Alternative Fates of Paused Ribosomes during Translation Termination*

The bacterial tmRNA·SmpB system facilitates recycling of stalled translational complexes in a process termed “ribosome rescue.” During ribosome rescue, the nascent chain is tagged with the tmRNA-encoded ssrA peptide, which targets the tagged polypeptide for degradation. Translational pausing also induces a variety of recoding events such as frameshifts, ribosome hops, and stop codon readthrough. To examine the interplay between recoding and ribosome rescue, we determined the various fates of ribosomes that pause during translation termination. We expressed a model protein containing the C-terminal Asp-Pro nascent peptide motif (which interferes with translation termination) and quantified the protein chains produced by recoding and ssrA-peptide tagging. The nature and extent of translational recoding depended upon the codon for the C-terminal Pro residue, with CCU and CCC promoting efficient +1 frameshifting. In contrast, ssrA-peptide tagging was unaffected by C-terminal Pro coding. Moreover, +1 frameshifting was not suppressed by tmRNA·SmpB activity, suggesting that recoding and ribosome rescue are not competing events. However, cells lacking ribosomal protein L9 (ΔL9) exhibited a significant increase in recoding and a concomitant decrease in ssrA-peptide tagging. Pulse-chase analysis revealed that pre-termination ribosomes turn over more rapidly in ΔL9 cells, suggesting that increased recoding alleviates the translational arrest. Together, these results indicate that tmRNA·SmpB does not suppress transient ribosome pauses, but responds to prolonged translational arrest.

Ribosomes usually decode mRNAs with high fidelity to ensure accurate protein synthesis. However, some messages contain recoding signals that instruct the ribosome to change translational reading frames (frameshifting) or to read stop codons as sense codons (stop codon readthrough). Translational recoding is exploited to expand the genetic code for the incorporation of selenocysteine at UGA stop codons (1). Additionally, the synthesis of some proteins is regulated by programmed recoding. In many bacteria, release factor-2 (RF-2) 2 synthesis requires a ϩ1 frameshift at a premature UGA stop codon (2). Because RF-2 itself mediates termination at UGA codons, high levels of the release factor suppress the frameshift, thereby providing a mechanism to autoregulate synthesis (2). A more dramatic recoding event occurs during translation of the gp60 topoisomerase subunit from bacteriophage T4. The T4 gene 60 message contains two open reading frames (ORF) separated by an intervening sequence of 47 nucleotides. Instead of terminating translation at the first stop codon, ribosomes "hop" over the intervening sequence and resume translation after "landing" on the downstream coding sequence (3). Frameshifting and ribosome hopping both involve detachment of peptidyl-tRNA from the P-site codon followed by rebinding to a new codon (4,5). Frameshifting occurs if the new codon overlaps the original P-site codon, whereas ribosome hopping entails translocation/scanning down the mRNA until a suitable landing codon is identified (1). Because detachment of the peptidyl-tRNA is kinetically unfavorable, programmed recoding events typically rely on translational pauses to promote alternative decoding.
Although ribosome pauses are exploited for recoding, they are also a manifestation of defective protein synthesis. In eubacteria, the tmRNA⅐SmpB quality control system "rescues" paused ribosomes and targets their nascent polypeptides for degradation. tmRNA is a bi-functional RNA that acts sequentially as tRNA and mRNA during ribosome rescue (6). SmpB is a tmRNA-binding protein required for both the delivery of tmRNA to the ribosome, and translation of the ssrA peptide tag (7)(8)(9). The tmRNA⅐SmpB complex binds to the A-site of paused ribosomes by virtue of the tRNA-like domain within tmRNA. After transfer of the nascent chain to tmRNA, the original mRNA is released from the ribosome and translation resumes using a short ORF within tmRNA. Thus, the nascent chain is tagged with the tmRNA-encoded ssrA peptide, which is recognized by a number of proteases (6,10). In this manner, tmRNA⅐SmpB facilitates ribosome recycling and promotes the degradation of truncated proteins. Because translational pausing promotes both recoding and ribosome rescue, it is conceivable that tmRNA⅐SmpB activity suppresses recoding. Such a function could be beneficial by preventing translational errors and thereby increasing the fidelity of protein synthesis. However, it may also be disadvantageous for tmRNA⅐SmpB to interfere with programmed translational pauses that regulate protein synthesis.
