Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

For organisms using a standard genetic code, translation termination occurs when one of the three stop codons, UAA, UGA, and UAG, enters the ribosomal A-site. In contrast to the recognition of sense codons by tRNA, stop codons are recognized by extra-ribosomal proteins called class I release factors. The efficiency of translation termination depends on competition between the recognition of the stop codon by a class I release factors and the decoding of the stop codon by a near-cognate tRNA. In wild-type situations there is no cognate tRNA for decoding the stop codon; near-cognate tRNAs recognize stop codons with low efficiency. This leads to a very low background of stop codon suppression. However, a number of viruses (mostly plant RNA viruses) and several cellular genes display a high level of natural suppression at particular sequences, allowing two related polypeptides to be produced from a single mRNA. This phenomenon is called “programmed stop codon readthrough” and will be hereafter referred to as “readthrough.” Readthrough depends on particular mRNA sequences and structures. In most cases, the elements involved in readthrough efficiency are located close to the stop, although at least one element lying hundreds of nucleotides downstream from the stop has been described (see below). Elucidating the precise mechanisms by which the sequence context surrounding the stop codon can influence termination is a major challenge in understanding readthrough. There are two main ways that readthrough can be increased: either by stimulating stop codon decoding by a near-cognate tRNA or by limiting the access of release factors to the stop codon. Finally, the difference between “standard stop codon readthrough” and incorporation of selenocysteine and pyrrolysine should be noted (see Chapters 1, 2, and 3). Incorporation of selenocysteine and pyrrolysine requires a specific cognate tRNA to read the stop codon, helped by a secondary structure downstream from the stop codon and several dedicated factors.

2 Translation Termination and Programmed Stop Codon Readthrough

In eukaryotes, two release factors eRF1 and eRF3 mediate translation termination. Either full or partial X-ray structures are available for both proteins and provide interesting insight into their function (Fig. 4.1) (Kong et al., 2004; Song et al., 2000). The overall shape of human eRF1 is similar to a tRNA, with functional motifs in both the peptidyl transferase center and the decoding site. eRF1 recognizes all three stop codons through the NIKS motif in its N-terminal domain. It triggers peptidyl–tRNA hydrolysis by activating the peptidyl transferase center of the ribosome through the highly conserved GGQ motif in the eRF1 central domain. The C-terminal domain of eRF1 is involved in binding eRF3, which is essential for in vivo termination (Eurwilaichitr et al., 1999). The molecular mechanisms underlying this process remain unclear in eukaryotes. It is possible that the binding of eRF1 to eRF3 induces a conformational change in eRF3 to stabilize the binding of GTP to eRF3 (Pisareva et al., 2006). In Saccharomyces cerevisiae, eRF1 and eRF3 are encoded by the genes SUP45 and SUP35, respectively. The eRF3 protein is composed of two different domains. Its N-terminal domain (called NM), specific to S. cerevisiae, is not necessary for termination activity but is a major determinant of eRF3 aggregation, involved in the formation of prion-like polymers known as [PSI +]. The C-terminal domain is highly conserved through evolution and is involved in both GTP and eRF1 binding. The affinity of free eRF3 for GDP is 100 times greater than its affinity for GTP; eRF1 binding to eRF3 is thus required for GTP binding. A recent study suggested that eRF3 acts as a proofreading factor for the termination process (Salas-Marco and Bedwell, 2004). The ternary complex eRF1:eRF3:GTP may bind the ribosomal A-site, but the binding of GTP to eRF3 prevents eRF1 from catalyzing termination. If a stop codon is located in the A-site, a conformational rearrangement occurs and activates eRF3 GTPase activity. GTP hydrolysis allows proper positioning of the eRF1 GGQ motif in the peptidyl transferase center to catalyze peptidyl–tRNA cleavage.

Fig. 4.1
figure 1

Structures of both eukaryotic release factors. Left panel shows the structure of the carboxy-terminal domain of eRF3 (PDB 1R5N) bound to GDP (in red). Right panel corresponds to the structure of the human eRF1 (PDB: 1DT9). Two important motifs highly conserved between species are shown in red. The NIKS motif can be seen at the top of domain 1, which is thought to directly interact with the stop codon, and the GGQ motif is at the end of domain 2, involved in cleavage of the last peptidyl–tRNA. The lower part of the figure represents a rotation of 90° of the upper region. The pocket where GDP binds eRF3 is clearly visible

Full size image

The only available structure of eRF1 is unlikely to represent its functional form. Indeed, the distance between the NIKS and the GGQ motifs is 97.5 Å, 22 Å larger than the distance between the anticodon and the CCA acceptor stem in tRNAs. Thus, in this form eRF1 is too large to enter the ribosomal A-site. A bioinformatics model of the conformational changes induced upon ribosomal binding has been proposed, fitting more closely to the structural constraints of the ribosome (Trobro and Aqvist, 2007).

