Conserved 5-methyluridine tRNA modification modulates ribosome translocation

Significance

Cells contain thousands of chemically modified RNAs. Decades of modification enzyme knockout studies, coupled with discovery of natural mutations that cause cellular dysfunction, reveal that modifications impact every step in the life cycle of an RNA. Nonetheless, despite their ubiquity throughout nature and importance for cellular health, the functional consequences of most individual RNA modification sites remain to be discovered. Emblematic of this, the conservation of m⁵U54 in transfer RNAs (tRNAs) has confounded scientists since the 1960s, as decades of studies have failed to reveal a role for the modification in protein synthesis. Our findings present the molecular-level insights demonstrating how m⁵U54 modulates protein synthesis to impact gene expression.

Abstract

While the centrality of posttranscriptional modifications to RNA biology has long been acknowledged, the function of the vast majority of modified sites remains to be discovered. Illustrative of this, there is not yet a discrete biological role assigned for one of the most highly conserved modifications, 5-methyluridine at position 54 in tRNAs (m⁵U54). Here, we uncover contributions of m⁵U54 to both tRNA maturation and protein synthesis. Our mass spectrometry analyses demonstrate that cells lacking the enzyme that installs m⁵U in the T-loop (TrmA in Escherichia coli, Trm2 in Saccharomyces cerevisiae) exhibit altered tRNA modification patterns. Furthermore, m⁵U54-deficient tRNAs are desensitized to small molecules that prevent translocation in vitro. This finding is consistent with our observations that relative to wild-type cells, trm2Δ cell growth and transcriptome-wide gene expression are less perturbed by translocation inhibitors. Together our data suggest a model in which m⁵U54 acts as an important modulator of tRNA maturation and translocation of the ribosome during protein synthesis.

Posttranscriptional modifications impact RNA structure, function, stability, and dynamics. Cells utilize these chemical modifications to control protein expression and to ensure the speed and accuracy of the ribosome. Their significance is underscored by observations that the dysregulation of RNA-modifying enzymes is linked to a myriad of pathologies including diabetes, neurological disorders, and many cancers (1–6). While over 150 different chemical modifications exist within thousands of RNAs across all three kingdoms of life (7), the molecular-level consequences of the vast majority of modified sites remain to be discovered. Even within the most well-studied class of RNA modifications, Escherichia coli and Saccharomyces cerevisiae tRNAs, < 50% of commonly modified positions have assigned functional roles within the protein synthesis pathway.

Emblematic of this, despite six decades of study, no function is ascribed to the universally conserved 5-methyluridine modification found at position U54 in tRNA T-loops (m⁵U54) (7–13) (Fig. 1). The conservation of m⁵U54 is a long-standing conundrum as bacterial and eukaryotic cells lacking uracil-5-methyltransferase (trmA in bacteria, trm2Δ in eukaryotes) do not exhibit a growth defect under normal laboratory conditions (14, 15). However, the ability of wild-type cells to outcompete trmA and trm2Δ strains suggests that the enzymes are somehow advantageous for cellular fitness (15, 16). The origin of the fitness advantage conferred by uracil-5-methyltransferases remains enigmatic. In vitro translation studies using m⁵U-depleted tRNAs demonstrate that loss of m⁵U54 from tRNAs does not slow peptide elongation by the ribosome (16–18). m⁵U has also recently been discovered in eukaryotic messenger RNAs (mRNAs) (19–22), providing the modification a possible non-tRNA mediated avenue to influence protein synthesis. Nevertheless, the low level of m⁵U in mRNAs [at least 10-fold less than N6-methyladenosine (m⁶A) and pseudouridine (Ψ) modifications] and its minor (zero-fold to two-fold) effect on the rate constant for amino acid addition make m⁵U in mRNAs unlikely to broadly enhance cellular fitness (19).

Fig. 1.

5-methyluridine structure and location. Structures of (A) uridine and (B) 5-methyluridine (m⁵U). (C) m⁵U54 is found in the T loop of tRNAs.

The most compelling evidence of a widespread role for tRNA uracil-5-methyltransferases is the modest destabilization of tRNAs purified from trmA and trm2Δ cell lines (23). However, it is unclear whether the enzymes’ effect on tRNA stability is directly attributable to their catalysis of m⁵U54 modification or the tRNA-folding chaperone activities reported for these enzymes (14–16). Further complicating matters, the “tRNA foldase” activity of TRMA does not fully account for the trmA fitness loss relative to wild-type cells; mutational studies reveal that the ability to catalyze m⁵U54 insertion is also required to rescue cellular growth in competition assays, suggesting the modification itself enhances fitness (16). Despite its conservation and apparent contributions to tRNA structure, the overall biological significance of the tRNA m⁵U54 modification and its impacts (if any) to protein translation are not defined.

In this work, we identify contributions of m⁵U54 to tRNA maturation, translation elongation, and gene expression. Our studies reveal that trm2Δ cells are both desensitized to translocation inhibitors and exhibit altered tRNA modification profiles. In vitro translation studies support these cellular observations, demonstrating that tRNA^Phe purified from trmA cells (tRNA^Phe,-m5U) has an altered modification profile and permits the translocation of the ribosome in the presence of hygromycin B. Together, our data reveal biological roles for m⁵U54—to promote the modification of tRNAs and modulate the speed of ribosome translocation. We find that even the subtle impact of m⁵U54 on translation has broad consequences for gene expression under cellular stress.

Results

Trm2 Impacts Cell Growth, Transcriptome Composition, and Reporter Protein Production under Translational Stress.

The significance of RNA modifications often becomes most apparent under cellular stress when protein synthesis (and not just transcription) is particularly important for controlling the composition of the proteome (24). With this in mind, we surveyed the impact of Trm2 on cell growth under sixteen different conditions using spot plating assays. We compared the growth of wild type and trm2Δ S. cerevisiae (BY4742) cells at varied temperature (22 °C, 30 °C, 37 °C), carbon source (glucose, sucrose, galactose), pH (4.5, 6.8, 8.5), salt concentration (NaCl, MgSO₄), and translation inhibitors (hygromycin B, cycloheximide, puromycin, paromomycin) (Dataset S1). Wild type and trm2Δ S. cerevisiae grew similarly regardless of temperature, carbon source, pH, or MgSO₄ concentration, while trm2Δ exhibited subtly enhanced growth over wild type under 1 M NaCl salt stress.

