Escherichia coli strains used in this study are shown in Supplementary Table 1. TransforMax™ EC100D™ pir + E. coli (LGC Biosearch Technologies) was used for construction of pRIPR plasmids and NEB5α (New England Biolabs) for all other plasmids. MFDpir (Ferrieres et al., 2010) was used as the donor strain for transfer of pRIPR plasmids by conjugation. All cells were made chemically competent using transformation and storage solution (Chung et al., 1989). Bacteria were grown using Miller Broth Base (Fisher Scientific) and supplemented with antibiotics ampicillin (Amp) (100 μg/mL), chloramphenicol (Cam) (30 μg/mL) and kanamycin (Kan) (30 μg/mL), (GoldBio). Other reagents included 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal), Isopropyl-β-D-1-thiogalactopyranoside (IPTG), glucose (40% w/v), diaminopimelic acid (Fisher Scientific), and anhydrotetracycline-hydrochloride (aTc) (Sigma-Aldrich). Q5 High-Fidelity polymerase for PCR and Nt.BspQI nickase were obtained from New England Biolabs.

An assay using a penicillinase from Bacillus cereus (Sigma-Aldrich) was used to test recombinants for antibiotic persistence and conducted as described previously (Slattery et al., 2013). DNA purifications were performed using Monarch PCR and DNA Cleanup Kit and Monarch Plasmid Miniprep Kit (New England Biolabs). Genomic DNA extractions were generated with MasterPure Complete DNA and RNA Purification Kit (formerly Epicentre) with Sanger sequencing carried out at the Iowa State University DNA Facility. Primers and synthetic DNA fragments were ordered from Integrated DNA Technologies, Coralville, IA, USA.

The R6Kγori plasmid pVIK107 (Kalogeraki and Winans, 1997) was modified by first deleting the lac operon genes on the vector by inverse PCR using primers VIKdel.S and VIKdel.AS (Supplementary Table 2). Primers were synthesized with a 5′ phosphate when used for inverse PCR reactions. The 6.3-kbp linearized PCR product was gel purified, re-circularized by ligation and transformed into EC100D™ pir+. Next, a 1-kbp cat cassette was amplified from pKD3 (Datsenko and Wanner, 2000) using primers KD3-Sce-SacI.S and KD3-Sce-SacI.AS, which introduced an I-SceI meganuclease recognition site at the 5′ end, along with SacI restriction sites at the 5′ and 3′ ends of the PCR products. The purified cassette was then introduced into the modified pVIK107 by digestion with SacI and ligation. Inverse PCR was then performed directly upstream and downstream of the neo gene selectable marker with primer pair VIK-NheI.S and VIK-NheI.AS to introduce flanking NheI restriction sites. The linear PCR product was gel purified, digested with NheI and re-circularized by ligation, resulting in the replacement of the neo gene with a single NheI site. We then generated a kan cassette using pKD4 (Baba et al., 2006) as a template using primers BspQI-KD4.S and BspQI-KD4.AS to introduce BspQI and StuI restriction sites. The PCR products were purified, digested with NheI and ligated into the modified pVIK107 vector to yield pRIPR (Figure 1).

To generate derivatives of pRIPR for genome editing, DNA fragments containing the desired mutation, including specific base pair changes or deletions flanked by homologous DNA, were created by synthetic gBlock gene fragments (Integrated DNA Technologies, Coralville, IA, USA) or by PCR amplification of E. coli genomic DNA. To facilitate cloning into pRIPR, we used ligation-independent cloning using Nt.BspQI nickase enzyme, as described previously (Oster and Phillips, 2011). For this, BspQI nickase recognition sites were introduced to the 5′ ends of gBlocks and PCR products using primers with 5′ extensions that included the enzyme sites (Supplementary Table 2). pRIPR was linearized with StuI to remove the Kan^R cassette and cloning reactions were transformed into EC100D™ pir+ and colonies selected on LB agar plates containing Cam. Transformants were patched on LB agar + Kan to test for loss of Kan^R. Plasmid DNA isolated from Kan^S colonies and the presence of an insert of correct size were confirmed by NheI digestion.

Since the RIPR technique does not require exogenous recombinases, the plasmid pSLTS (Kim et al., 2014) was digested with blunt-end restriction enzymes EcoRV and StuI and religated to remove the λ-Red recombineering genes (gam, beta, and exo) to yield pSceH.

pRIPR constructs were generated to target either lac or metG on the E. coli chromosome for mutagenesis. pRIPR-derivative plasmids were transformed into MFDpir by selecting colonies on LB agar plates supplemented with Cam and 150 mM DAP. Transformants were used as donors in conjugation reactions with BW30270 recipients as described (Ferrieres et al., 2010).

Cam^R recombinants were colony purified and transformed with the helper plasmid pSceH by selection on LB + Amp plates spread with 50 μL of anhydrotetracycline (100 μg/mL) and incubated overnight at 30°C. Single colonies were then purified by streaking onto LB and LB+Cam plates to screen for recombinants that had lost Cam^R. Colonies were again restreaked at 37°C and tested to ensure loss of pSceH. Recombinants were tested for new phenotypes by culturing on LB plates spread with X-Gal and IPTG (Supplementary Figure 1) to identify the desired recombinants and metG mutants tested by penicillinase persister assays (Slattery et al., 2013). To further confirm the phenotypes of high persister mutants, the tolerance disk test (TD test) was performed (Gefen et al., 2017). All recombinants were also tested by PCR and Sanger sequencing to confirm the predicted genetic changes.

