Inhibition of phospholipase D by amphiphysins.

Two distinct proteins inhibiting phospholipase D (PLD) activity in rat brain cytosol were previously purified and identified as synaptojanin and AP180, which are specific to nerve terminals and associate with the clathrin coat. Two additional PLD-inhibitory proteins have now been purified and identified as the amphiphysins I and II, which forms a heterodimer that also associates with the clathrin coat. Bacterially expressed recombinant amphiphysins inhibited both PLD1 and PLD2 isozymes in vitro with a potency similar to that of brain amphiphysin (median inhibitory concentration of approximately 15 nm). Expressions of either amphiphysin in COS-7 cells reduced activity of endogenous PLD as well as exogenously expressed PLD1 and PLD2. Coprecipitation experiments suggested that the inhibitory effect of amphiphysins results from their direct interaction with PLDs. The NH(2) terminus of amphiphysin I was critical for both inhibition of and binding to PLD. Phosphatidic acid formed by signal-induced PLD is thought to be required for the assembly of clathrin-coated vesicles during endocytosis. Thus, the inhibition of PLD by amphiphysins, synaptojanin, and AP180 might play an important role in synaptic vesicle trafficking.

Phosphatidylcholine (PC) 1 -specific phospholipase D (PLD) catalyzes the hydrolysis of PC to produce choline and phosphatidic acid (PA) (1). PA produced by PLD as a result of signaling activity is thought to play many roles as an intracellular messenger, and one of its functions is promoting formation of the clathrin coat in Golgi and endoplasmic reticulum, Golgi, trans-Golgi network, and lysosomes (2)(3)(4)(5)(6). PA can be further hydrolyzed to diacylglycerol by a specific phosphatase or to lysophosphatidic acid by phospholipase A 2 . Diacylglycerol activates protein kinase C, and lysophosphatidic acid is a potent mitogen (7). Although the basal activity of PLD in mammalian cells is low, the enzyme is activated in a variety of cells by a wide range of stimuli, including hormones, growth factors, neurotransmitters, and cytokines (1,8). Activation of PLD is thought to occur through its interaction with small GTP-binding proteins such as ADP-ribosylation factor (ARF) and Rho as well as through protein kinase C (9).
PLD has been cloned from a wide variety of species ranging from bacteria to humans. Two distinct mammalian PLD genes have been identified in humans (10 -12), mice (13,14), and rat (15,16), and the encoded proteins, PLD1 and PLD2, share ϳ50% amino acid sequence identity. Purified PLD1 exhibits a low basal activity but is markedly stimulated in a synergistic manner by protein kinase C-␣, ARF, or Rho in the presence of phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ), a cofactor that is required for PLD1 activation (1,17). PLD2 is also dependent on PI(4,5)P 2 but differs from PLD1 in that it exhibits constitutively high activity; it also appears to be weakly activated by ARF (12,18). PLD1 has been reported to be localized in endoplasmic reticulum, secretary granules, and lysosomes and PLD2 in plasma membrane (13,19). However, PLD1 upon stimulation was found to translocate to plasma membrane in RBL-2H3 cells (19).
We have previously shown that rat brain cytosol contains several distinct proteins that inhibit PLD activity as measured in the presence of PI(4,5)P 2 and ARF (20). Two such inhibitory proteins were purified to homogeneity and identified as synaptojanin and clathrin assembly protein 180 (AP180, previously named as clathrin assembly protein 3 or AP3) (21,22). The inhibition of PLD by synaptojanin was attributed to its ability to dephosphorylate PI(4,5)P 2 (21,23), whereas the inhibitory effect of AP180 is a result of its direct interaction with PLD (22). We have now purified a third type of PLD inhibitor from rat brain and have shown that it comprises amphiphysin I (AmphI) and AmphII, both of which are nerve-terminal proteins that are important for clathrin-mediated endocytosis (24,25).

