The cAMP–protein kinase A (PKA) pathway in the yeast Saccharomyces cerevisiae plays a major role in the control of metabolism, stress resistance and proliferation, in particular in connection with the available nutrient conditions. Extensive information has been obtained on the core section of the pathway, i.e. Cdc25, Ras, adenylate cyclase, PKA, and on components interacting directly with this core section, such as the Ira proteins, Cap/Srv2 and the two cAMP phosphodiesterases. Recent work has now started to reveal upstream regulatory components and downstream targets of the pathway. A G-protein-coupled receptor system (Gpr1–Gpa2) acts upstream of adenylate cyclase and is required for glucose activation of cAMP synthesis in concert with a glucose phosphorylation-dependent mechanism. Although a genuine signalling role for the Ras proteins remains unclear, they appear to mediate at least part of the potent stimulation of cAMP synthesis by intracellular acidification. Recently, several new targets of the PKA pathway have been discovered. These include the Msn2 and Msn4 transcription factors mediating part of the induction of STRE-controlled genes by a variety of stress conditions, the Rim15 protein kinase involved in stationary phase induction of a similar set of genes and the Pde1 low-affinity cAMP phosphodiesterase, which specifically controls agonist-induced cAMP signalling. A major issue that remains to be resolved is the precise connection between the cAMP–PKA pathway and other nutrient-regulated components involved in the control of growth and of phenotypic characteristics correlated with growth, such as the Sch9 and Yak1 protein kinases. Cln3 appears to play a crucial role in the connection between the availability of certain nutrients and Cdc28 kinase activity, but it remains to be clarified which nutrient-controlled pathways control Cln3 levels.
For several years, it has been thought that the Ras proteins replace the classical, heterotrimeric G-proteins for control of adenylate cyclase activity in the yeast Saccharomyces cerevisiae. (Broek et al., 1985; Toda et al., 1985). This has led to the concept of the Ras–adenylate cyclase pathway, in which the Ras proteins were presumed to mediate signalling to adenylate cyclase. The precise nature of these signals, however, has remained rather unclear. Yeast adenylate cyclase activity is not only stimulated by active Ras; in the absence of active Ras, basal cAMP synthesis in vivo is extremely low and insufficient for viability. Inactivation of Ras causes arrest at the same point in the cell cycle as nutrient depletion, and the cells permanently enter into the stationary phase G0. Several characteristics typically expressed in slow-growing and in stationary phase cells are controlled by the Ras—adenylate cyclase pathway. As these properties are also controlled by nutrient availability (in particular by the presence of glucose or related fermentable sugars), the Ras–cAMP pathway has generally been considered to play a role in signalling the nutrient status of the environment to the cellular machinery (for reviews, see Matsumoto et al., 1985; Wigler et al., 1988; Broach and Deschenes, 1990; Thevelein, 1991; 1994; Tatchell, 1993).
Elucidation of the nutrient-sensing mechanisms and their precise connection with the Ras–cAMP pathway has proved more difficult. Genetic evidence indicated a role for Cdc25 and Ras in signalling glucose availability (Mbonyi et al., 1988; Munder and Küntzel, 1989; Van Aelst et al., 1990; Van Aelst et al., 1991; Bhattacharya et al., 1995). Several yeast mutants isogenic with tps1 were defective in glucose activation of cAMP synthesis as well as other glucose-induced regulatory effects, and the gene product was therefore considered to be involved in glucose sensing. However, subsequent work revealed that the mutants were located in trehalose-6-phosphate synthase and that introduction of the hxk2 suppressor mutation in the tps1Δ mutant restored cAMP signalling (Thevelein, 1992; Hohmann et al., 1993; Thevelein and Hohmann, 1995). Recent work has shed new light on the mechanisms involved in glucose activation of cAMP synthesis. A G-protein-coupled receptor (GPCR) system, consisting of Gpr1 and its Gα protein Gpa2, is required for glucose stimulation of cAMP synthesis, suggesting that Gpr1 might itself function as a receptor for glucose (Colombo et al., 1998; Kraakman et al., 1999; F. Rolland et al., submitted). These results have brought yeast adenylate cyclase back in line with the classical concepts established for regulation of mammalian adenylate cyclase by GPCR systems (Gilman, 1987; Dohlman et al., 1991). At the same time, however, it has again brought up the question as to what is the true function of Ras-dependent control of adenylate cyclase in yeast.
The majority of downstream targets of PKA identified until recently were enzymes involved in intermediary metabolism and, in particular, carbon metabolism. Recently, four other targets have been identified, which play a regulatory role downstream of PKA. The Msn2 and Msn4 transcription factors in part mediate PKA-dependent regulation of the expression of STRE-(stress response element)-controlled genes (Martinez-Pastor et al., 1996; Schmitt and Mcentee, 1996; Boy-Marcotte et al., 1998; Smith et al., 1998). The Rim 15 protein kinase has been identified as a downstream target of PKA that acts as an activator of STRE-controlled gene expression upon entry into stationary phase. PKA activity also causes strong feedback inhibition on cAMP synthesis, which appears to play a role in downregulating cAMP signalling events (Nikawa et al., 1987; Mbonyi et al., 1990). The Pde1 cAMP phosphodiesterase has recently been identified as a PKA target specifically involved in downregulation of agonist-induced cAMP signalling (Ma et al., 1999).
