1. Genetics and Genomics
  2. Microbiology and Infectious Disease
Download icon

Organisms with alternative genetic codes resolve unassigned codons via mistranslation and ribosomal rescue

  1. Natalie Jing Ma
  2. Colin F Hemez
  3. Karl W Barber
  4. Jesse Rinehart  Is a corresponding author
  5. Farren J Isaacs  Is a corresponding author
  1. Yale University, United States
  2. Yale University School of Medicine, United States
Research Article
  • Cited 0
  • Views 856
  • Annotations
Cite this article as: eLife 2018;7:e34878 doi: 10.7554/eLife.34878

Abstract

Organisms possessing genetic codes with unassigned codons raise the question of how cellular machinery resolves such codons and how this could impact horizontal gene transfer. Here, we use a genomically recoded Escherichia coli to examine how organisms address translation at unassigned UAG codons, which obstruct propagation of UAG-containing viruses and plasmids. Using mass spectrometry, we show that recoded organisms resolve translation at unassigned UAG codons via near-cognate suppression, dramatic frameshifting from at least −3 to +19 nucleotides, and rescue by ssrA-encoded tmRNA, ArfA, and ArfB. We then demonstrate that deleting tmRNA restores expression of UAG-ending proteins and propagation of UAG-containing viruses and plasmids in the recoded strain, indicating that tmRNA rescue and nascent peptide degradation is the cause of impaired virus and plasmid propagation. The ubiquity of tmRNA homologs suggests that genomic recoding is a promising path for impairing horizontal gene transfer and conferring genetic isolation in diverse organisms.

https://doi.org/10.7554/eLife.34878.001

eLife digest

Usually, DNA passes from parent to offspring, vertically down the generations. But not always. In some cases, it can move directly from one organism to another by a process called horizontal gene transfer. In bacteria, this happens when DNA segments pass through a bacterium’s cell wall, which can then be picked up by another bacterium. Because the vast majority of organisms share the same genetic code, the bacteria can read this DNA with ease, as it is in the same biological language.

Horizontal gene transfer helps bacteria adapt and evolve to their surroundings, letting them swap and share genetic information that could be useful. The process also poses a threat to human health because the DNA that bacteria share can help spread antibiotic resistance. However, some organisms use an alternative genetic code, which obstructs horizontal gene transfer. They cannot read the DNA transmitted to them, because it is in a different ‘biological language’. The mechanism of how this language barrier works has been poorly understood until now.

Ma, Hemez, Barber et al. investigated this using Escherichia coli bacteria with an artificially alternated genetic code. In this E. coli, one of the three-letter DNA ‘words’ in the sequence is a blank – it does not exist in the bacterium’s biological language. This three-letter DNA word normally corresponds to a particular protein building block. Using a technique called mass spectrometry, Ma et al. analyzed the proteins this E. coli forms. The results showed that it has several strategies to deal with DNA transmitted horizontally into the bacterium. One method is destroying the proteins that are half-created from the DNA, using molecules called tmRNAs. These are part of a rescue system that intervenes when protein translation stalls on the blank word. The tmRNAs help to add a tag to half-formed proteins, marking them for destruction.

This mechanism creates a ‘genetic firewall’ that prevents horizontal gene transfer. In organisms engineered to work from an altered genetic code, this helps to isolate them from outside interference. The findings could have applications in creating engineered bacteria that are safer for use in fields such as medicine and biofuel production.

https://doi.org/10.7554/eLife.34878.002

Introduction

The standard genetic code allows faithful translation of proteins across nearly all living organisms and enables horizontally transferred genetic elements (HTGEs), such as conjugative plasmids and viruses, to exploit a host’s translational machinery (Krakauer and Jansen, 2002). Since naturally occurring exceptions to the standard genetic code exist (Ambrogelly et al., 2007; Knight et al., 2001), researchers have hypothesized that such alternative genetic codes might arise to escape viral predation (Shackelton and Holmes, 2008). Recent research supports this hypothesis, with modification to codon usage or the genetic code reducing the ability of viruses and conjugative plasmids to exploit their hosts (Coleman et al., 2008; Lajoie et al., 2013b; Ma and Isaacs, 2016). Given the medical, technological, and evolutionary importance of HTGE-mediated horizontal gene transfer (HGT) (Davies, 1994; Gogarten and Townsend, 2005; Moe-Behrens et al., 2013; Ochman et al., 2000), understanding the molecular basis for how alternative genetic codes impede HTGEs is vital.

At the molecular level, an alternative genetic code arises from reassignment of one or more codons in the genetic code, which stems from a change in the ability of an aminoacyl-tRNA or release factor (RF) to recognize codon(s) during translation. One possible alteration of the genetic code is the loss of a codon assignment through the deletion or modification of an aminoacyl-tRNA or release factor, removing the cell’s ability to decode that codon (Figure 1A). Such unassigned codons are found in alternative genetic codes in nature (Knight et al., 2001) and have been engineered into genomically recoded organisms (GROs) derived from Escherichia coli (Isaacs et al., 2011; Lajoie et al., 2013b). We recently demonstrated that a GRO with an unassigned UAG codon (i.e. lacking all instances of the UAG codon and release factor 1, RF1) impaired the propagation of HTGEs carrying UAG-ending genes, illustrating that alternative genetic codes can obstruct HGT (Ma and Isaacs, 2016) and establishing the GRO as an ideal model to study the molecular mechanisms that act at unassigned codons to impair HTGEs.

A UAG-ending transcript in the genomically recoded organism (GRO) may produce proteins with multiple differing C-termini.

(A) Unassigned codons arise when either the cognate tRNA or release factor recognizing a codon are removed. (B) Since the GRO lacks Release Factor 1 (RF1), ribosomal stalling at the UAG codons results in three possible fates for the nascent protein (blue): (1) suppression of the codon by a near-cognate or suppressor tRNA (yellow) and continued translation, (2) frameshifting of bases along the mRNA transcript into a new reading frame and continued translation (purple), or (3) ribosomal rescue by the ssrA-encoded tmRNA, ArfA, or ArfB proteins. If ribosomal rescue occurs via tmRNA, the resulting protein is tagged with a peptide sequence (red) for degradation, while rescue via ArfA or ArfB results in release of peptide without C-terminal modification.

https://doi.org/10.7554/eLife.34878.003

Encountering an unassigned codon during translation leads to ribosomal stalling, and without resolution, to cell death (Keiler and Feaga, 2014). However, the survival of organisms engineered to lack RF1 but retaining some UAG codons in their protein-coding sequences (Heinemann et al., 2012; Mukai et al., 2010) and the ability of GROs to resist exploitation by and continue growth in the presence of HTGEs (Ma and Isaacs, 2016) indicates that E. coli can resolve translation at unassigned UAG codons. We hypothesize that three mechanisms could resolve translation at prokaryotic ribosomes encountering these unassigned codons, each resulting in peptides with different C-terminal sequences (Figure 1B): (1) suppression of the codon by a near-cognate or mutated tRNA (e.g. amber suppressor) and continued translation, (2) frameshifting of bases along the mRNA transcript into a new reading frame and continued translation, or (3) stalling that elicits one of three ribosomal rescue pathways (tmRNA-SmpB, ArfA, or ArfB) in the cell (Keiler, 2015). The tmRNA-SmpB system acts as the primary rescue mechanism in prokaryotes, resolving ribosomal stalling that arises from the translation of mRNAs lacking a stop codon due to mRNA degradation, frameshifting, and stop codon read-through (Keiler, 2015). tmRNA-SmpB can also rescue ribosomes stalled on intact mRNAs for structural reasons (Cruz-Vera et al., 2011; Keiler, 2015; Li et al., 2012). The ssrA-encoded tmRNA associates with SmpB to form the tmRNA-SmpB complex, which adds a C-terminal degradation tag to peptides on stalled ribosomes (Tu et al., 1995). ArfA and ArfB, the secondary ribosomal rescue systems, alleviate stalling and release the stalled ribosome’s nascent peptide without modification (Chadani et al., 2012; Shimizu, 2012). tmRNA, ArfA, and ArfB all act on nonstop ribosomal complexes, which are stalled ribosomes that have reached the 3’ end of an mRNA because of stop-codon readthrough or because of the loss of a stop codon due to 3’ exonuclease degradation (Keiler, 2015). A possible fourth outcome identified from in vitro studies is loss of translational fidelity after the ribosome encounters rare or unassigned codons (Gingold and Pilpel, 2011), followed by untemplated termination by release factor 2 (RF2) (Zaher and Green, 2009).

Studies of ribosomal stalling arising at rare codons (Hayes et al., 2002) or in contexts of depleted or inefficient cognate decoding elements (George et al., 2016; Li et al., 2007; Roche and Sauer, 1999) suggest that a number of these mechanisms could resolve translation at unassigned codons, but a lack of well-characterized model organisms with an unassigned codon has precluded direct study of this question. Here, we use the GRO as a model to demonstrate that unassigned UAG codons in mRNA transcripts (1) elicit suppression, ribosomal frameshifting, and ribosomal rescue, (2) can induce ribosomal frameshifting from at least −3 to +19 nucleotides, and (3) lead to total loss of translational fidelity. By selectively deleting ribosomal rescue pathways in the GRO, we show that the tmRNA system is primarily responsible for rescuing ribosomes stalled at unassigned codons, with deletion of the tmRNA restoring expression of UAG-ending genes and re-enabling propagation of UAG-containing plasmids and viruses in the GRO. Our work reveals mechanistic details into how cells rescue ribosomes stalled at unassigned stop codons, providing insight into how alternative genetic codes act as barriers to HTGEs and demonstrating genomic recoding as a broadly applicable strategy to obstruct HGT in engineered organisms.

Results

Suppression, ribosomal frameshifting, and tmRNA-mediated peptide tagging occur at unassigned codons

In prior work, we constructed an Escherichia coli strain in which all UAG codons were mutated to UAA, permitting the deletion of release factor 1 (RF1) and resulting in an organism that lacks a codon assignment of UAG. This genomically recoded organism (GRO) (Isaacs et al., 2011; Lajoie et al., 2013b) exhibited resistance to multiple viruses and failure to propagate conjugative plasmids (Lajoie et al., 2013b; Ma and Isaacs, 2016) attributable to the unassigned UAG codon, but the molecular mechanisms that resolve unassigned UAG codons during translation remained unknown. In this study, we conducted two main experiments to uncover these mechanisms: (1) analysis of proteins translated from UAG-ending transcripts via mass spectrometry and western blots and (2) phenotypic assays to assess whether gene deletions of specific rescue factors restored the ability of conjugative plasmids and viruses to exploit the GRO. Since we hypothesized that the tmRNA-mediated response may resolve ribosomal stalling at the UAG codon, we also mutated the degradation tag encoded by the tmRNA from AANDENYALAA (AA-tag) to AANDENYALDD (DD-tag) for protein expression for mass spectrometry experiments. This mutation increases the half-life of protein products released by tmRNA (Keiler et al., 1996; Roche and Sauer, 1999), enabling their detection via mass spectrometry.

We assembled plasmids (pUAG-GFP and pUAA-GFP) encoding GFP genes with C-terminal 6x-His tags positioned immediately upstream of a UAG or UAA stop codon. We then expressed GFP from pUAG-GFP and pUAA-GFP in GRO cells containing the RF1-encoding prfA gene (GRO.DD.prfA+) or in GRO cells lacking prfA and consequentially without UAG assignment (GRO.DD) (Figure 2A; Table 1; see also Key Resources Table for a list of plasmids used in this study). We then purified proteins by nickel affinity chromatography, performed trypsin digest, and used tandem mass spectrometry to collect peptide mass data as described previously (Aerni et al., 2015; Amiram et al., 2015). To distinguish between mechanisms of ribosomal rescue and mistranslation at the UAG codon, we searched mass spectrometry data with theoretical peptide libraries detailed in Table 2 (see also Supplementary file 3 and 4) to identify evidence for suppression, ribosomal frameshifting, rescue via tmRNA tagging, and loss of translational fidelity.

UAG codons in the genomically recoded organism elicit suppression, frameshifting, and tagging for degradation by the tmRNA.

