Research ArticleSystems Biology

Phosphoproteomic analysis identifies proteins involved in transcription-coupled mRNA decay as targets of Snf1 signaling

See allHide authors and affiliations

Science Signaling  08 Jul 2014:
Vol. 7, Issue 333, pp. ra64
DOI: 10.1126/scisignal.2005000

Abstract

Stresses, such as glucose depletion, activate Snf1, the Saccharomyces cerevisiae ortholog of adenosine monophosphate–activated protein kinase (AMPK), enabling adaptive cellular responses. In addition to affecting transcription, Snf1 may also promote mRNA stability in a gene-specific manner. To understand Snf1-mediated signaling, we used quantitative mass spectrometry to identify proteins that were phosphorylated in a Snf1-dependent manner. We identified 210 Snf1-dependent phosphopeptides in 145 proteins. Thirteen of these proteins are involved in mRNA metabolism. Of these, we found that Ccr4 (the major cytoplasmic deadenylase), Dhh1 (an RNA helicase), and Xrn1 (an exoribonuclease) were required for the glucose-induced decay of Snf1-dependent mRNAs that were activated by glucose depletion. Unexpectedly, deletion of XRN1 reduced the accumulation of Snf1-dependent transcripts that were synthesized during glucose depletion. Deletion of SNF1 rescued the synthetic lethality of simultaneous deletion of XRN1 and REG1, which encodes a regulatory subunit of a phosphatase that inhibits Snf1. Mutation of three Snf1-dependent phosphorylation sites in Xrn1 reduced glucose-induced mRNA decay. Thus, Xrn1 is required for Snf1-dependent mRNA homeostasis in response to nutrient availability.

INTRODUCTION

The serine-threonine protein kinase Snf1 is activated by glucose depletion and other stresses in Saccharomyces cerevisiae (1). Snf1 is implicated in various processes, including metabolism of nonfermentable substrates (2, 3), regulation of phospholipid and fatty acid biosynthesis (4, 5), response to environmental stresses (6), cell cycle (7), invasive growth (8, 9), and meiosis and sporulation (10). Snf1 is a member of a family of adenosine monophosphate (AMP)–activated protein kinases (AMPKs). In animals and plants, AMPKs respond to fluctuations in the ratio of AMP to adenosine triphosphate (ATP) to regulate cellular metabolism and maintain energy homeostasis (1, 11). Inactivation of AMPK causes defects in mitosis and cell polarity in animals (1214) and prevents the developmental response to abscisic acid in plants (15). Human AMPK is implicated in various pathologies and is a potential target for pharmacological intervention (16).

Snf1 and AMPKs are heterotrimeric protein complexes consisting of a catalytic subunit (α) and two regulatory subunits (β and γ). The kinases Sak1, Tos3, and Elm1 in yeast or LKB1 and CaMKKβ in animals activate Snf1 and AMPKs by phosphorylating the activation loop (Thr210 in Snf1; Thr172 in animal AMPKs) of the kinase catalytic domain (1719). Mutation of either Thr210 or an essential residue in the active site of Snf1, Lys84, inactivates the enzyme (20). The binding of either AMP or ADP (adenosine diphosphate) to the γ (2123) or α (24) subunit prevents dephosphorylation of the activation loop and thus regulates the activity of the enzyme. In yeast, two protein phosphatases, protein phosphatase 1 (composed of the subunits Glc7 and Reg1) and protein phosphatase 2A (PP2A), inactivate Snf1 by dephosphorylating Thr210 in the activation loop (2527).

Snf1 and AMPKs influence several downstream pathways in mammals (16, 28) and yeast (1). Mass spectrometry (MS)–based phosphoproteomic studies have identified multiple direct and indirect substrates of AMPK family members (14, 15, 29, 30). Snf1-dependent phosphorylation of target proteins influences gene expression, including activation of genes that are repressed under glucose-replete conditions (also referred to as derepression) or induced by stress (1, 31, 32). Snf1 stimulates transcription by phosphorylating the zinc finger DNA binding protein Mig1, causing it to translocate to the cytoplasm from the nucleus, where it represses transcription of numerous genes (1, 3335). When glucose is depleted, Snf1 activates the transcription factors Adr1 and Cat8, causing transcription of more than 100 genes, many encoding enzymes that metabolize nonfermentable substrates (36). Moreover, artificially tethering Snf1 to a promoter by fusion to a DNA binding domain can activate gene transcription (37, 38), suggesting that Snf1 may have a direct role in transcription initiation.

Snf1 also plays a role in posttranscriptional regulation of gene expression. Culturing yeast in glucose-containing growth medium represses Snf1-dependent transcription of target genes and promotes turnover of the corresponding mRNAs in a process called glucose-induced mRNA decay (3942). Snf1-interacting proteins, in particular Reg1, are required for glucose-induced mRNA decay (43). Yeast with an analog-sensitive allele of SNF1, SNF1I132G (SNF1as) (44, 45), that are exposed to the inhibitory ATP analog 2-naphthylmethyl pyrazolopyrimidine 1 (2NM-PP1) exhibit rapid decay of Snf1-dependent transcripts in the absence of glucose (46). Thus, Snf1 is required to activate transcription and also to prevent mRNA decay after glucose depletion.

The synthesis of mRNA in the nucleus can be coupled to mRNA decay in the cytoplasm, a phenomenon known as transcription-coupled mRNA decay (4749). Xrn1, a yeast cytoplasmic 5′ to 3′ exoribonuclease, promotes mRNA homeostasis by balancing the rate of mRNA decay with the rate of mRNA synthesis (47, 48). Xrn1 has a widespread role in cytoplasmic mRNA decay (50, 51), including degradation of Snf1-dependent transcripts destabilized by glucose addition (43). However, whether the activity of Xrn1 is regulated, for example, to prevent degradation of Snf1-dependent transcripts during growth in the absence of glucose, is unknown.

To identify candidate Snf1 substrates involved in mRNA metabolism and other cellular processes, we performed proteomic analyses of wild-type and Snf1-inhibited yeast and studied changes in protein phosphorylation that occurred after glucose depletion. We discovered that Snf1 promoted phosphorylation of proteins involved in various pathways, including transcription-coupled mRNA decay and mRNA capping, deadenylation, degradation, and translation. We focused on the role of putative Snf1 substrates in mRNA homeostasis and showed that Snf1 genetically and functionally interacted with Xrn1 to promote transcription-coupled mRNA decay.

RESULTS

Phosphoproteomic analysis of yeast with inhibition of Snf1 activity

To identify Snf1-dependent phosphorylated proteins, we used MS to analyze phosphopeptide-enriched samples from snf1Δ yeast transformed with plasmids carrying wild-type SNF1, snf1K84R (kinase-inactive), or snf1T210A (activation-defective) or the parental plasmid (52) and incubated for 2 hours in low glucose–containing medium (glucose depletion) (fig. S1). Constitutive loss of Snf1 activity in snf1 mutant yeast could cause compensatory adaptations. Therefore, we also measured the abundance of phosphopeptides in yeast expressing an ATP analog–sensitive allele of SNF1 (SNF1as) and exposed to varying concentrations of the inhibitory ATP analog 2NM-PP1 for 30 min after 2 hours of incubation in low glucose–containing medium (fig. S1). We identified tandem MS (MS/MS) spectra corresponding to 2150 phosphopeptides at a false discovery rate (FDR) <1% in at least one replicate from at least one strain or 2NM-PP1 concentration analyzed (table S1).

To identify Snf1-dependent phosphorylation of proteins, we used label-free quantification of chromatograms based on full precursor scan (MS1) spectra to measure the relative abundances of phosphopeptides (52). We performed an analysis of variance (ANOVA) on the abundance of phosphopeptides in the wild-type, snf1 mutant, and SNF1as yeast data sets. In addition, to overcome the “missing data” problem frequently encountered in MS analyses, we used the G-test, which is a χ2 test modified to accommodate small numbers of sample replicates (53), to identify phosphopeptides having nonrandom missing blocks of data that were overlooked by the ANOVA due to the lack of sufficient data. Phosphopeptides with an ANOVA P value ≤0.05 and a twofold decrease in abundance in at least two replicates from at least two snf1 mutant strains and phosphopeptides with a G-test P value ≤0.05 that had missing data for all three replicates from two of three snf1 mutant strains and were present in at least two replicates from wild-type yeast were considered Snf1-dependent (table S2).

