Introduction

Antimicrobial resistance is an increasingly serious threat to global public health. Our recent finding that many tRNA modification genes are involved in the response to antibiotics from different families (1) led to further investigate the links between environmental factors (e.g. traces of antibiotics), tRNA modifications and bacterial survival to antibiotics. The regulatory roles of RNA modifications was first proposed for eukaryotes(2) and their importance in human diseases has recently emerged(3,4). In bacteria, while some tRNA modifications are essential(5), the absence of many RNA modification shows no growth phenotype in unstressed cells(6). At the molecular level, the roles of tRNA modifications in differential codon decoding have been described in various species(710). In most cases, no growth phenotype was associated with these variations in decoding in bacteria. Recent studies, however, do highlight the links between tRNA modifications and stress responses in several bacterial species(6,1117), and new modifications are still being discovered(18). Until recently, few tRNA modification factors have been clearly linked with resistance and persistence to antibiotics, via differential codon decoding in cell membrane and efflux proteins (TrmD(19), MiaA(8)). A link between stress and adaptation was described to occur via the existence of modification tunable transcripts, or MoTTs.

MoTTs were first (and mostly) defined in eukaryotes as transcripts that will be translated more or less efficiently depending on the presence or absence of tRNA modifications(20), namely upon stress(21). In bacteria, links between tRNA modifications and the response to several stresses are highlighted by studies focusing on the following MoTT/codon and tRNA modification couples (reviewed in(6)): differential translation of RpoS/leucine codons via MiaA (E. coli)(12); Fur/serine codons via MiaB, in response to low iron (E. coli)(13); MgtA/proline codons via TrmD, in response to low magnesium(14); catalases/phenylalanine and aspartate codons via TrmB, during oxidative stress (P. aeruginosa)(11). Mycobacterial response to hypoxic stress(15) also features MoTTs. In this latter study, specific stress response genes were identified in silico, through their codon usage bias, and then experimentally confirmed for their differential translation. tRNA modification-dependent translational reprogramming in response to antibiotic stress has not been the focus of a study so far in bacteria.

During studies in V. cholerae, we recently discovered that t/rRNA modifications play a central role in response to stress caused by antibiotics with very different modes of action (1), not through resistance development, but by modulating tolerance. The identified RNA modification genes had not previously been associated with any antibiotic resistance phenotype. The fact that different tRNA modifications have opposite effects on tolerance to different antibiotics highlights the complexity of such a network, and shows that the observed phenotypes are not merely due to a general mistranslation effect. Since tRNA modifications affect codon decoding and accuracy, it is important to address how differential translation can generate proteome diversity, and eventually adaptation to antibiotics.

In particular, deletion of the tgt gene encoding tRNA-guanine transglycosylase (Tgt) in V. cholerae confers a strong growth defect in the presence of aminoglycosides at doses below the minimal inhibitory concentration (sub-MIC)(1). Tgt incorporates queuosine (Q) in the place of guanosine (G) in the wobble position of four tRNAs with GUN anticodon (tRNA-Asp GUC, tRNA-Asn GUU, tRNA-Tyr GUA, tRNA-His GUG)(22). The tRNAs with “AUN” anticodons are not present in the genome, and thus each one of the four GUN tRNAs decodes two synonymous codons (aspartate GAC/GAT, asparagine AAC/AAT, tyrosine TAC/TAT, histidine CAC/CAT which differ in the third position). Q is known to increase or decrease translation error rates in eukaryotes in a codon and organism specific manner(22,23). Q was shown to induce mild oxidative stress resistance in the eukaryotic parasite Entamoeba histolytica, the causative agent of amebic dysentery, and to attenuate its virulence(24). In E. coli, the absence of Q modification was found to decrease mistranslation rates by tRNA-Tyr, while increasing it for tRNA-Asp(25,26). No significant biological difference was found in E. coli Δtgt mutant, except for a slight defect in stationary phase viability(27). Recent studies show that the E. coli tgt mutant is more sensitive to aminoglycosides but not to ampicillin nor spectinomycin and is more sensitive to oxidative stress but the molecular mechanisms were not elucidated(28).

We asked here how queuosine (Q) modification by Tgt modulates the response to sub-MIC aminoglycosides. We find that V. cholerae Δtgt displays differential decoding of tyrosine TAC/TAT. Molecular reporters, coupled to proteomics and in silico analysis, reveal that several proteins with codon usage biased towards TAT (versus TAC) are more efficiently translated in Δtgt. One of these proteins is RsxA, which prevents activation of SoxR oxidative stress response regulon(29). We propose that sub-MIC TOB treatment leads to increased expression of tgt and Q modification, which in turn allows for more efficient Sox regulon related oxidative stress response, and better response to sub-MIC TOB. Lastly, bioinformatic analysis identified DNA repair gene transcripts with TAT codon bias as transcripts modulated by Q modification, which was confirmed by decreased UV susceptibility of V. cholerae Δtgt.

Materials and methods

Media and Growth Conditions

Platings were done at 37°C, in Mueller-Hinton (MH) agar media. Liquid cultures were grown at 37°C in MH in aerobic conditions, with 180 rotations per minute.

Competition experiments were performed as described(1): overnight cultures from single colonies of mutant lacZ+ and WT lacZ-strains were washed in PBS (Phosphate Buffer Saline) and mixed 1:1 (500 μl + 500 μl). At this point 100 μl of the mix were serial diluted and plated on MH agar supplemented with X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) at 40 μg/mL to assess T0 initial 1:1 ratio. At the same time, 10 μl from the mix were added to 2 mL of MH or MH supplemented with sub-MIC antibiotics (concentrations, unless indicated otherwise: TOB: tobramycin 0.6 μg/ml; GEN: 0.5 µg/ml; CIP: ciprofloxacin 0.01 μg/ml, CRB: carbenicillin 2.5 μg/ml), PQ: paraquat 10 µM, or H2O2: 0.5 mM. Cultures were incubated with agitation at 37°C for 20 hours, and then diluted and plated on MH agar plates supplemented with X-gal. Plates were incubated overnight at 37°C and the number of blue and white CFUs was assessed. Competitive index was calculated by dividing the number of blue CFUs (lacZ+ strain) by the number of white CFUs (lacZ-strain) and normalizing this ratio to the T0 initial ratio. When a plasmid was present, antibiotic was added to maintain selection: kanamycin 50 µg/ml for pSEVA.

Construction of complementation and overexpression plasmids in pSEVA238

Genes were amplified on V. cholerae genomic DNA using primers listed in Table S4 and cloned into pSEVA238 (30) under the dependence of the Pm promoter(31), by restriction digestion with XbaI+EcoRI and ligation using T4 DNA ligase. Sodium benzoate 1 mM was added in the medium as inducer.

Survival/tolerance tests were performed on early exponential phase cultures. The overnight stationary phase cultures were diluted 1000x and grown until OD 600 nm = 0.35 to 0.4, at 37°C with shaking, in Erlenmeyers containing 25 mL fresh MH medium. Appropriate dilutions were plated on MH plates to determine the total number of CFUs in time zero untreated cultures. 5 mL of cultures were collected into 50 mL Falcon tubes and treated with lethal doses of desired antibiotics (5 or 10 times the MIC: tobramycin 5 or 10 μg/mL, carbenicillin 50 μg/mL, ciprofloxacin 0.025 μg/mL) for 30 min, 1 hour, 2 hours and 4 hours if needed, at 37°C with shaking in order to guarantee oxygenation. Appropriate dilutions were then plated on MH agar without antibiotics and proportion of growing CFUs were calculated by doing a ratio with total CFUs at time zero. Experiments were performed 3 to 8 times.

MIC determination

Stationary phase cultures grown in MH were diluted 20 times in PBS, and 300 μL were plated on MH plates and dried for 10 minutes. Etest straps (Biomérieux) were placed on the plates and incubated overnight at 37°C.

Quantification of fluorescent neomycin uptake was performed as described(32). Neo-Cy5 is the neomycin aminoglycoside coupled to the fluorophore Cy5, and has been shown to be active against Gram-bacteria(33,34). Briefly, overnight cultures were diluted 100-fold in rich MOPS (Teknova EZ rich defined medium). When the bacterial strains reached an OD 600 nm of ∼0.25, they were incubated with 0.4 μM of Cy5 labeled neomycin for 15 minutes at 37°C. 10 μl of the incubated culture were then used for flow cytometry, diluting them in 250 μl of PBS before reading fluorescence. WT V. cholerae, was incubated simultaneously without neo-Cy5 as a negative control. Flow cytometry experiments were performed as described(35) and repeated at least 3 times. For each experiment, 100,000 events were counted on the Miltenyi MACSquant device.

PMF measurements

Quantification of PMF was performed using the Mitotracker Red CMXRos dye (Invitrogen) as described(36), in parallel with the neo-Cy5 uptake assay, using the same bacterial cultures. 50 µL of each culture were mixed with 60 µL of PBS. Tetrachlorosalicylanilide TCS (Thermofischer), a protonophore, was used as a negative control with a 500 µM treatment applied for 10 minutes at room temperature. Then, 25 nM of Mitotracker Red were added to each sample and let at room temperature for 15 minutes under aluminium foil. 20 μL of the treated culture were then used for flow cytometry, diluted in 200 μL of PBS before reading fluorescence.

tRNA overexpressions

Synthetic fragments carrying the Ptrc promoter, the desired tRNA sequence and the natural transcriptional terminator sequence of VCt002 were ordered from IDT as double stranded DNA gBlocks, and cloned into pTOPO plasmid. Sequences are indicated in Table S4.

mRNA purification

For RNA extraction, overnight cultures were diluted 1:1000 in MH medium and grown with agitation at 37°C until an OD600 of 0.3-0.4 (exponential phase). 0.5 mL of these cultures were centrifuged and supernatant removed. Pellets were homogenized by resuspension with 1.5 mL of room temperature TRIzol Reagent. Next, 300 μL chloroform were added to the samples following mix by vortexing. Samples were then centrifuged at 4°C for 10 minutes. Upper (aqueous) phase was transferred to a new 2mL tube and mixed with 1 volume of 70% ethanol. From this point, the homogenate was loaded into a RNeasy Mini kit (Qiagen) column and RNA purification proceeded according to the manufacturer’s instructions. Samples were then subjected to DNase treatment using TURBO DNA-free Kit (Ambion) according to the manufacturer’s instructions.

mRNA quantifications by digital-RT-PCR

qRT-PCR reactions were prepared with 1 μL of diluted RNA samples using the qScript XLT 1-Step RT-qPCR ToughMix (Quanta Biosciences, Gaithersburg, MD, USA) within Sapphire chips. Digital PCR was conducted on a Naica Geode programmed to perform the sample partitioning step into droplets, followed by the thermal cycling program suggested in the user’s manual. Primer and probe sequences used in digital qRT-PCR reaction are listed in Table S4. Image acquisition was performed using the Naica Prism3 reader. Images were then analyzed using Crystal Reader software (total droplet enumeration and droplet quality control) and the Crystal Miner software (extracted fluorescence values for each droplet). Values were normalized against expression of the housekeeping gene gyrA as previously described(37).

tRNA level quantification by qRT-PCR

First-strand cDNA synthesis and quantitative real-time PCR were performed with KAPA SYBR® FAST Universal (CliniSciences) on the QuantStudio Real-Time PCR (Thermo Fischer) using the primers indicated in Table S4. Transcript levels of each gene were normalized to gyrA as the reference gene control(37). Gene expression levels were determined using the 2-ΔΔCq method (Bustin et al., 2009; Livak and Schmittgen, 2001) in respect to the MIQE guidelines. Relative fold-difference was expressed either by reference to antibiotic free culture or the WT strain in the same conditions. All experiments were performed as three independent replicates with all samples tested in duplicate. Cq values of technical replicates were averaged for each biological replicate to obtain the ΔCq. After exponential transformation of the ΔCq for the studied and the normalized condition, medians and upper/lower values were determined.

