Functional redundancy in tRNA dihydrouridylation

Abstract Dihydrouridine (D) is a common modified base found predominantly in transfer RNA (tRNA). Despite its prevalence, the mechanisms underlying dihydrouridine biosynthesis, particularly in prokaryotes, have remained elusive. Here, we conducted a comprehensive investigation into D biosynthesis in Bacillus subtilis through a combination of genetic, biochemical, and epitranscriptomic approaches. Our findings reveal that B. subtilis relies on two FMN-dependent Dus-like flavoprotein homologs, namely DusB1 and DusB2, to introduce all D residues into its tRNAs. Notably, DusB1 exhibits multisite enzyme activity, enabling D formation at positions 17, 20, 20a and 47, while DusB2 specifically catalyzes D biosynthesis at positions 20 and 20a, showcasing a functional redundancy among modification enzymes. Extensive tRNA-wide D-mapping demonstrates that this functional redundancy impacts the majority of tRNAs, with DusB2 displaying a higher dihydrouridylation efficiency compared to DusB1. Interestingly, we found that BsDusB2 can function like a BsDusB1 when overexpressed in vivo and under increasing enzyme concentration in vitro. Furthermore, we establish the importance of the D modification for B. subtilis growth at suboptimal temperatures. Our study expands the understanding of D modifications in prokaryotes, highlighting the significance of functional redundancy in this process and its impact on bacterial growth and adaptation.


Introduction
All RNA transcripts undergo a series of post-transcriptional processes tailored to optimize their functionality (1)(2)(3).These processes include the addition of various chemical groups appended to the base and / or ribose moieties at conserved positions within the RNA polymer, and catalyzed de novo by specific enzymes ( 4 ).Over 170 chemical modifications have been documented thus far, with ongoing advancements in transcriptome analysis, particularly through high-throughput sequencing technologies combined with chemical labeling and mass spectrometry, continually unveiling novel modifications ( 5 ).At the forefront of the most extensively modified RNA species lie tRNAs, small non-coding RNA molecules involved in decoding genetic information during translation ( 6 , 7 ).tR -NAs undergo complex modifications predominantly clustered at positions 34 and 37 within the anticodon loop.These modifications are not only acknowledged for their indispensable role in ensuring the accuracy and efficiency of translation processes ( 8 ), but are also emerging as vital regulatory elements ( 9 ,10 ).Equally significant, the chemical modifications located outside the anticodon and scattered throughout the polymer stabilize the peculiar and essential L-shaped tRNA structure formed by the kissing dihydrouridine (D) and ribothymidine (rT = m 5 U at position 54) loops (11)(12)(13), both of which represent conserved modified bases.
Unlike all other modified bases, D is a non-aromatic base that cannot participate in stacking interactions or engage in base pairing via hydrogen bonding.Nevertheless, dihydrouridine fulfills a distinctive role by promoting the flexible C2'endo conformation of the ribose ( 14 ).The exact function of D in RNA remains somewhat elusive, although several assumptions have been proposed.It is widely accepted that because the D base is not aromatic and thus disinclined to stacking interactions, it confers a certain degree of flexibility around its position, thereby allowing favorable tertiary interactions in the tRNA elbow region ( 14 ,15 ).This notion of flexibility finds support in studies showing that psychrophilic organisms, thriving in low-temperature environments, generally exhibit higher D content than thermophilic counterparts ( 16 ).In addition, higher D content may confer a growth advantage to cancer cells over healthy cells, perhaps by enhancing translational efficiency ( 17 ).However, the exact mechanisms underlying such effect remain unclear and require further exploration.
DusB emerged as the first Dus from the ancestral Dus, giving rise subsequently to DusA and DusC through duplication events ( 34 ).Our recent phylogenetic analysis revealed that Gram-positive bacteria exclusively carry DusB homologs, categorized into three subgroups: DusB1, DusB2 and DusB3 ( 33 ).While most of the examined genomes carry either a dusB1 or dusB2 gene, approximately 40% of these organisms contain both dusB1 and dusB2 genes, with dusB3 being restricted to a subset of Clostridia.Bacillus species generally retained both dusB1 and dusB2 (BSU00810 and BSU08030 annotated as dusB and dusC , respectively), while Mollicutes conserved only dusB1 and Staphylococcus species kept dusB2 .Both DusB subgroups likely originated from an ancestral DusB duplication event, which probably occurred in the common ancestor of the Firmicutes.In addition, the limited distribution of DusB3 suggests more recent origin.Biochemical characterization of DusB1 from M. capricolum (MCAP_0837) revealed its multisite specificity, catalyzing dihydrouridylation at U17, U20 and U20a positions, (Figure 1 ), consistent with sequenced tRNAs from this Mollicute species ( 33 ).The multi-site specificity feature of Gram-positive Dus, likely shared by both DusB1 and DusB2, is also supported by the tRNA modification profiles of three other bacteria: Lactococcus lactis , Streptomyces griseus and S. aureus , all displaying D17, D20 and D20a modifications.
All cases studied show that a given D residue is specifically synthesized by a single Dus.However, while several Dus have been reported to synthesize D at different positions, the redundancy of synthesis in terms of overlapping specificities has not yet been documented.In this investigation, we explore the contribution of Dus homologs, specifically DusB1 and DusB2, in D biosynthesis.Using B. subtilis as our model organism, we reveal a significant level of functional redundancy in D biosynthetic pathways catalyzed by both DusB homologs.

