Bidirectional control of airway responsiveness by endogenous cannabinoids

anandamide in lung control of In support of this conclusion, the CB1 enhances capsaicin-evoked bronchospasm and cough. Our results may account for the contrasting bronchial actions of cannabis-like drugs in humans, and provide a frame-work for the development of more selective cannabinoid-based agents for the treatment of respiratory pathologies.

Supplementary information is available on Nature's World-Wide Web site (http://www.nature.com) or as paper copy from the London editorial office of Nature.
analyses indicate that anandamide is synthesized in lung tissue on calcium-ion stimulation, suggesting that locally generated anandamide participates in the intrinsic control of airway responsiveness. In support of this conclusion, the CB1 antagonist SR141716A enhances capsaicin-evoked bronchospasm and cough. Our results may account for the contrasting bronchial actions of cannabis-like drugs in humans, and provide a framework for the development of more selective cannabinoid-based agents for the treatment of respiratory pathologies.
To explore the functional roles of the cannabinoid system in airway physiology, we investigated the effects of the endogenous cannabinoid, anandamide 6,7 , on the responsiveness of bronchial smooth muscle. Administration of capsaicin, the pungent component of chilli pepper 8 , produces a potent bronchoconstriction in anaesthetized guinea pigs (Fig. 1a) or rats (in g per animal, intratracheal: 10, 41 Ϯ 6% of maximal brochospasm; 30, 55 Ϯ 12%; 100, 81 Ϯ 19; means Ϯ s:e:m:, n ¼ 3). This response, which has been extensively studied both in animals and humans, is thought to result from the activation of capsaicin (or 'vanilloid') receptors on sensory C-fibres 9 . Accordingly, the constricting effects of capsaicin were blocked by the vanilloid antagonist capsazepine (Fig. 1a). When administered systemically before capsaicin, anandamide produced a dose-dependent attenuation of bronchospasm in guinea-pigs (Fig. 1b) and rats (capsaicin, 10 g per animal, intratracheal: 37:2 Ϯ 4:2% of maximal bronchospasm; capsaicin after anandamide, 1 mg per kg intravenous (i.v.), 14 Ϯ 7%, n ¼ 3). This effect was completely reversed by the selective CB1 cannabinoid antagonist SR141716A (ref. 10) but only slightly reduced with a maximal dose of the CB2 antagonist SR144528 (ref. 11) (Fig. 1b and data not shown). Palmitylethanolamide, a structural analogue of anandamide that inhibits nociception in mice, but does not interact with CB1 receptors 12 , was ineffective at alleviating capsaicin-evoked bronchospasm (data not shown). When given as an aerosol to conscious guinea-pigs, capsaicin stimulates C-fibre activity in the upper respiratory tract and triggers coughing 13,14 . Systemic or aerosolized anandamide reduced capsaicin-evoked coughing, an effect that was cancelled by CB1 receptor blockade (Fig. 1c).
Anandamide had no direct bronchomotor action, except at the highest dose tested (5 mg per kg), at which the compound elicited a small bronchoconstriction (11:8 Ϯ 5:9% of maximal; mean Ϯ s:e:m:, n ¼ 5). To investigate this response further, we examined the effects of anandamide in anesthetized rodents that were deprived of the bronchoconstricting tone conferred by the vagus nerve 9 . After vagotomy and atropine administration, which eliminate vagal influences, systemic application of anandamide produced a dose-dependent bronchoconstriction in guinea pigs (Fig. 2a) and rats (in mg per kg i.v.: 1, 0 Ϯ 0% of maximal bronch-ospasm; 3, 12 Ϯ 1:7%, 5, 18:3 Ϯ 1:2%; n ¼ 3). Similar effects were observed when anandamide was injected into the guinea-pig bronchi through a tracheal catheter (Fig. 2b), or applied to isolated strips of guinea-pig lung parenchyma ( Fig. 2c-e). The slow onset of the anandamide response in strips of guinea-pig lung is consistent with results obtained in other isolated tissues 6 ; the low potency of anandamide in this preparation may be accounted for by limited tissue penetration and/or rapid inactivation. In agreement with this possibility, the anandamide transport inhibitor AM404 enhanced anandamide-evoked contractions in isolated strips of guinea-pig lung (anandamide, 50 M, 0:336 Ϯ 0:07 dyne mg Ϫ 1 of tissue; anandamide plus AM404, 28 M, 0:638 Ϯ 0:06 dyne mg Ϫ 1 of tissue; P Ͻ 0:05, n ¼ 6). The CB1 antagonist SR141716A blocked whereas the CB2 antagonist SR144528 had no effect (data not shown). The cannabinoid agonist HU210 was also potent at eliciting guinea pig bronchial muscle constriction after tracheal administration (in g per animal: 10:0 Ϯ 0:6% of maximal bronchospasm; 1, 30 Ϯ 1:2%; 10, 60 Ϯ 2:2%; 30, 100%; n ¼ 6). Anandamide has been claimed to activate vanilloid receptors 15 . However, the vanilloid antagonist capsazepine had no effect on anandamideevoked bronchospasm at a dose that completely prevented the capsaicin response (0.2 mg per kg i.v.) (data not shown). These results indicate that removing the vagal excitatory tone unmasked a bronchoconstricting activity of anandamide mediated through CB1 receptors. The ability of anandamide to influence bronchial muscle contractility after local administration suggests that this compound may exert its effects by activating CB1 receptors located within the airways. To test this possibility, we examined the ultrastructural localization of CB1 receptors in rat lungs by electron microscopy, using an antibody directed against the intracellular carboxy terminus of the CB1 receptor protein. Immunogold staining revealed that CB1 receptors are present on nerve fibres distributed among bronchial and bronchiolar smooth muscle cells (Fig. 3a-c), or between the longitudinal and circular smooth muscle layers, where several axons are packed together into glial capsules (Fig. 3d, e). All bundles contained at least one CB1-receptor-positive axon. Detailed evaluation (20 bundles consisting of 91 axons followed