News and Publications

Automated Molecular Imaging

B. Carragher, C.S. Potter, D. Fellmann, F. Mouche, J. Pulokas, B. Sheehan, C. Suloway, Y. Zhu

Elucidating the structure and mechanism of action of molecular machines is an emerging frontier in understanding how the information in the genome is transformed into cellular activities. Molecular machines are associations of individual components (e.g., proteins, nucleic acids, lipids) in the form of large complexes; examples include ribosomes, transcription complexes, track-motor complexes, and membrane-embedded pumps and channels. These machines not only are large but also may be conformationally and compositionally dynamic or present in comparatively low numbers, factors that make them extremely challenging (or impossible) objects for study by x-ray crystallography and nuclear magnetic resonance methods. They are, however, ideal objects for using electron microscopy to determine their structures, and this approach has provided insights into the working of many macromolecular complexes at both high and moderate resolutions.

Although molecular microscopy has enormous potential, it is time-consuming and labor intensive. The techniques needed to produce a 3-dimensional electron density map of a macromolecular structure normally require manual operation of the electron microscope by a skilled operator and manual supervision of the sometimes complex software used for analysis and calculation of 3-dimensional maps. Typically, it takes weeks to months to collect and analyze a data set to reconstruct a map at 10- to 20-Å resolution. Scientists generally agree that increasing the resolution to a range where secondary structure will be discernable in the map (~7 Å) will require a 10-fold increase in the amount of data collected and analyzed. Clearly, molecular microscopy will not be practical as a mainstream technique unless the imaging and analysis can be automated and the overall throughput greatly improved.

We are trying to develop a completely automated system for molecular microscopy. Our goal is to automate all processes, from inserting the electron microscopy grid to calculating the final 3-dimensional map. During the past several years, we developed a system called Leginon to automatically acquire electron cyromicrographs. The system is designed to emulate all the decisions and actions of a highly trained microscopist in collecting data on a vitreous ice specimen. The tasks include identifying suitable areas of vitreous ice at low magnification, determining the presence and location of the specimen on the grid, automatically adjusting imaging parameters (e.g., focus, astigmatism), and acquiring images at high magnification. The system was tested with a variety of specimens that represent typical challenges in electron cyromicroscopy. The results indicated that the overall performance of the system is equivalent to that of an experienced microscopist.

During the past year, we integrated Leginon with an automated system for processing and reconstruction. We used tobacco mosaic virus as a test specimen to validate the overall system and show the feasibility of full automation. The 3-dimensional map of the virus at a resolution of 10 Å (Fig. 1) was generated without operator intervention and within 24 hours of inserting the specimen in the microscope.

Technologies that form the basis for this system include instrument control and development; robotics; and automated image processing, analysis, and reconstruction. Members of our group include engineers and computer scientists. Collaborations with other groups at TSRI provide excellent examples of biological research that helps drive the technologic developments.

Our goal for the next several years is to transform electron microscopy into a high-throughput method for structure determination. Success in this endeavor will do more than greatly facilitate molecular microscopy. It will also expand the scope of accessible problems and push experimental frontiers by making possible investigations that are presently deemed too high risk because of the enormous effort involved.

PUBLICATIONS

Fellmann, D., Pulokas, J., Milligan, R.A., Carragher, B., Potter, C.S. A relational database for cryoEM: experience at one year and 50,000 images, J. Struct. Biol. 137:273, 2002.

Rouiller, I., Pulokas, J., Butel, V.M., Milligan, R.A., Wilson-Kubalek, E.M., Potter, C.S., Carragher, B.O. Automated image acquisition for single-particle reconstruction using p97 as the biological sample, J. Struct. Biol. 133:102, 2001.

Zhu, Y., Carragher, B., Kriegman, D., Milligan R.A., Potter, C.S. Automated identification of filaments in cryoelectron microscopy images, J. Struct. Biol. 135:302, 2001.