|Home | About | Journals | Submit | Contact Us | Français|
Snf1 is the ortholog of mammalian AMP-activated kinase and is responsible for activation of glucose-repressed genes at low glucose levels in budding yeast. We show that Snf1 promotes the formation of phosphorylated α subunit of eukaryotic translation initiation factor 2 (eIF2α-P), a regulator of general and gene-specific translation, by stimulating the function of eIF2α kinase Gcn2 during histidine starvation of glucose-grown cells. Thus, eliminating Snf1 or mutating its activation loop lowers Gcn2 kinase activity, reducing the autophosphorylation of Thr-882 in the Gcn2 activation loop, and decreases eIF2α-P levels in starved cells. Consistently, eliminating Reg1, a negative regulator of Snf1, provokes Snf1-dependent hyperphosphorylation of both Thr-882 and eIF2α. Interestingly, Snf1 also promotes eIF2α phosphorylation in the nonpreferred carbon source galactose, but this occurs by inhibition of protein phosphatase 1α (PP1α; Glc7) and the PP2A-like enzyme Sit4, rather than activation of Gcn2. Both Glc7 and Sit4 physically interact with eIF2α in cell extracts, supporting their direct roles as eIF2α phosphatases. Our results show that Snf1 modulates the level of eIF2α phosphorylation by different mechanisms, depending on the kind of nutrient deprivation existing in cells.
Budding yeast cells reprogram gene expression in response to amino acid starvation by a two-pronged regulatory mechanism centered on the protein kinase Gcn2. Gcn2 has low activity in nonstarved cells and is activated by uncharged tRNAs that accumulate during starvation. Gcn2 has a single known substrate, Ser51 in the α subunit of eukaryotic translation initiation factor 2 (eIF2α). eIF2 delivers methionyl-tRNAMet in a ternary complex (TC) with GTP to the 40S ribosomal subunit, and eIF2α phosphorylation inhibits TC formation and subsequent steps in the initiation pathway. Although phosphorylated eIF2 inhibits general protein synthesis, it induces the translation of GCN4 mRNA, encoding a transcriptional activator of amino acid biosynthetic enzymes. Four short open reading frames (uORFs) in the GCN4 mRNA leader prevent scanning ribosomes from reaching the GCN4 start codon in nonstarved cells when TC is abundant. The reduction in TC levels provoked by eIF2α phosphorylation in starved cells allows ribosomes that have translated uORF1 and resumed scanning to bypass the more inhibitory uORFs 3 and 4 and reinitiate downstream at GCN4 instead. Hence, Gcn2 induces Gcn4 and its target amino acid biosynthetic genes to replenish amino acid pools while reducing general protein synthesis and amino acid consumption (reviewed in reference 10). This dual response to amino acid starvation also operates in mammalian cells, involving Gcn2 and transcription factors Atf4 (8, 28) and Atf5 (35).
Phosphorylation of the activation loop in the Gcn2 kinase domain (KD) is required for its activation. Thr-882 and Thr-887 are the only sites of autophosphorylation in vitro, and mutating them to Ala impairs (T882A) or abolishes (T887A) Gcn2 function (23). We previously isolated substitutions in the Gcn2 KD, R794G and F842L, that constitutively activate Gcn2 in the absence of amino acid starvation or the tRNA binding domain in Gcn2, which mediates activation of Gcn2 by uncharged tRNA (22). Kinase activation by these GCN2c mutations is associated with hyperphosphorylation of T882 (T882-P); however, it has not been demonstrated that physiological activation of wild-type (WT) Gcn2 in amino acid-starved yeast cells involves increased T882 phosphorylation.
Another layer of nutrient control over Gcn2 function involves the phosphorylation of a negative regulatory residue N terminal to the KD, Ser577, which is dependent on the Tor kinases. Yeast Tor1 and Tor2 regulate metabolic processes to coordinate cell growth with nutrient availability, and their inhibition by mutation or rapamycin provokes cellular responses characteristic of nutrient starvation (reviewed in reference 4). Tor kinases promote Gcn2 latency in nonstarved cells by enhancing Ser577 phosphorylation (S577-P), at least partly by inhibiting type 2A protein phosphatases that dephosphorylate S577-P (2). Thus, Tor proteins stimulate general protein synthesis in yeast in part by preventing eIF2α phosphorylation. As expected, rapamycin elicits dephosphorylation of S577-P by inhibiting Tor function; however, histidine starvation and other stress conditions that increase eIF2α phosphorylation do not reduce S577-P levels (2). Presumably, the inhibitory effect of S577-P can be overridden when uncharged tRNA accumulates to high levels during amino acid starvation to elicit Gcn2 activation (6).
Budding yeast Snf1, the founding member of the Snf1/ AMP-activated protein kinase (AMPK) family of protein kinases, is another key regulator of gene expression by nutrients. Its best-understood function is to orchestrate the transcriptional response to glucose limitation, involving induction of genes that regulate the metabolism of alternative carbon sources, gluconeogenesis, or respiration during growth in nonpreferred carbon sources. Snf1 kinase activity is stimulated at low glucose levels by upstream kinases that phosphorylate activation loop residue Thr-210, overcoming the inhibitory effect of the Reg1-Glc7 protein phosphatase 1 (PP1) complex that dephosphorylates T210-P under glucose-replete conditions (9). T210 phosphorylation and Snf1 activity are also stimulated by other environmental stresses, including high salinity and alkaline pH (16), and by inactivation of Tor kinases (9, 19). Snf1 also mediates resistance to heat shock (12), toxic cations (21), and hydroxyurea (5) in a manner that does not require increased T210-P or elevated kinase activity, indicating that Snf1 basal activity is sufficient for its function under certain stresses.
