Lack of 2'-O-methylation in the tRNA anticodon loop of two phylogenetically distant yeast species activates the general amino acid control pathway.

Modification defects in the tRNA anticodon loop often impair yeast growth and cause human disease. In the budding yeast Saccharomyces cerevisiae and the phylogenetically distant fission yeast Schizosaccharomyces pombe, trm7Δ mutants grow poorly due to lack of 2'-O-methylation of C32 and G34 in the tRNAPhe anticodon loop, and lesions in the human TRM7 homolog FTSJ1 cause non-syndromic X-linked intellectual disability (NSXLID). However, it is unclear why trm7Δ mutants grow poorly. We show here that despite the fact that S. cerevisiae trm7Δ mutants had no detectable tRNAPhe charging defect in rich media, the cells constitutively activated a robust general amino acid control (GAAC) response, acting through Gcn2, which senses uncharged tRNA. Consistent with reduced available charged tRNAPhe, the trm7Δ growth defect was suppressed by spontaneous mutations in phenylalanyl-tRNA synthetase (PheRS) or in the pol III negative regulator MAF1, and by overexpression of tRNAPhe, PheRS, or EF-1A; all of these also reduced GAAC activation. Genetic analysis also demonstrated that the trm7Δ growth defect was due to the constitutive robust GAAC activation as well as to the reduced available charged tRNAPhe. Robust GAAC activation was not observed with several other anticodon loop modification mutants. Analysis of S. pombe trm7 mutants led to similar observations. S. pombe Trm7 depletion also resulted in no observable tRNAPhe charging defect and a robust GAAC response, and suppressors mapped to PheRS and reduced GAAC activation. We speculate that GAAC activation is widely conserved in trm7 mutants in eukaryotes, including metazoans, and might play a role in FTSJ1-mediated NSXLID.

Introduction

During biogenesis, tRNAs acquire extensive post-transcriptional modifications that are important for their function as an adaptor molecule during translation. Modifications in the main body of the tRNA generally affect folding or stability of specific tRNAs [1-3], whereas modifications in and around the anticodon loop play crucial roles in translation, including promoting accuracy in charging [4, 5], reading frame maintenance [6-9] and decoding [10-13]. Indeed, modification is particularly extensive in the anticodon loop region comprising the loop itself and the 31–39 closing base pair, with an average of 2.72 modifications per eukaryotic cytoplasmic tRNA [14].

Defects in anticodon loop modification frequently lead to impaired growth in the yeast Saccharomyces cerevisiae and to a number of human disorders, particularly neurological disorders or mitochondrial syndromes [15, 16]. For example, yeast TAD2 and TAD3 are required for inosine modification of the wobble nucleotide A34 and are essential [10], and a mutation in the corresponding human ADAT3 gene is associated with intellectual disability and strabismus [17]. Similarly, yeast pus3Δ mutants have growth defects due to lack of pseudouridine (Ψ) at U38 and U39 and are temperature sensitive due to tRNAGln(UUG) [18], and a mutation in the corresponding human PUS3 gene is associated with syndromic intellectual disability and reduced pseudouridine [19]. In addition, yeast elongator mutants lacking the carbonylmethyl-U34 family of modifications (xcm5U34) have a number of phenotypes due to reduced function of two or three tRNA species [20-22], while Caenorhabditis elegans elongator mutants are associated with neurological and developmental dysfunctions [23], and human elongator mutations are linked to familial dysautonomia [24-26]. Although the molecular mechanisms linking tRNA modification defects to human diseases remain largely unknown, the causes are amenable to study in model organisms.

One such unsolved problem is why it is important for eukaryotes to have 2'-O-methylated C32 (Cm) and N34 (Nm) in their tRNAs, catalyzed by Trm7 family members. In S. cerevisiae, a trm7Δ mutant grows poorly due to reduced function of tRNAPhe, but not its other two substrates, tRNALeu(UAA) and tRNATrp(CCA), and in the phylogenetically distant yeast Schizosaccharomyces pombe, the near lethal phenotype of a trm7Δ mutant is rescued by overproduction of tRNAPhe [27-29]. In humans, seven different alleles of the human TRM7 homolog FTSJ1 have been linked to non-syndromic X-linked intellectual disability (NSXLID) [30-34], and lymphoblastoid cell lines (LCLs) derived from patients with two different FTSJ1 alleles had tRNAPhe with undetectable levels of Cm32 and Gm34 [34].

In eukaryotes, modification of tRNAs by Trm7 involves conserved partner proteins for each modification and a conserved circuitry for tRNAPhe anticodon loop modification. In S. cerevisiae, Trm7 interacts separately with Trm732 and Trm734 for formation of Cm32 and Nm34 respectively in each of its three tRNA substrates, and the presence of Cm32 and Gm34 in tRNAPhe drives the formation of wybutosine (yW) from 1-methylguanosine (m1G) modification at G37 [28]. S. pombe Trm732 and Trm734 have the same functions in Cm32 and Gm34 modification of tRNAPhe and, as in S. cerevisiae, Cm32 and Gm34 drive formation of yW37 in tRNAPhe [29]. Moreover, available evidence suggests that this circuitry is conserved in humans. tRNAPhe from patient LCLs with an FTSJ1 deletion or a splice site mutation had substantially reduced peroxywybutosine (o2yW37), as expected if o2yW37 formation is stimulated by Cm32 and Gm34 [34]. Furthermore, expression of either FTSJ1 or the TRM732 ortholog THADA complements the corresponding S. cerevisiae mutants, as does expression of S. pombe trm7 and the Drosophila TRM7 homolog ORF CG5220 [29].

