Growth-dependent tRNA reprogramming and codon bias link translation to metabolic state in Enterococcus faecalis

ABSTRACT

Enterococcus faecalis is a gram-positive commensal bacterium of the human gut microbiome and an opportunistic pathogen responsible for many hospital-acquired infections. Despite the clinical importance of E. faecalis, how gene and protein expression are coordinated with growth remains poorly defined. Here, we profiled transcript, protein, and tRNA pool dynamics across distinct phases of E. faecalis growth. Differences in protein abundance and corresponding mRNA levels suggested growth phase-dependent posttranscriptional regulation. Growth-associated genes exhibited biased synonymous codon usage, with ribosomal and glycolytic proteins enriched in low-abundance codons read by queuosine-modifiable tRNAs. Analysis of tRNA modification and tRNA isoacceptor abundance revealed growth phase-dependent changes, particularly in anticodon stem-loop modifications that influence synonymous codon translation. Changes in queuosine levels preceded shifts in ribosomal proteins, suggesting a contribution to codon-biased translation. Collectively, these findings reveal growth phase-associated remodeling of the E. faecalis tRNA pool and support a model in which queuosine-dependent translational reprogramming shapes protein expression during bacterial growth.

IMPORTANCE

Enterococcus faecalis is a common cause of hospital-acquired infections. Despite its clinical importance, a comprehensive understanding of the organism’s physiology and adaptation to environmental changes remains incomplete. Here, we characterized protein, transcript, and tRNA dynamics across bacterial growth phases, uncovering a role for posttranscriptional regulation marked by tRNA reprogramming and biased synonymous codon usage. These findings enhance our understanding of E. faecalis growth and support a model of translational reprogramming therein.

INTRODUCTION

Enterococcus faecalis is a common inhabitant of the human gastrointestinal tract that can be opportunistically pathogenic when mucosal immunity is compromised, the gastrointestinal barrier is disrupted, or the native microbiota is depleted following antibiotic treatment (1, 2). Consequently, E. faecalis is a frequent cause of hospital-acquired infections, including central-line-associated bloodstream infections, catheter-associated urinary tract infections, surgical site infections, and infective endocarditis (3, 4). Treatment is often complicated by antibiotic resistance (5, 6). Moreover, many E. faecalis are biofilm-associated (7), giving rise to phenotypic antibiotic tolerance, underscoring the need for new therapeutic strategies (8).

Despite this clinical imperative, the regulatory mechanisms that coordinate gene expression with growth and environmental adaptation in E. faecalis are incompletely understood. Bacteria must rapidly adjust protein synthesis in response to changes in temperature, pH, oxygen availability, nutrient levels, antibiotic exposure, and competition with other microorganisms. One mechanism that facilitates such rapid adaptation is remodeling of the tRNA pool, comprising both tRNA isoacceptors and their modifications. Changes in tRNA abundance or modification can promote selective translation of mRNA transcripts enriched in specific synonymous codons, thereby biasing protein synthesis toward transcripts that support distinct physiologic states (9, 10). Such codon-biased translation has been linked to cellular responses to growth conditions (11–20), oxidative and chemical stress (18, 19, 21–25), and antibiotic exposure (26–29). These dynamics arise through coordinated changes in the expression or activity of tRNA-modifying enzymes, as well as in the synthesis and degradation of tRNAs. Consistent with their regulatory importance, disruption of tRNA-modifying enzymes can impair bacterial fitness (30–33), sporulation (34, 35), biofilm formation (34), intracellular replication (19, 32), and host colonization (36–41). Whether tRNA pool remodeling contributes to growth-dependent translational regulation in E. faecalis remains unknown.

Here, we examined E. faecalis OG1RF growth by quantifying transcripts, proteins, tRNA modifications, and tRNA isoacceptors across distinct phases of bacterial growth (Fig. 1). Discordant changes between mRNA and protein abundance over time suggested growth-dependent posttranscriptional regulation. Analysis of protein-coding genes revealed strong codon biases in growth-associated proteins, particularly ribosomal proteins, which preferentially use C-ending codons read by queuosine (Q)-modifiable tRNAs. Moreover, tRNA modification and isoacceptor abundance changed dynamically during growth, with Q modification shifts preceding similar trends in ribosomal protein abundance. Collectively, these data provide an integrated understanding of E. faecalis growth and support a role for tRNA reprogramming and codon-biased translation in this process.

Fig 1.

Experimental approach to characterize E. faecalis growth. Biological triplicate samples were taken over 6 h, with spectrophotometric monitoring of cell density (bottom left, shaded region represents ± SD, sigmoidal curve fit). Analytes were isolated and quantified as indicated.