(A) Schematic of the GFP construct with a C-terminal 6x-His tag and a UAG stop codon, showing 102 nucleotides downstream of the UAG codon and the positions of other stop codons in the downstream tail. (B) Peptides identified from the C-terminus of a UAG-ending GFP construct expressed in the GRO (using libraries detailed in Supplementary file 3 and 4). Purified GFP protein was digested with trypsin, processed via MS/MS, and the resulting data were computationally searched using libraries encoding all possible suppressors and all possible subsequent reading frames. Peptides are mapped to the C-terminus of the original GFP construct and grouped by reading frame, with the number of bases skipped listed in the left column. Green text represents GFP, blue text represents the C-terminal 6xHis tag and unframeshifted readthrough, orange text represents the position of a UAG stop codon, purple text represents frameshifted readthrough, and red text represents the tmRNA tag. Black dashes represent ribosomal frameshifts (Figure 2—source datas 1 and 2). (C) MS-MS spectra for two peptides: the C-terminus of GFP with the appended degradation tag (LEHHHHHHAANDENYALDD) and the C-terminus of GFP demonstrating a + 10 base skip in translation (LEHHHHHHGDPMVR). The other spectra validated from UAG-GFP expressing GRO.AA are shown in Supplementary file 2.

https://doi.org/10.7554/eLife.34878.004
Table 1
Strains used in this study.
https://doi.org/10.7554/eLife.34878.007
Strain Abbreviation*Ancestor (source)Genotype# UAG CodonsRF1 Status§Ribosomal rescue gene deletionssrA tag Status#Investigated in
GRO.DD.prfA+GRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla]0+RF1n/aDDGFP expression for mass spectrometry (Figure 2)
GRO.DDGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1n/aDDGFP expression for mass spectrometry (Figure 2)
ECNR2.AAE. coli MG1655 (Wang et al., 2009)MG1655 ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla]321+RF1n/aAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AAECNR2.AA (Lajoie et al., 2013b)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1n/aAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AA.∆ssrAGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1ssrAAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AA.∆arfAGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1arfAAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AA.∆arfBGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1arfBAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AA.∆ssrA.arfBGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1ssrA,arfBAAFitness, conjugation, and viral infection (Figures 3 and 4)
GRO.AA.∆arfA.arfBGRO.AA (this study)ΔmutS:zeo.Δ(ybhB-bioAB):[λcI857.Δ(cro-ea59):tetR-bla], ΔprfA, ΔtolC0∆RF1arfA,arfBAAFitness, conjugation, and viral infection (Figures 3 and 4)
  1. *All strains derived from ECNR2, as described in Wang et al. (2009).

    †See Key Resources Table for additional information on strains and sources. The GenBank accession number for E. coli MG1655 is U00096, and the GenBank accession number for GRO.AA is CP006698.

  2. ‡ Out of a total of 321 in the original ECNR2 strain.

    §RF1 terminates translation at UAG and UAA. Deletion of RF1 eliminates recognition of UAG during translation; translational termination continues through RF2, which recognizes UAA and UGA.

  3. #The ssrA gene encodes the tmRNA, which appends the ssrA degradation tag to stalled ribosomes. The wild-type sequence is AANDENYALAA; mutation of the C-terminus to AANDENYALDD slows degradation of peptides to enable detection by mass spectrometry.

Table 2
Components of peptide library constructed to search and analyze tandem mass spectrometry data.

The LEHHHHHHXXX library was separate from the library that contained the entries of the first three rows of the table (see Supplementary file 3 and 4).

https://doi.org/10.7554/eLife.34878.008
Library componentExample peptides (from Figure 2A)Enables detection of…Complete peptide list
Any one of 20 canonical amino acids inserted at the UAG codonLEHHHHHHQGARNear-cognate suppressionSupplementary file 3
Any length of C-tail following UAG codon to the next non-UAG stop codon or to 38 amino acids downstream of the UAG codon, whichever came firstALGDPMVRReadthrough, frameshifting, and rescue by ArfA or ArfBSupplementary file 3
AANDENYALDD degradation tagLEHHHHHHGDAANDENYALDDRescue by tmRNA-SmpBSupplementary file 3
All peptides of form LEHHHHHHXXX, where X is any amino acidLEHHHHHHQLDLoss of translational fidelitySupplementary file 4

In the GRO lacking UAG assignment, the UAG codon elicited a combination of ribosomal rescue mechanisms and mistranslation events, including tmRNA-mediated tagging, near-cognate suppression, and frameshifting. The mutated ssrA DD-tag appended directly to the C-terminus of GFP (LEHHHHHHAANDENYALDD) appeared in both UAG- and UAA-ending transcripts in GRO.DD and GRO.DD.prfA+ (Figure 2, Supplementary file 1 – Table S1), consistent with previous reports that overexpressed proteins are targeted for degradation by the tmRNA (Baneyx and Mujacic, 2004; Li et al., 2007; Moore and Sauer, 2005; Tu et al., 1995). Both samples also contained the unmodified C-terminus of GFP (LEHHHHHH). In GRO.DD.prfA+, this is likely due to translational termination via RF1, while in GRO.DD this may represent rescue of nonstop ribosomes by ArfA/ArfB, release of nascent peptides undergoing translation at the time of cell lysis, or spontaneous dissociation of the ribosome, although this last event is estimated to occur fewer than once per 100,000 codon decoding events (Keiler and Feaga, 2014). While these were the only C-terminal fragments detected in GRO.DD expressing UAA-GFP and in GRO.DD.prfA+ expressing UAG-GFP, GRO.DD [pUAG-GFP] contained greater than 30 unique C-terminal sequences (Supplementary file 2).

The peptide fragments detected from GRO.DD [pUAG-GFP] demonstrate a combination of near-cognate suppression, ribosomal frameshifting, and tmRNA tagging (Figure 2B). We identified two previously known suppression events glutamine (Q) and tyrosine (Y) (Aerni et al., 2015; Lajoie et al., 2013b), and observed two new suppressors, aspartic acid (D) and valine (V). We detected ribosomal frameshifting of up to −3 (LEHHHHHHH) and +19 nucleotides (LEHHHHHHMVR), as determined by the presence of fragments from all three reading frames appended to the C-terminal peptide of LEHHHHHH. Additionally, the LEHHHHHHHH peptide may indicate a −6 frameshift, although it is impossible to determine whether this peptide arises from a −6 frameshift or two −3 frameshifts between histidine incorporation. We also detected peptides encoded as far downstream as +82 nucleotides after the UAG codon, illustrating that the ribosome can continue translation after encountering the unassigned UAG codon provided that stalling at the UAG codon is resolved. Lastly, we identified the modified ssrA DD-tag at both the site of the UAG codon and downstream on multiple peptides.

Prior research in vitro revealed that a mistranslation event increases the likelihood of subsequent mistranslation events and termination by release factor 2 (RF2) (Zaher and Green, 2009), and we investigated whether we could detect peptides representing such mistranslation events. Given the difficulty of distinguishing such peptides from suppression or frameshifting with one or two amino acids, we created a hypothetical peptide library (Supplementary file 1 – Table S2) containing all combinations of LEHHHHHHXXX, wherein X is any amino acid incorporated at the three residue positions directly downstream of the UAG codon (Supplementary file 4). The search with this library returned 23 unique peptides, 14 of which met our scoring threshold of 15 (Aerni et al., 2015). Five of these peptides (LEHHHHHHEKP, LEHHHHHHQLD, LEHHHHHHQQR, LEHHHHHHSLK, and LEHHHHHHYQR) could only arise from the mRNA transcript through two or more frameshift events after stalling at the UAG codon had already resolved (Supplementary file 1 – Table S2), suggesting they instead arise from loss of translational fidelity and spontaneous termination of translation following mistranslation at the UAG codon. We also had enough resolution in the data to manually verify the amino acid sequences of LEHHHHHHQQR and LEHHHHHHYQR, noting a 35 Da shift in mass between the Q and Y in the third position from the C-terminus.

Although several alternative hypotheses may explain these random tripeptides, these explanations are either incomplete or unlikely given our current understanding of prokaryotic translation. First, it is improbable that these fragments arose from routine errors in mRNA transcription because this would require at least two transcriptional errors in a nine-nucleotide span. The transcription error rate in E. coli is estimated to be ~1 in 10,000 bases (Blank et al., 1986; Rosenberger and Hilton, 1983) and our strains have no known mutations that would lead to greater error rates in transcription. Second, it is possible that ArfA or ArfB may have terminated translation in these peptides due to 3’ exonuclease shortening of the mRNA transcript as the ribosome is stalled at the UAG codon (Keiler and Feaga, 2014; Yamamoto et al., 2003). However, this does not explain the non-encoded tripeptides appended to the LEHHHHHH peptide. Lastly, the peptides LEHHHHHHQQR, LEHHHHHHSLK, and LEHHHHHHYQR may have been part of longer peptides that were cleaved off during trypsin digest. In this case, translation may have continued past the C-terminal R or K observed in these peptides, but this consideration would not apply to LEHHHHHHEKP and LEHHHHHHQLD and again does not explain the non-encoded tripeptide sequence observed appended to LEHHHHHH. Given this, we hypothesize that these five peptides result from loss of translational fidelity after stalling at the UAG codon that may lead to (1) spontaneous termination of translation due to the untemplated action of RF2 following mistranslation or (2) ArfA- or ArfB-mediated release predicated on 3’ exonuclease degradation of the mRNA. The rare event of spontaneous hydrolysis of the peptide from the ribosome is also possible.

ssrA and arfB mediate degradation of proteins containing unassigned UAG codons

Since mass spectrometry data indicated that a combination of mechanisms could resolve stalled translation at the unassigned UAG codon, we generated targeted deletions of the ribosomal rescue systems (ssrA, arfA, and arfB) in strains with wild-type ssrA sequence (GRO.AA) to determine whether protein production from UAG-ending transcripts in ΔRF1 cells could be restored to levels seen in +RF1 cells. Using recombineering (Sharan et al., 2009), we produced single and double deletions of the ssrA, arfA, and arfB genes that encode the ribosomal rescue systems. Efforts to generate a double deletion of ssrA and arfA failed (data not shown) because the resulting phenotype is synthetic lethal (Chadani et al., 2010). We transformed each deletion strain with the UAG-GFP construct under a highly expressing, inducible pLtetO promoter (Lutz and Bujard, 1997) and induced GFP expression for 20 hr, measuring the effect of protein expression on cellular growth through doubling time and maximum optical density at 600 nm (OD600) (Figure 3A and B, Supplementary file 1 – Table S3). To quantify protein expression, we then assayed whole-cell lysate from equal cell numbers, as determined by OD600, for abundance of protein via anti-GFP western blot alongside GFP standards of known concentration as described previously (Figure 3C, Figure 3—source data 6) (Pirman et al., 2015). We also included as positive controls (1) a wild-type strain (ECNR2) expressing the UAG-GFP construct and (2) GRO.AA expressing UAA-GFP.

Figure 3 with 1 supplement see all
Deletion of both ssrA and arfB restores protein production in the genomically recoded organism.

(A) Comparison of doubling times for WT and GRO strains carrying listed deletions with and without GFP induction. Error bars show standard deviation centered at mean, n = 3; data were analyzed using Source code 1 (Figure 3—source datas 1 and 2). (B) Change in maximum optical density at 600 nm (OD600) due to expression of UAG-GFP or UAA-GFP in wild-type (WT) and GRO strains carrying listed deletions. Error bars show standard deviation centered at mean, n = 3 (Figure 3—source datas 1 and 2). (C) Quantification of GFP abundance per 1 mL of cells at OD600 of 2.5 via western blot from biological replicates of indicated strains (Figure 3—source datas 36). Error bars show standard deviation centered at mean, n = 3 (Figure 3—source datas 35). See Figure 3—figure supplement 1 for linear calibration curves used to quantify GFP abundance for each replicate experiment. Image of representative western blot is below the graph. p-values are calculated in relation to the GRO containing the UAG-ending GFP (GRO – UAG) and are as follows: * is p≤0.05, ** is p≤0.01, *** is p≤0.001, and **** is p≤0.0001.

https://doi.org/10.7554/eLife.34878.009

Expression of UAG-GFP impaired GRO growth rate and cell density, generating a 54% increase in doubling time and 8% reduction in maximum OD600 compared to cells not expressing UAG-GFP, and a 25% greater doubling time and 14% lower maximum OD600 compared to cells expressing UAA-GFP. In contrast, ECNR2 exhibited only a 7% increase in doubling time and a 5% reduction in maximum OD600 when expressing UAG-GFP. Although deletion strains experienced reduced growth rate as measured by doubling time compared to the GRO.AA, they exhibited a less pronounced increase in doubling time when expressing UAG-GFP (increases in doubling time between 12% and 50%) as compared to the GRO.AA (54% increase in doubling time) (Figure 3A). However, deletion of ssrA reduced fitness during protein expression as measured by maximum OD600, with GRO.AA.∆ssrA demonstrating a 34% reduction in max OD600 and GRO.AA.∆ssrA.∆arfB demonstrating a 61% decrease in max OD600. This is potentially due to increased presence of misfolded or prematurely truncated peptides that are ordinarily tagged and degraded by the tmRNA. Interestingly, deletion of arfB produces a 50% increase in doubling time during protein expression, suggesting ArfB may play a role in ribosomal rescue during high levels of ribosomal stalling.

