Cloning of a Gene (PSDl) Encoding Phosphatidylserine Decarboxylase from Saccharomyces cerevisiae by Complementation of an Escherichia coli Mutant*

A gene (PSD1) encoding a phosphatidylserine decarboxylase of Saccharomyces cerevisiae was cloned by complementation of a conditional lethal mutation in the homologous gene in Escherichia coli strain EH150. Expression of the cDNA clone in EH150 corrected growth, phospholipid, and phosphatidylserine decar- boxylase activity defects. Expression of the genomic clone in wild type yeast resulted in 20-fold amplifica- tion of phosphatidylserine decarboxylase activity. A 1500-base pair open reading frame encodes a 56,558- Da protein with a potential mitochondrial targeting sequence. Upstream regulatory elements found in other enzymes of the phospholipid biosynthetic path-way are present in PSDl. The derived amino acid sequence shows 44 and 35% identity with the phosphatidylserine decarboxylases from Chinese hamster ovary cells and E. coli, respectively. Near the carboxyl terminus is an LGST sequence which, in E. coli, is the site of proteolytic cleavage of the proenzyme into the a and /3 subunits and formation of the pyruvate pros- thetic group (Dowhan, W., and Li, Q.-X. (1992) Methods Enzymol. 209,348-359). Disruption of the PSDl gene in a haploid strain of yeast resulted in loss of

The conversion of phosphatidylserine to phosphatidylethanolamine is catalyzed by phosphatidylserine decarboxylase in both eukaryotic and prokaryotic organisms (Bishop and Bell, 1988;Carman and Henry, 1989;Raetz and Dowhan, 1990). In Escherichia coli the gene ( p s d ) and gene product responsible for this essential step in phospholipid metabolism have been extensively studied (Dowhan et al., 1972;Hawrot and Kennedy, 1975;Satre and Kennedy, 1978). The psd gene encodes a proenzyme or *-subunit (MI = 35,893) which is cleaved post-translationally to a p subunit ( M , = 28,579) and an a-subunit (Mr = 7,332) (Li and Dowhan, 1988). The asubunit is derived from the carboxyl-terminal 69 amino acids of the ?r-subunit and is blocked at its amino terminus by an * This work was supported in part by United States Public Health Service Grant GM 35143. 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 thispaper has been submitted to the GenBankTM/EMBL Data Bank with accession numbeds) L20973.
$ T o whom reprint requests should be addressed. Tel.:  amide-linked pyruvate prosthetic group. Ser-254 of the Psubunit is essential for the formation of the two subunits and the pyruvoyl group. The @-hydroxyl of Ser-254 functions as a nucleophile attacking the carbonyl of the adjacent peptide bond to form an ester intermediate. This is followed by an a,@-elimination reaction and hydrolysis resulting in the formation of the two subunits. The resulting pyruvoyl group contributes the catalytic carbonyl group of the mature enzyme (Li and Dowhan, 1988;Dowhan and Li, 1992). This mechanism for activation has been observed in all pyruvoyl-dependent decarboxylases thus far investigated (van Poelje and Snell, 1990). Since pyruvoyl-dependent decarboxylases show extensive protein sequence homology within specific substrate groups and over several species (van Poelje and Snell, 1990), it is not completely surprising that the gene encoding phosphatidylserine decarboxylase from CHO' cells was originally identified by the similarity of the derived protein sequence with that of the enzyme from E. coli (Kuge et al., 1991). The predicted a-subunit for the CHO cell enzyme is 32 amino acids in length, about half the size of the a-subunit of the E. coli enzyme.
Considerably ,&is known concerning the phosphatidylserine decarboxglase from Saccharomyces cereuisiae since neither the gene nor the gene product has been isolated. Analogous to the mammalian enzyme, yeast decarboxylase activity is localized to the inner mitochondrial membrane (Kuchler et al., 1988) suggesting the requirement for a mitochondrial targeting sequence. Expression of the partial cDNA clone from CHO cells in yeast increases the phosphatidylserine decarboxylase activity and suggests functional homology between these enzymes (Kuge et al., 1991), but the cDNA does not include sufficient information at the amino terminus to confirm the presence of an amino-terminal targeting region. Yeast phosphatidylserine decarboxylase activity is responsive to the presence in the growth medium of precursors utilized in phospholipid biosynthesis (Lamping et al., 1991;Overmeyer and Waechter, 1991). This has been demonstrated for other phospholipid biosynthetic enzymes as well (Nikoloff and Henry, 1991;Carman and Henry, 1989). Regulatory sequences present 5' to the PSS (CHOl), ZNOl, PEMl, PEM2, and PIS genes appear to be transcriptionally responsive to these precursors (Bailis et al., 1992;Kodaki et al., 1991a;Nikoloff and Henry, 1991;Carman and Henry, 1989), but the precise mechanism of transcriptional regulation is not known.
The sequence of events leading to formation of the pros-thetic group and inner mitochondrial membrane assembly of yeast phosphatidylserine decarboxylase is of interest, and the further study of this enzyme is important because of its potential role in all eukaryotic cells in the regulation of phospholipid biosynthesis and interorganelle trafficking of phosphatidylserine. In eukaryotic cells phosphatidylserine is made in the endoplasmic reticulum, decarboxylated in the mitochondria, and translocated, as phosphatidylethanolamine, from the mitochondria to the other membrane systems of the cell (Voelker, 1991;Simbeni et al., 1991).
