The rise and fall of anandamide: processes that control synthesis, degradation, and storage

Abstract

Anandamide is an endocannabinoid derived from arachidonic acid-containing membrane lipids and has numerous biological functions. Its effects are primarily mediated by the cannabinoid receptors CB1 and CB2, and the vanilloid TRPV1 receptor. Anandamide is known to be involved in sleeping and eating patterns as well as pleasure enhancement and pain relief. This manuscript provides a review of anandamide synthesis, degradation, and storage and hence the homeostasis of the anandamide signaling system.

Introduction

The endocannabinoid anandamide (AEA) is an ethanolamide derivative of arachidonic acid (AA) that serves to activate primarily cannabinoid and vanilloid receptors. The resulting G-protein signaling initiates a number of biological pathways. The name given to it by its discoverers [1] is derived from the Sanskrit word ananda meaning bliss or joy in reference to their finding that this molecule competes with exogenous cannabinoids for their specific receptors in the brain, the first endogenous molecule found to do so. Subsequent research reveals that anandamide activates both CB1 in the central nervous system, CB2 in the peripheral tissues, including immune cells, and TRPV1, a non-selective cation channel ubiquitously expressed in all tissues. The resulting signaling involves a diverse set of biological results including sleep and eating patterns, short-term memory, mood, as well as modulation of the sensation of heat, acid, and proinflammatory stimulants. There are a number of excellent reviews on these subjects [2,3,4,5].

As is typical for signaling molecules, the in vivo concentration of AEA is maintained through the relative rates of synthesis and degradation, but also through a less common storage system, designed to meet high demand on short notice. This manuscript discusses the key enzymes in AEA homeostasis, in terms of structure, reaction specificity, enzymatic activity, regulation, and tissue and cellular expression patterns with a focus on the human isoforms involved.

Anandamide biosynthesis

AEA is synthesized through ligation of a previously membrane bound arachidonic acid and a membrane bound phosphatidylethanolamine forming an N-arachadonoyl phosphatidylethanolamine (NArPE). Release of AEA from the composite molecule is accomplished by a series of phospholipases (Fig. 1).

Fig. 1

Reactions involved in the formation of Anandamide Metabolites: AEA anandamide, glycero-p-AEA 1-acyl-sn-glycero-3-(N-arachidonyl) phosphoethanolamine, and NArPE N-acyl phosphatidylethanolamine, Enzymes: ABHD4 (lyso)-N-Acylphosphatidylethanolamine lipase, GDE1 Glycerophosphodiester phosphodiesterase 1, GDPD1,3 Lysophospholipase D isoforms 1 and 3, NAPE-PLD N-Acyl-Phosphatidylethanolamine-hydrolyzing Phospholipase D, PLAAT1,2,3,4,5 phospholipase A-Acyltransferases isoforms 1, 2, 3, 4, and 5, PLC phospholipase C, PTPN22 tyrosine-protein phosphatase non-receptor type 22, SHIP1 phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1, and sPLA₂ secreted Phospholipase A₂

Synthesis of N-acyl phosphatidylethanolamines (NAPEs)

Overview

The synthesis of NAPEs, including NArPE, is accomplished through the transfer of a fatty acyl group from a glycerophospholipid, phosphatidylcholine (PC) in particular, to the primary amine of a phosphatidylethanolamine (PE) by one of a series of N-acyltransferases that are either calcium-independent or calcium-dependent.

The calcium-independent group of N-acyltransferases are represented by the H-RAS-like suppressor (HRASLS) subfamilies of proteins and are members of the H-rev107 class of proteins. These proteins (PLA/AT subgroup) were originally classified as H-RAS-like tumor suppressors and later found to catalyze phospholipid hydrolysis and exhibit N-acyltransferase activity. All but PLAAT5 contains a putative transmembrane helix but is known to be membrane associated, nonetheless. Each of the enzymes catalyzes the hydrolysis of fatty acids from either the sn-1 or sn-2 position of glycerophospholipids, the sn-2 position being particularly important for anandamide synthesis, as arachidonic acid is typically found esterified to the sn-2 position in glycerophospholipids. Once the fatty acid is removed it is then a substrate for one of two secondary reactions, O-acyl transfer of the fatty acid to the free hydroxyl of a lysophospholipid or an N-acyl transfer to the primary amine on a phosphatidylethanolamine. These enzymes share a similar catalytic mechanism involving a thioester intermediate formed between an active site Cys and the fatty acid prior to transfer [6].

Phospholipase A and acyltransferase 1

There are two known splice isoforms of human phospholipase A/acyltransferase 1 (hPLAAT1, PLA/AT-1, UniprotKB-Q9HDD0) that are products of alternative splicing, Isoform 1 (Q9HDD0-1, PLAAT-1S, HRASLS, A-C1) and isoform 2 (Q9HDD0-2, PLAAT-1L) that contains an additional 105 residues—corresponding to exon 1—at the N-terminal [7]. Physical properties for all isoforms are given in the supplement Table S1. No coding single nucleotide polymorphism (SNP) variants are reported [8, 9] (https://genecards.org; https://www.ncbi.nlm.nih.gov/clinvar). Phobius web server predictions [10] (http://phobius.sbc.su.se/) suggest that both isoform 1 and isoform 2 are single-pass transmembrane proteins with large cytoplasmic N-terminal extracellular domains and short 8-residue lumenal C-terminal domains. There is some confusion in the literature regarding the expression of the two isoforms. Early work [11] suggests that Isoform 1 is the predominant species produced and is expressed in testes, muscle, brain, and heart. Later work suggests that only the longer Isoform 2 is the endogenous isoform in human tissue and is expressed in these same tissues, whereas both isoforms are known to coexist in murine tissues [7, 12]. The highest expression is found in the testis and skeletal muscle and it is ubiquitously expressed in most other tissues albeit at very low levels [13]. One difference between the two isoforms is the cellular distribution, where Isoform 1 is found in the cytosol, whereas Isoform 2 is found in the both the cytosol and the nucleus. The latter has been shown to be primarily associated with focal adhesion sites and in the cytosol to a lesser extent [13].

No X-ray structures have been reported for either isoform. There are, however, two potential structural model templates for both isoform 1 and 2 posted on the SWISS-MODEL website [14] (https://swissmodel.expasy.org/), template 4q95.1.A and template 2kyt.1.A. Predicted posttranslational modifications (PTMs) are given in the supplement Table S1 [15,16,17] and no confirmed sites have been reported.

Originally designated as a class II tumor suppressor protein, hPLAAT1 was later found to be involved in calcium-independent phospholipid metabolism with both lipase and acyltransferase activity [13]. The enzymatic activity of both isoforms is essentially the same [7]; however, most of the reported data refer to Isoform 1 (Table 1) [7, 11, 18]. There are three enzymatic activities to consider for this enzyme: 1) hydrolysis at the sn-1 position, 2) hydrolysis at the sn-2 position, and 3) the N-acyltransferase activity. Hydrolysis at the sn-1 position occurs at 7–9 times the rate for the sn-2 position and the hydrolysis rate of the arachidonic acid ester is slightly slower than observed for the palmitic acid ester (15–35% slower) [11]. The N-acyltransferase rate is 1.6–2.5 times faster than the overall phospholipase activity. Sequence comparison to other family members indicate that C119 is likely the active site nucleophile.

Table 1 Enzyme Activity for enzymes involved in the formation of N-acylphosphatidylethanolamine (NAPE)

Regulation of the hPLAAT1 phospholipase or acyltransferase activity by expression or by posttranslational modification has not been reported. However, it has been shown that this protein is significantly downregulated in tumor cells by unknown processes [19].

Phospholipase A and acyltransferase 2

Human phospholipase A and acyltransferase 2(hPLAAT2, PLA/AT-2, HRASLS2, UniprotKB-Q9NWW9) presents as a single isoform with a sequence homology to PLAAT1 isoform 1 of 47.6%. Physical properties are given in the supplement Table S1. No coding SNP variants are reported [8, 9]. As expected, structure predictions [10] suggest that it is a single-pass membrane protein with a large cytoplasmic N-terminal domain (residues 1-129) and a small lumenal C-terminal domain (residues 156-162). One partial X-ray structure is reported (PDB entry DPZ). hPLAAT2 is expressed primarily in the GI tract and blood but is also found ubiquitously expressed in most other tissues [13]. The cellular distribution is localized to mitochondria. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Also designated as a tumor suppressor protein, hPLAAT2 exhibits a calcium-independent N-acyltransferase reaction that is about 25% of the combined sn-1 and sn-2 phospholipase (PLA1/2) reaction rate in whole cells, whereas in cell homogenates the opposite is true. Hydrolysis rates for the sn-1 position compared to the sn-2 position vary from 0.7 to 1.9 (Table 1) [18, 20, 21]. Structural analysis reveals that C113 is likely the active site nucleophile [20]. Regulation of phospholipase or acyltransferase activity by expression or by posttranslational modification has not been reported.

