Inhibition of cytosolic phospholipase A2 by annexin V in differentiated permeabilized HL-60 cells. Evidence of crucial importance of domain I type II Ca2+-binding site in the mechanism of inhibition.

Annexin V belongs to a family of proteins that interact with phospholipids in a Ca2+-dependent manner. This protein has been demonstrated to have anti-phospholipase A2 activity. However, this effect has never yet been reported with the 85-kDa cytosolic PLA2 (cPLA2). We studied, in a model of differentiated and streptolysin O-permeabilized HL-60 cells, the effect of annexin V on cPLA2 activity after stimulation by calcium, GTPgammaS (guanosine 5'-O-(3-thiotriphosphate)), formyl-Met-Leu-Phe, or phorbol 12-myristate 13-acetate. Both recombinant and human placental purified annexin V inhibit cPLA2 activity whatever the stimulus used. The decrease of arachidonic acid release is of 40 and 50%, respectively, at [Ca2+] of 3 and 10 microM. The mechanism of inhibition was also analyzed. cPLA2 requires calcium and protein kinase C (PKC) or mitogen-activated protein kinase phosphorylation for its activation. As annexin V was shown to be an endogenous inhibitor of PKC, PKC-stimulated cPLA2 activity was analyzed. Using GF109203x, a specific PKC inhibitor, we demonstrated that this pathway is of minor importance in our model. cPLA2 inhibition by annexin V is not linked to PKC inhibition. To test the hypothesis of phospholipid depletion, mutants of annexin V were constructed using mutagenesis directed to Ca2+ site. We demonstrate that the Ca2+ site located in domain I is necessary for the inhibitory effect of annexin V on cPLA2 activity. The site in domain IV is also involved but with less efficiency. In contrast, mutations in site II and III do not modify this effect. Moreover, annexin V mutated on all sites does not inhibit cPLA2. Thus, we propose a predominant role of module (I/IV) in the biological action of annexin V, which, in physiological conditions, may control cPLA2 activity by depletion of the phospholipid substrate.

Annexins constitute a family of at least 13 structurally related cytoplasmic proteins in mammals, which bind to anionic phospholipids in a Ca 2ϩ -dependent manner. These proteins are widely distributed in eukaryotic cells and form a class of Ca 2ϩbinding proteins distinct from the "EF-hand" family and from other Ca 2ϩ -binding proteins containing the calcium and lipid binding domain such as conventional protein kinase C (cPKC) 1 and cytosolic phospholipase A 2 (cPLA 2 ) (1, 2). Each annexin consists of two different regions: (i) the N-terminal domain, diverse in sequence and length between the members of the family, which is likely to confer specific biological functions for each of them; and (ii) the C-terminal domain, named "core," and composed of four repeats (eight for annexin VI) of highly conserved 70 -80 amino acid sequence. The Ca 2ϩ -dependent membrane association of annexins relies in specific Ca 2ϩ -binding sites present in this common conserved core region (3). Although annexins have been implicated in many aspects of cell biology, their exact physiological functions are not yet known. However, it seems likely that their cellular and subcellular locations are related to their biological functions (2).
Annexin V is the most abundant of these proteins. Its cDNA encodes an unglycosylated protein of 320 residues with a molecular mass of 35,935 Da (4). Crystal structure of annexin V has been solved and serves as a structural prototype for the annexin family (5)(6)(7). The three-dimensional structure of annexin V has a slightly curved shape with four type II, high affinity Ca 2ϩ -binding sites located on the convex face, facing the membrane. Both N and C termini are on the opposite face (7,8). Moreover, the four repeats form four domains (I-IV), which are packed pairwise into two modules, (I/IV) and (II/III) (7,8). So far, several physiological functions of annexin V have been reported. This protein has been implicated in the control of blood coagulation (4). It was shown to form voltage-dependent Ca 2ϩ channels (9) and to be a high affinity inhibitor of protein kinase C (10,11). Moreover, it can regulate inflammation by inhibiting phospholipase A 2 activity (12)(13)(14)(15). However, neither the mechanism of PLA 2 inhibition by annexin V is fully understood, nor the isoform of PLA 2 involved in this effect defined.
Phospholipase A 2 enzymes catalyze the hydrolysis of the sn-2 fatty acyl chain of many different phospholipids to generate free fatty acids and lysophospholipids. Moreover, PLA 2 -catalyzed release of arachidonic acid (AA) is believed to be the rate-limiting event in the generation of proinflammatory mediators. PLA 2 enzymes can be separated into two main classes * 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.
