Association of the N- and C-terminal Domains of Phospholipase D

Rat brain phospholipase D1 (rPLD1) belongs to a superfamily defined by the highly conserved catalytic motif (H(X)K(X)4D, denoted HKD. rPLD1 contains two HKD domains, located in the N- and C-terminal regions. The integrity of the two HKD domains is essential for enzymatic activity. Our previous studies showed that the N-terminal half of rPLD1 containing one HKD motif can associate with the C-terminal half containing the other HKD domain to reconstruct wild type PLD activity (Xie, Z., Ho, W.-T. and Exton, J. H. (1998)J. Biol. Chem. 273, 34679–34682). In the present study, we have shown by mutagenesis that conserved amino acids in the HKD domains are important for both the catalytic activity and the association between the two halves of rPLD1. Furthermore, we found that rPLD1 could be modified by Ser/Thr phosphorylation. The modification occurred at the N-terminal half of the enzyme, however, the association of the N-terminal domain with the C-terminal domain was required for the modification. The phosphorylation of the enzyme was not required for its catalytic activity or response to PKCα and small G proteinsin vitro, although the phosphorylated form of rPLD1 was localized exclusively in the crude membrane fraction. In addition, we found that the individually expressed N- and C-terminal fragments did not interact when mixed in vitro and were unable to reconstruct PLD activity under these conditions. It is concluded that the association of the N- and C-terminal halves of rPLD1 requires their co-expression in vivo and depends on conserved residues in the HKD domains. The association is also required for Ser/Thr phosphorylation of the enzyme.

Phospholipase D (PLD) 1 is a ubiquitous enzyme found in bacteria, fungi, plants, and mammals (1). It hydrolyzes phosphatidylcholine to phosphatic acid and choline. Phosphatic acid is generally recognized as the signaling product of PLD and functions as an effector in different physiological processes. Phosphatic acid can be converted to diacylglycerol by phosphatidate phosphohydrolase or to lysophosphatidic acid by a specific phospholipase A 2 . Diacylglycerol is a well characterized activator for protein kinase C, while lysophosphatidic acid is a major extracellular signal that acts on specific cell surface receptors. PLD also catalyzes a phosphatidyl transfer reaction using primary alcohols as nucleophilic acceptors to produce phosphatidylalcohols. This reaction is used as a specific measure of PLD activity.
The physiological function of PLD is still under investigation. Several studies have shown that PLD plays a role in a variety of signaling pathways (2). Different factors, including protein tyrosine kinases, protein kinase C (PKC), heterotrimeric and small G proteins, and intracellular Ca 2ϩ regulate PLD activity directly or indirectly (2)(3)(4). Based on the wide involvement of PLD in signaling pathways and the actions of its products, multiple functions of PLD have been proposed (2), which include signal transduction, membrane trafficking, and cytoskeleton changes.
To date, two isoforms of mammalian PLD (PLD1 and PLD2) have been cloned. These isoforms share about 50% amino acid similarity, but exhibit quite different regulatory properties. PLD1 has a low basal activity and responds to PKC and to members of the Rho and ARF families of small G proteins (5)(6)(7)(8). The PKC interaction domain has been mapped to the N-terminal part of the molecule (9 -11), while the Rho interaction domain has been localized to the C-terminal part of the enzyme (12). PLD2, on the other hand, exhibits a high basal activity and generally shows little response to stimuli (13)(14)(15). However, a recent study has shown that PLD2 can be tyrosine phosphorylated and its activity up-regulated in HEK293 cells treated with epidermal growth factor (16). The precise cellular localization of the two isoforms is not well defined. However, recent reports have shown that both isoforms are present in a caveolin-rich membrane fraction (17,18).
Sequence analysis has revealed that PLD belongs to a superfamily defined by an invariant motif, HXK(X) 4 D, denoted "HKD" (19 -21). The enzymes within the family exhibit diverse functions and include phospholipid synthases, poxvirus envelope proteins, a Yersinia murine toxin (Ymt), and the Nuc endonuclease. Despite the distinct substrate specificities of the superfamily members, the consensus HKD motif appears to be essential for their enzymatic activity. PLD contains two copies of the HKD motif. Mutation of either HKD motif inactivates human PLD1 and mouse PLD2 (11,22). Biochemical and structural studies of the Nuc endonuclease and Ymt suggest that the histidine residue in the conserved motif is directly involved in the catalytic reaction by forming a phosphoenzyme intermediate (23)(24)(25).
We have cloned a PLD1 isoform from rat brain (rPLD1) and have studied its properties in vivo and in vitro (7,9,26,27). Our previous study of this isoform showed that N-terminal or C-terminal fragments containing only one of the HKD domains did not have PLD activity when expressed in COS7 cells (10). However, PLD activity was restored when the two fragments were coexpressed (10). Furthermore, the N-and C-terminal fragments were shown to physically associate (10). Thus, the two HKD domains are likely brought together to form an active catalytic center by interdomain association. In this study, we use site-directed mutagenesis to examine the importance of conserved amino acids in the HKD domains for the association of the N-and C-terminal halves of rPLD1. In the process of this investigation, we found that rPLD1 can be modified by Ser/Thr phosphorylation and this modification requires the interdomain association of the N-and C-terminal halves of the enzyme.
