These newly discovered pathways for NAE metabolism include: (a) members of the PLA/acyltransferase (PLA/AT) family; (b) multistep pathways via N-acylated lysophospholipid; and (c) NAAA. These enzymes/pathways were found in animal tissues as possible alternatives to Ca2+-dependent N-acyltransferase, NAPE-PLD, and FAAH, respectively.
PLA/AT family proteins as Ca2+-independent N-acyltransferases
In an attempt to identify Ca2+-dependent N-acyltransferase, we noticed that the reaction catalyzed by N-acyltransferase is similar to that catalyzed by lecithin retinol acyltransferase (LRAT). LRAT transfers an acyl chain from the sn-1 position of lecithin (PtdCho) to retinol (vitamin A alcohol), resulting in the formation of retinyl ester (the storage form of vitamin A). We speculated that the primary structure of N-acyltransferase might be analogous to that of LRAT. LRAT is a member of the NlpC/P60 thiol protease superfamily , and has considerable sequence homology with five members (HRASLS1–5) of the HRAS-like suppressor (HRASLS) family (Fig. 5) . HRASLS3 (also referred to as H-rev107) is a representative molecule of this family, and has been analyzed as a class II tumor suppressor that negatively regulates the activity of the oncoprotein Ras . Later, HRASLS1 (A-C1) , HRASLS2 , and HRASLS4 [tazarotene-induced protein 3 (TIG3)]  were also reported to have tumor-suppressing activity. As catalytically important residues, including cysteine and histidine as the putative catalytic dyad, are highly conserved among LRAT and HRASLS family members, HRASLS3 was earlier suggested to be an acyltransferase . However, the enzymatic properties of the family members were not examined until we reported that HRASLS5 (H-rev107-like protein 5) has an N-acyltransferase activity, forming NAPE by transferring an acyl group from PtdCho to PtdEtn [31, 82]. HRASLS5 was present mainly in the cytosolic fraction, and its N-acyltransferase activity was only slightly stimulated by Ca2+. Interestingly, the enzyme removed a fatty acyl group from both the sn-1 and sn-2 positions of the acyl donor PtdCho. These results strongly suggested that HRASLS5 is different from the known membrane-associated Ca2+-dependent N-acyltransferase. Our studies revealed that all other members of the HRASLS family (HRASLS1–4) also have N-acyltransferase activity [77, 83, 84]. Very recently, Golczak et al. also reported that purified HRASLS2, HRASLS3 and HRASLS4 generate NAPE from PtdCho and PtdEtn .
Figure 5. Primary structures of human LRAT and five members of the HRASLS (PLA/AT) family. Closed and shaded boxes indicate identity in all six and any four or five polypeptides, respectively. Dashes denote deletion of amino acids as compared with the other sequences. Asterisks indicate amino acids forming the catalytic triad.
Download figure to PowerPoint
Importantly, HRASLS1–5 showed not only N-acyltransferase activity, but also PLA1 and PLA2 (PLA1/2) activity, hydrolyzing the sn-1 and sn-2 ester bonds of PtdCho or PtdEtn, and O-acyltransferase activity, transferring an acyl group from PtdCho to the hydroxyl group of lyso-PtdCho. We thus proposed calling HRASLS1–5 PLA/AT-1–5 . PLA/AT-1, PLA/AT-2 and PLA/AT-5 showed relatively high N-acyltransferase activity relative to PLA1/2 activity, whereas PLA/AT-3 and PLA/AT-4 showed lower N-acyltransferase activity [77, 86]. When the homogenate of human embryonic kidney (HEK)293 cells, stably expressing PLA/AT-2, was examined, both the soluble and particulate fractions showed N-acyltransferase activity, with 1.4-fold higher activity in the soluble fraction. The N-acyltransferase activity was not stimulated by Ca2+ in either fraction. Thus, PLA/AT-2 was also presumed to be different from the known Ca2+-dependent N-acyltransferase.
