Targeting mitochondrial and cytosolic substrates of TRIT1 isopentenyltransferase: Specificity determinants and tRNA-i6A37 profiles

The tRNA isopentenyltransferases (IPTases), which add an isopentenyl group to N6 of A37 (i6A37) of certain tRNAs, are among a minority of enzymes that modify cytosolic and mitochondrial tRNAs. Pathogenic mutations to the human IPTase, TRIT1, that decrease i6A37 levels, cause mitochondrial insufficiency that leads to neurodevelopmental disease. We show that TRIT1 encodes an amino-terminal mitochondrial targeting sequence (MTS) that directs mitochondrial import and modification of mitochondrial-tRNAs. Full understanding of IPTase function must consider the tRNAs selected for modification, which vary among species, and in their cytosol and mitochondria. Selection is principally via recognition of the tRNA A36-A37-A38 sequence. An exception is unmodified tRNATrpCCA-A37-A38 in Saccharomyces cerevisiae, whereas tRNATrpCCA is readily modified in Schizosaccharomyces pombe, indicating variable IPTase recognition systems and suggesting that additional exceptions may account for some of the tRNA-i6A37 paucity in higher eukaryotes. Yet TRIT1 had not been characterized for restrictive type substrate-specific recognition. We used i6A37-dependent tRNA-mediated suppression and i6A37-sensitive northern blotting to examine IPTase activities in S. pombe and S. cerevisiae lacking endogenous IPTases on a diversity of tRNA-A36-A37-A38 substrates. Point mutations to the TRIT1 MTS that decrease human mitochondrial import, decrease modification of mitochondrial but not cytosolic tRNAs in both yeasts. TRIT1 exhibits clear substrate-specific restriction against a cytosolic-tRNATrpCCA-A37-A38. Additional data suggest that position 32 of tRNATrpCCA is a conditional determinant for substrate-specific i6A37 modification by the restrictive IPTases, Mod5 and TRIT1. The cumulative biochemical and phylogenetic sequence analyses provide new insights into IPTase activities and determinants of tRNA-i6A37 profiles in cytosol and mitochondria.

Introduction 45 different eukaryotic cytoplasmic (cy-) and about half as many mitochondrial (mt-) tRNAs contain unique sets of modifications that are considered in two groups [1][2][3]; those in the body or core, which contribute to folding and/or stability, and those in the anticodon loop (ACL) which contribute to mRNA decoding by cy-and mt-ribosomes. Modifications in the anticodon stem loop (ASL) are more concentrated and diverse than those in the tRNA body. One of these is isopentenylation of N 6 of adenosine, found only on tRNAs that decode UNN codons, at position 37 (i 6 A37) directly 3' to the anticodon. Bacterial i 6 A37 is hypermodified to 2-methylthio-N 6 -isopentenyl-A37 (ms 2 i 6 A37), whereas eukaryotic cy-tRNA-i 6 A37 is not. Also, while the mt-tRNAs of S. cerevisiae and S. pombe are not hypermodified, mammalian mt-tRNAs i 6 A37 are [4][5][6] (Table 1). i 6 A37 increases tRNA affinity for the ribosome, increases anticodon-codon pairing occupancy, can activate mRNA decoding and decrease frameshifting (reviewed in [7, see 8]). ASL modifications contribute to "circuits" [9], in which one is prerequisite for formation of others, and whose disruptions can cause disease [10,11]. I 6 A37 is required for m 3 C32 formation on the tRNAs Ser that decode UCN codons [12,13].
The blank white rectangle spaces denote that the tRNA is not expressed in this species [71]. # Position 34 of S. pombe mt-tRNA Trp is encoded as C [84] whereas in S. cerevisiae [50,104], bovine and human [105], it is encoded to be U, and was determined to be 5-taurinomethyluridine (tm 5 U) in bovine [4], and cmnm 5 U in S. cerevisiae [92] (see tRNAmodViz server [83]).

