Identification of Gene Encoding Plasmodium knowlesi Phosphatidylserine Decarboxylase by Genetic Complementation in Yeast and Characterization of in Vitro Maturation of Encoded Enzyme*

Background: Functional screening of malaria (P. falciparum) genes is a problem because of A + T content. Results: Use of a P. knowlesi cDNA library and a yeast mutant identifies an important parasite cDNA/gene and reveals new regulation of lipid synthesis. Conclusion: The P. knowlesi library will enable extrapolation to P. falciparum. Significance: Functional screening of malaria genes is now greatly enhanced. The 23-megabase genome of Plasmodium falciparum, the causative agent of severe human malaria, contains ∼5300 genes, most of unknown function or lacking homologs in other organisms. Identification of these gene functions will help in the discovery of novel targets for the development of antimalarial drugs and vaccines. The P. falciparum genome is unusually A + T-rich, which hampers cloning and expressing these genes in heterologous systems for functional analysis. The large repertoire of genetic tools available for Saccharomyces cerevisiae makes this yeast an ideal system for large scale functional complementation analyses of parasite genes. Here, we report the construction of a cDNA library from P. knowlesi, which has a lower A + T content compared with P. falciparum. This library was applied in a yeast complementation assay to identify malaria genes involved in the decarboxylation of phosphatidylserine. Transformation of a psd1Δpsd2Δdpl1Δ yeast strain, defective in phosphatidylethanolamine synthesis, with the P. knowlesi library led to identification of a new parasite phosphatidylserine decarboxylase (PkPSD). Unlike phosphatidylserine decarboxylase enzymes from other eukaryotes that are tightly associated with membranes, the PkPSD enzyme expressed in yeast was equally distributed between membrane and soluble fractions. In vitro studies reveal that truncated forms of PkPSD are soluble and undergo auto-endoproteolytic maturation in a phosphatidylserine-dependent reaction that is inhibited by other anionic phospholipids. This study defines a new system for probing the function of Plasmodium genes by library-based genetic complementation and its usefulness in revealing new biochemical properties of encoded proteins.

Malaria, a serious infectious disease responsible for over 800,000 deaths annually, is caused by intraerythrocytic proto-zoan parasites of the genus Plasmodium. Plasmodium falciparum is the major cause of human parasitic fatalities. The worldwide emergence of drug-resistant Plasmodium strains has made treatment of malaria increasingly difficult, thus emphasizing the need for new chemotherapeutic strategies to combat this disease (1). Following invasion of red blood cells, malarial parasites must increase lipid synthesis for membrane biogenesis and cell division. Inhibition of membrane lipid synthesis provides an attractive target for antimalarial chemotherapy (2,3). Lipid analysis of P. falciparum demonstrates a relatively high content (up to 35% of total phospholipid) of phosphatidylethanolamine (PtdEtn) 2 (4), which often functions as a nonbilayer hexagonal phase lipid (5,6).
The PtdEtn synthetic pathways in Saccharomyces cerevisiae share some similarity with those of the parasite (see Fig. 1). In the absence of exogenous Etn, the major route for the synthesis of PtdEtn originates with the synthesis of PtdSer in the endoplasmic reticulum or closely related membranes (mitochondrion-associated membrane) by PtdSer synthase (11)(12)(13). After its synthesis, PtdSer is decarboxylated to form PtdEtn by PtdSer decarboxylase 1 (PSD1) at the inner mitochondrial membrane (14) or PtdSer decarboxylase 2 (PSD2) at the Golgi (14,15). In the presence of Etn, PtdEtn is synthesized through the CDPethanolamine pathway. Unlike Plasmodium species, S. cerevisiae does not decarboxylate serine to form Etn. Previously, we have successfully used S. cerevisiae as a surrogate system to characterize P. falciparum genes and cDNAs involved in the synthesis and transport of lipid precursors (3,16,17). However, these studies required optimization of cDNAs because of the high A ϩ T content of P. falciparum genes (ϳ81%) (18). In contrast to P. falciparum, yeast have an A ϩ T content of 62%. To bypass the need for codon optimization and enable analysis of Plasmodium genes of unknown function, we constructed a cDNA library from P. knowlesi, which has an A ϩ T content of 63% (19). P. knowlesi causes malaria in both monkeys and humans (19). The goals of this study were to use the P. knowlesi cDNA library to complement well defined genetic defects in yeast lipid synthesis and characterize the parasite gene product.
This report demonstrates successful library-based complementation of yeast using the cDNA library and reveals new details about the properties and maturation of PkPSD.