RESULTS

Protein abundance during E. faecalis growth is not solely explained by transcription

We began by examining protein abundance across a 6-h E. faecalis growth time course. Using TMT-based quantitative proteomics, we detected 1,627 proteins, representing 65% proteome coverage (Table S1). Proteins with similar temporal patterns were grouped using fuzzy clustering (42), yielding eight distinct clusters (Fig. 2A and Table S2) enriched for specific KEGG and GO pathways (Fig. 2B and Table S3). Proteins involved in ATP biosynthesis, proton transmembrane transport, lysine biosynthesis, and pyrimidine metabolism (Cluster 3) decreased through log phase and then increased as cells approached stationary phase, suggesting a renewed demand for these pathways during slow growth. Carbohydrate transport proteins and associated phosphotransferase machinery (Cluster 7) declined rapidly upon entry into log phase and gradually increased thereafter, consistent with reduced uptake requirements during nutrient abundance followed by increased expression as resources become limiting. By contrast, glycolytic enzymes, ABC transporters, and ribosomal proteins (Clusters 4, 6, and 8, respectively) increased from lag to log phase, consistent with elevated metabolic and translational demands during rapid cell growth. Whereas glycolytic proteins remained elevated in stationary phase, levels of ABC transporters and particularly ribosomal proteins declined, consistent with reduced nutrient uptake and translational activity.

Fig 2.

Discordant trends observed between protein and corresponding transcript levels. (A) Fuzzy clustering analysis of quantitative proteomics data over time, with protein centroids shown as solid lines and mRNA centroids shown as dashed lines (n = 3 biological replicates). (B) Overrepresentation analysis of each cluster, with enriched KEGG and GO pathways indicated (P_adj < 0.05).

To determine the extent to which protein levels were transcriptionally driven, we quantified mRNA abundance over the same time course using RNA-seq (Fig. 2A and Table S4). In some clusters (e.g., Clusters 2 and 5), mRNA and protein profiles were similar, suggesting dominant transcriptional control. In others (especially Clusters 4, 6, 7, and 8), transcript and protein trajectories diverged, particularly at early time points. Such discordance is commonly observed and suggests the involvement of posttranscriptional regulatory mechanisms (43, 44). This raised the possibility that tRNA reprogramming and associated codon-biased translational regulation contribute to protein expression during E. faecalis growth.

Genes encoding growth-associated proteins exhibit distinct synonymous codon biases

If tRNA reprogramming and codon-biased translation contribute to posttranscriptional regulation in E. faecalis, genes encoding growth-associated proteins would be expected to exhibit biased synonymous codon usage. To test this possibility, we calculated codon usage and isoacceptor codon frequencies of all E. faecalis OG1RF protein-coding genes and performed principal component analysis (PCA) on the corresponding z-score data (Fig. 3A and B and Table S5) (9). Although many genes cluster closely in two-dimensional PCA space, a subset differs substantially in their codon usage compared to the rest of the genome. Overlaying KEGG and GO pathway annotations onto the PCA plot revealed that many growth-associated proteins occupy distinct regions of codon usage space (Fig. 3A). The corresponding loadings plot identified the synonymous codons contributing most strongly to this separation (Fig. 3B). Among these were codons for amino acids Asn, Asp, His, and Tyr, which are each encoded by two synonymous codons and decoded by single, modifiable GUN anticodons on their corresponding tRNAs (Table S6). Glycolytic and ribosomal proteins preferentially use C-ending Asn, Asp, His, and Tyr codons compared to their T-ending (U-ending in mRNA) counterparts (Fig. 3C through F). Ribosomal proteins in particular are strongly enriched in C-ending Asn and Tyr codons, and to a lesser extent, C-ending Asp and His codons, compared to the genome-wide median. Moreover, these C-ending codons are underrepresented in the genome, occurring nearly three times less frequently than their T/U-ending counterparts (Fig. 3C through F and Table S5). T/U-ending codons predominate in proteins important for bacterial survival in reduced-glucose environments, including phosphoenolpyruvate (PEP)-dependent sugar phosphotransferase system (PTS) and carbohydrate transport proteins. This preferential use of synonymous codons in growth-associated proteins suggested that their translation might be subject to tRNA reprogramming-based regulation.

Fig 3.

