The role of intracellular compartmentalization on tRNA processing and modification

ABSTRACT

A signature of most eukaryotic cells is the presence of intricate membrane systems. Intracellular organization presumably evolved to provide order, and add layers for regulation of intracellular processes; compartmentalization also forcibly led to the appearance of sophisticated transport systems. With nucleus-encoded tRNAs, it led to the uncoupling of tRNA synthesis from many of the maturation steps it undergoes. It is now clear that tRNAs are actively transported across intracellular membranes and at any point, in any compartment, they can be post-transcriptionally modified; modification enzymes themselves may localize to any of the genome-containing compartments. In the following pages, we describe a number of well-known examples of how intracellular compartmentalization of tRNA processing and modification activities impact the function and fate of tRNAs. We raise the possibility that rates of intracellular transport may influence the level of modification and as such increase the diversity of differentially modified tRNAs in cells.

Introduction

In eukaryotes, tRNAs are encoded in either the nucleus or one of the other genome-containing organelles; mitochondria, chloroplast or plastids, depending on the organism. Regardless of their site of synthesis, in all cases tRNAs are transcribed as premature molecules that cannot immediately partake in translation. As tRNAs are transcribed, they begin to form local structures that may include initial folding of the arms; the D-arm, followed by the anticodon arm, and finally the TΨC arm and the acceptor stem (Fig. 1A). Either during or soon after transcription, tRNAs also undergo various forms of “molecular pedicure” that trims their 5’ and 3’ ends, removes introns (if present) and adds their characteristic CCA tails, where eventually amino acids are attached. Concomitantly, the tRNA is also surveying various folding pathways and eventually assumes its canonical L-shape structure required for translation. At any point in this maturation pathway, post-transcriptional modifications may be added, and those that appear early, tend to influence and enforce proper folding. Studies using microinjection of tRNAs into Xenopus oocytes indicated that certain modifications appear before end maturation, in a somewhat specific order. For example, intron-containing pre-tRNA^Tyr with immature 5’ leader and 3’ trailer sequences acquire pseudouridines (Ψ) in the anticodon and TΨC loop, as well as 5-methylcytidine (m⁵C) in the variable loop and 1-methyladenosine (m¹A) in the TΨC loop.^1,2 As the 5’ and 3’ ends become matured, new modifications are added such as dihydrouridine (D) to the D-loop and additional pseudouridylations. Finally, after intron removal the remaining modifications are added to complete the required set for a particular tRNA¹, but in eukaryotes intracellular compartmentalization further imposes order to tRNA maturation and to the occurrence of modifications (Table 1).

Figure 1.

Tertiary contacts critical for global tRNA structure. A General tRNA cloverleaf structure highlighting the numbering scheme and important tertiary contacts between tRNA arms shown as dashed lines. Darker circles represent nucleotides, which when altered, disrupts the stability and export of the tRNA. Many nucleotide contacts that disrupt export are also those involved in inter-arm base pairing between the D and TΨC arms. B As tRNA tertiary structure is disrupted, certain modifications will be negatively affected while others may not. In blue are positions in which corresponding modification enzymes can tolerate global tertiary structural changes while those shown in red do not tolerate tertiary alteration of the tRNA.

Table 1.

The intracellular localization of a representative set of modification enzymes is shown. Whether a given modification occurs in intron-containing tRNAs (unspliced tRNA) is as indicated.

Modification	Enzyme(s)	Localization	Organism	Found in unspliced tRNA
Queuosine (Q)	TGT	Outer Mitochondrial Membrane 94	S. cerevisiae	No
Queuosine (Q)	TGT	Nucleus	T. brucei	No
1-methylguanosine (m1G37)	Trm5	Nucleus67	S. cerevisiae	No
Wybutosine (yW)	TYW1, TYW2, TYW3, TYW4	Cytoplasm/extra-nuclear 95	S. cerevisiae	No
2-thiouridine 34 (S2U)	Tuc1 (Ncs6), Tuc2 (Ncs2)	Cytoplasm 96	S. cerevisiae	No
2-thiouridine 34 (S2U)	Mtu1	Mitochondria 97	S. cerevisiae, Homo sapiens	No
N2,N2-dimethylguanosine (m2,2G)	Trm1	Nucleus, Mitochondria 98	S. cerevisiae	Yes
N2-methylguanosine (m2G10)	Trm11/Trm112	Cytoplasm 99	S. cerevisiae	No
1-methyladenosine (m1A58)	Gcd10/Gcd14	Nucleus8	S. cerevisiae	Yes
Threonylcarbamoyladenosine (t6A)	Sua5/KEOPS	Cytoplasm100	S. cerevisiae	No
Threonylcarbamoyladenosine (t6A)	Sua5/Qri7	Mitochondria100	S. cerevisiae	No
3-methylcitidine (m3C32)	Trm140	Cytoplasm101	S. cerevisiae	No
5-methylcytidine (m5C40)	Trm4 (NCL1)	Nucleus102	S. cerevisiae	Yes
5-methylcytidine (m5C34)	Trm4 (NCL1)	Nucleus102	S. cerevisiae	Yes
Pseudouridine (Ψ32)	Pus8	Cytoplasm103	S. cerevisiae	No
Pseudouridine (Ψ38 and Ψ39)	Pus3 (Deg1)	Nucleus, Cytoplasm104	S. cerevisiae	Yes
Pseudouridine (Ψ34 and Ψ36)	Pus1	Nucleus105	S. cerevisiae	Yes
N6-isopentenyladenosine (i6A37)	Mod5	Nucleus, Cytoplasm, Mitochondria106	S. cerevisiae	No
Wyosine (imG)	TYW1S,2,3B	Mitochondria107	T. brucei	No
Inosine (I34)	ADAT2/3	Cytoplasm108	T. brucei	No
3-methylcytidine (m3C32)	ADAT2/3,TRM140	Nucleus93	T. brucei	No
3-methyluridine (m3U32)	ADAT2/3, Trm140	Nucleus93	T. brucei	No
1-methylinosine (m1I37)	Trm5	Cytoplasm, Mitochondria109	S. cerevisiae	No
5-formylcytidine (f5C34)	Nsun3	Mitochondria110	Homo sapiens	No
N2-methylguanosine (m2G10)	Trm11/Trm112	Cytoplasm99	S. cerevisiae	Yes
5-carbamoylmethyl-2′-O-methyluridine (ncm5Um)	Trm7	Cytoplasm111	S. cerevisiae	No
5-carboxymethylaminomethyluridine (cmnm5U34)	Mss1	Mitochondria112	S. cerevisiae	No

