Acetylcholine receptors (nicotinic & muscarinic)

Start with the housekeeping, because the pharmacology hangs off it. Acetylcholine is built in a single step: choline acetyltransferase joins choline to an acetyl group borrowed from acetyl-CoA, and a vesicular transporter packs the product for release. Clearance is the unusual part. Most transmitters are pumped back into the terminal intact; ACh is destroyed where it lands, by acetylcholinesterase, one of the fastest enzymes in biology. The choline half is then recaptured by a high-affinity transporter and used again. That recycling loop, not fresh synthesis, is normally rate limiting, and it is the honest reason choline supply is discussed at all ^[18].

The cholinergic neurons fall into a few tidy populations. Every motor neuron driving skeletal muscle is cholinergic, as is the entire preganglionic autonomic outflow and the whole postganglionic parasympathetic branch, which is why antimuscarinic side effects read as a checklist of dry mouth, blurred near vision, constipation and a fast pulse. Inside the brain there are three sources: the basal forebrain projecting to cortex and hippocampus, the brainstem pedunculopontine and laterodorsal tegmental nuclei projecting to thalamus and midbrain, and cholinergic interneurons that live inside the striatum rather than projecting into it ^[1].

Functionally, ACh is not a go signal or a stop signal. It is a state-setter: it raises the weight given to incoming sensory input relative to internally generated feedback, supports theta-frequency rhythms, and biases cortical circuits towards encoding something new rather than replaying something old ^[2]. Release appears to run on two timescales at once, a slow tonic level tracking arousal and fast transients lasting seconds that line up with detecting a cue worth acting on ^[1][3].

The clinical anchor is loss. Basal forebrain cholinergic neurons degenerate early in Alzheimer's disease and the extent of that loss tracks cognitive decline, which is the core of the cholinergic hypothesis ^[18][19]. The oldest direct human demonstration is still the cleanest: healthy young volunteers given scopolamine produced a memory profile resembling normal ageing, and physostigmine partly reversed it ^[20].

One last piece of orientation. The two families are not two flavours of one protein. Nicotinic receptors belong to the Cys-loop pentamer superfamily, the same architecture as GABA-A, glycine and 5-HT3 receptors ^[4]; muscarinic receptors are class A GPCRs, structural cousins of the adrenergic and dopamine receptors ^[6][7]. Evolution wired two unrelated machines to one messenger, and nothing about a drug's behaviour at one family predicts its behaviour at the other.

Here is the whole system in two rows. Read the machinery column first; it sets everything else, because a channel and a GPCR cannot be fast and slow in the same way, cannot be desensitised in the same way, and cannot be drugged in the same way ^[4][6].

Notice what is missing from the ligand column: almost nothing on it is subtype selective. That is the defining practical fact about this system. Nicotine hits alpha4beta2 hardest but is not clean; atropine blocks all five muscarinic receptors more or less equally; a cholinesterase inhibitor raises transmitter at every receptor in the body at once. Subtype-selective cholinergic pharmacology is recent, difficult, and mostly allosteric ^[8].

A nicotinic receptor is a committee. Nine alpha subunits (alpha2 to alpha10) and three beta subunits (beta2 to beta4) can assemble in many combinations, but the brain leans hard on two of them ^[5]. Alpha7 is a homopentamer: five identical subunits, five potential binding sites, no beta subunit at all. Alpha4beta2 is a heteropentamer with two ACh sites at the alpha/beta seams, and it comes in two stoichiometries with different sensitivities, which is one reason its pharmacology looks inconsistent across papers ^[4][5].

The two behave almost oppositely. Alpha4beta2 binds nicotine with nanomolar affinity, carries mostly sodium, and is the receptor that a smoker's plasma nicotine level is actually sitting on; it is the target of Varenicline and Cytisine, both partial agonists chosen precisely because a partial agonist occupies the site while delivering less than a full response ^[5][14]. Alpha7 binds ACh weakly, is strikingly permeable to calcium, is blocked by alpha-bungarotoxin, and desensitises within milliseconds ^[4][15].

Desensitisation is the concept most often missed, and it is not blockade. A desensitised receptor has agonist bound and its channel shut. Flooding the system with agonist therefore does not produce a bigger effect; past a point it produces a smaller one, because the population settles into the desensitised state. This is why nearly every alpha7 drug programme chose a partial agonist or a positive allosteric modulator rather than a full agonist, and why continuously present nicotine from a patch feels nothing like the pulses from smoking ^[14][15].

