Inhibition of Phospholipase D Activity by Fodrin AN ACTIVE ROLE FOR THE CYTOSKELETON*

Phospholipase D (PLD) is a major enzyme implicated in important cellular processes such as secretion and proliferation. The knowledge of its regulation is essen- tial to understand the control of these phenomena. Several proteins activating PLD have been described in the last years. In this report, we chromatographed bovine brain cytosolic proteins to identify fodrin, the non- erythroid spectrin, as the first described inhibitor of PLD. A cytosolic fraction with an inhibitory effect on PLD activity loses its capacity after immunoprecipitation of fodrin. Moreover, at 1 n M , purified fodrin blocks fully and quickly PLD activity, whatever the stimuli used. In contrast, fodrin has no effect on adenylate cy- clase activity. Fodrin-analogous proteins like dimeric or tetrameric erythroid spectrin have the same inhibitory effect on PLD, at higher concentrations. Other cytoskeletal proteins, actin and vimentin, are inefficient on PLD inhibition. The mechanisms implicated in PLD modula-tion such as post-translational modifications of fodrin and the role of small G-proteins on the cytoskeleton regulation are discussed. In conclusion, this study re- veals that fodrin is formed instead of PA. Measurement of PtdEtOH is highly specific of PLD in comparison with choline measurement. We have previously checked that the inhibitory cytosolic factor provoked a decrease in the formation of both products of PC hydrolysis, PtdEtOH and choline (19). Thus, in this study, to identify the factor responsible for PLD inhibition and to show its effects, we choose to follow PLD activity by measuring its aqueous product, cho- line, as reported previously (21) because of the rapidity of this measurement. Briefly, as already mentioned choline-containing lipids were labeled by culturing cells for 3 days in a medium supplemented with [ methyl - 3 H]choline chloride. After labeling, cells were washed and re- suspended in isotonic buffer A, at 37 °C. 25 (cid:109) l of labeled cells, corresponding to 10 6 cells, were transferred to tubes containing an equal volume of the same buffer supplemented with the permeabilizing agent streptolysin O (0.4 unit/ml final), Mg-ATP (2 m M final), MgCl 2 (2 m M final), Ca 2 (cid:49) buffered with EGTA (3 m M final) at 10 (cid:50) 5 M ( p Ca 5), unless otherwise indicated and GTP (cid:103) S (20 (cid:109) M final). Fractions to be tested for PLD activity were added in 50- (cid:109) l aliquots. After incubation at 37 °C, samples were processed as reported previously (21). Choline present in the reaction was separated from phosphorus-containing choline metab-olites as described by Martin (23). PLD activation 4 (cid:98) -phorbol 12-myristate 13-acetate (PMA) or pervanadate was achieved by a 10-min preincubation of intact cells at 37 °C in the presence of 1 (cid:109) M PMA or 100 (cid:109) M pervanadate as reported previously (19). Cells were then washed, and PLD measurement was performed as described above. was performed to fully deplete fodrin from fractions. After centrifugation of beads, the cytosolic fractions were analyzed for their effect on PLD activity and by immunoblotting for their content in fodrin.

Phospholipase D (PLD) is a major enzyme implicated in important cellular processes such as secretion and proliferation. The knowledge of its regulation is essential to understand the control of these phenomena. Several proteins activating PLD have been described in the last years. In this report, we chromatographed bovine brain cytosolic proteins to identify fodrin, the nonerythroid spectrin, as the first described inhibitor of PLD. A cytosolic fraction with an inhibitory effect on PLD activity loses its capacity after immunoprecipitation of fodrin. Moreover, at 1 nM, purified fodrin blocks fully and quickly PLD activity, whatever the stimuli used. In contrast, fodrin has no effect on adenylate cyclase activity. Fodrin-analogous proteins like dimeric or tetrameric erythroid spectrin have the same inhibitory effect on PLD, at higher concentrations. Other cytoskeletal proteins, actin and vimentin, are inefficient on PLD inhibition. The mechanisms implicated in PLD modulation such as post-translational modifications of fodrin and the role of small G-proteins on the cytoskeleton regulation are discussed. In conclusion, this study reveals that fodrin is involved in the control of PLD activity, suggesting that the cytoskeleton could have an active role in control of secretion and proliferation.
Phospholipase D (PLD) 1 activity was shown to occur in a large number of intact cell types after triggering of receptors by agonists, including hormones, neurotransmitters, and growth factors. The involvement and the importance of PLD activity has been demonstrated in major physiological processes such as proliferation and secretion (reviewed in Refs. [1][2][3]. In mammalian cells, phosphatidylcholine (PC), the most abundant membrane phospholipid, appears to be the main substrate of phospholipase D (reviewed in Ref. 3). Thus, activation of this enzyme leads to the liberation of the headgroup, mainly choline, and of phosphatidic acid (PA). This highly negatively charged phospholipid has been shown to be biologically active (reviewed in Ref. 3), it possesses fusogenic properties (4), and it is involved in DNA synthesis and proliferation of fibroblasts (5,6). It is supposed to be a second messenger in secretory processes in several cell types, including pancreatic islets and adrenal glomerula cells (reviewed in Ref. 1).