Beyond the nucleus, tRNAs also undergo further modification in the cytoplasm. Yet the story does not end there, many cellular RNAs, and indeed tRNAs, may travel back and forth across cellular membranes and each movement to a new locale offers the potential to be further modified. If one accepts that many modifications may alter the structure of tRNAs locally and sometimes globally, then what modification enzymes encounter is an ever-changing substrate structural landscape, which in turn may be recognized differently by different enzymes. For example, it is possible that intracellular transport dynamics may lead to the appearance a of hypomodified tRNA in a cellular compartment (i.e., nucleus or cytoplasm) where such tRNA may now be recognized by a different set of enzymes, which normally do not encounter such a substrate. This may lead to the given tRNA acquiring new modifications that it normally does not get. Taken together, this raises the exciting possibility that the diversity of modifications in tRNA sets varies greatly in the life span of a cell and may be influenced not just by environmental cues but also by transport dynamics. In the following pages, we highlight a number of recent examples of tRNA movements within cells and how intracellular distribution of tRNAs, because of transport, is intricately linked to many processing events and most certainly modifications. In passing, we will also discuss the fates of tRNAs that are recognized by cells as not fully modified and highlight the fact that such surveillance pathways are not so exacting; an observation that should not be surprising given the often subtle effects that modifications have in their substrate targets.

Functionality checkpoints: Early steps in tRNA maturation in eukaryotes

With a starting point set by transcription, modification enzymes have been divided into two major groups based on the way they recognize their substrates: architecture-dependent enzymes requiring a fully folded tRNA for activity, and architecture-independent enzymes, which do not require a full-length tRNA for activity.³ The influence of tRNA structure on modifications has been examined by introducing a series of structural alterations to Saccharomyces cerevisiae tRNA^Asp followed by microinjection into Xenopus oocytes.^3,4 When mutations were introduced that changed the 3D structure of tRNA^Asp, several modifications such as 1-methylguanosine at position 37 (m¹G₃₇), Ψ₄₀ and Ψ₁₃ were not formed, leading to the suggestion that the enzymes catalyzing such reactions recognize a fully folded tRNA and are therefore architecture-dependent. On the other hand, global changes in tRNA structure had no major impact on m⁵C₄₉, 5-methyluridine at position 54 (m⁵U₅₄) and Ψ₅₅ (Fig. 1B). Unaffected modifications are thus catalyzed by enzymes, which recognize local structure such as stem loops or specific sequences; these motifs become available shortly after transcription and before tRNA assumes its L-shape. These enzymes are therefore architecture independent. These, however, still recognize limited structural features; for example, for Ψ₅₅ the presence of 4 G-C pairs in the TΨC loop, while a slightly longer stem was necessary for m⁵U formation.^4,5 Thus synthesis of modifications may be differently affected by changes to global vs. local structure of tRNA.³

Defects in the addition of modifications can impact tRNA stability. An interesting case occurs with m¹A₅₈; a modification conserved in Bacteria, Archaea, and Eukarya.⁶ In S. cerevisiae m¹A₅₈ is found in many tRNAs, but it is only essential for tRNA_i^Met. Mutation of the Trm6/Trm61 (GCD10/GCD14) complex, the methylase responsible for m¹A₅₈, results in increased tRNA_i^Met instability and a lethal reduction in tRNA_i^Met pools; a phenotype easily overcome by overexpression of tRNA_i^Met.^7-9 Further analysis, based on genetic screening, indicated that RRP44 (DIS3), a 3’-5’ exoribonuclease exosome subunit, and TRF4, DNA polymerase with poly(A) polymerase activity, had roles in the observed reduction of tRNA_i^Met. Upon closer investigation, it was apparent that mutation of either TRF4 or RRP44 resulted in restoration of tRNA_i^Met to levels close to normal, supporting their involvement in the reduction of tRNA_i^Met. Furthermore, deletion of the gene for the nucleus-exclusive exosomal subunit RRP6 restored tRNA_i^Met levels while overexpression of Trf4 caused an exacerbated reduction.⁹ Because of this, Trf4 was suggested to polyadenylate tRNA_i^Met, which would then serve as a signal for its targeted degradation by the nuclear exosome. Indeed, exosomal degradation of polyadenylated RNAs requires Trf4 polyadenylation activity.^10,11 The polyadenylation of tRNA and its subsequent exosomal degradation became known as the tRNA nuclear surveillance pathway. However, this discovery did not initially address how the hypomodified tRNA_i^Met was targeted, as Trf4 does not possess a recognizable RNA binding domain. To address this, affinity purification of Trf4, and a two-hybrid analysis of the nuclear exosome cofactor Mtr4, led to the identification of the TRAMP protein complex.^12,13 The TRAMP complex consists of three subunits: Trf4 or Trf5, Air1 or Air2, and Mtr4; When in a complex, Air1 or Air2 bind to their RNA target via their RNA binding domain. This is followed by polyadenylation by Trf4 or Trf5 and lastly, Mtr4 binds and facilitates unwinding of the RNA target via its helicase activity.¹⁴ Taken together, the lack of m¹A₅₈ destabilizes of tRNA_i^Met, which is then flagged for degradation by the activity of Trf4 and the TRAMP complex and later degraded by the exosome (Fig. 2A). Although m¹A₅₈ is the best known example of nuclear surveillance, other tRNAs are potentially monitored, since TRF4 deletion strains of yeast begin to accumulate precursor tRNAs in the nucleus. ¹⁵

