Modulation of anxiety through blockade of anandamide hydrolysis

The psychoactive constituent of cannabis, Δ9-tetrahydrocannabinol, produces in humans subjective responses mediated by CB1 cannabinoid receptors, indicating that endogenous cannabinoids may contribute to the control of emotion. But the variable effects of Δ9-tetrahydrocannabinol obscure the interpretation of these results and limit the therapeutic potential of direct cannabinoid agonists. An alternative approach may be to develop drugs that amplify the effects of endogenous cannabinoids by preventing their inactivation. Here we describe a class of potent, selective and systemically active inhibitors of fatty acid amide hydrolase, the enzyme responsible for the degradation of the endogenous cannabinoid anandamide. Like clinically used anti-anxiety drugs, in rats the inhibitors exhibit benzodiazepine-like properties in the elevated zero-maze test and suppress isolation-induced vocalizations. These effects are accompanied by augmented brain levels of anandamide and are prevented by CB1 receptor blockade. Our results indicate that anandamide participates in the modulation of emotional states and point to fatty acid amide hydrolase inhibition as an innovative approach to anti-anxiety therapy.

The psychoactive constituent of cannabis, ∆ 9 -tetrahydrocannabinol, produces in humans subjective responses mediated by CB1 cannabinoid receptors, indicating that endogenous cannabinoids may contribute to the control of emotion. But the variable effects of ∆ 9 -tetrahydrocannabinol obscure the interpretation of these results and limit the therapeutic potential of direct cannabinoid agonists. An alternative approach may be to develop drugs that amplify the effects of endogenous cannabinoids by preventing their inactivation. Here we describe a class of potent, selective and systemically active inhibitors of fatty acid amide hydrolase, the enzyme responsible for the degradation of the endogenous cannabinoid anandamide. Like clinically used anti-anxiety drugs, in rats the inhibitors exhibit benzodiazepine-like properties in the elevated zero-maze test and suppress isolation-induced vocalizations. These effects are accompanied by augmented brain levels of anandamide and are prevented by CB1 receptor blockade. Our results indicate that anandamide participates in the modulation of emotional states and point to fatty acid amide hydrolase inhibition as an innovative approach to anti-anxiety therapy.
tion of compound 4 with the active site of FAAH. Testing this hypothesis led to the design of the biphenyl derivative 5 (IC 5 0 = 63 ± 9 nM), which was further optimized by systematic modifications of the distal phenyl group, resulting in the potent inhibitor 6 (IC 50 = 4.6 ± 1.6 nM; Table 1). The lead optimization process will be reported elsewhere.
Kinetic analyses and dialysis experiments indicate that compounds 4 and 6 may inhibit FAAH activity through an irreversible interaction with the enzyme (data not shown), possibly due to a nucleophilic attack of an active serine residue on the carbamate group. This mechanism sets the present compounds apart from the α-keto heterocycle derivatives described previously 16 , which act as competitive FAAH inhibitors. A further indication of such a distinction is that in the α-keto heterocycle series potency is strongly dependent on the hydrophobicity of the flexible acyl chain, whereas in the carbamate series potency is modulated by the shape of the rigid aromatic moiety. Accordingly, when we replaced the biphenyl of compound 5 with a 5-phenylpentyl group, representing the most effective acyl chain in the α-keto heterocycle series, the inhibitory activity was lost (compound 7; Table 1).
Compounds 4 (URB532) and 6 (URB597) blocked the FAAHcatalyzed hydrolysis of exogenous [ 3 H]anandamide in primary cultures of intact cortical neurons, with IC 50 values that paralleled those obtained in membrane preparations (URB532, 214 ± 79 nM; URB597, 0.50 ± 0.05 nM; n = 8; Fig. 1a). By contrast, compound 8, an analog of URB532 that does not inhibit FAAH in membranes (Table 1), had no such effect (Fig. 1b). Moreover, URB532 and URB597 selectively impaired the breakdown of [ 3 H]anandamide without reducing its carrier-mediated uptake, causing non-metabolized [ 3 H]anandamide to accumulate in, and eventually exit from, the neurons. Thus, after a 4-minute incubation with [ 3 H]anandamide, the intracellular content of non-metabolized [ 3 H]anandamide was higher in inhibitortreated neurons than in control neurons (Fig. 1c). As expected, the anandamide transport blocker N-(4-hydroxyphenyl)arachidonamide (AM404) had an opposite effect, substantially reducing [ 3 H]anandamide internalization 4 (Fig. 1c). When neurons treated with URB597 were exposed for 4 minutes to [ 3 H]anandamide and then incubated for 15 minutes in an [ 3 H]anandamide-free solution, 42.6 ± 8.7% of the accumulated [ 3 H]anandamide was released back into the medium (n = 3; Fig.  1d). This process was linear with time ( Fig. 1e) and was not inhibited by AM404 (Fig. 1d), indicating that it occurred through passive diffusion rather than reverse transport. No such time-dependent release was observed in control neurons, the medium of which contained only residual levels of [ 3 H]anandamide carried over from the pre-incubation period. These studies identify a new class of carbamate inhibitors of FAAH activity, which potently block anandamide breakdown in intact brain neurons.

