Chapter 13

Ribosomes

HARRY F. NOLLER and MASAYASU NOMURA

INTRODUCTION

Ribosomes are large ribonucleoprotein particles that are responsible for translation of the genetic code. Their molecular architecture is vast and complex, as is the system of pathways leading to their assembly. During the past 30 years, a great deal has been learned about these particles, and this chapter is an attempt to summarize some highlights of the present state of our knowledge. As has often been the case in molecular biology, Escherichia coli serves as the reference organism for ribosome research. Any discussion of E. coli ribosomes will thus be virtually representative of the state of ribosome research in general. By contrast, comparatively few studies have utilized Salmonella ribosomes, which, however, are likely to resemble closely those of E. coli. In some cases, particularly in the structure area, ribosomes from systems other than E. coli have yielded the most information, and so some results from those studies are also discussed.

The first section reviews the properties of the individual macromolecular components of ribosomes: the ribosomal proteins (r-proteins) and rRNA. Ribosomal structure is discussed insofar as the folding of these individual macromolecules is understood. In the second section, ribosome assembly is discussed, as it has been studied by means of in vitro approaches. RNA-protein interactions in the ribosome and the process of in vitro assembly of functional ribosomes are summarized. In vivo assembly is discussed only in connection with information obtained from the in vitro assembly reaction. The present state of our understanding of ribosomal architecture is summarized in the third section. This includes the overall morphology of E. coli ribosomes and their subunits, as well as the locations of specific protein and RNA features in the structure. The final section brings together morphology, rRNA secondary structure, and rRNA–r-protein assembly, to give a sense of how information obtained from a number of quite different approaches has led to a model for the structure of the 30S ribosomal subunit that begins to account for some of the functional properties of the ribosome.

As a starting point for more extensive descriptions of these topics, the reader is referred to several classic volumes on ribosome structure and function (18, 42, 50, 88).

MACROMOLECULAR COMPONENTS OF E. COLI RIBOSOMES

Ribosomal Proteins

Identification, Isolation, and Stoichiometry.

The structural complexity of ribosomes first became evident in early studies on r-proteins. Waller (149) found dozens of distinct protein bands by using starch gel electrophoresis on E. coli 30S and 50S subunit proteins. The notion that these represented individual, distinct protein species was met with some skepticism because of the unprecedented structural complexity that was implied. Biochemical characterization of purified r-proteins finally established that the observed bands, for the most part, indeed represented individual protein species with distinct sizes and N-terminal and amino acid sequences (141). It is now generally accepted that the E. coli ribosome contains about 52 different proteins; 21 are found in the 30S subunit, and about 31 are found in the 50S subunit (156).

Isolation of r-proteins has usually been accomplished by cation-exchange chromatography in 6 M urea followed by gel filtration (reviewed in reference 155). Because r-proteins had been isolated independently in several laboratories, a number of competing nomenclature systems added to the complexity of these studies. By consensus, the presently used nomenclature system based on a widely used two-dimensional gel electrophoresis method (54) was adopted (159). In this system, small (30S) subunit proteins are called S1, S2, S3, . . . , S21, and large (50S) subunit proteins are called L1, L2, . . . , L34. This system provided a seemingly unambiguous method for identification of E. coli r-proteins. Subsequently, it was discovered that L8 is in fact an aggregate of proteins L7/L12 and L10 (100) and that L26 is identical to S20 (157). Furthermore, proteins L7 and L12 are identical, except for the N-terminal acetylation of L7 (136).

Once the proteins had been isolated and characterized, another important question arose. Do all ribosomes in a population have the same protein composition, or are there different classes of ribosomes with different structures and possibly different functional roles? Protein stoichiometries are difficult to measure, because of the ease with which certain r-proteins become dissociated from the particles during isolation, particularly during high-salt washing steps. It is generally believed that all of the proteins and RNA molecules are present in 1:1 stoichiometry with each other, except for L7/L12, which is probably represented at the level of four copies per ribosome (43, 131). This implies that all functional ribosomes have essentially equivalent structures and that "specialized ribosomes" probably do not exist in E. coli cells. This view is supported by the results of in vitro reconstitution studies (45). Proteins which bind independently to rRNA reach saturation at stoichiometries of 1:1 with the RNA (167). Single-component omission experiments show that omission of an r-protein appears to affect, in many instances, the assembly or activity of all particles in a population, rather than a specific fraction of them (45, 87).

Properties of r-Proteins.

All of the E. coli r-proteins have been sequenced (156), either by direct protein sequencing or by DNA sequencing of their genes, or both. Table 1 lists their calculated molecular weights, which average 14,800. Besides their generally small size, r-proteins tend to be quite basic, typically containing about 20% Lys + Arg, as might be expected for proteins that exist in a stable complex with RNA. Notable exceptions include the rather acidic proteins S1 (molecular weight, 61,000) and L7/L12 (molecular weight, 12,000), which are both strongly implicated in functional roles in the translation process.

Table 1E. coli r-proteinsa

With the sequences of all 52 r-proteins in hand, the structural complexity of the ribosome is even more apparent. Computer-assisted searches for sequence homology between different r-proteins have turned up surprisingly little (157, 158). Although it is difficult to rule out possible cryptic sequence homologies between proteins, this result suggests that each r-protein originated independently or at least diverged from the others at a very early evolutionary stage. Such a view is consistent with proposals that the ribosome originated as a structure that was composed mainly, if not completely, of RNA (81, 84). It is then easy to imagine the stepwise acquisition of individual proteins during evolution, gradually refining the process of assembly and translation. If the ribosome evolved in this way, we would not necessarily expect a priori to find any sequence homology between different r-proteins.

Concerning the overall shapes of the r-proteins, some studies have reported evidence supporting highly elongated structures, while other evidence seems to point to compact and globular shapes (reviewed in reference 65). This is a difficult question to resolve, since the relevant structure is that of an r-protein in its functional shape, i.e., in its complexed form within the ribosome. Thus, even the emerging X-ray crystal structures of r-proteins (see below) may be subject to the caveat that they may change when complexed in the ribosome. Short of a high-resolution crystal structure of the ribosome itself, neutron-scattering studies have provided one of the few direct approaches to this problem. Radii of gyration have been measured in situ by Ramakrishnan et al. (109) for several 30S r-proteins. Although the uncertainty of the measurements significantly limits interpretation of these data, most radii of gyration are consistent with globular shapes, except for proteins S1, S4, and S8, which appear to be significantly elongated.

Three-Dimensional Structures of Isolated r-Proteins.

