Inhibition of Transfer Messenger RNA Aminoacylation and trans -Translation by Aminoglycoside Antibiotics* □ S

Transfer messenger RNA (tmRNA) directs the modification of proteins of which the biosynthesis has stalled or has been interrupted. Here, we report that aminoglycosides can interfere with this quality control system in bacteria, termed trans -translation. Neomycin B is the strongest inhibitor of tmRNA aminoacylation with alanine ( K i value of (cid:1) 35 (cid:1) M ), an essential step during trans - translation. The binding sites of neomycin B do not over-lap with the identity determinants for alanylation, but the aminoglycoside perturbs the conformation of the acceptor stem that contains the aminoacylation signals. Aminoglycosides reduce the conformational freedom of the transfer RNA-like domain of tmRNA. Additional con-tacts between aminoglycosides and tmRNA are within the tag reading frame, probably also disturbing repro-gramming of the stalled ribosomes prior protein tagging. Aminoglycosides impair tmRNA aminoacylation in the presence of all of the transfer RNAs from Escherichia coli , small protein B, and elongation factor Tu, but when both proteins are present, the inhibition constant is 1 order of magnitude higher. SmpB and elongation factor Tu have RNA chaperone activities, ensuring that tmRNA adopts an optimal conformation during aminoacylation. In bacteria, transfer (tmRNA) and 50–330 n M purified E. coli AlaRS. When varying the pH from 5.5 to 9.5, 50 m M MES was used for pH 5.5, and 50 m M Tris-HCl was used for pH levels from 6.5 to 9.5. Usually, 1 (cid:3) M of tmRNA or tmRNA-TLD was denatured for 3 min at 75 °C followed by 10 min at room temperature. Then 1 (cid:3) M of purified E. coli His 6 -tagged SmpB and/or His 6 -tagged EF-Tu (cid:2) GDP and/or EF-Tu (cid:2) GTP were added and incubated for 15 min at room temperature. All of the proteins were at least 98% pure as judged by SDS-PAGE analysis. EF-Tu (cid:2) GDP was activated to EF-Tu (cid:2) GTP immediately before use by incubation at 37 °C for 15 min in 50 m M Tris-HCl (pH 7.6), 7 m M MgCl 2 , 60 m M NH 4 Cl complemented with 100 (cid:3) M GTP, 6 m M phosphoenolpyruvate, and 10 (cid:3) g/ml of pyruvate kinase. After the incubation, the mixture was kept on ice before use. Various concentrations of aminoglycosides were added, followed by the aminoacylation buffer, the labeled alanine, and the AlaRS. The aliquots were spotted onto 3MM Whatman papers at different times, and precipitated with 5% trichloroacetic acid. Kinetic parameters ( K m and V max ) in the presence and absence of aminoglycosides were performed under steady-state conditions of enzyme (50–330 n M AlaRS) and substrate concentrations of tmRNA (0.5–3.4 (cid:3) M ) or tmRNA-TLD (0.5–6 (cid:3) M ) and determined from Lineweaver-Burk plots. These experimental conditions were also applied for tmRNA aminoacylation in the presence of all tRNAs from E. coli . Ultraviolet Absorbance Melting Curves— The progressive melting of tmRNA and tmRNA-TLD was monitored by following their UV absor- bency at 258 nm as a function of temperature on a UVIKON XL (Bio-tek Instruments) equipped with a temperature regulator and with a six-cell holder. The temperature was increased gradually at 0.5 °C/min from 15 to 93 °C. The measurements were done in 20 m M potassium phosphate (pH 5.8), 0.5 m M EDTA, and 5 m M MgCl 2 for tmRNA-TLD and in 20 m M potassium phosphate (pH 5.8), 50 m M NaCl, and 5 m M MgCl 2 for tmRNA in the absence and presence of 500 (cid:3) M neomycin B, 10 m M tobramycin, and 2 m M paromomycin. Chemical Footprints— Labeling at the 5 (cid:2) ends of tmRNA and tmRNA-TLD were performed with [ (cid:1) - 32 P]ATP and phage T4 polynucle- otide kinase on RNA dephosphorylated previously with alkaline phosphatase. Labeling at their 3 (cid:2) ends was carried out by ligation of [ (cid:1) - 32 P]pCp using T4 RNA ligase. After labeling, tmRNA was gel-puri-fied (5% PAGE), eluted, and ethanol-precipitated. Labeled tmRNA was heated 2 min at 80 °C and slowly cooled down for 20 min at room temperature. The reaction mixtures (15 (cid:3) l) contained 200,000 cpm of 32 P-labeled tmRNA or tmRNA-TLD and increasing amounts of aminoglycosides (50–500 (cid:3) M of neomycin B, 0–1000 (cid:3) M of paromomycin, and 0–6 m M of tobramycin) in 50 m M Hepes (pH 7.5), 25 m M KCl, and 2 m M MgCl 2. After an incubation of 10 min at room temperature, ENU, lead-acetate, or iron-EDTA were added: 6.25 (cid:3) l of a solution of ENU

In bacteria, transfer messenger RNA (tmRNA) 1 known alternatively as SsrA RNA or 10Sa RNA, rescues stalled ribosomes and contribute to the degradation of incompletely synthesized peptides (for reviews, see Refs. [1][2][3]. In a process termed trans-translation, tmRNA acts first as a tRNA, being aminoacylated at its 3Ј end with alanine by alanyl-tRNA synthetase (AlaRS) (4,5) and adding an alanine to the stalled polypeptide chain. Resumption of translation ensues not on the mRNA on which the ribosomes were stalled but at an internal position in tmRNA. Translation termination occurs and permits ribosome recycling. trans-Translation plays at least two physiological roles in bacteria: removing ribosomes stalled upon mRNAs and tagging the resulting truncated proteins for degradation.
