The tRNA dihydrouridine synthase DusA has a distinct mechanism in optimizing tRNAs for translation

Abstract

Dihydrouridine (D) is one of the most highly conserved RNA modifications across all domains of life. D20 within the tRNA D loop is particularly conserved and is formed by DusA in Escherichia coli. However, the mechanisms and cellular functions of DusA and D20 remain poorly understood. Here, we characterize DusA’s role in tRNA binding, cofactor oxidation, and modification activity, along with its impact on tRNA maturation and translation. We find that DusA binds tRNA via a two-step mechanism involving a local structural rearrangement and exhibits a higher affinity for previously modified tRNA compared to unmodified tRNA. Unlike the T arm modifying enzymes TrmA and TruB, DusA does not broadly increase cellular aminoacylation for all tRNAs but enhances the charging of specific tRNA species. Despite limited alterations in overall tRNA charging and abundance in cells lacking DusA, DusA selectively improves translation at several specific codons, potentially indicating a direct contribution for dihydrouridine to the function of certain tRNAs on the ribosome. In conclusion, our findings suggest DusA acts nonredundantly with and complementary to TrmA and TruB in fine-tuning protein synthesis.

Graphical Abstract

Graphical Abstract.

Introduction

All transfer RNAs (tRNAs) are extensively and diversely modified [1]. The RNA modifying enzymes that introduce tRNA modifications play a variety of roles in tRNA maturation and function [2]. Across all domains of life, almost every tRNA features at least one dihydrouridine (D) modification, found most often within the eponymous tRNA D loop [3–5]. Whereas most RNA modifications stabilize the tRNA structure, the unique nonplanar nucleobase of dihydrouridine cannot participate in stabilizing base stacking interactions and primarily adopts the flexible C2ʹ endo ribose conformation (Fig. 1A) [4–6]. As such, dihydrouridine provides local flexibility within the tRNA structure, which may facilitate tertiary base pairing in the tRNA elbow, thereby stabilizing the overall tRNA structure [7, 8]. Supporting a functional role for dihydrouridine in increasing tRNA flexibility, tRNAs from psychrophilic organisms tend to contain more dihydrouridine than tRNAs from mesophiles and thermophiles [9, 10].

Figure 1.

Characterization of Escherichia coli DusA activity. (A) Structure of dihydrouridine, with the reduced C5–C6 bond as compared to uridine coloured in pink. (B) Structure of Thermus thermophilus Dus (TthDus) in complex with tRNA^Phe and flavin mononucleotide (FMN; PDB 3B0V) [33]. TthDus is grey, tRNA^Phe is light teal, and FMN is shown as spheres and coloured by atom (green: carbon, blue: nitrogen, red: oxygen). Residues C93 (equivalent to E. coli DusA C114) and K132 (equivalent to E. coli DusA K153) are shown as spheres, coloured in dark pink and light pink, respectively (residue numbering in figure is according to E. coli numbering). The target base, U20, is shown as spheres and coloured teal. (C) Zoomed in view of the TthDus active site, with residues coloured as in panel (B), but shown as sticks. (D) Dihydrouridine formation following 1 h incubation of equimolar concentrations of in vitro transcribed tRNA^Phe and purified DusA enzyme or no enzyme in the presence of FMN and NADPH (n = 3 independent modification reactions) determined by mass spectrometry. Escherichia coli tRNA^Phe contains 16 U nucleotides; therefore, 100% modification at one site (U20) corresponds to 6.25% D/total U (dashed line). (E) Absorbance spectra for purified E. coli DusA wildtype (black), and C114A (dark pink) and K153A (light pink) variants. (F) Oxidation of 500 µM NADPH (black squares) or NADH (grey circles) by 2 µM DusA. No enzyme controls for NADPH and NADH are shown as light grey squares and circles with dashed lines, respectively. (G) Oxidation of 500 µM NADPH to form NADP⁺ by 2 µM DusA wildtype (black), C114A (dark pink), and K153A (light pink). Table 2 lists initial velocities for each reaction in panels (F) and (G).

It is speculated that all dihydrouridine synthases (Dus) arose from duplications of a common ancestral Dus [11, 12]. Bacteria have three major families of tRNA dihydrouridine synthases, DusA, DusB, and DusC [13], in addition to a recently discovered ribosomal RNA (rRNA) dihydrouridine synthase, RdsA [14]. In E. coli, DusA, B, and C have nonoverlapping specificities, with DusA modifying U20 and U20a, DusB modifying U17, and DusC modifying U16 [13]. DusB is regarded to be the ancestral protein and DusA and DusC are thought to have arisen from gene duplication [11, 12]. However, dihydrouridine synthase evolution in prokaryotes is complex; for example, Firmicutes harbour up to three DusB subgroups (DusB1–3), wherein DusB1 homologues can have multisite specificity that varies between species [15, 16]. Eukaryotes have four families of dihydrouridine synthases with nonoverlapping specificities, Dus1–4, where Dus1 modifies both U16 and U17, Dus2 modifies U20, Dus3 modifies U47 within the tRNA variable loop, and Dus4 modifies U20a and U20b [17]. Finally, archaeal Dus constitutes a separate family of dihydrouridine synthases that are relatively diverged from each other and not well-characterized to date [11].

DusA modifies position 20 and/or 20a in over 35 of the 46 E. coli tRNAs [18], thereby contributing about half of the bulk cellular tRNA dihydrouridine content [13]. Like other tRNA modifying enzymes, the dusA gene can be deleted from E. coli and, even in combination with additional deletion of dusB and dusC genes, no growth defects are present for the triple knockout strain grown in ideal conditions [13] or for deletion of the dusA gene from Acinetobacter baumannii in several stress conditions [19]. Similarly, no significant growth phenotype was reported for yeast lacking dus1, dus2, dus3, or dus4 in solid and liquid media at various temperatures [17]. Since the dusA gene and D20/D20a modifications are highly conserved across Proteobacteria [11, 12] and DusB orthologs have been identified to instead introduce D20/D20a modifications in several Gram-positive species that lack DusA [15, 16], tRNA U20/U20a dihydrouridylation is likely to contribute to cellular fitness, as demonstrated previously for other tRNA modifying enzymes [20]. Interestingly, the dusA gene has been shown to serve as an integration site for prophages and genomic islands of diverse functions in over 200 sequenced Proteobacterial organisms [19, 21]. Although genetic element integration disrupts dusA, in investigated cases, a new promoter and restored reading frame is provided, allowing for continued expression of DusA protein [21, 22]. In a screen of several tRNA modifying enzymes, competition experiments and transposon insertion sequencing revealed that dusA inactivation is actually beneficial to E. coli and Vibrio cholerae fitness when grown in sub-minimal inhibitory concentrations of the aminoglycoside antibiotic tobramycin [23]. Similar enhancements in cell growth under certain stress conditions have also been described for other tRNA modifying enzyme knockouts including E. coli trmA and its homolog yeast trm2 [24]. This surprising finding indicates that environmental situations exists where these highly conserved genes decrease cellular fitness, rather than explaining their high conservation.

In humans, dihydrouridine has been long known to be overabundant in tRNAs from tumor cells compared to healthy tissue [25], and the U20-modifying enzyme Dus2 has been shown to be overexpressed in clinical lung cancer samples and nonsmall cell lung cancer cell lines [26]. Indeed, Dus2 is necessary for survival and growth of lung cancer cells, and Dus2 expression inversely correlates with patient survival time [26]. The mechanism underlying the involvement of Dus2 and cancer progression remains unclear. Interestingly, Dus2 has been shown to interact with glutamyl-prolyl tRNA synthetase and protein kinase R, which may have biological significance in cancer cells [26, 27]. In addition to cancer, Dus2 may be implicated in Alzheimer’s disease [28]. Moreover, human Dus4 has recently been shown to be upregulated in lung adenocarcinoma and to be necessary for cell proliferation of A549 cells [29]. Knockout or knockdown of Dus4 in two different human cell lines results in decreased D20a abundance [30, 31]. Intriguingly, this depletion of Dus4 and D20a results in increased D20 content, along with a decreased abundance of Dus4 substrate tRNAs, suggesting an interplay between D20 and D20a modification [31]. Recently, dihydrouridine has been found within a small number of messenger RNAs (mRNAs) and small nucleolar RNAs in budding yeast, fission yeast, and human cells, but these studies were unable to detect dihydrouridine outside of tRNA and rRNA in E. coli [30, 32]. Overall, a substantive body of research implicates dihydrouridines in tRNA to cellular fitness, health and disease.

