Technology

CSMP involves target selection, membrane protein expression, purification, characterization, crystallization, structure determination, and structure analysis. Target selection aims to identify the different representative classes of membrane proteins that cross the membrane at least 3 times (to eliminate signal directed secreted, single crossing, and membrane anchored proteins) for which a representative will be sought. Guided by Andrej Sali, this will be a dynamic process as structures emerge from efforts to determine membrane protein structures throughout the world.

Expression

The Center for Structure of Membrane Proteins (CSMP) at the University of California, San Francisco is dedicated to determination of the three dimensional atomic structures of integral membrane proteins of high biomedical impact, and their heteromeric complexes. Core capabilities support high efficiency application of target selection, cloning in cassette-based systems for tag-assisted purification, and fluorescent-protein fusions. Expression of eukaryotic and human MPs is focused on eukaryotic cells including human, insect and yeast cell culture. Purification using high affinity cleavable tags is guided by high quality biophysical characterization. Crystallization includes detergent/lipid based, and lipidic phase methods. Robotic crystal trials use detergent/lipid-based screens, and lipid-based methods. X-ray diffraction at the Advanced Light Source beamline 8.3.1 includes in-situ screening of crystallization trays at one of the world’s most advanced sources. The cores provide for collaborations, corporate collaborations and structural biology partnerships.

Structures

Following expression, solubilization in detergents, purification, crystallization trials, and protein crystallography will be pursued at beamlines 8.3.1 and 5.0.2 at the Advanced Light Source (ALS) in Berkeley. James Holton has developed state-of-the-art innovative procedures at the ALS beamline 8.3.1. An expert in automation, he developed new robotic systems for rapid evaluation of membrane crystals. Some membrane proteins may not crystallize in three dimensions, but may crystallize in two dimensions within a lipid bilayer. Electron diffraction will be championed by Henning Stahlberg for these cases. Electron crystallography will also be used to validate that the structures of 3-D crystalline samples are congruent with the bilayer form. Some membrane proteins may evade both structure determination schemes. NMR will be applied in these cases under direction of Roland Riek for small size targets.

Automated comparative modeling aids in target selection and target annotation of structural genomics targets, by evaluating the structural coverage of potential targets at target selection, and by evaluating the modeling leverage of a completed structure through template based modeling.

Modeling is performed by the automated modeling pipeline ModPipe (http://salilab.org/modpipe), that relies primarily on Modeller for fold assignment, sequence–structure alignment, model building and model assessment (http://salilab.org/modeller), and the results are deposited in the ModBase database of comparative protein structure models (http://salilab.org/modbase), which contains over 20 million reliable models from more than 3.5 million unique protein sequences. Approximately 800,000 models from 250,000 sequences were calculated for CSMP modeling tasks. ModBase, and the associated web-server ModWeb (http://salilab.org/modweb) are developed by Andrej Sali's group, and freely accessible to the academic community.

Modeling Leverage

One of the metrics guiding target selection in structural genomics is modeling leverage of a structure, which is defined as the number of protein sequences that can be modeled based on the structure. To determine the modeling leverage of each CSMP structure, template-based modeling as implemented in ModWeb (http://salilab.org/modweb) is performed, by using the sequence of the structure to find all related sequences in the UniProtKB database, and then modeling each of these sequences using the structure as a template. Finally, the resulting models are evaluated, and the modeling statistics of the calculation is summarized at http://salilab.org/csmp/csmp_models.cgi.

Functional Annotation

We have developed an integrated computational and experimental approach for functional annotation of membrane transporter targets with unknown structure. For high-priority transporter targets, initial models from ModBase are refined, by improving the template-target alignment, modeling the loops, repacking the side-chains, as well as by energy minimization using Molecular Dynamics simulations. Models are then evaluated based on their ability to discriminate known ligands from decoys using docking. We then perform virtual screening of small-molecule libraries, including approved drugs, metabolites, and fragment compounds, against the refined models. Finally, predicted small-molecule ligands are validated experimentally using kinetic measurements of uptake and site-directed mutagenesis. We applied this approach to characterize biomedically important transporters and their interactions with small molecule ligands.