Two-subunit enzymes involved in eukaryotic post-transcriptional tRNA modification

Abstract

tRNA modifications are crucial for efficient and accurate protein translation, with defects often linked to disease. There are 7 cytoplasmic tRNA modifications in the yeast Saccharomyces cerevisiae that are formed by an enzyme consisting of a catalytic subunit and an auxiliary protein, 5 of which require only a single subunit in bacteria, and 2 of which are not found in bacteria. These enzymes include the deaminase Tad2-Tad3, and the methyltransferases Trm6-Trm61, Trm8-Trm82, Trm7-Trm732, and Trm7-Trm734, Trm9-Trm112, and Trm11-Trm112. We describe the occurrence and biological role of each modification, evidence for a required partner protein in S. cerevisiae and other eukaryotes, evidence for a single subunit in bacteria, and evidence for the role of the non-catalytic binding partner. Although it is unclear why these eukaryotic enzymes require partner proteins, studies of some 2-subunit modification enzymes suggest that the partner proteins help expand substrate range or allow integration of cellular activities.

Keywords:

• cerevisiae
• modification
• Tad2
• tRNA
• Trm7
• Trm8
• Trm9
• Trm61
• Trm112
• Trm734

Introduction

The complexity of the post-transcriptional tRNA modification machinery is remarkable, with 63 genes known to be required for synthesis of the 25 chemically distinct modifications found in the cytosolic tRNA of the yeast Saccharomyces cerevisiae.¹ Over the past decade it has become apparent that formation of a significant number of these modifications requires a complex comprised of 2 different subunits in eukaryotes, but where known, only one protein subunit in bacteria.^1,2 Remarkably, these complexes include 2 tRNA methyltransferases that share the same scaffold (along with 2 other methyltransferases), each catalyzing formation of different modifications,^3-8 as well as one methyltransferase that uses 2 different partner proteins for the same modification on different residues.^9-11Here we discuss these modifications and the corresponding complexes, to shed light on the origin and conservation of the complexes and the functions of the subunits.

Post-transcriptional tRNA modifications occur in all domains of life, including the simplest organisms.^12,13 Modifications in and around the anticodon loop are often critical for translational fidelity and efficiency,¹⁴ whereas modifications in the body of the tRNA often contribute to tRNA folding and stability.^15-17 The importance of tRNA modifications is underscored by their high conservation in different organisms,¹⁸ by the frequent occurrence of a defined growth defect due to deletion of genes required for modifications,^2,19 and by the increasingly frequent association of human diseases with defects in modification.²⁰

Several modifications on specific tRNA residues are known to be formed by a conserved enzyme family within all 3 domains of life.^13,18 Some examples include formation of Ψ₃₈ and Ψ₃₉ (pseudouridine, standard tRNA numbering system) by the Escherichia coli TrmA/S. cerevisiae Pus3 family of pseudouridylases,^22-24 formation of Ψ₁₃ by the _Ec TruD/Sc Pus7 family of pseudouridylases,^25-27 and formation of t⁶A₃₇ (N⁶-threonylcarbamoyladenosine) by the core Sc Sua5/_Ec YrdC and Sc KaeI/_Ec YgjD families of proteins, along with other components depending on the organism.^28-32 By contrast, the highly conserved m¹G₃₇ (1-methylguanosine) modification is catalyzed by the Sc Trm5 family of Rossman fold methyltransferases in eukaryotes and archaea, and by the unrelated _Ec TrmD family of SPOUT methyltransferases in prokaryotes.^33,34

Of particular interest are the modifications that are catalyzed by enzymes comprised of a catalytic subunit and a partner subunit in eukaryotes, but not, apparently, in prokaryotes. In the yeast S. cerevisiae, there are 5 tRNA modification enzymes that are composed of 2 distinct subunits (Fig. 1, Table 1), whereas available evidence suggests only a single gene product is required in bacteria. Moreover, there are 2 additional modifications found in S. cerevisiae and other eukaryotes but not in bacteria that are catalyzed by 2-subunit enzymes.^1,2 Here we discuss the 7 S. cerevisiae tRNA modification enzymes that modify cytoplasmic tRNAs and are comprised of 2 distinct protein subunits. These two-subunit modification enzymes fall into 2 classes, one of which can be further subdivided into 3 subclasses based on common themes, as described below. For each enzyme, we will discuss the occurrence and biological role of the corresponding modification, evidence for the requirement of a partner protein for catalytic activity of the S. cerevisiae protein, evidence for the conservation of this requirement in other eukaryotes, evidence for a single catalytic component in bacteria, and what is known about the function of the non-catalytic binding partner.

Table 1. Eukaryotic tRNA modifications that require 2-subunit enzymes.

modification	S. cerevisiae enzyme^a,b	closest E. coli homolog^c	E. coli enzyme	references
I34	Tad2-Tad3	TadA	TadA	^44,45
m^1A58	Trm61-Trm6 (Gcd14-Gcd10)	TrmI	TrmI	^44,45,58,64,68
Nm32	Trm7-Trm732	FtsJ	TrmJ	^9,10,88
Nm34	Trm7-Trm734 (-Rtt10, Ere2)	FtsJ	TrmL	^9,10,90
m^7G46	Trm8-Trm82	TrmB	TrmB	^78,83
mcm^5U34	Trm9-Trm112	RlmA(I)	n/a^d	^5,6,107,110
m^2G10	Trm11-Trm112	YhdJ	n/a^d	^3

^aFirst protein listed is the catalytic subunit.

