The profound clinical consequences of Gram-positive toxic shock are hypothesized to stem from excessive Th1 responses to superantigens. We used a new superantigen-sensitive transgenic model to explore the role of TCRalphabeta T cells in responses to staphylococcal enterotoxin B (SEB) in vitro and in two different in vivo models. The proliferative and cytokine responses of HLA-DR1 spleen cells were 100-fold more sensitive than controls and were entirely dependent on TCRalphabeta T cells. HLA-DR1 mice showed greater sensitivity in vivo to two doses of SEB with higher mortality and serum cytokines than controls. When d-galactosamine was used as a sensitizing agent with a single dose of SEB, HLA-DR1 mice died of toxic shock whereas controls did not. In this sensitized model of toxic shock there was a biphasic release of cytokines, including TNF-alpha, at 2 h and before death at 7 h. In both models, mortality and cytokine release at both time points were dependent on TCRalphabeta T cells. Anti-TNF-alpha pretreatment was protective against shock whereas anti-IFN gamma pretreatment and delayed anti-TNF-alpha treatment were not. Importantly, anti-TNF-alpha pretreatment inhibited the early TNF-alpha response but did not inhibit the later TNF-alpha burst, to which mortality has previously been attributed. Splenic T cells were shown definitively to be the major source of TNF-alpha during the acute cytokine response. Our results demonstrate unequivocally that TCRalphabeta T cells are critical for lethality in toxic shock but it is the early TNF-alpha response and not the later cytokine surge that mediates lethal shock.