In this report, we document the existence of cross talk between Snf1, Gcn2, and two distinct eIF2α phosphatases in coupling eIF2α-P formation to amino acid and glucose availability. We found that Snf1 kinase activity is stimulated in histidine-starved cells, where it activates Gcn2. In contrast, Snf1 promotes eIF2α-P formation in the nonpreferred carbon source galactose by inhibiting dephosphorylation of eIF2α-P by protein phosphatases Glc7 and Sit4. Thus, we can now state that two major nutrient-signaling kinases that respond to the availability of different nutrients, Snf1 (carbon) and Tor (nitrogen), make critical contributions to setting the level of eIF2α-P, a key regulator of translation initiation in nutrient-limited cells.
All of the yeast strains and plasmids used in this study are listed in Tables Tables11 and and2,2, respectively. Strain HQY305 was constructed by transforming H1642 with a reg1Δ::hisG::URA3::hisG fragment, and deletion of REG1 in the resulting transformant was confirmed by PCR analysis, after which selection for growth on 5-fluoroorotic acid (5-FOA) was used to evict the URA3 marker. Similarly, HQY343, HQY344, HQY347, HQY356, and HQY357 were constructed from H1642, H1895, HQY305, H1515, and H1894, respectively, by transformation with the snf1Δ::hisG::URA3::hisG fragment from pHQ1160, confirming the deletion of SNF1 by PCR analysis, and selecting a Ura− segregant on 5-FOA medium. HQY1313 was constructed from H1515 by transformation with the GLC7::13myc::kanMX fragment, and the addition of myc13 coding sequences at the 3′ end of the GLC7 ORF was confirmed by colony PCR.
CY1194 was generated by crossing JC782-26C with HQY358, identifying a glc7-1 snf1Δ::hisG ascospore clone by its Ura+ phenotype and defective accumulation of glycogen (29), followed by selection on 5-FOA medium to evict the URA3 marker. Strain CY1323 was generated by transforming CY1194 with plasmid pCB227, linearized by digestion with BglII to direct integration to glc7-1; this was followed by selection of Ura− clones on 5-FOA medium. The presence of GLC7 in CY1323 was confirmed by staining for glycogen. Strains CY1297, CY1298, and CY1299 were constructed by replacing SIT4 with sit4Δ::kanMX4 in H1642, HQY343, and CY1194, respectively, followed by PCR verification of the replacements.
Plasmids pB11, pB34, pB61, pB107, pB115, pDH101, and pCB149 were described previously (2, 6, 23). Plasmid pHQ817 was made by inserting a 3.68-kb EcoRI-SalI fragment generated by PCR and containing 340 bp of 5′ noncoding sequences, the ORF, and 300 bp of 3′ noncoding sequences of REG1 between the EcoRI and SalI sites of YCplac111. To construct pHQ1160, a 590-bp XbaI-BglII fragment containing the SNF1 5′ noncoding region and a 500-bp BglII-XbaI fragment containing the SNF1 3′ noncoding region were generated by PCR and inserted at the XbaI site of pUC18 to produce pHQ1157, after which a 3.8-kb BamHI fragment containing hisG::URA3::hisG from pHQ221 was inserted at the BglII site of pHQ1157. Plasmid pHQ221 was constructed by ligating the 3.8-kb BamHI-BglII fragment containing hisG::URA3::hisG from pNKY51 (1) with XbaI-digested pUC18 after using the Klenow fragment to blunt end both fragments. pHQ1162 and pHQ1163 were constructed by inserting a 3-kb XbaI fragment generated by PCR, containing 590 bp of 5′ noncoding sequences, the ORF, and 500 bp of 3′ noncoding sequences of SNF1, into the XbaI sites of YCplac33 and YCplac111, respectively. pHQ1873 (snf1-K84R) and pHQ1874 (snf1-T210A) were constructed as follows. (i) PCR was employed with yeast genomic DNA as the template and primers 5′ CTC TCG ATC TTG CAG GCT ATG and 5′ CAGA TCT AGA CTT GGG ATT GTT TAG CGT C 3′ to amplify a SNF1 fragment containing 475 bp of the promoter and the first 44 codons after digestion with XhoI and XbaI (the latter is underlined in the reverse primer). (ii) Plasmids pSNF1-K84R and pMO19 were used as templates with primers 5′ CAAG ACT AGT TTA GCG GAT GGT GCA CAT ATC 3′ and 5′ CTT AAT CAA CAT GTA CGC GTC 3′ to produce 1,030-bp SNF1 fragments containing codons 45 to 382 after digestion with SpeI and MluI (the latter is underlined in the reverse primer). (iii) The digested PCR fragments from steps i and ii were ligated with pHQ1162 digested with XhoI and MluI. pHQ1774 was constructed as follows. (i) The HindIII-EcoRV fragment containing GLC7 obtained from p29.1 was inserted into YEplac195. (ii) An EcoRV-BglII GLC7::13myc fragment was amplified by PCR from the chromosomal DNA of HQY1313 and used to replace the cognate fragment in the plasmid obtained in step i. pCB227 was constructed by inserting the HindIII-XhoI fragment from p28.1 into pRS306 digested with SalI and HindIII.
