Anti-apoptotic Actions of the Platelet-activating Factor Acetylhydrolase I (cid:1) 2 Catalytic Subunit*

Platelet-activating factor (PAF) is an important mediator of cell loss following diverse pathophysiological challenges, but the manner in which PAF transduces death is not clear. Both PAF receptor-dependent and -independent pathways are implicated. In this study, we show that extracellular PAF can be internalized through PAF receptor-independent mechanisms and can initiate caspase-3-dependent apoptosis when cytosolic concentrations are elevated by (cid:1) 15 p M /cell for 60 min. Reducing cytosolic PAF to less than 10 p M /cell ter- minates apoptotic signaling. By pharmacological inhibition of PAF acetylhydrolase I and II (PAF-AH) activity and down-regulation of PAF-AH I catalytic subunits by RNA interference, we show that the PAF receptor-inde-pendent death pathway is regulated by PAF-AH I and, to a lesser extent, by PAF-AH II. Moreover, the anti-apo-ptotic actions of PAF-AH I are subunit-specific. PAF-AH I (cid:1) 1 regulates intracellular PAF concentrations under normal physiological conditions, but expression is not sufficient to m M NaCl, pH 7.5). Western analyses were performed using polyclonal anti-LIS1 (1:500, Chemicon), monoclonal anti- (cid:2) 1 (1:1000, Dr. H. Arai, University of Tokyo, Tokyo, Japan), monoclonal anti- (cid:2) 2 (1:1000, Dr. H. Arai), poly(ADP-ribose) polymerase (PARP, 1:10, 000, Clontech), and actin (1:1000, Sigma). Secondary antibodies were horseradish peroxi-dase-conjugated or biotin-conjugated anti-mouse IgG (1:2000, Jackson ImmunoResearch; 1:10, 000, Sigma) or anti-rabbit IgG antibodies (1: 5000, Jackson ImmunoResearch), and tertiary reagents were extravi- din alkaline phosphatase (1:300,000, Sigma) as appropriate. Immuno-reactive bands were visualized using SuperSignal West Pico (MJS BioLynx Inc.) or nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (Sigma). Densitometry was performed using ImageJ analy- sis software (National Institutes of Health) standardized to actin loading controls. PAF-AH Activity— PAF-AH activities in complete media, serum-free RPMI, PC12-AC cells, PC12-AC conditioned treatment media, and C57Bl/6 mouse brain (cid:1) 3 months of age (positive control) were determined using a commercial PAF-AH assay kit (Cayman Chemicals). Cells and tissue were homogenized in 250 m M sucrose, 10 m M Tris-HCl (pH 7.4), and 1 m M EDTA using a Tissue Tearor (Fisher). Samples were centrifuged at 600 (cid:2) g for 10 min and at 100,000 (cid:2) g for 60 min. Cytosolic supernatants were concentrated using an Amicon centrifuge concentrator with a molecular mass cut-off of 10,000 kDa (Millipore).

Platelet-activating factor (PAF, 1 1-O-alkyl-2-acetyl-sn-glycero-3-phosphocholine) is a key mediator of neuronal death in ischemia, encephalitis, epileptic seizure, meningitis, and human immunodeficiency virus-1 dementia in vivo and participates in etoposide-, prion-, and ␤-amyloid-induced cell death in vitro (1)(2)(3)(4)(5)(6)(7). In the periphery, pathological increases in PAF concentrations underlie cytotoxicity in chronic inflammatory dermatoses and lethality in systemic anaphylaxis (8,9). Although the majority of PAF effects are understood to be transduced by its G-protein-coupled receptor (PAFR) (10), PAFR signaling has been shown to be both pro-and anti-apoptotic. Ectopic PAFR expression exacerbates cell death induced by etoposide and mitomycin C but protects cells from tumor necrosis factor ␣, TRAIL, and extracellular PAF (6,7,11,12). These opposing effects likely depend upon the relative ratio of NF-B-dependent pro-and anti-apoptotic gene products elicited in different cell types in response to the combination of an external apoptotic inducer and PAF (6).
Accumulating evidence points to additional PAF signaling pathways transduced independently of PAFR (11,(13)(14)(15)(16)(17). PAFR-negative cells undergo apoptosis when extracellular PAF concentrations reach 100 nM and necrosis when PAF levels exceed the critical micelle concentration of 3 M (11). Little is known about how PAF signals cell death in the absence of PAFR. PAF activates NF-B, glycogen synthase kinase 3␤, and caspase 3 and triggers mitochondrial release of cytochrome c (6, 18 -20). Whether or not these effects are dependent on PAFR activation is not clear.
One approach to intervening in both PAFR-dependent and -independent cell death lies in reducing pathological increases in PAF. PAF is hydrolyzed by a unique family of serine esterases or PAF acetylhydrolases (PAF-AHs) that cleave the biologically sn-2 active side chain generating lyso-PAF. Three PAF-AH enzymes have been identified. Cytosolic PAF-AH I cleaves the acetyl group at the sn-2 position of PAF and PAFlike lipids with other phosphate head groups (21). The enzymatic complex is a G-protein-like trimer composed of two 29-kDa ␣ 1 and ␣ 2 catalytic subunits. The ␣ subunits form homodimers or heterodimers that complex with a noncatalytic 45-kDa regulatory ␤ subunit, LIS1. Mutations in the LIS1 gene * This work was supported in part by grants from the Alzheimer Society of Canada, Alzheimer Society of Saskatchewan, and the Canadian Institute of Health Research Joint Initiative (to S. A. L. B.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  are the genetic determinant of Miller-Dieker syndrome, a developmental brain disorder defined by type 1 lissencephaly (22). PAF-AH II is a single 40-kDa polypeptide (23). This isoenzyme recognizes both PAF and acyl analogs of PAF with moderate length sn-2 chains as well as short chain diacylglycerols, triacylglycerols, and acetylated alkanols (24). Ectopic expression reduces cell death triggered by oxidative stress (25,26). Plasma PAF-AH is a 45-kDa monomer secreted into circulation by endothelial and hematopoietic cells (27,28). The enzyme recognizes PAF and PAF analogs with short to medium sn-2 chains including oxidatively cleaved long chain polyunsaturated acyl chains (24). In vitro, recombinant plasma PAF-AH or ectopic expression protects cells from excitotoxicity or hypercholesterolemia (29 -31). In vivo, intravenous injection of human plasma PAF-AH reduces lethality in experimental models of anaphylactic shock (8). These findings provide compelling evidence that PAF-AH activity regulates PAF-mediated apoptosis. It remains to be determined whether these enzymes can, in fact, be targeted to inhibit PAF-mediated degeneration and disease.
