APC cytoprotective activities mediated by the cytoprotective protein C pathway are independent of APC anticoagulant activity. APC anticoagulant activity requires binding of the APC Gla-domain (green) (A) to negatively charged phospholipids on cell and/or platelet membranes or microparticles, whereas APC cytoprotective activity requires binding of the APC Gla-domain (green) to EPCR¹⁰⁹  (purple) (C). EPCR residues important for protein C/APC binding, as determined by alanine mutagenesis, are indicated in red (C).¹¹⁰  Interactions of APC with its macromolecular substrates (factor Va, factor VIIIa, and PAR-1) are dictated by exosite interactions of APC protease domain surface loops (A,C). Important exosite loops in APC for interactions with factor Va are the autolysis loop, the 37-loop (yellow), and the calcium-binding loop (yellow). In particular, residues Lys191, Lys192, and Lys193 (red; A,C) in the 37-loop and residues Arg229 and Arg230 (red; A,C) in the calcium-binding loop contribute significantly to interactions with factor Va, and mutation of these residues to Ala severely compromises APC anticoagulant activity (B). Exosite specificity is illustrated by the fact that both the 37-loop residues (red/yellow) and the calcium-binding loop residues (red/yellow) (A,C) are required for normal cleavage of factor Va at Arg506 and thus for normal anticoagulant activity (B), while they are not involved in proteolytic activation of PAR-1 that generates antiapoptotic, cytoprotective activity (D). Based on the exosite specificity, a molecular engineering approach of APC surface loops was used to design APC variants with reduced risk of bleeding due to reduced anticoagulant activity but with full cytoprotective activity.²⁹  Two such APC variants, 3K3A-APC and 229/230-APC, displayed increased ratios of cytoprotective (antiapoptotic) to anticoagulant activities (7- to 25-fold) compared with the reference ratio of 1.0 for recombinant wild-type APC (rwt-APC) (E). The ratio for a catalytically inactive mutant, S360A-APC (active site residue Ser360 mutated to Ala), is 0.0 (E). Data (B,D,E) were from Mosnier et al.²⁹  Deep View Swiss PDB viewer (http://www.expasy.org/spdbv/) was used for the structural alignment. (A) The model of full-length APC¹¹¹  is based on the serine protease domain structure of APC (Protein Data Bank entry 1AUT¹¹² ). (C) The Gla-domain of the model of full-length APC was aligned with the protein C Gla-domain peptide crystallized in complex with soluble EPCR (1LQV).^111,113  The model of full-length EPCR (kindly provided by Dr B. Dahlbäck, http://www.klkemi.mas.lu.se/dahlback/index.php) was aligned with the crystal structure of soluble EPCR to obtain a model representation of APC bound to EPCR on the endothelial cell membrane. (B, D) Error bars indicate SEM.