Endogenous Unsaturated C18 N-Acylethanolamines Are Vanilloid Receptor (TRPV1) Agonists*

The endogenous C18 N-acylethanolamines (NAEs) N-linolenoylethanolamine (18:3 NAE), N-linoleoylethanolamine (18:2 NAE), N-oleoylethanolamine (18:1 NAE), and N-stearoylethanolamine (18:0 NAE) are structurally related to the endocannabinoid anandamide (20:4 NAE), but these lipids are poor ligands at cannabinoid CB1 receptors. Anandamide is also an activator of the transient receptor potential (TRP) vanilloid 1 (TRPV1) on primary sensory neurons. Here we show that C18 NAEs are present in rat sensory ganglia and vascular tissue. With the exception of 18:3 NAE in rat sensory ganglia, the levels of C18 NAEs are equal to or substantially exceed those of anandamide. At submicromolar concentrations, 18:3 NAE, 18:2 NAE, and 18:1 NAE, but not 18:0 NAE and oleic acid, activate native rTRPV1 on perivascular sensory nerves. 18:1 NAE does not activate these nerves in TRPV1 gene knock-out mice. Only the unsaturated C18 NAEs elicit whole cell currents and fluorometric calcium responses in HEK293 cells expressing hTRPV1. Molecular modeling revealed a low energy cluster of U-shaped unsaturated NAE conformers, sharing several pharmacophoric elements with capsaicin. Furthermore, one of the two major low energy conformational families of anandamide also overlaps with the cannabinoid CB1 receptor ligand HU210, which is in line with anandamide being a dual activator of TRPV1 and the cannabinoid CB1 receptor. This study shows that several endogenous non-cannabinoid NAEs, many of which are more abundant than anandamide in rat tissues, activate TRPV1 and thus may play a role as endogenous TRPV1 modulators.

In the present study, we have examined the effects of naturally occurring C18 NAEs on native TRPV 1 in rodent blood vessels and the cloned hTRPV 1 expressed in HEK293 cells. Activation of TRPV 1 on perivascular sensory nerves in rat and mouse mesenteric arteries causes release of the potent vasodilator calcitonin gene-related peptide, the effect of which can be conveniently recorded in a preconstricted arterial segment (20,30). The flexible structures of NAEs differ greatly from the more rigid structures of capsaicin and the tricyclic cannabinoids. To gain insight into common structural features of these compounds, we used computational techniques to compare low energy conformations of NAEs and their putative pharmacophoric elements with those of the reference TRPV 1 and cannabinoid CB 1 receptor activators capsaicin and HU210, respectively.

MATERIALS AND METHODS
Quantification of NAEs-Rat dorsal root ganglia from all spinal levels and the mesenteric arterial bed were homogenized in 500 l of Tris buffer ( (antioxidant) was added to extract lipids. Standards for quantifications were obtained by addition of different amounts of the N-acylethanolamines to homogenates of dorsal root ganglia and mesenteric arterial bed, respectively. After centrifugation at 3000 rpm in 10 min (5°C), the supernatant was collected in polypropylene tubes and vacuum-evaporated. The extraction residue was reconstituted in 100 l of methanol with 0.3 mM ascorbic acid and stored at Ϫ20°C until analysis. The protein content of the pellet was determined with Coomassie Blue (Pierce) protein assay, using bovine serum albumin as a standard. A Perkin Elmer Series 200 liquid chromatography system with autosampler (Applied Biosystems, Norfolk, CT), coupled to an API 3000 LC-MS-MS (Applied Biosystems/MDS-SCIEX, Toronto, Canada) was used for the analysis. The column was a Genesis C 8 (20 ϫ 2.1 mm) with a particle size of 4 m (Jones, Lakewood, CO). Aliquots of 5 l were injected by the autosampler. Mobile phase was a water-methanol gradient containing 0.5% acetic acid, and the initial mobile flow was 75% methanol. A linear gradient to 100% methanol was applied in 6 min. The mobile flow rate was 0.2 ml/min. The turbo ion spray interface was set to 370°C, the declustering potential 40 volts and collision energy 35 volts. The analyses were performed in the positive ion multiple reaction monitoring mode, and the mass fragments used were for anandamide, m/z 348. 2 Recording of Vasorelaxation-Wistar-Hannover rats (250 g) of female gender and C57 BL/6 mice (30 gm) of either sex were killed by exsanguination under CO 2 anesthesia. TRPV 1 -deficient mice (16) were kindly provided by Prof. David Julius, UCSF. The first or second generation branches of the mesenteric artery was carefully dissected and flushed with physiological salt solution (composition in mM: NaCl 119, KCl 4.6, CaCl 2 1.5, MgCl 2 1.2, NaHCO 3 15, NaH 2 PO 4 1.2, and glucose 6). Ring segments, ϳ2-mm long, were suspended between two stainless wires in tissue baths, containing physiological salt solution. One of the wires was connected to a force-displacement transducer (model FT03C, Grass Instruments) for isometric tension recording. The physiological salt solution was continuously bubbled with carbogen (95% O 2 and 5% CO 2 ), resulting in a pH of 7.4.
