The Role of CDP-Diacylglycerol Synthetase and Phosphatidylinositol Synthase Activity Levels in the Regulation of Cellular Phosphatidylinositol Content*

The regulation of phosphatidylinositol synthesis was examined by cloning and expressing in COS-7 cells the human cDNAs encoding the two enzymes in the biosynthetic pathway. Human CDP-diacylglycerol synthetase (cds1) and phosphatidylinositol synthase (pis1) clones were identified in the human expressed sequence-tagged (EST) data base, and full-length cDNAs were obtained by library screening. The cds1 cDNA did not possess a recognizable mitochondrial import signal, and the activity of the expressed Cds1 protein was stimulated by nucleoside triphosphates in vitro, indicating that cds1 did not encode the mitochondrial-specific isozyme. There were two mRNA species (3.9 and 5.6 kilobases) detected on Northern blots hybridized with thecds1 probe that were expressed at distinctly different levels in various human tissues. Consistent with the presence of the two mRNAs, a cDNA predicted to encode a second human CDP-diacylglycerol synthetase (cds2) was also uncovered in the EST data base. In contrast to the two cds mRNAs, a single, 2.1-kilobase pis1 mRNA was uniformly expressed in all human tissues examined. Expression of the pis1 gene led to the overproduction of both phosphatidylinositol synthase and phosphatidylinositol:inositol exchange reactions, indicating that the Pis1 polypeptide catalyzed both of these activities. Phosphatase treatment of cell extracts abolished the CMP-independent phosphatidylinositol:inositol exchange reaction, and exchange activity was completely restored by the addition of CMP. Overexpression ofcds1 or pis1 alone or in combination did not enhance the rate of phosphatidylinositol biosynthesis. Also, overexpression did not result in a significant proportional increase in the cellular levels of CDP-diacylglycerol or phosphatidylinositol. These data illustrate that the levels of Cds1 and Pis1 protein expression are not critical determinants of cellular PtdIns content and argue against a determining role for the activity of either of these enzymes in the regulation of PtdIns biosynthesis.

The biosynthesis and metabolism of PtdIns 1 is of considerable interest due to the involvement of PtdIns and its phosphorylated derivatives in intracellular signal transduction. The breakdown products of the phosphoinositides are ubiquitous second messengers downstream of many G protein-coupled receptors and tyrosine kinases involved in mitogenesis, the regulation of calcium mobilization, and protein kinase C activation (1). Polyphosphoinositides are also involved in vesicular movement within cells (2), cytoskeletal organization (3)(4)(5)(6), and stimulation of protein kinase cascades (7). Much of the recent work on these signal transduction pathways has focused on the regulation of the kinases and phospholipases involved in generating polyphosphoinositide-derived second messengers. However, a rapid agonist-dependent burst of PtdIns biosynthesis was the first feature of the polyphosphoinositide signaling pathway to be discovered, illustrating that synthesis is tightly coupled to degradation (8). Little is known about the biochemical mechanisms responsible for this fine control over PtdIns synthesis.
There are two enzymes in the PtdIns biosynthetic pathway, CDP-DG synthetase (CTP:phosphatidate cytidylyltransferase, EC 2.7.7.41) and PtdIns synthase (CDP-diacylglycerol:myo-inositol 3-phosphatidyltransferase, EC 2.7.8.11). CDP-DG synthetase sits at a branch point in phospholipid metabolism where phosphatidic acid is partitioned between diacylglycerol or CDP-DG, the key intermediate in the formation of anionic phospholipids (9). CDP-DG synthetase genes have been cloned, and the proteins have been purified and biochemically characterized in Escherichia coli (10,11), Saccharomyces cerevisiae (12,13), and Drosophila (14). The enzyme is less characterized from mammalian sources but is thought to exist in at least two forms. The CDP-DG synthetase associated with the cytoplasmic aspect of the endoplasmic reticulum (15) is thought to operate in the PtdIns biosynthetic pathway, whereas the enzyme located on the matrix side of the inner mitochondrial membrane (16) appears to be involved in the synthesis of phosphatidylglycerol and cardiolipin. Neither form of mammalian CDP-DG synthetase has been purified although the enzymes from the two sources appear to be different in that the microsomal enzyme is stimulated by GTP, whereas the mitochondrial enzyme is not (17). Recently, human (18,19) and rodent (20) cDNA sequences for CDP-DG synthetase were reported (18). PtdIns synthase is located primarily on the cytoplasmic aspect of the endoplasmic reticulum (15) although the enzyme has also been detected in plasma membrane preparations (21)(22)(23)(24). PtdIns synthase has been purified from S. cerevisiae (25,26), and its gene has been cloned (27,28). PtdIns synthase has also been extensively purified from human placenta (29) and rat liver (30), yielding proteins with estimated molecular masses of 24 and 21 kDa, respectively. Recently, a rodent cDNA encoding rat PtdIns synthase was isolated (31). Imai and Gershengorn (22) proposed that cellular PtdIns content was regulated by feedback inhibition of PtdIns synthase by PtdIns. This conclusion was based on the inhibition of PtdIns synthase activity in cellular membrane preparations. However, this finding is different from Fischl et al. (26) who reported that PtdIns had a modest stimulatory effect on yeast PtdIns synthase activity. The reason for these differences is not clear but may relate to the assay systems employed. Alternatively, CDP-diacylglycerol synthetase may regulate PtdIns biosynthesis by analogy to CTP:phosphocholine cytidylyltransferase in the phosphatidylcholine biosynthetic pathway (9). The goal of this study was to clone and express the human cds 2 and pis genes to determine whether cellular PtdIns content was controlled by the expression levels of these two activities or whether regulation of PtdIns synthesis was independent of the cellular amounts of pathway enzymes.
