Reliable enzymatic endothelial cell (EC) harvest methods are required for clinical EC seeding of vascular prostheses by methods analogous to those demonstrated in dogs. But crude collagenases used for EC harvest vary in efficacy and cytotoxicity, and purified collagenases reportedly give low EC yields. To compare different harvest methods, we studied growth curves of primary adult human saphenous vein EC (HSVEC) harvests plated in replicate microwell cultures. The EC yield, defined as attachment-capable ECs obtained per square centimeter of vein lumen, was estimated from the lowest number of ECs counted in lag phase before exponential growth began. With the use of morphometric studies of HSVs that were perfusion-fixed at their original dimensions, the baseline in situ density of ECs available for harvest from HSV was estimated at 1.3 X 10(5) EC/cm2. Crude (CBC) and partially purified bacterial collagenase (PBC) solutions at concentrations with equal levels of basement membrane lysis activity (BMLA) were compared by the replicate microwell method in a series of 21 harvests (six CBC, eight PBC, and seven enzyme-free control harvests). All 14 enzymatic harvests produced confluent EC cultures with no significant difference in mean harvest efficiency between CBC (12% of in situ EC number) and PBC (15%). However, PBC caused less degradation of human fibronectin (p less than 0.0001) as measured by an enzyme-linked immunosorbent assay employing a fibronectin-specific monoclonal antibody. These data suggest that chemically defined mixtures of pure enzymes with BMLA equal to the BMLA of crude collagenase might allow reliable EC harvesting without sacrifice in EC yield but with improved preservation of structures at the EC periphery. EC losses during initial vein dissection may have contributed to the low 12% to 15% efficiency we observed.