RlmQ: a newly discovered rRNA modification enzyme bridging RNA modification and virulence traits in Staphylococcus aureus

S. aureus 23S rRNA contains a 7-methylguanosine at position 2601

Given the important roles of RNA modifications on bacterial physiology, we wondered whether the Gram-positive S. aureus bacteria harbor different sets of RNA modifications compared to the evolutionary distant Gram-negative bacteria E. coli. Here, we have focused the study on the modifications present in S. aureus 23S rRNA. After purification of the S. aureus 23S rRNA from the HG001 strain and isolation of specific fragments by RNase H cleavage, the identification of RNA modifications was done using RNase T1 digest followed by LC/MSMS analysis. Among the various sequenced oligoribonucleotides, a fragment with a mass of 1948.2 Da (m/z 973.62) corresponded to the UAC[mG]CG > p sequence. Subsequent MSMS fragmentation revealed the presence of a methylation at G2601 located in H90 (corresponding to E. coli G2574; Fig. 1A,B). To visualize the presence of the methylation and its position on the guanosine, we conducted an in-depth analysis of the high-resolution structure of S. aureus ribosome (Halfon et al. 2019) (map: EMD-10077; PDB: 6s0z; Fig. 1C). Within this map, we identified a density that extends into position 7 within the purine ring, in favor of the presence of a 7-methylguanosine (m⁷G), not modeled. Finally, the presence of m⁷G was validated by primer extension (PE) analysis, adapting a specialized protocol originally designed for genome-wide mapping of m⁷G (Marchand et al. 2018). The experiments showed a strong band corresponding to a specific termination of reverse transcription (RT) due to the m⁷G2601 in both the virulence-attenuated HG001 and the virulent USA300 (LAC) strains (Fig. 1D). Interestingly, this modified position has also been reported in B. subtilis (Hansen et al. 2002) suggesting that m⁷G2601 in the helix H90 is conserved among firmicutes. E. coli ribosome structure suggests that H90 serves as a corridor to facilitate the accommodation of the aminoacylated CCA-end of tRNA (aa-CCA) into the A-site of the peptidyl transferase center (PTC) (Fig. 1E; Whitford et al. 2010). Although the resolution of the cryo-EM structure of S. aureus 70S ribosome is lower than E. coli 70S and no A-site tRNA is present (Golubev et al. 2020), the positioning and structures of H90 and neighbor rRNA helices are highly similar. The accommodation process involves key conformational changes in both tRNA and ribosome that significantly impact the rate of cognate tRNA selection (Rodnina and Wintermeyer 2001). A kinetic study combined with dynamic simulation has revealed the relatively slow nature of this process due to spontaneous and reversible accommodation attempts, akin to a stochastic trial-and-error mechanism, encompassing various parallel pathways (Whitford et al. 2010). Notably, the corridor traversed by the aa-CCA of the tRNA is nearby m⁷G2601 in S. aureus. Hence, it is tempting to propose that this methylation might be one of the key players which facilitates the accommodation process. More experiments will be required to assess the impact of the m⁷G modification on translation dynamics, especially its selectivity and functional importance for different aminoacyl-tRNAs.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