We then investigated the impact of unassigned codons on protein production using western blot densitometry, and found that the GRO expressing UAG-GFP produced less than one-fourth of the protein amount than does ECNR2 expressing UAG-GFP (Figure 3C, 8.0 µg/ml for the GRO versus 35 µg/ml for ECNR2, p=0.0014). GRO.AA expressing UAA-GFP produced nearly nine times more protein than did GRO.AA expressing UAG-GFP (68 µg/ml for GRO.AA [pUAA-GFP] versus 8.0 µg/ml for GRO.AA [pUAG-GFP], p<0.0001), indicating that the UAG codon in pUAG-GFP is the cause of reduced protein expression in the GRO. Deletion of ssrA in the UAG-GFP-expressing GRO partially restored protein production to levels seen in its UAA-GFP-expressing counterpart with no knockouts (31 µg/ml for GRO.AA.∆ssrA [pUAG-GFP] versus 68 µg/ml for GRO.AA [pUAA-GFP]) and deletion of both ssrA and arfB fully restored protein production (70. µg/ml). These ssrA deletion strains likely demonstrate increased GFP expression and reduced growth rate (Figure 3A) and cell density (Figure 3B) because translation of GFP transcripts sequesters cellular resources at the expense of cellular replication, producing GFP peptides that are freed from nonstop ribosomes via ArfA or ArfB without addition of a degradation tag.

A deletion of arfB leads to strikingly low- protein abundances from UAG-GFP transcripts that approach the lower limit of detection of our assay, although this apparent reduction in protein production was not statistically significant in comparison to protein production by GRO.AA [pUAG-GFP]. These ArfB deletion data, together with the fitness reduction observed in the GRO, suggest that ArfB is constitutively expressed and relieving low levels of ribosomal stalling in E. coli. These data also suggest that while deletion of ssrA partially recovers protein production from UAG-ending transcripts in the GRO, deletion of both ssrA and arfB is necessary to fully recover protein expression from UAG-ending transcripts to levels seen from the translation of UAA-ending transcripts in the GRO.

Deletion of ssrA restores conjugative plasmid propagation and viral infection in the GRO

To determine whether deletions of of ssrA or arfB could restore propagation of horizontally-transferred genetic elements in the GRO, we assessed conjugation efficiency and growth rate from plasmids RK2 and F on GRO strains with single and double deletions of ssrA, arfA, and arfB. Previous research indicates that the UAG stop codon in the trfA gene on RK2 leads to impaired conjugation efficiency and replication in the GRO (Ma and Isaacs, 2016), likely because the TrfA protein is required to initiate plasmid replication (Pansegrau et al., 1994). Phenotypically, this manifests as reduced efficiency of plasmid transfer in conjugation experiments and increased doubling times for RK2+ strains in media selecting for plasmid maintenance due to loss of plasmid and concomitant antibiotic resistance genes. We found that deletion of ssrA increased the ability of the GRO to both receive (Figure 4A, Supplementary file 1 – Table S4) and replicate RK2 (Figure 4B, Supplementary file 1 – Table S5). RK2 conjugation efficiency in GRO.AA.∆ssrA improved to 99% (compared to 87% in GRO.AA), and the strain showed an increase in doubling time of only 6% compared to a 28% increase for GRO.AA (p<0.0001). We observed similar results for GRO.AA.∆ssrA.∆arfB. However, single deletion of arfB halved RK2 conjugative efficiency (Figure 4A, p=0.0002). This strain also exhibited a 38% increase in doubling time when bearing RK2, compared to the 28% increase in doubling time seen in the GRO with no ribosomal rescue gene deletions (Figure 4B, p<0.0001).

Deleting ssrA restores propagation of both viruses and conjugative plasmids in the genomically recoded organism.

(A) Percent transfer of conjugative plasmid RK2 from a wild-type donor into wild-type (WT), GRO, or GRO with designated deletions (KO) as recipients (Figure 4—source data 1). Data are obtained from technical triplicates generated from a single biological sample. (B) Percent increase in doubling time for strains carrying plasmid RK2 compared to strains lacking RK2 (Figure 4—source datas 2 and 3). (C) Number of conjugation events for conjugative plasmid F from wild-type, GRO, or GRO with designated gene deletions as donors to a wild-type recipient (Figure 4—source data 4). Data are obtained from technical triplicates generated from a single biological sample. (D) Relative titer on wild-type, GRO, and GRO with designated deletions of phage λ (Figure 4—source data 5). Error bars show standard deviation centered at mean, n = 3. p-values are calculated in relation to the GRO condition and are as follows: * is p≤0.05, ** is p≤0.01, *** is p≤0.001, and **** is p≤0.0001. (E) Effects of sequential deletions of ribosomal rescue mechanisms on conjugative plasmid transfer efficiency. (F) Effects of sequential deletions of ribosomal rescue mechanisms on viral susceptibility.

https://doi.org/10.7554/eLife.34878.017

For plasmid F (Figure 4C, Supplementary file 1 – Table S6), which contains UAG-ending genes traY and traL that are essential for conjugation between cells (Ma and Isaacs, 2016), we found that deletion of ssrA increased conjugation events from the GRO donor 1,000-fold to 3.56 × 107 (p=0.0015) compared to GRO.AA (3.30 × 104 events), arfA deletion (3.41 × 104 events), and arfB deletion (3.47 × 104 events). GRO.AA.∆ssrA.∆arfB and GRO.AA.∆arfA.∆arfB exhibited 5.2- and 2.3-fold decrease in conjugative efficiency when compared to GRO.AA.∆ssrA and GRO.AA.∆arfA single deletion strains, respectively (p<0.01 for each, Figure 4C). These reductions in RK2 and F conjugative efficiency attributable to arfB deletion indicate that ArfB likely contributes to relief of nonstop ribosomes when encoded in its native ribosomal context, supporting evidence of ArfB’s ribosomal rescue activity previously validated in vitro (Handa et al., 2011) and when over-expressed in the absence of ssrA and arfA in vivo (Chadani et al., 2010). However, deletion of ssrA is sufficient to restore both conjugation and propagation of RK2 and F in the GRO. We next attempted infection with phage λ on our suite of deletion strains (Figure 4D, Supplementary file 1 – Table S7). Although deletion of arfA or arfB does not recover viral infection, deletion of the ssrA gene—either alone (p=0.0016) or alongside deletion of arfB (p<0.0001)—recovers λ infection of the GRO to levels similar to wild-type, with about 108 plaque forming units per mL (PFU/mL) (Figure 4D). These results demonstrate that removal of ssrA has the greatest influence in restoring conjugative plasmid transfer efficiency and viral susceptibility in the GRO (Figure 4E and F).

Discussion

In this study, we use a genomically recoded organism (GRO) containing an unassigned UAG codon as a model to investigate the molecular mechanisms that obstruct the propagation of HTGEs in organisms with alternative genetic codes. We demonstrate that unassigned stop codons elicit near-cognate suppression, frameshifting, and the action of ribosomal rescue mechanisms (Figure 2). tmRNA-mediated ribosomal rescue prompted by the unassigned codon results in the degradation of nascent peptides translated from UAG-ending transcripts and obstructs the propagation of HTGEs (Figure 3, Figure 4). Additionally, ssrA deletion strains exhibit both significantly increased UAG-GFP yields (Figure 3C) and recovered propagation of HTGEs (Figure 4), consistent with evidence that deletion of ssrA removes inhibition of ArfA production and releases nascent peptides from stalled ribosomes without degradation (Chadani et al., 2011; Garza-Sánchez et al., 2011; Schaub et al., 2012). Our GRO model thus sheds light on the functional significance of previously described regulatory relationships while elucidating the unique mechanistic contributions of different ribosomal rescue systems in resolving translation at unassigned stop codons. These mechanistic outcomes that occur as a consequence of ribosomal stalling could be further investigated via ribosomal profiling in future work.

The mass spectrometry data collected from our GRO model demonstrate the striking proclivity for the ribosome to undergo un-programmed frameshifting at unassigned stop codons and represents, to our knowledge, the first in vivo study to examine such frameshifting. Prior studies have revealed programmed ribosomal frameshifting from −4 to +50 nucleotides (Atkins et al., 2016; Baranov et al., 2015; Huang et al., 1988; Yan et al., 2015), but these studies focused on frameshifts programmed into mRNA transcripts through combinations of four mechanisms: (1) use of rare codons to slow translation speed at the skip site, (2) weak base pairing of the P-site tRNA anticodon and mRNA codon, (3) strong base pairing of the P-site tRNA anticodon to the location where the ribosome will re-bind the mRNA, and (4) a region six bases upstream of the re-binding site that mimics a Shine-Dalgarno sequence and offsets the energetic cost of frameshifting (Pech et al., 2010). Although the UAG codon in our GFP transcript slows translation, the P-site codon-anticodon pair for the codon immediately upstream of UAG is exact (CAC codon and GUGHis-tRNA anticodon) (Hsu et al., 1984) and any frameshift except backward would incur greater mispairing between the P site codon and anticodon. Additionally, no Shine Dalgarno-like sequence (AGGAGG) (Shine and Dalgarno, 1974; Vimberg et al., 2007) exists upstream, suggesting that the GFP construct we use contains only one of the four elements required for programmed ribosomal frameshifting (Supplementary file 1). From our construct, we observed frameshifts of potentially up to −6 and +19 nucleotides in response to the unassigned UAG codon (Figure 2, Supplementary file 1 – Tables S1 and S2). Collectively, our work uncovers a wide variety of frameshifting events that can occur in response to ribosomal stalling in vivo, highlighting the capacity of the ribosome to continue translation despite missing an essential translational component.

Mass spectrometry analysis also revealed truncated mistranslation products that possibly represent loss of translational fidelity and termination by RF2 downstream of an initial mistranslation event at the UAG codon, known as post-peptidyl transfer quality control (Petropoulos et al., 2014; Zaher and Green, 2009), a result previously only observed in vitro. Although prior studies decades ago revealed premature truncation products in vivo (Manley, 1978), they lacked the technical capability to determine whether these peptides arose from a single mistranslation event or demonstrated loss of translational fidelity after the ribosome encounters a rare or unassigned codon. The mistranslation products we detect show repeated mistranslation events that could not have been produced by suppression, ribosomal rescue, or frameshifting, unless the ribosome frameshifted multiple times after resolving stalling at the UAG codon (Figure 2B, Supplementary file 1). These events may be followed by ribosomal rescue via ArfA or ArfB, spontaneous ribosomal dissociation, or termination via release factor 2, though our technique was not capable of distinguishing between these fates. Previous in vitro studies using purified ribosome complexes determined that a mistranslation event destabilized the P-site helix, reducing the ability of the A-site to discriminate between anticodons and resulting in further mistranslation events and rapid termination by RF2 with the assistance of release factor 3 (Zaher and Green, 2009; Zaher and Green, 2010). The researchers predicted that a single mistranslation event would also lead to prematurely truncated peptides with two or three miscoded C-terminal amino acids appended in vivo (Zaher and Green, 2009). These findings, together with our results, motivate future work to investigate the possibility of loss of translational fidelity after an initial translation error and highlight the GRO as a model for elucidating translational fidelity in vivo.

The GRO demonstrates that general ribosomal rescue mechanisms resolve ribosomal stalling at unassigned stop codons. As most sequenced bacterial species contain a homolog of the tmRNA, ArfA, or ArfB ribosomal rescue systems (Hudson et al., 2014; Keiler, 2015) and eukaryotic cells contain analogous pathways that rescue stalled ribosomes (Graille and Séraphin, 2012), we anticipate that translational stalling at unassigned codons can be resolved similarly in these organisms. Accordingly, we hypothesize that organisms beyond E. coli should tolerate unassigned codons as intermediates toward codon reassignments in genomic recoding, efforts for which are underway in numerous prokaryotic and eukaryotic species (Lau et al., 2017; Napolitano et al., 2016; Ostrov et al., 2016; Richardson et al., 2017). Additional barriers to codon reassignment exist, such as regulatory roles of codons in gene expression (Lajoie et al., 2013a), but our findings indicate that unassigned codons are tolerable in the absence of specialized translational machinery to address them, both as intermediate steps towards codon reassignment and as permanent parts of the genetic code.

