|Home | About | Journals | Submit | Contact Us | Français|
The PHO regulatory pathway is involved in the acquisition of phosphate (Pi) in the yeast Saccharomyces cerevisiae. When extracellular Pi concentrations are low, several genes are transcriptionally induced by this pathway, which includes the Pho4 transcriptional activator, the Pho80-Pho85 cyclin-CDK pair, and the Pho81 CDK inhibitor. In an attempt to identify all the components regulated by this system, a whole-genome DNA microarray analysis was employed, and 22 PHO-regulated genes were identified. The promoter regions of 21 of these genes contained at least one copy of a sequence that matched the Pho4 recognition site. Eight of these genes, PHM1–PHM8, had no previously defined function in phosphate metabolism. The amino acid sequences of PHM1 (YFL004w), PHM2 (YPL019c), PHM3 (YJL012c), and PHM4 (YER072w) are 32–56% identical. The phm3 and phm4 single mutants and the phm1 phm2 double mutant were each severely deficient in accumulation of inorganic polyphosphate (polyP) and Pi. The phenotype of the phm5 mutant suggests that PHM5 (YDR452w) is essential for normal catabolism of polyP in the yeast vacuole. Taken together, the results reveal important new features of a genetic system that plays a critical role in Pi acquisition and polyP metabolism in yeast.
Phosphate (Pi) is an essential nutrient for all organisms, used in the biosynthesis of diverse cellular components, including nucleic acids, proteins, lipids, and sugars. It is therefore essential for organisms to have evolved regulatory mechanisms for acquisition, storage, and release of this molecule (Torriani-Gorini et al., 1994 ).
In Saccharomyces cerevisiae, the PHO regulatory pathway regulates expression of the “PHO” genes, involved in the scavenging and specific uptake of Pi from extracellular sources (Johnston and Carlson, 1992 ; Oshima, 1997 ). The PHO regulatory system consists of at least five PHO-specific regulatory proteins, the Pho2 and Pho4 transcriptional activators, the Pho80-Pho85 cyclin-cyclin dependent protein kinase (CDK) complex, and the Pho81 CDK inhibitor (Figure (Figure1).1). Pho84 is a high-affinity Pi transporter localized on plasma membrane, which has been shown to contribute to Pi uptake from culture medium (Bun-ya et al., 1991 ). PHO84 gene expression is activated by a Pi-starvation signal mediated by the PHO regulatory system. Additionally, the PHO5 gene encodes a repressible acid phosphatase which is localized to the periplasmic space.
When the Pi concentration in the medium is low (~ 0.2 mM Pi), the Pho81 protein inhibits the Pho80-Pho85 kinase activity, which in its active state catalyzes a hyperphosphorylation of Pho4 (Schneider et al., 1994 ; Ogawa et al., 1995 ). The hypophosphorylated form of Pho4 is preferentially localized to the nucleus, where together with Pho2, it activates target gene transcription (Kaffman et al., 1998 ; Komeili and O'Shea, 1999 ). Alternatively, when the Pi concentrations are high (~ 10 mM Pi), the Pho80-Pho85 kinase phosphorylates Pho4. In addition to having a lower affinity for Pho2 and the nuclear import protein Pse1/Kap121, phosphorylated Pho4 is a preferred substrate of the nuclear export protein Msn5, resulting in extranuclear localization. Phosphorylated Pho4 is thus unable to activate target gene expression.
Besides PHO5 and PHO84, seven additional genes are known to be regulated by the PHO regulatory system; these include PHO11and PHO12 (homologs of PHO5), PHO8 (vacuolar alkaline phosphatase, Kaneko et al., 1987 ), PHO89 (Na/Pi cotransporter, Martinez and Persson, 1998 ), PHO86 (required for Pi uptake, Yompakdee et al., 1996 ), PHO81 and SPL2 (a homolog of PHO81, Flick and Thorner, 1998 ). The promoters of all nine previously recognized PHO-regulated genes have common motifs, CACGTG and/or CACGTT, as core sequences comprising the Pho4 binding site (Oshima, 1997 ). Both the regulating properties and the functions of the target genes point to the critical role played by the PHO regulatory system in Pi acquisition in yeast. Comprehensive identification and characterization of the PHO-regulated genes in the yeast genome is therefore likely to be an important step toward understanding the regulation and physiology of Pi metabolism.
DNA microarrays provide a systematic way to study the expression programs of the entire genome (DeRisi et al., 1997 ). Using DNA microarrays, we conducted an exhaustive search for yeast genes regulated by the PHO regulatory system. Several of the genes identified were of unknown function and were further characterized by gene disruption. Biochemical analysis of five of the novel PHO-regulated genes revealed them to be important in inorganic polyphosphate (polyP) metabolism.
