Trick Or Treat from Food Endocannabinoids?

scientific correspondence

3. Casselman, J. M. in Proc. 1980 North Am. Eel Conf. (ed. Loftus, NAEs (0.01–5.8 Ȗg per g) and oleamide of magnitude below those required, if K. H.) 74–82 (Ontario Ministry of Natural Resources, Ontario, (0.17–6.0 g per g), but no or very little administered by mouth, to reach the blood 1982). Ȗ 4. Radtke, R. L. Comp. Biochem. Physiol. A 92, 189–193 (1989). anandamide and no 2-AG. NAE levels are and cause observable ‘central’ effects. The 5. Kalish, J. M. J. Exp. Mar. Biol. Ecol. 132, 151–178 (1989). much lower in unfermented cocoa beans assays used here provide a gross evaluation of 6. Secor, D. H. Fish. Bull. US 90, 798–806 (1992). than in cocoa powder (which contained less cannabimimetic activity, and tests 7. Tzeng, W. N., Severin, K. P. & Wickström, H. Mar. Ecol. Prog. Ser. 149, 73–81 (1997). than 0.003 Ȗg per g anandamide). Tiny monitoring more subtle behavioural changes 8. Angino, E. E., Billings, G. K. & Anderson, N. Chem. Geol. 1, amounts of anandamide in cocoa could that might be induced by low oral doses of 145–153 (1966). therefore be explained as artefacts of pro- NAEs/oleamide are needed before the rele- 9. Nakai, I., Iwata, R. & Tsukamoto, K. Spectrochim. Acta B (in the cessing2. Like all higher plants, cocoa plants vance of these compounds to the purported press). 8 10. Otake, T., Ishii, T., Nakahara, M. & Nakamura, R. Mar. Ecol. cannot synthesize arachidonic acid or its mild rewarding and craving-inducing effects 7 Prog. Ser. 112, 189–193 (1994). derivatives . Notably, in their NAE and of cocoa can be dismissed. The presence in oleamide content, cocoa and dark chocolate milk of 2-AG (which may be released from are similar to other plant-derived foods triglycerides during digestion) and oleamide (including soybean, hazelnuts, oatmeal and also needs to be investigated. Trick or treat from food millet), in which we detected up to 2.3, 1.1 Vincenzo Di Marzo*, Nunzio Sepe*, and 2.8 Ȗg per g oleamide, N-oleoyl- Luciano De Petrocellis†, Alvin Berger‡, endocannabinoids? ethanolamine and N-linoleoyl-ethanol- Gayle Crozier‡, Ester Fride§, amine, respectively. Raphael Mechoulam§ The discovery of the endogenous cannabi- To establish whether endocannabinoids *Istituto per la Chimica di Molecole di Interesse noid N-arachidonoylethanolamine (anan- survive their passage through the digestive Biologico and †Istituto di Cibernetica, damide)1 and other N-acylethanolamines system, we assayed anandamide and 2-AG Consiglio Nazionale delle Ricerche, (NAEs) in chocolate2 has led to speculation after oral administration in a series of in Via Toiano 6, 80072 Arco Felice, Naples, Italy that the purported rewarding properties of vivo tests that are used widely to assess e-mail: [email protected] cocoa are due to the presence of cannabimimetic activity8 (Table 1). The two ‡Nestlé Research Centre, compounds “that could act as cannabinoid compounds were active in four of the five Nestec Ltd, Vers-Chez-Les Blanc, mimics”2. This observation raises some behavioural tests, but only at very high con- 1000 Lausanne 26, Switzerland important questions. First, are NAEs and centrations relative to those in foods. In §Department of Natural Products, anandamide, or the ‘endocannabinoid’ contrast, a smaller dose of the psychoactive David Bloom Centre for Pharmacy, 2-arachidonoylglycerol (2-AG)3, present in component of marijuana, (ǁ)-ǵ9-tetra- Faculty of Medicine, Hebrew University, widely consumed foods (such as milk) that hydrocannabinol (ǵ9-THC), was sufficient Jerusalem 91120, Israel are less ‘rewarding’ than chocolate? And to give a strong effect in all tests. 1. Devane, W. A. et al. Science 258, 1946–1949 (1992). second, to what extent do these compounds Because intraperitoneal anandamide 2. di Tomaso, E., Beltramo, M. & Piomelli, D. Nature 382, reach the bloodstream and exert pharmaco- and 2-AG are active in the same tests at 677–678 (1996). logical effects when consumed orally? We doses 20- to 60-fold lower3,9, these results 3. Mechoulam, R. et al. Biochem. Pharmacol. 