Hypomodified tRNA in evolutionarily distant yeasts can trigger rapid tRNA decay to activate the general amino acid control response, but with different consequences

All tRNAs are extensively modified, and modification deficiency often results in growth defects in the budding yeast Saccharomyces cerevisiae and neurological or other disorders in humans. In S. cerevisiae, lack of any of several tRNA body modifications results in rapid tRNA decay (RTD) of certain mature tRNAs by the 5’-3’ exonucleases Rat1 and Xrn1. As tRNA quality control decay mechanisms are not extensively studied in other eukaryotes, we studied trm8Δ mutants in the evolutionarily distant fission yeast Schizosaccharomyces pombe, which lack 7-methylguanosine at G46 (m7G46) of their tRNAs. We report here that S. pombe trm8Δ mutants are temperature sensitive primarily due to decay of tRNATyr(GUA) and that spontaneous mutations in the RAT1 ortholog dhp1+ restored temperature resistance and prevented tRNA decay, demonstrating conservation of the RTD pathway. We also report for the first time evidence linking the RTD and the general amino acid control (GAAC) pathways, which we show in both S. pombe and S. cerevisiae. In S. pombe trm8Δ mutants, spontaneous GAAC mutations restored temperature resistance and tRNA levels, and the trm8Δ temperature sensitivity was precisely linked to GAAC activation due to tRNATyr(GUA) decay. Similarly, in the well-studied S. cerevisiae trm8Δ trm4Δ RTD mutant, temperature sensitivity was closely linked to GAAC activation due to tRNAVal(AAC) decay; however, in S. cerevisiae, GAAC mutations increased tRNA loss and exacerbated temperature sensitivity. A similar exacerbated growth defect occurred upon GAAC mutation in S. cerevisiae trm8Δ and other single modification mutants that triggered RTD. Thus, these results demonstrate a conserved GAAC activation coincident with RTD in S. pombe and S. cerevisiae, but an opposite impact of the GAAC response in the two organisms. We speculate that the RTD pathway and its regulation of the GAAC pathway is widely conserved in eukaryotes, extending to other mutants affecting tRNA body modifications.

In S. cerevisiae, lack of any of several tRNA body modifications leads to decay of a subset of the corresponding hypomodified tRNAs, mediated by either of two tRNA quality control pathways, each acting on different hypomodified tRNAs and at different stages of tRNA biogenesis. First, the nuclear surveillance pathway targets pre-tRNA i Met lacking m 1 A, acting through the TRAMP complex and the nuclear exosome to degrade the pre-tRNA from the 3' end [17,[35][36][37]. The nuclear surveillance pathway also targets a large portion of wild type (WT) pre-tRNAs shortly after transcription, ascribed to errors in folding of the nascent tRNA or to mutations arising during transcription [38]. Second, the rapid tRNA decay (RTD) pathway targets a subset of the mature tRNAs lacking m 7 G 46 , m 2,2 G 26 , or ac 4 C 12 , using the 5'-3' exonucleases Rat1 in the nucleus and Xrn1 in the cytoplasm [18,21,22,39,40]. RTD is inhibited by a met22Δ mutation [22,39,41,42] due to accumulation of the Met22 substrate adenosine 3', 5' bisphosphate (pAp) [43,44], which binds the active site of Xrn1 and presumably Rat1 [45]. The RTD pathway also targets fully modified tRNAs with destabilizing mutations in the stems, particularly the acceptor and T-stem, which expose the 5' end [40][41][42]. The hypomodified tRNAs targeted by the RTD pathway also expose the 5' end, ascribed to destabilization of the tertiary fold [40]. There is limited evidence documenting tRNA quality control decay pathways that act on hypomodified tRNAs in other eukaryotes. A mouse embryonic stem cell line with a knockout of METTL1 (ortholog of S. cerevisiae TRM8) had undetectable m 7 G in its tRNA substrates and reduced levels of several METTL1 substrate tRNAs [46]. Similarly, knockdown of METTL1 and NSUN2 (homolog of S. cerevisiae TRM4) in HeLa cells led to reduced levels of tRNA Val(AAC) (abbreviated tV(AAC), as in the Saccharomyces genome database) at 43˚C in the presence of 5-fluorouracil (5-FU) [47], a known inhibitor of pseudouridine synthases and 5-methyluridine methyltransferase [48][49][50]. However, in both of these cases, the underlying mechanism is not known. It was also shown that WT mature tRNAi Met was subject to decay by Xrn1 and Rat1 after 43˚C heat shock in HeLa cells, although there was no change in the modification pattern in vivo or in the stability of the tRNA in vitro caused by this temperature shift [51].
The goal of the work described here is to determine if and to what extent tRNA quality control decay pathways are linked to hypomodified tRNAs in eukaryotes other than S. cerevisiae. To address this issue, we have studied the biology of the tRNA m 7 G 46 methyltransferase Trm8 in the fission yeast Schizosaccharomyces pombe, which diverged from S. cerevisiae~600 million years ago [52].
We chose to study S.pombe Trm8 because S. cerevisiae trm8Δ mutants were known to trigger decay by the RTD pathway. S. cerevisiae Trm8 forms a complex with Trm82 that is required for formation of m 7 G 46 in eukaryotic tRNAs [53,54]. S. cerevisiae trm8Δ and trm82Δ mutants are each modestly temperature sensitive [20], and trm8Δ or trm82Δ mutants also lacking any of several other body modifications had enhanced temperature sensitivity [18]. Moreover, the temperature sensitivity of trm8Δ mutants was suppressed by a met22Δ mutation and was associated with decay of tV(AAC) [22], and the more severe temperature sensitivity of trm8Δ trm4Δ mutants (lacking both m 7 G and m 5 C) was shown explicitly to be due to RTD of tV(AAC) [18,39]. In addition, in mammalian cells, Trm8 biology has other dimensions of complexity. The human TRM82 ortholog WDR4 was associated with reduced tRNA m 7 G modification and a distinct form of microcephalic primordial dwarfism [29]; METTL1 or WDR4 knock out mouse embryonic stem cells showed defects in self renewal and differentiation [46]; and METTL1 was also responsible for m 7 G modification of mammalian miRNAs and mRNAs [55,56]. This evidence emphasizes that Trm8/Trm82 (METTL1/WDR4) and/or its m 7 G modification product is important in S. cerevisiae and mammals, although the reasons are not yet known beyond S. cerevisiae.
We find here that S. pombe trm8Δ mutants have a temperature sensitive growth defect due primarily to decay of tRNA Tyr(GUA) (tY(GUA)) and to some extent tRNA Pro(AGG) (tP(AGG)) by the Rat1 ortholog Dhp1, demonstrating that a major component of the RTD pathway is conserved between S. pombe and S. cerevisiae. We also find an unexpected connection between the RTD pathway and the general amino acid control (GAAC) pathway in both S. pombe and S. cerevisiae. In both S. pombe trm8Δ mutants and S. cerevisiae trm8Δ trm4Δ mutants, the temperature sensitivity coincides with the onset of tRNA decay, which in turn triggers the GAAC activation, presumably due to the increased stress from the tRNA decay. However, in Sp trm8Δ mutants, GAAC activation is deleterious to growth, as mutations in the GAAC pathway restore growth and tRNA levels, whereas in S. cerevisiae trm8Δ trm4Δ mutants, GAAC pathway activation is beneficial, as GAAC mutations exacerbate the growth defect and accelerate tRNA loss. Thus, our results demonstrate a conserved GAAC response associated with tRNA decay by the RTD pathway, but opposite effects on cell physiology in the two organisms. These findings suggest the widespread conservation of the RTD pathway in eukaryotes, and its linkage to the GAAC pathway.

