The importance of post-translational modifications in regulating Saccharomyces cerevisiae metabolism


Correspondence: Ana Paula Oliveira, Institute of Molecular Systems Biology, ETH Zurich, Wolfgang-Pauli-Strasse 16, CH-8093 Zurich, Switzerland. Tel.: +41 44 633 3689; fax: +41 44 633 1051; e-mail:


Regulation of the flow of mass and energy through cellular metabolic networks is fundamental to the operation of all living organisms. Such metabolic fluxes are determined by the concentration of limiting substrates and by the amount and kinetic properties of the enzymes. Regulation of the amount of enzyme can be exerted, on a long-term scale, at the level of gene and protein expression. Enzyme regulation by post-translational modifications (PTMs) and noncovalent binding of allosteric effectors are shorter-term mechanisms that modulate enzyme activity. PTMs, in particular protein phosphorylation, are increasingly being recognized as key regulators in many cellular processes, including metabolism. For example, about half of the enzymes in the Saccharomyces cerevisiae metabolic network have been detected as phosphoproteins, although functional relevance has been demonstrated only in a few cases. Direct regulation of enzymes by PTMs provides one of the fastest ways for cells to adjust to environmental cues and internal stimulus. This review charts the so far identified metabolic enzymes undergoing reversible PTMs in the model eukaryote S. cerevisiae and reviews their underlying mechanistic principles – both at the individual enzyme level and in the context of the entire metabolic network operation.


A key challenge in biology, and a problem of particular interest for metabolic engineering and synthetic biology, is to understand how micro-organisms adjust their cellular machinery in response to genetic and environmental perturbations, and how that impacts metabolic operation. A quantitative measure of metabolic operation is the flux of metabolites through the network of metabolic reactions, which reflects the integrated response from all levels of cellular regulation (Nielsen, 2003; Sauer, 2006). The flux through each metabolic reaction is determined by the concentration of the reactants and by the amount and kinetic properties of the enzyme(s) that catalyze the reaction. The catalytic capacity of an enzyme can be modulated by the noncovalent binding of allosteric effectors and/or by post-translational modifications (PTMs) (Entian & Barnett, 1992; Gerosa & Sauer, 2011). Despite the knowledge that PTMs modulate the activity of some metabolic enzymes (Krebs & Beavo, 1979; Guan & Xiong, 2011), the contribution of PTMs to the regulation of fluxes has rarely been considered because of difficulties to experimentally quantify protein modifications and their effect on function. The advent of mass spectrometry (MS)-based proteomics enables now the identification and quantification of many PTMs at a nearly genome-wide level (Jensen, 2006; Witze et al., 2007). Such data are expected to contribute to a quantitative understanding of the role of PTMs on metabolic operation.

PTMs are regulatory processes involving an alteration of the original chemical composition of a protein, usually through the covalent addition of a small molecule to one of the amino acid residues. Each PTM requires a specialized protein that catalyzes the particular modification. When another protein able to reverse the chemical modification exists, the PTM is termed reversible. In eukaryotes, frequent PTMs include phosphorylation, acetylation, methylation, glycosylation, lipidation, flavinylation, ubiquitylation, and proteolysis. These general terms can comprise a variety of more specific chemical modifications, even within the same organism (see Walsh et al., 2005; Walsh, 2006; Heuts et al., 2009 for extensive reviews).

Protein phosphorylation is the best studied and arguably the most frequent PTM in Saccharomyces cerevisiae. According to phosphoproteomics data deposited in PhosphoPep (Bodenmiller et al., 2008), more than 2300 yeast proteins occur as phosphoproteins, covering nearly all cellular processes. Phosphorylation in yeast is orchestrated by a complicated network of about 160 protein kinases and phosphatases (Bodenmiller et al., 2010), and each kinase phosphorylates on average more than 70 protein targets (Bodenmiller et al., 2010; Sharifpoor et al., 2011). The functional relevance of protein phosphorylation on yeast metabolism has been evident from studies quantifying the effect of kinase and phosphatase deletions on gene expression (Usaite et al., 2009; van Wageningen et al., 2010; Livas et al., 2011; Zhang et al., 2011), protein levels (Usaite et al., 2009), phosphoprotein levels (Huber et al., 2009; Bodenmiller et al., 2010; Zhang et al., 2011), metabolite concentrations (Usaite et al., 2009; Zhang et al., 2011), and epistasis (Fiedler et al., 2009; van Wageningen et al., 2010). Protein phosphorylation can regulate metabolic operation in different manners. One way is by regulation of the total amount of available enzyme, which is determined by the balance between the rates of protein synthesis and degradation. At the forefront of regulation of protein synthesis is transcriptional regulation, which determines at the genetic level whether the protein is required or not. Transcription factor (TF) activity is known to be largely regulated by phosphorylation (Jackson, 1992; Everett et al., 2011), and 106 of the 164 yeast TFs listed in Yeastract (Teixeira et al., 2006) are described as phosphorylated in the PhosphoPep repository (Bodenmiller et al., 2008). The other determinant of the total amount of enzyme is the rate of protein degradation. Proteins are often flagged for degradation by the covalent addition of the small protein ubiquitin, a process sometimes controlled in combination with phosphorylation (Hemmings, 1980; Deshaies, 1997; Horak, 2003). Another, more targeted way of regulation by phosphorylation is through the direct modification of the enzyme structure and consequent alteration of its functioning. Several PTMs, most noticeably phosphorylation, are known to directly affect metabolic enzyme activity and this aspect will be the main focus of this review.

