5-Methyluridine is ubiquitous in Pseudomonas aeruginosa tRNA and modulates antimicrobial resistance and virulence

Abstract

Building on decades of work in characterizing the dozens of RNA modifications in the microbial epitranscriptome, recent advances in analytical technology and genetics have revealed systems-level functions for many tRNA modifications. The tRNA (uracil-5-)-methyltransferase TrmA and its product, 5-methyl uridine (m⁵U) at position 54 in the T-loop, however, has not been linked to a specific phenotype. Here, we defined the functional and biological roles of TrmA in Pseudomonas aeruginosa (PA14), a major multidrug-resistant pathogen. Surprisingly, though TrmA was found to site-specifically catalyze m⁵U54 on all PA14 tRNAs, loss of TrmA had no effect on the levels of any of 36 tRNA modifications except m⁵U and had minimal effects on multiple phenotypic parameters, including growth rate, morphology, motility, and biofilm formation. However, loss of TrmA conferred a striking polymyxin antibiotic resistance. mRNA and tRNA profiling and proteomics analyses revealed that TrmA regulates the expression of codon-biased gene families at the level of translation, including components of a type III secretion system (T3SS). Loss of TrmA upregulated T3SS, leading to increased macrophage IL-1β in bacterial challenge tests. Altogether, these results revealed novel biological functions of TrmA and its roles in modulating gene expression at multiple levels in P. aeruginosa.

Graphical Abstract

Introduction

More than 170 post-transcriptional modifications have been identified in the chemical diversity of RNA, with most found in transfer RNA (tRNA) molecules [1]. tRNA modifications function as local regulators of translational accuracy and efficiency and RNA secondary structure [2]. More recent evidence also points to a systems-level function for the tRNA epitranscriptome in regulating translation of proteins in response to environmental changes [3].

5-Methyluridine or ribothymidine (m⁵U) is one of the most common modifications found at position 54 in the T arm (TΨC) loop in nearly all tRNA species of most prokaryotes and most elongator tRNAs in eukaryotes [4]. The presence of m⁵U stabilizes tRNA secondary structure [5]. In Escherichia coli and Saccharomyces cerevisiae, m⁵U formation in tRNA is catalyzed by TrmA and Trm2, respectively. Mammalian tRNAs also possess the m⁵U modification, with TRMT2A and TRMT2B responsible for this modification in human cytosolic and mitochondrial tRNAs, respectively [6-8]. Previous studies have shown that TrmA preferably binds to and modifies unmodified tRNA [9], while more recent studies have demonstrated that m⁵U modification contributes to tRNA maturation [10, 11]. In addition to the m⁵U modification in tRNA, bacterial TrmA has been shown to methylate the T-loop UUC sequence in the tRNA-like domain (TLD) in tmRNA as well as 16S rRNA in vitro [12, 13]. Human homologs TRMT2B have dual functions in the formation of m⁵U in both tRNA and rRNA [8]. TrmA also possesses tRNA chaperone-like activity involved in L-shaped folding of tRNA in E. coli [14].

Despite the widespread presence of m⁵U in tRNA, the loss of trmA in E. coli has little impact on growth rate, tRNA binding to the ribosome, and codon recognition [15], though cells lacking trmA showed a lower capacity to survive during stationary phase when growing in the mixed population with wild-type cells containing m⁵U [15]. The absence of trm2 showed reduced sensitivity to salt stress, hygromycin B, and paramomycin, while increased sensitivity to cycloheximide in S. cerevisiae [10]. In mammalian cells, TRMT2A has been linked with breast cancer prognosis [16] and its effect on cell growth support the idea of its role as a cell cycle regulator [17]. Loss of TRMT2A also reduced translational fidelity and induced tRNA fragmentation [7, 18].

Here, we used a multi-omics approach to define the functions of m⁵U in tRNA in Pseudomonas aeruginosa. As an important Gram-negative pathogen [19, 20], transcriptional regulation of the P. aeruginosa response to environmental changes has been well studied [21-23] compared to post-transcriptional contributions to phenotype [24, 25]. We now report novel phenotypes of TrmA and its impact at multiple levels in P. aeruginosa, with a significant contribution of antibiotic resistance and translation of gene expression in virulence.

Results

TrmA is a tRNA methyltransferase that catalyzes m⁵U formation in all tRNAs of P. aeruginosa

Our first objective was to identify the enzyme catalyzing the formation of m⁵U in P. aeruginosa PA14. The PA14 genome possesses a 1,092 bp open reading frame (ORF), PA14_62450, which is annotated as a hypothetical gene encoding tRNA uracil-5-methyltransferase protein with a theoretical molecular weight of 41.3 kDa. PA14_62450 was aligned with established TrmA in E. coli using ESPript3.0 (https://espript.ibcp.fr/ESPript/ESPript/) and found to share 49.1% identity and 63.9% similarity, as shown in Figure 1B. PA14_62450 also contains several features that strongly support its identity as tRNA uracil-5-methyltransferase: (1) a GXGX motif at residues 220-223, which is the putative S-adenosyl methionine (SAM) binding site belonging to the Rossmann-fold SAM-dependent methyltransferase superfamily [26]; (2) a cysteine at position 324, which is a catalytic amino acid residue; and (3) conserved amino acids in the C-terminal catalytic domain including F188, Q190, G220, D299, R302, C324, and E358, which are involved in TrmA-tRNA intermediate formation and/or methyltransferase activity [27]. This level of homology was consistent with PA14_62450 encoding a tRNA (uracil-5-)-methyltransferase enzyme catalyzing m⁵U formation (Figure 1A), justifying annotation of PA14_62450 as TrmA.

Figure 1.

TrmA is a tRNA (uracil-C(5))-methyltransferase in PA14 and catalyzes the formation of m⁵U54 on all tRNA species. (A) TrmA catalyzes m⁵U formation in RNA, using SAM as a cofactor. (B) Sequence alignment between TrmA from Escherichia coli and Pseudomonas aeruginosa using ESPript 3.0 (http://endscript.ibcp.fr/). TrmA in P. aeruginosa shares 49.1% and 63.9% identity (red box) and similarity (white box) to TrmA in E. coli, respectively. Amino acid residues revealed to be associated to tRNA binding and catalytic activity are symbolized by *. Ec; Escherichia coli, Pa; Pseudomonas aeruginosa (coil = helix, arrow = β-strand, T = turn). (C) In vitro methyltransferase activity of TrmA. The recombinant TrmA was able to methylate total small RNAs extracted from trmA mutant. The reaction without TrmA served as a negative control. The quantity of SAH generated by TrmA methyltransferase activity was determined using MTase-Glo Bioluminescent Assay Kit (Promega). (D) Ribonucleoside analysis of tRNA isolated from wild-type and trmA mutant strains using LC-QQQ. The level of each modification was normalized using sum of UV signals for A, C, G and U. The normalized signal was calculated as a fold-change relative to wild-type tRNA. (E) Identification of TrmA substrates using synthetic tRNAs with recombinant TrmA and SAM. The reaction with tRNA isolated from wild-type cells served as a negative control. Data represent the mean ± SD for three experimental replicates. Asterisks denote a significant difference using Student’s t-test (**P<0.01, ***P<0.001).

