A tRNA modifying enzyme as a tunable regulatory nexus for bacterial stress responses and virulence

Abstract Post-transcriptional modifications can impact the stability and functionality of many different classes of RNA molecules and are an especially important aspect of tRNA regulation. It is hypothesized that cells can orchestrate rapid responses to changing environmental conditions by adjusting the specific types and levels of tRNA modifications. We uncovered strong evidence in support of this tRNA global regulation hypothesis by examining effects of the well-conserved tRNA modifying enzyme MiaA in extraintestinal pathogenic Escherichia coli (ExPEC), a major cause of urinary tract and bloodstream infections. MiaA mediates the prenylation of adenosine-37 within tRNAs that decode UNN codons, and we found it to be crucial to the fitness and virulence of ExPEC. MiaA levels shifted in response to stress via a post-transcriptional mechanism, resulting in marked changes in the amounts of fully modified MiaA substrates. Both ablation and forced overproduction of MiaA stimulated translational frameshifting and profoundly altered the ExPEC proteome, with variable effects attributable to UNN content, changes in the catalytic activity of MiaA, or availability of metabolic precursors. Cumulatively, these data indicate that balanced input from MiaA is critical for optimizing cellular responses, with MiaA acting much like a rheostat that can be used to realign global protein expression patterns.


INTRODUCTION
The translation of mRNA into protein by ribosomes and aminoacyl-transfer RNA (tRNA) complexes is an energyintensive process that is subject to multiple levels of complicated regulation. For example, tRNAs can be covalently modified by >100 different moieties that can influence the charging of tRNAs with amino acids, tRNA stability, codon usage, and reading frame maintenance (1)(2)(3)(4). In Escherichia coli and other organisms, the hypomodification of tRNAs can result in decreased growth rates, altered metabolic requirements, and reduced stress resistance (5)(6)(7)(8). Shifts in the prevalence of specific tRNA modifications are proposed to help optimize cellular responses to stress by affecting translational fidelity and selective protein expression (9)(10)(11). In other words, changing levels of tRNA modifications may control the codon-biased translation of select transcripts, providing a post-transcriptional programmable mechanism that distressed cells can use to facilitate beneficial changes in their proteomes. Although there is growing evidence from a variety of model systems in support of this idea (12,13), our understanding of how changing levels of a tRNA modifying enzyme actually impact the protein landscape is limited.
One of the most commonly modified tRNA residues in bacteria is adenosine-37 (A-37), which lies adjacent to the anticodon loop (8,14). In its final form in E. coli, A-37 of UNN-recognizing tRNA molecules is oftentimes prenylated and methylthiolated (15). The miaA gene of E. coli encodes a tRNA prenyltransferase that catalyzes the addition of a prenyl group onto the N 6 -nitrogen of A-37 to create i 6 A-37 tRNA (16,17) ( Figure 1A). The modified i 6 A-37 residue is subsequently methylthiolated by the radical-S-adenosylmethionine enzyme MiaB to create ms 2 i 6 A-37 (18). The bulky and hydrophobic ms 2 i 6 A-37 modification enhances tRNA interactions with UNN target codons, promoting reading frame maintenance and translational fidelity (5,8,19). Mutations in the miaA locus result in an unmodified A-37 residue, as prenylation is required for methylthiolation by MiaB. In laboratory-adapted K-12 E. coli strains, mutations in miaA impair attenuation of the tryptophan and phenylalanine operons (20,21) and diminish translation of the stationary phase sigma factor RpoS and the small RNA chaperone Hfq (7,22,23). Additionally, mutants lacking miaA are unable to effectively resolve aberrant DNA-protein crosslinks (24) and have somewhat elevated spontaneous mutation frequencies (10,25,26). The ms 2 i 6 A-37 modification is highly conserved in both prokaryotes and eukaryotes, though the specific enzymes that mediate this modification have diverged within evolutionarily distant organisms (8). However, in prokaryotes, MiaA and MiaB homologues are relatively well conserved, and the enzymes appear to function similarly in all tested bacterial species (27,28).
Given that the ms 2 i 6 A-37 modification is a well-defined regulator of many tRNA functions in lab-adapted K-12 E. coli strains, we sought to understand how this modification is co-opted in a pathogenic E. coli background. E. coli pathotypes display extensive genetic diversity and are usually much more resilient under stress than their lab-adapted counterparts (29). As such, pathogens can provide insight into stress resistance and adaptive mechanisms that may be difficult to discern in non-pathogenic lab strains. Extraintestinal Pathogenic E. coli (ExPEC) typically reside in the lower intestinal tract of mammals, where they are rarely associated with pathology (30). However, when these microbes spread outside the gut to other host sites they can cause a number of serious diseases, including urinary tract and bloodstream infections (29,31). Bacterial pathogens like ExPEC must be able to rapidly respond to a diverse array of stressors encountered within changing host environments. These include nutrient deprivation, oxygen and nitrogen radicals, extremes in pH, envelope damage, changing osmotic pressures, and a wide assortment of host immune effector cells and antimicrobial compounds (32)(33)(34)(35)(36).
Here, we show that ExPEC can tune MiaA expression in response to stress, and that varying levels of this enzyme can increase translational frameshifting and markedly alter the spectrum of expressed proteins. Our results place MiaA at the center of a regulatory network that can promote stark changes in the proteome via multiple processes, including the alteration of other RNA and translational modifiers and depletion of metabolic precursors.

Study approval
All animals used in this study were handled in accordance with protocols approved by the Institutional Animal Care and Use Committee at the University of Utah (Protocol number 19-01001), following US federal guidelines indicated by the Office of Laboratory Animal Welfare (OLAW) and described in the Guide for the Care and Use of Laboratory Animals, 8th Edition. Mice were purchased from The Jackson Laboratory, housed three to five per cage, and allowed to eat (irradiated Teklad Global Soy Protein-Free Extruded chow) and drink antibiotic-free water ad libitum.

Biological resources
Seven-to eight-week-old female C57Bl/6, CBA/J, or C3H/HeJ mice (The Jackson Laboratory) were used according to the experimental descriptions below. Cell culture experiments were performed using the human bladder epithelial cell line 5637 (HTB-9; ATCC) and are described further in the sections below (37). Bacterial strains used in this study are listed in Supplementary Table S1. Mutant strains were constructed in the reference ExPEC isolate UTI89 using the lambda Red recombination system and primers detailed in Supplementary Table S2, as previously described (38). The chloramphenicol resistance (Clm R ) cassette flanked by LoxP sites was amplified from plasmid pKD3 using primers that contain overhanging ends with ∼40 bp of homology near the 5 and 3 ends of each target locus. PCR products were introduced by electroporation into UTI89 carrying pKM208, which encodes an IPTGinducible lambda Red recombinase (39). Knockouts were verified by PCR using primers indicated in Supplementary  Table S2.
Expression and reporter constructs were generated using standard molecular biology approaches and primers listed in Supplementary Table S2. The miaA and miaB genes were amplified from UTI89 by PCR, digested, and ligated into pRR48 using Pst1 and Kpn1 restriction sites to create pMiaA Ptac and pMiaB Ptac . Sequences encoding Hfq fused with C-terminal Flag and 6xHis tags were cloned using a similar approach to make pHfq Ptac . Point mutations in pMiaA-Flag Ptac were introduced using the QuikChange  (28). HDMAPP, mono[(2E)-hydroxy-3-methyl-2-butenyl] ester diphosphoric acid; DMAPP, dimethylallyl diphosphate; SAM, S-adenosylmethionine; SAH, S-adenosylhomocysteine; Cys, cysteine. (B and C) To assess gut colonization, adult BALB/c mice were inoculated via oral gavage with ∼10 9 CFU of a 1:1 mixture of (B) UTI89 and UTI89 miaA or (C) UTI89 and UTI89 miaB. Fecal titers were determined at the indicated time points and used to calculate competitive indices (CI). ***P < 0.001; ****P < 0.0001 by one sample t-tests. n = 7-8 mice from two independent experiments. (D) The bladders of adult female CBA/J mice were inoculated via transurethral catheterization with ∼10 7 CFU of UTI89, UTI89 miaA, or UTI89 miaB. Mice were sacrificed 3 days later and bacterial titers within the bladders were determined by plating tissue homogenates. **P < 0.01 by Mann-Whitney U tests; n ≥ 19 mice per group from at least three independent experiments. In (B), (C) and (D), bars indicate median values; dots represent individual mice. (E) Kaplan-Meier survival curves of C57Bl/6 mice inoculated via i.p. injections with ∼10 7 CFU of UTI89 (black line), UTI89 miaA (blue) or UTI89 miaB (orange). ****P < 0.0001 by log-rank Mantel Cox test for UTI89 versus UTI89 miaA; n = 13 mice per group from two independent experiments. II site-directed mutagenesis kit (Agilent) with the primers noted in Supplementary Table S2. To create pMiaA nat , the UTI89 miaA coding sequence was amplified along with 200 bp of flanking sequences, including the miaA promoter, and then ligated into the EcoR1 site of pA-CYC184. The plasmid pMiaA-Flag nat , having the miaA promoter region upstream of sequences encoding MiaA with a C-terminal Flag-tag, was produced similarly. To create pMiaA-Flag Ptac , sequences for Flag-tagged MiaA were sub-cloned from pMiaA nat into the PstI and KpnI sites of pRR48. Of note, the Flag tag did not interfere with MiaA function in complementation assays.
The dual-luciferase reporter plasmids used for frameshifting assays were created using p2Luc plasmids as templates (40,41). The genes encoding the renilla and firefly luciferases were amplified by PCR along with intergenic Az1-or HIV-derived linker sequences. A Shine-Dalgarno ribosome binding site was incorporated into the forward primer (p2Luc F) to promote translation of the linked luciferases. PCR products were digested and ligated into the KpnI and HindIII sites of pBAD18 (Ap R ) and pBAD33 (Cam R ). Plasmids with different resistance cassettes were needed for use with UTI89 miaA (Cam R ) and UTI89 carrying pMiaA Ptac or the empty vector pRR48. Control plasmids in which the Az1 and HIV linkers are altered to place the two luciferases in-frame were generated in an analogous fashion using previously described p2Luc plasmids as templates (40,41).