Of the 924 enzymes that constitute the S. cerevisiae metabolism (Herrgard et al., 2008), the functional role of post-translational phosphorylation has been described for 23 of them. Recent phosphoproteomics studies have revealed, however, that many more metabolic enzymes are targets of the kinase/phosphatase regulatory network (Ptacek et al., 2005; Bodenmiller et al., 2010; Breitkreutz et al., 2010). In fact, about 52% of the enzymes in yeast metabolism have at least one phosphorylation site listed in the PhosphoPep or UniProt databases (Farriol-Mathis et al., 2004; Bodenmiller et al., 2008; UniProt-Consortium, 2011). How many of them are functionally relevant is an open question, and Landry et al. (2009) suggested that as much as 65% of the phosphorylation sites detected in yeast are nonfunctional, caused by off-target activity of protein kinases particularly in disordered regions of the proteins. The energetic cost for such nonfunctional phosphorylation was estimated to be actually negligible (Lienhard, 2008). Nevertheless, it seems likely that many phosphorylation events impacting enzyme functioning are yet to be unveiled.

Here we review the current knowledge on the mechanisms and function of enzyme regulation by PTMs in S. cerevisiae, with a particular focus on enzyme phosphorylation. By gathering information from traditional biochemistry literature and recent high-throughput proteomics studies, we chart on the yeast metabolic network all enzymes reported to undergo reversible PTMs. Furthermore, we discuss the implications of post-translational regulation on the functioning of individual enzymes in the context of the whole network operation. Future challenges toward the quantitative understanding of the impact of PTMs on the metabolic operation include the identification of actual functional PTM sites, evaluation of their impact on enzyme functioning and discovery of the responsible regulatory proteins catalyzing the modifications.

Detection and quantification of PTMs: from radioactive labeling to MS

High-throughput proteomics methods based on MS have proven to be particularly suited for quantification of covalently modified proteins at a nearly genome-wide scale, for organisms with a sequenced genome (Mann & Jensen, 2003; Witze et al., 2007). Compared to MS-based proteomics, classical approaches are low-throughput, laborious, and semi-quantitative. We summarize here different methods used for detection and quantification of PTMs. While these methods are generally applied to most PTMs, they are best known and developed for protein phosphorylation and they have all been applied to S. cerevisiae.

Classical approaches for detection of covalently modified proteins and identification of the modified sites typically include one or more of these techniques: (1) in vivo or in vitro incorporation of radioactive labeled elements to the covalently modified protein, (2) use of specific antibodies that recognize the modified residues in combination with Western blotting, (3) enzymatic or chemical removal of the covalent group, (4) separation of the modified protein by chromatographic methods or electrophoresis, and (5) degradation of the protein in combination with methods for detection of the modified amino acid or peptide, for example, by chromatographic methods, MS, or Edman amino acid sequencing (Yan et al., 1998; Walsh, 2006). Quantification of the modified amino acid by such classical approaches is hampered by the lack of real quantitative techniques (Yan et al., 1998). Currently, the most widespread technique for semi-quantification of PTMs in a high- or low-throughput manner is the usage of antibodies that recognize specific modified sites or epitopes, carried out in combination with Western blotting (Ptacek et al., 2005; Brumbaugh et al., 2011).

Proteomics based on 2D electrophoresis (2-DE) coupled with peptide mass fingerprinting offers higher throughput to identify and quantify PTMs in complex protein mixtures (Steinberg et al., 2003; Sa-Correia & Teixeira, 2010). Following protein separation based on charge and mass in 2D gels, relative quantification can be achieved through the intensities of the protein spots previously stained with detectable dyes. Spots can be excised and processed for protein identification by MS or MS/MS. Particularly relevant for detection of PTMs is the availability of fluorescent dyes that stain specific protein modifications, thereby aiding in the selection of modified protein spots for further identification (Sa-Correia & Teixeira, 2010). For example, the fluorescent dye ProQ Diamond allows the specific detection of phosphoproteins (Steinberg et al., 2003). In combination with other specific dyes, 2-DE can therefore facilitate the identification of potentially coexisting protein isoforms, that is, distinct covalently modified forms of the originally translated gene product. A shortcoming of 2-DE proteomics that interferes with quantification is the difficulty to confidently associate one protein with one gel spot across experiments. Another limitation lies within the range of proteins that can be detected, which excludes proteins that are extremely acidic or basic, have very low or high molecular weights, or occur in low abundance.

Methods based on liquid chromatography coupled with mass spectrometry (LC-MS) are also commonly used for high-throughput detection and quantification of PTMs (Jensen, 2006; Witze et al., 2007). Relative to 2-DE methods, they offer higher sensitivity, which improves quantification and allows detection of lower abundance proteins. They are also suited for detection of proteins independently of their pI or molecular weight. The pipeline of LC-MS-based proteomics typically includes the extraction and precipitation of the total protein from the cells, trypsinization, optional fractionation or enrichment of the digested peptides, separation of peptides by liquid chromatography and MS detection. Quantification of peptides can be achieved by label-free shotgun or differential isotope labeling using LC-MS/MS (Gstaiger & Aebersold, 2009). Particularly relevant for detection and quantification of PTMs is the availability of enrichment methods that specifically select for one type of modified peptides (Jensen, 2006; Witze et al., 2007). In the case of phosphorylation, enrichment for phosphopeptides is usually performed with metal affinity chromatography (IMAC) or titanium dioxide (TiO2) columns (reviewed in Bodenmiller & Aebersold, 2010). Detection of the same peptide sequence with different modification patterns indicates the existence of multiple protein isoforms. For example, the simultaneous detection of one nonphosphorylated, one single, and one doubly phosphorylated peptide of an otherwise identical peptide sequence indicates that the protein co-occurs in at least three isoforms. Compared to other methods, LC-MS/MS proteomics offers the sensitivity and reproducibility required for reliable quantification of total and modified proteins. In particular, several phosphoproteomics studies in S. cerevisiae based on LC-MS/MS have quantified phosphorylation changes across environmental conditions (Soufi et al., 2009), cellular stages (Keck et al., 2011), upon drug treatment (Huber et al., 2009) or between genetic variants (Bodenmiller et al., 2010), accumulating evidence for the metabolic relevance of phosphorylation.