EXPERIMENTAL PROCEDURES
Materials-Rat brains were obtained from Pel-Freez Biologicals (Rogers, AR); bovine brain PC and phosphatidylethanolamine (PE) were from Avanti Polar Lipids (Alabaster, AL); PI(4,5)P 2 was from Roche Molecular Biochemicals; [choline-methylϪ 3 H]dipalmitoyl-PC [(pam) 2 PC] (50 Ci/mmol) was from DuPont; and n-octyl-␤-D-glucopyranoside was from Calbiochem. ARF was purified from rat brain as described (20). Rabbit antibodies generated in response to the COOHterminal 12 amino acids of human PLD1, but which also recognize mouse PLD2, were kindly provided by Sung Ho Ryu (Postech, Pohang, Korea). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Recombinant PLD1 and PLD2-Human PLD1 and mouse PLD2 were isolated from Sf9 insect cells that had been infected with a recombinant baculovirus containing PLD1 or PLD2 cDNA (10,14). Partially purified PLD1 was obtained by fractionating detergent-solubilized Sf9 cell membrane proteins on a heparin-5PW column as described (21). PLD2-expressing Sf9 cells were lysed by sonication in a solution containing 20 mM Tris-HCl (pH 7.5), 1 mM EGTA, 1 mM MgCl 2 , 1 mM dithiothreitol, and 0.1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride. The lysate was then centrifuged at 50,000 ϫ g for 30 min, and the resulting supernatant was used as the source of PLD2.
In Vitro Assay of PLD Activity-ARF-dependent PLD1 activity was assayed by measuring the generation of 3 H-labeled choline from [choline-methylϪ 3 H](pam) 2 PC as described (21). Briefly, 25 l of lipid vesicles containing PE, PI(4,5)P 2 , and PC (with 200,000 cpm of [choline-methylϪ 3 H](pam) 2 PC) in a molar ratio of 16:1.4:1 were added to 125 l of a mixture containing partially purified PLD1, 100 nM ARF, 5 mM guanosine 5Ј-O-(3Ј-thiotriphosphate), 50 mM Hepes-NaOH (pH 7.5), 3 mM EGTA, 80 mM KCl, 2.5 mM MgCl 2 , and 2 mM CaCl 2 . The final concentration of PC was 3.4 mM, and the free Ca 2ϩ concentration was calculated to be 300 nM. The assay procedure for PLD2 was identical to that for PLD1 with the exception that ARF and GTP␥S were omitted from the reaction mixture. The reactions were stopped by adding 1 ml of a mixture of chloroform and methanol (1:1, v/v) followed by 350 l of 1 M HCl. The resulting upper layer (500 l) containing the released [ 3 H]choline was subjected to liquid scintillation spectrometry.
Purification of PLD Inhibitors-Fractions containing PLD inhibitory activity were obtained by subjecting rat brain cytosol to chromatography on a column of DEAE-Sephacel as described (20). Fractions corresponding to the second peak (peak II) of inhibitory activity (see Fig. 1 of Ref. 20) were pooled, dialyzed against purification buffer (20 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, leupeptin (1 g/ml), and aprotinin (1 g/ml)), and centrifuged to remove insoluble particles. The resulting supernatant (220 mg of protein) was applied to a TSK-gel DEAE-5PW column (21.5 ϫ 150 mm) that had been equilibrated with purification buffer. After washing the column with purification buffer, proteins were eluted at a flow rate of 5 ml/min with a 400-ml linear gradient of 0 -1.0 M NaCl in purification buffer. Fractions (5 ml) were collected and assayed for PLD1 inhibitory activity. Fractions corresponding to the peak of inhibitory activity were pooled, adjusted to 0.8 M ammonium sulfate by addition of a saturated solution, and centrifuged to remove insoluble particles. The resulting supernatant (84 mg of protein) was applied to a TSK-gel phenyl-5PW column (21.5 ϫ 150 mm) that had been equilibrated with purification buffer containing 0.8 M ammonium sulfate. After washing the column with purification buffer containing 0.8 M ammonium sulfate, proteins were eluted over 60 min at a flow rate of 5 ml/min with a decreasing linear gradient of ammonium sulfate from 0.8 -0 M in purification buffer. Fractions (5 ml) were collected, and those corresponding to the peak of PLD1 inhibitory activity were pooled, dialyzed against purification buffer, and centrifuged. The resulting supernatant (10 mg of protein) was applied to a TSK-gel heparin-5PW column (7.5 ϫ 75 mm) that had been equilibrated with purification buffer. Proteins were eluted at a flow rate of 1 ml/min with a linear NaCl gradient from 0 to 0.2 M for 10 min, from 0.2 to 0.4 M for 30 min, and from 0.4 to 1.0 M for 10 min. Fractions (1 ml) were collected and assayed for PLD1-inhibitory activity, two peaks of which, centered on fractions 23 and 28, were obtained (Fig. 1A). A pool of of fractions 22-25 from the heparin-5PW column was concentrated, and 25 g in 50 l was further applied to Superose 12 gel column (3.2 ϫ 300 mm). This sizing chromatography was performed on a Smart Chromatography system (Amersham Pharmacia Biotech) equipped with microseparation unit. Proteins were eluted at a flow rate of 40 l/min with the buffer for an hour. Collected fractions were assayed for the inhibitory assay, analyzed on 8% gel, and visualized with the silver staining method (Fig. 1D).