In total, we identified 177 phosphopeptides (128 proteins) that were reduced or absent in at least two snf1 mutant strains (table S2). There were 111 phosphopeptides in wild-type yeast that were not present in at least two snf1 mutant strains. These likely represent sites that are only phosphorylated by Snf1. There were 37 phosphopeptides that were present in all three snf1 mutant strains with abundances that were reduced at least twofold compared to wild type. This group likely represents sites that are phosphorylated by Snf1 and another kinase. There was extensive overlap among sets of proteins containing at least one phosphopeptide with altered abundance in the three snf1 mutant strains (Fig. 1A). The abundance of phosphopeptides for 38 proteins was absent or reduced in snf1K84R and snf1Δ yeast but not in snf1T210A yeast (Fig. 1A), suggesting that phosphorylation of these peptides may occur independent of Snf1 activation by upstream kinases. We found that 73 of 128 (57%) proteins corresponding to phosphopeptides that were absent or had reduced abundance in snf1Δ, snf1K84R, and snf1T210A yeast also had phosphopeptides that had at least a twofold decreased abundance in SNF1as yeast exposed to 10 μM 2NM-PP1 (Fig. 1B). The remaining phosphopeptides with decreased abundance in snf1 mutant strains that were unaffected in SNF1as yeast may not be dephosphorylated during the 30 min after addition of 2NM-PP1.

Fig. 1 The Snf1 phosphoproteome is linked to multiple cellular processes.

(A) Venn diagram illustrating the overlap of proteins containing phosphopeptides that changed in abundance in snf1 mutant yeast. (B) Venn diagram illustrating the overlap of proteins containing phosphopeptides with abundances that were reduced in at least two snf1 mutant yeast strains and at least threefold in SNF1as yeast exposed to 10 μM 2NM-PP1. For samples from snf1 mutant yeast with missing data, the phosphopeptides had a G-test P ≤ 0.05. For samples from snf1 mutant yeast with data, the phosphopeptides had a Wilcoxon rank sum test P ≤ 0.05 and abundance ratio (mutant/wild type) of ≤0.5. Snf1 was removed from these totals. (C) Interaction network of Snf1-dependent phosphorylated proteins (green) and known Snf1-interacting proteins (blue). Snf1-dependent phosphorylation is indicated by the blue dotted lines. RSC, RSC chromatin remodeling complex; SAGA/NuA4, histone acetyltransferase complexes; TORC, target of rapamycin complexes; PKA, protein kinase A; RNA polI & polIII, RNA polymerase I– and RNA polymerase III–associated proteins; SRP, signal recognition particle; Paf1/FACT, Paf1 and FACT (facilitates chromatin transcription) complexes; Phos. ino. met., phosphatidyl inositol metabolism; GARP, Golgi-associated retrograde protein complex; GID, glucose-induced degradation complex.

Phosphoproteomic analysis of yeast after glucose depletion followed by inhibition of Snf1

We predicted that direct Snf1 substrate proteins would be rapidly phosphorylated in glucose-depleted yeast; therefore, we monitored temporal changes in phosphopeptide abundance after glucose depletion. We collected samples from SNF1as yeast grown for 0, 5, 10, 20, or 40 min in low glucose–containing medium. In addition, to identify which of the resulting phosphopeptides were SNF1-dependent, we added 2NM-PP1 at 40 min and collected samples at 48 min (fig. S1). We identified 4967 phosphopeptides at an FDR <1% in at least one condition and one replicate (table S3). Although the abundance of most phosphopeptides was unchanged, 207 phosphopeptides increased in abundance at least threefold between 0 and 5 min, and 79 phosphopeptides decreased in abundance by at least threefold after exposing yeast to 2NM-PP1 for 8 min, starting at 40 min after glucose depletion. The abundances of phosphopeptides corresponding to 23% of proteins that were rapidly phosphorylated upon glucose depletion were also decreased by the subsequent 8-min exposure to 2NM-PP1 (fig. S2), suggesting that these proteins require Snf1 for phosphorylation.

We graphed the abundance of all of the Snf1-dependent phosphopeptides discovered from the analyses of wild-type and snf1 mutant yeast, after SNF1as inhibition, and after glucose depletion (File S1). For proteins that contained at least one Snf1-independent phosphosite, the abundance of the Snf1-independent phosphopeptide was also graphed. We estimated the half-life of the phosphosites for the Snf1-dependent phosphopeptides by comparing the abundance at 40 min, when 2NM-PP1 was added, to the abundance 8 min later, and found that the kinetics of dephosphorylation varied with apparent half-lives ranging from <1 to >12 min (table S2), indicating that different Snf1-dependent phosphorylated sites respond differently to Snf1 inactivation. Snf1-dependent phosphopeptides that were increased in abundance by glucose depletion but did not decrease with exposure to 2NM-PP1 might have slow rates of turnover or dephosphorylation.

Phosphopeptides corresponding to several known Snf1 substrate proteins, including Gal83 (54), Mig1 (33, 34), Rod1 (55), Sch9 (56), and Sip1 (57), increased in abundance more than threefold between 0 and 5 min after glucose depletion and decreased in abundance by at least threefold after inhibition of Snf1as by 2NM-PP1 for 8 min (fig. S3A). Phosphorylation of Gcn2, the nutrient-sensitive kinase that is activated by Snf1 under nitrogen-limiting conditions (58), was unaffected by glucose depletion (table S3). We found 33 phosphopeptides, corresponding to 17 proteins with abundances that met these criteria and were not detected in the first two MS experiments. We assumed that these proteins were likely Snf1-dependent phosphorylated proteins (table S2) (these were listed in boldface to indicate tentative assignment). Thus, in total in the three MS experiments, we identified 210 Snf1-dependent phosphopeptides from 145 proteins (table S2).

Not all known Snf1 substrates met all criteria for Snf1-dependent phosphorylation in our analysis. The abundance of a phosphopeptide corresponding to phosphorylation of Ser1157 on Acc1 was not increased after glucose depletion (fig. S3A) but was decreased in SNF1as yeast exposed to 2NM-PP1 (fig. S3, A and B) and in snf1 mutant yeast (fig. S3C). Thus, some of the putative Snf1 substrates identified in our study may have properties similar to Acc1.

Comparison of the Snf1 phosphoproteome with the Snf1 transcriptome

Decreased abundance of phosphopeptides in Snf1-inhibited yeast could be caused indirectly by decreases in total protein abundance resulting from changes in gene expression, mRNA stability, protein translation, or protein degradation. We found that 21 of 145 proteins found in our study are encoded by genes that exhibit decreased expression in the absence of Snf1 activity (36) (fig. S4), suggesting that the abundance of these phosphopeptides could be influenced by Snf1-dependent gene expression or by both Snf1-dependent gene expression and protein phosphorylation. Moreover, we found that 67 of 145 proteins with a Snf1-dependent phosphopeptide also had a phosphopeptide that was not affected by loss of Snf1 activity (file S1), suggesting that changes in the abundances of these phosphopeptides do not result from Snf1-dependent changes in gene expression or total protein abundance. These observations suggest that posttranslational modification, and not regulation of total protein abundance, explains the Snf1-dependent changes in the abundance of many of the identified phosphopeptides.

Functional annotation of Snf1-dependent phosphorylated proteins

We used bioinformatics analyses to characterize Snf1-dependent phosphorylated proteins. Gene ontology (GO) analysis using Saccharomyces Genome Database Slim Mapper (59, 60) revealed that the annotations “cytoplasmic mRNA processing body” and “cytoplasmic stress granule” were the most highly enriched among proteins corresponding to Snf1-dependent phosphopeptides (table S4), suggesting that Snf1 could be involved in mRNA processing and degradation. We created a protein interaction network composed of the interactions between Snf1 and the Snf1-dependent phosphorylated proteins that we identified by MS and protein interactions among these proteins retrieved from the STRING database (61). We found that 130 of the 145 Snf1-dependent phosphorylated proteins had interactions in addition to those identified here (Fig. 1C). Cluster analysis based on network connectivity and manual annotation identified these proteins as members of complexes involved in diverse cellular processes, including carbon metabolism, signal transduction, gene expression, growth, and membrane trafficking.