Cloning the Human cds1 and pis1 cDNAs-The yeast CDP-diacylglycerol synthetase protein sequence (accession number S45885) was used to search the EST data base using the BLAST algorithm (32) which identified a clone (accession number R27966) as a candidate for human cds1. This clone was sequenced, and a primer was synthesized to extend the sequence in the 5Ј direction by PCR RACE. A 1.1-kb product was generated, sequenced, and found to contain an ATG start codon. The RACE sequence was used to search the EST data base, and a clone (accession number N29532) was identified that matched the 5Ј RACE product. DNA sequencing, restriction mapping, and expression studies showed that the N29532 was a full-length cds1 cDNA derived from human placenta. The cds1 expression vector was constructed by digesting clone N29532 with SacII and SspI. The 1.643-kb fragment was blunt-ended with mung bean nuclease and cloned into pcDNA3 that had been digested with EcoRV. The correct orientation was confirmed by digestion with EcoRI, and the resulting plasmid, pPJ20, was confirmed to express a protein with an apparent molecular mass of 52 kDa using a coupled transcription/translation assay (not shown).
Human pis1 cDNA was cloned using a similar strategy. The yeast PtdIns synthase protein sequence (accession number J02697) was used to search the EST data base using the BLAST algorithm (32) which identified a clone (accession number T49269) as a candidate for human pis1. This clone was sequenced, and primers were synthesized to extend the sequence in the 5Ј direction by PCR amplification of a human testis cDNA library using pis1 gene-specific primers and the phage-specific long distance amplimers. A 650-bp product was generated, sequenced and found to predict an AUG start codon preceeded by stop codons in all upstream reading frames. This sequence was used to identify a second EST clone (accession number AA070158) that was sequenced to confirm the data obtained with the PCR RACE product. The AA070158 clone was sequenced and contained the start codon and additional upstream sequences. The pis1 open reading frame was assembled by ligating the 1.078-kb XhoI fragment from EST T49269 into XhoI-digested EST AA070158. Correct orientation was verified by digestion with BamHI. The 1.171-kb BamHI fragment was ligated into the BamHI site of pcDNA3 to generate plasmid pPJ28, and the correct orientation of the pis1 gene was confirmed by digestion with EcoRI.
CDP-diacylglycerol Synthetase Assay-The CDP-diacylglycerol synthetase assay was performed essentially as described (33). Cells were resuspended in 50 mM Tris-HCl, pH 8.0, incubated on ice for 30 min, and lysed by sonication in an ice bath for 3 ϫ 30 s. The assays contained 0.69 M [ 3 H]CTP (specific activity 14.5 Ci/mmol), 10 mM MgCl 2 , 2 mM phosphatidic acid (sonicated suspension in water), 100 mM BisTris-HCl, pH 6.5, and 5-25 g of protein in a final volume of 50 l. The MgCl 2 was the last component added to the reaction mixture. The assay mixtures were incubated for 5 min at 37°C, and the assays were terminated by the addition of 180 l of chloroform/methanol/concentrated HCl (1:2: 0.02, v/v). Next, 60 l of chloroform and 60 l of 2 M KCl were added, and following vortex mixing, the phases were separated by centrifugation. The amount of [ 3 H]CDP-diacylglycerol formed was determined by counting the radioactivity in 40 l of the organic phase. CDP-diacylglycerol formation was confirmed by thin-layer chromatography of the organic phase on Silica Gel 60 plates developed with chloroform/methanol/acetic acid/water (50:30:8:4, v/v). The radioactivity co-chromatographed with the CDP-diacylglycerol standard.
