Mechanism of Membrane Binding of the Phospholipase D1 PX Domain*

Mammalian phospholipases D (PLD), which catalyze the hydrolysis of phosphatidylcholine to phosphatidic acid (PA), have been implicated in various cell signaling and vesicle trafficking processes. Mammalian PLD1 contains two different membrane-targeting domains, pleckstrin homology and Phox homology (PX) domains, but the precise roles of these domains in the membrane binding and activation of PLD1 are still unclear. To elucidate the role of the PX domain in PLD1 activation, we constructed a structural model of the PX domain by homology modeling and measured the membrane binding of this domain and selected mutants by surface plasmon resonance analysis. The PLD1 PX domain was found to have high phosphoinositide specificity, i.e. phosphatidylinositol 3,4,5-trisphosphate (PtdIns-(3,4,5)P 3 ) >> phosphatidylinositol 3-phosphate > phos- phatidylinositol 5-phosphate >> other phosphoinositides. The PtdIns(3,4,5)P 3 binding was facilitated by the cationic residues (Lys 119 , Lys 121 , and Arg 179 ) in the putative binding pocket. Consistent

Mammalian phospholipase D (PLD) 1 catalyzes the hydrolysis of phosphatidylcholine to generate phosphatidic acid (PA) and choline (1,2). PA may act as a lipid mediator for various proteins involved in cell signaling and vesicle trafficking (3,4) and may also regulate the physical property of the cellular membranes (5,6). Two isoforms of mammalian PLDs, PLD1 and PLD2, have been implicated in numerous cellular processes, including vesicle trafficking, cytoskeletal rearrangement, and proliferation (1,3,4,7,8). PLDs are activated in many cell types in response to growth factors, hormones, and neurotransmitters (9). It has been reported that PLD activities are regulated through interactions with a wide variety of molecules, including small GTP-binding proteins, such as ADPribosylation factor (Arf), Rho, Rac, and Cdc42, and protein kinase C isoforms (10 -16).
In most mammalian cells, PLD activities have been found associated with the membrane fraction but PLDs show complex membrane localization patterns depending on cell types. While PLD2 is mainly found at the plasma membrane (17), PLD1 shows dynamic localization between the plasma membrane and the intracellular membranes of endocytic and secretory origin, including Golgi apparatus, endoplasmic reticulum, early and late endosomes, and multivesicular bodies (18 -22). The mechanisms underlying the subcellular localization and activation of PLDs are not fully understood. In particular, the role of lipid-protein interactions in the processes is not well defined. It has been shown that phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ) and phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P 3 ) can activate PLDs in vitro (14,23,24). All mammalian PLDs contain two membrane targeting domains, a Phox homology (PX) and a pleckstrin homology (PH) domain, in the amino-terminal region and a putative PtdIns(4,5)P 2 -binding polybasic motif in the catalytic domain. Although PH and PX domains from other proteins have been shown to bind various phosphoinositides (PIs) (25)(26)(27)(28)(29)(30), direct binding of any PI to neither isolated PH domain nor isolated PX domain of PLD has been unambiguously demonstrated. As such, the exact roles of PH and PX domains in the membrane binding and the regulation of mammalian PLDs remain controversial.
Sequence alignment of PX domains (see Fig. 1) shows that PLD1 contains a Lys (Lys 119 ) instead of Arg in the putative 3-phosphate-binding site. Also, PLD1 harbors an Arg (Arg 179 ) in the putative 4-phosphate interaction site. However, still unresolved is the phospholipid binding capability of the PLD1 PX domain. This study was undertaken to determine the membrane binding properties of the PLD1 PX domain, with an emphasis on elucidating the PI and/or anionic lipid binding specificity. Results from in vitro membrane binding measurements by surface plasmon resonance (SPR) analysis as well as structural modeling indicate that PLD1 PX specifically binds PtdIns(3,4,5)P 3 and also have a secondary binding site for anionic phospholipids, including PA and phosphatidylserine (PS). This information provides new insight into the membrane targeting and activation mechanism of PLD1.
