Proteolysis of Apolipoprotein A-I by Secretory Phospholipase A2

Background: During inflammation, increased levels of secretory phospholipase A2 (sPLA2) impact high density lipoprotein (HDL) cholesterol plasma concentration. Results: sPLA2 cleaves apolipoprotein A-I (apoA-I) in the central and C-terminal regions. Conclusion: sPLA2 selectively targets lipid-free apoA-I, whose levels are increased by HDL remodeling during inflammation. Significance: In acute and chronic (atherosclerosis) inflammation, the proteolytic action of sPLA2 may accelerate clearance of apoA-I, thereby reducing HDL biogenesis. In the acute phase of the inflammatory response, secretory phospholipase A2 (sPLA2) reaches its maximum levels in plasma, where it is mostly associated with high density lipoproteins (HDL). Overexpression of human sPLA2 in transgenic mice reduces both HDL cholesterol and apolipoprotein A-I (apoA-I) plasma levels through increased HDL catabolism by an unknown mechanism. To identify unknown PLA2-mediated activities on the molecular components of HDL, we characterized the protein and lipid products of the PLA2 reaction with HDL. Consistent with previous studies, hydrolysis of HDL phospholipids by PLA2 reduced the particle size without changing its protein composition. However, when HDL was destabilized in the presence of PLA2 by the action of cholesteryl ester transfer protein or by guanidine hydrochloride treatment, a fraction of apoA-I, but no other proteins, dissociated from the particle and was rapidly cleaved. Incubation of PLA2 with lipid-free apoA-I produced similar protein fragments in the range of 6–15 kDa, suggesting specific and direct reaction of PLA2 with apoA-I. Mass spectrometry analysis of isolated proteolytic fragments indicated at least two major cleavage sites at the C-terminal and the central domain of apoA-I. ApoA-I proteolysis by PLA2 was Ca2+-independent, implicating a different mechanism from the Ca2+-dependent PLA2-mediated phospholipid hydrolysis. Inhibition of proteolysis by benzamidine suggests that the proteolytic and lipolytic activities of PLA2 proceed through different mechanisms. Our study identifies a previously unknown proteolytic activity of PLA2 that is specific to apoA-I and may contribute to the enhanced catabolism of apoA-I in inflammation and atherosclerosis.

In the acute phase of the inflammatory response, secretory phospholipase A 2 (sPLA 2 ) reaches its maximum levels in plasma, where it is mostly associated with high density lipoproteins (HDL). Overexpression of human sPLA 2 in transgenic mice reduces both HDL cholesterol and apolipoprotein A-I (apoA-I) plasma levels through increased HDL catabolism by an unknown mechanism. To identify unknown PLA 2 -mediated activities on the molecular components of HDL, we characterized the protein and lipid products of the PLA 2 reaction with HDL. Consistent with previous studies, hydrolysis of HDL phospholipids by PLA 2 reduced the particle size without changing its protein composition. However, when HDL was destabilized in the presence of PLA 2 by the action of cholesteryl ester transfer protein or by guanidine hydrochloride treatment, a fraction of apoA-I, but no other proteins, dissociated from the particle and was rapidly cleaved. Incubation of PLA 2 with lipid-free apoA-I produced similar protein fragments in the range of 6 -15 kDa, suggesting specific and direct reaction of PLA 2 with apoA-I. Mass spectrometry analysis of isolated proteolytic fragments indicated at least two major cleavage sites at the C-terminal and the central domain of apoA-I. ApoA-I proteolysis by PLA 2 was Ca 2؉ -independent, implicating a different mechanism from the Ca 2؉ -dependent PLA 2 -mediated phospholipid hydrolysis. Inhibition of proteolysis by benzamidine suggests that the proteolytic and lipolytic activities of PLA 2 proceed through different mechanisms. Our study identifies a previously unknown proteolytic activity of PLA 2 that is specific to apoA-I and may contribute to the enhanced catabolism of apoA-I in inflammation and atherosclerosis.
Secretory phospholipase A 2 (sPLA 2 ) 3 is an acute phase protein whose concentration in plasma and inflammatory fluids can increase several hundred-fold during the acute phase response (1). The main reactivity of sPLA 2 is directed toward hydrolysis of the sn-2 ester bond of the glyceroacyl phospholipids present in lipoproteins and cell membranes. This reactivity induces structural and functional changes in lipoproteins and membranes and generates lysophospholipids and non-esterified fatty acids, which are proinflammatory signal molecules (2,3).
Atherosclerosis is increasingly recognized as a localized inflammatory condition within the vessel wall (7,8). The observation that patients with chronic inflammatory diseases are at higher risk of cardiovascular events confirms the inflammatory nature of atherosclerosis (9,10). Therefore, the proinflammatory properties of sPLA 2 enzymes and their localization in atherosclerotic lesions suggest a direct mechanistic link between the activity of sPLA 2 in the arterial subendothelial space and atherosclerosis development. Although recent human genetic association studies failed to detect a correlation between high sPLA 2 levels and increase in cardiovascular events (11), this link is supported by data from human sPLA 2 transgenic mice, in which high density lipoprotein cholesterol (HDL-C) and apolipoprotein A-I (apoA-I) plasma levels are significantly decreased compared with wild-type mice, and atherosclerotic lesions develop even on a chow diet (12)(13)(14)(15).
