Contributions of Pseudoknots and Protein SmpB to the Structure and Function of tmRNA in trans -Translation*

Bacteria contain transfer-messenger RNA (tmRNA), a molecule that during trans -translation tags incom-pletely translated proteins with a small peptide to signal the proteolytic destruction of defective polypeptides. TmRNA is composed of tRNA- and mRNA-like domains connected by several pseudoknots. Using truncated ribosomal protein L27 as a reporter for tagging in vitro and in vivo , we have developed exceptionally sensitive assays to study the role of Escherichia coli tmRNA in trans -translation. Site-directed mutagenesis experiments showed that pseudoknot 2 and the abutting helix 5 were particularly important for the binding of ribosomal protein S1 to tmRNA. Pseudoknot 4 not only facili-tated tmRNA maturation but also promoted tagging. In addition, the three pseudoknots (pk2 to pk4) were shown to play a significant role in the proper folding of the tRNA-like domain. Protein SmpB enhanced tmRNA processing, suggesting a new role for SmpB in trans -translation. Taken together, these results provide unanticipated insights into the functions of the pseudoknots and protein SmpB during tmRNA folding, maturation, and protein synthesis. An interruption of the elongation step of protein synthesis results in the production of truncated proteins and leaves ribosomes stalled at the 3 (cid:1) end of mRNA templates lacking a stop codon(s). To remove the defective polypeptides and recycle

An interruption of the elongation step of protein synthesis results in the production of truncated proteins and leaves ribosomes stalled at the 3Ј end of mRNA templates lacking a stop codon(s). To remove the defective polypeptides and recycle the ribosomes, bacteria have developed trans-translation, a quality control mechanism that tags the C termini of defective proteins with a short peptide recognized by housekeeping proteases. This peptide tag is encoded by a short open reading frame present in a small stable RNA molecule called 10Sa or transfermessenger RNA (tmRNA). 1 It has been shown that in addition to protein tagging, tmRNA facilitates the recycling of ribosomes by providing missing stop codons (1,2).
Structure probing of the Escherichia coli tmRNA and sequence comparisons demonstrated the presence of three domains ( Fig. 1A) (3)(4)(5). The 3Ј and 5Ј termini of the tmRNA form the tRNA-like domain (TLD) with a significantly reduced D arm. The resume codon and stop codon(s) demarcate the open reading frame in the mRNA-like domain (MLD). The TLD and the MLD are connected by a pseudoknot-rich domain consisting of four pseudoknots (pk1 to pk4) in most tmRNAs.
The three-dimensional model of E. coli tmRNA suggests a structure in which the TLD is connected to the circularly arranged MLD and pseudoknots through coaxially stacked helices (6). Recently, the entry of tmRNA into a stalled E. coli ribosome has been visualized by cryo-electron microscopy (7). At this particular step of trans-translation, the TLD, pk1, and the MLD contact the ribosome, whereas the pk2 to pk4 segment forms an arc that remains outside the ribosome.
Three proteins facilitate binding of tmRNA to the ribosome. Elongation factor Tu forms a ternary complex with GTP and aminoacyl-tmRNA, as in regular protein synthesis (8,9). Protein SmpB binds to the ternary complex in vitro as well as to stalled ribosomes in vivo (10 -12). Ribosomal protein S1 contacts the MLD and the pseudoknot-rich domain both on and off the ribosome (13).
Although pseudoknots are predominant tmRNA features, little is known about their contributions to tmRNA structure and function. Previous in vitro experiments suggested that pk1 is essential for tmRNA folding and protein tagging, whereas the three remaining pseudoknots, pk2 through pk4, are interchangeable and replaceable with stretches of single-stranded RNA (14 -16). However, these data were derived exclusively using an insensitive assay in which poly(U)-programmed ribosomes produce hydrophobic polypeptides (polyphenylalanine) that are inefficiently tagged and because of their heterogeneity are difficult to analyze qualitatively.
