In humans, tumor necrosis factor (TNF) treatment has been associated with characteristic changes in circulating white blood cell populations (leukopenia followed by leukocytosis) and increased cell-surface expression of integrins. A similar pattern of effects on leukocytes occurs with granulocyte-macrophage colony-stimulating factor (GM-CSF) and G-CSF treatment. To determine whether these effects were caused directly by TNF or as a result of secondary CSF release, G-GM-, and M-CSF levels were measured after TNF infusion (9.6 x 10(6) U/mg protein; < 5.0 endotoxin U/mg protein) in cancer patients during two phase I trials of TNF. One patient with aggressive fibromatosis was treated with TNF alone (200 micrograms/m2, days 1-5 every third week) and 10 patients (four colon cancer, four head and neck cancer; one melanoma; one sarcoma) received mitomycin C (15 mg/m2, day 1) followed by TNF (60-180 micrograms/m2, days 1-3) every sixth week. All treatments were given IV, mitomycin C over 5 minutes and TNF over 2 hours. Serum samples were collected at times 0 (before mitomycin C and TNF) and 1, 2, 4, 6, 12, and 24 hours after TNF initiation on day 1 and at similar times on subsequent treatment days. M-CSF samples were analyzed by radioimmunoassay (RIA) and G-CSF and GM-CSF by ELISA. The mean baseline M-CSF levels in normal control subjects (n = 12) was 158.4 +/- 36.2 (SD) U/mL, and in pretreatment cancer patients (n = 10) 235.7 +/- 60.9 U/mL (p = 0.004, Wilcoxon test). M-CSF levels increased 4 hours after TNF initiation (mean 354.7 +/- 96.3 U/mL; p = 0.020), remained elevated at 6 hours (305.6 +/- 45.4 U/mL; p = 0.004, Wilcoxon signed-rank test), and subsequently declined. This pattern was seen in all patients treated with TNF, whether treatment was TNF alone or TNF with mitomycin C. In patients treated with mitomycin C and TNF, G-CSF levels increased at 4 hours after TNF initiation (mean 3886 +/- 2009 pg/mL; p = 0.004), remained elevated at 6 hours (mean 2140 +/- 1131 pg/mL; p = 0.004), and subsequently declined. GM-CSF levels were not measurable before or after treatment with TNF. The changes in all three endogenous cytokines were not temporally related to the previously described leukopenia and integrin upregulation on circulating leukocytes and, therefore, appear to be unrelated to this event. However, release of endogenous G-CSF and M-CSF under the influence of TNF does temporally coincide with the previously described leukocytosis, suggesting a possible role for these endogenous cytokines in the release of bone marrow cellular stores.