Plasmid Construction-The N-terminal Xpress-tagged full-length rPLD1 or the N-or C-terminal fragments of rPLD1 with coding regions corresponding to amino acids 1-584 and 585-1036, respectively, were created by PCR amplification and subcloned at the KpnI/XbaI sites in the polylinker region of PcDNA3.1 vector (Invitrogen) as described (10). The C-terminal V5-tagged full-length rPLD1 containing amino acids 1 to 584 was also generated by PCR amplification followed by subcloning into the HindIII/XbaI site in the polylinker region of PcDNA3.1/V5 His A vector (Invitrogen). The forward PCR primer was 5Ј-AGGGTAAGCT-TACCATGTCACTAAGAAGTGAGGC. The reverse primer was 5Ј-TGCTCTAGAGGAGGCGCTGTCGACGCT. The PCR fragments were subcloned in-frame with the C-terminal V5-tag. The site-directed mutations of rPLD1 were generated as described in the QuickChange Site-Directed Mutagenesis Instruction manual from Stratagene. All the constructs were sequenced to verify the coding regions of rPLD1.
In Vivo PLD Assay-After 6 h transfection with FuGENE6, COS-7 cells in six-well plates were serum-starved (0.5% fetal bovine serum in Dulbecco's mdoified Eagle's medium) in the presence of 1 Ci/ml [ 3 H]myristic acid. After overnight starvation, the cells were washed with phosphate-buffered saline and incubated in serum-free medium supplemented with 0.3% bovine serum albumin for 50 min. PLD activity was then assayed as described (28). Briefly, cells were incubated in 0.3% 1-butanol for 25 min. Cells were then washed with ice-cold phosphate-buffered saline and stopped with methanol. Lipids were extracted, and the phosphatidylbutanol product was resolved by thin layer chromatography. Bands co-migrating with a phosphatidylbutanol standard were quantitated by scintillation counting.
Subcellular Fractionation-Two 10-cm dishes of COS-7 cells were harvested after transfection and overnight starvation as described above. The cells were washed twice with ice-cold lysis buffer (25 mM Hepes, pH 7.2, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 1 mM of dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture). The cells were then resuspended in lysis buffer and passed through a 27-gauge needle five times. The cell lysate was first spun at 500 ϫ g for 10 min to remove unbroken cells. The supernatant was then spun at 120,000 ϫ g for 45 min at 4°C to separate the cytosolic and crude membrane fractions. The particulate fraction was washed four times with the lysis buffer and then passed through a 27-gauge needle until the pellet was resuspended. The Bradford method was used for the quantitation of protein in the cytosolic and the crude membrane fractions. When the interaction of the N-and C-terminal domains was examined in vitro, the membrane fraction expressing either N-or Cterminal or both fragments were solubilized in the above lysis buffer with the supplementation of Triton X-100 to a final concentration of 0.5% and SDS to 0.1%. The suspension was first mixed as indicated under "Results" and spun at 120,000 ϫ g for 10 min to remove unsolubilized debris and then the supernatant was subjected to immunoprecipitation as described below.
Western Blotting-Protein samples were analyzed by SDS-PAGE on 8% gels and transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore). The blots were then blocked with 5% non-fat milk and incubated with appropriate primary antibodies followed by incubation with horseradish peroxidase-conjugated secondary antibody. Immunoreactive bands were detected using enhanced chemiluminescence.
Immunoprecipitation-COS-7 cells cultured on 6-well plates (35 mm well) were transfected and starved as described above. The cells were washed twice with ice-cold phosphate-buffered saline and then resuspended in immunoprecipitation buffer (IP buffer) containing 25 mM Hepes, pH 7.2, 10% glycerol, 1 mM EDTA, 1 mM EGTA, 50 mM KC1, 0.5% Triton X-100, 10 mM NaF, 10 mM Na 4 P 2 O 7 , 1.2 mM Na 3 VO 4 , 1 M microcystin, and 2 tablets of protease mixture. The cell suspension was then passed through a 27-gauge needle five times and the resulting cell lysate was spun at 15,000 rpm in an Eppendorf microcentrifuge for 10 min at 4°C to pellet the unbroken cells. The supernatant was then precleared by mixing it with 1 g of affinity purified mouse IgG and 20 l of a 1:1 slurry of protein A beads for 1 h at 4°C. The mixture was then spun and the resulting supernatant was incubated with 3 l of Xpress mouse antibody and 20 l of protein A beads overnight. The immunoprecipitates were washed four times with the IP buffer and then resuspended in SDS sample buffer and analyzed by Western blotting.
