Halting a cellular production line: responses to ribosomal pausing during translation


  • J. Ross Buchan,

    1. School of Medical Sciences, University of Aberdeen, Institute of Medical Sciences, Foresterhill, Aberdeen, Scotland AB25 2ZD, U.K.
    Search for more papers by this author
    • Department of Molecular and Cellular Biology, Life Sciences South Building, Room 414, 1007 E. Lowell Street, Tucson, AZ 85721-0106, U.S.A.

  • Ian Stansfield

    Corresponding author
    1. School of Medical Sciences, University of Aberdeen, Institute of Medical Sciences, Foresterhill, Aberdeen, Scotland AB25 2ZD, U.K.
    Search for more papers by this author

(email i.stansfield@abdn.ac.uk).


Cellular protein synthesis is a complex polymerization process carried out by multiple ribosomes translating individual mRNAs. The process must be responsive to rapidly changing conditions in the cell that could cause ribosomal pausing and queuing. In some circumstances, pausing of a bacterial ribosome can trigger translational abandonment via the process of trans-translation, mediated by tmRNA (transfer—messenger RNA) and endonucleases. Together, these factors release the ribosome from the mRNA and target the incomplete polypeptide for destruction. In eukaryotes, ribosomal pausing can initiate an analogous process carried out by the Dom34p and Hbs1p proteins, which trigger endonucleolytic attack of the mRNA, a process termed mRNA no-go decay. However, ribosomal pausing can also be employed for regulatory purposes, and controlled translational delays are used to help co-translational folding of the nascent polypeptide on the ribosome, as well as a tactic to delay translation of a protein while its encoding mRNA is being localized within the cell. However, other responses to pausing trigger ribosomal frameshift events. Recent discoveries are thus revealing a wide variety of mechanisms used to respond to translational pausing and thus regulate the flow of ribosomal traffic on the mRNA population.

Abbreviations used:

elongation factor


Echinococcus granulosus fatty-acid-binding protein


initiation factor


multidrug resistance


no-go decay


open reading frame


release factor


RNA polymerase


transfer—messenger RNA


upstream ORF


untranslated region


The process of protein synthesis lies at the heart of cellular metabolism and is required for the conversion of genetic information from mRNA into protein. As a complex biochemical polymerization process, translation is analogous to a factory production line. As the ribosome moves along the mRNA line, tRNA ‘assembly-line workers’ bring new amino-acid components to the growing protein. As with any production line, translation must be a robust process, accommodating unexpected pauses, and longer interruptions that could cripple productivity should elicit mechanisms to clear stalled complexes. Recent discoveries in the translation field have revealed how the ribosome deals with such interruptions. In this review, the new data are considered alongside previous understanding of how stalled translation intermediates are cleared in both eukaryotes and prokaryotes. As we describe below, such comparisons identify both similarities and differences in the way different domains of life have evolved to solve these biochemical blocks.

Pausing during elongation

The rate at which elongating ribosomes translate through ORF (open reading frame) sequences is codon-specific and in Escherichia coli varies from approx. 5 to 21 codons per s (Sorensen and Pedersen, 1991). This rate variation is partly determined by the properties of the tRNA itself; for example, GAA and GAG codons are decoded by the same E. coli tRNA, but at a 3-fold different rate (Sorensen and Pedersen, 1991). A second component of the translation rate is determined by the abundance of a given tRNA. Ribosomes paused at rare codons are on average forced to wait longer for the correct cognate, and correspondingly rare, tRNA. There are other cis-elements within the mRNA that affect ribosome transit rates; the presence of mRNA secondary structures can temporarily stall ribosomes while they attempt to unwind and translate through them. Finally, ribosomal elongation may also be stalled due to interactions between the nascent peptide and the ribosome occurring within the peptide tunnel.

Elongation pauses can have three distinct consequences for a translating ribosome. Elongation may resume as normal following a pause. Alternatively, delay may cause dissociation of tRNA from the mRNA within the ribosome, followed by movement of the ribosome relative to the mRNA and rebinding of the tRNA(s) elsewhere on the mRNA. Translation then continues with the production of a polypeptide distinct from that specified by a canonical decoding of the mRNA. The third and most severe fate is that translation is abandoned, with ribosomes leaving the mRNA (ribosome ‘drop off’), often followed by decay of the mRNA and the partly completed protein product. Examples of these distinct outcomes are described below, using examples from viral, prokaryotic and eukaryotic systems. In some cases the different mechanistic consequences of pausing are emergency responses to a potentially damaging metabolic block, whereas in others they have important regulatory roles in the cell.