Deletion of dusB1 and dusB2 of B. subtilis and complementation
The B. subtilis strains used in this study were derived from strain W168, obtained from Chastanet's lab (INRAE, Jouy en Josas, France), and listed in Supplementary Table S1 .All the primers used for mutant strain and plasmid constructions in this study are listed in Supplementary Table S2 .Mutant strains were obtained from Bacillus Genetic Stock Center.Double mutant dusB1::kan , dusB2::erm strain was generated by transforming single dusB1::kan with the PCR product amplified from the single dusB2::erm genome using BSU08030-5pL / BSU08030-3pR ( 35 ).B. subtilis strains expressing SadusB2 (SACOL0067) under the control of Bs-dusB1 promoter ( dusB1::SadusB2-kan , dusB2::erm ) was obtained by transforming dusB2::erm strain with a PCR fragment containing (i) the 5 BsdusB1 genomic sequence, (ii) SadusB2 CDS, (iii) a kanamycin resistance cassette and (iv) the 3 BsdusB1 genomic sequence.The same strategy was used to express McdusB1 (MCAP_0837).All Bacillus transformations were performed following the protocol described by Koo et al. ( 35 ).Strain selections were done on LB-agar containing kanamycin (40 μg ml −1 ) and / or erythromycin (5 μg ml −1 ).All strains were verified by PCR and sequencing.E. coli strains and growth conditions are detailed in previous studies ( 31 ,33 ).
Cloning dus B1 and dus B2 from B. subtilis , dus B1 from M. capricolum and dus B2 from S. aureus Plasmids containing dusB1 and dusB2 genes of B. subtilis (pEX-BsdusB 1 and pEX-BsdusB2 ) and dusB2 of S. aureus (pEX-SadusB2 ) were obtained from Eurofins.We used these plasmids to amplify by PCR dusB gene sequences using the primer pairs listed in Supplementary Table S2 .The dusB1 and dusB2 genes of B. subtilis were cloned as follow into pET15b with a sequence encoding for a 6-histidine tag and a thrombin protease site placed at the 5 end of the genes.After amplification, PCR fragments purified with QIAquick PCR purification kit (Qiagen) were cloned into PCR-linearized pET15b plasmid using the SLIC cloning method ( 36 ).Similarly, dusB1 and dusB2 genes of B. subtilis were cloned in pDG148 for overexpression in B. subtilis strains ( 37 ).In the case of dusB2 from S. aureus ( SadusB2 ), the gene was cloned into the pET28a plasmid containing a sequence encoding for a 6-histidine tag placed at 5 end of SadusB2 gene using the same strategy as described above.After cloning, dusB gene integrity was verified by DNA sequencing (Eurofins).

Activity assay and dihydrouridine quantification
In vitro activity was assayed for 1 h at 37 • C in 50 mM HEPES pH 7.5, 150 mM NaCl, 5 mM DTT, 10 mM MgCl 2 , 100 μM FMN and 15% glycerol under air.Bulk tRNAs (25 μM) issued from the dusB1::kan, dusB2::erm strain were incubated with various concentration of protein ranging from 0.05 to 50 μM in a total volume of 100 μl and reaction was started upon addition of NADPH at a final concentration of 2 mM.Quenching was performed by adding 100 μl of acidic phenol (Sigma-Aldrich) followed by centrifugation at 16 000 ×g for 10 min.tRNAs in the aqueous phase were ethanol precipitated and further purified using a MicroSpin G-25 column (GE-healthcare).Dihydrouridine quantification was carried out by LC-MS spectrometry analysis.

MALDI-TOF spectrometry analysis
For mass spectrometry analysis, about 50 μg of tRNAs were digested with either 10 μg of RNAse A (Euromedex) or RNAseT1 (Sigma-Aldrich), which generates 3 -phosphate nucleosides, in a final volume of 10 μl at 37 • C for 4 h.One microliter of digest was mixed with 9 μl HPA (40 mg / ml in water: acetonitrile 50:50) and 1 μl of the mixture was spotted on the MALDI plate and air-dried ('dried droplet' method) as previously described ( 31 ).MALDI-TOF MS analyses were performed directly on the digestion products using an Ultra-fleXtreme spectrometer (Bruker Daltonique, France).Acquisitions were performed in positive ion mode.An identical strategy was applied for RNase T1 digests (cleavage after G generating 3 -phosphate nucleosides).