Dephosphorylation-Full-length or N-and C-terminal halves of rPLD1 were immunoprecipitated by anti-Xpress antibodies as described above and washed twice with the IP buffer to remove proteins nonspecifically bound to the protein A beads and an additional two times with dephosphorylation buffer (25 mM Hepes, pH 7.2, 50 mM KCl, 10% glycerol, 1 mM dithiothreitol, and 2 tablets of protease inhibitor mixture) to remove the detergent in the IP buffer. The resulting products on the protein A-agarose beads were then suspended in the reaction buffer and treated with either no addition, or with the addition of Ser/Thr phosphatase mixture (29) containing mainly PP1 and PP2A or with the phosphatase mixture preinactivated with 1 M microcystin, a potent inhibitor of protein phosphatase type 1 and 2A (30). After 30 min incubation at room temperature, the reaction was either stopped by the addition of SDS sample buffer and analyzed by Western blotting, or the agarose beads were reisolated, washed with dephosphorylation buffer, and used for in vitro PLD assay.
In Vitro PLD Assay-The catalytic activity of the immunoprecipitated full-length or N-and C-terminal halves of rPLD1 was measured by the formation of [ 3 H]phosphatidylbutanol in vitro as described (7). Briefly, phospholipid vesicles generated from PE/PIP 2 /PC (16:1.4:1) containing [palmitoyl-3 H]PC (0.5 Ci/reaction) were used with butanol (1% v/v) as substrates. The reaction mixtures were incubated at 37°C for 30 min and stopped with chloroform/methanol/HCl (50:98:2). The lipids were extracted from the organic phase and resolved by thin layer chromatography. Bands co-migrating with a phosphatidylbutanol standard were quantitated by scintillation counting.

Conserved Amino Acids in the HKD Motifs Are Important for the Association of the N-and C-terminal Halves of PLD-Our
previous work showed that the N-terminal fragment of rPLD1 containing one of the HKD domains can form a functional complex with the C-terminal fragment containing the other HKD domain (10). Studies of human PLD1 have shown that mutation of amino acids in the conserved HKD motifs diminishes PLD activity (22) and recent characterizations of the Nuc endonuclease (25) and Ymt (23) have demonstrated that the conserved histidine residue in the HKD motif directly participates in the catalytic reaction by forming a phosphoenzyme intermediate. We tested the possibility that the HKD motifs and their surrounding regions are important for the interdomain association. Patches of hydrophobic amino acids containing mainly leucine, isoleucine, and valine found either within or close to the HKD domains are highly conserved among mammalian PLD1, PLD2, and yeast PLD SPO14 as outlined in Fig. 1A. We examined the roles of these amino acids in the interdomain interaction by mutating them to the less hydrophobic amino acid alanine. The mutations in the N-terminal half of rPLD1 were L467A, V468A, I469A, and I470A, and the mutations in the C-terminal half of rPLD1 were L861A, L862A, I863A, V869A, I870A, and I871A. In the following text, the wild type N-and C-terminal fragments of rPLD1 are referred to as rPLD1NT and rPLD1CT, respectively. In addition, the rPLD1NT used in the following experiments was tagged at its C terminus with a V5 epitope and rPLD1CT was tagged at its N terminus with an Xpress epitope. These two constructs have been shown to be able to associate and reconstruct wild type PLD activity when co-transfected in COS-7 cells (10).
We first examined the effect of the mutations of the hydrophobic amino acids on PLD activity in COS-7 cells (Fig. 1B). Co-transfection of wild type rPLD1NT with rPLD1CT in COS-7 cells reconstructed a PLD activity comparable to wild type. Mutation L467A in rPLD1NT decreased the reconstructed activity to 60% of that of wild type, while mutations V468A and I469A reduced it to 30% of wild type activity, and mutation I470A produced no activity above vector alone. When rPLD1CT carrying a V869A mutation was co-transfected with wild type rPLD1NT, 30% of wild type PLD activity was obtained. When mutation I871A was generated at the C-terminal half of rPLD1, residual activity was reproducibly observed when the mutated construct was co-transfected with the wild type Nterminal half in COS-7 cells. However, when mutations L861A, L862A, I863A, and I870A were examined, only background PLD activity comparable to vector control was observed. Thus, all the alanine mutations generated in either the N-or Cterminal halves decreased the PLD activity, although to different degrees. The loss of activity could not be ascribed to poor expression of the mutant proteins (Fig. 2, A and C). In addition to alanine mutations, we also mutated I470 to valine or glycine. Although the I470A mutation reduced the PLD activity to background level, wild-type activity was observed when Ile 470 was conservatively mutated to valine, a more hydrophobic amino acid than alanine (Fig. 1B, lane 6). However, mutation of Ile 470 to glycine totally abolished the PLD activity (data not shown). Similarly, mutating Ile 469 or Ile 871 to glycine diminished the residual PLD activity to vector control (data not shown). These results suggest that these patches of hydrophobic amino acids are important for PLD activity.
