Molecular Heterogeneity of Phospholipase D (PLD)

Phospholipase D (PLD) has emerged as an important enzyme involved in signal transduction, vesicle trafficking, and membrane metabolism. This report describes the cloning and expression of a new Arabidopsis PLD cDNA, designated PLDγ, and the regulation of PLDγ, -β, and -α by phosphatidylinositol 4,5-bisphosphate (PIP2) and Ca2+. The PLDγ cDNA is 3.3 kilobases in length and codes for an 855-amino acid protein of 95,462 Da with a pI of 6.9. PLDγ shares a 66% amino acid sequence identity with PLDβ, but only a 41% identity with PLDα. A potential N-terminal myristoylation site is found in PLDγ, but not in PLDα and -β. Catalytically active PLDγ was expressed inEscherichia coli, and its activity requires polyphosphoinositides. Both PLDγ and -β are most active at μm Ca2+ concentrations, whereas the optimal PLDα activity requires mm Ca2+concentrations. Binding studies showed that the PLDs bound PIP2 in the order of PLDβ > PLDγ > PLDα. This binding ability correlates with the degree of conservation of a basic PIP2-binding motif located near the putative catalytic site. The binding of [3H]PIP2 was saturable and could be competitively decreased by addition of unlabeled PIP2. Neomycin inhibited the activities of PLDγ and -β, but not PLDα. These results demonstrate that PLD is encoded by a heterogeneous gene family and that direct polyphosphoinositide binding is required for the activities of PLDγ and -β, but not PLDα. The different structural and biochemical properties suggest that PLDα, -β, and -γ are regulated differently and may mediate unique cellular functions.

Phospholipase D (PLD 1 ; EC 3.1.4.4) hydrolyzes phospholipids to produce phosphatidic acid, which may serve as a signaling messenger, a mediator of vesicular trafficking, or a key intermediate in glycerolipid metabolism and membrane remodeling (1,2). This enzyme was first discovered in plants 50 years ago, but a clear understanding of its regulatory mechanisms is still lacking. Calcium has been thought to be a regulator of plant PLD, but the requirement for nonphysiological mM levels of Ca 2ϩ by the PLDs previously purified from various plants has made Ca 2ϩ regulation an enigma. Recent advances in the molecular understanding of PLD have brought new insights into the control and cellular roles of PLD. An intracellular PLD was first cloned from castor bean (3), and the cloning of PLDs has since been reported from Arabidopsis (4,5), rice, maize (6), yeast (7), human (8), and mouse (9). Analysis of the PLD sequences has led to the identification of probable catalytic and regulatory domains of PLD (10,11). Two PLDs with distinct regulatory properties have been cloned from Arabidopsis: one is the conventional mM Ca 2ϩ -requiring PLD, designated PLD␣, and the other, named PLD␤, requires M Ca 2ϩ and polyphosphoinositides for activity (4,5,12).
The polyphosphoinositide activation of plant PLDs is a property shared by PLDs cloned from mammals and yeast (8,9,13). This property has been proposed to be physiologically important because, in addition to being precursors for signaling messengers, polyphosphoinositides themselves can also modulate the functions of proteins through direct binding. Such an interaction is believed to either recruit or activate proteins essential for signaling or membrane trafficking pathways, thus allowing coordinated regulation of different cellular processes (14). However, the mechanism by which polyphosphoinositides affect PLD activities has not been elucidated. Direct phosphatidylinositol 4,5-bisphosphate (PIP 2 ) binding has not been demonstrated previously for PLDs from any source.
In this report, we first describe the cloning and molecular analysis of a new Arabidopsis PLD, PLD␥, whose activity is regulated by polyphosphoinositides. To delineate the mechanisms of the polyphosphoinositide modulation of PLDs, we then provide evidence indicating that plant PLDs directly bind PIP 2 and that this binding is required for the activities of PLD␥ and -␤, but not PLD␣. Additionally, the regulation by Ca 2ϩ and the domain structures of plant PLD␥, -␤, and -␣ are compared. 3 H]inositol 4,5-bisphosphate and dipalmitoylglycero-3-phospho[methyl-3 H]choline were products of Amersham Corp. PC from soybean and PIP 2 from bovine brain were purchased from Sigma and American Radiolabeled Chemicals (St. Louis, MO), respectively. Other lipids were obtained from Avanti Polar Lipids.

