The localization of the cell adhesion molecules L1, neural cell adhesion molecule (N-CAM), and myelin-associated glycoprotein (MAG) was studied immunohistologically at the light and electron microscopic levels and immunochemically in the developing and adult mouse optic nerve and retina. The neural adhesion molecule L1 is strongly expressed on the shafts of fasciculating unmyelinated axons at all ages studied from embryonic day 15 through adulthood. Growth cones of retinal ganglion cell axons were weakly L1-positive or L1-negative when contacting glial cells. Unmyelinated axons were not only L1-positive when contacting each other but also when contacting glia, whereas contacts between glial cells were L1-negative at all developmental unmyelinated retinal nerve fiber layer or in the unmyelinated optic nerve head became L1-negative when enwrapped by myelin in the optic nerve proper. At all stages of development N-CAM showed profuse labeling on fasciculating axons, growth cones, and their contact sites with glial cells as well as contacts between glial cells. In contrast to L1, axons remained N-CAM-positive when becoming myelinated. Sometimes, N-CAM was found in compact myelin. However, N-CAM was absent from glial surfaces contacting basement membranes at the interface to meninges, blood vessels, and the vitreous body of the eye. MAG was first detectable intracellularly in oligodendrocytes associated with the endoplasmic reticulum and Golgi apparatus before it became apparent at the cell surface. There it was present on oligodendrocytes prior and during the first stages of ensheathment of axons, both on cell body and processes. After formation of compact myelin MAG remained strongly expressed periaxonally and was only weakly detectable in noncompacted myelin including inner mesaxon and paranodal loops. None of the adhesion molecules was detectable on extracellular matrix, in the meninges, or on endothelial cells. Immunochemical analysis of antigen expression at different developmental stages was in agreement with the immunohistological data. We infer from these observations that L1 is involved in stabilization not only of axon-axon, but also axon-glia contacts, while the more dynamic structure of the growth cone generally expresses less L1. A differential expression of L1 along the course of an axon--being present on its unmyelinated, but absent on its myelinated part--further supports the notion that L1 may be involved in the stabilization of axonal fascicles but not of axon-myelin contacts.(ABSTRACT TRUNCATED AT 400 WORDS)