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

Username: 

Password:  

 

Chapter 60

Suppression of Termination Codons

HANNA ENGELBERG-KULKA and RACHEL SCHOULAKER-SCHWARZ

 

INTRODUCTION

In both prokaryotes and eukaryotes, the termination of polypeptide chain synthesis is signaled by one of the three termination codons: UAG (amber), UGA (opal), or UAA (ochre). When the protein translating complex reaches a termination codon in the mRNA, protein release factors recognize the termination codon, causing the release of the polypeptide chain. There are two classically described means for overcoming (suppressing) translation termination: (i) nonsense suppression, where a tRNA reads a termination codon as though it codes for an amino acid, and (ii) frameshift suppression, a shift of the translation apparatus into a different (+1 or –1) reading frame. These two processes were discovered in Escherichia coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) (for reviews see references 54, 117, 118, and 123). Even today, pioneering research in the field of suppression is carried out in these two kinds of bacteria. Nonsense and frameshift suppression were originally described as low-efficiency processes caused by a mutation in the tRNA. We shall include in our discussion the significant advances in these areas that have been made over the last decade. This chapter is a review of the following topics: (i) recently discovered mutations in tRNA, and other components of the translation apparatus that affect suppression; (ii) the role of the context of the termination codon in nonsense suppression in general, and more specifically in the incorporation of the 21st amino acid, selenocysteine; (iii) naturally occurring suppression of termination codons by translational read-through, frameshifting, or bypassing of a large fragment of the mRNA (even in the absence of a mutation in the translation apparatus); and (iv) the regulatory implications of these processes.

There have been several excellent reviews on topics included in this chapter, such as those by Murgola (93), Fox (49), Eggertsson and Söll (34), Parker (108), and Atkins et al. (3). Several other recent reviews dealing with particular aspects of the topics covered here will be referred to at appropriate locations in the text.

 

RELEASE OF THE POLYPEPTIDE CHAIN

 

Release Factors

The mechanism of protein synthesis termination is not yet known precisely. In bacteria, it appears that when termination codons are in the ribosomal A-site (or very near to it), each of them is specifically recognized by a release factor protein which causes the hydrolysis of the peptide from the peptidyl-tRNA in the ribosomal P-site (108). In E. coli, two codon-specific release factors direct the termination of protein synthesis. Release factor 1 (RF1) catalyzes termination at UAG and UAA codons, and RF2 catalyzes termination at UGA and UAA codons (122). The gene encoding RF1, prfA, has been cloned and maps at 27 min on the E. coli chromosome; mutations in prfA have been shown to cause suppression of both UAG and UAA nonsense codons (33, 104, 105, 119, 120, 138). The RF2 gene, prfB, has been cloned and maps at 62 min on the E. coli chromosome; mutations in prfB have been shown to suppress UGA nonsense codons (24, 76, 77, 79). Some of these mutations have been found to be conditionally lethal (76). In both the E. coli and S. typhimurium chromosomes the operons specifying RF1 and RF2 seem to be similarly organized (36, 77). A third factor, RF3, which stimulates the activities of RF1 and RF2, has been purified (57) but has received little attention. However, recently an E. coli nonsense suppressor mutation (tos) has been located and characterized in a putative gene for RF3. tos maps at 99 min of the E. coli chromosome and suppresses all three nonsense codons (61a, 85). The suppression of the termination codons by mutations in release factors can be explained in two ways: (i) the release factors and suppressor tRNAs (either mutated or natural) compete for the termination codons, and/or (ii) the mutated release factors cause a delay in the release of the polypeptide chain from the ribosome, permitting a nonspecific incorporation of an amino acid at the termination codon.

 

Ribosomes

Various studies have been carried out to characterize the ribosomal requirements for release factor binding and function. An interaction of the release factors with the ribosomes is required for the release of the polypeptide chain. The nature of such an interaction was studied using several experimental approaches which included antibody inhibition, ribosomal protein reconstitution, protein cross-linking, and antibiotic inhibition. The results of all of these studies suggest that release factor binding is at the interface between the two ribosomal subunits (for review, see reference 30). The structural details of these interactions will be clearer when the appropriate crystallographic data have been gathered. Other important unsolved issues are the mechanisms involved in the termination event, in stop codon recognition, and in release factor binding. For example, are release factor binding and stop codon recognition a single step or two or more separate steps in the termination process? The results of early studies suggested that RF1 and RF2 bind directly to the stop codon (for review, see reference 30). The results of more recent investigations suggest that in addition to the protein release factors, 16S rRNA also participates in the termination event (59, 64, 81, 95). This role proposed for 16S rRNA is primarily based on the discovery that the deletion of base C1054 (ΔC1054) in 16S rRNA specifically suppresses UGA, but does not suppress the two other termination codons, UAA and UAG (95). Base 1054 is part of helix 34 of E. coli 16S rRNA (numbering according to reference 84), which contains two tandem 5'-UCA-3' triplets (1199-1204), complementary to the 5'-UGA-3' triplet termination codon (Fig. 1). It has been proposed that this region of the rRNA base pairs directly with the UGA stop codon; by coaxial stacking this paired region would form a helical element which could be the recognition signal for RF2-mediated termination (59, 94).

