequently demonstrated on the abluminal side of the retinal capillary endothelium (RCE) in intercellular spaces, basal laminae, pericytes, ganglion cells, astrocytes, and the perinuclear cytoplasm of cells in the inner nuclear layer. Albumin did not appear to cross RCE cell junctions; however, it was occasionally seen in RCE cytoplasm of galactosemic rats. In the outer retina, albumin was frequently detected in the subretinal space, in the intercellular space between photoreceptors, and in the perinuclear cytoplasm of photoreceptor cells, but was only infrequently found in the RPE cells constituting the barrier. Albumin derived from the choroidal vasculature did not appear to cross the tight junctions of the RPE. These findings suggest that specific sites of BRB compromise are infrequent but that once albumin has crossed the RCE or RPE it freely permeates the retinal tissue by filling intercellular spaces and permeating the membranes of cells not implicated in BRB formation. The diffuse cytoplasmic staining of some RCE and RPE cells suggests that the predominant means of BRB breakdown in diabetes and galactosemia involves increased focal permeability of the surface membranes of the RCE and RPE cells rather than defective tight junctions or vesicular transport.

The kinetics and distribution of ouabain binding in retinas of Rana pipiens were examined quantitatively by scintillation counting and freeze-dry autoradiography . Binding within the proximal receptor layers was also nonuniform . Several peaks were observed, with those at the inner segment and synaptic layers being especially noticeable . Assuming an absence of glial cell binding in the proximal receptor layers, we calculated there to be 13 X 10 6 ouabain or Na + ,K+ pump sites per rod receptor . Limited measurements suggest a BmaX of -8 X 10-6 M for the inner plexiform layer .