The entry of tmRNA into stalled ribosomes to serve first as a tRNA and then as an mRNA template poses two problems for our current understanding of translation. First, how does tmRNA get past the decoding center when it lacks an anticodon and enters an A-site devoid of mRNA? Secondly, how does the ribosome resume translation on tmRNA at the proper site and in the proper frame? The -1 triplet hypothesis of Lim and Garber [16
] suggests that three nucleotides immediately upstream of the resume codon mimic an codon-anticodon pair and bind directly to the ribosomal decoding center. Lim and Garber derived rules based on conformational analysis that predict which triplets can assume an A-form conformation as single-stranded RNA [16
Using the KanR assay, which ties the life of E. coli
cells to the function of tmRNA [11
], we directly assayed each -1 triplet mutant for tmRNA activity. We found that eight of the 18 forbidden mutants showed activity indistinguishable from wild-type. Conversely, of the 18 mutants that were completely inactive, only five were predicted to be forbidden. Our data support the idea that the -1 triplet sequence is important to tmRNA function, but they do not fit the rules developed by Lim and Garber (Table ). Instead, our data show that purines are highly favored at position 87 (the first of the -1 triplet) and that U and A are somewhat disfavored at position 89.
These characterizations broadly fit the natural distribution of -1 triplet sequences (Figure ). The preference for purines at position 87 matches the consensus obtained from all of the known natural tmRNA sequences [21
], in which A and G occur roughly equally with only a negligible number of pyrimidines. There seems to be no nucleotide preference at the middle position in either our data or the consensus sequence. At position 89, C and U are equally represented in natural tmRNA sequences and strongly preferred to A or G. In our data, U is underrepresented at position 89 in active sequences. Perhaps this is because seven of the nine inactive sequences containing U89 have a pyrimidine at position 87. For example, the UGU and CGU mutants display no detectable activity even in the low-stringency KanR assay. Going against both of these trends appears to be particularly deleterious. As natural -1 triplets nearly always begin with purines, U89 may be more acceptable in the natural pool of sequences.
We performed a second assay based on the efficiency of plaque formation by the hybrid phage λimmP22 c2-dis
. This phage can only form plaques on cells with active trans
-translation systems [3
]. We assayed five -1 triplet mutants ranging in the KanR assay from fully active to completely inactive under our lowest stringency conditions (UGU). We found that all five of these mutants had activities that were indistinguishable from wild-type tmRNA in the phage assay (Figure ). All five of these -1 triplet sequences were forbidden by Lim and Garber's rules. The results of these two assays suggest that the -1 triplet sequence does not serve as a codon-anticodon mimic nor interact directly with the decoding center in a manner essential to tmRNA accommodation in the A-site.
A recent study by Shimizu and Ueda provides further evidence that the tmRNA -1 triplet sequence, and indeed the whole template sequence, is not required for ribosome binding, A-site accommodation, or peptidyl transfer [22
]. Using a reconstituted translation system in vitro
, they detected polyalanine synthesis with the SmpB protein and a tmRNA consisting of only the TLD. In this system, the growing polyalanine chain is passed from one tmRNA TLD to the next. The ability of tmRNA lacking a template sequence to bind ribosomes and transfer Ala proves that only the TLD and SmpB protein are required for the tRNA activity of tmRNA. In addition, the revised cryo-EM structure of SmpB and tmRNA bound to the 70S ribosome shows the SmpB C-terminus binding near the decoding center in the 16S rRNA [23
]. Although the unstructured C-terminal tail is not visible in this structure, it is the most likely candidate for sending a signal equivalent to codon-anticodon pairing.
The -1 triplet is not required for legitimizing tmRNA entry into the ribosome, but together with U85 and A86, it does play a role in setting the correct frame for translation to resume. Analysis of tagging in the -1 frame reveals that both the deleterious UGU -1 triplet mutation and the U85A mutation increase translation in the -1 frame by a factor of 2.5 compared with wild-type (Figure ). The A86C mutation strongly induces +1 misreading. This tendency to select the wrong resume codon is one likely cause for the inactivity of these sequences in the KanR assay.
Analysis of the A86C mutation by phage assay (Figure ) and in vitro
tagging assays indicate that this mutation completely destroys tmRNA activity [13
]. No ribosome release or tagging is seen in this mutant. These data contradict our finding that in the misreading assays the A86C mutant shifts the resume codon strongly to the +1 frame. The explanation lies in the fact that different tag template sequences were used in these assays. Chemical probing experiments show that the upstream region in the A86C mutant assumes a different structure, presumably a base-paired structure and therefore impervious to chemical probes [24
]. A possible helical pairing surrounding the resume codon (between 85–90 and 95–100) would be further stabilized by the C:G pair created by the A86C mutation (Figure ). This structure could block access to a ligand that binds the upstream sequence. In our misreading assays, we used an altered tag sequence (ANDH6
D) that cannot form this secondary structure.
