Quantitative RNA pseudouridine landscape reveals dynamic modification patterns and evolutionary conservation across bacterial species.

Quantitative RNA pseudouridine landscape reveals dynamic modification patterns and evolutionary conservation across bacterial species - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice eLife . 2026 Jun 4;14:RP107545. doi: 10.7554/eLife.107545 Quantitative RNA pseudouridine landscape reveals dynamic modification patterns and evolutionary conservation across bacterial species Letong Xu 1, † , Shenghai Shen 2, † , Yizhou Zhang 1 , Zhihao Guo 3 , Beifang Lu 1 , Jiadai Huang 1 , Runsheng Li 4, 5 , Yitong Shen 2 , Li-Sheng Zhang 2, 6, ✉ , Xin Deng 1, 3, 4, ✉ Editors: Camilo Perez 7 , Bavesh D Kana 8 Author information Article notes Copyright and License information 1 Department of Biomedical Sciences, City University of Hong Kong, Hong Kong, China 2 Division of Life Science, The Hong Kong University of Science and Technology, Hong Kong, China 3 Shenzhen Research Institute, City University of Hong Kong, Shenzhen, China 4 Tung Biomedical Sciences Center, City University of Hong Kong, Hong Kong, China 5 Department of Infectious Diseases and Public Health, City University of Hong Kong, Hong Kong, China 6 Department of Chemistry, The Hong Kong University of Science and Technology, Hong Kong, China 7 University of Georgia, United States 8 University of the Witwatersrand, South Africa † These authors contributed equally to this work. ✉ Corresponding author. Roles Camilo Perez : Reviewing Editor Bavesh D Kana : Senior Editor Collection date 2026. © 2025, Xu, Shen et al This article is distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use and redistribution provided that the original author and source are credited. PMC Copyright notice PMCID: PMC13236023  PMID: 42241317 Previous version available: This article is based on a previously available preprint with doi: https://doi.org/10.1101/2025.05.13.653745 . Previous version available: This article is based on a previously available preprint with doi: https://doi.org/10.7554/eLife.107545.1 . Previous version available: This article is based on a previously available preprint with doi: https://doi.org/10.7554/eLife.107545.2 . Abstract Pseudouridine (Ψ) modifications are the most abundant RNA modifications; however, their distribution and functional significance in bacteria remain largely unexplored compared to eukaryotic systems. In this study, we present the first transcriptome-wide and quantitative mapping of Ψ modifications across five diverse bacterial species ( Bacillus cereus , Escherichia coli , Klebsiella pneumoniae , Pseudomonas aeruginosa , and Pseudomonas syringae ) at single-base resolution, utilizing the optimized baBID-seq method for bacterial RNA. Our analysis revealed growth phase-dependent dynamics of pseudouridylation in bacterial tRNA and mRNA, particularly in genes enriched in core metabolic pathways. Comparative analysis demonstrated evolutionarily conserved features of Ψ modifications, such as dominant motif contexts, Ψ clustering within operons, etc. Functional analysis indicated Ψ modifications affect bacterial mRNA stability, translation, and interactions with specific RNA-binding proteins in response to changing cellular demands during growth phase transitions. The integrated computational analysis on local RNA architecture was conducted to elucidate the structure-dependent Ψ modifications in bacterial RNA. Furthermore, we developed an integrated deep learning framework, combining LSTM-transformer-GNN-based neural networks (pseU_NN) to capture both RNA sequence and local structure features for effective prediction of Ψ-modified sites. Overall, our study provides valuable insights into the landscapes of bacterial RNA Ψ modifications and establishes a foundation for future mechanistic investigations on bacterial Ψ functions. Research organism: Other Introduction RNA modifications are a crucial layer of post-transcriptional regulation in biological systems, with approximately 170 distinct chemical modifications identified to date ( Roundtree et al., 2017 ). Pseudouridine (Ψ), often referred to as the ‘fifth nucleoside’, is one of the most prevalent and evolutionarily conserved RNA modifications ( Cerneckis et al., 2022 ; Rodell et al., 2024 ). This abundant modification arises from a specific isomerization process in which uridine undergoes a site-specific intramolecular rearrangement ( Yu and Allen, 1959 ). Ψ can thermodynamically stabilize RNA structures by enhancing base stacking and increasing the rigidity of the sugar-phosphate backbone, which helps maintain the structural folding of functional RNAs such as tRNA and rRNA ( Pan et al., 2003 ; Roovers et al., 2006 ). For example, Ψ at the 55th position of tRNA, a universally conserved modification in both eukaryotes and prokaryotes, is crucial for regulating tRNA stability and aminoacylation ( Ishida et al., 2011 ; Schultz et al., 2024 ). Ψ modifications in rRNA play important roles in rRNA biogenesis and function, as well as in mRNA translation, for both mammals and bacteria ( Leppik et al., 2017 ; Sloan et al., 2017 ; Zhao et al., 2023 ). The regulation of mRNA translation could be highly complicated, and mRNA stability remarkably impacts gene expression in mammals. Ψ-modified mRNAs demonstrate enhanced stability due to their resistance to RNase L-mediated degradation ( Anderson et al., 2011 ). Pre-mRNA is found to be pseudouridylated co-transcriptionally, with Ψ enriched near alternative splicing regions and RNA-binding protein (RBP) binding sites ( Martinez et al., 2022 ). Moreover, Ψ located within exon regions can alter codon properties to modulate translation, while Ψ modifications at stop codons promote ribosomal readthrough ( Hoernes et al., 2016 ; Karijolich and Yu, 2011 ). Ψ also facilitates the low-level synthesis of peptide products from individual mRNA sequences in human cells and increases the rate at which near-cognate tRNA Val interacts with ΨUU codons ( Eyler et al., 2019 ), suggesting a more complex regulatory role for Ψ in translation. Until recently, comprehensive investigations of bacterial RNA pseudouridylation have been limited due to technical challenges in precisely mapping and quantifying Ψ modifications at single-nucleotide resolution. While eukaryotic mRNA can be easily isolated and assessed using polyA + enrichment methods, bacterial transcripts lack polyA tails and are predominantly composed of ribosomal RNA, which accounts for over 95% of total RNA ( Liang et al., 2000 ). This methodological gap has hindered the functional exploration of Ψ roles in bacterial RNA. A previously reported CMC-based method can selectively label Ψ sites and induce truncation signatures during reverse transcription (RT). However, this technical strategy could exhibit several drawbacks, including relatively low sensitivity and limitations in quantifying Ψ modification levels. Recently, a new technique named Bisulfite-Induced Deletion sequencing (BID-seq) has emerged, utilizing unique deletion signatures induced at Ψ-modified sites to achieve base-resolution and quantitative characterization of Ψ sites transcriptome-wide ( Dai et al., 2023 ; Zhang et al., 2024 ). To address these challenges and enable future functional study of bacterial Ψ modifications, we developed an optimized BID-seq method for bacterial RNA, termed baBID-seq. We selected Klebsiella pneumoniae , Bacillus cereus , Pseudomonas aeruginosa , and Pseudomonas syringae based on their biological relevance and taxonomic diversity. K. pneumoniae , B. cereus , and P. aeruginosa are clinically important human pathogens responsible for many infectious diseases, yet transcriptome-wide pseudouridylation has not been systematically characterized in these organisms ( Ehling-Schulz et al., 2019 ; Kerr and Snelling, 2009 ; Wyres et al., 2020 ). P. syringae , a well-studied plant pathogen, was included to extend the analysis beyond human pathogens and to explore pseudouridine modification in a distinct ecological context ( Xin et al., 2018 ). Collectively, these species encompass both Gram-positive ( B. cereus ) and Gram-negative ( K. pneumoniae , P. aeruginosa , and P. syringae ) bacteria and exhibit substantial differences in genome size, GC content, and pathogenic lifestyle. This selection provides a comparative framework for investigating conserved and species-specific features of bacterial pseudouridylation across diverse lineages. By combining efficient rRNA depletion in baBID-seq, we expanded our quantitative Ψ analysis to these representative bacterial species, examining both exponential and stationary growth phases. Our analysis revealed the landscape of Ψ modifications on bacterial rRNA, tRNA, and mRNA, highlighting evolutionarily conserved Ψ features across bacterial strains. We investigated the sequence and structural properties of local mRNA regions that influence Ψ deposition. Dynamic Ψ modifications at specific sites in tRNA and mRNA were observed, showing distinct accumulation patterns during the stationary growth phase. In P. syringae , we explored the roles of Ψ under nutrient-deficient conditions and found a positive correlation between mRNA translation efficiency (TE) and Ψ intensity ( Hua et al., 2024 ). In P. aeruginosa , our data suggested the potential regulatory functions of Ψ in promoting mRNA interactions with the Hfq chaperone ( Trouillon et al., 2022 ). Furthermore, we employed a hybrid LSTM-attention-based graph neural network (GNN) classification approach, integrating RNA sequence and local structure features to predict Ψ modification sites. Collectively, our analysis revealed a dynamic landscape of Ψ modifications, uncovering their evolutionarily conserved features, alongside key motif contexts and structural elements that impact Ψ installation on bacterial RNA. Results baBID-seq quantitatively maps Ψ modification in bacterial rRNA, tRNA, and mRNA To investigate Ψ modifications in bacteria, we primarily applied the standard BID-seq protocol ( Zhang et al., 2024 ) to total RNA isolated from E. coli and P. aeruginosa during exponential and stationary growth phases. Ψ sites on rRNA were identified through deletion signatures at single-base resolution, and the observed deletion ratios were utilized to assess Ψ modification stoichiometry. By characterizing Ψ sites with significantly higher deletion ratios in ‘BID-seq treated’ samples compared to ‘Input’ samples, we detected 9 out of 10 known Ψ sites and 9 conserved Ψ sites, in 23S and 16S rRNA from E. coli and P. aeruginosa , respectively ( Figure 1a, b ). For example, Ψ781 in 16S rRNA of P. aeruginosa exhibited a distinct deletion signature in ‘BID-seq treated’ samples versus the input ( Figure 1—figure supplement 1c ). Although Ψ is known to regulate rRNA local structure and biogenesis ( Leppik et al., 2017 ), bacterial RNA BID-seq (baBID-seq) pointed out that the number of Ψ-modified sites is notably lower than the typical >100 Ψ sites found in mammalian rRNA ( Dai et al., 2023 ). We then compared the rRNA Ψ fraction in E. coli and P. aeruginosa across the two growth stages, and almost all Ψ sites are highly conserved, with only slight variations in deletion ratios at Ψ sites ( Figure 1a, b ). Figure 1. BID-seq identifies precise pseudouridine (Ψ) modification sites in ribosomal RNA and reveals dynamic Ψ modification patterns in transfer RNA. ( a, b ) Ψ modifications detected on 16S and 23S ribosomal RNA with baBID-seq of E. coli ( a ) and P. aeruginosa ( b ) total RNA during exponential and stationary growth phases. All Ψ sites in panels ( a ) and ( b ) were identified using filtration criteria of deletion fraction >0.02 and p-value ≤1 × 10 −4 ( c ) Pearson correlation analysis of Ψ modification fractions at individual sites between biological replicates. ( d ) Ψ modification fraction alteration pattern observed on specific sites in K. pneumoniae different tRNA regions. The left panel depicts the general tRNA secondary structure in K. pneumoniae . Structural regions are indicated in the legend and highlighted with corresponding colors. The right scatter plot’s y -axis depicts the Ψ fraction difference, calculated as (exponential phase Ψ fraction) − (stationary phase Ψ fraction). ( e ) Heatmap displaying the tRNA Ψ fraction alteration features detected across different conditions. Color represents the average Ψ fraction values (ranging from 2% to 100%) at specific sites within tRNA isoacceptors for each strain. The tRNA tags (labeled in y -axis) comprise the amino acids transferred by each tRNA, the corresponding anticodons, and the Ψ position on the tRNA molecules. ( f ) Venn plot shows the overlap of detected Ψ sites between biological replicates for P. aeruginosa and P. syringae during each growth phase. Figure 1—figure supplement 1. baBID-seq workflow and detection of Ψ sites on 16S rRNA. ( a ) Schematic overview of the baBID-seq workflow optimized for bacterial RNA, adapted from the original BID-seq protocol. ( b ) Distribution of detected Ψ sites on 16S rRNA across different bacterial strains, with the y -axis showing the deletion ratios in input and bisulfite-treated samples. ( c ) Integrative Genomics Viewer (IGV) snapshot illustrating the detected Ψ781 site in P. aeruginosa 16S rRNA. Figure 1—figure supplement 2. Reproducibility of Ψ detection and dynamic Ψ modification in tRNA. ( a ) Pearson correlation of Ψ modification fractions between two independent biological replicates for each strain. Changes in tRNA Ψ fractions in B. cereus ( b ) and P. syringae ( c ) between exponential and stationary growth phases. The y -axis represents the Ψ fraction difference (exponential phase minus stationary phase). ( d ) Overlap of identified mRNA Ψ sites between biological replicates for each strain. ( e ) Experimental validation of Ψ sites in P. aeruginosa PAO1 using the pseU-TRACE method. From right to left, the panels show validation of the Ψ site at position 944 on 23S rRNA, a negative control site from the guaA gene, a Ψ site in clpV1 (chromosomal position 109,883), and a Ψ site located in the intergenic region between guaA and guaB (chromosomal position 4,227,284). Error bars represent the mean ± SD from two independent biological replicates. Statistical significance was determined using Student's t-test. Open in a new tab To study Ψ modifications on bacterial RNA species beyond rRNA, we incorporated probe-based rRNA depletion ( Choe et al., 2021 ) into our optimized protocol for baBID-seq. We carefully established fragmentation conditions to generate RNA fragments of ~60–70 nt in length, ensuring adaptor ligation efficiency; meanwhile, size selection of the amplified library by PAGE gel minimized contamination