EXPERIMENTAL PROCEDURES
Materials-All chemicals, including amino acids for yeast media, were purchased from either Sigma or Fisher. Other components for yeast growth media were purchased from Difco. Phospholipids were obtained from Avanti Polar Lipids. Silica gel H plates and Silica gel 60 plates were purchased from Analtech Corp. and EMD, respectively. Radioactive L-[G-3 H]serine was from PerkinElmer Life Sciences. Reagents for protein determination were from Bio-Rad. Pre-cast SDS-polyacrylamide gels were purchased from Invitrogen. Mouse monoclonal antibodies against the V5 and His 6 epitope tags of the PkPSD fusion protein were obtained from Invitrogen and Clontech, respectively. Other reagents used for ligand blotting were obtained from Bio-Rad and Sigma.
cDNA Library-The cDNAs were generated from P. knowlesi mRNAs and inserted by directional cloning into a modified pBEVY vector (20), pBEVY-DS, containing a single SfiI site. The vector is a URA3-based multicopy Escherichia coli/yeast shuttle vector in which the cloned Pk-cDNAs are under the regulation of ADH1 promoter.
Cell Growth-psd1⌬psd2⌬dpl1⌬ strains harboring the PkPSD gene were cultured in SC-U or synthetic lactate uracil dropout (SL-U) medium (22) supplemented with 2 mM Etn. After washing twice with water, serial 5-fold dilutions of cells were plated onto the medium with or without 2 mM Etn. Plates were incubated at 30°C for 2-4 days.
Whole Cell Radiolabeling and Phospholipid Analysis-psd1⌬psd2⌬dpl1⌬ strains harboring the predicted PkPSD sequence and psd1⌬psd2⌬dpl1⌬ strains with empty vector were grown in synthetic complete medium plus 2 mM Etn with glucose as a carbon source (SC). Cells in mid-log phase were harvested by centrifugation and washed twice by resuspension in water and recentrifugation. The cells were suspended in SC medium at an A 600 of 0.35 in a volume of 2 ml. Radiolabeling was initiated by adding 10 Ci/ml L-[G-3 H]serine, and growth was continued at 30°C for 2 h with vigorous shaking. Labeled phospholipids were extracted as described previously (14,23), and the lipid classes were resolved by thin layer chromatography, and radioactivity was quantified by liquid scintillation spectrometry.
Measurement of PSD Activities-Cell-free extracts, membrane fractions, and soluble fractions were isolated from psd1⌬psd2⌬dpl1⌬ strains harboring the predicted PkPSD cDNA or empty vector. Endogenous yeast PSD1 activity was measured using MSY30 (PSD1DPL1psd2⌬) (24). Strains were FIGURE 1. PtdEtn synthesis pathways in Plasmodium and yeast. PtdEtn synthesis in Plasmodium occurs through two major pathways, the serine decarboxylase-CDP-ethanolamine pathway and the PtdSer decarboxylation pathway. The major enzymes executing reactions within these pathways are shown in boxes. In the serine decarboxylase-CDP-ethanolamine pathway, serine is decarboxylated by an unidentified serine decarboxylase (SD) to form ethanolamine. Serine decarboxylation is unique to malarial parasites and plants. Ethanolamine formed through this reaction is sequentially converted into phosphoethanolamine (P-Etn), CDP-ethanolamine (CDP-Etn), and PtdEtn by ethanolamine kinase (EK), ethanolamine-phosphate cytidylyltransferase (ECT), and ethanolamine phosphotransferase (EPT). In Plasmodium, P-Etn can be sequentially methylated to form phosphomonomethylethanolamine (P-MMe), phosphodimethylethanolamine (P-DMe), and phosphocholine (P-Cho) by a single phosphoethanolamine methyltransferase (PMT). The resultant phosphocholine is further metabolized by a choline-phosphate cytidylytransferase (CCT) and choline phosphotransferase (CPT) to produce CDP-Cho and PtdCho, respectively. In the PtdSer decarboxylation pathway, serine is incorporated into phosphatidylserine (PtdSer) by PtdSer synthase (PSS) and subsequently decarboxylated to form PtdEtn by PtdSer decarboxylase (PSD). In contrast to Plasmodium, yeast sequentially methylate PtdEtn via the action of the phospholipid methyltransferases, PEM1 and PEM2, to produce PtdCho. The black arrows show the parasite pathways, and the gray arrows show the yeast pathways. The enzymes catalyzing each step of the pathway are enclosed by boxes.
homogenized by glass bead beating, and subcellular fractions were prepared by differential centrifugation. The assay for PSD activity utilized Ptd[1Ј-14 C]Ser as the substrate, and the reaction product was trapped as 14 CO 2 on 2 M KOH-impregnated filter paper, as described previously (25). PSD activities in psd1⌬psd2⌬dpl1⌬ strains harboring the predicted PkPSD cDNA are presented as enzyme-specific activity (nmol/ mg⅐protein/45 min), or as the percentage of activity relative to the activity of yeast PSD1 wild type strains (100%), or as volume-normalized activity (nmol/ml/45 min) in TNT reactions. All enzyme reactions were performed at substrate (0.2 mM) and protein concentrations that produced a linear response with the amount of enzyme added to the reaction (usually 4 -20 g). In the most active preparations, the maximum substrate conversion to product did not exceed 10%.