Mutagenesis and Protein Expression-cDNA for rat PLD1 was kindly provided by Dr. Sungho Ryu of Pohang University. Constructs of PLD1-PX containing residues 76 -214 were obtained by the overlap extension polymerase chain reaction method (43). Each construct was subcloned into the pET28a vector containing a COOH-terminal hexahistidine tag (between restriction sites NdeI and HindIII) and transformed into DH5␣ cells for plasmid isolation. After verifying the DNA sequence of each construct, the plasmid was transformed into BL21(DE3) cells for protein expression. One liter of LB medium containing 50 g/ml kanamycin was inoculated with BL21(DE3) cells harboring each construct and grown at 37°C until absorbance at 600 nm reached 1.0. At this time, 1 mM of isopropyl 1-thio-␤-D-galactopyranoside was added, and cells were then incubated at 25°C for overnight. Cells were harvested for 10 min at 4000 ϫ g, and the resulting pellet was resuspended in 10 ml of 50 mM Tris-HCl, pH 8.0, containing 0.1 M NaCl, 1 mM phenylmethylsulfonyl fluoride, 0.4% Triton X-100, and 0.4% sodium deoycholate. The solution was then sonicated for 8 min using a 30-s sonication followed by 30-s cooling on ice. This was followed by centrifugation at 48,000 ϫ g to separate the soluble and insoluble fractions. The supernatant was filtered into a 50-ml tube, and 1 ml of nickel-nitrilotriacetic acid solution (Qiagen, Valencia, CA) was added. The mixture was incubated on ice with gentle shaking for 1 h. After this time, the mixture was poured onto a column, which was washed with 20 ml of 50 mM Tris-HCl, pH 8.0, containing 0.1 M KCl and 10 mM imidazole. Subsequently, the protein was eluted from the column in six fractions using 0.5 ml of 50 mM Tris-HCl, pH 8.0, containing 0.1 M KCl and 300 mM imidazole. Purity was checked on an 18% polyacrylamide gel and samples were pooled and concentrated to 1 ml. Protein concentration was then determined using the bicinchoninic acid method (Pierce).
Circular Dichroism Measurements-CD spectra of PX domains were taken in a 1-mm quartz cuvette in a Jasco J810 Spectropolarimeter. Protein concentration for all measurements was 0.3 mg/ml. Following dilution of sample to final concentration in a 10 mM sodium phosphate buffer solution, pH 7.0, four scans were taken from 190 to 280 nm and averaged.
SPR Analysis-All SPR measurements were performed at 23°C. A detailed protocol for coating the L1 sensor chip has been described elsewhere (44,45). Briefly, after washing the sensor chip surface, 90 l of vesicles containing various phospholipids (see Table I) were injected at 5 l/min to give a response of 5500 resonance units (RU). Similarly, a control surface was coated with vesicles, typically without the PI of interest, to give the same resonance unit response as the active binding surface. Under our experimental conditions, no binding was detected to this control surface beyond the refractive index change for the PX domains. Each lipid layer was stabilized by injecting 10 l of 50 mM NaOH three times at 100 l/min. Typically, no decrease in lipid signal was seen after the first injection. Kinetic SPR measurements were done at the flow rate of 30 l/min. 90 l of protein in 10 mM HEPES, pH 7.4, containing 0.16 M KCl was injected to give an association time of 90 s, whereas the dissociation was monitored for 500 s or more. The lipid surface was regenerated using 10 l of 50 mM NaOH. After sensorgrams were obtained for five different concentrations of each protein within a 10-fold range of K d , each of the sensorgrams was corrected for refractive index change by subtracting the control surface response from it. The association and dissociation phases of all sensorgrams were globally fit to a 1:1 Langmuir binding model: protein ϩ (protein-binding site on vesicle) 7 (complex) using BIAevalutation 3.0 software (Biacore) as described previously (45)(46)(47). The dissociation constant (K d ) was then calculated from the equation, K d ϭ k d /k a . Equilibrium (steady state) SPR measurements were performed with the flow rate of 5 l/min to allow sufficient time for the R values of the association phase to reach saturating response values (R eq ). R eq values were then plotted versus protein concentrations (C), and the K d value was determined by a nonlinear least squares analysis of the binding isotherm using an equation, R eq ϭ R max /(1 ϩ K d /C). Mass transport (48,49) was not a limiting factor in our experiments, since change in flow rate (from 2 to 60 l/min) did not affect kinetics of association and dissociation. After curve fitting, residual plots and 2 values were checked to verify the validity of the binding model. Each data set was repeated three times to calculate a standard deviation value.