PLD in mammalian cells, although not yet shown, seems to be tightly bound to plasma membranes. Due to the loss of enzyme activity during membrane protein solubilization, purification of PLD from mammalian cells has been unsuccessful for almost two decades. Over the past 2 years, proteins with PLD activity were partially purified with mild detergents (7,8), and ARF-stimulated PLD cDNA has been cloned recently (9).
Regulation of PLD activity is not yet fully understood. PLD activation pathways include stimulation by Ca 2ϩ , activation of PLC, and of different kinases (protein kinases C and tyrosine kinases) leading to phosphorylation cascades. Moreover, a positive regulation of PLD activity by G-proteins has been widely studied in different cell types but only indirect demonstration of heterotrimeric G-protein involvement in the enzyme activation has been made so far (reviewed in Ref. 1). In contrast, it has been clearly demonstrated that three small proteins with GTPase activity are stimulators of PLD activity: ARF (10,11), rhoA (12)(13)(14), and p21 ras in v-Src-transformed cells (15,16). Moreover, another cytosolic factor, identified as PKC␣, was recently demonstrated to be involved in the stimulation of PLD activity (17,18).
Thus, in mammalian cells, regulation of this enzyme seems to be under the control of neighboring elements.
The negative regulation of PLD has been reported to be under the control of cellular cytosol factor(s) not yet identified. During the initial step of purification of cytosolic PLD activators, a major increase in the total stimulating PLD activity was observed, suggesting the removal of an inhibitor. In a previous work, we have reported the presence in bovine brain cytosol of a high molecular mass factor, either a protein or a complex of proteins, which negatively regulates PLD activity (19). In this report, we identify this inhibitor as a single protein, fodrin, the non-erythroid spectrin. This protein is an actin-binding protein participating in the cytoskeleton. We also investigate the effect of fodrin on the enzyme activity. ** To whom correspondence should be addressed. 1 The abbreviations used are: PLD, phospholipase D; Pipes, piperazine-N,NЈ-bis(2-ethanesulfonic acid); PC, phosphatidylcholine; PA, phosphatidic acid; ARF, ADP-ribosylation factor; PtdEtOH, phosphatidylethanol; GTP␥S, guanosine 5Ј-3-O-(thio)triphosphate; PMA, 4␤phorbol 12-myristate 13-acetate; PIP 2 , phosphatidylinositol (4,5)-bisphosphate; AEBSF, 4-(2-aminoethyl)benzenesulfonyl fluoride; PAGE, polyacrylamide gel electrophoresis; PLA 2 , phospholipase A 2 .
Specific polyclonal antiserum against pig brain fodrin and pure lens vimentin were kindly given by Louise Anne Pradel, Institut de Biologie Physico-Chimique, Paris, and Karima Djabali, CNRS, Collège de France, Paris, France, respectively.
HL-60 cells given by Dr T. Breitman, NCI, Bethesda, MD, were cultured in suspension in RPMI 1640 medium as described (20). For labeling, medium 199 was chosen because of its low concentration in choline, cells were pelleted and resuspended at 0.5 ϫ 10 6 cells/ml in the medium containing 10% heat-inactivated fetal bovine serum, and 0.5 Ci/ml [methyl-3 H]choline chloride was added. Cells were cultured for 72 h at 37°C in a humidified incubator containing 5% CO 2 to reach isotopic equilibrium (21).

Cell Permeabilization with Streptolysin O and PLD Activity Measurement
After labeling, cells were washed three times in buffer A, pH 6.8, made of 137 mM NaCl, 2.7 mM KCl, and 20 mM Pipes and permeabilized at 4 ϫ 10 7 cells/ml in buffer A with 0.4 unit/ml streptolysin O.
