Genetic Perturbation of Glycolysis Results in Inhibition of de Novo Inositol Biosynthesis*

In a genetic screen for Saccharomyces cerevisiae mutants hypersensitive to the inositol-depleting drugs lithium and valproate, a loss of function allele of TPI1 was identified. The TPI1 gene encodes triose phosphate isomerase, which catalyzes the interconversion of dihydroxyacetone phosphate (DHAP) and glyceraldehyde 3-phosphate. A single mutation (N65K) in tpi1 completely abolished Tpi1p enzyme activity and led to a 30-fold increase in the intracellular DHAP concentration. The tpi1 mutant was unable to grow in the absence of inositol and exhibited the “inositol-less death” phenotype. Similarly, the pgk1 mutant, which accumulates DHAP as a result of defective conversion of 3-phosphoglyceroyl phosphate to 3-phosphoglycerate, exhibited inositol auxotrophy. DHAP as well as glyceraldehyde 3-phosphate and oxaloacetate inhibited activity of both yeast and human myo-inositol-3 phosphate synthase, the rate-limiting enzyme in de novo inositol biosynthesis. Implications for the pathology associated with TPI deficiency and responsiveness to inositol-depleting anti-bipolar drugs are discussed. This study is the first to establish a connection between perturbation of glycolysis and inhibition of de novo inositol biosynthesis.

Bipolar disorder, also called manic-depressive illness, is a severe psychiatric illness with a prevalence of about 1.5% (1). Lithium and valproate (VPA) 3 are two FDA-approved drugs for the treatment of bipolar disorder. Neither drug is completely effective, but the development of new therapies is hindered by the fact that the mechanisms underlying the therapeutic effects of lithium and VPA are not known. The inositol depletion hypothesis has been proposed to explain the therapeutic effects of lithium in the treatment of bipolar disorder (2). This hypothesis is based on the evidence that inositol monophosphatase is inhibited by therapeutic concentrations of lithium, which can, thus, disrupt the phosphoinositide cycle. More recently, VPA has also been linked to inositol depletion. VPA was found to decrease the concentration of myo-inositol in rat brain after chronic administration (3). Both lithium and VPA cause a decrease in intracellular inositol in yeast (4). Moreover, a recent study showed that lithium, VPA, and carbamazepine, another mood-stabilizing drug, decreased growth cone collapse and increased growth cone area in sensory neurons in culture. These effects were abolished by the addition of inositol (5). These studies suggested that inositol metabolism may be associated with the mechanism of action of lithium and VPA. Surprisingly, very little is known about the molecular control of inositol de novo biosynthesis in human cells.
The de novo biosynthesis of inositol has been extensively characterized in the yeast Saccharomyces cerevisiae. Isolation of spontaneous mutants unable to grow in the absence of inositol in S. cerevisiae was first carried out three decades ago (6). Inositol auxotroph mutants undergo inositol-less death, the abrupt decrease in viable cells when deprived of inositol (7). In the rate-limiting step of inositol synthesis, glucose 6-phosphate (G6P) is converted to myo-inositol 3-phosphate (MIP), catalyzed by the INO1-encoded MIP synthase (8). MIP is then dephosphorylated to myo-inositol by inositol monophosphatase (8). MIP synthase has been purified from a number of species (9 -14). Myo-2-inosose 1-phosphate, an intermediate in the reaction, is a strong competitive inhibitor of yeast MIP synthase (15). Further investigation also demonstrated that myo-2-inosose 1-phosphate analogues, including 2-deoxy-myo inositol 1-phosphate, 1-deoxy-1-(phosphonomethyl)myo-2-inosose, and dihydroxyacetone 1-phosphate (DHAP), are MIP synthase inhibitors (16). The extensive genetic and biochemical characterization of inositol metabolism in yeast make it an excellent model for investigating the mechanisms of action of inositol-depleting drugs. Furthermore, MIP synthase is highly conserved from yeast to humans, as expression of human MIP synthase complements the inositol auxotrophy of yeast ino1 null mutants (17). Consistent with the hypothesis that VPA depletes inositol by indirect inhibition of MIP synthase, human MIP synthase activity is decreased in vivo in the presence of VPA (17).