¶ To whom correspondence should be addressed. Tel.: 33-01-40-51-64-42; Fax: 33-01-40-51-77-49; E-mail: russo@icgm.cochin.inserm.fr. 1 The abbreviations used are: cPKC, conventional protein kinase C; PLA 2 , phospholipase A 2 ; cPLA 2 , group IV cytosolic phospholipase A 2 ; sPLA 2 , secretory phospholipase A 2 ; iPLA 2 , calcium-independent phospholipase A 2 ; PLD, phospholipase D; PLC, phosphoinositide-specific phospholipase C; AA, arachidonic acid; MAP, mitogen-activated protein; PKC, protein kinase C; PMA, phorbol 12-myristate 13-acetate; SLO, streptolysin O; fMLP, N-formyl-methionyl-L-leucyl-L-phenylalanine; GTP␥S, guanosine 5Ј-O- (3- (16). (i) The extracellular forms or 14-kDa secretory PLA 2 s (sPLA 2 s) that have been extensively characterized and structurally defined (17). Major features of sPLA 2 s include their requirement of millimolar [Ca 2ϩ ] and their sensitivity to reducing agents. The crystal structure of the Ca 2ϩ -binding loop of extracellular PLA 2 is similar to that found in annexin V (5). In vitro studies, performed at [Ca 2ϩ ] o , i.e. in the millimolar range, have indicated that annexins sequester the phospholipid substrate of sPLA 2 s thus preventing the enzyme access to its substrate (12,18). (ii) The intracellular class of PLA 2 is composed of a Ca 2ϩ -independent PLA 2 (iPLA 2 ) involved essentially in regulating the incorporation of AA into membrane phospholipids (19,20), and of the group IV cytosolic 85-kDa Ca 2ϩ -dependent PLA 2 (cPLA 2 ) responsible of intracellular AA release (16,20,21). cPLA 2 is activated by both micromolar [Ca 2ϩ ] i and Ser-505 phosphorylation by MAP kinases or PKC (21). Recently, an effect of annexin I on cPLA 2 activity was reported: annexin I would inhibit cPLA 2 activity by specific interaction between both proteins (15). This mechanism was also suggested by another group using an annexin I-derived peptide (22). Relationships between cPLA 2 and the other members of annexin family have never been under investigation. As annexin V and cPLA 2 are both ubiquitous and activated by similar [Ca 2ϩ ] i and as annexin V inhibits cPKCs, which could phosphorylate and activate cPLA 2 , we studied the effect of annexin V on AA release due to cPLA 2 activation in differentiated and permeabilized HL 60 cells and found that, in physiological conditions, annexin V was inhibitory for cPLA 2 activity. The mechanism of such inhibition was further investigated using Ca 2ϩ -binding site-directed mutagenesis of annexin V and specific enzyme inhibitors.
Culture and Differentiation of HL-60 Cells-HL-60 cells were cultured and differentiated into a neutrophil-like phenotype, as described previously (23). Briefly, cells were routinely grown in RPMI 1640 supplemented with 25 mM HEPES, 10% fetal bovine serum, glutamine 2 mM, 100 units/ml penicillin, and 100 g/ml streptomycin (pH 7.4) under a 95% air and 5% CO 2 atmosphere. Before differentiation, cells were adjusted to approximately 0.5 ϫ 10 6 cells/ml and grown for 24 h in the same medium to a density of about 0.8 ϫ 10 6 cells/ml. Differentiation into a neutrophil-like phenotype was then induced by culturing the cells for 72 h in the presence of 0.5 mM dibutyryl cyclic AMP.
Labeling of Cellular Phospholipids with [ 3 H]Arachidonic Acid (AA)-Differentiated HL-60 cells were counted and viability was monitored by trypan blue exclusion. It was always over 95%. Cells were harvested by centrifugation for 5 min at 1200 rpm at room temperature and washed three times in buffer A, made of 137 mM NaCl, 2.7 mM KCl, 20 mM HEPES, 1 mM CaCl 2 , 1 mM MgCl 2 , and 5.6 mM glucose, pH 7.2. Cells were then resuspended at a concentration of 2 ϫ 10 7 cells/ml in buffer A and labeled at 37°C for 1 h with 1 Ci/ml [ 3 H]arachidonic acid. Under these conditions, more than 70% of total radioactive molecules were incorporated into cellular phospholipids. At the end of labeling, cells were washed three times and resuspended at a concentration of 2 ϫ 10 7 cells/ml in buffer B containing 137 mM NaCl, 2.7 mM KCl, 20 mM PIPES, and 0.2% protease and fatty acid-free bovine serum albumin, pH 6.8. Typical value range for incorporated labeled arachidonate was 0.8 -1 ϫ 10 6 dpm/assay tube (10 6 cells).
