|Home | About | Journals | Submit | Contact Us | Français|
Bacterial ribosomes stalled by truncated mRNAs are rescued by transfer-messenger RNA (tmRNA), a dual-function molecule that contains a tRNA-like domain (TLD) and an internal open reading frame (ORF). Occupying the empty A site with its TLD, the tmRNA enters the ribosome with the help of elongation factor Tu and a protein factor called small protein B (SmpB), and switches the translation to its own ORF. In this study, using cryo-electron microscopy, we obtained the first structure of an in vivo-formed complex containing ribosome and the tmRNA at the point where the TLD is accommodated into the ribosomal P site. We show that tmRNA maintains a stable ‘arc and fork' structure on the ribosome when its TLD moves to the ribosomal P site and translation resumes on its ORF. Based on the density map, we built an atomic model, which suggests that SmpB interacts with the five nucleotides immediately upstream of the resume codon, thereby determining the correct selection of the reading frame on the ORF of tmRNA.
Translation of an mRNA that lacks a stop codon stalls the ribosome at the end of the message, and produces a defective polypeptide that may be toxic to the cell. In bacterial cells, this situation is resolved by a process called trans-translation (Tu et al, 1995; Keiler et al, 1996a). The key player in trans-translation is a highly structured molecule called the transfer-messenger RNA (tmRNA), which consists of a tRNA-like domain (TLD) that mimics a tRNA structurally and functionally (Gutmann et al, 2003; Bessho et al, 2007), an open reading frame (ORF) that encodes a degradation signal, and several pseudoknots and helices. Elongation factor Tu (EF-Tu) and an 18.3-kDa protein called small protein B (SmpB) also have important functions in trans-translation. Forming a complex with EF-Tu, GTP and SmpB, tmRNA occupies the empty A site with its TLD (Rudinger-Thirion et al, 1999; Valle et al, 2003). Acting like a tRNA, the TLD accepts the polypeptide from the P-site tRNA and is translocated to the P site, whereas the deacylated P-site tRNA moves to the E site. The ORF of the tmRNA is then placed into the ribosome, replacing the defective mRNA molecule (Williams et al, 1999; Ivanova et al, 2005). Translation resumes on the ORF until the ribosome reaches an internal stop codon on the ORF. Subsequently, the nascent polypeptide chain, which is now tagged by the signal encoded in the ORF, is released and targeted for degradation (Keiler et al, 1996b).
The initial stages of trans-translation, up to the point when TLD is accommodated to the ribosomal A site, have been visualized by cryo-EM. Structures of the tmRNA–ribosome complex solved to date include those of the pre-accommodation complex (Valle et al, 2003; Kaur et al, 2006), which reflects the state right before the accommodation of the TLD into the ribosomal A site, and the accommodation complex (Kaur et al, 2006; Weis et al, 2010), which reflects the state right after the accommodation of TLD in the A site. In the structures of the pre-accommodation complexes, the tmRNA pseudoknots form a solid arc around the head of the 30S subunit. Its TLD, positioned like an A/T-state tRNA, interacts with EF-Tu in a way similar to a canonical tRNA (Valle et al, 2003; Kaur et al, 2006). Thus, the cryo-EM structures confirmed the notion that the TLD is a structural and functional mimic of tRNA, and suggested that the pseudoknots most likely have a function in maintaining the complex topology of the tmRNA. In contrast to the solid structure of the pre-accommodation complex, the accommodation complex (Kaur et al, 2006; Weis et al, 2010) only showed fragmented density attributable to the tmRNA, suggesting structural variability resulting from flexibility of the complex and/or low occupancy of the tmRNA on the ribosome.
Two of the most intriguing questions are how the ribosome selects the correct reading frame on the ribosome and how tmRNA passes through the ribosome. Previous studies suggest that several nucleotides immediately upstream of the resume condon are important for establishing the correct reading frame (Williams et al, 1999; Lee et al, 2001; Miller et al, 2008). The interaction between SmpB and tmRNA also appears to have a function in this process (Konno et al, 2007; Metzinger et al, 2008; Watts et al, 2008, 2009). Biochemical and genetics methods have been used to investigate the manner in which the tmRNA passes through the ribosome, in the later stages of trans-translation (Ivanov et al, 2002; Wower et al, 2005; Ivanova et al, 2007; Bugaeva et al, 2008). Because of the complex topology of tmRNA, some researchers suggested that the tmRNA has to unfold extensively in order to pass through. However, without understanding of the structural aspects of these later processes, the answers to these questions remain elusive.