However, despite the extensive studies of Trm7 in different organisms, the biological consequences of lacking Cm32 and Gm34 modifications on tRNAPhe remain unclear. We investigate here why Cm32 and Gm34 modifications are critical for tRNAPhe function and healthy growth in yeast. We provide evidence that despite the lack of an obvious charging defect, trm7Δ mutants activate a robust general amino acid control (GAAC) response in both S. cerevisiae and S. pombe, each in a manner suggesting the sensing of uncharged tRNA. Moreover, in each organism we find that suppressors of the trm7Δ growth defect frequently map to subunits of phenylalanyl tRNA synthetase (PheRS) and reduce the GAAC response toward that in wild type cells. These results argue for a conserved Trm7 biology in eukaryotes and argue that subtle changes in tRNAPhe charging have dramatic effects on cell physiology.

Results

Suppressors of the growth defect of S. cerevisiae trm7Δ mutants map to PheRS, despite the lack of an obvious charging defect

To begin to elucidate why Trm7 and 2’-O-methylation at C32 and N34 of tRNAs were important, we isolated and analyzed spontaneous suppressors that improved the slow growth phenotype of S. cerevisiae trm7Δ mutants. This slow growth phenotype is apparent by analysis of growth of trm7Δ [URA3 CEN TRM7] mutants on media containing 5-FOA [28], and by growth analysis of trm7Δ mutants on rich media and minimal media immediately after loss of the [URA3 CEN TRM7] plasmid (S1 Fig), and in all of these conditions, the growth defect is fully suppressed by overproduction of tRNAPhe (S1 Fig, [28]). We isolated 21 genetically independent faster growing suppressors after plating trm7Δ cells on YPD (rich) medium, and found that 19 of them had a dominant mutation in either FRS1 or FRS2 (S1 Table), which encode the two subunits of PheRS [35].

This result was surprising since we had shown previously that tRNAPhe from trm7Δ mutants had no obvious charging defects, and was present at similar overall levels in WT cells [28]. Indeed, analysis of tRNA isolated under acidic conditions to preserve charging [36, 37] showed that tRNAPhe from three independent freshly derived trm7Δ isolates grown in rich media had no discernible charging defect (65 ± 1% charging), compared to tRNAPhe from WT cells (67 ± 2%) or tyw1Δ mutants (66 ± 2%) (Fig 1A). tyw1Δ mutants, like trm7Δ mutants, have m1G37 instead of yW37 [38], and migrate identically on acidic gels. Similarly, no charging defect was seen in the other two Trm7 substrates, tRNALeu(UAA) and tRNATrp, or in the non-substrate tRNAGly(GCC). Furthermore, no increase in charging was observed in each of three representative suppressors of the trm7Δ growth defect (frs1-E415K, frs1-A549T, and frs2-L265V) for any of the tRNA species examined (65 ± 1%, 64 ± 0%, 65 ± 1% respectively for tRNAPhe). By contrast, in synthetic minimal medium a more prominent charging defect was observed by acidic Northern analysis of tRNAPhe from trm7Δ cells (Fig 1B). Under this growth condition, we found that tRNAPhe charging levels were reduced to 55 ± 0% in trm7Δ mutants, substantially below those of WT cells (68 ± 1%) and tyw1Δ mutants (77 ± 2%). Moreover, the three trm7Δ suppressors all restored tRNAPhe charging to levels similar to charging observed in tyw1Δ mutants (75 ± 2%, 74 ± 1%, 73 ± 1% respectively for the frs1-E415K, frs1-A549T, and frs2-L265V mutants).

Fig 1

tRNAs from trm7Δ mutants have no obvious charging defect when grown in rich media, and a prominent charging defect when grown in synthetic minimal media.

Strains as indicated were grown in rich media or minimal media, and RNA was isolated under acidic conditions to maintain tRNA charging, and analyzed by Northern blot as described in Materials and Methods, with hybridization probes as indicated. Control samples (WT and tyw1Δ) were treated with mild base to de-acylate the tRNA. Upper arrows denote charged tRNA species, and lower arrows with dashed lines denote uncharged tRNA species. Numbers below each sample indicate percentage of charged tRNA. Strains as indicated were grown overnight in minimal complete media at 30°C, washed with water once, diluted to OD600 of ∼0.5, serially diluted 10-fold in water, and 2 μL was spotted onto SD-Phe media containing different concentrations of phenylalanine as indicated, and incubated at 30°C for 4 to 5 days.

Consistent with a tRNAPhe charging defect in minimal media, we found that limiting phenylalanine exacerbated trm7Δ growth defects. After deletion of PHA2 (encoding prephenate dehydratase) to confer phenylalanine auxotrophy [39], we found that trm7Δ pha2Δ mutants showed an exacerbated growth defect compared to trm7Δ mutants in the presence of 10 mg/L phenylalanine, and this defect was complemented by re-introduction of the PHA2 gene on a plasmid; by contrast, under the same conditions pha2Δ mutants showed no discernable growth defect compared to a WT (trm7Δ [TRM7]) strain (Fig 1C).

S. cerevisiae trm7Δ mutants activate a robust general amino acid control response through Gcn2

Since acidic Northern analysis of trm7Δ mutants revealed no detectable tRNAPhe charging defect in rich media, but a distinct charging defect in minimal media that was suppressed by each of three suppressors, we examined in vivo charging in both rich and minimal media by analysis of the general amino acid control (GAAC) response [40]. In yeast and other eukaryotes, uncharged tRNAs arising from amino acid starvation or lack of functional tRNA synthetases bind to Gcn2 and activate its kinase domain, resulting in phosphorylation of eIF2α, de-repression of GCN4 translation, and transcriptional activation of nearly one tenth of the yeast genome, including numerous amino acid biosynthetic genes [41-43]. We reasoned that if there was a subtle accumulation of uncharged tRNAPhe in trm7Δ mutants, this might result in a GAAC response. Indeed, RT-qPCR analysis of mRNA from cell pellets collected from the same cultures as those used in the acidic Northerns (Fig 1A and 1B) revealed that the mRNA levels of two known GCN4 target genes, HIS5 and LYS1, were significantly increased in trm7Δ mutants (relative to ACT1), compared to WT cells. In rich media, relative levels of HIS5 and LYS1 mRNA increased 27.8-fold and 90.9-fold respectively, and in minimal media relative levels increased 17.1-fold and 43.2-fold (Fig 2A, S2 Table). These GAAC activation levels in trm7Δ mutants were comparable to those in WT His+ cells treated for 1 hour with 10 mM or 100 mM 3-amino-1,2,4-triazole (3-AT) (Fig 2B, S2 Table), a competitive inhibitor of His3 that has been used extensively to induce the yeast GAAC response [40, 44]. This robust constitutive GAAC response in trm7Δ mutants provided initial evidence that charged tRNA was limiting in vivo.

