Mutational analysis of S12 protein and implications for the accuracy of decoding by the ribosome

Abstract

The fidelity of aminoacyl-tRNA selection by the ribosome depends on a conformational switch in the decoding center of the small ribosomal subunit induced by cognate but not by near-cognate aminoacyl tRNA. The aminoglycosides paromomycin and streptomycin bind to the decoding center and induce related structural rearrangements that explain their observed effects on miscoding. Structural and biochemical studies have identified ribosomal protein S12 (as well as specific nucleotides in 16S rRNA) as a critical molecular contributor in distinguishing between cognate and near-cognate tRNA species as well as in promoting more global rearrangements in the small subunit referred to as “closure”. Here we use a mutational approach to define contributions made by two highly conserved loops in S12 to the process of tRNA selection. Most S12 variant ribosomes tested display increased levels of fidelity (a “restrictive” phenotype). Interestingly, several variants, K42A and R53A, were substantially resistant to the miscoding effects of paromomycin. Further characterization of the compromised paromomycin response identified a probable second, fidelity modulating binding site for paromomycin in the 16S rRNA that facilitates closure of the small subunit and compensates for defects associated with the S12 mutations.

Introduction

The ribosome has evolved to select and incorporate cognate aminoacyl tRNA with very high fidelity (with an error frequency on the order of 10^-4-10^-3) during each elongation cycle of protein synthesis ^{1; 2}. Recent studies suggest that the fast but accurate protein synthesis observed during translation is best accounted for by a kinetic discrimination mechanism ^{3; 4}. According to this view, cognate aminoacyl tRNAs are preferentially selected during translation because they induce conformational rearrangements that result in the acceleration of forward rate constants for GTPase activation and tRNA accommodation, and thus acceptance into the ribosome. Consistent with this model, the miscoding antibiotics paromomycin and streptomycin stimulate the forward rate constants shown to be critical to the induced fit mechanism ^{5; 6}. Similarly, detailed analysis of a tRNA mutant that promotes miscoding indicated that it acts by accelerating forward rate constants in the tRNA selection scheme, arguing that the kinetically driven “induced fit” mechanism is the primary means by which selectivity is achieved in vivo ⁷.

Crystallographic studies on the ribosome have provided substantial independent evidence for the induced fit model of decoding ^{8; 9}. A number of local and global conformational changes are induced in the small ribosomal subunit specifically upon binding of a cognate anticodon stem loop (ASL) to the A site. Locally the Watson-Crick geometry of the minor groove of the codon:anticodon minihelix is recognized by 16S rRNA nucleotides A1492, A1493, G530, C518 and C1054. Amino acid residues P44 and S46 of S12 are involved in interactions that help position the 16S rRNA residues C518 and A1492 so that they can make these critical contacts with the codon:anticodon minihelix ⁸. In addition, S12 amino acid residues K42 and R53 make cognate tRNA dependent contacts with helices 27 and 44 of the decoding region. These striking changes in the decoding center are accompanied by more global changes in small subunit structure such that the head and shoulder more closely approach the platform region in a movement referred to as “domain closure” ⁹. Binding of near-cognate ASLs neither promotes these discrete local conformational changes (including the nucleotide rearrangements in the decoding center and the formation of specific contacts by S12 with the platform region of the rRNA) nor the more global closure movements of the subunit, unless the miscoding antibiotic paromomycin is also bound.

Previous detailed analysis of tRNA selection focused on specific contributions made by 16S rRNA nucleotides in the decoding center to the overall process ¹⁰. Here we focus on an analysis of the role played by several highly conserved elements of S12 that either directly contact the codon:anticodon helix upon recognition of a cognate tRNA species or are involved in specific interactions between the shoulder and the platform regions of the subunit upon global domain closure. S12 has been the focus of considerable interest in the ribosome field for many years because streptomycin resistance is most commonly associated with mutations in this protein ^{11; 12; 13}. Further, many of these genetically isolated S12 variants display a hyperaccurate decoding phenotype ¹⁴ that is referred to as “restrictive”. Such mutations typically map near or coincide with the residues described above that are involved in direct recognition of the codon:anticodon minihelix and/or small subunit domain closure. Here we take a general approach to identify specific amino acids in the highly conserved loop regions (PNSA and PGVRY) of S12 that are likely to play critical roles in the tRNA selection process. Characterization of the in vitro decoding properties of several of the most defective variants reveals a likely secondary binding site for the miscoding aminoglycoside paromomycin, which also modulates the fidelity of the decoding process.

Results

RpsL is an essential E. coli gene

As an initial step in characterizing the role of protein S12 in modulating translation, we constructed an E. coli strain with an rpsL (S12) genomic deletion. The experimental details of the strain construction are provided in the Materials and Methods section. Briefly, we used the Datsenko and Wanner system of bacterial gene replacement to insert a selectable kanamycin marker at the S12 locus while complementing for the genomic deletion with a plasmid-encoded S12 gene under the control of the IPTG inducible trp/lac (trc) promoter ¹⁵. The replacement of the genomic rpsL gene by the kanamycin expression cassette was confirmed by PCR analysis and sequencing (data not shown). The engineered rpsL-deleted genomic strain was viable only in the presence of the S12-bearing plasmid and IPTG, even after prolonged incubation at 37°C (Figure 1(a)). These results indicate that S12 is an essential gene for the growth of MG1655 E. coli cells.

Figure 1.