To investigate the roles of these amino acids in the association of the N-and C-terminal fragments of PLD, immunoprecipitation assays were performed. As mentioned above, the rPLD1NT and rPLD1CT fragments were differentially tagged with V5 and Xpress epitopes, respectively. The COS-7 cells were co-transfected with the indicated wild type or mutant Nand C-terminal fragments of rPLD1 and the cell lysates (Fig. 2, A and C) were immunoprecipitated with anti-Xpress antibodies, and analyzed by Western blotting with anti-V5 antibodies (Fig. 2, B and D). As a negative control, rPLD1NT was cotransfected with a nonspecific plasmid expressing Xpresstagged ␤-galactosidase. No protein band was detected by anti-V5 antibodies in this control experiment (lane 1 in Fig. 2, B and D). As a positive control, rPLD1CT was co-transfected with rPLD1NT and a protein corresponding to the size of rPLD1NT (ϳ66 kDa) was detected by anti-V5 antibodies (lane 2 in Fig. 2, B and D). A slower migrating band was also detected, and the significance of this is described in a later section. When mutated N-or C-terminal fragments were co-transfected with their corresponding wild type C-or N-terminal halves, we found that rPLD1NT (L467A) showed a comparable ability as wild type rPLD1NT to associate with rPLD1CT (Fig. 2B, lane  3). However, rPLD1NT (V468A), rPLD1NT (I469A), or rPLD1CT (V869A) retained only partial ability to associate with its corresponding wild type C-or N-terminal half (shown in Fig. 2, B and D), which is consistent with their limited ability to reconstruct a wild type PLD activity (Fig. 1B). As for the L861A, L862A, I863A, I870A, and I871A mutants of rPLD1CT, only very slight or no association with their N-terminal halves was observed (Fig. 2D). Consistent with its effect on PLD activity, rPLD1NT (I470A) lost its ability to associate with the A, alignment of the amino acids in the two HKD domains among rPLD1, rPLD2, and the yeast PLD1 isoform, SPO14. The amino acids that were mutated in the study are marked by asterisks. The four conserved domains are defined as described for human PLD1 (6). The domain I in rPLD1 is from amino acid 322 to 496; domain II is from 608 to 666; domain III is from 707 to 826; and domain IV is from 850 to 927. B, wild type or mutated N-and C-terminal fragments were co-transfected into COS-7 cells as indicated in the figure and the PLD activity was measured as described under "Experimental Procedures." NT, Cterminal V5-tagged N-terminal fragment of rPLD1. CT, N-terminal Xpress-tagged C-terminal fragment of rPLD1. The PLD activity is normalized to the percentage of wild type PLD activity. The results are representative of two experiments performed in triplicate. Mean values of three determinations are shown.
C-terminal fragment, whereas the I470V mutation did not affect the association between rPLD1NT and PLD1CT (Fig. 2B). In addition, mutation of Ile 469 , Ile 470 , Ile 863 , or Ile 871 to glycine all reduced the interdomain association to the background level ( Fig. 3 or data not shown). Examination of the expression of the mutated N-or C-terminal fragments in COS-7 cells showed that all the mutants were stably expressed at levels comparable with that of the corresponding wild type fragment (Fig. 2, A and C) and were localized in the crude membrane fraction (data not shown). Based on the above results, we conclude that these hydrophobic amino acids are important for the association between the N-and C-terminal halves of rPLD1. The reduced PLD activity caused by the various mutations (Fig. 1B) correlated closely with their effects on the interdomain association. Thus it seems very likely that the loss of catalytic activity is due to the decreased interdomain association caused by the mutations of these hydrophobic amino acids.
Several mutations in the HKD motifs, such as H464D, K860R, and G479V have been shown previously to inactivate the enzymatic activity of human PLD1 (22). We generated the corresponding H464D, H464E, and G479V mutations in the N-terminal half and K860R in the C-terminal half of rPLD1. Consistent with the findings with human PLD1, all these rPLD1 mutants lost the ability to reconstruct PLD activity when co-transfected with their corresponding wild type halves (data not shown). Furthermore, we found that these mutations also dramatically decreased the interaction between the N-and C-halves of rPLD1 (Fig. 2E), although these mutant constructs were stably expressed in COS-7 cells (Fig. 2F). Analysis of human PLD1 mutants in the conserved HKD domains has suggested that no significant conformational changes were generated by the mutations (22). In addition, evaluation of HKD mutations in Ymt by UV and CD spectral measurements indicated no disruption of secondary structure (23). Thus, the effects of the point mutations in the HKD motifs on the interdomain association and catalytic activity or rPLD1 were unlikely to be nonspecifically caused by dramatic conformational changes of the molecule.
In summary, the results shown above suggest that conserved amino acids in or adjacent to the HKD motifs are important for the association between the N-and C-terminal domains. Some amino acids, such as the histidines in the conserved motifs, may play dual roles in both the catalytic reaction and the The resultant products were analyzed by SDS-PAGE followed by Western blotting with anti-V5 antibodies. B, the N-terminal Xpress-tagged full-length rPLD1 was expressed and immunoprecipitated from COS-7 cells and treated with no addition or with phosphatase or preinactivated phosphatase as described in "A" and analyzed by Western blotting with anti-Xpress antibodies. rPLD1NT, N-terminal fragment of rPLD1 tagged at the C terminus with V5 epitope. rPLD1CT, C-terminal fragment tagged at the N terminus with Xpress epitope. association between the N-and C-terminal fragments.