Lipid Materials-Dipalmitoylglycero-3-phospho[inositol-2-
Cloning and Sequencing of the PLD␥ cDNA-An expression sequence-tagged cDNA clone, ϳ1 kilobase in length, was identified as a putative incomplete new Arabidopsis PLD cDNA by searching the Gen-Bank™ Data Bank against the cloned PLD sequences. A full-length PLD cDNA, named PLD␥, was isolated from a ZAPII cDNA library, constructed from 3-6-kilobase mRNA isolated from hypocotyls of 3-dayold Arabidopsis seedlings (15). The library was screened using the expression sequence-tagged cDNA clone as a probe with hybridization conducted at 65°C. The subsequent DNA manipulation of the positive clones, sequencing, and sequence analysis were based on previously described procedures (3). The final sequence was determined from both * This work was supported by National Science Foundation Grant IBN-9511623 and United States Department of Agriculture grants (to X. W.). This is Contribution 97-425-J from the Kansas Agricultural Experiment Station. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF027408.
strands. The expression sequence-tagged clone and cDNA library were kindly provided by the Ohio State University Arabidopsis Information Center.
Expression of PLD cDNAs in Escherichia coli-The expression of the PLD␥ cDNA was performed in two constructs: pBluescript SK(Ϫ) and the vector pGEX-2T (Pharmacia Biotech Inc.) containing the sequence encoding the glutathione S-transferase (GST) fusion protein. Oligonucleotides with an added BamHI site were used as primers to amplify a 2.9-kilobase fragment by polymerase chain reaction. The 5Ј-end primer corresponded to nucleotides 342-360 of the PLD␥ cDNA. After BamHI digestion, the insert was ligated into pBluescript SK or pGEX-2T, and the recombinant plasmids were transformed into E. coli JM109. The expression of PLD␥ was induced using the same procedure as described for PLD␣ and -␤ (3,5). The cells were lysed by sonication in the resuspension buffer, and cell debris was removed by centrifugation at 10,000 ϫ g for 5 min. Protein content in the supernatant was measured by the Bradford method according to the manufacturer's instructions (Bio-Rad). Aliquots of the supernatant were assayed for PLD activity or stored at Ϫ80°C until use.
In addition to the GST-PLD␥ fusion protein, GST-PLD␣ and GST-PLD␤ also were used in this study. The construct of GST-PLD␣ was reported previously (3), and the GST-PLD␤ fusion protein was generated by excising the PLD␤ cDNA insert in pBluescript SK (5) with EcoRI digestion and then ligating it into pGEX-2T. The procedures for induction and protein harvest were the same as those described for the pBluescript constructs (5), except that 0.2 mM isopropyl-1-thio-␤-Dgalactopyranoside was used for induction.
Purification of GST Fusion Proteins-GST-PLD␣, GST-PLD␤, GST-PLD␥, and GST alone were purified as described (16). Briefly, induced bacteria were pelleted and rinsed in a buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 2 mM EDTA by centrifugation at 3000 ϫ g for 10 min. The pellet was resuspended in rinsing buffer containing 2 g/ml antipain, 2 g/ml aprotinin, 2 g/ml pepstatin, and 0.5 mg/ml lysozyme and was incubated on ice for 15 min. Dithiothreitol was then added to a final concentration of 5 mM, and the samples were briefly sonicated on ice. To solubilize the expressed protein, N-laurylsarcosine was added to 1.5% (v/v), and then the samples were vortexed and centrifuged at 10,000 ϫ g for 5 min. The supernatant was transferred to a new tube, and Triton X-100 was added to a final concentration of 4% (v/v). Samples of GST-PLD␣, GST-PLD␤, and GST-PLD␥ with equal amounts of GST activity were then absorbed to swollen glutathioneagarose beads (50%, v/v) overnight at 4°C. The beads were rinsed with 15 bed volumes of phosphate-buffered saline. Both the GST activity and protein content of the rinsed beads were measured. GST activity was measured using a 1-chloro-2,4-dinitrobenzene assay according to the manufacturer's instructions (Pharmacia Biotech Inc.) and calculated in units of ⌬A 340 /min/ml. Protein content was determined by the Bradford method using purified GST fusion proteins eluted from the glutathione beads with 10 mM glutathione.