The crystal structure of T. thermophilus Dus (TthDus; a DusA-family dihydrouridine synthase) in complex with T. thermophilus tRNA^Phe and FMN has been solved [33] (Fig. 1B and C). In this structure, FMN is bound in the TthDus active site surrounded by four conserved residues including K132. In the active site, FMN forms π-stacking interactions with the target nucleobase, U20, which is flipped out from the D loop into the enzyme active site. In addition to recognition by FMN, the U20 base is directly recognized by the conserved residues C93, R178, and N90. When bound to TthDus, the tRNA^Phe D loop is distorted at U16 and U17 in addition to U20; however, the G18–U55 and G19–C56 tertiary base pairs are maintained, suggesting the tRNA elbow is undisturbed. Indeed, no other significant tRNA conformational changes are observed within the D stem, T loop, or anticodon stem although TthDus also binds these regions [33]. Thus, TthDus is likely to recognize properly folded tRNA nearing maturity. This is consistent with the lack of complex formation observed between TthDus and unmodified tRNA [33], and the fact that previous studies of purified DusA have utilized native tRNAs or bulk RNA extracted from a knockout strain rather than in vitro transcribed, unmodified tRNAs [12, 13, 18, 34].

The precise reaction mechanism of tRNA dihydrouridylation by DusA remains unknown, but it is proposed that in its reduced form, the DusA-bound flavin transfers a hydride from N5 of FMNH^- to C6 of the target uracil base, creating a nucleophilic centre at C5, which then attacks a conserved cysteine residue (C114 in E. coli, C93 in T. thermophilus) to obtain a proton [3, 4, 33]. In order to recycle the flavin prosthetic group, reduced nicotinamide adenine dinucleotide (phosphate) (NADH/NADPH) binds and reduces FMNH^- in an unknown manner [3, 33].

Despite the widespread conservation of D20/20a and its associated dihydrouridine synthase, the molecular mechanisms, relative timing of modification, and cellular impact of E. coli DusA remains poorly understood. Two similarly conserved tRNA modifying enzymes, TrmA and TruB, which form 5-methyluridine (m⁵U) 54 and pseudouridine (Ψ) 55, respectively, function as tRNA chaperones by disrupting tertiary interactions between the D and T loops to access their target base, thereby providing tRNA a second chance at correctly folding [35, 36]. TrmA and TruB are known to act early in tRNA biogenesis where they promote tRNA modification, folding, and aminoacylation to finetune mRNA translation across several specific codons [20]. In contrast to TrmA and TruB, studies of certain DusA homologs demonstrate a strong preference or even requirement for a modified tRNA substrate [33, 37–39] and a prerequisite for proper tertiary interactions between the tRNA D and T loops for tRNA binding [33]. Based on these observations we hypothesize that DusA functions in the later stages of tRNA maturation compared to the early-acting TrmA and TruB enzymes. To test this hypothesis, we characterized the molecular mechanisms of cofactor oxidation, tRNA binding, and tRNA modification using purified DusA enzyme. Additionally, we uncover the cellular functions of DusA in tRNA modification, aminoacylation, and protein translation by comparing a dusA deletion to its parental strain and compare the functions of DusA to previously characterized enzymes, TrmA and TruB.

Materials and methods

Buffers and reagents

Experiments were performed in TAKEM₄ buffer [50 mM Tris–HCl, pH 7.5, 70 mM NH₄Cl, 30 mM KCl, 1 mM ethylenediaminetetraacetic acid (EDTA), 4 mM MgCl₂]. Oligonucleotides were purchased from Integrated DNA Technology (IDT). Unless otherwise indicated, all reagents were purchased from Thermo Fisher Scientific.

DusA overexpression and purification

The dusA gene was amplified from E. coli DH5α and inserted into the expression vector pET28a(+) using the restriction enzymes NheI and BamHI to prepare pET28a-DusA wildtype. To prepare plasmids for the expression of DusA C114A and K153A variants, site directed mutagenesis was performed using pET28a-DusA wildtype as a template with overlapping primers specified in Table 1. The coding sequences of pET28a-DusA wildtype and variants were confirmed by Sanger sequencing (Azenta).

Table 1.

Oligonucleotide sequences used in this study

Primer name	Sequence
DusA C114A sense	AATGTCGGCGCCCCGTGTGACCG
DusA C114A antisense	AGACGGGGCGCCGACATTCAGGTTG
DusA K153A sense	GGTGACGGTGGCAACGCGTATTGGCAT
DusA K153A antisense	CGCGTTGCCACCGTCACCGGAATGC
T7 tRNA sense	GCTGCAGTAATACGACTCACTATAG
tRNAPhe antisense	mUmGGTGCCCGGACTCG
T7 tRNAGlyCCC fwd	GCTGCAGTAATACGACTCACTATAGGGCGCGTAGTTCAATGGTAGAACGAGAGCTT CCC
tRNAGlyCCC rvs	mUmGGAGGGCGCGAAGGGAATCGAACCCTCGTATAGAGCTTGGGAAGC TCTCGTTCTACCATTGAACTAC
tRNAGlyUCC fwd	GTTCCAGTAATACGACTCACTATAGGCGGCATCGTATAATGGCTATTACCTCAGCCT
tRNAGlyUCC rvs	mUmGGAGGCGGCAGCGGGAATCGAACCCGCATCAGCTTGGAAGGCTGAGGTA
dusA upstream	GCAAAATGCGAGATGCTGC
dusA downstream	CGCCAGCGGTCATAACGC

Plasmids were transformed into BL21(DE3) cells grown in Lysogeny broth (LB) supplemented with 50 µg/ml kanamycin at 37°C. Protein overexpression was induced at an OD₆₀₀ of ∼0.6 by adding Isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 1 mM, and the temperature was reduced to 18°C. Cells were collected by centrifugation at 5000 × g after 20 h, shock frozen, and stored at −80°C. Proteins were purified using their amino-terminal hexahistidine tag using nickel-sepharose chromatography followed by Superdex 75 size exclusion chromatography, similar to [40, 41]. Cells were resuspended to a final concentration in 0.2 g/ml Cell Opening Buffer (20 mM Tris–HCl, pH 8.1, 400 mM KCl, 1 mM β-mercaptoethanol, 30 mM imidazole, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), 5% (v/v) glycerol) and lysed for 30 min on ice by 1 mg/ml lysozyme. Sodium deoxycholate was added to a final concentration of 12.5 mg/g cells and incubated another 30 min. The solution was sonicated for 4 min total in cycles of 15 s on at 40% output and 45 s off (Fisher Scientific Sonic Dismembrator 500) and centrifuged for 30 min at 30 000 × g. The clarified lysate was loaded onto 3 ml of nickel sepharose resin (Cytiva), washed with 300 ml of Cell Opening Buffer, and DusA was eluted from the resin with Elution Buffer (same as Cell Opening Buffer except containing 500 mM imidazole and no PMSF). Following protein concentration with Amicon-15 Centrifugal Filters (Millipore), DusA was rebuffered and further purified by Superdex 75 chromatography (XK 26/100 column, Cytiva) into Superdex Buffer (20 mM HEPES–KOH, pH 7.5, 150 mM KCl, 1 mM β-mercaptoethanol, 0.5 mM EDTA, 5 mM MgCl₂, 20% (v/v) glycerol). Peak fractions were concentrated, aliquoted, flash frozen, and stored at −80°C. The concentration of purified DusA was determined photometrically at A₂₈₀ using an extinction coefficient of 41 370 M⁻¹ cm⁻¹ (calculated with Expasy ProtParam [42]) and confirmed using comparative sodium dodecyl sulphate–polyacrylamide gel electrophoresis.