Results
The S. pombe trm8Δ mutants lack m 7 G in tRNAs and are temperature sensitive As Trm8 is the catalytic subunit of the Trm8-Trm82 complex [20], we anticipated that tRNAs from S. pombe trm8Δ mutants would lack m 7 G. We purified tY(GUA) and tF(GAA), which had each been previously shown to have m 7 G 46 [57,58], and then analyzed their nucleosides by HPLC analysis. Purified tY(GUA) from S. pombe trm8Δ mutants had no detectable m 7 G levels (less than 0.03 moles/mole), compared to near stoichiometric levels in tY(GUA) from WT cells (0.93 +/-0.22 moles/mole), whereas levels of each of three other analyzed modifications (pseudouridine (C), m 5 C, and m 1 A) were very similar in trm8Δ and WT cells (Fig 1A). Similarly, purified tF(GAA) from trm8Δ mutants had no detectable levels of m 7 G compared to near stoichiometric levels in WT cells, but otherwise WT levels of C, 2'-O-methylcytidine (Cm) and m 2,2 G ( Fig 1A). These results suggest strongly that S. pombe trm8 + is the methyltransferase responsible for m 7 G formation in cytoplasmic tRNAs.
To understand the biology of S. pombe trm8Δ mutants, we examined the growth phenotypes of two genetically independent trm8Δ mutants. trm8Δ mutants were temperature sensitive starting at 37˚C on rich (YES) and minimal (EMM) media, and expression of P trm8 trm8 + on a plasmid restored WT growth in both media ( Fig 1B). Thus, the temperature sensitivity of trm8Δ mutants was due to lack of trm8 + .

S. pombe trm8Δ mutants have reduced levels of tP(AGG) and tY(GUA) at high temperatures
To determine if the temperature sensitivity of S. pombe trm8Δ mutants was associated with tRNA decay, we analyzed tRNA levels of trm8Δ mutants after an 8 hour temperature shift in YES media from 30˚C to 36.5˚C, 37.5˚C, and 38.5˚C, which progressively inhibited growth (S1 Fig). We measured tRNA levels of all 21 tRNAs in the Genomic tRNA Database [59] that had a 5-nt variable loop with a central guanosine residue (S1 Table), which is the signature for m 7 G modification [60]. We quantified levels of each tRNA at each temperature relative to the levels of that tRNA in WT cells at 30˚C, after normalization of each to the levels of the non-Trm8 substrate tG(GCC) at the corresponding temperature. We used tG(GCC) as the standard because, for unknown reasons, the usual standards 5S and 5.8S RNA each had temperature-dependent reduction in their levels in trm8Δ mutants (S2 Fig), as determined relative to input RNA levels. Note that with tG (GCC) as the standard, the levels of another non-Trm8 substrate, tL(UAA), were also unaffected.
Northern analysis showed that S. pombe trm8Δ mutants had significantly reduced levels of two of the 21 potential Trm8 substrate tRNAs as the temperature was increased. The levels of tP(AGG) were substantially reduced in trm8Δ mutants, from 70% of the levels in WT cells at 30˚C, to 50%, 31%, and 18% after temperature shift to 36.5˚C, 37.5˚C, and 38.5˚C respectively, whereas levels of tP(AGG) in WT cells remained constant as temperature increased (Fig 2A  and 2B). As expected, tP(AGG) is indeed a substrate of Trm8, since purified tP(AGG) from trm8Δ mutants had undetectable levels of m 7 G, but WT levels of each of three other modifications (S3 Fig). The levels of tY(GUA) were also reduced in trm8Δ mutants as temperature increased, albeit to a lesser extent than tP(AGG) levels. Levels of tY(GUA) in trm8Δ mutants were about the same as those in WT cells at 30˚C (119%), remained essentially unchanged at 36.5˚C and 37.5˚C (124%, and 98%), but were reduced to 67% at 38.5˚C, whereas tY(GUA) levels in WT cells were relatively constant at all temperatures. In contrast, none of the 19 other predicted Trm8 substrate tRNAs showed a temperature-dependent reduction in levels in trm8Δ mutants (Figs 2A and 2B and S4 and S5). Levels of 15 tRNAs were approximately constant in trm8Δ mutants as temperature increased, although the initial levels varied somewhat, and levels of the other four tRNAs (tR(CCU), tM e (CAU), tV(CAC), and tK(UUU)) were modestly increased at 38.5˚C. Thus, if the temperature sensitivity of trm8Δ mutants was due to loss of tRNAs, the likely candidates were tP(AGG) and tY(GUA).

