INTRODUCTION

Platelet-activating factor (PAF¹; 1-O-alkyl-2-acetyl-sn-glycero-3-phosphocholine) is a potent phospholipid mediator of inflammation (1, 2, 3, 4, 5) synthesized by many cell types, including macrophages, platelets, basophils, eosinophils, and endothelial cells on appropriate stimulation (6, 7, 8). It mediates a broad spectrum of biological activities, such as hypotension, smooth muscle contraction, and an increase in vascular permeability (9, 10, 11, 12). These actions of PAF are mediated mainly through specific cell surface receptors, although accumulation of intracellular PAF may also influence cell function (13, 14).

PAF is degraded by hydrolysis of the acetyl group at the sn-2 position of the glycerol backbone to produce the biologically inactive lyso-PAF and acetate. This reaction is catalyzed by a specific enzyme, PAF acetylhydrolase (15, 16, 17, 18). PAF acetylhydrolase is detectable in the cytosol of tissues and cells as well as plasma. The physiologic function of PAF acetylhydrolase has not yet been established, but several possible hypotheses have been proposed. Finally, PAF production may be regulated at the levels of both synthesis and degradation. This notion is based on the observation that PAF production was greatly enhanced in cells pretreated with phenylmethanesulfonyl fluoride, an inhibitor of intracellular PAF acetylhydrolase, on stimulation of platelets with thrombin (19). It was also demonstrated that plasma PAF acetylhydrolase effectively abolishes the inflammatory effects of PAF on leukocytes and the vasculature, indicating involvement of the enzyme in the maintenance of plasma PAF at certain levels (20). In addition to the degradation of PAF, PAF acetylhydrolase has the ability to hydrolyze short chain phospholipids and oxidized fragments of polyunsaturated fatty acids at the sn-2 position in a calcium-independent manner (21, 22, 23). From these biochemical properties, it was speculated that PAF acetylhydrolase may scavenge oxidized phospholipids produced inside or outside of the cells during oxidative stress. It is well known that during oxidation of low-density lipoprotein in vitro, significant amounts of lysophosphatidylcholine are produced in the low-density lipoprotein particles (23, 24, 25). Lysophosphatidylcholine may be formed by sequential oxidation of phosphatidylcholine in the surface coat of the low-density lipoprotein particle and subsequent hydrolysis of the oxidized phospholipids by plasma acetylhydrolase, which is usually associated with plasma lipoproteins (26, 27).

PAF acetylhydrolases from several sources have recently been purified and their cDNAs have been cloned. Plasma PAF acetylhydrolase is a 45-kDa monomeric enzyme (20) that is usually associated with plasma lipoproteins, such as low- and high-density lipoproteins (20, 26, 27). The predicted amino acid sequence deduced from its isolated cDNA is unique and unrelated to that of any known lipase or phospholipase (20, 28). We previously demonstrated that the PAF acetylhydrolase present in the soluble fraction of bovine brain cortex can be separated into three isoforms, designated isoforms Ia, Ib, and II (29). In contrast to plasma PAF acetylhydrolase, isoform Ib, which is the most abundant form in bovine brain, is a heterotrimeric enzyme composed of 29- (), 30- () and 45-kDa () subunits (29). A heterodimer of the  and  subunits forms a catalytic unit in the native complex. These two catalytic subunits are homologous with each other (63% identity), but share no homology with any other protein, including plasma PAF acetylhydrolase, except that a sequence of about 30 amino acids located 6 residues downstream from the active serine exhibits significant homology to the first transmembrane region of the PAF receptor (30, 31, 32). The  subunit has a 7 tandem WD-40 repeat (33, 34), which is often found in proteins that function through interaction with other protein components and is identical to the product of the causative gene for Miller-Dieker lissencephaly, a malformation of the brain cortex (35). Thus, this enzyme appears to play an important role in the signal transduction system for developement of the central nervous system. Very recently, we also succeeded in purifying isoform II and revealed that it differs from isoform Ib with respect to its polypeptide composition, substrate specificity, and tissue distribution, suggesting that it serves a different physiologic function from isoform Ib in tissues (36).

In this paper, we report the cloning of isoform II of tissue type PAF acetylhydrolase and demonstrate that it is distinct from isoform Ib but shows significant homology to the plasma PAF acetylhydrolase at the amino acid sequence level.