ARTICLES
cholinesterase, horse plasma butyryl cholinesterase and rat brain monoglyceride lipase (MGL; Table 2). The lack of MGL inhibition is particularly noteworthy in light of the proposed role of this enzyme in the biological inactivation of 2-arachidonoylglycerol 17 (2-AG), another endogenous cannabinoid present in brain [18][19][20] . Furthermore, URB532 and URB597 had no effect on anandamide transport in human astrocytoma cells or on the binding of a high-affinity ligand to CB1 and CB2 receptors ( Table 2). In addition, URB532 (10 µM) did not significantly interact with a panel of 21 receptors, ion channels and neurotransmitter transporters, which included adenosine A 1 , A 2A and A 2B ; adrenergic α 1A , α 2A , β 1 and β 2 ; dopamine D 1 and D 2 ; glutamate Nmethyl-(D)-aspartate; γ-amino-butyric acid (GABA) A agonist site; histamine H 1 ; opiate µ; muscarinic M 2 ; and brain nicotinic receptors (data not shown). This high selectivity for FAAH encouraged us to examine the effects of URB532 and URB597 in live animals.

FAAH inhibition in vivo
Intraperitoneal (i.p.) injection of either URB532 or URB597, but not the inactive analog 8, produced a profound, dose-dependent inhibition of brain FAAH activity (Fig. 2a). In six experiments, half-maximal inhibition was reached at 0.60 ± 0.09 mg kg -1 of URB532, and 0.150 ± 0.007 mg kg -1 of URB597. The discrepancy between potency ratios of URB532 and URB597 in vitro (100-fold) and in vivo (4-fold) presumably reflects differences in bioavailability or brain penetration of the two compounds. After injection of a maximal dose of URB597 (0.3 mg kg -1 , i.p.), FAAH inhibition was rapid in onset (<15 minutes), persistent (>6 hours; Fig.  2b) and accompanied by significant elevations in the brain levels of anandamide (Fig. 2c)  ). Parallel changes in FAAH activity and fatty acid ethanolamide levels were also measured in various peripheral tissues (data not shown). In agreement with the lack of MGL inhibition noted in our in vitro experiments (Table 2), URB597 did not change the brain content of 2-AG (Fig. 2d).
As previously observed in Faah -/mice 9 , FAAH inhibition was associated with increased sensitivity to the administration of exogenous anandamide. Accordingly, URB597 (0.3 mg kg -1 , i.p.) intensified and prolonged the decrease in body temperature elicited by a subthreshold dose of anandamide (5 mg kg -1 , i.p.), whereas it had no effect when injected alone (Figs. 2e and f). Underscoring the role of CB1 receptors in this response, the effect of anandamide plus URB597 was prevented by the CB1 antagonist SR141716A (rimonabant; Figs. 2e and f).

Pharmacological properties of FAAH inhibitors in vivo
Although URB532 and URB597 increased brain anandamide lev-  Values indicate the maximal concentrations of FAAH inhibitor tested on each target (in µM) and their corresponding selectivity index (SI). The SI is the ratio of maximal inhibitor concentration tested/IC50 for FAAH (from Table 1). els, they did not mimic the spectrum of pharmacological responses produced by exogenous anandamide. Systemic doses of URB532 (10 mg kg -1 , i.p.) or URB597 (0.3 mg kg -1 , i.p.) that maximally blocked FAAH activity produced no catalepsy (rigid immobility), hypothermia or hyperphagia (increased food intake), three typical signs of CB1 receptor activation 11 (data not shown). However, the compounds exerted moderate anti-nociceptive actions in the mouse hot-plate test, which measures response to noxious thermal stimuli. In this model, URB532 and URB597 significantly lengthened response latencies at a dose of 0.5 mg kg -1 (Figs. 3a and b), but not 0.1 mg kg -1 (data not shown). These effects were prevented by a dose of the CB1 antagonist rimonabant (0.2 mg kg -1 , intravenous (i.v.); Figs. 3a and b), which caused no hyperalgesia when administered alone (data not shown). Our results corroborate those obtained in Faah -/mice 9 , indicating that acute disruption of FAAH activity results in a mild CB1-mediated anti-nociception, but no hypothermia or catalepsy.