For detection of phosphorylated Gcn2 and eIF2α, whole-cell extracts (WCEs) were prepared from yeast cultures grown to mid-logarithmic phase in SD medium with minimal supplements (24) as previously described (2), except that the breaking buffer contained 40 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES; pH 6.0), 100 mM NaCl, 50 mM NaF, 40 mM β-glycerophosphate, 1 mM dithiothreitol (DTT), 5 mM phenylmethylsulfonyl fluoride (PMSF), EDTA-free protease inhibitor cocktail (Roche), and 1 μg/ml each leupeptin, aprotinin, and pepstatin. For analysis of eIF2α phosphorylation, aliquots of WCE were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and subjected to immunoblot analysis using antibodies specific for phosphorylated Ser51 in eIF2α (BIOSOURCE International). The blots were stripped and reprobed with polyclonal antibodies against eIF2α as described previously (2, 6). For analysis of Thr882 or Ser577 phosphorylation, aliquots of 4 to 8 mg of WCE were incubated with 30 μl anti-FLAG protein A-agarose beads (Sigma) at 4°C overnight. The immunoprecipitates were washed four times with 1 ml of breaking buffer, resolved by 4 to 12% SDS-PAGE, and subjected to immunoblot analysis using antibodies against Thr882-P (22) or Ser577-P (6). The blots were stripped and reprobed with polyclonal antibodies against Gcn2 (serum HL2523) (23). The immunocomplexes were visualized using anti-rabbit antibodies conjugated with horseradish peroxidase and SuperSignal West Dura (Thermo Scientific) or the enhanced chemiluminescence system (Amersham). NIH Image 1.63 software was used for quantification of Western signals. The Thr882 phospho-specific antibodies are not very efficient, and in some cases, the Thr882-P signal observed for the loading with the lesser amount of FL-Gcn2 was outside the linear range. For consistency, therefore, we always quantified the Thr882-P signals for the loadings with the greater amounts of FL-Gcn2 and ensured that the FL-Gcn2 signals exhibit a dose response between the two loadings.
For coimmunoprecipitation of Gcn2 with hemagglutinin (HA)-Snf1 or of eIF2α with Sit4-HA or Glc7-myc, WCE preparation and immunoprecipitation were done essentially as described in reference 27, with extraction buffer containing 50 mM HEPES (pH 7.5), 150 mM NaCl, 0.1% Triton X-100, 1 mM DTT, 10% glycerol, 1 mM PMSF, and complete protease inhibitor cocktail (Roche).
For analysis of Snf1 T210 phosphorylation, WCEs were prepared and subjected to immunoblot analysis with antibodies against AMPK T172-P (Cell Signaling Technology) as described previously (18).
HA-Snf1 was immunoprecipitated from WCEs (2 mg) prepared in the extraction buffer described above (27), using 30 μl of anti-HA agarose beads (Santa Cruz Biotechnology) for 2 h at 4°C. After immunoprecipitation, beads were washed three times with 1 ml of extraction buffer and two times with 1 ml kinase buffer (50 mM Tris-HCl [pH 7.5], 10 mM MgCl2, 1 mM DTT, 0.1% Triton X-100). Immunocomplexes bound to the beads were divided in half and subjected to Western analysis or kinase assays. The latter were conducted with 20 μl kinase buffer plus 250 μM ATP and 10 μCi [γ-32P]ATP at 30°C for 20 min and stopped by adding the same volume of 2× SDS sample buffer. The reactions were resolved by SDS-PAGE and analyzed by phosphorimaging.
In view of previous reports indicating regulatory roles for Snf1 basal kinase activity at high glucose levels (6, 15, 24), we were intrigued to find that deletion of SNF1 confers sensitivity to 3-aminotriazole (3AT), an inhibitor of the histidine biosynthetic enzyme His3, in glucose-containing medium (Fig. (Fig.11 A). Plasmid-borne SNF1 complemented the 3ATS phenotype of the snf1Δ mutant (Fig. (Fig.1B),1B), confirming that it results from the absence of Snf1. Sensitivity to 3AT (3ATS) is a hallmark of Gcn− mutants, evident for the SNF1 gcn2Δ mutant strain in Fig. Fig.1A.1A. To examine whether the 3ATS phenotype of snf1Δ mutant cells results from impaired translational induction of GCN4, we asked whether it could be suppressed by dominant alleles of GCN4 lacking the two most inhibitory uORFs (uORFs 3 and 4) or all four uORFs in the GCN4 mRNA leader (GCN4c alleles). Translation of GCN4 mRNA from these alleles is constitutively derepressed and independent of Gcn2 (17). Both GCN4c alleles only partially suppressed the 3ATS phenotype of the snf1Δ mutant (Fig. (Fig.1B).1B). As Gcn2 overexpression also is known to derepress GCN4 (30), we tested whether overexpressing Gcn2 from a low-copy-number (lc) or high-copy-number (hc) plasmid would suppress the 3ATS phenotype of snf1Δ mutant cells. However, this was not the case (Fig. (Fig.1B),1B), even though Gcn2 was overexpressed ~5-fold in the hc GCN2 transformants (data not shown). Finally, we found that snf1Δ does not lower the expression of a GCN4-lacZ construct in 3AT-treated cells; in fact, snf1Δ partially derepresses this reporter under nonstarvation conditions (Table (Table3).3). Similar results were reported recently by Arndt and colleagues (25). Hence, it appears that the 3ATS phenotype of snf1Δ does not involve reduced translation of GCN4. Consistent with this, Kuo and colleagues reported that eliminating Snf1 decreases the transcription of HIS3, whose product is targeted by 3AT, by impairing the transcriptional coactivator Gcn5 (14).
While investigating the 3ATS phenotype of snf1Δ mutant cells, we also explored whether the kinase function of Gcn2 is compromised in snf1Δ mutant cells by measuring the induction of Ser51 phosphorylation on eIF2α (eIF2α-P) by 3AT treatment. Indeed, the results showed clearly that snf1Δ reduces the abundance of eIF2α-P relative to the total amount of eIF2α (Fig. (Fig.1C).1C). This phenotype is complemented by SNF1 but is not suppressed by hc GCN2 (Fig. (Fig.1D).1D). As discussed below, the decreased level of eIF2α-P in snf1Δ mutant cells might not reduce GCN4 translation because eliminating Snf1 derepresses GCN4 independently of high-level eIF2α phosphorylation. Because eIF2α-P is an important negative effector of general translation, in addition to its importance in GCN4 control, we set out to elucidate the role of Snf1 in promoting eIF2α phosphorylation.