High HDL-C levels correlate with low incidence of atherosclerosis and are believed to be a marker of an efficient reverse cholesterol transport (16). In particular, the ability of HDL to mediate reverse cholesterol transport from lipid-laden macrophages in the vascular wall and transport cholesterol back to the liver for excretion is considered to be one of the mechanisms underlying HDL antiatherogenic function (17). Furthermore, HDL reduces atherosclerosis through its general antioxidant and anti-inflammatory action (18). In plasma, HDL undergoes a continuous remodeling process mediated by a variety of plasma factors. This remodeling determines the plasma concentration and subpopulation distribution of HDL (19). The effect of acute and chronic inflammation on HDL remodeling is of great importance because it may impact both the polydisperse particles and the equilibrium between HDL-associated and lipidfree apolipoproteins (20). Insights into the regulation of HDL remodeling have been obtained by several in vitro studies via chemical and thermal perturbation techniques (21)(22)(23)(24)(25). These approaches have established that dissociation of lipid-free apoA-I always occurs during HDL remodeling events. The ability of cholesteryl ester transfer protein (CETP) and phospholipid transfer protein to remodel HDL with dissociation of lipid-free apoA-I corroborates this notion (26), which bears important physiological implications (27)(28)(29). Although lipidfree apoA-I accounts for less than 10% of total plasma apoA-I (30,31), this form of the protein is central to HDL biogenesis, which requires interaction of lipid-free apoA-I with the ABCA1 (ATP-binding cassette A1) cell membrane transporter (27,32). Progressive lipidation of apoA-I via ABCA1 generates nascent HDL and efficiently reduces the plasma concentration of lipidfree apoA-I, which is otherwise easily catabolized through glomerular filtration. Therefore, plasma levels of apoA-I are ultimately determined by the rate at which lipid-free apoA-I is produced (taking into account its generation through dissociation from HDL) and catabolized (33). Interestingly, HDL-C and apoA-I levels are significantly decreased in both acute and chronic inflammatory states (15) at least in part due to accelerated apoA-I catabolism (13).
Although inflammation includes a wide variety of metabolic changes, two inflammation-dependent events that may affect HDL and apoA-I catabolism are the increase in plasma levels of serum amyloid A (SAA) protein and secretory (non-pancreatic) PLA 2 (12,13,34). Due to its high lipid binding affinity, SAA can displace up to 80% of apoA-I from HDL (35). In the acute phase response, when plasma levels of SAA rise severalfold, the protein composition of HDL is altered as lipid-free apoA-I is displaced form HDL by SAA (34). Increased catabolism of lipidfree apoA-I can than occur through normal glomerular filtration.
Studies in humans and animals have suggested that factors that affect HDL lipid composition modulate apoA-I catabolism (36). Secretory PLA 2 can remodel HDL in vitro by hydrolyzing HDL phospholipids, thereby reducing HDL particle size (19). However, unlike binding of SAA, sPLA 2 -mediated HDL remodeling does not generate lipid-free apoA-I (19,37). Hence, the mechanism that may link increased levels of sPLA 2 and decreased plasma levels of HDL-C and apoA-I in inflammation is unknown. The results reported here reveal a previously undetected selective proteolytic activity of sPLA 2 on lipid-free apoA-I. Our study suggests that in inflammation states, increased catabolism of apoA-I and, ultimately, reduction of HDL-C plasma levels may be mediated by direct interaction of sPLA 2 with lipid-free apoA-I. This novel function of sPLA 2 has potentially important implications for atherogenesis. In atherosclerotic lesions, high levels of sPLA 2 , in concert with increased levels of lipid-free apoA-I (30), could contribute to atherosclerosis progression by limiting the availability of functional antiatherogenic lipid-free apoA-I in the subendothelial space.

EXPERIMENTAL PROCEDURES
Materials-Porcine pancreatic PLA 2 (PP-PLA 2 ), bovine pancreatic PLA 2 (BP-PLA 2 ), bee venom PLA 2 (BV-PLA 2 ), and essentially fatty acid-free BSA were obtained from Sigma. Recombinant human secretory PLA 2 -IIA (Ͼ95% purity) was a generous gift from BioVendor Laboratories (Czech Republic). Unlike PP-PLA 2 , both BP-PLA 2 and BV-PLA 2 (i) are not subjected to exogenous trypsin digestion during extraction from the natural source (Sigma Technical Information Service) and (ii) are lyophilized, rather than prepared as a 3.2 M ammonium sulfate suspension. For PP-PLA 2 , the enzyme solution was extensively dialyzed against 10 mM Tris, pH 7.5 (buffer A), at 4°C and used within a week. Lyophilized BP-PLA 2 and BV-PLA 2 were solubilized in buffer A and stored at 4°C for no more than 1 week before use. Recombinant PLA 2 -IIA was reconstituted according to the manufacturer's instructions. Briefly, the lyophilized powder was dissolved in 0.1 M acetate buffer, pH 4.0, and diluted in buffer A to raise the pH to 7.5. The integrity and purity of PP-PLA 2 was checked by SDS-PAGE and matrix-assisted laser-desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) (data not shown), and no contamination was detectable. As an additional purity test, commercial PP-PLA 2 was purified by preparative size exclusion chromatography (SEC) on a Superdex 200 column. The lipolytic and proteolytic activities of the SEC-purified sample were not different from those of the unpurified sample. Therefore, it can be excluded that the observed activities of the commercial PP-PLA 2 are due to residual trace amounts of contaminants. For the experiments reported in this paper, commercial PP-PLA 2 was used without further purification.
Isolation of Human Very Low Density Lipoprotein (VLDL), HDL, and ApoA-I-Human lipoproteins were isolated from plasma of healthy volunteers. The blood was donated at the blood bank (Research Blood Components, LLC) according to the rules of the institutional review board and with written consent from the donors. Lipoproteins were isolated from fresh EDTA-treated plasma from a single donor by density gradient ultracentrifugation in the density range 0.94 -1.006 g/ml for VLDL and 1.063-1.21 g/ml for HDL, as described (38). Different lipoprotein fractions migrated as a single band on agarose gel electrophoresis. The density fractions containing VLDL and HDL were dialyzed against 10 mM sodium phosphate, pH 7.5 (buffer B), degassed, and stored in the dark at 4°C. The lipoprotein stock solutions were used (for experiments or for extraction of apoA-I from HDL) within 2-3 weeks, during which no protein degradation or changes in the lipoprotein electrophoretic mobility were detected by SDS-PAGE and agarose electrophoresis. ApoA-I was isolated and purified from HDL stock solutions as described (39). Purified apoA-I was refolded from 6 M GdnHCl after extensive dialysis against buffer B, stored at 4°C, and used within 1 month (25).
