To access the private area of this site, please log in.

Username: 

Password:  

 

Chapter 4

Structure and Function at the Subcellular Level

EDUARD KELLENBERGER

 

INTRODUCTION

In this chapter I will examine some of the historical developments that have led to our present state of understanding of the structure and function of the principal bacterial cell components. Much of our present knowledge is described in detail in individual chapters of this book. I will dwell on the interrelationship between morphological techniques and those associated with biochemistry and genetics. My purpose is to illustrate how imaging techniques, chiefly light and electron microscopy, can be interconnected with others, leading to an integrated approach to research on subcellular components.

Biochemical and genetic methods presently available are so efficient in yielding results that it is easy to forget the important role that imaging has played in formulating central concepts of modern biology. The concept of the cell per se is derived from light microscopy; that of cell organelles as sophisticated systems of intracellular membranes is derived from electron microscopy. Understanding the nature of membranes as true physical entities supplanted previous notions of "cellular interfaces" (in German, "Grenzflchen") or the "plasmalemma," thought by some to be a modified (denatured) part of the cytoplasm. The definitive physical proof for the existence of membranes is derived from the electron microscopic observation that the plasma membrane retains its "double track" appearance even when the cell has lost its content. Structural concepts have been equally important in genetics, where the light microscope permitted making the trip from the nucleus to the chromosomes. The electron microscope has allowed us to visualize DNA as a fibrous molecule. Together with X-ray structural studies, this led to a fundamentally new, "non-chemical" concept of specificity through a linear code (for a discussion of this basic change of paradigm see reference 66).

For the sake of clarity I will separate structural investigations into three phases, knowing that such a classification is necessarily artificial.

The first phase of any new imaging technique is descriptive. Organisms are reinventoried as finer and finer levels of imaging are achieved. In defense of imaging methods, it should be noted that an early descriptive phase is characteristic of any new method, as is obvious when considering, for example, DNA sequencing.

The second, "reductionist" phase consists in isolating and purifying the newly discovered structural elements or substructures by using various methods available, and then characterizing them biochemically. Going further, structural investigations with isolated particles are now possible, particularly by crystallographic methods that allow determination of the structure by averaging over large numbers of identical particles.

In the third, integrative phase, the functional activities of substructures and their regulation are investigated. This phase culminates in "holistic" attempts to understand coordinated, cooperative cellular and even organismal functions. Genetic tools, as provided by mutations and genetic engineering, have greatly facilitated physiological studies and allowed their integration with structure or biochemistry. Genetics has in this manner efficiently bridged structural and biochemical studies. Cogent examples of integration of biochemistry, morphology, and genetics are the studies of flagella and motility (chapter 10) and of phage assembly and form determination (see reference 71).

 

DESCRIPTIVE OR DISCOVERY PHASE

Compared to what was seen in eukaryotes, early examination of prokaryotic cells with the electron microscope was rather disappointing. Enterobacteria show an undifferentiated cytoplasm (Fig. 1), later called by some a "bagful of ribosomes," with some ribosome-free spaces containing the nuclear material (14). No nuclear or intracytoplasmic membranes or cytoskeletal fibers were observed. By cell fractionation, small spherical particles were seen early on and were found to contain protein and RNA (103). The same particles, single and in association, were seen at about the same time (but published only later) in in situ-lysed T4 phage-infected cells (72). Later on, these particles came to be called ribosomes and polysomes. In this context, it should be recalled that in the hands of eukaryotic cell fractionators, "microsomes" were an important if nebulous biochemical fraction of cell material. Direct electron microscopy of these particles and thin sections of eukaryotic cells rapidly clarified that this fraction consisted of a mixture of ribosomes and fragments of endoplasmic reticulum. Escherichia coli, with its absence of contaminating membranes, was then chosen as the source of the starting material for the detailed characterization of this important cellular constituent (113).

The development of methodology is sometimes helped by artifacts. For example, nucleoids, well known from Feulgen-like staining and light microscopy (for references, see reference 98), were readily identified with the electron microscope because of an artifact induced by the chemical fixatives used early on. Poor, or even absence of, fixation of the bacterial chromatin resulted, during dehydration, in easily observable aggregated "clumps" within an "empty nuclear vacuole." The many questions posed by this appearance led to the development of more sophisticated fixation techniques and, eventually, to the present state of knowledge (which I will discuss below).