Construction of gfp reporters with codon stretches

The positive control was gfpmut3 (stable gfp)(38) under the control of Ptrc promoter, the transcription start site, rbs and ATG start codon are indicated in bold and underlined.

TTGACAATTAATCATCCGGCTCGTATAATGTGTGGAATTGTGAGCGGATAACAATTTCACACAGGAAACAGCG CCGCATGCGTAAAGGAGAAGAACTTTTCACTGGAGTTGTCCCAATTCTTGTTGAATTAGATGGTGATGTTAATG GGCACAAATTTTCTGTCAGTGGAGAGGGTGAAGGTGATGCAACATACGGAAAACTTACCCTTAAATTTATTTG CACTACTGGAAAACTACCTGTTCCATGGCCAACACTTGTCACTACTTTCGGTTATGGTGTTCAATGCTTTGCGAG ATACCCAGATCATATGAAACAGCATGACTTTTTCAAGAGTGCCATGCCCGAAGGTTATGTACAGGAAAGAACT ATATTTTTCAAAGATGACGGGAACTACAAGACACGTGCTGAAGTCAAGTTTGAAGGTGATACCCTTGTTAATA GAATCGAGTTAAAAGGTATTGATTTTAAAGAAGATGGAAACATTCTTGGACACAAATTGGAATACAACTATAA CTCACACAATGTATACATCATGGCAGACAAACAAAAGAATGGAATCAAAGTTAACTTCAAAATTAGACACAAC ATTGAAGATGGAAGCGTTCAACTAGCAGACCATTATCAACAAAATACTCCAATTGGCGATGGCCCTGTCCTTTT ACCAGACAACCATTACCTGTCCACACAATCTGCCCTTTCGAAAGATCCCAACGAAAAGAGAGACCACATGGTCC TTCTTGAGTTTGTAACAGCTGCTGGGATTACACATGGCATGGATGAACTATACAAATAA

For the tested codon stretches, 6 repeats of the desired codon were added just after the ATG start codon of gfp. The DNA fragments were ordered as double stranded eblocks from Integrated DNA technologies (IDT), and cloned into pTOPO-Blunt using kanamycin resistance, following the manufacturer’s instructions.

For tests of sequence context surrounding tyrosine codons of rsxA, DNA was ordered from IDT and cloned into pTOPO as described for codon stretches above, based on the following amino acid sequences (tested sequences in bold):

VC1017RsxA V. cholerae

MLLLWQSRIMPGSEANIYITMTEYLLLLIGTVLVNNFVLVKFLGLCPFMGVSKKLETAIGMGLATTFVLTLASVCAYL VESYVLRPLGIEYLRTMSFILVIAVVVQFTEMVVHKTSPTLYRLLGIFLPLITTNCAVLGVALLNINENHNFIQSIIYGFG AAVGFSLVLILFASMRERIHVADVPAPFKGASIAMITAGLMSLAFMGFTGLVKL

RsxA E. coli

MTDYLLLFVGTVLVNNFVLVKFLGLCPFMGVSKKLETAMGMGLATTFVMTLASICAWLIDTWILIPLNLIYLRTLAFIL VIAVVVQFTEMVVRKTSPVLYRLLGIFLPLITTNCAVLGVALLNINLGHNFLQSALYGFSAAVGFSLVMVLFAAIRERL AVADVPAPFRGNAIALITAGLMSLAFMGFSGLVKL

Quantification of gfp fusion expression by fluorescent flow cytometry

Flow cytometry experiments were performed as described(35) on overnight cultures and repeated at least 3 times. For each experiment, 50,000 to 100,000 events were counted on the Miltenyi MACSquant device. The mean fluorescence per cell was measured at the FITC channel for each reporter in both WT and Δtgt strains, and the relative fluorescence was calculated as the ratio of the mean fluorescence of a given reporter in Δtgt over the mean fluorescence of the same reporter in the WT. Native gfp (gfpmut3) was used as control.

Transcriptional fusion: rsxA promoter sequence was amplified using primers ZIP796/ZIP812. gfp was amplified from pZE1-gfp(39) using primers ZIP813/ZIP200. The two fragments were PCR assembled into PrsxA-gfp using ZIP796/ZIP200 and cloned into pTOPO-TA cloning vector. The PrsxA-gfp fragment was then extracted using EcoRI and cloned into the low copy plasmid pSC101 (1 to 5 copies per cell). The plasmid was introduced into desired strains, and fluorescence was measured on indicated conditions, by counting 100,000 cells on the Miltenyi MACSquant device. Likewise, the control plasmid Pc-gfp (constitutive) was constructed using primers ZIP513/ZIP200 and similarly cloned in pSC101. For translational fusions, the constitutive Ptrc promoter, the rsxA gene (without stop codon) with desired codon usage fused to gfp (without ATG start codon) was ordered from IDT in the pUC-IDT vector (carbenicillin resistant). Native sequence of V. cholerae rsxA gene, called rsxATATgfp in this manuscript is shown below. For rsxATACgfp, all tyrosine TAT codons were replaced with TAC.

gfpTAC and gfpTAT (tyrosine 11 TAT instead of 11 TAC) were ordered from IDT as synthetic genes under the control of Ptrc promoter in the pUC-IDT plasmid (carbenicillin resistant). The complete sequence of ordered fragments is indicated in Table S4, tyrosine codons are underlined.

Construction of bla reporters

Point mutations for codon replacements were performed using primer pairs where the desired mutations were introduced and by whole plasmid PCR amplification on circular pTOPO-TA plasmid. Primers are listed Table S4.

Growth on microtiter plate reader for bla reporter assays

Overnight cultures were diluted 1:500 in fresh MH medium, on 96 well plates. Each well contained 200 μL. Plates were incubated with shaking on TECAN plate reader device at 37°C, OD 600 nm was measured every 15 min. Tobramycin was used at sub-MIC: TOB 0.2 μg/mL. Kanamycin and carbenicillin were used at selective concentration: CRB 100 μg/mL, KAN 50 μg/mL.

Protein extraction

Overnight cultures of V. cholerae were diluted 1:100 in MH medium and grown with agitation at 37°C until an OD 600 nm of 0.3 (exponential phase). 50 mL of these cultures were centrifuged for 10 min at 4°C and supernatant removed. Lysis was achieved by incubating cells in the presence of lysis buffer (10 mM Tris-HCl pH 8, 150 mM Nacl, 1% triton 100X) supplemented with 0.1 mg/mL lysozyme and complete EDTA-free Protease Inhibitor Cocktail (Roche) for 1 hour on ice. Resuspensions were sonicated 3×50 sec (power: 6, pulser: 90%), centrifuged for 1 h at 4°C at 5000 rpm and supernatants were quantified using Pierce™ BCA Protein Assay Kit (Cat. No 23225) following the manufacturer’s instructions. Proteins were then stored at -80°C.

Proteomics MS and analysis

Sample preparation for MS

Tryptic digestion was performed using eFASP (enhanced Filter-Aided Sample Preparation) protocol(40). All steps were done in 30 kDa Amicon Ultra 0.5 mL filters (Millipore). Briefly, the sample was diluted with a 8M Urea, 100 mM ammonium bicarbonate buffer to obtain a final urea concentration of 6 M. Samples were reduced for 30 min at room temperature (RT) with 5 mM TCEP. Subsequently, proteins were alkylated in 5 mM iodoacetamide for 1 hour in the darkness at RT and digested overnight at 37°C with 1 µg trypsin (Trypsin Gold Mass Spectrometry Grade, Promega). Peptides were recovered by centrifugation, concentrated to dryness and resuspended in 2% acetonitrile (ACN)/0.1% FA just prior to LC-MS injection.

LC-MS/MS analysis

Samples were analyzed on a high-resolution mass spectrometer, Q Exactive™ Plus Hybrid Quadrupole-Orbitrap™ Mass Spectrometer (Thermo Scientific), coupled with an EASY 1200 nLC system (Thermo Fisher Scientific, Bremen). One µg of peptides was injected onto a home-made 50 cm C18 column (1.9 μm particles, 100 Å pore size, ReproSil-Pur Basic C18, Dr. Maisch GmbH, Ammerbuch-Entringen, Germany). Column equilibration and peptide loading were done at 900 bars in buffer A (0.1% FA). Peptides were separated with a multi-step gradient from 3 to 22 % buffer B (80% ACN, 0.1% FA) in 160 min, 22 to 50 % buffer B in 70 min, 50 to 90 % buffer B in 5 min at a flow rate of 250 nL/min. Column temperature was set to 60°C. The Q Exactive™ Plus Hybrid Quadrupole-Orbitrap™ Mass Spectrometer (Thermo Scientific) was operated in data-dependent mode using a Full MS/ddMS2 Top 10 experiment. MS scans were acquired at a resolution of 70,000 and MS/MS scans (fixed first mass 100 m/z) at a resolution of 17,500. The AGC target and maximum injection time for the survey scans and the MS/MS scans were set to 3E6, 20ms and 1E6, 60ms, respectively. An automatic selection of the 10 most intense precursor ions was activated (Top 10) with a 35 s dynamic exclusion. The isolation window was set to 1.6 m/z and normalized collision energy fixed to 27 for HCD fragmentation. We used an underfill ratio of 1.0 % corresponding to an intensity threshold of 1.7E5. Unassigned precursor ion charge states as well as 1, 7, 8 and >8 charged states were rejected and peptide match was disable.

