The PEL1 Gene (Renamed PGS1 ) Encodes the Phosphatidylglycerophosphate Synthase of Saccharomyces cerevisiae *

Phosphatidylglycerophosphate (PG-P) synthase catalyzes the synthesis of PG-P from CDP-diacylglycerol and sn -glycerol 3-phosphate and functions as the committed and rate-limiting step in the biosynthesis of cardiolipin (CL). In eukaryotic cells, CL is found predominantly in the inner mitochondrial membrane and is generally thought to be an essential component of many mitochondrial functions. We have determined that the PEL1 gene (now renamed PGS1 ), previously proposed to encode a second phosphatidylserine synthase of yeast (Janitor, M., Jarosch, E., Schweyen, R. J., and Subik, J. (1995) Yeast 13, 1223–1231), in fact encodes a PG-P synthase of Saccharomyces cerevisiae . Overexpression of the PGS1 gene product under the inducible GAL1 promoter re-sulted in a 14-fold increase in in vitro PG-P synthase activity. Disruption of the PGS1 gene in a haploid strain of yeast did not lead to a loss of viability but did result in a dependence on a fermentable carbon source for growth, a temperature sensitivity for growth, and a petite lethal phenotype. The pgs1 null mutant exhibited no detectable in vitro PG-P synthase activity and no detectable CL or phosphatidylglycerol (PG); significant CL synthase activity was still present. The growth arrest phenotype and lack of PG-P synthase activity of a pgsA null allele of Escherichia and subsequent addition of Phos™ 530 produced the chemiluminescent signal visualized using x-ray film. Southern Hybridization Analysis of Genomic DNA— DNA samples were digested with restriction enzymes and separated by agarose gel electrophoresis. The DNA samples were transferred to positively charged nylon membranes by capillary action using 20 3 SSC (3 M NaCl, 0.3 M sodium citrate, pH 7.0) at room temperature. DNA was cross-2 to grow if it acquires a gene encoding PG-P synthase activity. plasmid pHD102 from isolated colonies was verified by sensitivity to chloram-phenicol and the absence of an uninterrupted copy of the wild-type pgsA gene of E. coli as determined by Southern hybridization (data Presence of the interrupted chromosomal copy of the pgsA gene was indicated by growth on kanamycin and Southern blot analysis. Cell-free extracts of two of the positive isolates were prepared by sonication (25) from cells grown at 42 °C in the presence of kanamycin, and these were assayed for in vitro PG-P synthase activity. Previous reports have established that the E. coli PG-P synthase in crude extracts retains up to 70% of its activity after incubation at 70 °C for 20 min (59), while the yeast PG-P synthase inactivated after 20 min at 60 °C Using optimal conditions for the yeast enzyme (which are suboptimal for the E. coli enzyme), the PG-P synthase in crude extracts of strain HD38/pEPG4 average for two isolates) after growth at 37 °C 18-fold greater than that of the wild-type control DH5 a (0.050 6 0.002 units/mg) grown under the same conditions. Diagnostic of the replace-ment E. coli yeast synthase the HD38 transformants the complete loss of synthase after treatment strain 20 50%

The synthesis of phosphatidylglycerol (PG) 1 and cardiolipin (CL) (1) utilizes CDP-diacylglycerol (CDP-DAG), a central intermediate of phospholipid metabolism in all organisms (see Fig. 1), which is synthesized by an enzyme exhibiting extensive homology over a broad spectrum of species (1). The committed and rate-limiting step in PG/CL biosynthesis in yeast (2,3) and Escherichia coli (4) is catalyzed by phosphatidylglycerophos-phate (PG-P) synthase. In yeast all of the steps after CDP-DAG formation appear to be associated with the mitochondrial inner membrane (3), while CDP-DAG is synthesized by a single gene product localized to both the mitochondria and the endoplasmic reticulum (5); however, trace amounts of PG-P synthase activity have been reported in the cytoplasmic membrane and secretory vesicles destined for this membrane (6).
PG-P synthases have been well characterized in several prokaryotic organisms and share significant amino acid homology along with a motif common to phosphatidyltransferases and enzymes that bind CDP-alcohols (7). The genes (pgsA) encoding these synthases have been biochemically verified in E. coli (8 -10), Rhodobacter sphaeroides (11), and Bacillus subtilis (12). No eukaryotic gene encoding PG-P synthase activity has been identified, and the open reading frame derived from the Saccharomyces cerevisiae genome sequence most homologous to bacterial PG-P synthases actually encodes CL synthase (13)(14)(15). Interestingly, there is a divergence between prokaryotic organisms and eukaryotic organisms including S. cerevisiae (16,17) in the final step of CL biosynthesis (Fig. 1), which would explain the lack of homology between bacterial and yeast CL synthases. Since there is also no homology in either amino acid sequence or reaction mechanism among eukaryotic and prokaryotic phosphatidylserine synthases (18 -20), there may also be a similar divergence between eukaryotic and prokaryotic PG-P synthases.
