03. GABA and Glycine: The Invention of No

The anaesthetist asks you to count backward from ten. Ten is clear. Nine is clear. Somewhere near eight the edges of the room soften, the sentence you were assembling loses its verb, and seven never arrives — not because you declined it, but because the part of you that would have said it is no longer permitted to speak. From outside this takes about twenty seconds. From inside it takes no time at all, because inside is precisely what has been switched off.

What entered the vein was propofol, and what propofol did was find a pocket. Its target is the GABA-A receptor, the fast brake of the vertebrate brain: a pentamer, five protein subunits standing in a ring around a central pore like staves around a barrel, opening to admit chloride ions and so pushing the cell further from firing. Where two neighbouring staves meet, a cleft forms. Propofol and etomidate bind one particular cleft, the interface between a β subunit's outward face and an α subunit's inward face, written β+/α–. Neither drug opens the channel itself; each makes it easier for GABA — gamma-aminobutyric acid — to open it. The syringe adds nothing to your brain. It leans on a lever already installed.

Which lever matters. Reviewing the subunit pharmacology in the British Journal of Anaesthesia in 2017, Hemmings and colleagues set out a dissociation that surgery quietly depends on: the β3 subunit carries the immobilising, hypnotic effect — the going under — while a different subunit, α5, carries the amnesia. Unconscious and will-not-remember-this are two effects, routed through two rooms of one very old house. The rooms themselves are not old at all.

Inhibition had to be invented, and the reason is arithmetic. A network whose only word is go has no stable state between silence and seizure: every excited cell excites its neighbours, and the endings available are exhaustion or catastrophe. Something has to be able to say no. What nature reached for is almost comically economical. Take glutamate — the go signal of the last chapter — remove one carboxyl group as carbon dioxide, and what remains is GABA. The enzyme that does the snipping, glutamic acid decarboxylase or GAD, works with pyridoxal-5′-phosphate, the active form of vitamin B6, on a catalytic fold shared across amino-acid chemistry in all three domains of life. One chemical step separates the brain's principal yes from its principal no. The brake is machined from the accelerator.

That economy is tidy enough to invite a false inference: that GAD must therefore be a single deep inheritance. The evidence says otherwise. In Escherichia coli, GAD runs glutamate-dependent acid resistance — each decarboxylation consumes a proton, buying roughly two hours of survival at gastric pH, as Capitani and colleagues showed structurally in The EMBO Journal in 2003 — and the same system runs in Shigella, Listeria and Lactococcus. Then the archaea break it open. A GAD homolog in the hyperthermophile Thermococcus kodakarensis, locus TK1814, was knocked out and found to decarboxylate aspartate instead, feeding β-alanine and coenzyme-A synthesis, with no connection to GABA or to signalling of any kind. Pyrococcus makes GABA; its neighbour has aimed the identical fold elsewhere. Plants aimed it elsewhere again, into the GABA shunt — cytosolic GAD plus two mitochondrial enzymes routing carbon out of glutamate and back into the citric-acid cycle, doubling as a stress and pH signal. This is not one commit inherited. It is a versatile tool picked up repeatedly, for local reasons.

Plants do something stranger too. In Nature Communications in July 2015, Ramesh and colleagues in Matthew Gilliham's Adelaide group showed GABA regulating wheat's aluminium-activated malate transporter TaALMT1 through a short motif resembling the mammalian GABA-A binding site; mutate the motif and GABA sensitivity vanishes while ordinary transport continues. Real, and specific — but not a "plant GABA receptor." ALMT is an anion transporter with no structural kinship to the pentameric channel. GABA tunes it. GABA does not gate it.

Who makes GABA and who listens for it are separate questions, and the listening side splits again. The fast branch is the Cys-loop superfamily of pentameric ligand-gated ion channels — five subunits around a central pore, each pinned by a signature loop stitched shut between two cysteine residues, which is all the name means — propofol's barrel. In Genome Biology in 2005, Tasneem, Iyer, Jakobsson and Aravind found that whole architecture, ligand-binding domain and pore together, already present in several bacteria and in the archaeal genus Methanosarcina, and proposed a single lateral transfer into the ancestral eukaryotic lineage followed by one duplication into a cation-selective branch (nicotinic acetylcholine, serotonin-3) and an anion-selective branch (GABA-A, glycine). Two years later Bocquet, Corringer, Changeux and colleagues gave the claim a body: GLIC, a working, proton-gated pentamer from the cyanobacterium Gloeobacter violaceus. The barrel is bacterial.