Phospholipid Binding-Phospholipid binding by the GST-PLD proteins and GST was performed according to a method adapted from Davletov and Sudhof (17). Stock vesicles of PIP 2 were prepared by mixing 0.9 Ci of [ 3 H]PIP 2 with 1.0 mol of unlabeled PIP 2 in chloroform, and the solvent was evaporated under a stream of N 2 . The phospholipids were dispersed in 0.5 ml of H 2 O by sonication at room temperature. The fusion proteins attached to the glutathione-agarose beads (30-l wet volume) were incubated with 20 nmol of [ 3 H]PIP 2 vesicles in buffer (50 mM MES (pH 7), 0.5 mM MgCl 2 , and 80 mM KCl) containing 1 mM EGTA and 0.1 or 5 mM Ca 2ϩ in a final volume of 100 l. After incubation at room temperature for 30 min with gentle agitation, PIP 2 bound to PLDs was pelleted with the affinity beads by centrifugation at 2000 ϫ g. The pellet was washed three times with 1 ml of the incubation buffer, and PIP 2 bound to PLDs was measured by scintillation counting of [ 3 H]PIP 2 . GST bound to glutathione-agarose was used as the control for background PIP 2 binding. Binding was expressed as dpm/unit of GST activity (dpm/⌬A 340 /min). The above method was also used to examine the binding of [ 3 H]PC to the GST fusion proteins.
In competition experiments, 20 nmol of [ 3 H]PIP 2 was incubated with GST-PLD fusion proteins attached to glutathione beads in the presence of 1 mM EGTA, and then 5, 20, or 40 nmol of unlabeled PIP 2 was added to a final volume of 100 l. The incubation, rinsing, and quantitation were performed as described above.
Generation of PLD Antibodies and Immunoblotting-A 14-amino acid peptide was synthesized that consisted of a cysteine and 13-amino acids corresponding to residues 838 -851 of PLD␥. The peptide was conjugated to keyhole limpet hemocyanin and used as an antigen to raise antibodies in rabbits (5). For immunoblot analysis, proteins were separated on 8% SDS-polyacrylamide gels, transferred onto a poly-vinylidene difluoride membrane, and incubated with antiserum that contained JM109 lysate to remove nonspecific reactive bacterial proteins. The immunoblot analysis was performed as described (5).
Assaying PLD Activity in the Presence of PIP 2 -PIP 2 -dependent PLD activity (see Figs. 3A, 4, and 5) was assayed using the conditions described previously (5) with some modifications. The basic assay mixture contained 100 mM MES (pH 7.0), 0.5 mM MgCl 2 , 80 mM KCl, 0.4 mM lipid vesicles, and 30 g of pBluescript SK-expressed protein in a total volume of 100 l. CaCl 2 was added to the reactions at the concentrations noted below. Lipid vesicles contained 35 nmol of PE, 3 nmol of PIP 2 , and 2 nmol of PC. PLD-mediated hydrolysis of PC was measured using either 1-palmitoyl-2-oleoyl[oleoyl-1-14 C]glycero-3-phosphocholine or dipalmitoylglycero-3-phospho-[methyl-3 H]choline as substrate. The reaction was initiated by addition of substrate and incubated at 30°C for 30 min in a shaking water bath. Choline, PA, and phosphatidylethanol produced in the PLD reaction were separated and quantitated as described previously (12). In the phospholipid activation experiments, PIP 2 was replaced with 3 nmol of PE, PA, PG, PS, PI, or PIP. In the Ca 2ϩ dependence experiments, varying amounts of Ca 2ϩ were added to the reaction mixture, and the free Ca 2ϩ was determined using arsenazo III, a Ca 2ϩ -sensitive dye (18), in reconstituted reaction mixtures. Control assays were performed using 30 g of protein from lysed bacteria harboring the pBluescript SK(Ϫ) plasmid without a PLD cDNA insert. The background activity from bacteria was very low (5) and was subtracted from the activity of the samples containing the recombinant PLDs.
Assaying PLD Activity in the Absence of PIP 2 -The PIP 2 -independent PLD assay (see Fig. 3, B and C) used a reaction mixture of 100 mM MES (pH 6.5), 0.5 mM SDS, 1% (v/v) ethanol, 30 g of protein, and 0.4 mM PC (egg yolk) containing dipalmitoylglycero-3-phospho[methyl-3 H]choline. CaCl 2 was added to reactions as noted below. The substrate preparation, reaction conditions, and separation of products were performed as described previously (3). Release of [ 3 H]choline into the aqueous phase was quantitated by scintillation counting. Control assays used 30 g of protein from lysed bacteria harboring the pBluescript SK(Ϫ) plasmid without a PLD cDNA insert. The background PLD activity was negligible (3) and was subtracted from the activity of the samples containing the recombinant PLDs.
