Transcriptional, proteomic, and metabolic responses to lithium in galactose-grown yeast cells.

Lithium is highly toxic to yeast when grown in galactose medium mainly because phosphoglucomutase, a key enzyme of galactose metabolism, is inhibited. We studied the global protein and gene expression profiles of Saccharomyces cerevisiae grown in galactose in different time intervals after addition of lithium. These results were related to physiological studies where both secreted and intracellular metabolites were determined. Microarray analysis showed that 664 open reading frames were down-regulated and 725 up-regulated in response to addition of lithium. Genes involved in transcription, translation, and nucleotide metabolism were down-regulated at the transcriptional level, whereas genes responsive to different stresses as well as genes from energy reserve metabolism and monosaccharide metabolism were up-regulated. Compared with the proteomic data, 26% of the down-regulated and 48% of the up-regulated proteins were also identified as being changed on the mRNA level. Functional clusters obtained from proteome data were coincident with transcriptional clusters. Physiological studies showed that acetate, glycerol, and glycogen accumulate in response to lithium, as reflected in expression data, whereas a change from respiro-fermentative to respiratory growth could not be predicted from the expression analyses.

Yeast cells respond rapidly to several types of stress. Of particular interest is the response to the nutritional status of the cell. Several strategies have been used in yeast, such as glucose deprivation, glucose limitation, and the use of nonmetabolizable glucose analogues (1,2). In Saccharomyces cerevisiae lithium toxicity is dependent on the carbohydrate source used for growth (3,4). During growth on glucose the IC 50 is 100 mM, whereas it is only 6 mM during growth on galactose. The much lower toxicity level during growth on galactose is not a consequence of osmotic stress. Galactose is metabolized in yeast, as in humans, by the Leloir pathway, and the main reason for lithium toxicity during growth on galactose is due to inhibition of phosphoglucomutase (3), an enzyme in the Leloir pathway that reversibly converts glucose 1-phosphate (Glu-1-P) to glucose 6-phosphate (Glu-6-P). Because phosphoglucomutase is an essential protein for the galactose metabolism (5), its inhibition impairs fermentation (3). In glucosegrown cells inhibition of phosphoglucomutase by lithium reduces the level of UDP-glucose, and consequently, the biosynthetic pathways, which use this key metabolite, are impaired (3).
In response to inhibition of fermentation, protein synthesis is inhibited at the translation initiation step (6). During the course of our studies involving the response of cells to lithium in galactose medium we have cloned SIT4, a putative Ser-Thr phosphatase, that when overexpressed confers lithium tolerance in galactose medium (4). SIT4 overexpression also protects translation initiation from inhibition by lithium, however, it does not alter the fermentation rate (6). We have previously observed that SIT4 acts in a pathway not involving induction of transcription of the Na ϩ transporter ENA1. It does not involve extrusion of lithium from the cells, but alters the monovalent cation homeostasis and internal pH (4).
In S. cerevisiae different signals have been reported to inhibit protein synthesis such as amino acid deprivation, depletion of purine (7,8), and depletion of glucose (9,10). Both inhibition of glycolysis by lithium and glucose deprivation lead to inhibition of protein synthesis (6,9), but only the first seems to be mediated by Sit4 as overexpression leads to recovery of protein synthesis when cells are grown in galactose in the presence of lithium but not when cells are deprived of glucose (6). These results suggest that each phenomenon could induce a different mechanism. Glucose sensing and repression are complex phenomena that involve different pathways, including several transcriptional regulators that lead to control of utilization of alternative carbon sources, glycolysis/gluconeogenesis, and respiration (1,11). Other pathways have also been implicated in glucose sensing such as the general amino acid control, the TOR 1 pathway (12)(13)(14), the hexose transporter induction pathway (11), and cAMP protein kinase (15).
In this work we have studied both the transcriptional and proteomic response of galactose grown cells to lithium, which is used to treat bipolar disorder (16), and therefore, therapeutically relevant concentrations of lithium (17) were used. The dynamic response of the intracellular metabolites to addition of lithium was used to correlate the response to the overall carbon metabolism.