from adaptor dimers or DNA of unexpected sizes ( Figure 1—figure supplement 1a ). We then applied the baBID-seq protocol to four bacterial species ( K. pneumoniae , B. cereus , P. aeruginosa , and P. syringae ). Due to rRNA depletion, rRNA-derived reads may not be suitable for comprehensive rRNA Ψ profiling. However, for benchmarking baBID-seq library quality, certain 16S rRNA Ψ sites with stable modification fractions can be used as internal indicators of baBID-seq performance ( Figure 1—figure supplement 1b ). baBID-seq further successfully captured various RNA species, including rRNA, tRNA, and mRNA. We characterized hundreds of Ψ sites on rRNA-depleted RNA isolated from four bacterial strains across exponential and stationary growth phases. The results from baBID-seq quantitatively demonstrated a strong correlation of deletion ratios at all detected Ψ sites, between biological replicates ( Figure 1c , Figure 1—figure supplement 2a ). Among these, while Ψ sites on rRNA and tRNA consistently showed stable modification levels across biological replicates, mRNA Ψ sites displayed greater variability. Given the conservation of tRNA Ψ sites across biological replicates, we used the average Ψ fraction at each specific tRNA Ψ site for downstream analysis. baBID-seq quantitatively maps Ψ modifications at various positions within tRNA, including the stem and loop of the T-arm, anticodon arm, and D-arm ( Figure 1d and Figure 1—figure supplement 2b, c ). To investigate tRNA Ψ dynamics during exponential versus stationary growth phases, we quantified Ψ fraction differences at each specific site across four strains. Most Ψ sites on bacterial tRNA consistently exhibited higher modification fractions in stationary phase, across all examined strains ( Figure 1e ). In K. pneumoniae , the Ψ sites within the T-arm, D-arm, and anticodon arm concordantly showed a reduced modification fraction in exponential phase, compared to stationary phase ( Figure 1d ). Similarly, in B. cereus and P. syringae , most tRNA Ψ sites within the T-arm displayed lower Ψ fractions under exponential phase ( Figure 1—figure supplement 2b, c ). Previous research has shown that the T-arm globally influences tRNA maturation and regulates translation in E. coli ( Schultz et al., 2024 ). Thus, this growth phase-dependent pattern of tRNA pseudouridylation suggests a coordinated regulatory mechanism that may fine-tune mRNA translation as bacteria adapt to changing environmental conditions. In addition to rRNA and tRNA, baBID-seq also identified highly conserved Ψ sites on mRNA, between biological replicates ( Figure 1f , Figure 1—figure supplement 2d ), providing strong evidence in identifying genuine Ψ modifications. To further verify site detection reliability, four Ψ sites were tested with pseU-TRACE ( Fang et al., 2024 ): a site at position 944 on 23S rRNA, a site within the clpV1 gene, an intergenic site located between guaA and guaB genes in P. aeruginosa , as well as a negative control site located within the guaA gene. All three positive sites were successfully detected by pseU-TRACE, while no signal was observed for the negative control ( Figure 1—figure supplement 2e ). For subsequent analysis, we focused exclusively on mRNA Ψ sites that were consistently detected across biological replicates. BID-seq profiles abundant Ψ modifications in bacterial mRNA With the identification of highly conserved Ψ modifications in bacterial mRNA enabled by baBID-seq, we proceeded to analyze their distribution patterns and quantitative features transcriptome-wide. In total, we detected over 3000 Ψ sites in the mRNA of four bacterial strains. Notably, the metagene plot revealed that most Ψ sites were enriched within the coding sequences (CDS), exhibiting a remarkable consistency across strains ( Figure 2a, b ) and a similar pattern to the observations in mammals ( Dai et al., 2023 ). Overall, the average Ψ modification in bacterial mRNA (mean Ψ fraction: 15%) was lower than that in 16S and 23S rRNA (mean Ψ fraction: 40%). Most mRNA Ψ sites were primarily distributed below a 25% Ψ fraction, while a smaller proportion reached fractions above 50% as highly modified sites ( Figure 2c , Figure 2—figure supplement 1a ). Figure 2. baBID-seq uncovers Ψ modification in bacterial mRNA CDS and untranslated regions (UTRs). ( a ) Density plot depicting the distribution of Ψ modifications in mRNA across different growth phases and conditions. ( b ) Distribution of mRNA Ψ fraction showing strain and growth phase-specific patterns of Ψ distribution (right Ψ density plot across each strain’s mRNA). ( c ) mRNA Ψ fraction and counts under different conditions. ( d ) Right pie charts show the proportion of Ψ sites in UTRs versus coding regions, and violin plots compare Ψ fraction values between UTRs (upstream and downstream UTRs) and coding regions. Statistical significance was determined using the Wilcoxon signed-rank test; ns, p-value ≥0.05; *p-value <0.05; **p-value <0.01; ***p-value <0.001; and ****p-value <0.0001. ( e ) The pie charts show the Ψ-modified gene overlap in two growth phases across four strains. ( f ) Density plot shows the Ψ site numbers per transcript. Figure 2—figure supplement 1. Distribution and fraction of mRNA Ψ modifications across transcript regions. ( a ) Distribution of mRNA Ψ fraction values in K. pneumoniae and P. syringae under minimal medium (MM) conditions. ( b ) Distribution of Ψ modifications between untranslated regions (UTRs) and coding sequences (CDSs) during the stationary growth phase and ( c ) exponential growth phase. ( d ) Distribution of Ψ fraction values in upstream and downstream UTRs of mRNAs across exponential and stationary growth phases. Open in a new tab Since Ψ modifications in mRNA untranslated regions (UTR) affect mRNA processing in eukaryotic cells ( Martinez et al., 2022 ; Rodell et al., 2024 ), and UTR plays functional roles in post-translational regulation in bacteria ( Adams et al., 2021 ), we conducted a detailed investigation of Ψ modifications in the upstream and downstream UTRs. UTR pseudouridylation accounted for 7.0–16.4% of total Ψ sites across the transcriptome, in both stationary and exponential phases ( Figure 2—figure supplement 1b, c ). Notably, in P. aeruginosa and B. cereus , we observed a significantly higher modification fraction for Ψ sites within downstream regions compared to CDS, in both growth phases ( Figure 2d , Figure 2—figure supplement 1d ). In contrast, P. aeruginosa exhibited a higher Ψ modification level in the upstream regions of mRNA during the exponential phase ( Figure 2d ). We defined a Ψ-modified gene as any mRNA containing one or more Ψ sites in any growth phase and identified both phase-shared and phase-unique genes across the four strains ( Figure 2e ). We also noted that individual genes frequently harbored multiple Ψ sites, with the number of sites varying dynamically across bacterial growth phases ( Figure 2f ). Motif contexts surrounding Ψ modifications in bacterial mRNA, rRNA, and tRNA Previous research has shown that different RNA species exhibit unique Ψ modification patterns, with Ψ synthases targeting specific sequence motifs in tRNA and rRNA (such as the RluA motif ΨURAA) ( Pan et al., 2003 ; Schaening-Burgos et al., 2024 ). To identify the sequence determinants of Ψ modifications and uncover RNA-type-specific Ψ motif contexts across bacterial transcriptomes, we conducted a motif analysis of Ψ-modified sites with a fraction above 2%, which were confidently identified in either growth phase. The sequence context analysis focused on 5-nucleotide motifs centered at each Ψ site. We calculated the frequency of Ψ motifs through comparing the count of each unique Ψ motif versus all Ψ motifs detected in mRNA, for a single bacterial strain ( Figure 3a ). Figure 3. Comparative analysis of Ψ modification motif across strains. ( a ) Comparison of overall mRNA 5-mer Ψ motif ratios across four strains. Motif ratios are calculated by dividing the count of each specific 5-mer motif centered on Ψ by the total number of motifs detected in each individual strain mRNA. Ψ-modified sites with a fraction above 2% are used here. ( b ) Distribution of Ψ fraction (ranging from 2% to 100%) for each motif detected. ( c ) Scatter pie chart shows the proportional distribution of top 10 (ranked by motif abundance) Ψ-containing motif counts categorized by RNA types and bacterial strains. ( d ) The scatter plot illustrates the relationship between the average modification fraction and abundance of motifs in P. aeruginosa in exponential ( P. aeruginosa _exp) and stationary ( P. aeruginosa _stat) growth phases. The average Ψ fraction was calculated as the sum of Ψ fractions for each individual motif divided by its frequency. Figure 3—figure supplement 1. Motif pattern of Ψ modification. ( a ) The Ψ motif ratio in tRNA across four strains. Motif ratios are calculated by dividing the count of each specific 5-mer motif centered on Ψ by the total number of motifs detected in each individual strain tRNA. ( b ) The Ψ motif distribution ratio in ribosomal RNA (rRNA) across four strains. Motif ratios are calculated by dividing the count of each specific 5-mer motif centered on Ψ by the total number of motifs detected in each individual strain rRNA. ( c ) Scatter plot displaying the average Ψ fraction values for different sequence motifs in ribosomal RNA. Each point represents a specific sequence motif, with its position on the y -axis indicating the average Ψ fraction value across all instances of that motif in rRNA. ( d ) Similar to ( c ), each point represents a specific motif, with its position on the y -axis indicating the average Ψ fraction value across all instances of that motif in tRNA. ( e–h ) Pseudouridylation motifs identified on mRNA transcripts from four bacterial strains using MEME. ( i ) Density plot displaying GC composition within 10-base pairs flanking Ψ sites in mRNA. ( j ) GC composition within 10-base pairs flanking Ψ sites in mRNA normalized to the background GC ratio. ( k ) Box plot comparing the GC ratio (normalized to genome background) around Ψ sites across different bacterial strains. ( l ) Hierarchical clustering plot showing gene orthology analysis of all pseudouridine synthases across bacterial strains. The color intensity represents clustered gene counts across genomes. Open in a new tab baBID-seq analysis revealed diverse Ψ motifs within bacterial mRNA. Notably, GCΨCG, GGΨCG, and CCΨCG were the most abundant motifs observed in the three Gram-negative bacteria, while GUΨGU and GGΨGU were the dominant motifs in B. cereus ( Figure 3a ). For quantitative features of Ψ sites within diverse motif contexts, the average Ψ fractions for different motifs ranged from 3.4% to 96.6%, indicating varying Ψ installation efficiencies in the presence of different Ψ synthases ( Figure 3b ). Overall, we summarized the top 10 frequent Ψ motifs for bacterial mRNA of P. aeruginosa and P. syringae : GUΨCG, (CC/CU/GC/GG/UC)ΨCG, (CU/GC/UC)ΨCC, and GCΨGG ( Figure 3c ). baBID-seq also reveals Ψ motif contexts in bacterial rRNA and tRNA. While a variety of Ψ motifs were identified in bacterial rRNA, GUΨCG motif stands out as the predominant one in tRNA across all tested strains ( Figure 3—figure supplement 1a–d ). Notably, GUΨCG motif is well-characterized within T-arm of tRNAs (at position 55) and specifically modified by TruB family ( Dai et al., 2023 ; de Crécy-Lagard et al., 2019 ; Hoang and Ferré-D’Amaré, 2001 ; Pan et al., 2003 ; Schultz et al., 2024 ; Veerareddygari et al., 2016 ). According to baBID-seq data, GUΨCG has been confirmed as the predominant motif in both tRNA and rRNA (mean fraction: 68%), as well as in mRNA (mean fraction: 19%) ( Figure 3c , Figure 3—figure supplement 1c, d ), suggesting a broad role of TruB across bacterial RNA species. Several other key motifs, including UUGC, UUGA, and UUAAA, correspond to the previously characterized RluA motif in E. coli ( Schaening-Burgos et al., 2024 ). Overall, no distinct sequence motifs were universally enriched in bacterial mRNA across strains ( Figure 3—figure supplement 1e–h ), likely due to the complex interactions among multiple Ψ synthases involved in U-to-Ψ conversion. To determine sequence preferences of Ψ modifications under varying growth conditions, we calculated both motif frequency and average Ψ fraction for each Ψ motif in P. aeruginosa . Notably, we observed a decrease in GCΨCG motif frequency ( Figure 3d ), as the most abundant Ψ motif among the three Gram-negative bacteria. For other top Ψ motifs, the associated modification levels showcased slight variations between the exponential and stationary phases. To gain insights into the sequence context features around Ψ modifications, we analyzed the nucleotide composition within a 10-nt window flanking Ψ sites on bacterial mRNA. While most strains exhibited non-unique differences in GC content, B. cereus displayed a notably higher GC ratio (normalized to the genomic background GC content), compared to the other three strains ( Figure 3—figure supplement 1i–k ). To decipher this distinct pattern in B. cereus , we conducted comparative orthology-based analyses for pseudouridine synthases among five bacterial strains ( Figure 3—figure supplement 1l ). A unique interactive pattern of pseudouridine synthases in B. cereus may explain its divergent sequence context pattern nearby Ψ sites, compared to other strains. Evolutionary conservation of clustered Ψ modifications in bacterial orthologous genes To investigate the evolutionary conservation of mRNA Ψ modifications among bacterial strains, we analyzed orthologous genes, focusing on both the preservation of modification sites and the functional characteristics of Ψ-modified genes. Based on clustered Ortho groups and genome annotations, our results revealed 225 homologous genes carrying Ψ modifications in at least two bacterial strains ( Figure 4a ). We then characterized the biological functions of these homologous Ψ-modified genes among four bacterial strains, through pathway enrichment analysis (Kyoto Encyclopedia of Genes and Genomes, KEGG), uncovering three distinct but interlinked metabolic clusters. The predominant cluster showed highly significant enrichment in central carbon metabolism, including glycolysis, TCA cycle, oxidative phosphorylation, and amino acid biosynthesis ( Figure 4a , Figure 4—figure supplement 1a ). Notably, gene clusters essential for ATP production, such as atpA and atpD , were also enriched. Among the four bacterial strains, the functional enrichment of Ψ-modified genes in core metabolic pathways, combined with the conservation of Ψ motif contexts ( Figure 3a ), provides evidence for the functional importance