Simple delays in elongation with resumption of translation

Pausing of elongating ribosomes caused by low tRNA availability is almost certainly a relatively frequent occurrence. In some circumstances this is part of a finely-tuned gene regulatory mechanism, known as transcription attenuation. This is employed within the bacterial operon leader sequences that regulate production of amino acid biosynthetic genes. In the case of the E. coli trp operon, the leader sequence contains an AUG initiation codon followed by trpL, a 15 codon ORF which includes region I (Figure 1) (Oxender et al., 1979). This ORF contains two key regulatory tryptophan-encoding codons. Next follows region II, followed by a pause site for the RNAP (RNA polymerase). Further downstream, regions III and IV are found. Region I can base-pair to form a stem loop with region II, and region III can form mutually exclusive stem loops with either region II (the anti-terminator stem loop) or IV (terminator stem loop). When the RNAP reaches the pause site, it stops and waits until a translating ribosome begins ‘melting’ the region I—region II stem loop (Winkler and Yanofsky, 1981; Fisher et al., 1985; Landick et al., 1985, 1987). At this point the RNAP is ‘synchronized’ with the elongating ribosome, and resumes transcription. A race is then effectively begun between the RNAP and elongating ribosome to determine whether transcription continues or is aborted. If tryptophan levels are low in the cell, availability of tryptophan-charged tRNA will also be low, leading to stalling of the ribosome at the key tryptophan-encoding codons in region I (Figure 1). If this occurs, RNAP finishes transcription through region IV before the ribosome has reached region II, causing the anti-terminator stem loop to form instead of the terminator stem loop. This allows RNAP to continue transcribing into the biosynthetic operon, allowing expression of tryptophan biosynthetic enzymes (Yanofsky and Horn, 1994; Elf and Ehrenberg, 2005). In contrast, when tryptophan levels are plentiful, ribosomes will translate quickly through region I and into region II, driving formation of the terminator stem loop by the time RNAP finishes transcribing region IV. This leads to immediate Rho-independent transcription termination (von Hippel, 1998; Elf and Ehrenberg, 2005).

Figure 1.

Transcription attenuation during expression of the E. coli trp operon

A representation of the attenuation process occurring during coupled transcription—translation of the trp operon, showing events in the instance of tryptophan scarcity and abundance. The regions I—IV are shown as open boxes within the transcripts. The black stars denote the tryptophan UGG codons within the trpL short ORF lying within region I (see text for details).

Another example of a beneficial elongation delay is found in the yeast Saccharomyces cerevisiae ASH1 gene. This encodes a transcription factor that is specifically localized to the tip of daughter cells during budding, resulting in inhibition of mating-type switching (Long et al., 1997). In the transcribed mRNA, four secondary structures, three of which lie within the ORF, act as docking sites for the protein She2p. She2p links via an adaptor protein to the myosin motor protein Myo4, which transports the mRNA—protein complex along actin filaments to the daughter cell (Bertrand et al., 1998; Beach et al., 1999; Bohl et al., 2000). Intriguingly, these secondary structures also appear to delay translation elongation and hence completion of protein synthesis until the mRNA is correctly localized, since, although relocating these secondary structures to the 3′ UTR (untranslated region) does not affect mRNA localization, it does result in Ash1p being erroneously expressed in the mother cell, as well as the daughter cell (Chartrand et al., 2002). Supporting this assertion, if translation initiation is inhibited on ASH1 mRNA containing the secondary structures shifted into the 3′ UTR, proper Ash1p localization is restored. Furthermore, in vitro and in vivo studies confirmed a ∼3-fold lower rate of protein synthesis using a wild-type ASH1 mRNA template compared with translation of the mutant ASH1 mRNA containing disrupted versions of the secondary structures (Chartrand et al., 2002).