In prokaryotes, three factors are involved in the termination process: two class I factors, RF1 and RF2, which recognize UAG and UAA or UGA and UAA, respectively, and one class II factor called RF3. Many structural (Petry et al., 2005; Rawat et al., 2003) and mutational studies (Ito et al., 2000; Mora et al., 2003; Trobro and Aqvist, 2007; Zavialov et al., 2002) have provided essential information about the translation termination process. Recently, high resolution crystal structures of the translation termination complex bound to the bacterial ribosome was published (Laurberg et al., 2008; Weixlbaumer et al., 2008; Korostelev et al., 2008). These allows detailed observation of release factors frozen in the act of termination (Fig. 4.2a). In a structure illustrated, the stop codon is bound in a pocket formed by interactions between the conserved PxT motif (the equivalent of eukaryotic NIKS motif) and conserved nucleotides of the 16S rRNA decoding site (Fig. 4.2b). The codon and the 30S subunit A-site undergo induced-fit binding resulting in stabilization of an RF1 conformation that promotes the positioning of Gln 230 (of the GGQ motif) in the peptidyl transferase centre of the ribosome to allow its direct participation in peptidyl–tRNA hydrolysis (Fig. 4.2c). Before this recent publication, the only high-resolution structure of a prokaryotic release factor (RF2) described did not fit the cryoEM density observed with RF2 bound to the ribosome (Klaholz et al., 2003; Vestergaard et al., 2001), making it difficult to compare the structures of prokaryotic and eukaryotic class I release factors. The high resolution structure of the terminating ribosome allows this comparison to be made. Figure. 4.3 shows both structures aligned by their conserved GGQ motifs. The overall shape of the two factors is clearly very different; however, both functional motifs (NIKS (green) and GGQ (blue)) of the closed conformation proposed by Tobro and colleagues display a similar orientation of PXT/GGQ to RF1, whereas these motifs have a very different orientation in the x-ray structure (Fig. 4.3).

Fig. 4.2
figure 2

Structure of prokaryotic RF1 bound to the ribosome. (a) the P-site tRNA is indicated in orange and RF1 in yellow. As in the previous figure, the two highly conserved motifs are in red. The close proximity of the ends of the tRNA and RF1 in the peptidyl transferase center is clearly seen. (b) two close views of the interactions between RF1, the stop codon (green) and the rRNA (dark blue) The PVT motif of RF1 is indicated red, as well as the identity of the important nucleotides of the rRNA. (c) the glutamine of the GGQ motif is shown in red; the last adenosine of the CCA t-RNA domain is also shown. This close view of this region also highlights the various interactions between RF1 and this region of the tRNA

Full size image
Fig. 4.3
figure 3

Comparison between prokaryotic and eukaryotic class I release factors. From left to right: prokaryotic RF1 (yellow), human eRF1 (orange), and a bioinformatics-derived model of a closed conformation of eRF1 (red). The distances between the GGQ (cyan) and NIKS/PVT (green) motifs are indicated for each structure. By comparison, the distance between the anticodon and the tRNA CCA motif is 75 Å. All the structures are aligned to allow direct comparison of GGQ domains

Full size image

3 Readthrough in Viruses and Phages

The majority of recoding events identified so far (including frameshifting) has been found in viruses or phages. The aim of this review is not to provide an exhaustive view of all the known viral readthrough sites, but to provide insights into stop codon readthrough mechanisms and into the biological relevance of readthrough products. Many readthrough sites have been identified simply by sequence analysis (presence of an in-frame stop codon); we will therefore only focus on well-characterized readthrough sites, discussing their biological function where known.

One of the earliest examples of programmed readthrough comes from bacteriophage Qβ. This single-stranded positive-strand RNA phage encodes four viral proteins from its own genome. These proteins (coat protein, replicase, and maturation/lysis) are all involved in viral replication. However, only the maturase and the coat protein are found in the mature particle, with just one protein – rather than two separate proteins – exhibiting maturation and lysis activity. Qβ also encodes a fourth protein obtained by readthrough of the UGA stop codon of the coat protein. The molecular weight of this recoded protein, called IIb, is 22 kDa greater than the coat protein and accounts for about 5% of the amount of coat protein produced (Weiner and Weber, 1971). Trp tRNA probably reads this UGA. Readthrough results in a considerably elongated product that is incorporated into the virion and is essential for infectivity (Hofstetter et al., 1974). This minor form of the coat protein could thus play a role in the assembly of the mature particle, like the frameshifted products involved in tail assembly in phage lambda (Levin et al., 1993) (see Chapter 11).