^bNames in brackets were original names prior to discovery of involvement in tRNA modification.

^cHomolog of the catalytic subunit of the S. cerevisiae enzyme.

^dModification is not known to occur in bacteria.

Figure 1. S. cerevisiae tRNA modifications formed by 2-subunit enzymes are located throughout the tRNA. Cloverleaf schematic depicting tRNA residues known to be modified (black) in S. cerevisiae, and those that are not modified (gray). Modifications (in brackets) formed by 2-subunit enzymes are labeled.

Heterodimers by Duplication and Divergence

As detailed below, there are 2 examples of eukaryotic 2-subunit enzymes that seem to have arisen by gene duplication events. For both of these complexes, it appears that the eukaryotic non-catalytic subunit has taken the place of one of the catalytic subunits of the bacterial homodimer,^35–37 perhaps to increase the substrate repertoire of the enzyme.^38,39

Tad2-Tad3, the A₃₄ deaminase

Conversion of the wobble residue A₃₄ to I₃₄ (inosine) is thought to occur on the majority of tRNA species that encode an A₃₄ residue in bacteria and eukaryotes, but is not known to occur in archaea, which lack tRNA genes encoding A₃₄.^13,40 In prokaryotes, I₃₄ is only known to occur on tRNA^Arg(ACG), and all 3 bacterial tRNA^Arg(ACG) species that have been examined contain I₃₄. All 6 (of 7) S. cerevisiae tRNA species with an encoded A₃₄ residue that have been examined have I₃₄, and all eukaryotic cytoplasmic tRNAs with an encoded A₃₄ that have been examined have the I₃₄ modification¹³, with the exception of wheat tRNA^Arg(ACG).⁴¹ I₃₄ increases the reading capacity of tRNAs, allowing decoding of codons ending in U and C, and often A,^42,43 and lack of the modification gene is lethal in S. cerevisiae, S. pombe, and E. coli.^44-46

In S. cerevisiae, I₃₄ is formed by the Sc Tad2-Tad3 protein pair.⁴⁴ Sc TAD2 was identified as being homologous to adenosine deaminases acting on RNA, the gene was determined to be essential, a temperature sensitive mutant was generated, extracts from the temperature sensitive mutant were shown to lack A₃₄ deaminase activity, and the activity was complemented by additional recombinant Sc Tad2 purified from E. coli.⁴⁴ However, purified recombinant Sc Tad2 itself lacked deaminase activity, and purification of Sc Tad2 from yeast cells resulted in co-purification of an additional polypeptide, called Sc Tad3. A₃₄ deaminase was concluded to be comprised of an Sc Tad2-Tad3 heterodimer, since Sc Tad2 and Sc Tad3 co-purified in stoichiometric amounts, and biochemical fractions containing both Sc Tad2 and Sc Tad3, but neither alone, could convert A₃₄ to I on a synthetic tRNA^Ala construct.⁴⁴

It is likely that TAD2 and TAD3 arose by a gene duplication event, followed by subsequent sequence divergence, since the 120 amino acid C-terminus of Sc Tad3 is 26% identical and 45% similar to Sc Tad2, and since both Sc Tad2 and Sc Tad3 have a conserved Zn²⁺ coordination motif, as well as a conserved proline that is generally required for ammonium group binding.⁴⁴ Although both proteins are required for binding to tRNA, mutational analysis showed that Sc Tad2 is almost certainly the catalytic subunit of the deaminase. An Sc Tad2 variant with an alanine substitution in the predicted catalytic residue E56 was not active, whereas Sc Tad3 has a valine (V218) in this position and is presumably inactive; moreover, a complex of the Sc Tad3-V218E variant and Sc Tad2 had wild type activity, but the Sc Tad3-V218E variant did not restore activity to the Sc Tad2-E56A variant.⁴⁴

Tad2 and Tad3 homologs are widely found in eukaryotes,⁴⁷ and evidence from eukaryotes other than S. cerevisiae further suggests that the Tad2-Tad3 complex is required for A₃₄ deaminase activity. Thus, a screen for temperature sensitive S. pombe mutants identified a mutant encoding an Sp Tad3 variant with greatly reduced binding of Sp Tad2 and associated with reduced levels of inosine in tRNA.⁴⁸ Moreover, the temperature sensitive phenotype of the Sp tad3 mutant could be suppressed by additional copies of Sp tad2⁺, suggesting that overexpression drove formation of catalytically active complexes.⁴⁸ Similarly, analysis of Trypanosoma brucei A₃₄ deaminase activity suggests the requirement of a Tad2-Tad3 complex, since a Tb Tad2-Tad3 complex has activity, but a Tb Tad2 homodimer lacks catalytic activity.³⁹

By contrast, A₃₄ deamination of tRNA^Arg(ACG) is catalyzed only by TadA (the homolog of Tad2 and Tad3) in prokaryotes,^13,40 based on the occurrence of only this homolog in bacteria,⁴⁷ and the activity of the purified proteins from E. coli and Agrobacterium tumefaciens.^38,45

TadA was shown to be a homodimer based on its crystal structure from Staphylococcus aureus, A. aeolicus, and A. tumefaciens.^36,38,49 The co-crystal structure of the S. aureus TadA homodimer bound to the 15-mer tRNA^Arg(AGC) anticodon stem-loop suggests that substrate binding occurs via an induced fit of the anticodon to the rigid interface between the homodimer via specific contacts predominantly with the 5 nucleotides of the anticodon loop and the C₃₂-A₃₈ pair at the top of the loop.³⁶

