The Role of Interfacial Binding in the Activation ofStreptomyces chromofuscus Phospholipase D by Phosphatidic Acid*

The Streptomyces chromofuscusphospholipase D (PLD) cleavage of phosphatidylcholine in bilayers can be enhanced by the addition of the product phosphatidic acid (PA). Other anionic lipids such as phosphatidylinositol, oleic acid, or phosphatidylmethanol do not activate this PLD. This allosteric activation by PA could involve a conformational change in the enzyme that alters PLD binding to phospholipid surfaces. To test this, the binding of intact PLD and proteolytically cleaved isoforms to styrene divinylbenzene beads coated with a phospholipid monolayer and to unilamellar vesicles was examined. The results indicate that intact PLD has a very high affinity for PA bilayers at pH ≥ 7 in the presence of EGTA that is weakened as Ca2+ or Ba2+ are added to the system. Proteolytically clipped PLD also binds tightly to PA in the absence of metal ions. However, the isolated catalytic fragment has a considerably weaker affinity for PA surfaces. In contrast to PA surfaces, all PLD forms exhibited very low affinity for PC interfaces with an increased binding when Ba2+ was added. All PLD forms also bound tightly to other anionic phospholipid surfaces (e.g. phosphatidylserine, phosphatidylinositol, and phosphatidylmethanol). However, this binding was not modulated in the same way by divalent cations. Chemical cross-linking studies suggested that a major effect of PLD binding to PA·Ca2+ surfaces is aggregation of the enzyme. These results indicate that PLD partitioning to phospholipid surfaces and kinetic activation are two separate events and suggest that the Ca2+ modulation of PA·PLD binding involves protein aggregation that may be the critical interaction for activation.

Mammalian phospholipase D enzymes have a complex intermediary role in many well characterized signal transduction pathways involving membrane-linked and cytosolic soluble signaling pathways (1). Most PLD 1 enzymes have a high affinity for anionic phospholipids, and all appear to require Ca 2ϩ for catalytic activity. Regulation of this class of enzymes is com-plex: protein kinase C (2), ARF (3), and Rho (2) proteins are activators of PLD. The lipophilic product of PLD cleavage, phosphatidic acid (PA), acts as a second messenger in cells and has been shown to activate phosphatidylinositol-specific phospholipase C-␥1 (4), inhibit adenylate cyclase (5), and mobilize intracellular Ca 2ϩ (6,7).
The PLD secreted by Streptomyces chromofuscus is considerably smaller than the eukaryotic enzymes, although like those enzymes it requires Ca 2ϩ for activity (8). The bacterial PLD is not involved in signal transduction but has a role in phosphate retrieval; it may also play a role in promoting infections of the organism. The S. chromofuscus PLD is an unusual phospholipase in that it does not exhibit "interfacial activation" (preference for micellar rather than monomeric short chain phospholipid substrate (9)) or "surface dilution" (dependence of enzyme specific activity on the mole fraction of substrate in a micelle or bilayer surface (10)) kinetics (11). A ping-pong-like ordered binding mechanism has been proposed for the enzyme in which the substrate (e.g. phosphatidylcholine) binds and is converted to a covalent phosphatidyl-enzyme with release of the free base (e.g. choline), followed by water attacking the PLD-PA covalent adduct to release phosphatidic acid (12). The nucleophilic attack of water on the distal phosphate ester bond results in cleavage of the P-O bond. The phosphatidyl-enzyme intermediate can be decomposed to a different phospholipid in the presence of a high concentration of a primary alcohol such as ethanolamine, serine, methanol, butanol, etc., a reaction used to generate different head group phospholipids (13,14).
S. chromofuscus PLD is dramatically activated for hydrolysis of PC packed in vesicles by the incorporation of PA (lyso-PA or phosphatidylinositol 4-phosphate) in the vesicle. The presence of PA increases the apparent V max and has little effect on the apparent K m (11). Other nonsubstrate anionic phospholipids such as PI or fatty acids do not activate the enzyme. This activation is allosteric because PA bilayers can activate the PLD toward diC 4 PC, a water-soluble substrate with no tendency to partition into bilayers (11). Two forms of S. chromofuscus PLD have been isolated from culture supernatants (15): intact PLD that is a monomer with a molecular mass of 57 kDa (PLD 57 ) and a tight complex of PLD that has been proteolytically cleaved near the C-terminal portion of the protein (PLD 42/ 20, named for the apparent size of the two subunits on SDS-PAGE). Interestingly, PLD 42/20 is more active than intact enzyme toward PC vesicles and toward monomeric substrates (dihexanoyl-phospholipids). Only the intact PLD 57 can be activated by PA. One possible explanation for PA activation is that the negatively charged product anchors the enzyme to the zwitterionic PC surface for processive catalysis, although this cannot be the only explanation because PA bilayers enhance PLD 57 cleavage of diC 4 PC (11).
