Translational initiation is a critical step in the regulation of gene expression that impacts which proteins are synthesized and to what extent. Unlike eukaryotic ribosomes, which scan from the 5’-end of messages and generally initiate at the first start codon, bacterial ribosomes can initiate at any position along an mRNA; this is a critical requirement because many bacterial mRNAs are polycistronic. Bacterial ribosomes must select the correct start codons amidst a vast excess of potential sites (AUG, GUG, and to some extent UUG) that have to be ignored. Not only does initiation determine where translation occurs (and therefore which proteins are made), in most cases the rate of initiation determines the level of protein output. In bacteria, a common strategy for regulating translation is to block ribosome recruitment to an mRNA through the action of small RNAs (Altuvia et al., 1998; Majdalani et al., 1998; Storz et al., 2004), small-molecule binding riboswitches (Winkler et al., 2002; Mandal and Breaker, 2004), and regulatory proteins (Moine et al., 1990; Babitzke et al., 2009).

Initiation rates vary in response to several mRNA features that determine how effectively an mRNA recruits 30S subunits to the start codon. Thermodynamically stable secondary structures surrounding the initiation site prevent 30S recruitment (Hall et al., 1982; de Smit and van Duin, 1990). The kinetics of RNA folding and unfolding are also critical (de Smit and van Duin, 2003; Espah Borujeni and Salis, 2016): some structures exist in an unfolded state for such a short period of time that 30S subunits cannot find the start codon quickly enough by diffusion alone. In several well-characterized examples, regions of single-stranded RNA known as standby-sites are found nearby, positioning 30S subunits in close proximity so that they can efficiently capture the start codon upon unfolding of the mRNA secondary structure (de Smit and van Duin, 2003; Espah Borujeni et al., 2014). Interactions of 30S subunits and single-stranded mRNA regions (especially those that are AU-rich) can be mediated through ribosome protein S1 (Boni et al., 1991; Komarova et al., 2005). Bound on the back of the 30S subunit, the S1 protein contains multiple RNA-binding domains that can recruit mRNA and melt secondary structures (Qu et al., 2012), facilitating hybridization of 16S rRNA with complementary mRNA sequences colloquially known as Shine-Dalgarno motifs.

Shine-Dalgarno motifs have the consensus sequence GGAGG and can base pair with as many as nine nt in the 3’ terminal sequence of 16S rRNA (ACCUCCUUA in E. coli) referred to as the anti-Shine Dalgarno or ASD (Shine and Dalgarno, 1974). Pairing of the SD-ASD sequences can recruit 30S subunits to the start codon 5–10 nt downstream (Steitz and Jakes, 1975). SD motifs that differ significantly from the consensus or that are positioned too close or too far from the start codon yield lower levels of initiation. Indeed, many experiments using reporter genes showed that raising the SD-ASD affinity increases protein output, demonstrating its importance for determining translation levels (Hui and de Boer, 1987; Jacob et al., 1987; de Smit and van Duin, 1990; Salis et al., 2009). In addition, the SD model serves as the foundation of practical bioengineering efforts ranging from optimizing expression of recombinant proteins to expansion of the genetic code (Rackham and Chin, 2005; Salis et al., 2009).

On the other hand, even though the ASD in 16S rRNA is almost universally conserved throughout the bacterial kingdom (Nakagawa et al., 2010), the percentage of genes with SD motifs varies widely between species. While well-characterized model species such as E. coli and B. subtilis have a high percentage of genes with SD motifs (54% and 78% respectively), there is little to no enrichment of SD motifs upstream of start codons in Bacteriodetes and Cyanobacteria (Nakagawa et al., 2010). In addition, although the majority of species in the phyla Firmicutes, Actinobacteria, and Proteobacteria have high percentages of SD-containing genes, several species have low percentages, arguing that the loss of this mechanism has occurred multiple times during evolution (Nakagawa et al., 2010; Hockenberry et al., 2017). These variations across the bacterial kingdom, despite the high conservation of the ASD element on the ribosome, raise questions as to how important the SD mechanism is for ribosome recruitment.

Ribosome profiling is a method for deep sequencing of ribosome-protected mRNA fragments that allows us to define the position and number of ribosomes bound across the transcriptome at nucleotide resolution (Ingolia et al., 2009). This information allows us to calculate the ribosome density on each mRNA as a proxy for the efficiency of translation initiation. In pioneering ribosome profiling studies in bacteria, the paradoxical observation was made that there is little or no correlation between the ribosome occupancy of a gene and the strength of its SD motif (calculated using thermodynamic algorithms for RNA pairing), as had been anticipated based on the SD model (Li et al., 2014; Schrader et al., 2014; Li, 2015; Del Campo et al., 2015). This surprising observation suggested that other mRNA features could effectively mask the effects of the SD correlation at the genome-wide level.

To isolate the effects of SD motifs on the global translational landscape, we expressed 16S rRNA mutants with altered (non-functional) ASD sequences, purified mutant ribosomes, and used ribosome profiling to ask how efficiently they translate each mRNA in the cell. Unlike previous studies that vary the SD motif and other mRNA-specific features, this approach allows us to specifically eliminate the SD-ASD interaction while keeping mRNA sequences and structures intact, so that we can specifically ask questions about the role the SD-ASD interaction plays in determining mRNA translation rates. Through this analysis, we observe for the first time the effects of SD motifs at the global level, revealing a linear correlation between SD strength and ribosome occupancy. We then combined our new profiling approach with retapamulin treatment to trap ribosomes at start codons (Meydan et al., 2019; Weaver et al., 2019) in order to study the role of SD motifs in selecting start codons. To our surprise, the ASD-mutant ribosomes selectively recognize the correct initiation sites as well as wild-type ribosomes, arguing that these sites are hard-wired for initiation independent of their SD-ASD pairing strength. We show that A-rich sequences recently identified by Fredrick and co-workers (Baez et al., 2019) are enriched at annotated start sites compared to other AUG codons in the transcriptome where initiation does not take place; these A-rich sequences are also found upstream of start codons in a wide variety of species across the bacterial kingdom. In addition, mRNA structure at annotated start sites is lower than at other AUG codons, facilitating 30S binding. Together, these studies refine our understanding of the role of SD motifs and other mRNA features in defining the proteomes of bacteria.

In this study, we performed ribosome profiling on mutant ribosomes purified using an RNA tag, the MS2 aptamer, a strategy we call MS2RP (Figure 1). Originally developed for in vitro studies of ribosomes containing lethal rRNA mutations (Youngman et al., 2004; Youngman and Green, 2005), MS2-tagged ribosomes also have potential to yield insights into the function of key rRNA sequences in vivo. In addition to the studies of the ASD sequence in 16S rRNA reported here, MS2RP could be employed to characterize the functions of rRNA domains on initiation, elongation, termination, and recycling at a genome-wide level in vivo. Because MS2RP can be performed on rRNA mutants expressed from plasmids, the method can be easily transferred to other bacteria or to eukaryotes without altering rDNA in the genome. Of particular interest are rRNA variants in bacterial genomes that are expressed differentially in response to changes in the environment and are proposed to have different specificities or functions (Kurylo et al., 2018; Song et al., 2019). Variant rRNA alleles have also been reported for eukaryotic cells (Parks et al., 2018); for example, different small subunit rRNA alleles are expressed in various developmental stages in Plasmodium (Gunderson et al., 1987). In addition, the functions of the highly variable rRNA expansion segments in eukaryotes are poorly understood (Spahn et al., 2001; Anger et al., 2013). MS2RP could be a powerful tool to elucidate the activities of various subpopulations of variant or mutated rRNAs.

Previous genome-wide studies in bacteria have shown little or no correlation between SD strength and ribosome occupancy (Li et al., 2014; Schrader et al., 2014; Del Campo et al., 2015). Using MS2RP, we are able for the first time to reveal the role of SD motifs in promoting initiation across the transcriptome. In our approach, we mutated the ASD on the ribosomes, thus maintaining mRNA sequence and structure, thus allowing us to isolate the effects of the SD:ASD interaction on translation. In the absence of SD:ASD pairing, we observed a strong negative correlation between ribosome occupancy and the SD strength (calculated by pairing with the wild-type ASD sequence). In other words, the mutant ribosomes translate genes with strong SD motifs worse than those with weak SD motifs (Figure 2B). There are two possible explanations for this negative correlation. It may be that the binding of wild-type ribosomes to mRNAs with strong SD motifs occludes their ribosome-binding sites, preventing mutant ribosomes from initiating and efficiently translating these genes. Alternatively, mRNA structure and other features may outweigh the impact of SD motifs, masking their effects, explaining why conventional ribosome profiling studies failed to observe correlations between SD strength and ribosome occupancy. Regardless of which of these explanations is correct, the MS2RP strategy allows us to subtract the cumulative contribution to ribosome occupancy of all of such other mRNA features, and thus to focus exclusively on the contribution to ribosome occupancy of the SD:ASD interaction genome-wide. In this analysis, we are now able to see a linear correlation between the SD strength of an mRNA and protein output (Figure 2C).