PolyP, a linear polymer of up to hundreds of Pi residues linked by high-energy phosphoanhydride bonds, is ubiquitous in nature, having been found in all organisms examined (Kornberg, 1999 ). S. cerevisiae is known to accumulate large amounts of polyP in vacuoles, comprising 37% of the total cellular phosphate (Urech, 1978 ). The enzyme primarily responsible for polyP synthesis in Escherichia coli is polyP kinase (PPK), and polyP is hydrolyzed to Pi by exopolyphosphatase (Kornberg, 1999 ). In S. cerevisiae, an exopolyphosphatase gene, PPX1, has been identified (Wurst et al., 1995 ), but a gene for PPK has not. Moreover, no gene homologous to the bacterial PPKs has been found in the genome databases of S. cerevisiae. Thus, the identify of the enzymes that mediate polyP metabolism in yeast and the other eukaryotes has been an important unsolved problem in metabolism of a centrally important nutrient.
S. cerevisiae strains used in this study are listed in Table Table1.1. YPAD medium (Adams et al., 1997 ) was used as rich media. A Pi-depleted YPAD medium (YPAD-Pi) was prepared as described (Kaneko et al., 1982 ), and a high-Pi YPAD (YPAD+Pi) medium was prepared by addition of sodium phosphate (10 mM, pH 5.8) to the YPAD-Pi medium. High-Pi, low-Pi, and Pi-free synthetic media were prepared as described (Yoshida et al., 1989 ). YPAD medium buffered at pH 7.5 supplemented with 50 mM CaCl2 was prepared as described (Zhang et al., 1998 ).
For comparison of the gene expression pattern between the low- and high-Pi conditions, yeast strain NBW7 or DBY7286 was grown in 25 ml of the YPAD+Pi medium overnight, the cells were collected, washed with the YPAD-Pi medium two times, and inoculated into 500 ml of the YPAD+Pi and -Pi media to an OD600 of 0.05. The two cultures were shaken at 30°C to an OD600 of 0.5, the grown cells were harvested, and then frozen by immersion in liquid nitrogen. The cells were stored in −80°C until RNA was prepared. For comparison of the gene expression pattern between the mutants and wild-type, yeast strains NBD82–1 (PHO4c-1), NBD80–1 (pho80Δ), NBD85A-1 (pho85Δ), NOF1 (PHO81c-1), and NBW7 (wild-type) were precultivated in 25 ml of YPAD overnight, the grown cells were collected washed with YPAD, and inoculated into 500 ml of YPAD to an OD600 of 0.1. The remainder of the procedure was the same as above.
The preparation of the yeast ORF DNA microarray, RNA preparation from yeast cells, probe preparation with fluorescence dye, hybridization, scanning of the hybridized array, and data processing were performed as described previously (Chu et al., 1998 ; Spellman et al., 1998 ). To compare gene expression patterns between the low- and high-Pi conditions, Cy3- or Cy5-labeled cDNA probes were prepared from the high- or low-Pi samples, respectively, by reverse transcription in the presence of Cy3- or Cy5-dUTP (Amersham Pharmacia Biotech., Piscataway, NJ), as previously described (DeRisi et al., 1997 ). To compare the gene expression patterns between mutant and wild-type cells, cDNA probes templated by mRNAs prepared from wild-type and mutant cells were labeled with Cy3-dUTP and Cy5-dUTP, respectively. The original array data are available on our web page on <http://cmgm.stanford.edu/pbrown/phosphate/>.
To make disruption mutants of PHM1, PHM2, PHM3, PHM4, PHM5, PHM6, and CTF19, PCR-mediated gene disruptions were performed as described (Sakumoto et al., 1999 ). pCgHIS3, pCgTRP1, pCgLEU2 harboring the Candida glabrata HIS3, TRP1, and LEU2 fragments on pUC19, respectively, were obtained from Dr. Y. Mukai (Osaka University), and used as templates to generate the PCR fragments for the gene disruptions. Deletion regions in the target genes were as follows: PHM1, +1 to +2387; PHM2, +1 to +2508; PHM3, +53 to +926; PHM4, +44 to +347; PHM5, +1 to +2025; CTF19, +1 to +1130; PHM6, +1 to +315 (nucleotide positions relative to A of initiation codon of ATG in each ORF). In each case, the deleted sequence was replaced with the marker genes indicated in Table Table1.1. The disruption junctions were verified by colony-PCR (Adams et al., 1997 ).
Polyphosphate overplus (Harold, 1966 ) culture was performed as follows: The yeast strains indicated were grown in the YPAD-Pi media overnight, the grown cells were collected, washed with water, and resuspended in 10 ml of the synthetic Pi-free media to an OD600 of 0.5. The cultures were shaken for 2 h. at 30°C, then potassium phosphate (pH 5.8) was added (10 mM final concentration). After 2 more hours of cultivation, the cells were harvested, and then frozen by immersion in liquid nitrogen. For the polyP overplus at low pH, yeast strains were grown in the YPAD-Pi media overnight, the grown cells were collected, washed with water, and resuspended in 10 ml of the YPAD media supplemented with 10 mM (final) KH2PO4 and/or 10 mM (final) sodium acetate buffer (pH 4.0) to an OD600 of 0.5. The cultures were shaken for 2 h at 30°C, then the cells were harvested, and frozen by immersion in liquid nitrogen.