The simplest readthrough motif is found in the Sindbis virus RNA. Indeed, a single cytidine residue immediately downstream from the termination codon is needed to give a readthrough efficiency of 10% (Li and Rice, 1993). However, the next residue (+2) also plays a role in readthrough efficiency, in both mammalian and insect cells (JPR and John F. Atkins, unpublished results).

Around one hundred plant RNA viruses use readthrough to express the replicase domain of their genome. The archetypal representative of this class is the tobacco mosaic virus (TMV), an RNA virus that infects more than 150 types of herbaceous, dicotyledonous plants including many vegetables, flowers, and weeds. The virus damages the leaves, flowers, and fruits leading to stunted plant growth. This virus uses UAG readthrough to produce its RNA-dependent RNA polymerase (Pelham, 1978) essential for its replication. A region of six nucleotides downstream from the stop codon promotes a high level of stop codon readthrough. Mutational analysis in plant cells has shown that this small unstructured region has the consensus motif CAR-YYA (Skuzeski et al., 1991). This motif can induce readthrough at all three stop codons. The near-cognate tRNA likely to mediate UAG readthrough in plants is Tyr tRNA. Tyrosine, lysine, and tryptophan (at a ratio of 4:2:1) have been found as products of a closely related readthrough site created by a nonsense mutation in the S. cerevisiae gene STE6 (Fearon et al., 1994). Several natural tRNAs are thus able to read the UAG stop codon in this sequence context. Other plant viruses use slightly different readthrough signals, most of which are not fully defined (Dreher and Miller, 2006). A second motif found in most plant RNA viruses displaying a readthrough event consists of two adenines located just 5′ of the stop. This motif is also associated with low termination efficiency in yeast (Tork et al., 2004). In the Luteoviridae, the cis-acting signals involved in readthrough are composed of two elements: a cytidine-rich repeat (CCNNNN)8–16 beginning about 20 nt downstream from the stop codon and a distal sequence found 700 nt downstream (Brown et al., 1996), which does not seem to exhibit significant secondary structure. This is a unique example of a stimulatory element located at such a distance from the site at which readthrough takes place. Deciphering the precise mechanisms involved would reveal the ways in which the ribosome can be forced to perform unconventional decoding.

More than a dozen animal viruses also use readthrough for replication (recode db). The Moloney murine leukemia virus (MoMulV), a retrovirus, produces the Gag-Pol polyprotein by readthrough of a UAG (Philipson et al., 1978). A frameshifting site is usually found at this position in most retroviruses (see Chapters 7 and 8). A glutamine tRNA reads the UAG codon with an efficiency of 5% (Yoshinaka et al., 1985). This maintains a precise ratio of Gag to Gag-pol protein for viral replication and in particular for nucleocapsid formation. Replacing the UAG codon by a sense codon leads to a defective viral cycle, whereas replacing UAG by another stop codon does not affect viral propagation (Jones et al., 1989).

The signal stimulating UAG readthrough in the MoMulV is far more complex than those found in either Qβ or TMV. Indeed, the sequence surrounding the UAG stop codon displays two alternative secondary structures: a pseudoknot and a stem-loop (Alam et al., 1999; Wills et al., 1991; Feng et al., 1992; Wills et al., 1994). The stop codon is located at the top of the stem-loop structure and comprises a 9 nt loop and a 14 nt stem. The stem overlaps the potential pseudoknot by 10 nt. The pseudoknot, but not the stem, has a stimulatory effect on readthrough, so when the stem-loop and the pseudoknot are present the level of readthrough is reduced by increasing the proportion of the stem-loop relative to that of the pseudoknot. This suggests that these two structures are in competition, but it is unknown if the proportion of time one structure forms at the expense of the other varies during the infective cycle with consequent modulation of readthrough efficiency and that a swap mechanism may regulate stop codon readthrough efficiency (Fig. 4.4a). The importance of this pseudoknot for UAG recoding is clearly established; in particular, nucleotides from loop 2 and from the spacer are important for readthrough efficiency (Wills et al., 1994). However, how the pseudoknot stimulates readthrough remains largely unknown. The pseudoknot probably induces a pause and/or a conformational change in the ribosome, but it is unclear how this is related to stop codon readthrough. A recent cryoEM analysis of a ribosome pausing at a −1 frameshifter pseudoknot (from IBV) provided structural information on the mechanisms of action of the pseudoknot. The pseudoknot blocks the ribosome during translocation, preventing a complete step to occur (Namy et al., 2006). The mRNA and tRNA are thus placed under tension, leading to a displacement and distortion of the translocating tRNA. These observations can be used to provide a model for the role of the MoMulV pseudoknot in stop codon readthrough. If this pseudoknot induces a pause in translocation at the same step as the IBV pseudoknot, the alteration of the ribosome structure would prevent eRF1 from entering the A-site, allowing more time for a near-cognate tRNA to read the stop codon. Obviously this model needs to be tested, but it could explain the importance of the pseudoknot in MoMulV readthrough. eRF1 may interact with the viral reverse transcriptase (RT) (product of the pol gene) (Orlova et al., 2003). This sequestering of eRF1 by the RT would stimulate stop codon readthrough, creating a positive feedback loop and increasing RT production (Fig. 4.4b). However, given that stop codon readthrough level is not modified during infection (i.e., whether in the presence or absence of RT) (Berteaux et al., 1991), the biological relevance of this observation remains unclear.