Given that the SD motif functions through a well-defined mechanism and is widely conserved throughout bacteria, it has been thought to provide an important mechanism for start codon selection and translational output. Consistent with such a view, SD motifs are underrepresented within ORFs in order to avoid spurious initiation at internal start codons (Hockenberry et al., 2018). Strikingly, however, we find that ribosomes with altered ASDs still find the correct start codons about as efficiently as wild-type ribosomes (Figure 3). Start peaks for all four ribosome types are observed at annotated start sites regardless of the affinity of the ribosome binding site for the ASD. This shows that initiation sites are hard-wired for initiation based on mRNA features separate from the potential for SD-ASD pairing. These observations also hold true at the occasional non-annotated AUG codons where some initiation occurs (Figure 4). These data are consistent with the conclusion that SD motifs are not essential for determining where translation starts on mRNAs genome-wide.

What, then, are other mechanisms that could be used for start codon selection? Local mRNA structure and RNA folding kinetics clearly must play a critical role in allowing ribosomes to find the start codon. A number of mechanistic studies have demonstrated that RNA structure around the start codon lowers translation levels (Hall et al., 1982; de Smit and van Duin, 1990; Osterman et al., 2013; Espah Borujeni et al., 2014). Studies of factors that alter the expression of simplified reporter genes (involving randomization of the 5’-UTR or coding sequences) show that lack of secondary structure surrounding the initiation site has the most significant correlation with protein output (Salis et al., 2009; Kudla et al., 2009; Goodman et al., 2013). Recent transcriptome-wide analyses of mRNA structure in E. coli confirm that annotated start sites have lower levels of mRNA structure, as seen by PARS on purified mRNA and DMS-seq in vivo (Del Campo et al., 2015; Burkhardt et al., 2017). mRNA structure is likely an important factor in start site selection: using high-resolution SHAPE-MaPseq data (Mustoe et al., 2018), we showed that annotated AUGs have lower levels of RNA structure 30 nt upstream and downstream whereas internal AUG are not surrounded by regions of lower structure (Figure 5E).

Interestingly, in comparing the sequence context of AUG codons that are annotated as initiation sites with those that are not, we found that A’s are enriched both upstream and downstream of annotated initiation sites (Figure 5A) and we confirmed their importance in reporter assays (Figure 5B–D). These results from endogenous initiation sites are reminiscent of observations of the over-representation of A’s in 5’-UTR sequences selected for strong affinity to the ribosome in vitro (Gao et al., 2016) and in 5’-UTRs selected from random sequences upstream of a reporter gene for high levels of translation in vivo (Evfratov et al., 2017). Comparison of annotated start sites and non-annotated AUGs across several bacterial genomes shows that this mechanism is widespread (Figure 5—figure supplement 1). Although enrichment of A’s is more subtle than enrichment of G’s in E. coli and B. subtilis, in organisms that lack SD motifs, such as Mycoplasma pneumoniae and Flavobacterium johnsoniae, A-rich motifs may play an important role in initiation. Indeed, in a recent study, Fredrick and co-workers used ribosome profiling in F. johnsoniae and observed enrichment of A’s upstream of start codons in mRNAs with high ribosome occupancy in comparison to genes with are translated less efficiently (Baez et al., 2019). We envision that this sequence, like the Shine-Dalgarno motif, acts as a translational enhancer, fine-tuning the efficiency of initiation.

The mechanism by which A-rich sequences enhance initiation is not clear. The prevalence of A’s may alter the mRNA dynamics; A-rich sequences tend to have less secondary structure than GC-rich sequences. We note however that replacing A’s with U’s in several reporters reduced translation levels even though the U’s are similarly not expected to yield strong structures. A second possibility is that ribosomal components may interact specifically with A’s close to the start codon that are bound inside the ribosome during initiation. Fredrick and co-workers used reporter assays to show that mutation of a particular A at the −3 position reduces expression; this result is intriguing because the classic Kozak sequence (GCC(A/G)CCAUG) that promotes high levels of translation in eukaryotes also contains a purine at position −3. A-rich sequences have been reported to enhance translation in a variety of eukaryotic contexts includingDrosophilaand wheat germ and reticulocyte lysates (Ranjan and Hasnain, 1995; Sano et al., 2002; Suzuki et al., 2006; Pfeiffer et al., 2012). It may be that A-rich sequences interact with conserved elements of the ribosome across the domains of life. A’s further from the start codon (10–20 nt upstream) may interact with bacteria-specific ribosomal protein S1. bS1 preferably binds to A/U-rich sequence elements upstream of SD sequences (Boni et al., 1991; Komarova et al., 2005) and is thought to unwind mRNA structure to induce initiation (Qu et al., 2012; Duval et al., 2013).

Our findings have broad implications for the evolution of translational mechanisms in bacteria. Not all bacteria utilize SD motifs to promote translational initiation—SD motifs are notably lacking in Bacteroidetes and Cyanobacteria. Because the prevalence of SD motifs is a feature of the genome in general and not of a single gene, it makes sense that evolutionary selective pressure for or against SD usage would act at the level of the transcriptome. The nature of these selective pressures remains unclear, although Hockenberry recently argued that bacteria with high levels of SD usage tend to have higher maximal growth rates (Hockenberry et al., 2017). Future studies will clarify the evolutionary relationship between the growth environment, levels of SD usage among bacterial species, and their transcriptome-wide effects.

Sequencing data have been deposited in the GEO under accession code GSE135906.

1. 1. Saito K
   2. Green R
   3. Buskirk AR

   (2019) NCBI Gene Expression Omnibus

   ID GSE135906. Translational initiation in E. coli occurs at the correct sites genome-wide in the absence of mRNA-rRNA base-pairing.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE135906

1. 1. Aleksashin NA
   2. Leppik M
   3. Hockenberry AJ
   4. Klepacki D
   5. Vázquez-Laslop N
   6. Jewett MC
   7. Remme J
   8. Mankin AS

   (2019) Assembly and functionality of the ribosome with tethered subunits

   Nature Communications 10:1–13.

   https://doi.org/10.1038/s41467-019-08892-w

   • PubMed
   • Google Scholar
2. 1. Altuvia S
   2. Zhang A
   3. Argaman L
   4. Tiwari A
   5. Storz G

   (1998) The Escherichia coli OxyS regulatory RNA represses fhlA translation by blocking ribosome binding

   The EMBO Journal 17:6069–6075.

   https://doi.org/10.1093/emboj/17.20.6069

   • PubMed
   • Google Scholar
3. 1. Andreeva I
   2. Belardinelli R
   3. Rodnina MV

   (2018) Translation initiation in bacterial polysomes through ribosome loading on a standby site on a highly translated mRNA

   PNAS 115:4411–4416.

   https://doi.org/10.1073/pnas.1718029115

   • PubMed
   • Google Scholar
4. 1. Anger AM
   2. Armache J-P
   3. Berninghausen O
   4. Habeck M
   5. Subklewe M
   6. Wilson DN
   7. Beckmann R

   (2013) Structures of the human and Drosophila 80S ribosome

   Nature 497:80–85.

   https://doi.org/10.1038/nature12104

   • Google Scholar
5. 1. Babitzke P
   2. Baker CS
   3. Romeo T

   (2009) Regulation of translation initiation by RNA binding proteins

   Annual Review of Microbiology 63:27–44.

   https://doi.org/10.1146/annurev.micro.091208.073514

   • PubMed
   • Google Scholar
6. 1. Baez WD
   2. Roy B
   3. McNutt ZA
   4. Shatoff EA
   5. Chen S
   6. Bundschuh R
   7. Fredrick K

   (2019) Global analysis of protein synthesis in Flavobacterium johnsoniae reveals the use of Kozak-like sequences in diverse Bacteria

   Nucleic Acids Research 47:10477–10488.

   https://doi.org/10.1093/nar/gkz855

   • PubMed
   • Google Scholar
7. 1. Boni IV
   2. Isaeva DM
   3. Musychenko ML
   4. Tzareva NV

   (1991) Ribosome-messenger recognition: mrna target sites for ribosomal protein S1

   Nucleic Acids Research 19:155–162.

   https://doi.org/10.1093/nar/19.1.155

   • PubMed
   • Google Scholar
8. 1. Burkhardt DH
   2. Rouskin S
   3. Zhang Y
   4. Li GW
   5. Weissman JS
   6. Gross CA

   (2017) Operon mRNAs are organized into ORF-centric structures that predict translation efficiency

   eLife 6:e22037.

   https://doi.org/10.7554/eLife.22037

   • PubMed
   • Google Scholar
9. 1. Davis MP
   2. van Dongen S
   3. Abreu-Goodger C
   4. Bartonicek N
   5. Enright AJ

   (2013) Kraken: a set of tools for quality control and analysis of high-throughput sequence data

   Methods 63:41–49.

   https://doi.org/10.1016/j.ymeth.2013.06.027

   • PubMed
   • Google Scholar
10. 1. de Smit MH
    2. van Duin J

    (1990) Secondary structure of the ribosome binding site determines translational efficiency: a quantitative analysis

    PNAS 87:7668–7672.