Before the advent of electron microscopic techniques, the envelope of E. coli was known biochemically to consist of several differentially extractable components thought to be interpenetrated and interconnected (120). At that time Weidel and his school called the whole envelope of E. coli "the membrane" and included murein as well as the lipid constituents in this term. Weidel hesitated to accept the existence of physically distinct layers even when electron microscopy clearly showed the envelope to have a separable bipartite nature, later defined as cytoplasmic membrane and cell wall (69). In gram-positive bacteria the situation was more readily intelligible because it was easy to remove a true cell wall with lysozyme (100). A true protoplast surrounded by the plasma membrane was the result (119). When applied to E. coli and other gram negatives, the same procedure resulted in the only partially denuded spheroplasts (15) still surrounded in part by the outer layer which we now know to be the outer membrane. The physical separation of the various layers could be attempted, helped by improved electron microscopy of thin sections (31, 70, 88). This help was limited because of the problem of interpreting where the heavy metals used in fixation and staining were deposited. This might be the right place to recall a famous fallacy of electron microscopy, the "unit membrane" concept (97). According to this model, the double-track appearance of biological membranes was due to lipid bilayers with osmium-stained membrane proteins located on both surfaces. This idea predominated despite the fact that the gene of the E. coli permease, responsible for the uptake and transport of lactose, had been found and that thus a protein accomplished the permeation (58), demonstrating that lipids alone could not perform the function of transmembrane transports. Finally, the Nicolson-Singer model (108) with integral membrane proteins provided a satisfactory explanation for membrane structure and function but also made it evident that heavy metal staining in electron microscopy, the basis for the unit membrane model, has to be interpreted with caution.

Another artifact, observed only in gram-positive cells, was the mesosome, an intracellular membranous structure connected to the plasma membrane. With cryofixation, this organelle is no longer observed, shattering hopes to find a mitochondrial analog. Today it is believed that the appearance of the mesosome is a consequence of the shrinkage of the cell, which is induced by processing for embedding. Lipid bilayers are not compressible (11), and neither are plasma membranes, with their large contents of lipids, as has been postulated by Koch in order to explain phenomena in plasmolysis of gram negatives (A. L. Koch and H. Schwarz, manuscript in preparation). Upon shrinkage of a gram-positive cell the membrane material has to escape in some manner and forms the inrolled mesosome by invagination.

Escherichia and Salmonella spp., and particularly the domesticated strains used in experimentation, show fewer intracellular particles or storage bodies than many other bacteria (106, 116). Particularly the widespread metachromatic or volutin granules, constituted of inorganic polyphosphates, are lacking. Quoting A. Kornberg (73): "Although E. coli, a major source of biochemical insight, lacks any visible content of polyP, it still proved to be a rich source of an enzyme which makes polyP..." In E. coli a polyglycoside storage particle was discovered and then identified as glycogen (22). Such bodies appear when the growth medium has an excess of carbon source and is deficient in other nutrients such as a nitrogen source. In E. coli C, electron-opaque bodies were found (69), but no attempt was made to identify them. For liver glycogen it had been shown that the particles could be lighter or darker than their surroundings, depending on the electron microscopy procedures used. This method of differentiation has unfortunately not yet been applied to the study of a bacterial glycogen. Observed "white" inclusion bodies can thus consist of either nonstained substances—including polysaccharides, such as glycogen, starch, and the cell walls of plants—or material dissolved during dehydration in organic solvents that leave an empty "hole" behind. The latter occurs most likely when inclusions consist of polyester oligomers, e.g., polyhydroxyalkanoates, which have acquired interest as the starting material for biodegradable resins (19, 56). A first representative of this class of polymers was polyhydroxybutyrate, discovered in 1926 (78) and observed by electron microscopy as inclusions in Bacillus cereus and Bacillus megaterium (81). Because the empty holes in E. coli have a similar appearance to those of polyhydroxybutyrate, the presence of glycogen was doubted. Recently it was shown that E. coli is able to synthesize small amounts of hydroxybutyrate, found in the membranes (105). It seemingly lacks the enzymes for polymerizing it into the storage form found as inclusion bodies. The whole set of enzymes needed for producing these compounds could be transferred into E. coli from Pseudomonas oleovorans (40).