One intriguing idea with some support is that, in certain instances, pauses in translation elongation may be deliberately selected to assist the folding of nascent protein domains. One of the first studies to suggest this involved replacement of 10 contiguous rare codons within an enzymatic domain of the S. cerevisiae TRP3 gene, with more abundantly used synonymous codons. Thus, although the encoded peptide remained identical, codon usage, and by implication tRNA decoding rates, were significantly altered in this region. Trp3p-specific protein activity reproducibly decreased 1.5-fold with the use of common codons (Crombie et al., 1992, 1994). Similarly, replacement of rare codons in the CAT (chloramphenicol acetyltransferase) gene with common synonymous codons at predicted pause sites led to reproducible decreases in specific enzyme activity of 20–30% in E. coli extract cell-free systems (Komar et al., 1999; Ramachandiran et al., 2002). In another instance, replacement of three consecutive rare codons in the EgFABP1 (Echinococcus granulosus fatty-acid-binding protein) gene with common synonymous codons led to an increased production of insoluble (and thus, by implication, unfolded) protein by 30% compared with wild-type EgFABP1 when expressed in E. coli (Cortazzo et al., 2002). The same silently mutated allele of EgFABP1 also caused a 6-fold increase in the expression of GroEL versus wild-type, as assessed by a lacZ reporter under control of GroEL promoter sequences. This is characteristic of an unfolded protein stress response. Finally, a recent study in human cell lines has shown that an allele of the MDR1 (multidrug resistance 1) gene with a silent synonymous mutation caused the encoded P-glycoprotein to fold differently from the wild-type gene. No differences in mRNA or protein levels were observed. The misfolding resulted in the P-glycoprotein encoded by the mutant MDR1 gene displaying an altered substrate specificity and sensitivity to inhibitor compounds. It was proposed the altered conformation resulted from mutation-induced differences in the timing of co-translational folding (Kimchi-Sarfaty et al., 2007). Codon usage may therefore be carefully selected so that certain parts of an mRNA are translated at differential rates, optimized for synthesis and folding of that particular protein.

The relative abundance of any given tRNA isoacceptor, determined largely by its gene-copy number, is an important factor determining translation of its cognate codon. Indeed, it has long been noted that in many microbial species, highly expressed genes make preferential use of codons decoded by highly abundant tRNAs (a phenomenon termed ‘codon bias’; Ikemura, 1981, 1982; Dong et al., 1996; Berg and Kurland, 1997). Codon bias has historically been explained as the result of selection to optimize the speed and fidelity of translation, thus stimulating protein expression, although probably rates of ribosome recruitment to an mRNA have a more influential effect on protein production than individual ribosome transit times (Vind et al., 1993).

One area where the importance of tRNA supply and codon demand is highlighted is under starvation conditions, when the fraction of tRNA isoacceptors charged with an amino acid can decrease, leading to lower tRNA decoding rates and a consequent increased likelihood of missense or frameshifting errors (Sorensen, 2001). Additionally, recent studies using E. coli have revealed that during amino acid starvation, different tRNAs within a given isoacceptor family can become uncharged at different rates determined by each tRNA's supply-to-demand codon ratio. For some tRNA families, the more abundant tRNAs are those that become exhausted first and are therefore more starvation sensitive, whereas rare tRNAs often retain a degree of starvation insensitivity (Elf et al., 2003; Elf and Ehrenberg, 2005; Dittmar et al., 2005). Examination of biological contexts, such as the aforementioned bacterial biosynthetic attenuation mechanisms, support these findings, since the E. coli regulatory codons upon which ribosomes pause within attenuation leaders in response to starvation are codons decoded by tRNAs predicted to be particularly starvation-sensitive (Elf et al., 2003; Elf and Ehrenberg, 2005). The resulting slow decoding gives sufficient time for the anti-terminator secondary structure to form, thus allowing transcription to continue. In contrast, the ORFs of amino acid biosynthetic enzymes themselves, where efficient translation is required to enable an amino acid biosynthetic response, make greater than expected usage of codons decoded by starvation-insensitive tRNAs (Elf et al., 2003; Elf and Ehrenberg, 2005).

Translation elongation can also be delayed and even arrested by nascent peptide sequences. Indeed, there are several examples where elongating ribosomes translating small uORFs (upstream ORFs) are stalled at or near the uORF termination codon through action of the nascent peptide. This prevents subsequent initiating (scanning) ribosomal subunits from reaching the main downstream ORF. For example, the AAP (arginine attenuator peptide), encoded by a uORF preceding the S. cerevisiae CPA1 and Neurospora crassa arg-2 genes, whose products function in arginine biosynthesis, stalls ribosomes at the uORF stop codon in response to high intracellular levels of arginine (Wang and Sachs, 1997a, 1997b; Gaba et al., 2001). Similarly, the gene encoding mammalian AdoMetDC (S-adenosylmethionine decarboxylase), which forms part of the polyamine biosynthetic pathway in most eukaryotes, stalls ribosomes in response to raised intracellular concentrations of polyamines (Hill and Morris, 1993; Ruan et al., 1996; Mize et al., 1998). Thus, in both instances, translation is homoeostatically governed by negative feedback inhibition control mechanisms in which nascent peptides specifically detect raised metabolite levels to stall elongating ribosomes.