Lipid Phosphorus Measurement-Yeast strains were cultured in SC-U medium and grown to mid-log phase at 30°C. The cells were harvested by centrifugation and washed twice with water. The lipids were extracted, as described previously (26). The phospholipids were separated by two-dimensional thin layer chromatography (TLC) on Silica 60 plates using chloroform/ methanol/ammonium hydroxide (65:35:5 v/v) followed by chloroform/acetic acid/methanol/water (75:25:5:2.2 v/v). Lipids were visualized with iodine vapor and quantified by measuring phosphorus (27). The results are shown as the percentage of total lipid phosphorus in each phospholipid fraction. Data are means Ϯ S.D. for three or four independent experiments.
In Vitro Expression of PkPSD-A TNT quick-coupled transcription/translation system (Promega Corp.) was used to express PkPSD proteins in vitro. Reactions were initiated with 0.5 l of 1 mM methionine and 2 l of plasmid (0.5 g/l) harboring PkPSD with an N-or C-terminal His 6 tag, added to 20 l of TNT quick master mix kit, which contained rabbit reticulocyte lysate, an amino acid mixture without methionine, T7 RNA polymerase, nucleotides, salts, and RNasin ribonucleotide inhibitor. To study the effect of phospholipids on the expression and processing of the PkPSD protein, 2 l of liposomes composed of dioleoyl phosphatidylserine (DOPS), dioleoyl phosphatidylethanolamine (DOPE), dioleoyl phosphatidylcholine (DOPC), dioleoyl phosphatidylglycerol (DOPG), dioleoyl phosphatidic acid (DOPA), or phosphatidylinositol (PI) from bovine liver or soybean were added to the TNT reaction and incubated for the indicated times at 30°C. Liposomes were prepared fresh for each experiment. To prepare liposomes, phospholipid in chloroform was transferred into an Eppendorf tube and dried under nitrogen gas and chased with methanol. The lipid pellets were resuspended in 20 l of methanol and dried using centrifugation under vacuum for 30 min to completely remove all organic solvents. The lipid pellets were resuspended in 0.1 M KCl, 10 mM Tris-Cl, pH 7.5, at 1 mg/ml, hydrated at 37°C for 30 min, and then mixed with a vortex mixer to create multilamellar liposomes. To create unilamellar liposomes, the suspension was bath-sonicated for 20 min. Defined size unilamellar liposomes were made by passing multilamellar liposomes through an Avanti LiposoFast with 50-, 200-, or 1000-nm filter sets. When the in vitro transcription/ translation step was temporally separated from processing steps, 0.2 mM cycloheximide was added to the TNT reaction to arrest translation, and the reactions were further incubated with liposomes. The expression and processing of PkPSD were monitored by Western blot analysis using anti-His 6 antibody and by PSD enzyme assay as described above.
Construction of Vectors to Express PkPSD in E. coli or Yeast-E. coli vectors harboring plasmids for PkPSD-His 6 or N-terminal 34-amino acid deleted PkPSD-His 6 (PkPSD⌬34-His 6 ) were created using a pET-45b(ϩ) vector. Briefly, specific primers for the individual constructs were generated and used to amplify DNAs from a template cDNA harboring PkPSD using PCR with Pfu-DNA polymerase (Stratagene). The PkPSD and PkPSD⌬34 constructs containing a 5Ј SpeI site and a 3Ј and XhoI site were purified by agarose gel electrophoresis. The pET-45b(ϩ) E. coli expression vector was digested with XbaI (compatible to SpeI ligation) and XhoI restriction enzymes, and the resultant DNA fragment was purified by electrophoresis. Appropriate ligation reactions yielded pET45-PkPSD-His 6 and pET45-PkPSD⌬34-His 6 . The plasmids were introduced in BL21 and Rosetta strain to express the constructs. The pET-45b(ϩ) vector was also utilized to create pET45-His 6 -PkPSD⌬34, which encodes an N-terminal His 6 epitope linked to PkPSD⌬34. The PCR construct, PkPSD⌬34, was digested with KpnI at the 5Ј end and XhoI at the 3Ј end and subsequently ligated into the linearized pET-45b(ϩ) vector, also cut with KpnI and XhoI restriction enzymes. The construct was introduced into BL21 and Rosetta strains to express the enzyme. The pET45-His 6 -PkPSD⌬34 construct was also used in vitro transcription/translation using a TNT kit.

Identification of P. knowlesi Phosphatidylserine Decarboxylase-An
Etn auxotrophic mutant yeast strain devoid of PSD activity and incapable of forming P-Etn from sphingolipids, HKY44 (psd1⌬psd2⌬dpl1⌬), was utilized to screen a multicopy Pk-cDNA library inserted into the vector pBEVY-DS. Approximately 650,000 uracil prototrophic pBEVY-DS library transformants were screened for Etn prototrophy, and 127 transformants were identified. Plasmids from nine uracil/ethanolamine yeast prototrophs were recovered using antibiotic resistance in E. coli. Sequencing of these plasmids revealed the presence of the ORF PKH_072580. The sequence was previously annotated as a putative phosphatidylserine decarboxylase. The remaining Etn prototrophs were also positive for PkPSD sequences when screened by PCR.
