To hydrolyze or not to hydrolyze: the dilemma of platelet-activating factor acetylhydrolase.

Mounting ambiguity persists around the functional role of the plasma form of platelet-activating factor acetylhydrolase (PAF-AH). Because PAF-AH hydrolyzes PAF and related oxidized phospholipids, it is widely accepted as an anti-inflammatory enzyme. On the other hand, its actions can also generate lysophosphatidylcholine (lysoPC), a component of bioactive atherogenic oxidized LDL, thus allowing the enzyme to have proinflammatory capabilities. Presence of a canonical lysoPC receptor has been seriously questioned for a multitude of reasons. Animal models of inflammation show that elevating PAF-AH levels is beneficial and not deleterious and overexpression of PAF receptor (PAF-R) also augments inflammatory responses. Further, many Asian populations have a catalytically inert PAF-AH that appears to be a severity factor in a range of inflammatory disorders. Correlation found with elevated levels of PAF-AH and CVDs has led to the design of a specific PAF-AH inhibitor, darapladib. However, in a recently concluded phase III STABILITY clinical trial, use of darapladib did not yield promising results. Presence of structurally related multiple ligands for PAF-R with varied potency, existence of multi-molecular forms of PAF-AH, broad substrate specificity of the enzyme and continuous PAF production by the so called bi-cycle of PAF makes PAF more enigmatic. This review seeks to address the above concerns.


ATHEROSCLEROSIS AND INFLAMMATION
The incidence of CVD in individuals without hypercholesterolemia in the recent past provides a compelling reason to look beyond the traditional risk factors of atherosclerosis. Also, the failure of lipid lowering agents to effectively reduce the risk associated with CVD universally has prompted scientists across the globe to investigate atherosclerosis in a novel dimension, namely infl ammation (25)(26)(27). A common feature of the various factors causing atherosclerosis is oxidative stress ( 28 ). Lipid molecules bearing PUFAs esterifi ed to phospholipids are energetically favored targets for oxidation. Thus, LDL particles generally implied in CVD are oxidized and are no longer native to the body. It is the nature of the human body to effectively clear molecules of foreign nature. The innate immune system polices this task by mounting an infl ammatory response ( 25,26 ). In order to effectively eliminate oxidized LDL, monocytes/macrophages are recruited. It is during this phase that the trapped LDL undergoes further oxidation leading to endothelial dysfunction, formation of foam cells ( 26 ), and setting into motion a cascade of events leading to atherosclerotic streak formation. Macrophages, a prolifi c source of PAF-AH, may have a vital role in curtailing the PAF-R-mediated responses by hydrolyzing PAF-R ligands and other oxidized phospholipids during these events ( 29 ). In this regard, it is worth mentioning that paraoxonase-1 (PON-1), an esterase associated exclusively with HDL, was once claimed to hydrolyze oxidized phospholipids including PAF ( 30,31 ). Later studies conclusively demonstrated that neither PAF nor oxidized phospholipids are hydrolyzed by PON-1, but trace amounts of the plasma form of PAF-AH copurifying with PON-1 during isolation was responsible for the observed hydrolysis ( 32 ). Moreover, recombinant PON-1, though containing signifi cant esterase activities toward synthetic esters, is totally devoid of the claimed PAF or oxidized phospholipid hydrolyzing ability ( 33 ). What PON-1 does in HDL remains elusive.