Elongation delays that disrupt anticodon—codon interactions

Slowed translation elongation can also cause ‘recoding’, defined as a non-standard translational event that alters the way in which the ribosome interprets the genetic code, changing the amino acid sequence of the translated polypeptide. One example is provided by +1 ribosomal frameshifting. At such a frameshift site, elongation is paused by the presence of either a rare codon or stop codon in the A-site. For Ty1 family retrotransposons in S. cerevisiae, a 7 nt sequence consisting of CUU-AGG-C is sufficient to drive +1 frameshifting at the junction of the gag (in the 0 frame) and pol reading (+1) frames (Figure 2) (Belcourt and Farabaugh, 1990). In this case, frameshifting depends on an unusual Leu tRNAUAG isoacceptor, which through lack of a modified ‘wobble’ nucleotide in the anticodon can potentially decode all CUN codons (Weissenbach et al., 1977). It was discovered that this tRNA can also base-pair in the +1 frame (UUA codon), leading to the proposal that delayed translation of the rare arginine AGG codon located in the ribosomal A-site causes an increased likelihood of slippage of the P-site Leu tRNAUAG into the +1 frame. Supporting this notion, decreasing levels of the Arg tRNACCU, and thus enhancing pausing, increases frameshifting (Kawakami et al., 1993). Conversely, over-expressing this tRNA reduces frameshifting (Belcourt and Farabaugh, 1990). In yeast Ty3 retrotransposons, a different mechanism is employed, as the 7 nt frameshift site differs (GCG-AGU-U), and the P-site alanine tRNAIGC that stimulates the frameshift is incapable of slippage into the +1 frame (Farabaugh et al., 1993; Sundararajan et al., 1999). For this type of frameshift, the unusual nature of the wobble nucleotide interaction by the P-site tRNA, in combination with low levels of a tRNA necessary to decode the zero frame A-site codon, stimulates binding in the +1 frame of a more abundant tRNA (Figure 2) (Sundararajan et al., 1999). Supporting this model, increasing or decreasing +1 frame binding tRNA levels caused increased or decreased Ty3 frameshifting (Sundararajan et al., 1999). Pause-induced translational frameshifting is not unique to eukaryote systems. In E. coli, rare arginine codons can induce translational pauses that result in +1 frameshifting in certain contexts, the frequency of which is reduced by overexpressing the rare tRNA argX (McNulty et al., 2003). Pausing at the ‘early’ in-frame stop codon in the E. coli RF2 (release factor 2) mRNA, the length of which pause is determined by RF2 availability, also triggers a programmed +1 frameshift event (Adamski et al., 1993). Pausing can also cause −1 frameshifting, which is employed by many retroviruses and other RNA virus. In this case, ribosomal pausing is engineered by the presence of an mRNA secondary structure (Jacks et al., 1988). This is commonly a complicated pseudoknot structure, which lies downstream of a heptameric slippery sequence that triggers the backward ribosomal slippage, coincident with EF2 (elongation factor 2)-mediated ribosomal translocation (Namy et al., 2006).

Figure 2.

Models of frameshifting

(A) The Ty1 ‘slippage’ mechanism depends on the slow arrival of tRNA (purple) in the A-site, and the ability of peptidyl-tRNA (red) to re-pair (red arrow) in the +1 frame. (B) The Ty3 ‘out-of-frame pairing’ mechanism depends on an imperfect wobble nucleotide interaction of P-site tRNA with its codon that disturbs (red circle) pairing in A-site codon in the 0 frame. Combined with slow arrival of the 0 frame tRNA (purple), this results in increased likelihood of decoding by an alternative tRNA (orange) in +1 frame. (C) The ‘tandem slippage mechanism’ typically depends on the presence of a pseudoknot (spiral) that inhibits EF2-driven translocation. The resulting stress on the mRNA can be resolved either by pseudoknot unwinding (not shown) or simultaneous re-pairing in the −1 frame by peptidyl (purple) and deacylated (brown) tRNAs. Repairing in all three models is indicated by relative movement of mRNA (blue box and lettering).

