Cannabinoid receptors (CB1 & CB2)

The order in which this system was discovered is worth knowing, because each step answered a question the previous one raised. The binding site came first, in 1988: a tritiated synthetic cannabinoid bound rat brain membranes with sub-nanomolar affinity, in a saturable, stereoselective way, and the rank order in which cannabinoids displaced it matched their potency in animals ^[1]. Then came the map, in 1990, when autoradiography showed where that site sits: densest in the outflow nuclei of the basal ganglia, the hippocampus and the cerebellum, and notably sparse in the brainstem areas that run breathing and heart rate. The authors drew the obvious conclusion, that this anatomy is probably why very high doses of THC are not lethal in the way opioids are ^[3]. Then came the gene ^[2], and only then the endogenous ligand, anandamide, isolated from pig brain in 1992 ^[4].

The second endocannabinoid arrived in 1995, and it arrived twice: one group isolated 2-AG from canine gut and showed it binds both receptors, another identified it in brain in the same year ^[6][7]. That double discovery is a useful reminder that the field converged rather than followed a single lab.

The full system has three parts: the receptors, the endocannabinoids that bind them, and the enzymes that build and destroy those messengers ^[21]. What sets it apart from classic neurotransmitters is timing. Dopamine and serotonin are made ahead of time and stored in vesicles, waiting to be released. Endocannabinoids are made on demand from membrane fat the moment a cell needs them, then destroyed within seconds to minutes. That is why a drug like URB597, which blocks the anandamide-clearing enzyme FAAH, raises anandamide indirectly rather than acting on the receptor at all.

The two endocannabinoids are not interchangeable, and treating them as one substance is a common error. 2-AG is present in brain at concentrations perhaps a thousandfold higher than anandamide, behaves as a full agonist, and does most of the retrograde signalling work. Anandamide is scarcer and is a partial agonist, so in some tissues it can even blunt a stronger signal. They are built by different enzymes and cleared by different ones, which means a drug can raise one without touching the other ^[21].

Because it is built into so many tissues, the system acts less like an on/off switch and more like a dimmer and thermostat; it fine-tunes appetite, pain, mood, memory, immune tone and inflammation, generally nudging things back toward a set point rather than driving them hard in one direction ^[10][21]. See the endocannabinoid system for the wider tour.

The cleanest way to hold the two subtypes in your head is by where they live. CB1 is a brain receptor first and foremost; it is one of the most abundant GPCRs in the mammalian central nervous system, densely packed in the cortex, hippocampus, basal ganglia and cerebellum ^[2][3]. It sits mostly on axon terminals, which matters enormously for how it works. CB2 is largely an immune receptor; in healthy tissue it sits mainly on immune cells, and in the brain it appears chiefly on microglia once they become activated ^[5][16].

That split explains the single most useful fact about cannabinoids: hitting CB1 tends to be psychoactive, hitting CB2 tends not to be. A drug that selectively targets CB2 can, in principle, calm inflammation without producing a high ^[10][15]. Both subtypes still share the same core wiring, coupling through Gi/o proteins and turning down adenylyl cyclase ^[10].

The CB1 knockout mouse is the experiment that pins this down, and it contains a surprise. Mice lacking CB1 are viable but hypoactive, hypoalgesic and die younger, and THC's catalepsy, hypothermia and reduced movement vanish completely in them. But THC-induced analgesia in the tail-flick test survives, along with some other responses ^[8]. In other words CB1 carries most of what THC does and demonstrably not all of it, which is the first hint that the two-receptor picture is incomplete.

CB2 has a longer-running problem, described in its own review as an identity crisis. Whether CB2 is genuinely expressed on neurons, and if so where and when, has been claimed, denied and re-claimed for years, largely because the antibodies used to detect it have been unreliable and because its expression is inducible rather than constitutive; it appears when tissue is inflamed and can be absent otherwise ^[16]. This is not a settled anatomy, and any confident statement about neuronal CB2 should be read with that in mind.

Both subtypes are Gi/o-coupled, so activating them inhibits adenylyl cyclase and lowers cAMP ^[2][10]. On neurons, CB1 does something more specific: it closes voltage-gated calcium channels and opens potassium channels, which together make the terminal less likely to release its own neurotransmitter ^[10]. Note what that means. CB1 does not excite or inhibit a cell in the ordinary sense; it turns down whatever that terminal was already releasing, whether the terminal was excitatory or inhibitory.