Peptide Purification-Proteins in fraction 28 of the heparin 5-PW column (Fig. 1A) were reductively denatured by treatment with 6 M guanidine hydrochloride and 2 mM dithiothreitol in 50 mM Tris-HCl (pH 8.0), after which sulfhydryl groups were labeled with 10 mM iodoacetamide. The labeled proteins were digested with V8 protease, and the resulting peptides were applied to a C 18 hydrophobic column (4.6 ϫ 250 mm; Vydac) that had been equilibrated with 0.1% trifluoroacetic acid and were eluted at a flow rate of 1 ml/min with a 60-ml linear gradient of 0 -60% (v/v) acetonitrile in 0.1% trifluoroacetic acid. Peptides were detected by measuring absorbance at 215 nm. Fractions corresponding to the highest four peaks were separately collected, dried, and subjected to digestion with trypsin, and the resulting peptides were purified on the Vydac C 18 column by applying the same gradient as before. The 83-kDa protein from fraction 23 of the heparin 5-PW column (Fig. 1B) was electroeluted from gel bands and digested with Lys-C protease. The resulting peptides were purified on the Vydac C 18 column as described above. Purified peptides were subjected to amino acid sequence analysis.
Preparation of GST-Amph Fusion Proteins-A cDNA (hAmphy-Z1A) encoding full-length (695-residue) human AmphI has been described (26), and a plasmid [pEXlox(ϩ)] that encodes a 434-residue variant of mouse AmphII was kindly provided by Dr. B. Kay (University of Wisconsin). The AmphI and AmphII cDNAs were separately ligated into the 5Ј BamHI and 3Ј EcoRI sites of the pGEX4T1 vector (Amersham Pharmacia Biotech) by standard techniques of subcloning and the polymerase chain reaction to produce expression vectors (pGEX-AmphI and pGEX-AmphII) encoding the respective GST fusion proteins. Similarly, truncated cDNAs encoding the NH 2 -terminal 373 residues or COOH-terminal 404 residues of AmphI were cloned into the 5Ј BamHI and 3Ј EcoRI sites of pGEX4T1 to produce pGEX-N-AmphI and pGEX-C-AmphI. Escherichia coli BL21 cells were transfected with the various expression vectors encoding the GST fusion proteins and grown at 37°C until the absorbance at 600 nm of the culture reached 0.5-0.6. Expression of GST-Amph proteins was then induced by incubation of the cells for 3 h in the presence of 100 M isopropyl-␤-D-thiogalactopyranoside. Cells were collected, washed with phosphate-buffered saline (PBS), sonicated in PBS containing 1% (v/v) Triton X-100, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, N-tosyl-L-phenylalanine chloromethyl ketone (1 g/ml), and N-tosyl-L-lysine chloromethyl ketone (1 g/ml), and centrifuged at 25,000 ϫ g for 30 min. GST fusion proteins were purified from the resulting supernatant with the use of glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The concentration of purified proteins was determined spectroscopically with extinction coefficients at 280 nm of 39,970 for GST, 97,870 for GST-AmphI, 81,700 for GST-AmphII, 70,140 for GST-N-AmphI, and 69,370 for GST-C-AmphI.
Antibodies to Amph Proteins-Rabbit monospecific antibodies to Am-phI were described previously (26). Rabbit antisera to N-AmphI or to C-AmphI were prepared by injection of recombinant proteins obtained by removal of the GST moiety from the corresponding GST fusion proteins with the use of thrombin.