Growth-associated pathway proteins are codon-biased. (A) PCA scores plot based on E. faecalis OG1RF protein-encoding gene codon usage z-scores. Ribosome, glycolytic process, PEP-dependent sugar PTS, and carbohydrate transport pathways are indicated in orange, teal, pink, and purple, respectively. Covariance ellipses are also depicted in pathway-specific colors. (B) PCA loadings plot with codons colored according to terminal nucleotide (A, blue; T, green; G, yellow; and C, pink). Codons read by Q-bearing tRNA isoacceptors are circled. Violin plot of the frequency of codons highlighted in panel B, namely, those for Asn (C), Asp (D), His (E), and Tyr (F), in proteins from the KEGG/GO pathways shown in panel A: ribosome (orange), glycolytic process (teal), PEP-dependent sugar PTS (pink), and carbohydrate transport (purple) versus in the entire genome (gray). Violin width represents density, thick gray bars the interquartile range (IQR), thin lines 1.5× IQR, and white dots the median. P-values were determined by Mann-Whitney U testing with Benjamini-Hochberg FDR correction.

tRNA modifications exhibit distinct growth-phase behaviors

Since tRNA reprogramming involves changes in both tRNA modification and tRNA isoacceptor abundance, we next characterized each across the E. faecalis growth curve. Purified small RNA (>85% tRNA) was digested to ribonucleosides, and modification levels were quantified by liquid chromatography-coupled tandem mass spectrometry (LC-MS/MS) (Table S7). Several modifications, including queuosine (Q), 5-methoxyuridine (mo⁵U), 4-thiouridine (s⁴U), and 2-methyladenosine (m²A), displayed pronounced growth-dependent dynamics (Fig. 4A and B).

Fig 4.

tRNA modification changes during growth. (A) Heatmap of small RNA modification levels over the growth time course, expressed as log₂(fold change) with respect to modification levels at t = 1 h. Values represent the average of three biological replicates. (B) Scatterplot of select dynamic modifications (indicated by arrows in panel A) over time, expressed relative to modification levels at t = 1 h. Shaded regions represent the mean ± SD. (C) tRNA isoacceptors and the nucleotide positions therein on which the modifications in panel B have been reported (45).

Q levels increased sharply in late lag phase and then decreased during log and stationary phases (Fig. 4A and B). Q is found at position 34, the first nucleotide of the anticodon, in Asn, Asp, His, and Tyr tRNAs, where it replaces G in their respective GUN anticodons (Fig. 4C; Fig. S1A) (45). Homology analysis suggests that Q is installed into E. faecalis tRNA through the sequential action of three enzymes: Tgt, QueA, and QueG (Fig. S1B) (46). Consistent with Q modification dynamics, Tgt protein levels decreased sharply during log phase (3 h) (Fig. S1C). Correlations with QueA and QueG were more nuanced; however, QueG levels increased during lag phase and decreased after log phase (Fig. S1D and E). These trends are consistent with the detection of the preQ₁ modification only at the 1-h time point, suggesting conversion to Q by late lag phase followed by a subsequent decline (Table S7). Levels of QueK, a queuosine hydrolase, and QueT, the queuine transporter, also increased from lag through log phase and then largely stabilized (Fig. S1F and G). These proteins have preQ₁ riboswitches upstream of their translational start sites (47); thus, their upregulation may be triggered by decreased intracellular preQ₁ levels as the molecule is incorporated into tRNA and subsequently converted to Q-tRNA. Since E. faecalis cannot synthesize Q-tRNA de novo but rather relies on Q/Q-precursor salvage from the environment, the bacteria may upregulate these proteins in an attempt to sustain Q-tRNA levels. Previous studies have linked Q levels to reactive oxygen species (ROS), suggesting that increased Q may occur under oxidative conditions (25, 48, 49). Consistent with earlier reports (50), we observed elevated H₂O₂ during E. faecalis lag phase that decreased through stationary phase in parallel with Q changes (Fig. S2). This observation raises the possibility that the transient Q spike is triggered by ROS accumulation early in growth.

Levels of mo⁵U and its precursor, 5-hydroxyuridine (ho⁵U), increased during lag phase and remained elevated over the time course (Fig. 4A and B). These modifications occur at position 34 in select Ala, Leu, Pro, Ser, Thr, and Val tRNAs (Fig. 4C; Fig. S3A) (45). ho⁵U is synthesized from uridine by TrhP1/TrhP2 and/or TrhO, which use prephenate and molecular oxygen as the hydroxyl source, respectively (Fig. S3B) (51). ho⁵U is then methylated to mo⁵U by the S-adenosylmethionine (SAM) methyltransferase TrmR (52). TrhP1 and TrhP2 homologs increased during lag phase, after which TrhP1 decreased slightly while TrhP2 continued to rise through log phase before stabilizing (Fig. S3C and D). These trends are consistent with the observed increase in ho⁵U in lag phase followed by a plateau in modification levels (Fig. 4B). TrmR abundance remained largely unchanged over the time course (Fig. S3E), suggesting that substrate availability may limit mo⁵U formation (Fig. 4A and B).