�
A designates an A37 that is not in an A36-A37-A38 context, containing C38 [4] instead of A as is found in its E. coli counterpart. The pattern listed in the human or bovine column is also the same for C. elegans, Drosophila melanogaster, Xenopus tropicalis, Gallus gallus, and Mus musculus [71]. yW = ybutosine.
All entries with subscript 37 have been verified as modified or unmodified; the A 37 in the S. cerevisiae MOD5 column indicates an exceptional unmodified A37 in a A36-A37-A38 context (see text).
Entries in parentheses show the DNA-encoded nucleotide but have not been validated as modified or not in RNA [4,43,59,60].
The mt-tRNAs Trp recognize 2 sense codons in budding yeast and animal mitochondria [109] [4]. We note that several yeast species encode cy-tRNAs Trp with G37 ( [71], see below), and for the two mitochondrial genomes we examined, Kluyveromyces lactis and The single gene-encoded Mod5 localizes to nuclei, nucleoli, cytoplasm and mitochondria [34][35][36]. Nuclear Mod5 is involved in tRNA gene-mediated silencing [37,38]; its localization [36] is consistent with this and early tRNA processing [39,40]. For some tRNAs, A36-A37-A38 is interrupted by an intron, and in yeast where tRNA splicing occurs on the surface of mitochondria, tRNA nuclear export is prerequisite to i 6 A37 formation [2,41, also see 42]. A Mod5 nuclear localization signal [36] is C-terminal to the eukaryote-specific IPTase Znfinger motif [7] that binds the tRNA backbone [24], and is required for the activity of the S. Pombe IPTase Tit1 [43]. I 6 A37 modification of mt-tRNA in the mitochondrial matrix is a critical TRIT1 function [14]. Pathophysiology of TRIT1-R323Q-associated neurodevelopmental disease is attributable to deficient mt-tRNA modification and consequent impairment of mitochondrial mRNA translation, even though cy-tRNAs Ser are also hypomodified [14]. Though S. cerevisiae and C. elegans use alternative translation initiation to target their IPTases to mitochondria [34,35,44], the mechanism used by TRIT1 would appear to differ (Fig 1) yet had not been subjected to experimental analysis. In the first part of this study we examine mitochondrial targeting of TRIT1 in human cells. We then extend this to TRIT1 mitochondrial function in S. pombe and S. cerevisiae model systems. Later, we examine modification of a diversity of tRNA substrates by TRIT1, Mod5 and Tit1, and the extent to which substrate-specific negative determinants may contribute to tRNAs-i 6 A37 profile variation (below).
A conserved positive determinant of all known IPTase substrates is the A36-A37-A38 recognition sequence required for i 6 A37 formation. Structures of bacterial and S. cerevisiae IPTases bound to their substrates and biochemical studies show that they directly recognize the ACLs of their target tRNAs and mediate i 6 A37 formation using dimethylallyl-pyrophosphate (DMAPP) as the isopentenyl donor [24,[45][46][47][48]. Intriguingly, despite these conserved features and effects of i 6 A37 on translation, variation exists in the tRNAs-i 6 A37 profiles among species (Table 1). Although i 6 A37 occurs widely on bacterial tRNAs for UNN codons, fewer cy-tRNAs bear i 6 A37 in S. pombe, S. cerevisiae and human (Table 1). tRNA-i 6 A37 identities also vary between mitochondria and cytosol, within and among species. Better understanding of a basis of this variation is a goal of our study.  The region between M1 and M12 of Mod5 that is required for mitochondrial targeting is indicated by a green rectangle above [34,35]. The second methionines that may be used as alternative translation start sites for cytoplasmic localization are boxed in blue. The invariant Threonine within the conserved P-loop that functions in catalysis is in the yellow rectangle. Asterisks indicate the conserved DSMQ sequence that forms a network of contacts that mediate A37 recognition; the D of which acts as a general base for catalysis [24] and the M of which is M57 of TRIT1(57-467). As predicted by MitoFates [53], the TOM20-binding site of human TRIT1 is underlined with a blue bar and the amphipathic helix of TRIT1 and Mod5 are indicated with green bars. The two key basic residues in the conserved amphipathic helix of mammalian MTS that were mutated in human TRIT1 are in red rectangles.
https://doi.org/10.1371/journal.pgen.1008330.g001 Table 1 emphasizes that while a single gene encodes the cytoplasmic and mitochondrial IPTases of the species listed, the cy-and mt-tRNA-i 6 A37 profiles differ. Specifically, while TRIT1 modifies only tRNAs Ser in cytosol, it modifies five tRNA types in mitochondria (Table 1). Table 1 represents a phylogenetic analysis that reveals substitution of A37 with G as a major source of variability among cy-tRNAs-i 6 A37, especially eukaryotes. tRNA Ser GGA is only one of ten A36-A37-A38 tRNAs in E. coli that lack i 6 A37; weak ASL base pairs decrease its substrate activity for MiaA [49]. Lack of i 6 A37 on S. cerevisiae cy-tRNA Trp CCA-A37-A38 is attributable to its CCA ACL as determined in vivo and in vitro, the latter with recombinant Mod5 on cell-derived cy-tRNA Trp and synthetic ASL substrates [43]. By contrast, S. pombe Tit1, which has an extended anticodon recognition loop relative to Mod5, readily modifies S. pombe cy-tRNA Trp CCA and the same ASL substrates [43]. Hence, Mod5 is denoted as a restrictive, and Tit1 as a nonrestrictive type eukaryotic IPTase. It is notable that while A37 of cy-tRNA Trp CCA-A37-A38 is unmodified in S. cerevisiae, the A37 of mt-tRNA Trp UCA-A37-A38 was reported as modified with i 6 A [50, see 51, Table 1, below]. Accordingly, the S. cerevisiae tRNA-i 6 A37 identity profile is defined in part by the A36-A37-A38 motif and in part by a distinct characteristic of its IPTase. Yet, the significance of the cy-tRNA Trp CCA exception in which the AC is a negative determinant in the presence of A36-A37-A38 was enigmatic and deserved further consideration.
Thus, it might appear that tRNA-i 6 A37 variation can be accounted for by a simple two component model, i) direct IPTase recognition of A36-A37-A38, ii) A37G substitution in tRNAs that decode UNN codons, and an exception, the unmodified A37 of cy-tRNA Trp C-CA-A37-A38 in S. cerevisiae. The exception is intriguing because its proposed mechanism involves the AC recognition loop of Mod5, mutagenesis of which altered the hierarchical substrate preference for tRNA Ser , tRNA Tyr , and tRNA Cys [43]. By this model, cy-tRNA Trp CCA would be a very poor competitive substrate for the restrictive Mod5. However, this model also raises the possibility that paucity of cy-tRNA-i 6 A37 in higher eukaryotes reflects an IPTase type hyper-restriction that expanded to additional cy-tRNA isoacceptors. In this case, ASLs with A36-A37-A38 that could not have been modified may have undergone A37G substitution during evolution (Table 1). For example, TRIT1 might be restrictive against human cy-tRNA-Tyr GUA-G37-A38 even if it had A37 because it bears other ASL determinants that negatively impact activity. In such a model, TRIT1 would modify the human mt-tRNAs-i 6 A37 because they do not bear negative determinants. An attempt to experimentally test a prediction of this model has not been performed. In addition, a restrictive vs. nonrestrictive type analysis had not been extended to a higher eukaryotic IPTase.
Here, we first establish that TRIT1 uses an N-terminal MTS in human cells, and that its function is required for modification of mt-tRNAs, but not cy-tRNAs, in S. pombe and S. cerevisiae. In addition, new phylogenetic and experimental analyses advance our understanding of tRNA-i 6 A37 identity determinants in eukaryotes. Presence of G37 rather than A in the tRNAs for UNN codons is the major determinant of the progressive paucity of cy-tRNAs-i 6 A37 in eukaryotes including human. A role for IPTases independent of A36-A37-A38 appears influential but limited to substrate-specific restriction against tRNA Trp CCA-A37-A38. We show that the generally nonrestrictive Mod5 and TRIT1 are restrictive against tRNA Trp C-CA-A37-A38, largely dependent on the source of the tRNA, S. cerevisiae or S. pombe which are different in their ACLs only at position 32. Additional data support that C32 is part of a tRNA Trp CCA substrate-specific inhibitory circuit for Mod5 (and TRIT1). The collective data advance knowledge of how tRNA sequence and IPTase activities contribute to cy-and mt-tRNA-i 6 A37 profiles.

TRIT1 has a mitochondrial targeting sequence (MTS) but is not expected to undergo cleavage
Mitochondrial targeting of human TRIT1 was predicted by the online tool MitoProt II [14] which also predicted a proteolytic cleavage site after amino acid 47 [52]. However, such cleavage would remove the conserved P-loop involved in DMAPP binding and catalysis (Fig 1, and  below). We used the prediction algorithm MitoFates which was designed as an improved method for predicting cleavable N-terminal MTSs, referred to as presequences, and their cleavage sites [53]. MitoFates predicted a MTS comprised of a TOM20 (translocase of the outer mitochondrial membrane-20) receptor binding site at amino acids 5-10 (AAARAV, Fig  1) followed by an amphipathic helix at position 11-23 (PVGSGLRGLQRTL), but with the low probability of cleavage of 0.106, well below the default cutoff of 0.5. We note that the cNLS mapper [54] predicted overlapping mono-and bi-partite importin-dependent nuclear localization signals at amino acids 420-453 of TRIT1.
We created multiple GFP fusion constructs to test for mitochondrial localization by confocal microscopy of human embryonic kidney (HEK)-293 cells (Fig 2A and 2B). To test for MTS activity of the TRIT1 N-terminal region we fused only its first 51 amino acids to GFP, as in . D) Western blot subcellular fractionation analysis. Transfected HEK293 cells were fractionated 48-hour later into supe (cytosol), mitochondria (mito) and mitochondria treated with Proteinase K (ProtK), the latter to digest proteins on the outer membrane of the mitochondria. GFP-TRIT1 WT and RR>EE mutant was detected using anti-GFP Ab. NDUFA9 is a mitochondrial inner membrane protein, La is a nuclear and cytosolic protein, TOMM20 is on the mitochondrial outer membrane. Total protein staining with Ponceau S is shown in the lower panel. Asterisks indicate BSA that was added during the mitochondrial isolation protocol (see text).
https://doi.org/10.1371/journal.pgen.1008330.g002 TRIT1(1-51)-GFP, and examined it for mitochondrial targeting. Two key point mutations, R17E, R21E, in the basic amphipathic helix component of the predicted MTS (Fig 1) were made in two contexts, the N-terminal fusion 1-51 construct, TRIT1(1-51:R17E/R21E)-GFP, and full length TRIT1(R17E/R21E)-GFP. These and other constructs were transfected into HEK293 cells. After 48 hr. the cells were exposed to MitoTracker to stain mitochondria red, fixed and treated with Hoechst to stain nuclei blue (Fig 2B). Cell lysates were examined for protein expression by western blot using anti-GFP antibody (Fig 2C). Because a significant fraction of TRIT1 localizes to nuclei, it was helpful to visualize the patterns in which the Hoechst stained nuclei were and were not included in the merged images as in rows 4 and 5 of Fig 2B respectively. Full length TRIT1-GFP was found localized to nuclei and cytosol with some evidence of mitochondrial localization reflected by overlap of GFP and MitoTracker as a stippling yellow-orange pattern in the merged panels ( Fig 2B, columns 1 and 2, lower two panels). This pattern was not observed with full length TRIT1(R17E/R21E)-GFP containing the MTS mutations, which instead accumulated mostly in nuclei and was not found to localize to mitochondria (Fig 2B,  columns 3 and 4). Even the least bright TRIT1-GFP expressing cells in the lower left corner of column 1, row 1 show orange stippling in column 1 row 5, whereas cells of comparable GFP brightness in column 3, row 1, show more red stippling in column 3 row 5.

