The Future of Imaging

WORKSHOP JULY 13–14, 2004 The Future of Imaging BAVARIAN ACADEMY OF SCIENCES MUNICH, GERMANY On July 13 and 14, 2004, 31 imaging experts from nine countries gathered at the Bavarian Academy of Sciences in Munich, Germany, to discuss the current state of and future trends in their field. The meeting, hosted by the Max Planck Society and the Howard Hughes Medical Institute and organized by David Agard, Wolfgang Baumeister, Stefan W. Hell, and Eva Nogales, focused primarily on challenges and opportunities with electron and optical microscopy. ELECTRON MICROSCOPY Detectors The electron microscopy portion of the meeting opened with a discussion of nonfilm detectors. Participants agreed that while digital detectors (charge-coupled devices, or CCDs) will likely replace film at some point, film currently has better resolution and sensitivity. At this time, CCDs have a resolution of 4000 x 4000 pixels and a pixel size of 15 microns. The fundamental challenge with fiber-optic coupled detectors has been the poor point spread function, which, at 300 kV, decreases resolution to about 50µm. At moderate resolutions up to 7 or 8 Å, or for diffraction studies or tomography, digital detectors are preferable because of their superior dynamic range, twofold higher contrast, and rapid readout. Use of an electron decelerator or a lens-coupled detector promises to offer better than a twofold improvement in resolution, yet it will come at a trade-off of complexity and cost. An alternative is to use detectors that directly detect the electrons, which would offer a superb signal-tonoise ratio and superior resolution. At this time, complementary metal oxide semiconductors (CMOS) are the most promising nonfilm detectors. The main drawback of this technology is that the imaging radiation damages the detector relatively rapidly; however, it should be possible to develop instruments with greater radiation damage tolerance. In response to this need, meeting participants proposed that the Medical Research Council Laboratory of 1 of 11 WORKSHOP Molecular Biology, the Max Planck Society, and perhaps HHMI form a consortium to fund the development of suitable CMOS prototypes that would allow rigorous determination of their utility at 300 kV. Electron Microscopy Contrast Enhancement An important issue in electron microscopy is how to enhance contrast in unstained materials. As in optical microscopy, this can be accomplished by phase contrast or differential interference contrast. Two technological approaches to phase contrast were discussed—a halfwave plate and an electrostatic phase-shift lens system. Dramatic enhancements in contrast have already been demonstrated using a phase plate; however, the lens approach promises to provide better control and eliminate scattering, improving the signal-to-noise ratio. While both these technologies need further development, the commercial sector appears to be making such efforts, and significant improvements are expected in the next few years. Cryoelectron Microscopy of Cellular Structures A key contribution of cellular electron microscopy imaging will be to bridge the gap between molecular structure and cell biology. Toward this end, cryoelectron tomography, in which cells are grown on electron microscopy grids on carbon films, frozen, and “sectioned” by electron tomography, will be an important tool. This technique provides outstanding preservation and high resolution. Additionally, it can be done with samples that have been minimally manipulated by fixation and dehydration, eliminating many of the artifacts associated with plastic-embedded transmission electron microscopy. One drawback of cryoelectron tomography is a low signal-to-noise ratio, but phase contrast techniques may improve this. A second problem is that because the tomography works best when the sample is thin, with an upper limit of about 500 nm, such as at the periphery of cells, it is impossible to see deeper structures at high resolution. The only general solution for imaging structures in cryopreserved cells is frozen sectioning to reduce the imaged thickness to 50 to 500 nm. There are a number of delicate steps in the protocol, each of which requires care. One set of problems comes from the action of the knife, which creates cracks, chatter, and distortion in the sections. Participants recognized that this technique has great potential but requires a high degree of expertise, and they proposed that it might be valuable to establish a technology center to provide expertise and training 2 of 11 WORKSHOP Electron Microscopy Approaches in Structural