To summarize, both +1 and −1 frameshifting are utilized as part of ‘programmed’ gene expression regulatory mechanisms. In these situations, frameshifting occurs at significant frequencies in optimized selected contexts, and is employed to precisely regulate both the amount of protein produced and the identity of the polypeptide expressed; normally both non-frameshift and frameshift encoded polypeptides have specific biological function. However, ribosomal frameshifting, particularly +1 frameshifts, can occur as a translational error in potentially many places within an ORF, depending upon the cis mRNA context and tRNA availability. There are reports from studies using E. coli that certain codons, if they become ‘hungry’ through aminoacyl tRNA limitation, can frameshift in error to generate a nonsense product (Fu and Parker, 1994; Gurvich et al., 2005). Frameshifting may therefore also be thought of as an adaptation mechanism that helps a stalled ribosome population to be released from the mRNA, since most accidental frameshifts, triggered by a hungry codon in an appropriate context, will be rapidly followed by a stop codon in the new reading frame.

Pauses during translation of rare codons and stop codons can also trigger translational bypassing in E. coli, in which ribosomes skip a section of mRNA sequence and resume translation downstream (Gallant and Lindsley, 1998) ‘Take-off’ is stimulated by a combination of weak Watson—Crick anticodon—codon interactions in the P-site at the ‘take-off’ codon, coupled with slow decoding in the A-site. ‘3′-Scanning’ then occurs, with the elongating ribosome still in a complex with the peptidyl tRNA (Herr et al., 2004). ‘Landing’ occurs when the peptidyl tRNA finds a synonymous codon where robust anticodon—codon interaction is possible (i.e. same as take off codon, or wobble-position variant) (Herr et al., 2004; Bucklin et al., 2005). Bypassing of this type occurs in mammalian cells at typically low frequencies (<1%; Chittum et al., 1998). However, in the case of the phage T4 gene60 mRNA (encoding a topoisomerase needed for T4 replication), E. coli ribosomes successfully bypass a 50 nt sequence at a remarkable frequency of 50% (Huang et al., 1988; Herr et al., 2000a). In this instance, bypassing is stimulated by a slowly decoded stop codon immediately following the take-off GGA codon, in combination with a downstream secondary structure (Herr et al., 2000a, 2001) and the properties of the nascent peptide (Herr et al., 2000b, 2001; Weiss et al., 1990). Bypassing thus resembles frameshifting, in the stimuli that induce it, in its potential use in gene expression regulatory mechanisms, and because it may represent a response to avoid unwanted ribosomal stalling.

Elongation delays that cause translational abandonment via mRNA degradation

Despite the importance of protein synthesis, elongating ribosomes do not always successfully complete translation (Jorgensen and Kurland, 1990), and indeed E. coli ribosomes possess a basal drop-off frequency of 4×10−4 per codon (Menninger, 1976). Although poorly understood, this ‘background’ level of translational abandonment is thought to result from dissociation of missense-decoded peptidyl-tRNAs from the mRNA due to their weak non-cognate interactions with the mRNA (Menninger, 1977). However, it is now clear that elongation pauses, caused by mRNA features such as strong ORF secondary structures or runs of rare codons, can lead to abandonment of translation via diverse mechanisms. In S. cerevisiae, strong ORF secondary structures trigger NGD (no-go decay), in which translation abandonment is coincident with mRNA decay (Doma and Parker, 2006). It was shown that insertion of a 31 bp stem-loop structure (ΔG −74.7 kcal·mol−1) into the ORF of two different reporter mRNAs induced an endonucleolytic cleavage event in the vicinity of the stalled ribosome, leading to turnover of the mRNA and release of the translational components. Interestingly, a pseudoknot or a run of rare codons also induced NGD, although to a far smaller extent, perhaps reflecting a lesser ability to stall ribosomes as effectively, or indicating that other mechanisms respond to these mRNA features (Doma and Parker, 2006). Although the endonuclease remains to be identified, two proteins, Dom34p and Hbs1p, which show similarity to the translation release factors eRF1 (eukaryotic RF1) and eRF3, were identified as key effectors of NGD, as far less mRNA cleavage occurs in their absence (particularly Dom34p). It has been suggested that Dom34p/Hbs1p may interact at the ribosomal A-site to release the stalled complex in a termination-like interaction (Doma and Parker, 2006). NGD may therefore act as a general regulatory mechanism that intervenes to rescue ribosomes stalled at difficult-to-translate mRNA sequences. Such sequences may have arisen due to transcriptional RNA damage, in which case NGD can be thought of as defective mRNA clean-up mechanism. Alternatively, NGD could also act upon ribosomes that have been deliberately paused at mRNA sequences in response to cell signals/environmental change, leading to a novel form of gene expression regulation.