By comparison, three translational inhibitors, hygromycin B, cycloheximide, and paromomycin, lead to more distinct phenotypes for wild type and trm2Δ strains. Relative to wild type, trm2Δ cells were more sensitive to cycloheximide, while they were less sensitive to hygromycin B and paromomycin (Fig. 2A and Dataset S1). Growth assays in media concur with our spot plating observations (SI Appendix, Fig. S1). Additionally, measurement of the minimum inhibitory concentration of hygromycin B in E. coli with plasmid-borne catalytically inactive trmA(C324A) expressed from its native promoter revealed that these cells were more sensitive to the drug than wild type cells (Fig. 2 B and C). Deletion mutants in yeast and E. coli also showed decreased sensitivity to hygromycin b (Fig. 2C and SI Appendix, Fig. S2), presumably because of the established chaperone activity of TrmA/Trm2 (14). We elected to continue follow-up studies with cycloheximide and hygromycin B because preliminary experiments suggested that these inhibitors were the most trm2Δ and trmA specific. Consistent with this observation, RNA sequencing (RNA-seq) reveals that the transcriptome profile of trm2Δ cells is less disrupted under hygromycin B stress relative to wild-type cells (Fig. 3 A–D), and the expression of a luciferase reporter in the presence of hygromycin B is increased trm2Δ cells (SI Appendix, Fig. S3).

Fig. 2.

Trm2 and TrmA modulate cell growth under cellular stress. (A) Spot plating assays comparing growth between WT and trm2Δ BY4741 S. cerevisiae with NaCl, hygromycin B, cycloheximide, and paromomycin. Each replicate was grown on the same plate, but WT and trm2Δ were not always in adjacent rows, so these cropped images have been placed next to each other. Full plate images are available in Dataset S1. Biological replicates were prepared from independent overnight cultures grown on different days. (B) Hygromycin B minimum inhibitory concentration curves for trmA E. coli (BW25113 background) bearing a plasmid containing the entire trmA structural gene (promoter, coding sequence, and terminator), either coding for the wild-type TrmA sequence or for catalytically inactive TrmA-C324A. For each replicate (n = 3), an independent transformant was used to prepare the inocula for the MIC₅₀ cultures. Each pair of replicates (WT and C324A) was conducted on a separate day. (C) MIC₅₀ values from panel B and also for BW25113 and trmA E. coli (n = 2 for the latter pair). Data from all replicates were fitted with a single curve and the error bars indicate the 95% CI on the fitted value of MIC₅₀.

Fig. 3.

mRNA gene expression is less impacted in trm2Δ than wild type S. cerevisiae following hygromycin B treatment. Log₂ fold change of the mean of normalized counts for each transcript detected by RNA-seq for (A) wild type vs. trm2Δ S. cerevisiae grown in unstressed (YPD, no hygromycin B) conditions; (B) wild type S. cerevisiae untreated with hygromycin B vs. wild type S. cerevisiae treated with hygromycin B; (C) trm2Δ untreated with hygromycin B vs. trm2Δ treated with hygromycin B; (D) wild type vs. trm2Δ S. cerevisiae treated with hygromycin B.

mRNA m⁵U Modification Does Not Drive Trm2-Dependent Changes in Cellular Fitness.

In yeast, m⁵U is incorporated into both tRNAs and mRNAs (19, 20, 22). Therefore, the Trm2-dependent differences in cellular fitness that we observed in the presence of hygromycin B could arise from changes to the modification landscape of either RNA species. We implemented a multiplexed liquid chromatography tandem mass spectrometry (UHPLC-MS/MS) assay developed in our lab to measure the levels of 50 different modifications within mRNAs isolated from untreated, hygromycin B- and cycloheximide- treated wild type cells (19). Total mRNA was purified using a previously described three-stage purification pipeline that treats total RNA to small RNA depletion, rRNA depletion, and two rounds of poly(A) RNA isolation (19). Subsequently, mRNA purity was confirmed using Bioanalyzer, RNA-seq, RT-qPCR, and LC-MS/MS (SI Appendix, Figs. S4–S6). We find the levels of a limited subset mRNA modifications increase (e.g., pseudouridine, 2’O-methylguanosine) or decrease (e.g., m⁵U) modestly (≤1.5 -fold) in response to sub-MIC concentrations of hygromycin B and cycloheximide (Fig. 4A). The levels of m⁵U in mRNAs isolated from antibiotic treated cells is extremely low (0.003 ± 0.001% m⁵U/U in hygromycin B and 0.0025 ± 0.0006% m⁵U/U in cycloheximide) and the modification is unlikely to be regularly encountered by the ribosome as there is only 1 m⁵U substitution per every ~35,000 Us (Dataset S2). Indeed, even the most abundant mRNA modifications are rare relative to m⁵U54 in tRNAs. While it is possible that changes in mRNA modifications might explain some effects of cycloheximide and hygromycin B, most effects are likely to be consequences of loss of m⁵U54 or downstream effects thereof. Taken together with our previous in vitro translation studies demonstrating that the inclusion of m⁵U into mRNA codons does not slow the peptide elongation significantly (19), our data suggest that the Trm2-dependent effects on cell growth that we observe do not originate from the action of the enzyme on mRNAs. Instead, they indicate that differential impact of hygromycin B on wild type and trm2Δ cellular fitness and protein production more likely results from the changes in m⁵U54-tRNA modification status.

Fig. 4.

The modification landscapes of both coding and noncoding RNAs are altered by translational inhibition and also by trmA. (A) Inhibition of translation by cycloheximide and hygromycin B modestly alters the mRNA modification landscape. Normalized modification abundance with purified mRNA (modification/main base %) in WT BY4741 S. cerevisiae without antibiotic (gray), 100 ng/mL cycloheximide (red), or 50 μg/mL hygromycin B (blue). Error bars indicate the SD of two biological replicates. * indicates a significant change where P < 0.05. (B) The sequence of E. coli tRNA^Phe with the sites of known modifications indicated. LC-MS/MS was used to determine the modification abundance (modification/total nucleoside %) for tRNA^Phe purified from either wild-type or trmA E. coli for (C) m⁵U, (D) Ψ, (E) i⁶A, and (F) acp³U. * corresponds to a significant alteration where P < 0.05. We estimate the i⁶A stoichiometry to be 0.8 modifications per tRNA^Phe,-m5U. (G) Normalized modification abundance (modification/main base %) in WT BY4741 S. cerevisiae grown without antibiotic (gray), 100 ng/mL cycloheximide (red), or 50 μg/mL hygromycin B (blue) for the subset of modifications found in the stem-loops of S. cerevisae tRNAs.

m⁵U54 in tRNA^Phe Does Not Broadly Influence the Rate Constant for Amino Acid Addition.