Phospholipase A and acyltransferase 3

Human phospholipase A and acyltransferase 3 (hPLAAT3, PLA/AT-3, HRASLS3, AdPLA, PLA2G16, H-Rev107, UniprotKB-P53816) presents as a single isoform with a sequence homology to PLAAT1 isoform 1 of 46.8%. Physical properties are given in the supplement Table S1. No coding SNP variants are reported [8, 9]. Structure predictions [10] suggest that it is a single-pass membrane protein with a larger cytoplasmic N-terminal domain (residues 1-132) and a small lumenal C-terminal domain (residues (155-162). Several partial X-ray structures are reported (e.g., PDB entry 4DOT). PLAAT3 is expressed primarily in the brain, adipose tissue, and blood, but is also found ubiquitously expressed in most other tissues [13]. Its cellular distribution appears to be primarily localized to the mitochondria [13] but has also been reported as cytosolic [22] and possibly plasma membrane and peroxisome membrane associated by comparison to the rat and mouse counterparts [21]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Also designated as a tumor suppressor protein, hPLAAT3 exhibits a calcium-independent activity. PLA_1/2 reaction rates (Table 1) are faster than the N-acylation reactions by 10–200-fold [18, 20, 21, 23]. There is a preference for the sn-1 site that varies from 1.2 to 14 fold (Table 1) and it exhibits tenfold faster reactions with PC substrates as compared to PE substrates [23]. Mutagenesis studies reveal that C113, H23, and H35 are necessary for activity [24]. Regulation of phospholipase or acyltransferase activity by expression or by posttranslational modification have not been reported. However, although PLAAT3 is active in the absence of intracellular calcium [Ca²⁺]_i, V_max for lipase activity increases up to two fold in the presence of calcium, suggesting that this enzyme is a special kind of calcium-dependent lipase [25]. Downregulation of the PLAAT3 gene through promotor methylation in cancer cells is well documented [6]. Regulation of phospholipase or acyltransferase activity by expression other than in cancer or by posttranslational modification has not been reported.

Phospholipase A and acyltransferase 4

Human phospholipase A and acyltransferase 4 (hPLAAT4, PLA/AT-4, HRSL4, RARRES3, HRALSL4, TIG3, UniprotKB-Q9UL19) presents as two isoforms with a sequence homology to PLAAT1 isoform 1 of 56.6% for Isoform 1 (Q9UL19-1) and 51.2% for isoform 2 (Q9UL19-2). Physical properties for all isoforms are given in the supplement Table S1. Two coding SNP variants are reported for the canonical isoform, V69L and A162V [8] but no biological ramifications are noted. Phobius web server predictions [10] indicate that Isoform 1 is a single-pass transmembrane protein with a large cytoplasmic N-terminal domain (residues 1-136) and a short 8-residue lumenal C-terminal domain, and Isoform 2 shows no transmembrane domains. Two NMR structures are reported for residues 1-125 of Isoform 1 (PDB entry 2LKT and 2MY9). The SWISS-MODEL website [14] selects the same model protein for both isoforms (2LKT), suggesting similar structures. PLAAT4 is highly expressed in the kidney, thyroid gland, GI tract, and blood, but is also found ubiquitously expressed in most other tissues, including the brain [13]. Its cellular distribution appears to be primarily localized to intracellular membranes [13]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally. PLA_1/2 reaction rates for purified protein are 50-fold faster than that for the N-acylation reactions with a threefold preference for the sn-1 site [18, 20, 23] (Table 1) and exhibits twofold faster reactions with PC substrates as compared to PE substrates [23]. Regulation of phospholipase or acyltransferase activity by expression or by posttranslational modification has not been reported. However, upregulation of hPLAAT4 mRNA indirectly in tumor cells by all-trans retinoic acid has been documented [26].

Phospholipase A and acyltransferase 5

Human phospholipase A and acyltransferase 5 (hPLAAT5, PLA/AT-5, HRSL5, HRALSL5, HRLP5, iNAT, UniprotKB-Q96KN8) present as three very similar isoforms with homologies to Isoform 1 of 96.7% and 96.4% for Isoform 2 and Isoform 3, respectively, with only 22.3% homology to PLAAT1 Isoform 1. Physical properties for all isoforms are given in the supplement Table S1. Four coding SNP variants are reported for isoform 1, S31G, A93P, Q214R, and A258V [8] but the clinical significance of these coding SNPs is unknown. Phobius web server predictions [10] indicate that no transmembrane domains exist. No X-ray structures are available, however, there are two potential structural model templates on the SWISS-MODEL website [14], template 2kyt.1.A and template 4q95.1.B. hPLAAT5 is highly expressed in the testis, granulocytes, pancreas, and expressed at lower amounts throughout the brain, and appears to be localized to the plasma membrane [13]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

PLA_1/2 reaction rates for the purified protein are about 25% of that for the N-acylation reactions [22, 27] (Table 1). Comparison of positional preferences for deacylation have not been reported. Regulation of phospholipase of acyltransferase reactions by expression or by posttranslational modification has not been reported.

Cytosolic phospholipase A2 epsilon

Human cytosolic phospholipase A2 epsilon (hPLA2G4E, Phospholipase A2 group IVE, UniprotKB-Q3MJ16) is a phospholipase with a calcium-requiring N-acyl transferase activity [28, 29]. As reported by Hussain et al. [28] hPLA2G4E is expressed in two isoforms. Isoform 2 is identical to Isoform 1 apart from the first 376 residues that are missing in this isoform, including a membrane associating C2 domain. Physical properties for all isoforms are given in the supplement Table S1. There are four coding SNP variants reported for isoform 1: R81Q, H213Y, K292N, N400S, M573V and A693T [8, 9]. All coding SNPs are reported to be benign. Phobius web server [10] analysis indicates a 70–80% likelihood for a single transmembrane sequence in each of the isoforms. No X-ray structures are available; however, there is one potential structural model template suggested on the SWISS-MODEL website [14] is template 5iz5.1.A. hPLA2G4E is expressed primarily in the skin and although found in other mammal brain tissue it has only been identified in the human pituitary tissue [13]. Sequence comparison to the murine enzyme indicates that S412 is the likely active site nucleophile. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Catalytic activity for both purified recombinant and recombinant cell lysates have been reported in terms of overall production of N-acyl fatty acid products [28, 29]. Catalytic activity is enhanced in the presence of phosphatidylserine (PS) by 25-fold in the presence of saturating amounts of calcium ion and the EC₅₀ for calcium is increased greater than eightfold in the presence of PS [28]. Other anionic phospholipids enhance the activity in the following order PS > phosphatidylinositol > phosphatidylglycerol > phosphatidic acid > phosphatidylcholine > phosphatidylinositol-4,5-bisphosphate [29]. Regulation by expression or by posttranslational modification has not been reported; however, phosphorylation of S800 is predicted by sequence comparison to the murine isoform, suggesting the possibility of regulation by phosphorylation.

The reaction has been characterized in the greatest detail for the recombinant murine isoform (UniProtK-Q50L42) with a 77.4% sequence homology to the human isoform. The brain protein expressed in HEK293 cells shows a strong preference (3.3:1 in mouse brain lysate) for transferring the sn-1 O-acyl chain to PE [30] and exhibits little PC-lipase activity in the absence of PE. The enzyme is strongly activated with increasing Ca²⁺, over 200-fold in cell lysates with recombinant protein and fourfold in murine brain tissue with maximal activity at 10 mM Ca²⁺. Catalytic activity is nearly fourfold higher in membrane fractions than cytosolic fractions obtained from transfected HEK293 cells suggesting that it is membrane associated. The N-terminal region of the murine protein does contain a C2 domain which upon calcium binding is known to target the protein to phospholipid membranes [31, 32]. However, unusual for a PLA₂-class enzyme, the murine protein appears to utilize an alternative method for membrane targeting where a region of the C-terminus containing a number of positively charges amino acid residues bind to the phosphate of several membrane phosphoinositides [33]. Characterization of a S420A mutant shows that S420 is the active site nucleophile. Phobius web server [10] analysis shows a 70% likelihood for a single transmembrane helix near the C-terminal end. The fact that the murine enzyme shows a 77.4% sequence homology with the human form strongly suggests similar structure and enzymatic activity.

Converting N-acylphosphatidylethanolamines (NAPEs) to N-acylethanolamines (NAEs)

Overview

Converting NAPEs to N-acylethanolamides (NAEs) such as AEA is accomplished through several pathways (Fig. 1). The dominant pathway (75%) [34] is catalyzed by a phospholipase D (NAPE-PLD) where a NAPE is converted directly into the corresponding NAE and a phosphatidic acid.