Cells in 150 μl of extraction buffer (50 mM Tris-HCl [pH 7.4], 100 mM KCl, 1 mM EDTA) were mixed in a vortex mixer with 150 mg of acid-washed glass beads (0.5-mm diameter) for 2 min at 4°C, and microcentrifuged at 14,000 rpm for 10 min at 4°C. The aqueous phase was extracted by phenol:chloroform followed by chloroform and ether extractions. After removing ether in the samples by evaporation, their A260 values were measured to calculate their total RNA concentrations.
PolyP analysis by PAGE was performed as described in Wurst et al. (1995) with the following modifications. Samples containing the indicated amounts of RNA were loaded with 7 μl of loading dye solution (1x TBE buffer [Sambrook et al., 1989 ], 10% sucrose, 0.05% bromophenol blue) on a 20% polyacrylamide gel (270-mm height x 165-mm wide x 0.5-mm thick) with 1x TBE buffer. Ten μg of sodium phosphate glass type 5, 15, or 35 (Sigma, St. Louis, MO) were loaded as a polyP size markers. Radioactive [γ-32P]ATP was also used as an additional size marker. The electrophoresis was run at 20 V/cm for ~ 1.5 or 3 h, and the ATP-loaded gel piece was analyzed by a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The nonradioactive piece of gel was soaked into 10% acetate, 10% methanol for 15 min, stained with the staining solution (0.5% Toluidine Blue O [Sigma], 25% methanol, 5% glycerol, 5% acetate) for 15 min, and destained with destaining solution (same as the staining without Toludine Blue O) for 10 min several times. ATP migrates at a position corresponding to polyP of between seven and eight Pi residues long. PolyP bands of sodium phosphate glass type P35 with 55 and 65 chain length on PAGE were extracted and used as markers to estimate longer chain lengths of polyP.
PolyP was assayed by the nonradioactive method as described (Ault-Richéet al., 1998 ), without the Glassmilk purification steps. Since the major population of yeast polyP is <60 Pi residues in length (Figure (Figure4B),4B), they are not able to be effectively trapped by Glassmilk. Concentrations of polyP in the samples were shown as mol of Pi residues per mg of total RNA.
pPHM2-GFP was constructed by insertion of the PCR fragment corresponding to nucleotide positions from −393 to +2505 of PHM2 (relative to A of initiation codon of ATG) into an EcoRI-HindIII gap of pTS395, a GFP-expressing yeast vector provided by Dr. D. Botstein. Yeast transformant NBM-4L19H1[pPHM2-GFP] was grown in synthetic low-Pi medium lacking uracil for >3 h at 30°C. Fluorescent dye, FM4–64 (Molecular Probes, Eugene, OR) , which has been reported to stain vacuolar membrane (Vida and Emr, 1995 ), was added to 30 μM, and incubated for an additional 15 min. Ten-microliter samples were examined by fluorescent microscope Axioplan2 (Carl Zeiss, Thomwood, NY) following the procedure of Vida and Emr (1995) and Adams et al. (1997) .
A Pi uptake assay was performed following the procedure of Bun-ya et al. (1991) . Yeast strains were grown in the YPAD-Pi media overnight, the grown cells were collected, washed with water, and resuspended in 50 ml of synthetic Pi-free media to an OD660 of 0.1. The cultures were shaken for 2 h at 30°C, and their OD660 values were remeasured. Potassium phosphate (0.1 mM, pH 5.8) and radioactive phosphate (32PO4; 1 μCi/ml) were added to cultures, and shaken. Five-milliliter samples were taken at indicated intervals and immediately filtered though a HA filter (3.5-mm diameter, Millipore). The cells trapped on the filter were washed with 10 ml of synthetic Pi-free media. The radioactivity trapped on the membrane filters was quantitated by a liquid scintillation counting.
In order to identify all of the genes regulated by the PHO regulatory pathway, we used DNA microarrays fabricated as described previously (DeRisi et al., 1997 , Chu et al., 1998 ; DeRisi et al., 1997 , Chu et al., 1998 ; Spellman et al., 1998 ). A cDNA probe prepared from poly(A)+ RNA isolated from a wild-type yeast strain, NBW7, cultivated in low-Pi media (YPAD-Pi), was labeled with Cy5 fluorescent dye, while a cDNA probe prepared from the same strain cultivated in high-Pi media (YPAD+Pi) was labeled with Cy3 fluorescent dye. The Cy5- and Cy3-labeled cDNA probes were mixed and hybridized to the microarray. As expected, the transcript levels of both PHO5 and PHO84 were elevated in low-Pi media in two independent experiments (Figure (Figure2A).2A). In a repeat experiment using strain DBY7286, derived from S288C, (the standard S. cerevisiae strain for the genome database), most of the highly derepressed genes we found in NBW7 were consistent with those identified in DBY7286 (Figure (Figure2)2) suggesting similar PHO regulation in the two strains.