Expression of Recombinant PLD and Amphiphysins in Mammalian Cells-pCGN expression vectors containing cDNAs for full-length PLD1 and PLD2 have been described previously (10,14). pCGN also encodes an influenza virus HA tag that becomes expressed in-frame at the NH 2 terminus of the PLD proteins. The XbaI-SmaI fragments of the pCGN vectors were subcloned into the mammalian expression plasmid pCDNA3.1 (Invitrogen), yielding pCDNA-PLD1 and pCDNA-PLD2. For preparation of amphiphysin expression vectors, the BamHI-EcoRI fragments from pGEX-AmphI and pGEX-AmphII were inserted into pCDNA3.1, yielding pCDNA-AmphI and pCDNA-AmphII, respectively. Subconfluent COS-7 cells were harvested by exposure to trypsin, centrifuged at 700 ϫ g for 5 min, and washed once with ice-cold transfection buffer (10 mM sodium phosphate buffer (pH 7.4), 1 mM MgCl 2 , 250 mM sucrose) (27). Cells (4 ϫ 10 6 in 400 l) were placed in a 0.4-mm gap cuvette and subjected to electroporation (two 99-s pulses at 850 V) in the presence of the indicated plasmids with a T820 electroporater (BTX, San Diego, CA). The cells were maintained on ice for 10 min and then mixed with 9 ml of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS).
A retroviral vector for the expression of AmphI (pSinRep5-AmphI) was constructed by inserting the NheI-EcoRV fragment of pCDNA-AmphI into a pSinRep5 virus vector (Invitrogen) that had been digested with XbaI and StuI. By using the empty pSinRep5 vector (Invitrogen) and the AmphI-expressing vectors, control and AmphIexpressing virus were generated with the Sindbis Expression System according to manufacturer's guide, and their titers were determined by measuring relative luminescent unit in BHKn-LUC cells (ATCC). The volumes of virus stocks were adjusted to equalize titers.
Measurement of PLD Activity in Intact Cells-PLD activity in cells was assayed by measuring the accumulation of 3 H-labeled phosphatidylethanol. Transfected COS-7 cells (7 ϫ 10 5 per well) were labeled by incubation for 18 h in 2 ml of DMEM containing 10 Ci of [ 3 H]myristic acid (ICN) and 1% FBS. The labeled cells were washed with DMEM, kept in 2 ml of DMEM containing 0.5% ethanol for 5 min, and then incubated for 30 min in the absence or presence of 100 nM PMA. The medium was then removed and replaced with ice-cold PBS to terminate cell stimulation. The cells were scraped into an ice-cold vial and then extracted with a mixture of chloroform and methanol (2:1). The organic phase was dried, redissolved with an equal mixture of chloroform and methanol, and spotted onto Silica Gel 60 thin layer chromatography plates (Merck). The labeled compounds were separated with the use of the organic phase of a mixture of ethyl acetate:n-octylbutane:acetic acid:methanol:water (13:2:3:10:10) (28). The phosphatidylethanol spots were detected by staining with iodine vapor, scraped into a vial, and sonicated in scintillation mixture. The amount of radioactivity associated with these spots was normalized on the basis of the radioactivity incorporated into total lipids of the cells and was then used as the measure of PLD activity.

RESULTS
Purification and Identification of PLD-inhibitory Proteins-We previously showed that chromatography of rat brain cytosol on a DEAE-Sephacel column yielded three partially resolved peaks of PLD inhibitory activity (20). Fractions corresponding to the first peak were further separated into two peaks, IA and IB, that yielded synaptojanin and AP180, respectively. Further purification of the second peak (peak II) from the DEAE-Sephacel column was performed by sequential high pressure liquid chromatography (HPLC) on DEAE-5PW, phenyl-5PW, and heparin-5PW columns. Fractions were assayed for their ability to inhibit PLD1 activity measured in the presence of ARF and PI(4,5)P 2 .
