Widespread occurrence of N⁶-methyladenosine in bacterial mRNA

Abstract

N⁶-methyladenosine (m⁶A) is the most abundant internal modification in eukaryotic messenger RNA (mRNA). Recent discoveries of demethylases and specific binding proteins of m⁶A as well as m⁶A methylomes obtained in mammals, yeast and plants have revealed regulatory functions of this RNA modification. Although m⁶A is present in the ribosomal RNA of bacteria, its occurrence in mRNA still remains elusive. Here, we have employed ultra-high pressure liquid chromatography coupled with triple-quadrupole tandem mass spectrometry (UHPLC-QQQ-MS/MS) to calculate the m⁶A/A ratio in mRNA from a wide range of bacterial species, which demonstrates that m⁶A is an abundant mRNA modification in tested bacteria. Subsequent transcriptome-wide m⁶A profiling in Escherichia coli and Pseudomonas aeruginosa revealed a conserved m⁶A pattern that is distinct from those in eukaryotes. Most m⁶A peaks are located inside open reading frames and carry a unique consensus motif of GCCAU. Functional enrichment analysis of bacterial m⁶A peaks indicates that the majority of m⁶A-modified genes are associated with respiration, amino acids metabolism, stress response and small RNAs, suggesting potential functional roles of m⁶A in these pathways.

INTRODUCTION

N⁶-methyladenosine (m⁶A) is the most frequent internal mRNA modification in a wide range of eukaryotes and certain viral RNAs (1–6). Modification of m⁶A is mediated by an N⁶-adenosine methyltransferase complex including a 70-kD SAM (S-adenosylmethionine)-binding subunit methyltransferase like 3 (METTL3, also called MT-A70), methyltransferase like 14 (METTL14) and Wilms tumor 1 associated protein (WTAP) (7). METTL3 and METTL14 form a heterodimer that catalyzes m⁶A RNA methylation, while WTAP interacts with the complex and affects the mRNA methylation (8). These methyltransferases play important roles in mouse embryonic stem cell differentiation and circadian rhythms (9–11) Fat mass and obesity-associated protein (FTO) and alkB homolog 5 (ALKBH5) are m⁶A RNA demethylases, which are involved in mammalian development, RNA metabolism and fertility (12,13). The human YTH-domain family 2 (YTHDF2) has recently been shown to specifically bind m⁶A-modified mRNA and to promote the decay of the bound mRNA (14,15). These discoveries present the first examples of reversible RNA modification and reveal the unique regulatory functions of reversible m⁶A methylation on mRNA and non-coding RNAs.

Transcriptome-wide profiling of m⁶A distributions in mammals and yeast further characterize the dynamic nature of m⁶A modification, which is enriched around the stop codon and at 3′ UTRs, as well as in long internal exons and at the transcription start site (11,15,16). m⁶A in Arabidopsis thaliana is also enriched around the stop codon, 3′ UTRs and around the start codon (17). The majority of m⁶A-peaks in these organisms harbor a consensus motif of RRACU (R = A/G).

Although m⁶A has been well documented in the rRNA in bacteria, its presence on mRNA is still elusive. In Escherichia coli, A1618 and A2030 of 23S rRNA are methylated by methyltransferases RlmF and RlmJ, respectively (18,19). Both deletion and overexpression of rlmF result in a loss of cell fitness and growth defect (18), while an rlmJ mutant shows mild phenotypes under various growth conditions (19). Interestingly, the modifications of m²A or m⁸A on A1607, A2503 and A2508 play important roles in antibiotic resistance, an extensively studied subject in microbiology during the last 10 years (20).

In order to investigate the potential presence and functions of m⁶A in bacterial mRNA, we calculated the m⁶A/A ratio in mRNA from seven diverse bacterial species, which reveal that m⁶A is an abundant mRNA modification in Gram-negative bacteria. High-resolution transcriptome-wide m⁶A profiling in two model bacteria E. coli and Pseudomonas aeruginosa reveal a conserved and distinct m⁶A distribution pattern. Most m⁶A-modified genes are involved in energy metabolism and small RNAs, suggesting potential functional roles of m⁶A in these processes.

