Variations in transfer and ribosomal RNA epitranscriptomic status can adapt eukaryote translation to changing physiological and environmental conditions.

The timely reprogramming of gene expression in response to internal and external cues is essential to eukaryote development and acclimation to changing environments. Chemically modifying molecular receptors and transducers of these signals is one way to efficiently induce proper physiological responses. Post-translation modifications, regulating protein biological activities, are central to many well-known signal-responding pathways. Recently, messenger RNA (mRNA) chemical (i.e. epitranscriptomic) modifications were also shown to play a key role in these processes. In contrast, transfer RNA (tRNA) and ribosomal RNA (rRNA) chemical modifications, although critical for optimal function of the translation apparatus, and much more diverse and quantitatively important compared to mRNA modifications, were until recently considered as mainly static chemical decorations. We present here recent observations that are challenging this view and supporting the hypothesis that tRNA and rRNA modifications dynamically respond to various cell and environmental conditions and contribute to adapt translation to these conditions.

Introduction

Co- or post-transcriptional modification of RNA is an evolutionarily conserved process that drastically increases the biological potential of these crucial molecules. In all life kingdoms, at least 143 distinct chemical modifications can occur on RNA molecules, 111 on transfer RNAs (tRNAs), 33 on ribosomal RNAs (rRNAs) and 17 on messenger RNAs (mRNAs) [1]. Considering only eukaryotes, these numbers are 68, 21 and 17 for respectively tRNAs, rRNAs and mRNAs. In eukaryotes, mRNA epitranscriptomic marks have a dramatic impact on mRNA splicing, transport, stability, storage and translation and can regulate genes involved in development and stress responses [2,3]. The chemical modification of mRNAs is a dynamic process, with enzymes responsible for setting up (writers), removing (erasers) and reading (readers) marks in response to various developmental and stress conditions [2,3].

Compared to mRNA, tRNA modifications are much more diverse, ranging from simple modifications (i.e. methylation of the nucleobase or the sugar moiety), to the addition of complex compounds (i.e. isopentenylation of adenosine or the formation of cyclopentendiol derivates or imidazopurines from guanosine). tRNA biogenesis and functions, including tRNA maturation, stability, structure, aminoacylation, interaction with ribosomes and mRNA decoding properties, can be modulated by these chemical modifications [4]. Not all tRNA ribonucleotides are equal in terms of modifications as some positions in the T-loop (54 and 55), the D-loop (16 and 20) and the anticodon-loop (32, 34 and 37) are more frequently targeted. The most complex modifications are mainly found in the anticodon-loop region while modifications in the tRNA core are usually simpler. On average, 17% of tRNA ribonucleotides are modified (representing 13 modifications on each tRNA) [5].

From 2% to around 3% of all ribonucleotides of the four eukaryote cytosolic rRNAs (28S (25S in yeast, 26S in nematodes), 18S, 5.8S and 5S), representing over 100 nucleotides in Saccharomyces cerevisiae (yeast) and 200 in human, are modified co- or post-transcriptionally [6]. Most of these modifications are 2ʹ-O-methylation of ribose (to generate 2ʹ-O-methyladenosine (Am), 2ʹ-O-methyluridine (Um), 2ʹ-O-methylguanosine (Gm) or 2ʹ-O-methylcytidine (Cm)) or uridine conversion to pseudouridine (Ψ), and are largely contributing to the biogenesis and stabilization of ribosomes [7]. Almost all rRNA modifications are found on either 18S and 28S rRNAs while 5.8S and 5S rRNAs are not modified or presenting a small number of modified positions. For example, in yeast, 5.8S rRNA is not modified and a single position is converted to Ψ in 5S rRNA [6]. In contrast with yeast, human 5S rRNA is not modified but four modifications (Um, Gm and two Ψ) are present on 5.8S rRNA [8]. The majority of 2ʹ-O-methylations and pseudouridylations are guided by C/D or H/ACA snoRNPs [9]. Other known rRNA modifications include 5-methylcytidine (m5C), 6-methyladenosine (m6A), N6,N6-dimethyladenosine (m6,6A), 7-methylguanosine (m7G), 1-methyladenosine (m1A), N4-acetylcytidine (ac4C), and 3-methyluridine (m3U) [1]. Some of these modifications (such as 25S/28S rRNA NOP2/NSUN1-dependent m5C [10] or the RRP8-dependent m1A [11] modifications) are clearly important for normal ribosome biogenesis, others (like human m6A or yeast m3U modifications) have no clear impact on this process [12,13]. In general, little is known on the impact of these modifications on ribosome function, although many of them are evolutionarily conserved in eukaryotes and localized within or near catalytic centres, suggesting that they should play important roles in translation [6].

A first line of evidence supporting the hypothesis that tRNA and rRNA modifications are not static comes from the observation that these chemical decorations vary in human diseases including cancer [14-22]. For example, two recent studies revealed a global reduction of rRNA 2ʹ- O-methylation at some rRNA sites (named variable sites) in two types of human tumours compared to their corresponding level in normal tissue [21,22]. These new cancer-related hypomethylated patterns were found to be tumour-specific and associated with tumour aggressiveness [21,22]. In contrast, the global level of tRNA modifications is generally higher in cancer cells compared to their corresponding level in normal tissue [23]. Based on these variations, both rRNA and tRNA modifications are now proposed as useful prognostic markers for cancer [14,21,22,24,25]. More generally, eukaryote tRNA modification levels have been shown to vary during cell cycle and in response to environmental stresses [5,26-31]. Although the kinetic of these variations has not been investigated in vivo, the fact that they can be observed from 15 min to 1 h after yeast cell exposure to stressing agents suggests that they result of either increased enzyme activity and/or variations in tRNA copy numbers [27,28,30,32,33]. For example, yeasts grown at high temperature (37°C) or exposed for 15 min to hyperosmotic conditions (0.4 M NaCl) show changes in their global tRNA modification levels, with induction of new chemical marks and elimination of others [28]. A global analysis of tRNA marks revealed that a combination of 14 modified ribonucleotides has strong predictive power to distinguish exposure of yeast to oxidative or alkylating agents, in a manner similar to transcriptional, proteomic and metabolomic profiling [30], although the exact profiles of stress-induced tRNA variations can be different among different yeast strains [34]. In plants, tRNA modification levels were also shown to vary in different stress situations. For example, in Arabidopsis thaliana (Arabidopsis) and Oryza sativa (rice), several tRNA marks increase in response to 1–5 days of exposure to drought, salt or cold temperature conditions [31]. Also, while tRNAs are stable molecules, with estimated half-lives from 9 h to days, stress-induced tRNA modifications can affect their stability in one way or the other [32,35].

