Salmonella Infections in naturally susceptible mice grow rapidly, with death occurring only after bacterial numbers in vivo have reached a high threshold level, commonly called the lethal load. Despite much speculation, no direct evidence has been available to substantiate a role for any candidate bacterial components in causing death, One of the most likely candidates for the lethal toxin in salmonellosis is endotoxin, specifically the lipid A domain of the lipopolysaccharide (LPS) molecule. Consequently, we have constructed a Salmonella mutant with a deletion-insertion in its waaN gene, which encodes the enzyme that catalyses one of the two secondary acylation reactions that complete lipid A biosynthesis. The mutant biosynthesizes a lipid A molecule lacking a single fatty acyl chain and Is consequently less able to induce cytokins and inducible nitric oxide synthase (iNOS) responses both in vivo and in vitro. The mutant bacteria appear healthy, are not sensitive to increased growth temperature and synthesize a full-length O-antigen-containing LPS molecule lacking only the expected secondary acyl chain. On intravenous inoculation Into susceptible BALB/c mice, wild-type salmonellae grew at the expected rate of approximately 10-fold per day in livers and spleens and caused the death of the infected mice when lethal loads of approximately 108 were attained in these organs. Somewhat unexpectedly, waaN mutant bacteria grew at exactly the same rate as wild-type bacteria in BALB/c mice but, when counts reached 108 per organ, mice infected with mutant bacteria survived. Bacterial growth continued until unprecedentedly high counts of 109 per organ were attained, when approximately 10% of the mice died. Most of the animals carrying these high bacterial loads survived, and the bacteria were slowly cleared from the organs. These experiments provide the first direct evidence that death in a mouse typhoid infection is directly dependent on the toxicity of lipid A and suggest that this may be mediated via proinflammatory cytokine and/or iNOS responses.