The putative PkPSD cDNA encodes a polypeptide of 354 amino acid residues with a deduced molecular weight of 41,524. The PkPSD protein shares strong sequence homology with the previously characterized PfPSD from P. falciparum with amino acid sequence identity of 72.5% and similarity of 88.2%. An MGSS sequence located at positions 306 -309 in the C terminus corresponds to a VGSS sequence located at positions 314 -317 of the PfPSD and appears related to the LGST motif, which is the endoproteolytic cleavage site of the E. coli and the S. cerevisiae PSD1 proenzymes ( Fig. 2) (14,28). The cleavage converts the inactive proenzyme to an active form containing a small ␣ subunit with a pyruvoyl prosthetic group essential for catalysis and a large ␤ subunit (29). Fig. 3a shows the growth of the psd1⌬psd2⌬dpl1⌬ strains harboring the PkPSD cDNA in minimal medium containing glucose. In the absence of Etn supplementation, the psd1⌬psd2⌬dpl1⌬ strain harboring the PkPSD cDNA grew with a pronounced lag but reached nearly the same saturation level after 2 days. Fig. 3b shows the growth on minimal glucose plates with or without Etn supplementation. The psd1⌬psd2⌬dpl1⌬ strains harboring the PkPSD cDNA grew robustly in the absence of Etn, whereas strains with empty plasmid vector failed to grow.
Next, we tested whether the presence of the PkPSD cDNA could support growth of the mutant strain under respiratory conditions in minimal medium with lactate as the carbon source, where PtdEtn synthesis is required for normal mitochondrial function. Fig. 3c shows that psd1⌬psd2⌬dpl1⌬ strains harboring PkPSD grew well under respiratory conditions in the absence of Etn. The ability of the PkPSD enzyme to support yeast growth under respiratory conditions indicates that the enzyme has access to pools of PtdSer located at the outer mitochondrial membrane. Collectively, the above data demonstrate that the Plasmodium PSD cDNA suppresses the growth defect of psd1⌬psd2⌬dpl1⌬ strains under both fermentative and respiratory conditions.
Characterization of PkPSD Expressed in Yeast-To characterize the PkPSD enzyme, catalytic activities were assayed both in vivo and in vitro. Yeast strains were labeled with [ 3 H]serine in minimal glucose medium, and the incorporation of the radiolabel into aminophospholipids was analyzed by thin layer chromatography (TLC) of the lipid extracts prepared from the cells (Fig. 4a). The mutant psd1⌬psd2⌬dpl1⌬ strain harboring an empty vector failed to generate significant levels of radiolabeled PtdEtn because of the lack of PtdSer decarboxylases. The mutant psd1⌬psd2⌬dpl1⌬ strain expressing the PkPSD, and yeast strains lacking PSD2 but containing the PSD1 gene, expressed from either single copy or multicopy plasmids, readily produced radiolabeled PtdEtn at normal levels. This indicates that nascent PtdSer is efficiently converted to PtdEtn by the PkPSD enzyme at levels comparable with that produced by the chromosomal copy of yeast PSD1.
Next, the enzyme activity was measured in cell extracts. Whole cell extracts, membrane fractions, and soluble fractions were incubated with Ptd[1Ј-14 C]serine substrate, and the decarboxylase activity was determined by measuring 14 CO 2 production. As shown in Fig. 4b, high enzyme activity was detected in the cell-free extracts of the psd1⌬psd2⌬dpl1⌬ strain harboring a PkPSD cDNA, whereas no activity was found in the psd1⌬psd2⌬dpl1⌬ strain harboring the empty vector. The levels of PkPSD expression produced nearly three times the catalytic activity of the endogenous yeast PSD1 gene. Interestingly, significant levels (49%) of PkPSD enzyme activity were detected in the soluble fractions prepared from cell extracts (Fig. 4c). Thus far, all eukaryotic PSD enzymes have been localized to membrane compartments (30 -32), and these findings suggest unusual properties of the malarial enzyme. The relatively large soluble population of the PkPSD protein could be an intrinsic feature of the enzyme that renders the molecule amphitropic.
The phospholipid compositions of the psd1⌬psd2⌬dpl1⌬ strains with the PkPSD cDNA were analyzed and compared with that of a psd2⌬ strain that contains wild type PSD1 and DPL1 genes. Lipids extracted from the cells grown on medium in the absence of Etn were separated by two-dimensional TLC and visualized by iodine staining. The appearance of a PtdEtn spot indicates that the lipid was synthesized by the action of PkPSD enzyme (Fig. 5a). Fig. 5b shows the quantification of each lipid extracted from the TLC plates. There was a slight increase in PtdEtn level and a 42% reduction in PtdSer in the psd1⌬psd2⌬dpl1⌬ strains expressing the PkPSD cDNA compared with that in the psd2⌬ strain. Taken together, the results from Figs. 4 and 5 demonstrate that the PkPSD enzyme expressed in the psd1⌬psd2⌬dpl1⌬ strains was fully functional and complemented the biochemical defect of the mutant strain.