DNA Isolation and Southern Blotting-Genomic DNA was isolated from Arabidopsis leaves and digested with different restriction enzymes. Full-length cDNAs of PLD␣, -␤, and -␥ were used as probes to hybridize the digested DNA at 65°C under the previously described conditions (3).
Protein Extracts from Arabidopsis-PLD␣-deficient plants were generated by introducing a PLD␣ antisense cDNA to Arabidopsis (12). Microsomal membranes were obtained from leaves of the PLD␣ antisense gene-suppressed and wild-type plants and then extracted with 0.44 M KCl following a described method (12). A previous study showed that most of the PIP 2 -dependent PLD activity from the antisense plants was recovered in the salt extract (12). The salt-extracted PLD was assayed in the presence of varying concentrations of neomycin.

Cloning of the PLD␥ cDNA and Sequence Comparison with
Other PLDs-An Arabidopsis expression sequence-tagged cDNA was identified as a putative, new PLD cDNA by searching the GenBank™ Data Bank against the plant PLD cDNAs. This cDNA, ϳ1 kilobase in length, was used as a probe to screen a ZAPII Arabidopsis cDNA library constructed using 3-6-kilobase mRNA from hypocotyls of 3-day-old seedlings (15). A full-length PLD cDNA, designated PLD␥, was identified and sequenced from both strands. The PLD␥ cDNA is composed of 3234 nucleotides, with the longest open reading frame from nucleotides 312 to 2876, and imparts a deduced protein sequence of 855 amino acids (Fig. 1). The PLD␥ polypeptide has a calculated molecular mass of 95,462 Da and a pI of 6.9. The previously cloned PLD␣ from Arabidopsis consists of 809 amino acids with a pI of 6.0, and PLD␤ has 968 amino acids and a pI of 7.9.
The amino acid and nucleotide sequences of PLD␥ are more closely related to PLD␤ than to PLD␣ from Arabidopsis. PLD␥ has a 66% amino acid identity to PLD␤ and an overall similarity of 81%. In contrast, PLD␥ displays only 41% identity and 60% overall similarity to PLD␣ from Arabidopsis. The C-terminal regions of the three PLDs share a higher degree of amino acid sequence similarities than their N-terminal residues. Phylogenetic alignments of PLD sequences from various sources show that PLD␥ and -␤ form a cluster that is divergent from PLD␣ cloned from Arabidopsis, castor bean, rice, and maize. PLD␣ from different plant species shares ϳ75-90% amino acid sequence identity. PLD␥ and -␤ are more closely related than PLD␣ to the PLDs cloned from human and yeast.
Domain Organization of PLDs-Sequence analysis indicates that PLD␥, -␤, and -␣ share some common structural motifs and differ in others (Fig. 1). PLD␥ possesses a myristoylation consensus sequence (19), MGXXXS, that is not present in PLD␣ or -␤. PLD␥ contains a Ca 2ϩ /phospholipid-binding C2 domain near its N terminus, and this domain is conserved in all cloned plant PLDs, but not in yeast and human PLDs. C2 domains have been identified in a number of proteins involved in signal transduction and membrane trafficking and are important in Ca 2ϩ -regulated binding to phospholipids (10,20). Ca 2ϩ binding is coordinated by four to five amino acid residues provided by bipartite loops within the C2 domain (21,22). PLD␥, like PLD␤, conserves all of the calcium-coordinating acidic amino acids, whereas two of the acidic residues in the C2 domain of PLD␣ are substituted by either positively charged or neutral amino acids, indicating a loss of affinity for Ca 2ϩ in PLD␣. Like all PLDs cloned from plants, yeast, and human, PLD␥ possesses two HXKXXXXD motifs (Fig. 1). It has been hypothesized that the conserved histidine, lysine, and aspartate residues form a catalytic triad responsible for the formation or hydrolysis of phosphoester bonds (11).
