tRNA modification dynamics from individual organisms to metaepitranscriptomics of microbiomes

Abstract

tRNA is the most extensively modified RNA in cells. On average a bacterial tRNA contains 8 and a eukaryotic tRNA contains 13 modifications per molecule. Recent studies reveal that tRNA modifications are highly dynamic and respond extensively to environmental conditions. Functions of tRNA modification dynamics include enhanced decoding of specific codons in response genes on demand and regulation of tRNA fragment biogenesis. This review summarizes recent advances in the studies of tRNA modification dynamics in biological processes, tRNA modification erasers, and in human associated bacteria. Furthermore, we use the term “metaepitranscriptomics” to describe the potential and approach of tRNA modification studies in the natural biological communities such as the microbiomes.

Short summary:

tRNA is highly modified in cells and tRNA modifications respond extensively to environmental conditions to enhance translation of specific genes and produce tRNA fragments on demand. We review recent advances in tRNA sequencing methods, tRNA modification dynamics in biological processes, and tRNA modification studies in the natural communities such as the microbiomes.

Introduction:

In all domains of cellular life, transfer RNAs (tRNAs) are the most extensively modified RNA family in cells. On average a bacterial tRNA contains 8 and eukaryotic tRNA contains 13 modifications per molecule, corresponding to ~1 in 10 to ~1 in 5 residues being modified with critical implications on cellular physiology. In broad categories, tRNA modifications in the anticodon loop are crucial to fine-tune mRNA decoding, whereas modifications outside of the anticodon loop play roles in tRNA stability, folding, localization, and quality control (Fig. 1).

Figure 1: Broad categories of tRNA modifications.

In the cloverleaf secondary structure of tRNA, the locations of the modified nucleotides are highlighted in orange. Dashed line separates the anticodon stem-loop from the rest of the tRNA. Broad categorization of tRNA modification function is on the right.

Chemically, the simplest tRNA modifications are base or 2’OH methylations. A human tRNA can contain up to 8 methylations including those ending in methyl groups and others followed by additional modifications (Boccaletto et al., 2018). Methylations are also primary markers for mRNA modification (e.g. N6-methyladenosine, m⁶A), chromosomal DNA modification (e.g. 5-methylcytosine, m⁵C), and proteins (e.g. N-methyl-lysine). On the other extreme, the wybutosine (W) modification in human tRNA^Phe requires 7 enzymatic steps of de novo synthesis, and human tRNA^Asp and tRNA^Tyr are first modified to queuosine (Q), then glycosylated with mannose and galactose. This wide range of chemical complexity indicates that tRNA modifications are also part of the extensive cellular networks of metabolism.

Dynamic changes in the tRNA modification fraction at each site are expected among varying cellular activities and physiologies. Dynamic tRNA modification level changes can regulate translational efficiency and fidelity of specific genes that depend on the codon usage. Dynamic tRNA modification level changes can also regulate tRNA fragment biogenesis that in turn affects many cellular processes through their interactions with mRNAs and proteins. Many excellent reviews on tRNA modifications have been recently published that extensively describe our recent advances in the identification of tRNA modification enzymes (called writers), their functions in health and disease, the discovery and function of numerous new tRNA modifications, and the biology and mechanism of action of tRNA fragments (Boccaletto et al., 2018; de Crecy-Lagard et al., 2019; Huber et al., 2019; Kirchner and Ignatova, 2015; Schimmel, 2018; Suzuki, 2021).

This review focuses on the aspect of dynamic tRNA modifications: defined here as variations and changes in the modification fraction at specific sites that are dependent on cell type and cell state. We describe the recent technological advances in the studies of tRNA modification fraction, the biological investigations of tRNA modification response to stress and other cellular conditions, tRNA modification fraction response to physiological conditions and their potential functions, and finally an outlook on ‘metaepitranscriptomics’, an emerging area of study that provides a framework to study tRNA modifications in complex biological systems, such as the microbiome. The literatures in this area are increasing rapidly, we apologize to those we did not have the opportunity to cite in this review.

Sequencing methods of detection and quantitation of tRNA modifications

While liquid chromatography-mass spectrometry (LC/MS) continues to be a powerful approach to study tRNA modifications, recent advances in sequencing strategies have enabled comprehensive access to the modification dynamics with unprecedented throughput. Variations in the library preparation steps, that target the different chemical properties or response to enzyme treatments, provide much flexibility in either detecting a broad group of modifications or targeting a specific common modification. Leveraging specific properties of tRNAs and their modifications is a common theme in recently developed sequencing methods while others aim to alleviate common issues during library preparation steps through clever use of enzymes and specific adapters (Table 1). Several methods also examine tRNA dynamics on a broader, encompassing scale, such as within tissues or in naturally occurring, complex microbiomes. However, the existing sequencing methods target only a subset of tRNA modifications that are open to detection and quantitative assessment by reverse transcription (Fig. 2). There is much room for future development for more comprehensive and more precise measurements of tRNA modifications as current methods have many limitations (Table 1).

Table 1.

