GABA receptors (GABA-A & GABA-B)

The two receptor families divide the work of inhibition by speed. GABA-A gives fast inhibition: it is an ion channel, so binding GABA physically opens a pore and chloride moves across the membrane in milliseconds, producing the crisp, moment-to-moment braking that shapes every thought and movement ^[1][2]. GABA-B gives slow inhibition: as a GPCR it triggers an internal cascade that opens potassium channels and closes calcium channels, an effect that builds over hundreds of milliseconds and lingers far longer than a synaptic event ^[22][24].

Inside the GABA-A family there is a second speed distinction that matters more than most textbooks admit. Receptors sitting in the synapse see brief, high concentration pulses of GABA and produce phasic inhibition, one clean inhibitory post-synaptic potential per release event. Receptors sitting outside the synapse see the low ambient GABA that escapes the cleft, and because they have higher affinity and desensitise slowly, they carry a standing tonic current that never switches off ^[3]. Tonic inhibition sets the baseline excitability of a cell rather than its individual responses, and it is disproportionately the target of neurosteroids, of some anaesthetics and possibly of alcohol ^[3][16].

One detail explains a great deal of confusing pharmacology: chloride does not always flow inward. GABA-A receptors do not hyperpolarise cells by magic, they open a channel and chloride follows its own gradient. In a mature neuron the KCC2 transporter keeps internal chloride low, so opening the channel lets chloride in and the cell hyperpolarises. In immature neurons, and in some injured or epileptic tissue, NKCC1 dominates and internal chloride is high, so the same receptor depolarises the cell. Even when the direction is wrong, opening a large conductance still shunts incoming current and damps the neuron, which is why GABA-A drugs remain broadly calming; but the assumption that GABA always means hyperpolarisation is wrong often enough to be worth holding loosely.

Why does inhibition matter so much? A brain that could only excite itself would seize. GABAergic braking sets the gain of neural circuits, carves out the timing that lets networks oscillate in rhythm, and quiets the neurons you are not using so that the ones you need stand out. Lose the brake and you get anxiety, insomnia and, in the extreme, seizures, which is exactly why this system is the workhorse target for anxiolytics, sleep aids, anticonvulsants and general anaesthetics ^[31].

The table below lays out the machinery, the timing and the ligands. Note the third row: the rho subunits form their own homopentamers, they are insensitive to bicuculline and to benzodiazepines, and they were called GABA-C for years before being folded back into the GABA-A family as GABA-A-rho ^[4][15]. Note also that a few compounds touch more than one family. Phenibut is primarily a GABA-B agonist, and the weak GABA-A activity often attributed to it is not well characterised at doses a person actually takes ^[26][27].

A GABA-A receptor is a pentamer: five protein subunits arranged in a ring around the central chloride pore. The brain draws those five from a menu of nineteen; six alpha (alpha1-6), three beta (beta1-3), three gamma (gamma1-3), three rho, plus delta, epsilon, pi and theta ^[1][5]. Most synaptic receptors settle on a common recipe of two alpha, two beta and one gamma, arranged in a fixed order around the ring, and the structural work of the last decade has confirmed that arrangement directly: a beta3 homopentamer in 2014 ^[34], the human alpha1beta2gamma2 receptor with flumazenil bound in 2018 ^[7], and a cryo-electron microscopy series with alprazolam, diazepam, picrotoxin and bicuculline in 2019 ^[8].

Those structures made two things concrete that used to be inferred. First, GABA binds at two sites, each at an interface between a beta and an alpha subunit. Second, the benzodiazepine site is not a separate invention; it sits at the equivalent interface between an alpha and the gamma2 subunit, the structural mirror of a GABA site occupied by a different chemistry ^[2][7]. That is why a benzodiazepine needs a gamma2 subunit to be present at all, and why a receptor built with alpha4 or alpha6 ignores it: those two subunits carry an arginine where alpha1, alpha2, alpha3 and alpha5 carry a histidine, and one amino acid decides the whole class ^[5][2].

How much any of this matters for behaviour was settled by an unusually clean set of experiments. Mice were engineered with a single histidine-to-arginine point mutation in one alpha subunit, which makes that subunit blind to diazepam while leaving GABA itself working normally. In alpha1 mutants, diazepam lost its sedative and amnesic actions and part of its anticonvulsant action, while the anxiety-reducing, muscle-relaxing and alcohol-potentiating effects survived intact ^[9]. In alpha2 mutants the anxiolytic effect disappeared, and in alpha3 mutants it did not ^[10]. That is about as direct an assignment of a drug effect to a molecular target as behavioural pharmacology gets, and it is the whole basis of the subtype-selective drug programmes described below ^[11][12].

The commercial payoff so far is modest but real. Zolpidem and Zaleplon act at the benzodiazepine site with a preference for alpha1, which is the sedation subunit, and in side-by-side recordings across the receptor subtypes tested, zaleplon's potency ran at about one third to one half of zolpidem's ^[14]. That preference is what makes them hypnotics rather than general anxiolytics; it is a partial selectivity, not a clean one, and at higher doses the distinction fades.

