Prediction and Validation Tools for Novel Membrane Interaction Surfaces from Protein Structures

点击翻译按钮获取中文摘要

It is generally thought that about one quarter of genomes encodes transmembrane proteins, many of which are important receptors and drug targets. Most of the remainder are typically assumed to be soluble proteins including signalling and metabolic enzymes. However an unknown number of these actually bind reversibly to membrane surfaces, and these interactions determine where these proteins are located inside cells, and regulate their enzymatic and signaling activities. In fact, it can be argued that a protein's location is just as important as its intrinsic activity, restricting its access to locally concentrated substrates, cofactors and ligands within the viscous and compartmentalized cell. Without an effective way to identify and test these membrane interaction surfaces, our knowledge of protein function will continue to be limited and our research progress in molecular biology and biochemistry will be restricted. Hence we are developing new computational and biophysical methods to accurately detect and validate protein-membrane interactions which localize proteins inside cells, providing new insights and tools for understanding cellular processes and disease mechanisms. A variety of protein modules including BAM, FYVE, PH and PX domains are known to bind membrane surfaces in response to changes cell stimulation, growth and differentiation. We propose that by analyzing the structural properties of such proteins, the principles of membrane binding can be elucidated and generalized, and entirely new classes of PMPs can be found. The interactions are diverse. Some proteins bind membranes specifically yet dynamically by reversible recognition of individual phospholipid headgroups, yet others bind tightly, being anchored to the bilayer where they help assemble molecular complexes and catalyze reactions. Yet our studies have revealed common themes including exposed hydrophobic loops, basic patches and polarized surfaces. These properties are integrated here by an algorithm that automatically identifies membrane binding sites from structures in seconds. Once suitably trained, this method will allow users to accurately predict the new types of PMPs, and experimental methods and lipid/micelle libraries will be available to allow researchers to efficiently validate such discoveries. The lack of fast and accurate tools to detect protein surfaces that interact with membranes has impeded progress in the fields of molecular and cellular biology, and has limited interactions between the fields of proteomics and lipidomics. The dearth of understanding about membrane protein interactions is compounded by technical difficulties of studying 'sticky' membrane interacting domains and delicate bilayer structures. Thus there is a real need for convenient and insightful computational and new experimental tools to analyze protein membrane recognition. Our solution aims to provide sufficient information to allow users to design and test how proteins are targeted to specific membrane domains, to predict their spatial orientations on membrane surfaces, and to reveal whether conformational changes could accompany binding events. Broad applicability is ensured by the fact that protein membrane interactions determine the organization and regulated activities of so many cellular organelles and molecular complexes. Some lipid binding domains influence cell proliferation, differentiation, survival, migration, adhesion and invasion. Others are involved in neurogenesis, angiogenesis, wound healing, immunity and developmental diseases. Our research will enable a deeper understanding of their interactions in sufficient detail to aid in the design of ligands and inhibitors, and may aid in the design of therapeutic agents where lipids normally bind. The tools will be developed and applied using selected human proteins to achieve high impact and disease relevancy, and will be standardized where possible to maximize applicability to any protei