Results
Contribution of Bs DusB1 and Bs DusB2 to tRNA dihydrouridylation in B. subtilis B. subtilis , complete modification profiles have been established for 24 tRNA sequences over a total of 35 different isoacceptors, allowing us to compile a more or less accurate distribution of the D sites present in this organism ( 4 ).The predominant positions where D is found include the canonical positions 17, 20 and 20a, along with position 47 for a single tRNA, tRNA Met CAU .A quick survey shows that residues D20 and D20a are the most frequent D residues, followed by D17 and D47 ( Supplementary Table S3 ).Notably, D20 stands out as the most prevalent across all tRNA sequences from all organisms ( 4 ).While the abundance of each specific tRNA still needs to be determined, it is reasonable to assume that D20 and D20a account for most of the D content in B. subtilis tRNAs.
The D content was determined using liquid chromatography-mass spectrometry (LC-MS) in tRNAs extracted from wild type B. subtilis W168 strain or from the isogenic single mutants ( dusB1::kan and dusB2::erm ) or double mutant (Figure 2 A).The double deletion led to a complete depletion of D content in bulk B. subtilis tRNA, indicating that one or both DusB enzymes cover all D biosynthesis in tRNA.However, intriguingly, in the dusB1::kan strain, D content decreased by 34%, while in the dusB2::erm strain, it only decreased by 18%.In other words, in the dusB1::kan strain, Bs DusB2 was responsible for 66% of the D content, whereas Bs DusB1 synthesized 82% of the D content in the dusB2::erm strain (Figure 2 B).These seemingly contradictory results may in fact be explained by an overlapping specificity shared by the two DusB enzymes.Complementation assays in the B. subtilis dusB1 ::kan, dusB2 ::erm strain showed that the expression of McdusB1 and SadusB2 from the dusB1 promoter restored 76% and 22% of the D content of wild type tRNAs, respectively (Figure 2 B).This indicates that both genes encode for Dus enzymes, and that the M. capricolum enzyme is more active than the S. aureus enzyme in the B. subtilis heterologous system.

Functional redundancy of the DusB enzymes in B. subtilis determined by MALDI-MS
The in vivo specificity of Bs DusB1 and Bs DusB2 dihydrouridylation sites was determined by comparing the D content in the tRNAs of four B. subtilis strains, including the W168 (wild type) and the single or double deletion strains.The approach involved a three-step workflow: (i) purification of specific tRNA types from various B. subtilis cells, (ii) fragmentation of the tRNA using RNAseA or RNAseT1 and (iii) analysis of the resulting fragments by MALDI-TOF.Deletion of dusB genes was expected to generate fragments containing U residues at the positions targeted by the corresponding enzymes, resulting in a −2Da shift relative to fragments in tR-NAs extracted from wild type cells.We selected three tRNAs to cover all D sites, namely, tRNA Phe  GAA for D17 and D20, tRNA  approach.Analysis of the D17 modification was made possible by monitoring the m / z 978 fragment corresponding to the UD 17 G trinucleotide generated by digestion of tRNA Phe GAA by RNAseT1 (Figure 3 A).This fragment (its corresponding intensity showing background level) was absent in the double mutant strain while the intensity of the m / z 976 peak increased.Similar results were observed for tRNA Phe  GAA from the dusB1::kan strain.In contrast, the UD 17 G fragment was detected in the dusB2::erm strain with intensity comparable to that of the W168 control.Therefore, these results suggested that Bs DusB1 was responsible for D17 biosynthesis.D20 was probed with two distinct fragments of D-containing tRNA Phe  GAA from two different digestions.The first digestion, performed with RNAseA, yielded the GGD 20 trinucleotide ( m / z 1017) (Figure 3 B).The second digestion, performed with RNAseT1, generated the trinucleotide D 20 AG ( m / z 1001).In both scenarios, these two fragments did not disappear in tRNA Phe  GAA from the two dusB single deletion strains, although a more consequent decrease in intensity was observed in the case of dusB2::erm .In contrast, in the case of the double mutant, the peak was no longer detectable.We concluded that D20 was inserted into tRNA Phe GAA using both Bs DusB1 and Bs DusB2, with a dihydrouridylation efficiency that appeared to be higher for Bs DusB2.D20a was detected in two different tRNAs: tRNA Arg ICG via the GGAD 20a ( m / z 1346) fragments obtained by RNAseA treatment and AD 20a AG generated by RNAseT1 ( m / z 1346) ( Supplementary Figure S1 ), and tRNA Met  CAU via the CD 20a AG fragment ( m / z 1306) obtained by RNAseT1 (Figure 3 C).In the case of D20a in tRNA Met CAU , both DusBs participated in its synthesis as neither mutant caused a substantial decrease in the intensity of the m / z 1306 peak, and their profiles were quite similar to that of the wild type.However, for D20a in tRNA Arg ICG , only the deletion of BsdusB2 or the double mutant led to a significantly decreased peak at m / z 1346, accompanied by an increase in the peak at m / z 1344 corresponding to the non-dihydrouridylated fragment ( Supplementary Figure S1 ).These results suggest that the involvement of the two Bs DusB paralogs in D20a biosynthesis may depend on the tRNA substrate.Lastly, D47 was assayed by following the D 47 CG fragment ( m / z 977), derived from treatment of tRNA Met  CAU with RNAseT1 (Figure 3 D).This analysis was carried out following the same analytical grid as before.The tRNA Met  CAU from wild type and dusB2::erm B. subtilis strains retained the prominent peak at m / z 977.In contrast, in the case of the dusB1::kan or the double mutant strains, the intensity of this peak drastically decreased concomitantly with the increase in the m / z 975 peak, suggesting that Bs DusB1 was also responsible for D47 biosynthesis.

