Physical and Genetic Interactions of Cytosolic Malate Dehydrogenase with Other Gluconeogenic Enzymes*

A truncated form ( (cid:1) nMDH2) of yeast cytosolic malate dehydrogenase (MDH2) lacking 12 residues on the amino terminus was found to be inadequate for gluconeogenic function in vivo because the mutant enzyme fails to restore growth of a (cid:1) mdh2 strain on minimal medium with ethanol or acetate as the carbon source. The (cid:1) nMDH2 enzyme was also previously found to be refractory to the rapid glucose-induced inactivation and degradation observed for authentic MDH2. In contrast, kinetic properties measured for purified forms of MDH2 and (cid:1) nMDH2 enzymes are very similar. Yeast two-hy-brid assays indicate weak interactions between MDH2 and yeast phosphoenolpyruvate carboxykinase (PCK1) and between MDH2 and fructose-1,6-bisphosphatase (FBP1). These interactions are not observed for (cid:1) nMDH2, suggesting that differences in cellular function between authentic and truncated forms of MDH2 may be related to their ability to interact with other gluconeogenic enzymes. Additional evidence was obtained for interaction of MDH2 with PCK1 using Hum-mel-Dreyer gel filtration chromatography, and for interactions of MDH2 with PCK1 and with FBP1 using surface plasmon resonance. Experiments with the latter technique demonstrated a much lower affinity for interaction of (cid:1) nMDH2 with PCK1 and no interaction between (cid:1) nMDH2 and FBP1. These results suggest that the interactions of MDH2 Conditions— Gene disruption mutants were constructed using the parental yeast strain S173-6B ( MAT a , leu2-3,112, his3-1, ura3-57, trp1-289 ) (20). Disruption of the MDH2 gene ( MDH2 :: HIS3 ) was previously described (2). PCK1 and FBP1 genes were independently disrupted by transformation with PCR prod- ucts containing (cid:4) 45 nucleotides from the 5 (cid:2) and 3 (cid:2) ends of each coding region fused with the kan MX4 coding region (21). Gene disruptions

Three differentially compartmentalized isozymes of malate dehydrogenase (MDH) 1 in Saccharomyces cerevisiae catalyze the NAD(H)-specific interconversion of malate and oxaloacetate. Mitochondrial MDH1 catalyzes a reaction of the tricarboxylic acid cycle, and disruption of the corresponding gene results in an inability to grow with acetate as a carbon source (1,2), a phenotype shared with yeast mutants containing disruptions in genes encoding other tricarboxylic acid cycle enzymes (3)(4)(5). Cytosolic MDH2 is a gluconeogenic enzyme and is required for growth on minimal medium with ethanol or acetate as the carbon source (6), as are the other yeast gluconeo-genic enzymes phosphoenolpyruvate carboxykinase (PCK1) (7) and fructose-1,6-bisphosphatase (FBP1) (8). Peroxisomal MDH3 is proposed to catalyze a step in the glyoxylate pathway, which permits formation of C 4 metabolites from C 2 precursors (9, 10), and has also been shown to be necessary for growth with oleate as the carbon source (11), suggesting an additional role in providing NADH for peroxisomal ␤-oxidation.
The yeast MDH isozymes are all homodimers with similar subunit M r values (33,500 for MDH1; 40,700 for MDH2; and 37,200 for MDH3), and they share primary sequence identities of 43 to 50%. Distinct differences among the enzymes include a 17-residue mitochondrial targeting sequence for MDH1 that is removed upon import (2), and a carboxyl-terminal tripeptide targeting sequence (Ser-Lys-Leu) required for peroxisomal import of MDH3 (11). In addition, MDH2 has a 12-residue extension on the amino terminus that is not present on mature MDH1 or on MDH3 (Fig. 1). We have previously shown that removal of this 12-residue extension, generating a truncated enzyme designated ⌬nMDH2, appears to have no effect on cellular levels of activity of the enzyme (12). However, the ⌬nMDH2 enzyme is resistant to catabolite inactivation, a regulatory phenomenon in yeast involving glucose-induced inactivation and degradation of gluconeogenic and glyoxylate cycle enzymes (13,14). In addition, when glucose is added to a culture of a strain expressing ⌬nMDH2, adaptation to utilization of glucose as a carbon source is slowed (12), suggesting that catabolite inactivation may provide a physiological advantage by rapidly removing enzymes that interfere with optimal rates of glycolysis. In this report, we also provide evidence that the catalytically active ⌬nMDH2 enzyme may not function efficiently in gluconeogenesis in vivo.
Because of the unfavorable equilibrium for formation of oxaloacetate from malate (⌬G 0Ј Х ϩ7 kcal/mol), it has been proposed that physical interactions between malate dehydrogenase and the next enzyme in the same metabolic pathway may be necessary to ensure direct transfer of oxaloacetate (15), a metabolite present in very low cellular concentrations. In support of this proposal, substantial experimental evidence has been presented for interactions between the mitochondrial tricarboxylic acid cycle enzymes, malate dehydrogenase and citrate synthase, in yeast and other organisms (16 -19). In the current study, we present evidence for physical interactions of MDH2 with PCK1 and FBP1. These results suggest an association of these cytosolic enzymes in vivo that may facilitate metabolic flux through gluconeogenesis.