PLA 2 Assays in Permeabilized and Differentiated HL-60 Cells-Labeled cells were equilibrated for 10 min at 37°C. Cytochalasin B (2 g/ml final) was added to the cells 2 min before transferring cells to assay tube. Assays were started by adding 50 l of the packed labeled cell suspension (2 ϫ 10 7 cells/ml) to the same volume of medium (prewarmed at 37°C) containing permeabilizing agent SLO and stimuli when indicated as described by Stuchfield and Cockcroft (24). Briefly, cells were incubated at 37°C with SLO (0.4 IU/ml), in the presence of 2 mM Mg-ATP, 2 mM MgCl 2 , and Ca 2ϩ at various concentrations, in the presence or absence of various stimuli such as GTP␥S or DTT. Incubations were carried out at 37°C for 10 min, under constant shaking. Preliminary experiments studying different incubation times (0 -20 min) indicated that maximal arachidonate release occurred at 10 min, confirming previous studies (23).
Reactions were stopped by 5-fold dilution with ice-cold 50 mM Tris/ HCl, 100 mM KCl, 5 mM EGTA and 5 mM EDTA, pH 7.5. After centrifugation at 4°C for 10 min at 1000 ϫ g, aliquots of the supernatants (300 l) were taken and associated radioactivity was quantified in a Beckman scintillation counter. Arachidonic acid released at the end of cellular reaction was measured by the amount of tritiated AA present in the supernatant and expressed as a percentage of total incorporated cellular radioactivity.
In experiments where PKC was stimulated by phorbol ester, 100 nM phorbol 12-myristate 13-acetate (PMA) was added during the 10 min preincubation period. Formyl-Met-Leu-Phe (fMLP) stimulation was started two minutes before permeabilization of cells.
Cell treatment with the different inhibitors (AACOCF3, MAFP, BEL, or GF109203x) was achieved by a 30-min preincubation at 37°C of labeled intact cells.
Calcium Buffers-Calcium buffers were prepared as described by Tatham and Gomperts from solutions of Ca-EGTA and EGTA of identical concentrations, at pH 6.8, which are combined in varying proportions (25,26). In the reaction, calcium buffers were used at 3 mM EGTA final, giving Ca 2ϩ concentrations ranging from 10 Ϫ8 to 10 Ϫ5 M (pCa 8 to pCa 5). At pCa 5 and 8, the final ratio of Ca[tot] to EGTA[tot] was 0.938 and O.014, respectively. Free Mg 2ϩ was maintained at 2 mM.
Phospholipase D and Phospholipase C Activity Measurements-Cells were labeled during the 72-h period of differentiation with 0.5 Ci/ml [methyl-3 H]choline chloride. Phospholipase D (PLD) activity was measured as described previously (27). Briefly, labeled cells were washed and resuspended in isotonic buffer C, containing 137 mM NaCl, 2.7 mM KCl, 20 mM PIPES, 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 SLO (O.4 units/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), and GTP␥S (10 M final). Fractions containing annexin V to be tested for PLD activity were added in 50-l aliquots. After incubation at 37°C, samples were proceeded as reported previously (27). Choline present in the reaction was separated from phosphorus-containing choline metabolites as described by T. W. Martin (28).
Phospholipase C activity was measured as described previously (29). Site-specific Mutagenesis-Mutants of annexin V were prepared as followed. Restriction digests, cloning, and isolation of the DNA fragments were performed according to standard procedures (30). The BamHI/HindIII fragment of the vector pGEX-2T containing the complete annexin V cDNA was cloned into the corresponding site of the Bluescript II KSϩ vector. The desired mutation was introduced by oligonucleotide-directed mutagenesis using the procedure of Kunkel et al. (31).
Initial structure analysis of annexin V revealed three high affinity calcium-binding sites located in domains I, II, and IV (6), composed of the structural motif GXGT-(38 residues)-D/E. In each site, the carboxyl acid residue D/E responsible for the bidentate attachment of calcium was suppressed by mutation toward N/Q residue. Thus, Glu-72, Asp-144, and Asp-303 were mutated to Gln, Asn, and Asn, respectively. The mutants Glu-72 3 Gln (M1), Asp-144 3 Asn (M2), and Asp-303 3 Asn (M4) were produced with the synthetic oligonucleotides 5Ј-P CCTGAA-ATCACAAGCTGACTGGAAAATTTG-3Ј, 5Ј-P GGAAGATGACGTCGT-GGGTAACACTTCAGG-3Ј, and 5Ј-P GATTTAAGGGAAATACTAGTGG-CGACTATAAG-3Ј respectively. Recently, another calcium-binding site, related to the presence of Trp-187 and located in domain III, appears to be essential (7). The related mutant Glu-228 3 Ala (M3) was produced with the synthetic oligonucleotide 5Ј-P CCATTGACCGCGCGACGTCT-GGCAATTTAGAGC-3Ј, which introduced a new AatII site for screening via restriction digests. The appropriate mutated cDNA forms were then selected on the evidence of the restriction site introduced during mutagenesis.
