tRNA hydroxylation is an epitranscriptomic modulator of metabolic states affecting Pseudomonas aeruginosa pathogenicity

Abstract

Post-transcriptional modification of transfer RNAs (tRNAs) represents an essential layer of translational regulation critical for bacterial adaptation to environmental changes. Increasing evidence links the tRNA epitranscriptome to pivotal roles in the regulation of gene expression and various cellular processes, including stress responses and establishment of virulence. In this study, we used mass spectrometry and nanopore sequencing to quantify and identify the sites of TrhPO-dependent tRNA hydroxylation in total and purified Pseudomonas aeruginosa tRNAs. Furthermore, transcriptome, ribosome profiling, and proteome data were integrated to demonstrate the post-transcriptional consequences of the absence of xo⁵U₃₄ modifications at the wobble position of selected tRNAs. We suggest that the impaired ability to infect host cells and attenuated virulence in Galleria mellonella are driven by changes in metabolic fluxes. In the absence of TrhPO-mediated tRNA modification, chorismate, the precursor for the biosynthesis of xo⁵U modifications, is funneled into alternative pathways, including the production of aromatic amino acids and phenazines. Our findings that metabolic rerouting, rather than changes in proteome profiles, attenuates P. aeruginosa virulence highlight the multifunctional roles of tRNA-modifying enzymes and suggest an underexplored role for these enzymes in monitoring and modulating metabolic fitness. These insights open new avenues for combatting the pathogenicity of this challenging opportunistic pathogen.

Graphical Abstract

Graphical Abstract.

Introduction

Transfer RNAs (tRNAs) are essential for decoding messenger RNA (mRNA) codons into proteins and thus play a critical role in translation. To ensure accurate decoding, tRNAs exhibit a diverse array of chemical modifications, introduced post-transcriptionally through well-orchestrated tRNA-modification pathways [1, 2]. These chemical features—collectively referred to as the tRNA epitranscriptome—vary in complexity and provide structural stability or influence the decoding capacity, thus improving the accuracy and fidelity of translation. tRNA modifications in the anticodon region, particularly at the wobble-position 34 or adjacent sites, can expand or limit wobble capacity beyond standard decoding rules [3, 4]. Recent studies underscore the pivotal role of the tRNA epitranscriptome in translation, protein homeostasis, and cellular fitness in both bacteria and eukaryotes [5–9]. Furthermore, dynamic reprogramming of tRNA modifications in response to environmental cues allows for modulated translation of mRNAs enriched with modification-sensitive codons, referred to as modification-tunable transcripts (MoTTs) [10–12].

In bacteria, tRNA modification systems have been demonstrated to influence virulence phenotypes, making them promising targets for novel therapeutic approaches [13–15]. Our recent work identified the role of the GidA-MmnE pathway in generating xm⁵(s²)U wobble uridine modifications in selected tRNAs, revealing how translation at the rare codons AGA and UUA in core genes drives virulence in the opportunistic pathogen Pseudomonas aeruginosa [16].

Another group of wobble U₃₄ tRNA modifications in gram-negative bacteria are hydroxylated derivatives (xo⁵U), crucial for efficient decoding of NYN family box codons. Recent studies have shown that xo⁵U modifications in tRNA anticodons facilitate non-Watson–Crick wobble-base pairing with guanidine (G) and pyrimidines (C, U) at the third position of mRNA codons, thus allowing single xo⁵U-modified tRNAs to decode multiple codons [17–21]. The biosynthetic pathway for U₃₄ hydroxylation of a subset of six tRNAs has recently been elucidated in Escherichia coli [22, 23] (Fig. 1A). The first step, the formation of 5-hydroxyuridine (ho⁵U), involves either prephenate-dependent ho⁵U synthesis, mediated by TrhP, or oxygen-dependent ho⁵U formation, catalyzed by TrhO. CmoB then carboxymethylates ho⁵U34 to form uridine 5-oxyacetic acid (cmo⁵U) using S-carboxymethyl-S-adenosyl-l-homocysteine (SCM-SAH), synthesized from S-adenosyl-methionine (SAM) and prephenate by CmoA. Further methylation by CmoM leads to the formation of uridine 5-oxyacetic acid methyl ester (mcmo⁵U) [24]. In E. coli, TrmL-dependent methylation of the ribose moiety of mcmo⁵U in tRNA^SerUGA results in the formation of 2′-O-methyluridine 5-oxyacetic acid methyl ester (mcmo⁵Um).

Figure 1.

Identification of enzymes involved in xo⁵U biosynthesis in P. aeruginosa tRNAs. (A) Pathway of xo⁵U biogenesis in E. coli. Both TrhP and TrhO hydroxylate the uridine at position 34 of tRNAs decoding NYN codons. TrhP-dependent ho⁵U34 formation depends on prephenate, while TrhO utilizes molecular oxygen as a donor for the hydroxyl group. CmoB requires SCM-SAH as a cofactor for carboxymethylation of ho⁵U, yielding cmo⁵U. CmoA, utilizing prephenate and SAM as precursors, mediates the formation of SCM-SAH. CmoM catalyzes methylation of cmo⁵U to mcmo⁵U. In E. coli, tRNA^SerUGA is further methylated by TrmL to yield mcmo⁵Um. tRNAs harboring the respective modification in E. coli are indicated. (B) Genomic organization of genes involved in xo⁵U biogenesis in E. coli K12 and identified homologs in P. aeruginosa PA14. Conserved functional amino acids are indicated according to their occurrence in the protein sequence. (C) Workflow for the purification of total tRNA from P. aeruginosa cultures, their hydrolysis into ribonucleosides, and quantification of modified ribonucleosides by targeted MRM LC-MS/MS analysis. (D) LC-MS/MS analysis of wobble base modifications ho⁵U, cmo⁵U, mcmo⁵U, and mcmo⁵Um in P. aeruginosa total tRNA. Left: extracted ion chromatograms (XIC) of calibration standards for each modification. Right: XICs of identified modifications in total tRNA extracted from PA14 WT and mutant strains lacking the depicted genes required for xo⁵U biogenesis (n = 4 for PA14 WT, n = 3 for mutant strains). The peak area of the detected modification was normalized to PA14 WT levels.

Emerging evidence suggests that xo⁵U modifications play a critical role in bacterial stress response. For example, elevated levels of cmo⁵U in tRNA^ThrUGU during early hypoxia of Mycobacterium bovis BCG were demonstrated to coincide with the translation of transcripts enriched in corresponding ACG codons, including DosR, a master regulator of hypoxic bacteriostasis [12].

Pseudomonas aeruginosais an opportunistic human pathogen, known for its intrinsic and acquired mechanisms of antimicrobial resistance. It thrives in diverse environments and is a leading cause of hospital-acquired infections, particularly in immunocompromised patients [25–29]. The exploration of the tRNA epitranscriptomic landscape in P. aeruginosa remains incomplete [30–32], but a deeper understanding promises to unravel novel molecular mechanisms that shape bacterial behavior and adaptation to host environments.

In this study, we identified the genes responsible for xo⁵U tRNA modifications in selected P. aeruginosa tRNAs and characterized the epitranscriptomic landscape using LC-MS/MS and nano-tRNAseq approaches. Furthermore, we applied a multi-omics approach on a P. aeruginosa mutant strain lacking xo⁵U modifications and described the molecular basis for hypomodification-induced phenotypic changes.

Materials and methods

Bacterial strains, plasmids, and growth conditions

Bacterial strains, plasmids, and primers used in this study are listed in Supplementary Table S1. Bacterial liquid cultures were cultivated in Luria-Bertani (LB) medium or modified M9 minimal medium supplemented with 20 mM glucose. Bacterial pre-cultures were inoculated from LB agar plates and incubated at 37°C and 180 rpm overnight in 3 ml liquid LB medium. Bacterial main cultures were inoculated at a starting optical density at 600 nm (OD₆₀₀) of 0.05. Plasmids were maintained using respective antibiotics. Gentamicin was used at a final concentration of 50 μg/ml and streptomycin, and kanamycin were used at a final concentration of 500 μg/ml.