The present work is aimed at examining the binding affinity of S. chromofuscus PLD for PA, PC, and other phospholipid surfaces. Two different types of surfaces were used to measure protein binding to phospholipids: (i) hydrophobic beads (composed of styrene divinylbenzene (SDVB)) coated with a monolayer of phospholipid and (ii) unilamellar vesicles. The results show that PLD 57 has a pH-dependent high affinity for POPA bilayers in the presence of EGTA that is weakened (at pH Ն 7) as Ca 2ϩ or other divalent cations are added to the system. The proteolytically clipped PLD 42/20 also binds tightly to POPA in the absence of metal ions at pH 8; isolated PLD 42 has a considerably weaker affinity for POPA surfaces. In contrast to PA surfaces, all PLD forms exhibited very low affinity for PC interfaces with an increased binding when Ba 2ϩ was added. The binding of PLD 57 to PC was enhanced by preincubation of the enzyme with diC 4 PA, consistent with PA binding to an allosteric site and causing a conformational change of the protein to a form with a higher affinity for bilayer surfaces. All PLD forms also bound tightly to other anionic phospholipid surfaces (e.g. PS, PMe, and PI); however, this binding was not modulated by the addition of divalent cations. These results indicate that PLD partitioning to phospholipid surfaces and kinetic activation are two separate events.

MATERIALS AND METHODS
Chemicals-Crude S. chromofuscus PLD (lyophilized powder) was purchased from Sigma. DiC 4 PC, POPC, POPA, POPS, PI, and dioleoylphosphatidylmethanol were purchased from Avanti and used without further purification. Hitrap HIC and Hitrap Q columns and phenyl-Sepharose were purchased from Amersham Pharmacia Biotech. DiC 4 PA and diC 4 PMe were synthesized from diC 4 PC as described previously (11,16). The reaction progress was monitored by 31 P NMR spectroscopy (17), and the final product was characterized by 1 H NMR spectroscopy. SDVB beads were purchased from Seradyn and prepared for coating with phospholipids as described by Cho and co-workers (18).
Purification of PLD Isoforms-Crude PLD from Sigma was fractionated to yield PLD 57 , PLD 42/20 , and PLD 20 using Hitrap HIC and Hitrap Q columns as described previously (15). In addition, chromatography on palmitoyl cellulose equilibrated with 10 mM Tris, pH 8.0, was used to isolate PLD 42 from the other species (20). Protein was eluted with a Triton X-100 gradient from 0 to 0.5 mM; PLD activity of fractions was assayed toward 5 mM diC 4 PC using the pH-stat technique. The PLD species responsible for the activity were identified by SDS-PAGE. Fractions containing PLD were pooled and dialyzed against 10 mM Tris, pH 8.0, to remove the Triton X-100. PLD 57 and PLD 42 were cleanly isolated from this column; PLD 20 was not detected in the eluants from the palmitoyl cellulose column. An alternate purification scheme to isolate PLD 57 and PLD 42 used chromatography of S. chromofuscus supernatant on a phenyl-Sepharose column. PLD fractions were eluted with a gradient of 1.0 to 0 M ammonium sulfate.
Preparation of Vesicles-An appropriate aliquot of phospholipid dissolved in chloroform was evaporated under argon for 10 -20 min. The resulting film was lyophilized and then resuspended in 1 mM EGTA (or EDTA), pH 8.0. For formation of small unilamellar vesicles, aqueous suspensions of the lipid were sonicated (using a Branson W-150 sonifier with a 1-cm-diameter probe) on ice in 3 ϫ 10-min intervals. NaCl was added to the vesicles to a final concentration of 100 mM. Varying amounts of divalent cations (Ba 2ϩ or Ca 2ϩ ) were then added to the vesicles. The vesicles were often stored overnight at 4°C and used the following day; PA vesicles could be stored for 1-2 days as long as the lipid concentration was Ͻ1 mM (vesicles aggregated and fused if stored longer). Higher concentration PA vesicles were used immediately. To control the maximum size of vesicles used in binding assays, all vesicle preparations were extruded through a 0.1-m syringe filter (Gelman Sciences Supor Acrodisc 32 syringe filters) and used directly in binding experiments.
