Workflow automation for image analysis of 2D crystals of membrane proteins

Membrane proteins carry out various functions essential to the survival of organ- isms. They transfer signals between the cell’s internal and external environments, move molecules and ions across the membrane, act as enzymes, and allow cell adhesion. This is why membrane proteins represent more than half of all drug targets. A deeper insight into the functional mechanisms of a protein can be gained from structural information. And so far only a fraction of membrane protein structures has been determined.
The topic of this thesis is structure determination of membrane proteins through electron crystallography focusing on the image processing of 2D crystals. The thesis combines both method development and structure studies. In the Methods part, state of the art processing of 2D crystal images is presented. The workflow em- bedding all the processing steps from the initial micrographs of 2D crystals to the resulting 3D electron density map of the reconstituted membrane protein is de- scribed. The possibility of autonomous high-throughput processing is discussed as the ultimate goal of automation of this workflow. An additional processing step of the workflow that captures the variation of tilt geometry in the 2D crystal is introduced. This is implemented as an iterative refinement of the local tilt geometry using a Single Particle processing approach.
A great benefit of electron crystallography is the fact that through reconstitution the purified protein is embedded in a natural environment, a membrane. Biochemical manipulations of this environment can lead to structural changes, which yields insight into the functional states of the protein. An new method of analyzing these structural changes in 2D projection maps is presented here. The method identifies significant changes in the protein by distinguishing them from noise derived artifacts.
The second part of this thesis covers applications of these methods in structural studies of unknown membrane proteins. In the study of the Secondary Citrate/Sodium Symporter CitS, the substrate binding domain was identified with help of the significant difference map method. The improvements of the image processing routines were directly applied in the analysis of the 2D crystals.
The structural studies of nucleotide-modulated potassium channel MloK1 also benefited from the automated image processing workflow and the significant difference map, while identifying structural changes through ligand binding. To gain a more detailed electron density map of MloK1, the local tilt geometry of the crystals were refined with the single particle 3D reconstruction for 2D crystal images method.