PLD␥ and the previously cloned PLD␤ are activated by polyphosphoinositides (Ref. 5 and results below). A motif rich in basic amino acids ((K/R)(X)XXXKX(K/R)(K/R)) has been found to be responsible for polyphosphoinositide binding in proteins such as the actin-severing proteins gelsolin and villin and phospholipase C (23)(24)(25). This motif (RXXXXKXRR) and an inverted sequence (RKXRXXXXR) are present in PLD␤ near the catalytic domain of the C terminus. Three of these four consensus basic residues are conserved in PLD␥, whereas PLD␣ shows the least conservation of amino acids, some of which are replaced by acidic residues (Fig. 1).
Distinct Genes for PLD␥, -␤, and -␣-PLD␥ is encoded by a gene distinct from those of PLD␣ and -␤ because hybridization of the digested Arabidopsis genomic DNA with the three different PLD probes yielded unique banding patterns (Fig. 2). The PLD␥ cDNA has one XbaI site and no site for KpnI, XhoI, and BamHI. One hybridization band was observed in the KpnI-, XhoI-, and BamHI-digested DNAs, whereas three bands were present in the XbaI-cut DNA (Fig. 2). The greater than predicted number of bands in the XbaI digestion could be caused by the presence of this restriction site in the intron sequences of the PLD␥ gene and/or by the presence of another PLD␥ or closely related gene.
Expression of Catalytically Active PLD␥ in E. coli-To verify that the cloned cDNA encodes a PLD, protein from this cDNA was produced in E. coli using pBluescript SK as an expression vector. After isopropyl-1-thio-␤-D-galactopyranoside induction, a unique immunoreactive band from bacterial extracts was recognized by antiserum generated against a synthetic peptide of the deduced PLD␥ amino acid sequence (data not shown). The expressed protein hydrolyzed PC in the presence of PE, PIP 2 , and M Ca 2ϩ (Fig. 3A). PLD␥ displayed the transphosphatidylation activity that is the signature of PLD enzymes (data not shown). In contrast, PLD␥ exhibited no PC hydrolytic activity when tested under a condition widely employed to assay the conventional plant PLD activity (Fig. 3B), which is now known to be PLD␣ (12). This assay used PC-only vesicles, 25 mM Ca 2ϩ , and 0.5 mM SDS. This property of PLD␥ was shared by PLD␤ (Fig. 3B). On the other hand, PLD␣ was active under such conditions (Fig. 3B). When these PLDs were assayed in the presence of 50 M Ca 2ϩ , SDS, and PC vesicles, none of them was active (Fig. 3C). This result indicated that the lack of PLD␤ and -␥ activities in the conventional PLD assay was not due to inhibition by mM Ca 2ϩ . On the other hand, the lack of PLD␣ activity at 50 M Ca 2ϩ was due to its requirement for higher Ca 2ϩ concentrations (Fig. 4).
Ca 2ϩ Dependence of PLD␥, -␤, and -␣-To compare the effect of Ca 2ϩ on different PLDs, PLD␥, -␤, and -␣ activities were measured in the presence of varying amounts of Ca 2ϩ . Compared with PLD␤ and -␣, PLD␥ displayed an ϳ2-fold higher basal activity in the absence of Ca 2ϩ , and it showed the smallest activation in response to Ca 2ϩ (Fig. 4). PLD␥ activity peaked at four times its basal level as Ca 2ϩ was increased to ϳ50 M. As Ca 2ϩ concentrations increased above 50 M, PLD␥ activity decreased and showed a near complete inhibition at 20 mM Ca 2ϩ . PLD␤ exhibited an optimal activation over a broader range of Ca 2ϩ concentrations than PLD␥, and its activity at 50 M Ca 2ϩ was ϳ10-fold higher than its basal activity. In addition, PLD␤ activity was not as completely inhibited by mM levels of Ca 2ϩ . PLD␣, in striking contrast to PLD␥ and -␤, displayed little activity at M Ca 2ϩ concentrations, but showed a marked increase in activity as Ca 2ϩ concentrations increased to the mM range.
The present Ca 2ϩ dependence study used an unbuffered system in which the free Ca 2ϩ was determined using the Ca 2ϩsensitive dye arsenazo III in reconstituted reaction mixtures. In this system, the Ca 2ϩ optimum of PLD␤ (50 M) was higher than that (0.5-1 M) reported in a previous study (5), in which the free Ca 2ϩ was calculated according to a Ca 2ϩ -EGTA buffer system. When the free Ca 2ϩ level in the buffered system was evaluated using arsenazo III, it was found that the presence of EDTA in the enzyme sample buffer interfered with the Ca 2ϩ buffering system, particularly at the low M Ca 2ϩ range. This interference resulted in a shift of the PLD␤ Ca 2ϩ optimum to an apparently lower free Ca 2ϩ concentration.