and evolutionary conservation of a specific group of Ψ-modified transcripts across bacterial systems. In addition to the homologous Ψ-modified genes, we identified 633 strain-specific Ψ-modified mRNAs, highlighting the dynamic nature of Ψ modifications and suggesting potential strain-specific regulatory mechanisms. Figure 4. Evolutionary conservation of clustered Ψ modifications in orthologous genes. ( a ) The bar plot depicts the Ψ-modified homologous genes among bacterial species (green box) and strain-specific genes (orange box). Functional networks generated by clustering Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment results of Ψ-modified homologous genes present across two or more strains. The dot size in the network indicates the gene number contained in specific KEGG pathways. The p-value for each pathway was calculated with Fisher’s exact test. ( b ) Regions with Ψ enrichment across the atp and eno operons in four strains. Dark red dots represent highly modified sites with Ψ fraction values exceeding 50%. Figure 4—figure supplement 1. Comprehensive analysis showing functional enrichment of Ψ-containing transcripts and Ψ-enriched region. ( a ) The bar plot illustrates the top 20 enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, ranked by p-value, derived from Ψ-modified homologous genes identified in at least two strains. ( b ) The Gene Ontology analysis of Ψ-mRNA in exponential and stationary phases of P. aeruginosa . ( c ) Comparison of Ψ intensity patterns between strain-specific and homologous gene sets during two different growth phases. ( d ) Analysis of pseudouridylation density showing the number of Ψ sites occurring within 500 nt windows centered on each modified position. ( e ) Ψ modification-enriched regions detected across bacterial strains. The x -axis represents positions on the bacterial chromosome. Dark red indicates highly modified Ψ sites, defined as positions where the Ψ fraction exceeds 50%. Open in a new tab We then conducted a functional investigation of Ψ-modified genes in the two growth phases of P. aeruginosa . The Gene Ontology (GO) enrichment analysis revealed that Ψ-modified genes were significantly enriched in GO terms related to energy metabolism under exponential phase, compared to stationary phase, highlighting distinct state-specific patterns ( Figure 4—figure supplement 1b ). The exponential phase is known to exhibit elevated levels of metabolism-related genes, reinforcing our findings regarding the biological relevance of Ψ modifications. Notably, Ψ-modified genes showed a preferential enrichment in type IV swarming motility during stationary phase. For instance, the key transcription factor algR , which regulates multiple virulence factors and promotes P. aeruginosa twitching motility, serves as a representative Ψ-modified gene ( Kong et al., 2015 ). This finding suggests that Ψ modifications may coordinate motility behaviors under stationary growth phase, potentially linking RNA modification to adaptive virulence mechanisms upon limited nutrient availability. Besides, we observed that homologous genes exhibited more stable Ψ fraction patterns, likely due to their enriched functions in fundamental metabolic processes ( Figure 4—figure supplement 1c ). To further explore the regional distribution of Ψ modifications, we conducted a 500-nt sliding window analysis across all detected mRNA transcripts in four bacterial strains, for identifying specific regions with multiple Ψ sites ( Figure 4—figure supplement 1d ). Further analysis revealed that most Ψ-enriched regions corresponded to operons or gene clusters. baBID-seq identified Ψ clusters containing two or more Ψ sites within operons such as the evolutionarily conserved atp operon ( Ventura et al., 2004 ) and eno operon, across all tested strains ( Commichau et al., 2009 ; Salgado et al., 2000 ; Figure 4b ). This phenomenon was also observed in other homologous gene clusters or operons, such as rpoA , fusA , groE , and rpc operons ( Salgado et al., 2006 ; Figure 4—figure supplement 1e ), suggesting Ψ role in bacterial post-transcriptional regulation ( Rodell et al., 2024 ). We observed distinct regional patterns of Ψ modifications, in operons and gene clusters: some Ψ sites accumulated within 3′ UTR regions of atp operon, for instance, in atpD of P. aeruginosa ( Figure 4b ); Ψ sites in atpD gene were predominantly located near stop codon in P. aeruginosa , whereas most Ψ gathered at the translation initiation region in B. cereus ( Figure 4b ). This regional analysis of Ψ clusters revealed strain-specific distribution patterns in the conserved operons. The flexible regional Ψ modifications in operons, such as atp , may regulate gene expression to align with bacteria-specific metabolic needs. Growth state-dependent dynamic Ψ modifications in bacterial mRNA Ψ modification levels on mRNA have been reported to fluctuate under stress conditions in human cells ( Li et al., 2015 ). baBID-seq also observed alterations in Ψ modifications across four bacterial strains under different growth conditions. According to a detailed comparative analysis of Ψ sites between two growth phases, we witnessed both newly emerged and diminished Ψ modification events, as well as alteration in modification fractions at conserved sites. The quantitative baBID-seq approach allowed us to pinpoint dynamic Ψ modifications in response to bacterial metabolic shifts and changes in growth states. We initially compared the distribution of modification fractions for all mRNA Ψ sites in exponential versus stationary phase. K. pneumoniae and B. cereus exhibited significantly higher global Ψ levels during the stationary phase ( Figure 5—figure supplement 1a, b ). In contrast, in either nutrient-enriched or minimal media (MM) condition, P. aeruginosa and P. syringae did not show significant changes in global Ψ fractions between two phases ( Figure 5—figure supplement 1c, d ). We then focused on changes in Ψ fractions at specific sites, setting a cutoff of 10% variation between two growth phases, across four bacterial strains. In P. aeruginosa and P. syringae , we identified numerous phase-specific Ψ sites that either emerged or disappeared, as well as many conserved Ψ sites exhibiting changes in Ψ fractions above 10% ( Figure 5a, b ). For instance, in P. aeruginosa , key genes linked to energy metabolism, such as sucB, sucC , and gltA , displayed Ψ sites of increased modification fraction during exponential phase of heightened metabolic activity. The stop codon region of secY , a protein essential for type II secretion systems in P. aeruginosa , contained a highly modified Ψ site that was only detected in exponential growth phase; similar Ψ fraction dynamics were observed for secY in K. pneumoniae and B. cereus ( Figure 5—figure supplement 1a–c ). Figure 5. Growth state-dependent dynamics of Ψ modification fraction. Heatmap showing the mRNA Ψ fractions alteration of each site in P. aeruginosa ( a ) and P. syringae ( b ). The color intensity reflects Ψ fraction at each site. Only sites with >10% absolute difference in Ψ fraction between exponential and stationary phases are displayed. Blank boxes signify either unmodified sites or those with Ψ fractions below 2%. The annotation combining position label and gene name indicates the precise location of Ψ modification within genes. ( c ) Scatter plot shows Ψ intensity alteration between two growth phases in P. aeruginosa . For each mRNA, Ψ intensity is calculated as the sum of all Ψ fractions throughout the transcript. ( d ) Similar to ( c ), the scatter plot shows Ψ intensity alteration in P. syringae under two growth phases. ( e ) The scatter plot shows Ψ intensity alteration in P. syringae under two growth phases in MM medium condition. The box plot shows the modified and unmodified mRNA transcripts per kilobase million (TPM) changing between exponential and stationary growth phases under different conditions: P. aeruginosa cultured in Luria-Bertani (LB) medium ( f ), P. syringae cultured in King’s B (KB) medium ( g ), and P. syringae cultured in MM medium ( h ). The y -axis shows log 2 (TPM at exponential phase/TPM at stationary phase) of each mRNA. The red color presents Ψ-mRNA, and the blue color indicates no-Ψ-mRNA. Wilcoxon signed-rank test; ns, p-value  ≥0.05; *p-value <0.05; **p-value <0.01; ***p-value <0.001, and ****p-value <0.0001. ( i ) The scatter plot illustrates the correlation between translation efficiency (TE) alteration (log 2 Foldchange of (TE_MM/TE_KB)) and Ψ intensity difference (Ψ intensity of MM-Ψ intensity of KB) in P. syringae cultured under MM medium (TE_MM) versus KB conditions (TE_KB). ( j ) The proportion of Hfq-bound Ψ-mRNA (red color for exponential growth phase mRNA and blue color for stationary phase mRNA) versus those non-Ψ-mRNA (gray color) across two growth conditions in P. aeruginosa . ( k ) The scatter plot shows a correlation between mRNA Ψ intensity and Hfq-binding score, where the Hfq-binding score is calculated as the sum of each mRNA peak binding strength (log 2 Foldchange value of each peak). Figure 5—figure supplement 1. Growth state-dependent dynamics of Ψ modification fraction. ( a–c ) Paired visualization showing Ψ dynamics across growth phases, with a violin plot depicting the distribution of Ψ fractions in mRNA during two distinct growth phases (left), and a corresponding heatmap displaying the specific Ψ modification sites dynamic detected across bacterial strains at these same growth phases (right). ( d ) The Ψ fraction distribution on mRNA in P. aeruginosa (left) and P. syringae (right) at two growth phases. ( e ) The violin plot quantifies the differential pseudouridylation intensity (exponential phase Ψ intensity minus stationary phase Ψ intensity) of specific genes at two growth phases. ( f ) The distribution plot illustrates the spatial relationship between Ψ modification sites and Hfq-binding peak centers during two distinct growth phases. The violin plot reveals that the distribution of minimum distances between Ψ sites and Hfq-binding centers varies with statistical significance (p-value = 0.0014) between growth phases, as determined by an F -test. ( g ) The scatter plot depicts the relationship between Ψ modification intensity and gene expression levels (log 10 (TPM)) under two distinct growth conditions. Open in a new tab In addition to analyzing Ψ sites, we also calculated Ψ intensity for each gene, defined as the sum of modification fractions for all Ψ sites across one single mRNA. During the transition from exponential to stationary phase, Ψ intensity of specific genes shifted significantly ( Figure 5c–e , Figure 5—figure supplement 1e ). In the exponential phase, many genes involved in metabolism, amino acid biosynthesis, and protein synthesis exhibited higher Ψ intensity, such as rpsD , rpsT , tufB , lon , and nuoD , reflecting their roles in supporting rapid growth; meanwhile, for these genes, their decreased Ψ intensity observed in the stationary phase may suggest a coordinated Ψ reprogramming that helps bacteria adapt to reduced nutrient availability and increased cell density, through downregulating energy-intensive processes. Overall, the dynamic nature of Ψ modifications—including newly emerged and diminished Ψ sites, as well as the ones with altered Ψ fractions between growth conditions—suggests their potential roles as responsive epitranscriptomic switches that facilitate bacteria to be adapted to varying environmental conditions and metabolic demands. Ψ correlates with bacterial mRNA metabolism and function Ψ can enhance mRNA stability and translation in mammals ( Karikó et al., 2008 ), parasites ( Li et al., 2025 ; Nakamoto et al., 2017 ), and plants ( Li et al., 2025 ). However, the effects of Ψ on bacterial mRNA have yet to be investigated. We normalized gene expression levels by calculating transcripts per kilobase million (TPM) for two growth phases and grouped the adequately expressed genes into Ψ-modified mRNA (Ψ-mRNA) and unmodified mRNA (non-Ψ-mRNA). The analysis revealed that the expression level of Ψ-mRNA was significantly higher than that of non-Ψ-mRNA in P. aeruginosa and P. syringae when cultured in a nutrient-sufficient medium ( Figure 5f, g , Figure 5—figure supplement 1f ). However, P. syringae exhibited more moderate changes in mRNA expression between two growth phases under MM conditions ( Figure 5h ). Our findings suggest that Ψ may stabilize mRNA in a growth phase-dependent manner in bacteria. We then conducted an analysis of TE in P. syringae with alterations in Ψ modifications under two distinct conditions (King’s B medium, KB, and MM) to examine whether Ψ impacts bacterial mRNA translation. Although we did not observe a strong global correlation between changes in Ψ modifications and TE for all genes under both conditions, a larger proportion of genes tend to show a positive correlation ( Figure 5i ). Our findings partially align with previous studies in mammals, which suggested that Ψ modifications may help enhance mRNA TE in bacteria, with complex roles to be further studied in translation regulation. Hfq is a major bacterial post-transcriptional regulator that functions as a pivotal RBP, orchestrating various cellular processes ( Trouillon et al., 2022 ). Its regulatory mechanisms have been extensively characterized, including the alteration of RNA structure and the facilitation of sRNA–mRNA interactions ( Chihara et al., 2019 ), highlighting its fundamental role in coordinating gene expression networks ( Dos Santos et al., 2019 ; Sobrero and Valverde, 2012 ). To investigate the potential association between Hfq and Ψ modifications, we performed an integrative analysis combining our data with previously published Hfq RIP-seq data ( Trouillon et al., 2022 ) in P. aeruginosa . We examined both exponential and stationary growth phases to evaluate whether Hfq targets Ψ-modified regions or whether Ψ modifications affect Hfq–RNA interactions. Our results indicated that the distance between Ψ sites and Hfq peak centers significantly decreased during the stationary phase ( Figure 5—figure supplement 1g ); meanwhile, Hfq-bound genes accounted for a substantial proportion of mRNA Ψ sites, with 12.5% and 27.3% during exponential and stationary phases, respectively ( Figure 5j ). By defining Hfq-binding score as the sum of enrichment scores at all Hfq peaks per transcript, we found that, in stationary phase, Hfq tends to exhibit a stronger binding affinity for genes carrying more Ψ modifications ( Figure 5k ). These results suggest that Ψ modifications may facilitate mRNA–Hfq interactions to some extent. Overall, our results suggest that dynamic Ψ modifications could influence bacterial mRNA