As discussed below, progress toward an X-ray crystallographic structure for the intact ribosome has been slow because of the many technical challenges posed by its large size and structural complexity. It has also proven difficult to obtain good crystals for the individual proteins from E. coli ribosomes, although those from thermophilic bacteria appear to be more amenable to crystallization.

Thus far, X-ray crystallographic structures have been determined for six ribosomal proteins, S5 (110), S6 (66), L6 (33), L7/L12 (63, 64), L9 (51), and L30 (63, 153). In addition, the structure of protein S17 has been determined by nuclear magnetic resonance spectroscopy (32). Although many of these structures were determined for proteins from Thermus or Bacillus ribosomes, the conservation of primary sequence between these proteins and their counterparts from E. coli is indicative that their three-dimensional structures will turn out to be similar. Although several of the structures are compact and globular, those of L7/L12 and L9 are highly elongated structures in which two essentially globular domains are connected via a slim tether.

rRNA

E. coli ribosomes are composed of 38% protein and 62% RNA. It is probably an important clue to the evolution and mechanism of action of ribosomes that, unlike other cellular polymerases, they contain RNA as a major structural component. There is now extensive evidence to support the view that rRNA is a functional rather than a merely structural component of ribosomes (80).

Three species of rRNA are found as integral structural components of the ribosomes of E. coli and other bacteria. The 30S subunit contains 16S rRNA, whereas 5S and 23S rRNAs are found in the 50S subunit. Primary and secondary structures are known for all three molecules. Many constraints on their tertiary folding are now known, and many of the sites of interaction with ribosomal proteins and functional ligands have been identified. These findings are summarized below.

Primary Structures

. There are seven independent rRNA transcriptional units in E. coli (see chapter 90 of this volume). Although some differences in primary structure are known to exist between the genes in different transcriptional units, the sequences are virtually (>99%) identical. The complete DNA sequence was determined for the rrnB transcriptional unit, which has served as the basis for extensive genetic studies on rRNA, in the form of the multicopy plasmid pKK3535 (12, 14).

The smaller rRNA, 5S RNA, was the first RNA molecule to be sequenced by isotopic methods (15). It is composed of 120 nucleotides and contains no posttranscriptionally modified nucleotides.

Over a decade elapsed before a complete sequence was obtained for 16S rRNA by DNA sequencing of the cloned gene (13) and by direct RNA sequencing (17). The 16S rRNA chain contains 1,542 nucleotides, 10 of which (positions 527, 966, 967, 1207, 1402, 1407, 1498, 1516, 1517, and 1518) are methylated and 1 of which is a pseudouridine (2). Analysis of phylogenetic conservation of sequence in 16S rRNA has provided insight into the relative biological significance of different regions of this molecule (40, 160). The occurrence of posttranscriptional modifications in regions of high sequence conservation suggests that they are functionally important. The fact that they show characteristic differences between the three major lines of descent would imply that they may be involved in fine tuning of crucial structural features (84). In some cases, methylation or nonmethylation is known to confer resistance to certain antibiotics.

The 23S rRNA consists of 2,904 nucleotides in E. coli (11) and, like 16S rRNA, contains numerous posttranscriptionally modified nucleotides, including pseudouridines as well as methylations. Many, but not all, of the modified positions have been placed in the sequence (3). Little or no significant sequence homology has been detected between the three different rRNAs; nor has any been found between different regions of the individual rRNAs, beyond that which would be anticipated to occur by chance, again arguing for independent evolutionary origins for the different rRNAs (81). The absence of significant structural symmetry in the ribosome is again apparent for rRNA, as it is for the r-proteins. As discussed below, this appears to extend to secondary and higher-order levels of structural organization.

Secondary Structures.

In the decade following the publication of the 5S rRNA primary structure, dozens of often quite different secondary-structure models were proposed for this molecule, in spite of the availability of considerable relevant biochemical and physical evidence and free energy prediction rules for estimation of RNA helix stabilities. Resolution of the dilemma was a result of the realization that, as in the case of tRNA, 5S rRNAs from different species are likely to have very similar structures, in spite of having significant sequence differences. Fox and Woese (27) used comparative sequence analysis explicitly for the first time to deduce the secondary structure of 5S rRNA (Fig. 1). Its four main helical elements are typical of the kinds of local features seen in the large rRNAs, including single- and double-base bulges, compound helices connected by internal loops, and multibranch loops.

Fig. 15S rRNA, the first RNA molecule whose secondary structure was derived explicitly by comparative sequence analysis (27).

Source:

Secondary structures for E. coli 16S and 23S rRNAs are presented in Fig. 2 and 3 (37, 38, 39, 41, 82, 84, 160). Evidence supporting the existence of each individual helix comes from several approaches, but the structures themselves were established by comparative sequence analysis. Significantly, every one of the helices that was considered proven by comparative analysis in the first published versions of the 16S and 23S rRNA structures (82, 84) has survived to the present versions, having been tested by more than 2,000 additional complete 16S-like rRNA sequences, more than 100 additional 23S-like rRNA sequences, and an abundance of new experimental information. Few possibilities for additional base pairing remain.

Fig. 216S rRNA (13, 40, 84). This RNA is divided into three major domains and one minor domain (see the text).

Source:

Fig. 323S rRNA (11, 38, 82). This RNA is divided into six major domains.

Source:

The 16S rRNA secondary structures is subdivided into three major structural domains and one minor domain by three sets of long-range base-paired interactions (Fig. 2). The 5' domain (residues ca. 26 to 557) is defined by the 27–37/547–556 helix; the central domain (residues ca. 564 to 912) is defined by the 567–569/881–883 helix; and the 3' major (residues ca. 926 to 1391) and 3' minor (residues ca. 1392 to 1542) domains are defined by the 923–933/1384–1393 helix. The extent to which these domains, defined here only in terms of secondary structure, correspond to true structural domains is relevant to our understanding of ribosomal architecture. Fragments of 16S rRNA, corresponding closely to the three major domains, were isolated in conjunction with early studies on ribosomal protein-binding sites, as discussed below. The ability of certain r-proteins to remain bound to these fragments or even to rebind to RNA fragments from which proteins have been removed (25, 76, 164, 168) supports the suggestion that these domains are true structural entities. Assembly of individual RNA domains with their cognate proteins provides further evidence in support of RNA structural domains (115, 151). This is not to imply that there may not be significant structural interactions between domains, however.