PLD activity can be studied by measurement of both products, choline or PA. PLD possesses a specific transphosphatidylation activity (22) and in presence of ethanol for example, the stable phospholipid phosphatidylethanol (PtdEtOH) is formed instead of PA. Measurement of PtdEtOH is highly specific of PLD in comparison with choline measurement. We have previously checked that the inhibitory cytosolic factor provoked a decrease in the formation of both products of PC hydrolysis, PtdEtOH and choline (19). Thus, in this study, to identify the factor responsible for PLD inhibition and to show its effects, we choose to follow PLD activity by measuring its aqueous product, choline, as reported previously (21) because of the rapidity of this measurement. Briefly, as already mentioned choline-containing lipids were labeled by culturing cells for 3 days in a medium supplemented with [methyl-3 H]choline chloride. After labeling, cells were washed and resuspended in isotonic buffer A, at 37°C. 25 l of labeled cells, corresponding to 10 6 cells, were transferred to tubes containing an equal volume of the same buffer supplemented with the permeabilizing agent streptolysin O (0.4 unit/ml final), Mg-ATP (2 mM final), MgCl 2 (2 mM final), Ca 2ϩ buffered with EGTA (3 mM final) at 10 Ϫ5 M (pCa 5), unless otherwise indicated and GTP␥S (20 M final). Fractions to be tested for PLD activity were added in 50-l aliquots. After incubation at 37°C, samples were processed as reported previously (21). Choline present in the reaction was separated from phosphorus-containing choline metabolites as described by Martin (23).
PLD activation by 4␤-phorbol 12-myristate 13-acetate (PMA) or pervanadate was achieved by a 10-min preincubation of intact cells at 37°C in the presence of 1 M PMA or 100 M pervanadate as reported previously (19). Cells were then washed, and PLD measurement was performed as described above.

Phospholipase A 2 and Phospholipase C Activity Measurements
Phospholipase A 2 activity was measured according to Xing et al. (24). Briefly, after labeling of intact cells with [ 3 H]arachidonate for 1 h at 37°C, phospholipase A 2 activity was estimated by measuring the radioactivity liberated in the supernatant by cells permeabilized with streptolysin O stimulated or not with 20 M GTP␥S, in the presence of pCa 5, MgCl 2 (2 mM), and Mg-ATP (2 mM).
Phospholipase C activity was measured as described previously (25). Incubations were performed under identical conditions to those for PLD measurements, i.e. without addition of LiCl and 2,3diphosphoglycerate.

HL-60 Cell Homogenization and Adenylate Cyclase Activity Measurement
HL-60 cells were washed and resuspended in a hypotonic buffer, pH 7.4, containing 10 mM Tris-HCl, buffered at pH 7.6, containing 1 mM dithiothreitol, 0.1 mM EDTA, 0.1 mM EGTA, and as protease inhibitors, 1000 IU/ml aprotinin, 1 g/ml pepstatin, 1 g/ml leupeptin, 1 M phenylmethylsulfonyl fluoride, 4 mM benzamidine, and 100 M N-tosyl-L-phenylalanine chloromethane. After 5 min on ice, cells were homogenized in a dounce (20 strokes) and centrifuged for 5 min at 800 ϫ g to remove nuclei and unbroken cells. Adenylate cyclase activity was measured according to Chneiweiss et al. (26). Fodrin effect was tested on (i) basal activity, (ii) in the presence of Mn 2ϩ , an activator of the catalytic unit of the enzyme (27), and (iii) in the presence of isoproterenol, a selective agonist of ␤-adrenergic receptor coupled to the enzyme through a Gs-protein.

Preparation of Bovine Brain Cytosol
All steps were carried out at 4°C. Gray tissue of one bovine brain (200 g) was cut into small pieces and put in 1 liter of a buffer consisting of 10 mM Tris-HCl, buffered at pH 7.6, containing 1 mM dithiothreitol, 0.1 mM EDTA, 0.1 mM EGTA, 2% glycerol, protease inhibitors as above plus 1 g/ml diisopropyl fluorophosphate (buffer B). Tissue was then homogenized in a Waring blender (10 ϫ 1 min). The homogenate was centrifuged for 14 h at 16,000 ϫ g, and fractionation was undertaken. After centrifugation, cytosolic proteins in the supernatant were precipitated for 30 min with 80% (NH4) 2 SO 4 by addition of solid salt and collected by centrifugation at 15,000 ϫ g for 30 min. Precipitated proteins were dissolved in 100 ml of buffer B.
Chromatography on DEAE-Sepharose Fast Flow-After extensive dialysis against buffer B, soluble proteins (100 ml) were applied on column of DEAE-Sepharose fast flow (5 ϫ 26 cm), an anion exchange gel equilibrated with buffer B. After extensive washing, retained proteins were eluted at 0.8 ml/min with a linear gradient of 1 liter of buffer A versus 1 liter of the same buffer containing 1 M NaCl. Fractions of 20 ml were collected.
Chromatography on Immobilized Heparin-The peak containing the PLD inhibitory activity was pooled and dialyzed against buffer B and loaded onto a heparin-agarose column (2.5 ϫ 20 cm) equilibrated with the same buffer. After column washing to remove unretained material, proteins with affinity for heparin were eluted at 30 ml/h with a linear NaCl gradient (0 -1 M, 1 liter), 8-ml fractions were collected.