Summary of next-gen tRNA-seq methods

Category	Method	Year Published	Description	Capabilities	Limitations	Reference
RT misincorporation signature	DM-tRNA-seq	2015	Uses E. coli AlkB demethylase mixture and the processive TGIRTa in library preparation Overcomes modification and tRNA structure issues during cDNA synthesis	Identification of Watson-Crick face methylations and several other modifications; can work with total RNA Generates higher proportion of full length tRNA reads; useful in tRNA isodecoder studies	Additional steps required such as demethylase treatment and gel purification of cDNA	Zheng et al., 2015
	ARM-seq	2015	Uses E. coli AlkB demethylase in library preparation Enables identification of methylations	Identification of Watson-Crick face methylations Improves tRNA fragment detection and quantitation	Uses conventional RTb that stops frequently Additional steps needed such as AlkB treatment	Cozen et al., 2015
	mim-tRNA-seq	2021	Uses TGIRT under new conditions that read through tRNA modification extensively Computation pipeline addresses mapping issues of incorporating modification-induced mutation results	Substantially increases full-length tRNA reads Computation pipeline improves tRNA isodecoder analysis	May need tRNA purification from total RNA	Behrens et al., 2021
Chemical treatment for specific modification	Ψ-seq; Pseudo-seq	2014 2014	CMCc reaction introduces RT stops at Ψd sites Ψ sites in mRNA and tRNA mapped	Specifically targets Ψ sites	Uses RT stop for detection Mostly short reads for tRNA	Schwartz et al., 2014; Carlile et al., 2014
	RNA bisulfite sequencing	2009 2012 2013	Bisulfite reaction changes C to U, while m5C remains as C in sequencing m5C sites in mRNA and tRNA mapped	Specifically targets m5C sites	tRNA sequencing uses conventional RT that stops frequently	Schaefer et al., 2009; Squires et al., 2012; Edelheit et al., 2013
	DM-Ψ-seq	2020	Uses AlkB demethylase treatment and CMC reaction to identify Ψ sites in tRNA	Significantly improves tRNA coverage by using demethylase	Ψ specific	Song et al., 2020
	RBS-seq	2019	Bisulfite treatment induces mutation and deletion signatures for m1A, Ψ and m5C	Simultaneous identification of multiple modifications	Ability to analyze modification levels unclear	Khoddami et al., 2019
	AlkAniline-Seq TRAC-seq HAC-seq	2018 2019 2021	Alkaline hydrolysis coupled with aniline cleavage identifies m7G and m3C sites in tRNA Borohydride reduction coupled with aniline treatment identifies m7G in tRNA Hydrazine treatment coupled with aniline treatment identifies m3C in tRNA	Specifically targets m7G and m3C sites	Ability to analyze modification levels unclear	Marchand et al., 2018; Lin et al., 2019; Cui et al., 2021
Library construction strategy	YAMAT-seq	2017	Uses ligation of a Y-shaped adapter to the 5′ and 3′ ends of tRNA, followed by PCR of cDNA from full-length tRNA	Specifically targets mature, full-length tRNAs	Misses tRNA modification or structure derived short cDNA that are the majority of cDNA products	Shigematsu et al., 2017
	Hydro-tRNAseq	2017	Uses standard small RNA sequencing method for tRNAs after partial alkaline hydrolysis	Shorter tRNA hydrolysis products have less structure and fewer modifications which enables the use of standard small RNA library preparations	Only short tRNA reads makes it difficult for tRNA isodecoder analysis Ability for comprehensive analysis of tRNA modifications unclear	Gogakos et al., 2017
	LOTTE-seq	2020	Ligation of a hairpin adapter to 3′ CCA allows for selective library construction for tRNA	Sequencing of full-length and prematurely terminated tRNA transcripts	Short tRNA reads due to modification or structure induced stops	Erber et al., 2020
	QuantM-tRNAseq	2020	Split ligation strategy allows for quantification of tRNA abundance and sequence variants Precise tRNA abundance measurements correlated with Northern blots	Sequencing of full-length and prematurely terminated tRNA transcripts	Short tRNA reads due to modification or structure induced stops	Pinkard et al., 2020
	AQRNA-seq	2021	Uses exonucleases to remove excess adapters, allows for rapid and uninterrupted library construction Direct correlation between read counts and copy number of miRNAs	Measures tRNA abundance and modification changes	Short 3′ tRNA reads due to modification/structure induced stops without demethylase treatment May need additional steps such as demethylase treatment	Hu et al., 2021

^aTGIRT: Thermostable group II intron reverse transcriptase

^bRT: Reverse transcriptase.

^cCMC: 1-Cyclohexyl-(2-Morpholinoethyl)Carbodiimide.

^dΨ: Pseudouridine.

Figure 2: Chemical structures of some tRNA modifications assessable by next-gen sequencing.

Modified chemical moieties are shown in red. (A) W-C face modifications: N1-methyladenosine (m¹A), N1-methylguanosine (m¹G), N2,2-dimethylguanosine (m²₂G), N3-methylcytidine (m³C), N3-(3-amino-3-carboxypropyl)uridine (acp³U). (B) After chemical treatment: 5-methylcytidine (m⁵C), pseudouridine (Ψ), dihydrouridine (D), N7-methylguanosine (m⁷G).

Reverse transcription misincorporation: m¹A, m¹G, m³C, m²₂G

Watson-Crick face methylations are abundant in tRNA; they introduce a major challenge in the reverse transcription step during cDNA synthesis. In eukaryotic tRNAs these include N1-methyladenosine (m¹A), 1-methylguanosine (m¹G), 3-methylcytidine (m³C), and N2,2-dimethylguanosine (m²₂G), which all cause some degree of misincorporation or truncated cDNA products during reverse transcription. However, several sequencing techniques take advantage of these observations of RT signatures derived from modifications. Ryvkin et al. developed the HAMR data analysis pipeline that identifies modifications through detection of nucleotides with significant sequencing error rates in Illumina sequencing results (Ryvkin et al., 2013). Detected modifications could be grouped into large modification families by similarities in their incorporation patterns.

The introduction of an enzymatic demethylation step in the sequencing library preparation allows for comparative analysis of treated and untreated samples to further pinpoint these modifications. ARM-seq (Cozen et al., 2015) and DM-tRNA-seq (Zheng et al., 2015) both take a similar approach in using the E. coli AlkB enzymes to demethylate RNA samples, and the latter also utilizes a highly processive thermostable group II intron reverse transcriptase (TGIRT) to increase read-through of these modifications in cDNA synthesis. These techniques solidify the characteristic properties of these methylations in producing specific RT signatures. Additional analysis of the DM-tRNA-seq results (Clark et al., 2016) introduces the modification index (MI) which combines the fraction of mutation and stop at each modification site to describe the semi-quantitative nature of measuring modification fractions.

Recently, mim-tRNAseq (Behrens et al., 2021) builds upon the analysis of modification-induced nucleotide misincorporation, with additional approaches to address cDNA synthesis issues and data analysis. mim-tRNAseq achieves a substantial increase in full length cDNA reads by using the TGIRT with DNA adapters at the 3’ end of tRNA and longer reaction time. Obtaining large portion of full-length tRNA reads is important in the precise mapping of mammalian tRNAs which are derived from many tRNA isodecoder genes with diverse sequences within the same anticodon family (Goodenbour and Pan, 2006).

Chemical treatment of specific modifications: Ψ, m⁵C, m⁷G, D, m³C.

Another fundamental theme in detecting and quantifying specific modifications by sequencing uses chemical treatments that cause targeted changes to select modification groups. Pseudouridine (Ψ) is a prevalent modification in all RNA families, and is also the most abundant modification in eukaryotic tRNA (Spenkuch et al., 2014). A common Ψ detection method uses the carbodiimide (CMC) reaction that specifically forms a chemical adduct for Ψ that can be detected by RT stops (Bakin and Ofengand, 1993). Ψ-seq (Schwartz et al., 2014) and Pseudo-seq (Carlile et al., 2014) use this CMC reaction approach to sequence Ψ in mRNAs and in tRNAs. DM-Ψ-seq (Song et al., 2020) combines demethylase treatment and CMC reaction which significantly enhances the data quality for Ψ analysis in tRNAs. Another Ψ detection method (RBS-seq) uses a bisulfite reaction that removes the Ψ base, thus producing a deletion signature in the RT reaction (Khoddami et al., 2019).