Here, we set out to study the structure of a functional tmRNA–ribosome complex in the later stages of trans-translation. We purified the complex directly from the cell in order to study the native states in the process of trans-translation. By introducing strategic mutations for stabilization of the tmRNA–ribosome complex and using a novel classification method (Scheres et al, 2007) for sorting out the coexisting conformations, we were able to obtain the structure of the complex at the point where the TLD has moved to the P site and translation is resumed on the ORF of tmRNA. Our results show that, similar as in the pre-accommodation complex, the tmRNA forms a stable, albeit differently positioned ‘arc and fork' structure on the ribosome at this point. Our atomic model based on the cryo-EM density map defines the contacts of the tmRNA with the ribosome, and sheds light on the process by which the ribosome selects the correct reading frame on the tmRNA.
To study the native states in the trans-translation processes, we formed the tmRNA–ribosome complex in vivo. We took several measures to address the anticipated problem of heterogeneity in the specimen. To improve the occupancy of the tmRNA on the ribosome, we inserted an MS2-binding hairpin (hp) into the helix 5 of tmRNA (Figure 1), so that only those ribosome complexes that were bound with tmRNA were purified. To reduce the structural heterogeneity, we mutated the 4th and 5th sense codons of the MS2-tmRNA to UAA stop codons, expecting that the complex formed by this tmRNA mutant and the ribosome would be less flexible because translation would be stopped at an earlier stage, thus reducing the number of different possible conformations of the complex.
Initial reconstruction of the complex showed a 70S ribosome structure with an extra density around the beak of the 30S subunit (Supplementary Figure S1A). Fragmented densities, which cannot be attributed to the 70S ribosome, show up at reduced threshold, suggesting the presence of structural heterogeneity (data not shown). Thus, we applied the maximum-likelihood (ML) classification method (Scheres et al, 2007) to sort the data set into multiple classes (Supplementary Figure S1). Separate reconstructions of the classes generated multiple distinct density maps with resolutions in the range of 13–18 Å (Supplementary Figures S1 and S2; Supplementary Table S1). For further structural analysis, we excluded those maps that contain fragmented densities for tmRNA (Structures 1–5 in Supplementary Figure S1). The remaining map is reconstructed from 20 873 projection images, and has a resolution of 13.6 Å. The crystal structure of the Escherichia coli 70S ribosome (Schuwirth et al, 2005) was fit into the cryo-EM density map using real-space refinement, a procedure which locally optimizes the cross-correlation between pieces of the atomic model and the density map (Gao et al, 2003). After obtaining an atomic model for this cryo-EM map, we were able to accurately subtract the densities attributed to the ribosome from the cryo-EM map. Part of the remaining density mass is situated at the A site and has the shape and size of a tRNA even when the map is displayed at high threshold, suggesting solid binding of a tRNA at the A site (Figure 2B).
The remaining density, after excluding the A-site tRNA, was compared with the tmRNA density isolated from the pre-accommodation complex (Kaur et al, 2006) (Figure 3). The similarity of the two maps can be readily seen: both maps contain the arc structure, which is connected to a fork structure that extends to the inter-subunit space of the ribosome. In the density of the pre-accommodation complex (Kaur et al, 2006), the fork structure extends to the TLD in the A/T position, whereas in the density we have now obtained, the fork structure extends to a ligand at the ribosomal P site. Based on the similarity of this density with the previous cryo-EM maps, and the connectivity of the TLD with the rest of the tmRNA, we conclude that this density must be attributed to the tmRNA, and specifically, the P-site ligand mass must be attributed to the TLD. In addition, when displayed at lower threshold, a thin strand of density, which is attributed to a highly stabilized single-stranded RNA, can be seen extending from the anticodon end of the A-site tRNA into the mRNA entry channel. This indicates the incorporation of the ORF into the ribosome (Supplementary Figure S3). Thus, the complex is apparently in the state where TLD has been translocated to the P site and translation has resumed on the first ORF codon. We confirmed this conclusion with northern blot analysis (see Materials and methods): the tRNAs in the tmRNA–ribosome complex is predominantly Ala-tRNA (Supplementary Figure S4), thus the first codon of the ORF on the mRNA-like domain must be at the A site. We term this state the resume state and the corresponding complex the resume complex.
