The anti-Shine-Dalgarno sequence drives translational pausing and codon choice in bacteria

Abstract

Protein synthesis by ribosomes takes place on a linear substrate but at non-uniform speeds. Transient pausing of ribosomes can affect a variety of co-translational processes, including protein targeting and folding. These pauses are influenced by the sequence of the messenger RNA. Thus, redundancy in the genetic code allows the same protein to be translated at different rates. However, our knowledge of both the position and the mechanism of translational pausing in vivo is highly limited. Here we present a genome-wide analysis of translational pausing in bacteria by ribosome profiling--deep sequencing of ribosome-protected mRNA fragments. This approach enables the high-resolution measurement of ribosome density profiles along most transcripts at unperturbed, endogenous expression levels. Unexpectedly, we found that codons decoded by rare transfer RNAs do not lead to slow translation under nutrient-rich conditions. Instead, Shine-Dalgarno-(SD)-like features within coding sequences cause pervasive translational pausing. Using an orthogonal ribosome possessing an altered anti-SD sequence, we show that pausing is due to hybridization between the mRNA and 16S ribosomal RNA of the translating ribosome. In protein-coding sequences, internal SD sequences are disfavoured, which leads to biased usage, avoiding codons and codon pairs that resemble canonical SD sites. Our results indicate that internal SD-like sequences are a major determinant of translation rates and a global driving force for the coding of bacterial genomes.

Fig. 1

Analysis of translational pausing using ribosome profiling in bacteria. a, Validation of the ribosome stalling site in the secM mRNA. b and c, Average ribosome occupancy of each codon relative to their respective tRNA abundance measured by Dong et al. For growth in Luria broth (b), elevated occupancy at serine codons (blue) likely reflects preferential depletion of this amino acid. In glucose-rich medium (c), the ribosome occupancy is independent of tRNA abundance. d, Plot of cross-correlation function between ribosome occupancy profiles and the presence of the indicated tri-nucleotide sequences.

Fig. 2

Relationship between ribosome pausing and internal Shine-Dalgarno sequences. a, Plot of correlation between ribosome occupancy and SD-like features for E. coli, B. subtilis, and Saccharomyces cerevisiae. b, Plot of the average ribosome occupancy of hexanucleotide sequences relative to their affinity to the anti-Shine-Dalgarno sequence. c, Re-programmed pausing by recoding the ompF mRNA. Ribosome occupancy (red) increases when the A-site is 8-11 downstream (arrow) from SD-like features (green). Synonymous mutations replacing the SD-like sequence (GGUGGUG) in wildtype ompF (top) with a sequence (GGCGGCG) with lower hybridization to the SD site (bottom) caused a corresponding decrease in ribosome pausing.

Fig. 3

Pausing of elongating ribosomes due to SD-aSD interaction. a, Two models could account for the excess ribosome density at internal SD-like sequences. b, Ribosome occupancy of lacZ mRNA translated by wildtype ribosome. Like other genes translated by the wildtype ribosome, the ribosome occupancy pattern along lacZ is correlated with the presence of SD-like sequences (left), and not with the orthogonal SD sequence (O-SD, right). c, Ribosome occupancy of lacZ mRNA translated by orthogonal ribosome (O-ribosome). Unlike other genes in the same cells, the specialized O-SD-lacZ has ribosome pausing at internal O-SD-like sequences (right), and not at SD-like sites (left).

Fig. 4

Selection against SD-like sequences and the constraint on protein coding. a, Rate of occurrence of hexanucleotide sequences in E. coli messages relative to their affinity to the anti-SD site. The orange line shows the average occurrence within a bin size of 0.5 kcal/mol. b, Occurrence of codon-pairs for glycine-glycine residues relative to their affinity to the anti-SD site. The color coding represents the enrichment in occurrence of codon pairs after correcting for the usage of single codons. c, Cross-correlation function of ribosome occupancy profiles between conserved genes in E. coli and B. subtilis. Zero offset means the two sequences are aligned at each amino acid residue.