Another group of investigators have suggested that helix 34 may have a different role in translation termination than recognizing stop codons (110). They found that ΔC1054 16S rRNA does not cause UGA suppression only, but that when the appropriate corresponding tRNAs are introduced into the cells, ΔC1054 16S rRNA also suppresses the UAG and UAA termination codons. Thus, ribosomes with an altered helix 34 can act as a "super suppressor." Apparently, the altered ribosomes do not require suppressor tRNAs in order to suppress the UGA termination codon because in E. coli only UGA, and not UAG or UAA, can be read at low efficiency by a normal tRNA (see below). Because base pairing with UAA or UAG stop codons is not possible in the vicinity of C1054, a nonsense codon-16S rRNA base-pairing model (95) is unlikely to be correct. An alternate hypothesis is that the role of helix 34 in translation termination may involve an interaction with the release factors (110, 111). To test this hypothesis, the combined action on nonsense codons of a nonsense tRNA suppressor with a series of base substitution mutations in the rRNA was examined (Fig. 1) (111). It was found that some of the ribosomal mutations had an effect on UGA, UAG, and UAA suppression: the magnitude of the effect was dependent on the specific nonsense codon examined and on the position and nature of the mutation within the helix. Two trends were evident: (i) mutations that affected UAG suppression did so by decreasing the efficiency of suppression, whereas (ii) mutations that affected UGA suppression did so by enhancing the efficiency of suppression. It was concluded that the two release factors RF1 (recognizing UAA and UAG) and RF2 (recognizing UAA and UGA) compete for an overlapping domain of the rRNA. In addition, altering helix 34 of the 16S rRNA may reduce the efficiency of RF2 binding and consequently cause an apparent increase in the efficiency of RF1 binding. Furthermore, this model implies that, as has previously been suggested (127), the initial interaction of the release factors with the ribosome is nonsense codon independent.

As revealed by immunoelectron microscopic studies (90), the release factor interacts with two separate domains on the 30S rRNA subunit. The facts that both RF1 and RF2 recognize the UAA nonsense codon, and that according to immunoelectron microscopic data RF1 binds lower on the small 16S rRNA subunit than does RF2, support the conclusion that the two release factors occupy distinct but overlapping binding sites (128). Since ribosomes carrying the ΔC1054 mutation increased the efficiency of the suppression of all three nonsense codons (94), it seems that this mutation affects the interactions of the ribosome with both RF1 and RF2.

 

NONSENSE SUPPRESSION

 

Termination Codons Can Be Read by Suppressor tRNAs and by Some Native tRNAs

Nonsense suppressors are tRNAs that have been modified, usually in the anticodon, such that in response to a termination codon they insert an amino acid. These suppressor tRNAs have been important tools in bacterial genetics (125), in studies of recognition of tRNAs by aminoacyl-tRNA synthetase (107), and in protein engineering (88, 99). In E. coli, five amber suppressors, which recognize the amber codon 5'-UAG-3', have been generated in vivo, mostly by a single base change resulting in the anticodon 5'-CUA-3' (58, 74, 124, 129, 142). These are the amber suppressors of genes supD, supE, supF, supP, and glyT, which, in response to an amber codon, insert serine, glutamine, tyrosine, leucine, and glycine, respectively (58, 74, 97, 124, 129, 142). Another gene, supG, encodes a suppressor tRNA that recognizes both ochre (UAA) and amber (UAG) codons, permitting the insertion of lysine (60). In S. typhimurium, some analogous suppressor tRNAs have been generated in vivo (12). Furthermore, over the last decade, use of in vitro site-directed mutagenesis, enabling more than one change to be made in the anticodon and thereby the conversion of numerous tRNAs to suppressor tRNA (78, 98, 99), has greatly increased the number of E. coli amber suppressors available. Recently, a method has been developed for the incorporation in vitro of modified amino acids into a polypeptide chain (37). It uses an amber suppressor tRNA charged with the modified amino acid, which is then incorporated at an amber codon at the desired position of the polypeptide. The ability to incorporate these modified amino acids site specifically allows for novel studies of protein structure (6, 27, 37).

In E. coli, there are two kinds of UGA suppressors derived from tRNA^Trp, the product of trpT. The anticodon of the first, Su7 (also sometimes designated genetically as supU) (5), has been modified from CCA to UCA and can read both tryptophan and UGA codons (112). The anticodon of the second, Su9, has not been modified; instead there is a base change in the dihydrouridine stem (a base substitution from G to A at position 24) (25, 69, 112) that enables tRNA^Trp to read UGA codons at high efficiency. Note, however, that in both E. coli and S. typhimurium, UGA codons can also be read by natural (unmutated) tRNA^Trp, although at low efficiencies (12, 70, 71, 91, 132). There is yet another natural tRNA in E. coli and, presumably, in other members of the Enterobacteriaceae that can read UGA codons. This is tRNA^Sec, which is specified by gene selC (68, 82) and mediates the UGA-directed incorporation of the modified amino acid selenocysteine into proteins which are called selenoproteins (for reviews see references 16 and 17). tRNA^Sec carries the anticodon UCA that matches the UGA termination codon. It is the longest known tRNA species and has a variable region of 22 nucleotides in which several normally invariant sequence positions deviate from the consensus. tRNA^Sec is charged with serine by seryl-tRNA synthetase (82). The serine moiety bound to tRNA^Sec is then further converted to selenocysteine in a reaction catalyzed by selenocysteine synthetase, the product of selA (45, 47). Another reaction, involving the selD product, provides an activated selenium donor compound for the conversion reaction (35, 131).

 