Intracellular membranes and fibers with well-defined functions are common features in higher cells. In prokaryotes they are rather the exception. With the advent of transgenic microorganisms it was observed that an overproduced gene product frequently leads to overproduction of membranous or fibrous materials. Many overproduced enzymes polymerize into tubular structures. Most often, however, the overproduced material collects in inclusion bodies or "lumps." A number of intracellular membranes with specific functions were found in photosynthetic and methane-oxidizing bacteria, but not yet in Escherichia and Salmonella. Intracellular fibers were observed in archaebacterial species (12), although it remains to be shown that they are functional and not just the polymerized product of the genes of a defective prophage, as was found in Proteus mirabilis (111, 116). If functional intracellular fibers are present in a small number per cell, as may suffice for the separation of dividing nucleoids (51) (part V of this book), they could never be detected by sectioning techniques. In this case, mutationally induced overproduction might be of help, as in the appearance of large fiber bundles documented by Okada et al. (90).

Extracellular flagella and fimbriae or pili of E. coli and other microorganisms were readily seen early on with the electron microscope. This served to clarify a dispute on the nature of flagella stemming from the suggestion by Pijper (94) that they represented trailing slime material. This controversy could not be resolved by light microscopy (for references, see references 32 and 55), and the definite proof that flagella are real fibers came from electron microscopy (85, 115; chapter 10, this volume).

Pili or fimbriae (chapter 11) were discovered and investigated with the electron microscope (16, 33, 55). One class of fimbriae was found to be important in bacterial adhesion and thus related to pathogenicity (10, 80). Another type of these structures was found to be essential in the process of conjugation and to be associated with the F-plasmid (18, 29). Filamentous appendices of a rather unusual cross-striated morphology were found in the Lisbonne strain of E. coli (76). When present in large numbers, fimbriae and other fibrous structures form a kind of capsule that can be detected in the light microscope by negative staining with India ink. The presence of a capsule is often reflected in the appearance of the colonies (i.e., rough or smooth) of solid-medium-grown cells (16). By electron microscopy, the slime of the gram-negative Acetobacter was very early on shown to consist of cellulose fibers (86). A capsule of E. coli that has receptor activity for a specific phage was of particular interest for electron microscopic studies (110). The spikes of the baseplate of this phage showed endoglycosidase activity (39), which allows it to dig its way through the capsule (8).

In general, the fibrous macromolecules that make up such capsules are too thin to be seen individually. These macromolecules exhibit a strong tendency to aggregate during the procedures conventionally used for thin sectioning. More appropriate techniques have become available, such as those depending on intermediate treatment with dimethyl formamide (7) or the use of cryofixation.

 

ISOLATION AND CHARACTERIZATION OF SUBCELLULAR STRUCTURES

Once discovered and morphologically defined, a subcellular structure can be isolated and purified using morphology as a criterion to check successive steps of fractionation. In its purified state, the structure can be not only biochemically characterized but also further investigated by structural methods. When two- or three-dimensional crystals are available, electron microscopy and atomic force microscopy (AFM) with image processing, or diffraction methods with electrons or X-rays, can be employed (for more details, see methods section below).

The isolation of flagella (118) led to successful studies of their fine structure (79) and laid the basis for the discovery of the flagellar hook (24) and the basal organelle or basal cap structure (74). With this background information, elegant studies on mobility could be performed (chapter 10), a fine example of the third or integrated phase.

Similarly, the fine structure and composition of capsular fibers and of fimbriae were investigated very early on by X-ray diffraction (17, 82, 84).

The principal structures of gram-negative bacterial envelopes that have been isolated and thoroughly analyzed are the inner membrane (chapter 7), the outer membrane (chapter 5), and the peptidoglycan layer (chapter 6). In addition, considerable knowledge has been gathered about the periplasm, occupying the space between the two membranes (chapter 8). Intensive study of the porins of E. coli (chapter 5) has provided much of the methodological basis for studying integral membrane proteins in general. Porins assemble relatively easily into regular, two-dimensional crystals that are highly suitable for study with the electron microscope (59). Processing the information obtained has resulted in imaging at a level of near-atomic fidelity (37). From the available studies it became evident that the function of these integral membrane proteins is reflected in their conformational changes (27). In these studies, electron microscopic analysis of two-dimensional crystals has proven to be an extremely useful complement to X-ray diffraction of three-dimensional crystals. Not only are three-dimensional crystals difficult to obtain per se, but frequently crystals of only one of several structural forms are obtained. On the other hand, X-ray diffraction permits a level of resolution that is more directly related to the visualization expressed by chemical structural formulas (28). AFM, which is also very suitable for the study of two-dimensional crystals (Fig. 2), has recently allowed investigators to progress towards increased resolution (43, 101, 102). Combining diffraction results with those of the various imaging procedures now available has a very high potential for the future. X-ray diffraction which yields atomic resolution, but restricted choice of conformational states, is complemented by images with lower resolution but easier access to functionally relevant conformations.