To understand how tmRNA facilitates the tagging of truncated proteins, we investigated the contributions of the pseudoknots pk2 to pk4 and protein SmpB. The key to achieving our goals was development of two exceptionally sensitive assays for testing the functionality of tmRNA in vivo and in vitro. A truncated ribosomal protein, L27, produced by ribosomes programmed with mRNA lacking stop codons, was tagged with a protease-resistant histidine-rich peptide in the presence or absence of protein SmpB. A stable fusion protein was easily detected in fractionated E. coli lysates either by staining with Coomassie Blue or by using anti-His-tag antibodies. We found that disrupting pseudoknots pk2 to pk4 had differential effects not only on tmRNA tagging activity and on binding of ribosomal protein S1 to tmRNA but also on tmRNA maturation. Taken together, these results indicated that pk2, pk3, and pk4 are important for the proper overall folding of the tmRNA. The ability of SmpB to reverse certain defects in tmRNA processing suggests that this protein plays a hitherto undiscovered regulatory role in trans-translation.

EXPERIMENTAL PROCEDURES
Bacterial Strains-E. coli strain XL1-B was the host for purification and maintenance of new plasmids. Expression strains IW410 and * This research was supported by National Institutes of Health Grant GM58267 (to J. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Construction of Plasmids-The two-step PCR cloning method described by Chen et al. (18) was used. Plasmid pACssrA(cam r ) was constructed by cloning of the E. coli ssrA gene with its flanking regions (133 bp upstream and 157 bp downstream) from E. coli K12 genomic DNA into plasmid pACYC184. The DNA segment was amplified using PR1 and PR2, digested with HindIII and BamHI and inserted into the pACYC184 vector.
Plasmid pETrpmA-At-1 contained the E. coli rpmA gene in which the stop codon was replaced with the trpA terminator sequence (19). To construct this plasmid, the rpmA gene obtained from plasmid pETrpmA was amplified using PR3 and PR4 and cloned into pET-23a (see Fig.  2A) (20). Plasmid pETrpmA-At-2 was a derivative of pETrpmA-At-1 into which the ssrA gene with its flanking regions (133 bp upstream and 157 bp downstream) was inserted. A PCR fragment containing the ssrA gene was cloned into the PvuII site in pETrpmA-At-1.
Plasmid pETrpmA-At-3 was prepared by inserting the smpB gene upstream of the ssrA gene in plasmid pETrpmA-At-2 to obtain their chromosomal configuration. Primers PR5 and PR6 were used to amplify the insert.
Mutagenesis-To replace the rpmA-trpAt gene with wild-type rpmA sequence in plasmid pETrpmA-At-3, mutagenic primer PR7 was used. The resulting plasmid pETrpmA-3 was suitable for expression of protein T7-L27 in vivo.
For replacing the sequence encoding the proteolytic peptide ANDE-NYALAA with a sequence encoding ANDEHHHHHH (H 6 tag) in E. coli ssrA gene in all pETrpmA-type plasmids, PCR-directed mutagenesis was carried out with PR8 and PR9. Similarly, for the replacement with ANHHHHHHHH (H 8 tag), PCR-directed mutagenesis was carried out with PR10 and PR11.
The base pairing in helix 5 of tmRNA(H 6 ) and tmRNA(H 8 ) was restored using PCR-directed mutagenesis with PR12. The resulting mutant ssrA genes encoded tmRNA(H 6 hp) and tmRNA(H 8 hp).
For the in vitro tagging experiments, the ssrA gene in plasmid ptmR was replaced with its derivative encoding tmRNA(H 8 hp) (13). All dele- tion mutations listed above were re-cloned into plasmid ptmRH 8 hp using restriction enzymes SphI and PvuII. The mutations were verified by DNA sequencing.
In Vitro RNA Synthesis and Analysis-tmRNAs were transcribed in vitro, purified, 32 P-labeled, aminoacylated, and tested for binding to protein S1 as described earlier (13,21).
Protein Tagging in Vivo-In initial in vivo tagging experiments the rpmA and ssrA genes were expressed simultaneously from the plasmids pETrpmA-At-1 and pACssrA in E. coli strains BL21(DE3) and IW410. Most of the in vivo tagging experiments were performed in E. coli IW363 cells transformed with either pETrpmA-At-2 or pETrpmA-At-3. Freshly transformed cells were grown at 37°C for 2 h in 2ϫYT broth supplemented with the appropriate antibiotics (kanamycin, 50 l/ml; chloramphenicol, 30 l/ml; ampicillin, 200 l/ml) to 0.3-0.4 A 600 . Cultures were diluted to an A 600 of 0.05 and incubated in the absence or presence of 1 mM IPTG for 3 h. 1-ml aliquots of each cell culture were collected for RNA and protein analysis.