Mitochondrial fractionation shows that the R17E, R21E mutations debilitate the TRIT1 MTS
We next performed subcellular fractionation on HEK293 cells after transfection of full length TRIT1-GFP and MTS mutant, TRIT1(R17E/R21E)-GFP. Western blots of total cell extracts, protein contents of their mitochondrial subfractions and supernatants, as well as of mitochondria after proteinase K-treatment, are shown in Fig 2D. TRIT1-GFP was detectable in total cell extract and supernatant (lanes 2 & 3), as well as in the contents of the mitochondrial and the proteinase K-treated mitochondria fractions (lanes 4 & 5), consistent with TRIT1 presence in the mitochondrial matrix, whereas TRIT1(R17E/R21E)-GFP was relatively lacking in the mitochondrial and proteinase K-treated mitochondria samples (lanes 8 & 9). This is consistent with localization of TRIT1 to nuclei and cytoplasm including in the mitochondrial matrix where its mt-tRNA substrates reside as previously determined by subcellular fractionation [14]. Additional analysis of the same fractions revealed the expected results for NDUFA9, an inner mitochondrial membrane protein, evident by its comparable detection in the mitochondrial and proteinase K treated mitochondria fractions in both wild-type TRIT1 and the MTS (R17E/ R21E) (RR>EE) mutant TRIT1 extracts (NDUFA9, lanes 4 & 5, and 8 & 9). Likewise, the expected fractionation was also observed for Tom20, a mitochondrial outer membrane protein, in both sample sets, in the mitochondrial (Tom20, lanes 4 & 8) but not the proteinase K treated mitochondria fractions (lanes 5 & 9), and for La, an abundant soluble nuclear and cytoplasmic protein ( Fig 2D, as indicated). We note that the slight mobility differences observed for the TRIT1-GFP bands in lanes 3 and 7 and their neighboring lanes is likely due to perturbation from BSA added to those samples as per the protocol (asterisks in Fig 2D bottom panel).
Our collective data support the idea that TRIT1 contains multiple trafficking elements that distribute it to nuclei, cytoplasm and inner mitochondrial matrix, similar to Mod5 [34-36, 38, 55, 56]. The MTS R17E/R21E mutations shifts distribution away from mitochondria, mostly to nuclei ( Fig 2B). Importantly, as will be shown below, the MTS mutations impair i 6 A37 modification of mt-tRNAs specifically, without impairing modification of the nuclear-encoded cy-tRNAs by TRIT1.

TRIT1 MTS-dependent i 6 A37 modification of mt-tRNAs in S. pombe
Mitochondrial protein import pathways have been functionally conserved in eukaryotes including mammals and yeasts [57,58]; S. cerevisiae, S. pombe, D. melanogaster and H. Sapiens share Tom20, five other TOM homologs, and twelve of the thirteen TIM (translocase of the inner mitochondrial membrane) system components [58]. Analysis of S. pombe mt-DNA by tRNAscan-SE predicts 25 mt-tRNAs (including 3 for Met), of which two, mt-tRNA Trp CCA and mt-tRNA Cys GCA contain A36-A37-A38. We examined the R17E/R21E point mutations that impaired mitochondrial localization in human cells for effects on mt-tRNA and cy-tRNA modification in S. pombe (Fig 3A-3E). For this, we used the unaltered nmt1 + promoter in the pRep4X expression vector as was done previously to examine mt-tRNA Trp CCA modification by Mod5 in S. pombe [6]. MOD5 uses alternative translation start sites, M1 to target the MTScontaining protein to mitochondria and cytoplasm, and M12 which produces the MTS-lacking protein that accumulates in S. cerevisiae cytoplasm and nuclei [34,35]. Fig 3A shows results of an i 6 A37-sensitive PHA6 blot assay for mt-tRNA Trp CCA, mt-tRNA Cys GCA and cy-tRNA Tyr-GUA. The PHA6 (Positive hybridization in Absence of i6A37) assay is based on loss of oligo-DNA probe hybridization to the ACL due to presence of i 6 A37, and is calibrated for quantitation by a body probe to the T stem loop on the same tRNA ( Fig 3A upper and lower panels) [6,43,59,60]. Here, pRep4X-TRIT1 led to~45% i 6 A37 modification of mt-tRNA Trp CCA and 65% modification of mt-tRNA Cys GCA ( Fig 3A, 3B & 3E). By contrast, TRIT1(R17E/R21E) led to~6% i 6 A37 modification of mt-tRNA Trp CCA and~10% modification of mt-tRNA-Cys GCA, a 7.5 and 6.5 fold reduction for these mt-tRNAs respectively, but showed high modification (~70%) of cy-tRNA Tyr GUA in the same cells, no reduction compared to TRIT1 ( Fig  3A, 3C and 3E). Thus, the R17E/R21E mutations that impaired TRIT1 MTS-mediated mitochondrial localization in HEK293 cells, also impaired modification of mt-tRNA Trp CCA and mt-tRNA Cys GCA in S. pombe, while they had no effect on modification of cy-tRNA Tyr GUA ( Fig 3C and 3E), the latter of which is evidence that the R17E/R21E mutations do not impair enzymatic activity. Furthermore, as the mtDNA-encoded tRNAs reside in the mitochondrial matrix and are modified in a TRIT1 MTS-dependent manner, these data provide evidence of functional localization of the TRIT1 enzyme to the mitochondrial matrix of S. pombe cells.
Substitution of invariant Thr-12 in the P-loop of Tit1 inactivates it for cy-tRNA Ser UCA TMS [43]. Tit1 T12A is homologous to T19A of E. coli IPTase which resulted in the largest decrease in k cat (600-fold) of all mutations tested [47]. Indeed, the homologous T23 in Mod5 is positioned for catalysis [24]. We show here that as expected, the corresponding TRIT1-T32A mutant was inactive on both the mt-and cy-tRNA substrates (Fig 3A-3E). We also examined a variant isoform, TRIT1(57-467) [61], missing the P-loop entirely as well as D55 (see Fig 1), which in Mod5 acts as a general base for catalysis [24]. This variant represents an isoform from which translation was proposed to start at the second methionine, M57, and could complement cy-tRNA Tyr UUA TMS in S. cerevisiae [61]. Our TRIT1(57-467) construct was inactive on the mt-tRNA and cy-tRNA substrates (Fig 3A-3E).

TRIT1 MTS-dependent i 6 A37 modification of mt-tRNAs in S. cerevisiae
We examined TRIT1 and other IPTases in the mod5Δ, S. cerevisiae strain, MT-8 [34]; ABL8 is a MOD5 replete strain. PHA6 blot analysis for multiple tRNAs are shown in Fig 4, including quantifications in replicate. Fig 4A shows the ACL probings for mt-tRNA Trp UCA and cy-tRNA Trp CCA, with the body probings for these tRNAs in Fig 4B. Results for cy-tRNA Ser AGA, cy-tRNA Tyr GUA and mt-tRNA Tyr GUA are shown in Fig 4C-4G. TRIT1 modified the three cy-tRNAs and both mt-tRNAs in S. cerevisiae whereas TRIT1-R17E/R21E was deficient only for the mt-tRNAs. TRIT1-R17E/R21E-mediated modification of cy-tRNAs but not mt-tRNAs supports the conclusion that the TRIT1 MTS is functional for delivery of TRIT1 IPTase activity to the inner mitochondrial matrix in S. cerevisiae.
Three additional points are relevant. First, substrate-specific restriction against cy-tRNA Trp CCA-A37-A38 by Mod5 relative to the efficient modification of the mt-tRNA Trp U-CA-A37-A38, mt-tRNA Tyr GUA-A37-A38, cy-tRNA Ser AGA, and cy-tRNA Tyr GUA, was demonstrated for the endogenous and ectopic Mod5 enzyme (compare both panels of Fig  4A, and upper panels of C, D, F, and panel G). Second, TRIT1 modified S. cerevisiae cy-tRNA Trp CCA-A37-A38 with efficiency similar to Tit1 and with efficiency similar to Mod5 for the other tRNAs ( Fig 4G). Third, Tit1 showed somewhat low activity on the mt-tRNAs ( Fig  4G). We suspect this may reflect a mitochondrial import mechanism that isn't fully efficient for Tit1 in S. cerevisiae, consistent with the alignment in We also examined the IPTases for cy-tRNA Tyr UUA-mediated TMS in S. cerevisiae which is dependent on efficient i 6 A37 modification. This revealed that while TRIT1, Tit1 and TRT1-R17E/R21E (TRIT1-RR/EE) were active, TRIT1-T32A and TRIT1(57-467) were inactive ( Fig 4H). We note that for TMS, Mod5, Tit1, TRIT1 and TRIT1-TRT1-R17E/R21E were expressed from a low copy plasmid whereas TRT1-T32A and TRT1(57-467) were from a high copy plasmid (Fig 4H).

