Identification of the Calmodulin-binding Domain of Recombinant Calcium-independent Phospholipase A2β

Calcium-independent phospholipase A2 (iPLA2) is the major phospholipase A2 activity in many cell types, and at least one isoform of this enzyme class is physically and functionally coupled to calmodulin (CaM) in a reversible calcium-dependent fashion. To identify the domain in recombinant iPLA2β (riPLA2β) underlying this interaction, multiple techniques were employed. First, we identified calcium-activated CaM induced alterations in the kinetics of proteolytic fragment generation during limited trypsinolysis (i.e. CaM footprinting). Tryptic digests of riPLA2β (83 kDa) in the presence of EGTA alone, Ca+2 alone, or EGTA and CaM together resulted in the production of a major 68-kDa protein whose kinetic rate of formation was specifically attenuated in incubations containing CaM and Ca+2 together. Western blotting utilizing antibodies directed against either the N- or C-terminal regions of riPLA2β indicated the specific protection of riPLA2β by calcium-activated CaM at a cleavage site ≈15 kDa from the C terminus. Moreover, calcium-activated calmodulin increased the kinetic rate of tryptic cleavage near the active site of riPLA2β. Second, functional characterization of products from these partial tryptic digests demonstrated that ≈90% of the 68-kDa riPLA2β tryptic product (i.e. lacking the 15-kDa C-terminus) did not bind to a CaM affinity matrix in the presence of Ca2+, although >95% of the noncleaved riPLA2β as well as a 40-kDa C-terminal peptide bound tightly under these conditions. Third, when purified riPLA2β was subjected to exhaustive trypsinolysis followed by ternary complex CaM affinity chromatography, a unique tryptic peptide (694AWSEMVGIQYFR705) within the 15-kDa C-terminal fragment was identified by RP-HPLC, which bound to CaM-agarose in the presence but not the absence of calcium ion. Fourth, fluorescence energy transfer experiments demonstrated that this peptide (694) bound to dansyl-calmodulin in a calcium-dependent fashion. Collectively, these results identify multiple contact points in the 15-kDa C terminus as being the major but not necessarily the only binding site responsible for the calcium-dependent regulation of iPLA2β by CaM.

The phospholipase A 2 -catalyzed release of arachidonic acid from its phospholipid storage depots is a critical component of intra-and intercellular signal transduction. In most noncirculating cells (e.g. cardiac myocytes, pancreatic islet ␤-cells, hippocampal neurons, and vascular smooth muscle cells), calciumindependent phospholipase A 2 (iPLA 2 ) 1 is the major but not the only phospholipase A 2 activity present (1)(2)(3)(4)(5)(6). Multiple lines of experimental evidence implicate iPLA 2 as an important mediator of arachidonic acid release in several cell types including: 1) the inhibition of the large majority of AVP-induced arachidonic acid release in A-10 smooth muscle cells by 1-2 M BEL (7); 2) the attenuation of the release of arachidonic acid in lipopolysaccharide-stimulated macrophages by either BEL or antisense DNA targeted to iPLA 2 ␤ (8,9); 3) the robust release of ligand-stimulated arachidonic acid in cells whose rise in cytosolic calcium ion content is ablated by EGTA and 1,2-bis(oaminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid tetra(acetoxymethyl)ester (i.e. the release of arachidonic acid occurs in the absence of Fura 2-detectable changes in intracellular Ca ϩ2 levels; Refs. 10 and 11); 4) the release of arachidonic acid by agents that deplete intracellular Ca ϩ2 stores in the absence of receptor occupancy (e.g. thapsigargin, cyclopiazonic acid, 2, 5-di-(t-butyl)-1,4-hydroquinone, and A23187) when alterations in [Ca 2ϩ ] i are prevented (10,11); and 5) the ionophore-induced release of arachidonic acid from HEK cells expressing riPLA 2 ␤ (but not in wild type HEK cells) with its subsequent directed conversion to prostaglandin E 2 by cyclooxygenase I but not cyclooxygenase II (12).