Plasmid Constructs and Yeast Two-hybrid Assays-For growth phenotype analyses, the ⌬mdh2 strain was transformed (22) with pRS314 plasmids (23) carrying authentic (MDH2) or truncated (⌬nMDH2) coding regions of MDH2. The pRS314MDH2 plasmid was previously described (24), and was used for mutagenesis to remove 12 codons following the AUG translation initiation codon (12) to produce the pRS314⌬nMDH2 plasmid.
For subcloning into two-hybrid vectors (Clontech, Inc.), PCR was used to introduce BamHI restriction sites onto the 5Ј and 3Ј ends of the coding region of MDH2 for in-frame fusions with the 3Ј end of the GAL4 DNA-binding domain sequence in pAS2-1 (carrying GAL4  and TRP1 for selection) and with the 3Ј end of the GAL4 transcriptionactivation domain sequence in pACT2 (carrying GAL4 (768 -881) and LEU2 for selection). This generated plasmids designated pASMDH2 and pACTMDH2. A truncated form of the MDH2 gene (⌬nMDH2) lacking codons for the first 12 amino acids was obtained by oligonucleotide-directed mutagenesis of pACTMDH2, and the truncated coding region was transferred to pAS2, generating both pACT⌬nMDH2 and pAS⌬nMDH2.
To construct "reverse" two-hybrid vectors to express fusion proteins with GAL4 domains on the carboxyl-terminal side, the EcoRI and BamHI sites in the multicloning sites of pACT and pAS2-1 were eliminated by oligonucleotide-directed mutagenesis. Subsequent mutagenesis was used to reinsert EcoRI and BamHI sites between the ADH1 promoter and the GAL4 domain coding sequences in these plasmids. The resulting reverse two-hybrid vectors permit in-frame cloning of coding sequences (carrying a 5Ј translation-initiator AUG codon) onto the 5Ј end of the GAL4 domain sequences. PCR fragments containing MDH2 and ⌬nMDH2 coding sequences with EcoRI restriction sites on the 5Ј and 3Ј ends were cloned into the reverse pACT and pAS2-1 plasmids as illustrated in Fig. 4.
To construct conventional two-hybrid vectors (pASPCK1 and pACT-PCK1) containing PCK1 sequences, PCR was conducted with a YEp352 plasmid carrying PCK1 (provided by Dr. E. Cardemil) (25) to generate the PCK1 coding region flanked by EcoRI restriction sites for subcloning. Similar vectors (pASFBP1 and pACTFBP1) containing FBP1 sequences were generated using PCR with yeast genomic DNA as the template to generate the FBP1 coding region with 5Ј and 3Ј XhoI restriction sites for ligation into plasmid SalI or XhoI sites. Coding sequences in two-hybrid plasmids were verified by DNA sequence analysis.
The two-hybrid plasmids were transformed singly and in pairwise pAS/pACT combinations into Y190, and transformants were selected by growth on YNB plates lacking tryptophan and/or leucine. A positive control strain was obtained by transformation with plasmid pCL1, a centromere-based plasmid carrying the full GAL4 gene and LEU2 for selection (26). Transformants were restreaked for colony filter assays of lacZ reporter gene expression as described in protocols provided by Clontech. The time required for colony color development was monitored over a 24-h period. Enzyme assays for ␤-galactosidase activity were conducted using extracts prepared by freeze-thaw cycles from transformants harvested from YP glucose cultures at measured densities (A 600 nm ϭ 1-2). ␤-Galactosidase activity was measured spectrophotometrically (A 578 nm ) at times (min) following color development with chlorophenol red-␤-D-galactopyranoside as the substrate (27). Units are expressed as A 578 nm ϫ 10 3 /min ml of A 600 nm .
Affinity Purification and Kinetic Analyses-To insert tags for affinity purification, MDH2 and ⌬nMDH2 genes on pRS426 plasmids, and the PCK1 gene on a YEp352 plasmid, were used for oligonucleotide-directed mutagenesis to add codons for six histidine residues between the final codon and the stop codon of each gene. Each plasmid contained ϳ500 bp of the authentic 5Ј promoter sequence for expression of each gene. A 2.0-kb DNA fragment containing the FBP1 coding region and adjacent promoter sequences was cloned using PCR from yeast genomic DNA and inserted into pRS426. Oligonucleotide-directed mutagenesis was subsequently used to insert nine codons for the residues in a streptavidin tag (28) into the 3Ј ends of PCK1 and FBP1 genes. Each plasmid was transformed into the yeast strain containing a disruption in the corresponding MDH2, PCK1, or FBP1 chromosomal locus.