To confirm that TrmA functions as uracil-5-methyltransferase enzyme, purified recombinant TrmA was incubated with total small RNA (>85% tRNA) from wild-type, trmA mutant, or trmA-complemented strains, with SAM as a methyl donor and production of S-adenosyl homocysteine (SAH) as the index of methyltransferase activity. As shown in Figure 1C, levels of SAH are barely detectable in the wild-type cells, which indicates that potential tRNA substrates of TrmA are fully modified in vivo. Small RNA from TrmA-deficient cells, however, led to significant SAH production as expected for hypomethylated tRNA, while the trmA-complemented strain partially restores the in vivo modification levels and reduced SAH levels (Figure 1C; Table S1). To investigate which specific tRNA modification is catalyzed by TrmA, small RNA from wild-type and trmA mutant cells was isolated; and modified ribonucleosides were analyzed by chromatography-coupled tandem quadrupole mass spectrometry. Of the 36 ribonucleosides analyzed in Figure 1D, only m⁵U was significantly reduced in the trmA mutant strain. These results establish PA14_62450 in P. aeruginosa as tRNA (uracil-5-)-methyltransferase TrmA. The presence of 5-methyluridine (m⁵U) in most tRNAs, as observed in E. coli and Eukarya [4], is supported by the observation of TrmA methyltransferase activity (i.e., SAH production) with synthetic versions of all 45 isoacceptors known to exist in PA14 [28] (Figure 1E).

Finally, we defined the location of TrmA activity in tRNA based on the observation that TrmA methylates uridine at position 54 in E. coli and most elongator tRNAs in eukaryotes [4]. Here we used chromatography-coupled quadrupole time-of-flight (LC-QTOF) high-resolution mass spectrometry to map the location of m⁵U in synthetic tRNA-Cys, tRNA-Phe, and tRNA-Sec reacted with TrmA and SAM. RNase T1 digestion was predicted to release a T-loop fragment m⁵UUCG (positions 54-57; MW 1,214.197 Da), which was confirmed by LC-QTOF as a doubly-charged negative ion with m/z 606.091 eluting at 4.8 min for tRNA-Cys (Figure 2A), tRNA-Phe (Figure 2D), and tRNA-Sec (Figure 2G). Mass spectra (Figure 2B, E, and H) and collision-induced dissociation fragmentation (CID) spectra (Figure 2C, F, and I) were all consistent with m⁵UUCG. These results indicate that P. aeruginosa TrmA catalyzes the methylation of the C5 atom of uridine at position 54 in the T-loop of tRNA. Interestingly, the CID spectrum for tRNA-Sec showed several product ions consistent with Um⁵UCG in addition to those for m⁵UUCG (Figure 2I), suggesting some flexibility in TrmA target recognition in this sequence.

Figure 2.

Localization of m⁵U on tRNA digested with RNase T1 using LC-QTOF analysis. Extracted ion chromatograms for m/z 606.091 released from tRNA-Cys-GCA (A), tRNA-Phe-GAA (D), and tRNA-Sec-UGA (G). Mass spectra showing (M-2H)⁻² ions with m/z 606.091 for tRNA-Cys-GCA (B), tRNA-Phe-GAA (E), and tRNA-Sec-UGA (H). CID spectra for the ion with m/z 606.091 from tRNA-Cys-GCA (C) tRNA-Phe-GAA (F), and tRNA-Sec-UGA (I). * and ** indicate product ions specific to m⁵UUCG and Um⁵UCG, respectively.

TrmA-dependent changes in gene expression are largely transcriptionally driven

We next explored the role of TrmA on gene expression at the level of transcription and translation. Three biological replicates of wild-type and trmA mutant cultures with an OD600 of 0.4–0.6, all from the same batch, were used for transcriptomics and proteomics analyses. Transcriptional profiling showed that 723 of 4,160 genes (~17%) were significantly altered by the loss of TrmA (P<0.05; Figure S1A, Supplementary Data S1), with 104 and 183 genes upregulated and downregulated, respectively, in the trmA mutant by >1.3-fold (Figure 3A). Gene ontology (GO) enrichment [29] of the differentially expressed genes revealed that TrmA-dependent upregulated mRNAs arose from genes involved in cellular hyperosmotic salinity response, polyamine catabolism, queuosine biosynthesis, and glycine betaine transport, while downregulated genes involved secondary metabolite biosynthesis and positive regulation of metabolism (Figure 3B; Table S2). KEGG pathway analysis of DEGs showed that the mRNA changes affected several pathways including metabolic pathways, nitrogen metabolism, glyoxylate and dicarboxylate metabolism, phenazine biosynthesis, carbon metabolism, and one carbon pool by folate (Figure S2).

Figure 3. Transcriptional profiling and proteomic comparisons of PA14 wild-type strain and the trmA mutant.

(A) Venn diagram of differentially expressed mRNA and proteins in the trmA mutant. (B) Biological process GO functional analysis of differentially expressed genes (DEGs) data in the trmA mutant relative to the wild-type strain. (C) Biological process GO functional analysis of differentially expressed proteins (DEPs) data in the trmA mutant relative to the wild-type strain. (D) Comparison of changes in mRNA and protein levels caused by loss of TrmA in the 80 genes with log₂ fold-change values with P<0.1. (E) Correlation plot for values in panel d, yielding R² = 0.32 with P<0.0001.

Steady-state protein levels were then assessed for the wild-type and the trmA mutant strains using TMT-based proteomics. A total of 2,343 proteins were detected and quantified, of which only 170 (7.2%) showed significantly differential expression (P<0.05; Figure S1B, Supplementary Data S2). Among the 27 proteins with fold-change values >1.2 (Table S3), 16 were upregulated, and 11 were downregulated in the trmA mutant compared to wild-type strain (Figure 3A). GO enrichment analysis of biological process revealed that the upregulated proteins were enriched in protein secretion by type III secretion systems (T3SS) and negative regulation of protein secretion, while the downregulated proteins were enriched in secondary metabolite biosynthetic processes and fatty acid biosynthetic processes (Figure 3C).

Comparison of mRNA and protein changes for 80 genes with fold-changes at the P<0.1 level showed similar behavior for most of the genes (Figure 3D), with a correlation plot revealing a strongly positive correlation (Figure 3E, P<0.0001). This correlation was strengthened in a comparison of the 62 genes with log₂ fold-change values at P<0.05 (Figure S3). These results are consistent with a contribution of transcriptional regulation of gene expression to loss of TrmA.

