III.F.1. Sample Preparation

a. Support films (see also Sec. II.A.1)

Most specimens are supported on a thin film of plastic or evaporated carbon (Sec. II.A.1). Micrographs, therefore, contain an image of the specimen superimposed with that of the support film. The added substructure of the support film constitutes a major source of noise that obscures fine specimen details (Sec. III.B.1). The support film also reduces specimen contrast. Film thickness should be kept as small as practical to achieve best results. Although not generally required, with certain specimens, ultra thin films (<5 nm) or no film at all may be useful.

Contrast is enhanced by suspending the specimen over holes in the support film. Several techniques for producing perforated films are available. When negatively-stained samples are suspended over holes, the thin layer of stained specimen usually breaks as it dries or when it is irradiated by the electron beam. Those areas which survive (remain stretched over the holes) usually contain distorted specimen due to the contraction of stain upon drying. Regions where the stain-specimen layer is broken are often uneven and do not lie flat and perpendicular to the electron beam (Josephs & Borisy (1972) J. Mol. Biol. 65:127-155). Uneven distortions in the specimen may not be apparent unless the specimen is tilted and viewed from at least two different directions. Regular (non-perforated) support films can also be uneven and induce distortions in the specimen (Kellenberger, et al. (1982) Ultramicrosc. 9:139-150). These effects are often ignored but should be recognized. In addition, although the specimen may remain intact, it may lie tilted relative to the electron beam due to support film unevenness. Precise knowledge of the specimen orientation is essential to assure that 3D structure is correctly determined (Sec. III.D.3).

b. Dehydration effects

Most biological specimens collapse and distort as a result of surface tension forces that occur during dehydration. Single, isolated particles tend to distort more than crystalline specimens in which there is extra support from the crystal lattice contacts between neighboring particles. Uneven distortion may occur if the specimen is supported anisotropically. Many specimens collapse to a greater extent on the side facing away from the grid. This presumably occurs if there is a thin layer of stain only supporting the side of the specimen in contact with the grid. Preparation procedures cause specimen flattening and shrinkage by as much as 30%. Some distortions, such as flattening in helical particles, are detected by optical diffraction methods (Sec. III.D.1) since such distortion leads to obvious loss of symmetry in the diffraction pattern. Similar results are found even for intact specimens if they are non-uniformly penetrated by stain.

Dehydration damage may be reduced by maintaining the aqueous medium or by replacing it with a suitable, substitute medium. Negative stains help reduce particle collapse by forming a protective, supporting cast around the specimen. This protective effect probably outweighs any distortion the stain itself induces in the specimen. Other established techniques, known to reduce dehydration damage, include critical point drying (Hayat & Zirkin, (1973) Princ. Tech. Elec. Microsc. 3:297-313) and freeze drying (Sec. II.A.6). Techniques for studying hydrated and frozen-hydrated specimens are among the best ways to eliminate dehydration damage (Sec. II.A.5).

Chemical cross-linking may be used to stabilize specimens (Sec. II.A.2), but this is generally considered a drastic procedure, one to be avoided if possible. Agents such as formaldehyde, gluteraldehyde, and dimethylsuberimidate have been used with some success, mostly to avoid specimen dissociation or denaturation. Cross-linking, of course, is required to stabilize specimens that are dehydrated and embedded for thin-sectioning work (Sec. II.A.2). Ethylene glycol is recommended as a dehydration solvent for sectioning studies because it is thought to be less denaturing to protein than agents like acetone.

Dehydration damage to unstained catalase and purple membrane specimens is reduced by adding glucose or other sugars as substitutes for the aqueous medium (Sec. II.A.5). Glucose, used in the same manner as negative stains, forms a glass-like, matrix that embeds and supports the specimen. Sugars differ from negative stains since, if anything, they reduce inherent specimen contrast.

c. Contrast enhancement

The low contrast inherent in biological specimens is typically enhanced by adding heavy atom compounds to the sample before or after it is deposited on the grid (Sec. II.A). Negative staining, shadowing, and positive staining of thin sections are common methods by which this is achieved. The negative staining technique is often preferred because it is probably the simplest method to learn and practice (Sec. II.A.3). Negative staining also generally provides the best resolution of specimen features compared with other conventional methods. Typically, details ~2-3 nm can be resolved in negatively-stained specimens, whereas resolution in shadowed or sectioned material is limited to about 5 nm. Some specimens, however, can not be examined successfully using negative staining techniques. For example, intracellular inclusions often lose crystallinity when isolated in vitro and must be studied by sectioning methods. In other instances, metal shadowing is preferred if details of one surface of the specimen structure are of most interest.