Data analysis

Acquired Raw data were analyzed using MaxQuant 1.5.3.8 version(41) using the Andromeda search engine(42) against Vibrio cholerae Uniprot reference proteome database (3,782 entries, download date 2020-02-21) concatenated with usual known mass spectrometry contaminants and reversed sequences of all entries. All searches were performed with oxidation of methionine and protein N-terminal acetylation as variable modifications and cysteine carbamidomethylation as fixed modification. Trypsin was selected as protease allowing for up to two missed cleavages. The minimum peptide length was set to 5 amino acids. The false discovery rate (FDR) for peptide and protein identification was set to 0.01. The main search peptide tolerance was set to 4.5 ppm and to 20 ppm for the MS/MS match tolerance. One unique peptide to the protein group was required for the protein identification. A false discovery rate cut-off of 1 % was applied at the peptide and protein levels. All mass spectrometry proteomics data have been deposited at ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD035297.

The statistical analysis of the proteomics data was performed as follows: three biological replicates were acquired per condition. To highlight significantly differentially abundant proteins between two conditions, differential analyses were conducted through the following data analysis pipeline: (1) deleting the reverse and potential contaminant proteins; (2) keeping only proteins with at least two quantified values in one of the three compared conditions to limit misidentifications and ensure a minimum of replicability; (3) log2-transformation of the remaining intensities of proteins; (4) normalizing the intensities by median centering within conditions thanks to the normalizeD function of the R package DAPAR(43), (5) putting aside proteins without any value in one of both compared conditions: as they are quantitatively present in a condition and absent in another, they are considered as differentially abundant proteins and (6) performing statistical differential analysis on them by requiring a minimum fold-change of 2.5 between conditions and by using a LIMMA t test(44) combined with an adaptive Benjamini-Hochberg correction of the p-values thanks to the adjust.p function of the R package cp4p(45). The robust method of Pounds and Cheng was used to estimate the proportion of true null hypotheses among the set of statistical tests(46). The proteins associated with an adjusted p-value inferior to an FDR level of 1% have been considered as significantly differentially abundant proteins. Finally, the proteins of interest are therefore the proteins that emerge from this statistical analysis supplemented by those being quantitatively absent from one condition and present in another. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD035297.

RNA purification for RNA-seq

Cultures were diluted 1000X and grown in triplicate in MH supplemented or not with 0.6 µg/ml of tobramycin, corresponding to 50% of the MIC in liquid cultures, to an OD 600nm of 0.4. RNA was purified with the RNAeasy mini kit (Qiagen) according to manufacturer’s instructions. Briefly, 4 ml of RNA-protect (Qiagen) reagent were added on 2 ml of bacterial cultures during 5 minutes. After centrifugation, the pellets were conserved at -80°C until extraction. Protocol 2 of the RNA protect Bacteria Reagent Handbook was performed, with the addition of a proteinase K digestion step, such as described in the protocol 4. Quality of RNA was controlled using the Bioanalyzer. Sample collection, total RNA extraction, library preparation, sequencing and analysis were performed as previously described (47). The data for this RNA-seq study has been submitted in the GenBank repository under the project number GSE214520.

Gene Ontology (GO) enrichment analysis

GO enrichment analyses were performed on http://geneontology.org/ as follows: Binomial test was used to determine whether a group of genes in the tested list was more or less enriched than expected in a reference group. The annotation dataset used for the analysis was GO biological process complete. The analyzed lists were for each condition (MH/TOB), genes (Table S3) with at least 2-fold change in RNA-seq data of WT strain compared to Δtgt, and with an adjusted p-value <0,05. The total number of uploaded gene list to be analyzed were 53 genes for MH and 60 genes for TOB. The reference gene list was Vibrio cholerae (all genes in database), 3782 genes. Annotation Version: PANTHER Overrepresentation Test (Released 20220712). GO Ontology database DOI: 10.5281/zenodo.6399963 Released 2022-03-22

Stringent response measurement

P1rrnB-gfp fusion was constructed using gfp ASV(48), and cloned into plasmid pSC101. P1rrnB-GFPasv transcriptional fusion was amplified from strain R438 (E. coli MG1655 attB::P1rrnB gfp-ASV::kan provided by Ivan Matic) using primers AFC060 and AFC055, thus including 42 bp upstream of rrnB transcription initiation site. PCR product was then cloned in pTOPOblunt vector and subcloned to pSC101 by EcoRI digestion and ligation. The final construct was confirmed by Sanger sequencing. The plasmid was then introduced by electroporation into the tested strains. Overnight cultures were performed in MH + carbenicillin 100 µg/mL and diluted 500x in 10 mL fresh MH or MH+ TOB 0.4 µg/mL, in an Erlenmeyer. At time points 0 min, and every 30 during 3 hours, the OD 600 nm was measured and fluorescence was quantified in flow cytometry. For each experiment, 50,000 to 100,000 events were counted on the Miltenyi MACSquant device.

tRNA enriched RNA extraction

Overnight cultures of V. cholerae were diluted 1:1000 in MH medium and grown in aerobic conditions, with 180 rotations per minute at 37°C until an OD 600 nm of 0.5. tRNA enriched RNA extracts were prepared using room temperature TRIzol™ reagent as described(49) and contaminating DNA were eliminated using TURBO DNA-free Kit (Ambion). RNA concentration was controlled by UV absorbance using NanoDrop 2000c (Thermo Fisher Scientific). The profile of isolated tRNA fractions was assessed by capillary electrophoresis using an RNA 6000 Pico chip on Bioanalyzer 2100 (Agilent Technologies).

tRNA-enriched sample digestion for quantitative analysis of queuosine by mass spectrometry

Purified tRNA enriched RNA fractions were digested to single nucleosides using the New England BioLabs Nucleoside digestion mix (Cat No. M0649S). 10μl of the RNA samples diluted in ultrapure water to 100 ng/μL were mixed with 1μL of enzyme, 2μl of Nucleoside Digestion Mix Reaction Buffer (10X) in a final volume of 20 μL in nuclease-free 1.5 mL tubes. Tubes were wrapped with parafilm to prevent evaporation and incubated at 37°C overnight.

Queuosine quantification by LC-MS/MS

Analysis of global levels of queuosine (Q) was performed on a Q exactive mass spectrometer (Thermo Fisher Scientific). It was equipped with an electrospray ionization source (H-ESI II Probe) coupled with an Ultimate 3000 RS HPLC (Thermo Fisher Scientific). The Q standard was purchased from Epitoire (Singapore).

Digested RNA was injected onto a ThermoFisher Hypersil Gold aQ chromatography column (100 mm * 2.1 mm, 1.9 μm particle size) heated at 30°C. The flow rate was set at 0.3 mL/min and run with an isocratic eluent of 1% acetonitrile in water with 0.1% formic acid for 10 minutes.

Parent ions were fragmented in positive ion mode with 10% normalized collision energy in parallel-reaction monitoring (PRM) mode. MS2 resolution was 17,500 with an AGC target of 2e5, a maximum injection time of 50 ms, and an isolation window of 1.0 m/z.

The inclusion list contained the following masses: G (284.1) and Q (410.2). Extracted ion chromatograms of base fragments (±5ppm) were used for detection and quantification (152.0565 Da for G; 295.1028 Da for Q). The secondary base fragment 163.0608 was also used to confirm Q detection but not for quantification.

Calibration curves were previously generated using synthetic standards in the ranges of 0.2 to 40 pmol injected for G and 0.01 to 1 pmol for Q. Results are expressed as a percentage of total G.

Queuosine detection by sequencing

The detection of queuosine was performed as described in (50). Briefly, 200 ng of total RNA were subjected to oxidation by 45 mM of NaIO4 in 50 mM AcONa pH 5.2 buffer for 1h at 37°C. The reaction was quenched by addition of 36 mM glucose and incubation for 30 min at 37°C and RNA was precipitated with absolute ethanol. After precipitation and two washes by 80% ethanol, the RNA pellet was resuspended in 3’ ligation reaction buffer 1x and subjected to library preparation using the NEBNext® Small RNA Library Prep Set for Illumina® (NEB, #E7330S). Specific primers for V. cholerae tRNAAsn_GUU1, tRNAAsn_GUU2, tRNAAsp_GUC, tRNATyr_GUA and tRNAHis_GUG were hybridized instead of RT primer used in NEBNext® Small RNA Library Prep kit, under the same hybridization conditions. The 5’-SR adaptor was ligated, and reverse transcription was performed for 1h at 50°C followed by 10 min at 80°C using Superscript IV RT (instead of Protoscript II used in the kit). PCR amplification was performed as described in the manufacturer’s protocol. Libraries were qualified using Tapestation 4150 and quantified using Qubit fluorometer. Libraries were multiplexed and sequenced in a 50 bp single read mode using NextSeq2000 (Illumina, San Diego).

Bioinformatic analysis was performed by trimming of raw reads using trimmomatic v0.39 to remove adapter sequences as well as very short and low-quality sequencing reads. Alignment was done by bowtie2 (v2.4.4) in End-to-End mode with --mp 2 --rdg 0,2 options to favor retention of reads with deletions, only non-ambiguously mapped reads were taken for further analysis. Coverage file was created with samtools mpileup and deletion signature extracted for every position using custom R script. Deletion score was calculated as number of deletions divided by number of matching nucleotides at a given position. Analysis of Q tRNA modification in V. cholerae strains was performed in triplicate for biological replicates with technical duplicate for each sample.

Analysis of Queuosine tRNA Modification Using APB Northern Blot Assay

Quantification of queuosine in tRNA-Tyr from purified tRNA-enriched RNA fractions was performed using a non-radioactive Northern blot method: the procedure for pouring and running N-acryloyl-3-aminophenylboronic acid (APB) gels was based on the method detailed in (51). tRNA-Tyr were detected using the following 3’-end digoxigenin (DIG)-labeled probe: 5’ - CTTTGGCCACTCGGGAACCCCTCC - 3’DIG.