Reversible PTMs affecting enzyme functioning

Reversible PTMs are particularly relevant signaling events because they offer the cell the possibility to dynamically modulate protein activity in fast response to environmental cues or internal stimuli. On the contrary, irreversible PTMs are long-term regulatory events, mostly associated with protein synthesis or degradation. We focus here on reversible PTMs that have been described to be relevant for post-translational modulation of enzyme functioning in S. cerevisiae. This includes, in particular, phosphorylation, ε-NH2 acetylation, S-methylation, and S-thiolation. Two other common reversible PTMs in yeast, O-mannosylation (Lehle et al., 2006) and S-palmytoylation (Bijlmakers & Marsh, 2003; Linder & Deschenes, 2003), are mainly involved in regulating and stabilizing protein localization in membranes and are not discussed here.

Reversible protein phosphorylation is probably the best studied PTM in terms of effect on protein structure, functional mechanisms of regulation, frequency of occurrence and evolutionary conservation. Its relevance in modulating yeast metabolism is also apparent from the large number of metabolic enzymes undergoing phosphorylation (Fig. 1, red circles). Given its significance, the impact of protein phosphorylation on enzyme functioning will be covered in detail in the next section.

Figure 1.

Post-translationally modified enzymes within central carbon metabolism. In color are Saccharomyces cerevisiae enzymes detected or verified to undergo phosphorylation (red), acetylation (blue), methylation (gray), and thiolation (green). Enzymes with functionally verified PTMs are marked with filled circles. Open circles represent detected but not yet functionally validated PTMs. Detected phosphorylation is based on data deposited in PhosphoPep (; as of February 2011), UniProt (; as of August 2011) and literature mentioned in the text. Detected acetylation and methylation are based on data deposited in UniProt and literature. Detected thiolation was taken from Le Moan et al. (2006). Metabolite abbreviations: 13BPG, 3-phosphoglyceroyl phosphate; 2PG, 2-phosphoglycerate; 3PG, 3-phosphoglycerate; 6PGC, 6-phosphogluconate; 6PGL, glucono-1,5-lactone 6-phosphate; ACAL, acetaldehyde; Accoa, acetyl-CoA; AKG, 2-oxoglutarate; CIT, citrate; DHA, dihydroxyacetone; DHAP, dihydroxyacetone phosphate; E4P, erythrose 4-phosphate; F26P, fructose 2,6-bisphosphate; F6P, fructose 6-phosphate; FDP, fructose 1,6-bisphosphate; FUM, fumarate; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate; GAP, glyceraldehyde 3-phosphate; GL3P, glycerol 3-phosphate; GLX, glyoxylate; ICIT, isocitrate; MAL, malate; OA, oxaloacetate; PEP, phosphoenolpyruvate; R5P, d-ribose 5-phosphate; RL5P, d-ribulose 5-phosphate; S7P, sedoheptulose 7-phosphate; SUCC, succinate; SUCCCoA, succinyl-CoA; TRE6P, trehalose 6-phosphate; UDPG, uridine diphosphate-glucose; X5P, xylulose-5-phosphate.

Protein acetylation involves the covalent addition of an acetyl group from acetyl coenzyme A to the target protein. In yeast, protein acetylation can occur in two distinct biological processes. One is the acetylation of the N-terminus of proteins, a frequent irreversible co-translational modification of unclear biological significance (Walsh, 2006) that is not covered here. The second process is the ε-NH2 acetylation of lysine residues, a PTM catalyzed by protein lysine acetyltransferases and reversed by deacetylases. Protein ε-NH2 acetylation was, until recently, mainly known for its relevance in regulating nuclear transcription processes, in particular through the modification of histones and TFs. Recent acetylproteome studies in Salmonella enterica (Wang et al., 2010) and human liver tissue (Zhao et al., 2010) have revealed extensive occurrence of acetylation, and more than 2000 proteins of mammalian cells have so far been found to undergo acetylation (Guan & Xiong, 2011). Surprisingly, a large number of metabolic enzymes were found acetylated in both S. enterica and liver tissue, including many involved in glycolysis, gluconeogenesis, the tricarboxylic acid (TCA) cycle, glycogen metabolism, oxidative phosphorylation, fatty acid oxidation, the urea cycle and amino acid metabolism (Wang et al., 2010; Zhao et al., 2010). Functional evaluation of acetylation sites for some of these enzymes, both in S. enterica and mammalian cells, demonstrated a clear role of acetylation in modulating enzyme activity, protein stability, and metabolic reaction directionality (reviewed in Guan & Xiong (2011)). So far, the acetylproteome of S. cerevisiae has not been analyzed, but several lines of evidence suggest that the extent and relevance of protein acetylation in regulating yeast metabolism may be akin to S. enterica and mammalian cells (Lin et al., 2008, 2009). In a study focusing on identifying the targets of the H4 acetyltransferase in S. cerevisiae, Lin et al. (2008) found 18 acetylated enzymes, mainly involved in carbohydrate and amino acid metabolism. Among these, they detected acetylation in the gluconeogenic enzyme phosphoenolpyruvate carboxykinase (Pck1) and verified that acetylation of Pck1 at Lys514 is required to assure maximal enzyme activity. In other studies, acetylation of the yeast glucose-6-phosphate dehydrogenase (Zwf1) at site Lys191 (Jeffery et al., 1985) and of the yeast 6-phosphofructo-2-kinase (Pfk26) at site Lys3 (Dihazi et al., 2005) was shown to inhibit enzyme activity. UniProt lists an additional 36 enzymes that undergo acetylation (Fig. 1, blue circles). In terms of biological significance, Guan & Xiong (2011) suggest that reversible acetylation, in particular when associated with a class of deacetylases that is NAD+-dependent (SIRTs), is a regulatory mechanism responding to the NAD+:NADH ratio, a measure of cellular redox and energy status.