The heparin-5PW column yielded two peaks of PLD inhibitory activity centered on fractions 23 and 28 (Fig. 1A). Analysis of fraction 23 by SDS-polyacrylamide gel electrophoresis (PAGE) on an 8% gel revealed two major broad bands corresponding to proteins with apparent molecular masses of 125 and 83 kDa, as well as a minor band that migrated slightly ahead of the 83-kDa protein (Fig. 1B). Fraction 28 yielded a single broad band corresponding to a protein of 125 kDa. This latter protein was digested with V8 protease, and the resulting fragments were fractionated by HPLC on a C 18 column. The FIG. 1. Purification of PLD-inhibitory proteins from rat brain cytosol. A, PLD-inhibitory proteins that had been partially purified from rat brain by chromatography on a DEAE-Sephacel column (peak II from Fig. 1 in Ref. 20), followed by sequential HPLC on DEAE-5PW and phenyl-5PW columns, were further subjected to HPLC on a heparin-5PW column, as described under "Experimental Procedures." B, the indicated fractions from the heparin-5PW column were subjected to SDS-PAGE on an 8% gel; the separated proteins were then either visualized by silver staining (upper panel) or transferred to a nitrocellulose sheet and subjected to immunoblot analysis with rabbit polyclonal antibodies to AmphI and horseradish peroxidase-conjugated antibodies to rabbit immunoglobulin G (lower panel). The positions of molecular size standards (in kilodaltons) are shown on the right. C, fractions 23 (open squares) and 28 (closed squares) from the heparin-5PW column (at the indicated amounts of protein) were assayed for their inhibitory effect on PLD1 activity. Data are expressed as a percentage of the PLD activity apparent in the absence of test protein and are means of two separate assays. D, a pool of fractions 22-25 of the heparin-5PW column was further separated by Superose gel column for the micropurification as described under "Experimental Procedures." Each fraction was tested for the inhibitory activity. Inset is SDS-PAGE analysis on 8% gel that was visualized by the silver staining method. E, fraction 25 from the Superose gel column was subjected to PAGE on a nondenaturing 6% gel (left panel). The protein corresponding to the major (240 kDa) band was eluted from the native gel and analyzed by SDS-PAGE on an 8% gel (right panel). The positions of molecular size standards are indicated on the right of each panel.
fractions corresponding to each of the four major peaks obtained each appeared to contain more than one peptide; these fractions were therefore incubated with trypsin and again subjected to chromatography on the same C 18  The human AmphI cDNA encodes a protein of 695 amino acids with a calculated molecular mass of 76,256 (26). The mobility of human AmphI on SDS-PAGE was previously shown to be less than that expected from its predicted size, which is likely a result of the acidic nature of the protein (isoelectric point, 4.50) (26). Identification of the 125-kDa protein in fractions 20 -28 from the heparin-5PW column as AmphI was further confirmed by immunoblot analysis with monospecific antibodies to AmphI (Fig. 1B). The broadness of the AmphI bands in Fig. 1B is likely attributable to the fact that AmphI is constitutively phosphorylated at multiple sites (29).
The potencies of fractions 23 and 28 from the heparin-5PW column with regard to inhibition of PLD1 were similar (Fig.  1C), suggesting that both the 83-and 125-kDa proteins in fraction 23 were able to inhibit PLD1 activity. However, neither antibodies to the NH 2 -terminal 373 residues of AmphI nor those to the COOH-terminal 404 residues of AmphI recognized the 83-kDa protein (data not shown), suggesting that the latter was not derived from AmphI. To separate further these two bands, a pool of fraction 22-25 from the heparin-5PW column was subjected to micropurification by Superose gel column, and each fraction was visualized by silver staining 8% on gel (Fig.  1D). Because the staining intensities of the 125-and 83-kDa proteins in the fractions from the gel column showed equal ratio (Fig. 1D, inset), we investigated whether the two polypeptides exist as a complex. PAGE analysis of fraction 25 on a nondenaturing gel yielded one major band corresponding to an apparent molecular mass of 240 kDa (Fig. 1E). When the 240-kDa protein was electroeluted from the native gel and then subjected to SDS-PAGE, two bands of 125 and 83 kDa were apparent (Fig. 1E), suggesting that the 125-and 83-kDa proteins indeed form a complex. The 83-kDa protein was then electroeluted from an SDS-PAGE gel and digested with Lys-C, and the resulting fragments were separated by HPLC on a C 18 column. The sequence of one purified peptide (SPSPPP-DGSPAATPEIRV) matched perfectly that of residues 296 -313 of rat AmphII (30). At least 10 splice variants of rat AmphII have been identified, with the largest such variant consisting of 588 amino acids. Unlike AmphI, AmphII proteins are not acidic and do not migrate abnormally on SDS-PAGE.