We first examined the possibility that snf1Δ decreases Gcn2 kinase activity by increasing the phosphorylation of Ser577 in Gcn2, a modification required for Gcn2 latency. Thus, we asked whether replacing Ser577 with Ala would suppress the defect in eIF2α phosphorylation in snf1Δ mutant cells, as S577A constitutively activates Gcn2 (6). As expected, GCN2c-S577A cells have constitutively induced levels of eIF2α-P; however, this phenotype was largely reversed by snf1Δ (Fig. (Fig.2A).2A). Furthermore, as described below, snf1Δ does not increase Ser577-P in a FLAG-tagged form of Gcn2 (e.g., Fig. Fig.2B,2B, lanes 1 to 6). Hence, snf1Δ reduces eIF2α-P accumulation in starved cells without altering the Ser577-P content of Gcn2.
The decrease in eIF2α-P observed in snf1Δ mutant cells could result from decreased Gcn2 kinase activity or increased eIF2α phosphatase activity. To distinguish between these possibilities, we assayed the phosphorylation level of Thr-882 in the Gcn2 activation loop as an indicator of Gcn2 activation. Similar to other kinases, Thr882 phosphorylation stimulates a conformational change in the KD that overcomes a nonproductive alignment of catalytic residues in the active site (7, 20). However, it was not known whether T882-P increases on the activation of WT Gcn2 in starved cells. If so, we could determine whether snf1Δ reduces T882-P and, hence, Gcn2 activity under these conditions.
Accordingly, we used antibodies specific for T882-P to probe FLAG-tagged Gcn2 (FL-Gcn2) immunoprecipitated from yeast. These antibodies were shown to react with a constitutively activated form of the Gcn2 KD but not with catalytically inactive or unphosphorylatable forms of the mutant KD harboring the K628R and T882A substitutions, respectively (22). The T882-P signal, normalized for the amount of FL-Gcn2 (or for the Ser577-P signal measured using the appropriate phosphospecific antibodies), increased >3-fold in response to 3AT (Fig. (Fig.2B,2B, cf. lanes 1 and 2 and lanes 7 and 8; also data not shown). Moreover, the T882-P/FL-Gcn2 ratio showed an obvious increase for activated FL-Gcn2c-S577A under nonstarvation conditions (Fig. (Fig.2B,2B, lanes 5 and 6 versus 1 and 2). Thus, T882-P increases with activation of Gcn2 elicited by histidine starvation or the absence Ser-577-P.
Importantly, the T882-P content of FL-Gcn2 was significantly reduced by snf1Δ in the presence of 3AT and possibly also in its absence (Fig. (Fig.2B,2B, cf. lanes 3 and 4 versus 1 and 2 and lanes 9 and 10 versus 7 and 8). This strongly suggests that Snf1 is a positive effector of Gcn2 activity and the reduction of eIF2α-P during starvation of snf1Δ mutant cells results at least partly from decreased Gcn2 function. As mentioned above, the level of S577-P in FL-Gcn2 does not vary significantly between WT and snf1Δ mutant cells in the presence or absence of 3AT (Fig. (Fig.2B,2B, middle blot), thus ruling out S577 dephosphorylation as a mechanism of Gcn2 activation by Snf1.
At high glucose concentrations, Snf1 activity is repressed by the PP1α Glc7 in complex with its regulatory subunit Reg1 (15, 16) such that Snf1 is activated in reg1Δ cells at high glucose levels. We asked, therefore, if activation of Snf1 by reg1Δ evokes greater Gcn2 activation and eIF2α phosphorylation during histidine starvation. Indeed, reg1Δ cells exhibit higher eIF2α-P levels than WT cells upon 3AT treatment (Fig. (Fig.3A,3A, lanes 14 and 15 versus 10 and 11), mirrored by a similar increase in the T882-P content of FL-Gcn2 (Fig. (Fig.3B,3B, lanes 13 and 14 versus 9 and 10). Importantly, deletion of SNF1 suppressed these effects of reg1Δ, yielding the reduced levels of eIF2α-P and T882-P in the reg1Δ snf1Δ double mutant characteristic of the snf1Δ single mutant (Fig. 3A and B, right panels). These data strongly support our conclusion that Snf1 stimulates Gcn2 kinase activity and attendant eIF2α phosphorylation under histidine starvation conditions and indicate that Reg1 constrains this Snf1 function in glucose medium. As expected, the S577-P/FL-Gcn2 ratio did not vary significantly among the WT, reg1Δ mutant, and snf1Δ mutant strains in these experiments (Fig. 3A and B).
Surprisingly, reg1Δ provoked a decrease, rather than an increase, in eIF2α-P under nonstarvation conditions (Fig. (Fig.3A,3A, left panel), even though it did not reduce the level of T882-P (Fig. (Fig.3B,3B, left panel). Moreover, the decrease in eIF2α-P conferred by reg1Δ was reversed by the deletion of SNF1 (Fig. (Fig.3A,3A, left panel). Both the decreased eIF2α-P under nonstarvation conditions and the increased eIF2α-P upon 3AT treatment in reg1Δ cells were complemented by plasmid-borne REG1 (Fig. (Fig.3C).3C). The fact that reg1Δ reduces eIF2α-P without decreasing T882-P (Gcn2 activity) implies that activating Snf1 in nonstarved reg1Δ mutant cells increases the dephosphorylation of eIF2α-P by a protein phosphatase. The relevant phosphatase might be activated by 3AT in reg1Δ mutant cells as well, but its effect on eIF2α-P accumulation could be masked by Snf1-mediated activation of Gcn2.