Chronic exposure then does something that looks backwards. Sustained nicotine increases the number of high-affinity binding sites rather than reducing it; mice infused chronically showed exactly that, with behavioural tolerance developing alongside the increase rather than against it ^[12]. The mechanism is mostly not transcriptional. Nicotine acts as a pharmacological chaperone inside the cell, stabilising assembled pentamers and slowing their turnover so more receptors reach the surface ^[13]. Receptor count rises while response falls, and the two are compatible because much of the enlarged population sits desensitised. Anyone reasoning about tolerance from receptor number alone gets this system exactly backwards.

Finally, nicotinic receptors are not confined to neurons. The alpha7 subunit on macrophages is required for the vagus nerve to suppress TNF release; knock it out and vagal stimulation stops working. That is the cholinergic anti-inflammatory pathway, and it is why alpha7 turns up in immunology papers with no cognition in them ^[17].

The five muscarinic genes were cloned in the late 1980s, and the sequencing arrived after the pharmacology, which is why the older tissue-based labels and the modern numbers do not always line up in old papers ^[6][7]. The organising rule is simple enough to memorise: odd numbers excite, even numbers inhibit. M1, M3 and M5 couple through Gq/11 to phospholipase C, releasing intracellular calcium and activating protein kinase C. M2 and M4 couple through Gi/o, lowering cAMP, opening GIRK potassium channels and closing calcium channels ^[6][10].

Assigning jobs to individual subtypes became one of the harder problems in receptor pharmacology, because the orthosteric pocket is almost identical across all five. Acetylcholine is a tiny molecule and the site that binds it has barely diverged, so classical agonists and antagonists are broadly non-selective, and most of what is known about which subtype does what came from knockout mice rather than from drugs ^[9]. The way out was to stop competing with acetylcholine and instead bind an allosteric site the five receptors do not share ^[8].

One older exception still appears in methods sections. Green and black mamba venoms contain muscarinic toxins: MT7 is a near-irreversible and genuinely M1-selective antagonist, and MT3 prefers M4. Before allosteric modulators existed these peptides were the only subtype-selective tools available ^[11].

The table below is the practical version. Note how much of the trouble with cholinergic drugs is peripheral: M2 in the heart and M3 in smooth muscle and glands are why a centrally intended drug produces bradycardia, cramping, sweating and blurred vision long before it produces a cognitive effect.

The classic nootropic pairing leans directly on the housekeeping described above. Choline donors such as Alpha-GPC and CDP-Choline supply substrate; they raise plasma choline, and both are broken down further before anything reaches a synapse, so what is delivered is a pool rather than a single defined molecule ^[18]. On their own they modestly enlarge the available supply, which then acts across both receptor families indiscriminately.

Be exact about the evidence, because it is thinner than the marketing. Alpha-GPC has one multicentre randomised placebo-controlled trial in mild to moderate Alzheimer's dementia reporting improvement on cognitive scales ^[23]. CDP-choline's Cochrane review found modest short-term effects on memory and behaviour in older people, no evidence on long-term outcomes, and generally weak trial quality ^[24]. Both populations are impaired; there is essentially no controlled evidence for either in healthy young adults, and the honest position is that the question has not been asked properly rather than that the answer is no.

The racetams are the other half of the pairing and they add no choline at all. Nefiracetam is the clearest illustration: it potentiates nicotinic acetylcholine receptor currents through two distinct intracellular signalling routes rather than by binding the channel as an agonist would ^[25]. That is a cholinergic mechanism with no choline anywhere in it, which is exactly why substrate and receptor-side compounds are not substitutes for each other.

The everyday racetam headache belongs here with a warning label attached. Users commonly report a dull, tension-type headache on a racetam that a dose of Alpha-GPC or CDP-Choline tends to relieve. It fits the picture of driving receptors faster than the supply chain refills; it is also folk pharmacology with no controlled test behind it, and should be read as a pattern rather than a finding.

Acetylcholinesterase inhibitors are the third lever and the most powerful of the three. Huperzine A, Galantamine, Donepezil and Tacrine add nothing to the system; they slow the destruction of what is already being released, so their effect scales with how much release is still happening, and in a system that has stopped releasing they have nothing to work with. The meta-analysis of huperzine A in Alzheimer's disease reports benefit on cognition and daily function while stating plainly that most included trials carried a high risk of bias, which is the whole finding and not a footnote ^[22]. Galantamine is worth separating from the group: it is also a positive allosteric modulator at nicotinic receptors, so it pulls two levers at once rather than being a dose-shifted donepezil ^[15].