PtdIns Synthase Assay-The PtdIns synthase assay was performed essentially as described (34). The PtdIns exchange reaction was assayed essentially as described previously (35). The assays contained 0.5 M PtdIns (sonicated suspension in water), 3.3 M myo-[ 3 H]inositol (specific activity 30 Ci/mmol), 2 mM MnCl 2 , 50 mM MgCl 2 , 100 mM Tris-HCl, pH 8.0, and 5-50 g of protein in a final volume of 50 l. The assay mixtures were incubated for 15 min at 37°C, and the assays were terminated by the addition of 180 l of chloroform/methanol/concentrated HCl (1:2:0.02, v/v). Next, 60 l of chloroform and 60 l of 2 M KCl were added, and following vortex mixing, the phases were separated by centrifugation. The amount of PtdIns formed was determined by counting the radioactivity present in 40 l of the organic phase. The formation of PtdIns was confirmed by thin-layer chromatography of the organic phase on Silica Gel 60 plates developed as described above in the PtdIns synthase assay. All of the radioactivity co-chromatographed with the PtdIns standard. Extracts of COS-7 cells transfected with pPJ28 (pis1) (34 g of protein) were assayed for PtdIns:Ins exchange activity as described above either in the presence or absence of 4 mM CTP or 20 units/ml calf intestinal phosphatase.
Expression of cds1 and pis1 in COS-7 Cells-COS-7 cells were grown in 100-mm dishes to 80% confluency in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% glutamine. Transfections were performed according to the manufacturer instructions. Briefly, 10 g of plasmid and 60 l of LipofectAMINE TM reagent were separately diluted into 0.8 ml of serum-free medium. The two solutions were combined and incubated at 25°C for 45 min. Next, 6.4 ml of serum-free medium was added to each tube, and the diluted solution was overlaid onto the COS-7 cells that had been previously rinsed with serum-free medium. The cells and reagents were incubated at 37°C for 5 h, and then 8 ml of growth medium containing twice the normal amount of serum was added. The medium was replaced 24 h after the start of the transfection procedure, and the cells were incubated for an additional 24 h at 37°C and then harvested for analysis.
Metabolic Labeling-The effect of cds1 and pis1 expression on the cellular levels of PtdIns and CDP-DG was determined in COS-7 cells transfected with a control plasmid, pPJ20, pPJ28 or pPJ20 ϩ pPJ28. The total plasmid amount in each of the transfections was 10 g. The cells were labeled with either [ 3 H]inositol (3 Ci/ml) or [ 3 H]cytidine (8 Ci/ml) at 24 h after transfection when the medium was changed, and the cells were incubated an additional 24 h in the presence of label. Cells were washed three times with 10 ml of phosphate-buffered saline and harvested in 10 ml of phosphate-buffered saline, and the cell pellets were extracted using the 2-phase system described in the PtdIns synthase assay section. The radioactivity in the soluble and phospholipid phases was quantitated. The presence and amount of CDP-diacylglycerol in the organic phase was determined by thin-layer chromatography 2 cds indicates the gene, whereas Cds denotes the protein.
on Silica Gel 60 plates developed with chloroform/methanol/acetic acid/ water (50:30:8:4, v/v). The inositol-labeled phospholipids were separated as described under the PtdIns synthase assay. Phosphoinositides and liponucleotides were identified by co-migration with standards.
The rate of [ 3 H]Ins incorporation into PtdIns in COS-7 cells (60-mm dishes) transfected with either the control plasmid (pcDNA3), plasmid pPJ20, or plasmids pPJ20 plus pPJ28 was measured. The total plasmid amount in each tranfection was 4 g as described above. Cells were allowed to express the proteins for 24 h, and the cells were then labeled with [ 3 H]Ins (3 Ci/ml of growth medium). Cells were harvested at different time points up to 6 h, and the phosphoinositides were extracted and analyzed as described above.
Total Cellular PtdIns Mass-The effect of cds1 and pis1 expression on the cellular mass of PtdIns was also determined in COS-7 cells transfected with a control plasmid, pPJ20, pPJ28, or pPJ20 ϩ pPJ28. The total plasmid amount in each of the transfections was 10 g. The cells were washed three times with 10 ml of Hepes-buffered saline (136.8 mM NaCl, 2.6 mM KCl, 10 mM Hepes, pH 7.2) and harvested in 10 ml of Hepes-buffered saline 48 h after transfection. Cell pellets were extracted using the 2-phase system described in the PtdIns synthase assay section, and the lower phase was backwashed to quantitatively recover the lipid in each sample. The entire lower phase containing the phospholipids was fractionated by thin-layer chromatography on Silica Gel 60 plates developed with chloroform/methanol/ammonium hydroxide (60:35:8, v/v). The plates were stained with iodine vapor, and the portion of each lane that co-migrated with a [ 3 H]PtdIns standard (American Radiochemical Co.) was recovered. Following elution of the phospholipid from the gel with chloroform and evaporation of the solvent, the lipid sample was digested with 72% perchloric acid and heating. The phosphorus content of the digested sample was determined using the modified micro-procedure of Bartlett (36).