The annexin V mutated cDNA inserts were purified by agarose gel and cloned into the pGEX-2T vector. Moreover, multiple calcium-binding sites mutants were produced by digesting and ligating of cDNA from each single mutant into another one. Double mutants M1M2, M1M3, M1M4, and M2M3; triple mutants M1M2M3, M2M3M4, and M1M2M4; and the quadruple mutant M1M2M3M4 were constructed. The cDNA of each mutant was sequenced in its full length to verify the presence of the desired mutation(s) and the absence of others.
Expression and Purification of Wild Type and Mutated Recombinant Annexin V-Wild type and mutated proteins were expressed, isolated, and purified according to previously described procedures (32).
Annexin V was also purified from human placenta using calcium precipitation and anion-exchange chromatography separation (FPLC system; Mono Q column) (32).
As analyzed by SDS-polyacrylamide gel electrophoresis, FPLC profiles, and Western blots, proteins were purified to homogeneity. The extinction coefficient of annexin V at 280 nm was used to determine protein concentrations accurately (⑀ 280 ϭ 0, 59 ml/mg⅐cm Ϫ1 ) (33) and to standardize the dye binding assay described by Bradford (34).
Expression of Data-All determinations were carried out in triplicate, and experiments were repeated on at least three occasions. All data shown are means Ϯ standard error of the means.

Release of [ 3 H]Arachidonic Acid from SLO-permeabilized HL-60 Cells Is Dependent on Ca 2ϩ and on the Stimulus Used-
The release of [ 3 H]arachidonic acid was measured as an index of cPLA 2 activity in differentiated HL-60 cells. This enzyme is present in this cell type and was shown to be regulated by calcium, and by kinases: PKCs and/or MAP kinases (21).
As shown in Fig. 1A, Ca 2ϩ induces an [ 3 H]AA release in a dose-dependent manner. In our model, up to 0.3 M of free calcium (pCa ϭ 6.5), no modification in the basal level of arachidonate release is observed. At 1 M Ca 2ϩ , AA liberation becomes significant. These data are consistent with previous studies on Ca 2ϩ requirement for cPLA 2 activity (21), demonstrating that relocation of cPLA 2 from cytosol to membrane requires micromolar Ca 2ϩ concentrations (35). Moreover, the non-hydrolyzable nucleotide, GTP␥S, which stimulates both small and trimeric G proteins, increases the effect of Ca 2ϩ on AA release. At low levels of calcium, this nucleotide is unable to generate the liberation of AA as observed by another group in electropermeabilized HL-60 granulocytes (Fig. 1A) (36).
The effects of the phorbol ester, PMA, and of the chemotactic factor, fMLP, were studied (Fig. 1B). PMA at 100 nM, by itself, has no effect on arachidonate release. However, at 10 M [Ca 2ϩ ], this tumor promotor increases AA release by 150%. At such calcium concentrations, it can be hypothesized that calcium-dependent conventional PKCs are activated to a level leading to AA release by cPLA 2 . A similar observation has been reported in Me 2 SO-differentiated U937 cells (37). Moreover, PMA increases to a large extent the effect of GTP␥S on arachidonate release as shown on Fig. 1 (A and B).
In comparison, fMLP added 2 min before permeabilization, allows a small but significant increase in AA release even at pCa 7. This powerful agonist possesses specific receptors on HL-60 differentiated by dibutyryl cAMP (38). fMLP stimulation is mediated by Ras and the MAP kinase cascade within minutes after its addition (39,40). fMLP potentiates the Ca 2ϩ ion effect and is as efficient as GTP␥S. Thus, our model of permeabilized differentiated HL-60 cell is adequate to study the regulation of enzyme activity involved in AA release.
Characterization of the PLA 2 Isoform Responsible for AA Release-Free arachidonic acid present in cells can result from the activation of different enzymes (41). This fatty acid can be liberated after a concerted action of PLC and diacylglycerol lipase (41). Different groups have shown that, in permeabilized HL-60 cells, AA is not the product of the activation of these enzymes, but is due to phospholipid hydrolysis after phospholipase A 2 activation (23,42).
This family contains two distinct groups of several enzymes, the secretory PLA 2 (sPLA 2 ), active in the extracellular compartment and the cytoplasmic PLA 2 s, active intracellularly (16). The first group is composed of 14-kDa enzymes, which are inactivated by thiol compounds reducing their essential disulfide bonds (12). Streptolysin O is a permeabilizing agent only in its reduced form requiring a reducing medium. Experiments in cells permeabilized with SLO were nevertheless performed with or without 2 mM DTT. As shown in Fig. 2A, DTT had no effect on calcium-dependent [ 3 H]AA release due to the presence of a reducing agent in SLO. Results similar to those reported above have been obtained with GTP␥S, PMA, or fMLP stimulations (not shown).