MATERIALS AND METHODS

Bacterial strains and mRNA purification

The strains and culture conditions used in this study are listed in Table 1. Total RNA was purified from bacterial pellets of 2-ml culture by using an RNeasy Mini Kit (Qiagen) that removes tRNA. Two micrograms of total RNA were applied to a MICROExpressTM Bacterial mRNA Enrichment Kit (Life technologies). A Ribo-Zero™ rRNA Removal Kit (Bacteria) (Epicentre) was used in order to further remove remaining rRNA. All procedures in the manufacturer's protocols were strictly followed. In order to verify the removal of rRNA, a qPCR (7300 Real-Time PCR System, Applied Biosystems) was done against the rRNA background in order to check relative enrichment levels. One nanogram of either total RNA or purified mRNA from E. coli was used per qPCR reaction (Power SYBR Real-Time PCR mater mix, Life technologies). The primers used were 5′-CTCCTACGGGAGGCAGCAG-3′ and 5′-GTATTACCGCGGCGGCTG-3′. The P. aeruginosa MPAO1 strain were cultured overnight at different temperatures (37, 40, 42 or 45°C) and then subjected to the mRNA purification protocol described above in the temperature variation studies.

Table 1. Strains and growth conditions.

Strain	Growth condition
Escherichia coli K-12 (CGSC)	LB, 37°C overnight
E. coli K-12 rlmJ mutant (CGSC)	LB, 37°C overnight
E. coli K-12 rlmF mutant (CGSC)	LB, 37°C overnight
E. coli K-12 ksgA mutant (CGSC)	LB, 37°C overnight
E. coli 5α	LB, 37°C overnight
E. coli XL-blue	LB, 37°C overnight
Pseudomonas aeruginosa MPAO1	LB, 37°C overnight
P. aeruginosa PA14	LB, 37°C overnight
Pseudomonas syringae pv. tomato DC3000	King's B medium, 28°C for 2 d
Staphylococcus aureus Newman	TSB medium, 37°C overnight
S. aureus USA100	TSB medium, 37°C overnight
S. aureus USA400	TSB medium, 37°C overnight
S. aureus USA700	TSB medium, 37°C overnight
S. aureus COL	TSB medium, 37°C overnight
S. aureus RN4220	TSB medium, 37°C overnight
Bacillus subtilis	LB, 37°C overnight
Anabaena sp. PCC 7120	Z8 medium, 25°C overnight
Synechocystis sp. PCC 6803	Z8 medium, 25°C overnight

Ultra-high pressure liquid chromatography coupled with triple-quadrupole tandem mass spectrometry (UHPLC-QQQ-MS/MS) analysis for m⁶A/A ratio

The highly purified bacterial mRNA was subjected to an UHPLC-QQQ-MS/MS (Agilent) analysis. Two hundred ng of mRNA or rRNA (on the beads of the mRNA Enrichment Kit) were digested by nuclease P1 (2 U) in 40 μl of nuclease buffer (25 mM of NaCl and 2.5 mM of ZnCl₂) at 37°C for 2 h, followed by the addition of NH₄HCO₃ (1 M, 2 μl) and alkaline phosphatase (0.5 U) at 37°C for 2 h. The nucleosides were separated by reverse phase ultra-performance liquid chromatography by a C18 column on an Agilent 6410 QQQ triple-quadrupole LC mass spectrometer in positive electrospray ionization mode. The nucleosides were quantified using the nucleoside-to-base ion mass transitions of 282 to 150 (m⁶A), 294 to 164 (m⁶₂A) and 268 to 136 (A). Quantification was performed by comparison with the standard curve obtained from pure nucleoside standards. Three biological repeats have been performed for all bacterial strains.

