Reciprocal regulation of TORC signaling and tRNA modifications by Elongator enforces nutrient-dependent cell fate

INTRODUCTION

Fission yeast sexual differentiation has evolved as a stress response, and almost any sudden change in growth conditions induces some cells to accelerate mitosis, become smaller, and enter gametogenesis, even in chemostat cultures (1). This is also observed on plated colonies where the heterogeneity of nutrient availability generates regions more or less conducive to differentiation, which is conveniently induced under laboratory conditions by altering, limiting, or withdrawing the source of nitrogen (2). More generally, in cultured cells or in whole organisms, the critical cell size for mitosis is determined by the growth conditions (3, 4). TOR (target of rapamycin) signaling is a conserved central mediator of this process. TOR regulates nutrient uptake, gene expression, protein synthesis, and metabolic pathways to ensure that the balance between cell growth and cell division is coordinated with nutrient availability. The process is evolutionarily conserved and notably shown in flies grown under nutrient-limiting conditions that develop to half of the size of well-fed animals without a reduction of the absolute cell number (3, 4). In fission yeast (Fig. 1A), TORC1 (TOR complex 1) defined by the Tor2 kinase and the raptor homolog Mip1 is sensitive to rapamycin, promotes cell growth, and inhibits entry into mitosis while repressing autophagy and cell differentiation. TORC2 defined by the Tor1 kinase and the Rictor homolog Ste20 promotes mitosis and is essential for differentiation (5–11). Consequently, TORC1 mutants uncouple nutrient availability from growth and have an “always starved” phenotype. These mutants eventually arrest as small rounded cell even when nitrogen is abundant. In contrast, TORC2 mutants uncouple starvation from gametogenesis and have a “never starved” phenotype resulting in sterility when nitrogen drops. On the basis of genetic analyses and the antagonistic relationship between TORC1 and TORC2, the concept of feedback loops connecting both complexes was suggested early on (12) but lacked mechanistic support until recently (13, 14).

Transfer RNAs (tRNAs) from all organisms contain modified nucleosides, and when a uridine is present in the wobble position (U₃₄), this residue is almost universally modified (15–17). In particular, the eukaryotic cytosolic tRNA^Lys_UUU is thiolated (s²) by the Ctu1 enzyme at the 2-carbon and contains the most critical methoxy-carbonyl-methyl (mcm⁵) modification at the 5-carbon on the wobble uridine, which is required to offset the translational inefficiency resulting from low effective stacking interactions of the A-U base interaction (18–20). The six-subunit Elongator complex synthetizes the cm⁵ moiety, which becomes a substrate for the Trm9 methylase to generate mcm⁵U (fig. S1A) (21–23). The catalytic subunit of Elongator is Elp3. Two copies of the Elp1-Elp2-Elp3 subunits form a two-lobe symmetric scaffold, which asymmetrically binds Elp4-Elp5-Elp6. In all species where it has been investigated, the absence of any subunits results in the inactivation of the whole complex, and mutants show an extremely diverse range of phenotypes (24). In fission yeast, a major phenotype resulting from the inactivation of Elongator is a cell cycle defect, and the cdr2 mRNA that encodes a key regulator of mitosis and cytokinesis was shown to harbor a skewed codon content for lysine, which places it under translational control by Elongator (25, 26). In the present study, we report a reciprocal regulation of TOR complex (TORC) signaling and Elongator, showing that tRNA modification by Elongator is at the nexus of cell growth and cell differentiation.