S. aureus ribosome contains m⁷G2601. (A) Mass spectrometry analysis of the RNase T1 derived fragment of S. aureus HG001 23S rRNA region spanning from U2598 to G2603 (corresponding to E. coli U2571 to G2576). The fragment has a mass of 1948.2 Da (m/z 973.62) and carries the sequence UAC[mG]CG > p based on MSMS analysis. (B) 2D representation of the entire S. aureus 23S rRNA, zooming on the region containing helix H90 with G2601 (corresponding to E. coli G2574). Red and black numbers indicate S. aureus and E. coli numbering, respectively. The red bar highlights the fragment analyzed by LC/MSMS. (C) Structural analysis of G2601 from the deposited S. aureus 50S cryo-EM map at 2.3 Å (Halfon et al. 2019) (map: EMD-10077; PDB: 6s0z). The density around N7 indicates the presence of the methyl group, which is not modeled in the atomic coordinates. (D) PE analysis of HG001 and USA300 (LAC) S. aureus strains, including controls for the AlkAlanine treatment. AGCU lanes represent dideoxy sequencing, NT stands for the nontreated sample, and the red arrow points to the reverse transcriptase stop due to the m⁷G-modified nucleotide. (E) Details of the accommodation corridor for aminoacyl-tRNA CCA-end on the E. coli ribosome structure, with the A-site tRNA (gray surface) in the accommodated state (Watson et al. 2020). The E. coli 23S rRNA helices (H69–H71, H90–H91) are shown: G2524 (red sphere), corresponding to m⁷G2601 in S. aureus, is located in H90 (green), while m⁵C1962 (yellow sphere), corresponding to C1989 in S. aureus, is in H70/H71 (bright orange). The mRNA (blue transparent surface) in the 30S decoding channel (DC) interacts with the anticodon of the A-site tRNA. (Sau) S. aureus, (Eco) E. coli.

In quest of the enzyme responsible for m⁷G2601

To identify the enzyme responsible for m⁷G2601 in S. aureus, we have first performed a bioinformatic analysis, primarily focusing on sequence similarities through BlastP and sequence alignments. Our approach has involved three representative methyltransferases from E. coli, S. aureus, and B. subtilis, which are known for their methylation activity on 16S rRNA (RsmG for m⁷G967—E. coli numbering), 23S rRNA (the carboxy-terminal half of E. coli RlmKL corresponding to B. subtilis RlmK, for m⁷G2069—E. coli numbering), and tRNA (TrmB for m⁷G46 in the variable loop of different tRNAs). We utilized BlastP with the sequences of the catalytic S-adenosyl-methionine (AdoMet) binding domains (MTase) from the selected m⁷G methylases to search for similar proteins in S. aureus NCTC8325, which is highly similar to the HG001 strain. The analysis revealed many members of the COG1092 methyltransferase family, encompassing m⁵C MTases and m⁷G MTases, due to the high similarity of their catalytic domains. Among the identified candidates in S. aureus, only seven are predicted enzymes with unknown functions. To ascertain whether one of these enzymes was responsible for the presence of m⁷G2601, we systematically examined the PE profile of 23S rRNA prepared from the mutant strains carrying individual gene disruption for each candidate (Fig. 2A). These strains were selected from the Nebraska Transposon Mutant Library collection and compared with the isogenic wild-type (WT) USA300 JE2 strain (Fey et al. 2013). The PE analysis showed that only the mutation at the locus SAUSA300_0981 in USA300 (HG001_00931 in HG001 strain) causes the loss of the specific RT pause at G2601.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