Protein methylation can occur through the covalent addition of one to three methyl groups from S-adenosylmethionine at different nucleophilic side chains, a PTM catalyzed by protein methyltransferases. The most common methylation in yeast is N-methylation, which can occur at the nitrogen atom of lysine and arginine residues. N-methylation is a frequent, irreversible PTM modifying mainly histones and proteins involved in translational processes (Walsh et al., 2005; Walsh, 2006), though it has not been found to impact metabolic enzymes post-translationally (Pang et al., 2010). The role of methylation in modulating yeast metabolic activity has been, indeed, only documented for three enzymes involved in trehalose metabolism (Sengupta et al., 2011). In their study, they showed that trehalose synthase (Tps1) is subjected to S-methylation in a cysteine residue by formation of a thiol ester, which enhances Tps1 activity. The neutral trehalase (Nth1) and the acid trehalase (Ath1) were also found to be methylated. UniProt lists three additional metabolic enzymes to undergo methylation (Fig. 1, gray circles). The biological causes of enzyme regulation by methylation are unclear, and its significance in yeast seems minor.

Lastly, S-thiolation is the modification of the protein SH groups of cysteines owing to the formation of disulfide structures binding to low-molecular-mass thiols such as glutathione, which can be chemically reversed inside the cell. S-thiolation prevents the irreversible oxidation of cysteine residues and has been observed in 37 yeast metabolic enzymes (Fig. 1, green circles), including enzymes of glycolysis, the TCA cycle, and amino acid metabolism (Le Moan et al., 2006). In terms of function, it was shown that enzymatic activity of the glyceraldehyde-3-phosphate dehydrogenase Tdh3, but not of its isoenzyme Tdh2, is inactivated by S-thiolation (Grant et al., 1999). The biological significance of S-thiolation of metabolic enzymes has been associated with regulatory mechanisms responding to oxidative stress.

Figure 1 charts the discussed PTMs in the yeast central carbon metabolism. Because the occurrence of a modification may not imply functionality (Lienhard, 2008; Landry et al., 2009), we distinguish between detected PTMs (Fig. 1, all circles) and those with verified functional impact (Fig. 1, filled circles). As it becomes evident, many enzymes in glycolysis, pentose-phosphate pathway, gluconeogenesis, the TCA cycle and carbohydrate storage metabolism are subject to PTMs, indicating that central carbon metabolism is heavily post-translationally regulated. Furthermore, several enzymes show multiple and/or tandem PTMs, suggesting that S. cerevisiae may use different PTMs to specifically regulate enzyme function in response to different environmental challenges.

Enzyme regulation by post-translational phosphorylation

Protein phosphorylation is the best characterized and understood PTM in eukaryotes and is involved in regulating practically all cellular processes. The covalent addition of a phosphate group from ATP to serine, threonine, or tyrosine residues is catalyzed by protein kinases and can be reversed by protein phosphatases. The structural consequences of post-translational phosphorylation have been elucidated for several proteins (Cohen, 2000; Johnson & Lewis, 2001; Serber & Ferrell, 2007; Johnson, 2009). In S. cerevisiae, two general mechanisms of how phosphorylation affects protein structure, and consequently function, have been described: (1) phosphorylation induces a conformational change in the protein structure, a process that can be classified as allosteric regulation (Goodey & Benkovic, 2008; Johnson, 2009) given the etymological origin of the word allosteric, from the greek allo-(‘other’) steric (‘arrangement of atoms in space’), and (2) phosphorylation modifies the bulk electrostatic charge of the protein (Strickfaden et al., 2007). The role of phosphorylation is usually described by a measurable impact on function rather than on structure. General methods to assess the impact of phosphorylation on function include, for example, measuring the in vitro activity (Fosset et al., 1971), determining the localization of the protein (Smith & Rutter, 2007) or identifying the participation in protein complexes (Palomino et al., 2006; Yachie et al., 2011). In the particular case of metabolic enzymes, functional impact is usually assessed at the level of the enzymatic catalytic capacity. This includes, for example, the determination of the in vitro enzyme activity, an approximate measure of the enzymatic capacity but that may not reflect the actual in vivo reaction rate. Functional assessment further requires the differential evaluation between phosphorylated and nonphosphorylated isoforms. This can be achieved by (1) chromatographic separation of protein isoforms (Fosset et al., 1971), (2) genetic substitution of the phosphorylated residues to mimic phosphosite loss or permanent phosphorylation (Hardy & Roach, 1993; Soulard et al., 2010), or (3) genetic encoding of a phosphocaged amino acid that can expose the phosphosite upon irradiation with light (Lemke et al., 2007).

