Mth1 receives the signal given by the glucose sensors Snf3 and Rgt2 in Saccharomyces cerevisiae

Abstract

We have determined that the mutant genes DGT1-1 and BPC1-1, which impair glucose transport and catabolite repression in Saccharomyces cerevisiae, are allelic forms of MTH1. Deletion of MTH1 had only slight effects on the expression of HXT1 or SNF3, but increased expression of HXT2 in the absence of glucose. A two-hybrid screen revealed that the Mth1 protein interacts with the cytoplasmic tails of the glucose sensors Snf3 and Rgt2. This interaction was affected by mutations in Mth1 and by the concentration of glucose in the medium. A double mutant, snf3 rgt2, recovered sensitivity to glucose when MTH1 was deleted, thus showing that glucose signalling may occur independently of Snf3 and Rgt2. A model for the possible mode of action of Snf3 and Rgt2 is presented.

Introduction

In the presence of glucose, Saccharomyces cerevisiae increases the rate of transcription of some genes encoding glycolytic enzymes and represses the transcription of genes encoding enzymes necessary for the utilization of alternative carbon sources. In addition, glucose triggers the activation of some enzymes and the inactivation of others (Gancedo and Serrano, 1989). All these processes determine the typical fermentative metabolism of S. cerevisiae and constitute the molecular basis of the so-called Crabtree effect (Barford and Hall, 1979). Not only the presence of glucose but also its concentration determine the nature or magnitude of the effect, as documented in the cases of the genes SUC2, encoding invertase (Lampen, 1968; Carlson and Botstein, 1982; Özcan et al., 1997), and HXT1, HXT2, HXT3 and HXT4, encoding glucose transporters (Özcan and Johnston, 1995).

Repression of different genes by glucose is correlated with the capacity of the yeast cells to take up glucose but not with the presence of any specific glucose transporter (Reifenberger et al., 1997). The induction of the HXT1 to HXT4 genes requires the protein Grr1 (Özcan and Johnston, 1995), which acts by regulating the DNA-binding protein Rgt1 (Li and Johnston, 1997). In grr1 (cat80) mutants, Rgt1 acts permanently as a repressor of the HXT genes, the capacity for glucose transport of the cell is low and glucose repression of a number of genes is relieved (Entian and Zimmermann, 1980; Bailey and Woodward, 1984).

The intimate mechanism of the system(s) that senses and transduces the message of ‘presence of glucose’ in the medium to the cell as well as the nature of the signal(s) is not well known. Evidence has accumulated recently on the role played in glucose sensing by the glucose receptors Snf3 and Rgt2 that act, respectively, at low and high external glucose levels (Özcan et al., 1996, 1997, 1998). Snf3 and Rgt2 are members of a large family of transmembrane proteins that includes many glucose transporters. Snf3 and Rgt2 do not transport glucose at a significant rate (Özcan et al., 1998) and have, as a distinctive feature with respect to sugar transporters, a cytoplasmic C-terminal extension of over 200 amino acids (Kruckeberg, 1996). The C-terminal cytoplasmic taiI of 303 amino acids found in Snf3 contains two nearly identical blocks of 17 amino acids, and at least one of these regions is required for Snf3 function (Celenza et al., 1988; Marshall-Carlson et al., 1990; Coons et al., 1997). A very similar sequence is also found in the 218-amino-acid-long C-terminal extension of Rgt2. These terminal domains are required for the transcriptional response to glucose of the genes HXT1 and HXT2 (Özcan et al., 1998). It was assumed that the C-terminal tails interacted with regulatory proteins responsible for further transmission of the glucose signal but such proteins had not been identified.

Looking for mutations that revert the toxic effects of glucose on certain glycolytic mutants, the dominant mutations HTR1-23 (Özcan et al., 1993), DGT1-1 (Gamo et al., 1994) and BPC1-1 (Blázquez et al., 1995) were identified. They confer similar characteristics to an otherwise wild-type strain: impaired growth on glucose, defects in the transcription of HXT genes and a marked decrease in glucose repression. The mutation HTR1-23 has been cloned (Schulte and Ciriacy, 1995) and found to be an allele of MTH1, a gene homologue to MSN3 (STD1) that partially relieves glucose repression of SUC2 when overexpressed (Hubbard et al., 1994; Tillman et al., 1995). [It should be noted that the name MTH1(CUP1) had also been given to an unrelated gene encoding a metallothionein (Naumov et al., 1992)].

We have studied the possible involvement of the products of DGT1-1 and BPC1-1 in the transmission of the glucose signal and show here that both DGT1-1 and BPC1-1 are allelic with MTH1. We present evidence demonstrating that Mth1 interacts with the C-tails of Snf3 and Rgt2 and that this interaction is responsive to the carbon source in the medium and is affected by the mutations DGT1-1 and BPC1-1.