Fig 2

trm7Δ mutants grown in either rich media or synthetic minimal media activate a robust general amino acid control (GAAC) response, mediated by Gcn2.

( Strains as indicated were grown in either rich media or minimal media at 30°C to mid-log phase, and mRNA was isolated and analyzed by RT-qPCR. The mRNA levels of GCN4-regulated genes, LYS1 and HIS5, were normalized to those of the nonregulated ACT1, and then normalized to WT grown in rich media. ( WT and trm7Δ strains were grown in rich media to mid-log phase. For 3-AT treatment, the WT His+ strains were grown in synthetic complete media to mid-log phase, spun down, resuspended in SD-His media containing different concentrations of 3-AT for 1h, and then mRNA was isolated and HIS5 mRNA was analyzed by RT-qPCR, as in (A). ( Strains as indicated were grown in rich media and analyzed for relative HIS5 mRNA levels as in (A). ( Strains as indicated were grown overnight in rich medium, and analyzed by serial dilution, spotting to rich (YPD) or minimal media (SD complete), and incubation for 4 d at 25°C. ( Strains with deletions of TRM732, TRM734, or TYW1, or with combinations of deletions, were grown in rich media at 30°C to mid-log phase, and mRNA was isolated and analyzed for relative HIS5 levels as in (A).

Further analysis showed that the trm7Δ-mediated induction of the GAAC response is occurring through Gcn2, which senses uncharged tRNA [45]. The GAAC pathway can be induced by Gcn2 or by a pathway independent of Gcn2 [46-48], which is not well understood. As expected of the Gcn2-mediated GAAC response, we found that trm7Δ gcn2Δ strains completely abolished transcriptional activation of the HIS5 gene, as did the control trm7Δ gcn4Δ mutants (Fig 2C, S2 Table). These results provided compelling evidence that the GAAC response observed in trm7Δ mutants arose from uncharged tRNA. We note that the slow growth of S. cerevisiae trm7Δ mutants appears to be due to both lack of available charged tRNAPhe and to activation of the GAAC response itself, since either a gcn2Δ or a gcn4Δ mutation partially improved growth of a trm7Δ strain in both rich and minimal media (Fig 2D). Nonetheless, the increased stress on trm7Δ mutants associated with activation of the GAAC response must be a secondary consequence of the lack of available charged tRNAPhe required to initiate the response.

Further analysis demonstrated that activation of the GAAC response was closely tied to the growth phenotype of trm7Δ related strains. Thus, as measured by HIS5 mRNA levels, the GAAC pathway was not activated by the lack of Cm32 in a trm732Δ mutant, by lack of Nm34 in a trm734Δ mutant, or by lack of yW37 in a tyw1Δ mutant, or by a trm732Δ tyw1Δ double mutant or a trm734Δ tyw1Δ double mutant (Fig 2E, S2 Table), all of which are healthy strains [28]. By contrast, a trm732Δ trm734Δ strain fully activated the GAAC response, with relative HIS5 mRNA levels comparable to those of trm7Δ mutants, consistent with our previous finding that trm732Δ trm734Δ strains phenocopied the growth defect of trm7Δ mutants [28].

Suppressors of the S. cerevisiae trm7Δ growth defect reduce activation of the GAAC response

Strikingly, each of 18 trm7Δ suppressors we examined reduced the magnitude of the GAAC response from relative HIS5 mRNA levels of 37.3-fold and 36.1-fold in trm7Δ mutants (Fig 3A, left side and right side respectively, S3 Table) to levels approaching those observed in WT cells (1.1- to 5.5-fold). These suppressors included 16 with mutations in PheRS subunits, as well as two that did not have mutations in PheRS. The co-reversion of the trm7Δ growth defect and the GAAC response further implied that lack of available charged tRNA was the cause of the growth defect.

Fig 3

Most suppressors of the S. cerevisiae trm7Δ growth defect map to PheRS subunits and all reduce activation of the GAAC response.

( Strains were grown in rich media at 30°C to mid-log phase, and mRNA was isolated and analyzed for relative HIS5 mRNA levels. ( WT or trm7Δ [URA3 TRM7] strains containing a high copy [2μ LEU2] plasmid expressing FRS1, FRS2, both, or neither as indicated under P control were grown overnight in S-Leu medium containing raffinose and analyzed by spotting to synthetic media containing 5-FOA, raffinose (raff), and galactose (gal), and incubated for 3 d at 30°C. Strains from the 5-FOA plate were then purified on the same medium, grown overnight in S-Leu medium containing raffinose and galactose, diluted, and spotted to rich (YP) media and minimal (S- leu) as indicated, and incubated for 3 d at 30°C. ( Strains containing a [LEU2] plasmid expressing FRS1, FRS2, both, or neither as indicated under P control were grown in S-Leu medium containing raffinose and galactose, and then mRNA was isolated and relative HIS5 mRNA levels were determined.