5-Methylcytosine (m⁵C) is another common modification in many RNA families. The most commonly used m⁵C detection method is the bisulfite treatment which converts unmodified C into U, but m⁵C remains as C in sequencing results. Several studies (Edelheit et al., 2013; Schaefer et al., 2009; Squires et al., 2012) apply this technique to study this modification in bacterial, archaeal and eukaryotic tRNAs.

Alkaline hydrolysis is not limited to specific modification changes, as shown by hydro-tRNAseq (Gogakos et al., 2017). Here, a limited alkaline hydrolysis step generates shorter, less structured RNA fragments containing fewer modifications, making them more amenable to cDNA library preparations. AlkAniline-Seq (Marchand et al., 2018) combines the alkaline hydrolysis step to generate abasic sites, followed by extensive 5’- and 3’-dephosphorylation and subsequent aniline cleavage to profile susceptible modifications of N7-methylguanosine (m⁷G), m³C and dihydrouridine (D).

Specific chemical reactions that target m⁷G (TRAC-seq) or m³C (HAC-seq) modifications result in highly efficient RT stops for their detection and quantitation (Cui et al., 2021; Lin et al., 2019).

A plethora of solutions to a common problem: Adapters, nanopore sequencing and new modification discovery

Several other approaches have been documented to overcome cDNA synthesis related issues to enrich for tRNA molecules, reduce truncated cDNA reads or increase full length tRNA reads, all resulting in higher quality sequencing results of tRNA. A common solution to these problems includes the construction of adapter molecules that select for tRNAs in the input sample, thus increasing the portion of sequencing data for tRNA in biological samples. Pang et al. (Pang et al., 2014) show the effects of stressors on tRNA levels in S. cerevisiae using two separate adapter ligation steps: the first adapter ligates selectively to the 3’ end of purified tRNA, while the second ligation to the 3’ end of the cDNA products enhances PCR amplification of truncated RT products. These two adapter ligation steps take into account modification induced falloffs during reverse transcription and increase read coverage. YAMAT-seq (Shigematsu et al., 2017) uses a Y-shaped adapter that specifically ligates to mature tRNA using T4 RNA Ligase 2. This high selectivity for full-length, mature tRNA is advantageous in comparing the tRNA expression levels in human cells using only a low amount of total RNA input. LOTTE-seq (Erber et al., 2020) addresses limitations of adapters that are selective for mature tRNA using an adapter capable of targeting both full length and prematurely terminated tRNA transcripts. This adapter is specifically ligated to the tRNA 3’-CCA end. Similarly, QuantM-tRNAseq (Pinkard et al., 2020) utilizes a split ligation strategy with complementary double-stranded adapters to monitor tRNA abundance and sequence variants in addition to tRNA modifications. AQRNA-seq (Hu et al., 2021) further addresses potential ligation biases towards specific RNA molecules in the library preparation steps. Using a two-step adapter ligation strategy with an optional AlkB treatment, Hu et al. compare their findings with commercially available small RNA sequencing kits and establishes an optimized pipeline with low level of length bias, as well as a potential direct correlation between sequencing read counts and RNA copy numbers.

These methods highlight the importance in examining and quantifying the potential biases and errors typical tRNA library preparation steps can introduce into the sequencing data, in which future methods and analysis pipelines should aim to further address, or at the very least, recognize (Table 1).

The emergence of nanopore sequencing technologies provides another framework in sequencing tRNA including the identification and quantification of modifications. One major benefit with nanopore sequencing is to bypass the issues associated with cDNA synthesis, as tRNAs can be sequenced directly. Nanopore also sequences single RNA molecules so that the coordination of multiple tRNA modifications can be studied simultaneously. However, the stable tRNA secondary structure can be problematic in translocating tRNA through the pores. Smith et al. (Smith et al., 2015) proposes possible methods to deal with linearizing tRNAs using specific adapters and polymerases to enhance processivity to avoid clogging the pores. Another limitation is that nanopore sequencing works poorly for short RNA, so that tRNA must be linked to long adaptor RNAs for sequencing runs. Thomas et al. (Thomas et al., 2021) sequence full length E. coli tRNAs by ligating double-stranded splint adapters to both ends of tRNA. The added bases to the 3’ and 5’ ends of tRNA enable efficient pore passthrough of both ends, which enables full-length tRNA to reach 76–92% of the aligned reads. In addition, systematic base miscalls indicate known tRNA modifications. With more nanopore tRNA sequencing methods currently in preprint, one can expect to see a boom of using nanopore to study tRNA modification dynamics.

With only a limited selection of complete tRNA modification profiles of mostly model organisms (Boccaletto et al., 2018), comparative analysis using these well described tRNA profiles is a useful approach for investigating the unknown tRNA profiles of other organisms. Kimura et al. (Kimura et al., 2020) leverages the known information on E. coli tRNA modifications to fully map and also identify novel V. cholerae specific tRNA modifications through combining tRNA sequencing and tRNA mass spectrometry.

tRNA modification dynamics in stress response and development

First global assessment

To investigate the dynamic modulation of tRNA modifications globally in yeast, Chan et al. developed a quantitative LC-MS/MS technique (Chan et al., 2010). Stress treatments cause a global tRNA modification profile reprogramming that is stress agent-specific and dose-dependent. Hydrogen peroxide, arsenite, methylmethane sulfonate (MMS), and hypochlorite induce distinct patterns of tRNA modification change. For example, the levels of 2’O-methyl-C (Cm), m⁵C, and m²₂G increase in response to H₂O₂ but decrease or are resistant to arsenite, MMS, and hypochlorite. As the concentration of H₂O₂ increases, the levels of m⁵C, Cm and m²₂G increase, while 5-methyl-U (m⁵U), m¹G, m²G, 5-methoxycarbonylmethyl-2-thio-U (mcm⁵s²U), N6-isopentenyl-A (i⁶A), wybutosine (yW) and m¹A decrease. MMS increases the level of m⁷G and decreases the level of m⁵C, mcm⁵s²U, Cm, i⁶A, and yW. tRNA 5-methoxycarbonylmethyl-U (mcm⁵U), m³C, m⁷G, mcm⁵s²U, i⁶A, yW, m⁵C, and Cm levels only decrease at highest concentration of NaAsO₂ of 60 μM. NaOCl also shows dose-dependent effect on the increased levels of 2’O-methyl-A (Am) and 2’O-methyl-U (Um) and decreased levels of m⁵C. These changes in tRNA modification levels can be derived from the induction of new modification enzymes, altered activity of existing modification enzymes, or selective degradation of modified or unmodified tRNAs.

tRNA modifications in the anticodon loop affect decoding of stress-response genes

The m⁵C34 modification in tRNA^Leu(CAA) is the anticodon wobble nucleotide whose installation is catalyzed by TRM4 in yeast. H₂O₂ treatment increases its level (Chan et al., 2012). Luciferase reporter and proteomic analysis show that elevated level of m⁵C34 enhances selective translation of UUG codon enriched mRNA (e.g. the ribosomal protein RPL22A). Deletion of TRM4 or RPL22A results in cytotoxic hypersensitivity to oxidative stress. Thus, the wobble m⁵C modification and the m⁵C dependent upregulation of RPL22A is crucial for oxidative stress response in yeast.