PLD Binding to Phospholipid Vesicles-Partitioning of PLD proteins to phospholipid vesicles was measured by adding 25-50 g of protein to an aliquot of the vesicle solution. The samples were immediately vor-texed, centrifuged, and filtered through an Amicon centricon-100 concentrator. PLD bound to vesicles stayed on the filter; free PLD passed through the filter and into the supernatant. There was always some protein "lost" during filtration (not accounted for on the filter or in the filtrate and presumably adhered to the filter and not easily eluted). This varied from 5% to at most 15% of the total protein. An equal volume of solution before and after filtration was analyzed by SDS-PAGE and by pH-stat for activity toward 5 mM diC 4 PC (the pK a2 of short chain PA is 6.8 so that most of the product PA is titratable at pH 8.0) to measure the amount of PLD bound to the vesicles. Bound PLD (E b ) was calculated by measuring total activity of PLD toward diC 4 PC (5 mM) in the absence of beads (E T ) and comparing it with free PLD activity in the supernatant after filtration (E f ): E b ϭ E T Ϫ E f . Alternatively, intensities of PLD bands on SDS-PAGE before (E T ) and after filtration (E f ) were compared to estimate E b . The K D was derived from fitting the data to the equation where L o is the total phospholipid concentration, and E b is bound enzyme.
PLD Binding to Phospholipid-coated SDVB Beads-Lyophilized beads (100 mg) were suspended in hexane:methanol (1:1) containing 420 M phospholipid. The bead suspension was gently vortexed; the coated beads were recovered by rotary evaporation over a warm (37°C) water bath. Dried coated beads were resuspended in water to a concentration of 60 M phospholipid, sonicated, and then washed until no floating beads remained. The coated beads were lyophilized overnight and then weighed out in tared Eppendorf tubes so that when resuspended in 250 l of buffer the final concentration of phospholipid would range between 60 and 600 M. For binding experiments where the ligand was varied, a fixed amount of enzyme was added to each tube (e.g. for PLD 57 2.5 g/500 l). PMe-, PI-, and POPA-coated beads were incubated with PLD with gentle rocking for 10 min at room temperature. PLD binding to POPC and POPS vesicles could not be examined in the presence of Ca 2ϩ , because the enzyme would be active and generate PA under those conditions. Instead, Ba 2ϩ was used and compared with vesicle binding of the enzymes in the presence of EGTA. Free PLD (E f ) was separated from bound PLD (E b ) by centrifugation; the supernatant was transferred to a new Eppendorf tube. Bound PLD (E b ) was calculated by measuring total activity of PLD toward diC 4 PC (5 mM) in the absence of beads (E T ) and comparing it with free PLD activity in the supernatant after centrifugation (E f ). When the enzyme concentration was fixed and phospholipid concentration was varied, the K D was derived as indicated above. In experiments with fixed phospholipid but varied enzyme, K D was derived from E b ϭ L o /(1 ϩ K D /E f ). Cooperative binding was not considered. All datum points were done in at least duplicate (and often in triplicate).
SDVB beads were not saturated with phospholipid at 60 M but were saturated when prepared in 600 M bulk phospholipid (bead surface saturation by phospholipids occurred between 120 -300 M bulk phospholipid). Therefore, a qualitative comparison of PLD binding to SDVB beads at comparable total phospholipid concentration but different surface saturation was used to check for large differences in PLD affinity that could indicate PLD binding directly to uncoated beads. For a fixed total phospholipid concentration of 420 M, there was only a slight decrease in PLD 57 specific activity in the supernatant (indicating slightly more PLD 57 binding to the beads) when beads were coated with 600 versus 60 M phospholipid. Although PLD 57 binds slightly more tightly to saturated beads, the difference is small and consistent with PLD 57 interacting with the phospholipid monolayer rather than the bead surface. The proteolytically clipped PLD 42/20 bound to beads with a much stronger dependence on the bead coverage. To avoid complications of the PLD 42/20 interacting differentially with the subsaturated beads, binding experiments were run at fixed ligand (420 M) and varying PLD 42/20 .
Cross-linking of PLD-Both PLD 57 and proteolytically cleaved fragments of PLD were examined for aggregate formation in the absence and presence of phospholipids and metal ions using the chemical crosslinking agent EDC, a heterobifunctional imide, obtained from Pierce. The cross-linking reaction was carried out in 100 mM MES, 100 mM NaCl, with the pH between 6 and 7 (average 6.5) in the absence and presence of cations. PLD, ranging from 25 to 75 g, and 10 mM EDC were incubated for various times at 30°C or room temperature. Crosslinking studies with PLD and POPA or PI vesicles used incubation times of 30 min; studies with POPC used 10 -30 min (to try and minimize production of PA by the high concentration of PLD, which is inhibited by Ba 2ϩ but not totally). Excess reagent was quenched by increasing the pH to 9. Samples were analyzed using either a 6.5% denaturing gel or a 7.5-15% gradient denaturing gel.
Protein Concentration-The amount of protein free and bound as well as protein lost (trapped in the centricon filters) was measured with NanoOrange Protein Quantitation Kit (Molecular Probes N-6666). The NanoOrange reagent (bound to protein) was excited at 480 nm, and the emission scanned from 570 -590 nm. A known amount of PLD 57 or some proteolytic fragment was used to construct a standard curve.