Batch Cultivation for Measurement of Metabolites and Transcription-A preculture for the batch cultivation was inoculated with a fresh colony of CEN.PK113-7D grown from a frozen stock on solid YPD (1% yeast extract, 2% bacto-peptone, 2% glucose) (19). The preculture was grown at 30°C and 150 rpm in a cotton-stopped, 500-ml Erlenmeyer flask with baffles containing 100 ml of a defined minimal media of pH 6.5 similar to that described previously (20) but containing different concentrations of galactose (20 g/liter), (NH 4 ) 2 SO 4 (7.5 g/liter), and KH 2 PO 4 (14 g/liter). The aerobic batch cultivation was subsequently inoculated with the exponential growing preculture to an initial concentration of 1 mg of DW/liter in a 4-liter laboratory bioreactor at 30°C with a stirrer speed of 800 rpm and an air flow rate of 4 liter/min. pH was kept at 5.0 by automatic addition of 4 M KOH. A defined medium was used with double concentrations of all components compared with that described previously (20) and contained (per liter): 30 g of galactose; 10.0 g of (NH 4 ) 2 SO 4 ; 6.0 g of KH 2 PO 4 ; 1.0 g of MgSO 4 ⅐7H 2 O; and trace metals and vitamins. 100 l/liter antifoam (Sigma A-8436) was added to avoid foaming. Galactose was autoclaved separately and afterward added to the bioreactor together with a sterile filtrated solution containing the vitamins.
The bioreactor was fitted with a cooled condenser to minimize evaporation of ethanol, and the off-gas was led to an acoustic gas analyzer (1311, Bruël & Kjaer, Naerum, Denmark) to measure the concentrations of CO 2 and O 2 . Production rates of CO 2 and consumption rates of O 2 were calculated as previously described (21). The culture was pulsed with 10 mM LiCl at a residual galactose concentration of 15 g/liter by addition of 1.5 M LiCl stock solution (pH 5.0).
Measurement of Biomass Concentration-The concentration of biomass in the aerobic batch cultivation was determined during the cultivation on a dry weight basis by filtering a known volume of culture through a pre-weighed 0.45-m nitrocellulose filter (Gelman Sciences, Ann Arbor, MI). The filter was washed with distilled water, dried in a microwave oven at 150 watts for 15 min, and, finally, weighed to determine its increase in dry weight.
Analysis of Extracellular Metabolites-Culture samples for determination of glucose, ethanol, glycerol, and acetate concentrations were filtered through a 0.45-m cellulose acetate filter (Osmonics, Minnetonka, MN) immediately after sampling, and the filtrate was frozen at Ϫ20°C until further analysis. The concentrations of the metabolites were determined by high pressure liquid chromatography on an Aminex HPX-87Hm column (Bio-Rad) kept at 65°C and eluted at 0.6 ml/min with 5 mM H 2 SO 4 . Acetate was detected spectrophotometrically by a Waters 486 turnable absorbance detector at 210 nm. Glucose, ethanol, and glycerol were detected refractometrically by a Waters 410 differential refractometer.
Sampling and RNA Isolation-Samples for RNA isolation were taken from the aerobic batch cultivation by rapidly sampling 20 ml of culture into a tube with 35-40 ml of crushed ice. Hereby the temperature decreased to below 2°C in less than 10 s. Cells were quickly pelleted (4500 rpm at 0°C for 2 min), instantly frozen in liquid nitrogen, and thereafter stored at Ϫ80°C. Total RNA was extracted by using the FastRNA kit, Red (BIO 101, Inc., Vista, CA) after thawing the samples on ice.
Probe Preparation and Hybridization to Arrays-mRNA extraction, cDNA synthesis, cRNA synthesis and labeling, as well as array hybridization to Affymetrix yeast S98 arrays were performed as described in the Affymetrix users' manual (Affymetrix GeneChip Expression Analysis Technical Manual (2000), Affymetrix, Santa Clara, CA). Briefly, poly(A) RNA was enriched from total RNA using the Qiagen Oligotex kit and applied for double-stranded cDNA synthesis employing the SuperScript Choice transcriptase II (Invitrogen) and a poly-T primer with a T7 RNA-polymerase promoter sequence. This cDNA was then used as template for in vitro transcription (ENZO BioArray High Yield IVT kit), which amplifies the RNA pool and incorporates biotinylated ribonucleotides required for the staining procedures after hybridization. 15 g of fragmented, biotinylated cRNA was hybridized to Affymetrix Yeast Genome S98 arrays at 45°C for 16 h as described in the Affymetrix users' manual. Washing and staining of arrays were performed using the GeneChip Fluidics Station 400 and scanning with the Affymetrix GeneArray Scanner.