The signature trick of the system is retrograde signalling. In a normal synapse the message flows forward, from presynaptic terminal to postsynaptic neuron. Endocannabinoids run the other way. When a postsynaptic neuron is strongly active, it synthesises 2-AG or anandamide on the spot, and these lipids drift backward across the synapse to hit CB1 on the terminal that just fired, telling it to ease off ^[9]. It is a built-in feedback brake: the receiving cell tells the sending cell to quiet down. In the hippocampus the classic demonstration is depolarisation-induced suppression of inhibition, where depolarising one pyramidal cell briefly silences the GABA input onto it, and blocking CB1 abolishes the effect ^[9].

That design is why the endocannabinoid system shows up everywhere in learning and synaptic plasticity. A signal that is made where it is needed, acts on the cell that just fired, and is destroyed within seconds is well suited to adjusting individual synapses rather than broadcasting a mood ^[21].

One consequence worth noting: because THC is only a partial agonist at CB1, the size of its effect depends heavily on how many receptors are present and how much endocannabinoid tone is already flowing, so the same dose reads very differently across brain regions and across people ^[14]. Partial agonists are context sensitive in a way full agonists are not.

The receptor was crystallised in 2016, bound to a stabilising antagonist, at 2.8 angstrom resolution. The structure shows a large, mostly hydrophobic binding pocket with an unusual entry route suited to lipids, which is a satisfying physical explanation for why molecules as different as a plant terpenoid, an arachidonic acid derivative and an indole synthetic all fit the same site ^[22].

Almost every cannabinoid question resolves into three properties: which receptor a molecule prefers, how much efficacy it has there, and whether it binds the main site at all. The last one sounds pedantic until you meet cannabidiol, which barely does.

Efficacy is where the real harm lives. THC is a partial agonist at CB1 and CB2, so there is a ceiling on how hard it can drive the receptor no matter how much is consumed. The synthetic cannabinoids sold as herbal blends are full agonists, often with much higher affinity, and several of their metabolites remain active agonists rather than being inactivated the way most of THC's are ^[19]. A full agonist at a receptor this abundant, plus active metabolites, plus no ceiling, is the mechanistic reason those products cause seizures, agitation, kidney injury and deaths that cannabis does not. They are also chemically unrelated to each other and change constantly, so a name like MDMB-4en-PINACA describes one generation of an ongoing series; see research chemicals for why that pattern is itself the hazard.

Cannabidiol is the opposite failure mode: not too strong, but barely binding. CBD has low direct affinity at CB1 and behaves as a negative allosteric modulator there, binding a site away from the main pocket and reducing the potency and the maximum effect of agonists like THC and 2-AG rather than activating anything itself ^[20]. That single fact explains a great deal of the confusing clinical picture around CBD, including why it can blunt some THC effects.

The last honest note on this map is that potency and effect size are different things. THCP binds CB1 far more tightly than THC because of a longer alkyl tail, and higher affinity does change the dose needed; it does not by itself make a molecule more efficacious, and the human data on these newer plant and semi-synthetic cannabinoids is very thin.

CB1 and CB2 are the two receptors everyone agrees on, but cannabinoids clearly touch more than that. The orphan receptor GPR55 is often floated as a putative CB3; it is activated by some cannabinoids and endocannabinoids and, notably, by cannabidiol and abnormal cannabidiol, which have little activity at CB1 or CB2. It couples through Galpha13 rather than Gi/o and remains formally unclassified ^[12]. Cannabinoids also engage the heat-sensing ion channel TRPV1, the nuclear PPAR receptors that regulate metabolism and inflammation, and, in the case of CBD, serotonin receptors as well ^[14][21].

This is the backdrop for the biggest surprise in the field. CBD is promiscuous and, oddly, a weak direct binder at the classic receptors, so much of what it does may run through these other targets rather than through CB1 and CB2 ^[14][20]. It is a good reminder that a cannabinoid is defined by its chemistry, not by hitting one receptor.

The strongest evidence that CBD does something real also happens to be the evidence that its mechanism is not cannabinoid. In a randomised, double-blind trial in Dravet syndrome, CBD reduced monthly convulsive seizures from 12.4 to 5.9 while placebo went from 14.9 to 14.1, with more diarrhoea, somnolence and abnormal liver enzymes in the treated group ^[24]. That is a genuine drug effect with an approved product behind it, and the mechanism responsible for it has still not been established.