s⁴U levels decreased from lag phase through stationary phase (Fig. 4A and B). This modification is present at position 8 between the 5′-acceptor stem and the D-loop in nearly all tRNAs (Fig. 4C; Fig. S4A) (45). E. faecalis utilizes two distinct sulfur-relay transfers to install s²U and s⁴U, respectively, in tRNA, and these are unique from the SUF system that is responsible for Fe-S-cluster protein biosynthesis (53–55). Two enzymes, NifZ and ThiI, carry out the coordinated transfer of sulfur from a donor Cys to form s⁴U in tRNA (Fig. S4B) (54). NifZ levels decreased from lag phase through log phase (Fig. S4C), while ThiI levels increased up to mid-log phase and then decreased thereafter (Fig. S4D). These changes are consistent with the overall decrease in s⁴U from lag phase through stationary phase (Fig. 4A and B).

Levels of m²A decreased from lag through log phase and then increased thereafter (Fig. 4A and B). This anticodon-adjacent modification occurs at position 37 in select Arg, Asp, Gln, Glu, and His tRNAs (Fig. 4C; Fig. S5A) (45). The dual-function enzyme RlmN installs m²A in both tRNA and rRNA in E. faecalis (Fig. S5B) (22, 56). Levels of RlmN increased from lag phase to log phase and then decreased from log phase to stationary phase, roughly the inverse of the observed m²A modification pattern (Fig. S5C). Two non-mutually exclusive explanations may account for this discordance between tRNA modification and RlmN protein levels. First, the increase in RlmN abundance parallels the increase in m²A levels in 16S/23S rRNA over the time course (Fig. S6), consistent with its dual substrate specificity. Second, we previously showed that ROS-inducing antibiotics reduce m²A levels in E. faecalis, likely via ROS-mediated inactivation of the Fe-S cluster required for RlmN activity (22). Posttranslational inactivation of RlmN by ROS during lag phase (Fig. S2) may therefore stimulate compensatory enzyme synthesis until oxidative stress subsides.

tRNA isoacceptor abundance does not fully explain modification dynamics

To determine whether changes in modification levels reflected altered tRNA abundance, we quantified tRNA isoacceptor abundance using absolute quantification RNA-sequencing (AQRNA-seq) (57), a sequencing method optimized for highly modified small RNAs (Fig. 5A and Table S8). Although some modification dynamics could be rationalized by changes in levels of the tRNAs bearing them, Q was a prominent counterexample (Fig. 5B). Q levels increased during late lag phase (2 h) without detectable changes in the abundance of the corresponding tRNA isoacceptors. Levels of Q then declined from log phase through early stationary phase, while Q-modifiable tRNA-Asn and tRNA-Tyr isoacceptor levels increased. These dynamics could reflect a combination of changes in levels or activity of Q writers and cryptic erasers, availability of Q precursors, degradation of Q-modified tRNAs, and synthesis of unmodified tRNAs. Regardless of the mechanism, these data indicate that the fraction of Q-modified tRNAs increases prior to log phase and decreases thereafter, with unmodified tRNA-Asn and tRNA-Tyr predominating in stationary phase.

Fig 5.

tRNA isoacceptor level dynamics during growth. (A) Heatmap of tRNA isoacceptor levels over the growth time course, expressed as log₂(fold change) with respect to tRNA isoacceptor levels at t = 1 h. Values represent the average of three biological replicates. (B) Select tRNA modification levels (black lines, from Fig. 4B) overlaid on their reported tRNA isoacceptor levels over the same time course, expressed relative to the t = 1 h time point. s⁴U was not included as it is present on almost all tRNAs. Shaded regions represent the mean ± SD.

DISCUSSION

The public health burden of Enterococcus faecalis and the need for new therapeutic approaches necessitate a comprehensive understanding of the organism’s physiology. A growing body of literature reveals that tRNA reprogramming and codon-biased translation are common regulatory mechanisms in bacteria, but whether they occur in E. faecalis is unknown. Therefore, we undertook a comprehensive characterization of E. faecalis growth that identified (i) changes in protein levels that were independent of their corresponding transcripts; (ii) biased synonymous codon usage among growth-associated proteins; and (iii) growth phase-dependent tRNA pool remodeling. These observations are consistent with a model in which the bacteria respond to environmental cues and modify their tRNAs to promote progression through various growth phases (Fig. 6).

