Exploring a unique class of flavoenzymes: Identification and biochemical characterization of ribosomal RNA dihydrouridine synthase

Significance

Posttranscriptional modifications of rRNA play a crucial role in ribosome biogenesis and fine-tuning of protein synthesis. These are predominantly concentrated at decoding and peptidyl transferase center (PTC) sites. To date, 35 of the 36 identified ribosomal modifications have been characterized, although the biosynthetic pathway for U2449 dihydrouridylation within the PTC has remained elusive. To fill this gap in knowledge, we developed a technique for monitoring dihydrourylation reaction of 23S rRNA. This method involves chemical labeling of dihydrouridine and RT-PCR blockage. By integrating this approach with epitranscriptomic, genetic, and biochemical methodologies, we successfully identified a unique class of flavin-dependent enzymes involved in rRNA dihydrouridylation. This finding opens up different avenues for exploring the biological significance of PTC dihydrouridylation in ribosomal function.

Abstract

Dihydrouridine (D), a prevalent and evolutionarily conserved base in the transcriptome, primarily resides in tRNAs and, to a lesser extent, in mRNAs. Notably, this modification is found at position 2449 in the Escherichia coli 23S rRNA, strategically positioned near the ribosome’s peptidyl transferase site. Despite the prior identification, in E. coli genome, of three dihydrouridine synthases (DUS), a set of NADPH and FMN-dependent enzymes known for introducing D in tRNAs and mRNAs, characterization of the enzyme responsible for D2449 deposition has remained elusive. This study introduces a rapid method for detecting D in rRNA, involving reverse transcriptase-blockage at the rhodamine-labeled D2449 site, followed by PCR amplification (RhoRT-PCR). Through analysis of rRNA from diverse E. coli strains, harboring chromosomal or single-gene deletions, we pinpoint the yhiN gene as the ribosomal dihydrouridine synthase, now designated as RdsA. Biochemical characterizations uncovered RdsA as a unique class of flavoenzymes, dependent on FAD and NADH, with a complex structural topology. In vitro assays demonstrated that RdsA dihydrouridylates a short rRNA transcript mimicking the local structure of the peptidyl transferase site. This suggests an early introduction of this modification before ribosome assembly. Phylogenetic studies unveiled the widespread distribution of the yhiN gene in the bacterial kingdom, emphasizing the conservation of rRNA dihydrouridylation. In a broader context, these findings underscore nature’s preference for utilizing reduced flavin in the reduction of uridines and their derivatives.

Posttranscriptional maturation processes shape RNA molecules into functional biomolecules. Among these processes, the incorporation of chemical groups at the nucleoside level, primarily within the bases, is catalyzed by modification enzymes. Over 170 modifications have been identified to date, with tRNAs and rRNAs being the most extensively modified RNA species (1). The chemical diversity of these modifications can significantly alter the physicochemical properties of nucleosides, exerting a profound impact on local and even global RNA structures. For the majority of modifications, the effect is stabilizing in nature (2, 3). For instance, the introduction of methyl groups tends to strengthen base stacking. One of the most abundant modifications across the transcriptome is dihydrouridine (D) (4, 5) (Fig. 1A). This modified base is evolutionarily conserved and is predominantly found in bacterial and eukaryotic tRNAs. Recently, D has also been detected in certain eukaryotic mRNAs (6–8) and long noncoding RNAs (6, 7, 9). In the case of rRNAs, sporadic observations of D have been reported, notably at position 2449 (D2449) in domain V of the 23S ribosomal RNA of Escherichia coli, in the peptidyl transferase center (PTC) (10, 11) (Fig. 1 B and C). D2449 has also been identified at positions 2449 and 2500 in Clostridium sporogenes (with D2449 being partially methylated to m⁵D) (12) and at position 1211 or 1212 in the 16S rRNA of Clostridium acetobutylicum (13). Comprehensive studies to investigate the presence of dihydrouridine in the rRNAs of bacteria and eukaryotes remain to be made.