Results

DGT1-1 and BPC1-1 are allelic forms of MTH1

Mutations DGT1-1 and BPC1-1 were isolated in independent screens aimed at obtaining suppressors of the effects of glucose on certain glycolytic mutants of S. cerevisiae (Gamo et al., 1994; Blázquez et al., 1995). The phenotype produced by them was similar to that caused by HTR1-23 (Özcan et al., 1993), itself allelic to MTH1 (Schulte and Ciriacy, 1995). We therefore examined whether DGT1-1 and BPC1-1, which appeared allelic or closely linked (Blázquez et al., 1995), were also allelic to MTH1. Although a strain with the mutation DGT1-1 is unable to ferment glucose and therefore does not grow on glucose in the presence of the respiratory inhibitor antimycin A, a mth1Δ mutant grows in these conditions. Therefore, we crossed a strain carrying the DGT1-1 gene with a mth1::URA3 mutant strain and examined the segregation of the inability to grow on glucose + antimycin A and uracil prototrophy in 32 complete tetrads. We found a 2⁺:2⁻ segregation for both phenotypes in all cases, and never a co-segregation of uracil prototrophy and inability to grow on glucose + antimycin A. This result indicates that DGT1-1 and MTH1 are either allelic or closely linked.

As next step, the MTH1 genes from a wild-type strain and from mutants carrying the DGT1-1 or the BPC1-1 mutations were isolated by polymerase chain reaction (PCR). The sequence of the corresponding DNAs shows that DGT1-1 and BPC1-1 are allelic forms of MTH1 carrying different mutations (Table 1). Whereas BPC1-1 carried a mutation in codon 85, the same as found in HTR1-23 (Schulte and Ciriacy, 1995), DGT1-1 carried an additional mutation in codon 102. The phenotype observed for DGT1-1 was stronger than that for HTR1-23 or BPC1-1 (Özcan et al., 1993; Gamo et al., 1994; Blázquez et al., 1995). This could be due either to the mutation in codon 102 or to a synergistic effect of the two mutations. We therefore constructed a version of MTH1 carrying only the mutation at codon 102, and tested the effects of this allele in a tps1 mutant that lacked trehalose-6-phosphate synthase and did not grow on glucose (Bell et al., 1992), and in a gpm1 mutant that lacked phosphoglycerate mutase and did not grow in glycerol + ethanol + glucose (Gamo et al., 1994). Whereas both BPC1-1 and DGT1-1 suppressed this phenotype, the allele with only the mutation at codon 102 was ineffective. On the other hand, although the DGT1-1 strain could not grow on glucose + antimycin A, both the BPC1-1 mutant and a strain transformed with a MTH1 allele carrying the 102 mutation grew on this medium. All these results show that the mutation in codon 102, although without visible effects by itself, reinforces those of the mutation in codon 85.

Table 1.  . Allelic forms of the MTH1 gene in S. cerevisiae^a (Hubbard et al., 1994; Schulte and Ciriacy, 1995). a. Different mutant MTH1 genes have been sequenced and compared with the wild-type gene. The amino acid encoded by the mutant codon(s) is indicated in parenthesis.

The fact that different search strategies to isolate mutants with altered responses to glucose yielded alleles of MTH1 almost exclusively (Özcan et al., 1993; Gamo et al., 1994; Blázquez et al., 1995) points to an important role for this gene in glucose signalling.

Effect of different MTH1 alleles on the expression of genes related with glucose transport

Transport of glucose is decreased in strains with a DGT1-1 or a HTR1-23 mutation (Özcan et al., 1993; Gamo et al., 1994). We have examined the influence of the different alleles of MTH1 on the expression of a glucose transporter of low affinity, Hxt1, a transporter of high affinity, Hxt2, and a glucose sensor, Snf3, which is repressed by glucose (Celenza et al., 1988; Bisson et al., 1993; Özcan and Johnston, 1995). The expression of the fusion genes HXT1–lacZ, HXT2–lacZ and SNF3–lacZ in a wild-type strain and in strains carrying the mutations DGT1-1, BPC1-1 or a complete deletion of MTH1, grown in 2% galactose or in media with 0.2% or 2% glucose, is shown in Table 2. The lack of Mth1 had no effect on the expression of HXT1, but strongly increased the expression of HXT2 in the absence of glucose; it also relieved the repression of SNF3 by 0.2% glucose. Dgt1-1 blocked the derepression of HXT1 at high glucose levels and decreased HXT2 expression in the absence of glucose and at 0.2% glucose. It made the expression of SNF3 inducible by glucose. In a BPC1-1 background, expression of HXT1, HXT2 and SNF3 became less sensitive to the presence of glucose and the level of expression was intermediate.

Table 2.  . Effect of different carbon sources on the expression of HXT1, HXT2 and SNF3 in strains with different alleles of MTH1^a. a. Yeasts, carrying the indicated allele of MTH1, were transformed with plasmids, carrying lacZ fused to the promoters of HXT1, HXT2 or SNF3 (see Experimental procedures). The yeasts were grown in minimal medium with the carbon sources indicated and harvested during the exponential phase of growth. β-Galactosidase was assayed as described in Experimental procedures. Values given are averages and standard deviations of three independent experiments.

From these results, it appears that the decrease in glucose transport rate caused by the dominant mutations in MTH1 can be mainly correlated with a decrease in HXT1 transcription.

