Exhaustive Sampling of Ribosome Binding Sites for Fast Translation

It is widely accepted that mRNA primary and secondary structure present near the start codon of a gene can influence protein expression. Although various methods have been used to study the effect of different sequences and structures on expression, the lack of exhaustive sampling by traditional high-throughput screening or in vivo directed evolution has left gaps in our understanding. To uncover new mRNA determinants of rapid translation, we constructed and exhaustively sampled an mRNA library containing all possible 18-base sequences situated directly before the AUG start codon of a well-expressed, robustly-folding model protein. We exploited the in vitro technique ribosome display and a reconstituted E. coli translation system to examine translation efficiency. Our unbiased mRNA library contained ~70 billion unique members, which could be exhaustively sampled using this powerful in vitro approach. We devised a selection strategy based on translation time and protein affinity separation to recover mRNA sequences that promoted fast translation of the full protein, presumably through fast ribosome binding, initiation complex assembly, and progression to elongation. Several selection rounds of increasing stringency (i.e., shorter translation times) were performed to enrich the most efficient sequences.

As expected, a large number of Shine-Dalgarno sequences were selected; however, we also observed a significant population of cytosine-rich sequences, which are not abundant in E. coli but have been implicated as translation initiation sites in distantly-related organisms. We performed 454 sequencing to more fully characterize the enriched library, particularly the cytosine-rich sequences, which often contained significant secondary structure. Intriguingly, a k-motif search (k = 4-7-base) revealed that highly-overrepresented non-Shine-Dalgarno motifs are often complementary to solvent-exposed portions of the 16S rRNA in the 30S ribosomal subunit near the mRNA tunnel. Some of these overrepresented motifs bear notable resemblance to known eukaryotic initiation signals, thus providing evidence for a conserved mechanism of translation initiation. Transient base-pairing between the 16S rRNA and the mRNA near the start codon seems to expedite initiation even in non-Shine-Dalgarno sequences. Additionally, a motif-correlation analysis revealed that certain short (~4-6-base) sequences may enhance the function of the Shine-Dalgarno sequence in expediting fast translation. We propose that these Shine-Dalgarno enhancers may work by base-pairing to the 16S rRNA, much like the cytosine-rich sequences. Furthermore, we have extensively characterized the preferred position of Shine-Dalgarno motifs in the context of our model mRNA.

The present investigation has used ribosome display as a discovery tool to generate new insights into the ribosome binding site. By exhaustively sampling a region of ~18 bases just prior to the start codon of a model gene and by utilizing large-scale sequencing, we were able to generate much richer data sets than in previous studies. Furthermore, in using a minimal, E. coli-based in vitro translation system, we have been able to identify the intrinsic molecular underpinnings of mRNA-ribosome interactions in the most insulated way possible. The presence of Shine-Dalgarno sequences validates our method, while the unexpected presence of many non-Shine-Dalgarno sequences provides evidence that the nature of the ribosome, a broad-specificity ribozyme, allows for many solutions to the fast-translation problem, including ones not currently known to be utilized in vivo. Insights derived from this work may aid in the development of algorithms to more deeply mine genomes for possible coding sequences. Finally, for the purpose of creating fast-translating constructs for biotechnological applications, the general design rules elucidated by our combinatorial method may be easily applied to optimize any mRNA of interest.