The DNMT2 dependent tRNA m⁵C38 modification also plays an important role in stress response. m⁵C38 modulates the stability and fragmentation of substrate tRNAs in the bone marrow. Loss of m⁵C38 in tRNA^Asp leads to misincorporation of near-cognate amino acids during translation of Asp codons in primary bone marrow cells (Tuorto et al., 2015). DNMT2 mutant fruit flies are more sensitive to oxidative stress. Under heat shock, DNMT2 re-localizes to stress granules in Drosophila ovaries (Schaefer et al., 2010).

The wobble anticodon uridines of tRNA are extensively modified. O² of U34 can be modified to 2-thiouridine (s²U) by the URM1 pathway proteins (Chowdhury et al., 2012; Marelja et al., 2008; Noma et al., 2009; Schlieker et al., 2008; Termathe and Leidel, 2018). Modification at C⁵ of U34 generates 5-carboxymethyluridine (cm⁵U), 5-methoxycarbonylmethyluridine (mcm⁵U), 5-(carboxyhydroxymethyl)uridine methyl ester (mchm⁵U), and 5-methoxycarbonylmethyl-2′-O-methyluridine (mcm⁵Um) by the elongator complex consisting of ELP1-ELP6, ALKBH8 or TRMT9L (TRM9 in yeast) (Dauden et al., 2018; Fu et al., 2010a; Fu et al., 2010b; Songe-Moller et al., 2010). The URM1 and ELP pathway dependent mcm⁵U and mcm⁵s²U modifications increase in nutrient starvation and oxidative stress. Translation assays with tRNAs lacking mcm⁵U and mcm⁵s²U modifications show impaired translation of genes enriched with AAA, CAA, and GAA codons in S. cerevisiae and C. elegans (Deng et al., 2015; Nedialkova and Leidel, 2015; Rezgui et al., 2013). Both mcm⁵U and mcm⁵s²U modulate translation by promoting binding of tRNA^Lys(UUU) to the ribosomal A-site. Loss of TRM9 causes decreased levels of wobble mcm⁵U and mcm⁵s²U modifications leading to decreased translation fidelity and increased levels of unfolded proteins which induces unfolded protein and heat shock responses (Nedialkova and Leidel, 2015; Patil et al., 2012a).

Hydroxyurea (HU)-induced S-phase cells have higher level of mcm⁵U which enables efficient translation of AGA codons and the RNR1 protein (Patil et al., 2012b). The level of mcm⁵U oscillates during the cell cycle. Loss of mcm⁵U modification causes a codon-specific downregulation of RNR1 and a delay in transition into S-phase after DNA damage (Patil et al., 2012b). Loss of mcm⁵s²U34 also induces multiple starvation responses in yeast (Bruch et al., 2020) and confers sensitivity to H₂O₂ stress (Garcia et al., 2016). Overexpression of tRNA^Lys(UUU) largely suppresses ribosome pausing and the sensitivity to oxidative stress and simultaneously restores protein homeostasis, indicating that the stress sensitivity is related to the U34 modified tRNAs-mediated translation (Nedialkova and Leidel, 2015). Loss of s²U34 thiolation in tRNA leads to impaired global translation and confers resistance to tunicamycin. The level of s²U34 is reduced at elevated temperatures (Alings et al., 2015; Damon et al., 2015) which is due to the synthesis of newly unmodified tRNAs, not the degradation of s²U34 containing tRNAs nor the removal of s²U34 from tRNA (Alings et al., 2015). During nitrogen deprivation, mcm⁵U and mcm⁵s²U modification-dependent reciprocal regulation of TORC signaling and tRNA modification is activated (Candiracci et al., 2019).

Depletion of mcm⁵s²U34 installation enzymes in human breast cancer cell lines (MCF7 and MDA-MB231) leads to translational defects including ribosome pause and accumulation of ribosomes on XAA codons enriched transcripts (Rapino et al., 2021). The U34 wobble modification status regulates the translation of specific penta-hydrophilic amino acid motifs that determines the fates of the protein product such as aggregation, degradation, and decreased expression.

m³C modification is present at positions 32 in the anticodon loop of tRNA^Ser/tRNA^Thr/tRNA^Arg (Clark et al., 2016). METTL2A/B and METTL6 catalyze the formation of m³C32 in cytosolic tRNAs, METTL2B may be responsible for the formation of m³C32 in mitochondrial tRNAs (de Crecy-Lagard et al., 2019; Ignatova et al., 2020; Xu et al., 2017). METTL6 methylates tRNA^Ser in vitro and in human cells. METTL6 knockout leads to global changes in translation profile and ribosome occupancy, as well as loss of mouse stem cell pluripotency (Ignatova et al., 2020). Stress response studies of m³C tRNA modification in yeast reveal a stress-specific reprogramming of modifications and selective expression of different classes of stress response proteins mediated by codon-biased mRNAs translation (Chan et al., 2015). Stress-induced tRNA modification and mRNA expression patterns can distinguish different types of stresses in yeast exposed to multiple oxidants and alkylating agents. For example, S_N2 alkylating agents cause increase of m³C32 in tRNA^Thr(IGU), which leads to selective expression of ACC/ACU codon enriched membrane protein genes (Chan et al., 2015).

N6-threonylcarbamoyladenosine (t⁶A) and 2-methylthio-N6-threonylcarbamoyladenosine (ms²t⁶A) are present at position 37 in the anticodon loop of tRNAs with NNU anticodons. t⁶A is catalyzed by a group of 6 proteins in the cytosol and YRDC and OSGEPL1 in mitochondria (de Crecy-Lagard et al., 2019; Thiaville et al., 2014; Zhou et al., 2020). CDKAL1 is responsible for the formation of ms²t⁶A37 in tRNA^Lys(UUU) (Wei et al., 2011). The ms²t⁶A37 modification is crucial for maintaining translation fidelity of AAA and AAG codons. Depletion of ms²t⁶A37 in tRNA^Lys(UUU) in mouse pancreatic β cell through CDKAL1 knockout causes impaired translation of proinsulin due to misincorporation of non-cognate amino acids at lysine codons (Wei et al., 2011). Loss of t⁶A modification in yeast confers sensitivity to stress treatments such as heat, ethanol, and salt and TOR pathway inhibitors, and leads to severe growth defects that cannot be rescued by the overexpression of any individual tRNA (Thiaville et al., 2016). t⁶A37 level is crucial for codon-biased mitochondrial translation under physiological conditions; it is also sensitive to environmental availability of CO₂/bicarbonate which is rate-limiting for t⁶A37 formation in human mitochondria (Lin et al., 2018). t⁶A37 depletion by knockout of OSGEPL1, one of the enzymes required for t⁶A, leads to respiratory defect and reduced mitochondrial translation.