Sequence motif analysis of Snf1-dependent phosphopeptides

We analyzed the sequence of the Snf1-dependent phosphopeptides. Using Seq2Logo (62), we found that the consensus motif (Fig. 2A) derived from all Snf1-dependent phosphopeptides (table S2) differed from that reported previously (63, 64). Further analysis using MDDLogo, which clusters sequences into multiple groups (65), revealed multiple motifs, including ones with (73%) and ones without (27%) Pro at the +1 position (fig. S5). Analysis of the sequences of phosphorylated sites that lacked Pro at the +1 position using Seq2Logo showed an enrichment of Arg or Lys at the −3, Asp at the +3, and Leu at the +4 position (motif I, Fig. 2B). Analysis of the sequences of phosphorylated sites with Pro in the +1 position also showed an enrichment of Arg or Lys at the −3 position, but lacked enrichment of Asp at the +3 and Leu at the +4 position (motif II, Fig. 2C). The sequence of motif I was consistent with a published consensus motif derived from in vitro studies using Snf1 and AMPKα with a peptide library (63, 64, 66), suggesting that the phosphorylated residues in peptides containing motif I may be directly phosphorylated by Snf1.

Fig. 2 Snf1-dependent phosphorylation occurs at multiple motifs.

(A to H) Graphs of the relative enrichment of the indicated amino acids in the sequences surrounding Snf1-dependent phosphorylated sites (table S2) created using Seq2Logo (62). (A) Consensus motif derived from all Snf1-dependent phosphopeptides. (B) Consensus motif derived from phosphopeptides that lack a Pro at the +1 position (motif I). (C) Consensus motif derived from phosphopeptides that have a Pro at the +1 position (motif II). (D) Consensus motif derived from phosphopeptides that showed a more than twofold increase in abundance between 0 and 5 min in glucose-depleted yeast. (E) Consensus motif derived from phosphopeptides that showed a more than twofold decrease in abundance after inhibition of Snf1as for 8 min. (F) Consensus motif derived from phosphopeptides that were not affected by glucose depletion. (G) Consensus motif derived from phosphopeptides that had a half-life of greater than 8 min in Snf1as yeast exposed to 2NM-PP1. (H) Consensus motif derived from phosphopeptides that were reduced less than twofold in abundance in Snf1as yeast exposed to 10 μM 2NM-PP1.

We reasoned that the abundance of phosphopeptides that correspond to sites directly phosphorylated by Snf1 should rapidly increase in response to glucose depletion and have a short half-life after Snf1as inhibition with 2NM-PP1. Therefore, we identified phosphopeptides with abundances that increased more than twofold from 0 to 5 min after glucose depletion and decreased more than twofold after Snf1as inhibition (table S2). We analyzed the sequences using Seq2logo and found that these phosphopeptides contained a consensus sequence similar to motif I (Fig. 2, D and E) and included proteins known to be directly phosphorylated by Snf1, such as Acc1, Gal83, Mig1, Rod1, Sch9, and Sip1 (table S2). We also analyzed the sequences of phosphopeptides with abundances that were not changed by glucose depletion, had a half-life greater than 8 min after inhibition of Snf1as, or decreased less than twofold in SNF1as yeast exposed to 10 μM 2NM-PP1 (file S1). Phosphopeptides meeting these criteria had a consensus sequence similar to motif II (Fig. 2, F to H). These residues in the corresponding proteins are likely not direct Snf1 substrates, consistent with the observation that Snf1-containing protein complexes purified from yeast show a strong bias against phosphorylating peptides with Pro in the +1 position in vitro (64).

Functional analysis of Snf1-dependent phosphorylated proteins involved in mRNA metabolism

Snf1-dependent phosphorylated proteins could play a role in Snf1-dependent transcription or glucose-induced mRNA decay. Thirteen of the proteins identified in our proteomics analysis (Ccr4, Cdc73, Dcs2, Dhh1, Leo1, Mpt5, Nab6, Ngr1, Pbp1, Puf3, Scp160, Whi3, and Xrn1) were functionally annotated in the Saccharomyces Genome Database (61) as involved in mRNA metabolism. Mpt5 (also known as Puf5) (67), Nab6 (68), Ngr1 (69, 70), Puf3 (71), Scp160 (72), and Whi3 (73) are RNA binding proteins that function in multiple pathways. Dcs2 (74), Dhh1 (75), Pbp1 (70), and Xrn1 (76) participate in various steps of mRNA degradation. Cdc73 (77) and Leo1 (78) are members of the Paf complex that interacts with RNA Pol I and II and participate in histone methylation and transcription elongation (79). The abundance of phosphopeptides corresponding to Ccr4, a component of the CCR4-NOT1 complex that is involved in transcriptional regulation and polyadenylate tail shortening (80, 81), was rapidly decreased after glucose depletion and increased in snf1 mutant yeast and in SNF1as yeast exposed to 2NM-PP1 (file S1), suggesting that its dephosphorylation may be regulated by Snf1. Therefore, we focused on whether these proteins were required for Snf1-dependent cellular functions.

Snf1 is required for metabolism of poorly fermented and nonfermentable carbon substrates (2). Therefore, we investigated whether Snf1-dependent phosphorylated proteins were required for the growth of yeast on glycerol, ethanol, or lactate (nonfermentable carbon sources). Consistent with previous studies (82, 83), deletion of SNF1 completely blocked growth on nonfermentable carbon sources (Fig. 3A). Similarly, ccr4Δ, dhh1Δ, scp160Δ, and xrn1Δ yeast grew more slowly than wild-type yeast on glycerol or ethanol (Fig. 3A). Similar to snf1Δ, dhh1Δ yeast did not grow on lactate (Fig. 3A). The growth of snf1Δ, dhh1Δ, and scp160Δ yeast on glucose was reduced by incubation at a higher temperature (Fig. 3A), indicating that, like Snf1 (6), Dhh1 and Scp160 are important for growth during heat stress. The similarities in the patterns of growth inhibition on different carbon sources among snf1Δ yeast and yeast deficient for Snf1-dependent phosphorylated proteins suggest that these proteins may function downstream of Snf1. Moreover, the time- and Snf1-dependent phosphorylation of Dhh1, Scp160, and Xrn1, or dephosphorylation in the case of Ccr4 (Fig. 3B), is consistent with the activity of these proteins being regulated during glucose depletion.

Fig. 3 Snf1-dependent phosphorylated proteins are required for growth and the metabolism of Snf1-dependent transcripts.

(A) Growth of yeast with the indicated deletions on plates containing the indicated carbon sources at the indicated temperatures. (B) Graph of the abundances of the phosphopeptides from the indicated proteins during glucose depletion from 0 to 40 min and inhibition of Snf1as by 2N-PP1 from 40 to 48 min. (C) Graph of the abundance of ADH2 mRNA in the indicated deletion mutant yeast after 4 hours of glucose depletion. The mRNA abundance was normalized to ACT1 and to normalized abundance in wild-type (WT) yeast (W303-CH1a or CKY19) at 0 min (t0) measured in the same experiment. Data are means ± SD of n ≥ 3 biological replicates. Two xrn1Δ strains were analyzed (KBY141 and KBY155). (D) Graph of glucose-induced decay of ADH2 mRNA in WT yeast and yeast with the indicated deletions. Data are from the same experiments described in (C). (E) Half-life of ADH2 mRNA calculated from experiments represented in (D). ***P < 0.01, Wilcoxon rank sum test (141).