To interpret the density of the tmRNA, we built an atomic model of the tmRNA using homology and ab initio modelling techniques, starting from 2D structural alignment and X-ray structures of homologous domains deposited in the PDB (see Experimental Procedure and Supplementary Figure S5). Based on biochemical data and structural evidence (Gutmann et al, 2003; Valle et al, 2003), we were able to determine the locations of these components unambiguously and place the atomic model into the density map. Despite the good fit of our model in the density map, several runs of molecular dynamic flexible fitting, which uses the density map as a force potential to drive models into the map during simulation, were additionally performed (Trabuco et al, 2008). The resulting conformation displays a better cross-correlation coefficient (CCC) and root mean square deviation (RMSD) with the density map than the initial model (Supplementary Figure S6).
With all the components of the tmRNA assigned, we can analyse the large conformational rearrangement that occurs during the transition from the pre-accommodation state to the resume state. Apparently, although one end of the arc, pseudoknot PK2, remains bound at similar positions on the 30S subunit, the other end, pseudoknot PK1, moves towards the inter-subunit space and down towards the codon recognition centre. Simultaneously, helix 2 and TLD moves all the way into the inter-subunit space (Figure 3). Evidently, the large conformational change enabled both the translocation of the TLD to the P site and the insertion of the ORF, which is downstream of pseudoknot PK1 but upstream of pseudoknot PK2.
As mentioned above, the TLD is located at the ribosomal P site. It is known that TLD and bound SmpB jointly mimic the structure of tRNA (Gutmann et al, 2003; Cheng et al, 2010). We observed that the P-site ligand has weaker density near the decoding centre. As protein results in weaker EM density than RNA, we conclude that in the resume state complex, SmpB mimics the anticodon arm of the normal P-site tRNA (Figure 4). We fitted the crystal structure of the Thermus thermophilus SmpB into our E. coli SmpB density. As shown in Figure 4, the density appears to be bigger than the crystal structure. We can explain this discrepancy as due to the fact that E. coli SmpB has an extra segment of 26 amino acids on the C-terminal tail.
The SmpB appears to have an instrumental function in establishing the correct reading frame during trans-translation. At the 30S ribosomal P site, the SmpB density has direct contact with at least five nucleotides upstream of the resume codon, comprising the −1 triplet the U85 and the A86 nucleotides (Figure 4). As a result, the resume codon is precisely placed at the ribosomal A site.
By reconstructing the structure of a tmRNA–ribosome complex formed in vivo, we obtained the first structures of the tmRNA–ribosome complex in the resume state, the state where translation is ready to resume on the ORF of the tmRNA. Our structure of the resume complex shows that tmRNA maintains a stable structure, at least up to this point.
Based on the cryo-EM density, we built an atomic model that allows for interpretation of the structure in great details. Generally, pseudoknots are known to be stable and robust structures. In the tmRNA model, successive pseudoknots are linked to each other by single-stranded regions and form an arc-like structure, giving the tmRNA the flexibility required for undergoing the observed large conformational changes from the pre-accommodated to the resume state, despite the structural robustness of the constitutive pseudoknots. The TLD displays clear structural resemblance with tRNA. It is connected to the arc of the pseudoknots by helix 2, which includes several mismatches and internal loops.
The fact that we observe only the resume complex suggests that the translation process is slowed down in the resume state, and once the complex has passed through this state, translation will continue until the tmRNA–ribosome complex dissociates. Our atomic model suggests that the insertion of the MS2 hp may account for the slowing down of the process. The MS2 hp appears to stabilize helix 5, keeping it from unwinding. At least five of the ORF codons, including the stop condon, are contained in the structure of the helix 5 and its apical loop. Therefore, part of the helix 5 needs to be unpaired in order to be rendered completely accessible for translation by the resume-state ribosome.
How the correct reading frame is established, at the point where the ribosome switches from the stalled mRNA to the ORF of the tmRNA, has remained a crucial question in the study of trans-translation. As described in the Results section, our cryo-EM density and atomic model show that the SmpB interacts with the upstream region of the resume codon, thus presenting the resume codon at the A site. This is consistent with several previous studies, which showed that mutations on the five nucleotides immediately upstream of the resume codon cause frame shifting and partial or complete abolition of the tmRNA activity (Williams et al, 1999; Lee et al, 2001; Trimble et al, 2004; Lim and Garber, 2005; Konno et al, 2007; Miller et al, 2008). On the other hand, SmpB, which has been shown to bind with the upstream region of the resume codon (Konno et al, 2007; Metzinger et al, 2008), appears to contribute to the reading frame selection. One study showed that certain mutations on SmpB restore the function of the tmRNA with mutation on its resume codon upstream sequence (Watts et al, 2008, 2009).