N-Acyl-phosphatidylethanolamine-hydrolyzing phospholipase D

Human N-acyl-phosphatidylethanolamine-hydrolyzing phospholipase D (hNAPE-PLD, UniprotKB-Q6IQ20) presents as a single isoform and is a member of the metallo-β-lactamase family. Physical properties are given in the supplement Table S1. Two coding SNP variants are reported: S152A and D389N [8]. Clinical ramifications for these SNPs are not reported. hNAPE-PLD is a cytosolic, zinc-requiring protein that is ubiquitously expressed in most tissues with particularly high expression in the brain, GI tract, kidney, and bladder [13]. One X-ray structure is available (PDB entry 4QN9) [22]. Acetylation on the N-terminal methionine has been reported [35]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Enzymatic activity of recombinant proteins for mouse, rat and human proteins have been reported for a variety of NAPE substrates (Table 2) [36, 37]. The mouse and rat proteins have specific activities that are an order of magnitude higher than the human enzyme against N-palmitoyl-phosphatidylethanolamine substrates. The active enzyme acts as a homodimer and requires the presence of a hydrophobic cofactor, a bile salt or n-octyl-glucoside [22, 37, 38]. Different bile acids enhance the reaction rate to different degrees depending on the location of hydroxyl groups on the steroid structure [38], suggesting that the enzyme activity may be modulated in vivo by the presence of specific bile acids. The enzyme also hydrolyzes N-arachidonoyl-phosphatidylethanolamine (NArPE) substrate 7 times faster than N-palmitoyl- phosphatidylethanolamine substrate in the presence of cholic acid (Table 2). Recently it was reported that a murine NAPE-PLD hepatocyte deficient strain develops a high-fat diet-like phenotype and that is much more sensitive to liver inflammation, likely due to the associated reduction in almost all endocannabinoid mediators [39]. The authors hypothesize that NAPE-PLD acts as a “hub” that controls a large array of liver lipids.

Table 2 Enzyme activity for enzymes involved in the degradation of N-acylphosphatidylethanolamine (NAPE)

Regulation via phosphorylation has not been reported; however, there is one report indicating a 4–14 fold increase in hNAPE-PLD at the protein level in lymphocytes taken from malignant endometrial carcinoma (EC) patients in an advanced disease state when compared to peripheral lymphocytes and without significant change in the transcription level [46]. This correlates well with the fact that N-acylethanolamine concentrations are higher in post-menopausal women with EC compared when with controls. A specific mechanism has not been proposed.

(lyso)-N-acylphosphatidylethanolamine lipase (ABHD4)

Human (lyso)-N-Acyl-phosphatidylethanolamine lipase (hABHD4, alpha/beta- hydrolase 4, UniprotKB-Q8TB40) is expressed as two splice isoforms. Physical properties for all isoforms are given in the supplement Table S1. No coding SNP variants are reported [8, 9]. Only Isoform 1 (hABHD4) will be discussed here. hABHD4 is a member of the peptidase S33 family, ABHD4/ABHD5 subfamily. No X-ray structures or posttranslational modifications have been reported. There are two potential structural model templates suggested on the SWISS-MODEL website [14], template 3nwo.1.A and template 5mxp.1.B. hABHD4 is produced ubiquitously in all tissue types studied, including brain, and is localized to the nucleoplasm with cytoplasmic expression in some tissues [13]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally. However, the rat isoform (UniprotKB-Q6QA69) with 94.5% sequence homology is known to be phosphorylated at S124 which corresponds to S122 on the human isoform [47].

The enzymatic activity of hABHD4 has not been reported on to date. However, there are several studies available for the mouse counterpart which shares a 96.5% sequence identity to the human protein suggesting a similar value for the human isoform (Table 2) [34]. ABHD4 catalyzes the removal of fatty acids from NAPE and the product 1-acyl-sn-glycero-3-(N-acyl) phosphoethanolamine (lyso-NAPE) and does so using an active site serine as the nucleophile that reacts readily with methoxy arachidonyl fluorophosphonate (MAFP) inhibitor. Only the latter reaction has been examined directly and only as the combined activity of ABHD4 and glycerophosphodiester phosphodiesterases (GDEs) [40]. Comparison of data from NAPE-PLD^−/− mice in the presence and absence of inhibitors of either the ABHD4/GDE arm or the PLC/phosphatase arm of the alternative pathways provides insight regarding the regulation of these two pathways [34]. By comparing results from 1 min incubation to 1 hr it is clear that the PLC/phosphatase pathway is responsible for a rapid production of AEA when the NAPE-PLD is blocked and that the ABHD2/GDE pathway is upregulated in response to low AEA. Comparison of data for the conversion of NAPE and lyso-NAPE to AEA [34, 40] also indicates that the ABHD2/GDE pathway exhibits a higher specific activity than the PLC/phosphatase pathway and is thus the dominant contributor to AEA production once upregulated.

Phospholipase C (PLC) and secreted phospholipase A₂ (sPLA₂)

Phospholipase C (PLC) represents a class of phospholipases that cleave the phospho-head group from phospholipids, the most studied of which involve the production of inositol 1,4,5-trisphosphate from a 1,2-diacyl-glycero-3-phospho-(1D-myo-inositol-4,5-bisphosphate). The production of phospho-AEA (pAEA) from NAPE has been observed in mouse brain homogenates, clearly involving a member of the PLC class [34, 45]. The identity of the particular isoform has yet to be reported. Secreted rat phospholipase A2 group IB (rat sPLA2-IB), secreted human calcium-dependent phospholipase A2 Group V (hPLA2G5) and human phospholipase A2 Group IIA (hPLA2G2A) all are known to produce lyso-NAPE from NAPE [48]. Specific activities for the human isoforms are given in Table 1.

Human phospholipase A2 Group IIA (hPLA2G2A, phospholipase A2, UniProtKB-P39877) is presented as a single isoform and is a secreted, calcium-dependent lipase. Physical properties are given in the supplement Table S1. There are 14 coding SNP variants reported [9]. Five of these are associated with familial fleck retina (G45C, G49S, Q128 frame shift, and two termination mutants W62 and R53) and the remaining with unknown significance (P6S, S14T, R73C, C103R, Y116C, R123Q, Q130C, Y131C and N134K). Its expression is limited but is found in the highest amounts in the liver followed by all muscle tissue, GI tract and female tissues [13]. There are numerous X-ray structures available (e.g., PBL entry 1DB4). Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Human phospholipase A2 Group V (hPLA2G5) is a secreted, calcium-dependent lipase. It is translated as 138 residue polypeptide that is processed to 118 residues following removal of the signal peptide. Physical properties are given in the supplement Table S1. There are 16 coding SNP variants reported [9]. Four of these are associated with familial fleck retina (G45C, Q128 frame shift, and two termination mutants W62 and R53), one is associated with late onset retinal degeneration (V99 frame shift), and the remaining with unknown significance (P6S, S14T, G45S, G49S, R73L, E76 deletion, C103R, Y116C, R123Q, Q130C, Y131C, and N134K). hPLA2G5 is expressed in most tissues but is highly expressed in the retina as well as heart and smooth muscle [13]. X-ray structures are not available, but a 3D model has been published (PDB entry 2GHN). Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Human phospholipase A2 Group IB (hPLA2G1B, PLA2, PLA2A, PPLA2, UniProtKB-P04054) is a secreted, calcium-dependent lipase. Four SNP variants are reported (R122H, D16A, N89T, and N89K) none of which has any known clinical significance [8, 9]. hPLA2G1B is expressed exclusively in the pancreas [13]. Two Xray (PDB entry 3ELO and 6Q42) and one model structure (PBD entry 1YSK) are available. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Specific activities have yet to be reported for purified hPLA2G2A, hPLA2G5 or hPLA2G1B. However, reaction rates of recombinant forms expressed in HEK293 cells have been reported as 0.37 nmol/min/10⁶ cells for hPLA2G2A and 5.5 nmol/min/10⁶ cells for PLA2G5 [48]. Regulation of hPLA2G5 expression is reduced in mast cells when treated with vitamin D (1,25-dihydroxyvitamin D₃) a known modulator of the inflammatory response, whereas other PLA2 isoforms are upregulated [49]. Regulation involving anandamide metabolism by expression or by posttranslational modification for any of these isoforms has not been reported.