The increased sensitivity of the trmA(C324A)-expressing E. coli for growth in sub-MIC concentrations of the translation elongation inhibitors raises the possibility that m⁵U54 modulates the elongation step in the protein synthesis pathway. We implemented a well-established fully reconstituted E. coli in vitro translation system to directly test this supposition (25). This system has long been used to conduct high-resolution kinetic studies investigating how the ribosome decodes mRNAs. The core mechanism of translation elongation is well conserved between bacteria and eukaryotes (26, 27), and prior studies demonstrate that mRNA modifications that slow elongation and/or change mRNA decoding elongation in the reconstituted E. coli system also do so in eukaryotes (28, 29). We first evaluated whether m⁵U54 influences the rate of amino acid addition by comparing the rate constants for a single amino acid addition to a growing polypeptide using E. coli tRNA purified from either wild type (tRNA^Phe) or trmA cell lines (tRNA^Phe,-m5U) (Fig. 5 A and B).

Fig. 5.

Loss of m⁵U54 and hygromycin B have antagonistic effects on translation in vitro. (A) Duplicate timecourses showing ^fMF dipeptide formation on mRNAs containing an unmodified UUU or UUC codon or with an m⁵U in the indicated positions, using tRNA^Phe purified from either wild-type or trmA E. coli. The color of each line corresponds with the color of the labeled bar in panel B. Two independent time courses were collected for each condition. (B) Fitted k_obs values for dipeptide formation from the curves in panel A for each combination of conditions (modified/unmodified codon, modified/unmodified tRNA^Phe). “N” or “native” refers to tRNA^Phe purified from wild-type E. coli. “KO” refers to tRNA^Phe purified from trmA E. coli. (C) ^fMFK tripeptide formation using tRNA^Phe purified from wild-type E. coli. (D) ^fMFK tripeptide formation using tRNA^Phe purified from trmA E. coli. For (C) and (D), triplicate experiments were performed. In both cases, tRNA^Lys from wild-type E. coli was used. (E) Global fitting was used to obtain observed rate constants for the formation of dipeptide (k_obs,1) and tripeptide (k_obs,2) in panels C and D. tRNA^Phe lacking m⁵U showed a decreased observed rate constant for tripeptide formation. Error bars are the 95% CI on the fitted values of k_obs,2. (F) The yield of ^fMFK tripeptide in the presence of hygromycin B is increased when tRNA^Phe from trmA E. coli is used. The amount of tripeptide formed is plotted for two independent experiments under each condition.

In our assays, E. coli 70S ribosome initiation complexes were formed on mRNAs encoding a ^fMet–Phe–Lys tripeptide, with ³⁵S-labeled ^fMet-tRNA^fMet bound in the ribosome P site. Initiation complexes were reacted with an excess of ternary complex containing a single aminoacylated tRNA (Phe-tRNA^Phe or Phe-tRNA^Phe,-m5U), elongation factor Tu (EF-Tu), and GTP. The formation of ³⁵S-^fMet–Phe dipeptide product is followed over time to determine the observed rate constant (k_obs) for Phe addition. tRNA^Phe,-m5U was aminoacylated as efficiently as tRNA^Phe by an excess of recombinant E. coli phenylalanine tRNA synthetase. Consistent with previous reports, the rate constants for Phe incorporation that we measured are comparable for tRNA^Phe and tRNA^Phe,-m5U (k_obs ~ 5 s⁻¹; Fig. 5 A and B).

Additionally, we investigated the possibility of cooperative effects between m⁵U containing mRNA codons and m⁵U54 in tRNA (19–22). To accomplish this, we interrogated how amino acid addition on m⁵U-containing codons (1st, 2nd, and 3rd position modified UUU or UUC codons) is impacted when decoded by tRNA^Phe,-m5U. These findings were compared with assays we previously published investigating amino acid addition on m⁵U-modified codons using a fully modified tRNA^Phe (19). As on the unmodified codons, the rate constants for amino acid addition were generally unchanged when tRNA^Phe and tRNA^Phe,-m5U were used in translation assays (Fig. 5 A and B), with the exception that tRNA^Phe,-m5U exhibits a slightly reduced (2.3 ± 0.4 fold) rate constant when m⁵U is present in the wobble position of the codon (19). Our data collectively demonstrate that loss of m⁵U54 in tRNA^Phe,-m5U does not substantially alter the overall rate constant for Phe addition on unmodified or m⁵U modified codons.

Loss of m⁵U54 in P Site tRNA^Phe Enables Ribosome Translocation in the Presence of Hygromycin B.

Hygromycin B blocks translation by interacting with the RNA in the ribosome A site to prevent translocation (30, 31). In the single amino acid addition assays described above, the ribosome does not need to translocate to form the ^fMet–Phe dipeptide. As such, these assays do not report on the step in protein synthesis for which we have a cellular phenotype (Fig. 2). Therefore, we next examined the impact of m⁵U54 by following the formation of ^fMet–Phe–Lys tripeptide, which requires the ribosome to translocate. In these experiments, we treated 70S initiation complexes with ternary complexes that included either EF-Tu:Phe-tRNA^Phe:GTP or Phe-tRNA^Phe,-m5U, EF-Tu:Lys-tRNA^Lys:GTP, and the elongation factor EFG required for efficient ribosome translocation. In the absence of hygromycin B, ^fMet–Phe–Lys synthesis robustly occurred with both Phe-tRNA^Phe and Phe-tRNA^Phe,-m5U (Fig. 5 C and D). The observed rate constant for dipeptide formation was unaffected, but the rate constant for tripeptide formation was slightly reduced (1.6 ± 0.4 fold) when tRNA^Phe,-m5U was used (Fig. 5E). When hygromycin B was included in the reaction mix, the formation of the ^fMet–Phe dipeptide was still rapid with tRNA^Phe and tRNA^Phe,-m5U. However, as expected, no ^fMet-Phe-Lys formation is observed in the presence of hygromycin B in assays performed with tRNA^Phe (Fig. 5F). In contrast, a small amount of ^fMet–Phe–Lys is generated when tRNA^Phe,-m5U is used instead (Fig. 5F), suggesting an explanation for our observation that trm2Δ S. cerevisiae cells produce more reporter protein than wild-type cells in the presence of hygromycin B (SI Appendix, Fig. S3). Similarly, we find that the propensity of E. coli 70S ribosomes to undergo frameshifting, a process that takes place during translocation, differs in the presence of total aminoacylated-tRNA pools purified from wild type and trmA E.coli cells (SI Appendix, Fig. S7). Together, these data suggest that inclusion of m⁵U54 in tRNA^Phe impacts the step in the elongation cycle between peptidyl transfer and subsequent aa-tRNA binding, translocation.