stability, translation, and RBP interactions, in response to altered cellular demands during growth phase transitions. Integrated computational analysis reveals structure-dependent Ψ modifications in bacterial RNA We observed diverse motif contexts at Ψ sites in bacterial mRNA ( Figure 3a, b ). This observation aligns with previous studies demonstrating that pseudouridine synthases, such as PUS1 and TruB, preferentially recognize RNA local structures beyond primary sequence motifs for Ψ installation ( Carlile et al., 2019 ; Lange et al., 2012 ; Pan et al., 2003 ; Safra et al., 2017 ). To computationally model the widespread Ψ modifications on bacterial mRNA, which are hypothesized to be structure-dependent, we incorporate Ψ modification determinants through clustering analyses of local RNA sequences and structural elements. We first calculated the predicted secondary structure at the 41-nt region centered by GUΨC motif, with the representative Ψ sites of 50–96% modification fraction ( Figure 6a ). Interestingly, all GUΨC motifs with varying Ψ fractions are predicted to occur within RNA loop structures. To determine whether RNA structural factors influence Ψ deposition and modification fractions, we compared two highly prevalent motifs, GUΨC and GCΨCG, by clustering predicted RNA structures across all RNA species in P. aeruginosa . Aside from tRNA and rRNA, which clustered together due to their distinct structural features, we observed small clusters within certain mRNAs, such as guaB , recA , and PA4943 . Overall, no characteristic structural signatures could completely discriminate GUΨC versus GCΨCG motif ( Figure 6b ). To gain deeper insights, we conducted structural clustering analyses of all RNA species across different strains ( Figure 6—figure supplement 1a–d ). Given that some pseudouridine synthases target-specific RNA structures, we anticipated highly distinguishable clustering results; however, such distinct clustering patterns were not observed. This suggests that certain pseudouridine synthases, such as RluA ( Schaening-Burgos et al., 2024 ), may not solely rely on structural features for RNA targeting. Notably, we identified clusters of Ψ sites with similar structures that exhibited higher modification fractions, including sucC and sucB in P. aeruginosa ( Figure 6—figure supplement 1a ), as well as sdhA , PSPPH_RS14750 , and atpB in P. syringae ( Figure 6—figure supplement 1c ). Combining these findings with the distinctive Ψ fraction patterns in various motifs ( Figure 3b ), our results suggest that both RNA sequence and local structure may affect Ψ installation. Figure 6. Structure-dependent patterns of Ψ modifications and transformer-graph neural network (GNN)-based deep learning network for Ψ prediction. ( a ) Predicted RNA secondary structures containing the GUΨCG motif with corresponding Ψ fraction values and gene identifiers annotated. MXfold2 is employed to model these structures using 20 nt flanking sequences extending from each modification site. ( b ) Sequence and structure clustering of 41-nucleotide RNA segments centered by Ψ sites with fraction values greater than 0.1 and containing either GCΨCG (black branch color) or GUΨC (red branch color) motifs. The circular visualization features three concentric layers: the inner layer displays the Ψ fraction value, the middle layer indicates RNA type, and the outer layer represents the GC ratio (%) of each 41 nt RNA segment. The red dot around the circle marks the position of RNA displayed in ( a ). ( c ) Architecture of pseU_NN. The model integrates sequence and structural information through two parallel pathways: (1) A sequence analysis branch with one-hot embedding followed by a multi-head transformer module, and (2) A structure analysis branch that processes RNA secondary structure adjacent matrices through a graph convolution module to extract structural features. The features extracted by the two modules are further weighted and merged as input for residual blocks (fully connected layers). ( d ) Bar plots summarize model performance with input sequences of 41 nt, 61 nt, and 81 nt, evaluated by PR-AUC, accuracy (ACC), F 1 score, Matthews correlation coefficient (MCC), precision, recall, and ROC-AUC. Overall, the three sequence lengths show comparable performance across metrics, with the 41 nt model achieving slightly higher PR-AUC and ROC-AUC, indicating that shorter sequence contexts are sufficient for robust pseudouridine prediction. ( e ) Multi-metric assessment showing precision–recall curve (AUC 0.906), F 1 curve (AUC 0.821), and ROC curve (AUC 0.89) of the pseU_NN model on 41 nt validation datasets, achieving a peak F 1 score of 0.804. ( f ) Distribution of pseU_NN prediction scores on 41 nt test datasets. Figure 6—figure supplement 1. The structure and sequence clustering analysis of RNA molecules containing Ψ modifications. ( a–d ) Sequence and structure clustering of 41-nucleotide RNA segments centered on Ψ modification sites with fraction values greater than 0.1. The circular visualization comprises three concentric layers: the inner layer displays the Ψ fraction value, the middle layer designates RNA type, and the outer layer represents the GC ratio of each 41-nucleotide RNA segment. ( e ) Multi-metric assessment showing the precision–recall curve (AUC = 0.899), F 1 score curve, and ROC curve (AUC = 0.879) of the pseU_NN model on the 61-nt validation dataset. The right panel ( f ) shows the distribution of pseU_NN prediction scores on the 61-nt validation dataset. ( g, h ) Similar to panels ( e, f ), multi-metric assessment showing the precision–recall curve (AUC = 0.888), F 1 score curve, and ROC curve (AUC = 0.869) of the pseU_NN model on the 81-nt validation dataset, and the right panel shows the distribution of pseU_NN prediction scores on the 81-nt validation dataset. Open in a new tab LSTM-transformer-GNN-based neural networks for prediction of Ψ-modified sites In next-generation sequencing data, variable read coverage dictated by gene expression patterns or limited sequencing depth can lead to missed Ψ sites. To address this limitation, we implemented a methodology that integrates RNA sequence and local structure for a transcriptome-wide scan of Ψ sites, resulting in a more comprehensive inventory of Ψ candidate sites. Previous studies have shown that the sequence context surrounding Ψ sites could serve as a reliable predictor ( Hoang and Ferré-D’Amaré, 2001 ; Song et al., 2021 ). Building on this, we developed a deep learning model that accurately captures both sequence and structural features surrounding known Ψ sites across various RNA species, allowing us to predict potential modification sites that may be condition-dependent or below baBID-seq detection thresholds. We extracted sequence segments of 41, 61, and 81 nucleotides (±20, ±30, and ±40 nt) centered at each Ψ site, applying window shifts of ±5, ±10, and ±15 nt, respectively. The input sequences were then embedded using one-hot encoding and processed through a multi-head transformer module, followed by convolution layers and bidirectional LSTM (Long Short-Term Memory) layers. Simultaneously, we utilized adjacency matrices representing local RNA structures predicted using MXfold2 ( Sato et al., 2021 ) as input for a GNN module. Features extracted from both modules were combined through weighted concatenation and subsequently processed using a residual block. We employed binary cross-entropy loss for predicting the likelihood of Ψ modifications. This hybrid LSTM-transformer-GNN architecture effectively integrated both RNA sequence and local structure characteristics across various transcripts ( Figure 6c ), termed pseU_NN. We used 3377 high-confidence Ψ sites (with fraction values >2%) as positive samples. The negative samples consisted of 3400 randomly selected U sites that contained the unique Ψ motif but without any evidence of Ψ deposition in baBID-seq. The dataset was then divided into 4744 training samples, 1016 test samples, and 1017 validation samples. The model consistently performed well across different input dimensions (41, 61, and 81 nucleotides), with all variants achieving AU-ROC scores exceeding 0.8 after convergence ( Figure 6d ). Using 41-nt inputs, our approach achieved impressive validation metrics, including an area under the precision–recall curve (AU-PRC) of 0.905 and an AU-ROC of 0.89 ( Figure 6e, f ). Models trained with alternative input sequence lengths also demonstrated strong performance metrics ( Figure 6—figure supplement 1e–h ). These results set the basis for further development of effective deep learning tools for transcriptome-wide Ψ prediction in bacteria and mammals. Discussion RNA modifications in bacteria, particularly Ψ, are less characterized than their well-studied eukaryotic counterparts. Leveraging recent advances in quantitative sequencing methods such as BID-seq ( Dai et al., 2023 ; Zhang et al., 2024 ), here we developed baBID-seq and presented single-base resolution maps of Ψ modifications, complete with stoichiometric information, across four bacterial strains under different growth phases. Our findings confirm the widespread occurrence of Ψ modifications in bacterial RNA and provide insights into their functional relevance. This extensive dataset serves as a valuable resource for understanding the evolutionary and functional significance of Ψ modifications in bacterial RNA. The Ψ modification plays regulatory roles in tRNA aminoacylation, stability, and the formation of functional structures ( de Crécy-Lagard et al., 2019 ; de Crécy-Lagard and Jaroch, 2021 ; Krutyhołowa et al., 2019 ; Schultz et al., 2024 ). Our analysis revealed that tRNA Ψ modifications are present in varying fractions, with a stronger modification level observed in the TΨC loop compared to the anticodon-arm and D-arm loops. Previous studies indicate that the tRNA T-arm is highly modified, not only by Ψ but also by other uridine modifications like 5-methyluridine (m 5 U) ( Chou et al., 2017 ). Both Ψ and m 5 U modifications globally enhance tRNA aminoacylation and also independently influence specific tRNA modifications, such as 3-(3-amino-3-carboxypropyl)uridine at position 47 ( Schultz et al., 2024 ). The TΨC loop is crucial for the interaction between tRNA and ribosome, facilitating the formation of the tRNA–ribosome complex ( Chou et al., 2017 ). Interestingly, we observed a novel phenomenon where Ψ modifications on the tRNA T-arm increase in the stationary growth phase compared to the exponential phase. Given that codon composition and mRNA expression are closely correlated ( Gouy and Gautier, 1982 ), dynamic Ψ modification within the TΨC loop may impact bacterial mRNA translation by modulating T-arm interactions with the ribosome. Besides, other RNA modifications on tRNA may be influenced by dynamic Ψ during growth phase transitions, potentially creating a feedback loop where existing modifications affect the biogenesis of subsequent modifications. Overall, this condition-dependent Ψ modification in tRNA may represent a new mechanism by which bacteria adapt to varying environmental conditions, anticipating future investigation. Previous studies have demonstrated that both a deficiency and an excess of pseudouridine can severely impair ribosomal translation and proper assembly in E. coli ( Leppik et al., 2017 ; O’Connor et al., 2018 ). The stable fraction of Ψ modification observed in E. coli rRNA across two different growth phases suggests that rRNA Ψ may be tightly regulated to maintain essential rRNA function. The role of Ψ in mRNA remains largely unclear across the three domains of life. Our results reveal a quantitative mRNA Ψ landscape in four bacterial strains. We found that the overall fraction of mRNA Ψ modifications was significantly lower than that of rRNA and tRNA, consistent with findings in plants and mammals ( Dai et al., 2023 ; Li et al., 2025 ). The distribution and stoichiometric patterns of mRNA Ψ between Gram-positive and Gram-negative bacteria exhibited similarities. Notably, we identified evolutionarily conserved Ψ modifications in mRNAs encoding proteins involved in energy generation, ATP binding, amino acid synthesis, and protein translation, mirroring the observations in mammals and plants ( Dai et al., 2023 ; Li et al., 2025 ). We also discovered clusters of multiple Ψ sites enriched in specific operons related to conserved functions, which were detected across multiple strains. To date, limited research has focused on Ψ modifications and their alterations in bacteria. In this study, we profiled and uncovered dynamic changes in Ψ modifications during growth phase transitions among four bacterial strains. During the metabolically active phase of P. aeruginosa , we observed increased Ψ modifications in many metabolism-related genes. In the stationary phase, our analysis revealed reduced pseudouridylation within the CDS of fimV , a gene that encodes an inner membrane protein in P. aeruginosa responsible for regulating intracellular cyclic AMP levels, type IV-mediated twitching motility, and type II secretion system genes ( Buensuceso et al., 2016 ; Semmler et al., 2000 ). Ψ modifications in CDS are known to alter codon properties on mRNA, leading to reconstituted translation and promoting the low-level synthesis of multiple peptides ( Eyler et al., 2019 ). These findings suggest a potential new mechanism for regulating bacterial metabolism and quorum sensing in P. aeruginosa . Previous studies have demonstrated that pseudouridylation enhances mRNA stability in mammals and plants ( Dai et al., 2023 ; Li et al., 2025 ; Zhang et al., 2024 ), and our research confirms and extends these observations to bacterial systems. It has been reported that methionine aminoacyl tRNA Met synthetase can target Ψ1074 in yeast ( Levi and Arava, 2021 ), and several Ψ sites overlap with RBP-binding regions ( Martinez et al., 2022 ). To systematically investigate dynamic Ψ modifications in bacterial RNA and their potential impact on RBP binding, we conducted an integrative analysis that combined our baBID-seq data with Hfq RIP-seq data from P. aeruginosa . Hfq is an RNA chaperone that recognizes 5-repeat AAN motifs in P. aeruginosa and plays a crucial role in regulating various post-transcriptional processes, including mediating sRNA–mRNA interactions ( Chihara et al., 2019 ). Hfq can trigger mRNA structural reprogramming, and RNA structural switches may facilitate or suppress pseudouridylation ( Carlile et al., 2019 ; Hua et al., 2024 ). We observed increased Hfq binding to Ψ-modified mRNAs during the stationary phase compared to exponential phase, suggesting that Ψ may modulate Hfq-mediated regulation via direct or indirect effects