We examined the relationship between the fine structure of lipid A and the toxicity of endotoxin or lipopolysaccharides as measured by the Limulus amebocyte lysate (LAL), rabbit pyrogenicity, chicken embryo lethal dose, and dermal Shwartzman reaction tests. Lipid A and lipid A-like compounds obtained from deep-rough mutants of Salmonella spp. and Escherichia coli had a wide range of structural variations. These compounds included native lipopolysaccharides, diphosphoryl and monophosphoryl lipid A's, and lipid X (a monosaccharide). The LAL test was positive for all lipids tested with lysates from Travenol Laboratories and from Associates of Cape Cod (2.9 X 10(3) to 2.6 X 10(7) endotoxin units per mg), except for O-deacylated and dephosphorylated lipid X, which were negative. The Mallinckrodt lysate gave negative tests for lipid X. In the rabbit pyrogenicity and chicken embryo lethal dose tests, only native lipopolysaccharide and diphosphoryl lipid A's were judged toxic. The Shwartzman reaction was positive for a specific purified diphosphoryl lipid A (thin-layer chromatography-3 fraction) but negative for the purified monophosphoryl lipid A (also a thin-layer chromatography-3 fraction). These results show that the LAL test is not a valid measure of all parameters of toxicity of a lipid A or lipid A-like compound and can yield false-positive results. However, these findings are not in conflict with the widespread use of the LAL assay for pyrogens in the pharmaceutical industry since a good correlation exists between LAL results and pyrogenicity when undegraded endotoxin is evaluated in parallel assays.