In this study, we used the PC12 cell model to investigate the anti-apoptotic actions of PAF-AH I and PAF-AH II in the PAFR-independent death pathway. We show that PC12 cells express all three PAF-AH I proteins (␣ 1 , ␣ 2 , and LIS1) as well as PAF-AH II but not plasma PAF-AH or PAFR. Expression of PAF-AH I ␣ 2 but not PAF-AH I ␣ 1 is induced when PC12 cells are deprived of serum. We found that this induction regulates the duration of apoptotic signaling initiated by PAF challenge. To enhance the endogenous anti-apoptotic activity of PAF-AH I ␣ 2 , we screened a panel of PAF antagonists, and we identified two compounds that blocked PAFR-independent death. Both compounds, the fungal derivative FR 49175 and the ginkgolide BN 52021, protected cells by accelerating PAF hydrolysis. Most surprisingly, both inhibitors suppressed ␣ 1 protein expression thereby promoting ␣ 2 /␣ 2 homodimer activity following PAF treatment. These findings point to a novel anti-apoptotic function for the ␣ 2 subunit of PAF-AH I and a potential means of pharmacologically targeting PAF-AH enzymes to reduce PAFmediated cell death.

EXPERIMENTAL PROCEDURES
Cell Culture-PC12-AC cells, a clonal derivative of the PC12 pheochromocytoma cell line (American Tissue Culture Collection), were cultured in complete media composed of RPMI 1640 containing 10% horse serum and 5% newborn calf serum at 37°C in a 5% CO 2 , 95% air atmosphere. Culture reagents were obtained from Invitrogen.
PAF-AH Activity-PAF-AH activities in complete media, serum-free RPMI, PC12-AC cells, PC12-AC conditioned treatment media, and C57Bl/6 mouse brain ϳ3 months of age (positive control) were determined using a commercial PAF-AH assay kit (Cayman Chemicals). Cells and tissue were homogenized in 250 mM sucrose, 10 mM Tris-HCl (pH 7.4), and 1 mM EDTA using a Tissue Tearor (Fisher). Samples were centrifuged at 600 ϫ g for 10 min and at 100,000 ϫ g for 60 min. Cytosolic supernatants were concentrated using an Amicon centrifuge concentrator with a molecular mass cut-off of 10,000 kDa (Millipore). Protein (30 -50 g) was incubated with C 16 -2-thio-PAF substrate for 30 min at room temperature. In some cases, lysates were pretreated for 15 or 30 min at room temperature with diisopropyl fluorophosphate (DFP) at the concentrations indicated in the text. Percent inhibition was calculated relative to lysates treated with vehicle (PBS) for the same time. Hydrolysis of the thioester bond at the sn-2 position was detected by conjugation with 5,5Ј-dithiobis(2-nitrobenzoic acid) (DTNB) at 405 nM. Control reactions included samples incubated without lysate or media and samples incubated without substrate.
Internalization Assay (Live Cell Imaging)-PC12-AC cells (5 ϫ 10 4 cells/well) were plated in complete media overnight in 24-well plates (VWR Scientific) coated with 0.1% gelatin. Cells were washed in 10 mM PBS and incubated with 1 M Bodipy FL C 11 -PAF (B-PAF) in RPMI 1640 containing 0.1% bovine serum albumin (BSA, Sigma). B-PAF was custom-synthesized for our laboratory by Molecular Probes. At each time point (0, 10, 20, 30, 40, 50, 60, 80, 90, 100, 110, and 120 min), incubation media were removed, and cells were washed with 10 mM PBS. Live cell imaging under phase and fluorescence of identical cell fields was performed using a DMR inverted microscope (Leica) equipped with a QICAM digital camera (Quorum Technologies) and captured using OpenLab software version 3.17 (Improvision). Following time-lapse imaging, the incubation media were replaced, and internalization was allowed to continue. Quantitation of fluorescence intensity of individual cells was performed using the Advanced Measurement Module of OpenLab version 3.17.
Lipid Extraction and TLC-PC12-AC cells seeded at 1 ϫ 10 5 cells onto 10-cm plates were maintained in complete media at 37°C in 5% CO 2 for 72 h. Cultures were incubated at 37°C with 1 M B-PAF in RPMI 1640 containing 0.1% BSA for 0, 5, 15, 30, 45, 60, and 75 min. At each time point, 4 plates/condition were removed from the incubator and were placed on ice. One ml of methanol acidified with 2% acetic acid was added to each plate, and the extracellular fraction was collected. This fraction contained B-PAF in the culture media and uninternalized B-PAF bound to cell surface proteins or associated with the plasma membrane. The remaining monolayer of cells was collected in acidified methanol by scraping the plate with a cell lifter (Fisher). Lipids were extracted from the extracellular milieu and cytosolic fractions by the Bligh and Dyer method (32) and developed on TLC plates (20 ϫ 20 cm Silica Gel 60 (Fisher) in a solvent system of chloroform/methanol/acetic acid/water (50:30:8:5, v/v). B-PAF and B-lyso-PAF (Molecular Probes) were used as authentic markers. Fluorescent lipids were visualized under UV light using AlphaImager-1220 software (Alpha Innotech Corp.). Fluorescence intensity corresponding to lipid yield was determined by densitometry using the Advanced Measurement Module of Metabolic activity was assessed based on the ability of mitochondrial dehydrogenases, active in viable cells, to reduce the formazan salt, WST-1 (measured at A 450 -A 690 nm , Roche Applied Science). DNA fragmentation was determined by terminal deoxytrans-ferase dUTP nick-end labeling (TUNEL, Roche Applied Science) of cultures fixed for 20 min in 4% paraformaldehyde in 10 mM as described previously (33). Cells, processed for TUNEL, were double-labeled with Hoechst 33258 (0.2 g/ml) for 20 min at room temperature for additional morphological evidence of apoptotic loss.