To gain insight into the biochemical mechanisms responsible for the activation of iPLA 2 in stimulated cells, an early observation was re-explored that demonstrated that the addition of calcium ion to myocardial cytosol inhibited iPLA 2 activity (1). Subsequently, this inhibition was shown to be due to a protein factor that was identified as calmodulin after purification to homogeneity (13). These results were confirmed by reconstitution of calcium-dependent iPLA 2 inhibition utilizing authentic calmodulin (13). Through detailed analysis of the interaction of myocardial iPLA 2 with CaM, we demonstrated the formation of a catalytically inactive ternary complex of CaM⅐Ca 2ϩ ⅐iPLA 2 that could be reversibly dissociated by chelation of calcium ion with EGTA to regain full catalytic activity. Subsequent work demonstrated that agents that deplete intracellular calcium * This work was supported jointly by Juvenile Diabetes Foundation International File Grant 996003 and National Institutes of Health Grants 1 PO1 HL 57278-02, 2 R02 HL 41250-06A1, and P60DK20579-22. 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: Washington University School of Medicine, Div. of Bioorganic Chemistry and Molecular Pharmacology, 660 South Euclid Ave., Campus Box 8020, St. Louis, MO 63110.Tel.: 314-362-2690; Fax: 314-362-1402. 1 The abbreviations used are: iPLA 2 , calcium-independent phospholipase A 2 ; iPLA 2 ␤, calcium-independent phospholipase A 2 ␤; riPLA 2 ␤; recombinant calcium-independent phospholipase A 2 ␤; BEL, (E)-6-(bromomethylene)-3-(1-naphthalenyl)-2H-tetrahydropyran-2-one; CaM, calmodulin; DTT, dithiothreitol; RP-HPLC, reverse-phase high performance liquid chromatography; PAGE, polyacrylamide gel electrophoresis; CAPS, 3-(cyclohexylamino)propanesulfonic acid; dansyl, 5-dimethylaminonaphthalene-1-sulfonyl. ion pools (e.g. thapsigargin and cyclopiazonic acid), even in the absence of receptor occupancy or alterations in cytosolic calcium ion concentration, resulted in arachidonic acid release in intact cells (10). Furthermore, agents that inhibited the interaction of calmodulin with its target proteins through prevention of calcium-induced conformational changes in calmodulin (e.g. W-7) resulted in the release of [ 3 H]arachidonic acid in resting A-10 smooth muscle cells (10). These observations indicate that the majority of iPLA 2 activity in cells is tonically inhibited and that dissociation of calmodulin from iPLA 2 is the major mechanism that transforms a calcium-independent catalytic activity into an enzyme that is responsive to alterations in intracellular calcium ion homeostasis. Collectively, these results gave rise to the hypothesis of "calcium pool depletionmediated iPLA 2 activation" (10). To further define the physiologic relevance of this hypothesis, it became important, from both a mechanistic perspective as well as a therapeutic strategy, to determine the site(s) of interaction of iPLA 2 ␤ with calmodulin.
Since the first x-ray crystal structure (14) and NMR solution structure (15) of CaM⅐peptide complexes became available, it was apparent that calcium-induced conformationally activated calmodulin interacted with its regulatory targets at multiple contact points. No universal consensus sequences mediating the interaction of calmodulin with its extremely diverse protein targets have been identified, although the importance of positionally conserved hydrophobic and basic amino acid residues has been demonstrated in some cases (16 -20). Accordingly, unambiguous identification of the loci of interaction of CaM with iPLA 2 ␤ cannot be recognized a priori from comparisons of the iPLA 2 ␤ primary sequence with other known calmodulinregulated target proteins. To identify the site of interaction of calcium-activated calmodulin with iPLA 2 ␤, analysis of direct protein-protein interactions (protein footprinting), functional interactions (ternary complex affinity chromatography), and biophysical interactions (fluorescence resonance energy transfer) were pursued. Herein, we demonstrate that incubation of iPLA 2 ␤ with trypsin in the presence of calmodulin and calcium ion protects against tryptic cleavage at a C-terminal site (at or near residue 630) and increases hydrolytic cleavage near the active site of riPLA 2 ␤. Moreover, complete tryptic digestion of riPLA 2 ␤ resulted in a peptide within this 15-kDa C-terminal region (AWSEMVGIQYFR corresponding to residues 694 -705) that binds to CaM in a calcium-dependent manner. Collectively, these results identify the 15-kDa C-terminal region of riPLA 2 ␤ as necessary and sufficient for the calcium-induced binding of calmodulin to riPLA 2 ␤ and conformational alteration of riPLA 2 ␤ near the active site.

Materials-
The peptide corresponding to residues 23-42 (CRVKEIS-VADYTSHERVREEG) near the N terminus of iPLA 2 ␤ (GenBank accession number I15470) was synthesized by Alpha Diagnostic International (San Antonio, TX). The peptide corresponding to residues 731-744 (CTEVYIYEHREEFQK) near the C terminus of iPLA 2 ␤ was synthesized by the Protein Chemistry Laboratory at Washington University. Both peptides contain N-terminal cysteine residues for covalent linkage to maleimide-activated keyhole limpet hemocyanin as a carrier protein for generating antibodies as well as to activated thiol-Sepharose 4B for subsequent affinity purification of the antibodies. The synthetic 12-amino acid peptide (AWSEMVGIQYFR) corresponding to amino acid residues 694 -705 of the iPLA 2 ␤ sequence was synthesized by Research Genetics. Sequencing grade modified trypsin used for trypsinolysis of iPLA 2 ␤ was purchased from Promega. Kaleidoscope prestained SDS-PAGE protein standards were obtained from Bio-Rad. The RPC C2/ C18 SC 2.1/10 column, CNBr-activated Sepharose resin, and ECL reagents were purchased from Amersham Pharmacia Biotech. L-␣-1-Palmitoyl-2-[1-14 C]arachidonyl-phosphatidylcholine, used for measurements of riPLA 2 ␤ activity, was purchased from PerkinElmer Life Sciences. High purity bovine brain calmodulin was purchased from Calbiochem. Calmodulin-agarose and most other reagents were obtained from Sigma.