For CRISPRi experiments, single guide RNA (sgRNA) plasmids were constructed using a derivative of pgRNA-bacteria (Qi et al., 2013) modified to replace the ColE1 replication origin with the ColE1-compatible replicon from RSF1030 (Phillips et al., 2000). Twenty base pair sequences specifying guide RNAs targeting the aaRS genes for alanine (alaS) and methionine (metG) were introduced by inverse PCR (Supplementary Table 2) and religation. Plasmids were confirmed by DNA sequencing and then transformed into LC-E75 (Cui et al., 2018) for phenotypic characterization.

Growth curves for wild-type E. coli and aminoacyl-tRNA synthetase CRISPRi knock down strains were performed by starting overnight cultures in Lysogeny broth (LB) medium with no antibiotic added. The next morning overnight cultures were diluted 1:100 into fresh LB medium and were aliquoted into a 96-well plate. Growth curves were started in a Tecan Spark microplate reader at 37°C, with the plate shaking, and optical density (OD) was measured at 600 nm every 15 min for 20 h. After 1 h of growth, anhydrous tetracycline (aTc) was added to the 96-well plate to final concentrations of 0, 0.25, 0.5, 1.0, 1.5, 2.0, 2.5, and 4.0 μg/mL. Growth curves for wild-type E. coli and metG variant strains were measured by growing cultures in LB medium with a 1:100 dilution from an overnight culture. The OD at 600 nm was measured every 30 min for 8 h. Colony forming units were measured against different concentrations of aTc to determine the optimal knock down efficiency of each individual guide RNA used for CRISPRi knock down. Briefly, overnight cultures were started the same as with the growth curves. Cultures were grown in LB medium with 1:100 dilution from an overnight culture. Bacterial cultures were grown at 37°C, shaking at 250 rpm. After 1 h of growth, aTc was added to the culture to a final concentration of 0, 0.5, 1.0, 1.5, 2.0, and 2.5 μg/mL. The cultures continued to grow for another 3 h with shaking. At this point the OD at 600 nm was measured and 1 mL of culture was taken out and serial diluted and plated on LB plates to count CFU/mL for each concentration of aTc used.

Minimum duration of killing (MDK) bacterial persistence assays were performed for wild-type and variant metG E. coli as previously described (Wood et al., 2021). Endpoint bacterial persistence assays were performed for the aminoacyl-tRNA synthetase CRISPRi knock down strains (wild-type, empty vector, alaS and metG) by starting an overnight culture in LB with no addition of antibiotic. The next morning cultures were started in fresh LB with a 1:100 dilution from the overnight culture. Cultures were grown at 37°C, shaking at 250 rpm, until the OD at 600 nm reached ∼0.1 (∼1 h). aTc was added to a final concentration of 2.0 μg/mL for wild-type and empty vector strains, and 0.5 μg/mL for the alaS and metG knock down strains (determined by growth curves and CFU/mL). The cultures continued to grow at 37°C with shaking for another 3 h. At this point 1 mL of culture was taken out and serial diluted and plated on LB plates to count colony forming units (CFU). HA filter disks were placed on LB plates supplemented with 100 μg/mL ampicillin and 100 mL of a 1:10 dilution was aliquoted onto the filter disk. These plates were incubated overnight at 37°C. The next day the filter disks were transferred to fresh LB plates and incubated overnight again at 37°C. Colonies were then counted to obtain CFU. To calculate persisters the CFU of survivors after ampicillin treatment was divided by the CFU of the culture with no ampicillin treatment.

Overnight cultures of the wild-type and metG varient strains were grown in M9 minimal media supplemented with 50 μM Methionine and back-diluted to an OD at 600 nm of 0.05 in the same media. Cultures were grown at 37°C until they reached an OD at 600 nm of 0.7, at which point 5 μL of 35S-Methionine was added (10.25 mCi/mL). From this point onward, aliquots of 50 μL were taken every 2 min, dispersed onto a filter pad soaked with 5% trichloroacetic acid (TCA) plus 5 μM methionine and washed 3 times in 5% TCA. Radioactivity was counted by a liquid scintillation counter (Beckman Coulter LS 6500 scintillation system).

Polysome profiles for wild-type and variant metG E. coli were performed as previously described (Qin and Fredrick, 2013). Briefly, cultures were grown overnight in M9 minimal media supplemented with 50 μM methionine and back diluted to an OD at 600 nm of 0.05 in the same media. Cultures were grown at 37°C to an OD at 600 nm of 0.7, cooled rapidly by mixing with crushed ice and collected by centrifugation. Cellular pellets were resuspended in 300 mL lysis buffer (10 mM Tris–HCl, pH 8.0, 1 mM MgCl2, 1 mg/mL Lysozyme) and flash frozen in liquid nitrogen. Samples were thawed, mixed with 10% sodium deoxycholate and flash frozen again. After thawing, samples were centrifuged at 10000 rpm, 4°C for 10 min and the supernatant was loaded on top of a 10–40% sucrose gradient.

The gradients were centrifuged at 35000 rpm, 4°C for 3 h in a SW41 swinging buckets rotor. After centrifugation, the gradient tubes were mounted onto a Brandel/ISCO syringe-pump system equipped with a UV detector and the contents were injected through the system at a rate of 1.5 mL/min. Absorbance at 254 nm was monitored during the process, yielding the polysome profile. Ribosomes corresponding to the 30S, 50S, 70S and polysome fractions were calculated as a percentage of the total area.