Activation of PLDs by Polyphosphoinositides-The inability of PLD␥ to hydrolyze PC in PC-only vesicles and its activity in the presence of mixed vesicles (Fig. 3) suggested that PLD␥ might require polyphosphoinositides for activity. To examine this possibility, PLD␥ was assayed with mixed lipid vesicles, PC/PE/X (2:35:3 mole ratio), where PE and PC were kept constant and X was PIP 2 , PIP, PI, PG, PA, PE, or PS. The highest PLD␥ activity was found in the PIP 2 -containing vesicles. PIP gave ϳ70% of the PIP 2 -stimulated activity, whereas PI, PS, PG, and PA showed much less of a stimulatory effect (Fig. 5A). PLD␤ showed a significant activation in the presence of the polyphosphoinositides PIP and PIP 2 , but not any other phospholipid. PLD␣, which has been well documented to be capable of hydrolyzing PC without any additional phospholipids (Fig.  3B), was also stimulated by PIP 2 and PIP at a suboptimal Ca 2ϩ concentration (5 mM) and in the absence of the detergent SDS.
The mode of PIP 2 stimulation of the PLDs was tested by examining their ability to be inhibited by neomycin, which is a high affinity ligand that selectively binds to polyphosphoinositides. Neomycin inhibited the PIP 2 -activated activities of PLD␥ and -␤ in a concentration-dependent manner, but did not inhibit the PIP 2 -stimulated PLD␣ activity (Fig. 5B). Such differential effects of neomycin on PLD isoforms were also observed in PLDs extracted from Arabidopsis. The expression of PLD␣ in leaves was suppressed by introducing a PLD␣ antisense gene into Arabidopsis (12). PLDs extracted from the PLD␣ antisense gene-suppressed leaves required PIP 2 for activity and were almost completely inhibited by neomycin. However, neomycin inhibited only ϳ30% of the PLD activity from wild-type Arabidopsis (Fig. 5C). Thus, the remaining 70% of the PIP 2 -stimulated activity came from PLD␣ and was refractory to neomycin. PIP 2 Binding by PLDs-One explanation for the different effects of neomycin on these PLDs is that PIP 2 stimulates PLD␥ and -␤ by directly binding to these enzymes as a cofactor, whereas the PIP 2 enhancement of PLD␣ is due to a physical alteration of the substrate vesicles. To establish whether or not there was a direct, specific interaction between PIP 2 and the PLD isoforms, recombinant proteins consisting of PLD␣, -␤, or -␥ fused to GST were purified and assayed for their PIP 2 binding ability (Fig. 6). These affinity-purified, glutathioneagarose-attached proteins were incubated with [ 3 H]PIP 2 in the absence or presence of Ca 2ϩ . GST bound to glutathione-agarose was used as a background control and showed little PIP 2 binding. The three GST-PLD fusion proteins displayed different PIP 2 binding abilities in the absence of Ca 2ϩ . PLD␤ bound ϳ2-fold more PIP 2 than PLD␣ and greater than one-third more than PLD␥ (Fig. 6A), which represented binding of ϳ9.8% of the total PIP 2 added for GST-PLD␤, 7.1% for GST-PLD␥, and 5.5% for GST-PLD␣. The PIP 2 binding did not require Ca 2ϩ , and inclusion of 5 mM Ca 2ϩ almost completely abolished PIP 2 binding by the PLDs. Addition of 100 M Ca 2ϩ decreased PIP 2 binding by PLD␤ and -␥, but not by PLD␣. PIP 2 binding by the GST-PLD fusion proteins increased with increasing amounts of PIP 2 (Fig. 6B). The initial binding response was linear for each of the GST-PLD fusion proteins. Above 10 -20 nmol of PIP 2 , these GST-PLD fusion proteins showed a decreased PIP 2 binding ability, indicating that this binding was saturable. Scatchard plot analysis showed that the PLDs were half-maximally saturated at 19, 26, and 13 nmol for GST-PLD␥, GST-PLD␤, and GST-PLD␣, respectively. The binding of [ 3 H]PIP 2 by the GST-PLD fusion proteins could be competitively decreased when unlabeled PIP 2 was added and the total PIP 2 approached saturating levels (Fig. 6C).