We report here the isolation of the PSDl gene of S. cereuisiae by complementation of a conditional lethal mutation in the E. coli psd gene. The yeast gene has been sequenced and compared to the related genes from other organisms. A potential mitochondrial targeting sequence and the predicted site of the proteolytic cleavage and formation of a pyruvate prosthetic group were found. We have also constructed a null allele in the yeast PSDl gene and characterized the phenotype of this mutant. The disruption of the gene resulted in loss of detectable phosphatidylserine decarboxylase activity but did not result in loss of viability or any major alterations in phospholipid composition.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were reagent grade or better. Radiochemicals and Hybond N nylon membranes were obtained from Amersham Corp. LiquiscintTM was purchased from National Diagnostics. Restriction endonucleases, DNA modifying enzymes, and the phage M13mp18 were obtained from New England Biolabs, Bethesda Research Laboratories, Boehringer Mannheim, Stratagene, and Promega Corp. Phospholipids were from Sigma. Agarose was from IBI and Sea Plaque low melt agarose was from FMC. The Gene Amp d PCR Reagent Kit was from Perkin-Elmer Cetus, and the SequenaseTM Version 2.0 Sequencing Kit was from U. S. Biochemical Corp. Long RangerTM polyacrylamide solution was from J. T. Baker Chemical Co. The GeniusTM 1 Kit (DNA Labeling and Detection Kit, Nonradioactive), digoxigenin-labeled DNA molecular weight markers, positively charged nylon membranes, and Lumi-PhosTM 530 were purchased from Boehringer Mannheim. Oligonucleotides were prepared commercially by Genosys Biotechnologies, Inc. or synthesized by the Molecular Genetics Core Facility, University of Texas Medical School at Houston using an Applied Biosystems 394 oligonucleotide synthesizer. The BCA kit was from Pierce. Yeast lytic enzyme (Arthrobacter luteus, 100,000 units/g) was from ICN Biochemicals. Pronase (Streptococcus griseus) was from Calbiochem. Components of bacterial and yeast growth media were purchased from Difco. Ultrafree-MC 0.45-pm filter units were from Millipore. Thin layer plates were from J. T. Baker Chemical Co., and HPTLC plates were from EPI. Phosphatidyl-L-[l-14C]serine was prepared enzymatically as previously described (Dowhan and Li, 1992) from phosphatidic acid derived from egg lecithin.
Growth Conditions, Strains, and Plasmids-Methods of yeast growth, sporulation, and tetrad analysis were as described by Sherman et al. (1986). YPD medium consisted of 1% Bacto-yeast extract, 2% Bacto-peptone, and 2% dextrose. In YPG, YPL or YPGal medium, 2% glycerol, lactate or galactose, respectively, replaced dextrose as the carbon source. YPGE medium contained 2% ethanol in addition to the glycerol in YPG medium. Minimal defined media were prepared from Yeast Nitrogen Base without amino acids and supplemented as described by Hirsch and Henry (1986). Inositol-free media were prepared according to Hirsch and Henry (1986), except that Vitaminfree Yeast Nitrogen Base was not available from suppliers, and components were supplied individually. Selection media omitted one component as required. Growth medium was supplemented with 1 mM ethanolamine and/or 1 mM choline as indicated. E. coli strains EH150, KK2186, and JF1754 were grown in LB medium or on M9 minimal medium (Maniatis et al., 1982) supplemented with amino acids required by the strain. Ampicillin (100-200 pg/ml) was added to cultures of E. coli strains carrying plasmids unless otherwise indicated. Agar plate versions of the above liquid media were made by adding 1.5% agar for bacterial media or 2% agar for yeast media. Yeast strains were grown at 30 "C, and bacterial strains were grown at 30, 37, or 42 'C as indicated. For the measurement of phosphati-dylserine decarboxylase activity in E. coli, cultures were grown in LB medium. For the measurement of phosphatidylserine decarboxylase activity in yeast strain DL1 transformed with genomic clones, cultures were grown to mid-log phase in inositol-free selection media; strain DL1 without plasmid was grown in inositol-free complete medium. For the measurement of phosphatidylserine decarboxylase activity in yeast strains not containing plasmids, cultures were grown to midlog phase in YPD medium.
A list of strains and plasmids used in this work is given in Table  I. E. coli strain EH150 cannot grow at the restrictive temperature of 42 "C due to a mutation (psd-2) in the gene encoding the phosphatidylserine decarboxylase. E. coli strain KK2186 was used to propagate derivatives of phage M13mp18. The AYES yeast cDNA library and XKC were kindly provided by Dr. Steve Elledge (Baylor College of Medicine). The library may be converted from X phage to plasmids in a suitable host strain. The resulting plasmids contain cDNAs in a vector (V-pSE936) which permits replication, expression, and selection in E. coli or S. cereuisiae. cDNA inserts, which can be excised by either XhoI or EcoRI endonuclease, are positioned between the lacOP promoter of E. coli and the GAL10 promoter of yeast directed in opposite orientations into the insert (Elledge et al., 1991). The yeast genomic DNA library (Nasmyth and Tatchell, 1980) from which plasmids pN9 and p N l l were isolated was made from a partial Sau3AI digest of yeast strain AB320 genomic DNA which was inserted into the tetracycline resistance gene of the plasmid pBR322 sequence carried within the plasmid vector YEpl3.
DNA Manipulations-Methods for plasmid and genomic DNA preparation, restriction enzyme digestion, and DNA ligations have been previously described (Maniatis et al., 1982;Ferbeyre et al., 1993). Prior to transformation, E. coli cells were washed with sterile water (resistance > 17 MR) and concentrated in 10% glycerol as described in the BTX (Biotechnologies & Experimental Research Inc.) protocol for E. coli. Concentrated cells were transformed by electroporation using a BTX Electroporation System at 10 KV/cm and resistance of 72 Q. Yeast strains were transformed with DNA using single-stranded sheared calf thymus DNA (Schiestl and Gietz, 1989) as carrier in the presence of lithium acetate.
DNA Sequencing-Single-stranded DNA sequencing of the yeast PSDl gene was performed by subcloning DNA fragments of a Hind111 digest of plasmid pN9 into phage M13mp18 by standard protocols (Sambrook et al., 1989). DNA sequencing using double-stranded DNA as a template was performed on plasmid pCC8. The dideoxy chain termination reaction using custom-tailored primers was performed as previously described (Sanger et al., 1977) following protocols outlined in the Sequenase@ Version 2.0 DNA Sequencing Kit. The reaction products were run on Long RangerTM polyacrylamide electrophoresis gels as in the protocol outlined by J. T. Baker Chemical Co. DNA sequence and putative amino acid sequence were analyzed by Genetics Computer Group Sequence Analysis Software Package, Version 7.2 (Genetics Computer Group, 1991), except for hydrophobicity profiles which were analyzed by DNA Strider (Marck, 1988).