The comparison of the translational initiation rate was based on the determination of the initial observation time of GFP. metG mutants and their parental strain were transformed with plasmid pS1, a low copy number encoding a gfp gene under an inducible arabinose promoter (Leiva et al., 2022). Briefly, saturated overnight cultures in LB media with ampicillin (100 g/mL) were washed and diluted to the same 0.1 OD₆₀₀ in fresh M9 media (7.7 mM Na₂HPO₄, 22.0 mM KH₂PO₄, 8.6 mM NaCl, 18.7 mM NH₄Cl, 2 mM MgSO₄, 0.1 mM CaCl₂, and 0.4% Glycerol) with ampicillin (M9 amp). In a 96-well optical-bottom plate, 100 μl of each strain were incubated for 30 min at 37°C, to then add 100 μl of M9 amp supplemented with arabinose (0.4% final concentration). Plates were further shaken at 37°C in a microplate reader (SPARK, TECAN) where OD₆₀₀ and GFP fluorescence intensity (Ex. 480 ± 20 nm, Em. 525 ± 20 nm) were determined. The square root of the fluorescence data and time (Schleif plot) were used to obtain the initial observation time of GFP (T_first GFP) in each strain (Schleif et al., 1973).

Escherichia coli tRNA^Met and tRNA^fMet encoding genes were individually cloned into a pUC18 plasmid. To generate the tRNA insert six overlapping synthetic DNA oligonucleotides that encode the T7 RNA promoter and the tRNA gene were heated to 90°C for 5 min and slowly cooled to room temperature. The annealed fragments were then ligated and cloned into pUC18 using the EcoRI and XbaI restriction digest sites. The first nucleotide in tRNA^fMet was mutated from cytosine to guanosine for T7 RNA polymerase to start transcription (G1-tRNA^fMet). Plasmid sequence was confirmed by DNA sequencing. A pUC18 plasmid with tRNA^fMet containing its natural first cytosine was ordered from GenScript Biotech Corp (Piscataway, NJ, USA). A hammerhead ribozyme sequence was placed between the T7 promoter and the tRNA sequence to generate a natural tRNA^fMet isoacceptor (C1-tRNA^fMet) (Fechter et al., 1998).

In vitro transcription of tRNA^Met and tRNA^fMet were performed as described (Steiner et al., 2019). Briefly, plasmids containing E. coli tRNA^Met and tRNA^fMet were used as template for PCR amplification of T7 promoter and tRNA. The PCR-amplified product was purified and used for T7 runoff transcription. In vitro transcription was performed using the purified PCR product and T7 RNA polymerase suspended in a buffer comprised of 40 mM Tris–HCl, Ph 8.0, 2 mM spermidine, 22 mM MgCl₂, 5 mM DTT, 50 μg/mL BSA, 4 mM ATP, 4 mM GTP, 4 mM CTP, 4 mM UTP, 20 mM 5′-GMP, pyrophosphatase and RNase inhibitor. Reactions were incubated overnight in a 42°C water bath. The tRNA products were purified using DEAE cellulose anion exchange resin and eluted with 1 M NaCl. The fractions that contained tRNA were monitored by absorbance at 260 nm and pooled together and ethanol precipitated overnight at −20C. The ethanol precipitated tRNA was centrifuged at 4,500 × g for 10 min at 4°C. The resulting pellet was dried and suspended in RNase free water. For in vitro transcription of tRNA^fMet with its natural cytosine at the first position and containing the hammerhead ribozyme between the T7 promoter and the tRNA sequence a few modifications were made to the purification of the tRNA (Fechter et al., 1998). After in vitro transcription was complete, 8 mM MgCl₂ was added to the reaction, which was then diluted fivefold with RNase free water and incubated at 60°C for 2 h. This allows for the hammerhead ribozyme to cleave before the first cytosine in tRNA^fMet, releasing its natural isoacceptor (C1-tRNA^fMet). This reaction was then ethanol precipitated. The resulting pellet was dried and suspended in RNase free water. The C1-tRNA^fMet was resolved on an 8% acrylamide, 9 M urea gel in Tris-boric acid-EDTA (TBE) buffer at 300 volts for 30 min. Under ultraviolet light the different tRNA fragments will illuminate. The middle fragment was excised from the gel (top fragment is the full transcript, the middle fragment is C1-tRNA^fMet, and the bottom fragment is the hammerhead ribozyme). The excised gel piece was cut up and added to a polyprep column containing 0.5 M ammonium acetate, 0.1% SDS, and 5 mM EDTA, pH 8.0, and incubated overnight at room temperature with agitation. The next day, the tRNA was eluted from the polyprep column and ethanol precipitated. The concentration of active tRNA was determined using aminoacylation kinetics as described below.