Bacterial growth analysis
UTI89 and its derivatives were grown from frozen stocks in 5 ml of LB (lysogeny broth), 100 mM MES-buffered LB (MES-LB; pH 5.0), or modified M9 medium (6 g/liter Na 2 HPO 4 , 3 g/l KH 2 PO 4 , 1 g/l NH 4 Cl, 0.5 g/l NaCl, 1 mM MgSO 4 , 0.1 mM CaCl 2 , 0.1% glucose, 0.0025% nicotinic acid, 0.2% casein amino acids and 16.5 g/ml thiamine in H 2 O) at 37 • C overnight in loosely capped 20-by-150-mm borosilicate glass tubes with shaking (225 rpm, with tubes tilted at a 30 • angle). Overnight cultures were brought to an OD 600 of ∼1.0 and then sub-cultured 1:100 into LB, MES-LB or M9 medium. Growth curves were acquired using a Bioscreen C instrument (Growth Curves USA) with 200l cultures in 100-well honeycomb plates shaking at 37 • C. Cultures included extra NaCl (5% w/v), 1 mM MV (Sigma-Aldrich), 1 or 2 mM ASN (Sigma-Aldrich), 25 or 50 M HDMAPP or DMAPP (Cayman Chemical), or IPTG, as indicated. Solutions containing MV, ASN, HDMAPP, or DMAPP were prepared fresh just before use. All growth curves were determined using quadruplicate samples with at least three independent replicates. Overnight cultures of strains carrying plasmids for complementation experiments were grown in the presence of antibiotics (100 g of ampicillin/ml or 50 g of tetracycline/ml) to maintain the plasmids, but antibiotics were not included in media used for the subsequent growth assays. Error bars are small for technical replicates in these experiments and were omitted for clarity. As a means to show levels of variance between each replicate, graphs with area under the curve (AUC) values for all growth curves are presented in Supplementary Figure S1.

Invasion, adhesion, and intracellular persistence assays
Bacteria were grown at 37 • C for 48 h in 20 ml static LB to induce expression of type 1 pili, which are important mediators of UPEC adherence and entry into host cells (42). Host cell association and gentamicin protection-based invasion and overnight intracellular persistence assays were performed as previously described using the human bladder epithelial cell line 5637 (HTB-9; ATCC) (37). Of note, UTI89 miaA is about 3-fold more sensitive to the host-cell impermeable antibiotic gentamicin, as determined by using Etest Strips (VWR) (Supplementary Figure S2). However, this likely had no effect on results from the cell culturebased invasion and intracellular survival experiments, as the concentrations of gentamicin (100 and 10 g/ml) used in these assays exceed those needed to effectively kill extracellular UTI89, UTI89 miaA, and UTI89 miaB.

Competitive gut colonization assays
For these assays, a kanamycin resistance cassette (Kan R ) was inserted into the attTn7 site of UTI89 to create UTI89::Kan R , which can be easily distinguished from the chloramphenicol resistant (Cam R ) miaA and miaB knockout mutants by plating on selective media. Previous work demonstrated that insertion of the Kan R resistance cassette into the attTn7 site does not impact ExPEC fitness within the gut (43,44). Individual cultures of UTI89::Kan R (standing in as the wild-type strain), UTI89 miaA, and UTI89 miaB were grown statically from frozen stocks for 24 h at 37 • C in 250-ml flasks containing 20 ml of modified M9 medium. Each knockout mutant was then mixed 1:1 with UTI89::Kan R (6 ml of each culture) and then pel-leted by centrifugation at 8000 × g for 8 min at room temperature. The bacterial pellets were then washed once with phosphate-buffered saline (PBS), pelleted again, and resuspended in 0.5 ml of PBS. Female SPF BALB/c mice aged 7 to 8 weeks were inoculated via oral gavage with 50 l PBS containing ∼10 9 CFU of each bacterial mixture. At the indicated time points post-inoculation, individual mice were placed into unused takeout boxes for a few minutes for weighing and feces collection. Freshly deposited feces were collected from the boxes and immediately added to 1 ml of 0.7% NaCl, weighed, and set on ice. The samples were then homogenized and briefly centrifuged at low speed to pellet any insoluble debris. Supernatants were serially diluted and plated onto LB agar containing either chloramphenicol (20 g/ml) or kanamycin (50 g/ml) for selective growth of UTI89::Kan R (wild type), UTI89 miaA, or UTI89 miaB. Fecal samples were also analyzed prior to the start of each experiment to ensure that there were no endogenous bacteria present that were resistant to chloramphenicol or kanamycin. Competitive indices (CI) were calculated as the ratio of knockout over wild-type bacteria recovered in the feces divided by the ratio of knockout over wild-type bacteria present in the inoculum (44,45). A total of seven to eight mice in two independent assays were used for each set of bacterial strains tested.

UTI model
The murine UTI model was used essentially as described by our group and others (42,46). Wild-type UTI89 and the miaA, and miaB knockout mutants were grown from frozen stocks in 20 ml LB in 250 ml Erlenmeyer flasks without shaking at 37 • C for 24 h. Bacteria were then pelleted by centrifugation (8 min at 8000 × g) and resuspended in PBS. Seven-to eight-week-old female CBA/J or C3H/HeJ mice were briefly anesthetized by isoflurane inhalation and slowly inoculated via transurethral catheterization with 50 l of PBS containing a suspension of ∼10 7 bacteria. Bacterial reflux into the kidneys using this procedure is rare, occurring in less than 1% of the test animals. At 0.25, 1, 3 or 9 days post-inoculation, mice were sacrificed and bladders were harvested aseptically, weighed and homogenized in 1 ml PBS containing 0.025% Triton X-100. Bacterial titers within the homogenates were determined by plating serial dilutions on LB agar plates. Nine or more mice in total, from two independent experiments, were used for each bacterial strain and time point examined.

Sepsis model
UTI89, UTI89 miaA, and UTI89 miaB were grown from frozen stocks in 20 ml M9 broth without shaking at 37 • C for 24 h, pelleted by centrifugation at 8000 × g for 8 min, and washed once with PBS, pelleted again, and resuspended in PBS. Seven-to eight-week-old female C57Bl/6 mice were briefly anesthetized by isoflurane inhalation and infected via intraperitoneal injection of ∼10 7 CFU within 200 ml PBS. Mice were monitored over a 72-h period for signs of morbidity and mortality. Alternatively, at 6 h postinoculation mice were sacrificed and the liver, kidneys, and spleens were harvested aseptically, weighed, and homogenized in 1 ml PBS containing 0.025% Triton X-100. Bacterial titers within the homogenates were determined by plating serial dilutions on LB agar plates.

Biofilm analysis
In vitro rugose biofilm assays were performed starting with cultures grown overnight at 37 • C shaking in LB, as described (47). Bacteria from each culture were then brought to an OD 600 of ∼1.0 and 10 l aliquots were spotted onto YESCA agar plates (12 g/l Casamino acids, 1.2 g yeast extract, 22 g agar) and incubated at RT (∼20-22 • C). After 14 days, biofilm images were acquired by focus stacking using an M.Zuiko Digital ED 60 mm lens mounted on an Olympus OM-D E-M1 Mark II camera.

