News & Views | Published:


A cellular basis for the munchies

Nature volume 519, pages 3840 (05 March 2015) | Download Citation


How does marijuana cause the irresistible hunger pangs known as the munchies? Paradoxically, the answer seems to involve an unusual mode of activation of a brain circuit best known for suppressing appetite. See Article p.45

According to data gathered by the United Nations1, 177 million people around the globe use marijuana. So some of you might be familiar with the munchies — that inexplicable drive to eat, stimulated by the active ingredients of marijuana, the cannabinoids. This connection has already led to the development of dronabinol, a synthetic version of the natural cannabinoid Δ-9-tetrahydrocannabinol, as a treatment for the metabolic disorder cachexia anorexia syndrome. But how, and where in the brain, do cannabinoids work to stimulate food intake? In this issue, Koch et al.2 (page 45) report that, when given in doses meant to simulate the effects of marijuana, cannabinoids surprisingly activate a subset of pro-opiomelanocortin (POMC) neurons, a cell group in the brain's hypothalamus that has a central role in inhibiting hunger.

Previous work3 has shown that ablation of POMC neurons, mutations in the gene encoding the POMC protein, and mutations in the melanocortin 4 receptor (MC4R) in downstream cellular targets of POMC neurons, all cause severe overeating (hyperphagia) and obesity. Conversely, experimental stimulation of the POMC cell group produces a slow-onset inhibition of food intake4,5. Cannabinoids bind to receptors dubbed CB1Rs (for cannabinoid receptor 1), and Koch and colleagues observe that, in mice, these receptors are found on nerve terminals that make synaptic connections not only to POMC neurons, but also on organelles called mitochondria, in POMC neurons themselves. This binding stimulates the specific release of an orexigenic (appetite-stimulating) neuropeptide called β-endorphin from the neurons, while somehow avoiding release of α-melanocyte-stimulating hormone (α-MSH), an appetite-suppressing peptide found in these same neurons (Fig. 1a).

Figure 1: Cannabinoid regulation of feeding circuits.
Figure 1

a, The conventional view of appetite circuits in the brain's hypothalamus involves Agouti-related protein (AgRP) neurons, which stimulate feeding, and pro-opiomelanocortin (POMC) neurons, which inhibit it. Both groups affect downstream neurons expressing melanocortin 4 receptors (MC4Rs). Koch et al.1 propose an alternative mechanism, in which cannabinoids stimulate food intake by causing a subset of POMC neurons to specifically activate β-endorphin-releasing boutons (release sites, not shown), which target downstream neurons expressing μ-opioid receptors. b, The authors suggest that, by activating cannabinoid receptor 1 (CB1R) at two sites, cannabinoids increase feeding behaviour. At one site, they could inhibit release of the neurotransmitter GABA from AgRP neurons onto POMC neurons, thereby enhancing the latter neurons' excitability. At the other site, CB1R activation in mitochondria could increase respiration, production of reactive oxygen species (ROS) and expression of mitochondrial uncoupling protein 2 (UCP2), which then acts as a switch to cause β-endorphin release selectively onto downstream neurons.

That cannabinoids can act at several brain regions to stimulate food intake is well established6,7,8. So the striking lesson here is not so much the orexigenic effect of cannabinoids through yet another of the many brain circuits involved in feeding behaviour. It is rather that cannabinoids can subvert an appetite-inhibitory (anorexic) circuit to become orexigenic, which indicates that the POMC circuit may be even more complex than previously thought. Another challenging concept arising from the present work is that, during the acute orexigenic response, cannabinoids stimulate these neurons partially through intracellular CB1Rs, rather than through the more usually observed actions of cannabinoids at nerve terminals to regulate the release of substances such as the inhibitory neurotransmitter GABA.

It is noteworthy that, at a neuroanatomical level, the POMC circuit is quite complex, with POMC neurons belonging to the arcuate cluster of cells sending axonal projections to more than 100 brain regions. To add further complexity, a subset of these neurons responds to the hormone insulin; a different subset is affected by the hormone leptin9, and roughly half of the cells may undergo inhibitory autoregulation by expressing the receptor MC3R (ref. 10). Moreover, almost all POMC neurons produce not only β-endorphin but also α-MSH.

These neurons may therefore be differentially secreting neuropeptides and neurotransmitters to either suppress or stimulate appetite. Parenthetically, administration of analogues of γ-MSH, another peptide released by POMC neurons, also has orexigenic effects11, suggesting that β-endorphin may not be the only product of POMC neurons that can stimulate food intake. Thus, the emerging picture of the arcuate POMC system is that of a circuit that can sense a wide array of signals and can then produce highly discriminatory responses through a differentiated set of circuits and molecular signalling mechanisms.