Northern Blots-Three human multiple tissue Northern blots were purchased from CLONTECH and were hybridized and washed according to the manufacturer instructions. The blots were first hybridized with a 32 P-labeled probe prepared from the 565-bp PstI-XhoI fragment of the cds1 cDNA. The blots were then stripped and hybridized with a 32 P-labeled probe prepared from the 582-bp SacI fragment of pis1.

RESULTS
Isolation and Structure of the cds1 and pis1 cDNAs-The EST data base was searched with the protein sequence of the known yeast gene encoding CDP-diacylglycerol synthetase, and human analogs of this protein were identified. The EST fragments were used to screen libraries and design primers for PCR RACE, and the complete cDNA sequences were isolated. The human cds sequence submitted by Heacock et al. (18) (accession number U65887) differs from the human cds1 sequence determined in this study and the human Cds sequence determined by Weeks et al. (19) (accession number U60808) in that the Heacock et al. sequence predicts a protein 17 amino acids shorter at the C terminus. The published rat cds sequence (20) (accession number AB000517) is the rodent homolog of human cds1 and is predicted to possess the C-terminal amino acid sequence predicted by us and Weeks et al. (19). The predicted protein is composed of 461 amino acids with a molecular mass of 53,226 daltons. We also identified a second cds gene in the EST data base (accession number AA040370), cds2, that is predicted to encode a highly related human CDP-diacylglycerol synthetase and also a murine sequence predicted to encode cds1 (accession number W30593). The sequences of the two human cds cDNAs are compared with the yeast, Drosophila, and E. coli enzymes in Fig. 1. There is significant similarity among these sequences throughout their coding regions. The most notable regions of homology among the Cds proteins of bacteria, yeast, flies, and mammals are between amino acids 273-284 and 361-407 and probably correspond to the active sites of the respective proteins. The human cds2 sequence is most likely derived from a second gene since the nucleic acid and amino acid sequences are unique and inconsistent with cds2 being an alternatively spliced form of cds1. The most notable differences between the predicted Cds1 and Cds2 proteins lie between residues 325 and 350 (Fig. 1). The Cds sequences are not related to the sequences of the cytidylyltransferases involved in the formation of CDP-choline (37) or CDP-ethanolamine (38), which are water-soluble intermediates in the biosynthesis of phosphatidylcholine and phosphatidylethanolamine.
A similar approach and methodology was used to identify, clone, and sequence the human pis1 cDNA. The human pis1 sequence predicts a protein composed of 213 amino acids with a molecular mass of 23,400 daltons (Fig. 2). The mammalian and yeast genes are highly related throughout their primary sequence, indicating that they have similar catalytic mechanisms and regulatory properties.
Chromosomal Location of the Human cds1 and pis1 Genes-The chromosomal locations of the two genes were determined by identifying exact matches in the human sequence-tagged FIG. 1. Comparison of the human cds1 predicted amino acid sequence with CDP-diacylglycerol synthetases from bacteria, yeast, and flies. The complete human cds1 sequence derived from this study is compared with the partial human cds2 sequence obtained from the EST data base (accession number AA040370), the partial murine cds1 sequence from the EST data base (accession number W30593), the complete Drosophila cds sequence (Dros.) (14) (accession number S52437), the complete yeast CDS1 sequence (13) (accession number S45885), and the complete E. coli cdsA sequence (10) (accession number M11330). The concensus sequence represents residues that are identical in at least three out of the four groups. Our human cds1 sequence is identical to the human cds sequence submitted by Weeks et al. (19) (accession number U60808). sites (STS) data base with the 3Ј-untranslated regions of the cds1 and pis1 cDNAs. These analyses revealed that the Cds1 gene was identical to the human STS WI-11924 (accession number G21972) that was mapped to human chromosome 1; however, a more precise localization was not available. The pis1 gene was identical to the human STS SHGC-11494 (accession number G14569). The pis1 STS was localized between 50 -56 centimorgan on human chromosome 16 between markers D16S3093 and D16S409. Also, a genomic clone for the human pis1 gene has been isolated. Cross et al. (39) used a methylated DNA binding column to purify MseI human genomic DNA fragments that contained non-methylated CpG islands. One of the clones isolated by this technique (accession number Z62620) contains the first coding exon (residues 1-15) and a portion of the second exon of the pis1 gene. CpG islands are typically located at the 5Ј end of "housekeeping" genes, a classification that is consistent with the universal expression of pis1 mRNA (see below).