Lysophospholipase D GDPD1

Lysophospholipase D GDPD1 is known to catalyze two reactions leading to the formation of AEA and other NAEs: (1) hydrolysis of lyso-NAPE and (2) hydrolysis of glycerophospho-N-acylethanolamine, each producing the same amide product [42]. Human lysophospholipase D GDPD1 (hGDPD1, GDE4, glycerophosphodiester phosphodiesterase 4, UniprotKB-Q8N9F7) is expressed as three splice isoforms. Physical properties for all isoforms are given in the supplement Table S1. Only the canonical sequence, Isoform 1, will be discussed here. No coding SNP variants are reported [8, 9]. hGDPD1 is a metallo-membrane protein containing two transmembrane helices with a preference for Mg²⁺ or Mn²⁺ [42, 50] and is inhibited by both Ca²⁺ and EDTA [44]. hGDPD1 is widely expressed in many tissues with the highest expression in the brain followed by a lower expression in endocrine tissue, the GI tract, and male and female tissues [13]. X-ray structures are not available, but a 3D model has been published [50]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Specific activities have been reported for both overexpressed mouse and human recombinant GDPD1 obtained from cell lysates (Table 2) [42]. These two enzymes share a 91.9% sequence homology. Both enzymes show a preference for N-acyl-phosphatidylethanolamine substrates over phosphatidylcholine substrates. However, for mGDPD1 the opposite is true for non-N-acylated substates [41]. Specific activities for both enzymes for the same substrates follow the following order: glycerophospho-N-palmitoylethanolamine (GP-NPE) > N-palmitoyl-lysophosphatidylethanolamine (N-Pal-lyso-PE) ≈ N-oleoyl-lysophosphatidylethanolamine (N-Ole-lyso-PE) > N-arachidonoyl-lysophosphatidylethanolamine (N-Ara-lyso-PE), all of which are orders of magnitude slower than the corresponding NAPE-PLD pathway competitor. Purified recombinant mGDPD1 shows a 5000-fold increase in specific activity for GP-NPE over that obtained from cell lysates with similar specific activities for lyso-PC, lyso-PE, and N-Pal-lyso-PE [44].

Although a specific catalytic mechanism has not been proposed, modeling reveals that D72, E74, and H87 are likely to be required for activity [50]. Further, a proteinase K protection assay reveals that the catalytic domain faces the lumen/extracellular space. Regulation involving anandamide metabolism by expression or by posttranslational modification has not been reported.

Lysophospholipase D GDPD3

Lysophospholipase D GDPD3 is known to convert lyso-NAPEs to AEA or other NAEs in competition with GDPD1 [41, 42]. Human lysophospholipase D GDPD3 (hGDPD3, GDE7, glycerophosphodiester phosphodiesterase 7, UniprotKB-Q7L5L3) is expressed as two splice isoforms. Isoform 2 is identical to Isoform 1 with the exception of missing residues 1-63. Physical properties for all isoforms are given in the supplement Table S1. Only the canonical sequence, Isoform 1, will be discussed here. No coding SNP variants are reported [8, 9]. hGDPD3 is a metallo-membrane protein containing two transmembrane helices with a preference for Ca²⁺ and is partially or completely inhibited by other divalent cations [41, 42]. No X-ray structures are available. There is, however, one potential structural model template suggested on the SWISS-MODEL website [14], template 5vug.1.B. hGDPD3 is widely expressed with the highest expression in tonsils and appendix, tongue, and esophagus, as well as being expressed in all areas of the brain [13]. Intracellular distribution is found on membrane structures throughout the cytosol and nucleoplasm. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Specific activities have been reported for both mouse and human recombinant GDPD3 [42] which share a 79.1% sequence homology. In contrast to GDPD1, both GDPD3 enzymes show a preference for phosphatidylcholine substrates over N-acyl-phosphatidylethanolamine substrates. The same is true for mGDPD1 hydrolyzing non-N-acylated substates [41]. Specific activities for both enzymes for the same substrates show the following order: N-Pal-lyso-PE ≈ N-Ole-lyso-PE >  > N-Ara-lyso-PE > GP-NPE, an order different than observed for GDPD1, and the reactions are orders of magnitude slower than the corresponding NAPE-PLD pathway competitor. Regulation of activity involving anandamide metabolism by expression or by posttranslational modification has not been reported.

Glycerophosphodiester phosphodiesterase 1

Glycerophosphodiester phosphodiesterase 1 (GDE1) is known to convert GP-NAEs to NAEs as well as exhibiting glycerophosphoinositol phosphodiesterase activity [32] and is a member of the glycerophosphoryldiester phosphodiesterase family (GDE phosphodiesterase family). Human glycerophosphodiester phosphodiesterase 1 (hGDE1, MIR16, RGS16-interacting membrane protein, UniprotKB-Q9NZC3) is expressed as a single isoform. Physical properties are given in the supplement Table S1. There are two reported coding SNP variants, R208Q and D328K [9]. Neither variant has a known clinical significance. hGDE1 is a metallo-membrane protein containing two transmembrane helices with a preference for Mg²⁺ and is inhibited by both Ca²⁺ and EDTA [44]. hGDE1 is widely expressed with the highest expression in the brain, endocrine tissues, GI tract, and muscle tissue. Intracellular distribution is found on membrane in the nucleoli, nucleoplasm, and vesicles throughout the cytosol. No X-ray structures are available for hGDE1, but 3D model has been proposed [51]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally. However, endo H treatment of rat GDE1 (87.9% sequence homology to the human isoform) confirms that 5–6 kDa of N-glycosylation is present, but specific sites of attachment are unknown [52].

Specific activities have not been reported for human GDE1 but have been reported for both rat and mouse [43, 44]. These enzymes have an 87.9% and 87.3% sequence homology to the human protein, respectively, and thus these specific activities can serve as an estimate for human values. Recombinant rat GDE1 shows a pH-dependent substrate specificity where lyso-phosphatidylcholine (lyso-PC) is the preferred substrate at pH 7.5 and lyso-phosphatidylinositol (lyso-PI) is the preferred substrate at pH 8.5 [43]. Specific activities for both enzymes for the same substrates follow the following order: GP-NPE > lyso-PE > N-Pal-lyso-PE > lyso-PC.

Although a specific catalytic mechanism for hGDE1 has not been proposed, site specific mutagenesis reveals that E97, D99, and H112 are required for activity [43]. Further detail has been gleaned from a 3D model study [52]. A crude mechanism has been proposed for rat GDE1 where H356 and E341 act as a general acid/base catalyst, generating a cyclic phosphate intermediate that is then hydrolyzed to form the respective products [43]. The catalytic domain has been shown to face the lumenal/extracellular space [43].

Murine GDE1 (mGDE1) was originally identified as a RGS16-interacting membrane protein, where RGSs are known GTPase activating proteins. This fact led to the discovery that the catalytic activity of mGDE1 is regulated by G-protein signaling. When HEK293T cells are transfected with mGDE1 they respond to isoproterenol, a known Gα_s coupled β-androgenic receptor agonist, with an ≈ 50% increase in GDE1 activity. When treated with phenylephrine, a known Gα_i/q coupled α-androgenic receptor agonist, however, GDE1 activity decreases by ≈ 30% [43]. These results indicate that GDE1 activity is both positively and negatively regulated by G-protein signaling. PRAF2 has also been shown to interact with RGS16, which in turn is known to interact with CCR5, a G-protein coupled chemokine receptor [51]. It has been proposed that a GDE1, RGS16, PRAF2, CCR5 complex may be involved in cell chemotaxis or perhaps desensitization, internalization, and recycling of CCR5 [52]. Immunoprecipitation studies and size exclusion chromatography reveals that both Gα_q/11 and Gβ proteins associate with rat GDE1 [53, 54]. In addition, Gα_q/11 knockdown Neuro2A cells exhibit lysoPLD activity that is reduced by about 50% compared to wildtype and the endogenous activity is partially rescued by overexpression of Gα_q in these cells. Taken together all of these results strongly suggest G-protein modulation of GDE1 activity. Regulation of activity involving anandamide metabolism by expression or by posttranslational modification has not been reported.