RESULTS

Our previous proteome-wide analysis of a fission yeast strain lacking the Elp3 subunit of Elongator revealed that four distinct functional Gene Ontology (GO) groups of proteins were down-regulated (25–27). Analysis of the codon usage within the corresponding mRNAs showed a bias for the AAA codon over AAG when encoding lysine, which rendered their expression highly dependent on the modification by Elongator (26). A detailed analysis of one GO group related to the response to nutrients and the TOR signaling pathway highlighted the presence of many autophagy-dependent (Atg) proteins, with autophagy controlled by the TOR pathway (Fig. 1B). In addition, the expression of positive effectors of TORC2 (Tor1, Ste20^Rictor, and Gad8) and inhibitors of TORC1 (Tsc1, Tsc2, Npr2, and Npr3) was also down-regulated in the absence of Elp3 (Fig. 1C and fig. S1B). The overexpression of the unmodified tRNA^LYS_UUU was sufficient to restore wild-type level of Ste20^Rictor and Tsc2 (Fig. 2A) while suppressing the rapamycin sensitivity of the elp3 mutant (Fig. 2B), supporting a close connection bewteen Elongator and the TORC1/TORC2 pathways. A broader phenotypic analysis showed that the sensitivity of the elp3 mutant to rapamycin was mediated through the Fkh1 protein and through the TORC1 complex, as a rapamycin-resistant tor2 allele (tor2 S1873E) completely suppressed the sensitivity while the corresponding rapamycin-resistant tor1 allele (tor1 S1834E) did not (Fig. 2B). This set of data suggested that the sensitivity of Elongator mutants to rapamycin resulted from the increased activity of TORC1 and unbalanced TORC signaling, as supported by the fact that the overexpression of tor2 also increased the sensitivity to rapamycin (Fig. 2B). In addition, the absence of Elongator resulted in resistance to canavanine (Fig. 2C), similar to the overexpression of tor2, or the deletion of tsc1, a gene encoding hamartin that negatively regulates TORC1 (Fig. 2C). Auxothrophic elp3 mutants grew poorly on supplemented minimal media, suggesting a deficient nutrient uptake that is also typical of unbalanced TOR signaling in benefit of TORC1 (Fig. 2D). Together, these results support that Elongator acts as a positive regulator of TORC2 and a negative regulator of TORC1 through translational controls of components of these complexes and may therefore regulate the balance between both complexes.

To test this possibility, we analyzed the behavior of the Elongator mutant during nutrient shifts between efficient (glutamate) and poor (proline) nitrogen sources, which alters the TORC1/TORC2 balance (28–30). Wild-type cells shifted from glutamate to proline were advanced into mitosis, as shown by a peak in the percentage of dividing cells about 1 hour after the nutritional shift (Fig. 3A). The proportion of dividing cells remained largely constant in an Elongator mutant (elp4 deletion), indicating that the advancement into mitosis was not efficient when Elongator was inactivated. Rapamycin was previously shown to mimic the nutritional shift-down (30). The use of rapamycin-resistant versions of tor1 and tor2 indicated that the mitotic advancement likely occurred by inhibiting TORC1 for the benefit of TORC2, as rapamycin had no effect on a strain harboring the rapamycin-resistant tor2 S1837E allele (Fig. 3B). In the reverse experiment where cells were shifted from the poor nitrogen source proline to glutamate, a drop in the percentage of dividing cells was observed within 1 hour, indicating a delayed onset of mitosis, which restores TORC1-dependent promotion of growth. While the elp4 mutant barely affected this response, cells remained delayed about 30 min longer, suggesting TORC1 hyperactivation (Fig. 3C). These data support the possibility that the Elongator complex is a critical regulator of the balance between the TORC1 and TORC2 complexes when environmental cues controlling the commitment to mitosis are changing.