The growth defect of S. pombe trm8Δ mutants is primarily due to loss of tY (GUA)
To evaluate the cause of the temperature sensitivity of S. pombe trm8Δ mutants, we analyzed growth after overexpression of tP(AGG) and/or tY(GUA) on leu2 + plasmids ( Fig 2C). S. pombe trm8Δ mutants lack m 7 G and are temperature sensitive. (A) trm8Δ mutants have no detectable m 7 G in their tY(GUA) and tF(GAA). S. pombe trm8Δ mutants and WT cells were grown in biological triplicate in YES media at 30˚C and tRNAs were purified, digested to nucleosides, and analyzed for modifications by HPLC as described in Materials and Methods. The bar chart depicts average moles/mol values of nucleosides with associated standard deviation; WT, green; S. pombe (Sp) trm8Δ, brown. (B) trm8Δ mutants are temperature sensitive due to lack of trm8 + . Strains with plasmids as indicated were grown overnight in EMM-Leu media at 30˚C, diluted to OD 600~0 .5, serially diluted 10-fold in EMM-Leu, and 2 μL was spotted onto plates containing EMM-Leu or YES media and incubated at 33˚C, 37˚C, and 38˚C. The two independent trm8Δ mutants were labeled as Sp trm8Δ V1 and V2. https://doi.org/10.1371/journal.pgen.1008893.g001 Surprisingly, on both YES media and EMM complete (EMMC) media lacking leucine, trm8Δ mutants expressing tY(GUA) grew almost as well as the trm8Δ [P trm8 trm8 + ] strain or the WT strain at elevated temperatures, whereas trm8Δ mutants expressing two tP(AGG) genes on a plasmid had little effect on the temperature sensitivity ( Fig 2D). As expected, northern analysis showed that trm8Δ [leu2 + tY(GUA)] strains had substantially more tY(GUA) than the trm8Δ [leu2 + ] vector control strain at 30˚C and 38.5˚C (3.2-fold and 6.8-fold more respectively) (S6 Fig). Similarly, trm8Δ mutants expressing two copies of tP(AGG) had more tP(AGG) at 30˚C and 38.5˚C than the vector control, and the levels of the control tT(AGU) were unchanged in all strains at both temperatures. We conclude that although levels of both tY(GUA) and tP (AGG) were reduced in trm8Δ mutants at elevated temperatures in both YES and EMMC media, tY(GUA) is the major physiologically important tRNA for these phenotypes.
Although tY(GUA) overexpression almost completely restored growth of S. pombe trm8Δ mutants in YES and EMMC media at 38˚C and 39˚C, expression of both tY(GUA) and tP (AGG) was required to completely suppress the growth defects in YES + glycerol media (S7 Fig). By contrast, overexpression of tY(GUA) and tP(AGG) had no effect on the known temperature sensitivity of trm8Δ mutants in YES media containing 5-FU [61,62] (S7 Fig), perhaps due to reduced levels of C and 5-methyluridine modifications, which could trigger decay of other hypomodified tRNA species in trm8Δ mutants. dhp1 mutations suppress the S. pombe trm8Δ growth defect and restore tY (GUA) and tP(AGG) levels To identify the mechanisms that restore growth to S. pombe trm8Δ mutants at elevated temperatures, we isolated and analyzed spontaneous suppressors of the temperature sensitivity. One major class of four trm8Δ suppressors were as temperature resistant as WT on YES and EMMC media, and nearly as resistant as WT on YES + 5-FU media (Figs 3A and S8). Genome sequencing revealed that these mutants each had distinct missense mutations in the RAT1 ortholog dhp1 + . The dhp1 mutations each occurred in highly conserved residues, based on an alignment of 18 RAT1/dhp1 + eukaryotic homologs from multiple phyla (Fig 3B), and are presumably partial loss of function mutations as S. pombe dhp1 + , like S. cerevisiae RAT1, is an essential gene [63,64].
Because we obtained four genetically independent S. pombe trm8Δ dhp1 mutants and very few other mutations in the whole genome sequencing, it was highly likely that the dhp1 mutations were responsible for the restoration of growth in trm8Δ dhp1 mutants. Consistent with this, a plasmid expressing dhp1 + complemented the S. pombe trm8Δ dhp1-1 suppressor, resulting in temperature sensitivity, but had no effect on WT or trm8Δ mutants (S9 Fig). Thus, we conclude that the dhp1 mutations were responsible for the rescue of growth at high temperature. Strains were grown in YES media at 30˚C, shifted to the indicated temperatures for 8 hours as described in Materials and Methods, and RNA was isolated and analyzed by northern blotting. n = 2 for WT cells, n = 3 for S. pombe trm8Δ mutants. (B) Quantification of tP(AGG), tY(GUA), and tT(AGU) levels in WT and trm8Δ mutants at different temperatures. The bar chart depicts relative levels of tRNA species at each temperature, relative to their levels in the WT strain at 30˚C (each value itself first normalized to levels of the control non-Trm8 substrate tG(GCC)). For each tRNA, lighter shades indicate progressively higher temperatures (30˚C, 36.5˚C, 37.5˚C to 38.5˚C) for tT(AGU), brown; tP(AGG), purple; tY(GUA), gray. Standard deviations for each tRNA measurement are indicated. The statistical significance of tRNA levels was evaluated using a two-tailed Student's t-test assuming equal variance. ns, not significant (p > 0.05); � , p < 0.05; �� , p < 0.01; ��� , p < 0.001. (C) Schematic of the secondary structure of tY(GUA)-1 and tP(AGG). Modifications of tY(GUA) are as annotated. WC base pairs, black lines; GU base pairs, black dots; mismatch C-A or C-U base pairs, red diamonds; presumed m 7 G 46 , red circle. (D) Overproduction of tY(GUA), but not tP(AGG), suppressed the temperature sensitive growth defect of trm8Δ mutants. Strains with plasmids as indicated were grown overnight in EMMC-Leu media at 30˚C and analyzed for growth as in Fig 1B on  As Dhp1 encodes a 5'-3' exonuclease [65], it seemed highly likely that the S. pombe trm8Δ dhp1 mutants prevented decay of tY(GUA) and tP(AGG) at non-permissive temperature. Indeed, we found that for each of two dhp1 suppressors, tY(GUA) levels were almost completely restored at 38.5˚C, from 29% in the trm8Δ mutant to 85% and 82% in the trm8Δ dhp1-1 and trm8Δ dhp1-2 strains respectively (Fig 3C and 3D). Similarly, tP(AGG) levels were virtually completely restored at 38.5˚C, from 16% in the trm8Δ mutant to 74% and 66% in the trm8Δ dhp1 suppressors, and the levels of the control tRNA (tT(AGU)) was unaffected. Similar restoration of tY(GUA) and tP(AGG) levels was also observed in the two other trm8Δ dhp1 suppressors at 38.5˚C (S10 Fig). We conclude that, as for RTD in S. cerevisiae modification mutants [22,39,40], tRNA decay in S. pombe trm8Δ mutants occurs by 5'-3' exonucleolytic degradation of tRNA, providing strong evidence for conservation of the RTD pathway in S. pombe.