Many residues shown to be important for tRNA binding in bacterial TadA are not conserved in Tb Tad2, suggesting a role for Tb Tad3 in substrate binding.⁵⁰ Indeed, some findings suggest that the eukaryotic heterodimer is important for recognition of a larger region on substrate tRNAs to catalyze modification of the multiple different tRNA targets, as opposed to local structural elements of the single tRNA recognized by homodimeric TadA. Thus, although bacterial TadA has activity toward only a stem-loop RNA construct, Sc Tad2-Tad3 requires a full tRNA construct and a proper 3-dimensional structure for activity.^38,44,51 Consistent with these findings, TadA can deaminate eukaryotic tRNA^Arg(ACG), but cannot deaminate other eukaryotic Tad2-Tad3 tRNA substrates.⁴⁵ Remarkably, the Tb Tad2-Tad3 enzyme has C to U ssDNA deaminase activity both in vivo and in vitro, further demonstrating the increased substrate repertoire of eukaryotic Tad2-Tad3 as compared to bacterial TadA.⁵² Alfonzo and colleagues have proposed a model wherein binding of the Zn²⁺ ion occurs intermolecularly, possibly granting increased ability to diversify substrates (as needed for the 7 substrate tRNAs in S. cerevisae) while still maintaining specificity for A₃₄.³⁹

Trm6-Trm61, the m¹A₅₈ methyltransferase

The m¹A₅₈ (1-methyladenosine) modification is commonly found in eukaryotic tRNAs, including 21 cytoplasmic S. cerevisiae tRNAs. This modification is also found less frequently in bacterial and archaeal tRNAs, with a greater frequency of occurrence in tRNA from thermophilic organisms¹³. m¹A₅₈ likely contributes to tRNA stability,⁵³ and S. cerevisiae mutants lacking this modification are inviable due to degradation of hypomodified tRNA_i^Met by the TRAMP complex (Trf4/Air2/Mtr4p polyadenylation complex) and the nuclear exosome.^16,54-57 This specific degradation of only tRNA_i^Met is likely due to loss of stability in an important substructure unique to eukaryotic initiator tRNAs, wherein m¹A₅₈ is involved in hydrogen bonding interactions with residues A₂₀, A₅₄, and A₆₀.^53,58 The requirement of m¹A₅₈ for viability does not extend to all eukaryotes, since S. pombe mutants lacking this modification are viable,⁴⁶ albeit with a slow growth defect.⁵⁹

In S. cerevisiae, m¹A₅₈ is formed by Trm6/Trm61.⁵⁸ Sc TRM6 (also named GCD10) and Sc TRM61 (GCD14) were first identified in screens selecting for mutations that increased GCN4 expression in an Sc gcn2–101 gcn3–101 mutant.^60-62 GCN4 expression is higher when there is less functional eIF2-GTP-tRNA_i^Met (eIF2: eukaryotic initiation factor 2) initiation ternary complex, and it was found that high copy tRNA_i^Met suppressed the temperature sensitive phenotype of Sc gcd10–504 (trm6) mutants and the lethality of Sc trm6Δ and Sc trm61Δ mutants by restoring levels of the initiator tRNA.^58,63 It was also shown that tRNA from Sc trm6Δ and Sc trm61 mutants lacked m¹A₅₈ modification,^58,64 that Sc Trm6 and Sc Trm61 form a complex with m¹A₅₈ catalytic activity, and that activity was dependent on the S-adenosyl methionine (AdoMet) binding domain of Sc Trm61, which is the catalytic subunit.^58,64 The Sc Trm6-Trm61complex appears to be a dimer of heterodimers based on size exclusion chromatography.⁶⁵

The m¹A₅₈ modification enzyme also appears to consist of a Trm6-Trm61 complex in other eukaryotes. Trm6 and Trm61 family proteins are found in yeast, plants, and animals,^66,67 and the human methyltransferase appears to require both Trm6 and Trm61. Thus, co-expression of human TRM6 and human TRM61 suppressed the temperature sensitive growth of Sc trm6–504 and Sc trm61–2 mutants, restored levels of m¹A₅₈ on tRNA in these mutants, and led to m¹A₅₈ formation on human tRNA^Lys introduced on a plasmid, as measured by an altered electrophoretic mobility of the tRNA.⁶⁷ Furthermore, expression of only human TRM6 or human TRM61 did not lead to a substantial increase in m¹A₅₈ modification in mutants, and a complex of human Trm6-Trm61 purified from yeast was able to specifically methylate yeast tRNA_i^Met.⁶⁷

It is likely that TRM6 and TRM61 arose from a gene duplication event followed by sequence divergence, based on the sequence similarity between predicted bacterial Trm61 homologs and eukaryotic Trm6, as well as the conservation of predicted secondary structural elements in eukaryotic Trm6, eukaryotic Trm61, and predicted bacterial Trm61 homologs.⁶⁶ This argument is further strengthened by the finding that many of the residues involved in Mycobacterium tuberculosis TrmI (a bacterial Trm61 homolog) homotetramer formation are also involved in the interaction between Sc Trm6 and Sc Trm61.^35,65

