Although it has long been recognized that lymphocytes have the capacity to reduce cortisol at the C3, C5, and C20 positions, the specificity and the physiological variation of these reactions have received little attention. We have shown that such reactions also occur with progesterone. Lymphocytes were isolated from whole blood using Percoll density gradient centrifugation. The cells were incubated for 20 h with tritiated progesterone as radioactive tracer. After extractions into ethyl acetate, the residue was subjected to high performance liquid chromatography, and the radioactivities of the separated compounds were determined. Without cells, 95-97% of the tracer added was recovered in the progesterone peak, while in the presence of 4 x 10(6) lymphocytes, this was reduced to 45-90%. The metabolites obtained included at least 10 different compounds, including those corresponding in their retention times to the neuroactive 5 alpha and 5 beta dihydroprogesterones and their 3 alpha- and 3 beta- tetrahydroprogesterone derivatives. The conversion decreased with the addition of other steroids such as testosterone, cortisol, and corticosterone, suggesting that these steroids are metabolized by the same enzymes. When the cells from two pregnant patients were combined and incubated with tracer, and with and without nonradioactive progesterone, no peaks were detected by two progesterone radioimmunoassays in the absence of added nonradioactive progesterone, while in its presence three peaks corresponding to 5 alpha-dihydroprogesterone, 3 alpha-hydroxy-5 alpha-pregnane-20-dione and 3 beta-hydroxy-5 alpha-pregnane-20-dione eluted before the P peak. Their identities were confirmed using the two different progesterone radioassays that cross-reacted with these metabolites. The highest mean conversion (44.7% +/- 3.2 SE) was found with the lymphocytes of pregnant women and with that of one lactating woman (50%). Conversions by lymphocytes of women in the follicular phase (29.3% +/- 1.3 SE) were significantly lower than those in pregnancy (P = 0.014) but did not differ significantly (P > or = 0.05) from those of women in the luteal phase (22.2% +/- 3.4 SE), those of postmenopausal women (23.5% +/- 4.9 SE), or of men (22.5% +/- 2.4 SE). Lymphocytes appear to provide a hitherto unrecognized but possibly important source of neuroactive steroids.