His₆-tagged wild-type and variant MetRS proteins were expressed from a pET28a plasmid in an E. coli BL21 (DE3) RIL strain. Protein expression and purification was performed essentially as described (Roy et al., 2005). The concentration of active MetRS was determined by active site titrations, as described (Ibba et al., 1994). To determine the steady-state amino acid activation kinetics for wild-type and variant MetRS ATP/PPi exchange assays were done as previously described (Roy et al., 2005). Briefly, reactions were performed at 37°C in 100 mM Na-HEPES, pH 7.2, 30 mM KCl, 10 mM MgCl₂, 2 mM NaF, 2 mM ATP, 2 mM ³²P-PPi (2 cpm/pmol) with various amounts of methionine (0–400 μM), and 25–100 nM enzyme. After 1–4 min, 25 μL of the reaction was quenched in a solution containing 1% activated charcoal, 5.6% HClO₄, and 75 mM PPi. The ³²P-ATP bound-charcoal was filtered through a 3 mm Whatman disc filter under vacuum and washed 3 times with 5 mL water. The filters were dried, and the radioactivity was counted by a liquid scintillation counter (Beckman Coulter LS 6500 scintillation system). Steady-state aminoacylation kinetics were performed as described (Francklyn et al., 2008). Briefly, reactions were performed in 100 mM Na-HEPES, pH 7.2, 20 mM KCl, 30 mM MgCl₂, 0.1 mg/mL BSA, 0.5 mM DTT, 40 mM ¹⁴C-L-methionine, and 50 nM enzyme. Various concentrations of tRNA^Met and tRNA^fMet (0 20 μM) and ATP (0–10 mM) were used depending on the substrate being tested for steady-state kinetics. Reactions were incubated at 30°C and at different minute intervals an aliquot of reaction was quenched on a filter pad soaked with 5% trichloroacetic acid (TCA) plus 5 μM methionine. The filters were washed 3 times in 5% TCA for 5 min each, dried, and the radioactivity was quantified by liquid scintillation counting. Kinetics measurements were calculated using GraphPad Prism.

S-NO-homocysteine was synthesized as previously described (Stamler et al., 1993; Jakubowski, 2000). Briefly, equimolar amounts of L-homocysteine and NaNO₂ were dissolved in 0.1 M HCl and incubated at 25°C. The reaction was neutralized with 0.5 M Na₂HPO₄, pH 7.4, and is complete when the reaction turns red. A small aliquot of the S-NO-homocysteine reaction and L-homocysteine were run on a silica TLC in n-butanol:acetic acid: ethyl acetate:water (1:1:1:1). After the TLC is complete and let dry, the amino acids are stained with a ninhydrin spray. S-NO-homocysteine will migrate faster than L-homocysteine. S-NO-homocysteine can be stored at −80°C for up to 2 weeks.

ATP consumption assays were performed to analyze pre-transfer editing of L-homocysteine and S-NO-homocysteine by wild-type and variant MetRS mutants as previously described (Steiner et al., 2019). Reactions contained 20 μM tRNA^Met or C1-tRNA^fMet (natural tRNA^fMet isoacceptor), 10 mM methionine, L-homocysteine, or S-NO-homocysteine, 3 mM ³²P-γ-ATP (∼5 CPM/pmole), 2 U/μL pyrophosphatase, and 1 μM wild-type or variant MetRS in a buffer containing 100 mM Na-Hepes, pH 7.2, 30 mM KCl, and 10 mM MgCl₂. Reactions were performed at 37°C and aliquots were quenched in an equal volume of glacial acetic acid at 3, 6, 10, 15, 20, and 30 min time points. A PEI cellulose TLC plate was pre-run in water, let dry, spot 1 μL of the quenched timepoint, and run in 0.7 M K₂HPO₄, pH 9 until the buffer reaches ∼1 cm from the top of the TLC plate. Let the TLC dry and expose to a phosphor screen and imaged on a phosphor imager (Typhoon). Quantification of the percentage of inorganic phosphate was completed using ImageQuant and GraphPad Prism.

For all growth experiments, overnight cultures of the wild-type and variant metG strains were grown to saturation in M9 minimal medium supplemented with 50 μM methionine. Cultures were collected by centrifugation, washed with fresh M9 media and backdiluted to an OD at 600 nm of 0.05 in M9 media supplemented with either Homocysteine. Cultures were then incubated at 37°C for 18 h, at which point the OD at 600 nm was measured. Four biological replicates were measured for each condition.

To facilitate construction of the metG mutants characterized in this study, we developed a gene editing system that allowed for specific base pair changes to be made to the E. coli chromosome. While several methods have been reported for genome sequence modifications, including λ Red-mediated homologous recombination (Datsenko and Wanner, 2000; Yu et al., 2000) and CRISPR-Cas9 technologies (Backes and Phillips, 2021) none is ideally suited for our specific needs. We, therefore, sought to develop a genetic system with no limitations as to the sequence changes made, constraints on surrounding sequences or host genotype that could be applied to genes essential for viability at high efficiency with a minimum number of steps.

Our approach was to build on the well-established R6K-based suicide vector system (Phillips, 2004), in combination with a proven counterselection strategy afforded by expression of the I-SceI meganuclease to generate lethal chromosomal breaks (Posfai et al., 1999; Datsenko and Wanner, 2000; Wong and Mekalanos, 2000; Lee et al., 2009). One limitation of R6K-based system is the requirement for relatively long regions of homology that cannot easily be incorporated into PCR primers. However, by taking advantage of the current availability of synthetic DNA constructs that can be custom made at low cost, along with methods that facilitate construction of recombinant plasmids (Casini et al., 2014), Since suicide vector systems offer several advantages for genome editing, we developed the R6K integrated-plasmid recombination (RIPR) system (summarized in Figure 1) for scarless editing of both essential and non-essential genes and to specifically investigate molecular mechanisms of MetRS activity in the phenomena of antibiotic persistence. The efficiency of the RIPR system for gene editing was validated by conducting several genome edits of the lac operon of E. coli (see Supplementary Information Results).