TrmA Modulates the Modification Landscape of tRNA^Phe.

It is becoming apparent that “modification circuits” exist in tRNAs, and a hierarchy has been observed for the installation of modifications in the T-loop region of yeast tRNAs, with m⁵U54 positively influencing the introduction of m¹A58 (32, 33). This led us consider that loss of m⁵U54 in trmA E. coli might change the stoichiometry of other tRNA modifications. To test this possibility, we used LC-MS/MS to quantitatively measure the levels of the m⁵U, 4-thiouridine (s⁴U), dihydrouridine (D), pseudouridine (Ψ), 7-methylguanosine (m⁷G), and 3-(3-amino-3-carboxypropyl)uridine (acp³U) modifications in tRNA^Phe purified from wild type and ΔtrmA E. coli (tRNA^Phe,-m5U). Additionally, we examined the levels the N6-isopentenyladenosine (i⁶A) component of the 2-methylthio-N6-isopentenyladenosine (ms²i⁶) modification, as we were unable to acquire a sufficiently soluble nucleoside standard required to directly measure ms²i⁶A levels in our multiplexed LC-MS/MS approach. We find that m⁵U54 is completely lost from tRNA^Phe,-m5U, as expected (SI Appendix, Fig. S8). Furthermore, while most other tRNA^Phe modifications were unaltered (Fig. 4G and SI Appendix, Fig. S8), the levels of two modifications did change. There was a moderate decrease in the inclusion of the variable loop modification acp³U (1.7-fold, Fig. 4F) and a significant increase in i⁶A (12-fold, Fig. 4E), consistent with observations made using orthogonal approaches (34). We posit that the increased level of i⁶A likely results from the depletion of the ms²i⁶A37 hypermodification. Thus, it appears that, like its yeast homolog Trm2, the action of TrmA influences the maturation of sites on tRNAs beyond m⁵U54.

S. cerevisiae tRNA Anticodon Stem Loop Modification Landscape Is Impacted by Antibiotic Treatment.

While tRNA modifications have long been thought to be stoichiometric and statically incorporated, evidence is emerging that many modifications are incorporated at sub-stoichiometric levels and that changes in tRNA modification status can modulate the sensitivity of cells antifungal, anticancer, and antibacterial drugs (35–38). Our dual observations that the loss of trmA changes the modification profile of E. coli tRNA^Phe and that lack of 5-methyluridine transferases in both bacteria (trmA) and yeast (trm2Δ) alter the sensitivity of cells to antibiotics led us to wonder whether the translational inhibitors that we investigated here impact the tRNA modification landscape. To test this, we again turned to our highly sensitive multiplexed UHPLC-MS/MS assay to measure the level of 50 RNA modifications in total RNA isolated from S. cerevisiae treated with cycloheximide and hygromycin B. These studies revealed that levels of tRNA-specific modifications, but not rRNA-specific modifications (e.g., m³U), fluctuate in response to the antibiotics (Fig. 4G and SI Appendix, Fig. S9). The largest effects were observed for modifications incorporated in tRNA anticodon stem loops at position 37 (e.g., m¹I, t⁶A, i⁶A, and m¹G) (Fig. 4 C–F and SI Appendix, Fig. S9). In general, cycloheximide and hygromycin B treatment had opposite effects on modification incorporation—with cycloheximide reducing modification levels, and hygromycin B promoting their inclusion. The most significant differences between conditions were seen for i⁶A, t⁶A, m¹I, and m¹G—which are roughly twice as prevalent in RNAs isolated from hygromycin-treated cells than those treated with cycloheximide. All four of these modifications (and their related derivative modifications) are well-established modulators of ribosome elongation speed and fidelity (39).

Discussion

There are few RNA modifications as highly conserved across biology as m⁵U54 is in tRNAs. The universal addition of this modification is a long-standing mystery, as the enzymes that incorporate m⁵U54 are not essential, and previous studies suggest that loss of m⁵U54 does not affect the step in translation that most tRNAs participate in, elongation (14–17). To address this conundrum, we employed genetic methods to study translation in cells lacking this modification and in vitro biochemistry using tRNAs without m⁵U54 in their TΨC loop. We find that m⁵U54 modulates translocation in vitro, consistent with published data indicating that the loss of m⁵U54 subtly alters translation efficiency in vitro and in cells (Fig. 2 and SI Appendix, Fig. S3) (39, 40). Furthermore, we reveal that the efficacy of the translocation inhibitors hygromycin B and cycloheximide is altered in strains lacking m⁵U54 (Fig. 2). Our data provide evidence that m⁵U54 influences gene expression, as hygromycin B induces substantial changes in the mRNAs present in wild type cells that are alleviated by loss of the m⁵U54 modifying enzyme (trm2Δ) (compare Fig. 3B vs. Fig. 3C). Finally, these studies reveal that m⁵U54 installation is critical for maintaining the global tRNA landscape, significantly promoting the inclusion of modifications in the anticodon stem loop (e.g., ms²i⁶A, t⁶A, m¹I; Fig. 4 C–F and SI Appendix, Fig. S8). Previous observations indicate that m⁵U54 limits the dynamics of the TΨC loop, suggesting that tRNAs lacking m⁵U54 are more flexible than their fully modified counterparts (23, 41). Together with our findings, this suggests a model wherein m⁵U54 rigidifies the TΨC loop to maintain the critical link between mRNA and peptide sequence, at the cost of decreased cellular fitness under some environmental stresses, and indeed, knockdown of TRM2A in human cells has been shown to reduce translational fidelity (42).

