Specimen Preparation And Electron Microscopy

Preparation of specimens for electron microscopy is a critical step that must take into consideration the goals of the project and the limitations imposed by the electron microscope. There are many books describing general techniques and methods for biological electron microscopy, and the reader should consult these for detailed protocols (7,22,40). Generally, the goal is to obtain the 2D or 3D structure of the protein at the highest possible resolution. The limitations imposed by the electron microscope are:

• The specimen should be thin, generally less than 1000 A. A related goal is to obtain a preparation of 2D crystals that are a single unit cell thick, rather than multiple layers forming small 3D crystals.

• The specimen must be stable at very high vacuum, approximately 1 x 10-7 Torr or better. This usually requires removing most or all of the water, unless the specimen is maintained at very low temperature, -160°C or below. Removal of water will generally destabilize the specimen conformation, so one is then faced with finding some mechanism to stabilize and support the specimen structure.

• Biological specimens yield very low contrast, as they are composed of light elements. Contrast is commonly increased by adding some form of heavy atom contrast agent (e.g., stain), but adequate contrast can be achieved on unstained specimens by adjusting the focus of the objective lens (see below).

3.1. Negative Staining

One of the most common techniques of specimen preparation is negative staining first developed in the 1960s. Negative staining works well with 2D membrane protein crystals, but the useful resolution that can be achieved is limited to about 10 A, and 20 A is more common. The specimen in an aqueous suspension of approximately 1 mg protein/mL is adsorbed onto a thin carbon film. Carbon films are generated by evaporation of carbon onto a suitable surface, for example freshly cleaved mica, from which it can be floated off onto a water surface and lowered onto the surface of electron microscope grids (commonly 400 mesh copper grids) previously placed on a support beneath the film. The carbon film can be lowered gently onto the grid surfaces by draining water out of the vessel. Alternatively, one can coat grids with a plastic film, generally Collodion, but Formvar is also used, after which a carbon film is evaporated onto the plastic film. These carbon—Collodion films can be used directly, but best results are usually obtained if the plastic is dissolved away with a solvent, since pure carbon films are thinner and more conductive giving less movement when irradiated with an electron beam. Ideally, one should use the thinnest carbon film possible in order to minimize its contribution to the image, but very thin carbon films are fragile, and 400 mesh grids are too coarse to support them adequately. In this case, one can first coat the grid with a holey film, a plastic film containing numerous circular or ellipsoidal holes of varying sizes. There are several methods by which holey plastic films can be formed (36,57,63), and they are generally stabilized to electron irradiation by evaporation of carbon onto them. Holey films make an excellent support for very thin carbon films and can even be used to support thin layers of stain over the holes without a carbon support film.

Freshly evaporated carbon films are generally hydrophilic and readily adsorb proteins and lipid vesicles. As they are stored, however, they quickly become more hydrophobic and then do not adsorb the specimen as well. This change is readily observed as one draws a liquid drop off of the grid with a piece of filter paper. If the liquid draws off evenly leaving a thin film, the grid is hydrophilic, but if it is all rapidly drawn off leaving none behind, the grid is hydrophobic and will probably not adsorb the specimen. One can overcome this change to some extent by increasing the concentration of the specimen, but it is often necessary to render the surface hydrophilic again by a process called glow discharge, in which the films are exposed to the ions formed by an electric discharge in the residual gases in a vacuum apparatus pumped down to several tens of microns of pressure. A glow discharge accessory is common in commercial vacuum evaporators found in electron microscopy laboratories, or an inexpensive glow discharge apparatus can be constructed from a plastic vacuum dessicator, a Tesla coil vacuum tester, and an inexpensive rotary pump (3). Procedure 2 describes a typical method to prepare negatively stained specimens.

❖ Procedure 2. Preparation of Negatively Stained Specimens

1. Allow a drop of the specimen to adsorb to a carbon-coated grid for 1 to 5 minutes.

2. Wash the grid with several drops of water or buffer followed by several drops of a suitable negative stain, commonly 1% to 2% uranyl acetate, but uranyl formate and phosphotungstic acid are also used. Uranyl stains generally provide higher resolution, but must be used at a pH of about 4.5, which may disrupt the structures of some specimens.

