Same circuit, two designer receptors — dial neural output up or down at will
Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) are engineered G-protein coupled receptors that respond only to a synthetic ligand, not to any endogenous neurotransmitter. By co-engineering a single circuit — or a single population tested across two sessions — with both an excitatory (hM3Dq) and inhibitory (hM4Di) DREADD, researchers gain a genetic switch that can push the same neurons in either direction on demand.
DREADDs were created by directed molecular evolution of native human muscarinic acetylcholine receptors. Researchers mutagenized the ligand-binding pocket of M3 and M4 receptors and screened thousands of variants in a yeast growth assay, selecting clones that had lost sensitivity to acetylcholine but gained sensitivity to clozapine-N-oxide (CNO) — a compound with negligible activity at any native receptor.
The result was two receptors built on the same seven-transmembrane GPCR chassis: hM3Dq, which retains the Gq-coupling of native M3, and hM4Di, which retains the Gi/o-coupling of native M4. Neither responds meaningfully to the brain's own chemistry; both respond robustly and selectively to their synthetic ligand.
Chemogenetics complements optogenetics rather than replacing it: DREADDs need no implanted fiber optic, can be activated brain-wide or in deep structures by a single systemic injection, and produce effects lasting hours rather than milliseconds — better suited to modulating slow processes like feeding, mood, or sleep than to millisecond-precise circuit dissection.
Armbruster et al. (2007, PNAS) created hM3Dq and hM4Di by screening mutant human M3 and M4 muscarinic receptors in a yeast growth assay, selecting variants that lost affinity for acetylcholine but gained sensitivity to the otherwise pharmacologically inert compound CNO — inventing a synthetic lock-and-key pair from a natural GPCR scaffold.
Getting hM3Dq into one neuron subset and hM4Di into another (or the same neurons across two sessions) relies on the standard chemogenetic toolkit: Cre-dependent AAV vectors injected into Cre-driver mouse lines restrict expression to a genetically defined cell type (e.g., AgRP neurons, D1-MSNs, a specific amygdala nucleus).
Two common circuit-switch designs exist. In a within-subject design, the same neurons are transduced with a single DREADD, and the "switch" is which ligand-receptor pairing is used across separate testing days with counterbalanced washout. In a between-population design, intersectional viral strategies (Cre + Flp dependence, or two spectrally distinct fluorophore-tagged constructs) allow hM3Dq and hM4Di to be expressed in genetically distinguishable subpopulations within the same animal, enabling simultaneous bidirectional readouts.
Before any ligand reaches the brain, hM3Dq- and hM4Di-expressing neurons behave exactly as they would without any transgene at all. This is the essential control state of every chemogenetic experiment: receptor expression alone, without its cognate synthetic ligand, should not perturb circuit function — a property that must be verified with vehicle-only sessions before any causal claim can be trusted.
hM3Dq and hM4Di share the identical seven-transmembrane GPCR fold inherited from their muscarinic parents, and both bind the same synthetic ligand with comparable affinity. What differs is a short stretch of the third intracellular loop — the region that determines which intracellular G-protein a GPCR recruits.
hM3Dq retains the Gq-coupling of native M3: agonist binding activates phospholipase Cβ, generating IP3 and diacylglycerol, releasing ER calcium, and ultimately depolarizing the cell. hM4Di retains the Gi/o-coupling of native M4: agonist binding inhibits adenylyl cyclase and, more acutely, liberates Gβγ subunits that open GIRK potassium channels, hyperpolarizing the cell.
This is the mechanistic root of bidirectional chemogenetic control: identical receptor architecture, identical ligand, opposite intracellular partner, opposite cellular outcome. The "switch" is not in the drug — it is in which G-protein the engineered receptor was built to recruit.
Clozapine-N-oxide (CNO) was originally described as a pharmacologically inert metabolite of the antipsychotic clozapine. Early DREADD studies used systemic CNO injection almost universally. But subsequent work showed that CNO undergoes measurable back-conversion to clozapine in rodents — and clozapine itself binds native serotonergic, dopaminergic, and muscarinic receptors at behaviorally relevant concentrations, raising concern about off-target confounds in some earlier studies.
The field has since largely shifted to deuterated clozapine (DCZ), a substantially more potent and selective DREADD agonist effective at doses as low as 0.1 mg/kg with minimal back-conversion, and to compound 21 (C21), another selective alternative. Rigorous designs still include vehicle-only sessions and CNO/DCZ-in-naive-animal control sessions to confirm the ligand itself is behaviorally silent absent the receptor.
Gomez et al. (2017, Science) showed that CNO undergoes measurable back-conversion to clozapine in rodents, which itself binds native receptors at behaviorally relevant doses — prompting the field's shift toward deuterated clozapine (DCZ), effective at just 0.1 mg/kg with minimal off-target and back-conversion liability.
When the ligand reaches hM3Dq-expressing neurons, it sets off the same Gq cascade that native Gq-coupled receptors use to depolarize cells — only now under the experimenter's explicit control. Within one to two minutes, intracellular calcium rises, firing rate climbs, and — in landmark cases like the hypothalamic AgRP circuit — a specific, quantifiable behavior follows almost immediately.
Ligand binding to hM3Dq activates Gαq, which stimulates phospholipase Cβ (PLCβ) to cleave membrane PIP2 into two second messengers: IP3, which triggers calcium release from the endoplasmic reticulum via IP3 receptors, and diacylglycerol (DAG), which activates protein kinase C.