People often speak of GABAergic drugs as one thing. They are not. A GABA-A receptor carries at least five distinct pockets, and the practical differences between drug classes fall almost entirely out of which pocket a compound occupies and whether that pocket can open the channel on its own ^[15][6].

That last question is the one that decides how dangerous a drug is. A benzodiazepine cannot open the channel by itself; it only increases the chance that GABA will. That imposes a ceiling, and it is why benzodiazepine overdose alone is usually survivable and why flumazenil exists as a specific reversal agent. A barbiturate potentiates GABA at low concentration but directly gates the channel at higher ones, so the dose-response curve has no ceiling, which is precisely why barbiturates were displaced as sedatives ^[15]. The neurosteroids behave the same way, potentiating at low concentration and opening the channel directly at higher ones ^[16].

Neurosteroids deserve more attention than they usually get, because they are the brain's own allosteric modulators rather than a foreign chemistry. Allopregnanolone, a metabolite of Progesterone, and THDOC bind a pocket in the transmembrane domain of the alpha subunit and potentiate GABA strongly, with the largest effects at extrasynaptic delta-containing receptors that benzodiazepines cannot touch at all ^[16]. That pharmacology is now a drug class: brexanolone is intravenous allopregnanolone, and in two phase 3 trials in post-partum depression it beat placebo on the 17-item Hamilton scale at 60 hours, by 5.5 points at the lower dose in the first study and 2.5 points in the second ^[17]. Those are real separations from placebo on a rapid timescale, and they are also a few points on a scale after a 60-hour infusion, which is the honest way to hold both facts at once. Zuranolone is the oral follow-up on the same mechanism, and Ganaxolone is a synthetic analogue developed for seizure disorders.

Alcohol is the messiest entry in the table. Enhancement of GABA-A function by low to moderate concentrations of ethanol, in the range of 3 to 30 millimolar, has been reported for more than thirty years, and which subunit combinations are responsible at those concentrations is still not agreed; recent work points toward extrasynaptic receptors, but it is a live question rather than a settled fact ^[18]. What is not in doubt is the clinical consequence: ethanol, benzodiazepines and barbiturates all push on overlapping GABA-A machinery, their effects on breathing add up, and combining them is the single most reliable way to turn a survivable dose into a fatal one.

At the gentler end, several plant flavonoids modulate GABA-A receptors, some through the benzodiazepine site and some through sites that flumazenil does not block ^[19]. That is the mechanism usually invoked for Apigenin, the chamomile flavone, and for parts of what Passionflower, Lemon Balm and Magnolia Bark are supposed to do. The receptor pharmacology is genuine in vitro; whether the concentrations reached after a cup of tea are anywhere near it is a separate question, and mostly unanswered.

GABA-B was the last major neurotransmitter receptor to be cloned, in 1997, and the sequence came as a surprise: it belongs to class C, the same small family as the metabotropic glutamate receptors and the calcium-sensing receptor, rather than to the large class A family that holds most monoamine receptors ^[20][25]. Class C receptors carry a large extracellular Venus flytrap domain that closes around the transmitter, which is a completely different binding architecture from a pocket buried inside the membrane helices.

Then came the second surprise. The cloned receptor barely worked. Expressed alone, GABA-B1 stays trapped inside the cell and never reaches the surface, and it binds agonists far more weakly than brain membranes do. The resolution was that the functional receptor is an obligate heterodimer: GABA-B1 and GABA-B2 must assemble together, GABA-B1 carries the flytrap that binds GABA, and GABA-B2 does the trafficking and the actual G-protein coupling ^[21][24]. This was the first clear demonstration that a GPCR could require heterodimerisation to function at all, and it changed how the whole field thought about receptor assembly ^[22].

There is a third layer. KCTD proteins (KCTD8, 12, 12b and 16) bind the tail of GABA-B2 as auxiliary subunits and set the kinetics and the agonist potency of the assembled receptor, so two anatomically distinct GABA-B receptors made of identical GABA-B1 and GABA-B2 can respond on quite different timescales ^[23]. This is the reason a single GABA-B pharmacology produces such varied physiology across brain regions.

What the receptor does depends on which side of the synapse it is on. Post-synaptically it opens GIRK potassium channels and produces the slow inhibitory potential that follows a fast one. Pre-synaptically it inhibits voltage-gated calcium channels and cuts transmitter release, acting as an autoreceptor on GABA terminals and as a heteroreceptor on glutamate terminals, so it can reduce excitation and inhibition depending on where it sits ^[22][24]. All of it runs through Gi/o, which also lowers cAMP.

The compound comparison is the useful part. Baclofen is the reference agonist, with an affinity around 6 micromolar in rat brain membranes. Racemic phenibut binds at about 177 micromolar and R-phenibut at about 92 micromolar, and the S enantiomer is essentially inactive in behavioural tests up to very high doses, so phenibut is roughly fifteen times weaker than baclofen at the receptor they share ^[27]. That single number explains the gram-scale dosing, and it is the reason phenibut has a reputation for a slow, hard-to-judge onset that invites redosing. There is also a wrinkle worth knowing: R-phenibut binds the alpha2-delta subunit of voltage-gated calcium channels with an affinity around 23 micromolar, about four times tighter than it binds GABA-B, and in nerve injury models its anti-nociceptive effect was not blocked by a GABA-B antagonist ^[28]. Phenibut is therefore not a pure GABA-B drug; part of it is doing something closer to what gabapentin does.