In vivo growth of chromosomal S12 knockout strain and S12 variant strains. (a) Chromosomal S12 knockout strain is viable only when S12 protein is expressed from the plasmid in the presence of IPTG. (b) Part of sequence alignment of S12 protein from bacteria, archaea and metazoans. The residues in dark and light blue show 100% and 80-90% conservation respectively. Arrowheads mark the sites chosen for alanine substitution mutations. The conserved PNSA and PGVR loops are enclosed in boxes. Ec, E. coli; Tt, T. thermophilus; Bs, B. subtilis; Hm, H. marismortui; Sp, S. pombe; Rn, R. norvegicus; Ms, M. musculus; Hs, H. sapiens; At, A. thaliana; Eg, E. gracilis. (c) The S12 residues mutated to alanine (shown as blue sticks) in the present study cluster near the decoding site (left panel). The right panel shows the S12 residues P44 and S46 (shown in blue) interacting with the decoding region upon binding of cognate anticodon-stem-loop (ASL). S12 protein is shown in light blue, neighboring helices (h27 in yellow and h44 including A1492 and A1493 in magenta), A-site codon in orange and ASL in green. (d) The dominant growth phenotype of wild type (WT) or mutant (K42A, P44A, N45A, S46A, R53A, R85A, D88A, L89A, P90A, G91A or R93A) S12. S12 protein was expressed from plasmid carrying a carbenicillin (Carb) resistance marker using IPTG in cells carrying an intact chromosomal copy of the S12 gene. For control cells were plated in the absence of IPTG. (e) The recessive growth phenotype of WT and variant S12 expressed from plasmid with Carb resistance gene. Cells with chromosomal S12 gene replaced with a kanamycin (Kan) resistance marker grow only when expression of the WT or a viable S12 mutant (K42A, N45A, S46A, R53A, R85A, L89A, P90A, G91A or R93A) is induced from the plasmid. Streptomycin (Strep) resistance was assessed by plating in the presence of the antibiotic.

Dominant and recessive growth phenotypes of S12 variant ribosomes

We next generated a series of alanine substitutions in the S12 protein and characterized the in vivo phenotype of the variant ribosomes as well as their in vitro decoding properties. The S12 protein sequences from bacteria, archaea, mammals and plants were aligned to look for highly conserved regions (Figure 1(b)). We found a number of S12 residues conserved across the three kingdoms of life and many of these are clustered in two loops, the PNSA and the PGVRY loops, near the decoding center of the small ribosomal subunit. Most of the streptomycin resistance mutations identified in various bacteria over the last four decades also map to these two loops. Given the likely functional significance of both loops, we selected several well-conserved residues in these regions as our targets for alanine scanning mutagenesis (see arrowheads marked in Figure 1(b) and blue residues in Figure 1(c), left panel) and introduced alanine substitutions into the plasmid-borne S12 gene (see Materials and Methods). In this list were included proline 44 and serine 46, found in the universally conserved PNSA loop, which form direct contacts with the codon:anticodon minihelix and with the universally conserved and centrally important 16S rRNA nucleotide A1492, respectively (Figure 1(c), right panel)). In addition, several amino acids, K42, R85 and P90, previously associated with streptomycin resistance ¹², were also subjected to the mutagenesis experiment.

The resulting S12-variant plasmids were transformed into the wild type E. coli strain MG1655 and the growth properties of the resulting strains expressing both wild type and variant S12 proteins were evaluated. In the absence of IPTG, where only chromosomally-encoded S12 is expressed, all transformants give rise to colonies that are similar to wild type MG1655 cells in appearance, size and number (Figure 1(d), +Carb). Similarly, when expression of wild type S12 is induced from the plasmid in the presence of IPTG, bacterial growth remains unaltered (Figure 1(d), WT, +Carb+IPTG). By contrast, expression of P44A and D88A substituted S12 confers a dominant lethal growth phenotype whereas expression of K42A and R53A substituted S12 confers somewhat milder dominant growth defects. Expression of the remainder of the S12 variants, including N45A, S46A, R85A, L89A, P90A, G91A and R93A, gave rise to colonies similar in size and appearance to the wild type control.

We were next interested in characterizing the growth phenotype of the mutants in the absence of wild type S12 protein. For this, we used the engineered rpsL-deleted genomic strain prepared above as the donor for a P1 lysate transduction into recipient strains carrying the various mutant S12 plasmids. By selecting for kanamycin resistance, it is possible to identify recipients that have acquired the kanamycin cassette (and thus are lacking the endogenous S12 gene). Transduction into most strains, excluding those carrying the dominant lethal S12 variants P44A and D88A, yielded kanamycin resistant colonies (Figure 1(e)). The kanamycin transductants depended on IPTG induction for growth, as expected if the plasmid-encoded variant S12 is the sole copy of the gene in the organism. The K42A and R53A variant S12 strains exhibited growth defects as evidenced by smaller colony size when compared to wild type, correlating well with the dominant phenotype of these mutants described in the previous section.

Many of the alanine substitutions constructed here are at positions that have previously been implicated in streptomycin resistance. To check if the alanine substitutions at these sites showed resistance to streptomycin, we plated the strains and observed growth in the presence or absence of streptomycin. Of the nine alanine substitutions examined here, only K42A grew in the presence of streptomycin. In fact, the K42A colony size in the presence of streptomycin was larger than in its absence, indicative of a streptomycin pseudo-dependent phenotype (Figure 1(e), K42A, +Carb+Kan+IPTG+Strep).

S12 variant ribosomes are modestly affected in in vitro decoding on cognate codons

We next characterized some of the biochemical deficiencies of the slow growing K42A and R53A strains as well as several other variants located in structurally interesting positions in S12 (N45A, S46A, P90A and R93A). Based on extensive structural and functional studies implicating these regions of S12 in decoding, we focused our biochemical characterization on this particular process. Variant S12-containing ribosomes were purified from the MG1655 rpsL deletion strain (above) by standard procedures ¹⁶. No subunit association defects were apparent during the sucrose gradient purification step of the procedure (data not shown).

