tRNAs are fundamental components of translation and emerging evidence places them more centrally in various other cellular processes. However, rather than being uniformly conserved, tRNA abundance is instead highly variable and adaptable. The amount of tRNA genes greatly differs among species. Moreover, even within the same genome, tRNA abundance shapes the proteome in a tissue- and cell-specific manner and is dynamically regulated in response to stress. Here, we review approaches for identification and quantification of tRNAs and their functional integrity. We discuss the resolution of each method and highlight new approaches with cell-wide resolution based on deep-sequencing technologies.
tRNAs are the most abundant ribonucleic entities among the small non-coding RNAs (4%–10% of all RNA). tRNAs are the connecting molecules between the nucleic acid moieties and the newly synthesized peptide chains. This dual function has shaped the structure of tRNAs; with the single-stranded anticodon loop, they pair to a codon of the mRNA and their 3′-termini are covalently extended with an amino acid (Figure 1). Besides their canonical role in translation, tRNAs participate in stress signalling, adaptive translation, protein degradation, cell membrane and wall modifications and priming of viral transcription . Various complex human pathologies are linked to mutations in tRNA genes or factors participating in their biogenesis [1,2].
The ‘clover leaf’ structure and identity elements of tRNA
The number of the tRNA genes among different organisms varies widely and increases with the complexity of the organism . Emerging evidence suggests that not all tRNA genes but rather different tRNA sets are expressed in various cell types most likely to respond to different proteome needs [4,5]. Thus, it is essential to define the sets of tRNAs expressed in each tissue and address their dynamics in response to stress and adverse environmental conditions. Recent reviews revealed unprecedented complexity of tRNA biosynthesis, modification patterns, regulation and function [1,6–8]. Here, we focus on the approaches for tRNA identification and quantification. We also discuss methods to determine tRNA aminoacylation level and the translation competence of tRNA.
Narrow physiological constraints shape tRNA structure
tRNAs are synthesized as precursors and processed in a sequence of maturation events which differ for eukaryotes and prokaryotes [6,8]. Mature tRNAs are prepared for their classic function in translation by attachment of an amino acid by the cognate aminoacyl-tRNA synthetases (aaRSs) to their common 3′-CCA ends (Figure 1), which by prokaryotes are encoded in the tRNA gene and in eukaryotes are enzymatically added post-transcriptionally. Among all RNA entities, tRNAs undergo by far the greatest number of post-transcriptional modifications [9,10]. Modifications in the stem loops are crucial for structural integrity and stability of the tRNA or as recognition for the aaRSs, whereas those in the anticodon loop maintain accuracy of decoding [6,11,12]. Modifications of the anticodon nucleosides (particularly at position 34, Figure 1) are also associated with increasing the diversity of the codon recognition through non-Watson–Crick base pairing between the third base in the codon and the first in the anticodon loop (i.e. wobbling), so that one tRNA can decode more than one codon [6,13]. Consequently, the number of tRNA iso-acceptors (that are different tRNA species with distinct anticodon sequence but carrying the same amino acid) to decode all 61 sense codons are usually much fewer than 61 .
The actual number of nuclear-encoded tRNA genes, particularly in the eukaryotes, is highly divergent; in humans, for example 513 nuclear-encoded tRNA genes encode 49 iso-acceptors . Each tRNA iso-acceptor is frequently encoded by an entire family of iso-decoders, which bear the same amino acids and anticodon but differ in the sequence of the tRNA body . Although the role of the tRNA iso-decoders remains unknown, they are not a result of neutral genome expansions of large genomes; different iso-decoders are expressed in tissue-specific manner and shape the tissue-specific tRNA sets [15,16].
The primary function of the majority of tRNAs is translation and interaction with the ribosome. Thus, in order to fit to the same ribosomal site, their length (e.g., tRNA length varies in a very narrow range of 71–89 nts in humans) and 3D architecture are constrained by common identity rules . This high sequence, length and structural homology impede the identification and quantification of iso-decoders.
Detection of single tRNA species
One of the pioneering techniques to identify tRNAs is northern blot hybridization , in which each tRNA can be separately detected with a radioactively- or fluorescently-labelled full-length tDNA probe. Since uncharged tRNAs have higher electrophoretic mobility than their aminoacylated counterparts, the method is suitable to determine both species [18–20]. Despite its common use, it is rather laborious approach; only one tRNA can be determined in one northern blot. Furthermore, the method relies on hybridization and for each tRNA the optimal conditions for hybridization of the probe need to be established.