Figure 2.

Two major surveillance pathways for defective tRNAs. A. tRNA_i^Met lacking m¹A₅₈ is polyadenylated by the TRAMP complex (Trf4/5, Air1/2). The helicase Mtr4 assists the TRAMP complex and targets TRAMP to the nuclear endonuclease complex consisting of several structural proteins and a 3’ to 5’ exonuclease. B. The rapid tRNA decay pathway (RTD) degrades hypomodified tRNA^Val and tRNA^Ser.

Export is the last major step that takes place in the nucleus and therefore serves as the final checkpoint tRNAs must clear before entering the cytoplasm. One of the major tRNA exporters, exportin-t, was first discovered when searching for nuclear transporters which interact with RanGTP, the GTPase involved in many transports functions.¹⁶ Exportin-t preferentially binds end-processed tRNAs in vitro but does not discriminate between the presence or absence of introns, instead relying on the recognition of a properly folded tRNA backbone.¹⁷ Lack of discrimination between intron-containing and intronless tRNAs is especially important in cases where the splicing machinery localizes to the cytoplasm, for example, in S. cerevisiae and Trypanosoma brucei.^18,19 In turn, deletion of the S. cerevisiae exportin-t homolog (Los1) causes end-matured, intron-containing tRNA to accumulate in the nucleus but paradoxically is non-essential for viability.^20-22 The plant ortholog is also non-essential indicating the existence of redundant pathways for export from the nucleus to the cytoplasm.^23,24 The existence of alternative tRNA exporters became evident with the discovery of exportin-5 in vertebrates and its homolog Msn5 in S. cerevisiae. Exportin-5 exports microRNAs and to a varying degree, aminoacylated tRNAs bound to eukaryotic elongation factor 1A.^25-27 Although tRNAs interact with exportin-5, they do so with different affinities; consequently, the role of exportin-5 in tRNA export varies among different species. Knockdown of exportin-5 in humans and plants does not affect tRNA levels in the nucleus, while a knockdown in Drosophila, which naturally lacks exportin-t, leads to nuclear tRNA accumulation.^27,28 In S. cerevisiae, Msn5 is the primary re-exporter for tRNAs, which have been retrogradely imported into the nucleus after splicing, and as such binds to matured aminoacylated tRNA.^29,30 It should be noted that the ability for Msn5 to export tRNAs that do not require splicing has not been ruled out. Although it is clear both Los1 and Msn5 share a role in tRNA export, deletion of Los1 and Msn5 simultaneously is still not lethal,³¹ indicating further built-in redundancy. However, in some systems nuclear export is not an all or nothing mechanism, but more of a kinetically controlled pathway, where tRNAs will be exported regardless, but exporters facilitate the export of “healthy” tRNAs.³²

When comparing the presence and absence of modifications, modified tRNAs bound with greater affinity to exportin-t than tRNA lacking modifications.¹⁶ Seeing how various modifications give rise to changes in both the structure and stability of tRNAs, it is conceivable that hypomodification could affect the export potential for tRNA from the nucleus. In fact, various point mutations interrupting critical tertiary contacts between the D and TΨC loops, caused decrease in affinity to exportin-t and subsequently nuclear export.³³ Disruption of the TΨC stem caused a similar reduction in the binding affinity of exportin-t for tRNA.³⁴ Although specific modifications themselves may individually be too small to influence export directly, hypomodification could influence the structure and rigidity of tRNA, which subsequently influences export. Along these lines, organisms have also adopted aminoacylation as a further checkpoint for export. For example, it was discovered that in Xenopus oocytes tRNAs are aminoacylated prior to nucleus-export to the cytoplasm.³² The lack of aminoacylation leads to a decrease in tRNA nuclear export and nuclear accumulation in yeast.^35,36 The adoption of tRNA aminoacylation as a prerequisite for nuclear export allows further screening of tRNA “healthiness” by aminoacyl tRNA synthetases, as a type of proofreading mechanism to ensure that only “good” tRNAs make it to the cytoplasm.

The cytosolic fate of tRNA modifications

After export from the nucleus, most tRNAs lack a complete set of modifications and still rely on numerous cytoplasmic modification enzymes for the final steps of maturation. In most cases, the intracellular distribution of modification enzymes is not clear, but certainly, many localize to the cytoplasm. The reasons why modification enzymes localize to specific compartments is not well understood, but in some cases localization is in line with the localization of other processing enzymes that are critical for tRNA maturation. For example, modification enzymes that are localized to the cytoplasm, may recognize tRNA substrates, which have previously undergone splicing, such as ribose methylations catalyzed by Trm44 (U_m44), Trm3 (G_m18), or Trm7 (C_m32, N_m34).³⁷ Here, it stands to reason that if splicing is cytoplasmic and the given enzyme requires the intron for substrate recognition, then it follows that the modifications will likely be cytoplasmic. However, this is not always the case, for there are examples of modifications that only occur after splicing but reside in the nucleus, as discussed in the following pages. In other cases, there are modification enzymes, which do not discriminate based on the presence or absence of an intron and thus their localization cannot be predicted by looking at the distribution of different processing events.