High-throughput and high-resolution m⁶A sequencing

Procedures were slightly modified from previously described protocols (21). In a 0.5-ml IP reaction, 5 μg purified bacterial mRNA and 15 μl of 0.5 mg/ml rabbit anti-m⁶A antibody (202003; Synaptic Systems) were incubated for 2 h at 4°C in IPP buffer (150 mM NaCl, 0.1% NP-40, 10 mM Tris–HCl, pH 7.4, 1 U/μl RNasin). The mixture was exposed to UV irradiation at 254 nm 3× (90 s each time), before RNase T1 (0.1 U/μl) digestion for 15 min at 22°C. After the digestion reaction was quenched on ice for 5 min, 200 μl pre-blocked protein A bead slurry was added into reaction for 1 h at 4°C. After washing thrice with IP wash buffer (50 mM HEPES-KOH, pH 7.5, 300 mM KCl, 0.05% NP-40, with proteinase inhibitor and RNasin), the beads were treated by a second round of RNase T1 digestion (15 U/μl) at 22°C for 15 min. The beads were cooled down on ice for 5 min and then thrice washed with high salt wash buffer (50 mM HEPES-KOH, pH 7.5, 500 mM KCl, 0.05% NP-40, with proteinase inhibitor and RNasin). The beads were then treated with Antarctic phosphatase (0.5 U/μl) for 20 min at 37°C. After dephosphorylation, beads were washed twice with phosphatase wash buffer (50 mM Tris–HCl, pH 7.5, 20 mM EGTA, 0.5% NP-40) and twice with PNK buffer without DTT (50 mM Tris–HCl, pH 7.5, 50 mM NaCl, 10 mM MgCl₂). Polynucleotide kinase (1 U/μl) and 200 μM adenosine triphosphate was then added to the beads at 37°C for 15 min. The RNA fragments were further purified by proteinase K digestion and TRIzol extraction. For IP samples, small RNA libraries were made by using NEBNext^® Small RNA Library Prep Set for Illumina^® (NEB). The input samples followed the above procedures without anti-m⁶A antibody pull-down and RNase T1 digestion. Libraries for input samples were made by using TruSeq RNA Sample Preparation Kits (Illumina, non-strand specific). Six libraries were constructed, containing one control sample and two duplicate IP samples for each strain.

Data analyses

All samples were sequenced using the HiSeq 2000 system (Illumina, with 50-bp and single end mode) at the Genomics Core Facility at the University of Chicago. FastQC was done to check the quality of each dataset; all datasets were obtained in high quality and can afford further reliable analyses. Two E. coli IP libraries obtained 1,718,364 mapped reads, and two P. aeruginosa libraries obtained 3,720,658 mapped reads. Sequence data were analyzed by following the procedures described previously (17). Briefly, Tophat (version 2.0.0, with the parameter: -p 8 –read-mismatches 2 –max-multihits 1) with Bowtie was run in order to align the input and IP-sequenced samples to the E. coli K-12 substr. MG1655 (ASM584v2, NC_000913.3) and P. aeruginosa PAO1 (ASM676v1, NC_002516.2) genomes and annotation files (22,23). In TopHat each read was only mapped to the genome once. The enriched peaks were identified using MACS software (version 2.0.0, with the parameter: callpeak -t ip.bam -c ck.bam -f BAM -g 6000000 –nomodel -n -p 1e-5) (24). Consensus sequence motifs were identified by using HOMER (version 4.7, with the parameter: -p 3 -rna -len 6) (25). A scrambled sequence was used as the background. Gene function analysis (GO enrichment) was performed with the online DAVID (version 6.7) tool (http://david.abcc.ncifcrf.gov/) (26). The m⁶A peaks were divided into three categories based on their relative positions in their corresponding genes: Overlap Start (±100 nucleotides around the start codon), Overlap End (±100 nucleotides around the stop codon) and Inside (other locations inside a coding region). The functional association with each gene was determined by NCBI annotation.