Elongating ribosomes that become stalled at rare codons in bacteria are rescued by the tmRNA (transfer—messenger RNA)–SmpB complex. tmRNA is an unusual tRNA/mRNA hybrid molecule whose tRNA-like domain is charged with alanine (Komine et al., 1994). Charged tmRNA binds to stalled ribosomes within a ribosomal A-site unoccupied by any mRNA template (Keiler et al., 1996). Ribosomes then ‘jump’ template, and switch from the defective mRNA in favour of tmRNA's own internal ORF of approx. 10 codons, including a terminal stop codon. Furthermore, the C-terminal tag encoded by the tmRNA targets partially made protein for turnover by ClpAP and ClpXP proteases (Gottesman et al., 1998). Bacterial ribosomes stalled at rare codons during elongation are also targets for tmRNA, after mRNA is endonucleolytically cleaved within or just downstream of the A-site within the paused ribosome complex (Roche and Sauer, 1999; Hayes and Sauer, 2003; Li et al., 2006). Such cleavage may be mediated by the RelE endonuclease (Christensen and Gerdes, 2003; Pedersen et al., 2003), although this may not always be the case (Hayes and Sauer, 2003; Sunohara et al., 2004). Clusters of rare codons, but not single rare codons, can also trigger tmRNA activity (Li et al., 2006). Studies with different numbers of contiguous rare codons, and tRNA overexpression, suggest rare codon runs trigger local charged tRNA depletion (Roche and Sauer, 1999). Intriguingly, pairs of rare codons are over-represented in all bacterial genomes (Buchan et al., 2006). Initially, this may seem counter-intuitive, given that rare codon clusters slow translation elongation (Sorensen et al., 1989; Irwin et al., 1995), and potentially induce tmRNA-mediated abandonment of translation, with mRNA turnover. However, it is thought that such codon usage patterns may form part of an amino acid starvation response in bacteria, whereby the action of RelE and tmRNA releases ribosomal resource from the majority of mRNA transcripts, allowing translation of mRNAs encoding amino acid biosynthetic enzymes (Pedersen et al., 2003). Lastly, tmRNA functions to free ribosomes paused for extended times at termination codons. In some cases, this is due to the combined properties of either a C-terminal proline or arginine residue together with a following inefficient stop codon (UGA; Hayes et al., 2002). Intriguingly, it has now been demonstrated that for ribosomes paused at stop codons and subject to tmRNA activity, tmRNA A-site entry is preceded by an unidentified endonucleolytic activity that is distinct from RelE, and, indeed, MazF and other endonucleases (Hayes and Sauer, 2003). To summarize, although mechanistically quite distinct from the eukaryotic NGD process (eukaryotes lack a tmRNA homologue), the prokaryote tmRNA-driven release of stalled ribosomes within ORFs seems similar in its requirement for an endonucleolytic primary trigger event.

Elongation delays that cause translational abandonment with peptidyl-tRNA release

The activity of factors such as tmRNA in bacteria and Dom34p/Hbs1p in yeast represents a way to neatly abandon translation during depletion of charged tRNA levels. However, there is evidence to suggest translation abandonment can occur in a far less structured way, with the ribosome somehow dropping off the mRNA. In E. coli, overexpression of very short ‘mini-genes’ (such as the bar genes in bacteriophage λ) causes toxicity in pth mutants. The pth gene encodes peptidyl-tRNA hydrolase, an enzyme acitivty that cleaves nascent peptides from non-ribosome bound tRNA molecules (Henderson and Weil, 1976; Garcia-Villegas et al., 1991). The process of translating mRNA transcripts of mini-genes is particularly sensitive to translation abandonment. In a pth temperature-sensitive mutant, the resulting peptidyl-tRNAs cannot be processed to release the tRNA, leading to a starvation for particular tRNAs in the cell. This starvation can be overcome by overexpression of the relevant tRNA (Heurgue-Hamard et al., 1996). Importantly, the toxicity of codons is not related to the abundance of the decoding tRNA, as the pause occurs at the stop codon terminating the mini-gene; it is the P-site tRNA that then ‘drops off’. The toxicity of codons is also determined by the efficiency of the RF apparatus, with defective RF1 causing an increased translational pause and thus opportunity for drop-off (Cruz-Vera et al., 2003). Curiously, the drop-off effect is highly length-dependent, and above a length of 6 codons there is a rapid reduction in peptidyl-tRNA release (Heurgue-Hamard et al., 2000).