Cells were lysed and fractionated on 12.5% SDS-polyacrylamide gels as described (23). Proteins were detected by staining with Coomassie Brilliant Blue R-250. When tagging was inefficient, tagged proteins were purified using Ni-NTA-Sepharose according to the manufacturer's procedure (Qiagen), fractionated by SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and visualized by Western blotting with anti-Histag antibodies, as suggested by the manufacturer (Bio-Rad).
Protein Tagging in Vitro-Tagging in vitro was performed with the E. coli T7 transcription/translation system for circular DNA following the instructions provided by Promega. A typical 20-l sample contained 2 g of plasmid pETrpmA-At-1, 20 pmol/l of purified tmRNA transcript, and 20 pmol/l of purified protein SmpB-H 6 (25). Tagged proteins were purified on Ni-NTA-Sepharose using the denaturing conditions recommended by Qiagen, resolved on a 12.5% SDSpolyacrylamide gel, and blotted onto a polyvinylidene difluoride membrane (Bio-Rad). Tagged proteins were identified with anti-His-tag monoclonal antibodies from Qiagen and horseradish peroxidase-conjugated goat anti-mouse IgG from Jackson ImmunoResearch Laboratories Inc. followed by detection with an ECL-Plus Western blotting system from Amersham Biosciences.

RESULTS
Monitoring tmRNA-dependent Tagging of Truncated Proteins in Vitro-Ribosomal protein L27, encoded by the rpmA gene, can be overexpressed in E. coli without detrimental effects (24). Thus, a truncated L27 encoded by mRNA lacking a stop codon(s) was expected to be suitable for studying tmRNA-dependent tagging in cell lysates by detection of polypeptides using SDS-PAGE and staining with Coomassie Blue. Stable expression of tagged L27 was assured by replacing the ANDENYALAA proteolysis tag with a protease-resistant peptide. We chose a string of six or eight histidines (His tag) to allow the isolation and optional detection of even minute quantities of the tagged fusion protein by affinity chromatography on Ni-NTA-Sepharose and anti-Histag antibodies, respectively.
The main components of the in vivo assay are the three plasmids, pETrpmA-At-1, -2, and -3 ( Fig. 2A), which direct the expression of truncated L27 either in the presence or absence of tmRNA and protein SmpB. Plasmid pETrpmA-At-1 was constructed by deleting the His-tag-encoding segment from pET-23a (Novagen) followed by inserting a modified rpmA gene under control of the T7 promoter. The 5Ј end of the rpmA gene was fused to the T7-tag, a DNA segment that encodes the N-terminal segment of the T7 major capsid protein. The stop codon of the rpmA gene was replaced with trpAt, an efficient trpA terminator (19). These manipulations yielded T7-rpmA-At, a modified rpmA gene encoding T7-L27-At, a 11,226-Da polypeptide consisting of 105 amino acids. T7-L27-At, when tagged by the tmRNA mutants with the protease-resistant ANDEHHHHHH (H 6 tag) or AANHHHHHHHH (H 8 tag) (see Fig. 1B), produced the 116-amino acid long proteins T7-L27-H 6 (M r 12,444) or T7-L27-H 8 (M r 12,474 Da). Because T7-L27-At and its tagged derivative can be separated by SDS-PAGE, tagging is easily monitored (Fig. 2C).
Expression of protein T7-L27-At in IW410, an E. coli strain lacking the ssrA gene, was prominent (Fig. 3). In contrast, in the ssrA gene-containing BL21(DE3) strain, T7-L27-At was degraded, demonstrating that this protein is suitable for monitoring tmRNA-directed tagging.
To prevent the degradation of tagged T7-L27-At, we constructed tmRNA(H 6 ) and tmRNA(H 8 ), encoding protease-resistant H 6 and H 8 tags (Fig. 1B). The alterations were introduced into plasmid pACssrA where the ssrA gene is under control of its own promoter. As shown in Fig. 4, tmRNA(H 8 ) was expressed less efficiently than the wild-type tmRNA. A similarly reduced level of expression was observed for tmRNA(H 6 ) (not shown). We attribute this effect to the disruption of base pairing in helix 5 because higher expression levels were observed with wild-type tmRNA and tmRNA(H 8 hp) in which helix 5 is restored (see Fig. 1B). The mechanism for how the disruption of helix 5 reduces the expression of the tmRNA(H 6 /H 8 )-encoded peptides remains to be resolved.
