Psychedelic-inspired drug discovery using an engineered biosensor

G protein-coupled receptors (GPCRs) are major drug targets, and ligands that stabilise particular receptor conformations can bias signalling toward therapeutically desirable pathways while avoiding side effects. Ligands at the serotonin 2A receptor (5-HT2AR) exemplify biased agonism and include clinically important compounds such as atypical antipsychotics, classical psychedelics (for example LSD), and so-called psychoplastogens that promote neural plasticity. Recent preclinical work suggests that neural plasticity and therapeutic benefit can be elicited by some 5-HT2AR-targeting compounds without necessarily producing overt hallucinatory effects, creating a need for assays that can predict whether a novel ligand will induce hallucinogen-associated receptor conformations. Dong and colleagues set out to build a genetically encoded fluorescent biosensor based on the human 5-HT2AR that reports ligand-induced receptor conformations and endogenous serotonin (5-HT) release in real time. Their goals were to: (1) create a sensor able to detect physiologically relevant 5-HT dynamics in vitro and in vivo, (2) determine whether sensor readouts distinguish hallucinogenic from non-hallucinogenic 5-HT2AR ligands, and (3) apply the sensor in a medium-throughput cellular assay to discover novel ligands, including non-hallucinogenic psychoplastogens with antidepressant-like properties.

The investigators engineered psychLight by inserting a circularly permuted GFP (cpGFP) into the third intracellular loop of the human 5-HT2AR, then optimised linker sequences, insertion sites and point mutations to maximise response to 5-HT. The top variant was termed psychLight1; addition of an endoplasmic reticulum export (ER2) motif produced psychLight2 with improved membrane localisation in neurons and HEK293T cells. A control sensor (psychLight0) carried a D155A mutation that abolishes agonist binding. Sensor characterisation combined cellular, slice and in vivo methods. In HEK293T cells psychLight dose–response and competition assays were run by confocal microscopy to obtain EC50 and Emax values and to compare responses to known agonists and antagonists. For ex vivo kinetics and sensitivity, organotypic cortical cultures and acute brain slices expressing psychLight2 were imaged using two-photon microscopy; RuBi-5-HT uncaging and electrical stimulation were used to evoke local 5-HT transients. In freely behaving mice, AAV9/8-mediated expression of psychLight2 plus implanted optical fibres enabled fibre photometry recordings during auditory fear conditioning and after systemic administration of agonists or antagonists. To create a medium-throughput assay, the authors generated a stable HEK293T cell line (PSYLI2) expressing psychLight2 under an EF1a promoter, and implemented wide-field high-content imaging in 96-well format. Assays were run in two modes: agonist mode (test compound alone) and antagonist mode (100 nM 5-HT plus test compound) to distinguish agonists, antagonists and non-binders. A combined ‘‘ligand score’’ was defined from agonist and antagonist readouts (calibrated to LSD and MDL100907). Screening libraries included 83 compounds of known or suspected 5-HT2AR activity and a separate set of 34 compounds with unknown hallucinogenic potential. Concentration–response assays, Schild regression for antagonist potency, and controls for compound autofluorescence (plate-reader checks) were incorporated. Behavioural and functional follow-up for selected hits included head-twitch response (HTR) assays and locomotion in mice, dendritogenesis assays in cultured embryonic cortical neurons, the forced swim test (FST) for antidepressant-like activity, and a sucrose preference test in VMAT2 heterozygous (VMAT2-HET) mice to assess anti-anhedonic effects. Standard statistical approaches were used (randomisation, blinding, ANOVA, repeated-measures ANOVA, Bonferroni post hoc tests) and key assay performance metrics (for example Z-factor) were reported.