Expression and Purification of ApoA-I Variants-The apoA-I variants in which residues 134 -145 were removed (apoA-I: ⌬134 -145) or all four native tryptophans were mutated to phenylalanines (⌬W-apoA-I) were produced using primer-directed PCR mutagenesis as described (40). The mutations were verified by dideoxy automated fluorescent sequencing. The protein variants were expressed in BL21 Escherichia coli (Agilent Technologies) and purified by nickel affinity chromatography (40). Protein purity (Ͼ95%) was confirmed by SDS-PAGE and mass spectrometry analyses.
PLA 2 Reaction with HDL-HDL was dialyzed against buffer A containing CaCl 2 (final concentration, 2 mM). HDL (5 mg, in total proteins) and PP-PLA 2 (25 g, 1.8 M) were incubated for 1 h at 37°C (41). The reaction was stopped by cooling on ice after adding an excess of EDTA (final concentration, 15 mM) to chelate Ca 2ϩ , an essential co-factor for PLA 2 lipolytic activity (42). These samples are termed PP-PLA 2 -HDL throughout this work. In control experiments, free fatty acids (FFAs) generated in PP-PLA 2 -HDL by the lipolytic action of PLA 2 on HDL phospholipids were removed by performing the incubation in the presence of essentially fatty acid-free bovine serum albumin (BSA) (final concentration, 2% (w/v)). The incubation conditions were the same as described before (41) (i.e. 1 h at 37°C and then 15 mM EDTA on ice). After incubation, FFAs were removed from PP-PLA 2 -HDL by separating the FFA-loaded BSA fraction by SEC on a Superose-6 10/300 GL column (elution: 10 mM PBS, 15 mM EDTA, pH 7.5, at 0.5 ml/min) to yield (PP-PLA 2 -HDL)-FFA. Complete removal of BSA from (PP-PLA 2 -HDL)-FFA was confirmed by SDS-PAGE analysis (data not shown). (PP-PLA 2 -HDL)-FFA was dialyzed against buffer B for further studies.
Remodeling of HDL by Guanidine Treatment-Intact HDL and PP-PLA 2 -HDL were incubated with 3 M GdnHCl at 37°C for 3 h and then dialyzed extensively against buffer B and stored at 4°C.
Remodeling of HDL by CETP-Plasma purified CETP was a generous gift from Prof. Kerry-Ann Rye (Sydney, Australia). Intact HDL and PP-PLA 2 -HDL (final 1 mg/ml, in total proteins) were incubated with VLDL (final 5 mM, in triglyceride) and CETP (final 5 units/ml) in buffer A at 37°C for 24 h. The reaction was terminated by cooling samples on ice to yield the samples termed CETP-HDL and CETP-PP-PLA 2 -HDL, respectively. The different components in the two samples were separated by SEC with a Superose-6 10/300 GL column (elution: 10 mM PBS, pH 7.5, at a flow rate of 0.5 ml/min). The isolated fractions were concentrated and used for further biochemical characterization.
PLA 2 Reaction with Lipid-free Proteins-Lipid-free proteins (0.5 mg) were incubated with PP-PLA 2 (2.5 g and 0.18 M) (final enzyme/substrate ratio 1:200 (w/w)) at 37°C for 1 h in the presence or absence of 15 mM EDTA. This concentration of PLA 2 is comparable with the plasma concentration of the enzyme under chronic inflammatory conditions (43). The reaction was terminated by freezing the sample until further use. For immediate electrophoresis analysis, reactions were stopped by adding SDS loading buffer and boiling for 10 min before loading the samples on gel. Because Ca 2ϩ is an essential co-factor for the lipolytic activity of PLA 2 (42), to inhibit Ca 2ϩ -dependent lipolysis in some experiments, apoA-I was incubated with PLA 2 in the presence of an excess (relative to Ca 2ϩ ) of EDTA (15 mM). These samples are termed "inactive PLA 2 ." Effect of the Protease Inhibitor Benzamidine on PLA 2 Activity-PP-PLA 2 was preincubated in the presence of benzamidine (final concentration, 2 mM) at room temperature for 15 min. Benzamidine-treated PP-PLA 2 was then incubated with lipidfree apoA-I as described above.
Non-denaturing Gradient Gel Electrophoresis (NDGGE)-Novex TM 4 -20% Tris-glycine gels (Invitrogen) were loaded with 3 g of protein/lane and run to termination at 1,500 V-h under non-denaturing conditions in Tris-glycine buffer. The gels were stained with Denville Blue protein stain (Denville Scientific).
SDS-PAGE-Novex TM 4 -20% Tris-glycine gels (Invitrogen) were loaded with 5 g of protein/lane and run at 200 V for 1 h under denaturing conditions in SDS-Tris-glycine buffer. The gels were stained with Denville Blue protein stain (Denville Scientific).
Size Exclusion Chromatography-SEC was performed with a Superdex 200 prep grade XK 16/100 column and/or a Superose 6 10/300 GL column (GE Healthcare) controlled by an AKTA UPC 10 FPLC system (GE Healthcare). Elution by PBS (10 mM sodium phosphate, 150 mM NaCl, pH 7.5) was carried out at flow rates of 1.0 ml/min and 0.5 ml/min for Superdex 200 and Superose 6 columns, respectively.
