Psychedelics & 5-HT2A

5-HT2A is a G protein-coupled receptor, and it is dense in the cortex, the outer sheet of tissue that does the brain's associative work. Its highest concentrations sit on the apical dendrites of layer 5 pyramidal neurons, the large output cells that project to other cortical areas and down to the thalamus ^[1]. Activating them raises their excitability and increases glutamate release onto the local cortical network, which is why a receptor that occupies a small fraction of brain tissue can reorganise perception so thoroughly.

Mechanically, 5-HT2A couples mainly to Gq, which switches on phospholipase C and releases calcium inside the cell. A cryo-electron microscopy structure of the receptor bound to a hallucinogenic agonist together with its Gq protein showed a deep, partly enclosed pocket with a side extension, which is why molecules as different as a flat indole and a bulky ergoline all fit ^[8]. It also explains a durable puzzle about LSD: bulky ligands sit under a lid formed by an extracellular loop and dissociate very slowly, and that slow off-rate matches the drug's extraordinary duration.

Here is the question that makes the receptor interesting rather than trivial. Serotonin itself is a 5-HT2A agonist, and serotonin is not psychedelic. Part of the answer is access; serotonin does not cross from blood into brain, and psychedelics are lipophilic enough to reach receptor pools that serotonin cannot. The rest of the answer appears to be signalling bias. In mouse cortex, drugs that are psychedelic in humans switch on a distinct 5-HT2A-linked signalling signature and a specific set of genes, while lisuride, which binds the same receptor with high affinity, does not, in the same cells at the same receptor ^[7]. Two agonists at one receptor produced two different downstream programmes.

The negative evidence is the strongest part of the case. Give a volunteer a 5-HT2A antagonist first and psilocybin becomes close to inert ^[4]; do the same before LSD and the sense of personal meaning that defines the state does not appear ^[5]. Blocking one receptor removes the entire phenomenon, which is a far higher bar than showing the drug binds there.

Chemically the classics fall into three groups, and the split is worth learning because it predicts duration, route and side character far better than the word psychedelic does.

Tryptamines are built on the indole ring, which is serotonin's own backbone; substituting position 4 of that ring gives psilocybin and psilocin, and substituting position 5 gives 5-MeO-DMT. Their defining pharmacokinetic fact is that most are destroyed by monoamine oxidase before they reach the brain, which is why DMT does nothing swallowed on its own and everything when swallowed alongside an MAO inhibitor, as in ayahuasca.

Phenethylamines share the backbone of dopamine and amphetamine. Adding two methoxy groups at positions 2 and 5 plus a lipophilic substituent at position 4 produces the 2C series and the longer, slower DOx series ^[9]. They resist monoamine oxidase, so they work orally without help, and they tend to carry more physical stimulation than the tryptamines.

Lysergamides are the ergolines, rigid four-ring systems that lock a tryptamine and a phenethylamine geometry into one molecule. That rigidity buys enormous potency and also broad promiscuity; LSD is active at many serotonin subtypes and at dopamine receptors as well, and separating which contributions matter is still unfinished work ^[10].

Two notes on the table below. Durations are typical ranges for the drug's action rather than a schedule, and they vary widely between people. The potency column is a hazard note rather than a specification; a compound active in micrograms is a compound where a weighing error is a medical emergency.

If 5-HT2A agonism were the whole explanation, every 5-HT2A agonist would be psychedelic. Several are not. Lisuride binds the receptor tightly and produces no psychedelic effect in humans; in cortical tissue it activates a different downstream programme from the one hallucinogens activate at the same receptor ^[7]. That single comparison broke the assumption that occupancy alone determines outcome and opened the field's most active question: which of the pathways a 5-HT2A agonist can recruit is the one that matters.

The question stopped being academic when it became a drug design programme. Psychedelics grow dendritic spines and increase synapse number in cultured cortical neurons and in vivo, and the effect depends on BDNF and mTOR signalling rather than on anything specific to the subjective experience ^[13]. Chemists then built tabernanthalog, an ibogaine-derived analogue that keeps the spine growth and the behavioural effects in rodent models while producing no head-twitch response, the standard rodent proxy for a psychedelic effect ^[14]. Whether an analogue that skips the experience also skips the benefit in a human being is not known; nothing of this class has a completed controlled trial.

It is worth being precise about what the rodent proxy proves. The head-twitch response is 5-HT2A dependent and correlates well with human psychedelic activity, but it is a twitch rather than an experience, and other receptors contribute to the full behavioural picture of these drugs, including 5-HT1A, 5-HT2C and trace amine receptors ^[10].

The honest summary: 5-HT2A agonism is necessary and is not sufficient. Blocking it removes the effect, so nothing else can substitute; but occupying it does not guarantee the effect, so something about how it is occupied carries the rest.

The receptor-level account explains almost nothing about why the experience feels the way it does. For that, the literature works at network scale, and the founding result was a surprise. When psilocybin was given during fMRI, blood flow and BOLD signal decreased in the brain's most connected hub regions, chiefly the medial prefrontal cortex and posterior cingulate cortex, and the coupling between those hubs fell apart ^[11]. Everyone had expected a drug that produces vivid experience to increase activity; it reduced it, and it reduced it selectively in the machinery that ordinarily constrains what the rest of the cortex is allowed to do.

