Gram-negative bacterial lipopolysaccharide (LPS) is a well known stimulus for cytokine production, particularly interleukin-1 (IL-1) and tumor necrosis factor alpha (TNF alpha). Polymyxin B (PMX-B) is a cationic polypeptide that binds to LPS, neutralizing its biological effects. PMX-B also disrupts gram-negative bacterial cell membrane phospholipids but is highly toxic to mammalian cells, therefore is of limited use. PMX-B is used as additive to media, as a way to handle LPS contamination. To derive benefit from the ability of PMX-B to neutralize lipid A in vivo while avoiding its systemic toxicity, PMX-B was covalently bound to polystyrene-derivative fibers, creating a hemoperfusion column (PMX-F) for the selective removal of circulating ET. In vitro PMX-F hemoperfusion studies have demonstrated effective ET removal, using either the Limulus amebocyte lysate assay or TNF alpha production by peripheral blood mononuclear cells (PBMC) as an index of ET removal. However, the question whether PMX-B itself could stimulate human PBMC to produce cytokines has not been adequately addressed. We examined the effect of increasing concentrations of PMX-B on cytokine production by PBMC in vitro. PBMC harvested from healthy volunteers were incubated for 24 hours at 37 degrees C with control (tissue culture media RPMI), or 5 microg/ml, 10 microg/ml, 20 microg/ml or 100 microg/ml PMX-B. At the end of 24 hours, PBMC were subjected to three freeze-thaw cycles, and total TNF alpha production (pg/2.5x10(6) PBMC) was measured by radioimmunoassay. Total TNF alpha production by PBMC was 163 +/- 3 pg, 171 +/- 9 pg, 164 +/- 4 pg, 323 +/- 63 pg and 331 +/- 58 pg, in the control, PMX-B 5 microg/ml, 10 microg/ml, 20 microg/ml and 100 microg/ml conditions, respectively. Compared to controls (RPMI), the percentage increase in TNF alpha production by PBMC was 5 +/- 6% (P=0.23), 1 +/- 3% (P=0.45), 99 +/- 40% (P=0.03) and 103 +/- 36% (P=0.02) in the presence of 5 microg/ml, 10 microg/ml, 20 microg/ml and 100 microg/ml of PMX-B, respectively. Furthermore, total TNF alpha production correlated significantly with increasing concentrations of PMX-B (R=0.53, P=0.007). We conclude that the use of PMX-B in in vitro studies as an LPS-neutralizing agent, or in the experimental treatment of endotoxic or septic shock can lead to erroneous interpretations of cytokine production by PBMC, and should be used cautiously in in vitro systems at high concentrations.