Lipid X, an early precursor in the biosynthesis of lipid A has been reported to directly induce cytokine release in macrophages but also to inhibit endotoxin-induced tumor necrosis factor (TNF) induction. In this report we provide evidence that these conflicting results could be due to contaminants present in different batches of lipid X used. Thus, in an apparently pure batch of crystalline lipid X as obtained by a published procedure (Macher, I. (1987) Carbohydr. Res. 262, 79-84) small amounts of N,O-acylated disaccharide-1-phosphates could be identified. Their isolation was achieved by gel filtration on Sephadex LH-20 and further analysis of fractions showing elevated limulus amebocyte lysate values by thin layer chromatography and reverse-phase high performance liquid chromatography (HPLC) in combination with bioassays. Identification of immunostimulatory by-products was possible by testing HPLC-fractions for TNF-induction in bone marrow-derived mouse macrophages. Applying these procedures a disaccharide-1-phosphate, containing four 3(R)-hydroxymyristic acids at positions 2, 3, 2', 3', was identified as the main immunostimulatory side product. Two isomeric hydrolysis products of this compound with only three 3(R)-hydroxymyristic acid moieties attached to the disaccharide-1-phosphate were also identified. Surprisingly, these compounds behave quite differently in the TNF induction test. The disaccharide-1-phosphate, acylated at positions 2, 2', 3', is a very potent inducer of TNF-release whereas the corresponding isomer containing the 3(R)-hydroxymyristic acids in positions 2, 3, 2', does not induce TNF release, but strongly inhibits TNF release as induced by the former compound. Thus, contamination of "pure" lipid X with immunostimulatory or immunoinhibitory impurities may explain the divergent pharmacological profiles which were attributed to synthetic lipid X.