Here, we examine the interplay between tmRNA⅐SmpBmediated ribosome rescue and translational recoding. Ribosomes were paused at stop codons using a C-terminal Asp-Pro nascent peptide motif, which interferes with translation termination (11)(12)(13). Biochemical and mass spectrometry analyses detected ssrA-tagged protein chains, as well as those produced by ϩ1 frameshifting and ribosome hopping. Recoding was dependent upon the coding of the C-terminal Pro residue, with CCC and CCU codons promoting efficient ϩ1 frameshifting. In contrast, the levels of ssrA-peptide tagging were not influenced by the C-terminal Pro codon. Cells lacking ribosomal protein L9 (⌬L9), which is required for frame maintenance during translational pauses (3)(4)(5), showed increased recoding and decreased ssrA-tagging compared with L9 ϩ cells. Pulse-chase analyses showed that paused ribosomes turned over more rapidly in the ⌬L9 background, presumably due to increased recoding during termination. The tmRNA⅐SmpB system did not suppress frameshifting in any of the examined strains, suggesting that recoding and ribosome rescue are not competing events. Because tmRNA⅐SmpB is only recruited to ribosomes that are paused on truncated mRNA (14), it appears that the rate of ribosome rescue is limited by mRNA cleavage or processing. According to this model, recoding occurs more rapidly than the nucleolytic processing required for tmRNA⅐SmpB activity. Taken together, these observations suggest that protein synthesis can be regulated by transient translational pauses without interference from the tmRNA⅐SmpB ribosome rescue system.
Protein Analysis-The synthesis of His 6 -YbeL proteins was induced with 1.5 mM isopropyl ␤-D-1-thiogalactopyranoside (IPTG). RF-1 production was induced with 0.4% L-arabinose. After 1.5 h of induction, cells were collected by centrifugation and the cell pellets frozen at Ϫ80°C. Frozen cells were broken by suspension in urea-lysis buffer [8 M urea, 150 mM NaCl, 10 mM Tris-HCl (pH 7.8)) followed by a freeze-thaw cycle. Lysates were clarified by centrifugation and His 6 -tagged proteins purified by Ni 2ϩ -affinity chromatography as described (20). Purified proteins were resolved on SDS-polyacrylamide gels and stained with Coomassie R-250. Samples for Western blot analysis were transferred onto nitrocellulose membranes and then incubated with primary rabbit anti-ssrA(DD) antibody overnight. Membranes were further washed and incubated with anti-rabbit secondary antibody (Rockland Immunochemicals), washed and scanned using a LI-COR infrared imager. Quantification of stained gels and Western blots was performed using Odyssey software. For mass spectrometry, proteins were further purified by reverse phase HPLC as described (21,22).
Pulse Chase and Northern Blot Analysis of Peptidyl-tRNA-Pulse-chase analysis of peptidyl-tRNA turnover was performed as described (13). Briefly, cells were grown in MOPS-buffered defined media supplemented with 0.4% D-glucose and all amino acids (except L-methionine and L-cysteine) at a concentration of 20 g/ml and L-arginine at 100 g/ml. Cultures were induced with 1.5 mM IPTG for 30 min, pulse-labeled for 1 min with L-[ 35 S]methionine/cysteine (20 Ci/ml, MP Biomedicals), then chased with 200 g/ml of unlabeled L-methionine/cysteine. Precipitated samples were run on acid-urea 6% polyacrylamide gels and visualized by phosphorimager. Data were quantified using Quantity One software (Bio-Rad) and exponential decay equations were fitted using DeltaGraph (RedRock Software). Northern blot analysis for peptidyl-tRNA was conducted as described (19). Oligonucleotides proK probe (5Ј-CTT CGT CCC GAA CGA AGT G) and proL probe (5Ј-CAC CCC ATG ACG GTG CG) were radiolabeled and used as probes for tRNA 1 Pro and tRNA 2 Pro , respectively.