For affinity purification, transformants were grown in 10-ml cultures of YNB glucose medium to maintain plasmid selection, then transferred to 250 -1000-ml cultures of YP ethanol medium to induce expression of gluconeogenic enzymes (6). The cells were harvested at A 600 nm ϭ 4 -5, and cell pellets were stored at Ϫ20°C. For purification of histidinetagged enzymes, the cell pellets were lysed with glass beads, and the lysate was diluted 1:1 with a buffer containing 100 mM sodium phosphate (pH 8.0), 600 mM sodium chloride, and 40 mM imidazole. The diluted lysate was incubated with 2 ml of Ni 2ϩ -NTA superflow resin (Qiagen) for 1 h at 4°C and loaded into a 1-cm diameter column. The column flow-through fraction was collected, and the column was washed with three column volumes of a buffer containing 50 mM sodium phosphate (pH 8.0), 300 mM sodium chloride, and 20 mM imidazole. Bound protein was eluted with 1.0 ml of the same buffer containing 250 mM imidazole. For purification of streptavidin-tagged enzymes, 2 g of avidin was added to the cell lysates, and the lysates were loaded onto 1.0-ml streptavidin-agarose columns (immunopure affinity Pak immobilized streptavidin, Pierce). The columns were washed with 5 ml of phosphate buffer (100 mM sodium phosphate, 50 mM sodium chloride, and 1 mM phenylmethylsulfonyl chloride). We found that the streptavidin-tagged enzymes bind weakly to these columns and are eluted with a second 5-ml phosphate buffer wash. Yields for the affinity purified enzymes were ϳ0.5 mg of histidine-tagged enzymes and 0.1-0.2 mg of streptavidin-tagged enzymes per 1.2-1.7 g of cells harvested from 250-ml cultures. Protein concentrations were determined using absorbance at A 280 nm and calculated extinction coefficients (29). Samples obtained during affinity purification were electrophoresed on 10% polyacrylamide/sodium dodecyl sulfate gels for Coomassie Blue staining or for immunoblotting.
For determination of kinetic parameters, 0.15 g of MDH2 His or ⌬nMDH2 His were used in 1.0-ml assays containing 50 mM potassium phosphate (pH 7.5). NADH concentrations were varied from 0 to 0.4 mM in the presence of 0.35 mM oxaloacetate, or oxaloacetate concentrations were varied from 0 to 1.0 mM in the presence of 0.4 mM NADH. Units are defined as micromole of NAD ϩ produced per minute.
Immunoblots and Co-immunoprecipitation-Immunoblot analysis of MDH2 or ⌬nMDH2 was conducted using a polyclonal antiserum prepared against MDH2 (6). The antiserum was preadsorbed with extracts from a ⌬mdh2 strain and used at dilutions of 1:250. A polyclonal antiserum against FBP1 was obtained from Dr. K. Eschrich (32) and was used at dilutions of 1:500. Histidine-tagged enzymes were also detected using a monoclonal penta-histidine antiserum obtained from Qiagen. Immunocomplexes were detected with the enhanced chemiluminescence method (Amersham Bioscience).
For co-immunoprecipitation experiments, cells were pelleted from YP ethanol cultures of the parental strain or, for controls, of gene disruption strains. Pellets of cells from 20 ml of the cultures (A 600 nm ϭ 1.5-2.0) were lysed in 400 l of 10 mM sodium phosphate (pH 7.5) containing a protease mixture (20 g/ml chymostatin, 10 g/ml pepstatin A, 0.5 g/ml leupeptin, 5 mM benzamidine, 20 g/ml E-64, and 2 mM phenylmethylsulfonyl chloride) and sodium chloride at concentrations ranging from 10 to 150 mM. The cleared lysates were incubated for 1 h at 4°C with 25 l of antiserum against MDH2 or FBP1. Protein A-Sepharose beads (50 l; Sigma) were added and the incubation continued for 30 min. The Sepharose beads were pelleted and washed three times with 400 l of 10 mM sodium phosphate buffer (pH 7.5) containing 50 mM sodium chloride, then resuspended and boiled in 50 l of SDS electrophoresis sample buffer. After centrifugation to remove the beads, the supernatants were used for gel electrophoresis and immunoblot analysis.

FIG. 1. Aligned amino-terminal sequences of yeast MDH isozymes.
Identical residues in aligned sequences are shaded. The 17-residue mitochondrial targeting sequence of MDH1 that is removed upon import is shown in parentheses. The 12-residue amino-terminal extension unique to MDH2 is underlined. The engineered ⌬nMDH2 enzyme lacks this 12-residue extension.
Hummel-Dreyer Analyses-Gel filtration chromatography was conducted with a Superdex 200 HR 10/30 column (24 ml bed volume) attached to an Applied Biosystems HPLC model 120A. The chromatography buffer contained 50 mM sodium phosphate (pH 7.4) and 50 mM sodium chloride. Samples and standard proteins (at concentrations of 4 M for calibration and otherwise as indicated in the text) were injected with a 50-l loop and chromatographed at a flow rate of 0.25 ml/min. For some experiments, 4.0 M MDH2 was included in the chromatography buffer.
Surface Plasmon Resonance Analyses-Experiments were performed on a BIAcore 2000 surface plasmon resonance (SPR) instrument using steptavidin and carboxymethyl dextran (CM5) research grade chips (BIAcore, Piscataway, NJ). Runs were conducted at 25°C with HBS buffer (10 mM HEPES, pH 7.0, 150 mM sodium chloride, 3 mM EDTA, and 0.005% surfactant P20) at a flow rate of 5 l/min. Proteins were immobilized on steptavidin chips by injection in 50 mM HEPES (pH 8.0) containing 300 mM sodium chloride. CM5 chips were activated by a 6-min injection of 50% N-hydroxysuccinimide and 50% N-ethyl-NЈ-(dimethylaminopropyl)carbodiimide. Proteins (ligands) were immobilized on activated CM5 chips at 1000 -3500 response units (RU; 1 RU ϭ a change of ϳ1 ng/mm 2 in surface protein concentration) at pH 7.5. After immobilization, the surfaces were blocked by a 6-min injection of 1.0 M ethanolamine (pH 8.5). Soluble analyte proteins or bovine serum albumin as the control were passed over the surfaces at the indicated concentrations for 3 min. The surfaces were regenerated between binding runs by a 3-min injection of 1 M sodium chloride or 4 M guanidine HCl.