Biology It is difficult to determine the structures of membrane proteins and large protein complexes by standard crystallographic techniques (these molecules are very hard to crystallize). An alternative is to image large numbers of identical particles and then process the images computationally to derive structural information at much higher resolution than each single image. A generic problem with this approach is that the particles are often randomly oriented and each image must be classified individually to overcome this. Another problem is beam-induced specimen movement. Some have suggested that this comes from sample charging, but it was also argued that mechanical deformation following radiolytic decay in the specimen causes most sample buckling and cracking. Overcoming this limitation could enable sufficient resolution (approximately 4 Å) to derive the correct fit of the aminoacid chain for a 100-kilodalton protein after imaging 50,000 particles. Participants agreed that a threefold reduction in sample motion should be feasible and would significantly improve imaging. This might be accomplished by brute-force scale-up to imaging 106 or more particles, with selection of the best 10 percent of images, or mechanically reducing motion, such as by improving support media or reducing the size of electron microscopy grid holes. One suggestion was to strengthen the ice, perhaps with embedded nanotubes. Also, imaging with short burst of the electron beam, rather than a continuous beam, might help. Another factor limiting the resolution obtainable with cryoelectron microscopy is sample heterogeneity. Heterogeneity may result from poor sample preparation or substoichiometric binding of ligands or partner proteins in a complex, but the predominant source of heterogeneity appears to be low-energy perturbations in local conformation. Users of cryoelectron microscopy face challenges to high-throughput imaging and automation. A high-resolution structure with 4-angstrom resolution requires as many as 106 particle images, so rapid, efficient imaging is critical. Whereas crystallographers have greatly accelerated their ability to produce finished structures by automating much of their sample preparation, this type of optimization is only just beginning in the cryoelectron microscopy field. Automated data collection recently passed a key milestone—the ability to collect 100,000 single-particle images of the GroEL protein complex in a 24-hour period. Using previous approaches, this task would have required many months. Another key limitation at this point is software: imagers are using four different, incompatible packages. Participants saw a clear need for a userfriendly, expert system to track samples, classify images, analyze the data, and render structures, and considered this a major, yet highly 3 of 11 WORKSHOP feasible project. Another exciting approach to protein structure determination was discussed—diffraction of protein molecules in a beam rather than a crystal. Crystals are generally required to obtain a diffraction pattern because the low radiation doses required to preserve the specimen produce only very weak signals per protein molecule. However, it is theoretically possible to supply a very high X-ray dose in a very short pulse (10 femptoseconds) and collect the data from an individual molecule before it is destroyed. Beams capable of producing X-ray doses adequate to generate high-resolution structures using this approach will soon be available at the Stanford linear accelerator and the Max Planck Institute for Quantum Optics. The group agreed that while single molecule imaging technology has yet to be tested, if it can be made practical it would have enormous impact. Another possibly revolutionary approach to obtaining structural data from single molecules was discussed. This is an electron diffraction approach based on producing a molecular beam in which elliptically polarized light orients a significant number of protein molecules flying in the same drop of solvent. This provides a good signal-to-noise ratio and greatly facilitates collection of data from all possible particle orientations. One challenge is that unlike in cryomicroscopy, the diffraction data would have to be computationally phased. Since the beams are continuous, data can accumulate until a good diffraction pattern is built up. As yet, this technique is untested on real molecules, but the first molecular beam apparatus is nearing completion. OPTICAL MICROSCOPY There was general agreement that since electron microscopy specimens must be imaged in a vacuum, optical microscopy