Our atomic model, combined with the aforementioned genetic and biochemical studies, suggests that the reduction, abolition and frame shifting of the trans-translation activity is due to the incorrect presentation of the resume codon to the A site, subsequent to the alteration of the interaction between the SmpB and the resume codon upstream region. To investigate the details of the interaction between the SmpB and tmRNA, we modelled the extra C-terminal segment, which is present in the E. coli SmpB but not in the T. thermophilus SmpB, as a long helix (see Materials and methods), and fitted this helix into the cryo-EM density along with the crystal structure of the T. thermophilus SmpB. Our fitted model suggests that the −1 triplet, the U85 and A86 nucleotides may interact with residues 134–140 of the SmpB. It is also possible that other contacts may be established, through amino acids 18–24 of the SmpB and U85 and A86 nucleotides of the tmRNA (Supplementary Figure S7). Because of the limited resolution of our cryo-EM density, the exact nature of the interaction between the upstream region and the SmpB has not been addressed. However, because of the relative tolerance to modifications of the upstream sequence (Miller et al, 2008), the interaction could rely partly on contacts with the nucleotide's backbone. The length of the upstream sequence, according to our model, may also be a determinant for the optimal interaction with the SmpB, and consequently, for the presentation of the resume codon at the A site.
Another major question to be answered in the study of trans-translation is how tmRNA passes through the ribosome despite its complex topology. Our structures show that tmRNA has a great degree of structural flexibility: as mentioned above, in five of the six structures obtained by ML classification, the tmRNA densities are fragmented, whereas the density of the tRNA and the ribosome are solid. As all our complexes have TLD in the P site, the fragmentation seems to suggest that tmRNA starts to unfold after the TLD is translocated to the P site. On the other hand, our atomic model shows that the tmRNA is composed of stable structural elements that are difficult to unwind. In addition, the atomic model does not seem to be subjected to significant topological constraints, at least up to the point where tmRNA is translocated to the E site. More structural studies need to be done to further investigate this question.
MS2-tmRNA(H8), a variant of E. coli tmRNAH8 that can bind the coat protein from MS2 bacteriophage, was constructed by PCR-directed mutagenesis as described earlier (Wower et al, 2004). The resulting mutant is called MS2-tmRNA-4 (see also Figure 1B and Supplementary Materials and methods).
E. coli strain IW363 was transformed with plasmid pETrpmA-At-3, which carries one copy of an rpmA gene encoding a truncated ribosomal protein L27, one copy of SmpB gene and one copy of modified ssrA gene encoding MS2-tmRNA-4 (Wower et al, 2004). Synthesis of truncated ribosomal protein L27 and following tagging activity was induced by addition of 1 mM IPTG to logarithmically growing cells and incubation was carried for 3 h. The expression of truncated protein L27 and its tagging were analysed by SDS–polyacrylamide electrophoresis as described in Wower et al (2004).
The identity of the tRNA was established by Northern hybridization with 32P-labelled probes directed against E. coli tRNAAla, tRNAAsn or tRNAHis. All RNAs were synthesized in vitro using T7 RNA polymerase and purified by denaturing PAGE. See Supplementary Materials and Methods for synthesizing tRNAAla, tRNAAsn and tRNAHis. Total RNA was extracted from two A260 units of purified tmRNA-70S ribosome complexes using the RNAqueous-Micro kit (Ambion). In all, 1/20 of the extract was resolved on a 10% urea-polyacrylamide gel with 50 ng of purified tRNA transcripts. RNAs were blotted onto Zeta-Probe membrane in a semi-dry cell (Bio-Rad). Hybridizations were carried out in UlraHyb-Oligo (Ambion) solution with 32P-labelled oligonucleotide probes: 5′-GAC CTC CTG CGT GC-3′ (for tRNAAla), 5′-GAC ATA CGG ATT AAC AG-3′ (for tRNAAsn) and 5′-GGA ATC ACA ATC CAG G-3′ (for tRNAHis). Northern analyses were quantified using Typhoon PhosphoImager and ImageQuant 5.2 software. E. coli tRNAAla was identified as the prominent species in the complexes. Only trace amounts of E. coli tRNAHis were observed.