Fig. 4.4
figure 4

This could be part of a swap mechanism controlling stop codon readthrough efficiency. (b) simplistic representation of the MoMulV replication cycle involving GAG and POL protein production. When ribosomes terminate at the stop codon, only the GAG protein is synthesized, whereas stop codon readthrough allows synthesis of the GAG-POL fusion protein. In turn, the POL protein can interact with eRF1, depleting release factors and thus stimulating readthrough and increasing GAG-POL protein synthesis Stop codon readthrough of MoMulV. (a) representation of the two potential structures; the stem-loop encompassing the stop codon, and the downstream pseudoknot. Sequences involved in the pseudoknot formation are shown in blue. The 3′ part of the stem can also form the pseudoknot.

Full size image

As discussed above, viruses often use new strategies to regulate the level of their own proteins. Initially found frequently in viral decoding, recoding events are now known to occur in all the kingdoms of life. The biological function of readthrough has been investigated in a few cases, but unfortunately remains a mystery in many cases.

4 Biological Relevance of Stop Codon Readthrough in Cells

In prokaryotes, the most common programmed readthrough event involves the insertion of the non-standard amino acid selenocysteine at the UGA codon (see Chapters 1 and 2). The only reported case of incorporation of a standard amino acid in place of a stop codon is the synthesis of the adherent CS3 pilus of the enterotoxigenic Escherichia. coli CFA/II strain, requiring the production of a 104 kDa protein by stop codon readthrough. A suppressor glutamine tRNA is necessary for full expression at this site (Jalajakumari et al., 1989).

In eukaryotes, stop codon readthrough, other than selenocysteine insertion, has been described in yeast and Drosophila. A few years ago, we reported that the yeast gene PDE2, encoding the high affinity cAMP-phosphodiesterase, is subjected to stop codon readthrough. The readthrough product is targeted to the 26S proteasome (Namy et al., 2002). The stop codon readthrough is influenced both by the nucleotide context of the stop codon and by the environment. Readthrough levels are increased either in the absence of glucose or in the presence of the prion [PSI +]. One phenotype associated with this increased level of stop codon readthrough is increased thermosensitivity of the cells. This suggests that stop codon readthrough of PDE2 can modify yeast fitness under stress conditions. Given that cAMP is a major secondary messenger in the cell, readthrough of the PDE2 stop codon may also affect other biological functions. Notably, the presence of the prion [PSI +] affects the regulation of gene expression in yeast (see below).

In Drosophila melanogaster, the expression of at least three genes is subject to readthrough. All three corresponding proteins are involved in developmental processes. This raises interesting questions about the origin of these genes. The oaf gene (out at first) contains a first open reading frame, which encodes a protein of 322 amino acids, separated from a second ORF by a UGA stop codon. Readthrough of the UGA codon will add 155 amino acids to the product of the first ORF, resulting in a protein of 477 amino acids. OAF is produced in nurse cells during oogenesis and is widely distributed throughout embryonic development (Bergstrom et al., 1995). This protein is found in the CNS but mutant larvae do not exhibit any obvious nervous system defects. However, some homozygous oaf mutants display peripheral nervous system defects with a penetrance depending on the mutant tested. These observations suggest that oaf is necessary for proper neuronal development and hatching. Homozygous mutants show a lethal phenotype late in embryogenesis or early during the first larval instar, including those showing no CNS defects. This suggests that additional roles of OAF remain to be identified.