To evaluate if these gene expression changes were dependent on the known PHO regulatory factors, further DNA microarray analyses were performed comparing the parental strain NBW7 to four strains carrying mutations in components of the PHO regulatory system (Figure (Figure1):1): 1) a gain-of-function mutation of the PHO4 gene encoding transcriptional activator named PHO4c-1 (Ogawa and Oshima, 1990 ); 2) a pho80 deletion mutant; 3) a pho85 deletion mutant; and 4) PHO81c-1 a gain-of-function mutation in the PHO81 gene, encoding the CDK inhibitor for Pho80-Pho85 (Ogawa et al., 1995 ). In all cases PHO5 was expressed (Figure (Figure2A)2A) irrespective of Pi concentration (YPAD media). A total of eight DNA microarray experiments were performed, including two independent repeated experiments. Cluster analysis of the combined data is shown in Figure Figure2B.2B. Transcript levels of >80 genes changed significantly in response to one or more of the conditions we total.
All nine genes that have previously been reported to be PHO-regulated (PHO5, PHO11, PHO12, PHO8, PHO84, PHO89, PHO86 PHO81 and SPL2) were successfully identified (Figure (Figure2B). 2B). PHO5, PHO11, PHO12, PHO84, PHO89, and SPL2 had high differential expression ratios (more than fivefold) whereas PHO8 and PHO86 had lower differential ratios (between two and fivefold), consistent with previous results from Northern analysis (Kaneko et al., 1987 ; Bun-ya et al., 1991 ; Yompakdee et al., 1996 ). The nine known PHO-regulated genes can be categorized into three groups according to their function (Table (Table2).2). A third member of the PHO81 family, YPL110c, also seems to be regulated by the PHO regulatory pathway (Figure (Figure2B)2B) in a manner similar to PHO81. Twelve additional genes showed clear evidence of PHO regulation in our experiments (Figure (Figure2B2B and Table Table2).2). Three of these genes (YPL019c, YJL012c and YER072w) had levels of differential expression similar to PHO5 and PHO84, in nearly all of the eight experiments. Interestingly, the putative proteins encoded by these three genes share homology with YFL004w as described later (Figure (Figure3A).3A). The expression of YFL004w was also regulated by the PHO pathway, with differential expression levels similar to PHO8 and PHO86 (Figure (Figure2B). 2B). Because of their homology and high differential expression ratios, we speculated that these four genes, which we have named PHM1 through PHM4 (phosphate metabolism genes), might have important functions in Pi metabolism.
Among the other newly identified PHO-regulated genes listed in Table Table2, 2, the functions of three had been previously described: HOR2/GPP2 (YER062c), CTF19 (YPL018w), and HIS1 (YER055c). The HOR2/GPP2 gene encodes glycerol phosphatase (Norbeck et al., 1996 ) which hydrolyzes the phosphate bond of glycerolphosphate, releasing free Pi. The His1 enzyme (ATP phosphoribosyltransferase)-catalyzed reaction can produce pyrophosphate from ATP and phosphoribosylpyrophosphate. While both HIS1 and HOR2 encode enzymes that catalyze reactions that release Pi, the special significance of these reactions for the cell's Pi economy is not obvious. The Ctf19 protein has previously been identified as a subunit of the centromere-binding complex (Ortiz et al., 1999 ), and has no recognized role in Pi metabolism or regulation. A possible explanation for the observed differential expression could be that CTF19 shares promoter sequences with the “head-to-head” oriented adjacent gene YPL019c/PHM2, which was highly differentially expressed under PHO regulation. More plausibly, it may have an additional, unrecognized role in Pi metabolism.
Two “core” Pho4 binding sequences, CACGTG and CACGTT have previously been described. At least one of these sequences is found within 500 bp upstream of the coding sequence of 21 of the 22 putative PHO-regulated genes listed in Table Table2,2, a fraction significantly higher than the 20% of all yeast ORFs that have one of these sequences within 500 bp upstream (1287 ORFs out of 6282). Of the 14 genes with more than two putative Pho4 binding sites, 9 had the high derepression ratios described above. Higher levels of differential expression showed some correlation to the presence of both of the two types Pho4 binding sites. Of the 22 putative PHO-regulated genes, only YER038c had no consensus Pho4-binding sites in its promoter region. The 3′ end of the predicted coding region of the adjacent gene, YER037w, is located only 4 bp from the YER038c ORF, which had a very similar observed expression pattern (Figure (Figure2). 2). It is therefore likely that the apparent PHO regulation of YER038c reflects hybridization to the 3′ untranslated region of YER037w.
The predicted coding regions Phm1 and Phm2 are similar in length (828 and 835 amino acid residues for Phm1 and Phm2, respectively) and 58% identical in amino acid sequence. The predicted coding region of Phm3 is smaller (648 amino acids) and has 33% identity to the N-terminal region of Phm1 (Figure (Figure3A).3A). Phm4 is predicted to encode a protein of 129 amino acids, with 32% identity to the C-terminal region of Phm1, which contains three putative transmembrane spanning regions. Phm3 and Phm4 share no homology. Interestingly, the N-terminal regions (amino acids 1–135, Figure Figure3A)3A) of Phm1, Phm2, and Phm3 have significant similarity to the N-terminal regions of Pho81, and its homolog Ypl110c (Figure (Figure3B).3B). Furthermore, this similarity is shared among a total of nine S. cerevisiae ORFs, including the putative phosphate transporter Pho87 (Bun-ya et al., 1996 ), two proteins homologous to Pho87, Yjl198w and Ynr013, and Syg1 (Spain et al., 1995 ), a multi-copy suppressor for a GPA1 deletion. The similarities were confined to the N-terminal region (<300 amino acids) in all of the nine proteins. Homologous sequences can also be found in the databases of Schizosaccharomyces pombe, Caenorhabditis elegans, Drosophila melanogaster, Arabidopsis thaliana, mouse, and human (Battini et al., 1999 ).