RNA Interference Transfection-To suppress expression of the PAF-AH I ␣ 2 subunit, we designed a double-stranded short interfering RNA (siRNA) to the ␣ 2 sequence (AATAAACATGCTTGTCACTCCC/ CTGTCTC) and a negative control scrambled sequence (AATGCATAG-GAGTTGGAGAGGC/CTGTCTC). Oligonucleotides were obtained from the Biotechnology Research Institute at the University of Ottawa. siRNA duplexes were generated using the Silencer siRNA construction kit (Ambion). Transfection of siRNA was performed with Lipo-fectAMINE 2000 (Invitrogen) and optimized to yield maximal transfection efficiency according to manufacturer's protocol. Briefly, 2 l of LipofectAMINE 2000 was diluted in 198 l of RPMI 1640 media for 5 min at room temperature. siRNA duplexes (PAF-AH ␣ 2 or scrambled) were suspended in 100 l of RPMI 1640 media. PC12-AC cells grown in complete media to 30% confluence in 4-well gelatin-coated Labtek wells were treated with 100 l of the LipofectAMINE-siRNA complex and incubated for 72 h at 37°C. Transfection efficiency was determined in separate cultures by counting EGFP-positive cells upon transfection with pEGFP-C1 as well as morphological changes using ␤-actin siRNA positive control (Ambion). Because we were unable to obtain Ͼ20% transfection efficiency, we co-transfected PAF-AH ␣ 2 or scrambled siR-NAs with 0.028 g/l of pEGFP-C1 to identify silenced cells. Cells were then exposed to 1 M PAF in treatment media (RPMI ϩ 0.025% BSA) for either 45 min or 24 h. Data were standardized to the number of EGFP-positive cells in vehicle-treated cultures transfected with pEGFP-C1 only.
Statistical Analysis-Data were analyzed by using ANOVA tests or unpaired Student's t tests, as applicable. Following detection of a statistically significant difference in a given series of treatments, post hoc Dunnett's t-tests or Tukey tests were performed where appropriate. p values under 0.05 were considered statistically significant (shown as * or †); p values under 0.01 were considered highly statistically significant (shown as ** or † †). (11) that PC12-AC cells do not express PAFR but undergo apoptosisassociated DNA fragmentation 24 h after treatment with Ն100 nM PAF and necrotic lysis when treated with Ͼ3 M PAF. In this study, we addressed the kinetics and underlying signaling mechanism of PAF-induced apoptosis in the absence of PAFR (Fig. 1A, inset). To determine whether cell death is mediated by PAF or downstream PAF metabolites, PC12-AC cells were treated with PAF, mc-PAF, the PAF-AH-resistant synthetic PAF analog (34), or lyso-PAF, the immediate PAF metabolite. Both mc-PAF and PAF triggered comparable concentration-dependent cell death 24 h after treatment (Fig. 1A). lyso-PAF had no significant effect on cell viability (Fig. 1A). To establish the kinetics of PAF-mediated cytotoxicity, cell survival was assessed at various times after phospholipid administration. PAF (1 M) elicited significant cell loss within 24 h of treatment; no further reductions in cell number were detected at 48 or 72 h relative to vehicle controls (Fig. 1B). mc-PAF (1 M) elicited incremental cell loss for up to 72 after treatment (Fig. 1B). Comparable kinetics were observed if PAF (1 M) was replenished in fresh media at 24-h intervals with cells repeatedly treated with PAF exhibiting sustained impairment of cellular metabolic activity (Fig. 1C) and an incremental reduction in cell survival (Fig. 1D).

PAF Can Elicit Apoptosis Independently of Its G-proteincoupled Receptor-We have demonstrated previously
To determine whether PAF activates caspases in the absence of PAFR, we examined cleavage of the caspase 3 substrate PARP. Caspase 3-mediated PARP cleavage from 116 to 85 kDa was observed 45 min after a single PAF treatment ( Fig. 2A). Cleavage markedly increased 3 h after phospholipid administration and was detected for up to 24 h after treatment ( Fig. 2A) but not at later time points (data not shown). TUNEL-positive cells, evidence of terminal DNA fragmentation, were first observed 24 h after PAF exposure with no evidence of additional death 48 h after exposure (Fig. 2B).
Collectively, these findings indicate the following. Cytosolic PAF-AH Enzymes Limit the Duration of PAF-induced Apoptotic Signaling-These kinetics implicate PAF degradation in the control of apoptotic signaling. To test this hypothesis, we first identified LIS1, PAF-AH I ␣ 1 , PAH-AH I ␣ 2 , and PAF-AH II but not plasma PAF-AH transcripts in PC12 cells by RT-PCR (Fig. 3, A-F). Amplicon integrity of the PAF-AH I subunits was verified by sequencing on both strands. We then confirmed protein expression by Western analysis. ␣ 1 protein was constitutively expressed in the presence or absence of PAF (Fig. 3G). ␣ 2 protein was present at extremely low levels (Fig. 3, H and I) with marked induction following exposure to PAF (Fig. 3, H and I). Closer examination revealed that ␣ 2 protein also increased when cells were cultured in serum-free treatment media suggesting a response to serum withdrawal rather than PAF-specific induction (Fig. 3I). LIS1 was constitutively expressed by PC12-AC cells (Fig. 3J).