What animals built on it was, for a long stretch, less than the textbooks imply. Tsang, Ng, Xu and Xue reported in 2007 that invertebrates carry only one ancestral GABA-A/glycine-receptor-like gene pair. Separate GABA-A and glycine receptors as distinct proteins, the α, β and γ families, the four human clusters on chromosomes 4, 5, 15 and X — all vertebrate, built on two whole-genome duplications plus one tandem duplication. β3 and α5 sit as neighbours in the chromosome-15 cluster. The dissociation the anaesthetist relies on is younger than jaws.

The slow branch runs deeper. GABA-B receptors are Family-C G-protein-coupled receptors, which conduct no ions but start an internal cascade instead — seconds, not milliseconds. Krishnan and Schiöth's group counted 33 Family-C GPCRs in the demosponge Amphimedon queenslandica in 2014, mGluR/GABA-B-type homologs among them, in an animal Vol. I already showed owns a postsynaptic scaffold and no neurons. In the sea anemone Nematostella vectensis, four GABA-B homologs carry conserved GABA- and baclofen-binding residues, and baclofen arrests the planula-to-polyp transition and blunts neurogenesis — work still a bioRxiv preprint from April 2019, its peer-review status unconfirmed. The ordering inverts the intuition: slow GABA sensing is confirmed in cnidarians and structurally present in sponges, while fast, synapse-grade, GABA-A-type inhibition has no confirmed instance anywhere below Bilateria.

Glycine's half was settled by poison and by cats. Strychnine — isolated from Strychnos nux-vomica seeds by Pelletier and Caventou in 1818–19, thereafter the reliable villain of detective fiction — is a competitive antagonist at the glycine receptor, and it kills by *dis*inhibition: it excites nothing, it merely lifts glycine's brake off motor neurons, and the body convulses itself to death on its own reflexes. Glycine's day job, dominant fast inhibition in spinal cord and brainstem, was pinned down by Aprison and Werman in 1965, measuring free amino acids in cat spinal cord and finding ventral-horn glycine above both GABA and glutamate exactly where inhibitory interneurons cluster. GABA's own confirmation came a year later, when Otsuka, Iversen, Hall and Kravitz caught it leaving an identified inhibitory axon on stimulation — in a lobster whose posture Vol. I's eleventh chapter would later turn on. Glycine works the far side of the wall as well, as a required co-agonist at the excitatory NMDA receptor, though in the forebrain that seat is largely taken by D-serine, identified by Wolosker and colleagues in 2000 and three to four times more potent there. Its maker, serine racemase, is far older than the job: the gene family spans at least eight invertebrate phyla and first duplicates at the eumetazoan ancestor, after sponges branched off. Old enzyme, late wiring — GABA's story told small.

When Homology Lies

The sharpest test arrived in 2025, and it failed the intuition badly. Nematostella's genome holds 42 channels of the GABA-A/glycine superfamily, and 37 of them fall in a cnidarian-specific radiation that looks, by every sequence metric available, like a set of GABA-A receptors; Hydra contributes ten more to the same radiation. Ojha and colleagues cloned seven testable members of the Nematostella set and expressed them in Xenopus oocytes — frog eggs, the standard host for asking a channel what it listens for.

None responded to GABA. None responded to glycine. Three responded to glutamate, and were renamed nvGluCl-1, -2 and -3. They are genuinely chloride-selective — substituting bulky gluconate for chloride shifted the reversal potential exactly as a chloride channel requires — and picrotoxin, the classic GABA-A blocker, blocks them like the real thing. The glutamate response runs through a single lysine, K229; mutate it to threonine and the response is gone. The work was posted as a bioRxiv preprint in July 2025 and published in Advanced Science in May 2026; the ledger below grades it as peer-reviewed.

The lesson is not that the channels are fake. They are real, ancient, pentameric, chloride-conducting and inhibitory. The lesson is that ligand identity — the actual word a channel listens for — is cheap to gain and cheap to lose, and one lysine can move it. Homology-based annotation, the workhorse that assigns names to genes in every newly sequenced genome, said GABA-A receptor and was wrong about the only thing that mattered. Function must be shown, never inferred from resemblance. Ctenophores sharpen the point: their genomes, as Vol. I's ninth chapter established, carry ionotropic glutamate receptors and no Cys-loop anion-channel superfamily at all.

The same discipline handles the supplement shelf, honestly rather than cheaply. The blood-brain-barrier literature has contradicted itself since 1958: studies in 1958, 1971 and 1988 found no crossing, four more between 1980 and 2002 found minuscule crossing, and the question has never been tested directly in humans. In mice, GABA leaves the brain roughly seventeen times faster than it enters — which alone could imitate impermeability in an organ that demonstrably carries GABA transporters. What the gut itself does with the molecule is better established: Philip Strandwitz's group isolated KLE1738, a human gut bacterium that grows on GABA and on nothing else they tested, depending on Bacteroides fragilis to release it — a resident of the human body whose only known food is the brain's word for no. On benefit, Hepsomali, Groeger, Nishihira and Scholey's review of fourteen placebo-controlled human trials found limited evidence for stress and very limited evidence for sleep. So: we do not know whether or how oral GABA reaches a brain, and the trial evidence is thin. Not fake. Not proven. Meanwhile the receptor family anaesthetists silence outright is the same one whose autoimmune destruction — anti-GAD65 antibodies, sometimes alongside glycine-receptor antibodies — produces the rigidity and spasm of stiff-person syndrome. This is not a mood molecule. It is a load-bearing dial, and bodies break when it slips.