5-oxyacetyluridine (cmo⁵U) is present at the wobble anticodon position of four tRNA families. cmo⁵U is crucial for expanding the ability of wobble U to read G, A, U, and sometimes C (Nasvall et al., 2007; Weixlbaumer et al., 2007). CMOB is responsible for converting 5-hydroxyuridine (ho⁵U) to either 5-methoxyuridine (mo⁵U) or cmo⁵U based on the availability of S-carboxymethyl-S-adenosylmethionine (carboxy-SAM). CMOM further modifies cmo⁵U to mcmo⁵U (Kim et al., 2015; Sakai et al., 2016). Under constant hypoxic stress, M. bovis BCG shows distinct global change in 40 tRNA modifications in the three classical phases of hypoxia induced persistence of aerated growth and non-replicating persistence stages 1 and 2 (Chionh et al., 2016). Early hypoxia specifically increases the level of cmo⁵U and the abundance of tRNA^Thr(UGU) which leads to genome-wide codon-biased translation of mRNA enriched in ACG/ACA codons. The codon-mediated translation of DOSR as the master regulator of hypoxic bacteriostasis is confirmed and analyzed by codon re-engineering of DOSR.

5-taurinomethyluridine (τm⁵U) and 5-taurinomethyl-2-thiouridine (τm⁵s²U) are present at the wobble anticodon position of several mitochondrial tRNAs in mammals (Suzuki and Suzuki, 2014; Suzuki et al., 2002). MTO1 and GTPBP3 are responsible for the formation of the τm⁵U modification, while MTU1 is responsible for the formation of 2-thio group of τm⁵s²U (Asano et al., 2018; Suzuki and Suzuki, 2014; Umeda et al., 2005). 5,10-methylene-tetrahydrofolate and taurine are the metabolic sources of τm⁵U34 in mitochondrial tRNA (Asano et al., 2018). Cells salvage taurine from the medium (Suzuki et al., 2002) and the availability of taurine affects the τm⁵U34 modification level. Loss of τm⁵U34 by GTPBP3-knockout leads to respiratory defects and reduced mitochondrial translation. Taurine depletion decreases τm⁵U(s²)34 levels in HeLa cells and animal tissues, and leads to the replacement of τm⁵U34 modification by 5-carboxymethylaminomethyluridine (cmnm⁵U) where the taurine moiety is replaced by glycine in mitochondrial tRNAs (Asano et al., 2018).

Queuosine (Q) is present at the wobble anticodon position of tRNA^Tyr/tRNA^His/tRNA^Asn/tRNA^Asp (Fergus et al., 2015). Q in tRNA^Tyr and tRNA^Asp is further glycosylated by galactose and mannose at the cis-diol group to generate galactosyl-Q (galQ) and mannosyl-Q (manQ) (Kasai et al., 1976; Kasai et al., 1975). QTRT1/QTRT2 heterodimer complex catalyzes the incorporation of queuine that is salvaged from diet and gut microbiome into tRNA (Boland et al., 2009). Q levels in tRNAs are controlled over development stages (White and Tener, 1973) and dynamically dependent on the availability of queuine in the diet (Reyniers et al., 1981) as well as the composition and functioning of the gut microbiome (Farkas, 1980). Q modification in tRNA^His enhances decoding of U-ending codons (Meier et al., 1985). In drosophila species, Q modification level changes correlate with genome-wide alteration of Tyr/His/Asn/Asp codons which is consistent with coordination of codon-mediated expression and translational fidelity (Zaborske et al., 2014). Queuine availability fine-tunes the translational speed and fidelity in S. pombe. Q modification in tRNA^Asp and tRNA^His increases the translation speed at cognate C-ending codons while Q modification in tRNA^Asn and tRNA^Tyr decreases the translation speed at cognate U-ending codons (Muller et al., 2019). Q enhances the translation fidelity through inhibiting second-position misreading of the glycine codon (GGC) by tRNA^Asp. The presence of Q modification moderately impairs mitochondrial functions under normal growth conditions by reducing the translation of mitochondrial proteins, while Q suppresses the inhibition effect on cell growth under CaCl₂ induced stress (Muller et al., 2019).

Q levels regulate DNMT2-dependent m⁵C38 methylation of tRNA^Asp. Q modification stimulates m⁵C38 installation in S. pombe and D. discoideum (Muller et al., 2015), in E. histolytica (Nagaraja et al., 2021), and in human cells (Tuorto et al., 2018).

Q modification increases the resistance of E. histolytica to oxidative stress and suppresses the downregulation of translation caused by oxidation stress through promoting the translation of oxidative stress response proteins, such as Hsp70, antioxidant enzymes, and DNA repair proteins (Nagaraja et al., 2021).

In human cells, queuine depletion causes reduced translation fidelity and formation of unfolded proteins that triggers endoplasmic reticulum stress and the unfolded protein response in HeLa and HCT116 cells and in germ-free mice (Tuorto et al., 2018). In HeLa cells, queuine depletion significantly impairs mitochondrial function, characterized by increased mitochondrial proton leak, decreased ATP synthesis, and reduction in cellular ATP levels. A breast tumor mouse model shows significant enrichment of microbes in the xenograft tumors in QTRT1 knockout tumor cells, indicating a possible role of Q modification in microbial recruitment to tumor sites (Zhang et al., 2020).

Inosine is present at positions 34 and 37 in the anticodon loop of eukaryotic tRNAs with ANN anticodons. ADAT2/ADAT3 are responsible for the formation of I at position 34 (Torres et al., 2014), and ADAT1 catalyzes the formation of I37 (Maas et al., 1999). In Arabidopsis, loss of the I37 modification due to the catalytically dead mutation of AtTAD1 shows no effect on tRNA^Ala(AGC) stability and no phenotype under normal conditions, but gains less biomass under cold and heat shock stress treatments (Zhou et al., 2013).