The tmRNA–ribosome specimen was diluted into a final concentration of 32 nM. A cryo-grid (carbon-coated Quantifoil 2/4 grid, Quantifoil Micro Tools GmbH, Jena, Germany) was prepared following standard cryo-procedures (Grassucci et al, 2007). Microscopy was performed under low-dose conditions on an FEI (FEI, Hillsboro, OR) Tecnai Polara instrument operating at 300 kV and a nominal magnification of × 59 000. Micrographs were recorded on a 4 × 4 k 16-bit TVIPS TemCam-F415 CCD camera, with a physical pixel size of 15 μm by using the automated data collection system AutoEMation (Lei and Frank, 2005). The effective magnification on the CCD was × 100 000 due to a post-magnification ratio of 1.7, thus making the pixel size 1.5 Å on the object scale. Micrographs were subsequently decimated to a 3 Å pixel size to boost the low-spatial frequency signal and reduce the processing time. Ribosome particles were windowed out from the micrographs using an automatic particle picking program (Rath and Frank, 2004). After manual verification, 90 894 particles were selected for single-particle classification and reconstruction, following standard SPIDER protocols for reference-based reconstruction (LeBarron et al, 2008; Shaikh et al, 2008). ML classification (Scheres et al, 2007) was performed using the XMIPP package (see Supplementary Materials and methods). The classification was done in a hierarchical manner as described in Supplementary Figure S1. The fitting of the atomic model was done in multiple steps using a combination of manual and real-space refinement fitting, as described in the text. The real-space refinement program used was RSREF (Chapman, 1995; Gao et al, 2003), and the initial atomic models were from the crystal structure of the E. coli ribosome (PDB codes: 2AVY and 2AW4 (Schuwirth et al, 2005)). The head of H38 was taken from PDB structure 1PNY (Vila-Sanjurjo et al, 2003).
Both homology and ab initio modelling techniques were used to build the 3D atomic model of the tmRNA. The initial structural alignment was retrieved on the Rfam website (http://rfam.sanger.ac.uk/), then realigned in reference to the E. coli tmRNA sequence with the S2S tool (Jossinet and Westhof, 2005), against 228 tmRNA sequences (Supplementary Figure S5). The 3D atomic model was built with Assemble (Jossinet et al, 2010) and regular helices were generated directly from the sequence. Pseudoknots were also built with Assemble, based on existing structures of pseudoknots found by browsing the non-redundant database of X-ray structures. The TLD region was modelled by homology by applying the fold of the TLD crystal structure from T. thermophilus (Bessho et al, 2007). The crystal structure of the T. thermophilus SmpB (PDBID: 2CZJ) was fitted into our cryo-EM density map after verification of the good sequence homology between the T. thermophilus and E. coli SmpB. The remainder of the E. coli C-terminal tail is predicted to be a long helical segment by the SYMPRED consensus secondary structure prediction web service (http://www.ibi.vu.nl/programs/sympredwww/) with the PROF (Rost, unpublished data), SSPRO (Pollastri et al, 2002), YASPIN (Lin et al, 2005) and PSIPRED (McGuffin et al, 2000) methods. A 20 amino acids long α helix was modelled using PyMol (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC) and fitted into the density to demonstrate the possibility of the interaction with the resume codon upstream region.
The system was set up in VMD by adding a solvent box of ~4 00 000 TIP3P water molecules by using the SOLVATE module implemented in VMD, and neutralized by adding K ions. A 0.2 mol/l of K+ ions has been added to the neutralized system. Ions were added using the ADDIONS module in VMD. CHARMM force field parameters were used (Combined CHARMM All-Hydrogen Topology File for CHARMM22 Proteins and CHARMM27 Lipids (MacKerell et al, 1998)), and the simulation was conducted using NAMD (Phillips et al, 2005). Trajectories of several nanoseconds were obtained for the cryo-EM density of the resume state. The final fitted structures were selected based on the RMSD and the CCC, after reaching a plateau within 400 ps of simulation time (Supplementary Figure S6).
All figures were rendered with UCSF Chimera (Pettersen et al, 2004).
PDB IDs: the coordinates for the modelled and refined E. coli tmRNA has been deposited in the Protein Data Bank with accession numbers: 3IZ4. The cryo-EM maps have been deposited at the 3D-EM database, EMBL with accession numbers: EMD-5234.
We thank Bob Grassucci for assistance with the microscopy, Lila Rubinstein and Melissa Thomas for assistance in the preparation of the illustrations, Fabrice Jossinet and Eric Westhof for their assistance in modelling the tmRNA molecule and Jiang Zhu for helpful discussions. This work was supported by HHMI and NIH R01 GM29169 (to JF), P01 GM064692 (to Robert M Glaeser) and R01 GM58267 (to JW).
The authors declare that they have no conflict of interest.