Because tmRNA is unique to prokaryotes and is required for viability of some bacteria, it has attracted the attention of those interested in novel targets for antibiotic therapy. tmRNA has to be aminoacylated before directing the addition of a peptide tag to the problematic protein. Moreover, in Neisseria gonorrhoea, and probably also in other bacteria, tmRNA aminoacylation is essential for resolving stalled translation complexes and preventing depletion of free ribosomes (6), whereas tagging for proteolysis is dispensable. tmRNA aminoacylation is therefore an attractive target for blocking trans-translation in the pathogenic bacteria responsible for infectious diseases.
The interactions of aminoglycosides with RNAs represent a paradigm in the use of small molecules as effectors of RNA function. Aminoglycosides bind and modulate the function of a variety of therapeutically useful RNA targets (for a review, see Ref. 7) and show antimicrobial as well as antiviral activities. Aminoglycoside antibiotics disrupt the bacterial membrane and induce miscoding during prokaryotic protein synthesis by binding to the ribosomal A site (8). They also interfere with translational control (9) and with human immunodeficiency virus replication (10) and inhibit the activity of several catalytic RNAs by displacing essential metal ions, such as selfsplicing group I introns (11,12), hammerhead (13), human hepatitis delta virus (14), and hairpin ribozymes (15), as well as the tRNA processing activity of RNase P RNA (16).
Recent structural (17) and functional (18) evidence indicates that aminoglycosides can bind and inhibit the aminoacylation of two canonical tRNAs: Escherichia coli tRNA Phe and yeast tRNA Asp . The crystal structure of yeast tRNA Phe in complex with neomycin B reveals that the aminoglycoside binds in the deep groove below the D-loop. Inhibition of aminoacylation of tRNAs is proposed to be either because the aminoglycoside interferes with the interaction between the aminoacyl-tRNA synthetase and its cognate tRNA through its binding to major recognition elements (the phenylalanine system) (17) or via a conformational change of the RNA (the acid aspartic system) (18). For the alanine system, aminoacylation is determined by a single G 3 ⅐U 70 pair (19) and also by minor identity elements including the discriminator base A 73 and a G 2 ⅐C 71 pair. This limited set of nucleotides (see Fig. 1, A and B, circled nucleotides) is conserved in all of the 293 tmRNA genes that have been sequenced (20). tmRNA is a partial structural mimic of canonical tRNAs thanks to the presence of an acceptor stem, a T-stem loop, and a D-analog (21). tmRNA is, however, ϳ5-fold bigger compared with canonical tRNAs, contains several pseudoknots and an internal open reading frame, probably lacks most of the key tertiary interactions present in tRNAs, and does not contain an anticodon stem-loop. Here, we report that several aminoglycosides interact with tmRNA with affin-ity and specificity, preventing its aminoacylation with alanine. The tRNA-like domain of tmRNA is altered upon aminoglycoside binding. Chemical footprints in solution have explored the structural basis of the interaction between tmRNA and several aminoglycosides. trans-Translation is therefore a novel target for aminoglycosides.

EXPERIMENTAL PROCEDURES
Enzymes and RNAs-Alkaline phosphatase and T4 polynucleotide kinase are from New England Biolabs (Beverly, MA). T4 RNA ligase was from Invitrogen. RNases S 1 , V 1 , U 2 , and T 1 were from Amersham Biosciences. [␥-32 P]ATP (3000 mCi/mmol) and [␣-32 P]pCp (3000 mCi/ mmol) were from PerkinElmer Life Sciences. E. coli tmRNA (22) and tmRNA-TLD 2 were overexpressed in E. coli cells and purified as described. Appropriate bands were electroeluted, and pure RNAs were recovered by ethanol precipitation. Total tRNAs from E. coli are from Sigma-Aldrich. Purified E. coli tRNA AlaUGC was a gift from Dr. R. Gillet (Medical Research Council, Cambridge, UK).