A mutant of Chinese hamster ovary cells in the synthesis of PG and CL (21,22), when grown at the restrictive temperature, displayed a 10-fold reduction in the level of PG-P synthase activity, a reduction in the rate of labeling of the PG and CL, and a 10-fold and 3-fold reduction in the steady state levels of PG and CL, respectively. This mutant has defects in electron transport and ATP production, shows reduced oxygen utilization, and has an increased rate of glycolysis. In addition there are distinct changes in mitochondrial morphology. Although anionic phospholipid synthesis is compromised in this mutant, there is no definitive evidence that the defect is in the gene encoding PG-P synthase or that the physiological consequences of this mutation are due to loss of PG and/or CL.
The mitochondrial defects noted in this somatic cell mutant are not surprising, given the biochemical evidence for the involvement of CL in many mitochondrial membrane associated processes (23,24). However, since yeast mutants lacking CL and CL synthase have elevated levels of PG but are only partially compromised for growth on non-fermentable carbon sources (13,15), PG may substitute for CL in many important mitochondrial functions. E. coli mutants (null in pgsA) unable to make PG and CL are not viable (25) and have been used to demonstrate a requirement for PG in protein translocation across the inner membrane (26 -29) and initiation of DNA replication (30). Interestingly, E. coli mutants lacking CL synthase have few remarkable phenotypes, indicating that PG can also substitute for all critical functions of CL in E. coli (31,32). * This work was supported in part by Grants GM20478 and GM54273 from the National Institutes of Health (to W. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) Z48162 (corrected sequence).
Therefore, isolation of mutants defective in both PG and CL biosynthesis in eukaryotic cells is necessary to study the role of these anionic phospholipids in cell function.
Several nuclear mutations in yeast have been isolated that are viable except when carried in mitochondrial petite mutant backgrounds (i.e. rho Ϫ or rho 0 mutants) and have been named pel mutants for petite lethal (33). Mutants in PEL1 (33, 34) are unique within this group in that they cannot grow on synthetic medium using a non-fermentable carbon source, are temperature-sensitive for growth on glucose, and have very low cytochome c oxidase content. The PEL1 gene is non-allelic with genes encoding processes known to be directly involved in oxidative phosphorylation, suggesting that this single nuclear mutation indirectly affects processes necessary for oxidative phosphorylation. These characteristics by themselves would not suggest a mutation in mitochondrial phospholipid metabolism, but these mutants also lack CL (35); lack of PG was not investigated. In addition the PEL1 gene product has some suggestive homology (36) with the E. coli pssA gene product (phosphatidylserine synthase, which also uses CDP-DAG as a substrate), with a number of phospholipases D, and even more remotely with the E. coli CL synthase (37). The PEL1 gene has been suggested to encode a second minor phosphatidylserine synthase activity in S. cerevisiae (and has also been denoted as PSS2; Ref. 38) based solely on sequence homology comparisons even though yeast cells carrying multiple copies of this gene in a cho1 (encodes the major yeast phosphatidylserine synthase) null mutant background still lack this synthase activity (35). Based on the above observations, we investigated the possibility that the PEL1 gene encodes PG-P synthase activity, which is consistent with the many of the above observations.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were reagent grade or better. Restriction endonucleases and DNA modifying enzymes were from Promega Corp., New England Biolabs, Stratagene, and Boehringer Mannheim. Thin layer chromatography Silica Gel 60 and HPTLC Silica Gel 60 plates were from EM Science. The polymerase chain reaction (PCR) was performed using PCR SuperMIX™ from Life Technologies, Inc. or Taq polymerase and reagents from Perkin-Elmer. Oligonucleotides were prepared commercially by Genosys Biotechnologies, Inc. QIAEX II™ gel extraction kit was from Qiagen. The Genius™ 1 kit (DNA labeling and detection kit, nonradioactive), digoxigenin-labeled DNA molecular weight markers, positively charged nylon membranes, and Lumi-Phos™ 530 were purchased from Boehringer Mannheim. All media for yeast and bacterial growth and selection were from BIO 101 and Difco. Radiochemicals were obtained from Amersham Pharmacia Biotech and American Radiochemicals. The BCA kit for protein determination was from Pierce. Phospholipids except for CDP-DAG were from Sigma. CDP-DAG (dioleolyl) was synthesized (39) and generously provided by Dr. George Carman (Rutgers University). Pronase (Streptococcus griseus) was from Calbiochem, and Zymolyase 100T was purchased from Seikagaku Corp. Universol scintillation fluid was from ICN Biomedicals, Inc.