Generation and Affinity Purification of Anti-iPLA 2 ␤-Antibodies-Recombinant calcium-independent phospholipase A 2 ␤ was purified to homogeneity from an Sf9 cell expression system by sequential affinity chromatographic steps with ATP-agarose and CaM-agarose affinity resins as described previously (21,22). Affinity purified polyclonal antibodies directed against riPLA 2 ␤ were prepared by initial injection of 50 g of purified riPLA 2 ␤ in 0.5 ml of Freund's complete adjuvant and boosted every 2 weeks with 50 g of riPLA 2 ␤ in Freund's incomplete adjuvant until seroconversion was detected by immunoblotting. Purified riPLA 2 ␤ was coupled to CNBr-activated Sepharose according to the manufacturer's instructions to generate an affinity column to purify the anti-iPLA 2 ␤ antibodies. Immunoreactive rabbit serum was loaded onto an riPLA 2 ␤ Sepharose column, and nonspecifically bound protein was removed from the column by exhaustive washing with 10 column volumes of 50 mM Tris-HCl, pH 7.4, containing 150 mM NaCl. Anti-iPLA 2 ␤ antibody was eluted by sequential application of solutions containing either 4.9 M MgCl 2 or 100 mM glycine, pH 2.5. After elution, the antibody was dialyzed against ammonium bicarbonate buffer prior to lyophilization for storage at Ϫ20°C.
Antibodies directed against the peptides corresponding to residues 23-42 (N-terminal) and residues 731-744 (C-terminal) regions of iPLA 2 ␤ were generated by immunizing rabbits with conjugates of the peptides with keyhole limpet hemocyanin. Briefly, maleimide-activated keyhole limpet hemocyanin (2 mg) was reacted with each peptide (2 mg) according to the manufacturer's instructions and dialyzed against 83 mM sodium phosphate, pH 7.2, containing 0.9 M NaCl. Rabbits were immunized/boosted with the conjugates using Freund's complete/incomplete adjuvants as described above. For affinity purification of the antibodies, each peptide (2 mg) was covalently linked to activated thiol-Sepharose 4B (1 ml) in the presence of an equal volume of 0.1 M sodium citrate buffer, pH 4.5. Following an overnight incubation with mixing at room temperature, the extent of the reaction was monitored spectrophotometrically by the displacement of the 2-pyridyl groups as 2-thiopyridone (⑀ 343 ϭ 8.08 ϫ 10 3 M Ϫ1 cm Ϫ1 ). Immunoreactive rabbit antisera were diluted 1:10 with 10 mM Tris-HCl, pH 7.5, prior to application to the appropriate affinity resin equilibrated with the same buffer. The resins were extensively washed with 10 column volumes of 10 mM Tris-HCl buffer (pH 7.5) containing 500 mM NaCl prior to elution of bound antibodies with 0.1 M glycine, pH 2.5, into collection tubes containing 1 M Tris-HCl, pH 9.0 (1:10 fraction volume), for investigation. Antibodies were dialyzed against phosphate-buffered saline and stored at 4°C prior to use.
CaM-Agarose Affinity Chromatography of Partially Trypsinized iPLA 2 ␤-Purified recombinant iPLA 2 ␤ (50 g) was partially digested with trypsin (1:30 w/w trypsin:iPLA 2 ␤) for 4 -20 min at room temperature in 25 mM imidazole buffer, pH 8.0, containing 0.1 mM EGTA, 1 mM DTT, and 50 -150 mM NaCl. The reaction was terminated by addition of 1 mM 4-(2-aminoethyl)benzene-sulfonylfluoride. Calcium chloride was added to a final concentration of 5 mM, and the sample was applied to a 0.5-ml CaM-agarose column equilibrated with 25 mM imidazole, pH 8.0, containing 5 mM CaCl 2 , 1 mM DTT, and 150 mM NaCl. The column was then washed with equilibration buffer containing 10-fold less (0.5 mM) CaCl 2 , followed by elution of the bound proteins with equilibration buffer (without CaCl 2 ) containing 10 mM EGTA. Column fractions were analyzed by Western analysis as described above using polyclonal antibodies directed against the iPLA 2 ␤ holoprotein.