    https://doi.org/10.1073/pnas.87.19.7668

    • PubMed
    • Google Scholar
11. 1. de Smit MH
    2. van Duin J

    (2003) Translational standby sites: how ribosomes may deal with the rapid folding kinetics of mRNA

    Journal of Molecular Biology 331:737–743.

    https://doi.org/10.1016/S0022-2836(03)00809-X

    • Google Scholar
12. 1. Del Campo C
    2. Bartholomäus A
    3. Fedyunin I
    4. Ignatova Z

    (2015) Secondary structure across the bacterial transcriptome reveals versatile roles in mRNA regulation and function

    PLOS Genetics 11:e1005613.

    https://doi.org/10.1371/journal.pgen.1005613

    • PubMed
    • Google Scholar
13. 1. Duval M
    2. Korepanov A
    3. Fuchsbauer O
    4. Fechter P
    5. Haller A
    6. Fabbretti A
    7. Choulier L
    8. Micura R
    9. Klaholz BP
    10. Romby P
    11. Springer M
    12. Marzi S

    (2013) Escherichia coli ribosomal protein S1 unfolds structured mRNAs onto the ribosome for active translation initiation

    PLOS Biology 11:e1001731.

    https://doi.org/10.1371/journal.pbio.1001731

    • PubMed
    • Google Scholar
14. 1. Espah Borujeni A
    2. Channarasappa AS
    3. Salis HM

    (2014) Translation rate is controlled by coupled trade-offs between site accessibility, selective RNA unfolding and sliding at upstream standby sites

    Nucleic Acids Research 42:2646–2659.

    https://doi.org/10.1093/nar/gkt1139

    • PubMed
    • Google Scholar
15. 1. Espah Borujeni A
    2. Salis HM

    (2016) Translation initiation is controlled by RNA folding kinetics via a ribosome drafting mechanism

    Journal of the American Chemical Society 138:7016–7023.

    https://doi.org/10.1021/jacs.6b01453

    • PubMed
    • Google Scholar
16. 1. Evfratov SA
    2. Osterman IA
    3. Komarova ES
    4. Pogorelskaya AM
    5. Rubtsova MP
    6. Zatsepin TS
    7. Semashko TA
    8. Kostryukova ES
    9. Mironov AA
    10. Burnaev E
    11. Krymova E
    12. Gelfand MS
    13. Govorun VM
    14. Bogdanov AA
    15. Sergiev PV
    16. Dontsova OA

    (2017) Application of sorting and next generation sequencing to study 5'-UTR influence on translation efficiency in Escherichia coli

    Nucleic Acids Research 45:3487–3502.

    https://doi.org/10.1093/nar/gkw1141

    • PubMed
    • Google Scholar
17. 1. Gao R
    2. Yu K
    3. Nie J
    4. Lian T
    5. Jin J
    6. Liljas A
    7. Su X-D

    (2016) Deep sequencing reveals global patterns of mRNA recruitment during translation initiation

    Scientific Reports 6:30170.

    https://doi.org/10.1038/srep30170

    • Google Scholar
18. 1. Goodman DB
    2. Church GM
    3. Kosuri S

    (2013) Causes and effects of N-terminal Codon bias in bacterial genes

    Science 342:475–479.

    https://doi.org/10.1126/science.1241934

    • PubMed
    • Google Scholar
19. 1. Gunderson JH
    2. Sogin ML
    3. Wollett G
    4. Hollingdale M
    5. de la Cruz VF
    6. Waters AP
    7. McCutchan TF

    (1987) Structurally distinct, stage-specific ribosomes occur in plasmodium

    Science 238:933–937.

    https://doi.org/10.1126/science.3672135

    • PubMed
    • Google Scholar
20. 1. Hall MN
    2. Gabay J
    3. Débarbouillé M
    4. Schwartz M

    (1982) A role for mRNA secondary structure in the control of translation initiation

    Nature 295:616–618.

    https://doi.org/10.1038/295616a0

    • Google Scholar
21. 1. Hockenberry AJ
    2. Stern AJ
    3. Amaral LAN
    4. Jewett MC

    (2017) Diversity of translation initiation mechanisms across bacterial species is driven by environmental conditions and growth demands

    Molecular Biology and Evolution 35:582–592.

    https://doi.org/10.1093/molbev/msx310

    • Google Scholar
22. 1. Hockenberry AJ
    2. Jewett MC
    3. Amaral LAN
    4. Wilke CO

    (2018) Within-Gene Shine-Dalgarno sequences are not selected for function

    Molecular Biology and Evolution 35:2487–2498.

    https://doi.org/10.1093/molbev/msy150

    • PubMed
    • Google Scholar
23. 1. Hui A
    2. de Boer HA

    (1987) Specialized ribosome system: preferential translation of a single mRNA species by a subpopulation of mutated ribosomes in Escherichia coli

    PNAS 84:4762–4766.

    https://doi.org/10.1073/pnas.84.14.4762

    • PubMed
    • Google Scholar
24. 1. Ingolia NT
    2. Ghaemmaghami S
    3. Newman JRS
    4. Weissman JS

    (2009) Genome-Wide analysis in vivo of translation with nucleotide resolution using ribosome profiling

    Science 324:218–223.

    https://doi.org/10.1126/science.1168978

    • Google Scholar
25. 1. Ingolia NT
    2. Lareau LF
    3. Weissman JS

    (2011) Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of mammalian proteomes

    Cell 147:789–802.

    https://doi.org/10.1016/j.cell.2011.10.002

    • PubMed
    • Google Scholar
26. 1. Jacob WF
    2. Santer M
    3. Dahlberg AE

    (1987) A single base change in the Shine-Dalgarno region of 16S rRNA of Escherichia coli affects translation of many proteins

    PNAS 84:4757–4761.

    https://doi.org/10.1073/pnas.84.14.4757

    • PubMed
    • Google Scholar
27. 1. Jiang H
    2. Lei R
    3. Ding SW
    4. Zhu S

    (2014) Skewer: a fast and accurate adapter trimmer for next-generation sequencing paired-end reads

    BMC Bioinformatics 15:182.

    https://doi.org/10.1186/1471-2105-15-182

    • PubMed
    • Google Scholar
28. 1. Komarova AV
    2. Tchufistova LS
    3. Dreyfus M
    4. Boni IV

    (2005) AU-rich sequences within 5' untranslated leaders enhance translation and stabilize mRNA in Escherichia coli

    Journal of Bacteriology 187:1344–1349.

    https://doi.org/10.1128/JB.187.4.1344-1349.2005

    • PubMed
    • Google Scholar
29. 1. Kudla G
    2. Murray AW
    3. Tollervey D
    4. Plotkin JB

    (2009) Coding-sequence determinants of gene expression in Escherichia coli

    Science 324:255–258.

    https://doi.org/10.1126/science.1170160

    • PubMed
    • Google Scholar
30. 1. Kurylo CM
    2. Parks MM
    3. Juette MF
    4. Zinshteyn B
    5. Altman RB
    6. Thibado JK
    7. Vincent CT
    8. Blanchard SC

    (2018) Endogenous rRNA Sequence Variation Can Regulate Stress Response Gene Expression and Phenotype

    Cell Reports 25:236–248.

    https://doi.org/10.1016/j.celrep.2018.08.093

    • Google Scholar
31. 1. Langmead B
    2. Trapnell C
    3. Pop M
    4. Salzberg SL

    (2009) Ultrafast and memory-efficient alignment of short DNA sequences to the human genome

    Genome Biology 10:R25.

    https://doi.org/10.1186/gb-2009-10-3-r25

    • Google Scholar
32. 1. Lee K
    2. Holland-Staley CA
    3. Cunningham PR

    (1996)

    Genetic analysis of the Shine-Dalgarno interaction: selection of alternative functional mRNA-rRNA combinations

    RNA 2:1270–1285.

    • PubMed
    • Google Scholar
33. 1. Lee S
    2. Liu B
    3. Lee S
    4. Huang SX
    5. Shen B
    6. Qian SB

    (2012) Global mapping of translation initiation sites in mammalian cells at single-nucleotide resolution

    PNAS 109:E2424–E2432.

    https://doi.org/10.1073/pnas.1207846109

    • PubMed
    • Google Scholar
34. 1. Li GW
    2. Burkhardt D
    3. Gross C
    4. Weissman JS

    (2014) Quantifying absolute protein synthesis rates reveals principles underlying allocation of cellular resources

    Cell 157:624–635.

    https://doi.org/10.1016/j.cell.2014.02.033

    • PubMed
    • Google Scholar
35. 1. Li GW

    (2015) How do Bacteria tune translation efficiency?