EXPERIMENTAL PROCEDURES

1-O-Hexadecyl-2-[³H-acetyl]-sn-glycero-3-phosphocholine was purchased from DuPont NEN (Boston, MA) and unlabeled PAF from Bachem Feinchemikalien AG (Bubendorf, Switzerland). Horseradish peroxidase-conjugated goat anti-mouse Ig polyclonal antibody and goat anti-rabbit Ig polyclonal antibody were obtained from Amersham Life Science. All other materials used were from Wako Pure Chemical (Osaka, Japan).

PAF acetylhydrolase II was purified from bovine liver as described previously (36). Approximately 0.4 mg of purified enzyme was reduced with 1 mg of dithiothreitol for 2 h at room temperature and then S-alkylated with 0.6% (w/v) 4-vinylpyridine for 2 h at room temperature. This reaction mixture was applied to a Sephadex G-25 gel filtration column equilibrated with 70% (v/v) formic acid. The protein solution was then dried using a Speedvac concentrator, and 0.2 ml of 70% formic acid solution containing 1% CNBr was added. The reaction mixture was incubated for 16 h at room temperature then dried again and dissolved in 0.1 ml of 70% formic acid. The fragments were applied to a reverse-phase high-performance liquid chromatography system with a Vydac 304-1251 C4 column pre-equilibrated with 5% acetonitrile containing 0.1% trifluoroacetic acid, and the peptides were eluted with a linear gradient of acetonitrile (5-70%, v/v) containing 0.1% trifluoroacetic acid. The amino acid sequence was determined using a Shimadzu automated sequencer. Lysyl endopeptidase fragments of purified isoform II were also prepared as described previously (36) and their amino acid sequence was determined as well as that of the cyanogen bromide fragment.

A monoclonal antibody against bovine isoform II was raised as follows. Immunization was performed by intrasplenic injection as described previously (37). Briefly, about 40 µg of antigen was resolved by SDS-PAGE using a 12% acrylamide gel and transferred to a 0.45 µm meshed nitrocellulose membrane. The membrane was then stained with Coomassie Brilliant Blue, and the band containing the antigen was cut out and homogenized in 800 µl of saline. Then 10 µg of antigen was injected into the spleens of BALB/c mice through a 26 gauge needle. After the intrasplenic injection, three further immunizations were performed by peritoneal injection every two weeks. Three days after the final immunization, 1.6 × 10⁸ spleen cells were fused with 4.0 × 10⁷ PAI mouse myeloma cells by treatment with 0.5 ml of 50% polyethylene glycol 1500 (Boeringer Mannheim) for 1 min at 37 °C. The cells were then washed with 10 ml of Dulbecco's modified Eagle's medium and resuspended in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and IL-6 as described previously (38). The cells were grown in 96-well culture plates. Ten days after fusion, the supernatants of the hybridomas were tested for production of anti-PAF acetylhydrolase II antibodies using an enzyme-linked immunosorbent assay. Positive cells were cloned by limiting dilution. One of the established monoclonal antibodies was named 7F7.

A polyclonal antibody against the  subunit of bovine isoform Ib was prepared as follows. The cDNA for the  subunit of isoform Ib was inserted into pET21a (Invitrogen) and transfected into BL21-competent cells. The recombinant  subunit was expressed in histidine-tagged fusion protein form. The fusion protein was then purified by manganese chelate affinity chromatography and hydroxylapatite column chromatography. One hundred micrograms of the purified recombinant protein was homogenized with an equal volume of Freund's adjuvant solution and injected intradermally into the back of an adult female New Zealand white rabbit. Two weeks after the first injection, a second injection was carried out using Freund's incomplete adjuvant. The third and fourth immunizations were carried out at 2-week intervals. One week after the final immunization, all blood was collected from the rabbit and the serum was used for polyclonal antibody analysis. The polyclonal antibody thus obtained was named 453.

A bovine kidney cDNA library was synthesized from poly(A) RNA using a cDNA synthesis kit from Life Technologies, Inc. The cDNA was ligated into pSPORT 1 and then transfected into Electro Max DH 10B-competent cells (Life Technologies, Inc.). The human brain library was obtained from Life Technologies, Inc.