Levels of cds1 and pis1 mRNA in Human Tissues-The relative levels of cds1 and pis1 mRNA in human tissues were compared using three multiple tissue Northern blots (Fig. 3). Two mRNAs with apparent sizes of 3.9 and 5.6 kb were detected with the cds1 probe. Different tissues have distinctly different levels of expression. For example, high levels of cds mRNA were detected in fetal kidney, lung, and brain, but the mRNA levels were significantly lower in fetal liver. Heacock et al. (18) also noted that two cds mRNAs were expressed in some of the human tissues they examined. It is possible that the two cds mRNA species in our blots correspond to the two cds cDNAs detected in our sequence analysis (Fig. 1) since the similarity between cds1 and cds2 at the nucleic acid level appears significant enough (not shown) to account for cross-hybridization with the 32 P-labeled probe used for detection. However, we do not have definitive information on this point, and obtaining the complete cds2 cDNA will be important to design the experiments needed to establish which one of the transcripts corresponds to cds1 and cds2. The pis1 probe detected a single mRNA species with an apparent size of 2.1 kb. In contrast to the pattern of cds expression, the expression of pis1 was more uniform among tissues although slightly higher levels of expression were detected in liver and skeletal muscle. The pattern of pis1 expression was consistent with the hypothesis of Cross et al. (39) who concluded that the promoter, upstream of what we now know is the pis1 gene, regulates the expression of a protein with a "housekeeping" function.
Expression and Activity of cds1-The specific activity of Cds1 in transfected cells increased from 15.7 Ϯ 1.8 pmol/min/mg in control cells to 120.5 Ϯ 5.5 pmol/min/mg in cells transfected with pPJ20 (cds1) (Fig. 4). The average increase in Cds-specific activity was 7.6-fold. The stimulation of Cds activity by GTP was previously proposed to distinguish the microsomal form of the enzyme, which is activated by GTP, from the mitochondrial form (17,40,41). GTP modulation of Cds activity reported by these investigators was potentially an exciting finding at the time since it suggested the possibility of G protein regulation of the enzyme. Therefore, we tested the ability of GTP and other nucleotides to stimulate the activity of Cds1 in extracts from COS-7 cells transfected with the cds1 expression vector. Cds activity was stimulated 2-2.8-fold by all nucleotides examined, including GTP, GDP, and ATP (Table I). Previous work did not explore this range of nucleotides, but our results are clearly different from the reports that ATP was without effect and that Cds activation was GTP-specific (17,40,41). The reason for the disagreement between the two results is not clear. Our data suggest that the effect of GTP on Cds activity was not GTPspecific. The presence of other nucleotides in the assay may stimulate the reaction by preventing the degradation or utilization of [ 3 H]CTP by competing reactions present in the crude cell extracts. The lack of specificity in the nucleotide studies (Table I) could also be due to the nucleoside diphosphate kinase activity likely present in crude cell extracts. This transphosphorylation reaction interconverts the various nucleoside triphosphates and obviates a clear interpretation of the data.
Nonetheless, our experiments are consistent with the biochemical properties attributed to the microsomal protein (17,40,41). There was no indication of a mitochondrial import signal in the cds1 sequence supporting the idea that cds1 encoded a Cds isoform that was not specifically localized to the mitochondria. A subcellular distribution study by Weeks et al. (19) found that a portion of the Cds1 activity was associated with the microsomal fraction, but the majority of activity was found in the nuclei/mitochondria pellet, suggesting that Cds1 was not exclusively localized to the microsomal fraction. Immunoelectron microscopy showed rodent Cds1 to be localized along the membranes of endoplasmic reticulum, vesicles, and nuclear envelopes (20). It will be interesting to complete the The complete human pis1 sequence derived from this study (GenBank TM /EBI accession number AF014807) is compared with the yeast (28) (accession number J02697) and rat (31) (accession number D82928) PtdIns synthase sequences. The concensus sequence represents the residues that are identical between the yeast and mammalian sequences. characterization of the cds2 clone to determine if the Cds2 isoform specifically localizes to the mitochondria. On the other hand, mitochondria possess a transport system for the uptake of CDP-DG from the cytosol (42), and the existence of this mechanism questions an absolute requirement for matrix Cds in order for mitochondria to synthesize acidic phospholipids. Immunological reagents to specifically localize Cds isoforms in cells will provide important tools to investigate the subcellular localization of Cds.