Lacking an antibody to PAF-AH II to verify protein expression, we sequenced the full-length transcript and assayed functional PAF-AH activity in the presence of cytosolic PAF-AH I and II inhibitors. Full-length transcript was amplified from Wistar rat brain and PC12-AC cells using a series of primer pairs homologous to conserved sequences in murine, bovine, and human genes (Table I). A 1173-bp open reading frame was identified in both PC12-AC cells and Wistar rat brain with no base pair mismatches between the sequences. 2 In functional assays, total cytosolic PAF-AH activity in cell lysates was comparable with enzymatic activity in adult mouse brain homogenates (Table II). PAF hydrolytic activity was not detected in treatment media harvested from cells after 24 h of culture confirming the RT-PCR analysis that PC12-AC cells do not express the secreted PAF-AH isoenzyme (Table II). To distinguish between PAF-AH I and PAF-AH II activity, cell lysates were treated with the active serine blocking agent DFP. 0.1 mM DFP inhibits PAF-AH I ␣ 1 /␣ 1 or ␣ 1 /␣ 2 dimer activity, whereas 1 mM DFP inhibits PAF-AH I ␣ 1 /␣ 1 , PAF-AH I ␣ 1 /␣ 2 , and PAF-AH II activity. PAF-AH I ␣ 2 /␣ 2 dimers are DFP-resistant (21,23,35). 68 -70% of the total cytosolic PAF-AH activity was blocked by 0.1 mM DFP indicative of PAFAH I ␣ 1 /␣ 1 or ␣ 1 /␣ 2 activity (Table II). Increasing the DFP concentration to 1 mM blocked an additional 4 -7% of PAF-AH activity attributed to PAF-AH II. DFP-resistant PAF-AH I ␣ 2 /␣ 2 homodimer activity composed 25% of cytosolic PAF-AH activity in lysates extracted from cells cultured in complete media.
The kinetics of PAF hydrolysis by PAF-AH I and II were established using the fluorescent PAF substrate B-PAF. Cells were treated with 1 M B-PAF to initiate apoptotic signaling. B-PAF and its metabolites were extracted at various time points using a modified Bligh and Dyer procedure and identi- 2 The rat PAF-AH II cDNA cloned from adult Wistar rat brain was deposited under GenBank TM accession number AY225592. fied by TLC. Fluorescent intensity was quantified by comparison to resolved fluorescent standards. Cytosolic and extracellular lipids were fractionated from B-PAF-treated cultures with an acidified methanol wash separating phospholipids bound to proteins on plasma membrane from internalized lip-ids (36). We found that 1 M B-PAF was stable in cell-free, serum-free treatment media at 37°C for at least 60 min; no degradation to B-lyso-PAF was observed in the absence of cells (Fig. 4A). B-PAF was rapidly internalized by PC12-AC cells (Fig. 4B, extracellular) with concentrations reaching ϳ16 -20 . Plasma PAF-AH mRNA was not detected (E). The absence of genomic DNA contamination or reagent contamination was confirmed by the control reactions: no RT during the RT reaction, no template during the PCR reaction, and no primers during the PCR reaction. Rat brain RNA was reverse-transcribed as a same-species positive control (Rat Brain lane). Template integrity of the random-primed PC12-AC RT product was verified by using GAPDH (F). Equal amounts of protein (30 g) from cells cultured in complete media (0) or cells treated for 60 min with 1 M PAF in treatment media (60) were separated under reducing conditions and immunoblotted for ␣ 1 (G), ␣ 2 (H and I), or LIS1 (J). Protein lysates from neonatal rat brain were used as a same-species positive control. Representative immunoblots of replicate experiments are depicted. PAF-AH ␣ 1 was constitutively expressed (G). Significant PAF-AH ␣ 2 protein was not detected until cells were treated with PAF (1 M) (H). Longer exposure times (overnight) indicated that PC12-AC cells expressed low levels of protein and that serum deprivation was sufficient to induce ␣ 2 protein expression (I). LIS1 was constitutively expressed (J). Blots were probed for actin as a loading control (G-J). pM/cell within 5 min of incubation (Fig. 4B, cytosolic). Internalized B-PAF levels remained constant for 60 min dropping to 6 -8 pM/cell by 75 min (Fig. 4B, cytosolic). B-PAF degradation was detected within 15 min of internalization (Fig. 4C, vehicle). A linear rate of metabolism (r 2 ϭ 0.97) was observed for up to 75 min (Fig. 4C, vehicle). Closer examination of the fate of B-PAF metabolite by fractionation revealed that B-lyso-PAF was released from cells in two stages (Fig. 4D). Between 5 and 15 min of incubation, 50% of internalized B-PAF was converted to B-lyso-PAF (Fig. 4D, Cytosolic) and secreted from the cells (Fig. 4D, extracellular). Between 30 and 60 min, B-lyso-PAF was not released and accumulated in the cell cytosol (Fig. 4D,  cytosolic). A delayed phase of secretion was observed between 60 and 75 min (Fig. 4D, extracellular).
To track the fate of secreted B-lyso-PAF, we performed quantitative time-lapse fluorescence microscopy (Fig. 4E). B-PAFcontaining media were removed, and cells were washed with PBS containing 1% BSA at various time points to remove free lipids loosely associated with the plasma membrane but not phospholipids bound to membrane proteins. Cells were photographed, and the B-PAF-containing media were replaced. A linear increase in cell-associated fluorescence was observed over the first 70 min of exposure (r 2 ϭ 0.97) followed by a plateau in cell-associated B-PAF (Fig. 4E).