What is remarkable about Koch and colleagues' findings is that cannabinoids seem to stimulate β-endorphin release selectively from POMC neurons. Consistent with this, the authors demonstrate that some 34% of the synaptic boutons, or release sites, on POMC neurons selectively express either β-endorphin or α-MSH.

There is, however, a nagging question about the proposed role of β-endorphin as the main mediator of the orexigenic actions of administered cannabinoids in POMC neurons. A previous study12 found that deletion of the portion of the POMC gene encoding β-endorphin produces hyperphagia and obesity rather than leanness, as might be expected if the primary role of the natural peptide is orexigenic. This issue can be readily addressed by testing whether there is a decrease in the orexigenic response to cannabinoid administration in mice carrying mutations in β-endorphin, or perhaps in mice lacking the μ-opioid receptor, the target of β-endorphin in downstream neurons.

The present data certainly make a clear case for the striking ability of cannabinoids to stimulate a small subset of arcuate POMC neurons and subsequently to increase food intake. Nonetheless, the precise mechanisms of CB1R-induced activation of POMC neurons and selective β-endorphin release remain to be fully elucidated.

The authors propose two potential routes by which CB1R activation could boost POMC-neuron activity to increase feeding behaviour (Fig. 1b). First, low doses of CB1R stimulators could increase firing in a subset of POMC neurons, most likely by reducing the release of incoming GABA signals that would otherwise dampen the neurons' activity13. Thus, these stimulators could modify the balance of excitation and inhibition in this neuronal subset. However, a previous study demonstrated14 that CB1R expression on neurons secreting the excitatory signal glutamate, rather than GABA, is required for the hyperphagic response to cannabinoids. Furthermore, increases in the firing rate of POMC neurons as a class reduces food intake4, and thus cannot explain the hyperphagic effects of cannabinoids. So, the possibility of an alternative mechanism to cannabinoid-induced synaptic activation of a specific subset of POMC neurons requires further investigation.

Koch et al. suggest that activation of mitochondrial CB1R represents just such an alternative. CB1R activation has been shown to block respiration in mitochondria by inhibiting the cAMP–PKA signalling pathway15. The authors now extend these findings to show that low levels of CB1R activation in fact increase mitochondrial respiration in POMC neurons as well as in neurons of the brain's hippocampus. They propose a signalling pathway involving CB1R-induced increases in mitochondrial respiration, contact between mitochondria and another cellular organelle called the endoplasmic reticulum, generation of reactive oxygen species and subsequent increased expression of mitochondrial uncoupling protein 2 (UCP2) — a regulator of both mitochondrial respiration in the hypothalamus and feeding (Fig. 1b). They show that UCP2 is essential for cannabinoid effects on mitochondrial respiration, β-endorphin release in the hypothalamus and feeding responses.

More definitive support for this provocative proposed mechanism could be provided by demonstrating that cell-impermeable inhibitors of CB1R do not block feeding induced by CB1R activators. Still unknown are the relative dominance of various CB1R sites in the central nervous system in the orexigenic action of administered cannabinoids, and the relative importance of cell-surface and mitochondrial CB1Rs. Regardless of this, Koch et al. provide another striking example of the complexity of the POMC circuits, and a new cellular mechanism by which cannabinoids stimulate feeding behaviour.



  1. 1.

    United Nations Office on Drugs and Crime. World Drug Report 2014 (UN, 2014).

  2. 2.

    et al. Nature 519, 45–50 (2015).

  3. 3.

    Nature Neurosci. 8, 571–578 (2005).

  4. 4.

    , & Nature Neurosci. 14, 351–355 (2011).

  5. 5.

    et al. J. Neurosci. 33, 3624–3632 (2013).

  6. 6.

    & Br. J. Pharmacol. 134, 1151–1154 (2001).

  7. 7.

    , , & Br. J. Pharmacol. 136, 550–557 (2002).

  8. 8.

    et al. Nature Neurosci. 17, 407–415 (2014).

  9. 9.

    et al. J. Neurosci. 30, 2472–2479 (2010).

  10. 10.

    et al. J. Neurosci. 19, RC26 (1999).

  11. 11.

    , , & Peptides 27, 259–264 (2006).

  12. 12.

    et al. Endocrinology 144, 1753–1760 (2003).

  13. 13.

    & Curr. Opin. Neurobiol. 29, 1–8 (2014).

  14. 14.

    et al. Nature Neurosci. 13, 281–283 (2010).

  15. 15.

    et al. Nature Neurosci. 15, 558–564 (2012).

Download references

Author information


  1. Sachin Patel and Roger D. Cone are at the Vanderbilt University Medical Center, Department of Molecular Physiology and Biophysics, Nashville, Tennessee 37232-0615, USA.

    • Sachin Patel
    •  & Roger D. Cone


  1. Search for Sachin Patel in:

  2. Search for Roger D. Cone in:

Corresponding author

Correspondence to Roger D. Cone.

About this article

Publication history



Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing