Functional Role of High-Affinity Anandamide Transport, as Revealed by Selective Inhibition

Anandamide, an endogenous ligand for central cannabinoid receptors, is released from neurons on depolarization and rapidly inactivated. Anandamide inactivation is not com-pletely understood, but it may occur by transport into cells or by enzymatic hydrolysis. The compound N -(4-hydroxyphenyl)arachidonylamide (AM404) was shown to inhibit high-affinity anandamide accumulation in rat neurons and astrocytes in vitro, an indi-cation that this accumulation resulted from carrier-mediated transport. Although AM404 did not activate cannabinoid receptors or inhibit anandamide hydrolysis, it enhanced receptor-mediated anandamide responses in vitro and in vivo. The data indicate that carrier-mediated transport may be essential for termination of the biological effects of anandamide, and may represent a potential drug target. Nerve growth factor (NGF) is a neurotrophic factor secreted by cells that are the targets of innervation of sympathetic and some sensory neurons. However, the mechanism by which the NGF signal is propagated from the axon terminal to the cell body, which can be more than 1 meter away, to influence biochemical events critical for growth and survival of neurons has remained unclear. An NGF-mediated signal transmitted from the terminals and distal axons of cultured rat sympathetic neurons to their nuclei regulated phosphorylation of the transcription factor CREB (cyclic adenosine monophosphate response element – binding protein). Internalization of NGF and its receptor tyrosine kinase TrkA, and their transport to the cell body, were required for transmission of this signal. The tyrosine kinase activity of TrkA was required to maintain it in an autophos-phorylated state upon its arrival in the cell body and for propagation of the signal to CREB within neuronal nuclei. Thus, an NGF-TrkA complex is a messenger that delivers the NGF signal from axon terminals to cell bodies of sympathetic neurons. T he growth and survival of many popula tions of neurons depends on trophic support provided by their target tissue ( 1 ). NGF is secreted by targets of sympathetic and some sensory neurons, and it is also expressed within discrete regions of the central ner vous system ( 1, 2 ). NGF belongs to a family of structurally related neurotrophic factors termed neurotrophins; this family includes brain - neurotrophic (BDNF),

is an endogenous lipid that activates brain cannabinoid receptors and mimics the pharmacological effects of ⌬ 9 -tetrahydrocannabinol, the active principle of hashish and marijuana (1). In humans, such effects include euphoria, calmness, dream states, and drowsiness (2). Depolarized neurons release anandamide (3) through a mechanism that may require the calcium-dependent cleavage of a phospholipid precursor in neuronal membranes (4). Like other modulatory substances, extracellular anandamide is thought to be rapidly inactivated, but the exact nature of this inactivating process is still unclear. A possible pathway is hydrolysis to arachidonic acid and ethanolamine, catalyzed by a membrane-bound fatty acid amide hydrolase (FAAH) highly expressed in rat brain and liver (5). Nevertheless, the low FAAH activity found in brain plasma membranes indicates that this enzyme may be intracellular (5), a possibility that is further supported by sequence analysis of rat FAAH (6). Although anandamide could gain access to FAAH by passive diffusion, the transfer rate is expected to be low because of the molecular size of this lipid mediator (7). In that other lipids including polyunsaturated fatty acids and prostaglandin E 2 (PGE 2 ) enter cells by carrier-mediated transport (8,9), it is possible that anandamide uses a similar mechanism. Indeed, the existence of a rapid, saturable process of anandamide accumulation into neural cells has been reported (3). This accumulation may result from the activity of a transmembrane carrier, which may thus participate in termination of the biological actions of anandamide. Accordingly, we developed drug inhibitors of anandamide transport and investigated their pharmacological properties in cultures of rat cortical neurons or astrocytes.