The reverse transcription polymerase chain reaction (PCR) was performed to clone the cDNA. Two degenerated oligodeoxyribonucleotides, CKRTGNGGDATCCAYTC and CARGARGCNGARGARAC, were synthesized based on peptide sequences EWIPHR and EEAEAEET, respectively. These oligodeoxyribonucleotides were used as PCR primers. Initial screening of total bovine kidney cDNA revealed the presence of a 380-base pair amplified product. This PCR product was sequenced and used as a basis for the synthesis of two oligodeoxyribonucleotides, GAATGGATCCCCCACCG and CAAGAGGCAGAGGAGAC. These specific primers were used for further screening, using PCR as described previously (39, 40) with a slight modification. Briefly, the cDNA plasmid library was distributed into 96-well plates (2,000 clones/well), and the supernatants were pooled in every column and row. PCR was carried out with the specific primers to identify the wells containing cDNA for isoform II. The positive pools were then plated and screened by colony hybridization method (41). The colonies were transferred to a Hybond N⁺ nylon membrane (Amersham Life Science) and then treated as described previously (41). The PCR product was used as a probe. A fragment of the PCR product was subcloned into pUC18 (Pharmacia Biotech Inc.) and labeled with ECL^TM probe-amp reagents (Amersham Life Science). The hybridized probe was detected using the Fluorescein Gene Images^TM labeling and detection system (Amersham Life Science).

The DNA sequence was determined by the method of Sanger et al. (42) using a Taq Dyedeoxy Terminal Cycle Sequence kit and an Applied Biosystems model 373A DNA sequencer.

The coding region of bovine isoform II cDNA was used as a probe to isolate human isoform II cDNA clones by colony hybridization. After recloning into M13 mp19 vector, the cDNAs of positive clones were deleted using a double-stranded nested deletion kit (Pharmacia). DNA sequencing was carried out using a Taq Dyedeoxy Terminal Cycle Sequence kit and an Applied Biosystems 373A fluorescence DNA sequencer.

Full-length isoform II cDNA was subcloned into pcDNA3 (Invitrogen) using a restriction enzyme (EcoRV/NotI). COS7 cells (7.5 × 10⁶ cells), grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum, were collected in K-PBS (30.8 mM NaCl, 120.8 mM KCl, 8.1 mM Na₂HPO₄, 1.46 mM KH₂PO₄, and 5 mM MgCl₂) and then transfected with 30 µg of pcDNA3 containing the cDNA for isoform II by electroporation using a Genepulser system (Bio-Rad). After 48 h of culture in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, the cells were collected in phosphate buffered saline. The cells were washed twice with buffer A (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 250 mM sucrose), suspended in 500 µl of buffer A, and disrupted by sonication in a vessel surrounded by ice for five periods of 5 s at 15-s intervals using a Branson sonifier. The same procedure was carried out for control cells transfected only with pcDNA3. The cell lysates were assayed for PAF acetylhydrolase activity as described previously (36). The cell lysate was then centrifuged at 100,000 × g for 1 h to separate the soluble and pellet fractions. The pellet fraction was suspended in buffer A using sonication to disrupt the cells. Western blot analysis was performed on the soluble (16 µg of protein) and pellet (10 µg of protein) fractions of the transfected cells using the monoclonal antibody 7F7.

The test samples were resolved by SDS-PAGE on a 12% acrylamide gel by the method of Laemmli (43) and blotted onto a nitrocellulose membrane. The membrane was blocked with PBS containing 5% skim milk for 2 h at room temperature and then incubated in the supernatant of the 7F7 hybridoma or rabbit anti- subunit serum diluted to 1/1,000 in PBS containing 5% skim milk for 12 h at 4 °C. The filter was washed four times with PBS containing 0.05% Tween 20 (T-PBS) and incubated for 1 h with horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit Ig polyclonal antibodies, diluted to 1/2,000 in PBS containing 5% skim milk. After washing the filter six times with T-PBS, the blots were detected by an enhanced chemiluminescence method using an ECL^TM Western blotting detection set from Amersham Life Science. The results were visualized by fluorography using Hyperfilm^TM-ECL (Amersham Life Science).

Fresh bovine tissues were obtained from a local slaughterhouse and were processed within 3 h of slaughter. All procedures were carried out at 0-4 °C. The tissues were homogenized with a Waring blender in 2 volumes of buffer A, and then the homogenates were centrifuged at 10,000 × g for 30 min to remove the bulk of the solid material. The protein concentrations of the homogenates were determined using BCA protein assay reagent with BSA as a standard. Western blot analysis was performed on the tissue homogenate (100 µg of protein) with the monoclonal antibody 7F7 or the polyclonal antibody 453.

