Membrane protein structure determination using NMR spectroscopy and computational techniques

Abstract

Membrane protein structures are very difficult to determine by solution NMR since severe line-broadening obstructs the measurement of restraints. To alleviate this problem we describe the measurement of paramagnetic restraints on membrane proteins, particularly Paramagnetic Relaxation Enhancements (PREs), Residual Dipolar Couplings (RDCs), and Pseudo-Contact-Shifts (PCSs). A paramagnetic center is introduced into the 12 kDa protein KCNE3. A single Cysteine residue binds an EDTA-derived chelating agent that coordinates a paramagnetic lanthanide ion. Computationally, a knowledge-based hydrophobicity scale is derived for both α-helical and β-barrel membrane proteins, that is used to train an Artificial Neural Network to predict the membrane environment from a protein sequence. The approach is extended to develop a method that is able to simultaneously predict the secondary structure as well trans-membrane spans. The novelty of the approach is the application to both α-helical proteins as well as β-barrels. The prediction accuracies are comparable or higher to other available state-of-the-art prediction tools.