The accumulation of exogenous [ 3 H]anandamide by neurons or astrocytes fulfills several criteria of a carrier-mediated transport ( Fig. 1) (10). It is a rapid process that reaches 50% of its maximum within about 4 min (Fig. 1A). Furthermore, [ 3 H]anandamide accumulation is temperature-dependent (Fig. 1A) and saturable ( Fig. 1, B and C). Kinetic analyses revealed that accumulation in neurons can be represented by two components of differing affinities (lower affinity: Michaelis constant, K m , ϭ 1.2 M, maximum accumulation rate, V max , ϭ 90.9 pmol/min per milligram of protein; higher affinity: K m ϭ 0.032 M, V max ϭ 5.9 pmol/min per milligram of protein) (Fig.  1B). The higher affinity component may reflect a binding site, however, as it is displaced by the cannabinoid receptor antagonist, SR-141716-A (100 nM) (11). In astrocytes, [ 3 H]anandamide accumulation is represented by a single high-affinity component (K m ϭ 0.32 M, V max ϭ 171 pmol/min per milligram of protein) (Fig. 1C). Such apparent K m values are similar to those of known neurotransmitter uptake systems (12) and are suggestive therefore of high-affinity carrier-mediated transport.
A primary criterion for defining carriermediated transport is pharmacological inhibition. To identify inhibitors of anandamide transport, we first examined compounds that prevent the cellular uptake of other lipids, such as fatty acids (phloretin, 50 M), phospholipids (verapamil, 100 M; quinidine, 50 M), or PGE 2 (bromcresol green, 0.1 to 100 M) (15). Among the compounds tested, only bromcresol green interfered with anandamide transport, albeit with limited potency and partial efficacy (Fig. 3, A and B). Bromcresol green inhibited [ 3 H]anandamide accumulation with an IC 50 (concentration needed to produce half-maximal inhibition) of 4 M in neurons and 12 M in astrocytes and acted noncompetitively (16). Moreover, bromcresol green had no significant effect on the binding of [ 3 H]WIN-55212-2 to rat cerebellar membranes (inhibition constant, K i , ϭ 22 M), FAAH activity in rat brain microsomes (IC 50 Ͼ 50 M), and uptake of [ 3 H]arachidonate or [ 3 H]ethanolamine in astrocytes (121 Ϯ 13% and 103 Ϯ 12%, respectively, at 50 M bromcresol green, n ϭ 3) (17). The sensitivity to bromcresol green, which blocks PGE 2 transport, raised the question of whether anandamide accumulation occurred by means of a PGE 2 carrier. That this is not the case was shown by the lack of [ 3 H]PGE 2 accumulation in neurons or astrocytes (18) and by the inability of PGE 2 to interfere with [ 3 H]anandamide accumulation ( Fig. 2A). Previous results indicating that expression of PGE 2 transporter mRNA in brain tissue is not detectable further support this conclusion (9).
To search for more potent anandamide transport inhibitors, we synthesized and tested a series of structural analogs of anandamide (19). From this screening, we selected the compound N-(4-hydroxyphenyl)arachidonylamide (AM404), which was both efficacious and relatively potent (Fig. 3, C and D; IC 50 was 1 M in neurons and 5 M in astrocytes). As we anticipated from its chemical structure, AM404 acted as a competitive inhibitor (20), suggesting that it may serve as a transport substrate or pseudosubstrate. In contrast, at the concentrations tested AM404 had no effect on FAAH activity (IC 50 Ͼ 30 M) or on uptake of [ 3 H]arachidonate or [ 3 H]ethanolamine (102 Ϯ 4% and 96 Ϯ 14%, respectively, at 20 M AM404, n ϭ 6). Furthermore, a positional isomer of AM404, N-(3-hydroxyphenyl)arachidonylamide (AM403), was significantly less effective than AM404 in inhibiting transport (Fig. 3, C and D). These data provide pharmacological evidence for the existence of a specific anandamide transporter and suggest (i) that neurons and astrocytes may act synergistically in the brain to dispose of extracellular anandamide and (ii) that the transport systems in these two cell types may differ kinetically and pharmacologically (Fig.  1, B and C, and Fig. 3, C and D).