Expression and Activity of pis1-The expression of pis1 in COS-7 cells led to an overproduction of Pis activity (Fig. 5A). The specific activity of Pis increased from 1.04 Ϯ 0.22 pmol/ min/mg of protein in nontransfected cells to 25.4 Ϯ 0.84 pmol/ min/mg of protein in transfected cells. There was also a corresponding increase in PtdIns:Ins exchange activity in cells transfected with pPJ28 (pis1) (Fig. 5B). The specific activity of the exchange reaction increased from 0.37 Ϯ 0.04 pmol/min/mg of protein in the nontransfected cells to 8.6 Ϯ 0.54 pmol/min/mg of protein in the transfected cells. Both activities increased approximately 24-fold over control values in the cells transfected with pPJ28 (pis1).
The PtdIns:Ins exchange reaction has been detected in cell extracts by numerous investigators (24,26,35,(43)(44)(45)(46)(47)(48), and there are two types of exchange activities that have been reported. The CMP-dependent exchange reaction is thought to be due to the reverse reaction of PtdIns synthase (26,34,48). There is also a "nucleotide-independent" exchange reaction that does not require the addition of exogenous CMP, and this activity has been attributed to the presence of a second enzyme (43,48). Recently, Klezovitch et al. (35) reported that expression of the yeast PtdIns synthase gene in E. coli resulted in an increase in both CMP-dependent and -independent exchange activities, and they suggested that these were distinctly different reactions catalyzed by the same enzyme. This interpretation was difficult to reconcile with the sequential ordered mechanism of yeast PtdIns synthase. The only exchange reaction catalyzed by purified yeast PtdIns synthase absolutely required the presence of CMP (26,34). Thus, we tested the idea that there was endogenous CMP bound to Pis1 in the cell extract that accounted for the nucleotide-independent exchange reaction. Extracts from COS-7 cells transfected with pPJ28 (pis1) were assayed for PtdIns:Ins exchange activity in the presence and absence of CMP and in the presence and absence of calf intestinal phosphatase (Fig. 6). Pis1 expression increased the basal PtdIns:Ins exchange rate (Fig. 5B) that was significantly stimulated by CMP (Fig. 6). The phosphatase was included in the incubations to degrade endogenous CMP, and indeed, treatment of the extract with phosphatase almost completely eliminated nucleotide-independent exchange activity (Fig. 6). The addition of CMP to the phosphatase-treated extract completely restored exchange activity. These data did not FIG. 3. Relative abundance of cds and pis1 mRNA in human tissues. Three human multiple tissue Northern blots were purchased from CLONTECH and were hybridized and washed according to the manufacturer instructions. The blots were first hybridized with 32 Plabeled probe prepared from the 565-bp PstI-XhoI fragment of the human cds1 cDNA. The blots were then stripped and hybridized with a 32 P-labeled probe prepared from the 582-bp SacI fragment of the pis1 cDNA.  The specific activity of Cds1 in COS-7 cell extracts in the absence of nucleotides was 78 pmole/min/mg. support the existence of both nucleotide-dependent and -independent exchange reactions for Pis1 but, rather, indicated that the nucleotide independent reaction was due to the association of endogenous CMP with the enzyme. Thus, we concluded that the only exchange reaction catalyzed by Pis1 was CMP-dependent, consistent with an ordered, sequential reaction mechanism where PtdIns dissociated from the enzyme before CMP.
Perturbation of PtdIns Metabolism by the Overexpression of Cds1 and/or Pis1-The effect of cds1 and pis1 expression on the rate of PtdIns synthesis was determined by metabolically labeling COS-7 cells transfected with plasmid pPJ20 (cds1), plasmids pPJ20 (cds1) plus pPJ28 (pis1), or a vector control with [ 3 H]inositol (Fig. 7). All three transfected cell populations readily incorporated [ 3 H]inositol into the cellular phospholipid fraction, and thin-layer chromatography confirmed that the phosphoinositides contained all of the label (not shown). There was little difference among the three transfected cell populations in the rate of [ 3 H]inositol incorporation although there was a slight increase in the rate of labeling in the cell population transfected with both cds1 and pis1. The slight difference in [ 3 H]inositol incorporation between the transfected cell populations and the controls did not correspond to the much higher levels of Cds1-and Pis1-specific activity in these cells (see above). Thus, the overexpression of either Cds1, or Cds1 plus Pis1, did not have a major impact on the rate of PtdIns biosynthesis.