As discussed above, our data indicate that one biological consequence of m⁵U54 is the modulation of translocation, both directly and indirectly (through subtly altering the tRNA modification landscape). In the protein synthesis pathway, translocation takes place following the transfer of the growing polypeptide chain from the P site peptidyl-tRNA to an aminoacyl-tRNA bound in the ribosome A site. Elongation factor G (EFG) is responsible for catalyzing the movement of the ribosome to reposition the A site peptidyl-tRNA in the P site. Hygromycin B blocks this process by flipping a key 16S rRNA nucleotide (A1493) between the A and P sites (30, 43, 44). Increased flexibility in the TΨC loop of m⁵U54-deplete tRNAs may provide a satisfying molecular rationale for our observation that trm2Δ and tRNA^Phe,-m5U partially overcome hygromycin B inhibition in cells and in vitro (Figs. 2, 3, and 5). Other hypothetical effects of hygromycin B treatment, such as changes to rRNA modifications (45), would not be overcome by loss of m⁵U54, consistent with our in vitro observations with untreated ribosomes. Such a model suggests that in unstressed circumstances m⁵U54 modulates translocation, as we observed with tRNA^Phe,-m5U (Fig. 6). This is consistent with long-perplexing findings from Nierhaus and coworkers that lysine peptides are generated more quickly in vitro from poly(A) transcripts using m⁵U54-depleted tRNAs but that neither tRNA binding or peptidyl-transfer are affected (40). Similarly, recent observations indicate that total protein production can be both decreased and increased from reporters containing select amino acid repeats in a trmA E. coli strain (34). Finally, comparison of available ribosome profiling for wild-type and trm2Δ yeast supports the supposition that m⁵U54 can subtly alter translation; there is a small, but statistically significant, global increase in the translational efficiency of ribosomes in trm2Δ cells (SI Appendix, Fig. S10) (39).

Fig. 6.

Model for mitigation of hygromycin B sensitivity by tRNA^Phe,-m5U. Hygromycin B acts to block translocation, using a noncompetitive mechanism to flip ribosome nucleotides and prevent movement of the A site peptidyl-tRNA into the P site. In this mechanism, when translation takes place with natively modified tRNAs that include m⁵U54, the observed rate constant for hygromycin B binding is much faster than that of translocation leading to translational stalling (k_{obs, hyg} >> k_{obs, translocation}). However, when m⁵U54 is missing from tRNAs the rate constant for translocation is altered sufficiently to raise the k_{obs, translocation} such that it now effectively competes with the k_{obs, hyg}, allowing for some translocation to occur and product to be generated (k_{obs, hyg} ~ k_{obs, translocation}). Given the modest levels of peptide observed in our assays, k_{obs, translocation} shifts to be within 10-fold of k_{obs, hyg} when reactions are performed with tRNA^Phe,-m5U.

This model lead us to wonder how the sometimes negative modulation of translational efficiency and translocation might be beneficial given that wild-type cells outcompete m⁵U54 depleted strains (trm2Δ and trmA) (16). Recent computational studies indicate that differences in the local dynamics of the TΨC loop in cognate and near cognate tRNAs during accommodation act as gatekeepers for ensuring accurate mRNA decoding (46). We speculate that perturbed tRNA dynamics may increase flux through a noncanonical pathway tRNA accommodation pathway that is independent of codon:anticodon interactions (46). Needless to say, increased decoupling of peptide sequence from mRNA sequence is likely to be detrimental to cells under optimal growth conditions. Additionally, by controlling translocation rates, m⁵U4 also has the potential to increase the accuracy of translation by limiting potentially deleterious processes such as frameshifting reliant on translocation (47, 48).

The implication of m⁵U54 in translocation is reasonable given that the elbow region of ribosome-bound tRNAs (formed by tertiary interactions between the TΨC- and D-loops) mediate their coordination with the ribosome (49). Therefore, residues in the elbow are likely to be critically important for efficient translocation. Consistent with this, the mutation of tRNA positions adjacent to m⁵U54 (Ψ55 and C56) decrease the rate of translocation (50). Notably, recent cryo-EM snapshots taken throughout translocation reveal that elbow moves sufficiently to alter the bend of the peptidyl-tRNA by up to 17.37° during translocation, further supporting the notion that tRNA elbow dynamics are critical (49). Our studies add to the idea that the TΨC plays a crucial role in translocation, suggesting that m⁵U54 helps to maintain the tRNA elbow region shape and dynamics necessary for mediating tRNA interactions with the ribosome during translocation.

Not all of the subtle changes in translation observed in cells lacking m⁵U54 likely arise from m⁵U54 directly tweaking translocation. Our LC-MS/MS analyses of tRNA modification landscapes indicate that m⁵U54 installation is correlated to alterations in the modification status of other tRNA positions, including acp³U and ms²i⁶A in E. coli tRNA^Phe (Fig. 4 C–F). These results agree with the findings of MSR-tRNA sequencing studies by Kothe et al. demonstrating that acp³U, ms²i⁶A, and s⁴U levels are altered in several (but not all) tRNAs, including tRNA^Phe, in trmA cells (34). Presumably, as not all tRNAs possess the same set of modifications, tRNA-specific changes may arise that have differential impacts on translocation.

The seemingly small changes in translocation or tRNA modification stoichiometry that we observe upon the loss of m⁵U54 can have an outsized impact on gene expression under stress conditions. Our work reveals that, under stress, Trm2 helps to shape the mRNA landscape (Fig. 3), supporting the idea that m⁵U54 alters translational efficiency given that translation rates are tied to mRNA stability (39). This is in line with recent studies suggesting that translocation, despite not being the overall rate limiting step for protein synthesis, nevertheless can impact protein production (49, 51). Although the effects that we observed on MIC were modest, the increased sensitivity of trmA(C324A) to hygromycin B augments the growing body of literature suggesting that tRNA-modifying enzymes play significant roles in altering gene expression to modulate the development of drug resistance (37).

Collectively, our findings address the decades long question of why m⁵U54 is conserved across tRNAs in all organisms, identifying multiple roles for the modification in the protein synthesis pathway. Our LC-MS/MS analyses reveal that loss of the enzyme responsible m⁵U54 installation alters the tRNA modification landscape in both S. cerevisiae and E. coli. Additionally, kinetic and cell-based assays provide evidence that m⁵U54 serves to modulate the translocation step in protein synthesis, to ultimately impact gene expression (especially when cells are under translational stress). Our data support structural work revealing that the tRNA TΨC stem loop forms critical interactions with the ribosome during translocation and demonstrates how even apparently small perturbations in the modification profiles and inherent dynamic landscape of tRNAs are important enough to drive the conservation of TΨC modifications throughout biology. As we seek to assign biological functions for other RNA modifications in the mechanism of protein synthesis, the findings of this study suggest that even relatively subtle changes in the dynamic processes that drive translation are likely significant in nature, when cells perform under nonoptimized (nonlaboratory) conditions.