Strains, Plasmids, and Growth Conditions-A list of strains and plasmids used in this work is given in Table I. Methods for E. coli growth and selection have been described previously (40), and all bacterial strains were grown at 37°C unless otherwise indicated. Bacterial strains were grown in LB medium (1% Bacto-tryptone, 0.5% Bactoyeast extract, 1% NaCl, pH 7.4) and supplemented with antibiotics when needed for plasmid selection as described previously (25). Strain HD38 carries a null allele of the pgsA gene and is absolutely dependent for growth on a gene encoding PG-P synthase activity. The presence of plasmid pHD102 will support growth of strain HD38 at 30°C but not 42°C because the plasmid itself is temperature-sensitive for replication, preventing daughter cells from inheriting a functional copy of the pgsA gene (25). Methods of yeast growth and selection were described previously (43,44), and all yeast strains were grown at 30°C unless otherwise noted. Yeast cells were grown in the following media. YPD medium consisted of 1% Bacto-yeast extract, 2% Bacto-peptone, and 2% dextrose. In YPGly or YPGal medium, 2% glycerol or galactose, respectively, replaced dextrose as the carbon source. Complete synthetic medium (CSM) was constituted as described previously (44) using Bio101 drop out medium plus required amino acids where appropriate, yeast nitrogen base with vitamins, and the indicated carbon source (D, dextrose; Gal, galactose; Gly, glycerol) at 2% (w/v). Yeast selection media contained the components of CSM except those noted for selection purposes (i.e. minus uracil, ϪUra; minus histidine, ϪHis). All the above media were supplemented with 2% agar (yeast) or 1.5% agar (E. coli) for growth on plates.
DNA Manipulations and Sequencing-Methods for plasmid and genomic DNA preparation, restriction enzyme digestion, DNA ligation, and E. coli transformation (CaCl 2 protocol) were performed as described previously (40,45). Yeast transformation was performed by the lithium acetate protocol (46). Chromosomal DNA was prepared as described previously (47). Plasmid DNA to be sequenced was purified by using the Wizard™ 373 kit (Promega), and DNA sequencing reactions were performed by the Taq Dye-deoxy Terminator (Applied Biosystems) method and run on an Applied Biosystems Sequenator as a service provided by the Molecular Genetics Core Facility, University of Texas Medical School, Houston, TX. DNA fragments to be sequenced were carried in plasmid pBluescript™ II KS (Stratagene) and were sequenced by using the T7 and T3 primers or specific primers derived from the determined sequence. Sequence analysis was carried out with the Pileup program in GCG (48).
Amplification of DNA by PCR-For both analytical and preparative purposes, PCR was performed after optimizing conditions as described previously (49). Amplification of the PGS1 gene from yeast chromosomal DNA employed primers based on the most recently updated DNA sequence (reported under GenBank™ accession number Z48162), which indicates that open reading frames YCL4w and YCL3w reported in the Saccharomyces Genome Data Base 2 are one open reading frame (36). Primer 1 (below) was designed to introduce both a HindIII site (underlined) for cloning and a prokaryotic ribosomal binding site (in italic type) 5Ј to the putative start codon (in bold) of the PGS1 gene; bases changed from the reported DNA sequence are indicated in lowercase. Primer 2 was designated in a similar manner except it was designed to begin priming 5Ј to the second methionine (Met-30) in the putative sequence (see Fig. 2), and it contains a BamHI site for cloning purposes. Primer 3 contains an EcoRI site and was designed to begin priming 846 bp from the 3Ј end of the PGS1 gene. Primer 1, 5Ј-AGCAagCTtAGGAT-AggAgATATTAATGACG-3Ј; primer 2, 5Ј-CCAGgAtCCCTTCAATAgg-AgAAGGCAGATGTCC-3Ј; primer 3, 5Ј-GCTATAATAGAAtTcATCGA-TCTATTTACGGGC-3Ј.
DNA Labeling and Detection-The Genius™ 1 kit was used according to the manufacturer's directions for preparation of and detection with nonradioactive DNA probes. The kit utilizes random priming of template DNA and incorporation of digoxigenin-dUTP into the probe. Templates were isolated by agarose gel electrophoresis, extracted by using the QIAEX™ II gel extraction kit, and used for the randomprimed labeling reaction. An alkaline phosphatase-coupled antibody directed against digoxigenin was used to detect hybridization of probe, and subsequent addition of Phos™ 530 produced the chemiluminescent signal visualized using x-ray film.
Southern Hybridization Analysis of Genomic DNA-DNA samples were digested with restriction enzymes and separated by agarose gel electrophoresis. The DNA samples were transferred to positively charged nylon membranes by capillary action using 20ϫ SSC (3 M NaCl, 0.3 M sodium citrate, pH 7.0) at room temperature. DNA was cross-FIG. 1. Cardiolipin biosynthetic pathway. The genes encoding the PG-P synthase (PGS1 and pgsA) and CL synthase (CLS1 and cls) of S. cerevisiae and E. coli, respectively, are indicated next to the reactions their gene products catalyze.
linked to the membrane by using a UV Stratalinker™ 1800. The membrane was placed in standard hybridization buffer (5ϫ SSC, 0.1% Nlauroylsarcosine, 0.02% SDS, and 1% Genius™ 1 kit blocking reagent) for 2 h at 68°C. After the addition of digoxigenin-labeled probe, hybridization was performed overnight at 68°C in standard hybridization buffer. Following hybridization, membranes were washed twice for 5 min in 2ϫ SSC, 0.1% SDS at room temperature and twice for 15 min in 0.1% SSC, 0.1% SDS at 68°C. Alkaline phosphatase activity was visualized by the addition of Phos™ 530.
