prepared by
Andreas Engel, Shirley A. Mueller and Thomas Walz
INTRODUCTION
In spite of major advances in X-ray crystallography , and recent developments in NMR spectroscopy , the structure of most membrane proteins has remained elusive. Of large diversity in function, these proteins share the feature of a hydrophobic belt by which they interact with the lipid bilayer. This attribute explains the slow progress of structural analyses for which membrane proteins need to be solubilized with detergents and purified. Solubilization tends to destabilize the proteins, particularly when short chained detergents as required for 3-dimensional (3-D) crystallization are used. Thus, solubilized proteins rarely form crystals suitable for X-ray analysis. A powerful alternative is the reconstitution of 2-dimensional (2-D) membrane protein crystals in the presence of lipids. In this approach, the native environment of membrane proteins is restored, as well as their biological activity [Walz et al., 1994a]. Electron crystallography [Henderson et al., 1990] can then be used to determine the protein structure at atomic resolution . In addition, the atomic force microscope (AFM) has demonstrated its potential to directly visualize conformational changes of membrane proteins at subnanometer resolution [for details see Müller, Aebi & Engel, this booklet]. These developments open the exciting possibility not only to analyze the 3-D structure of membrane proteins, but also to assess the relationship between their structure and function directly .
PRINCIPLE
2-D crystals assemble from the solubilized protein and mixed detergent-lipid micelles by slow removal of the detergent (Fig. 1a). The protein packing arrangement is controlled by the lipid-to-protein ratio (LPR), by the nature of the detergent and lipid, and by salts, pH, and small hydrophilic solutes. Detergent removal is achieved by dialysis, dilution or adsorption to Biobeads. Solubilization and isolation conditions must be tuned so as to maintain the minimum LPR required for the stability of the particular protein. In addition, the optimum conditions (LPR, salts, pH, temperature, etc.) for the reconstitution of 2-D crystals need to be established. Crystallization should be a slow and ordered process where individual units assemble one by one to form a highly ordered array. However, the size of the ternary complexes in a mixture of protein, lipid and detergent exhibits an abrupt change close to the critical micelle concentration (cmc) of the detergent when progressively diluted (Fig. 1b; Dolder et al., 1996). The formation of non-crystalline vesicles or proteoliposomes is characterized by a decrease of their size after the phase transition as dilution proceeds (open circles in Fig. 1b). In contrast, the size of the assembled structures remains at the maximum value attained when crystallization occurs (solid circles in Fig. 1b). In this case, the assemblies are crystalline just after their spontaneous and very rapid formation (Fig. 1c). Optimization of 2-D crystallization is thus intimately related to the abrupt phase transition .
Figure 1. 2-D crystallization of E. coli porin OmpF. (a) The hydrophobic belt of the membrane protein (dark shades) which ensures proper integration into the lipid bilayer is shielded by detergent molecules after solubilization. Elimination of detergent molecules by dialysis leads to the exposure of hydrophobic surfaces of both the proteins and the lipids. Hydrophobic forces then drive the assembly of 2-D crystals. (b) The assembly process starts just before the critical micelle concentration (vertical line) of the detergent is reached, and proceeds very fast. A one-step dilution (open circles) is too fast to allow the system to crystallize. Disordered small proteoliposomes form. During a slow dilution (total time 4 hours) crystallization occurs in spite of the abrupt phase transition (solid circles). Vesicle or sheet-type crystals reach their final size immediately. (c) They exhibit crystalline order directly after the phase transition, as documented by the hexagonal diffraction pattern (inset) characteristic of E. coli porin OmpF lattices. Scale bar, 1 mm.
