Bruce McEwen Laboratory
Cryo-electron Tomography
McEwen et al. J. Struct. Biol. 2002 138, 47-57.
Recent technical improvements in data collection have reduced the total dose requirements for electron tomography and have enabled applications to use frozen-hydrated specimens. Frozen-hydrated specimens are thought to be as close to the native structural state as is possible for electron microscopy. However, application of tomography to frozen-hydrated specimens is in its infancy, and many issues need to be resolved concerning feasible resolution levels and limitations due to beam damage. The above paper is a first attempt to quantify some of the relevant parameters, using sea urchin sperm axonemes.
Axonemes as test specimens for electron tomography of frozen-hydrated specimens. Axonemes are the core structure common to most eukaryotic cilia and flagella. These well-studied organelles make ideal test specimens for cryo-electron tomography for several reasons. First, at 0.25 micrometers in diameter, axonemes are large enough to require tomography, but not too large to obtain a reasonable level of resolution, i.e., 3 - 4 nm. The second major advantage is that axonemes are composed of a fairly well ordered arrangement of distinct components including dynein arms, radial spokes, 9 outer doublet microtubules, a pair of microtubules in the center, and the central sheath material surrounding the central microtubule pair (see Figure 1). All of these components are well characterized by previous studies.
Figure 1. Illustration of the arrangement of components in the cross-sectional view of a cilium. The microtubules, dynein arms, radial spokes, and central sheath are all indicated.
Figure 2. A 100 nm thick section cut from a preparation of plastic-embedded Chlamydomonas flagella. Individual protofilaments of the microtubules are visible in the cross-sectional view. Dynein arms, spokes, and the central sheath, are also discernable. The repeating units of the central sheath are seen in the longitudinal view in the upper left of the figure (see the black arrows).
Figure 3. Frozen-hydrated flagellum from a sea urchin sperm. Colloidal gold was added to serve as fiduciary markers for electron tomography. The suspension was then applied to a holy carbon EM grid and rapidly frozen by plunging it into liquid ethane. Images are recorded at 400 kV accelerating voltage on a JOEL 4000 IVEM (intermediate voltage electron microscope).
Resolution assessment. The complex arrangement of different sized components in axonemes provides a yardstick for quality assessment at multiple levels of resolution because the resolution requirement for unambiguous detection is: 1) 25-30 nm for outer doublet and central pair microtubules; 2) 15-20 nm for radial spokes; 3) 10-15 nm for central sheath subunits; 4) 5-10 nm for inner and outer dynein arms; and 3-6 nm for tubulin subunits. In our tomographic reconstructions, we readily detected doublet and central pair microtubules, radial spokes and central sheath components (Figures 3-7). Dynein arms were detected with more difficulty and the tubulin lattice was not detected. This indicates that our overall resolution was 6-8 nm.
Figure 4. A single x-y slice through the tomographic 3D reconstruction of the flagellum in Figure 3. Several components of the structure such as the central sheath (
), spoke heads (
), and dynein arms (
) are visible.
Figure 5. A single y-z slice that provides a longitudinal view of the central sheath (arrows).
Figure 6. A y-z slice showing the longitudinal repeat of the spokes.
Figure 7. An x-y slice containing four rows of dynein arms (arrows).
Layer lines. An additional advantage of the axoneme as a test specimen is that the dynein arms, radial spokes, central sheath components, and microtubule subunits are all arranged as repeating units along the axial direction (e.g. Fig. 2). The periodic and partially helical arrangement of these component structures gives rise to a series of layer lines in the diffraction patterns of both projection images and 2D slices from the tomographic reconstructions (Figure 8).
Figure 8. Diffraction patterns taken from single regions of (a) the projection image in Figure 3; and (b) the tomographic slice in Figure 4.
Assessing structural damage due to electron irradiation. Since these layer-lines arise from the averaging of regularly arrayed unit structures, they cannot be used to assess the resolution obtained for unique structures. Nevertheless, layer-lines can be used to assess the variation of image quality with imaging conditions and electron irradiation. This is illustrated in Figure 9 where the total amplitude of the 1/(8 nm) and 1/(16 nm) layers lines from projection images is plotted against the total electron exposure of the specimen. Note how lower resolution information (the 1/(16 nm) layer line) is more robust to electron irradiation than higher resolution information (1/(8 nm) layer line). Nevertheless, some structural information is detected at 8 nm even after exposure to 11,000 electrons/nm2.
Figure 9. Effect of cumulative electron dose on the layer-line intensity. Layer-line intensities were measured from projection images of an exposure series, as described in McEwen et al., 2002. Intensity of the 1/(16 nm) layer line remains steady throughout the series, while the 1/(8 nm) layer line shows an initial sharpo drop, followed by a more gradual decline throughout the exposure series.
Limits to resolution in frozen-hydrated electron tomography. Traditionally the resolution is described in terms of the size of the specimen and the number of projection images collected in the tilt series:
Where d is the obtained resolution, a is the tilt angle interval, and D the diameter of cylindrical specimen or the thickness of a slab-like specimen (the latter requires cosine tilt scheme; see the paper for a full discussion of resolution). This relationship indicates that resolution is limited by the number of tilt views that can be collected before the specimen is damaged by electron irradiation (Figure 9). However, the principle of dose fractionation (see McEwen et al., 1995) demonstrates that a given electron dose can be fractionated over any arbitrary number of projection images without loss of signal in the 3D reconstruction. Thus, in principle, the requirement for angular sampling is not a limitation because one can simply record tilt series images at a finer angular increment while keeping the total electron dose the same by reducing the electron exposure for each individual image.
With the development of automated data collection, extensive dose fraction has become more feasible because it is now possible to collect images at a finer tilt increment with little or no "wasted electron exposure" (specimen exposure not used for image recording). Nevertheless, Grimm et al. (Biophys. J., 74: 1013-1042 (1998)) pointed out another practical limitation to dose fractionation: sufficient electrons must be recorded for each image to give a statistically significant image on the recording device (CCD camera). According to Grimm et al., it will be possible to obtain resolutions approaching 2 nm for objects that are less than 100 nm in thickness because small objects do not require as fine of an angular sampling (see above equation), but specimens thicker than 500 nm will be limited to 6 - 10 nm resolution.
The analysis of Grimm et al., does not take into account the fact that as the desired resolution increases, so does the total electron dose required to insure statistical significance of each volume element at the resolution limit (Saxberg and Saxton, Ultramicroscopy, 6, 85-90 (1981)). This requirement, which also applies to 2D images, is independent of the size of the object or the recording media used. A poor recording media can impose further limits to resolution, but a superior media cannot overcome the statistical nature of the electron beam. Our analysis of Saxberg and Saxton's relationship shows that it will be difficult to push resolution past 5 nm without severe irradiation damage to the specimen because there is a fourth power dependence of desired resolution and required electron dose (Figure 10). The only way to overcome this limitation is to employ some form of averaging over identical, or near identical structural components.
Figure 10. Estimate of the electron dose (P) required for statistical significance plotted as a function of the desired resolution (r). The contrast factors (Cr) are estimates of the mass density, relative to water, that must be resolved for detection of a feature (see paper for details).