A sensitive and specific radioimmunoassay for human interleukin-8 (IL-8) was developed using isotopically labeled homogenous natural protein. The detection limit (20% inhibition of 125I-IL-8 binding) was 30 pg/100 microliters; 50% displacement occurred at 140 pg/100 microliters. There was no cross-reactivity with the structurally and functionally related neutrophil-activating peptides 2 and 3 up to 500 ng/100 microliters. The intra- and inter-assay coefficients of variation were 4 and 7%, respectively. In vitro experiments showed that human fibroblasts triggered by interleukin-1, double-stranded RNA or virus release immunoreactive and biologically active IL-8 in a dose- and time-dependent manner. Monocytes produce immunoreactive IL-8 in the 100 ng/ml range when exposed to plant mitogen, bacterial endotoxin, virus or IL-1. Although the radioimmunoassay was more sensitive than the chemotaxis assay (detection limit 0.6 ng/ml versus 10 ng/ml) a correlation between concentrations of immunoreactive IL-8 and neutrophil chemotactic activity in the supernatants from stimulated monocytes and fibroblasts was observed. In synovial fluids from patients with inflammatory joint disease, IL-8 was clearly demonstrable, but there was no correlation between IL-8 levels and general parameters of disease activity (erythrocyte sedimentation rate and serum levels of C-reactive protein). Synovial fluids from patients with rheumatoid arthritis, seropositive for rheumatoid factor, contained significantly higher concentrations of IL-8 than synovial fluids from seronegative rheumatoid arthritis patients and patients with non-rheumatoid arthritis joint inflammation. There was a highly significant correlation between IL-8 levels and serum titers of rheumatoid factor. These findings suggest that the molecular mechanisms underlying joint inflammation may be distinct in different types of arthritis.