Inhibition of in vitro cytokine production by human peripheral blood mononuclear cells treated with xenobiotics: implications for the prediction of general toxicity and immunotoxicity.

The use of human peripheral blood mononuclear cells (PBMC) as an in vitro system to predict in vivo toxicity was investigated. For 58 chemicals, the effect on cytokine secretion (IL-5, IFNgamma and TNFalpha) by phytohaemagglutinin-activated PBMC was measured, IC50 values were calculated and correlations of these endpoints with human LC50 values were determined. The best result was obtained with IFNgamma as an endpoint for which the calculated R(2) value was 0.58 which is comparable with the R(2) values for the classical neutral red uptake (NRU) assays using murine 3T3 cells and normal human keratinocytes (R(2)=0.56 and 0.59, respectively). When for each chemical the lowest IC50 value of the three endpoints was correlated with LC50 the calculated R(2) increased slightly to 0.63. A specific strength of our test is that it corrects several outliers (diazepam, digoxin, malathion and verapamil hydrochloride) which do not fit in the linear regression analysis for IC50 values obtained with the classical 3T3 NRU assay. Furthermore, 2,4-dichlorophenoxyacetic acid, cyclosporine A and pentachlorophenol had a 10 times lower IC50 value than the estimated human LC50 value and were identified as immunotoxic alerts. In conclusion, new endpoints investigated in this study contribute to the prediction of immunotoxic effects and correct outliers of classical cytotoxicity assays.