We have evaluated the TNF production by alveolar macrophages (AM) in 43 HIV-infected subjects in relation with 1) their clinical and biologic status; 2) the presence of lung opportunistic infections (OI); and 3) the expression of HIV by AM. This production was assessed in a standard chromium release test, using monocytic U937 cells as targets. The spontaneous TNF production by AM from patients without lung OI was higher than that from seronegative controls (p less than 0.02). This production by AM was similar to that of blood monocytes, suggesting that it was not related, in these subjects, to any particular lung status. The extent of TNF release by AM was correlated to the presence of a lymphocytic alveolitis (p less than 0.05), and not to the patients' clinical presentation nor to their CD4 cell count. Finally, AM from these subjects could be normally stimulated in vitro by IFN-gamma. On the other hand, it appeared that the spontaneous TNF release by AM shown in vitro to express HIV (p24+ AM) was significantly higher than that by their p24- counterparts (p less than 0.05) and by controls (p less than 0.01). In addition, contrasting with the marked increase of TNF release by p24- AM after their stimulation with IFN-gamma (p less than 0.001), p24+ AM appeared to be refractory to any stimulation by IFN, arguing for their activation in vivo. Finally, the spontaneous TNF release by AM was significantly increased during lung OI, compared with controls (p less than 0.01) as well as with AIDS patients without OI (p less than 0.01). In addition, the production of TNF by AM in these subjects was higher than that by the corresponding blood monocytes (p less than 0.02), suggesting a compartmentalization of this response within the lungs. In conclusion, it appears that the TNF production by AM of seropositive patients is highly related to the presence of lung OI as well as to the expression of HIV by these cells. In the context of the up-regulation of HIV expression induced by TNF in vitro, our data could suggest that the in vivo release of TNF by AM could participate in viral dissemination. Moreover, we hypothesize that the generation of activated AM refractory to any further stimulation could in turn lead to the development of additional pulmonary infections.