Quantitative RT-PCR (qRT-PCR) is a fast and specific method to detect nucleic acids. However, the reverse transcriptase is sensitive to the modifications in tRNA. The latter slow down or even completely arrest the reverse transcriptase and impede tRNA quantification. Thus, it might not be a method of choice for tRNA identification. A new twist of the qRT-PCR technology, four-leaf clover qRT-PCR (FL-PCR) , increases the applicability of reverse transcription-based approaches to determine modified nucleic acid entities. The method consists of three steps: (i) tRNA deacylation, (ii) ligation of DNA/RNA hairpin adapter which complements the 3′-NCCA ends of the mature tRNAs (similar to the one used in the tRNA microarrays , see below, Figure 2A) to generate a ‘four-leaf clover’ secondary structure and (iii) reverse transcription of the DNA–RNA hairpin using forward and reverse primers derived from the T- or D-stems (Figure 1). The T4 RNA ligase 2 (Rnl2) ligates the DNA/RNA adaptor at a 3′-OH/5′-phosphate nick present only in mature tRNAs, which ensures the specificity towards mature and not to pre-tRNAs or tRNA fragments . Moreover, the reverse transcription in FL-PCR is quantitative, as it transcribes the least modified parts of each tRNA, i.e. the acceptor stem. A drawback of FL-PCR is the high number of cells needed to extract tRNA, since the efficiency of the organic solvent extraction of structured small RNAs decreases by using a small number of cells .
Labelling of tRNAs
Modifications: another obstacle for tRNA quantification
The complete set of modifications is known only for few organisms [22–24]. In eukaryotes, more than 100 modifications have been identified so far . Since modifications influence quantification, the tRNA quantification should be coupled to identification of the modification pattern. tRNA modifications can be identified de novo by coupling enzymatic digestion to LC–MS/MS [25,26] or ESI–MS . With a cocktail of various nucleases, tRNAs are digested down to single nts and modified bases are identified by the retention times in LC and confirmed by their unique mass-to-charge (m/z) ratios with an MS/MS. The majority of the modifications are visible by this method except pseudouridine . To obtain further site- and sequence-specific information on the modified nucleoside, tRNAs are partially digested with a nuclease cocktail (for example RNase T1, U2 and A) and the fragments identified by MS/MS are mapped against the tRNA sequences [25,26]. Using this approach the tRNA modification patterns of Escherichia coli , Lactobacillus lactis  and yeast  have been identified. Some modifications may escape this detection due to inability to detect them in the positive ion mode of the MS/MS approach. For example in yeast, 2′-O-ribosyladenosine phosphate [Ar(p)] is invisible to this method most probably because of the negatively charged phosphate . This combined approach for detecting tRNA modifications requires knowledge on the tRNA sequences and is therefore only applicable for sequenced genomes.
Global attempts to quantify the tRNA concentration and aminoacylation level
2D gel electrophoresis is a classical method to analyse the amounts of all cellular tRNAs. Upon complete denaturation (7 M urea), tRNA iso-acceptors are firstly separated according to their molecular mass, followed by separation based on the secondary structure upon a partial refolding (4 M urea). This approach is suitable for quantifying small tRNA sets: for E. coli 44 out of 46  and for B. subtilis 30 out of 35  iso-acceptors were detected. tRNA of more complex organisms, e.g. multicellular eukaryotes, can be identified only partially.
Approaches using HPLC and MALDI–MS have been developed which are semi-quantitative with limited resolution for quantification. Using the HPLC methodology, differences in the tRNA pattern of proliferating compared with quiescent rat liver cells have been detected , although the method is restricted in detecting only iso-acceptor families (i.e., a family of all tRNAs carrying the same amino acid). Digestion of tRNAs with various RNases releases characteristic fingerprint products for each tRNA which then can be identified in MALDI–MS . Although highly specific, the resolution of MALDI–MS is limited to approximately 30 tRNA species.
Exploiting the advantages of microarray technology, Pan and co-workers [19,32] detected differences in the expression of tRNA set in various human tissues. Each tRNA is detected with its own tDNA probe. A radioactive labelling of the tRNAs at their 5′ or 3′-termini is used for their absolute quantification, whereas labelling with a fluorescent hairpin oligonucleotide at the common 3′-CCA ends (Figure 2A) is informative on relative changes of single tRNA iso-acceptors between two different conditions (thereby using hairpin oligonucleotides that bear two different fluorophores). The microarray approach does not provide a complete coverage of all tRNA iso-acceptors; only tRNAs with difference of at least eight nts can be reliably distinguished with this technology (in contrast, tRNAs that vary only by one or few nt, i.e. proline iso-acceptors with anticodons IGG, CGG and UGG, hybridize to the same probe on the microarray) [19,32]. Despite these limitations in resolution, divergent tRNA pools in proliferating and differentiating cells were detected with this approach . Furthermore, modifying the probes on the microarray (i.e. using fragments of full-length tRNAs) allows detection of stress-inducible tRNA-derived fragments  with crucial roles in stress signalling [34,35].
