Ribosomal Protein S12 and Aminoglycoside Antibiotics Modulate A-site mRNA Cleavage and Transfer-Messenger RNA Activity in Escherichia coli*

Translational pausing in Escherichia coli can lead to mRNA cleavage within the ribosomal A-site. A-site mRNA cleavage is thought to facilitate transfer-messenger RNA (tmRNA)·SmpB- mediated recycling of stalled ribosome complexes. Here, we demonstrate that the aminoglycosides paromomycin and streptomycin inhibit A-site cleavage of stop codons during inefficient translation termination. Aminoglycosides also induced stop codon read-through, suggesting that these antibiotics alleviate ribosome pausing during termination. Streptomycin did not inhibit A-site cleavage in rpsL mutants, which express streptomycin-resistant variants of ribosomal protein S12. However, rpsL strains exhibited reduced A-site mRNA cleavage compared with rpsL+ cells. Additionally, tmRNA·SmpB-mediated SsrA peptide tagging was significantly reduced in several rpsL strains but could be fully restored in a subset of mutants when treated with streptomycin. The streptomycin-dependent rpsL(P90K) mutant also showed significantly lower levels of A-site cleavage and tmRNA·SmpB activity. Mutations in rpsD (encoding ribosomal protein S4), which suppressed streptomycin dependence, were able to partially restore A-site cleavage to rpsL(P90K) cells but failed to increase tmRNA·SmpB activity. Taken together, these results show that perturbations to A-site structure and function modulate A-site mRNA cleavage and tmRNA·SmpB activity. We propose that tmRNA·SmpB binds to streptomycin-resistant rpsL ribosomes less efficiently, leading to a partial loss of ribosome rescue function in these mutants.

Translational pausing in Escherichia coli elicits a unique RNase activity that cleaves the A-site codon within paused ribosomes (1,2). A-site mRNA cleavage results in a ribosome that is arrested at the 3Ј end of a truncated transcript. In eubacteria, these stalled ribosomes are "rescued" by the tmRNA⅐SmpB 2 quality control system. tmRNA is a specialized RNA that functions as both a transfer RNA and a messenger RNA to remove stalled ribosomes from truncated messages. tmRNA acts first as a tRNA to bind the ribosomal A-site and add its charged Ala residue to the nascent peptide chain (3,4). The truncated message is then released, and the ribosome resumes translation using a small open reading frame within tmRNA to add the SsrA peptide tag to the nascent chain (3,4). SmpB is a tmRNA-binding protein that is required for both the delivery of tmRNA to the ribosome and translation of the SsrA peptide tag (5,6). The tmRNA system performs at least three distinct quality control functions. First, the SsrA peptide tag targets proteins for rapid degradation by ClpXP and other ATP-dependent proteases (7)(8)(9), ensuring that incomplete polypeptides do not accumulate in the cell. Second, tmRNA⅐SmpB mediates the recycling of stalled ribosomes into 50 and 30 S subunits, which are then able to reinitiate protein synthesis on other messages. Finally, tmRNA⅐SmpB facilitates truncated mRNA turnover by delivering RNase R to the released transcripts during ribosome rescue (10). Because tmRNA⅐SmpB is not recruited to ribosomes that are paused on full-length mRNAs (11), A-site cleavage is thought to produce the truncated transcripts required for tmRNA⅐SmpB-mediated ribosome rescue (1,2). In this manner, A-site mRNA cleavage and tmRNA⅐SmpB are proposed to collaborate in the rescue of distressed ribosomes.
A-site mRNA cleavage was first identified as an in vitro activity of the E. coli RelE protein (12). Subsequently, translational pauses were found to induce A-site cleavage in cells that lack RelE and all of its known homologs and paralogs (1,2,13,14). We have recently discovered that RelE-independent A-site cleavage requires at least two distinct RNase activities. RNase II first degrades mRNA to the leading edge of the paused ribosome, allowing for subsequent cleavage in the A-site codon (14). This ribosome border degradation is presumably required for subsequent A-site nuclease activity, because A-site cleavage does not occur in ⌬RNase II cells (14). None of the known E. coli RNases cleave the A-site codon during translational pausing, and the ribosome itself has been proposed to play a catalytic role (1,2,14). The 30 S ribosome subunit binds the A-site codon and is therefore appropriately positioned to catalyze A-site cleavage. However, the A-site is dynamic and could recruit an unknown RNase in the same manner in which it binds translation factors. Regardless of the catalytic scenario, the 30 S A-site defines the mRNA substrate, and the ribosome must, at a minimum, act as a scaffold for the A-site nuclease.
The A-site codon is held within the decoding center of the 30 S subunit, where base-pairing interactions between the codon and the incoming tRNA are monitored (15,16). Three 16 S rRNA residues, G 530 , A 1492 , and A 1493 (Fig. 1), make direct contact with the A-site codon. A 1492 and A 1493 bind in the minor groove of the codon-anticodon helix, ensuring that only Watson-Crick base pairs are allowed at the first two positions (15,16). Although 16 S rRNA is critical for the differentiation of cognate from near cognate tRNA, ribosomal proteins S12, S4, and S5 also play important roles in decoding. The ram (ribosome ambiguity) mutations in rpsD and rpsE (encoding S4 and S5, respectively) are thought to stabilize a closed A-site conformation characterized by high affinity for aminoacyl-tRNA (16 -18). The ram conformation stabilizes the binding of near cognate tRNA and thereby increases the frequency of decoding errors. The aminoglycoside antibiotic streptomycin binds to the 30 S A-site ( Fig. 1) and induces decoding errors, presumably by stabilizing the ram conformation (16,19). Streptomycin resistance is conferred by a variety of mutations in rpsL, which encodes ribosomal protein S12. Streptomycin-resistant S12 variants are predicted to stabilize an open A-site conformation and thereby counteract the closed conformation induced by streptomycin binding (16,18). In general, streptomycin-resistant rpsL alleles also confer an "error-restrictive" phenotype characterized by reduced A-site affinity for near cognate tRNA and hyperaccurate decoding (20). Neamine-containing aminoglycosides, such as paromomycin, also stabilize the closed A-site conformation but have ribosome binding sites that are distinct from that of streptomycin ( Fig. 1) (18,19).