Within each domain, the structure is organized into series of simple and compound helices, separated by various multibranch loops, interior loops, and bulges. Somewhat unexpected was the absence of long, regular helices in rRNA. Instead, the structure is formed by joining of many short helices, the junctions of which tend to create a variety of structural irregularities. The reason for this kind of architecture has been rationalized in two ways (84). (i) Multiple small helices allow a much more complex three-dimensional structure, which could approach that of a globular protein. (ii) Long, stable helices could result in kinetically trapping the RNA conformation in local energy minima, interfering with progress toward assembly into the final correct structure.

Figure 3 shows the secondary structure of E. coli 23S rRNA (38, 39, 82). As in the 16S rRNA, long-range base-paired interactions partition the chain into readily identifiable structural domains. There are six major domains in 23S rRNA, defined by their respective long-range interactions: domain I (16–25/515–524), domain II (579–585/1255–1261), domain III (1295–1298/1642–1645), domain IV (1656–1664/1997–2004), domain V (2043–2054/ 2615–2625, and domain VI (2630–2637/2781–2788). These domains project from a central loop created by pairing of the 5' and 3' ends of the molecule.

Several pseudoknot interactions have been identified in rRNA by comparative sequence analysis (39, 40, 41, 62). At least two have been found in 16S rRNA, involving pairing between positions 505 to 507 and positions 524 to 526 in the 5' domain and between positions 570 to 571 and positions 865 to 866 in the central domain, respectively. The 505–507/524–526 pseudoknot involves one of the most highly conserved structures in rRNA, the 530 loop. This interaction has been confirmed directly by genetic approaches (104). Genetic alteration of the 530 loop pseudoknot produces functional perturbations in the resulting mutant ribosomes, consistent with the role of this structure in translational accuracy. As many as six pseudoknot interactions have been identified in 23S rRNA (39, 41, 62). One of these involves pairing between positions 413 to 416 in domain I and positions 2407 to 2410 in domain V, a loop-loop interaction that provides a strong constraint on the three-dimensional folding of these two regions of 23S rRNA. Comparative analysis has also identified tertiary base-base interactions in rRNA. In all likelihood, only a fraction of the total tertiary interactions have been found so far. However, several clear instances of noncanonical base pairing, such as A-G, A-C, C-C, and U-U pairing, together with the pseudoknot interactions mentioned above, indicate that the higher-order structure of rRNA is rich with complexity and contains RNA structural features that have yet to be defined structurally (41).

Experimental Evidence for Higher-Order Structure.

Comparative analysis provides convincing evidence for the existence of a given helical element and even individual base pairs in vivo. However, ribosomes may be dynamic structures, in which interconversion of different conformers of rRNA occurs during ribosome assembly and in the course of protein biosynthesis. Thus, experimental approaches that provide information bearing on rRNA conformation not only help to provide tests of secondary structure but also may indicate whether all of the helical elements defined by the comparative approach actually exist at the same time. Currently available evidence indicates that all of the phylogenetically identified helices do, in fact, coexist in the ribosome, and convincing examples of mutually exclusive helices have not been identified so far, except for competition between the 5' helix of 16S rRNA and a putative precursor helix, which have a strand in common (23). The latter helical switch would presumably occur only at some step during maturation of 30S ribosomal subunits.

A wide variety of experimental methods have been used to study rRNA conformation. The most commonly used approaches include (i) the use of single- and double-strand-specific chemical and enzymatic probes, including cross-linking agents (see, e.g., references 8, 9, 10, 73, 143, 146, and 161); (ii) two-dimensional gel analysis, in which RNA fragments from partial nuclease digests, associated by base pairing in the first dimension, are resolved under denaturing conditions in the second dimension (56, 113); (iii) oligonucleotide probes, which presumably bind to free, unpaired regions of the rRNA (see, e.g., references 49 and 68); and (iv) nuclear magnetic resonance (see, e.g., references 134, 144, 152, and 154).

When using single-strand-specific chemical probes to test secondary-structure models, base-paired nucleotides should not show reactivity toward the probes. The converse does not necessarily hold, however; nucleotides shown as unpaired in the model may be resistant to attack by chemical probes because of tertiary interactions, tight binding of magnesium or other ions, or (as in the case of diethylpyrocarbonate) stacking interactions. Current methodology permits rapid, systematic chemical probing of bases at virtually every position of 16S and 23S rRNA. In one approach, ribosomes or rRNA is probed with a set of single-strand-specific chemical probes that attack the Watson-Crick pairing positions of all four RNA bases (73). Sites of chemical modification are then identified by primer extension with reverse transcriptase, which pauses or stops at modified bases. The results of probing naked 16S rRNA are in almost complete accord with the structure depicted in Fig. 2 and give further support to the power and accuracy of the comparative approach (73). In the remaining cases, discrepancies between the probe data and the deduced secondary structure are resolved when 16S rRNA is probed in 30S ribosomal subunits. Thus, certain aspects of the biologically relevant RNA conformation (which is identified by the comparative method) appear to depend on interactions between 16S rRNA and the 30S r-proteins.

Finally, certain chemical probing results appear to reflect the existence of tertiary or quaternary interactions and are not easily explained by consideration of secondary interactions alone. In some cases, the results support the occurrence of higher-order interactions predicted from comparative sequence analysis.

Szewczak et al. (134) have deduced the three-dimensional structure of the α-sarcin loop of 23S rRNA, using high-resolution nuclear magnetic resonance spectroscopy methods. This conserved 15-nucleotide "loop" is actually highly structured, as suggested by the lack of chemical reactivity of most of the bases in the loop toward chemical probes (72). It is characterized by noncanonical base-base interactions as well as base-phosphate interactions. This structure hints at the structural complexities that are yet to be elucidated in rRNA.

ASSEMBLY OF RIBOSOMES

In Vitro Reconstitution of Ribosomes

Even though the structure of the ribosome is very complex, functionally active ribosomes can be reconstituted from purified molecular components. The successful reconstitution of ribosomes was first achieved with E. coli 30S subunits by incubating 16S rRNA and a mixture of 30S r-proteins (138). Some important factors for the success were the use of moderately high ionic strength (the optimum being 0.37), presumably to prevent nonspecific ionic interactions between RNA and basic r-proteins and yet allowing specific RNA-protein interactions, high concentrations of Mg²⁺ ions, and moderately high incubation temperatures (the optimum being about 40°C) (139). Reconstitution of 50S subunits was achieved first with 50S subunits from a thermophilic bacterium, Bacillus stearothermophilus, under essentially the same conditions as those used for E. coli 30S subunits except for the use of higher temperatures, usually 60°C (85), and subsequently with E. coli 50S subunits by using a two-step incubation method (24, 79). The successful reconstitution of ribosomes demonstrates that all of the information needed for correct assembly of the complex ribosomal particles is contained in the structure of its molecular components. In this regard, it should be noted that the reconstitution of 30S subunits has been demonstrated with denatured 16S rRNA molecules (4, 133). Reconstitution systems have been used extensively to study the structure of ribosomes, the functional roles of individual molecular components, and the mechanism of assembly of ribosomes in vitro. Most of the studies on ribosome reconstitution have been reviewed previously (for 30S ribosome reconstitution, see reference 86; for the 50S ribosome reconstitution, see reference 78).