A twist of the microarray technology allows for evaluating the aminoacylation level of individual tRNAs. For this, the total tRNA is isolated under conditions to preserve the aminoacyl-tRNA (i.e. acidic buffers with pH of 4.8) and oxidized with periodate (Figure 2B). The aminoacyl moiety maintains the 3′-CCA ends intact, which upon deacylation ligate the fluorescently-labelled hairpin oligonucleotide used for detection. Deacetylated tRNAs are sensitive to oxidation and cannot be detected (Figure 2A).
The power of the deep-sequencing technologies has been used to determine the the entire set of tRNAs, tRNAome [24,36]. Because in its core it is based on reverse transcription of each tRNA to DNA, it bears the drawbacks of qRT-PCR. In particular, modifications that interfere with the quantitative PCR-based cDNA synthesis (such as N1-methyl-A and N1-methyl-G ) and the extensive tRNA structure make tRNA-Seq (tRNA sequencing) an inadequate approach for quantification, but rather are suitable for identification . Often, modifications are misread by the reverse transcriptase which appears as a mismatch by aligning the tRNA-derived reads. Allowing mismatches at the position of each modification when aligning the sequencing reads improves the quantification. This strategy allowed for quantification of 47 out of 49 tRNA species  and showed a reasonable overlap with earlier quantification of E. coli tRNAs by 2D gel electrophoresis . The applicability of this mapping strategy in the tRNA-Seq is, at least currently, limited, despite its large depth, since it requires knowledge of the exact position of all modifications to be considered by mapping of the sequencing reads. Moreover, mismatches in the sequencing reads can be used to identify putative positions of nt modifications, but since little information is available on the identity of the nt mismatch, it is not possible to determine the type of modification using the identity of mismatch nt .
Some reverse transcriptases have enhanced read through on modified ribonucleotides. Using a thermostable group II intron reverse transcriptase for the RNA-Seq library construction  have facilitated the generation of full-length cDNA copies of tRNA and enhanced the quantification potential of the tRNA-Seq technology. Moreover, the enzyme is insensitive to aminoacyl moiety at the 3′-terminus [40,41] and thus the results are reproducible independent on the tRNA isolation protocol (i.e., by acidic or alkalic pH, to preserve or not the aminoacyl group). Recent development has combined the power of the thermostable group II intron reverse transcriptase with engineered demethylases from E. coli, wild-type alpha-ketoglutarate-dependent dioxygenase, AlkB and D135S AlkB, to remove methylations at N1-methyladenosine, N3-methylcytosine and N1-methylguanosine . Although many bulky modifications still remain, the demethylase treatment has already resulted in longer transcripts which is crucial for adequate mapping, particularly of tRNAs of higher eukaryotes which differ only be few nts.
A recent development of the tRNA-Seq using two-step approach minimizes the influence of modifications and improves the applicability of the deep sequencing approaches to quantify tRNA sets . The crucial improvement of this approach is that it uses sequencing information of ∼30 nt from the 3′-end of tRNAs, which contain the fewest modified ribonucleosides in all tRNAs. Briefly, in the first step a pre-adenylated 20-nt long DNA linker (with a 3′-dideoxy end) is ligated to tRNA and the DNA linker is used as a priming site for the reverse transcription . This first reverse-transcription round minimizes the modification-induced fall-off or pausing of the reverse transcriptase to generate a quantitative ctDNA set that is then subjected to another round of linker ligation at the new 3′-end and subjected to deep sequencing . Although only the most 3′-ends of the tRNAs are reverse transcribed and amplified, the unique signature of each tRNA is however preserved in these 3′-end-derived fragments enabling quantification of all 76 uniquely expressed tRNAs in Saccharomyces cerevisiae, including also some iso-decoders . Similar, however, to other deep-sequencing approaches, the concentration of single tRNA species varies between biological replicates which can be attributed to variations in the efficiency of the ligation and PCR amplification steps .
With advances in experimental technologies, we are beginning to understand the variety of cellular functions of tRNA and different programmes that operate to co-ordinate their expression in tissue-specific fashion and the dynamics of tRNA as another regulatory layer to co-ordinate stress response. Mutation in tRNA genes and tRNA-biogenesis genes are linked to several human pathologies. Variations in tRNA abundance among different tissues may modulate the effect of tRNA-linked pathologies in a tissue-specific manner, underscoring the need for accurate quantification of tRNA expression and modification pattern for each specific cell type. This is a key area of study that will enable us to understand more clearly the genotype–phenotype relationship.
Furthermore, addressing differences in tRNA expression among various organisms will allow for improvement of the heterologous expression by tailoring the translation profile of the heterologous protein to the host-specific tRNAome.
This work was supported by the Deutsche Forschungsgemeinschaft [grant number FOR 1805 (to Z.I.)]; and the German Federal Ministry of Education and Research [grant number NatLife 2020, FKZ 031A206 (to I.F. and Z.I.)].
Translation UK 2015: Held at the University of Aberdeen, U.K., 7–9 July 2015.