Monolayer Measurements-Surface pressure () of solution in a circular Teflon trough (4 cm diameter x 1 cm deep) was measured using a Wilhelmy plate attached to a computer-controlled Cahn electrobalance (model C-32) as described previously (50). Five to ten l of phospholipid solution in ethanol/hexane (1:9 (v/v)) was spread onto 10 ml of subphase (20 mM Tris-HCl, pH 7.4, containing 0.16 M KCl) to form a monolayer with a given initial surface pressure ( 0 ). The subphase was continuously stirred at 60 rpm with a magnetic stir bar. Once the surface pressure reading of monolayer had been stabilized (after ϳ10 min), the protein solution (typically 100 l) was injected into the subphase through a small hole drilled at an angle through the wall of the trough and the change in surface pressure (⌬) was measured as a function of time. Typically, the ⌬ value reached a maximum after 30 min. It has been shown empirically that ⌬ caused by a protein is mainly due to the penetration of the protein into the lipid monolayer (50). The maximal ⌬ value at a given 0 depended on the protein concentration and reached a saturation value. Protein concentrations in the subphase were therefore maintained above such values to ensure that the observed ⌬ represented a maximal value. The critical surface pressure ( c ) was determined by extrapolating the ⌬ versus 0 plot to the x axis (51).
Homology Modeling and Electrostatic Potential Computation-Models for the PLD1 PX domain were constructed as described previously for the PH domain (52). Fold recognition programs (53,54) identified the structures of the p40 phox PX domain (Protein Data Bank code 1H6H) (26) and the p47 phox PX domain (Protein Data Bank code 1O7K) (29) as high confidence structural representations of the PLD1 and PLD2 PX domain sequences and were, therefore, used as structural templates in modeling. The sequence alignments between the PLD PX domains and its structural templates were manually edited and constructed by combining (a) the results of alignment algorithms (55, 56), (b) threading analysis (53,54), and (c) alignment of predicted (57-61) and known (62) secondary structure elements of the target and template, respectively. Homology models were constructed by overlaying the target sequence on the template structure according to the optimized sequence alignment using the program Nest (63). Alignments were further manually edited to produce models that maximized the fitness scores obtained in the structure evaluation program, Verify 3D (64). The electrostatic properties of the PX domains with and without bound lipid (PtdIns(3,4,5)P 3 and PS) were calculated with a modified version of the program DelPhi and visualized in the program GRASP (65). PtdIns(3,4,5)P 3 and PS were docked onto the modeled PX domains using the spatial coordinates of the bound sulfate ions, taken from the structure of p47 phox -PX (Protein Data Bank code 1O7K) (29), as guides in placing the lipid phosphate groups. Steric clashes were fixed manually.

Structural Modeling of the PLD1 PX Domain-Structural
analyses of the PX domains of p40 phox , p47 phox , cytokine-independent survival kinase, and Vam7p have shown that these PX domains have similar tertiary structures (26,29,66,67). Thus, we constructed a model structure of the PLD1 PX domain by means of the homology modeling to these PX domains, to gain better insight into its potential membrane binding properties. Fold recognition programs (53,54) identified the structures of the p40 phox PX domain (Protein Data Bank code 1H6H) (26) and the p47 phox PX domain (Protein Data Bank code 1O7K) (29) as high confidence structural representations of the PLD1 PX domain sequence and were, therefore, used as structural templates in modeling. Homology models were constructed by overlaying the target sequence on the template structure according to the optimized sequence alignment using the program Nest (63). Alignments were further manually edited to produce models that maximized the fitness scores obtained in the structure evaluation program, Verify 3D (64).