Methods

Spot Plating Growth Assay and Growth Curve Characterization under Stress.

Wild-type and trm2Δ cells were inoculated into 3 mL YPD and grown overnight. These cultures were diluted to OD₆₀₀ = 1, and 7 μL of 10-fold serial dilutions were spotted on fresh YPD agar plates including 0.75 to 1.0 M NaCl, 250 mM MgSO₄, 200 μM puromycin, 100 ng/mL cycloheximide, 25 to 50 μg/mL hygromycin B, 50 μM MG132, and 1.5 to 3 mg/mL paromomycin. Growth of the cells was also tested in the presence of different carbon sources including 2% glucose, 2% sucrose, 2% galactose, and 3% glycerol in YEP agar media (1% S. cerevisiae extract and 2% peptone). The plates were incubated for 2 to 5 d at 30 °C unless indicated.

For growth curves, wild-type and trm2Δ cells were inoculated into 5 mL YPD and grown overnight. Cultures were then diluted to a starting OD₆₀₀ = 0.05 to 0.1 in 100 mL YPD media containing either 1 M NaCl, 0.1 µg/mL cycloheximide, 50 µg/mL hygromycin B, or 3 mg/mL paromomycin. Cultures were grown in duplicate at 30 °C with shaking unless indicated, and growth was monitored by OD₆₀₀.

Minimum Inhibitory Concentration of Hygromycin B in S. cerevisiae and E. coli.

Wild-type BY4741 and trm2Δ BY4741 (Dharmacon) S. cerevisiae were inoculated in 5 mL of YPD and grown overnight at 30 °C shaking at 285 RPM. Overnight cultures were diluted to 0.02 OD₆₀₀ in YPD containing various concentrations of hygromycin B. The cells were cultured for 24 h at 30 °C shaking at 285 RPM. The growth was measured by OD₆₀₀.

The E. coli Keio knockout parental strain BW25113 (Dharmacon) and trmA E. coli JW3937 (Dharmacon) were inoculated in 5 mL of LB and grown overnight at 37 °C shaking at 250 RPM. Overnight cultures were diluted to 0.002 OD₆₀₀ in LB containing various concentrations of hygromycin B. The cells were cultured for 24 h at 37 °C shaking at 250 RPM. The growth was measured by OD₆₀₀. The trmA structural gene and the trmA(C324A) variant were synthesized by Twist in the pTwist medium-copy vector, transformed into chemically competent trmA E. coli, and assayed for hygromycin B inhibition according to the same protocol. Plasmid sequences are available in Dataset S4.

Luciferase Reporter Assay.

Dual-luciferase reporter plasmid pJDRaugFaug (52, 53) was transformed into wild-type and trm2Δ S. cerevisiae by the standard lithium-acetate/PEG protocol. The plasmid sequence is included in Dataset S4. The cells were streaked onto CSM-URA agar plates to isolate single colonies (54). CSM-URA media (30 mL) were inoculated with a single colony and allowed to grow overnight at 30 °C and 250 RPM. The cells were diluted to an OD₆₀₀ of 0.05 with 500 mL of CSM-URA medium and were grown to an OD₆₀₀ of 0.5 at 30 °C and 250 RPM. At this point, the cells were stressed with hygromycin B or cycloheximide and were allowed to continue to grow. At time points of 0 min, 20 min, and 150 min after the translational stress, 10 mL of culture was pelleted at 8,000 × g for 10 min. The cell pellet was washed with 1 mL of water prior to storage at −80 °C until the assay was performed.

The cell pellets were washed with 500 μL of breaking buffer (50 mM sodium phosphate pH 7.4, 2 mM EDTA, 10% glycerol, 2 mM PMSF) prior to resuspension in 200 μL of breaking buffer. An equal volume of acid-washed glass beads were added, and the tubes were vortexed for 30 s followed by 30 s on ice. The vortexing and ice incubation were repeated three times for a total of four times. The lysates were centrifuged for 21,000 × g for 15 min at 4 °C. The supernatant (30 μL per replicated) was characterized by the Promega dual luciferase assay kit.

S. cerevisiae Cell Growth and mRNA Purification.

Wild-type BY4741 and trm2Δ S. cerevisiae (Dharmacon) were grown in YPD medium as previously described (55). trm2Δ Saccharomyces cerevisiae were grown in the presence of 200 μg/mL Geneticin. Briefly, 10 mL of YPD medium was inoculated with a single colony selected from a plate and allowed to grow overnight at 30 °C and 250 RPM. The cells were diluted to an OD₆₀₀ of 0.1 with 200 mL of YPD medium and were grown to an OD₆₀₀ between 0.6 and 0.8 at 30 °C and 250 RPM. Translational stress S. cerevisiae were grown with 50 μg/mL hygromycin B or 100 ng/mL cycloheximide. Hygromycin B S. cerevisiae were grown to an OD₆₀₀ of 0.4 to ensure cells were in mid-log phase growth. This cell culture was pelleted at 15,000 × g at 4 °C and used for the RNA extraction.

S. cerevisiae cells were lysed as previously described with minor alterations (19). The 200 mL cell pellet was resuspended in 8 mL of lysis buffer (60 mM sodium acetate pH 5.5, 8.4 mM EDTA) and 800 μL of 10% SDS. One volume (8.8 mL) of phenol was added and vigorously vortexed. The mixture was incubated at 65 °C for 5 min and was again vigorously vortexed. The incubation at 65 °C and vortexing was repeated once. Then, the mixture was rapidly chilled in an ethanol/dry ice bath and centrifuged for 15 min at 15,000 × g. The total RNA was extracted from the upper aqueous phase using a standard acid phenol-chloroform extraction. The extracted total RNA was treated with 140 U RNase-free DNase I (Roche, 10 U/μL) at 37 °C for 30 min. The DNase I was removed through an acid phenol-chloroform extraction. The resulting total RNA was used for our UHPLC-MS/MS, bioanalyzer, and RNA-seq analyses.

mRNA was purified through a three-step purification pipeline (19). First, small RNA (tRNA and small rRNA) was diminished using the MEGAclear Transcription Clean-Up Kit (Invitrogen) to purify RNA >200 nt. Then, two consecutive Dynabeads oligo-dT magnetic bead selections (Invitrogen, USA) were used to purify poly(A) RNAs from 140 μg of small RNA depleted RNA. The resulting RNA was ethanol precipitated and resuspended in 14 μL. Subsequently, we used the commercial riboPOOL rRNA depletion kit (siTOOLs Biotech, Germany) to remove residual 5S, 5.8S, 18S, and 28S rRNA. The Bioanalyzer RNA 6000 Pico Kit (Agilent, USA) was used to evaluate the purity of the mRNA prior to UHPLC-MS/MS analysis.

qRT-PCR.

The RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, USA) was used to reverse transcribe DNase I-treated total RNA and three-stage purified mRNA (200 ng) using the random hexamer primer. The resulting cDNA was diluted 5,000-fold and 1 μL of the resulting mixture was analyzed using the Luminaris Color HiGreen qPCR Master Mix (Thermo Scientific, USA) with gene-specific primers (19).

RNA-seq.

The WT S. cerevisiae total RNA and mRNA were analyzed by RNA-seq as previously described by paired-end sequencing using 2.5% of an Illumina NovaSeq (S4) 300 cycle sequencing platform flow cell (0.625% of flow cell for each sample) (19). All sequence data are paired-end 150 bp reads. The reads were trimmed using Cutadapt v2.3 (56). The reads were evaluated with FastQC (57) (v0.11.8) to determine the quality of the data. Reads were mapped to the reference genome Saccharomyces_cerevisiae (ENSEMBL) using STAR v2.7.8a and assigned count estimates to genes with RSEM v1.3.3 (58, 59). Alignment options followed ENCODE standards for RNA-seq (Dobin). QC metrics from several different steps in the pipeline were aggregated by multiQC v1.7 (60). Differential expression analysis was performed using DESeq2 (61).https://github.com/BioinfoHR/coRdon

RNA Enzymatic Digestion and UHPLC-MS/MS Ribonucleoside Analysis.

Total RNA and mRNA (125 ng) were digested for each condition. The RNA was hydrolyzed to composite mononucleosides using a two-step enzymatic reaction and quantified using LC-MS/MS as previously described with no alterations (19).

E. coli Ribosomes, and Translation Factors tRNA and mRNA for In Vitro Assay.

Ribosomes were purified from E. coli MRE600 as previously described (28). All constructs for translation factors were provided by the Green lab unless specifically stated otherwise. Expression and purification of translation factors were carried out as previously described (28). Unmodified transcripts were prepared using runoff transcription of a DNA template. Modified mRNA sequences containing m⁵U were purchased HPLC purified from Dharmacon. The mRNA sequence used was 5′-GGUGUCUUGCGAGGAUAAGUGCAUUAUGUUCUAAGCCCUUCUGUAGCCA-3′, with the coding sequence underlined. The m⁵U modified position was always the first position in the UUC phenylalanine codon.

Native E. coli tRNA^Phe was purified as previously described with minor alterations (25). Bulk E. coli transfer RNA was purified in E. coli from HB101 containing pUC57-tRNA^phe provided by Yury Polikanov. Two liters of enriched Terrific Broth (TB) media [TB, 4 mL glycerol/L, 50 mM NH₄Cl, 2 mM MgSO₄, 0.1 mM FeCl₃, 0.05% glucose and 0.2% lactose (if autoinduction media was used)] were inoculated with 1:400 dilution of a saturated overnight culture and incubated at 37 °C overnight shaking at 250 RPM with 400 mg/mL of ampicillin. Cells were harvested the next morning by 30 min centrifugation at 5,000 RPM and then stored at −80 °C until lysis was performed. To lyse the cells, the pellet was resuspended in 200 mL of resuspension buffer (20 mM Tris-Cl, 20 mM Mg(OAc)₂ pH 7) and 100 mL of phenol:chloroform:isoamyl alcohol pH 4.5 (125:24:1) was added. The mixtures were incubated at 4 °C shaking at 250 RPM for 1 h. After incubation, the lysate was centrifuged at 3,220 × g for 60 min at 4 °C, and the aqueous supernatant was transferred to a fresh tube. The phenol:chloroform:isoamyl alcohol was washed with 100 mL of 18 MΩ water, the mixture was centrifuged at 3,220 × g for 60 min at 4 °C, and the aqueous supernatant was transferred to the same tube. The pooled aqueous supernatant was washed with 100 mL chloroform, centrifuged at 3,220 × g for 60 min at 4 °C, and the supernatant was transferred to a fresh tube. DNA was precipitated with 150 mM NaOAc pH 5.2 and 20% isopropanol. The DNA was removed through centrifugation at 13,700 × g for 60 min at 4 °C. The supernatant was transferred to a fresh tube, the isopropanol content was raised to 60%, and the bottle was stored at −20 °C to precipitate the RNA. The precipitation was centrifuged at 13,700 × g for 60 min at 4 °C and the supernatant was removed. The pellet was resuspended in 10 mL 200 mM tris-acetate pH 8.0 and incubated at 37 °C for 2 h. After incubation, the RNA was precipitated through the addition of 1/10th volume of 3 M NaOAC pH 5.2 and 2.5 volumes of 100% ethanol. The precipitate could either be stored at −20 °C or centrifuged immediately at 16,000 × g for 60 min at 4 °C. The pellet was washed with 70% ethanol, resuspended in 5 to 10 mL of water, and desalted using a 3K MWCO Amicon centrifugal concentrator.

Total tRNA was isolated by anion-exchange chromatography on a FPLC using a Cytiva Resource Q column. The mobile phase A was 50 mM NH₄OAc, 300 mM NaCl, 10 mM MgCl₂, and mobile phase B was 50 mM NH₄OAc, 800 mM NaCl, 10 mM MgCl₂. The resuspend RNA was injected (5 mL) and eluted with a linear gradient from 0 to 50% B over 18 column volumes. Fractions were pulled and ethanol precipitated overnight at −20 °C.