Identification of homologs

For the identification of homologs in P. aeruginosa PA14, corresponding protein sequences of described E. coli enzymes were retrieved from the UniProt database for E. coli [33] and served as queries in Basic Local Alignment Search Tool (BLAST). The BLASTp algorithm was employed with the default parameters against P. aeruginosa protein sequences available in the Pseudomonas.com database [34]. The selection of homologous enzymes was based on the following criteria: significant sequence similarities determined by E-value (≤1^e−5) by BLASTp; conservation of functional domains or motifs crucial for enzymatic activity based on predictions by the BioCyc database [35]; and high sequence coverage ensuring a comprehensive alignment between the query and the subject sequences. The putative homologous enzymes identified through bioinformatics analyses were further validated through experimental assays (see below).

CRISPR-Cas9 engineering of mutant strains

To generate clean knockout strains of genes of interest, a CRISPR-Cas9-based recombination protocol previously developed in our laboratory was used [36]. Oligonucleotides used for deletions are listed in Supplementary Table S2.

Extraction of total RNA

Isolation of total RNA from P. aeruginosa strains was conducted with phenol in acidic conditions as previously described [37]. Briefly, bacterial liquid cultures (50 ml) were centrifuged, and pellets were dissolved in resuspension buffer, followed by two extractions with equal volumes of cold acidic phenol. Total RNA was recovered, and the pellet was resolved in 100 μl sodium acetate (NaAC), followed by precipitation at −20°C overnight. Total RNA was recovered by centrifugation, and pellets were washed twice with ethanol (70% v/v). Pellets were dried and resuspended in RNase-free H₂O. RNA concentration was determined by a NanoDrop spectrophotometer (DeNovix DS-11 FX+).

Extraction of total tRNA

Total tRNA was isolated from total RNA by solid-phase chromatography using anion exchange columns with a quaternary ammonium-modified silica membrane (Chromabond SB 1 ml, Macherey-Nagel) as described previously [32]. Total tRNA was precipitated at −20°C overnight. Total tRNA was recovered by centrifugation and washed three times with ethanol (70% v/v). The pellets were dried and suspended in RNase-free H₂O. The concentration of total tRNA was determined using a BioTek Synergy H1 plate reader with a Take3 Microvolume Plate (Agilent), and RNA integrity was checked using the Agilent Bioanalyzer (RNA 6000 Nano Kit; Agilent).

Purification of individual tRNAs

Individual tRNAs were purified from total tRNA pools using a magnetic bead-based selective hybridization method with highly specific DNA oligonucleotides complementary to variable regions of the target tRNA sequence (Supplementary Table S2) [16, 38]. RNA concentration was determined using a BioTek Synergy H1 plate reader with a Take3 Microvolume Plate (Agilent). RNA integrity was checked using the Agilent Bioanalyzer (small RNA Kit; Agilent).

Liquid chromatography-mass spectrometry analysis of ribonucleosides

Total tRNA as well as individual tRNAs were enzymatically hydrolyzed prior to the LC-MS/MS analysis as described previously [32]. Tenofovir [100 ng/ml] was used as an internal standard. Nucleosides were analyzed by reverse-phase chromatography coupled to a Linear Ion Trap Quadrupole mass spectrometer (MS) (QTRAP5500, Sciex). The MS was operated in multiple reaction monitoring (MRM) mode. External standard calibration with synthetic reference standards for each analyte was used to quantify modified ribonucleosides. Commercially available synthetic reference standards were purchased or synthesized. The concentrations of the external standard mixtures ranged from 13.4 pM up to 5 μM with 2.5-fold serial dilutions. Raw data were analyzed using the Analyst 1.6 software (Sciex) with the following parameters: smoothing width three points, noise percent 90%, and peak-splitting factor 2. For the calibration curve, quadratic regression with 1/× weighting and accuracy with 100 ± 20% was determined to detect the linear range. Based on the calibration curve of each reference standard, the lowest limit of quantification and upper limit of quantification were determined. The modification amount was calculated by the ratio of the peak area from the modified ribonucleoside and the peak area from the IS.

Metabolomics analysis

Bacterial liquid cultures were harvested at an OD₆₀₀ of 2. Biomass was extracted with 250 μl methanol supplemented with 1% U-13C-ribitol (0.2 mg/ml) for 15 min in an ultrasonic bath. Subsequently, 250 μl of water was added, and the samples were vigorously mixed. After addition of 500 μl dichloromethane, the samples were mixed and centrifuged (10 000 × g, 5 min). The polar phase was collected and dried under vacuum. Metabolite analysis was performed on an Agilent GC-MSD system (7890B coupled to a 5977 GC) equipped with a high-efficiency source and a PAL RTC system according to [39]. A two-step derivatization with a methoxyamine hydrochloride solution (20 mg/ml in pyridine) and N-methyl-N-(trimethylsilyl)-trifluoracetamide was automatically performed with the PAL RTC system. One microliter of the sample was injected into a multimode inlet in both split mode (split ratio of 10:1, split flow of 12 ml/min) and pulsed splitless mode (30 psi until 0.75 min, 50 ml/min at 1 min). Separation was conducted on an Agilent VF-5ms column with a helium flow of 1.2 ml/min. The oven temperature was hold at 70°C for 6 min and then linearly increased with 6°C/min up to 325°C. Ions were detected in scan mode from 70 to 700 m/z with 2.3 scans/s. Data analysis of intracellular metabolites was performed as previously described [40, 41].

Nano-tRNAseq library preparation and sequencing

Nano-tRNAseq of P. aeruginosa tRNA samples was performed as recently described [42]. Briefly, for the nano-tRNAseq library preparation, tRNA samples were deacetylated and ligated to the pre-annealed 5′ and 3′ splint adapters. Next, 5′ and 3′ ligated tRNAs were ligated to the pre-annealed RTA adapters. Afterwards, a reverse transcription master mix was added directly to the ligation reaction before ONT RMX sequencing adapters (Direct RNA Sequencing Kit, SQK-RNA002, Oxford Nanopore Technologies) were ligated to the tRNAs. MinION flow cells (FLO-MIN-106) underwent quality control, priming, and loading following the standard ONT SQK-RNA002 protocols.

Nano-tRNAseq data processing and analysis

Sequencing runs were performed for 24 h without live basecalling, and bulk output files were obtained for each run. Fast5 data were basecalled in high-accuracy mode without trimming using Guppy v6.5.7 (https://community.nanoporetech.com/downloads) and aligned against a non-redundant P. aeruginosa PA14 tRNA database [34] with BWA mem v0.7.17 [43] using the nano-tRNAseq parameters recommended in [42]. SAMtools v1.15 [44] was used in downstream processing to retain only primary alignments with a mapQ score >0 in the forward orientation [samtools view -F 2324 -q 1] and to sort and index data files. Abundance counts for each tRNA were generated using [samtools view infile.bam | cut -f3 | grep -v LN | sort | uniq -c | sed “s/^[ \t]*//” | sed “s/ / \t/g”]. Scatter plots were generated using the ggplot2 package [45] in R studio [46]. Differential expression analysis was performed using the DESeq2 package [47].