Although PLA/AT proteins show N-acyltransferase activity in vitro, it remained unclear whether these molecules generate NAPE in living cells. When we metabolically labeled COS-7 cells transiently expressing recombinant PLA/AT-1–5 with [14C]ethanolamine, we found that PLA/AT-1, PLA/AT-2, PLA/AT-4 and PLA/AT-5 generated significant amounts of [14C]NAPE and [14C]NAE . Liquid chromatography–tandem MS demonstrated that the stable expression of PLA/AT-2 in HEK293 cells greatly increased endogenous levels of NAPEs and NAEs. Furthermore, additional expression of NAPE-PLD in the PLA/AT-2-expressing cells led to efficient conversion of the increased NAPE to NAE. RT-PCR revealed that human HeLa cells expressed endogenous PLA/AT-2, and the knockdown of this protein by small interfering RNA lowered the endogenous level of NAPE. Taken together, these results suggested that the PLA/AT proteins produce NAPE, which serves as a precursor of NAE in living cells. In the cells overexpressing recombinant PLA/AT proteins, the generation of NAPE proceeded without any cellular stimuli. However, we could not rule out the possibility that the N-acyltransferase activity of native PLA/AT proteins is regulated by intracellular Ca2+ or other signaling molecules.
We reported that PLA/AT-3 (H-rev107) functions mainly as PLA1/2, with low N-acyltransferase activity [83, 84]. However, Duncan et al. termed this molecule adipose-specific PLA2, and classified it as group XVI PLA2 (PLA2G16) . Recently, two groups demonstrated that PLA/AT-3 has PLA1 activity that is as potent as its PLA2 activity [85, 88]. PLA/AT-3 attracted attention because of its abundant expression in adipose tissue, its induction during adipose differentiation, and its ability to suppress lipolysis [89, 90]. We also showed that the overexpression of PLA/AT-3 in HEK293 cells caused dysfunction of peroxisomes and a remarkable decrease in ether-type triglycerides and plasmalogens . It remains to be determined whether the N-acyltransferase activity of PLA/AT proteins is related to their activities regarding tumor suppression, obesity, and dysfunction of peroxisomes. Recently, the N-terminal catalytic domain of PLA/AT-3 was characterized by solution NMR structure analysis  and X-ray crystallography [85, 88]. Together with site-directed mutagenesis studies, these studies increased our knowledge of the structure and function of PLA/AT-3, including the following: (a) the catalytic triad in the active site is composed of His23, His35, and Cys113 (His35 is replaced by asparagine in PLA/AT-1); (b) the acylation site in the acyl–protein complex is Cys113; and (c) the C-terminal transmembrane domain is required for the interfacial catalysis.
Multistep pathways via N-acylated lysophospholipid
Whereas NAPE-PLD directly releases NAE from NAPE, multistep pathways for the biosynthesis of NAE from NAPE via N-acylated lysophospholipid were also suggested in the 1980s (Fig. 2) . We showed that the secretory PLA2 isoforms of groups IB, IIA and V hydrolyzed NAPE to 1-acyl-lyso-NAPE, which was further converted to NAE by a membrane-associated lyso-PLD-like enzyme existing in brain and other tissues of rat . Although this ‘lyso-PLD’ was distinguished from NAPE-PLD by its catalytic properties, its further characterization has not been carried out. The presence of NAEs in the tissues of NAPE-PLD−/− mice clarified that NAE can be formed by an alternative pathway(s) in vivo . The double O-deacylation of NAPE via lyso-NAPE and further hydrolysis of the resultant metabolite glycerophospho-NAE (GP-NAE) to NAE and glycerol 3-phosphate were proposed as one alternative pathway (Fig. 2) . GP-NAE was actually detected in mouse brain tissue . In the analysis of NAPE-PLD−/− mice, we found a remarkable increase in endogenous brain levels of lyso-NAPE and GP-NAE as well as of NAPE . These results suggested that NAPE accumulates in the brain because of the deficiency of NAPE-PLD, and is degraded in the alternative pathway via lyso-NAPE and GP-NAE. By the functional proteomic isolation method with fluorophosphonate-biotin probe, α/β-hydrolase 4 [α/β-hydrolase domain-containing protein (ABHD)4] was demonstrated to be responsible for the double O-deacylation that generates GP-NAE from NAPE via lyso-NAPE . ABHD4 thus hydrolyzes both NAPE and lyso-NAPE by PLA1/2 (or PLB) and lysophospholipase activities. ABHD4 preferentially hydrolyzed lyso-NAPE among various lysophospholipids, including lyso-PtdEtn, lyso-PtdCho, and lysophosphatidylserine. Regarding the N-acyl species of lyso-NAPEs, ABHD4 did not distinguish between saturated and polyunsaturated acyl chains. Ser146, in the consensus sequence GXSXG, was presumed to be the catalytic nucleophile. It was recently reported that knockdown of ABHD4 inhibits anoikis (cell death in response to loss of cell–cell and cell–matrix interactions) in prostate epithelial cells . Its relationship to NAE metabolism is unclear.