To determine whether Snf1-dependent phosphorylated proteins were involved in Snf1-dependent gene expression, we measured the abundance of transcripts from ADH2, ACS1, and FBP1, which are Snf1-regulated genes activated by glucose depletion (84). Four hours after glucose depletion, the expression of ADH2 (Fig. 3C), ACS1, and FBP1 (table S5) was reduced 8- to 30-fold in ccr4Δ and xrn1Δ compared to wild-type yeast. This result in ccr4Δ yeast is consistent with the observation that the originally identified defective ccr4 allele has reduced activity of alcohol dehydrogenase II, the product of ADH2 (85). The expression of FBP1 was reduced 35-fold in whi3Δ yeast compared to wild-type yeast, but ADH2 and ACS1 expression was less affected (table S5). Deletion of the genes encoding Cdc73, Dcs1, Dcs2, Leo1, Nab6, Ngr1, Puf3, and Pbp1 had a smaller or no effect on the expression of ADH2, ACS1, and FBP1 (table S5). Thus, Snf1-dependent transcription may depend on proteins involved in posttranscriptional events, including mRNA decay.

To determine whether the stability of Snf1-dependent transcripts was affected in yeast strains with deletions of Snf1-dependent phosphorylated proteins involved in mRNA metabolism, we grew yeast under glucose-depleted conditions for 4 hours and then added glucose to induce mRNA decay and measured the abundance of the ADH2 transcripts (86, 87). In wild-type yeast, the addition of glucose reduced ADH2 transcript abundance by more than 90% within 20 min (Fig. 3D). In ccr4Δ and dhh1Δ yeast, glucose-induced ADH2 mRNA decay was reduced, and in xrn1Δ yeast, the abundance of ADH2 mRNA did not decrease during the 20 min after glucose addition (Fig. 3D). The apparent half-life of ADH2 mRNA was increased at least twofold in ccr4Δ and dhh1Δ and was greater than 20 min in xrn1Δ, compared to a 5-min half-life in wild-type yeast (38) (Fig. 3E). Deletion of the genes encoding Scp160 (Fig. 3, D and E), Cdc73, Dcs2, Leo1, Mpt5, Nab6, Ngr1, Pbp3, Puf3, and Whi3 had little or no effect on ADH2 mRNA decay (fig. S6, A and B). Thus, Xrn1 and to a lesser degree Ccr4 and Dhh1 are involved in promoting mRNA accumulation and glucose-induced mRNA decay of Snf1-regulated genes.

Validation of the role of Xrn1 in transcription and glucose-induced mRNA decay of Snf1-dependent genes

We focused on Xrn1 because deletion of XRN1 had the strongest effect on ADH2 mRNA decay in the presence of glucose and Xrn1 is involved in transcription-coupled mRNA decay (47, 48). We found that deletion of XRN1 influences the transcript abundance during glucose depletion and the glucose-induced mRNA decay of multiple Snf1-dependent genes (fig. S7). We also assayed Snf1-dependent transcription in wild-type, xrn1Δ, and adr1Δ strains using an Adr1- and Snf1-dependent (88) artificial reporter gene in which the promoter of ADH2 was fused to the coding region of lacZ (89). Consistent with previous results (89), activation of ADH2-lacZ by glucose depletion was decreased in adr1Δ yeast (Fig. 4A). Activation of ADH2-lacZ by glucose depletion was also reduced in xrn1Δ yeast (Fig. 4A), suggesting that the ability of Xrn1 to modulate the abundance of ADH2 mRNA is linked to transcription from the ADH2 promoter and not to potential effects on ADH2 mRNA degradation. We also asked whether activation of ADH2-lacZ in the presence of glucose was dependent on Xrn1. We activated ADH2-lacZ by transforming adr1Δ yeast with a plasmid encoding glucose-insensitive Adr1 (ADR1c) or by overexpressing ADR1 from the ADH1 promoter (oeADR1) (9094). Unlike adr1Δ yeast expressing wild-type Adr1, those with glucose-insensitive Adr1 or overexpressing Adr1 activated the ADH2-lacZ reporter in the presence of glucose (Fig. 4B). Moreover, deletion of XRN1 did not significantly affect activation of ADH2-lacZ by expression of glucose-insensitive Adr1 or overexpression of Adr1 in the presence of high glucose (Fig. 4B), suggesting that the effect of Xrn1 on ADH2 mRNA abundance is affected by the availability of glucose.

Fig. 4 Xrn1 requires its exonuclease activity to promote transcription and decay of Snf1-regulated genes.

(A) Graph of β-galactosidase activity from the ADH2-lacZ reporter induced by depletion of glucose in WT, adr1Δ, or xrn1Δ yeast. Data are means ± SD of two experiments. (B) Graph of β-galactosidase activity from the ADH2-lacZ reporter in adr1Δ xrn1Δ yeast transformed with plasmids encoding WT Adr1 or glucose-insensitive Adr1 in the form of constitutively active Adr1 (ADR1S230A, ADR1c) or Adr1 expressed from the ADH1 promoter (oeADR1). Data represent the means ± SD of three transformants per plasmid in two experiments. (C) Graph of the abundance of ADH2 transcripts in xrn1Δ yeast (KBY155) derived from W303-CH1a × KBY212 diploid yeast transformed with the indicated plasmids and grown in synthetic medium (SM)–Leu with high glucose (5%) or depleted of glucose (0.05%). Data are the means ± SD of four transformants per plasmid in two independent experiments. (D) Graph of glucose-induced ADH2 mRNA decay in xrn1Δ (KBY155) yeast transformed with the indicated plasmids. Data are expressed as a percent of normalized abundance of ADH2 in yeast 5 min (t5) after transfer from low glucose (0.05% for 4 hours) to high glucose (5%). Data are representative of two independent experiments. (E) Growth of xrn1Δ (KBY155) yeast transformed with the indicated plasmids grown on the indicated carbon source at 30°C. (F) xrn1Δ yeast transformed with the indicated plasmids [derived from a KBY155 (xrn1Δ) × CHY36a (reg1Δ) diploid] were grown in YPD (yeast extract, peptone, and dextrose) for about 30 generations and plated onto YPD and SD-Leu plates. Plasmid loss is reported as the percentage of colonies that did not grow on SD-Leu. Data are means ± SD of two biological replicates.

We further validated the role of Xrn1 in the metabolism of Snf1-dependent transcripts by testing whether the defects in xrn1Δ yeast were caused solely by the loss of Xrn1 activity. A yeast strain with an XRN1 deletion (KBY155) was transformed with a plasmid expressing wild-type XRN1 or the parent plasmid as a negative control. Similar to the results in Fig. 3C and table S6, xrn1Δ yeast transformed with the parent plasmid had a reduced abundance of ADH2 transcripts after glucose depletion (Fig. 4C), and ADH2 mRNA decay was slower in the presence of glucose (Fig. 4D). In contrast, xrn1Δ yeast transformed with a plasmid encoding wild-type Xrn1, but not Xrn1 with a Glu176-to-Gly mutation in its active site (XRN1E176G), which abolishes its 5′ to 3′ exonuclease activity (75), had increased abundance of ADH2 transcripts during glucose depletion (Fig. 4C) and rapid decay of ADH2 transcripts in the presence of glucose (Fig. 4D). Thus, the production and decay of the Snf1-dependent ADH2 transcripts depends on the exonuclease activity of Xrn1.

Rat1 is a nuclear localized 5′ to 3′ exoribonuclease that is essential for the survival of yeast (95) and shares homology to Xrn1 in its exonuclease domain (96). To determine whether Rat1 could functionally substitute for Xrn1 in the metabolism of Snf1-dependent transcripts if localized to the cytoplasm, we transformed xrn1Δ yeast with a plasmid encoding Rat1ΔNLS, which lacks its nuclear localization sequence (96), and found that ADH2 expression (Fig. 4C) and decay (Fig. 4D) were similar to that in xrn1Δ yeast transformed with wild-type Xrn1, supporting a role for exonuclease activity in the metabolism of Snf1-regulated transcripts.

We analyzed the growth of xrn1Δ yeast expressing XRN1, XRN1E176G, or RATΔNLS. The xrn1Δ yeast (derived from an xrn1Δ × reg1Δ diploid) transformed with the parent plasmid or a plasmid encoding XRN1E176G grew slower than those transformed with plasmids encoding Xrn1 or Rat1ΔNLS on both fermentable (glucose) and nonfermentable (glycerol) carbon sources (Fig. 4E), consistent with Xrn1-dependent mRNA degradation playing an important role in carbon metabolism.