For the generation of modification-induced mismatch profiles, for each position in each tRNA, the number reads supporting A, C, G, U, or N (representing insertions/deletions) was determined using bamreadcount (https://github.com/genome/bam-readcount) [-w 5 -d 10 000 000] and parsed into a readable format with a custom python script (variant_caller_v1.3.py). Output files were imported into R studio and processed with a custom script (barplot_trna_name_plotting.R) making use of the R packages data.table, ggplot2, dplyr, tidyr, and patchwork. This script was used to generate barplots showing the frequency represented as a fraction between 0 and 1 of A, C, G, U, and N at each position within the tRNA of interest. Sequencing results were compared and aligned to a non-redundant set of reference tRNA sequences for P. aeruginosa PA14 obtained from the genomic tRNA database.

RNA-sequencing and transcriptome analysis

RNA extraction for RNAseq was performed using the RNeasy Mini Kit (Qiagen) in combination with Qiashredder™ columns as described previously [48]. Briefly, bacterial main cultures (10 ml) were inoculated with a start OD₆₀₀ of 0.05 and harvested in exponential phase at an OD₆₀₀ of 0.5 and directly mixed with an equal volume of RNA protect (Qiagen). RNA was extracted according to the manufacturer’s instruction with slight modifications. DNA removal was performed using the DNA-free™ Kit (Thermo Fisher Scientific, Waltham, MA, USA). RNA integrity was checked using the Agilent Bioanalzyer (RNA 6000 Nano Kit; Agilent). Ribosomal RNA was removed using the Ribo-Zero Bacteria Kit (Illumina) and sequencing of the samples was performed in paired-end mode (2 × 50 bp reads) on an Illumina NovaSeq 6000 device. Sequencing reads were mapped to the PA14 reference genome using bowtie2 [49], and the number of reads per gene was assessed with FeatureCounts [50]. Differential gene expression analysis between the PA14 WT and mutant strains was performed with the edgeR package [51] using the function glmTreat (fold-change 1.2). Genes were filtered using edgeR function filterByExpr, and reads were normalized using the edgeR function calcNormFactors (trimmed mean of M values). Adjusted P-values were calculated for multiple tests using the Benjamini–Hochberg adjustment. The significance threshold was set to a false discovery rate (FDR) of ≤ .05 to identify differentially expressed genes. Functional enrichment analysis of Gene Ontology (GO) terms [52] was carried out using the hypergeometric test [R function phyper, adjusted P-value (FDR) < .05].

Ribosome profiling

Ribosome profiling was performed as described previously [16]. Size-selected RNA was converted to complementary DNA libraries using a NEBNext Multiplex small RNA library prep for Illumina (E7300) using custom multiplex primers. Sequencing was performed on an Illumina NovaSeq 6000 platform as a 50 bp paired-end sequencing at the Genome Analytics Research Group at the Helmholtz Centre for Infection Research (HZI) of Braunschweig.

The sequencing depth of the ribosomal footprint of PA14 WT and ΔtrhPO was between 20.9 and 26.6 million reads. For gene-level comparison, data were processed the same way as transcriptome data. Differences in ribosome occupancy between the PA14 WT and ΔtrhPO mutant were calculated using edgeR comparing (WT_RBF - ΔtrhPO _TR) - (WT_RBF - ΔtrhPO _TR). Reads per gene were filtered and normalized in the same way as the transcriptome data (filterByExpr, calcNormNactors). The replicates were analyzed using the edgeR function plotMDS, based on which one ΔtrhPO replicate was excluded from the analysis as a strong outlier. For codon enrichment analysis, genes were ordered by their adjusted P-value and the top 25% genes with the highest and the lowest ribosome occupancy (log₂ fold change higher/lower 0) in ΔtrhPO were selected. Enrichment analysis was performed for each codon using the R function phyper and information about the presence/absence of codons within each gene. The codon enrichment score (fold enrichment of the number of genes bearing a codon compared to the number of genes expected by chance) was calculated using the formula Cs/Cb, where Cs is the proportion of these gene-subsets containing a codon and Cb is the codon proportion in the total number of genes used in the edgeR analysis. Data were plotted using the R library ggplot2.

Proteome analysis

Bacterial cells were harvested at the same conditions as for transcriptome and RiboSeq analysis. Cell pellets were resuspended in a sodium dodecyl sulfate lysis. Mild sonification was used for cell lysis. Protein concentration was quantified using the Pierce Modified Lowry Protein Assay Kit (Thermo Fischer Scientific). Proteins were extracted by phenol-chloroform extraction. The precipitated proteins were dissolved in a digestion buffer, and samples were acidified to pH 2.3 with formic acid, followed by centrifugation. The supernatant was transferred to a new tube, and the supernatant was evaporated at 30°C in a Speed-Vac®. Peptides were desalted using the OASIS elution plate (Waters), with elution performed by using 60% acetonitrile and 0.1% formic acid. Peptide concentration was determined using Pierce Quantitative Fluorometric Peptide Assay (Thermo Fischer Scientific). LC-MS analyses were conducted utilizing an Exploris 240 orbitrap mass spectrometer (Thermo Fischer Scientific) with a resolution set to 60.000 and a scan range of 350 to 1900 m/z. The samples were separated before measurement using a Thermo Scientific COL-nano050G1B column (PharmaFluidics) with a gradient of acetonitrile from 2.5% up to 95%. Raw files were processed using MaxQuant (Version 2.4.10.0) [53]. Protein searches were executed by employing the integrated Andromeda search engine [54], at a 1% FDR. Further data processing and statistical analyses were performed using R studio involving log₂ transformation of peptide values. Data were filtered for unique peptides, contaminants were removed, and a two-sample t-test was conducted on the mean log₂ values between conditions.

Fluorescence reporter assays and flow-cytometry

Oligomers used for cloning (Supplementary Table S2) into pCDN2 reporter plasmids were ordered from Eurofins Genomics. Sequences were verified by Sanger sequencing (Microsynth). P. aeruginosa strains carrying pCDN2 derivatives were plated on LB agar supplemented with 500 μg/ml kanamycin. Isolated colonies were picked and grown overnight in 3 ml LB + 500 μg/ml kanamycin. Aliquots of 1 ml from each subculture were washed with phosphate buffered saline (PBS) and grown in 10 ml LB without antibiotic in 50 ml Erlenmeyer flask. To avoid carry-over of fluorescent proteins from pre-cultures, the initial OD₆₀₀ was adjusted to 0.005. Samples were taken in early exponential phase (OD₆₀₀ 0.5), appropriately diluted in filtered PBS (0.22 nm pore size) containing 30 μg/ml chloramphenicol to ∼10⁶ cells/ml, and measured on a BD FACS Fortessa flow cytometry system. Measurements were taken with a 561-nm laser excitation source/610-nm (20-nm bandpass filter) fluorescence acquisition channel in the case of mCherry fluorescence and an excitation/610-nm (20-nm bandpass filter) emission channel for msfGFP. Calibration of the experiments was performed as follows: forward and side scatter density plots were used to identify the bacterial cell population and to exclude debris. Moreover, detection of the fluorescent populations was calibrated using non-fluorescent PA14 WT. All flow cytometry data were processed using FlowJo, excluding the cells that did not produce mCherry signal. The msfGFP and mCherry signals were calculated using population mean fluorescence intensities and used to calculate the msfGFP/mCherry fluorescence ratio.

Competition assays

Bacterial strains were transformed with pSEVA2313 plasmids harboring monomeric superfolder green fluorescent protein (msGFP) or mScarlet. Main cultures were inoculated to a start OD₆₀₀ of 0.05 and grown at 37°C with shaking at 180 rpm. At indicated time points, cultures were diluted to 10⁻⁴–10⁻⁷ and spread on agar plates with appropriate antibiotics. Colony forming units (CFUs) expressing the different fluorescent proteins were counted under blue light.