Taken together, these data indicate that PAF administered to cells is metabolized by PAF-AH I and II with a t1 ⁄2 of 75 min (where t1 ⁄2 is the time to reach half-maximal concentrations). The PAF metabolite, lyso-PAF, is secreted from cells but remains bound, apparently to carrier proteins, on the extracellular surface of the plasma membrane.
Cytosolic PAF-AHs Can Be Targeted Pharmacologically to Promote Cell Survival-The kinetics of PAF-induced apoptosis (Figs. 1 and 2) and PAF degradation (Fig. 4) suggest that cytosolic PAF concentrations must remain elevated for at least 60 min to elicit apoptosis. This hypothesis predicts that compounds capable of inhibiting PAF internalization or increasing cytosolic PAF-AH activity will block PAF-mediated death triggered independently of PAFR. To test this hypothesis, we evaluated the anti-apoptotic actions of four PAF antagonists (CV-3988, CV-6209, BN 52021, and FR 49175). These compounds were chosen because of their affinities for different PAF-binding sites identified pharmacologically. CV-3988 and CV-6209 are competitive PAF antagonists that preferentially interact with synaptosomal and microsomal PAF-binding sites (17). CV-3988 competes for PAF at the plasma membrane, likely PAFR, as well as interacting with intracellular microsomal binding sites, likely internalized PAFR (17). CV-6209 preferentially interacts with a single binding site in microsomal membranes (17). BN 52021 (also known as ginkgolide B) is a noncompetitive PAF antagonist with affinity for three discrete PAF-binding sites (17) as well as potent antioxidant activity (37). FR 49175 is a fungal metabolite derivative that inhibits PAF-induced biological activity through unknown mechanisms (38,39). Treatment of PC12-AC cells with CV-3988, BN 52021, or FR 49175 had no effect on cell survival (Fig. 5, A-C); however, CV-6209 was toxic to cells at concentrations above 1 M (Fig. 5D). We observed significant concentration-dependent anti-apoptotic activity when cells were preincubated for 15 min with BN 52021 (1-100 M) or FR 49175 (0.5-50 M) and then exposed to PAF (Fig. 5E). Both BN 52021 and FR 49175 inhibited PAF-mediated caspase 3 activation as assessed by PARP cleavage (Fig. 5F). CV-3988 (0.1-10 M) and CV-6209 (0.01-1 M) did not exhibit anti-apoptotic activity (Fig. 5E).
We next tracked B-PAF fate by lipid extraction and TLC in the presence or absence of BN 52021 and FR 49175 to establish whether these antagonists affect phospholipid internalization or PAF-AH activity. FR 49175 and BN 52021 did not alter the rate or extent of B-PAF internalization (Fig. 6, A and D, extracellular). In both antagonist-and vehicle-treated cultures, maximal cytosolic B-PAF levels were attained within 5 min of B-PAF exposure (Fig. 6, A and D, cytosolic 5 min). We did not observe an increase in total B-PAF hydrolysis. Cytosolic B-PAF concentrations, estimated at 8 pM/cell, in BN 52021-and FR 49175-treated cultures were comparable with vehicle-treated cells by the end of the 75 min test period (Fig. 6, A and D,  cytosolic, 75 min). Significantly, we found that FR 49175 and BN 52021 accelerated both the kinetics of B-PAF degradation (Fig. 6, A and D, 45 min) and the release of B-lyso-PAF into the media (compare Fig. 6, B and E, cytosolic 45 min and C and F,  extracellular, 45 min). By 45 min, 39 (FR 49175) or 41% (BN 52021) of exogenous B-PAF had been converted to B-lyso-PAF compared with 29% in vehicle-treated cells. This acceleration dropped cytosolic B-PAF levels more rapidly in antagonisttreated cultures from 16 pM/cell (5 min) to 10 pM/cell (45 min) in the presence of FR 49175 or 8 pM/cell (45 min) in the presence of BN 52021 (Fig. 6, B and E). Maximal release of B-lyso-PAF from cells was observed within 45 min in antagonist-treated cultures compared with 75 min in vehicle-treated cultures (Fig.  6, C and F). 50% of the released metabolite diffused or was transported back into the cells within 75 min in FR 49175 treated cultures; 32% of B-lyso-PAF re-entered cells in BN 52021-treated cultures (Fig. 6, B-F).
PAF-AH I ␣ 2/ ␣ 2 Activity Is Anti-apoptotic-To determine how BN 52021 and FR 49175 accelerate PAF-AH hydrolysis, we examined PAF-AH I ␣ 1 and ␣ 2 expression. Cells were preincubated in serum-free treatment media with vehicle (Me 2 SO), BN 52021, or FR 49175 for 15 min before addition of PAF (Fig. 7). We found PAF-AH ␣ 1 expression remained constant whether cells were cultured in complete media, serum-free treatment media, pretreated with Me 2 SO, or exposed to PAF (Fig. 7A, . Surprisingly, PAF-AH ␣ 1 protein levels decreased progressively over the first 30 min of PAF treatment and remained low until the end of the 90-min test period (Fig. 7A, BN 52021). Comparable results were observed in FR 49175-treated cultures (Fig. 7A, FR 49175). ␣ 2 expression was not altered by BN 52021 or FR 49175 in that protein expression increased dramatically during the 15-min preincubation period in serumfree treatment media regardless of antagonist or vehicle administration and remained elevated over the 90 min PAF test period (Fig. 7B).