Data Acquisition and Analysis of Array Images-Acquisition and quantification of array images as well as primary data analysis were performed by using the Affymetrix software package Microarray Suite version 5.0. All arrays were globally scaled to a target value of 500 using the average signal from all gene features using Microarray Suite 5.0. The Affymetrix S98 yeast microarrays contain probe sets representing 9335 distinct transcription features. By excluding all probe sets not assigned yeast open reading frame (ORF) abbreviations identified in the Saccharomyces Genome Database (SGD) (available at genomewww.stanford.edu/Saccharomyces/) and all probe sets representing groups of genes that were already represented as singletons, 6072 probe sets remained for analysis. Data were used for all probe sets representing transcripts, called "present" by Microarray Suite 5.0, in at least one array. Genome-wide transcription analysis was performed on cells harvested immediately before (denoted 0 min in the text) and 20, 40, 60, and 140 min after addition of LiCl. From the cells harvested at 0 min the transcription analysis was performed in triplicate allowing an estimation of the standard deviation on the intensity for each gene. Genes were considered as having a changed expression if, when compared at 0 min, they had an mRNA concentration that changed 3-fold or more in at least one time point or 2-fold or more in a minimum of two time points. In addition, -fold changes were only considered if the difference in the intensity between the time point and 0 min was more than three times higher than the standard deviation calculated for each gene at 0 min. The -fold changes were calculated by using the average of the triplicate at 0 min.
Gene functions were annotated in March of 2003 according to SGD (genome-www.stanford.edu/Saccharomyces/), and over-represented gene groups with common biological processes were found by using the SGD Gene Ontology Term Finder (genome-www4.stanford.edu/cgi-bin/ SGD/GO/goTermFinder) and applying a cut-off of p Ͻ 0.01.
Promoter analysis was performed on upstream sequences (from Ϫ800 to Ϫ1) of co-regulated genes using the web-based software Regulatory Sequence Analysis Tools (rsat.ulb.ac.be/rsat/) (22). The oligonucleotide analysis and dyad analysis were used to search for over-represented penta-and hexanucleotides and patterns of two 3-bp sequences spaced by 0 -20 nucleotides. The relative abundance of each over-represented promoter element was determined from a new enquiry of the analyzed gene promoters and the entire set of the 6451 yeast promoters in the genome.
Hierarchical clustering of transcription data was performed with log-transformed -fold changes using Gene Cluster (rana.lbl.gov/Eisen-Software.htm). The algorithm employed by Gene Cluster for hierarchical clustering is called Average Linkage as described previously (23). A centered Pearson correlation was used as the distance (in metric), and the resultant dendrogram was visualized with TreeView (rana.lbl.gov/ EisenSoftware.htm).
Measurement of Intracellular Metabolites-The cell extracts from aerobic batch cultivations and cleanup of sugar phosphates (Gal-1-P, Glu-1-P, Glu-6-P, Man-6-P, and Fru-6-P), were obtained by applying a method described previously (24). Cell extracts were obtained using cold methanol as quenching agent and chloroform as extraction solvent. Furthermore, cleanup of sugar phosphates was done using solid-phase extraction (24). Determination of sugar phosphates was done by anionexchange chromatography with pulsed amperometric detection as described previously (24).
