The intracellular route followed by viral envelope glycoproteins in polarized Madin-Darby canine kidney cells was studied by using temperature-sensitive mutants of vesicular stomatitis virus (VSV) and influenza, in which, at the nonpermissive temperature (39.5 degrees C), the newly synthesized glycoproteins (G proteins) and hemagglutinin (HA), respectively, are not transported out of the endoplasmic reticulum. After infection with VSV and incubation at 39.5 degrees C for 4-5 h, synchronous transfer of G protein to the plasma membrane was initiated by shifting to the permissive temperature (32.5 degrees C). Immunoelectron microscopy showed that under these conditions the protein moved to the Golgi apparatus and from there directly to a region of the lateral plasma membrane near this organelle. G protein then seemed to diffuse progressively to basal regions of the cell surface and, only after it had accumulated in the basolateral domain, it began to appear on the apical surface near the intercellular junctions. The results of these experiments indicate that the VSV G protein must be sorted before its arrival at the cell surface, and suggest that passage to the apical domain occurs only late in infection when tight junctions are no longer an effective barrier. In complementary experiments, using the temperature-sensitive mutant of influenza, cultures were first shifted from the nonpermissive temperature (39.5 degrees C) to 18.5 degrees C, to allow entrance of the glycoprotein into the Golgi apparatus (see Matlin, K.S., and K. Simons, 1983, Cell, 34:233-243). Under these conditions HA accumulated in Golgi stacks and vesicles but did not reach the plasma membrane. When the temperature was subsequently shifted to 32.5 degrees C, HA rapidly appeared in discrete regions of the apical surface near, and often directly above, the Golgi elements, and later diffused throughout this surface. To ensure that the anti-HA antibodies had access to lateral domains, monolayers were treated with a hypertonic medium to dilate the intercellular spaces. Some labeling was then observed in the lateral plasma membranes soon after the shift, but this never increased beyond 1.0 gold particle/micron, whereas characteristic densities of labeling in apical surfaces soon became much higher (approximately 10 particles/micron). Our results suggest that the bulk of HA follows a direct pathway leading from the Golgi to regions of the apical surface close to trans-Golgi cisternae.