Layering a donor, a racetam and an esterase inhibitor stacks effects and stacks side effects. The ceiling is almost always peripheral muscarinic: nausea, cramping, salivation, sweating, a slowed pulse. None of that is a brain effect, and none of it is a sign the stack is working.

Alzheimer's disease is why this system was studied so hard for so long. Basal forebrain cholinergic neurons degenerate early, the loss correlates with cognitive decline, and for years the only symptomatic drugs available were the esterase inhibitors that squeeze more out of what remains ^[18][19]. The Cochrane synthesis puts the size of that benefit at roughly 2.7 points on the ADAS-Cog at six months, consistent across donepezil, galantamine and rivastigmine ^[21]. That is real, it is small, and it is symptomatic; nothing about it slows the disease.

The nicotinic alpha7 receptor became a target in its own right because activating it enhances long-term potentiation, drives calcium signalling and carries the anti-inflammatory role described above ^[15][17]. Results have been stubbornly mixed. GTS-21, a partial alpha7 agonist also known as DMXB-A, went into a phase 2 trial in schizophrenia; it did not separate from placebo on the cognitive battery that was its primary interest, but it did improve negative symptoms, among the least treatable features of the illness ^[16]. That shape of result, a miss on the intended endpoint with a signal somewhere adjacent, has repeated across the alpha7 field.

The muscarinic story is the one that finally landed, and it took twenty-five years. Xanomeline was designed as an M1/M4-preferring agonist for Alzheimer's disease; it improved cognitive and psychotic symptoms, but peripheral cholinergic side effects drove so much dropout that the programme stalled ^[26]. The fix was mechanical rather than chemical: pair it with Trospium chloride, a quaternary antimuscarinic whose permanent positive charge keeps it out of the brain, so it cancels much of the periphery and leaves the central agonism intact. That combination beat placebo on total PANSS score in phase 2 ^[27] and again in phase 3 ^[28], with no dopamine receptor blockade anywhere in the mechanism ^[29]; the first genuinely new antipsychotic mechanism in decades, out of this receptor family.

At the other extreme sit the deliriants. Because the muscarinic arm is required for ordinary cognition, blocking it hard does not produce a pleasant altered state; it produces confusion, dense amnesia, and true hallucinations that the person cannot tell from the room. Trihexyphenidyl at proper doses treats Parkinsonism and dystonia; well above them it is a deliriant, alongside Scopolamine and plants such as Datura. The point is not that the dose is high. The point is that switching off these receptors removes the machinery that keeps thought coherent.

The quieter version of the same risk is anticholinergic burden, and it applies to ordinary prescriptions rather than to misuse. A nested case-control study of 58,769 people with dementia and 225,574 matched controls found risk rising with cumulative exposure to strongly anticholinergic drugs, reaching an adjusted odds ratio of 1.49 in the highest exposure band ^[30]. It is observational, so confounding by indication is a live objection: some of these drugs are prescribed for symptoms that precede a dementia diagnosis. It is also the largest dataset on the question, and the direction has been consistent across independent cohorts.

Whether extra choline does anything in a healthy person. Every controlled trial of a choline donor with a positive result was run in an impaired or elderly population ^[23][24]. Extrapolating from a failing cholinergic system to an intact one is exactly the inference the cholinergic hypothesis warns against, since esterase inhibitors also work best where transmission is still partly present ^[21].

Whether the racetam headache is cholinergic at all. Nobody has run a randomised comparison of a racetam with and without a choline source, with headache as the endpoint.

Tonic tone versus fast transients. Whether cortical acetylcholine acts mainly as a diffuse background level or as fast, spatially precise, second-scale releases tied to cue detection is still contested ^[1][3]. This is not a detail: if the signal is transient and precise, globally raising acetylcholine is closer to noise than to enhancement, and the two models predict different things about every drug on this page.

Why alpha7 agonists keep failing. Fast desensitisation, signalling that does not depend on ion flux at all, and species differences between rodent and human receptors have all been offered ^[15][16]. None is established, so each new alpha7 candidate is a fresh bet rather than an iteration.

Which muscarinic subtype carries the antipsychotic effect. M1 and M4 are both mechanistically plausible and the clinical agent activates both ^[8][29]. A subtype-selective successor would answer it; none has completed a trial.

Whether anticholinergic drugs cause dementia or mark people already on the way to it ^[30]. Only a randomised deprescribing trial could separate the two, and none has reported.

What tolerance to a nicotinic agonist actually is. Receptor number rises, responsiveness falls, and desensitisation state, trafficking and circuit adaptation are all in play at once ^[12][13]. No single one accounts for the whole effect, and any explanation resting on receptor count alone has the sign backwards.