TrmA-dependent changes in the tRNA landscape

Since m⁵U is present in all tRNAs in P. aeruginosa, we set out to assess the effect of loss of TrmA on the RNA landscape and tRNA pool. TrmA-dependent changes in the proportions of tRNA, 16S rRNA, or 23S rRNA were assessed by quantitative comparison of Agilent Bioanalyzer traces for wild-type and trmA mutant strains (Figure S4), which revealed no significant differences. The effect of TrmA loss on the levels of individual tRNA isoacceptors and tRNA fragments was quantified using AQRNA-seq [30, 31] (Supplementary Data S3). As shown in Figure 4A, loss of TrmA caused few changes in the levels of 47 isoacceptors, of which 7 tRNAs comprising nearly 45% of total tRNA abundance (Gly-GCC, Arg-ACG-1, Ala-GGC, Asp-GUC-2, Gln-UUG, Leu-CAG, and Tyr-GUA), while Arg-ACG-2 had the lowest expression (<0.01%). Consistent with studies in yeast [32], the correlation between tRNA abundance and the genome usage of the corresponding codon(s) read by the tRNAs for wild-type and trmA mutant strains was found to be modest, with Spearman’s rank correlation coefficients of 0.50 and 0.46 for wild-type and trmA mutant data sets, respectively (Figure S5). AQRNA-seq data were also mined for tRNA fragments, which revealed no apparent differences for the 47 tRNA isoacceptors in the trmA mutant (Figure S6). tRNA fragmentation is apparent in the stack plots in Figure S6 as reads that do not start at the 3’-end of the sequence (far right x-axis). For example, tRNA-Glu-UUC shows ~50% of reads starting at position 59 in the T-loop while full length tRNAs start at position 76 (Figure S6).

Figure 4. TrmA loss alters several tRNA isoacceptors and causes codon-biased translation of type III secretion system proteins.

(A) tRNA landscape of PA14 wild-type and trmA mutant strains shown as percentage of the tRNA pool. (B-E) Partial least-squares regression analysis (PLSR) of TrmA-dependent mRNAs and proteins and codon usage patterns of their genes. The scores plots of PLSR analysis demonstrated the analysis of upregulated (blue) and downregulated (red) genes (B) and proteins (D) in the trmA mutant and the wild-type cells. Ellipse in the scores plot denotes the Hoteling T² limit. The loading plots of PLSR analysis demonstrated the analysis of codon usage pattern of differentially expressed genes (C) and proteins (E) in the trmA mutant and the wild-type strains. The outer and inner ellipses in the correlation loadings plot depict 100% and 50% explained variance, respectively. Blue color depicts synonymous codons over-represented in the upregulated proteins, whereas red color depicts the respective synonymous codons in the downregulated proteins in the trmA mutant. Arrows indicate the quantitative expression and altered direction of tRNA expressed genes in the trmA mutant for cognate codons. (F) tRNA abundance of Ser-GCU in wild-type and trmA mutant strains. (G) Codon usage in mRNAs for upregulated proteins. AGU is cognate codon for tRNA-Ser-GCU. (H) tRNA abundances of Ile-GAU in wild-type and trmA mutant strains. (I) Codon usage in mRNAs for downregulated proteins. AUC is cognate codon for tRNA-Ile-GAU. Z-scores for codon usage relative to genome average. Dot plot demonstrated data for three biological replicates with bars for mean ± SD. Asterisks denote significant difference using Student’s t-test (*P<0.1).

However, several differences were noted in the tRNA pool of the trmA mutant compared to wild-type strain: increased Pro-CGG, Pro-GGG, Ser-GGA, Ser-GCU, Ser-CGA, Ser-UGA, Sec-UCA, and Und-NNN-2, and decreased Ile-GAU, Gly-UCC, Arg-CCU, Arg-CCG-2, fMet-CAU-1, fMet-CAU-2, and Met-CAU (P<0.1) (Figure 4A and Supplementary Data S3). These results point to a small but important number changes in the tRNA pool caused by loss of a universal tRNA modification, with implications for codon-biased translation, as discussed next.

tRNA pool reprogramming caused by loss of TrmA correlates with codon-biased translation of type III secretion system

The observed changes in the tRNA pool in the TrmA mutant raised questions about contributions of translational regulation of gene expression in the TrmA-dependent phenotype. Based on previous studies in prokaryotes showing that tRNA modification-dependent changes in the tRNA pool regulate translation of mRNAs with biased codon usage matching the altered tRNA pool [3, 33], we assessed the link between tRNA m⁵U54 and the translation of mRNAs with biased codon usage. We first quantified codon usage in 5,901 PA14 genes (Supplementary Data S4; Z-scores for isoacceptor frequencies calculated with the gene-specific codon counting algorithm [34]). We then performed partial least squares regression (PLSR) analyses to assess codon biases in TrmA-dependent changes in mRNA and protein levels. PLSR analysis showed a scores plot with few distinctions between upregulated and downregulated mRNAs and no obvious codon biases in the loadings plot (Figure 4B, C; Supplementary Data S4). However, for differentially expressed proteins (Table S3), the scores plot shows a clear discrimination of upregulated and downregulated proteins based on the codon usage biases of their genes (Figure 4D; Supplementary Data S4). Moreover, the loadings plot revealed pairs of synonymous codons differentially used in upregulated versus downregulated proteins (Figure 4E). These data suggest that loss of TrmA causes a translational response involving changes in codon-biased translation.

We next tested the association between altered tRNA levels in the trmA mutant and codon usage patterns in genes for TrmA-dependent proteins. We found that 7 of upregulated proteins overused codon AGU read by tRNA Ser-GCU that was increased in the trmA mutant (Figure 4F, G). Additionally, 8 of 11 downregulated proteins in the trmA mutant overused the AUC codon read by tRNA Ile-GAU that was decreased in the trmA mutant (Figure 4H, I). These results suggest that the lack of TrmA altered tRNA levels to regulate translation of specific mRNAs containing codon biases matching the altered tRNAs.

GO analysis of the 16 upregulated proteins in the trmA mutant revealed a striking enrichment (10 of 16) of proteins related to T3SS: export proteins PscE, PscF and PscD, outer membrane protein PopN and PopD, transcription regulator ExsA, exotoxins T and U (ExoT, ExoU), ExoU chaperone SpcU, and a chaperone protein (Table S3) [35]. Given previous reports of T3SS-mediated modulation in the pro-inflammatory cytokine IL-1β in P. aeruginosa-infected macrophages [35-37], we compared IL-1β production in LPS-induced Raw264.7 macrophages infected with wild-type, trmA mutant, and trmA-complemented P. aeruginosa strains. The trmA mutant-infected macrophages showed higher IL-1β levels than macrophages infected with the wild-type or trmA-complemented strain (Figure 5A), suggesting that the absence of trmA affected the host immune response or inflammatory response.

Figure 5. Phenotypic changes in PA14 caused by loss of TrmA.

(A) A type III secretion system (T3SS) in PA14 modulates the host inflammatory response. Secretion of IL-1β from LPS-pretreated Raw264.7 macrophages infected with wild-type, trmA mutant, and trmA-complemented strains. Data represent mean ± SD for three biological replicates. (B) Growth effects caused by TrmA-dependent polymyxin B resistance. Growth curves for wild-type, trmA mutant, and trmA-complemented strains in MHB medium with or without 2 μg/mL polymyxin B. Data represent mean ± SD for three biological replicates. (C) Detection of ROS production using DCF in wild-type, trmA mutant, and trmA-complemented strains treated with 1 μg/mL polymyxin B for 20 h. Data represent mean ± SD for four biological replicates. Asterisks denote significant difference using Student’s t-test (*P<0.05).