Motility assays
Cultures of UTI89, UTI89 miaA, and UTI89 miaB grown overnight shaking in LB or M9 medium were brought to OD 600 of 1.0. Swim motility plates, containing 0.2% agar in LB or M9 medium, were inoculated with 2 l of each bacterial suspension delivered just below the agar surface. The diameter of bacterial spreading was measured every 1-2 h over the course of an 8-10 h-incubation at 37 • C. Swim rates were calculated during logarithmic growth. To assess the effects of MiaA and MiaB overexpression on motility, tryptone soft agar plates (48) containing 50 g/ml ampicillin and 100 M IPTG were inoculated with UTI89/pRR48, UTI89/pMiaA P tac and UTI89/pMiaB Ptac from overnight shaking cultures. Plates were imaged after a 6-h incubation at 37 • C.

Acid resistance assays
Bacterial strains from overnight cultures were diluted 1:100 in fresh LB and grown shaking at 37 • C for 3 h. Concentrated HCl was then added to each culture to adjust the pH to 3.0 and incubations were continued for another 30 min. Bacteria from 1 ml of each culture were then pelleted at 16 000 × g for 5 min and washed in PBS. Surviving bacteria were enumerated by plating serial dilutions on LB agar and normalized to input titers.

Osmotic stress resistance assays
UTI89/pACYC184, UTI89 miaA/pACYC184, UTI89 miaA/pMiaA nat , and UTI89 miaB were grown shaking overnight at 37 • C in 5 ml LB broth with 20 g/ml tetracycline and then back diluted 1:100 into 5 ml fresh LB (+ tetracycline). After 5 h shaking at 37 • C, a 1-ml aliquot of each culture was pelleted, resuspended in 1 ml of sterile water with or without 0.1% glucose, and incubations were continued for another 2 h with shaking at 37 • C. Viable bacteria present at 0, 30, 60, 90, and 120 min after resuspension in water were quantified by dilution plating and normalized to input titers. Growth curves in LB ± 5% NaCl were acquired as described above.

RT-qPCR analysis
UTI89 was diluted 1:100 from overnight cultures into fresh LB, grown shaking for 2.5 h at 37 • C prior to resuspension in LB or LB + 5% NaCl. After another 1-h incubation bacteria were pelleted and total RNA was extracted using the miRNeasy mini kit (QIAGEN). RNA samples were treated with RNase-Free DNase (QIAGEN) and cDNA was made using SuperScript IV VILO Master Mix (Invitrogen) according to the manufacturer's protocol. Quantitative PCR (qPCR) was carried out using primers listed in Supplementary Table S2 with the PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) on a QuantStudio 3 Real-Time PCR Instrument (Applied Biosystems). Replicas were made for each cDNA sample and miaA and miaB levels were normalized to rpoD, which is known to be stably expressed across a variety of growth stages and conditions (50,51). Products were resolved in 1.5% agarose gels, stained with ethidium bromide, and visualized using a GelDoc system (BioRad Technologies) to help verify the specificity of the RT-qPCR results.

UTI89,
UTI89 miaA, UTI89/pRR48, and UTI89/pMiaA Ptac carrying one of the dual-luciferase reporter plasmids (see Supplementary Table S1) were grown overnight in LB supplemented with chloramphenicol (20 g/ml) or ampicillin (100 g/ml). The cells were sub-cultured 1:100 into 6 ml LB with and without 1 mM IPTG. At an OD 600 of 0.2, arabinose (0.2%) was added to all of the cultures to induce expression of the luciferases. Cells were allowed to continue growing until reaching an OD 600 of 0.5, at which point the cultures were adjusted to an OD 600 of 1.0 and pelleted by spinning at 8000 × g for 1.5 min. The pellets were then subjected to one freeze-thaw cycle before being resuspended in Passive Lysis Buffer (Promega; E1910). One scoop of 0.15 mm Nucleic Acids Research, 2022, Vol. 50, No. 13 7575 zirconium oxide beads (Next Advance; ZrOB015) was added to each tube and bacteria were lysed using a Bullet Blender (Next Advance) set at speed 8 for 3 min. After a 30-s spin in a microfuge to pellet beads and any large debris, the supernatants were collected, and luciferase activities were analyzed as previously described (40). Briefly, the Dual-Luciferase Reporter Assay System (Promega) was used in combination with the Veritas Microplate Luminometer from Turner Biosystems to quantify activities of the two luciferases. Frameshifting was calculated by first determining the ratio of firefly to renilla luciferase activity for each sample, and then normalizing each out-of-frame construct (pCWR43 and pCWR45) with their associated in-frame control (pCWR42 and pCWR44, respectively).

Analysis of relative i 6 A and ms 2 i 6 A levels
UTI89 and UTI89 miaA were grown from frozen stocks shaking at 37 • C overnight in LB. UTI89/pRR48 and UTI89/pMiaA Ptac were grown similarly using LB supplemented with ampicillin (100 g/ml). The bacteria were subcultured 1:100 into 6 ml of LB ± 1 mM IPTG and then grown shaking to an OD 600 of 0.5. After adjusting the cultures to OD 600 of 1.0, the bacteria were pelleted by spinning at 8000 × g for 1.5 min. Pellets were then resuspended in 1 ml of RNAlater Stabilization Solution (ThermoFisher) and stored at 4 • C overnight prior to extraction of RNA using a Norgen Total RNA Extraction Kit.
RNA samples were analyzed using a Hypersil GOLD C18 column (2.1 mm × 150 mm, 1.9 m particle size; Thermo Fisher) attached to a Thermo Scientific Dionex UltiMate 3000 UHPLC instrument in line with an LTQ-OrbiTrap XL instrument (Thermo Fisher). The LC-MS parameters were based upon a procedure described previously (52,53), with the following adjustments. The UHPLC column was pre-equilibrated in 100% Buffer A [50 mM ammonium acetate (Fisher) in LC-MS Optima water]. Buffer B consisted of 60% (v/v) LC-MS Optima acetonitrile (Fisher) and 40% LC-MS water (Fisher). The reaction components were eluted at a rate of 0.2 ml/min with the following program: 0% B from 0 to 3.46 min, 0 to 0.9% B from 3.46 to 3.69 min, 0.9 to 1.5% B from 3.69 to 3.92 min, 1.5 to 3% B from 3.92 to 4.25 min, 3 to 20% B from 4.25 to 6.5 min, 20 to 25% B from 6.5 to 7 min, 25 to 40% B from 7 to 8.5 min, 40 to 45% B from 8.5 to 9.25 min, 45 to 60% B from 9.25 to 9.95 min, 60 to 100% B from 9.95 to 10.45 min, 100% B from 10.45 to 16 min, 100 to 0% B from 16 to 16.1 min, and 0% B from 16.1 to 20 min. The flow from the column was diverted to the mass spectrometer from 3.5 to 17 min during the UH-PLC program. The mass spectrometer was operated in positive ion mode, and authentic guanosine material (Sigma-Aldrich) was used to generate a tune file for the instrument. The observed m/z values of the +1 charge states of the i6a and ms2i6a RNA bases were 336.1658 and 382.1535, respectively. The observed retention times for i 6 A and ms 2 i 6 A were determined from the center of their extracted ion chromatogram peaks to be 15.55 and 16.45 min respectively. The retention time of i 6 A from biological extracts was consistent with the retention time of authentic i 6 A material (Cayman Chemical). Absolute intensities of the i 6 A ions were retrieved from the mass spectrum scanned between 15.4 and 16.1 min, while the absolute intensities of the ms 2 i 6 A ions were retrieved from the mass spectrum scanned between 16.2 and 17.0 min. The scan range was chosen to include the entire peak of an EIC trace, excluding mass spectral data recorded out of these bounds. This ensured that the intensities of ions 336.17 and 382.15 arise from eluted i 6 A and ms 2 i 6 A material and did not include background during the rest of the run. The scan windows were wide enough to account for any small drift in retention that might occur from sample to sample. Because total RNA concentration varied by sample, the samples were normalized against the total RNA concentration of each sample, as estimated via NanoDrop measurements at 260 nm.