RESULTS

m⁶A is presented in mRNA of a wide range of bacterial species

Although m⁶A is the most abundant internal mRNA modification in eukaryotes, its potential presence in the kingdom of bacteria has yet to be investigated. To this end, we selected seven diverse model bacterial species (E. coli, P. aeruginosa, Pseudomonas syringae, Staphylococcus aureus, Bacillus subtilis, Anabaena sp. PCC 7120 and Synechocystis sp. PCC 6803) to grow in a common laboratory environment and measured their m⁶A/A ratios in purified mRNA. Unlike eukaryotes, bacterial mRNA lacks a poly(A) tail, which makes it challenging to purify mRNA. By following the protocols described in the Method section, we were able to remove >90% rRNA in the purified mRNA sample (Supplementary Figure S1).

We then tried to use an UHPLC-QQQ-MS/MS approach in order to quantify the m⁶A/A level in the bacterial mRNA samples that contain residual rRNA (<10%). Given that two m⁶A (catalyzed by RlmF and RlmJ) and two N⁶,N⁶-dimethladenosine (m⁶₂A, catalyzed by KsgA) are known to be present in rRNA of E. coli and other related bacterial species (18,19,27), the values of m⁶₂A levels can be used as an internal reference for the m⁶A level from the residue of rRNA presented in the purified mRNA. We first determined the m⁶₂A/m⁶A ratio of rRNA as 1.30 in the wild-type strain and 2.04 in either an rlmF mutant or an rlmJ mutant. As a negative control, the m⁶₂A modification was not detectable in a ksgA mutant (Supplementary Figure S2). Based on the m⁶₂A/m⁶A ratio in rRNA, we were able to accurately calculate the real m⁶A/A level as ([m⁶A-m⁶₂A/1.30]/A) in the purified mRNA samples.

We observed the presence of m⁶A in all of the tested bacterial species, whose m⁶A/A ratio varied within the range of 0.02–0.28% (Figure 1A). We obtained the m⁶A/A ratios in mRNA from three Gram-negative bacteria (E. coli, P. aeruginosa and P. syringae) (>0.2%) and from two Gram-positive bacteria (S. aureus and B. subtilis) (<0.08%). Unlike E. coli and Pseudomonas spp., two other Gram-negative cyanobacteria (Anabaena sp. PCC 7120 and Synechocystis sp. PCC 6803) showed low m⁶A/A ratios (<0.04%). In order to test if m⁶A/A ratios also vary among different strains in the same species, three strains of E. coli (K-12, 5α and XL-blue), two strains of P. aeruginosa (MPAO1 and PA14) and six strains of S. aureus (Newman, USA100, USA400, USA700, RN4220 and COL) were tested, all of which revealed a constant ratio in the same species (Figure 1B). These results clearly demonstrate the widespread occurrence of m⁶A in bacterial mRNA. Gram-negative bacteria tend to have higher m⁶A/A ratios in mRNA than Gram-positive bacteria. The high m⁶A/A ratio (>0.2%) in mRNA from E. coli and Pseudomonas spp. is comparable to that from eukaryotes (1), suggesting that m⁶A could be an important mRNA modification playing functional roles in these and other bacteria.

Figure 1.

Presence of m⁶A in bacterial mRNA. (A) The m⁶A/A ratios of mRNA isolated from seven bacterial species. (B) The m⁶A/A ratios of mRNA isolated from different strains of Escherichia coli and Pseudomonas aeruginosa. Error bars represent standard deviations, which were calculated from three replicates.

m⁶A distribution exhibits a distinct topology in E. coli

To obtain the transcriptome-wide m⁶A map of E. coli, we employed an m⁶A-specific antibody for pull-down coupled with high-throughput sequencing (15,16). In order to obtain a high-resolution m⁶A-map, bacterial mRNAs were subjected to a modified photo-crosslinking-assisted m⁶A-seq approach (PA-m⁶A-seq), which significantly improves the m⁶A peak resolution from ∼200 nt to around 23 nt (21). In total, we identified 265 m⁶A peaks representing the transcripts of 213 genes in E. coli (Supplementary Table S1).