Functional analysis of Snf1-dependent phosphorylation of Xrn1

We examined whether the Snf1-dependent phosphorylated sites in Xrn1 were required for its role in glucose-induced mRNA decay. We mutated Xrn1 Ser1330, Ser1505, and Ser1510 to Ala (XRN1-3A) to prevent phosphorylation or to Asp (XRN1-3D) to mimic phosphorylation. Unlike the expression of wild-type XRN1 or XRN1-3A, the expression of XRN1-3A did not restore glucose-induced ADH2 mRNA decay (Fig. 5A).

Fig. 5 Xrn1 requires Snf1-dependent phosphorylation to promote glucose-induced ADH2 mRNA decay.

(A and B) Graphs of the normalized abundance of ADH2 mRNA in yeast grown in high glucose (5%) for the indicated times after transfer from low glucose (0.05% for 4 hours). (A) xrn1Δ yeast (KBY155) were transformed with the indicated plasmids: XRN1-3A (S1329A, S1505A, S1510A), XRN1-3D (S1329D, S1505D, S1510D), XRN1-7A (S1329A, S1330A, S1500A, S1501A, S1505A, T1506A, S1510A), and XRN1-7D (S1329D, S1330D, S1500D, S1501D, S1505D, T1506D, S1510D). Data are means ± SD of two independent experiments. (B) WT (W303-CH1a) or XRN1ΔC (KBY144) yeast. Data are means ± SD of two biological replicates. For (A) and (B), t0 = 0 min and t5 = 5 min in high glucose.

Because Ser1330, Ser1505, and Ser1510 are located in peptides that have adjacent Ser or Thr, our MS-based analyses did not unambiguously identify the specific site of phosphorylation. In addition, mutation of a residue that is normally phosphorylated could result in compensatory phosphorylation of adjacent residues. Therefore, we simultaneously mutated Ser1330, Ser1505, and Ser1510 and four surrounding residues that were annotated as putative phosphorylation sites in the PhosphoGRID database (97), including Ser1330, Ser1501, Thr1506, and Ser1510, to Ala (XRN1-7A) or Asp (XRN1-7D). Similar to XRN1-3A, but unlike wild-type XRN1 or XRN1-7D, XRN1-7A expression did not rescue glucose-induced ADH2 transcript decay (Fig. 5B) in xrn1Δ yeast. Because the XRN1 mutants with seven mutated residues exhibited the ability (Asp mutants) or inability (Ala mutants) to rescue xrn1Δ to a similar extent as the Xrn1 mutants with three mutated residues, it is likely that the three residues identified by the phosphoproteomic analysis (Ser1330, Ser1505, and Ser1510) are the relevant sites for Snf1-dependent regulation.

The C terminus of Xrn1 is not required for its exonuclease activity, but when a protein fragment representing the C terminus is overexpressed, it inhibits the growth of yeast (96). Because Snf1-dependent phosphorylated sites were located in the C terminus of Xrn1, we tested whether this region was required for the decay of Snf1-dependent transcripts. We found that yeast with a deletion of the C terminus of XRN1 (XRN1-ΔC, amino acids 1242 to 1528) showed glucose-induced ADH2 mRNA decay (Fig. 5B) similar to the wild-type yeast, suggesting that Snf1-dependent phosphorylation of the C terminus of Xrn1 could relieve inhibition of glucose-induced mRNA decay.

Epistatic analysis of Snf1 and Xrn1

Our MS data showed that Snf1 enhanced phosphorylation of Xrn1, and expression of Xrn1 with mutations that inhibit Snf1-dependent phosphorylation did not complement the deletion of XRN1 in assays of glucose-induced mRNA decay, implying that Xrn1 functions downstream of Snf1. To further assess the epistatic relationship of SNF1 and XRN1, we attempted to delete REG1, which encodes a phosphatase that inhibits Snf1 activity (25, 27, 98), in xrn1Δ yeast. However, we could not isolate reg1Δ xrn1Δ haploid yeast by direct transformation (99), suggesting that the double mutant was not viable. Therefore, we created reg1Δ xrn1Δ diploid yeast and genotyped individual cells dissected from the tetrad cluster that forms after meiosis and sporulation. We found no reg1Δ xrn1Δ cells; however, we found that about one-third of cells were wild type for XRN1 and REG1, and about two-thirds of cells were either reg1Δ or xrn1Δ (Fig. 6A), confirming that the double mutant was synthetic lethal.

Fig. 6 Simultaneous deletion of XRN1 and activation of Snf1 by deletion of REG1 is synthetic lethal.

(A to C) Analysis of the genotypes of individual spores dissected from tetrads resulting from sporulated xrn1Δ-XRN1 reg1Δ-REG1 (KBY141 × CHY36a) diploid yeast. The numbers in parenthesis indicate the total number of tetrads analyzed, and the numbers in the tables show the number of individual spores with the indicated genotype. (B) One copy of SNF1 was deleted. (C) Yeast were transformed with the indicated plasmids, and the fractions indicate the number of viable xrn1Δ reg1Δ spores divided by the total number of spores expected.

To address whether the synthetic lethality of reg1Δ and xrn1Δ depended on Snf1, we deleted SNF1 in reg1Δ xrn1Δ diploid yeast and performed tetrad analysis. We found that triple mutant snf1Δ xrn1Δ reg1Δ cells were viable at the expected ratio relative to cells with other genotypes (Fig. 6B), suggesting that loss of Snf1 conferred resistance to the synthetic lethality of XRN1 and REG1 deletion.

We also asked whether cytosolic exoribonuclease activity was required for viability in reg1Δ yeast. We transformed heterozygous reg1Δ xrn1Δ diploid yeast with the parent plasmid or plasmids encoding Xrn1, Xrn1E176G, or Rat1ΔNLS and performed tetrad analysis. Expression of XRN1, and to a lesser degree XRN1E176G and RAT1ΔNLS, restored the viability of reg1Δ xrn1Δ cells (Fig. 6C), suggesting that there are exoribonuclease-dependent and exoribonuclease-independent functions for Xrn1 in yeast with activated Snf1.

We directly tested whether Xrn1 affects the abundance or activity of Snf1. The total abundance and phosphorylation of Thr210 of Snf1 after 2 hours of glucose depletion was similar in wild-type and xrn1Δ yeast (fig. S8A). Moreover, the in vitro kinase activity of purified Snf1as from xrn1Δ yeast was similar to that in wild-type yeast (fig. S8B), suggesting that Xrn1 is not required for the transcription, translation, stability, or activation of Snf1 and consistent with a role for Xrn1 downstream of Snf1-dependent phosphorylation.

DISCUSSION

We used phosphoproteomics to identify Snf1-dependent phosphorylated proteins. We compared the proteins corresponding to Snf1-dependent phosphopeptides identified in our study with those from other MS-based studies of snf1Δ yeast (30, 100, 101). We found that 7.6 to 30% of the proteins identified here (table S2) overlapped with those found in other studies (figs. S9 to S13). Only two proteins, Pfk2 and Mig1, were common among all four studies. Methodological differences, experimental variability, and differences in growth conditions and yeast strains may partially account for the differences among these data sets. We found that 12 proteins, including Pfk2 and Mig1, identified here were also present in a list of 80 proteins that are potential direct targets of Snf1 (66) (fig. S13), suggesting that Pfk2 and Mig1 are central targets of Snf1 phosphorylation in multiple growth conditions.