To determine how lithium and VPA affect the inositol metabolic pathway, we used a genetic approach to identify mutants hypersensitive to lithium and VPA in the absence of inositol. We identified a yeast mutant defective in TPI1, which encodes triose phosphate isomerase. This enzyme catalyzes the interconversion of DHAP and glyceraldehyde 3-phosphate (G3P). In humans, TPI deficiency is a rare multisystem disorder characterized by autosomal recessive inheritance (18). It is associated with chronic hemolytic anemia, recurrent infections, cardiomyopathy, progressive neurologic dysfunction, and death in childhood (18). The crystal structures of Tpi1p from different species indicate that the enzyme is a dimer of two identical subunits (19 -22). The subunit interface is of particular interest because it is critical for dimer stability (23,24). In humans, 80% of TPI-deficient patients carry an inherited substitution of aspartate for glutamate at residue 104 (E104D). This mutation is located at the subunit interface of the Tpi1p dimer, which leads to loss of enzyme activity due to instability of the mutant dimer (25).
In this report we show that a single mutation (N65K) in the dimer interface region of TPi1p completely abolished enzyme activity and led to a 30-fold increase in the intracellular DHAP level. Interestingly, tpi1 exhibited the inositol auxotrophy phenotype. Consistent with inositol auxotrophy, MIP synthase was inhibited by DHAP, and other metabolites of carbohydrate metabolism also inhibited the enzyme. This is the first report indicating that the perturbation of glycolysis inhibits the inositol biosynthetic pathway.
Chemical Mutagenesis-Ethylmethane sulfonate mutagenesis was carried out according to the method of Lindegren et al. (26). Cells were mutagenized to 50 -90% kill.
Screen for Lithium-and VPA-hypersensitive (lvs) Mutants-Mutagenized cells were plated on YPD plates and replica-plated to complete synthetic inositol-free medium with or without 60 mM LiCl or 6 mM VPA. lvs mutants were identified, patched to YPD plates, and replica-plated to medium containing or lacking lithium, VPA, or myo-inositol (75 M). Mutants in which hypersensitivity to lithium and VPA was reversed in the presence of inositol were chosen for further analysis.
Identification of Mutant Genes-Mutants containing mutations in a single nuclear gene were transformed by electroporation with a genomic library in the YCp50 plasmid (pBR236) containing ϳ10 -15 kilobases of yeast genomic DNA or a cDNA library (YES), and transformants were selected by uracil prototrophy. The transformants were replica-plated onto plates containing lithium or VPA to select for increased drug resistance. To confirm that colonies acquired resistance from the plasmid, plasmid DNA was extracted from yeast using a Zymoprep yeast plasmid miniprep kit, amplified in Escherichia coli, extracted using a Wizard Miniprep DNA purification kit, and retransformed into the mutant.
DNA Sequencing-The plasmid inserts of the complementary clones were sequenced using ABI fluorescence technology with primers that flank the insertion site. The nucleotide sequences obtained from the forward and reverse primers were used to identify the cloned genes. Primers were designed that flanked the insertion site (the BamHI site, nucleotide 378) of the genomic fragment in the YCp50 vector. The forward primer begins at nucleotide 329, 5Ј-TTGGAGCCACTATC-GACTACG-3Ј, and the reverse primer begins at nucleotide 399, 5Ј-AT-GCGTCCGGCGTAGAGGATC-3Ј. The primer designed for the YES vector flanked the XhoI site, the forward primer beginning at nucleotide 7767, 5Ј-GGAATTACCAAGACCATTGGC-3Ј, and reverse primer beginning at nucleotide 95, 5Ј-ATGTTGTGTGGAATTGTGAGCG-3Ј. Cycle sequencing was performed with ABI PRISM Dye terminators. The sequences obtained were analyzed by a data base search using Blast in the Saccharomyces Genomic Database to search for sequences showing homology to the sequenced clones.
The TPI1 alleles were isolated from genomic DNA of wild type SMY15 and lvs5 strains. Several internal primers were designed to sequence the entire coding regions of the TPI1 gene and 1-kilobase sequences upstream of the start codon. Cycle sequencing using ABI PRISM Dye terminators was carried out, and data base searching and alignment were performed.
Growth Conditions-Cells were precultured in the presence of 75 M inositol for 24 h. Precultured cells grown in inositol were washed twice and suspended in fresh medium for inoculation. Liquid cultures were inoculated to an A 550 of 0.1 and grown to the indicated growth phase. Cell number was determined by microscopic counting using a hemocytometer, and the desired numbers of cells were spotted on the indicated plates. Cell viability was determined by serial dilution and plating on YPD plates.