In this report, we ask whether structural perturbations to the ribosomal A-site influence A-site mRNA cleavage and tmRNA⅐SmpB activities. We show that two aminoglycoside antibiotics, streptomycin and paromomycin, inhibit A-site cleavage of stop codons during inefficient translation termination. Aminoglycosides also induced stop codon read-through, suggesting that these antibiotics reduce A-site cleavage by alleviating the underlying translational pause. A-site cleavage and tmRNA⅐SmpB-mediated SsrA peptide tagging activities were significantly reduced in several streptomycin-resistant rpsL mutants. In general, SsrA peptide tagging was reduced in cells containing error-restrictive ribosomes. However, streptomycin was able to fully restore peptide tagging activity in a subset of rpsL strains, without significantly affecting error restriction in these mutants. It appears that tmRNA may be more sensitive than tRNA to structural changes in the A-site, perhaps reflecting the unique manner in which tmRNA⅐SmpB binds the ribosome. Based on these results, we propose that rpsL mutations specifically interfere with the recruitment of tmRNA⅐SmpB to the ribosome.
Plasmids pKW1, pKW11, pKW23, pPW500, pCH201, and pAD8 have been described previously (8,(22)(23)(24). DNA fragments encoding the trc promoter and the N-terminal domain of phage cI repressor (N) were PCR-amplified from plasmid pPW500 using oligonucleotide primers containing restriction sites (underlined bases). The N-flag-His 6 (PP) construct was generated with oligonucleotides lacI-HpaI (5Ј-TAT CCC GCC GTT AAC TAG TAT CAA ACA GGA TTT TCG C) and His 6 (PP)-SacI (5Ј-AAT GAG CTC AAT TAG GGC GGA TGA  TGG TGA TGA TGG TGC). The N-flag-His 6 (LA) fragment was generated with oligonucleotides His 6 (LA)-SacI (5Ј-ACC  GAG CTC AAT TAT GCC AGA TGA TGG TGA TGA TGG) and lacI-HpaI. The resulting PCR fragments were digested with HpaI and SacI and ligated to plasmid pTrc99A (GE Healthcare). The FLAG-N expression constructs were generated by PCR using oligonucleotide N-NdeI (5Ј-CAA TTT CAC ACA GGA AAC AGC ATA TGG GCA CAA AAA AGA AAC C) in con-FIGURE 1. Aminoglycoside binding sites in the ribosomal A-site. The decoding center of the 30 S subunit is depicted with helix 44 (h44) of 16 S rRNA and ribosomal protein S12. 16 S rRNA residues G 530 , A 1492 , and A 1493 are indicated along with the ribosome binding sites of streptomycin (red) and paromomycin (green). Mutations that change S12 residues Lys 42 and Pro 90 are able to confer both streptomycin-resistant and streptomycin-dependent phenotypes. These data were taken from PDB accession number 1FJG (19) and rendered with PyMol. junction with N(PP)-SacI (5Ј-GCG GAG CTC TCG AAT TAG GGC GGA GAC ATG CTA ACC GCT TCA TAC), N(LA)-SacI (5Ј-TTG AGC TCT CGA ATT ATG CCA GAG ACA TGC TAA CC), and N(PPQ)-SacI (5Ј-GCG GAG CTC TCG AAT TGG GGC GGA GAC ATG CTA ACC GCT TCA TAC). The resulting PCR products were digested with NdeI and SacI and ligated to plasmid pFG501. Plasmid pFG501 is a modified version of pTrc99A (GE Healthcare) encoding the FLAG epitope between NcoI and NdeI restriction sites, facilitating the construction of N-terminal FLAG fusion proteins (25).
mRNA Expression and Analysis-E. coli strains were grown overnight at 37°C in LB medium supplemented with the appropriate antibiotics (150 g/ml ampicillin, 25 g/ml tetracycline, or 50 M streptomycin). The following day, cells were resuspended at an A 600 of 0.05 in 15 ml of fresh medium and grown at 37°C with aeration. Once cultures reached A 600 of ϳ0.5, mRNA expression was induced with 2 mM isopropyl ␤-D-thiogalactopyranoside, and simultaneously treated with paromomycin or streptomycin at the indicated concentrations. After further incubation for 15 min, the cultures were poured into 15 ml of ice-cold methanol and collected by centrifugation, and cell pellets were frozen at Ϫ80°C. Total RNA was extracted from cell pellets using 1.0 ml of a solution containing 0.6 M ammonium isothiocyanate, 2 M guanidinium isothiocyanate, 0.1 M sodium acetate (pH 4.0), 5% glycerol, 40% phenol. The disrupted cell suspension was extracted with 0.2 ml of chloro-form, and the aqueous phase was removed and added to an equal volume of isopropyl alcohol to precipitate total RNA. RNA pellets were washed once with ice-cold 75% ethanol and dissolved in 10 mM sodium acetate (pH 5.2), 1 mM EDTA.