The effect of Cds1 and Pis1 protein levels on the total cellular content of PtdIns was investigated by labeling transfected COS-7 cells with [ 3 H]Ins for 24 h to uniformly label the PtdIns pool (Fig. 8). Despite a 7.6-fold increase in Cds activity in the cell population, PtdIns increased an average of only 15.8%. Pis1 overexpression led to an 8.2% increase. Overexpression of both Cds1 and Pis1 together led to a greater increase in PtdIns labeling (59.6%), but this 1.5-fold increase in PtdIns detected by this method was significantly less than the 7.6-and 25-fold increases in Cds1 and Pis1 protein expression, respectively. Also, we did not observe differences between the amount of label in the soluble Ins pool in the transfected compared with control cell populations. These results were corroborated by labeling the transfected cell populations to equilibrium with [ 32 P]orthophosphate for 24 h which failed to show a significant increase of labeled PtdIns (not shown). We also measured the cellular mass of PtdIns, including its phosphorylated derivatives, following transfection of COS-7 cells with either the cds-1or the pis-1 cDNA, or the two cDNAs together. The amount of lipid phosphorus per mg of protein was determined by a colorimetric method in duplicate 48 h after the transfections. Two independent experiments revealed that the control value of 45.1 Ϯ 7.8 ng of P i /mg of protein obtained using cells transfected with the pcDNA3 vector did not vary significantly with overexpression of the Cds protein (49.6 Ϯ 2.1 ng of P i /mg of protein) or the Pis protein (50.9 Ϯ 3.2 ng of P i /mg of protein). Both the Cds and Pis proteins were co-expressed to ensure that neither the supply of CDP-diacylglycerol nor Pis protein was limiting to the rate of flux through the biosynthetic pathway, and we found that the PtdIns content was 54.6 Ϯ 1.5 ng of P i /mg of protein under this experimental condition. Thus, the overexpression of either Cds1, Pis1, or both did not significantly alter the steady-state amount of PtdIns.
We also examined the effect of Cds1 overexpression on the levels of intracellular CDP-DG by metabolic labeling with [ 3 H]cytidine. We detected little difference between the amount of [ 3 H]cytidine in the lipid phase of control and cds1-transfected cell populations (36.8 Ϯ 0.05 compared with 40.3 Ϯ 0.05 dpm/mg of protein). Thin-layer chromatography showed that Ͼ80% of the [ 3 H]cytidine label in the lipid extract co-migrated with CDP-DG standard. Thus, the overexpression of Cds1 by an average of 7.6-fold did not result in an increased intracellular concentration of its product, CDP-DG. DISCUSSION Cytidine intermediates play a key role in the biosynthesis of all classes of glycerol phospholipids, and the enzymes catalyzing their formation are thought to catalyze the rate-controlling steps in their respective pathways (9). The biosynthetic pathways to PtdIns, phosphatidylcholine, and phosphatidylethanolamine are all 2-component systems composed of a cytidylyltransferase followed by a synthase (9). CTP:phosphocholine cytidylyltransferase is the most studied enzyme and is the rate-controlling step in phosphatidylcholine biosynthesis (9). Accordingly, transient overexpression of this cytidylyltransferase in COS cells yielded a 300 -500% increase in the incorporation of [ 3 H]choline into phosphatidylcholine and a significant elevation in cellular CDP-choline (49). By analogy, Cds has been proposed as a regulatory point in PtdIns biosyn-thesis (9), and this idea predicts that the transient overexpression of Cds would accelerate PtdIns synthesis. However, Cds overexpression did not result in significantly enhanced incorporation of Ins into PtdIns or result in higher cellular levels of cellular PtdIns. The same conclusion arises from our analysis of the PtdIns labeling pattern in cells overexpressing Pis1, or Cds1 plus Pis1.
We conclude that the expression levels of Cds1, Pis1, or both are not determining factors in controlling the de novo rate of PtdIns synthesis or in establishing the cellular PtdIns content. One explanation for this result may be that Cds1 and/or Pis1 are subject to very tight regulation. Another possible explanation that we considered was that the PtdIns:Ins exchange activity of Pis1 acts to reduce an excessive cellular PtdIns content. The exchange activity is due to the reverse reaction of PtdIns synthase and is dependent on CMP, which is tightly bound to Pis1 protein. The degree of reduction of PtdIns would not only be governed by the equilibrium constants of the forward and reverse reactions and the concentration of PtdIns but also by the amount of CMP-bound protein. Previous work with cell membranes indicated that PtdIns was an inhibitor of the PtdIns synthase forward reaction (22). Our data suggest that PtdIns inhibition correlates with heightened exchange activity. However, a mechanism that places Pis1 as a rate-controlling reaction predicts that CDP-DG would accumulate in cells when Cds1 activity is overexpressed. In contrast, we found that the intracellular levels of CDP-DG were the same in control and Cds-overexpressing cell populations. These data illustrate that CDP-DG pools are strictly maintained and strongly argue against a significant role for Pis1 in controlling the cellular PtdIns content. Cds1 activity is inhibited by polyphosphoinositides in vitro (20), suggesting that these end products of the pathway may function as feedback inhibitors of PtdIns biosynthesis in vivo. If this regulatory mechanism is physiologically important, then the levels of polyphosphoinositides would need to increase in Cds-transfected cell populations to effectively regulate the elevated levels of Cds. However, we did not observe an increase in the level of polyphosphoinositides in Cds1-overexpressing cell populations compared with controls, thus arguing against this mechanism for Cds regulation. CDP-DG pools may also be controlled by a CDP-DG hydrolase. Although an E. coli CDP-DG hydrolase (cdh) is known (50), a specific enzyme that hydrolyzes CDP-DG has not been characterized in metazoan cells.