The differential distribution of modification enzymes within cells could potentially lead to the accumulation of immature tRNAs in the cytoplasm, which could cause a problem during protein synthesis. However, analogous to the nucleus, the cytoplasm also has mechanisms to deal with hypomodified tRNAs. Studies examining the importance of modifications were at first puzzling, as many genes encoding evolutionarily conserved modifications were found non-essential. For example, deletion of genes encoding the enzymes responsible for the synthesis of m⁵C (at positions 34, 40, 48 and 49), D (at positions 16, 17, 20 and 47), Ψ (at positions 13, 31, 35 and 55), 7-methylguanosine (m⁷G) at position 46, and m¹G₉ had little to no effect on growth.^38-43 However, a number of nonessential genes encoding for modification enzymes when deleted in tandem lead to synthetic lethality. For example, deletion of several genes encoding “nonessential” modification enzymes in yeast, led to a significant reduction in the steady-state levels of tRNA^Val _AAC.⁴⁴ This observation led to the discovery of the rapid tRNA decay pathway (RTD), which works independently of the nuclear surveillance pathway described previously (Fig. 2A). Work to identify the components of RTD uncovered three genes which affected tRNA degradation: XRN1, RAT1, and MET22. Xrn1 and Rat1 are both 5’ to 3’ exonucleases, which can mediate degradation of destabilized tRNA, while Met22 indirectly modulates Rat1 and Xrn1 activity via accumulation of the inhibitor adenosine 5’,3’ bisphosphate (Fig. 2B).^45,46 Not all tRNAs are affected equally by RTD after multiple “non-essential” modification genes are deleted. For example, loss of m⁷G₄₆ and m⁵C₄₉ in both tRNA^Val _AAC and tRNA^Val _CAC, results in the targeted degradation of only tRNA^Val _AAC. This specificity is also observed for other m⁷G₄₆ and m⁵C₄₉ containing tRNAs such as tRNA^Met and tRNA^Phe neither of which is subsequently degraded following the loss of these specific methylations. Similarly, loss of N4-acetylcytidine at position 12 (ac⁴C₁₂) and Um₄₄ leads to the specific degradation of tRNA^Ser _UGA/CGA while tRNA^Ser _IGA/GCUis unaffected.^45,47 These results point out that simply lacking a set of modifications is not enough to cause RTD but likely leads to a destabilization or disruption of the tRNA structure in certain tRNAs. In favor of this, the nucleotide sequence of the RTD susceptible tRNA^Ser _CGAwas altered to more closely resemble that of tRNA^Ser _IGA in the TΨC and acceptor stem. Unlike before, the resulting mutant tRNA^Ser _CGA was not degraded by RTD as the altered nucleotide sequences allowed for the stability required to compensate for the loss of ac⁴C₁₂ and Um₄₄.⁴⁸ The data suggests that RTD can be triggered by the absence of modifications that alter the structural stability in the TΨC and acceptor stems. The presence of certain pairs of modifications becomes more crucial for tRNAs that may be inherently more structurally unstable and rely on those modifications more heavily for proper folding. The existence of a cytoplasmic tRNA monitoring pathway helps ensure that only properly modified tRNAs participate in translation, while limiting the availability of hypomodified tRNAs, which may cause problems of translational efficiency, accuracy or both.

In general, it is difficult to predict or even tease out the significance of the intracellular compartmentalization for a given modification enzyme. Although, examples above highlight aspects of specific systems where the localization of a given processing event establishes modification enzyme distribution, in reality, there are not many hard and fast rules for localization prediction. In the end, it may well be that intracellular localization is dynamic, if not transitory, and cells may exploit intracellular partitioning as a way to control enzyme function and certainly impact substrate availability and recognition.

