We evaluated P388D1 macrophagelike cells as model host cells for studying the intracellular survival and strain-specific virulence of Histoplasma capsulatum. Previously characterized strains which were virulent for mice destroyed monolayers of these cells within a few days. In contrast, related avirulent "smooth" variants failed to do so even after 20 days, although they persisted within P388D1 cells for at least 7 days. On the basis of this observation, we developed a quantitative radiolabel assay to use as an initial screen for virulence. Another cell type lining the respiratory tract was then examined as a potential host for H. capsulatum. Hamster trachea epithelial (HTE) cells readily internalized a variety of strains lacking alpha-(1,3)-glucan in their cell walls; however, the tracheal cells were only rarely infected by organisms possessing this polysaccharide. We subsequently inoculated HTE cells with alpha-(1,3)-glucan-positive strains and enriched for the few yeasts infecting these cells. The progeny resembled smooth variants in terms of colony morphology, the absence of alpha-(1,3)-glucan in their cell walls, and their inability to kill macrophages. Did the HTE cells select for these variant yeasts from the parent inoculum or instigate a change from the parental phenotype? Following a 3-h uptake period, only 2% of the ingested yeasts lacked alpha-(1,3)-glucan. One day later, nearly half of the intracellular organisms lacked this polymer. This rapid conversion of a large proportion of the inoculum suggests some type of environmentally triggered change, perhaps analogous to phase variation seen in many other pathogens. Infection of epithelial cells or some other nonprofessional phagocyte during natural histoplasmosis might give rise to similar variants, thus establishing a reservoir of organisms capable of causing chronic or latent infections.