Consistent with the interpretation that the poor growth of trm7Δ mutants is caused by defective charging, we found that overproduction of PheRS on a [P-FRS1 P-FRS2] plasmid improved trm7Δ growth, compared to that of a trm7Δ strain with a plasmid expressing either subunit of PheRS, or an empty vector (Fig 3B). Furthermore, overexpression of both FRS1 and FRS2 improved tRNAPhe charging (S2 Fig) and reduced relative HIS5 mRNA levels in trm7Δ mutants from 17.8 to 4.9, while overexpression of either FRS1 or FRS2 had no effect (Fig 3C, S3 Table).

Other genetic manipulations expected to modulate availability of charged tRNAPhe levels also suppress the growth defect and reduce GAAC activation

Since elongation factor 1A (EF-1A) binds to and delivers aminoacylated tRNA to the ribosomes A-site, we speculated that its overexpression might result in more charged tRNAPhe available for use in translation, thereby improving trm7Δ growth. To test this hypothesis, we introduced an extra copy of TEF1 or TEF2, which encode identical copies of EF-1A, into a trm7Δ strain. We found that elevated levels of EF-1A moderately rescued the growth defect (Fig 4A), and partially suppressed the GAAC activation, with p values of 0.012 and 0.055 respectively (Fig 4B, S4 Table), while deletion of TEF2 in a trm7Δ strain exacerbated the slow-growth phenotype (Fig 4C). The improved growth of trm7Δ mutants with increased EF-1A levels presumably reflects increased availability of aminoacylated tRNAPhe for translation after charging by PheRS, rather than increased tRNAPhe charging, which is not significantly altered (S3 Fig).

Fig 4

EF-1A levels commensurately affect both trm7Δ growth and activation of the GAAC response.

( trm7Δ mutants containing a [CEN LEU2] plasmid with TEF1, TEF2, or TRM7 as indicated were grown overnight in SD-Leu at 30°C, analyzed by spotting to plates as indicated, and incubated for 2 d at 30°C. ( WT or trm7Δ strains containing a [CEN LEU2] plasmid with TEF1 or TEF2 as indicated were grown as in (A), and relative HIS5 mRNA levels were determined. p values are shown above. ( WT and trm7Δ [CEN URA3 TRM7] strains with or without a tef2Δ mutation, as indicated, were grown overnight in rich media, and analyzed by spotting to media containing 5-FOA, and incubation for 2 d at 30°C. Strains from the 5-FOA plate were then purified on medium containing 5-FOA, grown overnight in YPD, diluted, spotted on rich (YPD) and minimal (SD complete) plates as indicated, and incubated for 2 d at indicated temperatures.

Similar results are obtained by treatments expected to increase the population of charged tRNAPhe. Thus, overexpression of tRNAPhe on a high copy plasmid, which is known to suppress the trm7Δ growth defect (S1 Fig, [28]) and results in 4.2-fold overproduction of tRNAPhe (S4A Fig), reduced the relative HIS5 mRNA levels to values similar to WT cells, while overexpression of other control tRNAs had no effect (Fig 5A, S5 Table). Furthermore, whole genome sequencing showed that trm7Δ suppressor 12 (Fig 3A) had a mutation in MAF1, a negative regulator of pol III transcription [49]. Since this maf1-C299Y mutation alters a highly conserved residue in the Box C region of Maf1, we inferred that this mutation behaved as a null mutation [49-51]. To test this inference, we introduced a MAF1 deletion into the trm7Δ [URA3 TRM7] strains and tested for growth on media containing 5-FOA to select against the URA3 plasmid. The resulting trm7Δ maf1Δ strain grew better than the control trm7Δ mutants (Fig 5B), and had increased levels of tRNAPhe and tRNAPhe charging, in both log phase and stationary phase (S4B Fig).

Fig 5

Manipulations expected to increase tRNAPhe levels link suppression of the trm7Δ growth defect with reduction of GAAC induction.

( reduces induction of the GAAC response. WT or trm7Δ strains containing a high-copy [2μ LEU2] plasmid expressing tF(GAA), tL(UAA), tW(CCA) or a vector as indicated were grown in SD-Leu at 30°C to mid-log phase, and mRNA was isolated and analyzed for relative HIS5 mRNA levels. ( WT and trm7Δ [CEN URA3 TRM7] strains with or without a maf1Δ mutation, as indicated, were grown overnight in rich media, and analyzed by spotting to media containing 5-FOA, and incubation for 2 d at 30°C. Strains from the 5-FOA plate were then purified on medium containing 5-FOA, grown overnight in YPD, diluted, spotted on plates as indicated, and incubated for 2 d at 30°C.

Robust activation of the GAAC response is not routinely observed in mutants defective in anticodon loop modifications

To determine if GAAC activation is a common theme among tRNA anticodon loop modification mutants, we examined the GAAC response in several other mutants, including strains lacking isopentenyladenosine (i6A37), due to a mod5Δ mutation [52]; 3-methylcytidine (m3C32), due to a trm140Δ mutation [53, 54]; the cm5U moiety of xcm5U34, due to a kti12Δ mutation [55]; the 2-thiouridine moiety (s2U) of mcm5s2U34, due to a uba4Δ mutation [56]; and Ψ38 and Ψ39, due to a pus3Δ mutation [57]. Among these mutants, only pus3Δ mutants had a substantial increase in relative HIS5 mRNA levels (15.7-fold), albeit much less than in trm7Δ mutants (116-fold in this experiment), whereas other modification mutants had only slightly increased HIS5 mRNA levels (1.8- to 3.2-fold) (Fig 6). Thus, robust GAAC activation, as observed in trm7Δ mutants, is not a general theme among modification mutants.

Fig 6

Robust GAAC activation in trm7Δ mutants is not a general theme among anticodon loop modification mutants.

Strains with trm7Δ, mod5Δ, trm140Δ, kti12Δ, uba4Δ, or pus3Δ mutations in anticodon loop modification genes, as indicated, were grown in rich media at 30°C to mid-log phase, and then mRNA was isolated and analyzed for relative HIS5 mRNA levels.

