Control of pain initiation by endogenous cannabinoids

The potent analgesic effects of cannabis-like drugs 1–4 and the presence of CB1-type cannabinoid receptors in pain-processing areas of the brain and spinal cord 5,6 indicate that endogenous cannabinoids such as anandamide 7 may contribute to the control of pain transmission within the central nervous system (CNS) 8 . Here we show that anandamide attenuates the pain behaviour produced by chemical damage to cutaneous tissue by interacting with CB1-like cannabinoid receptors located outside the CNS. (PEA), which is released together with anandamide from a common

The potent analgesic effects of cannabis-like drugs [1][2][3][4] and the presence of CB1-type cannabinoid receptors in pain-processing areas of the brain and spinal cord 5,6 indicate that endogenous cannabinoids such as anandamide 7 may contribute to the control of pain transmission within the central nervous system (CNS) 8 . Here we show that anandamide attenuates the pain behaviour produced by chemical damage to cutaneous tissue by interacting with CB1-like cannabinoid receptors located outside the CNS. Palmitylethanolamide (PEA), which is released together with anandamide from a common phospholipid precursor 9 ,exerts a similar effect by activating peripheral CB2-like receptors. When administered together, the two compounds act synergistically, reducing pain responses 100-fold more potently than does each compound alone. Gas-chromatography/mass-spectrometry measurements indicate that the levels of anandamide and PEA in the skin are enough to cause a tonic activation of local cannabinoid receptors. In agreement with this possibility, the CB1 antagonist SR141716A and the CB2 antagonist SR144528 prolong and enhance the pain behaviour produced by tissue damage. These results indicate that peripheral CB1-like and CB2-like receptors participate in the intrinsic control of pain initiation and that locally generated anandamide and PEA may mediate this effect.
Although pain perception is thought to be controlled mainly by neurotransmitter systems that operate within the CNS 10 , antinociceptive (pain-relieving or -preventing) mechanisms also occur in peripheral tissues. For example, endogenous opioid peptides released from activated immune cells during inflammation inhibit pain transmission by interacting with opioid receptors on peripheral sensory nerve endings 11 . To determine whether endogenous cannabinoids have an analogous function to that of opioids, we used the formalin test, a behavioural model of injury-induced acute and tonic pain 12 . Injection of dilute formalin into the hind paws of freely moving rodents evokes a pain behaviour consisting of two temporally distinct phases of licking and flexing o f t he injected limb 12 . An early phase involving acute activation of pain-sensing C fibres 13 begins immediately after formalin administration, reaches a peak within 5 min, and then rapidly declines. After an interval of 10-15 min, a second phase of sustained pain behaviour appears, in which sensory fibre activity is accompanied by i nflammation 14 and central sensitization 15 . In mice, the early phase of pain behaviour was blocked when anandamide was injected into the paw together with formalin, whereas both phases were blocked by the synthetic cannabinoid agonists WIN-55212-2 and HU-210 ( Fig. 1a; data not shown). These analgesic effects were prevented by systemic administration of the CB1 antagonist SR141716A (Fig. 1a), but not of the CB2 antagonist SR144528 (Fig. 1a) 16 or of the opioid antagonist naloxone (0.2 mg per kg, intravenous; data not shown). The lack of effect of anandamide on late-phase pain behaviour may be explained by the short lifespan of this compound, which undergoes rapid biological inactivation in tissues 17,18 . In support of this suggestion, the inactivation-resistant analogue methanandamide inhibited pain behaviour during the entire test (Fig. 1a).
Local anandamide injections were not accompanied by central signs of cannabimimetic activity, indicating a peripheral site of action. To test this possibility, we measured the antinociceptive potency of anandamide following local (intraplantar, i.pl.), intravenous (i.v.) or intraperitoneal (i.p.) administrations. Anandamide was 100 times more potent in preventing formalin-evoked pain behaviour when injected locally rather than intravenously, with half-maximal inhibitory doses (ID 50 values) of 0.1 mg per kg and 10 mg per kg, respectively (Fig. 1b). Anandamide had no significant effect when injected intraperitoneally (Fig. 1b). As a further test, we determined the biodistribution of 3 H-labelled anandamide 10 min after i.pl. injection in rats. In three experiments, we found that 94% of recovered 3 H-labelled anandamide remained associated with the injected paw (6648 Ϯ 820 d:p:m: per g, mean Ϯ s:e:m:), whereas little or no radioactivity above background was detected in forebrain, cerebellum and spinal cord (79 Ϯ 19; 165 Ϯ 67; and 90 Ϯ 48 d:p:m: per g, respectively). These results indicate that anan-damide inhibits nociception after formalin injection by activating CB1-like receptors, which may be located on peripheral endings of sensory neurons involved in pain transmission 5,6 .