To tag T7-L27-At in vivo, E. coli IW410 cells were transformed simultaneously with pETrpmA-At-1 and derivatives of pACssrA encoding either wild-type tmRNA or tmRNA(H 6 /H 8 ). The synthesis of truncated L27 was induced in logarithmically growing cells by the addition of IPTG. Cells continued to grow for 3 h, and their protein content was analyzed by SDS-PAGE. As shown in Fig. 5, similar sets of proteins were apparent in uninduced and induced cells, indicating that the truncated L27 was tagged by plasmid-borne wild-type tmRNA and then proteolytically degraded. In contrast, a prominent band corresponding to T7-L27-At was visible in lysates of the IPTGinduced cells expressing either tmRNA(H 6 ) or tmRNA(H 8 ). Using Coomassie Blue staining, tagged proteins could not be detected.
To demonstrate that tmRNA(H 6 ) and tmRNA(H 8 ) were able to tag in vivo, it was necessary to use 35 S-labeled T7-L27-At, affinity chromatography, and anti-His-tag antibodies. IW410 cells were transformed simultaneously with plasmids pETrpmA-At-1 and pACssrA and treated with IPTG in a medium containing [ 35 S]Met as described (26). The 35 S-labeled, His-tagged T7-L27-At proteins were captured on Ni-NTA-Sepharose and analyzed by SDS-PAGE. Protein T7-L27-At was tagged very poorly by tmRNA(H 6 ) and tmRNA(H 8 ) (see Fig. 6). Northern blot analysis revealed that a substantial portion of the mutant tmRNAs was not processed (Fig. 7) and was likely to account for some of the observed poor tagging of protein T7-L27-At.
E. coli strains BL21(DE3) and IW410, when transformed simultaneously with plasmids pETrpmA-At-1 and pACssrA, displayed an undesirable variability in their protein expression profiles (not shown). To increase the reproducibility of these experiments, we inserted the ssrA gene into plasmid pETrpmA-At-1 to form pETrpmA-At-2 ( Fig. 2A). Because the rmpA and ssrA genes were part of the same molecule, their products were expected to be expressed at a constant ratio. To increase the expression levels of tmRNA(H 6 ) and tmRNA(H 8 ), base pairing in helix 5 was restored by introducing compensatory mutations (see Fig. 1B). These modifications resulted in tmRNA(H 6 hp) and tmRNA (H 8 hp), which expressed as efficiently as wild-type tmRNA (not shown). Precursors of tmRNA(H 6 hp) and tmRNA(H 8 hp) were trimmed to their mature form as efficiently as the precursors of wild-type tmRNA (compare Figs. 7 and 15).
Tagging of protein T7-L27-At by tmRNA(H 6 hp) and tmRNA(H 8 hp) was tested in IW363, a derivative of E. coli strain BL21(DE3)/pLysS lacking the ssrA gene. As shown in Fig. 8A, inserting the ssrA and T7-rpmA-At genes into a single plasmid and restoring helix 5 in tmRNA(H 6 hp) and tmRNA(H 8 hp) improved the tagging efficiency of protein T7-L27-At to a level that allowed the detection of the fusion protein by Coomassie Blue staining. Effective His-tagging of protein T7-L27-At was confirmed by its capture on Ni-NTA-Sepharose and the use of anti-His-tag antibodies (Fig. 8B). Both tmRNA(H 6 hp) and tmRNA(H 8 hp) tagged protein T7-L27-At efficiently. Together, these findings demonstrated that, at least in E. coli, the disruption of helix 5 inhibits maturation of precursor tmRNAs and tmRNA-dependent tagging. Therefore, helix 5 constitutes a functionally important element of tmRNA structure and function.
Monitoring tmRNA-dependent Tagging of Truncated Proteins in Vitro-As shown above, certain alterations in tmRNA affected tmRNA-dependent protein tagging by preventing the processing of tmRNA precursors. To differentiate between defects in protein tagging and tmRNA maturation, we developed an in vitro assay that employed plasmids pT7tmRNAH 8 hp and pETrpmA-At-1. pT7tmRNAH 8 hp linearized with restriction enzyme BstNI provided a template suitable for in vitro synthesis of tmRNA(H 8 hp) by T7 RNA polymerase. pETrpmA-At-1, when added to an in vitro transcription/translation system, was used for the synthesis of stop-free mRNA encoding truncated protein L27. Although the in vitro tagging of truncated protein L27 by tmRNA(H 8 hp) was less efficient than the in vivo reaction, formation of the fusion protein could be easily monitored by fractionation of the reaction products on SDS-polyacrylamide gels followed by Western blotting with anti-His-tag antibodies.