The honest standing is a gap with a definite shape. Both halves of inhibition are old, and old separately: a promiscuous decarboxylase making GABA in some lineages and quite different things in others, and a bacterial barrel borrowed for purposes nobody has recovered. What no one has dated is the coincidence — the moment a lineage that happened to synthesise GABA also happened to evolve a pocket that recognised it, in the same cells, at the same time. That conjunction, not either part alone, is the commit worth finding, and the log is blank exactly where it should read. The next molecule made the opposite bargain: it committed early, spread everywhere, and settled most of itself somewhere with nothing to do with thinking at all.

Commit Ledger

Overturned if: a GABA-A/glycine-receptor-like gene from a basal bilaterian invertebrate is functionally cloned and shown to respond to GABA itself, pushing true ionotropic GABA-gated inhibition into or below early Bilateria rather than leaving it, as the evidence now does, effectively undated there.

Sources

• 2017 · Hemmings et al., "Role of GABA-A receptor subtypes in the behavioural effects of intravenous general anaesthetics" — British Journal of Anaesthesia 119:i167
• 2003 · Capitani et al., "Crystal structure and functional analysis of E. coli glutamate decarboxylase" — EMBO Journal
• 2012 · "Glutamate decarboxylase-dependent acid resistance in orally acquired bacteria" — Molecular Microbiology
• 2014 · Knockout of Thermococcus kodakarensis TK1814, showing aspartate-decarboxylase function — Journal of Bacteriology
• 2015 · Ramesh et al., "GABA signalling modulates plant growth by directly regulating the activity of plant-specific anion transporters" — Nature Communications 6:7879
• 2005 · Tasneem, Iyer, Jakobsson & Aravind, "Identification of the prokaryotic ligand-gated ion channels and their implications for the mechanisms and origins of animal Cys-loop ion channels" — Genome Biology 6:R4
• 2007 · Bocquet, Carvalho, Cartaud, Neyton, Poupon, Taly, Grutter, Changeux & Corringer, "A prokaryotic proton-gated ion channel from the nicotinic acetylcholine receptor family" — Nature 445:116–119
• 2007 · Tsang, Ng, Xu & Xue, "The Evolution of GABAA Receptor–Like Genes" — Molecular Biology and Evolution 24(2):599–610
• 2014 · Krishnan, Dnyansagar, Sällman Almén, Williams, Fredriksson, Manoj & Schiöth, Family-C GPCRs in Amphimedon — BMC Evolutionary Biology
• 2019 · GABA-B receptors and baclofen-arrested metamorphosis in Nematostella — bioRxiv 621060 — preprint; peer review unconfirmed
• 2026 · Ojha et al., "Functional Analysis of Ligand-Gated Chloride Channels in a Cnidarian Sheds Light on the Evolution of Inhibitory Signaling" — Advanced Science; first posted as bioRxiv 2025.07.26.666939
• 2010 · "Early History of Glycine Receptor Biology in Mammalian Spinal Cord Circuits" (synthesising Aprison & Werman 1965) — Frontiers in Molecular Neuroscience 3:8
• 1966 · Otsuka, Iversen, Hall & Kravitz, "Release of gamma-aminobutyric acid from inhibitory nerves of lobster" — PNAS 56:1110
• 2000 · Wolosker et al., "D-Serine is an endogenous ligand for the glycine site of the N-methyl-D-aspartate receptor" — PNAS 97:4926
• 2015 · Serine/aspartate racemase across invertebrate phyla — Amino Acids, with 2022 follow-up
• 2020 · Hepsomali, Groeger, Nishihira & Scholey, "Effects of Oral Gamma-Aminobutyric Acid (GABA) Administration on Stress and Sleep in Humans: A Systematic Review" — Frontiers in Neuroscience 14:923
• 2019 · Strandwitz et al., "GABA-modulating bacteria of the human gut microbiota" — Nature Microbiology 4:396–403
• 2018 · McVean, "GABA Supplements: Glorious, Gimmicky or Just Garbage?" — McGill Office for Science and Society — science communication, not primary literature
• 2015 · "Antibodies to Inhibitory Synaptic Proteins in Neurological Syndromes Associated with Glutamic Acid Decarboxylase Autoimmunity" — PMC4361655