Fig. 1.

Dihydrouridylation of the PTC of the E. coli ribosome: (A) Biosynthetic reaction of dihydrouridine involves the reduction of the C5=C6 double bond of uridine. This redox reaction requires 2 electrons + 2 protons. (B) Secondary structure of the V domain of E. coli 23S rRNA. D2449 is highlighted in red, and the question mark indicates that the associated modification enzyme was unknown before our current study. (C) High-resolution structure of the E. coli ribosome in complex with tRNA (magenta). The PDB associated with this structure is 8B0X (11). The Inset on the Right provides a close-up of the PTC, showing the positioning of D2449 (pink) in proximity to A2451 and m²G2445.

The study of D synthases involved in tRNA dihydrouridylation has revealed that the conversion of U to D is a redox-dependent reaction relying on NADPH as a reducing agent (4, 14–21) (Fig. 1A). Dus enzymes are flavoenzymes, utilizing FMN as a coenzyme for redox reactions. In the initial step, the flavin is reduced to an anionic hydroquinone (FMNH⁻), which then transfers a hydride to the C5–C6 double bond of uridine, forming an enolate intermediate (4, 22). This intermediate accepts a proton from the strictly conserved catalytic cysteine, leading to the breakdown of D’s aromaticity, resulting in a nonplanar base and a more flexible ribose conformation. The significance of D’s base flexibility is highlighted by the fact that psychrophilic organisms generally have more D in their tRNAs than thermophiles (23). Structurally, D destabilizes small stem loops, promoting the formation of larger loops (24). The presence of D2449 in the central large loop of domain V in E. coli’s 23S rRNA could potentially serve this purpose, although addressing this issue is challenging due to the unidentified gene encoding ribosomal Dus. Beyond its presumed structural role, the biological significance of D remains obscure. Recent discoveries of D residues in yeast and human mRNAs, also introduced by tRNA dihydrouridine synthases, suggest potential roles in meiosis and translational efficiency through yet-to-be-determined molecular mechanisms (6, 8). These studies indicate that Dus enzymes are promiscuous, although, at least in the case of E. coli, they do not act on rRNA.

The E. coli rRNA encompasses 36 modifications, and to date, all modification enzymes for E. coli rRNA have been identified, except for the rRNA D synthase (1) (SI Appendix, Table S1). In this study, we identified this elusive enzyme, hereafter referred to as RdsA (Ribosomal Dihydrouridine Synthase A), responsible for the biosynthesis of D2449 in 23S rRNA. Biochemical characterizations of RdsA unveiled its identity as a flavin-dependent D synthase with redox activity reliant on FAD and NADH. The proposed chemical mechanism of dihydrouridylation, derived from our findings, outlines a likely hydride transfer mechanism, underscoring nature’s preference for utilizing flavin to reduce uridine or its derivatives. Moreover, our phylogenetic data demonstrate that the RdsA gene is widely distributed among bacteria, including many pathogens, while absent in eukaryotes. This observation suggests that ribosomal D may be pervasive in microbial rRNAs, rendering it a potential target for selective antibiotic intervention.

Results

Development of a Rapid D-ribosomal Detection Test.