The RNA was centrifuged at 16,000 × g for 30 min at 4 °C, washed with 70% ethanol, and resuspended in approximately 500 μL of water. tRNA^Phe was purified from total tRNA using a Waters XBridge BEH C18 OBD Semiprep column (10 × 250 mm, 300 Å, 5 μm). Mobile phase A was 20 mM NH₄OAc, 10 mM MgCl₂, 400 mM NaCl at pH 5 and mobile phase B was 20 mM NH₄OAc, 10 mM MgCl₂, 400 mM NaCl at pH 5 with 60% methanol. Total tRNA was injected (400 μL) and separated with a linear gradient of buffer B from 0 to 35% was done over 35 min. The MPB% was increased to 100% over 5 min and held at 100% MPB for 10 min. The column was then equilibrated for 10 column volumes before the next injection at 0% MPA. Each fraction was analyzed by A₂₆₀ absorbance and amino acid acceptor activity to identify fractions that contain tRNA^Phe.

Formation of E. coli Ribosome Initiation Complexes.

Initiation complexes (IC’s) were formed in 1× 219-Tris buffer (50 mM Tris pH 7.5, 70 mM NH₄Cl, 30 mM KCl, 7 mM MgCl₂, 5 mM ß-ME) with 1 mM GTP as previously described (25). Then, 70S ribosomes were incubated with 1 μM mRNA (with or without modification), initiation factors (1,2,3) all at 2 μM final and 2 μM of radiolabeled ^fMet-tRNA for 30 min at 37 °C. After incubation, MgCl₂ was added to a final concentration of 12 mM. The ribosome mixture was then layered onto 1 mL cold buffer D (20 mM Tris-Cl, 1.1 M sucrose, 500 mM NH₄Cl, 10 mM MgCl₂, 0.5 mM disodium EDTA, pH 7.5) and centrifuged at 69,000 rpm for 2 h at 4 °C. After pelleting, the supernatant was discarded into radioactive waste, and the pellet was resuspended in 1× 219-tris buffer and stored at −80 °C.

In Vitro Amino Acid Addition Assays: Dipeptide Formation.

Initiation complexes were diluted to 140 nM with 1× 219-tris buffer. Ternary complexes (TCs) were formed by first preloading EF-Tu with GTP (1× 219-tris buffer, 10 mM GTP, 60 μM EFTu, 1 μM EFTs) at 37 °C for 10 min. The EF-Tu mixture was incubated with the tRNA mixture (1× 219-tris buffer, Phe-tRNA^Phe (1 to 10 μM), 1 mM GTP) for another 15 min at 37 °C. After TC formation was complete, equal volumes of IC complexes (70 nM) and ternary complex (1 μM) were mixed either by hand or using a KinTek quench-flow apparatus. Discrete time-points (0 to 600 s) were taken as to obtain observed rate constants on m⁵U-containing mRNAs. Each time point was quenched with 500 mM KOH (final concentration). Time points were then separated by electrophoretic TLC and visualized using phosphorimaging as previously described (25, 28). Images were quantified with ImageQuant. The data were fit using the equation below:

$$\mathrm{Fraction} \mathrm{product}=A·\left(1-e^{k_{\mathit{obs}}t}\right).$$

In Vitro Translation Amino Acid Addition Assays for Tripeptide Formation.

Initiation complexes were diluted to 140 nM with 1× 219-Tris buffer. Ternary complexes (TCs) were formed by first preloading EF-Tu with GTP (1× 219-tris buffer, 10 mM GTP, 60 μM EFTu, 1 μM EFTs) at 37 °C for 10 min. The EF-Tu mixture was incubated with the tRNA mixture [2 μM aminoacyl-tRNA Phe/Lys(s), 24 μM EF-G, 60 μM EF-Tu] with ICs (140 nM) in 219-Tris buffer (50 mM Tris pH 7.5, 70 mM NH₄Cl, 30 mM KCl, 7 mM MgCl₂, 5 mM βME) for 15 min at 37 °C. These experiments are done with both native phenylalanine tRNA or our Δm5U phenylalanine tRNA. After TC formation was complete, equal volumes of IC complexes (70 nM) and ternary complex (1 μM) were mixed using a KinTek quench-flow apparatus. Discrete time-points (0 to 600 s) were taken to obtain observed rate constants on nonmodified mRNAs, containing a UUC phenylalanine codon. Each time point was quenched with 500 mM KOH (final concentration). Time points were then separated by electrophoretic TLC and visualized using phosphorimaging as previously described (25, 28). Images were quantified with ImageQuant. The data were fit using Eq. 1 as previously described.

In Vitro Translation Amino Acid Addition Assays for Tripeptide Formation in the Presence of Hygromycin B.

Initiation complexes were diluted to 140 nM with 1× 219-tris buffer. Ternary complexes (TCs) were formed by first preloading EF-Tu with GTP (1× 219-tris buffer, 10 mM GTP, 120 μM EFTu, PEP, 12 mM, PK .40 μM, 40 μM EFTs) at 37 °C for 10 min. The EF-Tu mixture was incubated with the tRNA mixture [20 to 60 μM aminoacyl-tRNA Phe/Lys(s), 24 μM EF-G, 60 μM EF-Tu] with ICs (140 nM) in 219-tris buffer (50 mM Tris pH 7.5, 70 mM NH₄Cl, 30 mM KCl, 7 mM MgCl₂, 5 mM βME) for another 15 min at 37 °C. These experiments are done with both native phenylalanine tRNA or our Δm5U phenylalanine tRNA. After TC formation was complete, 50 μg/mL of Hygromycin B was added to the IC complex. Then, by hand equal volumes of IC complexes (70 nM) and ternary complex (1 μM) were mixed, and discrete time-points (0 to 600 s) were taken to obtain observed rate constants on nonmodified mRNAs, containing a UUC phenylalanine codon. Each time point was quenched with 500 mM KOH (final concentration). Time points were then separated by electrophoretic TLC and visualized using phosphorimaging as previously described (25, 28). Images were quantified with ImageQuant. The data were fit using the equation as described above.

Fitting Observed Rate Constants and Global Analysis Simulations of Amino Acid Addition.

Fits to obtain observed rate constant k₁ was done using a single differential equation in KaleidaGraph for products in dipeptide formation. When multiple peptide products were formed, the disappearance of ^fMet product was fit using a single exponential equation in KaleidaGraph to get an observed rate constant k₁. This value was then used in KinTex Explorer to measure subsequent rate constant k₂ using simulations. Simulations were modeled against the equation:

$${}^f\text{M}+\text{F}{\to }^f\text{MF}+\text{K}{\to }^f\text{MFK}$$

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Dataset S02 (GZ)

Dataset S03 (XLSX)

Dataset S04 (TXT)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. RNA-seq data is available in the GEO database (GSE273506) (62).