PLD Assays-Two methods were used to measure PLD specific activity. To quantify free PLD in the binding studies, a pH-stat assay was used. Hydrolysis of 5 mM diC 4 PC to diC 4 PA was monitored with a Radiometer pH-stat model VIT90 as described previously (15) using 5 mM NaOH as the titrant. For each phospholipid concentration, assays were run in duplicate or triplicate. For assays with unilamellar vesicles as the substrate, either 31 P or 1 H spectra were acquired to monitor PLD activity as described previously (11,15). The assay buffer used was 50 mM imidazole in D 2 O buffer, pH 7.2 (meter reading).

RESULTS
PLD 57 Binding to PA Interfaces-In the presence of EGTA and at pH 8.0, PLD 57 bound quite tightly to the PA-coated SDVB beads; no activity for free PLD 57 was detected in the supernatant under these conditions. Given the error in determining PLD specific activity (15%) and the smallest concentration of PA examined (10 M), K D Ͻ 2 M (Table I). Inclusion of Ca 2ϩ (or Ba 2ϩ ) in the binding buffer dramatically weakened the interaction of PLD with the PA-coated beads, and more free enzyme could be detected in the supernatant (Fig. 1). At a fixed metal ion concentration but with varying PA, the amount of enzyme bound to the beads was quantified and used to calculate a K D (Fig. 2). Increasing the Ca 2ϩ concentration from 2 to 5 mM (in the presence of 1 mM EGTA) increased the PLD 57 K D for the PA from 12 to 215 M (Fig. 2). In the presence of 5 mM Ba 2ϩ instead of Ca 2ϩ , the PLD 57 K D for PA-coated beads was 462 Ϯ 21 M (Table I). However, not all divalent ions were effective in releasing PLD 57 from PA surfaces. In the presence of Mg 2ϩ , PLD remained tightly bound to the PA-coated beads. The apparent affinity of Ca 2ϩ for PA has been described in the literature as ranging from 0.5 to 5 mM. A large number of factors affect this interaction including pH, solution ionic strength, and the presence of divalent cations or cationic peptides. The change in PLD binding to PA in the presence of Ca 2ϩ suggested that Ca 2ϩ may compete very effectively with PLD for binding to PA surfaces.
The tight interaction of PLD 57 with PA surfaces in the presence of EGTA at pH 8.0 was also examined with POPA vesicles and a filtration assay. In Fig. 3, lane 1 shows total enzyme (PLD 57 ) prior to filtration, and lane 5 shows free enzyme after incubation with 0.5 mM POPA vesicles. Clearly, most of the PLD 57 was bound to the vesicle surface in the absence of divalent metal ions.
The effects of pH and ionic strength on PLD partitioning to the bilayer surface were also monitored. The amount of free PLD was estimated from residual PLD activity in the filtrate toward diC 4 PC or by SDS-PAGE of the filtrate compared with sample prior to filtration. As the solution pH was decreased from 8 to 6, there was a significant decrease in the amount of PLD bound to the PA surface (Table II). Because pK a2 for PA is ϳ8.5 in pure PA bilayers (21), the majority of the PA is monoanionic over this pH range and not changing significantly. Between pH 5 (16% bound) and 7 (78% bound), the PA is nearly all monoanionic; hence the large change in binding is not likely to result from changes in PA ionization. A more likely explanation is that a group on the enzyme (possibly a histidine) must be deprotonated for efficient binding of PLD to PA in the absence of Ca 2ϩ . Adding Ca 2ϩ to the solution of PA vesicles enhanced PLD binding at acidic pH values (Table II). This suggests that a group(s) on the enzyme interacts with the metal  ion to lower the pK a of the group that must be deprotonated for optimal binding of the protein to PA.
PLD 57 binding to POPA in the absence of divalent cations was also examined as a function of added NaCl. Increasing the NaCl from 0.1 to 1.0 M caused a large decrease in the amount of enzyme bound to the PA vesicles (Table III). The ability of high NaCl to inhibit PLD binding to PA indicates that the interaction has a large electrostatic component.
PLD 57 Binding to Other Anionic Phospholipid Interfaces-Intact PLD also bound tightly to other anionic phospholipid surfaces, both phospholipid-coated SDVB beads and unilamellar vesicles, in the presence of 1 mM EGTA (Table I). PLD 57 bound to PI, PMe, and PS surfaces with K D Ͻ 2 M. Again, as Ca 2ϩ (or Ba 2ϩ in the case of substrates PS and PMe) was added, the PLD binding affinity decreased for the phospholipid. With the nonsubstrate and noninhibitor PI, PLD 57 binding at pH 8.0 was the same whether Ca 2ϩ or Ba 2ϩ was added (e.g. K D values of 294 Ϯ 20 and 235 Ϯ 21 M were obtained for PI at 4 mM (total cation minus EGTA concentration) Ba 2ϩ and Ca 2ϩ , respectively). As with PA, the interaction of the enzyme with PI was weaker at acidic pH and metal ions enhanced binding at low pH (Table II). Interestingly, PI, PMe, and PS do not activate PLD 57 toward PC vesicles (11). Thus, binding to phospholipid surfaces alone is not sufficient for kinetic activation.