The vessel segments were allowed to equilibrate for 1 h under a passive load of 1 mN and 2 mN for mouse and rat mesenteric arteries, respectively. Arteries were contracted with 3 M phenylephrine to induce stable and submaximal contractions. Increasing concentrations of test drugs were added cumulatively to determine concentration-response relationships. Relaxant responses are expressed as percentage reversal of the phenylephrine-induced contraction. All experiments were performed at 37°C in the presence of 3 mM N G -nitro-L-arginine and 10 M indomethacin as previously described (20,30). Some arteries were pretreated with 10 M capsaicin for 60 min to cause desensitization and/or neurotransmitter depletion of sensory nerves. Capsaicin, capsazepine, and the NAEs were dissolved in ethanol and added cumulatively to the organ baths in volume of 2.5 l. The final ethanol concentration in the organ bath never exceeded 1%. The incubation time with capsazepine was 20 min.
Electrophysiology-HEK293 cells were transfected with hTRPV 1 cDNA, kindly provided by Dr. Sven-Eric Jordt (UCSF), using Lipofectamine (Invitrogen Life Technologies, Inc.). After 24 h, whole cell currents were recorded at a holding potential of Ϫ50 mV. The bath solution contained (in mM): NaCl 140, KCl 5, MgCl 2 2, glucose 10, and TES 10 adjusted to pH 7.4. The pipette solution contained (in mM): CsCl 140, EGTA 5, and TES 10 adjusted to pH 7.4. All experiments were carried out at room temperature (20 -22°C). The test drugs were dissolved in ethanol. The final ethanol concentration in the bath solution never exceeded 0.2%. Responses are calculated as a percentage of the response to 2 M capsaicin. For further details regarding experimental procedure and data acquisition, see Ref. 31.
Fluorometric Calcium Imaging-HEK293 cells were transfected with hTRPV 1 cDNA as described above. After 24 h, transfected cells were plated on poly-D-lysine-coated 384-well optiplates (Corning) at a density of ϳ40,000 cells/well and were allowed to proliferate for 24 h. Prior to start of the assay, the cells were incubated with 2 M fluo-4/AM for 30 min at 37°C. Dye not taken up by cells was removed by aspiration followed by washing three times with 25 l of a HEPES-buffered ringer solution (composition in mM: NaCl 145, KCl 5, MgCl 2 1, CaCl 2 1, and HEPES 10 adjusted to pH 7.4). The assay was performed in the HEPESbuffered ringer solution at room temperature. Fluorescence measurements were performed at 1-s intervals using a 384-well fluorometric imaging plate reader (FLIPR; Molecular Devices, Sunnyvale, CA). Cellular responses were quantitated by calculating the difference between peak increases in fluorescence over baseline. Responses are calculated as a percentage of the response to a saturating concentration of anandamide (100 M). The lipids were dissolved in ethanol. The final concentration of ethanol in the wells was 0.05% in all experiments.
Calculations and Statistics-The Ϫlog of the agonist concentration eliciting half-maximal response (pEC 50 ) was determined by nonlinear regression (GraphPad Prism 3.0). E max refers to the maximal response achieved. When the concentration-response curve did not reach a plateau, and hence E max and EC 50 could not be determined, the area under curve (AUC) was calculated (GraphPad Prism version 3.0) and used for evaluation of drug effects. Two-tailed, unpaired Student's t test or analysis of variance (ANOVA) followed by Dunnett's post hoc test (Graph-Pad Prism 3.0) was used for statistical comparison. The content of NAEs is expressed as mol per mg protein. These values were log-transformed before statistical comparison, using ANOVA followed by the Bonferroni's post hoc test (GraphPad Prism 3.0). Statistical significance was accepted when the p value was less than 0.05.