BETWEEN THE SUBSTRATE AND PRODUCT OF PAF-AH
PAF and truncated oxidized phospholipids, including PAF mimetics, are among the commonly identifi ed substrates of the plasma form of PAF-AH ( 2-4, 7, 9, 32, 33 ). activity, which allows it to be constitutively active. Another feature of PAF-AH is its inability to hydrolyze intact long chain fatty acids at the sn -2 position of the phospholipids. This allows the plasma PAF-AH to circulate without causing any damage to the cellular components or to lipoproteins ( 2,7 ). With a better understanding of PAF biology and the process of infl ammation, it is now clear that the substrate specifi city of PAF-AH is vastly relaxed beyond PAF to an array of related molecules. An extreme example includes complex phospholipids such as long chain phospholipid hydroperoxides and isoprostane-containing PCs ( Fig. 2 ) ( 10 ). Such modifi ed phospholipids have been identifi ed in oxidized LDLs (11)(12)(13); in models of oxidative insult such as alcoholic blood ( 14 ), smokers blood ( 15 ), and electronegative LDLs ( 16 ); and in models of cutaneous infl ammation ( 17 ). Notable among these lipids are the oxidatively truncated phospholipids butanoyl/butenoyl PAF/PC ( 11 ), azelaoyl PAF/PC ( 18,19 ), palmitoyl glutaroyl PC, palmitoyl oxovaleryl PC ( 12 ), kodia PC ( 13 ), and many more ( 19 ) ( Table 1 ). Although PAF is recognized by PAF receptor (PAF-R) at subnanomolar concentrations ( 11,20 ), some of the above mentioned molecules also bind to PAF-R with Fig. 2. Chemical structure of esterifi ed F 2 isoprostane PC. A bulky sn-2 group derived from arachidonic acid is present in this phospholipid. PAF-AH can also recognize this as a substrate, but is a poor ligand to PAF-R. barrier through juxtacrine signaling ( 34,35 ). Being potent, PAF exerts its effects at subnanomolar concentrations through a single cell surface G protein-coupled receptor (GPCR) ( 11,20 ). Thus, not only PAF synthesis but also its subsequent hydrolysis to the biologically inactive PAF is an early mediator of infl ammation that is produced rapidly upon appropriate proinfl ammatory stimulus ( 1,2 ). PAF activates a variety of cells of the innate immune system where it enhances the migratory and adhesive behavior of these cells, enabling them to transmigrate the endothelial Free radical-mediated oxidative attack generates a host of truncated phospholipids. Oxidatively modifi ed LDL, apoptotic cells are rich sources of these phospholipids. All of them are sensitive to PAF-AH and also recognize PAF-R with varying potency. to oxidation of phospholipids or may reach to millimolar levels in the case of hyperlipidemic subjects ( 53,54 ). Most of this lysoPC is bound to albumin and to lipoprotein particles. However, it is important to note that the optimal concentration of lysoPC required to elicit reported biological responses is in the range of 10-50 M. Thus, the concentration of lysoPC in the plasma is already higher than its action range and further addition of lysoPC should logically be ineffective. As mentioned, lysoPC binds to albumin, other plasma proteins, and even to cells; the real free lysoPC concentration in vivo will always be lower than the measured concentration. Second, the proatherogenic properties of lysoPC probably stem from the experiments that utilized commercial preparations of the molecule that are likely to be contaminated with PAF or PAF mimetics ( 55 ). Given the high potency of PAF and trace amount of PAF contaminants, lysoPC may manifest many of the PAF actions through PAF-R. In fact, lysoPC responses are blocked by PAF-R antagonists ( 55,56 ). In a carefully performed study ( 55 ), it was proved that PAF present in trace amounts as contaminants in these commercial preparations was the reason for the infl ammatory properties of lysoPC/lysoPAF. Moreover, the credibility of lysoPC receptor is seriously questioned ( 57,58 ).

RISK FACTOR OR A RISK MARKER?
An important observation that came from WOSCOPS (West of Scotland Coronary Prevention Study) caused increased ambiguity concerning the anti-infl ammatory nature of the plasma PAF-AH. This study reported a correlation between elevated levels of PAF-AH and the severity of CVDs ( 59 ). On the contrary, additional trials failed to reproduce these fi ndings [reviewed in ( 28 )]. We believe that the increased level of the enzyme serves a protective function based on both in vitro and in vivo experiments and observation from PAF-AH-defi cient subjects. For example, retroviral introduction of the plasma form of PAF-AH reduces atherogenesis in a murine model ( 60 ). Endothelial cells exposed to electronegative LDL pretreated with PAF-AH were protected from undergoing apoptosis, suggesting again the protective role of PAF-AH ( 16 ). Elevating the circulating levels of PAF-AH by exogenous administration was also found to be benefi cial ( 61 ). Conversely, transgenic mice overexpressing the PAF-R exhibited increased bronchoconstriction ( 62 ) and susceptibility to develop spontaneous melanoma ( 63 ), while the PAF-R-null mice were less susceptible to systemic infl ammation and acid inspiration-induced lung injury ( 64,65 ). More importantly, in a recently concluded phase III STA-BILITY (Stabilisation of Atherosclerotic Plaque by Initiation of DarapladibTherapy) trial of 16,000 patients by GlaxoSmithKline involving a tightly controlled multi-center study with chronic coronary heart diseases, darapladib, a specifi c PAF-AH inhibitor, did not yield promising results. The drug failed to produce a statistically signifi cant improvement in the risk of heart attack, stroke, or death, though it added greater reductions for some of the secondary product, lysoPAF/lysoPC, by the action of PAF-AH is an important mechanism employed to prevent an exaggerated infl ammatory response, and hence PAF-AH is aptly termed as "signal terminator" ( 36,37 ).