Proteolytically Clipped PLD Binding to PA Interfaces-The proteolytically clipped PLD where the two fragments are still associated was not activated by PA for hydrolysis of PC vesicles (15). How it interacts with PA surfaces could shed light on the mechanism for the kinetic activation. When incubated with 0.5 mM PA vesicles in the absence of divalent metal ions, PLD 42/20 bound well to the PA (Fig. 3, compare lane 4 with lane 8). Using PA-coated SDVB beads (Table I) Addition of Ba 2ϩ reduced the affinity of the PLD fragments for these anionic phospholipid surfaces as it did for the intact PLD.
The 42-kDa fragment alone exhibited much weaker binding to anionic lipid surfaces irrespective of phospholipid identity and surface (Fig. 3, for vesicles compare lane 2 (total PLD 42 ) with lane 6 (PLD 42 in the filtrate); for SDVB beads see Table I). For example, the PLD 42 K D for a PS monolayer in the presence of EGTA was 760 Ϯ 75 M compared with Ͻ2 M for both PLD 57 and clipped but associated PLD 42/20 (Ͻ50 M). In the presence of 5 mM Ba 2ϩ , the K D for PLD 42 binding to PA-coated SDVB beads was 2.0 Ϯ 0.2 mM; the K D for PLD 42/20 was 0.46 Ϯ 0.03 mM. Thus, removal of the 20-kDa fragment weakened PLD binding to anionic phospholipid surfaces. The 20-kDa fragment exhibited the weakest interaction with PA vesicles (Fig. 3, compare lane 3 (total PLD 42/20 ) with lane 7 (PLD 42/20 in the filtrate)) and PA-coated SDVB beads (no binding could be measured). These differences suggest that part of the PA activation phenomenon involves selective binding to PA in the presence of Ca 2ϩ and that an intact C-terminal part of the protein is involved.
PLD Binding to PC Interfaces-In contrast to the tight binding to anionic phospholipid surfaces, either monolayers on beads or vesicles, PLD exhibited poor binding to PC surfaces in the presence of EGTA or EDTA. The binding was sufficiently weak that estimates of the PLD K D for PC-coated SDVB beads were Ͼ1 mM for intact PLD 57 and associated PLD fragments (PLD 42/20 ). The interaction of the enzyme with PC surfaces was examined in detail using PC unilamellar vesicles and monitoring the amount of PLD bound to the PC bilayer (Table IV). In the absence of divalent metal ions, isolated PLD fragments     (Table III) or over the pH range from 6 to 8 (Fig. 4A). A direct comparison of PLD binding to PC and PS vesicles in the absence or presence of Ba 2ϩ emphasizes this behavior (Table  IV). High Ba 2ϩ basically prevented PLD 57 from binding to the PS surface, whereas that amount of metal ion enhanced PLD 57 binding to PC surfaces about 4-fold. Only PLD 57 bound to PS exhibited this sensitivity to Ba 2ϩ ; clipped PLD or the isolated fragments still bound tightly to the PS surface whether Ba 2ϩ was present or not. Thus, PLD 57 has different interactions with anionic lipids and divalent cations than the fragments or even the associated but clipped PLD 42/20 . If a small amount (10 mol %) of anionic phospholipid, either PA or PS, was included in the PC vesicles, partitioning of all forms of PLD to PC surfaces was enhanced (Table IV). Increasing the PA content enhanced this partitioning still further and was most effective for PLD 57 . The proteolytically clipped version of the enzyme did not bind to the mixed PC/PA phospholipid interfaces as well as intact protein. However, PS was more effective than PA with proteolytically clipped or purified fragments (PLD 42 and PLD 20 ). This is consistent with the PA activation of PLD involving enhanced binding of the protein to substrate interfaces. However, this surface binding cannot be the only effect of the PA.
The inclusion of either PA or PI at 10 mol % in PC vesicles also altered the pH behavior of PLD 57 binding to PC surfaces. In the absence of metal ion, the enzyme partitioning reflected binding of PLD 57 to the anionic phospholipid (e.g. binding decreased with decreasing pH) and not the PC (partitioning was invariant between pH 6 and 8). In the presence of 5 mM Ba 2ϩ (Fig. 4), PLD 57 binding to the PC/PI surface (Fig. 4C) had the same profile as for PC without PI (Fig. 4A); PLD 57 binding in the presence of Ba 2ϩ was strongest to the PC/PA vesicle, particularly at pH 5 (Fig. 4B).