N-terminal Deletion of PkPSD Yields Active Forms of Enzyme-
The N-terminal 137 amino acids of yeast PSD1 contain mitochondrial targeting and inner membrane sorting sequences. The sequence is not only required for the mitochondrial targeting but also for the correct processing into ␤ and ␣ fragments (33). Previously, plant PSD sequences from tomato and Arabidopsis were identified and expressed in the yeast psd1psd2 double mutant strain (10). The N-terminal mitochondrial targeting sequence of plant PSD protein was inhibitory to its functional expression in yeast. A deletion construct that removed the plant-specific mitochondrial targeting sequence and a chimeric construct in which the yeast PSD1 targeting sequence replaced the plant sequence could readily suppress the Etn auxotrophy of the yeast double mutant strain.
The N-terminal 55-amino acid sequence of the PkPSD protein does not show any significant amino acid homology with any other PSDs. When the sequence was analyzed with TMpred software, a trans-membrane domain was predicted at positions 18 -34. To investigate the role of the N-terminal sequence, two truncated PkPSD cDNAs were constructed to encode proteins with deletions of residues 2-34 (PkPSD⌬34) and 2-55 (PkPSD⌬55) (Fig. 6a). The constructs were placed under control of the GAL1 promoter in an episomal vector (pYES2.1, Invitrogen) and introduced into the psd1⌬psd2⌬ double mutant strain. Enzyme assay showed that the PkPSD with a 34-amino acid deletion produced robust activity, whereas the PkPSD with a 55-amino acid deletion produced very weak activity (Fig. 6b). The catalytic activity of the PkPSD with the 34-amino acid deletion was also evenly distributed between membrane and soluble fractions (data not shown). This indicates that the PkPSD residues between 2 and 34 were not required for enzyme maturation, catalysis, or membrane association in the yeast expression system. In contrast, the amino acids between positions 35 and 55 appear crucial for the functional activity of PkPSD.
Expression of PkPSD Fusion Proteins in Bacteria and Yeast-To further characterize the activity and processing of PkPSD, we generated PkPSD-His 6 and PkPSD-V5 fusion proteins expressed in E. coli and yeast, respectively. A PkPSD-His 6 fusion protein was expressed in BL21 and Rosetta strains of E. coli. Protein extracts were separated on SDS-PAGE, and the fusion protein was detected with anti-His 6 antibody. As seen in  Fig. 7a, the fusion protein was detected only with the Rosetta host strain, which provides six tRNAs for codons that are rarely used in E. coli but common in eukaryotic organisms, thus enabling the efficient translation of DNA from P. knowlesi. Two immunoreactive bands were detected with the anti-His 6 antibody. The upper 43-kDa band corresponds to the unprocessed proenzyme and the 7-kDa lower band corresponds to the ␣ subunit containing the active site of the enzyme (28). The majority of the E. coli-expressed PkPSD was in the form of proenzyme, which indicates that the endoproteolytic processing of the PkPSD fusion protein was not efficient. The Rosetta strain expresses its endogenous PSD, but the total catalytic activity in cell extracts increased 2-fold with PkPSD cDNA expression (Fig. 7b).
A multicopy yeast vector encoding a PkPSD-V5 fusion protein under control of the GAL1 promoter was constructed and transformed into a yeast psd1⌬psd2⌬ strain. Fig. 8a shows the growth of the psd1⌬psd2⌬ strains harboring either an empty vector, a multicopy vector with the yeast PSD1, or a multicopy vector containing the PkPSD-V5 cDNA fusion construct. As expected, the psd1⌬psd2⌬ strains harboring empty vector failed to grow in the absence of Etn, because the strains had no PSD enzymes. Expression of the PkPSD fusion protein readily suppressed the growth defect of the mutant strain in the absence of Etn. Extracts from the yeast cells were analyzed by SDS-PAGE followed by Western blotting using an anti-V5 antibody (Fig. 8b). Two protein bands, corresponding to the proenzyme and the active form of the ␣ subunit, were detected demonstrating the maturation of the enzyme in vivo. These findings . PkPSD is a functional PtdSer decarboxylase enzyme. a, incorporation of radioactive serine into amino glycerophospholipids. Labeled phospholipids were extracted from log phase cells grown for 2 h in synthetic medium containing 10 Ci/ml L-[G-3 H]serine. The lipid classes were resolved by thin layer chromatography, and radioactivity was quantified by liquid scintillation spectrometry. Data are the mean Ϯ S.D. from 2 experiments performed in duplicate. Results are the percentage of total radiolabel incorporated into each phospholipid. The trace labeling in PtdIns is found primarily in fatty acids and co-migrating sphingolipids. b, PSD enzyme assays were performed with cell-free extracts from a parental strain (psd2⌬), a psd1⌬psd2⌬dpl1⌬ mutant strain harboring either an empty vector, or pBEVY-PkPSD. c, PSD enzyme assays were performed with membrane (black bar) and soluble (gray bar) fractions from a parental strain (psd2⌬) and a psd1⌬psd2⌬dpl1⌬ mutant strain harboring pBEVY-PkPSD. Assay for PSD utilized Ptd[1Ј-14 C]Ser as the substrate, and the reaction product was trapped as 14 CO 2 on 2 M KOH impregnated filter paper. Data are means Ϯ S.D. for two experiments each performed in duplicate. with the fusion proteins demonstrated that epitope-tagged versions of PkPSD could undergo processing in vivo and maturation to active enzyme.