The bacterial nucleoid has no limiting membrane, and transcription and translation are coupled events in prokaryotes. It is not surprising therefore that the isolation of this organelle, if it can be designated as such, has not been very successful and reproducible. To visualize the DNA-containing material only, the adhering ribosomes with the mRNA have to be stripped off by forces able to rupture any noncovalent bond involved in the organization of the bacterial chromosome. The folded nucleoid inside the cell is extended linearly by more than a factor of 10; in the isolated state it occupies a volume that is 1,000 times larger than within the cell (see Fig. 5 in reference 64). This situation was documented by the beautiful micrographs of Kavenoff et al. (60, 61) They showed "exploded" nucleoids with some 200 independent loops, each tethered to a structure that may serve as a scaffold. To identify proteins by using immunocytochemistry, similar preparations were urgently needed. Unfortunately it has not been possible to reproduce them with similar quality in other laboratories. For the closest repeat, albeit without supercoils, see reference 23.

In vivo psoralen binding studies have shown that bacterial chromatin is negatively supercoiled (93). The existence of individually supercoiled loops has been demonstrated by X-ray-induced single-strand breaks that led to their relaxation (107; also see chapter 12). Approximately 160 single-strand breaks were needed to relax the whole genome. Relaxation can apparently also be induced by milder treatment, as is shown by the fact that even on Kavenoff’s micrographs not all the 10-fold extended loops retained all or part of their supercoil, which is not solenoidal but plectonemic, i.e., stretched out as one is used to see in vitro with supercoiled circular plasmids. Plectonemic supercoiling is not able to significantly compact the DNA, quite in contrast to the solenoidal supercoil observed with eukaryotic chromatin, associated with the binding to a core of histones (Fig. 3). Recent electron microscopic examinations of sections of cryofixed bacteria show that the DNA-containing plasm has a granular aspect, suggesting strongly the presence of solenoidal or stochastic supercoils, which would have the required high ability of compaction. The problem here is the insufficient amount of histonelike proteins that are needed to form a compact nucleosomic core for a stable solenoidal supercoil, as is the case in eukaryotes. The prokaryotic supercoils must therefore be of a quite different type than the eukaryotic ones (67, 91; B. Arnold-Schulz-Gahmen, B. Bohrmann, L. Falquet, R. Gyalog, R. Johansen, M. Maeder, J. Pelzer, and E. Kellenberger, submitted for publication; preprints are available from the authors).

Cytoplasmic fractions from a number of bacteria contained relatively large particles that were easily resolved even by old-time electron microscopy. These particles were first proposed to be the ribosomes and later, when the real ribosomes were identified, to be polymerases of various types. By better electron microscopy they were later shown to have the shape of cylinders with a diameter of approximately 13 nm and a length of approximately 12 nm. Uncertainties regarding their functions persisted until one kind of these particles with an unusual sevenfold symmetry was shown to be composed of the GroE protein (53). E. coli groE mutants lead to incorrect assembly of phages (25, 41); by cloning the gene in bacteriophage λ, its product could be identified as a protein of 65 kDa (42, 50). Genetically induced overproduction of this gene product then allowed for a clear identification of the function associated with this particle, made of 14 GroE protein subunits (53). GroE is the first representative of the large class of chaperonins, involved in the correct folding of particular proteins (47, 49; also see chapter 61). Besides GroE, morphologically similar particles of about the same dimensions have been found. They can barely be distinguished, even pushing techniques of image processing, except for a few cases where the particles have other than sevenfold symmetry. In most cases their functions remain to be described. The prosomes (9) and the proteosomes (92, 95) have so far only been found in archaebacteria and eukaryotes. A proteolytic function has been clearly demonstrated for the proteosomes.