Monomer PA Effect on PLD Binding and Hydrolysis of PC Surfaces-The results discussed above indicate that intact PLD 57 has a higher affinity for PC surfaces containing PA. Is this change in surface binding the main reason for the interfacial activation by PA? DiC 4 PA has a critical micelle concentration of Ͼ100 mM and should not partition into PC bilayers. If it can bind as a monomer to PLD, it may affect partitioning of the enzyme to vesicle surfaces. Therefore, partitioning of intact PLD 57 to POPC vesicles was examined in the absence of Ba 2ϩ (which by itself can enhance binding to PC surfaces). As shown in Fig. 5A, under the conditions used, the bulk of the enzyme was free and not bound to PC in the absence of the short chain PA. However, as diC 4 PA was titrated into the system, more of the enzyme was partitioned onto the vesicle surface (notice the increasing loss in intensity for free PLD 57 in Fig. 5 for lanes 4 -6, which reflect incubation with 1, 2.5, and 5 mM diC 4 PA, respectively). 5 mM diC 4 PA was not quite as efficient at driving the enzyme to the POPC surface as a vesicle composed of 0.9 mM POPC and 0.1 mM POPA (lane 7). Thus, a much higher concentration of water-soluble diC 4 PA than long chain PA (which is already localized in the bilayer) is needed to translocate PLD 57 to the bilayer surface. The observation that watersoluble diC 4 PA drives the PLD to the membrane surface is consistent with the PA acting allosterically. In the presence of Ba 2ϩ , the diC 4 PA was less effective at partitioning PLD to PC surfaces (similar to what was observed for long chain PA species).
The same PLD 57 binding experiments were carried out in the presence of water-soluble diC 4 PMe (Fig. 5A, lanes 9 -12). In contrast to diC 4 PA, high concentrations of water-soluble diC 4 PMe did not induce any detectable partitioning of PLD 57 to the PC bilayers. Thus, the enhanced surface binding effect was specific to PA. There must be a discrete binding site on the enzyme that interacts with PA head groups and promotes a change in the PLD that enhances productive binding to zwitterionic interfaces.
If this binding is a key parameter in PA-activation of PLD 57 , then the presence of diC 4 PA should activate the enzyme toward POPC vesicles. This was monitored by 1 H NMR spectroscopy (11,15). The initial rate (first 15 min) of hydrolysis of 10 mM POPC vesicles in 50 mM imidazole with 5 mM Ca 2ϩ , pH 7.2, was increased ϳ4-fold when 10 mM diC 4 PA was present (Fig. 5B), although the soluble diC 4 PA was not as effective an activator as long chain PA incorporated in the vesicle (11). Furthermore, the hydrolysis of POPC without the short chain PA was nonlinear. The rate was quite low for the first 5-10 min and then exhibited a nonlinear increase to where the rates for the two samples were within 30%. At the point where PLD showed an increased hydrolysis rate toward POPC vesicles, 5% of the long chain PA activator was generated in situ. The biphasic behavior for pure POPC vesicles suggests that as long chain PA is generated, the enzyme becomes activated. Because activation by interfacial PA is more effective than soluble PA, the PLD hydrolysis rates become more similar. Clearly, maximum activation by PA requires interfacial PA. What does the surface PA do to PLD 57 that soluble diC 4 PA cannot?
PLD Aggregation State Bound to Vesicles-One possible explanation for PA activation of PLD 57 toward PC vesicles is that the enzyme becomes oligomerized on the vesicle surface through its interactions with PA and Ca 2ϩ . EDC, a heterobifunctional cross-linker that forms amide bonds from closely juxtaposed aspartate/glutamate and lysine side chain amino groups, was used to explore the aggregation state of PLD when bound to different phospholipid vesicle surfaces (Fig. 6A). In the presence of Ba 2ϩ (to supply a divalent cation but inhibit PLD hydrolysis of the PC) and absence of any phospholipids, PLD 57 (typically 30 g/ml) formed only a small proportion of oligomers (Ͻ5%) when treated with EDC (Fig. 6A, lane 1). When PLD 57 was incubated with PC/PA (9:1) vesicles and 1 mM EGTA (lane 2), there was very little cross-linking of the protein.
However, when 5 mM Ba 2ϩ was present along with the PC/PA vesicles and EGTA, there was extensive cross-linking of the protein subunits to dimers and tetramers (lane 3). PLD 57 incubated with PC/PI (9:1) vesicles, and Ba 2ϩ also showed crosslinking by EDC (lane 4), but it was not as extensive (dimers were the preferred species formed) as with PA. Oligomerization of the PLD 57 by EDC absolutely required an interface because incubation of PLD 57 with the monomeric substrates diC 4 PC and diC 4 PA in the absence or presence of Ba 2ϩ yielded no cross-linked protein oligomers (data not shown).