RESULTS
Translational Recoding during Inefficient Termination-The C-terminal Asp-Pro nascent peptide motif interferes with translation termination and has been shown to promote stop codon readthrough in E. coli (23,24). We reported previously that the Asp-Pro motif also induces ssrA-peptide tagging of the E. coli YbeL-DP protein, but observed no evidence of readthrough or other recoding events in that study (11). To determine whether mRNA sequence influences recoding in our system, we produced His 6 -YbeL-DP from messages that encode the C-terminal proline residue as CCA, CCG, CCC, and CCU ( Fig. 1A). All constructs were initially expressed in cells lacking tmRNA (⌬tmRNA) so that recoding could be examined in the absence of ssrA peptide tagging. SDS-PAGE analysis showed higher molecular weight products in purified His 6 -YbeL-DP expressed from the CCC and CCU messages (Fig. 1B). Mass spectrometry revealed that the larger chains were produced by a ϩ1 frameshift during termination ( Fig. 1C and Table 1, rows 3 and 4). These findings are in agreement with a previous study (25) and are consistent with the P-site tRNA slippage model for ϩ1 frameshifting (4,26). In this scenario, tRNA 2 Pro (which decodes both CCC and CCU) transiently detaches from the P-site codon and slips forward by one nucleotide into the A-site UAA stop codon (e.g. CCC-UAA to CCC-UAA) to establish a new reading frame for continued translation into the 3Ј-untranslated region (UTR). This model accounts for the lower levels of frameshifting observed with a P-site CCU codon ( Fig.  1B), because such a shift forms a G:U wobble pair in the second position. CCU is also decoded by tRNA 3 Pro , but this isoacceptor would also form unfavorable G:U and U:U basepairs at the second and third positions, respectively. The same reasoning could explain the low level of recoding with P-site CCA and CCG codons, because the frameshifted base pairs would be unfavorable compared with the original codon-anticodon interaction. Based on the tRNA slippage model, UUU is only other P-site codon capable of supporting an efficient ϩ1 frameshift into a stop codon (UUU-UAA to UUU-UAA). However, we saw no evidence of ϩ1 frameshifting during synthesis of His 6 -YbeL-DF, in which the C-terminal Phe residue was encoded by UUU ( Fig. 1, A and B). Additionally, there was no apparent recoding with His 6 -YbeL-DA, whose C-terminal Asp-Ala peptide supports efficient translation termination (11,13). These results show that ribosome pausing is required for high-level recoding during termination.
Although we detected ϩ1 frameshift products with mass spectrometry, other less abundant recoding products may not have been identified due to signal swamping from the fulllength His 6 -YbeL-DP protein. To enrich for rare recoding products, we expressed FLAG-YbeL-DP (which contains an N-terminal FLAG epitope) from plasmids that encode His 6 tags in each reading frame downstream of the stop codon (supplemental Fig. S1A) (19). Therefore, His 6 -tagged proteins are only produced when recoding allows translation to proceed into the 3Ј-UTR (supplemental Fig. S1). Western blot against the His 6 epitope showed translation in the 0 and ϩ1 reading frames of the 3Ј-UTR (supplemental Fig. S1B). These recoding products were purified by Ni 2ϩ -NTA affinity chromatography and identified by mass spectrometry. Stop codon readthrough during translation with the CCC-and CCG-containing constructs (supplemental Fig. S1 and Table S2), accounting for much of the 0 frame recoding. An additional 0 frame product was formed by a ϩ12-nucleotide ribosome hop during translation of the CCG construct (supplemental Table S2). During ribosome hopping, peptidyl-tRNA Pro disengages from the P-site codon and reattaches to a downstream CCG codon in the 3Ј-UTR and resumes translation (Figs. 1A, supplemental Fig.  S1A and Table S2). Taken together, these results show that the Asp-Pro motif promotes recoding during termination, with the P-site codon-anticodon interaction determining the nature and extent of recoding.
tmRNA-mediated Ribosome Rescue Does Not Suppress ϩ1 Frameshifting-Because tmRNA tagging activity and recoding both occur during termination of YbeL-DP synthesis, we asked whether these two events influence one another. We produced His 6 -YbeL-DP in cells expressing tmRNA(DD), which encodes the protease resistant ssrA(DD) peptide tag (27); and then quantified the proportions of recoded and tagged proteins by SDS-PAGE and Western blot analyses. Although the same relative propensity for recoding was observed (CCC Ͼ CCU Ͼ CCG Ϸ CCA), there were more recoding products (relative to full-length chains) in tmRNA(DD) cells compared with ⌬tmRNA cells (Fig. 2, A and C). These data show that tmRNA⅐SmpB activity does not suppress recoding when ribosomes pause during termination. Moreover, the His 6 -YbeL-DP protein was tagged to the same extent irrespective of the C-terminal Pro codon (Fig. 2, A and B), indicating that increased recoding does not lead to a concomitant decrease in tmRNA⅐SmpB activity. Mass spectrometry showed that ssrA(DD) peptide tags were added after the C-terminal Pro residue (data not shown), confirming that tmRNA⅐SmpB acted during termination. In contrast, neither of the negative control proteins, His 6 -YbeL-DA or His 6 -YbeL-DF, were tagged in tmRNA(DD)expressing cells (Fig. 2, A and B). Taken together, these results suggest that translational recoding and tmRNA⅐SmpBmediated ribosome rescue are not competing processes.
