Evolving specificity of tRNA 3-methyl-cytidine-32 (m3C32) modification: a subset of tRNAsSer requires N6-isopentenylation of A37.

Post-transcriptional modifications of anticodon loop (ACL) nucleotides impact tRNA structure, affinity for the ribosome, and decoding activity, and these activities can be fine-tuned by interactions between nucleobases on either side of the anticodon. A recently discovered ACL modification circuit involving positions 32, 34, and 37 is disrupted by a human disease-associated mutation to the gene encoding a tRNA modification enzyme. We used tRNA-HydroSeq (-HySeq) to examine (3)methyl-cytidine-32 (m(3)C32), which is found in yeast only in the ACLs of tRNAs(Ser) and tRNAs(Thr) In contrast to that reported for Saccharomyces cerevisiae in which all m(3)C32 depends on a single gene, TRM140, the m(3)C32 of tRNAs(Ser) and tRNAs(Thr) of the fission yeast S. pombe, are each dependent on one of two related genes, trm140(+) and trm141(+), homologs of which are found in higher eukaryotes. Interestingly, mammals and other vertebrates contain a third homolog and also contain m(3)C at new sites, positions 32 on tRNAs(Arg) and C47:3 in the variable arm of tRNAs(Ser) More significantly, by examining S. pombe mutants deficient for other modifications, we found that m(3)C32 on the three tRNAs(Ser) that contain anticodon base A36, requires N(6)-isopentenyl modification of A37 (i(6)A37). This new C32-A37 ACL circuitry indicates that i(6)A37 is a pre- or corequisite for m(3)C32 on these tRNAs. Examination of the tRNA database suggests that such circuitry may be more expansive than observed here. The results emphasize two contemporary themes, that tRNA modifications are interconnected, and that some specific modifications on tRNAs of the same anticodon identity are species-specific.

INTRODUCTION

Nascent transcripts from tRNA genes undergo extensive post-transcriptional processing that includes nucleolytic cleavage, splicing, 3′ CCA addition, and chemical modifications of numerous nucleotides which play important roles in tRNA function (Phizicky and Hopper 2010). While modifications are observed throughout the tRNA, two nucleotides in the anticodon loop (ACL), at positions 34 (wobble) and 37 (the extended anticodon Yarus 1982), are the most variably and extensively modified as these contribute to decoding precision, efficiency, and reading frame maintenance during translation (Phizicky and Hopper 2010). Deamination of A34 to inosine expands codon recognition capability, especially in eukaryotes (Torres et al. 2014b). mcm5U34 enhances decoding and is used to regulate translation during stress conditions (Begley et al. 2007). Modifications of A37 to N-isopentenyl-A37 (i6A37) and N-threonylcarbamoyl-A37 (t6A37) enhance tRNA activity (Lamichhane et al. 2011, 2013; Thiaville et al. 2015a). In some bacteria, yeasts, protists, and plants, t6A37 exists mostly as a cyclized ester with an oxazolone ring requiring the activity of the TCD enzymes (Miyauchi et al. 2013). Although several tRNA modification enzymes are nonessential in yeast, mutations in the human gene homologs are associated with diseases including cancer and neurodegeneration (Torres et al. 2014a).

Spatiotemporal order of tRNA processing and modification is dictated by subcellular location (Phizicky and Hopper 2010; Hopper 2013). The isopentenyltransferases (IPTase) responsible for i6A37 recognize single stranded, A36–A37–A38 in the context of a stem–loop motif (Motorin et al. 1997; Soderberg and Poulter 2000). However, some i6A37-tRNAs are derived from precursor-tRNAs with introns inserted between A37 and A38 (Marck and Grosjean 2002); for this subset, modification must follow nuclear export because in budding and fission yeast, tRNA splicing occurs in the cytoplasm (Intine et al. 2002; Yoshihisa et al. 2003, 2007; Hopper et al. 2010; Cherkasova et al. 2011). Also, retrograde transport of spliced tRNAPhe back into the nucleus is required to complete the G37 modification known as wybutosine (yW) (Shaheen and Hopper 2005; Hopper and Shaheen 2008; Ohira and Suzuki 2011).