Tolerance to cannabis is not vague folklore; it has been measured directly in human brains. Positron emission tomography in chronic daily cannabis smokers found downregulation of cortical CB1 receptors, the degree of it correlating with years of smoking, and it was regionally selective rather than global. After roughly four weeks of continuously monitored abstinence on a research unit, CB1 density returned to normal levels ^[18]. That is an unusually clean answer to a question people usually argue about: the receptor loss is real, it is measurable, and it reverses.

It also sets a realistic expectation for a tolerance break. Weeks, not days, and the recovery is of receptor number rather than of anything mystical. See tolerance and dependence for the general form of this process across drug classes; the cannabis version is milder than the opioid one but follows the same logic of a system adapting to constant stimulation.

The mirror image of that experiment is rimonabant, and it is the most instructive drug failure in this field. Rimonabant is a CB1 inverse agonist, meaning it does not merely block the receptor but pushes it below its resting activity. As a weight-loss drug it worked: pooled trials found about 4.7 kg more weight loss than placebo at one year. It also produced significantly more adverse events, and patients on it were about 2.5 times more likely to stop treatment because of depressed mood, with anxiety adding further discontinuations, in trials that had excluded depressed patients at entry ^[13]. It was withdrawn. The lesson generalises: the resting tone this system carries is doing something for mood, and switching it off has a cost.

One more clinical oddity belongs here because it confuses people badly. Cannabinoid hyperemesis is a cyclical vomiting illness in heavy long-term users, first described in a South Australian case series where the vomiting resolved on stopping cannabis, returned on rechallenge, and was accompanied in nine of ten cases by compulsive hot bathing ^[11]. It is counterintuitive because cannabis is also an antiemetic, and the mechanism remains unresolved.

Almost every cannabis compound can be read off the CB1 and CB2 map. THC is a CB1 and CB2 partial agonist, and its CB1 activity is what makes it psychoactive ^[14]. THCv flips with dose, acting as a CB1 antagonist at low doses and an agonist at higher ones while working as a CB2 partial agonist ^[14]. CBN, the oxidation product of aged THC, leans CB2. And CBD barely activates either directly, instead tuning the whole system from the side ^[20].

The plant is not the only route in. Beta-caryophyllene, a peppery terpene in black pepper, cloves and cannabis itself, is a selective CB2 agonist with a binding affinity around 155 nM; it is a genuinely dietary cannabinoid, and its anti-inflammatory effect in mice disappears in animals lacking CB2, which is the control that makes the claim credible ^[15]. You can also raise your own cannabinoids without any external agonist: URB597 blocks FAAH so anandamide lingers, amplifying tone the system was already generating.

Finally, keep the naming honest. Palmitoylethanolamide and oleoylethanolamide are built by the same enzymatic route as anandamide and are constantly sold as cannabinoids, but they are not cannabinoid receptor agonists; they act mainly at PPAR-alpha. Understanding the two subtypes is what lets you predict, rather than memorise, how each of these behaves. For the wider signalling picture see neurotransmitters 101; the endocannabinoids are the clearest example of a messenger that runs the synapse in reverse.

Whether CB2 is expressed on neurons. The claim has been made, refuted and remade for over a decade, hampered by unreliable antibodies and by the fact that CB2 expression is inducible; the answer matters commercially, because a CB2-selective analgesic is only attractive if CB2 is where the reviews say it is ^[16].

Whether raising your own endocannabinoids is therapeutic. The idea is elegant and the human record is poor. A FAAH inhibitor that reduced enzyme activity by more than 96 percent and clearly raised its substrates produced no analgesia at all in knee osteoarthritis, stopping at interim analysis for futility, a plain disconnect from the animal data ^[17]. A different, non-selective FAAH inhibitor caused an acute and severe neurological syndrome in a phase 1 trial, leaving one participant brain dead and two with lasting deficits; that compound's off-target profile is the leading suspect rather than FAAH inhibition itself, and the mechanism was never established ^[23].

How CBD works when it works. The Dravet syndrome result is real and reproducible, and no proposed mechanism, CB1 modulation, GPR55, TRPV1, adenosine reuptake or serotonin receptors, has been shown to be the one responsible ^[20][24].

The entourage effect. The idea that terpenes and minor cannabinoids meaningfully shape the effect of THC is popular and mostly untested; beta-caryophyllene is the one clear case of a terpene with a real cannabinoid receptor action, and it was demonstrated with a knockout control ^[15]. Treat broader entourage claims as unproven rather than disproven.

What downregulation costs. CB1 receptors recover after about four weeks of abstinence ^[18], but whether the period of reduced density explains any of the cognitive findings reported in heavy long-term users, and whether those findings themselves fully reverse, is not established.