on RNA–protein affinity and mRNA structure remodeling. To establish causality, targeted perturbation of specific pseudouridine synthases or in vitro Hfq–RNA interaction assays with Ψ-modified versus unmodified RNAs are needed for confirmation. Ψ modification is specifically recognized in RNA local structures or motif contexts by TruB pseudouridine synthase ( Machnicka et al., 2014 ). Our analysis of Ψ-containing sequences in mRNA revealed conserved motif signatures that closely resemble those found in tRNA and rRNA. This conservation pattern suggests that tRNA and rRNA Ψ synthases may directly or indirectly recognize similar sequence and structural elements in mRNA, as evidenced by PUS1 and PUS6, which can both add Ψ to tRNA and mRNA ( Carlile et al., 2019 ; Levi and Arava, 2021 ). Through an integrated clustering analysis combining sequence and structural features, Ψ site location and modification fraction were suggested to link with RNA local structures ( Carlile et al., 2019 ). This may explain why certain potential modification sites with appropriate motifs remain unmodified or exhibit dynamic Ψ levels under different growth conditions. Deep learning methods are increasingly utilized for RNA modification prediction. Building on previous concepts, such as attention-based multi-label neural networks that predict multiple RNA modifications using sequence context ( Song et al., 2021 ), we developed a hybrid LSTM-transformer-GNN architecture that integrates RNA structural features with multi-head attention mechanisms to predict potential Ψ sites (pseU_NN), particularly those on transcripts of low expression levels. pseU_NN enables the prediction of potential Ψ sites in different bacterial contexts, providing a more comprehensive Ψ map across bacterial transcriptomes. Overall, this study presents the first quantitative landscape of Ψ modifications across diverse bacterial strains. baBID-seq revealed Ψ stoichiometry in tRNA, rRNA, and mRNA under exponential and stationary growth conditions, as well as nutrient-deficient conditions. The motif analysis provides insights into pseudouridine synthase activity on bacterial mRNA. Evolutionarily conserved patterns of Ψ-enriched modifications were identified in operons involved in metabolic pathways across bacterial strains. In summary, our study enhances the understanding of Ψ modifications and their functions in bacteria, paving the way for future mechanistic investigations. Materials and methods Bacteria strains and growth conditions The wild-type P. syringae pv. phaseolicola 1448A strain was cultured in KB medium ( King et al., 1954 ) (20 g/l proteose peptone, 1.5 g/l K 2 HPO 4 , 1.5 g/l MgSO 4 ·7 H 2 O, and 10 ml/l glycerol) at 28°C for 12 hr (overnight) until reaching an optical density at 600 nm (OD₆₀₀) of 1–2, corresponding to stationary phase. B. cereus ATCC 14579, P. aeruginosa PAO1, Escherichia coli K-12 MG1655, and K. pneumoniae CR-HvKP4 were grown in Luria-Bertani (LB) broth at 37°C until achieving an OD₆₀₀ of 1–2 for stationary phase samples. For exponential phase samples, all bacterial strains were first cultured to the stationary phase as described, then subcultured into fresh medium and incubated under the same conditions until reaching an OD₆₀₀ of 0.5–0.6. For P. syringae in MM, exponential phase cells were harvested, washed three times with freshly prepared MM ( Huynh et al., 1989 ), resuspended in MM at an OD₆₀₀ of 0.1–0.2, and cultured for an additional 6 hr. baBID-seq library construction RNA was extracted using RNA isolation Kit V2 (Vazyme, #RC112-01). The extracted RNA was processed using DNase I (RNase-free, NEB # M0303S) and collected using RNA Clean & Concentrator-5 (Zymo #R1014). The RiboRID technique was used for rRNA depletion ( Choe et al., 2021 ). For strains B. cereus and K. pneumoniae , rRNA was depleted with NEBNext rRNA Depletion Kit (Bacteria) (#E7850L). RNA concentration in each step was tested using Qubit RNA Assays. One hundred nanograms of RNA generated from the ribosome removal process from each biological replicate were used for the library construction. The fragmentation was optimized to 4 min at 70°C with the fragmentation buffer used (Invitrogen). For the following steps, we strictly followed the BIDseq protocol ( Zhang et al., 2024 ). The final amplified cDNA was collected and then optimal library size is selected using native PAGE gels. 40% Acrylamide/Bis (29:1) 10% native polyacrylamide gel was used. The optimal band (175–200 bp) used collected. The gel was soaked in 400 μl of 1× TE buffer at 37°C for 1 hr on the thermal shaker at 600 rpm. The gel was crushed and snap-freezed using liquid nitrogen and incubated on a thermal shaker at 37°C at 600 rpm for 12 hr. The supernatant was collected using Spin-X, followed by DNA precipitation. The constructed libraries were sequenced on the Illumina NovaSeq sequencing platform in paired reads mode and single-end reads were used for following BID-seq data processing. pseU-TRACE verification RNA samples (500 ng each for input and bisulfite-treated conditions) were subjected to bisulfite treatment using freshly prepared bisulfite (BS) reagent containing 2.4 M Na 2 SO 3 and 0.36 M NaHSO₃, followed by incubation at 70°C for 3 hr. Treated RNA was purified using the Zymo RNA Clean & Concentrator-5 kit. For RT, 1 µl of 5 µM site-specific RT primer (for either the target Ψ site or the negative control site) was added, and samples were incubated at 65°C for 5 min and immediately placed on ice. RT was then performed using SuperScript IV reverse transcriptase (Thermo Fisher #18090050), following the same RT conditions as in the BID-seq protocol. The resulting cDNA was treated with RNase H (NEB #M0297S) at 37°C for 20 min, followed by heat inactivation of RNase H at 70°C for 5 min. For splint-ligation, 1 µl of cDNA was mixed with upstream and downstream primers (final concentration 0.01 µM each), and the mixture was annealed using a temperature gradient (90°C for 1 min, 80°C for 1 min, 70°C for 1 min, 60°C for 1 min, 50°C for 1 min, and 40°C for 6 min). 2 µl of SplintR ligase (NEB #M0375S) was then added, and ligation was carried out at 40°C for 60 min, followed by denaturation at 95°C for 5 min and holding at 12°C. The reaction was diluted with 40 µl RNase-free H₂O. Quantitative real-time PCR (qPCR) was performed using a QuantStudio 3 PCR System. Each 20 µl reaction contained 2×SYBR Green qPCR Master Mix (MCE #HY-K0501), qPCR forward and reverse primers, diluted ligation product, and RNase/DNase-free water. The qPCR cycling conditions were as follows: 95°C for 5 min; 40 cycles of 95°C for 10 s and 60°C for 35 s; followed by 95°C for 15 s and 60°C for 1 min (fluorescence acquisition at a ramp rate of 0.05°C/s). Ct values were normalized to the negative control site within each replicate, and the normalized treated signal was further normalized to the corresponding input sample to quantify the Ψ modification level. Primer sequences used for detection are listed in Supplementary file 2 . Sequencing data processing and analysis Sequencing data were subjected to a refined bioinformatic workflow adapted from previously established protocols. Raw sequencing reads underwent adapter trimming via Cutadapt (v.3.5) and PCR duplicate elimination using BBMap tools (v.38.73). The filtered reads were initially aligned to a curated repository of non-coding RNA sequences (including rRNAs, tRNAs, and other small RNAs). Subsequently, unmapped reads were subjected to genome alignment with optimized mapping parameters tailored for Ψ detection. The resultant alignment files were processed with Samtools (v.1.13) to generate strand-specific BAM files, which were then interrogated using bam-readcount (v.1.0.1) to quantify nucleotide deletion events and calculate coverage metrics. Ψ modification sites were identified by integrating deletion rate profiles, site coverage, and background signal correction derived from ‘input’ libraries. Bacterial samples from distinct temporal phases were analyzed as discrete entities to minimize batch effects. The quantitative assessment of Ψ levels was achieved by transforming raw deletion ratios according to previously reported calibration curves, yielding high-confidence Ψ fractions at single-nucleotide resolution. To ensure robust and reliable detection of low-level Ψ sites in mRNA, we applied the following stringent BID-seq filtration criteria to all candidate sites (all sites reported in the Supplementary file 1 passed these thresholds): (1) total sequencing coverage >20 reads in both the bisulfite-treated (BID-seq) libraries (Σd t >20) and untreated input libraries (Σd i >20); (2) average deletion number >5 in the treated libraries; (3) average modification fraction >0.02 (2%) in the treated libraries; and (4) average deletion ratio in treated libraries at least twofold higher than in input libraries. We consider sites with stoichiometry thresholds >0.5 as highly modified sites. TPM calculation The TPM value for a specific transcript i is calculated as: T P M i = ( q i l i ∑ j ( q j l j ) ) × 10 6 q i is the number of reads mapped to transcript i . l i is the length of transcript i (in nucleotides). 10 6 is the scaling factor to express the result as transcripts per million, j is unique transcript number ( Zhao et al., 2021 ). Evolutionary analysis Orthologous gene analysis was performed using OrthoFinder ( Emms and Kelly, 2019 ) software to identify homologous gene clusters across the five bacterial strains. The analysis incorporated complete genome sequences and their corresponding GFF (General Feature Format) annotation files with default parameters. The resulting orthologous gene clusters were subsequently utilized for KEGG pathway analysis. Genes that were not clustered in the orthology analysis were excluded from the KEGG pathway mapping to ensure reliable results. Gene functional annotations were performed using eggNOG-mapper v2 ( Cantalapiedra et al., 2021 ). RNA structure analysis Pseudouridine sites with a modification fraction exceeding 2% were retained for downstream analysis. To capture the local sequence context surrounding each modification, 20 nucleotides upstream and downstream of each pseudouridine site were extracted, yielding 41-nucleotide sequences centered on the modification position. These sequences were subsequently subjected to RNA secondary structure prediction using MXfold2 ( Sato et al., 2021 ) and clustering analysis using RNAclust ( Engelhardt et al., 2010 ). All downstream result processing and visualization were performed using custom scripts implemented in R. pseU_NN Datasets preparation The negative control dataset was constructed by scanning bacterial genomes for pseudouridylation-compatible sequence motifs that were absent from the experimentally verified modification sites identified in baBID-seq data. Sequence segments of 41, 61, and 81 nucleotides were then generated, with each unmodified motif positioned at the center. MXfold2 ( Sato et al., 2021 ) was used to predict the secondary structure of all sequences as input for training and validation process. A total of 3377 high-confidence Ψ sites with fraction values exceeding 2% were used as positive samples, while 3400 randomly selected sites carrying the unique Ψ motif but showing no experimental evidence of Ψ deposition were designated as negative samples. For each sequence length, the complete dataset was partitioned into 4744 training samples, 1016 test samples, and 1017 validation samples. Model architecture The model architecture comprises three main components. First, two dense graph encoders were used to extract RNA secondary structure features derived from MXfold2 predictions and from one-hot-encoded RNA sequences, respectively. Second, the one-hot-encoded sequence was further processed by a convolutional layer for local sequence feature extraction, followed by two bidirectional LSTM layers to capture long-range dependencies. Positional encoding was subsequently applied before the representations were forwarded to a single-layer multi-head Transformer block. Finally, features extracted from the structure- and sequence-based modules were integrated via weighted concatenation and passed through a residual block composed of fully connected layers, with a sigmoid activation function applied to the final output to predict the probability of pseudouridine modification. Training and evaluation The pseU_NN was trained using binary cross-entropy loss with the Adam optimizer. We implemented a dynamic learning rate scheduler that adjusted the rate based on validation AUC-ROC performance. For the final evaluation, we utilized an independent test dataset, with F 1 scores and precision–recall curves guiding prediction threshold selection. This balanced approach ensured optimal sensitivity and specificity—critical in biological classification systems where both false positives and false negatives carry significant consequences. The F 1 score is the harmonic mean of precision and recall. The definition of true positive (TP), true negative (TN), false positives (FP), and false negative (FN) with shifting was adopted in previous studies ( Wang et al., 2023 ; Yu et al., 2021 ). p r e c i s i o n = T P T P + F P r e c a l l = T P T P + F N F 1 s c o r e = 2 × p r e c i s i o n × r e c a l l p r e c i s i o n + r e c a l l The accuracy calculated using standard definitions: a c c u r a c y = T P + T N T P + F P + T N + F N The Matthews correlation coefficient (MCC) is calculated using the following formula: M C C = T P × T N − F P × F N ( T P + F P ) ( T P + F N ) ( T N + F P ) ( T N + F N ) Code availability The code for pseU_NN used for this paper is available at https://github.com/Dylan-LT/pseU_NN , copy archived at Dylan-LT, 2026 . Acknowledgements This study was funded by the Guangdong Major Project of Basic and Applied Basic Research (2020B0301030005), National Natural Science Foundation of China (32172358), General Research Funds of Hong Kong (21103018, 11101619, and 11102720), and Early Career Scheme of Hong Kong (26103623). The funders had no role in study design, data collection, interpretation, or the decision to submit the work for publication. Funding Statement The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication. Contributor Information Li-Sheng Zhang, Email: zhangls@ust.hk. Xin Deng, Email: xindeng@cityu.edu.hk. Camilo Perez, University of Georgia, United States. Bavesh D Kana, University of the Witwatersrand, South Africa. Funding Information This paper was supported by the following grants: Guangdong Major Project of Basic and Applied Basic Research

2020B0301030005 to Xin Deng. General Research Funds of Hong Kong

21103018 to Xin Deng. Early Career Scheme of Hong Kong