We next analyzed the distribution of m⁶A in the whole transcriptome of E. coli. We determined the distribution of m⁶A reads along transcripts in both the m⁶A-IP and non-IP (input) samples. Intriguingly, we found that reads from m⁶A-IP tend to be equally distributed throughout a gene, with a peak in the middle of open reading frames (ORFs) (Figure 2A). The prevalence is quite different from that observed in mammals, which accumulates around the stop codon and within 3′ UTRs (15,16). To further confirm the preferential locations of m⁶A on transcripts, we investigated the metagene profiles of m⁶A peaks. Consistent with the distribution of reads, m⁶A peaks are abundant inside ORF (72%), followed by regions at the start of gene (15%) and then the end of gene (13%) (Figure 2B).

Figure 2.

Overview of m⁶A methylome in Escherichia coli. (A) The m⁶A peak distribution within different gene contexts. The y-axis represents (number of reads/length unit)/(number of total reads), which is an indicator of the extent to which sequencing reads are enriched in different segments across the entire transcript. (B) Accumulation of m⁶A reads along transcripts. Each transcript is divided into three parts: Overlap Start, Inside and Overlap End. (C) The UGCCAG conserved sequence motif for m⁶A-containing peak regions. (D) Pie chart displaying the percentage of genes containing m⁶A peaks with functional categories. (E) GO-enrichment analysis of all the genes with m⁶A peaks. The effect size (number of enriched genes/total genes in the GO category) for each category is 15/45 (aerobic respiration), 15/122 (amino acids biosynthesis), 6/138 (response to abotic stimulus), 5/61 (cell wall biosynthesis) and 6/39 (anaerobic respiration), respectively. The statistical test (P-value) used by DAVID was the Fisher Exact test.

We then used the HOMER tool to identify a leading m⁶A consensus sequence (UGCCAG, P < 1e-14), which could be found in more than 41.2% of all m⁶A peaks (Figure 2C and Supplementary Table S1) (25). This motif is different from the conserved one (RRACU, R = A/G) found in eukaryotes. The unique feature of the m⁶A distribution suggests a likely unique role of m⁶A in perhaps bacteria-specific pathways.

m⁶A-containing mRNAs in important biological pathways in E. coli

Diverse functions are encoded by m⁶A-containing genes, which include metabolism (52%), transportation (12%), gene regulation (11%), cell envelope (5%), ribosome (3%), nucleic acids (1%), stress response (5%), genes with unknown function/annotation (6%) and pseudogenes (2%) (Figure 2D). To further uncover potential functional insights on m⁶A in E. coli, we selected 213 m⁶A-containing transcripts and identified the enriched gene ontology (GO) terms using the DAVID tool (26). We found that these genes are highly enriched in aerobic respiration, amino acids biosynthesis, response to abiotic stresses, cell wall biosynthesis and anaerobic respiration (Figure 2E). The first two classes belong to housekeeping genes that are involved in central energy production and metabolism, while the latter three classes are bacteria-specific categories. The m⁶A distribution pattern suggests that m⁶A may play roles in these important biological pathways. For instance, hydrogenase 1 mediates hydrogen uptake and transport in the process of anaerobic respiration. Four (hyaA, hyaB, hyaC and hyaD) genes of the six-gene-operon encoding hydrogenase 1 contain multiple m⁶A peaks inside the transcript, suggesting a clustering of m⁶A in this operon (Figure 3A) (28). We also observed concentrated m⁶A peaks in gabD and gabT genes, which encode succinate-semialdehyde dehydrogenase and 4-aminobutyrate aminotransferase in the pathway of amino acid metabolism (Figure 3B) (29).