To address the function of the PHM1, PHM2, PHM3, and PHM4 genes, strains carrying deletions of each of these genes were constructed (see MATERIAL AND METHODS). All five mutant strains, including a phm1Δ phm2Δ double mutant, were viable in rich media (YPAD), and no significant growth defects were observed in low-Pi synthetic media. All five were able to produce acid phosphatases in response to low Pi. When we analyzed polyP accumulation, however, the mutants showed striking phenotypes.
It is known that S. cerevisiae accumulates a large amounts of polyP in vacuoles under conditions of high Pi preceded by a period of Pi starvation. This is referred to as the “polyphosphate overplus” phenomenon (Harold, 1966 ). PolyP chains in extracts from yeast were analyzed by PAGE followed by staining with a Toluidine Blue dye, which stains polyP as well as nucleic acids. The extract from wild-type cells, NBW7, after the polyP overplus treatment (see MATERIAL AND METHODS) resulted in two distinct populations of stained molecules as is shown in Figure Figure4A.4A. The upper population represents RNA, whereas the lower “ladder-like” population is polyP, confirmed by pretreatment with either RNase A or exopolyphosphatase (our unpublished results). The chain lengths of the polyP bands were determined by comparing them with polyP marker ladders run in a nearby lane. The distribution of the polyP chain length in the NBW7 strain was broad (~100 Pi residues) with a median length of ~ 60 Pi residues (Figure (Figure4B).4B). The cellular concentration was measured at 19.4 μmol (1.55 mg) of polyP/mg total RNA by an enzymatic assay (Ault-Richéet al., 1998 ) (Table (Table3). 3).
The effects of PHM1-PHM4 on polyP accumulation following the polyP overplus treatment were analyzed by gel electrophoresis (Figure (Figure4)4) and enzymatic assay (Table (Table3).3). Total polyP was slightly reduced in the phm1Δ mutant. The phm2Δ mutant had significantly reduced total polyP (14% of wild-type), and the median size of the polyP molecules was significantly smaller than that in the wild-type. PolyP was undetectable in the phm1Δ phm2Δ double mutant. This clearly demonstrates that the PHM1 and PHM2 genes have redundant functions in polyP accumulation. Both the phm3Δ and phm4Δ mutants lacked detectable polyP. These results suggest that Phm1, 2, 3 and 4 proteins are involved in the accumulation of polyP and that the products of Phm3, Phm4 and either Phm1 or Phm2 are required for polyP accumulation.
More than 90% of total polyP in yeast is localized to the vacuole (Urech, 1978 ). Since the Phm1–Phm4 proteins are involved in polyP synthesis, we speculated that the Phm1–Phm4 proteins were also localized to the vacuole. To address this question, a PHM2-GFP fusion gene, which encodes the full PHM2 ORF fused to the GFP ORF, was constructed and expressed in wild-type and phm1Δ phm2Δ double mutant strains. The fusion protein was active; PHM2-GFP was able to complement the polyP deficient phenotype of the phm1Δ phm2Δ mutant (our unpublished results). Phm2-GFP indeed appeared to be localized to the vacuoles (Figure (Figure5)5) by colocalization with the vacuolar membrane marker FM4–64 (Vida and Emr, 1995 ).
While this study was in progress, Cohen et al. (1999) reported on a gene family involved in a vacuolar transporter chaperon, VTC1, VTC2, VTC3, and VTC4 in S. cerevisiae. These four genes are identical to PHM4, PHM1, PHM2, and PHM3, respectively. In their report, the Phm4/Vtc1 protein was originally found in the same fraction as a Vma10 protein, a subunit of vacuolar H+-ATPase (v-ATPase) localized on the vacuolar membrane. Furthermore, Phm3/Vtc4, which does not have a putative transmembrane domain, was found in the membrane fraction only in the presence of Phm4/Vtc1 and either Phm1/Vtc2 or Phm2/Vtc3. These finding support our hypothesis that the Phm1 through 4 proteins form a complex on the vacuolar membrane. In addition, an S. pombe protein, Nrf1, with an amino acid sequence 78% identical to that of Phm4/Vtc1, was recently reported as a vacuolar membrane protein (Murray and Johnson, 2000 ).