To facilitate the swift detection of D2449, we have developed a methodology termed RhoRT-PCR. This innovative approach involves the blocking of reverse transcriptase (RT) through dihydrouridine labeling with rhodamine 110, following a well-established mechanism (SI Appendix, Fig. S1). Subsequently, this step was followed by PCR amplification of RT-products employing a primer pair to target a specific region (Fig. 2A). The primer pair 2F/3R was meticulously designed to amplify the region downstream of position 2449 in the 23S rRNA, yielding a 248 bp fragment. It functions as a positive control for RT activity. Conversely, the region encompassing position 2449 was amplified using primers 1F/3R, generating a 429 bp product. Therefore, the absence of rhodamine labeling on D2449 or the lack of dihydrouridylation at position 2449 (indicative of U2449) should result in the presence of both 248 and 429 bp fragments, discernible on an agarose gel. In contrast, the presence of D2449 would result solely in the 248 bp band. The validity of this approach was tested using purified rRNA from wild-type strains of E. coli (known to contain D2449) and Bacillus subtilis (known to harbor a uridine at the corresponding rRNA position). As depicted in Fig. 2B, in the case of E. coli, the absence of Rho labeling showed the anticipated two fragments. However, prelabeling D2449 with Rho resulted in the disappearance of the 429 bp band, strongly indicating the presence of D2449. Conversely, in the case of B. subtilis, whether labeled or not, both bands persisted, suggesting the absence of D in this region of rRNA. Recently, it has been demonstrated that the three Dus enzymes in E. coli (DusA, B, and C) are not responsible for the synthesis of D2449 (7). To confirm this result, we applied our RhoRT-PCR method to purified rRNA from the E. coli triple mutant strain ΔdusABC. As shown in SI Appendix, Fig. S2A, the result revealed a single band at 248 bp, clearly demonstrating the presence of D2449 in this mutant. This indicates that the classical Dus enzymes are not involved in the dihydrouridylation of E. coli 23S rRNA. Thus, the RhoRT-PCR technique emerges as an effective method for the precise detection of D2449.

Fig. 2.

Principle of RhoRT-PCR. (A) Diagram showing the main steps of RhoRT-PCR and the expected results depending on the presence of D in the analyzed region. The red scrambled oval depicts specific D-labeling with rhodamine-110. (B) Representative results obtained with rRNA containing D at position 2449 (E. coli BW25113, 14 repeats) or U at the same position (B. subtilis W168, 2 repeats). RT was performed with species-specific 4R oligonucleotides. PCR amplification was performed with species-specific 1F/3R (1F) and 2F/3R (2F) oligonucleotides.

Identification of the Gene Encoding rRNA Dihydrouridine Synthase.

In the quest to pinpoint the gene responsible for dihydrouridine synthesis in rRNA, we applied the RhoRT-PCR technique to rRNA extracted from the E. coli strain ME5125 (25, 26). This strain is distinguished by 15 major deletions in its chromosomal DNA, making it an ideal candidate due to its ability to maintain the smallest genome among E. coli strains while retaining sufficient viability for rRNA isolation. As illustrated in Fig. 3A, the deletions exhibited both fragments, whether treated with rhodamine or not, strongly suggesting the probable absence of D. To substantiate these findings, we subjected BW25113 and ME5125 rRNA to MALDI mass spectrometry after RNAse T1 treatment. The analysis revealed the ²⁴⁴⁸ADAACAGp²⁴⁵⁴ fragment with an m/z of 2293.3 in the case of BW25113, while in ME5125, the corresponding fragment displayed an m/z of 2291.3, indicative of the non-dihydrouridylated fragment (Fig. 3B). Moreover, these results were validated by AlkAniline-Seq, underscoring the presence of D2449 in WT rRNA and its conspicuous absence in the ME5125 strain (Fig. 3C). Individual testing of the 15 deletion strains using RhoRT-PCR and AlkAniline-Seq showed that only the ME5079 strain, harboring a genomic deletion encompassing 77 genes (SI Appendix, Fig. S3), lacked D2449 (Fig. 3 A and C). A similar approach was used for the analysis of 10 strains bearing single-gene knockouts that were annotated with unknown functions but potentially linked to redox processes or ribosome biogenesis (SI Appendix, Table S2 and Fig. S2). Comprehensive testing across all these strains unequivocally demonstrated that only the rRNA from the ΔyhiN strain lacked D2449 (Fig. 3C and SI Appendix, Fig. S2A). Notably, yhiN is annotated as a putative oxidoreductase with an FAD/NAD(P)H domain and operates outside an operon (SI Appendix, Table S2). However, it is situated in proximity to two rRNA methyltransferase genes, namely rsmJ and rlmJ, and a putative ribosome-associated ATPase gene (rbbA) (SI Appendix, Table S2 and Fig. S3).