Codon Context and Nonsense Suppression

The influence of the codon context on the efficiency of nonsense suppression has been thoroughly investigated, particularly in E. coli and S. typhimurium (for references see reference 79). It has been found that the efficiency of nonsense suppression is influenced by the nucleotides surrounding the termination codon (the codon context). The nucleotides immediately at the 3' side of the termination codon have been most strongly implicated as responsible for this codon context effect; nonsense suppression is affected when either a purine (high level of suppression) or a pyrimidine (low level of suppression) is located at this position (4, 18, 19, 44, 87). However, some data indicate that the nucleotides at the 5' side are also involved (22, 44, 92). It should be mentioned that some of these studies were carried out by using nonsense codons in different immediate contexts located at various sites of a gene (22, 44, 87, 92). Thus, in addition to variations in the immediate context of the nonsense codons, there were variations in a broader context which may also affect nonsense suppression. Bossi and Roth (19) were the first to take an approach in which the broad context remained constant and the only change was the immediate context of the termination codon. They have shown by elegant experiments that the suppression efficiency of the UAG codon by amber suppressor tRNA supE of S. typhimurium is significantly increased due to the alteration to A of the nucleotide following UAG. More recently, a systematic study was carried out in E. coli on the influence of the codon context on UGA suppression by the UGA suppressor tRNA Su9 of E. coli (79). This was achieved by a series of constructs in which only the immediate context of the TGA codon was varied by only one nucleotide at a time. The influence of the UGA codon context obeyed the following rules. (i) The nature of the nucleotide immediately adjacent to the 3' side of the UGA is an important determinant: at that position, the level of UGA suppression is influenced by the nucleotides in the order A > G > C > U (a similar 3' codon context effect has recently been observed in the case of the E. coli tRNA^Trp UAG suppressor [109]). (ii) At extremely high or low levels of UGA suppression, the influence of the adjacent 3' nucleotide is not seen. Finally, (iii) in all cases, the nature of both the nucleotide immediately adjacent to the 5' side of the codon and that following the base adjacent to the 3' side of the codon has little effect, if any, on UGA suppression.

Several models have been proposed to explain the effect of codon context on nonsense suppression (4, 19, 126; reviewed in reference 140). The hypothesis originally suggested by Taniguchi and Weissman (126) implicated a role for the U residue invariably at position 33, adjacent to the 5' side of the tRNA anticodon. This U residue could possibly base pair with the purine residue at the 3' side of the codon and would thus form either a U-A or a U-G pair. The possibility of such base pairing was rejected based on the results of direct experimentation on the UAG suppressor variant tRNA^Trp Su7-UAG (4) and the UGA suppressor variant tRNA^Trp Su9-UGA (56). By constructing and testing every possible nucleotide combination of bases between position 33 of the tRNAs and the nucleotide immediately at the 3' side of the nonsense codons, it has been shown that the codon context effect does not require a fourth base pair. Based on these results, another hypothesis was proposed by Ayer and Yarus (4). This hypothesis suggests that a purine at the 3' side of the codon would increase the stability of codon-anticodon interaction. This would be analogous to the "dangling-end effect" on the stability of pairing (50, 51); it follows that the nucleotide at the 3' side of the codon would be the most effective in the stabilizing role and that a purine would stabilize better than a pyrimidine. As has been shown (79), at least in UGA suppression by Su9, the effect of the nature of the nucleotide at the 3' side of the UGA is more subtle than simply purine > pyrimidine: the nucleotide in this position influences the level of UGA suppression in the order A > G > C > U. Such a hierarchy is not seen in the dangling-end effect on the stability of pairing of short duplexes in solution, for which only purines and pyrimidines can be differentiated (50, 51). This is probably because the effect of the 3' nucleotide on mRNA-tRNA interaction is more complicated than is the base pairing of short duplexes in solution. Alternatively, it may be that the suppressor tRNA^Trp Su9, which has its mutation in the stem of the dihydrouracil loop (rather than in the anticodon) and which can base pair only to the first two nucleotides of the UGA codon, represents a unique case of mRNA-tRNA interaction.

 

Codon Context and UGA Read-Through

The process permitting a natural tRNA to read a termination codon is called translational read-through, and in the case of the UGA codon it is called UGA read-through (for reviews see references 40 and 108). So far, in E. coli, only two natural tRNAs have been implicated in UGA read-through: tRNA^Trp causes UGA-directed tryptophan incorporation into a polypeptide, and tRNA^Sec causes selenocysteine incorporation (see above). It is obvious to ask, in E. coli genes carrying TGA codons, what causes the selection between the termination of the polypeptide chain by RF2 and UGA read-through by either tryptophan or selenocysteine? The results of recent experiments have revealed that the UGA codon context is an important determinant in the selection process between these three options. Initial examination of the nucleotide sequences of genes coding for UGA read-through proteins in E. coli has already revealed the presence of an A residue at the 3' side of UGAs that permits the incorporation of tryptophan; it has been hypothesized that it is the A residue at this position that permits the UGA read-through process by charged tRNA^Trp (38). This hypothesis has been further supported by the results of direct experiments showing that the nature of the nucleotide at the 3' side of the UGA is an important determinant not only in UGA suppression by the suppressor tRNA^Trp Su9 (see above), but also in the UGA read-through process which probably takes place by natural charged tRNA^Trp (79); the nucleotides at the 3' side of the UGA affect this process in the order A > G > C > U, the same order as when Su9 is used as a UGA suppressor. Thus, UGA read-through by tryptophan is determined by the nature of the nucleotide at the 3' side of the UGA. In contrast, a more elaborate UGA codon context is required for the incorporation of selenocysteine.