Additional tRNA modifying enzymes (TrmA, TruB, IscS, and ThiI) used to prepare partially modified tRNAs were overexpressed and purified similar as above and previously published [20, 36, 40].

tRNA in vitro transcription and modification

The DNA template for tRNA^Phe was prepared by polymerase chain reaction (PCR) of plasmid pCF0 [43] using primers specified in Table 1. The DNA templates for tRNA^Gly_CCC and tRNA^Gly_UCC were obtained by extending two overlapping primers containing T7 promoter and tRNA sequence by PCR (Table 1). Subsequently, in vitro transcription was carried out using the PCR template [10% (v/v)] in transcription buffer [40 mM Tris–HCl, pH 7.5, 15 mM MgCl₂, 2 mM spermidine, 10 mM NaCl, 10 mM dithiothreitol (DTT)] with 3 mM NTPs (ATP, CTP, GTP, and UTP; Sigma), 5 mM GMP, 0.01 U/ml inorganic pyrophosphatase (Sigma), 0.3 µM T7 RNA Polymerase (purified in-house), and 0.12 U/ml RiboLock RNase inhibitor at 37°C for 4 h. DNA template was degraded by addition of 2 U/ml DNase I for 2 h at 37°C. tRNA was purified by phenol/chloroform extraction to remove enzymes followed by Superdex 75 (XK 26/100) chromatography with TAKEM₄ buffer to remove unincorporated NTPs. tRNA concentrations were determined photometrically using an extinction coefficient of 500 000 M⁻¹ cm⁻¹ for tRNA^Phe [44] and 713 400 M⁻¹ cm⁻¹ for tRNA^Gly_CCC (IDT Oligoanalyzer).

To prepare tRNA^Phe and tRNA^Gly containing s⁴U8, m⁵U54, and Ψ55 modifications, purified tRNA (6 µM) was incubated in TAKEM₄ buffer with 1 µM each of TruB, TrmA, IscS, and ThiI enzymes, 50 µM S-adenosylmethionine (New England Biolabs), 0.5 mM L-cysteine, 250 µM ATP, 40 µM pyridoxal 5ʹ-phosphate (Sigma), 1 mM DTT, and 0.01 U/µl RNase Inhibitor at 37°C for 3 h within a total reaction volume of 7 ml. To purify partially modified tRNA, enzymes were first removed by phenol/chloroform extraction and small molecules were separated from tRNA using Superdex 75 (10/300 GL) size exclusion chromatography [45].

DusA NAD(P)H oxidation assays to determine DusA activity and cofactor preference

To determine the NAD(P)H oxidation activity of DusA wildtype and variants, 2 µM DusA with 2 µM FMN was incubated with 500 µM NADH or NADPH in TAKEM₄ buffer in the presence or absence of 50 µM in vitro transcribed tRNA^Phe. Reactions were started by addition of NAD(P)H, and oxidation to NAD(P)⁺ was followed by monitoring the decrease in absorbance at 340 nm (ε_NAD(P)H = 6220 mM⁻¹ cm⁻¹). Since reactions took place in aerobic conditions, oxygen acts as the final electron acceptor. To determine initial velocities (v₀), the NAD(P)⁺ formation was plotted versus time, and the slope of the linear portion of each reaction was determined.

DusA tRNA dihydrouridylation assays

In vitro transcribed tRNA^Phe (5 µM) was incubated with 5 µM DusA wildtype or variant in the presence of 5 mM NADPH and 1 mM FMN in TAKEM₄ buffer containing 1 mM DTT and 0.04 U/µl RiboLock for 1 h at 37°C. Reactions were stopped by phenol extraction and modified tRNA was purified by Superdex 75 size exclusion chromatography as described above.

Assessing DusA activity with liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS)

We conducted liquid chromatography - tandem mass spectrometry (LC-MS/MS) assays to quantitatively assess the ability of our DusA protein to modify tRNA using a previously described method [46]. The levels of all four standard nucleosides (A, U, G, C) and dihydrouridine (D) were measured in in vitro transcribed tRNA^Phe incubated with either no DusA, wild-type DusA, or a DusA mutant (K153A, C114). Each treated tRNA^Phe was hydrolyzed to mononucleotides with 300 U/μg Nuclease P1 (NEB, 100 000 U/ml), 100 mM ammonium acetate, and 100 μM zinc sulfate at 37°C overnight. The resulting mononucleotides were then dephosphorylated with 50 U/μg bacterial alkaline phosphatase (BAP, Invitrogen, 150 U/μl), 100 mM ammonium bicarbonate, and 50 mM zinc sulfate at 37°C for 5 h. The samples were then lyophilized and resuspended in 9 μl of water and combined with 1 μl of 400 nM ¹⁵N₄-inosine internal standard. Samples were separated on a Waters Acquity UPLC HSS T3 column (100 Å, 1.8 μm, 1.0 × 100 mm) heated to 35°C on an Agilent 1290 Infinity II liquid chromatography system coupled to an Agilent Technologies 6460 Triple Quadrupole mass spectrometer and ionized via electrospray ionization. Mobile phase A was 0.01% (v/v) formic acid in water and mobile phase B was 0.01% (v/v) formic acid in acetonitrile. Sample injection volume was 5 μl. Samples were run in positive mode with 4000 kV capillary voltage. The gas temperature was 350°C, the gas flow was 10 l/min, the nebulizer gas pressure was 25 psi, the sheath gas temperature was 350°C, and the sheath gas flow was 11 l/min. Calibration curves were prepared of the four canonical bases and modified nucleosides which were used to quantify sample nucleoside levels. Samples were prepared in triplicate and averaged.

Nitrocellulose filter binding to determine affinity of DusA for tRNA

In vitro transcribed tRNAs were dephosphorylated with calf intestinal phosphatase (New England Biolabs) and radiolabelled using [γ-³²P]ATP (Revvity) and T4 polynucleotide kinase (New England Biolabs). To remove unincorporated nucleotide, radiolabelled tRNA was purified through 200 µl Sephadex G-25 (Cytiva) resin in spin columns. Specific activity was determined by scintillation counting (Revvity Tri-Carb 4910TR).

Prior to the binding reaction, tRNA was refolded in 1× TAKEM₄ buffer by heating to 65°C for 5 min and cooling to room temperature for 10 min. Refolded tRNA (160 nM) was incubated with increasing concentrations of DusA wildtype or variant (0–50 µM) in TAKEM₄ buffer. For indicated reactions, NADPH (1 mM) or FMN (250 µM) were added to the reactions. After incubation for 10 min at room temperature, tRNA-enzyme mixtures were filtered through a nitrocellulose membrane under vacuum, and percent bound tRNA was determined by scintillation counting. Dissociation constants (K_D) were determined by plotting percent tRNA bound (Bound) as a function of enzyme concentration [enzyme] and fit with a hyperbolic equation using GraphPad Prism (version 10.4):

Determination of DusA tRNA binding kinetics using stopped-flow

Fluorescein-labelled s⁴U8 tRNA^Phe was prepared as described previously [47, 48] by first introducing the s⁴U8 modification into in vitro transcribed tRNA^Phe with ThiI and IscS enzymes, removing reaction components from s⁴U8-tRNA^Phe using phenol extraction followed by Superdex 75 (10/300 GL) size exclusion chromatography, and subsequently labelling the tRNA thiol group with 5-iodoacetamidofluorescein (5-IAF; Sigma). Herein, s⁴U8-tRNA^Phe (60 µM) was incubated with 3.2 mM 5-IAF in 12 mM HEPES–KOH, pH 8.2 containing 80% (v/v) dimethyl sulfoxide at 65°C in the dark for 4 h. To remove unincorporated fluorescent dye, successive phenol extractions were performed until the organic layer no longer appeared yellow (at least eight extractions). Trace phenol was removed by two chloroform extractions followed by ethanol precipitation, and the final fluorescent tRNA product was dissolved in water. tRNA concentration and fluorescein labelling efficiency was determined using spectrophotometry at 260 and 492 nm.