Wild type and variant ribosomes were evaluated in several different assays that report on the process of tRNA selection. A minimal kinetic model for this multistep process is depicted schematically in Figure 2(a) ⁴. Variant ribosomes were first evaluated in a dipeptide formation assay that reports on the proofreading stage of tRNA selection ³. This experiment provides two types of information: the observed rate of peptide bond formation reports on the rate-limiting step of tRNA accommodation while the endpoint of the reaction reports on the rejection of aminoacyl-tRNA during the proofreading stage (see kinetic scheme in Figure 2(a)). Reduced levels of dipeptide product are indicative of increased rates of aminoacyl-tRNA rejection following GTP hydrolysis. For this assay, ribosome initiation complexes were formed carrying radiolabeled initiator f-[³⁵S]Met-tRNA^fMet in the AUG-programmed P site and a UUC phenylalanine codon in the A site. To these complexes was added sub-stoichiometric amounts of cognate Phe-tRNA^Phe:EF-Tu:GTP ternary complex. Rates of dipeptide formation were determined from single exponential fits of the data as described previously ³. The data in Figure 2(b) and Table 1 indicate that neither the rate constant nor the endpoint of the peptidyl transfer reaction are dramatically affected in the six variant ribosomes tested. The largest effects that we observe on the rate constant are 2- to 3-fold decreases. The measured rate constants in our experiments with wild type ribosomes match those previously reported under the same high fidelity conditions ³.

Figure 2.

In vitro biochemical characterization of tRNA selection by wild type and S12 variant ribosomes on cognate and near-cognate codons. (a) Kinetic scheme of tRNA selection on the ribosome ⁴. (b) Time courses of dipeptide (fMet-Phe) formation of wild type and S12 variant ribosomes with cognate (UUC) codon in the A site. The data were fit to single exponential kinetics to obtain the respective rates and endpoints of accommodation, reported in Table 1. The endpoints of accommodation of tRNA^Phe ternary complex with a near-cognate (AUC) codon in the A site in HiFi (c) or LoFi (d) buffer. Each bar represents the average of three or more independent experiments.

Table 1.

Kinetic parameters of accommodation and GTP hydrolysis of the cognate ternary complex on wild type (WT) or S12 variant (K42A, N45A, S46A, R53A, R85A, L89A, P90A, G91A or R93A) ribosomes. n.d. not determined.

S12	kobs (acc.), s-1	Endpoint(acc.)	kobs (GTPase), s-1
WT	4.1 ± 0.6	0.96 ± 0.04	22 ± 8.0
K42A	1.6 ± 0.3	0.90 ± 0.08	11.2 ± 2.6
N45A	2.9 ± 0.4	0.82 ± 0.06	6.5 ± 1.7
S46A	1.8 ± 0.2	0.96 ± 0.06	n.d.
R53A	3.0 ± 0.4	0.78 ± 0.08	15 ± 4.2
P90A	2.1 ± 0.3	0.92 ± 0.06	n.d.
R93A	3.9 ± 0.3	0.94 ± 0.04	19.9

Next the rates of GTP hydrolysis by EF-Tu were measured for wild type and four of the variant ribosomes (K42A, N45A, R53A and R93A) in order to evaluate the rate-limiting GTPase activation step during the initial selection phase of decoding. Time courses of GTP hydrolysis were obtained by mixing ribosome initiation complexes (again carrying initiator fMet-tRNA^fMet, though in this case not radioactively labeled, bound to the AUG-programmed P site and neighboring phenylalanine codon, UUC, in the A site) with sub-stoichiometric amounts of FPLC-purified ternary complex (comprised of EF-Tu, Phe-tRNA^Phe and [γ-³²P]-GTP) and measuring the amount of GTP converted to GDP over time under high fidelity buffer conditions ³. Single turnover rates of GTP hydrolysis derived from single exponential fits of the data indicated that the variant ribosomes tested had relatively minor deficiencies ranging from 2- to 3-fold (Table 1). These effects were very consistent with those observed in the peptidyl transfer assay described above.

S12 mutations are error restrictive in vitro

As discussed, streptomycin resistance mutations generally map to protein S12 and most of these variants exhibit increased levels of discrimination in the tRNA selection process ¹⁴. Because our site-directed mutations in S12 lie intermingled with the previously isolated streptomycin resistance variants, we decided to evaluate the in vitro miscoding properties of the S12 variants relative to wild type ribosomes. As above, initiation complexes were formed on wild type and variant ribosomes, except that in this case the message carries an AUC codon in the A site which is near-cognate for tRNA^Phe (cognate codon UUC). Sub-stoichiometric ternary complex was again added to these initiation complexes and the amount of dipeptide (f-Met-Phe) formed at the completion of the reaction was evaluated (the “endpoint”). The experiment was performed both in the standard high fidelity buffer system used above (Figure 2(c)), as well as in lower stringency buffer conditions (Figure 2(d)). Both conditions revealed fidelity phenotypes associated with the variant ribosomes. Under the more permissive conditions, all S12 variants exhibited markedly greater fidelity than wild type ribosomes with endpoint differences ranging from 2- to 8-fold lower (Figure 2(d)). Of the variant ribosomes, K42A, R53A and P90A showed the most accuracy whereas N45A, S46A and R93A showed an intermediate degree of accuracy. These data are consistent with models suggesting that mutations in the loop regions of S12 might disrupt ribosomal interactions required for small subunit closure during tRNA selection.