Association of the N-and C-terminal Domains Is Required for Ser/Thr Phosphorylation of the N-terminal Half of rPLD1-Several interesting phenomena were observed in the characterization of the above mutations in the HKD motifs. First, when the N-terminal half of rPLD1 was co-transfected with its C-terminal half, the N-terminal half of rPLD1 was partly modified as demonstrated by the appearance of bands with decreased electrophoretic mobility when the cell lysates (lane 2 in Fig. 2, A and C) or the immunoprecipitates (lane 2 in Fig. 2, B  and D) were analyzed. However, no modification of the Nterminal fragment was observed when it was transfected alone (data not shown) or co-transfected with a nonspecific plasmid expressing the ␤-galactosidase (lane 1 in Fig. 2A). In addition, the C-terminal fragment migrated as a single band, i.e. showed no evidence of modification. When the mutations generated above in the HKD motifs were examined, it was found that mutations in either the N-or C-terminal domains that abolished or greatly diminished the interdomain interaction (I470A, L861A, L862A, or I870A) also eliminated the modification of the N-terminal fragment (Fig. 2, A and C). On the other hand, the mutants that did not affect or slightly decreased the interdomain association (L467A and I470V) still retained full or reduced modification compared with wild type (Fig. 2, A and  B). In the case of those mutants in which slight association was evident (V468A, I469A, and V869A), a low level of modification was detectable (Fig. 2, A to D). These results suggest that rPLD1 can be modified in the N-terminal domain, and this modification requires not only the presence of the C-terminal half but, more importantly, association with the C-terminal domain.
In order to elucidate the nature of the modification, the Nand C-terminal halves were coexpressed and immunoprecipitated from the COS-7 cells using anti-Xpress antibodies as described under "Experimental Procedures." We first examined whether the modification observed in the N-terminal half of rPLD1 was due to phosphorylation. The isolated immunoprecipitates were incubated at room temperature for 30 min with a Ser/Thr phosphatase mixture (29) or with the phosphatase mixture preinactivated by microcystin, a potent inhibitor of phosphatase type 1 and 2A (30). As a control, the immunoprecipitates were also incubated with no additions for 30 min on ice or at room temperature. The reaction products were analyzed by Western blotting with anti-V5 antibodies to detect the N-terminal fragment of rPLD1. As shown in (Fig. 3A, lanes 1  and 2), the pattern of modification was not altered when the immunoprecipitates were incubated with no additions. However, when the immunoprecipitates containing the N-and Cterminal halves were incubated with the Ser/Thr phosphatase mixture, the band with lower electrophoretic mobility disappeared and there was increased intensity of the protein with faster electrophoretic mobility, which corresponded to the size of the unmodified N-terminal fragment (Fig. 3A, lane 3). When the Ser/Thr phosphatase mixture was first mixed with microcystin and then added to the immunoprecipitates, the effect of the phosphatase was blocked (lane 4 in Fig. 3A).
To see if the modification was also seen with full-length rPLD1 and to examine the effects of Ser/Thr phosphatase, full-length rPLD1 tagged with Xpress epitope was precipitated and analyzed by Western blotting. Modification of the enzyme was also detected as demonstrated by the existence of a protein band with lower electrophoretic mobility than that of unmodified rPLD1 (lane 1 and 2 in Fig. 3B). The band with a low electrophoretic mobility was sensitive to treatment by the Ser/ Thr phosphatase mixture which converted it to a single band which migrated like the unmodified form (Fig. 3B, lane 3). This effect of the phosphatase mixture was also inhibited by micro-cystin (Fig. 3B, lane 4). The modification detected in the Nterminal half or full-length rPLD1 was also sensitive to treatment by alkaline phosphatase (data not shown), but not by treatment with the catalytic domain of the protein tyrosine phosphatase SHP-1 (31) (data not shown). These results suggest that the modification observed in rPLD1 in COS-7 cells is Ser/Thr phosphorylation and that it occurs on the N-terminal half of the enzyme. The modification likely occurs on multiple Ser/Thr residues, since it resulted in an electrophoretic mobility change of the N-terminal fragment of rPLD1 corresponding to 11 kDa (Fig. 3). The association between the N-and Cterminal halves is necessary for the modification of rPLD1. However, the modification was never complete because a significant fraction of unmodified N-terminal fragment was always associated with the C-terminal half in the immunoprecipitates (Fig. 2). Although the C-terminal domain obviously plays an important role in the Ser/Thr modification of the N-terminal half, the mechanism is not clear. One explanation is that the C-terminal fragment recruits the kinase that carries out the Ser/Thr phosphorylation. However, it is possible that a conformational change induced by the association of N-and C-terminal domains of rPLD1 exposes the phosphorylation sites to the kinase.