Mutations in the GAAC pathway suppress the S. pombe trm8Δ growth defect and restore tRNA levels
A second major group of six S. pombe trm8Δ suppressors was temperature resistant on YES and EMMC media, but sensitive on YES + 5-FU media, and genome sequencing showed that these suppressors each had distinct mutations in elements of the GAAC pathway (Figs 4A and S11). Among these, we found three trm8Δ suppressors with gcn2 mutations, one with a gcn1 mutation, and two with tif221 mutations, encoding the translation initiation factor eIF2Bα (S2 Table). Each of these genes in S. cerevisiae is known to be critical for the GAAC pathway [66][67][68], which is widely conserved in eukaryotes, including S. pombe and mammals [69][70][71][72][73][74]. In this pathway, amino acid starvation leads to uncharged tRNAs that bind Gcn2 to activate its kinase domain, phosphorylation of eIF2α by Gcn2, global repression of translation, and derepression of translation of the transcription factor Gcn4, resulting in increased transcription of nearly one tenth of the S. cerevisiae genes [66,[75][76][77][78]. A similar massive transcription program change occurs in S. pombe after amino acid starvation [79].
Consistent with their role as S. pombe trm8Δ suppressors, all six of the trm8Δ GAAC mutants had increased levels of tY(GUA) and tP(AGG) at high temperature. After growth in YES media at 38.5˚C, all trm8Δ GAAC mutants showed a 1.7-fold to 2.1-fold increase in tY (GUA) levels compared to the parent trm8Δ mutant, and tP(AGG) levels were increased~3 fold, whereas the controls tT(AGU) and tV(AAC) did not have increased levels (Figs 4C and 4D and S13). These results provided strong evidence that the effect of the GAAC mutants was to increase tRNA levels to restore growth.  Fig  1B. (B) Mutations in dhp1 that restored growth of S. pombe trm8Δ mutants reside in evolutionarily conserved residues. The amino acid sequence of Sp Dhp1 was aligned with putative Rat1/Dhp1 orthologs from 17 evolutionarily distant eukaryotes, using MultAlin (http://multalin.toulouse.inra.fr/multalin/) [135]. red, > 90% conservation; blue, 50%-90% conservation. Alleles of dhp1 mutations are indicated at the top. (C) Each of two trm8Δ dhp1 mutants had restored tRNA levels in YES media at 38.5˚C. Strains were grown in YES media at 30˚C and shifted to 38.5˚C for 8 hours, and RNA was isolated and analyzed by northern blotting as in Fig 2A.  To test if the temperature resistance of S.pombe trm8Δ GAAC mutants extended further downstream within the GAAC pathway, we deleted the GAAC transcription factor fil1 + , the functional equivalent of S. cerevisiae Gcn4 [82]. We found that trm8Δ filΔ mutants were  Fig 1B. (B) Expression of gcn2 + and tif221 + complemented the suppression phenotype of trm8Δ gcn2-1 and trm8Δ tif221-2 mutants respectively. S. pombe trm8Δ gcn2-1 and trm8Δ tif221-2 mutants expressing gcn2 + and tif221 + respectively, or a vector, were grown overnight in EMMC-Leu media at 30˚C, and analyzed for growth as in Fig 1B. Note that expression of gcn2 + was kept to modest levels by adding thiamine to the media to partially suppress overexpression from the P nmt1 �� promoter. (C) gcn1, gcn2, and tif221 mutations each partially restored tY(GUA) and tP(AGG) levels of trm8Δ mutants. Strains were grown in YES media at 30˚C and shifted to 38.5˚C for 8 hours as described in Materials and Methods, and RNA was isolated and analyzed by northern blotting. distinctly more temperature resistant than trm8Δ mutants in EMMC-His media, but not as temperature resistant as a trm8Δ gcn2Δ mutant (S14 Fig). We thus infer that suppression of the trm8Δ temperature sensitivity observed in S. pombe trm8Δ GAAC mutants is due in part to transcription activation by Fil1. We note that the filΔ mutation did not rescue the temperature sensitivity of trm8Δ mutants in YES media; this could be due in part to the temperature sensitivity of fil1Δ mutants in YES media.
The temperature sensitivity of S. pombe trm8Δ mutants coincides precisely with the onset of GAAC activation and tY(GUA) decay Since S. pombe trm8Δ suppressors in different components of the GAAC pathway all restored tY(GUA) and tP(AAG) levels, we inferred that trm8Δ mutants activated the GAAC pathway at non-permissive temperatures, and that this activation somehow promoted further loss of the tRNA. To establish the precise connection between growth, tRNA levels, and GAAC activation in trm8Δ mutants, we measured each parameter during liquid growth in rich media at a permissive temperature (30˚C), and at three elevated temperatures: 36.5˚C, 37.5˚C, and 38.5˚C. In this experiment, trm8Δ mutants grew virtually identically to WT control strains at 36.5˚C and 37.5˚C, and the growth defect was only obvious at 38.5˚C (S15 Fig).
Strikingly, S. pombe trm8Δ mutants activated the GAAC pathway only at 38.5˚C, the lowest temperature at which the growth defect was obvious. We measured GAAC activation by measuring mRNA levels of the known GAAC targets lys4 + and aro8 + (SPAC56E4.03) [79], which we had previously used [83]. At 38.5˚C in trm8Δ mutants, we observed a 7.1-fold increase in lys4 + mRNA levels (relative to the standard act1 + ), compared to that from WT or trm8Δ mutants at 30˚C (5.3 vs 0.74 and 0.74) (Fig 5A). By contrast, we found no measurable change in lys4 + mRNA levels in trm8Δ mutants grown at 36.5˚C and 37.5˚C, relative to that observed in trm8Δ mutants at 30˚C, or in WT cells at any temperature. Moreover, the increase in relative lys4 + mRNA levels in trm8Δ mutants in YES media at 38.5˚C was almost as high as that observed in WT cells induced with 3-AT, and was completely eliminated in trm8Δ gcn2-1 mutants. Examination of relative aro8 + mRNA levels gave a similar result (S16 Fig): a substantial Gcn2-dependent increase in relative aro8 + mRNA levels at 38.5˚C in trm8Δ mutants, relative to 37.5˚C (1.5 vs 0.45), and no change in relative aro8 + mRNA levels at 37.5˚C in trm8Δ mutants compared to WT (0.45 vs 0.44). Consistent with the appearance of the S. pombe trm8Δ growth defect and the GAAC activation only at 38.5˚C, tY(GUA) decay was only significant in YES media at 38.5˚C, and at that temperature the gcn2-1 mutation significantly restored tY(GUA) levels ( Fig 5B and 5C).
As anticipated, phosphorylation of eIF2α tracked with GAAC activation. We grew WT and S. pombe trm8Δ mutants at 30˚C and 38.5˚C, and measured both eIF2α phosphorylation levels and GAAC activation of lys4 + and aro8 + mRNA expression. We observed much more pronounced levels of eIF2α phosphorylation in trm8Δ mutants at 38.5˚C, compared to modest phosphorylation levels in WT at 38.5˚C, and much reduced phosphorylation in both trm8Δ mutants and WT at 30˚C (S17A Fig). Analysis of the same samples by RT-qPCR showed a substantial increase of lys4 + and aro8 + mRNAs in trm8Δ mutants at 38.5˚C, but not in WT at 38.5˚C or in either trm8Δ or WT at 30˚C (S17B Fig), just as we observed in Figs 5A and S16. These results provide evidence that the Gcn2 mediated GAAC activation of expression of lys4 + and aro8 + mRNAs occurs through eIF2α phophorylation in trm8Δ mutants at 38.5˚C.
As tY(GUA) was the major physiologically relevant substrate in YES media (Fig 2D), we speculated that at 38.5˚C, tY(GUA) decay might be driving the GAAC activation associated with the trm8Δ growth defect. Alternatively, GAAC activation could be a consequence of both tY(GUA) and tP(AAG) decay, reduced tRNA charging associated with trm8Δ mutants at high temperature, or partly as a consequence of temperature stress itself, as a number of different stress conditions are known to activate the GAAC pathway [72,74,84,85].
To determine the extent to which reduced tRNA levels activated the GAAC pathway, we examined GAAC induction of S. pombe trm8Δ strains after overproduction of tY(GUA) and tP(AGG), using the same samples we used to show that overproduction of tY(GUA) suppressed the trm8Δ temperature sensitivity (Figs 2D and S6). As expected, relative lys4 + mRNA levels were increased in trm8Δ [vector] strains grown at 38.5˚C, compared to this strain at 30˚C (4.2 vs 0. 57, 7.4-fold) or to WT at 38.5˚C (7.0 fold), indicating GAAC activation ( Fig  5D). Notably, relative lys4 + mRNA levels were reduced 4.0-fold in the trm8Δ [tY(GUA)] strain compared to the corresponding trm8Δ [vector] strain (from 4.2 to 1.04), and were not reduced in the trm8Δ [tP(AGG)] strain. Based on these results, we conclude that reduced function of tY(GUA) is the primary cause of GAAC activation and temperature sensitivity of trm8Δ mutants at 38.5˚C. As charging of tY(GUA) was distinctly but marginally reduced at 38.5˚C relative to WT (S18 Fig), we cannot determine if the GAAC activation was due to reduced levels of tY(GUA), or to a combination of reduced levels and charging [69,76,80,83,86].
To determine the effects of Trm8 and Gcn2 on processing and transcription, we examined expression of the four pre-tY(GUA) species (1-1, 1-2, 1-3, and 2, as described [59]) in the WT, trm8Δ, and trm8Δ gcn2-1 strains, using appropriate intron-specific probes. With a probe specific for the pre-tY(GUA)-2 species, we found that at all temperatures trm8Δ mutants accumulated significantly more of the 3' end-extended and the end-matured pre-tY(GUA) species than the WT strains (S19 Fig). Similarly, with the pre-tY(GUA)-1-3 probe, we observed accumulation of the end-matured pre-tY(GUA) in trm8Δ mutants at all temperatures. These results suggest a processing defect due to lack of m 7 G for these pre-tY(GUA) species. We also found that levels of the primary transcript, corresponding to the largest pre-tY(GUA) species, were slightly elevated in trm8Δ gcn2-1 mutants relative to trm8Δ mutants, at both 37.5˚C and 38.5˚C, but not at 36.5˚C (S19A and S19B Fig). This result suggests that tRNA transcription might play some role in restoring tY(GUA) levels in a trm8Δ gcn2-1 mutant at 37.5˚C and 38.5˚C, although it is not clear yet if this effect accounts for all of the suppression.