Chromatography on a Hi-Trap Q Column-The active fractions were pooled and dialyzed against the same buffer as for DEAE. Proteins of the supernatant were chromatographed on a 5 ml Hi-Trap Q column (Pharmacia Biotech Inc.), an anion exchange column, equilibrated with the same buffer, and eluted at 1 ml/min with a linear NaCl gradient (0 -500 mM, 200 ml). Fractions of 2.5 ml were collected.
Chromatography on a Phenyl-Superose Column-Fractions containing the activity were pooled, dialyzed against buffer B, except that glycerol was omitted, adjusted to 0.8 M (NH4) 2 SO 4 with 4 M concentrated solution, and applied to a phenyl-Superose high performance column (1.6 ϫ 8 cm) equilibrated in a Tris buffer with the same final salt concentration as the sample. After washing, retained proteins were eluted with a decreasing linear salt gradient, and 2.5-ml fractions were collected and desalted on PD 10 column (Pharmacia).
Gel Filtration Chromatography-The peak containing PLD inhibitory activity after chromatography on phenyl-Superose was eluted in the absence of ammonium sulfate salt. The active peak was pooled, concentrated, and dialyzed against a 10 mM Tris-HCl buffer, pH 7.6, and 0.5 ml was loaded on a Superdex 200 column (1.6 ϫ 30 cm, Pharmacia) and eluted at a flow rate of 0.3 ml. Fractions of 0.5 ml were collected.
After each step of chromatography, each fraction was tested for adsorbance at 280 nm, and 50 l of each eluted fraction was tested for its ability to inhibit PLD activity.
The rapidity in the inhibitor purification was essential as we observed that this factor was not stable although a large range of protease inhibitors were always present in buffers.

Preparation of Brain Fodrin
Fodrin was extracted from bovine brain membranes at low ionic strength with dithiothreitol (1 mM) and purified according to the molecular mass on a gel filtration column (Bio-Gel A-5m 100 -200 mesh) followed by chromatography on a weak anion exchange column (DEAE-Sepharose) according to Davis and Bennett (28).

Fodrin Treatment with Phospholipid Vesicles
Vesicles containing various concentrations of PC with or without 125 M PIP 2 were made according to Brown and Sternweis (29). Vesicle aliquots of 25 l were incubated with 25 l of either buffer, pH 7.6, made of 20 mM Tris and protease inhibitors as above or fodrin (10 nM in the same buffer) in the presence of MgCl 2 (2 mM) Mg-ATP (2 mM) and Ca 2ϩ (10 M). After 30 min at room temperature, PLD activity was measured in the absence and in the presence of the vesicles with or without fodrin in permeabilized cells as described above.

Protein Estimation
Protein content in the different fractions from bovine brain cytosol was estimated by light absorbance at 280 nm for column elution and using the technique described by Bradford (30) when a precise amount of pure proteins was required.
Fraction Analysis and Immunoblotting 30 l of each fraction eluted after chromatography were mixed with 15 l of SDS-PAGE sample buffer (three times) and boiled for 2 min. Proteins were separated by SDS-PAGE (31). Proteins were transferred onto nitrocellulose membranes by semidry Western blotting according to Burnette (32) except that the second antibody was peroxidase-linked anti-rabbit IgG and the blot was developed using ECL reagents.

Immunoprecipitation Assay
Cytosolic fractions (1 ml) from DEAE chromatography containing PLD inhibitory activity were precleared with protein A-Sepharose beads and then incubated at 4°C for 2 h, on a rotating wheel, with affinity-purified fodrin antibodies (5 g/1.4 mg of proteins) or with the same amount of rabbit Ig. The incubate was then adsorbed onto protein A-Sepharose beads. A sequential run of immunoprecipitation was performed to fully deplete fodrin from fractions. After centrifugation of beads, the cytosolic fractions were analyzed for their effect on PLD activity and by immunoblotting for their content in fodrin.

RESULTS
Identification of the Inhibitor of PLD Activity-We took advantage of cell treatment with streptolysin O, which produces major holes in the plasma membrane, to isolate and identify a cytosolic inhibitor for PLD. To detect any inhibition, PLD activity was measured in whole permeabilized cells, i.e. in the presence of their cytosol which contains PLD activators.