Nucleotide and predicted amino acid sequences were analyzed using GENETYX programs (Software Development, Tokyo, Japan). Computer searches for protein sequences were performed using the Genome Net World-Wide Web server (address, http://www.genome.ad.jp/).

RESULTS

As previously reported, the N-terminal amino acid sequence of the purified 40-kDa polypeptide could not be determined. One fragment obtained by digestion of the polypeptide with lysyl endopeptidase showed no perfect match with any sequence ever reported, indicating that this is a new enzyme (36). In this study, purified PAF acetylhydrolase II was again digested with lysyl endopeptidase and cyanogen bromide, and the peptide fragments were separated by reverse-phase high-performance liquid chromatography. Five new peptide sequences were determined.

Using the peptide fragment sequences as a base, degenerated oligonucleotides were synthesized and a series of PCRs was carried out, using bovine liver and kidney cDNA libraries as a template. One set of primers (CKRTGNGGDATCCAYTC and CARGARGCNGARGARAC) from a kidney cDNA library yielded an amplified product containing sequences corresponding to the peptides. This amplified product was used to screen a bovine kidney cDNA library to identify the clone containing the full sequence. A total of four independent clones were isolated. The clone possessing the longest insert was subcloned and sequenced. The nucleotide and predicted amino acid sequences of bovine PAF acetylhydrolase II cDNA are shown in Fig. 1. The full-length clone contained a 2.4-kilobase insert. The amino acid sequence deduced from the nucleotide sequence contained all the peptide fragment sequences. The ATG codon (Fig. 1, 1~3) was designated as the translated initiation codon, since no other ATG codon existed between the methionine codon and the first in-frame stop codon TGA (Fig. 1, 89~87). The cDNA contained an open reading frame encoding 392 amino acids, with a calculated molecular mass of 43,864 Da, which compares favorably with the molecular mass of 40 kDa estimated by SDS-PAGE. The predicted protein sequence has a Gly-X-Ser-X-Gly motif at Ser²³⁶, which is characteristic of lipases and serine esterases (44). The presence of this sequence explains the inhibition of the activity of this enzyme by the serine esterase inhibitor diisopropyl fluorophosphate (36). Replacement of Ser²³⁶ in the recombinant isoform II by Cys using site-directed mutagenesis yielded a protein with no enzyme activity (data not shown), supporting the hypothesis that this serine residue is a catalytic center.

The cDNA for human PAF acetylhydrolase II was subsequently cloned from a human brain cDNA library by colony hybridization using the coding region of bovine isoform II cDNA as a probe. The predicted amino acid sequence of human isoform II is shown in Fig. 2. The cloned human PAF acetylhydrolase II cDNA encoded the same number of amino acids and exhibited 83.5% nucleotide and 88% predicted amino acid identity to bovine isoform II.

pcDNA3 was used to express PAF acetylhydrolase II in COS7 cells. The cDNA generated a protein that migrated to the same position as purified PAF acetylhydrolase II when subjected to SDS-PAGE followed by Western blotting (Fig. 3B). The recombinant protein showed significant PAF acetylhydrolase activity (Fig. 3A). In contrast to the catalytic subunits of PAF acetylhydrolase Ib, both of which existed almost exclusively in the cytosolic fraction when expressed in COS7 cells (data not shown), recombinant isoform II existed in both the soluble and pellet fractions as determined by Western blot analysis (Fig. 3B).

Interestingly and unexpectedly, computer searches of protein sequences revealed that the predicted amino acid sequence for intracellular PAF acetylhydrolase II showed notable homology with plasma PAF acetylhydrolase (20) (Fig. 4). Human isoform II exhibited 43% predicted amino acid identity to human plasma PAF acetylhydrolase. A hydrophobic stretch comprising the first 20 residues of plasma PAF acetylhydrolase appears to form the signal sequence for secreted proteins. Mature protein purified from human plasma, however, starts with Ile⁴² or Lys⁵⁵, indicating cleavage by posttranslational modification or fragmentation during purification, although the mature protein showed full enzyme activity. These results indicate that the 54 N-terminal residues are not essential for expressing enzyme activity. Consistent with this, a mutation study also revealed that deletion of the first 53 amino acids rather enhances enzyme activity (28). When the amino acid sequences of both the plasma and intracellular enzymes were aligned using a computer program to maximize matching, intracellular isoform II was found to lack the first 50 residues of the plasma enzyme. Moreover, isoform II possessed no putative signal sequence motif in its N terminus, confirming the intracellular location of the enzyme. In human plasma acetylhydrolase, removal of 21 C-terminal amino acids caused only a slight loss of activity, but a 30-residue deletion reduced catalysis to below the detectable limit. Interestingly, as shown in Fig. 4, the homologous regions ended with 22 C-terminal amino acids (plasma) and 8 amino acids (intracellular isoform II).