We used the RIPR system to construct E. coli mutants expressing different variants of metG to assess their contributions to bacterial antibiotic persistence. Given that bacterial cells that have entered an antibiotic persistent state represent only a small subpopulation of a culture, genetic approaches to study the phenomenon have included isolation of high persister mutants. Interestingly, metG mutants have been isolated in multiple studies seeking high persisters, including in E. coli (Girgis et al., 2012; Fridman et al., 2014; Levin-Reisman et al., 2017), Salmonella enterica (Slattery et al., 2013), Shigella flexneri (Zhu et al., 2018), Burkholderia thailandensis (Yi et al., 2018), and Vibrio cholerae (Silva-Valenzuela et al., 2017). While the function of MetRS in protein synthesis has been well characterized (Dardel et al., 1984; Ibba and Soll, 2000), its function in persistence is not well understood (Girgis et al., 2012; Harms et al., 2016).

For our investigation, we used RIPR to engineer four metG alleles identified from multiple sources into wild type E. coli K-12 (Supplementary Table 1). The metG mutations include metGΔETIT, which originated from an E. coli mutant isolated in a screen for antibiotic tolerance that results in an internal deletion of amino acids 569–572 (Levin-Reisman et al., 2017) and metG630, a frameshift mutation near the 3’ end of the gene that is comparable to a metG variant isolated as a high persister mutant in S. enterica (Slattery et al., 2013). In addition, alleles metG83 (Blumenthal, 1972) and metG87 (Greene et al., 1973) were characterized. DNA sequence analyses revealed metG83 corresponds to amino acid changes G263S and S264F and metG87 with S264F. These mutants were originally isolated as methionine auxotrophs and have not previously been characterized for antibiotic persistence. To construct these mutants, we included the 3’ end of the metG gene as part of PCR products (metG83 and metG87) and synthetic DNA constructs (metGΔETIT and metG630), as shown graphically in Figure 1. Mutants were generated, as described, and their genotypes confirmed by PCR and DNA sequencing.

To survey the level of persistence for each metG mutant, we conducted initial assays as described in see section “2 Materials and methods.” The metG630 and metGΔETIT mutants produced persisters at 2–3 orders of magnitude higher frequency than wild type E. coli (Figure 2A). The metG83 and metG87 mutants also showed an order of magnitude increase in persistence, indicating that mutations throughout the metG coding region can influence the levels of antibiotic persistence. To determine if the metG alleles conferred elevated persistence to other antibiotics, we performed plate assays, as described (Gefen et al., 2017). As shown in Figure 2B, both the metG630 and metGΔETIT mutants showed increased levels of persistence to ciprofloxacin, oxytetracycline, and cephalosporin compared to the wild type control.

Examination of growth kinetics of the metG mutants revealed that only the metG87 mutant exhibited an increase in lag time on rich medium before entering exponential phase (Supplementary Figure 2). metG87 encodes a S264F replacement near the catalytic site, metG83 a G263S/S264F double replacement. We concluded that the S264F replacement in metG resulted in a slow growing phenotype, and the second G263S replacement found in metG83 acts to suppress this growth defect. Increased lag times have been reported in persistent and tolerant phenotypes and can be the mechanism behind the persistent phenotype (Fridman et al., 2014; Vulin et al., 2018; Simsek and Kim, 2019). Since we are interested in exploring the effects of MetRS and translation on bacterial antibiotic persistence we excluded metG87 and focused the rest of our studies on the three metG mutants that exhibited similar growth kinetics to WT.

Further characterization of the metG alleles effects on antibiotic persistence was performed using minimal duration of killing (MDK) assays (Figure 3). The MDK assays showed the characteristic biphasic killing curves that is attributed to bacterial antibiotic persistence (Figure 3; Brauner et al., 2016; Balaban et al., 2019). The rate of killing and the higher number of persisters for the metG mutants compared to wild type E. coli followed the same pattern as the initial characterization of the metG alleles (Figure 2).

We next wanted to investigate if the overall reduction in intracellular concentration of MetRS influenced bacterial antibiotic persistence. We used CRISPRi to reduce the transcript levels of metG, which lead to a reduction of intracellular MetRS concentration. Transcription of alaS, which encodes for the elongator alanyl-tRNA synthetase, was also knocked down by CRISPRi to ascertain the impact of persistence of an unrelated aaRS (Supplementary Table 2). Growth curves kinetics and colony forming unit experiments were conducted to determine the optimal concentration of anhydrous tetracycline (aTc) that would result in the same knock down efficiency of the alaS and metG transcripts by the specific small guide RNA (Supplementary Figures 3, 4).

Bacterial antibiotic persister endpoint assays were performed to study the effect that metG and alaS transcription knock down had on persistence (Figure 4). Bacterial antibiotic persistence increased significantly when transcription of metG was knocked down compared to wild type E. coli and the empty vector (Figure 4). However, when alaS was knocked down there was not a significant increase in the level of antibiotic persistence.

We next used pulse-chase labeling to measure rates of protein synthesis of the metG mutants in minimal media (Figure 5). In general, slower translation rates, while not sufficient to impact growth rates on complete medium of the metG mutants, nevertheless correlated to increases in bacterial antibiotic persistence. This suggests that there may be a connection between translation rates within a cell and its ability to enter a persistent phenotype. To further characterize the translation process, we conducted polysome profiling on each of the mutants (Supplementary Figure 5). The percentage of ribosomes within the 70S fraction was not significantly reduced in three of the mutants (metG83, metG630, metGΔETIT). These results demonstrate that the metG mutants do not have an impact on ribosome assembly but do affect translation rates, consistent with increased disassembled subunits and decreased polysomes. This indicates that the reduced translation rates may be caused by either a defect in translation initiation or the intracellular concentration of methionyl-tRNA^Met if the mutation within MetRS affects its ability to aminoacylate initiator or elongator tRNA^Met, respectively.