Purification of the inhibitor was attempted by precipitation of cytosolic proteins by 80% ammonium sulfate followed by chromatography on a weak anion exchange resin, DEAE-Sepharose fast flow. One peak with inhibitory activity was eluted and corresponding fractions were pooled, concentrated, and further purified on immobilized heparin followed by chromatography on a strong anion resine, Hi-Trap Q, and on phenyl-Superose. Separation of proteins according to their molecular weight was performed by gel filtration on a Superdex 200 column. Fractions containing the inhibitory activity for PLD were eluted in one peak in the void volume, indicating a mass over 700 kDa for the inhibitory factor. Protein analysis by SDS-PAGE showed that fractions with inhibitory activity contained several proteins with molecular masses ranging from 40 to 260 kDa. However, only high molecular mass proteins, particularly a doublet at around 250 kDa and a band at 150 kDa, fitted with the peak of inhibition (Fig. 1, A and B). We have previously eliminated that this factor was the adaptin or the COP complex as none of these proteins were found in the inhibitory fraction (19). Considering that PLD is located at the plasma membrane level, near the cytoskeleton, we suspected that the inhibitor could belong to this cellular structure and that the doublet could be the non-erythroid spectrin, fodrin, which shares these characteristics. Using specific antibodies against fodrin, it was confirmed by immunoblot analysis that the doublet in fractions containing the inhibitory effect on PLD activity corresponded to ␣ and ␤ chains of fodrin. A band at 150 kDa was also detected by polyclonal specific antibodies to fodrin (Fig. 1C). Fodrin is easily cleaved during the preparation, giving a product of about 150 kDa (28). Fodrin was absent in fractions without inhibitory effect.
To confirm the involvement of this protein in PLD regulation, immunoprecipitation using specific fodrin antibodies was performed. As shown in Fig. 2a, fodrin was not detected by immunoblotting in the specifically immunoprecipitated fraction which has lost its ability to inhibit PLD activity (Fig. 2b).
Purification of Fodrin-Fodrin responsibility in PLD inhibition was also confirmed using purified protein extracted from bovine brain membranes according to Davis and Bennett (28). Briefly, membranes from bovine brain were prepared by homogenizing the brain in a buffer at low ionic strength at 4°C. The cytosol was then removed, and membranes were washed several times at low ionic strength (4°C). This procedure reduced the efficiency of proteases. Fodrin was then extracted from membranes at 37°C for 1 h in the same hypotonic buffer in the presence of a reducing agent, dithiothreitol. The extracted fodrin was purified under its tetrameric form with an apparent molecular mass of approximately 1,000 kDa on gel filtration. Fig. 3, A and B, shows that the peak of protein with FIG. 1. Partial purification of PLD inhibitory proteins and identification of fodrin in the partially purified cytosolic fraction. After four different steps of chromatography (DEAE-Sepharose, heparin-agarose, Hi-trap Q, and phenyl-Superose) as reported under "Methods," and a 0.5-ml fraction with inhibitory activity for PLD was further purified by gel filtration chromatography on a Superdex 200 column. 0.5-ml fractions were collected, and one aliquot of each fraction was tested for its effect on PLD activity (q) and absorbance at 280 nm (----) (A); B, fractions containing the peak with inhibitory activity were analyzed by SDS-PAGE (7.5%); C, electrophoresed proteins were also transferred onto a membrane of nitrocellulose and tested by immunoblotting for their content in brain fodrin using specific antibodies. Arrows on the left indicate fodrin doublet and fodrin degradation product.
Fodrin Inhibits PLD Activity at Nanomolar Concentrations-The inhibitory effect of pure fodrin on PLD activity was studied on HL-60 permeabilized cells either in the absence or in the presence of 20 M GTP␥S. The level of activation of PLD is mainly dependent on the number of cells present. This number determines the amount of PLD itself and of cytosolic and membrane PLD activators, such as the different small G-proteins present in the assay. To study the relative amount of fodrin necessary to inhibit PLD activity, the experiments were performed using regularly 0.5 ϫ 10 6 cells/assay in the presence of different concentrations of pure fodrin. Fig. 4 reports the doseresponse curve of PLD inhibition. As fodrin is purified as a tetramer (28), the concentration of the protein able to inhibit half of the GTP␥S-stimulated PLD activity was estimated to be equivalent to 0.8 nM. PLD activity was fully inhibited by fodrin at concentrations higher than 1 nM. In the absence of GTP␥S, PLD was only slightly active; however, the same concentrations of fodrin also inhibit, to a small extent, basal PLD activity (Fig. 4). After boiling, fodrin at 10 nM has lost all its inhibitory effect on PLD.
Effect of Preincubation of Fodrin with PLD Substrate (PC) and/or PLD Cofactor (PIP 2 )-We tested the hypothesis that fodrin may prevent the access of PLD to either its substrate (PC) or its cofactor (PIP 2 ). As shown in Fig. 5A, preincubation of fodrin with vesicles made of PC are able to prevent PLD inhibition by fodrin only at concentrations of PC in the millimolar range, whereas PC vesicles in the absence of fodrin have no effect on choline release. Other authors have reported that the presence of PIP 2, in phospholipid vesicles, is necessary for the enzyme activity in vitro. Fodrin ␤ chain possesses a PH domain known to bind to PIP 2 . The hypothesis that fodrin inhibits PLD activity by binding PIP 2 and, thus, would sequester the cofoactor for PLD activation was investigated by preincubating fodrin with vesicles containing various concentrations of PC and a constant concentration of PIP 2 (125 M). As reported in Fig. 5B, such vesicles have no effect on choline release (control). Only millimolar concentrations of phospholipids inhibit fodrin effect on phospholipase D activity.