Fig 6.

Proposed model of how tRNA reprogramming contributes to E. faecalis growth. In the nutrient-rich lag phase, increased expression of tRNA-modifying enzymes TrhP1, TrhP2, TrmR, and QueG results in elevated ho⁵U, mo⁵U, and Q levels. These modifications support overall translational efficiency, specifically of C-ending Q-dependent codon-enriched transcripts encoding glycolytic and ribosomal proteins, which characterize log-phase bacterial growth. In nutrient-poor late log phase, decreased expression of tRNA-modifying enzymes QueG and ThiI results in decreased levels of Q and s⁴U. Lack of these modifications results in reduced overall translational efficiency, as well as the preferential translation of alternative metabolism-associated U-ending Q-dependent codon-enriched transcripts, which characterize stationary-phase bacterial growth.

During nutrient-rich lag phase, Q-modified tRNA levels rise in E. faecalis. The translational consequences of Q modification appear complex, with studies reporting preferential translation of either U-ending or C-ending codons, depending on the organism, amino acid, and sequence context (18, 27, 34, 58–62). Here, Q dynamics precede similar trends in glycolytic and ribosomal proteins, which increase through log phase and then decrease thereafter. These proteins are enriched in C-ending Q-dependent codons, suggesting that Q may favor translation of these codons early in E. faecalis growth. These protein trends are consistent with a prioritization of translational capacity and glycolytic flux during early growth when preferred carbon sources like glucose are abundant. E. faecalis is primarily fermentative and uses this seemingly wasteful strategy (i.e., lower ATP yield per glucose molecule than respiration) to allow rapidly growing bacteria to optimize proteome efficiency, or energy produced per unit protein mass invested, in nutrient-rich environments (63–65). Levels of ho⁵U and mo⁵U also rise quickly during lag phase and then stabilize. Structurally, the hydroxyl group of ho⁵U stabilizes the anticodon loop in canonical Watson-Crick geometry, enabling pairing with U- and C-ending codons, whereas methylation to mo⁵U is required to stabilize the enol form of U to promote pairing with G-ending codons (66). Early increases in these modifications could therefore expand the decoding capacities of tRNA-Ala-UGC, tRNA-Leu-UAG, tRNA-Pro-UGG, and tRNA-Val-UAC to include their corresponding C-, G-, and/or U-ending codons and support overall translational efficiency during rapid growth.

In many microorganisms, when nutrients begin to be depleted (on transition to stationary phase), carbon catabolite repression is lifted to enhance bacterial survival in the absence of preferred carbohydrates. During this time, Q modification levels decline, and subsequent increases in unmodified Q-dependent tRNAs may then facilitate translation of less-biased and U-ending codon-biased transcripts, which are more abundant in proteins associated with metabolism of other (non-glucose) carbon substrates. m²A levels decrease in log phase and then increase through stationary phase. The fact that RlmN installs m²A in both tRNA and rRNA complicates understanding the role of the modification in the different RNA types. Plants, however, have separate m²A writers, permitting interrogation of the function of each (67). In Arabidopsis, m²A37 helps maintain a relaxed tRNA conformation and facilitates decoding of tandem m²A-bearing-tRNA-dependent codons (67). Although there is no reported preference that m²A bestows for one synonymous codon or another, Watson-Crick-pairing codons read by m²A-modifiable tRNAs are correlated with ribosomal and glycolytic proteins. Higher levels of m²A in stationary phase may stabilize codon-anticodon pairing in general, possibly promoting translational efficiency despite decreasing levels of ribosomes. Levels of s⁴U generally decrease over time. These trends are consistent with those previously reported in Escherichia coli, in which the fraction of s⁴U-modified tRNAs for select isoacceptors decreased with increasing growth rate (11). This modification plays a role in the intracellular stability of tRNAs, with hypomodification shown to result in rapid tRNA decay by the RNA degradasome (68). Decreased s⁴U levels in late stages of growth may reflect or contribute to translational slowdown.