Analysis of Galactitol-The quenching and extraction were done as previously described (25) using a quenching method similar to the one described above for measurement of intracellular metabolites, except that the cell pellet was not washed before extraction of galactitol in buffered boiling ethanol. Furthermore, we used 70 mM Tricine as buffer instead of 70 mM Hepes, and the cell pellets were stored at Ϫ80°C in 2.5 ml, 100% methanol with 70 mM Tricine. Determination of galactitol was done by anion-exchange chromatography with pulsed amperometric detection using 612 mM NaOH for elution and a Carbopac MA1 column (Dionex). The presence of galactitol was determined through retention time measurements and sample spiking with a galactitol standard. 35 S Labeling-Yeast cells strain FY833 was grown to mid-logarithmic phase (A 600 ϭ 1.0) on YNB-galactose medium (6.3 g/100 ml yeast nitrogen base without amino acids, 2% galactose) supplemented with uracil, leucine, histidine, lysine, tryptophan, and adenine and incubated with 15 mM LiCl for 40 or 120 min. An aliquot of 2 ml was withdrawn, 200 Ci of [ 35 S]methionine (200 Ci/mmol) was added, and the cells were lysed for protein expression.
Analysis of Protein Expression-Protein extraction and two-dimen-sional gel electrophoresis were performed as previously described (26) using the Millipore Investigator apparatus. The radioactive gels were recorded by using a PhosphorImager and were analyzed with twodimensional gel analysis software (Melanie II, Bio-Rad). The spot intensities were obtained in pixel units and normalized to the total radioactivity of the gel or using actin as standard. Both methods gave similar profiles. The lithium stimulation index was calculated as the ratio of spot intensity between lithium and standard conditions. The proteins reported have a stimulation index higher than 1.5 or inhibition index lower than 0.6. Proteins were identified by matching two-dimensional gels with a reference gel containing more than 450 previously identified proteins (27). Data presented are the mean of two independent experiments.
Extraction and Analysis of Glycogen and ATP-Yeast cells (FY833) were grown to mid-logarithmic phase in YP-galactose medium to midlogarithmic phase, and none or 15 mM lithium was added during 2 h. For analysis of glycogen, aliquots of the culture were collected by centrifugation at 3000 ϫ g for 10 min. Cells were immediately transferred to 80% ethanol and washed twice. They were further resuspended in 1 ml of 0.25 M Na 2 CO 3 and incubated for 90 min at 95°C. 0.05 ml of 3 M acetic acid and 0.7 ml of 0.2 M sodium acetate, pH 4.8, were added to a 0.2-ml aliquot of this extract. 2 units of amyloglucosidase (Sigma) was added, and the mixture was incubated for 22 h at 37°C. Afterward, samples were boiled during 5 min, and glucose was assayed with a commercial glucose oxidase-peroxidase-coupled assay (Doles reagents, Brazil). For analysis of intracellular ATP, cells were collected by vacuum filtration in a 0.45-m Millipore filter and rapidly transferred to 1 ml of 6% perchloric acid. The solution was neutralized with KOH, and ATP was determined by the development of NADH using an enzymatic assay based on coupling of hexokinase and glucose-6-phoshate dehydrogenase (Sigma).

RESULTS
Genome-wide Transcription Analysis of LiCl-pulsed Galactose-grown S. cerevisiae-The prototrophic wild type strain CEN.PK113-7D was grown on a minimal media. This growth condition was chosen to avoid addition of various amino acids that could affect transcription of genes involved in the amino acid metabolism. At a residual galactose concentration of 15 g/liter the exponential growing culture was pulsed with LiCl to a concentration of 10 mM, because lithium in this concentration range has been shown to strongly affect galactose grown cells (4). We harvested cells for genome-wide transcription analysis immediately before (denoted 0 min in the text) and 20, 40, 60, and 140 min after addition of LiCl. Three-cell samples were harvested at 0 min to get a precise estimation of the gene expression at this time point and assess the variance of each gene. We then defined genes that had changed expression from a set of arbitrary selected criteria described under "Experimental Procedures." The addition of lithium had an immediate and drastic effect on the specific growth rate, substrate uptake, metabolite production, and gene expression. Of the 5723 ORFs measured as being expressed in at least one of the time points, 1390 ORFs were identified as having a changed mRNA level after addition of LiCl to the culture. A cluster analysis identified two main clusters (see supplementary material available at www.cpb.dtu.dk/data/licl.html for the full data set) one with 664 down-regulated genes and one consisting of 725 up-regulated genes during one or several of the four sampled time points according to the definitions used.