Syntheses of some tRNA anticodon loop modifications are interdependent. Paabo and co-workers showed that Q formation in the anticodon of marsupial mitochondrial tRNAAsp occurs only after C to U editing elsewhere in the anticodon (Morl et al. 1995). In a cytosolic tRNAThr in Trypanosoma brucei, C to U editing at position 32 stimulates the efficiency of A to I editing at 34 (Rubio et al. 2006). In Thermus thermophilus, 7methyl-G46 (m7G46) positively affects Gm18 and m1G37 modifications (Tomikawa et al. 2010) whereas U55-pseudouridine negatively affects Gm18, m5s2U54, and m1A58 modifications (Ishida et al. 2011). A three nucleotide circuitry was demonstrated in the ACL of S. cerevisiae tRNAPhe in which yW37 synthesis depends on 2′-O-methylation of C32 and G34, which requires TRM7 (Guy et al. 2012). Genetic dissection of this ACL circuitry provided insight into the basis of a neurologic deficiency caused by heritable mutations in FTSJ1, encoding the human homolog of yeast Trm7 (Guy et al. 2015).

In the ACL, base identities at one position covary with modifications at other positions as a mechanism to fine-tune or compensate for anticodon:codon base-pairing strengths so that all tRNAs achieve uniform binding affinity (Olejniczak et al. 2005; Ledoux and Uhlenbeck 2008). Identity and modification of the extended anticodon position 37 is tuned to the anticodon base identity at position 36 which pairs with the first nucleotide of the codon (Yarus 1982). With few exceptions, the first ACL nucleotide, position 32, is a pyrimidine, and 33 is almost always U (Fig. 1A; Olejniczak et al. 2005; Ledoux et al. 2009). Pyrimidines at 32 make a noncanonical hydrogen bond with the base at 38 (Auffinger and Westhof 1999), the presence or absence of which in different tRNA contexts help fine-tune their affinity for the ribosome (Olejniczak et al. 2005). Known modifications at position 32 of eukaryotic tRNAs are pseudouridine, N-C methylation (m3C), or 2′-O-methylation (Cm) (Phizicky and Hopper 2010). m3C32 has been observed only on tRNAsSer and tRNAsThr in yeast and these plus tRNAsArg in higher eukaryotes. Trm140 is responsible for m3C32 in S. cerevisiae (D'Silva et al. 2011; Noma et al. 2011). While a trm140Δ mutant exhibits no significant phenotype, a trm140Δ trm1Δ double mutant (also lacking tRNA m22G26 modification) is sensitive to cycloheximide, suggesting impaired translocation on the ribosome (D'Silva et al. 2011).

FIGURE 1.

The S. pombe tRNAs for serine and threonine have m3C32 modification. (A) A cartoon of schematized tRNA anticodon loop with different nucleotide positions indicated as referred to in the text. (B) The tRNASerUGA and tRNAThrUGU track profiles from IGV display software showing misincorporations as colored bars (Arimbasseri et al. 2015). IGV introduces color by default if ≥15% mismatch is detected relative to the reference gene sequence. Color key: green, A; red, T; orange, G; blue, C. Sequence read counts are indicated on the y-axis. (C) Plot showing C32 misincorporation levels for all S. pombe tRNAs that have C at position 32. Error bars indicate standard deviation of four replicates. The tRNAsSer and tRNAsThr are further grouped according to base identity at position 36; note that three of the four tRNAsSer contain A36 while the other tRNASer contains U36 as do all of the tRNAsThr. (D) Heat map showing fractions of each nucleotide misincorporated at C32 for the seven tRNA anticodons to the right.

We recently developed tRNA-HySeq, which detects certain base modifications as nucleotide misincorporations (Arimbasseri et al. 2015). By this approach, all tRNAsSer and tRNAsThr in S. pombe indicate m3C32, which we confirmed here by deletion of the methyltransferase genes responsible. We report that unlike S. cerevisiae, in which a single gene, TRM140, is responsible for modifying both tRNAsThr and tRNAsSer, in S. pombe, two closely related genes, trm140+ and trm141+, are specific for tRNAsThr and tRNAsSer, respectively. A third homolog in mammals, METTL8, is correlated with emergence of m3C32 in tRNAsArg. More significantly, analysis of other S. pombe tRNA modification mutants uncovered a new ACL m3C32 circuitry. All three tRNAsSer with A at position 36 and i6A37, show no m3C32 in tRNA isopentenyltransferase-1 mutants (tit1Δ) which lack i6A37. Inspection of the tRNA modification database suggests that this circuitry may occur widely in nature and possibly extend to other A37 modifications.

RESULTS

tRNA-HySeq misincorporations indicate m3C32 in all tRNAsSer and tRNAsThr

Deep sequencing of full length tRNAs can be hindered by structures that impede adapter ligation as well as various base modifications that limit processivity of reverse transcriptase. Fragmentation of purified tRNAs by limited hydrolysis prior to adapter ligation can generate high yield tRNA sequencing by tRNA-HySeq (Arimbasseri et al. 2015). By this and similar approaches, modifications that alter base-pairing are detected as misincorporations (Arimbasseri et al. 2015 and references therein). Figure 1B shows tRNA-HySeq output data for tRNASerUGA and tRNAThrUGU using IGV display in which gray bars indicate match to genomic sequence and colored bars indicate mismatch at ≥15% (default setting of IGV program) (Arimbasseri et al. 2015). We note that these two tRNAs share misincorporations at C32 reflecting their common m3C32 modification but differ at other modification sites (Fig. 1B). When combined with deletion of a suspected modification enzyme gene, tRNA-HySeq can verify that a misincorporation is indeed due to the modification of interest (Arimbasseri et al. 2015).