To identify regulators of the activity of Elongator, we next set up a genetic screen based on dual-color luciferase reporters harboring skewed codon content for lysine. We generated a strain expressing the green-emitting click beetle luciferase (CBG) expressed from the ura4 locus and the red-emitting click beetle luciferase (CBR) expressed from the leu1 locus (31), which is suitable for Chroma-Glo assays carried out directly in the culture medium (fig. S2A). The natural codon bias for lysine (CBR: 35% Lys^AAA/65% Lys^AAG and CBG: 54% Lys^AAA/46% Lys^AAG) was sufficient to render their expression sensitive to the absence of Elongator, with CBG being the most affected. However, CBR was also affected and therefore was not suitable as internal control. We therefore modified the reporter strain by using a CBG_AAG version, where all lysine codons are encoded by the AAG codon, and a CBR_AAA version, where all lysine codons are encoded by the AAA codon. In addition, CBG_AAG was fused to the Ade3 protein, which we have shown to be expressed independently of Elongator, and CBR_AAA was fused to the Cdr2 protein, which we have shown to require Elongator for efficient translation (26). Both reporters were stably expressed, and the steady-state levels of their mRNAs were independent of the elp3, ctu1, and trm9 genes required for the synthesis of the mcm⁵s²U modification, while the bioluminescence of CBR_AAA, but not CBG_AAG, was affected in the absence of these three genes (fig. S2, B to D). The reporter strain was crossed to the fission yeast deletion library, and the resulting strains were screened for altered dual-color luminescence ratio (fig. S3A). An increased or decreased CBR/CBG ratio was found in 417 clones, suggesting that the translation of the reporter of Elongator activity can be either up- or down-regulated (table S1). Considering this large number of candidates and the possibility of cell cycle effects resulting from using the Cdr2 fusion, the positive clones were screened for altered expression of Atf1, another well-characterized but unrelated target of Elongator (32). The set of 100 clones that behaved similarly in both screens was further analyzed. We found that the tsc1, tor1, and ekc1 genes encoding key players of the TOR pathways were recovered in both screens, and this list was further extended to tsc2, npr2, ste20, and hhp2 when considering only screen 1 (fig. S3B). The casein kinase Hhp2 (Hrr25) and the Ekc1 (Sit4) phosphatase subunit were previously shown to be required for the activity of Elongator in budding yeast (fig. S3C) (23, 33), further validating the screens. The fact that inhibitors of TORC1 (Tsc1, Tsc2, and Npr2) and components of TORC2 (Tor1 and Ste20) were identified as positive regulators of Elongator suggested that, while Elongator is required for proper balance between TORC1 and TORC2, TORCs could in turn regulate the synthesis of the modification by Elongator (fig. S3C).

To explore this possibility, we first adapted the previously described liquid chromatography–coupled mass spectrometry (LC-MS) quantitative approach (34) to fission yeast. Analysis of the elp3 and ctu1 single mutants and the double elp3/ctu1 mutant confirmed our previous work (35) showing the loss of canonical Elongator-dependent modifications in the absence of elp3 and the specific loss of s² when ctu1 is deleted, while a set of unrelated modifications was barely affected (Fig. 4A). We next focused on the complete set of detectable modified nucleosides dependent on Elongator activity (fig. S1A) and analyzed them in strains overexpressing either tor1 or tor2 or grown in the presence of rapamycin. This experiment revealed that the inhibition of TORC1 by rapamycin or the overexpression of the TORC2 kinase Tor1 resulted in an increase in the cm⁵U and cm⁵Um nucleotides, while the overexpression of the TORC1 kinase Tor2 had the opposite effect (Fig. 4A and fig. S4, A to C). These data indicate that the artificial manipulation of TOR signaling affects the level of the Elongator-dependent modifications, with TORC1 and TORC2 showing an opposite effect. The increased levels of cm⁵U or cm⁵Um are reminiscent of the deletion of trm9, suggesting that the methylase may become limiting when Elongator activity is increased.

Phosphoproteome-wide analyses in budding yeast have highlighted that the Elp4 subunit of Elongator undergoes rapamycin-dependent phosphorylation on serine-221 (36). The same region in fission yeast Elp4 (fig. S5A) was reported to be phosphorylated in two independent large-scale analyses (37, 38). In this latter case, the targeted sites are S114 and S118 in the S¹¹⁴S¹¹⁵P¹¹⁶P¹¹⁷S¹¹⁸ sequence, which are conserved within the fission yeast clade (fig. S5B). In addition, the deletion of the elp4 gene in Schizosaccharomyces japonicus also resulted in increased sensitivity to rapamycin, which supports a conservation of the TOR-Elongator connection (fig. S5C). Serine-114 constitutes a glycogen synthase kinase 3 (Gsk3) consensus site, as this kinase requires a priming phosphorylation at the +4 residue. These findings were of interest, as Gsk3 is known to be negatively regulated by the Akt kinase downstream of the TORC pathway (39, 40).