Northern blot and S1 nuclease protection analyses of all mRNAs were performed as described (1). Oligonucleotide Trc-RBS (5Ј-CAT GGT CTG TTT CCT GTG TGA AAT TG) was 5Ј-end-labeled with [␥-32 P]ATP and used as a probe for Northern blot hybridizations. Oligonucleotides cI(PP) S1 (5Ј-GCA GGT CGA CTC TAG AGG ATC CCC GGG TAC CGA GCT CGA GTT AGG GCG GAG ACA TGC TAA CCG CTT CAT ACA TCT CGT AG) and cI-His 6 (PP) S1 (5Ј-GGT CGA CTC TAG AGG ATC CCC GGG TAC CGA GCT CAA TTA GGG CGG ATG ATG GTG ATG ATG GTG CTT GTC ATC GTC GTC CTT GT) were 3Ј-end-labeled with [␣-32 P]dideoxy-ATP and used as probes for nuclease S1 protection assays. Northern blots were visualized by phosphorimaging, and A-site mRNA cleavage was quantified using Quantity One software (Bio-Rad). A-site cleavage efficiency was determined as a percentage of total transcripts, defined as full-length transcripts plus A-site-truncated transcripts.
Protein Expression and Western Blot Analysis-Strains were cultured as described above for RNA analysis. Total protein was extracted from frozen cells in 8 M urea, 10 mM Tris-HCl (pH 8.0), 150 mM NaCl, and His 6 -tagged proteins purified by Ni 2ϩnitrilotriacetic acid affinity chromatography as described previously (23). SsrA(His 6 )-tagged protein was further purified by reverse-phase high pressure liquid chromatography as described (26), for electrospray ionization mass spectrometry analysis. Ni 2ϩ -nitrilotriacetic acid-purified proteins were resolved on SDS-polyacrylamide gels, followed by staining with Coomassie Blue. Stained gels were scanned using the LI-COR Odyssey infrared imaging system, and the percentage of SsrA(DD)-tagged chains was quantified as described (13). Reported tagging efficiencies are the means Ϯ S.E. for three independent experiments. Western blot analysis was performed using the LI-COR Odyssey infrared imaging system according to the manufacturer's instructions with minor modifications. Briefly, 10 g of total urea-soluble protein was resolved by SDS-PAGE, followed by electrotransfer to nitrocellulose membranes. Membranes were blocked with 2% (w/v) bovine serum albumin in phosphate-buffered saline (2.7 mM KCl, 1.8 mM KH 2 PO 4 , 137 mM NaCl, 10.1 mM Na 2 HPO 4 (pH 7.4)), followed by incubation overnight with anti-SsrA(DD) polyclonal and anti-FLAG M2 monoclonal antibodies (Sigma). IRDye TM 800-conjugated anti-mouse (Rockland Immunochemicals) and Alexafluor 680-conjugated anti-rabbit (Invitrogen) secondary antibodies were used for fluorescence detection.
Dual-Luciferase Assay-E. coli strains carrying plasmid pAD8 (24) were grown overnight in LB medium supplemented with 150 g/ml ampicillin. The following day, cells were diluted 1:200 into fresh LB containing 150 g/ml ampicillin and 50 M streptomycin as indicated. Cultures were grown at 37°C to midlogarithmic phase, collected by centrifugation, and resuspended in 200 l of lysis buffer (1 mg/ml lysozyme, 10 mM Tris-HCl (pH 8.0), 1 mM EDTA). Cell suspensions were incubated on ice for 10 min and then frozen in a dry ice/ethanol bath. Samples were then thawed on ice, and 5 l of each lysate was assayed for firefly and Renilla luciferase activities using the Dual-Luciferase reporter assay system (Promega). For each reaction, luminescence (expressed as counts/s) was collected over a 10-s interval using a model 1420 Victor 3 V plate reader with injectors (PerkinElmer Life Sciences). Each lysate was assayed in triplicate, and all rpsL mutants were independently tested at least three times.

Aminoglycoside Antibiotics Inhibit A-site mRNA Cleavage-
We examined the effect of streptomycin on A-site mRNA cleavage using a plasmid-borne construct that expresses the N-terminal domain of phage cI repressor containing a C-terminal Pro-Pro sequence ( Fig. 2A). The C-terminal Pro-Pro motif interferes with normal translation termination and is sufficient to elicit A-site cleavage at stop codons (1,27). This construct also encodes an N-terminal FLAG epitope to facilitate tracking of the reporter protein by Western blot (Fig. 2A). Expression of FLAG-N(PP) resulted in the accumulation of truncated mRNA in cells lacking tmRNA (⌬tmRNA) but not in tmRNA ϩ cells (Fig. 2B). A-site-cleaved transcripts do not typically accumulate in tmRNA ϩ cells, presumably because tmRNA⅐SmpB releases the truncated messages from stalled ribosomes, thereby facilitating their rapid degradation (1,10,28). S1 nuclease protection analysis confirmed that these transcripts were truncated in the stop codon (data not shown), consistent with A-site mRNA cleavage during inefficient translation termination (1,2). Truncated mRNA was not detected in cells expressing N with a C-terminal Leu-Ala sequence (Figs. 1A) (data not shown), in agreement with previous studies showing that this C-terminal peptide sequence does not induce ribosome pausing during translation termination (14,27). Treatment of ⌬tmRNA cells with streptomycin led to a dosedependent decrease in A-site truncated mRNA without affecting the levels of full-length transcript (Fig. 2B). These results suggest that streptomycin inhibits A-site mRNA cleavage during inefficient translation termination.