Growth kinetics analysis

Bacterial overnight cultures were diluted to a start OD₆₀₀ of 0.05 in fresh LB or M9 medium. Growth kinetics of all strains were monitored in 96-well plate format using a BioTek Synergy H1 plate reader (Agilent) by measuring the OD₆₀₀ in a 30 min interval at 37°C with orbital shaking (180 rpm) for a total time period of 24–48 h. Growth metrics were calculated using the R QurvE package [55].

Quantification of pyocyanin

Pyocyanin quantification was performed as described previously by chloroform extraction [56]. Samples for pyocyanin extraction were collected after 8 h. Pyocyanin quantification was performed by measuring the absorption at 520 nm in 96-well format in a BioTek Synergy H1 plate reader (Agilent). Pyocyanin concentration in μg/ml was determined by multiplication of the absorption with the pyocyanin-specific extinction factor 17.072. For each time point, the OD₆₀₀ was determined and used for normalization.

Quantification of pyoverdine

Quantification of the siderophore pyoverdine was performed as described previously [57]. Briefly, bacterial pre-cultures were cultivated in metal-scarce casamino acid (CAA) medium. Bacterial main cultures were inoculated to a start OD₆₀₀ of 0.05 in 10 ml CAA medium. The cultures were grown for 24 h at 37°C at 180 rpm to induce the production of pyoverdine. Afterwards, the OD₆₀₀ was determined, and 2 ml of the culture was centrifuged. Pyoverdine in the supernatant was measured by UV–visible spectrophotometry at λ = 400 nm in a quartz cuvette and normalized to growth.

Galleria mellonella infection assay

Larvae of the Galleria mellonellawax moth (Fauna Topics, Germany) were visually inspected and sorted before infection experiments. Bacterial cells from overnight cultures of all tested strains were washed in sterile PBS and used to prepare the inoculum in a serial dilution in PBS. For each biological replicate, 10 larvae were infected by injecting 20 μl of the inoculum corresponding to a multiplicity of infection (MOI) of 10 with a 500 μl Hamilton syringe equipped with a 30 G needle into the last proleg of each larvae. Injection of PBS was used as a control. Infected larvae were incubated at 37°C, and survival of the larvae was monitored after 16 h post-infection (p.i.) for a total time period of 72 h. Death was defined as loss of movement and melanization of the cuticle.

In vitro cytotoxicity assay

Briefly, adenocarcinoma human alveolar basal epithelial cells (A549) were maintained in Dulbecco’s modified Eagle medium (DMEM, Gibco, Life Technologies; supplemented with 2 mM L-glutamine, 10% fetal bovine serum). A549 cells were grown in 24-well plates at 37°C and 5% CO₂ to 90% confluence before infection with different bacterial strains. The inoculum was prepared from bacterial cultures in DMEM and adjusted to an MOI of 20. The plates were incubated, and supernatants were collected after 3 h p.i. For the determination of cytotoxicity effects, Pierce lactate dehydrogenase (LDH) assay was performed according to the manufacturer’s instructions (Thermo Scientific). As a positive control, wells without bacteria were treated with lysis buffer 45 min prior to the collection of the supernatants (maximum cytotoxicity). Cytotoxicity for each tested strain was calculated as percentage of maximum cytotoxicity.

Biofilm microscopy

Biofilm characteristics were monitored by confocal laser-scanning microscopy as described previously [58]. Overnight LB cultures were adjusted to an OD₆₀₀ of 0.05 and 100 μl of this suspension were transferred into a well of a microtiter plate (96 well, half-area, Greiner). For each strain, eight replicates were prepared, and the experiment was performed three independent times. The microtiter plate was covered with an air permeable foil and incubated at 37°C for 24 h in a humid atmosphere. LIVE/DEAD staining (BacLight™ Bacterial Viability kit, Molecular Probes/Invitrogen) was applied, and the strains were incubated for an additional 24 h before images including z-stacks were acquired using a confocal microscope (TCS SP8, Leica Microsystems) with a 40×/NA 1.1 water-immersion objective.

Results

Identification of genes responsible for the biosynthesis of xo⁵U derivatives at the wobble position of tRNAs in P. aeruginosa PA14

Previous studies have shown that genes involved in the biosynthesis of xo⁵U tRNA modifications are widespread in γ- and β-proteobacteria [22]. To identify potential homologs responsible for xo⁵U tRNA modification in P. aeruginosa, we used the sequences of E. coli proteins previously implicated in xo⁵U biogenesis as references for BLAST searches against the P. aeruginosa PA14 protein sequence database [34] (Supplementary Fig. S1). This analysis led to the identification of PA14_71820, a putative peptidase U32 family protein homolog of TrhP_Ec (71.5% sequence identity and 82.9% similarity), and PA14_53180, a putative TrhO_Ec homolog, classified as a rhodanese family protein (29.4% sequence identity and 44.7% similarity). The latter contains the characteristic CTGGTR (CXGGXR) motif, including a highly conserved cysteine within a glycine-rich loop in its active site, as described for TrhO_Ec [22]. We also identified PA14_63310 as a putative CmoM_Ec homolog (39.06% sequence identity and 56.6% similarity) containing a distinct class I SAM-dependent methyltransferase domain involved in the terminal methylation of cmo⁵U to mcmo⁵U. Additionally, PA14_54270 was identified as a potential homolog to CmoB_Ec (53.6% sequence identity and 69.6% similarity), with a conserved lysine residue at position 91 essential for Cx-SAM recognition. PA14_54270 is predicted to form an operon with PA14_54260, which we identified as a putative CmoA_Ec homolog (52.6% sequence identity and 68.1% similarity). As in E. coli, these candidate genes are scattered across the PA14 genome (Fig. 1B). The identified homologs were named based on their corresponding E. coli annotations within the xo⁵U biogenesis pathway, as shown in Fig. 1B. A reanalysis of previously recorded transcriptional profiles of PA14, which were collected hourly during its growth cycle [59], revealed that trhO, cmoA, cmoB, and cmoM are co-regulated and exhibit an induced expression during early growth phases, with a slight delay in the expression of trhP (Supplementary Fig. S2).

Dual deletion of trhP and trhO abolishes xo⁵U formation in P. aeruginosa tRNAs

Using a CRISPR-Cas9-assisted recombineering toolkit [36], we deleted individual and combinations of the putative genes involved in xo⁵U biogenesis in the PA14 wild-type (WT) strain and analyzed the functional consequences. We extracted total tRNA from PA14 WT and the respective single or double deletion strains, hydrolyzed the tRNA into ribonucleosides, and applied a targeted LC-MS/MS approach [32] to investigate changes in the levels of ho⁵U, cmo⁵U, mcmo⁵U, and mcmo⁵Um modifications, respectively (Fig. 1C). We used synthetic standards of all derivatives as references for the identification and quantification in our LC-MS/MS approach and normalized the peak areas corresponding to modified ribonucleosides in each mutant strain to PA14 WT levels (Fig. 1D). The relative abundance of ho⁵U was slightly decreased in ΔtrhP and ΔtrhO, whereas ho⁵U significantly accumulated upon cmoB deletion. The levels of cmo⁵U were reduced by ∼40% in ΔtrhP and ∼20% in ΔtrhO, whereas none of the modifications could be detected in a ΔtrhP ΔtrhO double mutant (ΔtrhPO), underscoring the redundancy of the TrhP- and TrhO-dependent hydroxylation pathways [22]. Consistent with previous findings, we could not detect mcmo⁵U modifications in a ΔcmoM mutant, while in contrast to earlier findings in E. coli [24], no mcmo⁵Um modifications were identified in P. aeruginosa tRNAs. From this we conclude that the PA14 TrmL homolog is unlikely to target tRNAs.

