We characterized the release and the protease composition of high m.w. complexes released from dispersed human skin mast cells, under conditions that did not disrupt the binding of proteases to proteoglycan. The net percent release ratio of tryptase to histamine, after anti-IgE and calcium ionophore A23187 stimulation was higher than those for chymase or carboxypeptidase. This was explained by the greater cell association of carboxypeptidase and chymase, compared with tryptase, after mast cell degranulation and/or differential cosedimentation of the proteases with mast cells, because treatment of activated mast cells with 1 M NaCl increased the release ratios of chymase and carboxypeptidase more than that of tryptase. Tryptase, after release, was stable in 0.12 M NaCl and had a molecular mass of approximately 200 to 250 kDa, suggesting that it was bound to proteoglycan. We demonstrated that complexes containing chymase and carboxypeptidase were separable from tryptase-containing complexes by gel filtration and by affinity chromatography. First, on fast protein liquid chromatography, released tryptase filtered at a molecular mass of approximately 200 to 250 kDa, compared with chymase and carboxypeptidase at 400 to 560 kDa. Second, by using affinity chromatography with immobilized antitryptase mAb in 0.15 M NaCl, carboxypeptidase and chymase activities were recovered primarily in the effluent and washes of an antitryptase antibody affinity column and cofiltered at 400 to 560 kDa. Tryptase was recovered only in the eluate. Finally, by using potato tuber carboxypeptidase inhibitor-Sepharose affinity chromatography, tryptase activity was found primarily in the effluent and washes, filtered at a molecular mass of 200 kDa on fast protein liquid chromatography, and was stable in 0.12 M NaCl buffer at 37 degrees C. Carboxypeptidase and chymase activities were found primarily in the eluate. These findings suggest that tryptase and carboxypeptidase/chymase reside in distinct macromolecular complexes. Separate complexes containing these proteases may help explain previous ultrastructural observations in which the distributions of chymase and tryptase within a single granule did not always coincide.