Almost everything in the calming half of the pharmacopeia routes through this system. Fast anxiolytics and sleep aids lean on the GABA-A benzodiazepine and neurosteroid sites; barbiturates, propofol and much of alcohol's effect run through overlapping GABA-A machinery, which is why mixing depressants is so dangerous, since their effects on breathing add rather than compete ^[15][18]. GABA-B agonists such as Baclofen treat spasticity, and phenibut is a genuinely effective GABA-B anxiolytic that carries genuine dependence risk ^[26][29].

The catch across the whole system is tolerance, and the interesting finding is that tolerance is effect-specific rather than drug-specific. Reviewing the benzodiazepine literature, tolerance develops quickly to the sedative and anticonvulsant actions, while tolerance to the anxiolytic and amnesic actions may not develop at all ^[30]. That is a strange result if you picture a single receptor being turned down; it makes more sense once you accept that different effects run through different subtypes in different circuits, each adapting on its own schedule. The proposed mechanisms include changes in subunit expression, uncoupling of the benzodiazepine site from the GABA site, receptor trafficking away from the membrane, and compensating changes in glutamate signalling ^[30].

Withdrawal is the mirror image. Lean on the brake for long enough and the brain rebalances by reducing its own GABAergic tone and raising excitatory tone; remove the drug and that compensation is suddenly unopposed, producing rebound anxiety, insomnia, tremor and, with the strongest agents, seizures. This is why benzodiazepine and alcohol withdrawal are medically dangerous in a way that stimulant withdrawal is not, and why they show cross-tolerance with each other. The same logic applies on the GABA-B side; a survey of online phenibut use found an average dose around 2.4 grams and named tolerance and withdrawal as the commonly reported adverse effects, with emergency presentations for both heavy sedation and withdrawal ^[29].

The practical reading of all of this is not abstinence, it is restraint with the strongest tools. The compounds that work fastest and hardest on GABA-A are the ones with the steepest tolerance curves and the worst withdrawal, and the gentler options are gentler partly because they never engage the receptor as forcefully. Nothing in the pharmacology suggests that pushing harder on inhibition produces a durably calmer brain; the receptor adapts, and the adaptation is the problem. See Anxiolytics & GABA for the applied version of this, and Tolerance and dependence for the general case.

How many GABA-A subtypes actually exist in a brain. The arithmetic of nineteen subunits taken five at a time gives an absurd number; the real count is far smaller, because assembly is constrained. A careful attempt to list the native subtypes produced 26 candidates sorted into three tiers: identified, existence with high probability, and tentative ^[4]. That tiering is the honest state of the field. Sources that state a specific number of GABA-A subtypes as fact are compressing an open question.

Whether subtype selectivity translates from animals to people. This is the most instructive failure in the field, because the preclinical case was excellent. Compounds selective for alpha2 and alpha3 over alpha1 were non-sedating anxiolytics in rodents and primates, exactly as the knock-in mouse work predicted. In humans, MRK-409 produced sedation at receptor occupancy below 10 percent, which nobody expected; TPA023 and TPA023B avoided that by having no efficacy at alpha1 at all, and TPA023 did show an anxiolytic-like signal, but its clinical trials were terminated early because of preclinical toxicity findings ^[13]. Three decades after the target was identified, no subtype-selective anxiolytic is on the market. The mechanism was right and the drugs still did not arrive.

Which receptors alcohol actually acts on at drinking concentrations. Enhancement of GABA-A function at 3 to 30 millimolar ethanol is old and reproducible, but the identity of the sensitive subunit combinations in an intact brain remains contested, with extrasynaptic delta-containing receptors the current leading candidate rather than a settled answer ^[18]. This matters practically, because it is the difference between alcohol having a specific receptor target and alcohol acting diffusely across several.

Whether swallowed GABA does anything. GABA is sold as a supplement, and a review of the evidence concluded that the blood brain barrier question is genuinely unresolved rather than closed in either direction, that the studies reporting calming effects were disproportionately produced by researchers with a commercial interest, and that any real effect might be routed through the enteric nervous system rather than the brain at all ^[33]. Neither confident answer is supported. What is clear is that GABA at body pH carries both a positive and a negative charge, and doubly charged molecules cross into the brain poorly, which is the whole reason prodrug approaches like Picamilon were attempted in the first place.

The compounds that borrowed the name. Pregabalin and Gabapentin are structural analogues of GABA that do not bind GABA-A or GABA-B in any meaningful way; their target is the alpha2-delta auxiliary subunit of voltage-gated calcium channels, where they reduce transmitter release ^[32]. Homotaurine and Taurine are frequently described as GABAergic on the strength of in vitro activity at concentrations far above anything an oral dose produces. Naming is not mechanism, and in this family the two come apart constantly. See Neurotransmitters 101 for the wider map, and Glutamate receptors for the accelerator this system is braking.