Differential Function of MDH2 and ⌬nMDH2
in Vivo-Removal of the amino-terminal 12-residue extension of yeast MDH2 was previously shown to have little effect on activity of the enzyme, because levels of malate dehydrogenase activity and immunodetectable MDH2 were similar in extracts from cells expressing the authentic or truncated enzyme (⌬nMDH2) (12). However, unlike authentic MDH2, ⌬nMDH2 was found to be resistant to glucose-induced degradation, as were the activities of fructose-1,6-bisphosphatase and isocitrate lyase in cells expressing ⌬nMDH2. The latter phenotype suggested a dominant effect on enzymes normally subject to catabolite inactivation (13,14). To further investigate potential growth phenotypes that might be associated with a general effect on gluconeogenesis, we constructed yeast mutants with disruptions in MDH2, FBP1, or PCK1 genes encoding the major gluconeogenic enzymes. As previously reported (6 -8), these mutant strains were found to grow well on plates containing rich medium with C 2 carbon sources (ethanol or acetate), but the strains fail to grow on plates of minimal medium with the same carbon sources. This phenotype was attributed to the requirement for gluconeogenesis for growth in minimal medium that lacks carbon compounds sufficiently abundant in rich medium to allow for growth of these strains. (Note, no growth phenotypes are observed with glucose media because gluoneogenic enzymes are not expressed or functional under this growth condition. ) We additionally found that transformation of the ⌬mdh2 strain with a centromere-based plasmid containing the MDH2 gene restores the ability to grow on minimal medium plates with ethanol or acetate as carbon sources. In contrast, transformation of the same strain with a similar plasmid bearing the ⌬nMDH2 gene fails to restore growth under these conditions. These results were extended by measuring culture doubling times in minimal medium. When shifted from minimal glucose to minimal ethanol medium, the parental strain and the ⌬mdh2 strain expressing MDH2 exhibit a lag of ϳ24 h before attaining doubling times of 5.9 and 6.9 h, respectively (Table I).
In contrast, all three gluconeogenic gene disruption strains and the ⌬mdh2 strain expressing ⌬nMDH2 fail to grow under these conditions. Similar results were obtained when strains and transformants were shifted to minimal medium with acetate as the carbon source (Table I), although doubling times for the parental strain and for the ⌬mdh2 strain expressing MDH2 were significantly longer on this medium. These results suggest that ⌬nMDH2 does not compensate for the normal gluconeogenic function of MDH2.
Similar levels of malate dehydrogenase activity (7-8 units/mg of protein) were measurable with cellular extracts from the ⌬mdh2 strains expressing MDH2 or ⌬nMDH2. However, because of expression of other MDH isozymes in those strains, immunoblot analysis was conducted to specifically examine levels of expression of MDH2 and ⌬nMDH2. As illustrated in Fig. 2, this analysis indicates no significant difference between levels of MDH2 (lane 2) and levels of ⌬nMDH2 (lane 3) in extracts from respective transformants. Levels of the plasmid-expressed enzymes exceed that of chromosomally expressed MDH2 in extracts from the parental strain (lane 1) by 2-3-fold, and MDH2 is clearly absent in the ⌬mdh2 strain (lane 4). Fig. 2 also illustrates the differences in electrophoretic mobilities of authentic MDH2 (lanes 1 and 2) relative to ⌬nMDH2 (lane 3) and relative to purified MDH2 carrying a histidine tag (lane 5), as described below.
To examine activity and kinetic properties of the MDH2 and ⌬nMDH2 enzymes in more detail, we constructed plasmids for production of carboxyl-terminal histidine-tagged versions of both enzymes. The enzymes were affinity purified with similar yields using Ni 2ϩ -NTA chromatography. Basic kinetic properties of the purified enzymes were compared using standard assay conditions. In assays with increasing concentrations (0 -300 M) of oxaloacetate, both enzymes exhibit hyperbolic kinetic curves, as illustrated in Fig. 3A. Higher concentrations of oxaloacetate produce an inhibition of activity (inset in Fig. 3A), resulting in velocities lower than the apparent V max as also reported by others (33). Using data from the hyperbolic regions of the curves, similar apparent V max and K m values for oxaloacetate were obtained for MDH2 and ⌬nMDH2 (Table II). With NADH as the varied assay component, the wild-type and mutant enzymes exhibit similar kinetic curves (Fig. 3B), and both enzymes are inhibited at high concentrations of NADH (inset in Fig. 3B). On average, the ⌬nMDH2 enzyme exhibits slightly higher apparent K m and V max values with NADH than the MDH2 enzyme (Table II). The kinetic parameters obtained for the affinity purified wild-type enzyme are similar to those previously reported for the conventionally purified enzyme (9), indicating that the histidine tag has little effect on catalytic properties of the enzyme.