S. pombe Trm7 depletion and suppressors of the trm7Δ growth defect parallel charging and GAAC effects in S. cerevisiae

To investigate the evolutionary implications of our results, we examined the charging status and GAAC response in S. pombe trm7Δ mutants, which as in S. cerevisiae, grow poorly due to lack of sufficient tRNAPhe [29]. Since Sp trm7Δ strains are barely viable, we assayed tRNA charging and GAAC induction after growth of an Sp trm7Δ [P trm7+] strain in minimal (EMM) medium, followed by addition of thiamine to repress Sp Trm7 expression [29]. As in S. cerevisiae trm7Δ mutants grown in rich medium, we found that Sp tRNAPhe charging levels were comparable in Sp trm7Δ [P trm7+] grown in repressing conditions to deplete Trm7 (77.0 ± 2.6%), compared to the same strain in permissive conditions (79.3 ± 3.5%) or to WT strains (76 ± 5.6%) (Fig 7A). (Note that a similar S. pombe Trm7 depletion experiment could not be done in rich (YES) medium due to the presence of thiamine in this medium.) However, examination of mRNA from cell pellets collected in parallel from the same cultures revealed that Sp trm7Δ [P trm7+] strains grown in repressing conditions induced the GAAC response, with significantly increased relative mRNA levels of three Gcn2 dependent GAAC-regulated genes (lys4+, aro8+ (SPBC1773.13), and aro8+ (SPAC56E4.03)) [58], compared to WT cells (14.7-fold, 6.9-fold and 22.8-fold increase respectively); whereas Sp trm7Δ [P trm7+] strains grown under permissive conditions had relative mRNA levels very similar to WT cells. The GAAC induction levels in the Sp trm7Δ [P trm7+] strains grown in repressing conditions were similar to those when WT S. pombe cells were treated with 10 mM or 30 mM 3-AT for 4 hours (Fig 7B, S7 Table). Thus, depletion of Trm7 in S. pombe resulted in little, if any, detectable defect in tRNAPhe charging, but a robust induction of the GAAC pathway.

Fig 7

S. pombe strains depleted of Trm7 have no detectable tRNAPhe charging defect but induce a robust GAAC response, which is reduced in trm7Δ suppressors.

( ] strains grown under repressive conditions have no obvious tRNA charging defect. S. pombe trm7Δ [P trm7] strains were grown in EMM with thiamine (repressive conditions) or without thiamine, along with WT S. pombe grown in EMM, and then RNA was isolated under acidic conditions and analyzed for charging as in Fig 1(. ( ] strains grown under repressive conditions induce the GAAC response. Left side: S. pombe WT and trm7Δ [P trm7] strains were grown in EMM with thiamine (repressive conditions) or without thiamine to log phase and then mRNA was isolated and analyzed by RT-qPCR for mRNA levels of GAAC-regulated genes, lys4, aro8 (SPBC1773.13), and aor8 (SPAC56E4.03), normalized to those of nonregulated act1, and then normalized to WT without thiamine. Right side: GAAC induction of WT cells treated with different concentrations of 3-AT as indicated for 4 hours, and evaluated in parallel to the Left side. ( S. pombe strains as indicated were grown in EMM as in (A). mRNA was isolated from each strain and analyzed by RT-qPCR as in (B) for relative mRNA levels.

As in S. cerevisiae, S. pombe mutants lacking Cm32 or Gm34 of tRNAPhe have GAAC responses that tracked with the growth defect. Sp trm734Δ strains grow relatively poorly [29], but not nearly as poorly as trm7Δ mutants, and had a partially activated GAAC response, with 6.5-fold and 6-fold increased relative mRNA levels of lys4+ and aor8+ (SPAC56E4.03) respectively, compared to 9.0-fold and 12.9-fold for Sp trm7Δ [P trm7+] strains grown under repressive conditions. By contrast, Sp trm732Δ mutants have no obvious growth defect at 30°C—37°C [29], and had near wild type relative mRNA levels for lys4+ and aor8+ (SPAC56E4.03) (Fig 7C, S7 Table).

Further analysis showed that rescue of the growth defect of S. pombe trm7 mutants reduced the GAAC response toward WT levels. As in S. cerevisiae, overproduction of Sp tRNAPhe reduced the GAAC response as measured by relative mRNA levels of lys4+ and aor8+ (SPAC56E4.03) (Fig 7C, S7 Table), consistent with the rescue of the Sp trm7Δ growth defect we previously observed [29]. Furthermore, suppressors of the Sp trm7Δ growth defect behaved as in S. cerevisiae. We isolated Sp trm7Δ suppressors by plating Sp trm7Δ [P trm7+] cells on media containing FOA, and each of two suppressors we analyzed had mutations in PheRS, and the suppressor strains in each case reduced the induction of the GAAC response (Fig 7C, S7 Table). These results suggest that lack of sufficient charged tRNAPhe is also the main problem causing slow growth in Sp trm7Δ cells, despite the lack of detectable charging defect in acidic Northerns.

Based on the conserved induction of the GAAC response in S. cerevisiae and S. pombe trm7Δ mutants, we examined human lymphoblastoid cell lines with mutations in FTSJ1 for an induced GAAC response by measuring mRNA levels of two Gcn2 dependent GAAC-regulated genes, CTH and GADD153 [59]. Although WT control cell lines treated with the prolyl tRNA synthetase inhibitor halofuginone induced a significant GAAC response (S5A Fig) [60, 61], we obtained equivocal and inconclusive results for GAAC induction in the human lymphoblastoid FTSJ1 cell lines, compared to the WT cell lines (S5B Fig).