RlmQ is a m⁷G methyltransferase responsible for the enzymatic modification of m⁷G2601. (A) AlkAniline PE analysis of S. aureus 23S rRNA from USA300 JE2 WT and seven selected strains from the Nebraska collection with disrupted genes for methyltransferases of unknown function. AGCU dideoxy sequencing lanes. (Left) Lane 1, USA300 WT; lane 2, NE553; lane 3, NE1671; lane 4, NE1442; lane 5, NE189; lane 6, NE807; lane 7, NE1546 (corresponding to rlmQ-disruption mutant); lane 8, NE1337. (Right) Analysis of m⁷G restoration in the complemented mutant strain. Lane 1, S. aureus USA300 Lac WT-pCN38:wP:TT; lane 2, NE1546-pCN38:wP:TT; lane 3, NEB1546-pCN38:wP:rlmQ:TT. Red arrows indicate m⁷G modification. (B) Sequence alignment of S. aureus RlmQ and E. coli RlmI. Distinct domains are highlighted with colored bars as follows, starting from the amino terminus: PUA domain (blue bar over the sequence), EEHEE domain (magenta bar), β-hairpin (green bar), and MTase domain (red bar). Colored boxes spotlight the most significative signatures discriminating RlmQ and RlmI enzymes, as determined by BIS² analysis (Dib and Carbone 2012). Red boxes represent cluster 1, with a P-value cutoff at 4.3 × 10⁻⁶, revealing the substitution of the conserved catalytic cysteine, typically found in m⁵C methyltransferases, by asparagine in RlmQ. Within this cluster, another substitution (His/Phe) correlates with the former, contributing to the distinctive motifs for the two classes of methyltransferases. Blue boxes denote cluster 2, with a P-value cutoff at 2.6 × 10⁻⁵, encompassing Lys/Met, Pro/Gly, and Arg/Trp covariations in RlmI/RlmQ, respectively. The brown box indicates cluster 3, with a P-value cutoff at 3.7 × 10⁻⁵, associating the fragment sequences DRSDVAVRKK/EKVRFK (RlmI/RlmQ) with the covariations of cluster 1. (C) Structure analysis illustrating the specific covariation clusters characterizing RlmQ (on the right) and RlmI (on the left). The structures of RlmQ and RlmI are sourced from the pdb 3VSE (Kita et al. 2013) and pdb 3C0K, respectively. The domains and motifs in the figure follow the same color code as in B. At the bottom, a comparison of the MTase active sites is presented for S. aureus RlmQ (orange background structure) and E. coli RlmI (cyan background structure). In the motif FSCSG, E. coli RlmI contains a cysteine (Cys), whereas S. aureus RlmQ has an asparagine (Asn) in the motif CTNAS. To facilitate comparison, the substrate S-adenosyl-l-homocysteine (SAH), present only in the RlmQ crystal structure, has also been modeled in the RlmI structure.

To ensure that the differences observed were solely attributable to SAUSA300_0981 disruption, we sequenced and analyzed the genomes of both WT and the mutant strains. The presence of the transposon at position 451 of the SAUSA300_0981 gene sequence (1173 pb long) was confirmed and no other sequence change in the genome was detected. At the point of insertion, a stop codon in the frame with the coding region was generated (Supplemental Fig. 1) preventing the translation of full-length protein. Through RT-PCR, we demonstrated that the transposon insertion led to the synthesis of a shorter transcript, lacking the sequence encoding the catalytic MTase domain (Supplemental Fig. 1). This strongly suggests that the mutation cannot lead to the synthesis of an active enzyme. Subsequently, we conducted a complementation of the mutant strain by transforming it with a plasmid expressing the WT SAUSA300_0981 gene under the control of a constitutive mild promoter. A mass spectrometry analysis of total proteins from the WT, the mutant strain, and the complemented strain confirmed the absence of spectra for the SAUSA300_0981 protein in the mutant. In contrast, several spectra corresponding to different regions of the protein were observed in the WT and complemented strains (Supplemental Fig. 1).

The PE analysis showed that the premature RT stop was restored in the complemented strain, as shown for the WT JE2 strain (Fig. 2A). All in all, these experiments show that SAUSA300_0981 encodes the methyltransferase responsible for the 7-methylation of G2601. In accordance with the established bacterial nomenclature for rRNA methyltransferases, we have named this newly identified methyltransferase “RlmQ.”