Dihydrouridylation redundancy targets several tRNAs as in vestig ated by deep-sequencing based AlkAnilineSeq method
An analysis of B. subtilis Bs DusB in vivo specificities was performed using the AlkAnilineSeq method (see supplementary methods for details) ( 42 ).This method exploits the Dring's instability under alkaline conditions ( 20 ), leading to its cleavage and the formation of β-ureidopropionic acid.This instability results in aniline-driven RNA cleavage, generating a 5 -phosphate group (5 -P) on the neighboring N + 1 residue, which serves as an input for highly selective ligation of sequencing adapters.Alongside D-residue detection, AlkAnilineSeq also allows parallel detection of 7-methylguanosine (m 7 G), 3-methylcytidine (m 3 C) and 5hydroxycytidine (ho 5 C), which share some degree of fragility in their base rings and / or N-glycosidic bonds, present in these modified residues.Mapping was achieved for all D containing tRNAs from the four B. subtilis strains, including the W168 strain, as well as the single and double dus deletion strains.It is important to emphasize that none of the D residues detected by this method was present at stoichiometric levels, suggesting partial dihydrouridylation of the target uridines.Importantly, the results obtained by AlkAnilineSeq were consistent with the MALDI-MS mapping experiments.For example, the disappearance of D17 in tRNA Ala  GGC and tRNA Ala UGC of B. subtilis W168 was observed only in dusB1::kan and double deletion strains, suggesting that Bs DusB1 was involved in the reduction of U17 in these tRNAs.In tRNA Arg ACG , the loss of both D17 and D20a was seen in the double deletion strain, whereas in the dusB1::kan strain, only the loss of D17 was observed ( Supplementary Figure S2 ).In contrast, in the dusB2::erm strain, the signal attributed to D20a declined when compared to the signal observed in dusB1::kan , while D17 remained unchanged ( Supplementary Figure S2 ).This is consistent with the fact that Bs DusB1 was responsible for the formation of both D17 and D20a, whereas Bs DusB2 formed only D20a in this tRNA ( Supplementary Figure S2 ).Moreover, Bs DusB1 was implicated in the biosynthesis of all three D17 / D20 / D20a residues in tRNA Asp GUC , whereas Bs DusB2 participated only in the latter two positions.In the case of tRNA Glu UUC , Bs DusB1 was only capable of forming D20, while Bs DusB2 could form both D20 and D20a.These findings suggested that the two enzymatic dihydrouridylation activities did overlap.
To gain a comprehensive view of both Bs Dus enzymes' activity, we generated an activity profile heatmap, as presented in Figure 4 .The heatmap clearly demonstrates that only the double mutant lacked all D residues in tRNAs, consistent with both LC-MS and MALDI-TOF data.This supports the earlier observation that both Dus enzymes are essential for dihydrouridylation across the full range of tRNA substrates.Moreover, it is evident from the heatmap that only the Bs DusB1 enzyme was involved in the formation of D17, whereas both enzymes contributed to the formation of D20 and D20a.Further analysis of the D signal intensities revealed that while most of the D20 and D20a residues were synthetized by both Bs DusB1 and Bs DusB2, a few dihydrouridylation events preferentially used Bs DusB2 (such as for D20 in tRNA Also the AlkAniline-Seq method did not detect the presence of D47, unlike the experiments performed by MALDI-MS on tRNA Met CAU .This discrepancy could be explained by interference caused by m 7 G46, which produces a strong AlkAnilineSeq signal.

Bs DusB1 and Bs DusB2 are flavoproteins characterized by a distinct polarity of their active site
The Bs DusB1 and Bs DusB2 proteins share a relatively low sequence identity of 26% ( Supplementary Figure S3 ).To characterize these two proteins in vitro , the genes encoding Bs DusB1 (BSU00810, Uniprot Id P37567) and Bs DusB2 (BSU08030, Uniprot Id O31546) were cloned into expression vectors, expressed in E. coli , and subsequently purified to homogeneity ( Supplementary Figure S4 ).To determine their oligomeric state in solution, gel filtration on a Superdex increase 75 10 / 300 column was performed, revealing that both proteins exist as monomers with an estimated molecular weight (Mw) of approximately 40 kDa for Bs DusB1 (elution volume ∼ 11.2 ml) and 39 kDa for Bs DusB2 (elution volume ∼ 11.7 ml).Bs DusB proteins were found to be copurified with their flavin coenzyme, evident from the yellowish color of the protein samples and characteristic absorbance spectra (Figure 5 A).The latter featured two absorption bands typical for flavin: the S0-S2 bands exhibited a maximum at 372 nm, while the S0-S1 band in Bs DusB1 and Bs DusB2 showed a maximum at 450 and 458 nm, respectively.The difference in the wavelength maximum of the S0-S1 transition between the two proteins suggests dissimilarity in the polarity of their active sites.Upon the addition of sodium dodecyl sulfate (SDS), the proteins denatured, releasing flavin into the solution.The resulting flavin in solution displayed an absorption spectrum similar to that of free FMN, confirming that both Bs DusB enzymes are flavoproteins with the FMN non-covalently bound to the apoprotein.FMN fluorescence in both holoproteins was also monitored and showed a slight red shift in the maximum fluorescence emission band of Bs DusB2, at 530 nm, compared to that of Bs DusB1 observed at 527 nm, supporting the existence of distinct environments for the two FMN coenzymes ( Supplementary Figure S5 ).This polarity contrast is substantiated by our analysis of the active sites in the holoprotein forms of Bs DusB1 and Bs DusB2 Alphafold models (see supplementary results and Figure 5 B).