The dynamic, and not static, nature of tRNA and rRNA modifications suggests that it represents a sensing system linking cellular and environmental stimuli to translation and metabolism. For this system to work, variations in tRNA and rRNA chemical status must be exploited to adapt ribosome mRNA decoding potential to particular cellular conditions. One way to do this is to target tRNA subsets, especially in the anticodon loop [26,29,30,36,37], in order to better translate physiologically relevant mRNAs preferentially decoded by these tRNAs. Another way to modulate translation is to synthesize new ‘specialized’ ribosomes with variable levels of key rRNA modifications [6,38]. These specialized ribosomes could then preferentially bind and translate mRNA subsets involved in responding to specific cellular and/or environmental conditions [39,40]. We review here evidence supporting the idea that variations in tRNA and rRNA epitranscriptomic marks are not only passive consequences of different cellular conditions but can fine-tune translation to adapt cellular activities and the physiology of organisms to environmental changes. The principal tRNA- and rRNA-modified positions discussed in the following sections are presented in Figure 1, and a short summary of major information concerning these positions and their impact on translation is presented in Supplementary Table S1. In addition, Figure 2 gives an outline of the translation adjustment through tRNA and rRNA epitranscriptomic landscape modifications leading to adaptation to new environmental conditions.

Figure 1.

Schematic representation of tRNA and rRNA molecules with the position of major modifications discussed in the main text: 5-methylcytidine (m5C), 3-methylcytidine (m3C), 1-methyladenosine (m1A), 6-methyladenosine (m6A), 1-methylguanosine (m1G), 7-methylguanosine (m7G), queuosine (Q), wybutosine (yW), 5-methoxycarbonylmethyluridine (mcm5U), 5-methoxycarbonylmethyl-2-thiouridine (mcm5s2U), Inosine (I), Pseudouridine (Ψ), 1-methylinosine (m1I), N6-threonylcarbamoyladenosine (t6A), N6-isopentenyladenosine (i6A) and 1-methyl-3-amino-carboxyl-propyl pseudouridine (m1acp3 Ψ). Positions of each modified tRNA ribonucleotides are indicated by their corresponding number. For 18S and 28S rRNAs, only close-ups of the different regions are presented. The numbering of the different 18S and 28S positions and helices (H) are taken from human sequences. A short summary of key information concerning these modifications is presented in Table S1

Figure 2.

Reprogramming of tRNA and rRNA epitranscriptomic landscape to fit nutritional variation and stress condition. Nutrient availability and/or various stress conditions can reprogram the tRNA and/or rRNA epitranscriptomic landscape. This epitranscriptomic reprogramming than contributes to focus translation on different subsets of mRNAs facilitating acclimation and/or adaptation processes. Mechanisms by which various changes in tRNA and rRNA epitranscriptomic marks impact translation are summarized in Table S1

Stress-modulated tRNA modifications and impact on translation

A key tRNA function is to decode mRNAs in the context of ribosome-directed translation. The 20 amino acids are encoded by 61 sense codons so that, for each amino acid, several ‘isoacceptor’ tRNAs (i.e. charged with the same amino acid but having distinct anticodons) are required. For example, the plant Arabidopsis has 597 tRNA genes, producing 198 unique sequences and 46 isoacceptors [41]. tRNA position 34 is named ‘wobble’ as it allows non-Watson and Crick parings with codon’s third positions. In yeast, stress-responsive mRNAs named MoTTs for Modified Tunable Transcripts are biased for codons that necessitate the presence of corresponding tRNAs decorated with the proper chemical modification at the wobble position to be efficiently decoded [26,29,30,36,37]. Stress-induced variations of chemical modifications at the wobble position can therefore directly impact MoTTs (as well as other mRNAs) translation elongation speed [42], potentially influence co-translational mRNA decay [43], and affect the capacity of yeasts to survive stress. Apart from position 34, other stress-responsive modifications in the anticodon loop, such as the ones in positions 32 and 37, can also directly impact translation. Finally, modifications outside the anticodon loop can also affect cell translational capacities during stress. The next sections will focus on the description of these stress-responsive tRNA modifications involved in modulating the translational potential of cells.

Stress-responsive modifications of tRNA ‘wobble’ position 34

Uridine in position 34 of eukaryote tRNAs almost invariably carries a modification. In many instances, U34 is transformed to 5-methoxycarbonylmethyluridine (mcm5U) or 5-methoxycarbonylmethyl-2-thiouridine (mcm5s2U) by a complex enzymatic process [1]. The elongator complex, composed of six subunits (ELP1–ELP6), is needed to transform U34 into 5-carboxymethyluridine (cm5U) [44]. Next, the TRM9/TRM112 complex can methylate cm5U34 to form mcm5U34 [45]. Finally, thiolation of mcm5U34, to generate mcm5s2U34, is a multistep process involving the ubiquitin-related protein modifier 1 (URM1)-like proteins URM11 and URM12, and the CTU1/CTU2 complex [46]. tRNA arginine (UCU), glycine (UCC), glutamine (UUG), glutamic acid (UUC) and lysine (UUU) can be mcm5U-modified while the last three tRNAs can be further thiolated [15,37]. The loss of ELP3, the catalytic subunit of the elongator complex [47] or TRM9 [48] completely abolishes the presence in tRNAs of both mcm5U34 and mcm5s2U34. This situation is embryo lethal in Drosophila melanogaster (drosophila) and mouse [49,50], which results in developmental defects in Caenorhabditis elegans (nematode) and plants [51,52] and generates stress hypersensitive phenotypes in yeast [37,42]. Alternatively, overexpressing Arabidopsis ELP3 or ELP4 in tomato and strawberry enhances resistance to pathogens [53,54]. In yeast, the loss of TRM9, drastically reduces the amount of polysomal mRNAs enriched in AGA and GAA codons, and corresponding protein production, without affecting global translation [37]. Deficiency in mcm5U34 modification also leads to amino acid incorporation errors due to improper pairing of arginine tRNAs (UCU) with near-cognate serine codons in the ribosome [37]. Mutation of URM11, URM12, CTU1 or CTU2 prevents the formation of mcm5s2U34 [55,56]. In yeast, this results in a slower propagation speed at high temperature of mutant stains compared to wild type (wt) [57]. More generally, the level of U34 thiolation is closely associated with thermotolerance properties of different yeast strains [34]. Dysfunction of CTU1 in nematodes also leads to a thermosensitive phenotype [46]. In Arabidopsis, the urm11/urm12 double mutant is more sensitive to drought stress and produces leaf cells with altered ploidy levels and less chlorophyll content [55,56]. In rice, mutation of CTU2 impairs the heat-stress response while its overexpression enhances tolerance to high temperature [58]. Also, in both rice and Arabidopsis, the ctu1 or ctu2 mutations are associated with a root-deficient phenotype [58-60].