MALDI-TOF MS-MALDI-TOF MS spectra were recorded on a Reflex-IV spectrometer (Bruker Daltonics, Billerica, MA) equipped with a 337-nm nitrogen laser. The instrument was operated in the positive ion reflection mode at 20-kV accelerating voltage with time lag focusing enabled. Calibration was performed using a standard calibration mixture containing oxidized B-chain of bovine insulin, equine cyctochrome c, equine apomyoglobin, and bovine serum albumin in linear mode. The matrix, cyano-4-hydroxycinnamic acid (␣-cyano, M r 189), was prepared as a saturated solution in 70% acetonitrile and 0.1% trifluoroacetic acid in water.
Liquid Chromatography Mass Spectrometry (LC-MS)-Intact apoA-I and PP-PLA 2 -apoA-I were prepared as described above and subjected to SDS-PAGE (Fig. 4B). Gel fragments corresponding to the protein bands stained with Denville Blue protein stain were excised from the gel slab and digested overnight with trypsin (0.02 g/l; Promega) as described (44). Tryptic digests were analyzed by nanoflow HPLC microelectrospray ionization on a Thermo Scientific LTQ Orbitrap XL. Samples were first trapped on a trap column (100 m ϫ 5 cm; Polymicro Technologies, Phoenix, AZ) packed with Poros 10R2 (Applied Biosystems, Foster City, CA) and subsequently eluted onto the analytical column (75-m inner diameter ϫ 10-cm fused silica analytical column, self-packed with Magic C18AQ, 5 m; Michrom BioResources, Auburn, CA). Elution was performed at 200 nl/min with the following gradient: 99% to 30% Buffer A in 25 min and then 30% to 10% Buffer A in 30 min and re-equilibration to 99% Buffer A in 40 min (Buffer A: 0.1% formic acid in water; Buffer B: 0.1% formic acid in acetonitrile). The column outflow was ionized and sprayed into the LTQ Orbitrap by way of a 8-m SilicaTip (New Objective) and Picoview source (New Objective). Spectra were acquired in a data-dependent mode throughout the gradient, and a full MS scan in the Orbi-trap analyzer was obtained, followed by seven subsequent MS/MS scans in the ion trap based on the seven most intense peaks identified in the full scan. Collision-induced dissociation fragmentation was achieved with a collision energy of 35%, a capillary temperature of 150°C, and an electrospray voltage of 1.9 kV. Once an MS/MS spectrum was obtained two times, it was put on the exclusion list for 3 min to allow for lower intensity peptides to be analyzed. Data were analyzed using the Mascot algorithm by searching against the updated non-redundant database from NCBI.
Gel electrophoresis and SEC data reported in this paper are representative results of at least three independent experiments. Mass spectrometry data are the average of 3-5 independent experiments.

PLA 2 -mediated Hydrolysis of HDL under HDL-destabilizing
Conditions-During the acute and chronic phase of inflammation, the plasma concentration of acute phase sPLA 2 rises sharply (1). To investigate the effect of this increased load of sPLA 2 on HDL and its resident proteins, we incubated HDL with PP-PLA 2 (PP-PLA 2 -HDL), as described under "Experimental Procedures." Furthermore, to mimic HDL remodeling that occurs in vivo, PP-PLA 2 -HDL was incubated with CETP or was exposed to chemical denaturation. Both treatments lead to HDL remodeling with concomitant release of lipid-free proteins (21,45).
First, HDL was incubated with PP-PLA 2 for 1 h at 37°C. Under these conditions, nearly 10% of phospholipids were converted into lysophosphatidylcholine and FFA, as reported previously (41). By NDGGE and SEC analyses of PP-PLA 2 -HDL, surface hydrolysis of HDL phospholipids by PP-PLA 2 produced a reduction in HDL particle size without dissociation of apoA-I from HDL (Fig. 1, A and B). Furthermore, no protein degradation was observed by SDS-PAGE (Fig. 1C). MALDI-TOF MS analysis of PP-PLA 2 -HDL confirmed the presence of only fulllength apoA-I, apoA-II, and apoC proteins (Table 1).
To investigate whether the presence of PLA 2 on HDL affects the products of HDL remodeling by chemical denaturation, we incubated HDL and PP-PLA 2 -HDL in 3 M GdnHCl, as described under "Experimental Procedures." After guanidine removal by extensive dialysis, the relative NDGGE mobilities of the dissociated proteins in guanidine-treated HDL (Gd-HDL) 4 and Gd-PP-PLA 2 -HDL were significantly different (Fig. 1D), indicating that protein modifications must have occurred during guanidine treatment in the presence of PP-PLA 2 . Interestingly, a new SEC peak at higher elution volumes (V e Ͼ 125 ml) than full-length lipid-free apoA-I (V e ϭ 125 ml; Fig. 4A) was present in Gd-PP-PLA 2 -HDL (Fig. 1E). Consistent with this observation, SDS-PAGE analysis of Gd-PP-PLA 2 -HDL revealed protein bands in the 15 to 5 kDa size range that were not present in Gd-HDL (Fig. 1F).