For 1 gel, ABP gel buffer was prepared by mixing 4.2g urea, 50mg 3-(Acrylamido) phenylboronic acid (Sigma Aldrich Cat No. 771465), 1ml 10X RNase-free TAE, 3.2ml 30% acrylamide and bis-acrylamide solution 37.5:1 and adding water to adjust the final volume to 10ml. After stirring to facilitate dissolution and right before pouring, 10µl TEMED and 60 μL 10% APS were added to the 10ml ABP buffer to catalyze and initiate polymerization respectively. Gels were casted using the Mini-PROTEAN® Bio-Rad handcast system, short plates (70×100 mm), 0.75 mm spacers and 10-well gel combs. Gels were left to polymerize at room temperature for 50 minutes.

Alkaline hydrolysis in 100mM of Tris-HCl pH 9 of our tRNA-enriched RNA extracts was carried out at 37°C for 30 min to break the ester bonds between tRNAs and their cognate amino acids. 10µl of the deacylated tRNA-enriched RNA samples were mixed with 8 µl of 2X RNA loading dye (Thermo Scientific™ Cat No. R0641), denatured for 3 min at 72°C and the whole volume was loaded onto the gel. Electrophoresis of the gels were carried out in the Mini-PROTEAN® Tetra Vertical Electrophoresis Cell for 30 min at 85V at room temperature followed by 1h30 at 140 V at 4°C. Gels were incubated at room temperature for 15 min with shaking in 50 ml 1X RNAse-free TAE mixed with 10 µl of 10000X SYBR Gold nucleic acid staining solution (Invitrogen™ S11494) and nucleic acids were visualized using a transilluminator. Transfers of the nucleic acids to positively charged nylon membranes were performed at 5 V/gel for 40 min at room temperature using a semi dry blotting system. RNAs were crosslinked to the membrane surface through exposure to 254 nm UV light at a dose of 1.2 Joules.

Membranes were transferred into glass bottles containing 5ml of pre-warmed hybridization buffer and incubated for 1 hour at 42°C at a constant rotation in a hybridization oven. Hybridization buffer was obtained by mixing 12.3 ml of 20X SSX, 1ml of 1M Na2HPO4 pH7.2, 17.5ml of 20% SDS, 2ml of 50X Denhardt’s solution and 17ml RNase-free H2O. 3µl of the DIG-labeled probe solution at 100 pmol/µl were then added into the 5ml hybridization buffer and the bottles were rotated in the hybridization oven at 42°C overnight. Membranes were washed 2 time with 2X SSC/5% SDS for 15 min at 42°C and 1 time with 1X SSC/1% SDS for 15min at 42°C.

Nucleic acids wash and immunological detection of the DIG-labeled probes were performed using the DIG Wash and Block Buffer Set (Roche, Cat No. 11585762001) according to the manufacturer’s protocols. Membranes were placed in a plastic container filled with 15ml of the blocking solution for nonspecific binding sites blocking. After 30min incubation at room temperature with rotation, 3µl of the alkaline phosphatase-coupled anti-DIG antibody (Fab fragments, Roche, Cat No. 11093274910) were added to the buffer and incubation at room temperature on a belly-dancer was allowed for 30 more min. Next, membranes were washed 3 times 15 min with DIG-wash buffer and once with DIG-detection buffer for 5min. For chemiluminescence visualization of the probe, 1ml of CDP-Star® Chemiluminescent Substrate (Disodium 2-chloro-5-(4-methoxyspiro[1,2-dioxetane-3,2′-(5-chlorotricyclo[3.3.1.13.7]decan])-4-yl]-1-phenyl phosphate) (Roche, Cat No. 11685627001) was added to 9ml of DIG-detection buffer and membranes were then incubated with the substrate for 5 min. The chemiluminescent signal was detected with the iBright Imaging Systems.

Ribosome profiling (Ribo-seq)

Pellet from 200 ml of Vibrio cholerae at 0.25 OD600nm WT or mutant Δtgt in triplicates, with or without tobramycin were flash frozen in liquid nitrogen and stored at -80°C. The polysomes were extracted with 200 µl of extraction buffer (20 mM Tris pH8-150 mM Mg(CH3COO)2-100 mM NH4Cl, 5 mM CaCl2-0,4% Triton X100-1% Nonidet P40) added of 2x cocktail anti proteases Roche and 60U RNase Inhibitor Murine to the buffer, DNase I and glass beads (diameter <106 micrometers), vortexed during 30 min at 4°C. The supernatant of this crude extract was centrifugated 10 min at 21 krcf at +4°C. The absorbance was measured at 260nm on 1 µl from 1/10 extract. After 1 hour of digestion at 25°C with 0.75 U MNase/0.025 UA260nm of crude extract, the reaction was stopped by the addition of 3 µl 0.5 M EGTA pH8. The monosomes generated by digestion were purified through a 24% sucrose cushion centrifuged 90 min at 110 krpm on a TLA110 rotor at +4°C. The monosomes pellet was rinsed with 200 µl of resuspension buffer (20 mM Tris-HCl pH 7,4 - 100 mM NH4Cl - 15 mM Mg(CH3COO)2 - 5 mM CaCl2) and then recovered with 100 µl. RNA were extracted by acid phenol at 65°C, CHCL3 and precipitated by Ethanol with 0.3M CH3COONa pH 5.2. Resuspended RNA was loaded on 17% polyacrylamide (19:1); 7 M urea in 1x TAE buffer at 100 V during 6 hours and stained with SYBRgold. RNA fragments corresponding to 28-34 nt were retrieved from gel and precipitated in ethanol with 0.3 M CH3COONa pH 5.2 in presence of 100 µg glycogen. rRNA were depleted using MicrobExpress Bacterial mRNA Enrichment kit from Invitrogen. The supernatant containing the ribosome footprints were recovered and RNA were precipitated in ethanol in presence of glycogen overnight at -20°C. The RNA concentration was measured by Quant-iT microRNA assay kit (Invitrogen). The RNA was dephosphorylated in 3’ and then phosphorylated in 5’ to generate cDNA libraries using the NebNext Small RNA Sample Prep kit with 3’ sRNA Adapter (Illumina) according to the manufacturer’s protocol with 12 cycles of PCR amplification in the last step followed by DNA purification with Monarch PCR DNA cleanup kit (NEB). Library molarity was measured with the Qubit DNAds HS assay kit from Invitrogen and the quality was analyzed using Bioanalyzer DNA Analysis kit (Agilent) and an equimolar pool of the 12 libraries was sequenced by the High-throughput sequencing facility of I2BC with NextSeq 500/550 High output kit V2 (75 cycles) (Illumina) with 10 % PhiX. Sequencing data is available at GSE231087.

Analysis of ribosome profiling data

RiboSeq analysis was performed using the RiboDoc package (52) for statistical analysis of differential gene expression (DEseq2). Sequencing reads are first trimmed to remove adaptors then aligned to the two V. cholerae chromosomes (NC_002505 and NC_002506). Reads aligned uniquely are used to perform the differential gene expression analysis. MNase shows significant sequence specificity at A and T (53). Due to this specificity and A-T biases in V. cholerae genome, ribosome profiling data exhibit a high level of noise that prevents the obtention of a resolution at the nucleotide level.

UV sensitivity measurements

Overnight cultures were diluted 1:100 in MH medium and grown with agitation at 37°C until an OD 600 nm of 0.5-0.7. Appropriate dilutions were then plated on MH agar. The proportion of growing CFUs after irradiation at 60 Joules over total population before irradiation was calculated, doing a ratio with total CFUs. Experiments were performed 3 to 8 times.

Quantification and statistical analysis

For comparisons between 2 groups, first an F-test was performed in order to determine whether variances are equal or different between comparisons. For comparisons with equal variance, Student’s t-test was used. For comparisons with significantly different variances, we used Welch’s t-test. For multiple comparisons, we used ANOVA. We used GraphPad Prism to determine the statistical differences (p-value) between groups. **** means p<0.0001, *** means p<0.001, ** means p<0.01, * means p<0.05. For survival tests, data were first log transformed in order to achieve normal distribution, and statistical tests were performed on these log-transformed data. The number of replicates for each experiment was 3<n<6. Means and geometric means for logarithmic values were also calculated using GraphPad Prism.

Bioinformatic analysis for whole genome codon bias determinations

Data

Genomic data (fasta files containing CDS sequences and their translation, and GFF annotations) for Vibrio cholerae (assembly ASM674v1) were downloaded from the NCBI FTP site (ftp://ftp.ncbi.nlm.nih.gov).

Codon counting

For each gene, the codons were counted in the CDS sequence, assuming it to be in-frame. This step was performed using Python 3.8.3, with the help of the Mappy 2.20 (54) and Pandas 1.2.4 (55,56) libraries.

Gene filtering

Genes whose CDS did not start with a valid start codon were excluded from further computations. A valid start codon is one among ATA, ATC, ATG, ATT, CTG, GTG, TTG, according to the genetic code for bacteria, archaea and plastids (translation table 11 provided by the NCBI at ftp://ftp.ncbi.nlm.nih.gov/entrez/misc/data/gc.prt). Further computations were performed on 3590 genes that had a valid start codon.

Codon usage bias computation

The global codon counts were computed for each codon by summing over the above selected genes. For each gene as well as for the global total, the codons were grouped by encoded amino acid. Within each group, the proportion of each codon was computed by dividing its count by the sum of the counts of the codons in the group. The codon usage bias for a given codon and a given gene was then computed by subtracting the corresponding proportion obtained from the global counts from the proportion obtained for this gene. Codon usage biases were then standardized by dividing each of the above difference by the standard deviation of these differences across all genes, resulting in standardized codon usage biases “by amino acid” (“SCUB by aa” in short). All these computations were performed using the already mentioned Pandas 1.2.4 Python library.

Associating genes to their preferred codon

For each codon group, genes were associated to the codon for which they had the highest “SCUB by aa” value. This defined a series of gene clusters denoted using the “aa_codon” pattern. For instance, “*_TAT” contains the genes for which TAT is the codon with the highest standardized usage bias among tyrosine codons.