So far, the second group of PLA 2 family consists in most cells of two different cytosolic enzymes. The first one is the group IV calcium-dependent 85-kDa cytosolic phospholipase A 2 (cPLA 2 ), selective for arachidonic acid. The second one is the calciumindependent PLA 2 (iPLA 2 ), which seems to be essentially involved in reincorporation of part of the free-fatty acid into membranes (19,20). This enzyme has not yet been demonstrated to be present in HL-60 cells. The respective importance of both intracellular PLA 2 s was studied using selective inhibitors of each enzyme. Involvement of group IV cPLA 2 was investigated by using AACOCF3 (an analogue of arachidonate; Ref. 43) and MAFP (an irreversible inhibitor of this enzyme; Ref. 44). These inhibitors display a specificity for cPLA 2 versus sPLA 2 s (43,44). Fig. 2B represents the effect of a pretreatment of HL-60 cells by 10 M AACOCF3 on arachidonate release at pCa 5 after stimulation by different molecules. Whatever the stimulus used, AACOCF3 decreases the arachidonic acid liberation to 56% (from 49% to 59%). At pCa 6 and 5.5, the inhibition is 51 and 59% respectively, with no significant difference between GTP␥S, fMLP, and GTP␥S ϩ PMA. The inhibition of AA release by AACOCF3 in HL-60 cells is of the same order of magnitude as that reported in platelets activated by thrombin or calcium ionophore (45). Dose-response curve of MAFP on GTP␥S-stimulated AA release at pCa 5 is shown in Fig. 2C. This compound decreases the release of [ 3 H]AA by 60, 69, and 71% at 10, 20, and 30 M, respectively. Moreover, both AA-COCF3 and MAFP do not change the AA release at pCa 8 and 7 (data not shown).
However, recent evidence demonstrated that AACOCF3 and MAFP are also iPLA 2 inhibitors (46,47). Although iPLA 2 has never been found in HL-60 cells, the responsibility of this enzyme in the observed AA release was tested using BEL, which inhibits iPLA 2 irreversibly. This compound was shown to be a poor inhibitor for cPLA 2 activity (20,48). In our model, whatever the stimuli used, BEL, at concentrations up to 50 M, was found to be inefficient in reducing AA release.
Thus, in our experimental conditions, AACOCF3-and MAFP-sensitive AA release should be due to cPLA 2 activation.
Inhibition of Cytosolic Phospholipase A 2 by Annexin V-To study the intracellular effect of a cellular component, we used a model of cells permeabilized with SLO. This agent generates persistent membrane permeabilization, which allows the flux of proteins of molecular mass up to 400 kDa, including 3-phosphoglycerate kinase (45 kDa) and lactate dehydrogenase (140 kDa) (25). The efflux of these proteins is often used to follow the permeabilization process. Addition of recombinant annexin V to the medium during SLO permeabilization gives a dose-dependent inhibition of AA release in differentiated HL-60 cells stimulated by 10 M GTP␥S at pCa 5 (Fig. 3A). The EC 50 of the inhibitory effect was found to be 1 M with a plateau occurring at 2 M annexin V. In such experimental conditions, the addition of 1 molecule of annexin V, which binds 4 molecules of Ca 2ϩ , does not reduce the amount of free Ca 2ϩ in the reaction: in the presence of 2 M annexin V, Ca 2ϩ decreases from 10 to 9.6 M, which is not different from the control and excludes a Ca 2ϩ buffering effect of annexin V. Identical results in experiments performed with purified human placental annexin V have been obtained (data not shown). Both proteins have been shown to share the same biological characteristics (49). Thus in our model, annexin V appears to strongly inhibit cPLA 2 activation. Moreover, at 3 M, recombinant annexin V is able to inhibit cPLA 2 activity whatever the stimulus used (Fig. 3B). Relatively, annexin V inhibits fMLP-stimulated cPLA 2 activity to a lesser extent than the enzymatic activity stimulated by GTP␥S or GTP␥S ϩ PMA. This can be explained by the two minute delay between addition of the chemotactic factor and permeabilization allowing annexin V effect. The inhibitory activity of annexin V was suppressed by heating the protein at 100°C for 5 min, confirming that this activity depends on the integrity of the structure. To analyze the effect of Ca 2ϩ on cPLA 2 inhibition by annexin V, experiments were performed at various [Ca 2ϩ ] (10 Ϫ7 , 10 Ϫ6 , and 3 ϫ 10 Ϫ6 M). As shown in Fig. 3C, 3 M annexin V has no effect on arachidonate release at 10 Ϫ7 and 10 Ϫ6 M Ca 2ϩ . In contrast, at 3 M Ca 2ϩ , a decrease in AA release of 38%, 48%, and 33% for control cells, GTP␥S-, and fMLP-stimulated cells, respectively, was observed. Thus, in our cell line, annexin V is an inhibitor of cPLA 2 when Ca 2ϩ is higher than 1 M. [Ca 2ϩ ] i concentration in the micromolar range can physiologically occur in subcellular compartment after cell activation (50) and has been reported to be required for annexin V binding to artificial phospholipid vesicles (51).