Tyrosine-protein phosphatase non-receptor type 22 (PTPN22)

As the name suggests, tyrosine-protein phosphatase non-receptor type 22 is known primarily for its ability to remove phosphates from tyrosines [55]. However, several reports clearly show that PTPN22 also catalyzes the dephosphorylation of pAEA to AEA [34, 45]. Human PTPN22 (hPTPN22, PEP, LyP, 70Z-PEP, PTPN8, UniprotKB-Q9Y2R2) is expressed as six splice isoforms. Sequence alignment reveals that Isoforms 5 and 6 lack both the known active site nucleophile [56] and key sections of the active site structure of the canonical sequence and can thus be discounted as biologically functional in terms of phosphatase activity. Isoform 3 lacks a key portion of the substrate binding site and thus is also likely not to be biologically competent for phosphatase activity. It is unclear from the literature which or if any combination of the three remaining isoforms is responsible for the phosphatase activity for substrate pAEA. Phobius web server [10] analysis indicates a 50% likelihood for a single transmembrane sequence in each of the three phosphatase competent isoforms noted above. However, hPTPN22 is known to be membrane associated rather than an integral membrane protein. The proposed transmembrane helix may instead represent a helical structure that is partially imbedded in the membrane. Physical properties for all isoforms are given in the supplement Table S1. There are 26 reported coding SNP variants of which only one has known clinical significance [8, 9]. The variant R620W is implicated in susceptibility to insulin dependence, diabetes, rheumatoid arthritis, and other disorders. hPTPN22 is widely expressed with the highest expression in bone marrow, lymphoid tissues and blood as well as being expressed in all areas of the brain. Intracellular distribution is found in the nucleoplasm, vesicles, and on the plasma membrane [13]. Numerous Xray structures are available for the first 300 or so residues which contains the N-terminal through the active site regions for all three of the isoforms noted above (e.g., PDB entry 2P6X). One report suggests the presence of second catalytic site could account for the diversity of substrate, lipid vs. peptide [57]. Predicted PTMs are given in the supplement Table S1, of which only phosphorylation at S751 [58] and S35 [59, 60] on isoform 1 have been confirmed experimentally. Comparison of the human to murine PTPN22 (UniProtKB-P29352; 70.7% sequence homology) sequences suggest potential phosphorylation at S449, S635, S684, and S692 as well.

Specific activities have only been reported for mPTPN22 but since there is a 70.7% sequence homology between the human and murine protein, the reported value should serve as a rough estimate of the human enzyme activity (Table 2). Only two studies have reported specific activities related to this enzyme. One study examined mouse brain homogenates from NAPE-PLD^−/− mice in the presence of an inhibitor for ABHD4 and EDTA to knock out any GDE or GDPD activity [34]. The resulting specific activity (Table 2) can thus be attributed to a combination of a yet-to-be identified PLC, PTPN22, and SHIP1 (see below) with values in the low pmol/mg/min range (discussed below). A second study examined mouse macrophages (RAW264.7) in culture, comparing the conversion of pAEA to AEA by wildtype to PTPN22^−/− knockout strains. The difference in specific activity between the two represents the combined activity of PTPN22 and SHIP1 (see below) [45].

Regulation of this protein is multifold. One of the first descriptions of the involvement of PTPN22 in anandamide metabolism was obtained from a study in which mouse macrophages were treated with the inflammation initiator lipopolysaccharide (LPS) [45]. Here it was found that LPS upregulates anandamide synthesis and PTPN22 expression while at the same time downregulates NAPE-PLD expression. Further, treatment of these cells with phosphatase inhibitors or the specific PLC inhibitor neomycin reduces but does not eliminate LPS-induced pAEA dephosphorylation. These results indicate that the NAPE-PLD pathway is likely the constitutive pathway to AEA production, whereas the PLC pathway is invoked during inflammation. A second regulatory pathway for PTPN22 involves the active site C227 and two other Cys residues where C129 protects C227 from irreversible oxidation in an oxidizing environment through the formation of a disulfide bond and concomitant reversible loss of activity [56]. On the other hand, formation of a disulfide with the local C231 also results in inactivation of the enzyme but suppresses reactivation in a reducing environment, providing negative regulation. A third regulatory pathway involves phosphorylation of specific Ser residues. Phosphorylation is known to prolong the half-life of hPTPN22 by inhibiting K48-linked ubiquitination and also inhibits its recruitment to the plasma membrane, a requirement for its inhibition of T cell receptor signaling [58]. Phosphorylation of hPTPN22 at S35 by protein kinase C has been shown to occur in vivo and in vitro and inhibits the ability to downregulate T cell receptor signaling [48]. Further, phosphorylation impairs hPTPN22’s ability to inactivate Src family kinases. Similar studies on AEA production have not been reported.

Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 (SHIP1)

As the name suggests, the primary function of Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 (SHIP1) is the removal of the 5′-phosphate from phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P3). However, its involvement in AEA metabolism has been observed where SHIP1 competes with PTPN22 for the hydrolysis of pAEA. Human SHIP1 (hSHIP1, INPP5D, SIP-45p, 150Ship, hp51CN, UniprotKB-Q92835) is expressed as three alternative splicing isoforms. Isoform 2 lacks V117 and is otherwise identical to Isoform 1 and thus is very likely to be involved in the AEA pathway. Isoform 3, however, lacks both the SH2 and SH3 binding domains essential for its PtdIns(3,4,5)P3 activity and for this reason is not likely to be a biologically competent isoform for this process. Physical properties for all isoforms are given in the supplement Table S1. Two coding SNP variants for the canonical isoform have been reported: H1169Y and V685D [8]. None of the variants have any known clinical significance. One X-ray structure for residues 397-857 (PDB entry 6IBD) and one nmr structure for residues 1-112 (PDB entry 2YSX) are available. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

The only direct reference to the pAEA activity for SHIP1 comes from a comparison of murine control and mSHIP1^−/− knockout macrophages (Table 2) [45]. The difference in specific activities between the knockout and controls is 0.77 nmol/mg/min which is quite comparable to the 1.0 nmol/mg/min determined for PTPN22. Like mPTPN22, mSHIP1 is also upregulated in response to LPS and is tyrosine phosphorylated [61]. Although specific phosphorylation sites stimulating upregulation were not determined, there are three known mSHIP1 tyrosine phosphorylation sites, Y918, Y945, and Y1021 [62], corresponding to Y915, Y944, and Y1022 in hSHIP1. There are also five known mSHIP1 serine/threonine phosphorylation sites, S246, S935, T964, S967, and S972 [63, 64], corresponding to S243, S934, T963, and S971 in hSHIP1. There is no human counterpart for S967. The 87.4% sequence homology suggests a similar likelihood of phosphorylation in human cells. Regulation activity involving anandamide metabolism by expression has not been reported.

Anandamide loss

Anandamide concentrations are reduced through three general pathways: (1) degradation to AA and ethanolamine, (2) conversion to other molecules, and (3) sequestration through transport to intracellular lipid microvesicles. Figure 2 depicts the known reaction pathways, the enzymes, and other proteins known to be involved.

Fig. 2

Reactions involved in the degradation of anandamide Metabolites: 5,6-EET-EA N-(2-hydroxyethyl)-5,6-epoxyeicosatrienamide, 11,12-EET-EA N-(2-hydroxyethyl)-11,12-epoxyeicosatrienamide, 12S-HPETE-EA N-(2-hydroxyethyl)-12S-Hydroperoxyicosatetraenamide, 14,15-EET-EA N-(2-hydroxyethyl)-14,15-epoxyeicosatrienamide, 15S-HETE-EA N-(2-hydroxyethyl)-15S-Hydroxyicosatetraenamide, 15S-HPETE-EA N-(2-hydroxyethyl)-15S-Hydroperoxyicosatetraenamide, 20-HETE-EA N-(2-hydroxyethyl)-20-hydroxyeicosatetraenamide, 8,9-EET-EA N-(2-hydroxyethyl)-8,9-epoxyeicosatrienamide, AA arachidonic acid, AEA anandamide, EA ethanolamine, EXA₄-EA eoxin A₄ ethanolamide, EXC₄-EA eoxin C₄ ethanolamide, EXD₄-EA eoxin D₄ ethanolamide, EXE₄-EA eoxin E₄ ethanolamide, PGD₂-EA prostaglandin D₂ ethanolamide, PGE₂-EA prostaglandin E₂ ethanolamide, PGF2a-EA prostaglandin F2a ethanolamide, PGH₂-EA prostaglandin H₂ ethanolamide, PGI₂-EA prostaglandin I₂ ethanolamide, TXA₂-EA thromboxane A2 ethanolamide, TXB₂-EA thromboxane B₂ ethanolamide. Enzymes: AKR1C3 aldo–keto reductase family 1 member C3, ALOX 15 arachidonate 15-Lipoxygenase, ALOX12 arachidonate 12-Lipoxygenase, COX-2 cyclooxygenase-2, CYP450s cytochrome P450s (see text for specific isoforms), FAAH-1,2 fatty-acid amide hydrolase isoforms 1 and 2, GPX glutathione peroxidase, NAAA N-acylethanolamine-hydrolyzing acid amidase, PRXL2B prostamide/prostaglandin F synthase, PTGDS prostaglandin D2 synthase, PTGES1,2,3 prostaglandin E2 synthase isoforms 1, 2, and 3, PTGIs prostaglandin I2 synthase, and TBXAS1 thromboxane A synthase 1

Degradation to free AA and ethanolamine

Fatty-acid amide hydrolase 1 (FAAH-1)