Amplification of DNA by PCR-For both analytical and preparative purposes, PCR was performed after optimizing conditions as described by Innis and Gelfand (1990). Amplification of insert DNA from the AYES clones employed the DNA sequencing primers described by Elledge et al. (1991) which anneal to vector sequences adjacent to the cloning site. Primers 1 (5"CCAGTTAA-GAACGCCTTG-3') and 2 (5'-ATAACCTGTAACCCCTA-3)') used for amplification of yeast genomic sequences were chosen from the series of oligonucleotides developed during sequencing of the PSDl gene. DNA Labeling and Detection-The GeniusTM 1 Kit was used according to the manufacturer's directions for preparation and detection of nonradioactive DNA probes. The kit utilized random priming of template DNA and incorporation of digoxigenin-dUTP into the probe. Template DNA was produced by PCR and isolated by agarose (low melt) gel electrophoresis, excision of the desired band, and centrifugal filtration (0.45-pm filter unit). An antibody to digoxigenin coupled to alkaline phosphatase, which in the presence of Lumi-PhosTM 530 produces a chemiluminescent signal, was used to permit detection of hybridized probe by x-ray film.
Screening of the Genomic DNA Library-E. coli colonies bearing a yeast genomic DNA library carried on the E. coli yeast shuttle vector YEpl3 (Nasmyth and Tatchell, 1980) were transferred to positively charged nylon membranes and screened for hybridization to the labeled cDNA insert derived from plasmid pCC8. Transfer of colonies to membranes, hybridization, and development of blots were carried out using the manufacturer's instructions for use of positively charged Phosphatidylserine Decarboxylase in S. cerevisiae nylon membrane and the GeniusTM 1 Kit. SSC dilutions were prepared from 20 X SSC (3 M NaC1, 0.3 M sodium citrate, pH 7.0). Hybridization was performed overnight at 68 "C in hybridization solution (5 X SSC, 0.5% GeniusTM 1 Kit blocking reagent, 0.1% Nlauroyl-sarcosine, 0.02% SDS) containing the labeled cDNA probe (10 ng/ml). Following hybridization, membranes were washed twice for 5 min in 2 X SSC, 0.1% SDS at room temperature and twice for 15 min in 0.1 X SSC, 0.1% SDS at 68 'C. Colonies corresponding to the positive regions of the blot were picked and restreaked for single colonies. These were transferred to membranes and a second round of hybridization screening was used to identify the positive colonies. Southern Hybridization Analysis of Genomic DNA-Genomic DNA was digested with restriction enzymes and separated by agarose gel electrophoresis. DNA was then transferred to positively charged nylon membranes by capillary transfer using 20 X SSC, and the membranes were baked 2 h at 80 "C under vacuum. The labeled cDNA probe used for screening the genomic library was also used for hybridization to Southern blots. Methods for hybridization and development of blots were the same as those described for library screening.
Preparation of Cell Fractions and Measurement of Phosphutidylserine Decarboxylase Actiuity-All cell fractionation procedures were carried out at 4 "C. S. cereuisiae were grown to mid-log phase and the cells collected by centrifugation in tared containers. Wet weight of cells was calculated and the membrane fraction prepared as follows: cells were washed in 0.1 M Tris, pH 8.4, 0.25 M sucrose (2.5 ml/g cells), centrifuged, and resuspended in 50 mM Tris-HCI, pH 7.5, containing 0.3 M sucrose, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride (2 ml/g cells); following mechanical shearing in the presence of glass beads, unbroken cells and glass beads were removed by centrifugation at 1,500 X g. , for 15 min. The total particulate membrane fraction was separated from the cytoplasmic fraction by centrifugation at 100,000 X g. , for 1 h; the membrane pellet was suspended in 100 mM Tris-HC1, pH 7.5, containing 1 mM phenylmethylsulfonyl fluoride and 0.6 M sucrose (0.1 ml/g cells). Lysates of E. coli were prepared in 25 mM Tris-HC1, pH 8.2, containing 2.5 mM EDTA and 100 rg/ml lysozyme as described by Hawrot and Kennedy (1975) and assayed for phosphatidylserine decarboxylase activity after a 15-min preincubation at 30 or 42 "C as indicated. Phosphatidylserine decarboxylase activity was determined by the release of 14C0, from phosphatidyl-~-[l-"C]serine as previously described (Dowhan and Li, 1992) using the assay conditions reported for the enzyme from either yeast (Lamping et al., 1991) or E. coli (Dowhan and Li, 1992) as indicated. Protein concentration was determined by the protocol described in the BCA kit.
Complementation of E. coli Strain EH150 by Yeast cDNA-Complementation was carried out by a modification of the procedure described by Elledge et al. (1991). Briefly, E. coli strain EH150 (psd-2) was infected with phage AKC and grown on LB agar plates containing 40 pg/ml kanamycin at 30 "C. A colony of strain EH150/ AKC was grown overnight at 30 "C in LB medium containing 0.2% maltose and 10 mM IPTG, centrifuged at 3500 X g. , for 10 min, and resuspended in 0.5 volume of 10 mM MgS04. Then 4 X lo' phage from the AYES-yeast cDNA library were incubated with 5 X lo9 cells for 30 min at 30 'C without shaking. Five ml of M9 minimal medium supplemented with 0.2% mannitol and 1 mM IPTG were added, and the mixture was incubated for 2 h at 30 "C with shaking. Dilutions of cells were plated on either LB medium or LB medium containing ampicillin and IPTG and grown at 30 "C to assess cell viability or the number of total transformants, respectively. To select for transformants which complement the growth phenotype of strain EH150/AKC, 4 ml of the mixture was pelleted and resuspended in 0.4 ml of LB medium. Aliquots (0.1 ml) were plated on LB plates containing ampicillin and 1 mM IPTG and incubated at 42 "C.