The connection between tRNA splicing and modifications

In all domains of life, subsets of tRNAs are interrupted by introns; these must be removed before the tRNA can be used for translation.^49,50 There is no conservation regarding the number and distribution of tRNA introns across different species, with the range of intron-containing tRNA genes varying from a single tRNA in Azoarcus sp. and T. brucei, to 5% of all tRNAs genes in humans, mouse and drosophila, 20% in S. cerevisiae, 48% in Crenarchaeota and 92% in the yeast Cryptococcus neoformans.^51-53 The mechanism of tRNA splicing also varies between domains of life: bacterial introns are removed through a self-splicing mechanism, while archaeal and eukaryotic introns are removed strictly through protein-catalyzed reactions involving specialized enzymes that work independently from the mRNA spliceosomes.^49,51,54 Despite the fact that tRNA introns are ubiquitous, ascertaining the reasons why they are maintained has proven challenging: some studies indicate certain introns can be removed from the genome with little effect on the organism, while in other cases introns serve as important substrate recognition elements for splicing and modification enzymes.^37,55 Early studies in Xeonopus oocytes were used to classify tRNA modifications under three categories, according to their dependence on the presence of an intron.^1,56 The first category includes modifications that are added only to intron-containing tRNAs, including Ψ (at positions 34-36) and m⁵C (at positions 34 and 40) (Fig. 3A). The second involves modifications that are added only to spliced tRNAs, including Ψ₃₂, m³C₃₂, Um₄₄ and Gm₁₈ (Fig. 3B). The third, and final category, contains modifications that are added to tRNA regardless of the presence of an intron, but do not depend on the intron as a recognition element; these include m¹G₉, m²G_{6, 10}, D_{16, 17, 20}, and 5-methoxycarbonylmethyl-2-thiouridine at position 34 (mcm⁵s²U₃₄).⁵⁶ Interestingly, studies in which the intron sequence was altered, in an effort to elucidate its role in substrate recognition by the tRNA splicing endonuclease and/or modification enzymes, showed that most mutations do not prevent splicing or modification, as long as they do not interfere with the overall cloverleaf structure of the pre-tRNA.^57,58 Two noteworthy studies on this regard were performed in yeast, taking advantage of naturally occurring suppressor tRNAs. In the first study, the genomic copy of the tRNA^Tyr gene SUP6 was replaced with a version that lacked the 14 nucleotide-long intron.² SUP6 is a tRNA^Tyr ochre suppressor, responsible for tRNA-mediated nonsense suppression during translation. Analysis of the mutant tRNA showed that it lacked the modification Ψ at the second position of the anticodon (Ψ₃₅), indicating that either the intron was necessary for substrate recognition by the appropriate modification enzyme, or that the splicing pathway was a pre-requisite for the modification to occur.² This mutant tRNA was defective at nonsense suppression, and present in lower concentrations in the cell.² In a following study, the 32 nucleotide-long intron sequence of the tRNA^Leu gene SUP53, another nonsense suppressor, was either mutated or removed, and the mutant tRNAs analyzed.⁵⁷ The mutant tRNA, transcribed from the intronless gene, lacked the modification m⁵C₃₄, and was defective at tRNA-mediated nonsense suppression.⁵⁷

Figure 3.

Several modifications depend on the presence of an intron in tRNA. A Certain modifications (as indicated) can only be added to intron-containing tRNA; the intron is an essential recognition element for their respective enzyme. B Some modifications are only added after splicing (as indicated).

Unusual examples are the cases of T. brucei and Haloferax volcanii, in the former, the intron sequence must be first edited in two or three positions in order for splicing to take place.⁵⁹ In the latter, the tRNA^Trp intron once cleaved becomes a small guide RNA (sRNA) required for ribose methylations to form Cm₃₄ and Um₃₉ on the cognate spliced tRNA^Trp.⁶⁰ These two examples provide a rationale for intron maintenance and in both cases the splicing of the intron is intricately connected to an additional activity. Notably, in the T. brucei system, and other kinetoplastids, tRNA^Tyr is the only intron-containing tRNA. Therefore, either at some point during their evolution other tRNAs contained introns in these organisms and the intron-containing genes were systematically replaced by their spliced equivalent, or most tRNAs never contained introns. Importantly, despite the variability in the numbers and types of intron-containing tRNAs in many eukaryotes, in all cases tRNA^Tyr always contains an intron, arguing for the importance of that particular intron.

As mentioned above, since in most eukaryotes tRNA splicing takes place in the nucleus, modifying enzymes that act on intron-containing tRNAs also localize to the nucleus, while those that act on spliced tRNAs may also localize to the cytoplasm.⁵⁶ There is at least one significant exception to this rule, however, in S. cerevisiae the tRNA splicing machinery localizes to the surface of the mitochondria facing the cytoplasm. In this organism, tRNA splicing takes place in the cytoplasm, and many modifying enzymes that act on both pre-tRNA and spliced tRNA localize to the cytoplasm as well.^37,56,18 These include the cases of SUP6 and SUP53 described above, in which the modifying enzymes (Pus7 for Ψ₃₅ and Trm4 for m⁵C) localize to the cytoplasm and require the intron for substrate recognition.⁶¹

Intracellular transport dynamics that set the order of modifications

It was always assumed that tRNAs, moved unidirectionally from their site of transcription in the nucleus to their site of action in the cytoplasm, and the cytoplasm was the final destination. Work in the last years, however, demonstrated that the cytoplasm is not the last stopping point for tRNAs. Not only can tRNAs be reimported to the nucleus (by retrograde transport) but, in a growing number of organisms, tRNAs are also imported into the organelles, for example mitochondria.⁶² The latter, as discussed below, serve to complement or complete the tRNA set needed for organellar translation.

The concept of tRNA retrograde transport was first described in S. cerevisiae, based on the observation of significant levels of spliced tRNAs in the nucleus, despite splicing being a cytoplasmic event.^31,62 It was difficult to explain why mutations of MES-1 or CCA-1, which led to lack of aminoacylation, caused the accumulation of spliced tRNA within the nucleus.^35,63 The retrograde nuclear transport pathway was then discovered by means of an elegant heterokaryon assay, where a karyogamy deficient kar1-1 mutant, which prevents nuclear fusion after mating was used. This leads to heterokaryon cells, which harbor two nuclei in a shared cytoplasm. Using heterokaryon assays in combination with fluorescence in situ hybridization (FISH) to label tRNA, the movement of tRNA from a donor to a target nuclei was observed.^31,62 Furthermore, spliced endogenous tRNAs accumulated in the nucleus after treatment with the RNA polymerase inhibitor thiolutin, again supporting the movement of tRNA from the cytoplasm to the nucleus. Since then, multiple lines of evidence have supported these initial studies while several nutritional conditions have been outlined that cause nuclear accumulation of tRNA.^36,64,65 The use of heterokaryon assays proved to be instrumental in demonstrating the existence of bidirectional movement of tRNA, further highlighting tRNA intracellular transport dynamics.