The hydrolysis of fatty-acid amides to their corresponding fatty acid and amine is facilitated by fatty-acid amide hydrolase 1 (FAAH-1). Human fatty-acid amide hydrolase 1 (hFAAH-1, FAAH, anandamide amino hydrolase 1, oleamide hydrolase 1, UniprotKB-O00519) is expressed as a single isoform. Physical properties are given in the supplement Table S1. There are 6 reported coding SNP variants: G110E, P129T, R295Q, S356V, A476G, and A345D [8, 9]. The P129T isoform is known to have reduced cellular stability and has been linked to problem drug use [65]. A345D may also have clinical significance, as it was obtained from a breast cancer sample. There are no crystal structures for hFAAH-1; however, the rat counterpart (rFAAH-1) with an 82.7% sequence homology has numerous published structures to serve as a model. The crystal structure of a truncated rFAAH-1 (residue 30-579, PDB entry 1MT5) indicates that it is a dimeric enzyme, each monomer containing two intramembrane hydrophobic helices that are thought to provide a hydrophobic membrane cap (residues 404-433). Phobius web server predictions [10] indicate a single transmembrane helix located at the N-terminus (residue 9-29) and a potential second intramembrane helical structure (residue 404-433) for both hFAAH-1 and rFAAH-1 the former is more likely an intramembrane structure. PSORT analysis [66] predicts that the catalytic domain is located in the cytosolic compartment. Further analysis of rFAAH-1 reveals a unique catalytic triad in the active site consisting of S241-S217-K142 for which S241 is the catalytic nucleophile. Sequence comparison shows that the same residues apply to the human isoform. hFAAH-1 is widely expressed with the highest expression in the brain as well as high expression in male and endocrine tissues [13]. Intracellular distribution includes its presence on membrane structures as well as throughout the cytosol. Treatment with PNGaseF does not alter gel migration, indicating the absence of N-linked glycosylation [66]. Predicted PTMs are given in the supplement Table S1.

Specific activities have been reported for both human and rat isoforms with the latter being ≈10 times larger (Table 3) [66,67,68]. FAAH-1 has a catalytic preference for the particular fatty acid present where anandamide > oleamide > N-oleoyl ethanolamine > N-palmitoyl ethanolamine and also has the ability to hydrolyze N-acyltaurines (NATs) but at lower rates than ethanolamides [66].

Table 3 Enzyme activity for enzymes involved in the degradation of anandamide (AEA) and phosphoanandamide (pAEA)

Regulation of FAAH-1 occurs at both transcriptional and post-transcriptional levels. Physiological levels of progesterone have been shown to upregulate hFAAH-1 at both the transcriptional and translational levels in T-lymphocytes, up to a 270% increase over controls resulting in a 60% decrease in anandamide concentrations [69]. It was subsequently shown that the increase in hFAAH-1 results from an increase in the transcription factor Ikaros that subsequently binds to the FAAH-1 promotor site. These findings are consistent with a role for FAAH-1 in modulating immunoendocrine interactions in human pregnancy. Membrane bound non-raft cholesterol in AEA/cholesterol rich membranes such as the ER have been shown to stabilize dimeric FAAH-1 and directly enhances the enzymatic activity by increasing the accessibility of the membrane port to AEA [70]. Regulation of FAAH-1 by phosphorylation has not been reported.

Fatty-acid amide hydrolase 2 (FAAH-2)

The hydrolysis of fatty-acid amides to their corresponding fatty acid and amine is also facilitated by fatty-acid amide hydrolase 2, but with different substrate preferences than FAAH-1. Human fatty-acid amide hydrolase 2 (hFAAH-2, AMDD, anandamide amino hydrolase 2, oleamide hydrolase 2, UniprotKB-Q6GMR7) is expressed as a single isoform. Physical properties are given in the supplement Table S1. There are 7 reported coding SNP variants (R42W, R192Q, G342R, A458S, A388S, and A418S) of which none have known clinical significance [8, 9]. There are no crystal structures for hFAAH-2 and only a 32.1% sequence homology to hFAAH-1. However, a threaded model for hFAAH-2 based on rFAAH-1 has been reported and clearly shows the expected Ser-Ser-Lys catalytic triad involving S206, S230, and K131 [71] in which S230 appears to be the active site nucleophile. Interestingly, Phobius web server predictions [10] indicate a single transmembrane helix located near the C-terminus (residue 394-414) rather than at the N-terminal as observed for hFAAH-1 but in the same region where a potential intramembrane helix is predicted for hFAAH-1. In addition, a potential N-terminal intramembrane helix is predicted for hFAAH-2, albeit with low probability, in the same region the transmembrane helix is predicted for hFAAH-1. PSORT analysis predicts that the catalytic domain is located in the lumenal compartment [66]. hFFAH2 is widely expressed throughout most tissues and is particularly abundant in the pancreas and endocrine tissues but is also expressed in the brain with highest amounts in the cerebellum [13]. Intracellular distribution appears to be localized to lipid droplet membranes [72]. Predicted PTMs are given in the supplement Table S1, none of which have been confirmed experimentally.

Specific activity has been reported for hFAAH-2 with a value for AEA hydrolysis that is only 1–2% of that exhibited by FAAH-1, but comparable activity for hydrolysis of oleamide (Table 3). In addition, FAAH-2 has a catalytic preference for the fatty acid present where oleamide >> N-oleoyl ethanolamine > anandamide > N-palmitoyl ethanolamine and is unable to hydrolyze NATs [66]. These data suggest that FAAH-1 is primarily involved in anandamide and NAT hydrolysis and both FAAH-2 and FAAH-1 contribute to the hydrolysis of monosaturated lipid amides [66]. Regulation of FAAH-2 at any level has yet to be reported.

N-acylethanolamine-hydrolyzing acid amidase (NAAA)

The hydrolysis of fatty-acid amides to their corresponding fatty acid and amine is also facilitated by an N-acylethanolamine-hydrolyzing acid amidase which exhibits a preference for small saturated fatty acids. Human N-acylethanolamine-hydrolyzing acid amidase (hNAAA, ASAHL, PLT, acid ceramidase-like protein, UniprotKB-Q02083) is expressed as three isoforms where isoform 2 is missing the last 35 residues present in isoform 1 and isoform 3 is missing the last 159 residues of isoform 1. Physical properties for all isoforms are given in the supplement Table S1. Only Isoform 1 exhibits demonstrated activity [82] and thus will be the only isoform discussed further. There are 4 reported coding SNP variants (K75R, N107K, V151I, and F334L) none of which has a known clinical significance [8, 9]. Isoform 1 is processed further to remove the signal peptide, the first 28 residues and a single cleavage between residues 125 and 126 forming a hetero-dimer consisting of the α-subunit (residue 29-125, 14.6 kDa) and the β-subunit (residue 126-359, 33.3 kDa) [74]. There are two crystal structures for hNAAA, one without the 125–126 cleavage (residue 29-359, PDB entry 6DXW) and one with the 125–126 cleavage (PDB entry 6DXX). hNAAA is a glycoprotein for which N37, N107, N309, N315 and N333 are confirmed sites of glycosylation [73,74,75, 83]. hNAAA is widely expressed in most tissues with highest expression in blood, monocytes in particular, lymphoid tissues, spleen, and lymph nodes, and to a lesser extent in the gastrointestinal tract and brain [13]. Intracellular distribution favors lysosomes [75, 84]. Predicted PTMs are given in the supplement Table S1, but only the N-glycosylations noted above have been confirmed experimentally.

hNAAA is a cysteine amidase where the active site C126 is responsible for not only substrate hydrolysis, but also facilitates auto-cleavage of the 125–126 bond [83] along with residues R142, D145, N287 [84], and L325 and T335[82]. Although NAAA is considered to be a cytosolic protein, it must also penetrate the lipid bilayer in order to access substrate. It is thought that helix a3 on the α-subunit and helix a6 from the β-subunit are attracted to the anionic surface of lysosomal vesicles through their associated cationic residues leading to imbedding of the associated hydrophobic residues [83]. In addition, its activity is stimulated by both DTT and the detergent Nonidet P-40, the latter shown to initiate a conformational change in the structure leading to the formation of the active site.

Specific activity has been reported for hNAAA with a value for AEA hydrolysis that is only 30–50% of that exhibited by FAAH-2 and only 3% of that for PEA hydrolysis (Table 3). hNAAA has a catalytic preference for the smaller fatty-acid amides where PEA >  > N-myristylethanolamine >  > N-steroylethanolamine ≈ N-laurylethanolamine > AEA [73]. Clearly the primary biological function of this enzyme is the hydrolysis of PEA but can readily compete with FAAH-2 for AEA.

Regulation of NAAA by lipid and thiol compounds has been reported [85]. The presence of phosphatidylcholine (PC), phosphatidylethanolamine, or sphingomyelin increases the hydrolysis rate of PEA by fivefold at 100 µM and increases with phospholipid concentration based on PC results. This compares to the sixfold increase with a similar amount of the nonionic detergent Nonidet P-40. Numerous endogenous and non-endogenous thiol compounds were tested where DTT and the endogenous dihydrolipoic acid have the greatest effect. These compounds presumably help keep the catalysis C126 in the reduced state that is required for activity. Neither additive influences substrate preference.