LiCl-pulse Changed the Metabolism from Respiro-fermentative to Respiratory-Addition of LiCl led to a halt in the formation of ethanol (Fig. 1), which could be explained by a lower glycolytic flux. However, the presence of lithium did not have any detectable repressing effect on genes encoding enzymes in the central carbon metabolism, but it led to increased expression of several genes in glycolysis/gluconeogenesis and the pentose phosphate pathway (Fig. 2). Concomitantly, a dramatic drop in the specific production rate of CO 2 (see Fig. 3) was observed, but the consumption of O 2 and transcription of genes in oxidative phosphorylation were not negatively affected by lithium. The CO 2 yield on galactose during the first 6 h of the lithium pulse was unchanged, which is explained by a decrease in the specific galactose uptake rate (see Table I). The ratio between the production rate of CO 2 and consumption rate of O 2 was 4.09 before the pulse, and it diminished to 1.2 after lithium addition showing that cells changed from respiro-fermentative growth to respiratory growth (Table I).
Glycerol and Acetate Accumulated as a Response to Addition of Lithium-When S. cerevisiae is exposed to osmotic stress glycerol accumulation is initiated by an up-regulation of the genes encoding the enzymes catalyzing synthesis of glycerol (28,29). We observed that the glycerol synthesis genes (GPD1, GPD2, and HOR2) are up-regulated 2-to 5-fold in response to lithium, which is reflected in an increased yield of glycerol on galactose (Table I). This indication of a lithium-induced stress response was further supported by the observed up-regulation of the genes, which encode two of the cytosolic acetate-synthesizing enzymes Ald2 and Ald3 (see Fig. 2). ALD2 and ALD3 have active stress response elements (STREs) in their promoters and are both induced in response to osmotic stress and diauxic shift (30). In general, many of the genes, which encode enzymes that catalyze the conversion of cytosolic pyruvate into cytosolic acetyl-CoA and are involved in transport of acetyl-CoA over the mitochondrial membrane, were up-regulated as depicted in Fig. 2. During the first 6 h after addition of LiCl, the percentage of carbon directed toward glycerol and acetate increased significantly. After 1 h 14% of the used galactose was converted into acetate, whereas the yield of glycerol had increased 300% and continued to increase throughout the experiment (see Table I and Fig. 1). The large accumulation of acetate decreased 6 h after the LiCl pulse and was coincident with a halt in biomass formation (Table I).
Genes in the Pathways Leading to Glycogen and Trehalose Are Up-regulated-Most genes encoding enzymes for trehalose and glycogen metabolism were found to be up-regulated in response to the lithium pulse (see Fig. 2). Up-regulation of most of these genes is normally seen in response to different stress conditions (31)(32)(33) and leads to rapid accumulation of glycogen and trehalose (32). A study of the dynamic response of biomass formation and galactose uptake showed that the initial decrease in specific galactose uptake was not accompanied by a decrease in the specific growth rate measured from the cell dry weight (DW) ( Table I). This resulted in a dramatic increase in biomass yield, which almost doubled, so that ϳ47% of the carbon taken up from galactose was converted to biomass in the first hour after addition of lithium (see Table I). About 1 h after addition of the lithium pulse the specific growth rate decreased to a value of 0.025 h Ϫ1 , but around 25% of the galactose was still converted into biomass up to 6 h after addition of lithium. We hypothesized that the initially unchanged specific growth rate and the high yield of biomass on galactose was due to accumulation of trehalose and glycogen, whereas synthesis of proteins and other biomass constituent decreased as a consequence of the lithium pulse. Therefore, it was subsequently tested if glycogen accumulates in galactose-grown cells in response to lithium addition, and as expected we found that the cellular content of glycogen increased significantly. Treatment with just 6 mM LiCl led to a glycogen content of 18.5 Ϯ 2.4 mg/g DW compared with only 9.0 Ϯ 0.1 mg/g DW in non-treated cells (given as mean Ϯ S.D. of triplicate determinations). Apparently, the inhibition of phosphoglucomutase and the resulting reduction of flux down the glycolysis result in redirection of Glu-1-P to glycogen and trehalose. This is consistent with the FIG. 3. Lithium changes metabolism from respiro-fermentative to respiratory. Specific production rate of CO 2 (E) and consumption rate of O 2 (q) were followed before and after addition of 10 mM LiCl to the exponentially galactose-grown culture. Addition of LiCl is denoted as time zero. up-regulation of the genes involved in the synthesis of glycogen and trehalose. Lithium Down-regulates Genes in Protein Biosynthesis-Lithium has previously been shown to inhibit