We also examined kinetic parameters for the reverse reaction as would be required for function in gluconeogenesis. With NAD ϩ as the variable assay component  ⌬nMDH2), and the mutant enzyme exhibits a higher V max value (130 units/mg) than the wild-type enzyme (44 units/mg). Collectively, these data suggest that the authentic and truncated MDH2 enzymes are kinetically quite similar, and that any differences in function in vivo cannot be attributed to inherent catalytic differences or to differences in steady state levels of expression in ethanol-grown cultures. Two-hybrid Analyses of MDH2 and ⌬nMDH2-Among possible explanations for differential function of the authentic and truncated forms of MDH2 in vivo despite their kinetic similarities are differences in interactions with other cellular components. Because of the dominant effect of MDH2 truncation on apparent glucose-induced turnover of other gluconeogenic enzymes (12), and the inability of the ⌬nMDH2 enzyme to support gluconeogenesis as described above, we hypothesized that co-localization and/or interactions among gluconeogenic enzymes in the cytosol might facilitate pathway function and/or regulated turnover, and that the amino-terminal extension of MDH2 might be essential for these processes. To investigate these possibilities, we initially tested possible interactions among gluconeogenic enzymes using the yeast two-hybrid system because this provides a sensitive in vivo assay.
For two-hybrid assays, MDH2 and ⌬nMDH2 were expressed as fusion proteins with DNA-binding and transcriptional activation domains of GAL4 using conventional two-hybrid plasmids. These constructs result in GAL4 domains fused to the amino termini of MDH2 and ⌬nMDH2. However, anticipating the possibility that such fusions might block potential interactions of the amino-terminal extension of MDH2 (34), we also constructed reverse two-hybrid plasmids that permit expression of MDH2 and ⌬nMDH2 with free amino termini and GAL4 domains fused to the carboxyl termini (Fig. 4).
For initial tests of fusion proteins in two-hybrid assays, we determined the ability to detect MDH2 and ⌬nMDH2 homodimeric interactions. As shown in Table III, positive results are obtained with colony filter assays with both conventional and reverse MDH2 and ⌬nMDH2 fusion proteins. Transformants expressing the conventional pairs of MDH2 and ⌬nMDH2 fusion proteins give less robust results (2 h for development of blue color indicating ␤-galactosidase activity) than does a transformant expressing full-length GAL4 (40 min for development of blue colony color). Enzyme activity measurements were generally consistent with colony color except for extracts from the ⌬nMDH2 pair, which gave measurable but much lower levels of activity for ␤-galactosidase. In contrast, both colony filter assays and enzyme assays indicate approximately equivalent ␤-galactosidase expression in transformants expressing the reverse pairs of MDH2 and ⌬nMDH2 fusion proteins, and these levels are similar to those obtained with full-length GAL4. We also tested and found that MDH2 and ⌬nMDH2 fusion proteins can be paired together in various combinations to form "heterodimers" in these two-hybrid assays (data not shown). These results suggest that (a) both the authentic and truncated MDH2 polypeptides expressed with these plasmids are appropriately folded in vivo for homodimeric interactions, (b) that truncation of MDH2 does not alter the strength of polypeptide interactions in the ho-  modimer, and (c) that addition of the GAL4 domains onto the amino termini may slightly retard homodimeric interactions.
To test possible heteromeric interactions of MDH2, we constructed conventional two-hybrid plasmids containing PCK1 and FBP1 coding regions fused to the 3Ј-end of GAL4 domain sequences. These plasmids were co-transformed in all possible pairwise combinations with the conventional and reverse twohybrid plasmids containing MDH2 and ⌬nMDH2 gene fusions. As illustrated in Table IV, most of the assays were negative. However, positive results were observed with two pairs of constructs. Both contain the GAL4 DNA-binding domain fused to the amino terminus of MDH2, which appears to interact with the GAL4 transcription-activation domain fused to either PCK1 or FBP1. With both types of assays, the levels of ␤-galactosidase activity in these transformants are lower than those detected in a transformant expressing full-length GAL4 or the (⌬n)MDH2 homodimers (Table III), but they are substantially higher than background levels detected with nega-tive controls obtained with pairs of empty two-hybrid plasmids or with pairs expressing only one fusion protein. The MDH2/ FBP1 pair exhibits lower levels of ␤-galactosidase activity in enzyme assays than does the MDH2/PCK1 pair. We also tested and found no evidence with this method for an interaction between PCK1 and FBP1 (data not shown).
Because putative interactions for MDH2 with PCK1 or with FBP1 can be detected only when MDH2 is fused to the carboxyl terminus of the GAL4 DNA-binding domain (Table IV, section A), this suggests that the GAL4 transcription-activation domain in this position may block the interactions. With similar conventional two-hybrid constructs, no interactions are detectable for ⌬nMDH2 with PCK1 or FBP1 (Table IV, section B), suggesting that the amino-terminal extension of authentic MDH2 is essential for such interactions. Neither MDH2 nor ⌬nMDH2 in reverse two-hybrid fusion proteins exhibits interactions with PCK1 or FBP1 in these assays (Table IV, sections C and D), suggesting for MDH2 that relatively large GAL4 domains fused to the carboxyl terminus may block interactions observed when the GAL4 DNA-binding domain is fused to the amino terminus of MDH2. The latter suggestion seems reasonable in light of the proximity of amino and carboxyl termini in three-dimensional structures of related malate dehydrogenases (35,36).