APPLICATION I: Photosystem I reaction center
A fundamental process of life on earth is the conversion of light energy into chemical energy . In higher plants and cyanobacteria water is split to liberate oxygen, a proton gradient is established which drives ATP synthesis, and NADP+ is reduced. These processes are catalyzed by membrane-bound protein complexes , the photosystem II (PS II), the cytochrome b6/f complex , and the photosystem I (PS I). Electron crystallography provided the first insight into the structure of the PS I reaction center [Ford et al., 1990]. X -ray crystallography has subsequently yielded a 6 Å map of the PS I complex from Synechococcus sp. [Krauss et al., 1993], and most recently a 4.5Å map. We have now improved the 2-D crystallization of the PS I complex so as to obtain precise information on its integration into the lipid bilayer (Fig. 2a and b). Cryo-electron microscopy of these 2-D crystals produced a projection map at 8Å resolution (Fig. 2; Karrasch et al., 1996).
Figure 2. The photosystem I reaction center is an elongated complex with two distinct hydrophilic surfaces. (a) Highly ordered negatively stained crystals exhibit a clear periodic pattern. (b) The 3-D map displays the stromal (S) and the lumenal (L) surface of the complexes. The lipid bilayer is schematically presented by a solid planar surface. (c) The 8 Å projection map of vitrified crystals. The white contour is from the 6 Å map obtained by X-ray crystallography. The correlation with the map from cryo-electron microscopy is excellent, except for the arm ( *) that forms the contacts in the trimeric complex present in the 3-D crystals. The black contour outlines the stromal protrusion revealed in the 3-D map of negatively stained crystals (see b). The unit cell size is 139 Å by 145 Å, the complex has a thickness of 90 Å, and the 21 screw axes are marked by arrows. Scale bar, 0.1 mm.
APPLICATION II: Aquaporin-1, the water channel of the human erythrocyte
membrane
The plasma membranes of specialized mammalian cells, plant cells and bacteria are highly permeable to water molecules as result of water-selective pores known as "aquaporins". The water conducting pore of human erythrocytes (AQP1) was the first aquaporin shown to transport water after expression of its RNA in Xenopus laevis oocytes [Preston et al., 1992]. Aquaporins must contain a highly specific water channel structure, since the passage of ions and small hydrophilic solutes is blocked, whereas the passage of water is unhindered .
AQP1 solubilizes as a tetrameter in detergent and readily forms 2-D tetragonal crystals when reconstituted in the presence of phospholipids (Fig. 3a; Walz et al., 1994a). A 3-D map of the crystallographic unit cell has been determined from tilt series of negatively stained 2-D crystals [Walz et al., 1994b]. As revealed in Fig. 3b, the crystal symmetry is p4221, with two tetramers per unit cell packed in opposite orientations in the lipid bilayer. The asymmetry of the tetramer with respect to the membrane plane contradicts the sequence-based structure prediction. The surface relief of the crystals was therefore determined by EM of freeze-dried/metal -shadowed specimens (Fig. 4). The computed surface relief (Fig. 5b) corroborates the isocontoured 3-D map from negatively stained samples (Fig. 5a) and has also been confirmed by surface topographs of AQP1 in buffer solution (Fig. 5c) recorded with the AFM [see Müller, Aebi & Engel, this booklet]. Cryo-electron microscopy of glucose and trehalose embedded 2-D crystals has also been started (Fig. 6; Walz et al., 1995), and progress toward the atomic structure of AQP1 is documented by the projection maps shown in Fig. 7.
Figure 3. Solubilized aquaporin (AQP1) tetramers assemble into square arrays when reconstituted in the presence of lipids. (a) A faint tetragonal pattern is characteristic of negatively stained crystalline vesicles. (b) The 3-D map of negatively stained AQP1. The information of many projections from 2-D crystals tilted in the microscope with respect to the electron beam was combined to calculate the stain distribution. This dense heavy-metal salt envelope outlines the hydrophilic surface of the protein-lipid crystal. The map is displayed as a stack of horizontal sections, separated vertically by 3 Å, and comprising four unit cells of 96 Å by 96 Å. Scale bar, 0.1 mm.
Figure 4. Surface relief of AQP1. (a) Freeze-dried crystalline vesicles were coated unidirectionally with a 5 Å thick layer of tantalum/tungsten. Polystyrene spheres seen on the left side are used to measure the shadowing elevation and azimuth (arrow). (b) As the metal coat is thin, averaging is required to reveal the shadow cast by the protrusions of the AQP1 tetramers. Scale bar, 1 mm; the box comprising one unit cell has a side length of 96 Å.