Figure 1C shows the misincorporation levels for S. pombe tRNAs with C at position 32 which ranged from 11–65% for tRNAsSer and tRNAsThr but ≤1% for the 15 others (Fig. 1C, also see Supplemental Table 1). The known C32 modifications of eukaryotic tRNAs are 2′-O-ribose methylation (Cm) and N3-methylation (m3C) (Phizicky and Hopper 2010). Since m3C is predicted to interfere with Watson–Crick base-pairing and eukaryotic tRNAsSer and tRNAsThr were known to contain m3C32 (Jühling et al. 2009), it was likely that these misincorporations reflected m3C. Consistent with this, tRNAPheGAA, which is known to contain Cm32 (McCutchan et al. 1978), did not show C32 misincorporation (Fig. 1C; Supplemental Table 1). The majority of C32 misincorporations were to T (Fig. 1D), as also detected by another tRNA-Seq method (Ryvkin et al. 2013).

Two distinctions are notable at this point. Ser is coded for by six codons, four begin with U and two begin with A, whereas Thr is limited to four codons, and all begin with A. Three tRNAs decode the four U-beginning Ser codons and a different tRNASer decodes the A-beginning codon, while three tRNAsThr decode all four of its A-beginning codons. The three tRNAsThr and one tRNASer that decode A-beginning codons and thus share U as their position 36 base, showed higher levels of C32 misincorporations than the three tRNAsSer with A as their position 36 base (Fig. 1C). Second, the three tRNAsSer with A at position 36 contain i6A37, while the tRNASer with U at position 36 contains t6A37, as do the tRNAsThr with U at 36 (Jühling et al. 2009).

S. pombe has two distinct homologs of TRM140

The gene responsible for m3C modification of tRNAsSer and tRNAsThr in S. cerevisiae is TRM140/ABP140 (D'Silva et al. 2011; Noma et al. 2011) which contains two open reading frames that produces a single polypeptide due to a +1 frame shift (Fig. 2A; Farabaugh et al. 2006). The N-terminal region is an actin binding domain (Asakura et al. 1998) that is dispensable for tRNA C32 methylase activity which is mediated by the AdoMet-MTase C-terminal domain (Fig. 2A; D'Silva et al. 2011; Noma et al. 2011). Deletion of TRM140 from S. cerevisiae abolished all detectable m3C in total RNA (Noma et al. 2011).

FIGURE 2.

Two sequence homologs for TRM140 in S. pombe. (A) Cartoon showing the general domain architecture of S. cerevisiae TRM140 and the two S. pombe homologs. Position of +1 frame shift in S. cerevisiae gene is indicated as +1 FS, which corresponds to amino acid 301; amino acid numbering is further indicated in panel B. (B) Alignment of the two S. pombe homologs of TRM140 with the S. cerevisiae protein. Only the C-terminal domain of the S. cerevisiae protein (starting from 301) is shown. Asterisks indicate amino acids D466, D547, and the horizontal line indicates the C-terminal 20 amino acids important for catalytic activity of TRM140 (Noma et al. 2011). (C) Phylogenetic tree of trm140+ and trm141+ homologs in different species generated from an alignment created using MUSCLE in Jalview (Edgar 2004; Waterhouse et al. 2009).

BLASTp analysis of the S. cerevisiae Trm140 amino acid sequence against the S. pombe genome returned two highly significant homologies, SPBC21C3.07c (hereafter, trm140 and SPBC3H7.11 (hereafter, trm141+), neither with an actin binding domain (Fig. 2A), with e-values of 1 × 10−102 and 6 × 10−36, respectively, while the next gene homology value was 4.7. Sequence alignment with the AdoMet-MTase domain of S. cerevisiae Trm140 shows that D466 and D547, which are important for catalytic activity (Noma et al. 2011), were conserved in both of the S. pombe proteins (Fig. 2B). As will be shown in a later section, trm140+ and trm141+ encode functionally distinct tRNA m3C modification activities. We also performed more extensive analyses that led to the phylogenetic tree shown in Figure 2C.