The identity of Snf1-dependent phosphorylated proteins suggests diverse roles for Snf1 in cell physiology and metabolism. Several pathways contained multiple previously uncharacterized Snf1 target proteins (fig. S14), including proteins involved in phosphoinositide biosynthesis, an important source of signaling metabolites (102104), and glycolysis. For example, two enzymes important for proper physiological regulation of the glycolytic pathway, phosphofructokinase (Pfk1 and Pfk2) and pyruvate dehydrogenase (the subunit Pda1) (105108), were phosphorylated in a Snf1-dependent manner (Ser163 of Pfk2 and Ser313 of Pda1). In animal cells, AMPK phosphorylates and activates 6-phosphofructose-2-kinase (PFK-2), the isoform of phosphofructokinase that synthesizes fructose 2,6-bisphosphate, a potent stimulator of glycolysis (109). The orthologous yeast enzyme Pfk26 is activated by phosphorylation of Ser644 by PKA (110). Phosphorylation of a different residue, Ser667, was increased 12-fold in snf1Δ yeast grown in high glucose (29). During glucose depletion, when PKA activity is reduced (45), phosphorylation of Ser667 was unaffected by the loss of Snf1 activity (table S1), but phosphorylation of Ser644 increased more than 10-fold in wild-type yeast (table S3). These results suggest that Snf1 could influence the phosphofructokinase step in glycolysis through different mechanisms during fermentative versus gluconeogenic growth. These examples suggest that Snf1 participates in reducing the activity of the glycolytic cycle by phosphorylating, directly or indirectly, key regulatory enzymes.

Snf1 could affect cell growth and physiology through multiple signaling pathways. We identified two subunit proteins, Kog1 and Tco89, of the target of rapamycin complex 1 (TORC1) (111, 112) and multiple TORC1 effectors, including Kns1, Npr1, Tip41, Rtg3, Gat1, Gis1 (113), and Sch9, which is a known Snf1 substrate (56), as being phosphorylated in a Snf1-dependent manner. Thus, Snf1 may influence TORC1 activity directly by promoting phosphorylation of Kog1 and Tco89, and indirectly by promoting phosphorylation of proteins acting downstream of TORC1. AMPK inhibits the activity of the mammalian target of rapamycin (mTOR) by phosphorylating Raptor, an ortholog of Kog1, making it a substrate for 14-3-3 proteins (63). Two proteins, Bcy1 and Cyr1, which regulate PKA (45), and several other protein kinases, including Akl1, Hrk1, Kin82, Psk1, Ptk2, and Tda1, were also phosphorylated in a Snf1-dependent manner (table S2).

Snf1 may also affect the activity of protein phosphatases. Four regulatory subunits of Glc7 (Reg1, Gip2, Scd5, and Ypi1) (114); Tip41, a component of the PP2A pathway (115); and two regulatory subunits of calcineurin, Rcn1 and Rcn2 (116), were phosphorylated in a Snf1-dependent manner. Phosphorylation of one of the Snf1-dependent sites in Rcn1 (Ser113) by Mck1, a protein kinase related to GSK3 (glycogen synthase kinase 3), is important in calcineurin signaling (116), suggesting that Snf1 may activate Mck1 during glucose depletion. Snf1 is inactivated by Glc7 and PP2A (25, 26), suggesting that Snf1-dependent phosphorylation of associated proteins may prevent dephosphorylation of Snf1. AMPK promotes the completion of mitosis by phosphorylating a regulatory subunit of the PP1 in mammals (14). Thus, Snf1, and by analogy AMPK in plant and animal cells, may regulate a complex network of reactions in the cell by influencing the activity of numerous protein kinases and protein phosphatases.

Network analysis of Snf1-dependent phosphorylated proteins and their first-order interacting proteins showed a high degree of interaction. The affinity of direct protein-protein interactions could be modified by Snf1-dependent phosphorylation of one or more partner proteins. In addition, indirect interactions in this network could be the result of Snf1-dependent phosphorylation of RNA Pol II coactivators, corepressors, and transcription factors. For example, Cyc8 (also known as Ssn6) forms a conserved corepressor complex with Tup1 and is recruited to gene promoters by DNA binding proteins, including Mig1 (1). Multiple motif II–containing phosphopeptides from Cyc8 decreased in abundance in snf1K84R and snf1Δ yeast, but not in snf1T210A yeast, suggesting that Snf1-dependent phosphorylation of Cyc8 is independent of phosphorylation of the Snf1 activation loop. Consistent with this model, Cyc8 phosphopeptides were present in yeast grown in glucose and did not increase when Snf1 was activated by glucose depletion (file S1). Cyc8 phosphorylation in the presence of high glucose depends on both Snf1 and the LAMMER kinase Kns1 (29). The Kns1 ortholog in Schizosaccharomyces pombe phosphorylates and inactivates two Tup1 orthologs (117). Kns1 was phosphorylated in a Snf1-dependent manner (table S2 and file S1), suggesting that it may be activated by Snf1 and subsequently phosphorylate Cyc8. Thus, in addition to the genetic interaction of SNF1 with CYC8 and TUP1 (114), Snf1 may have a direct role in regulating the activity of the Cyc8-Tup1 corepressor complex.

The numerous proteins that were phosphorylated (or dephosphorylated in the case of Ccr4) in a Snf1-dependent manner suggest a link between mRNA decapping and Xrn1-mediated mRNA decay. Xrn1 is involved in transcription-coupled decay, a process that balances mRNA synthesis and decay (4749). On the basis of evidence from metabolic labeling and extensive microarray analyses, Sun et al. (48) concluded that Xrn1 is unique in this regard among ~40 genes encoding proteins involved in mRNA processing. Deletion of XRN1 reduced the rate of mRNA decay and enhanced the rate of mRNA synthesis, leading to a fourfold increase in mRNA abundance. These authors concluded that Xrn1 indirectly inhibits mRNA synthesis by repressing expression of the gene encoding the global repressor Nrg1. Haimovich et al. (47) analyzed mRNA decay and synthesis using deletion mutants of about six genes involved in mRNA processing and concluded that Xrn1 is involved in transcription-coupled mRNA decay. In contrast to Sun et al. (48), Haimovich et al. (47) concluded that Xrn1 was a direct activator of mRNA synthesis. Their different conclusions may be due to using different techniques to study mRNA synthesis. Both studies used yeast grown in the presence of high glucose, and thus, their studies would not have included Snf1-dependent genes. We found that Xrn1 was required to maintain abundance of Snf1-regulated transcripts in glucose-depleted yeast, supporting a role for Xrn1 in promoting transcription of these genes.

Our genetic and biochemical studies suggest an important role for Snf1-dependent phosphorylation of Xrn1. Yeast expressing XRN1 mutants lacking the Snf1-dependent phosphorylation sites (XRN1-3A and XRN1-7A) had reduced glucose-induced ADH2 mRNA decay. Yeast expressing phosphomimetic XRN1 mutants (XRN1-3D and XRN1-7D) or those with XRN1 with a C-terminal deletion were similar to wild-type yeast in transcript decay, suggesting that phosphorylation of the C terminus of Xrn1 is important to maintain its function in glucose-induced mRNA decay. Xrn1 is homologous to Rat1 except that Xrn1 has an extended C terminus. Rat1 localized to the cytoplasm rescued growth, mRNA synthesis, and decay phenotypes in yeast with XRN1 deletion, suggesting that the C terminus of Xrn1 confers a distinct function that may be regulated by Snf1-dependent phosphorylation. The C terminus of Xrn1 is not required for its exoribonuclease activity, but when overexpressed, it inhibits the growth of yeast (118). An important genetic interaction between SNF1 and XRN1 was revealed by synthetic lethality. We were not able to delete REG1, which encodes a protein phosphatase regulatory subunit that inhibits Snf1, in the absence of XRN1. The synthetic lethality of xrn1Δ reg1Δ yeast was rescued by deleting SNF1, demonstrating that yeast containing activated Snf1 require the presence of Xrn1 to maintain viability. This observation is consistent with the growth defect of xrn1Δ mutant yeast in nonfermentable carbon sources because Snf1 is active in these conditions. Thus, Xrn1 may provide an essential function related to transcription-coupled mRNA decay when Snf1 is activated.