Our findings suggest that we can use unassigned codons to engineer organisms with broad resistance to HTGEs and impart genetic isolation, increasing engineered organisms’ stability in biotechnology applications. Since tmRNA homologs are found in >99% of all sequenced bacterial genomes (Hudson et al., 2014; Keiler, 2015), we would expect other organisms engineered to contain unassigned codons to exhibit immunity to horizontally transferred genetic elements. As researchers pursue further efforts in whole genome recoding (Boeke et al., 2016; Lau et al., 2017; Napolitano et al., 2016; Ostrov et al., 2016; Richardson et al., 2017) and engineer organisms for use in open environments, we require strategies to genetically isolate such organisms from their surrounding environment to ensure robust function, both individually (Moe-Behrens et al., 2013) and as members of microbial communities (Grosskopf and Soyer, 2014; Hillesland and Stahl, 2010). Genomically recoded organisms with unassigned codons would possess reduced susceptibility to exploitation by HTGEs, increasing their stability in open environments. Although this work demonstrates that an unassigned stop codon acts as a barrier to HGT, this current barrier can be breached by mutation or deletion of the tmRNA to produce a functional protein. In contrast, we expect that an organism with an unassigned sense codon would have even greater barriers to HGT, as premature termination at an unassigned sense codon would likely produce a nonfunctional, truncated peptide. We thus anticipate that further genomic recoding to engineer additional unassigned sense and nonsense codons may be a broadly applicable strategy to confer genetic isolation in living systems, facilitating the safe use of engineered organisms in complex open environments.

Materials and methods

Key resources table
Genetic reagents, bacterial strains, antibodies, and software used in this study.
https://doi.org/10.7554/eLife.34878.023
Reagent
type (species)
or resource
DesignationSource or
reference
IdentifiersAdditional
information
Isaacs
Lab
Reference
#
Full
genotype
of strains
# UAG
Codons
RF1
status
Ribosomal
rescue
gene
knockout
ssrA
tag
status
Gene
(Escherichia
coli)
pUAG-GFPthis papereGFP-6xHis
-UAG; Plasmid
NJM88;
Strain
NJM1242
eGFP protein
with a C-terminal
6-His tag for protein
purification,
terminating
translation in a
UAG codon.
Plasmid
NJM88;
Strain
NJM1242
N/AN/AN/AN/AN/A
Gene
(E. coli)
pUAA-GFPthis papereGFP-6xHis
-UAA; Plasmid
NJM89;
Strain
NJM1249
eGFP protein with
a C-terminal 6-His
tag for protein
purification,
terminating
translation in a
UAA codon.
Plasmid
NJM89;
Strain
NJM1249
N/AN/AN/AN/AN/A
Genetic
reagent
(E. coli)
RK2410.1126/science
.1205822;
10.1016/j.cels
.2016.06.009
pRK24;
Strain NJM699
Conjugative RK2
plasmid (10.1006/
jmbi.1994.1404),
but lacks functional
AmpR gene.
Strain
NJM699
N/AN/AN/AN/AN/A
Genetic
reagent
(E. coli)
FYale University
Coli Genetic
Stock Center
(CGSC),
Strain #4401
pF; Strain
EMG2; Strain
CGSC#4401;
Strain
NJM426;
Strain
NJM473
Conjugative F
plasmid, as
described by
PMID: 4568763.
Obtained from
the Yale CGSC.
Strain
NJM426;
Strain
NJM473
N/AN/AN/AN/AN/A
Genetic
reagent
(E. coli)
pZE21_
UAG-GFP
this paperpZEtR-eGFP
-cHis-TAG-
v02; Plasmid
NJM88;
Strain
NJM1242
pZE21 plasmid
with pLtetO
promoter driving
inducible expression
of eGFP with a
C-terminal 6-His
tag and terminating
in UAG codon.
Inducible with
anhydro-tetracycline.
Plasmid
NJM88;
Strain
NJM1242
N/AN/AN/AN/AN/A
Genetic
reagent
(E. coli)
pZE21_
UAA-GFP
this paperpZEtR-eGFP
-cHis-TAA-v02
; Plasmid
NJM89;
Strain
NJM1249
pZE21 plasmid
with pLtetO
promoter driving
inducible expression
of eGFP with a
C-terminal 6-His
tag and terminating
in UAA codon.
Inducible with
anhydro-tetracy
cline.
Plasmid
NJM89;
Strain
NJM1249
N/AN/AN/AN/AN/A
Genetic
reagent
(Enteroba
cteria
phage λ)
λ.CI857Coli Genetic
Stock Center
(CGSC), Yale
University
(contact John
Wertz directly)
λ.CI857; λ
phage;
Phage NJM102
Phage λ with
temperature-
sensitive CI
repressor gene;
when incubated
at 37° C, phage
becomes obligate
lytic
Phage
NJM102
N/AN/AN/AN/AN/A
Cell line
(E.
coli)
GRO.DDthis paperC31GIB.
tmRNA-DD;
Strain #987
MG1655-derived
strain with all 321
UAG codons
mutated to UAA,
deletion of RF1,
and tmRNA tag
C-terminal amino
acids mutated from
AA to DD. Retains
lambda red cassette
for recombineering.
Investigated in
Figure 2.
Strain
#987
ΔmutS:zeo.
Δ(ybhB-
bioAB)
:[λcI857.
Δ(cro-ea59)
:tetR-bla].
ΔprfA.ΔtolC
.tmRNADD
0+RF1n/aDD
Cell line
(E. coli)
GRO.
DD.prfA+
this paperC31GIB.
prfA+.tmRNA
-DD; Strain
#996
MG1655-derived
strain with all 321
UAG codons
mutated to UAA,
retains RF1,
and tmRNA tag
C-terminal amino
acids mutated from
AA to DD. Retains
lambda red cassette
for recombineering.
Investigated in
Figure 2.
Strain
#996
ΔmutS:zeo.
Δ(ybhB-
bioAB)
:[λcI857.
Δ(cro-ea59):
tetR-bla].
ΔtolC.tm
RNADD
0∆RF1n/aDD
Cell line
(E. coli)
ECNR210.1016/j.cels
.2016.06.009
ECNR2.Δmut
S:zeocin.Δ
λRed; Strain
#795
MG1655-derived
strain that contains
321 UAG codons
and retains RF1.
Investigated in
Figures 3 and 4.
Strain
#795
ΔmutS:zeo321+RF1n/aAA
Cell line
(E. coli)
GRO.AA10.1016/j.cels
.2016.06.009
C31.final.
ΔmutS:
zeocin.ΔprfA
.ΔλRed;
Strain #796
MG1655-derived
strain with all 321
UAG codons
mutated to UAA,
deletion of RF1.
Investigated in
Figures 3 and 4.
Strain
#796
ΔmutS:
zeo.ΔprfA
(GenBank
ID:
CP006698)
0∆RF1n/aAA
Cell line
(E. coli)
GRO.
AA.∆arfB
this paperC31GIB.arfB:
tolCorf.
ΔλRed;
Strain #1230
MG1655-derived
strain with all 321
UAG codons
mutated to UAA,
deletion of RF1,
and deletion of arfB.
Investigated in
Figures 3 and 4.
Strain
#1230
ΔmutS:
zeo.ΔprfA
.arfB:tolC
0∆RF1∆ssrAAA
Cell line
(E. coli)
GRO.
AA.∆ssrA
this paperC31GIB.ssrA
:tolC.ΔλRed;
Strain #1231
MG1655-derived
strain with all 321
UAG codons
mutated to UAA,
deletion of RF1,
and deletion
of ssrA.
Investigated in
Figures 3 and 4.
Strain
#1231
ΔmutS:
zeo.ΔprfA.
ssrA:tolC
0∆RF1∆arfAAA
Cell line
(E. coli)
GRO.
AA.∆arfA
this paperC31GIB.arfA
:tolC.ΔλRed
; Strain #1232
MG1655-derived
strain with all
321 UAG codons
mutated to UAA,
deletion of RF1,
and deletion of
arfA. Investigated
in Figures 3 and 4.
Strain
#1232
ΔmutS:
zeo.ΔprfA.
arfA:tolC
0∆RF1∆arfBAA
Cell line
(E. coli)
GRO.AA
.∆ssrA.∆arfB
this paperC31GIB.ΔarfB
.ssrA:tolC.Δ
λRed; Strain
#1233
MG1655-derived
strain with all
321 UAG codons
mutated to UAA,
deletion of RF1,
and deletion of
ssrA and arfB.
Investigated in
Figures 3 and 4.
Strain
#1233
ΔmutS:
zeo.ΔprfA
.ΔarfB.ssrA:tolC
0∆RF1∆ssrA.
∆arfB
AA
Cell line
(E. coli)
GRO.AA
.∆arfA.
∆arfB
this paperC31GIB.Δarf
B.arfA:tolC.
ΔλRed;
Strain #1234
MG1655-derived
strain with all
321 UAG codons
mutated to UAA,
deletion of RF1,
and deletion of
arfA and arfB.
Investigated in
Figures 3 and 4.
Strain
#1234
ΔmutS:
zeo.ΔprfA
.ΔarfB.arfA
:tolC
0∆RF1∆arfA.
∆arfB
AA
Antibodymouse
anti-GFP
antibody
otherInvitrogen
(Ref#: 332600,
Lot#:
1513862A)
Invitrogen
(Ref#: 332600,
Lot#: 1513862A);
(5.5 μL antibody
in 3 mL Milk
 + TBST)
N/AN/AN/AN/AN/AN/A
Antibodygoat
anti-mouse
antibody
otherAbCam (Ref#:
ab7023, Lot#:
GR157827-1)
AbCam (Ref#:
ab7023, Lot#:
GR157827-1);
(2.2 μL antibody
in 10 mL Milk
 + TBST)
N/AN/AN/AN/AN/AN/A
Recombinant
DNA reagent
ssrA:tolCthis paper; for
use, see
tolC positive
/negative
selection in
10.1038/nprot
.2014.081
dsDNA
NJM111
The E. coli native
tolC gene used to
delete ssrA gene via
recombineering
(10.1038/nprot.
2008.227).
dsDNA
NJM111
N/AN/AN/AN/AN/A
Recombinant
DNA reagent
arfA:tolCthis paper; for
use, see
tolC positive
/negative
selection in
10.1038/nprot
.2014.081
dsDNA
NJM112
The E. coli native
tolC gene used to
delete arfA gene
via recombineering
(10.1038/nprot.
2008.227).
dsDNA
NJM112
N/AN/AN/AN/AN/A
Recombinant
DNA reagent
arfB:tolCthis paper; for
use, see
tolC positive
/negative
selection in
10.1038/nprot
.2014.081
dsDNA
NJM113
The E. coli native
tolC gene used to
delete arfB gene
via recombineering
(10.1038/nprot.
2008.227).
dsDNA
NJM113
N/AN/AN/AN/AN/A
Software,
algorithm
Doubling
time
algorithm
10.1126/
science.1241459
Growth_
Analyze_
GK.m
Doubling time
used in 10.1126
/science.1241459,
written by Gleb
Kuznetsov in the
lab of Dr. George
Church.
N/AN/AN/AN/AN/AN/A
Software,
algorithm
MaxQuant
v1.5.1.2
otherN/ACommercial
software for
mass
spectrometry
analysis.
N/AN/AN/AN/AN/AN/A
Software,
algorithm
Graphpad
Prism 7
otherN/ACommercial
software for
statistical
analysis and
graphing,
provided
through Yale
University.
N/AN/AN/AN/AN/AN/A

Strains and media

All bacteria used in this study are derived from E. coli ECNR2, which is in turn derived from E. coli MG1655 (GenBank ID: U00096) in which mutS is replaced by a zeocin resistance cassette (Wang et al., 2009; Lajoie et al., 2013b). Additionally, the native bioAB genes found in MG1655 are replaced by the lambda red cassette in ECNR2. This strain is designated ECNR2.AA (see Table 1 for full genotype). For experiments expressing UAG-GFP and UAA-GFP for mass spectrometry, strains with all 321 UAG codons changed to UAA (designated ‘GRO’ strains) were used to control for potential differences in protein expression arising from these mutations (GenBank ID for GRO.AA: CP006698). For all other experiments, control strains labeled wild-type (WT) are MG1655 derivatives retaining all 321 UAG codons. All deletions of ssrA, arfA, and arfB were generated with a tolC resistance cassette via recombineering (Sharan et al., 2009). Modification of the ssrA tag from AANDENYALAA to AANDENYALDD (AA->DD) to increase stability of tagged proteins was performed with MAGE as described previously (Gallagher et al., 2014; Wang et al., 2009). All modifications to strains made in this study were validated through Sanger sequencing (GeneWiz; South Plainfield, NJ).