In E. coli, a selenocysteine-specific UGA codon context has been identified that permits the UGA to be read as a selenocysteine codon, rather than a termination codon (143). This codon context is 40 nucleotides long, can form a stem-and-loop structure (Fig. 2a), and is located immediately downstream from the UGA in the E. coli fdhF (143) and fdnG (11) genes in the mRNAs specifying for the selenocysteine-containing enzymes formate dehydrogenase H and N, respectively. In E. coli, inserting this specific UGA codon context into a heterologous gene like lacZ results in the incorporation of selenocysteine into the corresponding polypeptide (26, 113, 143); this incorporation process is dependent on several specific components, such as the product of selC (tRNA^Sec), and the presence of simple selenium derivatives in the medium (26, 113, 143). Such a lacZ construct in E. coli provides a specific and sensitive reporter system for determining the concentration of several selenium derivatives, including selenocysteine itself and some of its precursors, and thus can be used as a selenium bioassay (113). The unique secondary structure of the mRNA (Fig. 2a) immediately downstream from the UGA codon has been found to be essential for selenocysteine incorporation (143). Based on mutagenesis experiments, it seems that each part of the construction of this stem-and-loop structure is highly specific, including the primary sequence in the loop and stem region, the correct folding of the stem, and the size of the stem (67). The loop region is particularly important: it has been suggested that this loop region may serve as a recognition element for the SELB protein (67). SELB is a specialized elongation factor that can transport selenocysteine-tRNA^Sec to the ribosome and thus can alternate for elongation factor EF-Tu (46, 48). It has recently been confirmed that SELB does bind specifically to the loop region of the mRNA, forming a complex with selenocysteinyl-tRNA^Sec at the mRNA (7). Based on these results, a mechanism for the UGA-directed selenocysteine incorporation has been proposed (7) which is described in Fig. 2b. According to this model, SELB, the selenocysteine-specific elongation factor, in the presence of GTP forms a complex with selenocysteinyl-tRNA^Sec which binds to the stem-and-loop recognition element at the 3' side of the UGA codon of the mRNA. A ribosome approaching from the 5' side partially melts the hairpin structure, and as the ribosomal A site reaches the UGA codon, selenocysteinyl-tRNA^Sec is correctly positioned for codon-anticodon recognition (see Fig. 2b).

How the machinery for selenocysteine incorporation may compete for the recognition of UGA codons with both RF2 and UGA suppressors is not clear. Possibly, this UGA codon is shielded either by the hairpin structure of the mRNA by itself or when it is complexed with SELB and tRNA^Sec. Thus shielded, this UGA would not be recognized either by RF2 or by a UGA suppressor tRNA. In recent experiments, it has been asked whether the position of the UGA codon relative to the hairpin structure of fdhF mRNA affects the efficiency of selenocysteine incorporation (26). It appears that the efficiency of selenocysteine incorporation is highest when the UGA codon is located immediately upstream from the hairpin structure. Furthermore, when the UGA codon is in this position, nonspecific suppression is least efficient. These attractive but incomplete models for the competitive interaction of the numerous components involved in the recognition apparatus of the selenocysteine-specific codon context still have many unanswered questions.

 

FRAMESHIFT SUPPRESSION

A termination codon can also be overcome by a shift of the translational apparatus into a different (+1 or –1) reading frame. Such a frameshifting event can occur close to or far from the termination codon. Most cases of translational frameshifting are due to the slippage of the ribosome forward or backward by one nucleotide from the reading frame of the initiation codon (for reviews see references 2 and 108). However, as recently discovered, frameshifting can also be caused by the bypassing of a larger segment of the mRNA (2, 8, 39, 137). Thus, a termination codon in the right place could be simply bypassed by the translational apparatus. Frameshifting can be either programmed or unprogrammed (for reviews see references 2, 3, and 108). In programmed frameshifting, the ribosomes shift frames in response to specific signals in the mRNA, in some cases with an efficiency of 50% or more (3, 108). In the absence of such specific signals for programmed frameshifting, the maintenance of the normal triplet decoding is highly efficient (the frequency of frameshifting is 10^–4 to 3 × 10^–5 per codon) (108, 135). However, this efficiency can be drastically changed by mutations in the translation components causing frameshifting (up to an efficiency of 20%) at sites where it does not normally occur (3). Here we shall mainly consider some interesting mutations in various translation components affecting frameshifting in E. coli and S. typhimurium. In addition, we shall also discuss programmed frameshifting as a regulatory mechanism of gene expression.

 

Mutations in Translation Components Affecting Frameshifting Mutant tRNAs

Mutations in the translation apparatus that affect frameshifting are detected by virtue of their ability to compensate for (to suppress) the defect caused by frameshift mutations at the coding level. They have therefore been called frameshift suppressors (for reviews see references 3 and 118). The most numerous and most efficient frameshift suppressors are altered tRNAs that occasionally induce a shift in reading frames. Based on their alterations, they are here divided into the following groups.

Group 1: tRNAs with an Extended Anticodon.

tRNAs with an extended anticodon were the first extragenic frameshift suppressors described and were isolated in S. typhimurium (115; for reviews see references 117 and 118). A common feature of these +1 frameshift suppressors is a base insertion next to the anticodon: G in sufA and sufB and C in sufD (Fig. 3a). These modified tRNAs occasionally read a run of 4 bases, e.g., a run of C’s (sufA and sufB) or a run of G’s (sufD). In light of this fact, a quadruplet codon-anticodon pairing seems to be the expected mechanism for these +1 suppressions. Although its anticodon does not include a run of 4 similar bases, the more recently isolated frameshift suppressor, sufT621 of S. typhimurium (Fig. 3b) (73), also belongs in this same category. In this case, the extra base, G, has been inserted adjacent to the 3' side of the anticodon (3'-GCI-5' of a CGU coding

. Presumably this results in the anticodon 3'-GGC(I)-5' (Fig. 3b), which can promote quadruplet translocation at the proline codon 5'-CCG(U)-3' (73).

Group 2: tRNAs with an Extended Anticodon Loop.

The tRNA +1 frameshift suppressor sufJ128 of S. typhimurium belongs to this group. It is a mutant

in which the size of the anticodon loop has been increased from 7 to 8 nucleotides by the insertion of an extra C residue (20, 21). This C residue has been inserted outside the anticodon at the boundary between the anticodon loop and the anticodon stem (Fig. 3c) and not within the anticodon itself. Thus, sufJ is unlike the frameshift suppressor tRNAs in group 1 above. The functional difference between the two groups is seen in the particular decoding specificity of the sufJ suppressor, which can read any ACCN quadruplet regardless of the identity of the base at the fourth codon position. Thus, the use of a 4-base anticodon does not seem to be involved. Instead, this suppressor appears to recognize the standard triplet ACC and then cause the ribosome to "translocate over" 4 bases (20). According to a model for this "three out of four" reading method (21), a tRNA with an 8-nucleotide anticodon loop can either pair to a quadruplet (group 1) or pair to a triplet and ignore 1 base (independent of its identity) and read the next available triplet in the +1 frame (group 2). Originally isolated as a missense suppressor (96), the E. coli frameshift suppressor glyT can also be explained by this model. glyT has an 8-base anticodon loop (Fig. 3d) and has also been found to act as a frameshift suppressor (130).