Doses of phosphatidylcholine able to overcome fodrin effect on phospholipase D represent about 1000-fold the amount present within the cells during the reaction as estimated by measurement of phosphate present in cell aliquots and in vesicles according to Galliard et al. (33). Thus it seems unlikely in cellular experiments that fodrin acts on PLD by substrate or cofactor depletion.

Effect of Fodrin on Other Enzymes Involved in Signal Transduction: Phospholipase A 2 , Phospholipase C, and Adenylate
Cyclase-To check the specificity of fodrin on phospholipase D activity, we looked at the possible effect of this protein on other enzymes involved in signal transduction. Other phospholipases were chosen as they translocate toward the membrane after A and B) in Fig. 2b.   FIG. 3. Purification of non-erythroid fodrin and identification of its inhibitory effect on PLD activity. Membranes from bovine brain were prepared and fodrin was then extracted at low ionic strength. Fodrin was purified as reported (Davis and Bennett (28)) by two chromatography steps, by size exclusion on a Bio-Gel A-5m, followed by a weak anion exchange filtration (DEAE-Sepharose). Fractions eluted from the DEAE-Sepharose column were analyzed for PLD activity (A), protein (6% SDS-PAGE) (B), and fodrin (C) content as in Fig. 1. cell activation. Adenylate cyclase was chosen due to its location in the plasma membrane and to its role in signal transduction via G-proteins. Table I shows that basal, to a small extent, and mostly GTP␥S-stimulated PLA 2 and PLC activities are inhibited in the presence of 3 nM fodrin: 90 and 70%, respectively. In contrast, basal and Mn 2ϩ -stimulated enzyme activity in HL-60 homogenates are not modified in the presence of fodrin at a concentration maximally inhibiting PLD activity. In HL-60 cells, adenylate cyclase was stimulated to a small extent (42 Ϯ 4.7%) by isoproterenol, a ␤-adrenergic receptor agonist. No inhibition of the receptor-coupled adenylate cyclase stimulation was observed in the presence of fodrin (data not shown).

FIG. 2. Depletion of PLD inhibitory activity with fodrin antibodies. A cytosolic fraction containing PLD inhibitory activity, obtained after DEAE chromatography, was treated with affinity-purified fodrin antibodies (lane B) or with control rabbit Ig (lane A). The content of fodrin in both cytosolic fractions was visualized by immunoblotting in a. PLD activity was tested in the absence of cytosolic fraction (Control) and in the presence of cytosolic fractions (columns
Action of Other Cytoskeletal Proteins-We also tested the effect of two other proteins belonging to the cytoskeleton: vimentin and actin under its G and F forms on PLD activity. None of these proteins was able to inhibit PLD activity (Table  II) at concentrations much higher than fodrin concentrations giving a PLD inhibition.
Erythroid Spectrin Is Also Able to Inhibit PLD Activity-Fodrin belongs to the family of spectrin proteins and consists of two long flexible chains associated side to side in an antiparallel way to form a heterodimer. Two dimers are associated head to head in a tetramer. To check the specificity of fodrin, we investigated the effect of human erythrocyte spectrin under its tetrameric form. Fig. 4 shows that its inhibitory effect on the enzyme activity is similar to fodrin; however, human erythrocyte spectrin is less efficient than fodrin from bovine brain. 50% inhibition of PLD activity is obtained with 30 nM of spectrin versus 0.8 nM for fodrin. Using erythroid spectrin, we investigated whether a difference in activity could be observed between spectrin under its tetrameric and its dimeric form. Dimeric spectrin was prepared by extraction at 37°C. Same results were found with both forms of spectrin (data not shown).
PLD Inhibition by Fodrin Is Stable-As reported in Fig. 6, fodrin at 3 nM leads to a complete inhibition of PLD activity, which can be observed already after 2.5 min (the earliest period studied) and remains stable for at least 30 min (the longest period studied). Thus, in the presence of GTP␥S, which activates G-proteins, PLD can be maintained in an inactive state when fodrin is present at a sufficient concentration.
Fodrin Inhibits Different Pathways of PLD Activation-Ca 2ϩ in the micromolar range stimulates, to a small extent, the basal level of PLD in permeabilized HL-60 cells. In the presence of GTP␥S, this activity is highly stimulated at Ca 2ϩ concentrations from 10 Ϫ6 to 10 Ϫ5 M. Fodrin at 1 nM has no effect on the basal level stimulated by 10 Ϫ5 M Ca 2ϩ , whereas it has a marked inhibitory effect on the GTP␥S-stimulated PLD activity, which is almost totally abolished (Fig. 7A).