Preparation of Mitochondrial-enriched Fraction-All cell fractionation procedures were performed at 4°C by modification of a previously described procedure (50). Cultures of yeast cells were grown to late log phase in CSMD or CSMGal with auxotrophic selection when appropriate. Cells were pelleted and washed with homogenization buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.3 M sucrose, 10 mM 2-mercaptoethanol), resuspended in one volume of homogenization buffer, and frozen at Ϫ80°C for later use. Cell suspensions (1 ml) were mixed with an equal volume of 0.5-mm silica/zirconium beads (Biospec Products), and cells were disrupted at 4°C by mechanical shearing using a Mini-Bead-Beater™ 8 (Biospec Products) for five 1-min bursts with intermittent 2-min pauses on ice. Unbroken cells and debris were pelleted twice by centrifugation at 1,500 ϫ g for 10 min. An enriched mitochondrial pellet was obtained by subjecting the supernatant to centrifugation at 27,000 ϫ g for 10 min in Sorvall SS-34 rotor, and the resulting pellet was washed twice with homogenization buffer. The final pellet was resuspended in 0.25 ml of 20% glycerol, 50 mM Tris, pH 7.5, and 10 mM 2-mercaptoethanol per ml of starting cell suspension using a Dounce homogenizer. The mitochondrial preparations were stored at Ϫ80°C.
Enzyme Assays-PG-P synthase activity was measured by the incorporation of radiolabeled substrate sn-[ 14 C]glycerol 3-phosphate into the chloroform-soluble product PG-P (50). The assay was performed at 30°C in the presence of 50 mM MES-HCl buffer (pH 7.0), 0.3 mM MgCl 2 , 0.2 mM CDP-DAG, 0.5 mM sn-[ 14 C]glycerol 3-phosphate (2,000 -4,000 cpm/nmol), and 1 mM Triton X-100 in a total volume of 100 l. The reaction was stopped after 20 min by the addition of 0.5 ml of 0.1 N HCl in methanol, and the lipids were extracted with 2 ml of chloroform and 3 ml of 1 M MgCl 2 . The chloroform phase was evaporated followed by addition of Universol scintillation fluid and determination of radioactivity using a scintillation counter. A unit of enzymatic activity is defined as the amount of enzyme that catalyzes the formation of 1 nmol of product/min under the assay conditions described above. The specific activity is defined as unit/mg of protein. CL synthase activity was measured by the conversion of [ 32 P]PG to CL dependent on CDP-DAG as described previously (17). [ 32 P]PG was prepared by thin layer chromatography (51) from a chloroform/methanol extract (25) of E. coli strain ADC90 labeled with [ 32 P]orthophosphate during growth at 42°C.
Labeling and Analysis of Phospholipids-For steady state labeling cells were grown in 5-ml aliquots of CSMD with auxotrophic selection when appropriate. [ 32 P]Orthophosphate was added to a final concentration of 10 Ci/ml, and the cells were grown for 16 h (six generations beginning at an A 600 of 0.05) before harvesting by centrifugation at 1,500 ϫ g for 10 min. The cells were washed with 5 ml of H 2 O, and the resulting pellet was resuspended in 0.6 ml of chloroform/methanol/0.1 N HCl (1:2:0.8, v/v) and approximately a 200-l volume of 0.5-mm silica/ zirconium beads was added. Carrier lipids (30 g of an equal-ratio mixture of phospholipids including phosphatidylcholine, phosphatidylinositol, phosphatidylethanolamine, phosphatidylserine, phosphatidic acid, PG, and CL) were added. Cells were broken using a Mini-Bead-Beater™ (Biospec Products) at maximal setting for 2 min. Cell debris was pelleted, and the supernatant was extracted after addition of 0.2 ml of chloroform and 0.2 ml of 0.1 N HCl, 0.5 M NaCl (52). The organic phase was separated by centrifugation and taken to dryness. Isolated radiolabeled phospholipids (ca. 10,000 cpm) were dissolved in chloroform and applied to HPTLC Silica Gel 60 plates for two-dimensional thin layer chromatography (53). The first dimension was developed from right to left (Fig. 4) with chloroform/methanol/acetic acid (65:25: 10, v/v). The second dimension was developed from bottom to top with chloroform/methanol/88% formic acid (65:25:10, v/v). The position of each spot was determined both by authentic standards (visualized by iodine staining and ninhydrin spray) as well as comparison to previous reports (53). The results were visualized by exposure to x-ray film, and for quantitative results the radioactivity was counted in a Packard Instant Imager (Packard Instruments Inc.). Alternatively, one-dimensional thin layer chromatography was performed using boric acid-impregnated Silica Gel 60 plates (51) developed in chloroform/methanol/ water/ammonium hydroxide (60:37:5:3:1, v/v). Phospholipid content is expressed as mol % of total phospholipid based on the radiolabel in each spot and its respective molar phosphate content.