    Current Opinion in Microbiology 24:66–71.

    https://doi.org/10.1016/j.mib.2015.01.001

    • PubMed
    • Google Scholar
36. 1. Lodish HF

    (1970) Secondary structure of bacteriophage f2 ribonucleic acid and the initiation of in vitro protein biosynthesis

    Journal of Molecular Biology 50:689–702.

    https://doi.org/10.1016/0022-2836(70)90093-8

    • Google Scholar
37. 1. Majdalani N
    2. Cunning C
    3. Sledjeski D
    4. Elliott T
    5. Gottesman S

    (1998) DsrA RNA regulates translation of RpoS message by an anti-antisense mechanism, independent of its action as an antisilencer of transcription

    PNAS 95:12462–12467.

    https://doi.org/10.1073/pnas.95.21.12462

    • PubMed
    • Google Scholar
38. 1. Mandal M
    2. Breaker RR

    (2004) Gene regulation by riboswitches

    Nature Reviews Molecular Cell Biology 5:451–463.

    https://doi.org/10.1038/nrm1403

    • PubMed
    • Google Scholar
39. 1. Meydan S
    2. Marks J
    3. Klepacki D
    4. Sharma V
    5. Baranov PV
    6. Firth AE
    7. Margus T
    8. Kefi A
    9. Vázquez-Laslop N
    10. Mankin AS

    (2019) Retapamulin-Assisted ribosome profiling reveals the alternative bacterial proteome

    Molecular Cell 74:481–493.

    https://doi.org/10.1016/j.molcel.2019.02.017

    • PubMed
    • Google Scholar
40. 1. Mohammad F
    2. Woolstenhulme CJ
    3. Green R
    4. Buskirk AR

    (2016) Clarifying the translational pausing landscape in Bacteria by ribosome profiling

    Cell Reports 14:686–694.

    https://doi.org/10.1016/j.celrep.2015.12.073

    • Google Scholar
41. 1. Moine H
    2. Romby P
    3. Springer M
    4. Grunberg-Manago M
    5. Ebel JP
    6. Ehresmann B
    7. Ehresmann C

    (1990) Escherichia coli threonyl-tRNA synthetase and tRNA(Thr) modulate the binding of the ribosome to the translational initiation site of the thrS mRNA

    Journal of Molecular Biology 216:299–310.

    https://doi.org/10.1016/S0022-2836(05)80321-3

    • PubMed
    • Google Scholar
42. 1. Mustoe AM
    2. Busan S
    3. Rice GM
    4. Hajdin CE
    5. Peterson BK
    6. Ruda VM
    7. Kubica N
    8. Nutiu R
    9. Baryza JL
    10. Weeks KM

    (2018) Pervasive regulatory functions of mRNA structure revealed by High-Resolution SHAPE probing

    Cell 173:181–195.

    https://doi.org/10.1016/j.cell.2018.02.034

    • PubMed
    • Google Scholar
43. 1. Nakagawa S
    2. Niimura Y
    3. Miura K
    4. Gojobori T

    (2010) Dynamic evolution of translation initiation mechanisms in prokaryotes

    PNAS 107:6382–6387.

    https://doi.org/10.1073/pnas.1002036107

    • PubMed
    • Google Scholar
44. 1. Neumann H
    2. Wang K
    3. Davis L
    4. Garcia-Alai M
    5. Chin JW

    (2010) Encoding multiple unnatural amino acids via evolution of a quadruplet-decoding ribosome

    Nature 464:441–444.

    https://doi.org/10.1038/nature08817

    • PubMed
    • Google Scholar
45. 1. Orelle C
    2. Carlson ED
    3. Szal T
    4. Florin T
    5. Jewett MC
    6. Mankin AS

    (2015) Protein synthesis by ribosomes with tethered subunits

    Nature 524:119–124.

    https://doi.org/10.1038/nature14862

    • PubMed
    • Google Scholar
46. 1. Osterman IA
    2. Evfratov SA
    3. Sergiev PV
    4. Dontsova OA

    (2013) Comparison of mRNA features affecting translation initiation and reinitiation

    Nucleic Acids Research 41:474–486.

    https://doi.org/10.1093/nar/gks989

    • PubMed
    • Google Scholar
47. 1. Parks MM
    2. Kurylo CM
    3. Dass RA
    4. Bojmar L
    5. Lyden D
    6. Vincent CT
    7. Blanchard SC

    (2018) Variant ribosomal RNA alleles are conserved and exhibit tissue-specific expression

    Science Advances 4:eaao0665.

    https://doi.org/10.1126/sciadv.aao0665

    • PubMed
    • Google Scholar
48. 1. Pfeiffer BD
    2. Truman JW
    3. Rubin GM

    (2012) Using translational enhancers to increase transgene expression in Drosophila

    PNAS 109:6626–6631.

    https://doi.org/10.1073/pnas.1204520109

    • PubMed
    • Google Scholar
49. 1. Qu X
    2. Lancaster L
    3. Noller HF
    4. Bustamante C
    5. Tinoco I

    (2012) Ribosomal protein S1 unwinds double-stranded RNA in multiple steps

    PNAS 109:14458–14463.

    https://doi.org/10.1073/pnas.1208950109

    • PubMed
    • Google Scholar
50. 1. Rackham O
    2. Chin JW

    (2005) A network of orthogonal ribosome x mRNA pairs

    Nature Chemical Biology 1:159–166.

    https://doi.org/10.1038/nchembio719

    • PubMed
    • Google Scholar
51. 1. Ranjan A
    2. Hasnain SE

    (1995) Influence of Codon usage and translational initiation Codon context in the AcNPV-based expression system: computer analysis using homologous and heterologous genes

    Virus Genes 9:149–153.

    https://doi.org/10.1007/BF01702657

    • PubMed
    • Google Scholar
52. 1. Rex G
    2. Surin B
    3. Besse G
    4. Schneppe B
    5. McCarthy JE

    (1994)

    The mechanism of translational coupling in Escherichia coli. Higher order structure in the atpHA mRNA acts as a conformational switch regulating the access of de novo initiating ribosomes

    The Journal of Biological Chemistry 269:18118–18127.

    • PubMed
    • Google Scholar
53. Software

    1. Saito K

    (2020) 2019_SDASD, version 6afaf41

    Github.

    https://github.com/greenlabjhmi/2019_SDASD
54. 1. Salis HM
    2. Mirsky EA
    3. Voigt CA

    (2009) Automated design of synthetic ribosome binding sites to control protein expression

    Nature Biotechnology 27:946–950.

    https://doi.org/10.1038/nbt.1568

    • PubMed
    • Google Scholar
55. 1. Sano K
    2. Maeda K
    3. Oki M
    4. Maéda Y

    (2002) Enhancement of protein expression in insect cells by a lobster tropomyosin cDNA leader sequence

    FEBS Letters 532:143–146.

    https://doi.org/10.1016/S0014-5793(02)03659-1

    • PubMed
    • Google Scholar
56. 1. Schrader JM
    2. Zhou B
    3. Li GW
    4. Lasker K
    5. Childers WS
    6. Williams B
    7. Long T
    8. Crosson S
    9. McAdams HH
    10. Weissman JS
    11. Shapiro L

    (2014) The coding and noncoding architecture of the Caulobacter crescentus genome

    PLOS Genetics 10:e1004463.

    https://doi.org/10.1371/journal.pgen.1004463

    • PubMed
    • Google Scholar
57. 1. Shine J
    2. Dalgarno L

    (1974) The 3'-terminal sequence of Escherichia coli 16S ribosomal RNA: complementarity to nonsense triplets and ribosome binding sites

    PNAS 71:1342–1346.

    https://doi.org/10.1073/pnas.71.4.1342

    • PubMed
    • Google Scholar
58. 1. Song W
    2. Joo M
    3. Yeom JH
    4. Shin E
    5. Lee M
    6. Choi HK
    7. Hwang J
    8. Kim YI
    9. Seo R
    10. Lee JE
    11. Moore CJ
    12. Kim YH
    13. Eyun SI
    14. Hahn Y
    15. Bae J
    16. Lee K

    (2019) Divergent rRNAs as regulators of gene expression at the ribosome level

    Nature Microbiology 4:515–526.

    https://doi.org/10.1038/s41564-018-0341-1

    • PubMed
    • Google Scholar
59. 1. Spahn CM
    2. Beckmann R
    3. Eswar N
    4. Penczek PA
    5. Sali A
    6. Blobel G
    7. Frank J

    (2001) Structure of the 80S ribosome from Saccharomyces cerevisiae--tRNA-ribosome and subunit-subunit interactions

    Cell 107:373–386.

    https://doi.org/10.1016/S0092-8674(01)00539-6

    • PubMed
    • Google Scholar
60. 1. Steitz JA
    2. Jakes K

    (1975) How ribosomes select initiator regions in mRNA: base pair formation between the 3' terminus of 16S rRNA and the mRNA during initiation of protein synthesis in Escherichia coli

    PNAS 72:4734–4738.

    https://doi.org/10.1073/pnas.72.12.4734

    • PubMed
    • Google Scholar
61. 1. Storz G
    2. Opdyke JA
    3. Zhang A

    (2004) Controlling mRNA stability and translation with small, noncoding RNAs

    Current Opinion in Microbiology 7:140–144.

    https://doi.org/10.1016/j.mib.2004.02.015

    • PubMed
    • Google Scholar
62. 1. Suzuki T
    2. Ito M
    3. Ezure T
    4. Kobayashi S
    5. Shikata M
    6. Tanimizu K
    7. Nishimura O