26103623 to Xin Deng. National Natural Science Foundation of China

32172358 to Xin Deng. General Research Funds of Hong Kong

11101619 to Xin Deng. General Research Funds of Hong Kong

11102720 to Xin Deng. Additional information Competing interests No competing interests declared. Author contributions Conceptualization, Data curation, Software, Formal analysis, Validation, Visualization, Methodology, Writing – original draft. Data curation, Software, Visualization, Methodology. Methodology. Software. Data curation, Writing – original draft, Writing – review and editing. Writing – review and editing. Writing – review and editing. Software. Writing – original draft, Writing – review and editing. Supervision. Additional files Supplementary file 1. Modification sites. elife-107545-supp1.csv (452.1KB, csv) Supplementary file 2. Primer. elife-107545-supp2.csv (774B, csv) MDAR checklist elife-107545-mdarchecklist1.docx (88.7KB, docx) Data availability All sequencing data files have been submitted to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database with the reference code of GSE292335 . The following dataset was generated: Deng X, Xu L. 2026. Quantitative RNA pseudouridine maps reveal functional insights into pseudouridylation in bacteria. NCBI Gene Expression Omnibus. GSE292335 References Adams PP, Baniulyte G, Esnault C, Chegireddy K, Singh N, Monge M, Dale RK, Storz G, Wade JT. Regulatory roles of Escherichia coli 5’ UTR and ORF-internal RNAs detected by 3’ end mapping. eLife. 2021;10:e62438. doi: 10.7554/eLife.62438. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Anderson BR, Muramatsu H, Jha BK, Silverman RH, Weissman D, Karikó K. Nucleoside modifications in RNA limit activation of 2’-5’-oligoadenylate synthetase and increase resistance to cleavage by RNase L. Nucleic Acids Research. 2011;39:9329–9338. doi: 10.1093/nar/gkr586. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Buensuceso RNC, Nguyen Y, Zhang K, Daniel-Ivad M, Sugiman-Marangos SN, Fleetwood AD, Zhulin IB, Junop MS, Howell PL, Burrows LL. The conserved tetratricopeptide repeat-containing C-terminal domain of Pseudomonas aeruginosa FimV is required for its cyclic AMP-dependent and -independent functions. Journal of Bacteriology. 2016;198:2263–2274. doi: 10.1128/JB.00322-16. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cantalapiedra CP, Hernández-Plaza A, Letunic I, Bork P, Huerta-Cepas J. eggNOG-mapper v2: functional annotation, orthology assignments, and domain prediction at the metagenomic scale. Molecular Biology and Evolution. 2021;38:5825–5829. doi: 10.1093/molbev/msab293. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Carlile TM, Martinez NM, Schaening C, Su A, Bell TA, Zinshteyn B, Gilbert WV. mRNA structure determines modification by pseudouridine synthase 1. Nature Chemical Biology. 2019;15:966–974. doi: 10.1038/s41589-019-0353-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cerneckis J, Cui Q, He C, Yi C, Shi Y. Decoding pseudouridine: an emerging target for therapeutic development. Trends in Pharmacological Sciences. 2022;43:522–535. doi: 10.1016/j.tips.2022.03.008. [ DOI ] [ PubMed ] [ Google Scholar ] Chihara K, Bischler T, Barquist L, Monzon VA, Noda N, Vogel J, Tsuneda S. Conditional Hfq association with small noncoding RNAs in Pseudomonas aeruginosa revealed through comparative UV cross-linking immunoprecipitation followed by high-throughput sequencing. mSystems. 2019;4:mSystems. doi: 10.1128/mSystems.00590-19. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Choe D, Szubin R, Poudel S, Sastry A, Song Y, Lee Y, Cho S, Palsson B, Cho B-K. RiboRid: A low cost, advanced, and ultra-efficient method to remove ribosomal RNA for bacterial transcriptomics. PLOS Genetics. 2021;17:e1009821. doi: 10.1371/journal.pgen.1009821. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Chou HJ, Donnard E, Gustafsson HT, Garber M, Rando OJ. Transcriptome-wide analysis of roles for tRNA modifications in translational regulation. Molecular Cell. 2017;68:978–992. doi: 10.1016/j.molcel.2017.11.002. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Commichau FM, Rothe FM, Herzberg C, Wagner E, Hellwig D, Lehnik-Habrink M, Hammer E, Völker U, Stülke J. Novel activities of glycolytic enzymes in Bacillus subtilis: interactions with essential proteins involved in mRNA processing. Molecular & Cellular Proteomics. 2009;8:1350–1360. doi: 10.1074/mcp.M800546-MCP200. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Dai Q, Zhang LS, Sun HL, Pajdzik K, Yang L, Ye C, Ju CW, Liu S, Wang Y, Zheng Z, Zhang L, Harada BT, Dou X, Irkliyenko I, Feng X, Zhang W, Pan T, He C. Quantitative sequencing using BID-seq uncovers abundant pseudouridines in mammalian mRNA at base resolution. Nature Biotechnology. 2023;41:344–354. doi: 10.1038/s41587-022-01505-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] de Crécy-Lagard V, Boccaletto P, Mangleburg CG, Sharma P, Lowe TM, Leidel SA, Bujnicki JM. Matching tRNA modifications in humans to their known and predicted enzymes. Nucleic Acids Research. 2019;47:2143–2159. doi: 10.1093/nar/gkz011. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] de Crécy-Lagard V, Jaroch M. Functions of bacterial tRNA modifications: from ubiquity to diversity. Trends in Microbiology. 2021;29:41–53. doi: 10.1016/j.tim.2020.06.010. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Dos Santos RF, Arraiano CM, Andrade JM. New molecular interactions broaden the functions of the RNA chaperone Hfq. Current Genetics. 2019;65:1313–1319. doi: 10.1007/s00294-019-00990-y. [ DOI ] [ PubMed ] [ Google Scholar ] Dylan-LT PseU_NN. swh:1:rev:229aabf75073512903eefe9c8a39dabfc502b334Software Heritage. 2026 https://archive.softwareheritage.org/swh:1:dir:146a633819c82e4e27fdb612acf39129e531f985;origin=https://github.com/Dylan-LT/pseU_NN;visit=swh:1:snp:a2803fe7cec74826e9a0af7d8f5554732c54d590;anchor=swh:1:rev:229aabf75073512903eefe9c8a39dabfc502b334 Ehling-Schulz M, Lereclus D, Koehler TM. The bacillus cereus group: bacillus species with pathogenic potential. Microbiology Spectrum. 2019;7:2018. doi: 10.1128/microbiolspec.gpp3-0032-2018. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Emms DM, Kelly S. OrthoFinder: phylogenetic orthology inference for comparative genomics. Genome Biology. 2019;20:238. doi: 10.1186/s13059-019-1832-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Engelhardt J, Heyne S, Will S, Reiche K. RNAclust Documentation. Bioinformatics Group, Department of Computer Science, University of Leipzig; 2010. [ Google Scholar ] Eyler DE, Franco MK, Batool Z, Wu MZ, Dubuke ML, Dobosz-Bartoszek M, Jones JD, Polikanov YS, Roy B, Koutmou KS. Pseudouridinylation of mRNA coding sequences alters translation. PNAS. 2019;116:23068–23074. doi: 10.1073/pnas.1821754116. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Fang X, Zhao R, Wang Y, Sun M, Xu J, Long S, Mo J, Liu H, Li X, Wang F, Zhou X, Weng X. A bisulfite-assisted and ligation-based qPCR amplification technology for locus-specific pseudouridine detection at base resolution. Nucleic Acids Research. 2024;52:e49. doi: 10.1093/nar/gkae344. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Gouy M, Gautier C. Codon usage in bacteria: correlation with gene expressivity. Nucleic Acids Research. 1982;10:7055–7074. doi: 10.1093/nar/10.22.7055. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hoang C, Ferré-D’Amaré AR. Cocrystal structure of a tRNA Psi55 pseudouridine synthase: nucleotide flipping by an RNA-modifying enzyme. Cell. 2001;107:929–939. doi: 10.1016/s0092-8674(01)00618-3. [ DOI ] [ PubMed ] [ Google Scholar ] Hoernes TP, Clementi N, Faserl K, Glasner H, Breuker K, Lindner H, Hüttenhofer A, Erlacher MD. Nucleotide modifications within bacterial messenger RNAs regulate their translation and are able to rewire the genetic code. Nucleic Acids Research. 2016;44:852–862. doi: 10.1093/nar/gkv1182. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hua C, Huang J, Sun Y, Wang T, Li Y, Cui Z, Deng X. Hfq mediates transcriptome-wide RNA structurome reprogramming under virulence-inducing conditions in a phytopathogen. Cell Reports. 2024;43:114544. doi: 10.1016/j.celrep.2024.114544. [ DOI ] [ PubMed ] [ Google Scholar ] Huynh TV, Dahlbeck D, Staskawicz BJ. Bacterial blight of soybean: regulation of a pathogen gene determining host cultivar specificity. Science. 1989;245:1374–1377. doi: 10.1126/science.2781284. [ DOI ] [ PubMed ] [ Google Scholar ] Ishida K, Kunibayashi T, Tomikawa C, Ochi A, Kanai T, Hirata A, Iwashita C, Hori H. Pseudouridine at position 55 in tRNA controls the contents of other modified nucleotides for low-temperature adaptation in the extreme-thermophilic eubacterium Thermus thermophilus. Nucleic Acids Research. 2011;39:2304–2318. doi: 10.1093/nar/gkq1180. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Karijolich J, Yu YT. Converting nonsense codons into sense codons by targeted pseudouridylation. Nature. 2011;474:395–398. doi: 10.1038/nature10165. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Karikó K, Muramatsu H, Welsh FA, Ludwig J, Kato H, Akira S, Weissman D. Incorporation of pseudouridine into mRNA yields superior nonimmunogenic vector with increased translational capacity and biological stability. Molecular Therapy. 2008;16:1833–1840. doi: 10.1038/mt.2008.200. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kerr KG, Snelling AM. Pseudomonas aeruginosa: a formidable and ever-present adversary. The Journal of Hospital Infection. 2009;73:338–344. doi: 10.1016/j.jhin.2009.04.020. [ DOI ] [ PubMed ] [ Google Scholar ] King EO, Ward MK, Raney DE. Two simple media for the demonstration of pyocyanin and fluorescin. The Journal of Laboratory and Clinical Medicine. 1954;44:301–307. [ PubMed ] [ Google Scholar ] Kong W, Zhao J, Kang H, Zhu M, Zhou T, Deng X, Liang H. ChIP-seq reveals the global regulator AlgR mediating cyclic di-GMP synthesis in Pseudomonas aeruginosa. Nucleic Acids Research. 2015;43:8268–8282. doi: 10.1093/nar/gkv747. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Krutyhołowa R, Zakrzewski K, Glatt S. Charging the code - tRNA modification complexes. Current Opinion in Structural Biology. 2019;55:138–146. doi: 10.1016/j.sbi.2019.03.014. [ DOI ] [ PubMed ] [ Google Scholar ] Lange SJ, Maticzka D, Möhl M, Gagnon JN, Brown CM, Backofen R. Global or local? Predicting secondary structure and accessibility in mRNAs. Nucleic Acids Research. 2012;40:5215–5226. doi: 10.1093/nar/gks181. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Leppik M, Liiv A, Remme J. Random pseuoduridylation in vivo reveals critical region of Escherichia coli 23S rRNA for ribosome assembly. Nucleic Acids Research. 2017;45:6098–6108. doi: 10.1093/nar/gkx160. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Levi O, Arava YS. Pseudouridine-mediated translation control of mRNA by methionine aminoacyl tRNA synthetase. Nucleic Acids Research. 2021;49:432–443. doi: 10.1093/nar/gkaa1178. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Li X, Zhu P, Ma S, Song J, Bai J, Sun F, Yi C. Chemical pulldown reveals dynamic pseudouridylation of the mammalian transcriptome. Nature Chemical Biology. 2015;11:592–597. doi: 10.1038/nchembio.1836. [ DOI ] [ PubMed ] [ Google Scholar ] Li H, Wang G, Ye C, Zou Z, Jiang B, Yang F, He K, Ju C, Zhang L, Gao B, Liu S, Chen Y, Zhang J, He C. Quantitative RNA pseudouridine maps reveal multilayered translation control through plant rRNA, tRNA and mRNA pseudouridylation. Nature Plants. 2025;11:234–247. doi: 10.1038/s41477-024-01894-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Liang ST, Xu YC, Dennis P, Bremer H. mRNA composition and control of bacterial gene expression. Journal of Bacteriology. 2000;182:3037–3044. doi: 10.1128/JB.182.11.3037-3044.2000. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Machnicka MA, Olchowik A, Grosjean H, Bujnicki JM. Distribution and frequencies of post-transcriptional modifications in tRNAs. RNA Biology. 2014;11:1619–1629. doi: 10.4161/15476286.2014.992273. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Martinez NM, Su A, Burns MC, Nussbacher JK, Schaening C, Sathe S, Yeo GW, Gilbert WV. Pseudouridine synthases modify human pre-mRNA co-transcriptionally and affect pre-mRNA processing. Molecular Cell. 2022;82:645–659. doi: 10.1016/j.molcel.2021.12.023. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Nakamoto MA, Lovejoy AF, Cygan AM, Boothroyd JC. MRNA pseudouridylation affects RNA metabolism in the parasite Toxoplasma gondii. RNA. 2017;23:1834–1849. doi: 10.1261/rna.062794.117. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] O’Connor M, Leppik M, Remme J. Pseudouridine-free Escherichia coli ribosomes. Journal of Bacteriology. 2018;200:e00540-17. doi: 10.1128/JB.00540-17. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Pan H, Agarwalla S, Moustakas DT, Finer-Moore J, Stroud RM. Structure of tRNA pseudouridine synthase TruB and its RNA complex: RNA recognition through a combination of rigid docking and induced fit. PNAS. 2003;100:12648–12653. doi: 10.1073/pnas.2135585100. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Rodell R, Robalin N, Martinez NM. Why U matters: detection and functions of pseudouridine modifications in mRNAs. Trends in Biochemical Sciences. 2024;49:12–27. doi: 10.1016/j.tibs.2023.10.008. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Roovers M, Hale C, Tricot C, Terns MP, Terns RM, Grosjean H, Droogmans L. Formation of the conserved pseudouridine at position 55 in archaeal tRNA. Nucleic Acids Research. 2006;34:4293–4301. doi: 10.1093/nar/gkl530. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Roundtree IA, Evans ME, Pan T, He C. Dynamic RNA modifications in gene expression regulation. Cell. 2017;169:1187–1200. doi: 10.1016/j.cell.2017.05.045. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Safra M, Nir R, Farouq D, Vainberg Slutskin I, Schwartz S. TRUB1 is the predominant pseudouridine synthase acting on mammalian mRNA via a predictable and conserved code. Genome Research. 2017;27:393–406. doi: 10.1101/gr.207613.116. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Salgado H, Moreno-Hagelsieb G, Smith TF, Collado-Vides J. Operons in Escherichia coli: genomic analyses and predictions. PNAS. 2000;97:6652–6657. doi: 10.1073/pnas.110147297. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Salgado H, Gama-Castro S, Peralta-Gil M, Díaz-Peredo E, Sánchez-Solano F, Santos-Zavaleta A, Martínez-Flores I, Jiménez-Jacinto V, Bonavides-Martínez C, Segura-Salazar J, Martínez-Antonio A, Collado-Vides J. RegulonDB (version 5.0): Escherichia coli K-12 transcriptional regulatory network, operon organization, and growth conditions. Nucleic Acids Research. 2006;34:D394–D397. doi: 10.1093/nar/gkj156. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Sato K, Akiyama M, Sakakibara Y. RNA secondary structure prediction using deep learning with thermodynamic integration. Nature Communications. 2021;12:941. doi: 10.1038/s41467-021-21194-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Schaening-Burgos C, LeBlanc H, Fagre C, Li GW, Gilbert WV. RluA is the major mRNA pseudouridine synthase in Escherichia coli. PLOS Genetics. 2024;20:e1011100. doi: 10.1371/journal.pgen.1011100. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Schultz SK, Katanski CD, Halucha M, Pena N, Fahlman RP, Pan T, Kothe U. Modifications in the T arm of tRNA globally determine tRNA maturation, function, and cellular fitness. PNAS. 2024;121:e2401154121. doi: 10.1073/pnas.2401154121. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Semmler ABT, Whitchurch CB, Leech AJ, Mattick JS. Identification of a novel gene, fimV, involved in twitching motility in Pseudomonas aeruginosa. Microbiology. 2000;146 ( Pt 6):1321–1332. doi: 10.1099/00221287-146-6-1321. [ DOI ] [ PubMed ] [ Google Scholar ] Sloan KE, Warda AS, Sharma S, Entian KD, Lafontaine DLJ, Bohnsack MT. Tuning the ribosome: the influence of rRNA modification on eukaryotic ribosome biogenesis and function. RNA Biology. 2017;14:1138–1152. doi: 10.1080/15476286.2016.1259781. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Sobrero P, Valverde C. The bacterial protein Hfq: much more than a mere RNA-binding factor. Critical Reviews in Microbiology. 2012;38:276–299. doi: 10.3109/1040841X.2012.664540. [ DOI ] [ PubMed ] [ Google Scholar ] Song Z, Huang D, Song B, Chen K, Song Y, Liu G, Su J, Magalhães JPD, Rigden DJ, Meng J. Attention-based multi-label neural networks for integrated prediction and interpretation of twelve widely occurring RNA modifications. Nature Communications. 2021;12:4011. doi: 10.1038/s41467-021-24313-3. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Trouillon J, Han K, Attrée I, Lory S. The core and accessory Hfq interactomes across Pseudomonas aeruginosa lineages. Nature Communications. 2022;13:1258. doi: 10.1038/s41467-022-28849-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Veerareddygari GR, Singh SK, Mueller EG. The pseudouridine synthases proceed through a glycal intermediate. Journal of the American Chemical Society. 2016;138:7852–7855. doi: 10.1021/jacs.6b04491. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Ventura M, Canchaya C, van Sinderen D, Fitzgerald GF, Zink R. Bifidobacterium lactis DSM 10140: identification of the atp (atpBEFHAGDC) operon and analysis of its genetic structure, characteristics, and phylogeny. Applied and Environmental Microbiology. 2004;70:3110–3121. doi: 10.1128/AEM.70.5.3110-3121.2004. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Wang F, Alinejad‐Rokny H, Lin J, Gao T, Chen X, Zheng Z, Meng L, Li X, Wong K. A lightweight framework for chromatin loop detection at the single‐cell level. Advanced Science. 2023;10:2303502. doi: 10.1002/advs.202303502. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Wyres KL, Lam MMC, Holt KE. Population genomics of Klebsiella pneumoniae. Nature Reviews. Microbiology. 2020;18:344–359. doi: 10.1038/s41579-019-0315-1. [ DOI ] [ PubMed ] [ Google Scholar ] Xin XF, Kvitko B, He SY. Pseudomonas syringae: what it takes to be a pathogen. Nature Reviews. Microbiology. 2018;16:316–328. doi: 10.1038/nrmicro.2018.17. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Yu CT, Allen FW. Studies on an isomer of uridine isolated from ribonucleic acids. Biochimica et Biophysica Acta. 1959;32:393–406. doi: 10.1016/0006-3002(59)90612-2. [ DOI ] [ PubMed ] [ Google Scholar ] Yu M, Abnousi A, Zhang Y, Li G, Lee L, Chen Z, Fang R, Lagler TM, Yang Y, Wen J, Sun Q, Li Y, Ren B, Hu M. SnapHiC: a computational pipeline to identify chromatin loops from single-cell Hi-C data. Nature Methods. 2021;18:1056–1059. doi: 10.1038/s41592-021-01231-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhang LS, Ye C, Ju CW, Gao B, Feng X, Sun HL, Wei J, Yang F, Dai Q, He C. BID-seq for transcriptome-wide quantitative sequencing of mRNA pseudouridine at base resolution. Nature Protocols. 2024;19:517–538. doi: 10.1038/s41596-023-00917-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhao Y, Li MC, Konaté MM, Chen L, Das B, Karlovich C, Williams PM, Evrard YA, Doroshow JH, McShane LM. TPM, FPKM, or normalized counts? A comparative study of quantification measures for the analysis of RNA-seq data from the NCI patient-derived models repository. Journal of Translational Medicine. 2021;19:269. doi: 10.1186/s12967-021-02936-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zhao Y, Rai J, Li H. Regulation of translation by ribosomal RNA pseudouridylation. Science Advances. 2023;9:adg8190. doi: 10.1126/sciadv.adg8190. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] eLife. doi: 10.7554/eLife.107545.3.sa0 eLife Assessment Camilo Perez 1 Author information Article notes Copyright and License information 1 University of Georgia, United States Roles Camilo Perez : Reviewing Editor Keywords: Solid Keywords: Valuable PMC Copyright notice This study illustrates a valuable application of BID-seq to bacterial RNA, allowing transcriptome-wide mapping of pseudouridine modifications across various bacterial species. The evidence presented includes solid data and analyses that would benefit from additional experimental validation. The work will interest a specialized audience involved in RNA biology. eLife. doi: 10.7554/eLife.107545.3.sa1 Reviewer #1 (Public review): Anonymous Author information Copyright and License information Roles Anonymous : Reviewer PMC Copyright notice Summary: The manuscript by Xu et al. reported base-resolution mapping of RNA pseudouridylation in five bacterial species, utilizing recently developed BID-seq. They detected pseudouridine (Ψ) in bacterial rRNA, tRNA, and mRNA, and found growth phase-dependent Ψ changes in tRNA and mRNA. They then focused on mRNA and conducted comparative analysis of Ψ profiles across different bacterial species. Finally, they developed a deep learning model to predict Ψ sites based on RNA sequence and structure. Strengths: This is the first comprehensive Ψ map across multiple bacterial species, and systematically reveals Ψ profiles in rRNA, tRNA, and mRNA under exponential and stationary growth conditions. It provides a valuable resource for future functional studies of Ψ in bacteria. Weaknesses: Ψ is highly abundant on non-coding RNA such as rRNA and rRNA, while its level on mRNA is very low. The manuscript focuses primarily on Ψ on mRNA, which is prone to false positives. Many conclusions in the manuscript are speculative, based solely on the sequencing data, but not supported by additional experiments. eLife. doi: 10.7554/eLife.107545.3.sa2 Reviewer #2 (Public review): Anonymous Author information Copyright and License information Roles Anonymous : Reviewer PMC Copyright notice Summary: In this study, Xu et al. present a transcriptome-wide, single-base resolution map of RNA pseudouridine modifications across evolutionarily diverse bacterial species using an adapted form of BID-Seq. By optimizing the method for bacterial RNA, the authors successfully mapped modifications in rRNA, tRNA, and, importantly, mRNA across both exponential and stationary growth phases. They uncover evolutionarily conserved Ψ motifs, dynamic Ψ regulation tied to bacterial growth state, and propose functional links between pseudouridylation and bacterial transcript stability, translation, and RNA-protein interactions. To extend these findings, they develop a deep learning model that predicts pseudouridine sites from local sequence and structural features. Strengths: The authors provide a valuable resource: a comprehensive Ψ atlas for bacterial systems, spanning hundreds of mRNAs and multiple species. The work addresses a gap in the field - our limited understanding of bacterial epitranscriptomics, by establishing both the method and datasets for exploring post-transcriptional modifications. Weaknesses: The main limitation of the study is that most functional claims (i.e. translation efficiency, mRNA stability, and RNA-binding protein interactions) are based on correlative evidence. While suggestive, these inferences would be significantly strengthened by targeted perturbation of specific Ψ synthases or direct biochemical validation of proposed RNA-protein interactions (e.g., with Hfq). Additionally, the GNN prediction model is a notable advance. eLife. doi: 10.7554/eLife.107545.3.sa3 Reviewer #3 (Public review): Anonymous Author information Copyright and License information Roles Anonymous : Reviewer PMC Copyright notice Summary: This study aimed to investigate pseudouridylation across various RNA species in multiple bacterial strains using an optimized BID-seq approach. It examined both conserved and divergent modification patterns, the potential functional roles of pseudouridylation, and its dynamic regulation across different growth conditions. Strengths: The authors optimized the BID-seq method and applied this important technique to bacterial systems, identifying multiple pseudouridylation sites across different species. They investigated the distribution of these modifications, associated sequence motifs, their dynamics across growth phases, and potential functional roles. These data are of great interest to researchers focused on understanding the significance of RNA modifications, particularly mRNA modifications, in bacteria. eLife. 2026 Jun 4;14:RP107545. doi: 10.7554/eLife.107545.3.sa4 Author response Letong Xu 1 , Shenghai Shen 2 , Yizhou Zhang 3 , Zhihao Guo 4 , Beifang Lu 5 , Jiadai Huang 6 , Runsheng Li 7 , Yitong Shen 8 , Lisheng Zhang 9 , Xin Deng 10 Author information Article notes Copyright and License information 1 City University of Hong Kong, Hong Kong, Hong Kong 2 Division of Life Science, The Hong Kong University of Science and Technology, China, Hong Kong 3 City University of Hong Kong, Hong Kong, Hong Kong 4 City University of Hong Kong, Hong Kong, Hong Kong 5 City University of Hong Kong, Hong Kong, Hong Kong 6 City University of Hong Kong, Hong Kong, Hong Kong 7 City University of Hong Kong, Hong Kong, Hong Kong 8 Department of Chemistry, The Hong Kong University of Science and Technology, Hong Kong, Hong Kong 9 Department of Chemistry, The Hong Kong University of Science and Technology, Hong Kong, Hong Kong 10 City University of Hong Kong, Hong Kong, Hong Kong Roles Letong Xu : Author Shenghai Shen : Author Yizhou Zhang : Author Zhihao Guo : Author Beifang Lu : Author Jiadai Huang : Author Runsheng Li : Author Yitong Shen : Author Lisheng Zhang : Author Xin Deng : Author Collection date 2026. PMC Copyright notice The following is the authors’ response to the original reviews. Public Reviews: Reviewer #1 (Public review): The manuscript by Xu et al. reported base-resolution mapping of RNA pseudouridylation in five bacterial species, utilizing recently developed BID-seq. They detected pseudouridine (Ψ) in bacterial rRNA, tRNA, and mRNA, and found growth phase-dependent Ψ changes in tRNA and mRNA. They then focused on mRNA and conducted a comparative analysis of Ψ profiles across different bacterial species. Finally, they developed a deep learning model to predict Ψ sites based on RNA sequence and structure. This is the first comprehensive Ψ map across multiple bacterial species, and systematically reveals Ψ profiles in rRNA, tRNA, and mRNA under exponential and stationary growth conditions. It provides a valuable resource for future functional studies of Ψ in bacteria. We thank Reviewer 1 for the supportive and positive comments, particularly for highlighting the novelty and value of our comprehensive pseudouridine landscapes across multiple bacterial species as a valuable resource for the scientific community. Ψ is highly abundant on non-coding RNA such as rRNA and tRNA, while its level on mRNA is very low. The manuscript focuses primarily on mRNA, which raises questions about the data quality and the rigor of the analysis. Many conclusions in the manuscript are speculative, based solely on the sequencing data but not supported by additional experiments. We appreciate the insightful comments of Reviewer 1. We fully agree that Ψ is highly abundant on rRNA and tRNA, while its fractions on mRNA are generally lower. Ψ is highly conserved at specific positions in rRNA and tRNA, such as Ψ within tRNA T‑arm (position 55), where it plays essential roles in tRNA structural folding, tRNA stability, and mRNA translation, across plants, mammals, and bacteria[1–3]. However, most Ψ sites in mRNA exhibit lower fractions compared to rRNA and tRNA. This phenomenon is also widely observed in HeLa cell mRNA and plant mRNA, as evidenced by bisulfite-induced deletion sequencing and 2-bromoacrylamide-assisted cyclization sequencing[3–5]. In bacteria, the modifications on mRNA are harder to map and quantify, due to its low abundance in total RNA and difficulty in bacterial rRNA removal. This highlights the significance of our study. To prove our data quality and analytical rigor, we first present the most convincing sites in bacteria, as benchmark sites. Specifically, we detected 9 out of 10 known conserved pseudouridine (Ψ) sites in E. coli across two biological replicates [6], displaying notable modification fraction. Ψ516 site in E. coli 16S rRNA, which serves as a benchmark site, consistently exhibited a high modification fraction (~100%) under multiple growth conditions, underscoring the robustness of our method. In other strains, we also observed conserved 16S rRNA Ψ sites. To further demonstrate strong reproducibility and sensitivity. We selected three positive Ψ sites from two independent biological replicates for experimental validation, alongside one negative control site, using pseU‑TRACE method[6]. Ct values were first normalized to the corresponding Ct value of the negative control site, and the treated samples were then further normalized to their corresponding input controls (new Supplementary Fig. 2e). Four Ψ sites were tested with pseU‑TRACE: Ψ site at position 944 on 23S rRNA, a negative control site located within guaA gene, a Ψ site within clpV1 gene, and an intergenic Ψ site located between guaA and guaB genes. We successfully validated these Ψ sites in P. aeruginosa . The detailed pseU‑TRACE experimental procedures and corresponding data figures have been added to the revised manuscript, in either Results or Methods sections (Line 171-175, 594–617). Previous transcriptome-wide mapping of Ψ have primarily relied on CMC-based methods to induce RT truncation signatures at the modified sites, exhibiting a limited Ψ detection sensitivity caused by low labeling efficiency[5]. In