The resulting model (see Fig. 2) predicts the presence of two basic pockets as potential lipid-binding sites: a primary pocket similar to the PI-binding sites of other PX domains and a shallower secondary site. A predicted PI-binding pocket contains three cationic residues, Lys 119 , Lys 121 , and Arg 179 . As shown in Fig. 1, Lys 119 of the PLD1 PX domain aligns to Arg 58 of the p40 phox PX domain and Arg 43 of the p47 phox PX domain, respectively, both of which are involved in 3Ј-phosphate binding (26,29). On the other hand, Arg 179 of the PLD1 PX domain aligns to Arg 105 of the p40 phox PX domain and Arg 90 of the p47 phox PX domain, respectively. The former interacts with the 4-and 5-hydroxyl groups of PtdIns(3)P bound to the domain (26), whereas the latter may be involved in interaction with the 4Ј-phosphate (29). Thus, these cationic residues in the putative PI-binding pocket are good candidates for binding to PIs containing 3Ј-and 4Ј-phosphate (and possibly 5Ј-phosphate). However, Lys 119 in place of Arg may attenuate the specificity and/or affinity for 3Ј-phosphorylated PIs, since Lys cannot provide the bidentate interaction of Arg observed in the crystal structure of the p40 phox PX-PtdIns(3)P complex.
The location of a secondary lipid-binding pocket was predicted for the PLD1 PX domain (Fig. 2), based upon the crystal structure of the p47 phox PX domain with a secondary lipidbinding site (29). In this case, however, the exact residues involved in forming the second site could not be predicted with confidence, since our modeling procedure could not reliably predict the orientation of loops. Nevertheless, the model suggests that a large cluster of cationic residues in this region, including Lys 132 , Lys 144 , Arg 145 , Arg 149 , Arg 150 , and Arg 159 , might be involved in interaction with an (or more) anionic phospholipid.
Our model structure also positions two hydrophobic residues Ile 171 and Phe 148 near the PI-binding pocket and the secondary binding site, respectively. Hydrophobic residues surrounding the PI-binding pocket of the PX domains of p40 phox and p47 phox have been shown to participate in partial membrane penetration (47), while corresponding residues in Vam7p PX domain have been shown to undergo chemical shifts in NMR in the presence of phospholipids (35). A hydrophobic residue near the secondary lipid-binding site of p47 phox PX has also been shown to penetrate the membrane (47). Thus, Ile 171 and Phe 148 might be involved in partial membrane penetration and hydrophobic interactions with the membrane.
To validate our model structure for the PLD1 PX domain, we bacterially expressed the domain and measured its CD spectrum. In general, the bacterial expression and purification of the PLD1 PX domain and its mutants gave a lower yield (Ͻ0.5 mg from 1 liter of growth medium) than other PX domains previously characterized, due to their tendency to aggregate. As shown in Fig. 3, the CD spectrum of the PLD1 PX domain is similar to that of the p47 phox PX, which was used as a template for the homology modeling of the PLD1 PX domain. This indicates that the bacterially expressed PLD1 PX domain is functionally folded and its folded conformation is similar to that of the p47 phox PX, as predicted by our homology modeling.
Phosphoinositide Specificity of the PLD1 PX Domain-To see if the two putative lipid-binding sites actually interact with phospholipids, we measured the binding of the PLD1 PX domain to vesicles with different lipid compositions. We first measured by kinetic SPR analysis the binding of wild type PLD1 PX domain to various PI-containing lipid vesicles immobilized to the sensor surface. Representative sensorgrams are shown in Fig. 4A for binding of the wild type to POPC/POPE/ PtdIns(3,4,5)P 3 (77:20:3) vesicles. To validate the K d value determined from the kinetic SPR analysis, we also determined K d by equilibrium SPR analysis. The K d value (20 Ϯ 0.4 nM) calculated from the equilibrium binding isotherm (Fig. 4B) agreed well with K d determined from the kinetic analysis (K d ϭ 18 Ϯ 4 nM).