Tryptic Digestion, Isolation, and Characterization of the CaM-binding Peptide from riPLA 2 ␤-A 25-g sample of highly purified riPLA 2 ␤ was exhaustively digested with 1 g of N-tosyl-L-phenylalanine chlo-romethyl ketone-modified trypsin in 1.5 ml of 25 mM imidazole buffer, pH 8.0, containing 1 mM EGTA for 18 h at 37°C. The resultant proteolytic fragments were adjusted to a final calcium ion concentration of 5 mM and loaded onto a 0.5-ml column of CaM-agarose. The column was washed with 5 column volumes of a buffer containing 25 mM imidazole, pH 8.0, and 0.5 mM CaCl 2 prior to the elution of bound peptides with a buffer containing 25 mM imidazole, pH 8.0, and 4 mM EGTA. Aliquots of the tryptic peptides (50 l) in the CaM-agarose load and void fractions were directly analyzed by RP-HPLC utilizing a RPC C2/C18 SC 2.1/10 Smart System column with UV absorbance detection at 215 nm. The peptides from the EGTA eluent of the CaM-agarose column (2.0 ml) were concentrated by lyophilization, resuspended in 100 l of deionized H 2 O containing 0.075% trifluoroacetic acid (buffer A), and loaded onto a RPC C2/C18 SC 2.1/10 RP-HPLC column. Peptides were resolved utilizing a discontinuous linear gradient from 0 to 38% buffer B (80% acetonitrile and 0.06% trifluoroacetic acid) for 60 min, 38 to 75% buffer B for 30 min, and 75 to 100% buffer B for 15 min. The affinity purified peptide from the RP-HPLC purification of the calmodulin-agarose eluent (denoted in Fig. 4B) was collected, lyophilized, and sequenced by automated Edman degradation utilizing an Applied Biosystems Procise Sequencer. RP-HPLC of the synthetic 12-amino acid peptide corresponding to the obtained sequence (vida infra) was conducted employing the chromatographic conditions described above.
Fluorescence Spectrometry of Dansyl-CaM-Experiments employing dansyl-calmodulin were conducted utilizing an Aminco SLM 4800C fluorescence spectrometer (SLM Instruments, Inc.) using established methods (24 -27). Briefly, the indicated amounts of synthetic peptide were added to a 3-ml cuvette containing 2 ml of dansyl-CaM (1 g/ml) in the presence of 20 mM HEPES, pH 7.2, 130 mM KCl, and 500 M CaCl 2 or 1 mM EGTA. Emission spectra were recorded from 400 to 550 nm utilizing an excitation wavelength of 340 nm.
Deletional Mutagenesis of iPLA 2 ␤-Mutants of iPLA 2 ␤ were prepared by polymerase chain reaction-directed mutagenesis techniques as previously described (28). Three truncated mutants of iPLA 2 ␤ containing amino acid residues 1-381, 1-600, and 1-690 were prepared. Briefly, antisense primers containing a stop codon after amino acids 381, 600, and 690 followed by an SphI restriction site were prepared for use in polymerase chain reaction with a sense primer containing an EcoRI restriction site prior to the ATG start site of the iPLA 2 ␤ coding sequence. EcoRI/SalI-digested polymerase chain reaction products were cloned into similarly digested pFAST vectors for expression of recombinant protein in the baculovirus expression system as previously described (22). All constructs were sequenced on both strands to ensure fidelity of the sequence within the constructs prior to use in the baculoviral expression system.

RESULTS
Previous work has demonstrated that riPLA 2 ␤ reversibly binds to calmodulin through a ternary complex comprised of calcium, calmodulin, and unidentified domains of riPLA 2 ␤ (22). Accordingly, we sought to identify the specific regions of riPLA 2 ␤ involved in calmodulin binding and modulation of iPLA 2 ␤ enzymic activity. Historically, one approach through which protein-protein contact points have been identified is through binding of the regulatory protein to the target protein and protection of the target protein from proteolysis at the binding site (i.e. protein footprinting). Accordingly, we analyzed the kinetics of the production of individual proteolytic fragments formed during timed incubations of trypsin with riPLA 2 ␤, calcium, and calmodulin. Incubation of trypsin with riPLA 2 ␤ produced six major proteolytic fragments (A-F) within 9 min under the conditions employed (Fig. 1). Similar kinetic analysis of tryptic digests of riPLA 2 ␤ in the presence of EGTA alone, Ca ϩ2 alone, or CaM and EGTA together produced similar amounts and ratios of proteolysis products. Remarkably, tryptic digests of riPLA 2 ␤ in the presence of CaM and calcium together resulted in the nearly complete disappearance of the major proteolytic product at 68 kDa and the increased intensity of the band at 40 kDa (Fig. 1, band E). These results demonstrate that calcium-activated CaM specifically protected riPLA 2 ␤ from trypsinolysis at a single cleavage site within ϳ15 kDa of either its N or C terminus. Moreover, they suggest that a site near the center of riPLA 2 ␤ (i.e. near the active site) undergoes a conformational alteration in the presence of calcium and calmodulin together but not with either alone. To determine the locus of the protected site, immunoaffinity purified polyclonal antibodies directed against either the N-or C-terminal regions of iPLA 2 ␤ were generated as described under "Experimental Procedures." Western blotting of the products from the limited trypsinolysis of riPLA 2 ␤ using the Nterminal antibody demonstrated bands at 83 kDa (starting material) and fragments at 68, 59, and 43 kDa ( Fig. 2A). In contrast, Western blotting with the C-terminal antibody identified the 83-kDa (starting material) as well as 53-and 40-kDa proteolyzed products (Fig. 2B). These results demonstrate that calcium-activated CaM tightly binds iPLA 2 ␤ at a specific site Ϸ15 kDa from the C-terminal region (i.e. the production of the 68-kDa product was dramatically attenuated in the presence of calcium-activated CaM but not with calmodulin or calcium ion alone). Furthermore, the results also demonstrate the conformational modification of iPLA 2 ␤ by calcium-activated CaM at a site 40 kDa from the C terminus (i.e. near the active site).