Proteomics
The ExPEC isolate UTI89 and its derivatives UTI89 miaA, UTI89/pRR48, and UTI89/pMiaA Ptac were grown from frozen stocks in 5 ml of LB at 37 • C overnight in loosely capped 20-by-150-mm borosilicate glass tubes with shaking (225 rpm, with tubes tilted at a 30 • angle). UTI89/pRR48 and UTI89/pMiaA Ptac were grown overnight in the presence of ampicillin (100 g/ml) to help maintain the plasmids, but the antibiotic was not included in subsequent steps. Overnight cultures were brought to an OD 600 of ∼1.0 and then sub-cultured 1:100 into LB and grown to mid-log phase (OD 600 ∼ 0.5) in LB shaking at 37 • C. IPTG (1 mM) was included for UTI89/pRR48 and UTI89/pMiaA Ptac . About 1 × 10 9 CFU from each culture was pelleted at 8000 × g for 1.5 min. Supernatants were then removed, and cells from three independent replicates were plunged into liquid nitrogen. Shotgun proteomic analysis of cell lysates was performed with the MSRC Proteomics Core at Vanderbilt University by first partially resolving 20 g of protein lysates about 1.5 cm using a 10% Novex precast gel, excising the protein region, and then performing in-gel tryptic digestion to recover peptides. These peptides were analyzed by high-resolution data dependent LC-MS/MS. Briefly, samples were autosampled onto a 200 mm by 0.1 mm (Jupiter 3 micron, 300A) self-packed analytical column coupled directly to a Q-exactive plus mass spectrometer (ThermoFisher) using a nanoelectrospray source and resolved using an aqueous to organic gradient. Both the intact masses (MS) and fragmentation patterns (MS/MS) of the peptides were collected in a data-dependent manner utilizing dynamic exclusion to maximize depth of coverage. Resulting peptide MS/MS spectral data were searched against the UTI89 protein database using MaxQuant-LFQ along with subsequent MS1-based integrations and normalizations (54).
Label-free quantification (LFQ) values were loaded into Prostar software for statistical analysis and visualization. The data set was filtered by requiring all conditions to contain at least two values. Imputation for partially observed values was done with the Structured Least Square Adaptative algorithm. Imputation for conditions in which values were missing for a specific protein in all three biological replicates used the DetQuantile algorithm with the settings Quantile:2.5 and Factor:1. Statistical analysis was performed using the 1vs1 settings and Student's t-tests. Differentially expressed proteins were categorized (Supplementary Dataset S1) based on literature searches and information drawn from EcoCyc ((55); http://ecocyc.org/), STRING Protein-Protein Interaction Networks Functional Enrichment Analysis ((56); https:// string-db.org/), and Phyre2 ((57); http://www.sbg.bio.ic.ac. uk/~phyre2/html/page.cgi?id=index).

Statistical analyses
The proteomics data were analyzed as described above. Two-sided Spearman's rank correlation coefficients (rho) and FDR-corrected P values were calculated using R version 4.0.5. For all other data, P values were determined by Log-Rank (Mantel-Cox), Mann-Whitney U tests, ANOVA, or two-tailed Student's t-tests performed using Prism 9.0.0 software, with corrections as indicated (GraphPad Software). AUC values for growth curves were also calculated and graphed using Prism software. Data distribution normality (Gaussian) was not assumed, such that non-parametric tests were used for the mouse experiments. P values of less than or equal to 0.05 were defined as significant.

MiaA promotes ExPEC fitness and virulence in vivo
To assess the importance of MiaA and MiaB for Ex-PEC within varied host environments, we employed wellestablished mouse models of gut colonization, urinary tract infection (UTI), and bloodstream infection (58). For these and subsequent experiments, miaA and miaB were independently deleted from the ExPEC reference strain UTI89 to generate the isogenic knockout mutants UTI89 miaA and UTI89 miaB (38,39).
Gut colonization. The mammalian gastrointestinal (GI) tract serves as a major reservoir for ExPEC that can seed extraintestinal infections (59)(60)(61)(62)(63). Roles for MiaA and MiaB in ExPEC colonization of the GI tract were probed using competitive assays in which ∼10 9 colony forming units (CFU) of a 1:1 mixture of UTI89 and either UTI89 miaA or UTI89 miaB were introduced into adult specific-pathogen-free (SPF) BALB/c mice via intragastric gavage (43)(44)(45). In this model system, the levels of ExPEC recovered from the feces reflect ExPEC titers within the large intestines (44). For these assays, UTI89 and the miaA and miaB knockout mutants were engineered to express either kanamycin (Kan R ) or chloramphenicol (Cam R ) resistance cassettes so that the strains could be readily identified by plating fecal homogenates on selective media. Feces were collected at the indicated time points and the numbers of viable bacteria were enumerated to determine competitive indices (CI). UTI89 miaA was significantly outcompeted by wild-type UTI89 as early as day 3 post-inoculation ( Figure  1B). By day 10, there was about a 25 000-fold reduction in the relative levels of UTI89 miaA recovered from the feces, correlating with a median CI of -4.39. At this time point, UTI89 miaA titers in the majority of mice were below the limit of detection. In contrast, there were no notable differences in titers between UTI89 miaB and UTI89 in the feces at any time point ( Figure 1C). These results indicate that the loss of MiaA, but not MiaB, greatly impairs the fitness of UTI89 within the gut.
UTI. During the course of a UTI, ExPEC is able to bind and invade the host epithelial cells that comprise the bladder mucosa (64). Once internalized into bladder cells, ExPEC can traffic into late endosome-like compartments where it may form quiescent reservoir populations that promote long-term bacterial persistence. Alternatively, ExPEC can enter the host cytosol and rapidly multiply, forming large intracellular bacterial communities that eventually rupture the epithelial cell. In cell culture-based assays using a bladder epithelial cell line, we found that UTI89 miaA and UTI89 miaB are able to bind, invade, and survive intracellularly in overnight assays much like wild-type UTI89 (Supplementary Figure S3).
To investigate MiaA and MiaB requirements during UTI, 10 7 CFU of wild-type UTI89, UTI89 miaA, and UTI89 miaB were independently inoculated via transurethral catheterization into adult female SPF CBA/J mice and bacterial titers in the bladders were determined after 3 days. In this analysis, UTI89 miaB showed no statistically significant defect relative to the parent strain, whereas the miaA strain was clearly attenuated ( Figure 1D). Deficiencies in bladder colonization by UTI89 miaA were apparent by 6 h post-inoculation (Supplementary Figure S4A) and were still significant after 9 days (Supplementary Figure  S4B). The differences observed between wild-type UTI89 and UTI89 miaA at 3 days post-inoculation of CBA/J mice were also manifest in C3H/HeJ mice (Supplementary Figure S4C). Due to defects in Toll-like receptor 4 (TLR4) signaling and other innate defenses, C3H/HeJ mice have attenuated inflammatory responses and increased susceptibility to UTI (65)(66)(67)(68). Our results indicate that the decreased capacity of UTI89 miaA to colonize the bladder is not attributable to an inability of the miaA knockout to handle TLR4-dependent innate host defenses. Collectively, these results indicate that MiaA is required for maximal fitness in mouse UTI models, while MiaB is less critical.
Bloodstream infection. ExPEC is a leading cause of bloodstream infections, which too often trigger discordant systemic inflammatory responses that can result in a lifethreatening condition known as sepsis (69). To examine the contributions of MiaA and MiaB to ExPEC virulence and fitness in a model of sepsis, adult SPF C57Bl/6 mice were inoculated via intraperitoneal (i.p.) injections with ∼10 7 CFU of wild-type UTI89, UTI89 miaA, or UTI89 miaB. Following i.p. injection, the bacteria enter the bloodstream and disseminate (43,70,71). In our experiments, only 15% (2/13) of the mice infected with wild-type UTI89, and 0% (0/13) of the mice injected with UTI89 miaB, were viable after 48 h ( Figure 1E). In sharp contrast, 84% (11/13) of the mice infected with UTI89 miaA survived. At 6 h post-injection, significantly lower numbers of bacteria were recovered from the spleens and kidneys of UTI89 miaA-infected mice, relative to mice infected with wild-type UTI89 or UTI89 miaB (Supplementary Figure S5A and B). While not significant, titers in the liver also trended lower in UTI89 miaA-infected mice (Supplementary Figure S5C). Combined, these data demonstrate that MiaA is important Nucleic Acids Research, 2022, Vol. 50, No. 13 7577 for the virulence of ExPEC and its survival during systemic infections, while MiaB appears dispensable.