By contrast, m¹A₅₈ modification of bacterial and archaeal tRNA is formed by only TrmI (the homolog of Trm61), based on the occurrence of only one Trm61 homolog and no obvious Trm6 homolog in bacterial and archaeal species,^66,68,69 and the activity of the purified Thermus thermophilus, M. tuberculosis, and Pyrococcus abyssi proteins.^68–70 Based on electrospray ionization mass spectrometry of the native complex, Tt TrmI is a homotetramer,³⁷ and it is also likely that the M. tuberculosis protein is homotetrameric, based on gel filtration analysis.⁷⁰

One of the major functions of Trm6 in the Trm6-Trm61 complex appears to be tRNA binding. Thus, Trm6 contains an RNA recognition motif, and wild type Sc Trm6-Trm61 binds tRNA, as does a complex containing Sc Trm6 and an Sc Trm61 variant with mutations in the AdoMet binding domain, whereas Sc Trm61 by itself does not.⁶⁴ Moreover, mutations of conserved residues predicted to be involved in the interface between Sc Trm6 and Sc Trm61 abrogate tRNA binding and m¹A₅₈ activity (based on the TrmI crystal structure³⁵), but do not appear to disrupt the heterotetrameric complex or AdoMet binding.⁶⁵ These results have led to the speculation that these conserved residues are required to form a Trm6/Trm61 interface that is required for tRNA binding, rather than for complex formation of the proteins themselves.⁶⁵

Acquisition of an Unrelated Partner Protein

As detailed below, there are 5 eukaryotic methyltransferases that have acquired partner proteins unrelated to the catalytic subunit. We further divide these methyltransferases into 3 distinct subclasses.

Acquisition of an unrelated subunit by eukaryotes for the same bacterial modification

This class of modification enzymes catalyzes conversion of G₄₆ to m⁷G₄₆ (7-methylguanosine). The m⁷G₄₆ modification occurs in bacterial and eukaryotic tRNAs,¹³ but to date has not been found in tRNA from archaea, although m⁷G₄₉ is found on tRNA^Leu(UAG) from Thermoplasma acidophilum.⁷¹ m⁷G₄₆ plays a role in stabilizing the tertiary fold of the tRNA, and is part of a commonly occurring base triple with N₁₃ and N₂₂.^72,73 S. cerevisiae strains lacking m⁷G₄₆ have a mild growth defect when grown at 38°C on synthetic media containing glycerol,⁷⁴ and mutants lacking m⁷G₄₆ and m⁵C (5-methylcytidine) are temperature-sensitive due to degradation of tRNA^Val(AAC) by the rapid tRNA decay pathway, which also affects strains lacking m⁷G₄₆ in combination with lack of any of several other modifications in the body of the tRNA.^17,75,76 Additionally, m⁷G₄₆ was shown to be required for growth of T. thermophilus at high temperature, and to be required for subsequent Gm₁₈ (2′-O-methylguanosine) and m¹G₃₇ modification on tRNA^Phe as part of a tRNA modification network.⁷⁷

In S. cerevisiae, m⁷G₄₆ is formed by the Sc Trm8-Trm82 protein pair.⁷⁸ Sc Trm8-Trm82 was discovered using a biochemical genomics approach⁷⁹ when it was found that protein purified from S. cerevisiae strains expressing tagged open reading frames for either Sc TRM8 or Sc TRM82 yielded m⁷G formation activity on pre-tRNA^Phe.⁷⁸ Evidence indicating that the enzyme is composed of the Sc Trm8-Trm82 complex includes the observation that deletion of either gene results in lack of m⁷G modification on tRNA, that the 2 proteins form a stoichiometric complex, that recombinant Sc Trm8 purified from E. coli has low in vitro m⁷G formation activity that is increased 250-fold when co-expressed with Sc Trm82, and that co-translation is required for an active complex.⁷⁸ Sc Trm8 contains a methyltransferase domain and is the catalytic subunit of the enzyme, whereas Sc Trm82 is a WD-repeat protein.

In other eukaryotes, the m⁷G₄₆ modification enzyme also appears to consist of the Trm8-Trm82 protein pair. Trm8 and Trm82 homologs are found in yeast, plants, and animals,⁷⁸ and co-expression of human METTL1 (human TRM8) and WDR4 (human TRM82) complemented the m⁷G₄₆ modification defect in trm8Δ or trm82Δ mutant cells, whereas expression of only human METTL1 or human WDR4 did not.⁷⁸

In humans, it was also recently reported that HeLa cells with reduced levels of human METTL1 and human NSUN2 (required for m⁵C) are sensitive to 5-fluorouracil, resulting in a decrease in tRNA^Val(AAC) levels.⁸⁰ Interestingly, human METTL1 was found to be inactivated by phosphorylation at Ser27 by protein kinase B, suggesting a possible mechanism to regulate m⁷G modification levels in the cell.⁸¹

By contrast, in bacteria m⁷G₄₆ is formed by the Trm8 homolog TrmB alone, based on the apparent absence of Trm82 homologs,⁸² the activity of purified E. coli TrmB,⁸³ and the ability of E. coli TrmB to complement the lack of Sc Trm8-Trm82 in S. cerevisiae trm8Δ trm82Δ double mutants.⁷⁴ _Ec TrmB is monomeric,⁸³ whereas Bacillus subtilis TrmB is homodimeric in solution and in its crystal structure, which has been proposed as a first evolutionary step in the requirement for a dimeric enzyme.⁸⁴