Previous work has shown that polyP accumulation is dependent upon v-ATPase activity (Wurst et al., 1996 ). A vma4 mutant strain, in which v-ATPase activity is completely deficient (Zhang et al., 1998 ), was reproducibly found to have no accumulation of polyP under the polyP overplus conditions (Figure (Figure6A).6A). Cohen et al., (1999) reported that the v-ATPase activity in a phm4Δ/vtc1Δ mutant was 10–30% of that in wild-type cells. Cells with mutations in v-ATPase are characteristically deficient in respiration and sensitive to media containing 60 mM CaCl2 at pH 7.5 (Zhang et al., 1998 ). To address the possibility that the Phm1 through 4 proteins could be necessary for v-ATPase activity, and thus indirectly for the accumulation of polyP, the phm disruptants were tested for growth on nonfermentable carbon sources and in the presence of CaCl2. Growth of the phm1Δ phm2Δ, phm3Δ, and phm4Δ mutants in 60 mM CaCl2 was similar to that of the wild-type strain (Figure (Figure7),7), whereas the vma4 mutant did not grow. Moreover, phm1Δ phm2Δ, phm3Δ, and phm4Δ mutants were able to grow on rich media with ethanol or glycerol as a sole carbon source, while the vma4 mutant was not (our unpublished results). These data suggest that the phm mutants retain v-ATPase activity.
It is known that incubating yeast in low pH media can, in part, reverse the phenotype seen with v-ATPase mutants by lowering the intravacuolar pH via fluid-phase endocytosis (Nelson and Nelson, 1990 ). A low but significant level of polyP accumulation was detected when the vma4 mutant was incubated at low pH (Figure (Figure6B).6B). In contrast, the phm1Δ phm2Δ, phm3Δ and phm4Δ mutants failed to accumulate detectable polyP with the low pH incubation (Figure (Figure6B).6B). These results strongly suggest that the PHM1-PHM4 gene products are directly involved in polyP accumulation and that v-ATPase activity is not strictly essential for polyP synthesis.
Phm1 through 4 proteins clearly play a role in polyP synthesis. The apparently paradoxical increase in the cell's ability to convert Pi into polyP in response to Pi starvation might represent a strategy for accumulating and holding precious Pi. We therefore measured Pi uptake in the phm1-phm4 mutants (Figure (Figure8).8). After cultivation in Pi-free synthetic media for 2h, NBW7 wild-type cells displayed a linear (nonsaturable) Pi uptake for up to 25 min when incubated in 0.1 mM Pi (twofold less concentration of synthetic low-Pi medium). In contrast, the pho84Δ mutant, which is deficient in H+/Pi cotransporter activity, showed very low uptake, as reported previously (Bun-ya et al., 1991 ). The phm1, 2, 3, and 4 mutants each had unique saturable Pi-uptake profiles. The phm1Δ single mutant had an uptake profile similar to wild-type. The phm2Δ single mutant showed a rate of uptake similar to wild-type from 0 to 8 min, but after 8 min uptake was minimal. The Pi-uptake profiles of the phm1Δ phm2Δ double mutant, the phm3Δ, and the phm4Δ mutants were similar to wild-type for the first 5 min, after which uptake appeared to cease. This cessation of uptake in the phm mutants correlated well with the overall ability of each mutant to accumulate polyP. These data suggest that polyP accumulation is required, presumably as a sink, to sustain a high rate of Pi uptake.
The previously uncharacterized ORFs YDR452w, YDR281c, YOL084w, and YER037w were also found to be PHO-regulated. They are therefore named PHM5 through PHM8, respectively. YDR281c/PHM6 and YER037w/PHM8 were previously noted by Gray et al. (1998) to be induced by treatment with drugs that inhibit Cdc28 and Pho85 kinases. Among those four genes, Phm5 has significant homology to human and C. elegans acid sphingomyelinases (Figure (Figure3C)3C) which function in hydrolyzing phosphodiester bonds in sphingomyelin. When phm5Δ, ctf19Δ, and phm6Δ single disruptants were constructed and tested for polyP accumulation levels, we observed that the phm5Δ mutant had a unique distribution of longer chain length polyP when compared with wild-type (Figure (Figure4A).4A). PolyP levels and size in the ctf19Δ and phm6Δ mutants were indistinguishable from wild-type. The average polyP chain length in the phm5Δ mutant was over 150 Pi residues (Figure (Figure4B).4B). By enzymatic measurement, the total amount of polyP in the phm5Δ mutant was similar to that of wild-type (Table (Table3),3), indicating that polyP synthesis activity in phm5Δ was unaffected. These data suggest that the Phm5 protein is associated with, or is a polyphosphatase.
The deduced amino acid sequence of Phm5 (Figure (Figure3C)3C) includes a putative N-terminal membrane spanning domain, similar in size (27 amino acids), position and flanking amino acid sequences (rich in serine and positive charged residues) to that of the Pho8 protein, a PHO-regulated vacuolar alkaline phosphatase, which is anchored to the vacuolar membrane by a transmembrane domain at its N-terminus (Klionsky and Emr, 1989 ).