Isolation and Characterization of Protein Fragments-To characterize the PP-PLA 2 -mediated modifications of HDL proteins, Gd-HDL and Gd-PP-PLA 2 -HDL were separated by SEC in two and three fractions, respectively ( Fig. 2A). By thin layer chromatography, HDL lipids were detectable only in fraction 1 (Fr1) of Gd-HDL and Gd-PP-PLA 2 -HDL (data not shown). NDGGE of the SEC fractions confirmed that no lipid-free proteins were present in Fr1 of both Gd-HDL and Gd-PP-PLA 2 -HDL, whereas no HDL-size-like particles were detected in Fr2 or Fr3 (Fig. 2B). These analyses support the conclusion that Fr1 contains HDL-associated proteins, whereas Fr2 and any fraction eluting at volumes higher than Fr2 are predominantly composed of lipid-free proteins. By SDS-PAGE analysis, bands corresponding to full-length apoA-I and apoA-II occurred in Fr1 of Gd-HDL, whereas only one band, corresponding to fulllength lipid-free apoA-I, was present in Fr2 (Fig. 2C). Results for Gd-PP-PLA 2 -HDL were more complex (Fig. 2D). The HDLassociated proteins in Fr1 were full-length apoA-I, full-length apoA-II, and a smaller protein fragment with electrophoretic mobility corresponding to ϳ10 -12 kDa. Fr2 consisted of fulllength apoA-I, a band with mobility consistent with the size of full-length apoA-II, and at least two protein fragments within ϳ15-10 kDa. In Fr3, no full-length apoA-II and almost no fulllength apoA-I but several protein fragments that ranged from ϳ26 to 5 kDa were detected.
The molecular weight of the protein fragments in guanidinetreated samples was estimated by MALDI-TOF MS analysis. Although only apoA-I and apoA-II were detected in a stained SDS-polyacrylamide gel (Fig. 2, C and D), intact forms of all major HDL proteins (apoA-I, apoA-II, and apoCs) were identified in intact HDL, PP-PLA 2 -HDL, and Gd-PP-PLA 2 -HDL (total sample) ( Table 1). Upon SEC fractionation, masses corresponding to full-length apoA-II and apoCs were detected only in Fr1 of Gd-PP-PLA 2 -HDL (Table 1). Importantly, by Western blot analysis of Gd-PP-PLA 2 -HDL, only one band compatible with the size of full-length apoA-II (ϳ17 kDa) was detected by anti-apoA-II antibodies (data not shown). The absence of anti-apoA-II reactive bands for sizes Ͻ17 kDa suggests that the protein bands observed by SDS-PAGE in the 15 to 5 kDa molecular mass range in Fr2 and Fr3 are lipid-free proteolytic fragments of apoA-I.
As shown in Fig. 1C, no apoA-I proteolysis is observed in PP-PLA 2 -HDL in the absence of guanidine treatment. FFAs produced during PLA 2 -mediated lipolysis can inhibit the lipolytic action of the enzyme (46). To test whether the absence of 4 Samples generated by guanidine treatment are indicated by the prefix "Gd-" throughout.  (Table  1). This observation suggests that a "lipid-free form" of PLA 2 mediates proteolysis of lipid-free apoA-I in solution rather than on the surface of the HDL particles.
Furthermore, analysis of two major HDL subclasses isolated by density gradient ultracentrifugation, HDL 2 and HDL 3 , showed similar protein fragments upon guanidine treatment of PP-PLA 2 -HDL (data not shown), suggesting that PLA 2 -medi-

TABLE 1 MALDI-TOF MS analysis of intact HDL, PP-PLA 2 -HDL, and HDL and PP-PLA 2 -HDL after incubation in the presence of GdnHCl (Gd-HDL and Gd-PP-PLA 2 -HDL)
For Gd-samples, analyses of total samples and SEC-separated fractions are reported. ated proteolysis of apoA-I does not depend on HDL particle size, HDL protein or lipid composition, or the conformation of apoA-I on HDL. PLA 2 -mediated Proteolysis of ApoA-I upon HDL Destabilization by CETP-To test whether a physiologically relevant process that mediates HDL destabilization and lipid-free apolipoprotein release from HDL promotes PLA 2 -dependent proteolysis as observed in Gd-PP-PLA 2 -HDL, PP-PLA 2 -HDL was incubated with CETP as described under "Experimental Procedures." Consistent with previous reports (45), CETP promoted the release of lipid-free apoA-I form HDL (Fig. 3), but no apoA-I proteolysis was observed under these conditions. In contrast, in the presence of PP-PLA 2 , proteolysis was observed in the lipid-free protein fraction (Fr2) of CETP-PP-PLA 2 -HDL (Fig. 3B). Interestingly, the proteolytic patterns observed in these reaction conditions and in Gd-PP-PLA 2 -HDL were similar (Figs. 2D and 3B).

Molecular mass
Upon guanidine treatment of PP-PLA 2 -HDL, PP-PLA 2 was detected only as non-HDL-associated protein in SEC Fr3 of Gd-PP-PLA 2 -HDL (Table 1). To test whether a similar HDL remodeling occurs by action of CETP, the SEC fractions from CETP-PP-PLA 2 -HDL were analyzed by MALDI-TOF MS (Table 2). Interestingly, PP-PLA 2 was detected only in the HDL-associated fraction (Fr1). Therefore, not only lipid-free but also HDL surface-bound PLA 2 can cleave apoA-I, with similar substrate specificity.
In conclusion, destabilization of HDL by chemical or enzymatic means, in the presence of PP-PLA 2 , promotes proteolysis of apoA-I. This proteolytic activity can be mediated by both lipid-free and HDL-associated PLA 2 .