The I34 modification is conserved in eukaryotic tRNAs, it occurs in 8 human tRNA isoacceptors of all 3–4 codon boxes of Ala, Arg, Leu, Ile, Pro, Ser, Thr, Val with the exception of Gly (Novoa et al., 2012). Inosine expands the decoding potential in bacterial tRNA^Arg(ICG) to U/C/A ending codons. In human, I34 modification is catalyzed by ADAT2/ADAT3 heterodimer complex (hetADAT) which co-localizes in the nucleus. The level of I34 modification is dynamically controlled in response to the hetADAT protein levels (Torres et al., 2015). To our knowledge, dynamic changes of I34 modification in response to cellular conditions have not been characterized.

tRNA modifications outside of anticodon loop affect tRNA stability and localization

m⁵C is present at many positions in tRNA. NSUN2 is responsible for the formation of m⁵C at positions 40, 48–50 in the anticodon stem and T stem. (Blanco et al., 2014; Tuorto et al., 2012). NSUN6 catalyzes the formation of m⁵C at position 72 in the acceptor stem (Haag et al., 2015). m⁵C modification plays crucial roles in stress resistance and neurological development. Deletion of NSUN2 leads to the loss of m⁵C in specific tRNAs in mice and humans (Blanco et al., 2014). Loss of NSUN2-specific m⁵C modification does not reduce the abundance of specific tRNAs but causes an accumulation of 5’ tRNA-derived small RNA fragments (5’ tRF) by promoting angiogenin-mediated tRNA cleavage. During UV radiation and NaAsO₂ treatment, NSUN2 re-localizes to nucleoplasm and cytoplasmic granules, leading to decreased methylation activity of NSUN2 on tRNA and increased 5’ tRF. 5’ tRF co-localizes with NSUN2 in cytoplasmic stress granules. NSUN2 depleted mouse and human cells are sensitive to UV radiation and NaAsO₂ mainly due to the loss of NSUN2-mediated tRNA methylation. The accumulation of 5’ tRFs caused by NSUN2 loss inhibits translation, and this effect is sufficient to trigger cellular stress response pathways that lead to impaired survival of cortical, hippocampal and striatal neurons.

N2,N2-dimethylguanosine (m²₂G) is installed by TRMT1 and present at position 26 at the junction of anticodon and D stems in about half of cytosolic tRNAs and mitochondrial mt-tRNA^Ile (Clark et al., 2016; de Crecy-Lagard et al., 2019; Dewe et al., 2017). Loss of m²₂G modification due to TRMT1 knockout leads to impaired cell proliferation, disrupted global protein translation, and redox homeostasis perturbations, which leads to increased endogenous ROS levels and confers hypersensitivity to oxidative stress (Dewe et al., 2017).

3-(3-amino-3-carboxypropyl)uridine (acp³U) is present at position 20/20a in the D loop of human tRNAs, and position 47 in the variable loop of E. coli tRNA (Takakura et al., 2019). Acp³U47 in E. coli is installed by tapT (yfiP) and is crucial for tRNA thermal stability. Loss of acp³U47 in E. coli ΔtapT mutant also leads to genome instability under continuous heat shock stress. The human tapT homologues (DTWD1 and DTWD2) are shown to mediate the formation of acp³U in cytosolic tRNAs. DTWD1 and DTWD2 double knockout decreases cell growth (Takakura et al., 2019).

tRNA modifications and tRNA-derived fragments (tRFs)

tRNA fragments (tRF) are cleavage products derived from pre-tRNAs and mature tRNAs that have many biological functions such as regulating translation, enhancing cell survival, and mediating cell differentiation (Anderson and Ivanov, 2014). tRNA modifications play an important role in the regulation of tRNA cleavage and fragment generation. For example, DNMT2 dependent m⁵C38 modification (Blanco et al., 2014; de Crecy-Lagard et al., 2019; Schaefer et al., 2010; Tuorto et al., 2012) protects substrate tRNAs from angiogenin-mediated cleavage under oxidative stress and heat shock (Schaefer et al., 2010). The dynamic modulation of tRNA modifications by stress treatments leads to a dynamic pool of tRFs. The roles of tRNA modifications in tRF biogenesis and function have been summarized in several recent reviews (Durdevic and Schaefer, 2013; Fagan et al., 2021; Guzzi and Bellodi, 2020; Lyons et al., 2018). tRNA modifications such as m¹A, Ψ, m⁵C, m³C, and Q can protect full-length tRNAs from stress-induced ribonucleases cleavage (Blanco et al., 2014; Chen et al., 2019; Schaefer et al., 2010; Wang et al., 2018). Other modifications such as mcm⁵s²U34 and Q may promote tRF generation (Donovan et al., 2017; Jablonowski et al., 2006). Besides, modifications in the tRNA fragments may be crucial for tRF stability and function. For example, Ψ is crucial for fine-tuning the activities of tRNA fragments in stem cells (Guzzi et al., 2018).

Erasing tRNA modifications

tRNA methylations are reversible and dynamically controlled through the action of tRNA methylases and demethylases. Currently, three tRNA demethylases, called erasers have been identified in humans: ALKBH1, ALKBH3, and FTO (Chen et al., 2019; Liu et al., 2016; Ueda et al., 2017; Wei et al., 2018). These erasers help control the dynamic level of tRNA methylations which fine-tune protein translation and tRNA fragment generation.

ALKBH1 demethylates m¹A58 modification in ~10 tRNA isoacceptor families (Liu et al., 2016). M¹A58 is located in the T loop of tRNA and adds a positive charge to the tRNA. ALKBH1 depletion leads to increased m¹A58 levels in the target tRNA. ALKBH1 depletion also leads to an increase of tRNA^iMet level. Coupled with the elevated protein translation and cell proliferation, these results suggest that m¹A58 demethylation regulates translational initiation and elongation. In response to glucose deprivation, ALKBH1 protein level increases and the m¹A level in target tRNA decreases which can explain the downregulation of translation under glucose starvation (Liu et al., 2016). ALKBH1-mediated tRNA demethylation also shows stress-specific patterns. Depletion of ALKBH1 improves both apoptosis and necrosis and reduced tRNA cleavage under arsenite stress. ALKBH1 depletion partially rescues tRNA fragment generation under arsenite and antimycin A stress, and conversely ALKBH1 overexpression enhances tRNA fragment generation through modulating m¹A58 demethylation (Rashad et al., 2020).

In mitochondria, ALKBH1 can demethylate m¹A in the bodies of tRNA^Arg and tRNA^Lys (Kawarada et al., 2017), although the function of mt-tRNA m¹A demethylation is unclear. ALKBH1 plays a dual role in tRNA modification, as it is also the writer of the cytosolic tRNA^Leu and mt-tRNA^Met 5-formyl-C (f⁵C); depletion of ALKBH1 reduces mitochondrial translation and respiratory complex activities (Haag et al., 2016; Kawarada et al., 2017).