RESULTS
Isolation of the PGS1 Gene-Identification and initial characterization of the PEL1 gene has been reported previously (33-36). The PEL1 gene expressed from its endogenous promoter and carried on a plasmid complements the multiple phenotypes of pel1 mutants. Although the PEL1 gene was postulated to encode a second phosphatidylserine synthase, it appeared to us that PEL1 plays an important role in CL biosynthesis since a null mutant of PEL1 has no detectable CL (35). The following biochemical and genetic evidence supports PEL1 (PGS1) as encoding the PG-P synthase of S. cerevisiae.
The PEL1 gene has been mapped to chromosome III of yeast and encompasses two open reading frames (YCL3w and YCL4w), 2 which have been reported to define a single open reading frame (36). We isolated the complete open reading frame of the putative PGS1 gene (including more extensive 5Ј coding sequence not previously reported) from genomic DNA of yeast strain YP501 using PCR and primers 1 and 3 as outlined under "Experimental Procedures." The PCR product was subcloned into the HindIII and EcoRI sites of the plasmid pBlue-script™ II KS generating plasmid pBA53. This insert in plasmid pBA53 was excised and ligated between the HindIII and EcoR I sites of plasmid pYES2 generating plasmid pYPGS10 -2 thus placing the PGS1 gene under the regulation of the GAL1 promoter (P GAL1 ). A truncated version of the putative PGS1 gene made by first employing primers 2 and 3, which intro- Phosphatidylglycerophosphate Synthase of S. cerevisiae duced a BamHI site 5Ј to the second methionine (Met-30) of the putative sequence (see Fig. 2) and an EcoRI site 3Ј of the gene. This fragment was subcloned between the BamHI and EcoRI sites of plasmid pUC19 placing the gene under the control of lacOP. Three (plasmids pEPG1, pEPG4, and pEPG5) independent isolates from a single PCR experiment were analyzed further. The PGS1 gene from the independently generated plasmids pBA53 (complete gene) and pEPG1, pEPG4, and pEPG5 (beginning near the second methionine of the putative gene) was completely sequenced except where noted below. All four PGS1 isolates showed the same three discrepancies (noted in Fig. 2) with the reported sequence for the PEL1 gene (36) and the more complete sequence information annotated under Gen-Bank™ accession number Z48162. However, these three regions of the sequence agree with the sequence reported in the Saccharomyces Genome Data Base, which itself has multiple discrepancies with our sequence as well as the reported PEL1 sequence. Our sequence data only extends through bp 1526 of Fig. 2, but the Data Base and PEL1 sequences agree beyond this point. Therefore, the DNA sequence and derived amino acid sequence shown in Fig. 2 most likely represents the genomic sequence of the PGS1 gene. The full length PGS1 gene in plasmid pBA53 also showed a silent change of a T to G and a conserved change of A to T (valine to alanine), which were not found in any of the other clones or reported sequences suggesting errors introduced by PCR (not shown in the consensus sequence). Plasmid pEPG1 and pEPG5 also had single base discrepancies not found in any other sequences and which were not included in the final consensus sequence.
The PGS1 gene defines a putative 1,563-bp open reading frame encoding a 521-amino acid protein with a predicted molecular weight of 59,333. Sequence homology with the E. coli pssA gene product is only weakly suggestive of some functional relationship. The first 300 amino acids of the E. coli enzyme show about 23% sequence identity and 45% similarity with the amino acid segment 29 -375 (beginning near Met-30) of the yeast enzyme; homology is scattered and there are no extensive highly homologous regions. Analysis for potential subcellular location of the gene product by PSORT strongly indicates a mitochondrial matrix or inner mitochondrial membrane (54) location for the synthase. A hydropathy plot of the predicted amino acid sequence shows no strongly hydrophobic domains in the primary sequence, in contrast to the overall hydrophobicity of the E. coli PG-P synthase (9), suggesting a mitochondrial matrix or inner membrane peripheral protein. Previous reports indicated "solubilization" of synthase activity from isolated mitochondria using detergents (50), but no studies have been The underlined amino acids are the result of changes introduced by the PCR primer used to generate this derivative of the PGS1 gene. The last 3 amino acids of this line begin the sequence of the PG-P synthase in these plasmids. The second and third lines begin the complete sequence of the PGS1 gene and its product with nucleotide and amino acid numbering shown at the right, respectively. The underlined and bold nucleotides at positions 132 (silent change) and 1493 (results in amino acid change noted) are those that differ from the previously reported sequence for PEL1 as noted in the text; lack of the previously reported "A" (shaded) between nucleotides 1506 and 1507 results in a frameshift and a change from previous reports in the C-terminal amino acid sequence noted by italics.
reported on release of activity by permeabilizing mitochondria without disruption of membrane bilayer structure. Cleavage of a potential mitochondrial targeting sequence was predicted at or near Met-30. The first 29 amino acids of the protein contain 6 basic amino acids (4 of which are arginines) spaced about one helix turn apart, no acidic amino acids, only 1 proline, and 9 hydrophobic amino acids suggesting a possible amphipathic helix with a high positive charge consistent with mitochondrial targeting sequences (55). Therefore, the predicted properties of this gene product are consistent with that of a protein targeted to either the mitochondrial inner membrane or matrix.