Extracting most positively biased genes from each cluster

Within each cluster, the distribution of “SCUB by aa” values for each codon was represented using violin plots. Visual inspections of these violin plots revealed that in most cases, the distribution was multi-modal. An automated method was devised to further extract from a given cluster the genes corresponding to the sub-group with the highest “SCUB by aa” for each codon. This was done by estimating a density distribution for “SCUB by aa” values using a Gaussian Kernel Density Estimate and finding a minimum in this distribution. The location of this minimum was used as a threshold above which genes were considered to belong to the most positively biased genes. This was done using the SciPy 1.7.0 (57) Python library. Violin plots were generated using the Matplotlib 3.4.2 (58) and Seaborn 0.11.1 (59) Python libraries.

Code availability

All codes to perform these analyses were implemented in the form of Python scripts, Jupyter notebooks (60) and Snakemake(61) workflows, and are available in the following git repository: https://gitlab.pasteur.fr/bli/17009_only. Data are available for whole genome codon usage of V. cholerae in excel sheet and V. cholerae codon usage biased gene lists at zenodo public repository with the following doi: 10.5281/zenodo.6875293.

Results

Tobramycin tolerance is decreased in Δtgt without any difference in uptake

We confirmed V. cholerae Δtgt strain’s growth defect in sub-MIC tobramycin (TOB) (Fig. 1A) and that expression of tgt in trans restores growth in these conditions (Fig. 1B). We further tested tolerance to lethal antibiotic concentrations by measuring survival after antibiotic treatment during 15 minutes to 4 hours. As expected, Δtgt is less tolerant than WT to TOB (Fig. 1CD), but had no impact in ciprofloxacin (CIP) or carbenicillin (CRB) (Fig. 1EF). In specific cases, the levels of a given tRNA modification can depend on the presence of other ones(62). One example is the m5C38 modification of tRNAAsp which is favored by the presence of Q in eukaryotes(22,63). However, bacterial tRNAs do not harbor m5C and neither do they harbor mal-Q or gal-Q hypermodifications, unlike in eukaryotes (64). In E. coli, tRNATyr is modified by RluF which introduces a pseudouridine (Ψ) at position 35 of the anticodon, next to the G/Q at position 34 (65). We tested whether the presence or absence of this second modification has an impact on Q-dependent fitness phenotypes. Competition experiments showed no effect of the rluF deletion in any condition (Sup. Fig. S1), showing that the effect of tgt is not linked to an effect of a Ψ modification possibly made by RluF in V. cholerae.

V. cholerae Δtgt shows decreased aminoglycoside tolerance.

A. Competition experiments between WT and Δtgt with the indicated antibiotic or oxidant at sub-MIC concentration. NT: non-treated. TOB: tobramycin 0.6 µg/ml; CIP: ciprofloxacin 0.01 µg/ml; CRB: carbenicillin 2.5 µg/ml; PQ: paraquat 10 µM; H2O2: 2mM. B. Competition experiments between WT and Δtgt carrying the indicated plasmids. MH: non-treated. CDEF. Survival of exponential phase cultures after various times of incubation (indicated in minutes on the X-axis) with the indicated antibiotic at lethal concentration: 5MIC: 5 times the MIC; 10MIC: 10 times the MIC. G. PMF measurement of exponential phase cultures using fluorescent Mitotracker dye, measured using flow cytometry. H. Neomycin uptake measurement by flow cytometry using fluorescent Cy5 coupled neomycin. I. Competition experiments between WT and Δtgt carrying either empty plasmid (p0), or a plasmid overexpressing tRNA-Tyr with the native GUA anti-codon, or with the synthetic AUA anticodon. The anticodon sequence is indicated (e.g. tRNATyrAUA decodes the TAT codon). NT: non-treated. TOB: tobramycin 0.6 µg/ml. For multiple comparisons, we used one-way ANOVA. **** means p<0.0001, *** means p<0.001, ** means p<0.01, * means p<0.05. ns: non-significant. Number of replicates for each experiment: 3<n<8.

We asked whether the growth defect of Δtgt is due to increased aminoglycoside entry and/or a change in proton-motive force (PMF) (32,66). We used a ΔtolA strain as a positive control for disruption of outer membrane integrity and aminoglycoside uptake(67). No changes either in PMF (Fig. 1G), nor in uptake of the fluorescent aminoglycoside molecule neo-Cy5 (Fig. 1H)(33) were detected in the Δtgt strain, indicating that the increased susceptibility of Δtgt to TOB is not due to increased aminoglycoside entry into the V. cholerae cell.

Overexpression of the canonical tRNATyrGUA rescues growth of Δtgt in TOB

We next investigated whether all four tRNAs with GUN anticodon modified to QUN by Tgt, are equally important for the TOB sensitivity phenotype of the Δtgt mutant: Aspartate (Asp)/Asparagine (Asn)/Tyrosine (Tyr)/Histidine (His). The absence of Q could have direct effects at the level of codon decoding but also indirect effects such as influencing tRNAs’ degradation(68). qRT-PCR analysis of tRNATyr levels showed no major differences between WT and Δtgt strains, making it unlikely that the effect of Q modification on codon decoding is caused by altered synthesis or degradation of tRNATyr (Sup. Fig. S2A). The levels of the other three tRNAs modified by Tgt also remained unchanged (Sup. Fig. S2A). These results do not however exclude a more subtle or heterogeneous effect of Q modification on tRNA levels, which would be below the detection limits of the technique in a bacterial whole population.

We next adopted a tRNA overexpression strategy from a high copy plasmid. The following tRNAs-GUN are the canonical tRNAs which are present in the genome: TyrGUA (codon TAC), HisGUG (codon CAC), two isoforms of AsnGUU (codon AAC), AspGUC (codon GAT). The following tRNAs-AUN are synthetic tRNAs which are not present in the genome: TyrAUA, HisAUG, AsnAUU, AspAUC. tRNAPheGAA was also used as non Tgt-modified control. Overexpression of tRNATyr, but not tRNATyr rescues the Δtgt mutant’s growth defect in sub-MIC TOB (Fig. 1I). Overexpression of tRNAHisAUG also seemed to confer a benefit compared to empty plasmid (p0), but not as strong as tRNATyr (Sup. Fig. S2B). We do not observe any major rescue of TOB sensitive phenotypes when the other tRNAs are overexpressed, suggesting that changes in Tyr codon decoding is mostly responsible for the Δtgt mutant’s TOB-susceptibility phenotype.

Q modification influences amino acid incorporation at tyrosine codons

We decided to measure the efficiency of amino acid incorporation at corresponding codons in Δtgt, using gfp reporters. First, we confirmed that GFP fluorescence from native GFP (encoded by gfpmut3) is not affected in Δtgt compared to WT (gfp+ in Fig. 2), indicating that there are no major differences on expression or folding of the GFP in Δtgt. We next constructed gfp fluorescent reporters by introducing within their coding sequence, stretches of repeated identical codons, for Asp/Asn/Tyr/His. This set of reporters revealed that the absence of Q leads to an increase of amino acid incorporation at Tyr TAT codons, both without and with sub-MIC TOB (Fig. 2A NT and TOB). This was not the case for Asp (Fig. 2B), nor for Asn (Fig. 2D), and we observed a slighter and more variable change for His (Fig. 2C). No significant effect of tgt was observed for 2nd near-cognate codons obtained by changing 1 base of the triplet for TAC and TAT codons (Fig. 2E): Phe TTC/TTT, Cys TGT/TGC, Ser TCT/TCC (the 3rd near-cognate stop codons TAA and TAG were not tested in this setup). Thus, Q modification strongly impacts the decoding of Tyr codons, and to a lesser extent His codons in this reporter system.

Codon decoding differences for V. cholerae WT and Δtgt. A. to E. Codon specific translation efficiency in WT and Δtgt using 6xcodon stretches inserted in GFP.

Y axis represents the relative fluorescence of a given GFP in Δtgt over the same construct in WT. NT: non-treated; TOB: tobramycin at 0.4 µg/ml. Each specified codon is repeated 6x within the coding sequence of the GFP (e.g. TACTACTACTACTACTAC). For multiple comparisons, we used one-way ANOVA. **** means p<0.0001, *** means p<0.001, ** means p<0.01, * means p<0.05. ns: non-significant. Number of replicates for each experiment: 3<n, each dot represents one replicate.

The absence of Q decreases misincorporation at Tyr TAT

GFP reporters tested above with codon stretches were pivotal for the identification of codons for which decoding efficiency differs between WT and Δtgt, even though it’s not a natural setup. We next developed a biologically relevant β-lactamase reporter tool to assess differences in the decoding of the tyrosine codons in WT and Δtgt strains. The amino acid Tyr103 of the β-lactamase, was previously shown to be important for its function in resistance to β-lactam antibiotics, such as carbenicillin (6971). We replaced the native Tyr103 TAC with the synonymous codon Tyr103 TAT (Fig. 3). While in the WT, both versions of β-lactamase conferred similar growth in carbenicillin with or without sub-MIC TOB (Fig. 3AB), in the Δtgt strain the Tyr-TAT version grows better than the Tyr-TAC version upon exposure to TOB stress (Fig. 3CD). This suggests a more efficient translation of the Tyr103-TAT β-lactamase mRNA, compared to the native Tyr103-TAC version, in stressed Δtgt strain.

Differential translation at tyrosine codons evaluated by growth on carbenicillin of mutated β-lactamase reporters.

The catalytic tyrosine at position 103 was tested in its native sequence (TAC), or with the synonymous TAT mutation. A to D. Growth on microtiter plates. Growth was followed by measuring the OD 620 nm every 15 minutes during 800 minutes. CARB was used at 100 µg/ml. A and C: not TOB. B and D: TOB at 0.2 µg/ml (20% of MIC).

Presence or absence of a modification could also affect aminoacylation of the tRNA. Both TAT and TAC codons are decoded by the same and only one tRNA, tRNATYRGUA. In this case, a defect in aminoacylation of this tRNA would impact the decoding of both codons. Our results do not support such an aminoacylation problem, because the efficiency of decoding of TAT but not TAC increases in Δtgt, and the difference is clearer in TOB. These results are also consistent with rescue of TOB-sensitive phenotype by tRNATyrGUA but not tRNATyrAUA overexpression in Δtgt.

Altogether, results suggest that in the presence of sub-MIC TOB stress, the absence of Q namely leads to better translation of the TAT vs TAC codon, leading to differential translation of proteins with codon usage biases towards TAC or TAT codons.