through at least 25 consecutive sections) revealed that 36% of the axons were labelled with the CB1 receptor antibody. The gold particles that labelled CB1 receptors were attached to the inner surface of the axon plasma membrane, either at the release site or in the preterminal segments. This is consistent with the fact that the antibody we used recognizes the intracellular C terminus of the CB1 receptor protein. Axon terminals bearing CB1 receptor immunoreactivity were in close proximity to smooth muscle cells (0.2-0.5 m), and contained a large number of small agranular vesicles, along with few dense-core vesicles (Fig. 3a, b). In some cases, CB1 receptor immunoreactivity was adjacent to clusters of vesicles accumulated at the plasma membrane, which most probably represent neurotransmitter release sites (Fig. 3c). Next, to determine whether CB1 receptors are localized on noradrenaline-containing and/or non-noradrenaline-containing fibres, we used a combination of immunogold staining for CB1 receptors, and immunoperoxidase staining for neuropeptide Y (NPY), a co-transmitter in sympathetic neurons 9 . We found that 63% of NPY-bearing axons were also CB1-receptor-positive (Fig. 3f, g). Notably, however, extensive labelling was observed on many NPY-negative axons (data not shown), suggesting that both noradrenaline-containing and/or non-noradrenaline-containing nerves may express CB1 receptors.
The finding that CB1 receptors are found predominantly, if not exclusively, on axon terminals of airway nerves indicates that anandamide may regulate bronchial smooth muscle tone through a prejunctional mechanism. Inhibition of excitatory neurotransmission in the airways may provide a parsimonious explanation for the ability of anandamide to oppose capsaicin-evoked bronchospasm and cough. This interpretation is further supported by the ability of anandamide and other cannabinoid agonists to inhibit neurotransmitter release in both peripheral tissues [16][17][18] and the central nervous system 19 . The mechanism underlying the constricting actions of anandamide in animals lacking acetylcholinemediated control is currently unknown. One possibility, which is consistent with the co-localization of CB1 receptors with NPY, is that anandamide inhibits the release of a bronchodilating mediator. Alternatively, anandamide may interact with CB1 receptors on smooth muscle. Our failure to detect CB1 receptor immunoreactivity in lung smooth muscle may have been caused by insufficient sensitivity of our technique, or by the presence in smooth muscle of a receptor variant that is not recognized by our antibody. Interestingly, northern blot analyses suggest that alveolar type II cells in the lung may express two different CB1 receptor messenger RNA species 20 .
To test the possibility that endogenous cannabinoids regulate airway responsiveness, we determined the intrinsic effects of CB1 and CB2 antagonists on bronchospasm and cough in guinea-pigs. CB1 receptor blockade with SR141716A had no bronchomotor consequences per se, but significantly enhanced the bronchoconstriction and coughing evoked by capsaicin administered through a tracheal catheter (Fig. 4a, b), even in the presence of anandamide (Fig. 1b, c). This response did not depend on the capsaicin administration route, as it was also seen after i.v. injection of the drug (30 mg per kg capsaicin alone, 55:3 Ϯ 8:2% of maximal bronchospasm; capsaicin after SR141716 (0.5 mg per kg i.v.), 92:3 Ϯ 3:4%; P Ͻ 0:05, n ¼ 3). The CB2 antagonist SR144528 had no effect on capsaicin-induced bronchoconstriction and cough (data not shown).
Although the bronchomotor actions of the CB1 antagonist may be accounted for by its inverse agonist properties 21 , evidence suggests that this drug acted by opposing an ongoing cannabinoid modulation. First, the lack of effect seen with the CB1 antagonist in the absence of capsaicin is incompatible with an inverse agonist behaviour. Second, analyses by high-performance liquid chromatography (HPLC) coupled to positive-ionization electrospray mass spectrometry (MS) revealed that anandamide is synthesized in rat lung tissue through a Ca 2+ -ion-activated mechanism (Fig. 5). Rat lung membranes produced on average 0:6 Ϯ 0:2 pmol of anandamide per mg of protein in the presence of the Ca 2+ chelator EGTA (1 mM); and 1:6 Ϯ 0:2 pmol of anandamide per mg of protein in the presence of Ca 2+ (3 mM) (mean Ϯ s:e:m:, n ¼ 4; P Ͻ 0:05 between EGTA and Ca 2+ ; Student's t-test) (Fig. 5, inset). Guinea-pig lung membranes produced 0:9 Ϯ 0:3 pmol of anandamide per mg of protein in the presence of EGTA; and 8:8 Ϯ 1:2 pmol of anandamide per mg of protein in the presence of Ca 2+ (n ¼ 4; P Ͻ 0:0001; Student's t-test).
Our results indicate that activation of CB1 receptors by locally released anandamide may participate in the control of bronchial contractility. How anandamide exerts such control may depend, however, on the state of the bronchial muscle. When the muscle is contracted, as during capsaicin-evoked bronchospasm, anandamide may counteract this contraction, possibly by inhibiting the prejunctional release of excitatory neurotransmitters and neuropeptides. In contrast, when the smooth muscle is relaxed, as seen after removal of the constricting influence of the vagus nerve, anandamide may cause bronchoconstriction. These state-dependent effects may account for the variable bronchomotor actions of acutely administered cannabis-like agents in humans, and might contribute to the long-term toxicity associated with cannabis smoking 24 . Furthermore, the existence of an intrinsic cannabismediated control of airway function may open perspectives for the development of new antitussive and anti-asthmatic agents. For example, drugs targeting the mechanisms of anandamide inactivation 25 may exert more selective pharmacological effects on bronchial responsiveness than those elicited by direct-acting CB1 agonists.

