Search

A system is presented for semi-automatic specimen selection and data acquisition for protein electron crystallography, based on a slow-scan CCD camera connected to a transmission electron microscope and control from an external computer. Areas of interest on the specimen are localised at low magnification and subsequently imaged on the CCD camera, using a dose which is small compared to the dose used in the exposure mode. The crystalline quality of the area is evaluated from the appearance of diffraction peaks in the calculated image Fourier transform. If the quality is considered good, images can then be recorded in different modes, both on film and using the CCD camera. Using this system a significant gain, both quantitatively and qualitatively, can be obtained in acquiring data for electron crystallography of beam-sensitive materials.

Structural differences between oxy-hemocyanin and deoxy-hemocyanin from the spiny lobster P. interruptus were studied by electron microscopy and image analysis of negatively stained preparations. Projections of the hexameric P. interruptus hemocyanin from electron microscopy were compared with simulated projections of the high-resolution structure determined by X-ray crystallography previously. Image analysis was focused on the hexagonal views. Four independent data sets of hexagons (two of oxy- and two of deoxy-preparations) were analyzed and compared. Despite the small size of the projections, a resolution in the average images of better than 1.5 nm was obtained. A significant difference in stain modulation could be noticed between the oxy- and deoxy-preparations in two areas of the projections. The first difference in stain exclusion occurred in the center of the hexagonal views. The second is the shape of the domain 3 region. In the deoxy-images this region has a triangular shape, in the oxy-images it appears to be more rounded. The overall differences suggest a “breathing” of the hexamer upon oxygenation/deoxygenation.

The secretion apparatus known as the needle complex (NC) from the bacterium Shigella flexneri was studied by single particle electron microscopy. The isolated intact NC appears in projection to be composed of a basal body consisting of seven rings and a protruding needle appendage. A comparison of averaged projections of the intact NC and its fragments revealed the organization of the NC into several major subcomplexes. One of these lacks an inner membrane ring of the basal body but still presents the needle appendage attached to four upper rings. The position of the needle appendage within these rings is variable, suggesting that the dissociated component is necessary for stabilizing the needle appendage. Averaged images of the subcomplex lacking the inner membrane basal rings show a thicker extension at the base of the needle appendage, called the socket. This socket was also found to be present in images of the basal body fragment isolated from mutants lacking the mxiH and mxiI genes. This suggests that the socket is not composed of MxiH and MxiI subunits, which form the needle appendage. A symmetry analysis of the basal body top view projections indicated that a peripheral protein component of the inner membrane ring is present in a ring with 24 copies, in contrast to the Salmonella typhimurium NC. A model is presented in which the NC is only associated to the outer- and inner-membranes with its first and seventh ring, respectively.

The Na+-pumping V-ATPase complex of the thermophilic bacterium Caloramator fervidus was purified and dissociated under controlled conditions. The structure of purified V-1-ATPase subcomplexes differing in subunit composition was analyzed by electron microscopy and single particle analysis of 50 000 projections. Difference mapping of subcomplex projections revealed the presence and position of two subunits in the central stalk. A density with an elongated shape similar to the gamma subunit of F-ATPases is partly located within V-1 and corresponds, most likely, to subunit E. Subunit E is connected to the membrane-bound part V-0 via subunit C, a spherical density that is connected to the center of V-0. The presence of subunit C makes the central stalk substantially longer in comparison to the F-ATPases, in which the subunit connects directly to F-0.

Electron-microscopic studies revealed that two types of subunits of Panulirus interruptus haemocyanin crystallize in different ways. Homohexamers of subunit a give close-packed two-dimensional crystals whereas homohexamers of subunit c form open two-dimensional arrays. We applied computer-image analysis to these arrays and studied the differences in crystallization properties by combining the electron-microscopic data with amino acid sequence information and the X-ray diffraction model of subunit a.

SUMMARY Our experience in the growth of two-dimensional crystals of different proteins is presented. Polyethylene glycol was used to produce two-dimensional arrays of haemocyanin from O. vulgaris and of cholera toxin. The arrays showed a hexagonal close-packed structure of only randomly oriented molecules. The increase in protein concentration probably occurred too quickly to allow complete crystallization. Different two-dimensional arrays of hexameric haemocyanin molecules (from P. interruptus) were obtained by microdialysis through the specimen supporting film. A comparison was made with X-ray data. Two-dimensional tetrameric arrays of molecules, possibly rhodopsin, were seen in samples of bovine retinal rod outer segments in the presence of ammonium sulphate. Two-dimensional crystals of complex I (from bovine mitochondria) were prepared by dialysis in the presence of ammonium sulphate. A three-dimensional reconstruction was made from two tilt-series by computer filtration using the direct SIRT procedure. Finally, the possibility of computer crystallization using correlation techniques in combination with correspondence analysis is discussed.