Absence of trmA confers polymyxin B resistance

In spite of the near-ubiquitous presence of m⁵U54 in prokaryotic tRNAs [38] (Figure 1E), the disruption of trmA in E. coli [15] does not cause obvious fitness cost. Here we broadly surveyed TrmA-dependent phenotypes in P. aeruginosa PA14, starting with bacterial growth rate, morphology, biofilm formation, and motility. The absence of trmA had no impact on any of these parameters (Figure S7). To explore other TrmA effects using Biolog plates (PM11-20), we performed phenotypic profiling of wild-type and trmA mutant cells with 236 compounds (Table S4), including drugs, oxidants, and disinfectants. We found only one striking phenotype caused by loss of TrmA: resistance to polymyxin B. The resistant phenotype was verified using a broth microdilution assay, which revealed a minimum inhibitory concentration (MIC) of 4 μg/mL polymyxin B for the trmA mutant compared to 2 μg/mL for wild-type and trmA-complemented strains. A growth study demonstrated that the trmA mutant was able to grow in MHB containing 2 μg/mL polymyxin B, whereas the wild-type and the complemented strains were not (Figure 5B).

We next undertook an assessment of the basis for the TrmA-dependent polymyxin B resistance. Mechanisms of polymyxin resistance mainly involve alterations of the cell surface including changes in ultrastructure and cationic chemical modifications of the outer membrane lipopolysaccharide (LPS) that reduce drug binding [39, 40]. We used a Zetasizer to measure bacterial surface charge (zeta potential, ZP) of wild-type and trmA mutant cells in exponential growth phase, stationary phase, and biofilm stage with or without polymyxin B. We also examined bacterial ultrastructure using transmission electron microscope (TEM) with the wild-type and trmA mutant strains. However, our results revealed no significant differences in the zeta potential or cell membrane structure (Figure S8).

Since previous studies showed that polymyxin B exposure resulted in increased levels of potentially cytotoxic reactive oxygen species (ROS) in E. coli [41] and Acinetobacter baumannii [42], we hypothesized that polymyxin resistance in the trmA mutant might result from low levels of drug-induced ROS in the mutant relative to the wild-type. Indeed, treatment of the trmA mutant with polymyxin B caused lower intracellular ROS levels than wild-type P. aeruginosa, with the ROS level restored in the trmA-complemented strain (Figure 5C). These results indicate that the trmA-dependent polymyxin B resistance is related to reduced levels of ROS in polymyxin B treatment but unrelated to changes in the bacterial cell surface or cell structure.

Discussion

Here we showed that P. aeruginosa TrmA is a bona fide tRNA methyltransferase that is responsible for the formation of m⁵U in tRNA at position 54, with loss of the enzyme linked to limited but striking phenotypes of antibiotic resistance and virulence mediated by tRNA-linked codon-biased translation. In this study, we analyzed 57 modified ribonucleosides in our LC-MS/MS MRM table but detected only 36 (Table S5). Among the three publications in which tRNA modifications have been analyzed in P. aeruginosa [43, 44], including the present work, a total of 60 modified ribonucleosides have been analyzed but only 18 (cmnm⁵s²U, acp³U, ct⁶A, ho⁵U, inm⁵U, m¹A, m²₂G, m³U, m⁶t⁶A, mcmo⁵U, mo⁵U, ms²i⁶A, ms²io⁶A, oQ, preQ1, Q, rNA, s²U) were detected by the three reporting groups and 25 (cmnm⁵U, cmo⁵U, D, io⁶A, m²G, m³C, m⁵C, m⁶₂A, mnm⁵s²U, mnm⁵U, s²C, Am, Cm, Gm, I, i⁶A, m¹G, m²A, m⁵U, m⁶A, m⁷G, s⁴U, t⁶A, Um, Ψ) were detected in common by at least two groups (Table S5). While the latter represents a presumptive set of modifications in PA14, some of the modifications in Table S6 have not been validated with chemical standards and some almost certainly exist but were not detected, so further work is needed. What has been validated here, however, is that all tRNA isoacceptors in P. aeruginosa PA14 are substrates for TrmA, within a two-fold variance in efficiency (Figure 1E), and all contain the m⁵U consensus sequence UUC at position 54-56 in the T loop (Supplementary Data S5). This observation is consistent with the conclusion that all tRNAs in E. coli and most elongator tRNAs in mammalian cells contain m⁵U [4]. It should be noted that m⁵U54 is absent in some organisms, such as Mycoplasma capricolum, Methanococcus vannielli, and Mycobacterium smegmatis [45-47], a lack of essentiality consistent with the paucity of phenotypes observed here in the trmA mutant of PA14.

While there was no alteration in the overall tRNA modification profile, except for m⁵U, in the P. aeruginosa trmA mutant (Figure 1D), recent publications have demonstrated changes in other tRNA modifications beyond m⁵U in the E. coli trmA mutant [10, 11]. Jones et al. reported a decrease in acp³U and an increase in i⁶A in tRNA-Phe in the ΔtrmA E. coli [10]. Schultz et al. documented reduced levels of (1) acp³U47 in tRNA-ArgACG-1, tRNA-IleGAU-1, tRNA-LysUUU-1, tRNA-PheGAA-1, and tRNA-ValGAC-1; (2) ms²i⁶A37 in tRNA-CysGCA-1 and tRNA-SerCGA-1; and (3) s⁴U in tRNA-Asp in the E. coli ΔtrmA [11]. These differences in detected tRNA modifications is likely due to variations in sample types and detection methods used in the experiments. In our studies, total tRNA was analyzed using LC-MS/MS, whereas the Koutmou group focused on tRNA-Phe using LC-MS/MS, and the Kothe group studied total tRNA using multiplex small RNA sequencing (MSR-seq) [10, 11]. Further investigation is required to gain deeper insights into the impact of TrmA on tRNA modification landscape, particularly in specific tRNA species, in P. aeruginosa PA14.

In addition to the tRNA modifications, we also quantified expressed tRNA isoacceptors in P. aeruginosa PA14. The expression pattern of tRNA isoacceptors was different from that of Mycobacterium bovis Bacille Calmette-Guérin (BCG) and E. coli [11, 30], which might be expected based on the different codon usage patterns in the three bacteria (Table S7-S10). In P. aeruginosa, there were approximately 7 tRNAs expressed in a similar level, which accounted for approximately 45% of tRNA pool, whereas tRNA Lys-CUU has shown the highest expression (20%) among tRNAs in the BCG (Table S8, S9 and Supplementary Data S6). In E. coli, approximately 7 tRNAs accounts for 50 % of the tRNA pool, however, with different set of the tRNAs from P. aeruginosa (Table S8, S10). The lowest expressed tRNA in BCG and E. coli tRNA pool was Ser-GGA and Thr-CGU-2 [11, 30], respectively, while tRNA Arg-ACG-2 showed the lowest expression in P. aeruginosa (Supplementary Data S3, S6; Table S8-S10). As shown in Figure S5, there is a modest correlation between the tRNA read count, which correlates directly with the number of tRNA copies in AQRNA-seq [30] for PA14 and MSR-seq for E. coli [11], and the usage frequency of the cognate codons read by the tRNAs. A modest correlation was similarly observed in the trmA mutant, though the tRNA expression levels differed from respective wild-type forms of PA14 and E. coli (Figure S5). The correlation between the tRNA abundance and the usage frequency of the cognate codons in BCG was much weaker due to the high abundance of 4 tRNA isoacceptors (Figure S5). The significantly altered tRNAs in the P. aeruginosa trmA mutant and E. coli trmA mutant were not the tRNAs responsible for reading optimal codons (Figure S5; Table S8, S10 and Supplementary Data S6). These results are consistent with published studies showing modest correlations between tRNA abundance and optimal codons [30, 48].

