Where Cannabinoid Receptors Actually Live: A Working Map of CB1 and CB2 Across the Human Body

Two cannabinoid receptors have been cloned and formally classified: CB1, reported by Matsuda and colleagues in Nature in 1990, and CB2, reported by Munro and colleagues in Nature in 1993. They share roughly 48 percent amino acid sequence identity and both signal through inhibitory G proteins, according to the International Union of Pharmacology classification published in Pharmacological Reviews in 2002.

CB1 is the most abundant G protein-coupled receptor in large parts of the mammalian brain and is also present, at far lower levels, in a long list of peripheral tissues. CB2 is concentrated in immune cells and lymphoid tissue. That division is real, but it is a gradient rather than a wall, and the exceptions are where most of the current scientific argument lives.

Receptor location explains clinical behavior. CB1 density is high in the basal ganglia, hippocampus and cerebellum and low in the brainstem nuclei that run breathing and circulation, which is the anatomical reason a large dose of THC produces disorientation and tachycardia rather than respiratory arrest. Herkenham and colleagues made exactly that point in the 1990 paper that first mapped the receptor.

It also explains the limits. A patient who reads that CB2 receptors regulate inflammation and concludes that a CB2-targeted product will treat their arthritis has skipped several steps, because expression of a receptor in a tissue is not evidence that a drug acting on it changes a clinical outcome.

CB1 came first. L. A. Matsuda and colleagues at the National Institute of Mental Health cloned a complementary DNA encoding a G protein-coupled receptor that inhibited adenylate cyclase in a dose-dependent, stereoselective and pertussis toxin-sensitive manner, and that responded more strongly to psychoactive than to non-psychoactive cannabinoids. The paper appeared in Nature on August 9, 1990.

CB2 came three years later, from a different direction. Sean Munro and colleagues at the MRC Laboratory of Molecular Biology in Cambridge were looking for a receptor that could explain the non-psychoactive effects of cannabinoids. They found one that was not expressed in brain but was expressed in macrophages in the marginal zone of the spleen, and published it in Nature in September 1993.

The 2002 International Union of Pharmacology classification, authored by Allyn Howlett with Raphael Mechoulam, Roger Pertwee, Ken Mackie, Miles Herkenham and others, formalized the two-receptor scheme. That document also made a point worth repeating: the committee considered it premature to rename cannabinoid receptors after an endogenous agonist, because it was unclear whether the known endocannabinoids are the only or even the primary natural ligands. Twenty-four years later that caution still reads as correct.

The foundational anatomy comes from autoradiography, meaning a radiolabeled cannabinoid is applied to tissue sections and the binding is photographed. Herkenham and colleagues used tritiated CP 55,940 across several mammalian species including human and reported a conserved distribution: densest in the outflow nuclei of the basal ganglia, specifically the substantia nigra pars reticulata and globus pallidus, and in hippocampus and cerebellum.

Michelle Glass, Michael Dragunow and Richard Faull at the University of Auckland did the detailed human work, published in Neuroscience in 1997, using quantitative autoradiography on one fetal, two neonatal and eight adult human brains. Their findings are more granular than most summaries admit. Binding was very high in the dentate gyrus, Ammon’s horn and subiculum of the hippocampal formation, high in entorhinal cortex and amygdala, and highest within neocortex in the association areas of the frontal and limbic lobes rather than in primary sensory or motor cortex.

Within cortex the labeling followed the laminar architecture, concentrating in a narrow superficial band corresponding to lamina I and a broader deep band corresponding to laminae V and VI. In the midbrain, substantia nigra pars reticulata carried some of the highest densities in the entire brain. The spinal cord, by contrast, showed very low binding. Fetal and neonatal brains showed the same overall pattern at generally higher density, particularly in basal ganglia.

S. Galiegue and colleagues at Sanofi Recherche in Montpellier quantified CB1 and CB2 transcripts across human tissues and leukocyte subpopulations using a sensitive PCR method, publishing in the European Journal of Biochemistry in 1995. CB1 was mainly a central nervous system transcript but was detectable at lower levels in adrenal gland, heart, lung, prostate, uterus, ovary, testis, bone marrow, thymus and tonsil.