Plasmid Construction-A 2012-bp sequence (Chromosome IV, 555639 -557651) containing the TPI1 gene was excised from the isolated genomic plasmid clone by the restriction enzymes HindIII and ClaI and cloned into the YCp50 vector. The resulting recombinant plasmid YCp50-TPI1 contains the complete TPI1 coding sequence with flanking sequences 1187 bp upstream and 82 bp downstream. A 426-bp TPI1 sequence was amplified from yeast chromosomal DNA using the TPI1 5Ј primer, 5Ј-AGAAGCCACAAGTCACTGTCG-3Ј, and TPI1 3Ј primer, 5Ј-CCAACTTGGAAGCCAAGAAC-3Ј. The PCR reaction products were cloned into the pGEM-T EASY vector. The resulting recombinant plasmid pGEM-426 was linearized with NdeI for synthesis of the TPI1 riboprobe.
Measurement of TPI1 and INO1 Expression-Cells were harvested, and RNA was isolated by hot phenol extraction (27), fractionated on an agarose gel, and transferred to a nylon membrane. The blots were hybridized with 32 P-labeled TPI or INO1 riboprobes followed by a riboprobe for the constitutively expressed ribosomal protein gene TCM1 to  (28) using bovine serum albumin as the standard.
Triose Phosphate Isomerase Assay-The specific activity of triose phosphate isomerase was determined in cell-free extracts as described by Maitra and Lobo (29).
Determination of Intracellular DHAP Concentration-Determination of intracellular DHAP concentration was performed in cell-free extracts as described by Compagno et al. (30). Briefly, cells were rapidly collected by filtration through a 0.8-m filter and resuspended in precooled 9% HClO 4. The acidic cell suspension was frozen in liquid nitrogen and thawed on ice three times. Neutralization of the cell-free extract was carried out with 2 M KHCO 3 at pH 7. The DHAP assay was performed in 0.1 M triethanolamine buffer, pH 7.4, 0.3 mM NADH, and 1 unit/ml glycerol-3-phosphate dehydrogenase. To calculate the intracellular DHAP concentration, a yeast cytosolic volume of 1.67 l per mg of dry yeast biomass was assumed (31).
Purification of Recombinant MIP Synthase-Yeast (32) and human (17) MIP synthases were isolated and purified as previously described.
MIP Synthase Assay-Two well established methods were used to measure the MIP synthase activity; the paper chromatographic method of Chen and Charalampous (33) and the colorimetric method of Barnett et al. (34). The effect of high concentrations of DHAP on yeast MIP synthase was assayed by the chromatographic method of Chen and Charalampous (33), because contaminating phosphate in high concentrations of DHAP interfered with the latter procedure. Activity was measured in the presence of 100 mM Tris acetate, pH 7.0, 20 mM NH 4

RESULTS
Identification of tpi1-To identify common targets of lithium and VPA, we implemented a genetic screen to isolate mutants hypersensitive to both drugs in the absence of inositol, in which hypersensitivity was rescued by exogenous inositol. Ethylmethane sulfonate mutagenesis was carried out according to the method of Lindegren et al. (26). Mutagenized wild type cells (35,000) were screened, and 26 mutants, termed lvs mutants (lithium and valproate sensitive), were isolated. All mutants were recessive, and 12 complementation groups were identified, including 6 in which mutants displayed a temperature-sensitive (ts) growth phenotype. One of the ts mutants, lvs5, was characterized further. Phenotypic analysis revealed that lvs5 is auxotrophic for inositol and temperature-sensitive (Fig. 1A). The mutant was unable to grow at 30°C on synthetic media lacking inositol, but 5 M inositol supported growth. The mutant was hypersensitive to lithium or VPA in the presence of 5 M inositol. Supplementation with 75 M inositol could partially restore growth of the mutant in the presence of lithium and VPA (Fig. 1B) but did not restore growth of the mutant at 39°C (Fig. 1A). Further analysis indicated that the mutant exhibited inositol-less death, the precipitous decrease in viability when cells are starved for inositol (7). As seen in Fig. 2, lvs5 cells shifted from Iϩ to IϪ media lost viability similar to ino1⌬ cells, in which MIP synthase is deleted.   DECEMBER 23, 2005 • VOLUME 280 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 41807
To clone the gene mutated in lvs5, the lvs5 mutant was transformed with genomic and cDNA libraries, and transformants were screened for the ability to grow at 30°C in the absence of inositol, in the presence of 60 mM lithium or 6 mM VPA, and at the non-permissive temperature of 39°C on synthetic uracil drop out media. A genomic library plasmid bearing a 7.4-kilobase DNA sequence containing the TPI1 locus and a cDNA library plasmid containing the TPI1 gene complemented the mutant phenotypes, including drug sensitivity, inositol dependence, and temperature-sensitive growth. The TPI1 gene was subsequently cloned into the YCp50 plasmid and transformed into the lvs5 mutant. Expression of the TPI1 gene restored growth in the absence of inositol, in the presence of lithium and VPA, and at 39°C (Fig. 3). Sequencing analysis revealed a single C-to-A mutation in the TPI1 locus of lvs5, resulting in a missense codon leading to a single amino acid change, Asn to Lys, at residue 65 (N65K). The Asn-65 residue is at the dimer interface and is conserved in all species for which the gene has been sequenced (NCBI GenBank TM ) (Fig. 4). The lvs5 mutant will be referred to as tpi1.