The present study has provided a comprehensive understanding of the macromolecular dynamics and tRNA reprogramming that occur during E. faecalis growth, yet certain limitations must be acknowledged. First, the model presented here suggests that the observed correlation between Q-tRNA dynamics and translation of C-ending Q-dependent codon-rich ribosomal and glycolytic proteins represents an active, cell-based reprogramming of its tRNAs to facilitate growth-related protein translation. However, the current data do not exclude the possibility of passive, nutrient deficiency-based tRNA reprogramming wherein media depletion of Q/Q precursors contributes to the observed decrease in Q-modified tRNAs. Thus, while the observed trends are definitive, elucidation of their mechanistic underpinnings will require further analyses. These could include media supplementation/metabolomics to elucidate Q/Q precursor levels and reporter assays or ribosome footprinting coupled with genetic perturbation to determine the direct consequences of Q-modification on U- vs C-ending Q-dependent codon translation. Second, while this study has revealed molecular changes across canonical bacterial growth phases, caution should be taken in extrapolating this data beyond the time points explicitly tested. Given the extent of culture dilution and timespan investigated, early time points may reflect some lingering features of stationary-phase growth, and late time points only the very beginnings of quorum sensing. Comprehensive characterization of the transition between growth and biofilm formation warrants further study. Third, the present study was performed using E. faecalis OG1RF, a widely used laboratory strain, which exhibits rifampicin resistance due to a H486R mutation in the β subunit of RNA polymerase (RpoB) (69, 70). Mutations in rpoB have been reported to influence the transcriptional profiles of several bacterial species, including E. faecium (71, 72), and the cephalosporin sensitivity of E. faecalis (73). Therefore, the data sets presented here may differ from those of wild-type E. faecalis or other strains. Nevertheless, all analyses were performed within the same genetic background across growth phases, allowing robust relative comparisons among conditions.

It remains to be seen whether this observed growth phase-dependent tRNA reprogramming is unique to E. faecalis or is also present in other bacteria. Vibrio cholerae and Escherichia coli were also recently reported to have ribosomal proteins enriched in C-ending Q-dependent codons, despite a genome-wide preference for their U-ending counterparts (27). Regardless, the data presented here suggest a role for tRNA reprogramming and codon-biased translation in E. faecalis growth and proffer a framework on which future mechanistic studies, in this organism or others, can be designed. Furthermore, the comprehensive analyses and data sets provided herein will serve as a resource for the microbiology community.

MATERIALS AND METHODS

Bacterial growth conditions

E. faecalis OG1RF (69, 74) cultures were grown statically overnight (14–18 h) at 37°C with ambient air in BBL Brain Heart Infusion broth (BHI; Becton Dickinson, 211059). For growth time-course experiments, overnight cultures of equivalent optical density (OD₆₀₀ ~1.5) were diluted 1:25 directly into fresh BHI media. Cultures were grown statically at 37°C with ambient air for 6 h, with absorbance measurements (OD₆₀₀) recorded every hour.

Bacterial lysis for RNA and protein extraction

Bacterial cultures (10–35 mL) were harvested by centrifugation at 5,000 × g at 4°C for 5 min. Cell pellets were then washed once with 35 mL of ice-cold PBS and pelleted again by centrifugation (5,000 × g at 4°C for 5 min). The wash supernatant was decanted, and the cell pellets were immediately resuspended in 1 mL of ice-cold TRIzol Reagent (Thermo Fisher, 15596026) and added to Lysing Matrix B tubes (MP Biomedicals, 116911100) on ice. Bacteria were lysed by bead-beating (3 × [40 s beat, 2 min rest on ice]) using a FastPrep-24 5G instrument (MP Biomedicals, 116005500) at a rate of 6.0 m/s. Samples were then centrifuged at 12,000 × g at 4°C for 5 min. The supernatants were transferred to new tubes to which 200 μL of chloroform was added. Samples were shaken vigorously for 15 s and then allowed to settle at ambient temperature for 2 min. The resulting mixtures were centrifuged at 12,000 × g at 4°C for 15 min, enabling separation of the aqueous (containing RNA) and organic (containing protein) phases.

Large and small RNA isolation

Large and small RNAs were isolated from a portion (200 μL) of the extracted aqueous phase by solid-phase extraction using a PureLink miRNA Isolation Kit (Thermo Fisher, K157001) according to the manufacturer’s protocol with minor modification. Prior to transfer to the first cartridge, ethanol was added to the aqueous phase to a final concentration of 15% (vol/vol). This ethanol concentration was determined empirically to minimize retention of small RNAs on the first cartridge while still ensuring complete retention of large RNAs on the first cartridge. Ethanol was added to the flow-through of the first cartridge to give a final concentration of 70% (vol/vol), prior to the addition to the second cartridge. Samples on both cartridges were washed twice with wash buffer, spun dry, and then eluted in 50 μL pre-warmed (70°C), nuclease-free (non-DEPC-treated) water. RNA concentrations were determined using a NanoDrop spectrophotometer, and RNA quality and purity were assessed by Agilent 2100 Bioanalyzer pico and small RNA microfluidic chips.