During the optimization of the in vitro protein tagging assay, we found that in addition to plasmids pETrpmA-At-1 and tmRNA(H 8 hp), the transcription/translation reactions had to be supplemented with protein SmpB (Fig. 9). The most efficient tagging was observed when tmRNA(H 8 hp) and protein SmpB were provided in equimolar quantities. This finding was consistent with the earlier demonstration that protein SmpB is an essential component of the trans-translational apparatus in vitro (25,26).
Functions of Pseudoknots in tmRNA-dependent Protein Tagging-TLD and MLD are mimics of tRNA and mRNA, respectively, and function in trans-translation as expected (1,2,27,28). Pseudoknots, on the other hand, form in many of the larger RNAs where they are involved in a wide variety of biological processes (29 -31).
To determine the roles of the pseudoknots, we first tested the tagging potential of tmRNAs that lacked two or more pseudoknots. Three derivatives of tmRNA(H 8 hp) missing pk2 and pk3 (tmRNA⌬pk2/pk3), pk3 and pk4 (tmRNA⌬pk3/pk4), or pk2-pk4 (tmRNA⌬pk2-pk4) were unable to tag protein T7-L27-At in vivo (not shown). The tmRNA⌬pk3/pk4 construct was able to mediate tagging of protein T7-L27-At in vitro ( Fig.  10) but with greatly reduced efficiency. This result suggested that the loss of pk2 is more detrimental for tmRNA-dependent tagging than the depletion of pk3 or pk4.
Previously, ribosomal protein S1 was shown to be associated with the MLR and pseudoknots pk2 to pk4 (17). Therefore, we investigated the ability of purified E. coli S1 to bind to mutant tmRNAs that lacked various pseudoknots. TmRNAs were 32 Plabeled at their 3Ј ends with yeast ATP/CTP tRNA nucleotidyltransferase in the presence of [␣-32 P]ATP and aminoacylated with alanine as described (13). Typically 80 -95% of tmRNA⌬pk2/pk3, tmRNA⌬pk3/pk4, and tmRNA⌬pk2-pk4 could be charged (not shown), indicating that the TLDs of the mutant tmRNAs were properly folded (32).
To avoid widespread structural changes in the tmRNA molecule that may result from the pseudoknot excisions, we replaced pk2, pk3, and pk4 with single-stranded RNAs to form tmRNApk2L, tmRNApk3L, and tmRNApk4L, respectively. When expressed from plasmid pETrpmA-At-2 in vivo, all three mutant tmRNAs displayed levels of tagging that were undetectable by Coomassie Blue staining (not shown). Affinity purification followed by the use of anti-His-tag antibodies demonstrated that tmRNApk2L, tmRNApk3L, and tmRNApk4L tagged T7-L27-At at an ϳ20-fold reduced level compared with the tagging by tmRNA(H 8 hp) (Fig. 12). The decrease was only 5-fold lower than the tagging by tmRNA(H 8 hp) when the transcription/translation reactions were furnished with purified SmpB (Fig. 13).
The boost in tagging activity afforded by the addition of protein SmpB allowed us to test the tagging properties of tmRNAs in which pseudoknot pairs pk2 and pk3 (tmRNApk2/3L) and pk3 and pk4 (tmRNApk3/4L) as well as the pseudoknot triplet pk2-pk4 (tmRNApk2-4L) were replaced with single-stranded RNAs. As suspected, multiple mutations were more detrimental for protein tagging than disruptions of a single pseudoknot in tmRNApk2L, tmRNApk3L, and tmRNApk4L. Only tmRNApk3/4L displayed detectable tagging activity ϳ20-fold lower than the tagging by tmRNA(H 8 hp) (Fig.  13). Inactivity of both tmRNApk2/3L and tmRNApk2-4L highlighted the importance of pk2 for tagging.