In the absence of PI, the PLD1 PX domain showed extremely low affinity for POPC/POPE (80:20) vesicles (K d Ͼ 30 M). The addition of PI to the vesicles enhanced the vesicle affinity of the PLD1 PX domain in a concentration-dependent manner (data not shown), but the extent varied widely among PIs. When compared at 3 mol % PI (see Table I), the PLD1 PX domain shows the highest affinity for PtdIns(3,4,5)P 3 -containing vesicles, modest affinity for PtdIns(3)P-and PtdIns(5)P-containing vesicles, and much lower affinity for vesicles with other PIs. Specifically, its affinity for PtdIns(3,4,5)P 3 is 8-fold higher than that for PtdIns(3)P and Ͼ200-fold higher than that for PtdIns(3,4)P 2 , demonstrating specificity for PtdIns(3,4,5)P 3 . It should be noted that the PLD1 PX domain binds mono-phos-phorylated species, PtdIns(3)P or PtdIns(5)P, much better than bis-phosphorylated species, PtdIns(3,4)P 2 or PtdIns(3,5)P 2 . This precludes the possibility that the PtdIns(3,4,5)P 3 specificity of the PLD1 PX domain is simply due to a nonspecific electrostatic effect caused by highly negative PtdIns(3,4,5)P 3 .
In the case of the p47 phox PX domain, the secondary lipidbinding pocket was shown to bind with moderate affinity an anionic phospholipid with a smaller headgroup, such as PA and PS, but not PIs (29). Also, simultaneous occupation of the two sites by PI and PA (or PS), respectively, synergistically en- Thus, it appears that PA (or PS) binding to the secondary site (see below) and PtdIns(3,4,5)P 3 binding to the primary PI-binding site synergistically enhance the membrane affinity of the PLD1 PX domain. Interestingly, the addition of 3 mol % PtdIns(3,4)P 2 , which has extremely low affinity for the primary PI-binding site, to POPC/POPE/PtdIns(3,4,5)P 3 (77:20:3) vesicles had a similar enhancing effect. This indicates that unlike the secondary lipid-binding site of the p47 phox PX domain, the secondary site of the PLD1 PX domain can also interact with some PIs, presumably due to its strongly cationic nature and conformational flexibility.
Specificities of the Two Lipid-binding Sites-The lipid specificities of the two phospholipid-binding pockets were further clarified by measuring the binding of a series of site-specific mutants to vesicles containing PtdIns(3,4,5)P 3 and different anionic lipids. All putative lipid-binding residues except Lys 119 were mutated to Ala; for Lys 119 , the K119Q mutation was made due to instability of the K119A mutant. We first measured the binding of mutants to POPC/POPE/PtdIns(3,4,5)P 3 (77:20:3) vesicles by SPR analysis. Results are summarized in Table II. Clearly, mutation of Arg 179 in the putative PI-binding pocket abrogated binding to PtdIns(3,4,5)P 3 -containing vesicles, demonstrating its critical role in PtdIns(3,4,5)P 3 binding. This was not due to deleterious structural changes caused by the mutation, as evidenced by similar CD spectra of wild type and R179A (see Fig. 3). Mutation of Lys 119 and Lys 121 , which are also located in the same pocket as Arg 179 , resulted in a 13-and 7-fold reduction, respectively, in vesicle affinity. This suggests that Lys 119 and Lys 121 are also involved in PtdIns(3,4,5)P 3 binding. In contrast to mutations in the PI-binding pocket, mutations of cationic residues located in the putative secondary site (K144A/R145A, R149A/R150A, and R159A) resulted in little decrease in affinity for POPC/POPE/PtdIns(3,4,5)P 3 vesicles, indicating that PtdIns(3,4,5)P 3 does not interact with the secondary anion-binding pocket with high affinity.
We then measured the binding of mutants to POPC/POPE/ POPS (POPA) (50:20:30) vesicles in the absence of PtdIns(3,4,5)P 3 to assess the roles of the two binding pockets in PA or PS binding. As shown in Table II, mutations in the PI-binding pocket (K119Q, K121A, or R179A) did not significantly affect the binding to POPA or POPS-containing vesicles, demonstrating the specificity of this site. However, the mutations in the secondary lipid-binding pocket (K144A/R145A and R149A/R150A) caused a 4 -5-fold decrease in PA or PS affinity, indicating that the secondary site is solely responsible for binding to PA or PS.