Relative distribution and positioning of xo⁵U modifications in individual tRNAs

We further aimed to identify xo⁵U modifications on the level of individual P. aeruginosa tRNA isoacceptors and to map identified modifications onto reference tRNA sequences. We therefore employed a combination of LC-MS/MS-dependent quantification and a nano-tRNAseq approach [42], which enables the detection of modification sites through systematic basecalling errors caused by modified ribonucleosides [42, 60–62]. In E. coli, six tRNAs, tRNA^LeuUAG, tRNA^ProUGG, tRNA^SerUGA, tRNA^ValUAC, tRNA^AlaUGC, and tRNA^ThrUGU, responsible for the decoding of NYN codon family boxes, have been described to be targeted by either TrhP or TrhO-dependent hydroxylation (Fig. 1A). We isolated and purified the respective P. aeruginosa tRNAs, hydrolyzed them into ribonucleosides, and applied our targeted LC-MS/MS approach to determine xo⁵U modification levels (Fig. 2A). The quantities of the different xo⁵U derivatives varied significantly among the examined isoacceptors, suggesting that each tRNA carries a unique set of modifications introduced by the same biosynthetic pathway. We also generated a comprehensive modification profile of a selected PA14 tRNA, tRNA^AlaUGC, and identified 11 different modifications, including the commonly observed ones such as pseudouridine (Ψ), dihydrouridine (D), 7-methylguanosine (m⁷G), and 4-thiouridine (s⁴U), in addition to hydroxylated wobble uridine derivatives cmo⁵U and mcmo⁵U, respectively (Fig. 2B and C). We mapped the identified modifications to specific positions within the reference sequence based on public databases [63] and experimental data from various organisms, including E. coli [64, 65] (Fig. 2D). Next, we performed nano-tRNAseq to systematically investigate basecalling errors caused by the occurrence of modified ribonucleosides in P. aeruginosa tRNAs (Fig. 2E). Therefore, we sequenced two selected individual tRNAs, tRNA^AlaUGC and tRNA^ValUAC, purified from PA14 WT and the ΔtrhPO mutant, and compared the modification-induced mismatch profiles to that of a synthetic, unmodified reference tRNA (Fig. 2F, upper panel and Supplementary Fig. S3). We detected increased mismatch frequencies at positions 31–35 for the PA14 WT tRNAs, which were absent in the hypomodified mutant and the synthetic tRNA variant, indicating the presence of xo⁵-modified uridines at the wobble position. To examine the consistency of these results, we additionally isolated and sequenced total tRNA pools of PA14 WT and the ΔtrhPO mutant, extracted modification-induced mismatch profiles of single tRNAs, and compared the profiles of tRNA^AlaUGC and tRNA^ValUAC to those obtained from individual tRNA sequencing (Fig. 2F, lower panel and Supplementary Fig. S3). In line with our previous observations, we detected consistent profiles featuring characteristic mismatches at xo⁵ modified wobble uridines in tRNA^AlaUGC and tRNA^ValUAC as well as in tRNA^ThrUGU derived from total tRNA sequencing (Supplementary Fig. S4). To further validate these findings, we analyzed the mismatch profiles of tRNA^ArgUCU, a tRNA not targeted by TrhPO-dependent hydroxylation but carrying GidA/MnmE-dependent mnm⁵s²U at the wobble position [16] (Supplementary Fig. S5). Notably, no differences in mismatch frequencies were detected between tRNA^ArgUCU variants from PA14 WT and the ΔtrhPO mutant, highlighting the reliability of nano-tRNAseq for detecting wobble base modifications.

Figure 2.

Relative distribution and positioning of xo⁵U modifications in individual P. aeruginosa tRNAs using LC-MS/MS and nano-tRNAseq analysis. (A) Quantities of ho⁵U, cmo⁵U, and mcmo⁵U modifications in PA14 WT tRNA^ValUAC, tRNA^AlaUGC, tRNA^SerUGA, tRNA^ThrUGU, tRNA^ProUGG, and tRNA^LeuUAG determined by targeted MRM LC-MS/MS with external standard calibration using synthetic reference standards (n = 3). (B) XIC of identified modifications in PA14 WT tRNA^AlaUGC. (C) Quantities of modifications in PA14 tRNA^AlaUGC assessed by targeted MRM LC-MS/MS with external standard calibration (n = 3). Modification levels (modified ribonucleoside [fmol] × number of parent ribonucleoside in tRNA sequence/parent ribonucleoside [fmol] + detected modified ribonucleosides that descends from the respective parent nucleosides [fmol]) for panels (A) and (C) were calculated according to Grobe et al. [32]. (D) Secondary structure of PA14 WT tRNA^AlaUGC with indicated modification sites. (E) Schematic representation of ion current changes depending on signal interferences between canonical and modified nucleotides during nanopore sequencing. (F) Nano-tRNAseq analysis of PA14 tRNA^AlaUGC. Upper panel: Modification-induced mismatch profiles of individually sequenced PA14 WT and ΔtrhPO tRNA^AlaUGC; a synthetic, unmodified tRNA^AlaUGC served as a reference. PA14 WT and ΔtrhPO tRNA^AlaUGC were purified from total tRNA pools by selective hybridization before sequencing. Total read count and number of aligned reads to the reference sequence can be found in Supplementary Fig. S6. Lower panel: Modification-induced mismatch profiles of PA14 WT and ΔtrhPO tRNA^AlaUGC derived from total tRNA pool sequencing. Positions with mismatches to the reference sequence are indicated by colors (green = A, blue = C, orange = G, red = T, and dark gray = N; N equals either insertions or deletions at the respective position). On the right are close-up views from the anticodon region of each sequenced tRNA.

Abolishment of xo⁵U biosynthesis alters virulence and metabolic fluxes

Next, we explored the phenotypic consequences associated with xo⁵U hypomodification (Fig. 3 and Supplementary Fig. S7). We first monitored the growth behavior of the ΔtrhPO mutant lacking xo⁵U modifications compared to PA14 WT. We did not observe a reduction in growth in LB medium or in M9 minimal medium supplemented with 20 mM glucose (Fig. 3A and B). Nevertheless, in a direct competition assay, the ΔtrhPO mutant exhibited a fitness disadvantage in LB medium (Fig. 3C). Next, we infected A549 cells with both strains and determined cytotoxicity after 3 h post-infection (p.i.) using lactate dehydrogenase assays. After 3 h p.i., a significant reduction of cytotoxicity was observed in the double deletion strain compared to the PA14 WT (Fig. 3D). Furthermore, pathogenicity was tested by infecting the larvae of the wax moth G. mellonella, a commonly used model for studying host-pathogen interactions [66]. We monitored survival rates of the infected larvae and noted significantly higher survival of the larvae infected with the ΔtrhPO mutant compared to PA14 WT (PA14 WT ∼10% survival, ΔtrhPO mutant ∼40% survival after 40 h p.i.; Mantel-Cox test P < .0001) (Fig. 3E). We also assessed the effect of trhPO deletion on the ability to form biofilms but did not detect any discernible differences in biovolume when compared to the PA14 WT (Fig. 3F).

Figure 3.