Interruption of the Genomic Copy of the PSDl Gene-A 2.1-kb HindIII fragment from the genomic clone pN9, containing the open reading frame of the PSDl gene, was subcloned into plasmid pUC19. The single NcoI site within the gene was converted to a BglII site by ligation of the oligonucleotide 5"CATGAGATCT-3' to a NcoI-digested plasmid. The yeast LEU2 gene, isolated from plasmid YEpl3 by digestion with BgZII, was inserted into the newly created BglIJ site. The psd::LEU2 construct was used to transform E. coli strain JF1754 (kuB), and positive colonies were selected by their ability to grow at 37 'C on M9 medium supplemented with ampicillin (100 pg/ml) in the absence of leucine. In plasmids isolated from these colonies, the absence of an NcoI site and the presence of a 3-kb BglII fragment indicated the insertion of the LEU2 gene into the coding region of the PSDl gene. This construct was cut with HindIII to generate the 5.1-kb psdl::LEU2 gene which was used to transform the diploid yeast strain YPH501 (leu2/leu2). Potential chromosomally interrupted colonies were selected by growth on minimal defined medium lacking leucine.
Labeling and Analysis of Phospholipids-Five-ml aliquots of minimal defined media lacking leucine and either containing or lacking 1 mM ethanolamine were inoculated with the psdl::LEU2 haploid S. cereuisiae strain YCDlA or YCDlD to an Aero of approximately 0.05. Analogous 5-ml aliquots of culture medium containing leucine were inoculated with the PSDl haploid strain YCDlB or YCDlC. Follow-ing the addition of 50 pCi of [3ZP]orthophosphate, cells were grown for at least six generations to assure uniform labeling before harvesting by centrifugation at 1500 X gav. The cell pellets were suspended in 1 ml of 80% ethanol at 80 "C and incubated at 80 "C for 15 min to inactivate degradative enzymes (Henschke and Rose, 1991). After centrifugation at 1500 X gav, the ethanol supernatants were discarded and the pellets resuspended in 1 ml of chloroform/methanol/O.l N HC1 (1:2:0.8, v/v) containing 100 wg of a mixture of unlabeled carrier phospholipid (PC/PE/PS/PI, in the ratio of 15:10:2:2). Cells were lysed using glass beads, and the phospholipids were extracted as described by Homann et al. (1987). Isolated radiolabeled phospholipids were applied to boric acid-impregnated silica gel plates (Fine and Sprecher, 1982) which were developed in one dimension with chloroform/methanol/water/ammonium hydroxide (120:75:6:2) as the solvent system. Labeled phospholipids were detected and quantified directly from the thin layer plate using a Betascope (Betagen Corp.). The areas corresponding to phosphatidylinositol, phosphatidylserine, and phosphatidylethanolamine were scraped from the plate, eluted, and concentrated as described by Fine and Sprecher (1982) and applied to silica gel HPTLC plates. Plates were developed in the solvent system chloroform/methanol/acetic acid (65:25:10) or chloroform/methanol/formic acid (65:25:10).
One-ml cultures of E. coli were inoculated to AG,,,, < 0.01, grown at 30 "C in the presence of 5 pCi of [3ZP]orthophosphate for 1 h, then shifted to 42 "C for 2.5 h. Lipids were extracted using the method described by DeChavigny et al. (1991). Isolated radiolabeled phospholipids were applied to silica gel plates which were developed in one dimension with chloroform/methanol/water (65:25:4) as the solvent system. Labeled phospholipids were detected and quantified directly from the thin layer plate using a Betascope.

Complementation of Growth Phenotype of Strain EH150-
Incubation of 5 X lo9 EH150/XKC cells with 4 X lo7 phage from the AYES yeast cDNA library resulted in 2.5 x lo6 total transfectants, six of which grew at the nonpermissive temperature. Plasmids isolated from the six colonies were used to retransform strain EH150. All plasmids corrected the conditional lethal growth phenotype of strain EH150 indicating that in each case the correction was due to the presence of plasmid and not to a reversion event. The cDNA inserts ranged in size from 0.6 to approximately 5 kb. By restriction mapping, three of the inserts appeared to be the same and are represented by plasmid pCC8, resulting in a panel of four unique positive plasmids (pCC8, pCC9, pCC15, and pCC21).
As shown in Fig. lA, these four positive plasmids, when carried in strain EH150, elevated the level of phosphatidylserine decarboxylase activity (E. coli assay conditions) in cell extracts when compared to strain EH150 carrying no plasmid or carrying the control plasmid (V-pSE936). This result was independent of growth temperature of cells or preincubation temperature of the cell extracts. As expected, all cells grown at 30 "C and cell extracts preincubated at 30 "C showed phosphatidylserine decarboxylase activity; however, cells with the four positive plasmids showed overexpression of the decarboxylase activity. When the cells were grown at 30 "C and the cell extracts preincubated at 42 "C, all extracts showed some decrease in phosphatidylserine decarboxylase activity consistent with the loss of temperature-sensitive E. coli activity expressed at 30 "C. When cells were grown at 42 "C and preincubated at 30 "C, lysates from control cells displayed little phosphatidylserine decarboxylase activity, while lysates from cells carrying the positive plasmids showed significant phosphatidylserine decarboxylase activity. When cells were grown at 42 "C, preincubation of lysates at 42 "C completely abolished phosphatidylserine decarboxylase activity in the control lysates while the lysates from cells carrying the positive plasmids showed an increase in phosphatidylserine decarboxylase activity. These results are consistent with the presence of phosphatidylserine decarboxylase activities from the host strain as well as from the plasmid. Similar results A. duplicate 5-ml aliquots of LB medium were inoculated to A m < 0.01 with EH150 or with EH150 transformed with the indicated plasmids. Cultures were grown 1 h at 30 "C. Half the cultures were shifted to 42 "C and incubation of all cultures continued for 2.5 h. Extracts were prepared as described and preincubated at 30 or 42 "C for 15 min prior to measurement of phosphatidylserine decarboxylase activity at 30 "C under E. coli conditions. I , no temperature shift of culture; preincubation of extract at 30 "C. ZZ, no temperature shift of culture; preincubation of extract at 42 "C. ZIZ, shift of culture to 42 "C; preincubation of extract at 30 "C. ZV, shift of culture to 42 "C; preincubation of extract at 42 "C. B, 1-ml aliquots of LB containing 5 pCi of [32P]orthophosphate were inoculated and grown as described above except all cultures were shifted to 42 "C for 2.5 h. Phospholipids were isolated and analyzed as described.