Deletion of Ribosomal Protein L9 Increases Translational Recoding-Ribosomal protein L9 plays a role in reading frame maintenance during translational pauses (4, 5, 28 -32). Therefore, we sought to modulate the frequency of recoding through deletion of L9 (⌬L9). In ⌬tmRNA cells, we found that the L9 deletion increased recoding with all four Pro codon constructs (Fig. 3, A and B, compare L9 ϩ and ⌬L9 samples). ϩ1 Frameshifting was the predominant recoding event when the C-terminal Pro residue was encoded as CCC or CCU (Table 1, rows  7 and 8). Additionally, we detected the ϩ12 ribosome hop product when the CCA, CCG and CCU constructs were expressed in the ⌬L9 background ( Fig. 3A and Table 1, rows 5, 6, and 8). Examination of recoding in ⌬L9 cells using the plasmids that encode His 6 tags in the 3Ј-UTR yielded similar results as in L9 ϩ cells; stop codon readthrough and ribosome hopping occurred during translation of the CCC and CCG constructs, but ϩ1 frameshifting was predominant with the CCC construct (supplemental Fig. S1 and Table S2). Western blot analysis suggested that stop codon readthrough and ribosome hopping occur at roughly equal frequencies during translation of the CCG construct in ⌬L9 cells (supplemental Fig. S1, compare the (0) and (ϩ1) reading frames), which likely explains the two distinct recoding products seen by SDS-PAGE (Fig. 3, A and C). We suspect that similar recoding processes occur with the CCA construct, leading to the same protein profile as the CCG sample. In contrast, deletion of L9 had no discernable effect on the synthesis of the His 6 -YbeL-DA and His 6 -YbeL-DF control proteins (Fig. 3, A and B, Ala and Phe samples), again suggesting that ribosome pausing is required for recoding.
tmRNA Tagging of His 6 -YbeL-DP Is Reduced in ⌬L9 Cells-Because ⌬L9 cells showed increased recoding during inefficient termination, we asked whether tmRNA⅐SmpB activity is also affected in this mutant. His 6 -YbeL-DP was produced from the four Pro codon constructs in L9 ϩ and ⌬L9 cells that express tmRNA(DD). Western blot analysis showed that His 6 -YbeL-DP purified from the ⌬L9 background was tagged at lower levels compared with protein isolated from L9 ϩ cells (Fig.  3C, compare L9 ϩ and ⌬L9 samples). Although reduced, ssrA(DD)-peptide tagging was roughly equivalent for all four Pro codon constructs in ⌬L9 cells (Fig. 3C). These data suggest a Proteins were purified as described in "Experimental Procedures" and analyzed by electrospray ionization mass spectrometry. b Strains were also deleted for tmRNA (⌬tmRNA). c Mass calculation was based on the oxidized YbeL protein, which contains two disulfide bonds. Additionally, the N-terminal Met residue is removed posttranslationally. that increased recoding in ⌬L9 cells may relieve the ribosome pause that induces tmRNA⅐SmpB recruitment. However, it is also possible that L9 is required for optimal tmRNA⅐SmpB activity, such that tagging is inhibited independently of the effects on ribosome pausing. To test this possibility, we examined ssrA-peptide tagging in ⌬L9 cells using a well-characterized non-stop mRNA that encodes the N-terminal domain of phage cI protein (17,27). In addition, we deleted the ClpP protease from the ⌬L9 strain to prevent proteolysis of ssrAtagged cI repressor proteins. Because all ribosomes arrest at the 3Ј-end of the non-stop message, recoding products cannot be produced from this transcript. Instead, the nascent chain is either tagged by tmRNA⅐SmpB or released from the stalled ribosome in a process that requires ArfA (6,33). We found that non-stop encoded cI protein was ssrA-tagged at the same levels in L9 ϩ and ⌬L9 cells (Fig. 3D, compare lanes 2 and 3), demon-strating that L9 is not generally required for tmRNA⅐SmpB activity. This result is consistent with the findings of K. Williams and co-workers (34), who showed that L9 plays no role in resume codon selection during trans-translation. Together, these results suggest that increased recoding in ⌬L9 cells alleviates ribosome pausing during termination, and thereby specifically decreases tagging of the full-length His 6 -YbeL-DP protein.
