Mutations were introduced in selected helices of the Escherichia coli 16S rRNA by a two-step PCR using as template the pKK3535 plasmid (19) carrying the wild-type rrnB rRNA operon under the control of the native constitutive promoter P₁P₂. First, portions of the gene upstream and downstream of the mutation site were amplified with complementary extensions (Table 1), which were then used as a composite template for amplification of the entire mutant fragment. Mutant 16S rDNA fragments flanked by AflIII/BglII or BglII/XbaI restriction sites were ligated with the remaining portions of pKK3535 and transformed into the DH5α strain. Plasmids isolated from viable clones were sequenced and retransformed into the TA531 strain, which lacks all seven rrn operons and expresses its rRNA off the pHK-rrnC plasmid (20). Transformants were grown on carbenicillin media to select for pKK3535 derivatives, and loss of pHK-rrnC was confirmed by their sensitivity to kanamycin.

Ribosomes were purified from TA531 cells expressing wild-type or mutant pKK3535 following published protocols with slight modifications (21,22). Cell extracts for translation assays were prepared from MRE600 cells following the same protocols to remove ribosomes from clarified lysates. Ion exchange chromatography on DEAE cellulose (Whatman) was used to prepare tRNA-free extract for tRNA aminoacylation.

Initiation factors IF-1, IF-2 and IF-3 were purified from overexpressing E.coli strains following published protocols (23,24). Elongation factors EF-Tu, EF-Ts and EF-G were expressed and purified as described previously (8). Short 5′-biotinylated mRNA containing a spacer region, strong Shine–Dalgarno sequence and the upstream portion of the T4 gene 32 was chemically synthesized (Dharmacon Research). Aminoacylation and formylation of tRNA^fMet (Sigma) followed standard methods (25) using tRNA-free cell extract and 10-formyltetrahydrofolate as the formyl donor (26). The acp³U47 position of tRNA^Phe (Sigma) was labeled with Cy5-NHS ester (Amersham Biosciences) following published protocols (27). Labeled tRNA^Phe was purified by high-performance liquid chromatography using a TSK-phenyl 5-PW column (Tosohaas) and aminoacylated as described previously (28). Short synthetic DNA oligonucleotides (IDT DNA) were 5′ labeled with ³²P and purified by gel filtration. Synthetic DNA oligonucleotides with amino modification at the 5′ end (IDT DNA) were tagged with Cy3-NHS ester (Amersham Biosciences) according to the manufacturer's recommendations and purified by elution from polyacrylamide gels.

Binding assays were carried out with 10 nM ³²P-labeled oligonucleotides and 100 nM purified ribosomes preheated at 42°C and incubated for 2 h at 37°C (unless indicated otherwise) in Tris-polymix buffer (8). Reactions were resolved on native composite gels containing 0.5% agarose, 2.75% acrylamide, 25 mM Tris OAc, pH 7.5, 6 mM KCl, 2 mM MgCl₂, 1 mM DTT, 1% sucrose and 0.45% DMAPN (29), and quantified on PhosphorImager (Molecular Dynamics).

Translation assays were carried out according to the published protocols (30) using 1 μM purified ribosomes prehybridized with excess unlabeled oligonucleotide in Tris-polymix buffer. Reaction products were resolved on 12% SDS–PAGE and visualized on PhosphorImager (Molecular Dynamics). Ribosomes prepared in the same manner were used for toeprinting analysis as described previously (8).

Purified ribosomes prehybridized with excess Cy3-labeled oligonucleotide were assembled onto 5′-biotinylated synthetic mRNA in the presence of fMet-tRNA^fMet and initiation factors in Tris-polymix buffer following published protocols with slight modifications (31). Initiation complexes were purified using 10–40% (w/v) sucrose gradients in Tris-polymix buffer with 20 mM Mg(OAc)₂ and immobilized on streptavidin-derivatized quartz surfaces as described previously (8). The preparation and delivery of EF-Tu/GTP/Phe-tRNA^Phe (Cy5-acp³U47) ternary complex was according to the earlier protocols (8). Single-molecule fluorescence data collection using total internal reflection microscopy was carried out as described previously (8).