Fig. 3.

Identification of D2449 deficient E. coli strains. (A) Representative RhoRT-PCR results with the indicated strains (BW25113 and ME5125, 14 repeats; ME5079 and BW25113 ΔyhiN::kan, 3 repeats). Primers used are the same as in Fig. 2A. The absence of 1F/3R amplification with BW25113 rRNA as a substrate indicates the presence of D2449, which was confirmed by MALDI-TOF analysis of RNAse digestion fragments (2 repeats) (B) and AlkAniline-Seq (C) (for the latest, BW25113, ME5125, and ME5079, 3 repeats; BW25113 ΔyhiN::kan, 2 repeats).

The Gene yhiN Encodes for an FAD Binding Protein.

The E. coli yhiN gene encodes a 400-amino acid protein with a theoretical mass of 43.7 kDa (SI Appendix, Fig. S4). To elucidate the structural organization of YhiN, an AlphaFold model was generated (Fig. 4A). Remarkably, YhiN showcases a distinct fold compared to classical Dus proteins (4, 15–17, 19, 20). A PDB search revealed homology between YhiN and various flavoproteins, with crystallographic structures available for holoprotein forms (2GQF, 3V76, 4CNJ, 2I0Z), including Streptococcus oligofermentes L-aminoacetone oxidase. Structural alignments of the YhiN model with homologous structures revealed RMSD values below 2.3 Å (SI Appendix, Fig. S5), signifying significant structural conservation within this flavoprotein class. YhiN comprises three distinct domains with an intricately complex topology (Fig. 4A). The Rossmann fold domain (FBD), housing the flavin and putatively the catalytic site, accommodates the insertion of the other two domains between G190 and A337, resulting in the Nt and Ct ends being borne by this domain. The second domain (ins1D), homologous to the Ct domain of E. coli EF-Tu, is interrupted at residue L257 by the insertion of the third domain (ins2D), exhibiting structural homology with the Helix-2-turn-helix domain of the topoisomerase VI-B subunit of Sulfolobus shibatae. Ins2D spans L258 to L322, succeeded by a segment of ins1D and FBD. Structural alignment of YhiN with flavoprotein homologs facilitated the identification of the conserved FAD binding site (Fig. 4B and SI Appendix, Fig. S5). No clashes were observed between residues forming the FAD pocket and the coenzyme. The FAD assumes an elongated conformation upheld by a series of interactions, with the isoalloxazine, the redox center, situated within a solvent-accessible crevice. To assess the flavoprotein nature of E. coli YhiN (RdsA), recombinant protein production was undertaken. As depicted in Fig. 4 C and D, the UV/visible spectrum of freshly purified YhiN indicated the presence of oxidized flavin in 61% of holoprotein. The addition of SDS induced coenzyme release, yielding a UV/visible spectrum akin to free FAD. Additionally, mutating key residues in this identified FAD binding pocket led to a lower ratio of holoproteins ranging from 37%, when affecting residues interacting with the adenine moiety, to almost 0% when affecting residues in the vicinity of the ribityl (Fig. 4D). These findings conclusively establish YhiN as a noncovalently bound FAD flavoprotein characterized by a sophisticated structural topology.

Fig. 4.