It is clear that part of the role of Sc Trm82 is to maintain levels of Sc Trm8 in yeast, and that Sc Trm82 is also required for other reasons, since Sc Trm8 levels are greatly reduced in Sc trm82Δ mutants, but restoration of levels in S. cerevisiae through Sc Trm8 overexpression only marginally restores m⁷G₄₆ activity.⁷⁴ In the crystal structures of members of the _Ec TrmB/Sc Trm8 family, the B4-αD loop region of unbound Sc Trm8 is in a much different conformation than that of the Sc Trm8-Trm82 complex and that of the B. subtilis TrmB, which have a similar conformation to one another.^84,85 In Sc Trm8, the distinct conformation of the unbound form is stabilized by a salt bridge between R195 and E204, which is unable to form in the Sc Trm8-Trm82 dimer due to steric constraints imposed by Sc Trm82.⁸⁵ The equivalent residue to Sc Trm8 R195 (R129) in B. subtilis TrmB points in the opposite direction compared to unbound R195 in Sc Trm8,⁸⁴ and alanine substitution of the equivalent arginine residue in E. coli TrmB results in loss of more than 90% of its methyltransferase activity,^85,86 suggesting that this residue is important for Trm8/TrmB activity.

Sc Trm82 does not appear to be involved in tRNA binding since tRNA^Phe cross-links only to Sc Trm8, and not to Sc Trm82,⁷⁴ and since the best fit small-angle X-ray scattering (SAXS) model of the Sc Trm8-Trm82 complex bound to tRNA suggests that only Sc Trm8 is involved in tRNA binding.⁸⁵ This SAXS model also suggests that Sc Trm8 binds the tRNA through the local structure around the variable region, especially the D-stem and T-stem,⁸⁵ which is consistent with the finding that deletion of these stems leads to complete loss of methylation activity by Sc Trm8-Trm82, whereas deletion of the acceptor or anticodon stems does not.⁸⁷

Acquisition of 2 different unrelated subunits by eukaryotes for the same modification at different locations

The Trm7 methyltransferase is an example of this subclass, wherein a catalytic subunit engages 2 distinct partner proteins to direct the same 2′-O-methylation activity to different residues: to C₃₂ and N₃₄ to form Cm₃₂ and Nm₃₄. Nm₃₂ and Nm₃₄ modifications are found in tRNAs from eukaryotes, bacteria, and archaea.¹³ Cm₃₂ and Nm₃₄ occur in tandem on 3 tRNA species from S. cerevisiae, and Cm₃₂ and Gm₃₄ modification of tRNA^Phe appears to be highly conserved in eukaryotes, occurring in 16 of 17 eukaryotic tRNA^Phe species that have been examined.¹³ Although Cm₃₂ and Gm₃₄ are found in tandem on tRNA^Phe, they are in chemically distinct environments from each other.⁷² Other Nm₃₂ and Nm₃₄ modifications are found in mammalian, insect, plant, bacterial, and archaeal tRNAs, although not always in tandem,¹³ and Nm₃₂ and Nm₃₄ modifications appear to be important in eukaryotes, although their roles are not well understood. Thus, S. cerevisiae and S. pombe mutants lacking Cm₃₂ and Gm_34, and S. pombe mutants lacking Gm₃₄, have a severe growth defect due to reduced function of tRNA^Phe, whereas S. cerevisiae or S. pombe mutants lacking only Cm₃₂ are healthy.^9-11 Furthermore, Cm₃₂ and Gm₃₄ on tRNA^Phe are also important for the formation of wybutosine (yW₃₇) from m¹G₃₇ in S. cerevisiae and S. pombe^10,11 (Fig. 2). In bacteria and archaea, there are no reported deleterious phenotypes associated with lack of Nm₃₂,^88,89 but lack of Nm₃₄ in E. coli causes a defect in amber (UAG) suppression by tRNA^Leu(CUA), which was suggested to implicate Nm₃₄ in wobble codon:anticodon pairing.⁹⁰

Figure 2. Schematic of FtsJ methyltransferase stem-loop substrates. S. cerevisiae Trm7 requires Trm732 for Cm₃₂ modification and Trm734 for Gm₃₄ modification of the anticodon loop of tRNA^Phe which, as for all tRNAs, has 7 bases. These modifications then drive formation of yW₃₇ from its m¹G precursor. The thicker arrow from Gm₃₄ indicates that yW formation is more dependent on this modification. Other FtsJ family members modify the 5-base A-loop in the rRNA large subunit in different organisms and organelles, as indicated.

In S. cerevisiae, Cm₃₂ is formed by the Sc Trm7-Trm732 protein pair, and Nm₃₄ is formed by the Sc Trm7-Trm734 pair.^9,10 Sc TRM7 was identified by searching for S. cerevisiae homologs of the E. coli 2′-O-methyltransferase FtsJ (RrmJ),⁹ which 2′-O-methylates U₂₅₅₂ residue of the 23S rRNA subunit.^91,92 It was found that tRNA^Phe, tRNA^Leu(UAA), and tRNA^Trp from S. cerevisiae trm7Δ mutants lacked Cm₃₂ and Nm₃₄, and that tagged Sc Trm7 purified from yeast cells was able to form Cm₃₂, and to a lesser extent, Gm₃₄ on tRNA^Phe.⁹ It was later shown that Sc Trm732 is required for Cm₃₂ formation and that Sc Trm734 is required for Nm₃₄ formation by showing that all 3 Sc Trm7 tRNA substrates from Sc trm732Δ or Sc trm734Δ mutants completely lacked their respective Cm₃₂ or Nm₃₄ modifications, that extracts from Sc trm732Δ or Sc trm734Δ mutants were unable to form their respective Cm₃₂ or Nm₃₄ modifications on synthetic substrates, and that Sc Trm7 forms a distinct complex with Sc Trm732, and a distinct separate complex with Sc Trm734.¹⁰ The requirement of Sc Trm732 for Cm₃₂ modification on tRNA^Phe and of Sc Trm734 for Gm₃₄ modification was further demonstrated by the failure of overexpressed Sc Trm7 to suppress the slow growth phenotype of Sc trm732Δ trm734Δ mutants.¹⁰ Sc Trm732 is an armadillo repeat protein that contains a domain of unknown function (DUF2428), and Sc Trm734 is a WD-repeat protein.