Anandamide is the ethanolamide of arachidonic acid and is thought to be produced by phosphodiesterase-mediated cleavage of N-arachidonyl phosphatidylethanolamine 9,19 . Enzymatic cleavage of other N-acyl phosphatidylethanolamines may give rise to additional fatty acylethanolamides, the physiological roles of which are still poorly understood 20 . PEA, an acylethanolamide found in neural and non-neural tissues, inhibits mast-cell activation and reduces inflammatory r esponses 21,22 b y a m echanism t hat may involve binding to CB2-like receptors 23 . The molecular identity of these receptors is unknown, although they are likely to be distinct from the CB2 receptors whose encoding genes have been cloned, as PEA shows little or no affinity for these receptors 24 . We found that PEA, but not two closely related analogues, inhibited both early and late phases of formalin-evoked pain behaviour after i.pl. injection in mice (Fig. 2a). This effect may not be explained by the antiinflammatory properties of PEA; 30 min after injection, the analgesic effects of PEA were not accompanied by a reduction in inflammatory oedema (volumes injected into the paw in ml were: control, 0:18 Ϯ 0:003; formalin, 0:37 Ϯ 0:006; formalin plus 50 g PEA, 0:35 Ϯ 0:006), which became apparent only 1 h after formalin administration 22 .
The analgesia produced by PEA was reversed by administration of the CB2 antagonist SR144528 (Fig. 2a), whereas the CB1 antagonist SR141716A and the opioid antagonist naloxone were ineffective ( Fig. 2a; data not shown). In addition, PEA was more potent when administered locally (i.pl.) than systemically (i.v. or i.p.) (Fig. 2b). Together, these results indicate that PEA exerts antinociceptive effects that are mediated by peripheral CB2-like receptors. The cellular localization of such receptors and their possible structural relationship with the CB2 receptor whose gene has been cloned, which is primarily expressed in immune cells 25 , are unknown.
The fact that anandamide and PEA activate pharmacologically distinct receptors and that these two substances can be produced simultaneously in tissues 9 prompted us to examine their possible interactions in vivo. When injected together in equal amounts, anandamide and PEA inhibited the early phase of formalin-evoked pain behaviour with a potency that was approximately 100-fold greater than each of the compounds separately (Fig. 3a). A similar synergistic potentiation occurred in the late phase, on which anandamide had no effect when given alone (Figs 1a and 3b). Earlier administration of either CB1 or CB2 antagonists entirely blocked the response (Fig. 3c). This interaction did not appear to involve pain-processing structures within the brain: injection of PEA in the cerebral ventricles did not affect the behavioural responses to acute thermal stimuli, assessed in the hot-plate test, and did not enhance the inhibitory activity of anandamide administered by the same route (Table 1). These results indicate that the parallel activation of peripheral CB1-and CB2-like receptors by anandamide and PEA results in a synergistic inhibition of peripheral pain initiation.
To test this idea further, we determined the intrinsic effects of CB1 and CB2 antagonists on formalin-evoked pain behaviour (Fig. 4). Blockade of CB1 receptors with SR141716A produced significant hyperalgesia (Fig. 4a) 8 . This effect was particularly pronounced after local injection of the drug, which resulted in a 10-min prolongation of the early nociceptive phase and in a two-to threefold increase in pain behaviour during the entire testing period (Fig. 4b). In contrast, systemic administration of the CB2 antagonist SR144528 caused a selective enhancement of early-phase, but not of latephase, responses (Fig. 4a). The selectivity of this effect may not result from a rapid elimination of SR144528 after early phase, as the drug reversed PEA-induced antinociception during both early-and late-phase pain behaviour (Fig. 2). SR144528 could not be administered locally because of its limited solubility in the injection vehicle.
Although the hyperalgesic effects of CB1 and CB2 antagonists may be accounted for by their inverse agonist properties 16,26 , two lines of evidence suggest that these drugs acted by removing an endogenous cannabinoid tone. First, gas-chromatography/massspectrometry analyses showed that anandamide and PEA are present in  rat paw skin (Fig. 5). By comparison with internal deuterated standards, we measured 49 Ϯ 9 pmol of anandamide and 692 Ϯ 119 pmol of PEA per g of tissue (n ¼ 8). These amounts are five-to tenfold higher than those measured by the same method in rat brain and plasma 27,28 , and are probably sufficient to activate cannabinoid receptors 7,23 . Second, the CB2 antagonist SR144528 enhanced nociception selectively during the early phase of the formalin response (Fig. 4a), a result inconsistent with an inverse agonist effect. Thus, a parsimonious interpretation of our findings is that endogenous PEA acting at CB2-like receptors may participate in attenuating the early stages of nociception, whereas endogenous anandamide acting at CB1-like receptors may have a more sustained modulatory effect on both acute and tonic pain.