Loss of Tpi1p Enzyme Activity and Accumulation of DHAP in tpi1-The specific activity of triose phosphate isomerase and intracellular DHAP levels were measured in cell-free extracts of tpi1 and isogenic wild type (SMY15) cells. As seen in TABLE TWO, triose phosphate isomerase activity is not detectable in tpi1 mutant cells, which accumulate about 30-fold more DHAP than wild type cells. This is consistent with levels found in a previously identified tpi1 mutant (30). The presence of a plasmid-born TPI1 gene in tpi1 restored wild type levels of enzyme activity and intracellular DHAP. The loss of triose phosphate isomerase activity in the tpi1 mutant is not due to decreased expression of TPI1, the levels of which were similar in tpi1 and wild type cells (Fig. 5).
High Levels of DHAP Inhibit MIP Synthase Activity-To determine whether inositol auxotrophy could be explained by inhibition of MIP synthase by DHAP, we assayed the activity of purified yeast MIP synthase in the presence of DHAP. As seen in Fig. 6, DHAP inhibits MIP synthase activity. At 22 mM DHAP, the concentration observed in the tpi1 mutant, yeast MIP synthase activity was completely abolished. This result suggested that inositol auxotrophy of tpi1 is due to inhibition of MIP synthase by increased DHAP levels and predicted that other glycolysis mutants that accumulate DHAP are also inositol auxotrophs. The glycolysis mutant pgk1 is defective in the conversion of 3-phosphoglyceroyl phosphate to 3-phosphoglycerate, and accumulation of DHAP has been demonstrated in this mutant (35). Consistent with this prediction, pgk1 was also unable to grow in the absence of inositol (Fig. 7A) and exhibited inositol-less death when deprived of inositol (Fig. 7B). Inositol auxotrophy of tpi1 could not be explained by the inability to derepress INO1. As seen in Fig. 8A, INO1 is derepressible in the tpi1 mutant when deprived of inositol. INO1 abundance was quantified by Northern analysis of RNA extracted from cells after the shift from Iϩ to IϪ media. INO1 expression was similar in tpi1 and wild type cells (Fig. 8A). Furthermore, when assayed in the absence of DHAP, MIP synthase was FIGURE 3. TPI1 complements the inositol auxotrophic, temperature-sensitive, and lvs phenotypes of lvs5. Isogenic lvs5 and wild type (WT) (SMY15) cells transformed with YCp50 (vec) or with YCp50 carrying the TPI1 gene (TPI1) were precultured in synthetic uracilϪ media plus 75 M inositol at 30°C for about 24 h. Cells were washed twice, quantified, serially diluted, and spotted on uracilϪ plates with or without 75 M inositol, 60 mM lithium, or 6 mM VPA as indicated. Cells were incubated at the indicated temperature for 4 days.

Triose phosphate isomerase activity and intracellular DHAP concentration in tpi1
Cells were grown in Iϩ SC media to the middle logarithmic phase of growth. The specific activity of triose phosphate isomerase and intracellular DHAP concentration were determined in cell-free extracts. ND, not detectable. Data represent the average of three independent experiments.