Kinetics of Paused Ribosome Recycling-To test whether translational pausing is indeed alleviated in ⌬L9 cells, we used pulse-chase analysis of peptidyl-tRNA turnover to monitor ribosome pausing during termination. Immediately prior to release, YbeL-DP nascent chains are covalently linked to tRNA Pro in the P-site of pre-termination ribosomes. Inefficient termination results in an accumulation of peptidyl prolyl-tRNA Pro , which serves as a biochemical marker of paused ribosomes (13). To facilitate the analysis of peptidyl-tRNA by gel electrophoresis, we used mini-gene constructs that encode the C-terminal 49 residues of YbeL-DP fused to an N-terminal FLAG epitope. Because the C-terminal Asp-Pro motif is sufficient to pause ribosomes during termination (11), the smaller FLAG-(m)YbeL-DP nascent chain pauses ribosomes in the same manner as full-length YbeL-DP (data not shown) (13).
Cells expressing FLAG-(m)YbeL-DP were labeled with [ 35 S]methionine/cysteine, and the tRNA fraction isolated for acid-urea gel analysis. Autoradiography showed a ladder-like pattern of radiolabeled species corresponding to peptidyl-tRNAs (Fig. 4, A and B). Peptidyl-tRNA was not detected in uninduced cells, nor in cells overproducing the FLAG-(m)-YbeL-DA peptide (data not shown), indicating that ribosome pausing is required for peptidyl-tRNA accumulation (13). To identify the species corresponding to peptidyl prolyl-tRNA Pro bound to pre-termination ribosomes, we blotted unlabeled and 35 S-labeled samples onto the same membrane, then split the blot and performed Northern blot analysis for tRNA 1 Pro and tRNA 2 Pro on the unlabeled samples. Realignment of the membranes after northern hybridization showed that the peptidyl-tRNA species with the lowest gel mobility corresponded to peptidyl prolyl-tRNA Pro (Fig. 4A). Therefore, we quantified the turnover of this species to determine the half-life of paused ribosomes. The turnover of peptidyl prolyl-tRNA Pro in ⌬tmRNA cells expressing either the CCC or CCG coded constructs was essentially identical (Fig. 4B and Table 2). These data indicate that increased frequency of recoding seen with the CCC construct has little effect on ribosome pausing during termination. In contrast, peptidyl prolyl-tRNA Pro turnover was accelerated in ⌬tmRNA/⌬L9 cells compared with ⌬tmRNA/L9 ϩ cells (Table 2 and Fig. 4B, compare L9 ϩ with ⌬L9 samples). Consistent with previously published findings (13), we found that tmRNA also accelerated peptidyl prolyl-tRNA 2 Pro turnover in both L9 ϩ and ⌬L9 cells ( Table 2). These results support the hypothesis that increased recoding in ⌬L9 cells suppresses translational pausing during termination.

RF-1 Overexpression Suppresses Recoding during Inefficient
Translation Termination-Overproduction of release factor-1 (RF-1) has been shown to suppress ssrA-tagging of YbeL-DP by increasing the rate of translation termination (11)(12)(13). Because ribosome pausing also appears to induce recoding during translation termination, we examined the effects of RF-1 on recoding in ⌬tmRNA cells. RF-1 overproduction significantly reduced ϩ1 frameshifting and ribosome hopping, but was unable to completely suppress recoding in ⌬tmRNA/⌬L9 cells (Fig. 5, A and B). These data indicate that recoding is suppressed under conditions that alleviate ribosome pausing.