Complementation of cy-tRNA Ser UCA-mediated suppression in S. pombe
The monitoring of codon-specific insertion of an amino acid in the β-galactosidase active site was used to show that i 6 A37 increases tRNA decoding activity of the two different tRNAs tested by 3 to 4-fold in S. pombe, consistent with i 6 A37 leading to general increased tRNA affinity for ribosome decoding [60]. TMS makes use of a codon-specific reporter to examine TRIT1 mutants for modification activity of cy-tRNA Ser UCA in S. pombe deleted of tit1 + (yNB5 tit1Δ, Fig 5A). Cy-tRNA Ser UCA is dependent on efficient i 6 A37 modification for serine insertion at codon 215 of ade6-704 mRNA [62] with consequent decrease in cellular red pigment during growth in limiting adenine [33]. The pathogenic TRIT1-R323Q variant exhibits substantially reduced modification activity relative to TRIT wild type in the S. pombe TMS assay using the pRep82X expression vector [14]. In this assay, the positive control strain yYH1 (tit1 + ) is pink (moderately suppressed) whereas yNB5(tit1Δ) transformed with empty vector is unsuppressed, red ( Fig 5A, sectors 1, 2). Transformation of yNB5 with pRep-Tit1 under control of the tit1 + promoter, and pRep82X-TRIT1, led to suppression (sectors 3, 4). The catalytic

TRIT1 MTS-dependent modification of mt-tRNAs in S. pombe. A-C)
The PHA6 northern blot assay (Positive Hybridization in Absence of i6A37) uses i 6 A37-sensitive ACL probes. This assay is based on loss of oligo-DNA probe hybridization to the ACL probe due to presence of the bulky i 6 A modification, and is calibrated for quantitation by a body probe to the T stem loop on the same tRNA. The panels show sequential probing (after stripping, Methods) for mt-tRNA Trp CCA, mt-tRNA Cys GCA, and cy-tRNA Tyr GUA with ACL and body probes as indicated; lanes are numbered below panels A. Samples in lanes 1-4 are from tit14 cells (yNB5) transformed with IPTases indicated above the lanes or empty vector (lane 5); lane 6 is from positive control, tit1 + (yYH1) cells. D) shows the small RNA region of the gel after staining with EtBr. E) Quantitation of % i 6 A37 modification of the tRNAs in A-C from biological duplicate, or triplicate PHA6 blots as indicated; error bars indicate standard deviation (SD).

TRIT1 exhibits substrate-specific restriction against modification of S. pombe cy-tRNA Trp CCA
We note that while comparison of different IPTases in a heterologous 'in vivo' system is not physiological, it provides an experimental model with which to gauge activity of an ectopic IPTase for a particular tRNA relative to other tRNA substrates in the cells and compare the overall tRNA substrate activity pattern to that of a different ectopic IPTase in the same system. We did that for Fig 3 by using S. cerevisiae as the heterologous system and in Fig 6 using S. pombe as the heterologous system. Unexpectedly, the cy-tRNA Trp CCA-A37-A38 in S. pombe was a very poor substrate of TRIT1 as compared to cy-tRNA Ser AGA and cy-tRNA Tyr GUA, and Tit1 (Fig 6). Unexpected because TRIT1 efficiently modified the S. cerevisiae cy-tRNA Trp CCA-A37-A38 in S. cerevisiae. Similarly, it was unexpected that Mod5 would modify the S. pombe cy-tRNA Trp CCA-A37-A38 to as high as 60% (Fig 6E) based on its restriction against S. cerevisiae cy-tRNA Trp CCA-A37-A38.
Effects of increasing TRIT1 levels on tRNA modification were also examined. Two additional IPTase constructs were included in this analysis, Mod5-M1A and Mod5MTS-TRIT1, the latter of which was created by replacing the first twenty-eight codons of TRIT1 with the first nineteen codons of MOD5 which includes sequences in its MTS and the second translation initiation site at M12, as schematized in the left and right panels of Fig 6A. In addition to usefulness for assessing mt-tRNA modification, the MOD5 leader sequence also produces increased levels of cytoplasmic isoforms of the IPTases because of the known high favorability of the translation initiation context of the second methionine, M12 relative to MOD5 M1 [68] and TRIT1 M1 (Fig 6A, S1A-S1D Fig). Mod5 and Mod5-M1A were only modestly deficient for cy-tRNA Trp CCA modification, whereas TRIT1 and Mod5MTS-TRIT1 were severely deficient for cy-tRNA Trp CCA modification relative to cy-tRNAs Tyr GUA and Ser AGA (Fig 6B, 6C  & 6E). Hypomodification of cy-tRNA Trp CCA is reflected by strong hybridization signals in lanes 1 and 4 relative to lanes 2 and 6 of the PHA6 ACL probe blots in Fig 6B, upper panel, and by the biological replicate data in Fig 6E. The data indicate that while Mod5 sees the S. pombe cy-tRNA Trp CCA as the least preferred of the cy-tRNA substrates, this is modest compared to the more severe substrate-specific restriction against this tRNA by TRIT1. The cy-tRNA Trp CCA is efficiently modified by Tit1 consistent with previous data [43].