In summary, these data from two-hybrid analyses suggest that free amino termini enhance homodimer formation for both MDH2 and ⌬nMDH2. The apparent interactions between MDH2 and PCK1 or FBP1 are observed only in certain pairs of possible domain combinations and orientations. These interactions appear to require the amino-terminal 12-residue extension of MDH2.
Physical Assays for Protein Interactions Involving MDH2-To initiate physical approaches to examine possible interactions among yeast gluconeogenic enzymes, we generated constructs to permit affinity purification of PCK1 and FBP1. Plasmids were constructed for expression in yeast of PCK1 carrying a carboxyl-terminal pentahistidine tag, and for expression of PCK1 and FBP1 carrying nine-residue streptavidin tags on their carboxyl termini. Electrophoretic analyses of fractions FIG. 4. Reverse two-hybrid constructs and proteins. Shown are the relative orientations of coding sequences for (⌬n)MDH2 and for GAL4 domains in the reverse two-hybrid plasmids constructed in this study. MCS* indicates the position of the residual multicloning site sequences from pAS2-1 and pACT plasmids following repositioning of EcoRI and HindIII sites onto the 5Ј side of GAL4 domain sequences as described under "Experimental Procedures." These constructs result in free amino termini for MDH2 or ⌬nMDH2 in the expressed fusion proteins. In contrast, conventional two-hybrid plasmids result in fusion of these amino termini to the carboxyl termini of GAL4 domains. The diagrams of fusion proteins are proportional to relative sizes for MDH2 (376 amino acid residues) and for the GAL4 DNA-binding domain (147 amino acid residues). a Filter assays were conducted as described under "Experimental Procedures." Times were reproduced in two experimental determinations.
b Enzyme assays were conducted as described under "Experimental Procedures" using extracts obtained from pellets of equivalent cell numbers. Values represent averages of two to experimental determinations.
c Positive controls were yeast colonies transformed with a plasmid expressing the full-length GAL4 protein.
-ND a Assays, controls, and abbreviations are as described in the legend for Table III. b "Ϫ," no detectable interaction. c "ND," not determined.
obtained during affinity purification of MDH2 His , FBP1 Strep , PCK1 His , and PCK1 Strep are illustrated in Fig. 5 (panels A-D, respectively). We found that the histidine-tagged enzymes are efficiently purified with good yields, whereas the steptavidintagged enzymes are purified with significantly lower yields. All purified enzymes are active in standard enzyme assays, confirming retention of the native states during purification. Our first test for physical associations among the gluconeogenic enzymes was to assess the possibility of co-purification. As shown in Fig. 5, denaturing electrophoretic analysis of each of the affinity purified enzymes produced no evidence for copurification of significant amounts of other gluconeogenic enzymes. For more sensitive assays, we assayed wash and elution fractions from each affinity column for activity of the other gluconeogenic enzymes. Results of these assays, as well as results of immunoblot analyses of the same fractions with antisera recognizing the other gluconeogenic enzymes, provided no evidence for co-purification (data not shown). We conclude that interactions between MDH2 and PCK1 or FBP1 that are detectable with two-hybrid assays are insufficient to withstand the relatively stringent conditions of affinity chromatography.
Co-immunoprecipitation techniques were also tested as described under "Experimental Procedures" using polyclonal antisera against MDH2 and FBP1. We found some evidence for co-immunoprecipitation of MDH2 with the anti-FBP1 antiserum from extracts of the parental strain, and MDH2 was not precipitated with the same antiserum when extracts were pre-pared from the ⌬FBP1 strain (data not shown). However, these results were difficult to routinely reproduce, and the converse co-immunoprecipitation of FBP1 from parental cell extracts with anti-MDH2 serum was not detected. We therefore focused on using the affinity purified enzymes for other physical assays for protein interactions.
With histidine-tagged forms of MDH2 and PCK1, it was possible to purify sufficient quantities of both enzymes to apply the classical Hummel-Dreyer chromatography method for binding analyses (37)(38)(39). For this, we developed a gel filtration system for resolution of MDH2 and PCK1 and determined their elution patterns relative to a set of standard proteins (inset, Fig. 6). MDH2 eluted with a calculated M r of 71,000 and PCK1 eluted with a calculated M r of 212,000, values, respectively, consistent with the native dimeric and tetrameric forms of these enzymes. The gel filtration column was then equilibrated with buffer containing 4.0 M MDH2 His and loaded with a sample containing 1.0 M PCK1 His . The elution profile showed, in addition to the peak for PCK1 His , a trough in the basal absorbance (data not shown), indicating binding of MDH2 His in the buffer to PCK1 His (39). Similar chromatography runs were conducted with the same buffer and samples containing 1.0 M PCK1 His plus MDH2 His at concentrations ranging from 0 to 14 M. The size of the PCK1 peak eluting in each column run increased with increasing concentrations of MDH2 in the sample. The increases in the size of the PCK1 peak were attributed to binding of MDH2 by PCK1 and were used to calculate the concentration of bound MDH2 (Fig. 6). The slope of the plot indicates a binding to PCK1 of ϳ0.14 mol of MDH2/mol of total MDH2 within these concentration ranges. Because of limitations with absorbance measurements, it was not possible to achieve saturation concentrations. However, an interaction between these two gluconeogenic enzymes is clearly measurable using this technique.