It is now clear that retrograde transport occurs constitutively. Thus retrograde transport constantly cycles tRNA in and out of the nucleus,²⁹ raising the possibility that it may also play other roles in tRNA maturation. An interesting example is offered by wybutosine (yW) a bulky hypermodified guanosine found at position 37 in tRNA^Phe of most eukaryotic organisms; it provides an important frameshifting-prevention mechanism.⁶⁶ Wybutosine biosynthesis involves the formation of a new heterocycle in an otherwise standard guanosine base, this is followed by several enzymatic steps that add various methyl groups and an ACP (aminocarboxypropyl) side chain. It all starts with methylation of G₃₇ to form m¹G, a reaction catalyzed by Trm5 in eukaryotes. This is followed by serial enzymatic reactions catalyzed by Tyw1, Tyw2, Tyw3, and Tyw4 (and Tyw5 in some organisms). Although Trm5 methylates several tRNAs, only tRNA^Phe gets wybutosine. In yeast, Trm5 localizes to the nucleus, while the remaining Tyw enzymes are found in the cytoplasm, thus formation of yW involves two different cellular compartments. The story is, however, even more complicated as Trm5 is not able to methylate intron-containing tRNA^Phe and since the splicing endonuclease is tethered to the outer membrane of the mitochondria facing the cytoplasm, the tRNA has to forcibly go back and forth from the nucleus to get wybutosine. Thus, the separation of these maturation steps by intracellular compartmentalization causes a situation whereby yW synthesis is dependent on retrograde transport. First, intron-containing tRNA^Phe is transcribed in the nucleus, receives some modifications and undergoes end maturation. Then intron-containing tRNA^Phe is exported to the cytoplasm for splicing, followed by retrograde transport to the nucleus to get m¹G₃₇, and finally it is re-exported to the cytoplasm to complete the synthesis of yW (Fig. 4).⁶⁷

Figure 4.

The biosynthesis of wybutosine (yW) in S. cereviase requires retrograde transport of tRNA^Phe to the nucleus. Intron-containing tRNA^Phe is exported to the cytoplasm from the nucleus, where the intron is cleaved by the heterotetrameric splicing endonuclease, which localizes to the outer surface of the mitochondria facing the cytoplasm endonuclease. The exons are then processed by the tRNA-splicing ligase (Trl1) and the tRNA phosphotransferase (Tpt1) to complete the splicing reaction. The newly spliced tRNA^Phe is imported back into the nucleus (retrograde transport) where a nucleus-localized 1-methyltransferase Trm5 forms m¹G₃₇. The methylated tRNA is then re-exported to the cytoplasm, where a series of reactions catalyzed by Tyw1, Tyw2, Tyw3, Tyw4 ultimately creates yW₃₇.

In the trypanosome system, we have discovered a second case of retrograde transport impacting tRNA^Tyr and the modified nucleotide queuosine (Q) (Kessler and Alfonzo, unpublished results). This occurs at position 34 of the anticodon of tRNA^Tyr, ^−Asp, ^−Asn and ^−His in all Eukarya, except for S. cerevisae, which naturally lacks this modification pathway. Q formation is catalyzed by tRNA guanine-transglycosylase (TGT), which is strictly a sequence-specific enzyme and only modifies anticodon loops containing the trinucleotide UGU (positions 33-35).^68,69 TGT has a known substrate preference for spliced tRNA as the transglycosylation reaction performed when replacing G with Q is inhibited by the presence of an intron.¹ In mammals, tRNA splicing is confined to the nucleus and TGT has been localized to the outer membrane of the mitochondria thus clearly Q formation occurs in a spliced tRNA. In T. brucei, however, tRNA splicing, like in yeast, localizes to the cytoplasm,¹⁹ while TGT localizes to the nucleus. In T. brucei as mentioned before, tRNA^Tyr contains an intron, which must be removed before TGT can use it as a substrate. This again creates a similar situation as tRNA^Phe in yeast; intron-containing tRNA^Tyr must leave the nucleus for splicing and then re-enter the nucleus for Q addition (Fig. 5). Although retrograde transport is necessary for Q formation, its current role in T. brucei is not fully understood. In other organisms, however, studies conducted on the role Q may have on translation suggest the presence of Q can alter the codon bias of Q-containing tRNA. For example, Xenopus microinjections of either G₃₄-containing or Q₃₄-containing tRNA^His aminoacylated with [³H]-labeled histidine indicated a clear preference toward CAC over CAU codons when Q was absent.⁷⁰ Subsequent work has argued that Q can alter codon preference by altering the speed and accuracy of a given Q-containing tRNA codon, which ultimately shapes the proteome.⁷¹

Figure 5.

The biosynthesis of queuosine (Q) in T. brucei requires retrograde transport of tRNA^Tyr to the nucleus. The only intron-containing tRNA in T. brucei, tRNA^Tyr, is exported to the cytoplasm for splicing by the tRNA splicing-specific endonuclease, ligase (Trl1) and phosphotransferase (Tpt1). After splicing tRNA^Tyr travels back to the nucleus where the nucleus-localized tRNA guanosyl transglycosylase (TGT) forms Q₃₄.