protein synthesis at the initiation step in galactose grown cells (6). Among the down-regulated genes we observed an over-representation of genes encoding proteins in protein biosynthesis such as ribosomal proteins as well as proteins in ribosome biogenesis and TABLE II Over-represented functional clusters of down-regulated genes The table shows gene groups with common biological processes found to be over-represented (p Ͻ 0.01) according to the SGD Gene Ontology Term Finder (genome-www4.stanford.edu/cgi-bin/SGD/GO/goTermFinder) among the transcripts with down-regulated mRNA levels. Some genes may be represented in several groups. The hierarchies of the groups are illustrated: the main over-represented groups are in bold with the overrepresented subcategories indented. Only over-represented main and subgroups are shown. assembly, RNA metabolism, transcription, nucleotide metabolism, and nucleobase metabolism (Table II). Hence, we found that protein synthesis is also repressed at the transcriptional level as proteins in purine and pyrimidine metabolism and structural proteins in RNA and DNA polymerases are downregulated (see Fig. 2 and Table II). This general down-regulation of the transcriptional and translational apparatus at the level of gene expression was very pronounced showing a clear inhibition of cell growth. Besides these, 280 of the down-regulated ORFs (corresponding to 42%) have unknown molecular function according to SGD (available at genome-www.stanford.edu/Saccharomyces/). In comparison more than half (57%) of the ORFs found to be up-regulated did not have a known molecular function (Table III).
Proteomic Response of S. cerevisiae to Lithium in Galactose Medium-A two-dimensional gel analysis was performed to study the changes of protein expression and to control if changes in mRNA levels are transmitted to the protein level. From the more than 450 proteins that have been identified in the yeast proteome, 27 were found to be down-regulated more than 2-fold when cells were treated with lithium (Table IV), whereas 21 proteins were up-regulated more than 1.5-fold (Table V). From this a total of 26% of the down-regulated and 48% of the up-regulated proteins were also identified as changed on the mRNA level. This discrepancy between the microarray and proteome data could be due to the different thresholds that were necessary to use in selection of altered genes or expressed proteins in the experiments. It might also be an indication of post-transcriptional regulation of protein expression, differences between the two S. cerevisiae strains used. However, both strains have the same LiCl sensitivity and phenotype upon LiCl addition. The trend that protein synthesis, purine and pyrimidine metabolism, as well as methionine metabolism are inhibited, whereas proteins involved in carbohydrate metabolism and stress responses increased, is clearly shown on both the mRNA and protein levels (Tables IV and V). In this regard the proteome and transcription data correlate very well.
Lithium Induces STRE-regulated Genes-Analysis of the promoter regions upstream the coding gene sequences, which have similar regulation, can reveal motifs that are common among the analyzed group of genes. Such promoter elements might be binding sites for regulatory transcriptions factors that are able to repress or activate transcription. Genes containing the so-called stress related element (STRE) in their promoter regions were over-represented among the genes with up-regulated mRNA levels. 29% of the 725 up-regulated genes contained at least two STRE in comparison to only 16% of all ORFs suggesting that a significantly part of the transcriptional response was mediated by STRE up-regulation. This observation fits with the many stress response genes, which were overrepresented in the up-regulated ORFs (see Tables III and V). We identified that STREs were especially over-represented in one large cluster of 418 ORFs, which were up-regulated at 20 min, by dividing the up-regulated genes into smaller groups (not shown) using cluster analysis. In another large cluster of 189 members, up-regulation in response to the lithium pulse was somewhat delayed and only observed in samples harvested later than 40 min or more after addition of lithium. In this group of genes STREs were not over-represented, showing that the STRE-regulated genes responded immediately to the presence of lithium. Among the down-regulated genes we identified three over-represented promoter motifs with unknown function: CTCATC, AAATTT, and CGATGA, which are present in at least two copies of 18, 36, and 14%, respectively, of the 664 down-regulated ORFs compared with 10, 23, and 8%, respectively, in all ORFs. We also found a fourth motif GAAAAAT to be slightly over-represented with 21% in the down-regulated cluster against 13% of all ORFs. GAAAAAT has similarities to the Hxk2/Med8 binding motif (C/A)(G/A)(G/A)AAAAT, which is involved in response to the presence of glucose (34).