The present studies showed that loss of TrmA did not broadly affect PA14 phenotype consistent with what was observed in E. coli [11]. Nonetheless, loss of TrmA in PA14 caused a shift in the tRNA pool that correlated with selective translation of codon-biased mRNAs for T3SS genes. Among the proteins differentially regulated by P. aeruginosa TrmA (Table S3), upregulated proteins significantly overused the Ser-AGU codon, which coincided with increased levels of its cognate tRNA-GCU (Figure 4F, G). At the same time, nearly all downregulated proteins in the P. aeruginosa trmA mutant overused codon Ile-AUC, which correlated with reduced levels of its cognate tRNA-GAU (Figure 4H, I). In E. coli, loss of trmA does not affect global translation but affects translation of specific codons, for instance, decrease in Gln CAA, Leu UUG, Arg CGG and AGG, and Cys UGU codons and increase in Leu UUA codon [11]. Furthermore, most of the downregulated proteins in E. coli ΔtrmA overused the less efficient codons such as Arg CGG codon in translation [11]. How TrmA causes differential regulation of specific tRNA isoacceptor levels is not clear from the results presented here. The differential impact on protein expression might be from the differential effect of m⁵U on each tRNA despite its presence in all tRNAs. As described by Kersten et al., translation of poly-A peptide dependent poly(Lys) was increased when the tRNA-Lys-UUU lacked TrmA-dependent m⁵U54 modification, whereas the same loss of m⁵U54 in tRNA-Phe-AAA had no effect on translation of poly-U-dependent poly(Phe) [49]. One possible mechanism to explain m⁵U54 effects on tRNA pool composition lies in the functional duality of TrmA as both a methyltransferase and a tRNA-folding chaperone [14]. Keffer-Wilkes et al. showed that the RNA-binding domain of TrmA disrupts interactions between the D and T arms of tRNA, which confers a chaperone activity [14]. The loss of the chaperone function of TrmA reduces cellular fitness [14]. A parallel observation was made for TruB, which catalyzes the formation of pseudouridine at position 55 (Ψ55) in all tRNAs in E. coli [50-52] and functions as a chaperone required for tRNA folding [50-52]. With parallels to TrmA, Ψ55 improved the translation efficiency of the CGA codon but not CGU in spite of its presence on all tRNAs [51]. We do not know if Ψ55 is present in PA14 tRNAs. In total, these findings all point to a contribution by TrmA in regulating the composition of the tRNA pool and thus regulating protein expression by selective translation of codon-biased mRNAs.

Limited as the TrmA-dependent phenotypic changes were, loss of TrmA did significantly alter PA14 host immune response and antibiotic sensitivity. Phenotypic changes directly related to TrmA-dependent codon-biased translation were striking for proteins in a T3SS as well as the ExsA master activator of T3SS (Table S3 and Supplementary Data S4F). This system has been studied for its ability to activate the caspase-1 pathway, induce IL-1β secretion, and cause cell death through pyroptosis in infected macrophages [53]. Previous studies have reported that macrophages infected with P. aeruginosa deficient in T3SS translocon components (PopB or PopD) showed reduced secretion of IL-1β, whereas macrophages infected with P. aeruginosa deficient in T3SS effector (ExoS or ExoU) exhibited increased production of IL-1β [36, 37]. Furthermore, a significantly lower abundance of IL-1β was observed in macrophages infected with the exoS/popB mutant [37]. However, the present study found that TrmA-dependent upregulation of both T3SS components and effector proteins caused increased secretion of IL-1β in Raw264.7 macrophage cells infected with the trmA mutant (Figure 5A). Interestingly, T3SS-suppressing behavior of TrmA in PA14 is opposite to that observed for other T3SS-regulating, tRNA-modifying enzymes in P. aeruginosa. For example, MiaB mediates the final step of ms²i⁶A synthesis in tRNA and acts as an activator of T3SS in the ExsA-dependent manner [54]. Further, TruA catalyzes the formation of Ψ at positions 38, 39, and 40 in tRNA, and its absence has been reported to impair the expression of T3SS [55]. The effect of loss of TruB-mediated pseudouridylation of U55 in PA14 tRNA [56] on T3SS has not been determined.

Another striking phenotype caused by loss of trmA was resistance to the antibiotic polymyxin B (Figure 5B), one of the last-line treatments for P. aeruginosa infections [40]. This led us to investigate the mechanistic basis for the drug resistance. There are multiple resistance mechanisms for the cationic peptide structure of polymyxin B and related antibiotics (e.g., colistin), termed the cationic antimicrobial peptide (CAMP) resistance pathway^† (ko01503) in the Kyoto Encyclopedia of Genes and Genomes (KEGG) [57]. For example, the most common mechanism of resistance to polymyxin B involves cell surface charge alterations [58], with cationic polymyxin B peptide structure binding to anionic lipopolysaccharide (LPS) residues on the cell surface of Gram-negative bacteria [58]. Resistance is induced by adding cationic groups to the outer membrane LPS, such as 4-amino-4deoxy-L-arabinose (L-Ara4N) catalyzed by arnBCADTEF and PEtN. This decreases the overall negative charge of the membrane and reduces polymyxin B binding [58]. Another study demonstrated that loss of ornithine decarboxylase or SpeD led to increased susceptibility to polymyxin B in Burkholderia cenocepacia [59]. This enzyme converts ornithine into putrescine, which is a common polyamine found in this bacterium and reported to increase resistance to antimicrobial peptides in P. aeruginosa [59, 60]. Finally, several two-component signal transduction systems have been reported to influence LPS-mediated polymyxin resistance, including PhoPQ, PmrAB, ColSR, CprSR, and ParSR [40]. However, these mechanisms are apparently not operational in the trmA mutant since we did not observe a change in surface charge or cell morphology upon loss of TrmA (Figure S7). Further, none of the genes noted here changed in expression upon loss of TrmA, either at the transcriptional or translational levels (Supplementary Data S1, S2). However, GO analysis revealed that mRNAs for polyamine catabolism were significantly upregulated by the loss of trmA (Figure 3B; Table S2), which is consistent with their role in polymyxin B resistance [59, 60]. Another possible mechanism for TrmA-dependent polymyxin B resistance involves a role for ROS. The general idea that ROS production serves as the final pathway for antibiotics was supported by our studies with PA14, showing that the loss of TrmA resulted in reduced ROS levels following polymyxin B treatment (Figure 5C), which adds to the controversy surrounding this proposed mechanism [41, 42, 61, 62].