Proteomics study identifies RsxA among factors for which translation is most impacted in Δtgt

These observations show a link between Q modification of tRNA, differential decoding of Tyr codons (among others) and susceptibility to aminoglycosides. We hypothesized that proteins that are differentially translated according to their Tyr codon usage could be involved in the decreased efficiency of the response to aminoglycoside stress in Δtgt. We conducted a proteomics study comparing WT vs Δtgt, in the absence and presence of sub-MIC TOB (proteomics Table S1 and Fig. 4). Loss of Q results in generally decreased detection of many proteins in TOB (shift towards the left in the volcano plot Fig. 4AB), and in increases in the levels of 96 proteins. Among those, RsxA (encoded by VC1017) is 13-fold more abundant in the Δtgt strain compared to WT in TOB. RsxA is part of an anti-SoxR complex. SoxR is an oxidative stress response regulator(29) that controls sodA (VC2694, superoxide dismutase) and acrA (VC0913, efflux), among other genes of the regulon. The Rsx complex reduces and inactivates SoxR, preventing the induction of the regulon. Consistently, we find that the levels of SodA and AcrA proteins are decreased in Δtgt compared to WT in TOB (indicated in Fig. 4B). With 83% of Tyr-TAT codons, instead of the expected 53% average, RsxA has a clear codon usage bias. To test whether some of the differentially abundant protein groups in the Q deficient mutant show similar biases, the Tyr codon usage was calculated for the 96 more abundant and 195 less abundant proteins expressed in TOB. More abundant proteins in Δtgt TOB with a codon usage bias towards TAT vs TAC are represented as light blue dots in Fig. 4AB. No statistically significant difference was detected for TAT codon usage in neither sets of proteins. Thus, one cannot draw conclusions or infer predictions about codon decoding efficiencies in a tRNA modification mutant such as Δtgt from the proteomics data alone.

Post-transcriptional regulation in Δtgt and Tyr codon usage bias.

A. and B. Volcano plots showing less and more abundant proteins in proteomics analysis (performed in triplicates) in Δtgt compared to WT during growth without antibiotics (A. where MH is the growth medium), or in sub-MIC TOB at 0.4 µg/ml (B.). C. Codon usage bias calculation example with VC1017 rsxA tyrosine TAC and TAT codons. D. to G. Plots showing codon usage bias for genes lists that are up or down in ribosome profiling analysis (performed in triplicates), for codons decoded by tRNAs with Q modification. Each dot represents one gene of the lists.

We thus performed Ribo-seq (ribosome profiling) analysis on extracts from WT and Δtgt strains grown in presence of sub-MIC TOB. Unlike for eukaryotes, technical limitations (e.g. the RNase which is used displays significant sequence specificity), do not allow to obtain codon resolution in bacteria (72). However, we determined 159 transcripts with increased and 197 with decreased translation in Δtgt and we plotted their standardized codon usage bias for codons of interest (Fig. 4DEFG). The calculation of this value is explained in details in the materials and methods section, and shown for rsxA as example in Fig 4C. Briefly, we took as reference the mean proportion of the codons of interest in the genome (e.g. for tyrosine: TAT= 0.53 and TAC= 0.47, meaning that for a random V. cholerae gene, 53% of tyrosine codons are TAT). For each gene, we calculated the proportion of each codon (e.g. for rsxA: TAT=0.83 and TAC=0.17). We next calculated the codon usage bias as the difference between a given gene’s codon usage and the mean codon usage (e.g. for rsxA, the codon usage bias for TAT is 0.83-0.53=+0.30). Finally, in order to consider the codon distribution on the genome and obtain statistically significant values, we calculated standardized bias by dividing the codon usage bias by standard deviation for each codon (e.g. for rsxA, 0.30/0.25 = +1.20). This is done to adjust standard deviation to 1, and thus to get comparable (standardized) values, for each codon.

Ribo-seq data (Table S2) shows that TAT (and GAT) codon usage was decreased in the list of transcripts with decreased translation in Δtgt, while it was increased in the list of transcripts with increased translation (Fig. 4DE). No difference was detected for AAT and CAT (Fig. 4FG). Data is thus consistent with more efficient translation of TAT codons in Δtgt.

In addition to mistranslation and codon decoding efficiency, other factors also influence detected protein levels, such as transcription, degradation, etc. Moreover, the localization and sequence context of the codons for which the efficiency of translation is impacted, may be important. Nevertheless, as translation of proteins with a codon usage bias towards TAC or TAT may be impacted in Δtgt, and as the most abundant protein RsxA in Δtgt in TOB shows a strong TAT bias, we decided to evaluate whether RsxA is post-transcriptionally regulated by the Q modification and whether it may affect fitness in the presence of TOB.

RsxA is post-transcriptionally upregulated in Δtgt due to more efficient decoding of tyrosine TAT codons in the absence of Q modification

Transcriptomic analysis comparing at least 2-fold differentially expressed genes between V. cholerae Δtgt and WT strains (Table S3) showed that, respectively, 53 and 26 genes were significantly downregulated in MH and sub-MIC TOB, and 34 were up in sub-MIC TOB. Gene ontology (GO) enrichment analysis showed that the most impacted GO categories were bacteriocin transport and iron import into the cell (45- and 40-fold enriched) in MH, and proteolysis and response to heat (38- and 15-fold enriched) in TOB. In both conditions, the levels of rsxA transcript remained unchanged.

RsxA carries 6 tyrosine codons among which the first 5 are TAT and the last one is TAC. RsxA is 13-fold more abundant in Δtgt than WT, but transcript levels measured by digital RT-PCR are comparable in both strains (Fig. 5A), consistent with RNA-seq data. We constructed transcriptional and translational gfp fusions in order to evaluate the expression of rsxA in WT and Δtgt strains. As expected from digital RT-PCR results, no significant differences in fluorescence were observed for the transcriptional fusion of the rsxA promoter with gfp (Fig. 5B), excluding transcriptional regulation of rsxA in this context. For translational fusions, we used either the native rsxA sequence bearing 5 TAT + 1 TAC codons, or a mutant rsxA allele carrying all 6 TAC codons (hereafter called respectively RsxATAT and RsxATAC). Confirming the proteomics results, the RsxATAT-GFP fusion was more fluorescent in the Δtgt mutant, but not the RsxATAC-GFP one (Fig. 5C and detailed flow cytometry data in Sup. Fig. S3ABC). Since increased rsxA expression appeared to be somewhat toxic for growth, and in order to test translation on a reporter which confers no growth defect, we chose to test directly the translation of gfp, which originally carries 4 TAT (36%) and 7 TAC (64%) codons in its native sequence. We constructed two synonymous versions of the GFP protein, with all 11 tyrosine codons either changed to TAT or to TAC. Similar to what we observed with rsxA, the GFPTAT version, but not the GFPTAC one, generated more fluorescence in the Δtgt background, (Fig. 5C and detailed flow cytometry data in Sup. Fig. S3DEF).

Post-transcriptional upregulation of RsxA in Δtgt due to a Tyr codon bias towards TAT and toxicity in sub-MIC TOB.

A. rsxA mRNA levels measured by digital RT-PCR. B. Transcriptional expression levels from the rsxA promoter measured by flow cytometry. C. Translational fusion of rsxA to gfp and gfp alone, with differences in codon usage. *** means p<0.001, ** means p<0.01. ns: non-significant. D. Relative OD 620 nm of WT strain carrying either empty plasmid or plasmid overexpressing RsxA comparing growth in sub-MIC TOB divided by growth in the absence of treatment.

Since not all TAT biased proteins are found to be enriched in Δtgt proteomics data, the sequence context surrounding TAT codons could affect their decoding. Overall, our results demonstrate that RsxA is upregulated in the Δtgt strain at the translational level, and that proteins with a codon usage bias towards tyrosine TAT are prone to be more efficiently translated in the absence of Q modification, but this is also dependent on the sequence context.

Increased expression of RsxA hampers growth in sub-MIC TOB

We asked whether high levels of RsxA could be responsible of Δtgt strain’s increased sensitivity to TOB. rsxA cannot be deleted since it is essential in V. cholerae (see our TN-seq data (1,73)). We overexpressed rsxA from an inducible plasmid in WT strain (Fig. 5D and Sup. Fig. S4ABCD). In the presence of sub-MIC TOB, overexpression of rsxA in the WT strain strongly reduces growth (Sup. Fig. S4B, black curve compared to blue), while overexpression of tgt restores growth of the Δtgt strain (Sup. Fig. S4D, green curve). This shows that increased rsxA levels can be toxic during growth in sub-MIC TOB and is consistent with decreased growth of the Δtgt strain.

Unlike for V. cholerae, rsxA is not an essential gene in E. coli, and does not bear a TAT bias. It has however the same function. In order to confirm that the presence of RsxA can be toxic during growth in sub-MIC TOB, we additionally performed competition experiments in E. coli with simple and double mutants of tgt and rsxA. Since Δtgt strain’s growth is more affected than WT at TOB 0.5 µg/ml (indicated with an arrow in Sup. Fig. S4E), we chose this concentration for competition and growth experiments. The results confirm that inactivation of rsxA in Δtgt restores fitness in sub-MIC TOB (Sup. Fig. S4F), and that overproduction of RsxA decreases growth in TOB.

tgt transcription is repressed by CRP and induced by tobramycin in V. cholerae

tgt was previously observed to be upregulated in E. coli isolates from urinary tract infection(74) and in V. cholerae after mitomycin C treatment (through indirect SOS induction(47)). We measured tgt transcript levels using digital RT-PCR in various transcriptional regulator deficient mutants (iron uptake repressor Fur, general stress response and stationary phase sigma factor RpoS and carbon catabolite control regulator CRP), as well as upon exposure to antibiotics, particularly because tgt is required for growth in sub-MIC TOB. We also tested the iron chelator dipyridyl (DP), the oxidant agent paraquat (PQ) and serine hydroxamate (SHX) which induces the stringent response.

Among all tested conditions, we found that sub-MIC TOB and the stringent response increase tgt transcript levels, while the carbon catabolite regulator CRP appears to repress it (Fig. 6A). We found a sequence between ATG -129 to -114: TTCGCAGGGAAACGCG which shows some similarity (in blue) to the V. cholerae CRP binding consensus (T/A)1(G/T)2(T/C)3G4(A/C)5NNNNNN(T/G)12C13(A/G)14(C/A)15(T/A)16. However, CRP binding was not previously detected by ChIP-seq in the promoter region of tgt in V. cholerae(75). CRP binding could be transitory or the repression of tgt expression by CRP could be an indirect effect.

Regulation of tgt expression and tRNA Q levels.