In S. cerevisiae, mutation of the GAAC pathway exacerbates the effects of the RTD pathway
To investigate the evolutionary implications of the GAAC pathway on RTD, we examined the consequences of deletion of GAAC components in S. cerevisiae trm8Δ trm4Δ mutants, which are highly temperature sensitive due to substantial decay of tV(AAC) by the RTD pathway, compared to the modest RTD-dependent temperature sensitivity and the limited tV(AAC) decay of S. cerevisiae trm8Δ mutants [18,39]. In contrast to our results in S. pombe trm8Δ  WT, trm8Δ, and trm8Δ gcn2-1 mutants at different temperatures. tRNA levels were quantified as described in Fig 2B. (D) tY(GUA) overproduction repressed the GAAC induction of trm8Δ mutants at 38.5˚C. Strains as indicated with plasmids expressing tY(GUA) and/or tP (AGG) were grown in EMMC-Leu media at 30˚C and shifted to 38.5˚C for 8 hours, and then RNA was isolated and lys4 + mRNA levels were analyzed by RT-qPCR as described in Fig 5A. Right side: GAAC induction of WT cells grown at 30˚C in EMMC-His media, and treated with 10 mM of 3-AT for 4 hours, evaluated in parallel to other samples. GAAC mutants, deletion of GCN1 or GCN2 exacerbated the temperature sensitivity of S. cerevisiae trm8Δ trm4Δ mutants in both rich (YPD) and minimal complete (SDC) media, and this exacerbated temperature sensitivity was also observed upon deletion of the GAAC transcription factor GCN4 (Fig 6A). Moreover, we found that a met22Δ mutation, known to prevent RTD, reversed the enhanced temperature sensitivity of a trm8Δ trm4Δ gcn2Δ strain relative to a trm8Δ trm4Δ mutant (S20 Fig). Furthermore, a similar exacerbated temperature sensitivity due to mutation of the GAAC pathway was also observed in other S. cerevisiae modification mutants known to be subject to RTD [22,39], including trm8Δ, trm1Δ, tan1Δ, and tan1Δ trm44Δ mutants (S21 Fig). To determine if the exacerbated growth defect of S. cerevisiae trm8Δ trm4Δ GAAC mutants was due to exacerbated loss of tV(AAC), we analyzed tRNA levels after a four-hour temperature shift from permissive to non-permissive temperature (27˚C to 32˚C). Consistent with the exacerbated temperature sensitivity caused by the gcn1Δ and gcn2Δ mutations, tV(AAC) levels were further reduced in both trm8Δ trm4Δ gcn1Δ and trm8Δ trm4Δ gcn2Δ mutants at 32˚C, compared to the trm8Δ trm4Δ mutant (12% and 17% vs 43%, relative to the values at 27˚C) (Fig 6B and 6C). Using a pre-tV(AAC) probe specific for 7 of the 14 tV(AAC) genes in S. cerevisiae, we found that levels of the primary pre-tV(AAC) transcript were modestly reduced at 32˚C in the trm8Δ trm4Δ gcn1Δ mutant, relative to the trm8Δ trm4Δ mutant (S22 Fig). This result suggests that reduced pre-tV(AAC) transcription could be responsible for some of the exacerbated loss of tV(AAC) in the trm8Δ trm4Δ gcn1Δ mutant and for its exacerbated temperature sensitivity, although the magnitude of the change may not account for all of the additional loss of tV(AAC).

The temperature sensitivity and tV(AAC) decay of S. cerevisiae trm8Δ trm4Δ mutants coincides with GAAC activation
Since GAAC mutations exacerbated the RTD growth defect and enhanced the decay of an S. cerevisiae trm8Δ trm4Δ mutant at 32˚C, it seemed likely that the GAAC pathway was activated in the trm8Δ trm4Δ mutant. To evaluate GAAC activation, we measured mRNA levels of the Gcn4 target genes LYS1 and HIS5 after the 4 hour temperature shift of trm8Δ trm4Δ mutants to 32˚C, using the same RNA as in the northern analysis of decay (Figs 6B and 6C and S22). RT-qPCR analysis showed that trm8Δ trm4Δ mutants had a large increase in relative LYS1 mRNA levels at 32˚C, compared to 27˚C (22.2 vs 0.81, 27.4-fold), or to WT cells at either 27˚C or 32˚C (Fig 6D), showing that the GAAC activation was specific to the trm8Δ trm4Δ mutant at 32˚C. We observed a similar activation of the GCN4 target HIS5 in the trm8Δ trm4Δ mutant at 32˚C. This GAAC activation was correlated with a modest but distinct increase in uncharged tV(AAC) commensurate with the reduced tV(AAC) levels (S23 Fig). Furthermore, we found that overproduction of tV(AAC) in trm8Δ trm4Δ mutants suppressed induction of the GAAC pathway (Fig 6E), showing that GAAC activation in trm8Δ trm4Δ mutants was due to the reduced function of tV(AAC). Thus, in both S. cerevisiae trm8Δ trm4Δ mutants and S pombe trm8Δ mutants, degradation of a single biologically relevant tRNA is the cause of GAAC induction, which then either promotes further loss of tRNA in S. pombe, or restores tRNA in S. cerevisiae (Fig 7).