ALKBH3 demethylates m¹A58 in 12 tRNA isoacceptor families and m³C in 4 tRNA families (Chen et al., 2019; Ueda et al., 2017). Unlike ALKBH1, ALKBH3 knockdown reduces cell proliferation through downregulation of translation (Chen et al., 2019; Ueda et al., 2017). ALKBH3 dependent tRNA demethylation also affects tRNA cleavage and fragment generation by angiogenin. These demethylation-dependent tRFs interact with the cytochrome c protein to reduce cell apoptosis and play a role in ALKBH3-induced cancer cell progression (Chen et al., 2019).

The third tRNA demethylase is FTO which also demethylates N6-methyladenosine (m⁶A) in mRNA. FTO demethylate m¹A58 in many tRNA isoacceptor families, and codon-specific luciferase reporter assays indicate that FTO depletion leads to upregulation of translation (Wei et al., 2018).

These results from tRNA eraser studies suggest a role of the tRNA demethylation in the dynamic regulation of cellular functions in response to environmental stimuli. ALKBH1, ALKBH3, and FTO can all demethylate m¹A58 in tRNAs. How these three demethylases interact and cooperate on m¹A58 demethylation and cellular function needs further exploration.

Recently, ALKBH7 was shown to demethylate m²₂G in mitochondrial tRNA^Ile and m¹A in mitochondrial tRNA^Leu1 pre-tRNA regions within nascent mitochondrial polycistronic RNA. Demethylation by ALKBH7 is crucial for normal processing of polycistronic mitochondrial RNA, which regulates steady-state mitochondrial RNA levels and downstream protein translation. Depletion of ALKBH7 leads to downregulation of mitochondrial proteins and decreases mitochondrial activity in human cancer cell lines and mouse tissues (Zhang et al., 2021).

tRNA modifications in the microbiome: a new world

A myriad of bacteria is present in a vast variety of environments ranging from the human gut to the bottom of the ocean, yet only a modest number of studies have been conducted beyond the laboratory cultures. Below we will briefly review some studies framed within the context of the interactions with human cells or in human health, highlighting another biological axis for future investigations.

Gut bacteria respond to food and antibiotic intake

The mammalian gut microbial communities live in multiple different environments, depending on the type of diet and metabolite intake by the host. Bacteria in the gut respond to these changes, either to utilize the influx of new nutrients, resist changes to the existing environment or respond in different manners to ensure survival.

One such example is the regulation of tRNA 2’-O-methylation (Nm) in E. coli tRNAs upon the onset of mild antibiotic stress and limited nutrient supply. The regulation of Nm modification does not significantly influence translation, but rather, the immune-stimulatory properties of the bacteria (Galvanin et al., 2020). tRNA 2’-O-methlyations, specifically at position G18 in the D loop, effectively suppress Toll-like receptors (TLR7) of the innate immune system in mammalian hosts (Gehrig et al., 2012). Under typical laboratory growth conditions, the lack of Gm18 minimally affects translation efficiency. However, Gm18 increases upon exposure to mild antibiotic conditions. E. coli tRNA isolated from mild antibiotic stress conditions causes lower activation of human plasmacytoid dendritic cells than those from conditions without antibiotic exposure. This tRNA modification dependent bacterial adaptation in response to antibiotics improves its survival within a human host.

Food intake also greatly influences gut bacterial physiology (Teng et al., 2018). The gut bacterium Lactobacillus rhanosus GG increases in abundance in the mouse gut in response to lemon-derived exosome-like nanoparticles (LELNs), likely through increased resistance to bile (Lei et al., 2021). The underlying mechanism may be specific tRNA decay, in which downregulation of proteins Msp1 and Msp3 leads to decreases in cell wall hydrolysis which reduces bile accessibility into cells. The minor change of Msp1 and Msp3 mRNA levels suggests a potential mechanism at the translational or post-translational level. Indeed, the tRNA processing enzyme RNase P seems to be involved in mediating tRNA^Ser decay which decreases the expression of Msp1/3 proteins.

Pathogens utilize modifications to increase survival within hosts

Pathogenic bacteria also alter their tRNA dynamics to ensure survival during infection. On a global tRNA modification scale, high levels of tRNA methylations in E. coli are necessary in ensuring a general resistance against antibiotics (Masuda et al., 2019). In line with the use of modifications to control translation efficiency, the m¹G37 modification status in the anticodon loop significantly affects the translation of several membrane proteins through proline decoding in their open reading frames (Masuda et al., 2019). The lack of the m¹G37 writer impairs membrane structure of E. coli and Salmonella enterica, sensitizing them to antibiotics.

Upon onset of infection within a human host, reactive oxygen species production is a common general response in mitigating growth and spread of pathogenic bacteria. As a result, these pathogens have adapted specific methods to deal with this environmental stressor. Mycobacteria have been observed to change their tRNA patterns upon exposure to hypoxia, a commonly induced environment due to the inflammatory response by the immune system (Chionh et al., 2016). Each stage of hypoxia-induced change is associated with distinct patterns in tRNA modification. Specifically, the Doc regulon genes, which control the hypoxia response, have altered translation efficiency depending on their codon composition bias, particularly through decoding by tRNA^Thr(UGU). Pseudomonas aeruginosa modulates their expression of detoxifying genes such as KatA and KatB at the translational level upon exposure to hydrogen peroxide stress (Thongdee et al., 2019). Here, the m⁷G46 writer, TrmB is necessary to effectively translate Phe- and Asp-enriched mRNAs including the catalase enzymes KatA and KaB. Exposure to hydrogen peroxide stress is associated with increases in m⁷G levels, while the lack of TrmB leads to decreases in KatA and KatB expression and reduced resistance to hydrogen peroxide stress. These studies highlight specific mechanisms in which tRNA dynamic changes can be a main driver in mediating bacterial survival responses to human hosts defenses.

tRNA modification dynamics in gut microbiomes

All studies described above dealt with single organisms. However, complex biological communities also respond to the environment through tRNA modifications. Schwartz et al. (Schwartz et al., 2018) perform tRNA sequencing of the gut microbiome of mice fed with high-fat or low-fat diets to investigate how tRNA modifications respond to dietary conditions. In contrast to standard tRNA-seq where sequencing reads can be readily assigned to the tRNA genes in a reference genome, microbiomes do not have a readily available tRNA references for sequencing data alignment. To overcome this technical obstacle, for microbiome tRNA-seq analysis Schwartz et al. developed a pipeline that identifies each tRNA de novo using a defined signature of canonical tRNAs (Fig. 3A). Each tRNA is then assigned to a taxonomic group obtained from a database from >4,000 microbial genomes. This analysis also generates tRNA “seed” sequences that are most closely related to the tRNA genes in the microbial genomes. In the case of mouse gut microbiome on average ~10,000 seed sequences are obtained in each sample.