Assembly Mapping of r-Proteins

In the case of 30S subunits, the in vitro assembly reaction takes place with 16S rRNA and a mixture of 21 purified r-proteins under a fixed optimum condition. It was first shown that the assembly reaction is cooperative and sequential (89, 139). The cooperative nature of the assembly reaction and the sequence of events during the reaction were subsequently studied with purified proteins. Since the reaction proceeds to completion to produce the final product, the 30S subunit, which has all the r-proteins stably bound to 16S rRNA, a simple logic was used to construct an "assembly map"; if protein B will not bind well to RNA (or an RNA-protein complex) until protein A does, then protein B must follow protein A in the assembly reaction. It was found that under reconstitution conditions, only six (see the legend to Fig. 4 and below) of the 30S r-proteins, termed primary binding proteins, bind individually to 16S rRNA to make a stable complex. Certain other proteins (secondary binding proteins) bind only after some of the first six proteins are bound. Binding of the remaining (tertiary binding) proteins requires the presence of several proteins in addition to some or all of the initial binding proteins. In this way, the "sequence" of addition of proteins to 16S rRNA has been analyzed, and a map has been constructed (Fig. 4) (44, 69).

Fig. 4Assembly map for the 30S ribosomal subunit (44, 69). Arrows between proteins indicate the facilitating effect of one protein on the binding of another; a thick arrow indicates a major facilitating effect. The thick arrows from 16S RNA to S4, S8, S15, S17, and S20 indicate that each of these proteins binds directly to 16S RNA in the absence of other r-proteins. The thin arrow pointing toward S7 indicates that S7 binds directly to 16S RNA, but its binding is weaker and is stimulated by other proteins as indicated in the map. However, later studies showed that S7 binding to 16S RNA is sufficiently strong, so that the S7 binding site on 16S RNA can be studied without other r-proteins. Proteins above the dotted line are those either required for the formation of RI* particles or found in the isolated 21S RI particles. (The actual isolated 21S RI particles contained only traces of S5, S19, and S12; it was suggested that these proteins are important for the temperature-dependent rate-limiting reaction but bind very weakly to RI particles at low temperatures (46).

Source:

The assembly map shows interdependence of r-proteins in the assembly reaction and does not necessarily show the temporal sequence of the assembly. Nevertheless, studies on intermediate particles in the in vitro assembly reaction gave results which are consistent with the assembly map (46). More recently, kinetic studies of the in vitro assembly reaction show a close correlation with the order implied in the assembly map. By using chemical probing of 16S rRNA, it could be seen in certain instances that assembly proceeds from primary to secondary to tertiary binding proteins (103). By carrying out assembly at 30°C, at which reconstitution takes about 2 h to go to completion, intermediate stages of the process could be monitored. Certain regions of the RNA were protected almost immediately, while others were protected only later during assembly. Some bases became more reactive during early stages and later became unreactive, indicating formation of transient assembly intermediates. These studies also revealed an aspect of assembly that was not anticipated by the assembly map. Assembly was found to proceed sequentially with a 5'-to-3' polarity, along the 16S rRNA chain. This result implies that the 16S rRNA is somehow designed to fold in a 5'-to-3' direction, presumably because of coupling of ribosome assembly with transcription.

Interdependence of binding of two proteins observed during the construction of the assembly map does not necessarily indicate a physical proximity of the two proteins. Protein A may help the binding of protein B by directly interacting with B or may help it indirectly by creating a correct binding site for protein B through some conformational alteration. Specific examples of possible mechanisms of this kind have been identified in studies of protein-RNA interactions (102, 123, 124, 125). Yet, as described in a later section, a good correlation has been observed between assembly interdependence and physical proximity in most cases (as determined directly by physical measurement).

An assembly map for E. coli 50S r-proteins was constructed by Nierhaus and his coworkers by the same approach (48, 111, 112, 114) and is shown in Fig. 5. It is evident that construction of a 50S assembly map is inherently more complex and difficult. First, many more components are involved in the assembly. Second, the standard procedure for reconstitution of functionally active 50S particles involves two steps with different reaction conditions, and hence, the assumption that the interactions observed in the first incubation conditions are all pertinent to a "genuine" assembly reaction may have to be taken somewhat more cautiously than in the case of constructing a map for 30S assembly. In connection with this second problem, it should be noted that Spierer et al. (120, 121) used different incubation conditions, the conditions used for the reconstruction of 30S subunits, and constructed a partial assembly map which showed some discrepancies from, as well as agreements with, the results obtained by Röhl and Nierhaus (111). The discrepancies may be at least in part a result of the differences in the conditions used for the studies of assembly interactions by the two groups. It is possible that the interactions observed by Spierer et al. but not shown in Fig. 5 are still "genuine" ones and take place during the second step of the standard reconstitution system (and in the finished ribosome structure).

Fig. 5Assembly map for the 50S ribosomal subunit (111, 112, 114). The three main fragments of 23S RNA (13S, 8S, and 12S RNAs) are indicated, and proteins are arranged according to their binding regions on 23S RNA. Proteins above the dashed line are those important for the conformational change RI50 (I) ? RI*50 (I). L5, L18, and L15, circled by the dotted line, are the proteins important for mediating the binding of 5S RNA to 23S RNA.

Source: Reprinted from reference 48 with permission.

Intermediate Particles in Assembly

The assembly of 30S ribosomal subunits involves 22 molecular reactants. However, the overall reaction follows simple first-order reaction kinetics, indicating that the rate-limiting step is a unimolecular reaction (139). In fact, by carrying out the reaction at suboptimal temperatures, e.g., at 30°C, and analyzing the reaction mixture by sedimentation techniques, one can observe the appearance of intermediate particles which sediment at about 21S (called reconstitution intermediate [RI] particles) and which are slowly converted to the final products, the reconstituted 30S particles (46, 89). The presumed RI particles were isolated from reaction mixtures kept at low temperatures (10°C or below) as functionally inactive 21S particles which contain some but not all of the 30S r-proteins. When these particles are heated (40°C) in the absence of free r-proteins, new particles (RI* particles) are formed which sediment at about 26S and are now capable of binding the remaining r-proteins to form active 30S subunits at lower temperatures. Thus, the following reaction sequence can be recognized:

The step from RI to RI* is the major rate-determining step that requires the high activation energy and represents a unimolecular rearrangement of the intermediate particle (46, 139). Studies on the physical properties of the RI and RI* particles demonstrate that the initial binding of the proteins to 16S rRNA at low temperatures does not markedly affect the overall shape of the RNA molecule but that the step from RI to RI*, which takes place at higher temperatures, involves significant folding of RNA, resulting in a structure considerably more compact than that of the naked 16S rRNA or the RI particles (135).