Effect of Annexin V on PLD and PLC Activities-PLC and PLD have been shown to be activated by Ca 2ϩ . In SLO-permeabilized HL-60 cells, the three phospholipases cPLA 2 , PLC, and PLD respond to Ca 2ϩ in a similar way; at pCa 6 only a small activation of each enzyme is observed, which is maximal at pCa 5. Higher Ca 2ϩ concentrations, which are not any more physiological, lead to a decrease in the enzyme activity (24,29). Table I reports the effect of 20 M annexin V at 10 M Ca 2ϩ on GTP␥S-activated PLC, PLD, and cPLA 2 . As shown, annexin V has no effect on PLC and PLD activities, whereas it inhibits cPLA 2 under the same conditions. Thus, in permeabilized HL-60 cells, annexin V inhibits solely cPLA 2 .
Importance of Each Calcium Site of Annexin V in cPLA 2 Inhibition-We have demonstrated above that calcium is essential for both cPLA 2 activation and annexin V effect on the enzyme. It is also known that annexins are structurally related with a core composed of four domains containing each a type II Ca 2ϩ -binding site (6,7). To understand the mechanism of inhibition of annexin V on cPLA 2 activity and to determine which specific Ca 2ϩ -binding site is responsible for the effect on cPLA 2 , we constructed a series of mutants in a similar manner to what has been reported for annexin IV (52).
Construction of annexin V mutants was done, according to crystallographic data, by destroying high affinity Ca 2ϩ -binding sites (6). All mutants were purified with yields of 3-5 mg/liter of bacteria culture. SDS-polyacrylamide gel electrophoresis and Western blot analysis showed no difference between wild type and mutants of annexin V. Circular dichroism analysis of the recombinant wild-type protein and the M1, M2, M1M2, M1M2M3, and M1M2M3M4 mutants were conducted, revealing no difference between proteins (not shown). This observation demonstrates that mutations have not modified the overall structure.
The mutants containing the defective Ca 2ϩ -binding site located in domain II or III (mutants M2 or M3) do not modify the inhibitory effect of annexin V (Fig. 4A). In contrast, mutant M1 loses the inhibitory effect, inhibiting only 20% of AA release after cPLA 2 activation by GTP␥S and 10 M Ca 2ϩ . Intermediate results are obtained with mutant M4 giving an inhibition of 35% of cPLA 2 activity.
Annexin V mutants containing more than one defective Ca 2ϩ -binding site were also produced. Experimental analysis of these multiple mutants demonstrates a hierarchy in the calcium-binding sites of annexin V, as reported for annexin IV  (52), for inhibition of cPLA 2 activity (Fig. 4, B and C). Thus, a series of mutants have lost their ability to inhibit AA release: M1, M1M2, M1M3, M1M4, M1M2M3, M1M2M4, and M1M2M3M4. All these mutants contain a defect in the Ca 2ϩ -binding site of domain I. Mutations of Ca 2ϩ -binding sites of one of other domains already with M1 does not increase the loss of inhibition obtained (Fig. 4B). Moreover, both mutants M1M2M4 and M1M2M3M4 have completely lost the ability to inhibit cPLA 2 activity even at doses up to 20 M.
These results indicate that the Ca 2ϩ -binding site located in domain III is not important for cPLA 2 inhibition. No significant differences were observed for AA release inhibition obtained for each mutant whatever the stimulus used, GTP␥S, fMLP, or GTP␥S ϩ PMA.
These results are the first demonstration of the importance of the Ca 2ϩ -binding site located in the first domain of annexin V in the inhibition of cPLA 2 . Moreover, this demonstration has been carried out in permeabilized cells not too far from physiological conditions. Does Annexin V Modulate cPLA 2 Activity by Inhibiting PKC?-It has been clearly demonstrated using recombinant proteins that cPKCs phosphorylate cPLA 2 in vitro (53). In vivo, cPLA 2 phosphorylation by cPKCs was also observed in some cell types (21). Our group has recently reported that annexin V is an endogenous inhibitor of cPKCs activity (11). However, the mechanism of this inhibition is not clear and several hypotheses can be proposed, including direct interaction between annexin V and PKC or competition between both proteins for phospholipids or calcium (10,54). The importance of cPKC inhibition by annexin V in cPLA 2 regulation was investigated using different approaches. 1) We used GF109203x, a potent and selective inhibitor of protein kinase C (55). To evaluate the relative importance of PKC in signal transduction after fMLP or GTP␥S ϩ PMA stimulation, HL-60 cells were pretreated with 10 M GF109203x for 30 min at 37°C. As shown in Fig. 5, 18, 23, or 19% inhibition of, respectively, Ca 2ϩ -, fMLP-, or GTP␥S ϩ PMA-stimulated cPLA 2 activity was observed. These results argues for a MAP kinase predominant pathway in cPLA 2 activation in HL-60 cells and is in accordance with the results of Worthen et al., who demonstrated that neutrophil stimulation with the chemotactic factor (fMLP) activates predominantly this pathway (39). In GF109203x-treated cells, annexin V inhibits to a further extent arachidonic acid release whatever the stimulus used (Fig. 5). This inhibition leads to the release of the same amount of AA as that found in nontreated cells (Fig. 3B).