Dynamic Response of Intracellular Metabolites to Lithium-It has previously been shown that addition of LiCl to galactose grown cells results in accumulation of Glu-1-P and Gal-1-P (4). To study the dynamic alterations of these metabolite concentrations, as well as other intracellular metabolite concentrations they were measured at 0, 20, 60, and 140 min after addition of LiCl. As expected Gal-1-P accumulated in response to the lithium pulse, but surprisingly a significant increase of Glu-1-P was not observed (see Fig. 4). For Glu-6-P, fructose TABLE III Over-represented functional clusters of up-regulated genes The table shows gene groups with common biological processes found to be over-represented (p Ͻ 0.01) according to the SGD Gene Ontology Term Finder (genome-www4.stanford.edu/cgi-bin/SGD/GO/goTermFinder) among the transcripts with up-regulated mRNA levels. Some genes may be represented in several groups. The hierarchies of the groups are illustrated: the main over-represented groups are in bold with the overrepresented subcategories indented. Only over-represented main and subgroups are shown. 6-phosphate (Fru-6-P), and mannose 6-phosphate (Man-6-P), which are downstream of phosphoglucomutase, the expected decrease was first seen after 60 min indicating that the immediate decrease in the flux through glycolysis was not due to a lower concentrations of these metabolites.
Galactitol, which is one of the suspected toxic compounds formed from galactose in connection with galactosemia (35,36), initially increased 2-fold from 27 to 48 mol/g DW 20 min after addition of LiCl but returned to the concentration found before the pulse after 60 min (results not shown) indicating that formation of galactitol had no significant effect during the first hours after addition of lithium. This conclusion was also supported by measurements of the extracellular concentration of galactitol, which changed from 0.29 to only 0.33 mM after the first 140 min after addition of lithium. However, from 6 to 22 h a significant amount of galactitol was formed accounting for 3% of all carbon according to measurements of extracellular galactitol (1.58 mM galactitol 22 h after the LiCl pulse).
We have also measured if the intracellular content of ATP is diminished in cells treated with lithium, because this could be a metabolite sensed by the signaling pathways. However, our results show that the ATP content is not changed by lithium treatment (results not shown). DISCUSSION In this study, we show that cells grown on galactose and treated with therapeutically relevant concentrations of lithium suffer a dramatic drop in flux through glycolysis resulting in an immediate shift from fermentation to respiration and an increase in the glycogen content. Thus, the effect of lithium has similarities to phenomena of nutrient limitation, which can be observed at the diauxic shift, where glucose becomes limited (37) or as a result of rapamycin treatment (38).