For stopped-flow experiments, 1 µM fluorescein-s⁴U8-tRNA^Phe was rapidly mixed with excess DusA wt or variant enzymes at indicated final concentrations (3–20 µM) in a KinTek SF-2004 stopped-flow at 20°C. Fluorescein was excited at 480 nm and emission was monitored from 505 nm onwards. Relative fluorescence (Y) was plotted against time (t) and fit to a two-exponential function to determine apparent rates (k_app) using TableCurve 2D (version 5.01):

Data shown are averages of at least eight independent replicates. For DusA wt, apparent rates (k_app) were plotted against the enzyme concentration [enzyme] and fit with a linear equation to determine the association rate constant, k₁:

Multiplex small RNA sequencing to determine tRNA abundance, charging, and modification

Bulk RNA was extracted using TRIzol (Invitrogen) under acidic conditions (pH < 5) to preserve the tRNA aminoacyl bond from early-log phase E. coli BW25113 wildtype and ΔdusA cells [49] grown in LB medium. Strain identity was confirmed by Sanger sequencing of the dusA locus using primers specified in Table 1. After shock freezing wet cell pellets from 15 ml culture grown to ∼0.4 OD₆₀₀, cells were opened with 3 ml TRIzol and 0.6 ml chloroform was added after a 3-min incubation, according to the manufacturer’s protocol. After centrifugation (5000 × g, 45 min), the aqueous phase was transferred to a new tube and ethanol precipitated several times. Pure RNA was resuspended in 10 mM NaOAc (pH 4.8) and stored at −80°C. Concentrations were determined using ultraviolet spectrometry (NanoDrop 2000c). Three biological replicates from each strain were isolated.

Library preparation, MSR-tRNA-seq, and bioinformatics were performed by MesoRNA (Chicago, IL, USA). Reads were demultiplexed and trimmed as appropriate in accordance with the multiplex small RNA sequencing (MSR-seq) library preparation protocol [50]. Quality control was performed with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were aligned to tRNA sequences adapted from GtRNAdb [51] with bowtie2 [52]. Further analysis and plots were generated with custom software developed by MesoRNA. Read counts are normalized to 5S rRNA.

Tandem codon GFP reporter assays

BW25113 wildtype and ΔdusA strains were transformed with reporter plasmids carrying an arabinose-inducible fluorescent transcriptional cassette encoding superfolder green fluorescent protein (sfGFP) followed by mCherry fluorescent protein (kind gift from Assaf Katz, Universidad de Chile, Santiago, Chile) [53]. Each reporter contains a set of four tandem repeats of a codon following the third sfGFP codon, such that sfGFP expression relies on translation of the four repeated codons. To account for subtle differences in translation between strains, plasmid S1 was used as a control, which lacks any additional codons in sfGFP. sfGFP and mCherry fluorescence were determined similarly as previously described [20, 53]. In brief, overnight cultures were diluted 1:20 to an OD₆₀₀ of ∼0.1 and grown in LB medium supplemented with 100 µg/ml ampicillin until reaching an OD₆₀₀ between 0.4–0.6. Cells were then diluted 1:4 into fresh LB medium containing 100 µg/ml ampicillin and 0.4% (w/v) arabinose in black, optical bottom 96 well plates and grown at 37°C with shaking at 100 rpm. Three-hours post induction, OD₆₀₀, sfGFP fluorescence (excitation: 470–15 nm, emission: 515–20 nm), and mCherry fluorescence (excitation: 570–15 nm, emission: 620–20 nm) were measured using a CLARIOstar Plus plate reader (BMG Labtech). sfGFP/mCherry ratios for each test codon were normalized to the sfGFP/mCherry ratio of plasmid S1 for the respective strain:

For each codon, translation between strains was compared using two-way ANOVA. Significant differences (P <.05) between strains are indicated.

Results

Although dihydrouridine is one of the most frequent tRNA modifications, the enzymes that form dihydrouridine remain understudied in comparison to other tRNA modifying enzymes. Here, we examined the molecular mechanism of E. coli DusA and clarify the roles of this enzyme and D20/D20a modifications in vivo during tRNA maturation and protein synthesis.

DusA can form dihydrouridine in in vitro transcribed tRNA^Phe

Previous work has shown that some tRNA dihydrouridine synthases require the presence of pre-existing tRNA modifications for tRNA modification and/or NAD(P)H reduction activity [33, 37], and previous studies of DusA have utilized native tRNAs rather than unmodified transcripts [12, 13, 18, 34]. In contrast, a recent study has found that purified human Dus2 quickly modifies in vitro transcribed tRNA^Val_CAC, suggesting not all dihydrouridine synthases require a modified substrate [54]. Asking whether E. coli DusA requires a partially modified tRNA substrate for activity, we incubated a high concentration (5 µM) of purified DusA wildtype or variant with a stoichiometric amount of unmodified tRNA for 1 h, purified the resulting DusA-modified tRNA and determined the amount of dihydrouridine formed by mass spectrometry. We found that DusA wildtype can modify in vitro transcribed tRNA^Phe under the conditions tested, with an end level of 5.65% D/total U content (Fig. 1D). DusA has previously shown to be specific to U20 in tRNA^Phe [13, 18], which has 16 U nucleotides in E. coli tRNA^Phe, i.e. 5.65% D/total U accounts for >90% of U20 being converted to D20. Similar to previous reports, substitution of the proposed catalytic cysteine residue to alanine (DusA C114A) resulted in a drastic decrease in dihydrouridylation by DusA (Fig. 1D). Previous work has shown that the DusA K153A variant or its equivalent in other organisms cannot form D20 when expressed in cells [34, 55]. In contrast to these in vivo experiments, in our in vitro assay, DusA K153A was able to form dihydrouridine in in vitro transcribed tRNA, although with a lower end-level than that of the wildtype enzyme, suggesting that dihydrouridine formation is slower for this variant compared to the wildtype enzyme. Based on previous studies of the equivalent variant of TthDus, we hypothesize that catalysis by DusA K153A is slow due to decreased affinity for FMN and tRNA [33]. Indeed, whereas DusA C114A co-purifies with a flavin cofactor similar to the wildtype enzyme, no absorbance at ∼375 nm and ∼450 nm, characteristic of a flavoprotein, was observed for purified DusA K153A, suggesting impaired FMN interaction for this variant (Fig. 1E).

Characterization of NAD(P)H oxidation by DusA

Whereas most characterized prokaryotic and eukaryotic tRNA dihydrouridine synthases oxidize NADPH faster than NADH [15, 16, 38], the preferred FMN reduction substrate for DusA has not yet been reported. To address this, we monitored the oxidation of NAD(P)H by measuring the loss of absorbance at 340 nm as NAD(P)H is oxidized to NAD(P)⁺. Under multiple turnover conditions with 500 µM NAD(P)H, DusA quickly oxidizes NADPH in the absence and presence of tRNA, with initial velocities of 85 ± 3 µM min⁻¹ and 68 ± 4 µM min⁻¹, respectively (Fig. 1F and Table 2). Thus, DusA can oxidize NADPH, with air acting as the final electron acceptor, even in the absence of tRNA. In contrast, the oxidation of NADH is significantly slower, but still above the background (Fig. 1F and Table 2).

Table 2.

Initial velocity (v₀) of NAD(P)H oxidation by DusA wildtype and variants

	NADPH oxidation v0, µM min−1	NADH oxidation v0, µM min−1
DusA wt	85 ± 3	3.8 ± 0.7
DusA wt + tRNAPhe	68 ± 4	ND
DusA C114A	17 ± 0.5	ND
DusA K153A	3.5 ± 0.2	ND

Next, we asked if the DusA C114A and DusA K153A variants can oxidize NADPH although DusA C114A was not active in tRNA modification and DusA K153A purified without the FMN cofactor (Fig. 1G). DusA C114A retained the ability to oxidize NADPH, albeit with a five-fold slower initial velocity compared to DusA wildtype (Table 2). NADPH oxidation by DusA K153A was even slower than that for DusA C114A, but still above background with a 25-fold slower initial velocity compared to the wildtype enzyme, suggesting some residual NADPH reduction activity remains for this variant (Table 2).

Characterization of tRNA binding by DusA

We next investigated the affinity of DusA for two of its known substrate tRNAs: tRNA^Phe and tRNA^Gly_CCC. Previous work with TthDus has suggested this enzyme requires its substrate tRNA to contain modifications for efficient binding and/or catalysis, at least at high temperatures [33, 39], whereas yeast Dus2 requires tRNA to be modified for efficient activity [37]. Human Dus2 does not require tRNA to be modified for binding, and studies are conflicted whether this enzyme requires previous modifications for enzymatic activity [38, 54]. Since all previous studies of E. coli DusA utilized tRNA substrates that were isolated from a knockout strain [13, 18, 34], we sought to clarify whether modifications are important for the binding of DusA to tRNA. For this reason, we also prepared partially modified tRNA^Phe and tRNA^Gly_CCC that contain s⁴U8, m⁵U54, and Ψ55 by modifying in vitro transcribed tRNAs with ThiI, IscS, TrmA, and TruB [45]. For tRNA^Gly_CCC, these constitute all its known native modifications except for the DusA-introduced D20 modification, and for tRNA^Phe these are three of its nine total non-D20 modifications [56]. The affinity of wildtype DusA for unmodified tRNA^Phe was found to be 7.3 ± 0.8 µM, whereas the affinity of DusA for unmodified tRNA^Gly_CCC was about two-fold lower at 14 ± 2 µM (Fig. 2A and B, and Table 3). Presence of modifications within tRNAs slightly but significantly increase the affinity of DusA for tRNA^Phe and tRNA^Gly_CCC with dissociation constants of 5.9 ± 0.3 µM and 9.3 ± 0.7 µM, respectively (Fig. 2A and B, and Table 3).