MiaA enhances ExPEC growth and stress resistance
Earlier studies showed that K-12 E. coli and Salmonella mutants lacking miaA are moderately impaired in nutrientrich broth, but less so in nutrient-limited media (19,26,72). Using in vitro growth assays, we found that UTI89 miaA grew normally in modified M9 minimal media but failed to reach densities as high as the wild-type strain in more complex, nutrient-rich lysogeny broth (LB) (Figure 2A and B). In contrast, the miaB knockout exhibited no overt growth defects in either type of media. These data suggest that UTI89 miaA has reduced metabolic flexibility relative to wild-type UTI89 and the miaB mutant. This may contribute to the decreased fitness of UTI89 miaA in our mouse models, where the bacteria likely encounter marked shifts in nutrient availability (58,73,74). However, within different host environments ExPEC will face a wide variety of additional challenges that might be countered by MiaA-dependent processes. We investigated this possibility by examining the effects of MiaA and MiaB on ExPEC resistance to nitrosative, oxidative, and osmotic stress.
Oxidative and nitrosative stress. During the course of an infection, both host and bacterial cells can produce reactive oxygen and nitrogen radicals that can damage lipids, proteins, and nucleic acids (75,76). The contributions of MiaA and MiaB to nitrosative and oxidative stress resistance were assessed using acidified sodium nitrite (ASN) and methyl viologen (MV), respectively. When added to low pH morpholineethanesulfonic acid (MES)-buffered LB (MES-LB; pH 5.0), sodium nitrite dismutates to form nitrous acid which in turn generates NO and other harmful reactive nitrogen species (77). In un-supplemented MES-LB, UTI89 miaA reached a lower maximal density than wild-type UTI89 ( Figure 2C), similar to results obtained using standard LB ( Figure 2B). The addition of 1 mM ASN delayed entry of UTI89 miaA into exponential growth phase by close to 4 h ( Figure 2D), while 2 mM ASN delayed growth by more than 15 h relative to wild-type UTI89 ( Figure 2E). The addition of 1 mM MV, which produces superoxide radicals (78), had even stronger inhibitory effects on growth of UTI89 miaA ( Figure 2F). In contrast, UTI89 miaB grew much like the wild-type strain in the presence of ASN or MV ( Figure 2D-F). Complementation with pMiaA nat , a low copy plasmid that encodes MiaA under control of its native promoter, restored growth of UTI89 miaA to near wild-type levels in both 2 mM ASN ( Figure 2G) and in 1 mM MV ( Figure 2H).
Osmotic stress. During a UTI, osmotic pressure within the bladder can shift from 50 to >1400 mOsm/kg due to varying concentrations of solutes like sodium and urea (79,80). By comparison, the normal osmolarity of blood ranges from about 275 to 295 mOsm/kg. To test the sensitivities of wild-type UTI89 and the knockout strains to hypoosmotic stress, we diluted the bacteria from early stationary phase cultures into ddH 2 O, and then quantified the numbers of viable bacteria every 30 min over the course of 2 h. Titers of UTI89ΔmiaA carrying the empty vector pACYC184 were greatly reduced following exposure to hypoosmotic stress, whereas the levels of UTI89/pACYC184 and UTI89 miaB/pACYC184 remained mostly unchanged ( Figure 3A). Survival of UTI89 miaA was restored by complementation with pMiaA nat . To ensure that reduced survival of UTI89 miaA was attributable to hypoosmotic stress and not starvation, cells were resuspended in ddH 2 O containing 0.1% glucose, which is comparable to the glucose levels within our M9 medium. Viable bacteria measured after 120 min indicated that the death of UTI89 miaA was not due to nutrient deprivation ( Figure 3B). We also observed that UTI89 miaA grew poorly in hyperosmotic conditions, created by addition of 5% NaCl to standard LB, whereas UTI89 miaB behaved more like the wild-type strain ( Figure 3C). Growth of UTI89 miaA was restored to wild-type levels by complementation with pMiaA nat ( Figure 3D). These results indicate that the miaA knockout has decreased resistance to both hypo-and hyperosmotic stresses.

Hyperosmotic stress attenuates MiaA translation
We next examined how MiaA levels in UTI89 change in response to environmental cues, focusing on hyperosmotic stress. For these assays, we employed a low-copy number plasmid (pMiaA-Flag nat ) that encodes C-terminal FLAGtagged MiaA under control of the native miaA promoter. Mid-logarithmic phase cultures of UTI89/pMiaA-Flag nat were resuspended in LB ± 5% NaCl and levels of MiaA-Flag were then assessed by western blot at 30-min intervals over the course of 1.5 h ( Figure 4A). Interestingly, MiaA levels in UTI89 exposed to high salt broth were decreased at all time points in comparison with bacteria grown in standard LB. We observed a similar phenomenon if overnight cultures of UTI89/pMiaA-Flag nat in standard LB were back-diluted into high salt broth and then grown to mid-logarithmic phase (OD 600 ≈ 0.5, Figure 4B). Importantly, in these assays we observed no loss of the pMiaA-Flag nat construct. In addition, miaA transcripts were often elevated following exposure of UTI89 to high salt ( Figure  4C), suggesting that the downregulation of MiaA protein levels in response to this osmotic stress occurs via a posttranscriptional mechanism. The transcription of miaB was generally reduced under the same conditions ( Figure 4D).
To determine if lower amounts of the MiaA protein detected in high salt broth culture affected i 6 A or ms 2 i 6 A levels, we employed liquid chromatography-coupled mass spectrometry (LC-MS). Normalized amounts of the i 6 A modification in wild-type UT89 grown to mid-logarithmic phase in LB were similar to those measured in UTI89 grown in high salt broth ( Figure 4E). However, hyperosmotic stress caused a marked reduction in ms 2 i 6 A levels ( Figure 4F), possibly due to reduced transcription of miaB ( Figure 4D). i 6 A and ms 2 i 6 A were undetectable in UTI89 miaA, regardless of high salt exposure, confirming that MiaA is required for both modifications ( Figure 4E and F). In contrast, deletion of miaB prevented formation of ms 2 i 6 A, but led to greatly elevated levels of i 6 A (Supplementary Figure S6). Cumulatively, these data indicate that in response to hyperosmotic stress UTI89 can post-transcriptionally downreg-   UTI89/pMiaA-Flag nat was diluted from overnight cultures into fresh LB and grown shaking for 2.5 h at 37 • C prior to resuspension in LB or LB + 5% NaCl. Incubations were continued for the indicated times before samples were collected and analyzed by western blots using anti-Flag and anti-E. coli (loading control) antibodies (bottom panel). (B) Top panel summarizes the experimental setup in which overnight cultures of UTI89/pMiaA-Flag nat were diluted directly into fresh LB or LB + 5% NaCl and grown shaking to OD 600 of 0.5 prior to processing for western blot analysis (bottom panel). (C and D) UTI89 from mid-logarithmic cultures in LB was resuspended in LB or LB + 5% NaCl and 1 h later the levels of miaA and miaB transcripts were determined by RT-qPCR. Bars indicate mean values from 9 independent replicates, each with two technical replicates. *P < 0.05; ***P < 0.001 by Mann-Whitney U tests. (E and F) Graphs show normalized levels of i 6 A and ms 2 i 6 A recovered from UTI89 and UTI89 miaA following growth to OD 600 of 0.5 in LB or LB + 5% NaCl, as determined by LC-MS. **P < 0.01; ***P < 0.001; ****P < 0.0001 by unpaired t tests. Bars indicate median values from four to six independent replicates. (G(i)) Schematic of the experimental setup in which 1 mM (ii) MV or (iii) EDTA were added to UTI89/pMiaA-Flag nat cultures for 1 h prior to processing for western blot analyses. ulate MiaA, coordinate with reduction of both miaB messages and ms 2 i 6 A levels. We observed that MiaA levels were also reduced upon exposure to MV ( Figure 4G-i, ii), indicating that MiaA levels can be varied in response to distinct stressors. Interestingly, MiaA levels were intensified in the presence of the divalent metal chelator EDTA ( Figure 4Giii), raising the possibility that the quantities of this tRNA modifying enzyme are controlled by one or more EDTAsensitive metalloproteases as previously suggested for the tRNA modifier GidA (81). This idea requires further testing, as the chelating activity of EDTA can potentially impact multiple cellular processes that require divalent metals.

Overexpression of MiaA is detrimental under stressful conditions
Since it was unexpected that high salt stress would lead to a decrease in the levels of MiaA and the ms 2 i 6 A modification, we set out to determine if overexpression of MiaA would affect bacterial growth during environmental stress. To overexpress MiaA, we utilized a plasmid (pRR48) with miaA under control of an IPTG-inducible Ptac promoter in wild-type UTI89. By LC-MS, normalized intensities of i 6 A were significantly higher in UTI89/pMiaA Ptac induced with 1 mM IPTG and grown to mid-logarithmic phase in LB compared to UTI89 carrying the empty vector pRR48, whereas the amounts of ms 2 i 6 A were only modestly elevated ( Figure 5A).
Next, overnight cultures of UTI89/pMiaA Ptac were backdiluted into LB, LB + 1 mM MV, LB + 5% NaCl, MES-LB, or MES-LB + 1 mM ASN, and grown in the presence of increasing IPTG concentrations ( Figure 5B-F). Lower levels of MiaA protein induction caused no overt defects and the bacteria grew much like UTI89/pRR48. However, higher levels of IPTG-induced MiaA expression hindered growth of UTI89/pMiaA Ptac in the presence of 1 mM MV, 5% NaCl, MES-LB, and 1 mM ASN. In contrast, over expression of MiaB did not affect bacteria growth in these in vitro assays (Supplementary Figure S7). These findings indicate that too much MiaA can be detrimental to bacterial fitness, similar to the complete absence of the enzyme.