The finding that both compounds accelerate PAF degradation by reducing PAF-AH I ␣ 1 subunit expression was unexpected. Two alternative, but not mutually exclusive, explanations lie in the possibility that this reduction in ␣ 1 expression promotes formation of PAF-AH I ␣ 2 /␣ 2 homodimers and/or that PAF-AH II activity is increased. Following exposure to PAF in serum-free media, the catalytic composition of PAF-AH I would be expected to change from a predominance of ␣ 1 /␣ 1 homodimers in control cultures (Table I) to ␣ 1 /␣ 2 heterodimers and ␣ 2 /␣ 2 homodimers given the 8 -10-fold increase in ␣ 2 protein expression (Figs. 3 and 7). In cultures pretreated with BN 52021 or FR 49175, suppression of ␣ 1 protein expression likely favors a more a rapid transition from ␣ 1 /␣ 1 to ␣ 2 /␣ 2 homodimers following PAF exposure. This hypothesis is supported by the dramatic increase in the ␣ 2 to ␣ 1 ratio observed 30 -90 min after PAF exposure in BN 52021 and FR 49175treated cultures relative to vehicle (Fig. 7C). Alternatively, in addition to effects on PAF-AH ␣ 1 , BN 52021 and/or FR 49175 may alter PAF-AH II activity, an enzyme with proven antiapoptotic properties (26).
To address these possibilities, we performed three loss of function studies. First, we treated cells with 0.1 mM DFP to inhibit PAF-AH I ␣ 1 /␣ 1 and ␣ 1 /␣ 2 catalytic activity without changing the relative ratio of ␣ 1 to ␣ 2 protein expression (21,23,35). In the unlikely event that the anti-apoptotic effects of BN 52021 and FR 49175 are mediated by a loss of ␣ 1 catalytic activity, then 0.1 mM DFP should be cytoprotective. Cells were exposed for 15 min (Fig. 4C and Fig. 8A) before addition of B-PAF. DFP did not affect B-PAF internalization (data not shown) but reduced B-PAF degradation by 38% 75 min after B-PAF exposure (Fig. 4C, 0.1 mM DFP, 75 min). To ensure intracellular concentrations of DFP reached 0.1 mM by the time cells were exposed to PAF, we extended the preincubation time point, lipids were extracted from the extracellular and cytosolic fractions and were separated by TLC. A-C, data represent relative fluorescence of B-PAF in each fraction expressed as a percentage of the total fluorescent lipids recovered. Data are the mean of 2-4 independent experiments conducted in replicate. A, B-PAF was not degraded by treatment media in the absence of PC12 cells. B, in cell cultures, B-PAF was rapidly internalized within 5 min (extracellular fraction) after which levels remained stable until 60 min and dropped by 75 min (cytosolic fraction). C, conversion of B-PAF to B-lyso-PAF was monitored to determine the degree of PAF hydrolysis in the presence or absence of the PAF-AH I ␣ 1 and PAF-AH II inhibitor DFP. Cells were pretreated for 15 min in treatment media with vehicle (PBS) or DFP before addition of B-PAF. D, the fate of the B-PAF metabolite B-lyso-PAF was followed by fractionation. E, to determine whether B-lyso-PAF secreted from cells remains loosely associated with the lipid bilayer or bound to membrane proteins, cell-associated fluorescence was tracked by live-cell imaging. The B-PAF containing media were removed every 10 min, and cells were washed with PBS ϩ 1% BSA to remove phospholipids at the extracellular face of the plasma membrane but not lipids associated with membrane proteins. Data are reported as the mean increase in fluorescence intensity standardized to background fluorescence in untreated cells Ϯ S.E. (n ϭ 16 -25 per data point). Cell-associated fluorescence increased 3-fold over the first 75 min of exposure and remained constant between 75 and 120 min despite repeated washes indicating that the secreted material removed by the acidified methanol wash (D, extracellular) was likely bound to proteins at the plasma membrane (E). Scale bars, 5 m. FIG. 4. Internalization, kinetic analysis, and metabolic fate (Fig. 8A, vehicle (0.1%  EtOH)) and did not intensify PAF-mediated cell death (Fig. 8A,  PAF). These data suggest that PAF-AH I ␣ 1 does not play a significant role in regulating PAF-induced apoptosis.
Second, to determine the role of PAF-AH II in control of PAF-mediated cell death, we treated cultures with PAF (1 M) or PAF vehicle (0.1% EtOH) in the presence of DFP, the sulfhydryl blocking reagent (DTNB), or vehicle (PBS) (Figs. 4C and 8A). Cells were pre-exposed to 1 mM DFP or 1 mM DTNB for 15 (Figs. 4C and 8A) or 30 min (data not shown) before treatment with PAF or vehicle. Individually, DFP and DTNB were found to have no effect on the survival of vehicle-treated cells (Fig.  8A, vehicle (0.1% EtOH)). Increasing the DFP concentration from 0.1 to 1 mM inhibited 55% of B-PAF degradation within 75 min indicating that 17% of DFP-sensitive PAF-AH activity was PAF-AH II (Fig. 4C, 1 mM DFP, 75 min). Despite this substantial inhibition of PAF hydrolysis (55%), PAF-mediated cell loss was intensified by only 8% (Fig. 8A, 1 mM DFP). To confirm these results, we used a pharmacological agent that, unlike DFP, inhibits PAF-AH II but not PAF-AH I through different mechanisms. Exposure to 1 mM DTNB elicited the same results as 1 mM DFP (Fig. 8A, DTNB). To ensure complete inhibition of PAF-AH II, we exposed cells to both DFP and DTNB. Combination treatment reportedly abolishes activity of purified PAF-AH II (35). Preincubation in DFP and DTNB followed by 24 h of exposure to EtOH decreased vehicle-treated cell survival by 12% (Fig. 8A, vehicle (0.1% EtOH)). Combination of treatments reduced PAF-treated cell survival by 20%, likely the cumulative results of inhibitor toxicity (12%) and PAF-AH II inhibition (8%) (Fig. 8A, PAF). These findings suggest that (a) PAF-AH I ␣ 1 activity is not anti-apoptotic and (b) that although PAF-AH II activity is cytoprotective, PAF-AH II plays a secondary role to PAF-AH I ␣ 2 in regulating PAF-mediated apoptosis.