We first tested the ability of Gsk3 to phosphorylate an Elp4 peptide encompassing the targeted region. Immunoprecipitated Gsk3 only phosphorylated the Elp4 peptide prephosphorylated at position 118 and harboring a serine at position 114, which confirmed the dependency of Gsk3 for a priming phosphorylation at +4 (Fig. 4B). To test the ability of Gsk3 to phosphorylate the endogenous Elp4, we precipitated Gsk3 and Elp4 from fission yeast cells and mixed the precipitated products for a subsequent kinase assay. Gsk3 phosphorylated multiple proteins from the Elp4 immunoprecipitation (IP), with one band fully dependent on the presence of a serine at position 114 on Elp4 (Fig. 4, C and D). Together, these data strongly support that Elp4 is a target of Gsk3 in fission yeast.

We next confirmed that the fission yeast Akt kinase Gad8 phosphorylated the fission yeast Gsk3 in vitro using a recombinant glutathione S-transferase (GST)–Gsk3 as a substrate. A previous study had identified S335 as a major, albeit not unique, inhibitory phosphorylated site (41), and the phosphorylation of Gsk3 by Gad8 was indeed reduced when a GST-Gsk3 S335A protein was used as a substrate (fig. S6, A to C). These data supported a conservation of the TOR → Akt → Gsk3 connection in fission yeast. Moreover, recombinant GST-Gsk3 harboring a S335E mutation was less active than the wild-type version (fig. S6D). In addition, the activity of Gsk3 on Elp4 was increased in a gad8 mutant (fig. S6D) and strongly decreased when nitrogen was removed (fig. S6E), a condition known to activate the TORC2-Gad8 branch of the TOR pathway. Despite the fact that, in this case, the level of the Gsk3 protein dropped, the expected net consequence is a decreased phosphorylation of Elp4 on S114.

To analyze the consequence of Elp4 phosphorylation, we generated unphosphorylatable and phospho-mimic S114 and S118 versions of Elp4 and examined their phenotypes. As an intermediate step in creating the elp4 mutant, we generated a strain disrupted for elp4 by retaining only the first 91 amino acids. Unexpectedly, this truncated version behaved as a dominant negative, as the expression of wild-type elp4 could not fully complement the rapamycin sensitivity of the strain (Fig. 5, A and B). The S114A mutant grew much better than the wild type, bypassing the dominant-negative phenotype, which suggested that the S114 phosphorylation by Gsk3 is inhibitory and that its removal in the S114A mutant generates a hyperactive version of Elongator. In contrast, mutating S114, the phospho-mimic glutamic acid, decreased the ability of the strain to grow in the presence of rapamycin (Fig. 5B), suggesting a negative effect of the phosphorylation. The status of position 114 directly affected the level of the Elp4 protein (Fig. 5C), which is a typical feature of Gsk3-phosphorylated sites (42). The effect of mutating S118 was less clear and difficult to interpret, as either alanine or glutamic acid versions behaved closer to the wild type with a clear stabilization of the protein. In addition, the kinase phosphorylating S118 is unknown, and we therefore focused on the S114 site.

We next analyzed the effect of the Elp4 S114A mutation on the TORC1 and TORC2 complexes using the level of phosphorylation of Psk1 (43) and Gad8 as specific reporters of the activity of these complexes (Fig. 5D and fig. S7). This experiment revealed that the Elp4 S114A mutation has an opposite effect on TORC1 (Psk1) and TORC2 (Gad8) in favor of TORC2 activity. Similar to a TORC2 mutant, an elp4 mutant is defective for gametogenesis (Fig. 5E). The defect was complemented by the S114A version of Elp4, confirming the close relationship between Elongator and TORC2. Together, these data suggest that the S114 phosphorylation of Elp4 decreases the activity of Elongator.