Lipid-free ApoA-I Is the Substrate of PLA 2 -mediated
Proteolysis-The proteolytic action of PLA 2 on apoA-I occurred exclusively after PP-PLA 2 -HDL destabilization by guanidine or CETP treatments, which promote dissociation of lipid-free apoA-I from the HDL particle. We hypothesized that, in these HDL-destabilizing conditions, lipid-free rather than HDL-associated apoA-I is the substrate of PLA 2 -mediated proteolysis. To test this hypothesis, purified lipid-free apoA-I was incubated with PP-PLA 2 under the same conditions used for plasma HDL to generate PP-PLA 2 -apoA-I. By SEC analysis, lipid-free apoA-I eluted as a single peak at V e ϭ 125 ml, whereas no chromatographic signal was detectable in the 120 -140-ml range (Fig. 4A). In the chromatogram of PP-PLA 2 -apoA-I (Fig.  4A), the intensity of the peak corresponding to full-length lipidfree apoA-I was reduced, compared with pure lipid-free apoA-I. Importantly, a new protein fraction eluted at V e ϳ135 ml. This elution volume is similar to the one observed in the SEC chromatogram of Gd-PP-PLA 2 -HDL (Fig. 1B). Furthermore, proteolytic fragments similar to those present in Gd-PP-PLA 2 -HDL (Fig. 2C, total and Fr2 ϩ 3) were observed in PP-PLA 2 -apoA-I by SDS-PAGE analysis (Fig. 4B). When lipidfree apoA-I was incubated in the absence of PP-PLA 2 for the same time period, no proteolysis was detected by SDS-PAGE analysis (Fig. 4B, apoA-I control). Therefore, lipid-free apoA-I is the substrate of the proteolytic action of PLA 2 . To further confirm this conclusion, the lipid-free protein fraction produced by GdnHCl treatment of HDL (Gd-HDL-Fr2; Fig. 2A) was incubated in the presence of PP-PLA 2 under the same experimental conditions used for lipid-free apoA-I. Notably, PLA 2 -mediated proteolysis of apoA-I in PP-PLA 2 -Gd-HDL-Fr2 was similar to the proteolytic pattern observed for lipid-free apoA-I (data not shown).
Proteolysis and Lipolysis Mechanisms-To test whether the mechanisms of PLA 2 -mediated proteolysis of apoA-I and of the normal lipolytic activity of the enzyme are related, PP-PLA 2 was incubated with apoA-I in the presence of EDTA (inactive PP-PLA 2 ). EDTA inhibits the lipolytic activity of PLA 2 by chelating Ca 2ϩ , a co-factor essential for PLA 2 -mediated lipolysis (42). The apoA-I proteolytic pattern generated by inactive-PP-PLA 2 was not significantly different from that observed in the absence of EDTA (Fig. 4B), indicating that the proteolytic activity of PP-PLA 2 is not Ca 2ϩ -dependent.
When lipid-free apoA-I and PP-PLA 2 were incubated in the presence of the protease inhibitor benzamidine, apoA-I proteolysis was completely suppressed (Fig. 4B). Notably, benzamidine had no effect on the lipolytic activity of PLA 2 , as confirmed by the reduction in HDL size upon incubation of HDL with PP-PLA 2 and benzamidine (data not shown). Taken together, these results indicate that the proteolytic and lipolytic activities of PLA 2 proceed through different mechanisms.
Analysis of Proteolyzed ApoA-I-The protein fragments generated during incubation of lipid-free apoA-I with PP-PLA 2 for 1 h (PP-PLA 2 -apoA-I) were separated by SDS-PAGE (Fig. 4B). The gel bands (numbered 1-6) were excised from the gel and digested with trypsin, and the tryptic digest was analyzed by LC-MS (Table 3). The mass of the main protein component in the undigested samples was determined by MALDI-TOF MS (Table 3).  Notably, when an apoA-I variant in which residues 134 -145 were removed (apoA-I:⌬134 -145) was incubated with PP-PLA 2 , only bands 1 and 2 were detected by SDS-PAGE analysis of PP-PLA 2 -apoA-I:⌬134 -145 (Fig. 4C). Thus, deletion of residues 134 -145 completely inhibited cleavage at Lys 133 , which accounts for the absence of band 3 (fragment 1-133) and band 5 (fragment 134 -230). Furthermore, absence of bands 4 and 6 in PP-PLA 2 -apoA-I:⌬134 -145 suggests that fragments 19 -126 and 178 -230 are produced by secondary proteolytic reactions after the protein structure is destabilized by the primary cuts at Lys 133 and Leu 230 . Taken together, these results suggest that PP-PLA 2 has specific proteolytic activity on two regions of lipid-free apoA-I: the C-terminal region (positions 191 and 230) and a segment in the protein central domain (position 133).
Structural Requirements at ApoA-I C-terminal Domain for PLA 2 -mediated Proteolysis-The C-terminal domain of apoA-I is essential not only for lipid binding during HDL assembly (47), but it also drives apoA-I self-association in the lipid-free state (40). To investigate the role of the structure of the C-terminal domain of apoA-I in PP-PLA 2 -mediated proteolysis, a tryptophan-null apoA-I variant (⌬W-apoA-I) that exhibits unique self-association behavior was incubated with PP-PLA 2 . The ␣-helical content of the higher order self-association state of ⌬W-apoA-I (⌬W[high]) is significantly higher compared with WT-apoA-I (85 versus 60%, respectively) (40). Intermolecular protein-protein interactions in ⌬W[high] drive the folding and thermodynamic stability of the otherwise largely unstructured C-terminal domain (48) into an ␣-helical structure that accounts for the overall increase in protein ␣-helical content. Thermal treatment of ⌬W[high] dissociates the higher order self-associated species to tetramers (60°C, ⌬W[high]60C) and monomers (90°C, ⌬W[high]90C) and proportionally reverses the ␣-helical content of ⌬W-apoA-I to that of WT-apoA-I.
Interestingly, upon incubation of ⌬W[high] with PP-PLA 2 , none of the proteolysis bands observed with WT-apoA-I were detected by SDS-PAGE analysis (Fig. 5A). In contrast, some of the bands previously identified as the products of proteolytic cleavage at sites in the central domain of the protein were observed in PP-PLA 2 -⌬W[high]60C. Importantly, the proteolytic patterns of monomeric ⌬W-apoA-I (PP-PLA 2 -⌬W[high]90C) and wild-type apoA-I (PP-PLA 2 -apoA-I) were indistinguishable (Figs. 5A and 4B, respectively). Taken together, the ⌬W-apoA-I data and the results from Gd-and CETP-PP-PLA 2 -HDL indicate that when the C-terminal domain of apoA-I is engaged in either protein-lipid or protein-protein interactions, the proteolytic activity of PLA 2 is inhibited.