It has been shown that Mg-ATP is not necessary for PLD activation, but it amplifies the GTP␥S response (34). We studied the effect of fodrin on the PLD activity stimulated with different concentrations of this nucleotide in the presence or in the absence of GTP␥S. As reported in Fig. 7B, the highest concentration of Mg-ATP studied (3 mM) stimulates the basal level of PLD at 10 Ϫ5 M Ca 2ϩ . Fodrin, at 1 nM, inhibits this stimulation. Increasing the Mg-ATP level leads to a major GTP␥S-stimulated enzyme activity. The stimulated PLD activity by GTP␥S and/or Mg-ATP is fully inhibited by fodrin.
PLD is also stimulated via PKC or tyrosine kinase activation. We investigated the effect of purified fodrin on both pathways stimulated by PMA and pervanadate, respectively. As reported in Fig. 8, fodrin at 3 nM, which is inhibitory on Ca 2ϩand Mg-ATP-stimulated PLD activity. Moreover, this cytoskeletal protein inhibits, to a similar extent, PMA-and pervanadate-stimulated PLD activity.
Thus, fodrin appears to inhibit PLD activated by different stimuli and is likely to maintain the enzyme under an inactive state.
Effect of Fodrin Hydrolyzed in Vitro by Calpain 1 and Calmodulin on PLD Activity-It is known that Ca 2ϩ /calmodulin binds to fodrin and allows its proteolysis by calpain 1. The hydrolysis of fodrin ␣ chain by calpain 1 occurs only in the presence of Ca 2ϩ (35), whereas the ␤ chain hydrolysis requires both Ca 2ϩ and calmodulin (36). Thus, fodrin was submitted to proteolysis by incubation at 30°C for 30 min either in the presence of EGTA, Ca 2ϩ , calpain, and calmodulin or with different combinations of these components. The inhibitory effect of the intact or proteolyzed protein was then tested on PLD activity fully stimulated by GTP␥S, Mg-ATP, and Ca 2ϩ . No difference in the inhibition of PLD activity was observed between intact fodrin and this protein cut by calpain 1 with or without calmodulin (data not shown). Therefore, it is unlikely that either lysis by calpain 1 or calmodulin in the presence of Ca 2ϩ would be a way for removing inhibition by fodrin in cells.

DISCUSSION
In this study, we identify fodrin, the non-erythroid spectrin, as the first described natural inhibitor of phospholipase D. This cytoskeletal protein, which is mostly associated with plasma membrane in resting cells, is likely to play an active role in the control of a phospholipase activity involved in signal transduction.
Fodrin is a high molecular mass protein, purified as a dimer with a molecular mass of about 1000 kDa. We wondered whether the inhibitory effect detected at the earliest time of PLD measurement could be the consequence of a nonspecific steric occupancy related to the size of the protein. This hypothesis can be ruled out for the following reasons: (i) the concentration of fodrin provoking the maximal inhibition is very low, in the nanomolar range; (ii) erythroid spectrin, which shares the same molecular mass, also inhibits PLD activity but at a higher concentration (30-fold), indicating the higher specificity of fodrin in non-erythroid cells; (iii) after complete cleavage of both chains of fodrin in the presence of calpain 1, Ca 2ϩ , and calmodulin, proteolyzed proteins are still inhibitors for PLD activity (data not shown); (iv) the activity of adenylate cyclase, another enzyme involved in signal transduction and located at the plasma membrane, was not modified in the presence of fodrin in conditions which almost completely inhibit PLD activity. Thus, fodrin does not appear to inhibit PLD unspecifically by steric hindrance due to its large size. Complex interrelationships exist between the three classes of phospholipases. PA, the product of PLD activation, is stimulatory for PLA 2 (37) and arachidonate products, including lysophosphatidic acid (reviewed in Ref. 38), leukotriene B 4 , prostaglandin F 2␣ , and 12-hydroxyeicosatetraenoic acid (39 -41) stimulate PLD. This enzyme activity requires PIP 2 , the PLC substrate (42), and is stimulated by diacylglycerol, the product of PLC activity (reviewed in Ref. 1). In the present work we report that the inhibitory effect of fodrin is not exclusive to PLD activity but affects also PLA 2 and PLC activities. This observation might be   explained by one or both of the two following mechanisms. 1) Fodrin might sequester phospholipids necessary for different enzyme activities. In favor of this hypothesis is the observation that all activation pathways studied are inhibited by this protein. However, our results (Fig. 5) appear to exclude the hypothesis of substrate or cofactor depletion. 2) Fodrin may act by preventing phospholipases translocation to the membrane, maintaining the enzymes under an inactive state.
