Insulin-like growth factor (IGF)-binding protein-3 (IGFBP-3) is produced by prostate epithelial and stromal cells and either enhances or inhibits the effects of IGF on prostate epithelial cells. The levels of this protein in the male reproductive tract may be determined in part by proteases, including prostate-specific antigen (PSA), produced by the prostate epithelium. In this study we examined the proteolytic activity of human seminal fluid on IGFBP- 3. Seminal fluid and prostate massage fluid (PF) were examined for IGFBP-3 or its fragments by use of an IGFBP-3 RIA that detects intact IGFBP-3 as well as fragments, a two-site immunoradiometric assay (IRMA) that detects intact IGFBP-3 and the larger fragments, Western ligand blots (WLB), and immunoblots (WIB). In seminal fluid, IGFBP-3 was readily detectable by RIA, but was detected in only 50% of the samples assayed by IRMA. No detectable IGFBP-3 was observed by WLB with [125I]IGF-I as the ligand, but with IGF-II as the ligand, IGFBP-3 fragments at 16-17 kDa were noted. On WIB, the 16-kDa fragment of IGFBP-3 was most abundant, with a smaller amount of the 29-kDa fragment, but no intact IGFBP-3. These results indicated that most of the IGFBP-3 detected in seminal fluid was in small (≤ 16-kDa) fragments. When three or more seminal fluid samples collected 1 month apart were available from the same individual, the coefficient of variation was 10.0 ± 1.26% (± SE) for IGFBP-3 by RIA vs. 73.3 ± 11.2% for sperm counts in the same samples. In a group of 42 PF samples, the IGFBP-3 levels measured by either RIA or IRMA were approximately 3-fold higher than those in seminal fluid. Intact IGFBP-3 was detected by both WLB and WIB. There was a significant inverse correlation between PSA and IGFBP-3, measured by IRMA, in PF (r = - 0.526; P ≤ 0.004). Finally, in the PF of African-American men, PSA was significantly lower, and IGFBP-3 determined by IRMA was significantly higher compared to those in Caucasian men.