Effects of streptomycin and paromomycin on variant S12 ribosomes during decoding

Earlier work established that the addition of certain aminoglycoside antibiotics can suppress defects in tRNA selection associated with closure defects ^{5; 6; 10}. We examined whether the observed in vitro restrictive biochemical properties of the variant S12 ribosomes are suppressed by the aminoglycosides paromomycin and streptomycin. For these experiments, initiation complexes were again formed with the phenylalanine near-cognate codon, AUC, in the A site and these complexes were incubated in high fidelity buffer conditions in the presence or absence of streptomycin (30 μM) or paromomycin (20 μM). These concentrations of aminoglycoside were chosen because they are well above the presumed K_d of these compounds for wild type ribosomes based on earlier footprinting experiments ¹⁷. The treated initiation complexes were subsequently allowed to react for a short time (30 seconds at 20°C) with sub-stoichiometric amounts of Phe-tRNA^Phe ternary complex and the extent of dipeptide formation was determined. With wild type ribosomes, the extent of near-cognate tRNA accommodation was very low in the absence of antibiotics (Figure 3(a)) and was dramatically increased in their presence (either with streptomycin or paromomycin) such that essentially all of the Phe-tRNA^Phe was incorporated into the dipeptide product. As described above, in the absence of antibiotics the variant S12-containing ribosomes incorporated less near-cognate tRNA than wild type. However, in the presence of the antibiotics, the variants displayed rather interesting behaviors. First, all of the mutants except K42A responded rather uniformly and robustly to streptomycin (Figure 3(a)), allowing for complete incorporation of the near-cognate Phe-tRNA^Phe into dipeptide at the chosen short time point. The streptomycin-refractive phenotype of the K42A variant is consistent with the fact that K42 makes direct contacts with streptomycin in the crystal structure, and that mutations at this position have often been associated with streptomycin resistance. The refractive behavior of K42A was also consistent with the survival of this variant on streptomycin-containing media in vivo (Figure 1(e)).

Figure 3.

Modulation of the decoding properties of wild type and S12 variant ribosomes by miscoding antibiotics. (a) The endpoint of dipeptide (fMet-Phe) formation with a near-cognate (AUC) codon in the A site of wild type (WT) or S12 variant ribosomes (K42A, N45A, S46A, R53A, P90A, R93A), in the absence or presence of streptomycin (Strep) and paromomycin (Paromo). A similar endpoint analysis was carried out with varying concentrations of paromomycin (0-2 mM). The fraction TC converted to fMet-Phe was plotted as a function of paromomycin concentration and fit to single (a - WT, K42A, R53A) or double (c - N45A, S46A, P90A, R93A) hyperbolic equations. The inset in (c) shows the data points from 0 to 20 μM paromomycin, fit to a single hyperbolic equation.

The responses of the variant ribosomes to paromomycin were more varied. None of the six variants responded as fully as the wild type to paromomycin in the miscoding assay and two variants (K42A and R53A) were almost completely refractive (Figure 3(a)). While the amount of paromomycin used for this endpoint assay (20 μM) was well above presumed (but not directly measured) K_d for wild type ribosomes ¹⁷, it seemed possible that the mutations in S12 might disrupt the binding sites for the aminoglycosides, thus making them less effective at inducing miscoding. We note that paromomycin is known to bind to isolated 16S rRNA elements with considerably higher affinity (K_d of around 100 nM) ¹⁸. To test this possibility, we repeated the endpoint assays using variable concentrations of the antibiotic to see if paromomycin-mediated miscoding could be observed. Indeed, each of the variant ribosomes does eventually respond to increasing levels of paromomycin (Figure 3(b) and (c)). Interestingly, the data for wild type, K42A and R53A ribosomes fit nicely to simple hyperbolic functions yielding K_1/2 values of 1.0, 1160 and 1344 μM, respectively (Figure 3(b)), where the K_1/2 is defined as the amount of paromomcyin required for half-maximal miscoding under these chosen conditions. The data for the remaining four variants (N45A, S46A, P90A and R93A) were not fit well by a single, but rather by a double, hyperbolic function (Figure 3(c)). In each case, the first phase of the curve yielded a relatively low K_1/2 value in the micromolar range (Figure 3(c), inset) whereas the second phase yielded a substantially higher K_1/2 value in the millimolar range. The term K_1/2 is used loosely here and is meant to provide a means for comparing the responses of a set of variant ribosomes to the miscoding effects of the aminoglycoside paromomycin.

At least one potential explanation for the poor responses of the variant ribosomes to paromomycin is that binding to this compound is compromised. Careful inspection of the structural data for the small subunit however suggests that the sites of mutation on S12 being studied are physically remote from the well-described paromomycin binding site in helix 44 (h44) ¹⁹. These observations led us to suspect that the observed K_1/2 values may not be simple reporters of paromomycin binding affinity.

Direct measurement of paromomycin binding to S12 variant ribosomes

Paromomycin binding to the small subunit of the ribosome can be directly assessed using chemical modification analysis and a previously described dimethylsulfate (DMS) protection at position 1408 of helix 44 of 16S rRNA ¹⁷. To evaluate binding of paromomycin to wild type and S12 variant ribosomes, we prepared ribosomes programmed with a model mRNA and deacylated initiator tRNA^fMet in the P site and then added paromomycin at 5 μM. These complexes were subsequently treated with DMS and primer extension analysis was used to evaluate A1408 protection patterns (Figure 4(a)). As seen in Figure 4(a), paromomycin (5 μM) strongly protects A1408 from DMS modification on wild type and three chosen variant ribosomes (K42A, R53A and R93A). Notably, we chose to analyze two of the variant ribosomes (K42A and R53A) that were most defective in their response to paromomycin. Similar modification experiments were next performed using varying concentrations of paromomycin to determine the K_ds of the variant ribosomes for this aminoglycoside. Binding curves for wild type, K42A and R53A ribosomes were indistinguishable from one another, each yielding K_d values for paromomycin in the range of 100 nM (Figure 4(b)). These data argue that binding of paromomycin in the well-defined site in helix 44 (which A1408 protection likely reports on) is unperturbed in all of the variant ribosomes.