RNA digestion and LC-MS/MS-based quantification of modifications

RNAs were hydrolyzed and dephosphorylated using a cocktail of enzymes, including benzonase, alkaline phosphatase, and phosphodiesterase I, at 37°C for 6 h. Ribonucleoside analysis was performed using a Thermo Hypersil Gold aQ C18 column (100 × 2.1 mm, 1.9 µm) on an Agilent 1290 Infinity II HPLC system with an in-line UV detector coupled to an Agilent 6495 triple quadrupole mass spectrometer. Dynamic MRM was used for analyte detection and quantitation, with modification levels normalized to the sum of the UV signals (260 nm) for the unmodified nucleosides (A, U, G, and C). Additional details are available in Method S1.

Absolute quantification RNA-sequencing

To quantify tRNA isoacceptor levels, small RNA (75 ng) was used to prepare cDNA libraries for AQRNA-seq as previously described (57, 75), with targeted modifications to mitigate polymerase fall-off and primer dimer formation (76). Specifically, polymerase fall-off was minimized by replacing AlkB demethylation with a high-processivity reverse transcriptase (SuperScript IV; Thermo Fisher, 18091050), which enhances read-through across diverse RNA modifications and increases full-length cDNA yield. In addition, primer dimer formation was minimized by incorporating a post-reverse transcription digestion using Escherichia coli exonuclease I (ExoI, NEB, and M0293S) to remove excess reverse transcription primers, thereby eliminating the need for gel purification. Libraries were quantified using a LightCycler 480 Real-Time PCR System (Roche), and library quality was assessed using the AATI Fragment Analyzer (Agilent). Pooled samples were cleaned using 0.9× volume PCRClean DX beads (Aline Biosciences) prior to 75 bp paired-end sequencing on an Illumina MiSeq platform (v3 reagent kit) with custom forward and reverse sequencing primers (57, 75). A list of 37 mature, sequence-unique tRNAs present in E. faecalis OG1RF was curated from the Genomic tRNA Database (GtRNAdb). Raw sequences were processed using a custom AQRNA-seq analytical pipeline to yield tRNA abundances (76).

Total RNA isolation

Total RNA was isolated from a portion (120 μL) of the extracted aqueous phase (see Bacterial lysis for RNA and protein extraction) using a RNeasy Mini Kit (Qiagen, 74106) according to the manufacturer’s protocol, including on-column DNase I treatment, with elution in 50 μL nuclease-free water. RNA quality was assessed using an Agilent 2100 Bioanalyzer pico microfluidic chip, and RNA concentrations were determined using a Qubit RNA High-Sensitivity Assay Kit (Thermo Fisher, Q32852) with a Qubit 2.0 fluorimeter. DNA contamination in the samples was determined to be <10% using QuantiFluor dsDNA and RNA Systems (Promega, E2671 and E3310, respectively).

RNA-sequencing

Purified total RNA was processed using the Illumina Stranded Total RNA Library Preparation protocol (Standard, without rRNA depletion). Libraries were prepared using the IDT for Illumina RNA UD Indexes, and 150 bp paired-end sequencing was performed on an Illumina NextSeq 2000 platform using XLEAP-SBS chemistry. Raw sequenced reads were processed using standard open-source packages and tools for bacterial RNA-sequencing (RNA-seq) data processing. Briefly, sequence quality of the raw FASTQ files was assessed using FastQC (77), after which the paired-end reads were adapter- and quality-trimmed using fastp (78) with automatic paired-end adapter detection and a minimum post-trimming read length of 40 nt. Trimmed reads were aligned to the E. faecalis OG1RF reference genome (GenBank CP002621.1) using BWA-MEM (79) with default parameters. The resulting SAM output was converted to BAM using samtools (80), then coordinate-sorted and indexed, and read counts per coding sequence (CDS) were generated from sorted BAMs using the featureCounts package (81). Per-sample counts attributed to CDS by locus tag values were finally assembled for downstream analysis.

Protein isolation

Ethanol (300 μL) was added to the entire volume of the extracted organic phase (see Bacterial lysis for RNA and protein extraction), and samples were centrifuged at 2,000 × g for 5 min at 4°C to precipitate any remaining DNA. The supernatants were transferred to fresh tubes to which 2 volume equivalents of isopropanol were added. Samples were mixed well and allowed to stand at ambient temperature for 10 min. Samples were then centrifuged for 5 min at 12,000 × g at 4°C, and the supernatant was removed carefully by pipette. Protein pellets were washed three times with 2 mL of 95% (vol/vol) ethanol (in water) and once more with 100% ethanol, with centrifugation at 7,500 × g for 5 min at 4°C. Supernatants were removed, and the protein pellets were air-dried at ambient temperature for 10 min.