50, 83–90 (1995). 4. Cravatt, B.F. et al. Science 268, 1506–1509 (1995). believe that the content of endocannabi- indicate that only 1.6–5% of the orally 5. Mechoulam, R. et al. Nature 389, 25–26 (1997). noids in foods, and in cocoa in particular, is administered compounds enter the blood- 6. Bisogno, T. et al. Eur. J. Biochem. 254, 634–642 (1998). not sufficient to produce cannabis-like stream, probably owing to the high levels in 7. Cook, H. W. Biochemistry of Lipids, Lipoproteins and Membranes effects in mammals. the gastrointestinal tract of the enzyme fatty 141–169 (Elsevier, Amsterdam, 1991). 10 8. Martin, B. R. et al. Pharmacol. Biochem. Behav. 40, 471–478 (1991). We purified NAEs and 2-AG from acid amide hydrolase , which catalyses the 9. Fride, E. & Mechoulam, R. Eur. J. Pharmacol. 231, 313–314 (1993). various foods (including mature human, hydrolysis of both compounds. Therefore, it 10. Katayama, K. et al. Biochim. Biophys. Acta 1347, 212–218 (1997). bovine and goat milk, and cocoa at various is unlikely that the amounts of anandamide stages of processing) and quantified them and 2-AG found in food are sufficient to Beltramo and Piomelli reply — The two by gas chromatography and mass spectrom- produce observable psychotropic effects. most abundant N-acylethanolamines etry. Milk was the only food analysed in We did not test other unsaturated NAEs (NAEs) found in chocolate (N-oleyl- which an endocannabinoid, 2-AG, was that have been suggested to enhance ethanolamine and N-linoleylethanolamine) found in relatively high concentrations endogenous anandamide activity by do not activate brain cannabinoid receptors (0.33Ǆ0.11 Ȗg mlǁ1 in human milk, for inhibiting its inactivation2, but instead but effectively inhibit anandamide degrada- example). The level of NAEs varied between assayed oleamide — for which this kind of tion1,2. This observation suggested to us that 0.003 and 0.024 Ȗg mlǁ1, with anandamide action is also supported by pharmacological these compounds might contribute to the being the least abundant compound. data5,6. High oral doses are again necessary hedonic properties of chocolate by causing Oleamide, a sleep-inducing substance4 that for activity to be observed in vivo (Table 1). non-metabolized anandamide to accumu- inhibits the hydrolysis of anandamide5,6, Our results show that the amounts of late at its sites of action2. was also present (0.055–1.27 Ȗg mlǁ1). anandamide, 2-AG and oleamide in foods, How can this hypothesis be tested? One In cocoa-derived samples, we detected including milk and cocoa, are several orders possibility is to determine whether antago- Table 1 Effects of test substances on performance of behavioural tests in mice nists of cannabinoid receptors prevent or Olive oil Anandamide 2-AG ǵ9-THC Oleamide reduce chocolate craving. However, because (vehicle) (300 mg kgǁ1) (400 mg kgǁ1) (90 mg kgǁ1) (200 mg kgǁ1) cannabinoid antagonists also reduce the Ambulation 77Ǆ11 51Ǆ8* 13Ǆ9** 34Ǆ7** 39Ǆ6** intake of other palatable foods and drinks3–5, Rearing 29Ǆ5 17Ǆ4* 3Ǆ2** 3Ǆ3** 12Ǆ2** the results of this experiment would be diffi- Immobility 22Ǆ7 70Ǆ20* 157Ǆ19** 146Ǆ18** 85Ǆ19* cult to interpret. Another approach would Analgesia 12Ǆ0.4 16Ǆ3 10Ǆ1 25Ǆ2** 15Ǆ3 be to test whether N-oleyl-ethanolamine Hypothermia ǁ0.2Ǆ0.25 ǁ2.30Ǆ0.42** ǁ3.2Ǆ0.5** ǁ4.1Ǆ1.0** ǁ2.8Ǆ0.56** and N-linoleylethanolamine act as reinforc- Effects of orally administered olive oil, anandamide, 2-AG, ǵ9-THC and oleamide on mouse ambulation (number of ing stimuli in behavioural experiments. It is square crossings) and rearing (number of offspring reared) in an open field, immobility (time spent motionless on a ring, in seconds), analgesia (response latency to hind paw lick on a hot plate (54 °C), in seconds), and hypothermia possible, for example, to determine in ani- (decrease in rectal temperature, in °C)6,7,9,10. The procedures are not harmful