Anxiolytic effects of FAAH inhibitors
To identify intrinsic actions of anandamide that might be significantly magnified by FAAH inhibition, we turned, for three reasons, to the regulation of emotional reactivity. First, CB1 receptors are expressed at high levels in brain regions such as the amygdala which are implicated in the control of anxiety and fear [21][22][23] . Second, acute administration of cannabinoid drugs pro-duces emotional responses in rodents 11 and humans 10,24 . Third, pharmacological 25,26 or genetic 27,45 thus, disruption of CB1 receptor activity elicits anxiety-like behaviors in rodents, suggestive of the existence of an intrinsic anxiolytic tone mediated by endogenous cannabinoids. We used two pharmacologically validated animal models of anxiety: the elevated zero-maze test and the isolation-induced ultrasonic emission test. The zero maze consists of an elevated annular platform with two open and two closed quadrants and the test is based on the conflict between an animal's instinct to explore its environment and its fear of open spaces, where it may be attacked by predators 28,29 . Clinically used anxiolytic drugs such as the benzodiazepines increase the proportion of time spent in, and the number of entries made into, the open compartments. Similarly, URB532 (0.1-10 mg kg -1 , i.p.) and URB597 (0.05-0.1 mg kg -1 , i.p.) evoked anxiolytic-like responses at doses that corresponded to those required to inhibit FAAH activity in vivo (F = 24.7, df = 4/41, P < 0.001; F = 7.7, df = 2/27, P < 0.01; Figs. 4a and b). In keeping with an involvement of endogenous anandamide, the anxiolytic-like effects of both compounds were attenuated by a non-anxiogenic dose of rimonabant (2 mg kg -1 , i.p.; F = 14.87, df = 3/28, P < 0.001; F = 15.2, df = 3/28, P < 0.001; Figs. 4c and d). Moreover, the effects were apparently dissociated from overall changes in motor behavior. Indeed, although URB532 elicited in adult rats a modest decrease in ambulation (which was also antagonized by rimonabant; data not shown), it did so at doses that were higher than those needed to cause anxiolysis (≥10 mg kg -1 ; F = 3.57, df = 2/22, P < 0.05; Fig. 4e). We confirmed this dissociation by testing URB532 and URB597 in the ultrasonic vocalization emission model, which measures the a b Fig. 3 Anti-nociceptive actions of URB532 and URB597. Shown are effects of URB532 (a) and URB597 (b) (both at 0.5 mg kg -1 , i.p.) on response latencies in the mouse hot-plate test, in the absence or presence of the CB1 antagonist rimonabant (Ri, 0.2 mg kg -1 , i.v.). FAAH inhibitors and rimonabant were injected 60 min and 40 min before tests, respectively. Rimonabant alone had no effect on response latencies (not shown). *P < 0.05; ANOVA followed by Dunnett's test (n = 12). , vehicle; í, urb597. Error bars, s.e.m. number of stress-induced vocalizations emitted by rat pups removed from their nest [30][31][32] . As seen with anxiolytic drugs, the FAAH inhibitors reduced ultrasonic calls (F = 10.8, df = 5/33, P < 0.001; F = 19.3; df = 4/25, P < 0.001; Figs. 4f and g) at doses (URB532, 1 and 5 mg kg -1 ; URB597, 0.1 mg kg -1 ) that had no effect on pup movement (data not shown). These anxiolytic-like responses were blocked by rimonabant (2 mg kg -1 ; Figs. 4f and g).