Interestingly, 284 bp 5Ј to the open reading frame (not shown), there is a sequence (5Ј-CAAGTGAAT-3Ј) that matches well with the consensus upstream activating sequence UAS INO , 5Ј-CATGTGAAAT-3Ј (56,57), which has been found associated with genes whose expression is repressed by inositol in the growth medium. Inositol has been shown to regulate PG-P synthase activity levels (58), but the mechanism remains uncharacterized.
Complementation of an E. coli pgsA Null Mutant-Based on the scheme used to construct the pEPG series of plasmids, each should express a chimeric protein with a short N terminus derived from the N terminus of ␤-galactosidase plus additional amino acids resulting from the PCR primer fused to the putative PG-P synthase sequence beginning with Arg-28 (Fig. 2), which is near a possible posttranslational processing site. If PGS1 encodes a PG-P synthase, then these plasmids should complement E. coli mutant strains such as strain HD38 lacking PG-P synthase activity.
Strain HD38/pHD102 was transformed with plasmid pEPG4 (lacks any differences in the consensus DNA sequence (Fig. 2) for the PGS1 gene) and complementation was scored by growth of cells at 42°C on LB agar plates containing kanamycin and ampicillin. Under these conditions strain HD38 should lose plasmid pHD102 and will only be able to grow if it acquires a gene encoding PG-P synthase activity. Loss of plasmid pHD102 from isolated colonies was verified by sensitivity to chloramphenicol and the absence of an uninterrupted copy of the wild-type pgsA gene of E. coli as determined by Southern hybridization (data not shown). Presence of the interrupted chromosomal copy of the pgsA gene was indicated by growth on kanamycin and Southern blot analysis. Cell-free extracts of two of the positive isolates were prepared by sonication (25) from cells grown at 42°C in the presence of kanamycin, and these were assayed for in vitro PG-P synthase activity. Previous reports have established that the E. coli PG-P synthase in crude extracts retains up to 70% of its activity after incubation at 70°C for 20 min (59), while the yeast PG-P synthase is completely inactivated after 20 min at 60°C (50). Using optimal conditions for the yeast enzyme (which are suboptimal for the E. coli enzyme), the PG-P synthase activity in crude extracts of strain HD38/pEPG4 (0.88 Ϯ 0.04 units/mg, average for two isolates) after growth at 37°C was 18-fold greater than that of the wild-type control DH5␣ (0.050 Ϯ 0.002 units/mg) grown under the same conditions. Diagnostic of the replacement of the E. coli synthase by the yeast synthase in the HD38 transformants was the complete loss of synthase activity after treatment of extracts of strain HD38/pEPG4 at 65°C for 20 min as compared with the retention of 50% of the activity by the wild-type control. Introduction of plasmid pEPG4 into strain DH5␣ also results in a similar increase of PG-P synthase activity over wild-type levels; interestingly, plasmid pEPG5, which has an Ala-456 3 Pro change in the predicted gene product sequence apparently introduced by PCR, conferred no increase in synthase activity. Complementation of a pgsA null mutant of E. coli by plasmid pEPG4 definitively establishes that the PGS1 gene of yeast encodes a PG-P synthase.
Expression of the PGS1 Gene in Yeast-Cultures of wild-type strain DL1 with or without plasmid pYPGS10 -2 (P GAL1 -PGS1) were grown to late log phase in CSMϪUra or CSM, respectively, with 2% glucose or 2% galactose as carbon source. Mitochondrial-enriched fractions were prepared and examined for PG-P synthase activity as described under "Experimental Procedures." Wild-type strain DL1 routinely displayed a higher specific activity when grown in glucose (0.12 Ϯ 0.005 units/mg) versus galactose (0.066 Ϯ 0.002 units/mg). PG-P synthase activity in strain DL1 transformed with the plasmid pYPGS10 -2 was 0.25 Ϯ 0.008 units/mg when grown under non-induced conditions (glucose) versus 0.94 Ϯ 0.020 units/mg when grown under induced conditions (galactose). The 14-fold increase in PG-P synthase activity dependent on plasmid pYPGS10 -2 in galactose grown cells is consistent with the PGS1 gene encoding PG-P synthase activity. The 14 C-labeled chloroform-soluble product formed by extracts of DL1 was verified by one-dimensional thin layer chromatography to be predominantly PG; a minor amount of PG-P and no CL were found.