Ser/Thr Phosphorylation Modification of rPLD1 Is Not Required for Its Enzymatic Activity in Vitro-Our previous study (10) showed that either the N-or C-terminal half of rPLD1 containing one HKD domain did not exhibit PLD activity when transfected into COS-7 cells. The enzymatic activity of rPLD1 required the association of the N-and C-terminal domains (Ref. 10 and this study). The data shown above reveal that interdomain association is also required for Ser/Thr phosphorylation of the enzyme. The question arises whether the phosphorylation is required for catalytic activity of rPLD1. To address this question, we expressed and immunoprecipitated N-terminal Xpress-tagged full-length rPLD1 in COS-7 cells. The immunoprecipitated rPLD1 was divided into three parts and treated with no addition or was dephosphorylated by incubation with Ser/Thr phosphatase mixture, pretreated or not with microcystin at room temperature for 30 min. After the treatment, the immunoprecipitates were reisolated, washed with dephosphorylation buffer, and examined for in vitro transphosphatidylation activity. The Xpress-tagged C-terminal half of rPLD1, which has no catalytic activity, was also expressed and immunoprecipitated from COS-7 cells and used as a background control in the in vitro PLD assay. As shown in Fig. 4, transphosphatidylation activity above background control was observed for the immunoprecipitated full-length rPLD1 that was incubated with no additions at room temperature for 30 min. This PLD activity was stimulated by addition of PKC␣ and PMA. When the immunoprecipitates were first treated with the phosphatase mixture or mock-treated with inactivated phosphatase, transphosphatidylation activity above background was also detected, which was also responsive to treatment with PKC␣ and PMA. Thus, the Ser/Thr phosphorylation of rPLD1 was not essential for the basal catalytic activity of the enzyme or its response to PKC␣ stimulation in vitro. In fact, the phosphatase treatment consistently increased the activity of the enzyme in the presence or absence of PKC␣ plus PMA. The extent of the increase varied from experiment to experiment, with an average of 70%. When the response to the small G proteins was examined, we found that the Ser/Thr phosphorylation was not required for the response of rPLD1 to RhoA or Arf in vitro (Fig. 4B). Compared with the untreated or mocktreated enzyme, the phosphatase-treated rPLD1 had a similar fold response to the stimulation of RhoA or Arf in the presence of GTP␥S (Fig. 4B).
The Association of the N-and C-terminal Domains Required the Coexpression of the Two Domains in Vivo-The combination of two enzyme fragments to restore catalytic activity of a fulllength enzyme has been observed not only for rPLD1, but also for some well characterized enzymes such as ribonuclease S and ␤-galactosidase (32,33). Recent studies have also revealed that the catalytic activity of adenylyl cyclase and phospholipase C-␥1 can be reconstructed by the combination of two protein fragments that contain catalytic domains (34,35). In the case of adenylyl cyclase, the two cytosolic catalytic domains (C 1A and C 2A ) can be expressed separately in Escherichia coli and mixed in vitro to restore the enzymatic activity (34). To examine if mixing the N-and C-terminal domains of rPLD1 could restore catalytic activity in vitro, we transfected the N-and C-terminal halves of rPLD1 alone or in combination in COS-7 cells. As stated before, the N-terminal fragment was tagged with V5 epitope and the C-terminal fragment was fused with an Xpress epitope. After 24 h of expression, the cells were harvested and membrane and cytosolic fractions were prepared. The N-and C-terminal fragments were mainly localized in the crude membrane fraction. Therefore, the membrane fraction was suspended in lysis buffer containing 0.5% Triton X-100 and 0.1% SDS to extract membrane proteins, and used for studying the interaction of the N-and C-terminal domains in vitro. As a negative control, the membrane fraction expressing the V5-tagged N-terminal fragment was mixed with Xpress-tagged ␤-galactosidase during the immunoprecipitation, and no N-terminal fragments were detected by V5 antibodies in the immunoprecipitates by Xpress antibodies (Fig.  5A, lane 1). A protein band corresponding to the size of the N-terminal fragment was detected by V5 antibodies in the Xpress immunoprecipitates from the membrane fraction of COS-7 cells co-transfected with the N-and C-terminal fragments (Fig. 5A, lane 3). However, when the membrane fractions from cells expressing only the N-or C-terminal fragment were mixed, no apparent band was detected by V5 antibodies in the Xpress immunoprecipitates (Fig. 5A, lane 2), although the separately expressed N-and C-terminal fragments were mixed at comparable amounts to those co-expressed in COS-7 cells (Fig. 5B). It is possible that the detergents present in the immunoprecipitation solution might disrupt the potential interaction between the individually expressed N-and C-terminal fragments in vitro. We thus examined the in vitro association between the two halves using the cytosolic fractions which did not contain detergents. Again, the N-terminal fragment was detected by V5 antibodies in the Xpress immunoprecipitates only when the N- The washed immunoprecipitates were then aliquoted equally and further treated with either no addition (rPLD1), with phosphatase (rPLD1 ϩ phosphatase), or with microcystin preinactivated phosphatase (rPLD1 ϩ phosphatase*) for 30 min at 25°C. The resultant products were reisolated by two additional washes with dephosphorylation buffer and resuspended in the dephosphorylation buffer and used for in vitro PLD assay. rPLD1 