Figure 2.

Affinity of DusA for tRNA. Nitrocellulose filtration titration for DusA wt binding to unmodified (black circles) or partially modified (s⁴U8, m⁵U54, and Ψ55; teal squares) tRNA^Phe (A) or tRNA^Gly (B). (C) Nitrocellulose filtration for DusA C114A (dark pink squares) and K153A (light pink circles) binding to unmodified tRNA^Phe. For comparison, the binding curve for wildtype DusA shown in panel (A) is repeated here (black circles). N = 3 for each curve and error bars represent standard deviations. Dissociation constants for each data set were determined by hyperbolic fitting and are displayed in Table 3.

Table 3.

Dissociation constants (K_D) for DusA wt and variants binding to tRNA^Phe and tRNA^Gly

	tRNAPheKD, µM	tRNAGlyCCCKD, µM
Unmodified tRNA	7.3 ± 0.8 µM	14 ± 2 µM
Modified tRNA	5.9 ± 0.3 µM	9.3 ± 0.7 µM
DusA C114A	17.6 ± 1.1 µM	ND
DusA K153A	>50 µM	ND

We next examined other aspects of tRNA binding by DusA. Wondering if DusA specifically recognizes only substrate tRNAs, we measured the affinity of DusA for tRNA^Gly_TCC, which has not previously been found to be modified by DusA in vivo [18]. The affinity of this interaction was found to be 14 ± 3 µM, which is similar to that of the affinity for the known substrate tRNA^Gly_CCC (Supplementary Fig. S1 and Supplementary Table S1). We also assessed whether cofactor presence may stabilize tRNA binding, thereby increasing the affinity. DusA co-purifies with FMN, but it is unknown whether this cofactor is present at stoichiometric amounts. Thus, we examined whether DusA binding to tRNA^Phe and tRNA^Gly_CCC increased in the presence of saturating FMN or NADPH concentrations. Surprisingly, addition of either cofactor instead significantly reduced tRNA binding, with addition of NADPH lowering the affinity up to ∼1.5-fold and FMN lowering the affinity >5-fold (Supplementary Fig. S1 and Supplementary Table S1). Corresponding with this increase in K_D, the end level of tRNA binding is lower upon cofactor addition compared to the absence of cofactors.

Finally, we examined whether the DusA variants are deficient in tRNA binding. DusA C114A binds to tRNA^Phe with a K_D around 18 µM (∼2-fold weaker than DusA wildtype) with approximately the same end level of binding as the wildtype enzyme. In contrast, DusA K153A was found to bind tRNA^Phe with a much lower affinity, with a K_D above 50 µM and observed end level approximately half of that for the wildtype enzyme (Fig. 2C and Table 3).

Rapid kinetic dissection of DusA binding to tRNA

In order to monitor the binding of DusA to tRNA^Phe in real time, we utilized a previously established fluorescent tRNA stopped flow assay [47, 48]. Herein, we introduced the s⁴U8 modification within in vitro transcribed tRNA^Phe and subsequently labelled the thiol group with fluorescein (Fig. 3A). Rapid mixing of DusA wildtype (20 µM) with fluorescent tRNA (1 µM) results in a two-phase decrease fluorescence within one second (black trace, Fig. 3B). Fitting this data with a two-exponential function reveals a fast phase with an apparent rate of 146 ± 17 s⁻¹ and a slow phase with an apparent rate 17 ± 2 s⁻¹, respectively (Fig. 3B). Similar to wildtype DusA, rapid mixing of 20 µM DusA C114A and K153A variants with fluorescent tRNA also results in a two-phase fluorescence decrease (Fig. 3B and C); however, the amplitudes of these fluorescence decreases are significantly smaller than that for the wildtype enzyme. Fitting with a two-exponential function yields apparent rates equal to 91 ± 11 s⁻¹ and 15 ± 1 s⁻¹ for DusA C114A and 29 ± 2 s⁻¹ and 1 ± 3 s⁻¹ for DusA K153A (Fig. 3B and C).

Figure 3.

Rapid kinetic dissection of DusA binding to tRNA^Phe. (A) Structure of T. thermophilus Dus (grey) bound to tRNA^Phe (light teal) (PDB 3B0V) [33] showing the locations of the target nucleotide of DusA (U20) as teal spheres and of the fluorescently labelled nucleotide (s⁴U8) as orange spheres. (B) Rapid mixing of 20 µM DusA wt (black curve), DusA C114A (dark pink curve), or K153A (light pink curve) with 1 µM fluorescent tRNA^Phe at 20°C. Curves were fit to two-exponential equations. (C) Zoomed-in view of rapid binding of DusA C114A and K153A variants to fluorescently labelled tRNA^Phe to better visualize the small decrease in fluorescence. (D) Pre-steady state binding of increasing DusA wt concentrations to 1 µM fluorescent tRNA^Phe to determine apparent rates. Plotting k_app against time shows that k_app1 is concentration dependent (E), whereas k_app2 is independent of DusA concentration (F). (G) Kinetic mechanism for tRNA binding to DusA. Following binding of tRNA by DusA, conformational changes take place.

To gain a deeper understanding of the kinetics of wildtype DusA binding to tRNA, the stopped-flow experiments were repeated at different concentrations of DusA (Fig. 3E and F), and apparent rates were plotted against DusA concentration (Fig. 3E and F). We observed that the apparent rate corresponding to the fast phase (k_app1) is dependent on DusA concentration, consistent with this phase representing a bimolecular binding event between DusA and tRNA. Linear fitting reveals a k₁ of 5 ± 1 µM⁻¹ s⁻¹ and an estimated k_-1 equal to 55 ± 18 s⁻¹ (Fig. 3E), which agrees with the determined K_D of ∼7 µM (Fig. 2A and Table 3). The apparent rate corresponding to the slow phase (k_app2) was found to be not dependent upon the DusA concentration, with an average k_app of ∼18 s⁻¹ (Fig. 3F). This second phase therefore reflects a unimolecular event such as a conformation change of tRNA. Taken together, this stopped-flow data suggests that DusA binds tRNA with a two-step mechanism, with the first step encompassing binding and the second step encompassing a conformational change (Fig. 3G).

Impact of DusA on tRNA abundance and charging in vivo

To date, there are no reported growth defects for the E. coli ΔdusA strain, and the conserved biological functions for DusA binding and modifying many tRNAs remains unclear. To address this question, we utilized MSR-seq, to compare the abundance, aminoacylation, and modification status for each tRNA isoacceptor between the wildtype and ΔdusA strain [50]. Herein, bulk RNA isolated in acidic conditions was subjected to periodate oxidation and β-elimination followed by library preparation and Illumina sequencing. Comparison of the abundances of individual tRNA species between the wildtype and ΔdusA strain revealed only one tRNA was significantly upregulated in ΔdusA (tRNA^Leu_CAA, Fig. 4A). Although no other tRNAs were found to be individually significantly different in the knockout strain compared to wildtype, on a global scale, more tRNAs are upregulated upon dusA deletion compared to downregulated (Fig. 4A). Indeed, on average, each tRNA is 1.3-fold more abundant in the ΔdusA strain than in wildtype (Fig. 4B).

Figure 4.

Abundance and aminoacylation levels of certain tRNAs are affected by DusA. (A) tRNA abundance change in ΔdusA compared to the wildtype strain (n = 3 for each strain). Horizontal dashed line indicates P = 0.05. One tRNA with significant changes in the knockout strain was identified and is indicated by a large pink circle, tRNA^Gly isoacceptors are indicated by grey colour (B) Box and whisker plot showing relative abundance level changes (left) and relative charging level changes (right) for all tRNAs in wildtype and ΔdusA strains. (C) Heatmap of relative charging level for each tRNA in all replicates for wildtype and ΔdusA strains. Stars indicate significant (P <.05) differences in tRNA charging between the wildtype and ΔdusA strains.

