The possibility that lymphokines such as TNF-alpha produced by CD8+ CTL are responsible for acute (short term) target cell damage induced by CTL has been debated for many years. However, the slow kinetics of TNF-induced target cell death stands in sharp contrast to the rapid target cell lysis mediated by CTL. We find that cloned CD8+ CTL activated through their TCR secrete TNF-alpha. On the other hand, our cloned CTL also have a membrane form of TNF-alpha, and they kill TNF-alpha-sensitive target cells not recognized through the TCR in a slow (18-h) lytic reaction using this surface-associated TNF-alpha. There is no secreted TNF-alpha release during this interaction. Cyclosporin A and protein synthesis inhibitors block TNF-alpha secretion, but have no effect on slow lysis mediated by the CTL. On the other hand, TNF-alpha-resistant variants are greatly resistant to slow lysis, and antibodies to TNF-alpha strongly inhibit this slow lysis. Thus, although secreted TNF-alpha does not seem to be the mechanism behind slow lysis, some form of TNF-alpha, most likely the membrane-associated form, must be involved. Not only does surface TNF-alpha appear to be biologically active in these CTL, but its expression is enhanced severalfold upon activation of the CTL through the TCR. This may be important in vivo, where surface TNF-alpha could preserve the localized nature of cytolysis and endow a CTL with an additional, albeit slower, mechanism of cell lysis. Finally, we find that although activated CTL clearly use the membrane form of TNF-alpha in slow lysis, they appear not to use TNF-alpha, in any form, during acute lysis, even under conditions in which degranulation and perforin assembly are blocked.