Regulation of NAAA expression involving anandamide metabolism has not been reported. N-glycosylation is necessary for transport of hNAAA from the Golgi, but not required for catalytic activity [86].

Conversion of anandamide to other compounds

There are numerous pathways involved in the conversion of anandamide to other compounds that display a wide variety of different biological effects. The reactions center on the enzyme catalyzed oxidation of the AA moiety itself. The reactions discussed below examine the first steps in such conversions which involve enzymes that are also associated with metabolism of AA along numerous eicosanoid metabolic pathways (Fig. 2). Greater detail on these enzymes and the pathways are found elsewhere [87, 88].

Prostaglandin G/H synthase 2 (COX-2)

The primary biological role for prostaglandin G/H Synthase 2 (COX-2) and prostaglandin G/H Synthase 1 (COX1) is the conversion of AA to prostaglandin H₂ (PGH₂) via prostaglandin G₂ (PGG₂) through the reaction of substrate with molecular oxygen. PGH₂ is then rapidly converted enzymatically to prostaglandin D₂ (PGD₂), prostaglandin E₂ (PGE₂) and prostaglandin F_2a (PGF_2a), prostaglandin I₂ (PGI₂) and thromboxane A₂ (TXA₂) [87]. PGH₂ is quite unstable and is also rapidly converted to PGD₂, PGE₂, levuglandin E₂ (LGE₂) and levuglandin D₂ (LGD₂) non enzymatically in aqueous solution [88, 89]. COX-2 and not COX1 oxygenates anandamide to PGH₂ ethanolamide (PGH₂-EA) which is then rapidly converted to PGD₂ ethanolamide (PGD₂-EA), PGE₂ ethanolamide (PGE₂-EA), PGF_2a ethanolamide (PGF_2a-EA), PGI₂ ethanolamide (PGI₂-EA), and TXA₂ ethanolamide (TXA₂-EA) enzymatically (Fig. 2) [76, 90,91,92,93].

Human cyclooxygenase-2 (COX-2, PTGS2, PGHS-2, cyclooxygenase-2, UniProtKB-P35354) is a heme-requiring enzyme that is expressed as a single isoform. Its physical properties and PTMs have been discussed in detail elsewhere [87, 94,95,96]. During inflammation, COX-2 is significantly upregulated in the affected tissue with the exception of brain, testes, and the macula densa of the kidney where it is constitutively expressed [97]. Intracellular distribution favors the endoplasmic reticulum.

Specific activity for partially purified hCOX-2 towards AEA has been reported and found to be 70% of that observed for the more typical substrate AA (Table 3). Clearly hCOX-2 is a formidable contributor to AEA degradation when expressed in appropriate quantities, for example, during inflammation. Regulation of COX-2 at the transcriptional level has been reported where a variety of cytokines and growth factors have been shown to stimulate induction [98]. Tyrosine phosphorylation at Y120 and Y446 are known to enhance the activity of hCOX-2 towards AA substrate [95, 99]. Nitrosylation of C526 by an inducible nitric oxide synthase, a major mediator of inflammation, results in a twofold enhancement of activity towards AA [96, 100].

Cytochrome P450s

The primary biological role for cytochrome P450 enzymes is the oxidation of steroids, lipids, and various xenobiotics. Cytochrome P450 (CYP) represents a large class of heme-requiring monooxygenases, too vast to discuss in detail here. However, it is worthwhile to examine a few that are known to be involved in anandamide catabolism. There are five major products (Fig. 2) [77]—10 in total—for anandamide metabolism by CYP enzymes: N-(2-hydroxyethyl)-20-hydroxyeicosatetraenamide (20-HETE-EA), N-(2-hydroxyethyl)-5,6-epoxyeicosatrienamide (5,6-EET-EA), N-(2-hydroxyethyl)-8,9-epoxyeicosatrienamide (8,9-EET-EA), N-(2-hydroxyethyl)-11,12-epoxyeicosatrienamide (11,12-EET-EA), N-(2-hydroxyethyl)-14,15-epoxyeicosatrienamide (14,15-EET-EA), each of which is formed at different rates by different CYP isoforms (Table 3). Isoforms CYP2D6, CYP3A4, CYP2J2, CYP4F4 are known anandamide monooxygenases and are found in the human brain as well as other tissues [13, 77,78,79, 101]. hCYP2D6 and hCYP3A4 have been shown to be major contributors to anandamide metabolism in the brain where incubation with isoform-specific inhibitory antibodies reduces AEA metabolism by 88% and 42–66%, respectively [77]. Interestingly, one of the metabolites, 5,6-EET-EA, is a potent endocannabinoid CB2 receptor agonist with comparable binding efficiency to AEA as well as being more stable than AEA [101], the production of which thus increases the potency of the original anandamide signaling. In addition, 5,6-EET-EA is also an agonist for the endocannabinoid CB1 receptor, albeit with > 1000-fold weaker binding. 20-HETE-EA and 14,15-EET-EA also bind to the CB1 receptor with 3.6-fold and 5.7-fold lower efficiency than anandamide, respectively, and are more rapidly degraded in the brain than anandamide [80]. The biological impact of these and the other P450 metabolites of AEA needs to be clarified.

Lipoxygenases

Lipoxygenases represent a class of non-heme iron-requiring enzymes responsible for oxygenation of a wide array of polyunsaturated fatty acids (PUFAs). The primary biological role is to oxygenate PUFAs at specific positions in a stereospecific manner, producing the corresponding mono-hydroperoxy products that are readily converted to mono-hydroxy derivatives by ubiquitous glutathione peroxidases (GPX). Individual lipoxygenases are responsible for hydroperoxidation at specific positions and associated stereochemistry. For example, of the 12 possible peroxy positions on arachidonic acid, 10 are found in nature and 5 found in humans. Both 12S and 15S peroxidations of AEA catalyzed by 12S-lipoxygenase and 15S-lipoxygenase, respectively, have been reported [81, 102, 103]. However, 5S and 12R peroxidation, and 15R hydroxylation catalyzed by 5S-lipoxygenase, 12R-lipoxygenase, and aspirin-acetylated COX-2, respectively, have been observed for AA substrate, but not for AEA.

12S-lipoxygenase (ALOX12)

12S-lipoxygenase catalyzes the oxygenation of AEA to 12(S)-hydroperoxyarachadonoylethanolamide (12(S)-HPETE-EA) which is rapidly converted to 12(S)-hydroxyarachadonoylethanolamide (12(S)-HETE-EA) as well as numerous other products (Fig. 2) [103, 104]. Human12S-lipoxygenase (hALOX12, 12S-LOX, 12LO, LOG12, UniprotKB-P18054) is translated as a single isoform. Physical properties and PTMs are discussed elsewhere [104]. The specific activity for hALOX12 with AEA was determined for solubilized protein precipitates obtained from human leukocyte/platelet homogenates and found to be ≈ 70% of that found for AA [81] (Table 3).

As observed for CYP enzymes, the hydroxylated product also binds to the endocannabinoid receptors CB1 and CB2. 12(S)-HETE-EA binds to rat brain membranes containing CB1 with lower efficiency (≈ 53%) than AEA and human CB2 expressed in CHO cells also binds with lower efficiency than AEA (≈ 60%) [103]. A second report indicates that 12(S)-HETE-EA binds to both CB1 in rat brain membranes and CB2 in rat spleen membranes with lower efficiency than AEA, 60% and 72%, respectively [105]. The binding of 12(S)-HETE-EA to mixed CB1 and CB2 receptors has been shown to be agonistic but less effective than AEA with an EC₅₀ that is 2.5 times that observed for AEA for the inhibition of electrically initiated twitch response in murine deferens [81].

The catalytic activity of hALOX12 is sensitive to the redox conditions [106] within the cell and may be regulated by phosphorylation at S246 based on the known phosphorylation of the rat isoform [47] for which it shares an 84.5% sequence homology. Upregulation of ALOX12 in porcine aortic vascular smooth muscle cells by the proinflammatory cytokines IL-1β, IL-4, and IL-8 has also been reported [107].

15S-lipoxygenase (ALOX15)

15S-lipoxygenase catalyzes the oxygenation of AEA to 15(S)-peroxyarachadonoylethanolamide (15(S)-HPETE-EA) which is rapidly converted to 15(S)-hydroxyarachadonoylethanolamide (15(S)-HETE-EA) by ubiquitous GPXs (Fig. 2). ALOX15 can also convert 15(S)-HPETE-EA to eoxin A₄ ethanolamide (EXA₄-EA) which is then readily converted to the ethanolamides of eoxin C₄ (EXC₄-EA) and eoxin D₄ (EXD₄-EA) [108] through the action of leukotriene C₄ synthase (LTC4S) and gamma-glutamyl transaminase (GTT1) in a sequential manner (Fig. 2) [104]. Like hALOX12, human 15S-lipoxygenase (hALOX15, 15A-LOX, 15-LOX, 15-LOX-1, 12/15-lipoxygenase, UniProtKB-P16050) is a non-heme, iron-requiring cytosolic protein that becomes membrane associated in the presence of calcium. It is transcribed as two isoforms, but only the canonical form, Isoform 1, has been observed at the protein level. The physical properties and PTMs are discussed elsewhere [104, 109].