In yeast, mcm5U34 and mcm5s2U34 levels vary in the presence of oxidative and alkylating agents or in heat and salt stress conditions [5,27-29]. Variations in the level of U34 modifications in the five concerned tRNAs have been shown to impact translation of stress-responsive mRNAs enriched in their cognate codons [37,61]. This is also true in Schizosaccharomyces pombe where, in response to H2O2, mcm5s2U34-containing tRNA lysine (UUU) better translates AAA codon-rich stress-responsive genes [62]. In addition, a global reduction of yeast mcm5U34 and mcm5s2U34 levels in tRNA lysine (UUU), glutamine (UUG) and glutamic acid (UUC) induces ribosome pausing at cognate codons for a subpopulation of mRNAs and triggers the proteotoxic stress response [37,42]. The level of U34 thiolation is also critical in yeast to adjust translation and growth to the amount of sulphur amino acids [63]. Under nutritional stress, low levels of sulphur-containing amino acids directly impact negatively the thiolation status of U34. This results in the reduced translation of key mRNAs coding for important translation and growth factors that are enriched in lysine, glutamine and glutamate codons. This, in turn, slows down cell growth and stimulates the synthesis and salvage of methionine and cysteine [63].

Interestingly, a crosstalk between the Target of Rapamycin (TOR) pathway and mcm5U34 and mcm5s2U34 tRNA levels has been observed in yeast, human and plants [60,64-67]. In yeast, the urm11 mutant is hypersensitive to the TOR kinase inhibitor rapamycin and reduction in mcm5s2U34 levels impacts the TOR pathway negatively by a yet-to-describe feedback mechanism [64,65]. In human cancer cells, higher expression of CTU1 is associated with increased cell growth and TOR activation [67]. In Arabidopsis, ctu1 and ctu2 mutants as well as the urm11/urm12 double mutant phenocopy the root phenotype of hypomorph tor mutants [56,59]. Also, the Arabidopsis ctu1 mutant is hypersensitive to TOR inhibitors [60]. These results suggest that stress-induced variations of mcm5U34 and mcm5s2U34 could also impact translation initiation through the TOR pathway. Overall, these observations indicate that mcm5U34 and mcm5s2U34 deposition is likely an evolutionarily conserved process that regulates translation initiation and elongation of a subset of transcripts involved in stress, nutrition and development.

Uridine in position 34 (as well as 6 other positions) can also be transformed to Ψ by the stand-alone pseudouridine synthase PUS1 [68,69]. In yeast, the pus1 mutation causes a growth defect at high temperature. It also leads to synthetic lethality in combination with the loss of other pseudouridine synthases or in the presence of destabilized tRNA variants [69]. This suggests that PUS1-dependent modifications become essential when other aspects of tRNA biogenesis or modifications are disturbed. More specifically, Ψ in position 34 is proposed to stabilize codon-anticodon interactions, preventing the formation of unconventional pairs at non-synonymous near-cognate tRNAs [69]. The stabilizing property of Ψ is attributed to the presence of an extra NH moiety able to make more hydrogen bonds than uridine [70]. Also, Ψ in position 34 prevents 5-carmoylmethyluridine (mcm5U34) modification of intron-containing tRNA Isoleucine (UAU), a modification that would jeopardize its normal decoding capacity [71]. Although yeast tRNA Ψ levels have been shown to vary upon exposure to oxidative and alkylating agents [5,27] and during cold and heat stress [72], it is not clear at the moment if these modifications affect especially Ψ in position 34 (or other tRNA positions that are also converted to Ψ, see Table S1) and are used to reprogram translation in stress situations.

The nucleobase queuine (q) is a cyclopentendiol derivative of 7-aminomethyl-7-deazaguanine. This micronutrient, and its corresponding nucleoside queuosine (Q), cannot be synthesized by eukaryotes and must be acquired from the environment [73]. In most eukaryotes (with a few exceptions, such as S. cerevisiae and plants from the Brassicaceae family including Arabidopsis [74,75]), Q replaces G in position 34 of tRNAs having a GUN anticodon. The level of Q-tRNAs was shown to vary during development and the absence of Q-modified tRNAs is concordant with, and relevant to, the replicative undifferentiated cellular state. Accordingly, tRNAs of human primary tumours are hypomodified with respect to Q, with decreased levels correlating with disease progression and poor patient survival such that Q hypomodification is proposed to be a deliberate and advantageous adaptation of cancer cells [76]. In mice, Q modification of tRNA is also required for normal tyrosine production such that animals made deficient in Q died within 18 days of withdrawing tyrosine from the diet [77]. Queuine deficiency also impacts the activity of several antioxidant systems [78] and phosphorylation levels of tyrosine phosphoproteins involved in cell signalling [79]. In plants other than Brassicaceae, the degree of Q-modification in tRNAtyr(GUA) impacts the capacity of this tRNA to suppress UAG termination codons present on tobacco mosaic virus mRNAs [80]. All these observations suggest that Q modification of tRNA may affect severalf biological processes through broad changes in protein translation profiles.

The exact molecular impact of Q-tRNA on translation is still debated and possibly varies among species. In Schizosaccharomyces pombe, Q modifications enhance the translational speed of C-ending codons for aspartate (GAC), histidine (CAC), asparagine (AAC) and tyrosine (UAC), and reduce that of synonymous U-ending codons thus equilibrating the genome-wide translation of these codons [81]. Furthermore, Q prevents translation errors by suppressing second-position misreading of the glycine codon GGC. The absence of Q causes reduced translation of mRNAs involved in mitochondrial functions, and accordingly, lack of Q modification causes a mitochondrial defect [81]. In drosophila, the quantity of Q-containing tRNAs was shown to vary across development and translation to be adapted to these variations [82]. For example, at the embryonic stage, the level of Q-tRNA is high, and gene highly expressed at this stage are enriched in C-ending codons for fast and very accurate translation. However, at the larval and pupal stages, when low levels of Q-tRNAs are available, U-ending synonymous codons are favoured in highly expressed genes. In mammals, the translation of Q-decoded codons is slowed down in the absence of Q modifications [83]. This dysregulation of translation results in the accumulation of misfolded proteins and aggregates that triggers the activation of endoplasmic reticulum (ER) stress and the unfolded protein response. Consistent with reduced rates of protein translation, Q-deficient mice had a substantially reduced body weight. Overall, these observations reveal a route by which environment nutrients (including those generated by the gut microbiome of animals or the endophytic bacteria of plant tissues) can adapt protein translation. Furthermore, it is interesting to note that methylation by the DNMT2 methyltransferase of cytidine in position 38 of tRNAs to generate m5C is strongly stimulated by the Q34 modification in S. pombe [84]. As m5C38 has been shown to have a function in the control of tRNA cleavage (see below) and translational accuracy [85], this observation suggests another way by which nutritional factors could modulate mRNA decoding and translation. Also, drosophila lacking DNMT2 are more sensitive to heat and oxidative stress, further suggesting a role for Q34, in combination with m5C38, in stress tolerance at least in this species [86].