Translation initiation and ribosome recycling after termination are both inhibited at high aminoglycoside concentrations (29,30). Ribosomes must translate to the stop codon in order to elicit A-site cleavage in our system; therefore, it is possible that streptomycin reduced A-site cleavage indirectly by shutting down protein synthesis. To test whether streptomycin inhibited protein synthesis under these conditions, we measured the accumulation of FLAG-N(PP) as a function of time in the absence and presence of streptomycin (Fig. 2C). FLAG-N(PP) synthesis was not inhibited by 7.5 M streptomycin during the first 20 min of treatment and continued over the entire time course (Fig. 2C). Streptomycin significantly inhibited protein synthesis after 40 min of treatment (data not shown). However, we note that A-site cleavage was assessed after 15 min of streptomycin treatment, a point at which there was no decrease in reporter protein synthesis. Although protein synthesis was not inhibited, a prominent FLAG-reactive alternative translation product was observed in cells treated with streptomycin ( Fig.  2C). Streptomycin increases the frequency of miscoding events, including stop codon read-through (31,32), which could account for this product. To ascertain whether streptomycin induced read-through, we mutated the flag-N(PP) stop codon to a Gln codon to mimic the read-though product and found that the resulting protein co-migrated with the streptomycin-induced product on SDS-polyacrylamide gels (Fig. 1, A  and C). Moreover, the alternative product was not detected in streptomycin-treated cells expressing the control FLAG-N(LA) protein (Fig. 2C). Taken together, these results suggest that streptomycin-induced miscoding occurs specifically when ribosomes pause during translation termination.
To determine whether neamine-containing aminoglycosides also influence A-site mRNA cleavage, we examined the effects of paromomycin using an additional N reporter construct. The N-flag-His 6 (PP) construct is derived from the previously characterized pPW500 plasmid (8), which encodes internal FLAG and His 6 peptide epitopes that allow for immunodetection and affinity purification (Fig. 3A). Like the flag-N(PP) construct described above, the N-flag-His 6 (PP) message also undergoes A-site cleavage in ⌬tmRNA cells (Fig. 3B) (data not shown). A-site mRNA cleavage was significantly inhibited in ⌬tmRNA cells treated with increasing concentrations of paromomycin (Fig. 3B). Western blot analysis showed that N-FLAG-His 6 (PP) synthesis was not dramatically reduced in ⌬tmRNA cells treated with up to 30 M paromomycin (Fig. 3C). Moreover, time course analysis showed that N-FLAG-His 6 (PP) accumulated at similar rates in either the absence or presence of 20 M paromomycin (Fig. 3C). Although N-FLAG-His 6 (PP) synthesis was not inhibited, paromomycin induced significant stop codon read-through (Fig. 3C). Similar to the findings with streptomycin, paromomycin-induced read-through was not observed during expression of the control N-FLAG-His 6 (LA) protein, which undergoes efficient translation termination (Fig. 2,  A and C). These results show that aminoglycosides inhibit A-site mRNA cleavage at concentrations that still permit efficient protein synthesis in ⌬tmRNA cells. The accompanying increase in stop codon read-through suggests that aminoglycosides may inhibit A-site cleavage by relieving the underlying translational pause during termination.
Aminoglycosides and SsrA Peptide Tagging Activity-A-site cleavage is thought to be required for tmRNA⅐SmpB recruitment to ribosomes during inefficient translation termination (1,2). We reasoned that if aminoglycosides alleviate ribosome pausing via stop codon read-though, then they should also reduce the attendant SsrA peptide tagging activity. We first confirmed peptide tagging at the C terminus of FLAG-N(PP) using tmRNA(His 6 ), which encodes the protease-resistant SsrA(His 6 ) tag (23,33). FLAG-N(PP) was expressed in tmRNA(His 6 ) cells, and SsrA(His 6 )-tagged proteins were purified for mass spectrometry analysis. The major tagged product had a mass of 13,365 Da, corresponding to SsrA(His 6 ) tag addition after the C-terminal proline residue of FLAG-N(PP) (Fig.  4A). We also used mass spectrometry to confirm C-terminal tagging of the N-FLAG-His 6 (PP) reporter protein with the stable SsrA(DD) tag (data not shown). These results confirm previous work showing that A-site cleavage at stop codons in ⌬tmRNA cells is correlated with SsrA peptide tagging of fulllength proteins in tmRNA ϩ cells (1).