Glycerophosphodiester phosphodiesterase 1 (GDE1) was shown to hydrolyze GP-NAE to NAE and glycerol 3-phosphate . On the basis of tissue distribution and catalytic properties, including Mg2+ requirement, the brain enzyme activity hydrolyzing GP-NAE to NAE was attributed to GDE1. Thus, the combination of ABHD4 and GDE1 was considered to form a NAPE-PLD-independent pathway. Initially, this pathway was expected to be responsible for anandamide formation, as the brain level of anandamide was not altered in NAPE-PLD−/− mice, in contrast to the decrease in the levels of saturated NAEs . However, neither ABHD4 nor GDE1 showed a preference for N-arachidonoyl species of lyso-NAPE and GP-NAE (precursors of anandamide). It remains unclear whether or not the ABHD4–GDE1 pathway is involved in the selective formation of anandamide. Alternatively, in our analysis of NAPE-PLD−/− mice, the brain level of anandamide was significantly lower, together with the levels of other NAEs .
GDE1 is an integral membrane glycoprotein, and was originally discovered as MIR16, a protein interacting with RGS16 (a regulator of G protein signaling) . Later GDE1 was shown to have a phosphodiesterase activity, preferentially hydrolyzing glycerophosphoinositol . GDE1−/− mice were born at the expected Mendelian frequency, were viable and healthy, and showed no abnormal cage behavior . Expectedly, the formation of NAE from GP-NAE or lyso-NAPE was hardly detected in the brain homogenates. However, endogenous brain levels of NAEs were not significantly different between the homozygous and heterozygous mice. Furthermore, no significant difference was seen between NAPE-PLD−/− mice and mice with double knockout of GDE1 and NAPE-PLD. These results suggest that enzymes or pathways other than NAPE-PLD and the ABHD4–GDE1 pathway are involved in NAE formation. Further analysis of GDE1−/− mice demonstrated that brain levels of glycerophosphoinositol, glycerophosphoserine and glycerophosphoglycerate were highly elevated . In agreement with these findings, the brain level of free serine, which should be released from glycerophosphoserine by GDE1, was significantly reduced.
The fact that the brain contains plasmalogen-type ethanolamine phospholipid (plasmenylethanolamine) in abundance  suggests that pNAPE also exists in the same tissue and serves as a precursor of NAE (Fig. 6). In rat brain, 65% of N-arachidonoylethanolamine phospholipids were of the plasmalogen type . In agreement with an earlier report using rat heart microsomes , recombinant NAPE-PLD hydrolyzed N-palmitoyl-plasmenylethanolamine to PEA at 70% of the rate of N-palmitoyl-PtdEtn hydrolysis . The brain levels of pNAPE and its lyso form (lyso-pNAPE) in NAPE-PLD−/− mice were much higher than those in wild-type mice, as shown by liquid chromatography–tandem MS. Furthermore, the brain homogenate of NAPE-PLD−/− mice converted pNAPE to NAE, and the homogenate also released NAE from lyso-pNAPE. As lyso-pNAPE has a vinyl ether bond rather than an ester bond at the sn-1 position, lysophospholipases such as ABHD4 should be inactive with this lysophospholipid. Therefore, it was likely that a lyso-PLD-type phosphodiesterase directly releases NAE from lyso-pNAPE. The lyso-PLD-type enzyme found in the brain was active at neutral pH, and converted N-palmitoyl-lysoplasmenylethanolamine, N-oleoyl-lysoplasmenylethanolamine and N-arachidonoyl-lysoplasmenylethanolamine to their corresponding NAEs at similar rates. The activity was stimulated by 2 mm Mg2+ and inhibited by 0.1% Triton X-100. We found that recombinant GDE1 showed weak lyso-PLD activity in hydrolyzing N-palmitoyl-lysoplasmenylethanolamine in addition to the aforementioned GP-NAE-hydrolyzing activity. As GDE1 is expressed in brain, GDE1 may be, at least in part, responsible for the brain lyso-PLD activity.
PLC-mediated hydrolysis of NAPE to NAE phosphate and dephosphorylation to generate NAE is another multistep route for the conversion of NAPE to NAE that does not go through lyso-NAPE (Fig. 2). Treatment of RAW264.7 mouse macrophages with LPS potently enhanced anandamide levels, despite the downregulation of NAPE-PLD [39, 105]. Protein tyrosine phosphatase, nonreceptor type 22 was identified as one of the gene products that increase the anandamide level, and exhibited a phosphatase activity that generates anandamide from anandamide phosphate. Src homology 2 domain-containing inositol-5-phosphatase 1 showed the same phosphatase activity. Although the PLC–phosphatase pathway was suggested to function in the brain of NAPE-PLD−/− mice , the PLC-like enzyme responsible remains uncharacterized.