To investigate if the metG mutants primarily affect either the initiation stage of translation or elongation, we used an inducible fluorescent reporter system to evaluate protein synthesis. Under the premise that none of the metG mutations affect the rate of transcription or folding of GFP, we assumed that transcription and maturation times are constant for all mutants. By comparing the initial GFP observation time, which corresponds to the minimum time needed to reach the critical mass of fluorescent reporter to be detected, we see an ∼1.5-fold increase in the time required for GFP synthesis for all metG mutants (Figure 6A). This supports the results presented previously in Figure 5 in which an impairment in the translation rate was shown. Additionally, a non-fluorescent domain was fused to the N terminus of GFP, nearly doubling the size of the protein and the number of methionines involved in elongation (244 to 477 amino acids and 4 to 8 methionine, respectively) (Figure 6B). The initial observation time for the long GFP variant (GFP_Long) is the same as that of normal-sized GFP for wild type E. coli, and the metG83 and metGΔETIT mutants, demonstrating that the changes generated due to the additional domain that must be elongated do not affect the total translation time. This suggests that while mutations in metG83 and metGΔETIT mainly affect the initiation step, metG630 affects both the initiation and elongation steps as evidenced by an increase in the initial observation time for GFP_Long.

We next investigated the steady-state kinetics of the MetRS variants to further understand how the different mutations in metG affect aminoacylation. Recombinant protein of wild type MetRS and the three variant MetRSs (metG83, metG630, metGΔETIT) and in vitro transcribed elongator tRNA^Met and initator tRNA^fMet was used for steady-state kinetic analyses (Table 1). In general, there was not any significant difference in the aminoacylation kinetics or specific activity of the variant MetRSs when compared to wild type MetRS. These results indicate that the different hyperpersistent mutations found within the metG gene do not significantly affect the kinetics for methionine, ATP, tRNA^Met, or tRNA^fMet.

To further investigate if the MetRS variants have any effect on enzyme chemistry we looked at the pre-transfer editing mechanism. Homocysteine (Hcy) is a non-cognate amino acid for aminoacylation of tRNA^Met by MetRS (Jakubowski, 2003, 2017). If Hcy enters the active site of MetRS it will be pre-transfer edited forming homocysteine thiolactone and releasing tRNA^Met, AMP, and inorganic pyrophosphate (Jakubowski, 2003, 2017). To study the kinetics of the pre-transfer editing mechanism we used recombinant wild type MetRS and three variant MetRSs (metG83, metG630, metGΔETIT) and in vitro transcribed elongator tRNA^Met and initiator tRNA^fMet with three different amino acids, the cognate methionine, and non-cognate homocysteine and S-NO-homocysteine, a natural variant of homocysteine (Jakubowski, 2000). The first step of MetRS’s editing mechanism is to adenylate the amino acid (Ibba and Soll, 2000). Adenylated-homocysteine has a short half-life because the thiol group of the homocysteine is positioned in a way that allows for easy attack of the carboxyl group which releases the more stable metabolite homocysteine thiolactone and AMP (Jakubowski, 2017). However, if homocysteine is modified with a nitrosyl-group it stabilizes the thiol group making adenylated-S-NO-homocysteine more stable than its unmodified amino acid form (Supplementary Figure 6; Jakubowski, 2000, 2003). Therefore, we synthesized S-NO-homocysteine so that we could investigate if MetRS and its mutant variants are able to perform pre-transfer editing of adenylated-homocysteine rather than random hydrolysis of homocysteine to homocysteine thiolactone (Jakubowski, 2000). To study aaRS editing rates we employed ATP futile cycling assays which measure the depletion of ATP to AMP by calculating the amount of P_i produced (Steiner et al., 2019). As more editing of non-cognate amino acid occurs more ATP is utilized, which in turn produces more inorganic phosphate.

We first looked at editing with all three amino acids tested with and without elongator tRNA^Met (Figure 7). As expected, with wild type MetRS we observed no editing with methionine, ∼80–90% editing with homocysteine, and ∼25% editing with S-NO-homocysteine. There was no significant difference in the percentages of editing with and without tRNA^Met, which indicates that wild type MetRS can perform pre-transfer editing with both homocysteine and S-NO-homocysteine. There were no indications of pre-transfer editing with the MetRS 83 variant. Interestingly, MetRS 630 was only able to edit homocysteine once tRNA^Met was added to the reaction. Also, MetRS ΔETIT can edit homocysteine without tRNA^Met present, but its editing activity increases with tRNA^Met. Next, we looked at editing with these three amino acids with and without initiator tRNA^fMet (Figure 8). Intriguingly, there was no editing observed for any of the MetRS variants with initiator tRNA^fMet. Wild type MetRS showed ∼25% editing with S-NO-homocysteine but no editing was observed once tRNA^fMet was added to the reaction. These results indicate that the MetRS variants used in this study have different capabilities of pre-transfer editing.

Hcy is a key intermediate in central metabolism and necessary for several different pathways such as methionine biosynthesis (Jakubowski and Glowacki, 2011; Ferla and Patrick, 2014). Changes in metabolism and increases in intracellular deacylated-tRNA levels have both been linked to underlying mechanisms of bacterial antibiotic persistence (Amato et al., 2014; Radzikowski et al., 2017; Wood et al., 2021). Since Hcy is important in both metabolism and tRNA^Met aminoacylation we wanted to investigate if Hcy had any effects on the growth of the three different metG mutant strains. Overall, all the metG mutants (metG83, metG630, metGΔETIT) were hypersensitive to increasing concentrations of Hcy compared to wild type E. coli (Figure 9), suggesting a link between the metG mutants, Hcy, and antibiotic persistence.