Structural and biochemical characterization of E. coli rRNA dihydrouridine synthase. (A) Structural organization of E. coli RdsA. At the Top, a schematic diagram depicting the modularity of RdsA and delineating various domains—Rosman fold domain in pink, Ins1 domain in blue, and Ins2 domain in green. At the Bottom, an AlphaFold model of E. coli RdsA in a cartoon representation with the same color code as above. The Ins1 domain is homologous to the C-terminal domain of E. coli EF-Tu (PDB: 1EFC, colored in blue), while the Ins2 domain is homologous to the helix-2-turn-helix domain of S. shibatae DNA topoisomerase VI-B subunit (PDB: 1MU5, in green). (B) Holoprotein model of RdsA showing the binding of the FAD coenzyme (ball sticks, yellow) at the Rossman fold domain and key residues potentially interacting with FAD. (C) UV-Visible spectrum of native (red) and 0.1% SDS-denatured (blue) recombinant RdsA protein from E. coli. (D) FAD/protein molar ratio of recombinant RdsA harboring no change (WT) or mutations changing key residues interacting with FAD shown in B. This ratio ± SEM of 3 repeats was determined by spectrophotometric quantification of released FAD from a given amount of SDS-treated protein. For each mutant, the targeted interaction with FAD is indicated. (E) Michaelis–Menten curve for the NADH oxidation by RdsA. (F) Sequence and potential structure of the in vitro transcript Mini-PTC-U (Top) and Mini-PTC-C (Bottom) established with MXFold2 server. Nucleotide at position 21 is the anticipated target of dihydrouridylation. (G) Dihydrouridylation activity of RdsA on a Mini-PTC-U (blue) and Mini-PTC-C (green). D was quantified by LC–MS after incubation in reaction buffer of the designated transcript without or with 25 µM of RdsA for the indicated time. Results are expressed as the percentage of quantified D per quantified A ± SEM of 3 to 4 repeats (%mod/A).

The Flavoprotein YhiN: Unraveling Its Role as the rRNA Dihydrouridine Synthase.

The conversion of uridine to dihydrouridine is a redox reaction requiring 2 electrons + 2H⁺ (or hydride + H⁺) (Fig. 1A) (4). In the case of classical Dus proteins, their RNA substrates are reduced by FMNH⁻, produced through the prior reduction of FMN by NADPH (4, 22). Initially, we investigated whether NADPH could serve as a reducer of flavin by monitoring the spectrophotometric NADPH oxidase activity at 340 nm of recombinant YhiN. However, no NADPH consumption was observed. Conversely, we observed efficient NADH oxidation with a k_cat of 0.027 ± 0.003 s⁻¹ and a K_M of 6.5 ± 1.5 µM (Fig. 4E). With a conclusive reducing source identified, our focus shifted to detecting potential dihydrouridylation activity of rRNA by YhiN. In our initial endeavors, we aimed to reconstitute in vitro activity using rRNA isolated from an E. coli strain with the yhiN deletion. This involved coincubating the rRNA with recombinant YhiN and an excess of NADH. Subsequently, the generated rRNA product was enzymatically digested to its nucleoside level, and the presence of D was analyzed using LC/MS. Despite our concerted efforts, definitive detection of activity proved challenging, primarily due to notable contamination of D most probably originating from tRNA (SI Appendix, Fig. S6). To overcome this challenge, we designed and produced an in vitro transcribed 40-mer RNA (Mini-PTC-U), encompassing nucleotides 2438 to 2459 and mimicking stems and loops of E. coli’s 23S rRNA (Fig. 4F), serving as a refined substrate in subsequent activity assays. Our results unequivocally established that YhiN dihydrouridylates this mini-RNA, strategically mimicking a segment of the ribosomal peptidyl-transferase center (Fig. 4G). The D level increased with prolonged incubation time (Fig. 4G) and, after 24 h of incubation, 0.04% of D per quantified adenosine was formed (%mod/A). As the synthetic RNA contains 10 adenosines, the number of D per molecule is then 0.4%. However, this artificial substrate might adopt alternative structures prone to dihydrouridylation of other uridines found in this Mini-PTC as a much lower, but detectable, activity was measured when changing the target uridine by a cytidine (Mini-PTC-C in Fig. 4 F and G). Nonetheless, these findings conclusively position YhiN as the dihydrouridine synthase for E. coli’s ribosome, catalyzing the reduction of U2449 to D2449, leveraging FAD as a redox coenzyme and NADH as a source of reducing equivalents.

Predicted RdsA Are widespread in Bacteria and Members of a Nonisofunctional Superfamily.