That finding has been extended into a general framework, usually summarised as a loosening of high-level priors: under a psychedelic the brain's top-down models exert less control, sensory and associative regions that normally do not communicate begin to, and the resulting state is more variable and more open to revision ^[12]. This framework is the field's best current organising story, and it should be read as a framework. It is an interpretive model built on top of imaging results, not a measured mechanism, and its authors present it that way ^[12].

Two caveats a careful reader should keep. Psychedelic imaging studies are small, and a drug that causes head motion and changes cerebral blood flow makes fMRI harder than usual to interpret ^[2]. And a change in network organisation during the drug's few hours does not by itself explain a change in mood measured weeks later; connecting the acute state to the durable outcome is the largest open gap between the neuroscience and the clinical work ^[12].

Psilocybin has more controlled human data behind it than any other classic psychedelic, and the record is genuinely encouraging without yet being conclusive. The table sets out the four studies that most people are actually referring to when they cite this literature, along with what each one cannot establish.

The recurring methodological problem is functional unblinding. A high dose of a psychedelic announces itself, so participants and raters both know which arm they are in, and expectancy in this field is unusually strong because volunteers arrive with strong beliefs about the drug. A detailed review of the trial literature argues that blinding failure is close to universal and that effect sizes should be read with that in mind ^[18]. The cleanest natural experiment came from outside formal trials: a self-blinded citizen-science study of microdosing, in which participants prepared their own capsules so nobody knew which weeks were active, found that placebo and microdose groups both improved and did not differ from each other ^[19].

Two further limits belong on the record. The drug is not the whole intervention; every positive trial pairs the dose with hours of preparation and psychological support, and no completed study separates the pharmacology from the therapy. And harms have been collected unevenly. A systematic review of trials with serotonergic psychedelics and MDMA found that most studies recorded adverse events unsystematically and that long-term effects were rarely tracked at all ^[22]. That is not evidence of hidden harm; it is an absence of the evidence that would show harm if it were there.

Start with what is not a risk here. Classic psychedelics do not produce a dependence syndrome, and the reason is built into the receptor: repeated dosing downregulates cortical 5-HT2A receptors, so tolerance appears within a day or two and a second dose soon after the first does very little ^[20]. Tolerance is also cross-tolerance across the whole class, which is a pharmacological fingerprint of the shared mechanism as much as a safety fact.

The main risks are psychological. Acute anxiety, panic and disorganised thinking are common enough that the published safety guidelines for human research are built almost entirely around managing them: screening, preparation, an attended setting, and a person present throughout ^[21]. Those guidelines are also the reason trial safety data does not transfer to unsupervised use. Trials exclude people with a personal or family history of psychotic illness, so the group at highest theoretical risk is systematically absent from the dataset that gets quoted as evidence of safety.

Hallucinogen persisting perception disorder is the one long-term psychological outcome with its own name. It describes visual disturbances that continue after the drug has cleared, sometimes for years. It appears to be uncommon, its true incidence is unknown because it is diagnosed retrospectively from self-report, and no treatment has been established ^[23].

Then the sharp chemical exceptions, which are the ones worth memorising because they are not judgement calls.

NBOMe compounds are not interchangeable with LSD. Adding an N-benzyl group to a 2C drug raises 5-HT2A affinity by more than an order of magnitude ^[24], and the clinical record for the resulting compounds includes seizures, severe hyperthermia, vasoconstriction and deaths ^[25]. They are sold on blotter, which is visually indistinguishable from LSD. The one crude discriminator available is taste; NBOMe compounds are bitter and numb the mouth, and LSD is tasteless.

MAO inhibitors. Irreversible MAOIs such as phenelzine and tranylcypromine, and the reversible MAO-A inhibitors in ayahuasca brews such as harmine, block the enzyme that clears tryptamines. Combining one with a serotonergic psychedelic risks serotonin syndrome, and with the 5-MeO tryptamines the combination has been fatal.

Lithium. This one is poorly known and worth flagging. An analysis of a large body of self-reported psychedelic experiences found that coadministration with lithium was associated with seizures, while lamotrigine was not ^[27]. The data are self-reported and cannot establish a rate, but the signal is specific enough that it should be treated as a hard stop rather than a caution.

Repeated low doses are a different risk question from single high doses. LSD and several tryptamines are agonists at 5-HT2B, and chronic 5-HT2B agonism is the established mechanism behind the valvular heart disease caused by fenfluramine and by pergolide ^[26]. No study has demonstrated valvulopathy from a psychedelic in a human being; the concern is a mechanistic inference from a well-characterised class effect, and it applies to daily or near-daily dosing rather than occasional use. It is the clearest example of a risk that enthusiasm for microdosing has outrun.

This page is educational. For the wider class of substances sold with no human safety data behind them, see research chemicals; for what tolerance and downregulation mean generally, see tolerance and dependence.