Components of the cAMP–PKA pathway
The first components of the cAMP pathway cloned in S. cerevisiae were the RAS1 and RAS2 genes. The similar phenotype of Ras mutants and cAMP–PKA mutants led to the discovery that adenylate cyclase is controlled by the Ras proteins in yeast cells. A major argument for a role of Ras as activator of adenylate cyclase was the phenotype of a strain containing the dominant Ras2val19 allele, the equivalent of the rasval12 mammalian oncogene (Toda et al., 1985). It displayed the same phenotype as mutants with an elevated cAMP level. Subsequent to the identification of the Ras proteins as controlling elements of adenylate cyclase, several other major components of the cAMP pathway were identified and their genes cloned (for reviews, see Wigler et al., 1988; Broach and Deschenes, 1990; Thevelein, 1992; 1994; Tatchell, 1993) (Fig. 1). Deletion of the Ras–guanine nucleotide exchange protein, Cdc25, causes the same lethal phenotype as deletion of Ras or inactivation of adenylate cyclase. A functional homologue of Cdc25, Sdc25, was discovered, but its deletion is not lethal. The activity of the Ras proteins is also controlled by the Ira1 and Ira2 proteins, which are activators of Ras-GTPase activity. Inactivation of the Ira proteins causes overactivation of the cAMP pathway. Adenylate cyclase is encoded by the CYR1/CDC35 gene and an associated protein by CAP/SRV2. cAMP is hydrolysed by a low-affinity (Pde1) and a high-affinity (Pde2) cAMP phosphodiesterase. Deletion of the Pdes, especially Pde2, causes the same phenotype as that observed in mutants with an overactive cAMP pathway, in spite of the fact that it increases the cAMP level in the cells about two- to threefold. The catalytic subunits of cAMP-dependent protein kinase are encoded by three genes, TPK1, TPK2 and TPK3, and the regulatory subunits by the BCY1 gene. Of all known mutations in the cAMP pathway, deletion of BCY1 causes the strongest downregulating effect on the targets of the pathway. The three TPK gene products apparently have an overlapping substrate recognition, at least for several targets, and also display different activities. It appears that the main components of the core section of the Ras–cAMP pathway have been identified. The main lack of knowledge for many years was concerned with the upstream and downstream components. In spite of many attempts, no clear triggers or signal transducers located upstream of Cdc25 could be identified and, except for metabolic enzymes, knowledge about possible downstream regulators was also very limited.
G-protein control of adenylate cyclase
Control of adenylate cyclase by heterotrimeric Gs- and Gi-proteins is one of the earliest and best established concepts in the field of signal transduction (Gilman, 1984). It came as somewhat a surprise, therefore, that in the yeast S. cerevisiae, the Ras proteins rather than homologues of the heterotrimeric G-proteins were involved in the control of adenylate cyclase (Wigler et al., 1988). Later work has shown that, in other yeast species, such as Schizosaccharomyces pombe (Fukui et al., 1986; Nadin-Davis et al., 1986), the Ras proteins did not act on adenylate cyclase. However, the G-protein nature of the Ras proteins, the strong effects of their deletion and overexpression on cAMP-controlled phenotypic properties and the results of extensive genetic and biochemical experiments established the strong paradigm that the Ras proteins were regulators of adenylate cyclase in S. cerevisiae (Wigler et al., 1988; Broach and Deschenes, 1990; Thevelein, 1992; 1994; Tatchell, 1993). Apparently, it was also tacitly assumed that the Ras proteins actually replaced the heterotrimeric G-proteins as regulators of adenylate cyclase and that signalling to adenylate cyclase would run through the Ras proteins. The main focus of research in the field subsequently concentrated on the nature of the signals transmitted by the Ras proteins and, in particular, on possible upstream activators of Cdc25 that would transmit these signals to the Ras proteins.
Phenotypic properties controlled by the cAMP–PKA pathway
Depletion of cAMP or inactivation of PKA causes yeast cells to arrest proliferation and to enter permanently into the stationary phase G0. Before entering G0, the cells arrest in the G1 phase of the cell cycle at the same point as nutrient-starved cells. This has led to the concept that the cAMP–PKA pathway in some way signals nutrient availability to the cell cycle machinery. When the cells enter G0, they acquire all the characteristics typical of stationary phase, e.g. strong accumulation of trehalose and glycogen, high stress resistance, low cell wall lyticase sensitivity, strong expression of a variety of genes controlled by STRE elements in their promoter and low expression of ribosomal protein genes. When yeast cells grow on non-fermentable carbon sources, they also display these stationary phase characteristics to some extent. Mutants with reduced activity of the pathway also display these properties when they grow on glucose or other fermentable carbon sources, whereas mutants with enhanced activity of the pathway fail to develop these properties even upon nutrient starvation for glucose or other nutrients. These observations have led to the concept that the cAMP–PKA pathway signals the availability of an optimal growth medium, i.e. a medium with glucose or another rapidly fermented sugar. In cells growing in such a medium, the activity of the pathway is presumed to be high, whereas in cells growing on non-fermentable carbon sources or in stationary phase, the activity of the pathway is considered to be low (Wigler et al., 1988; Broach and Deschenes, 1990; Thevelein, 1992; 1994; Tatchell, 1993). Consistent with this idea, regulatory enzymes of glycolysis are stimulated by high PKA activity, while regulatory enzymes of gluconeogenesis are inhibited. The concept that high PKA activity prevails in cells growing on glucose, whereas low PKA activity prevails in cells deprived of glucose is well accepted. However, there has been great controversy as to whether this difference in PKA activity is caused by stimulation of cAMP accumulation in the presence of glucose (see below).
Cells in which the cAMP–PKA pathway is inactivated arrest at the start point of the cell cycle and do not synthesize cyclins in spite of the presence of a full growth medium (Hubler et al., 1993). Hence, it is clear that one of the direct or indirect targets of PKA must be a factor that is required for cyclin synthesis. Whether this factor is activated by the presence of nutrients through PKA or whether PKA is only required for the activity of this factor and therefore the synthesis of cyclins is unclear. The connection between nutrients, cAMP and cyclin synthesis is also more complex than a simple stimulation of cyclin synthesis by cAMP in response to nutrient availability. Repression of cyclin synthesis by the cAMP–PKA pathway has also been reported (Baroni et al., 1994; Tokiwa et al., 1994). Although the correlation between nutrient control and PKA control of growth, cell cycle progression and several growth-connected characteristics was recognized early on, it has not yet led to the identification of a clear signalling function for the Ras proteins in connecting nutrient availability with growth and cell cycle progression.