We performed all protein expression assays and conjugation assays in LB Lennox at pH 7.5. We performed all phage assays in Tryptone-KCl (TK) media as described previously (Jaschke et al., 2012; Ma and Isaacs, 2016; Valentine et al., 2002).

Phages and plasmids

For viral relative titers, we used phage λ cI857 obtained from Dr. John Wertz at the Yale Coli Genetic Stock Center (CGSC) because it is obligately lytic at 37°C, preventing possible confounding factors from lysogeny. We used the conjugative plasmid RK2 described in Isaacs et al. (2011), which is a derivative of the RK2 plasmid described in Pansegrau et al. (1994) carrying blaR instead of kanR. The complete nucleotide sequence for the plasmid is available in NCBI database, Accession L27758.1 and GI 508311. We obtained the F plasmid from the Yale CGSC (NCBI Accession AP001918.1, GI: 8918823) and added KanR from plasmid pZE21 for antibiotic selection.

To create the UAG-GFP and UAA-GFP constructs for protein expression, we cloned an eGFP construct with a C-terminal 6xHis tag downstream of pLtetO into a modified pZE21 vector with kanamycin resistance (kanR)carrying a copy of the tet repressor gene (tetR) to prevent leaked gene expression. We then modified the stop codon of the eGFP construct to end in either a UAG or UAA stop codon.

Protein expression and purification

To obtain GFP for analysis via mass spectrometry, we transformed UAG-GFP and UAA-GFP constructs into wild-type and GRO strains carrying the AA->DD modification in the ssrA tag to prolong the half-life of tagged peptides. Experiments in the absence of the AA->DD modification yielded no peptides with ssrA degradation tags (data not shown). We then grew 50 mL cultures of each strain at 33°C in LB Lennox with 30 μg/mL kanamycin to an OD600 of 1.0 and induced protein expression with the addition of 30 ng/uL anhydrotetracycline (aTC). After incubation overnight, we pelleted cells and resuspended them in sterile phosphate buffer solution, then lysed cells via sonication. Cell debris was then pelleted by centrifugation and GFP purified from supernatant via a nickel resin affinity column. To concentrate protein and exchange buffer for subsequent trypsin digest, we then concentrated GFP via Millipore Amicon spin columns.

For whole western blots on whole cell lysates, we transformed UAG-GFP and UAA-GFP constructs into wild-type, GRO, and GRO strains with deletions of the ribosomal rescue systems. We then grew 5 mL cultures of each strain at 33°C in LB Lennox with kanamycin overnight, then diluted all cultures OD600 of 0.15 in fresh media containing 30 μg/mL kanamycin and 30 ng/uL aTC for 20 hr. To quantify protein expression and compare across strains, we normalized the OD600 of all cultures to 2.5 and pelleted 1 mL of this culture, which we placed in the −80C for 2 hr. We then re-suspended cell pellets in lysis buffer described previously (Aerni et al., 2015), incubated for 10 min on ice, centrifuged lysate, and ran 1:10 dilutions of resulting supernatant on gels for western blot analysis. Overnight starter cultures were diluted to an OD600 of 0.15 into three separate culture tubes, and cells within each tube were induced in parallel for GFP expression. GFP was purified from each of these cultures in parallel.

Mass spectrometry and proteomic analysis

Trypsin digest, sample preparation for mass spectrometry, and liquid chromatography elution gradients were performed as described previously (Aerni et al., 2015). Desalted peptides were injected onto a 75 μm ID PicoFrit column (New Objective) packed to 50 cm in length with 1.9 μm ReproSil-Pur 120 Å C18-AQ (Dr. Maisch). Samples were eluted over a 90 min gradient using an EASY-nLC 1000 UPLC (Thermo) paired with a Q Exactive Plus (Thermo), using the following parameters: (MS1) 70,000 resolution, 3 × 106 AGC target, 300–1700 m/z scan range; (MS2) 17,500 resolution, 1 × 106 AGC target, top 10 mode, 1.6 m/z isolation window, 27 normalized collision energy, 90 s dynamic exclusion, unassigned and +1 charge exclusion. Peptide identification from collected spectra was performed using MaxQuant v1.5.1.2 (Cox and Mann, 2008). Samples were searched using custom databases representing potential translational outcomes in response to the UAG codon within the GFP reporter construct (Supplementary file 3 and 4), as well as the E. coli proteome (EcoCyc K-12 MG1655 v17). The searches considered carbamidomethyl (Cys) as a fixed modification and the following variable modifications: acetyl (N-terminal), oxidation (Met), deamidation (Asn, Gln), and phosphorylation (Ser/Thr/Tyr). Discovered peptides had a minimum length of five amino acids and could contain up to three trypsin miscleavage events. A 1% false discovery rate was used. The mass spectrometry proteomics data and the custom search databases have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository (Vizcaíno et al., 2014) with the dataset identifier PXD009643. Mass spectrometry spectra were manually validated by identifying all spectra with an MS/MS score over 15 and verifying the presence sufficient b- and/or y-ion series.

Western blot experiments and analysis

Western blots were run as described previously using SDS-PAGE gels (Pirman et al., 2015). We ran GFP-6xHis standards of known amount (1, 10, 50, and 100 ng) alongside experimental samples and used these standards to generate linear-range calibration curves to quantify protein abundance in experimental samples (Figure 3—figure supplement 1). Because the antibody signal appeared sublinear in the 0–10 ng regime when we performed linear regression using all standards, we generated separate linear fits using the 1–10 ng standards and the 10–100 ng standards. We then determined experimental sample concentrations using these linear approximations. 20 of the 24 experimental samples quantified fell within or slightly above the 10–100 ng range (with the highest-intensity sample quantified as 136 ng), and 3 of the 24 samples fell within the 1–10 ng range. The one remaining sample, which had a weaker intensity than that of the 1 ng standard, was quantified through a linear approximation between the intensity of the 1 ng sample and of a blank lane with an assumed intensity of zero.

We expressed GFP-6xHis as described above, normalized cell cultures to an OD600 of 2.5, and lysed cells using BugBuster protein extraction reagent (Merck, Darmstadt, Germany). We then ran 10 µl of 1/150 diluted lysate per lane of the SDS-PAGE gel. We obtained primary mouse anti-GFP antibody from Invitrogen (Ref#: 332600, Lot#: 1513862A; RRID:AB_2234927) and goat anti-mouse antibody from AbCam (Ref#: ab7023, Lot#: GR157827-1; RRID:AB_955413). Western blots were developed using Bio-Rad Clarity Western ECL Blotting Substrate and Imaged on a GE Amersham Imager 600. We performed quantification of western blot bands as described previously (Pirman et al., 2015). We repeated three western blots in parallel for each strain induced in separate culture tubes (i.e. biological triplicates, see Protein expression and purification).

Viral relative titers

To quantify relative titers, we mixed 100-fold dilutions of phage with 300 µL of mid-log (OD600 = 0.5) cells in 3 mL of TK soft agar and poured onto TK solid agar plates. Starter cultures of cells were diluted to an OD600 of 0.5 into three separate culture tubes, and cells within each tube were infected with phage lambda in parallel (i.e. biological triplicate). Each tube was plated on a separate TK solid agar plate. We incubated plates overnight at 37°C, and counted plaques the next day.

Quantifying conjugation

We used conjugation conditions described previously (Ma and Isaacs, 2016; Ma et al., 2014). Briefly, we grew cultures of donor and recipient cells to late log in antibiotics selecting for plasmid or recipient and then rinsed and re-suspended in media to remove antibiotics. After concentrating cells to an OD600 of 20, we mixed donors and recipients in 1:1 ratio and spotted onto pre-warmed LB Lennox agar plates in 2 × 20 uL and 6 × 10 uL pattern. For F, we incubated plates at 37°C for 2 hr, then rinsed cells off plate, diluted serially 10-fold, and plated serial dilutions on plates containing antibiotic selecting for conjugants and incubated overnight at 37°C. For RK2, we incubated plates at 37°C for 1 hr, then plated on agar plates selecting for the recipient. To quantify the rate of transfer, we then picked 86 colonies from plates selecting for the recipient strain and patched them onto plates selecting for both recipient and conjugative plasmid, incubated plates overnight at 37°C, and counted the number of patched colonies that grew. After the conjugation, colonies were plated three times to generate technical triplicates.

Statistical and data analysis

We performed all t-tests and one-way ANOVA tests for statistical significance in GraphPad Prism 7. We calculated doubling times and maximum OD600 values from growth curve data using MATLAB (Newton, MA) code that we generated (Source code 1).

Experimental replicates

We used the definitions for biological and technical replicates outlined in Blainey et al., 2014. Biological replicates consist of parallel measurements of different biological samples subjected to the same experiment, and technical replicates are parallel measurements of a single biological sample subjected to experimentation. Data represented in (Figures 3, 4B and D) are biological replicates; data represented in (Figure 4A and C) are technical replicates. Data for all 96-well plate assays (Figures 3A, B and 4B) were obtained as biological replicates: One well of each sample was grown overnight as a starter culture in a 96-well plate. Starter cultures were then inoculated into three separate wells in a separate 96-well plate.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
    Structure of an Escherichia coli tRNA operon containing linked genes for arginine, histidine, leucine, and proline tRNAs
    1. LM Hsu
    2. HJ Klee
    3. J Zagorski
    4. MJ Fournier
    (1984)
    Journal of Bacteriology 158:934–942.
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
    The E Site and Its Importance for Improving Accuracy and Preventing Frameshifts
    1. M Pech
    2. O Vesper
    3. H Yamamoto
    4. DN Wilson
    5. KH Nierhaus
    (2010)
    In: J. F Atkins, R. F Gesteland, editors. Recoding: Expansion of Decoding Rules Enriches Gene Expression. New York: Springer. pp. 345–362.
    https://doi.org/10.1007/978-0-387-89382-2_16
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73

Decision letter

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Organisms with alternative genetic codes resolve unassigned codons via mistranslation and ribosomal rescue" for consideration by eLife. Your article has been reviewed by Gisela Storz as the Senior Editor, a Reviewing Editor, and three reviewers. The following individuals involved in review of your submission have agreed to reveal their identity: Kenneth Keiler (Reviewer #1); Yitzhak Pilpel (Reviewer #2); Alexander Mankin (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

In this manuscript, Ma et al. investigate how bacteria contend with unassigned codons by expressing genes with a UAG stop codon in a genetically recoded E. coli strain lacking UAG codons and RF1. This question is important because there are a number of examples in nature of codon reassignment, but it is not clear how organisms would make the transition from a canonical code if there is a large penalty for having an unassigned codon. It has also been suggested that reassignment might limit transfer of foreign DNA. The authors used mass spectrometry to identify the proteins produced from a reporter gene with a UAG codon and find that multiple mechanisms help cells deal with the unassigned codon, including suppression, frameshifting, and general mistranslation. The authors also tested the effect of inactivation of three ribosome rescue systems (tmRNA, ArfA and ArfB) on the reporter expression and on propagation of conjugative plasmids or phages that have genes terminating in UAG. These data suggested that tmRNA accounts for the resistance of their E. coli mutant strain to horizontal gene transfer.

This is an interesting study that expands our knowledge of how cells deal with difficult-to-translate codons. It is also useful for the future use of genetically recoded organisms.

Essential revisions:

1) The manuscript consistently refers to unassigned codons, implying that the work addresses this general issue. For example, in the Introduction: "Our work reveals mechanistic details into how cells rescue ribosomes stalled at unassigned codons". However, the effects observed when ribosome rescue pathways are deleted are almost certainly unique to unassigned stop codons. Trans-translation activity on ribosomes stalled at unassigned stop codons will remove the protein, but when ssrA is deleted, ArfA will release a complete polypeptide identical to what would be produced by RF1. If a sense codon were unassigned, neither trans-translation nor ArfA activity could produce active proteins. The authors should either provide an explanation for why the ribosome rescue pathways would have the same impact on unassigned codons within a gene, or they should clarify that their interpretation is restricted to unassigned stop codons.