Both deletion and overexpression of MiaA increase frameshifting
Previous research in K-12 E. coli and Salmonella showed that deletion of miaA can compromise translational fidelity, resulting in increased ribosomal frameshifting (82,83). To determine the effects of MiaA on frameshifting in UTI89, we utilized dual-luciferase reporter plasmids that consist  Table S1). Linker sequences, derived from either Antizyme 1 (Az1) or HIV gagpol, were placed between the two luciferase genes ( Figure  6A). The Az1-derived linker sequence contains a stop codon positioned in-frame so that a + 1 frameshift must occur for read-through expression of firefly luciferase (41). In contrast, a -1 frameshift is required for expression of firefly luciferase downstream of the HIV-derived linker (40). Importantly, upstream of the in-frame stop codons in both linkers are UNN codons that can be recognized by MiaAmodified tRNAs. The firefly and renilla luciferases act on distinct substrates, which are used to sequentially assess levels of expression of each enzyme (40,41). Control plasmids in which the two luciferases are in-frame were used to normalize the data by accounting for ribosome drop-off.
To examine the consequences of MiaA expression on frameshifting, the dual-luciferase reporter constructs were used in combination with wild-type UTI89, UTI89 miaA, UTI89/pMiaA Ptac , and UTI89 carrying the empty control vector pRR48. After overnight growth in LB, UTI89 and UTI89 miaA were back-diluted into LB while UTI89/pMiaA Ptac and UTI89/pRR48 were back-diluted into LB + 1 mM IPTG to induce MiaA expression. After reaching mid-log growth, the enzymatic activities of the two luciferases were quantified. Both the lack of MiaA and MiaA overexpression caused notable increases in frameshifting in both the +1 and -1 directions ( Figure  6B and C). These results confirm that loss of MiaA can increase frameshifting and show that elevated MiaA levels can likewise impact the fidelity of translation.

Changing levels of MiaA alters the spectrum of expressed proteins
To determine how deletion and overexpression of MiaA affect translation we used multidimensional protein identi-fication technology (MudPIT; LCMS/MS) with wild-type UTI89 and UTI89 miaA cultures grown to mid-log phase in LB, and UTI89/pMiaA Ptac and UTI89/pRR48 similarly grown in LB + 1 mM IPTG. Of 1524 proteins detected in UTI89 and UTI89 miaA, 105 were picked up only in the wild-type strain and 23 were unique to the miaA knockout mutant ( Figure 7A). 1,471 proteins were identified in UTI89/pRR48 and UTI89/pMiaA Ptac , with 42 being exclusive to UTI89/pRR48 and 20 seen only in the MiaA overexpression strain ( Figure 7B). 115 proteins were significantly downregulated in UTI89 miaA relative to wild-type UTI89, while 34 proteins were upregulated in the knockout mutant ( Figure 7C). Notably fewer proteins were significantly altered when MiaA was overexpressed. Relative to the control strain UTI89/pRR48, 20 proteins were downregulated in UTI89/pMiaA Ptac , whereas nine (including MiaA) were upregulated ( Figure 7D).
The specific proteins detected, including those that were differentially expressed due to miaA deletion or overexpression, are detailed in Supplementary Dataset S1. The differentially expressed proteins were assigned to one or more of 14 functional categories (see Categories worksheet and embedded graph in Supplementary Dataset S1). A majority of the altered proteins were linked with metabolic pathways, secondary metabolites, and functions associated with the bacterial envelope. These included several proteins involved in sugar and fatty acid metabolism and the biosynthesis and regulation of electron transport chains (e.g. UbiC, WrbA, ChrR, Qor, NuoM, NudJ and CyoC). The dysregulation of these factors likely contributed to the various phenotypic defects observed in our in vitro and in vivo assays and suggested MiaA involvement in other important processes.
In particular, many of the differentially expressed proteins were shown in previous studies to directly or indirectly affect motility or biofilm development. The former group comprised the chemotaxis protein CheA and the flagella-associated proteins FliF, FlhA and FlgH. Accord-

(B and C) Graphs
show results from + 1 and -1 frameshifting assays with UTI89, UTI89 miaA, UTI89/pRR48, or UTI89/pMiaA Ptac carrying one of the dual luciferase reporter constructs. Bacteria were grown shaking at 37 • C in LB, with 1 mM IPTG included for UTI89/pRR48 and UTI89/pMiaA Ptac . After reaching an OD 600 of ∼0.2, 0.2% arabinose was added to induce expression of the luciferases. At an OD 600 of 0.5, translational error rates were quantified by determining the ratio of firefly to renilla luciferase activities in bacteria carrying the +1 (Az1) and -1 (HIV) reporter constructs. Results were normalized using the ratio of firefly to renilla luciferase activity in bacteria carrying control plasmids in which the luciferases are in-frame. **P < 0.01; ****P < 0.0001 by two-tailed unpaired t tests; n = 10-14 independent replicates. ingly, both deletion of the miaA gene and MiaA overexpression markedly decreased UTI89 motility on swim plates (Supplementary Figure S8). MiaB did not affect motility in these assays. Factors linked with biofilm development include YoaB, the type 1 pilus-associated regulator FimB and periplasmic chaperone FimC, the acid stress-response chaperone HdeB, the cellulose synthase catalytic subunit BcsA, and the cytochrome bo subunit CyoC. Using yeast extractcasamino acids (YESCA) medium, which promotes the development of elaborate rugose-colony biofilms (47,84), we found that UTI89 miaA, but not UTI89 miaB, formed atypical biofilms with notably less rugosity than the parent strain (Supplementary Figure S9). Interestingly, the biofilms formed by UTI89 miaA were architecturally similar to those formed by a UTI89 mutant lacking the CyoCinteracting partners CyoAB (47).
Our MudPIT results also indicated that MiaA can regulate numerous proteins that have been associated with redox and bacterial responses to nitrosative, oxidative, and more generally, genotoxic stresses (Supplementary Dataset S1). Aberrant expression of these factors, including proteins like GadB, CadA, Dps, glutathione S-transferase Gst, and the glutatredoxins GrxB and GrxC, may account for increased sensitivity to oxygen and nitrogen radicals (see Figure 2D-H and Figure 5D-F). Some of these factors, and others like HdeA and HdeB, can also guard against acid stress.
Follow-up experiments confirmed that UTI89 miaA, but not UTI89 miaB, is notably less resistant to acid stress than the wild-type strain (Supplementary Figure S10). On average, relative to wild-type UTI89, UTI89 miaA titers were reduced over 6,000-fold following exposure to acidic conditions in LB.
The sensitivity of both UTI89 miaA and UTI89/pMiaA Ptac to osmotic stress may arise due to the significant downregulation of proteins like SLP, BetB, YggT, ProP, and YnaI (Supplementary Dataset S1). Other differentially expressed proteins that probably contribute to the varied phenotypes associated with miaA deletion or MiaA overexpression in our assays include multiple transcriptional regulators, several ribosome-and RNAassociated factors, and the tRNA ligases LysU, TyrS, and PheS. These findings indicate that MiaA is tied into a complex web of factors that can have direct and indirect effects on translation. Driving this point home is the observation that MiaA overexpression suppresses the production of TadA, an enzyme that catalyzes the deamination of adenosine-to-inosine (A-to-I) in Arg2 tRNA and a select set of mRNAs (85,86). In K-12 E. coli, the A-to-I editing function of TadA can recode at least 12 mRNAs, which results in the generation of proteins with altered activities that can impact bacterial cell physiology (86). Among the known TadA-edited transcripts is one encoding IlvC,  A and B) Venn diagrams indicate numbers of unique and shared proteins detected in wild-type UTI89 versus UTI89 miaA, or in UTI89/pRR48 versus UTI89/pMiaA Ptac , following growth to OD 600 of 0.5 in LB. IPTG (1 mM) was included in the UTI89/pRR48 and UTI89/pMiaA Ptac cultures. Relative protein levels were determined by MudPIT. (C and D) Volcano plots show relative protein levels (Log2-fold change) versus P values (-Log10). Vertical dashed lines denote 2-fold change cutoffs, while the horizontal dashed lines indicate a P value of 0.05. Blue dots, proteins that were significantly changed (P < 0.05) by at least 2-fold. P values were determined by Student's t tests; n = 4 independent replicates for each group. (E-J) Plots show relative protein levels (Log2-fold change as noted in C and D) versus UNN codon usage ratios for Leu, Ser, or total UNN codons per open reading frame. For I and J, total codons refer specifically to those for Leu, Ser, Cys, Phe, Trp and Tyr. Purple circles, proteins that were significantly changed (P < 0.05, by Student's t tests) by at least 2-fold in UTI89 miaA or UTI89/pMiaA Ptac relative to their controls. Positions for Hfq and RpoS are highlighted (yellow). The vertical dotted lines are placed at the 2-fold change cutoffs. Black and green horizontal dashed lines specify the mean UNN codon usage ratios for all open reading frames encoded by UTI89 and the K-12 strain MG1655, respectively. an enzyme involved in isoleucine and valine biosynthesis which, like TadA, is downregulated ∼3.5-fold in UTI89 when MiaA is overexpressed ( Figure 7D).