An unusual behavior of Dus pyrimidine discrimination and dihydrouridylation activity of tRNA
Dus enzymes share a highly conserved catalytic mechanism that involves two redox reactions ( 22 ,24 ).NADPH reduces FMN to yield FMNH − , which is then oxidized to upon reduction of uridine to dihydrouridine.We measured the NAD(P)H oxidase activity of the two Bs DusB enzymes independently by monitoring the consumption of NADH or NADPH under aerobic conditions using absorbance spectrophotometry at 340 nm and steady-state conditions.The data were analyzed using the Michaelis-Menten formalism and the related kinetics parameters are presented in Table 1 .The results revealed that Bs DusB1 oxidized NADPH and NADH with identical catalytic constants ( k cat ∼ 0.013 s −1 ) and comparable K M values, indicating that the enzyme did not discriminate between NADH or NADPH and could use both equally.This result was unexpected, because all previously studied Dus enzymes, both prokaryotic and eukaryotic, showed a preference for NADPH over NADH ( 30 , 33 , 43 ).In contrast, for Bs DusB2, NADPH was a better substrate than NADH due to a lower K M for NADPH (2 μM) than for NADH (22 μM) and ∼ a 3-fold higher catalytic constant for NADPH than for NADH.Overall, NADPH exhibited a 5-fold higher catalytic efficiency than NADH.
To examine the Bs DusB activity of B. subtilis , in vitro dihydrouridylation assays were performed with bulk tRNAs from the double deletion strain, and the reaction products were traced using LC / MS.In the presence of 1 μM protein, Bs DusB2 was able to restore a 40% higher D level compared to Bs DusB1 after 1 hour, indicating that Bs DusB2 is the more active enzyme (Figure 5 C).

Structural characterization of DusB enzymes and RNA binding
The structural models of Bs DusB1, Bs DusB2, Ec DusB and Mc DusB1 were examined using models generated through Alphafold Colab2 (Figure 5 D).The derived models exhibited per-residue confidence scores exceeding 90% across most of their respective regions, as illustrated in Supplementary Figure S6 A. As anticipated, these enzymes display a conserved canonical folding of the Dus family, i.e. (i) a catalytic domain adopting a TIM-Barrel type structure (TBD) where the flavin coenzyme binding site lies at the entrance of the barrel, (ii) a helical domain (HD) composed of a 4-helix bundle, and (iii) a short linker of about 10 amino acids connecting the two domains.Conducting a structural alignment and comparing the models revealed low RMSD values within the Bs DusB1 subfamily ( Supplementary Figure S6 ).This supports the notion that the models for Bs DusB1, Mc DusB1 and Ec DusB1 exhibit highly similar structures.A broad distribution of positive surface charges accessible to the solvent, most likely engaged in interactions with the tRNA substrates, can be distinguished (Figure 5 E).This distribution is arranged on both sides of a line of demarcation (LOD) that can be drawn from the left extremity of the TBD throughout the active site cavity, ending at the lower tip of the HD at the C-terminus.Several interesting points can be observed based on this spatial arrangement.Bs DusB1 has a continuous, positive electrostatic surface stretched on both sides of the LOD, whereas Bs DusB2 is distinguished by a positive surface forming an elongated stripe parallel to the LOD and spanning almost on all its length, but primarily found on the proximal side of this line.In Ec DusB, a significant portion of the positive area forms an off-center globular area on the distal edge of the LOD, involving predominantly the apical region of the TBD.Mc DusB1 shows a certain similarity to Bs DusB1 but with a distinctive feature, namely the presence of several rather isolated positive charge patches.Thus, each of the studied DusB seems to have its own tRNA binding pattern, likely adapted to its site specificity.Likewise, each Dus will probably orientate the tRNA in a distinct way to allow the active site of the enzyme to gain access to the correct uridine substrate to be modified ( 21 ,44 ).To evaluate whether this difference in positive surface area affects the stability of the enzyme / tRNA complex, we examined the ability of Bs DusB1 and Bs Dus2 to bind to tRNA by specifically monitoring the impact of tRNA titration on flavin fluorescence.Addition of tRNA resulted in an increase in FMN fluorescence of both Bs DusB describing a cooperative process