 

INTEGRATED STUDIES

In my introduction I defined the goal of integrated research to be the establishment of functional relationships by using a combined morphological, biochemical, and genetic approach. Both in vivo and in vitro work have their foibles and must be interpreted with caution. Thus, we must keep in mind that the ionic and small solute conditions used in biochemical studies are far from what they are intracellularly. Within the actively metabolizing cell the K⁺ concentration is approximately three times that of the Na⁺ concentrations of most of the common growth media and buffers. Intracellularly, Na⁺ and Cl^– are virtually absent, where the main small-molecular-weight ionic compounds are K⁺, balanced with glutamate and other amino acids, and organic phosphates (20, 21, 30, E. Kellenberger and A. Kuhn, submitted for publication). The amount of intracellular Mg²⁺ is also not negligible (83), nor are those of putrescine and spermidine, which are present in nearly equimolar concentrations (75). Noncovalent interactions are very dependent on ionic concentrations and temperature. For example, hydrophobic interactions, which are prevalent in protein-protein binding, are promoted by high salt and high temperature. The assembly of protein subunits is thus shifted towards dissociation when intracellular (multi-subunit) particles are released by cellular lysis into a surrounding buffer. In addition, by the law of mass action alone, dilution also acts in favor of dissociation. Protein-nucleic acid interactions, in contrast, are hindered by high salt, but to a lesser extent with glutamate than with chloride (77). Kuhn has introduced the notion of "experiments with open cells," where the intracellular conditions of still metabolizing cells are altered by manipulations from the outside in order to allow for experiments half way between in vitro and in vivo. Experiments along this line have been carried out with bacteriophage multiplication in E. coli (for references, see reference 75 and Kellenberger and Kuhn, submitted).

For light microscopy, cytochemical staining procedures have provided a fair link with biochemistry. Nothing like this existed for ultrathin sections until the introduction of electron microscopy-autoradiography, using metabolites radioactively labeled with ³H, ¹⁴C, ³⁵S, or ³⁶P. Later on, the powerful techniques of immunocytochemistry were introduced (for technical references, see reference 48), and important applications contributed toward the understanding of the arrangement of components of the nucleoid (14, 34, 52, 112) (see also chapter 12). At this writing, the possibilities offered by immunolabeling with colloidal gold and particularly by the new procedure of immunostaining on thin sections (13) are far from being exploited to their full potential.

The elucidation of the mechanisms of replication (89) and transcription (122) is a prominent example of an integrated approach that used conditional lethal mutants of coliphage T4. Although derived from studies with phage, the resulting models proved to be of general relevance (chapters 50 and 55). Curiously, no structural methods were involved in the construction of these structural models!

 

SOME UNSOLVED, CURRENT PROBLEMS

I believe that for some time to come E. coli will continue to provide the ideal experimental system for many fundamental life processes. Both in vivo and in vitro experiments can be carried out with this organism with nearly equal precision, although sometimes requiring unequal efforts. Such complementary experiments impose themselves more and more because, as we have seen above, they are carried out under quite different conditions. Obviously, the in vitro study of many regulatory processes does not fully reflect the in vivo situation. Examples are assembly processes, e.g., phage-merids (subassemblies, like capsids or tails) that are assembled in vitro from protein subunits using temperature upshifts and/or increases in salt concentrations. Nothing of the sort occurs in vivo, where new precursor particles are initiated randomly in time. Each particle has its own morphogenetic clock which decides upon the uptake of an additional protein selected out of the "soup" of precursor molecules (for references, see references 63 and 71).

Our knowledge of the integration ("assembly") of proteins into lipid bilayers is quite advanced (chapter 7), as is that on the genetic and biochemical identification of certain membrane proteins, for instance the pumps (chapter 72). However, investigations of their structure-function relationships, e.g., the mechanism of pumping, are still at a timid start. Best understood is the comparatively simple function of the porins (104). The study of the water pores with their most important regulatory roles has just begun for eukaryotic cells (38, 117). Protein transport through biological membranes is also fairly well understood, particularly from studies with bacteriophage fd and with the "honorary coli," the mitochondrion. In contrast to the passage of proteins, the mechanisms of uptake and secretion of DNA still remain a riddle (chapter 132).

The problems of protein transport get more complicated when two biological membranes have to be crossed, as with gram negatives (chapter 7) and mitochondria (44). In E. coli a periplasmic gel keeps the distance between the two membranes constant, although discussions about the actual width of the space are far from settled. The bacterial periplasmic gel and the murein layer make for a situation that is certainly more complicated than with mitochondria, where apparently both are lacking and where the two membranes are most likely only regionally and temporarily in close contact (54). A similar, although more complicated situation is imaginable also for E. coli and other gram-negative bacteria, where for the transport across the two membranes, permanent connections between the inner and outer membranes were postulated and visualized as "adhesion sites" which appear after strong plasmolysis of the cells (1, 2, 3, 4, 5). However the occurrence and aspect of these fibrillar, slimelike connections are not reproducible features. Using what is thought to be the same approach, the results obtained by the same persons can vary significantly, as can be seen in the above cited publications. Thus, immunolabeling of proteins postulated to migrate through connecting zones has not yielded convincing results. Most importantly, it has not yet been possible to distinguish between ubiquitous present periplasmic proteins and those transported across the periplasm. Our own unpublished results show that when the conventional procedures are used, a majority of the periplasmic oligosaccharides and periplasmic proteins are released into the medium. This release is not significantly inhibited by the currently used chemical fixations. The evidence thus accumulates that these adhesion sites do not correctly represent the sites and structures of the transperiplasmic transports (for references, see references 62 and 123). This point remains controversial (6) and efforts are urgently needed to better understand the mechanisms underlying plasmolysis (68; Koch and Schwarz, manuscript in preparation; H. Schwarz and A. L. Koch, manuscript in preparation).