The ribosome is a multicomponent assembly of rRNA and proteins, both of which are potential targets for attachment of fluorescent dyes. To ensure homogeneity and functionality of our ribosomal particles, we devised a method that involves the expression of pure mutant ribosomes that can be labeled by hybridization of fluorescent oligonucleotides to extensions in rRNA hairpins. The oligonucleotide hybridization approach also ensures maximum labeling specificity and stability, while allowing a wide distribution range of target sites throughout the ribosome structure.

Sites for oligonucleotide hybridization were introduced into phylogenetically variable regions of the E.coli 16S rRNA by cloning 22–23 nt extensions that replaced the loops of surface-exposed rRNA helices. Using atomic resolution models of the prokaryotic ribosome as a physical map (32–34) and phylogenetic conservation analysis (based on the comparative RNA website ), seven sites were selected for mutagenesis in the initial approach: helices 6, 10, 17 and 44 in the lower body of the 30S subunit, helices 33a and 39 in the head region, and helix 26 on the platform (Figure 1). The sequence of the rRNA insertion was designed to form a stable, partially mismatched helical extension that would not interfere with the native rRNA fold but would readily unfold and hybridize with an oligonucleotide to achieve efficient rRNA labeling (hairpin 1 shown in Figure 2). The insertions were cloned into rrnB rRNA operon-carrying plasmid pKK3535 (19) and transformed into the TA531 strain. In this strain, all seven chromosomal rrn operons have been deleted and wild-type rRNA is expressed from the pHK-rrnC plasmid (20). All mutant rRNA plasmids, except for those with helix 17 and helix 26 extensions, encoded viable functional rRNAs as demonstrated by the efficient replacement of pHK-rrnC with mutant pKK3535 derivative. The doubling times of E.coli expressing mutant ribosomes were comparable to wild-type strains (data not shown).

Mutant ribosomes expressed from plasmid pKK3535 containing hairpin 1 extensions in helices 10, 33a, 39 and 44 were isolated as pure populations from the TA531 strain; wild-type ribosomes were also expressed and isolated from this plasmid system. To test for binding of the oligonucleotide to the designed ribosomal extension and to optimize annealing conditions, we performed hybridization with ³²P-labeled DNA oligonucleotides targeting various portions of hairpin 1 extension (Table 2). Relative binding affinities were measured using electrophoretic mobility shift assays on native composite gels. The results demonstrate that all sites are accessible for oligonucleotide hybridization, and that ribosome-oligonucleotide interaction is specific, as wild-type ribosomes do not bind oligonucleotides at the concentrations tested (Figure 3A). However, oligonucleotide binding is not very efficient since only small portion of oligonucleotide binds to excess ribosomes. We estimate a K_D of ∼1 μM, which is orders of magnitude greater than that predicted by simple base-pairing thermodynamics of single strands (35). These results indicate that under our conditions the thermodynamics of the target hairpin favors its native helical conformation rather than the oligonucleotide-hybridized form. Increasing the hybridization temperature and time, and testing various salt conditions, only slightly improved binding affinity and was not optimal for ribosome activity (data not shown). Binding affinity was somewhat higher for oligonucleotides targeting the 3′ region of the target hairpin and leaving its 5′ portion single stranded and flexible. Coaxial stacking of the native rRNA helix and the rRNA–oligonucleotide duplex might increase the stability of the complex (data not shown).

To improve binding affinity, a series of target hairpins were designed, in which more mismatches were introduced into the hairpin stem (selected constructs, hairpin 2 through 5, are shown in Figure 2). We expected that destabilization of the rRNA hairpin structure should shift the ribosome–oligonucleotide binding equilibrium towards the bound form. The insertions were cloned into helix 33a and functionality of mutant ribosomes was again tested in vivo in the TA531 strain. Mutants containing insertions with too many mismatches (e.g. hairpin 2) were lethal, possibly due to unfolding, misassembly or degradation of rRNA. Functional ribosomes purified from viable mutants (hairpins 3, 4 and 5) showed very efficient hybridization of oligonucleotide with K_D of ∼15 nM (Figure 3B) even under gentle non-inactivating conditions (Figure 3C).