2) For the experiment in Figure 2, the basis for some of the assignments are unclear. For example, how is it known that LEHHHHHHMVR results from a +19 skip instead of from loss of fidelity like LEHHHHHHYQR? The presence of two additional His residues in the His-tail of the GFP-His6 reporter may indeed indicate a -6 frameshift, as the authors propose, but may also indicate two consecutive -3 frameshifting events. Authors do not discuss the second possibility and suggest that they have detected "the furthest frameshift backward". Without strong evidence that they are indeed dealing with a -6 frameshift, instead of two -3 frameshifts, this seems to be an overstatement. Some of the peptides could also result from transcription errors rather than translation mistakes. A more quantitative summary of the types of peptides found in the mass spec, including the types of peptides found in the strain expressing the UAA-containing construct, would be very useful.

3) Were the experiments in Figure 3 and Figure 4 done with strains containing ssrA-DD (as in Figure 2) or wild-type ssrA? This is critical for the interpretation. All strains and plasmids need to be described at a level of detail that would allow others to reproduce the work.

4) Figure 3A is completely not clear and perhaps even misleading since it shows the effect of the different mutations on the ratio between the expressed and non-expressed constructs. It is hard to know whether the effects shown are due to effects of the mutations on the expressed or non-expressed construct. To resolve this issue, the authors should show the effects of the different mutations on the actual max OD600 and doubling time values of each strain separately (i.e. not just on the ratio).

5) The results shown for some of the strains (Figure 3) are surprising and not expected (for example, the large decrease in max OD600 ratio of the ssrA arfB double mutant – Figure 3A, the opposite trend in GFP values shown for the ssrA arfB compared to arfB and ssrA alone – Figure 3C etc.). There is no clear explanation to these peculiarities in the text. A more elaborate discussion in the context of epistatic interactions shown is needed, such a discussion could address the effects of double mutations compare to single ones. Likewise, the authors should try to comment on the observation that knockout of arfA increases GFP production (Figure 3C), whereas ArfA is expected to facilitate termination of the GFP translation at the UAG codon.

6) In Figure 4B, the authors interpreted the change in doubling time as an indication for the ability of RK2 plasmid to replicate – some explanation to this should be added to text.

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

Thank you for resubmitting your work entitled "Organisms with alternative genetic codes resolve unassigned codons via mistranslation and ribosomal rescue" for further consideration at eLife. Your revised article has been favorably evaluated by Gisela Storz (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) For the experiments in Figure 3 and Figure 4, explain whether the replicates are biological replicates or technical replicates and at what step the replication occurred. From the source data, it appears that all of the growth data is from a single 96 well plate. Were replicate wells in the plate inoculated from independent overnight cultures or from the same culture? This distinction is important to determine if the small changes observed are likely to be physiologically relevant. It is confusing that there are 3 plates of data in the source file, but data in the figure all seems to come from a single plate. Likewise, for the conjugation experiments, do the data come from one mating that was plated three times, from three matings made with the same cultures, or from three matings performed with independently grown cultures?

2) Figure 3C shows negative protein concentration, and the explanation is that the band was quantified, and the value was plugged into a formula from a standard curve, giving a negative result. Because it is physically impossible to have negative protein concentration and there is quite clearly a band on the blot, the explanation provided suggests that the standard curve was not accurate. From the source data, it appears that the standard curves are derived from two points at 1 ng and 100 ng fit to a line. Unfortunately, almost all of the samples are outside the 1-100 ng range. It appears the negative value is the result of a large negative number from blot 3 which also seems to have an anomalously low value for 100 ng in the standard curve. The differences in protein production from the different samples are clear, but the quantification is clearly inaccurate. This problem could potentially be addressed by running another replicate of the experiment or reporting the data normalized to one of the samples such as GRO.AA [pUAG-GFP]. We will not publish a negative concentration.

Text clarifications:a) In subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons”: Spontaneous termination of translation could refer to untemplated termination by RF2 or to spontaneous (nonenzymatic) hydrolysis of the peptidyl-tRNA.

b) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: Fitness is best used in relation to competitive growth experiments because it is possible for a strain to grow to a lower OD600 in monoculture but outcompete other strains and therefore have higher fitness. It would be clearer here to refer to growth rate or OD600 instead of fitness.

c) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: What is the knockout of arfB being compared to here? In Figure 3A, the induced/uninduced ratio for the arfB strain looks very similar to that for the isogenic arfA+ strain, which would seem to suggest that ArfB does not play a role.

d) The sentence in the last paragraph of subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons” is unclear: "Interestingly, a single knockout of arfB significantly reduced production of protein from UAG-GFP to low levels similar to those mapped to quantified GFP standards."

e) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”, the comparison for "fully restore protein expression from UAG-ending transcripts" appears to be GRO.AA[pUAA-GFP] instead of ECNR2[pUAG-GFP], but "restore" suggests it should be the latter. The comparison and the meaning behind which strain is used for the comparison should be clarified.

f) In subsection “Deletion of ssrA restores conjugative plasmid propagation and viral infection in the GRO”, it is not clear what comparison was used for the 2.4-fold increase in doubling time. The graph in Figure 4B shows a 38% increase for the arfA strain versus 28% for the isogenic wild type.

g) Discussion section: What is the evidence supporting the demonstration of ribosome stalling? It seems that ribosome stalling was assumed based on the addition of the SsrA tag. Experiments such as ribosome profiling could demonstrate ribosome stalling, but these were not done. I think stalling is part of the model here but has not been demonstrated.

h) Discussion section: Similar to (g) above, the regulatory relationship between tmRNA and ArfA was used to explain the data, so it would be circular reasoning to then use this explanation to validate the regulatory relationship.

i) Discussion section: "extensive" suggests there is a large amount of frameshifting, but the frameshifting events cannot be quantified using the techniques in this work. Perhaps something like "a wide variety of frameshifting events" would be more accurate.

j) Figure 1: The cartoon shows the SsrA-tagged protein going into the protease N terminus first, but all the proteases recognize the tag and start at the C terminus (this is not critical). In the legend, the word "hypothesized" should be removed from the last sentence – the lack of modification has been observed.

k) Figure 3 legend: For panels A and B, the legend does not match the labels. A is doubling time and B is max OD.

https://doi.org/10.7554/eLife.34878.033

Author response

In this manuscript, Ma et al. investigate how bacteria contend with unassigned codons by expressing genes with a UAG stop codon in a genetically recoded E. coli strain lacking UAG codons and RF1. This question is important because there are a number of examples in nature of codon reassignment, but it is not clear how organisms would make the transition from a canonical code if there is a large penalty for having an unassigned codon. It has also been suggested that reassignment might limit transfer of foreign DNA. The authors used mass spectrometry to identify the proteins produced from a reporter gene with a UAG codon and find that multiple mechanisms help cells deal with the unassigned codon, including suppression, frameshifting, and general mistranslation. The authors also tested the effect of inactivation of three ribosome rescue systems (tmRNA, ArfA and ArfB) on the reporter expression and on propagation of conjugative plasmids or phages that have genes terminating in UAG. These data suggested that tmRNA accounts for the resistance of their E. coli mutant strain to horizontal gene transfer.

This is an interesting study that expands our knowledge of how cells deal with difficult-to-translate codons. It is also useful for the future use of genetically recoded organisms.

We thank the reviewers for recognizing the significance of our work and relevance to both the translation of rare codons found in nature and to the translation of codons engineered into genomically recoded organisms.

Essential revisions:

1) The manuscript consistently refers to unassigned codons, implying that the work addresses this general issue. For example, in the Introduction: "Our work reveals mechanistic details into how cells rescue ribosomes stalled at unassigned codons". However, the effects observed when ribosome rescue pathways are deleted are almost certainly unique to unassigned stop codons. Trans-translation activity on ribosomes stalled at unassigned stop codons will remove the protein, but when ssrA is deleted, ArfA will release a complete polypeptide identical to what would be produced by RF1. If a sense codon were unassigned, neither trans-translation nor ArfA activity could produce active proteins. The authors should either provide an explanation for why the ribosome rescue pathways would have the same impact on unassigned codons within a gene, or they should clarify that their interpretation is restricted to unassigned stop codons.

It is true that our work is limited to unassigned stop codons and we agree that ribosomal rescue through trans-translation or ArfA at an unassigned sense codon would likely produce a truncated, nonfunctional protein and think this greatly supports our argument made in the Discussion section that further modifying the genetic code in engineered organisms could lead to even greater barriers to horizontal gene transfer. However, we also recognize prior literature that reports that these ribosomal rescue mechanisms would not be restricted to unassigned stop codons. As seen previously, these ribosomal rescue mechanisms are active at rare sense codons (Hayes, Bose and Sauer, 2002) and during periods of amino acid starvation (George et al., 2016; Li et al., 2007; Roche and Sauer, 1999), as we note in the Discussion section, suggesting they would also rescue stalled ribosomes at unassigned sense codons. In this manuscript, we empirically examine this in the context of an unassigned stop codon for two reasons: (1) when ssrA is deleted, fully-functional protein is made that is detectable both by western blot for GFP and phenotypic rescue for conjugative plasmids and viruses, which simplifies study of the molecular mechanism and (2) our model remains the only organism to date engineered to have an unassigned codon. Given that recoding of sense codons is actively underway (Napolitano et al., 2016; Ostrov et al., 2016; Lau et al., 2017), such GROs with unassigned sense codons would offer opportunities to conduct a similar study on sense codons to directly answer such questions.

2) For the experiment in Figure 2, the basis for some of the assignments are unclear. For example, how is it known that LEHHHHHHMVR results from a +19 skip instead of from loss of fidelity like LEHHHHHHYQR? The presence of two additional His residues in the His-tail of the GFP-His6 reporter may indeed indicate a -6 frameshift, as the authors propose, but may also indicate two consecutive -3 frameshifting events. Authors do not discuss the second possibility and suggest that they have detected "the furthest frameshift backward". Without strong evidence that they are indeed dealing with a -6 frameshift, instead of two -3 frameshifts, this seems to be an overstatement. Some of the peptides could also result from transcription errors rather than translation mistakes. A more quantitative summary of the types of peptides found in the mass spec, including the types of peptides found in the strain expressing the UAA-containing construct, would be very useful.

It would be difficult to determine using mass spectrometry whether LEHHHHHHMVR results from a + 19 skip or a loss of fidelity, but given that the skips of up to +50 bases have been validated in vivo (Huang et al., 1988) and loss of fidelity has only been observed in vitro prior to this manuscript, we hypothesize that this peptide arises from skipping. In addition, a key difference between LEHHHHHHMVR and the 5 tripeptides described in subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons” is that LEHHHHHHMVR could be genetically encoded with a +19 skip, while the other peptides could not be encoded unless multiple mistranslation events occurred. With regard to LEHHHHHHHH resulting from a -6 frameshift or two -3 frameshifts, it would be impossible to distinguish between these two scenarios using mass spectrometry data. As we have not found anything in the literature to denote which scenario is more likely, we have modified both the Abstract and text in subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons”, to read: “We detected ribosomal frameshifting of up to -3 (LEHHHHHHH) and +19 nucleotides (LEHHHHHHMVR), as determined by presence of fragments from all three reading frames appended immediately following the C-terminal peptide of LEHHHHHH. Additionally, the LEHHHHHHHH peptide may indicate a -6 frameshift, although it is not possible to determine whether this peptide arises from a single -6 frameshift or two -3 frameshifts between histidine incorporation.”

Although some of the tripeptides we identified could be the result of transcription mistakes, the number of transcriptional mistakes required to produce the degenerate tripeptides we identified would be between 2 and 6 transcriptional errors in a 30-nucleotide span. Given that the transcriptional error rate is 1 in 10,000 bases (Blank et al., 1986; Rosenberger and Hilton, 1983) and our strains have no mutations that would lead to greater error rates in transcription, we find this hypothesis to be less likely than loss of translational fidelity. We have revised the text to clarify this in subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons”, which reads: “it is unlikely these fragments arose from routine errors in mRNA transcription because this would require ≥2 transcriptional errors in a 30-nucleotide span. The transcription error rate in E. coli is estimated to be ~1 in 10,000 bases (Blank et al., 1986; Rosenberger and Hilton, 1983) and our strains have no known mutations that would lead to greater error rates in transcription.”