Search for proteins that interact with Mth1

To identify proteins that could interact with the wild-type Mth1 or with some of its mutated forms, we carried out a two-hybrid screen, using Gal4_BD–Mth1 or Gal4_BD–Dgt1-1 as baits against a library of fusions between the Gal4 activation domain and yeast genomic DNA fragments. We failed to detect specific interactions with Mth1 (about 2 × 10⁷ colonies screened). However, using Gal4_BD–Dgt1-1, one colony from 10⁷ colonies screened showed a specific interaction that allowed the expression of the three reporter genes used, HIS3, ADE2 and lacZ. Sequence analysis of the plasmid carried by this colony revealed that it had a DNA insert with the sequence of the C-terminal domain of Snf3 (amino acids 643–884).

We observed that this Snf3 tail [Snf3(T)] was also able to interact with Mth1 and with Bpc1-1. A quantitative measure of the interaction is shown in 1Fig. 1A. The level of expression of β-galactosidase was dependent on the carbon source; when the Mth1 fusion protein was used it decreased strongly in the presence of both 0.2% and 2% glucose, the effect of glucose was weaker in the case of the Bpc1-1 or Dgt1-1 mutant proteins.

Figure 1. . Interaction of the C-tails of Snf3 and Rgt2 with different variants of Mth1. S. cerevisiae strains pJ696, SNF3 RGT2 (A), or MJL59, snf3 rgt2 (B), transformed with plasmids carrying the corresponding two-hybrid constructs, were grown in SC-Leu-Trp with either lactate, 0.2% glucose or 2% glucose as a carbon source and collected during the exponential phase of growth. The activities of β-galactosidase are means from at least two different cultures.

As the other glucose sensor identified, Rgt2, also has a long C-terminal domain, we used the Rgt2 tail [Rgt2(T)] to examine whether the different Mth1 variants could interact with it (Fig. 1A). At 0.2% glucose the interaction was similar for the different Mth1 variants. It decreased at 2% glucose, except for Dgt1-1. In the absence of glucose, the interaction of Mth1 with Rgt2(T) was markedly increased.

The results shown in 1Fig. 1A suggest that glucose strongly decreases the interaction between Mth1 and Snf3(T) or Rgt2(T). However, it should be taken into account that the interaction measured between a pair of proteins such as Gal4_BD–Mth1 and Snf3–Gal4_AD in different metabolic conditions does not reflect only the capacity of Gal4_BD–Mth1 to bind to Snf3(T) but may be affected by a competition for binding between the native proteins, Mth1, Rgt2 and Snf3, and the introduced fusion proteins. In addition, the Snf3 and Rgt2 tails could show a different response to glucose when fused to Gal4_AD or when part of the plasma membrane proteins Snf3 and Rgt2, able to bind glucose. Therefore, we also performed the two-hybrid experiments, using a yeast strain with both SNF3 and RGT2 genes interrupted as receptor for the protein fusions. Figure 2 shows the positive interactions revealed by the growth in a medium lacking adenine. The levels of β-galactosidase expressed in the different conditions are given in 1Fig. 1B. In the snf3 rgt2 background, the data for the interaction of Mth1, Bpc1-1 or Dgt1-1 with Snf3(T) or Rgt2(T) in the different media do not show large differences. From all the data in Fig. 1, we can conclude that the mutations in Mth1 have a much stronger effect in the interaction of this protein with the tails of the native Snf3 and Rgt2 than with the soluble tails, and that the interaction of Mth1 with these tails shows little sensitivity to the presence of glucose. The experiments in the wild-type background were carried out with a diploid strain and the snf3 rgt2 strain is haploid. However, the differences observed are not influenced by the ploidy, as we verified that the interaction of Mth1 and Snf3(T) in a haploid SNF3 RGT2 strain, grown on high or low glucose, was also much weaker than during growth in lactate (results not shown).

Figure 2. . Interactions between different alleles of MTH1 and the C-terminal tails of the glucose sensors Snf3 and Rgt2. S. cerevisiae ML59 (snf3 rgt2) was transformed with the plasmids indicated to the left of and above the panels. The left-hand picture shows growth on a permissive medium. Positive interactions in the two-hybrid system are shown by the growth on the medium lacking adenine (right-hand panel).

The phenotype of a snf3Δ rgt2Δmutant depends on the presence of MTH1

As Mth1 interacted with Snf3 and Rgt2, we looked at the epistatic relationships between the snf3, rgt2 and mth1 mutations. A double mutant, snf3Δrgt2Δ, grows poorly on glucose, does not grow on glucose plates containing antimycin A and is defective in glucose repression of GAL1 and SUC2 (Özcan et al., 1998). As shown in Table 3, in the snf3 rgt2 mutant, repression of fructose-1,6-bisphosphatase by glucose was also much reduced, and the expression of NAD-dependent glutamate dehydrogenase and isocitrate lyase was even higher in cells grown on glucose. Disruption of MTH1 in the snf3 rgt2 mutant restored growth on glucose + antimycin A (Fig. 3), showing that the ability to ferment glucose had been recovered, and suppressed the defect in catabolite repression of a series of enzymes (Table 3).

Table 3.  . Specific activities of some glucose repressible enzymes in wild type, snf3Δrgt2Δ and snf3Δrgt2Δmth1Δ strains^a. a. The yeasts were grown in rich medium with the indicated carbon sources, as described in Experimental procedures, and harvested during the exponential phase of growth. Enzymes were assayed as described in Experimental procedures. Values given are averages and standard deviations of three independent experiments.