Molecular Modeling-Monte Carlo conformational searches, using the MacroModel (32) suite of software (version 8.6), were conducted to identify low energy families of conformers within 3 kcal/mol of the global energy minimum of each compound in water. Non-bonded interactions within 8 Å for van der Waals interactions and 20 Å for electrostatic interactions were included in the calculations. The XCluster (33) program implemented in the MacroModel package was used to group C18 NAEs, 20:4 NAE, capsaicin, and HU210 into geometrically similar families. The MacroModel software (version 8.6) was used to find optimal alignments, i.e. the minimum root mean square deviation (RMSD) between the pharmacophoric elements (see supplementary data for further details).

RESULTS
The amounts of NAEs in the rat mesenteric artery and dorsal root ganglia were measured by LC-MS-MS (TABLE ONE). The levels of 18:2 NAE, 18:1 NAE, and 18:0 NAE in the mesenteric artery are 24, 20, and 26 times higher, respectively, compared with anandamide (p Ͻ 0.001, n ϭ 12). No differences could be detected between the amounts of 22:4 NAE and 18:3 NAE compared with anandamide. The levels of 18:1 NAE and 18:0 NAE in dorsal root ganglia are 7 and 8 times higher, respectively, compared with anandamide (p Ͻ 0.001, n ϭ 12). The amount of anandamide was 37 times higher than 18:3 NAE (p Ͻ 0.001), whereas no differences were detected between the levels of anandamide compared with 22:4 NAE and 18:2 NAE (n ϭ 12).
The unsaturated C18 NAEs N-linolenoylethanolamine (18:3 NAE), N-linoleoylethanolamine (18:2 NAE), and N-oleoylethanolamine (18:1 NAE) all induce a concentration-dependent relaxation in the rat mesenteric artery. Oleic acid at a concentration of 10 M is unable to cause a relaxation in this artery (TABLE TWO), which excludes that oleic acid formed by hydrolytic cleavage of 18:1 NAE is responsible for the vasodilator response to this NAE. The relaxation induced by these unsaturated NAEs are inhibited by the TRPV 1 receptor antagonist capsazepine in a concentration-dependent manner (Fig. 1, A-C) and absent in arteries pretreated with the neurotoxin capsaicin (10 M for 60 min;    The saturated 18:0 NAE N-stearoylethanolamine can neither cause relaxation (Fig. 1D) nor enhance the relaxation induced by the TRPV 1 receptor agonist anandamide. The pEC 50 and E max values for anandamide are 6.7 Ϯ 0.1 and 96 Ϯ 1% in the absence and 6.9 Ϯ 0.1 and 96 Ϯ 1% in the presence of 18:0 NAE, respectively (n ϭ 6).
Unsaturated 18 NAEs evoke concentration-dependent inward whole cell membrane currents in HEK293 cells expressing hTRPV 1 (Fig. 2). The pEC 50  As shown in experiments with the calcium fluorometric imaging technique, all unsaturated C18 NAEs, 20:4 NAE, and 22:4 NAE are able to evoke an increase in intracellular calcium in a concentration-dependent manner (Fig. 3). In contrast, 18:0 NAE and oleic acid are inactive. The pEC 50 and E max values obtained in these experiments are shown in  TABLE TWO. A conformational analysis of highly flexible ligands, such as the long chain NAEs (Fig. S1, supplement), generates a very large number of low energy conformations within 3 kcal/mol of the lowest energy minimum. The multitude of unique minima found in these systems (1783 for NAE 18:0, 600 for NAE 18:1, 1604 for NAE 18:2, 2217 for NAE 18:3, and 2485 for NAE 20:4) are, as a consequence of their great flexibility, not all structurally distinct. Similar conformational structures were therefore grouped in clusters by XCluster calculations (33).