The lipid binding specificity of the PLDs was examined by comparing [ 3 H]PIP 2 and [ 3 H]PC binding in the presence of the same specific radioactivity and 1 mM EGTA. The GST-PLD fusion proteins bound very little PC (Fig. 6D). After subtracting the GST background, the PLDs bound 8 -10-fold more PIP 2 than PC. As a known substrate, PC must bind to these enzymes, but the binding may be Ca 2ϩ -dependent as their activities require Ca 2ϩ (Fig. 4). Under the present conditions tested, the lack of PC binding can not be attributed to PC hydrolysis by the enzymes because PLD␣, -␤, and -␥ are inactive in the absence of Ca 2ϩ and PIP 2 (Figs. 3 and 4). This result indicates that the binding of PIP 2 is not due to a nonspecific hydrophobic interaction between the phospholipid and the GST-PLD fusion proteins.

DISCUSSION
This study has identified a new PLD, PLD␥, that is encoded by a gene distinct from the previously cloned PLD␣ and -␤. PLD␥ is more closely related to PLD␤ than to PLD␣ in terms of overall similarities in nucleotide and amino acid sequences and activation by PIP 2 and Ca 2ϩ . Differences do exist between PLD␥ and -␤ in size, isoelectric point, phospholipid activation, Ca 2ϩ concentration dependence, and PIP 2 binding ability. The amino acid sequences in the catalytic C-terminal regions of the three PLDs display a higher degree of similarity than their regulatory N-terminal residues.
A feature unique to PLD␥ is the presence of a putative myristoylation site, MGXXXS, near its N terminus. Myristoylation of many proteins is thought to represent a mechanism for reversible membrane association (26 -28). It is believed that the hydrophobic nature of myristate allows it to associate directly with membrane lipids, but there are proteins, such as mammalian cAMP-dependent kinase, in which myristoylation has no membrane-targeting function, but is instead important for structural stabilization (29). In some cases, it has been reported that myristate interacts with a specific membrane protein, thus mediating a targeted protein-protein interaction (30). The putative myristoylated N-terminal glycine is ϳ30 amino acids away from the first ␤-sheet strand in the C2 domain of PLD␥. Since myristoylation and C2 domains of many proteins are involved in protein-membrane interactions and since the putative myristoylation site and C2 domain of PLD␥ are very close together, it is possible that a covalently linked myristate and the C2 domain coordinately regulate the association of PLD␥ with membranes. Similar myristoylation consensus sequences have been observed in other plant proteins such as certain Ca 2ϩ -dependent protein kinases (31). However, this type of covalent modification is not well understood in plants. It is known that for myristoylation to occur by the yeast Nmyristoyltransferase, the initiator methionine preceding the targeted glycine residue has to be removed (19). The MGXXXS sequence in PLD␥ is 12 amino acids downstream of the first methionine in the continuous open reading frame, but the true initiation site for PLD␥ remains to be experimentally determined. Efforts are underway to determine whether PLD␥ is myristoylated and, if so, what the functional consequences of this covalent modification are.
The results have demonstrated clearly that the activities of PLD␥, -␤, and -␣ are regulated differently by Ca 2ϩ and polyphosphoinositides. PIP 2 is a proposed regulator of PLDs from plants (5,12), animals (8,9), and yeast (13), but the mode of this PIP 2 effect is unclear. This study has provided evidence that plant PLDs bind PIP 2 and that this binding is required for PLD␤ and -␥ activities. This conclusion is supported by several lines of evidence. A direct indication is the association of radioactive PIP 2 with these PLDs (Fig. 6). That inhibition of PIP 2 binding by neomycin or mM Ca 2ϩ also inhibits PLD␤ and -␥ activities (Figs. 4 and 5) provides correlative evidence for this conclusion. The sensitivity of PIP 2 binding by PLD␤ and -␥ to cationic molecules such as neomycin and Ca 2ϩ suggests that these molecules compete with the basic amino acids of the polyphosphoinositide-binding motif. The requirement for Ca 2ϩ and direct PIP 2 binding for the activity of PLD␥ and -␤ suggests that Ca 2ϩ may serve as both a positive and negative regulator of these enzymes. At concentrations at or below 100 M, Ca 2ϩ is a positive regulator of PLD␥ and -␤ activities. Above 100 M, Ca 2ϩ inhibits the binding of PIP 2 to PLD␥ and -␤, thus negatively regulating their activities.