In Vitro Processing of PkPSD Demonstrates PtdSer Post-translationally Enhances Enzyme Maturation-The retention of enzyme activity by PkPSD⌬34 (see Fig. 6b) indicated that the predicted membrane binding N terminus of the protein was not required for processing of the proenzyme to its mature form, and it raised the possibility that processing of a soluble form of the enzyme could be examined in vitro using epitope-tagged versions of the enzyme. We utilized a coupled in vitro transcription-translation (TNT) system to examine nascent enzyme formation and its subsequent processing to mature enzyme. We empirically determined that an N-terminal His 6 tag appended to PkPSD⌬34 produced the most robust catalytic activity from the in vitro TNT reaction. In our initial experiments, we observed significant production of the PkPSD⌬34 proenzyme (39 kDa), with modest processing to the mature enzyme (seen as the 33-kDa ␤ subunit), detected by both immunoblotting and measurement of catalytic activity (see Fig. 9, a and b). We tested whether the presence of a membrane compartment, canine pancreatic microsomes, could augment processing of the enzyme, and these experiments did not demonstrate any influence of added membranes upon the process (Fig. 9, a and b).
Because several pyruvoyl enzymes have been reported to be stabilized via a Schiff base conjugate with their substrates and reaction products (34), we first tested whether inclusion of the substrate (PtdSer) for the reaction might be required to promote proenzyme processing. As detailed in Fig. 10a, the inclu-   5 and 6). b, PSD enzyme assays were performed with cell-free extracts from Rosetta strain harboring pET45 and pET45-PkPSDHis 6 . Data are means Ϯ S.E. for three experiments each performed in duplicate. sion of increasing concentrations of DOPS in the TNT reaction, greatly enhanced the processing of the PkPSD enzyme to its mature form as shown by the reduction in the proenzyme content and increased appearance of the ␤ subunit, detected by immunoblotting. The enhanced processing of the proenzyme was also accompanied by significant quantitative increases (3-4-fold) in the amount of detectable catalytic activity as shown in Fig. 10b.
Next, we examined if enzyme processing is a co-translational event or occurs in a time-dependent manner after translation. First, in vitro synthesis of PkPSD was conducted for 60 min in the absence of DOPS. Subsequently, the translation reactions were halted by addition of cycloheximide, and the reaction aliquots were incubated further to allow processing of PkPSD in the absence or presence of DOPS. The reaction aliquots were removed at increasing times of incubation after translation arrest and analyzed as shown in Fig. 10c. The data reveal that during the initial 60-min translation reaction, significant precursor PkPSDs were made, but with very low levels of processed ␤ subunits. With increasing time after translational arrest, some mature forms were produced even in the absence of DOPS, indicating that PkPSD processing did not need to be co-translational. The lipid supplementation after translational arrest revealed that DOPS significantly enhanced processing of the newly synthesized proenzyme, which resulted in a time-de-pendent decline in the proenzyme content and an increase in the mature ␤ subunit content.
The DOPS used to stimulate PkPSD processing was prepared as a sonicated suspension of unilamellar vesicles. We also examined whether the diameter of the vesicles affected the processing reaction by preparing liposomes of various sizes, using extrusion through polycarbonate filters. Liposomes of 50, 200, and 1000 nm diameter all produced similar levels of  enzyme processing (Fig. 10d). In addition, large multilamellar liposomes also produced robust enhancement of PkPSD proenzyme processing.
Processing of PkPSD Is Down-regulated by the Anionic Lipids, DOPA, DOPG, and PI-We next investigated if regulation of PkPSD processing is PtdSer-specific. PtdSer is an anionic phospholipid, and we compared its activity with other anionic phospholipids, including DOPA, DOPG, and bovine liver PI. Comparisons were also made with the zwitterionic phospholipids, DOPE, and DOPC. Fig. 11a shows that supplementation of the in vitro processing reactions with DOPC or a DOPC/DOPE mixture had only a modest stimulatory effect on the process. In contrast, the anionic lipids, DOPA and DOPG, exerted a strong inhibitory effect upon maturation of PkPSD, which resulted in concomitant reduction of ␤ subunit formation and catalytic activity. Strong inhibition of the processing of the proenzyme by DOPG and DOPA resulted in reduced PSD activities (90 and 85% reduction by DOPG and DOPA, respectively) as seen in Fig. 11b. DOPC and a DOPC/DOPE mixture show only mild increases in the catalytic activities consistent with their mild stimulatory effects on the processing of the proenzyme. Interestingly, although other lipids did not affect overall synthesis of the PkPSD proenzyme, PI strongly inhibited it, indicating disruption of either transcription or translation by the in vitro reaction.