Oxidative stress leads to an uncontrolled nonenzymatic chemical attack of the PUFAs in the phospholipids generating a pool of PAF mimetics and other oxidized phospholipids ( 11,20,38 ). Amid a sea of various other oxidation products produced during LDL oxidation, butanoyl and butenoyl species ( Table 1 ) account for the bulk of the PAF activity, and these PAF mimetics were identifi ed to be up to 1/10th as potent as PAF ( 11 ). In a case control study, mean plasma PAF levels of 23.8 pg/ml were reported in healthy subjects while CVD patients had elevated PAF levels of 49.7 pg/ml ( 39 ). Studies by Ninio and her coworkers detected the presence of PAF preferentially in intermediate LDL during LDL oxidation, where it reached up to 8.6 ± 5.7 pmol/mg in 3 h of oxidation ( 40 ). In another report, serum PAF levels were directly correlated with severity of anaphylaxis, where the PAF levels rose up to 805 ± 595 pg/ml in patients while control subjects had a basal levels of 127 ± 104 pg/ml ( 41 ).
The oxidized phospholipids could be a part of the cellular membrane or the lipoprotein particle. In either case, monitoring its effective removal by the action of PAF-AH is important due to the damage they cause by forming whiskers in membranes ( 42 ). Considering the proinfl ammatory properties of the substrates of PAF-AH, it appears important that the enzyme be maximally active to protect cells from uncontrolled oxidative/infl ammatory damage. In this regard, it is not surprising that the enzyme is constitutively active. Moreover, overexpression of the enzyme confers cell survival, which otherwise would undergo cell death in response to an oxidative insult ( 43 ). Consistent with the concept that this enzyme serves a protective role, some forms of PAF-AH even change their subcellular location to potentially allow the removal of PAF/PAF mimetics at the very site of their origin. For example, measurable translocation of the type 2b enzyme from cytosol to plasma membrane in models of oxidative stress has been shown ( 44,45 ).
LysoPC, the product of PAF-AH action when acyl PAF/ diacyl phospholipid oxidation products are the substrates, has gained much attention in the past ( 46 ). Although acyl PAF is a moderate PAF-R ligand (it is just several fold less active than its alkyl counterpart), its hydrolytic product, lysoPC, is not an effective PAF-R agonist. However, lysoPC is believed to possess a variety of activities such as induction of cytokine synthesis ( 47 ), augmented migration of monocytes ( 48 ), chemoattraction in smooth muscle cells ( 49 ), etc.; all processes are potentially proatherogenic. Presently, a proinfl ammatory (hence proatherogenic) property is also ascribed to lysoPC/lysoPAF that not only questions the very anti-infl ammatory nature of the PAF-AH, but also the proinfl ammatory nature of PAF/PAF mimetics ( 50,51 ). However, a "proinfl ammatory" status cannot be assigned to lysoPC for two simple reasons. First, the normal serum concentration of lysoPC is between 140 and 150 M ( 52 ); this value might increase by 40-50% owing infl ammation ( 77 ), for example the promoter of the PAF-AH gene is positively regulated by PAF and negatively regulated by interferon ␥ and lipopolysaccharide [reviewed in ( 28 )]. Because the fi nal levels of PAF-AH activity are due to the result of both positive and negative modulators, it is diffi cult to assign precisely whether PAF-AH is a risk factor or a risk marker. The elevated levels of the enzyme probably help in curtailing the ill effects of the increased PAF and oxidized phospholipids during atherosclerosis by hydrolyzing it to a less harmful lysoPC/lysoPAF. Therefore, the increased serum concentration of the enzyme is most likely to be a potential risk marker.