A summary of the C-terminal peptides identified from our mass spectrometry data is available on Supplementary file 1 and Supplementary file 2, and this has been modified to include the source strain for each peptide identified. As noted in the manuscript in subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons”, the only detectable GFP C-terminal peptide from UAA-ending constructs (and from strains containing RF1) were LEHHHHHH and LEHHHHHHAANDENYALDD. Mass spectrometry datasets that include ion intensity scores are available in Supplementary file 1 and Supplementary file 2. We agree that the quantification of the relative abundances of the various observed translational products would be very useful. Unfortunately, the amino acid sequence of each unique tryptic peptide confers different ionization properties, meaning that some peptides are much more amenable to observation by mass spectrometry than others. By the same token, the intensity of the peptides observed by mass spectrometry cannot be used to directly quantify difference in abundance between different peptides and mistranslated proteins. We therefore view our experiments as a “scouting” technique to identify the possible translational outcomes in response to the unassigned UAG codon, but further experimentation using a targeted collection of isotopically labeled peptides would be necessary for absolute peptide quantification, which is beyond the scope of the current work. However, even rigorously quantitative measures pose challenges because protein turnover rate is difficult to assess, and quickly-degraded proteins such as those with the tmRNA tag may be undervalued by simply comparing peptide abundances. A detailed description of mass spectrometry and proteomic analysis was added to the Materials and methods section of the manuscript.

3) Were the experiments in Figure 3 and Figure 4 done with strains containing ssrA-DD (as in Figure 2) or wild-type ssrA? This is critical for the interpretation. All strains and plasmids need to be described at a level of detail that would allow others to reproduce the work.

We thank the reviewers for their comment on the ssrA variants used in Figure 3 and Figure 4, and for pointing out the need to describe the strains and plasmids used in this study in greater detail. Strains used for the protein yield (Figure 3) and conjugation/infection (Figure 4) experiments contain wild-type ssrA (ssrA-AA). We added the following sentence to the manuscript to clarify this (subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”): “Since mass spectrometry data indicated that a combination of mechanisms could resolve stalled translation at the unassigned UAG codon, we constructed targeted deletions of the ribosomal rescue systems (ssrA, arfA, and arfB) in strains with wild-type ssrA sequence (GRO.AA) to determine whether protein production from UAG-ending transcripts in ΔRF1 cells could be restored to levels seen in +RF1 cells.”

We have also compiled a new table of strains (Table 1) and a list of plasmids (see Key Resources Table) used in this study. The strain table includes information on the status of the ssrA tag, as well as the figure attribution for each strain.

4) Figure 3A is completely not clear and perhaps even misleading since it shows the effect of the different mutations on the ratio between the expressed and non-expressed constructs. It is hard to know whether the effects shown are due to effects of the mutations on the expressed or non-expressed construct. To resolve this issue, the authors should show the effects of the different mutations on the actual max OD600 and doubling time values of each strain separately (i.e. not just on the ratio).

We thank the reviewers for this comment and agree with their assessment. We have thus modified Figure 3 to consist of bar graphs demonstrating the doubling times with and without GFP expression (Figure 3A) and max OD600with and without GFP expression (Figure 3B). The quantitative values that comprise these figures are also available on Supplementary file 3.

5) The results shown for some of the strains (Figure 3) are surprising and not expected (for example, the large decrease in max OD600 ratio of the ssrA arfB double mutant – Figure 3A, the opposite trend in GFP values shown for the ssrA arfB compared to arfB and ssrA alone – Figure 3C etc.). There is no clear explanation to these peculiarities in the text. A more elaborate discussion in the context of epistatic interactions shown is needed, such a discussion could address the effects of double mutations compare to single ones. Likewise, the authors should try to comment on the observation that knockout of arfA increases GFP production (Figure 3C), whereas ArfA is expected to facilitate termination of the GFP translation at the UAG codon.

We thank the reviewers for raising these points and would hypothesize that, given limited cellular resources, cells are capable of expressing high levels of protein or achieving high fitness (via low doubling times or high max OD600) but not both simultaneously. In ssrA knockout strains, GFP expression increase at the expense of cellular fitness because GFP peptides stalled on ribosomes are freed by ArfA or ArfB and not tagged for degradation. We have clarified this point in the manuscript in subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: “These ssrA knockout strains likely demonstrate increased GFP expression and reduced fitness (Figure 3A and 3B) because translation of GFP transcripts sequester cellular resources at the expense of cellular replication, producing GFP peptides that are freed from stalled ribosomes via ArfA or ArfB without addition of a degradation tag.”

We have also addressed the unusual results in the arfB knockout strain by stating in subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: “Interestingly, a single knockout of arfB significantly reduced production of protein from UAG-GFP to an undetectable level when calibrated to quantified GFP standards (p = 0.0311). These ArfB deletion data, together with the fitness reduction observed in the GRO, suggest that ArfB is constitutively expressed and relieving low levels of ribosomal stalling in E. coli.”

Since the increased GFP expression observed in the arfA knockout strain is not statistically significant when compared to the GRO strain expressing the same UAG-GFP construct, we do not think emphasizing such nuanced differences is warranted and would require detailed follow-up in future work.

6) In Figure 4B, the authors interpreted the change in doubling time as an indication for the ability of RK2 plasmid to replicate – some explanation to this should be added to text.

We thank the reviewers for identifying the need to clearly articulate our interpretation of doubling time data. The basis for interpreting changes in doubling time as an indication of RK2’s ability to replicate stems from a prior study we conducted that demonstrated the UAG-ending trfA gene on RK2, which initiates plasmid replication, reduces cell growth and increases doubling time in the GRO unless its UAG stop codon is recoded to UAA when grown in media selecting for plasmid maintenance (Ma and Isaacs, 2016). To clarify this point, we changed the sentence in which we discuss these results (in subsection “Deletion of ssrA restores conjugative plasmid propagation and viral infection in the GRO”) to the following: “Previous research indicates that the UAG stop codon in the trfA gene on RK2 leads to impaired conjugation efficiency and replication in the GRO (Ma and Isaacs, 2016), likely because the trfA protein is required to initiate plasmid replication(Pansegrau et al., 1994). Phenotypically, this manifests as reduced efficiency of plasmid transfer in conjugation experiments and increased doubling times for RK2+ strains in media selecting for plasmid maintenance due to loss of plasmid and concomitant antibiotic resistance genes.”

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

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) For the experiments in Figure 3 and Figure 4, explain whether the replicates are biological replicates or technical replicates and at what step the replication occurred. From the source data, it appears that all of the growth data is from a single 96 well plate. Were replicate wells in the plate inoculated from independent overnight cultures or from the same culture? This distinction is important to determine if the small changes observed are likely to be physiologically relevant. It is confusing that there are 3 plates of data in the source file, but data in the figure all seems to come from a single plate. Likewise, for the conjugation experiments, do the data come from one mating that was plated three times, from three matings made with the same cultures, or from three matings performed with independently grown cultures?

Thank you for pointing out the need to distinguish between biological and technical replicates in our experiments. We used the definitions for biological and technical replication outlined in Blainey et al., (2014): Biological replicates are parallel measurements of different biological samples subjected to the same experiment, and technical replicates are parallel measurements of a single biological sample subjected to experimentation. The data represented in all figures except for Figure 4A and 4C are the result of biological replicates. Figure 4A and 4C represent technical replicates (see below for more details).

The source data file for Figure 3A and 3B (Figure 3—source data 2) contains data from three identical 96-well plate experiments conducted using different 96-well plate reader models. Each experiment tested each sample in biological triplicate. Although data from the first two experiments are consistent with data from the third experiment, variability among plate reader machines precludes us from averaging the data obtained in the different experiments to draw conclusions of biological relevance. As such, we only represent the data obtained from “Plate 3” in Figure 3A and 3B. We modified the legend for Figure 3—source data 2 to clarify this: “Analysis of doubling times and maximum OD600’s of indicated strains. File contains doubling times and maximum OD600’s for three separate experiments conducted on different plate reader machines. Each experiment tested each sample in biological triplicate. Only the biological triplicate data from Plate 3 is represented in Figure 3A and 3B.”

To clarify that the data obtained from our conjugation experiments (represented in Figure 4A and 4C) are the result of technical replication, we have added the following sentence to the legend for Figure 4A and 4C: “Data are obtained from technical triplicates generated from a single biological sample.” We also note that these data are the consequence of technical replication in the legends for the appropriate source data files (Figure 4—source data 1 and Figure 4—source data 4).

We have written a new paragraph within our Materials and methods section to outline our definitions of biological and technical replication, and have clarified the nature of replication for each experiment within relevant paragraphs of the Materials and methods section. We updated our Transparent Reporting Form to include the above-mentioned definitions of biological and technical replication.

2) Figure 3C shows negative protein concentration, and the explanation is that the band was quantified, and the value was plugged into a formula from a standard curve, giving a negative result. Because it is physically impossible to have negative protein concentration and there is quite clearly a band on the blot, the explanation provided suggests that the standard curve was not accurate. From the source data, it appears that the standard curves are derived from two points at 1 ng and 100 ng fit to a line. Unfortunately, almost all of the samples are outside the 1-100 ng range. It appears the negative value is the result of a large negative number from blot 3 which also seems to have an anomalously low value for 100 ng in the standard curve. The differences in protein production from the different samples are clear, but the quantification is clearly inaccurate. This problem could potentially be addressed by running another replicate of the experiment or reporting the data normalized to one of the samples such as GRO.AA [pUAG-GFP]. We will not publish a negative concentration.

We agree with the reviewers that reporting a negative GFP value does not make sense, and we have amended our data analysis to provide better quantitation of the amount of GFP present in each sample. Images of the blots were provided in our initial submission as Figure 3—source data 3-5. The raw values for our reanalysis are provided in an updated version of Figure 3—source data 6.

For the western blots represented in Figure 3C, 1ng, 10ng, 50ng and 100ng GFP-6xHis standards were run on each blot next to the experimental samples. The majority of the samples fell within this range of GFP standards because the y-axis of Figure 3C, in µg/mL GFP, corresponds to a 6x correction factor based on the loading volume of lysate in each lane of the SDS-PAGE gels, to convert the ng of protein detected to µg/mL in the total lysate. Our previous data analysis, which yielded negative values for the GRO.AA.∆arfB samples and the 1ng standard, suggested that the antibody signal is sublinear in the 0-10ng regime. To address this problem, we revised the data analysis as follows: we created new standard curves using only the 10ng, 50ng and 100ng standards for each replicate western blot (R2 > 0.97 for all three blots), and fit data points with an intensity greater than that of the 10 ng standard to this curve. 20 of 24 samples fell within or slightly above this range, with the highest-intensity sample quantified as 136ng. We created a second set of standard curves using the 1ng and 10ng standards to quantify the protein abundances of three samples whose intensities fell within this range. The one remaining sample had a weaker intensity than that of the 1ng standard, and we quantified protein in this sample through a linear approximation between the 1ng sample intensity and a blank lane with an intensity of zero.

The results of this analysis are not substantively different than our previous analysis, but this method is clearly superior in calculating the GFP concentration from low-intensity samples and allows us to accurately report non-negative protein abundances. The standard curves from our reanalysis are provided in Figure 3—figure supplement 1.

We have revised the subsection “Western blot” to reflect these changes and to describe our analysis procedure in greater detail. The subsection now reads: “Western blots were run as described previously using SDS-PAGE gels (Pirman et al., 2015). […] We repeated three western blots in parallel for each strain induced in separate culture tubes (i.e., biological triplicates, see Protein expression and purification).”

Text clarifications:a) In subsection “Suppression, ribosomal frameshifting, and ssrA tagging occur at unassigned codons”: Spontaneous termination of translation could refer to untemplated termination by RF2 or to spontaneous (nonenzymatic) hydrolysis of the peptidyl-tRNA.

We thank the reviewers for pointing this out. The nonenzymatic hydrolysis of peptidyl-tRNA is posited to be a rare event (occurring about once per 100,000 codon decoding events), and we refer to this possibility in subsection “Suppression, ribosomal frameshifting, and tmRNA-mediated peptide ssrA tagging occur at unassigned codons”. However, we recognize that our methods do not distinguish between these possible events and have thus modified subsection “Suppression, ribosomal frameshifting, and tmRNA-mediated peptide ssrA tagging occur at unassigned codons” to read: “Given this, we hypothesize that these five peptides result from loss of translational fidelity after stalling at the UAG codon that may lead to (1) spontaneous termination of translation due to the untemplated action of RF2 following mistranslation or (2) ArfA- or ArfB-mediated release predicated on 3’ exonuclease degradation of the mRNA. The rare event of spontaneous hydrolysis of the peptide from the ribosome is also possible.”

b) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: Fitness is best used in relation to competitive growth experiments because it is possible for a strain to grow to a lower OD600 in monoculture but outcompete other strains and therefore have higher fitness. It would be clearer here to refer to growth rate or OD600 instead of fitness.