We also measured the binding of wild type and two representative mutants, R179A and R149A/R150A, for POPC/POPE/ PtdIns (3,4,5) Roles of Hydrophobic Residues in Membrane Penetration-It has been shown that hydrophobic residues near the PI-binding pockets of PX, FYVE, and ENTH domains participate in partial membrane penetration that contributes significantly to overall membrane affinity of these domains (46,47,68). To assess the role of Ile 171 and Phe 148 near the PI-binding pocket and the secondary binding site, respectively, in membrane binding of the PLD1 PX domain, we measured the binding of I171A and F148A mutants to two different types of vesicles. When POPC/ POPE/PtdIns(3,4,5)P 3 (77:20:3) vesicles were used, I171A and F148A had 4-and 1.6-fold lower affinity than the wild type (Table II), respectively. Conversely, I171A and F148A had 1.5and 2.5-fold, repectively, lower affinity than the wild type for POPC/POPE/POPS (47:20:30) vesicles. These results indicate that Ile 171 and Phe 148 contribute to the membrane binding of the PLD1 PX domain in response to the ligand binding of their neighboring lipid-binding pockets. More importantly, the observed reduction in affinity by the mutations (i.e. 1.5-4-fold decrease) is about 10 times smaller than that caused by similar mutations in the p47 phox PX domain (29), which suggests that   the PLD1 PX domain may not significantly penetrate the membrane.
To test this notion, we measured the interactions of the PX domain with lipid monolayers of different lipid compositions. We first measured the effects of PtdIns(3,4,5)P 3 and other anionic lipids on the monolayer penetration of the PLD1 PX domain (see Fig. 5). When the penetration of the PLD1 PX domain into the POPC/POPE (80:20) monolayer of varying o was monitored, the PLD1 PX domain showed low penetrating power with a c value of 26 dyne/cm, implying that PLD1 PX has low intrinsic membrane penetrating capability. Incorporation of 3 mol % PtdIns(3,4,5)P 3 into the monolayer (i.e. POPC/ POPE/PtdIns(3,4,5)P 3 (77:20:3)) modestly elevated the monolayer penetrating capability of the domain, with the c value near 30 dyne/cm. However, this monolayer penetration was significantly less than that of the p47 phox PX domain that has the c value above 35 dyne/cm with the POPC/POPE/ PtdIns(3,4)P 2 (77:20:3) monolayer. Furthermore, 3 mol % POPA, POPS, PtdIns(3)P, PtdIns(5)P, or PtdIns(4,5)P 2 exhibited essentially the same effect on the monolayer penetration as PtdIns(3,4,5)P 3 , indicating that this enhancing effect is due not to the specific PI-induced membrane penetration reported for the p47 phox PX domain (47) but to nonspecific electrostatic attraction.
Electrostatic Potential Calculations-The above studies indicate that although the PLD1 PX domain and the p47 phox PX domain are similar with respect to the tertiary structure and dual lipid binding, they also have noticeably different membrane binding properties, particularly with respect to PI-mediated membrane penetration. To understand the basis of these differences, we calculated the electrostatic potentials of the PLD1 PX domain model structure in the presence and absence of PtdIns(3,4,5)P 3 and PS, respectively. Our previous electrostatic potential calculations of p47 phox PX showed that binding of PtdIns(3,4)P 2 and PS reduces the electrostatic potentials surrounding their respective binding sites, which in turn allows favorable partitioning of hydrophobic residues in these regions into the membrane due to the reduced desolvation penalty associated with the process (29,47).