The cAMP–PKA pathway also negatively influences the capacity to sporulate and stimulates pseudohyphal growth. Yeast mutants with high PKA activity are unable to sporulate, whereas mutants with strongly reduced PKA activity display constitutive sporulation even in rich medium (Shilo et al., 1978; Toda et al., 1987a,b). The transition of diploid cells to pseudohyphal growth on specific starvation media is stimulated by Ras2val19 (Gimeno et al., 1992), and the three different Tpk catalytic subunits play a specific role in establishing pseudohyphal growth (Robertson and Fink, 1998).
Another important target of the PKA pathway is cAMP synthesis itself. PKA exerts a very strong feedback inhibition on cAMP synthesis, which explains why cAMP levels in pde1 pde2 double mutants are only minimally elevated. Although Ras and several proteins controlling Ras activity have been proposed as direct targets of PKA for the feedback inhibition, recent work has shown that the GTP/GDP ratio of the Ras proteins is not affected in mutants with reduced or enhanced feedback inhibition (Colombo et al., 1998). Hence, adenylate cyclase itself seems to be a more likely candidate for the target of the feedback inhibition mechanism.
Some of the protein substrates of PKA have been identified. These include, in particular, enzymes of carbon and lipid metabolism: trehalase, glycogen synthase, glycogen phosphorylase, phosphofructokinase 2, fructose-1,6-bisphosphatase, isocitrate lyase, phosphatidylserine synthase and phosphatidate phosphatase. Post-translational control of the enzymes of reserve carbohydrate metabolism (in conjunction with control at the transcriptional level; see below) can explain the strong effect of PKA on the levels of trehalose and glycogen.
The yeast Ras proteins as possible signal mediators
There are only two conditions known in vegetative yeast cells that trigger a rapid increase in the cAMP level (Thevelein, 1991) (Fig. 1). The first is the addition of a rapidly fermented sugar to derepressed yeast cells, which triggers a rapid, transient spike in the cAMP level. It is important to emphasize that there is no convincing evidence that the cAMP level would stay higher than the original basal level present before the glucose-induced spike. All evidence indicates that, during growth on glucose, the basal cAMP level is similar compared with that during growth on a non-fermentable carbon source (Eraso and Gancedo, 1984; Ma et al., 1997). Hence, with the term ‘glucose-induced cAMP signalling’, we always refer to the transient cAMP increase observed after the addition of glucose to derepressed cells and not to any possibly more permanent effect on the basal cAMP level in the presence of glucose. Rapidly fermented sugars are sugars such as glucose and fructose that are rapidly and preferentially converted to ethanol, while derepressed cells are cells that have been growing on non-fermentable carbon sources, such as ethanol or glycerol, using respiration. Stationary phase cells are usually derepressed cells, because ethanol and acetate are commonly the last substrates consumed. The second condition is intracellular acidification, which triggers a more pronounced and long-lasting increase in the cAMP level. Experiments with mutants in the Ras–cAMP pathway indicated that both Cdc25 and Ras were required for stimulation of cAMP synthesis by glucose and by intracellular acidification (Thevelein, 1991). The Ras2val19 allele appeared to be unable to mediate glucose-induced cAMP signalling (Mbonyi et al., 1988). Deletion of CDC25 in a Ras2ile152 strain strongly reduced cAMP signalling (Van Aelst et al., 1990), and specific mutations were identified in Cdc25 that abolished glucose-induced cAMP signalling but not viability (Schomerus et al., 1990). A mutant allele of Ras unable to be targeted to the plasma membrane supported viability but not glucose-induced cAMP signalling (Bhattacharya et al., 1995). As the Ras proteins were the only G-proteins known with a clear regulatory effect on adenylate cyclase, these and other results were interpreted as indicating a role for Ras in transmission of the glucose-induced cAMP signal. However, these experiments do not allow us to distinguish between a requirement for active Ras proteins and signal transmission through the Ras proteins. The interpretation is often also compromised by the requirement for suppressor mutations or multicopy suppressor genes that must be present to maintain viability. These may well influence cAMP accumulation through the potent feedback inhibition of PKA on cAMP synthesis, or they could influence the cellular localization of components of the Ras–cAMP pathway, including adenylate cyclase itself.
In mammalian cells, direct evidence for a function of the Ras proteins as signal mediators was obtained by measurements of the ratio of GTP/GDP bound to Ras as a function of time after the addition of several agonists (Downward et al., 1990; Gibbs et al., 1990; Satoh et al., 1990a,b). Recently, we have performed similar measurements of the ratio of GTP/GDP bound to the yeast Ras proteins upon challenge with glucose or intracellular acidification to distinguish between a mere requirement for Ras or signalling through Ras (Colombo et al., 1998). These experiments revealed that intracellular acidification caused a rapid increase in GTP on the Ras proteins. This increase was not dependent on the Cdc25 and Sdc25 proteins and was absent in ira1 ira2 deletion mutants, suggesting that it might be mediated by inhibition of the Ira proteins (Fig. 1). In ira1Δ ira2Δ mutants, intracellular acidification still stimulated cAMP accumulation, indicating the involvement of another target downstream of Ira (Fig. 1). In contrast, no increase in Ras-bound GTP could be detected after the addition of glucose, either in wild-type strains or in strains with reduced feedback inhibition on cAMP synthesis where glucose causes a large and long-lasting increase in the cAMP level. Hence, we concluded that the Ras proteins are apparently not involved in transmission of the glucose signal to adenylate cyclase.
The physiological role of the stimulation of cAMP synthesis by intracellular acidification has always remained rather enigmatic. Suppression of the transient glucose-induced drop in the intracellular pH does not prevent glucose-induced activation of cAMP synthesis (Thevelein et al., 1987). Hence, intracellular acidification does not serve a role as mediator of the glucose effect on cAMP synthesis. We have suggested a possible function for this intracellular acidification control in the maintenance of a proper intracellular pH and ATP level during carbon starvation (Colombo et al., 1998). This is explained in Fig. 2. During carbon starvation, intracellular acidification might link ATP regeneration through mobilization of storage carbohydrates to ATP shortage. In the absence of external carbon sources, the resulting drop in the ATP level will cause a decrease in the intracellular pH. Through its stimulation of cAMP synthesis, this intracellular acidification will cause mobilization of trehalose and glycogen. The resulting increase in activity of the fermentation and respiration pathways will allow recovery of the ATP level. In this way, activation of the Ras proteins by low intracellular pH could help the cells to maintain an appropriate ATP level under starvation conditions. Experimental evidence for this sequence of events includes the slow consumption of trehalose and glycogen during starvation, which is essential for the maintenance of viability (Lillie and Pringle, 1980). Stimulation of intracellular acidification in starved yeast cells by protonophores is well known to cause mobilization of trehalose and enhancement of endogenous fermentation (Stickland, 1956; Berke and Rothstein, 1957; Brady et al., 1961), presumably because of cAMP-triggered activation of trehalase (Thevelein, 1984a,b).