Nutrient limitation also leads to repression of protein biosynthesis both at the transcriptional and translational levels, whereas transcription of STRE-regulated genes are induced (38). Similarly, we found that genes involved in protein biosynthesis had down-regulated mRNA levels. In addition, repression at the post-transcriptional level is seen in the uncorrelated mRNA and protein levels for some genes. These results fit a previous study showing that translation initiation is inhibited by lithium (6). Furthermore, genes containing at least two STRE elements were found to be up-regulated already 20 min after addition of LiCl, indicating a global stress response, which was further supported by the increased accumulation of acetate, glycerol, and glycogen, and the up-regulation of genes related to these pathways. This finding, combined with the observation that Sit4 phosphatase is a suppressor of lithium toxicity (3), could indicate that the observed repression of the flux through glycolysis by lithium involves the TOR pathway. However, rapamycin treatment, diauxic shift, and adaptation to low glucose lead to a down-regulation of glycolytic genes (13), whereas lithium treatment leads to an up-regulation. More- over, nitrogen catabolite-repressed (NCR) genes, which normally are up-regulated in a TOR response to nutrient limitation (38), were not found to be over-represented among the up-regulated transcripts. Only 11 genes (CPS1, URA10, ZRT1, YHR029C, GDH3, GDH2, UGA1, NPR1, DIP5, YBR139W, and YDR380W), among 56 genes reported to be NCR genes (39 -41), were found to be up-regulated, whereas 4 of these NCR genes were found to be down-regulated. TOR is also known to control the synthesis of rRNA and tRNA in a positive fashion (42). Hence TOR may play a positive role in the regulation of rRNA and tRNA genes in response to lithium. We did not include expression of rRNA and tRNA in the analysis due to their lack of poly-A tails, but it was found that the genes encoding the structural constituents of RNA polymerases I and III, which are responsible for transcription of rRNA and tRNA, were down-regulated.
The increased production of glycerol, which was observed in response to lithium, led us to consider a possible high osmolarity glycerol response (HOG response). However, we do not consider this to be likely, because the HOG-independent ALD2 and ALD3 (30) were found to be up-regulated, whereas ALD6, which is up-regulated in a HOG-dependent manner in response to osmotic stress (43), was down-regulated 1.9-fold on the mRNA level and up to 4.2-fold on the protein level (Table IV).
We observed that the mechanism that cells use to sense lithium in galactose medium is mediated by a strong stimulation of STRE genes, an increment of glycolytic genes, and an inhibition of translation, transcription, and DNA synthesis. A question that arises from the data is whether this response could be mediated by an intracellular metabolite. Metabolite analysis showed that the initial lithium-induced responses are not due to absence of galactose, Glu-6-P, or any of the other intracellular metabolites measured, because unchanged or increased amounts of these metabolites are present at the first 20 min after addition of LiCl. Furthermore, inhibition of galactose fermentation in yeast by lithium did not change the intracellular ATP concentration either. The stress signal could also be mediated by accumulation or depletion of intracellular metabolites such as Gal-1-P. Gal-1-P has been shown to be toxic in human galactosemic cells by inhibiting UDP-hexoses phosphorylases (44) and inositol monophosphatase (45).
The observed disparity between changes in mRNA and protein levels found for some of the protein with changed expression were not surprising due to several studies, which have shown that the overall correlation between mRNA and protein expression is only weakly positive (46 -49). However, it was recently shown that protein expression is transcriptionally controlled for some entire protein pathways such as purine and pyrimidine nucleotide biosynthetic pathways and pathways for amino acids such as methionine (49). This corresponds with our findings for the purine and pyrimidine metabolism where the four proteins found to have changed expression in this part of the metabolism also were identified as changed on the transcriptional level. For proteins involved in protein biosynthesis,  such as ribosomal proteins, there are several reports describing a poor correlation for many of these genes (46, 48, and 49). But even though many of the proteins in protein synthesis, which we found to have changed expression, were not identified as having changed mRNA levels, we found a general transcriptional and translational down-regulation of genes related to protein synthesis. Based on the proteome and genome-wide transcription analyses, we have mapped the global response of galactose-grown cells to treatment with lithium in therapeutically relevant concentrations. Our studies show both a transcriptional and translation response that embraces a large part of the cellular metabolism and that results in some clear physiological effects observed for both growth and metabolite production. The general down-regulation of the machinery for cell growth may be a consequence of the diminishing amounts of free available energy supplied by catabolism due to redirection of resources toward stress defense. In addition, a substantial set of genes with unknown function were found to respond at the mRNA level, which, in combination with the clear up-regulation of stress response genes and down-regulation of growth-related genes, may help elucidating some of their functions. A more complete understanding of the response induced by addition of lithium to galactose-grown cells may enhance the understanding of the basic mechanisms underlying how a cell can rapidly control its protein expression both at the transcriptional and translational level in response to rapid down-regulation of glycolytic flux. Furthermore, the responses, which were observed, may be related to the galactose toxicity seen in GAL7-deficient yeast and in humans with galactosemia.