We found a lysosomal enzyme hydrolyzing NAEs, first in CMK human megakaryoblastic cells , and later in the lung and other tissues of rats . cDNA cloning of this enzyme, referred to as NAAA, showed it to be a cysteine hydrolase belonging to the N-terminal nucleophile hydrolase superfamily (Fig. 7) [109-111]. No sequence homology was seen between NAAA and FAAH. Prior to our identification, NAAA was recognized as acid ceramidase-like protein from sequence homology . Acid ceramidase is a lysosomal enzyme that hydrolyzes ceramide to sphingosine and fatty acid. In agreement with its localization in lysosomes, NAAA is active only at acidic pH, and hydrolyzes various NAEs, with a preference for PEA in vitro. Human NAAA is a glycoprotein with four N-glycosylation sites [113, 114]. Like acid ceramidase , recombinant NAAA is produced as an inactive proenzyme, and is converted by autocatalytic cleavage between Phe125 and Cys126 to a catalytically active heterodimer composed of α and β subunits. The N-terminal 28 amino acids form a signal peptide, which is not contained in the α subunit . Although the β subunit was purified from rat lung , native αβ heterodimer has not been isolated. It remains unclear whether native NAAA stably exists as the heterodimer. On the basis of the sequence homology among the family members, the N-terminal cysteine of the β subunit (Cys126 in human NAAA) has been presumed to be the catalytic nucleophile. This was recently demonstrated by showing that β-lactones, which inhibit NAAA, acylate Cys126 [116, 117].
Figure 7. The primary structures of human NAAA and acid ceramidase. Asterisks and dots denote identity and similarity of the two sequences, respectively. Dashes indicate deletion of amino acids as compared with the other sequence. Diamonds and the circle indicate N-glycosylation sites and the catalytic nucleophile of NAAA, respectively. The arrow and the arrowhead denote a site of cleavage by signal peptidase (NAAA) and that between the α and β subunits (both NAAA and acid ceramidase), respectively.
Download figure to PowerPoint
NAAA is expressed in various human and rodent tissues, with predominant expression in macrophages [118, 119]. The expression of NAAA was highest in the prostate among various human tissues . Nonionic detergents (Triton X-100 and Nonidet P-40) and the thiol-reducing agent dithiothreitol have been used as NAAA stimulators in vitro. As endogenous stimulators, choline-containing or ethanolamine-containing phospholipids (PtdCho, PtdEtn, and sphingomyelin) and dihydrolipoic acid (the reduced form of α-lipoic acid) could substitute for Nonidet P-40 and dithiothreitol, respectively . These endogenous molecules may function by keeping NAAA active in lysosomes. As it is likely that FAAH and NAAA contribute to the degradation of NAEs in vivo, we investigated whether or not compensatory induction of NAAA mRNA occurs in the tissues of FAAH−/− mice. However, the expression levels in various tissues were not significantly different from those in wild-type mice .
Considering that NAAA preferentially hydrolyzes PEA over other NAEs, selective NAAA inhibitors that may increase local levels of endogenous PEA were expected to be anti-inflammatory and analgesic drugs . To date, many compounds have been reported to selectively inhibit NAAA (Fig. 8) [117, 123-130]. The most potent NAAA inhibitors thus far reported are lactone derivatives such as (S)-N-(2-oxo-3-oxetanyl)-3-phenylpropionamide [(S)-OOPP], (S)-N-(2-oxo-3-oxetanyl)biphenyl-4-carboxamide, and (2S,3R)-2-methyl-4-oxo-3-oxetanylcarbamic acid 5-phenylpentyl ester (URB913/ARN077), whose IC50 values were 420, 115 and 127 nm, respectively [125, 126, 128]. (S)-OOPP normalized the decreased PEA levels in carrageenan-stimulated leukocytes and LPS-treated RAW264.7 macrophage cells, and led to a reduction in neutrophil migration and inhibition of carrageenan-induced plasma extravasation . Recently, we reported that lipophilic amines such as pentadecylamine and tridecyl 2-aminoacetate inhibited NAAA with IC50 values of 5.7 and 11.8 μm, respectively, and showed much weaker effects on FAAH . These simple structures may provide a scaffold for further improvement.
Figure 8. Chemical structures of specific NAAA inhibitors. (A) Lactone derivatives. (B) Other compounds. IC50 values are shown in parentheses.
Download figure to PowerPoint