To substantiate the functional importance of the 15-kDa C-terminal region of riPLA 2 ␤ in the binding of calcium-activated calmodulin to iPLA 2 ␤, ternary complex affinity chromatography was employed. Previously, it has been established that riPLA 2 ␤ (both crude and purified) binds to CaM in the presence of calcium ion and that it can be reversibly dissociated by EGTA (22). Accordingly, the specific interactions engendered from ternary complex calmodulin affinity chromatography were utilized to discriminate between the proteolytic products of riPLA 2 ␤ that retained calcium-activated CaM binding elements and those that did not. Two approaches were employed. In the first approach, partial tryptic digests of riPLA 2 ␤ (4 min) were prepared in the absence of calcium-activated CaM to generate (predominantly) the 68-kDa polypeptide cleaved at or near residue 630. Following the addition of calcium ion, the partially trypsinized mixture was loaded onto a CaM-agarose affinity column previously equilibrated with buffer containing 5 mM CaCl 2 . After washing, the column was eluted with buffer containing 10 mM EGTA. Western analysis of column eluates of CaCl 2 ) and CaM (ϳ9 g). Aliquots from each reaction were taken at the indicated times, and proteolysis was terminated by boiling each aliquot in SDS-PAGE loading buffer for 3 min. Proteolytic products of riPLA 2 ␤ (and undigested riPLA 2 ␤) were resolved by electrophoresis on 10% polyacrylamide gels followed by transfer to polyvinylidene difluoride membranes. Immunoreactive bands were detected by ECL utilizing immunoaffinity-purified polyclonal antibodies directed against the riPLA 2 ␤ holoprotein and an anti-rabbit IgG-horseradish peroxidase conjugate as described under "Experimental Procedures." The first lane is riPLA 2 ␤ alone incubated in buffer containing EGTA for 9 min at 22°C in the absence of trypsin (ϪT). Similar profiles were obtained in three independent preparations. Molecular masses of marker proteins are shown on the right in kilodaltons.
proteolytic fragments from the 4-min digestion reaction demonstrated that the overwhelming majority (ϳ90%) of the 68-kDa polypeptide did not bind to the affinity matrix, whereas Ͼ95% of the holoenzyme (83 kDa) avidly bound to the CaM column (Fig. 3A). Interestingly, the 40-kDa fragment (band E in Figs. 1 and 2B) corresponding to the C-terminal half of riPLA 2 ␤ containing the active site appeared to bind as tightly to the Ca ϩ2 ⅐CaM resin as did the 83-kDa holoprotein (Fig. 3A). The small residual amount of binding of the 68-kDa fragment could be due to the presence of iPLA 2 ␤ heteromultimers, which still possess at least one intact C-terminal portion of iPLA 2 ␤ (either as an intact 83-kDa component or as a noncovalently associated C-terminal fragment). To further address this issue, riPLA 2 ␤ was trypsinized for 20 min to eliminate (as much as possible) the 83-kDa holoprotein and then subjected to CaMagarose chromatography as described above. Under these conditions, virtually all of the 68-kDa fragment failed to bind to the column (Fig. 3B, upper panel). In addition, N-terminal fragments D and F were also not retained by CaM-agarose in the presence of calcium ion (Fig. 3B, upper panel), although untrypsinized riPLA 2 ␤ bound tightly to calmodulin-agarose in control experiments (Fig. 3B, lower panel). Notably, the 68-kDa fragment retained the ability to bind to ATP-agarose, demon-strating that its folding and structural integrity had not been significantly compromised during limited trypsinolysis (data not shown). We specifically point out that the present results do not exclude the possibility that some portions of the 68-kDa N-terminal polypeptide do contact calmodulin. However, the results do demonstrate that the C-terminal portion is both necessary (i.e. Ͼ 90% of the 68-kDa peptide was not bound by calmodulin) and sufficient (i.e. the 40-kDa C-terminal peptide bound in the presence of calcium ion and was released by EGTA) for binding of iPLA 2 ␤ to calmodulin in a calcium-dependent manner. Collectively, these results demonstrate that the major domain responsible for riPLA 2 ␤ binding to calciumactivated calmodulin is located between residues 620 and 752.