3. Draw off excess stain with a piece of filter paper leaving a very thin layer that dries down forming a glass-like electron dense replica surrounding the specimen molecules supporting and contrasting them.

In the electron microscope, the actual protein structure appears light against a dark background formed by the stain, hence the term "negative" stain. It is important to realize that negative staining contrasts the 3D surface of the molecule with atoms that are approximately 7 Â in diameter. Thus, the resolution in negative-stained specimens is limited to approximately 10 Â at best and more commonly 20 Â. Furthermore, the internal structure of protein domains are not contrasted.

3.2. Other Methods

Images of negatively stained specimens are a projection of the electron density of the 3D stain replica onto the 2D image plane. Ultimately, one would wish to calculate a 3D structure from multiple images of tilted specimens (see below), but information on the 3D configuration of the molecule can often be quickly obtained by other common specimen preparation techniques. Indeed, this information is often essential to confirm the molecular packing of a new crystal form, information that is required in calculating the 3D structure. One of the most useful techniques is heavy metal shadowing of freeze-dried specimens (Procedure 3).

❖ Procedure 3. Heavy Metal

Shadowing

1. Adsorb the specimen to a flat hydrophilic surface, such as a carbon film or freshly cleaved mica.

2. Rinse with water or volatile buffer to remove sample that is not adsorbed to the surface.

3. Freeze by plunging into a cryogen; liquid nitrogen is acceptable, but ethane or propane cooled in liquid nitrogen freezes much more rapidly.

4. Place the sample in a freeze fracture—etch instrument and remove ice by sublimation at approximately 70°C.

5. Contrast the surface by evaporating a heavy atom, generally platinum evaporated with carbon, but tungsten-tantalum may produce smaller metal grains yielding higher resolution (12). This creates a shadow effect that highlights the surface topography and can also be used to measure the thickness of the specimen.

This technique has been applied to contrast selectively the surface of vesicle crystals of cytochrome oxidase dimers indicating that the enzyme protrudes 20 to 30 A beyond the bilayer surface on the exterior surface (corresponding to the matrix side of the inner mitochondrial membrane) (33) and to measure the thickness of a number of 2D crystals (37). Berriman, Leonard, and coworkers used shadowing to demonstrate that crystals of the mitochondrial cytochrome bq complex thinned markedly during electron irradiation (5), and Smith and Ivanov have published a procedure to compute the surface relief structure from images of shadowed specimens (67). A related technique, freeze fracture-replication, can also be used to study the structure of membrane protein crystals in order to help determine the molecular packing (13). Heuser has adapted the technique of rapid slam freezing and freeze fracture-etch to look at molecules adsorbed to a slurry of small mica chips (44). Conventional plastic embedding and thin sectioning can also be used to evaluate the gross structure of a new crystal preparation of a membrane protein and to help confirm the molecular packing model (31,51,80).

3.3. Low Dose Electron Microscopy

The challenges of specimen preparation for electron microscopy are compounded by the sensitivity of biological specimens to electron irradiation. The exposure required to record a single image at moderate resolution, approximately 1 nm, can be as high as 100 to 300 electrons/A2. However, measurements of electron damage at much lower exposures paint a gloomy picture for prospects of achieving even this modest resolution. An exposure of even 20 to 30 electrons/A2 results in loss of 20% to 30% of the mass of a typical biological specimen (20), exposure of less than 10 electrons/A2 disrupts the higher order features of a protein crystal (38), and an electron dose of approximately 0.5 electrons/A2 is sufficient to inactivate enzymes (39). The primary function of stain is to increase the contrast of biological specimens so lower electron doses can be used to achieve useful images. Furthermore, heavy atom stains are more resistant to damage by electron irradiation. Nevertheless, studies in the 1970s demonstrated the efficacy of recording electron micrographs using minimal electron exposure even for negatively stained specimens (77,83). Now, most electron microscopes allow one to focus and correct astigmatism on an area of the specimen grid adjacent to the specimen and then record an image of the specimen, exposing it to only the electrons required to expose the photographic film. This procedure is absolutely essential when recording images of unstained specimens that are much more sensitive to electron irradiation than are stained specimens.

