J. Clin. Invest. 108: 341–347. DOI: 10.1172/JCI200113662. to a lesser extent, with their conditioned medium (5). The putative cell surface proteinase that activates the latent heparanase has not been characterized, but preliminary studies indicate that the heparanase precursor may bind to the cell surface, most likely to HS, and is then converted to its highly active 50 kDa form in a process accompanied by endocytosis of the processed form (Katz, B.-Z., et al., our unpublished results). Heparanase activity is readily obtained after transfection of mammalian cells with cDNAs encoding the entire heparanase precursor (5–10). Attempts to express the truncated 50-kDa (Lys158 to Ile543) protein failed, however, to yield active enzyme, suggesting that regions N-terminal to Lys158 are required for expression and/or function of the protein. In fact, the active enzyme has been postulated to be a heterodimer of the 50-kDa subunit noncovalently associated with an 8-kDa peptide (Gln36 to Glu109), which arises from proteolytic processing of the pre-proheparanase protein (9). It has not been determined whether association of the 50-kDa polypeptide with the 8-kDa fragment is essential for heparanase activity (9). The predicted amino acid sequence of heparanase includes six putative N-glycosylation sites, five of which cluster in the first 80 amino acids of the 50-kDa mature protein. Removal of N-glycosylation does not affect the enzyme activity (5). The sequence also contains a putative N-terminal signal peptide sequence (Met1 to Ala35) and a candidate transmembrane region (Pro515 to Ile534)(5, 6, 12). Alignment of the human, mouse, and rat heparanase amino acid sequences corresponding to the 50-kDa human mature enzyme (Lys158 to Ile543) revealed 80–93% identity (6); 61% homology was found between the recently cloned chicken heparanase (19) and the human enzyme. A prominent difference between the chicken and the mammalian enzymes resides in their signal peptide sequence, accounting for the chicken enzyme being secreted and localized in close proximity to the cell surface. In contrast, the human heparanase is mostly intracellular, localized in perinuclear granules (19). The fact that highly homologous cDNA sequences were derived from different species and types of normal and malignant cells is consistent with the notion that one dominant HS-degrading endoglycosidase is expressed by mammalian cells (5–10, 12). Thus, unlike the large number of proteases that can solubilize polypeptides in the ECM, one major heparanase appears to be used by cells to degrade the HS side chains of HSPGs. A putative heparanase 2, which shows 35% homology with the heparanase 1 described above, was recently cloned, although no enzymatic activity has been associated with this gene product (20). Unlike heparanase 1, heparanase 2 mRNA expression shows a wide distribution in normal tissues (20). Secondary structure predictions suggest that heparanase contains an (α/β) 8 TIM-barrel fold (residues 411–543), characteristic of the clan A glycosyl hydrolase families (11). Site-directed mutagenesis reveals that, as with other TIM-barrel glycosyl hydrolases, heparanase’s catalytic mechanism involves two conserved acidic residues, a putative proton donor at Glu225 and a nucleophile at Glu343. Conserved basic residues are found in proximity to the proposed cat-