April 15, 2006 (Vol. 26, No. 8)

Applying TEM-based Protein Tomography

Knowledge of the molecular processes that constitute biological systems has grown explosively in recent years. Translating that knowledge into useful therapies, however, has fallen short of early expectations. Translational researchers confront a stubborn, multidimensional gap in which technological deficiencies along at least three different axes coincide.

The simplest of these is the gap in imaging and analytical spatial resolution falling between the atomic resolution of NMR and XRD and the cellular resolution of light microscopy, a critical spatial regime that includes large molecules and multi-molecular complexes. The second gap occurs between the natural, living context of in vivo analysis and the artificial context of isolated, purified, crystallized molecular preparations. The final gap exists between model systems, used to discover and develop new drugs and the biological systems the models are intended to imitate.

Recent experience provides evidence of the difficulties posed by these gaps. Attrition rates in drug discovery programs remain above 90&#37, despite massive investments in R&D. Most estimates place the cost of bringing a new drug to market at well over $1 billion. The emphasis on brute force approaches, such as high-throughput screening and extensive automation, has done little to improve the picture and may actually be an important contributor to the negative correlation between R&D investments and productivity.

TEM-based Protein Tomography

Protein Tomography, which is based on advanced transmission electron microscopy (TEM), has the potential to bridge these gaps in investigative capabilities and support a more rational and, presumably, more productive approach. A TEM illuminates a thin sample with a broad beam of high energy electrons and focuses the transmitted electrons into an image. The optical path is analogous to that of a light microscope using transmitted light. In nonbiological materials, advanced TEMs have resolutions better than 0.1 nm. However, the lower beam exposure required for biological specimens reduces image contrast and limits resolution.

Protein Tomography mathematically combines a sequence of TEM images, each acquired from an incrementally different perspective, into a 3-D reconstruction of the specimen. The computation methods of tomography are well-known and used routinely in other medical imaging techniques.

The low signal-to-noise ratio of the biological images, however, requires an additional step to optimize performance. Sidec&#180s (www.sidec.com) Protein Tomography applies an algorithm that starts with an initial 3-D reconstruction, then iteratively maximizes the entropy of the model under the constraint of a chi squared fitting parameter, essentially producing the most featureless reconstruction that still fits the observed projection data.

Fig.1: Human growth hormone receptor dimer shown in situ(left) and in vitro (middle) using Protein Tomography, and an atomic level resolution model (right) obtained by XRD analysis of the crystallized receptor bound to human growth hormone (right)

Bridging the Gap

Protein Tomography bridges the gap in each of the senses described above. It&#180s resolution of 2&#821110 nm falls between the sub-Angstrom resolution of XRD and NMR and the sub-micrometer scale of light microscopy (Figures 1 & 2). It is sufficient to resolve the tertiary and quaternary structure of proteins and protein complexes and is not limited to smaller molecules. Unlike NMR and XRD, Protein Tomography examines molecules one at a time, permitting observations of structural differences between individual molecules, such as detecting a shift in equilibria between biologically active and inactive conformation.

Although TEM cannot look at living systems, it can look at biological systems in a near natural state. In particular, vitrification freezes materials so rapidly that ice crystals do not form, preserving delicate biological structure in an almost pristine state. Molecules are seen in their natural context with associated proteins and other molecules and in juxtaposition to functionally related cellular structures.

Finally, the natural context helps to bridge the gap between model and natural biological systems by allowing researchers to see more clearly the correspondence or lack thereof between them.

Fig.2: Structures of the Fc and Fab fragments as determined by XRD of the crystallized fragments correlate well with larger-scale structure as determeined by Protein Tomography (wire frame).

Membrane Proteins

Membrane proteins are difficult to express in large quantities and difficult to crystallize for X-ray analysis or purify for NMR. In nature, some portions of the molecule exist in aqueous intracellular and extracellular environments, while the transmembrane structures are embedded within the hydrophobic interior of the cell membrane. Any structural information obtained outside the natural membrane context may not reflect their true native conformation. Protein Tomography permits analysis of membrane proteins in their native state in situ.

Flexible Proteins

Many proteins are flexible, and conformational changes play an essential role in many protein interactions. Flexible proteins are difficult to crystallize for X-ray analysis, and if they are crystallized, their flexibility cannot be observed. Moreover, both NMR and XRD provide average structures, calculated from a large number of molecules. Protein Tomography can distinguish structural variations among individual molecules within a population of compositionally identical proteins and analyze their dynamic properties.

Multimolecular protein complexes comprise many of the most important biological structures. Most of these complexes form only in a natural context and are destroyed by typical isolation and purification procedures. Protein Tomography, however, can image ion channels that are transmembrane proteins composed of multiple subunits. Such image-based analysis provides valuable insights into their mechanisms of interaction and dynamic properties and can help validate target selection early in the drug discovery process.


Protein Tomography bridges a critical gap in our ability to visualize proteins, understand functional relationships, molecular associations, and intermolecular interactions in a natural context, as well as validate the model systems used to discover new drugs. It offers significant opportunity to enhance the rational aspects of the drug discovery process, eliminate unsuccessful candidates earlier to reduce unproductive investments, and ultimately yield more successful drugs at a reduced cost.

Previous articleASSAY and Drug Development Technologies Editor-in-Chief James Inglese, Ph.D., Director – Biomolecular Screening and Profiling Division, NIH Chemical Genomics Center
Next articleJohn Sterling talks with Bruce H. Littman, M.D., Vice President and Head, Global Translational Medicine, Pfizer Global Research and Development about translational medicine.