The GPCR system, Gpr1–Gpa2, is required for glucose activation of cAMP synthesis
The failure to observe any glucose-induced increase in GTP on Ras led us to reinvestigate the possible involvement of another G-protein, Gpa2, in glucose activation of cAMP synthesis. The GPA2 gene was cloned based on its homology with mammalian heterotrimeric Gα-proteins. Although overexpression of the gene affected cAMP levels, a gpa2Δ strain still showed a glucose-induced cAMP signal, and the precise function of Gpa2 remained unclear (Nakafuku et al., 1988; Papasavvas et al., 1992). We reasoned that the failure to observe Gpa2 involvement in glucose-induced cAMP signalling might have been caused by interference with the effect of intracellular acidification. Therefore, we added a few mM of glucose shortly before the high glucose concentration. This raises the ATP level and the intracellular pH and normally causes only a minor increase in cAMP. Under such conditions, the cAMP signal triggered by 100 mM glucose was entirely dependent on the presence of the Gpa2 protein. Deletion of Gpa2 also caused effects on several targets of the cAMP–PKA pathway consistent with a reduction in the cAMP level and lowered PKA activity (Colombo et al., 1998). Evidence consistent with a role for Gpa2 in control of the cAMP level was also provided by studies on the effect of Gpa2 on pseudohyphal growth. Gpa2 deletion mutants are deficient in pseudohyphal growth, and this defect can be rescued by conditions that enhance the cAMP level. It has been suggested that Gpa2 was part of the nitrogen-sensing mechanism responsible for the induction of pseudohyphal growth under nitrogen starvation conditions and that its effect is mediated by cAMP (Kubler et al., 1997; Lorenz and Heitman, 1997). Up to now, however, no evidence has been provided that nitrogen starvation lowers the cAMP level in yeast. Deletion of Gpa2 is lethal in the absence of Ras2, which is also consistent with a role for Gpa2 as stimulator of adenylate cyclase (Kubler et al., 1997).
Using the yeast two-hybrid system and Gpa2 as a bait, we and others isolated part of the C-terminus of a seven-span membrane protein with structural and functional homology to GPCRs (Yun et al., 1997; Xue et al., 1998; Kraakman et al., 1999). Growth experiments with deletion mutants in Gpr1 and Gpa2 and determination of PKA-controlled characteristics in such strains are consistent with a role for Gpr1 as activator of Gpa2 in the control of cAMP synthesis. We have also identified Gpr1 in a screen for mutants with delayed glucose-induced loss of heat resistance, and we demonstrated that deletion of Gpr1 abolished glucose-induced cAMP signalling (Kraakman et al., 1999). The effect of Gpr1 deletion as well as Gpa2 deletion on the PKA targets, however, is clearly confined to the transition period to growth on glucose. The general fluctuation in the PKA-controlled characteristics during diauxic growth is not affected (Colombo et al., 1998; Kraakman et al., 1999). This has important consequences for the mechanism that is responsible for establishing the differences in PKA-controlled characteristics between cells growing on glucose and cells deprived of glucose (see below).
The results obtained with the Gpr1 and Gpa2 mutants indicate that the Gpr1 and Gpa2 proteins constitute a GPCR system that activates adenylate cyclase in response to glucose (Fig. 1). Hence, it would function as a sensor for glucose. Up to now, GPCR systems have been identified for a wide variety of ligands: hormones, neurotransmitters, pheromones, light, odorants and chemoattractants (Dohlman et al., 1991; Strader et al., 1994). If a glucose-sensing function for Gpr1 can be confirmed, it would be the first GPCR involved in the detection of a nutrient. Glucose sensing is a process of widespread importance in nature. Well-known systems include glucose sensing in pancreatic beta cells (Matschinsky et al., 1998) and in photosynthetic plant cells (Jang and Sheen, 1997; Smeekens and Rook, 1997). Glucose-sensing GPCRs and possibly GPCRs involved in the sensing of other nutrients might be widespread in eukaryotes. In S. pombe, glucose causes repression of the fbp1 gene through activation of the cAMP pathway, and a homologue of the heterotrimeric Gα-protein family, gpa2/git8, has been implicated in this process (Nocero et al., 1994). Its deletion abolishes glucose-induced cAMP signalling (Isshiki et al., 1992). A GPCR that binds to the gpa2/git8 protein has not yet been reported.
The main challenge at present is to demonstrate that Gpr1 itself binds glucose and therefore constitutes a real glucose receptor. At present, it cannot be excluded that Gpr1 is only required for the expression or functioning of a glucose-sensing system. For instance, Gpr1 could act as an anchoring protein for an intracellular sugar sensor such as a sugar kinase, recruiting it to the plasma membrane and thereby allowing interaction with adenylate cyclase or an associated protein. Other studies on yeast glucose sensors face the same challenge. It has been proposed that Snf3 and Rgt2 act as glucose sensors, respectively, for high and low glucose levels in the regulation of expression of glucose carrier genes (Özcan et al., 1996; 1998). However, there is no evidence yet that Snf3 and Rgt2 actually bind glucose. They could also function as anchoring proteins for the recruitment of an intracellular sugar sensor, such as hexokinase, to the plasma membrane. Neither Snf3 nor Rgt2 is required for glucose activation of cAMP synthesis (F. Rolland et al., submitted).