Recoding at Rare Codons-Ribosomes also pause during the translation of tandem rare Arg codons (19,27), which are decoded by low abundance tRNA Arg species (35). To examine recoding at rare codons, we expressed constructs encoding His 6 -YbeL-DPRRLA, a variant of His 6 -YbeL-DP containing an Arg-Arg-Leu-Ala sequence at its C terminus. The tandem Arg residues were encoded as either AGG (rare codon) or CGU (common codon) (Fig. 6A). Expression of the AGG codon construct in ⌬tmRNA cells produced two recoding products, each arising from ϩ1 frameshifting onto an A-site AGG codon (Fig.  6B, supplemental Fig. S1 and Table S2). In contrast, the tandem CGU construct produced full-length His 6 -YbeL-DPRRLA in ⌬tmRNA cells (Fig. 6B). Deletion of L9 led to a small increase in frameshifting at the rare Arg codons (Fig. 6, B and C, compare ⌬tmRNA/L9 ϩ to ⌬tmRNA/⌬L9). Frameshifting was suppressed by the overproduction of tRNA 5 Arg , which decodes AGG codons (Fig. 6, B and C). These results indicate that ribosome pausing at the rare AGG codons is required for the ϩ1 frameshifts.
We next tested whether tmRNA⅐SmpB activity influences recoding at rare Arg codons. Quantification of recoding products showed more ϩ1 frameshifting in tmRNA(DD) cells compared with ⌬tmRNA cells (Fig. 6C). This trend was observed in both L9 ϩ and ⌬L9 backgrounds (Fig. 6C). Therefore, the tmRNA⅐SmpB system does not suppress frameshifting at rare codons. We also quantified the levels of ssrA(DD) peptide tag-ging in this system and found that tagging at rare codons was unaffected by the L9 background (Fig. 6C). These results are generally consistent with the data obtained using ribosomes paused during termination, and indicate that recoding and tmRNA⅐SmpB activity are not competing events.

DISCUSSION
There are at least four general fates for paused ribosomes: (i) RF-mediated termination, (ii) translational recoding, (iii) tmRNA⅐SmpB-mediated rescue, and (iv) tmRNA-independent ribosome recycling. Translation termination is probably the major pathway because most of the released chains correspond to full-length protein. However, we note that full-length chains ] blots were realigned and imaged to identify the peptidyl-tRNA that corresponds to prolyl-tRNA Pro bound to the pre-termination ribosome. B, pulse-chase analyses of peptidyl-tRNA turnover. ⌬tmRNA cells (with and without ribosomal protein L9) expressing CCG-and CCC-coded constructs were pulse labeled with [ 35 S]methionine/cysteine, then chased with an excess of unlabeled amino acids (chase t ϭ 0 s). The peptidyl-tRNA Pro signals were quantified by phosphorimaging and a double exponential decay was fitted to the data to calculate composite half-lives (T 1/2 ) of peptidyl-tRNA turnover.  could also be released by the recently discovered ArfA peptide, which mediates tmRNA-independent ribosome rescue (33). ArfA is normally tagged for degradation by tmRNA⅐SmpB (36), but this alternative ribosome rescue system could significantly contribute to chain release in ⌬tmRNA backgrounds. In accord with reports from Isaksson and colleagues (23,24), we find that the C-terminal Asp-Pro nascent peptide motif promotes a variety of recoding events including ϩ1 frameshifting, stop codon readthrough and ribosome hopping. The propensity for recoding is well described by the P-site slippage model for frameshift-ing (26,37). This model predicts that recoding frequencies are related to the energy of the new codon-anticodon interactions. Consistent with this model, we find ϩ1 frameshifts occur at a higher frequency with the CCC-coded construct because P-site peptidyl-tRNA 2 Pro forms cognate base-pairs after advancing into the ϩ1 reading frame. In contrast, ϩ1 frameshifting is not favored during translation of the CCG-coded construct, and instead the paused ribosomes undergo stop codon readthrough and ribosome hopping. Quantification of recoding products indicates that ϩ1 frameshifting during translation of the CCC construct occurs more frequently than readthrough/hopping with the CCG construct. This difference in recoding has no discernable effect on peptidyl-tRNA Pro turnover as measured by pulse-chase analysis, presumably because only a small proportion of paused ribosomes undergo recoding.