Although thus far only yW and Q require retrograde transport, this limited set is more representative of the difficulty in evaluating the intracellular distribution of modifications, rather than an exhaustive list. For example, an interesting possibility exists with the dually localized (nucleus and cytoplasm) modification enzyme Mod5, which generates isopentenyladenosine (i⁶A), a common modification important for translational accuracy in all domains of life. Addition of i⁶A₃₇ only happens after intron-containing tRNAs are spliced, which as discussed previously occurs in the S. cerevisiae cytoplasm.⁷² The possibility then exists that after splicing, a portion of the spliced tRNA travels to the nucleus to receive i⁶A. However, given that significant amounts of Mod5 are detectable in the cytoplasm, it may well be that retrograde transport is not really necessary for i⁶A formation.

Aside from a direct requirement for retrograde transport to facilitate modifications, such as the case of yW and Q, it is possible that dual localization of modification enzymes serves a purpose as a repair mechanism. One could envision a scenario where the tRNA export rate to the cytoplasm may be faster than the rate of nuclear modification, depending on environmental conditions. It then may be possible that hypomodified tRNAs may make it to the cytoplasm; these either get degraded by the RTD pathway or alternatively may be remodified by dually localized enzymes to set them once more into a fully modified, fully functional state. In this realm, constitutive retrograde transport may serve a similar “rescue” function. Perhaps the dual localization of Mod5 reflects such a role. Likely, many tRNAs may take advantage of this “second chance” at modification, which may be advantageous if repairing modifications becomes preferable to degradation and re-synthesis.

Modifications and the mitochondria

Over the course of evolution, mitochondria have undergone loss of genetic material, becoming dependent on nucleus-encoded genes and protein transport from the cytoplasm for function.⁷³ The number of tRNAs encoded by this organelle varies considerably between species and includes a presumably minimal but complete set in humans, all except those for six amino acids in Arabidopsis thaliana and only three tRNAs in Chlamydomonas reinhardtii.⁷³ In the most extreme cases, such as the kinetoplastids (e.g. Leishmania tarentolae and T. brucei) and other protists, no tRNAs are encoded in the mitochondrial genome and the complete set must be imported from the cytosol.^74,75 This led to the question of the evolution of mitochondrial tRNA import and how it relates to tRNA demands set by mitochondrial translation. To answer this question, work performed in C. reinhardtii replaced the often-used GGC and GGT codons with the seldom-used GGG in the mitochondrial genome, increasing the percentage of GGG codons from 0.1 to 0.45%. Strikingly, no increase in import of the tRNA^Gly _CCC, now presumably needed in higher amounts, was observed; the mutants showed decreased respiratory rates caused by low activity of complexes I and IV, increased doubling time and reduced mitochondrial protein synthesis. These results suggest that the correlation between mitochondrial tRNA import and codon usage is most probably a result of co-evolution of both import and translation, and that tRNA import cannot be regulated to quickly cope with changes in mitochondrial genome content.⁷⁶

Overall, tRNA import seems to functionally complement what is encoded in the mitochondrial genome by providing the missing tRNA species necessary for mitochondrial translation. There are cases, however, in which the imported tRNAs are already encoded in the mitochondrial genome: in the plant Marchantia polymorpha, tRNA^Val _AAC is imported from the cytoplasm, where it coexists with the mitochondrial-encoded tRNA^Val _UAC, which alone was considered sufficient for translating all Val codons through wobble pairing.⁷⁷ In yeast, rat and human mitochondria, the nucleus-encoded tRNA^Gln _UUG was shown to co-exist with the mitochondria-encoded tRNA^Gln _UUG after being imported through an ATP-dependent mechanism.^78,79 Finally, in S. cerevisiae, the imported tRNA^Lys _CUU co-exists with the mitochondria-encoded tRNA^Lys _UUU, which was considered sufficient for translating all Lys codons through wobble pairing.⁸⁰ Interestingly, recent work shed some light on to the yeast tRNA^Lys import; under high temperatures (37 °C, as opposed to the normal 30°C for yeast), the mitochondria-encoded tRNA^Lys _UUU becomes under modified and incapable of decoding the AAG codon, with the imported tRNA^Lys _CUU becoming necessary to correct the translational deficiency, indicating that, perhaps the seemingly redundant imported tRNAs become particularly important under certain growth conditions.⁸¹

The determinants for mitochondrial import are still poorly understood and do not appear conserved even within similar organisms, nevertheless, sequence specificity, editing and modification have been described as determinants in some systems.⁸² One noteworthy study compared the cytosolic and mitochondrial populations of tRNA^Glu, tRNA^Gln and tRNA^Lys in L. tarentolae. In this organism, populations of tRNAs are not always evenly distributed among the cell compartments, and can be classified accordingly: group I tRNAs are mostly cytosolic, group II, which includes tRNA^Lys, are mostly mitochondrial and group III, which includes tRNA^Glu and tRNA^Gln, are equally distributed between compartments.⁸³ Their analysis revealed that the cytosolic tRNA^Glu and tRNA^Gln populations contained mcm⁵s²U₃₄, while the mitochondrial populations of the same tRNAs contained only 5-methoxycarbonylmethyl-2′-O-methyluridine (mcm⁵Um). Furthermore, in vitro assays with isolated mitochondria showed that the thiolated tRNAs were not as efficiently imported into the organelle as the non-thiolated ones, suggesting that L. tarentolae uses s²U₃₄ as a negative determinant for mitochondrial import (Fig. 6A ).⁸³ Interestingly, similar work performed in the closely related organism T. brucei showed a different result. Knocked down by RNAi of the cysteine desulfurase (TbNfs), essential for tRNA thiolation, led to disappearance of s²U₃₄, as expected, but had no effect on mitochondrial tRNA import in vivo or in vitro.⁸⁴

Figure 6.