Next, we examined the fraction of each tRNA charged between the wildtype and ΔdusA strains. In contrast to our previous study with trmA and truB [20], deletion of dusA does not globally affect tRNA aminoacylation in the ΔdusA strain (Fig. 4B). Instead, the charging fraction of only a few tRNAs was changed as we observed a significant decrease for tRNA^Gly_GCC and tRNA^Gly_TCC and a significant increase for tRNA^Tyr_GTA charging within the ΔdusA strain compared to wildtype (Fig. 4C). Intriguingly, one of the two affected tRNA^Gly isoacceptors, tRNA^Gly_UCC, has not been previously identified to be modified by DusA, and similarly tRNA^Tyr_GTA is not known to be a substrate for DusA [18, 56].

To determine whether the translation of specific codons is affected by the altered levels of tRNA abundance or charging of specific tRNAs in the ΔdusA strain, we used a sfGFP-based codon reporter, wherein four tandem repeats of a specific codon are present near the beginning of the sfGFP open reading frame and mCherry is independently translated as a control (Fig. 5A) [53]. First, we examined translation of Leu codons, as the abundance of tRNA^Leu_CAA is significantly higher in the knockout strain compared to wildtype. No significant difference in sfGFP expression for the reporter containing its cognate Leu codon, TTG, was observed between the two strains (Fig. 5B). However, we noticed a decrease in sfGFP expression in ΔdusA for the reporter containing tandem repeats of the Leu codon CTA, read by DusA-modified tRNA^Leu_TAG, which did not display a change in abundance or charging in the knockout strain (Fig. 5B).

Figure 5.

Changes in tRNA abundance and aminoacylation are not correlated with translational changes in the ΔdusA strain. Measurements of codon specific translation using a sfGFP codon reporter library wherein four tandem repeats of a specific codon are present near the beginning of the sfGFP open reading frame. sfGFP and mCherry expression is induced in early log phase. Whereas sfGFP expression is dependent on readthrough of the four repeated codons, mCherry is independently expressed and is used for normalization. SD: Shine Dalgarno sequence, figure adapted from [53] (A). sfGFP/mCherry expression ratios for all Leu (B), Tyr (C), and Gly (D) codons. For each codon, the data for E. coli wildtype (grey, left bar) are compared to data for the ΔdusA strain (pink, right bar). The cognate tRNA(s) are indicated underneath each codon, coloured in pink if known to be modified by DusA and black if not known to contain D20/D20a [18]. For each codon, at least eight biological replicates were measured. * Indicates P <.05, ** indicates P <.01, and *** indicates P <.001.

Additionally, we examined translation of the four Gly codons, as tRNA charging was found to be reduced in the ΔdusA strain for two of the three tRNA^Gly isoacceptors. Indeed, the translation of the Gly codon GGC is significantly decreased in ΔdusA compared to wildtype (Fig. 5C). This codon is read by tRNA^Gly_GCC, which is one the isoacceptors with reduced aminoacylation in the deletion strain, and which is known to be modified by DusA [18]. This tRNA additionally reads Gly GGT as a cognate codon; however, the translation of Gly GGT is not altered in ΔdusA. Although charging tRNA^Gly_TCC was also found to be decreased in the knockout strain, the translation of both of its cognate codons is not affected by loss of dusA (Fig. 5C). Likewise, the translation of both Tyr codons is not affected by dusA deletion, although the charging of the cognate tRNA is slightly but significantly increased in the knockout strain (Fig. 5D).

From the MSR-seq experiment, we were additionally able to determine relative changes in certain modification levels between the two strains for modifications that leave a reverse transcriptase signature [50]. For example, acp³U47 is located nearby to D20 in the tRNA elbow (Fig. 6A and B) and this modification results in deletions and mutations during reverse transcription that are read during sequencing (Fig. 6C and D). We examined the sum of the relative fraction of tRNA reads that contain deletion and mutations at position 47 for the seven tRNAs known to contain acp³U47 [57, 58] (Fig. 6E). Of these tRNAs, six are known to be also modified by DusA [18]. For each of these tRNAs except tRNA^Val_GAC, the sum of deletions and mutations at position 47 was lower for tRNAs extracted from the ΔdusA strain compared to wildtype, with tRNA^Met_CAT, tRNA^Ile_CAT, and tRNA^Arg_ACG showing statistically significant reductions (Fig. 6E). This observation indicates that the absence of D20 also reduces the presence of acp³U47 in several tRNAs.

Figure 6.

Changes in acp³U47 content in ΔdusA E. coli. (A) Structure of unmodified E. coli tRNA^Phe (PDB 3L0U) [59] showing relative locations for D20 (teal spheres) and acp³U47 (burgundy spheres). (B) Chemical structure of acp³U with the modified portion of the nucleobase highlight in pink. Relative mutation fraction (C) and deletion fraction (D) comparing wildtype (grey) and ΔdusA (pink) at each position in tRNA^Arg_ACG. (E) Heatmap showing the sum for mutation and deletion fractions relative to the wildtype strain at position 47 for all E. coli tRNAs known to contain acp³U47. Pink stars indicate significant differences in mutation and deletion sum at position 47 between the wildtype and the ΔdusA strain (P <.05).

Moreover, we examined changes in the abundance of I34 in tRNA^Arg_ACG. This tRNA editing event causes reading of position 34 as a G rather than an A, resulting in an apparent mutation during tRNA mapping (Fig. 6C). No change in abundance for this essential tRNA anticodon modification was observed between the wildtype and ΔdusA strains, with nearly 100% of this position read as a G in both strains (Fig. 6C). We additionally examined for changes in mutations and deletions at position 37 but did not identify any potential changes in tRNA modifications at these sites (Supplementary Fig. S2).

DusA affects the translation of certain codons

Intriguingly, when we examined specific translation for codons that had cognate tRNAs with altered abundance or aminoacylation in ΔdusA compared to wildtype, we found that not all of these codons display significant changes in translation, whereas certain codons read by unaffected tRNAs show differences in translation (Fig. 5). Together with the fact that levels of at least acp³U47 are changed in other tRNAs (Fig. 6), we wondered if the presence of the dusA gene may affect overall translation or the specific translation of additional codons. Thus, we examined codon specific translation using the sfGFP reporter as above with several additional codons (Fig. 7).

Figure 7.

Translation of several codons read by DusA-modified tRNAs are decreased in the ΔdusA strain. Relative sfGFP/mCherry expression ratios for indicated codons. For each codon, the data for E. coli wildtype (grey, left bar) are compared to data for the ΔdusA strain (pink, right bar). The cognate tRNA(s) are indicated underneath each codon, coloured in pink if known to be modified by DusA and black if not known to contain D20/D20a [18]. For each codon, at least eight biological replicates were measured. * Indicates P <.05, ** indicates P <.01, and *** indicates P <.001.

To assess whether there is a difference in global translation between the wildtype and dusA knockout strain, we compared sfGFP expression between the two strains in the absence of additional codons as a control. No significant difference in sfGFP translation was observed (Fig. 7). As previously reported, introduction of four identical codons often results in increased translation in E. coli wild-type compared to the control sfGFP [20, 53, 60]. However, we did observe several significant codon-specific decreases in translation in the ΔdusA strain compared to the wildtype strain for several codons including Phe UUU and UUC (tRNA^Phe_GAA), Asp GAC and GAT (tRNA^Asp_GTC), Met ATG (tRNA^eMet_CAT), Ser TCT and TCC (tRNA^Ser_GGA), Ile ATT, ATC, and ATA (tRNA^Ile_GAT and/or tRNA^Ile_GAT), Gln CAA and CAG (tRNA^Gln_TTG and/or tRNA^Gln_CTG), and Thr ACA and ACG (tRNA^Thr_TGT, and/or tRNA^Thr_CGT) codons (Fig. 7). Notably, for these tRNAs, the translation of each of their cognate codons is decreased when dusA is deleted. For other tRNAs, including tRNA^Val_GAC, tRNA^His_GTG, tRNA^Cys_GCA, tRNA^Arg_ACG, and tRNA^Thr_GGT, translation of only one of the two or three cognate codons is affected by the loss of dusA (Fig. 7). In other cases, none of the cognate codons read by tRNAs known to be modified by DusA are affected by dusA knockout, such as for tRNA^Asn_GTT, tRNA^Trp_CCA, tRNA^Ser_GTC, tRNA^Ser_TGA, tRNA^Ser_CGA, tRNA^Arg_TCT, and all tRNA^Pro isoacceptors (Fig. 7). Importantly, all these listed tRNAs are known to be modified by DusA [18]. We additionally tested several codons read by tRNAs that are not modified by DusA including tRNA^Val_TAC, tRNA^Arg_CCT, tRNA^Arg_CCG, and tRNA^Tyr_GTA. As expected, knockout of dusA does not affect the translation of any of these cognate codons (Figs 5 and 7). Taken together, our results suggest that DusA positively impacts the translation of many, but not all, cognate codons for DusA-modified tRNAs.