UTI89 miaA phenotypes are not entirely due to aberrant RpoS or Hfq expression
In K-12 E. coli, the deletion of miaA results in decreased translation of the alternate Sigma factor RpoS ( S ) and the small RNA chaperone Hfq (7,22,23). Both of these factors are important for the stress resistance and virulence potential of ExPEC (68,87). In line with results from K-12 E. coli, our proteomics analysis indicated that RpoS and Hfq levels were reduced 2.5-and 2.8-fold, respectively, in UTI89 miaA relative to the wild-type strain ( Figure  7C and Supplementary Dataset S1). RpoS downregulation in the absence of miaA was also confirmed by western blot analysis (Supplementary Figure S11A) These observations suggest that the phenotypic defects associated with UTI89 miaA might be attributable to aberrant expression of RpoS or Hfq. However, despite some similarities, the phenotypes that we previously observed with UTI89 mutants lacking either rpoS or hfq are distinct from one another and from those that we report here with UTI89 miaA (68,87). Furthermore, the induced expression of recombinant RpoS or Hfq failed to rescue growth of UTI89 miaA under hyperosmotic conditions (Supplementary Figure S11B-E). The pRpoS Ptac and pHfq Ptac expression constructs used in these assays can complement UTI89 mutants lacking rpoS or hfq, respectively (68,87). Cumulatively, these data indicate that the phenotypes seen with UTI89 miaA are not entirely due to attenuated expression of either RpoS or Hfq. Also of note, our ability to complement UTI89 miaA with MiaA expression constructs (see Figures 2 and 3, and Supplementary Figure S9) demonstrates that the phenotypic defects associated with this knockout mutant are not caused by off target mutations or polar effects on hfq, which lies immediately downstream of miaA.

UNN codon usage by MiaA-sensitive transcripts
Messages like those encoded by rpoS and hfq are classified as Modification Tunable Transcripts (MoTTs), which are identifiable by 1) codon usage different from that of average transcripts and 2) translation that is sensitive to changing levels of tRNA modifications (13,23). Studies in the K-12 E. coli strain MG1655 of rpoS, hfq, and other transcripts suggest that MiaA-sensitive MoTTs have higher than average ratios of UNN-Leu codons relative to total Leu codons (22,23). This led us to ask if UNN-Leu codon usage correlates with protein expression levels in UTI89 when MiaA is either absent or over-produced. Plotting results from our MudPIT analysis versus UNN-Leu codon usage (Supplementary Dataset S1) showed that just over 60% of the proteins that are differentially expressed in UTI89 miaA or UTI89/pMiaA Ptac have UNN-Leu codon usage ratios that are greater than the K-12 average ratio of 0.22 ( Figure  7E-F, green dashed line). In agreement with previous findings (22,23), RpoS and Hfq are among the differentially expressed proteins with UNN-Leu codon usage ratios of greater than 0.22. However, the average UNN-Leu codon usage ratio in UTI89 is somewhat higher than that in K-12 E. coli. Using this value, which is 0.28, less than half of the proteins that are differentially regulated in UTI89 miaA or UTI89/pMiaA Ptac have greater than average UNN-Leu codon usage ratios ( Figure 7E-F, black dashed line). Furthermore, among the proteins that are not significantly altered by either deletion or overexpression of miaA, about 30% have UNN-Leu codon usage ratios greater than 0.28. By calculating Spearman's rank correlation coefficients ( ), we found a significant, though modest negative correlation between protein abundance and UNN-Leu codon usage ratios for UTI89 miaA ( = -0.07, P = 0. Building on these observations, we next assessed if changes in protein abundance due to MiaA deletion or overexpression were affected by the numbers of single or consecutive UNN codons within each open reading frame, but not UNN codon usage ratios per se (see Supplementary Dataset S1). Spearman's rank correlation coefficients indicated significant negative correlations between protein abundance in UTI89 miaA and the presence of one, two, or three consecutive UNN codons (Supplementary Table S3). Similar trends were detected in UTI89/pMiaA Ptac . The correlations were slight when considering the less frequent instances in which four or more consecutive UNN codons were present. Negative correlations between protein abundance and total UNN codons per open reading frame were also evident in UTI89 miaA and especially UTI89/pMiaA Ptac , independent of UNN positions within the coding sequences. Cumulatively, these results indicate that higher numbers of UNN codons, especially singlet and doublet UNN codons, may render translation of specific transcripts more sensitive to varying MiaA levels.

Conditional negative effects of DMAPP binding when MiaA levels are elevated
To better understand why overexpression of MiaA, but not MiaB, rendered UTI89 more sensitive to stress (see Figure  5 and Supplementary Figure S7), we introduced a series of point mutations into MiaA that are known to alter the substrate binding and catalytic activities of the enzyme (88). A crystal structure-based model of MiaA in complex with tRNA and the prenyl group precursor DMAPP is shown in Figure 8A, with the mutated residues indicated. The T19A mutation is expected to decrease the catalytic activity (K cat ) of MiaA by ∼600-fold relative to the wild-type enzyme, with little effect on the Michaelis constants (K m ) for either tRNA (K m RNA ) or DMAPP (K m DMAPP ) (88). In contrast, the T24A mutation increases the K m RNA 10-20 times higher than wild-type MiaA and elevates K m DMAPP by more than 200-fold. The K280A mutation causes >30-fold increases in both K m RNA and K m DMAPP values. We also generated a double T19A/K280A mutant that is expected to bind tRNA and DMAPP poorly and completely lack catalytic activity. Mutations were introduced into plasmid pMiaA-Flag Ptac , encoding IPTG-inducible MiaA with a C-terminal Flag tag. The wild-type and mutant MiaA variants were all expressed at similar levels in UTI89 following IPTG induction (Supplementary Figure S13).
As expected from results presented in Figure 5, IPTGinduced overexpression of Flag-tagged wild-type MiaA markedly attenuated growth of UTI89 in the presence of both high salt and MV ( Figure 8B). In high salt, overexpression of the K280A and catalytically inactive T19A MiaA mutants were nearly as detrimental to bacterial growth as wild-type MiaA. In contrast, overexpression of the T24A and T19A/K280A mutants had no adverse effects on UTI89 growth in high salt. In MV, the overexpression of all four MiaA mutants impaired growth of UTI89, but none were as detrimental as overexpression of wild-type MiaA. These results suggest that high-level expression of MiaA can impair bacterial fitness via distinct mechanisms, depending on the specific environmental stressors present.
Data presented in Figure 8B argue that the attenuated growth of UTI89 caused by MiaA overexpression in high salt is mostly attributable to increased binding of DMAPP. In addition to serving as a precursor for MiaA-mediated tRNA modifications, DMAPP is also an important cellular metabolite for a variety of other biosynthetic pathways (89,90). To assess the possibility that reduction of DMAPP precursor pools contributed to some of the observed defects caused by MiaA overexpression, we performed growth curve analyses of UTI89/pMiaA Ptac using media supplemented with DMAPP or its precursor HDMAPP. The addition of DMAPP, which is highly membrane impermeable (91), had no discernable effects. However, in high salt media with IPTG present to induce MiaA overexpression, addition of the DMAPP precursor HDMAPP rescued growth of UTI89/pMiaA Ptac in a dose-dependent manner ( Figure  8C). Added HDMAPP did not have an effect on growth of UTI89/pMiaA Ptac in the absence of IPTG, and failed to rescue bacterial growth when MiaA was overexpressed in the presence of MV. These results are in line with observations made with the MiaA point mutants ( Figure 8B), suggesting that the UTI89 growth defects associated with MiaA overexpression in high salt media, but not with MV, are in large part caused by depletion of DMAPP pools.