The occurrence of reversible phosphorylation was first described for the enzyme glycogen phosphorylase of rabbit skeletal muscle (Krebs & Fischer, 1955), and many of the first characterized targets of protein phosphorylation were other metabolic enzymes (Krebs & Beavo, 1979). In S. cerevisiae, reversible phosphorylation was first described also for glycogen phosphorylase (Gph1) (Fosset et al., 1971), which was detected to incorporate radioactive phosphorus in the presence of [γ-32P]ATP and of a specific yeast protein kinase. The phosphorylated protein isoform showed higher in vitro activity compared to the dephosphorylated isoform, indicating that phosphorylation activates Gph1 activity (Fosset et al., 1971). The site of phosphorylation was later identified as Thr31, following digestion of 32P radioactively labeled protein, chromatography separation of radioactively labeled peptides, and amino acid sequencing (Lerch & Fischer, 1975). The structural mechanism of how phosphorylation regulates Gph1 has also been elucidated, both in yeast (Lin et al., 1996) and in mammalian cells (Barford et al., 1991). In yeast, activation by phosphorylation involves a competition between the phosphoamino acid and the allosteric effector glucose-6-phosphate, a relief of steric blocking of the catalytic site and a conformational reorganization of the protein (Lin et al., 1996).

Besides Gph1, eight other enzymes of the yeast central carbon metabolism are known to be regulated by post-translational phosphorylation (Fig. 1, filled red circles). Using chromatographic separation of protein isoforms followed by in vitro enzymatic assays, phosphorylation was reported to activate Pfk26 (Francois et al., 1984) and to inhibit Fbp1 and Pda1 activity (Mazon et al., 1982; Uhlinger et al., 1986). Phosphorylation of Fbp1 was later suggested to be a signal for degradation (Holzer & Purwin, 1986). By mimicking phosphosite loss through the genetic substitution of phosphoserine by alanine residues, in combination with in vitro enzymatic assays or localization studies, phosphorylation was further found to activate Nth1 and Cdc19 (Pyk1) (Wera et al., 1999; Portela et al., 2006), to inhibit Gsy2 (Hardy & Roach, 1993), to regulate the role of Hxk2 in glucose repression of invertase (Randez-Gil et al., 1998) and to induce a conformational change that regulates the location of the reaction catalyzed by Ugp1 (Smith & Rutter, 2007). Of these nine central carbon metabolism enzymes, four are concentrated within carbohydrate storage pathways and two are at the pyruvate branch-point (Fig. 1). Except for Ugp1, all enzymes are involved in catalyzing irreversible reactions and four of them have isoenzymes. Outside central carbon metabolism, ten other enzymes sparsely located in the metabolic network (Herrgard et al., 2008) have functionally verified phosphosites – Cki1, Dpm1, Fur4, Lcb4, Pde1, Pma1, Slm1, Tor2, Ura7 and Ycf1 [see PhosphoGrid (Stark et al., 2010)]. At least four additional metabolic proteins – Acc1 (Mitchelhill et al., 1994), Bap2 (Omura & Kodama, 2004), Gap1 (Stanbrough & Magasanik, 1995), and Gdh2 (Uno et al., 1984) – are reportedly regulated by phosphorylation, though their functional sites have not yet been identified. Noticeably, four of the phosphoregulated proteins are membrane transporters, including the amino acid permeases Bap2 and Gap1, the uracil permease Fur4, and the membrane proton pump Pma1. Membrane transporters are particularly known to be post-translationally regulated by ubiquitylation (Horak, 2003), and in the cases above, phosphorylation is known to occur in association with ubiquitylation, playing a key role in signaling the proteins for degradation (Stanbrough & Magasanik, 1995; Horak, 2003; Omura & Kodama, 2004).

It is likely that many more than the currently known 23 metabolic enzymes are functionally regulated by phosphorylation, given that at least 486 enzymes of the yeast metabolic network can occur phosphorylated according to several phosphoproteomics databases (Farriol-Mathis et al., 2004; Bodenmiller et al., 2008; UniProt-Consortium, 2011). Further substantiating the relevance of phosphorylation in metabolism is the observation that duplicated genes arising from the duplication of the S. cerevisiae genome are more phosphorylated than the others (Amoutzias et al., 2010). The authors of this work suggested that the acquisition of different enzyme functionalities by phosphorylation contributed to the retention of the duplicated genes.

Protein phosphorylation is catalyzed by protein kinases, which can recognize specific peptide sequences (Miller et al., 2008; Mok et al., 2010) and consequently phosphorylate their target proteins in few milliseconds (Papin et al., 2005). Responses to stimuli are often transducted through a cascade of several kinases and therefore it may take several seconds or minutes to observe protein phosphorylation following an internal or external stimulus (Gross et al., 1992; Salazar et al., 2010). There are about 120 protein kinases in S. cerevisiae, of which only 40 have clearly identified target substrates reported in the literature (see positive-training set of Sharifpoor et al., 2011). Of the 23 phosphoregulated metabolic enzymes listed above, the yeast kinases responsible for phosphorylation are known for only seven of them: Gph1 and Gsy2 can be phosphorylated by the cyclin-dependent kinase Pho85 (Wilson et al., 2005); Cdc19 and Pfk26 can be phosphorylated by the PKA subunit Tpk1 (Rayner et al., 2002; Dihazi et al., 2003); Pda1 can be phosphorylated by either of the mitochondrial protein kinases Ptk1 or Ptk2 (Gey et al., 2008); Ugp1 can be phosphorylated by either of the PAS kinases Psk1 and Psk2 (Smith & Rutter, 2007); and Acc1 can be phosphorylated by the AMP-activated protein kinase Snf1 (Mitchelhill et al., 1994). While protein phosphorylation can either activate or inhibit a particular enzyme activity, protein phosphatases are able to remove the phosphate group, thereby reversing the phosphorylation effect. Protein kinases are typically rather specific, while the about 40 protein phosphatases in yeast exhibit somewhat lower specificity for their targets.