Discussion

Herein, we uncover novel insights into the mechanisms and cellular functions of E. coli DusA. These results illuminate key differences between DusA and two other, well characterized highly conserved tRNA modifying enzymes, TrmA and TruB. In summary, our experiments demonstrate for the first time that E. coli DusA can bind and modify unmodified tRNA, though it binds more tightly to tRNA that already contains modifications. tRNA binding by DusA occurs via a two-step mechanism, with the second step constituting a conformational rearrangement in tRNA. Finally, we find that DusA only subtly affects cellular tRNA abundance and charging but increases translation of at least 21 codons. We contextualize our findings for DusA with the characteristics of TrmA and TruB, as three of the most well-characterized E. coli tRNA modifying enzymes.

Molecular determinants of DusA substrate binding and activity

To clarify the molecular determinants of DusA, we examined the roles of two residues, C114 and K153, previously found to be necessary for overall dihydrouridine formation [33, 34]. Despite being strongly impaired in tRNA modification, DusA C114A remains able to oxidize NADPH and bind tRNA albeit with reduced activity and affinity, respectively. In contrast, we found that DusA K153A can still generate dihydrouridine in tRNA in vitro even though it is drastically impaired in NADPH oxidation and tRNA binding and does not co-purify with FMN in agreement with its low affinity for the flavin cofactor [33]. However, in the cellular context, this variant is unable to form detectable amounts of dihydrouridine [34, 55]. Taken together, these findings validate distinct roles for C114 and K153, with C114 necessary for catalysis, likely acting as a general acid as previously proposed, whereas K153 is required for FMN binding and thereby stable tRNA binding [33].

Additionally, for the first time we clarify that NADPH is preferred source of reducing equivalent for FMN regeneration by DusA (Fig. 1F). This finding is in line with experiments with several dihydrouridine synthases, with only one enzyme identified so far to prefer NADH (Bacillus subtilis DusB2) [15, 16, 38]. Unlike for tRNA methyltransferases such as TrmA and TrmB [36, 48], we find that pre-incubating DusA with its cofactors before tRNA binding does not increase, but rather decreases the affinity of DusA for tRNA (Supplementary Fig. S1A and B, and Supplementary Table S1). We speculate this may be due to product (D20) formation during the incubation period prior to filtration, potential competition between tRNA and NADPH, or due to a redox state preference of DusA for tRNA binding. Additionally, we note that we used cofactor concentrations above the enzyme concentration of our assay, which are above the cellular concentrations of NADPH and free flavin [61, 62] and thus do not represent the physiological state.

tRNA modifications contribute to DusA binding and activity during the intermediate stages of tRNA maturation

Although all previous studies of E. coli DusA have utilized native tRNAs containing all modifications except for D20 [12, 13, 18, 34], we demonstrate here that DusA can bind and modify in vitro transcribed tRNA^Phe (Figs 1D and 2), similar to human Dus2 [54]. Notably, we find that DusA consistently binds tRNA^Phe more than 1.5-fold tighter that tRNA^Gly_CCC regardless of its modification status. This suggests that DusA engages different tRNA species with distinct affinities, perhaps reflecting differences in primary sequence or intrinsic structural stability of the elbow region, which DusA homologs are known to recognize [33, 63]. We found that tRNA^Gly_UCC binds DusA with a similar affinity as tRNA^Gly_CCC (Supplementary Fig. S1C and Supplementary Table S1). As two dihydrouridine mapping studies have identified D20 in tRNA^Gly_CCC but not in tRNA^Gly_TCC [18, 30], our results demonstrate DusA binds this nonsubstrate tRNA with a similar affinity as a substrate tRNA in vitro and may be capable of binding other nonsubstrate tRNAs. Further studies will be required to determine whether this interaction exists in vivo, to identify the exact tRNA sequence and/or structural determinants for DusA binding and investigate the potential biological role of nonsubstrate tRNA binding to DusA. For both substrate tRNAs examined, we find that presence of s⁴U8, m⁵U54, and Ψ55 modestly but significantly enhance the affinity of DusA. The binding preference of DusA for a modified tRNA substrate contrasts TrmB, which is not sensitive to tRNA modification status, and TruB, which prefers binding unmodified tRNA over modified forms [41]. TrmA behaves similarly to DusA in terms of binding, preferring to bind tRNA with Ψ55 over unmodified tRNA; however TrmA displays slower steady-state methylation kinetics for Ψ55 tRNA [41]. Because we measured the end-level of dihydrouridine formation in unmodified tRNA, it remains to be determined whether the presence of previous modifications affects how fast DusA modifies tRNA.

Based on our observations and previous studies of DusA homologs, we speculate that E. coli DusA will modify tRNA that already contains modifications faster than unmodified tRNA, and DusA may exhibit an even stronger preference for modifying an already modified tRNA substrate than it does for tRNA binding. Thermus thermophilus and Saccharomyces cerevisiae U20-dihydrouridylating enzymes have shown strong preferences for modifications in their tRNA substrate for efficient binding and/or modification. Similarly, modifications have been suggested to be a requirement for TthDusA•tRNA complex formation [33] and/or increase reaction velocities at high temperatures [39]. Whereas purified yeast Dus2 has dihydrouridine activity on in vitro transcribed precursor (pre)-tRNA^Tyr and pre-tRNA^Leu [64], the affinity of yeast Dus2 for native mature tRNA^Leu lacking D20 was stronger than that for the pre-tRNA transcript, and the oxidation of yeast Dus2 is >600-fold faster in the presence of native, modified tRNA^Leu compared to an in vitro transcript [37].

These findings suggest D20 is added to tRNA in the later stages of tRNA maturation. However, D20 is formed prior to intron removal in Xenopus oocytes and yeast indicating further modifications are added following D20 onto spliced tRNA [64–66]. Further to this, we observe a lower proportion of tRNAs contain acp³U47 in E. coli lacking dusA, implying TapT may modify U47 at an even later stage of tRNA maturation. In conclusion, the introduction of D20 seems to occur in the intermediate stages during tRNA maturation across bacteria and eukaryotes.

DusA binds folded tRNA using a two-step mechanism involving a local conformational change in the D arm

Our kinetic experiments reveal that DusA binds tRNA with a two-step mechanism including a tRNA conformational change. Upon TthDus binding to Tth tRNA^Phe, significant conformational changes are observed for U16, U17, and the target base U20, but G18 and G19 remain stably base paired with U55 and C56 in the T arm, respectively, and there are no significant tRNA conformational changes outside of the D loop [33]. Moreover, tRNA variants with disrupted G18-U55 and/or G19-C56 interactions cannot be modified by yeast Dus2 [63]. Therefore, we speculate that the conformational change we observe corresponds to these local conformational rearrangements within the D loop changing the environment of the fluorophore attached at U8.

This local conformational change within the D loop upon DusA binding contrasts the major rearrangements within the tRNA elbow that must take place for other tRNA elbow modifying enzymes, including TrmA and TruB to bind their respective target bases, U54 and U55. Upon binding TrmA, U54 flips out of the T loop leading to a base stack containing G53, A58, G57, C56, and U55 and disruption of the G18–U55 and G19–C56 base pairs [67]. Similarly, U55, C56, and G57 flip out of the T loop upon binding TruB, again disrupting tertiary base pairing between these nucleobases and G18 and G19 in the D arm [68, 69]. Accordingly, partially unfolded tRNA substrates with disrupted G18–U55 and/or G19–C56 interactions are modified by TrmA with a higher catalytic efficiency [70], and display faster T arm base flipping kinetics upon interaction with TruB [35]. Taken together, these structural and biochemical experiments reveal a fundamental mechanistic difference between DusA and TrmA/TruB: whereas TrmA and TruB enzymes must break tertiary interactions in the tRNA elbow to access their substrate, DusA instead requires a stable elbow structure in order to modify tRNAs.