will remain the leading technology for live-cell imaging. Methods for bridging the gap between high-resolution electron microscopy and live-cell biology are a major need—for example, techniques that allow a particular region of a cell, identified by optical microscopy, to be imaged at high magnification by cryoelectron microscopy. Currently, optical microscopy using fluorescent probes provides the only method for following specific molecules and processes in vivo. Participants agreed that there is a need for continuing fluorescent probe development, in particular for probes that can be detected by both optical and electron microscopy. One probe type discussed by the group was quantum dots (QD)—semiconductor particles in which the size of the core determines the optical properties. They are typically 4 of 11 WORKSHOP 6 to 10 nm in diameter and are much brighter than other common probes. However, they are inherently not genetically encodable, making their precise targeting difficult. One approach for overcoming this difficulty is to attach peptides to their surface that can target either endogenous proteins or genetically encoded tags. Participants saw a need for faster (nonscanning) confocal microscopy for in vivo applications. A “programmable array microscope” (PAM) is under development that is 100-fold faster than scanning. The PAM will allow full-field optical sectioning and can yield spectroscopic information. The PAM has no moving parts, and researchers should be able to use it as an attachment to existing microscopes. Another approach is optical microspectroscopy of single gold nanoparticles. Gold particles do not suffer from “blinking” or photobleaching and can be detected optically. This requires photothermal microscopy: a spot close to the image focus is heated by 20°C using a laser, allowing 5 nm particles to be imaged with 10 ms per pixel integration. Smaller gold particles can be resolved with less heating. Optical Microscopy Resolution The wavelength of light has been thought to place a firm limit on the resolution of optical microscopy. However, recent advances suggest that this theoretical Abbe’s limit can be surpassed, and that resolution down to 10 nm might be possible. Twofold improvements in resolution can be obtained in either confocal (4Pi) or widefield (I5M) technologies. In either case, improved axial resolution uses two lenses and interference to improve resolution along the z-axis, and widefield approaches use structured illumination to improve the lateral resolution. Resolution beyond this twofold enhancement (approximately 100 nm) requires exploitation of nonlinear properties in the fluorescent dyes. Such super resolution has now been demonstrated using either dye saturation coupled with structured illumination or stimulated emission depletion (STED). STED employs light-induced quenching of fluorescence, at a wavelength longer than the excitation energy, to “squeeze” the illuminated spot in a scanning microscope. There is no theoretical lower limit to this, but in practice the high-level quenching light can bleach the fluorophore and will require the development of very bright and bleaching-resistant probes. The future of these methods will require the development of dyes optimized for their nonlinear properties. 5 of 11 WORKSHOP Optical Coherence Tomography Optical coherence tomography (OCT) allows for deep in vivo imaging. The technique is similar in principal to ultrasound in that it is based on measuring echo time delay. OCT uses light rather than sound to measure the delays, using interferometry to measure reflected optical waves. A third-generation OCT imaging device is available commercially for ophthalmologic applications. This approach allows clinicians to produce retinal cross-sections. Improvements in resolution down to 1 to 5 µm and up to 100-fold increased speed are anticipated, through the use of femptosecond-pulsed lasers. Active and Adaptive Optics One challenge to optical microscopy is variations in thickness and optical index within a sample. In principle, this can be overcome by first probing the specimen to detect inherent aberrations, and then rescanning the image with the optics dynamically adjusted to correct for the observed aberrations. Astronomers routinely use such methods to correct for effects of the earth’s atmosphere, but sophisticated methods are required to accurately generate arbitrary correction fields, such as pupil phase modification, deformable mirrors, spatial light modulation, and digital mirror devices. This is clearly a promising direction, especially for in vivo microscopy