The electrostatic potential calculation of the PLD1 PX domain showed a highly cationic surface in the vicinity of the secondary binding site, due to the presence of a large number of surface cationic residues, whereas the PI-binding site has a less pronounced positive electrostatic potential (see Fig. 6A). This suggests that the secondary site may play a role of initially bringing the domain to the anionic membrane through nonspecific electrostatic interactions. The binding of PtdIns(3,4,5)P 3 (3.7 Ϯ 0.6) ϫ 10 4 (8. to the putative PI-binding site caused a dramatic change in electrostatic potential (i.e. electrostatic switch) surrounding the binding site (Fig. 6B). However, the negative potential surrounding Ile 171 remains relatively unchanged after PtdIns(3,4,5)P 3 binding. Thus, it is not expected that PtdIns(3,4,5)P 3 binding will facilitate the membrane penetration of Ile 171 by reducing the desolvation penalty associated with the membrane insertion. This is consistent with the finding that PtdIns(3,4,5)P 3 had a relatively modest effect on the monolayer penetration of the PLD1 PX domain (see Fig. 5). Fig.  6C illustrates that the docking of PS to the secondary pocket does not significantly reduce the electrostatic potential surrounding this site and Phe 148 . This again explains the modest effect of the F148A mutation on the membrane affinity for POPC/POPE/POPS (57:20:30) vesicles. Last, we calculated the change in electrostatic potential of PLD1 PX in the presence of both PtdIns(3,4,5)P 3 and PS. As shown in Fig. 6 (B and D), the docking of PS does not significantly affect the potential of the PtdIns(3,4,5)P 3 -bound PLD1 PX domain. Collectively, our electrostatic calculations on the PLD1 PX domain suggest that neither PtdIns(3,4,5)P 3 binding to the PI-binding site nor PS (or PA) binding to the secondary site has an electrostatic unmasking effect on the neighboring surface hydrophobic residues, which is consistent with its lower degree of monolayer penetration when compared with the p47 phox PX domain (see Fig. 5). The exceptionally positive electrostatic potential of this PX domain suggests that its membrane binding be driven primarily by electrostatic forces. DISCUSSION PLD1 has at least three structural parts that might play a role in its membrane targeting: PX domain, PH domain, and the polybasic motif in the catalytic domain. It has been reported that PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 stimulate PLD activities, whereas other PIs, including PtdIns(3,4)P 2 and PtdIns(4)P, are ineffective (14,23,24). Previous studies have indicated that both the PH domain (24) and the polybasic motif (69) interact with PtdIns(4,5)P 2 , albeit with different affinities, and are therefore involved in PtdIns(4,5)P 2 -mediated membrane targeting and activation of PLD1. A recent study (19) suggested that the PLD1 PX domain might bind PtdIns(5)P and mediate targeting of the protein to endocytic vesicles. The present study clearly shows that the PX domain of PLD1 has high affinity and specificity for PtdIns(3,4,5)P 3 , due to the presence of three cationic residues in the PI-binding pocket that might be able to interact with the 3Ј-, 4Ј-, and 5Ј-phosphate groups in PtdIns(3,4,5)P 3 . High affinity and specificity of the PLD1 PX domain for PtdIns(3,4,5)P 3 suggest that it can mediate the reported action of PtdIns(3,4,5)P 3 and allow PLD1 to function as a downstream effector of PtdIns(3,4,5)P 3 . This notion is supported by findings that the inhibition of PI 3-kinase results in the down-regulation of PLD activity (70 -72) and that PLD is involved in insulin-like growth factor-I-induced activation of extracellular signal-regulated kinase (ERK) in Chinese hamster ovary cells (73). The study also indicates that anionic lipids, such as PS or PA, in the membrane can synergistically enhance the PtdIns(3,4,5)P 3 -mediated membrane binding of the PLD PX domain.
Our previous studies on ENTH, FYVE, and PX domains have shown that these domains initially bind the anionic membrane containing their cognate PIs through nonspecific electrostatic interaction, which is followed by the PI binding and concomitant (or subsequent) PI-induced membrane penetration of surface hydrophobic residues (46,47,68). PIs induce the membrane penetration of their binding domains by causing local conformational changes of proteins and by neutralizing the positive electrostatic potential surrounding the hydrophobic loop residues that restricts the membrane penetration of the domains. For the p47 phox PX domain, binding of PA (or PS) to the secondary lipid-binding site synergistically enhances the PI-mediated membrane penetration (29,47). Our vesicle binding, monolayer penetration, and electrostatic calculation studies indicate that the PLD1 PX domain has a different membrane binding mechanism. It is not clear from our studies whether the secondary binding site binds a single anionic lipid molecule in the pocket or interacts nonspecifically with anionic membrane surface. In either case, however, the strongly cationic region surrounding the secondary lipid-binding site would seem to make initial contact with the anionic membrane during the membrane binding of the PLD1 PX domain, which is followed by the binding of PtdIns(3,4,5)P 3 to its pocket. Unlike the p47 phox PX domain, neither PI binding nor PS (or PA) binding has a large effect on the membrane penetration of the PLD1 PX domain. Instead, the membrane interactions of the PLD1 PX domain appear to be primarily electrostatic in nature, as is the case with many PH domains.