In addition to uridine and guanosine, cytidine at the tRNA wobble position 34 is another stress-sensitive ribonucleotide. Cytidine in position 34 can be modified to m5C by the action of TRM4(NSUN2). In yeast, m5C34 increases following oxidative stress, leading to the selective translation of UUG-enriched mRNAs among which are represented ribosomal and stress responsive genes [87]. Accordingly, yeast trm4 mutants are hypersensitive to H2O2 [87]. The global tRNA m5C level also varies in yeast following exposure to alkylating agents and during heat and salt stress [5,27,28] and following plant exposure to cold, drought and salt [31]. However, since m5C can be introduced in six tRNA positions (34, 38, 48–50, 72), it is not known to which proportion these variations affect position 34. In nematodes, m5C34 modification of tRNA leucine (CAA) facilitates the translation of leucine UUG codons upon heat stress and supports the animal fitness at high temperature suggesting that this modification is involved in the adaptation to heat stress [88].

Finally, adenosine in position 34 can be deaminated to inosine (I) by the TAD2(ADAT2)/TAD3(ADAT3) complex. Inosine at position 34 (I34) expands the tRNA decoding capacity as it can pair with U-, C- and A-ending codons [89]. Lack of I34 modifications is associated with several human diseases [90] and, in plants, to slower chloroplast translation, thereby affecting development [91]. In yeast, the global tRNA inosine level varies in the presence of oxidative or alkylating agents [5,27,29,30], but it is not clear if these variations impact positions 34, 37 or both and how they affect translation.

Stress-responsive modifications of tRNA position 37

Position 37, with position 34, are the two major modified positions in tRNA anticodon loops. All tRNAs harbour a purine at position 37 that is often modified into more complex derivates. A modified base in position 37 is proposed to stabilize, by base stacking, weaker (A:U) interactions between tRNA’s position 36 and mRNA’s first codon position [92]. Also, modification of position 37 impairs interactions with position U33 that would otherwise negatively affect the anticodon loop [92]. Modifications at position 37 are mainly known to be important for proper mRNA decoding and to prevent frameshift during translation [93].

N6-threonylcarbamoyladenosine (t6A), N6-isopentenyladenosine (i6A), as well as related ribonucleotides [1], are universally conserved stress-sensitive modifications found at position 37 that are crucial to translational accuracy. t6A37 is found in nearly all tRNAs that decode ANN codon [94] and its general level was shown to vary following yeast exposure to oxidative and alkylating agents [5,27], as well as during heat stress [28]. In drosophila, variations in the amount of t6A37 impact protein synthesis homoeostasis and can favour or inhibit translation of specific open reading frames [93]. In human cell lines, the amount to t6A37 is regulated by the intracellular levels of CO2 and bicarbonate, and hypomodification of t6A37 in mitochondrial tRNAs was shown to downregulate mitochondrial translation in a codon-specific manner [95]. Variations in tRNA t6A37 content in relation to CO2 levels are proposed to regulate oxidative phosphorylation under hypoxic conditions, a process particularly important for solid tumour cell proliferation [95]. As such, t6A37 is used as a prognostic marker for breast cancer [24]. Surprisingly, lowering the proportion of t6A37-modified initiator methionyl-tRNA (tRNAiMeti) in drosophila was shown to downregulate TOR kinase activity, inhibiting translation and growth [96]. This led to the suggestion that t6A37-modified tRNAiMeti could be a limiting factor for growth that is under the control of external stimuli [96].

i6A37 is another important tRNA stress-sensitive modification, helping to properly decode the first codon position forming an A:U or U:A base pair [92]. In yeast, i6A37 levels vary following salt and heat stress and exposure to oxidative and alkylating agents [5,27,28,30] and strains deficient in i6A37 fail to sporulate [97]. In S. pombe, strains deficient in i6A production present a mitochondrial dysfunction, mainly related to a lower content of i6A-modified cytosolic tRNA tyrosine [98], and a reduced amount of several polysomal mRNAs enriched in i6A-dependent codons [99]. Also, these strains are hypersensitive to rapamycin suggesting again, as for t6A, mcm5U and mcm5s2U, the existence of a crosstalk between the level of i6A-containing tRNA and the TOR pathway [99]. Accordingly, in nematodes, the loss of i6A results in slower growth and development [100].

In plants, tRNA isopentenyltransferase 2 and 9 (IPT2 and IPT9) can generate i6A in position 37 of tRNA recognizing codons beginning with U [101]. AtIPT2 and 9 are indispensable for the biosynthesis of cis-zeatin, a stress-regulating plant cytokinin [101]. This is due to the fact that cis-zeatin can only be produced by the degradation of tRNAs containing the hydroxylated form of i6A (io6A) [101]. Cis-zeatin is important to maintain minimal cytokinin activity under growth-limiting conditions, including abiotic stress [102-104]. In these conditions, cis-zeatin replaces trans-zeatin for a lower promotion of cell division activities and an efficient set-up of the stress-responsive genetic programme. Heat, cold, drought and salt stress as well as nitrogen deficiency all lead to peaks of cis-zeatin with a strong decrease in trans-zeatin [102-104]. Accordingly, plants with higher levels of i6A and io6A in their tRNAs are proposed to be more resistant to abiotic stress as they can produce higher amounts of free cis-zeatin following stress-induced tRNA turnover [102]. Based on these observations, it has been suggested that plant stress tolerance could potentially be improved by increasing the content of endogenous i6A/io6A tRNAs [103].

It is interesting to note that the amount of tRNAs containing t6A and i6A (including their derivates) can also vary in relation to nutritional signals. The formation of t6A (and derivates) uses threonine, an essential amino acid that must be salvage from bacteria in eukaryotes. Therefore, as for Q, this is another way by which environmental nutrients can feedback on the regulation of protein translation. The biosynthesis of i6A (and derivates) depends on dimethylally pyrophosphate that itself is derived from acetyl-CoA [1]. Since acetyl-CoA levels depend on glycolysis or fatty acid beta-oxidation, this suggests that modifying tRNA i6A levels could be a way to adjust the translation of specific mRNAs to the cell metabolic status. Taking into consideration that Q34, mcm5s2U34 and m1A58 (see below) levels are also controlled by the cell nutrient status, a picture is emerging in which connections between tRNA modifications and the cell translational output is not limited to environmental exceptional situations but is also part of the cell’s normal metabolism and growth programmes [105].