We also tested a second sensitive method for detection of protein-protein interactions, SPR, which requires relatively small amounts of purified proteins. The affinity tags of PCK1 Strep and FBP1 Strep were used initially to couple the proteins (ligands) to the surfaces of strepavidin chips. MDH2 His as the soluble analyte was found to bind to both chips and not to a control chip carrying biotinylated bovine serum albumin. However, the kinetics of binding were slow, precluding accurate measurements of saturation. Therefore, PCK1 Strep and FBP1 Strep were covalently immobilized on carboxymethyl dextran (CM5) chips, which significantly increased the kinetic response to MDH2 binding.
The binding of MDH2 His and ⌬nMDH2 His to CM5 chips carrying PCK1 Strep was measured at equilibrium using a range of concentrations of each analyte protein. The binding of MDH2 by PCK1 was found to be dose-dependent and saturable (q in Fig. 7A), whereas no binding of bovine serum albumin by PCK1 was observed in parallel experiments (data not shown). A calculated maximum of 10,590 RU of MDH2 could be bound to a chip loaded with 1,560 RU of PCK1. Saturation was achieved with Ն8 M MDH2 His as ligand. Binding to PCK1 was also observed using ⌬nMDH2 His as the analyte (E in Fig. 7A). Whereas the calculated maximum binding (10,640 RU of ⌬nMDH2 His by a chip loaded with 996 RU of PCK1) was similar to that obtained with the authentic enzyme, the saturation curve suggests a significantly lower affinity. These binding data were used for Scatchard plot analysis (Fig. 7B)  Collectively, these SPR results provide further evidence for physical interactions among yeast gluconeogenic enzymes. PCK1 and FBP1 exhibit similar affinities but different holoenzyme stoichiometries for binding MDH2. The affinity of PCK1 for ⌬nMDH2 appears to be substantially lower than that for FIG. 6. Hummel-Dreyer analysis of MDH2 binding to PCK1. Gel filtration chromatography was conducted as described under "Experimental Procedures," and elution patterns were determined for MDH2 His and PCK1 His relative to those for indicated protein standards (inset). The size of the PCK1 His peak was used to calculate the concentration of MDH2 His bound following chromatography of samples con- MDH2, and no interaction between FBP1 and ⌬nMDH2 was detectable. Thus, these results from SPR experiments are generally consistent with those obtained with yeast two-hybrid assays using the authentic and truncated forms of MDH2.

DISCUSSION
Cytosolic malate dehydrogenase (MDH2) performs an essential function in gluconeogenesis in yeast, because this enzyme is required for assimilation of C 2 carbon sources such as ethanol or acetate. We have found that a truncated form of this enzyme (⌬nMDH2) lacking the first 12 amino acids fails to provide this function, despite retention of kinetic properties that are essentially indistinguishable from those of the authentic enzyme. Previous studies also indicated that the ⌬nMDH2 enzyme is refractory to catabolite inactivation (12), which normally results in rapid elimination of MDH2 and other gluconeogenic enzymes when glucose is added to a yeast culture growing on non-fermentable carbon sources (13,14). In addition, expression of ⌬nMDH2 appears to have a dominant inhibitory effect on catabolite inactivation of other cellular enzymes (12). These observations led us to hypothesize that some attribute of MDH2, in addition to catalytic activity, is important for normal cellular function, and that this attribute is dependent on the amino-terminal region of the protein. We therefore investigated the possibility that MDH2 may normally interact with the other gluconeogenic enzymes, and that these interactions might be dependent on the presence of the normal amino terminus of the protein. In support of this hypothesis, we have detected interactions of MDH2 with PCK1 and FBP1 using three distinct experimental techniques. Our data also suggest that these interactions are substantially reduced by truncation of the amino terminus of MDH2.
Positive results from yeast two-hybrid assays are potentially the most convincing evidence for interactions of MDH2 (but not of ⌬nMDH2) with PCK1 and with FBP1, because these assays are conducted in vivo. Based on colony filter and enzyme assays for reporter gene expression, these interactions are weak relative to interactions between identical polypeptides to form homodimers of MDH2 (and of ⌬nMDH2). The weakness of heteromeric interactions among gluconeogenic enzymes may in part be because of localization in the nucleus in these assays. The heteromeric gluconeogenic enzyme interactions were also found to be highly dependent on orientation and the type of GAL4 domain fused to MDH2, because positive results are obtained only when the GAL4 DNA-binding domain is fused to the amino terminus of MDH2. This orientation dependence is contrary to our initial prediction that interactions might be stronger with a free amino terminus of MDH2. However, because of the proximity of amino-and carboxyl-terminal ends of malate dehydrogenase polypeptides in crystallographic analyses (35,36), we assume that fusion of GAL4 domains onto the carboxyl terminus of MDH2 blocks heteromeric two-hybrid interactions with PCK1 and FBP1. This is not the case for homodimeric interactions, which appear to be enhanced by the free amino termini on MDH2 (and ⌬nMDH2) subunits.