The fact that the increased levels of T882-P and eIF2α-P produced by reg1Δ in histidine-starved cells depend on SNF1, combined with the known function of Reg1 in negatively regulating Snf1, suggested that activation of Snf1 by the phosphorylation of T210 in the activation loop is required for its ability to stimulate Gcn2. To test this prediction, we examined the ability of snf1-T210A and catalytically inactive snf1-K84R to complement snf1Δ mutant cells for impaired eIF2α-P induction by 3AT. Both snf1 alleles were indistinguishable from the empty vector (Fig. (Fig.4A).4A). Furthermore, in the snf1Δ reg1Δ background, plasmid-borne SNF1, but not snf1-T210A, increased both eIF2α-P and T882-P levels above those in transformants with the empty vector under starvation conditions (Fig. (Fig.3D,3D, upper and lower panels). These results imply that the activated level of Snf1 function requiring T210-P is necessary for stimulating Gcn2 activity in histidine-starved cells.
To test this conclusion further, we asked whether 3AT increases the phosphorylation of T210 in Snf1 by using phosphospecific antibodies (19, 26). Consistent with previous findings (16), the T210-P content of immunoprecipitated HA-tagged Snf1 was increased by reg1Δ and undetectable in the strain expressing HA-Snf1-T210A (Fig. (Fig.4B,4B, lanes 7 to 10 versus 1 and 2). Treatment with 3AT consistently provoked a moderate increase in the T210-P content of HA-Snf1 (cf. lanes 3 and 4 and lanes 5 and 6), suggesting that Snf1 function is stimulated by histidine starvation.
Having found that Snf1 kinase activity is required for its role in activating Gcn2, we wondered whether it might function directly by phosphorylating Gcn2. If so, we would expect to find that Gcn2 and Snf1 are physically associated in vivo. Indeed, Gcn2 coimmunoprecipitated specifically with HA-Snf1, being absent in immunocomplexes from cells expressing untagged Snf1 (Fig. (Fig.5A,5A, top blot, lanes 3 and 11). Similarly, Snf1-HA coimmunoprecipitated specifically with Gcn2, being absent in immunocomplexes from gcn2Δ mutant cells expressing Snf1-HA (Fig. (Fig.5A,5A, bottom blot, lanes 3, 7, and 11). We were unable to determine whether association of Gcn2 with Snf1 is induced by histidine starvation because Snf1 is activated during cell harvesting by the g stress of centrifugation or nutrient limitation in the dense cell pellet (32). Moreover, Gcn2 is activated in extracts from nonstarved cells (31). Hence, it is currently impossible to prepare native cell extracts for coimmunoprecipitation experiments in which Snf1 and Gcn2 are at basal levels of activation. Thus, whereas activated Snf1 and Gcn2 can clearly associate (Fig. (Fig.5A),5A), it is not known if the nonactivated forms of these proteins can do so.
Having detected complexes containing native Gcn2 and Snf1, we explored whether phosphorylation of T882 in the Gcn2 activation loop is mediated by Snf1. Although T882 is autophosphorylated by Gcn2 in vitro (23), it was not known whether autophosphorylation is the primary source of T882-P in vivo. To address this, we examined the effect of inactivating Gcn2 by the K628R substitution in the active site on the level of T882-P in the mutant protein. If T882 is phosphorylated by Snf1 in vivo, then inactivation of Gcn2 should not eliminate T882-P. Indeed, the equivalent inactivating substitution in Snf1 (K84R) does not diminish T210-P because this residue is phosphorylated by upstream kinases (13, 16). At odds with this possibility for Gcn2, the K628R substitution eliminates T882-P in cells starved with 3AT (Fig. (Fig.5B).5B). This makes it unlikely that Snf1 activates Gcn2 directly by T882 phosphorylation.
It was possible that Snf1 phosphorylates Gcn2 on a residue outside the activation loop to stimulate kinase activity, indirectly increasing the T882-P content. To examine this possibility, we coimmunoprecipitated catalytically inactive gcn2-K628R with either functional HA-Snf1 or nonfunctional HA-snf1-T210A and carried out immunocomplex kinase assays with radiolabeled ATP. We observed no labeling of gcn2-K628R in complexes with functional or nonfunctional Snf1-HA (Fig. (Fig.5C,5C, lanes 1 and 2 versus 3 and 4). Consistent with this, a glutathione S-transferase-Snf1 protein purified from yeast, which was functional for autophosphorylation, did not produce any detectable phosphorylation of purified, catalytically inactive gcn2-K628R (data not shown). Finally, we tested the possibility that association with active Snf1 might increase the ability of Gcn2 to autophosphorylate its activation loop. Indeed, we consistently observed an increase in the labeling of WT Gcn2, by a factor of 2.3 ± 0.19, in immunocomplexes with HA-Snf1 versus HA-snf1-T210A (Fig. (Fig.5C,5C, lanes 5 and 6 versus 7 and 8; data not shown). These last results suggest an indirect mechanism of Gcn2 activation by Snf1, which is considered further below.
Because Snf1 is activated in nonpreferred carbon sources and eIF2α phosphorylation increases at low glucose concentrations (33), we went on to investigate whether Snf1 regulates Gcn2 activity during growth in galactose. When WT and snf1Δ mutant strains were shifted from glucose to galactose medium, eIF2α-P declined in both strains within 20 min (Fig. (Fig.6A);6A); however, whereas eIF2α-P began to recover in WT cells after 4 h, eIF2α-P remained at low levels in the snf1Δ mutant strain throughout the experiment. Thus, Snf1 promotes eIF2α phosphorylation after cells are shifted from glucose to galactose medium.