The functionality of purified mutant ribosomes containing hairpin 5 extension in helix 33a prehybridized with the sp5 oligonucleotide was demonstrated using in vitro translation assays. Comparison of the radioactive protein product yields in translation reactions with and without the oligonucleotide confirms that its presence on the ribosomes does not affect translational activity. Mutant ribosomes translate at more than 90% activity of wild-type ribosomes both in the presence and absence of the oligonucleotide (Figure 4A). In addition, mutant ribosomes prehybridized with the oligonucleotide and assembled into initiation complexes with mRNA and initiator tRNA were shown to bind aminoacyl-tRNA and translocate using standard toeprinting analysis (Figure 4B). In toeprinting assays, reverse transcription of the mRNA is blocked when the polymerase encounters the ribosome. Thus, the length of the reaction product is indicative of the position of the ribosome on the mRNA and can be used to monitor ternary complex binding and translocation. These results demonstrate that our hybridization approach does not interfere with the proper function of the ribosomal particle.

To examine the applicability of the oligonucleotide hybridization method for tagging ribosomal particles for use in single-molecule fluorescence experiments, mutant ribosomes with the hairpin 5 extension in helix 33a were hybridized with the Cy3-labeled sp5 oligonucleotide. Labeled ribosomes were then assembled into initiation complexes with 5′-biotinylated or non-biotinylated mRNA and initiator tRNA, and immobilized on streptavidin-derivatized quartz surfaces, as described previously (8). Laser illumination of quartz surfaces resulted in significant biotin-dependent Cy3 fluorescence from spatially separated mutant ribosomes (Figure 5A, panel 3, on average 275 particles per 60 μm × 120 μm field of view) but only background fluorescence from wild-type ribosomes (Figure 5A, panel 1, on average six spots per field of view). This is consistent with over 98% specificity of our labeling method.

To observe dynamic distance changes using FRET, two fluorophores must be introduced on the particle of interest. Excitation of the donor fluorophore (Cy3 in this study) with laser light will result in non-radioactive transfer of energy and fluorescence emission from the acceptor fluorophore (Cy5 in this study) if the two dyes are within certain distance determined by their spectral properties (7). The activity of our mutant ribosomes was verified using single-molecule FRET. Mutant ribosomes carrying a Cy3-labeled sp5 oligonucleotide prehybridized to the hairpin 5 extension in helix 33a were immobilized on the surface as described above, and a cognate Cy5-labeled tRNA in ternary complex with EF-Tu/GTP was added using a stopped-flow delivery apparatus. Based on structural models of ribosomal complexes, we predicted a 70–80 Å distance between Cy5 on the accommodated tRNA and Cy3 on the 30S subunit. Upon excitation with 532 nm laser light, anti-correlated changes in fluorescence intensity of Cy3 and Cy5 were observed in a population of particles that were indicative of a FRET signal defined as FRET = I_Cy5/(I_Cy3 + I_Cy5), where I_Cy3 and I_Cy5 are fluorescence intensities of both dyes (Figure 5B). The average FRET value was found to be 0.12 ± 0.04. This corresponds to a distance of ∼70 Å between the two fluorophores [assuming R₀ = 50 Å and random dye orientation as determined previously (8)]. The observed transitions to zero FRET result from photophysical event known as blinking, which involves temporary loss of Cy5 fluorescence.

Fluorescent labeling of surface-immobilized ribosomes had no effect on their function. The efficiency of tRNA binding was 50% as visualized by FRET in 143 out of 287 ribosome complexes observed in four separate experiments. The rate of tRNA delivery was monitored as the increase in the number of ribosomes with FRET ≥ 0.10 over time (Figure 5C). The labeled mutant ribosomes bind 10 nM ternary complex with a rate constant of 9 × 10⁷ M⁻¹ s⁻¹. This value is consistent with the tRNA binding rate reported previously for wild-type ribosomes (8), and provides further evidence that our labeling method does not interfere with ribosome activity. These preliminary results demonstrate the applicability of our oligonucleotide hybridization approach to label ribosomal particles for single-molecule FRET studies of translation.