Furin, a member of the proprotein convertase family, is a multi-domain proteinase, the catalytic domain of which is similar in structure to bacterial subtilisin. Furin functions in the Golgi apparatus, in the secretory vesicles and, potentially, also on cell surfaces. This unique specificity proteinase cleaves after basic residues many functionally important cellular proteins, including soluble and membrane-tethered metalloproteinases, integrins, signaling receptors, growth factors, hormones and neuropeptides, and transforms them into their respective mature forms 
. In addition to processing cellular precursor proteins, furin is also exploited by numerous bacterial and viral pathogens. Pathogenic viruses and bacterial toxins employ host furin to become fully functional and to allow entry into host cells and to cause disease onset.
In this study, we developed an approach for proteinase target identification and used it to identify approximately 490 probable targets of furin proteolysis in the human proteome (). This number significantly exceeds the number of previously identified and experimentally confirmed substrates and demonstrates the power of our highly multiplexed, proteome-wide approach. We confirmed that a majority of the known furin substrates reported in CutDB and Merops databases were identified by our pipeline and included in our furin substrate dataset.
Furthermore, since we have accomplished our experimental studies and our manuscript was under preparation, others obtained valuable experimental confirmations of our predictions. For example, it has been recently shown in the direct laboratory experiments that glypican 
, bone morphogenetic protein 9 and 4 
, growth differentiation factor 11 
, endothelin and adrenomedullin 
, Kazal type 5 
, and angiopoietin-like protein 4 
are furin substrates as we predicted by using our in silico pipeline. We are confident that additional experimental confirmations will follow in a near future. In principle, our approach to discover proteinase targets can be applied to other proteinases in eukaryotic, prokaryotic and viral proteomes, and may ultimately enable the identification of new therapeutic targets.
We also established a method for measuring the efficiency of furin proteolysis, identified the effect of the residues at P7-P4’ positions on the overall cleavage efficiency of the substrate, and determined the precise cleavage preferences of furin. We demonstrated that the multi-basic amino acid motif is not a sufficient requirement for the efficient furin proteolysis of the substrate protein. We showed that amino acids located at the P7, P6, P5, P3, and P1’-P4’ oppositions are very important for modulating furin cleavage efficiency. This preference information can be used in the design of inhibitors, which, especially for proteinase antagonists, frequently starts with a known substrate. For example, because several furin-like individual PCs and furin exhibit similar P4-P1 preferences, furin inhibitors with P4-P1 occupancy alone will cross-react with PCs, which is a known phenomenon 
. However, our results indicate that cross-reactivity with PCs might be decreased by designing inhibitors that occupy not only the S4-S1 but also the S6-S7 sub-sites in the furin active site. This specificity requirement, however, may increase the size of furin inhibitors and, as a result, negatively affect their drug-likeness parameters including cell permeability. Nevertheless, inhibitors designed in this way could be useful research tools.
Because of its high positive charge, the multi-basic cleavage motif is predominantly localized on the protein surface and is readily accessible. We initiated the analysis of the structural parameters determining accessibility of the cleavage site to a proteinase. The structural parameters may play a significant role in affecting the proteolysis efficiency 
. Accordingly, our pipeline could be improved by including a structural analysis of the cleavage site vicinity. This task could be accomplished by either including the prediction of the secondary structure elements involving the cleavage sites 
, by inspecting available experimental structures from the Protein Data Bank 
, or by analyzing three-dimensional protein models built using homology modeling approaches 
. The results obtained in this study can be used for both development of a better mathematical model for detecting proteinase cleavage sites and for improvement of the existing methods of cleavage site prediction 
Our combined in vitro
and in silico
analysis culminated with the assignment of probable furin substrates to key signaling and metabolic pathways, molecular networks and biological processes, suggesting directly that furin proteolysis plays a highly significant role in normal development and embryogenesis, and, specifically, in regulating axonal guidance and cardiogenesis, and maintenance of stem cell pluripotency. These findings are consistent with the phenotype observed in the furin knockout mice 
. We believe that our results can help guide experiments to elucidate the molecular mechanisms of furin-associated human pathologies, which may lead to the design of new pharmacological interventions.