The connection between tRNA thiolation and mitochondrial import in L. tarentolae and T. brucei. A. Thiolation of tRNA^Glu andtRNA^Gln acts as a negative determinant of mitochondrial import in L. tarentolae, but not in T. brucei. B. A portion of tRNA^Trp is kept for cytoplasmic translation, while another is imported into the mitochondria. In the mitochondrial lumen, tRNA^Trp is subjected to thiolation at the unusual position U₃₃ and edited from C to U at position 34. In L. tarentolae, only the edited tRNA gets thiolated whilst in T. brucei, thiolation acts as a negative determinant for editing. The same enzyme Nisf1, is responsible for thiolation of cytoplasmic tRNA^Glu andtRNA^Gln at position U₃₄.

Mitochondrial reliance on nucleus-encoded tRNAs has also lead to an interesting situation beyond simply supplying tRNAs: Reflecting its bacterial ancestry, the mitochondrial genome is not universal and differs from the nuclear code. For example, in most eukaryotes, with the exception of plants, UGA has been reassigned from a stop codon to now mean tryptophan. In turn, most mitochondrial genomes encode a tRNA^Trp with anticodon UCA dedicated to reading the reassigned codon. But this raised an interesting question in kinetoplastid protists, which do not encode any tRNAs in their mitochondria, import all tRNAs from the cytoplasm, and encode no gene in the nuclear genome that could act as a potential suppressor. It was first found in L. tarentolae that these organisms import the standard tRNA^Trp _CCA into mitochondria and following import approximately 50% of the tRNA is edited from C₃₄ to U₃₄ generating tRNA^Trp _UCA that can now decode UGA as tryptophan, while maintaining a significant level of tRNA^Trp _CCA presumably to decode the UGG codons, which also exist in mitochondria.^85,86 The question then is how is the 50/50 ratio between edited and unedited tRNA maintained. It was found that in L. tarentolae, tRNA^Trp was also thiolated at U₃₃, a position that was supposed to be unmodified in all tRNAs in all organisms. However, only the edited tRNA was thiolated, suggesting a connection between editing and thiolation.⁸⁷ A similar situation was later found in T. brucei, but in this case both the edited and unedited tRNAs were thiolated and in addition there was significant amounts of edited but not thiolated tRNA^Trp, suggesting that this modification may not be a requirement for editing.⁸⁸ Separate studies in T. brucei then showed that thiolation acts as a negative determinant for editing, with the down-regulation of tRNA thiolation in this organism, achieved by knockdown of the conserved cysteine desulfurase Nsf1, leading to an increase in C to U editing levels that resulted in almost 90% of tRNA^Trp being edited.⁸⁶ Curiously, mitochondrial thiolation relies on Nsf1, which is also required for cytoplasmic thiolation, but in addition, mitochondrial thiolation requires the mitochondrial enzymes tRNA-specific 2-thiouridylase, Mtu1, and iron-sulfur biogenesis desulfurase interacting protein, Isd11 (Fig. 6).^85,89 Overall, these findings indicate considerable differences between the cytoplasmic and mitochondrial thiolation pathways, and how thiolation and editing are specifically handled between these two closely related organisms.

Finally, there are cases in which tRNA modifications help with naturally occurring extreme structural changes in tRNAs, such as the truncated mitochondrial tRNAs found in nematodes. These tRNAs are unlike cytoplasmic tRNAs, with many lacking the entire D or T-arms. Recognition by tRNA binding proteins is thus dependent on unique proteins including mitochondria-specific elongation factors (EF-Tu1 and 2) that recognize armless tRNAs only.^90,91 In the nematode Ascaris suum, EF-Tu1 recognition of tRNA^Met, which lacks the T-arm, is achieved through unique interactions between the protein C-terminus and the D-arm.⁹² This interaction is dependent both on conserved residues in the D-arm, and in the overall structure of the tRNA, which is altered by the presence of the modification m¹A₉. This methylation leads to a different folding pattern in the D-arm and the small loop region that replaces the missing T-arm. Moreover, m¹A₉ was also shown to be important for efficient aminoacylation, as the structural changes it generates also change the distance between the CCA and the anticodon, affecting the binding efficiency of the corresponding aminoacyl tRNA synthetase.⁹⁰

Conclusions

We have previously proposed that post-transcriptional modifications do not occur in isolation but in fact may be part of largely organized and well-orchestrated chemical cascades.⁹³ One modification may in a subtle manner change the local structure of a target substrate creating a transitory substrate structure that will then be slightly changed by a subsequent modification and so on. Such ideas become even more provocative when one considers intracellular transport dynamics and the permeability barriers set forth by cellular membrane systems. It is our view that cellular transport of tRNAs across membranes may be influenced by many factors and transport rates may change in response to environmental cues. We also suggest that such changes will no doubt impact tRNA processing and tRNA modifications. Environmental sensing will no doubt be at the front and center of how cells couple translational responses with metabolite availability via tRNA modifications. In this review with have highlighted instances where transport dynamics affect modifications. We surmise that elucidating the many levels at which transport, environmental signals and modifications are integrated will be indeed a challenging task, but one that should yield the most interesting observations.

Disclosure of conflicts of interest

The authors declare no conflicts of interest.

Acknowledgment

This work was supported in part by NIH grant GM084065 grant to J.D.A. and a Graduate Fellowship from The Ohio State University Center for RNA Biology to G.S.D. We wish to thank all members of the Alfonzo laboratory for useful comments and discussions.