S. aureus RlmQ and E. coli RlmI methyltransferases are paralogues with distinct activities

Based on its amino acid sequence, HG001_00931 (or SAUSA300_0981 in USA300) gene locus was previously annotated as RlmI (Caldelari et al. 2017), the RNA methyltransferase responsible for m⁵C1962 in 23S rRNA from E. coli (corresponding to S. aureus C1989) and various bacteria (Purta et al. 2008). Indeed, they share 46% sequence similarity and signatures for similar domain organization (Fig. 2B). To perform a more detailed analysis of the similarities and differences between these two 23S rRNA methyltransferases, we conducted an extensive search in the PDB database using RlmQ as the reference (Holm et al. 2023). Remarkably, we identified the crystal structure of an S. aureus methyltransferase named SAV1081 (Kita et al. 2013), with no attributed function but which displayed 100% coverage and sequence similarity to RlmQ. This structure displays three main domains: a Rossman-Fold methyltransferase domain (MTase), the methyltransferase conserved RNA recognition module EEHEE domain, and the RNA binding domain PUA, which is responsible for the recognition of specific RNA sequences (Fig. 2C; Sunita et al. 2008). The EEHEE and MTase domains are interconnected by a conserved β-hairpin structure, a feature shared with several other methyltransferases. Superposition of E. coli RlmI and S. aureus RlmQ structures (pdb files 3VSE for RlmQ and 3C0K for RlmI) shows a root mean square deviation (RMSD) of 1.45 Å and a TM-score of 0.4254 (Zhang and Skolnick 2004), indicating a strong overall structural similarity. Both enzymes belong to the COG1092 family, featuring a well-conserved AdoMet binding site in the MTase domain. This domain shares 28.5% of sequence identity, 45.2% of sequence similarity, and a TM-score of 0.2518 between the two enzymes. Additionally, their EEHEE domains exhibit a high degree of conservation with 32.7% sequence identity and 54.8% sequence similarity. Structure comparison also reveals similarity (TM-score = 0.321). In contrast, their amino-terminal PUA domains display the most pronounced sequence and structural divergence with 21.3% sequence identity and 41.3% sequence similarity. The TM-score for this region is 0.2518. Within the MTase active site, the principal distinction arises from a specific amino acid substitution: a catalytic cysteine (C) in RlmI is substituted by an asparagine (N) in RlmQ (Fig. 2C). The presence of this cysteine in eubacteria, archaea, and certain plants (Kita et al. 2013) confers m⁵C RNA methyl transferase activity (Liu and Santi 2000; King and Redman 2002), while the presence of an asparagine has been found in a separate subgroup of the COG1092 family in methylases in several Firmicute members (Sunita et al. 2008). To find sequence and structural motifs characteristic of RlmQ and RlmI enzymes, we searched for coevolution signals among pairs of amino acids or protein sequence fragments using the BIS² software (Dib and Carbone 2012). We initially obtained homologous sequences from distinct bacterial taxa for RlmQ and RlmI independently. This retrieval was performed from the manually annotated section of UniprotKB using psiblast (Altschul et al. 1997) at the website (https://www.ebi.ac.uk/Tools/sss/psiblast/) with default setting. We subsequently conducted a multiple sequence alignment (MSA) on a representative collection of 19 homologous protein sequences (Supplemental Fig. 2A) taken from different taxa and used it for BIS² analysis. The analysis, allowing a maximum of one exception among the sequences, has identified different coevolution clusters which have been mapped on the sequence alignment (Fig. 2B; Supplemental Fig. 2A), and on RlmQ and RlmI structures (Fig. 2C). The most significant cluster (cluster 1, with a P-value cutoff at 4.3 × 10⁻⁶) provides additional evidence for the species-specificity of Asn (RlmQ) and Cys (RlmI) residues, together with their coevolved cluster partners, the structurally close Phe (RlmQ) and His (RlmI) residues within their MTase domain. In addition, a slight reduction in the significance cutoff (P-value cutoff at 3.7 × 10⁻⁵) revealed correlated variations in the PUA, EEHEE, and β-hairpin domains as well (Fig. 2B,C; Supplemental Fig. 2A). Our analysis also revealed two intriguing exceptions. In the γ-proteobacterium Pseudomonas aeruginosa, both RlmQ and RlmI modification enzymes coexist. Whether both modifications (m⁵C and m⁷G) could occur on the same 23S rRNA remained to be demonstrated. Among the Firmicute phylum, Clostridium perfringens has a methylase enzyme carrying all the peculiarities of the RlmI class.