Trm140+ duplication extends to other fission yeasts, worms, flies, and vertebrates

We performed BLAST analysis of a range of species using S. pombe Trm141 and Trm140 amino acid sequences as queries (Supplemental Table 2). Trm141 and Trm140 yielded nearly identical results for several different budding yeasts (Supplemental Table 2) which indicate a single TRM140 gene (Noma et al. 2011). In contrast, all fission yeast and metazoa examined revealed distinct homologs for Trm141 and Trm140 (Supplemental Table 2).

A multiple sequence alignment was generated from extensive BLAST analyses (not shown) and used to generate a phylogenetic tree (Fig. 2C). This suggested three major branches comprised of Trm140 homologs, Trm141 plus metazoan METTL6 homologs, and the METTL2 and METTL8 homologs (Fig. 2C). As shown below, the fission yeast gene duplication yielded functionally distinct tRNA m3C modification activities.

Distinct m3C modification systems for tRNAsSer and tRNAsThr in S. pombe

We deleted trm140+ and trm141+ individually and in combination, and subjected total RNA from the deletion strains to mass spectrometric analysis for m3C and m5C. This revealed that trm140Δ and trm141Δ separately showed reduced m3C/m5C levels (40% and 37%, respectively; Fig. 3B,C) relative to the wild-type (WT) strain (Fig. 3A), whereas deletion of both trm140Δ and trm141Δ abolished all m3C (Fig. 3D). We note that deletion of trm140+ and trm141+ each decreased m3C/m5C levels to <50% of WT levels. This and other observations noted below and in the Discussion suggest some cross-talk between these modification systems.

FIGURE 3.

S. pombe trm140Δ and trm141Δ are each partially deficient in m3C and can be rescued by trm140+ or trm141+. (A–H) Mass chromatograms of m3C and m5C from total RNA from WT (wild type), trm140Δ, trm141Δ and the double mutant, trm140Δ trm141Δ, cells transformed by empty plasmid pRep4X (A–D), or with Rep4X expressing trm140+ or trm141+ (E–H). The y-axes indicate abundance and the numbers associated with the peaks indicate quantities for m3C and m5C, which were used to calculate the m3C/m3C value given in each panel; the value of 0.2091 in the WT cells was set as 100%. (The high levels of both m3C and m5C in G reflect a generally higher concentration of free nucleosides in that digest as indicated by its adenosine content relative to others samples [not shown]). (I) Bar plot showing C32 misincorporations for individual tRNAsSer and tRNAsThr in trm140Δ, trm141Δ, and WT cells. Error bars indicate standard deviation of four replicates for WT and two replicates each for trm140Δ and trm141Δ.

We performed rescue of the single trm140Δ and trm141Δ mutants as well as the trm140Δ trm141Δ double mutant using a pRep4X expression plasmid containing either trm140+ or trm141+ under the control of the nmt1+ promoter (Fig. 3E–H). Expressing trm140+ in the single trm140Δ mutant led to increased m3C/m5C levels relative to trm140Δ transformed with the empty pRep4X plasmid (compare m3C/m5C levels in Fig. 3B,E). Likewise, expressing trm141+ in trm141Δ led to increased m3C/m5C relative to trm141Δ with empty pRep4X (compare m3C/m5C levels in Fig. 3C,F). We note that while the expression plasmids more than doubled the m3C/m5C levels in the single deletion mutants, they did not restore levels to 100% relative to WT, perhaps because this is an ectopic expression system under control of a heterologous promoter. Expressing either trm140+ or trm141+ in the trm140Δ trm141Δ double mutant increased m3C/m5C levels from zero with empty pRep4X (Fig. 3D) to levels comparable to that found in each of the single mutants (compare Fig. 3G,H,D with B,C). The results fit a model in which Trm140 and Trm141 each have a limited capacity for m3C modification. Individual tRNA analysis in the next section revealed the tRNA-specific basis of this.

tRNA-HySeq reveals quantitative information on individual tRNAs including the modification efficiency of several methylated bases (Arimbasseri et al. 2015). For the present study, we used this to focus on m3C in the trm140Δ and trm141Δ strains. As expected, based on results from S. cerevisiae, our analysis of S. pombe revealed that all of the m3C was found in tRNAsThr and tRNAsSer. Deletion of trm140+ abolished C32 misincorporations from the tRNAsThr but not the tRNAsSer, while deletion of trm141+ abolished misincorporations only from tRNAsSer (Fig. 3I). These data indicate that distinct but related modification enzymes are responsible for C32 methylation in S. pombe. Moreover, this occurs in a tRNA-specific manner; trm141+ for tRNAsSer, and trm140+ for the tRNAsThr. Notably, while C32 misincorporations in the principal target tRNAs were eliminated in their respective mutants, misincorporations in the other tRNAs were nonuniformly decreased relative to WT (Fig. 3B), suggesting some cross-talk between the modification systems (see Discussion).