MAKING AND BREAKING PAF AND PAF MIMETICS: BI-CYCLE OF PAF AND PAF MIMETICS
Despite the fact that lysoPC is not proatherogenic and PAF-AH is not a true risk factor, whether PAF-AH is a friend or a foe is still a debated issue ( 78,79 ). It may be possible that overwhelming levels of acyl PAF and its mimetics may have a role in curtailing PAF-R activation. In fact, endothelial cells predominantly make acyl PAF when stimulated with proinfl ammatory agonists ( 80,81 ). In any cell engaged in PAF biosynthesis, acyl PAF constitutes a major part of the PAF pool under normal conditions and alkyl PAF accounts for a very minor fraction. This is also true in cell free systems undergoing oxidation, such as oxidation of LDL ( 11 ). The acyl PAF can also bind and activate PAF-R at high concentration and is hydrolyzed by the same PAF-AH that also hydrolyzes alkyl PAF ( 1,2,4 ). The difference in the relative abundance of the two species of PAF is subjected to the availability of the precursor molecules. It is now believed that PAF (acyl and alkyl) is continuously produced at basal levels irrespective of a stimulus due to membrane remodeling ( 82 ). But the continuous PAF synthesis is counteracted by constitutive PAF-AHmediated degradation ( 82 ). Also noteworthy is that acyl PAF is far less potent than alkyl PAF, and upon binding to PAF-R, presumably silences the cells. This is evident from the observation that PLA 1 treatment of a lipid extract possessing PAF-like activity increases its activity several fold after the removal of diacyl PCs including acyl PAF ( 11 ). These two features are among a variety of remarkable mechanisms that the nature employs to ensure homeostasis . Thus, acyl PAF and acyl PAF mimetics may act as biological antagonists of PAF-R under normal conditions to keep PAF-sensitive cells quiescent.
Making and breaking alkyl and acyl PAF can be thought of as a bi-cycle of PAF ( Fig. 3 ) involving two different cycles each producing the respective species of PAF. The rate of the reactions producing acyl PAF is higher than that of the alkyl PAF. Thus, it is apparent that under normal conditions, the two wheels of the PAF bi-cycle operate under different rates. Considering the potency of alkyl PAF acting at subnanomolar concentrations upon a suitable proinfl ammatory stimulus, it is obvious that the rate of production of alkyl PAF increases only by a small degree while the increased rate of acyl PAF production is endpoints . Currently, GlaxoSmithKline is thoroughly investigating this data for subgroup differences and an additional SOLID-TIMI 52 (Stabilization of Plaques using Darapladib-Thrombolysis in Myocardial Infarction 52) trial of 13,000 patients is ongoing. Better understanding of the complex infl ammatory disorders in general, especially those that involve the PAF signaling system, is a prerequisite prior to developing novel drugs targeting PAF-AH ( 66 ).
The relevance of PAF and PAF-AH in health and disease is observed not just in CVDs, but in a host of diseases affecting the respiratory, dermal, gastrointestinal, pancreatic, and other systems. Both elevated and decreased activities of PAF-AH have been reported in a variety of diseased conditions. For example, dynamic variations in endogenous levels of PAF-AH occur over time in both experimental sepsis and in critically ill sepsis patients ( 67 ). In one study involving genetically defi cient plasma PAF-AH mice, initially mice were protected from mortality when exposed to bacteria, but later developed signifi cant necrotizing enterocolitis when compared with wild-type mice ( 68 ). On the other hand, decreased levels of the enzyme are associated with a number of diseases such as asthma, systemic Lupus erythematosus , and Crohn's disease [reviewed in ( 4 )]. More than the murine examples, humans from Asian populations, who are defi cient in circulating plasma PAF-AH due to a common mutation at position 994 of the PLA2G7 gene resulting in valine to phenylalanine substitution (V 279 F), are at an increased risk to develop a range of infl ammatory disorders ( 69 ). Unexpectedly, using recombinant PAF-AH in sepsis patients did not decrease the mortality rate, as reported by Opal et al. ( 70 ) in their phase III clinical trial . This study was carried out in comparison with activated protein C (APC), the only critical care drug that was available for sepsis. However, the population that Opal et al. ( 70 ) studied was at a low mortality risk when compared with the previous APC study. Unfortunately, the recombinant form of APC (commercially sold as Xigris) that was in use until now, has recently been withdrawn, leaving the critical care physicians without a drug to treat sepsis ( 71 ). Outcomes from previous clinical sepsis trials using anti-PAF agents also were not of much promise. This has raised the question of whether the PAF signaling pathway should be targeted in sepsis ( 72 ). More recently other targets have also been identifi ed, for example GPCRs coupled to endothelial G ␣ q / G ␣ 11 signaling and spingosine-1-phosphate-dependent Gi signaling may play a critical role in mediating lethal responses to anaphylactic mediators such as PAF ( 73,74 ).