The tmRNA⅐SmpB quality control system is responsible for recycling stalled ribosomes and therefore could play a role in suppressing unwanted translational recoding events. However, the proportion of protein chains produced by recoding actually increases in tmRNA(DD) cells compared with the ⌬tmRNA background. The mechanism underlying this seemingly paradoxical effect is unclear, but may be related to the observation that protein synthesis is more efficient in tmRNA ϩ cells. The ribosome rescue pathway is probably unable to suppress recoding because tmRNA⅐SmpB is only recruited to ribosomes arrested on truncated mRNA (6,14). Therefore the 3Ј-UTR of the ybeL-DP transcript must be cleaved or degraded prior to tmRNA⅐SmpB activity. Recoding is likely to occur more rapidly than this mRNA processing because we did not detect any protein chains produced by frameshifting on truncated transcripts. Moreover, His 6 -YbeL-DP chains were ssrAtagged to the same extent when produced from the CCCand CCG-coded constructs despite the different recoding frequencies. Together, these observations suggest that ssrApeptide tagging and recoding are not directly competing processes, and that only a subset of recalcitrant ribosomes is destined for tmRNA⅐SmpB-mediated ribosome rescue.
The C-terminal Asp-Pro sequence is under-represented in the proteome of E. coli and many other eubacteria (11). E. coli encodes only one predicted protein with this C-terminal sequence (YabN), but ϳ21 proteins are expected based on the frequency of the Asp-Pro dipeptide in E. coli proteins (11). Boycheva et al. have shown that certain Pro:stop codon combinations are not found in the E. coli genome including CCC:UAG, CCU:UAG and CCA:UAG (38). There also appears to be selection against Pro:Arg codon combinations that include either AGG or AGA rare Arg codons (38). Together with a variety of experimental results, these observations suggest that certain codon pairs are underrepresented because they induce translational pauses that induce tmRNA⅐SmpB activity and/or recoding events (11,24,25,39). However, the CCC:UGA codon pair is not underrepresented in E. coli and has been shown to induce significant ϩ1 frameshifting during the translation of the PheL leader peptide (25). Ribosome pausing during PheL synthesis regulates transcription of the phe operon through an attenuation mechanism. Although this recoding could conceivably modulate attenuation, Gurvich et al. recently showed that ϩ1 frameshifting has no effect on phe operon expression (40). Additionally, C-terminal Pro residues can be exploited to regulate gene expression. The E. coli TnaC leader peptide contains a critical C-terminal residue that is required for a programmed ribosome arrest that regulates expression of the tryptophanase (tna) operon through a novel attenuation mechanism (41,42). Although the 3Ј-end of tnaC gene contains a CCU:UGA codon pair that supports ϩ1 frameshifting, recoding is not likely to occur at appreciable levels because TnaC-arrested ribosomes have RF-2 bound stably in the A site (41,43). Therefore, although some frameshifting signals are fairly common in E. coli genes, the actual frequency of recoding at these sites tends to be low (25).
In accord with work from Atkins, Gesteland, Björk and their co-workers (4, 29, 31), we find that ribosomal protein L9 plays a role in reading-frame maintenance during translational arrest. This function was first described by Gesteland & Atkins, who discovered mutations in rplI gene (encoding L9) that modulate the frequency of ribosome hopping during translation of the bacteriophage T4 topoisomerase IV protein. High-resolution crystal structures have revealed the position of L9 on the ribosome (32,44), but the role of L9 in frame maintenance is still uncertain. L9 is a component of the large subunit, which does not make contacts with the mRNA during translation. Moreover, L9 is bound near the E site and extends a domain away from the subunit. One model postulates that L9 influences release of deacylated tRNA from the E site in response to the binding of ternary complex to the A site. Presumably, ribosomes with empty E and A sites have increased mobility on mRNA, allowing them to advance into the unliganded A-site codon. Regardless of mechanism, deletion of L9 enhances ϩ1 frameshifting during inefficient termination to such an extent that tmRNA⅐SmpB activity is reduced and peptidyl-tRNA turnover is increased. However, loss of L9 has no discernable effect on protein synthesis in the absence of translational pausing. This is consistent with the relatively subtle phenotype of E. coli ⌬L9 mutants, which grow slightly more slowly in rich media under laboratory conditions. Moreover, L9 is not present in either eukaryotic or archaeal ribosomes (45), suggesting that this mode of reading frame maintenance is not a central function in protein synthesis. Perhaps L9 acts as a fail-safe system that holds the reading frame in response to stress or other untoward conditions. One such stress is amino acid starvation, which is known to slow translation rates and induce ribosome pausing at some codons.