There may be a link between the C-terminal region of Xrn1 (where Snf1-dependent phosphorylation occurs), mRNA decapping, and transcription termination. Recent studies indicate that Drosophila melanogaster and human orthologs of Xrn1 interact through their C termini with the decapping protein Dcp1 (119). We identified Dhh1 as a potential substrate of Snf1. Dhh1 promotes decapping by various mechanisms (120124) and interacts with the Ccr4-Not complex, a global regulator of deadenylation and transcription (125). Ccr4 was dephosphorylated in a Snf1- and glucose depletion–dependent manner, suggesting that its function may be indirectly regulated by Snf1. Dcs2, an inhibitor of Dcs1, a scavenger decapping enzyme that is an activator of Xrn1 both in vivo and in vitro (126), is a Snf1 target, again associating Snf1 and mRNA decapping. Nuclear mRNA decapping and degradation inhibits transcription elongation in human cells and promotes premature termination (127). We also identified Snf1-dependent phosphorylation of several translation initiation factors involved in cap metabolism and recognition, including Cet1, Eap1, Tif4631, Tif4632, and Tif5. Thus, these results strongly suggest a connection between Snf1 and molecular mechanisms involving the mRNA 5′-cap. Future studies will undoubtedly clarify the mechanistic details of Xrn1 in Snf1-dependent regulation of mRNA metabolism.

MATERIALS AND METHODS

Yeast strains and culture conditions

Yeast strains were constructed by one-step gene disruption (99, 128) or by mating and tetrad dissection (129) (table S6). The parental strain was W303-CH1a (130). In the case of one-step gene disruptions, the open reading frame was replaced by a gene encoding proteins that enable resistance to either hygromycin or kanamycin and confirmed by polymerase chain reaction (PCR). Deletion of the C-terminal 286 amino acids of XRN1 was created by PCR and epitope tagging (128) with 3×HA (hemagglutinin) at amino acid 1242. Yeast were grown on a rotary shaker at 250 rpm and 30°C in either yeast extract–peptone or SM (129) supplemented with 5% glucose (high glucose) or 0.05% glucose (low glucose, glucose depletion). Cultures of yeast containing replicating plasmids (table S7) were maintained in SM supplemented with the appropriate amino acids, adenine, and uracil. To maintain yeast with plasmids containing TRP1, 0.1% casamino acids (Sigma) were used.

To generate xrn1Δ yeast expressing XRN1 or RAT1, MATa wild-type (W303-CH1a) yeast were transformed with the plasmids shown in table S7 by a standard protocol (131) and mated with MATα xrn1Δ::hphMX (KBY212). The resulting diploid yeast were sporulated, and strains were identified by tetrad dissection and phenotypic analysis on drop-out plates (129). The parental plasmid was pRS315 (132). Commercial deletion strains were obtained from Research Genetics. Colony growth assays were performed by making serial fivefold dilutions of yeast cultures in microtiter plates and spotting 3 μl onto agar plates containing various carbon sources and incubating the plates for 3 days at either 30°C or 37°C.

Plasmid construction

All plasmids and their sources are shown in table S7. The plasmids encoding XRN1 with mutations in putative phosphorylated residues were created by site-directed mutagenesis of the plasmid encoding HA-tagged Xrn1 (pAJ152) using the QuikChange XL protocol (Agilent Technologies, cat. no. 200521). Plasmids were sequenced for the mutations, and the C-terminal fragment of XRN1 containing the mutations was subcloned using Aat II and Hind III back into the parental vector (pAJ152).

Phosphopeptide enrichment, MS, and statistical analyses

For experiment 1 with snf1 mutant yeast (fig. S1), three replicate cultures of strain CKY18 (snf1Δ) carrying URA3-CEN plasmids with no SNF1 (parental plasmid, pRS316), wild-type SNF1, snf1K84R, or snf1T210A (table S7) were grown in supplemented SM with 5% glucose to an absorbance at 600 nm of about 0.8, pelleted, and resuspended in supplemented SM with 0.05% glucose for 2 hours.

For experiment 2 using Snf1as yeast cultured with different concentrations of 2NM-PP1 (fig. S1), three isogenic SNF1as strains (TYY923, TYY924, and TYY925) were grown in SM with 5% glucose to an absorbance at 600 nm of about 0.8, pelleted, and resuspended in SM with 0.05% glucose. After 2 hours, each culture was divided into four 50-ml aliquots. The vehicle control (dimethyl sulfoxide, 1%) or 2NM-PP1 was added to a final concentration of 0.50, 1.0, and 10 μM for 30 min.

For experiment 3 with the glucose depletion time course in SNF1as yeast (TYY923) (fig. S1), three replicate cultures were grown in SM (300 ml) with 5% glucose to an absorbance at 600 nm of about 0.8. Aliquots (50 ml) were removed from each culture for analysis. The remaining cells were pelleted and resuspended in SM (250 ml) with 0.05% glucose. Aliquots (50 ml) were removed from each culture at 5, 10, 20, and 40 min. At 40 min, 2NM-PP1 (10 μM) was added for 8 min.

Yeast were mixed with trichloroacetic acid (TCA; 6.25%) and collected by centrifugation. Protein extracts were prepared and digested with trypsin (Sequencing Grade Modified Trypsin, Promega), and phosphopeptides were enriched using metal affinity chromatography with titanium dioxide resin (Titansphere TiO 5 μm, GL Sciences) as described previously (63).

Phosphopeptide-enriched samples from both the snf1 mutant yeast experiment and the SNF1as yeast experiment with different concentrations of 2NM-PP1 were separated by nanoscale reversed-phase high-pressure liquid chromatography with a Proxeon Easy-nLC (Thermo Fisher Scientific) system, ionized with a Thermo Nanospray source, and analyzed on an LTQ-FT mass spectrometer (Thermo Fisher Scientific). A 90-min elution gradient from 2 to 24% acetonitrile in acidic water (0.1% formic acid) was used. Peptides of 350 to 1600 mass/charge ratio (m/z) were analyzed, and the five most intense ions from the precursor scans were selected for fragmentation and identification. The third set of samples from the glucose depletion time course experiment was analyzed in an LTQ-Orbitrap XL (Thermo Fisher Scientific). Phosphopeptides were separated by a Proxeon Easy-nLC high-pressure liquid chromatography system. A 90-min gradient from 3 to 23% acetonitrile was used for the elution. The four most intense ions detected in each MS1 measurement, within a range of 350 to 1600 m/z, were selected for MS/MS fragmentation.

MS spectra were searched against the database of S. cerevisiae strain S288C proteins obtained from the Saccharomyces Genome Database (http://www.yeastgenome.org/download-data/sequence; orf_trans_all.fasta.gz) updated on February 2011, which we extended to include a reversed sequence for each of the protein sequences included in the database as decoys. Spectra searches were performed by Sorcerer Sequest version 4.2.0 search algorithm (133) run on Sorcerer2 computer (Sage-N Research, Thermo Electron). In silico trypsin digestion was performed after lysine and arginine (unless followed by proline) tolerating up to three missed cleavages in fully tryptic peptides. Database search parameters were set for carboxyamidomethylation [57.021465 atomic mass units (amu)] of cysteine residues as a static modification; serine, threonine, and tyrosine phosphorylation (79.966331 amu) and methionine oxidation (15.99492 amu) were set as differential modifications. The mass tolerance was set to 50 parts per million (ppm) for the monoisotopic masses of precursor ions. Search results were evaluated with the Trans Proteomic Pipeline (134) using the Peptide Prophet version 4.5.2. The software OpenMS version 1.8 (135) was used to identify peptide elution curves from precursor ion scans, align them between all the MS runs, and quantify relative peptide abundances. Probability scores from analysis of peptides by Peptide Prophet were used to filter OpenMS results at an FDR threshold less than 1%. Peptides with identical sequence and phosphorylation state but different charge states were merged for subsequent analysis. Only phosphopeptides were considered for further analysis. G-tests (53) were implemented in MATLAB for the detection of differentially regulated phosphopeptides among all the measured samples (version 7.10.0.499, R2010a).

Bioinformatic analysis of phosphorylated proteins and comparison to existing data sets

The 145 Snf1-dependent phosphorylated proteins (table S2) were input into STRING version 9.05 (136) to identify known gene and protein associations with confidence scores greater than 0.75 including an additional 150 associated first-order interacting proteins. This network of 296 proteins was combined with a set of edges representing a binary network between Snf1 and the Snf1-dependent phosphorylated proteins identified here. The combined networks were analyzed using Cytoscape version 3.0.1 (55). Groups of highly interconnected proteins were identified using ClusterOne version 1.0 (60) with the default parameters. These groups were then manually annotated on the basis of information from the Saccharomyces Genome Database (23 November 2013 release).