As an emergency ‘bailing out’ procedure, ribosomal drop-off raises some interesting questions, not least how the E. coli ribosome and peptidyl-tRNA are removed from the mRNA. There are suggestions of two distinct mechanistic routes. In the first, IF1 (initiation factor 1) and IF2 synergistically remove the peptidyl-tRNA from the ribosome, creating a substrate for peptidyl-tRNA hydrolase (Karimi et al., 1998). This IF-driven process competes directly with translation elongation, and occurs during the synthesis of short peptides. As the ORF length progressively increased from three codons to seven, the in vitro drop-off frequency decreased (Karimi et al., 1998). Overexpression of these two initiation factors in vivo enhances pth temperature-sensitive mutant phenotypes, suggesting that IF1/IF2 overexpression enhances the release of peptidyl-tRNAs from the ribosome (Karimi et al., 1998). The second factor- driven drop-off mechanism also operates on short nascent polypeptides. This second mechanism involves RF3, EF-G and RRF (ribosome recycling factor; also called RF4) instead of IF1 and IF2 (Heurgue-Hamard et al., 1998; Karimi et al., 1998).

The factor-driven stimulation of ribosomal drop-off during translation of short peptides suggests the translating ribosome is particularly sensitive to drop-off immediately post-initiation. However, premature ribosome drop-off can occur during translation of longer ‘normal-length’ protein-encoding ORFs. Codons of the form NGG reduce the efficiency of translation of a long ORF if located at the +2 to +5 positions (relative to the AUG codon designated +1). Thereafter, NGG codons have no detectable negative effect (Gonzalez de Valdivia and Isaksson, 2004). Overexpression of a long ORF with NGG codons at +3 to +5 caused growth inhibition of a pth mutant strain, suggesting peptidyl-tRNA drop-off was being triggered (Gonzalez de Valdivia and Isaksson, 2005). Other researchers have shown that pairs of AGA AGG arginine codons close to the initiation codon of the λ int gene can cause inhibition of growth due to peptidyl-tRNAArg accumulation (Olivares-Trejo et al., 2003). Finally, overexpression of normal-length ORFs can be toxic in pth mutants, with the toxicity dependent upon the nature of the last sense codon, suggesting translation pausing while waiting for termination may trigger ribosome drop-off and a demand for pth activity (Menez et al., 2000). In summary, the peptidyl-tRNA hydrolase activity is clearly essential in E. coli to respond to a significant basal level of ribosome drop off in the bacterial cell. Curiously, codons that exhibit enhanced peptidyl drop-off frequencies in mini-genes, such as AA(A/G) and AA(U/C), are over-represented at the beginning of E. coli ORFs (Chen and Inouye, 1994) and yet paradoxically enhance translation of ORFs when placed at positions +2 or +3 of a reporter (Cruz-Vera et al., 2003). It has been suggested these may act as sensors of translational health, preventing translation of an ORF if availability of charged tRNA is low (Cruz-Vera et al., 2003). Eukaryote species also have genes for peptidyl-tRNA hydrolase activity. In yeast, two have been identified, one of which may encode a (partly) mitochondrial-located protein. However, a yeast double PTH gene deletant is viable, suggesting either that there is less of a requirement for this activity in eukaryotes, or that there exists an additional redundant gene(s) still to be identified (Rosas-Sandoval et al., 2002).

Conclusions and overview

Given the wide range of effects translational pauses can induce, a fundamental question is which mechanistic route is followed when elongating ribosomes stall. For example, for protein folding to be assisted by selected pauses during elongation, there must be a means by which these pauses are tolerated without activating events such as frameshifting, bypassing or RelE—tmRNA/NGD-mediated mRNA turnover. Is there a critical length of pause time tolerated before some form of translational abandonment or frameshifting is triggered, or does the formation of a ribosomal queue represent the trigger for some form of emergency action? What is clear from the well-characterized examples of ‘programmed’ ribosomal frameshifting and bypassing is that cis-features, such as slippery nucleotide sequences, the presence of secondary structure, as well as nascent peptide effects, can favour one specific mechanism over others.