Phenotypic characterization of the ΔtrhPO mutant lacking xo⁵U₃₄ tRNA modifications. (A) Growth curves of PA14 WT and ΔtrhPO in LB medium and M9 medium supplemented with 20 mM glucose. Growth was monitored for 24 h in LB and for 48 h in M9 medium, respectively. Data represent means ± SD of individual experiments (n = 9 for LB and n = 6 for M9 samples). (B) Doubling times (time it takes the cells to double in number) for both conditions. (C) Competition assay for growth: the relative fraction [%] of the population of PA14 WT and ΔtrhPO after co-culturing for 24 h in LB medium is depicted (n = 3). (D) Cytotoxicity of both strains was assessed in vitro using human epithelial A549 cells at 3 h (n = 6) post-infection (p.i.). Cytotoxicity was measured via lactate dehydrogenase assay, with cells infected at an MOI of 20. (E) Kaplan–Meier survival curves of G. mellonella following infection at a MOI of 10 CFU per larvae. Ten larvae per biological replicate were infected (n = 9). Injection of PBS served as a control. Survival of the larvae was monitored for 72 h p.i. Significant differences between the survival rates of larvae infected with PA14 WT or ΔtrhPO were determined by log-rank (Mantel-Cox) test with ****P ≤ .0001. (F) Quantification of biofilm volume formed in LB medium after 48 h growth in a microtiter plate and stained with BacLight™ Viability Kit. Eight wells were quantified by strain, and the experiment was repeated three times. For each replicate, values were normalized to the average biovolume value of the WT strain. (G) Quantification of pyoverdine production in CAA medium after 24 h. (H) Quantification of pyocyanin in LB medium after 8 h. (I) Metabolic pathways linked to the biosynthesis of xo⁵U derivatives, phenazines, and aromatic amino acids with chorismate as a central precursor. Chemical structures of relevant metabolites and responsible genes catalyzing each step are depicted. Dashed lines represent pathways leading to metabolites with increased production in the ΔtrhPO mutant. (J) Quantification of shikimate, shikimate-3-phosphate, chorismate and (K) aromatic amino acids L-tryptophan, L-phenylalanine and L-tyrosine and amino acids not linked to chorismate metabolism using GC-MS analysis. Data for (C), (D), (G), (H), (J), and (K) represent means ± SD of at least three biological replicates. Statistical significance for (B), (D), (F), (G), (H), (J), and (K) was determined by two-tailed Student’s t-test (unpaired) with ns = not significant, *P ≤ .05, **P ≤ .01, ***P ≤ .001, ****P ≤ .0001.

Additionally, we quantified the levels of pyoverdine, a siderophore required for iron acquisition and pathogenesis in P. aeruginosa [67] and observed a severe reduction of pyoverdine levels in the ΔtrhPO mutant (Fig. 3G). Concomitantly, increased production of the blue-colored virulence factor pyocyanin was detected after 8 h of growth (Fig. 3H). Both, the trhPO-dependent tRNA modification pathway, as well as the siderophore and phenazine biosynthesis, rely on chorismate as a central precursor (Fig. 3I). Chorismate, synthesized from shikimic acid via the aro pathway, serves as a key precursor for downstream metabolic pathways that also include the biosynthesis of the Pseudomonas quinolone signal and aromatic amino acids. We quantified the levels of selected metabolites related to shikimate and chorismate metabolism, as well as the levels of aromatic amino acids as direct downstream products of chorismate using GC-MS analysis with synthetic reference standards. Consistent with the increased pyocyanin production, the ΔtrhPO mutant exhibited elevated levels of shikimate, shikimate-3-phosphate, chorismate, and aromatic amino acids (Fig. 3J and K). In contrast, the biosynthesis of nonaromatic amino acids remained largely unaffected by disruption of the tRNA hydroxylation pathway. Collectively, these findings indicate that xo⁵U hypomodification results in reduced virulence and metabolic imbalances due to altered metabolic fluxes within the chorismate pathway. Although individual gene deletions within the tRNA hydroxylation pathway did not result in pronounced defects in virulence or pyocyanin production, they were nonetheless associated with metabolic shifts affecting shikimate, chorismate, and aromatic amino acid biosynthesis (Supplementary Fig. S7). Notably, a ΔtrhP ΔcmoB double mutant phenocopied the ΔtrhPO strain with respect to increased pyocyanin production (Supplemental Fig. S8).

xo⁵U-dependent modulation of translation efficiency at single codon level

Since xo⁵U modifications have been shown to enhance base pairing with pyrimidine- and G-ending codons in four codon boxes [17–19, 21, 22] (Fig. 4A), we hypothesized that the hypomodified tRNAs in the ΔtrhPO mutant exhibit compromised decoding capabilities. To ensure that disruption of xo⁵U biogenesis does not impact the composition of the tRNA pool, we quantified tRNA levels from the ΔtrhPO mutant as compared to the PA14 WT reference using nano-tRNAseq (Fig. 4B). Low variability was observed between biological replicates (Supplementary Fig. S9), and the tRNA levels correlated well between the ΔtrhPO mutant and the PA14 WT (r²= 0.973), indicating that xo⁵U hypomodification does not impact the abundance of individual tRNAs.

Figure 4.

Decreased translation efficiency at individual codons in the absence of xo⁵U tRNA modifications. (A) Schematic representation of the expanded decoding capabilities using the example of the xo⁵U-modified tRNA^LeuUAG. (B) Scatter plot showing the abundance of tRNAs in ΔtrhPO compared to PA14 WT (Pearson correlation r²= 0.973, n = 2) quantified by nano-tRNAseq. Read counts were normalized by dividing the abundance count for each tRNA by the summed abundance count for all tRNAs and multiplied by a million (depicted as transcripts per million [TPM]). (C) MDS plot showing differences between RiboSeq and RNAseq datasets for ΔtrhPO and PA14 WT. Symbols represent the respective method. (D) Codon enrichment analysis (codon presence versus absence) in transcript groups categorized by higher or lower translation efficiency (ratio of ribosome footprints to mRNA) in ΔtrhPO compared to PA14 WT. Translation efficiency was calculated for each transcript in both strains, and the top 25% genes exhibiting a log₂FC higher in ΔtrhPO versus PA14 WT or lower are depicted. The codon enrichment factor (y-axis) for each of the two gene groups was determined by calculating the ratio of transcripts containing a specific codon to those without that codon. Significance was determined using hypergeometric testing; codons with a codon enrichment score <1.01 are not shown. Asterisks indicate Benjamini–Hochberg adjusted P values with *P ≤ .05, **P ≤ .01, ***P ≤ .001. Compared to other codons, xo⁵U-dependent codons exhibited a higher mean difference (codon enrichment factor of genes with high versus low translation efficiency, P= .065, t-test, two-tailed). Codons decoded by xo⁵U-modified tRNAs are underlined.

Next, we used ribosome profiling (RiboSeq) [16] to identify the positions of actively translating ribosomes by performing deep sequencing of ribosome-protected mRNA footprints [68]. Multidimensional scaling analysis revealed low variability between biological replicates and distinct separation between RiboSeq and RNAseq profiles in both strains, indicating differences between transcriptional and translational dynamics (Fig. 4C). We assessed the translation efficiency, defined as the ratio of ribosomal footprints to mRNA abundance (ribosome density), for each transcript in both strains by calculating a codon enrichment factor (Fig. 4D) [16]. Although transcripts containing the three xo⁵U-dependent codons Val GUU, GUA, and Ala GCU exhibited a higher translation efficiency in the ΔtrhPO mutant, transcripts enriched in 11 xo⁵U-dependent codons, including Leu CUA, CUC, CUU, and CUG; Pro CCC and CCA; Ser UCA and UCU; and Thr ACA, ACG, and ACU displayed a decreased translation efficiency. While this trend was not statistically significant (P = .065, two-tailed t-test), our findings indicate a role of xo⁵U tRNA modifications in efficient recognition of NYN codons ending in all four canonical bases, particularly in Leu, Pro, Ser, and Thr codon boxes. We also found various xo⁵U-independent codons that exhibited increased or decreased translation efficiency in the absence of the tRNA hydroxylation pathway.

xo⁵U hypomodification induces expression of the mexCD-oprJ efflux pump gene cluster

