This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
This Major Research Instrumentation-Recovery and Reinvestment (MRI-R2) award funds the development of an 8kx8k pixel direct detection CMOS camera with single electron quantized detection for high-resolution single particle cryo electron microscopy (cryoEM) at the University of California-San Francisco (UCSF). The new camera system will enable high frame readout rates, with extreme sensitivity and high resolution. Because it uses so little material and does not require crystals, single particle cryoEM has become an indispensable tool for studying the three-dimensional (3D) structures of complex biological assemblies, providing critical information not obtainable by more traditional methods such as x-ray crystallography or NMR spectroscopy. The camera development project will be coordinated through collaborations among UCSF, the Lawrence Berkeley National Laboratory (LBNL), and Gatan, Inc., a company specializing in EM cameras and peripherals. The camera will benefit a large number of research projects and improve the training prospects for 15-25 students and postdoctoral fellows in advanced cryoEM technology. UCSF PIs are also active participants in summer research programs and have been hosting undergraduates in their labs to provide research opportunities with advanced technologies. Performance and availability of the camera will be widely disseminated through a web site, in publications, and through presentations at local and international meetings. Commercialization of the new technology may lead to even wider dissemination of similar devices.
The goal of the project was to develop a large format (8kx8k pixel) camera for cryo electron microscopy (cryoEM) that would directly detect the electrons in the silicon chip of the CMOS sensor instead of needing a scintillator to convert the electrons into photons before detection, as in traditional EM cameras. The project was designed to proceed in several phases starting with an initial pixel design test chip built with HHMI funds (Phase I), followed by a 4kx4k complete sensor with high speed counting electronics (Phase 2) and finally, to use Phase II as a basis for scaling up to an 8kx8k pixel sensor with associated counting electronics (Phase 3). As part of the design goals, there were two ways of using the sensor – i) conventional analog integration mode and ii) single electron detection mode. In the former, signal is accumulated on the chip during the exposure and the resultant total accumulated image is then read out. In the counting mode, the goal is to readout at such high rates that each primary electron event is spatially distinct from its neighbors, allowing the entire distribution of imager electrons arising from the primary event to be individually recognized, and its centroid and intensity determined. The result is a single digital count instead of a cloud of electrons and background noise. Because the response cloud can be centroided to sub pixel accuracy, the effective size of the camera can be increased by a factor of 2 in each direction (super-resolution) to ~7500x7500 apparent pixels. The Phase II development of an ~ 4kx4k pixel radiation hardened sensor and camera optimized for single electron counting has been a spectacular success. The camera has a winning combination of exceptional detection efficiency, near zero noise and very fast readout. While we had always intended that we could use the fast readout to correct for sample drift in the microscope, it turns out apparent motion induced by the electron beam itself, has been a major limiting factor in cryoEM. We now record super low dose movies throughout the exposure and, because the noise is so low, are able to re-register each movie frame to correct for motion. This has led to a game-changing advance in attainable resolutions (3.3Å resolution on a 700KDa protein complex). At these resolutions individual side chains can be visualized allowing the determination of macromolecular atomic structure directly from the experimental density maps. This is a very exciting breakthrough for the field and has led to the rapid adoption of the commercial version of this camera, known as the Gatan K2 Summit. Based on the analysis of the Phase II camera, it was decided that a simple size scale up as planned should be delayed in favor of other improvements. While the design part of that effort is underway, actual development of the new sensor has been delayed by legal problems related to switching to a different chip fabrication facility located outside the US. Since the timing wasn’t workable, we requested, and were granted, permission to use the remaining funds to develop a next generation imaging energy filter (GIF) that would be optimized for use with the new K2 camera, and greatly improve the quality of imaging of larger, unique cellular samples that are tilted to collect 3D information. Because cellular samples are generally quite thick, and get even thicker when tilted to 60-70 degrees as required by tomography, it is necessary to use an imaging energy filter to remove inelastically scattered electrons – selecting only the purely elastic electrons to form the image. This is required because the high degree of chromatic aberration in the EM leads to electrons of different energies being brought to different focal planes, and hence adding noise rather than coherent signal to every image recorded. Even though the time has been quite limited, this effort has also been quite successful. The new BioQuantum GIF, with a completely redesigned prism and new lenses has been prototyped and was shown to be working at the end of September at Gatan’s factory in Pleasanton. The improvement in all aspects of the specifications was quite impressive, but final testing can only happen here at UCSF. Once the revised control electronics are available (scheduled by the end of October), the GIF will be installed at UCSF with a K2 camera both before and after the GIF, allowing rigorous testing for improved resolution, lower distortions, etc. The predictions are it should have a significant impact on tomographic data collection and may also aid in single particle work.