Figure 5. Perspective views of the AQP1 surface topography. (a) A solid surface has been calculated by isocontouring (for details see Stoffler & Henn, this booklet) the 3-D map shown in Fig. 3b. The major tetrameric protrusion and the stain -filled indentation are distinct even at a resolution of 20 Å. (b) Much higher resolution (9 Å) is achieved in the surface relief reconstruction from the average shown in Fig. 4b. (c) Surface topography directly recorded with the AFM in buffer solution also exhibits 9 Å resolution.
Figure 6. Crystalline AQP1 lattices are highly ordered as demonstrated by cryo-electron microscopy of glucose embedded membranes. (a) Overview of a flattened vesicle, ~3 mm in diameter. (b) Optical diffraction pattern from the same vesicle but recorded at high magnification. Sharp spots extend to a resolution of 7.5 Å, and the circular zones of the contrast transfer function are clearly depicted. (c) The electron diffraction pattern reveals spots beyond 5 Å resolution.
Figure 7. Progress in resolution of glucose (a), or trehalose (b, c) embedded AQP1 2-D crystals. (a) The 6 Å map shows eight density peaks per AQP1 monomer, most of which probably represent a-helical membrane spans. (b) The 3 Å map has been calculated by merging 10 electron micrographs exhibiting diffraction spots to at least 3 Å resolution. The p422 1 symmetry has been enforced, and the amplitudes have been weighted to compensate for the loss of contrast at high resolution. (c) Significantly finer structural detail is revealed by the map calculated from the phases of map (b) and the electron diffraction intensities shown in Fig. 6c. Atomic interpretation of this map will be possible once the 3-D structure is determined at atomic detail from projections of tilted crystals. Unit cells of 96 Å side length are displayed.
REFERENCES
Dolder, M., Engel, A. & Zulauf, M. (1996). The micelle to vesicle transition of lipids and detergents in the presence of a membrane protein: towards a rationale for 2D crystallization. FEBS Letters 382, 203-208.
Ford, R.C., Hefti, A. & Engel, A. (1990). Ordered arrays of the photosystem I reaction center after reconstitution: projections and surface reliefs of the complex at 2 nm resolution. EMBO Journal 9, 3067-3075.
Henderson, R., Baldwin, J.M., Ceska, T.A., Zemlin, F., Beckmann, E. & Downing, K.H. (1990). Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. Journal of Molecular Biology 213, 899-929.
Karrasch, S., Typke, D., Walz, T., Miller, M., Tsiotis, G. & Engel, A. (1996). Highly ordered 2-dimensional crystals of photosystem I reaction center from Synechococcus sp.: functional and structural analyses. Journal of Molecular Biology, in press.
Krauss, N., Hinrichs, W., Witt, I., Fromme, P., Pritzkow, W., Dauter, Z., Betzel, C., Wilson, K.S., Witt, H.T. & Saenger, W. (1993). 3-Dimensional structure of system-I of photosynthesis at 6 Ångstrom resolution. Nature 361, 326-331.
Preston, G.M., Carroll, T.P., Guggino, W.B. & Agre, P. (1992). Appearance of water channels in Xenopus oocytes expressing red cell CHIP28 protein. Science 256, 385-387.
Walz, T., Smith, B.L., Zeidel, M.L., Engel, A. & Agre, P. (1994a). Biologically active 2-dimensional crystals of aquaporin CHIP. Journal of Biological Chemistry 269, 1583-1586.
Walz, T., Smith, B.L., Agre, P. & Engel, A. (1994b). The 3-dimensional structure of human erythrocyte aquaporin CHIP. EMBO Journal 13, 2985-2993.
Walz, T., Typke, D., Smith, B.L., Agre, P. & Engel, A. (1995). Projection map of aquaporin-1 determined by electron crystallography. Nature Structural Biology 2, 730-732.