That peripheral presence is the reason cannabinoid effects are not confined to perception. It is also the reason a CB1 blocker turned out to be a systemic drug rather than a brain drug, with consequences discussed further below.

One location deserves separate mention because it changed how the field thinks about the receptor. Giovanni Benard, Giovanni Marsicano and colleagues in Bordeaux reported in Nature Neuroscience in 2012 that CB1 is present on the membranes of mouse neuronal mitochondria, where activation reduced cyclic AMP concentration, protein kinase A activity, complex I enzymatic activity and respiration. A receptor that regulates cellular energy production from inside the organelle is a different kind of object than a receptor that only sits on the plasma membrane, and this work has not been replicated in human tissue.

The same 1995 Galiegue study gives the cleanest quantitative picture of CB2 that exists in human tissue. CB2 was not detected in brain. It was abundant in immune tissue, at an expression level 10 to 100 times higher than that of CB1. In spleen and tonsil, CB2 messenger RNA content was equivalent to the CB1 content of the central nervous system, which is a useful way to calibrate what abundant means here.

Among human blood cell subpopulations the rank order was B cells, then natural killer cells, then a substantial drop to monocytes, polymorphonuclear neutrophils, CD8 T cells and CD4 T cells. The same ordering held in myeloid, monocytic and lymphoid cell lines. Immunohistology on tonsil sections localized CB2 protein to B-lymphocyte-enriched areas of the mantle of secondary lymphoid follicles, not diffusely across the tissue.

Caroline Turcotte and colleagues at Universite Laval reviewed the functional consequences in Cellular and Molecular Life Sciences in 2016. The recurring finding across models is that CB2-deficient mice show an exacerbated inflammatory phenotype, which positions CB2 as a brake on immune activation rather than an on switch. Whether that brake can be pressed therapeutically in humans is a separate question that the animal work does not answer.

The neat division of CB1 in brain and CB2 in immune cells started breaking down in 2005, when Marja Van Sickle, Keith Sharkey and colleagues reported CB2 messenger RNA and protein on brainstem neurons in Science. Hai-Ying Zhang and colleagues at the National Institute on Drug Abuse reported in PNAS in 2014 that CB2 receptors are expressed in ventral tegmental area dopamine neurons and that a CB2 agonist inhibited their firing, with the effects absent in CB2 knockout mice.

Brady Atwood and Ken Mackie titled their 2010 British Journal of Pharmacology review CB2: a cannabinoid receptor with an identity crisis, and the title was earned. The central problem is technical. In 2018 the Zhang group tested four anti-CB2 antibodies against partial knockout mice and concluded that none of them were highly specific for mouse CB2. Y. Li and J. Kim, using ultrasensitive in situ hybridization in mouse hippocampus in 2015, detected CB2 messenger RNA in subsets of excitatory and inhibitory neurons but rarely in microglia, while noting that immunological detection of the protein remained unreliable.

So the honest position is this. CB2 transcripts are detectable in some central neurons. The protein is present at levels low enough that the antibodies available cannot reliably find it. Anyone who tells you confidently that CB2 is or is not a neuronal receptor is ahead of the data. A related open question concerns GPR55, which E. Ryberg and colleagues at AstraZeneca characterized in 2007 as binding and being activated by cannabinoid ligands including anandamide and cannabidiol. It has been proposed as a third cannabinoid receptor, and the international nomenclature committee has not adopted that designation.

Anandamide, chemically arachidonylethanolamide, was isolated from porcine brain by William Devane, Lumir Hanus and Raphael Mechoulam’s group and reported in Science in December 1992. Two independent groups identified 2-arachidonoylglycerol in 1995: Mechoulam’s laboratory found it in canine intestine, and Takayuki Sugiura’s group at Teikyo University identified it in rat brain.

The signaling geometry is unusual and worth understanding, because it explains why cannabinoid drugs behave the way they do. Rachel Wilson and Roger Nicoll showed in Nature in 2001 that endocannabinoids released by depolarized hippocampal pyramidal neurons travel backward across the synapse to CB1 receptors on presynaptic terminals and suppress transmitter release. Endocannabinoids are not stored in vesicles and released on command. They are synthesized on demand from membrane lipids at the site and moment they are needed, then broken down locally by fatty acid amide hydrolase and monoacylglycerol lipase.