The data point to PAF-AH I ␣ 2 as a potential therapeutic target to reduce apoptogenic PAF concentrations under diverse pathophysiological conditions. To confirm this role directly, we acutely suppressed ␣ 2 expression using an RNAi strategy. PAF-AH I ␣ 2 -specific RNAi but not the scrambled control or transfection reagent alone reduced ␣ 2 protein expression (Fig.   8B). Because we were unable to achieve Ͼ20% transfection efficiency, we could not knockdown PAF-AH I ␣ 2 expression in all cells. To identify transfected cells, we co-transfected cultures with EGFP in combination with ␣ 2 -specific or scrambled siRNAs (Fig. 8C). To control for possible cytotoxic effects of siRNAs, parallel experiments were performed in which cultures were transfected with EGFP alone, and survival was expressed relative to EGFP-positive vehicle-treated cells. Most importantly, PAF-AH I ␣ 2 RNAi down-regulation significantly enhanced PAF-mediated cytotoxicity (Fig. 8C). DISCUSSION PAF-like lipids have been implicated as key apoptotic second messengers induced by a variety of pathological stressors (6,19,20,25,26,29,30,33,40). Converging evidence points to PAFR activation of the conserved mitochondrial death pathway (6, 12, 18 -20, 41). Our previous work has demonstrated that PAF can also transduce cell death independently of its Gprotein-coupled receptor (11), but it is not known how cell death is regulated in the absence of PAFR. The importance of this apoptotic cascade is underlined by the fact that neurons express low levels to no PAFR (33,42), and yet PAF is a principal mediator of neuronal loss in ischemia, encephalitis, epileptic seizure, meningitis, and human immunodeficiency virus-1 dementia (1)(2)(3)(4)(5). To provide mechanistic insight into this PAFR-independent pathway, we show the following. 1) PAF can initiate caspase-dependent cell death in the absence of it G-protein-coupled receptor.
2) The duration of PAF apoptogenic signaling is regulated by the ␣ 2 subunit of PAF-AH I and, to a lesser extent, by PAF-AH II. 3) PAFR-independent cell death can be inhibited by two PAF antagonists: the gingkolide BN 52021 and the fungal derivative FR 49175 but not by CV-3988 or CV-6209. 4) Both BN 52021 and FR 49175 protect cells from PAF-induced cell death by reducing PAF-AH I ␣ 1 protein levels indirectly promoting PAF-AH I ␣ 2 homodimer activity.
Apoptotic PAF Signal Transduction in the Absence of PAFR-To provide mechanistic insight into how PAF transduces cell death in the absence of PAFR, we followed the internalization and metabolic fate of Bodipy fluorophore-conjugated PAF in PAF-sensitive cells (11,43). PC12 cells express all of the components of both cytosolic PAF-AH enzymes (I and II) but not plasma PAF-AH or PAFR. Extracellular PAF was internalized by PC12 cells with a t1 ⁄2 of ϳ5 min. Downstream activation of caspase 3 was initiated when cytosolic PAF concentrations were elevated by ϳ15-20 pM/cell. These findings complement previous work demonstrating that exposure to oxidative stressors triggers apoptogenic remodeling of membrane phospholipids into PAF-like lipids (26) and provide strong evidence that cytosolic accumulation of PAF and PAFlike lipids can trigger apoptotic death independently of PAFR. In fact, internalization of PAF mediated by PAFR endocytosis (36) may protect cells from this cell death cascade. Endocytosis of the PAF⅐PAFR complex triggers release of plasma PAF-AH from macrophages thereby reducing extracellular ligand concentrations (36).
Although we have yet to identify the effector proteins responsible for transducing increases in intracellular PAF concentration into downstream caspase activation, our data indicate that the temporal kinetics of PAF accumulation regulate the duration of apoptotic signaling and that PAFR-independent cell death is triggered by PAF and not by its immediate metabolite lyso-PAF. We found that intracellular PAF levels must remain elevated by approximately 60 min to elicit significant apoptotic death. Decreasing cytosolic concentrations to less than 10 pM/ cell of exogenous PAF stops the cell death signaling. lyso-PAF, the immediate PAF metabolite, does not transduce downstream apoptotic induction given that equimolar concentrations of lyso-PAF are not cytotoxic and that the endogenous metabolite remains cell-associated without detectable effect. Accelerating the kinetics of PAF degradation with BN 52021 or FR 49175 reduces PAF concentration to sub-apoptotic levels (approximately Ͼ8 pM/cell) within 45 min and prevents PAFinduced caspase activation. The PAFR-specific antagonists CV-3988 and CV-6209 have no effect on PAF-mediated PAFRindependent cell death when administered at concentrations up to 20 times that of their reported IC 50 antagonist activities (44 -46). Most interestingly, the anti-cell death activities of the antagonists tested in this study (FR 49175 Ͼ BN 52021 Ͼ CV-3988 or CV-6209) are in direct opposition to their PAF antagonist potency in other biological assays (CV-6209 Ͼ CV-3988 Ͼ BN 52021 Ͼ FR 49175) (38, 44 -47), suggesting that their PAFR-independent anti-apoptotic actions are inversely proportional to their affinity for PAFR.
We do not know how PAF is internalized by PC12 cells in the absence of its G-protein-coupled receptor. In addition to endocytosis as a PAF⅐PAFR complex, active trafficking of PAF across the plasma membrane is accomplished by a PAF-specific transglutaminase and by interaction with low affinity binding sites yet to be identified at the molecular level (14,48). Passive PAF internalization is regulated by transbilayer movement (flipping) across the plasma membrane occurring as a result of physicochemical changes in membrane properties accompanying cellular activation (15). It has been suggested that PAF internalization in the absence of PAFR is not rapid enough to elicit acute biological activity in hematopoietic cells (36,49). In this study, we present evidence that PAFR-independent PAF internalization by nonhematopoietic cells is indeed sufficient to trigger apoptotic cell loss.