Figure 4.

Paromomycin binding analysis of wild type and S12 variant ribosomes. (a) Wild type or mutant (K42A, R53A and R93A) ribosome complexes were incubated in the presence or absence of Paromomycin (Par) before carrying out DMS footprinting analysis of the h44 region of the 16S rRNA. A wild type control reaction was carried out in the absence of DMS (lane 3). (b) Wild type, K42A or R53A ribosomes complexes were incubated with varying concentrations of paromomycin before carrying out the DMS footprinting analysis. The normalized band intensity of the A1408 band was plotted as a function of paromomycin concentration and fit to hyperbolic equations to obtain the K_d of 100 nM for the wild type as well as K42A and R53A ribosomes. (c) Wild type ribosome complexes were incubated with 0 to 2 mM Paromomycin (lanes 1-9), modified with DMS and primer extension analysis was carried out focusing on the h27 region of the 16S rRNA. (d) The normalized intensity of the DMS modification bands A908 and A892 bands was plotted as a function of paromomycin concentration for WT and K42A variant ribosomes. The A892 band intensity was fit to a hyperbolic equation to obtain a K_d of 300 μM for wild type and K42A ribosomes.

Because the variant ribosomes apparently bind paromomycin in the helix 44 site with high affinity, and yet respond to paromomycin in the miscoding assay only at much higher concentrations of the antibiotic, we wondered whether a second lower affinity binding site for paromomycin might exist to which the variants were responding. Importantly, this second binding site should also affect ribosome fidelity. There are several other known binding sites for related aminoglycosides on the ribosome ²⁰. For example, hygromycin binds in helix 44 of the 16S rRNA at a site distinct from the paromomycin binding site while streptomycin binds in nearby helix 27. Hygromycin primarily inhibits translocation whereas streptomycin influences tRNA selection. A secondary binding site for paromomycin was recently identified in helix 69 of the 23S rRNA and it is thought that this binding site may explain some of the observed effects of this compound on ribosomal recycling ²¹.

We reasoned that paromomycin might bind with lower affinity in one of these alternative sites and were particularly interested in the streptomycin binding site based on its known effects on tRNA selection and miscoding. We searched for chemical modification protection patterns in the helix 27 and 44 regions of the 16S rRNA (as well as helix 69 of the 23S rRNA) in wild type ribosomes with increasing amounts of paromomycin. Indeed, in addition to the characteristic protection of A1408 with paromomycin in the micromolar range, we observed that A908 reactivity with DMS is first enhanced by paromomycin in the micromolar range and then is protected by this compound in the millimolar range (Figure 4(d)). We also observe an enhancement in the reactivity of A892 towards DMS at these same concentrations of paromomycin. Similar protection and enhancement profiles for nucleotides A908 and A892 by paromomycin were observed for wild type and K42A ribosomes (Figure 4(d)). We failed to see any new protections of either helix 44 of 16S rRNA (other than A1408) or helix 69 of 23S rRNA that titrate in the millimolar concentration range of paromomycin. We note that this failure to observe paromomycin-dependent protections may simply indicate that there are no convenient DMS-sensitive reporter nucleotides in these regions.

Discussion

The analysis of variant ribosomes carrying substitutions in two highly conserved loops of ribosomal protein S12 provides us with new insights into the mechanism of decoding and accuracy during translation. Substitution of amino acids in both conserved loops of S12 resulted in general in vivo growth defects as well as in specific defects in tRNA selection. Two variants, P44A and D88A, had severe dominant growth defects. These two residues reside at the tips of the two loops that project towards the decoding center of the small subunit (Figure 5(a)). Moreover, the backbone carbonyl of P44 is known to directly coordinate a Mg²⁺ ion together with the O2 of C518, a conserved base of the 16S rRNA, and the 2’ hydroxyl of the third nucleotide of the codon during cognate tRNA recognition ⁸. Given the severity of the phenotype associated with the substitution of an alanine at P44, it seems likely that coordination of this Mg²⁺ is critical to ribosome function. D88 is a universally conserved residue and is β-methylthiolated in E. coli and T. thermophilus ^{22; 23}. Consistent with the dominant lethal phenotype of the D88A mutant analyzed here, Carr et al. ²⁴ were previously unable to obtain mutants at this position of S12. Other substitutions flanking the PNSA loop, K42A and R53A, resulted in milder dominant growth defects that permitted subsequent in vitro characterization. A recent study in T. thermophilus isolated streptomycin resistant, pseudo-dependent and dependent (R, P and D) strains that identified a number of mutations (P41S, K42R, R85C, R85H, K87R, K87E and P90L) in these same regions of S12 ¹². However, in our study alanine substitutions at some of these positions (R85 and P90) failed to confer resistance to streptomycin (Figure 1(e)). Substitution by bulkier side chains is predicted to be more disruptive than substitution by alanine, consistent with the recent genetic analysis of streptomycin resistance variants in S12 in T. thermophilus ²⁴. Of the variants we studied, only K42 makes direct contact with streptomycin in the ribosome structure. Interestingly, K42A was the only variant in our study that exhibited streptomycin pseudo-dependent growth on plates (Figure 1(e)).

Figure 5.