For 18:0 NAE, two clusters of extended shapes, in which the ends are far from each other, accounted for 65% of the low energy conformations (Fig. 4). Thus, it is clear that this compound prefers to exist in an extended shape in water. However, the situation changes dramatically after introduction of a cis double bond in the hydrocarbon chain. In the major low energy clusters of the unsaturated NAEs, the ends of the molecules are brought together in U-shaped structures (Fig. 4). These clusters accounted for 71% (18:1 NAE), 72% (18:2 NAE), 63% (18:3 NAE), and 41% (20:4 NAE) of the conformations. A low energy cluster of helical shapes was also identified for each of the polyunsaturated NAEs, accounting for 12% (18:2 NAE), 12% 18:3 NAE, and 29% (20:4 NAE) of the conformations, whereas a small cluster of extended shapes (2%) was demonstrated for 18:1 NAE (Fig. 4). All other low energy clusters identified for each compound included less than 2% of conformers.
A conformational analysis of HU210 revealed two major conformational families with the dimethylheptyl side chain extending either axial (58%) or perpendicular (42%) to the tricyclic system (supplemental Fig.  S3). The latter conformation is considered to confer activity (34,35). Optimal RMSD alignment of representative helical conformers of 18:2, 18:3, and 20:4 NAEs to a representative conformer of HU210 revealed a good fit of the pharmacophoric elements for 18:3 NAE (RMSD, 0.7 Å) and 20:4 NAE (RMSD, 0.6 Å), but not for 18:2 NAE (RMSD, 1.3 Å) (Fig.  S3, supplement). However, the terminal trans carbon chain of C18 NAEs was two carbons shorter than the dimethylheptyl side chain of HU210. Our modeling further implicates that the imperfect fit of 18:2 NAE also originates from the propensity of the acyl chain to adopt a wider curvature, hindering the saturated tail to extend in a direction preferred by the dimethylheptyl side chain of the "bioactive" conformation of HU210. Tong et al. (36) arrived at a similar pharmacophore alignment of 20:4 NAE using another cannabinoid as template.

DISCUSSION
The discovery that the endocannabinoid anandamide (20:4 NAE) can directly activate TRPV 1 (20) put focus on NAEs as potential endogenous TRPV 1 modulators. In the present study, we show that endogenous C18 NAEs, having at least one double bond, cause a TRPV 1 -mediated vasorelaxation in rat mesenteric arteries and activate the human orthologue of TRPV 1 expressed in HEK293 cells. Among these lipids, N-linolenoylethanolamine (18:3 NAE) and linoleoylethanolamine (18:2 NAE) have not previously been shown to interact with TRPV 1 . Whereas these lipids activate native TRPV 1 at submicromolar concentrations (present study), they are poor ligands at cannabinoid CB 1 and CB 2 receptors (25,26). Structure-activity relationship studies of NAEs at cannabinoid CB 1 receptors have shown that an acyl chain length of at least 20 carbons and a minimum of three double bonds are required for optimal activity (37). As shown in the present study, 18:3 NAE, 18:2 NAE, and 18:1 NAE are almost as potent as anandamide and 22:4 NAE to elicit TRPV 1 -mediated vasorelaxation (20). Collectively, these findings indicate clear dif-ferences in ligand recognition properties between TRPV 1 and cannabinoid CB 1 receptors with respect to this class of endogenous lipids.
N-Linolenoylethanolamine (18:3 NAE), 18:2 NAE, 18:1 NAE, and 22:4 NAE are all endogenous lipids (24). Interestingly, unsaturated C18 NAEs are considerably more abundant than anandamide in many tissues, including brain and small intestine (24, 38 -40). In the present study, we show that all four C18 NAEs and 22:4 NAE together with anandamide are present in sensory ganglia and blood vessels. In both these tissues, the levels of 18:1 NAE are substantially higher than those of anandamide. Interestingly, activation of cultured neurons, including dorsal root ganglion neurons, leads to the formation of TRPV 1 ligands, such as anandamide and 18:1 NAE (41,42). Likewise, products of 12-lipoxygenase, such as 12-(S)-hydroperoxyeicosatetraenoic acid and 12-(S)-hydroxyeicosatetraenoic acid, are also synthesized in cultured dorsal root ganglion neurons and represent another group of potential endogenous TRPV 1 active eicosanoids (43,44). Anandamide and capsaicin share the same binding site on the cytosolic domain of TRPV 1 (20,(45)(46)(47), and anandamide is more potent when administered intracellularly (48). This raises the interesting possibility that NAEs and lipoxygenase products generated within sensory neurons act as intracellular modulators of TRPV 1 .