To identify specific portions of the 15-kDa C-terminal fragment (or potentially other regions) that interacted with CaM in a calcium-dependent manner, riPLA 2 ␤ was exhaustively trypsinized. SDS-PAGE of these samples demonstrated the completeness of trypsinolysis (Fig. 4A). RP-HPLC of the exhaustively digested riPLA 2 ␤ revealed the presence of ϳ44 peptide peaks (Fig. 4B, Load). When riPLA 2 ␤ peptides generated by exhaustive trypsinolysis were applied to CaM-agarose in the Aliquots from each reaction were taken at the indicated times, boiled in SDS-PAGE buffer, electrophoresed, and transferred to polyvinylidene difluoride membranes as described above. Immunoreactive bands were detected by ECL utilizing immunoaffinity-purified polyclonal antibodies directed against either the iPLA 2 ␤ N terminus (residues 23-42) (A) or the iPLA 2 ␤ C terminus (residues 731-744) (B) as described under "Experimental Procedures." The first lane is riPLA 2 ␤ alone incubated in buffer containing EGTA for 9 min at 22°C in the absence of trypsin (ϪT). Similar results were obtained in three independent preparations. Molecular masses of marker proteins are shown on the right in kilodaltons. presence of calcium ion, a single peptide (retention time ϭ 62 min) was absent in the void volume (Fig. 4B). Moreover, when peptides that bound to the CaM matrix were subsequently eluted with buffer containing EGTA, the same unique peptide (retention time ϭ 62 min) was selectively concentrated. Of the Ϸ44 peaks in the load and void fractions that did not interact with CaM (i.e. they eluted in the void volume), the peak at 62 min was the only peak enriched in the EGTA elute (Fig. 4B). The amino acid sequence of the unique calmodulin binding peptide was determined by automated Edman degradation and identified as the 12-amino acid fragment AWSEMVGIQYFR, which is within the previously identified 15-kDa C-terminal fragment from calmodulin footprinting experiments (Figs. 1 and 2). This peptide corresponds to tryptic cleavage between residues Arg 693 and Arg 705 .
To determine whether this peptide interacted with calmodulin in a calcium-dependent fashion, fluorescence resonance energy transfer experiments were performed. First, a peptide with this 12-amino acid sequence was synthesized and demonstrated to possess an elution profile by RP-HPLC similar to that of the peptide resulting from exhaustive trypsinolysis of riPLA 2 ␤ (Fig. 5). Next, the 12-amino acid synthetic peptide was incubated with dansyl-CaM and shown to interact in a calciumdependent manner as demonstrated by fluorescence energy transfer (Fig. 6). The increase in fluoresence and blue shift of the emission maxima of dansyl-CaM, which occurs only in the presence of calcium ion and the peptide, is indicative of their close association, as has been observed with other calcium-dependent CaM-binding peptides (24 -27).
To further substantiate the results from the partial trypsinolysis experiments, attempts were made to generate iPLA 2 ␤ deletion mutants lacking various portions of the C terminus. After expression, these constructs were found to have barely measurable levels of iPLA 2 ␤ protein mass (Ͻ100-fold protein mass relative to wild type). Western blot analyses demonstrated the presence of lower molecular mass bands presumably because of rapid proteolytic degradation of the deletion mutants lacking the C terminus (data not shown). Accordingly, we considered the possibility that strict conservation of the 15-kDa C-terminal region was necessary for appropriate folding and retention of protein conformation to resist proteolysis of the incompletely folded protein. One way to gain insight into the importance of specific regions of primary sequence in protein structure and function is to identify highly conserved areas of amino acid homology across species lines. Absolute complexity alignment analysis (29) demonstrated that the CaM-binding domain was the longest conserved region in the protein over four known species lines (Fig. 7). Thus, this region may be necessary for the proper folding and tertiary structure of iPLA 2 ␤, and its absence may make the protein more susceptible to Sf9 cell degradative proteases. DISCUSSION Intracellular phospholipases A 2 are comprised of two distinct families of proteins that catalyze the serine-mediated nucleophilic attack of ester linkages in cellular phospholipids resulting in the release of arachidonic acid from its endogenous phospholipid storage depots. One family, the cytosolic phospholipase A 2 family, of enzymes possesses a GXSGS sequence, FIG. 4. RP-HPLC analysis of exhaustive tryptic digests of recombinant iPLA 2 ␤ after CaM-agarose chromatography. Peptides from the trypsin digestion of riPLA 2 ␤ were prepared and subjected to CaM-agarose ternary complex affinity chromatography as described under "Experimental Procedures." A illustrates the completeness of tryptic digestion by SDS-PAGE. In B, peptides from exhaustive trypsinolysis were applied to a CaM-agarose affinity column in the presence of calcium ion. After washing the column, peptides possessing an affinity for calcium-activated calmodulin were eluted by application of buffer containing EGTA. Peptides in each fraction were concentrated, and prepared for RP-HPLC as described under "Experimental Procedures." RP-HPLC was performed on an RPC C2/C18 SC 2.1/10 RP-HPLC