The protein composition of the RI particles was inferred from the isolated 21S particles and other experiments that determined the identity of the proteins required for the essential conformational change in the RI → RI* reaction. These are the proteins appearing above the dotted line in the assembly map shown in Fig. 4 (see the comments in the legend). Although some of the interactions shown in the assembly map are really essential (e.g., omission of protein S7 in the reconstitution mixtures leads to formation of particles which are missing proteins S19, S14, and S10 completely and probably S9, S3, and S2 at least partially [69]), some other interactions might be dispensed with, thanks to the extensive cooperative interactions among ribosomal components. For example, even though S16 plays an important role in the assembly reaction, as evidenced in the map, 30S-like particles with almost full activity are slowly formed even in the absence of S16 (47). Another example is that although S6 plays an important role in the binding of S18 as demonstrated in the process of construction of the assembly map (69), omission of S6 from the reconstitution mixture does not have any significant effects on assembly kinetics or the function of the assembled particles in various functional assays (87). It appears that many other cooperative interactions enable binding of S18 and other "downstream" proteins, such as S11 and S21.

The sequence of events during in vitro E. coli 50S reconstitution was also studied in a way similar to that used for 30S reconstitution, and the following reaction scheme was proposed (24, 119):

(Essentially the same reaction scheme was proposed earlier for the reconstitution of B. stearothermophilus 50S ribosomal subunits [26].) Proteins that participate in the formation of RI₅₀ (I) and those participating in the later reaction were classified by Nierhaus and coworkers (111, 112, 114) as indicated in the assembly map shown in Fig. 5. As for S6 and S16 in the case of the 30S assembly reaction (see above), proteins L24 (122) and L20 (90) were concluded to be essential for the assembly reaction but not for the activity of the assembled 50S subunits in protein synthesis.

RNA-PROTEIN INTERACTIONS

Our understanding of the mechanism of RNA-protein recognition is in a stage of rapid development. Although many DNA-protein interaction mechanisms are now well understood, there is reason to believe that recognition of RNA differs in several fundamental respects. DNA, with rare exceptions, has a simple helical structure; recognition of specific sites on a DNA molecule by a protein thus relies by necessity to a large degree on sequence recognition and possibly, to a lesser extent, on subtle, sequence-dependent variations in helical geometry. In contrast, RNA structure is complex, containing not only helical elements but also such intricacies as hairpin loops, bulges, interior loops, multibranch loops, and pseudoknots, and has the ability to achieve compact, globular structures by means of rather complex tertiary interactions, exemplified by tRNA (55). Furthermore, RNA helices are believed to be confined largely to the A or A' geometry, in which the major groove is deep and narrow, interfering with penetration by a protein α-helix, the means by which DNA sequences appear to be "read" by their cognate proteins in several cases (97). However, recent studies indicate that irregularities in RNA structure, such as bulged bases and internal loops, may facilitate the accessibility of bases in the major groove of RNA helices (150). It may thus be significant that such irregularities are found throughout the rRNA structure, and so part of their function may be to permit major groove recognition by ribosomal proteins.

Ribosomes provide a wealth of model systems for the study of specific RNA-protein interactions. Such studies not only provide important information concerning the mechanism of ribosome assembly but also contribute to an understanding of the general problem of RNA-protein recognition. Several protein-binding sites have been localized on rRNA, and it is worthwhile to review briefly what has been learned from these studies.

Discovery of the in vitro reconstitution method (138) stimulated the first studies on the mechanism of binding of r-proteins to rRNA. The fact that fully functional particles were obtained argued that any intermolecular interactions observed with this system were likely to be authentic. Studies on assembly performed with purified r-proteins revealed that only a subset of the 30S proteins (the primary binding proteins) have the capacity to bind independently to 16S rRNA in the absence of others. Criteria for the specificity of binding are as follows: (i) a given protein should bind to only one species of RNA (e.g., 16S but not 23S rRNA); (ii) in protein excess, the molar ratio of protein to RNA should plateau at a stoichiometric ratio (e.g., 1:1); (iii) interaction of one protein with the RNA should not hinder the binding of another; and (iv) a complexed protein should be incorporated into the mature ribosomal structure upon addition of the remaining proteins without prior dissociation from the RNA (166).

By using these criteria, there is general agreement that proteins S4, S7, S8, S15, S17, and S20 bind directly and independently to 16S rRNA; that L1, L2, L3, L4, (L12)4-L10, L11, L15, L20, L23, and L24 bind to 23S rRNA; and that L5, L18, and L25 bind to 5S rRNA. Localization of the RNA-binding sites for these proteins has been pursued by several approaches. (i) The classical "bind-and-chew" method utilized gentle nuclease digestion of protein-RNA complexes followed by isolation of ribonucleoprotein particles containing subfragments of the RNA (reviewed in reference 167). In several cases, the ability of the protein to rebind to the purified RNA fragment confirmed that at least part of its RNA-binding site was contained in the fragment. This approach proved to be quite generally acceptable and useful in early studies. However, a number of potential shortcomings were recognized. In some cases, the protected RNA fragment was obtained even in the absence of protein, suggesting that the structure of the RNA itself, rather than its interaction with protein, stabilized the protected fragment (29). Conversely, features of RNA-binding sites that are inherently unstable may not survive nuclease treatment and so cannot be identified by this approach. (ii) Protection of RNA from chemical or enzymatic probes by bound protein ("footprinting") has been successfully exploited, particularly in studies of 16S rRNA (125). (iii) "Damage/selection" experiments have the potential to identify RNA sites whose chemical modification interferes with protein binding (98). (iv) Chemical or UV cross-linking of bound proteins to RNA indicates sites that are proximal to the protein; in some cases, the cross-linked site is not part of the binding site proper (8, 10, 162). (v) Electron microscopy of protein-RNA complexes has provided low-resolution information about protein-RNA interaction sites (19). (vi) Site-directed mutagenesis of cloned rRNA genes in sequences corresponding to protein-binding regions in the RNA can identify structural and sequence requirements for protein recognition (116, 163). Mutagenesis approaches, however, have the inherent danger that the mutated RNA may fold incorrectly, so that the failure of a protein to bind may be caused indirectly by the effects of the mutation on the RNA structure itself.