An actual glucose-sensing function for Gpr1 has become more apparent with the demonstration that Gpr1 is essential for the sensing of extracellular glucose. Elucidation of the precise mechanism of glucose sensing is complicated in yeast and most other systems, because of the requirement for partial metabolism of the glucose. This creates a possibility for sensing the level of glucose present through the activity of glucose-metabolizing enzymes or the level of glucose-derived metabolites, which could function as ‘metabolic messengers’. Also, for activation of the cAMP pathway by glucose, phosphorylation of the glucose is required (Fig. 1). However, neither a straightforward correlation between the increase in sugar phosphates and the increase in cAMP nor a specific sugar carrier seems to be required (Beullens et al., 1988; F. Rolland et al., submitted). It is unclear why glucose phosphorylation is required for glucose activation of cAMP synthesis and at precisely what point it interferes with the cAMP pathway. Recent work in our laboratory has shown that the glucose phosphorylation requirement can be separated from the requirement for glucose activation of the Gpr1–Gpa2-dependent system. The addition of glucose to cells of a multiple glucose carrier deletion strain, lacking detectable glucose transport, does not result in an increase in the cAMP level. However, this is apparently because of the absence of glucose phosphorylation. When the cells are preincubated with a low level of maltose (resulting in phosphorylation of glucose intracellularly derived from maltose), the addition of external glucose causes an immediate spike in the cAMP level (F. Rolland et al., submitted). This increase is dependent on Gpr1 and Gpa2, indicating that the Gpr1–Gpa2 system is required for the detection of extracellular glucose. This set-up has been very useful in determining the ligand specificity and the apparent Ka of the Gpr1-dependent glucose-sensing system independent of the metabolism of the sugar. As glucose activation of the cAMP pathway requires glucose phosphorylation, unmetabolizable sugars, such as l-glucose, or sugars causing ATP depletion, such as 2-deoxyglucose, cannot be evaluated for their potency to stimulate the pathway with a normal set-up in wild-type cells. Using preaddition of maltose to a glucose transport-deficient strain, it was shown that cAMP synthesis is stimulated twice as much by β-d-glucose compared with α-d-glucose, and not by any of the glucose analogues or other sugars tested, except for sucrose (F. Rolland et al., submitted; K. Lemaire et al., unpublished results). The apparent Ka for the activation of cAMP synthesis by glucose, as measured in this system, is high (about 75 mM). In wild-type cells too, a high apparent Ka of about 20–25 mM for glucose activation of cAMP accumulation has been determined previously (Beullens et al., 1988). This glucose concentration is in the same range as that at which yeast cells switch fully from respiration to fermentation. Hence, the Gpr1–Gpa2 system may function in signalling the availability of an optimal concentration of glucose for fermentation. We concluded that glucose activation of the cAMP pathway requires two independent inputs: extracellular Gpr1–Gpa2-dependent detection of glucose and an intracellular glucose-sensing process that is dependent on glucose phosphorylation (F. Rolland et al., submitted). An important issue to be resolved now is precisely how glucose phosphorylation interferes with the Gpr1–Gpa2-dependent activation pathway of adenylate cyclase.
Gpa2 is a member of the heterotrimeric Gα-protein family and therefore is expected to form a complex with β- and γ-subunits like the other Gα members of the family. However, Gpa2 is unusual in that it contains a very long N-terminal extension, which might completely or partially take over the function of a β–γ dimer. On the other hand, the yeast genome sequence contains several genes encoding proteins with the typical size and WD repeats set-up of G-protein β-subunits, which might be partners of Gpa2. A third possibility is that Gpa2 shares β- and γ-subunits (Ste4 and Ste18) with Gpa1, the Gα-protein of the mating pathway (Marsh et al., 1991). This could explain interferences observed previously between activation of the mating factor pathway and glucose-induced cAMP signalling (Arkinstall et al., 1991; Papasavvas et al., 1992). Thus far, neither model has been conclusively discarded or confirmed (Lorenz and Heitman, 1997; J. H. de Winde et al., unpublished results).
Is Ras required for signalling through adenylate cyclase?
The results on the control of yeast adenylate cyclase by the Gpr1–Gpa2 GPCR system bring S. cerevisiae back in line with the general concept of G-protein control of adenylate cyclase in eukaryotes. However, they also raise new questions with respect to Ras control of adenylate cyclase. Obviously, the first issue is how previous results indicating a role for Ras in glucose-induced cAMP signalling can be explained. The inability of Ras2val19 to mediate glucose-induced cAMP signalling can apparently be explained by a strong inhibitory effect of Ras2val19 on cAMP signalling. Introduction of Ras2val19 in a yeast strain with reduced feedback inhibition of PKA on cAMP synthesis strongly inhibits both the glucose- and intracellular acidification-induced cAMP increase (Colombo et al., 1998). Results obtained with cdc25 mutants can be explained by the fact that active Ras and therefore active Cdc25 is required for cAMP signalling by the Gpr1–Gpa2 system, but the signal does not run through Ras. Alternatively, inactivation of Cdc25 drastically reduces the amount of membrane-bound adenylate cyclase (Pardo et al., 1993). Deletion of the RAS genes also results in cytosolic localization of adenylate cyclase, but overexpression of Cdc25 in a ras1Δ ras2Δ bcy1Δ mutant again restores adenylate cyclase plasma membrane localization (Engelberg et al., 1990). Consequently, most of the adenylate cyclase in Cdc25-deficient cells, such as in the cdc25 RAS2ile152 strain (Van Aelst et al., 1990), may no longer be accessible for activation by the Gpr1–Gpa2 system. Bhattacharya et al. (1995) reported that palmitoylation of the yeast Ras2 protein is not required to sustain viability but is required for the glucose-induced cAMP increase. Although they also reported evidence that the palmitoylation apparently somewhat reduced Ras2 activity in vivo, farnesylation of the Ras2 protein alone appears to restore most of the adenylate cyclase activity (Kuroda et al., 1993). Farnesylated, unpalmitoylated Ras2 protein is largely located in the cytosol (Bhattacharya et al., 1995). As deletion of Ras causes delocalization of adenylate cyclase to the cytosol, delocalization of Ras to the cytosol might easily have the same effect. This would then also abolish the accessibility of adenylate cyclase to activation by the Gpr1–Gpa2 system. Hence, the effect of Cdc25 or Ras inactivation on glucose-induced cAMP signalling might be largely explicable by delocalization of adenylate cyclase to the cytosol. An important function of the Cdc25–Ras system in addition to sustaining basal adenylate cyclase activity might therefore be to localize adenylate cyclase to the plasma membrane for optimal interaction with the Gpr1–Gpa2 system. It will not be easy to make a definite distinction between a simple requirement of the Ras proteins for cAMP signalling through the Gpr1–Gpa2 system and signalling through the Ras proteins themselves. Direct demonstration of an increase in GTP bound to Ras after the addition of an agonist appears to be essential in this respect. Mere demonstration that a Ras mutant is deficient in a glucose-induced signalling event (Jiang et al., 1998) does not suffice to assign a signal transmission role to the Ras protein.