Biological assays
For bronchospasm, we anaesthetized Dunkin-Hartley guinea-pigs (Charles-River, weighing 200-400 g) or Wistar rats (Charles-River, weighing 200-300 g) with pentobarbital (40 mg per kg intraperitoneal (i.p.)) and fentanyl (25 g per kg intramuscular), catheterized the trachea and carotid artery to measure airway obstruction and systemic blood pressure, and catheterized the jugular vein to administer drugs. Pancuronium bromide (4 mg per kg i.v.) was administered to prevent spontaneous breathing. The animals were ventilated with room air using a rodent ventilator (U. Basile, Comerio, Italy) run at 60 strokes min −1 ; stroke volume was 3-7 ml. Airway resistance was measured by using a differential pressure transducer (U. Basile) connected by the side-arm of the tracheal catheter to a bronchospasm transducer. Bronchospasm was expressed as a per cent of the maximal response, which was determined by clamping the tracheal catheter before and after each experiment. Drugs were dissolved in saline containing 10% DMSO, and injected through the jugular vein; responses were evaluated at their peak. Arterial blood pressure was measured continuously with a pressure transducer connected to a recorder (U. Basile). To abolish vagal influences on bronchial musculature, in some experiments we bilaterally transected the vagus nerves and administered atropine sulphate (2 mg per kg i.v.). Anandamide was administered 15 min before capsaicin; the antagonists were given 15 min before anandamide.
For coughing, we individually exposed conscious guinea-pigs to aerosolized capsaicin (0.3 mM) for 4 min while recording coughs by using a microphone placed in the exposure chamber 26 . Each animal was treated only once with capsaicin. Anandamide was administered either systemically (i.p. 30 min before capsaicin) or locally (by aerosol, 10 mg per ml, 15 min before capsaicin). Aerosols were prepared with an Air Lister Basic apparatus (Hatù , Italy), regulated at an emission flow rate of 6 l min −1 . For isolated lung strips, guinea-pig parenchymal strips were prepared essentially as described 27 . The strips were placed in 10-ml organ baths containing Krebs' buffer (in mM: NaCl, 118; KCl, 4; K 2 HPO4, 1.2; MgSO 4 , 1.2; CaCl 2 , 2.5; NaHCO 3 , 25.0; glucose, 11.2; supplemented with 7 M atropine sulphate and 15 M indomethacin) at 37 ЊC and aerated with an oxygen/ carbon dioxide mixture (95/5%). Muscle contractions were recorded with an isometric force transducer (U. Basile) and are expressed in dyne per mg of fresh tissue.