Protein binding site regions in 16S rRNA obtained by approach (i) (the "bind-and-chew" method) correspond closely to domains or subdomains of the RNA structure. This observation argues for the existence of these domains, derived mainly from comparative sequence analysis, as true physical entities; their ability to rebind to their cognate proteins, in many cases, suggests that they possess some degree of structural autonomy in the ribosome. This has been borne out in experiments in which independent structural domains have been reconstituted from discrete domains of 16S rRNA in vitro, as described below (115, 151).

Which structural features of rRNA are actually recognized and/or bound by the proteins? It is doubtful that all 400 or so nucleotides of the S4-binding fragment are important for binding, since the mass of the RNA moiety outweighs that of S4 itself by more than a factor of 6. Chemical and nuclease footprinting experiments, in contrast, localize the region of interaction of S4 to a relatively compact region centered on the confluence of five helical elements around position 500 of the 5' domain of 16S rRNA (123, 127). Deletion experiments, on the other hand, have led to the proposal that S4 binding requires RNA elements outside the footprinted region and, conversely, that some of the footprinted elements are dispensable (116). More recent experiments (106) suggest the following partial resolution of this dilemma. When bound at low temperature, all of the bases protected by S4 are found within the region defined by the deletion experiments. After brief incubation at 40°C, the remaining bases become protected, indicating a heat-dependent conformational rearrangement of the S4-RNA complex. Since the binding reactions were carried out at low temperature in the deletion studies (116), it seems likely that only the initial, low-temperature complex was being tested. The apparent requirement for RNA sequences outside of the footprinted region might be explained by disruption of normal folding in some of the deleted RNAs.

This footprinting strategy has also been used to locate the sites of interaction of the rest of the 30S proteins on 16S rRNA (102, 107, 123, 124, 125, 132). This was done by stepwise in vitro assembly of the proteins and 16S rRNA, following the 30S assembly map (Fig. 4). At each step, the RNA in the complex was submitted to probing with a set of base-specific chemical probes, and the bases protected by assembly of each protein were localized (Fig. 6). It should be recognized that some of the protein-dependent protections could be due to indirect interactions, in which assembly of a given protein results in a conformational change in the RNA. Thus, some of the protected bases could be remote from the binding site of the protein. Close agreement between these results and those obtained by cross-linking (below) are reassuring, however.

Fig. 6Effects of assembly of individual 30S subunit r-proteins on the reactivities of bases in 16S rRNA (102, 107, 123, 124, 125, 127, 132). Protein-dependent protection from (?) or enhancement of (?) attack by chemical probes is indicated, as is protection or enhancement of nuclease attack (arrows and arrows with triangles, respectively). Positions where individual proteins have been cross-linked to 16S rRNA (see Fig. 7) are shown by large open arrows. The proteins are grouped approximately according to whether their main effects are on the 3? (top), central (middle), or 5? (bottom) domain of 16S rRNA. Note how assembly map proximities between proteins (Fig. 4) are often reflected in structural proximities of their effects on the RNA.

Source:

Several groups of investigators, most notably Brimacombe and coworkers, have used chemical cross-linking methods to probe the three-dimensional proximities between protein and RNA, as well as between different regions of the RNA, in the ribosome. Figure 7 summarizes the results obtained for the 30S subunit by the Brimacombe laboratory (1, 36, 57, 96, 128, 162, 169). Comparison with Fig. 6 shows that many of the same protein-RNA proximities can be inferred by using these two quite different approaches.

Fig. 7RNA-protein and RNA-RNA cross-links found for 16S rRNA in E. coli 30S ribosomal subunits (1, 36, 57, 96, 128, 162, 169). Arrows point to sites cross-linked by the indicated proteins. Dots joined by lines show positions in the 16S rRNA that are cross-linked to each other.

Source:

ARCHITECTURE OF RIBOSOMES AND THEIR SUBUNITS

General Morphology of Ribosomal Subunits

From the preceding discussion, it is clear that structural information concerning the individual molecular components, particularly the rRNAs, provides important clues to the overall structural organization of the E. coli ribosome. At the same time, physical studies have provided information on the overall shape of ribosomal subunits and the location of specific structural features therein. Our current perception of the shape of ribosomes derives mainly from electron microscopy (EM) and neutron-scattering studies.

Several groups have studied the shapes of ribosomal subunits by EM (5, 28, 93, 129, 145). Although the details of models proposed by the various groups of investigators differ because of differences in interpretation as well as methodology, they are typified for the most part by models for the 30S and 50S subunits similar to those described by Lake (58) (Fig. 8). Image reconstruction methods have also been used, in combination with EM, to produce three-dimensional computer-averaged structures for the ribosome and its subunits. One such reconstruction for the E. coli 70S ribosome, obtained by Frank and coworkers (28, 99), is shown in Fig. 9.

Fig. 8A three-dimensional model of the E. coli ribosome derived from electron microscopy (91). This model gives an asymmetric shape to the two subunits. The small subunit (left) includes a head, a body, and a platform. The large subunit (second from left) includes a central protuberance flanked by a ridge on one side and a stalk on the other. Two orientations of the model are shown. The length of a ribosome is about 250 Å.

Source:

Fig. 9Three-dimensional reconstruction of the E. coli ribosome, based on EM of single particles embedded in ice (28, 99). (Top) View from the 30S subunit side, similar to the lower right of Fig. 8. (Middle) View from the right-hand side. (Bottom) View from the back of the 50S subunit.

Source:

The 30S subunit (Fig. 8) is divided into a "body" containing ca. two-thirds of the mass of the particle and a "head" containing the other one-third of the mass. A "platform"-like projection extends from the body of the particle near the head-body junction, and a "cleft" is seen (or not seen, depending on the particular method of specimen preparation and/or interpretation) to be formed between the head and platform. Some workers identify a second platform-like feature on the 30S subunit, located on the side away from the main platform (28, 129). The largest dimension of the particle is about 240 to 250 Å (1 Å is 0.1 nm).

The 50S subunit has a more hemispherical or globular shape (Fig. 8) and contains three recognizable projections: the L7/L12 stalk on the right, the central protuberance in the middle, and the L1 ridge on the left (58, 83). There is generally better agreement between the different groups concerning the shape of the 50S particle. The largest dimension of the 50S subunit is again about 240 to 250 Å.