We used several methods to study interactions of gluconeogenic enzymes in vitro. Whereas MDH2 and FBP1 could be co-immunoprecipitated, this technique was not sufficiently reliable for further analyses. A physical interaction between MDH2 and PCK1 was demonstrated with Hummel-Dreyer chromatography. Physical interactions between MDH2 and PCK1 and between MDH2 and FBP1 were also observed using SPR. In contrast, SPR results indicated a significant reduction in affinity for binding of ⌬nMDH2 by PCK1 and no binding of ⌬nMDH2 by FBP1. Consistent with two-hybrid results, SPR data indicate that interactions of PCK1 and FBP1 with MDH2 are relatively weak, with K d values in the low micromolar range. In fact, these K d values are near the limit for detection with two-hybrid assays (38,40). This limit would explain the inability of two-hybrid assays to detect even weaker ⌬nMDH2/ PCK1 interactions. The stoichiometries obtained for MDH2/ PCK1 and MDH2/FBP1 interactions by SPR are very different and imply differences in individual subunit interactions (10:1 for MDH2 and PCK1 subunits and 1:1 for MDH2 and FBP1 subunits). The stoichiometry for the MDH2/FBP1 interaction is clearly closer to that expected for a physiologically relevant interaction. However, it will be of interest to determine how these values correlate with relative cellular concentrations of the enzymes, because one prediction for association of enzymes in a multienzyme complex is the maintenance of specific stoichiometries within the complex.
Our evidence for physical associations, particularly between MDH2 and PCK1, is consistent with a prediction by Srere (15) that close proximity with the next enzyme within a metabolic pathway may be important to ensure direct delivery of oxaloacetate generated in a reaction catalyzed by malate dehydrogenase. An interaction between sequential enzymes in a given metabolic pathway is frequently cited as evidence for the possibility of "channeling" of intermediates within the pathway FIG. 8. SPR analysis of MDH2 binding to FBP1. A, binding of soluble MDH2 His with FBP1 Strep immobilized on a CM5 chip was measured as the RU value obtained at equilibrium (Req MDH2 ) with different concentrations of MDH2 His . Data were fit by a rectangular hyperbolic function. B, Scatchard plot analysis of MDH2 binding by FBP1 was conducted using an RU value of 8710 for FBP1 (R FBP1 ) immobilized on the CM5 chip. (41). To our knowledge, ours are the first data demonstrating such interactions involving the cytosolic gluconeogenic isozyme of malate dehydrogenase. However, there is abundant experimental evidence for interactions between the mitochondrial tricarboxylic acid cycle isozyme of malate dehydrogenase and citrate synthase. This evidence has been obtained using a variety of methods including gel filtration chromatography and polyethylene glycol precipitation (17). In addition, kinetic models, based on yeast or mammalian fusion proteins containing citrate synthase fused to the amino terminus of malate dehydrogenase, have been presented to support the potential for channeling of oxaloacetate between the soluble enzymes (17,18,42). Vélot and Srere (19) also used a peptide "purtubagen," an in vivo competitor, to identify a region (residues 353-366) of yeast citrate synthase (CIT1) that appears to be important in functional interactions with MDH1 in vivo. In preliminary experiments, we have been unable to detect interactions between MDH1 and CIT1 using yeast two-hybrid assays, but results cited above obtained for the fused CIT/MDH enzymes suggest that such interactions may be highly dependent on orientation and on linker spacing between MDH1 or CIT1 and GAL4 domains.
At present, we have no evidence for a kinetic advantage of physical associations among the gluconeogenic enzymes, but their co-localization would theoretically reduce diffusion of intermediates (e.g. oxaloacetate) away from the pathway. In the current study, we have focused upon the enzymes unique to gluconeogenesis. Clearly, however, this pathway also requires several enzymes that also function in glycolysis. Therefore, it will be of interest to examine potential associations with some of the glycolytic enzymes.
A particular reason for focus on the enzymes unique to gluconeogenesis in yeast is the dramatic coordinate degradation of these enzymes during catabolite inactivation. Within minutes after addition of glucose to cells growing on non-fermentable carbon sources, the unique gluconeogenic enzymes are catalytically inactivated by a process that involves phosphorylation (43,44), then degraded by a mechanism that occurs at a much faster rate than normal cellular protein turnover (13,14). The mechanism for such rapid and specific glucose-induced degradation is of significant interest, but is currently somewhat controversial. Evidence exists both for a mechanism involving a vacuolar protease (45) and/or for a ubiquitin-dependent mechanism involving the cytosolic proteosome (46,47). The amino termini of MDH2 (i.e. the residues missing in ⌬nMDH2) and FBP1 share features including several hydroxylated and charged amino acids. Phosphorylation of Ser-11 of FBP1 appears to be involved in glucose-induced catalytic inactivation but not degradation (43). MDH2, but not ⌬nMDH2, is also phosphorylated in response to glucose (44). However, an aspartate replacement of Ser-12 of MDH2 produced an enzyme still subject to phosphorylation and inactivation but not to rapid degradation. Thus, although details and requirements for catabolite inactivation are not identical for the two enzymes, amino termini of both MDH2 and FBP1 are clearly involved in the process. According to a recent report (48), the most important shared feature in this region may be the presence of an amino-terminal proline residue that is essential for glucoseinduced degradation of FBP1. Structural features important for catabolite inactivation of PCK1, which lacks an aminoterminal proline residue, have not been analyzed. However, it will obviously be of interest in the future to examine the importance of the amino-terminal proline residue of MDH2 in interactions with PCK1 and FBP1 using methods described in this report.