A major topic that requires an integrated approach is the way the genetic material of bacteria distributes itself equally between the daughter cells. Functionally, chromosome equipartition represents nothing more than a primitive form of mitosis. Since the bacterial genome consists only of one chromosome (a single linkage group) that does not undergo cycles of condensation and decondensation, the morphological situation is obviously different from that of the eukaryotes. In addition, bacterial chromosome replication is continuous, so that a chromosome of actively dividing cells is "branched" by a number of replication forks, depending on the growth rate. In addition, transcription and translation need not to be cyclic and, at least in rich media, protein synthesis takes place continuously over the cell cycle. All this, taken together with the chromatin being supercoiled, leads to a very complicated topological situation. How can these branched chromosomes disentangle and separate into two daughter nucleoids? How can the replication and transcription forks move along the solenoid?

For the segregation of the bacterial chromosomes, Jacob et al. (57) formulated the attractive hypothesis of the attachment of the dividing bacterial chromosome to the cell envelope. The attachment points of two daughter chromosomes would be separated by intervening envelope growth. Experiments in this field have produced contradictory and controversial results and alternative hypotheses are presently being studied intensively (chapters 100, 101, and 105).

As mentioned above, eukaryotic HP-chromatin (called HP because of its high protein content, 1:1 histones/DNA by weight) cannot be used as a model for prokaryotic chromatin because it must differ in the type of compacting supercoiling (Fig. 3). The amount of the most abundant of the histonelike proteins in E. coli, HU, is at least five times too small to produce—for the whole chromosome—the solid nucleosomic cores seen in eukaryotes. The smaller relative amount of these proteins explains the aggregation sensitivity of bacterial low-protein chromatin (called LP) observed in electron microscopy studies. Analogously, when the weight ratio of histones to DNA is below 0.5, eukaryotic chromatin assembled in vitro becomes as aggregation sensitive as that of bacteria (67, 91; Arnold-Schulz-Gahmen et al., submitted; see Isolation and Characterization of Subcellular Structures, above). "Compactosomes," formed with possibly solenoidally, but more likely with stochastically, supercoiled LP-chromatin (Fig. 3), were observed with in situ-lysed bacteria (45). Unfortunately, they were interpreted as being proof for the presence of nucleosomes. Similar structures can be obtained by treating protein-free DNA with ethanol (35, 67). The putative bacterial compactosomes are known to be so highly unstable that, in contrast to eukaryotic nucleosomes, their isolation has not yet been achieved.

Several experimental observations, discussed in reference 98, strongly suggest that the LP-chromatin of energy-deprived cells readily undergoes profound structural changes. A stretched-out chromatin appears to be able to form liquid crystals. This is what is observed on ultrathin sections of chemically fixed DNA-containing plasms: in contrast to the rather granular aspect after extremely rapid cryofixation followed by freeze-substitution, here we see fibers arranged in some sort of nicely ordered bundles and whorls (Fig. 4). Their occurrence is explained as follows: in a first step only the cytoplasm is fixed; the cell is killed and no longer produces the energy required to maintain supercoiling. The nonfixed chromatin has time to stretch out and to crystallize before it is postfixed with uranyl salts hours later (67; E. Kellenberger et al., unpublished data).