column, and absorbance of column eluates was monitored at 215 nm. The large peaks from 80 -95 min are not peptides and result from concentrating uv absorbing buffer contaminants during RP-HPLC. The peak indicated by the asterisk represents the unique peptide derived from riPLA 2 ␤ that bound to the calmodulin-agarose resin in a calciumdependent manner. whereas the other family, the iPLA 2 family, possesses a GX-STG sequence at their active site. The iPLA 2 family originated as an evolutionary distant archetype (i.e. iPLA 2 ␣) (30) that subsequently developed specialized domains to fulfill specific intracellular functions. For example, iPLA 2 ␤ contains eight N-terminal ankyrin repeat sequences (31) and is modulated by calmodulin (13), whereas iPLA 2 ␥ contains a C-terminal peroxisomal localization sequence (32). The present study demonstrates that the calmodulin-binding domain of iPLA 2 ␤ is comprised of multiple contact points in the 15-kDa C-terminal region. Calmodulin footprinting of riPLA 2 ␤ with trypsin dramatically diminished the rate of generation of the 68-kDa riPLA 2 ␤ proteolysis product (which lacks the C-terminal 15-kDa polypeptide) in the presence of calcium and CaM together but not with either alone. This strongly suggests that calciumactivated CaM directly binds to this region of iPLA 2 ␤, protecting it from proteolysis. Moreover, the large majority (ϳ90%) of the 68-kDa trypsinolysis product (lacking the 15-kDa C-terminus) failed to bind CaM in the presence of Ca ϩ2 (Fig. 3). Importantly, the calmodulin binding peptide identified after exhaustive trypsinolysis is also present within this 15-kDa fragment and directly interacts with CaM as demonstrated by fluorescent energy transfer experiments. Collectively, these results identify the calmodulin-binding domain in iPLA 2 ␤ as the 15-kDa C-terminal portion through both direct physical protein-protein interactions and the calcium-dependent functional association of this region with calmodulin.
The molecular mechanisms underlying the regulation of intracellular phospholipases A 2 has been an area of intense investigation. Alterations in cellular calcium ion flux are an integral part of signal transduction processes in most cell types. In the case of cytosolic phospholipase A 2 ␣, an internal C2 domain binds calcium and facilitates its translocation to specific membrane compartments after cellular activation (33)(34)(35). In prior studies, we have demonstrated that recombinant iPLA 2 ␤ reversibly binds to CaM in the presence of calcium ion and calcium-activated calmodulin modulates iPLA 2 ␤ enzymic activity (13). Calcium-dependent binding of CaM to iPLA 2 ␤ is an intrinsic property of this polypeptide, which was previously exploited to obtain highly purified riPLA 2 ␤ via CaM-agarose affinity chromatography (22). Thus, iPLA 2 ␤ differs from cytosolic phospholipase A 2 ␣ by requiring an exogenous protein, calmodulin, to integrate alterations in cellular calcium ion homeostasis during cellular activation with release of fatty acids and the generation of lysolipids. In the present study, multiple contact points in the 15-kDa C-terminal portion of iPLA 2 have been identified (Fig. 8) as important determinants of the calcium-dependent interaction of iPLA 2 ␤ with CaM. The demonstration of specific Ca ϩ2 -dependent CaM footprinting during limited trypsinolysis clearly identifies one site of calciumdependent binding of calmodulin at or near residue 630 in iPLA 2 ␤. Similar CaM footprinting experiments with myosin light chain kinase (36), calcineurin A (37), and with endothelial (38) and neuronal nitric oxide synthases (38 -40) have each provided insight into the domain structure and CaM binding sites of these proteins.
Remarkably, addition of calcium-activated CaM to riPLA 2 ␤ induced a conformational change near the active site as demonstrated by an increase in the kinetic rate of the appearance of band E in the presence of calcium and calmodulin together but not with either alone (Figs. 1 and 2B). Production of band E results from hydrolysis at a site ϳ200 amino acids from the CaM-binding domain. The most likely reason underlying this increased rate of production of band E is that this site interacts through space with the CaM-binding region. Accordingly, we propose a model in which the active site of iPLA 2 ␤ interacts with the CaM-binding domain (in the absence of CaM) leading to a catalytically competent enzyme, whereas the reversible disruption of this interaction by binding of CaM to the 15-kDa C-terminal region abrogates this interaction with resultant loss of enzyme activity. In this model, association of calciumactivated CaM to the C terminus of riPLA 2 ␤ putatively prevents important catalytic interactions between the CaM-binding domain and the active site. Results from site-directed mutagenesis near this region have suggested its importance for catalytic activity of iPLA 2 ␤. 2 Intriguingly, absolute complexity analysis of the four known iPLA 2 ␤ sequences (human, rat, mouse, and hamster) reveals that the amino acids residues between the 1-9-14 (vide intra) and IQ motifs represent one of the most highly conserved regions in the entire iPLA 2 ␤ protein (Fig. 7). Collectively, these results provide the first specific evidence of a calcium-activated CaM-induced conformational alteration of the active site of iPLA 2 ␤.