If the Gpr1–Gpa2 GPCR system is responsible for glucose control of the cAMP pathway, then what is the main function of the Ras proteins in the control of adenylate cyclase? First of all, it is important to emphasize that the interaction of Ras with adenylate cyclase involves two important aspects: (i) GTP-bound Ras is an activator of adenylate cyclase; and (ii) in the absence of GTP-bound Ras, adenylate cyclase loses most of its activity. Hence, adenylate cyclase is not only activated by GTP-bound Ras, it is also largely dependent on GTP-bound Ras for activity. This complicates the study of any signalling function of Ras. Only in a strain in which the two phosphodiesterase genes are deleted is adenylate cyclase able to synthesize enough cAMP in the absence of Ras to sustain viability. It is unclear why adenylate cyclase activity in yeast is so strongly dependent on Ras. It forms a sharp contrast with Gpa2, the deletion of which apparently has only little effect on basal adenylate cyclase activity. Deletion of Gpa2 has very little effect on the growth rate, whereas deletion of Ras is lethal. Up to now, most attention has been paid to the first aspect of the interaction between adenylate cyclase and Ras: Ras as an activator of adenylate cyclase. It has always been tacitly assumed that Ras mediates agonist-induced signal transduction to the cyclase. The second aspect, why the basal adenylate cyclase activity is so strongly dependent on Ras activity, has received only little attention.
In S. cerevisiae cells, the capacity to synthesize cAMP is very high. The addition of glucose and intracellular acidification trigger a severalfold increase in the cAMP level within seconds. Maybe the association with Ras increases the responsiveness of adenylate cyclase to stimulation by the normal heterotrimeric G-protein. Apparently, it is advantageous to S. cerevisiae to be able to increase its cAMP level rapidly. Possibly, a cAMP-triggered protein phosphorylation cascade rapidly adjusts metabolism from gluconeogenic/respirative to fermentation of the preferred carbon source and therefore stimulates ethanol production (Thevelein, 1994; Thevelein and Hohmann, 1995). Such a rapid adaptation mechanism might have a selective advantage, as the ethanol produced by yeast cells inhibits the growth of competing microorganisms. In stationary phase and during growth on non-fermentable carbon sources, yeast cells display a phenotype indicative of low activity of the cAMP–PKA pathway. Under these conditions, however, adenylate cyclase activity, as measured in isolated plasma membranes, is not reduced. Mutants with an overactive cAMP pathway die very rapidly in stationary phase and are unable to grow on non-fermentable carbon sources (for reviews, see Wigler et al., 1988; Broach and Deschenes, 1990; Thevelein, 1992; 1994; Tatchell, 1993). Hence, it appears to be essential for yeast to restrict and control its high-capacity adenylate cyclase precisely under these conditions. The dependency on Ras proteins may allow a more stringent downregulation of adenylate cyclase activity under conditions in which even slight elevation of the activity would be detrimental.
In spite of many efforts, no clear upstream activator of Cdc25 or Sdc25 has been found. Some proteins have been shown to bind to Cdc25, one of these being Hsp70 chaperone. Geymonat et al. (1998) have proposed that recruitment of Hsp proteins to denatured proteins in response to stress conditions would reduce interaction with Cdc25 and, hence, lead to reduced activity of the cAMP pathway. This would enhance the general stress response, as it would lead to enhanced synthesis of heat shock proteins and trehalose. This model actually proposes a cellular homeostasis function for Cdc25, rather than a function in the transmission of extracellular signals. Whatever the precise function of Ras in yeast, at present it cannot be excluded that Ras could function in the transmission of other extracellular signals besides glucose.
Glucose activation of cAMP synthesis and control of PKA targets in cells growing on glucose
It is now well established that several phenotypic properties that show striking differences between yeast cells growing on glucose and cells growing on non-fermentable carbon sources or in stationary phase are controlled by the cAMP–PKA pathway. However, controversy exists in the literature as to whether this difference is caused by a higher cAMP level in the presence of glucose or whether it is the result of another mechanism, i.e. a glucose-dependent signalling pathway that activates PKA activity in a cAMP-independent way. Evidence in favour of the former hypothesis includes counteracting downregulation of PKA targets during nitrogen starvation by the addition of cAMP in the medium (Boy-Marcotte et al., 1987; Boy-Marcotte et al., 1996), the claim that cAMP levels are higher in cells growing on glucose (François et al., 1987; Russell et al., 1993) and the activation of cAMP synthesis upon addition of glucose to derepressed cells (Thevelein, 1991). We have proposed an alternative hypothesis to explain the differences in PKA-controlled properties between the two cell types. We have suggested that the basal level of cAMP is essentially the same in the two cell types, but that the apparently higher PKA activity in vivo in glucose-growing cells results from an alternative, novel pathway that is only activated by a combination of glucose and all other nutrients essential for growth. This alternative pathway would finally result in activation of the catalytic subunits of PKA in a cAMP-independent way (Thevelein, 1991). We have called it the ‘fermentable growth medium-induced (FGM) pathway’, because it apparently requires both a fermentable carbon source and a complete growth medium for full and sustained activation (Thevelein, 1994). Evidence in favour of this alternative pathway includes glucose activation of cAMP synthesis not being observed in cells growing on glucose, apparently because it is downregulated by the main glucose repression pathway (Beullens et al., 1988), and known PKA targets can be affected by the addition of nitrogen, phosphate or sulphate in a cAMP-independent way in the presence of glucose (Belazzi et al., 1991; Hirimburegama et al., 1992). The existence of the FGM pathway has not been proposed on the basis of molecular genetic data concerning components of the pathway, as is the case for many signalling pathways in yeast. It has been proposed on the basis of the physiological data that indicate that only a full growth medium containing glucose or a related fermentable carbon source is able to cause maintenance of the glucose effect on the targets of the PKA pathway. Hence, there should be one or more pathways controlled by this combination of nutrients that finally converge either on PKA or on one or more systems controlling the same targets as PKA.
