Characterization of Promiscuity and Specificity in Proteases

Proteins form the central machinery of life. They interact with other biomolecules to perform nearly all important tasks in living cells. Some of these interactions have to be precisely governed specifically towards individual interaction partners whereas others are much more universal. In this way proteins are the essentials tools of our body. Some of these tools are very specialized whereas other serve much more general purposes. This project focuses on proteases. Proteases are protein-tools in this machinery that are responsible for cutting other proteins and peptides. Evolution has reinvented proteases many times, surprisingly coming up with similar solutions over and over again. Nearly three percent of our genes code for the 560 different proteases in our body. Many drugs inhibit proteases, e.g., to control blood coagulation, to treat hypertension and diabetes, to fight cancer and to combat viral diseases like HIV or Hepatitis C. On the one hand, there are proteases, like the ones responsible for digestion, which are designed by evolution to be nearly universal cutting tools. On the other hand, there are proteases, which are so extremely specialized, that only a single substrate, which is cut by these proteases, could be found yet. Surprisingly in a static view, these very universal and very specialized proteases do have nearly identical structures. There are strong hints that major differences can only be identified by looking at dynamic properties of these proteases, i.e., their flexibility. Flexible proteases might be more promiscuous, as they could adapt to a larger number of interaction partners, whereas rigid proteases could be more specific. Ideal techniques to analyze flexibility are computer simulations of the proteases dynamics. The project aims at systematically characterizing flexibility and other important physics-based properties which are responsible for promiscuity and specificity of two sets of proteases, which have a very similar structure but very different binding specificities. Ultimately we envisage that this characterization can be transferred to other biomolecular interfaces to predict their promiscuity and specificity. This would help to understand many biochemically processes and might be used to optimize the specificity of biopharmaceuticals.

A varying degree of promiscuity and specificity of protein interfaces for binding partners is fundamental to many biological processes. Prototypic large protein families that are well known to comprise both rather promiscuous and rather specific members are, e.g., antibodies, kinases and proteases. In all these cases, the degree of specificity and promiscuity is decisive and fundamental for the proteins' biological functions. In the project we have focused on proteases and the study of their interface properties. Proteases cleave the peptide bond between amino acids and perform a wide variety of biochemical functions, comprising central roles in seemingly very different aspects of life ranging, e.g., from signaling cascades over key aspects of the immune system and programmed cell death to digestion. Although these very different functions imply vast differences in specificity and promiscuity, they are usually performed by remarkably similar proteases. Efficient and reliable techniques have been developed to characterize substrate preferences of proteases, resulting in the construction of MEROPS, a large proteases' substrate database. Proteases bind their peptide substrates in a rather linear binding cleft in a beta-strand like conformation. This linear binding cleft recognizes side chains of the peptide towards the N- and C-terminal end of the peptide with a varying degree of tolerance. We have developed methods to quantify and compare this tolerance and to localize the areas of promiscuity and specificity in the binding cleft. We have developed different computational techniques to unravel the physicochemical principles underlying differences in promiscuity within the binding cleft of individual proteases and differences in substrate specificities of seemingly similar proteases. We have focused on enthalpic and entropic features of the binding interfaces as well as on differences in solvation in order to characterize and rationalize differences within and inbetween proteases. Therefore, we have combined enhanced sampling techniques for computer simulations and Markov state models with grid-based methods and inhomogeneous solvation theory. The results have helped to provide detailed understanding and a comprehensive picture of characteristics of protease promiscuity and specificity.