We are studying the molecular organization of desmosomes by electron tomography. These investigations began with mouse skin, which was preserved by high-pressure freezing and freeze-substitution. 3D structures of individual desmosomes were determined after collecting image tilt series from thin sections through the epidermis. X-ray coordinates for cadherin were then fitted to molecular densities in the intercellular region of the map to elucidate the intermolecular interactions that give rise to this form of cellular adhesion (He et al. with supplementary material). More recently, we have been working with cultured keratinocytes to study the architecture of the intracellular plaque. This plaque has several constituents, which link to the intermediate filament network and thus stabilize the intercellular junction. We have studied the role of plakoglobin by culturing wild-type and knockout keratinocytes and imaging their respective desmosomes by tomography. In wild-type desmosomes, the site of intercellular adhesion is anchored to the intermediate filament network by a dense plaque of molecules including plakoglobin, plakophilin and desmoplakin. Genetic knockout of plakoglobin disrupts this plaque, fails to anchor intermediate filaments and thus produces desmosomes incapable of withstanding mechanical stress. We are currently characterizing the structural consequences of desmoplakin knockout as well as the architecture of the wild-type desmosome in the frozen-hydrated state. The latter involves cryoultramicrotomy of skin and imaging of whole mount keratinocytes grown directly on EM grids.
Click image to see higher resolution version

We are studying mechanisms of ion transport by P-type ion pumps using electron crystallography. We started by characterizing the structure and conformational changes that drive Ca transport by Ca-ATPase. We fitted an x-ray crystal structure (Toyoshima et al, 2000. Nature 405:647) to our lower resolution map from helical reconstruction, thus revealing large movements of cytoplasmic domains as a result of Ca binding to the transmembrane domain (Xu et al, 2002 , Stokes & Green, 2003 ). We have also determined a structure of the related Na/K-ATPase from duck salt glands and built an atomic model based on its homologies with Ca-ATPase (Rice et al, 2001). Currently, we are studying the physical interaction of Ca-ATPase and its regulator phospholamban, which is an important modulator of Ca concentrations in cardiac muscle. In particular, we are collaborating with a group involved in chemical crosslinking (Chen et al., 2006) and also studying two-dimensional co-crystals composed of CaATPase and phospholamban (Stokes et al., 2006). Recently we have solved the structure from a bacterial copper pump from Archaeoglobus fulgidis. This class of "PIB" pumps have specialized N-terminal domains that resemble soluble metallochaperones. By comparing structures of two CopA constructs, we have revealed the location of this N-terminal domain, allowing us to propose a model for how Cu-mediated regulation of transport.

In a new development, we have joined the NIH Protein Structure Initiative in an effort to add to the database of membrane protein structures. To date, such efforts have revolved primarily around X-ray crystallography, which has indeed proven extremely successful for soluble proteins. Membrane proteins, however, are far more refractory to forming the 3D crystals required for X-ray crystallography and we have therefore initiated a project to facilitate the use of electron microscopy for structure determination (Vink et al., 2007). Generally speaking, 2D crystallization within the membrane environment requires fewer constraints and provides a more native environment for this class of proteins. Although there have been several atomic structures determined by cryoelectron microscopy of 2D membrane protein crystals, this technique has not hit the mainstream. A major stumbling block is the lack of high-throughput methods for evaluating large numbers of 2D crystallization trials. Indeed, X-ray crystallographers routinely screen thousands of conditions in order to obtain suitable crystals and our project starts with the development of comparable methods for electron crystallography. We have teamed up with the New York Consortium for Membrane Protein Structure to obtain candidate proteins and are implementing methods for parallel microdialysis trials on a 96-well format followed by robotic imaging in the electron microscope. Ultimately, we hope not only to determine structures for a variety of membrane proteins, but also to establish fundamental parameters that govern the process of 2D crystallization.