Discussion
The results described here provide strong evidence that the RTD pathway is conserved between the distantly related species S. cerevisiae and S. pombe. We have shown that the temperature sensitivity of S. pombe trm8Δ mutants is due to reduced levels of tY(GUA) and to some extent tP (AGG), and is efficiently suppressed by mutations in the 5'-3' exonuclease Dhp1 that concomitantly restore the levels of these tRNAs, strongly suggesting decay of tY(GUA) and tP (AGG) by the RTD pathway. As RTD is triggered in S. cerevisiae strains lacking m 2,2 G 26 or ac 4 C 12 , as well as in strains lacking m 7 G 46 , [22,39], we speculate that the RTD pathway will also act in S. pombe strains lacking other body modifications. Furthermore, given the large evolutionary distance between S. cerevisiae and S. pombe, we speculate that the RTD pathway is conserved throughout eukaryotes. The existence of a mammalian RTD pathway could explain the reduced levels of specific tRNA species in mouse strains lacking m 5 C in their tRNAs [33,87] and in mouse embryonic stem cells lacking m 7 G in their tRNAs [46], and might explain other phenotypes associated with mutations in METTL1 or WDR4 [29,30,46,88].
It is puzzling that although tP(AGG) is substantially more degraded than tY(GUA) in S. pombe trm8Δ mutants at elevated temperatures, the temperature sensitivity of the mutants in both rich and minimal media is primarily due to decay of tY(GUA). One possible explanation of this result is that tY(GUA) levels might be more limiting in the cell than tP(AGG) levels at elevated temperature, relative to the number of their respective cognate codons requiring decoding. This type of argument was advanced as a possible explanation for why the growth defects of i 6 A-lacking S. pombe tit1Δ mutants were rescued by increased expression of tY (GUA), but not by any of the other four Tit1 tRNA substrates [89]. A second, and less likely, interpretation is that tP(AGG) decoding might be compensated by other tRNA Pro isodecoders specific for the CCN codon box. This explanation is based on the finding that in S. cerevisiae, deletion of the two tP(AGG) genes is viable, implying that the 10 tP(UGG) isodecoders can decode all four proline CCN codons [90]. However, as S. pombe has six tP(AGG) genes and only two tP(UGG) genes (as well as one tP(CGG)), it seems unlikely that the loss of almost all of the tP(AGG) in trm8Δ mutants can be efficiently compensated by the small number of tP (UGG) species (assuming that tRNA expression from each gene is comparable).
It is not immediately clear why tP(AGG) and tY(GUA) are the specific tRNAs subject to RTD in S. pombe trm8Δ mutants. Based on current understanding of RTD determinants in S. cerevisiae, RTD substrate specificity is determined by stability of the stacked acceptor and Tstem, with contributions to stability from the tertiary fold that are reduced in modification mutants, and some contributions from the other two stems, but not the anticodon loop [18,22,[39][40][41][42]. tP(AGG) may be an RTD substrate because it is predicted to have a less stable acceptor and T-stem than most Trm8 substrates (S1 Table) [91]. Furthermore, the destabilizing C 4 -A 69 mismatch in the middle of the tP(AGG) acceptor stem might be expected to lead to increased local breathing at the 5' end, which is likely important for recognition by the 5'-3' exonucleases Xrn1 and Rat1, as the Xrn1 active site binds the three most 5' nucleotides [92]. However, it is more difficult to rationalize why tY(GUA) is a substrate for RTD in trm8Δ mutants, as its acceptor and T-stem are predicted to be moderately stable among Trm8 substrates. However, tY(GUA) does have a destabilizing N 27 -N 43 pair (C 27 -U 43 for 3 isodecoders, grown overnight in YPD media at 27˚C and analyzed for growth on YPD and SDC plates at different temperatures. (B) Deletion of GCN1 or GCN2 exacerbated tV(AAC) decay of trm8Δ trm4Δ mutants at 32˚C. Strains were grown in YPD media 27˚C, shifted to 32˚C and harvested after 4 hours, and then bulk RNA was isolated and analyzed by northern blotting. Fig  6B. The bar chart depicts levels of tRNA species at 27˚C or 32˚C, relative to levels of that tRNA in the WT strain at 27˚C (each value itself first normalized to levels of the control 5S rRNA). tRNA levels are indicated by diagonal hatching for WT (green); S. cerevisiae (Sc) trm8Δ trm4Δ (brown); Sc trm8Δ trm4Δ gcn2Δ (light blue); and Sc trm8Δ trm4Δ gcn1Δ (dark blue). (D) trm8Δ trm4Δ mutants induced the GAAC pathway at 32˚C. Bulk RNA from the growth done for Fig  6B was used for RT-qPCR analysis of levels of LYS1 and HIS5 mRNA, normalized to ACT1. mRNA levels are indicated by horizontal lines for LYS1 and hatching for HIS5, WT (green); Sc trm8Δ trm4Δ (brown). Right side: Relative levels of LYS1 mRNA of WT cells grown at 30˚C in SD-His media, and then treated with 10 mM 3-AT for 1 hour, evaluated in parallel to other samples. (E) tV(AAC) overproduction repressed the GAAC induction of trm8Δ trm4Δ mutants at 36˚C. Strains with plasmids as indicated were grown in SD-Ura media 27˚C and shifted to 36˚C for 1 hour, and then RNA was isolated and relative LYS1 mRNA levels were analyzed by RT-qPCR as described in Fig 6D. Right side: GAAC induction of WT cells grown and induced with 3-AT as described in Fig 6D. https://doi.org/10.1371/journal.pgen.1008893.g006 and U 27 -U 43 for one isodecoder), which might reduce the stability of the tertiary fold by affecting stability of the adjacent tertiary 26-44 interaction [93][94][95].