To further investigate whether modifications at the C-terminal of apoA-I can affect the proteolytic activity of PLA 2 at other apoA-I sites, we used SEC to isolate the 1-230 fragment of apoA-I (apoA-I:⌬231-243) generated by a first reaction of fulllength apoA-I with PLA 2 . The purity and mass of this fragment was confirmed by MALDI-TOF. When apoA-I:⌬231-243 was incubated in the presence of PP-PLA 2 (PP-PLA 2 -apoA-I:231-243), no proteolysis was observed (Fig. 5B). These results indicate that the two cleavage sites at Lys 133 and Leu 230 are not independent. Cleavage at Lys 133 occurs only on full-length apoA-I because cleavage at Leu 230 inhibits further proteolysis at Lys 133 . Therefore, to obtain the observed proteolytic pattern, cleavage at Lys 133 must precede cleavage at Leu 230 . In conclusion, modifications at the extreme C terminus of apoA-I, either by truncation of the terminal amino acids or by structural modifications promoted by self-association, impair PLA 2 proteolysis at the central domain of apoA-I (Lys 133 ). Therefore, specific structural and dynamic features at the C-terminal domain of monomeric lipid-free apoA-I are required for PLA 2 -mediated proteolysis of apoA-I to occur.
Substrate Specificity of PLA 2 -Characterization of the protein fragments in Gd-PP-PLA 2 -HDL suggested that only apoA-I, and not apoA-II, is the substrate of PLA 2 -mediated proteolysis. To further investigate the substrate specificity of the proteolytic action of PP-PLA 2 , lipid-free apo-A-II or BSA were incubated with PP-PLA 2 under the same conditions described for lipid-free apoA-I. By SDS-PAGE analysis, no fragmentation of apo-A-II or BSA was observed, even for long incubation times (12 h) (Fig. 5C), indicating that the proteolytic activity of PLA 2 is specifically directed toward lipid-free apoA-I.
ApoA-I Proteolysis Is Replicated by Different PLA 2 Isoforms-To investigate whether the proteolytic activity of PLA 2 against apoA-I is a general property of all PLA 2 enzymes or a specific activity of the pancreatic isoforms, porcine pancreatic (PP-PLA 2 , type IB), bee venom (BV-PLA 2 , type III), and human secretory PLA 2 (sPLA 2 , type IIA) were incubated with lipid-free apoA-I, as described under "Experimental Procedures." Although small variations in the nature and distribution of the proteolytic fragments were observed, all three enzyme isoforms cleaved apoA-I (Fig. 5D). Notably, both BV-PLA 2 and sPLA 2 -IIA needed longer incubation times to achieve levels of protein cleavage similar to those observed with PP-PLA 2 (ϳ12-24 h versus 1 h). In contrast, pancreatic PLA 2 enzymes from different animal sources (porcine (PP-PLA 2 ) or bovine (BP-PLA 2 )) had similar proteolytic efficiencies (Fig. 5D). These results suggest that the ability of PLA 2 to cleave the lipid-free form of apoA-I is a general property of all PLA 2 isoforms.

DISCUSSION
In inflammation states, plasma sPLA 2 levels increase, whereas HDL-C levels diminish. Because sPLA 2 is primarily associated with HDL in plasma (49), direct causality between these two events has been previously hypothesized, but a linking mechanism is yet to be established (13,50). Similar to HDL-C, apoA-I levels are also reduced during inflammation, leading to the hypothesis that the decrease in HDL-C originates from a net loss of apoA-I, perhaps due to accelerated catabolism (13,33,51,52). It is plausible that the combined increase in plasma levels of sPLA 2 and SAA in response to inflammatory stimuli promotes HDL remodeling and generation of lipid-free apoA-I that can be rapidly catabolized, with overall reduction in plasma HDL-C levels. Notably, the fractional catabolic rate of apoA-I in human sPLA 2 -transgenic mice is higher than in wildtype mice, whereas the catabolic rate of apoA-II is unaffected (12). This raises the possibility that human sPLA 2 specifically targets apoA-I.
Interestingly, sPLA 2 is also found in atherosclerotic lesions, which are sites of chronic inflammation (53). Similar to the acute phase response, the concerted presence of SAA and sPLA 2 in the extracellular matrix of atherosclerotic lesions is likely to increase apoA-I catabolism by promoting local HDL remodeling, thereby potentially reducing the antiatherogenic power of HDL (53). We previously investigated HDL remodeling strategies for the in vitro production of lipid-free/lipid-poor apoA-I (23)(24)(25). In the current study, we examined the direct effect of sPLA 2 on HDL and/or its molecular components when HDL remodeling events were induced by in vitro chemical denaturation or the action of CETP, in the absence or presence of PLA 2 . PLA 2 -mediated surface hydrolysis of HDL phospholipids resulted in a reduction of HDL particle size without dissociation of apolipoproteins (Fig. 1). When HDL remodeling was stimulated through partial chemical denaturation by guanidine treatment, apoA-I, but not apoA-II, apoC-I, or apoC-III, dissociated from HDL ( Fig. 2 and Table 1). Importantly, when the same remodeling experiment was performed in the presence of PLA 2 , apoA-I, but none of the other HDL-associated

LC-MS and MALDI-TOF MS analysis of control-apoA-I and PP-PLA 2 -apoA-I after SDS-PAGE separation
The gel bands numbered 1-6 in Fig. 4B were excised from the gel and digested with trypsin, and the tryptic digest was analyzed by LC-MS. The sequence of the tryptic peptides identified by LC-MS is highlighted in red. In a parallel SDS-PAGE experiment, the mass of the main protein component in undigested samples from the same gel bands was determined by MALDI-TOF MS (see "Experimental Procedures" and Table 1) and is reported in parenthesis. The sequence of the apoA-I peptide with calculated mass closest to the mass of the undigested protein fragment, as determined by MALDI-TOF, is shown in boldface type.
proteins, was extensively fragmented (Fig. 2 and Table 1). Similar results were observed when HDL remodeling was promoted by reaction with CETP ( Fig. 3 and Table 2).