The relative orientation of the 30S and 50S subunits in the 70S particle has also been a matter of controversy (see the discussion in reference 108). Current consensus appears to favor the arrangement originally proposed by Lake (59) (Fig. 8). In this model, the platform side of the 30S subunit faces the concave side of the 50S subunit, with the head of the 30S particle positioned between the L1 ridge and the central protuberance. This arrangement is supported by single- and double-antibody labelling of 70S ribosomes (59).

Both X-ray and neutron-scattering methods have been employed to study the shape of ribosomes and their subunits in solution (65, 74). Although the results of solution scattering experiments are much more subtle to interpret than those of EM, they are nevertheless of great importance because they measure the properties of fully hydrated particles under conditions that are approximately physiological. It is therefore reassuring that a model for the 50S subunit derived from neutron-scattering data (130) bears striking resemblance to that obtained by EM.

Besides shape, information concerning the relative spatial distribution of RNA and protein can be obtained from measurements of their respective aggregate radii of gyration in contrast-matching experiments. There is presently general agreement that the protein and RNA distributions in the 50S subunit have roughly coincident centers of mass but that the protein component has a much larger radius of gyration (90 to 100 Å than the RNA component (55 to 60 Å) (21). In the 30S subunit, the distributions of the protein and RNA components are again roughly concentric, with a tendency for the surface to be protein rich; the radii of gyration are ca. 78 Å for the protein and 61 Å for the RNA (reviewed in reference 22).

Structural details of macromolecules are traditionally solved by X-ray crystallography. Unfortunately, ribosomes, particularly those from E. coli, have proven difficult to crystallize. In recent years, however, significant progress has been made in crystallization of ribosomal subunits from halophilic (31, 147) and thermophilic (77, 137, 165) bacteria. Recently, Von Bohlen et al. (147) have obtained crystals of 50S subunits from Haloarcula marismortui that diffract to less than 3 Å. Deducing phase information for these crystals will probably require isomorphous crystals containing heavy-atom clusters at defined positions of the particles; this should prove to be a challenging task. However, the final goal—the molecular structure of the 50S ribosomal subunit at nearly atomic resolution—will more than justify the many years of intensive effort that will have been expended.

Internal Architecture

. Possible clues to internal structural features have been suggested by results from electron microscopy and from low-angle X-ray scattering. Three-dimensional reconstructions of electron micrographs show channel-like interior structural features, which may correspond to positively stained segments of the rRNA. It has long been known that in low-angle X-ray scattering studies on ribosomes, weak maxima are observed at about 45, 30, and 25 Å (60). Using protein-depleted particles and contrast variation methods, Serdyuk et al. (118) have shown that these maxima are due to structural features of the rRNA. The data are consistent with parallel packing of RNA helices within the ribosome, as originally suggested by Langridge (60).

Location of Proteins and RNA in Ribosomes

During the past decade, much has been learned about the spatial arrangement of ribosomal proteins in the ribosome. Most crucial has been information obtained from immunoelectron microscopy (IEM) and neutron diffraction studies (reviewed in references 16, 93, and 129). Important contributions have also come from singlet-singlet energy transfer (52), protein-protein cross-linking (140), and protein-RNA cross-linking (8). Positioning of the RNA in the ribosome represents a different sort of problem. Apart from positioning the relatively small 5S rRNA, location of the rRNA component might at first be considered trivial, because its mass probably coincides roughly with that of the whole particle. However, the information of greatest interest is the location of specific sites or regions of the rRNA in the ribosome. In effect, this amounts to determination of aspects of its tertiary structure. Among the relevant experimental approaches to this problem are IEM (see, e.g., reference 101), singlet-singlet energy transfer (94), RNA-protein or RNA-RNA cross-linking (8), protein-binding site studies, and model building, using the aforementioned data and stereochemical rules as constraints (7, 67, 126).

Immunoelectron Microscopy.

The principle of IEM is based on the fact that individual antibody molecules can be visualized by EM. Antibodies to purified ribosomal proteins or to haptens are bound to ribosomes or subunits and visualized by EM (93, 95, 129). The relative position of the protein in the ribosome structure is inferred by the position of the antibody attachment site in relation to recognizable structural features on the ribosome. Although this method is straightforward in principle, there was a great deal of controversy over various results in previous years. This appeared to be attributable to a number of problems, both experimental and interpretational. Happily, this situation has largely been resolved (see the discussion in reference 108). Many of the ribosomal proteins have been located by this method on their respective subunits, and the placements are in quite good agreement with the results obtained from other approaches.

Features of the RNA have also been placed in the EM model by using EM-based methods. Antibodies to methylated nucleotides have allowed positioning of the dimethyladenosines 1518 and 1519 of 16S rRNA on the platform (101) and ⁷m G-527 on the body (142). Derivatization of the 5' and 3' termini of the rRNAs with haptens has made possible the location of the 5' and 3' ends of 16S rRNA and the 3' ends of 5S and 23S rRNA (reviewed in reference 83). Another approach is the use of synthetic DNA oligonucleotides that hybridize to accessible regions of 16S rRNA. Their locations can be identified by EM methods with an antibody to a hapten attached to the oligonucleotide (61) or by binding of multiple avidin molecules to a biotin polymer attached to the oligonucleotide (92). Together, these EM-based approaches have been used to localize several important features of 16S rRNA in the 30S subunit structure.

Neutron Diffraction.

A three-dimensional map of the positions of the 21 30S subunit proteins has been determined by neutron diffraction (16). Selective deuterium labelling of pairs of proteins that are reconstituted with the other (nondeuterated) components into a 30S subunit allows measurement of the distance between the centers of mass of the two deuterated proteins. Measurement of numerous pairwise distances eventually permits triangulation of their spatial positions relative to one another. This three-dimensional map (Fig. 10) correlates well with the results of the IEM approach and permits correlation of the neutron data with the structure for the 30S particle obtained by EM (Fig. 8 and 9). In most cases, there is good agreement between the positions of proteins as determined by IEM and neutron diffraction. However, for certain proteins, significant discrepancies exist, not only between the IEM results from different laboratories but also between the results of IEM and neutron diffraction. These proteins include S16, S19, and S20. S19 has been placed in two different positions in the head of the 30S subunit by IEM and in a third by neutron diffraction (16, 93, 129). Early IEM results were equivocal for S16 and S20, but recent studies (117) place these proteins much lower in the 30S subunit than their respective positions in the neutron map. Resolution of this dilemma may require devising yet other methods for localization of r-proteins.