(C) Quantification of tRNA tV(AAC), tR(UCU), and tW(CCA) levels from the northern in
It is not yet clear how Dhp1 degrades tY(GUA) in S. pombe trm8Δ mutants. In S. cerevisiae, Rat1 is nuclear [96] and catalyzes a substantial amount of the decay of mature tV(AAC) in trm8Δ trm4Δ mutants [39], suggesting that the retrograde transport pathway is required to . The further reduced levels of tRNA resulting from GAAC induction is in part due to transcription upregulation of Fil1 target genes (solid lines), and may also be due in part to the global reduction in translation (dotted lines). Right: S. cerevisiae trm8Δ trm4Δ mutants (green) trigger RTD of tV(AAC), leading to GAAC induction. This results in inhibition of further loss of tV (AAC), which is due in part to the the transcription upregulation of Gcn4 target genes (solid lines), and may also be due to global reduction in translation (dotted lines).
https://doi.org/10.1371/journal.pgen.1008893.g007 deliver the tV(AAC) substrate to the nucleus [5,[97][98][99][100]. However, we do not know the exact species of tY(GUA) that is degraded by Dhp1 in S. pombe trm8Δ mutants. If mature tY(GUA) is the actual Dhp1 substrate, it is almost certainly subject to retrograde transport back to the nucleus for the subsequent decay, because Dhp1 is known to be nuclear [64,101] and S. pombe pre-tRNA splicing initiates in the cytoplasm on the mitochondrial surface [102]. Such a retrograde transport mechanism for hypomodified tY(GUA) lacking m 7 G 46 in S. pombe trm8Δ mutants would be similar to that shown for hypomodified tRNAs lacking m 2,2 G 26 in S. cerevisiae trm1Δ mutants [100]. However, it is also possible that Dhp1 acts to degrade unspliced pre-tY(GUA) that accumulates in trm8Δ mutants, analagous to the recently described Met22-dependent pre-tRNA decay pathway [103].
It seems likely that the 5-FU sensitivity of S. pombe trm8Δ mutants is due to decay of multiple tRNA species in the presence of the drug, caused by the reduced levels of C and m 5 U [48][49][50], in addition to the lack of m 7 G. This interpretation is consistent with the lack of suppression of the 5-FU sensitivity of trm8Δ mutants by tY(GUA) and tP(AGG), and its almost complete suppression by dhp1 mutations, and is consistent with the enhanced 5-FU sensitivity of a number of tRNA body modification mutants in S. cerevisiae [61].
Our finding that loss of function of tRNA due to tRNA decay is itself the trigger for induction of the GAAC response in both S. pombe trm8Δ mutants and S. cerevisiae trm8Δ trm4Δ mutants suggests an intimate relationship between reduced tRNA function and GAAC activation. The loss of functional tRNA below some presumed threshold level is the proximal cause of GAAC induction, because in each organism the GAAC pathway is activated at the lowest temperature at which tRNA decay and a growth defect is observed, and in each organism overproduction of the physiologically relevant tRNA represses GAAC induction. The GAAC pathway has previously been implicated in the biology of a number of anticodon loop modifications [83,104,105]. Robust constitutive GAAC induction is observed in S. cerevisiae and S. pombe trm7Δ mutants (lacking Nm 32 and Nm 34 ) and S. cerevisiae pus3Δ mutants (lacking C 38 and C 39 ) [83,105], each of which has a constitutive growth defect [106,107], and GAAC induction is known to be Gcn2-dependent in S. cerevisiae trm7Δ mutants [83]. By contrast, S. cerevisiae mutants lacking either the mcm 5 U or the s 2 U moiety of mcm 5 s 2 U induce the GAAC pathway independently of Gcn2 at 30˚C [104] and are temperature sensitive at 37˚C [108]. Our finding that S. pombe trm8Δ and S. cerevisiae trm8Δ trm4Δ mutants each trigger Gcn2-dependent GAAC induction only at the temperature that the growth defect is observed is consistent with Gcn2-dependent GAAC induction in S. cerevisiae anticodon loop modification mutants with a constitutive growth defect.
It is striking that the induction of the GAAC response due to tRNA decay in S. pombe trm8Δ mutants and S. cerevisiae trm8Δ trm4Δ mutants has opposite consequences in each organism. In S. pombe trm8Δ mutants, activation of the GAAC response exacerbates the growth defect, as mutation of any of four components (gcn1, gcn2, tif221, or fil1) protects against loss of tRNA. Activation of the GAAC pathway is also part of the reason that S. cerevisiae trm7Δ mutants grow poorly [83], and defects in the integrated stress response pathway (ISR) in humans are implicated in disease phenotypes [109][110][111]. By contrast, in S. cerevisiae trm8Δ trm4Δ mutants, activation of the GAAC response rescues the growth defect, as deletion of any of three GAAC components (gcn1Δ, gcn2Δ, or gcn4Δ) exacerbates the growth defect. Furthermore, this result extends to multiple S. cerevisiae modification mutants with an RTD phenotype, since a gcn2Δ mutation also exacerbated the growth defects of trm8Δ, trm1Δ, tan1Δ, and tan1Δ trm44Δ mutants. Although the rescue of RTD in S. cerevisiae by GAAC activation is opposite to the exacerbating effect of GAAC activation in S. pombe trm8Δ mutants, it is consistent with a concerted stress response. Moreover, the finding that GAAC effects on RTD extended to fil1Δ/ gcn4Δ mutations in S. pombe and S. cerevisiae has mechanistic implications. Deletion of GCN1 or GCN2 each prevent sensing of tRNA status, and the consequent eIF2α phosphorylation, reduced translation initiation, reduced global translation, and massive downstream transcription activation. However, as fil1 + /Gcn4 is downstream of the sensing machinery, but upstream of the transcription activation, we infer that the GAAC effects on RTD in S. pombe and S. cerevisiae are in part due to lack of transcription activation by fil1 + /Gcn4.
The opposite effects of GAAC activation on growth and RTD in S. pombe trm8Δ mutants and in several S. cerevisiae RTD mutants is likely due to a differential GAAC response. Although in each organism the GAAC pathway is known to regulate the transcription of more than 500 genes [75,77,79,82], there are distinct differences in the GAAC response in the two species. For example, it is known that the GAAC response to amino acid starvation results in repression of methionine synthesis genes in S. pombe but induction of these genes in S. cerevisiae [79]. As a met22Δ mutation is known to inhibit RTD in S. cerevisiae [22,39], this opposite GAAC activation effect on methionine genes in the two organisms is in the wrong direction to explain the opposite RTD effects of the GAAC pathway. Other possible explanations for the differential effects of the GAAC pathway on RTD include differential regulation of the synthesis or biochemical activity of RTD regulators such as EF1A, aminoacyl tRNA synthetases, pol III transcription [22,112], or 5'-3' exonucleases [39,42], as well as changes in any number of indirect effectors affecting overall levels or availability of tRNA and/or nucleases. In addition, the overall stress response pathways are substantially different between S. cerevisiae and S. pombe. In S. cerevisiae, Gcn2 is the sole eIF2α kinase regulating stress responses [78,113,114], whereas in S. pombe three different eIF2α kinases (Gcn2, Hri1 and Hri2) [115] each respond to a diverse set of stress treatments [72,74,84,85,109]. The differences in kinases affecting eIF2α phosphorylation implies substantial differences between the two species in regulation of all sorts of combinations of stress response, which might be occurring at elevated temperature when tRNA decay is occurring [116].
The results outlined here underscore that GAAC activation occurs in S. cerevisiae and S. pombe trm8Δ modification mutants precisely at the point of observed growth stress due to tRNA decay, albeit with different effects in S. pombe trm8Δ mutants and S. cerevisiae trm8Δ trm4Δ mutants. These results, coupled with the constitutive GAAC activation in S. pombe and S. cerevisiae trm7Δ mutants and S. cerevisiae pus3Δ mutants [83,105], fuel speculation that the GAAC response will also be activated in mammals and other eukaryotes with tRNA modification mutations or other mutations that result in reduced tRNA function. Given that GAAC activation at the onset of reduced tRNA function regulates RTD in opposite ways in S. pombe and S. cerevisiae, it would be interesting to determine the GAAC effect on tRNA decay and tRNA levels in mammalian systems. Based on the observation that GAAC activation in mice attenuates the growth defects caused by the combination of reduced tRNA Arg(UCU) levels and a defect in the ribosome recycling component GTPBP2 [86], we speculate that GAAC activation by reduced tRNA function in mammals will likewise attenuate RTD and promote survival.
As Trm8 is phosphorylated and likely inactivated by treatment of HEK293 cells with insulin-like growth factor-1 [117], it seems plausible that if RTD is conserved in mammals, m 7 G modification dynamics could be used to regulate tRNA levels physiologically. There is growing evidence that levels of a number of modifications are under dynamic control in different conditions [118,119]. There is also evidence that regulation of tRNA expression plays an important role in differentiation and proliferation, and is also a characteristic of breast cancer [120][121][122][123][124]. For example, tRNA Arg iso-acceptors have different expression in differentiated vs proliferating cells [123] and tR(UCU) iso-decoders show tissue specific regulation of expression [125]. It remains to be determined if regulated changes in expression, phosphorylation, or biochemical activity of METTL1 or WDR4 result in altered m 7 G levels and consequent changes in tRNA levels that physiologically regulate expression.

Plasmids
Plasmids used in this study are listed in S4 Table. AB553-1 was constructed by insertion of a NotI restriction site between the P nmt1 promoter and the PstI site of the pREP3X plasmid. The S. pombe plasmid expressing S. pombe P trm8+ trm8 + (ETD 67-1) was constructed by inserting PCR amplified DNA genomic DNA (including 1000 bp upstream and 1000 bp downstream) into the NotI and XhoI sites of AB 553-1, removing the P nmt1 promoter. Plasmids expressing S. pombe P dhp1 dhp1 + or S. pombe tRNA genes were constructed using the same approach, including~300 bp upstream and 300 bp downstream for the tRNA genes. S. pombe plasmids expressing P nmt1 � � gcn2 + (low strength, no message in thiamine) or P nmt1 � tif221 + (medium strength, no message in thiamine) were constructed by PCR amplification of the respective coding sequence from S. pombe WT genomic DNA (including introns), and insertion into the XhoI and BamHI sites of the pREP81X or pREP41X vectors respectively.