Structural implications for the modulation of decoding by S12 and paromomycin. (a) Structural view of S12 protein (light blue) contacting helix 44 (h44, magenta) and helix 27 (h27, yellow) (PDB accession no. 1FJG) ¹⁹. The sites of lethal mutations in S12 (P44 and D88) are shown in black. S12 residues K42 and R53 (dark blue) make hydrogen bonds with the backbone of A1412 (h44) and A913 (h27) respectively to facilitate domain closure as induced either by the miscoding antibiotics or the cognate ASL. The miscoding antibiotics streptomycin (Strep, red) and paromomycin (Par, green) bind near the decoding region at distinct sites. (b) The interface surface of small ribosomal subunit (PDB accession no. 1N32) ⁹. The boxed region in the left panel is depicted as a cartoon in the right panel. The color coding is essentially as above, except that helix 18 (h18, dark pink), A site codon (orange) and anticodon-stem-loop (ASL, green) are also shown. Streptomycin forms contacts with h27/h44 in the platform domain and h18/S12 (K42) in the shoulder domain of the small subunit. Paromomycin binding to the canonical binding site in h44 (Par1) results in the flipping out of A1492 and A1493 that facilitates binding of near-cognate ASL and domain closure. S12 mutations (notably at residues K42 and R53) are refractory to the miscoding induced by paromomycin binding in h44, stressing the functional importance of these interactions in mediating paromomycin induced domain closure. A secondary, lower affinity binding site for paromomycin (Par2) is documented near the streptomycin binding site in h27 which compensates for the S12 mutations that are refractory to the miscoding induced by h44 paromomycin binding (Par1). Figure rendered using PyMOL ²⁹.

The in vitro biochemical properties of the variant ribosomes provide us with insight into the potential functional role of these conserved loop regions of S12. First, variant ribosomes were modestly defective (around 2- to 3-fold, Table 1) in two standard assays with cognate EF-Tu ternary complexes that measure the rate constants for critical steps reporting on the tRNA selection process (GTPase activation and accommodation) (Table 1). These effects are consistent with those reported previously for streptomycin resistant mutations in S12 where similar approximately 2-fold slower overall rates of acceptance were observed ²⁵. Interestingly, all six variants displayed more obvious perturbations from wild type in their increased ability to distinguish against near-cognate tRNA species as evidenced by lower endpoints in a peptidyl transfer assay (Figure 2(c) and (d)). These apparently disparate effects on cognate and near-cognate species can be understood using the logic of synthetic lethality. The cumulative (synthetic) effects of disruptions in the S12 closure interface and in the decoding helix lead to markedly reduced incorporation of near cognate tRNAs, whereas cognate tRNAs acceptance is only marginally affected by disruption of the S12 closure interface (Figure 5(b)).

The hyper-accurate phenotype (known as “restrictive”) has previously been associated with a number of S12 mutations that confer streptomycin resistance. These mutations cluster in a region of S12 that interacts with helices 18, 27 and 44 of the 16S rRNA (Figure 5(b)). Interactions in this region are believed to secure the closed configuration of the small subunit that is an essential feature of tRNA selection. The S12 mutations studied here very likely disrupt this same interface (and associated closure) and thus yield similar hyper-accurate phenotype. Here, the K42A, R53A and P90A substitutions had the largest effects on fidelity whereas the N45A, S46A and R93A variants had more intermediate effects (Figure 2(c) and (d)). However, only one of these restrictive mutants, K42A, shows resistance to streptomycin both in vivo and in in vitro miscoding assays (Figure 1(e), Figure 3(a)). An interesting outcome of our studies is that it is relatively simple to generate hyper-accurate ribosomes as they are merely a consequence of loss of function (closure) mutations in a region where interactions must exist -- the restrictive phenotype is not necessarily associated with streptomycin resistance.

We note that in our previous analysis of mutations in the universally conserved nucleotides of the decoding center, the rate defects of the variant ribosomes on cognate tRNA species were in the range of 20- to 50-fold, whereas their activities on near-cognate species were so low as to be undetectable ¹⁰. The S12 variants studied here are by comparison considerably less compromised in their biochemical properties, consistent with the fact that they are able to support in vivo growth. That said, we note that while we were able to measure endpoints of the peptidyl transfer reactions with variant ribosomes and near-cognate tRNA species, obtaining accurate rates for the processes was technically challenging under the more interesting high fidelity conditions (because of the very low endpoints and thus signal for the reaction).

Analysis of the response of the S12 variant ribosomes to aminoglycosides led to a surprising set of observations. The interactions of the aminoglycoside antibiotics paromomycin and streptomycin with the ribosome are structurally and biochemically well understood (Figure 5(b)). Streptomycin binds between the platform and shoulder regions of the small subunit near rRNA helices 18 and 27 and ribosomal protein S12. Streptomycin is thought to facilitate closure in this region of the small subunit, thus promoting tRNA acceptance on the ribosome. Paromomycin on the other hand binds entirely within h44 and is thought to induce miscoding by facilitating local conformational rearrangements critical to tRNA selection. Consistent with their propensity to promote these functionally related closure events, these antibiotics have been shown to suppress the defects associated with mutations in universally conserved decoding nucleotides A1492/3 ¹⁰. We were surprised to discover that the S12 variant ribosomes in this study responded robustly to streptomycin (stimulating miscoding at wild type ribosome levels), but that as a group, they responded rather weakly to paromomycin (Figure 3(a)). Based on the location of K42 and R53 at the interface of the shoulder and platform region, it is not surprising that these two S12 variants were the most defective in responding to paromomycin. However, paromomycin titrations revealed that all variants did ultimately respond in the miscoding assay, but that the responses were strikingly different for the various mutants -- the wild type ribosome response was robust at a concentration of paromomycin of 1 μM whereas the K42A and R53A variants only responded at concentrations in the millimolar range. The other variants (N45A, S46A, P90A and R93A) had biphasic responses to paromomycin, as though two separate binding sites were critical to the reporting miscoding response.