DusA’s biological role for translation in comparison to TruB and TrmA

Upon deletion of dusA, codon reporter data reveal a decrease in cellular translation for at least 21 codons decoded by 17 tRNAs (Figs 5 and 7). However, there is no general correlation between the small number of changes in tRNA abundance or charging with translation of specific codons by these tRNAs in our assay. Except for tRNA^Gly isoaccpetors, almost every other tRNA has a higher abundance in the dusA knockout as determined relative to 5S rRNA, which indicates that on a global scale, tRNA abundance is increased (Fig. 4). The biological relevance for these increased tRNA levels may reflect compensation for hypomodified tRNAs or changes in cellular redox homeostasis, but these tRNA level changes do not seem to affect translation.

As discussed above, several D20 synthases like DusA show a conserved preference for acting in the intermediate stages of tRNA maturation acting on folded tRNA whereas the tRNA chaperones TrmA and TruB prefer to act in the early stages where they unfold tRNA and provide a second chance at correct tRNA folding. These differences in molecular mechanisms between DusA and TrmA/TruB result in distinct cellular functions for these enzymes. Previously, we have shown that tRNA chaperones TrmA and TruB enhance aminoacylation globally across all tRNA species in E. coli while not affecting steady-state tRNA abundances [20]. In contrast, we discover here that deletion of dusA does not affect global aminoacylation, and the charging levels of only two tRNA^Gly isoacceptors are significantly reduced. Unlike the deletion of trmA and/or truB, which causes both decreases and increases in codon-specific translation, deletion of dusA consistently negatively impacts the translation of several specific codons or does not impact the translation of a given codon at all. A recent study has revealed the dihydrouridine level of tRNA does not affect tRNA recruitment to polysomes in E. coli [18], suggesting tRNA binding by EF-Tu is likely not impaired in the absence of dusA. Thus, dihydrouridylation of tRNAs by DusA likely directly impact the function of specific tRNAs on the ribosome during mRNA translation for specific codons. Related to a proposed function of m⁵U54 formed by TrmA [24], we speculate that D20 affects the flexibility of tRNA required to move through the ribosome during translocation from the A to the P site. Interestingly, we observe that certain tRNAs are impaired in reading one, but not all its cognate codons, which is prominent for tRNA^Leu_TAG, tRNA^Gly_GCC, tRNA^Arg_ACG, but also observed for tRNA^His_GTG, tRNA^Val_GAC, tRNA^Thr_GGT. In all these cases, only the codon that depends on Watson-Crick rather than wobble decoding (G-U / U-G or I-U and I-A for tRNA^Arg_ACG) at the third codon position is impaired (Figs 5 and 7). Given this trend, we posit that codon-specific effects of dusA deletion could be related to the codon-anticodon interaction, where removing D20 results in similar reading of all codons by the same tRNA in contrast to the enhanced reading of a particular codon observed in E. coli wild-type for Leu CTA, Gly GGC, and Arg CGC. However, further mechanistic studies are required to understand this observation fully as other tRNAs lacking D20 show decreased reading of both or none of their cognate codons irrespective of the third codon position. Further supporting a role for U20 modification for translation, addition of purified human Dus2 to a rabbit reticulocyte in vitro translation system increased synthesis of certain proteins [26]. Alternatively, modification of U20 by DusA may affect a later acting tRNA modification enzyme such as TapT generating acp³U47, whose resulting modification may instead or additionally affect the function of tRNA on the ribosome. Future studies are required to precisely define how D20 affects function of individual tRNAs in translation.

Potential additional biological functions of DusA

Interestingly, some of our data indicate that DusA may have additional biological roles beyond directly impacting translation. We show that DusA can bind tRNA^Gly_TCC and upon dusA deletion, the aminoacylation of this tRNA^Gly_TCC is decreased, although this tRNA has not been previously shown to contain U20/U20a [18]. These observations may suggest a more specific and perhaps indirect role for DusA in modulating the aminoacylation and possibly abundance of tRNA^Gly species without broadly affecting the charging or cellular stability of other tRNAs. Intriguingly, previous work has shown that upon oxidative stress, the cellular aminoacylation level is significantly reduced for all three Gly isoacceptors [60]. Since DusA utilizes reducing equivalents, deletion of dusA may affect cellular redox homeostasis and/or play a role underlaying the loss of active tRNA^Gly in oxidative stress which could be investigated in future studies. Surprisingly, we also observe enhanced aminoacylation of tRNA^Tyr_GTA, which is not modified by DusA, which could possibly be mediated by phenylalanine hydroxylase converting phenylalanine to tyrosine and linked to cellular redox homeostasis [71]. We also note that cells display a noticeably yellow phenotype when DusA is overexpressed in E. coli (Supplementary Fig. S3), suggesting that FMN production might be upregulated by so far unknown mechanisms. Supporting the hypothesis that DusA plays a role in cellular redox homeostasis, recent work in V. cholerae found that DusB is essential for survival in oxidative stress through its NADPH oxidase activity, rather than tRNA modifying activity [72], and it will be very interesting to conduct similar studies for DusA in the future.

Conclusion

In conclusion, our results suggest that DusA acts nonredundantly with TrmA and TruB to fine-tune tRNA function during mRNA translation (Fig. 8). Whereas TrmA and TruB disrupt interactions between the D and T arms to modify all tRNAs thereby providing all tRNAs a second chance at properly folding, DusA instead requires the tRNA elbow to be already folded properly prior to dihydrouridylation of specific tRNAs. As tRNA modifications are known to stabilize the tRNA structure [5] and TthDus does not form any specific interactions with tRNA bases that undergo modification [33], DusA likely prefers to bind an already modified tRNA substrate due to the reinforced tertiary structure. Because TrmA and TruB disrupt the tRNA tertiary structure, these enzymes provide misfolded tRNAs second chances at properly folding [35, 36]. In contrast, DusA is unlikely to be a tRNA chaperone as it requires folded tRNA, which retains its tertiary structure when interacting with DusA. Accordingly, TrmA and TruB are known to act early in tRNA maturation [41, 73, 74], whereas DusA most likely acts in the intermediate or late stages. In line with the role of TrmA and TruB as tRNA chaperones, these enzymes increase aminoacylation of all tRNAs [20], while DusA does not affect tRNA charging on a global scale. Like for TrmA and TruB, many changes in codon specific translation are observed in the absence of DusA in E. coli; however, several codon-specific changes for TrmA and TruB may be accounted for by aminoacylation changes, whereas most codon-specific alterations in the dusA knockout strain are likely due to differences in tRNA activity on the ribosome. Since DusC also does not disrupt tertiary interactions in the tRNA elbow [75], we speculate that DusC likewise may not function as a tRNA chaperone and instead acts similarly to DusA in tRNA maturation. In addition to its functions for protein synthesis, DusA may also fulfill additional cellular roles, e.g. linked to redox homeostasis. In conclusion, sophisticated in vitro studies combined with global in vivo analyses reveal the diverse and complementary functions of different tRNA modification enzymes that collectively fine-tune tRNAs for optimal protein synthesis.

Figure 8.

Comparison of DusA with T arm modifying enzymes TrmA and TruB. Although all three enzymes modify the tRNA elbow region, DusA differs in its molecular mechanism and biological function.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the Natural Sciences and Engineering Research Council of Canada [U.K.: Discovery Grant RGPIN-2020-04965 and Discovery Accelerator Supplement RGPAS-2020-00010] and National Science Foundation [NSF CAREER 2045562 to K.S.K.]. S.K.S. is supported by the RNA Innovation NSERC CREATE program. Funding to pay the Open Access publication charges for this article was provided by Natural Sciences and Engineering Research Council of Canada Discovery [RGPIN-2020-04965].

Data availability

MSR-seq data is available at the NCBI GEO database under the accession GSE304803.