recovered on protein A-agarose beads from each treatment was analyzed by Western blotting and quantitated by imaging densitometer (Bio-Rad, Model GS-670). The PLD activity was normalized according to the amount of rPLD1. A, the in vitro PLD assays were carried out either with no further addition or in the presence of 50 nM PKC␣ and 1 M PMA for 30 min at 37°C. B, the response of rPLD1 to the small G proteins, RhoA and mArf3, prepared as described in Ref. 26 N-and C-terminal fragments of rPLD1 were expressed in COS-7 cells either alone or in combination. After 24 h of expression, the cells were harvested and fractionated. The membrane fraction was resuspended in lysis buffer with the supplementation of Triton X-100 to a final concentration of 0.5% and SDS to 0.1%, and immunoprecipitated by anti-Xpress antibodies with the indicated combination of membrane or cytosolic fraction mixture. The immunoprecipitates from the membrane fractions (A) or cytosolic fractions (C) were analyzed by SDS-PAGE followed by Western blotting with anti-V5 antibodies. The starting material for the immunoprecipitations from either the membrane (B) or the cytosolic fraction (D) was probed by anti-V5 and anti-Xpress antibodies. Control, the membrane fraction expressing the V5-tagged N-terminal fragment was mixed with Xpress-tagged ␤-galactosidase; "NT" ؉ "CT", the fraction expressing only the N-terminal half of rPLD1 was mixed with that expressing only the C-terminal half; NT/CT, the fraction coexpressing the N-and C-terminal fragments of rPLD1. rPLD1NT or NT, N-terminal fragment of rPLD1 tagged at the C terminus with V5 epitope. CT, C-terminal fragment tagged at the N terminus with Xpress epitope. and C-terminal fragments were coexpressed in the COS-7 cells (Fig. 5C, lane 3). In the control experiments, it was shown that the cytosolic N-or C-terminal fragments were also mixed at comparable amounts to those co-expressed in COS-7 cells (Fig.  5D). Consistent with the inability to associate in vitro, no PLD activity was reconstructed when the individually expressed Nand C-terminal fragments from either the membrane or cytosolic fractions were mixed and assayed by in vitro transphosphatidylation or choline release assay (data not shown). The negative results are unlikely to be due to denaturation of the fragments since the co-expressed N-and C-terminal fragments were subjected to the same procedures, yet exhibited association and PLD activity. Thus, these experiments indicate that the association of the N-and C-terminal fragments of rPLD1 requires their coexpression in vivo.
We observed during our studies of the fractionation of the Nand C-terminal halves of rPLD1 that the phosphorylated form which migrated with slower electrophoretic mobility was only detected in the membrane fraction (Fig. 6, lane 1). This form was also observed in immunoprecipitates from the membrane fraction (Fig. 5A, lane 3). In contrast, the N-terminal fragment that was coexpressed with the C-terminal fragment was detected as a single, faster migrating band in the cytosolic fraction (Fig. 6, lane 2). When the full-length rPLD1 was examined, the modified form that exhibited slow electrophoretic mobility was also observed in the membrane fraction (Fig. 6, lane 3), but not in the cytosolic fraction (Fig. 6, lane 4). Since only a small portion of the N-and C-terminal fragments of rPLD1 appeared in the cytosolic fraction, either more total cytosolic protein was analyzed or the Western blot was exposed longer to ensure that the form with a lower electrophoretic mobility could be detected from the cytosolic fraction. However, these procedures did not detect the modified form of rPLD1 indicating that it is located exclusively in the membrane fraction.
Previous studies of cytosolic rPLD1 in vitro have shown that it has PLD activity and responds to effectors such as small G proteins (7). Transphosphatidylation activity that was stimulated by PKC␣ plus PMA could be detected in the cytosolic fraction of COS-7 cells transfected with full-length rPLD1 or co-transfected with the N-and C-terminal halves of rPLD1 (Fig. 7). Since the rPLD1 and N-terminal half of rPLD1 present in the cytosol are not covalently modified (Fig. 6) these results are consistent with the immunoprecipitation data of Fig. 4 in that they indicate that the phosphorylation modification of rPLD1 is not essential for catalytic activity of the enzyme in vitro.

DISCUSSION
Our previous studies of rPLD1 showed that the N-terminal fragment containing one HKD motif can associate with the C-terminal fragment containing the other HKD motif to form a functional complex (11). However, the domains involved in the association and the nature of the interaction were not clear. By site-directed mutagenesis, we found that mutation to alanine of conserved hydrophobic amino acids within or close to the conserved HKD domains (Fig. 1A) all decreased the association between the N-and C-terminal domains (Fig. 2), although to various extents. The effects of these mutations on the interdomain association correlated closely with their effects on the ability of the domains to reconstruct PLD activity in COS-7 cells (Fig. 1A). Consistent with our previous studies (10) these data demonstrate that the association of the N-and C-terminal domains is important of the catalytic activity of rPLD. Furthermore, they indicate that the HKD motifs contribute to this interdomain interaction. This is consistent with results obtained from the crystal structure of Nuc endonuclease, where the buried interface between the two monomers includes the consensus amino acids in the HKD motifs (24). This association probably organizes the two HKD domains to form a catalytic center.