The hdc gene (headcase) has a 3241 nt ORF interrupted by an in-frame UAA codon at position 2981 (Steneberg et al., 1998). The short and long polypeptides are both efficient in inhibiting terminal branching in the trachea. However, the longer product is more efficient than the peptide from ORF1 alone (Steneberg and Samakovlis, 2001). A secondary structure is predicted downstream from the UAA codon. This structure is sufficient to induce readthrough whatever the identity of the stop codon (Steneberg and Samakovlis, 2001). This, together with MoMulV, is the only known example with secondary structure involved in stop codon readthrough. This stem-loop structure has a similar shape to the PYLIS structure involved in the incorporation of pyrrolysine in archaea (Namy et al., 2004; Namy et al., 2007) (and see Chapter 3). It could be speculated that a rare amino acid would be inserted at this position in the hdc protein.

Kelch is an essential protein for the production of viable eggs in Drosophila ovaries. It is a structural component of the ring canals that act as intercellular conduits through which cytoplasm is transported from nurse cells to the oocyte in an egg chamber. The gene encodes a 76 kDa protein from one open reading frame (ORF1; 689 aa) and a 160 kDa product (ORF1 + ORF2; 782 aa) from the same mRNA. This stop codon readthrough is conserved among Drosophila species. The ratio of long to short product is regulated during embryonic development to give a maximum ratio during metamorphosis (Robinson and Cooley, 1997). The amino acid incorporated at the UGA codon is not known, but the reduced activity of a mutant generated by a deletion of the in-frame UGA indicates that this amino acid is important for the proper function of the protein. However, expression of the ORF1 is sufficient for Drosophila development. By contrast to hdc, the UGA stop codon is essential and cannot be replaced by another stop codon. This strongly suggests that both genes use different readthrough mechanisms to suppress the stop codon. It has been proposed that a selenocysteine is incorporated at the site of the Kelch readthrough event. However, this is unlikely given that no selenocysteine insertion sequence (SECIS) is found in the 3′ UTR and no 75Se incorporation has been demonstrated (Robinson and Cooley, 1997). Alternatively, mRNA editing could take place at this stop codon. RNA editing involves the adenosine deaminase (ADAR) enzyme, which catalyzes the deamination of adenosine to inosine (Bass, 2002). Due to its base-pairing properties, inosine is recognized as a guanosine. This modifies the genetic information carried by the mRNA, changing a stop codon to a sense codon.

As we can see, the biological relevance of stop codon readthrough in Drosophila remains to be clearly determined.

5 Identification of Readthrough Sites in Genomes

Several genomes have been screened extensively for recoding events, namely frameshifting and selenocysteine incorporation (Castellano et al., 2008; Kryukov et al., 1999; Lescure et al., 1999; Mix et al., 2007). There is no doubt that incorporation of a standard amino acid during stop codon readthrough can be used as a regulatory mechanism (Fujita et al., 2007). However, such events have received much less attention, probably because they are far more difficult to identify due to the absence of a consensus motif.

Two approaches are commonly used to identify readthrough sites:

  1. (i)

    searching for readthrough motifs within a particular genome. In S. cerevisiae, this approach led to the identification of genes with inefficient stop codons (Namy et al., 2002; Williams et al., 2004). A given 3′ readthrough motif is probably specific to a subset of organisms, so the nature of this motif is likely to change from an organism to another. This method of identifying new programmed readthrough genes is powerful but needs prior systematic analysis to identify inefficient termination sequence context.

  2. (ii)

    searching for genes with expression controlled by stop codon suppression, without a priori knowledge of the sequences involved. This approach has been developed for the S. cerevisiae genome (Namy et al., 2003) and inefficient termination signals have been searched for in the Oryza sativa genome (Liu and Xue, 2004). Other methods developed to identify rare amino acids incorporated at a stop codon could also be used to identify programmed readthrough sites (Fujita et al., 2007). These methods do not require prior knowledge about the motifs or the mechanisms involved. It is therefore an efficient method of identifying new recoding events involving a stop codon. However, the main challenge remains the biological validation of the candidates identified by bioinformatics methods. Indeed, many factors can influence the quality of the results. In prokaryotes, the presence of operons with many ORFs separated by a single stop codon leads to a high number of false-positive candidates. In eukaryotes, the complexity of genome readout is a major challenge, with alternative splicing making it difficult to search for recoding sites. A clear example is provided by the large-scale analysis of 12 Drosophila genomes (Lin et al., 2007). This study identified 149 new candidates for stop codon readthrough. The analysis took into account conservation of amino acids downstream from the stop. Indeed, sequences that do not encode a protein tend to be less conserved. Similarly, the ENCODE project identified a number of potential candidate readthrough sites (Birney et al., 2007). However, it is difficult to determine the biological significance of these observations without functional validation. A large proportion of these genes may use alternative splicing, mRNA editing or a different recoding event (i.e. hopping?). For all these approaches, the results obtained must be considered with caution until the genes identified are characterized further.