Figure 3. . Effect of the disruption of MTH1 on the growth on glucose + antimycin A of a snf3 rgt2 mutant. The MTH1 gene was disrupted in a snf3 rgt2 background and the triple mutant was streaked on YPD and YPD + antimycin A plates.

Repression of fructose-1,6-bisphosphatase and isocitrate lyase by galactose was unaffected by the different mutations (Table 3). The slight derepression of NAD-dependent glutamate dehydrogenase in the two mutant strains appears marginal and remains unexplained. In any case, it can be concluded that catabolite repression by galactose is independent of the SNF3–RGT2–MTH1 pathway.

In a snf3 rgt2 mutant, HXT1 and HXT2 are no longer induced by glucose and the expression of SUC2 at low concentrations of glucose is reduced (Özcan et al., 1998). In the triple mutant snf3 rgt2 mth1, derepression of SUC2 at low glucose is nearly as marked as in the wild type (Table 4), and repression of SUC2 at 2% glucose is also increased with respect to that observed in the snf3 rgt2 mutant.

Table 4.  . Effect of deleting MTH1 in the expression of invertase in snf3Δrgt2Δ mutants^a. a. Cells were grown on YP medium, containing 2% glucose, to mid-log phase and then transferred to YP medium, containing either 2% or 0.05% glucose, for 3 h and assayed for invertase. Values given are averages and standard deviations of three independent experiments.

Pyruvate decarboxylase is induced by glucose in wild-type cells of S. cerevisiae (Boles and Hollenberg, 1997). We found that induction was impaired in the double mutant, but was recovered in the triple mutant (results not shown). All these results suggest that glucose signalling may operate independently of Snf3 and Rgt2 as long as Mth1 is absent. However, expression of HXT1–lacZ in a triple mutant, snf3 rgt2 mth1, was very low (around 10 mU mg⁻¹ of protein) both in the presence and in the absence of glucose in the medium. Therefore, induction of HXT1 by glucose seems to require at least one of the glucose sensors.

Discussion

We have shown that the mutations DGT1-1 and BPC1-1, which reduce glucose transport and alleviate catabolite repression, are different allelic forms of MTH1. Both have a substitution of Ile⁸⁵ by Asn, and DGT1-1 presents an additional replacement of Ser¹⁰² by Gly. This last mutation, without effect by itself, reinforces the effect of the other and renders the yeast unable to grow on glucose + antimycin A. These results clearly indicate a role for Mth1 in glucose metabolism. This idea was reinforced by the results of the two-hybrid experiments that showed that Mth1 interacts with the C-terminal domains of Rgt2 and Snf3 (Fig. 1).

To study the effect of changes in metabolic conditions on the interaction between two proteins using the two-hybrid approach, some important points should be considered. The strength of the interaction, measured by the quantity of β-galactosidase produced, is a function of the amounts of interacting proteins and of their conformation. As the genes encoding the fusion proteins used in the two-hybrid assay are often under the control of the strong promoter from ADH1, it should be kept in mind that this promoter is activated at least fourfold by glucose (Müller et al., 1995). On the other hand, the response of a fusion protein to changing metabolic conditions could be different from that of the corresponding endogenous protein. In the case of membrane proteins, the whole protein cannot be used in a two-hybrid system requiring nuclear localization of the fusion proteins. In our study, we used only the cytoplasmic C-terminal domain of Snf3 or Rgt2, but the conformation of these isolated tails does not respond to external glucose in the same way as it does when forming part of a membrane protein able to bind glucose. To circumvent this problem, we examined how the presence or absence of the complete proteins Snf3 and Rgt2 affected the interactions between the different Mth1 variants and the C-tails of Snf3 and Rgt2 (Fig. 1). Using such a system, we have observed that, in a snf3 rgt2 background, the interactions between Mth1 and Snf3(T) or Rgt2(T) were strong and not much affected by the mutations in the first protein. Moreover, the carbon source had only a moderate effect on such interactions, indicating that glucose does not directly affect the conformation of the isolated tails.

In the SNF3 RGT2 background, during growth on lactate, the presence of Snf3 and Rgt2 did not interfere with the interactions between the fusion proteins. In most cases, there was a higher level of β-galactosidase for which we do not have an explanation. At 0.2% glucose, the presence of the glucose sensors did not markedly disrupt the interaction between Rgt2(T) and Mth1, Bpc1-1 or Dgt1-1, whereas at 2% glucose there was a clear competition between Rgt2(T) and Rgt2 (and/or Snf3) for Mth1 or Bpc1-1, but not for Dgt1-1. In contrast, both at 0.2% and 2% glucose, the presence of Snf3 and Rgt2 considerably decreased the interaction between Snf3(T) and Mth1. The effect was still seen but was weaker for the interaction between Snf3(T) and Bpc1-1 or Dgt1-1(see Fig. 1A).

Although other interpretations for these observations are possible, a plausible one is that when Snf3 and Rgt2 bind strongly to Mth1 they compete with Snf3(T) and Rgt2(T) for the Mth1 fusion protein available. From the whole set of data, it may be concluded that the interaction of Snf3 and/or Rgt2 with Mth1 is weak in the absence of glucose but strong in its presence. The interaction of the glucose sensors with Bpc1-1 and Dgt1-1 is also increased by glucose but to a smaller degree. The fact that the competition between the glucose sensors was much greater for Snf3(T) than for Rgt2(T) indicates that Mth1 has the highest affinity for Rgt2(T), an intermediate one for Rgt2 and Snf3 and the weakest for Snf3(T).