Inhibition of PLD1 and PLD2 by GST Fusion Proteins of AmphI and AmphII-We prepared glutathione S-transferase (GST) fusion proteins containing full-length (695 residues) human AmphI (GST-AmphI), the NH 2 -terminal 373 amino acids (GST-N-AmphI), or the COOH-terminal 404 amino acids (GST-C-AmphI) of AmphI, or a full-length 434-residue variant of mouse AmphII (GST-AmphII) and then tested their abilities to inhibit PLD1 and PLD2. The bacterially expressed GST fusion proteins were isolated with the use of glutathione-Sepharose beads, and their purity was verified by SDS-PAGE ( Fig. 2A). Both GST-AmphI and GST-AmphII inhibited PLD1 activity in a concentration-dependent manner which was measured in the presence of ARF and PI(4,5)P 2 (Fig. 2B); the two fusion proteins showed similar potencies, with half-maximal inhibition apparent at 10 -15 nM, similar to the value of 5 nM calculated from the data in Fig. 1C for Amph proteins purified from rat brain. GST alone had no effect on PLD1 activity. The inhibitory potency and efficacy of GST-N-AmphI were similar to those of GST-AmphI, whereas GST-C-AmphI had virtually no effect on PLD1 activity (Fig. 2B). The NH 2 -terminal 373 residues, but not the COOH-terminal 404 residues, of AmphI thus appear critical for inhibition of PLD1. PLD2 activity measured in the presence of PI(4,5)P 2 was also inhibited by GST-AmphI and by GST-AmphII with similar potencies (Fig. 2B).
We also investigated the possibility that the Amph-dependent inhibition of PLD was attributable to blockage of the access of PLD to PI(4,5)P 2 . The ARF-dependent activity of PLD1 in the absence of PI(4,5)P 2 is about 5% of that measured in the presence of PI(4,5)P 2 (21). Inhibition of PLD1 by various concentrations of GST-AmphI was measured with substrate vesicles containing or not containing PI(4,5)P 2 , with the amount of PLD1 used in the absence of PI(4,5)P 2 being 20 times that used in its presence. The concentration dependence of inhibition of PLD1 by GST-AmphI was similar in the absence or presence of PI(4,5)P 2 (Fig. 3), indicating that AmphI does not inhibit PLD activity by interfering with the interaction of the enzyme with PI(4,5)P 2 . The observation that GST-AmphI and GST-AmphII potently inhibited the activity of PLD2 measured in the absence of ARF also suggests that the inhibition of PLD1 activity is not due to interference with ARF binding.
Physical Association of Amphiphysins with PLD-We next investigated whether the amphiphysin proteins physically interact with PLD enzymes. GST-AmphI or GST-AmphII was mixed with partially purified PLD1 or PLD2, and the GST fusion proteins were then precipitated with glutathione-Sepharose 4B. The resulting precipitates and supernatants were subjected to immunoblot analysis with antibodies that recognize both PLD1 and PLD2. Both GST-AmphI and GST-AmphII associated tightly with PLD1 and PLD2 as indicated by the presence of the PLD proteins in the fusion protein precipitates and their absence from the supernatants (Fig. 4). GST alone did not interact with either PLD enzyme. Similar precipitation experiments revealed that PLD1 associated with GST-N-AmphI but not with GST-C-AmphI (Fig. 4), consistent with the observation that PLD1 activity was inhibited by GST-N-AmphI but not by GST-C-AmphI (Fig. 2B).
To determine whether AmphI forms a complex with PLD enzymes in intact cells, we cotransfected COS-7 cells with vectors encoding AmphI and either hemagglutinin epitope (HA)-tagged PLD1 or HA-tagged PLD2. Cell lysates were subjected to immunoprecipitation with a monoclonal antibody to HA, and the resulting precipitates were subjected to immunoblot analysis with antibodies to AmphI. AmphI was detected in both the PLD1 and PLD2 immunoprecipitates; AmphI immunoreactivity was not apparent in precipitates prepared from control cells not expressing either recombinant AmphI or HA-tagged PLD (Fig. 5). These results demonstrate that AmphI tightly associates with PLD1 and PLD2 in intact cells.
Inhibition of PLD by Amphiphysins in Intact Cells-The effects of amphiphysins on PLD activity in intact cells were investigated by transient expression of AmphI or AmphII in COS-7 cells with the use of pCDNA3.1 vector. Because endogenous PLD activity in unstimulated COS-7 cells is too low for reliable measurement, activity was measured after stimulation of the cells with phorbol 12-myristate 13-acetate (PMA). The PMA-stimulated PLD activity was reduced by ϳ25% in cells expressing AmphI or AmphII (Fig. 6), suggesting that the endogenous PLD activity was inhibited by the exogenous amphiphysins. In Fig. 6, inset, the AmphI blot was shown but not AmphII because the antibody was not available. This relatively low percentage of inhibition, given the marked inhibitory efficacy and potency of AmphI and AmphII apparent in vitro (Fig.  2B), is likely attributable to the fact that only a fraction (usually 20 -40%) of the cells actually became transfected, and the PLD activity of the nontransfected cells thus gives rise to a high background.