Discussion
We have developed a new class of agents that prevents anandamide inactivation by targeting the intracellular enzymatic activity of FAAH. URB597, the most potent member of this class, inhibited FAAH activity with an IC 50 value of 4 nM in brain membranes and 0.5 nM in intact neurons, and an ID 50 value of 0.15 mg kg -1 after systemic administration in the rat. This compound had much greater selectivity for FAAH than other cannabinoidrelated targets, including cannabinoid receptors (selectivity index, >25,000) and MGL, an enzyme involved in the deactivation of the endogenous cannabinoid ester 2-AG (selectivity index, >7,500). Such target discrimination was matched by a lack of overt cannabimimetic effects in vivo. Thus, at doses that almost abolished FAAH activity and substantially raised brain anandamide levels, URB597 and its analog URB532 did not evoke catalepsy, reduce body temperature or stimulate feeding, three key symptoms of cannabinoid intoxication in the rodent 11 . Nevertheless, the compounds did elicit marked anxiolytic-like responses, which paralleled their ability to inactivate FAAH and were prevented by the CB1 receptor antagonist rimonabant. We interpret these findings to indicate that URB597 and URB532 may selectively modulate anxiety-like behaviors by enhancing the tonic actions of anandamide on a subset of CB1 receptors which may normally be engaged in controlling emotions. Forebrain sites that might be implicated in such actions include the basolateral amygdala, the anterior cingulate cortex and the prefrontal cortex, all key elements of an "emotion circuit" 33 that contains high densities of CB1 receptors 21,22 . CB1 receptors in these structures are localized to the axon terminals of a subpopulation of GABAergic interneurons, which also express the peptide cholecystokinin 23,34 (CCK). The anxiogenic properties of CCK 35 and the ability of CB1 agonists to inhibit K + -evoked CCK release from hippocampal slices 36 indicate that interactions between this peptide and anandamide may participate in the control of anxiety.
In addition to their anxiolytic-like actions, URB597 and URB532 exerted modest anti-nociception in a model of acute pain, which also was sensitive to CB1 receptor blockade. These findings are similar to those reported for Faah -/mice 9 and support the proposed roles of anandamide in the intrinsic modulation of pain 37 . However, as emotional states may strongly influence pain sensation, it is possible that anxiolysis might have contributed to the mild anti-nociceptive effects of the FAAH inhibitors.
URB597 and URB532 increased brain anandamide levels without modifying those of the second endogenous cannabinoid, 2-AG. It is therefore likely that the pharmacological actions of these compounds, which are sensitive to the CB1 antagonist rimonabant, are primarily due to anandamide accumulation. But the FAAH inhibitors also produced large elevations in the levels of two anandamide analogs, palmitoylethanolamide and oleoylethanolamide, whose biological effects are independent of CB1 receptors 7,8 . Thus, we cannot exclude the possibility that additional properties of URB597 and URB532, mediated by these fatty ethanolamides, remain to be discovered.
Our results define a novel class of inhibitors of FAAH activity, which enhance endogenous anandamide signaling without directly interacting with cannabinoid receptors. The behavioral profile of these agents-characterized by anxiolysis and mild analgesia-reveals a key role for anandamide in the regulation of emotional states and indicates a new mechanistic approach to anti-anxiety therapy.

Methods
Animals and cells. We used Wistar rats (200-350 g) and Swiss mice (20 g). All procedures met the National Institutes of Health guidelines for the care and use of laboratory animals and those of the Italian Ministry of Health (D.L. 116/92). We prepared cultures of cortical neurons from 18-day-old Wistar rat embryos and maintained them as described 38 . We purchased astrocytoma cells from American Type Culture Collection (Manassas, Virginia).
Chemicals. Anandamide and related lipids were synthesized in the laboratory 39 . SR141716A (rimonabant) was provided by RBI (Natick, Massachusetts) as part of the Chemical Synthesis Program of the National Institutes of Health. AM404 was from Tocris (Avonmouth, UK) and other drugs from Sigma.
In other experiments, after having exposed the neurons to [ 3 H]anandamide for 4 min, we rinsed the cells and measured [ 3 H]anandamide release as described above.
High-performance liquid chromatography/mass spectrometry. We extracted lipids with methanol-chloroform and fractionated them by column chromatography 39 . Anandamide and other fatty acid derivatives were quantified by high-performance liquid chromatography/mass spectrometry 39 .
Body temperature and catalepsy. We administered compounds (in saline/Tween 80/polyethylene glycol, 90/5/5, i.p.) immediately before tests. We measured body temperature with a rectal probe connected to a digital