Figure 3.

Accumulation of m⁶A reads in hyaABCD genes (A), gabDT (B) and lacZI(C) in Escherichia coli transcriptome. CK represents the control sample and IP represents ChIP-seq sample.

We next sought to determine if the unique m⁶A position patterns are related to bacteria-specific GO categories. As aforementioned, we classified genes into three subgroups according to the location of m⁶A peaks on a gene: Overlap Start (m⁶A peaks within 100-bp from the start codon), Overlap End (m⁶A peaks within 100-bp from the stop codon) and Inside (m⁶A peaks inside the coding region) (Figure 2B). We then performed GO-enrichment analysis for each subgroup. As expected, the same five GO categories were enriched in the Inside subgroup that consists of 72% of all m⁶A peaks. Two GO categories (aerobic respiration and stress responses) were enriched in the Overlap Start subgroup, while amino acids biosynthesis and response to stimulus were enriched in the Overlap End subgroup (Supplementary Figure S3).

Beside these five GO categories, we identified high m⁶A peaks in a group of functionally important genes, such as lacZ and lacI (Figure 3C). LacI negative regulates the classic lac operon (lacZYA) that is required for transport and metabolism of lactose (30). Interestingly, we also noticed 15 small RNAs carrying m⁶A modification (Supplementary Table S1). These newly found m⁶A marks in these transcripts could open a new angle to study novel regulatory roles in well-established pathways.

Unique patterns of P. aeruginosa methylome

Given that P. aeruginosa, a widely-spread human opportunistic pathogen, also possesses a high m⁶A/A ratio in mRNA, we applied the same modified photo-crosslinking-assisted m⁶A-seq approach to obtain a high-resolution map of its m⁶A methylome. We identified 109 m⁶A peaks representing the transcripts of 68 genes in P. aeruginosa (Supplementary Table S2). The m⁶A-modified transcripts identified are around half of those in E. coli.

We next determined the distribution of m⁶A reads along transcripts in both the m⁶A-IP and non-IP (input) samples. Like in E. coli, we found that reads from m⁶A-IP are equally distributed throughout a gene, with two peaks in the middle of ORFs as well as in the beginning of genes (Figure 4A). m⁶A peaks are abundant inside ORF (77%), followed by the start of gene regions (15%) and end of gene regions (8%) (Figure 4B). We were also able to identify an m⁶A consensus sequence (GGYCAG, Y = C/U P < 1e-16), which were found in >70% of all m⁶A peaks (Figure 4C and Supplementary Table S2). This motif is almost identical to the one in E. coli (UGCCAG). A similar feature of the m⁶A distribution in both E. coli and P. aeruginosa indicates that m⁶A possesses functions unique to these bacteria.

Figure 4.

Overview of m⁶A methylome in Pseudomonas aeruginosa. (A) The m⁶A peak distribution within different gene contexts. The y-axis represents (number of reads/length unit)/(number of total reads), which is an indicator of the extent to which sequencing reads are enriched in different segments across the entire transcript. (B) Accumulation of m⁶A reads along transcripts. Each transcript is divided into three parts: Overlap Start, Inside and Overlap End. (C) The GGCCAG conserved sequence motif for m⁶A-containing peak regions. (D) Pie chart displaying the percentage of genes containing m⁶A peaks with functional categories. (E) KEGG-enrichment analysis of all the genes with m⁶A peaks. The effect size (number of enriched genes/total genes in the KEGG category) for each category is 7/148 (amino acids metabolism), 4/37 (glycolysis) and 3/56 (TCA cycle), respectively.