Aminoacylation Experiments-Recombinant alanyl-tRNA synthetase was purified on Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen), and purity was confirmed on a 10% SDS-PAGE. Five independent protein purifications were required. The final protein concentration ranged from 0.5 to 3.5 M. The assays were performed at 20 or 37°C in a medium containing 50 mM Tris-HCl (pH 7.5), 10 mM KCl, 20 mM ␤-mercaptoethanol, 10 mM MgCl 2 , 2 mM ATP ([MgCl 2 ]/[ATP] ϭ 5), 0.05 mg/ml of bovine serum albumin, 42-59 M L-[ 14 C]alanine (170 mCi/ mmol), and 50 -330 nM purified E. coli AlaRS. When varying the pH from 5.5 to 9.5, 50 mM MES was used for pH 5.5, and 50 mM Tris-HCl was used for pH levels from 6.5 to 9.5. Usually, 1 M of tmRNA or tmRNA-TLD was denatured for 3 min at 75°C followed by 10 min at room temperature. Then 1 M of purified E. coli His 6 -tagged SmpB and/or His 6 -tagged EF-Tu⅐GDP and/or EF-Tu⅐GTP were added and incubated for 15 min at room temperature. All of the proteins were at least 98% pure as judged by SDS-PAGE analysis. EF-Tu⅐GDP was activated to EF-Tu⅐GTP immediately before use by incubation at 37°C for 15 min in 50 mM Tris-HCl (pH 7.6), 7 mM MgCl 2 , 60 mM NH 4 Cl complemented with 100 M GTP, 6 mM phosphoenolpyruvate, and 10 g/ml of pyruvate kinase. After the incubation, the mixture was kept on ice before use. Various concentrations of aminoglycosides were added, followed by the aminoacylation buffer, the labeled alanine, and the AlaRS. The aliquots were spotted onto 3MM Whatman papers at different times, and precipitated with 5% trichloroacetic acid. Kinetic parameters (K m and V max ) in the presence and absence of aminoglycosides were performed under steady-state conditions of enzyme (50 -330 nM AlaRS) and substrate concentrations of tmRNA (0.5-3.4 M) or tmRNA-TLD (0.5-6 M) and determined from Lineweaver-Burk plots. These experimental conditions were also applied for tmRNA aminoacylation in the presence of all tRNAs from E. coli.
Ultraviolet Absorbance Melting Curves-The progressive melting of tmRNA and tmRNA-TLD was monitored by following their UV absorbency at 258 nm as a function of temperature on a UVIKON XL (Bio-tek Instruments) equipped with a temperature regulator and with a six-cell holder. The temperature was increased gradually at 0.5°C/min from 15 to 93°C. The measurements were done in 20 mM potassium phosphate (pH 5.8), 0.5 mM EDTA, and 5 mM MgCl 2 for tmRNA-TLD and in 20 mM potassium phosphate (pH 5.8), 50 mM NaCl, and 5 mM MgCl 2 for tmRNA in the absence and presence of 500 M neomycin B, 10 mM tobramycin, and 2 mM paromomycin.
Chemical Footprints-Labeling at the 5Ј ends of tmRNA and tmRNA-TLD were performed with [␥-32 P]ATP and phage T4 polynucleotide kinase on RNA dephosphorylated previously with alkaline phosphatase. Labeling at their 3Ј ends was carried out by ligation of [␥-32 P]pCp using T4 RNA ligase. After labeling, tmRNA was gel-purified (5% PAGE), eluted, and ethanol-precipitated. Labeled tmRNA was heated 2 min at 80°C and slowly cooled down for 20 min at room temperature. The reaction mixtures (15 l) contained 200,000 cpm of 32 P-labeled tmRNA or tmRNA-TLD and increasing amounts of aminoglycosides (50 -500 M of neomycin B, 0 -1000 M of paromomycin, and 0 -6 mM of tobramycin) in 50 mM Hepes (pH 7.5), 25 mM KCl, and 2 mM MgCl 2. After an incubation of 10 min at room temperature, ENU, lead-acetate, or iron-EDTA were added: 6.25 l of a solution of ENU saturated in 100% EtOH supplemented with 1 g of total tRNA; 0.3 mM of lead acetate supplemented with 2.5 g of total tRNA and 1 mg/ml of Fe(NH 4 ) 2 (SO 4 ) 2 , 5 mM EDTA, and 12.5 mM dithiothreitol; and 0.25% H 2 O 2 supplemented with 2 g of total tRNA for iron-EDTA mapping (a quick centrifuge spin mixed all four chemicals deposited at four locations inside the Eppendorf tube). The incubation times were 4 h at 37°C for ENU, 7 min at 37°C for lead acetate, and 10 min at 0°C for iron-EDTA footprints. RNAs are ethanol-precipitated, and the pellets are washed twice with 80% EtOH, dried, and counted. Then the RNA fragments are submitted to 8 or 12% PAGE. The results were analyzed on a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). During lead mapping of tmRNA in the presence of either tobramycin or paromomycin, we noticed that gradually increasing the concentration of the aminoglycosides increases the ratio of the uncleaved versus the cleaved RNA. Modifying the lead concentration, the incubation times, or the amount of total RNA added during the reaction could solve the issue. Because the overall conformation of tmRNA changes in the presence of both aminoglycosides, the conformation of tmRNA in the presence of the aminoglycosides might be less sensitive to the action of the structural probe.