We asked next if the reduced level of eIF2α-P in WT and snf1Δ mutant cells evoked by shifting to galactose is associated with reduced Gcn2 activity. In contrast to the reduced level of eIF2α-P seen in WT cells 90 min after a shift to galactose (Fig. (Fig.6A),6A), T882-P in FLAG-Gcn2 was not diminished under these conditions (Fig. (Fig.6B,6B, lanes 5 and 6 and lanes 1 and 2). Furthermore, the even larger reduction in eIF2α-P seen in snf1Δ mutant cells after 90 min on galactose (Fig. (Fig.6A)6A) is associated with an ~1.5-fold increase in T882-P (Fig. (Fig.6B,6B, lanes 7 and 8 versus 3 and 4). Hence, the reduced level of eIF2α-P seen upon a shift to galactose does not result from lower Gcn2 activity. This implies that a higher level of eIF2α-P phosphatase activity exists in galactose versus glucose medium and in snf1Δ mutant versus WT cells on galactose medium.
Interestingly, the increased T882-P observed in snf1Δ mutant cells after 90 min in galactose is associated with an ~2-fold decrease in Ser577 phosphorylation compared to the levels observed in WT or snf1Δ mutant cells on glucose (Fig. (Fig.6B,6B, lanes 7 and 8 versus 3 to 6). These data suggest that, in galactose medium, Snf1 promotes S577-P and thereby reduces Gcn2 activity. However, the fact that snf1Δ mutant cells exhibit lower levels of eIF2α-P despite increased Gcn2 activity (T882-P level) implies that the inhibitory effect of Snf1 on eIF2α-P phosphatase activity (deduced from Fig. Fig.6B)6B) exceeds Snf1's inhibitory effect on Gcn2 in galactose medium.
Having obtained evidence that Snf1 inhibits an eIF2α-P phosphatase (eIF2α-PPase) in galactose medium, we wondered whether eliminating Reg1 would elevate eIF2α-P by activation of Snf1 and attendant downregulation of eIF2α-PPase activity. Consistent with this idea, reg1Δ evoked higher levels of eIF2α-P in WT cells on galactose (Fig. (Fig.6C,6C, upper panels, lanes 5 and 6 versus 1 and 2), similar to those above for histidine starvation. Furthermore, snf1Δ suppressed this phenotype in the snf1Δ reg1Δ double mutant and reduced eIF2α-P to the low level observed in snf1Δ REG1 mutant cells (Fig. (Fig.6C,6C, lanes 7 and 8 versus 3 to 6). Thus, reg1Δ provokes a Snf1-dependent increase in eIF2α-P in galactose. Importantly, reg1Δ produced only a modest increase in T882-P relative to that measured in WT cells on galactose, comparable to that seen in snf1Δ mutant cells (Fig. (Fig.6C,6C, lower panels). Thus, an ~5-fold higher level of eIF2α-P in reg1Δ versus snf1Δ mutant cells occurs with no significant change in T882-P. This implies that, on galactose, an eIF2α-PPase is activated when Snf1 function is enhanced by eliminating Reg1.
Previous data suggested that the essential PP1α Glc7 acts in opposition to Gcn2 in regulating levels of eIF2α phosphorylation on glucose medium. Consistent with this, we found that the hypomorphic glc7-1 allele confers increased eIF2α-P on GCN2 mutant cells in glucose (Fig. (Fig.7A,7A, lanes 2 and 3 versus 10 and 11). We concluded above that Snf1 inhibits an eIF2α-PPase on galactose medium. If Glc7 is an eIF2α-PPase inhibited by Snf1 under these conditions, then glc7-1 should dampen the reduction in eIF2α-P that normally occurs in snf1Δ mutant cells on galactose. The results in Fig. Fig.7A7A support this prediction, as the eIF2α-P/eIF2α ratio is higher in the glc7-1 snf1Δ double mutant than in the GLC7 snf1Δ mutant strain (cf. lanes 8 and 9 versus 16 and 17). Moreover, glc7-1 does not increase Gcn2 activity on galactose (or glucose) medium, as glc7-1 and GLC7 mutant cells exhibit no significant differences in the T882-P content of FL-Gcn2 (Fig. (Fig.7B,7B, lanes 1 to 8 versus 9 to 15). Thus, the moderately higher levels of T882-P on galactose versus glucose, and in snf1Δ versus SNF1 cells, described above for GLC7 cells, also occur in glc7-1 mutant cells. The fact that glc7-1 elevates eIF2α-P in snf1Δ mutant cells without activating Gcn2 supports our conclusion that the eIF2α-PPase function of Glc7 is inhibited by Snf1 on galactose medium.
In the experiment just described, we observed that snf1Δ still evokes a significant decrease in eIF2α-P on galactose in the glc7-1 background (Fig. (Fig.7A,7A, lanes 8 and 9 versus 6 and 7). This could be explained by noting that glc7-1 does not completely inactivate Glc7, so that releasing its residual activity from inhibition by Snf1 (in snf1Δ mutant cells) should lead to a further reduction in eIF2α-P. However, an alternative possibility is that Snf1 inhibition of a second eIF2α-PPase contributes to the increased eIF2α-P conferred by snf1Δ in the glc7-1 background. Indeed, we obtained evidence previously that the type 2A-related phosphatase Sit4 contributes to the dephosphorylation of eIF2α on glucose, as sit4Δ mutant cells have elevated eIF2α-P levels under amino acid-replete conditions during growth in glucose medium (2). We confirmed this finding here (data not shown) and also observed that sit4Δ evokes increased eIF2α-P in SNF1 mutant cells on galactose (Fig. (Fig.7D,7D, lanes 6 and 7 versus 2 and 3). However, sit4Δ does not significantly alter T882-P (or S577-P) levels in glucose or galactose medium (Fig. (Fig.7C7C and data not shown), indicating that Sit4 does not regulate Gcn2 activity. Thus, Sit4, in addition to Glc7, contributes to the dephosphorylation of eIF2α-P in galactose-grown cells.