    (2006) Performance of expression vector, pTD1, in insect cell-free translation system

    Journal of Bioscience and Bioengineering 102:69–71.

    https://doi.org/10.1263/jbb.102.69

    • Google Scholar
63. 1. Weaver J
    2. Mohammad F
    3. Buskirk AR
    4. Storz G

    (2019) Identifying small proteins by ribosome profiling with stalled initiation complexes

    mBio 10:e02819.

    https://doi.org/10.1128/mBio.02819-18

    • PubMed
    • Google Scholar
64. 1. Winkler W
    2. Nahvi A
    3. Breaker RR

    (2002) Thiamine derivatives bind messenger RNAs directly to regulate bacterial gene expression

    Nature 419:952–956.

    https://doi.org/10.1038/nature01145

    • PubMed
    • Google Scholar
65. 1. Woolstenhulme CJ
    2. Guydosh NR
    3. Green R
    4. Buskirk AR

    (2015) High-precision analysis of translational pausing by ribosome profiling in Bacteria lacking EFP

    Cell Reports 11:13–21.

    https://doi.org/10.1016/j.celrep.2015.03.014

    • PubMed
    • Google Scholar
66. 1. Wu X
    2. Bartel DP

    (2017) kpLogo: positional k-mer analysis reveals hidden specificity in biological sequences

    Nucleic Acids Research 45:W534–W538.

    https://doi.org/10.1093/nar/gkx323

    • PubMed
    • Google Scholar
67. 1. Youngman EM
    2. Brunelle JL
    3. Kochaniak AB
    4. Green R

    (2004) The active site of the ribosome is composed of two layers of conserved nucleotides with distinct roles in peptide bond formation and peptide release

    Cell 117:589–599.

    https://doi.org/10.1016/S0092-8674(04)00411-8

    • PubMed
    • Google Scholar
68. 1. Youngman EM
    2. Green R

    (2005) Affinity purification of in vivo-assembled ribosomes for in vitro biochemical analysis

    Methods 36:305–312.

    https://doi.org/10.1016/j.ymeth.2005.04.007

    • PubMed
    • Google Scholar

Acceptance summary:

This paper uses an innovative twist on ribosome profiling to investigate the importance of Shine-Dalgarno sequences in bacterial translation initiation. Surprisingly, the data show that strong base-pairing between the 16S ribosomal RNA and the mRNA Shine-Dalgarno sequence is neither necessary nor sufficient for translation initiation. This suggests that start-codons are "hard-wired" into the genome, largely independent of Shine-Dalgarno sequence.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Shine-Dalgarno sequences fine-tune translation genome-wide but are not the primary determinants of start-site selection" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Joe Wade as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jim Manley as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Shura Mankin (Reviewer #2).

Our decision has been reached after an extensive discussion involving the three reviewers. The reviewers were enthusiastic about parts of the manuscript, in particular the method itself; however, there was some disagreement as to the significance of the work. Much of the discussion focused on the data in Figure 6, which the reviewers considered to present potentially the most important result. We felt that further analysis is needed to fully address the key questions of (i) whether annotated start codons are inherently good at binding ribosomes, independent of the SD, and (ii) whether an SD alone (e.g. next to an ATG in the middle of an ORF) is insufficient to bind a ribosome. Put more simply, we felt that further analysis is required to show that start-codons are "hard-wired" into the genome, largely independent of SD sequence. We are therefore rejecting the paper because the outcome of the new analysis is unclear. Nonetheless, we would be willing to consider a revised version if the analyses suggested below, or something equivalent, provide stronger support for the idea that start-codons are hard-wired, independent of the SD sequence.

The main concerns with Figure 6 are that (i) the control set of ORF-internal ATGs is not the best control because many (most?) of the ATGs don't have a good SD sequence for the modified ribosomes; and (ii) ribosome density at annotated start codons for the modified ribosomes could be due to a subset of start codons that have decent matches to the modified ASD sequence. We suggest that a more appropriate control set of ATGs would be those with a good predicted match to the altered ASD sequence. We also suggest limiting the analysis in Figure 6C to start codons that have a poor match to the modified ASD. Another way to look at this would be to compare which annotated start codons are recognized by the different modified ribosomes; if all three types of ribosome recognize the same subset of start codons, it's safe to conclude that this is occurring independent of the SD. If these (or other) analyses can provide stronger support for the "hard-wired" model, that would likely be sufficient for publication. In addition to the re-analysis of data from Figure 6, it's important to improve the clarity of the paper, which was at times confusing (see the detailed reviews for more information on this). Additionally, reviewer #3 makes some important points about the calculation of hybridization energies, such as considering a full, 9 nt SD sequence with variable spacing. Lastly, the manuscript would benefit from a clearer description of what is already known about features other than SD sequence that contribute to translation initiation (see comments from reviewer 3).

Reviewer #1:

This paper describes an innovative approach to probe the importance of Shine-Dalgarno (S-D) sequences in translation initiation in Escherichia coli. By performing ribosome profiling on modified ribosomes, the authors are able to observe translation by ribosomes with altered anti-S-D sequences. This method reveals that despite no correlation between S-D strength and translation levels, there is a contribution of S-D strength that is apparent when all confounding factors have been controlled for. Interestingly, this effect of S-Ds is lost during other growth conditions, although for cold shock that is largely consistent with previous work, and it is unclear what the mechanism is in stationary phase. While I think the topic is interesting and the primary method is ingenious, I'm not convinced that the authors have learned much about the relative importance of S-Ds in translation initiation. As they acknowledge, previous studies have failed to see a correlation between S-D strength and translation initiation levels, and the importance of secondary structure and of A-rich sequences has been described previously. The fact that predictions of S-D strength correlate with translation initiation levels once factors other than S-D have been accounted for indicates that these predictions are fairly accurate. This is important, since it accounts for the possibility that the lack of correlation between predicted S-D strength and translation initiation is because of our inability to predict S-D strength. However, the impact of this advance is small. I also have concerns about the interpretation of Figures 6 and 7 that impact the overall conclusions.

- The presentation of the cold shock data is confusing. The overall conclusion is that S-D-dependence is lost at almost all genes during cold shock. However, a small subset of genes appears to depend strongly on the S-D. The distinction between the effect on the majority of genes and the effect on a small subset should be explained more clearly. The simplest interpretation of these data is that most start codons are highly structured during cold shock, but those that are do not rely on their S-Ds. This model is largely consistent with previous work.

- I disagree with the interpretation of Figure 6. The data show that for the altered ribosomes, annotated start codons are used far more efficiently than the collection of all other ATG sequences within ORFs. However, there are many more ATGs within ORFs than annotated start codons, and even if translation relies heavily on S-D sequences, you would expect that most ATGs within ORFs would not be selected by alternative ribosomes because only a small subset will have appropriate S-D sequences, and many may be weakly expressed. My interpretation of these data is that alternative ribosomes do use annotated start codons, but there is no way to tell how selectively they do this. A more appropriate comparison would be of (i) annotated start codons to (ii) ATGs within ORFs where the ATG is associated with a sequence that is predicted to function as a good S-D for the alternative ribosome.

- Another concern I have with Figure 6 is that presumably some, and perhaps many of the annotated start codons will have good SD matches for the alternative ribosomes. Figure 2C suggests that the number with good matches will be fairly high. Is the ribosome density at annotated start codons simply due to the subset of start codons that have reasonable SD matches to the altered ASD? Another way to think about this is to ask whether the start codons contributing to the signal in Figure 6C are the same start codons that contribute to the signal in Supplementary Figure 5A-B.

- Figure 7E shows the importance of an A-rich sequence in the context of start codons lacking a good S-D. Similar to Figure 6, these data highlight the contribution of non-SD sequences to translation initiation, but they do not provide any information about the relative importance of the different sequence elements.

Reviewer #2:

Major findings:

The paper of Saito and al. examines the contribution of Shine-Dalgarno sequence (SD) to the translation efficiency in bacteria. Using a clever approach, the authors use ribosome profiling to compare mRNA occupancy by wt ribosomes and ribosomes with the altered anti-SD sequence (ASD). In confirmation of previous findings from the Weissman lab, they find that the general translation efficiency does not correlate with the predicted strength of SD-ASD interactions. However, when all the other factors are masked, they observe a strong dependence of the initiation rate on the strength of SD-ASD pairing. They also noted that a subset of genes expressed in the stressed cells depend heavily on recognition of the SD sequence by the ribosomes. One of the unexpected, but highly important findings is the observation that the ribosomes with the altered ASD can nevertheless correctly and selectively initiate translation at the known start sites underscoring the importance of factors other than SD-ASD interactions in the start codon selection. Importantly, the reported work reveals the prevalence of A-rich motifs in the ribosome binding sites of the genes with weak SD sequences in E. coli and other bacteria. This trend becomes especially prominent in the bacterial species that do not rely on SD-ASD interactions for translation initiation.

Critique:

This is an interesting, intriguing and important study. The results are nice and clean and the implications are important for unraveling the fundamental mechanism of translation initiation in bacteria. Although the paper is generally well written, it was hard at times to follow the authors logic and I strongly encourage the authors to try to clarify the message, which often was hard to extract.