Bacterial translation quality control is important to maintain the homeostasis of the proteome (Bullwinkle and Ibba, 2016; Mohler and Ibba, 2017; Kelly et al., 2019). When translation quality control is absent tRNAs may be misacylated, which can result in protein synthesis errors. This was once thought to be detrimental to the cell, however, it is now understood that altered quality control can sometimes be beneficial as bacterial cells adapt to new environments (Ribas de Pouplana et al., 2014; Rathnayake et al., 2017; Kelly et al., 2019). For example, previous studies using quality control-modified PheRS variants showed that deacylated-tRNA is a trigger for changes in bacterial antibiotic persistence independent of the stringent response (Wood et al., 2021). A genetic screen performed in E. coli identified several hyper-persistent mutations in metG which encodes for MetRS (Fridman et al., 2014). Specifically, the mutants found were metG87, which has a point mutation near the catalytic site (S264F) and metG83, a double mutation also near the catalytic site (G263S/S264F). Two additional hyper-persistent mutations that were originally isolated in S. Typhimurium were also investigated in this study, metG630 which is a −4 frameshift mutation resulting in the change of 4 amino acids at the end of a truncated protein near the anticodon binding domain, and metGΔETIT, a deletion mutation removing amino acids 569–572 near the anticodon binding domain (Slattery et al., 2013). Variants in MetRS are of special interest to study how defects in translation can affect bacterial antibiotic persistence because MetRS is responsible for aminoacylation both elongator tRNA^Met and initiator tRNA^fMet (Ibba and Soll, 2000; Laursen et al., 2005). MetRS is also able to perform pre-transfer editing to form homocysteine thiolactone from homocysteine, an important intermediate in central metabolism (Jakubowski, 2003, 2017; Jakubowski and Glowacki, 2011; Ferla and Patrick, 2014). Perturbations in metabolism and consequently in intracellular homocysteine levels could affect the editing activities of MetRS leading to changes in the cellular concentrations of deacylated tRNA^Met. The numerous roles of MetRS in protein synthesis, and indirectly in metabolism, make metG mutants of particular interest to investigate bacterial antibiotic persistence.

In our initial characterization of the metG mutants, metG630 and metGΔETIT showed 2–3 orders of magnitude more persisters than wild type E. coli, and increased levels of persistence to ciprofloxacin, oxytetracycline, and cephalosporin (Figure 2). The MDK assays of the metG mutants showed the characteristic biphasic killing curves and the rate of killing compared to wild type E. coli followed the same pattern as the initial characterization of the metG alleles (Figure 3). This indicates that mutations throughout metG can affect levels of bacterial antibiotic persistence. To see if the increased levels of persistence are linked to defects in translation, we characterized the overall translation rates of the metG mutants, along with their polysome profiles. It was observed that slower rates of translation of the metG mutants generally correlated with higher levels of antibiotic persistence (Figure 5). However, the 70S ribosome profile fraction was not significantly reduced for the metG mutants compared to wild type E. coli, indicating that there is not an impact on ribosome assembly but that these mutants do affect overall translation rates. It is worth noting that there are differences amongst the metG mutants for the 30S, 50S, and polysome profile fractions. For example, metG630 has the highest rate of antibiotic persister formation (Figures 2, 3.), the second slowest rate of translation (Figure 5), the highest percentage of 30S and 50S ribosome fractions but the lowest percentage of polysomes (Supplementary Figure 5). This may imply that while there are no differences in the 70S ribosomes the differences in the other fractions of the polysome profile may have a downstream affect that contributes to overall translation defects and bacterial antibiotic persistence increases. This further suggests that the reduced translation rates are caused by a defect in either translation initiation or the intracellular concentration of methionyl-tRNA^Met. However, there were not any significant differences in the aminoacylation kinetics or specific activity of the variant MetRSs when compared to wild type MetRS (Table 1). These results indicate that the different hyperpersistent mutations found within metG do not affect the steady-state kinetics of MetRS or its substrates.

Since MetRS is involved in both translation initiation and elongation we further investigated how metG mutants may affect these two stages of translation. It was observed that all the metG mutants had an increase in time of ∼1.5-fold for translation initiation compared to wild type E. coli (Figure 6A). Only the metG630 mutant also showed an increase in the time for translation elongation when compared to wild type (Figure 6B), consistent with it being the mutant with the largest changes in polysome profile. Overall, the mutants that display the most significant decreases in protein synthesis rates and active translation correlated with higher persistence.

Mutations within the metG coding region have a strong correlation between translation defects and bacterial antibiotic persistence, but a poor correlation between MetRS aminoacylation kinetics and persistence. We therefore considered if changes in the intracellular concentration of MetRS affected antibiotic persistence levels or if these effects were due to the changes in the function or activity of MetRS. This was achieved by knocking down the transcription levels of metG via CRISPRi technologies. When metG was knocked down this resulted in a significant increase in the level of bacterial antibiotic persistence when compared to wild type E. coli and an alaS knock-down control (Figure 4). These results indicated that some but not all the 20 aminoacyl-tRNA synthetases are implicated in an underlining mechanism of bacterial antibiotic persistence. In a previous study where several PheRS variants were used it was observed that these variants did increase the level of bacterial antibiotic persistence as a result of changes in cellular deacylated tRNA levels (Wood et al., 2021). Reducing the transcription levels of alaS produced increased levels of antibiotic persistence broadly comparable to those for PhRS variants; however, the effects on persistence from the metG mutations are much stronger, likely due to additive effects on translation initiation.

MetRS can perform pre-transfer editing of homocysteine, an important non-proteogenic amino acid that is involved in central metabolism (Jakubowski, 2003, 2017; Jakubowski and Glowacki, 2011; Ferla and Patrick, 2014). The canonical reaction of MetRS is to esterify both tRNA^Met and initiator tRNA^fMet with methionine in an ATP dependent manner (Ibba and Soll, 2000; Laursen et al., 2005). However, if non-cognate homocysteine becomes adenylated, MetRS will perform pre-transfer editing releasing homocysteine thiolactone and deacylated tRNA^Met/fMet (Supplementary Figure 6; Jakubowski, 2003, 2017). A natural modification of homocysteine is S-NO-homocysteine which usually occurs under nitric oxide stress (Braun et al., 2007). S-NO-homocysteine has been detected in human endothelial cells and it has been shown to be able to be translationally incorporated into proteins in E. coli, because the S-nitrosyl group stabilizes the thiol group of homocysteine and prevents pre-transfer editing by MetRS (Stamler et al., 1993; Jakubowski, 2000). We studied the kinetics of the pre-transfer editing mechanism of wild type and the three variant MetRSs used in this study along with both elongator tRNA^Met and initator tRNA^fMet. Both tRNAs were in vitro transcribed to ensure that they contained their natural first base; tRNA^Met starts with a guanosine base and tRNA^fMet starts with a cytosine base (Laursen et al., 2005). In E. coli, tRNA^fMet contains a 1–72 mismatched base pair with is important for the formylation of Met-tRNA^fMet by methionyl-tRNA formyltransferase (Mayer and RajBhandary, 2002; Giege and Eriani, 2023).

As expected, with wild type MetRS and either tRNA^Met or tRNA^fMet there was no editing with methionine observed, ∼80–90% editing with homocysteine, and ∼25% editing with S-NO-homocysteine (Figures 7, 8). There was no editing observed with the MetRS 83 variant with any of the amino acids tested with either tRNA^Met or tRNA^fMet. Intriguingly, the product of the metRS630 variant was only able to edit homocysteine in the presence of tRNA^Met specifically. MetRSΔETIT was able to modestly edit homocysteine alone, but editing was increased in the presence of tRNA^Met specifically. These MetRS variants have changes to the amino acid sequence near the anti-codon binding domain. This observation may indicate that tRNA^Met binding is required for a tRNA-specific conformational change to induce editing of homocysteine and S-NO-homocysteine. The MetRS630 and MetRSΔETIT variants were unable to edit these non-cognate amino acids in the presence of tRNA^fMet. These results indicate that the MetRS variants used in this study have different capabilities of pre-transfer editing and that the high rate of persistence observed with the metG mutant E. coli strains are supported by editing deficiencies. The differences observed with initiator tRNA and tRNA^Met in pre-transfer editing also indicate that translation initiation and elongation may be contributing in various ways toward the higher rates of bacterial antibiotic persistence.

We observed that the different metG mutants have increased levels of bacterial antibiotic persistence. These increases may be due to changes in translation initiation and/or translation elongation and the differences that were observed in homocysteine pre-transfer editing activities of the MetRS variants. Since homocysteine is important in both central metabolism and tRNA^Met aminoacylation we investigated if this non-proteogenic amino acid had any effects on the growth of the different metG mutant strains used in this study. We observed that all the metG mutants were hypersensitive to increasing concentrations of homocysteine compared to wild type E. coli (Figure 9). The sensitivity to homocysteine supplementation to growth of the metG mutant strains and the defects in editing of the MetRS variants suggest that there is a strong correlation between the metG mutants, homocysteine, and antibiotic persistence. Isoleucyl-tRNA synthetase and leucyl-tRNA synthetase are also able to edit homocysteine to homocysteine thiolactone and have both previously been linked to increases in bacterial antibiotic persistence (Garoff et al., 2018; Khare and Tavazoie, 2020). It is interesting to speculate that the levels of homocysteine and homocysteine thiolactone could cause shifts in central metabolism and thus could be an underlining factor contributing to antibiotic persistence.

In conclusion, several different pathways are implicated in the elevated levels of bacterial antibiotic persistence in the metG mutant E. coli strains used in this study. We observed defects in translation initiation and elongation rates, different defects in the editing capabilities of the MetRS variants, and hyper-sensitivity to the growth of the metG mutant E. coli strains when supplemented with homocysteine. Intriguingly, all these changes in the E. coli cell could have a compounding effect resulting in the high levels of bacterial antibiotic persistence that were observed. Several aaRSs have been found to impact the levels of bacterial antibiotic persistence (Germain et al., 2013; Kaspy et al., 2013; Levin-Reisman et al., 2017; Garoff et al., 2018; Hatch and Ouellette, 2020; Huang et al., 2020; Khare and Tavazoie, 2020; Wood et al., 2021). However, any contributions from translation initiation to the increased levels of antibiotic persistence would be specific to MetRS. Taken together, these data suggest that while there is a general translation effect, as with many aaRSs, the MetRS variants have a stronger effect on antibiotic persistence that may be due to their editing defects and the resulting changes in cellular homocysteine and homocysteine thiolactone levels.