Wybutosine (yW) is a complex stress-responsive guanine modification found only in position 37 of eukaryotes tRNA phenylalanine (GAA) [1]. In yeast, this modification requires the retrograde nuclear import of tRNA phenylalanine to be synthetized [106]. In yeast and plants, the level of yW depends on cell growth conditions [107,108] and, in yeast, varies upon exposure to oxidative and alkylating agents as well as following heat and salt stress [5,27,28,30]. These observations suggest that the level of yW is modulated in response to environmental conditions and the cell’s metabolomic requirements. yW is important to limit frameshifting particularly at ‘U’ stretches [107] and its loss results in a fourfold increase in −1 frameshift in yeast and animals [109,110]. It remains to be seen if the translation of some stress-responsive mRNAs could benefit from ‘programmed’ frameshifting as this is the case for many viruses that use this strategy to generate multiples viral proteins [110]. This ‘frameshifting potential’ that depends on the amount of yW37-modified tRNA phenylalanine was proposed as the evolutionary driving force behind the emergence of this modification from the m1G37 platform [105].

Deamination of adenosine in position 37 by TAD1(ADAT1) results in the formation of inosine (I) that can be further transformed to 1-methylinosine (m1I) by TRM5. TRM5 can also directly modify guanosine in position 37 to generate 1-methylguanosine (m1G). The I, m1I and m1G tRNA content was shown to vary upon yeast exposure to oxidative and alkylating agents [5,27,29,30] and, for m1I, in heat and salt stress conditions [28]. In plants, the tad1 mutant is hypersensitive to heat and cold stress [111] suggesting that I37 and/or m1I37 are playing a role in plant acclimation to these temperature variations. Also, in Arabidopsis, the trm5 mutant is slow-growing, late flowering, has reduced lateral roots, and accumulates fewer proteins involved in photosynthesis and ribosome biogenesis [112]. At the molecular level, the loss of AtTRM5 leads to aberrant protein translation and disturbed hormone homoeostasis. Since TRM5 is responsible for m1G37 and m1I37 methylation, it is not known if the loss of one, the other or both modifications are leading to the observed phenotypes. Nevertheless, based on the impact of m1G/m1I deficiencies at position 37, it was suggested that hypomodified tRNA would be unable to efficiently decode their cognate codons or induce frameshifts, resulting in a global reduction of protein output [113]. In yeast, TRM5 is downregulated by exposure to alkylating agents [30] and m1I was shown to oscillate throughout the cell cycle [114], suggesting a role for m1I in cell cycle regulation and response to at least this stress condition.

Stress-responsive modifications of tRNA position 32

In the anticodon-loop region, 3-methylcytidine (m3C) in position 32 was recently shown to be an important stress-sensitive modification directly regulating mRNA translation in yeast and mouse [30,115]. m3C occurs at position 32 of all tRNAs decoding serine and threonine and of two tRNAs decoding arginine, and its global level was found to vary upon yeast exposure to oxidative and alkylating agents as well as in heat and salt stress conditions [5,27,28,30]. The upregulation of m3C content upon yeast exposure to alkylating agents led to the selective translation of mRNAs enriched in four specific threonine codons, suggesting that these codons are differentially recognized by m3C-modified tRNAThr [30,116]. TRM140(METTL2) and TRM141(METTL6) are responsible for the m3C32 modification [115,117]. In human, METTL6 was identified as a crucial regulator of tumour cell growth, and its deletion in mouse stem cells results in changes in mRNA ribosome occupancy and impairs pluripotency [115]. These results suggest that m3C32 is a key modification required to adapt translation to various cell growth and stress conditions. Interestingly, the m3C32 modification of three S. pombe tRNAs is dependent on the synthesis of i6A37 [118], suggesting that these two modifications can be interconnected. Also, in Trypanosoma brucei, m3C32 can be further converted to m3U32 by the action of the TAD2/TAD3 complex [119].

Stress-responsive modifications outside of the tRNA anticodon-loop region

Not all stress-sensitive tRNA modifications directly impacting translation occur in the anticodon-loop region. One example is the modification by the TRM61/TRM6 complex of tRNAiMeti adenosine 58 to generate m1A. In all eukaryotes, this modification is critical to ensure the stability of tRNAiMeti and is a way to control translation initiation [4,5,120,121]. In mammals, the ALKBH1 demethylase can demethylate m1A58 in response to variations in nutritional conditions [122]. For example, in glucose deprivation condition, ALKBH1 expression is up-regulated leading to a reduction of m1A58, tRNAiMeti and translation [122]. On the contrary, the knockdown of ALKBH1 results in higher m1A58 levels of specific tRNAs (including tRNAiMeti) and favour translation initiation and elongation from corresponding codons [122]. In addition of being more stable, m1A58-methylated tRNAs are preferentially recognized and delivered to actively translating ribosomes [122]. In eukaryote, m1A can also be present at two other tRNA positions (9, and 14) [1]. In yeast, global m1A levels vary upon exposure to oxidative and alkylating agents [5,27] and in plants upon exposure to cold, drought and salt stress [31]. It is not known if m1A in position 58 is mainly affected in these conditions, but if this is the case, then translation could also be efficiently modulated by m1A58 levels following stress.

In addition to its presence in position 37, m1G is also found in position 9 of many cytosolic and mitochondrial tRNAs. In human, the TRM10A methylase is responsible for installing this modification on a large number of tRNAs (21) [123]. Recently, a very interesting link was established between tRNA m1G9 and mRNA m6A modifications [123]. TRM10A and the m6A demethylase FTO were shown to collaborate to target a specific subset of m6A-containing mRNAs whose efficient translation requires the presence of m1G9-containing tRNAs. In the presence of both FTO and TRM10A, the m6A level of targeted mRNAs is maintained low, preventing efficient binding of the m6A reader protein YTHDF2. This situation contributes to maintain the targeted mRNAs stable while their translation is favoured by the presence of m1G9-modified tRNAs. In the absence of TRM10A, FTO would less efficiently demethylate these mRNA targets, leading to hyper m6A methylation, YTHDF2 binding and mRNA instability. In addition, these unstable hypermethylated mRNAs would be poorly translated due to the absence of m1G9-modified tRNAs. Recently, the lack of m1G9 methylation was also shown to cause a decrease in the steady-state amount of human tRNAiMeti suggesting that this modification, as for m1A58, could be involved in regulating translation initiation [124]. Global m1G levels are known to vary in different environmental situations (see earlier) and it remains to be determined if these modifications affect positions 9, 37 or both.