Several resources recently available allow hypothesizing the in vivo targets of protein kinases and phosphatases in S. cerevisiae. The Yeast Kinase Interaction Database (KID) is a repository of high- and low-throughput phosphorylation data and also gives a confidence score on the likelihood of the kinase-target interaction (Sharifpoor et al., 2011). Although useful, KID does not detail which phosphosites are functionally regulated and by which specific kinase. This information, when available, can be found in PhosphoGrid (Stark et al., 2010), a repository of phosphorylation sites with functional annotation based on literature information and peptide motifs. The specificity of kinase recognition motifs can indeed help identifying new protein targets, and recently Mok et al. (2010) determined consensus phosphorylation motifs for 61 yeast kinases. Another study particularly relevant for hypothesizing in vivo targets of kinases/phosphatases is the analysis of phosphoproteome responses to nearly every single kinase/phosphatase deletion in yeast (Bodenmiller et al., 2010). This study, however, does not distinguish between direct and indirect targets of the deleted kinases/phosphatases. Generally, it remains a challenge to confidently identify the biological relevant kinases and phosphatases targeting particular protein sites, as well as to assess the impact of phosphorylation sites in activating or inhibiting protein function.

Protein phosphorylation as a mechanism that modulates the catalytically competent enzyme

The knowledge that enzyme phosphorylation generally activates or inhibits catalytic function is a qualitative description but does not bring quantitative insights into how phosphorylation actually modulates the functioning of the enzyme pool. Recent studies have combined the quantification of phosphorylation abundance with total protein abundance to determine the degree of phosphorylation, that is, the percentage of a particular phosphorylated isoform that exists within the total protein pool, at a particular time/cellular state (Wu et al., 2009; Olsen et al., 2010, Wu et al., 2011ab). Importantly, these studies made clear that phosphorylation is not binary, that is, pools of phosphorylated and nonphosphorylated protein coexist, and the fraction of phosphorylated protein can largely vary across conditions or cellular stages.

Knowing that the degree of phosphorylation is not binary, but rather fractional, helps to rationalize how enzyme functioning is modulated by post-translational phosphorylation. For example, considering the simplest scenario of two coexisting isoform pools, one of nonphosphorylated protein and one of phosphorylated protein, two relationships between isoforms and activity are possible (Fig. 2). In cases where phosphorylation increases enzyme activity, the phosphorylated isoform should represent the pool of catalytically competent enzyme, while the pool of nonphosphorylated isoform should show low or no activity (Fig. 2a). On the contrary, in cases where phosphorylation inhibits activity, the nonphosphorylated pool should represent the catalytically competent enzyme, while the phosphorylated pool should show low or no activity (Fig. 2b). The described model is a relatively straightforward formalization of the observation that phosphorylation activates or inhibits function. It helps, however, establishing a quantitative relationship between phosphorylation and enzyme activity. In a schematic example, and considering that (1) the enzyme is phosphorylated at a single site, (2) phosphorylation affects catalytic activity, and (3) both phosphorylated and nonphosphorylated isoforms coexist across conditions and their total abundances are known, then the catalytic activity should be proportional to the total amount of catalytically competent enzyme, whether this is the phosphorylated or the nonphosphorylated pool (Fig. 2). This model is valid if the catalytic activity is determined by the amount of active enzyme rather than by the limiting substrate concentration and if no other post-translational regulation occurs, such as other PTMs, noncovalent binding of allosteric effectors or differential protein complex assembly.

Figure 2.

Phosphorylation determines the pool of catalytically competent enzyme. The figure depicts a schematic case in which the exact amounts of phosphorylated and nonphosphorylated protein P are known across five growth conditions. Phosphorylated protein is depicted with a red phosphate group. Protein P is an enzyme that catalyzes a metabolic reaction not limited by the concentration of reactants. The catalytic activity (Act.) of the enzyme, given in arbitrary units, is also known across the five conditions. (a) When the phosphorylated protein is the catalytically competent pool, the enzymatic activity should be proportional to the amount of phosphorylated protein. In such cases, phosphorylation globally activates the entire pool of protein P. (b) When the nonphosphorylated protein is the catalytically competent pool, the enzymatic activity should be proportional to the amount of nonphosphorylated protein. In such cases, phosphorylation globally inhibits the entire pool of protein P.

The view that phosphorylation modulates the pool of catalytically competent enzyme able to realize flux suggests that it is important to include post-translational regulatory events that determine the active protein when considering the flow of information from genotype to phenotype. That is, genetic information flows from gene to mRNA to enzyme to active enzyme, which then is a determinant of the metabolic flux. Identification and quantification of the impact of PTMs on enzyme functioning are therefore fundamental steps toward a more complete and quantitative understanding of the metabolic network operation.