RNA dihydrouridylation broadening specificity depends on enzyme concentration
Our complementation results in the E. coli triple dus mutant strain ( dusA ::kan, dusB ::Ø, dusC ::Ø) with Bs-dusB1 or BsdusB2 demonstrated that both enzymes could dihydrouridylate positions U17, 20 and 20b, acting as both Ec DusB and Ec DusA (see supplementary results & Supplementary Figure S7 ).While the outcomes for Bs DusB1 were anticipated, the unexpected capability of Bs DusB2 to catalyze D17 formation in E. coli was intriguing.This finding suggested several possibilities in a heterologous context: (i) Bs DusB2 lost its substrate specificity due to the differences in tRNA nature (sequence and modification profile) between both organisms; (ii) a protein partner, RNA, or other compounds in B. subtilis controlled the site specificity; or (iii) the intracellular concentrations of Dus proteins differed between E. coli and B. subtilis .Indeed, complementation assays in E. coli were performed with BsdusB1 or BsdusB2 under the control of an arabinose-inducible promoter with concentration of inducer adjusted to allow for the detection of dihydrouridylation.In contrast, in B. subtilis both genes are expressed from the chromosome by their own promoter.
To further explore these possibilities, we assessed the effect of increasing enzyme concentrations on Bs DusB's dihydrouridylation activity in vitro using tRNA from the B. subtilis double deletion strain as a substrate.Additionally, we performed the experiments in the presence of B. subtilis dusB1::kan , dusB2::erm cell extract to examine the existence of a potential partner for Bs DusB2 that might be essential for its site specificity.AlkAnilineSeq quantifications showed that the level of D17 inserted by Bs DusB1 and Bs DusB2 increased with enzyme concentration (Figure 6 A), confirming that Bs DusB2 can synthetize D17 in vitro on tRNA from B. subtilis .AlkAnilineSeq also provided insights into the dihydrouridylation efficiency for all D-sites (Figure 6 B).Dihydrouridylation efficiency seemed to depend on the nature of the tRNA and the modified position.As expected, Bs DusB1 formed D17 / D20 / D20a.Except for tRNA Lys UUU and tRNA Ala GGC , the dihydrouridylation efficiency was higher at positions 20 and 20a than at position 17.Experiments conducted with crude B. subtilis extracts revealed that Bs DusB2 retained its ability to synthesize D17 even at higher enzyme concentrations (data not shown), suggesting the absence of a cellular partner that regulates the specificity of this Dus enzyme.To validate these findings in vivo , both wild type and mutant strains were transformed with plasmids overexpressing either Bs DusB1 or Bs DusB2.AlkAnilineSeq profiles from these strains clearly demonstrated that overexpression of Bs DusB2 in dusB1-deficient strains or Bs DusB1 in dusB2-deficient strains was able to restore the dihydrouridylation profile for a significant subset of tRNAs ( Supplementary Figure S8 ).Moreover, Bs DusB2 exhibited the capability to introduce D17 residues into several tRNAs, indicating its functional equivalence to Bs DusB1 upon overexpression ( Supplementary Figure S8 ).Taken together, these results demonstrate that specificity likely depends on both the nature of the tRNAs and the enzyme concentration.

Effect of Bs DusB deletions on cell growth
The optimal growth temperature of B. subtilis ranges from 35 to 37 • C. The influence of the lack of Bs DusB and by extrapolation of D on the growth of B. subtilis was investigated in LB medium at 23, 30 and 37 • C (Table 2 ).At the standard growth temperature of 37 • C, B. subtilis W168 exhibited a generation time of 21 minutes.However, in the case of the three strains with deletions in either one or both dus genes, there was a slight increase in generation time.The effect was slightly more visible when cells were grown at 30 • C, with the generation time rising from 31 min for the wild type to 39 and 40 min for ΔdusB1 and ΔdusB2 , respectively.The effect was even more pronounced in the double mutant strain, where this doubling-time increased to 43 min.A more significant difference in growth was observed when the temperature was lowered to 23 • C. Here, the generation time increased from 49 minutes for W168 to 87 min for the three mutant strains.Thus, the absence of D does not seem to have too great of an impact on B. subtilis at physiological growth temperatures, but becomes significant at low temperature such as 23 • C.This observation aligns with the role of this modified base in promoting structural flexibility at the tRNA level, a feature that is more crucial at lower temperatures than at higher ones.
Generation time is just one among several growth parameters for bacteria, serving as an indicator of potential fitness loss.Therefore, we conducted competition experiments between mutants and the wild-type strain to evaluate the impact of tRNA dihydrouridylation loss on mutant fitness.Surprisingly, all mutant strains exhibited decreased fitness compared to the wild type, even at 37 • C, with the ΔdusB1 strain showing the lowest competitive index ( Supplementary Figure S9 ).However, observed differences in fitness among mutants were not statistically significant ( t -test, P > 0.03), suggesting a potential role of the redundancy in specificity of Bs DusB enzymes.