The modification of guanosine to m7G in position 46 is one of the most prevalent tRNA modifications found in eukaryotes and concerns a large number of different tRNAs (11 in yeast, 22 in human) [125]. The m7G46 level was found to vary in yeast upon oxidative and alkylating treatments [5,27] and, in plants, in cold, drought and high salt conditions [31]. TRM8/TRM82 in yeast and METTL1/WDR4 in human are responsible for this modification. m7G46 is important to stabilize tRNAs, accordingly the trm8 or trm82 mutants have an increased sensitivity to high temperature [126]. In human cells, depletion of METTL1 results in the loss of m7G46 and causes a global reduction in translation. Furthermore, in this mutant, mRNAs having low translation efficiency (TE) compared to wt have a significantly higher frequency of codons decoded by m7G46-modified tRNAs [125]. A ribosome occupancy study in the mettl1 mutant also revealed an increase ribosome pausing at codons whose translation is dependent on the presence of m7G46-containing tRNAs [125]. These results suggest that the amount of m7G46 in tRNAs can regulate the translation of a subpopulation of mRNAs in response to various environmental conditions.

Several other common tRNA modifications are stress-sensitive, like 2-methylguanosine (m2G) and N2,N2-dimethylguanosine (m2,2G) (in positions 10 and 26), 5-methyluridine (m5U) (in position 54), dihydrouridine (D) (in positions 16–20) and (Ψ in 14 possible tRNA positions apart from position 34) [5,27,31]. These modifications are mainly associated with the stabilization of the tRNA tertiary structure and to prevent tRNA misfolding [1]. Dihydrouridine is more specifically important to maintain tRNA conformational flexibility, especially in low-temperature conditions [127]. Increased D levels are also observed in several cancer cell types [25]. On the contrary, Ψ is able to stabilize RNA, improving base-stacking by forming additional hydrogen bonds with water through its extra imino group [128]. For example, Ψ in position 55 was proposed to stabilize the tertiary structure of tRNA, particularly in extremely high-temperature conditions [1]. Ψ as D, has been found to increase in some cancers [129]. It remains to be seen if stress-induced variations of these modifications can adapt translation to specific cellular and environmental conditions.

Chemical modifications affecting the production of biologically functional tRNA fragments: an indirect way to adapt translation to stress?

tRNAs are generally considered as stable molecules, but this high stability greatly relies on the acquisition of a proper tertiary structure. Miss-folded tRNA molecules are targeted for degradation by the nuclear surveillance pathway or by the rapid tRNA decay (RTD) pathway [130,131]. Several chemical marks are important for tRNA proper folding (see above for numerous examples), so that hypomodified tRNAs are preferential targets of tRNA decay pathways [132-134]. For example, precursors of initiator tRNA methionine lacking m1A58 are rapidly turned over by the nuclear surveillance pathway [135], while mature tRNAs valine (AAC) lacking both m7G and m5C (in position 34 and/or 48 and 49) are degraded by the RTD pathway at high (37°C) temperature in yeast [132].

While the nuclear surveillance and RDT pathways are expected to completely degrade tRNAs, relatively stable tRNA-derived RNA fragments (tRFs) of different sizes have been found in many different organisms [130,134]. Half-tRNAs (5ʹ-halves (tRF5A) and 3ʹ-halves (tRF3A)) are produced primarily upon different stresses by an endonucleotidic cleavage in the anti-codon loop. Also, in normal physiological conditions, numerous short tRFs (around 15–25 nucleotides) are produced from mature tRNAs, mainly by cleavage in the D (tRF-5D) and T loops (tRF-3 T). tRFs have many proposed biological activities, one of which is to modulate translation [134]. Mechanisms by which such regulation can be achieved are still under study, but they could involve blocking access of eIF4F to the mRNA cap structure [136], directly interacting with the small ribosomal subunit to inhibit translation [137] or more specifically acting on specific mRNAs, in a microRNA-like manner [138,139]. We focus in the following on the few known situations in which key tRNA chemical decorations were shown to influence the biogenesis and/or function of tRFs and, doing so, potentially regulate translation.

Angiogenin (ANG) is a stress-inducible, vertebrate-specific, endonuclease of the RNase A family that can cleave the anticodon loop of a subset of tRNAs (i.e. those with a CA-motif in the anticodon) to generate tRF5A and tRF3A fragments [140,141]. ANG cleavage is inhibited by the presence of m5C in position 38 (deposited by DNMT2) or in position 34 and/or 48 and 49 (deposited by TRM4(NSUN2)) [134]. Therefore, stress conditions leading to variations in tRNA m5C levels can not only directly adapt ribosome mRNA decoding properties (see above) but, by influencing ANG cleavage activity, also produce variable amounts of tRFs. In turn, this stress-specific tRF population could potentially contribute to adapt translation to the situation. Another enzyme responsible for generating tRFs is RNase L, a mammalian-specific endonuclease that can cleave the anticodon loop of tRNA histidine depending on the presence of a specific tRNA modification [142]. RNase L is activated upon detection of double-stranded RNA, a hallmark of viral infection, and cleaves single stranded viral and cellular RNAs at the very promiscuous recognition site UNN. The presence of Q in position 34 is protecting most tRNAs from RNase L cleavage, except for tRNA histidine where, on the contrary, it stimulates cleavage at position 37, leading to the synthesis of stable tRNA histidine fragments and to a general decrease in protein synthesis [142]. Modifications in tRNA Q34 levels could therefore affect this situation, leading to the synthesis of new populations of tRNA fragments with different impacts on translation. tRNAs decorated with specific chemical marks can also be the target of eukaryotic toxin ribonucleases for competitive purposes. For example, the subunit of the zymocin toxin from Kluyveromyces lactis is a ribonuclease that, once introduced in competitor yeast cells, can down-regulate translation by cleaving several tRNAs presenting the mcm5s2U modification in position 34 [143]. In response, adjusting tRNA mcm5s2U34 levels in yeast could potentially dampen this negative influence on translation. RNases from the T2 and (to a lesser extent) DICER families are also responsible for generating long and short stable tRNA fragments [139,144,145]. Certain tRNAs are more susceptible than others to cleavage by these enzymes and one largely understudied determinant of this selectivity could be the nature of chemical decorations on each tRNAs. Whether tRFs biological activity could be influenced by the nature of chemical marks they contain, is also largely unexplored. Indeed, in a single case, the presence of a modified ribonucleotide (Ψ in position 8) on small tRFs, issued from three different tRNAs, was shown to be essential to exert translation inhibition in a human stem cell line [146]. Clearly, more work is needed to establish the global impact of tRNA epitranscriptomic marks on the production and biological activity of tRFs and their impact on translation.