3.4. Unstained Specimens

Ideally, one would like to record electron micrographs of unstained specimens, since the resulting images would be of the actual biological molecules rather than the distribution of heavy atom stains around them. There are several problems in achieving this goal, however, beginning with the low contrast afforded by biological specimens and by their sensitivity to exposure to high energy electrons. Through the use of low dose techniques, one can record electron micrographs of unstained specimens at electron doses low enough to minimize radiation damage. The disadvantage is that although these images may technically be high resolution, the signal-to-noise ratio (S/N) is very low, often less than 1.0, and cannot be interpreted. The S/N can be increased by averaging many images of identical structures such as the unit cells of a crystal; statistically the S/N is increased by a factor equal to the square root of the number of structures averaged. This can be a very powerful tool in the case of 2D crystals, where a very small area might contain 100 unit cells giving an increase in the S/N of tenfold. A more typical situation would be a crystal containing several thousand unit cells giving and an increase in S/N of thirty- to fiftyfold. This was first demonstrated by Unwin and Henderson with their images of unstained purple membrane containing thousands of bacteri-orhodopsin molecules; these electron micrographs appear featureless, but the averaged image was clearly interpreted at 7 A resolution as resulting from the presence of transmembrane a-helices (78).

The remaining problem is how to prepare unstained specimens for the high vacuum conditions in an electron microscope. Unwin and Henderson dried their purple membrane specimens in a thin layer of 1% glucose in order to surround them with a hydrophilic substance that could also support them structurally. This approach has worked very well with purple membrane and with a number of other examples, but has the disadvantage that glucose has a density very similar to protein and thus actually reduces the already low contrast rather than increasing it. This was acceptable in the case of bacteriorhodopsin molecules as nearly all of the protein lies within the lipid bilayer surrounded by the lower density hydrocarbon tails of the lipid molecules. Cytochrome oxidase crystals, however, project much of their structure beyond the lipid bilayer surface, and these portions of the structure are virtually invisible above 10 A resolution if the crystals are embedded in glucose (17,18,42). Better results have been obtained with auroth-ioglucose, a glucose derivative containing gold atoms, or with glucose mixed with uranyl acetate (79). These mixtures provide low resolution contrast of the hydrophilic domains of membrane proteins while embedding the structure in a hydrophilic substance. Kuhlbrandt and others have obtained excellent results using tannic acid rather than glucose (54).

3.5. Frozen Hydrated Specimens

The ideal method is to maintain an aqueous environment around the crystal as is the case with 3D protein crystals studied by X-ray diffraction. Although environmental chambers have been constructed that maintain significant partial pressure of water around the specimen by differential pumping, these proved too unstable for high resolution imaging. Taylor and Glaeser demonstrated another approach, freezing the specimen in a thin layer of ice and keeping it frozen at low temperature, -130°C or less, with a specially designed cryoelectron microscope stage (72). The early attempts, particularly by Dubochet's group and Unwin's group, provided very useful results but suffered from stage vibrations that limited resolution. This spurred efforts to construct more stable cryospeci-men holders now marketed by Gatan

(Pleasanton, CA, USA) and by Oxford Instruments (Concord, MA, USA) for the popular side entry electron microscopes. The techniques of cryoelectron microscopy have been described in an excellent review by Dubochet et al. (21), and I will only summarize the important points here. The key to this technique is to freeze the specimen very rapidly in a thin layer of water. With freezing velocities above approximately 10 000 degrees/second, the water is transformed to vitreous ice, a noncrys-talline ice form that has a density and structure similar to liquid water. It is very difficult to freeze a thick specimen this rapidly, but a thin layer of water clinging to an electron microscope grid can be readily frozen to vitreous ice by plunging it into a suitable cryogen such as liquid propane or liquid ethane cooled by liquid nitrogen. The specimen grid is held in a pair of fine tweezers clamped into a simple device for plunging, and Procedure 4 is followed (see Figure 43 in Reference 21).

❖ Procedure 4. Preparation of Frozen Hydrated Specimens

1. Apply 1 to 5 pL of the specimen suspended in a suitable buffer at relatively high concentration, approximately 5 to 20 mg/mL, to a grid with a hydrophilic substrate, either continuous carbon film or holey film that has been recently glow discharged.

2. Blot the grid by pressing it firmly by hand between two layers of filter paper.

3. Plunge immediately into the cryogen; this step is facilitated if the plunge device has a foot pedal release.

4. Transfer very quickly to liquid nitrogen, quickly flicking off excess cryogen, and store under liquid nitrogen until use.