In another experiment, COS-7 cells were transiently cotrans- . After the addition of 10 l of glutathione-Sepharose 4B slurry (50%, w/v), the mixtures were incubated for an additional 5 min and then centrifuged at 5,000 ϫ g for 5 min. The resulting precipitates were washed three times with the incubation buffer, after which proteins bound to the glutathione beads were released by incubation at 95°C with SDS sample buffer. Both the precipitate (PPT) and supernatant (SUP) of the bead incubations were subjected to SDS-PAGE on an 8% gel and immunoblot analysis with rabbit antibodies that recognize both PLD1 and PLD2. The positions of PLD1 and PLD2 are indicated. fected with vectors encoding Amph (AmphI or AmphII) and PLD (PLD1 or PLD2), and PMA-stimulated PLD activity was measured. Expression of recombinant PLD1 or PLD2 alone resulted in an approximately 2-fold increase in PLD activity compared with that of cells transfected with the empty vector (Fig. 7). Approximately 70 -80% of this increase in PLD activity was inhibited by coexpression of AmphI or AmphII.
To overcome the low transfection efficiency, we prepared a retrovirus vector harboring AmphI cDNA. Infection of baby hamster kidney (BHK) cells with this virus revealed that the PMA-stimulated activity of endogenous PLD was inhibited by the expression of AmphI in a dose-dependent manner, with virtually complete inhibition apparent at the highest tested dose of the virus (Fig. 8). DISCUSSION AmphI was originally identified as a brain protein associated with synaptic vesicles (31). Subsequently, AmphII (of which various splice variants have been named BIN-1, SH3P9, BRAMP-2, or ALP) was cloned by several laboratories (32)(33)(34). The largest isoform of AmphII, comprising 588 amino acids, appears to be brain-specific, whereas smaller isoforms show a widespread tissue distribution. In the brain, AmphI and AmphII are highly concentrated in nerve terminals. The two proteins were shown to form a stable heterodimer, which is the predominant form in brain (30). Both AmphI and AmphII contain an Src homology 3 (SH3) domain in their COOH-terminal regions (30,35,36).
Amphiphysins are implicated in clathrin-mediated endocytosis in nerve terminals, and AmphII is also implicated in endocytosis outside synapses (24,25). Assembly of clathrin coats is thought to be initiated by binding of the AP2 adapter protein complex to membrane receptors. Through its interaction with the AP2 complex, clathrin assembles and polymerizes to form a polygonal lattice, eventually leading to the formation of a coated bud. Dynamin is then recruited to the neck of the invaginating clathrin-coated vesicle, possibly to effect vesicle closure by activating its downstream effector, endophilin I which has lysophosphatidic acid acyl transferase activity to form PA around neck squeezed by dynamin. PA thus produced facilitates a drastic change in the membrane curvature (37). It is important to note that PA derived by the action of endophilin I was proposed to contain much higher content of arachidonic acid compared with PA derived by the action of PLD. The AmphI-AmphII heterodimer is thought to be responsible for dynamin recruitment as a result of its ability to bind dynamin and AP2 simultaneously. The heterodimer interacts with multiple dynamin molecules through its two SH3 domains and simultaneously binds to AP2 through a region distinct from its SH3 domains. Amphiphysins also bind through their SH3 domains to synaptojanin. This protein negatively regulates the recruitment of dynamin by competing with it for binding to the SH3 domains of amphiphysins as well as by hydrolyzing PI(4,5)P 2 , which serves as a binding site for the pleckstrin homology domain of dynamin (23,36,38). Clathrin-mediated endocytosis in nerve terminals requires many other protein components, including AP180, which promote coat assembly by They were then rinsed three times with ice-cold PBS and scraped into Eppendorf microcentrifuge tubes. The cells were isolated by a brief centrifugation and sonicated in the presence of 1 ml of lysis buffer (1% n-octyl-␤-D-glucopyranoside, 150 mM NaCl, 100 mM Tris-HCl (pH 8.5), 1 mM EGTA, 1 mM dithiothreitol, leupeptin (1 g/ml), aprotinin (1 g/ml), and 1 mM phenylmethylsulfonyl fluoride). The lysates were centrifuged at maximum speed for 20 min, and the resulting supernatants were pretreated with protein A and then subjected to immunoprecipitation (IP) with a monoclonal antibody to HA (␣HA) and protein A. Immune complexes were collected by a brief centrifugation, fractionated by SDS-PAGE on an 8% gel, and subjected to immunoblot (IB) analysis with rabbit polyclonal antibodies to AmphI and horseradish peroxidase-conjugated antibodies to rabbit immunoglobulin G (upper panel). The same membrane was subsequently also probed with the monoclonal antibody to HA and horseradish peroxidase-conjugated antibodies to mouse immunoglobulin G (middle panel). Cell lysates were also subjected to direct immunoblot analysis with antibodies to AmphI (lower panel). The position of an 83-kDa molecular size standard is indicated. interacting with clathrin as well as with AP2.