While this paper was being written, Schmidt et al. (1999) reported an interaction of Mth1 with Snf3. A protein homologue to Mth1, Std1, was also shown to interact with Snf3 and Rgt2. In their work, no data on the influence of the carbon source on those interactions were presented and, in contrast to our results, no interaction between Mth1 and Rgt2 C-tails was found. A possible explanation for this difference could be that we used a longer fragment of the C-terminus than that used by Schmidt et al. (203 amino acids versus 116 or 146). Using proteins fused with GST and tagged with the HA epitope, they did not observe interactions in vitro between Mth1 or Std1 and Snf3 or Rgt2, suggesting that the interactions detected in vivo are only transient, or dependent on particular internal conditions.

The results from the two-hybrid experiments reported above led us to the following hypothesis. The free tails of Snf3 and Rgt2 probably have a loose conformation independent of the concentration of glucose in the medium. However, when they form part of the membrane proteins, Snf3 and Rgt2, their conformation becomes dependent on the external glucose.

Snf3 and Rgt2 are not likely to send a positive signal for the expression of the glucose transporters, as the levels of these transporters are high and similar in the double mutant mth1 std1 and in the quadruple mutant snf3 rgt2 mth1 std1 (Schmidt et al., 1999). It appears more likely that the role of the glucose sensors is to avoid the negative effects of Mth1 and Std1.

The mode of action of Mth1 and Std1 is not yet known, but these proteins are at least partially localized within the nucleus (Schmidt et al., 1999) and Std1 has been shown to suppress defects caused by overexpression of the C-terminus of the TATA binding protein (Ganster et al., 1993). Therefore, Mth1 and Std1 may interfere with the operation of the RNA polymerase II at the HXT promoters. There is also preliminary evidence for the binding of an allelic form of Mth1, Htr1-23, to HXT promoters (Schulte and Ciriacy, 1995). A weak interaction of Std1 with the protein kinase Snf1 has been reported (Hubbard et al., 1994), and repression of HXT1–lacZ transcription by overexpression of STD1 has been shown to be Snf1 dependent (Schmidt et al., 1999). However, Std1 does not contain a consensus sequence for phosphorylation by Snf1 (Smith et al., 1999) and the functional relationship between both proteins is not yet clear. Snf3 and Rgt2 could block the action of Mth1 and Std1 simply by sequestering these proteins and maintaining them out of the nucleus or they could facilitate some covalent modification of them. The free tails Snf3(T) and Rgt2(T), which bind Mth1, do not relieve the repression of HXT1 and HXT2 (Özcan et al., 1998) and do not restore catabolite repression of glutamate dehydrogenase in a snf3 rgt2 strain (our unpublished results). These observations could be explained if the complexes between Mth1 and the tails of the glucose sensors can still repress the HXT genes, as long as they are able to enter the nucleus.

With the available evidence, we propose the following model (Fig. 4). In the absence of glucose, a ‘tight’ conformation of Snf3 and Rgt2 makes them unable to bind Mth1 or Std1. Mth1 strongly inhibits the expression of HXT1, HXT2, HXT3 and HXT4, whereas Std1 has a strong inhibitory effect on HXT1 transcription and only a weak one on that of HXT2, HXT3 and HXT4. This would account for the fact that a deletion of MTH1 results in a large increase of HXT2 transcription in the absence of glucose (Table 2). Mutant alleles such as SNF3-1 or RGT2-1 encode proteins with a more ‘relaxed’ conformation (Fig. 4B). Snf3-1 and Rgt2-1 can bind Mth1 and, though only weakly, Std1, thus allowing a high expression of HXT2 (probably also of HXT3 and HXT4) and a low expression of HXT1 (Özcan et al., 1996, 1998). The fact that overexpression of SNF3 or RGT2 also causes low expression of HXT1 and high expression of HXT2 in the absence of glucose (Özcan et al., 1998) could be the result of a low basal capacity of the uncharged plasma membrane proteins to bind to Mth1. From the data in Table 2, it would appear that Bpc1-1 does not have a marked inhibitory effect itself and can even counteract repression by Std1, something that Dgt1-1 cannot do.

Figure 4. . A possible model for the effects of the interaction of Mth1 and Std1 with the glucose sensors Snf3 and Rgt2. It is proposed that the conformation of the cytoplasmic tails of Snf3 and Rgt2 affects their capacity to bind Mth1 or Std1 and that this conformation is influenced by external glucose. In turn, the free levels of Mth1 or Std1 determine the rate of expression of the HXT genes. The situation in a wild-type yeast is shown in (A), the situation in a yeast with mutated versions of Snf3 or Rgt2 in (B) and the situation in a yeast with a mutant form of Mth1 (Dgt1-1) in (C) (See text for details). The thickness of the lines indicates the strength of repression or the level of transcription. Glucose is shown as a black hexagon and a star shows the position of the mutation in the Snf3-1 and Rgt2-1 proteins. Broken lines around Std1 or Dgt1-1 indicate that only a small proportion of the protein is bound to the glucose sensors; the fact that the amount of Snf3 is low at high glucose is indicated by a cross over the protein.