We next tested whether aminoglycosides inhibited SsrA(DD) peptide tagging at the C terminus of the reporter proteins. We note that tmRNA ϩ cells are more resistant to streptomycin than ⌬tmRNA cells (34), and therefore higher streptomycin concentrations were used in these tagging experiments (compared with the data shown in Fig. 2B). Western blot analysis showed that streptomycin and paromomycin both inhibited SsrA(DD) peptide tagging and concomitantly increased stop codon read-through (Fig. 4B). However, total N reporter protein synthesis also appeared to be inhibited by the aminoglycosides, particularly in paromomycin-treated cells (Fig. 4B). The inhibition of N expression in these experiments was perplexing, because both aminoglycosides had little effect on protein synthesis in ⌬tmRNA cells (see Figs. 1C and 2C). To determine whether SsrA(DD) peptide tagging was inhibited to a greater extent than N synthesis, we quantified and compared the relative decreases in SsrA(DD)-and FLAG-dependent fluorescence from Western blots. We found that streptomycin inhibited SsrA(DD)-tagging to a greater extent than it inhibited FLAG-N(PP) synthesis (Fig. 4B) (data not shown). However, SsrA(DD) peptide tagging and N-FLAG-His 6 (PP) synthesis were inhibited to the same extent by paromomycin treatment (Fig.  4B) (data not shown). Therefore, although these results suggest that streptomycin inhibits SsrA tagging via stop codon readthrough, tmRNA⅐SmpB activity was also inhibited due to a general decrease in protein synthesis in paromomycin-treated cells.
In principle, it is possible that aminoglycosides directly inhibit tmRNA⅐SmpB activity. Indeed, paromomycin has been Ribosomal Protein S12 and SsrA Peptide Tagging NOVEMBER 13, 2009 • VOLUME 284 • NUMBER 46 reported to interfere with distinct stages of the tmRNA⅐SmpB activity cycle (35,36). To determine whether aminoglycosides inhibit SsrA peptide tagging independent of stop codon read-through, we examined tmRNA⅐SmpB activity with ribosomes stalled on nonstop mRNA. Nonstop messages lack inframe stop codons. Therefore, ribosomes translate to the 3Ј-end of these "truncated" transcripts, where they stall with no codon in the A-site (Fig. 5A). In tmRNA ϩ cells, these stalled ribosomes are rapidly rescued by tmRNA⅐SmpB, and a large proportion of the protein chains are SsrA-tagged (3,22). In ⌬tmRNA cells, the untagged nascent chains are released from the stalled ribosome by an uncharacterized process (Fig. 5A). Thus, tmRNA⅐SmpB activity can be assessed by measuring the ratio of tagged to untagged protein synthesized from nonstop mRNA (Fig.  5A). Expression of N-FLAG-His 6 from nonstop mRNA in ⌬tmRNA cells resulted in the accumulation of untagged protein, whereas almost all of the N-FLAG-His 6 chains were tagged in tmRNA(DD) cells (Fig. 5B). When tmRNA(DD) cells were treated with increasing concentrations of streptomycin or paromomycin, we again saw that reporter protein synthesis was somewhat inhibited (Fig. 5B). However, aminoglycoside treatment did not increase the proportion of untagged N-FLAG-His 6 chains (Fig. 5B). These results indicate that under these experimental conditions, aminoglycosides do not specifically inhibit the tmRNA⅐SmpB system.
A-site mRNA Cleavage in Streptomycin-resistant rpsL Mutants-In addition to their well described effects on the ribosome, aminoglycosides also bind and modulate the activity of other catalytic RNAs (37)(38)(39). To determine whether aminoglycosides influence A-site cleavage due to their effects on translation, we examined streptomycin-resistant rpsL mutants. The ribosomes from rpsL mutants are altered in ribosomal protein S12 and typically have much lower affinity for streptomycin (40 -42). To generate a variety of rpsL mutations, we used phage Red-mediated recombination and oligonucleotide libraries to randomize codons corresponding to S12 residues Lys 42 and Pro 90 (see Fig. 1) and then selected the mutagenized cells for streptomycin-resistant mutants. We focused on Lys 42 and Pro 90 because changes in these residues are most commonly associated with streptomycin resistance (20,43). We sequenced the rpsL gene from 100 streptomycinresistant mutants and identified 16 different missense mutations: nine Lys 42 alleles and seven Pro 90 alleles ( Table 2). Because oligonucleotide recombineering allows the routine isolation of unusual missense mutations (21), many of these mutations have not been previously identified in E. coli. Examination of rpsL mutant growth rates in the presence and absence of streptomycin led to the identification of two streptomycin-dependent mutants, rpsL(P90K) and rpsL(P90R), and one streptomycin pseudodependent mutant, rpsL(P90Q), which grew almost 2-fold faster in the presence of streptomycin (Table 2).