Specific activity for hALOX15 with AEA has not been reported; however, one publication gives a K_m for AEA of 108 µM [108]. It should be noted that hALOX15 also produces 12S-HETE from AA in a 1:9 12S:15S ratio [110]. The production of 12S-HETE-EA by this enzyme is to be expected but has yet to be reported. 15(S)-HETE-EA binds to rat brain CB1 with lower efficiency than AEA (14%) and does not bind to human CB2 expressed in CHO cells [103]. A second report indicates that 15(S)-HETE-EA binds to CB1 in rat brain membranes with a lower efficiency than AEA (15%) and does not bind to CB2 in rat spleen membranes [105]. The binding of 15(S)-HETE-EA to mixed CB1 and CB2 receptors has been shown to be agonistic but less effective than AEA with an EC₅₀ that is 4.9 times that for AEA for the inhibition of electrically initiated twitch response in murine deferens [81].

Regulation via posttranslational modification has not been reported. However, upregulation of hALOX15 expression by the proinflammatory cytokines IL-4 and IL-13 in human monocytes is well documented [111].

Anandamide transport and storage

AEA is a well-known paracrine and autocrine signaling molecule. The fact that it is uncharged and hydrophobic necessitates that AEA be transported both extracellularly and intracellularly by carrier molecules. Further, both the biosynthetic and degradation machinery are located within the cell, necessitating the existence of some process or processes to move AEA out of the cell for extracellular signaling and import for degradation. Until recently this process was thought to operate solely on a supply and demand mode. However, it has become clear that AEA is also stored in lipid droplets, thus complicating the supply demand model with a “buffer” system that can modulate available AEA. The following sections outline each of the processes noted above, including characterizations of the proteins involved.

Extracellular AEA transport proteins (AETs)

Transport of hydrophobic molecules in aqueous solution in living systems is typically facilitated by specific binding proteins. In plasma, albumin, and members of the lipocalin family are involved in transporting lipophilic molecules in plasma and may possibly be involved in extracellular AEA transport. Proteins historically shown to be involved in intracellular transport [112] have more recently been shown to exist extracellularly and one is known to be involved in extracellular AEA transport and others may be involved as well [113, 114]. Several fatty acid binding proteins, FABP3, FABP5, and FABP7 are known to be involved in intracellular trafficking of AEA [115]. FABP5 was recently shown to mediate retrograde endocannabinoid transport, suggesting that it may serve as a synaptic carrier for AEA [113]. heat shock proteins (HSPs) are also involved in intracellular transport of AEA [112]. HSP70, HSP90, HSP60 as well as several other HSPs (e.g., Grp75, Grp 80) have been identified extracellularly [114]. All of these species have been shown to bind lipids and thus may potentially act as additional AETs.

Microvesicles and exosomes are also well documented as purveyors of both intracellular and extracellular signaling through their transport of proteins, peptides, nucleic acids, and lipids [116], the latter making them potential candidates for both intracellular an extracellular transport of AEA. One report supporting vesicular transport of AEA shows that microglia produce extracellular vesicles with AEA on their surface and in turn stimulate type-1 endocannabinoid receptors (CB1), facilitating inhibition of presynaptic transmission [117].

Transport of AEA across plasma membranes

Endocannabinoids are readily transported across plasma membranes. The mode for transport of AEA has been a hotly debated subject for many years. Model proponents fall into three camps, diffusion, facilitated transport, or an endocytotic pathway. It is clear, however, that AEA does not pass through membranes by active transport [118].

Support for diffusion control of AEA transport is multifold. Some cite the lack of direct evidence for a proteinaceous transporter and a simpler model where sequestration of cytosolic AEA by proteins or vesicles necessarily leads to the AEA gradient necessary for diffusion across the plasma membrane [119]. Indeed, external addition of AEA to lipid vesicles containing internalized FAAH results in rapid AEA transport and hydrolysis [115]. Further, this transport is enhanced by cholesterol and coprostanol but is not affected by cholesterol sulfate. The equality of cholesterol and coprostanol transport indicates that the sterol ring enhances transport and coprostanol facilitated transport indicates that lipid rafts are not necessary for this process to occur. The lack of enhancement with cholesterol sulfate supports a mechanism whereby sterol ring flip-flopping may be involved in the transport process. Earlier work shows that both AEA insertion into lipid membranes and transport across them is cholesterol-dependent [120]. In related work, AEA present in resealed red blood cell membranes (ghosts) rapidly exits to external solutions containing bovine serum albumin (BSA), a known AEA binder, in a BSA-dependent manner [118]. Similarly, treatment of HeLa cells expressing mouse FABPs, known AEA binders, with FABP inhibitors reduces uptake of [¹⁴C] AEA as do FABP knockout cells in the absence of inhibitors [121].

Much of the support for AEA membrane transporter proteins (AMT) comes from observations that a variety of pharmaceuticals can reduce AEA transport presumably by interacting with an undiscovered AMT. On the other hand, these inhibitors may instead be interacting with intracellular proteins involved in sequestration or destruction of AEA. A more recent study clearly shows the existence of an AMT that is capable of bi-directional trafficking of AEA [122]. Using specific inhibitors of both AMT and FAAH they show that inhibition of FAAH results in an increase in intracellular AEA and inhibition of the putative AMT results in a decrease in intracellular AEA, clearly showing the existence of an AMT. In addition, they also show that the same AMT inhibitors produce a concentration-dependent decrease in AEA export, clearly supporting the existence of a yet-to-be identified AMT.

A third possibility for transmembrane trafficking involves endocytosis. Treatment of rat basophil cells (RBL-2H3) with inhibitors of caveola-related (clatherin-independent) endocytosis reduces AEA transport by ≈ 50% compared to controls [123]. Further, treatment of these cells with inhibitors of clatherin-dependent processes has no effect on AEA transport, indicating that only the caveola-related transport is involved. The same study found that AEA is concentrated in caveolin-rich membranes following internalization, further supporting the clatherin-independent pathway.

Storage of AEA

The long-held belief that AEA is synthesized on demand has been seriously challenged by more recent reports (see review [2]). It has been shown that once AEA is taken up by cells it is rapidly targeted to adiposomes [112, 124]. Further, it is also observed that larger adiposome compartments have an increased capacity to both store and metabolize AEA [124]. Adiposomes not only house intracellular AEA, but also contain degradation enzymes FAAH-1&2 [72, 112] and thus these lipid bodies serve as modulators of intracellular accumulation and degradation of AEA.

Both mobilization and degradation of AEA stored in liposomes requires a mechanism by which to move the internalized AEA from within the liposome to the surface. One mode for this transport has been shown to involve sterols inserted into the liposome membrane [115]. Both lipid-raft associating cholesterol and lipid-raft avoiding coprostanol enhance transport of AEA across the liposome bilayer indicating that lipid rafts are not necessary for this process to occur. The mode of action is thought to involve a competition for shielding the small headgroups from solvent water resulting in increased surface exposure of AEA to FAAH [115] or to specific transport proteins.

Conclusions

Anandamide homeostasis is achieved through the rates of synthesis, degradation, and storage of this molecule. This manuscript discusses the known enzymes and other proteins that are major contributors to this system. Biosynthesis occurs along four parallel pathways, each of which is catalyzed by its own set of enzymes and under independent control. Degradation can be simple, involving a hydrolysis of the amide by enzymes, or complex involving the eicosanoid biosynthesis machinery. Lastly, there is a storage system involving lipid microvesicles that modulate available AEA through sequestration, regulated release, and degradation. One additional item for consideration is the fact that AEA can be converted to other molecules that are known to activate the same CB receptors that AEA stimulates, thus continuing the signaling in the absence of AEA itself.

There are more than a few missing aspects that have yet to be uncovered in the laboratory and are required to provide a comprehensive picture of anandamide homeostasis. For example, a relative comparison of the synthesis degradation enzyme rates within particular tissues under defined metabolic conditions (e.g., healthy vs. inflammatory state) would be most enlightening. In addition, exploration of potentially new endocannabinoid products formed by oxidation catalyzed by eicosanoid enzymes other than ALOX12 and ALOX15 is in order. Lastly, a thorough investigation of how each pathway enzyme is regulated would help to better define the condition-dependent roles of each arm of the multi-armed pathways.