Group 3: tRNAs with Altered CCA Ends.

The trinucleotide CCA is found universally at the 3' end of tRNA molecules, where it serves as the site of amino acid attachment. Despite this extreme conservation, mutants of E. coli

with GCA or ACA ends have been isolated on the basis of their ability to promote frameshifting (103). These same mutants also caused read-through of termination codons located one or two codons downstream from the valine codons decoded by the mutant tRNA. It is known that, as the ribosome progresses through its cycle, the interaction of the CCA end of tRNAs with various sites on the 23S rRNA ribosomal subunit serves to anchor the tRNA at one or two alternate spots on the ribosome (89). It has therefore been proposed that disrupting the interactions between the large ribosomal subunit and the CCA end of a tRNA molecule would promote aberrant codon-anticodon interactions on the small ribosomal subunit. In the case of , these aberrations are manifested by frameshifting and the subsequent read-through of stop codons (103).

Group 4: tRNA Modifications.

1-Methylguanosine (m¹G) is one of the most evolutionarily conserved nucleosides. It is found at position 37 in tRNAs that read codons starting with C like tRNA^Pro (CCN), tRNA^Leu (CUN), and tRNA^Arg (CGG) (14). In these tRNAs, in S. typhimurium, the methylase encoded by gene trmD modifies G37 to m¹G (15). The methyl group of m¹G prevents the standard base pairing between G and C. The mutant trmD3 lacks m¹G in tRNA^Pro and tRNA^Leu at high temperatures. This deficiency has been correlated with +1 frameshifting activity for these tRNAs (63). In the case of tRNA^Pro, it has been shown by protein sequencing that a quadruplet translocation had occurred and that, probably because the unmodified G37 base pairs to the first C of the proline codon, a proline was inserted at these sites. It has thus been suggested that the methylation of G37 is important for reading-frame maintenance (63).

Group 5: Hopping tRNAs.

The hopR mutants are a special group of tRNA suppressors that cause tRNAs to "hop." The anticodon on the mRNA is the "take-off site" from which they can dissociate and then hop further downstream on the mRNA to an identical "landing site" (73, 101). When the landing site triplet codon is in a new frame, a frameshift occurs (73). Two hopR mutants have been investigated (101). These are both mutants of tRNA^Val in which an A or a U has been inserted between positions 34 and 35 of their anticodons. The hopR mutants were isolated as –1 frameshift suppressors that caused a single valine residue to be inserted for the 5-base sequence GUGUG, which has overlapping codons. Detachment from the 5'-GUG valine codon (underlined) leads to bypassing 2 bases to land on the overlapping GUG valine codon (overlined). Hopping also occurs when the valine codons are separated by a termination codon. However, the two different hopR mutants overcome the termination codon differently (101). The mutant with the anticodon sequence 3'-CAAU-5' hops over the termination codon UAA in the mRNA sequence GUGUAAGUU by inserting a single valine residue. In contrast, the mutant with the anticodon 3'-CAUU-5' overcomes the termination codon in that same sequence by hopping over it and inserting two valine residues. It seems that in some instances of hopping, alternate extended anticodon bases are used at the take-off and landing sites. The maximum hopping efficiency is about 10%, and it declines with increasing distance of the hop; hopping tRNAs bypass a maximum of two pairs of codons. It is interesting that wild-type tRNA^Val has also been found to hop, though only at low efficiencies of hopping (up to 1%) (101, 134). Hopping over larger segments of mRNA has also been described (8, 72, 137). Long-distance hopping is a programmed event that probably has regulatory implications (see below).

 

Mutant Ribosomes

All these ribosomal mutants increase the frequency of +1 and –1 frameshifting and can also act as termination codon suppressors. They all occur either in a ribosomal protein or in regions of rRNA involved in the formation of the translational complex.

Ribosomal ambiguity mutants (ramA mutants, now designated as rpsD) have been known for a long time. These mutations, found in ribosomal protein S4, act as low-level suppressors of both nonsense and frameshift mutations; they are restricted by a second class of mutants, strA or rpsL, having an altered ribosome protein S12 (1). Two other ribosomal ambiguity mutants, rpsE and rplL, have been shown to increase missense and nonsense suppression due to changes in ribosomal proteins S5 and L7/L12, respectively. Their ability to cause frameshifting has not yet been tested (for review see references 3 and 108). Recently, Dahlberg and coworkers have isolated several mutants in rRNAs which are of special interest because they confirm the involvement of definite regions of rRNAs in the maintenance of accuracy of translation (61, 100, 102). These mutants act as general suppressors of nonsense codons and of +1 and –1 frameshift mutations. These are (i) mutant G517 (either deletion or base substitution) in the 530 region of E. coli 16S rRNA, which is probably involved in binding tRNAs with the A- and P-sites of the ribosome (102); (ii) mutant U2555 in 23S rRNA (substitution by G or A but not C)—this U residue is normally protected by the aminoacyl residue of A-site-bound tRNA (100); and (iii) mutant G2253 in 23S rRNA. This last residue probably interacts with the 3'-terminal CCA of P-site-bound tRNA (61).