The plasma PAF-AH is susceptible to oxidant attack and suffers inactivation from modifi cation of the residues that contribute to enzymatic activity ( 75 ). Whether highly variable levels of circulating PAF-AH arise from oxidation in the general population is not currently known. In an elegant study ( 76 ), variation in PAF-AH levels is not due to variations in the effi ciency of transcription, translation, and mRNA stability, but due to the presence or absence of the N-terminal domain. Additionally, expression of plasma PAF-AH is transcriptionally modulated by the mediators of incompletely informed of the biology of this enigmatic molecule.
unlikely to make a signifi cant difference as far as PAF-R activation is concerned ( Fig. 3 ). Thus, the sheer abundance of acyl PAF enables it as the obvious target of PAF-AH action. Under normal conditions, though PAF-AH is engaged in hydrolyzing acyl PAF, low levels of alkyl PAF may transiently bind to the PAF-R. But, these low levels are unlikely to elicit an infl ammatory response . Upon proinfl ammatory stimulation, the rate of acyl PAF synthesis goes up as does the synthesis of alkyl PAF . Now PAF-AH is busy in hydrolyzing overwhelming levels of acyl PAF, leaving alkyl PAF to effectively bind to PAF-R, resulting in a transient infl ammatory response. Under oxidative insult conditions, free radical-mediated oxidative attack generates a plethora of oxidized phospholipids, all of which are PAF-AH substrates and some of which are also excellent PAF-R ligands (PAF mimetics). These oxidized phospholipids amplify the infl ammatory responses. Failure of the PAF-AH to effi ciently hydrolyze alkyl PAF and PAF mimetics due to their low abundance amid a pool of abundant acyl PAF and acyl PAF mimetics might lead to an exaggerated but unnecessary infl ammatory response. This is evident in murine models of infl ammation wherein massive amounts of recombinant PAF-AH (approximately 10-fold over the endogenous PAF-AH levels) had to be administered in order to generate a noninfl ammatory phenotype ( 36,60 ). A dedicated PAF synthase is yet to be described ( 83 ). Thus we propose that the ambiguity is not concerning the nature of the products formed from PAF-AH hydrolysis or the elevated levels of PAF-AH during CVD, but rather the heterogeneous group of molecules that can act as agonists/antagonists for the PAF-R and as substrates for PAF-AH. We remain Fig. 3. Bi-cycle of PAF. Resting cells make minute amounts of alkyl PAF (smaller circle) and relatively more acyl PAF (bigger circle) from alkyl/acyl PC and diacyl PC, respectively, via PLA 2 action. A nonspecifi c PAF synthase/acetyltransferase catalyzes the formation of acyl PAF and alkyl PAF. Continuous hydrolysis of both analogs by PAF-AH makes PAF undetectable in resting cells. Upon proinfl ammatory stimulus, relative abundance of alkyl/acyl PAF goes up. This elevated amount of PAF may not be susceptible for effective hydrolysis by PAF-AH leading to the accumulation of PAF. Although acyl PAF may act as PAF-R antagonist, transient increased accumulation of alkyl PAF over basal levels may evoke an infl ammatory response. Under oxidative insult conditions, oxidant attack (dashed lines) may inactivate PAF-AH, leading to generation of truncated oxidized phospholipids all of which are PAF-AH substrates and some of which are also excellent PAF-R ligands (PAF mimetics). Hence, PAF-AH is a real signal terminator and unlikely to be a risk factor.