where sample-induced aberrations significantly degrade the resultant images. CONCLUSION Imaging by light and electron microscopy will play increasingly important roles in biology. The participants were optimistic that the near future will see significant improvements in both electron and optical microscopy in terms of resolution, efficiency, and data analysis. These advances will likely require financial investment in instrument development beyond what can be expected from commercial manufacturers and will result in constantly evolving and sophisticated instruments that might be more efficiently maintained in centers rather than in individual laboratories (or even universities). Simultaneously, organized efforts to develop and evolve robust software packages for data collection, reconstruction, and image analysis are clearly needed. The scope of these efforts is also likely to be beyond the reach of individual laboratories. 6 of 11 WORKSHOP WORKSHOP PARTICIPANTS David Agard Howard Hughes Medical Institute University of California Donna Arndt-Jovin Max Planck Institute for Biophysical Chemistry Wolfgang Baumeister Max Planck Institute of Biochemistry Bridget Carragher The Scripps Research Institute David Clayton Howard Hughes Medical Institute Winfried Denk Max Planck Institute for Medical Research Ken Downing University of California Jacques Dubochet Laboratoire d´Analyse Ultrastructurale Université de Lausanne Joachim Frank Howard Hughes Medical Institute Wadsworth Center James G. Fujimoto Massachusetts Institute of Technology Robert Glaeser University of California Nikolaus Grigorieff Howard Hughes Medical Institute Brandeis University Mats Gustafsson University of California 7 of 11 WORKSHOP Maximilian Haider Corrected Electron Optical Systems GmbH (CEOS) Stefan Hell Max Planck Institute for Biophysical Chemistry Richard Henderson MRC Laboratory of Molecular Biology Grant Jensen California Institute of Technology Zvi Kam Weizmann Institute of Science Ferenc Krausz Max Planck Institute for Quantum Optics Werner Kühlbrandt Max Planck Institute of Biophysics Vadim Mesyanzhinov Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry Kevin Moses Emory University School of Medicine Kuniaki Nagayama Center for Integrative Bioscience Okazaki National Research Institutes Eva Nogales Howard Hughes Medical Institute University of California M.A.G.J. Orrit Wiskunde & Natuurwetenschappen Gerry Rubin Howard Hughes Medical Institute John C.H. Spence Arizona State University 8 of 11 WORKSHOP Holger Stark Max Planck Institute for Biophysical Chemistry Shimon Weiss University of California, Los Angeles Tony Wilson University of Oxford Willy Wriggers University of Texas WORKSHOP AGENDA The Future of Imaging July 13–14, 2004 Bavarian Academy of Sciences Munich, Germany Tuesday July 13, 2004 Introductory Remarks David Agard, Wolfgang Baumeister, Stefan Hell, and Eva Nogales Electron Microscopy: Detectors Richard Henderson EM: Contrast Enhancement by Electron Optical Means Kuniaki Nagayama Optical Microscopy: Marker and Probes—Bleaching, Q-dots, Single Molecules Shimon Weiss OM: Imaging Molecular Interactions—FLIM, FRET Donna Arndt-Jovin Working Group Discussions EM: Detectors Moderator: Ken Downing EM: Contrast Enhancement by Electron Optical Means Moderator: Maximilian Haider 9 of 11 WORKSHOP OM: Marker and Probes, FLIM, FRET Moderator: M.A.G.J. Orrit EM: Factors Limiting Resolution Robert Glaeser EM: Conformational Heterogeneity Joachim Frank OM: Overcoming the Resolution Limits Stefan Hell OM: Optical Coherence Tomography/Microscopy James Fujimoto Working Group Discussions EM: Factors Limiting Resolution Moderator: Nikolaus Grigorieff EM: Conformational Heterogeneity Moderator: Holger Stark OM: Overcoming the Resolution Limits Moderator: Mats Gustafsson OM: Optical Coherence Tomography/Microscopy; Thick Objects Imaging Moderator: Winfried Denk Wednesday July 14, 2004 Introduction to the Resolution Limits of Light Microscopy Mats Gustafsson OM: Ultrashort Pulses, Novel Light Sources Ferenc Krausz OM: Wavefront and Aberration Control/Adaptive Optics Tony Wilson EM: High Throughput Tactics Bridget Carragher 10 of 11 WORKSHOP Working Group Discussions EM: High Throughput Tactics Moderator: Eva Nogales EM: Correlative LM and EM Moderator: Winfried Denk OM: Ultrashort Pulses, Novel Light Sources Moderator: James Fujimoto OM: Wavefront and Aberration Control/Adaptive Optics Moderator: Zvi Kam EM: Cryoelectron Tomography/Cryosectioning Wolfgang Baumeister and Jacques Dubochet EM: Electron Diffraction John Spence Reports from working groups and discussion 11 of 11