Variations in rRNA modifications and impact on translation

Diverse ribosome populations exist within cells and this heterogeneity can be due to variations in their rRNA and protein compositions, but also to post-transcriptional/translational modifications of these components (reviewed in Guo [147]). To what extent these variations in composition influence ribosome properties, thereby changing the output of translation, with some specialized ribosomes displaying differential affinities for particular mRNAs, is the subject of intense debates (reviewed in Ferretti and Karbstein [148]). We specifically review here evidence for ribosome functional specialization linked to variations in their rRNA chemical composition. Since rRNA modifications are generally installed during ribosome assembly and are considered to be irreversible [6,13], ribosome turnover would be needed to remove an existing ribosome population to the profit of a different one. This implies that ribosome functionalization by changing rRNA chemical composition would be a slow process, potentially useful as a long-term strategy of cell adaptation but not pertinent for rapid acclimation processes [148,149].

Variation in rRNA 2ʹ-O-methylation and pseudouridylation levels and impact on translation

In certain situations, 2ʹ-O-methylated nucleotides and pseudouridines are deposited by snoRNAs in substochiometric amounts, generating heterogeneous ribosome populations [6,38,150]. In human, the recent remapping of all 2ʹ-O-methylated nucleotides allowed the identification of ‘vulnerable’ sites, particularly affected by fibrillarin and anti-tumoural p53 levels, and most likely to undergo specific regulation [150]. In human, snoRNAs are differentially expressed in cancers and some of them have been associated with oncogenesis [151,152]. Accordingly, levels of 2ʹ-O-methylated nucleotides and Ψ are altered in cancer lines [17,18], although it is not clear at the moment if these changes always result of corresponding changes of snoRNA expression. Also, whether 2ʹ-O-methylation and pseudouridylation rRNA profiles can vary under non-pathological conditions remains to be determined. Interestingly, ribosomes with reduced amount of 2ʹ-O-methylation levels present a fourfold reduction in their capacity to initiate translation using internal ribosome entry sites (IRES) [17,39]. Reducing the amount of pseudourydilated rRNAs in ribosomes was also shown to impact IRES-dependent translation, either by increasing or reducing its efficiency depending on the studied mRNA [18,153,154]. IRES-dependent translation concerns many important cellular mRNAs, including growth factors and receptors, apoptosis regulators, oncogenes and tumour suppressors [155], and modulating its efficiency, by increasing or reducing 2ʹ-O-methylation and pseudouridylation levels, could represent a way to adapt translation to specific cellular conditions. In addition to IRES-containing mRNAs, modulating the amount of 2ʹ-O-methylation was also shown to impact (positively or negatively) the cap-dependent TE of several mRNAs [39]. However, since removing 2ʹ-O-methylation over a certain threshold can impact ribosome biogenesis [156], it is possible that these effects are simply due to a general reduction of ribosome availability and not to ribosome specialization per see [148]. Therefore, it is unclear at the moment if global variations in 2ʹ-O-methylation and pseudouridylation contribute or not to generate specialized ribosomes affected in their capacity to initiate cap-dependent translation on mRNA subsets.

A key evolutionary conserved Ψ present on 18S rRNA (position 1248 in human and 1191 in yeast) and structurally located at the ribosome P site, can be further modified by a methyltransferase and an aminocarboxyl propyl transferase (respectively named EMG1 and TSR3 in human) to generate 1-methyl-3-amino-carboxyl-propyl Ψ (m1acp3 Ψ) [19]. This modification is involved in 18S rRNA processing [13]. Also, by interacting with the 40S ribosomal protein RPS16 and tRNA, it can directly impact the ribosome P site function [157]. Recently, a large subset of human tumours was found to possess hypo-m1acp3Ψ-modified ‘onco-ribosomes’[19]. In these tumour cells, while global protein translation was unaffected, the TE of a subset of mRNAs coding for ribosomal proteins increased leading to the higher accumulation of corresponding proteins [19]. The same observation was made for tsr3 mutant cell lines presenting low m1acp3Ψ levels [19]. This suggests that ribosomes lacking m1acp3Ψ can specifically promote the translation of a subset of mRNAs. It remains to be determined if non-pathological conditions can also result in the production of specialized m1acp3Ψ-free ribosomes.

m5C rRNA variations and impact on translation

The yeast RNA methyltransferase RCM1 is responsible for converting the cytidine in position 2278 of 25S rRNA in m5C [149]. The nematode-corresponding position (m5C2381) is also targeted by NSUN5, the ortholog of RCM1. In these two organisms, as well as in drosophila, the loss of RCM1(NSUN5) confers increased lifespan and resistance to different types of stress [149]. In RCM1-knockout cells, ribosome lacking m5C2278 are more efficient to translate several stress-responsive genes [149]. Since RCM1 is localized in the nucleoli [10], a stress-mediated response involving the downregulation of RCM1 enzymatic activity is likely to be slow, involving the exchange of methylated ribosomes for unmethylated ones so that this mechanism would preferentially modulate long-term chronic stress [149]. In human, high NSUN5 expression promotes the progressing of cancer cells through cell cycle regulation [158]. Accordingly, the epigenetic silencing of NSUN5 in human glioma cells generates ribosomes lacking 28S C3761 methylation (the equivalent of yeast 25S C2278) that promote the selective translation of stress-responsive genes and limit global protein synthesis and cell growth [159]. Also, NSUN5 knockout in mice leads to reduce body weight and reduced protein synthesis in many tissues [160].

Another conserved methyltransferase, NOP2(NSUN1), is involved in converting a second large subunit rRNA cytidine (in position 2870 for yeast, 2982 for nematodes and 4417 for human) in m5C. In yeast and human, NOP2(NSUN1) is essential for rRNA processing and synthesis of the large ribosome subunit, a function that is independent of its m5C modification activity [161-163]. Accordingly, NOP2(NSUN1) is essential for yeast growth and mammalian embryo development [161,162]. This is not the case for nematodes, as nsun1 worms are viable and not significantly affected in ribosome biogenesis nor in global translation [164] (a situation analogous to what is observed in nsun5 worms [149]). nsun1 worms have a longer lifespan and were shown to remodel the translation of specific mRNA transcripts [164]. Therefore, the level of NSUN1 in nematodes is likely important for the synthesis of specialized ribosomes more or less adapted to the translation of different mRNA subsets.