Regulation of metabolism by phosphorylation: post-translational vs. transcriptional regulation of enzymes

In addition to modulating enzyme activity by modifying its protein structure, regulation of metabolism by protein phosphorylation can also be exerted by altering the amount of enzyme via modulation of TF activity (Jackson, 1992; Everett et al., 2011). In S. cerevisiae, phosphorylation is known to affect the nuclear activity of several TFs by regulating their cytosolic–nuclear translocation. Examples of TFs regulated by phosphorylation include Mig1, Msn2, and Pho4 (De Vit et al., 1997; Komeili & O'Shea, 1999; Gorner et al., 2002). In yeast, there are about 120 TFs that regulate expression of genes coding for metabolic enzymes (Fendt et al., 2010b). Of these, 69 TFs, targeting about 660 metabolic genes, are found phosphorylated in PhosphoPep (Bodenmiller et al., 2008), suggesting that their TF activity is modulated by phosphorylation. Several studies evaluating the effect of kinase deletion/inhibition on mRNA expression have indeed found many gene targets in metabolism (Slattery et al., 2008; Usaite et al., 2009; van Wageningen et al., 2010; Livas et al., 2011; Zhang et al., 2011). For example, Usaite et al. (2009) observed that deletion of Snf1 has extensive effects on the mRNA levels of genes associated with carbon metabolism, stress, redox homeostasis, β-oxidation of fatty acids and lipid biosynthesis. Using Reporter Effectors as an indicator of differential TF activity (Oliveira et al., 2008), they further identified 49 effectors likely to regulate most changes in gene expression, 39 of which we find phosphorylated in PhosphoPep. In another study, van Wageningen et al. (2010) profiled the transcriptome of all viable kinase and phosphatase deletions in yeast and observed that about 70% of the single deletions did not cause any defects in gene expression under the tested condition. However, following up on double mutants with negative epistatic interactions, whose single mutants had no defects in gene expression relative to the wild type, they observed many transcriptional changes in metabolic genes. Their study further yielded interesting insights into genetic buffering relationships among pairs of kinase/kinase, kinase/phosphatase and phosphatase/phosphatase, including the discovery of functional redundancy cases. We note that both studies measured gene expression in yeast growing under pseudo-steady state, implying that the observed changes in gene expression may not be all directly attributed to the kinase removal but reflect instead the cellular adaptation to the new genotype.

When comparing how many of the 924 metabolic enzymes (Herrgard et al., 2008) are affected by phosphorylation, we find that 486 enzymes can be directly targeted at post-translational level and that 661 can be regulated at gene expression level by phosphorylatable TFs (Fig. 3a and b). Of these, 367 enzymes can be regulated at both levels (Fig. 3b). More generally, we find that 856 metabolic enzymes can be transcriptionally regulated by at least one TF listed in Yeastract (Teixeira et al., 2006) (Fig. 3a and c), 460 of which can also be regulated at post-translational level (Fig. 3c). A remaining of 68 enzymes in the yeast metabolic network are not associated with any transcriptional regulator, suggesting that they are constitutively expressed. Of these, 26 can be post-translationally phosphorylated, including the pyruvate dehydrogenase component Pda1 (Fig. 3c). Globally, our analysis indicates that about half of the metabolic enzymes can be regulated by at least two different processes, that is, at gene expression level and by post-translational phosphorylation. These two regulatory processes allow the cell to react to internal or external stimulus at different time scales. The large number of enzymes regulated at both levels not only highlights the crucial importance of metabolism for cell functioning but also makes clear how important is to identify and understand the several layers of regulation linking genotype to metabolic operation.

Figure 3.

Regulation of metabolism by phosphorylation and at gene expression level. (a) In response to a stimulus, the phosphorylation signal (in red) is transducted through a signaling cascade of kinases (in purple). The final receivers of the phosphorylation signal include, among others, metabolic enzymes and TFs. Phosphorylation can regulate metabolism either by direct PTM of the metabolic enzyme (in light blue) or by modulation of the activity of TFs regulating the gene expression of metabolic enzymes (in yellow). More generally, transcriptional regulation can be exerted by any TF, regardless of it being or not phosphorylated. For the yeast metabolic network (Herrgard et al., 2008), we find 486 enzymes to be post-translationally phosphorylated (in dark blue), 661 enzymes to be transcriptionally regulated by phosphorylatable TFs (in brown), and 856 enzymes to be transcriptionally regulated by any TFs (in green). Phosphorylation data are based on PhosphoPep (; as of February 2011) and UniProt (; as of August 2011), while the transcriptional regulatory network is based on Yeastract (; as of February 2011). (b) Venn diagram representing how many enzymes can be regulated by post-translational phosphorylation (PTP) and/or at gene expression level by phosphorylatable TFs. Listed are the enzymes of central carbon metabolism mapped in Fig. 1 and that are exclusively post-translationally phosphorylated or transcriptionally regulated. Thus, enzymes regulated at both levels (overlapping area) correspond to the Fig. 1 enzymes not listed in the diagram. (c) Venn diagram representing how many enzymes can be regulated by post-translational phosphorylation and/or at gene expression level by at least one TF listed in Yeastract; otherwise similar to (b) description.