Because protein SmpB enhanced protein tagging in vitro (see Fig. 13), we speculated that co-expression of tmRNA mutants and SmpB may also improve protein tagging in vivo. In pETrpmA-At-3 the smpB gene was inserted upstream of the ssrA gene in pETrpmA-At-2 to mimic the chromosomal organization of these two genes ( Fig. 2A). TmRNApk2L, tmRNApk3L, and tmRNApk4L, when co-expressed with protein SmpB, tagged so efficiently that protein T7-L27-H 8 could be easily detected by staining with Coomassie Blue (Fig. 14A). As much as 50% of protein T7-L27-At was tagged by tmRNA(H 8 hp). The levels of tagging by tmRNApk2L and tmRNApk3L were similar. However, the expression of protein T7-L27-At in cells producing tmRNApk3L was consistently lower than in cells expressing tmRNA(H 8 hp), tmRNApk2L, and tmRNApk4L. The tagging by tmRNApk4L, tmRNApk2/3L, and tmRNAp3/4L was significantly impaired and was detected only with anti-His-tag antibodies (Fig. 14B). The tmRNApk2-4L could not tag at all. Together, the observations indicated that pk4 is very important for protein tagging.
The tagging patterns seen in Figs. 13 and 14 suggested that certain defects displayed by tmRNAs with disrupted pseudoknots might be related to tmRNA maturation. To test this possibility we isolated total RNA from IW363 cells before and after induction of protein T7-L27-At synthesis with IPTG. The RNAs were examined by Northern blot analysis with an oligonucleotide probe complementary to nucleotides 54 -67 of E. coli tmRNA. As seen in Fig. 15, E. coli coped very well with processing of tmRNA precursors before IPTG induction. More than 90% of precursor tmRNA(H 8 hp), tmRNApk2L, tmRNApk3L, and tmRNApk2/3L were trimmed to their mature forms. Maturation of precursor tmRNApk4L, tmRNApk3/4L, and tmRNApk2-4L was less effective, but still, more than 70% of the molecules were processed. Upon IPTG induction, the processing of precursor tmRNApk4L, tmRNApk3/4L, and tmRNApk2-4L was affected the most. More than 50% of these tmRNA derivatives remained unprocessed. Because pk4 was disrupted in these tmRNA mutants, the maturation pattern shown in Fig. 15 indicated that pk4 is   FIG. 10. In vitro tagging of truncated protein L27 by  tmRNA(H 8 hp) mutants lacking pk2-pk3, pk3-pk4, and pk2-pk4. Each tagging reaction contained equimolar amounts of tmRNA and His-tagged protein SmpB. Tagging was visualized by Western blotting with anti-His-tag antibodies. ⌬pk2/pk3, ⌬pk3/pk4, and ⌬pk2-pk4 denote mutant tmRNA lacking pk2-pk3, pk3-pk4, and pk2-pk4, respectively. Lane M shows molecular mass markers. For more details, see the legend for Fig. 9. C, control.  12. In vivo tagging of truncated protein L27 by  tmRNA(H 8 hp), tmRNApk2L, tmRNApk3L, and tmRNApk4L. Logarithmically growing E. coli IW363 harboring plasmid pETrpmA-At-2 were induced with 1 mM IPTG. Cell lysates were fractionated by electrophoresis on a 12.5% SDS-polyacrylamide gel. Tagging was visualized by Western blotting with anti-His-tag antibodies. Two independent clones were tested for each mutant to highlight reproducibility of the assay. tmRNA* denotes tmRNA(H 8 hp). more important for tmRNA processing than pk2 and pk3.
Because protein SmpB improved tmRNA-dependent tagging in vivo, we compared the maturation of mutant tmRNAs with a single pseudoknot disrupted in the presence and absence of SmpB. Total RNA was isolated from the IPTG-induced IW363 cells harboring plasmids pETrpmA-At-2 or pETrpmA-At-3 and examined by Northern blot analysis. This analysis demonstrated that the majority of the precursor tmRNApk2L and tmRNApk3L and 90% of tmRNApk4L remained unprocessed when SmpB was absent (Fig. 16). In contrast, in the presence of protein SmpB, more than 85% of precursor tmRNA(H 8 hp), tmRNApk2L, and tmRNApk3L was processed to their mature forms. Most of the tmRNApk4L mutant remained unprocessed, confirming that pk4 plays a prominent role in tmRNA maturation. DISCUSSION For protein tagging, the 457-nucleotide precursor of E. coli tmRNA must be trimmed to its mature form, aminoacylated, delivered to the ribosome, and then perform its dual tRNA-and mRNA-like functions (2,17,28,33). We have studied these processes in vivo and in vitro by using suitably modified tmRNA derivatives with focus on the relative contribution of pseudoknots pk2 to pk4, helix 5, and proteins SmpB.