 

Hungry Codons

Limiting aminoacyl-tRNA promotes frameshifting at certain sites which are called "hungry" codons. This phenomenon was first described in outline about a decade ago (83) and has since been extensively investigated by Gallant and coworkers (52, 83; for review see reference 53). The critical sequence for such a frameshifting event is a heptanucleotide which includes the hungry codon itself; the position of the hungry codon within the heptanucleotide determines the direction as well as the magnitude of the frameshifting. It has been found that at a hungry codon for lysine (AAG), the sequence GCCAAGC leads to rightward frameshifting, whereas the sequence CTTCAAG leads to leftward frameshifting. Thus, for frameshifting to the right, the critical heptanucleotide includes 3 bases to the left of the hungry codon and 1 to the right. In contrast, for frameshifting to the left, the critical heptanucleotide includes 4 bases to the left and none to the right of the hungry codon. The importance of the presence of 3 or 4 bases to the left of the frameshift site suggests that the adjacent P-site plays an important role in shifting. Since frameshifting at a hungry codon occurs only when the appropriate aminoacyl-tRNA is in limited supply, the hungry codon may dictate that the ribosome be stalled for lack of enough cognate aminoacyl-tRNAs. In fact, since release factors are much rarer than most tRNA molecules, ribosomes may also stall at termination codons. This raises the possibility that, at certain termination codons, ribosomes may occasionally shift reading frames instead of terminating. Studies by Weiss and colleagues (133) have shown that when termination codons are located immediately at the 3' side of certain slippery sequences (i.e., sequence of RNA at which ribosomes tend to get out of frame), frameshifting by ribosomes is indeed enhanced (134, 135).

 

REGULATORY IMPLICATIONS OF TRANSLATIONAL READ-THROUGH, FRAMESHIFTING, AND BYPASSING

As described above, a termination codon can be overcome at the translational level by three different processes. These processes, nonsense suppression, frameshift suppression, and hopping, have mainly been studied when a component of the translation apparatus was mutated. However, these same processes also occur naturally in unmutated cells when they are programmed into the mRNA sequence and/or structure. These natural processes are called translational read-through, frameshifting, and bypassing. They are interesting in relation to regulatory processes, since they provide regulatory mechanisms for gene expression by permitting a quantitative differential production of more than one polypeptide from a single mRNA. Furthermore, in these mechanisms, tRNAs and release factors compete for binding sites (23, 32). Thus, physiological conditions which would preferentially affect each of these competing elements would also affect regulation by translational read-through (41), frameshifting, and bypassing.

 

Translational Read-Through

UGA read-through was first discovered in the E. coli RNA phage Qβ (71, 91, 132). The major Qβ capsid protein, the coat protein, and the minor capsid protein IIb are translated from a common initiation site. A UGA codon signals the termination of the synthesis of the coat protein. When this UGA codon is occasionally read by a normal tRNA^Trp, the protein synthesis machinery continues on to synthesize the minor capsid protein IIb. Thus, UGA read-through provides a regulatory mechanism for the synthesis of these two capsid proteins in relative different amounts. An A residue is located at the 3' side of the UGA of the Qβ coat cistron (38). As described, the presence of an A residue at this position permits the greatest spectrum of options for UGA read-through by a natural tRNA, probably tRNA^Trp (38, 79). A TGA codon followed by an A residue is also located in several other E. coli genes and their phages, i.e., gene O of bacteriophage λ (121), gene rpsK for ribosomal protein S7 (106), and the trp leader, trpL (13), providing reason to believe that UGA read-through may be involved in their expression (38). The extended λO' protein has been shown to be synthesized in vitro (141) and S7' in vivo (106). In the case of the E. coli trp leader, two lines of evidence indicate that UGA read-through may have a role in trp attenuation for which the trp leader RNA is responsible. trp attenuation is increased by reducing the level of UGA read-through in the presence of the E. coli mutant rpsL (in ribosomal protein S12) (80). Conversely, transcription termination at the trp operon attenuator is increased by an increase in UGA read-through, caused by a reduction in the level of RF2 (116). UGA read-through is also permitted in E. coli by tRNA^Sec, which, together with SELB protein, recognizes an elaborate UGA codon context having a specific stem-and-loop structure (see Fig. 2a). According to the available data for the fdhF gene (which specifies for the selenocysteine-containing protein formate dehydrogenase H), the efficiency of this UGA read-through process is at most 25% (113, 143). It remains to be clarified whether there is a biological role for the 140-amino-acid-long product of the fdhF gene that is translated until the UGA codon that specifies for selenocysteine is reached.

 

Translational Frameshifting

The synthesis of RF2, which is encoded by the E. coli prfB gene, is a very interesting example of how translational frameshifting can provide an autogenous regulatory mechanism (29). On the one hand, the role of the RF2 protein requires it to recognize the UGA termination codon. On the other hand, synthesis of RF2 requires a frameshift event for overcoming the internal UGA codon located at position 26 of the prfB mRNA (Fig. 4a). This UGA codon forms the slippery site of a +1 frameshift event: since the alternative to prfB frameshifting is translation termination at this UGA codon, RF2 can regulate its own production through a choice between translation termination and frameshifting (Fig. 4b). When adequate RF2 is present, RF2 synthesis is prematurely terminated at the 26th UGA codon, and thus it limits its own production. On the other hand, when the level of RF2 is low, the frameshift mechanism can permit the increased expression of RF2. This model by Craigen et al. (29) is supported by more recent genetic evidence (86). As in other described programmed frameshifting events, in the case of RF2, frameshifting involves two cis elements in the mRNA (Fig. 4a). The first is the slippery site itself (CUU-UGA), at which the +1 frameshifting occurs (29). The second is a Shine-Dalgarno sequence (AGG-GGG) located several nucleotides upstream from the slippery site. Base pairing between the Shine-Dalgarno sequence and the 16S rRNA is required for RF2 frameshifting to occur (136). It has recently been shown that the CUU codon in the slippery site is particularly frameshift prone (31). The maximum level of prfB frameshifting is about 50% (28).