Other mechanisms of post-translational regulation of metabolism

For a more complete overview of post-translational regulation directly targeting enzymes, we briefly mention here other post-translational regulatory mechanisms beyond PTMs, that is, noncovalent binding of allosteric metabolites and modulation of enzyme activity by mRNA binding. Noncovalent binding of allosteric effectors is a well-known post-translational regulatory mechanism impacting metabolism. Allosteric effectors alter the activity of an enzyme by binding to a protein region other than the catalytic center and consequently inducing a conformational change (Peracchi & Mozzarelli, 2011). Several metabolic enzymes in S. cerevisiae are known to be allosterically regulated by noncovalent binding of metabolites. In glycolysis, 6-phosphofructo-1-kinase (Pfk1/Pfk2) and pyruvate kinase (Cdc19, also known as Pyk1) activities are allosterically modulated by metabolites (Reibstein et al., 1986; Jurica et al., 1998). Both are regulated by fructose-2,6-bisphosphate, while Pfk has several other effectors. The contribution of these allosteric effectors to the metabolic operation has been incorporated into a kinetic model of yeast central carbon metabolism (Teusink et al., 2000). Within the amino acid biosynthetic pathways, several yeast enzymes are also reportedly regulated by allosteric binding of amino acids (Jones & Fink, 1982). This information was incorporated into a model of flux prediction that was used to show that flux predictions from protein/mRNA levels can be improved when accounting for the density of allosteric regulation (Moxley et al., 2009). In general, identification of allosteric metabolites and quantification of their effect on the enzyme activity represent a challenge. Recent studies have introduced technologies to systematically identify metabolite–protein interactions and further showed that several lipids bind to metabolic enzymes (Gallego et al., 2010; Li et al., 2010). The functional implications of this binding remains unclear, but at least a fraction of these interactions are expected to have a regulatory role (Li & Snyder, 2011). We anticipate that in the next few years many new metabolite-enzyme interactions will be mapped, though the evaluation of their impact on enzyme functioning will remain a challenging task.

Much less characterized is regulation based on RNA-binding proteins (RBPs). Many proteins, including several metabolic enzymes, are known to bind to RNA molecules and thereby increase or decrease their half-lives (Ciesla, 2006; Scherrer et al., 2010; Tsvetanova et al., 2010). Regulation by RBP is therefore generally referred as a post-transcriptional regulatory mechanism. Many RBP–RNA interactions have been mapped in yeast, and several enzymes of central carbon metabolism were found to bind RNA (Scherrer et al., 2010; Tsvetanova et al., 2010). Surprisingly, it has been observed that binding of mRNA to some enzymes modulates their catalytic capacity (Ciesla, 2006). For example, in human cells, binding of the cytosolic aconitase to mRNA encoding for iron-regulatory elements modulates the activity of the aconitase (reviewed in Ciesla (2006) and Hentze & Preiss (2010)). In the presence of iron, there is no binding and the enzyme is active, while during iron starvation, binding occurs and the aconitase is inactive. In yeast, the enzymatic activity of the NAD+-dependent isocitrate dehydrogenase (Idh1/Idh2) was found to be allosterically inhibited by mitochondrial mRNA (Anderson et al., 2000). This was suggested to represent a mechanism to control the flux through the TCA cycle (Ciesla, 2006). In both examples, it is unclear whether regulation occurs explicitly because of the mRNA binding or rather because of co-occurring PTMs or allosteric metabolites, but some authors speculate enzyme modulation by mRNA binding to be a potential mechanism for rapid sensing and adjustment of metabolism through gene expression regulation (Ciesla, 2006; Hentze & Preiss, 2010).


Metabolic enzymes are apparently very frequently post-translationally modified. Phosphorylation in particular is a highly frequent PTM, although functional evidence is available for < 25 of the several hundred reported cases. Bridging this gap requires either to demonstrate functionality of enzyme phosphorylation or to show that this is simply noise resulting from kinase promiscuity. Toward this endeavor, technological developments along three major lines are required: methods for more quantitative and comprehensive monitoring of PTMs, large-scale methods to specifically perturb particular PTM events, and methods to resolve the functional consequences of these perturbations.

While different solutions for the three challenges of monitoring, perturbing, and functional characterization are possible, we comment here on some with already demonstrated potential. On the monitoring end, MS-based proteomics methods must be developed that depart from the current global and less sensitive profiling of peptides to more targeted and quantitative methods that can comprehensively and reproducibly detect prespecified PTMs. Such targeted methods exist already for total protein abundance monitoring (Picotti et al., 2009) as used, for example, for the comprehensive monitoring of all yeast proteins in central carbon and amino acid metabolism (Costenoble et al., 2011). Regarding perturbation of PTMs, methodologies will undoubtedly continue to be based on evaluating the effect of different environmental conditions or of responses to systematic kinase/phosphatase deletions (Bodenmiller et al., 2010). In particular, the information content of perturbation studies will increase as more rapid monitoring methods become available. Additionally, large-scale methods for systems genetics are necessary to probe causality by removing particular PTM sites, for example, large numbers of phosphosites in yeast enzymes. Lagging behind so far are scalable methods to detect the functional consequences of PTM events. Here we believe that in particular high(er)-throughput 13C-flux methods and metabolomics will make the difference (Zamboni & Sauer, 2009). Monitoring the integrated flux response in TF mutants has already demonstrated to identify the condition dependent relevance of transcriptional regulation (Fendt et al., 2010b; Haverkorn van Rijsewijk et al., 2011). To complement flux methods that detect primarily large metabolic effects, metabolomics methods can reveal more subtle and local changes in enzyme activity (Fendt et al., 2010a). Because metabolomics methods are particularly amenable to high-throughput experimentation (Fuhrer et al., 2011), we anticipate a major role for metabolomics in functional validation of PTMs.

Once identified and functionally understood in the metabolic context, the network of PTM sites of enzymes and their involved regulatory proteins will contribute to our understanding of metabolic regulation and operation. In particular, it will offer novel opportunities for fine-tuning of metabolism by modification of individual PTM sites, thus reducing regulatory side effects that are inherent to deletion or over-expression of entire genes. Because yeast is an important eukaryotic model, at least some of the identified principles will also be useful to lead discoveries for human health and disease.


We thank Christina Ludwig and Juliane Schulz for critical comments on the manuscript. This work was financially supported through the YeastX grant of