A low concentration of glucose in the medium will change the conformation of Snf3 and its cytoplasmic tail, which could then bind Mth1, releasing its repression of HXT2, HXT3 and HXT4. Std1 is bound weakly or not at all, and HXT1 remains fully repressed. Rgt2 does not bind glucose and its tail remains in a tight conformation (Fig. 4A). However, the mutant form, Rgt2–1, may bind the sugar, relax its tail and bind Std1 as well as Mth1 (Fig. 4B). As a consequence, the genes HXT1, HXT2, (HXT3 and HXT4) are well expressed (Özcan et al., 1998).

A high glucose level is sensed by Rgt2, its relaxed tail binds Std1 and the repression of the HXT genes is greatly alleviated (Fig. 4A). However, as expression of HXT2 and HXT4 is repressed by Mig1 at high glucose concentrations (Özcan and Johnston, 1996), only HXT1 and HXT3 are expressed. Although MTH1 is repressed by high glucose concentrations (Schmidt et al., 1999), Mth1 still has some repressing effect in this medium, which can be neutralized by the charged Rgt2. As SNF3 is also repressed by high glucose levels (Celenza et al., 1988; Gamo et al., 1994), Snf3 is not likely to play an important role in these conditions.

Mutant variants of Mth1 cause alterations in the interactions described. Dgt1-1 and Bpc1-1 have a lower affinity than Mth1 for the Snf3 or Rgt2 tails bound to glucose. Therefore, at low glucose concentrations, a relevant proportion of the protein is not bound to Snf3 and the expression of HXT2 (and probably that of HXT3 and HXT4) is decreased (Fig. 4C). At high glucose concentrations, Dgt1-1 does not bind Rgt2 efficiently and is therefore able to repress HXT1 and HXT3. Repression by Bpc1-1 is only partial. Although MTH1 is repressed by high glucose concentrations, mutations in this gene, such as HTR1-1 (BPC1-1), relieve this repression (Schulte and Ciriacy, 1995).

It is interesting to note that repression by high glucose concentrations of HXT2 and HXT4 is low in a mth1Δ mutant and that it disappears in a mth1Δstd1Δ double mutant (Schmidt et al., 1999). It is possible that in these conditions Mig1, which binds upstream of the UASs from the HXT2 and HXT4 promoters (Özcan and Johnston, 1996), is no longer able to block the transcription of these genes. In contrast, in the SUC2 gene, the UAS elements and the sites binding Mig1 are overlapping (Bu and Schmidt, 1998), and SUC2 is still repressed by high glucose concentrations in a mth1std1 background (Schmidt et al., 1999).

Although the role of Snf3 and Rgt2 as glucose sensors in yeast is clearly established, we observed that, in a snf3 rgt2 mth1 strain, changes in external glucose concentration still elicited definite responses. Low glucose induced invertase, whereas high glucose induced pyruvate decarboxylase and repressed several enzymes tested. This observation suggests the existence of a glucose-sensing mechanism independent of Snf3 and Rgt2. The lack of effect of glucose in a snf3 rgt2 mutant could be merely a consequence of impaired glucose transport, or may indicate that the glucose sensors have a direct effect on the expression of glucose-regulated genes. The fact that in a snf3 rgt2 strain catabolite repression by galactose is not impaired (Table 3) supports the first interpretation. In addition, in the triple mutant snf3 rgt2 mth1 efficient glucose metabolism is recovered (Fig. 3) as several glucose transporters are constitutively expressed (Schmidt et al., 1999) and the yeast becomes responsive to glucose. A plausible mechanism for glucose signalling independent of the glucose sensors Snf3 and Rgt2 is a change in the concentration of some intermediary metabolite(s). For instance, the concentration of hexose phosphates and of fructose-1,6-bisphosphate increases strongly in a yeast growing on glucose (Gancedo and Gancedo, 1973). This increase in the concentration of intermediary metabolites depends on the flux through the glycolytic pathway, and this flux in turn will be limited by the capacity of the glucose transport.

The signal produced when glucose is being utilized triggers the activation of some genes and the repression of others (Rodríguez and Gancedo, 1999). Although cAMP may play a role in such signalling (Zaragoza et al., 1999), other partially redundant mechanisms should be operative.

Experimental procedures

Plasmids

Plasmid pH 7.5 (Hubbard et al., 1994) was digested with EcoRV and the 3.9 kb fragment containing the MTH1 gene was subcloned in the SmaI site of pUC19, yielding pMJ1. pMJ1 was digested with AflII and NcoI, blunt-ended and ligated to a 1.1 kb fragment containing the URA3 gene to yield plasmid pMTH1::URA3. Plasmids pMTH1::TRP1 and pMTH1::HIS3 were constructed similarly, using the YDp plasmids (Berben et al., 1991).

pFJG054, carrying a disrupted SNF3 gene (provided by F. J. Gamo, this laboratory), was constructed by replacing the 1.4 kb BglII–BamHI fragment from plasmid YEp351-SNF3 (Gamo et al., 1994) by a 1.7 kb BamHI–BamHI fragment, containing the HIS3 gene from S. cerevisiae (Rothstein, 1991).