Discussion

Although standard acidic Northern analysis did not reveal significant reduced tRNAPhe charging in S. cerevisiae trm7Δ mutants grown in rich media, we provided four lines of evidence supporting the conclusion that the growth defect of trm7Δ mutants is caused by reduced available charged tRNAPhe. First, contrary to our observations in rich media, in minimal media acidic Northern analysis revealed distinctly reduced tRNAPhe charging in trm7Δ cells, charging was restored in each of three suppressors analyzed, and limiting phenylalanine exacerbated the trm7Δ growth defect. Thus, it seemed plausible that there was a subtler charging defect in rich media. Second, trm7Δ mutants activated a robust GAAC response in both rich media and minimal media, and activation of the GAAC response in rich media depended on Gcn2, which is known to sense uncharged tRNA [45, 62]. Third, each of 18 tested trm7Δ suppressors isolated in rich media suppressed the activation of the GAAC response found in trm7Δ mutants, and the vast majority had mutations that mapped to PheRS, arguing for the importance of increased charging for suppression of both the trm7Δ growth defect and the GAAC activation. Fourth, overproduction of PheRS also suppressed both the trm7Δ growth defect and GAAC activation, further implying that more charged tRNAPhe could overcome the phenotypes of trm7Δ mutants. Ascribing the trm7Δ growth defect to reduced tRNAPhe charging is also consistent with our previous observation that the steady state levels of tRNAPhe were normal in trm7Δ mutants [28].

The effects of manipulation of EF-1A, MAF1, or tRNAPhe gene dosage on suppression of S. cerevisiae trm7Δ phenotypes can also be interpreted in terms of tRNAPhe charging or availability. The rescue of both the trm7Δ growth defect and GAAC activation by an extra copy of TEF1 or TEF2 could be due to the increased availability of the EF-1A:phe-tRNAPhe complex for the translation machinery, achieved by increased overall binding of charged tRNAPhe to EF-1A relative to PheRS, due to the tight binding constant of EF-1A for charged tRNA [63], or by preventing spontaneous deacylation of charged tRNAPhe not bound by EF-1A, as demonstrated for EF-Tu [64]. The rescue of both the trm7Δ growth defect and GAAC activation by a maf1 mutation is likely due to the observed increase in tRNAPhe levels, consistent with the role of Maf1 as a negative regulator of pol III [49, 65], resulting in more charged tRNAPhe. Similarly, the rescue of both the trm7Δ growth defect [28] and GAAC induction by overexpression of tRNAPhe is due to the 4.2-fold increase in tRNAPhe, and the commensurate increase in charged tRNAPhe. We note that there is also an increase in the ratio of charged:uncharged tRNAPhe that occurs when tRNAPhe is overexpressed or in a maf1Δ mutation; this likely results from the decreased relative usage of tRNAPhe during translation when it is overproduced. We also note that the increase in uncharged tRNAPhe that occurs when tRNAPhe is overexpressed or in a maf1Δ mutation does not provoke the GAAC response. This result is consistent with the prevailing model that Gcn2 activation occurs in concert with the Gcn1-Gcn20 complex at the ribosome, triggered by entry of uncharged cognate tRNA at the A site independent of EF-1A [66-68]. Based on this model, the increased pools of charged tRNA would effectively outcompete the increased pool of uncharged tRNA for binding at the A-site when both are available, thus preventing activation of the GAAC response.

We have also shown that depletion of Trm7 in S. pombe resulted in a severe growth defect, and induced a robust GAAC response with no obvious alteration of tRNAPhe charging as measured by acidic Northerns, and that suppressors of the growth defect reduced induction of the GAAC response and mapped to PheRS. Since the genes we assayed respond to the GAAC pathway when it is activated by uncharged tRNA, but not by other stimuli [58], we infer that S. pombe trm7Δ mutants, like S. cerevisiae trm7Δ mutants, behave as if they have uncharged tRNA.

It is intriguing that there was no discernible tRNAPhe charging defect detected in acidic Northerns from S. cerevisiae trm7Δ mutants grown in rich media and in S. pombe trm7Δ [P trm7+] mutants grown in repressing conditions, whereas mRNA levels analyzed from the same cultures showed robust induction of the GAAC response. There are at least three reasonable explanations of this observation. First, the tRNAPhe charging defects may be too subtle to be detected by the acidic Northern assay, but can be effectively captured by the sensitive GAAC response. Acidic Northern analysis has been used extensively to measure charging since its initial description [36, 69]. However, quantification of uncharged tRNA might be particularly difficult for tRNAPhe because of the higher background of uncharged tRNAPhe in most RNA preps (Fig 1A; [70, 71]) and because of the possibility of incomplete yW modification of tRNAPhe in WT cells grown in different conditions [72], which could interfere because of small mobility differences between uncharged tRNAPhe with yW, and charged tRNAPhe without yW. In this regard, it is not clear how much uncharged tRNA in the cell is required to activate the GAAC response for a given tRNA species [40]. Second, it is possible that tRNAPhe is efficiently charged in vivo, but is sequestered from use in translation by some tRNA binding proteins, resulting in an increased probability that uncharged tRNAPhe will bind at the ribosome A site and trigger the GAAC response. The tRNAPhe might be sequestered in the nucleus by retrograde tRNA nuclear import [73, 74] or as an Msn5:EF-1A:phe-tRNAPhe complex [75] somehow triggered by lack of the modifications, or perhaps sequestered in a stress granule [76]. However, it seems unlikely that charged tRNAPhe is sequestered by binding as a product to PheRS, since overproduction of PheRS suppresses the growth defect and the GAAC response. Whether tRNAPhe is subtly undercharged or is charged but effectively sequestered, the concordance of the trm7Δ growth defect and the robust GAAC response is striking in both S. cerevisiae and S. pombe, and suggests that they have the same root cause: lack of available charged tRNAPhe. A third explanation is that lack of Trm7 modifications causes ribosome stalling independent of uncharged tRNA, as reported in mouse mutants deficient in tRNAArg(UCU) and GTPBP2, a ribosome rescue factor [77].