We have previously shown that synaptojanin and AP180 each inhibit PLD activity (21,22). We have now demonstrated that AmphI and AmphII each inhibit the activities of both PLD1 and PLD2 as a result of direct interaction with these enzymes. Our observation that the COOH-terminal 404 residues of AmphI were not required for inhibition of PLD indicates that the SH3 domain of AmphI does not contribute to this effect. Although the NH 2 terminus of AP-180 and AmphI showed PLD inhibition, their NH 2 termini did not show strong common denominators. Similar to our results, GST fusion proteins of full-length and NH 2 -terminal AmphI showed tubulation of liposomes, but the COOH-terminal ones could not (39).
Recent evidence suggests that PA and PI(4,5)P 2 also play pivotal roles in the initiation of vesicle complex formation in several types of coated membrane (2,40), probably by interacting with AP2, AP180, dynamin, and synaptotagmin (41)(42)(43)(44)(45). The production of PA on the lysosomal membrane induced by treatment with bacterial PLD was reported to be sufficient to trigger AP2 translocation and limited coat assembly (6). An additional role of PA was proposed to change the membrane curvature of neck already constricted by dynamin and to accelerate vesicle fission (37). PIP2 formed by PI(4)P 5-kinase was shown to be necessary for Ca 2ϩ -activated secretion in neuroendocrine cells (46). The formation of constitutive secretory vesicles from trans-Golgi network was activated by PI transport protein (47). In yeast enzymes related to PI metabolism were demonstrated to be critical in the vesicular transportation (48,49). The production of these two acidic lipids is closely linked. Thus, PI(4,5)P 2 is a cofactor for PLD and is generated by the sequential action of PI 4-kinase and PI(4)P 5-kinase. PA, produced by the action of PLD, is a potent activator of PI(4)P 5-kinase (50 -52) and thereby increases the synthesis of PI(4,5)P 2 , which, in turn, leads to further stimulation of PLD activity. This positive feedback loop would be expected to result in a rapid increase in the cellular concentrations of PA and PI(4,5)P 2 , as well as a consequent marked change in the microdomain architecture of the lipid bilayer that would facilitate coat assembly (21,22,53,54). According to this model, it would be necessary to halt the feedback loop before initiation of the disassembly of the coated vesicle and its subsequent fusion with the target membrane. We propose that inhibition of PLD by amphiphysins and AP180, together with the hydrolysis of PI(4,5)P 2 by synaptojanin, provides a mechanism to break the feedback loop. Given that AmphI, AP180, and synaptojanin are all nerve terminalspecific proteins, such a mechanism would be specific to synaptic vesicles. Such redundant inhibition of PLD by multiple proteins might be necessary during synaptic transmission to ensure a rapid response; the rate of response is likely less critical for endocytotic events in other cell types, where AmphII might suffice by itself. Finally, it is important that PLD inhibition occurs after coat assembly to prevent further hydrolysis of PC. Such timing might be related to the facts that synaptojanin and dynamin compete for binding to the SH3 domains of amphiphysins and that AmphI, AP180, synaptojanin, and dynamin are all phosphoproteins that undergo stimulation-dependent dephosphorylation in nerve terminals (55).  5 ) that had been grown in 6-well plate with ␣minimum Eagle's medium containing 5% FBS were infected by incubation for 12 h at 37°C in 2 ml of the same medium containing 0.5% FBS and the indicated volumes (in microliters) of control and AmphI viruses; the cells were labeled for the same period by inclusion of 10 Ci of [ 3 H]myristic acid in the infection mixture. PMA-stimulated PLD activity was then measured as described in Fig. 6. Data are expressed as a percentage of PMA-stimulated PLD activity measured in cells infected with the control virus alone and are means of two independent experiments. Inset, immunoblot analysis of cell lysates with antibodies against AmphI. The order of three immunoblot samples is the same as that for PLD activity measurement.