Figure 3: Gut microbiome tRNA-seq.

(A) Canonical tRNA signature used in de novo identification of tRNA reads. (B) tRNA modification identification by mutation in a bacterial culture. Demethylase (DM) treatment removes the mutation at nucleotide 22, assigning it to the known m¹A modification. Position 8 mutation is in-sensitive to demethylase treatment; it corresponds to the well-characterization s⁴U modification in bacterial tRNA. (C) Same as panel B in a gut microbiome sample. The two demethylase removable peaks correspond to the m¹A modification in clostridia and bacilli around nucleotide 22 (c/b), and in actinobacteria around nucleotide 77 in this type II tRNA.

tRNA modifications in the microbiome are identified by the misincorporation or “mutation” from the most closely related seed sequence. To distinguish mutation signature from actual tRNA modification versus those from single nucleotide polymorphism (SNP) in the microbial population, the same sample is sequenced with and without demethylase treatment using the DM-tRNA-seq procedure (Zheng et al., 2015). The most obviously verifiable tRNA modification is m¹A which is “mutated” in high levels without and converts back to the seed sequence with demethylase treatment (Fig. 3B, 3C). Another modification identified through mutation signature is 4-thio-U (s⁴U) at the junction between the acceptor and D stems. Even though s⁴U does not respond to demethylase treatment, its position in tRNA is always U in canonical tRNA. Overall, among the 10 bacterial taxonomic classes found in the mouse gut microbiome in this study, m¹A is present mostly in 3 and s⁴U present in 8 classes. Among the two most abundant bacterial classes in the gut microbiome, clostridia tRNAs have both m¹A and s⁴U, bacteriodes tRNAs have neither. As measured by mutation fractions, the s⁴U levels are about the same in both high-fat and low-fat fed mice. A known function of s⁴U is response to UV radiation (Kramer et al., 1988). The gut microbiome samples are not exposed to light which can explain the low dependency of s⁴U level on diets.

In the gut microbiome, the tRNA m¹A modifications respond strongly to the dietary conditions: higher in the same microbial tRNAs in the high-fat fed than the low-fat fed mice. In order to understand this result, Schwartz et al analyzed the published meta-proteome data from the high-fat and low-fat mice (Zhang et al., 2016). First, they found ~650 genes in the clostridia class that are differentially expressed. Performing the codon usage analysis of the protein group overexpressed in high-fat fed mice versus another protein group overexpressed in low-fat fed mice, they find that high-fat overexpressed proteins have strongly enriched Glu-Glu codon pair or other codon pairs that include Glu. Since tRNA^Glu is one of the 4 tRNAs containing m¹A and m¹A level is higher in high-fat fed mice, this result is consistent with higher m¹A level in tRNA^Glu enhancing translation in a codon pair dependent manner. As for why only Glu codon, but not the codons read by other 3 tRNAs are affected, one possible explanation is that Glu codon response is specific for the dietary condition. One can hypothesize that codons read by the other 3 m¹A-modified tRNAs, tRNA^Cys, tRNA^Gln, and tRNA^Ser may be preferred in translation under other conditions encountered by the gut microbiome such as inflammation or oxidative stress.

Metaepitranscriptomics

Applications of our evolving understanding of tRNA modifications to study microbial responses to environmental change within short timescales represents a new and exciting frontier. We live on a microbial planet (Flemming and Wuertz, 2019): thanks to their astonishing metabolic potential and diversity, microbes occupy all niches Earth has to offer, where they form complex and dense communities undertaking the recycling of critical ingredients of life through global biogeochemical cycles (Falkowski et al., 2008). Due to the profound impact of microbes on Earth’s habitability, understanding their ecology, evolution, and responses to environmental change has always been one of the most significant endeavors of the life sciences.

Emerging ‘omics approaches along with advances in sequencing strategies offer unprecedented insights into microbial ecosystems and lifestyles. For instance only during the past few years metagenomics, the study of the entire DNA content of a given environment, has enabled the discovery of new branches of life previously missed by culture-based studies (Castelle and Banfield, 2018), revealed new clades of human virome and their biogeography (Edwards et al., 2019), and resulted in key ecological (Delmont et al., 2018), evolutionary (Spang et al., 2015), and biotechnological (van Kessel et al., 2015) insights. However, as widely used high throughput ‘omics strategies typically target genetic information encoded in DNA or RNA molecules, studies that aim to understand environmental microbes lack insights into translation and its regulation. Only a small number of publications so far has ventured into this direction (Fremin et al., 2020; Schwartz et al., 2018). As a result, accents that fine-tune the complex metabolic orchestration of environmental bacteria beyond the information encoded in their genome are systematically absent in state-of-the-art analyses.

A bacterial cell has 30–45 different tRNA species, up to 100,000 tRNA transcripts, and 6–10 modifications per molecule. Based on the summary of bacterial tRNA modification types and locations (Fig. 4), the current strategies to sequence tRNA transcripts can reveal ~25% of sites on average of all bacterial tRNA modifications based on mutation signatures in cDNA synthesis. The dynamic range of tRNA modifications in complex environmental microbial populations is currently unknown. Yet, insights from the ‘epitranscriptomics’ of cells in culture which reveal responses to new environmental conditions within timescales of minutes, foreshadow not only the complexity of such signal in natural habitats but also the utility of ‘metaepitranscriptomics' of microbial communities to study temporal and spatial impact and magnitude of change. Since the codon composition of genomes and the tRNA genes they encode form a natural framework to study dynamics of tRNA transcripts, there is an inherent yet unexplored connection between popular ‘omics approaches and tRNA sequencing. Thus, the inclusion of metaepitranscriptomics into existing software platforms that currently offer data integration opportunities for popular ‘omics approaches (Eren et al., 2021) is critical to bring a currently missing perspective into microbial lifestyles by extending the computational tools available to environmental microbiologists and enable deeper insights into microbial life.

Figure 4. Bacterial tRNA modifications.

Modified positions are shown in yellow with the corresponding modification abbreviations. Anticodon nucleotides are in gray. Modification types and sites data from Modomics (Boccaletto et al. 2018).

In summary, current studies on dynamic tRNA modifications have already generated many exciting insights into how biological systems utilize them for their adaptive benefits. As an emerging area, future research shall reveal unprecedented and new insights not only for the biology of single organisms but also for the microbial and microbial-host communities.