Wurst et al. (1995) reported that a mutant deficient in three vacuolar proteinases: proteinase A, proteinase B, and carboxypeptidase Y, has a longer polyP chain length distribution than the wild-type. This phenotype was reproducibly observed in our PAGE assay (Figure (Figure6A).6A). Furthermore, the phenotype was observed in a pep4 single mutant, which has a defect only in Proteinase A, which is required for maturation of several vacuolar enzymes, including Pho8. Together with Phm5′s similarity to sphingomyelinases, these data suggest that PHM5 encodes a polyphosphatase, which is matured by the vacuolar proteinase. Kumble and Kornberg (1996) purified a processed endopolyphosphatase of 35 kDa from yeast. Recently, it has been found that the 35-kDa protein contains amino acid sequences identical to the deduced Phm5 sequence (Sethuraman and Kornberg, personal communication).
Since the genes for polyP processing are regulated by the PHO regulatory system, polyP accumulation should be affected by mutations which affect this system. To investigate this, polyP levels under polyP overplus conditions were measured in pho4Δ, pho80Δ, and pho84Δ mutants (Figure (Figure4A).4A). The pho4 mutant, which is incapable of derepression of its target genes (Figure (Figure1)1) accumulates a lower level of polyP than wild-type. The pho84Δ mutant, lacking a high-affinity Pi transporter, accumulates a very low but detectable level of polyP. The pho80Δ mutant, surprisingly, did not accumulate detectable levels of polyP, despite the fact that PHO84, and PHM1 through 4 are highly expressed in this mutant. This paradoxical phenotype may be a consequence of an abnormal vacuole in this mutant (Nicolson et al., 1995 ).
Using DNA microarray technology, we have identified at least 22 genes, including 13 novel genes, whose expression is regulated by the PHO pathway. Based on the premise that PHO-regulated genes are integral components of Pi metabolism, further molecular genetic approaches were undertaken to search for the function of the previously-uncharacterized genes. Our work has identified five of these genes as being involved in polyP metabolism. Thus, a total of 17 genes have now been found to function in Pi acquisition in yeast (Figure (Figure9). 9).
It is interesting to consider how the products of those 17 genes work together as a physiological system for Pi acquisition and storage. When yeast encounter conditions of Pi starvation, the low Pi signal initiates Pho81 activity, which suppresses Pho80-Pho85 kinase activity. The ankyrin domain of the Pho81 protein is sufficient to inhibit Pho80-Pho85 activity (Schneider et al., 1994 ; Ogawa et al., 1995 ). The PHO-regulated Pho81 homologs YPL110c and SPL2 share this domain, but their involvement in Pho80-Pho85 inhibition remains obscure (Flick and Thorner, 1998 ). Inhibition of the Pho80-Pho85 kinase results in an active Pho4 protein (hypophosphorylated form), which is localized to the nucleus where it acts as a specific transcriptional activator of PHO-regulated genes. Transcribed genes include PHO81, providing a positive feedback loop, which acts to keep Pho4 in its active form (Ogawa et al., 1995 ), resulting in high, continued expression of all PHO-regulated genes.
The Pi starvation signal triggers increased production of at least four types of phosphatases; 1) the acid phosphatases Pho5, Pho11, and Pho12, which are localized in periplasmic space; 2) the alkaline phosphatase Pho8, which is localized to the vacuole, 3) the glycerol phosphatase Hor2; 4) the putative polyphosphatase Phm5, localized in the vacuole. The acid and alkaline phosphatases are nonspecific, and hydrolyze a variety phosphorylated substrates, including nucleic acids, phosphosugars, phospholipids, and phosphoproteins. The glycerol phosphatase hydrolyzes the phosphate ester bond of glycerolphosphate, which is found in many sugar and lipid metabolites. All of these enzymes can contribute to increased levels of free Pi.
Under conditions of Pi starvation, expression of genes encoding the phosphate transporters, Pho84 and Pho89, are induced. Their optimal conditions are quite different: Pho84 transports Pi optimally at pH 4 and cotransports H+. Pho89, on the other hand, has optimal activity at pH 9.0 and cotransports Na+. This pair of Pi transporters work well in a wide range of environmental conditions in which yeast live. Expression of PHO86, a Pi transporter-related gene, is also increased in Pi starvation. The Pho86 protein was originally thought to form a complex with Pho84 and Pho87 (Bun-ya et al., 1996 ), however Lau et al. (2000) recently found that Pho86 is localized to the endoplasmic reticulum, where it functions in the proper localization of the Pho84 protein to the plasma membrane. Thus, Pho86 is now thought to act indirectly in Pi uptake.
The PHM1 through 4 genes, which we have shown to be involved in polyP synthesis, contribute to the Pi accumulation by a unique mechanism. Our results suggest that polyP synthesis is required for proper Pi accumulation. When polyP synthesis is critically slow, it can control the rate at which Pi is taken up by Pho84 membrane transporter. When polyP synthesis is slow, intracellular free Pi levels become high, which in turn acts as a direct negative feed back on the Pho84 membrane transporter. This critical intracellular Pi level is achieved after approximately 5 min of incubation in media containing 0.1 mM Pi. The phm mutants that lacked detectable polyP synthesis activity (phm1Δ phm2Δ double, phm3Δ and phm4Δ single disruptants) showed rapid initial uptake of Pi (like wild-type) but were incapable of further Pi uptake after ~ 5 min. Interestingly, the phm2 mutant, which had some residual polyP synthesis activity (~ 10% of wild-type) could continue to accumulate Pi after the initial 5-min period, but did so at a very reduced rate. In this mutant, it appears that the rate-determining step in Pi uptake in the first 5 min was controlled by the Pho84 membrane transporter, and after this time the rate was controlled by Pi to polyP conversion.