Position 32 influences tRNA Trp CCA substrate-specific restriction by Mod5 and TRIT1 IPTases
Our data indicate that Mod5 and TRIT1 exhibit differential restriction against the cy-tRNAs Trp CCA of S. cerevisiae and S. pombe which differ at C32 vs. U32 as well as other positions more distant to their ACLs (S2 Fig). Given the modification circuit relationship between position 32 and i 6 A37 among tRNAs Ser [12,13] and other potential effects of position 32 on ACL structure (and function [69,70]), we suspected position 32 as a possible determinant of the differential activities of TRIT1 and Mod5 on the S. cerevisiae and S. pombe cy-tRNAs Trp CCA.
To attempt to gain broader perspective we collected tRNA gene sequences encoding the ACLs of cy-tRNAs Trp CCA in representative related yeasts as well as animal cells available in  [101][102][103], the former of which was derived for S. cerevisiae but applies to other yeasts including S. pombe, at the -3 position. B) The i 6 A37-sensitive PHA6 northern blot assay for three tRNAs from S. pombe expressing IPTases from pRep82X plasmids indicated above the lanes. The blot was sequentially probed with the i 6 A37-sensitive ACL probes to the cy-tRNAs indicated to the left of the panels. C) The body probes for each of the tRNAs in B. D) The same blot as in B-C, probed for mt-tRNA Trp CCA using the ACL and body probes indicated. E) Quantitation of % i 6 A37 modification of the cy-tRNAs from biological replicate PHA6 northern blots as in panels B-C; n = 5 for cy-tRNA Trp CCA and all others except n = 1 for cy-tRNA Tyr in Mod5 and Mod5M1A. F) Quantitation; % i 6 A37 modification of mt-tRNA Trp CCA from biological duplicate PHA6 northern blots; error bars indicate standard deviation (SD). an appropriate database [71] and subjected them to multiple sequence alignment (Fig 7A). Three patterns emerged. The related Saccharomyces and Zygoaccharyomyces bailii species, at the top of the alignment, revealed a C32-A37 pattern different from the other fungi. The remaining nineteen fungal species exhibited two different covariations; eight with T32-G37 and eleven with C32-A37 (Fig 7A). The C32-G37 pattern was predominant in the animal cell species, with the exceptions being the primitive Leishmania and Dictyostelium. For comparison, ACL alignments of cy-tRNA Cys GCA was also done which revealed a less clear modal 32-37 covariation among the fungi than for cy-tRNAs Trp CCA (S3 Fig). For all fungal cy-tRNAs-Tyr GUA examined, the C32-A37 pattern was set, whereas the C32-G37 pattern was set for all eukaryotes examined (S3 Fig). The data are consistent with the idea that potential for i 6 A37 modification of cy-tRNAs Trp CCA was relatively variable in yeasts due to A37 vs. G37 identity and became progressively less variable and then fixed by G37 identity in higher eukaryotes. The observed differences in relative modification of S. cerevisiae and S. pombe cy-tRNAs Trp CCA by TRIT1 and Mod5 in the two yeast systems might be due to differences in the tRNA sequences themselves, posttranslational modifications of the IPTases, or subcellular distributions. Considering the wide range of functional transferability of IPTases from multiple distant phylogenetic groups to another (Introduction), we suspected that a basis of the modification differences may be in the cy-tRNAs Trp CCA themselves. We set out to look into this by examining IPTase activity for the S. pombe cy-tRNA Trp CCA carrying its native encoded U32 or with this position mutated to C32, expressed in mod5Δ MT-8 S. cerevisiae. We chose this approach because S. pombe tRNA genes can be readily expressed in S. cerevisiae, but not vice versa [72], and because this tRNA can be distinguished from the S. cerevisiae cy-tRNA Trp CCA [43]. A PHA6 ACL probed blot for the S. pombe cy-tRNA Trp CCA with U32 and C32 is shown in the Right and Left panels of Fig 7B, and the body probe blot is shown in Fig 7C. The ACL probe for the endogenous S. cerevisiae cy-tRNA Trp CCA is shown in Fig 7D, upper panel, with the body probe in the lower panel. Fig 7E shows the lower part of the gel from which the blot was made. Panel F shows quantitation of the PHA6 data for all three tRNAs. The PHA6 analysis of the S. pombe cy-tRNA Trp CCA with U32 and C32 was consistent with Figs 4 and 6 in so far as TRIT1 was less active than Mod5 on the cy-tRNA Trp CCA with its native U32 and more active than Mod5 on cy-tRNA Trp CCA substituted with C32 (Fig 7B and 7C) which better matches the S. cerevisiae cy-tRNA Trp CCA ACL. The data showed that substrate-specific restriction by TRIT1 against S. pombe cy-tRNA Trp CCA could be demonstrated in S. cerevisiae, while maintaining efficient modification of endogenous S. cerevisiae cy-tRNA Trp CCA (Fig 7F), with a trend of general relative activities similar to observed for TRIT1 in the two yeasts. Moreover, the single point mutation, U32C, converted S. pombe cy-tRNA Trp CCA from a poor to better substrate of TRIT1, while Tit1 modified both substrates to comparable efficiencies (Fig 7B and 7F).
However, this experiment also produced unexpected results that raised a new hypothesis. Mod5 was more active on the C32 tRNA Trp than expected, which bears an encoded ACL that matchs that of the endogenous S. cerevisiae cy-tRNA Trp CCA, while it remained unable to modify the latter (Fig 7D and 7F). This suggests that another difference(s) in the cy-tRNAs Trp CCA of the two yeasts may be responsible, for example, G26 vs. U26 respectively (see S2 Fig).

Evidence against a hyper-restriction model for TRIT1
Finally, we designed an experiment to address if a hypothetical hyper-restriction IPTase model may apply to TRIT1 or if the alternative simple model of A36-A37-A38 as the principal positive determinant of activity applies. As noted in the Introduction, the hyper-restriction model would predict that TRIT1 would not modify human cy-tRNA Tyr GUA-G37-A38 even if it had A37 because other negative determinants would prevail. Whereas in the simple model, cy-tRNA Tyr GUA-A37-A38 would be a ready substrate of TRIT1. We therefore made a point mutation, G37A, in a human cy-tRNA Tyr GUA gene and examined it in MT-8 cells. This converted the tRNA Tyr GUA into a substrate that was modified with comparable efficiency by both enzymes (Fig 8). This was as expected of the simple model of A36-A37-A38 recognition in substrates that do not have a YYA anticodon.

Discussion
In this paper, two related topics concerning i 6 A37 modification of cytoplasmic and mitochondrial tRNAs were advanced. First, a newly predicted MTS in the amino-terminal sequence of TRIT1 comprised of a TOM20 receptor binding site followed by an amphipathic helix was shown to be functional by the mutation of two key basic residues in the amphipathic helix. The mutations disabled targeting of TRIT1-GFP to the inner mitochondria of human cells, and impaired TRIT1-mediated i 6 A37 modification of mt-tRNAs in S. pombe and S. cerevisiae without affecting i 6 A37 modification of cy-tRNAs in the cells. Use of these yeast in vivo systems was critical for demonstrating function of the TRIT1 MTS in mt-tRNA modification. Understanding TRIT1 mitochondrial function is important because human TRIT1-deficiency

PLOS GENETICS
Conserved IPTases select different subsets of tRNAs for i6A37 modification leads to i 6 A37-hypomodification of mt-tRNAs and consequent impaired translation of mt-mRNAs that encode respiratory complex components [14].
That TRIT1 can access and modify cy-and mt-tRNAs in two evolutionarily diverged yeasts suggests that no specialized posttranslational modification of TRIT1 is required for function, or if so, it is highly conserved [73]. A model of IPTase action of direct recognition of the tRNAs as observed in cocrystal structures is consistent with this and the ability of phylogenetically divergent species' IPTases to complement tRNA-mediated TMS in yeast (Introduction).
This study also advanced insight into determinants of tRNA-i 6 A37 identity variation. It should be noted that tRNA-i 6 A37 variation in Table 1 is in contrast to that for threonylcarbamoyl-A37 (t 6 A37) which occurs nearly without exception on all tRNAs for ANN codons in all species [74, 75, reviewed in 76].
Although biological relevance of tRNA-i 6 A37 variation is yet unknown, that i 6 A37 is a functionally important modification is clear. I 6 A37 tRNA hypomodification that results from pathogenic mutations in TRIT1 causes debilitating childhood neurodevelopmental disorders [14,15]. Deletion of TRIT1 is embryonic lethal in mice [77] and RNAi-mediated IPTase knockdown in silkworms leads to death [28]. Studies in S. pombe support the idea that changes in i 6 A37 levels can affect translation of specific mRNAs rich in i 6 A37-cognate codons [6,60]. Because i 6 A37 formation is linked to basal metabolism, it is plausible that tRNA-i 6 A37 levels could vary under physiologic conditions and regulate the translation of specific mRNAs [78, see 79]. As i 6 A37 is a unique modification that increases tRNA affinity for the ribosome and specific activity for decoding (Introduction), it may serve its translation systems distinctively and plastically [80]. Full appreciation of i 6 A37 function will require better understanding of individual IPTase-tRNA systems. With the R17E,R21E mutations, researchers can attempt to specifically rescue cy-tRNA i 6 A37 modification in patient-derived cells while maintaining mt-tRNA hypomodification to begin to dissect the potential contribution of the cy-tRNAs to the phenotype.
This study represents comparative analyses of three IPTases and multiple tRNA substrates in two heterologous 'in vivo' model organisms, S. pombe and S. cerevisiae deleted of their endogenous IPTases. Quantitative modification analysis used the PHA6 assay, i 6 A37-sensitive northern blot probing of more than ten individual tRNAs followed by tRNA-specific internal calibration for each tRNA species, five endogenous substrates in S. cerevisiae, four endogenous substrates in S. pombe, and multiple heterologous tRNAs. We showed for the first time, the large quantitative extent to which cy-tRNA Trp CCA is a poor substrate of Mod5 (endogenous and ectopic) in S. cerevisiae, while other substrates, including mt-tRNA Trp UCA, are efficiently modified in the same cells (Fig 4A-4G). IPTase activities were also examined in S. pombe ( Fig  6). A conclusion is that the three divergent IPTases exhibit largely unrestrained i 6 A37 modification activity on a range of tRNAs-A36-A37-A38 with the exception of tRNA Trp CCA which is a least favored substrate for Mod5 and TRIT1, albeit conditional on position 32, whereas Tit1 does not discriminate against this type of substrate. Thus, although restrictive type IPTase discrimination against recognition of certain anticodons does exists in some eukaryotes, this appears to have been limited to cy-tRNA Trp CCA. This discriminatory feature was retained by TRIT1 even though its cy-tRNA Trp CCA contains G37 and is not a substrate for i 6 A modification. Data in Fig 7 suggest that elements beyond the ACL of S. cerevisiae cy-tRNA Trp C-CA-A37-A38 contribute to the remarkable extent to which this tRNA is a specifically poor substrate of its cognate IPTase (Fig 4A and 4G).