1) Here are several examples:

- Abstract: The statement "We reveal a genome-wide correlation between the SD strength and translational efficiency" is followed by "this global correlation is lost and a subset of genes […] becomes [dependent] on SD motifs for translation". This is hard to digest.

- Figure 4C legend ("the strength of the SD motifs determines whether wild-type or ASD mutants are recruited to messages") is supposed to contrast Figure 4F legend ("the unstructured SD motifs can recruit wild type ribosomes more effectively than they recruit ASD mutants"). However, they sound nearly identical and thus, do not accurately communicate the point the authors apparently are trying to make.

- "genes with strong SD motif are translated better by ribosomes with canonical ASD": better in comparison with the ASD-mutant ribosomes or better in comparison with the genes with weak SD?

2) Aleksashin et al., 2019, have shown that altering ASD in 16S rRNA compromises rRNA maturation. Although the presence of unprocessed sequences at the 5' and 3' end of the ASD-mutant 16S rRNA would not likely change the general conclusions of the paper, hypothetically it could affect the functionality and elongation rate of the mutant ribosomes. I am wondering whether authors have checked how well their mutant 16S rRNAs are processed. Irrespectively, I believe a more detailed discussion of the general functionality of the ribosomes with altered ASD, especially in relation to the elongation rate, would be beneficial.

3) Subsection “Gene-specific roles of SD motifs under stress”. The readers need a better explanation why the authors switched from ΔlogTE to ΔlogRPKM metrics when they move to the experiments in the stressed cells.

4) The influence of the competition between wt and mutant 30S subunits for the translation start sites on the conclusions drawn from ribosome profiling should be discussed.

Reviewer #3:

The authors introduce mutated 16S ribosomal RNAs into E. coli strains, altering their anti-Shine Dalgarno sequences, to investigate how these perturbations affect translation rates across the E. coli genome. To do this, they carry out ribosome profiling experiments to measure genome-wide ribosome densities on aSD-modified strains, including during exponential, stationary, and cold shock growth phases. Overall, they find that changing the last 9 nucleotides of the 16S rRNA has a significant effect on the transcriptomes' translation rates.

Overall, the collected measurements are interesting and potentially useful. However, the analysis suffers from a terribly incomplete knowledge of what controls a mRNA's translation rate. The statistics applied are tailored for a 1-factor problem, when in fact, there are many factors that control translation rate. There are also inconsistencies and errors in the authors' calculations that should be corrected. The authors' conclusions are not well supported by their analysis. The manuscript requires significant work for it to productively add to our knowledge of what controls translation rate in bacteria.

1) The author focuses primarily on the importance of the sequence colloquially known as the Shine-Dalgarno in controlling a mRNA's translation initiation rate. The authors write that "Initiation rates vary depending on how well an mRNA recruits 30S subunits to the start codon, and in bacteria, the working model is that this is accomplished primarily by Shine-Dalgarno (SD) motifs." This is incorrect. The current working model is that a mRNA's translation initiation rate is controlled by at least five important molecular interactions, only one is the hybridization between the last 9 nucleotides of the 16S rRNA and the mRNA. They include:

a) the hybridization between the last 9 nucleotides of the 16S rRNA and the mRNA;b) the unfolding of mRNA structures that overlap with the ribosome footprint;c) the differences in spacing (physical distance) between the 16S rRNA binding site and the start codon;d) the standby site's accessibility, as determined by the length of available single-stranded RNA;e) the start codon and its hybridization to the tRNA.

The free energy needed to unfold inhibitory mRNA structures is also affected by the dynamics of RNA folding (RNA folding kinetics) as well as the rate of ribosome binding.

If the authors better understood how translation rate was controlled, they could more productively use their measurements to push the real state-of-the-art forward. Their current conclusions are already subsumed within the state-of-the-art (i.e., nothing new).

2) Any discussion of "which translation rate interaction is most important" or "which translation rate interaction is responsible for X whereas the interaction Y only fine-tunes Z" is not productive and can easily be contradicted by selecting a real counter example. Overall, it is the binding free energy of the 30S ribosome to the mRNA that determines its translation initiation rate. Each of these interactions contributes free energy to this process and the magnitude of the contributed free energies can be roughly equal across a selection of real mRNA examples. There are unstructured mRNAs where there is little penalty for unfolding inhibitory mRNA structures. There are highly structured mRNAs that have consensus SDs sequences. There are mRNAs that have consensus SD sequences far away from the start codon. All of these mRNAs could have the same translation rate. Which interaction is most important? That's not the right question to ask, because it's meaningless.

3) The manuscript's main topic is the Shine-Dalgarno sequence, but the authors should be made aware that at least the last 9 nucleotides of the 16S rRNA can contact the mRNA and hybridize to it. In E. coli, the anti-Shine Dalgarno sequence is 5'-ACCUCCUUA-3' and the "consensus" Shine-Dalgarno sequence is therefore 5'-TAAGGAGGT-3'. The manuscript text and the authors' calculations should reflect this.

4) The authors are mis-using the ribosome profiling measurements in their analysis. Ribosome profiling measurements do not directly measure translation rates. They measure mRNA-bound ribosome densities. A mRNA's ribosome density will depend on both its translation initiation rate AND its translation elongation rate. Specifically, in steady-state conditions, the ribosome density will be the ratio between these two quantities (initiation rate over elongation rate). In the initial applications of ribosome profiling, researchers assumed that all mRNAs have the same translation elongation rate in order to conclude that ribosome density measurements were proportional to translation initiation rates. This is not true. Coding sequences in mRNAs have very different translation elongation rates, due to differences in synonymous codon usage. Unless each mRNAs' translation elongation rates are predicted or directly measured, ribosome density measurements cannot be used to infer their translation initiation rates. Therefore, when the authors write "In pioneering ribosome profiling studies in bacteria, the paradoxical observation was made that there is little or no correlation between the translational efficiency of a gene and the strength of its SD motif (calculated using thermodynamic algorithms for RNA pairing), as had been anticipated based on the SD model." there is no actual paradox. The ribosome profiling measurements were not used correctly to test how mRNA sequences control translation rate.

5) Getting to the authors' main conclusions, they write that "These data indicate that the ASD mutant ribosomes translate genes with weak SD motifs better than genes with strong SD motifs, exactly the opposite of what wild-type ribosomes are expected to do." This statement is confusing given the real conclusion of the authors, that all other factors being equal a "strong" SD motif does result in higher translation than a "weak" SD motif. It's only because of other confounding factors that the initial analysis did not yield a positive correlation. An incorrect analysis (excluding confounding variables) cannot lead to a correct conclusion.

6) Figure 2C shows a very interesting and productive result, that the difference in translation efficiency between the wild-type "C" ribosomes and the A-ribosomes correlates to some degree with the hybridization free energy between the mRNA and (a portion of) the anti-SD sequence. This is a productive approach towards eliminating key confounding variables because, in principle, the strengths of the four other interactions that control translation initiation rate should not change when the 16S rRNA aSD sequences are changed. However, it's not apparent in the manuscript text, but the authors are using the modified 16S rRNAs to “eliminate the SD-aSD interaction as a contribution to the mRNA's translation rate”. So when they subtract the contribution from the modified A-ribosome's translation rates from the C-ribosome's translation rate, they are observing more directly the contribution from the SD-aSD interaction. The manuscript text should more clearly explain this experimental design. This is a creative and valid way of using ribosome profiling measurements.

7) However, the hybridization free energy calculations could be improved. First, as mentioned previously, the wild-type aSD sequence in E. coli is ACCUCCUUA. Second, the hybridization free energy calculation was only performed on the region from 15 to 6 nucleotides upstream of the start codon, but the aSD sequence can hybridize at other locations. Third, the hybridization between the mRNA and aSD can accommodate 1 or 2-nucleotide bulges or internal loops.

8) The measurements in cold shock are greatly confounded by the higher expression levels of RNA chaperones that are unfolding mRNA structures “at specific mRNAs” where the RNA chaperones recognize binding motifs. The conclusion here should be that RNA chaperones bind specific mRNAs, unfold their inhibitory mRNA structures, and increase their translation rates during cold shock. This is all independent of the Shine-Dalgarno sequence. This process also does not depend on many other uninteresting factors.

9) The use of ORF-wide GINI values is odd because it's generally only the region surrounding the start codon that affects its translation initiation rate, and not the structure of the entire ORF (which this coefficient is quantifying). Also, using the SHAPE reactivity around a start codon as a proxy for RNA structure is a bit misleading as ribosomes actively unfold RNA structures during translation initiation. A highly structured mRNA with a consensus SD sequence will have a high SHAPE reactivity (i.e., low RNA structure) because the ribosomes can rapidly bind to the mRNA and unfold the mRNA structure. SHAPE reactivity is measuring the effect of rapid ribosome binding and not the cause of it. Rapid ribosome binding can also be facilitated by slow RNA refolding kinetics, called "Ribosome Drafting" in the literature.