In summary, we discovered that P. aeruginosa TrmA catalyzes the formation of m⁵U on all tRNAs and, while its loss produces few phenotypic changes, those that do occur reveal functions for TrmA in the host inflammatory response infection and polymyxin B resistance. The results revealed a contribution of tRNA-mediated codon-biased translation to regulating expression of virulence genes. These findings expand our understanding of links between the TrmA as a part of tRNA epitranscriptome and microbial physiology.

Material and methods

Bacterial strains, growth condition, and plasmids

Bacterial strains and plasmids used in this study are listed in Table S11. Bacteria were cultured in LB at 37 °C. Media were supplemented with antibiotics, if necessary, as follows: for E. coli, 100 μg/mL ampicillin, 15 μg/mL gentamicin; for Pseudomonas aeruginosa, 200 μg/mL carbenicillin, 75 μg/mL gentamicin, and 200 μg/mL tetracycline.

Construction of a PA14 strain lacking trmA

A PA14 strain lacking PA14_62450 or trmA (ΔtrmA) was constructed using a clean deletion technique. Upstream and downstream DNA fragments of trmA were amplified by PCR using primers BT7325 (5’-GGCGCTCGAGGCCGGTATCGGCCTGT-3’) BT7326 (5’-CGTACATATGCTTCTCGTTGCCGGTT-3’) and BT7327 (5’-GGCATGGATCCGGACACCTGC-3’) BT7328 (5’-GACCGAGCACGTGGCCTTCAC-3’), respectively. The downstream fragment was cloned into pKNOCK-Ap-loxPGmloxP vector digested with Ecl136II. Then, the upstream fragment was cloned into pKNOCK-Ap-loxPGmloxP containing the downstream fragment digested with XhoI and NdeI. The recombinant plasmid was delivered to P. aeruginosa PA14 by conjugation with E. coli BW20767. The resulting strain contains the insertion of gentamycin resistance cassette into the trmA gene. pCM157 expressing Cre recombinase enzyme was transferred into the PA14 ΔtrmA::Gm to remove gentamicin resistance cassette, yielding PA14 with clean deletion of trmA (PA14ΔtrmA) harboring pCM157. Then, the strain was sub-cultured in LB to lose the pCM157 plasmid. The trmA mutant was verified using PCR and Southern blot analysis.

Construction of trmA complemented strain

The full length PA14_62450 gene was amplified using primers BT5008 (5’-CAACCCATGGGCCGTCCACAGTTCG-3’) and BT5009 (5’-CACTCTCGAGGCGGCGTTCGAGCA-3’) and cloned into SmaI-digested broad host range vector, pUC18mini-Tn7T-Gm. This plasmid was co-transformed with helper plasmid, pTNS2, into the trmA mutant by electroporation. pTNS2 plasmid contains genes for site-specific transposition necessary for the insertion of the target fragment at a neutral site, downstream of glmS gene [63]. The trmA complemented strain was selected on LB agar plate containing 75 μg/mL gentamicin.

Expression and purification of the recombinant TrmA protein

Primer BT7308 (5’-ACTGGGAAGACATCACCGTC-3’) containing NcoI site and primer BT7309 (5’-ACGAGATCGATCCAGCCTAT-3’) containing XhoI site were used to amplify the full length trmA gene. The trmA fragment was digested with NcoI and XhoI and cloned into pETBlue-2 (Novagen) cut with the same restriction enzymes prior to transforming into E. coli DE3 (BL21) for protein expression. E. coli DE3 (BL21) harboring pETBlue-2-trmA-6X-histag was cultured in terrific broth containing 500 mM glycylglycine. The late log-phase cells were induced with 200 μM isopropyl-β-D-thiogalactopyranoside (IPTG) for 16-18 h at 37 °C. Cell pellet was resuspended in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM Imidazole, pH 8.0 with 1 mM PMSF and 1 mM DTT) followed by sonication. Lysate was collected by centrifugation. TrmA protein was purified using Ni-NTA agarose (Qiagen). Protein purity was determined using SDS-PAGE analysis. The purified recombinant TrmA was dialyzed against storage buffer (50 mM NaH₂PO₄, 300 mM NaCl, pH 8.0, 1 mM PMSF, 1 mM DTT, and 15% glycerol) and concentrated using 10 kDa filter column (Amicon). The concentrated-purified recombinant TrmA was aliquoted and stored at –20 °C. The purity of the recombinant purified TrmA is shown in Figure S9.

Extraction and purification of in vivo tRNA

Total RNA was extracted from exponential phase cells using Trizol reagent (Thermo Fisher Scientific). Then, the total RNA solution was precipitated using 35% ethanol to get rid of large RNAs. Small RNA species including tRNAs in aqueous solution was transferred into new tube and precipitated by adding absolute ethanol at −70 °C for 2 h. tRNAs were then purified using HPLC as described previously [33, 44].

Preparation of in vitro tRNA

tRNA was synthesized using MEGA Short Transcript Kit (Thermo Fisher Scientific). Synthesized single-stranded DNA containing a T7 promotor located at 5’ end of each tRNA species and the specific reverse primers listed in Thongdee et al. [33] and (Table S12) were used to amplify double-stranded DNA by Phusion DNA polymerase PCR. The double stranded DNA was then used as a template in the MEGA short transcript reactions according to manufacturer’s protocol to generate tRNA. The tRNA transcripts were further purified using HPLC as described previously [44].

Analysis of modified ribonucleosides by LC-QQQ

Modified ribonucleosides in RNA from P. aeruginosa strains were identified and quantitated using HPLC-coupled mass spectrometry as described previously [44, 64]. Briefly, total tRNA was incubated with benzonase nuclease (Sigma), phosphodiesterase, alkaline phosphatase (Invitrogen), deaminase inhibitors (5 μg/mL tetrahydrouridine and 20 nmol/μL erythron-9-(2-hydroxy-3-nonyl)adenine), and antioxidants (50 μM desferrioxamine and 50 μM butylatedhydroxytoluene) at 37 °C for 6 h. The reaction mixture was filtered using 10K MWCO column (Ambion) to remove proteins. Ribonucleosides were fractionated on a Water acquity UPLC BEH C18 column (130Å, 1.7 μm, 2.1 mm x 50 mm) using a flow rate of 0.35 mL/min with gradient of acetonitrile in 0.02% formic acid as follows: 0-2 min, 0%; 2-4 min, 0-5.7%; 4-5.9 min, 5.7-72%; 5.9-6 min, 72-100%; 6.7-9 min, 0%. The ribonucleosides were introduced into an electrospray ionization triple quadrupole mass spectrometer (Agilent 6495) operated in positive mode. The modified ribonucleosides were identified in comparison with chemical synthetic standards using HPLC retention time, CID fragmentation patterns, fragmentor voltages, and collision energies. Multiple reaction monitoring (MRM) was used to quantify the level of each modified ribonucleoside based on the retention time, the m/z of the transmitted parent ion, and the m/z of the monitored product, as noted in Table S5. The mass spectrometer signals were normalized against the total UV absorbance of adenine, guanosine, uridine, and cytidine using an in-line spectrophotometer. Raw LC-MS data for ribonucleosides are available from the Chorus Project database as accession number ID1849. Processed LC-MS/MS data are present in Supplementary Data S7.