The finding that a gcn2Δ or a gcn4Δ mutation partially improved growth of an S. cerevisiae trm7Δ strain indicates that some combination of the massively re-programmed expression pattern during the GAAC response [78] increases the stress on the trm7Δ mutants. This interpretation is consistent with models suggesting that constitutive activation of the GAAC response is deleterious to yeast [79, 80], as it may also be in metazoans based on the observation that inactivation of the GAAC response relieves TDP-43 toxicity in Drosophila and in mammalian neurons [81].

The GAAC activation we observed in S. cerevisiae trm7Δ mutants was more robust than each of the other anticodon loop modification mutants tested. The much more modest GAAC activation found in kti12Δ or uba4Δ mutants was very similar in magnitude to the Gcn2-independent GAAC activation found previously for disruption of the same mcm5s2U34 modification [82], and a similar modest GAAC activation was also detected in mod5Δ and trm140Δ mutants. These more modest GAAC activation levels are associated with mutants that have no obvious growth defect under these conditions; by contrast, the more substantial GAAC induction found in pus3Δ mutants is consistent with the known growth defect of pus3Δ mutants [18, 57]. Since a pus3Δ mutation impairs function of at least 3 of its 19 or more tRNA substrates in S. cerevisiae [18], it is possible that more than one tRNA is responsible for the GAAC induction. The extent of GAAC induction observed in these anticodon loop modification mutants is consistent with a recent study on transcriptome-wide analysis of roles for tRNA modifications by ribosome profiling [83].

It is unclear from our results why the frs1 or frs2 mutations that we identified from S. cerevisiae trm7Δ suppressors were all genetically dominant. Dominant PheRS mutations would be expected if trm7Δ mutants had a charging defect, since gain of function mutations are expected to be dominant. However, the frs1 mutations map throughout the body of the protein, based on the human PheRS structure [84], bringing up the question of how scattered mutations all improve the function of the synthetase in trm7Δ mutants. As none of the frs1 mutations localized to the editing domain, it is unlikely that these PheRS mutations reduce PheRS editing to inhibit GAAC induction, as observed for an frs1 editing mutant grown under conditions of excess tyrosine relative to phenylananine [70, 71]).

Two models of PheRS function could explain the widespread locations of dominant frs1 mutations among the S. cerevisiae trm7Δ suppressors. First, PheRS function could be reduced in trm7Δ mutants because of decreased recognition and binding of the hypomodified tRNA to PheRS, in which case the scattered frs1 gain-of-function mutations would all act to improve interactions with tRNA. This model is plausible, and consistent with the principle of weak binding for efficient catalysis [85]. It is also formally possible in this model that the frs1 mutations improve interaction between the two PheRS subunits, or that they improve stability or expression of the PheRS subunits, but these possibilities seem less likely to us because the mutations map all over the Frs1 subunit. Second, the charging activity of PheRS could be reduced in trm7Δ mutants because of increased binding of PheRS to hypomodified phe-tRNAPhe and the consequent slow release of product, reducing the rate of multiple turnover reactions. Although in bacteria rate limiting product release is found in class I synthetases rather than class II synthetases like PheRS [86], this mechanism is in principle plausible for eukaryotic PheRS acting on tRNAPhe lacking 2'-O-methylation. In this case, the scattered gain-of-function frs1 mutations would all reduce interactions between PheRS and hypomodified tRNAPhe, promoting more effective release of charged tRNAPhe from PheRS and increased overall charging. Both of these models call for specific interactions between the anticodon loop and PheRS, consistent with the known PheRS recognition of G34 [87], but the effects of Cm32 and Gm34 have not been tested [88, 89].

It is remarkable that in both S. pombe and S. cerevisiae the poor growth of trm7Δ mutants is associated with apparently complete tRNAPhe charging but a robust GAAC response. Since these species diverged ∼330 to 420 million years ago [90], this result implies its generality among eukaryotes, to go along with the previously established conserved importance of tRNAPhe as a Trm7 substrate in S. pombe and S. cerevisiae, the conserved anticodon loop modification circuitry of tRNAPhe in S. pombe, S. cerevisiae, and humans, and the conserved favored importance of Gm34 in S. pombe and humans [28, 29, 34]. Moreover, all eukaryotic PheRS species appear to have similar recognition sets, since human and S. cerevisiae PheRS each recognize the same five residues [91], and tRNAPhe from wheat germ or S. pombe is charged by S. cerevisiae PheRS nearly as effectively as the native substrates [92]. Although our preliminary analysis of the GAAC response in human lymphoblastoid cell lines with mutations in FTSJ1 yielded equivocal results, this analysis might require specialized cell types to explain the non-syndromic nature of NSXLID [30-34], or the cell lines may have accumulated secondary lesions that mask the GAAC induction. Based on this high degree of conservation of the biology of Trm7 and PheRS, we speculate that GAAC activation will be widely conserved in trm7 mutants in eukaryotes, including metazoans, and might play a role in NSXLID due to lesions in human FTSJ1.

Materials and methods

Yeast strains

Yeast strains used in this study are listed in S1 and S8 Tables. trm7Δ supp 1 to 10 were isolated from yMG105 (MATa, trm7Δ::ble) strain, and supp 11 to 21 from yMG107 (MATα, trm7Δ::ble) strain. For all other experiments, trm7Δ mutants were freshly derived from yMG348-1 trm7Δ::ble [CEN URA3 TRM7] each time before use, by growing yMG348-1 in YPD media overnight followed by streaking on media containing FOA to select against the URA3 plasmid. The WT His+ strains were derived from BY4741 by PCR amplification of HIS3 with its 5' and 3' flanking sequence, followed by linear transformation, selection on SD-His and PCR verification. PHA2, GCN2, KTI12, UBA4, MAF1, and TEF2 were deleted by PCR amplification of DNA from the appropriate YKO collection kanMX strains using oligomers containing sequences 5' and 3' of the gene [93], followed by linear transformation and selection on YPD media containing 300 mg/L geneticin. GCN4 was deleted by PCR amplification of the hyg marker, followed by linear transformation and selection on YPD media containing 300 mg/L hygromycin B. All trm7Δ double-mutant strains were constructed similarly by PCR amplification, linear transformation into yMG348-1 trm7Δ::ble [CEN URA3 TRM7], and selection against the URA3 plasmid by streaking on media containing FOA.