Fodrin with other cytoskeletal proteins, forms a two-dimensional network at the cytoplasmic surface of plasma membranes to which it gives a structural support, a mechanical stability and determines the shape of the cells. It is also re-sponsible for the flexibility and elasticity of the cells, necessary to achieve different physiological processes including secretion and division. Moreover, fodrin seems to play a role in specialized cell domains, in the regulation of vesicle-membrane interactions and in formation of cell-cell junctions (reviewed in Refs. [43][44][45]. How this protein modulates PLD activity is not yet understood. Different mechanisms of action can be hypothesized.
Our study indicates that fodrin maintains PLD under an inactive state. To restore an active enzyme, post-translational modifications of the cytoskeletal protein and/or of phospholipase D itself, including proteolysis and/or phosphorylations are probably necessary.
Whether one or several domains of fodrin are involved in the regulation of phospholipase D is still unknown.
Fodrin is a rod-shaped protein composed of several domains. The complete sequences of non-erythroid ␣ and ␤ chains are known. Both chains of erythroid and non-erythroid spectrin are mainly comprised of tandem homologous 106 residue units, flanked by nonhomologous NH 2 -and COOH-terminal sequences. These nonhomologous areas contain particular domains or modules such as (i) the actin binding domain at the NH 2 -terminal end of ␤ chain; (ii) a PH domain (pleckstrin homologous) present only in the COOH end of non-erythroid spectrin; (iii) two EF hand domains (calcium-binding domain) at the COOH-terminal end of ␣ chain; and (iv) a SH 3 domain (Src homologous). PH and SH 3 domains have been found in numerous proteins, most of them involved in signal transduction (46). In addition, it has been demonstrated recently that fodrin SH 3 domain binds to amiloride-sensitive Na ϩ channel in epithelial cells (47), suggesting a potential effect of fodrin in maintaining proteins in specific membrane localization. It seems unlikely that the PH domain of fodrin would be the main chain sequence involved in PLD regulation, first because of the lack of modification of the fodrin inhibitory effect after incubation with PIP 2 and second because of our observation that erythrocyte spectrin, which does not possess a PH domain, also exerts an inhibitory effect on PLD activity. We are currently investigating which specific domain could be responsible for PLD regulation.
Thus, fodrin appears to be a potent inhibitor of PLD activity. Over the past 2 years, several factors were shown to be involved in the control of PLD activity. Several small G-proteins have been reported to activate PLD, including ARF (10,11), rhoA in different cell types and p21 ras (16).
The first demonstration of rhoA involvement in the activation of PLD was indirect through the observation of rhoGDI acting as an inhibitor of PLD (12). Activation of the enzyme with pure recombinant rhoA was then shown (13). This small G-protein also leads to phosphatidylinositol-4-phosphate 5-kinase activation (48), provoking an increase in phosphatidylinositol 4,5-bisphosphate, a cofactor for PLD activation (42). Thus, rhoA may act on PLD by its effect on the local ratio of phospholipids. Moreover, rhoA plays an important role in the regulation of actin polymerization and stress fibers (49,50). Its effect on the actin network has also been demonstrated in the secretion process (51).
In the present paper, we demonstrate that fodrin, a cytoskeletal protein, is an important inhibitor of PLD activity. This actin-binding protein has also been shown to be involved in secretion through modifications occurring at the level of actinfodrin interaction (52)(53)(54). Moreover, we have observed that cytochalasin D treatment of cells, which leads to actin depolymerization, increases PLD activity (data not shown). The enzyme appears to be under the control of proteins both involved in the regulation of actin polymerization and bound to F-actin. Thus, the state of the actin network might be important for PLD activity. If this is the case, the actin-binding domain of fodrin is likely to be important in PLD regulation.
ARF-activated PLD has been cloned recently. The recombinant protein was expressed in baculovirus where it is cytosolic (9). Regulation of PLD activity in relationship with the cytoskeleton might reflect the involvement of this cellular structure in the enzyme activity control.
At the light of the present knowledge, it appears that the enzyme regulation is due to complex relationships between PLD itself, fodrin, and cytosolic proteins such as small Gproteins. A new concept for the understanding of PLD regulation could be proposed: PLD would be normally maintained inactive in resting cells by fodrin. After cell activation, changes at the level of several cytoskeletal or related proteins would 1) modify the inhibitory effect of fodrin on PLD by removing either a direct effect of fodrin on the enzyme or its access to the plasma membrane, 2) allow the generation of PIP 2 , the necessary cofactor of PLD activity, 3) lead to the access of PLD to its substrate and/or to a firm association of the enzyme with the plasma membrane via a local reorganization of the cytoskeleton.