were obtained when activity measurements were made under yeast assay conditions. Labeling of phospholipids with 32P04 also showed that the accumulation of phosphatidylserine and the decrease in phosphatidylethanolamine which occurs in control cells after a shift of growth temperature to 42 "C does not occur in cells with the positive plasmids (Fig. 1B). Therefore, all four plasmids corrected the temperature-dependent loss of viability, loss of decarboxylase activity, and increase in phosphatidylserine levels. Screening of the Genomic DNA Library-The insert derived from plasmid pCC8 was chosen as the probe used to screen a yeast genomic DNA library carried in E. coli. Plasmid pCC8 was selected from the panel of four positive clones because half of the original six colonies detected by complementation of the growth phenotype of strain EH150 contained this cDNA. The probe showed hybridization to a dot-blot of yeast genomic DNA (strain DL1) but not E. coli genomic DNA (strain EH150) confirming its yeast origin (data not shown). Southern hybridization analysis (using the same probe) of strain DL1 genomic DNA revealed two hybridization-positive DNA fragments after digestion with NcoI, EcoRI, or HpaI, but only one hybridization-positive DNA fragment after digestion with either XhoI or BglII (data not shown). These results were consistent with the restriction map of plasmid pCC8 shown in Fig. 24 and indicated a sequence in yeast genomic DNA that was structurally related to plasmid pCC8. Screening of the yeast genomic DNA library carried in E. coli was performed as described under "Experimental Procedures.'' From 20,000 colonies screened, two colonies (containing plasmids pN9 and pN11) were identified as potentially carrying a genomic copy of the yeast PSDl gene. Restriction mapping of the two clones suggested that they were identical. As shown in Fig. 2B, the restriction endonuclease map of plasmid pN9 reveals a HindIII fragment that contains the entire cDNA insert in plasmid pCC8.
Phosphatidylserine Decarboxylase Activity of Genomic Clones pN9 and pNll -Total membrane fractions from midlog cultures of yeast strain DL1 transformed with plasmids pN9 and p N l l were examined for phosphatidylserine decarboxylase activity under yeast assay conditions. Phosphatidylserine decarboxylase activity in strain DL1 (0.45 nmol/mg/ min) was increased 25-or 18-fold when carrying plasmid pN9 (11 nmol/min/mg) or plasmid p N l l (8.2 nmol/min/mg), respectively. Activity was linear with respect to time for at least 60 min in assays with 60-100 pg of membrane protein (data not shown). Since both plasmids exhibited identical restriction patterns (data not shown) and brought about a similar overproduction of enzymatic activity, they appeared to carry the same PSDl structural gene.
DNA Sequence of the PSDl Gene-Single-stranded sequencing of the genomic clone pN9 was performed as described under "Experimental Procedures." Both DNA strands were sequenced using overlapping primers as summarized in Fig. 2B. An open reading frame of 1500 bp corresponding to A.

B.
a predicted 500-amino acid protein (56,558 Da) was observed (Fig. 3). The region 5' of the open reading frame contained three sequences homologous to the consensus sequence 5'-ATGTGAAAT-3', which is postulated to confer inositol and choline regulation on several yeast phospholipid biosynthetic genes (Bailis et al., 1992;Kodaki et al., 1991b;Carman and Henry, 1989). Downstream of these sequences were two potential TATA promoter elements (TATAAA consensus sequence). Within the expected range of 40 to 120 bp downstream of the TATA promoter elements (Struhl, 1989), two sequences conformed to either consensus sequence RRYRR (where R = purines and Y = pyrimidines) or TC(A/G)A, which account for more than half of the known yeast transcription initiation sites (Guarente, 1987). Using either of these potential initiation sites, the predicted length of the 5'untranslated region would be consistent with the average of 52 nucleotides reported for many yeast genes (Cigan and Donahue, 1987). The common preference for adenine nucleotides was observed 5' to the ATG start codon, and the region surrounding the start codon loosely matched the yeast translation initiation site consensus sequence, (A/Y)A(A/ T ) A a T C T (Cigan and Donahue, 1987). In the 3"untranslated region, sequences similar to consensus sequence, (TAA)/ .TTT (Zaret and Sherman, 1982), and motifs, TAG. . . TATATA and TAG. . . TATGTA (Russo et al., 1991), postulated to be involved in transcription termination were found. Finally a polyadenylation site (Wickens and Stephenson, 1984) appeared near the end of the determined sequence. The genomic library containing the plasmid pN9 was generated from a partial Sau3AI digest of yeast genomic DNA (Nasmyth and Tatchell, 1980). Since there were no Sau3AI sites in the sequence left of the left-most HindIII site (Fig.  2B), the possibility that this region is comprised of randomly ligated yeast genomic DNA fragments from the partial Sau3AI digest is unlikely. The sequence of the 2.1-kb region between the two HindIII sites contained five Sau3AI sites, but also contained NcoI, EcoRI, and HpaI sites in complete agreement with the plasmid pCC8 restriction map. Further-  (Elledge et al., 1991). The restriction pattern was derived from restriction endonuclease digestion followed by agarose gel electrophoresis as described under "Experimental Procedures." B, the restriction map of the genomic DNA insert from plasmid pN9. The restriction pattern is based on sequencing information where possible, otherwise on restriction endonuclease digestion. The sequencing strategy is depicted by arrows representing DNA sequence obtained with each primer used. Except for the extreme flanking regions, both DNA strands including overlapping regions between primers were sequenced.   more, results of Southern and PCR analyses of yeast genomic DNA (see below) are consistent with the size and sequence reported here. Therefore, the entire sequenced region should correspond to a contiguous yeast genomic sequence. Partial double-stranded sequencing of the 5' and 3' ends of the pCC8 insert was performed using AYES sequencing primers (Elledge et al., 1991). Sequences identical to bp -53 to 89 and 1275 to 1555 of the genomic clone were found. Therefore, plasmid pCC8 contained the complete 1500-bp open reading frame encoding phosphatidylserine decarboxylase.