The result is a specimen embedded in a material very similar to its native environ ment at a very low temperature where movement is inhibited. The specimen may be adsorbed to a thin carbon film as for negative staining, or it may be suspended over the holes of a holey film. The latter method has the advantage that the specimen is not in contact with a solid support prior to freezing, but the holey film has a significantly smaller area suitable to record images. The specimen grid should be frozen immediately after blotting, but the thin film of water supporting the specimen may still dry significantly if the relative humidity of the environment is low. To minimize drying, one can maintain higher humidity around the specimen by:

• Blowing humidified air across it (21).

• Freeze in a cold room where humidity is high.

• Use a specially constructed freezing device that incorporates a humidity chamber (64).

Adrian et al. have adapted this procedure to incorporate heavy atom salts in the vitrified water layer in order to increase the contrast of the frozen specimens (1). A very important benefit of cry-oelectron microscopy is the reduction of electron beam damage at low temperature. In measurements of loss of higher resolution information as a function of electron irradiation, specimens at -170°C can be exposed to 5 to 10 times the number of electrons as those at room temperature (38).

The low contrast provided by the relatively small density differences between vitreous ice and protein can be enhanced by appropriate choice of focus of the objective lens. In the brightfield mode of a transmission electron microscope, contrast is generated by two mechanisms: (i) amplitude contrast is generated when electrons are scattered by the specimen at a wide enough angle to cause them to be intercepted by the objective aperture subtracting them from the image. This is analogous to absorption contrast in the light microscope and contrasts relatively low resolution details; and (ii) phase contrast is generated when the phases of electrons are retarded as they pass through the specimen. The phase of these electrons are further modified by the objective lens, and the extent of this phase shift depends upon the:

• Angle of diffraction.

• Spherical aberration of the objective lens.

• Focus of the objective lens.

Thus, it is possible to control the amount of phase shift of the diffracted electrons by changing the focus of the objective lens, and with an appropriate level of underfocus, some of the diffracted electrons can be further phase-shifted by approximately 90°, generating appropriate phase contrast when combined with undif-fracted electrons at the image plane. But for each choice of underfocus, only electrons diffracted at particular angles are phase-shifted by 90°, generating proper contrast. Electrons diffracted at other angles are phase-shifted by smaller amounts, generating less contrast, or are phase-shifted in the wrong direction, generating inverted contrast.

Diffraction angle correlates with resolution, and electrons diffracted at higher angles contain higher resolution information. The changes in the phase of electrons as a function of their diffraction angle is described by the contrast transfer function (CTF) (23,26). In order to visualize individual macromolecules, one must adjust the objective lens to a relatively large underfocus, one micron or more. This generates contrast of lower resolution, features making the molecules visible, but may introduce contrast reversals at high resolution. Figure 3c is an optical diffraction pattern (equivalent to a plot of Fourier transform intensities) of the cytochrome oxidase crystal in Figure 3 a, and the effects of the CTF can be seen in the concentric rings of high background noise. The regions of the Fourier transform, where phases have been shifted in the wrong direction generating reversed contrast, are shown in the plot of the CTF displayed as an insert in Figure 3c on the same scale as the as the diffraction pattern. By definition, CTF values greater than zero represent incorrect phase shifts,

Figure 3. (a) An electron micrograph of a frozen hydrated crystal of cytochrome oxidase dimers; one unit cell is outlined. (b) A Fourier-filtered image with dramatically increased S/N calculated from 5 electron micrographs similar to panel a. One unit cell is outlined with unit cell axes of a = 100 A and b = 125 A. (c) Optical diffraction pattern of the crystal in panel a; the optical diffraction pattern is equivalent to a plot of the intensities of the Fourier transform. The reciprocal lattice vectors, a* and b*, are indicated. The inset is a plot of the phase CTF, x(a), f°r this defocus shown on the same scale, and the zeros in the CTF are indicated by horizontal lines showing regions of minimal contrast in the diffraction pattern.