Most biological specimens studied by image processing have been prepared by conventional, negative staining methods. Among the many available stains, 1-2% aqueous solutions of the uranyl or tungstate salts are most commonly used. Dilute (0.1-0.5%) stains are sometimes used to reduce stain inhomogeneities in structures with very low inherent contrast. Specimen and stain may be applied to the grid separately or as a mixture. If applied separately, the specimen should not be allowed to dry before the stain is added. Ideally, stains surround and coat the accessible surfaces of the specimen in a non-specific manner, and when dry, form a cast that maps out the surface topology. If large enough cavities in the specimen are present, some stain may penetrate and thereby raise the contrast of internal features (e.g., central hole of the TMV helical structure). Specimen structure is inferred from the regions where stain is excluded. The 2-3 nm resolution limit of specimen features partly results from the finite size of the stain molecules (0.5-1 nm) and partly reflects the fidelity with which the stain penetrates and coats the specimen. Resolution is also limited by radiation effects (Sec. II.B).

An ideal stain does not interact chemically or physically with the specimen nor does it cause the specimen to dissociate, denature or lose activity. Because biological specimens are inherently sensitive to changes in pH, stains are customarily used in a pH range in which they are soluble and compatible with specimen stability and activity. When specimen structure is unknown, it is important to try several different stains under varying conditions of pH and concentration. Stains which give reproducible results often, but not always, give the most faithful rendition of structure. Stains sometimes interact favorably with the specimen, stabilizing it for microscopy. For example, microtubules depolymerize in the presence of phosphotungstate stains but appear intact when uranyl stains are used. Apparently, microtubules are fixed at the low pH values characteristic of most uranyl salt solutions.

Specimen appearance depends on which stain is used and the conditions of staining. Catalase platelet crystals and gap junction membranes, for example, have been studied with a variety of negative stains. Which images most faithfully represent the native structures? Different stains often alter specimen structure in undetermined ways. Uranyl acetate seems to stain catalase in a consistent manner with resulting images closely resembling the structure determined by x-ray crystallography. When a new or unfamiliar specimen is examined, the choice of which stain to use may not be obvious.

Some (negative) stains also act as positive stains if they interact with specific groups on the specimen. Uranyl compounds are often used in x-ray crystallography of proteins because the compounds form isomorphous derivatives specifically bound to carboxylic acid groups. Molecules that act as positive stains are advantageous for locating specific chemical moieties in macromolecules. Nevertheless, specific binding of negative stains to specimens can bias structural interpretation since the positively-bound stain may remain undetected due to the dominant negative stain effect.

Unfortunately, the accuracy by which artificial contrast methods reveal genuine specimen features always raises some doubt. At best, only those features in contact with negative stain can be revealed at 2-3 nm resolution. Features not contrasted with stain (therefore not observed in the image), can not be detected. Artifacts from uneven staining (surfaces near the grid often stain more heavily than those away from the support film) lead to incorrect representation of specimen detail. Some of these artifacts can be detected by image processing. Conversely, uneven staining may be useful in determining the handedness of specimen features (Sec. III.E.1).

Methods developed to examine unstained, hydrated specimens (Sec. II.A.5) reveal surfaces and internal structural details and avoid staining artifacts. The types of specimen that can be studied with these techniques seem virtually limitless with the application of newer and more powerful computer processing methods (Sec. III.D.3). No matter what method is used to prepare specimens, it is prudent to examine several images and ascertain what features are reproducible and thus are likely to represent genuine specimen structure.

III.F.2. Imaging Conditions

a. Contrast enhancement (see also Sec. I.C)

Images in the electron microscope form when incident electrons are scattered by the specimen and focused by one or more electromagnetic lenses. Electrons scatter elastically, without energy loss (velocity and wavelength remain unchanged) if they encounter the nuclei of specimen atoms, or inelastically with loss of energy (velocity decreases and wavelength increases), when the orbital electrons of the specimen atoms are encountered. Elastic collisions are non-destructive and result in deflection of electrons through a wide range of angles. Inelastic events generally involve deflections through small angles (<10^-4 radians) and cause specimen damage (Sec. III.2.B).

The amount of scattering is proportional to the mass thickness of the specimen (product of density and thickness). Thick specimens, or those containing a large number of heavy atoms, scatter more electrons than ones that are thin or have low average atomic number. Typical specimens are thinner than 50 nm, consequently, most electrons pass through them without scattering (recall, atoms are mainly empty space).