We further tested this possibility using LC-MS to measure the impact of the S114 phosphorylation on the subset of Elongator-dependent modifications. The S114A mutation behaved as a deletion mutant of gsk3, mostly resulting in an increase of the major Elongator-dependent modified nucleotides (Fig. 6A). A previous study in fission yeast reported that the Ppe1 phosphatase opposes Gsk3 (44). We therefore tested the ppe1 mutant in the same setup, which indicated that it behaved oppositely to Gsk3 and was required for proper level of the Elongator-dependent modifications (Fig. 6A). It will therefore be of interest to test whether Ppe1 can directly dephosphorylate Elp4 when an assay becomes available. To get a condition where rapamycin affects the growth of fission yeast, we combined it with caffeine as previously reported (45). Under these conditions, the behavior of the gsk3 mutants was again opposite to that of the ppe1 mutant, showing slow growth reminiscent of an elp4 mutant, while the deletion of gsk3 suppressed the growth defect of the wild type (Fig. 6B). In addition, the Elp4 S114A mutant substantially suppressed the growth defect of the ppe1 deletion strain on rapamycin (Fig. 6B).

Together, these data indicate that TORC2 negatively regulates Gsk3 and that, in turn, Gsk3 negatively regulates Elongator through the phosphorylation of Elp4 S114, while Ppe1 likely opposes this phosphorylation. In that context, the structural partner of Ppe1, Ekc1 (46), behaved as an activator of Elongator in the genetic screen described in fig. S3. These data also provide a mechanistic view of the role of Elongator in regulating the nutrient-dependent switch between TORC1 and TORC2, as presented in the model shown in fig. S9. To further test this model, we directly analyzed the effect of the deletion of gsk3 or the inactivation of Elongator on the activities of TORC1 and TORC2 using the phosphorylation of Psk1 and Gad8 as direct specific reporters of TOR signaling (47, 48) (Fig. 6, C and D, fig. S8, and table S2). Within an hour, nitrogen starvation leads to the switch from high TORC1 activity to high TORC2 activity. In the absence of Gsk3, while the inactivation of TORC1 occurred normally, the phosphorylation of Gad8 increased 30 min earlier. The absence of Elp4 led to an opposite effect with persistent TORC1 activity and the inability to activate TORC2. In the reciprocal experiment in which nitrogen is added to starved cells, high TORC2 activity decreased within 30 min, in addition to an increase in TORC1 activity. In the absence of Gsk3, the activation of TORC1 was delayed by about 30 min with persistent TORC2 activity. In the absence of Elongator, TORC1 activity was still detected even during nitrogen starvation and quickly increased when nitrogen was added, while the activity of TORC2 was barely detectable during the entire course of the experiment.

DISCUSSION

This work shows that, by promoting the translation of Tor1 and Ste20 among others, Elongator is a critical activator of the TORC2 complex. In turn, TORC2 releases the inhibitory phosphorylation of Elp4 by Gsk3, therefore creating a positive TORC2-Elongator feedback loop, which is required for the nutrient-dependent control of mitosis onset and cell fate (fig. S9). In contrast, Elongator negatively controls TORC1 by sustaining the translation of negative regulators of the complex including the hamartin Tsc1 and Tsc2. Recent work has revealed that the Ppe1-Eck1 phosphatase is inhibited by TORC1 (49). Here, we have identified Eck1, an activating subunit of Ppe1, as an activator of Elongator. Together with previous data indicating that Ppe1 counteracts Gsk3 in fission yeast (44), the possibility that Ppe1 dephosphorylates Elp4 on S114 exists, therefore creating a negative TORC1-Elongator feedback loop, which we are currently investigating.

The central position of Elongator as a regulator of the balance between TORC1 and TORC2 is also supported genetically. Similar to strains with hyperactive TORC1, Elongator mutants display resistance to canavanine, slow growth when bearing auxotrophic markers, defect in gametogenesis, and the inability to adapt cell size in response to change in the nitrogen source (11). Elongator mutants also share some phenotypes with TORC2 inactivation including moderate cell elongation and misplaced septa (50). In addition, it was previously noted on that the levels of Ste20 and Sin1, two components of TORC2, are reduced when the TORC1 kinase Tor2 is overexpressed (6), which likely reflects the fact that Elongator is down-regulated under these conditions.