Yeast media and growth conditions
S. pombe strains were grown at desired temperatures in rich (YES) media (containing 0.5% yeast extract, 3% glucose, and supplements of 225 mg/l of adenine, uracil, leucine, histidine and lysine), or Edinburgh minimal media (EMM) containing glucose and the same supplements, as well as similar amounts of relevant auxotrophic requirements. Minimal complete (EMM-C) media was supplemented with 225 mg/l of all amino acids, adenine, and uracil, as well as 100 mg/l of para-amino benzoic acid and inositol, and 1125 mg/l of leucine for Leuauxotrophs [79]. For temperature shift experiments, cells were grown in YES or EMMC media at 30˚C to OD 600~0 .5, diluted to~0.1 OD in pre-warmed media at the desired temperature, grown to OD~0.5, harvested at 4˚C, washed with ice cold water, frozen on dry ice, and stored at -80˚C. To select spontaneous suppressors of S. pombe trm8Δ mutants, cells were grown overnight in YES media at 30˚C and~10 7 cells were plated on YES media plates at 38˚C and 39˚C. S. cerevisiae strains were grown in rich (YPD) media (containing 1% yeast extract, 2% peptone, 2% dextrose, and 80 mg/L adenine hemisulfate), or minimal complete (SDC) media [129] as indicated, and temperature shift experiments were performed as described for S. pombe. All experiments with measurements were performed in biological triplicate, unless otherwise noted.

Quantitative RT-PCR analysis
Strains were grown in triplicate to log phase and bulk RNA was prepared from 2-5 OD pellets using acid washed glass beads and phenol. Then, RNA was treated with RQ1 RNase-free DNase (Promega), reverse transcribed with Superscript II Reverse Transcriptase, and quantitative PCR was performed on the cDNA as previously described [131].

Isolation and purification of tRNA
S. pombe WT and trm8Δ mutant strains were grown to~0.5 OD in YES media at 30˚C. Then bulk low molecular weight RNA was extracted from~300 OD of pellets by using hot phenol, and tRNAs were purified using 5'-biotinylated oligonucleotides complementary to the corresponding tRNAs (S6 Table) as previously described [132].

HPLC analysis of nucleosides of purified tRNA
Purified tRNAs (~1.25 μg) were digested to nucleosides by treatment with P1 nuclease, followed by phosphotase, as previously described [132], and nucleosides were analyzed by HPLC at pH 7.0 as previously described [133].

Whole genome sequencing
Whole genome sequencing was performed by the University of Rochester Genomics Center at a read depth of 20-110 per genome nucleotide.

Crude extracts and western blot analysis
Crude extracts of S. pombe WT and trm8Δ mutants were prepared by lysis with glass beads as described [79]. Then 25 μg of crude extract proteins were resolved on 4-20% SDS-PAGE gels (Criterion TGX, Bio-Rad), transferred to a 0.2 μm nitrocellulose membrane (Bio-Rad), and probed with antibodies as described [134], using anti-phosphorylated eIF2α (Cat. # 44-728G, Thermofisher; diluted 1:6000) and anti-α-tubulin (Cat. # T-5168 Sigma, diluted 1:6000).  trm8 dhp1-1 mutant. WT, trm8Δ, and trm8Δ dhp1-1 cells expressing P dhp1 dhp1 + or a vector were grown overnight in EMMC-Leu media at 30˚C, and analyzed for growth. (PDF) S10 Fig. A. S. pombe trm8Δ dhp1-3 and trm8Δ dhp1-4 mutants also restored tY(GUA) and tP(AGG) tRNA levels at 38.5˚C. Strains were grown in YES media at 30˚C and shifted to 38.5˚C for 8 hours, and RNA was isolated and analyzed by northern blotting as in Fig 2A.  Strains were analyzed for growth on YES media, EMM-C-His or EMMC-His media containing 10 mM 3-AT, as described in Fig 1B. Reconstructed S. pombe trm8Δ mutant was labeled as V3. (PDF) S13 Fig. A. Northern analysis of WT, S. pombe trm8Δ, trm8Δ gcn2-2, trm8Δ tif221-1, and  trm8Δ gcn2-3 cells. Strains were grown in YES media at 30˚C and shifted to 38.5˚C for 8 hours, and RNA was isolated and analyzed by northern blotting as in Fig 2A.  WT and S. pombe trm8Δ strains were grown in YES media at 30˚C, shifted to 38.5˚C for 8 hours and cells were harvested. Then crude extracts were prepared and analyzed by western blotting as described in Materials and Methods, using anti-phosphorylated eIF2α and anti-α-tubulin. Controls: WT and gcn2Δ mutants were grown at 30˚C in EMMC-His media and, where indicated, treated with 20 mM 3-AT. Then extracts were prepared and evaluated by blotting in parallel to the experimental samples. B. Increased levels of phophorylated eIF2α in S. pombe trm8Δ mutants at 38.5˚C were associated with increased expression of lys4 + and aro8 + mRNAs. Bulk RNA was prepared from the growth done for S17A Fig, and levels of aro8 + (SPAC56E4.03) and lys4 + mRNAs were quantified relative to act1 + , using RT-qPCR, as in Fig 5A. (PDF) S18 Fig. A. Analysis of charging levels of tY(GUA) or tP(AGG) in S. pombe trm8Δ mutants at 38.5˚C. Strains were grown in YES media at 30˚C and shifted to 38.5˚C, and samples were harvested after 8 hours. Then bulk RNA was isolated and resolved by denaturing PAGE under acidic conditions (to preserve tRNA charging), transferred, and then analyzed by hybridization as described in Materials and Methods. Control samples (WT and S. pombe trm8Δ mutants) were treated with 1 mM EDTA and 0.1 M Tris-HCl (pH 9.0) for 30 min at 37˚C to de-acylate the tRNA. b, base treated bulk RNA; Upper arrows, charged tRNA species; lower arrows with dashed lines, uncharged tRNA species. B. Quantification of tY(GUA) or tP(AGG) charging and tRNA levels. The percent charging was calculated as the ratio of aminoacylated species to the total for each tRNA. Relative levels of tP(AGG) and tY(GUA) were quantified as in Fig 2B, relative to the non-Trm8 substrate tL(UAA). (PDF) S19 Fig. A. Northern analysis of pre-tY(GUA) levels in S. pombe WT, trm8Δ, and trm8Δ  gcn2-1 mutants. The northern blot shown in Fig 5B was continued to analyze levels of pre-tY (GUA) species in WT and trm8Δ, and trm8Δ gcn2-1 mutants at different temperatures, using appropriate gene-specific probes (S5 Table) for the introns of the different tY(GUA) genes [59]. Cartoons at the right indicate exons, heavy bars; 5' leaders, 3' trailers, and introns, light bars. The primary pre-tY(GUA) transcript has 5' leader, 3' trailer, and intron, and the end-