Monoclonal antibodies, specific for the beta A and beta B subunits of activin, were used to develop a new two-site ELISA for activin-AB. The assay had a detection limit of 0.19 ng/ml. High concentrations of activin-AB were found in bovine, ovine and porcine follicular fluids (FF), with less in human FF (1310, 1730, 688 and 7 ng/ml respectively). Recovery of spiked activin-AB standard from human, bovine and ovine FFs and from homogenized human placental extracts averaged 91%, 115%, 115% and 94% respectively. Within-plate coefficients of variation for different concentration of activin-AB were between 1.3% and 2.67%. The between-plate coefficient of variation was 5.5%. Cross-reactivity experiments showed the high specificity of the assay for activin-AB, with inhibin-A, inhibin-B, follistatin, activin-A and activin-B all cross-reacting < 0.2%. Incubation with high concentrations of follistatin (500 ng/ml) prior to assay did not affect the recovery of activin-AB. Samples of bovine, porcine, ovine and human FF gave dose responses parallel to that of the standard, as did bovine granulosa cell-conditioned media. In human and porcine FF, levels of activin-A and activin-AB were similar whereas, in bovine and ovine FF, activin-A levels were approximately threefold higher than activin-A, nearly all of the endogenous activin-AB in bovine FF was detected in the eluate from gel permeation chromatography with an M(r) of > 700000 indicating its association with higher molecular weight binding protein(s). By contrast, after denaturation, immunoreactive activin-AB was detected with an M(r) of approximately 25000 consistent with the complete dissociation from binding proteins. Activin-A was detected in relatively high concentrations in human FF (approximately 5 ng/ml), homogenized placental extracts (4.35-95.5 ng/g), sera from pregnant women (> 4 ng/ml) and amniotic fluid (3-13 ng/ml), and in much lower concentrations in postmenopausal serum (500 pg/ ml), normal cycle serum (100-200 pg/ml), serum from gonadotrophin-treated women (200 pg/ml), and normal adult male serum (225 pg/ml). Activin-A was also found in the culture media from explants of human amnion, chorion, maternal decidua and placenta. In marked contrast, activin-AB was undetectable (< 0.19 ng/ml) in all of these samples with the exception of human FF (approximately 7 ng/ml). In conclusion, we have developed a sensitive and specific ELISA to measure total (bound+free) activin-AB. Preliminary results show a more restricted distribution of this isoform compared with activin-A. The presence of high levels of both activin-A and activin-AB in FF suggests a function for both isoforms in the developing ovarian follicle.