Expansion of m3C in mammals to m3C32 in tRNAArgCCU and tRNAArgUCU, and m3C47:3 in the variable arms of tRNAsSer

We analyzed mouse embryonic fibroblast (MEF) and human A549 cells using tRNA-HySeq. Unlike in yeast, vertebrate gene copies for a tRNA with any given anticodon tend to contain one or more nucleotide differences, referred to as isodecoders (Goodenbour and Pan 2006; Chan and Lowe 2015). In mouse, there are 26 distinct tRNAs of Ser and Thr identity all of which contain C at position 32. Moreover, there are 121 total distinct mouse tRNA sequences with C at position 32, thirty-four of which showed misincorporation at significant levels (Fig. 4A,B). In addition to all of the tRNAsSer and tRNAsThr, we found isodecoders for both tRNAsArg with U at position 36, tRNAArgCCU and tRNAArgUCU, with substantial C32 misincorporations (Fig. 4B), but no misincorporations at C32 for the many tRNAsArg, with G at position 36.

FIGURE 4.

m3C32 modification in mouse and human cells. (A) Heat map of 121 unique mouse embryonic fibroblast (MEF) tRNA sequences with C at position 32, 34 of which have significant, ≥0.1, misincorporation at position 32. (B) Bar plot of the 34 tRNAs from A that have ≥0.1 misincorporation at C32. (C) Bar plot for all unique tRNAs that show significant C32 misincorporation in human A549 cells. (D) Bar plots for all unique tRNAs from MEFs and A549 cells that show significant misincorporation at C47:3. Error bars indicate standard deviation of two replicates each for MEFs and A549.

Similar to MEFs, human A549 cells also have m3C32 on tRNAArgCCU and tRNAArgUCU in addition to all of the tRNAsSer and tRNAsThr (Fig. 4C). While the total number of isodecoders is greater in human as compared to mouse for these tRNA species, the m3C32 was limited to the same tRNA identities as in mice, and no m3C32 on any of the many isodecoders of the three tRNAsArg with G at 36. The cumulative results are in good agreement with data from Bos taurus in which both tRNAArgCCU and tRNAArgUCU were found to have m3C32 (Keith 1984) while neither tRNAArgCCG from B. taurus (Miller et al. 1983) nor tRNAArgACG from mouse showed m3C32 (Jühling et al. 2009).

In S. pombe and other yeast, m3C32 is limited to tRNAsSer and tRNAsThr while we additionally found it on tRNAArgCCU and tRNAArgUCU in mouse and human cells, consistent with data from various higher eukaryotes (Ginsberg et al. 1971; Keith 1984; Capone et al. 1985; Cribbs et al. 1987). tRNA-HySeq of mouse and human tRNAs also found m3C in the variable arm, at position 47:3 of all four tRNAsSer (Fig. 4D) but not other tRNAs, consistent with prior data (Jühling et al. 2009). This revealed that all tRNAsSer with the UCU sequence at positions 47:2-47:3-47:4 show misincorporations at C47:3 in mouse and human cells.

Loss of i6A37 from tRNAsSer causes their loss of m3C32

I6A37 is synthesized by tRNA isopentenyltransferases in bacteria and eukaryotes. In S. pombe, tit1+ synthesizes i6A37 and tit1Δ deletion strains lacking it have been characterized (Lamichhane et al. 2011, 2013). The substrates of Tit1 include the three tRNAsSer with A at position 36 but not the tRNASer with U at position 36 nor the tRNAsThr (Lamichhane et al. 2011). N-isopentenyl-A37 retains an N hydrogen that can donate to base-pairing and does not cause misincorporation, as reported previously (Fig. 5A; Arimbasseri et al. 2015). Comparison of tit1+ and tit1Δ strains revealed loss of C32 misincorporation in tit1Δ for the usual Tit1 tRNASer substrates (compare WT and tit1Δ in Fig. 5A) but not for the non-Tit1 substrates, tRNAThr (compare WT and tit1Δ in Fig. 5B).

FIGURE 5.

Loss of i6A37 from tRNAsSer causes their loss of m3C32. (A,B) IGV display tracks for tRNASerUGA (A) and tRNAThrUGU (B); colors as in Figure 1; note the absence of C32 misincorporations from the tit1Δ cells of A but not B. (C) Bar plot of C32 misincorporations in the mutants indicated. Brackets with asterisks indicate the three Tit1 substrates, all of which have A at position 36 and differ from the four other m3C32-containing tRNAs which contain U at position 36. Error bars indicate standard deviations of four replicates each for WT and tit1Δ, and two replicates each for tit1Δ+tit1-T12A and tit1Δ+tit1+. (D) Mass chromatograms of m3C and m5C in WT and tit1Δ cells; numbers indicate abundances. (E) Pie chart showing the A37 modification status of tRNAs that carry m3C32; compiled from data in the tRNA modification database (UK, unknown) (Jühling et al. 2009).