Discussion
We investigated the role of the two homologs, DusB1 and DusB2, in D base biosynthesis in B. subtilis tRNAs.Both Bs DusB enzymes are FMN-dependent flavoenzymes with a conserved canonical structure of bacterial Dus, retaining key catalytic residues (Figure 5 A,B).However, they differ in the polarity of their active sites and preference for the reducing agent, NAD(P)H (see supplementary discussion and Table 1 ).Most modification enzymes are highly site-specific and modify only one position.However, a small number of enzymes exhibits promiscuous site specificity, targeting either adjacent bases, or multiple positions scattered along the nucleotide sequence of their RNA substrate, or even have both capabilities ( 45-52 ) (see also Supplementary Table S4 ).The Dus enzymes also display the two cases of targeting juxtaposed uridines as observed with bacterial DusA ( 31 ) and Bs DusB2 for U20-U20a, and with eukaryotic Dus1 (U16-U17) and Dus4 (U20a-U20b) ( 20 ).Gram + Dus enzymes show a wider multisite specificity as seen with the Mc DusB1 that modifies the U17-U20-U20a triplet ( 33 ) and reinforced here with the discovery that Bs DusB1 modifies not only the same bases as Mc DusB1 but also the U47 (Figure 3 ).D47 is located in the variable loop which, in eukaryotes, is catalyzed by Dus3, an enzyme that differs from all Dus by its size and complex modularity ( 22 ).
Remarkably, we uncovered an unprecedented property in modification enzymes namely, functional redundancy.This property remains very enigmatic since Bs DusB1 can intro- duce almost the entire D content while Bs DusB2 provides a backup activity for positions 20-20a with an efficiency largely in favor of this enzyme.It is worth mentioning that this overlap in activity concerns most tRNAs (Figure 4 ).Nevertheless, Bs DusB2 also have its proper tRNA substrates not shared by Bs DusB1 suggesting that this D20-D20a redundancy in dihydrouridylation activity targets a specific set of tRNAs.Surprisingly, Bs DusB2 has also the ability to modify U17 only at a certain enzyme concentration, which could probably be consistent with a lower dihydrouridylation efficiency for this site (Figure 6 ).Of note, Dus enzymes involved exclusively in D20 (or D20-D20a) modification seem to be always more ac- The physiological significance of this redundancy in B. subtilis raises intriguing questions.In general, homolog-based functional redundancy can provide functional resilience or flexibility to cope with varying conditions or stresses (53)(54)(55).This could indeed apply to Bs DusB taking the advantage of having one enzyme more efficient than the other, especially when dealing with redox reactivity issues.It is tempting to propose that this backup functionality could be a more efficient way to dihydrouridylate tRNAs under conditions or events leading to significant tRNA damage requiring rapid maturation of newly transcribed pools of tRNAs to afford the cell to cope with abrupt environmental changes notably under limiting NADPH concentration for example.In such a scenario, up-regulation of Bs DusB2 could also be an additional mean by which the cell boosts tRNA-dihydrouridylation activity but also extends its site specificity to compensate for the low Bs DusB1 activity .Interestingly , such type of regulation has precedent as exemplified by the downregulation of the gene coding for the mesophilic Clostridium botulinum DusB homolog during a heat shock stress at 45 • C ( 56 ).In that specific case, D has probably no utility at high temperatures, and thus this bacterium would naturally require less D and would therefore decrease the expression of its cognate enzyme.DusC is also differentially regulated in response to the growth temperature in the thermophilic B. manusensis ( 57 ).In B. subtilis , our studies revealed a visible impact of the absence of Bs DusB1 or Bs DusB2 on the growth phenotype of this organism (Table 2 ), suggesting that loss of D can have significant effects on cell physiology.
Another speculative yet intriguing possibility for this functional redundancy is related to the evolutionary process of these enzymes.Both Bs DusB1 and Bs DusB2 originated from a duplication event of an ancestral Dus enzyme likely multisite specific.Bs DusB1 has retained the functional features of this ancestral enzyme, while Bs DusB2 might be undergoing a process of functional speciation.This could explain why Bs DusB2's dihydrouridylation activity at position 17 is detectable only under high enzyme concentration (Figure 6 ).Comparative analysis of the presumed tRNA binding interfaces on Bs DusB models suggests that the Bs DusB2 interface is clearly different from the others with nonetheless some positive charges that remain common to these enzyme systems (Figure 5 D).This agrees with the fact that Bs DusB2 may preferentially bind its tRNAs according to its own recognition mode.Previous phylogenetic analyses proposed that DusB / DusB1 was the common ancestor to all bacterial Dus proteins ( 34 ), a finding that we reproduce in a small-scale analysis with reference bacterial genomes (Figure 7 ).While the exact timing of the branching of the DusB2 subgroup from the DusB group remains uncertain, it is clearly distinct from both the DusA and DusC subgroups.Further comprehensive phylogenetic analyses would be needed to understand this evolutionary relationship.However, this data suggests that Bs DusB2 might be converging towards DusA-type activities, specifically modifying the 20 / 20a position, while potentially losing its capacity to modify the U17 positions like its DusB homologues.This suggests a possible transitional state in the evolution of Bs DusB2's enzymatic specificity.

Figure 1 .
Figure 1.Location of D-sites in tRNA and the corresponding enzyme involved in site dihydrouridylation determined experimentally.Schematic representation of the secondary str uct ure of tRNA, showing the location of D residues and the corresponding Dus enzyme responsible for their synthesis in E. coli , T. thermophilus and M. capricolum for eubacteria and S. cerevisiae for eukaryotes.In the lower panel is shown the sequence of B. subtilis tRNAs used to analyze the D-sites in the MALDI-MS experiments.
ArgICG for D20a and tRNAMet  CAU for D20a and D47 (Figure1).The mass profiles of these tRNAs are depicted in Figure3A-D and Supplementary FigureS1.Analysis of tRNAs from the W168 strain confirmed the presence of all distinct Dcontaining fragments at the expected positions, validating the

Figure 2 .
Figure 2. Quantification of D-le v el in tRNA from B. subtilis .( A ) Extracted ion chromatograms of dih y drouridine in tRNAs isolated from B. subtilis WT strain (W168 in light green), dusB1::kan , dusB2::erm double deletion strain (orange) and dusB1::kan (blue) and dusB2::erm single mutant strains (cy an).T he signals w ere normaliz ed to the respectiv e UV signal of A denosine.( B ) D le v els determined in bulk tRNAs of B. subtilis WT strains (W168 in light green), dusB1 (blue) and dusB2 (cyan) single mutants, dusB1 dusB2 (orange) or double deletion complemented with either DusB1 of M. capricolum (magenta) or DusB2 from S. aureus (green).The strains were grown in LB media at 37 • C. Results are shown as average of three biological replicates in relation to the wild type strain W168.