In this study, we have shown that polyP plays an important role in Pi accumulation and metabolism in yeast. The evidence for this involvement is not only metabolic but also genetic. Similar genetic interactions between polyP and Pi metabolism have been observed in E. coli. The promoter of the ppk-ppx operon, containing the genes for polyP kinase and exopolyphosphatase, includes a pho box, the response element for the Pi starvation signal mediated by the phoB-phoR two-component regulator (Kato et al., 1993 ). This suggests that the corresponding bacterial genes are regulated in a manner analogous to PHM1 through 4 and PHM5, respectively. Vibrio cholerae, a Gram-negative bacterium, has a similar ppk-ppx operon, with a pho box in its promoter, and its ppk mutant was unable to sustain a high rate of Pi accumulation (Ogawa et al., 2000 ). Thus, regulation by Pi appears to be a physiologically conserved feature of the genes for polyP metabolism, in both bacteria and yeast. Moreover, these results suggest that a major physiological role of polyP may be to promote long-term uptake and accumulation of Pi.
Prior to this study, the PPX1 gene, encoding an exopolyphosphatase, was the only yeast gene implicated in polyP processing (Wurst et al., 1995 ). The expression of the PPX1 gene in this study showed no detectable response to Pi levels or perturbation of PHO regulation. Ppx1 protein is believed to be localized to the cytoplasm, which contains negligible amounts of polyP. Since >90% of total cellular polyP is accumulated in vacuoles (Urech et al., 1978 ), the principal physiological polyphosphatase is likely to be vacuolar. The predicted vacuolar localization and PHO-regulation of Phm5 suggest that it is likely to contribute significantly to polyP degradation in vivo. Indeed, the phm5 mutation resulted in a marked increase in polyP chain lengths, whereas the ppx1 mutation resulted in a much smaller change (Figure (Figure66A).
Every Gram-negative bacterium for which the genome has been sequenced, has genes homologous to E. coli PPK (Tzeng and Kornberg, 1998 ). No genes homologous to PPK have been found in the genomes of Gram-positive bacteria, archea, or eukaryotes. This PPK is the only reported enzyme capable of synthesizing polyP, despite the fact that polyP has been found in every organism in which it has been sought, including Gram-positive bacteria, eukaryotes, and archea (Kornberg, 1999 ). Kornberg (1999) has speculated that mammalian cells synthesize polyP directly from the incorporated Pi without an ATP intermediate. Our results suggest that eukaryotic cells have an enzyme system for polyP synthesis completely different from the PPK of Gram-negative bacteria. PolyP synthesis in yeast requires v-ATPase activity, which produces proton motive force across the vacuolar membrane. We therefore speculate that the high-energy phosphoanhydride bonds in polyP are directly synthesized from Pi, using the proton motive force as a source of energy, by a vacuolar membrane-bound enzyme(s), analogous to the F-type ATPase in mitochondria. Based on this hypothesis, the Phm1 through 4 protein complex is the best candidate for this polyP synthesis enzyme in yeast.
Does a Pi acquisition system similar to the yeast system exist in higher eukaryotic cells? Kido et al. (1999) reported that expression of type II Na+/Pi cotransporter gene, NPT2, in rat kidney is derepressed by a dietary Pi starvation, and that the Pi signal-responsive element in its promoter contains the sequence, CACGTG, which is identical to the Pho4-binding site in yeast. Moreover, the N-terminal regions of Phm1, Phm2, and Phm3 contain conserved domains shared with six additional S. cerevisiae proteins, and similar protein structures can be found in genomes of many eukaryotes. Eight of the nine yeast proteins with this domain are related to Pi metabolism. We therefore refer to this domain as the “phosphate (Pi) domain” (Figure (Figure3B).3B). Since most of the homologous genes in other species have little or no functional characterization (except for human XPR1, gene encoding xenotropic and polytropic retrovirus receptor [Battini et al., 1999 ]), the function of the Pi domain may provide a useful lead for further functional investigations. Thus, further studies of the yeast system for Pi metabolism are likely to provide fundamental insights into Pi metabolism in all eukaryotes.
We thank Dr. Y. Mukai in Osaka University for generously providing plasmids, Dr. A. Kornberg in Stanford University for his fruitful advice and discussion, Dr. D. Botstein in Stanford University for generously providing plasmids and his helpful discussion, and Dr. J. Krise for help with manuscript and discussion in course of this study. This work was supported by grant HG00983 from the NHGRI and by the Howard Hughes Medical Institute. P.O.B. is an associate investigator of the Howard Hughes Medical Institute.