Despite these considerations, the main problem, common to all genomes, in identifying readthrough sites seems to be the quality of the DNA sequence. Indeed, a search for readthrough sites cannot rely only on the presence of an in-frame stop codon because of the high frequency of sequencing errors. These sequencing errors are almost indistinguishable from a bona-fide programmed readthrough site in a coding sequence. Even incorporating other criteria in the analysis like protein motifs or coding sequence probability (HMM) – does not solve the problem.

As mentioned above, the identification and characterization of programmed readthrough sites give significant insight into cell physiology and could lead to the discovery of new pathways regulating translation. However, readthrough sites and recoding events in general are also interesting, providing powerful tools for the study of ribosome function.

6 Programmed Readthrough as a Tool to Study Translation Termination

A powerful approach for deciphering the molecular mechanisms underlying a phenomenon is to identify “exceptions to the rule” and look for defects in the regulation and function of regular components behaving in a non conventional way. Programmed stop codon readthrough is very useful in studying translation termination because it decreases the natural high termination efficiency. As mentioned above, the analysis of cis-acting sequences involved in readthrough has allowed further characterization of the role of surrounding nucleotide sequence in controlling termination efficiency of the stop codon.

The analysis of stop codon context in a large number of organisms clearly shows a bias. In the sequence 5′ of the stop codon, this effect may depend on the nature of the P-site tRNA pairing with the adjacent codon, on the amino acid present on this tRNA, or on the nucleotides present in the mRNA. It is difficult to distinguish between these possibilities due to the intricacies of these signals. One study has shown P-tRNA structure to influence stop codon readthrough efficiency (Smith and Yarus, 1989). The authors suggest that interactions take place with the release factor through the anticodon loop of the P-tRNA. This is consistent with recent structural data showing direct contact between RF1 and the P-tRNA (Figs. 4.2a and b). However, the effect exerted by the 5′ nucleotide sequence is not limited to interactions between the P-tRNA and RF1. In E. coli, the last amino acid on the P-tRNA can interfere with the termination process (Björnsson et al., 1996; Mottagui-Tabar and Isaksson, 1997; Mottagui-Tabar and Isaksson, 1998). These initial observations in E. coli have been extended to other organisms (Arkov et al., 1995; Bonetti et al., 1995; McCaughan et al., 1995).

The role of 3′ nucleotides remains unclear despite many attempts to elucidate their potential function. A strong bias has been found at position +1 following the stop codon in both prokaryotes and eukaryotes (Brown et al., 1990; Cridge et al., 2006; Poole et al., 1995). The importance of the 3′ nucleotide sequence is not limited to the nucleotide immediately following the stop codon; the signal can be extended up to six nucleotides downstream from the stop. To identify efficient readthrough sequences, we set up a SELEX-like screen in yeast. This allowed us identify the motif CAR-NBA as the readthrough consensus 3′ motif. This general motif (including the plant CAR-YYA motif), first identified in the TMV readthrough site, is functional in S. cerevisiae. It is also sufficient to drive a high level of stop codon readthrough in mammalian cells (Cassan and Rousset, 2001). Many hypotheses have been proposed to explain the role of the 3′ sequence (Bonetti et al., 1995; Cassan and Rousset, 2001; Namy et al., 2001; Skuzeski et al., 1991; Tate and Mannering, 1996) but none of them have been confirmed.