Inhibition of tmRNA Aminoacylation by Aminoglycoside
Antibiotics-Aminoacylation experiments with purified E. coli alanyl-tRNA synthetase were performed on two RNAs purified in vivo: tmRNA from E. coli (363 nucleotides; Fig. 1A) and tmRNA-TLD (61 nucleotides; Fig. 1B), a shorter RNA recapitulating the tRNA-like domain of tmRNA. tmRNA-TLD is capable of being aminoacylated with alanine in vitro. 2 The structures of the six aminoglycosides tested are shown in Fig. 1C. Aminoglycosides are subdivided into three major families (23), and representatives of each family were tested. Paromomycin and neomycin B belong to the neomycin family; amikacin, tobramycin, and kanamycin A are members of the kanamycin family; and gentamicin represents the gentamicin family. All six aminoglycosides inhibit aminoacylation of tmRNA with alanine, in a concentration-dependent manner (Fig. 2). In the presence of each aminoglycoside, the decrease of aminoacylation is not a consequence of tmRNA or tmRNA-TLD degradation (not shown). Neomycin B is the most efficient inhibitor, with apparent inhibition constants (K i ) of ϳ35 M for tmRNA-TLD and ϳ70 M for tmRNA ( Fig. 2 and Table I). Neomycin B inhibits the aminoacylation of tmRNA-TLD with alanine at a ϳ30-fold lower concentration than that required to inhibit tRNA Ala aminoacylation (Table I). Paromomycin, gentamicin, and amikacin are also potent inhibitors, whereas kanamycin A and tobramycin are modest ones, with a substantial difference in activity between the two RNAs (kanamycin A and tobramycin have no measurable K i on tmRNA-TLD alanylation). Compared with tmRNA, aminoacylation of tmRNA-TLD is half-inhibited at ϳ2-fold higher concentrations of paromomycin, gentamicin, and amikacin (the opposite effect is observed for neomycin B). Aminoacylation with alanine of a tRNA Ala transcript can also be altered by these aminoglycosides, but the inhibition constants are 3-15-fold higher than that of tmRNA. Amikacin, kanamycin A, and tobramycin have related structures (Fig. 1C), although different inhibition constants (Table I), especially between amikacin and the two others. Between amikacin and tobramycin, the only differences are at positions R1 and R2 (Fig. 1C). Therefore, the chemical groups substituted at positions R1 and/or R2 of amikacin are responsible for its lower K i . Replacing a hydroxyl group (paromomycin) by an amino group (neomycin B) decreases the K i for tmRNA 3-fold (Table I), indicating that a minor modification on the chemical structure of an aminoglycoside can significantly increase its inhibitory activity on tmRNA aminoacylation. All of the subsequent functional and structural investigations were conducted using three representatives, neomycin B, and paromomycin as potent inhibitors and tobramycin as a modest inhibitor of tmRNA aminoacylation.
The number of charged amino groups is important for inhi-bition of RNA function by aminoglycosides. Neomycin B has six protonatable amino groups; their corresponding pK a values are indicated in Fig. 1C. In most cases, inhibition of RNA function by aminoglycosides depends on the number of charged amino groups (24). tmRNA-TLD inhibition by neomycin B (at its K i of 50 M) was monitored at five different pH levels (5.5, 6.5, 7.5, 8.5, and 9.5), centered on pH 7.5, which was used in our assays (not shown). At pH 7.5, neomycin B inhibits 50% of aminoacylation; when pH increases up to 9.5, neomycin B has no inhibitory effect on aminoacylation (the amino groups are mostly deprotonated). At pH 8.5, 70% of aminoacylation is detected; when pH decreases to 6.5 and 5.5, inhibition of aminoacylation by neomycin B is more efficient, leaving only 35 and 20% residual aminoacylation, respectively. Therefore, inhibition of tmRNA aminoacylation by neomycin B depends on the number of charged amino groups and suggests that complex formation between tmRNA and neomycin B is mainly driven by electrostatic interactions. ATP is required during the first step of all the aminoacylation reactions to form the activated aminoacyl-adenylate (in our case an alanyl-adenylate). Therefore, aminoglycosides could bind specifically to the catalytic site of E. coli AlaRS as structural analogs of ATP. To test this hypothesis, increasing concentrations of ATP from 1 to 64 mM were added in the aminoacylation reaction of tmRNA-TLD with and without 500 M paromomycin and in the aminoacylation reaction of tmRNA with or without 2 mM of tobramycin ([MgCl 2 ] is kept constant at 20 mM). Increasing the concentration of ATP does not relieve the inhibitory effect caused by the two aminoglycosides on the aminoacylation of both RNAs (not shown). Thus, neither paromomycin nor tobramycin function as structural analogs of ATP in the aminoacylation reaction of tmRNA-TLD and tmRNA with alanine, respectively.
The Michaelis-Menten parameters of tmRNA-TLD and tmRNA aminoacylation with alanine in the absence and presence of neomycin B, paromomycin, or tobramycin have been determined (Table II). In the absence of aminoglycosides, the K m for tmRNA-TLD and tmRNA are essentially the same (5-6 M), but the V max for tmRNA aminoacylation is 1 order of magnitude lower than that of tmRNA-TLD (Table II). Compared with tmRNA, the smaller size of tmRNA-TLD may allow a faster turnover of the substrate during the aminoacylation reaction. For both RNAs, gradually increasing the concentration of either neomycin B (from 15 to 60 M for tmRNA-TLD), paromomycin (from 50 to 500 M for tmRNA and from 100 to 1000 M for tmRNA-TLD), or tobramycin (from 1 to 3 mM for tmRNA and from 3 to 10 mM for tmRNA-TLD) increases the K m up to 30-fold (Table II), but the V max also increases (not shown). Alanylation is not completely inhibited at saturating antibiotic concentration; there is significant residual aminoacylation level at antibiotic saturation with tmRNA-TLD and in the presence of amikacin and kanamycin A with tmRNA (Fig. 2). Thus, the inhibition of tmRNA aminoacylation by aminoglycosides is neither competitive nor non competitive; therefore the K i values could not be measured.