To bypass the inhibitory effect on expression of precursor PkPSDs by PI, the reaction was conducted for 40 min in the absence of lipid, followed by translational arrest with cycloheximide and further incubation with the various lipids. The Western blot analysis showed that DOPE and DOPC did not alter the basal level of processing, but PI along with DOPA and DOPG strongly inhibited processing (Fig. 11c). Taken together, the data demonstrate that anionic lipids, other than PtdSer, have a strong inhibitory effect on the post-translational processing of the PkPSD precursor. These findings suggest that membrane lipid composition plays an important regulatory role in PSD processing.
Regulating Lipids Do Not Bind to PkPSD Precursors with High Affinity-Because there was strong activation/inhibition by anionic phospholipids of the processing of the PkPSD precursor, we investigated whether there was any stable physical association between lipids and the proenzyme. Using multilamellar liposomes, we tested if the precursor PkPSDs generated in the TNT reactions could be sedimented with the lipid by centrifugation. The data in Fig. 12 shows that there is no strong physical association between any of the lipids and the PkPSD precursor. These results suggest that both DOPS stimulation of proenzyme processing and DOPG, DOPA, and PI-mediated inhibition of proenzyme processing occur by low affinity interactions between the protein and lipids.

DISCUSSION
The successful cloning of the PkPSD cDNA using a librarybased genetic complementation approach in yeast demonstrates the utility of this system for assigning function to numerous genes within the Plasmodium genus. We screened ϳ130 P. knowlesi genome equivalents and identified 127 isolates of PkPSD. These data indicate that the library will likely cover very rare cDNAs expressed at very low levels. The boot- FIGURE 11. Anionic phospholipids inhibit maturation of PkPSD. The influence of multiple phospholipids upon the maturation of PkPSD was examined by Western blotting, and enzyme assay was performed in conjunction with TNT reactions that used pET45-His 6 -PkPSD⌬34 as a template. a, TNT reactions were performed for 90 min at 30°C in the presence of empty vector, or plasmid harboring His 6 -PkPSD⌬34 in the absence of lipids (No lipid), or the presence of 100 g/ml DOPC, or 50 g/ml DOPC ϩ 50 g/ml DOPE, or 100 g/ml DOPG, or 100 g/ml DOPA. Following these incubations, aliquots of the reactions were analyzed by Western blotting. a also contains incubations that included soybean PI. b, same reactions were performed as shown in a with the corresponding plasmid and lipid additions indicated in the figure and then added to PSD assays, and the enzyme activity was quantified. c, TNT reaction using a pET45-His 6 -PkPSD⌬34 template was performed for 60 min and then arrested by the addition of 0.2 mM cycloheximide. Subsequently, the maturation reaction was conducted at 30°C for an additional 40 min in either the absence of lipid or the presence of DOPA, DOPG, PI, DOPC, DOPE, or DOPS. Aliquots from the maturation reaction were examined by Western blotting. FIGURE 12. Regulatory lipids do not bind to the His 6 -PkPSD⌬34 precursor with high affinity. The TNT reaction for in vitro expression of His 6 -PkPSD⌬34 was carried out for 40 min at 30°C. Translation was arrested by addition of 0.2 mM cycloheximide, and maturation reactions were continued for an additional 40 min in either the absence (None) or presence of 100 g/ml lipids (DOPS, DOPC, PI, and DOPG) as indicated. The lipids were in the form of multilamellar liposomes that could be sedimented by centrifugation. Following the maturation reaction, aliquots were shifted to 0°C and either unprocessed (R) or centrifuged at 13 ϫ 10 4 ϫ g for 20 min to recover liposomes (P). Volume normalized aliquots of the total reaction and the liposome pellet were analyzed by Western blotting.
strapping approach of functional genetic screening with PkPSD libraries and cross-correlating to the P. falciparum genome effectively bypasses the problem of the A ϩ T-rich genome of P. falciparum. We foresee this work as an important stepping stone toward large scale functional annotation of the P. falciparum genome. In addition, our approach has immediately yielded important new insights into the biochemistry of the parasite PSD, which have not otherwise been tractable with other eukaryotic PSD enzymes. More detailed understanding of the mechanism of PkPSD processing has the potential to reveal new and unsuspected vulnerabilities of the parasite to disruption of membrane biogenesis.