pRGT2 was constructed as follows. The entire RGT2 coding sequence was obtained using PCR with primers 5′-CGTCCATCGGCATAGAGATGTCGC-3′ and 5′-CGGTGGT CTTACCGGCACCATCTA-3′ and genomic DNA as a template. The 3.8 kb amplified fragment was digested with SphI and XbaI and introduced into the SphI and XbaI sites of pUC19. pRGT2 was digested with Nsi I and HpaI and a 2.2 kb fragment, containing the LEU2 gene, was ligated into it to yield pRGT2::LEU2.

pSNF3–lacZ, carrying the promoter of the SNF3 gene fused in frame to E. coli lacZ was constructed by inserting the 0.8 kb fragment SalI(blunt-ended)–BamHI from pJF177 (given by J. M. François, Toulouse, France) into the EcoRI(blunt-ended)–BamHI sites from plasmid YEp353 (Myers et al., 1986). pHXT1–lacZ, containing the promoter of HXT1 fused to lacZ, was provided by F. Portillo (Madrid, Spain) and pBM2717, containing the promoter of HXT2 fused to lacZ (Özcan and Johnston, 1995), by M. Johnston (St. Louis, USA).

pGBT9 (containing the GAL4 DNA-binding site and the TRP1 marker) and pACT2 (containing the GAL4 activating domain and the LEU2 marker) were from Clontech.

Strains and media

Yeast strains used in this study are listed in Table 5. PJ696 is a derivative of PJ69-4 (James et al., 1996) constructed by D. Locksohn and S. Fields (WA, USA). Strains with a disrupted MTH1 gene were constructed by the transformation of a W303 strain with plasmids pMTH1::URA3, pMTH1::TRP1 or pMTH1::HIS3 previously digested with EcoRI.

Table 5.  . Yeast strains used in this work (Thomas and Rothstein, 1989; James et al., 1996).

To disrupt the chromosomal copy of the SNF3 gene, the 3.6 kb EcoRI–SalI fragment from plasmid pFJG054 was used to transform strains W303, CLF61 and pJ69-4.

Disruption of the RGT2 gene in strain W303 was performed by transforming the yeast with a 3.6 kb HindIII–ScaI fragment isolated from pRGT2::LEU2. As an alternative means of disrupting RGT2, the oligonucleotides 5′-TTGCACAGAAAC CACTATATATATATGGAAATATCTCGAATATTGCTTGTA TGCGTACGCTGCAGGTCGACG-3′ and 5′-ATATAAAAC GGTTTATAAGACCTCGAACGATCGTAAGATGCTATTGG TTTTTAATCGATGAATCGAGCTCGTT-3′ were designed; both contain a RGT2 flanking sequence fused to a KanMX4 flanking sequence. Using these oligonucleotides and DNA from plasmid pF6A-KanMX4 (Wach et al., 1994) as a template, a 1.4 kb PCR product was obtained. This fragment was used to disrupt the chromosomal copy of RGT2 in strain MJL58 to produce strain MJL59.

In all cases, deletions were confirmed by Southern analysis of the transformants.

Strain CJM 284 (snf3 rgt2) was derived from a cross between a snf3::HIS3 strain and a rgt2::LEU2 strain in the W303 background. Strain CJM285 (snf3 rgt2 mth1) was derived from a cross between a snf3::HIS3 mth1::TRP1 strain and a rgt2::LEU2 strain in the W303 background.

All media used, including rich medium (YP), synthetic complete medium (SC), minimal medium (YNB) and sporulation medium were made as has been described by Rose et al. (1990). As carbon sources, glucose (at the concentrations indicated), 2% galactose, 2% lactate or a mixture of 3% glycerol + 2.5% ethanol were used. Antimycin A was added at a final concentration of 2 μg ml⁻¹. To select for the kanamycin resistance marker (KanMX4), yeasts were grown on YPD plates containing 200 mg l⁻¹ of G418 (Geneticin, Sigma). Standard genetic methods for mating, sporulation and tetrad analysis were used throughout this study. Yeast cells were transformed using the lithium acetate method (Ito et al., 1983).

Escherichia coli strains DH5α (supE44 ΔlacU169 (ø80lacZΔM15) hsdR17 recA1 endA1 gyrA96 thi-1 relA1) and HB101 [Δ(gpt-proA)62 leuB6 thi-1 lacY1 hsdS_B20 recL20 (Str^r) ara-14 galK2 xyl-5 mtl-1 supE44 mcrB_B] were used for plasmid propagation and isolation. Transformants were selected by growth on Luria–Bertani (LB) or M9 supplemented with 50 μg ml⁻¹ ampicillin as described by Sambrook et al. (1989). Rescue of the leuB6 in HB101 by the yeast LEU2 was assayed on M9 supplemented with 0.02% proline and 1 mM thiamine-HCl.

Isolation of the genes MTH1, DGT1-1 and BPC1-1

The MTH1 gene and its alleles DGT1-1 and BPC1-1 were amplified with the primers 5′-GATGTCCATGGGATAGTA TTG-3′ and 5′-GCTATCACATCACGTTCTCACCGTGG-3′, using genomic DNA from the wild type and the corresponding mutant strains. The resulting 2.8 kb fragments were subcloned into the centromeric plasmids pRS315 and pRS416 (Sikorski and Hieter, 1989).