A: tgt transcript levels measured by digital-RT-PCR. Y-axis represents relative abundance compared to the non-treated (NT) condition in WT. For multiple comparisons, we used one-way ANOVA (for A. C. E.). **** means p<0.0001, *** means p<0.001, ** means p<0.01. Only significant differences are shown. B. Stringent response induction measured with P1rrn-gfp reporter shown as fluorescence (Y-axis) as a function of growth (X-axis: OD 600 nm). C. Q levels in tRNA enriched RNA extracts, measured by mass spectrometry. D. Northern blot and quantification of Q levels in tRNATyr. The lower band in the gel corresponds to unmodified tRNATyr. The upper band corresponds to Q-modified tRNA-Tyr. Δtgt is the negative control without modification of tRNATyr. Histograms show the quantification of Q-modified tRNATyr over total tRNATyr, as follows: Q-modified tRNATyr/(Q-modified tRNATyr + tRNATyr without Q)= upper band/(upper band + lower band). 2.5 µg of in tRNA enriched RNA extracts were deposited in lanes WT NT, WT TOB and Δcrp. 0.9 µg was deposited in lanes Δtgt. Number of replicates for each experiment: 3

Regarding induction by sub-MIC TOB, the mechanism remains to be determined. We previously showed that sub-MIC TOB induces the stringent response(1,76). Since induction of tgt expression by SHX and by TOB seems to be in the same order of magnitude, we hypothesized that sub-MIC TOB could induce tgt through the activation of the stringent response. Using a P1rrnB-gfp fusion (1), which is down-regulated upon stringent response induction(77) (Fig. 6B), we found that the stringent response is significantly induced by sub-MIC TOB, both in WT and Δtgt. This indicates that sub-MIC TOB possibly induces tgt expression through the stringent response activation.

Q modification levels can be dynamic and are directly influenced by tgt transcription levels

We have identified conditions regulating tgt expression. We next addressed whether up/down-regulation of tgt affects the actual Q modification levels of tRNA. We measured Q levels by mass spectrometry in WT and the Δcrp strain, where the strongest impact on tgt expression was observed (Fig. 6C). We find a significant 1.6-fold increase in Q levels in Δcrp. We also tested the effect of sub-MIC TOB, but smaller differences are probably not detected using our approach of mass spectrometry in bulk cultures.

In order to get deeper insight into modification level of V. cholerae tRNAs potentially having Q34 modification, we decided to adapt a recently published protocol for the detection and quantification of queuosine by deep sequencing (50). This allowed us to validate the presence of Q34 modification in the V. cholerae tRNAs Asp, His and Asn_GTT2 and precisely measure its level and modulation under different growth conditions (Fig. S5). We also showed that Q34 detection is robust and reproducible, and reveals increased Q34 content for tRNAHis and tRNAAsn in Δcrp strain where tgt expression was induced, while for tRNAAsp Q34 level remains relatively constant. V. cholerae tRNAAsn_GTT1 is very low expressed and likely contains only sub-stoichiometric amounts of Q34, while analysis of tRNATyr is impeded by the presence of other modifications in the anticodon loop (namely i6A37 or its derivatives), which prevents the correct mapping and quantification of Q34 modifications using deletion signature. In order to evaluate Q34 levels in tRNATyr more specifically, we performed APB northern blots allowing visualization and quantification of Q-modified and unmodified tRNAs (51). As anticipated from increased tgt expression, Q-modified tRNA levels were strongly increased in Δcrp strain. Sub-MIC TOB also increases the proportion of Q containing tRNATyr compared to the non-treated condition (Fig. 6D). However, this result was variable suggesting a subtle fine-tuning of the Q levels depending on growth state (optical density) and TOB concentration. These results show that tRNA Q modification levels are dynamic and correlate with variations in tgt expression, depending on the tRNA.

DNA repair genes are TAT-biased

We further analyzed in silico the codon usage of V. cholerae genome, and for each gene, we assigned a codon usage value to each codon (Fig. 4C and doi:10.5281/zenodo.6875293). This allowed the generation of lists of genes with divergent codon usage, for each codon.

For genes with a tyrosine codon usage bias towards TAT in V. cholerae, gene ontology enrichment analysis (Sup. Fig. S6CDE) highlights the DNA repair category with a p-value of 2.28×10-2 (Sup. Fig. S6C.). Fig. 7A shows Tyr codon usage of V. cholerae DNA repair genes. We hypothesized that translation of DNA repair transcripts could be more efficient in Δtgt, and that such basal pre-induction would be beneficial during genotoxic treatments as UV irradiation (single stranded DNA breaks). UV associated DNA damage is repaired through RecA, RecFOR and RuvAB dependent homologous recombination. Five of these genes, recO, recR, recA and ruvA-ruvB, are biased towards TAT in V. cholerae (Fig. 7A, red arrows), while their repressor LexA bears a strong bias towards TAC. DNA repair genes (e.g. ruvA with 80% TAT, ruvB with 83% TAT, dinB with 75% TAT) were also found to be up for Δtgt in the Ribo-seq data, with unchanged transcription levels. V. cholerae Δtgt appears to be 4 to 9 times more resistant to UV irradiation than the WT strain (Fig. 7B). This is consistent with increased DNA repair efficiency in the V. cholerae Δtgt strain.

DNA repair after UV irradiation is more efficient in V. cholerae Δtgt.

A and C. Tyrosine codon usage of DNA repair genes A: in V. cholerae. C: in E. coli. Red indicates positive codon usage bias, i.e. TAT bias. Blue indicates negative codon usage bias for TAT, i.e. TAC bias. B and D. Survival of Δtgt relative to WT after UV irradiation (linear scale) B: in V. cholerae. D: in E. coli. For multiple comparisons, we used one-way ANOVA. **** means p<0.0001, ns: non-significant.

We also analyzed tyrosine codon usage for the DNA repair genes in E. coli, and did not observe the same bias (Fig. 7C), with 51% TAT bias, i.e. the expected level for a random group of genes of the E coli genome, and with a TAC bias for recOR and recA (red arrows) and strong TAT bias for lexA (blue arrow) (Sup. Fig. S6B, whole genome E. coli). These genes thus show the exact opposite bias in E. coli (Fig. 7C). Strikingly, unlike for V. cholerae, E. coli Δtgt mutant did not show increased UV resistance (Fig. 7D). This is consistent with the hypothesis that modification-tuned translation of codon biased transcripts can be an additional means of regulation building upon already described and well characterized transcriptional regulation pathways.

Discussion

We show here that Q modification levels can be dynamic in bacteria and respond to external conditions; and that Q levels on V. cholerae tRNATyr correlate with tgt expression. This is clearer in conditions where tgt transcription is highly induced (Δcrp), and more variable in conditions where this induction is intermediate (sub-MIC TOB). As summarized in Fig. 8, we propose that exposure to sub-MIC aminoglycosides increases tgt expression in V. cholerae, and impacts the decoding of tyrosine codons. The TAT biased transcripts RsxA is an anti-SoxR factor. SoxR controls a regulon involved in oxidative stress response and sub-MIC aminoglycosides trigger oxidative stress in V. cholerae(78). A link between Q and oxidative stress has also been previously found in eukaryotic organisms (24). Better decoding at TAT versus TAC codons in the absence of Q observed here in V. cholerae is also consistent with recent findings in human tRNAs, where the presence of Q increases translation of transcripts biased in C-ending codons (79). Our findings are in accordance with the concept of the so-called modification tunable transcripts (MoTTs)(20). We thus show that a proteins’ codon content can influence its translation in a Q modification dependent way, and that this can also impact the translation of antibiotic resistance genes (here β-lactamase). Note that, our results do not exclude the involvement of additional Q-regulated MoTTs in the response to sub-MIC TOB, since Q modification leads to reprogramming of the whole proteome. Finally, we show that we can predict in silico, candidates for which translation can be modulated by the presence or absence of Q modification (e.g. DNA repair genes), which was confirmed using phenotypic tests (UV resistance).

Model.

Upon exposure to sub-MIC aminoglycosides, the expression of tgt is up-regulated in V. cholerae and influences the decoding of tyrosine TAC vs TAT codons. This leads to differential translation from transcripts bearing a codon usage bias for tyrosine codons. The rsxA transcript bears a tyrosine codon bias and its translation can be tuned by tRNA Q modification. RsxA is an anti-SoxR factor. SoxR controls a regulon involved in oxidative stress response. When RsxA levels are high, SoxR is increasingly inactivated and oxidative stress response efficiency decreases. It has been previously shown that sub-MIC aminoglycosides trigger oxidative stress in V. cholerae(78). Increasing RsxA levels thus reduces fitness in TOB by hampering efficient oxidative stress response. As a corollary, decreased RsxA would lead to increased expression of the SoxR regulon, which would allow for more efficient response to oxidative stress, and increased fitness in the presence of sub-MIC TOB. We propose that when Tgt/Q levels in tRNA increase, RsxA synthesis is low and active SoxR levels are high, facilitating the bacterial response to aminoglycoside dependent oxidative stress.

Essential/housekeeping genes are generally TAC biased (Sup. Fig. S6AB), as well as ribosomal proteins, which carry mostly tyrosine TAC codons both in V. cholerae and E. coli. It has been proposed that codon bias corresponding to abundant tRNAs at such highly expressed genes, guarantees their proper expression and avoids titration of tRNAs, allowing for efficient expression of the rest of the proteome(80). Induction of tgt by stress could also possibly be a signal for the cell to favor the synthesis of essential factors. Our results are also consistent with the fact that synonymous mutations can influence the expression of genes(81).

V. cholerae is the model organism for different species of Vibrio. We have previously shown that V. cholerae’s response to sub-MIC antibiotic stress is transposable to other Gram-negative pathogens (78,82), while there are differences between E. coli and V. cholerae, in the response to sub-MIC antibiotics and oxidative stress phenotypes (78,82,83). Here, we have also addressed some of the effects of TOB in E. coli Δtgt mutant. In E. coli, the deletion of tgt has a less dramatic effect on the susceptibility to TOB ((1) and Sup. Fig S4F). V. cholerae and E. coli globally show similar tyrosine codon usage in their genomes (Sup. Fig. S6AB). However, E. coli rsxA does not display a codon bias towards TAT, and neither do DNA repair genes. One can think that in regard to MoTTs, different organisms have evolved according to the environments where they grow, selecting the integration of specific stress response pathways under specific post-transcriptional regulations.