Studies of Nuc exonuclease and Ymt have suggested that the conserved histidine in the HKD motifs may directly participate in the catalytic reaction by functioning as a nucleophile in the catalysis (23)(24)(25). This mechanism probably applies to PLD as well, since a mutation of the conserved histidine (H464D) in the HKD motif also abolishes the activity of human PLDs. When the corresponding mutation was made in the rPLD1, we found that this mutation not only diminished the catalytic activity of rPLD1, but also dramatically reduced the association between the N-and C-terminal halves (Fig. 2E). Thus, the conserved amino acids in the HKD motifs may have dual roles in both the catalytic reaction and the interdomain association. It should be pointed out that other amino acid sequences besides those within or near the two HKD domains may also be involved in the interdomain interaction. More detailed studies of PLD will be required to define whether such sequences exist.
Interestingly, we found that the association of the N-and C-terminal domains was also required for Ser/Thr phosphorylation modification of rPLD1 at the N-terminal half (Fig. 2). Mammalian PLD1 and PLD2 have been found to be tyrosinephosphorylated in different cell types when treated with agonists (16,36,37). The Ser/Thr phosphorylation of rPLD1 identified in this study occurs in the absence of agonist treatment and the function of the modification needs further investigation. The experiments of Fig. 2 indicate that Ser/Thr phosphorylation is not essential for the association of the N-and Cterminal halves since a major fraction of the N-terminal fragments found in the immunoprecipitates were non-modified. The experiments of Figs. 4 and 7 also indicate that the phosphorylation is not needed for the catalytic activity of rPLD1 or its response to PKC␣ in vitro. In fact, dephosphorylated rPLD1 was reproducibly found to have a higher basal activity ( Fig. 4 and data not shown). Ser/Thr phosphorylation of PLD2 has also been observed when the HeLa cells were treated with okadaic acid, a Ser/Thr protein phosphatase inhibitor (38). The phosphorylation suppressed the catalytic activity of PLD2, while dephosphorylation relieved the suppression (38). Thus, Ser/Thr phosphorylation of PLD1 and PLD2 may play a role in regulating the catalytic activity of these enzymes.
Another property of rPLD1 that was affected by the modification was the localization of the enzyme. As shown in Fig. 6, Ser/Thr-phosphorylated rPLD1 was located exclusively in the membrane fraction. A phosphorylation modification has been detected in SPO14, a yeast PLD isoform (39). This modification did not affect its catalytic activity in vitro, however, it specifically regulated the localization of SPO14, and was found to be essential for the formation of the new spore membrane during meiosis. PLD1 has been found not only in intracellular membranes but also in caveolae (18), which are specific domains on plasma membranes that are enriched in signaling molecules. Thus it is possible that the phosphorylation of rPLD1 may be essential for the physiological function of the enzyme by specifying its cellular location in vivo. A rPLD1 mutant that lacks the Ser/Thr phosphorylation sites could be used in future studies to examine whether the phosphorylation affects the catalytic activity and function of the enzyme in vivo.
Studies of the association of the N-and C-terminal domains in vitro showed that the individually expressed N-and Cterminal halves did not interact in vitro when mixed. Instead, the association was only detected when the N-and the Cterminal domains were coexpressed in the COS-7 cells (Fig. 5,  A and C). Thus the association of the N-and C-terminal domains likely occurs during their translation and may involve specific folding processes of the two fragments. This property has been observed for PLC␥, where the conserved catalytic domains of the enzyme, X and Y, can interact and reconstruct the catalytic activity only when coexpressed in cells (35). Reconstruction of catalytic activity by two fragments has also been documented for adenylyl cyclase, although the two catalytic domains, C 1A and C 2A , can be individually expressed and mixed in vitro to restore enzymatic activity (34). In addition, the enzymatic activity of adenylyl cyclase can be regulated through changes in the affinity of the association of the two domains induced by effectors of the enzyme (34). It is possible that the activity of PLD may also be regulated by the association between N-and C-terminal fragments of PLD. Since the amino acids required for catalysis and association overlap at the HKD regions, a slight change in the association may alter catalytic activity. Effectors of PLD could regulate its enzymatic activity by modifying the interdomain association.
Although our studies have shown that conserved hydrophobic amino acids in the HKD domains are important for both the catalytic activity and interdomain association of rPLD1, structural studies will be needed to fully understand how the domains associate and construct a catalytic center. The recognition that the enzyme is Ser/Thr phosphorylated on its N-terminal half and that this modification is only seen in the membraneassociated form also requires further work to elucidate the functional significance of the phosphorylation and to identify the sites and protein kinase(s) involved, and also to define the role of the C-terminal half in promoting the phosphorylation.