A version of the MTH1 gene containing the mutation in the codon 102 was constructed using the oligonucleotides 5′-GCTGAATGCATACTGCCTGAATGATGTG-3′ and 5′-GCT ATCACATCACGTTCTCACCGTGG-3′. The amplified PCR product was cloned into pGEM-T (Promega). A 1.7 kb Nsi I–SpeI fragment was then released and ligated into pMJ1 and digested with Nsi I and SpeI to replace the corresponding fragment of the wild-type gene. All constructions were verified by sequencing; for DGT1-1 two clones corresponding to two different PCRs were used.

Two-hybrid screen

The plasmids that contain the fusion of MTH1, or its dominant alleles DGT1-1 or BPC1-1, in frame with the Gal4 DNA binding domain (Gal4_BD) were constructed as follows. The complete region of MTH1 or its alleles, from 63 bp before the ATG to 43 bp after the stop codon, was amplified using PCR employing the following oligonucleotides as primers: 5′-CTCAATAGCGGATCCACAAGCAGC-3′ and 5′-CGAAGAGTCTGCAGAAAAACCATCG-3′. The 1.4 kb PCR product was digested with BamHI and PstI (underlined sequences) and the resulting fragments were ligated into plasmid pGBT9, previously cut with BamHI and PstI, to generate MTH1-pGBT9, BPC1-pGBT9 and DGT1-pGBT9.

The C-terminal domain of RGT2 (from the codon corresponding to amino acid 570 to the end) was fused in frame with the Gal4 activation domain (Gal4_AD). The RGT2 fragment was amplified using PCR with the primers 5′-GCA AATGGATCCAAAAAATAAGGAAAAGGTGC-3′ and 5′-CA TCTTTTCTCTATGAGATCTTAATTCC-3′. The 1.4 kb PCR product was digested with BamHI and BglII and introduced into pUC19 cut with BamHI. The resulting plasmid was digested with BamHI and SacI and the 1.4 kb fragment cloned into the BamHI and SacI sites of pACT2 to yield RGT2–pACT2.

The yeast two-hybrid system was used to assay for in vivo protein–protein interactions (Fields and Song, 1989), using the yeast mating procedure (Bendixen et al., 1994; Fromont-Racine et al., 1997). Strain PJ696/Matα, bearing a plasmid expressing either a MTH1 fusion (MTH1–pGBT9) or a DGT1-1 fusion (DGT1–pGBT9), was mated to strain PJ696 transformed with a library of fusions between the Gal4 activation domain and yeast genomic DNA fragments (R. Poirey and J.-C. Jauniaux, www.mips.biochem.mpg.de/proj/eurofan_1/b5/b5_results-b.html).

Diploids with putatively interacting fusion proteins were identified on two sets of selective plates, one lacking histidine, tryptophan and leucine and containing 20 mM 3-aminotriazole and the other lacking adenine, tryptophan and leucine. The diploids selected in each set of plates were tested for expression of the other reporter genes. The different pACT2 fusion plasmids from the diploids positive for all reporter genes were rescued by complementation of a leuB6 E. coli strain. The fusion genes were sequenced using the oligonucleotide 5′-GGCTTACCCATACGATGTTC-3′ as primer.

Extracts and enzymatic assays

Cell extracts were prepared by shaking with glass beads in 20 mM imidazole, pH 7, as has been described by Blázquez et al. (1993). Fructose-1,6-bisphosphatase, NAD-dependent glutamate dehydrogenase, isocitrate lyase and pyruvate decarboxylase were measured spectrophotometrically (Gancedo and Gancedo, 1971; Doherty, 1970; Dixon and Kornberg, 1959; Maitra and Lobo, 1971). β-Galactosidase was assayed as described by Miller (1972), using non-centrifuged extracts and performing the centrifugation step before reading the absorbance. Invertase was assayed as described by Goldstein and Lampen (1975), but using whole cells and measuring the glucose formed in the reaction with hexokinase and glucose-6P dehydrogenase. Enzymatic activities are expressed in mU mg⁻¹ of protein. One milliunit is defined as the amount of enzyme that transforms 1 nm of substrate per min in the conditions used. Protein was determined using the Pierce Protein Assay Reagent with bovine serum albumin as a standard.

Acknowledgements

We thank M. A. Blázquez, Marian Carlson, S. Fields, Carmen-Lisset Flores, J. M. François, F. J. Gamo, M. Johnston, D. Locksohn and F. Portillo for providing plasmids or yeast strains. Critical reading of the manuscript and useful comments by O. Zaragoza, Carmen-Lisset Flores and Cristina Rodríguez are warmly acknowledged. This work was supported by grants PB094-0091-CO2-01 and PB97-1213-CO2-01 from the Spanish Dirección General de Investigación Científica y Técnica. María J. Lafuente had short-term FEBS and EMBO fellowships to work in the laboratory of J.-C. Jauniaux.

Footnotes

1. ^*Present address: Human Genetics Unit, Molecular Medicine Centre, Edinburgh EH4 2XU, UK.

2. ‡We dedicate this work to the memory of our friend Michael Ciriacy.