The requirement for the Trm7-Trm732 protein pair for Nm₃₂ formation, and of the Trm7-Trm734 protein pair for Nm₃₄ formation is likely conserved throughout eukaryotes. Thus, analysis of 25 eukaryotic genomes comprising all 5 eukaryotic supergroups readily identified Trm7 homologs in all 25 organisms, Trm732 homologs in 22 organisms, and Trm734 homologs in 14 organisms.^11,93 In S. pombe, Sp trm7Δ mutants lack Cm₃₂, Gm₃₄, and yW₃₇ on their tRNA^Phe, Sp trm732Δ mutants lack Cm₃₂, and Sp trm734Δ mutants lack Gm₃₄.¹¹ Furthermore, expression of human FTSJ1 (the predicted TRM7 human homolog) suppressed the growth defect of Sc trm7Δ mutants by forming Cm₃₂ on tRNA^Phe, and suppression and modification by human FTSJ1 required either Sc TRM732 or human THADA (the predicted TRM732 human homolog).¹¹ These findings implicate defective Nm₃₂ and Nm₃₄ modifications in non-syndromic X-linked intellectual disability, since mutations in FTSJ1 are strongly linked to this disease.^94–97

The formation of Nm₃₂ in E. coli and in the archaeon Sulfolobus acidocaldarius requires members of the homodimeric TrmJ SPOUT methyltransferase family,^88,89 which are not obviously related to Trm7, Trm732, or Trm734. _Ec TrmJ appears to require elements in the D-stem and loop for modification activity, whereas Sa TrmJ appears to require elements solely in the anticodon loop.⁸⁹

The formation of Nm₃₄ requires distinct genes in bacteria and archaea, neither of which are related to components of the TRM7 modification machinery. Thus, in E. coli, Cm₃₄ and Um₃₄ on certain tRNA^Leu species are formed by the SPOUT methyltransferase TrmL, which recognizes its substrates by interactions with specific residues, including the N⁶-(isopentenyl)-2-methylthioadenosine modification formed from A₃₇ in the anticodon loop of substrate tRNAs.^90,98 By contrast, in the archaeon Haloferax volcanii, Cm₃₄ formation on tRNA^Trp and on elongator tRNA^Met require box C/D snoRNPs (small nucleolar ribonucleoprotein) specific to each corresponding pre-tRNA.^99,100

Comparison of Trm7 to other FtsJ family proteins suggests a possible reason for the requirement of additional proteins for Trm7 activity, since the stem-loop tRNA substrates modified by Trm7 are slightly different than the rRNA stem-loop substrates modified by other proteins in this family (Fig. 2). Thus, E. coli FtsJ (which is ∼34% identical to Sc Trm7) methylates the first residue (Um₂₅₅₂) of the 5-base A-loop in the 23S rRNA,^91,92 S. cerevisiae Mrm2 (∼29% identical to Sc Trm7) methylates the first residue (Um₂₇₉₁) of the 5-base A-loop in the the 21S mitochondrial rRNA,¹⁰¹ and S. cerevisiae Spb1 (∼34% identical to Sc Trm7) methylates the second residue (Gm₂₉₉₂) of the same 5-base A-loop of cytoplasmic 27S pre-rRNA,¹⁰² each apparently acting alone. We therefore speculate that Sc Trm732 may help Sc Trm7 to recognize and modify the first residue of the 7-base loop of tRNA (as opposed to the 5-base loop in rRNA), and that Sc Trm734 may help Sc Trm7 recognize the N₃₄ residue, which is the third residue in the anticodon loop, and chemically distinct from the substrate residues of the other known FtsJ family members.

S. cerevisiae Trm734 has also been implicated in endoplasmic recycling, seemingly unrelated to tRNA modification. Sc TRM734 (ERE2) was identified in a genome-wide screen for deletion mutants with increased canavanine resistance due to defects in endoplasmic recycling.¹⁰³ It was suggested that Sc Trm734 regulates the function of Sc Ere1, which was identified in the same screen, since it interacts with Trm734 in membrane bound fractions on a glycerol gradient, co-immunoprecipitates with Sc Trm734, and since some Trm734 colocalized with Sc Ere1 in endosomal compartments in ESCRT (endosomal sorting complexes required for transport) mutant cells. Thus, it may be that cytosolic Sc Trm734 functions for tRNA modification, and that membrane-bound Sc Trm734 plays a role in endoplasmic recycling.¹⁰³

Sc TRM734 (RTT10) was also identified in a screen for S. cerevisiae deletion mutants that showed increased Ty1 retrotransposition.¹⁰⁴ In retrospect, however, this defect is likely due to decreased expression of an important protein(s) involved in repression of Ty1 transposition caused by lack of Nm₃₄ modification, since Sc TRM7 was identified in the same screen.