contrast, BID-seq method used in this study provides substantially higher sensitivity of Ψ detection, particularly the low-stoichiometry Ψ sites within mRNA. The high reliability and quantitative performance of BID-seq have been extensively validated in prior work using mammalian cells and synthetic Ψ-containing oligonucleotides[4]. To further ensure robustness and minimize false positives—when identifying low-level mRNA Ψ sites through bioinformatic analysis—we have applied stringent and uniform filtration criteria to all candidate sites on mRNA (new Supplementary Table 1): (1) Total sequencing coverage >20 reads in both ‘Treated’ (BID-seq; Σd t > 20) and ‘Input’ libraries (Σd i > 20); (2) An average deletion count >5 in ‘Treated’ libraries; (3) An average modification fraction >0.02 (2%) in ‘Treated’ libraries; (4) A deletion ratio in ‘Treated’ libraries at least two-fold higher than that in ‘Input’ libraries. Sites with a Ψ stoichiometry >0.5 (50%) were classified as highly modified. These filtration criteria have now been explicitly described in Methods section (Lines 739–745). We strictly adhered to these Ψ site identification standards, leading to all subsequent analysis and functional studies. Finally, to address concerns regarding reproducibility, we calculated mRNA Ψ site overlap and correlation of Ψ fractions, between two biological replicates, which has been presented in (new Supplementary Fig. 2a,d). Overall, we have revised the manuscript to clarify these methodological strengths, and validate mRNA Ψ detection. We also tone down all speculative conclusions, with more clear linkage to the actual sequencing data, which await future functional validation. Reviewer #2 (Public review): Summary: In this study, Xu et al. present a transcriptome-wide, single-base resolution map of RNA pseudouridine modifications across evolutionarily diverse bacterial species using an adapted form of BID-Seq. By optimizing the method for bacterial RNA, the authors successfully mapped modifications in rRNA, tRNA, and, importantly, mRNA across both exponential and stationary growth phases. They uncover evolutionarily conserved Ψ motifs, dynamic Ψ regulation tied to bacterial growth state, and propose functional links between pseudouridylation and bacterial transcript stability, translation, and RNA-protein interactions. To extend these findings, they develop a deep learning model that predicts pseudouridine sites from local sequence and structural features. Strengths: The authors provide a valuable resource: a comprehensive Ψ atlas for bacterial systems, spanning hundreds of mRNAs and multiple species. The work addresses a gap in the field - our limited understanding of bacterial epitranscriptomics, by establishing both the method and datasets for exploring post-transcriptional modifications. We thank Reviewer 2 for the supportive and positive comments. We appreciate the reviewer’s recognition of the novelty and value of our work in providing a comprehensive pseudouridine atlas across multiple bacterial species. Weaknesses: The main limitation of the study is that most functional claims (i.e., translation efficiency, mRNA stability, and RNA-binding protein interactions) are based on correlative evidence. While suggestive, these inferences would be significantly strengthened by targeted perturbation of specific Ψ synthases or direct biochemical validation of proposed RNA-protein interactions (e.g., with Hfq). We thank Reviewer 2 for the constructive feedback. We fully agree that our functional claims regarding translation efficiency, mRNA stability, and RNA-binding protein interactions rely primarily on correlative evidence from existing datasets rather than a direct experimental validation. We agree that the perturbation of specific pseudouridine synthases and direct biochemical validation of proposed RNA-protein interactions (for instance, Hfq) would substantially strengthen the conclusions on bacterial Ψ function. In Discussion section, we have added a discussion on this limitation of our current study (Line 517–523). Considering the scope of our current work, we anticipate such validation experiments in future research. Additionally, the GNN prediction model is a notable advance, but methodological details are insufficient to reproduce or assess its robustness. In response to methodological concerns regarding our pseU_GNN prediction model, we have undertaken substantial improvements to address these issues comprehensively. We have updated the complete codebase on GitHub ( https://github.com/Dylan-LT/pseU_NN.git ) with comprehensive documentation and a user-friendly prediction tool specifically designed for Ψ site prediction across the four bacterial species examined in this study. We further systematically evaluated multiple neural network architectures and implemented critical architectural refinements. Specifically, we incorporated bidirectional LSTM (bid-LSTM) layers upstream of the transformer block to more effectively capture sequential dependencies and contextual information in RNA sequences. This enhanced architecture demonstrates substantially improved predictive performance, achieving an AUC-ROC of 0.89 on independent test datasets using 41-nucleotide input sequences (new Figure 6). We have revised Figure 6 and Supplementary Fig. 7, along with their corresponding content and figure legends (Lines 428-430, 434–436, 440-447, 1065-1073), to reflect these architectural improvements and performance enhancements. We have detailed the methods part (Lines 679–708), including model architecture, validation methods and evaluation score calculation. Additionally, we have provided detailed documentation of the evaluation score calculation methodology to ensure reproducibility and transparency. Reviewer #3 (Public review): Summary: This study aimed to investigate pseudouridylation across various RNA species in multiple bacterial strains using an optimized BID-seq approach. It examined both conserved and divergent modification patterns, the potential functional roles of pseudouridylation, and its dynamic regulation across different growth conditions. Strengths: The authors optimized the BID-seq method and applied this important technique to bacterial systems, identifying multiple pseudouridylation sites across different species. They investigated the distribution of these modifications, associated sequence motifs, their dynamics across growth phases, and potential functional roles. These data are of great interest to researchers focused on understanding the significance of RNA modifications, particularly mRNA modifications, in bacteria. We thank Reviewer 3 for the supportive and positive assessment. We are particularly grateful for the reviewer’s acknowledgment of the value of our analyses on modification distribution, sequence motifs, growth‑phase dynamics, and potential functional roles, which we hope will be of broad interest to researchers studying bacterial RNA modifications, particularly mRNA Ψ. Weaknesses: (1) The reliability of BID-seq data is questionable due to a lack of experimental validations. We thank Reviewer 3 for the constructive feedback. We have undertaken comprehensive revisions to address the concerns regarding manuscript structure and information organization. We have incorporated pseU‑TRACE experiments and data quality results to provide orthogonal validation of Ψ detection, strengthening the robustness of our work. Here we copied the response in Reviewer 1 section: “To further demonstrate strong reproducibility and sensitivity. We selected three positive Ψ sites from two independent biological replicates for experimental validation, alongside one negative control site, using pseU‑TRACE method[6]. Ct values were first normalized to the corresponding Ct value of the negative control site, and the treated samples were then further normalized to their corresponding input controls (new Supplementary Fig. 2e). Four Ψ sites were tested with pseU‑TRACE: Ψ site at position 944 on 23S rRNA, a negative control site located within guaA gene, a Ψ site within clpV1 gene, and an intergenic Ψ site located between guaA and guaB genes. We successfully validated these Ψ sites in P. aeruginosa . The detailed pseU‑TRACE experimental procedures and corresponding data figures have been added to the revised manuscript, in either Results or Methods sections (Line 171-175, 594–617).” (2) The manuscript is not well-written, and the presented work shows a major lack of scientific rigor, as several key pieces of information are missing. We thank Reviewer 3 for the suggestion. We restructured the main text to present a clearer logical flow, with key objectives (Lines 83–96, 171–175, 428–447, 517-523) explicitly stated in Introduction section and Conclusions section, with data figures directly addressing these stated aims (Supplementary Fig. 1–7). (3) The manuscript's organization requires significant improvement, and numerous instances of missing or inconsistent information make it difficult to understand the key objectives and conclusions of the study. We thank Reviewer 3 for the constructive feedback. All supplementary figures have been updated with detailed figure legend, methodology description, and consistent formatting. We also systematically inspected and resolved instances of missing or inconsistent information throughout the main text and supplementary materials (Supplementary Fig. 1–7; Supplementary Table 1). To enhance computational reproducibility, we have updated our GitHub repository with well-documented code and developed user-friendly prediction tools for Ψ identification across the four bacterial species examined in this study. (4) The rationale for selecting specific bacterial species is not clearly explained, and the manuscript lacks a systematic comparison of pseudouridylation among these species. We thank Reviewer 3 for the constructive feedback. The bacterial species analyzed in this study were selected based on both diversity and significance. K. pneumoniae , B. cereus , and P. aeruginosa are top model human pathogens responsible for a wide range of clinically significant infections, yet transcriptome-wide pseudouridylation has not been systematically explored in these organisms[7–9]. P. syringae , the most important model plant pathogen, was included to extend our analysis beyond human pathogens and to examine Ψ modification in a distinct ecological and evolutionary context, where epitranscriptomic regulation also remains poorly characterized[10]. Importantly, the selected species represent both Gram-positive ( B. cereus ) and Gram-negative ( K. pneumoniae , P. aeruginosa , and P. syringae ) bacteria, spanning substantial differences in genome size, GC content, lifestyle, and pathogenic strategies. This diversity enables a comparative framework for examining conserved and species-specific pseudouridylation patterns across bacterial lineages. To address the reviewer’s concern, we have revised the manuscript to more clearly articulate the rationale for species selection and have added a comparative analysis highlighting similarities and differences in Ψ site distribution and modification levels among these species (Lines 83–96). We systematically compared Ψ-carrying motif for analyzing sequence context of 10 bases flanking Ψ sites in bacterial mRNA, with Supplementary Fig. 4 added. Reference (1) Leppik, M., Liiv, A. & Remme, J. Random pseuoduridylation in vivo reveals critical region of Escherichia coli 23S rRNA for ribosome assembly. Nucleic Acids Res. 45, (2017). (2) Rajan, K. S. et al. A single pseudouridine on rRNA regulates ribosome structure and function in the mammalian parasite Trypanosoma brucei . Nat. Commun. 14, (2023). (3) Li, H. et al. Quantitative RNA pseudouridine maps reveal multilayered translation control through plant rRNA, tRNA and mRNA pseudouridylation. Nat. Plants 11, 234–247 (2025). (4) Dai, Q. et al. Quantitative sequencing using BID-seq uncovers abundant pseudouridines in mammalian mRNA at base resolution. Nat. Biotechnol. 41, 344–354 (2023). (5) Xu, H. et al. Absolute quantitative and base-resolution sequencing reveals comprehensive landscape of pseudouridine across the human transcriptome. Nat. Methods 21, 2024–2033 (2024). (6) Fang, X. et al. A bisulfite-assisted and ligation-based qPCR amplification technology for locus-specific pseudouridine detection at base resolution. Nucleic Acids Res. 52, (2024). (7) Wyres, K. L., Lam, M. M. C. & Holt, K. E. Population genomics of Klebsiella pneumoniae. Nature Reviews Microbiology vol. 18 Preprint at https://doi.org/10.1038/s41579-019-0315-1 (2020). (8) Kerr, K. G. & Snelling, A. M. Pseudomonas aeruginosa : a formidable and ever-present adversary. Journal of Hospital Infection vol. 73 Preprint at https://doi.org/10.1016/j.jhin.2009.04.020 (2009). (9) Ehling-Schulz, M., Lereclus, D. & Koehler, T. M. The Bacillus cereus Group: Bacillus Species with Pathogenic Potential . Microbiol. Spectr. 7, (2019). (10) Xin, X. F., Kvitko, B. & He, S. Y. Pseudomonas syringae: What it takes to be a pathogen. Nature Reviews Microbiology vol. 16 Preprint at https://doi.org/10.1038/nrmicro.2018.17 (2018). Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Citations Deng X, Xu L. 2026. Quantitative RNA pseudouridine maps reveal functional insights into pseudouridylation in bacteria. NCBI Gene Expression Omnibus. GSE292335 Supplementary Materials Supplementary file 1. Modification sites. elife-107545-supp1.csv (452.1KB, csv) Supplementary file 2. Primer. elife-107545-supp2.csv (774B, csv) MDAR checklist elife-107545-mdarchecklist1.docx (88.7KB, docx) Data Availability Statement All sequencing data files have been submitted to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database with the reference code of GSE292335 . The following dataset was generated: Deng X, Xu L. 2026. Quantitative RNA pseudouridine maps reveal functional insights into pseudouridylation in bacteria. NCBI Gene Expression Omnibus. GSE292335 Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd ACTIONS View on publisher site PDF (5.8 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Comparative genomics approaches to identify genomic regions associated with the antimicrobial activity of Pseudomonas protegens PBL3.#653568
• Pseudouridine increases ribosome stability in a thermophilic eukaryote#147551
• Alternative nucleic acid structures in microbial biology: regulatory functions and antimicrobial opportunities.#452930
• Dataset: Repurposing Chlorhexidine as a Potential Anti-Virulence Agent in Pseudomonas aeruginosa [Molecular Docking Raw Data and Analysis]#658809
• Beyond protist generalizations: different evolutionary trajectories in closely related microbial species.#481041