High-efficiency programmed translational frameshifting is also a regulatory device in the expression of the E. coli dnaX gene that codes for subunits γ and δ of DNA polymerase III (for review see reference 42). In this case, however, frameshifting provides no means for overcoming an in-frame termination codon. Instead, a UGA termination codon in the –1 frame causes the formation of the truncated γ subunit. In this case, the result of the frameshifting is a shortened rather than a lengthened protein. Thus, frameshifting may have yet further regulatory roles beyond the scope of this chapter.

 

Translational Bypassing

Translational bypassing is a most intriguing frameshifting process for overcoming a termination codon: rather than slipping over a single nucleotide or hopping over a few, the protein synthesis machinery bypasses a large segment of mRNA. Bypassing was first found in E. coli in the expression of bacteriophage T4 gene 60, which encodes a subunit of DNA topoisomerase (72, 137). In this case, the cellular translation apparatus bypasses a 50-nucleotide gap which results in a –1 frameshifting event. Programmed into the mRNA sequence, this frameshifting/bypassing event requires several elements including matched GGA (glycine) codons before and after the gap and a stop codon (in frame 0 immediately in the beginning of the gap) located in a stem-and-loop structure (Fig. 5a). The requirement for a matched set of codons at the 5' and 3' junctions of the gap suggests that a single tRNA reads the codons at either side of the gap, probably by a hopping mechanism (137). However, unlike the hopping over short gaps by E. coli hopping tRNAs (see above) in T4 gene 60, a large gap (50 nucleotides) is bypassed and the process is extremely efficient (100%). In the case of T4 gene 60, both the mRNA structure within the bypassed gap and the nascent polypeptide are stimulatory signals (55) which appear to be particularly specialized for the frameshifting/bypassing event. However, the exact mechanism of translational bypassing of T4 gene 60 still awaits elucidation.

Though the bypassing in the synthesis of T4 gene 60 appears to be 100% efficient (137), could the bypassing event have a regulatory function nevertheless? We suggest that the UAG termination codon located in frame 0 at the beginning of the untranslated region (Fig. 5a) could be a clue to such a regulatory function. This UAG termination codon is required for bypassing to occur. When this UAG codon is replaced by a sense codon, bypassing is drastically reduced (137), probably because this process requires a "paused ribosome." It seems that the efficiency of T4 bypassing may be a function of how well RF1 recognizes that UAG codon so that it can act as a termination signal. Thus, physiological conditions or cellular factors that would reduce the efficiency of recognition, like a reduced RF1 concentration or activity or an increase in competing suppressor tRNAs, would lower the efficiency of T4 bypassing to less than 100%.

Another example of the E. coli translation apparatus bypassing a large segment of an mRNA is the case of E. coli trpR. The trpR gene codes for the trp repressor (62), which regulates the biosynthesis of tryptophan and its transport (for review see reference 139). It has been shown that in the expression of a trpR ₊₁-lac'Z gene fusion a 55-nucleotide gap is bypassed, resulting in a +1 frameshifing event (8). The mechanism of trpR ₊₁-lac'Z frameshifting/bypassing differs totally from that of T4 gene 60 (39, 42). As shown in Fig. 5b, there is no pair of matched codons at the border of the gap, no essential secondary structure, no stop codon in frame 0 of the gap, and no requirement for the synthesis of a specific region of the nascent polypeptide. Instead, two adjacent cis elements are required: (i) a specific segment of 15 nucleotides of trpR, which must be preceded by (ii) a nonspecific 5' end longer than 10 translatable codons. In addition, only about 5% of the translating ribosomes bypass the gap in the +1 frame, while 95% continue in frame 0. Moreover, translation of the gap in frame 0 is essential for frameshifting/bypassing in that system (8). In summary, although the mechanism of this bypassing process has not yet been elucidated, it seems to be different from, and less specialized and less efficient than, that of T4 gene 60. Thus, E. coli trpR ₊₁-lac'Z bypassing may be representative of a more general translational control mechanism that enables the ribosomes to skip over a large segment of an mRNA. It is interesting and important to note that trpR frameshifting/bypassing can be modulated (10). The frequency of bypassing is inversely proportional to the rate of translation initiation; when the rate of translation initiation is low, the +1 and 0 frame products of trpR are synthesized in similar amounts.

Though as yet trpR bypassing has only been demonstrated through the use of a trpR ₊₁-lac'Z gene fusion, there is some evidence suggesting that a similar frameshifting/bypassing mechanism also occurs in the natural trpR gene itself when it is not fused to lac'Z mRNA. For example: (i) a 10-kDa +1 frameshift product of trpR has been identified by using peptide-directed antibodies in immunoprecipitation experiments (9); (ii) the frequency of the synthesis of both trpR ₊₁-lac'Z and the 10-kDa +1 trpR product is regulated by the level of translation initiation (10); and (iii) an E. coli mutant has been isolated in which trpR bypassing is specifically increased, such that the synthesis of both trpR ₊₁-lac'Z and trpR +1 products are increased by about threefold (R. Hazan, R. Schoulaker-Schwarz, and H. Engelberg-Kulka, unpublished data). Recently, a regulatory role in trp repression and in the biosynthesis of tryptophan has been suggested for the trpR ₊₁ bypassing product (42).