to mice and meet with NIH ethical mals whether these compounds possess dis- standards. Substances were administered to Sabra female mice by mouth by gavaging. Doses lower than those shown did not produce any statistically significant effect. Results are meanǄs.e.m. from 4–9 mice. *, P<0.01; criminative properties, a measure of their **, P<0.005; assessed by analysis of variance followed by Newman–Keuls post-hoc tests for group differences. ‘subjective’ effects. In either case, the aim is 636 Nature © Macmillan Publishers Ltd 1998 NATURE | VOL 396 | 17 DECEMBER 1998 | www.nature.com scientific correspondence to ascertain what component of the psycho- yellow exit. When the entrance was lemon pharmacological effects of chocolate, if any, scented, they preferred the blue exit (Fig. is mediated by an indirect activation of the Honeybees link sights 1b, experiment 1). The scent of mango evi- endogenous cannabinoid system. dently evoked recall of yellow, and lemon Di Marzo and co-workers have tested to smells triggered recall of blue. Bees could also be the role of endogenous cannabinoids in trained to make the opposite associations: chocolate craving by investigating the It is common for a smell or a sound to trig- yellow with lemon, and blue with mango extent to which NAEs and anandamide ger a vivid recollection of an associated event (Fig. 1b, experiment 2). reach the bloodstream and exert their phar- in the past, even if it involves a different sen- Bees trained on this task sometimes hes- macological effects after oral administra- sory modality and the episode occurred a itate to choose a colour in chamber B, the8n tion. To test whether anandamide or long time ago1,2. The human brain displays return to the entrance of chamber A and oleamide (which is not an NAE, but inhibits impressive cross-modal associative recall. hover in front of the scented vial with the enzymatic hydrolysis of anandamide) Here we investigate whether this capacity extended antennae, as if to sample the scent elicit overt cannabis-like effects in mice, extends to insects. We find that trained once more, before returning to chamber B they gave anandamide or oleamide orally in honeybees can recall a specific colour when to make their choice. We also found the quantities similar to those found in choco- they encounter a particular scent. reverse phenomenon, that bees can be late, and assessed their psychotropic effects To investigate whether recall of a colour trained to recall a specific scent when they by a standard behavioural procedure that is can be triggered by exposure to a scent, the see a particular colour (data not shown). used in the screening of cannabinoid-recep- two kinds of stimulus should be presented ‘Associative learning’ describes the tor ligands. From their results, they con- sequentially. Furthermore, the colour process by which an organism learns to clude that the content of NAEs and other should not be present when the scent is associate a particular sensory stimulus with cannabinoid-related compounds in cocoa is encountered, and vice versa. We met these a reward or a punishment3–9. The associa- insufficient to elicit cannabis-like effects. requirements by using the apparatus shown tive recall that we describe here, however, is These results lend themselves to two in Fig. 1a. As the bees entered the first part more elaborate than this. It involves both considerations. First, it seems illogical to of the apparatus, chamber A, they received the grouping together of reward and sig- presume that compounds present in an olfactory stimulus from a vial placed at nalling sensory stimuli of different modali- chocolate should display a pharmacological the entrance. They then had to fly to cham- ties in memory, and the recall of one profile similar to that of cannabis. This ber B, which has two exits, one marked with member of a grouped pair when the other would be the same as assuming that cocoa the colour blue and the other with yellow. member is