Disruption of the Genomic PGS1 Gene-The yeast PGS1 gene was disrupted in vitro and introduced into the genome by homologous recombination. This disruption was accomplished by removing 413 bp from the center of the open reading frame (beginning with nucleotide 599 from the beginning of the open reading frame) by digestion of plasmid pBA53 with AsuII and PstI. The HIS3 gene derived from an AhaII-NsiI digestion of vector pRS303 (42) was inserted at this position. The pgs1::HIS3 gene was released from the resulting plasmid by digestion by AseI and EcoRI digestion (resulting in fragment ends homologous with sequences internal to the PGS1 gene) and used to transform strain DL1 (41). Transformants were selected on CSMDϪHis medium to produce the pgs1::HIS3disrupted haploid strain YCD4.
Disruption of the PGS1 gene was confirmed by Southern hybridization analysis (Fig. 3). Genomic DNA from wild-type strain DL1 and pgs1 null mutant YCD4 was isolated and digested with EcoRI as described under "Experimental Procedures." Probe specific for the PGS1 gene was generated using the HindIII-EcoRI insert of the plasmid pYPGS10 -2 serving as template. In strain DL1, a band consistent with the predicted 3,452-bp fragment for the wild-type allele was seen, and in the strain YCD4, a band consistent with the predicted 4,376-bp disrupted allele was seen. Phenotypic Characterization of Strain YCD4 -The pgs1 null strain YCD4 was viable but exhibited significantly different growth phenotypes compared with the parental strain DL1. From overnight cultures, fresh cultures of strains YCD4 and DL1 were inoculated into YPD, CSMD, or CSMDϪHis medium. To induce the appropriate metabolic enzymes, cells were also allowed to grow overnight in either rich or minimal media containing 0.2% glucose and either 2% glycerol or 2% galactose. After serial dilutions, cells were seeded on the following agar plates: YPD; YPGal; YPGly; and CSM or CSMϪHis with either 2% glucose, 2% glycerol, or 2% galactose. The cells were grown at both 30°C and 37°C. To test whether the pgs1 null strain exhibited the petite lethal phenotype, cells were also seeded to plates (CSMD, CSMDϪHis, or YPD) containing 25 mM ethidium bromide; cells grown in the presence of ethidium bromide acquire a high frequency of deletions in their mitochondrial DNA generating rho mutants. When compared with the strain DL1, strain YCD4 formed small colonies on dextrosecontaining plates and no colonies on glycerol-containing plates consistent with a defect in mitochondrial oxidative phosphorylation; the mutant also did not form colonies on galactosecontaining plates. Strain YCD4 did not form colonies when plated in the presence of ethidium bromide (petite lethal) or at 37°C, even when grown on glucose consistent with the previously observed properties of pel1 mutants (33, 34). Strain YCD4 grew in liquid media significantly slower and reached a lower cell density in stationary phase than strain DL1.
An enriched mitochondrial preparation made from strain YCD4 grown on CSMD contained no detectable PG-P synthase activity although the lower limit of detectability was estimated at about 10% of the wild-type levels for cells grown in glucose; CL synthase activity was detectable at about 50% the normal level in strain DL1. We were not able to reproducibly complement the pgs1 null strain with a plasmid borne copy of the PGS1 gene under P GAL1 regulation. The basis for the inability of the pgs1 null mutant to grow on galactose is not known, but this inability to metabolism galactose may make induction of PGS1 expression under P GAL1 regulation variable and difficult. Difficulty in inducing expression from genes under P GAL1 regulation in cells compromised in mitochondrial energy metabolism has been noted in some genetic backgrounds (60,61). However, PEL1 (PGS1) expressed on a plasmid from its endogenous promoter does complement the pel1 null mutant (36), verifying that the pleiotropic phenotypes of pgs1/pel1 mutants are due to a single nuclear lesion and not caused by additional secondary mutations in either nuclear or mitochondrial DNA.