2) C-terminal tail of annexin V (C-ter) shares an analogous sequence (named the annexin-like domain) with 14-3-3 proteins, endogenous inhibitors of PKC (56). Moreover, crystal structures of both families of proteins reveal that this domain is accessible for direct interaction with PKCs. To test this potential mechanism of PKC inhibition by annexin V, a mutant of annexin V deleted in the last 14 residues was constructed. However, this protein was insoluble (even with detergents) and could not be purified. Thus, another strategy was chosen, which used a 16-amino acid synthetic peptide derived from the Cterminal tail of annexin V. Such an approach has been successfully used by another group to reproduce 14-3-3 protein capacity to inhibit phosphorylation by PKC in permeabilized platelets (57). In our model, the C-ter annexin V peptide was unable to modify AA release at all doses used (Table II). DISCUSSION Most members of the annexin family are known to possess an antiphospholipase A 2 activity. This activity had only been described on the 14-kDa secretory PLA 2 s and the proposed mechanism of action generally accepted was substrate depletion (12). Recent advances on PLA 2 s have revealed the complexity of this enzyme family, including at least the extracellular PLA 2 s (sPLA 2 s) and the intracellular PLA 2 s (cPLA 2 and iPLA 2 ) (16). The discovery of cPLA 2 , the knowledge of its sequence and regulation have permitted a new approach of the annexin inhibitory effect on this enzyme. Using pure cPLA 2 , Kim et al. have shown that this enzyme was inhibited trough a specific interaction by annexin I (15). This report was also suggested using an annexin I-derived peptide (amino acids 13-25) (58). Moreover, a monoclonal antibody to annexin I was shown to abolished anti-PLA 2 activity without altering the capacity of annexin I to bind phospholipids (59).
Annexin V is also involved in the control of inflammation but was not previously investigated in the control of cPLA 2 activity. Indeed, annexin V present in the extracellular medium was reported to control inflammation. Our group has demonstrated that the degree of inflammation in rhumatoid arthritis was strongly correlated with annexin V antibody level present in synovial fluid (60). Moreover, intracerebroventricular injection of an annexin V-derived peptide (amino acids 204 -212) has been reported to possess anti inflammatory and antipyretic properties (61). However, annexin V is not mainly an extracellular protein. It is mostly found within cells, in the cytoplasm, and represents the most abundant and ubiquitous annexin. We then addressed the issue of its possible interaction with cPLA 2 , which shares the same properties (cytoplasmic, ubiquitous, and activated by similar [Ca 2ϩ ] i ).
In our model of differentiated and permeabilized HL-60 cells, both native placental and recombinant annexin V were found to be potent inhibitors of cPLA 2 . This effect is shown to be mediated through the protein calcium-binding sites, specially the site located in domain I.
Annexin V-cPLA 2 Inhibition: Physiological Relevance-Our original observation that annexin V inhibits cPLA 2 is likely to be also relevant in physiological conditions for several reasons.
(i) Both proteins have been shown to relocate in the same subcellular compartment, the nuclear membrane, after activation in many cell types (62)(63)(64)(65). This location may be relevant to their related biological function.
(ii) Annexin V relationship with phospholipids, substrates of cPLA 2 , is complex and depends on subcellular Ca 2ϩ concentrations (1,2). At a concentration that allows cPLA 2 membrane translocation (0.8 M) (35), a recent study on platelets has demonstrated that annexin V is firmly bound to membranes (51). A major part of membrane-bound annexin V (85%) is resistant to EGTA extraction that could be due to a still unidentified membrane protein. It is proposed that membrane binding of annexin V at such [Ca 2ϩ ] i would provoke a destabilization of the lipid bilayer (66). This could favor cPLA 2 activity (67). At higher [Ca 2ϩ ] i (10 M), membrane binding of annexin V becomes sensitive to EDTA (51) and could stabilize the lipid bilayer (66), thus decreasing the activity of membrane-associated cPLA 2 .
(iii) In the present study, annexin V inhibitory effect is shown to be Ca 2ϩ -dependent with a maximum at 10 M, the highest calcium concentration used. In vivo, during cellular activation, calcium oscillations may reach 10 M [Ca 2ϩ ] i in specific subcellular domains located at the nuclear membrane and at the sarcoplasmic membrane (50). These locations of high [Ca 2ϩ ] i correspond with those of cPLA 2 and annexin V after cell stimulation (62)(63)(64)(65).