Sequence alignment between PkPSD and E. coli PSD reveals that the PkPSD sequence starting at position 56 shares homology with E. coli sequence at position 16 (Fig. 2). The sequence homology between PkPSD and ScPSD1 begins at position 67 of the parasite protein and position 138 of the yeast protein. The N-terminal 137 amino acids of yeast PSD1 contain mitochondrial targeting and inner membrane sorting sequences. The truncated yeast PSD1 with a 137-amino acid deletion was reported as a nonfunctional enzyme because the encoded construct failed to be processed to functional ␣ and ␤ fragments (33). Unlike PSDs from other organisms, including E. coli, yeast, mammals, and plants, that contain a strong hydrophobic transmembrane domain in the C terminus of their ␤ fragments, there is no trans-membrane domain in the corresponding region of PkPSD. Instead, PkPSD contains a predicted transmembrane domain at the N terminus at positions 18 -34. When the PkPSD cDNA was expressed in yeast, nearly half of the enzyme activity was recovered in the soluble fraction. Western blot analysis of epitope-tagged versions of the full-length proenzyme did not reveal any discernible size differences between the soluble and membrane-associated proteins, suggesting that the protein was amphitropic. The amphitropic properties of the enzyme are likely to allow access to its substrate in multiple membranes, including the outer mitochondrial membrane. Decarboxylation of PtdSer at the outer mitochondrial membrane is the likely reason that expression of PkPSD in yeast enables strain growth under respiratory conditions. The factors that dictate membrane association remain unidentified.
Truncated versions of the PkPSD lacking the N terminus and predicted transmembrane domain (amino acids 2-34) were expressed in yeast, and the PkPSD⌬34 version resulted in a highly active form of the enzyme. Epitope-tagged versions of PkPSD also resulted in catalytically active forms of the enzyme in both bacteria and yeast. Epitope-tagged forms of PkPSD⌬34 provided a convenient variant of the enzyme for examination of processing events in vitro, and the His 6 -PkPSD⌬34 cDNA was transcribed and translated with a TNT system that contains reticulocyte lysate components for protein synthesis. In the absence of lipid supplementation, a modest level of the nascent enzyme is processed to the mature form. Inclusion of DOPS in the transcription-translation reaction increases the processing of proenzyme 3-4-fold, as evidenced by the diminution in the amount of unprocessed form, and a concomitant increase in the amount of mature ␤ subunit. The increased processing observed on Western blotting correlates with definitive quan-titative increases in catalytic activity of the enzyme detected in PSD assays. The action of DOPS appears specific, insofar as DOPE and DOPC have little or no effect upon enzyme maturation, whereas DOPG, DOPA, and PI are inhibitory. The observed low level of maturation of the PkPSD in the absence of lipid addition may be a consequence of endogenous levels of phosphatidylserine present in the reticulocyte lysate.
The application of the TNT system also enabled us to determine that the P. knowlesi PSD undergoes processing after translation is complete. This timing may be of important regulatory value insofar as it provides the nascent proenzyme with a time window for sampling membrane content of PtdSer and other anionic phospholipids. Membrane environments rich in PtdSer will promote processing, whereas those with high content of other anionic phospholipids will inhibit processing. A schematic summary of lipid modulation of PkPSD proenzyme processing appears in Fig. 13.
The maturation of PSD enzymes broadly appears to take place in three steps (35). In the first step, the hydroxyl group of a critical serine, which is destined to form the active site pyruvoyl group, attacks the proximal (N-terminally located) peptide bond, effecting serinolytic cleavage that results in the ␤ subunit being attached to the ␣ subunit by an ester bond. In the second step, the ␤ subunit is liberated by an elimination reaction, which also yields an ␣ subunit harboring an N-terminal dehydroalanine. In the third step, the reactive dehydroalanine undergoes water addition and ammonia elimination to form the pyruvoyl prosthetic group on the ␣ subunit. The mechanism of action of DOPS in promoting proenzyme processing is not yet known, but two possibilities seem likely. One mechanism could involve DOPS forming a Schiff base with the newly formed carbonyl moiety of the ␣ subunit immediately as it is formed, which would function to shift the equilibrium of the processing reaction in the direction of mature enzyme. A second mechanism might involve DOPS binding to the ␤ subunit as an allosteric enhancer of the auto-endoproteolytic reaction. The availability of soluble versions of PkPSD should enable testing of these two mechanisms in the future. PkPSD is synthesized as a proenzyme that undergoes auto-endoproteolytic processing to form a mature enzyme containing a large ␤ subunit and a small ␣ subunit with an N-terminal pyruvoyl prosthetic group. An engineered form of PkPSD lacking the N-terminal transmembrane domain and containing a His 6 epitope tag undergoes maturation in vitro. The maturation reaction is a post-translational event that is enhanced by the presence of the substrate for the enzyme (DOPS) and inhibited by other anionic phospholipids PI, DOPG, and DOPA. These findings suggest that intracellular substrate abundance and local membrane phospholipid composition play an important role in regulating the quantity of active PSD within the cell.