G T T A T A C A C~G A T T A C A C C T T A C C C C A G T T T T~A A C C T T T T~G C T A G T C T T A G A T A C C C~
The predicted amino acid sequence and its homology with the phosphatidylserine decarboxylase from CHO cells (Kuge et al., 1991) and E. coli (Li and Dowhan, 1988) are depicted in Fig. 4. From computer analysis of the protein sequence, the amino acid sequence from S. cereuisiae shared 44% identity and 60% similarity with the CHO cell enzyme and 35% identity and 55% similarity with E. coli enzyme, while the enzyme from CHO cells and E. coli were 33% identical and 53% similar. There were two distinct areas of high homology among the three enzymes. Both the amino-and carboxylterminal ends of the proteins showed significant homology, with little homology among the central domains of the enzymes. Hydrophobic analysis by the method of Kyte and Doolittle (1982) showed a very similar homology profile for all three proteins (Fig. 5). The significance of the homology at the amino-terminal end was not clear but homology at the carboxyl-terminal end included the highly conserved LG$T(V/I)(V/I) sequence corresponding to the post-translational processing site in E. coli which results in Ser-254 being converted to the pyruvoyl prosthetic group at the amino terminus of the a subunit of 7,332 Da (Li and Dowhan, 1988). Similar post-translational activation of the yeast enzyme would result in a pyruvoyl-containing a subunit of 4,192 Da and a @-subunit of 52,367 Da; the predicted size for the asubunit of the CHO cell enzyme is 3,547 Da (Kuge et ul., 1991). The pyruvoyl prosthetic group appears to lie in a hydrophobic environment in all three enzymes.
The amino-terminal sequence (57 amino acids) was examined for potential as a mitochondiral targeting sequence.
Computer analysis showed a weak a-helical but a large 8sheet hydrophobic moment (0.74) by the method of Eisenberg et al. (1984). Previous work has suggested that an a-helical hydrophobic moment is important and the preferred structure for a targeting structure, but a @-sheet hydrophobic moment also appears to be functional (Lemire et al., 1989;Schatz, 1987). In addition, this region exhibited an abundance of positively charged residues and an absence of either acidic amino acids, two adjacent "helical breakers," or extended hydrophobic regions (Lemire et al., 1989;Hart1 et al., 1989). Hence, the overall characteristics of this region are suggestive of a mitochondrial targeting sequence. Interruption of the Genomic Copy of the PSDl Gene-The diploid yeast strain YPH501 (leu2/leu2) was transformed with a linear DNA construct carrying the PSDl gene interrupted by insertion of the LEU2 gene as described under "Experimental Procedures." The resulting diploid strain YCDl was sporulated, and 20 sets of tetrads were isolated on dissection plates made of YPD medium. Four spores from each tetrad produced colonies. These colonies were streaked to complete inositol-free medium and to inositol-free medium without leucine. In addition, the colonies were streaked to analogous plates supplemented with ethanolamine and choline, to YPD medium or to YPGE medium. None of the 20 sets of tetrads demonstrated auxotrophy for ethanolamine or choline. Eighteen of the tetrads demonstrated 2:2 segregation of the leu2LEU2 markers as shown by growth of haploids on inositol-free media with and without leucine. Two of the tetrads showed 3:l (LEUZleu2) segregation and were not considered further. Those colonies prototrophic for leucine should represent cells derived from the spores carrying the interrupted psdl::LEU2 gene. All colonies prototrophic for leucine grew slowly on YPGE plates relative to the colonies auxotrophic for leucine. There was only a slight difference in growth on YPD plates between any of the colonies. Colonies from the four spores of one tetrad were also streaked to the following media: YPGal, YPL, YPG, and YPGE. In all cases, colonies from spores containing the psd1::LEUZ grew more slowly than colonies containing wild type PSDl gene.
Genomic DNA was prepared from four haploid cultures derived from one tetrad, Fig. 6A shows the results obtained when PCR was performed using genomic DNA from these haploid strains as template and the oligonucleotide primers 1 and 2. Primers 1 and 2 should produce a fragment extending from nucleotide 693 to 1164 of the sequence shown in Fig. 3. The resulting fragment should be 471 bp in length, the length of the PCR fragment observed when genomic DNA from strain YCDlB or YCDlC (PSDl, leu2) was used as template. Since primers 1 and 2 flank the NcoI site (nucleotides 1143-1148) which was used for the insertion of LEU2 gene into the PSDl sequence, the interruption of PSDl by LEU2 should result in a 3.5-kb PCR fragment, the length of the fragment  I   I I I I I I I I I I I I I I  I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I   IO0   200  300  CHO cell, and E. coli were used to generate hydrophobicity profiles by the method of Kyte and Doolittle (1982), where a positive number represents hydrophobic regions. The hydropathic index is plotted against amino acid position, using 11 amino acid windows. The amino acid positions for the CHO cell and E. coli enzymes have been shifted to match the homology plot (Fig. 4) to emphasize topological similarities. The arrows indicate the position of the serine residue which should be the precursor to the pyruvoyl prosthetic group. observed when genomic DNA from strain YCDlA or YCDlD (psdl::LEU2, leu2) was used as template. These results confirmed that the genomic copy of the PSDl gene has been interrupted by the LEU2 gene in strains YCDlA and YCDlD and that no other sequence existed in these strains which could be detected by PCR using primers 1 and 2. In a second experiment, genomic DNA was digested with HindIII and subjected to Southern hybridization analysis using the labeled insert of plasmid pCC8 as a probe. As shown in Fig. 6B, DNA from the strains YCDlB and YCDlC contained a 2.1-kb HindIII fragment which hybridized to the probe. In contrast, DNA from strains YCDlA and YCDlD contained a 5.1-kb HindIII fragment which hybridized to the probe. The 5.1-kb fragment corresponds to the wild type 2.1-kb fragment into which a 3-kb LEU2 fragment has been inserted.