 

SOME METHODS IN STRUCTURAL BIOLOGY

 

Studies of Entire Nondissociated Cells

The conventional methods (staining or phase-contrast light microscopy and staining for electron microscopy of ultrathin sections, freeze-fractures, and negatively stained specimens) remain the choice for the characterization and comparative studies of previously unstudied species. Thin-sectioning procedures additionally provide for a very simple test for deciding between the aggregation-sensitive LP-chromatin of the known prokaryotes and the insensitive HP-chromatin of most of the eukaryotes (for references, see reference 98). After a conventional, cytological fixation, preferably with aldehydes, the material is split into two aliquots, only one of which is treated by postfixation with uranyl acetate (note that the osmium fixation under strict RK conditions [i.e., without phosphates present, which remove the added Ca²⁺] is not considered a conventional cytological fixative, because it is able to cross-link bacterial chromatin without postfixation in uranyl acetate [for references, see reference 98]). With LP-chromatin the postfixed aliquot will show a fine fibrillar structure of DNA-containing plasm (Fig. 4c and d), while the other will have coarse aggregates. With HP-chromatin no significant difference will be detected. It should be noted that in eukaryotes mitochondria, plastids, and kinetosomes also contain LP-chromatin, as do the chromosomes of the eukaryotic dinoflagellates.

With the ability to produce vitreous ice, cryotechniques have recently received much attention (96, 109). Cryofixation followed by freeze-substitution has proven to have substantial advantages (65), as mentioned above. Immobilization by rapid freezing into amorphous, vitreous ice is much faster (microseconds) than with chemical fixatives. All of the commonly used chemical fixatives induce leakage of intracellular ions and small solutes long before cross-linking is achieved (75; Kellenberger and Kuhn, submitted), with the possibility of consequential structural alterations. Cryosections of the frozen material can be produced and viewed while still in the frozen state, using an electron microscope equipped with a cryostage. The production of cryosections, although improved substantially during the last years, is still not as routinely achievable, with consistently high quality and required thinness, as with resin embedded material. By freeze-substitution, the ice of the cryofixed material can easily be replaced by the resin. Immobilization is maintained during this process by a sort of physical cross-linking (briefly described in reference 65 and by B. Bohrmann and E. Kellenberger, in preparation), so that in the organic fluids involved in substitution and in the uncured resin, biological macromolecules do not precipitate as the usual aggregates seen with non-cross-linked, not gelled, material.

Immunocytochemistry allows the identification of antigens by reaction on thin sections, bypassing the problem of the lack of penetration of antibodies into microbial cells. Antibodies that react with antigens accessible on the surface of the section are easily observable by an additional reaction with gold coated with protein A which binds to the Fc domain of the antibody. These techniques, and their variations, are described in detail in several edited volumes (48). Recently it was shown that the gold label can be replaced by a stain obtained by reacting the specific antibody with an antibody against it. This protein complex, on the surface of resin sections, can afterwards be amplified by heavy metal stain (13) as illustrated in Fig 1C for the DNA plasms of bacterial nucleoids (14). For reasons that are not yet understood, this immunostain is in most cases more sensitive than the gold label.

The commonly used negative stain is still the method of choice for observing extracellular structures, although sections might also be required for capsules (7, 26). The well-established freeze-fracture technique is best suited for the study of extracellular and intracellular surfaces.

 

Studies of Isolated Substructures

By integrating various complementary approaches, spectacular progress has been made in the last two decades with methods of structural biology that allow determination of structure-function relationships. X-ray diffraction methods allow construction of three-dimensional models at atomic resolution, provided that the amino acid sequence of the protein is available and that the protein can be crystallized into a three-dimensional crystal. The first is easily obtained through the base sequence of the corresponding gene; the latter is more problematic. Not every protein yields a three-dimensional crystal with the required high order of the subunits and with a successful solution to the phase problem, for example by isomorphic replacement. In addition, frequently only one of several possible conformational states of a protein is amenable to crystallization. The biological functions of most of the interesting proteins are conformationally controlled, and knowledge of the shifts between conformations is thus of utmost importance. In this respect membrane proteins are particularly interesting. They transmit signals, modulate ion currents, convert light energy into chemical energy, and transport ions and even bulky solutes, all problems that involve conformational changes. Since membrane proteins are in general not stable when in detergent micelles and are thus difficult to crystallize, the structures of only some 10 of them have been determined, contrasting with those of more than 1,000 soluble proteins. Fortunately, membrane proteins have a propensity to form two-dimensional crystals in the presence of lipids. These crystals are accessible to studies by electron diffraction and imaging as well as with the modern method using AFM (46), of which an example is shown in Fig. 2. The detailed information obtainable with these methods on biological material is not at the atomic level and is only in the range of 1-to 2-nm resolution. When the restricted information on the surface topography obtainable by AFM is combined with an already known atomic structure of one of the several conformational states, the structure of a conformational variant(s) can be determined from two-dimensional crystals. By this combination of methods the fundamentally important problem of regulatory controls by conformational changes can now be approached (102).