Importantly, in the sit4Δ background, deletion of SNF1 still confers a strong reduction in eIF2α-P (Fig. (Fig.7D,7D, lanes 8 and 9 versus 6 and 7), comparable to the effect of snf1Δ on SIT4 mutant cells. This suggests that Snf1 promotes eIF2α-P in galactose primarily by inhibiting the eIF2α-PPase activity of Glc7. If Sit4 and Glc7 together make up the bulk of the eIF2α-PPase activity downregulated by Snf1, then deletion of SNF1 in the sit4Δ glc7-1 background should evoke little decrease in eIF2α-P. This prediction was fulfilled by the results in lanes 14 to 17 of Fig. Fig.7D.7D. Hence, Glc7 and Sit4 are both downregulated by Snf1, and Glc7 is responsible for the majority of the eIF2α-P dephosphorylation in galactose.
Finally, having found that Sit4 and Glc7 both promote the dephosphorylation of eIF2α, we sought to demonstrate that eIF2α associates with Sit4 and Glc7 in cell extracts. Indeed, we found that eIF2α specifically coimmunoprecipitates with epitope-tagged forms of Sit4 and Glc7 in both glucose and galactose media (Fig. 7E and F). These findings strongly support the idea that Sit4 and Glc7 function directly as eIF2α phosphatases in vivo.
We have shown here that histidine starvation elicits Gcn2 activation via increased phosphorylation of T882 in the activation loop and that eliminating Snf1 or impairing its activation by upstream kinases (with the T210A substitution) reduces both T882-P in Gcn2 and phosphorylated eIF2α in 3AT-treated cells. In addition, histidine starvation evokes a moderate increase in T210-P levels in Snf1, and deletion of REG1 confers elevated Gcn2-Thr882-P and eIF2α-P in 3AT-treated cells in a manner dependent on Snf1. Thus, Snf1 is activated by histidine starvation and stimulates Gcn2 activity in glucose-grown cells (model in Fig. Fig.8A).8A). As T882-P and eIF2α-P are still induced by 3AT in snf1Δ mutant cells, albeit to lower levels than in WT cells, there is also a Snf1-independent component of Gcn2 activation evoked by uncharged tRNA.
Our finding that Snf1 and Gcn2 are associated with one another in cell extracts is consistent with a direct role for Snf1 in stimulating Gcn2 activity. However, T882-P is undetectable in catalytically inactive gcn2-K628R immunopurified from yeast, implying that T882-P formation in vivo depends primarily on autophosphorylation by Gcn2. Consistent with this, when examining the fraction of Gcn2 that specifically coimmunoprecipitates with Snf1, we found no Snf1-dependent phosphorylation of catalytically inactive gcn2-K628R. Interestingly, however, we observed a higher level of phosphorylation of active Gcn2 in such assays containing WT Snf1 versus those containing the snf1-T210A mutant.
One explanation for these last findings could be that the Snf1 KD interacts with the Gcn2 KD to mediate allosteric activation of Gcn2 autokinase activity or to block the access of a coimmunoprecipitating protein phosphatase that dephosphorylates the Gcn2 activation loop. Snf1 catalytic activity would be required for these putative regulatory functions involving physical interactions between the KDs of Snf1 and Gcn2, as the Snf1-T210A substitution reduces Gcn2 activation and lowers eIF2α-P but does not reduce Snf1 binding to Gcn2 in extracts (Fig. (Fig.5C).5C). Another possibility is that Snf1 phosphorylates Gcn2 in a manner that enhances Gcn2 autokinase activity, but the hypothetical Snf1-phosphorylated residue is labile in vitro and, hence, was undetectable in the gcn2-K628R of Fig. Fig.5C.5C. Additional work is required to elucidate the mechanism of Gcn2 activation by Snf1 in histidine-starved cells.
Although Snf1 stimulates Gcn2 function in histidine-starved cells on glucose, it might also negatively regulate an eIF2α-PPase to increase eIF2α-P under these conditions (Fig. (Fig.8A).8A). Our previous results indicated that Glc7 (29) and Sit4 (2) both downregulate eIF2α-P in glucose, and we demonstrated here that this occurs without reducing Gcn2 activity (T882-P level). This provides evidence that both enzymes function directly as eIF2α phosphatases, rather than dephosphorylating the Gcn2 activation loop, which is supported by our finding that eIF2α is physically associated with Glc7 and Sit4 in extracts. Thus, if Snf1 inhibited one of these enzymes during histidine starvation, it would contribute to the Snf1-dependent increase in eIF2α-P under that condition (Fig. (Fig.8A).8A). Although it appears that Snf1 stimulates, rather than inhibits, an eIF2α-PPase in the absence of histidine starvation on glucose medium, as reg1Δ reduces eIF2α-P without changing the activation level of Gcn2 (T882-P) (Fig. 3A and B), it is possible that the situation is reversed during histidine starvation.
We found that Snf1 is also required to maintain normal levels of eIF2α phosphorylation in cells shifted to galactose. Interestingly, this Snf1 function does not involve stimulation of Gcn2 activity, as T882-P levels are somewhat higher in snf1Δ mutant cells on galactose. Hence, Snf1 must promote eIF2α-P by downregulating eIF2α-PPase activity on galactose (Fig. (Fig.8B).8B). This function of Snf1 is also negatively regulated by Glc7/Reg1, as reg1Δ provokes a strong increase in eIF2α-P on galactose (without elevating T882-P) that is abolished in the snf1Δ reg1Δ double mutant.