Initially, we examined A-site mRNA cleavage in three previously characterized mutants: rpsL(K42T), rpsL(K42R), and rpsL(K42A). The rpsL(K42T) mutation is a classical streptomycin resistance allele that produces error-restrictive ribosomes. In contrast, rpsL(K42R) is unique among known streptomycin resistance alleles in that its phenotype is non-restrictive (20). Green and co-workers (44) recently described the rpsL(K42A) mutation, reporting that it conferred a streptomycinpseudodependent phenotype. However, we observed no change in the growth rate of rpsL(K42A) cells when cultured in medium containing streptomycin (Table 2). A-site mRNA cleavage was reduced in all three of these mutants compared with rpsL ϩ cells (Fig. 6). Growth in the presence of 50 M streptomycin had little effect on A-site mRNA cleavage in rpsL(K42A) and rpsL(K42T) cells but decreased cleavage effi- The flag-N(PP) transcript was expressed in ⌬tmRNA cells containing rpsL mutations that encode the indicated streptomycin-resistant S12 variants. Total RNA was isolated from cells grown in the absence and presence of streptomycin (50 M) as indicated. Samples from tmRNA ϩ and ⌬tmRNA cells containing wild-type S12 (rpsL ϩ ) were included in each blot as a reference control for A-site cleavage. The rpsL(P90K) and rpsL(P90R) mutants are streptomycin-dependent and therefore were not tested in the absence of streptomycin. Similarly, the rpsD mutations confer streptomycin sensitivity to the parental rpsL(P90K) strain, and therefore these mutants were not tested with streptomycin. The positions of full-length and A-site-truncated transcripts are indicated. The percentage of A-site-truncated transcripts, with respect to total transcript (full-length ϩ truncated), was determined from phosphorimaging data as described under "Experimental Procedures." ciency in the rpsL(K42R) mutant (Fig. 6). We also examined the novel rpsL(P90F) mutant, which was the most frequently identified mutation in the selection (Table 2). A-site mRNA cleavage was reduced in rpsL(P90F) cells but could be restored when the mutant was grown with streptomycin ( Fig. 6). Taken together, these results strongly suggest that streptomycin modulates A-site cleavage by virtue of its effects on the ribosome and translation. We also examined the two streptomycin-dependent rpsL(P90R) and rpsL(P90K) alleles. Because streptomycin is presumably bound to streptomycin-dependent ribosomes, we used these mutants to examine whether ribosome-bound aminoglycoside influences A-site mRNA cleavage. Both streptomycin-dependent mutants exhibited lower A-site cleavage levels compared with rpsL ϩ cells (Fig. 6). We next asked whether mutations that suppress streptomycin dependence could restore A-site cleavage to rpsL(P90K) cells. These suppressor mutations typically occur in the rpsD and rpsE genes, which encode ribosomal proteins S4 and S5, respectively (20). We isolated and identified four rpsD mutations that allowed rpsL(P90K) cells to grow in the absence of streptomycin. The rpsD-1, rpsD-2, and rpsD-3 alleles encode Q53K, N85Y, and Y203amber missense mutations, respectively ( Table 1). The rpsD-4 mutation is a single-nucleotide deletion in codon 193, resulting in a frameshift that replaces the C-terminal Asp 193 -Lys 205 residues of S4 with the TLTNT pentapeptide ( Table 1). Each of the rpsD mutants was streptomycin-sensitive but still contained the original rpsL(P90K) mutation. In addition to con-ferring streptomycin sensitivity, the rpsD mutations also appeared to restore some A-site cleavage activity compared with the parental rpsL(P90K) strain (Fig. 6). These results demonstrate that structural perturbations to the A-site can significantly modulate A-site cleavage activity.
SsrA Peptide Tagging Activity in Streptomycin-resistant rpsL Mutants-Given the effects of rpsL mutations on A-site cleavage, we next asked whether these alleles also influence tmRNA⅐SmpB activity. We expressed N-FLAG-His 6 (PP) in rpsL mutants containing tmRNA(DD) and then purified the reporter protein by Ni 2ϩ affinity chromatography to determine the percentage of SsrA(DD)tagged chains. SsrA(DD) tagging was largely unaffected in most of the rpsL(K42) mutants, although rpsL(K42V) cells showed significantly less tagging (28 Ϯ 1.2% tagged) than rpsL ϩ cells (43 Ϯ 0.6% tagged) (Fig. 7). In contrast, most of the rpsL(P90) mutants displayed significant SsrA(DD) peptide tagging defects. The greatest effect was observed in rpsL(P90Q) cells, in which only 19 Ϯ 1.1% of the N-FLAG-His 6 (PP) chains were tagged (Fig. 7). Notably, SsrA(DD) tagging was increased when the rpsL(P90N), rpsL(P90Y), rpsL(P90F), and rpsL(P90Q) mutants were grown in 50 M streptomycin (Fig. 7). Surprisingly, there was little correlation between SsrA tagging efficiency and A-site mRNA cleavage in the rpsL mutants. For example, the rpsL(K42A) mutation had no effect on peptide tagging activity in streptomycin-treated tmRNA(DD) cells (Fig.  7) yet reduced A-site cleavage in the ⌬tmRNA background (Fig.  6). A similar disparity between A-site cleavage and peptide tagging was observed with the rpsL(P90R) streptomycin-dependent mutation.
SsrA tagging of full-length N-FLAG-His 6 (PP) chains results from inefficient translation termination. Therefore, tagging is inversely related to the efficiency of stop codon decoding by release factors (23,27). To assess tmRNA⅐SmpB activity without competition from release factors, we examined SsrA(DD) tagging of N-FLAG-His 6 expressed from nonstop mRNA. As outlined above, decreased tmRNA⅐SmpB activity leads to the accumulation of untagged products, which are readily distinguished from tagged protein by gel electrophoresis. We focused on the rpsL(P90) mutants because these cells exhibited the most pronounced decrease in SsrA(DD) tagging during inefficient termination (Fig. 7). In addition, the previously characterized rpsL(K42R), rpsL(K42T), and rpsL(K42A) mutants were also analyzed. All of the examined strains had significant tagging defects, except for the rpsL(K42R) and rpsL(P90R) mutants, in which SsrA(DD) tagging was actually more efficient than in rpsL ϩ cells (Fig. 8). Streptomycin treatment increased tmRNA⅐SmpB activity in all of the non-dependent rpsL(P90) mutants. In fact, the rpsL(P90H), rpsL(P90F), and rpsL(P90Y) mutants regained wild-type tagging efficiency or better when treated with streptomycin (Fig. 8). In contrast, the rpsL(K42A) and rpsL(K42T) mutants showed no change in peptide tagging in response to streptomycin (Fig. 8). We also examined the effects of the four rpsD mutations and found that the rpsD-1, -2, and -4 alleles decreased SsrA(DD) tagging efficiency compared with the parental rpsL(P90K) strain (Fig. 8).