In vitro methyltransferase assay

Methyltransferase activity was determined using MTase-Glo Bioluminescent Assay Kit (Promega). The reaction mixture contained 50 mM Tris-HCl (pH 8.0), 5 mM MgCl₂, 50 mM KCl, 50 μM SAM, 3 μg tRNA, 5 μM recombinant TrmA, and 1X MTase-Glo reagent. The reaction mixture was incubated at 37 °C for 40 min. MTase-Glo detection solution was added to detect light signal using microplate luminometer (Thermo Fisher Scientific) after incubation at ambient temperature for 30 min. The light signal is proportional to an amount of S-adenosyl homocysteine (SAH) catalyzed by TrmA methyltransferase activity. The reactions without TrmA for each tRNA species were used for background subtraction (Table S1).

Analysis of modified ribonucleoside position by HPLC-coupled QTOF

Methyltransferase reactions composing of 10 μg of synthesized tRNA transcript, 5μM recombinant TrmA protein, 50 mM KCl, 50 mM Tris-HCl pH 8.0, 50 μM purified SAM (Sigma), and RNase free water up to 80 μl were mixed and incubated at 37 °C for an hour. Five micrograms of the methylated in vitro tRNAs were then digested into fragments with RNaseT1 and dephosphorylated with bacterial alkaline phosphatase (Invitrogen) in 10 mM ammonium acetate buffer, pH 7.0 containing deaminase inhibitors and antioxidants as mentioned above at 37 °C for 4 h. The RNA fragments were purified by solid phase extraction using Strata^™ (Phenomenex).

The purified RNA fragments were analyzed by Agilent G6550B Q-TOF system using ACQUITY Premier BEH Amide column (2.1 mm ID x 150 mm, 1.7 μm particle size). The LC system was conducted at 50 °C and a flow rate of 0.3 ml/min, with a gradient starting with 90% solvent A (10 mM ammonium formate in 75% Acetonitrile) and 10% solvent B (10 mM ammonium formate in 25% Acetonitrile), followed by 0-20 min, 10%-80% B; 20-23 min, 80% B; 23-24 min, 80%-10% B. The LC-MS system used an electrospray ionization source in negative mode with the following parameters: gas temperature, 250 °C; gas flow, 15 L/min; nebulizer, 40 psi; sheath gas temperature, 250 °C; sheath gas flow, 12 L/min; capillary voltage, 3,500 V. The acquisition mode is data dependent MS/MS with the following parameters: MS1 range, 500-2,000 m/z; MS2 range, 100-2,000 m/z; MS/MS Scan Rate, 2; Max Precursors Per Cycle, 5; Collision energy, (1.6 x m/z of precursor mass)/100 + 25.

Transcriptional analysis

For RNA-seq library preparation, total RNA was extracted from the same samples as for proteomics analysis using Purelink RNA mini kit. Library preparation and sequencing were performed by MIT BioMicro Center (BioMicroCenter - OpenWetWare). Briefly, the quality and quantity of RNA were examined using fragment analysis. NEBNext^® rRNA depletion and NEBNext Ultra II Directional RNA library prep kits were used to remove ribosomal RNA and prepare the library, respectively. Sequencing was performed using MiSeq (Illumina) with 50 bp Single End sequencing.

Sequencing reads were then processed using a P. aeruginosa reference strain UCBPP-PA14 genome fasta and gtf files (GCF_000014625.1) downloaded from the Pseudomonas Genome Database website (http://www.pseudomonas.com). The genome was indexed followed by mapping using STAR [65]. HTSeq count script was used to obtain read counts. Reads per kilo base per million mapped reads (RPKM) was obtained by normalizing read counts with transcript length. Differential expression was evaluated using DESeq2 package [66]. Raw RNA-seq data are available from the GEO database as accession number GEO252633. Processed RNA-seq data are present in Supplementary Data S1.

Proteomics analysis

For protein extraction, digestion, and labeling, three cultures of PA14 wild-type strain and trmA mutant were grown independently in LB broth and harvested at mid-log phase by centrifugation at 500 xg for 10 min at 4 °C. The cell pellets were washed with cold PBS followed by resuspension with lysis buffer (20 mM sodium phosphate buffer pH 6.8, 10 mM DTT, 1 mM EDTA pH 8.0, 0.1% Tween, 1 mM PMSF, 1 U/μL benzonase, 1 mg/mL lysozyme, and Roche protease inhibitor). The cell suspension was incubated at 30 °C with shaking for 30 min. Cells were disrupted using tip sonication; 50% duty cycle, 8 pulsed with 30 s intervals for 10 times. Cell lysates were collected using centrifugation at 200 xg, for 15 min at 4 °C. Proteins were precipitated by 80% acetone/water for overnight at −20 °C. The protein pellet was centrifuged at 16,100 xg for 10 min at 4 °C, washed with cold 80% acetone/water and subsequently reconstituted in 10 mM triethylammonium bicarbonate (TEAB) pH 8.5. The Bicinchoninic acid (BCA) method was used to measure protein concentration.

Each sample (~10 μg) was treated with a proteomics-grade trypsin (Thermo-Scientific) to generate peptides. Peptides from each sample were then tagged with a unique isobaric marker using a TMT 6-plex kit (Thermo-Scientific). Thereafter, all samples were pooled, and then fractionated into eight tubes. For LC-MS/MS analysis, an EASY-nLC1000 was used interfaced to a QExactive Orbitrap mass spectrometer (Thermo-Scientific). Each fraction of the TMT-labeled peptides was loaded onto an Acclaim PepMap 100 pre-column (Thermo-Scientific #164946) to concentrate and wash, followed by transfer to a custom-made analytical capillary column (120 mm, New Objective #PF360-50-10-N-5) packed with C18 beads (YMC gel, ODS-AQ, 12nm, S-5μm, AQ12S05) [67]. Solvent A (0.1% formic acid) was delivered using a flow rate of 300 nL/min with a gradient of solvent B (0.1% formic acid in acetonitrile) as follows: 0-5 min, 2-5%; 5-105 min, 5-25%; 105-125 min, 25-37%; 125-129 min, 37-50%; 129-130 min, 50-95%; 130-146 min, 95%. MS data was acquired in positive ion mode using the following parameters for full-scan MS: a range from 350–2,000 m/z at a resolution of 70,000, automatic gain control (AGC) set to 3e6, and a maximum injection time (IT) of 60 ms. The full MS scan was followed by MS/MS for the top 10 precursor ions in each cycle with a normalized collision energy (NCE) of 34 eV and dynamic exclusion of 30 s.