Inhibition of tmRNA Aminoacylation in the Presence of All tRNAs from E. coli-We tested whether aminoglycosides can inhibit tmRNA aminoacylation in the presence of all 46 tRNAs from E. coli (Table III). For this experiment, we selected paromomycin at a concentration corresponding to its apparent in-hibition constant for tmRNA alanylation (Table I). Native tRNA Ala isoacceptors represent ϳ6% of the total tRNA population in E. coli (25). In all of the tRNAs from E. coli, up to 60% of the tRNA Ala isoacceptors are able to charge alanine (100% of charging is not observed probably because our calculated number of tRNA Ala present in the mixture is overestimated; alternatively, a fraction of tRNA Ala might be degraded). In the presence of all tRNAs from E. coli, there is no inhibition on the aminoacylation of the tRNA alanine isoacceptors (Table III). This result is consistent with the fact that purified tRNA Ala UGC transcript has a ϳ4-fold higher K i for paromomycin, compared with tmRNA (Table I). When tmRNA is mixed with all of the tRNAs from E. coli, paromomycin decreases the aminoacylation level, even in the presence of a 10-fold excess of total tRNAs (Table III). Because paromomycin at that concentration has no effect on the alanylation of the tRNAs, the decrease in tmRNA aminoacylation with alanine is responsible for the overall decrease in the charging levels of the RNA mixture containing both tRNAs and tmRNA. Thus, paromomycin can inhibit tmRNA alanylation in the presence of all of the tRNAs from E. coli.  Magnesium Ions Cannot Reverse the Inhibitory Effect Caused by the Aminoglycosides-Positively charged ammonium groups of aminoglycosides match negatively charged metal ion-binding pockets in RNA three-dimensional structures, displacing divalent metal ions (26). Recent structural (17) and functional (18) data indicate that the rule also applies for specific interactions between canonical tRNAs and aminoglycosides. If true for tmRNA, inhibition of tmRNA aminoacylation by aminoglycosides might be overcome by increasing the magnesium concentration. The [Mg 2ϩ ]/[ATP] ratio was kept constant at 5, as in the previous aminoacylation assays, because it affects the specificity and efficiency of the aminoacylation reaction (27). The magnesium ion concentration was gradually increased, and the aminoacylation of tmRNA-TLD and tmRNA, with and without paromomycin and tobramycin, respectively, was measured (Fig. 3, A and B). In the presence of paromomycin, aminoacylation of tmRNA-TLD slightly increases up to 20 mM [Mg 2ϩ ], whereas in the presence of tobramycin, aminoacylation of tmRNA show almost no difference up to 10 mM [Mg 2ϩ ]. At higher magnesium concentrations, the aminoacylation plateau decreases for both RNAs with and without aminoglycosides, probably because elevated magnesium concentrations start degrading the RNAs (not shown). Therefore, for both RNA-aminoglycoside interactions, increasing the magnesium concentration during the aminoacylation reaction cannot rescue the inhibitory effect caused by the aminoglycosides on aminoacylation. Using lead acetate as a probe, we performed a footprinting experiment between 500 M neomycin B and labeled tmRNA-TLD at various magnesium concentrations (1, 2, 5, and 10 mM; not shown). The result is that the footprints do not depend on magnesium concentration and therefore strengthen our claim that magnesium does not rescue aminoglycoside inhibition.

Inhibition of tmRNA Aminoacylation by Neomycin B in the Presence of Specific Ligands-SmpB (small protein B) and EF-
Tu⅐GTP both enhance alanine accepting activity of tmRNA, and SmpB protects the 3Ј end of Ala-tmRNA against enzymatic degradation (28 -30). 1 M of tmRNA with post-transcriptional modifications charges alanine to 26% on average, but there is a half-increase in charging when either 1 M of purified SmpB or 1 M of purified EF-Tu⅐GTP is present (Fig. 4, inset). A 2-fold increase is observed when both proteins are present at 1 M. This indicates that in our aminoacylation assays, both SmpB and EF-Tu⅐GTP bind tmRNA and enhance its aminoacylation with alanine. With EF-Tu⅐GDP, the charging level of tmRNA is not significantly affected, but when in the presence of SmpB, the aminoacylation plateau also increases 2-fold (Fig. 4, inset). We verified by native gel shift assays that SmpB, EF-Tu⅐GTP, and EF-Tu⅐GDP bind deacylated tmRNA (not shown). In the presence of an equimolar ratio of SmpB, neomycin B can still inhibit aminoacylation, but the K i increases 1.5-fold, compared with tmRNA alone. In the presence of EF-Tu⅐GTP, the K i increases 5-fold; in the presence of both proteins, the K i increases 10-fold, compared with tmRNA alone (Fig. 4). In the presence of 1 M EF-Tu⅐GDP, the K i increases 2.5-fold; in the presence of 1 M SmpB and 1 M EF-Tu⅐GDP, the K i also increases 10-fold, compared with tmRNA alone (Fig. 4, inset). These data indicate that in the presence of SmpB and EF-Tu, tmRNA aminoacylation is significantly protected against the inhibitory effect of aminoglycosides.