Two-dimensional crystals of LH2 (B800–850) light-harvesting complexes from Ectothiorhodospira sp. and Rhodobacter capsulatus were obtained by reconstitution of purified protein into phospholipid vesicles and characterized by electron microscopy. The size of the crystals was up to several micrometers. The crystals of LH2 from Ectothiorhodospira sp. were with a square unit cell (p42 1 2 plane group symmetry) and a repeat of 9.9 urn. By image processing of negatively stained crystals a projection map was obtained at 1.8 run. It shows complexes with an outer diameter of about 6.5 nm and with an octagonal appearance. Crystals of Rb. capsulatus have a hexagonal unit cell with a repeat of 8.1 nm and a resolution of 2.7 nm. A ring-like complex with an outer diameter of about 7 nm and an inner stain-filled indentation with a diameter of 2 nm was obtained. This structure is similar to the one of LH2 from Rhodopseudornonas acidophila , determined at atomic resolution (McDermott et al. (1995) Nature 374, 517–521), which is a multimer of 9 α- and (β-subunits. Taken together, the data indicate that isolated LH2 from Rb. capsulatus is a nonamer, whereas LH2 from Ectothiorhodospira sp. is an octamer.

Electron microscopy can give information about the quaternary and tertiary structure of protein molecules by application of negative-staining, glucose-embedding or freezing procedures. Several methods are known to filter or reconstruct the structural data from electron microscopy (Misell, 1978). To apply this technique on proteins successfully, regular arrays of molecules should be available. Protein crystals can meet these requirements, but in most cases they are too thick for electron microscopy.

Bovine rhodopsin has been clustered into two-dimensional crystals in highly purified native rod disk membranes and studied with negative staining and transmission electron microscopy. The lattice is P2 1 with dimensions of 8.3 × 7.9 nm and interaxis angles of 86 ± 3 degrees. 110 images of ordered areas were digitized and aligned with computer-correlation methods to calculate an average image with diffraction to the fourth order. The images were computer-filtered and reconstructed to approx. 2 nm resolution. When crystals appeared they covered 20–40% of the surface of the preparation and, since rhodopsin is at least 95% of the protein, there is no doubt that the crystals were due to rhodopsin. There appear to be two rhodopsin dimers per unit cell. Each rhodopsin molecules takes up about 7.5 nm 2 of membrane area and is estimated to be associated with about 12 lipids on each side of the membrane. The membrane area found for bovine rhodopsin supports the rhodopsin origin of rarely seen but more highly ordered two-dimensional crystals found in detergent-treated frog rod membranes (Corles, J.M., McCaslin, D.R. and Soctt, B.L. (1982) Proc. Natl. Acad. Sci. USA 79, 1116–1120). Furthermore, the rhodopsin membrane area is close to that of bacteriorhodopsin and is consistent with a seven transmembrane helix structure proposed for rhodopsin (for references see Dratz E.A. and Hargrave, D.A. (1983) Trends Biochem. Sci. 8, 128–131). Crystallization was accomplished by loweing the pH to 5.5 near the isoelectric point of rhodopsin, raising the salt concentration of 2 M (NH 4 ) 2 SO 4 , adding 5% glucose and 0.02% Hibitane (Ayerst), a cationic amphipathic antiseptic that favored crystal growth.

Two-dimensional crystalline arrays of NADH:Q oxidoreductase preparations have been obtained by microdiffusion of protein dissolved in detergent against a 15 mM sodium acetate buffer of pH 5.5 containing 10% (w/v) ammonium sulphate. Electron microscopy was used to study the structure of negatively stained crystals. Computer-reconstructed images were obtained by the Fourier peak filtering method. The crystals have p4 symmetry and a square unit cell with dimensions of 15.2 +/- 0.5 nm. The four asymmetric units in the unit cell form a single tetrameric molecule with a dimension in the third direction of 8.2 nm. It is concluded on the basis of the estimated molecular mass that each tetramer cannot contain more than only one FMN molecule. This implies that the tetramers possibly are only a part of Complex I, since there is much evidence that one functional enzyme molecule of Complex I contains two FMN molecules.