Our findings confirm earlier work showing that A86 is the most important determinant for frame choice on the tmRNA template [14
]. It is the most highly conserved nucleotide in the upstream sequence. U85 is tolerant to mutation, except for U85A. We propose that this mutation causes the A86 binding ligand to recognize A85 as well, shifting the frame back by one nucleotide. This explains the -1 shift in frame in U85A. The A86G mutation shows very low activity and the C and U mutations are totally inactive. When the masking effect of an inhibitory structure is removed, A86C causes +1 misreading because the A86 ligand binds to G87 instead, moving the resume codon ahead by one nucleotide. A86U has also been reported to induce +1 misreading in vitro
]. The ligand for A86 prefers purines over pyrimidines. The -1 triplet also affects the frame but in a more subtle manner. The preference for a purine over a pyrimidine at the first base of this triplet (G87), the strongest trend for these three nucleotides, likely involves an additional interaction with the A86 ligand.
In contrast with an earlier model that suggested that the resume codon is positioned inside the ribosome by the global fold of tmRNA, and the frame fine-tuned by the upstream sequence [14
], we propose that the A86-binding ligand acts independently to set the frame. The only ribosome binding element on tmRNA that can provide such positioning is the TLD. The four pseudoknots in tmRNA can be replaced with unrelated sequences and structures as can the tag template [11
]. The distance from pseudoknot 1 to U85 is not critical for tagging, but insertions or deletions between A86 and the resume codon (G90) cause misreading of the resume codon [13
]. It would seem that the A86-binding ligand is a ruler that establishes the frame four nucleotides down from A86 by placing the resume codon in the A-site to act as a template.
The refutation of the -1 triplet hypothesis discredits the most plausible model for direct binding of the ribosomal decoding center to the upstream sequence of tmRNA. The frame misrecognition results are best explained by a separate ligand binding to A86 and playing the dominant role in establishing the frame. What is the A86-binding ligand? One candidate is ribosomal protein S1, which was shown to crosslink to U85 in the upstream sequence (as well as pseudoknots 2 and 3) [26
]. Visualization of tmRNA inside 70S ribosomes by cryo-electron microscopy revealed changes in the structure of the template sequence in ribosomes with or without the S1 protein [27
]. Specifically, the template sequence is structured and removed from the decoding center in ribosomes lacking S1. The authors speculate that free S1 binds tmRNA and stabilizes a functional, open conformation of the template that is then passed to stalled ribosomes [27
]. In support of this model, one study shows that S1 is dispensable for tmRNA entry and Ala transfer but required for its mRNA activity in vitro
]. On the other hand, the expression of dominant negative S1 mutants in vivo
does not interfere with tmRNA function although it inhibits normal translation [30
]. Bacillus subtilis
and other Gram-positive bacteria lack an S1 protein and yet have tmRNA and functional trans
-translation systems. Two recent in vitro
studies using E. coli
or Thermus thermophilus
reconstituted translations systems show that S1 does not affect tmRNA function [31
]. S1 cannot interact directly with tmRNA on the ribosome, as S1 binds to the back of the head of the 30S subunit.
A second candidate is SmpB, a protein that enhances aminoacylation of tmRNA and is required for entry into the ribosome (as discussed above). The best characterized binding site of SmpB is on the TLD of tmRNA, the interaction visualized by two co-crystal structures of SmpB and short fragments of tmRNA [33
]. Multiple copies of SmpB can bind tmRNA simultaneously [35
], however, even in the context of the 70S ribosome [23
]. SmpB binding also alters the accessibility of the upstream sequence and pseudoknot 1 to nucleases in probing experiments, leading to the proposal that it plays a role in frame choice [36
]. This possibility is further strengthened by the recent finding that SmpB binding protects U85 from chemical modification and that this protection shifts by one nucleotide in tmRNA mutants that induce misreading of the resume codon [24
]. On the other hand, several crosslinking, chemical probing, and hydroxyl-radical cleavage assays have failed to detect such an interaction [35
]. It is not known which amino acids on SmpB may interact with the upstream sequence. Through this interaction, SmpB may orient the template sequence with respect to the TLD, causing conformational changes in the template, upstream sequence, and pseudoknot 1, positioning the resume codon in the A-site. If this model is correct, SmpB is solely responsible for allowing tmRNA to enter stalled ribosomes, tricking the decoding center, and it also plays a crucial role in its interaction with the upstream sequence on tmRNA to set the frame for translation of the tmRNA template.