Fig. 10Stereo view of the neutron diffraction map of the three-dimensional positions of the 30S r-proteins (16). Each protein is represented as a dotted sphere, scaled to the anhydrous mass of each protein, centered on the measured position of its center of mass.

Source:

Singlet-Singlet Energy Transfer.

Distance measurements can also be obtained from radiationless singlet-singlet energy transfer between two fluorescent probes. Two different fluorescent groups, an acceptor and a donor, are attached covalently to two different positions in the ribosome. Particles are irradiated at a wavelength at which the donor, but not the acceptor, absorbs. The excited acceptor molecule decays, transferring its fluorescent energy to the acceptor, which absorbs in the region of the emission spectrum of the donor. The donor then fluoresces, and the emitted energy is measured. Distances can be inferred because the quantum yield of the transfer is proportional to the inverse sixth power of the distance between the probes. These measurements have been performed for many of the possible pairwise combinations of 30S subunit proteins (52). The results are in quite good agreement with neutron diffraction measurements that have been performed on the same pairs (Table 2), with the possible exception of the S4 to S15 and the S4 to S20 distances. Because S4 is known to have an elongated structure in the ribosome (109), the discrepancies could be accounted for by an asymmetric distribution of the fluorescent probes on S4. One drawback of this approach is that the fluorescence methods tend to underestimate the longer distances (75).

Table 2Some protein-protein proximities in the 30S ribosomal subunita

Singlet-singlet energy transfer has also been used to measure inter- and intramolecular distances between the ends of the rRNAs (94). Distances between probes attached to the 3' ends of the 16S rRNA and the 5S or 23S rRNA were estimated to be about 55 and 71 Å, respectively. The corresponding distance between the 5S and 23S rRNAs was too large to be measured accurately but was estimated to be greater than 65 Å.

Cross-Linking.

Neighboring molecules in the ribosome have been identified by covalent cross-linking, giving protein-protein, protein-RNA, or RNA-RNA proximity information. Both chemical and photochemical strategies have been employed, and results have come from many different groups of investigators. Protein-protein cross-linking has been reviewed in reference 141, and RNA-RNA and RNA-protein studies have been reviewed in references 8 and 10.

Among the bifunctional reagents that have been employed, 2-iminothiolane (140) is one of the more useful and elegant examples. This cyclic reagent reacts with amino groups and, in so doing, opens to provide an available sulfhydryl moiety. Oxidation then creates disulfide bridges between neighboring sulfhydryls, forming covalent cross-links. Cross-linked proteins are then identified by two-dimensional diagonal gel electrophoresis, in which they migrate as pairs in off-diagonal positions. Some results of protein-protein cross-linking are summarized in Table 2, in comparison with results from other protein-proximity studies. There is a strong correlation with the neutron diffraction and fluorescence studies; proximal proteins are cross-linked, and distal ones (>40 Å apart) are not.

Molecular Model for the 30S Subunit

Several groups have derived molecular models for the folding of 16S rRNA in the 30S ribosomal subunit (7, 53, 67, 126). One model (126) is based on the positions of the 30S proteins, as determined by neutron diffraction. Starting with the phylogenetically derived secondary structure for 16S rRNA, individual helical elements were docked on their cognate proteins while maintaining stereochemically feasible distances between helices. The footprinting and cross-linking data summarized in Fig. 6 and 7 were used to determine which elements of the 16S rRNA were to be docked with which probes. A more recent version of this model (H. F. Noller, T. Powers, G. Heilek, S. Mian, and B. Weiser, unpublished data), incorporating an additional set of data obtained from hydroxyl radical probing of protein-RNA interactions (T. Powers and H. F. Noller, unpublished data), is shown in Fig. 11. The resemblance between this model and those derived from EM (Fig. 8 and 9) is clear. The head, body, and platform elements are readily identifiable. One apparent difference is that there is somewhat less mass in the lower part of the body of the molecular model than in the EM model. This may be due to the fact that only about 75% of the 16S rRNA structure has so far been incorporated into the molecular model; possibly, the remaining 25% of the RNA occupies the bottom of the subunit.

Fig. 11Model for the folding of 16S rRNA in the 30S ribosomal subunit (H. F. Noller, T. Powers, S. Mian, S. Stern, and B. Weiser, unpublished data).

Source:

One important prediction of this model is that the three major domains of the 16S rRNA secondary structure fold into largely autonomous three-dimensional structural domains. Thus, the 3' domain defines the head, the 5' domain defines the body, and the central domain defines the platform and cleft region. Recently, the 3' domain has been assembled with its cognate proteins into a well-defined globular ribonucleoprotein particle (115), providing direct evidence for the structural autonomy of this domain.

Another consequence of the model is that nearly all of the tRNA-protected bases in 16S rRNA (70, 71) are clustered around the cleft region, on the head and platform. This is consistent with the localization of tRNA in the 30S subunit by EM methods (34, 148). However, at least one set of tRNA-protected bases, in the 530 loop region, is located remote from the cleft; the implication that protection of these bases is caused by an allosteric mechanism has been the subject of extensive debate (6, 105).

Models of this kind are useful in several respects. Besides being a valid approach (and, so far, perhaps the only approach) to deducing ribosome structure, albeit at low resolution, model building readily identifies what kinds of new information are needed to test, confirm, or extend our present structural knowledge. In addition, it serves to sharpen and focus discussions of ribosome structure and function, as the field more and more explicitly addresses the molecular mechanisms underlying protein synthesis. Finally, interpretation of X-ray crystallographic results will most probably benefit from these model-building exercises.

CONCLUSION

Early hints of an absence of structural symmetry for the ribosome have by now been amply confirmed. Primary structures for the 52 r-proteins and the three rRNAs and secondary structures for the latter show a staggering level of structural complexity. Solution of the three-dimensional structure of the ribosome will therefore probably prove to be a difficult task. An understanding of the mechanism of ribosome assembly, both in vitro and in vivo, will rely on an improved understanding of its structure and of the numerous protein-RNA interactions that make up this particle.

Our knowledge of ribosome structural organization is presently most advanced for the E. coli 30S subunit. The extent to which the various approaches are coalescing to give a coherent view of this particle is evident from Fig. 11.

Corresponding data for the 50S subunit are less advanced, but in this case recent progress suggests that the large subunit may yield to crystallographic analysis before the 30S subunit. However, efforts on elucidating its structure will very probably continue to benefit from approaches used in studying the small subunit. Even when high-resolution crystallographic data are forthcoming, interpretation will undoubtedly rely heavily on the kinds of results that are now being obtained with the less direct biochemical approaches described above. In any case, the coming years should provide many exciting insights into the nature of this ancient and ubiquitous biological structure.