(iv) Moreover, cPLA 2 activation plays a role in the regulation of [Ca 2ϩ ] i increase. Indeed, arachidonate mobilization was recently demonstrated to be coupled to activation of the capacitive pathway of calcium entry (68), suggesting that cPLA 2 may promote its own negative control by increasing [Ca 2ϩ ] i to levels necessary for annexin V inhibitory effect.
Mechanism of Inhibition of cPLA 2 by Annexin V: Substrate Depletion-Several hypotheses can be proposed for the observed inhibitory effect of annexin V on cPLA 2 activity: inhibition of cPKC by annexin V preventing cPLA 2 phosphorylation essential to its activation, substrate depletion necessary for cPLA 2 activity, and/or direct interaction between these two proteins.
i) In the present study we observed that cPLA 2 activity stimulated by PMA, at 10 M Ca 2ϩ , can be inhibited by annexin V. However, it has been established that PMA stimulates cPLA 2 via both pathways, PKC and the MAP kinase cascade (69). Hence, involvement of PKC in PMA-stimulated cells could be partial. Moreover, the relative importance of the PKC pathway involved in cPLA 2 activation has been demonstrated to depend on the cell line used (22). In our model, a pretreatment with the specific PKC inhibitor GF109203x only decreases by 20% arachidonic acid release due to all stimuli. Thus, this pathway appears to be of minor importance in HL-60 cells. Therefore, as annexin V still inhibits AA release in GF109203xtreated cells, PKC inhibition by annexin V is unlikely to be the mechanism responsible for the inhibitory effect of annexin V on cPLA 2 activity in this cell type. However, other studies have reported a cPLA 2 phosphorylation exclusively by a PKC-dependent pathway in other cell types (21). Thus, action of annexin V on cPLA 2 activity via PKC inhibition should be studied in these cell types.
ii) Like cPLA 2 , annexins are supposed to bind to phospholipids by their Ca 2ϩ -binding sites. We hypothesized that, as for secretory PLA 2 s activity inhibition, cytosolic PLA 2 activity could be controlled by annexin V through a substrate depletion mechanism. Site-directed mutagenesis of annexin II and IV has been realized to study their Ca 2ϩ -binding properties to artificial vesicles; both studies demonstrate that these sites are responsible of the attachment of annexins to the vesicles (52,70). In the present study, a similar strategy of site-directed mutagenesis of the four Ca 2ϩ -binding sites of annexin V was chosen. Here, we demonstrate, for the first time, that domain I plays the most important role in the inhibitory effect of this protein on cPLA 2 activity. In contrast to sites located in domains II and III, the fourth Ca 2ϩ -binding site also participates in the inhibition of AA release. The prevalence of the module comprising domains I and IV for membrane binding was proposed by one of the first study of crystal structure of annexin V analyzing the calcium-binding sites (71). In contrast, recent electron microscopy analysis of annexin V bound to artificial lipid monolayer seemed to indicate that all four domains participate in membrane binding (72). Both results are compatible. Indeed, after membrane binding of module (I/IV), a change in orientation of module (II/III) may occur, which allows a shift of the latter closer to the membrane (72), reconciling our data with the electron microscopy results. A complete loss of inhibition appears only in M1M2M4 or in the quadruple mutant, indicating a possible cooperativity between the two modules. Our data show the prevalence of domain I in all annexin V mutants defective in an inhibitory effect on cPLA 2 . The dynamic aspect of the two modules, (I/IV) and (II/III), has been shown to be disturbed by a mutation on one of residues Glu-17 and Glu-78, which alters the charge of module (I/IV) (8). In the same way, the mutation E72Q in repeat I might also affect the charge of module (I/IV) and thus the control of the intermodule angle. This could explain the prevalence of domain I. However, charge modification induced by mutation is not sufficient to explain the decrease in the inhibitory effect. Indeed, mutant M2 has the same charge modification as M1. No modification in its inhibitory property was observed.
In our experiments, mutation of the third domain was completely silent, indicating no involvement of this site in calciumtriggered interaction with phospholipid membranes.
Thus, this study confirms that annexins differ in the position and importance of their Ca 2ϩ -binding sites and demonstrates that in annexin V the main sites are located in domains I and IV. In contrast, domains II, III, and IV are predominant in annexin I and II (2,70).
The loss of inhibitory effect of M1M2M4 and M1M2M3M4 on cPLA 2 activity strongly suggests that annexin V acts on the enzyme through its Ca 2ϩ -binding sites.
The present study clearly demonstrated for the first time an inhibitory effect of annexin V on cPLA 2 activation. Ca 2ϩ -binding sites present in module (I/IV) are necessary for this property. Thus, the hypothesis of substrate depletion is likely the inhibitory mechanism of annexin V on cPLA 2 regulation.