Since readthrough sites are generally located in the middle of the mRNA, the stop codon is equivalent to a non-sense mutation appearing in a normal gene; the stop can thus be viewed as a premature termination codon. Premature termination codons have recently received considerable interest due to their being potential therapeutic targets (see Chapter Chapter 6). Underlying this interest in therapeutic benefit, a more fundamental issue needs to be addressed, to determine whether a premature termination codon behaves like a standard stop codon. In eukaryotes, when a ribosome encounters a premature termination codon, the “Nonsense mRNA decay” (NMD) pathway is activated, leading to the decapping and rapid degradation of the mRNA, whereas a stop codon in its normal position does not induce mRNA decapping (Frischmeyer et al., 2002; Mitchell and Tollervey, 2003). Recent studies have shown that eRF3 can bind either UPF proteins or PABP bound to the polyA tail (Ivanov et al., 2008; Singh et al., 2008). Thus, two termination complexes should exist, one of them able to activate the NMD pathway. In this case, we would expect signals to be present on the mRNA, indicating whether the stop codon is a premature termination codon or not. Moreover, if the nucleotide context of stop codons is under selective pressure to confer a high termination efficiency, the premature termination codon is unlikely to be in an appropriate nucleotide context for termination. One major factor determining NMD activation is the distance between the premature termination codon and the 3′UTR/ polyA tail (Amrani et al., 2004; Silva et al., 2008). However, as mRNAs are always highly folded, it is unclear how the distance between the premature termination codon and the 3′UTR/polyA tail is measured by the cell. Stop codon readthrough directly affects NMD efficiency; indeed, a threshold level of stop codon readthrough antagonizes NMD in yeast (Keeling et al., 2004) and in human cells (Allamand et al., 2008).

Identifying the precise mechanisms by which the local nucleotide context can influence the balance between release factors and suppressor tRNA remains a major challenge. The analysis of readthrough signals will be very helpful in elucidating the function of these nucleotides.

Programmed readthrough is also useful in deciphering the role of trans-acting factors in termination. One major question rarely addressed is the identity of the tRNA reading the stop codon. In eukaryotes, several naturally occurring cytoplasmic tRNAs have been shown to recognize stop codons involved in programmed translational readthrough events (Beier and Grimm, 2001; Lecointe et al., 2002). In all cases, stop codon recognition implies non-orthodox base pairing between the second or the third base of the anticodon and the first or second base of the codon. The ability of these tRNAs to compete with release factors by reading a stop codon depends largely on their modified nucleotide content, particularly in their anticodon branch (Beier and Grimm, 2001). The study of viruses has allowed the identification of several new tRNAs that use different modified nucleotides for decoding. For example, the plant tRNAArg is a natural suppressor of UGA in the PEMV (pea enation mosaic virus) (Baum and Beier, 1998). As discussed above, the tRNATyr decodes the UAG codon in the TMV. Two modified nucleotides have been shown to play an important role for UAG readthrough: (i) Y35 in the middle of the anticodon is a major determinant for UAG readthrough by tRNATyr; (ii) The queosine modification at position 34 of the same tRNA counteracts the effect of Y35 (Zerfass and Beier, 1992). Surprisingly, the absence of pseudouridinylation at positions 38 or 39 of the anticodon branch decreases readthrough efficiency of a programmed stop codon in S. cerevisiae (Lecointe et al., 2002). Indeed, deletion of the PUS3 gene responsible for these modifications affects readthrough of all three stop codons in the TMV. Interestingly, all three known near-cognate tRNAs able to decode the stop codon in S. cerevisiae (tRNATrp, tRNATyrand tRNALys) harbor a pseudouridine at position 39. It is thus possible that this modification allows a stronger interaction between the codon and the anticodon, which is particularly important for decoding a stop codon. This modification also increases +1 translational frameshifting efficiency (Lecointe et al., 2002).

As described above, stop codon readthrough can be stimulated by modifying the decoding efficiency of the suppressor tRNAs. Alternatively, readthrough can be increased by decreasing the efficiency of release factors. One way to achieve this is to modify the concentration of either eRF1 or eRF3. In S. cerevisiae, [PSI +] is the prion form of eRF3. The conformational change impairs eRF3′s termination activity through aggregation of free active molecules to form inactive polymers. This consequently increases stop codon readthrough, resulting in the production of proteins with carboxy-terminal extensions (Serio and Lindquist, 1999). Although termination is affected at all the stop codons, stop codons in a weak termination context are more sensitive to the presence of [PSI +]. These conditions thus provide a weak termination background in which to study the incorporation of near-cognate tRNAs. Moreover, [PSI +] has interesting physiological functions. Indeed, many phenotypes, dependent on the genetic background of the host, are associated with the [PSI +] status of the cell (Namy et al., 2008; True and Lindquist, 2000). These phenotypes reflect the broad effects of modifying termination efficiency on yeast physiology. Indeed, these phenotypes may be connected to the general disruption of the yeast proteome. However, we have recently shown that [PSI +] increases frameshifting at the programmed shifty stop present in the ornithine decarboxylase antizyme-encoding gene (Namy et al., 2008). Frameshifting in turn stimulates the expression of the antizyme, which negatively regulates the ornithine decarboxylase, leading to a general decrease of polyamines in cells. This reduction of polyamine concentration is responsible for about half of the [PSI +]-related phenotypes.