The N-terminal MTS of TRIT1
The mechanisms by which Mod5 localizes to mitochondria, nuclei and cytoplasm in S. cerevisiae have been well investigated [34-36, 68, 81, 82]. Mod5 and GRO-1 of the round worm, C. elegans use alternative translation initiation to target to mitochondria or cytoplasm [34,35,44]. Our data indicate that TRIT1 uses a single translation start site to produce mitochondrial, cytoplasmic and nuclear protein. Use of alternative translation initiation start sites is attractive because it could theoretically provide opportunity for regulation of the relative amount of the isoforms. The second methionine of the conserved metazoan IPTase is M57 of human TRIT1, part of a DSMQ motif involved in A37 recognition (Fig 1). A cDNA representing TRIT1(57-467) was isolated from human lung cDNA (referred to as Tr1 in ref [61]). This variant was found to be inactive for TMS in S. pombe and S. cerevisiae when compared to full length TRIT1 (Figs 4 and 5). It should be noted that although TRIT1(57-467)/Tr1 appeared active for TMS relative to other variants Tr2-8, it was not compared to full length TRIT1 [61] as was done here. Based on data here including TRIT1-T32A, we suggest that TRIT1 variant activities should be compared to TRIT1.
Because mt-tRNA i 6 A37 modification is critical to TRIT1 function we wanted to examine its mitochondrial import mechanism. As noted, MitoProt II predicted post-import cleavage at position 47, yet such a protein would lack a P-loop and the invariant T32 whose homologous mutations in TRIT1, Tit1-T12A, and MiaA inactivated them. MitoFates [53] is a newer and improved algorithm that predicted a new N-terminal MTS for TRIT1 but with low probability of cleavage. Prior subcellular fractionation showed TRIT1 localizes to the mitochondrial matrix [14]. Our data confirmed and extended this by point mutagenesis of two residues of the MTS which debilitated localization in human cells as well as i 6 A37 modification of mt-tRNAs, but not cy-tRNAs, in S. pombe and S. cerevisiae. In summary, our data do not support a model in which translation initiation at the second methionine, M57, nor cleavage after mitochondrial import at position 47, can produce efficient tRNA i6A37 modification activity.
Interestingly, it was concluded that the N-terminal 11 amino acids of Mod5 is required as part of a longer MTS that is probably not cleaved after import [34,35]. This is supported by MitoFates analysis that predicts positively charged amphiphilicity in the Mod5 N-terminal region but with a very low probability of cleavage, 0.026. The earlier studies also indicate that Mod5 MTS is inefficient at import, consistent with a functional cytoplasmic pool of N-terminal extended protein [34]. Our data with Mod5MTS-M12A-TRIT1 are consistent with this as it retained significant i 6 A37 modification of cy-tRNAs. Perhaps related, biochemical and subcellular fractionation suggest a similar feature of TRIT1 [14]. Yarham et al. could not exclude the possibility that a significant fraction of cytoplasmic TRIT1 may be associated with the outer mitochondrial membrane [14]. This is consistent with microscopy that showed a substantial amount of the total cytoplasmic TRIT1-GFP and TRIT1(1-51)-GFP appeared to be mitochondria-associated (Fig 2B).
Future experiments may address the possibility that TRIT1 and Mod5 localize on the outer mitochondrial surface. Such localization has recently been noted for other tRNA ACL modification enzymes as well as processing factors for tRNA and other RNAs in yeast and metazoa [reviewed in 2]. Specifically, the enzyme responsible for formation of cyclic t 6 A37 [74] localizes to S. cerevisiae mitochondrial surface [see 2]. Similar IPTase localization might augment tRNA i 6 A37 formation via ready access to DMAPP, a product of the HMG-CoA reductase pathway, as it exits mitochondria, and enhance the link between i 6 A and basal metabolism [78, see 79].

Different TRIT1 modification efficiencies for cy-vs. mt-tRNA Trp CCA-A37-A38 in S. pombe
At different TRIT1 expression levels, mt-tRNA Trp CCA-A37-A38 was modified more efficiently than cy-tRNA Trp CCA-A37-A38 (Fig 6, S1F and S1G Fig). Although the data support that the differences are significant yet do not provide mechanistic resolution, some points can nonetheless be considered. While the encoded ACL nucleotides of these cy-and mt-tRNAs Trp are identical (S1G Fig), the difference in i 6 A37 efficiency might reflect differential modification at another site(s) or some other feature(s). Related to this we note that all eukaryotic cy-tRNAs Trp CCA in the database representing mammals, chicken, plants and S. cerevisiae, contain 2'-O-methyl-cytidine (Cm34) at position 34 [83]. An encoded C34 in mt-tRNAs Trp is found only in plants [83] and Schizosaccharomyces species [84]. The U34 of mt-tRNA Trp UCA in metazoan and many yeasts is hypermodified to decode numerous UGA (Trp) codons in their mt-mRNAs [83,85,86]. It was suggested that the wobble C34 of S. pombe mt-tRNA Trp CCA is modified to facilitate recognition of the UGA (Trp) codon found in mt rps3 [84]. However, the only known modified C that can decode A is lysidine, which has so far been found in tRNAs Ile CUA [87]. We did not detect a difference at C34 of S. pombe cy-and mt-tRNAs Trp CCA by primer extension using low (0.01 mM) dNTPs [88].

A Mod5-type restriction against CCA anticodon recognition appears inherent to TRIT1
Biological relevance of the exceptional absence of i 6 A37 modification of S. cerevisiae tRNA Trp CCA-A37-A38 had been enigmatic. Because human mt-tRNA Trp UCA is one of several mt-tRNAs with i 6 A37, it might have been unexpected that TRIT1 would exhibit strong substrate-specific restriction against S. pombe tRNA Trp CCA-A37-A38 (Fig 6) [43]. This is important because it is evidence that a restrictive type IPTase is not a peculiarity of Mod5. Further, it indicates that restriction toward a tRNA Trp CCA ACL is an IPTase characteristic in extant higher eukaryotes even though their cy-tRNAs Trp CCA contain G37. Yet, additional findings suggest that TRIT1 exhibits a restrictive type recognition that is conditional on features of the YYA ASL that is distinct from Mod5 type recognition. This raises possibilities for future studies of potentially new modes of ASL recognition.