This might be the place to explain why even with the imaging procedures crystals are needed. There are two reasons. First, X-rays and electrons damage individual proteins by radiochemical effects. Yet, to achieve a certain resolution, a minimum of radiation quanta is required per molecule. If several identically positioned proteins are available, as in a crystal, the amount of radiation per molecule can be reduced inversely proportional to

of N identical subunits seized. If a crystal is not available, other methods for using multiple images of individual molecules have recently proven to be applicable. The second reason is that proteins are not rigid "steel balls" but elastic, deformable structures which get easily distorted by the physical forces involved in specimen preparation or even imaging. The latter can be nicely observed in the AFM, when the stylus, a mechanical probe scanned over the specimen, is too near the protein (87). It is easy to understand that, also in this imaging mode, it is important to be able to average over "redundant" subunits in a crystal.

The scanning tunnel microscope, which is similar to AFM, gives, in addition to the surface structure, information on the atomic composition of a sample. This has been used with enormous success in solid-state physics. For biological material this has been limited by the very low conductivity of biomacromolecules and the current requirement of the scanning tunnel microscope to operate in air or in vacuum (36).

For atomic structure analysis, nuclear magnetic resonance (124) is potentially very interesting, because it works on proteins in solution. For the time being, it is limited to small proteins, below approximately 30 kDa.

 

HISTORICAL REMARKS

It is interesting to follow how newly available instruments and methods have influenced the direction taken by subsequent scientific investigations. Besides the naked eye, the light microscope has been, for several centuries, the main instrument of all biologists. Small wonder that morphology and comparative anatomy were so completely dominant until early in this century. Necessarily, morphology was purely descriptive until studied in parallel with physiology.

Impelled by medicine, chemistry was introduced into biological research through the bias of physiological chemistry, which first concerned itself nearly exclusively with the environment of cells (the blood!). Out of this, biochemistry arose early in this century, leading to the study of enzymes inside cells. It took longer for macromolecules to be investigated, initially only by a few outsiders, until the discipline that was later named molecular biology appeared in the forties. Almost simultaneously, a new flowering of morphology came into being with electron microscopy, which proved particularly helpful in the field of cytology. The number of newly discovered and rapidly christened organelles and features of the eukaryotic cell and tissues reached vertiginous heights. The functional roles of the most important ones were discovered, although, when consulting the literature and lecturing to students, one is aware that many more of these structures still await a functional role. This situation is akin to that of comparative anatomy and taxonomy in the world of old!

Molecular biology, on the other hand, has brought a fundamentally new stimulus into biochemical and biological research via its subdiscipline, molecular genetics. Using the revolutionary new concept of high specificity through a (genetic) code based on a linear sequence of only four different signals (for references, see reference 66), a field of discoveries was opened, leading to genetic engineering and the subsequent sequencing of genes. As everybody knows, these techniques presently dominate biological, biochemical, and biomedical research.

The term "molecular biology" became more and more restricted to mean genetic engineering or in vitro recombination of DNA. The part of molecular biology concerned with protein structure had to change its name to "structural biology." Amino acid sequences have become essential for constructing three-dimensional images of proteins in connection with data furnished from diffraction procedures with X-rays and electrons and Fourier transforms of images obtained by electron microscopy and AFM. Helped enormously by computer technology, information on three-dimensional protein structure has become fundamental for understanding life processes.

In an interesting study, West (121) elaborated on the "visual thinking" that reemerged after an eclipse of nearly half a century, during which it was replaced by an abstract sort of algebraic reasoning. It was customary on all levels of teaching physics to insist that "anything visually, i.e., also intuitively, understandable is always wrong." This widespread misconception was not shared by many outstanding scientists. Adequate computer programs have apparently brought back this visual thinking, introducing a new area of scientific endeavor.

ACKNOWLEDGMENTS

I am particularly indebted to Moselio Schaechter, for his comments and extensive editorial work; without them the paper would not be readable. I am grateful for discussions, comments, and references from Bernhard Witholt, Rosetta N. Reusch on the alkanoates, Andreas Engel on AFM and crystallographic methods, Gottfried Schatz on protein translocation in mitochochondria, and Jacques Dubochet on the supercoiling and liquid crystals of DNA. Last but not least, I thank Hedi Frefel and Marlies Zoller of the Biocenter in Basel for the fine photographic work in preparing the figures, and Margrit Jggi for the drawing of Fig. 3.