To investigate the potential involvement of RlmQ in the modification of m⁵C1989 (E. coli C1962), we conducted a meticulous analysis using mass spectrometry on the relevant 23S rRNA region from S. aureus. We then compared it with the corresponding region of E. coli 23S rRNA, where modification m⁵C1962 by RlmI was anticipated (Fig. 3A). The 23S rRNA, isolated from S. aureus HG001 strain and E. coli MRE600 strain, was first hybridized to two specific oligonucleotides and digested with RNase H. This treatment led to specific RNA fragments of 25 nt, which include rRNA helices H70 and H71 and the expected target cytosine (Fig. 3B). After gel electrophoresis purification, the gel bands containing the 25 nt long RNA fragments were further treated with RNase T1. Subsequent LC/MSMS analysis of S. aureus sample revealed the presence of a short fragment with a mass of 1606.21 Da corresponding to the unmodified sequence ACCCGp (Fig. 3A, upper panel). In contrast, for E. coli, we detected a fragment of 1621.2 Da corresponding to the modified sequence AC[m5C]UGp (Fig. 3A, lower panel). Comparison of the density maps from the reported high-resolution structures of E. coli and S. aureus ribosomes reveals clearly the presence of a methyl group at position 5 of C1962 in E. coli 23S rRNA, while this methyl is likely absent at position 5 of the equivalent C1989 in S. aureus (Fig. 3C).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

RlmQ/RlmI evolution changes 23S methylation maps in S. aureus and E. coli. (A) Detection of nonmodified C1989 in S. aureus (upper panel) and m⁵C1962 in E. coli (lower panel) using mass spectrometry. The RNase H generated fragment including H70 and H71 of the 23S rRNAs has been further digested with RNase T1 before analysis. S. aureus MSMS spectrum (upper panel) contains three fragments including the sequence of interest ACCCGp showing no methylation. E. coli MSMS spectrum (lower panel) contains only the methylated fragment of interest AC[m⁵C]UGp. (B) 2D representation of the entire S. aureus 23S rRNA, zooming on the region containing H70 and H71, with C1962 (E. coli numbering, corresponding to S. aureus C1989). Red and black numbers indicate S. aureus and E. coli numbering, respectively. The red bar highlights the fragment analyzed by LC/MSMS. (C) Structure analysis of C1962 from the deposited S. aureus 50S cryo-EM map at 2.3 Å (Halfon et al. 2019) (map: EMD-10077; PDB: 6s0z) and E. coli deposited 70S cryo-EM map at 2 Å (Watson et al. 2020) (map: EMD-22586; PDB: 7k00). While a clear density for methylation of position 5 on C1962 is observed in E. coli 23S allowing building this modification in the fitted model, no extra density is present on S. aureus 23S.

Taken together, we show in vivo that S. aureus RlmQ methylates the N7 position of G2601 (equivalent to G2574 in E. coli), while C1989 is not modified (equivalent to C1962 in E. coli). Because this methylation has also been observed in B. subtilis, we propose that Firmicutes share the same features. Conversely, in Gram-negative bacteria, RlmI is the enzyme which methylates C1962 at its position 5, while no existing modification was attributed to G2574. It is conceivable that these two paralogous enzymes may have originated from a gene duplication event and divergent evolution in these distant species (Supplemental Fig. 2B). Interestingly, E. coli m⁵C1962 and S. aureus m⁷G2601 are both situated along the pathway taken by tRNAs during their transit into the ribosome (Fig. 1E; Watson et al. 2020). In this scenario, the tRNAs do not make direct contact with helices H70 and H71, although these helices are in close proximity to the tRNA acceptor arms and variable loops. It is plausible that Gram-negative and Gram-positive ribosomes have evolved slightly different solutions to facilitate the smooth transition of tRNAs within the ribosome, both involving rRNA methylation, albeit targeting distinct residues.