At the moment, it is not clear exactly what environmental signals could regulate NOP2(NSUN1) and RCM1(NSUN5) levels. The fact that the human NSUN5 CpG island promoter can be epigenetically regulated [159] may be a way environmental cues could achieve such regulation. Overall, these observations suggest that RCM1(NSUN5) and NOP2(NSUN1) (at least in nematodes), by adjusting ribosome m5C levels, adapt translation to different physiological conditions such that, in fast growth conditions, highly m5C-modified ribosomes are preferred while m5C-hypomodified ribosomes are favoured in stress situations.

m6A rRNA variations and impact on translation

Another important rRNA modification that affects ribosome translation in animals (globally and/or on mRNA subsets) is m6A [165-167]. In human and nematodes, 18S and 28S/26S rRNAs are decorated each by one m6A, installed by METTL5(METL5) on 18S [12,165,166] and by ZCCHC4 on 28S/26S [12,167,168]. In nematodes, the loss of METL5 does not globally affect translation while the TE of cyp-29A3, a transcript coding for a cytochrome P450, is reduced 10-fold, suggesting that ribosome 18S m6A methylation levels can regulate the translation of specific mRNAs [166]. metl-5 worms have an increased lifespan and are more resistant to several abiotic stresses [166]. These phenotypes are proposed to be the direct result of reduced CYP-29A3 translation and CYP-29A3-dependent synthesis of eicosanoid lipids in the mutant. In contrast to nematodes, in human cells, knocking out METTL5 results in a general decrease in TE, but whether some specific mRNAs are more affected than others has yet to be investigated [165]. In human cells, knocking out ZCCHC4, also results in a general decrease in TE (of about 25%), but in that case, a subset of 311 mRNAs were shown to be much more affected than others, including transcripts coding for membrane protein targeting, mRNA catabolic process, ER localization and translation initiation, here again suggesting that 28S m6A levels can affect the translation of specific mRNAs [167]. The identification of physiological and/or environmental conditions leading to the accumulation of ribosome population with substochiometric amount of m6A-modified rRNA is still needed to firmly established rRNA m6A level as a new regulation layer of the animal stress response.

m1A rRNA variations and impact on translation

Another key rRNA modifying enzyme is the methylase RRP8 that generates m1A in position 645 of yeast 25S rRNA [169]. The loss of m1A645 results in the production of ribosomes altered in their general ability to initiate translation, possibly linked to a reduced competence for the 60S subunit lacking m1A645 to bind to the 40S subunit [169]. Surprisingly, despite having a reduced translation initiation efficiency, most proteins are produced in similar amounts in wt and rrp8 mutant lines. Exceptions to this rule concern several enzymes involved in carbohydrate metabolism that are either up or down regulated in rrp8 mutant compared to wt, suggesting that ribosomes lacking m1A645 translate corresponding mRNAs more or less efficiently [169]. This suggests that, under some growth conditions, ribosomes lacking m1A645 could be synthesized to specifically regulate the translation of mRNAs involved in producing key carbohydrate metabolism enzymes. In human, m1A-modified nucleotides are elevated in the urine of cancer patients [20] and lowering the level of 28S m1A1309 (the equivalent of yeast 25S m1A645) leads to the downregulation of cell proliferation in a p53-dependent manner [170]. These results suggest that methylation at this position is a way to control cell proliferation in mammals. Finally, in nematodes, T07A9.8 (the orthologue of the yeast RRP8 enzyme) methylate position A674 of 26S rRNA (the equivalent of yeast 25S m1A645) [171]. Impairing this function leads to an extending life span for nematodes, again linking m1A modification at this position to cell cycle regulation [171].

m7G rRNA variations and impact on translation

BUD23 is an important protein involved in the biosynthesis of the translational apparatus, firstly by processing the pre-18S RNA into its mature form and secondly by modifying an 18S rRNA guanosine (in position 1639 for human and 1575 for yeast) to m7G [172,173]. Since the methyltransferase activity is not needed to process pre-18S RNA, these two functions are considered to be independent of each other [173,174]. In human cell lines, 18S m7G has been proposed to be present in substochiometric amount in ribosome populations, which may indicate a selective role in ribosome function [174]. Furthermore WBSCR22, the human ortholog of BUD23, is found to be overexpressed in breast cancer and has been proposed as a cancer biomarker [174]. In human cells, 48 h after knocking down BUD23, no global impact on protein translation rate was observed. Yet, the TE of more than 700 mRNAs was affected in this condition, including a strong TE decrease for mRNAs coding for mitochondrial proteins [172]. Also, mRNAs with contrasting GC 5ʹUTR content were differentially translated following BUD23 knockdown. Transcripts having a low GC 5ʹUTR content also had a low TE, while the opposite was true for transcripts having a high GC 5ʹUTR content [172]. These results suggest that variations in 18S m7G levels could generate ribosomes with different affinities for mRNAs having contrasting GC 5ʹUTR content. However, it is not clear for the moment if cell translation occurring 48 h after knocking down BUD23 mainly results from ribosomes lacking m7G or if the loss of BUD23 could impact translation independently of ribosome m7G levels.

Conclusion

A large and convincing body of evidence exists to conclude that tRNA and rRNA chemical modifications are critical for the biogenesis, stabilization and proper decoding functions of the constitutive translation apparatus (reviewed in Sloan et al. and Sharma and Lafontaine [6,13]). However, whether variations in the level of these modifications, in some cell and environmental conditions, can be used to adapt this apparatus to target physiologically relevant mRNA subsets, possibly still await further experiments. At the molecular level, demonstrations that changing the nature and stoichiometry of many tRNA modifications, in yeast, nematodes, drosophila, plants and mammals, can indeed favour the translation of specific mRNAs are more and more numerous and convincing. Although less numerous, reports of heterogeneous ribosome populations, specialized to target specific mRNAs due to changes in the stoichiometry of one or several rRNA chemical marks, have been lately published using different eukaryotic systems. So, what is missing to firmly establish tRNA and rRNA epitranscriptomic variations as a new layer of eukaryote gene regulation? First, it is not always clear in which (non-pathological) physiologically relevant conditions these variations can occur and what could be the impact of these changes at the organism level. In other words, can these variations significantly impact nutritional, developmental or stress-responses leading to acclimation and/or adaptation of individuals? Also, studies of these variations in natural populations coming from contrasting environments are clearly missing to ensure that this regulatory process is indeed under selection in natura and therefore meaningful as a regulatory process. Plants as complex organisms that can be studied at the physiological and molecular levels, as well as in natural environments, may represent good systems to try to solve these issues.

Click here for additional data file.