The resulting rise in intracellular Ca²⁺ opens depolarizing conductances (including TRPC-family cation channels), pushing the membrane potential toward threshold and increasing both spontaneous and evoked firing rate. Because this cascade runs through second messengers rather than a direct ionotropic channel, onset is slower than optogenetics — roughly one to two minutes — but the elevated state persists for the pharmacokinetic lifetime of the ligand, typically one to several hours.
The arcuate nucleus AgRP neuron is one of the most thoroughly validated chemogenetic circuits in neuroscience. These neurons are known from lesion and optogenetic studies to promote food intake; hM3Dq expression let researchers ask whether activating them alone, in a fully sated animal with no other manipulation, is sufficient to trigger feeding.
It was. CNO injection into fed, satiated AgRP::hM3Dq mice produced feeding behavior within minutes, with total intake rising roughly ten-fold over vehicle sessions — establishing chemogenetics as a way to test causal sufficiency of a defined neural population for a specific, measurable behavior, without touching any other part of the brain.
Krashes et al. (2011, J Clin Invest) showed that Gq-DREADD activation of AgRP neurons in fed, satiated mice triggered feeding within minutes, with intake rising roughly ten-fold over baseline — one of the first demonstrations that chemogenetic excitation alone is sufficient to drive a complex, motivated behavior.
Deliver the identical ligand to hM4Di-expressing neurons in the same circuit, and the outcome reverses completely. Gi/o signaling suppresses cAMP production and, more immediately, opens GIRK potassium channels — hyperpolarizing the membrane and silencing firing. This is the necessity half of the bidirectional experiment: instead of asking whether activating a circuit is sufficient for a behavior, it asks whether the circuit is required for that behavior to occur at all.
Ligand binding to hM4Di activates Gαi/o, which inhibits adenylyl cyclase and lowers intracellular cAMP, damping PKA-dependent excitability over minutes to hours. The more immediate and often dominant effect, however, comes from the liberated Gβγ subunits, which directly bind and open G-protein-coupled inwardly rectifying potassium (GIRK / Kir3) channels.
The resulting outward K⁺ current hyperpolarizes the neuron by roughly 10–15 mV, raising the threshold for action potential firing and suppressing both spontaneous and evoked activity — frequently silencing the neuron almost entirely for the duration of ligand exposure.
hM4Di activation opens GIRK channels, driving a hyperpolarizing shift of roughly 10–15 mV — sufficient to silence spontaneous firing in most central neurons within minutes of ligand exposure. The same manipulation of AgRP neurons nearly abolishes CNO/DCZ-independent food intake, mirroring the ~90% suppression reported by Krashes et al. (2011).
Gain-of-function experiments (activating hM3Dq) test sufficiency: does turning this circuit on produce the behavior? Loss-of-function experiments (activating hM4Di) test necessity: does turning this circuit off eliminate the behavior when it would otherwise occur?
A circuit that is both sufficient and necessary — where excitation drives the behavior and inhibition abolishes it — satisfies the strongest available standard of causal evidence in systems neuroscience, well beyond what correlational recording, static lesion, or single-direction manipulation can offer. Chemogenetics is particularly suited to this pairing because the same viral targeting strategy, the same animal, and often the same session structure can be reused for both directions simply by swapping which DREADD is expressed or which cohort is tested.
When the same anatomically defined circuit produces opposite, quantifiable behavioral phenotypes purely as a function of which designer receptor was engaged, the result is among the strongest causal claims available in neuroscience. This bidirectional logic — excite it, the behavior appears; inhibit it, the behavior disappears — has been repeated across feeding, anxiety, and motor circuits, and is now being explored for circuit-based psychiatric interventions.
AgRP feeding circuit: hM3Dq excitation triggers voracious feeding in sated mice within minutes; hM4Di inhibition of the same population blunts fasting-induced feeding — establishing AgRP neurons as both sufficient and necessary for hunger-driven food seeking.
Amygdala anxiety circuits: excitatory DREADD activation of basolateral-to-central amygdala projections increases anxiety-like avoidance behavior in standard assays (open field, elevated plus maze), while inhibitory DREADD silencing of the same projection reduces it — demonstrating bidirectional control of an affective state from a single, genetically defined pathway.
Striatal direct/indirect pathway switching: Gq activation of D1-expressing direct-pathway medium spiny neurons increases locomotion and reward-related behavior, while activating Gi in D2-expressing indirect-pathway neurons produces the opposite motor and motivational effect — a textbook case of two molecularly distinct but anatomically interleaved populations pushing the same downstream output in opposite directions.
Because the same anatomical circuit can be pushed in either direction depending solely on which designer receptor is expressed, bidirectional DREADD experiments satisfy both necessity and sufficiency criteria in a single genetic background — widely regarded as the gold-standard evidentiary standard for establishing that a defined neural population causally controls a specific behavior.
The same logic that lets researchers dial a mouse circuit up or down is now motivating circuit-based interventions for human psychiatric and neurological disease. Rather than a fixed lesion or a blunt systemic drug, a chemogenetic-style approach could in principle titrate the activity of a specific, disease-relevant circuit up or down as needed — a form of pharmacologically reversible, cell-type-specific neuromodulation.
Orthogonal ligand-gated ion channel platforms (PSAM/PSEM chemogenetic tools) and next-generation DREADD variants are being explored as translatable analogues, with early-stage work targeting circuits implicated in epilepsy, treatment-resistant depression, and addiction. The core appeal is the same one demonstrated at the bench: reversible, titratable, bidirectional control of a single circuit, without the permanence of a lesion or the invasiveness of an implanted device.