The results shown in Fig. 8 suggest that SsrA tagging in rpsL mutants may be related to the error-restrictive phenotype. For example, the restrictive rpsL(K42T) mutation decreased SsrA tagging, whereas the non-restrictive rpsL(K42R) mutation allowed efficient tagging (Fig. 8). Additionally, streptomycin counteracts the restrictive phenotype (20), which could conceivably lead to more efficient tagging in the streptomycintreated rpsL(P90) mutants. Therefore, we assessed miscoding in several of the rpsL mutants to determine whether a correlation exists between error restriction and reduced tmRNA⅐ SmpB activity. Stop codon read-through was measured using a Renilla luciferase-firefly luciferase fusion construct, in which the firefly luc gene contains an in-frame UGA stop codon at position 417 (24). This construct only produces firefly luciferase activity when the UGA codon is inappropriately decoded as a sense codon. In the absence of streptomycin, most of the rpsL mutants had lower firefly luciferase/Renilla luciferase activity ratios than rpsL ϩ cells, indicative of reduced read-through and the error-restrictive phenotype (Fig. 9). As expected, the rpsL(K42R) mutant was non-restrictive but surprisingly became more restrictive when grown in medium supplemented with streptomycin (Fig. 9). Streptomycin slightly increased read-through in the rpsL(K42T) and rpsL(P90Q) mutants, but firefly luciferase/Renilla luciferase ratios were unaffected by streptomycin in the other mutants (Fig. 9). Streptomycin-dependent rpsL(P90R) cells were moderately restrictive, whereas the rpsL(P90K) mutant was more restrictive, similar to the other rpsL mutants we examined (Fig. 9). The rpsD mutations had little effect on miscoding in the rpsL(P90K) background, although we note that it was necessary to test these strains under different conditions due to their respective streptomycin-dependent and streptomycin-sensitive phenotypes (Fig. 9). Taken together, we found no direct correlation between error restriction and efficiency of SsrA peptide tagging (compare Figs. 7 and 8). These results indicate that some other property of rpsL ribosomes influences tmRNA⅐SmpB activity.

DISCUSSION
The results presented here show that alterations in ribosomal A-site structure and function have significant effects on A-site mRNA cleavage and tmRNA⅐SmpB activity. Streptomycin and paromomycin both inhibited A-site cleavage during inefficient translation termination. A-site cleavage during translational pauses is a complex process that requires at least two RNase activities. RNase II processively degrades downstream mRNA in a 3Ј 3 5Ј direction until it encounters the leading edge of the paused ribosome, and this activity appears to be required for subsequent cleavage in the A-site codon by another RNase (14). The A-site nuclease has yet to be identified and could be an activity of the ribosome itself. Because streptomycin and paromomycin bind very near the A-site codon and dramatically alter A-site structure and function (15,16,19,20,45,46), these aminoglycosides could directly inhibit A-site nuclease activity. However, aminoglycosides induced stop codon read-through, which occurred concomitantly with the decrease in A-site mRNA cleavage. Although correlative, this finding suggests that aminoglycoside-induced miscoding alleviates the translational pause required for A-site cleavage. Other indirect evidence also suggests that aminoglycosides influence ribosome pausing in this system (31,32,47). When A-site cleavage is inhibited, downstream mRNA is still degraded to the ribosome border, resulting in a slightly larger truncated transcript that indicates the "toeprint" of the paused ribosome (14,25). The lack of A-site and ribosome border-truncated transcripts in aminoglycosidetreated cells suggests that ribosomes no longer pause during translation termination. We note that paromomycin has the potential to actually increase A-site cleavage during translation termination, because it inhibits release factor binding to A-site stop codons (48). However, paromomycin also stabilizes near cognate tRNA in the A-site (19), providing an outlet for arrested ribosomes via stop codon read-through. Taken together, it appears that aminoglycoside-induced miscoding accounts for the observed decrease in A-site mRNA cleavage.
Aminoglycosides also inhibited the SsrA peptide tagging during inefficient translation termination. In our system, this inhibitory effect appears to be the result of increased stop codon read-through and decreased protein synthesis. Aiba and co-workers (49) have previously shown that aminoglycoside induced stop codon read-through can result in translation through the 3Ј-untranslated region to the end of the message and thereby increase SsrA peptide tagging. Here, we see that the same miscoding event can actually prevent peptide tagging associated with inefficient translation termination. We note that the constructs used in our study have additional in-frame stop codons in the 3Ј-untranslated region, which prevent ribosomes from reaching the 3Ј-end of the transcript. Paromomycin has also been reported to inhibit the aminoacylation of tmRNA and to shift the tmRNA reading frame into the Ϫ1 frame in vitro (35,36). In those studies, tmRNA aminoacylation was inhibited at ϳ225 M paromomycin, and the Ϫ1 frameshift effect became apparent at 55 M paromomycin. We do not know the intracellular concentration of aminoglycosides in our experiments, but there was no evidence of Ϫ1 frame translation by Western blot analysis or mass spectrometry. 3 Moreover, there was no indication that tmRNA aminoacylation was specifically inhibited by either paromomycin or streptomycin. A lack of tmRNA aminoacylation would presumably be manifested as a ⌬tmRNA phenotype, yet we observed no increase in untagged chains during aminoglycoside treatment. Therefore, although aminoglycosides can clearly inhibit distinct stages of the tmRNA activity cycle, these effects appear to require higher concentrations than those used in our study. It is still unclear why the synthesis of our reporter proteins was more sensitive to aminoglycoside treatment in tmRNA(DD) cells compared with ⌬tmRNA cells, particularly because ⌬tmRNA cells have been shown to be more sensitive to aminoglycoside antibiotics (34). However, we note that protein synthesis was also inhibited in ⌬tmRNA cells after about 40 min of aminoglycoside treatment.