Our footprinting analysis revealed that paromomycin binding to its canonical h44 binding site was unperturbed even in the most defective S12 variants K42A and R53A (Figure 4(b)). It also revealed previously unreported alterations in chemical accessibility induced in h27 upon paromomycin binding in the micromolar and millimolar range. First, an enhancement in reactivity of A908 was observed in the micromolar range that likely reports on the substantial conformational rearrangements taking place in h44 on binding this aminoglycoside. In addition, a paromomycin-dependent protection of A908 (as well as an enhancement at A892) titrated in the millimolar concentration range both on wild type and K42A variant ribosomes (Figure 4(d)). The paromomycin responses that we observe in the millimolar concentration range can be simply explained by the postulate of paromomycin binding to a secondary site around helix 27 where streptomycin binds and promotes miscoding. Allen and Noller ²⁶ previously reported altered DMS reactivity of A908 in various fidelity modulating ribosome variants.

We note that the primary effects on miscoding from paromomycin in vivo surely result from binding to the high-affinity site in helix 44, and its downstream consequences. Here, the differential responses of the variant ribosomes to paromomycin provided a tool that yielded new insights into the molecular mechanism underlying closure in the small subunit during tRNA selection. Binding of paromomycin to such a secondary binding site can help to reconcile the miscoding assay results as follows. S12 variant ribosomes have defects in closure because of compromised interactions between the shoulder and platform region of the small subunit. Paromomycin binding in helix 44 promotes decoding center closure but is unable to compensate for closure defects in the S12-interface region (K42A and R53A). While earlier studies argued for significant cooperativity in the decoding center itself to bring about localized closure ^{10; 27}, these results argue that the global closure critical to tRNA acceptance depends on multiple independent closure contacts (Figure 5(b)). Closure at both the decoding center and at the S12 interface must take place for tRNA selection to proceed. Binding of paromomycin at high concentrations near the canonical streptomycin site mimics streptomycin function, making contributions to ribosome closure, such that even the most defective variant ribosomes can robustly miscode. The kinetically driven process of tRNA selection thus depends on the formation of multiple distinct interactions, each playing a critical role in the process.

Materials and Methods

Plasmids

The open reading frame (ORF) of the S12 gene (rpsL) was cloned into the NcoI and BamHI sites of pTRC99b to yield pTRCS12wt. The same PCR product was cloned into a pTRC99b variant with a spectinomycin resistance gene to yield pTRCS12wt-Spec. Mutagenesis of the pTRCS12wt plasmid was performed using the QuickChange method (Stratagene), to generate the alanine substitution mutations in the S12 gene. The resulting plasmids are called pTRCS12XnA (where X stands for the single letter code for the amino acid residue at position n in the S12 gene mutated to an alanine).

RpsL gene replacement and drop test

Initially the rpsL gene of strain BW25141 containing pKD46 and pTRS12wt-Spec plasmids was disrupted by a kanamycin cassette. The method of gene replacement reported by Datsenko and Wanner was followed ¹⁵, with the following modifications. The kanamycin cassette was produced by PCR of the pKD4 vector with primers S12ORFP1 (5’-GAGGACGTTTTATTACGTGTTTACGAAGCAAAAGCTAAAACCAGGAGC TATTTAATGATTGAACAAGATGGATTGC-3’) and S12ORFP2 (5’-TTTAGTTTGACATTTAAGTTAAAACGTTTGGCCTTACTTAACGGAGAACCATT AGAAGAACTCGTCAAGAAGGCG -3’) and was transformed into the BW25141 strain as described to yield the strain S12koBW in which the chromosomal S12 ORF was replaced by the kanamycin cassette complemented by expression of the copy of wild type S12 gene from the plasmid pTRCS12-Spec. The resultant colonies were screened for placement and orientation using genomic PCR and analysis of the PCR product by sequencing. This rpsL knockout strain S12koBW was used as a donor stain in the following P1 transduction experiments in order to replace the rpsL ORF of MG1655 RA+ strain with the kanamycin cassette.

A generalized P1 transduction experiment was performed using S12koBW as the donor strain and MG1655 RA+ carrying the plasmid pTRCS12wt or pTRCS12XnA (see previous section for plasmid nomenclature) as the recipient strain and selection was carried out on kanamycin, carbenicillin and IPTG containing plates to yield S12:KanMGwt or S12:KanMGXnA strains respectively. Repeated streaking on plates containing carbenicillin, kanamycin and IPTG was performed to purify the transformants. The overnight cultures of the purified strains were stored as glycerol stocks.

S12:KanMGwt cells were streaked on carbenicillin and kanamycin containing plates, in the presence or absence of IPTG and incubated overnight at 37°C in order to assess if S12 protein expression was essential for growth. The drop tests were performed as follows. To assess the dominant growth phenotype, single colonies from overnight streaks of MG1655 RA+ cells carrying the plasmid pTRCS12wt or pTRCS12XnA were grown in LB in the presence of carbenicillin and IPTG. Serial dilutions of overnight cultures were plated in the presence of carbenicillin alone or carbenicillin + IPTG and incubated overnight at 37°C. In order to assess the recessive growth phenotype, single colonies from overnight streaks of S12:KanMGwt or S12:KanMGXnA, were grown in LB in the presence of carbenicillin + kanamycin + IPTG. Serial dilutions of the overnight liquid cultures were grown overnight at 37°C on plates containing carbenicillin + kanamycin, carbenicillin + kanamycin + IPTG or carbenicillin + kanamycin + IPTG + streptomycin.

