Depletion of Shine-Dalgarno Sequences Within Bacterial Coding Regions Is Expression Dependent

Abstract

Efficient and accurate protein synthesis is crucial for organismal survival in competitive environments. Translation efficiency (the number of proteins translated from a single mRNA in a given time period) is the combined result of differential translation initiation, elongation, and termination rates. Previous research identified the Shine-Dalgarno (SD) sequence as a modulator of translation initiation in bacterial genes, while codon usage biases are frequently implicated as a primary determinant of elongation rate variation. Recent studies have suggested that SD sequences within coding sequences may negatively affect translation elongation speed, but this claim remains controversial. Here, we present a metric to quantify the prevalence of SD sequences in coding regions. We analyze hundreds of bacterial genomes and find that the coding sequences of highly expressed genes systematically contain fewer SD sequences than expected, yielding a robust correlation between the normalized occurrence of SD sites and protein abundances across a range of bacterial taxa. We further show that depletion of SD sequences within ribosomal protein genes is correlated with organismal growth rates, supporting the hypothesis of strong selection against the presence of these sequences in coding regions and suggesting their association with translation efficiency in bacteria.

Keywords: gene expression; growth regulation; translation initiation.

Figure 1

The possible dual impacts of Shine-Dalgarno (SD) sequences on protein synthesis. (A) SD sequences in the 5′ untranslated region (UTR) of mRNA (messenger RNA) are known to facilitate translation initiation in bacteria via binding to the anti-SD sequence on the 3′ tail of the 16S ribosomal RNA. (B) Recent research suggests that SD sequences within coding sequences may regulate the rate of translation elongation.

Figure 2

Quantifying aSD sequence binding within coding regions. (A) We estimate the free energy of binding for each hexamer within a gene to the core aSD (anti-Shine-Dalgarno) sequence (5′-CCUCCU-3′). (B) Free energy (top) and affinity (bottom) profiles for a typical E. coli gene (b3055). The affinity profile amplifies the contribution from strongly binding regions within the gene. nt, nucleotides.

Figure 3

Depletion of SD occurrence in genomes compared to expectation from 1000 randomly generated genomes using our codon-shuffled null model. (A) the canonical SD (Shine-Dalgarno) sequence 5′-AGGAGG-3′ is depleted within coding sequences in most genomes (175 of 187). (B) The genome aSD (anti-SD) binding score $S_{\text{genome}}$ is lower for most organisms (172 of 187). Both distributions are centered significantly to the left of 0, showing that the majority of organisms avoid SD sequences according to both metrics.

Figure 4

aSD binding scores negatively correlate with gene expression in E. coli. (A) An example dataset showing negative correlation between protein abundance and aSD (anti- Shine-Dalgarno) binding scores for individual E. coli genes ($R_{\mathrm{adj}}^2=0.175,$ $p<{10}^{-18}$). Specifically, coding sequences containing fewer SD sequence motifs have higher protein abundances. (B) Multivariate regression shows that expression changes cannot be fully explained by codon usage bias, and that additional predictive power is offered by $S_{\mathrm{gene}}.$ We chose five datasets that provide independent measurements of mRNA (messenger RNA), protein, and translation efficiency levels in order to test the robustness of our findings (Lu et al. 2007; Taniguchi et al. 2010; Shiroguchi et al. 2012; Li et al. 2014).

Figure 5

Shine-Dalgarno (SD) sequence depletion is correlated with protein abundances in a diverse set of bacterial taxa. Distribution of differences between the $R_{\mathrm{adj}}^2$ for models which do and do not contain the S score. For 23 of the 26 organisms, inclusion of aSD (anti-SD) binding score as an independent variable enhances predictive power. The full data table, including organism names and values, is available in Table S2.

Figure 6

Depletion of SD (Shine-Dalgarno) sequences within ribosomal protein coding genes is widespread throughout the bacterial kingdom and associated with organismal growth. (A) Distribution of aSD (anti-SD) binding scores of ribosomal protein coding sequences in E. coli, compared to that of all other protein coding sequences. We characterize SD sequence usage bias in a genome with Equation (3). (B) Distribution of genome SD bias index for 187 bacteria genomes. Ribosomal proteins have significantly lower aSD binding scores, as compared to the rest of the genome, in the majority of bacterial species. (C) SD bias is correlated with minimum generation time in 187 organisms (Spearman-rank: $\mathrm{\rho }=0.530,$ $p<{10}^{-14}$). Depletion of internal-SD sequences in ribosomal protein genes is associated with faster growth. The full data table for this analysis, including organism names, growth rate, and B values, is provided as (Table S3).