PAF-AH ␣ 2 Activity Is Anti-apoptotic-Administration of recombinant PAF-AH II or plasma PAF-AH and ectopic overexpression protects cells from death elicited by low density lipoprotein, glutamate, or oxidative stress (25, 26, 29 -31). In this study, we show that PAF-AH I exerts similar cytoprotective effects with three important distinctions. First, the antiapoptotic actions of PAF-AH I are subunit-specific. Under normal culture conditions, we found PAF hydrolysis in PC12 cells to be primarily mediated by the PAF-AH I ␣ 1 subunit and to a lesser extent by PAF-AH II. Following serum deprivation and exposure to pathological PAF concentrations, PAF-AH I ␣ 2 but not PAF-AH I ␣ 1 , protein expression is acutely up-regulated. By pharmacological inhibition using DFP and DTNB and by RNA interference, we found that PAF-AH I ␣ 2 but not PAF-AH I ␣ 1 activity reduces the duration of PAF-mediated apoptotic signaling. When cells were cultured under normal conditions, we found that cytosolic PAF-AH activity is PAF-AH 1 ␣ 1 /␣ 1 and ␣ 1 /␣ 2 (ϳ70%) Ͼ PAF-AH I ␣ 2 /␣ 2 (ϳ25%) Ͼ PAF-AH II (5%). When cells were deprived of serum and challenged with PAF, this activity profile shifted in favor of PAF-AH I ␣ 2 /␣ 2 (45%) Ͼ PAF-AH ␣ 1 /␣ 1 or ␣ 1 /␣ 2 (38%) Ͼ PAF-AH II (17%). Most surprisingly, pharmacological inhibition of ␣ 1 enzymatic activity had no effect on PAF-mediated cell death, whereas suppression of ␣ 2 induction by RNA interference significantly enhanced cell loss providing strong evidence for anti-apoptotic subunit specificity.
Second, PAF-AH I ␣ 2 is mobilized as part of an endogenous cell survival response. These findings complement studies documenting the anti-apoptotic actions of plasma PAF-AH and PAF-AH II (8, 25, 26, 29 -31). Moreover, we find that PAF-AH II is not able to compensate for a loss of PAF-AH I ␣ 2 function. In fact, we were surprised to find that a 55% reduction in PAF hydrolysis observed 75 min after PAF challenge following DFP treatment only moderately enhanced PAF-mediated cell death. The kinetics of PAF-AH activation may explain the lack of significant PAF-AH II protection. PAF-AH ␣ 2 is mobilized acutely in response to serum withdrawal, but an increase in DFP-or DTNB-sensitive PAF-AH II activity is only observed FIG. 7. BN 52021 and FR 49175 increase the relative ratio of PAF-AH I ␣ 2 to ␣ 1 expression following PAF challenge. PC12 cells were treated with PAF (1 M) following a 15-min preincubation in the presence or absence of BN 52021 (50 M), FR 49175 (50 M), or vehicle (0.1% Me 2 SO (DMSO)). Immunoblotting was performed for ␣ 1 (A) or ␣ 2 (B). BN 52021 and FR 49175 reduced ␣ 1 expression (A) without affecting the ␣ 2 induction (B). Blots were reprobed for actin as a loading controls (A and B). ␣ 1 (A) and ␣ 2 (B) protein levels were standardized to actin and expressed as a percent of the basal expression in cells cultured in complete media. C, the relative ratio of ␣ 2 to ␣ 1 protein following BN 52021 or FR 49175 revealed a stoichiometric increase in the levels of ␣ 2 compared with ␣ 1 relative to cells exposed to PAF in the absence of antagonist. after 30 -60 min of PAF exposure. Only PAF-AH I ␣ 2 was found to reduce apoptogenic concentrations of cytosolic PAF within the first 30 min of exposure. PAF-AH II is mobilized as part of a delayed cell survival program possibly to ensure that "subapoptotic" PAF concentrations are maintained.
Third, and perhaps most importantly, the anti-apoptotic response afforded by shifting the PAF-AH I ␣ 1 subunit expression in favor of ␣ 2 can be enhanced by PAF antagonists. Pharmacological suppression of ␣ 1 by BN 52021 or FR 49175 accelerates the kinetics of PAF deacetylation to lyso-PAF and protects cells from PAF-mediated apoptosis. Although it would at first appear counterintuitive that a reduction in PAF-AH subunit expression enhances PAF hydrolysis, these data are consistent with previous reports that BN 52021 inhibits PAF degradation under nonapoptotic culture conditions (48,50). PAF-AH I ␣ 2 /␣ 2 is known to hydrolyze PAF more efficiently than ␣ 1 /␣ 1 homodimers, and it is likely that reducing PAF-AH ␣ 1 coincident with PAF-AH ␣ 2 induction would promote more rapid formation of ␣ 2 /␣ 2 homodimers, enhance PAF hydrolysis, and increase cell survival. Most interestingly, this phenomenon models the graded reduction in ␣ 1 observed over the course of cerebral development by changing the predominant PAF-AH I catalytic composition from ␣ 1 /␣ 2 heterodimers in embryonic central nervous system to ␣ 2 /␣ 2 homodimers in adult brain (51). Our data suggest that this shift may render adult brain more resistant to PAF challenge than embryonic brain.
Summary-In summary, this study provides mechanistic ev-idence that sustained exposure to elevated intracellular PAF concentrations is sufficient to elicit apoptosis through a PAFRindependent cell death pathway. Previous studies have demonstrated that oxidative modification of membrane lipids during atherosclerotic injury, treatment with chemotherapeutic agents, ultraviolet B exposure, or excitoxicity is sufficient to produce PAF-like lipids (6,19,30,31,40). Our data suggest that these effects can occur in the absence of PAFR. To intervene in PAF-mediated death, we describe a novel anti-apoptotic function for PAF-AH I ␣ 2 and II, and we identify two PAF antagonists that accelerate cytosolic PAF-AH I activity to inhibit PAF-mediated apoptosis.