Quantitative proteomics

Protein pellets were digested with Lys-C and trypsin and peptides labeled using the TMTpro 18-plex Label Reagent set (Thermo Fisher, A52045). Combined samples were subjected to high-pH fractionation prior to mass spectrometric analysis. Chromatographic separation was performed using an EASY-nLC 1000 column (Thermo Fisher) coupled to an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher) using data-dependent mode. Raw mass spectra were searched using Sequest available in Proteome Discoverer (v3.1, Thermo Fisher) against E. faecalis OG1RF primary protein sequences retrieved from NCBI (NCBI Reference Sequence: NC_017316.1). Additional details are available in Method S2.

Hydrogen peroxide measurements

Bacteria-produced hydrogen peroxide levels were determined using a method adapted from Boonanantanasarn et al. (82) and Gaca et al. (50), wherein hydrogen peroxide production is measured using a horseradish peroxidase-coupled fluorometric assay. Additional details are available in Method S3.

Data mining and statistical analysis

All data mining and statistical analyses were performed in either Python or R, unless otherwise indicated. Soft clustering of proteomics data was performed in R using the Mfuzz package (42). Raw protein intensities were normalized by total protein abundance per sample, normalized across samples by arithmetic mean, and then log2-transformed. Time-matched biological replicates were subsequently averaged, and protein z-scores were determined prior to clustering. The optimal number of clusters was determined from a minimum centroid distance (D_min) curve. Overrepresentation analysis was performed on the resulting protein clusters using Bioconductor’s clusterProfiler package and publicly available KEGG and in-house-curated GO databases (83–87). Raw mRNA counts were normalized using DESeq2 (median-of-ratios normalization) (88), followed by variance-stabilizing transformation. Time-matched biological replicates were then averaged and transcript z-scores determined. Only proteins with corresponding transcripts (and vice versa) were included in calculations of the centroid values (Fig. 2A). R scripts and associated data used to generate Fig. 2 are available at https://github.com/kline-unige/multiomics-clustering. Gene-specific and genome-based codon analytic measures were made using previously described methods (9). Each protein-coding gene was analyzed from start to stop codon, and the number of each of 64 codons was counted. From these data, gene-specific isoacceptor codon frequencies and corresponding z-scores were computed. Raw tRNA abundances were normalized using DESeq2 (median-of-ratios normalization) in R (88).

DATA AVAILABILITY

All source data have been deposited in public repositories: LC-MS/MS RNA modification data have been deposited in the MetaboLights repository (89) with the study identifier MTBLS14602; RNA-seq and AQRNA-seq data have been deposited in NCBI’s Gene Expression Omnibus (90) and are accessible through GEO Series accession numbers GSE334041 and GSE331534, respectively; and mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository (91) with the data set identifier PXD079013. Processed data and all other supporting files are included in the supplemental information.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.01474-26.

Supplemental Information. mbio.01474-26-s0001.pdf.

Table of contents for all supplemental figures, tables, and methods; supplemental figures; supplemental methods.

Table S1. mbio.01474-26-s0002.xlsx.

Raw data and subsequent calculations/normalizations for E. faecalis OG1RF proteomics data across the growth time course.

Table S2. mbio.01474-26-s0003.xlsx.

Mfuzz soft-clustering analysis of z-score-normalized proteomics data.

Table S3. mbio.01474-26-s0004.xlsx.

Overrepresentation analysis of protein clusters using KEGG and GO databases.

Table S4. mbio.01474-26-s0005.xlsx.

Raw data, subsequent calculations/normalizations, and cluster centroids for E. faecalis OG1RF transcriptomics data across the growth time course.

Table S5. mbio.01474-26-s0006.xlsx.

Gene-specific codon usage, isoacceptor codon frequencies, and corresponding isoacceptor frequency-based z-scores for all E. faecalis OG1RF protein-coding genes.

Table S6. mbio.01474-26-s0007.xlsx.

tRNA isoacceptor decoding of codons in E. faecalis OG1RF.

Table S7. mbio.01474-26-s0008.xlsx.

Raw data, subsequent normalization, and log₂(fold change) calculations for E. faecalis OG1RF tRNA modification data across the growth time course.

Table S8. mbio.01474-26-s0009.xlsx.

Raw data, subsequent normalization, and log₂(fold change) calculations for E. faecalis OG1RF tRNA isoacceptor data across the growth time course.

Table S9. mbio.01474-26-s0010.xlsx.

LC-MS/MS detection parameters for RNA modifications identified in E. faecalis OG1RF.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.