The E. coli YhiN protein is a member of the COG2081 and PF03486 (HI0933-like protein) families. An initial analysis of the family in the EggNog database (http://eggnog6.embl.de/search/ogs/COG2081/) suggested that it was not isofunctional. Indeed, the 9,168 proteins listed in this family are from 6,723 species, hence 2,245 are paralogs (Dataset S1). The only two experimentally characterized protein in this family are 3-dehydro-bile acid delta(4,6)-reductase (BaiN) from Clostridium scindens involved in bile degradation (27) and aminoacetone oxidase (AAO) from Streptococcus cristatus, involved in resistance to endogenous metabolites-generated ROS and mutagens (28, 29). Hence, the RdsA activity of YhiN must be shared by only a subset of the COG2081/PF03486 members. To identify the RdsA subgroup more precisely, we constructed a sequence similarity networks (SSN) of PF03486 with different alignment score thresholds (AST, see SI Appendix, Supplementary Materials and Methods). An AST of 80 allowed the separation of the cluster containing YhiN_Ec from the cluster containing bonafide BaiN (indicated by a yellow line in Fig. 5A) and the cluster containing AAO_So (indicated by a green line in Fig. 5A). Here, we analyzed the rRNA of Vibrio cholerae species by AlkalinineSeq and showed that its 23S rRNA indeed harbors D2432 while carrying a RdsA (SI Appendix, Fig. S7). These bacteria do encode proteins highly similar to YhiNEc (cyan nodes in Fig. 5A), supporting their function as RdsA enzymes. In contrast, B. subtilis 168 rRNA does not show any D residues (1), but its genome does encode a PF03486 family protein (YtfP or BSU30060). This protein is not located in the same SSN cluster as YhiNEc (Fig. 5A) and the gene is in an operon with the opuD gene encoding the glycine betaine/arsenobetaine transporter (Fig. 5A and SI Appendix, Fig. S8A). Most Acinetobacter species encode two members of the PF03486 family (Dataset S2). These separate nicely into two groups in the SSN (Fig. 5A). Group 1 proteins (YhiN1) are part of the YhiNEc cluster and hence would be predicted to catalyze the formation of D in rRNA. Group 2 proteins (YhiN2) are part of separate cluster, but the corresponding genes are located downstream trmB genes encoding tRNA (guanine(46)-N(7))-methyltransferase) (Fig. 5B) still indicating a possible link to RNA modifications for this YhiN2 group. Other physical association with RNA modifications were observed: yhiN homolog are located next to dusA genes in organisms such as Synechococcus elongatus (yellow nodes in Fig. 5 and SI Appendix, Fig. S9), or to genes encoding ribosomal pseudouridine synthase RsuA-like proteins (orange nodes in Fig. 5 and SI Appendix, Fig. S8B). In summary, the COG2081/PF03486 family is clearly not isofunctional and, because gene neighborhood information does not give definitive functional clues, further experimental characterization of the proteins from the different subgroups is required to establish whether they harbor RdsA activity.

Fig. 5.

Protein SSN of PF03486 family proteins. (A) SSN of 7,930 PF03486 family proteins with the cutoff of alignment score of 80. Each node in the network represents one or multiple PF03486 family proteins that share 90% or more identity. An edge (represented as a line) is drawn between two nodes with a BLAST E-value cutoff of better than 10^–80. The nodes are colored based on taxonomy or the gene neighborhood context of the corresponding genomes. The cluster containing YihN_Ec is boxed. (B) Examples of gene neighborhoods organization. Uniprot IDs of PF03486 proteins: C. scindens BaiN, B0NAQ4; S. cristatus AAO, E8JVH0; B. subtilis YhiN, Q795R8; Acinetobacter calcoaceticus YhiN1, A0A0A8XKZ3; A. calcoaceticus YhiN2, A0A446ZNN4; Peptococcaceae bacterium YhiN, A0A942G053; Synechococcales bacterium YhiN, A0A354AWX7.