Comparing responses to translational pausing in prokaryote and eukaryotic systems reveals both similarities and differences. The RelE—tmRNA system found in bacteria has no direct counterpart in eukaryotes, and yet the use of the RelE endonuclease in bacteria to allow tmRNA into the ribosomal A-site of a paused ribosome is clearly analogous to the as-yet-unidentified endonuclease in yeast that participates in the NGD mechanism. However, in yeast, NGD is apparently more sensitive to a pause induced by mRNA structure than that induced by rare codons. It is possible that translational systems in bacteria have evolved to be hypersensitive to ribosome pausing at poorly translated codons, in order to rapidly respond to blocks in translation. It has yet to be reported whether the RelE—tmRNA system responds to pausing caused by mRNA secondary structures, although in E. coli palindromic sequences are rare, being efficiently removed from the genome by the activity of SbsC/SbsD (Chalker et al., 1988; Gibson et al., 1992). In contrast, a number of palindromic sequences exist in yeast, with longer ones over-represented (Lisnic et al., 2005). Consequently, responses to this type of translational pause may be required more frequently in eukaryotes. It seems likely that the pause response systems will be tuned to the properties of the host translation system. Ribosome drop-off, creating pth substrates, is clearly an important translational event in bacteria, but the extent to which it occurs in eukaryotes is unknown. More work on the activity and requirement for the yeast pth proteins will be required before this assessment can be made.

It is clear that in response to changing translation conditions, the action of frameshifting, bypassing, NGD/RelE-tmRNA and ribosomal drop-off within ORFs all operate as partially redundant rescue mechanisms to avoid the sequestering of elongating ribosomes on poorly translatable mRNAs (summarized in Figure 3). In E. coli, for example, stalled ribosomes can be substrates for drop-off, with peptidyl-tRNAs removed by pth, or alternatively be substrates for RelE—tmRNA. Recently, it has been reported that overexpression of tmRNA can partly suppress the phenotypes of a conditional lethal pth mutant, whereas the temperature-sensitive pth phenotypes are enhanced by tmRNA gene deletion (Singh and Varshney, 2004). This implies that stalling which would normally produce drop-off can be rescued instead by the action of tmRNA, presumably in concert with RelE. The rescue pathways are thus at least partly redundant, although even in an ssrA (encoding tmRNA) overexpression background, pth is still an essential gene (Singh and Varshney, 2004).

Figure 3.

Mechanisms and contexts of ribosomal drop-off

(A) Ribosomal drop-off occurs at a high frequency during translation of prokaryote uORF stop codons. The process can be catalysed by two different groups of factors, either IF1 and IF2, or by EF-G with the termination-associated RF3 and RRF (ribosome recycling factor). The released peptidyl-tRNA is a substrate for the essential peptidyl-tRNA hydrolase (pth) enzyme. (B) During translation of a longer ORF, translational pauses caused by stop codons, rare codon (pairs) or mRNA secondary structures can trigger abandonment of translation through a variety of mechanisms, some specific to either prokaryotes or eukaryote systems (see text for details). The drop-off mechanism employed is determined by the nature of the ribosomal pause. The Figure represents the current state of knowledge of drop-off mechanisms in both prokaryotes and eukaryotes.

In conclusion, the process of translation elongation is a focal point for a wide array of gene expression regulatory mechanisms that respond to the diverse sets of conditions experienced by an elongating ribosome, governed by tRNA availability, mRNA sequence and secondary structure, and nascent peptide sequence. Responses by the translating ribosome can trigger anything from a local enhancement of protein folding of the encoded polypeptide, through mRNA and nascent protein turnover, to translational recoding events, such as frameshifting. The way in which the ribosome, mRNA and tRNA populations interact clearly has the potential to fundamentally influence the relationship between transcriptome and proteome, and much of this potential remains to be investigated.


Research in I.S.'s laboratory is supported by funding from Tenovus Scotland, BBSRC (Biotechnology and Biological Sciences Research Council) and EPSRC (Engineering and Physical Sciences Research Council). We thank Roy Parker for reading the manuscript and offering helpful suggestions for improvement.


  1. Transcription attenuation: This is the regulated abandonment of transcription. Translation-mediated transcription attenuation, described here, is mediated by interplay between the activities of the RNA polymerase and ribosomes translating the transcript.

  2. Rho-independent transcription termination: This is defined as the termination of bacterial transcription, directed by a stable RNA hairpin loop secondary structure and downstream U-rich sequence. This form of transcription termination occurs independently of Rho protein activity.

  3. Ribosomal frameshifting: A 5′ or 3′ movement by the ribosome relative to the mRNA that changes the codon reading frame translated by the ribosome.

  4. Hungry codon: A codon whose cognate aminoacyl-tRNA is in short supply. This arises either because the tRNA itself is rare, or the relevant amino acid is in short supply, thus leading to a shortage of the charged form of an otherwise abundant tRNA.

  5. RelE: An E. coli endonuclease that cleaves the mRNA within a ribosomal A-site unoccupied for extended periods.