Thermal UV Melting Profiles tmRNA-TLD in the Presence of Aminoglycosides-UV absorbance melting curves were performed to assay the effects of aminoglycosides on tmRNA and tmRNA-TLD structures. In the presence of 5 mM MgCl 2 , the melting profile of tmRNA is multiphase, with a first transition around 40°C likely corresponding to the unfolding of its tertiary structure; a series of smaller transitions from 60 to 90°C correspond to the progressive unfolding of the secondary structure (not shown). Because the unfolding pathway of tmRNA is so intricate, no clear picture emerges, e.g. stabilization or destabilization, when aminoglycosides are present (not shown). For tmRNA-TLD, however, UV melting experiments have provided information about the effects of aminoglycosides on the RNA structure (Fig. 5). In the absence of aminoglycosides, tmRNA-TLD structure unfolds in two transitions; the first one is centered on a calculated melting temperature (T m ) of around 35°C; the second transition occurs at higher temperature, with a T m close to 80°C (Fig. 5A). That second transition is broad and noncooperative, suggesting that several conformations of the RNA coexist in solution and unfold independently when the temperature increases. 10 mM tobramycin does not affect the lower transition, whereas the T m corresponding to the second transition is considerably reduced (a decrease of 18°C), to 60°C (Fig. 5B). Also, the second transition becomes sharper and cooperative. A similar destabilizing effect on tmRNA-TLD structure is observed in the presence of 2 mM paromomycin, with no significant effects on the lower transition, but a 16°C drop of the T m corresponding to the second transition (Fig. 5C). As for tobramycin, the second transition, although shifted to a lower temperature, is nevertheless sharper and more cooperative compared with tmRNA-TLD alone. Therefore, both aminoglycosides reduce the conformational freedom of tmRNA-TLD and stabilize a conformation that is no longer aminoacylatable with alanine. In the presence of 500 M neomycin B, the T m corresponding to the first transition is not affected, but the T m of the second increases of about 10 degrees, up to 90°C (not shown). Therefore, upon binding tmRNA-TLD, neomycin B stabilizes its conformation.
Chemical Footprints of tmRNA-TLD with Neomycin B-To gain structural information about the interaction between tmRNA-TLD and neomycin B, lead acetate footprints, in the absence and presence of increasing concentration of neomycin B, were performed (Fig. 3, C and D). Nine nucleotides (A 8 -U 12 and A 15 -C 18 ) from the D-analog of tmRNA-TLD and nucleotides G 31 and A 32 are cleaved by lead but become protected in the presence of increasing concentrations of neomycin B (Fig. 3, C  and D). Six nucleotides involved in the three base pairs U 6 -A 52 , G 7 -C 51 , and G 34 -C 50 are cleaved by lead in the absence of the aminoglycoside, suggesting that these pairs are unstable in tmRNA-TLD; when neomycin B is present, all of the six cleavage sites disappear or are significantly reduced, suggesting that neomycin B, upon binding tmRNA-TLD, stabilizes the three pairings (result consistent with the increase of the T m corresponding to the second transition). There are no visible changes in the conformation of stem-loop H5, the five upper pairs in H1, and the four upper pairs in the T-stem and the T-loop, with and without the aminoglycoside.
Chemical Footprints of tmRNA with Paromomycin and Tobramycin-The phosphates (ENU) or nucleotides (lead) of tmRNA that are protected or that become accessible in the presence of the aminoglycosides are indexed on the right-hand sides of the four upper panels (A-D) in Figs. 6 and 7. The footprints of both aminoglycosides are summarized on the secondary structure of E. coli tmRNA. Mapping of ribose accessibility with iron-EDTA shows no differences in the absence and presence of increasing concentrations of both aminoglycosides (Figs. 6 and 7, panels E are representatives; additional experiments were performed with longer migration times and 3Ј labeling of the RNA to get resolution of the upper part of the gels), suggesting that they do not contact any sugars from tmRNA backbone.
Mapping the phosphates from tmRNA by ENU in the absence of aminoglycoside shows that of 363 phosphates, 39 are cleaved by the probe and therefore accessible in the native conformation of tmRNA (Figs. 6 and 7, A and B). Paromomycin induces 29 concentration-dependent chemical footprints onto tmRNA structure, 15 against ENU and 14 against lead (Fig. 6,  A-D). With lead, two positions that are not accessible in tmRNA structure, C 44 in helix H2 and G 150 in the helical portion of pseudoknot PK2, become reactive in the presence of paromomycin. Because of the conformational flexibility of  ENU (A and B), lead acetate (C and  D), and iron-EDTA (E). Autoradiograms of 8% denaturing PAGE of cleavage products of 5Ј and 3Ј labeled tmRNAs, with similar migration times. Lanes C, incubation controls; lanes G L , RNase T 1 hydrolysis ladder; lanes A L , RNase U 2 hydrolysis ladder. The sequence is indexed on the left sides. Nucleotides with black and white triangles are Gs and As, respectively. Nucleotides indexed on the right sides of the autoradiograms are the identified footprints. Footprints of the aminoglycoside are indicated onto tmRNA secondary structure, shown schematically. Only the protections or enhancements of reactivity of nucleotides that vary according to the concentration of the aminoglycoside were considered as reliable data. The black dots are the phosphates protected against ENU cleavages in the presence of the aminoglycoside. The gray dots are the nucleotides protected against lead cleavage in the presence of the aminoglycoside. The black stars are the positions that are protected by both ENU and lead in the presence of the aminoglycoside. The gray squares are the nucleotides that become accessible to lead cleavage in the presence of the aminoglycoside. With ENU, some nucleotides become accessible in the presence of the aminoglycoside, but all of these cleavage sites are already present in the control lane; thus, they were omitted on purpose. The footprints that are specific to the aminoglycoside are underlined (those that are common to paromomycin and tobramycin are not). H5), and G 333 (at one end of H5). When paromomycin binds tmRNA, these accessible sites become protected (direct contact between the aminoglycoside and tmRNA or indirect effect of paromomycin on the conformation of tmRNA). Increasing concentrations of paromomycin modify the reactivity of nucleotides or phosphates located in the tRNA-like domain, H5-PK4, PK1-H3-H4, and PK2 but not in H6 and PK3. Within the tRNA-like domain of tmRNA, seven positions (U 9 , U 17 , U 328 -U 330 , G 333 , and C 353 ) become protected by paromomycin, as tobramycin also does but at much higher concentrations (Fig. 7).