To survey the taxonomic distribution of the PF03486 proteins conservatively predicted to be isofunctional with YhiNEc, we further analyzed the proteins within the same cluster (boxed in Fig. 5A and Dataset S3) in an SSN with a more stringent alignment score threshold (SI Appendix, Fig. S9) and constructed a phylogenetic tree of the same protein set (SI Appendix, Fig. S10). These analyses show that members of the YhiNEc subgroup are mostly found in bacteria with only two eukaryotic homologs from dinoflagellates (Uniprot IDs: A0A812V8P8 and A0A9P1GL58) and a small group found in Archaea (mainly Methanobacteriales and Methanomicrobiales, Dataset S3). The protein tree diverges from the species tree in several places (for example, the inclusion of Thermodesulfobacteriota in the middle of the gammaproteobacteria proteins) suggesting horizontal gene transfer events. The gene neighborhood information does not link YhiN to translation or RNA modifications as genes encoding universal stress protein, transporters, and RNA helicase-like proteins are mostly found surrounding these sets of YhiN encoding genes (SI Appendix, Fig. S10).

Discussion

In the course of this investigation, we uncovered a unique dihydrouridine synthase, referred to as RdsA, responsible for catalyzing the NADH and FAD-dependent reduction of U2449 in the 23S rRNA of E. coli. Traditional Dus enzymes, encompassing bacterial DusA, B, C, and eukaryotic Dus 1 to 4, also operate as flavoenzymes but rely on FMN and NADPH (4). This parallels their counterparts, dihydropyrimidine deshydrogenase (DHPD) (30), and dihydroorotate deshydrogenases (DODH) (31), which facilitate the reduction of uracil to dihydrouracil and orotate to dihydroorotate (a uracil derivative), respectively. This implies that nature utilizes reduced flavin for uridine reduction. However, in the tertiary structure generated by AlphaFold, RdsA distinguishes itself from these FMN-dependent flavoenzyme systems due to its predicted distinct structure, displaying a unique topology. Unlike Dus, DHPD, and DODH, where the catalytic site binding FMN is carried by a TIM BARREL, RdsA’s flavin is held by a Rossman fold domain (FBD) housing two additional inserted domains (ins1D and ins2D). Surface charge representation on the AlphaFold model unveils several patches of positive charges concentrated on the three domains, arranging around the active site pocket and likely defining the RNA binding site (SI Appendix, Fig. S11). This model, if proved to be correct, would suggest that FBD, ins1D, and ins2D probably play a role in substrate recognition. RNA modifications follow a nonrandom, stepwise progression at distinct stages during ribosome biogenesis. Despite the complexity of this process, our observation of dihydrouridylation activity on a mini-ARN containing a small PTC-like region by RdsA should imply a potential role for this enzyme in acting on a ribosome precursor.

Phylogenetic analysis reveals the widespread distribution of YhiN among bacteria, implying the presence of D in the rRNA of numerous bacterial species. Indeed, our analyses of V. cholerae rRNA demonstrate the presence of D2432 in this organism (corresponding to the D2449 in E. coli), and furthermore, this species harbors a gene encoding RdsA. Nevertheless, our study brings attention to a perplexing observation: the absence of YhiN in the genome of C. sporogenes, despite the detection of m5D2449 and D2500 in its 23S rRNA (12). This suggests the potential existence of another unidentified type of Dus. In contrast, B. subtilis harbors a yhiN gene, yet its rRNA modification profile does not show any D residues (1), hinting at potential alternative biological functions for this flavoprotein. This proposition gains support from our findings, which reveal that RdsA belongs to a nonisofunctional superfamily. This is further underscored by the identification of a homologous gene to E. coli yhiN, operating as an aminoacetone oxidase in Streptococcus oligofermentas (28), albeit with measured activities that are notably weak.

