Specific length and structure rather than high thermodynamic stability enable regulatory mRNA stem-loops to pause translation

Abstract

Translating ribosomes unwind mRNA secondary structures by three basepairs each elongation cycle. Despite the ribosome helicase, certain mRNA stem-loops stimulate programmed ribosomal frameshift by inhibiting translation elongation. Here, using mutagenesis, biochemical and single-molecule experiments, we examine whether high stability of three basepairs, which are unwound by the translating ribosome, is critical for inducing ribosome pauses. We find that encountering frameshift-inducing mRNA stem-loops from the E. coli dnaX mRNA and the gag-pol transcript of Human Immunodeficiency Virus (HIV) hinders A-site tRNA binding and slows down ribosome translocation by 15-20 folds. By contrast, unwinding of first three basepairs adjacent to the mRNA entry channel slows down the translating ribosome by only 2-3 folds. Rather than high thermodynamic stability, specific length and structure enable regulatory mRNA stem-loops to stall translation by forming inhibitory interactions with the ribosome. Our data provide the basis for rationalizing transcriptome-wide studies of translation and searching for novel regulatory mRNA stem-loops.

Fig. 1. Model mRNAs containing either an FSS stem-loop or a bimolecular helix.

a Sequences of model mRNAs containing the Shine-Dalgarno (SD) sequences (green) and the FSS. The dnaX_Slip mRNA contains native slippery sequence (magenta). In other mRNA constructs, the slippery sequence is replaced with non-slippery codons. The first nucleotide of P-site codon and the last nucleotide bound in mRNA tunnel are labeled +1 and +11, respectively. b–d Secondary structures of frameshift-stimulating stem-loops (b, c) or bimolecular RNA helices (d). Dashed boxes indicate replacements of native GC basepairs with less stable A–U or G–U pairs. ΔG° values represent the overall thermodynamic stability of corresponding structures. ΔΔG° values of FSS variants indicate the changes in stabilities relative to the corresponding WT hairpins. Secondary structures and ΔG°/ΔΔG° values (b–d) were determined by the RNAstructure software using the default settings. Uncertainties in the folding free energy changes are determined using the variance and covariance of the nearest-neighbor parameter terms, determined by propagating the experiment uncertainties.

Fig. 2. Experimental design.

a The effects of FSS variants on A-site tRNA binding were determined by filter-binding assays. 70S bacterial ribosome (gray) bound with P-site peptidyl-tRNA and mRNA was incubated with aminoacyl-tRNA, which had been charged with tritium-labeled amino acid and complexed with EF-Tu•GTP. b The effects of FSS variants on ribosome translocation coupled to intersubunit rotation were studied by smFRET. Cy5 (red hexagram) and cy3 (blue hexagram) fluorophores were attached to the 30S protein S6 and 50S protein L9, respectively. The pre-translocation S6-cy5/L9-cy3-labeled ribosome contained deacylated and peptidyl-tRNAs that adoped P/E and A/P hybrid states in the R conformation of the ribosome. EF-G•GTP was delivered to the ribosome by injection. The amino acids bonded with cognate tRNAs are shown in purple and orange. A, P and E sites of the ribosome are indicated by orange, yellow, and blue, respectively.

Fig. 3. The stem length and the presence of the loop are more important for the FSS-induced inhibition of tRNA binding than FSS local and overall thermodynamic stabilities.

Kinetics of EF-Tu-catalyzed [³H]Tyr-tRNA^Tyr binding to the A site of the 70S ribosome. The ribosomes contained N-Ac-Phe-tRNA^Phe in the P site and were programmed with HIV (a) or dnaX (b) mRNAs, which contained an FSS variant (Fig. 1) 11 nucleotides downstream of the P-site codon as indicated. The binding of radio-labeled [³H]Tyr-tRNA^Tyr to ribosomes programmed with FSS-containing mRNA is shown relative to that observed in ribosomes programmed with corresponding ΔFSS mRNA. Apparent pseudo-first-order tRNA binding rates were deduced from single-exponential fitting as shown by the line graphs. Data are presented as mean values ± standard deviations of three independent measurements. Source data are provided as a Source Data file.

Fig. 4. Extension of the stem length eliminates the FSS-induced inhibition of tRNA binding.

The ribosomes, which were programmed with HIV stem-extended FSS mRNA and contained P-site N-Ac-Met-tRNA^Met, were incubated with EF-G GTP, EF-Tu GTP, Phe-tRNA^Phe, and Tyr-tRNA^Tyr for 5 min before loading ribosomes onto a nitrocellulose filter and washing away unbound aa-tRNA. The experiment was repeated three times with one of the three aminoacyl-tRNAs radio labeled, i.e. using N-Ac-[³H]Met-tRNA^Met (blue bar), [³H]Phe-tRNA^Phe (orange bar), or [³H]Tyr-tRNA^Tyr (red bar). Similar experiments were performed with ribosomes programmed with HIV ΔFSS mRNA. The binding of each radio-labeled tRNA to ribosomes programmed with the HIV stem-extended FSS mRNA is shown relative to that observed in ribosomes programmed with the HIV ΔFSS mRNA. Data are presented as mean values ± standard deviations of three independent measurements. Source data are provided as a Source Data file.