Figure 3. (a) An electron micrograph of a frozen hydrated crystal of cytochrome oxidase dimers; one unit cell is outlined. (b) A Fourier-filtered image with dramatically increased S/N calculated from 5 electron micrographs similar to panel a. One unit cell is outlined with unit cell axes of a = 100 A and b = 125 A. (c) Optical diffraction pattern of the crystal in panel a; the optical diffraction pattern is equivalent to a plot of the intensities of the Fourier transform. The reciprocal lattice vectors, a* and b*, are indicated. The inset is a plot of the phase CTF, x(a), f°r this defocus shown on the same scale, and the zeros in the CTF are indicated by horizontal lines showing regions of minimal contrast in the diffraction pattern.

and the circled diffraction spots lie within rings of the diffraction pattern that have been phase-shifted in the wrong direction. In order to calculate a high resolution image of a biological specimen, one must correct for the effects of the CTF. Since different values of underfocus optimally contrast features at different levels of resolution (different diffraction angles), it is sometimes advantageous to record more than one image of the same specimen, the first at lower defocus for high resolution information, and the second at greater defocus for lower resolution information

3.6. Collecting Tilt Data

Transmission electron micrographs are 2D projections of the 3D electron density of the specimen. While these projections reveal important information about the structure of the specimen, much detail is lost when structural features are projected upon one another. If the specimen is tilted and another micrograph recorded, a different set of features will be superimposed, providing new information about structure. This is most readily seen if the two projections are viewed as a stereo pair. A full 3D reconstruction requires many images of the specimen tilted at different angles, greatly exceeding the number that can be recorded from a single specimen without very significant radiation damage to unstained specimens (14,43). Thus, low dose images of many different but identical specimens must be recorded in order to sample the 3D Fourier transform (see section 3.6 below and Figure 4). All images must be translated to bring them into alignment at a common origin, so it is usually necessary to record images at a variety of tilt angles and merge the data, beginning with images recorded with the smallest tilt and then adding images in order of increasing tilt angle. In the case of negatively stained specimens, one normally records several images at different tilt angles for each specimen, since these specimens are more resistant to radiation damage. Unstained specimens are much more sensitive to electron irradiation, and generally, only one high resolution image is recorded from each. Specimens are more stable to irradiation at low temperature (10,38), so it is possible to record more than one micrograph from a single specimen at low temperature, particularly if the highest resolution is not required.

3.7. Specific Labeling

Most studies of protein structure by electron crystallography do not yield resolution sufficient to construct an atomic model, and methods to identify the locations of functionally important sites and/or components are needed in order to exploit fully lower resolution structures. There are two general approaches to identify specific sites on low resolution structures: (i) specific labeling with molecules visible by electron microscopy; and (ii) comparison with structures lacking one or more components.

Both of these approaches have been applied with success to low resolution structures of heme proteins derived from electron microscopic data.

The most common method of specific labeling in electron microscopy of biological structures is the use of specific antibody molecules. Antibodies are most commonly used to determine the distribution proteins in cells and organelles, but can also be used to identify the position of antibody epi-topes on molecular structures. Frey et al. used subunit-specific antibodies to determine that the surface exposed in vesicle crystals of cytochrome oxidase dimers corresponded to the matrix surface of the inner mitochondrial membrane, concluding that the interior surface of the vesicles corresponds to the surface exposed to the intermembrane space (30). They subsequently prepared monovalent antibody fragments, Fabs, to label subunit IV on the surface of these crystals, concluding that this subunit lies 20 to 30 A from the 2-fold axis of the dimer and near the a crystal axis (32). Based upon the prediction of a transmembrane a-helix from residues 80 to 97, the volume of the N-terminal domain of subunit IV could account for the 20 to 30 A structure projecting from the bilayer surface that was detected in freeze-dried and shadowed specimens (34). Fab fragments have also been used as bulky affinity labels to identify their corresponding epitope binding sites by electron microscopy in many other specimens (2,82).

In many cases, the protein being studied binds another protein with sufficient affinity and specificity that the protein ligand can be used to label its binding site. This is the case with cytochrome c, one of the substrates for cytochrome c oxidase. Frey and Murray (35) incubated crystals of cytochrome oxidase monomers with cyto-chrome c, which binds to cytochrome oxidase with an affinity comparable to that of specific antibodies. After extensive image processing, the site of cytochrome c binding to cytochrome oxidase monomers was deduced from difference images (Figure

Figure 4. Lattice lines of the 3D Fourier transform of a 2D crystal. The positions of the reflections in the diffraction pattern of an untilted crystal are shown as black ellipses. In the 3D Fourier transform, these reflections extend perpendicular to the plane of the crystal as shown by the lines of periodically varying intensity. The Fourier transform of an electron micrograph of a crystal that has been tilted is a central section of the 3D Fourier transform that intercepts the lattice lines at the points shown by the X's.