Our genetic studies show that a functional RdsA enzyme is not essential for E. coli survival, and this is in keeping with the majority of rRNA modification enzymes. However, the biological role of ribosomal D remains elusive. Multiple studies have demonstrated that the lack of rRNA modifications can result in changes of ribosomal active sites (32), leading to a reduction in translation speed, an increase in mRNA reading errors by the ribosome (33), reduced virulence in pathogenic bacterial species, and a disrupted response to metabolites and antibiotics (34, 35). Limited to microbes, including certain pathogens, our research lays the groundwork for exploring the impact of dihydrouridylation on bacterial ribosome maturation and the potential targeting of this process by inhibitors.

Materials and Methods

Strains Media and Oligonucleotides Used in This Study.

All strains are listed in SI Appendix, Table S3. BW25113 were obtained from the Keio collection, and genomic deletion strains were obtained from NBRP (Japan). Cells were usually grown in LB medium at 37 °C. Oligonucleotides used for RhoRT-PCR and MALDI-TOF analysis are listed in SI Appendix, Table S4.

Rhodamine Labeling, Reverse Transcription, and PCR.

The method for rhodamine-110 labeling was derived from the one described in ref. 7. From the RNA stock, 50 μg was extracted, to which 16 μL of 1 M Tris HCl pH 7.5 and 40 μL of 100 mg/mL NaBH4 diluted in 10 mM KOH were added, reaching a final volume of 400 µL. After 1 h of incubation at room temperature (open tubes in the dark), the reaction was neutralized by adding 80 µL of 6 M acetic acid. The treated RNAs were then precipitated with 0.3 M NaCl and two volumes of 100% ethanol. The rRNAs were subsequently resuspended in 550 µL of 0.1 M formate buffer pH 3.15, containing 0.5 mM Rhodamine 110, and incubated at 37 °C for 1.5 h for fluorochrome conjugation. The reaction was stopped by adding 0.2 M Tris-HCl, pH 8.5, and the excess fluorochrome was removed by acidic phenol extraction (pH 4.3). Finally, the RNA was precipitated and resuspended in RNase-free H₂O. Reverse transcription and PCR conditions are detailed in SI Appendix.

Analysis of Ribosomal Dihydrouridylation at Position 2449.

Dihydrouridine quantification was done by tandem LC–MS as detailed in supplementary information. Ribosomal modification profiles were analyzed by the AlkAniline-Seq method (36). This method exploits the instability of the D-ring under alkaline conditions leading to its cleavage and formation of β-ureidopropionic acid (21). This instability leads to 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.

Biochemical Characterization of RdsA.

E. coli yhiN gene was cloned in pET15b for production and purification in BL21-DE3 E. coli cells. The NADH oxidase activity was determined by following decrease of A₃₄₀ under aerobic conditions and dihydrouridylation assays were performed by incubating the recombinant RdsA with the target RNA in a buffer containing 50 mM HEPES pH 7.5, 150 mM NaCl, 5 mM DTT, 10 mM MgCl2, 100 µM FAD, and 15% glycerol. The detailed procedure for LC–MS analysis of the reaction is detailed in SI Appendix.

General Bioinformatic resources.

For literature and sequence retrievals, the resources at NCBI (https://www.ncbi.nlm.nih.gov/) (37), UniProt (https://www.uniprot.org) (38), EggNog (http://eggnog6.embl.de/) (39), and BV-BRC (https://www.bv-brc.org) (40) were routinely used. PaperBlast was used to find published papers on members of the COG2018 family (papers.genomics.lbl.gov/) (38). The SubtiWiki/CoreWiki (http://corewiki.uni-goettingen.de/) (41), the String database (https://string-db.org/) (42) and the EFI Gene Neighborhood webtool (43) were used to explore physical clustering. Protein sequences were aligned using MUSCLE v5.1 (44) and visualized using Weblogo3 (45). SSN and phylogenetic analysis are detailed in supplementary information.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. All NGS data associated to this manuscript are deposited and made publicly available in ENA (The European Bioinformatics Institute EMBL-EBI) under the accession number PRJEB77571 (46).