The extent of cross-linking was dependent on the identity of the divalent cation as well as the concentration of PLD 57 . The gel in Fig. 6B shows how Ba 2ϩ (5 mM) or Ca 2ϩ (2 mM) affect cross-linking of the enzyme to pure PA or PI vesicles (in the presence of 1 mM EGTA). Binding of the enzyme to both anionic phospholipid vesicles in the presence of metal ions leads to oligomers trapped by EDC. PA was more effective at inducing PLD 57 trimers and tetramers than PI (Fig. 6B, compare lanes  1 and 2 with lanes 3 and 4). Furthermore, a larger proportion of higher order aggregates (tetramers and trimers) was formed with Ca 2ϩ than with Ba 2ϩ for PLD 57 binding to both PA and PI vesicles (Fig. 6B, lane 1 versus lane 2). The PLD 57 aggregates formed on Ca 2ϩ ⅐PA vesicles were still cross-linked by EDC (10 mM) when the enzyme was diluted 5-, 10-, or 20-fold in the presence of 10 mM PA vesicles (lanes 5-7). In fact, at the lowest dilution of the enzyme, the tetramer is the darkest band of PLD 57 (lane 7). This may suggest that activation by PA involves not only binding of the PLD 57 to the vesicle surface but aggregation of the protein as well.
Interestingly, none of the other PLD forms (PLD 42/20 or PLD 42 ) formed multimers that could be cross-linked by EDC when both Ca 2ϩ or Ba 2ϩ and POPA were present. These observations suggest that proteolytic cleavage alters the orientation of acidic and basic groups that are covalently linked by the EDC. They also are consistent with an oligomer as the kinetically activated form of PLD 57 .

DISCUSSION
Interfacial substrate hydrolysis by lipolytic enzymes can have two distinct binding modes for the enzyme: an initial interfacial interaction (that can be rather nonspecific and rely on electrostatics or hydrophobic interactions) that anchors the enzyme to the surface and a specific binding of a single substrate molecule to the catalytic site. Either step can be ratelimiting and affect the observed rate of vesicle hydrolysis. Phospholipase D from S. chromofuscus is an unusual phospholipase in that it displays neither interfacial activation nor surface dilution kinetics with micellar substrates. However, with monomeric diC 4 PC as the substrate, interfaces can enhance the specific activity of the intact protein (but not the proteolytically clipped enzyme whose activity is about four times higher) toward this water-soluble substrate (11). POPC vesicles are also poor substrates for this enzyme; enzyme activity increases if PA is incorporated into the PC bilayers (11). These kinetic observations suggest an allosteric role for PA (or similar interfacial molecules). At least in the case of the PC vesicles, such an allosteric role could involve increasing the amount of enzyme that binds to the vesicle surface, the first binding step in processing substrate at an interface.
PLD 57 (or any of the PLD fragments) binding to pure PC vesicles is extremely weak but enhanced by Ba 2ϩ ; enzyme partitioning on PC vesicles with Ba 2ϩ present is not pH-dependent over the pH range of 6 -8 and is not affected by salt. In contrast, PLD 57 binding to anionic phospholipid vesicles in the absence of divalent metal ions is pH-dependent with higher affinity as the pH is increased (K D Ͻ2 M at pH 8). PLD 57 binding to PI vesicles in the absence of divalent cation exhibits the same pH dependence. The binding of PLD 57 to PA surfaces can also be inhibited by high NaCl, implying that the PA⅐PLD interaction has a large electrostatic component. Although PLD 57 should have a net negative charge (pI ϭ 5.1 (8)), there must be distinct cationic sites on the enzyme that mediate this binding. When 10 mol % of anionic lipids are incorporated into PC vesicles, PLD binding (in the presence of EGTA) to the predominantly PC vesicle is dramatically enhanced, presumably through binding to the anionic lipid. However, the addition of divalent metal ions decreases binding of the enzyme to PA (alone or mixed with PC in vesicles) at pH 8. Millimolar divalent metal ions enhance PLD binding to PC but not PA interfaces.
PLD 57 absolutely requires Ca 2ϩ for activity; the K D for Ca 2ϩ is 75 M (15) with monomeric diC 4 PC as the substrate, but higher Ca 2ϩ (Ͼ1 mM) is required for activity of the enzyme toward PC in vesicles (11). The excess Ca 2ϩ will interact with the PA head groups (causing clustering of the PA in the membrane) as well as acidic groups on the protein. At these higher levels of Ca 2ϩ (e.g. 10 mM), a large fraction of PLD 57 should be bound to PC vesicles (assuming the binding will be similar to that measured with Ba 2ϩ ). Yet the enzyme is not maximally active under these conditions and requires PA in the bilayer for a higher specific activity. This clearly indicates that partitioning of the enzyme to a bilayer surface can be separate from optimal active site binding and catalysis. Furthermore, under low Ca 2ϩ conditions, PLD 57 would bind very tightly to 10 mol % PA (or other anionic phospholipid) that is incorporated into PC membranes. However, the rate of PLD 57 catalyzed hydrolysis of POPC with 10 mol % PA in the bilayer, and 0.5 mM Ca 2ϩ was inhibited when compared with pure POPC vesicles (11). Again, the binding studies indicate all the enzyme would be on the PC/PA bilayer surface. Clearly, more than just binding to the interface is involved in the kinetic activation of PLD by PA.