10) The data in Figure 6 just says that A-ribosomes can initiate translation rate at other start codons because they now have more negative binding free energies to those start codons, compared to the annotated ones. The authors could perform hybridization calculations using the A-ribosome's aSD sequence to investigate whether these "new start codons" have a nearby "SD" sequence that is complementary to the A-ribosome's aSD. That would be interesting.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Translational initiation in E. coli occurs at the correct sites genome-wide in the absence of mRNA-rRNA base-pairing" for further consideration by eLife. Your revised article has been evaluated by James Manley as the Senior Editor, and three reviewers, including Joe Wade as the Reviewing Editor and Reviewer #1.

The reviewers and editors agree that the revised manuscript is greatly improved, and we are pleased to provisionally accept the manuscript. We ask that you make a few small changes in response to the reviewers' comments. First, reviewer 2 has two minor concerns that are easily addressed. Second, based on reviewer 3's comments, the conclusions regarding the importance of A-rich sequences should be softened a little. The reviewers' comments are listed below:

Reviewer #1:

The authors have done an excellent job improving the manuscript. Removing the data on cold-shock has improved the focus and readability. Moreover, the new analyses in Figures 3 and 4 make a more compelling case that Shine-Dalgarno sequences are neither necessary nor sufficient for start site selection.

Reviewer #2:

The streamlined paper of Saito et al. reads much better than the original version and delivers a clear and impactful message.

I believe it can be published after authors address two remaining issues:

The authors refer to "the number of elongating ribosomes per mRNA as a proxy of initiation rates". This is incorrect: there would be twice as many ribosomes on an mRNA that is twice as long as another one, even if those two would have the same initiation rate. The correct metrics is not the number of ribosomes per mRNA but the ribosome density (their number normalized by mRNA length). This does not affect conclusions of the paper because authors normalize RiboSeq reads by RNASeq reads. Yet, I would try to avoid this confusion.

The authors write: "Interestingly, in comparing internal AUG codons that support initiation in our ribosome profiling data to those that do not, we found that A's are enriched both upstream and downstream of initiation sites (Figure 5A).…. This results from endogenous initiation sites… ". However, Figure 5A does not deal with the internal initiation sites, but with the annotated sites lacking SD.

Reviewer #3:

With the revisions, the authors have greatly improved the manuscript's introductory description of translation and the overall analysis of their dataset, leading to a more laser-focused and well-supported set of conclusions. These results provide an excellent and clarifying view of the sequence determinants and interactions that control translation initiation rate within natural (highly evolved) mRNAs by cleanly separating the role of the SD:aSD interaction from other factors, including the presence/absence of inhibitory mRNA structures.

The Introduction provides a more comprehensive description of the several sequence determinants and factors that control translation initiation rate, which is essential towards understanding the authors' excellent dataset. The analysis clearly explains how their dataset provides comparative measurements with vs. without the SD:aSD interaction and how those measurements quantify its effect on translation rate. Start codon selection is a property of all the factors that control translation initiation rate, and is also likely a property of ribosome-ribosome dynamics along the mRNA.

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Reviewer #1:

This paper describes an innovative approach to probe the importance of Shine-Dalgarno (S-D) sequences in translation initiation in Escherichia coli. […] I also have concerns about the interpretation of Figures 6 and 7 that impact the overall conclusions.

- The presentation of the cold shock data is confusing. The overall conclusion is that S-D-dependence is lost at almost all genes during cold shock. However, a small subset of genes appears to depend strongly on the S-D. The distinction between the effect on the majority of genes and the effect on a small subset should be explained more clearly. The simplest interpretation of these data is that most start codons are highly structured during cold shock, but those that are do not rely on their S-Ds. This model is largely consistent with previous work.

We have removed the figures dealing with cold shock and other stresses. We agree that they are largely consistent with previous work. The confusion raised by the cold shock story appears to have taken away from the main story.

- I disagree with the interpretation of Figure 6. The data show that for the altered ribosomes, annotated start codons are used far more efficiently than the collection of all other ATG sequences within ORFs. However, there are many more ATGs within ORFs than annotated start codons, and even if translation relies heavily on S-D sequences, you would expect that most ATGs within ORFs would not be selected by alternative ribosomes because only a small subset will have appropriate S-D sequences, and many may be weakly expressed. My interpretation of these data is that alternative ribosomes do use annotated start codons, but there is no way to tell how selectively they do this. A more appropriate comparison would be of (i) annotated start codons to (ii) ATGs within ORFs where the ATG is associated with a sequence that is predicted to function as a good S-D for the alternative ribosome.

We have added analyses to the new Figure 4 and Figure 4—figure supplement 1 showing initiation at internal AUG codons predicted to have high affinity for the mutant ASD sequences. These data show that initiation occurs with all four ribosome types regardless of the SD strength or specificity, but that initiation is most efficient when the SD and ASD are complementary.

- Another concern I have with Figure 6 is that presumably some, and perhaps many of the annotated start codons will have good SD matches for the alternative ribosomes. Figure 2C suggests that the number with good matches will be fairly high. Is the ribosome density at annotated start codons simply due to the subset of start codons that have reasonable SD matches to the altered ASD? Another way to think about this is to ask whether the start codons contributing to the signal in Figure 6C are the same start codons that contribute to the signal in Supplementary Figure 5A-B.

We added analyses to the new Figure 3 showing that initiation occurs with all three mutant ribosomes at annotated start sites that have no affinity for the mutant ASD sequences.

- Figure 7E shows the importance of an A-rich sequence in the context of start codons lacking a good S-D. Similar to Figure 6, these data highlight the contribution of non-SD sequences to translation initiation, but they do not provide any information about the relative importance of the different sequence elements.

We have removed claims about the relative importance of different sequence elements.

Reviewer #2:

Critique:

This is an interesting, intriguing and important study. The results are nice and clean and the implications are important for unraveling the fundamental mechanism of translation initiation in bacteria. Although the paper is generally well written, it was hard at times to follow the authors logic and I strongly encourage the authors to try to clarify the message, which often was hard to extract.

1) Here are several examples:

- Abstract: The statement "We reveal a genome-wide correlation between the SD strength and translational efficiency" is followed by "this global correlation is lost and a subset of genes […] becomes [dependent] on SD motifs for translation". This is hard to digest.

- Figure 4C legend ("the strength of the SD motifs determines whether wild-type or ASD mutants are recruited to messages") is supposed to contrast Figure 4F legend ("the unstructured SD motifs can recruit wild type ribosomes more effectively than they recruit ASD mutants"). However, they sound nearly identical and thus, do not accurately communicate the point the authors apparently are trying to make.

- "genes with strong SD motif are translated better by ribosomes with canonical ASD": better in comparison with the ASD-mutant ribosomes or better in comparison with the genes with weak SD?

We removed the section on stress conditions that was hard to follow and taking away from the main point of the manuscript.

2) Aleksashin et al., 2019, have shown that altering ASD in 16S rRNA compromises rRNA maturation. Although the presence of unprocessed sequences at the 5' and 3' end of the ASD-mutant 16S rRNA would not likely change the general conclusions of the paper, hypothetically it could affect the functionality and elongation rate of the mutant ribosomes. I am wondering whether authors have checked how well their mutant 16S rRNAs are processed. Irrespectively, I believe a more detailed discussion of the general functionality of the ribosomes with altered ASD, especially in relation to the elongation rate, would be beneficial.

RNA-seq analyses of rRNA (prior to nuclease treatment) is now shown in Figure 1—figure supplement 1 and discussed early in the Results section (subsection “Selective profiling of ribosomes with mutant ASD sequences”).

3) Subsection “Gene-specific roles of SD motifs under stress”. The readers need a better explanation why the authors switched from ΔlogTE to ΔlogRPKM metrics when they move to the experiments in the stressed cells.

This section was removed.

4) The influence of the competition between wt and mutant 30S subunits for the translation start sites on the conclusions drawn from ribosome profiling should be discussed.

This possibility was added to the Discussion.

Reviewer #3:

1) The author focuses primarily on the importance of the sequence colloquially known as the Shine-Dalgarno in controlling a mRNA's translation initiation rate. The authors write that "Initiation rates vary depending on how well an mRNA recruits 30S subunits to the start codon, and in bacteria, the working model is that this is accomplished primarily by Shine-Dalgarno (SD) motifs." This is incorrect. […] Their current conclusions are already subsumed within the state-of-the-art (i.e., nothing new).

We added a more detailed description of the factors that affect initiation rates to the Introduction, including the points listed above.

2) Any discussion of "which translation rate interaction is most important" or "which translation rate interaction is responsible for X whereas the interaction Y only fine-tunes Z" is not productive and can easily be contradicted by selecting a real counter example. Overall, it is the binding free energy of the 30S ribosome to the mRNA that determines its translation initiation rate. Each of these interactions contributes free energy to this process and the magnitude of the contributed free energies can be roughly equal across a selection of real mRNA examples. There are unstructured mRNAs where there is little penalty for unfolding inhibitory mRNA structures. There are highly structured mRNAs that have consensus SDs sequences. There are mRNAs that have consensus SD sequences far away from the start codon. All of these mRNAs could have the same translation rate. Which interaction is most important? That's not the right question to ask, because it's meaningless.