Phosphatidylserine Decarboxylase Activity of Haploid Strains Derived from the Diploid Strain YCD1-Total membrane preparations were made and assayed for phosphatidylserine decarboxylase activity under yeast conditions as described under "Experimental Procedures." The phosphatidylserine decarboxylase activities for the two haploid strains carrying the wild type PSDl gene were linear with time up to 100 min and had an average specific activity over this time scale of 0.82 and 1.24 nmol/min/mg (Fig. 7A). The two haploid strains with the interrupted psdl::LEU2 gene showed no reproducible or convincing activity over the background level.
To more fully investigate the presence of residual activity %S.I kb 5.0 kb C Z . 0 kh FIG. 6. Analysis of genomic DNA from sibling haploid strains derived from strain YCD1. Sporulation of diploid strain YCDl (PSDI/psdI::LEU2, leu2) yielded leucine prototrophs YCDlA and YCDlD (psdZ::LEUZ) and leucine auxotrophs YCDlB and YCDlC (PSDI). Mobility of DNA standards is indicated to the right. A, PCR was performed as described under "Experimental Procedures" using DNA from each of the sibling haploid strains as template and primers 1 and 2 which flanked the site of LEU2 insertion into the PSDI gene. Products were run on an 0.7% agarose gel and visualized with ethidium bromide. B, the labeled insert of plasmid pCC8 was used to probe a HindIII digest of genomic DNA from each of the four haploids strains by Southern hybridization as described under "Experimental Procedures." and to rule out an inhibitor in the extract from mutant cells, membrane preparations from one of the psdl::LEU2 haploid strains was supplemented on a percent protein basis (w/w) with a membrane preparation from a PSDl haploid strain. This mixture of membrane preparations was assayed for 60 and 120 min (Fig. 7 B ) . Membrane preparations from the PSDl haploid strains exhibited a specific activity of 0.55 nmol/min/mg, which was comparable to wild type haploid strain YPH499 (0.24 nmol/min/mg). The membranes from the psdl::LEU2 haploid strains again showed no significant activity. The supplementation of the psdl::LEU2-derived membrane preparation with the PSDl -derived membrane preparation showed activity above background consistent with the amount of supplementation by wild type membrane preparations. Based on the above supplementation experiment, a residual activity of less than 5% of wild type would not have been convincingly detected. This experiment also ruled out an inhibitor of the phosphatidylserine decarboxylase activity in the mutant membrane preparations. No significant activity was observed in the cytosolic fraction of any of the four haploids (data not shown), ruling out the presence of a stable non-membrane associated form of the activity.
Phospholipid Composition of Haploid Strains Derived from the Diploid Strain YCDl-Labeling of phospholipids with 32P04 showed that even in the absence of detectable phosphatidylserine decarboxylase activity, the percentage of phospholipids represented by PS and PE did not change when the PSDl gene was interrupted by LEU2 (Table 11). PI and PC levels appeared to be slightly higher in cells bearing the interrupted PSDl gene. Rechromatography of recovered PI, PE, and PS on HPTLC plates in solvent systems which would Phosphatidylserine decarboxylase activity in haploid strains derived from diploid strain YCDl. All assays of membrane extracts contained 0.1 mg of protein/assay. The specific activities of radiolabeled phosphatidylserine in the assays for panels A and B were 133 disintegrations/min/nmol and 191 disintegrations/ min/nmol, respectively. A , total membrane preparations of P S D l (0 and 0 ) and psd1::LEUZ (m and 0) haploid strains were made and assayed as described under "Experimental Procedures." These experiments were repeated twice with similar results. B, the release of radiolabel (as disintegrations/min above a background of 240 k 1%) during the assay (60 min 0, 120 min 0) of total membrane preparations from a wild type control ( Y P H 4 9 9 ) and the above four haploid strains. The three columns on the right represent the release of radiolabel during assay of membrane preparations from one of the psd::LEU2 haploid strains supplemented on a percent protein basis (w/w) with a membrane preparation from a P S D haploid strain. Each sample was assayed in duplicate. These experiments were repeated with similar results. detect heterogeneity in the phospholipid spots confirmed the identity of the phospholipids and the absence of any comigrating phospholipids. When the cell cultures were grown in minimal defined medium supplemented with ethanolamine, the changes in phospholipid concentration were not great and did not appear to correlate with the function of the PSDl gene.

DISCUSSION
The evidence presented supports the cloning and disruption of the gene encoding the major phosphatidylserine decarbox- Combination of several minor radioactive spots; includes PA.
ylase activity thus far identified in yeast. The complementation of the E. coli mutant, the sequence homology with the E. coli enzyme, and the overproduction in yeast of phosphatidylserine decarboxylase activity by multiple copies of the genomic clone establish that a gene encoding yeast phosphatidylserine decarboxylase activity was isolated rather than a gene responsible for regulation of PSDl gene expression. The genetic analysis of haploid strains following sporulation of the PSDl/psdl::LEU2 strain, the Southern hybridization and PCR analysis of the genomic DNA from the haploid strains, and the lack of phosphatidylserine decarboxylase activity in the two haploids prototrophic for leucine support the disruption of the gene encoding the major activity.