We thank the reviewers for making this distinction. We have replaced the term “fitness” in subsection “Suppression, ribosomal frameshifting, and tmRNA-mediated peptide ssrA tagging occur at unassigned codons” to “growth rate and cell density.”

c) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”: What is the knockout of arfB being compared to here? In Figure 3A, the induced/uninduced ratio for the arfB strain looks very similar to that for the isogenic arfA+ strain, which would seem to suggest that ArfB does not play a role.

In the sentence that the reviewers mention, we are comparing GFP production from UAG-ending transcripts between a strain with ssrA knockout and a strain with both an ssrA and arfB knockout. We intend to point out that the knockout of both ssrA and arfB is needed to recover GFP production to levels seen in GRO cells that produce GFP from UAA-ending transcripts. We revised this sentence (and the sentence before it) to clarify the nature of our comparisons, and to make explicit the nature of the p-values we are reporting (all of which are calculated using GRO.AA [pUAG-GFP] as the null hypothesis). The paragraph beginning in subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons” now reads:

“We then investigated the impact of unassigned codons on protein production using western blot densitometry, and found that the GRO expressing UAG-GFP produced less than one-fourth of the protein amount than does ECNR2 expressing UAG-GFP (Figure 3C, 8.0 µg/ml for the GRO versus 35 µg/ml for ECNR2, p = 0.0014). […] These ssrA deletion strains likely demonstrate increased GFP expression and reduced growth rate and cell density (Figure 3A and 3B) because translation of GFP transcripts sequesters cellular resources at the expense of cellular replication, producing GFP peptides that are freed from nonstop ribosomes via ArfA or ArfB without addition of a degradation tag."

The reviewers are also correct to point out that, in Figure 3A, knockout of arfB does not seem to affect the induced/uninduced doubling time ratio. Indeed, ArfB does not seem to play a substantial role in influencing the growth rate of the GRO when it is expressing protein from UAG-ending transcripts. However, our data on protein expression represented in Figure 3C suggests that ArfB has a significant role to play in protein production. We emphasize this point on in subsection “Suppression, ribosomal frameshifting, and tmRNA-mediated peptide ssrA tagging occur at unassigned codons “.

d) The sentence in the last paragraph of subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons” is unclear: "Interestingly, a single knockout of arfB significantly reduced production of protein from UAG-GFP to low levels similar to those mapped to quantified GFP standards."

We thank the reviewers for pointing this out. Our intention is to demonstrate that UAG-GFP expression in the ∆arfB GRO strain leads to very low protein production levels (although there is no statistically significant difference in protein production between this strain and the GRO strain with no knockouts), and that the resulting protein abundance from this strain is near the lower limit of detection of our assay. We have revised subsection “Suppression, ribosomal frameshifting, and tmRNA-mediated peptide ssrA tagging occur at unassigned codons” to read: “A deletion of arfB leads to strikingly low protein abundances from UAG-GFP transcripts that approach the lower limit of detection of our assay, although this apparent reduction in protein production was not statistically significant in comparison to protein production by GRO.AA [pUAG-GFP].”

e) In subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons”, the comparison for "fully restore protein expression from UAG-ending transcripts" appears to be GRO.AA[pUAA-GFP] instead of ECNR2[pUAG-GFP], but "restore" suggests it should be the latter. The comparison and the meaning behind which strain is used for the comparison should be clarified.

The reviewer is correct, and we are indeed comparing protein production from GRO.AA.∆ssrA.∆arfB [pUAG-GFP] with GRO.AA [pUAA-GFP]. Because of the genetic differences between ECNR2 strains and GRO strains (Table 1), it would be inaccurate to compare protein production from GRO.AA.∆ssrA.∆arfB [pUAG-GFP] with that of ECNR2 [pUAG-GFP]. Our intention in making this comparison is to show that the knockout of both ssrA and arfB is needed to enable the GRO to translate peptides from UAG-ending transcripts at abundances similar to the translation of peptides from UAA-ending transcripts. We have revised subsection “ssrA and arfB mediate degradation of proteins containing unassigned UAG codons” to clarify the nature of our comparison, and it now reads: “These data also suggest that while deletion of ssrA partially recovers protein production from UAG-ending transcripts in the GRO, deletion of both ssrA and arfB is necessary to fully recover protein expression from UAG-ending transcripts to levels seen from the translation of UAA-ending transcripts in the GRO.”

f) In subsection “Deletion of ssrA restores conjugative plasmid propagation and viral infection in the GRO”, it is not clear what comparison was used for the 2.4-fold increase in doubling time. The graph in Figure 4B shows a 38% increase for the arfA strain versus 28% for the isogenic wild type.

We thank the reviewers for pointing this out. We revised the sentence in question to accurately reflect the data shown in the figure. Subsection “Deletion of ssrA restores conjugative plasmid propagation and viral infection in the GRO” now reads: “RK2 conjugation efficiency in GRO.AA.∆ssrA improved to 99% (compared to 87% in GRO.AA), and the strain showed an increase in doubling time of only 6% compared to a 28% increase for GRO.AA (p < 0.0001). We observed similar results for GRO.AA.∆ssrA.∆arfB. However, single deletion of arfB halved RK2 conjugative efficiency (Figure 4A, p = 0.0002). This strain also exhibited a 38% increase in doubling time when bearing RK2, compared to the 28% increase in doubling time seen in the GRO with no ribosomal rescue gene deletions (Figure 4B, p < 0.0001).”

g) Discussion section: What is the evidence supporting the demonstration of ribosome stalling? It seems that ribosome stalling was assumed based on the addition of the SsrA tag. Experiments such as ribosome profiling could demonstrate ribosome stalling, but these were not done. I think stalling is part of the model here but has not been demonstrated.

We appreciate the reviewer’s question and believe that clarifying the terminology used in the manuscript will resolve the issue. The term “ribosomal stalling” refers to the situation in which a ribosome reaches the 3’ end of an mRNA without encountering a stop codon, or when translation slows or pauses in the middle of an mRNA (Keiler, 2015). Stalling at the 3’ end of an mRNA results in ribosomal rescue (Keiler, 2015). Slowed translation within an mRNA can arise when the ribosome encounters a rare codon or a codon with depleted or inefficient cognate decoding elements, and can result in near-cognate suppression, frameshifting, or mRNA cleavage followed by ribosomal rescue (Aerni et al., 2015; George et al., 2016; Hayes et al., 2002; Keiler, 2015; Li et al., 2007; Roche and Sauer, 1999).

The reviewer is correct to point out that we do not present data to support the claim that unassigned codons result in ribosomal stalling as we have defined the term above. However, our mass spectrometry data does reveal that unassigned codons elicit translational responses (i.e., ribosomal rescue, frameshifting, and near-cognate suppression) consistent with resolution mechanisms for stalled ribosomes.

We have modified the logic of the sentence in question to better reflect our experimental data. The sentence in the Discussion section now reads: “We demonstrate that unassigned stop codons elicit near-cognate suppression, frameshifting, and the action of ribosomal rescue mechanisms (Figure 2).” We conclude this paragraph with the additional sentence: “These mechanistic outcomes that occur as a consequence of ribosomal stalling could be further investigated via ribosomal profiling in future work.”

h) Discussion section: Similar to (g) above, the regulatory relationship between tmRNA and ArfA was used to explain the data, so it would be circular reasoning to then use this explanation to validate the regulatory relationship.

We thank the reviewers for pointing this out, and we agree that our data does not explicitly validate previously observed regulatory relationships between tmRNA and ArfA (as elucidated by Chadani et al., 2011; Garza-Sanchez et al., 2011; and Schaub et al., 2012). Our data does, however, reveal the physiological relevance of this relationship. In order to make this point more explicit, we have revised the sentence in the Discussion section to read: “Our GRO model thus sheds light on the functional significance of previously-described regulatory relationships while elucidating the unique mechanistic contributions of different ribosomal rescue systems in resolving translation at unassigned stop codons.”

i) Discussion section: "extensive" suggests there is a large amount of frameshifting, but the frameshifting events cannot be quantified using the techniques in this work. Perhaps something like "a wide variety of frameshifting events" would be more accurate.

We agree with the reviewer on this point. We have changed “extensive frameshifting” to “a wide variety of frameshifting events.”

j) Figure 1: The cartoon shows the SsrA-tagged protein going into the protease N terminus first, but all the proteases recognize the tag and start at the C terminus (this is not critical). In the legend, the word "hypothesized" should be removed from the last sentence – the lack of modification has been observed.

We thank the reviewers for this observation. We have modified the Figure to show the protease hydrolyzing the C-terminal end of the peptide and have removed the word “hypothesized” from the figure legend.

k) Figure 3 legend: For panels A and B, the legend does not match the labels. A is doubling time and B is max OD.

We thank the reviewers for pointing this out. We have altered the figure legend accordingly.

https://doi.org/10.7554/eLife.34878.034

Article and author information

Author details

  1. Natalie Jing Ma

    1. Department of Molecular, Cellular & Developmental Biology, Yale University, New Haven, United States
    2. Systems Biology Institute, Yale University, West Haven, United States
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Colin F Hemez and Karl W Barber
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7452-9482
  2. Colin F Hemez

    1. Systems Biology Institute, Yale University, West Haven, United States
    2. Department of Biomedical Engineering, Yale University, New Haven, United States
    Contribution
    Investigation, Visualization, Writing—original draft, Writing—review and editing
    Contributed equally with
    Natalie Jing Ma and Karl W Barber
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3445-7706
  3. Karl W Barber

    1. Systems Biology Institute, Yale University, West Haven, United States
    2. Department of Cellular & Molecular Physiology, Yale University School of Medicine, New Haven, United States
    Contribution
    Resources, Investigation, Visualization, Writing—review and editing
    Contributed equally with
    Natalie Jing Ma and Colin F Hemez
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0672-8409
  4. Jesse Rinehart

    1. Systems Biology Institute, Yale University, West Haven, United States
    2. Department of Cellular & Molecular Physiology, Yale University School of Medicine, New Haven, United States
    Contribution
    Resources, Supervision, Funding acquisition, Validation, Methodology
    For correspondence
    jesse.rinehart@yale.edu
    Competing interests
    No competing interests declared
  5. Farren J Isaacs

    1. Department of Molecular, Cellular & Developmental Biology, Yale University, New Haven, United States
    2. Systems Biology Institute, Yale University, West Haven, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    farren.isaacs@yale.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8615-8236

Funding

Defense Advanced Research Projects Agency (N66001-12-C-4211)

  • Farren Isaacs

U.S. Department of Energy (152339.5055249.100)

  • Farren Isaacs

National Institutes of Health (R01GM117230)

  • Jesse Rinehart
  • Farren Isaacs

National Institutes of Health (R01GM125951)

  • Farren J Isaacs
  • Jesse Rinehart

National Science Foundation (Graduate Research Fellowship DGE-1122492)

  • Karl W Barber

Gruber Foundation (Graduate Research Fellowship)

  • Natalie Jing Ma

Arnold and Mabel Beckman Foundation (Young Investigator Award)

  • Farren Isaacs

DuPont (Young Professor Award)

  • Farren Isaacs

National Institutes of Health (Graduate Training Grants T32GM007499,T32GM007223)

  • Natalie Jing Ma

Defense Advanced Research Projects Agency (HR0011-15-C-0091)

  • Farren Isaacs

Arnold and Mabel Beckman Foundation (Beckman Scholar Award)

  • Colin F Hemez

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the members of the Isaacs and Rinehart labs for helpful feedback on this study. We also thank Paul Turner for valuable discussions and experimental advice. The authors gratefully acknowledge support from DARPA (N66001-12-C-4211, HR0011-15-C-0091 to FJI), DOE (152339.5055249.100 to FJI), NIH (R01GM117230, R01GM125951 to FJI and JR, T32GM007499 and T32GM007223 to NJM), and NSF (DGE-1122492 to KWB). The authors also thank the Gruber Foundation (NJM), the Arnold and Mabel Beckman Foundation (FJI and CFH), and DuPont Inc. (FJI) for funding.

Publication history

  1. Received: January 7, 2018
  2. Accepted: August 26, 2018
  3. Version of Record published: October 30, 2018 (version 1)

Copyright

© 2018, Ma et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 856
    Page views
  • 153
    Downloads
  • 0
    Citations

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

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Chromosomes and Gene Expression
    2. Genetics and Genomics
    Jürgen Jänes et al.
    Tools and Resources Updated
    1. Genetics and Genomics
    Niko Välimäki et al.
    Research Article Updated