ABSTRACT. Glycosidases, which cleave the glycosidic bond between a carbohydrate and another moiety, have been classified into over 63 families. Here, a variety of computational techniques have been employed to examine three families important in normal and abnormal pathology with the aim of developing a framework for future homology modeling, experimental and other studies. Family 1 includes bacterial and archaeal enzymes as well as lactase phlorizin-hydrolase and klotho, glycosidases implicated in disaccharide intolerance II and ageing respectively. A statistical model, a hidden Markov model (HMM), for the family 1 glycosidase domain was trained and used as the basis for comparative examination of the conserved and variable sequence and structural features as well as the phylogenetic relationships between family members. Although the structures of four family 1 glycosidases have been determined, this is the first comparative examination of all these enzymes. Aspects that are unique to specific members or subfamilies (substrate binding loops) as well those common to all members (a (beta/alpha)(8)barrel fold) have been defined. Active site residues in some domains in klotho and lactase-phlorizin hydrolases differ from other members and in one instance may bind but not cleave substrate. The four invariant and most highly conserved residues are not residues implicated in catalysis and/or substrate binding. Of these, a histidine may be involved in transition state stabilization. Glucosylceramidase (family 30) and galactosylceramidase (family 59) are mutated in the lysosomal storage disorders Gaucher disease and Krabbe disease respectively. HMM-based analysis, structure prediction studies and examination of disease mutations reveals a glycosidase domain common to these two families that also occurs in some bacterial glycosidases. Similarities in the reactions catalyzed by families 30 and 59 are reflected in the presence of a structurally and functionally related (beta/alpha)(8) barrel fold related to that in family 1.

Keywords: Hidden Markov model, Klotho, ageing, Gaucher disease, Krabbe disease, protein evolution, TIM barrel, glycosidase.

Reprint requests to: I.S. Mian, Life Sciences Division (Mail Stop 29-100), Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, phone: (510) 486-6216, fax: (510) 486-6949, e-mail: Smian@lbl.gov.

ABSTRACT: Human protein S binds to C4b-binding protein (C4BP) both in plasma and in a system using purified proteins. Amino acid residues 420-434 of the first disulfide loop of the sex hormone binding globulin-like domain of protein S are involved in the interaction of protein S with C4BP. To define the involvement of specific polar amino acids within residues 420- 434, we studied in parallel synthetic protein S peptides and recombinant protein S variants containing the same amino acid replacements, K423E, E424K, Q427E and K429E. Synthetic peptide analogs of peptide PSP-420 (residues 420-434) were assayed for binding C4BP and as inhibitors of complex formation. The PSP-420 peptide and the analogous peptide with the substitution E424K, but not the peptides containing the substitutions K423E and K429E, were able to bind C4BP. Recombinant proteins with mutations of K423E, Q427E and K429E showed reduced affinity for C4BP compared to plasma protein S, recombinant wild type protein S, or E424K-protein S. These results suggest that Lys-423, Gln- 427 and Lys-429 of protein S are important for normal binding to C4BP. The anti-protein S monoclonal antibody LJ-56, raised against peptide PSP-420, recognizes only free protein S and inhibits complex formation with C4BP. Antibody LJ-56 recognized the E424K and Q427E peptides but not the K423E or K429E peptides. Similarly, the E424K and Q427E protein S mutants were recognized by LJ-56, whereas the K423E and K429E protein S mutants were not recognized. This suggests that both in the peptide PSP-420 and in protein S, Lys-423 and Lys-429 significantly contribute to binding to antibody LJ-56. These results demonstrate that protein S residues 423, 427 and 429, but not residue 424, are involved in binding to both the antibody LJ-56 and to C4BP. When peptides PSP 420 and SL-6 (residues 447-460) with carboxy- terminal amide or carboxylate moieties were compared to their ability to inhibit C4BP-protein S complexation, PSP-420-amide was the most potent. This finding together with the other results described here supports the hypothesis that the residues 420 and 434 in protein S provides a major binding site for C4BP.

Keywords: Protein S, C4b-binding protein, protein C, blood coagulation, complement.

Reprint requests to: Bonno N. Bouma, Ph.D., The Scripps Research Institute, Department of Molecular and Experimental Medicine, SBR5, 10550 North Torrey Pines Road, La Jolla, CA 92037 USA, phone: (619) 784-8220, fax: (619) 784-2243, e-mail: bnbouma@scripps.edu.

Communicated on April 13, 1998, by Bjorn Dahlbäck, M.D., University of Lund, Malmo General Hospital, Malmo, SWEDEN.