Insight into tRNA-i 6 A37 identity profiles of eukaryotes
Phylogenetic analysis of gene sequences encoding cy-tRNAs for UNN codons other than Ser in representative advanced animal species shows that G is uniformly present at position 37 as noted for Table 1. By contrast, primitive animals, Dictyostelium discoidium and Strongylocentrotus purpuratus, as well as yeasts S. pombe, S. cerevisiae, Aspergillus and Candida subspecies, have G or A at positions 37 of tRNAs Trp and tRNAs Cys [71,89], also noted for Table 1 We set an experiment to compare the IPTases from three divergent species, on S. pombe cy-tRNA Trp CCA-A37-A38 substrates that differ at one encoded nucleotide, C32 vs. U32, one of the differences between the S. cerevisiae and S. pombe cy-tRNAs Trp CCA-A37-A38. When the S. pombe cy-tRNA Trp CCA with either C32 or U32 was expressed in S. cerevisiae the results were consistent with our prior experiments in which the IPTases were compared in each of the two yeasts. The substrate-specific restriction observed for TRIT1 against S. pombe cy-tRNA Trp CCA -U32, was demonstrated in S. cerevisiae while this IPTase maintained efficient modification of the endogenous S. cerevisiae cy-tRNA Trp CCA. Moreover, the point mutation, U32C, converted the S. pombe cy-tRNA Trp CCA from a poor to a better substrate for TRIT1, while Tit1 modified both substrates to comparable efficiencies. The results also raised a new hypothesis, that another difference in the cy-tRNAs Trp CCA of the two yeasts, G26, may be responsible for the higher activity of Mod5 for the S. pombe cy-tRNA Trp CCA -C32, while remaining relatively inactive for the endogenous cy-tRNA Trp CCA which encodes U26 (S2 Fig,  Fig 7B-7D). The hypothesis that modification of G26 to m 2 2 G26 may promote i 6 A37 modification of certain tRNAs, including S. pombe cy-tRNA Trp CCA which carries both modifications [90] is plausible because m 2 2 G26 can influence ASL structure [91]. This is a testable hypothesis fit for future study.
We note that our findings also have additional implications. Human mt-tRNA Trp U-CA-A37-A38 carries U32 and U34. This and other observations suggest the possibility that the cmnm 5 U34 and tm 5 U34 modifications on S. cerevisiae and human mt-tRNAs Trp UCA [4,83,92] enhance activities of their substrate tRNAs for the respective IPTases. It is tempting to speculate that bulky modified uridines would also enhance other YYA AC IPTase substrates, the suppressors tRNA Tyr UUA, tRNA Ser UCA, and the tRNA Ser[Sec] UCA, which contain mcm 5 U (or hypermodified versions) [93][94][95][96].

Materials and methods
S. pombe and S. cerevisiae yeast transformations were by standard methods. All strains were published previously as noted. MT-8 (mod5Δ) lacks codons 1-250 of Mod5 and 35 bp upstream [34].

DNA constructs
The different TRIT1 mutants and wild-type for mammalian transfection were cloned in the HindIII and AgeI sites of the pEGFP-N1 vector. A linker of 5 glycine residues was introduced between the TRIT1 sequences and GFP.
All primers and oligos were obtained from Eurofins Genomics (Tables 2 and 3). For S. pombe, Tit1 and Tit1-T12A, both with a C-terminal HA tag were expressed from the tit1 + promoter in a pRep vector in which the nmt1 + promoter was excised [43]. TRIT1 and Mod5 were expressed from the nmt1 + promoters in pRep82X or 4X, cloned such that their ATG immediately followed the Xho I site. Mutants were made using site-directed mutagenesis or PCR. The Mod5MTS-M12A-TRIT1 construct was generated by site directed mutagenesis using the Q5 Site-Directed Mutagenesis kit (New England Biolabs). All constructs were sequence verified. Cloning of IPTases in S. cerevisiae expression plasmids. TRIT1, Tit1 and Mod5 ORF nucleotide sequences were PCR amplified using gene specific forward and reverse primers containing HindIII/Xho1 restriction sites. For Tit1 and Mod5, an existing ATG in pYX141 was disrupted prior to insertion of the ORF; pYX141 (low copy, 786 promoter) was digested with EcoRI and BamHI, followed by treatment with Klenow fragment and ligation. Sequencing confirmed disruption of the ATG. The HindIII/Xho1 restriction sites were then used for ORF insertion. For TRIT1, TRIT1-R17E/R21E, TRIT1-T32A, and TRIT1(57-467), the sequences were PCR amplified from the corresponding pREPx plasmid constructs using primers that contained Nco1/Xho1 sites, and these were cloned into the corresponding sites of pYX141 or pYX242 (high copy, TP1 promoter).

Cell culture and transfection
HEK293 cells were maintained in DMEM plus Glutamax (Gibco) supplemented with 10% heat-inactivated FBS in humidified 37˚C, 5% CO 2 incubator. The cells were transfected using Lipofectamine 2000 according to manufacturer's instructions using 2.5 ug plasmid DNA and 6.25 ul lipofectamine on cells that were 80% confluent in 6-well plates. The next day, cells were seeded onto coverslips for confocal microscopy (see below), or divided over multiple 10 cm plates for mitochondrial preparation (see below) and western blot analysis.

Mitochondrial preparation and cell subfractionation
Subcellular fractions were isolated as described previously [98], with a few modifications. Forty-eight hours post transfection, HEK293 cells (8 x 10 cm plates at 80% confluency per condition) were washed with 10 ml warm PBS per plate and scraped in 5 ml ice-cold PBS per plate and combined in a 50 ml Falcon tube. Cells were spun down for 3 mins at 1200 rpm and the pellets were washed twice in 10 ml ice-cold PBS. An aliquot of cell pellet was kept for isolation of total protein in RIPA buffer (Pierce) containing protease inhibitors (Roche). Pellets were suspended in 2 ml of homogenization buffer (HB, 0.6 M Mannitol, 1 mM EGTA, 10 mM Tris-HCl pH 7.4) +0.1% BSA) and transferred to a hand-held 10 ml glass-Teflon homogenizer. After 25 strokes, the suspension was transferred to a 2 ml Eppendorf tube and centrifuged at 400g for 10 min at 4C. The supernatant, containing crude mitochondria was kept. The pellet was resuspended in 2 ml HB and homogenization was repeated using 15 strokes. Suspension was transferred to a 2 ml Eppendorf tube and centrifuged at 400g for 10 min at 4C. Again, the supernatant, containing crude mitochondria was kept. This process was repeated one more time with 15 strokes. All three supernatants were centrifuged again at 1000g for 10 mins at 4C. Supernatants were transferred into fresh Eppendorf tubes without disturbing the pellets and spun at 11,000g for 10 mins at 4C to pellet the mitochondria. The supernatant of the first homogenization was kept as the soluble cytoplasmic fraction. Mitochondria were washed twice with 1ml HB without BSA and PMSF and combined into one tube and centrifuged again at 11,000g for 5 mins at 4C to generate a single pellet. This was taken up in 500 ul of HB without BSA and PMSF and split into 2 tubes (200 ul/tube). One tube was put on ice (ProtK fraction) and the other one (mitochondrial fraction) was spun down and the pellet taken up in 40 ul RIPA buffer (Pierce) containing protease inhibitors (Roche). Protein concentrations of the total protein-and mitochondrial fractions were determined using BCA (Pierce). 40 ug of the protK fraction in a total of 190 ul HB without BSA and PMSF was treated with 1.6 ug proteinase K (Viagen) by incubating on ice for 15 mins and subsequently at room temperature for 15 mins. 5 mM PMSF was added to inactivate the Proteinase K and the mitochondria were spun down at 11,000g for 5 mins at 4C. Mitochondria were washed with 1 ml HB including PMSF, but without BSA and taken up in 40 ul RIPA buffer (Pierce) containing protease inhibitors (Roche). For the cytosolic fraction, equal protein concentration between wild-type and mutant was determined using nanodrop using HB as blank.17.5 ug of each fraction was separated on SDS-PAGE and analyzed on western blot.