We have removed language that focuses on the relative contribution of the individual factors that affect translational initiation. We agree that our analyses do not allow us to determine their relative contributions.

3) The manuscript's main topic is the Shine-Dalgarno sequence, but the authors should be made aware that at least the last 9 nucleotides of the 16S rRNA can contact the mRNA and hybridize to it. In E. coli, the anti-Shine Dalgarno sequence is 5'-ACCUCCUUA-3' and the "consensus" Shine-Dalgarno sequence is therefore 5'-TAAGGAGGT-3'. The manuscript text and the authors' calculations should reflect this.

We revised the Introduction to explicitly state that up to 9 bp can form. Drawing on previous work, we refer to the “consensus” as GGAGG because it is the G’s that are overrepresented upstream of start codons (see the data for E. coli in Figure 5—figure supplement 1).

4) The authors are mis-using the ribosome profiling measurements in their analysis. Ribosome profiling measurements do not directly measure translation rates. They measure mRNA-bound ribosome densities. A mRNA's ribosome density will depend on both its translation initiation rate AND its translation elongation rate. Specifically, in steady-state conditions, the ribosome density will be the ratio between these two quantities (initiation rate over elongation rate). In the initial applications of ribosome profiling, researchers assumed that all mRNAs have the same translation elongation rate in order to conclude that ribosome density measurements were proportional to translation initiation rates. This is not true. Coding sequences in mRNAs have very different translation elongation rates, due to differences in synonymous codon usage. Unless each mRNAs' translation elongation rates are predicted or directly measured, ribosome density measurements cannot be used to infer their translation initiation rates. Therefore, when the authors write "In pioneering ribosome profiling studies in bacteria, the paradoxical observation was made that there is little or no correlation between the translational efficiency of a gene and the strength of its SD motif (calculated using thermodynamic algorithms for RNA pairing), as had been anticipated based on the SD model." there is no actual paradox. The ribosome profiling measurements were not used correctly to test how mRNA sequences control translation rate.

We revised the Introduction to explicitly state that up to 9 bp can form. Drawing on previous work, we refer to the “consensus” as GGAGG because it is the G’s that are overrepresented upstream of start codons (see the data for E. coli in Figure 5—figure supplement 1).

5) Getting to the authors' main conclusions, they write that "These data indicate that the ASD mutant ribosomes translate genes with weak SD motifs better than genes with strong SD motifs, exactly the opposite of what wild-type ribosomes are expected to do." This statement is confusing given the real conclusion of the authors, that all other factors being equal a "strong" SD motif does result in higher translation than a "weak" SD motif. It's only because of other confounding factors that the initial analysis did not yield a positive correlation. An incorrect analysis (excluding confounding variables) cannot lead to a correct conclusion.

The sentence, “These data indicate that the ASD mutant ribosomes translate genes with weak SD motifs better than genes with strong SD motifs” describes the observations in Figure 2B, the result of all the other factors except for SD-ASD pairing. We removed the phrase “exactly the opposite of what wild-type ribosomes are expected to do” that seems to have caused the confusion.

6) Figure 2C shows a very interesting and productive result, that the difference in translation efficiency between the wild-type "C" ribosomes and the A-ribosomes correlates to some degree with the hybridization free energy between the mRNA and (a portion of) the anti-SD sequence. This is a productive approach towards eliminating key confounding variables because, in principle, the strengths of the four other interactions that control translation initiation rate should not change when the 16S rRNA aSD sequences are changed. However, it's not apparent in the manuscript text, but the authors are using the modified 16S rRNAs to “eliminate the SD-aSD interaction as a contribution to the mRNA's translation rate”. So when they subtract the contribution from the modified A-ribosome's translation rates from the C-ribosome's translation rate, they are observing more directly the contribution from the SD-aSD interaction. The manuscript text should more clearly explain this experimental design. This is a creative and valid way of using ribosome profiling measurements.

We revised the language at the end of the Introduction and the beginning of the Results section to better explain our experimental design. The fact that we can isolate the effects of SD-ASD interactions from the other factors that set initiation rates explains why we use statistics for single variables and focus primarily on the SD mechanism of initiation.

7) However, the hybridization free energy calculations could be improved. First, as mentioned previously, the wild-type aSD sequence in E. coli is ACCUCCUUA. Second, the hybridization free energy calculation was only performed on the region from 15 to 6 nucleotides upstream of the start codon, but the aSD sequence can hybridize at other locations. Third, the hybridization between the mRNA and aSD can accommodate 1 or 2-nucleotide bulges or internal loops.

The fact that we see a strong correlation between our calculated SD affinities and differences in ribosome occupancy (WT – mutant) argues that the calculations are basically reliable. We see the highest correlation when affinities are calculated using the 10 nt between -15 and -6 from the AUG and we show the data for various windows with different SD distances in Figure 2—figure supplement 2. The calculations are quite robust to changes in parameters: we see little or no differences in SDRO correlations if we use 9 nt of ASD sequence to calculate affinities instead of 7 nt, or if we allow the ASD to pair anywhere between -20 to 0 upstream of AUG, or if we use the RBS calculator to generate the ΔG values.

8) The measurements in cold shock are greatly confounded by the higher expression levels of RNA chaperones that are unfolding mRNA structures “at specific mRNAs” where the RNA chaperones recognize binding motifs. The conclusion here should be that RNA chaperones bind specific mRNAs, unfold their inhibitory mRNA structures, and increase their translation rates during cold shock. This is all independent of the Shine-Dalgarno sequence. This process also does not depend on many other uninteresting factors.

We removed the section on the role of SD motifs under stress because it generated confusion and ornithological references without strengthening the main point of our manuscript.

9) The use of ORF-wide GINI values is odd because it's generally only the region surrounding the start codon that affects its translation initiation rate, and not the structure of the entire ORF (which this coefficient is quantifying). Also, using the SHAPE reactivity around a start codon as a proxy for RNA structure is a bit misleading as ribosomes actively unfold RNA structures during translation initiation. A highly structured mRNA with a consensus SD sequence will have a high SHAPE reactivity (i.e., low RNA structure) because the ribosomes can rapidly bind to the mRNA and unfold the mRNA structure. SHAPE reactivity is measuring the effect of rapid ribosome binding and not the cause of it. Rapid ribosome binding can also be facilitated by slow RNA refolding kinetics, called "Ribosome Drafting" in the literature.

Carol Gross and colleagues showed that ORF-wide GINI values are highly correlated with translational efficiency genome-wide. This is true whether the DMS probing is done in vivo (where ribosomes could affect structure by unwinding the RNA) or to a lesser extent with purified RNA in vitro. Kevin Weeks also showed that RNA structures are correlated in vivo and in vitro using the SHAPE reagent. These data argue that at least to some extent mRNA structure is driving translation rates. This was clarified in the Results section in the discussion of Figure 5E.

10) The data in Figure 6 just says that A-ribosomes can initiate translation rate at other start codons because they now have more negative binding free energies to those start codons, compared to the annotated ones. The authors could perform hybridization calculations using the A-ribosome's aSD sequence to investigate whether these "new start codons" have a nearby "SD" sequence that is complementary to the A-ribosome's aSD. That would be interesting.

The new Figures 3 and 4 now include analyses of sets of initiation sites with various affinities for the WT or mutant ASD sequences showing more clearly that translation occurs at start codons even without strong SD-ASD pairing.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Reviewer #2:

The streamlined paper of Saito et al. reads much better than the original version and delivers a clear and impactful message.

I believe it can be published after authors address two remaining issues:

The authors refer to "the number of elongating ribosomes per mRNA as a proxy of initiation rates". This is incorrect: there would be twice as many ribosomes on an mRNA that is twice as long as another one, even if those two would have the same initiation rate. The correct metrics is not the number of ribosomes per mRNA but the ribosome density (their number normalized by mRNA length). This does not affect conclusions of the paper because authors normalize RiboSeq reads by RNASeq reads. Yet, I would try to avoid this confusion.

The language was changed to “ribosome density” instead of “the number of ribosomes.”

The authors write: "Interestingly, in comparing internal AUG codons that support initiation in our ribosome profiling data to those that do not, we found that A's are enriched both upstream and downstream of initiation sites (Figure 5A).…. This results from endogenous initiation sites… ". However, Figure 5A does not deal with the internal initiation sites, but with the annotated sites lacking SD.

Thanks for catching this mistake; the Discussion was updated to reflect this.

Author details

1. Kazuki Saito

   Department of Molecular Biology and Genetics, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   Conceptualization, Formal analysis, Investigation, Methodology, Writing - original draft

   Competing interests

   No competing interests declared

2. Rachel Green

   1. Department of Molecular Biology and Genetics, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   Conceptualization, Funding acquisition, Writing - review and editing

   Competing interests

   Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0001-9337-2003

3. Allen R Buskirk

   Department of Molecular Biology and Genetics, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing - review and editing

   For correspondence

   buskirk@jhmi.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-2720-6896

• 12,605

  Page views

• 1,232

  Downloads

• 39

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.