Acquisition of a common unrelated partner protein for different methyltransferases

The Trm9 and Trm12 methyltransferases are examples of proteins in this subclass, wherein different catalytic subunits engage the same partner protein Trm112 to direct different chemical modifications on different residues. The first tRNA methyltransferase that requires Trm112 is responsible for formation of the terminal methyl group of mcm⁵U₃₄ (5-methoxycarbonylmethyluridine) and related modifications. The mcm⁵U family of modifications is found only in eukaryotes and is implicated in efficient reading of AGA and AAG codons by tRNA^Arg(UCU) and tRNA^Glu(UUC), respectively in S. cerevisiae.¹⁰⁵ Lack of the modification is associated with sensitivity to DNA damaging agents in yeast and in humans,^105,106 as well as sensitivity to aminoglycosides at high temperature and resistance to zymocin-mediated tRNA cleavage and cell death in yeast.^107-109

In S. cerevisiae, the terminal methyl group of mcm⁵U₃₄ and mcm⁵S²U₃₄ (5-methoxycarbonylmethyl-2-thiouridine) is formed by the Sc Trm9-Trm112 protein pair.^{5,6,107,109,110} Sc TRM9 was identified as a putative methyltransferase by bioinformatics, tRNA from S. cerevisiae trm9Δ mutants was shown to lack the mcm⁵U and mcm⁵S²U modifications, and tagged Sc Trm9 protein purified from yeast was shown to methylate a saponified tRNA extract, demonstrating that Sc Trm9 is required for formation of the terminal methyl group of mcm⁵U.¹⁰⁷ Subsequently, several suppressors of zymocin toxicity in S. cerevisiae were identified as mutants encoding Sc Trm9 variants lacking the C-terminal domain required for interaction with Sc Trm112.^109,110 The requirement of Sc Trm112 for mcm⁵U and mcm⁵S²U was later determined explicitly by showing that tRNA from Sc trm112Δ mutants lacked the terminal methyl group of the mcm⁵U modification, and that purified Sc Trm9- Trm112 from E. coli had this methylation activity, whereas Sc Trm9 alone did not.^5,6 Interestingly, lack of either Sc Trm9 or Sc Trm112 gives rise to ncm⁵U (5-carbamoylmethyluridine) instead of the expected cm⁵U modification in yeast, raising questions about the biochemistry of mcm⁵U formation.⁶

The requirement of Trm9-Trm112 for mcm⁵U is likely conserved in other eukaryotes. Thus, predicted Trm9 proteins are found in yeast, worms, flies, and humans,¹⁰⁷ and predicted Trm112 proteins are found in fungi, animals, and plants.⁴ Additionally, the Mus musculus Trm9 and Trm112 homologs are required for mcm⁵U formation in mice,¹¹¹ and the Trm9 homolog is required for this modification in human cells.¹⁰⁶ However, the role that Trm112 plays for the methyltransferase activity of Trm9 is not yet clear.

Trm112-Trm11, the m²G₁₀ methyltransferase

The second tRNA methyltransferase that requires Trm112 is responsible for formation of the m²G₁₀ (N²-methylguanosine) modification, which is found in eukaryotes and archaea, but not in bacteria.¹³ Although the precise role of this modification is not clear, it is likely involved in tRNA stability since lack of this modification in combination with lack of m^2,2G₂₆ (N²,N²,-dimethylguanosine) in S. cerevisiae results in slow growth.³

In S. cerevisiae, m²G₁₀ is formed by the Sc Trm11-Trm112 pair.³ Sc TRM11 was identified by bioinformatics approaches as a putative methyltransferase, and it was found that extracts from S. cerevisiae cells lacking Sc TRM11 were unable to catalyze formation of m²G on synthetic tRNA^Ile(UAU). However, although purified Sc Trm11 from yeast yielded m²G activity, Sc Trm11 purified from E. coli did not, and it was shown that Sc trm112Δ mutants lacked m²G activity, that tagged Sc Trm112 purified from yeast exhibited m²G activity, and that Sc Trm11 and Sc Trm112 formed a complex.³ Sc Trm112 was later shown explicitly to be required for Sc Trm11 activity using a wheat germ cell free system.¹¹²

Since Sc Trm11 and Sc Trm112 homologs are found widely in eukaryotes and in archaea,³ it is likely that they are required for m²G₁₀ activity in other organisms, although to our knowledge this has not been tested experimentally. The precise role of Sc Trm112 for Sc Trm11 activity is not known, but Sc Trm11 stability does not appear to be affected by Sc Trm112 levels.³

Trm112 has distinct roles in other methyltransferase complexes

Trm112 is also required for the function of 2 other 2-subunit methyltransferase enzymes. We will briefly discuss these enzymes because studies on their structure and function suggest several possible roles for Trm112 in Trm9 and Trm11 function, and because the relative amounts of each complex may affect the methylation activity of the other Trm112-containing complexes, since Trm112 appears to be limiting in S. cerevisiae.^8,113

Analysis of the Mtq2-Trm112 complex suggests that Trm112 functions to solubilize Mtq2 and to place it in an active conformation.^4,114 In S. cerevisiae, Sc Mtq2-Trm112 is the enzyme responsible for N⁵ methylation of the glutamine residue of the GGQ tripeptide motif of eukaryotic release factor 1 (eRF1).^4,115,116 The crystal structure of Mtq2-Trm112 from Encephalitozoon cuniculi consists of a single heterodimer, and suggests that E. cuniculi Trm112 may function to solubilize E. cuniculi Mtq2 by masking a hydrophobic patch involved in the dimer interface.¹¹⁴ Indeed, the presence of Sc Trm112 increases the solubility of Sc Mtq2 when the proteins are expressed in E. coli.⁴ Analysis of the crystal structure also suggests that E. cuniculi Trm112 may be important for the conformation of the AdoMet binding domain of E. cuniculi Mtq2, since a loop motif is involved in both interaction with E. cuniculi Trm112 and with the AdoMet cofactor.¹¹⁴