We found that mutations impairing Glc7 or Sit4 elevate eIF2α-P in galactose, showing that both enzymes contribute to eIF2α dephosphorylation and, hence, are candidates for inhibition by Snf1 in this carbon source. If Sit4 was the sole eIF2α-PPase downregulated by Snf1 in galactose, then sit4Δ should abolish the effect of removing Snf1 on eIF2-αP levels. This was not the case, however, as sit4Δ snf1Δ mutant cells exhibited markedly lower levels of eIF2α-P compared to sit4Δ single mutants on galactose. This finding suggests that the eIF2α-PPase activity of Glc7 is also downregulated by Snf1 on galactose, which is supported by our finding that glc7-1 blunts the effect of snf1Δ in lowering eIF2α-P. The residual reduction in eIF2α-P levels produced by snf1Δ in glc7-1 mutant cells likely results from inhibition of Sit4 by Snf1 because snf1Δ evokes no reduction in eIF2α-P in glc7-1 sit4Δ mutant cells, where both phosphatases are impaired. Together, our results indicate that Snf1 inhibits the eIF2α-PPase activities of both Sit4 and Glc7 in galactose medium and that Glc7 is responsible for most of the eIF2α-P dephosphorylation seen under these conditions (Fig. (Fig.8B).8B). It is unknown what regulatory subunit Glc7 might utilize for its eIF2α-PPase activity (Fig. (Fig.8B).8B). It is unlikely to be Reg1, considering that the increases in eIF2α-P produced by reg1Δ in histidine-starved cells, or after a transfer to galactose, are fully suppressed by snf1Δ, indicating that Reg1 acts upstream of Snf1 in regulating eIF2α-P.
Can we rationalize the fact that Snf1 employs distinct mechanisms to promote eIF2α phosphorylation in histidine-starved versus galactose-grown cells, activating Gcn2 in the former and downregulating eIF2α-PPases in the latter? The fact that Snf1 stimulates Gcn2 activity only in histidine-starved cells might be explained by noting that the principal activating ligand for Gcn2, uncharged tRNA, accumulates to high levels under these conditions (34), but probably not after the cells are shifted to galactose. Hence, the Snf1 stimulation of Gcn2 might be observed only when uncharged tRNA accumulates and Gcn2 can be activated by this important stimulatory ligand (Fig. (Fig.8A).8A). On the other hand, the ability of Snf1 to effectively inhibit an eIF2α-PPase, as it does in galactose, might require the elevated Snf1 kinase activity that prevails only in the absence of glucose.
Besides its principal role in helping yeast cells adapt to glucose limitation and use alternative carbon sources, Snf1 has been implicated in responses to environmental stress (9). Our results extend Snf1's purview to include a key translational response to amino acid starvation. By stimulating eIF2α phosphorylation, Snf1 contributes to the reduction in general protein synthesis evoked by eIF2α-P, a common response to starvation and stress that conserves amino acids and facilitates the reprogramming of gene expression. Snf1's role in promoting eIF2α-P accumulation in galactose could be viewed as a means of reducing the energy consumption in protein synthesis during growth on a nonpreferred carbon source.
The other important consequence of eIF2α phosphorylation is induction of GCN4 translation and attendant transcriptional activation of amino acid biosynthetic enzymes by Gcn4. However, Arndt and colleagues reported (and our findings confirm) that eliminating Snf1 function does not reduce GCN4 translation in amino acid-starved cells. In fact, the translation of a GCN4-lacZ reporter is partially derepressed, as is that of numerous Gcn4 target genes, in cells lacking Snf1 function on glucose medium without amino acid limitation, conditions under which eIF2α-P is at basal levels (25). Thus, it seems likely that inactivation of Snf1 alters a step in translation initiation in a way that reduces the inhibitory effects of the GCN4 uORFs and partially derepresses GCN4 translation independently of high-level eIF2α-P. As noted above, the fact that snf1Δ leads to increased Gcn4 synthesis but confers a 3ATs phenotype is explained by the fact that Snf1 also stimulates the ability of Gcn4 to activate the transcription of HIS3 (14).
Together, the available data indicate that Snf1 interfaces with translational and transcriptional responses to amino acid starvation at multiple levels in budding yeast. In cells starved for histidine in minimal glucose medium, Snf1 kinase function is needed for the high-level activation of Gcn2 and eIF2α-P accumulation that downregulates general translation initiation under starvation conditions. In WT cells, this function of Snf1 is expected to also enhance the translational induction of GCN4. However, detecting the absence of this function in snf1 mutants might be confounded by the fact that Snf1 inactivation derepresses GCN4 translation by an unknown mechanism. In histidine-starved cells, Snf1 not only is required for high-level eIF2α phosphorylation by Gcn2 but also promotes transcriptional activation by Gcn4. In galactose medium, Snf1 maintains the appropriate level of eIF2α-P by inhibiting the eIF2α-PPase activities of Glc7 and Sit4. The phosphorylation levels of other substrates for these protein phosphatases are also likely enhanced by the activated form of Snf1 present in galactose and other nonpreferred carbon sources.
We thank Marian Carlson, Martin Schmidt, Sergei Kuchin, Kim Arndt, George Sprague, and John Cannon for gifts of strains and plasmids and Richard Maraia for kindly providing laboratory space for several experiments. We thank Karen Arndt for discussing results prior to publication.
This work was supported in part by the Intramural Program of the National Institutes of Health.
Published ahead of print on 19 April 2010.