To identify conserved sequences in our set of Snf1-dependent phosphopeptides (table S2), an 11–amino acid sequence containing a centrally located Ser or Thr was analyzed by MDDLogo, version 1.0, a program that clusters sequences into conserved subgroups (65). To identify consensus sequences in the two most prominent subgroups identified by MDDLogo, we used version 2.0 of Seq2Logo, a program that generates information on both enrichment and depletion of amino acids at each position (62).

mRNA isolation and RT-qPCR

To measure transcript abundance during glucose depletion and glucose-induced mRNA decay, duplicate cultures were grown in 5% glucose–containing medium to an absorbance at 600 nm of about 0.8, and aliquots were removed for RNA analysis. The remaining cells were pelleted by centrifugation and resuspended in 0.05% glucose–containing medium. After 4 hours, aliquots of each culture were removed for RNA analysis, glucose was added to a final concentration of 5%, and aliquots were collected at multiple time points for RNA analysis. Aliquots (10 ml) were added to 50-ml Falcon tubes containing 10 g of ice, pelleted, and frozen at −80°C. Total RNA was extracted using either the acid phenol method (137) or an RNeasy kit (Qiagen). Residual DNA in the RNA preparation was removed by treatment with DNase I (Ambion) following the manufacturer’s recommendations. Complementary DNA (cDNA) synthesis was performed with SuperScript III (Invitrogen) or an iScript kit (Bio-Rad) following the manufacturer’s protocol. Quantitative real-time reverse transcription PCR (RT-qPCR) for measuring mRNA levels was performed using a 1:300 dilution of the cDNA in SsoFast EvaGreen Supermix (Bio-Rad). Primer sequences used for RT-qPCR are listed in table S8.

β-Galactosidase assays

Yeast containing either an integrated ADH2-lacZ reporter (88) or the same reporter on a CEN plasmid (138) were grown as described for RNA analysis, and 2-ml aliquots were collected for β-galactosidase activity assays at various times after depleting the cells of glucose. β-Galactosidase assays were conducted on permeabilized whole cells (139) using a Shimadzu spectrophotometer to measure the absorbance. β-Galactosidase activity was determined using three to five transformants of each deletion strain.

Protein purification, in vitro kinase assay, and Western blot

Soluble His-tagged Mig1 (52) was prepared from Escherichia coli as described in the Qiagen protein purification manual. For in vitro kinase assays, Snf1as-TAP (52) was purified from yeast by binding to immunoglobulin G magnetic beads (GE Healthcare). The beads were washed three times with kinase buffer containing 10% glycerol. Kinase reactions (usually 20 μl) contained 50 ng of Mig1, 3 to 5 μl of Snf1as-TAP bound to magnetic beads, 1.0 mM 6-cyclophenyl-γS-ATP, 5 mM MgCl2, and 50 mM tris-HCl (pH 7.5). The kinase reactions were incubated for 60 min at 30°C. Simultaneous methylation of proteins phosphorylated by γS-ATP, and thus containing a thiol-phosphate modification, and termination of the reaction were achieved by addition of p-nitrobenzyl mesylate (50 μM) and EDTA (20 mM) and incubation for 60 min at 30°C. Sample buffer was added and samples were heated at 90°C for 5 min. A portion was run on a 4 to 20% polyacrylamide gel, transferred onto a nylon membrane (Millipore), processed for Western blot using a rabbit monoclonal antibody (Epitomics) to detect methylated thiophosphorylated proteins (140), and detected with a Odyssey imaging system (Licor).

To prepare extracts for the phospho-Snf1 Western blot (fig. S10A), TCA was added directly to cultures to a final concentration of 20%. After 15 min on ice, the cells were pelleted and frozen. The cell pellets were resuspended in 1:1 20% TCA and TCA buffer [20 mM tris-HCl (pH 8), 50 mM ammonium acetate, 2 mM EDTA, protease inhibitors] and lysed with glass beads. The proteins were precipitated by centrifugation at 16,060g for 10 min, resuspended in TCA-Laemmli buffer (3.5% SDS, 14% glycerol, 120 mM tris-base, 0.005% bromophenol blue, 8 mM EDTA, 5% β-mercaptoethanol, protease inhibitors), and incubated at 99°C for 10 min, and insoluble material was removed by centrifugation at 16,060g for 10 min in a microcentrifuge. Protein samples were analyzed on a 4 to 20% polyacrylamide gel. After transfer and staining with Ponceau S, phospho-T210-Snf1 was detected with an antibody against phospho-Thr172-AMPK (Cell Signaling 2531), and total Snf1 was detected with an antibody against poly-His (H1029, Sigma).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/7/333/ra64/DC1

Fig. S1. Experimental design of Snf1 phosphoproteomics study.

Fig. S2. Comparison of phosphopeptides increased by glucose depletion and decreased by inhibition of Snf1as.

Fig. S3. Summary of phosphoproteomics data for known Snf1-substrate proteins.

Fig. S4. Comparison of the Snf1-dependent phosphoproteome and transcriptome.

Fig. S5. Groups of consensus sequences identified by MDDLogo.

Fig. S6. Glucose-induced ADH2 mRNA decay in yeast with deletions of genes involved in mRNA metabolism.

Fig. S7. Snf1-dependent transcript abundance and decay in the absence of Xrn1 activity.

Fig. S8. Snf1 activation in vivo in xrn1Δ yeast and Snf1 kinase activity in xrn1Δ extracts.

Fig. S9. Comparison of Snf1-dependent phosphorylated proteins to published Snf1-dependent phosphoproteomics data sets.

Fig. S10. Comparison of Snf1-dependent phosphorylated proteins to published Snf1-dependent phosphoproteomics data sets.

Fig. S11. Comparison of Snf1-dependent phosphorylated proteins to published Snf1-dependent phosphoproteomics data sets.

Fig. S12. Comparison of Snf1-dependent phosphorylated proteins to published Snf1-dependent phosphoproteomics data sets.

Fig. S13. Comparison of Snf1-dependent phosphorylated proteins to published Snf1-dependent phosphoproteomics data sets.

Fig. S14. Metabolic, cell signaling, and other pathways containing multiple Snf1-dependent phosphoproteins.

Table S1. Phosphopeptide abundance in snf1 mutant yeast and Snf1as yeast exposed to 2NM-PP1 for 30 min.

Table S2. Summary of the characteristics of Snf1-dependent phosphorylated proteins.

Table S3. Time course of phosphopeptide abundance in Snf1as yeast during glucose depletion and exposure to 2NM-PP1.

Table S4. GO categories of Snf1-dependent phosphorylated proteins.

Table S5. Abundance of Snf1-target genes in yeast with deletions of genes involved in mRNA metabolism.

Table S6. Yeast strains used in this study.

Table S7. Plasmids used in this study.

Table S8. qPCR primers used in this study.

File S1 (.PDF format) Graphs of the abundances of Snf1-dependent phosphopeptides.

REFERENCES AND NOTES

Acknowledgments: We thank M. Schmidt for plasmids and advice; C. Zhang for the gift of 2NM-PP1; R. Rossi for constructing pRR01; A. Johnson for the XRN1, XRN1-E176G, and RAT1ΔNLS plasmids; B.-J. Webb Robertson for help in applying the G-test algorithm; and V. L. Price for editorial assistance. Funding: This research was supported by NIH grant GM26079 to E.T.Y. and by SystemsX.ch, the Swiss initiative for systems biology, to R.A. R.A. was also supported by the FP7 project UnicellSys by the European Union. Author contributions: K.A.B., K.M.D., and E.T.Y. conceived the experiments, analyzed the data, and wrote the manuscript. S.V. and R.A. purified the phosphopeptides and performed the MS and preliminary data analysis. S.P., K.A.B., F.F., and E.T.Y. performed quantitative PCR, reporter activity assays, growth assays, in vitro kinase assay, and Western blot. K.M.D. generated the Venn diagrams and performed statistical analyses of the data. K.A.B. and S.P. generated the yeast strains. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Yeast mutant strains and plasmid constructs are available from E.T.Y. upon request. Raw MS files can be obtained from R.A. by request.
View Abstract

Navigate This Article