Analysis of the overall structure of the E. cuniculi Mtq2-Trm112 complex suggests that Trm112 may also function in substrate binding.¹¹⁴ Although the structure of the methyltransferase domain of E. cuniculi Mtq2 is most similar to that of E. coli PrmC (the protein that N⁵ methylates the Gln residue of the GGQ motif of bacterial release factors RF1 and RF2^117,118), the overall structure of the E. cuniculi Mtq2-Trm112 complex most closely resembles that of E. coli RlmA(I),¹¹⁹ which methylates 23S rRNA residue G₇₄₅.^120,121 The Zn²⁺ binding domain of E. cuniculi Trm112 superimposes well with the Zn²⁺ binding domain of E. coli RlmA(1), which has been implicated in rRNA binding,¹¹⁹ suggesting that Trm112 may be important for substrate recognition or binding.¹¹⁴

In addition, analysis of RlmA(I) structure suggested a role for Trm112 in binding and activity of Trm9, since RlmA(I) is the closest E. coli homolog of S. cerevisiae Trm9, and a model of Trm112 binding to Trm9 correctly predicted residues important for both Trm9-Trm112 protein-protein interactions and activity.¹¹⁴

Analysis of the Bud23-Trm112 complex indicates that Trm112 maintains Bud23 protein levels.⁷ In S. cerevisiae, the Sc Bud23-Trm112 complex is required for the m⁷G modification of G₁₅₇₅ on the 18S subunit of rRNA in yeast.^7,8,122 Sc trm112Δ mutants have severely reduced levels of Sc Bud23,⁷ and surprisingly, the Sc bud23Δ slow growth phenotype can be rescued by expression of a catalytic null Sc Bud23 variant,¹²² further indicating that the role of Sc Trm112 in this complex is not for catalysis. Moreover, the slow growth phenotype of the Sc trm112Δ mutant is almost completely due to Bud23 defects, since the Sc bud23Δ trm112Δ double mutant grew as poorly as the Sc bud23Δ or Sc trm112Δ single mutants.^5,8

It appears that Sc Trm112 may also act to integrate methylation signals between tRNA, the ribosome, and release factors (Fig. 3).¹¹⁴ Because Sc Trm112 is likely present in non-stoichiometric amounts compared to its methyltransferase partners,^8,113 the activity of each methyltransferase partner is sensitive to the relative amounts of the other Sc Trm112 methyltransferase partners. For instance, overexpression of Sc Trm11 or Sc Mtq2 results in resistance to zymocin toxicity and reduced binding between Sc Trm112 and Sc Trm9. Furthermore, overexpression of Sc Mtq2 leads to cell growth defects, presumably by titrating Sc Trm112 away from Sc Bud23, and this defect is partially rescued by overexpression of Sc Bud23.⁷

Figure 3. Trm112 partners with several methyltransferases involved in diverse translational processes. Schematic of the S. cerevisiae Trm112 physical interactions known to affect the activity or stability (in brackets) of its partner methyltransferases. (*) The Sc Bud23 protein, but not its methyltransferase activity, is required for m⁷G₁₅₇₅ formation on the cytoplasmic 18S rRNA subunit.

Concluding Remarks

Several themes emerge upon analysis of the protein pairs involved in eukaryotic tRNA modification. As described above, tRNA modification enzymes have acquired a second subunit either by duplication of a gene or by acquisition of an unrelated partner protein, but with multiple different properties. One functional theme that has emerged is that the addition of a partner protein may allow for an expansion of the substrate repertoire for a given enzyme, as for the eukaryotic Tad2-Tad3 enzyme, which modifies an additional 6 tRNA species (in S. cerevisiae) in contrast to its homodimeric bacterial TadA counterpart, which appears to only modify tRNA^Arg(ACG). A second theme is the expanded location of modifications facilitated by the acquisition of different partners for eukaryotic Trm7 activity, allowing modification of 2 different residues in chemically distinct environments in the same tRNA species. A third theme that emerges is the distinct possibility of cross-pathway regulation by modification. This is most clear for the different methyltransferases that each require association with limiting amounts of Trm112 for crucial methylations involved in translation termination, tRNA modifications, and ribosome biogenesis. The modulation of several methyltransferase activities by one protein in theory allows the integration of multiple signals to fine-tune translation. Similarly, it is possible that Trm734 could also act to integrate different cellular functions because of the involvement of Trm734 in endoplasmic recycling and retrotransposition.

While we have pointed out some of the common themes in these 2-subunit eukaryotic tRNA modification enzymes, it remains unclear precisely how the non-catalytic subunits function in the modification complexes. Although we have cited evidence describing roles of these proteins in maintaining stability, altering conformations of proteins, expanding substrate repertoire, and promoting substrate binding, in many cases the evidence is indirect and limited. Furthermore, the mechanism by which unrelated partners were acquired is not clear in most cases. Further research will undoubtedly lead to insight into these and other questions regarding the roles of partner proteins in these important and conserved modification reactions.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank E. Grayhack for discussions.

Funding

This work is supported by NIH grant GM052347 to EMP.