The identification of inhibitors allowed us to examine whether transmembrane transport participates in terminating anandamide responses mediated by cannabinoid receptor activation. Cannabinoid receptors of the CB1 subtype are expressed in neurons (21) where they are negatively coupled to adenylyl cyclase activity (22). Accordingly, in cultures of rat cortical neurons the cannabinoid receptor agonist WIN-55212-2 inhibited forskolin-stimulated adenosine 3Ј,5Ј-monophosphate (cAMP) accumulation (control: 39 Ϯ 4 pmol per milligram of protein; 3 M forskolin: 568 Ϯ 4 pmol per milligram of protein; forskolin plus 1 M WIN-55212-2: 220 Ϯ 24 pmol per milligram of protein), and this inhibition was prevented by the antagonist SR-141716-A (1 M) (555 Ϯ 39 pmol/mg of protein, n ϭ 9) (23). Anandamide produced a similar effect, but with a potency (IC 50 , 1 M) that was 1/20 of that expected from its binding

REPORTS
www.sciencemag.org ⅐ SCIENCE ⅐ VOL. 277 ⅐ 22 AUGUST 1997 constant for CB1 cannabinoid receptors (K i Ϸ 50 nM) (1) (Fig. 4A). The transport inhibitor AM404 bound to CB1 receptors with low affinity (K i ϭ 1.8 M) (19) and did not reduce cAMP concentrations when applied at 10 M (Fig. 4B). Nevertheless, the drug enhanced the effects of anandamide, increasing the potency (by a factor of 10) and decreasing the threshold (by a factor of 1/100), an effect that was prevented by SR-141716-A (Fig. 4A). Thus, a concentration of anandamide that was below threshold when applied alone (0.3 M) produced an almost maximal effect when applied with AM404 (Fig. 4B). Bromcresol green and AM403, which were less effective than AM404 in inhibiting anandamide transport (Fig. 3), were also less effective in enhancing the anandamide response (Fig.  4B). Furthermore, the decreases in cAMP concentrations produced by WIN-55212-2 (which stimulates CB1 receptors but is not subject to physiological clearance) or glutamate [which stimulates metabotropic receptors negatively coupled to adenylyl cyclase (24) and is cleared by a selective transporter (25)] are not affected by any of the anan-damide transport inhibitors tested (26).
These results suggest that pharmacological blockade of carrier-mediated transport protects anandamide from physiological inactivation, enhancing the potency of anandamide to nearly that expected from its affinity for CB1 cannabinoid receptors in vitro. To find out whether this potentiation occurs in vivo, we tested the effects of AM404 on the antinociceptive activity of anandamide in mice. Intravenous anandamide (20 mg per kilogram of body weight) elicited a modest but significant analgesia, as measured by the hot plate test (27) (P Ͻ 0.05, Student's t test); this analgesia disappeared 60 min after injection and was prevented by SR-141716-A (Fig. 4C) (28). Administration of AM404 (10 mg/kg, intravenously) had no antinociceptive effect within 60 min of injection but significantly enhanced and prolonged anandamide-induced analgesia (Fig. 4C) (P Ͻ 0.01, Student's t test).
Our findings indicate that a high-affinity transport system present in neurons and astrocytes has a role in anandamide inactivation by removing this lipid mediator from the extracellular space and delivering it to intracellular metabolizing enzymes such as FAAH (5,6). Therefore, the identification of selective inhibitors of anandamide transport should be instrumental in understanding the physiological roles of the endogenous cannabinoid system and may lead to the development of therapeutic agents. Results are expressed as mean Ϯ SEM of nine independent determinations. One asterisk indicates P Ͻ 0.05 and two asterisks P Ͻ 0.01 (ANOVA followed by Bonferroni test). (C) Effects of AM404 on the analgesic activity of anandamide in the hot plate test. Three groups of six mice received AM404 (10 mg/kg, intravenous), anandamide (20 mg/kg, intravenous), or anandamide plus AM404. The hot plate test (55.5°C) was performed at the times indicated, and latency to jump (in seconds) was measured before (control) and after the drugs were injected. In all groups, latency to jump before injections was 21 Ϯ 0.6 s (n ϭ 18). A fourth group of mice received injections of vehicle alone (saline containing 20% dimethyl sulfoxide), which did not affect latency to jump. One asterisk indicates P Ͻ 0.05 compared with uninjected controls (ANOVA followed by Bonferroni test), and one cross indicates P Ͻ 0.01 compared with anandamide-treated animals (Student's t test). SCIENCE ⅐ VOL. 277 ⅐ 22 AUGUST 1997 ⅐ www.sciencemag.org