Fig. 5. The FSS-induced inhibition of translocation is influenced by length of the mRNA spacer between P-site codon and FSS.

The pre-translocation S6-cy5/L9-cy3-labeled ribosomes were programmed with either HIV WT FSS (a, b, e, f, g, h) or HIV ∆FSS (c, d) mRNA (Fig. 1). The HIV FSS was positioned either 11 (a, b), 12 (e, f) or 13 (g, h) nucleotides downstream of the P-site codon, respectively. a, c, e, g Representative FRET traces show cy3 donor fluorescence (blue), cy5 acceptor fluorescence (red), and FRET efficiency (orange) fitted by two-state HHM (brown). Arrows show the injection of EF-G•GTP. The EF-G-catalyzed translocation corresponds to the transition from R (0.4 FRET) to a stable (i.e. lasting over 4 s) NR (0.6 FRET) state of the ribosome. b, d, f, h Histograms (2 s binning size) show the distributions and median values of the dwell time between the injection and the translocation, τ_trl. N indicates the number of traces assembled into each histogram. To extend lifetime of the acceptor fluorophore in the experiments with 11 nucleotide spacer mRNA, the excitation laser was turned off after the EF-G injection and switched back on either 40 s (a) or 60 s later. τ_trl distribution in (b) was combined from data sets acquired with 40 s (Supplementary Fig. 4a) and 60 s (Supplementary Fig. 4b) laser shut-off times. Source data are provided as a Source Data file.

Fig. 6. The stem length and the presence of the loop are more important for the FSS-induced inhibition of translocation than FSS local and overall thermodynamic stabilities.

The S6-cy5/L9-cy3 ribosomes were programmed with mRNA containing either an HIV (a) or dnaX (b) FSS variant as indicated (Fig. 1). Bar graphs show median values of τ_trl. Respective mean values of τ_trl together with standard error of the mean (SEM) and standard deviations (SD) are shown in Supplementary Fig. 5. For all FSS variants but the HIV stem-extended FSS, τ_trl represents the dwell time between injection of EF-G•GTP and ribosome translocation in smFRET experiments. For the HIV stem-extended FSS, τ_trl corresponds to the dwell time between tRNA binding and translocation (Fig. 7). Source data are provided as a Source Data file.

Fig. 7. The FSS-induced inhibition of translocation is alleviated by extension of the stem length.

The S6-cy5/L9-cy3-labeled ribosomes containing P-site N-Ac-Met-Phe-tRNA^Phe were programmed with either HIV stem-extended FSS (a–c) or HIV ∆FSS (d–f) mRNA (Fig. 1). The FSS was positioned 11 nucleotides downstream of the P-site codon. a, d Representative FRET traces show cy3 donor fluorescence (blue), cy5 acceptor fluorescence (red) and FRET efficiency (orange) fitted by two-state HHM (brown). Arrows mark the injection of EF-Tu•GTP•Tyr-tRNA^Tyr and EF-G•GTP. τ_bd is the dwell time between the injection and the Tyr-tRNA^Tyr binding to the ribosomal A site, which corresponds to the transition from NR (0.6 FRET) to R (0.4 FRET) state of the ribosome. τ_trl is the dwell time between the A-site tRNA binding and the EF-G-catalyzed ribosome translocation, which corresponds to the transition from R to the stable (i.e. lasting over 4 s) NR state of the ribosome. b, c, e, f Histograms (2 s binning size) show the distributions and median values of τ_bd and τ_trl. N indicates the number of traces assembled into each histogram. The full-length views of the FRET traces are shown in Supplementary Fig. 8. Source data are provided as a Source Data file.

Fig. 8. The bimolecular helix formed in the m291 mRNA at the entry to mRNA tunnel mildly inhibits translocation.

Kinetics of translocation was measured by smFRET experiments with pre-translocation S6-cy5/L9-cy3 ribosomes programmed with m291 mRNA containing a 16-bp bimolecular RNA helix (Fig. 1) positioned 11 nucleotides downstream of P-site codon (a) or with single-stranded m291 mRNA (b). Histograms (2 s binning size) show the distributions and median values of τ_trl, which represents the dwell time between injection of EF-G GTP and ribosome translocation. N indicates the number of FRET traces compiled into each histogram. Source data are provided as a Source Data file.

Fig. 9. The presence of the loop is more important for the FSS-stimulated −1 PRF than the local and overall thermodynamic stabilities of the FSS.

Ribosomes containing P-site N-Ac-Val-tRNA^Val were programmed with dnaX_Slip mRNA containing an FSS variant (Fig. 1). The ribosomes were incubated with EF-G•GTP, EF-Tu•GTP, Lys-tRNA^Lys and [³H]Glu-tRNA^Glu (decodes GAG codon in −1 frame) for 5 min. For each FSS variant, [³H]Glu-tRNA^Glu binding to the ribosomal A site was measured by filter-binding assays. Frameshifting efficiency is represented by A-site occupancy by [³H]Glu-tRNA^Glu normalized by the P-site occupancy of N-Ac-[³H]Glu-tRNA^Glu, which non-enzymatically binds to the ribosome programmed with dnaX_Slip ΔFSS mRNA. Data are presented as mean values ± standard deviations of three independent measurements. Source data are provided as a Source Data file.