Our results show that the tonic activation of local CB1-like and CB2-like receptors may regulate pain initiation in cutaneous tissue. These findings s upport t he p ossibility t hat e ndogenous cannabinoids, in addition to their spinal and supraspinal sites of action 8 , may participate in buffering emerging pain signals at sites of tissue injury. The results also indicate that selective agonists for the CB2-like receptor activated by PEA, or peripherally administered CB1-like/CB2-like agonists, may reduce pain without the dysphoric side effects and perceived abuse potential typical of centrally acting cannabimimetic or opiate drugs. Note added in proof: After submission of this paper, Jaggar et al. 29 have observed an analgesic effect of systemically administered PEA in formalin-induced pain. Synthesis Program of the NIMH; all other drugs were from Tocris. The drugs were dissolved in dimethylsulphoxide (DMSO) and administered in physiological saline containing 10% DMSO; volumes injected were 0.2 ml i.v. and i.p. and 0.01 ml i.pl. Nociceptive tests. Saline (10 l) containing 5% formalin and 10% DMSO was injected subcutaneously into the hind paws of male Swiss mice (20-25 g in weight; Nossan). The duration of paw licking was monitored by an observer blind to the experimental treatment for periods of 0-15 min (early phase) and 15-30 min (late phase) after formalin administration. Initial tests determined that the presence of 10% DMSO in the injection vehicle did not significantly affect formalin responses (data not shown). For the experiment in Fig. 4, pain behaviour was monitored for 5-min periods during the 30 min following formalin injection. Time delays to escape jumping or hind-paw licking were measured on a plate heated to 55.5 ЊC, according to standard procedures 18 . Inflammation. I nflammatory oe demas we re pr oduced in th e hi nd pa ws of Swiss mice by injection of saline (10 l) containing 5% formalin and 10% DMSO, and were measured with a plethysmometer 22 (Ugo Basile). Gas chromatography/mass spectrometry (GC/MS). Hairless paw skin tissue was excised from Wistar rats anaesthetized with nembutal, and immediately immersed in cold methanol to prevent alterations in tissue lipids. After homogenization and chloroform/methanol extraction, highperformance liquid chromatography and quantitative analysis of the trimethyl silyl derivatives of anandamide and PEA by isotope dilution GC/MS were carried out as described 28 . Data analysis. Results are expressed as means Ϯ s:e:m: The significance of differences between groups was evaluated using analysis of variance with a subsequent Dunnett's test.  Substantial changes in the ear's anatomy and in gross cochlear potentials occur around the onset and subsequent rapid maturation of auditory function 7,[9][10][11] . Very little is known, by contrast, about how the membrane currents of the inner and outer hair cells (IHCs and OHCs) develop. We have studied, under whole-cell voltage clamp, time-and voltage-dependent conductances of IHCs, the cells responsible for signalling the reception of sound to the brain, as a function of development between postnatal day (P) 0 and 68.

Excitable cells use ion channels to tailor their biophysical proper
Before P11-P12, the outward membrane currents of IHCs rise slowly and are fairly small (Fig. 1a). By P13, the outward currents of all cells have acquired an additional fast component (Fig. 1b) which becomes larger over the next few days, as shown for a P19 cell in Fig.  1c. The current-voltage curves of Fig. 1d (filled symbols) show the appearance and the increase in magnitude of the fast component of the outward current that occurs between P11 and P19; this fast component is largely responsible for the increase in total current (open symbols).
The outward currents recorded in mouse IHCs from a few days after the onset of hearing closely resemble the membrane currents described before for mature guinea-pig IHCs, which also have a fast and a slow component equated with two different potassium currents, I K,f (probably carried by BK channels 1,12,13 ) and I K,s (a delayed rectifier), r espectively 14,15 . T he f ast c omponent o f the current in mouse IHCs is reversibly suppressed by tetraethylammonium (TEA; half-blocking concentration (IC 50 ) estimated from the dose-response curve as 0:30 Ϯ 0:03 mM; n ¼ 5), to which the slow component is comparatively resistant (IC 50 ¼ 31 Ϯ 8 mM, n ¼ 4; measured between P17 and P20). The fast component is also blocked by charybdotoxin with an IC 50 of 21 Ϯ 2 nM (n ¼ 13) and by the apparently selective BK-channel-blocker iberiotoxin 16 (IC 50 Ͻ 10 nM; n ¼ 3). Taken together, this evidence identifies the fast component in mouse IHCs with I K,f . The slow component of the mouse K + currents is like I K,s in cells studied after the onset of hearing, but the slow current observed at earlier times, in neonatal cells, appears different 15 : it inactivates (slowly) and is considerably more sensitive to block by TEA (IC 50 4:6 Ϯ 0:9 mM, n ¼ 6; P4-P11); we therefore call it I K,neo . The development over time of the different currents is shown in Fig. 2. The total current increases sevenfold between P12 and P16 (Fig. 2a), coincident with the onset and rapid maturation of hearing. Figure 2b shows that this increase is not simply due to the increase in cell capacitance (or membrane area), which is less than twofold, and has a different time course. The increase in total current is mostly contributed by the appearance of I K,f (Fig. 2c): there is an increase in the slow current as well, but it is more gradual (Fig. 2d). In the maturing cochlea, develop-