encountered. and cannabis have comparable psychoactive The bees had to choose the yellow exit if For a foraging honeybee, cross-modal effects, which is evidently not the case. It they had encountered the scent of mango at associative recall can facilitate the search for would have been more informative to com- the entrance to chamber A, and the blue a food source; for example, detecting the pare anandamide and oleamide to cocoa in exit if they had encountered the smell of scent of lavender can initiate a search for a standard drug discrimination test. Sec- lemon. The blue and yellow labels in cham- purple flowers. Indeed, one might surmise ond, the effects of N-oleylethanolamine and ber B were interchanged every 10 minutes, that the nectar samples received by a poten- N-linoleylethanolamine were not investi- and the reward moved with the appropriate tial foraging recruit from a dancing bee may gated: not only are these NAEs present in colour, to make sure the bees found the stimulate the recruit into recalling many of chocolate in greater amounts than anan- reward by associating each scent with the the host flower’s other attributes, such as its damide, but they are also produced in neu- appropriate colour, rather than with a par- colour, shape and texture, and perhaps even rons through an activity-dependent ticular exit in chamber B (left or right). the route to the flower patch if the recruit mechanism that is similar to the one After one day’s training, the bees were has previously visited it. implicated in anandamide formation6. The tested in the apparatus one at a time. When In nature, the smell and appearance of possibility that these compounds act syner- the bees encountered the scent of mango at flowers are often (but not always) detected gistically to prevent anandamide degrada- the entrance to chamber A, they showed a at the same time. In our experiments we tion in vivo therefore remains to be strong and significant preference for the presented the olfactory and visual stimuli investigated. In conclusion, although the results of Di b Scent-to-colour P < 0.001 P < 0.001 a n = 80 P < 0.001 association n = 81 n P < 0.001 Marzo et al.’s study will reassure manufac- = 87 n =76 turing companies that the risks of chocolate Feeder consumption do not include cannabis-like 100 ) e c

% 80 intoxication, they provide little new Colour 1 Colour 2 i (

B o y information on the intriguing psychophar- h c 60 Baffle c

n r macology of cocoa. This substance remains, e Fan Fan u

u 40 o l in R. J. Huxtable’s apt words, “more than a q o e r f food but less than a drug”7. A C 20 Massimiliano Beltramo, Daniele Piomelli 0 Lemon Yellow Department of Pharmacology, 360 Med Surge II, Entrance Scent Mango University of California at Irvine, Mango Blue 25 cm Experiment 1 Lemon Irvine, California 92697-4625, USA Experiment 2 Scent at entrance 1. Desarnaud, F., Cadas, H. & Piomelli, D. J. Biol. Chem. 270, 6030–6035 (1995). Figure 1 Cross-modal associative recall in honeybees. a, Experimental set-up. Bees had to choose the 2. di Tomaso, E., Beltramo, M. & Piomelli, D. Nature 382, appropriate colour in chamber B (blue or yellow) according to the scent (lemon or mango) they experienced 677–678 (1996). 3. Arnone, M. et al. Psychopharmacol. 132, 104–106 (1997). from a vial at the entrance to chamber A. The fans created a slight positive pressure in chamber A, ensur- 4. Colombo, G. et al. Life Sci. 63, PL113–PL117 (1998). ing that the bees encountered the scent only as they entered that chamber, and not in chamber B. The baf- 5. McGregor, I. et al. Proc. Int. Cannabinoid Res. Soc. Meet. 76 fle ensured that the bees did not see the colours until they entered chamber B. b, Results of tests; n (1997). 10 6. Di Marzo V. et al. Nature 372, 686–691 (1994). indicates the number of choices analysed and P represents the confidence level for choice frequencies 7. Huxtable, R. J. Nature 382, 411–412 (1994). being significantly different from random choice (red line).