The purpose of this study was to characterize the role of tumour necrosis factor (TNF) and neutrophils (PMN) in the pathogenesis of pulmonary oedema induced by endotoxin (lipopolysaccharide (LPS)). Intraperitoneal administration to BALB/c mice of 0.6-1 mg of LPS caused pulmonary oedema and lethality. This was associated with production of TNF in serum and bronchoalveolar lavage fluid and with accumulation of PMN in the lung. In this experimental model, we could block TNF production by different means: pretreatment 30 min before LPS with 4 mg/kg of i.p. chlorpromazine (CPZ), 3 mg/kg of i.p. dexamethasone (DEX), 1 g/kg p.o. of N-acetylcysteine (NAC, an antioxidant precursor of glutathione), or an anti-TNF MoAb. CPZ, DEX and anti-TNF completely prevented LPS lethality but not pulmonary oedema or pulmonary PMN infiltration, indicating that: (i) lung oedema is not the main cause of death after LPS; and (ii) lung oedema induced by LPS is not mediated by TNF. Pretreatment with NAC not only inhibited TNF production but also protected against LPS-induced pulmonary oedema, indicating that reactive oxygen intermediates are implicated. NAC also blocked TNF production in blood and in bronchoalveolar lavage. We also tested the effect of PMN depletion induced with cyclophosphamide (CP) or 5-fluorouracil (5-FU). While no pulmonary PMN infiltrate was observed in PMN-depleted mice, neutropenia did not prevent LPS lethality or oedema, indicating PMN do not play an important role in the toxic effects of LPS in this experimental model.