Our assay conditions could have detected at least 5% of normal phosphatidylserine decarboxylase activity if it were present in the extracts from the disrupted haploid strains. There was no evidence of any activity in either the total membrane or cytoplasmic fractions which eliminates a mistargeted fragment of the enzyme or the presence of a nonmembrane-associated form of the enzyme; given what is known about the enzymological properties of the closely related enzyme from E. coli (Warner and Dennis, 1975), it is unlikely that the enzyme could catalyze decarboxylation if it did not have affinity for membranes or detergent micelles. No indication of an inhibitor of decarboxylase activity was found in the mutant extracts. Therefore, the fact that disruption of this gene in haploid strains did not result in ethanolamine auxotrophy, elimination of phosphatidylethanolamine, or the accumulation of phosphatidylserine was unexpected. The possibility exists that the disrupted gene could still express a truncated enzyme from a new promoter originating within the LEU2 insert. However, partial expression of the PSDl gene is an unlikely explanation for the phenotype of our mutant since a deletion of the PSDl gene including the region encoding the putative catalytic subunit (Trotter et al., 1993) does not result in either ethanolamine or choline auxotrophy. Since phosphatidylserine levels did not dramatically increase and phosphatidylethanolamine levels were normal in cells with no exogenous supply of ethanolamine, the most likely explanation for these results is a second phosphatidylserine decarboxylase activity which is below the level of detection of our in vitro assay.
The possibility of two genes encoding phosphatidylserine decarboxylase activity is supported by our isolation of several other cDNAs which complement the E. coli mutant, but which show little homology with the insert in plasmid pCC8. There is precedence in yeast for a minor isozyme of a subunit of the inner mitochondrial membrane-associated cytochrome oxidase which can provide sufficient activity for mitochondrial function (Cumsky et al., 1985). CDP-diacylglycerol synthase activity, which is required for synthesis of phospholipids both in the endoplasmic reticulum and the mitochondria, is also found associated with both of these organelles  although it is not known if these activities are encoded by the same or different genes. Therefore, the question of whether a phosphatidylserine decarboxylase activity is essential in yeast still remains unresolved, but our results explain why it has been difficult to isolate a mutant in this activity based on screening for a conditional lethal phenotype or ethanolamine auxotrophy resulting from a mutation in phosphatidylserine decarboxylase activity.
We have used complementation of a conditional lethal mutation in E. coli phospholipid biosynthesis to clone a gene responsible for encoding the same activity in yeast mitochondria. This approach is not novel for cloning eukaryotic genes. Yeast genes encoding lipid biosynthetic activities associated either with the endoplasmic reticulum such as phosphatidylinositol synthase (Nikawa et al., 1988) and squaline synthase (Jennings et al., 1991) or the cytoplasm such as cholinephosphate cytidylyltransferase (Tsukagoshi et al., 1991) and choline kinase (Hosaka et al., 1989) have been functionally expressed in E. coli. However, successful application of our approach to mitochondrially localized, membrane-associated enzymes of phospholipid metabolism has implications beyond this work. Given the membrane targeting and post-translational events necessary to assemble a functional enzyme into the inner mitochondrial membrane (Glick and Schatz, 1991), it is somewhat surprising that complementation in E. coli occurred. Plasmid pCC8 contains a full-length cDNA including a putative mitochondrial targeting sequence. Either this sequence does not function as predicted or does not interfere with enzymatic activity in E. coli, or the protein is properly processed in E. coli; alternatively, the targeting sequence is not removed in yeast. Secondly, the additional postulated complex post-translational modification of the yeast protein necessary for formation of the pyruvate prosthetic group also must occur in E. coli. Either this event is autocatalytic as suggested for other pyruvoyl enzymes (van Poelje and Snell, 1990) or other gene products required for modification also recognize the highly homologous yeast enzyme. Finally, extension of this successful approach seems warranted for the cloning of other genes responsible for phospholipid metabolism in the mitochondria. Mutations, along with the respective cloned genes, of many of the yeast enzymes of phospholipid metabolism localized to the endoplasmic reticulum are currently available (Carman and Henry, 1989); however, the isolation of mutants in the gene products localized to the mitochondria and cloning of these genes have met with considerable difficulty. Since mutants are available for E. coli in all steps of anionic phospholipid metabolism starting with the formation of lysophosphatidic acid and resulting in the formation of phosphatidylglycerol and cardiolipin , a similar approach as used for the PSDl gene has potential in cloning several genes of central importance to mitochondrial phospholipid metabolism.
The high sequence homology among the three phosphatidylserine decarboxylases from CHO cells, yeast, and E. coli is noteworthy and follows a similar finding for t h e p y r u~~y ldependent histidine decarboxylases, but not the pyruvoyldependent S-adenosylmethionine decarboxylases where the E. coli enzyme shows no homology with either the yeast or CHO cell enzyme (van Poelje and Snell, 1990). Given the sequence homologies, the eukaryotic phosphatidylserine decarboxylases most likely undergo a similar post-translational activation step as does the E. coli enzyme. It is not clear whether this event occurs before or after membrane assembly, but in E. coli a mutant phosphatidylserine decarboxylase which cannot undergo activation (Ser-254 replaced by alanine) or mutants in which activation is very slow (Ser-254 replaced by cysteine or threonine) accumulate unprocessed enzyme in the membrane and not in the supernatant (Li and Dowhan, 1990). This would suggest that membrane assembly does not require enzyme activation, and membrane assembly may even be required for activation. In the histidine decarboxylases, the activation event requires a folded proenzyme form (van Poelje and Snell, 1990). Since translocation of proteins targeted to the inner mitochondrial membrane occurs while the precursor is in a largely an unfolded state (Glick and Schatz, 1991) and activation of the decarboxylase results in a heterologous dimer as the minimum catalytic unit, the activation step probably occurs after translocation and assembly.
Now that a psdl mutant and a PSDl gene of yeast are available, several future experiments are possible. First, the issue of the mechanism by which phosphatidylethanolamine is synthesized in the absence of the major phosphatidylserine decarboxylase and ethanolamine supplementation must be resolved. Further characterization of the remaining cDNAs which correct the E. coli mutant may aid in answering this problem. Interesting questions concerning the mechanism of targeting and assembly of this inner mitochondrial membrane enzyme can be addressed. Further resolution of the genetics and biochemistry of this step in phospholipid metabolism will shed light on the synthesis and regulation of phospholipids in both the mitochondria and throughout the cell, particularly if two decarboxylases exist. Finally, a more direct approach to the complex problem of phosphatidylserine/phosphatidylethanolamine trafficking between the endoplasmic reticulum and the mitochondria can be formulated.