All three unsaturated C18 NAEs produced consistent activation of TRPV 1 in rat blood vessels and the FLIPR assay, but evoked TRPV 1mediated currents in only 59% of capsaicin-sensitive HEK293 cells. Non-responding cells cannot be detected in the FLIPR assay, because the calcium signal represents the average signal from thousands of cells. The observation that capsaicin-sensitive cells (both sensory neurons and TRPV 1 -transfected cells) may display variability in the responsiveness to 18:1 NAE has been reported previously (27,28). This variability has been attributed to different levels of TRPV 1 expression and phosphorylation among individual cells (27,28). Indeed, protein kinase activation enhances TRPV 1 -dependent responses to both 18:1 NAE and anandamide (19, 49 -52). Differences between bioassays with respect to post-translational modifications of TRPV 1 and the presence of efficient amplification systems in intact tissues could also explain why the potencies of the unsaturated NAEs were consistently higher in blood vessels than in TRPV 1 -expressing HEK293 cells.
In contrast to the unsaturated NAEs, the saturated 18:0 NAE N-stearoylethanolamine did not activate TRPV 1 in any of the bioassays    (20:4 NAE). The six key atoms chosen for the pharmacophore overlay were as follows: the carbonyl oxygen, the carboxamide nitrogen, C 3 , C 5 , C 7 , and the C 9 carbon. The dashed lines show the six atoms applied for optimal RMS superposition. used. Moreover, 18:0 NAE did not displace the selective TRPV 1 ligand [ 3 H]resiniferatoxin from mouse cortical slices (53). The saturated 16:0 NAE N-palmitoylethanolamine failed to activate TRPV 1 in rat blood vessels and HEK293 cells (20). With the exception of 16:0 NAE, which produced a small activation, saturated long chain NAEs (C12-18) did not induce calcium responses in HEK293 cells expressing hTRPV 1 (29). However, these lipids consistently enhanced anandamide-induced calcium responses in the same bioassay system (29). Although such an "entourage" effect was not observed with 18:0 NAE in the present study, 16:0 NAE did enhance the vasodilator response to anandamide in rat mesenteric arteries (54).
Molecular modeling, using capsaicin as template, provided a rational explanation why unsaturated, but not saturated, NAEs are able to activate TRPV 1 . Our findings are in line with the previous observation that the activity at TRPV 1 is lost when the midchain unsaturation in olvanil (N-oleoylvanillylamine) is removed (55). According to SAR studies of saturated N-acylvanillylamines, the in vivo biological activity peaks at an acyl chain length of 9 carbons (56), which corresponds to the length of the acyl chain before the U-turn of the unsaturated long chain C18 NAEs and anandamide. Thus, the unsaturation at C9 of long chain NAEs may have a conformational effect, allowing the molecules to reach the binding site, rather than participating in the interaction with the channel protein.
In addition to the low energy U-shaped conformational families of C18 NAEs and anandamide, a second large helical conformational family of 18:2 NAE, 18:3 NAE, and anandamide was found. A similar low energy helical conformation of anandamide has been reported previously using computational methods (36). As shown previously, the cannabinoid CB 1 receptor recognizes NAEs with fatty acid chains having at least (i) three homoallylic double bonds and (ii) five terminal unsaturated carbons (25). The lack of one or both of these structural features in the C18 NAEs could explain the weak effect of these lipids on cannabinoid CB 1 receptors. As shown by the present study, only anandamide, in its helical form, displays a high degree of matching of important pharmacophoric elements with the potent cannabinoid CB 1 receptor agonist HU210. Interestingly, it has recently been shown that anandamide prefers an extended conformation in membrane phospholipid bilayers (57). Although it remains to be shown which of these configurations is important for binding to the cannabinoid CB 1 receptor, the ability of anandamide to adopt multiple low energy conformations is likely to explain the intriguing pharmacology of anandamide, being a dual activator of both TRPV 1 and cannabinoid receptors.
In conclusion, we have identified novel endogenous TRPV 1 activators, belonging to the unsaturated C18 NAE class of lipids. Molecular modeling have demonstrated that these unsaturated C18 NAEs adopt U-shape conformations, which fit with several of the pharmacophoric elements of capsaicin. This provides an explanation for the pivotal role of acyl chain unsaturation for activity at TRPV 1 . These NAEs, which are poor ligands at the cannabinoid CB 1 receptor, may act in concert with anandamide and lipoxygenase products as intracellular TRPV 1 modulators in capsaicin-sensitive sensory neurons.