Specific PLA 2 -mediated proteolysis of lipid-free apoA-I was confirmed by direct incubation of the enzyme with the lipidfree protein, upon which apoA-I was cleaved in fragments similar to those observed after HDL chemical denaturation or reaction with CETP in the presence of PLA 2 . This activity was specific because no proteolysis was observed upon incubation of PLA 2 with apoA-II or BSA. Mass spectrometric analysis of the apoA-I fragments generated by PLA 2 -mediated proteolysis identified two main cleavage sites in the central (Lys 133 ) and C-terminal (Leu 230 ) regions of apoA-I.
Amino acid 133 is proximal to a largely unstructured region of apoA-I, both in the HDL-associated (54,55) and in the lipidfree form of the protein (48). Notably, deletion of residues 134 -145 completely abolished proteolysis by PLA 2 at Lys 133 . This result suggests that a certain degree of structural freedom in the proximity of the cleavage site is a stringent condition for proteolysis by PLA 2 . This notion was confirmed by the results obtained with ⌬W-apoA-I, a Trp-Phe mutant in which the otherwise largely unstructured C-terminal domain of wild-type apoA-I folds into an ␣-helical structure due to increased protein self-association (48). Self-associated ⌬W-apoA-I was completely protected from PLA 2 -mediated proteolysis. These results indicate that the structural and dynamic properties of  (40). B, effect of PP-PLA 2 on the 1-230 fragment of apoA-I (apoA-I:⌬231-243) generated by a first reaction of full-length apoA-I with PLA 2 and isolated by SEC. ApoA-I:⌬231-243 was incubated in the presence of PP-PLA 2 for 1 h at 37°C (PP-PLA 2 -apoA-I:231-243) before SDS-PAGE analysis. C, to investigate the substrate specificity of PP-PLA 2 proteolytic action, lipid-free apoA-II and bovine serum albumin (BSA, essentially fatty acid-free) were incubated with PP-PLA 2 in the same conditions as described for apoA-I but for long incubation times (12 h). By SDS-PAGE analysis, no fragmentation of apo-A-II or BSA was observed. D, comparison of the proteolytic activities against apoA-I of different isoforms of PLA 2 and from different animal sources, by SDS-PAGE. Porcine pancreatic (PP-PLA 2 , type IB) and bovine pancreatic (BP-PLA 2 , type IB) PLA 2 were incubated with apoA-I for 1 h. Bee venom (BV-PLA 2 , type III) and human secreted (sPLA 2 , type IIA) PLA 2 were incubated with apoA-I for 24 h. For all panels, 5 g of protein was loaded per lane, and molecular weight standards were Precision Plus Protein Dual Xtra standards (Bio-Rad).
lipid-free apoA-I, which is highly flexible in the central and C-terminal regions, are likely to contribute to the susceptibility of the protein to PLA 2 -mediated proteolysis.
Important information about the mechanism of proteolysis was derived by the observation that the proteolytic but not the lipolytic activity of PLA 2 was Ca 2ϩ -independent. Similarly, the proteolytic but not the lipolytic activity of PLA 2 was inhibited by benzamidine. Recently, Cavazzini et al. (56) reported that activation of the fungal sPLA 2 enzyme TbSP1 is mediated by an autoproteolytic reaction. Based on the observation that autoproteolysis of TbSP1 is Ca 2ϩ -independent and inhibited by chemicals that do not affect the Ca 2ϩ -dependent lipolytic activity of the enzyme, the authors concluded that lipolysis and proteolysis are catalyzed by two independent sites. In analogy with the TbSP1 case, our data suggest that lipolysis and apoA-I proteolysis are also mediated by different catalytic sites and/or different conformational states of PLA 2 .
Notably, similar proteolytic activities directed toward apoA-I were observed with different classes of PLA 2 , including pancreatic (group IB), human plasma-secreted (group IIA), and bee venom (group III) PLA 2 . The susceptibility of the apoA-I C terminus to proteolysis by an array of proteases (e.g. metalloproteinases, chymotrypsin, trypsin, elastase, subtilisin, staphylococcus V8 protease, and arginine C endopeptidase) is well documented (57)(58)(59)(60) and derives from the relatively unstructured nature of the C-terminal domain of lipid-free apoA-I (48). Because the C-terminal domain of apoA-I is critical for lipid binding (61) and deletion of the terminal 23 residues (positions 221-243) abolishes apoA-I-mediated cholesterol efflux in cultured cells (62), proteolytic degradation of the apoA-I C-terminal domain by increased levels of PLA 2 may account for the production of dysfunctional apoA-I with diminished HDL biogenesis capacity, thus leading to the observed reduction in HDL-C levels during inflammation.
The current opinion is that plasma levels of enzymes of the sPLA 2 family are positively associated with the pathogenesis of atherosclerosis by various mechanisms, including the generation of atherogenic lipoprotein particles (63) and the activation of several proinflammatory pathways (64). In fact, some physiological functions of sPLA 2 enzymes are unrelated to their lipolytic activity and can be ascribed to the engagement of specific receptors on target cells (65). Our findings identify a direct link between PLA 2 and low HDL-C levels in inflammation through generation of fragmented apoA-I that can be rapidly catabolized. This novel property of PLA 2 may provide a mechanistic explanation for the contribution of sPLA 2 to the development of atherosclerosis, a disease tightly linked to inflammation.