Thus, it appears that the inhibition of protein synthesis is merely delayed in ⌬tmRNA cells, perhaps reflecting differences in the rate of aminoglycoside uptake between the two genetic backgrounds.
A-site mRNA cleavage was significantly decreased in several streptomycin-resistant rpsL mutants, yet there was no correlation between error restriction and A-site cleavage in these strains. The non-restrictive rpsL(K42R) mutation and the weakly restrictive rpsL(P90R) mutation both reduced A-site cleavage to the same extent as restrictive rpsL alleles. Streptomycin-resistant ribosomes tend to suppress stop codon readthrough, suggesting that translation termination may occur more efficiently in these mutants. Of course, increased fidelity during termination does not necessarily indicate more rapid stop codon decoding. Indeed, SsrA(DD) tagging of full-length FLAG-N(PP) in the rpsL(K42A) and rpsL(P90R) mutants was essentially identical to that of rpsL ϩ cells, suggesting comparable translational pausing during termination in these backgrounds. It seems more likely that reduced read-through is due to low A-site affinity for suppressor tRNA rather than an 3 L. E. Holberger and C. S. Hayes, unpublished results. increase in the rate of termination (50 -52). Alternatively, reduced A-site cleavage may result from processivity errors, which are characteristic of both streptomycin-resistant and -dependent ribosomes (45,46,53). Processivity errors occur when translating ribosomes fail to reach the stop codon and instead produce truncated peptide chains. Most of these errors have been proposed to be the result of "drop-off" (20), in which peptidyl-tRNA dissociates from the ribosome. Although the restrictive P-site has high affinity for peptidyl-tRNA, the restrictive A-site has a corresponding low affinity for peptidyl-tRNA, and drop-off is hypothesized to occur from the A/P hybrid state prior to translocation (17,46). We did not detect incomplete FLAG-N(PP) chains in rpsL mutants, but perhaps drop-off occurred while ribosomes paused during termination. Non-canonical release of full-length nascent chains in rpsL mutants would be indistinguishable from normal termination and therefore very difficult to study in vivo. We are currently examining the kinetics of ribosome pausing in various rpsL mutants to determine whether pretermination ribosomes are recycled more rapidly in these cells.
Finally, our data show that ribosomal protein S12 plays an important role in tmRNA⅐SmpB-mediated ribosome rescue. This effect was most apparent during ribosome arrest on nonstop mRNA, in which a significant proportion of chains were not tagged in tmRNA(DD) cells. One possible explanation for these findings is that rpsL ribosomes do not support efficient tmRNA⅐SmpB activity. In principle, a defect in any stage of the tmRNA activity cycle could produce the partial loss of function observed in these mutants. But given the well characterized role of S12 in tRNA selection, a defect in the initial binding of tmRNA to the A-site seems most likely. A-site binding of tmRNA differs from that of canonical tRNAs in several respects. tmRNA lacks an anticodon and readily binds ribosomes that contain no A-site codon. Additionally, SmpB appears to mimic the missing A-site codon-anticodon helix and has recently been shown to make contacts with 16 S rRNA residues G 530 , A 1492 , and A 1493 in the decoding center (54 -56). Perhaps these atypical interactions render tmRNA⅐SmpB more sensitive to perturbations in the A-site. Although rpsL ribosomes tend to have lower A-site affinity for canonical tRNA, this phenomenon cannot completely account for our data. SsrA peptide tagging was significantly reduced in several error-restrictive rpsL(P90) mutants. However, streptomycin was able to fully restore tagging in several of these mutants while having no effect on error restriction. These results suggest that the rpsL mutations have specific effects on tmRNA⅐SmpB recruitment, distinct from the effects on tRNA binding. Although we favor a model in which rpsL mutations interfere with tmRNA⅐SmpB binding to the A-site, we recognize that there is a tmRNA-independent pathway that releases nascent chains from nonstop-arrested ribosomes. It is possible that rpsL mutations accelerate this tmRNA-independent ribosome-recycling pathway, thereby giving the appearance of defective tmRNA⅐SmpB function. These two models can be distinguished by measuring the rates of peptidyl-tRNA turnover from nonstop mRNA-arrested ribosomes in tmRNA ϩ and ⌬tmRNA cells. Peptidyl-tRNA accumulates on nonstopstalled ribosomes in ⌬tmRNA cells but not in tmRNA ϩ cells, presumably due to rapid ribosome rescue. 4 If rpsL mutations interfere with tmRNA⅐SmpB recruitment to paused ribosomes, then peptidyl-tRNA should be detectable in these mutants. Alternatively, if the rpsL mutations facilitate tmRNA-independent ribosome recycling, then peptidyl-tRNA should turn over more rapidly in ⌬tmRNA cells carrying these mutations. We are currently measuring the kinetics of peptidyl-tRNA turnover in rpsL mutants to uncover the basis of this ribosome rescue phenotype.