Electron microscopy
The lungs were removed from three rats perfused with a phosphate-buffered (0.1 M) fixative containing 4% paraformaldehyde, 0.2% picric acid and 0.05% glutaraldehyde, and were further fixed for 24 h. We carried out immunohistochemical analyses as described 19 . Rabbit C-terminal anti-CB1 and anti-NPY antibodies were used at 1:5,000 and 1:20,000 dilution, respectively. The specificity of the NPY antibody was reported previously 28 ; the specificity of the CB1 antibody has been verified in several ways, including staining of CB1 knockout mice, and will be detailed elsewhere (N. Hájos, I. K., C. Ledent, K. M. and T. F. F., manuscript in preparation).

Membrane preparation and lipid extraction
Lung particulate fractions were prepared as described 29 . We carried out incubations for 1 h at 37 ЊC in Tris buffer (50 mM, pH 7.4) containing CaCl 2 (3 mM) or EGTA (1 mM) and 2 mg ml −1 of membrane protein. Reactions were stopped by adding cold methanol, and the lipids were extracted with chloroform. Before HPLC/MS analysis, we fractionated anandamide and NAPE by silica gel column chromatography 23 .

HPLC/MS
Anandamide was identified and quantified by reversed-phase HPLC coupled to positive ionization electrospray MS, by using an isotope-dilution method as described 30 . We purified NAPE species by reversed-phase HPLC on a C 18 Bondapak column (300 ϫ 3:9 mm internal diameter 5 m) (Waters) maintained at 20 ЊC and interfaced with an Agilent HP1100 model mass spectrometer. HPLC conditions consisted of a linear gradient of methanol in water (from 75 to 100% methanol in 30 min) with a flow rate of 1 ml min −1 . Under these conditions, different NAPE species were eluted from the column as a group of peaks at retention times comprising 27-29 min. MS analyses were performed with the electrospray ion source set in the negative ionization mode; the V cap set at 5 kV; and the fragmentor voltage set at 200 V. Nitrogen was used as a drying gas at a flow rate of 12 l min −1 . The drying gas temperature was set at 350 ЊC, and the nebulizer pressure at 30 PSI. For quantitative purposes, we extracted diagnostic ions (deprotonated molecules, ½M Ϫ Hÿ Ϫ ) from full scan data and quantified them by comparison with an external standard (1-palmityl-2-oleyl-sn-glycero-phosphoethanolamine-N-arachidonyl, Avanti Polar Lipids). NAPE 2 was eluted from the column as a doublet, and the areas under both peaks were combined for quantification.

Data analysis
Results are expressed as means Ϯ s:e:m: The significance of differences among groups was evaluated using Student's t-test or analysis of variance followed by Dunnett's test.