Tobramycin induces 31 concentration-dependent chemical footprints onto tmRNA structure, 14 against ENU and 17 against lead (Fig. 7, A-D). As for paromomycin, C 44 in helix H2 became reactive toward lead cleavage. Three nucleotides, G 61 in PK1, G 150 in PK2, and G 288 in PK4 are protected against both lead and ENU cleavages (Fig. 7, black stars). The footprints are in the tRNA-like domain, PK1-H3-H4, and PK2-PK3-PK4. Paromomycin and tobramycin share 20 common footprints located for the most part in the tRNA-like domain, the tag reading frame, and PK2. Of the common footprints between tmRNA and both aminoglycosides, seven are in the tRNA-like domain, indicating that tight binding requires full-length tmRNA. This result is in agreement with a higher K i for tmRNA-TLD aminoacylation by both aminoglycosides, compared with tmRNA ( Fig. 2 and Table I). However, there are significant differences between both aminoglycosides (underlined nucleotides in Figs. 6 and 7): paromomycin, but not tobramycin, modifies specifically the reactivity of a cluster of nine nucleotides centered on an internal bulge in H5 and in PK4. Tobramycin, but not paromomycin, modifies specifically the reactivity of 11 nucleotides, in PK1, in and around the tag, in PK3 and PK4. Altogether, our probing data suggest that when aminoglycosides bind tmRNA, there is a significant conformational rearrangement of the RNA, as already suggested by our UV melting data collected on tmRNA-TLD. DISCUSSION Functional and structural evidence is provided to demonstrate that aminoglycoside antibiotics interact in vitro with tmRNA from E. coli and modify its conformation in solution, resulting in its inability to be efficiently aminoacylated with alanine by alanyl-tRNA synthetase. tmRNA aminoacylation was most strongly inhibited by neomycin B and by paromomycin followed by, in descending order, gentamicin, amikacin, kanamycin A, and tobramycin (Fig. 2). When tmRNA is reduced to its tRNA-like domain, the concentration of paromomycin, gentamicin, and amikacin required for half-inhibition of aminoacylation increases ϳ2-fold (Table I), indicating that structural domains of tmRNA outside the tRNA-like core are required for optimal inhibition of aminoacylation. This functional result is confirmed by our structural analysis of tmRNA in complex with two different aminoglycosides, because paromomycin and tobramycin both induce chemical protections of accessible bases and phosphates located outside the tRNA-like domain (Figs. 6 and 7). Therefore, with the exception of neomycin B, efficient binding of aminoglycosides requires fulllength tmRNA. Neomycin B interacts specifically with yeast tRNA Phe (17), rationalizing that the tRNA portion of tmRNA is sufficient for efficient binding.
Between tmRNA-TLD and tRNA AlaUGC , the concentrations of paromomycin, gentamicin, and amikacin required for halfinhibition of aminoacylation also increase ϳ2-fold (neomycin B is an extreme case, with a 30-fold difference). Specific sequences and/or structural features of tmRNA-TLD that are not present in tRNA Ala are responsible. These differences are located in the D-analog and the C 21 -G 30 stem-loop of tmRNA-TLD; there are also a few sequence variations in the acceptor stem and the T stem-loop of canonical tRNA Ala , compared with tmRNA-TLD. Neomycin B protects accessible nucleotides in the D-analog of tmRNA-TLD against lead cleavages that are not present in tRNA Ala , rationalizing the binding specificity of the aminoglycoside on tmRNA-TLD. For an aminoglycoside to inhibit tmRNA aminoacylation, it could prevent the correct recognition of the limited set of nucleotides that specify the alanine identity, all located in the acceptor stem, by the AlaRS; alternatively, they could stabilize a nonfunctional state of tmRNA. Cocrystallization of yeast tRNA Phe with neomycin B reveals that the aminoglycoside is positioned in the deep groove below the D-loop, possibly interfering with the interaction between PheRS and its cognate tRNA through its binding to major tRNA Phe charging determinants (17). The structural data reported here indicate that neomycin B interacts with the D-analog of the tRNA-like portion of tmRNA but not in the upper portion of the acceptor stem where the major recognition determinant for alanylation, a GU pair, is located. Upon binding, neomycin B strengthens three base pairs at the junction between the acceptor and the T stems, perturbing the conformation of the acceptor stem that contains the aminoacylation signals. When tmRNA-TLD aminoacylation is performed at 20°C instead of 37°C, the K i for neomycin B increases about 4-fold (from 35 to 145 M). Neomycin B is a better inhibitor when the temperature increases, perhaps allowing the RNA to adopt an optimal conformation for binding.
