A human subject (NR) was identified whose eosinophils and neutrophils failed to respond to TNF in vitro in 29 of 33 experiments, using several biological assays. There was a response rate to TNF of 100% among 37 control subjects whose leukocytes were tested in parallel. NR serum contained an activity that inhibited the cytotoxic function of TNF- and GM-CSF-stimulated normal human eosinophils. A similar activity was detected in 4 of 122 control sera and in sera of two subjects with hypereosinophilia. This activity (ECI) had an apparent molecular weight of 80,000-100,000 and was sensitive to heating at 80 degrees C or to trypsin treatment. HPLC sizing chromatography increased the titer of ECI by a factor of 50 to 2,000 in experiments using NR serum or other sera with detectable inhibitory activity. In seven experiments using sera with no inhibitory activity, HPLC generated ECI of the same apparent molecular weight. The effect of HPLC on ECI activity required the separation of serum components and did not result from exposure to HPLC system components or other sample processing methods. This suggests that ECI in serum can be stabilized in an inactive or partially active form and that HPLC removes the stabilizing component. ECI suppressed TNF-stimulated eosinophil cytotoxic function when added to cultures up to 4 h after exposure of eosinophils to cytokine. However, ECI did not protect L929 cells from the toxic effects of TNF. Thus, ECI did not act by preventing the initial interaction of TNF with eosinophils or by interfering with the binding of TNF to its receptor on L929 cells. The results suggest that ECI is a component of a feedback mechanism that suppresses functions of cytokine-activated eosinophils in inflammation.