Phospholipid Composition of Wild-type and Mutant Strains-Strains DL1 and YCD4 were labeled by growth in [ 32 P]orthophosphate and analyzed as described under "Experimental Procedures" (Fig. 4). Visual inspection of the autoradiograms indicated no detectable CL or PG in strain YCD4 compared with wild-type levels of 1.0% and 0.5%, respectively (Table II). Separation of phospholipids by a one-dimensional system, which clearly separates PG and CL from phosphatidic acid and phosphatidylethanolamine, also showed no evidence of CL or PG in the mutant strain (data not shown). The only other significant difference in the lipid composition between strains DL1 and YCD4 was higher phosphatidylinositol content in the latter strain, which was also noted in a pel1 null strain (35). These results are consistent with the PGS1 gene encoding the major PG-P synthase, if not all of the PG-P synthase, and mutations in the PGS1 gene being responsible for the changes observed in total phospholipid composition. DISCUSSION The results reported here conclusively demonstrate that the PGS1 gene, formerly named PEL1 and proposed to encode a second phosphatidylserine synthase activity (33, 35,36), encodes the major PG-P synthase of yeast. Quite surprising is the lack of amino acid homology between the prokaryotic and yeast PG-P synthases, the significant homology especially over extended regions (26% identity and 52% similarity) between the E. coli PG-P synthase and the yeast CL synthase (13)(14)(15), and the weak but significant homology between the yeast PG-P synthase and the E. coli phosphatidylserine synthase (36). Given the similarity in the substrates of the PG-P, phosphatidylserine, and CL synthases, these unpredictable relationships are not completely unreasonable. Such results should be a warning against reliance on sequence homology relationships alone without biochemical verification in the assignment of gene product function. On the other hand, now that the eukary- FIG. 4. Phospholipid analysis of the pgs1 null strain. Strains DL1 (panels A and C) and YCD4 (panels B and D) were grown in CSMD to late log phase in the presence of [ 32 P]orthophosphate. Total membrane phospholipid was extracted and separated on HPTLC plates as described for two-dimensional thin-layer chromatography under "Experimental Procedures." Panels A and B were short exposures (2 h), and panels C and D were long exposures (12 h) of the same thin layer chromatography plates. Standard phospholipids were used to assign the identity of each spot as follows: CL, cardiolipin; PA, phosphatidic acid; PG, phosphatidylglycerol; PE, phosphatidylethanolamine; PS, phosphatidylserine; PI, phosphatidylinositol; PC, phosphatidylcholine. otic genes have been properly identified, rapid progress should be possible in isolating the mammalian homologues of these genes.
The disruption of the PGS1 gene in yeast is not lethal to cells but does seriously compromise mitochondrial function, indicating a requirement for PG and/or CL in critical functions of this organelle. The lack of a significant growth phenotype for the cls1 null mutant of yeast lacking CL but accumulating PG is surprising (13)(14)(15), given the diverse functions postulated for CL (23,24). However, cells lacking CL do enter stationary phase at a lower cell density and grow slower on non-fermentable carbons sources, indicating some compromise in optimal mitochondrial function. The acquisition of more severe defects in mitochondrial function by the pgs1 null mutant indicates that PG can substitute for CL in critical mitochondrial processes and may explain the lack of a more severe growth phenotype for the cls1 mutant. A similar absolute requirement for PG and limited effects upon loss of CL have been found in E. coli (4).
The yeast PGS1 and CHO1 gene products (35,36) like the E. coli cls and pssA gene products (62) appear to be interrelated in their requirement for cell viability. Single null mutants of either gene pair are viable, but double null mutants are not viable; we do not yet know if a double null cls1 cho1 mutant of yeast is viable. Such results suggest an overlapping or synergistic role between anionic phospholipids and phosphatidylethanolamine (derived by decarboxylation of phosphatidylserine in the inner membrane of E. coli; Ref. 63). In yeast phosphatidylserine is decarboxylated in the mitochondria by the PSD1 gene product (64,65) and in an extra-mitochondrial organelle by the PSD2 gene product (66,67), although there appears to be mixing of phosphatidylethanolamine between these subcellular compartments. Construction of mutations in either the PGS1 or CLS1 genes in strains carrying mutations in either of the above two PSD genes should shed light on whether the incompatibility of cho1 and pgs1 mutants is due to simultaneous loss of phosphatidylserine and PG/CL or loss of mitochondrial phosphatidylethanolamine and PG and/or CL.
The first reported pel1 mutants had low levels of cytochrome c oxidase as determined by spectral measurements (34), which would certainly explain the lack of growth on non-fermentable carbon sources but not the molecular basis for the pleiotropic defects in pgs1 mutants such as the lack of growth on galactose, temperature sensitivity for growth, and the petite lethal phenotype. Anionic phospholipids have been suggested as important participants for the efficient import of nuclear encoded proteins into the mitochondria (68 -70), as has been established for protein translocation across membranes in E. coli (4). Since several of the subunits of cytochrome c oxidase are encoded by nuclear genes and must be imported into the mitochondria (71), lack of anionic phospholipids may be the basis for lack of oxidase function and spectral signature. The mitochondrial encoded subunits that impart the characteristic spectrum to the oxidase are rapidly degraded in the absence of stoichiometric amounts of the nuclear encoded subunits (71). The effect on the assembly of multiple mitochondrial proteins coupled with the complete loss of mitochondrial encoded subunits (petite lethal phenotype) of these proteins may lead to cumulative effects, resulting in loss of mitochondria integrity or minimal functions required for cell viability as reported for pel1 mutants (34).
Now that the precise defect in the pel1 mutant has been identified, the characterization of the molecular basis for the role of anionic phospholipids in eukaryotic cells is approachable. It should be possible to design "biological reagents" with regulated expression of both the PGS1 and CLS1 genes to study the role of anionic phospholipids in mitochondrial processes in much the same manner as has been done in E. coli to study the roles of PG, CL and phosphatidylethanolamine in cell function (4).