 

CONCLUSIONS AND PERSPECTIVES

Cells have two opposite requirements: accuracy and flexibility. Of the processes and components involved in the transfer of the genetic information from DNA to proteins, the translational apparatus seems to be the most flexible. Flexibility in the translation process probably depends on the diversity, plasticity, and complexity of the components involved. Much of the research done over the last decade has examined the flexible nature of the translation machinery. This flexibility is manifested through three basic properties that provide the central dogma of the classical genetic code (for reviews see references 2, 39, 42, 49, and 108). These are (i) the classification between sense and nonsense codons, (ii) maintenance of the reading frame, and (iii) the sequential reading of the mRNA. Three natural processes, translational read-through, frameshifting, and bypassing, can be seen as deviations from these "rules." These three processes are now considered as alternative ways for reading the genetic code (for review see references 43 and 108). In this chapter, we have discussed the existence of such alternatives only in E. coli and S. typhimurium. However, translational read-through and frameshifting, at least, reflect more general phenomena which also take place in yeasts, plants, and mammals and particularly in retroviruses and retrotransposons (for review see references 2, 16, 43, 66, 75, and 108). Specific cis elements, like the primary sequence and/or secondary or tertiary mRNA structures, are the means through which these translational processes are programmed in the mRNA (for review see reference 43). In the examples illustrated here, such cis elements are the slippery site itself and the Shine-Dalgarno sequence required for prfB frameshifting (136) and the specific stem-and-loop structure required for the UGA-directed selenocysteine incorporation (143). Specific cis elements are also involved in examples in E. coli of translational bypassing over a large segment of the mRNA (8, 39, 72, 137). In the case of the UGA-directed selenocysteine incorporation, specific trans elements are also required, such as the specific elongation factor SELB and the specific tRNA^Sec (for review see references 16 and 17). Thus, it is possible, and perhaps even likely, that other translational programmed alternatives for reading the genetic code require some specific trans elements as well.

The variety of processes for reading the genetic code provides a powerful means for increasing the ability of the cell to produce and to regulate the production of several proteins from a single gene. For some of these examples of translational read-through, frameshifting, and bypassing, it appears that a termination codon is an important regulatory element. Recall that there is an antagonistic relationship between the release factors and the tRNAs. The fact that the termination codon can be recognized by these competing elements, the release factor(s) and a specific tRNA, provides a subtle means of control for some of these processes. That the translational machinery is so flexible that genetic information can be transferred by several known alternative processes suggests that in both prokaryotes and eukaryotes additional genes will be found that are expressed and regulated through such processes. Translational bypassing is the most extreme representative of these processes. It has been suggested that a large bypassed segment of the mRNA be called "translational intron" (39). So far, translational bypassing has been only been found to occur in E. coli, operating by shifting from frame 0 into a different frame by skipping over a range of about 50 nucleotides (8, 72, 137). In any case, it has been suggested that a frameshift event may not be obligatory and, furthermore, that the size of the bypassed segment may vary and may be limited by the ability of the translational apparatus to skip over information carried by the mRNA (39).There are two ways of facilitating the detection of additional genes expressed and regulated by the alternatives in reading the genetic code: characterizing the nature of gene-encoded proteins by direct amino acid sequencing rather than according to the DNA sequence, and selecting for cellular mutants that specifically affect these novel mechanisms of gene expression. E. coli and S. typhimurium are particularly suitable for the second approach, which has several advantages: (i) the use of "up" mutations will facilitate the detection of products that are produced in low amounts, at least under usual growth conditions; (ii) trans elements involved will be more easily identified; and (iii) it will be easier to characterize both the translational components that permit this reading flexibility and those components required to maintain its accuracy. These two groups of components are opposing but necessary requirements for the transfer, regulation, and evolution of genetic information.

To date, the best-characterized and most outstanding alternative in reading the genetic code is the UGA-directed incorporation of selenocysteine in E. coli, where both cis and trans elements have been identified as required (for review see references 16 and 17). Based on this example and on other UGA-directed systems in E. coli, the similarities of the basic principles of the genetic code and of human languages can be further compared.

 

The UGA Codon: a Homonym

During the early 1960s, when the main rules of the genetic code were established, it was natural to compare them to the rules of human language. As a result, linguistic nomenclature has been borrowed to describe elements of the genetic code. For example, in linguistics, different words having the same meaning are called "synonyms," a term which is used to describe different codons specifying for the same amino acid. Similarly, here we shall borrow the linguistic term "homonym" for another characteristic of the genetic code. In linguistics, the term homonym describes two or more words which are identical in spelling but different in meaning (65). There are many examples in human languages of such homonyms like "rest" (remainder; relax), "like" (similar; love), and "fair" (beautiful; exhibition). In every case, the selection of the right meaning of the word is determined by its context in the sentence. Analogously, the UGA codon is a "bio-homonym": it is multifunctional, signaling for either RF2, tRNA^Trp, or tRNA^Sec, which will respectively cause either polypeptide chain termination or translation by either tryptophan or selenocysteine. In addition, as in language, the selection of the appropriate meaning of the UGA codon is determined by the codon context. The nature of the nucleotide at the 3' side of the UGA codon determines whether it will be recognized by either RF2 or tRNA^Trp, and the existence of a unique stem-and-loop structure immediately downstream from the UGA codon permits it to be recognized by tRNA^Sec with the subsequent incorporation of the modified amino acid selenocysteine. In addition to its unique UGA codon context, the incorporation into a polypeptide of selenocysteine, the 21st amino acid in an extended genetic code, also requires the specific translation elongation factor SELB (16, 17). The modification of amino acids is presumed to occur in the polypeptide chain at a posttranslational stage. In fact, as in the case of selenocysteine, some such modification events may occur at the translational level facilitated by the existence of specific cis and/or trans elements. Through further experimentation additional homonyms for UGA and/or other codons may be discovered, which would extend the genetic code even more.

ACKNOWLEDGMENTS

We thank Hanita Khaner and Ronen Hazan (Jerusalem, Israel) for graphical assistance, Myriam Reches (Jerusalem, Israel) for her comments, and F. R. Warshaw-Dadon (Jerusalem, Israel) for her critical reading of the manuscript. In addition, we are grateful to Rachel Kleiman (Jerusalem, Israel) for helpful linguistic discussions. This research was supported by grants from the German-Israel Foundation for Scientific Research and Development (G.I.F.) and by the endowment fund for Basic Research, Foundation in Life Sciences (The Dorot Science Fellowship Foundation), administered by the Israel Academy of Sciences and Humanities.