The presence of basic polyphosphoinositide-binding motifs near the putative catalytic sites (Fig. 1) may provide a structural basis for the observed PIP 2 binding by the PLDs. The extent of PIP 2 binding exhibited by the different PLDs appears to be correlated with the degree of amino acid conservation in the putative PIP 2 -binding motifs. The PIP 2 -binding sequence is also present in the PLDs cloned from human (amino acids 549 -556) and yeast (1421 to 1428), both of which are also stimulated by PIP 2 (8,13). Direct evidence for PIP 2 binding by this basic motif comes from a recently solved crystal structure of phospholipase C␦. This binding motif in phospholipase C␦ stretches from amino acids 434 to 441 as the sequence KILLKGKK (24). Lysines 438 and 440 serve as direct ligands to phosphate groups at the 4-and 5-positions of inositol 1,4,5trisphosphate (22). For phospholipases C, the basic binding motif is part of their catalytic sites because PIP 2 is a substrate. On the other hand, work in this laboratory has shown that PLD␥, -␤, and -␣ do not hydrolyze PIP 2 , 2 and thus, the binding of PIP 2 by PLDs is part of PLD regulation and activation. A detailed mechanism for the role of PIP 2 binding in PLD-mediated hydrolysis remains to be elucidated.
In addition to direct binding, polyphosphoinositides may also enhance PLD activity by an alteration of substrate vesicle or membrane structure. This is evident from the PIP 2 stimulation of PLD␣ activity. It has been well documented that PLD␣ is most active in the presence of mM Ca 2ϩ and detergents, and this activity is independent of polyphosphoinositides (3,32). At suboptimal Ca 2ϩ (5 mM) and with no detergent, PLD␣ is stimulated significantly by PIP 2 . This stimulation does not require direct PIP 2 binding because it is insensitive to neomycin (Fig.  5B) and is not correlated with the extent of PIP 2 bound to PLD␣ (Fig. 6A). Although PLD␣ binds PIP 2 at lower Ca 2ϩ concentrations, this binding is almost absent at 5 mM Ca 2ϩ (Fig. 6A) whereas PLD␣ activity increases at this Ca 2ϩ concentration (Fig. 4). One explanation for these results is that there are different locations for the PIP 2 -binding sites on the PLDs. One is the polyphosphoinositide-binding motif near the C termini of PLD␤ and -␥, which is absent in PLD␣. PIP 2 binding at this site may be required for catalysis. The other is the C2 domain, which is known to bind acidic phospholipids (20,21). PIP 2 binding at this less specific site may be involved in regulating the membrane association of all three PLDs (20). Furthermore, PLD␣ contains a basic amino acid motif in the middle of its C2 domain, stretching from amino acids 58 to 66 as the sequence KARVGRTRK (Fig. 1) The present results also suggest that, in addition to PIP 2 , other inositol-containing lipids may be involved in PLD regulation. All three plant PLDs are stimulated by PIP, albeit to an lesser extent than by PIP 2 (Fig. 5A). The potential exists for inositol phospholipids to serve as a key point of intersection for several phospholipid signaling pathways. For instance, in addition to being an activator of PLD, PIP 2 is a substrate for phospholipase C. It is known that PA, the lipid product of PLD, stimulates the synthesis of PIP 2 from PIP by PI-4-phosphate 5-kinase (33). Such an activation may have a dual signaling effect by providing an activator of PLD and a substrate for PLC.
Prior findings from this laboratory have demonstrated that two types of PLD activities can be selectively assayed in plant extracts: one is polyphosphoinositide-independent and mM Ca 2ϩ -requiring, and the other requires polyphosphoinositides and M Ca 2ϩ (12). The present results show that while the former activity comes from PLD␣, the latter activity can result from PLD␤, PLD␥, or both. The finding of PLD as a family of heterogeneous enzymes indicates that PLD in the cell is regulated by complex mechanisms and that the physiological roles of PLD are diverse. It is possible that the different PLDs are involved in distinct cellular processes such as membrane degradation, lipid mobilization, vesicular trafficking, and signal transduction. Further investigations on the substrate specificity, spatial and temporal distribution, in vivo activations, and genetic manipulation of these isoforms will be important to better understand the control and function of PLDs.