Compiled data for all seven tRNAs show that loss of m3C32 in tit1Δ was specific to the three tRNAsSer that are usually substrates for i6A37 modification (Fig. 5C, compare WT and tit1Δ). Because some tRNA modifying activities involve multisubunit complexes (e.g., see Guy et al. 2012, 2015) we asked if dependency of m3C32 on tit1+ requires i6A37 modification activity per se or just the presence of the Tit1 protein. For this, we expressed ectopic tit1+ or a previously characterized tit1-T12A point mutant that encodes a catalytically inactive protein (Lamichhane et al. 2011, 2013) in tit1Δ cells and performed tRNA-HySeq. Ectopic tit1+ restored m3C32 on the Tit1 targets while tit1-T12A did not (Fig. 5C). The data strongly suggest that i6A37 itself is required for m3C32 modification by the Trm141 enzyme on the Tit1 substrates. Notably, tRNASerGCU does not bear i6A37 (Lamichhane et al. 2011) and its m3C32 content was not significantly affected by tit1+ deletion (Fig. 5C).

Mass spec analysis of m3C in WT and tit1Δ RNA revealed that total m3C levels in tit1Δ were reduced to about one-third of WT relative to m5C (Fig. 5D), consistent with the results from tRNA-HySeq. Independent analysis by mass spec also confirmed the complete absence of i6A in the tit1Δ cells as expected (not shown).

Finally, we report that deletion of trm140 and trm141 either separately or together had no overt phenotype nor caused loss of tRNA-mediated suppression (not shown). We could find no difference in growth relative to our wild-type S. pombe cells by serial dilution spot assays in rich or minimal media, nor with glycerol as the carbon source or in the presence of rapamycin, in each condition either at 30°C or 37°C (not shown).

DISCUSSION

The major conclusion of this study is that m3C32 modification of the three tRNAsSer with A at position 36, which decode Ser codons that begin with U, is specifically dependent on i6A37 modification. This indicates a new ACL modification circuitry in addition to activities described including 2′-O-ribose methylation of C32 and G34, and modification of m1G37 to yW of tRNAPhe (Guy and Phizicky 2015; Guy et al. 2012, 2015). A second conclusion is that the genes responsible for tRNA m3C32 modification evolved isoacceptor specificity in fission yeast such that trm140+ and trm141+ are principally responsible for tRNAsThr and tRNAsSer, respectively. Phylogenetics suggests that the one- versus two-enzyme system is a distinguishing feature of the lineages to the budding yeast and to the fission yeasts and animals, respectively.

The results emphasize a current theme, that tRNA modifications are interconnected, especially in the ACL. In addition, the results contribute to an appreciation of the species-specificity of tRNA modification patterns, even for tRNAs with the same anticodon identity.

Evolution of C32 methyltransferases

m3C32 is not found on bacterial or archaeal tRNAs and thus appears to be eukaryote-specific (Jühling et al. 2009). A significant aspect of this work is the evolutionary divergence and expansion of C32 methyltransferases with distinct tRNA specificities in fission yeast (and presumably higher eukaryote) lineages as compared to one in budding yeast (Fig. 2C; D'Silva et al. 2011; Noma et al. 2011). The ability of tRNA-HySeq to monitor m3C32 of individual tRNAs combined with gene deletion and mass spec clearly indicates related but distinct m3C32-methyltransferases in S. pombe: trm140+ for tRNAsThr and trm141+ for tRNAsSer. BLAST analyses found both homologs in the three other fission yeast and multiple invertebrates examined, and three homologs in vertebrates. Since mammals have m3C32 on tRNAArgUCU and tRNAArgCCU in addition to tRNAsSer and tRNAsThr, it is tempting to speculate that the third homolog may methylate the tRNAsArg. Primates have four sequence homologs, including two versions of METTL2, A and B. METTL2B appears to be a true homolog since its knockdown reduced m3C32 levels (Noma et al. 2011).