That architecture is why the endocannabinoid system behaves like a local dimmer switch rather than a broadcast signal, and why flooding the whole system with an exogenous agonist produces effects that native signaling never would.

The evidence is not uniform across this page, and it helps to grade it. CB1 localization in the human brain is about as solid as neuroanatomy gets: multiple independent methods, replicated across species, with quantitative human data from Auckland covering fetal through adult tissue. Treat the regional CB1 map as settled.

CB2 expression in immune tissue is also well supported, with concordant messenger RNA quantification, Northern blot and immunohistology in the same 1995 study. CB2 in neurons is a different matter entirely: the transcript evidence is reasonable, the protein evidence rests on antibodies that failed specificity testing against knockout tissue in 2018, and the field has not resolved it.

Most of the anatomical work here is rodent, and the human data that exist come from small autopsy series. The Glass study, which remains the best quantitative human map, used eight adult brains. That is enough to establish a pattern and not enough to characterize variation between people, which is precisely the variation a clinician cares about when one patient tolerates a dose that floors another.

Receptor messenger RNA is not receptor protein, and receptor protein is not receptor function. A great deal of published expression data conflates these three, and the CB2 antibody problem is the clearest illustration of how far wrong that can go. Expression studies also tend to be performed in healthy tissue, while CB2 in particular is inducible and rises during inflammation, so a baseline map understates what is available for a drug to act on in a diseased tissue.

Nothing on this page shows that cannabis treats any condition. Receptor anatomy establishes where a drug can act, not whether acting there helps. The step from expression map to clinical benefit requires randomized trials, and for most conditions those trials either have not been done or have not been positive.

This material also does not support product marketing that assigns specific effects to specific receptors. Claims of the form that a product targets CB2 for inflammation without touching CB1 are not supported by the pharmacology of plant cannabinoids, which bind promiscuously and, in the case of cannabidiol, act substantially through targets other than either receptor.

The most instructive episode in this field is rimonabant. A selective CB1 blocker was developed for obesity, worked for weight loss, and was withdrawn. Robin Christensen and colleagues pooled four randomized trials totaling 4,105 participants in The Lancet in 2007: 20 mg daily produced 4.7 kg more weight loss than placebo at one year, and patients were two and a half times more likely to discontinue for depressive mood disorders. That is what happens when a drug engages a receptor expressed across association cortex and limbic structures.

In the other direction, CB1 is not static. Jussi Hirvonen and colleagues at the National Institute of Mental Health used positron emission tomography to show reversible, regionally selective downregulation of cortical CB1 receptors in chronic daily cannabis smokers, with density returning toward normal after roughly four weeks of monitored abstinence, published in Molecular Psychiatry in 2012. Receptor anatomy is a map of a system that adapts to what you put into it.

I keep coming back to the brainstem finding from 1990, because it answers a question patients ask in almost every consultation. The reason a cannabis overdose is frightening rather than fatal is that the nuclei controlling breathing and blood pressure carry very few CB1 receptors. That single anatomical fact does more to calm a nervous new patient than any amount of reassurance, and it has the advantage of being true.

The part I try to be careful about is the opposite direction. Patients hear that receptors are everywhere and conclude that cannabis must therefore do everything. Receptors being present tells you a tissue can respond. It does not tell you the response is beneficial, or that you can steer it with a dispensary product. I would rather a patient leave my office understanding one receptor properly than leave believing a system diagram they cannot use.

CB1 concentrates in basal ganglia outflow nuclei, hippocampus, cerebellum and association cortex, and is sparse in cardiorespiratory brainstem. CB2 concentrates in immune tissue, with B lymphocytes carrying the most. Both receptors appear outside their headline locations, the neuronal CB2 question remains open, and none of this anatomy by itself establishes that any cannabis product treats any condition.

Read this as a map with variable resolution. The CB1 regions are drawn to high confidence, the CB2 immune distribution to good confidence, and the edges of the map, meaning neuronal CB2, mitochondrial CB1 and GPR55, are sketched in pencil. A page that presents all three at the same confidence level is telling you something other than the truth.