DISCUSSION
The results presented here demonstrate that MiaA is crucial for ExPEC fitness and virulence, and that changing MiaA levels can impact the translation of a broad spectrum of pro-Nucleic Acids Research, 2022, Vol. 50, No. 13 7585 teins. Our findings are in agreement with previously published work showing that tRNA modifying enzymes can influence the fitness and virulence potential of a variety of microbial pathogens (6,11,(92)(93)(94)(95)(96)(97)(98)(99)(100)(101)(102)(103)(104). The attenuation of bacterial virulence-related phenotypes in the absence of a specific tRNA modifying enzyme can, in some cases, be explained by sub-optimal translation of specific toxins or key regulatory factors (11,(102)(103)(104). For example, deletion of miaA in the diarrheagenic bacteria Shigella flexneri ablates translation of the transcriptional master regulator VirF, resulting in the reduced expression of downstream virulence factors (95,105). Overexpression of recombinant VirF alone is sufficient to rescue the miaA mutant, suggesting that lowlevel production of VirF is in large part responsible for the virulence-related defects caused by the deletion of miaA in S. flexneri. In contrast, our work indicates that the diverse phenotypes affected by MiaA expression in the ExPEC isolate UTI89 are not attributable to any single factor, but rather arise due to the altered expression and dysregulation of multiple proteins and pathways downstream of MiaA.
The ms 2 i 6 A modification is understood to affect the fidelity of translation (5,8,19). Earlier work in lab-adapted K-12 E. coli and Salmonella strains showed that bacteria lacking miaA have an increase in the + 1 direction of frameshifting, but not in the -1 direction (82,83). In these studies, the i 6 A modification was found to be a major contributor to ribosome fidelity. In UTI89, significant increases in frameshifting were seen in both the + 1 and -1 directions when miaA was knocked out. When MiaA was overproduced, we also observed marked elevation of frameshifting in the -1 direction, while frameshifting levels in the +1 direction were more modest. Most reports to date indicate that tRNA modifications typically affect frameshifting primarily in one direction (5,83). At first glance, our data seemingly counter this trend. However, we note that the expression of firefly luciferase downstream of the HIV-derived linker in our reporter system may also occur as a consequence of a +2 frameshift, rather than a -1 frameshift, which would more closely mirror what was observed in previous studies with K-12 E. coli and Salmonella strains (82,83). It is also possible that the apparent increases in both -1 and +1 frameshifting observed in our assays reflect the presence of MiaA-sensitive regulatory circuits in UTI89 that are different from those in lab-adapted K-12 E. coli or Salmonella strains.
During the course of this study, we were surprised to observe that MiaA levels in the ExPEC reference strain UTI89 were substantially decreased in response to high salt and other stressors like MV (see Figure 4). Due to the sweeping phenotypes seen in the absence of MiaA, we hypothesized that the levels of MiaA would have stayed the same or increased in response to stressors like high salt and MV. However, our data indicate that MiaA levels are finetuned within ExPEC such that too much or too little enzyme can have similarly detrimental consequences. The effects of MiaA expression in our growth curve assays were dose-dependent, with high-level expression of MiaA being nearly as disruptive as the deletion of miaA. For instance, low-level expression of MiaA restored the resistance of UTI89 miaA to high salt, whereas overexpression of MiaA resulted in greatly increased sensitivity (see Figure   5). Defective growth of UTI89 caused by overexpression of MiaA during high salt stress was offset if the enzyme was mutated to disrupt DMAPP binding or if the media was supplemented with the DMAPP precursor HDMAPP (see Figure 8A-C). These observations suggest that the growth defects associated with MiaA overexpression in high salt media are largely attributable to depletion or sequestration of DMAPP. This molecule feeds into a number of other critical pathways, including the biosynthesis of quinones within the electron transport system (90). Interestingly, the growth defects caused by MiaA overexpression in the presence of MV were not linked with DMAPP availability and were only partially countered if the DMAPP binding and catalytic activities of MiaA were ablated. In these assays, the overexpression of DMAPP-binding and catalytically inactive MiaA mutants may still impair ExPEC growth in the presence of MV due to the ability of MiaA to form multimers (28). It is feasible that the formation of mixed wildtype and mutant MiaA complexes could in effect take the wild-type enzyme out of action, partially mimicking the phenotype of a miaA deletion mutant. Overall, these observations indicate that MiaA overexpression can interfere with bacterial growth via distinct mechanisms, dependent upon the types of environmental stressors that are present.
MiaA is part of a complex superoperon and is regulated directly and indirectly by a variety of factors, including RNaseE, Hfq, RpoS and the heat shock sigma factor RpoH (106)(107)(108). However, the control of MiaA expression, and the regulation of tRNA modifying enzymes in general, is not clearly understood (10,109). Though we did not investigate MiaA regulation in detail here, our RT-qPCR experiments indicate that MiaA levels are reduced in response to high salt stress via a post-transcriptional mechanism (Figure 4A-D). Interestingly, miaA has a higher-than-average UNN Leu codon usage ratio of 0.46, suggesting that MiaA may help regulate the translation of its own transcripts (23). Furthermore, we note that MiaA levels are also reduced in response to oxidative stress caused by MV, but intensified in the presence of the metal chelator EDTA ( Figure 4G). The latter observation suggests that MiaA quantities can be controlled by one or more EDTA-sensitive metalloprotease, as proposed for the tRNA modifier GidA (81). Additional work is needed to decipher how proteolytic degradation and other post-transcriptional processes might modulate MiaA levels in response to specific stressors like MV and high salt.
By adjusting the levels of tRNA modifying enzymes like MiaA, ExPEC and other organisms may be able to vary the diversity of translated proteins and thereby optimize adaptive responses to stressful stimuli (9)(10)(11)(12)81,(110)(111)(112). Indeed, the overexpression and deletion of miaA led to the generation of distinct proteomes by UTI89 (Figure 7) and compromised the ability of this ExPEC strain to deal with multiple stressors (Figure 2, 3 and 5). Because tRNA modifications can have pleiotropic effects, it is not always easy to distinguish the direct and indirect effects that tRNA modifying enzymes like MiaA have on translation (12). For example, pioneering work in K-12 E. coli indicates that the efficient translation of RpoS and Hfq relies on MiaA for proper decoding of UNN-Leu codons (7,22,23), but these factors can themselves regulate the expression of numerous other proteins, including MiaA itself (106,(113)(114)(115). The ca-pacity for MiaA to have additional, indirect effects on the fidelity and specificity of translation is further highlighted by our proteomics data showing that MiaA impacts the expression of multiple ribosome-and RNA-associated factors, tRNA ligases, and the RNA editing enzyme YfhC (TadA). These findings suggest the existence of a complex network of RNA and translational modifiers that can regulate the expression of one another. Layered on top of this is the potential for MiaA to affect the biosynthesis and availability of specific metabolites used by other tRNA modifying enzymes (10,98).
Increases in frameshifting due to changing levels of MiaA may also allow for more error-prone translation and the subsequent diversification of expressed proteins, which could allow bacteria to better deal with stressful stimuli. The ability of cells to actively regulate frameshifting and other translational errors in order to generate mutant proteins that deviate from those encoded by the genome is gaining appreciation as an adaptive response to stress (116)(117)(118)(119). Ongoing studies aim to utilize ribosomal profiling along with RNA-seq and proteomics to determine if offframe and mutant proteins are being produced by ExPEC via translational modifiers like MiaA in response to stressful stimuli. Similar lines of research may also shed light on the somewhat more cryptic functions of the MiaB-catalyzed tRNA modification.
In the absence of miaA, the i 6 A and ms 2 i 6 A modifications are not detectable ( Figure 4E and F), as expected from previously published work (109). When MiaA is overproduced, high levels of i 6 A are observed, while ms 2 i 6 A modification levels remain relatively stable ( Figure 5A). Interestingly, high salt stress reduces both MiaA expression and ms 2 i 6 A levels but does not significantly affect i 6 A levels ( Figure 4E and F). In contrast, the disruption of miaB prevents the formation of ms 2 i 6 A and causes marked increases in i 6 A levels (see Supplementary Figure S6), but this had no phenotypic effect in any of our assays. These findings present a conundrum--why does high-level production of i 6 A due to MiaA overexpression attenuate the stress resistance of UTI89 while even higher levels of i 6 A that build up in the absence of MiaB had no overt phenotypic effects in our assays? In considering this issue, it should be noted that we quantified normalized amounts of i 6 A and ms 2 i 6 A, and not specific tRNAs, leaving open the possibility that changing levels of MiaA differentially affect distinct subsets of cognate or potentially non-cognate tRNAs. Either possibility could help account for the contrasting phenotypic effects linked with elevated i 6 A levels due to MiaA overexpression versus those caused by miaB deletion. However, our data indicate that during high salt stress, the growth defects associated with MiaA overexpression are mostly due to disruption of DMAPP pools, as noted above. The deletion of miaB, by causing a buildup of i 6 A rather than accelerated production of this modified residue, may have less of an abrupt impact on the availability of substrates like DMAPP. Finally, it is feasible that MiaA has moonlighting function(s), affecting non-tRNA targets and compromising bacterial fitness when produced in excess.
Cumulatively, the findings presented here, and summarized in Figure 8D, highlight the central and complex roles that core metabolic genes like miaA can have on the elaboration and fine-tuning of pathogen stress resistance and virulence-associated phenotypes. By varying levels of MiaA, our results show that ExPEC can markedly alter frameshifting frequencies and the spectrum of expressed proteins, directed in part by the prevalence of UNN codons. MiaA-mediated changes in the proteome can in turn modify bacterial stress resistance, but varying levels of MiaA may also influence the availability of key metabolites like DMAPP that can have broad-ranging effects on multiple pathways. The MiaA-dependent phenotypes observed in this study with ExPEC parallel some of those associated with MiaA homologues in eukaryotes. For example, both MiaA and the human homologue TRIT1 can modulate the expression of electron transport chain complex subunits (120). Mutations in TRIT1 are associated with mitochondrial disease as well as the progression of some cancers (120)(121)(122). Only a limited number of tRNA substrates for TRIT1 have been identified in human cells (123). However, if TRIT1 acts akin to MiaA in ExPEC, it may also be able to directly and indirectly impact multiple regulatory factors and metabolites, either enhancing cellular health when TRIT1 activities are properly tuned or promoting disease when out of balance.

DATA AVAILABILITY
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (124) with the dataset identifier PXD024782.