The m⁶A-containing genes cover different gene categories in P. aeruginosa, including metabolism (40%), gene regulation (8%), transportation (1%), virulence (4%), ribosome (12%), stress response (4%) and genes with unknown function/annotation (31%) (Figure 4D). DAVID analysis revealed significant GO enrichments in amino acids metabolism, glycolysis, and tricarboxylic acid (TCA) cycle (Figure 4E), all of which belong to housekeeping genes that are involved in central energy production and metabolism. The m⁶A pattern suggests that m⁶A may play roles in these essential pathways in P. aeruginosa. For instance, three adjacent genes (PA3415-PA3417 encoding branched-chain alpha-keto acid dehydrogenase, pyruvate dehydrogenase β and α subunit, respectively) involved in glycolysis and TCA carry numerous m⁶A peaks (Figure 5A) (31). We also observed multiple m⁶A peaks in the next downstream gene ldh, which encodes leucine dehydrogenase in the pathway of amino acid metabolism (Figure 5A). Beside these housekeeping genes, we identified high m⁶A peaks in two important small RNAs, namely RsmY and RsmZ, as well as in two key virulence genes rhlA and rhlB (Figure 5B).

Figure 5.

Accumulation of m⁶A reads in PA3415–3417 and ldh (A), rsmY (B) and rsmZ (C) in Pseudomonas aeruginosa transcriptome. CK represents control sample and IP represents ChIP-seq sample.

Temperature tunes m⁶A level in P. aeruginosa

In humans and mice, dynamic changes of certain m⁶A peaks have been observed under dif­ferent stress conditions, indicating a link between m⁶A and stress response (15). In order to test if this trend also exists in bacteria, we measured the m⁶A/A ratios under a variety of growth environments or stress conditions (such as varying temperatures, different growth media, exposure to different antibiotics and oxidative stresses) for both E. coli and P. aeruginosa. For most tested conditions, we did not observe a significant difference in the m⁶A/A level compared to the normal condition (LB, mid-log phase, 37°C) for both bacteria. Interestingly, we noticed that increasing the culture temperature (from 37–45°C) led to a clear decrease in the m⁶A/A ratio in P. aeruginosa (Figure 6). Although P. aeruginosa still slowly grew, particularly at 45°C, the m⁶A modification was almost abolished.

Figure 6.

Growth temperature significantly affects the m⁶A/A ratio in Pseudomonas aeruginosa. Error bars are calculated from three replicates.

DISCUSSION

Recent discoveries and characterization of m⁶A erasers (demethylase), binders (m⁶A-specific binding protein) and writers (methyltransferase) as well as advances in profiling the m⁶A methylomes in eukaryotic systems reveal that m⁶A is a reversible and dynamic modification with important regulatory functions (1). On the other hand, the m⁶A methylomes in bacterial mRNA remain poorly studied. Here, we report the presence of m⁶A in a wide range of bacterial species and the ratio of m⁶A/A in mRNA from diverse bacterial strains vary within the range of 0.02–0.28%. We noticed that S. aureus and B. subtilis showed very low m⁶A/A ratios, which suggests that they may not possess an m⁶A methylase that could be present in the Gram-negative bacteria. Based on the genome annotation in NCBI, there are at least 43 proteins containing a S-adenosyl methionine (SAM)-binding domain in E. coli K-12, but only 24 in S. aureus USA300. We further present the high-resolution, transcriptome-wide m⁶A distributions in E. coli and human pathogen P. aeruginosa, which contain 265 and 109 peaks, respectively.

In order to provide additional insights into the overall m⁶A patterns in the bacterial kingdom, we compared these two newly identified methylomes. E. coli and P. aeruginosa share many similarities in their m⁶A distributions that are distinct from those of mammals: (i) a similar motif GCCAG instead of RRACU motif in mammals; (ii) most peaks are in the middle of coding region, while mammalian m⁶A peaks enrich around the stop codon and at 3′ UTRs; (iii) enrichment of GO categories of energy and amino acids metabolism; (iv) many small noncoding RNAs were found to carry m⁶A for both organisms. These shared characteristics suggest that other bacterial species, especially Gram-negative bacteria, may have similar m⁶A characteristics in mRNA. The new consensus sequence (GCCAG) is different from known rRNA methylation sites, including the two m⁶A sites on rRNA (CACA*GGU for RlmF and GUGA*AGA for RlmJ) and one on tRNA^val (UACA*AGG for YfiC). Given that our recent study demonstrates that m⁶A and its specific binding protein, YTHDF2, affect the translation status and lifetime of mRNA in eukaryotes (14), m⁶A may play a similar role in bacteria.