Buffers and Reagents

All experiments were performed in the HiFi buffer A (50 mM Tris HCl (pH 7.5), 70 mM NH₄Cl, 30 mM KCl, 3.5 mM MgCl₂, 0.5 mM spermidine, 8 mM putrescine, and 2 mM DTT) at 20°C except those described in Figure 2 (d) which were done in the LoFi buffer B (50 mM Tris HCl (pH 7.5), 70 mM NH₄Cl, 30 mM KCl, 10 mM MgCl₂, and 2 mM DTT). The footprinting analysis was performed in buffer C (80mM K-HEPES at pH 7.8, 15mM MgCl₂, 100mM NH₄Cl). Chemicals were purchased from Sigma-Aldrich. Radioactive compounds were purchased from Perkin-Elmer Life and Analytical Sciences.

Kinetic analysis

Ribosomes were isolated essentially as described before ¹⁶ and stored at -80°C in convenient aliquots. The slow growth properties of the K42A and R53A strains required long growth periods (12-16 hours) to yield sufficient cell density for the preparative procedure whereas the P44A and D88A variant ribosomes could not be prepared since the cells were not viable. E. coli initiator tRNA (tRNA^fMet) and the phenylalanyl specific tRNA (tRNA^Phe) were procured from Sigma-Aldrich. E. coli Initiation factors (IF1, IF2 and IF3) and Elongation Factor-Tu (EF-Tu) were expressed and isolated as described ¹⁰. The mRNAs were derived from gene 32 and contained a Shine-Dalgarno sequence, an AUG start codon, and a UUC or an AUC as the second codon as in ¹⁰.

GTPase rate assay was performed essentially as described ¹⁰, with the following modifications. The final concentration of the Initiation Complex was 0.8 μM for wild type as well as the mutant ribosomes.

Accommodation rates were determined as follows. The initiation complexes were formed by incubating ribosomes (1 μM) with mRNA (4 μM) and f-[³⁵S]-Met-tRNA^fMet (1.5 μM) in the presence of initiation factors IF1, IF2 and IF3 (1.5 μM each) in HiFi buffer A (described above) for 60 min at 37°C. The initiation complexes were purified from the unbound factors, mRNA and tRNA by pelleting over a 1.1 M sucrose cushion in buffer A without spermidine and putrescine at 70K rpm for 3 hours in a Beckman-Coulter Optima ultracentrifuge. The purified initiation complexes were resuspended in buffer A and the binding of initiator tRNA to the P site assessed by scintillation counting of pre-and post-pelleting complexes, and diluted to 1.6 μM in the same buffer. The ternary complexes were formed by incubating 0.4 μM Phe-tRNA^Phe with 5 μM EF-Tu and 1 mM GTP in buffer A. The reaction for accommodation rates and end points was carried out at 20°C, by mixing equal volumes of 1.6 μM initiation complex with 0.4 μM ternary complex in a quench apparatus. The reaction was stopped by addition of 0.2 N KOH and incubation for 10 minutes at 37°C. This treatment released f-[³⁵S]-Met and f-[³⁵S]-Met-Phe from the tRNAs allowing them to be resolved electrophoretically. Phosphorimager analysis was done using a Typhoon phosphorimaging system. The reaction plots were fit to a first order exponential equation using the Prizm software.

Miscoding Assay

Pelleted initiation complexes and ternary complexes were form essentially as described in the previous section, with the following modifications. A gene 32 derived mRNA with an AUC codon, instead of a UUC codon in the second position was used. For Figure 2(d) initiation complexes were resuspended and diluted in the LoFi buffer B and ternary complexes were formed in buffer B. For the remaining miscoding assays (Figure 2(c) and Figure 3) HiFi buffer A was used throughout.

For the antibiotics miscoding assays 20 μL of 1.2 μM near cognate (AUC) initiation complex were pre-incubated with 0 to 3 mM paromomycin or 45 μM streptomycin in buffer A at 20°C, before mixing with 10 μL of 0.6 μM ternary complex and the incubation continued at 20°C. The reaction was stopped at 30s by diluting 6 μL aliquots with 2 μL of 1 N KOH to stop the reaction and analyzed as described in the previous section.

DMS Footprinting

Ribosome complexes were formed by incubating 20 pmol of wild type or mutant ribosomes with 80 pmol of gene 32 mRNA and 30 pmol of deacylated tRNA^fMet in 25 μL of buffer C (80 mM K-Hepes (pH 7.8), 15 mM MgCl₂, 100 mM NH₄Cl) at 37°C for 30 minutes. The reaction complexes were then mixed with 3 μL of paromomycin (final concentration ranging from 2 μM to 2 mM) in buffer C and incubation continued at 37°C for another 10 min. DMS modification of the ribosome complexes was performed as described ²⁸, with the following modifications. The modification reactions were carried out at 37°C for 10 min in the presence of 70 mM DMS (by adding 1 μL of 1/6, v/v dilution of DMS in ethanol). The RNA was extracted using Ambion RNAqueous kit and analyzed by primer extension as described ²⁸. Primer extension gels were analyzed using a Molecular Dynamics Phosphorimager. The density of the altered bands was normalized to a fixed band in the same primer extension lane, the data was plotted against paromomycin concentration and the plots were fit to a hyperbolic equation to obtain the respective K_ds.

In order to titrate the lower concentrations of paromomycin the protocol described above was modified as follows. The ribosomal complexes were formed by incubating 6 pmol of ribosomes with 24 pmol of gene 32 message and 10 pmol of deacylated tRNA^fMet in 100 μL of buffer C. 5 μL of paromomycin in buffer C was added to a final concentration ranging from 0.1 μM to 5 μM and for DMS modification 4 μL of 1/6, v/v dilution of DMS in ethanol was added. The reaction was stopped, rRNA extracted and analyzed by primer extension as described above.