2D crystallization and image processing of membrane proteins

The goal of structural biology is to determine the structure of biological molecules such as proteins,
lipids, DNA or bigger complexes consisting of these basic building blocks. The determination of
structure is an important step to gain insight into the physical and biological functioning of these
molecules and complexes. Structure, in the general sense of physical and chemical composition,
determines lastly the function.
Several techniques are known and used in the field of structure determination. X-rays can be used
to determine structures of crystallized biological molecules or complexes to a high precision, as high
as atomic resolution. The drawback of the structure determination using the X-ray technique is that
the molecules are not anymore in their biological environment and that they are forced in the crystal
packing to conformations they would not adapt in nature. NMR spectroscopy is an other method used
to determine the structure of biological molecules at atomic resolution. This method is preferentially
used for soluble compounds. The structure of larger complexes could recently be analyzed with both
X-ray crystallography (ribosomes [45]) and NMR spectroscopy (GroEL GroES complex [14]), but
the determination of such big complexes remains a very difficult issue.
Single particle analysis of electron microscopy images on the other side allows the structural determination
of bigger complexes. The drawback of this method is that a much lower level of detail
(resolution) can be achieved.
Every technique has its field of application, its advantages but also its drawbacks. It is often the
problem one wants to solve, that determines the the way to solve it. Bigger complexes of molecular
machines are investigated using the single particle method, small soluble molecules using NMR or
membrane bound proteins are investigated using electron crystallography.
Membrane proteins are important in cells, as they can select what passes through a membrane and
therefore what moves in or out of a cell or an organelle. Since membrane proteins have a hydrophobic
surface, they are stabilized to a great extent by the membrane they are in. The isolated protein is
rather unstable and easily loses its functionality when removed from the lipid bilayer. This is the
main reason why membrane proteins are so difficult to crystallize in 3D in order to be analyzed
with the X-ray crystallography and why they are too unstable in solution for the analysis with NMR
spectroscopy. These limitations explain the observation, that only very few structures of membrane
proteins are resolved to a high resolution until now.
The generation of 2D crystals and the subsequent analysis with the electron microscope using images
or direct electron diffraction is a another way to get structural data of membrane proteins up to a very
high resolution as it has been demonstrated by solving the structure of AQP-1 to a resolution of 3.8
Å [34] and then 3.2 Å [10] even before the structure was solved by X-ray crystallography [48]. More
recently the structure of AQP-0 was solved by electron crystallography to 1.9 Å, revealing not only
the protein, but also the lipids surrounding it [18]. Hence, the big advantage of this method is that the
proteins are in a lipid bilayer and therefore in a close to native environment. However, its drawback
is that only few steps in the whole process of getting structural information from the purified protein
are automated. A systematic screening for ideal crystallization conditions is mandatory but time consuming.
The image acquisition with the electron microscope demands highly skilled and experienced
users for the sample preparation and the microscope handling in order to get high resolution data.
Image analysis and data processing as last step in the structure determination process is of great importance
as the information present in the images or in the diffraction data needs to be extracted and
interpreted. The advance of data processing was slow for a long time and could hardly cope with the
enormous amount of data generated. Only recently more resources are available for developments in
this field ([17], [38] and [37]).
Advances in all these presented fields, from protein production up to image processing, are pushed by
our group. In this thesis I will present my contributions to the development of a novel method for 2D
crystallization, the production of 2D crystals and the development of tools for image processing.
The methods used, the biological background and the insights gained from the performed experiments
will be described and discussed. An overview and some insights gained through collaborative work
in an interdisciplinary team will also be given.