Image contrast arises from a combination of two processes. Highly deflected, elastically scattered electrons are physically blocked by the objective aperture in the back focal plane of the objective lens (see Handout #25). Dense or thick specimen regions appear dark because a large portion the electrons they scatter never reach the image plane. This is called scattering (amplitude) contrast. The heavy atoms in negative stains increase the scattering contrast in images of biological specimens.

The phase of an incident electron wave (recall that moving electrons behave as waves and particles) is changed by inelastic scattering and interference (phase) contrast is produced when this wave interferes with unscattered electrons. Specimens that are thin or mainly contain atoms of low atomic number are considered to be phase objects. Thus, unstained biological specimens are essentially pure phase objects. It is estimated that thin, negatively-stained biological specimens are imaged with 35% scattering and 65% interference contrast (Erickson & Klug, 1971).

Microscopists are well aware how focusing the objective lens alters image contrast. Slight defocusing enhances phase contrast because the scattered and unscattered rays recombine with maximum phase difference. Best results are obtained for negatively-stained objects with an underfocus of 300-500 nm.

Multiple (dynamical) scattering becomes a factor if the mass thickness is so large that a beam electron scatters more than once as it passes through the specimen. The resultant image artifacts can lead to incorrect interpretation of specimen structure. Image processing is often helpful in detecting the presence of these artifacts (Voter & Erickson (1974) Elec. Microsc. Soc. Amer. Proc. 32:400-401). Thin, negatively-stained objects (<50 nm thick) generally scatter kinematically (single scattering) so multiple scattering effects can be ignored.

Dark field imaging (Sec. I.F.2) dramatically enhances specimen contrast by removing unscattered rays. Although there are only a few examples of image processing applied to dark field images (e.g. Ottensmeyer, et al. (1977) J. Microsc. 109:259-268), this imaging technique, especially in conjunction with scanning transmission electron microscopy, may provide an important alternative to conventional studies by bright-field, transmission microscopy. Further discussion of image processing applications, with particular reference to dark field images, is not given here.

b. Radiation effects (see also Sec. II.B)

The destructive effect of the electron beam on specimens poses severe restrictions in the study of biological structure at high resolution. The single, most significant advance in specimen preparation and imaging techniques was the development by Unwin and Henderson (1975) of very low-dose electron microscopy of unstained, unfixed crystalline specimens. Their remarkable study of the 3D structure of the purple membrane at 0.7 nm resolution revealed the alpha-helical secondary structure in protein subunits. This low-dose technique has now been used successfully to study large, periodic arrays of unstained specimens, while minimal-dose methods based on the original Williams and Fisher (1970) method are widely practiced with negatively-stained specimens. Normal, high-dose microscopy, in which no precaution is taken to minimize the total dose, leads to stain migration and crystallization and results in lower resolution of specimen details in negatively-stained samples. High-dose microscopy of unstained specimens is impractical because it leads to massive structural damage, leaving little or no discernible features in the micrographs.

The main advantages of low-dose imaging techniques are: i) minimization of structural changes due to beam damage, ii) manifestation of genuine contrast in unstained structural features, and iii) reduction of dehydration damage in specimens where the aqueous medium is replaced with glucose or other low-molecular weight sugars or the aqueous medium is retained in a frozen state. These techniques offer an optimum way to study the native structure of specimens in the electron microscope at high resolution. Unfortunately, only a few studies of this type have been successful at high resolution (<1 nm) because the specimen must be well-ordered over large areas (1-10 µm). This severely limits the number of specimens suitable for study by these methods. Unless ways are developed for inducing in vitro formation of large, periodic arrays, most specimens, including helical and single, isolated asymmetric particles, will be difficult to study at such high resolution.

c. Non-linearities and other artifacts (also see Secs. I.B-E)

Image processing results are influenced by the conditions of microscopy at the time the images are recorded. Instrumental aberrations are a major source of image artifacts which can lead to misrepresentation of structural features in specimens. Digital image processing methods can be used to compensate for correctly diagnosed artifacts. The contrast transfer characteristics of the electron microscope (defocus, spherical aberration, beam coherence, astigmatism, etc.) are routinely detected by optical diffraction and can be quantified and corrected with digital processing procedures. Such analysis is especially important for high resolution (<2 nm) work.

One problem not yet sufficiently resolved concerns the fact that some specimens display high resolution (0.3-0.4 nm) information in low-dose electron diffraction patterns, but similar resolution can not be obtained from optical diffraction patterns of images recorded with low dose techniques. Instead, the resolution in images is limited to 0.7-1.0 nm. Electron diffraction results clearly show that specimens can be preserved in the microscope at near-atomic resolution. However, this information is not recorded in the image. Specimen movement and distortion, and inelastic scattering are thought to be among the major factors which reduce image contrast, and hence detection of high resolution features.