All the developments described earlier have combined in the last few years to produce the most fruitful era of 3D-EM so far (Baumeister and Steven 2000; Henderson 2004). tte first successes came from samples that form 2D crystalline sheets. 2D crystals contain a large number of copies of the same object, already put into register thanks to the specimen intrinsic order. In this way, it was possible to trace the peptide chain of the light-harvesting complex (Kuhlbrandt et al. 1994) and of bacteriorhodopsin (Henderson et al. 1990; Kimura et al. 1997), in cryo-EM maps with 3.0-3.5-A resolution. After that, about 20 more proteins were reconstructed to resolutions in the 4-8-A interval, including aquaporin (Murata et al. 2000) and other membrane channels, specimens particularly refractory to X-ray diffraction techniques. Soluble proteins that form crystalline sheets under certain conditions can also be approached by this methodology, as in the successful determination of the tubulin structure (Nogales et al. 1998).
Reconstruction of 2D crystals suffers from a strong resolution anisotropy. Owing to the physical impossibility of tilting the sample holder in the microscope to angles larger than approximately 60°, there is a lack of data in some orientations, ttis results in a significant decrease of resolution in the direction perpendicular to the sample plane. Samples that form tubular crystals or highly ordered helical assemblies are particularly advantageous in this respect, since in those cases the images provide a more complete set of projections for the object, ttus, atomic models have recently been obtained using 4-Ä-resolution 3D-EM maps for the acetylcholine receptor (Miyazawa et al. 2003) and the bacterial flagellum (Yonekura et al. 2003).
Although they give excellent results regarding resolution, crystalline samples suffer from an important drawback: they require crystallization of the sample. Fortunately, with 3D-EM we can also approach single particle specimens, i.e., isolated biological objects that display low or no symmetry. However, this kind of object usually adopts random orientations on the specimen support, and the difficulty in determining the relative orientations of the thousands of particles one needs to average reduces the practical resolution. Nevertheless, even at 7-Ä resolution it was possible to trace the polypeptide chain of the hepatitis B virus core protein (Böttcher et al. 1997; Conway et al. 1997). Icosahedral viruses represent especially favorable cases, because their high symmetry (60-fold) decreases the number of particles required to calculate the reconstruction. In this manner, resolutions better than 10Ä have also been attained for several other viruses, including herpes, one of the largest known viruses (Zhou et al. 2000), the enveloped dengue virus (Zhang et al. 2003), and cytoplasmic polyhedrosis virus (Zhou et al. 2003). tte single particle approach applied to low-symmetry samples can reach subnanometer resolutions too, and the results have been pushed to limits difficult to imagine few years ago. ttat is the case for the astonishing 6-Ä-resolution map obtained for the GroEL chaperone in native conditions (Ludtke et al. 2004). tte 10-Ä objective is also getting closer for particles without any symmetry, as proved by works on ribosome structure (Valle et al. 2003a, b) and the small ribonucleoprotein U1 (U1 snRNP), a particle fundamental for the correct gene reading in eukaryotic organisms, which is in the lower limit of size approachable by 3D-EM (Stark et al. 2001).
In everyday work, the potential of cryo-EM is limited by instabilities of the sample, radiation damage, aberrations of the magnetic lenses, and inaccuracies during the digital processing of the images. In most of the studies, these and other factors constrain the 3D maps to moderate resolutions. In this situation, the combination of 3D-EM reconstructions of whole complexes (at medium resolution) with the structures of each individual component (at atomic resolution) appears as a critical tool. Available atomic coordinates of protein domains or subdomains can be fitted to EM maps of large biological macromolecules, as if solving a jigsaw puzzle (Fig. 10.3). Indeed, this hybrid methodology has already produced models with a precision of around 4Ä for several viruses (Grimes et al. 1997; San Martin et al. 2001; Mukhopadhyay et al. 2003), including the one that causes dengue fever (Kuhn et al. 2002), or the bacteriophage T4 baseplate, composed by 15 different proteins (Kostyuchenko et al. 2003), as well as for the cytoskeleton components tubulin and actin (Steinmetz et al. 1998; Nogales et al. 1999). Subsequent resolution of the same specimens by X-ray crystallography has experimentally confirmed the accuracy of
Fig. 10.3. The hybrid approach: a 3D-EM map for bacteriophage PRD1 (San Martin et al. 2001) at moderate resolution (a) supplies enough constraints to fit the atomic coordinates of P3 (b), the major coat protein, and to construct a quasi-atomic model of the viral capsid (c). m indicates the presence ofan inner membrane
Fig. 10.3. The hybrid approach: a 3D-EM map for bacteriophage PRD1 (San Martin et al. 2001) at moderate resolution (a) supplies enough constraints to fit the atomic coordinates of P3 (b), the major coat protein, and to construct a quasi-atomic model of the viral capsid (c). m indicates the presence ofan inner membrane these models (Grimes et al. 1998; Abrescia et al. 2004). tte power of this hybrid approach has been demonstrated in difficult scenarios, such as the flexible docking of ribosomal RNA and dozens of ribosomal proteins within cryo-EM maps for the ribosome under large conformational changes (Gao et al. 2003).
Was this article helpful?