The discovery of the Gpr1–Gpa2 system opens a new method of assessing these two alternatives of cAMP-dependent and -independent control. Deletion of GPA2 or GPR1 abolishes glucose activation of cAMP synthesis, but this only slightly delays the establishment of the PKA-controlled characteristics in glucose-growing cells (Colombo et al., 1998; Kraakman et al., 1999). Hence, glucose activation of cAMP synthesis is not essential for the difference in PKA-controlled properties between cells growing on glucose and cells deprived of glucose. This is further supported by the finding that a specific mutation in adenylate cyclase, cyr1met1876, which abolishes all cAMP signalling, causes the same phenotype as the deletion of Gpa2 or Gpr1 (Vanhalewyn et al., 1999). Moreover, reinvestigation of the basal cAMP level during growth on glucose has shown that the previously reported higher cAMP level on glucose medium is entirely associated with the lag phase of the culture and that there is no correlation between the decrease in cAMP and the drop in the glucose level of the medium (Ma et al., 1997). More data have become available on the composition and functioning of the FGM pathway, which support the notion that this pathway is different from the Ras–cAMP pathway. Nitrogen activation of trehalase does not require the Ras proteins or the regulatory subunit of PKA (Durnez et al., 1994). Nitrogen activation of several FGM pathway targets is largely dependent on the presence of glucose, but the short-term activation is largely independent of glucose phosphorylation (Pernambuco et al., 1996). This is in contrast to glucose activation of cAMP synthesis, which is entirely dependent on glucose phosphorylation (Beullens et al., 1988; F. Rolland et al., submitted). The Sch9 protein kinase is required for short-term activation of the FGM pathway but not for glucose activation of PKA targets in derepressed cells (Crauwels et al., 1997). The finding that both Sch9 and PKA catalytic subunit activity are required for the FGM pathway raises the issue as to which of the two kinases is actually phosphorylating the PKA substrates upon activation of the FGM pathway. The problem has become even more intricate with the finding that the deletion of Sch9 results in an increase in PKA activity, as measured in vitro, which explains the higher basal activity of PKA targets observed in vivo in a sch9Δ strain (Crauwels et al., 1997). Apparently, Sch9 acts directly or indirectly as an inhibitor of PKA. As deletion of Sch9 causes slow growth that is suppressed by very high PKA activity, deletion of Sch9 in itself may be lethal in the absence of the partial increase in PKA activity. Also, in fission yeast Schizosaccharomyces pombe, strong evidence for cAMP-independent and even PKA-independent control of classical PKA targets such as neutral trehalase has been reported (Soto et al., 1995a,b; 1997).
PKA control of transcription
A large number of yeast genes are downregulated by PKA through one or more 5′-CCCCT-3′ (and vice versa) elements in their promoter (Ruis and Schuller, 1995; Moskvina et al., 1998; Treger et al., 1998). These elements are called STRE elements (stress responsive elements), because they also mediate induction of these genes by a variety of stress conditions. Expression of STRE-controlled genes is low in cells growing on glucose and high in cells growing on non-fermentable carbon sources and in stationary phase. Similar promoter elements discovered in this context have been called PDS (post-diauxic shift elements; Boorstein and Craig, 1990). Nutrient regulation of gene expression has been interpreted as a function of stress control: expression is lowest in the best growth medium and increases during growth under non-optimal conditions and under nutrient starvation. Whether the same mechanism is involved in nutrient and stress regulation, however, remains unclear. For example, the addition of non-phosphorylatable sugars can initiate repression of STRE-controlled genes but obviously cannot lift nutrient starvation (Pernambuco et al., 1996). Recently, two transcription factors that bind to STRE elements have been identified, Msn2 and Msn4 (Martinez-Pastor et al., 1996; Schmitt and Mcentee, 1996). In msn2Δ msn4Δ mutants, the induction of STRE-controlled genes is reduced; however, a significant induction remains, indicating the involvement of additional transcription factors in the control of STRE elements. Interestingly, the cellular localization of Msn2 and Msn4 is strongly influenced by stress conditions and PKA activity (Görner et al., 1998). A variety of stress conditions causes transfer of Msn2 and Msn4 from the cytosol to the nucleus, while high PKA activity counteracts this process. In mutants with reduced PKA activity, these factors were constitutively located in the nucleus. These results are consistent with the involvement of stress and PKA-regulated intracellular localization of the Msn transcription factors in the general stress response and its modulation by the cAMP–PKA pathway (Görner et al., 1998). Expression of PKA targets after the diauxic shift is significantly dependent on Msn2 and Msn4, suggesting that PKA control is mediated through the STRE elements (Boy-Marcotte et al., 1998; Smith et al., 1998).