7 What’s Next? Remaining Questions and Objectives

As mentioned above, the identification of programmed readthrough sites using bioinformatics is very challenging, because it is very difficult to distinguish bona-fide readthrough sites from sequencing errors. Although the number of sequenced genomes in the database is rapidly increasing, there are unfortunately no efficient approaches in place to identify these sites. Moreover, in the most favorable cases, automatic annotations systematically indicate such recoding events as pseudogenes and in some cases the sequence is corrected to get rid of the stop codon. Correction of the sequence in such cases leads to a loss of the primary information, making it impossible to study the potential existence of a translational recoding event. A better understanding of the mechanisms of stop codon readthrough would help to identify these sites. Clearly, in addition to searching for in-frame stop codons, other factors such as the presence of protein motifs or conservation among different species also need to be considered; however, this does not allow the specific identification of translational readthrough events.

The study of readthrough will improve our understanding of translation termination mechanisms, which remain largely unknown (summarized in Fig. 4.5). The role of the GGQ motif is still a matter of debate. Indeed, it has been suggested that the glutamine coordinates a water molecule in the peptidyl transferase center (Heurgue-Hamard et al., 2005; Song et al., 2000). However, glutamine may play a more direct role in the hydrolysis of the last peptidyl–tRNA (Seit-Nebi et al., 2001). This glutamine residue is methylated both in prokaryotes and in eukaryotes (Figaro et al., 2008). Although this modification is necessary for regulating termination efficiency in prokaryotes, its role in eukaryotes is unclear. Very little is known about the intramolecular interactions between the different domains and their role in the activity of class I release factors. As discussed above, it is highly likely that eRF1 undergoes a large conformational change upon binding to the ribosome. Recent high-resolution NMR structure shows the overall folding of eRF1 to be similar in solution and in the crystal. However, there are noticeable differences in the GGQ loop between the two structures (Ivanova et al., 2007). It is therefore possible that the crystal conformation of eRF1 does not represent a biologically relevant conformation.

Fig. 4.5
figure 5

The remaining questions on translation termination. Schematic representation of a P-site tRNA and an A-site stop codon being decoded by eRF1. The main questions are summarized in the boxes

Full size image

In prokaryotes, RF3 binding induces conformational changes in the ribosome, breaking the interactions between RF1/2 and the decoding and peptidyl transferase centers and thus leading to the release of the class I release factor (Gao et al., 2007). The function of eRF3 is unclear. This factor seems to play a very different role from its prokaryotic counterpart. As recently suggested, the binding of eRF1 to eRF3 may induce a conformational change in eRF3 to stabilize the binding of GTP to eRF3 (Pisareva et al., 2006). eRF3 might also trigger eRF1 conformational changes to couple stop codon recognition and peptide release (Fan-Minogue et al., 2008).

Recognition of the stop codon by release factors has been the subject of many studies. A major challenge was to understand how these factors distinguish between a stop codon like UGA and a sense codon (UGG) with such high efficiency (Chavatte et al., 2003). Organisms that use an alternative genetic code have been very useful in studies of eRF1 regions involved in distinguishing between the different stop codons (Alkalaeva et al., 2006; Kervestin et al., 2001; Lekomtsev et al., 2007; Salas-Marco et al., 2006). Several models have been proposed to explain how the release factor binds the stop codon. The first model proposed was the tripeptide anticodon, based on the similarity between class I release factors and tRNA (Nakamura and Ito, 2002). A cavity model was later proposed by Stansfield’s group (Bertram et al., 2000) and has gained support more recently (Fan-Minogue et al., 2008). Moreover, this proposal is consistent with the recently published structure of RF2 bound to the ribosome (Laurberg et al., 2008) (see Fig. 4.2b).

Termination efficiency is modulated by the local nucleotide context of the stop codon. However, the molecular mechanisms underlying these observations have not been determined. This is currently a major challenge facing researchers in the field. These nucleotides also modulate the action of aminoglycoside antibiotics, strongly limiting their use in therapeutics protocols for “stop codon diseases” (see Chapter 6). We believe that the identification of new natural readthrough sites, together with further analysis of the role of the nucleotides surrounding the stop codon, will help to understand these striking observations. Last but not least, interactions between class I release factors with rRNA or with the P-tRNA probably play an important role in stop codon readthrough (Poole et al., 2007). Currently, this translational step requires eukaryotic structural data to complement the elegant and numerous biochemical and genetic analyses performed so far. The analysis of stop codon readthrough mechanisms should be very useful in addressing all these issues.