The footprints of tobramycin and paromomycin on tmRNA and the fact that aminoglycoside binding can be described in two slopes (at least in some cases; Fig. 2) suggest that several antibiotic moieties bind to each tmRNA molecule. Both paromomycin and tobramycin induce the formation of a favored conformation of tmRNA-TLD, but 7-fold lower concentrations of paromomycin are required, compared with tobramycin (Table I). Without aminoglycosides, the conformation of tmRNA-TLD is poorly defined in solution (Fig. 5), but aminoglycosides trap tmRNA-TLD into an inactive conformation that is either stabilized (neomycin B) or destabilized (paromomycin and tobramycin). Aminoglycosides displace a conformational equilibrium toward nonfunctional tmRNAs; once the conformational switch is triggered, increasing the magnesium concentration cannot restore the aminoacylation capacities of tmRNA (Fig. 3).
tmRNA aminoacylation is required for trans-translation. It is not known whether or not aminoglycosides bind tmRNA in vivo. trans-Translation will be impaired in vivo only if the intracellular concentration of the aminoglycoside, in our case neomycin B, is sufficient to interact with some of the ϳ500 copies of tmRNA/cell (31), in addition to its other RNA targets. In bacteria, aminoglycosides impair various cellular processes; they mostly disturb ribosome decoding and cause misreading of the genetic code (32). Paromomycin binds RNA constructs containing the ribosomal A site with dissociation constants of ϳ1.5 M (33), indicating that the major target of aminoglycosides is the ribosome. They also inhibit RNase P RNA cleavage in vitro. The strongest inhibitor of E. coli RNase P RNA cleavage in the presence of its associated protein is also neomycin B (16), with a K i value of 60 M (paromomycin has a K i value of 190 M on RNase P RNA). Therefore, the K i values of inhibition of RNase P RNA cleavage and tmRNA alanylation by both neomycin B and paromomycin are similar.
tmRNA aminoacylation with alanine can be impaired by an aminoglycoside in the presence of all tRNAs from E. coli (Table  III). In its natural context, tmRNA has specific ligands. Two proteins that interact with tmRNA, SmpB (34) and EF-Tu (35), stimulate tmRNA aminoacylation in vitro (28). When both proteins are added to deacylated tmRNA prior neomycin B, they prevent tmRNA charging from being inhibited by low concentrations of aminoglycoside (Fig. 4). Whereas SmpB has a small protecting effect, EF-Tu, in both the GDP and the GTP forms, has a significant protecting effect against the inhibition of tmRNA aminoacylation by neomycin B. Protection is maximal when both SmpB and EF-Tu are present, suggesting that tmRNA aminoacylation might not be impaired in vivo. RNase T 1 footprints of SmpB on a tmRNA transcript indicate that nucleotides G 333 (G 31 in tmRNA-TLD) and G 336 (G 34 in tmRNA-TLD) are protected by SmpB against RNase T 1 cleavage (28). In the presence of neomycin B, G 31 from tmRNA-TLD is protected against lead cleavage and the G 34 -C 50 pairing is reinforced, indicating that there are some overlaps between the binding sites of SmpB and neomycin B on tmRNA. Deacylated tmRNA forms a complex with either EF-Tu⅐GDP or EF-Tu⅐GTP, and two UV cross-links with EF-Tu⅐GDP are located outside the tRNA part, in H5 (U 268 ) and PK4 (U 308 ) (36). Binding sites between EF-Tu and neomycin B onto tmRNA might overlap. Alternatively, SmpB and EF-Tu could modify the conformation of deacylated tmRNA, with high efficiency when both proteins are present, such as tmRNA becomes an efficient substrate for aminoacylation but an inefficient ligand for neomycin B. In the presence of an equimolar ratio of purified ribosomal protein S1 that also interacts with tmRNA, the inhibition constant of tmRNA aminoacylation with neomycin B increases 3-fold (not shown).
All of the known 298 tmRNA gene sequences (20) possess the recognition determinants for efficient aminoacylation with alanine, suggesting that in all these species the first amino acid of the tag is alanine. The secondary and probably also tertiary structures of these tmRNA sequences might be sufficiently conserved to provide the proper recognition determinants for aminoglycoside binding. Therefore, aminoglycosides are probably also able to interfere with trans-translation in bacterial species other than E. coli. Mutating the acceptor stem of tmRNA to confer histidine acceptance retains its ability of protein tagging in vitro, suggesting that the first alanine of the tag can be substituted by another amino acid (37). Therefore, after specific mutations within its nucleotide sequence, tmRNA could potentially be chargeable by amino acids other than alanine. Paromomycin and tobramycin modify the reactivity of the tag reading frame toward structural probes, likely disturbing reregistration from the stalled ribosome to the tag and preventing tagging of the truncated proteins, even if aminoacylation can proceed. The overall pharmacological effect of aminoglycosides during the treatment of infectious diseases may result from a combination of actions prior and during protein synthesis and also when protein synthesis has stalled or has been interrupted. Aminoglycosides primarily cause misreading of mRNA that leads to the synthesis of nonsense or truncated peptides. If tmRNA function is also impaired by aminoglycosides, the nonsense or truncated peptides will accumulate, speeding up cell death.