Recently, the Rim 15 protein kinase has been identified as a component acting downstream of PKA and required for the development of stationary phase characteristics (Reinders et al., 1998). Rim 15 was previously known as a regulator of meiotic gene expression (Vidan and Mitchell, 1997). In addition, it has now been found to be required for the accumulation of trehalose and glycogen, expression of STRE-controlled genes, acquirement of heat and starvation resistance and for proper G1 arrest. Interestingly, deletion of Rim 15 suppresses the lethality caused by PKA inactivation, suggesting that some of the stationary phase-induced genes may be inhibitory for growth. PKA inhibits Rim 15 protein kinase activity in vitro by phosphorylation, indicating that inactivation of Rim 15 by PKA abrogates the expression of genes inhibitory for growth (Reinders et al., 1998). As STRE-controlled genes are strongly expressed during growth on non-fermentable carbon sources, the question remains as to why putative growth-inhibitory genes are not expressed or their gene products not active under these conditions. Rim 15 does not seem to be required for stress induction of STRE-controlled genes, supporting the fact that nutrient control and stress control of STRE-controlled genes use different regulatory pathways (Reinders et al., 1998).
Another protein that may function downstream of PKA is Sok2. The SOK2 gene was isolated as a suppressor of PKA deficiency. It is required for PKA-controlled characteristics, such as glycogen accumulation and expression of STRE-controlled genes, and deletion of Sok2 exacerbates the phenotype of a tpk-attenuated strain (Ward et al., 1995). The precise position of Sok2 downstream of PKA, in particular with respect to Msn2, Msn4 and Rim 15, has not been determined yet.
PKA activity is known to antagonize sporulation. Recent work has shown that the expression of a central regulator of meiosis, Ime1, is antagonized by glucose and PKA activity via an upstream activating element that is controlled by Msn2 and Msn4. This suggests that PKA and nutrient control of sporulation involves the same factors as nutrient control in vegetative cells (Sagee et al., 1998).
Downregulation of cAMP signalling through PKA-mediated activation of Pde1
Recently, the low-affinity cAMP phosphodiesterase, Pde1, has been identified as a target of PKA. Because of its very low affinity for cAMP (Km = ± 50 μM) and the absence of clear phenotypic effects caused by its deletion or overexpression, a physiological function for this phosphodiesterase has remained obscure. We have shown recently that deletion and overexpression of Pde1 specifically produces strong effects on agonist-induced cAMP signalling and has only little, if any, effect on the basal cAMP level. The opposite is observed for the high-affinity cAMP phosphodiesterase, Pde2, deletion and overexpression of which has very little effect on cAMP signalling but drastically influences the basal cAMP level. The activity of Pde1 in vivo is apparently regulated by PKA phosphorylation on serine-252, suggesting that agonist-induced cAMP signalling is downregulated through PKA activation of Pde1 (Ma et al., 1999).
PKA and control of cell proliferation
Inactivation of PKA blocks growth and causes arrest in the G1 phase of the cell cycle at the same point as nutrient depletion. Interestingly, deletion of Msn2 and Msn4 overcomes growth arrest caused by PKA inactivation, indicating that these transcription factors mediate the expression of proteins causing growth inhibition (Smith et al., 1998). As deletion of the Rim 15 protein kinase also overcomes lethality caused by PKA inactivation and abolishes the acquirement of stationary phase characteristics (Reinders et al., 1998), Rim 15 might act on or in the same pathway as Msn2 and Msn4. One gene whose expression was abolished in a msn2Δ msn4Δ mutant is YAK1, which encodes a protein kinase that is known to inhibit growth (Garrett et al., 1991). The inhibitory effect of Yak1 is independent of Sch9 (Hartley et al., 1994). Restoration of growth in a PKA-inactivated mutant by the deletion of Yak1 requires the presence of Sok1. As overexpression of Sok1 also suppresses lethality caused by PKA inactivation, Sok1 appears to be a positive factor located downstream of Yak1 (Ward and Garrett, 1994).
Growth arrest upon nutrient depletion or PKA inactivation is associated with loss of cyclin synthesis. Evidence has been reported pointing to Cln3 as a possible mediator of nutrient control on the G0/G1 transition, and translational control of Cln3 synthesis by PKA has been proposed as a mechanism for the link with nutrient availability (Hall et al., 1998; Mendenhall and Hodge, 1998).
The recent discovery of a GPCR system, consisting of Gpr1 and Gpa2, being involved in glucose activation of cAMP synthesis has brought new perspectives to the cAMP–PKA signalling pathway in yeast. If Gpr1 itself binds glucose, it would be the first nutrient-sensing GPCR discovered in eukaryotes. Interesting questions to be resolved are the identity of the other heterotrimeric subunits interacting with Gpa2 and possible connections with heterotrimeric G-protein signalling in the pheromone pathway. A central issue is the connection with the requirement for glucose phosphorylation, apparently a nearly universal feature in glucose sensing by eukaryotic cells. The discovery of Gpr1–Gpa2 raises new questions concerning the precise function of the Ras proteins. Previously, the primary issue was the nature of the signals mediated by Ras and the mechanisms involved in the sensing and transmission of these signals via Cdc25 through Ras to adenylate cyclase. At present, the main question is whether the Ras proteins in yeast truly function as signal transmitters or whether they have another role, for instance in cellular homeostasis. More and more evidence indicates that PKA is only part of a complex of protein phosphorylation cascades running both downstream of and parallel to the PKA pathway (Fig. 3). It will be an impressive challenge to untangle the regulation and connections between these phosphorylation cascades and their precise contribution to the control of the many PKA targets, including the control of growth. The discovery of Msn2 and Msn4 has set the pace for the elucidation of transcriptional control of the many STRE-regulated genes by PKA and related pathways. The underlying regulatory network is likely to consist of several overlapping and cross-acting mechanisms, in line with the various environmental conditions that yeast cells encounter and to which they adapt gene expression and metabolism through regulation of PKA.
Part of this review was written during a sabbatical stay by J.M.T. at the Universities of Stellenbosch and the Orange Free State (Bloemfontein) in the framework of an International Scientific and Technological Cooperation project with South Africa (BIL96/27), Ministry of the Flemish Community. Original research in the laboratory in Leuven has been supported by the Fund for Scientific Research — Flanders, the Research Fund of the Katholieke Universiteit Leuven (Concerted Research Actions), Interuniversity Attraction Poles Network P4/30, the Flemish Ministry of Economy through the Institute for Scientific and Technological Research (IWT/EUREKA EU/1431) and the Flanders Interuniversity Institute for Biotechnology (VIB).