Trm140 and Trm141 may operate as multisubunit complexes with shared and distinct cofactors for tRNAsSer and tRNAsThr. With regard to this, we note that deletion of either trm141+ or trm140+ reduced m3C32 levels in the others’ substrates (Fig. 3B). In particular, tRNAThrAGT m3C32 was unusually and unexpectedly sensitive to trm141-deletion (Fig. 3B). Also, tRNAThrAGT m3C32 levels were unexpectedly lowered by both ectopic tit1+ and tit1-T12A (Fig. 5C). These observations suggest dominant negative effects of ectopic Tit1 when overexpressed from the strong promoter used, consistent with involvement in a multisubunit complex. Finally, our unpublished data show that i6A levels measured by mass spectrometry were significantly increased in trm141Δ and trm140Δ deletion strains relative to wild type, but returned to wild type levels in the trm141Δ trm140Δ double deletion mutant.

A new ACL modification circuit

Another unexpected finding was dependence of the three tRNASer substrates of trm141+, each with A at position 36, on i6A37 in the same tRNAs. A strength of the data is the specificity provided by tRNASerGCU, which is a substrate of trm141+ but naturally does not carry i6A37, and does not lose m3C32 upon tit1+ deletion (Fig. 5C). We can exclude a differential role for introns in this specificity because tRNASerGCU, tRNASerCGA, and tRNASerUGA contain introns while tRNASerAGA does not; from this can also be inferred that m3C32 modification must occur in the cytoplasm or require retrograde transport of spliced tRNA to the nucleus (Hopper and Huang 2015). Specificity provided by tRNASerGCU allows another conclusion. Because tRNASerGCU is a Trm141 substrate but retains m3C32 in tit1Δ (Fig. 5C), we can conclude that i6A37 is not required for Trm141 activity per se.

One possibility is that m3C32 became uniquely dependent on i6A37 in the tRNAsSer that naturally bear i6A37, perhaps associated with the evolutionary divergence of the two TRM140 tRNA substrate specializations reported here. There is support for this but to a limited extent. In vitro modification using T7 RNA polymerase-transcribed tRNAThr as substrate showed that recombinant S. cerevisiae Trm140 could modify C32 indicating no absolute prerequisite for i6A37 (D'Silva et al. 2011; Noma et al. 2011). Other attempts using T7-transcribed tRNASer were unsuccessful whereas T7-transcribed tRNAThr was less efficiently methylated when compared to tRNA substrate purified from Trm140-deficient S. cerevisiae; this indeed led the authors to suggest that another modification may positively affect C32 methylation (Noma et al. 2011). However, we emphasize that these studies were done with S. cerevisiae Trm140.

Another possibility is that S. cerevisiae Trm140 is dependent on either of two bulky A37 modifications, i6A37 or t6A37 or their derivatives, for efficient methyltransferase activity, and that as part of their substrate specificity, S. pombe Trm140 and Trm141 became dependent on either A37 modification. The tRNAsThr and tRNASerGCU share U at position 36, part of the U36–A37–A38 consensus for t6A37 (Morin et al. 1998; Thiaville et al. 2015b), and are known to contain t6A37 (Jühling et al. 2009).

We therefore analyzed existing tRNA modification data (Jühling et al. 2009) and found that virtually all known tRNAs with m3C32 also carry i6A37, t6A37 or their derivatives (Fig. 5E). However, it should be noted that several others that carry t6A37, namely tRNAs for Met, Ile, Asn Lys, and Arg (Thiaville et al. 2015b), do not contain m3C32 even though several have C at position 32 (see Fig. 1C), suggesting that t6A37 alone is not a sufficient signal for m3C32 modification.

Further analysis revealed another finding of potential relevance. While tRNASerGCU shares U36 and t6A37 with tRNAsThr, it is a Trm141 substrate, while the tRNAsThr are Trm140 substrates. A more unique distinction for tRNASerGCU is that of all seven tRNAs with m3C32, it is the only one with C rather than G in anticodon position 35 (see Fig. 1D). In this regard, it matches the higher eukaryote m3C32 expansion substrates, tRNAArgCCU and tRNAArgUCU at positions C35, U36, and t6A37 (Thiaville et al. 2015b).

Finally, we note that the findings reported here may be relevant to a heritable disorder caused by mutation of human tRNA isopentenyl transferase, TRIT1 and associated i6A37 hypomodification of cytoplasmic and mitochondrial tRNA (Yarham et al. 2014; Lamichhane et al. 2016). The data suggest that human i6A37 hypomodification may be accompanied by m3C32 hypomodification. Notably, using yeast, Guy et al. (2015) could dissect the tRNAPhe ACL yW37 modification pathway and discover an allele of FTSJ1, the human homolog of yeast TRM7, that provided unique insight into the circuitry wiring.

MATERIALS AND METHODS

S. pombe strains, growth conditions, and cloning

A list of S. pombe strains used is given in Table 1. Deletion of trm140 and trm141 was executed as described (Longtine et al. 1998). The growth media used were EMM (minimal media) or YES (rich media).

TABLE 1.