The haploid S. pombe trm7Δ::kanMX [ura4 P trm7] (yMG1052A) strain was generated as previously described [29] and used for isolation of suppressors. The haploid trm7Δ::kanMX [LEU2 P trm7] (yMG1541) strain was generated by transformation of yMG1052A with a LEU2 P sp trm7 plasmid (pMG527B), and selection against the ura4 plasmid by streaking on media containing FOA, and was used for experiments in which Trm7 was depleted with thiamine.

For all experiments in which two or more strains with the same genotype are analyzed, these samples are biological replicates.

Plasmids

Plasmids used in this study are listed in S9 Table. Plasmids for FRS1 and/or FRS2 expression were derived from pBG2619, which is a [2μ P LIC] dual ORF expression plasmid. In this plasmid, expression of one ORF is under P control with a C-terminal PT tag, containing 3C site-HA epitope-His6-ZZ domain of protein A, and expression of the second ORF is under P control with no tag [94]. CEN plasmids were constructed by ligation-independent clone (LIC) of genes containing their own 5' and 3' flanking sequence into pAVA581 (LEU2) or pAVA579 (URA3) [94].

Northern blot analysis

S. cerevisiae strains were grown at 30°C to mid-log phase in either rich media or minimal media as indicated. S. pombe strains were grown at 30°C to mid-log phase in EMM supplemented with 225 mg/L adenine, lysine, histidine, leucine, and uracil. To analyze WT and trm7Δ [P trm7] strains under repressive growth conditions, thiamine was supplemented to the media at 5 mg/L. For either S. cerevisiae or S. pombe, bulk RNA was prepared from ~4 OD pellets using glass beads, and RNA was resolved on acrylamide gels and analyzed by hybridization as previously described [3]. For analysis of charging, RNA was prepared and resolved under acidic conditions as described [3].

Real-time quantitative PCR

Strains were grown in triplicate to mid-log phase as described above for Northern blot analysis. Bulk RNA was prepared from 5–10 OD pellets using glass beads, treated with DNase, reverse transcribed, and the resulting cDNA was amplified and analyzed as previously described [95].

Whole genome sequencing

Whole genome sequencing was done at the UR Genomics Research Center at a read depth of greater than 100-reads.

Overproduction of tRNAPhe fully rescues trm7Δ growth defect on both rich and minimal media.

WT or trm7Δ strains containing a high-copy LEU2 plasmid expressing TRM7, tRNAPhe, tRNATrp, tRNALeu(UAA), or a vector as indicated were grown in SD-Leu, analyzed by spotting to plates as indicated, and incubated for 2 d at 30°C.

(TIF)

Click here for additional data file.

Overproduction of PheRS restores tRNAPhe charging levels in trm7Δ mutants to WT levels.

WT and trm7Δ strains containing a high-copy [2μ LEU2] plasmid expressing FRS1and FRS2 under control of the P promoter, or a vector control, were grown in S-Leu medium containing raffinose and galactose, and then RNA was isolated under acidic conditions and analyzed for charging as in Fig 1(.

(TIF)

Click here for additional data file.

An additional copy of EF-1A does not alter tRNAPhe charging in trm7Δ mutants.

WT or trm7Δ strains containing a [CEN LEU2] plasmid expressing TEF1 or TEF2, or a vector control, were grown in SD-Leu, and then RNA was isolated under acidic conditions and analyzed for charging as in Fig 1(.

(TIF)

Click here for additional data file.

Overproduction of tRNAPhe or a maf1 deletion results in increased levels of tRNAPhe and tRNAPhe charging in trm7Δ mutants.

(A) WT, trm7Δ, or tyw1Δ strains containing a high-copy [2μ LEU2] plasmid expressing tF(GAA), or a vector control, were grown in SD-Leu, and then RNA was isolated under acidic conditions and analyzed for charging as in Fig 1(. a, b, 1.5 μg RNA analyzed; b/4, 0.375 μg analyzed. Relative expression of tF(GAA) represents tF(GAA) expression normalized to that of tG(GCC), and then normalized to expression in WT [vec], itself normalized to tG(GCC). WT, trm7Δ, and tyw1Δ strains overexpressing tF(GAA) have 3.7-, 4.2-, and 2.6-fold more tRNAPhe respectively than the corresponding vector control strains. (B) Strains as indicated were grown in minimal (SD complete) media to log phase or stationary phase, and then RNA was isolated under acidic conditions and analyzed for charging as in Fig 1(.

(TIF)

Click here for additional data file.

GAAC induction in human lymphoblastoid FTSJ1 cell lines is inconclusive.

(A) A WT control cell line treated with halofuginone induces a significant GAAC response. A WT control cell line was grown as previously described [34] and then treated with halofuginone (HF) at indicated concentrations for 4 hours. Bulk RNA was then extracted, and analyzed by RT-qPCR for mRNA levels of Gcn2-dependent GAAC-regulated genes, CTH and GADD153, normalized to those of nonregulated GAPDH. (B) GAAC induction in human lymphoblastoid FTSJ1 cell lines. WT control cell lines and FTSJ1 cell lines as indicated [34] were examined for GAAC induction as in A.

(TIF)

Click here for additional data file.

frs1 and frs2 mutations identified in trm7Δ suppressors.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 2.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 3.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 4B.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 5A.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 6.

(PDF)

Click here for additional data file.

Relative mRNA levels in Fig 7.

(PDF)

Click here for additional data file.

Strains used in this study.

(PDF)

Click here for additional data file.

Plasmids used in this study.

(PDF)

Click here for additional data file.