To assess the global impact of xo⁵U hypomodification on protein homeostasis, we performed label-free quantitative proteomics of PA14 WT and the ΔtrhPO mutant. This approach led to the identification of 1743 proteins in total, with 51 showing differential regulation upon trhPO deletion (24 upregulated and 27 downregulated; FDR ≤ 0.05, log₂FC ≤ −0.5, and ≥ 0.5) (Fig. 5A). Most notably, among the upregulated proteins, we observed elevated levels of MexC and MexD. Both are components of the MexCD-OprJ efflux pump system belonging to the Resistance-Nodulation-Division family (RND) transporters that confer resistance to antimicrobials and other xenobiotics [69, 70] (Fig. 5B). In line, the global repressor of the MexCD-OprJ operon, NfxB, was significantly less abundant in the double mutant. The mexCD-oprJ operon is typically inactive unless the cell is exposed to membrane-damaging agents or antibiotics, triggering an envelope stress reaction [69]. Once activated, the pump exhibits nonspecific translocation activity and can expel a wide range of molecules, including non-antibiotic compounds such as quorum-sensing molecules and other metabolic byproducts [70, 71]. Overexpression of the efflux pump in nfxB-deficient strains has been previously associated with reduced virulence, decreased production of siderophores, impaired cell invasion, and disrupted type three secretion system (T3SS) activity [72–75]. Our proteomic findings were supported by a subsequent transcriptomic analysis, which also identified components of the mexCD-oprJ operon as being significantly upregulated (Fig. 5C). However, in the transcriptome we found elevated transcript levels of the nfxB repressor as well as the PA14 esrC homolog (PA14_60810), representing a second repressor of the operon previously demonstrated to be exclusively expressed under envelope stress conditions to modulate mexCD-oprJ activity [76].

Figure 5.

TrhPO-dependent changes of protein and mRNA levels in P. aeruginosa reveal overexpression of the MexCD-OprJ efflux system. (A) Volcano plot depicting fold changes in protein levels comparing the ΔtrhPO mutant to PA14 WT (n = 3). Red dots represent proteins that are significantly upregulated [−log₁₀(FDR) ≥ 1.3, log₂FC ≥ 0.5], while blue dots represent proteins that are significantly downregulated [−log₁₀(FDR) ≥ 1.3, log₂FC ≤ −0.5]. (B) Genetic organization of the mexCD-oprJ operon with the two repressors (red) and illustration of the architecture of the active MexCD-OprJ efflux pump. (C) Volcano plot depicting fold changes in mRNA levels comparing the ΔtrhPO mutant to PA14 WT (n = 3). Color coding and thresholds are consistent with panel (A). Dark red dots represent selected genes that do not reach the stringent significance threshold. (D) Functional enrichment analysis of differentially expressed genes assigned to GO and PseudoCAP (denoted by asterisk) categories (hypergeometric test, FDR ≤ 0.05). (E) The log₂ protein-to-mRNA ratio of PA14 WT was plotted against the log₂ protein-to-mRNA ratio of ΔtrhPO (Pearson correlation r² = 0.945). Colored dots represent proteins outside the 95% confidence interval (CI) of the difference between ΔtrhPO (proteome/transcriptome) and PA14 WT (proteome/transcriptome) and were defined as exhibiting increased (red) or decreased (blue) protein-to-mRNA ratio upon trhPOdeletion.

In accordance with the observed attenuated virulence phenotype of the ΔtrhPO mutant (Fig. 3), many of the transcriptionally downregulated genes were associated with functional categories attributed to pathogenesis, production of the T3SS, translocation of proteins and peptides, and anaerobic respiration (Fig. 5D). We furthermore found a significant enrichment of genes functionally categorized as transcriptional regulators among the upregulated genes in the ΔtrhPO mutant. Next, we calculated the mRNA-to-protein ratio for PA14 WT and the ΔtrhPO mutant (Fig. 5E). A strong positive correlation (r² = 0.945) was observed between proteomic and transcriptomic datasets of both strains, indicating their overall high similarity. However, among the 89 proteins that displayed an either increased or decreased mRNA-to-protein ratio (outside the 95% CI), we identified the NfxB repressor as exhibiting a reduced protein production from the nfxB transcript in the ΔtrhPO mutant. These data suggest that nfxB might be a MoTT, whose translation efficiency is influenced by xo⁵U-modification and biased codon usage.

MexCD-OprJ induction is not linked to xo⁵U-biased codon usage of nfxB

To test the hypothesis of nfxB being a xo⁵U-MoTT, we cloned the full PA14 WT nfxB gene into a translational reporter system consisting of a strong P_EM7 promotor driving the constitutive expression of an mCherry gene, along with a msGFP reporter fused in frame with the transcript of interest. We then assessed differences in translation efficiency of the nfxB transcript between the xo⁵U hypomodified ΔtrhPO mutant and PA14 WT using flow cytometry analysis (Fig. 6A). The normalized fluorescence ratio remained consistent across both strains, indicating no significant changes in translation efficiency of the nfxB transcript due to the absence of xo⁵U modifications. This suggests that nfxB is not a xo⁵U-MoTT.

Figure 6.

MexCD-OprJ plays a nonessential role in detoxification upon xo⁵U-hypomodification-induced metabolic dysregulation. (A) FACS analysis of PA14 WT and ΔtrhPO carrying the reporter plasmid harboring the PA14 WT nfxBgene sequence. The fluorescence ratio of msGFP signal versus the constitutively expressed mCherry was calculated. Statistical significance was determined by Student’s t-test (unpaired) with ns = not significant. (B) Growth curves of PA14 WT, ΔtrhPO, and ΔtrhPO ΔmexCD in LB medium. Growth was monitored for 24 h. Data represent means ± SD of individual experiments (n = 3). (C) Competition assay for growth: the relative fraction [%] of the population of ΔtrhPO and ΔtrhPO ΔmexCD after co-culturing for 24 h in LB medium is depicted. (D) Kaplan–Meier survival curves of G. mellonella following infection at a MOI of 10 CFU per larvae. Ten larvae per biological replicate were infected (n = 4). Injection of PBS served as a control. Survival of the larvae was monitored for 72 h pi. Significant differences between the survival rates of larvae infected with PA14 WT, ΔtrhPO, or ΔtrhPO ΔmexCD were determined by log-rank (Mantel-Cox) test with ****P ≤ .0001. (E) PAβN (phenylalanine-arginine-β-naphtylamide) is a selective inhibitor of the three main P. aeruginosa RND efflux pumps MexAB-OprM, MexCD-OprJ, and MexEF-OprN. (F) Growth curves of the three strains in LB medium supplemented with increasing concentration of PAßN. Data represent means ± SD of individual experiments (n = 3). (G) Quantification of pyocyanin in LB medium after 8 h. Data represent means ± SD of different biological replicates (n = 3). (H) Quantification of shikimate, shikimate-3-phosphate, chorismate, and aromatic amino acids L-tryptophan, L-phenylalanine, and L-tyrosine using GC-MS analysis. Data represent means ± SD of different biological replicates (n = 4). Statistical significance for panels (G) and (H) was determined by ANOVA (one-way) with Tukey’s multiple comparison test with ns = not significant, *P ≤ .05, **P ≤ .01, ***P ≤ .001, ****P ≤ .0001.

Alternatively, the upregulation of mexCD-oprJ could be a secondary effect triggered by increased intracellular concentrations of accumulated toxic intermediates of the chorismate pathway, which cannot be metabolized due to the disruption of the xo⁵U tRNA modification pathway. To unravel the potential role and importance of MexCD-OprJ in detoxification and thus fitness of the ΔtrhPO mutant, we deleted mexCD in the ΔtrhPO mutant background and monitored its growth. However, no growth impairment was observed in the ΔtrhPO ΔmexCD quadruple mutant compared to both the ΔtrhPO mutant and PA14 WT (Fig. 6B). Furthermore, no fitness disadvantage was observed in direct competition assays with its reference ΔtrhPO mutant in LB medium (Fig. 6C), and deletion of the pump did not affect in vivo virulence in the G. mellonella infection model (Fig. 6D). These results demonstrate that upregulation of the efflux pump is not required to retain fitness or virulence under the tested conditions. To further exclude that alternative efflux pumps compensate for MexCD–OprJ–mediated detoxification due to overlapping substrate specificities, we employed PAβN (phenylalanine-arginine-β-naphtylamide), a selective inhibitor of the three P. aeruginosa RND efflux pumps MexAB-OprM, MexCD-OprJ, and MexEF-OprN [77] (Fig. 6E). However, although overall growth was adversely affected in a dose-dependent manner, no significant differences were observed between the ΔtrhPO ΔmexCD quadruple mutant, the ΔtrhPO mutant, and the PA14 WT treated with PAβN (Fig. 6F).