Figure 4. Lattice lines of the 3D Fourier transform of a 2D crystal. The positions of the reflections in the diffraction pattern of an untilted crystal are shown as black ellipses. In the 3D Fourier transform, these reflections extend perpendicular to the plane of the crystal as shown by the lines of periodically varying intensity. The Fourier transform of an electron micrograph of a crystal that has been tilted is a central section of the 3D Fourier transform that intercepts the lattice lines at the points shown by the X's.

5 a), consistent with the cytochrome c binding site in the atomic structure, determined by X-ray diffraction and from biochemical studies (see section 4.1 and Figure 5b).

Other labeling studies have taken a different approach, using heavy atom cluster molecules that have been modified to react selectively with certain functional groups of proteins, generally reactive sulfhydryl groups of cysteine residues (29). A special issue of The Journal of Structural Biology is devoted to results from this approach using gold cluster compounds (1999, volume 127, issue 2). Crum et al. used a mono-maleimide derivative of an undecagold cluster compound to label specifically Cys-115 of cytochrome oxidase subunit III in crystals of cytochrome oxidase dimers. They then identified the binding site by low dose cryoelectron microscopy of specimens embedded in glucose and uranyl acetate (16).

A different approach to identify the various components of a macromolecular com plex is to compare structures of the intact complex with structures of subcomplexes. This approach was used in the study of the mitochondrial cytochrome bcj complex. Weiss, Leonard, and coworkers crystallized Neurospora mitochondrial cytochrome c reductase (cytochrome bcjor Complex III) by reconstituting it with purified lipids and adsorbing excess detergent with Bio-Beads. This produced crystals of the type shown in Figure 1c, although these were generally formed in the two layers of a collapsed vesicle giving two overlapping crystalline layers (56,84). Their low resolution 3D reconstruction of the intact complex is shown in Figure 6b. Hovmoller et al. subsequently formed crystals of the purified subcomplex lacking two large "core" proteins, and comparison of the 3D structure with that of the intact complex allowed them to identify the functional components as shown in Figure 6 (46,47). The core subunits, which probably function in facilitating the assembly of the complex, were later purified, and their

Figure 5. A comparison of the structure of cytochrome oxidase monomers determined by electron crystallography (a) and by X-ray crystallography (b). (a) A low resolution structure in projection calculated from electron micrographs of frozen hydrated crystals of cytochrome oxidase monomers. The dark peak outlined in white contour lines is the position of cytochrome c binding calculated from difference images. (b) A ribbon diagram produced from the atomic coordinants calculated from the high resolution X-ray structure and displayed by the program RasMol. The cytochrome c binding site is placed between Cys-115 of subunit III and the acidic residues of subunit II as determined biochemically.

Figure 5. A comparison of the structure of cytochrome oxidase monomers determined by electron crystallography (a) and by X-ray crystallography (b). (a) A low resolution structure in projection calculated from electron micrographs of frozen hydrated crystals of cytochrome oxidase monomers. The dark peak outlined in white contour lines is the position of cytochrome c binding calculated from difference images. (b) A ribbon diagram produced from the atomic coordinants calculated from the high resolution X-ray structure and displayed by the program RasMol. The cytochrome c binding site is placed between Cys-115 of subunit III and the acidic residues of subunit II as determined biochemically.

low resolution structure was determined from helical aggregates confirming the assignment in Figure 6 (48). The cytochrome bfcomplex found in the thylakoid membrane of chloroplasts has a function in photosynthesis very similar to that of cytochrome bc1 in mitochondria, but lacks the core subunits. The low resolution projection structure of the cytochrome b6f complex purified from Chlamydomonas reinhardtii was determined by electron crystallography of 2D crystals grown by reconstitution and found to be very similar to the cytochrome bc1 subcomplex lacking the core subunits (8).

0 0

Post a comment