The computer search also revealed that isoform II showed homology in three distinct regions (81~140, 200~260, and 321~380) with other proteins of mammalian and bacterial origin. A region surrounding Ser²³⁶ in the Gly-X-Ser-X-Gly motif of isoform II exhibited high homology with the active sites of other lipases and esterase, suggesting that it is a member of the serine esterase family. A highly homologous region was observed in the N-terminal half (101~142) of plasma and intracellular isoform II (Fig. 4), which is located upstream of the putative catalytic center of both enzymes. This region is included in the 81~140 region, which has some homology with lipases of bacterial origin (Fig. 5). In particular, the sequence LASXGFVV (124~131) of isoform II almost completely matched a sequence from plasma PAF acetylhydrolase and other bacterial lipases. Finally, the C-terminal region of isoform II showed some homology with other, apparently unrelated proteins, such as NK-4 protein (45) and dihydrolipoamide acetyltransferase (46) (Fig. 5). The latter enzyme is a subunit of the pyruvate dehydrogenase complex. The pyruvate dehydrogenase complex in all organisms studied to date is a large multi-enzyme complex that catalyzes the oxidative decarboxylation of pyruvate, the transfer of the acetyl unit to coenzyme A, and ultimately the reduction of NAD⁺. Dihydrolipoamide acetyltransferase catalyzes the second reaction. This enzyme also has some homology with isoform II around the catalytic serine residue. NK-4 protein was identified as the product of a gene the expression of which is increased after activation of T cells by mitogens or activation of NK cells by IL-2. The transcript, which encodes a protein with a mass of 27 kDa, appears to have a function common to the activation pathways of both NK cells and T cells, although its precise role in the cell remains unknown. The N-terminal region of this protein exhibits less, but significant, homology with the N-terminal region of isoform II.

Next, we examined the tissue distribution of isoform II by Western blot analysis using an established monoclonal antibody against purified bovine isoform II. As shown in Fig. 6, among the bovine tissues tested, isoform II was most abundant in the liver and kidney, although other tissues expressed detectable levels of the protein. We also performed Western blot analysis with a polyclonal antibody against the  subunit of isoform Ib. Significant levels of the  subunit were expressed in the brain, kidney, adrenal, ovary, and intestine. Interestingly, the catalytic subunit of isoform Ib was almost undetectable in bovine liver, in which isoform II is most abundant. These data thus indicate that these two intracellular PAF acetylhydrolases show a marked contrast in tissue distribution.

DISCUSSION

We previously demonstrated that intracellular PAF acetylhydrolase isoforms Ib and II are different from each other with respect to substrate specificity, tissue distribution, and polypeptide composition (36). In the current study, it was shown by cDNA cloning that these two intracellular PAF acetylhydrolases are also distinct from each other at the amino acid sequence level, indicating a different origin for the two enzymes. Interestingly, the amino acid sequence of isoform II exhibits 43% identity to that of plasma PAF acetylhydrolase. The plasma form of acetylhydrolase has been cloned from humans (20) and several other species (28). The human enzyme encodes a 441-amino acid protein that includes a predicted signal peptide for secretion. The deduced amino acid sequence is unique except for the Gly-X-Ser-X-Gly motif found around the active serine residue of most serine estrases and lipases. Ser²⁷³ of the human plasma PAF acetylhydrolase was identified as an active site. The active serine residue of intracellular isoform II was inferred to be Ser²³⁶, as this apparently corresponds to Ser²⁷³ of the plasma enzyme and the Gly-X-Ser-X-Gly motif occurs around this amino acid (Fig. 4). Complete loss of enzyme activity after mutation of this Ser residue to Cys also supports this hypothesis. The region surrounding the active serine residue (231~243 of isoform II) exhibited an almost complete match with plasma acetylhydrolase (Fig. 4). The active site of plasma PAF acetylhydrolase, as well as those of other lipases, forms a catalytic triad comprising the nucleophilic residues Ser, Asp, and His. Using site-directed mutagenesis, it was proposed that Ser²⁷³, Asp²⁹⁶, and His³⁵¹ form a catalytic triad in human plasma PAF acetylhydrolase. Although we have not yet confirmed this, Asp²⁵⁹ and His³¹⁴ of isoform II correspond to the amino acids comprising the catalytic triad of plasma PAF acetylhydrolase. Interestingly, the sequences surrounding these Asp and His residues exibit high homology between the plasma and intracellular enzymes, suggesting that the structures of the catalytic domains of these two enzymes are highly homologous. Crystallization and x-ray structural analysis of both enzymes should provide a definite conclusion.