Strains
Triose phosphate isomerase activity Intracellular DHAP  (37) have been reported. These levels can be extrapolated to ϳ1-3 mM, which is in the range that can inhibit human MIP synthase activity in vitro.

DISCUSSION
The demonstration of inositol auxotrophy in the tpi1 mutant intriguingly links perturbation of glycolysis to the inositol biosynthetic pathway. Intracellular DHAP increased about 30-fold in tpi1, consistent with DHAP levels reported for a previously isolated tpi1 mutant (30). DHAP is a competitive inhibitor of MIP synthase, probably due to its structural similarity to the ketone and phosphate moieties of myo-2-inosose 1-phosphate, an intermediate in the MIP synthase reaction. In the tpi1 mutant the INO1 gene is derepressed, and MIP synthase protein is expressed when inositol is limiting, as shown in Fig. 6. Therefore, inhibition of MIP synthase by DHAP is the most likely explanation for inositol auxotrophy of tpi1. Consistent     DECEMBER 23, 2005 • VOLUME 280 • NUMBER 51 with this, the glycolysis mutant pgk1, which also accumulates DHAP, was also auxotrophic for inositol. Purified yeast and human MIP synthases were inhibited by DHAP as well as by other metabolites of carbohydrate metabolism, including G3P and OAA. These results suggest that genetic perturbation of glycolysis leads to inhibition of inositol de novo biosynthesis by direct inhibition of MIP synthase. This report demonstrates a direct link between glycolysis and inositol biosynthesis and differs mechanistically and qualitatively from previous studies linking regulation of carbohydrate utilization to INO1 expression. Expression of INO1 is affected by the glucose response pathway, an energysaving mechanism that turns off expression of genes required for utilization of carbon sources other than glucose when glucose is present, and turns them on when glucose is limiting (38,39). In yeast, the glucose response pathway is mediated by the Snf1p⅐Snf4p kinase complex and the Reg1p⅐Glc7p phosphatase complex, which regulate Mig1p, a repressor of genes required for utilization of non-glucose carbon sources (39). When glucose is limiting, the Snf1p⅐Snf4p kinase inactivates Mig1p. When glucose is abundant, Mig1p is activated by Reg1p⅐Glc7p phosphatase. Interestingly, the glucose repression pathway affects inositol synthesis at the level of INO1 expression. The snf1⌬ and snf4⌬ mutants display low levels of INO1 expression and are inositol auxotrophs (40,41), whereas the reg1 mutant exhibits increased expression of INO1 and has an inositol excretion phenotype (41,42). Thus, expression of INO1 is altered when the response to glucose is genetically perturbed.

Regulation of Inositol Biosynthesis by Glycolytic Intermediates
The results reported here have implications for the pathology associated with TPI deficiency, one of several well studied glycolytic enzyme deficiencies (18,25,43). In contrast to other deficiencies, TPI-deficient patients have a progressive, severe neurologic disorder (44). The neurological defects may be, in part, a result of inositol deficiency. The brain possesses ϳ6 mM inositol, and alterations in brain inositol concentrations have been reported in a number of pathological conditions, including bipolar disorder, Down syndrome, and diabetic peripheral neuropathy (45). A recent study indicated that MIP synthase activity is inhibited by VPA in crude homogenates of human prefrontal cortex (46), which may explain the observed decrease in intracellular inositol in the brain (3). In patients with TPI deficiency, DHAP levels of 1-3 mM are as much as 100-fold higher than normal (36,37). Human MIP synthase activity is inhibited by 25% in the presence of 1 mM DHAP (data not shown). It is possible that inhibition of MIP synthase by increased DHAP in TPI deficiency leads to inositol depletion in the brain, which may be alleviated by inositol supplementation. It is noteworthy in this regard that inositol has been documented to be an effective treatment for several central nervous system disorders, such as depression, Alzheimer disease, panic disorder, and obsessive-compulsive disorder (47).
This study may have implications for responsiveness to the inositol depleting drugs, lithium and VPA. For example, diabetes-related hyperglycemia is associated with a 3-10-fold increase in DHAP and G3P levels (48 -50). The accumulation of both DHAP and G3P might affect inositol de novo biosynthesis by inhibition of MIP synthase activity and could, thus, have a potential synergistic effect with inositol depleting drugs. Further analysis of the interconnection between glycolysis and inositol de novo biosynthesis may enable us to better understand and treat disorders affecting these metabolic pathways.