Notably, we observed increased levels of pyocyanin, shikimate, and chorismate in the ΔtrhPO ΔmexCD quadruple mutant compared to the ΔtrhPO mutant, supporting a role of the efflux pump in exporting excess metabolites (Fig. 6G and H). Our data suggest that the loss of efflux capacity further enhances the redirection of shikimate and chorismate derivatives into the production and secretion of pyocyanin. This pathway appears to serve as an alternative and sufficient means of detoxification, as we did not observe exacerbated fitness defects in the ΔtrhPO ΔmexCD quadruple mutant.

Discussion

Wobble-base modifications in tRNA anticodons play crucial roles in ensuring accurate and efficient translation. Unlike other well-studied wobble base modifications [16, 78], the importance of xo⁵U modifications in bacterial physiology and their influence on virulence regulation remains poorly understood [79, 80]. In this study, we identified the P. aeruginosa genes responsible for xo⁵U biogenesis and characterized the epitranscriptomic landscape, along with the functional roles of hydroxylated derivatives in selected P. aeruginosa tRNAs. We applied a combination of LC-MS/MS-dependent quantification and a nano-tRNAseq approach to identify and quantify xo⁵U modifications at the level of individual P. aeruginosa tRNA isoacceptors. We furthermore mapped the identified modifications onto reference tRNA sequences by detecting increased non-random mismatch frequencies around the xo⁵U-modified wobble positions in PA14 WT tRNAs. These mismatches were distinct and not observed in hypomodified tRNAs of the ΔtrhPO mutant and a chemically synthesized tRNA. Notably, we observed consistent mismatch profiles between individual tRNAs sequenced after purification and those from total tRNA sequencing. This consistency enables accelerated evaluation of modification-induced mismatch profiles across multiple tRNAs simultaneously, eliminating the more time- and cost-intensive process of sequencing each one individually.

Previous studies on the functional roles of xo⁵U tRNA modifications have demonstrated that they enhance structural stability in the anticodon stem-loop prior to codon binding in the A-site of the 30S ribosomal subunit [81, 82] and expand the decoding capacity of the respective modified tRNAs [18–22]. In E. coli, disrupted xo⁵U biosynthesis, coupled with deletion of cognate tRNAs, resulted in severe translational defects [22].

In this study, genome-wide analyses of codon-specific translation efficiency using ribosome profiling confirmed the role of xo⁵U modifications for maintaining accurate translation of P. aeruginosa transcripts enriched in codons sensitive to the modification status of tRNAs. However, with regard to global changes in the translational program, these effects were subtle.

In line with subtle effects on translation efficiency, we did not observe significant shifts in the proteome of the ΔtrhPO mutant lacking xo⁵U modifications, which is in stark contrast to the substantial effects of other tRNA wobble base-modifying systems [16]. One exception was the low expression level of NfxB, the repressor of the MexCD-OprJ efflux pump, in the xo⁵U-hypomodified mutant, despite the presence of nfxB transcript. However, reporter assays using the PA14 WT nfxB transcript showed no xo⁵U-dependent differences in translation efficiency, thus excluding that nfxB is a MoTT.

Given the substantial amounts of xo⁵U-modified tRNAs in the cell, which require high levels of prephenate for both hydroxylation and carboxymethylation [22, 30], it is reasonable to infer that disruption of the xo⁵U tRNA modification pathway leads to the accumulation of metabolic precursors and byproducts. This is supported by our metabolomics analyses, which revealed elevated levels of intermediates associated with chorismate-dependent pathways. The resulting metabolic overflow appears to be redirected into adjacent downstream pathways, resulting in the increased production of aromatic amino acids and phenazines in the ΔtrhPO mutant. The increased levels of these products, without corresponding elevations in transcripts or proteins involved in their biosynthesis, suggest an enhanced flow of metabolites through alternative downstream pathways, rather than an upregulation of these pathways as a response to increased demands.

The upregulation of the MexCD-OprJ efflux pump may also be a secondary response to shifted metabolic states in the ΔtrhPO mutant, as this efflux system is involved in detoxifying accumulating metabolites. Increasing evidence indicates that efflux pumps represent a critical component of the bacterial stress response, functioning to expel toxic compounds, antibiotics, and metabolic byproducts [71–73, 83–87]. MexCD-OprJ is induced in response to envelope stress, whereas other efflux systems, such as the MexAB efflux pump, which shares similar substrate specificity, are constitutively expressed [71, 88]. However, we found that the activity of MexCD-OprJ is not critical for bacterial fitness under the tested environmental conditions, as neither the additional deletion of genes encoding the efflux pump in the ΔtrhPO mutant nor treatment with a pump inhibitor aggravated its fitness defect. It rather seems that a redirected metabolism alone may be sufficient to stabilize bacterial fitness. In line with this, we observed significantly elevated levels of shikimate and chorismate along with a modest further increase in pyocyanin production in the ΔtrhPO ΔmexCD quadruple mutant compared to the ΔtrhPO mutant, suggesting that the detoxification of chorismate derivatives via pyocyanin production, followed by its export from the cell, stabilizes fitness of the ΔtrhPO mutant and fully compensates for the inactive efflux pump.

Collectively, our results indicate that the observed phenotypes of the ΔtrhPO mutant—ranging from decreased fitness in in vitro cultures to an impaired ability to infect host cells and attenuated virulence in the G. mellonella model—are driven by changes in metabolic fluxes rather than shifts in proteome profiles. These metabolic changes are marked by the accumulation of chorismate derivatives, including aromatic amino acids and phenazines, alongside reduced production of the iron-scavenging siderophore pyoverdine. The metabolic imbalance, rather than differences in translation efficiency, likely accounts for the strongest proteomic effect: the upregulation of MexCD-OprJ. Our findings underscore that xo⁵U hypomodification-induced metabolic alterations are main drivers of P. aeruginosa virulence shifts, highlighting the multifunctional roles of tRNA-modifying enzymes in maintaining metabolic balance and regulating bacterial pathogenicity.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy—EXC 2155 “RESIST”—Project ID 390874280. S.H. received funding within the SFB/TRR-298-SIIRI—Project-ID 426335750 and in the SPP 2389 (HA 3299/9–1, AOBJ: 687646), from the Ministry of Science and Culture of Lower Saxony (Niedersächsisches Ministerium für Wissenschaft und Kultur), BacData ZN3428, and from the Novo Nordisk Foundation (NNF 18OC0033946). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Funding to pay the Open Access publication charges for this article was provided by the corresponding author’s funding and support of the host institute.

Data availability

RNA-seq, RiboSeq, Nano-tRNAseq, and Scripts used for base calling, alignment, and subsequent processing steps data have been deposited in NCBI GEO database (GSE283553), European Nucleotide Archive (PRJEB83028), and GitHub (https://github.com/DepledgeLab/tRNA-studies), respectively. Proteomics data is available at the PRIDE website with project accession PXD060249; Project Name: tRNA hydroxylation is an epitranscriptomic modulator of metabolic states affecting P. aeruginosa pathogenicity.