In light of this data, we propose that the supply of phosphatidic acid limits net PtdIns synthesis via the Cds/Pis pathway. Phosphatidic acid phosphohydrolase lies at a branch point in glycerolipid biosynthesis and is responsible for converting phosphatidic acid to diacylglycerol, which in turn is used for the biosynthesis of phosphatidycholine, phosphatidylethanolamine, and triacylglycerol. There are two forms of this enzyme. The Mg 2ϩ -independent form (PAP2) is thought to play a role in signal transduction (51,52) and is therefore unlikely to play a role in controlling de novo biosynthesis. The Mg 2ϩ -dependent isoform (PAP1) is thought to control the biosynthetic dephosphorylation of phosphatidic acid and is regulated by reversible association with membranes (53). Thus, the amount of phosphatidic acid available to Cds may be severely limited by PAP1, shunting the bulk of the phosphatidic acid to diacylglycerol and the major membrane phospholipid classes. This conclusion is not entirely satisfying in light of the results of Balsinde et al. (54) who reported that the inhibition of PAP1 with bromoenol lactone led to a decrease in triacylglycerol and phosphatidylcholine synthesis and an increase in phosphatidic acid levels. However, there was no change in the rate of incorporation of FIG. 8. Perturbation of PtdIns metabolism by the overexpression of CDP-diacylglycerol synthetase and/or PtdIns synthase. COS-7 cells were transfected with plasmids that expressed either cds1 (pPJ20), pis1 (pPJ28), or with the two plasmids combined. The cells were incubated for 24 h to allow expression of the genes, the media was changed, and the PtdIns pool was labeled with [ 3 H]inositol for an additional 24 h. The cells were harvested and extracted, and the amount of label incorporated into PtdIns was determined. labeled precursors into PtdIns as would be expected if the supply of phosphatidic acid was limiting.
Alternatively, our data are consistent with diacylglycerol kinase as a potential rate controlling enzyme in the production of PtdIns. Diacylglycerol kinase would provide phosphatidic acid for utilization by Cds and downstream incorporation into PtdIns. There are several isoforms of diacylglycerol kinase, each of which contain distinct domains postulated to regulate its activity and mediate protein:protein interactions at specfic cellular locals (34,55). The existence of multiple diacylglycerol kinase isoforms suggests a diversity of functions for this enzyme, one of which may be to divert diacylglycerol to phosphatidic acid for utilization by the Cds/Pis pathway. Thus, the diacylglycerol pool is partitioned between the major phospholipid classes, phosphatidylcholine and phosphatidylethanolamine, by the regulated synthesis of CDP-choline and CDPethanolamine, whereas a diacylglycerol kinase activity may control the partitioning of diacylglycerol for PtdIns synthesis.
There are two studies that indicate Cds is important in controlling the activity of the PtdIns cycle in signal transduction systems. Wu et al. (14) showed that Drosophila mutants defective in the isoform of Cds expressed in photoreceptors cannot sustain a light-activated current due to rapid depletion of phosphatidylinositol-4,5-bisphosphate. Furthermore, Cds overexpression increased the amplitude of the light response, indicating that Cds activity regulated the availability of phosphatidylinositol-4,5-diphosphate to phospholipase C. Weeks et al. (19) constructed stable cell lines that expressed twice the normal levels of human Cds1. Such cells exhibited increased secretion of tumor necrosis factor-␣ and interleukin-6 in response to stimulation with interleukin-1␤, suggesting that even modest levels of Cds overexpression amplify cellular signaling systems. Our results, together with these observations from other groups, point to a determinant role for Cds enzyme level in regulating the rate of PtdIns turnover rather than controlling the rate of de novo PtdIns biosynthesis. It will be important to determine if the other Cds isoform, Cds2 (Fig. 1), is responsible for regulating the de novo component of PtdIns metabolism.