Figure 3 .
Figure 3. MALDI-TOF analysis of position 17, 20, 20a and 47 in tRNAs from B. subtilis WT and Dus deletion mutants.( A ) D17-containing MS relative isotope patterns of derived oligonucleotides after RNAse T1 treatment of tRNA Phe #AA isolated from wild type, dusB1 dusB2 , dusB1 and dusB2, respectively.( B ) D20-containing MS relative isotope patterns of derived oligonucleotides after RNAse A treatment of tRNA Phe #AA isolated from wild type, dusB1, dusB2 and dusB1 dusB2 , respectively.( C ) D20a-containing MS relative isotope patterns of derived oligonucleotides after RNAse T1 treatment of tRNA Met CAU isolated from wild type, dusB1, dusB2 and dusB1 dusB2 , respectively.( D ) D47-containing MS relative isotope patterns of derived oligonucleotides after RNAse T1 treatment of tRNA Met CAU isolated from wild type, dusB1, dusB2 and dusB1 dusB2 , respectively.Further details of the tRNA-derived oligonucleotide fragments and their sizes ( m / z ) used for the identification of DusB specificities are shown in supplementary figures .
tRNA Ile CAU , tRNA Ser UGA , tRNA Arg CCG ).We did not detect D-signal for three tRNA, namely tRNA Ile GAU , tRNA Pro UGG and tRNA Val UAC .

Figure 4 .
Figure 4. Heatmaps for the assessment of dihydrouridylation changes in individual modified sites in tRNAs from B. subtilis and its DusB mutants.The heatmap displa y s one specific D-modification's stoichiometry across the dif ferent samples (in X-axis) and the dif ferent D-sites retained f or analy sis (in Y-axis).The stoichiometry is blue-coded and relies on through stop ratio of the AlkAnilineSeq detection method, which detects m 7 G, m 3 C and D. R1, R2 and R3 represent the results for the three different replicas.

Figure 5 .
Figure 5. Str uct ural and functional characterization of B .subtilis DusB .( A ) UV-visible absorption spectra of Bs DusB1 (blue) and Bs DusB2 (teal) holoproteins.( B ) Comparative str uct ural models of the active sites of Bs DusB1, Bs DusB2 and DusB of E. coli ( Ec DusB).The active site view is centered on the o v erla y of a section encompassing the FMN isoalloxazine (yellow) of the respective active site of the three DusB ( Bs DusB1 in pink, Bs DusB2 in blue and Ec DusB in deep olive).Residues around the FMN are shown in stick in the respective color codes of the Dus.( C ) In vitro dihydrouridylation activity test of recombinant Bs DusB at 1 μM of enzyme after 1 hour incubation at 37 • C. Dih y drouridine le v els w ere determined b y LC-MS / MS and normalized to the UV signal of adenosine.To compare the activity of Bs DusB, the activity of Bs DusB1 was set to 100%.Results are shown as average of biological duplicates.( D ) Str uct ural models of the DusB holoenzymes from B. subtilis , E. coli and M. capricolum .Except for Ec DusB, which is a crystallographic str uct ure (PDB , 6EI9), the other three models are from Alphafold.TBD = TIM Barrel Domain, HD = Helical domain.The FMN is shown in y ello w stick.( E ) Electrostatic surface of the Dus model.The dashed line represents the line of demarcation (LOD) mentioned in the text.( F ) Isotherm of tRNA binding to Bs DusB.F529nm is the change in FMN fluorescence at 529 nm resulting from tRNA titration to Bs DusB1 (blue) and Bs DusB2 (teal).

Figure 6 .
Figure 6.In vitro biosynthesis of D in B. subtilis tRNAs catalyzed by the recombinant Bs DusB1 and Bs DusB2 proteins.( A ) Recombinant enzymes expressed in E. coli and purified were incubated with D-unmodified B. subtilis total RNA fraction extracted from dusB1::kan , dusB2::erm strain.Quantification of D17 le v el w as done using NormCount score of AlkAnilineSeq (the signal normaliz ed to median of back ground clea v ages in the surrounding 10 nucleotides).NormCount score (as well as other AlkAnilineSeq Scores) does not show linear dependence from D content, but provides good compromise between sensitivity and specificity of detection for low D levels in tRNA.Only 8 best modified tRNA sites are shown (out of 18 altogether).Concentration of the recombinant Bs DusB1 and Bs DusB2 is expressed in μM.Identity of tRNA substrates analyzed is shown at the right.( B ) Modification efficiency of the D-sites measured at 25 μM of enzymes.Quantification of D le v el w as done using NormCount score of AlkAnilineSeq.

Figure 7 .
Figure 7. Phylogenetic analysis of DusB1, DusB2, DusC and DusA proteins in 120 reference and complete Bacteria.DusA proteins are in blue.DusC proteins are in green.DusB / DusB1 proteins are in black.DusB2 proteins are in red.The B V -BRC annotations seem to correctly group the proteins with one e x ception, the Caur_0210 protein annotated as DusB but clustering with the DusB2 proteins.T his section of the tree has ho w e v er v ery lo w bootstrap values as the thickness of the tree branches are reflective of the bootstrap percentage values.E. coli proteins are highlighted in yellow and B. subtilis proteins in purple.

Table 1 .
Kinetic parameters for NAD(P)H oxidase activity of B. subtilis Dus

Table 2 .
Effect of dus deletion on the generation time of B.

subtilis 37 • C 30 • C 23 • C
*The generation times are expressed in minutes.