Phenotypic characterization of S. aureus rlmQ mutant

S. aureus is an opportunistic pathogen, which has evolved many strategies to regulate the synthesis of numerous virulence factors in response to the host, stress, and various environmental changes. To assess the impact of RlmQ enzyme modification on the physiology of S. aureus, the WT USA300 JE2 strain was compared to a mutant deficient for the rlmQ enzyme (rlmQ⁻) in various phenotypic tests including growth rate, biofilm, and cytotoxicity on human cells (Fig. 4).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Phenotypic characterization of S. aureus rlmQ mutant. (A–C) Growth of the WT strain and the mutant inactivated for rlmQ (rlmQ⁻) was compared under different media and temperature conditions. Bacterial growth was measured during 24 h in BHI medium at 37°C (A), 42°C (B), and in RPMI medium at 37°C (C), n = 2–6. (D) Biofilm was quantified by crystal violet staining of the extracellular matrix after 24 h of culture in BHI medium, n = 8. (E) Bacterial supernatant cytotoxicity was quantified by measuring propidium iodide (PI) incorporation into C5aR1-transfected U937 monocytes, n = 7. Mann–Whitney tests performed: (*) P-value < 0.05, (**) P-value < 0.01, (***) P-value < 0.001. (F) Mutation rate of protein sequences from 301 clinical strains was estimated with a ratio: number of missense mutations divided by sequence match length.

We first analyzed whether the mutation had an effect on S. aureus growth. Under optimal growth conditions (BHI medium, at 30°C and 37°C), no difference was observed between the WT and the rlmQ⁻ mutant strains (Fig. 4A; Supplemental Fig. 3). This is in agreement with the fact that the polysome profiling in the WT and mutant strains remained identical in the two strains under these conditions of growth (Supplemental Fig. 4). These experiments also show that the maturation process of the ribosome is not significantly affected by the mutation. Only at 42°C in BHI, the rlmQ⁻ mutant reached a slightly higher OD at the stationary phase than the WT strain (Fig. 4B). The two strains were cultured in a less rich environment (RPMI medium) closer to the conditions encountered during infection. Under these conditions, we observed a significant reduction in the growth of the mutant strain compared to the WT strain with a reduced doubling time at the exponential phase (rlmQ⁻ 65 min/WT 45 min), and a diminution of the maximal OD (rlmQ⁻ 0.38/WT 0.56) (Fig. 4C). We then monitored the ability of S. aureus WT and mutant strains to form biofilm which play major roles in an infectious context. The amount of biofilm formed on plates by the rlmQ⁻ mutant was significantly greater than that of the WT strain (Fig. 4D). Another important feature of bacterial–host interaction in infectious conditions is the cytotoxic activity of S. aureus against immune cells. The cytotoxic activity of the WT and mutant strains was assessed on human monocytes expressing the C5aR1 receptor targeted by two pore-forming toxins, the Panton–Valentine leucocidin (PVL) and the γ toxin CB (HlgCB). Monocytes were exposed to overnight culture supernatants (which contain bacterial toxins but no live bacteria). The cytotoxic activity of these supernatants was assessed by the incorporation of PI into the DNA of cells with altered plasma membranes. The rlmQ⁻ mutant showed a significant reduction in cytotoxic activity compared to the WT strain (Fig. 4E).

We finally addressed whether RlmQ is conserved among clinical infection isolates. To this end, the rlmQ gene was searched in the genomes of a collection of 301 well-characterized clinical S. aureus strains from previously published cohorts. The data showed that the rlmQ gene was detected in all 301 strains and had the same size in 298 strains (Fig. 4F). Furthermore, the protein sequence was extremely well conserved with a mutation rate of 0.008 comparable to the highly conserved GyrB (MR = 0.002) but much lower than that of coagulase (MR = 0.263), which is known to be highly variable among strains. Using a Random Forest model based on horizontal oligonucleotide frequency, a conserved promoter sequence was predicted between −432 and −475 nt upstream of the initiator codon for the 301 strains (Supplemental Fig. 5).

Taken together, our data suggest that the highly conserved RlmQ methylase is playing an important role in human infections. Slower growth and increased biofilm formation are indicative of a chronic infection mode, and conversely, toxin production is a marker of acute infection. As RlmQ is associated with a better growth under physiological conditions, with an increase in cytotoxic activity, and a decrease in biofilm formation, we suggest that RlmQ might be important for the establishment of acute infection.