Yeast strains used for experimental work in this study

The genes trm140 and trm141 were amplified from the S. pombe chromosomal DNA of the wild-type strain yAS99 and cloned into the Xho1/BamH1 sites of pRep4X, which puts their expression under the control of the nmt1 promoter. The oligo-DNA primers used for amplification cloning were: TRM140-FOR: 5′-AGTCGAGATCctcg-agATGGATACAACACCAGATAATTC-3′; TRM140-REV: 5′-AGATTTGACAggatc-cTCATTTTTGAAACTTGGCTTGTA-3′; TRM141- FOR: 5′-AGTCGAGATCctcgag-ATGAGTGGTATATCAAAATTAAG-3′; TRM141-REV: 5′-AGATTTGACAggatcc-TCACAGCTTTTTCCAAACTCCTTG-3′. The constructs were confirmed by restriction digestion analysis and sequencing.

“Mouse embryonic fibroblasts” and A549 cells were grown in DMEM media (Gibco) with 10% serum at 37°C in 5% CO2. A549 cells were obtained from the Division of Cancer Treatment and Diagnosis, NCI Tumor Repository (https://dtp.cancer.gov/organization/btb/docs/DCTDTumorRepositoryCatalog.pdf). Cells were grown in 150 mm2 flasks to ∼70% confluency before harvesting for RNA isolation.

tRNA isolation

Total RNA from S. pombe was isolated as described previously (Arimbasseri et al. 2015) using hot-phenol. Total RNA from mammalian cells was prepared using Tripure RNA isolation reagent (Roche, Basel).

Mass spectrometry

Total RNAs isolated from WT (wild type), tit1Δ, trm140Δ, trm141Δ, and trm140Δ trm141Δ cells were digested with 2.5 U of Nuclease P1 (Sigma-Aldrich) and 0.2 U alkaline phosphatase (Takara) in 5 mM ammonium acetate pH 5.3 and 20 mM HEPES-KOH pH 7.0 for 3 h at 37°C. Nucleotides were separated by C18 reverse-phase column (GL Science) and directly injected to Agilent 6460 Triple Quadrupole Mass spectrometry. MRM parameters for m3C and m5C were as follows: m3C/m5C: precursor ion, m/z 258, product ion, m/z 126.

tRNA-HySeq

A high-throughput tRNA sequencing library was prepared from gel-purified tRNA as previously described (Arimbasseri et al. 2015) and modified from Karaca et al. (2014). tRNA was excised after electrophoretic resolution on 6% urea–polyacrylamide gels, and subjected to the crush and soak extraction method in 0.3 M NaCl at 4°C. The eluate was filtered and RNA was precipitated using ethanol. In short, purified tRNA was subjected to partial alkaline hydrolysis in 10 mM bicarbonate pH 9.7 at 90°C for 5 min. After quenching on ice for 2 min, the RNA fragments were subjected to dephosphorylation to remove 3′ phosphates with calf intestinal phosphatase (CIP; NEB). The 5′ ends were then phosphorylated using T4 polynucleotide kinase (T4 PNK; NEB). To the 5′ phosphorylated fragments, preadenylated, 3′ blocked 3′ adapters were ligated using the truncated T4 RNA ligase known as T4 RNL2K227Q (NEB). The ligated fragments were gel purified from 15% polyacrylamide–urea gel followed by ligation of 5′ adapters using T4 RNA ligase I (Thermo Scientific). Ligated fragments were gel purified and subjected to reverse transcription using SSIII RT (Life Technologies) followed by PCR amplification of the library and sequencing using an Illumina HiSeq machine.

Analysis of sequence data

Reads were mapped to a reference library comprised of all predicted tRNA genes in S. pombe, mouse or human. Misincorporations were calculated from the alignment files for the positions studied (Arimbasseri et al. 2015).

Phylogenetics

To identify C32 methyltransferases in fission yeast, BLASTp analysis of S. cerevisiae Trm140 protein sequence was queried against the S. pombe genome using the default settings at NCBI. For trm140+ and trm141+, we analyzed genomes of budding (ncbi taxid:4892) and fission yeast (Schizosaccharomyces; taxid: 4895), Caenorhabditis elegans (taxid: 6239), Drosophila melanogaster (taxid:7227), Danio rerio (taxid:7955), Xenopus laevis (taxid:8355), Gallus gallus (taxid:9031), Mus musculus, chimp (taxid:9598), and human (taxid:9606) using NCBI BLAST with Trm140 or Trm141 protein sequence as query with default settings. Sequences were aligned using MUSCLE (Edgar 2004) and the output was used to create a phylogenetic tree using Jalview software (average distance using BLOSUM62) (Waterhouse et al. 2009).

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.