The affinity (K d ϭ 18 nM) of the PLD1 PX domain for POPC/ POPE/PtdIns(3,4,5)P 3 (77:20:3) vesicles is comparable with that of p47 phox PX domain (K d ϭ 38 nM) for its preferred vesicles (i.e. POPC/POPE/PtdIns(3,4)P 2 (77:20:3)) (29). Since the isolated p47 phox PX domain is not localized at cell membranes (47), the PLD1 PX domain may not be able to autonomously translocate to cell membranes unless the localization concentration of PtdIns(3,4,5)P 3 is high. Also, it was reported that the full-length PLD1 binds PtdIns(4,5)P 2 -containing membranes with K d Ϸ 2 nM via its PH and/or polybasic motif (24). In conjunction with the higher cellular concentration of PtdIns(4,5)P 2 than PtdIns(3,4,5)P 3 in the plasma membrane and other endocytic membranes, these data indicate that the PLD1 PX domain would not drive the binding of PLD1 to these membranes but rather provide secondary interactions. It should be noted, however, that the membrane affinity of the PLD1 PX domain is synergistically enhanced by other anionic lipids and that all cellular membranes contain a considerable amount of anionic lipids. Thus, in the case that the local concentration of PtdIns(3,4,5)P 3 in the plasma membrane and endocytic membranes rises in response to specific stimuli, the PX domain would make significant contribution to the binding of PLD1 to these membranes. Furthermore, the PX domain might allow the PLD molecule to perform the catalysis processively by interacting with PA locally produced by the PLD catalysis. This notion is consistent with our finding that PtdIns(3,4,5)P 3 and PA greatly slow the membrane dissociation of the PLD1 PX domain (see Table I).
It is also possible that the PX domain plays a role of activating the PLD molecule after the protein is targeted to the membrane by other parts of the molecule, such as the PH domain or the polybasic motif. That is, the interaction of the PX domain with PtdIns(3,4,5)P 3 in the membrane may induce a conformational change of PLD1 that leads to the enzyme activation. In fact, the PX domain has been implicated in PLD1 activation and PLD1-induced activation of other proteins. For instance, the PLD1 PX domain has a phosphorylation site for protein kinase C, phosphorylation of which renders the enzyme responsive to agonist, such as phorbol ester and growth factors (11-13, 15, 74 -79). Also, it was reported that the interaction of PXXP motif in the PX domain of PLD2 with the Src homology 3 domain of phospholipase C-␥1 (80), which might be allosterically regulated by the lipid binding of the PX domain, is important for the epidermal growth factor-induced phospholipase C-␥1 activation. Thus, it is plausible that the PX domain of PLD1 serves as a module that turns on and off its enzyme activity and interaction with other proteins in response to PtdIns(3,4,5)P 3 binding.
It should be noted that the PLD1 PX domain has the highest affinity for PtdIns(3,4,5)P 3 among PIs when the same concentration of PIs are incorporated in the vesicles. Since the affinity of the PX domain to PI-containing vesicles sharply depends on the PI concentration, it would bind comparably or even better to PtdIns(5)P-(or PtdIns(3)P)-containing vesicles than to PtdIns(3,4,5)P 3 -containing vesicles if the PI content of the former is considerably higher than that of the latter. By the same token, the PLD1 PX domain could respond equally well or even more favorably to PtdIns(3)P or PtdIns(5)P than to PtdIns(3,4,5)P 3 in the cell, depending on the local concentration of PI. This may account for the observation that the PLD1 PX domain mediates targeting of PLD1 to endocytic vesicles via PtdIns(5)P binding (19). In summary, the present investigation demonstrates that PLD1 PX binds PtdIns(3,4,5)P 3 with high affinity and specificity and that this binding is synergistically enhanced by binding of anionic lipids to the secondary site. The membrane interaction of the PLD1 PX domain will significantly enhance the affinity of PLD1 for cell membranes under appropriate conditions and may also play a role of activating the membranebound enzyme. This study opens up a new avenue to study the role of the PX domain in subcellular targeting and activation of mammalian PLDs.