A possible specific role of PA⅐Ca 2ϩ in activating PLD 57 is to cause a conformational change in the enzyme that alters its accessibility to substrate. Whatever the change, it occurs with intact PLD 57 and not with PLD 42/20 (which is already optimally active toward PC vesicles or diC 4 PC) or PLD 42 . This suggests that the PA⅐Ca 2ϩ interaction occurs with the C-terminal domain of the protein. Lipases have "lids" that are opened in the presence of appropriate substrates or cofactors (22,23), and perhaps a similar lid is formed by the C-terminal domain of PLD 57 . Substrate accessibility is enhanced, and, more intriguingly, a surface-induced aggregation of PLD 57 is observed. Aggregation of PLD 57 is particularly interesting because the crystal of a small nuclease that is a member of the PLD superfamily shows a dimer and shared active site (24). However, such aggregation was not detected with monomeric substrates (EDC did not cross-link PLD 57 in the presence of diC 4 PC/Ba 2ϩ or diC 4 PA/Ca 2ϩ ) or with PLD 42/20 where enzyme activity is high. Nonactivating anionic lipids (e.g. PI) also induce PLD 57 aggregation in the presence of Ca 2ϩ . The one striking difference in enzyme aggregation induced by PA⅐Ca 2ϩ and PI⅐Ca 2ϩ is that a larger fraction of tetramers are generated with PA. Perhaps the surface activated form of PLD 57 is a tetramer.
The results generated thus far with S. chromofuscus PLD suggest a model for PA activation of the enzyme. A key component of the model is that there are two sites on the enzyme: an active site and a surface binding site. Regions near the active site of PLD, if similar to the nuclease, are likely to have significant cationic character; hence anionic lipids (such as PA) would bind with a high affinity (Fig. 7A). Low concentrations of FIG. 7. Model for PA activation of PLD 57 toward PC aggregates. A, in the absence of Ca 2ϩ , PLD 57 binds tightly to PA containing vesicles through electrostatic interactions. When excess Ca 2ϩ is added, the POPA clusters and the PA⅐Ca 2ϩ complex binds to PLD 57 and alters its conformation at the interface (inducing aggregation of the enzyme subunits) such that more facile hydrolysis of PC occurs. B, monomeric diC 4 PA can enhance binding of PLD 57 to a PC surface and enhance hydrolysis as well (although it is not as effective as long chain PA in the bilayer). C, isolated PLD 20 and PLD 42 bind more weakly to anionic phospholipids such as PA. Proteolytically cleaved but associated PLD 42/20 still binds fairly tightly to PA. The active sites of PLD 42 and PLD 42/20 are accessible to PC substrate when anchored to the membrane with or without PA. Ca 2ϩ binding to the catalytic site would have small effects on ligand binding. As Ca 2ϩ is increased, secondary Ca 2ϩ sites on the protein are saturated, and PA becomes clustered in the membrane. The PA⅐Ca 2ϩ binds to an allosteric site on the enzyme that serves to expose the catalytic site to the bilayer surface for easy diffusion of PC substrate. Protein aggregation also occurs under these conditions and may be a component of the kinetic activation of PLD 57 . That monomeric diC 4 PA enhanced both binding of PLD 57 to PC surfaces and hydrolysis, although not as effectively as long chain PA in the bilayer, confirms the PA⅐Ca 2ϩ interaction as allosteric (Fig. 7B). The lower activation by diC 4 PA could correlate with the lack of protein aggregates cross-linked by EDC. The weaker binding of PLD 42 to anionic vesicles and its higher specific activity suggest that, upon proteolytic cleavage, there is a change in the orientation of the C-terminal domain with respect to the catalytic site such that substrate accessibility is not a problem for PLD 42/20 (Fig. 7C). That PLD 20 and PLD 42 alone bind to PA surfaces with weaker affinity (Fig. 7C) would be expected if the protein is anchored via two distinct domains (C-terminal portion and catalytic domain). Clearly, a detailed structure of PLD will be needed to understand the relationship of C-terminal portion of the protein to the active site. Also critical to sorting out the spatial relationship of the PA allosteric site to the active site is the number and location of Ca 2ϩ sites. Because roughly 10% of all the amino acids in PLD are glutamate, it is likely that multiple Ca 2ϩ surface sites exist.