Using novel sensitive assays for monitoring tmRNAdependent tagging of truncated protein L27, we demonstrated that deletions and disruptions of pseudoknots pk2-pk4 and helix 5 impair protein tagging in vitro and in vivo. The level of tagging was different for each modification. For example, tmRNA⌬pk2/pk3 and tmRNA⌬pk2-pk4, mutants lacking pk2, did not tag in vitro. TmRNApk3L tagged well but inhibited the expression of the truncated protein L27 in vivo. TmRNApk4L, tmRNApk3/4L, and tmRNApk2-4L, all of which contained a disrupted pk4, showed no tagging activity in vivo. However, tmRNApk4L was as active as tmRNApk2L and tmRNApk3L in tagging in vitro.
Differences between the in vitro and in vivo tagging profiles suggested that different functions of tmRNA may be impaired (compare Figs. 13 and 14). Indeed, further testing using gel mobility shift assays demonstrated that the deletion of pk2 in particular weakened the interactions between tmRNA and ribosomal protein S1, suggesting that pk2 may be an important S1 binding site in tmRNA. This result is consistent with footprinting studies, which demonstrated a decreased reactivity of pk2 toward single-strand-specific probes when tmRNA is bound to protein S1 in vitro (34). Furthermore, cryo-electron microscopy revealed that pk2 remains available for interac- . Northern blot analysis of 1 g of total RNA extracted from IPTG-induced E. coli IW363 cells separated on a 5% denaturing polyacrylamide gel and blotted to a Zeta-probe membrane. 32 P-Labeled oligonucleotide TM4 was hybridized to both precursor tmRNA (p-tmRNA) and mature tmRNA. The lower panel shows the graphical representation of PhosphorImager-derived data from the Northern blot. Open bars, p-tmRNA; filled bars, tmRNA. tions with protein S1 when tmRNA enters the ribosome (7).
Our work demonstrated that mutant tmRNAs with alterations in helix 5 and the pk2-pk4 region adversely affect maturation of tmRNA precursors. Because formation of the doublestranded acceptor stem is a prerequisite for trimming the termini of the precursor tmRNA (17,33), this finding implied that disruption of pseudoknots impairs the folding of the 3Ј and 5Ј termini of tmRNA into a tRNA-like structure. However, pk2 to pk4 were not essential for tmRNA maturation. As shown in Fig. 15 (upper panel, last lane), even upon disruption of base pairing in the pk2-pk4 segment E. coli was able to process ϳ50% of the tmRNApk2-4L precursors to mature, taggingincompetent RNA molecules.
Because we observed similar cellular levels of vigorously tagging mature tmRNA(H 8 hp) and tagging-impaired mature tmRNA mutants, at least one pseudoknot, preferably pk4, is required for tmRNA-directed protein tagging. In contrast, it was suggested previously that pk2, pk3, and pk4 are interchangeable and can be replaced with single-stranded RNAs without detriment to tmRNA tagging (16). However, these data were derived exclusively from inefficient in vitro tagging experiments. The differences in protein tagging patterns shown in Figs. 12 and 14 underscore the need to investigate the structure and function of tmRNA in vitro as well as in vivo.
Co-expression of protein SmpB with tmRNA significantly improved the maturation of tmRNA precursors, suggesting an unanticipated role for this protein in trans-translation. Because SmpB forms a tight complex with the TLD (21,35), it may facilitate the formation of the acceptor arm in precursor tmRNA and, thus, assist in the formation of a structure that is suitable for trimming by RNases P, T, and PH (16,33). Moreover, because pre-binding of SmpB to stalled ribosomes triggers trans-translation (12), overexpression of truncated proteins is likely to recruit SmpB to ribosomes and might limit its availability for tmRNA maturation as observed (see Fig. 15).
In summary, our data provide the first evidence that pseudoknots pk2-pk4 play an important role in folding and maturation of tmRNA. Moreover, the presence of at least one pseudoknot, optimally pk4, is required for tmRNA tagging. This work also highlights the important role of helix 5, for which phylogenetic support is weak (5). Finally, evidence is presented that protein SmpB helps to process tmRNA precursors. Further studies will investigate how protein SmpB facilitates maturation of tmRNA, why pk4 is especially important for tmRNA maturation, why disruption of pk3 inhibits synthesis of truncated protein, and why pk2 is more important for protein S1 than the other pseudoknots. Powerful assay systems to address these critical questions are now available.