MS/MS spectra were searched against 5,886 P. aeruginosa PA14 protein coding sequences deposited in the Uniprot database and embedded into Proteome Discoverer using Sequest HT engine with trypsin as enzyme name. Parameters used in the software were as follows: peptide mass tolerance, 10 ppm; fragment mass tolerance, 0.02 Da; and max missed cleavages, 2. Carbamidomethylation of cysteine and TMT6plex of lysine and N-terminus were specified as static modifications. Oxidation, acetyl of the N-terminus, and methionine loss at N-terminus were specified as dynamic modifications. False discovery rate was set at <0.01. Total sum abundance value for each channel was normalized by total peptide amount. The channel with the highest total abundance was used as a reference. All abundance values were corrected in the other channels by a constant factor per channel. Log₂ fold-change (trmA mutant:PA14 wild type strain) <−0.263, >0.263 with P<0.05 were used for downregulated and upregulated proteins, respectively. Gene Ontology enrichment analysis of differentially expressed proteins was performed based on DAVID. Raw proteomics data are available from the PRIDE ProteomeXchange database as accession number PXD048253. Processed proteomics data are present in Supplementary Data S2.

Quantitative analysis of tRNA landscapes

Absolute quantification RNA sequencing (AQRNA-seq) [30] was performed to quantify tRNA molecules in both the wild-type and trmA mutant strains. Sequencing libraries were constructed following a revised version of the AQRNA-seq protocol in which an alternative strategy is used for mitigating the carry-over of primer dimers into the constructed cDNA libraries [68]. Specifically, after the reverse transcription (RT) reaction, exonuclease I is employed to digest excess RT primers, thereby inhibiting the formation of primer dimers (i.e., ligation of DNA linkers to the 3’ end of RT primers) from the first place. As a result, the gel purification step is eliminated in the revised protocol. The constructed libraries were submitted to the MIT BioMicro Center for quality assessments using fragment analysis and real-time PCR, followed by 100-bp Paired-End sequencing on an Illumina MiSeq platform with the v2 reagent kit. The raw sequence reads were subsequently processed to obtain raw tRNA abundance data using an in-house data analytical pipeline.

To identify tRNA species that were significantly upregulated or downregulated in the trmA mutant strain as compared to the wild-type strain, differential expression analysis was performed using the DESeq2 v 1.36.0 package [66] in R Statistical Programming Environment v 4.2.1 (R) [69]. Briefly, the workflow includes the following steps: (i) normalize the raw tRNA abundances using the median of ratios method; (ii) estimate dispersion parameters for each tRNA using empirical Bayes shrinkage; (iii) fit the abundance data for each tRNA with a negative binomial regression model; (iv) estimate logarithmic fold changes using Approximate Posterior Estimation for generalized linear model [70]; and (v) perform Wald test for significant testing (with the Benjamini-Hochberg method for multiple correction). Raw AQRNA-seq data are available from the GEO database as accession number GEO252634. Processed AQRNA-seq data are present in Supplementary Data S3.

Codon usage analysis

The Gene-specific Codon Utilization Tool [34] was used to identify frequencies of codon usage, relative to a genome average, across the P. aeruginosa PA14 genome from NCBI database. Codon analytics are detailed in Supplementary Data S4. Codon usage patterns in genes encoding upregulated and downregulated proteins in the trmA mutant relative to wild-type strain were assessed by PCA and PLSR (UnscramblerX version 10.4.1).

Macrophage cell culture and in vitro bacterial infection

Raw264.7 macrophage cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, and 2 mM L-Glutamine at 37 °C/5% CO₂. Cells (2 x 10⁶) were seeded in a 6-well plate and stimulated with 1 μg/mL LPS from S. enterica serotype typhimulium (Sigma) for 18 h. The medium was replaced by antibiotic-free DMEM supplemented with 10% FBS and 2 mM L-glutamine. The activated macrophages were infected with the mid-log phase of P. aeruginosa at a MOI of 50. The infected cells were incubated at 37 °C in 5% CO₂. After 2 hours, cell culture supernatants were collected. The supernatants were assayed for IL-1β according to manufacturer’s protocol (BiolLegend).

Phenotypic microarray analysis

P. aeruginosa wild-type and trmA mutant strains were tested against various types of chemicals using Biolog Phenotypic Microarrays number 11-20 (Biolog Inc., USA) using the manufacturer’s protocol. The tested chemicals are listed in Table S4. The bacterial strains were grown overnight on LB agar plate at 37 °C. The cultures were swabbed from the plate and suspended in 1x IF-0 inoculating fluid. The cell suspension was adjusted to transmittance of 85%. The 85%T cell suspension was diluted at 1:200 with 1x IF-10 and dye mix A (Biolog). Finally, 100 μL of the cell suspension was inoculated into each well of PM11-20 microplates. Reduction of tetrazolium dye was kinetically measured every 15 min using Omnilog reader (Biolog). The results were analyzed using Omnilog Phenotype Microarray Software (OL_OM_12 release 1.20.02). Area under respiration curve was used for statistical analysis.

Growth curve analysis

An overnight cell culture was diluted in Mueller-Hinton Broth (MHB) to a turbidity of 0.5 McFarland standard. The culture was then diluted by a factor of 1:20 with 0.9% sterile normal saline. The diluted cultures were used to inoculate each well of a 96-well plate containing different concentrations of the test chemical solution prepared in MHB. Cultures without inoculated cells were used as a negative control, while inoculated cells in MHB without any test chemical were used as a positive control. The 96-well plate was incubated at 37 °C. The turbidity in each well was kinetically measured at OD_600nm for 24 h using a microplate spectrophotometer. Gen5 software was used to control the device according to the manufacturer’s guide.

Detection of intracellular ROS

Flow cytometry was performed to measure the level of ROS in P. aeruginosa wild-type, trmA mutant, and trmA-complemented strains. Four biological replicates of each strain were sub-cultured in LB to OD₆₀₀ 0.4-0.6. For antibiotic treatment, cells were diluted in MHB to an OD 0.1. A hundred microliter of the diluted cells was added to 96-well plate containing 100 μL of two-fold higher than the final concentration of the antibiotic. After treatment, 20 μL of the treated cells were stained by adding 180 μL of PBS containing 10 μM 2’,7’-dichlorofluorescin diacetate (Sigma, USA). H₂O₂-treated cells (1 mM) were used as positive control. For H₂O₂ treatment, cells were washed once with PBS pH 7.4 and diluted to OD₆₀₀ 0.1 in PBS. Cells with the final OD₆₀₀ of 0.01 were incubated with the fluorescence reporter dyes during the treatment for 30 min at 37 °C in the dark. Samples were analyzed on BD Accuri C6 Plus Flow cytometer using CSampler plus (BD Biosciences), with 488 blue argon laser, and emission filter of 530/30 nm for DCF. For each sample, total of 50,000–100,000 events were collected and analyzed using BD CSampler Plus software. Unstained cells were used to gate population of bacteria using forward-scatter (FSC) and size-scatter (SSC) of light.

Highlights.

• In P. aeruginosa, TrmA catalyzes m⁵U formation at position 54 on all tRNAs.

• The loss of TrmA increases resistance to polymyxin B, a last-resort drug.

• TrmA regulates the translation of T3SS via a codon-biased mechanism.

• This study reveals the expression profile of tRNA genes in P. aeruginosa.

Data availability

Proteomics data were deposited in PRIDE ProteomeXchange with accession number PXD048253. Transcriptional profiling data and AQRNA-seq data were deposited in Gene Expression Omnibus with accession numbers GEO252633 and GEO 252634, respectively. Mass spectrometry data for ribonucleoside analyses were deposited in Chorus with accession number ID1849.