Western blots
Total protein was extracted from transfected HEK293 cells lysed in RIPA buffer (Pierce) containing protease inhibitors (Roche) 48 hours post transfection and size separated using SDS-PAGE. Proteins were then transferred to a nitrocellulose membrane. Antibodies used were anti-GFP (Santa Cruz, sc-8334), anti-Tom20 (Santa Cruz, sc-11415), anti-NDUFA9 (Abcam, ab14713) and polyclonal human anti-La (Go) was a gift of J. D. Keene and D. J. Kenan (Duke Medical Center) [99]. Anti-TRIT1 Ab was generated in rabbits using the peptide KLHPHDKRKVARSLQVFEE as an antigen and was affinity purified using the same peptide (Pierce protein research). Primary antibodies were detected by secondary antibodies from LI-COR Biosciences, which were conjugated to either IRDye 800CW or 680RD and western blot were scanned and quantified using the Odyssey CLx imaging system (LI-COR Biosciences). Western blots for S. pombe protein: For protein isolation from S. pombe, cells were transferred to a 1.5 mL microcentrifuge tube, an equal volume of 0.7 N NaOH solution was added, mixed, then incubated at room temperature for 3 minutes. The tube was centrifuged at 5,000 rpm for one min and the supernatant discarded. SDS-PAGE sample buffer was added, mixed well and the sample was heated at 95˚C for 5 min centrifuged again and the resulting supernatant recovered for analysis. Proteins were then transferred to a nitrocellulose membrane and subsequently analyzed. Anti-β-actin was from Abcam (mAbcam 8224) and used at a 1:5000 dilution. Blots were scanned as described above using the Odyssey CLx imaging system (LI-COR Biosciences).

Confocal microscopy
24 hrs. after transfection cells were seeded onto coverslips. The next day the live cells were treated with MitoTracker (ThermoFisher) at 200 nM in PBS for 20 min. at 37˚C. Cells were then fixed with 4% paraformaldehyde for 15 min. at room temperature and washed with PBS. PBS containing 1 ug/ml Hoechst 33342 (Sigma) was added and allowed to sit for 10 min, followed by a final PBS wash. Imaging was with a Leica DM IRE2 confocal microscope using a HCX PL APO CS 63.0x1.32 oil UV objective. Images were scanned in sequential mode in Hoechst, GFP and MitoTracker channels.

PHA6 northern blotting
We recently found that increasing the post-hybridization wash temperature, to as high as 15˚C above the Ti (hybridization incubation temperature, below) increased the difference in signal between IPTase-positive and IPTase-negative sample tRNAs, especially, mt-tRNAs. We note that the optimal wash temperature for ACL probes varies among individual tRNAs possibly due to other ACL modifications and other characteristics. We found that subjecting each tRNA-ACL probing to sequential washes at progressively higher temperatures yielded more reliable and higher apparent % modification efficiencies than subjecting all tRNA-ACL probe combinations to a final wash at the same temperature. After the maximum difference in IPTase-positive and -negative sample tRNAs is recorded, the next high temperature signal from all samples may substantially or drastically decrease. The results reported and quantified in this paper are from blot washings that reveal the greatest difference in ACL probe signal between IPTase positive and IPTase negative sample tRNAs that were obtained using this approach. After the maximum difference in IPTase-positive and -negative sample tRNAs occurs, the next high temperature wash may drastically decrease signal from all samples and may not be useful for quantification.
Novex 10% TBE-urea polyacrylamide gels and transferred to GeneScreen plus nylon membranes using Invitrogen iBlot transfer apparatus. Blots are UV crosslinked then baked in a vacuum oven at 80˚C for 1-2 hrs. The blot is incubated in 10-15 ml of PHA6 hybridization buffer contains 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) plus 0.2% SDS, 1× Denhardt's solution [Life Technologies], 100 μg/ml sheared salmon sperm DNA (supplier) at the oligonucleotide DNA probe hybridization temperature for 1 to 2 h. A 32 P-end-labeled antisense DNA oligonucleotide at >10 8 cpm/μg DNA was added to the hybridization buffer (final concentration, 10 5 cpm/ml) and left overnight for hybridization at the Ti. The Ti is calculated as follows: Ti = T m -15˚C where T m is equal to 16.6 log[M] + 0.41[P gc ] + 81.5 − P m − (B/L) − 0.65, where M is the molar salt concentration, P gc is the percent G+C content in the oligonucleotide DNA probe, P m is the percentage of mismatched bases, if any, B is 675, and L is the oligonucleotide DNA probe length [100]. Blots were washed with 2× SSC, 0.2% SDS two times at room temperature (10 min each) followed by 30 min at Ti +5˚C, +7.5˚C, +10˚C, +12˚C or +15˚C (determined empirically, described above). Optimal wash temperatures for different tRNAs has ranged from Ti to T1+15˚C.
Before the next hybridization, blots were stripped with 0.1 × SSC, 0.1% SDS at 85-90˚C, and the nearly complete removal (�90%) of 32 P was confirmed by Phosphor Imager analysis. In general, multiple ACL probings are done followed by the body probes. Table 2 provides sequences of the oligo-DNAs used as ACL and body probes.
The formula used to calculate % i 6 A37 modification is [1-(ACLtit1-Δ (transformed with IPTase constructs)/BPtit1-Δ (transformed IPTase constructs))/(ACLtit1Δ (empty vector)/ BPtit1Δ (empty vector))] x 100, where % modification in yNB5 (tit1-Δ) = 0, "ACL" indicates the ACL probe and "BP" indicates the body probe. In cases where MT-8 cells were used rather than yNB5, mod5Δ would substitute for tit1Δ. As described, the formula is internally normalized by the tit1Δ or mod5Δ samples.  [103]. Lines 3 and 4 are for the alternative ATGs that encode MOD5 M1 and M12 in their native context; lines 5 and 6 are for the MOD5 M1 and TRIT1 M1 after cloning into the Xho1 site of S. pombe expression vectors pRep4X or pRep82X (pRepXho1). Line 7 shows the native TRIT1 M1 ATG context. B) Nucleotide sequence of MOD5 including its M1 and M12 ATG sites and the M12A substitution mutation in the Mod5MTS-M12A-TRIT1 construct. C) Western blot of TRIT1 protein developed using anti-TRIT1 Ab, from cells with constructs containing the strong (pRep4X) or weak (pRep82X) nmt1 + promoter (see text), indicated above the lanes as 4X or 82X. Lanes are numbered below. MW markers are indicated in kDa. The shared band below 25 kDa is an internal control. Lower panel shows β-actin as loading control used for quantitative normalization. D) Determination of TRIT1 levels in C by quantitative Odyssey CLx imaging (Methods); numbers above bars indicate TRIT1/β-actin levels in each sample; The pRep vectors used are indicated as 4X and 82X are indicated along the X-axis. E) Northern blot of 2 cy-and 2 mt-tRNAs by TRIT1 and Mod5MTS-TRIT1 each from the strong promoter, pRep4X and weak promoter pRep82X, as indicated above the lanes, as 4X and 82X. The top four panels show the ACL probings as indicated to the left, and the bottom four panels show the corresponding body probings. F) Quantitation of % i 6 A37 modification of the mt-tRNAs and the cy-tRNAs; the pRep vectors used are indicated as 4X and 82X along the X-axis. G) Clover leaf representations of S. pombe cy-tRNA Trp CCA and mt-tRNA Trp CCA as encoded by the nuclear and mitochondrial DNA and folded by tRNAscan-SE [112] (Table 1