It was recently shown that E. coli Δtgt strain is more resistant than the WT to nickel toxicity, most certainly because the nickel importer genes nikABCDE are less expressed, but the underlying molecular mechanism had not been elucidated(28). As NikR, the repressor of the nik operon is enriched in TAT codons (100%), a more efficient translation of the nikR gene in the absence of Q would lead to the observed repression phenotype. In addition, the nickel exporter gene rcnA is also enriched is TAT (100%), while one of the genes for subunits of the nickel importer nikD is enriched in TAC codons (100%). In combination this could explain the clear resistance of the tgt strain to high levels of nickel.

However, protein levels are not always in line with the codon bias predictions. The positions of the codons of interest and their sequence context may also be important for differential translation. The presence of the codon of interest in the 5’-end vs 3’-end of a transcript could have a bigger impact on the efficiency of translation (84,85). A recent study testing TAC/TAT codons placed between two genes in a translational fusion yielded different results compared to our constructs with the tested codons at the 5’ of the transcript(86). Similarly, the distance between two codons of interest, or the identity of the nearby codons may be important. The translation of highly transcribed genes and genes with low levels of mRNAs could be dissimilar. Codon usage may also directly impact gene expression at mRNA levels with an effect on transcription termination(87), especially for constitutive genes. Thus, the search for MoTTs could be facilitated by comparing transcriptomics to proteomics data, and additional experiments need to be performed to elucidate post-transcriptional regulation-related phenotypes, but the differential expression of specific TAT/TAC biased proteins finally allows to propose a model for the pleiotropic phenotype caused by Q deficiency in E. coli.

Studies, mostly in eukaryotes, reveal that tRNA modifications are dynamic and not static as initially thought (8890). Modification levels depend on growth (91,92), environmental changes (93) and stress (reviewed in (62)). Stress regulated tRNA modification levels have an impact on the translation of regulators, which in turn trigger translational reprogramming and optimized responses to stress (2,94,95). We show here that tgt expression is regulated by sub-MIC TOB, the stringent response and CRP, and that tRNATyr Q modification levels increase with tgt expression. The fact that such correlation between tgt expression and Q levels does not occur for all tRNAs (e.g. tRNAAsp), indicates that other parameters also influence Q modification levels. One possibility is that other modifications, such as those on the anticodon loop of tRNATyr, may influence the way Tgt modifies these tRNAs, as documented for other modification circuits (22,96). Tgt may also bind these tRNAs differently (for a review on modification specificity (62)). Since sub-MIC TOB triggers the stringent response, one hypothesis could be that sub-MIC TOB induces tgt through stringent response activation. The stringent response is usually triggered upon starvation, for example when amino acids are scarce. tgt expression was also recently shown to be regulated by tyrosine levels and to affect tRNA-Tyr codon choice in Trypanosoma brucei(97).

Regarding CRP, the carbon catabolite regulator, it represses transcription of tgt. Interestingly, V. cholerae crp carries only TAC codons for tyrosine, and is strongly down-regulated in Δtgt in our Ribo-seq data, while its transcription levels remain unchanged. The downregulation of CRP translation when Q modifications are low (as in the Δtgt strain), could be a way to de-repress tgt and increase Q modification levels. Note that CRP is involved in natural competence of V. cholerae, during growth on crustacean shells where horizontal gene transfer occurs. One can thus speculate that during exogenous DNA uptake, tgt repression by CRP could lead to better decoding of AT rich (i.e. TAT biased) mRNAs. Thus, modulation of tgt levels during natural transformation may modulate the expression of horizontally transferred genes, which by definition may bear different GC content and codon usage. Moreover, if tgt expression is repressed by CRP during competence state, this would favor the translation of TAT-biased DNA repair genes and possibly recombination of incoming DNA into the chromosome. Translational reprogramming in response to DNA damage can thus be an advantageous property selected during evolution.

Our results also demonstrate that we can identify other Q-dependent MoTT candidates using in silico codon usage analysis. In fact, since we now have extensively calculated the codon usage biases at all codons for V. cholerae and E. coli genes, this approach is readily adaptable to any tRNA modification for which we know the differentially translated codons. Such regulation may be a possible way to tune the expression of essential or newly acquired genes, differing in GC-content. It may also, in some cases, explain antibiotic resistance profiles in bacterial collections with established genome sequences, and for which observed phenotypic resistance does not always correlate with known resistance factors(98). Further studies are needed to characterize the determinants of tRNA modification-dependent translational reprogramming.

Acknowledgements

We thank Dr. Francesca Tuorto for sharing the protocol for APB northern blots. Many thanks to Dr. Paola Arimondo for discussions about setting up the queuosine mass spectrometry experiments. We also thank, for RNA-seq experiments, E. Turc, L. Lemée, T. Cokelaer, Biomics Platform, C2RT, Institut Pasteur, Paris, France, supported by France Génomique (ANR-10-INBS-09) and IBISA. We would like to acknowledge Valérie Bourguignon (IMoPA, UMR7365 CNRS-UL) for her help in validation of queuosine modification protocol, and Enora Corler for her help with Ribo-seq analysis.

This research was funded by the Institut Pasteur, the Centre National de la Recherche Scientifique (CNRS-UMR 3525), ANR ModRNAntibio (ANR-21-CE35-0012), ANR-LabEx [ANR-10-LABX-62-IBEID], the Fondation pour la Recherche Médicale (FRM EQU202103012569), the Institut Pasteur grant PTR 245-19 and by the National Institute of General Medical Sciences (NIGMS) grant GM70641 to V dC-L. AB was funded by Institut Pasteur Roux-Cantarini fellowship. The authors acknowledge a DIM1Health 2019 grant from the Région Ile de France to the project EpiK for the LCMS equipment.

Author contributions

Conceived and designed the analysis: ZB, DM; Collected the data: ZB, LF, AB, AC, ML, MD, BL, QG-G, FB, CF, GS, IH; Contributed data or analysis tools: BL, ON, CF, MM, MD, QG-G, FB, HA, VdC-L, VM, YM; Performed the analysis: ZB, LF, AB, AC, ML, BL; Wrote the paper: ZB, VdC-L.

Supplementary Figures

No effect of rluF deletion on fitness in V. cholerae WT and Δtgt strains.

In vitro competition experiments of V. cholerae WT and mutant strains in the absence or presence of sub-MIC aminoglycosides tobramycin TOB and gentamicin (GEN) (50% of the MIC, TOB 0.6 μg/ml, GEN 0.5 µg/ml). MH: no antibiotic treatment. The Y-axis represents log10 of competitive index calculated as described in the methods. A competitive index of 1 indicates equal growth of both strains. The X-axis indicates growth conditions. For multiple comparisons, we used one-way ANOVA on GraphPad Prism. **** means p<0.0001, * means p<0.05. Number of replicates for each experiment: 3<n<10.

Impact of tRNA overexpression on fitness during growth in sub-MIC TOB.

A. Absence of tgt does not visibly affect endogenous tRNA-GUN levels. qRTPCR. A: relative tRNA abundance in Δtgt compared to WT strain in the absence of treatment (NT) and in the presence of sub-MIC tobramycin (TOB). B. In vitro competition experiments between V. cholerae WT and Δtgt strains, carrying a plasmid overexpressing the indicated tRNA; in the absence (NT) or presence of sub-MIC tobramycin (50% of the MIC, TOB 0.6 μg/ml). NT: no antibiotic treatment. The Y-axis represents log10 of competitive index calculated as described in the methods. A competitive index of 1 indicates equal growth of both strains. The X-axis indicates which tRNA is overexpressed, from the high copy pTOPO plasmid. The anticodon sequence is indicated (e.g. tRNATyrAUA decodes the TAT codon). The following tRNAs-GUN are the canonical tRNAs which are present in the genome, and modified by Tgt: TyrGUA, HisGUG, AsnGUU, AspGUC. The following tRNAs-AUN are synthetic tRNAs which are not present in the genome: TyrAUA, HisAUG, AsnAUU, AspAUC. tRNAPheGAA is used as non Tgt-modified control. p0 is the empty plasmid. For multiple comparisons, we used one-way ANOVA on GraphPad Prism. **** means p<0.0001, * means p<0.05. Number of replicates for each experiment: 3<n<8. In A, only significant differences are indicated. ns means non-significant.

Post-transcriptional upregulation of RsxA in Δtgt due to a Tyr codon bias towards TAT and toxicity in sub-MIC TOB. ABC.

Translational fusion of TAC and TAT versions of rsxA to gfp (A. 4 hours exponential phase cultures, B. overnight stationary phase cultures). DEF. TAC and TAT versions of gfp alone (C. 4 hours exponential phase cultures without antibiotics, D. in the presence of sub-MIC TOB), measured by flow cytometry. Y-axis represents mean fluorescence. C and F show representative examples of fluorescence observed with indicated reporters.

rsxA levels impact growth in the presence of sub-MIC TOB.

ABCD. Growth curves in microtiter plate reader, in V. cholerae in indicated conditions. P-tgt+: pSEVA expressing tgt. P-rsx+: pSEVA expressing rsxA. P-control: empty pSEVA. E. Growth curves in microtiter plate reader, in E. coli WT and Δtgt in MH or in the presence of tobramycin at the indicated concentrations (in µg/mL). F. Left: Competition experiments in E. coli, with indicated strains. Right: Relative OD 620 nm of indicated strain comparing growth in TOB (0.4 µg/ml) divided by growth in the absence of treatment.

Q detection and quantification by IO4- oxidation coupled to deep sequencing. The Y-axis represents deletion score, which correlates with the Q34 level in tRNA. Results for tRNAs His, AsnGUU2 and Asp are shown. Analysis was performed for three biological replicates (shown as colored dots) and at least two technical replicates for each strain (only mean of technical replicates is shown). p-values are calculated by two-tailed T-test, error bars correspond to standard error of the mean.

Codon usage.

A and B: Tyrosine codon usage in A: V. cholerae and B: E. coli in the whole genome and selected groups of genes. Red indicates positive codon usage bias, i.e. TAT bias. Blue indicates negative codon usage bias for TAT, i.e. TAC bias. CDE. Gene ontology enrichment analysis in V. cholerae of C: overrepresented gene categories with TAT usage bias. D: underrepresented gene categories with TAT usage bias. E: overrepresented gene categories with TAC usage bias.

Proteomics identify differentially abundant proteins

Ribosome profiling.

RNA-seq V. cholerae WT/Δtgt, in MH and TOB. Only significant differences higher than 2-fold are shown.

Primers, plasmids and strains