Proteins that interact with CaM have evolved to preserve critical residues necessary for the recruitment and binding of calcium-activated calmodulin. Based on the analysis of calmodulin binding sites (16,18,19), various CaM-binding sequences have been identified that, when modeled as an ␣-helix, form a predicted amphipathic structure with hydrophobic and basic amino acid residues positioned on opposite sides of the helix. Several calcium-dependent CaM-binding proteins (e.g. Ras-GRF (41), CDC25 Mm (42), and IRS-1 (43)) contain an IQ CaMbinding motif (IQXXXRGXXXR) (19,44). Notably, the 12-residue CaM-binding peptide (AWSEMVGIQYFR) identified in 2 D. J. Mancuso and R. W. Gross, unpublished observations. FIG. 7. Absolute complexity analysis of iPLA 2 ␤ of the four known sequences from different species. Amino acid sequences of iPLA 2 ␤ from hamster, rat, mouse, and human were aligned utilizing the AlignX program (Clustal W algorithm (47)) of Vector NTI Suite 6.0 (29). From this alignment, the sum of all pairwise residue substitution scores (using an identity score matrix) normalized by the total number of pairs were calculated for a given 20-amino acid residue window. A schematic of the iPLA 2 ␤ sequence is presented above with the ankyrin repeat, ATP, lipase, and putative CaM-binding domains as indicated.
FIG . 8. Amino acid sequences of the 1-9-14 and IQ motifs of iPLA 2 ␤. The upper sequence (residues 615-640) contains a putative 1-9-14 calmodulin-binding motif comprised of hydrophobic residues (underlined) with interspersed positively charged amino acids. This region is protected from trypsinolysis by calcium and calmodulin. The lower sequence (residues 691-716) includes the sequence of the CaM binding peptide (boxed) that overlaps with the IQ motif (defining residues are underlined). Positively charged residues are as indicated.
this study as an important determinant in CaM-iPLA 2 ␤ interactions possesses a shortened version of the first half of the consensus IQ motif (I 701 Q 702 XXR 705 ) (Fig. 8). The apparent close proximity of the tryptophan in the 12-mer peptide (Trp 695 in iPLA 2 ␤) to dansyl-calmodulin as demonstrated by fluorescence resonance energy transfer further substantiates the presence of a contact point between CaM and iPLA 2 ␤ mediated by this peptide. Interestingly, iPLA 2 ␤ contains arginine residues at positions 691 and 693 that contribute to the overall positive charge of this region and likely participate in electrostatic interactions with CaM that contribute to the stability and high affinity binding of the complex.
Analysis of the iPLA 2 ␤ sequence at the CaM protected site shows the presence of a cluster of positively charged residues ( 623 RKGQGNKVKK 632 ), one or more of which is likely a site of trypsinolysis (Fig. 8). Extensive digestion of iPLA 2 ␤ would be expected to destroy this site, thus explaining why it is not present in the RP-HPLC profile of the exhaustively digested protein loaded onto the CaM affinity column and absent in the EGTA Elute fraction (Fig. 4). Further analysis of this region indicates a 1-8-14 pseudo-consensus sequence (I 622 X-XXXXXXV 630 XXXXI 635 ϭ 1-9-14) in which the numbers refer to the relative position of hydrophobic residues within the sequence (19). Notably, Ile 622 and Ile 635 are separated by 12 amino acids, a distance that is believed to be critical for anchoring these residues to the two Ca ϩ2 -binding lobes of CaM (15,16). Some variability has been demonstrated in the location of the central hydrophobic residue of the 1-8-14 sequence, suggesting that the flexible central helix of CaM is able to accommodate minor alterations in the position of this residue (45,46). The high net positive charge (ϩ5) of this putative 1-9-14 motif in iPLA 2 ␤ and the positioning of these positively charged residues (Arg 623 , Lys 624 , Lys 629 , Lys 631 , and Lys 632 ) adjacent to two of the hydrophobic residues (Ile 622 and Val 630 ) may function to stabilize the iPLA 2 ␤⅐Ca 2ϩ ⅐CaM complex through electrostatic charge pairing with the two glutamate clusters of calcium-activated CaM (14,46).
In conclusion, these results illustrate: 1) the direct physical interaction of calmodulin with riPLA 2 ␤ within the 15-kDa Cterminal region; 2) calcium-activated calmodulin-induced conformational alterations near the iPLA 2 ␤ active site; and 3) the conservation of hydrophobic and positively charged motifs identified in other proteins present in the calmodulin binding motif of iPLA 2 ␤. It is hoped that identification of the domain of riPLA 2 ␤ that interacts with CaM and the demonstration of calmodulin-induced conformational alterations of the active site will facilitate the development of therapeutic strategies that will modulate the activity of iPLA 2 ␤ in disease processes.