The other region showing extensive homology between plasma and intracellular isoform II was observed in the N-terminal half (101~142; Fig. 4). A computer search showed that this region also exhibited some homology with lipases of bacterial origin (Fig. 5). Although the function of this region is unknown at present, it may be assumed to be involved in substrate binding or recognition of lipid-water interfaces, since it is relatively hydrophobic on hydropathy plot analysis (data not shown).

As described above, isoform II and plasma PAF acetylhydrolase appear to be typical members of the serine esterase family and are possibly derived from a common ancestral gene. In contrast, several features of isoform Ib, the other form of intracellular PAF acetylhydrolase, suggest that it does not belong to this family. First, it showed no significant homology with other lipases and esterases on computer search analysis. Second, it does not contain the typical Gly-X-Ser-X-Gly pentapeptide motif around its catalytic serine residue. We have recently succeeded in crystallizing the  subunit, one of the catalytic subunits of isoform Ib, and have revealed its structure on x-ray.² According to this analysis, Ser⁴⁷, Asp¹⁹², and His¹⁹⁵ form a catalytic triad. In most serine esterases, the Asp and His are separated by more than 20 amino acid residues (28, 42), whereas the distance in the  subunit is unusually short (only 4 residues). Finally, the tertiary folding of the catalytic subunit is strikingly similar to that of p21^ras and the GTP-binding domain of the  subunit of trimeric G proteins. All of these experimental data support the notion that isoform Ib is a novel serine esterase.

Western blot analysis of recombinant bovine isoform II expressed in COS7 cells revealed that the enzyme was distributed in the membrane fraction as well as in the soluble fraction in the transfected cells. In our preliminary experiments, endogenous isoform II was also found to be distributed in both cytosolic and membranous fractions in cultured MDBK cells. Although PAF acetylhydrolase activity in rat liver is highest in the cytosolic fraction, a significant level of activity was also detected in the light and heavy membrane fractions (16). Although the PAF acetylhydrolase activity present in membrane fraction has not yet been characterized, isoform II associated with the membrane fraction may at least in part contribute this activity. In contrast, the catalytic subunits of isoform Ib, which is expressed in MDBK cells, are exclusively located in the cytosolic fraction (data not shown). It is notable that the majority of PAF acetylhydrolase in human plasma is associated with plasma lipoproteins such as low- and high-density lipoproteins. Thus, both intracellular isoform II and plasma PAF acetylhydrolase appear to be relatively hydrophobic in nature and to have the ability to associate with membranous components inside or outside cells.

It has been proposed that one of the physiologic functions of plasma PAF acetylhydrolase is to degrade oxidized phospholipid produced in circulating lipoproteins (21). The substrate specificity of isoform II resembles that of plasma PAF acetylhydrolase, which can hydrolyze oxidatively fragmented fatty acyl chains attached to phospholipid molecules as well as to PAF. The liver, in which isoform II is most abundant, is known to lack the ability to produce PAF (4, 5), indicating that isoform II in liver cells is not involved in the regulation of PAF production. Several investigators have reported that oxidized phospholipids are preferentially hydrolyzed by cellular phospholipase A2 (47, 48, 49, 50, 51). Housekeeping protective enzymes are thought to be necessary both inside and outside cells, since reactive oxygen species can be formed both intracellularly and extracellularly. The molecular nature of the enzyme responsible for this activity, however, remains to be determined. Intracellular PAF acetylhydrolase II and plasma PAF acetylhydrolase may be possible candidates for this reaction.