On the other hand, there are clear species-specific features of the m⁶A distribution between these two bacteria. Although the two species share major GO categories (energy and protein metabolism), we observed a very low rate of overlapping genes. Besides rRNA genes that are previously known to carry m⁶A modification, only one gene (aceA encoding isocitrate lyase) was shared between E. coli and P. aeruginosa. Each bacterium has distinct functional categories of mRNAs that carry m⁶A. For example, genes involved in cell wall biosynthesis and anaerobic respiration are enriched in the E. coli methylome only, while a group of virulence genes (RsmYZ and rhlAB) are enriched in the P. aeruginosa methylome (Figure 5B and C). RsmYZ binds to RsmA and dissociates RsmA away from its mRNA targets, which in turn tunes a group of important virulence pathways including Type III secretion system (T3SS) and biofilm formation (32,33). rhlAB encodes rhamnosyltransferase, producing rhamnolipid biosurfactants that are involved in uptake of hydrophobic substrates, virulence, biofilm and antibiotic resistance (34). m⁶A marks in these virulence genes could connect RNA modification to bacterial pathogenesis. Our result also suggests a relationship between m⁶A and P. aeruginosa adaption to temperature changes. Alternatively, the putative m⁶A methylase in P. aeruginosa may be inactive at high temperature.

Given that multiple known rRNA or tRNA adenine methylases have been characterized in E. coli, we measured the m⁶A/A ratios in two methylase mutants, rlmF and rlmJ. However, the ratios were not significantly lower than the wild-type strain, suggesting that they are not mRNA methyltransferases (Supplementary Figure S4). As a negative control, the ksgA mutant lost the m⁶₂A modification and showed a higher m⁶A/A ratio than the other strains, suggesting a small content of rRNA in the mRNA sample. We also tried to look for bacterial homologs of mammalian m⁶A methyltransferases (METTL3, METTL14 and WTAP), but could not identify one, reminiscent of a distinct bacterial m⁶A motif that is different from that of mammals. These results suggest that bacterial m⁶A modification in mRNA is possessing of a mechanism that differs from eukaryotes.

Recent m⁶A profiling in yeast revealed dynamic changes in methylation during meiosis, which led us to test if bacterial m⁶A patterns also vary in different growth stage (11). To this end, we measured the m⁶A/A ratios during lag phase, log phase, stationary phase and death phase for both E. coli and P. aeruginosa. No significant difference was recorded throughout the bacterial growth curve, which indicates that m⁶A is a stable modification during bacterial growth.

The first bacterial m⁶A maps presented here provide a starting roadmap for uncovering bacterial distinct m⁶A functions in the future. Recent breakthroughs in the charac­terization of m⁶A-associated proteins as well as in the development of high-throughput assays in mammals present a very useful toolbox for us to study m⁶A in bacteria. Given the high abundance of m⁶A in numerous bacterial species, we foresee unique functions of m⁶A modification in mRNA in the wide bacteria kingdom.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

ACCESSION NUMBER

m⁶A-seq data are available in the National Center for Biotechnology Information Gene Expression Omnibus under series GSE65347.

FUNDING

US National Institutes of Health [AI074658 to C.H.]. Project of Tianjin, China (13TXSYJC40100). Funding for open access charge: US National Institutes of Health [AI074658 to C.H.].

Conflict of interest statement. None declared.