Both TORCs also critically regulate the level of Elongator-dependent tRNA modifications, as shown by MS-based analyses of modified uridine in strains overexpressing either tor1 or tor2 (Fig. 4A). Although it was anticipated that Elongator could be regulated on the basis of the phosphorylation of Elp1 (33), and despite the requirement of the Hrr25 kinase and the Sit4 phosphatase for the synthesis of the mcm⁵ modification (23), our work shows that tRNA modification by Elongator not only is a basic requirement for proteome integrity (51) but also is regulated by the central growth control machinery of the eukaryotic cell.

The mechanism of the control of Elongator by TORC2 relies on a critical phosphorylation of Elp4, which is regulated by the TORC2-Gad8 branch of the TOR pathway through Gsk3. The site is conserved within the fission yeast clade, and a closely related position was shown to undergo TOR-dependent phosphorylation in the distantly related species Saccharomyces cerevisiae (36), which suggests that the phosphorylation we report here likely plays a major conserved function. The S114 and flanking consensus are located within a region with overall low conservation that was shown to make intimate contacts with both Elp1 and Elp3 (52), which may underlie its importance. Moreover, the fact that two Elp4-Elp5-Elp6 subcomplexes can be loaded onto the Elp1-Elp2-Elp3 two-lobe symmetric scaffold may allow the possibility of double rate synthesis of tRNA modification under some conditions, and the phosphorylation of Elp4 could be critical for this regulation. The Elp4 S114 phosphorylation site is found within a Gsk3 consensus, and our data indeed show that it requires prephosphorylation at the +4 position (S118). In addition, an S114A mutation markedly stabilizes Elp4, which is another well-conserved feature of Gsk3 substrates (42). As the phosphorylation of S118 is a prerequisite for efficient S114 phosphorylation, one may anticipate that a S118A mutant would behave similarly to a S114A mutant. However, the S118A mutant rather behaved as the wild type (Fig. 5B). One cannot exclude that S118 phosphorylation has a function on its own, as suggested by the detrimental effect of the S118E mutation, which complicates the interpretation of the behavior of the S118 mutant.

Our data also support that Gsk3 is itself phosphorylated by Gad8, which negatively regulates its activity. This may occur at S335, whose phosphorylation was previously shown to decrease the activity of Gsk3 in fission yeast (41). However, this site is not in a perfect Akt-Gad8 consensus, but additional consensus sites, including S282, are present and were not tested here. In addition, the level of Gsk3 phosphorylation by Gad8 in vitro remains weak, which may indicate that the assay is not optimal or that additional kinases function in vivo.

Genome-wide screens for rapamycin sensitivity in fission yeast have shown that the phenotype is not common but specific to Elongator mutants (53). Our data show that the sensitivity depends on Tor2, as a tor2-resistant allele but not a tor1-resistant allele bypasses it. A possibility to explain the sensitivity of Elongator mutant to rapamycin is that, upon addition of the compound, TORC1 activity is decreased but TORC2 cannot be activated because of the low level of Ste20 and Tor1. The resulting uncoordinated TOR activity would be lethal. It was previously reported that a tor1/tor2 double mutant is not viable despite the opposite functions of both complexes (11). Therefore, the concomitant decrease of TORC1 and TORC2 activities when Elongator mutants are grown on rapamycin could similarly result in lethality.

Rapamycin mimics the shortening of the cell cycle that results from a change in the nitrogen source that affects cell fate. The present work shows that Elongator is essential for that process and functions as a central coordinator of the nutrient-dependent control of cell division and cell fate. These findings shed new light on the increasing reports of prominent roles of Elongator in tumorigenesis (54–56).

MATERIAL AND METHODS

Supplementary Material