Sensor development and in vitro pharmacology: Through linker and insertion-site screening Dong and colleagues produced psychLight1 and the improved membrane-trafficked psychLight2. In HEK293T cells psychLight1 responded to 5-HT with an EC50 comparable to conventional G protein and β-arrestin assays. Known 5-HT2AR agonists increased fluorescence to differing degrees, whereas classical antagonists (ketanserin, MDL100907) produced minimal fluorescence or slight quenching and blocked 5-HT responses. Ex vivo and in vivo serotonin dynamics: Two-photon uncaging of RuBi-5-HT on psychLight2-expressing layer 2/3 pyramidal neurons evoked rapid fluorescence transients that decayed with a Tau off of 5.4 ± 0.9 ms. In acute BNST slices single-trial electrically evoked 5-HT release was detectable (sensitivity metric d0 reported as 234.2 for those experiments) and two response types were observed: a smaller fast-decaying response (DF/F = 4.7% ± 1.5%, Tau off fast = 0.997 ± 0.038 s) and a larger slow-decaying response (DF/F = 9.7% ± 1.2%, Tau off slow = 3.998 ± 0.610 s). Escitalopram increased response amplitude (DF/F = 18.4% ± 4.3%), while tetrodotoxin, the 5-HT3 antagonist granisetron, and ketanserin blocked electrically evoked psychLight signals. During auditory fear conditioning recorded by fibre photometry, psychLight2 reported region-specific 5-HT dynamics: in the dorsal raphe nucleus (DRN) a robust fluorescence increase coincided immediately after foot-shock onset followed by a sharp decline; in the BNST an immediate decrease that returned to baseline within ~4 s was observed; in the basolateral amygdala (BLA) and orbitofrontal cortex (OFC) an initial decrease was followed by a substantial post-shock rise. psychLight0 (D155A) showed no significant changes, supporting ligand specificity. Trial-level discriminability metrics (d0) for each region were reported (example d0 ≈ 12 for DRN, BNST, BLA, OFC in fear conditioning). Differentiation of hallucinogens and non-hallucinogens: In HEK293T assays psychLight1 was activated by canonical serotonergic hallucinogens with these reported potencies and efficacies: LSD EC50 = 18.8 nM, Emax = 20.0%; 5-MeO-DMT EC50 = 157 nM, Emax = 48.4%; DOI EC50 = 35.5 nM, Emax = 52.9%; DMT EC50 = 627 nM, Emax = 12.4%. Structurally similar non-hallucinogenic congeners (for example lisuride, 6‑MeO‑DMT) bound the receptor but failed to increase sensor fluorescence even at 10 μM. PsychLight1 potencies (EC50) correlated strongly with hallucinogenic potencies in humans (r2 = 0.9), whereas Emax values, not potencies, separated hallucinogens from non-hallucinogens in this assay. PsychLight readouts also differed from traditional downstream assays (phosphoinositide hydrolysis, Gq activation, calcium mobilisation), suggesting psychLight reports distinct ligand-induced conformational states. Medium-throughput screening and ligand scoring: The PSYLI2 cell line and high-content imaging produced a robust assay (Z-factor = 0.6). Screening 83-compound libraries showed that serotonergic hallucinogens produced responses > +1 SD from vehicle in agonist mode, whereas non-hallucinogenic 5-HT2AR ligands decreased fluorescence in antagonist mode. Autofluorescent artefacts were detected for two compounds (BOL-148, bromocriptine) and controlled by cell-free fluorescence measurements. The combined agonist/antagonist ‘‘ligand score’’ categorised ligands: LSD had a ligand score of 23.0, lisuride −42.3; non-serotonergic hallucinogens/dissociatives (salvinorin A, ketamine, PCP) scored near 0. Schild analyses indicated lisuride, apomorphine and benztropine act as competitive psychLight antagonists, whereas 6‑MeO is less potent. Prediction and in vivo validation of designer drugs: From a set of 34 compounds of unknown hallucinogenic potential psychLight predicted that smaller halogenated DMTs (5‑F‑DMT, 5‑Cl‑DMT) would be hallucinogenic and that larger 5‑Br‑DMT would not. Three-point dose–response HTR assays confirmed that 5‑F‑DMT and 5‑Cl‑DMT elicited robust HTRs while 5‑Br‑DMT did not. Locomotor effects did not correlate with HTR frequency, emphasising functional specificity of the assay. Discovery and characterisation of AAZ‑A‑154: The screen identified AAZ‑A‑154 as a candidate non-hallucinogenic 5-HT2R ligand occupying novel chemical space. In cellular assays AAZ‑A‑154 behaved as a competitive antagonist of psychLight (Schild regression). It was selective for 5-HT2 receptors across a panel of GPCR sensors and produced no HTRs up to 100 mg/kg, although high doses reduced locomotion. In cultured rat embryonic cortical neurons AAZ‑A‑154 increased dendritic arbor complexity to an extent comparable to ketamine; this psychoplastogenic effect was blocked by ketanserin, implicating 5-HT2 receptor involvement. Antidepressant-like and anti‑anhedonic effects: In C57BL/6J mice AAZ‑A‑154 reduced immobility in the forced swim test at 30 min and maintained effects 1 week after a single administration. In VMAT2‑HET mice, which show baseline anhedonia, a single dose of AAZ‑A‑154 (15 mg/kg) restored sucrose preference to wild-type levels immediately and the effect persisted for at least 12 days; total fluid consumption did not differ between genotypes, and WT mice showed no change in sucrose preference with the compound.

Dong and colleagues interpret psychLight as a 5-HT2AR-based fluorescent sensor that bridges in vitro pharmacology and in vivo behaviourally relevant readouts by directly reporting ligand-induced receptor conformations and endogenous 5-HT dynamics. The sensor's millisecond off-kinetics allowed detection of rapid 5-HT release/reuptake ex vivo and temporally precise signals in vivo during fear conditioning. PsychLight showed higher apparent affinity than an alternative serotonin sensor (iSeroSnFR) but a smaller dynamic range, a property the authors suggest may be advantageous for reporting low-concentration 5-HT events while risking saturation during massive release. A central claim is that psychLight fluorescence correlates with hallucinogenic receptor conformations: hallucinogenic serotonergic agonists produced robust sensor activation, whereas structurally similar non-hallucinogenic congeners bound the receptor without activating the sensor. Dong and colleagues propose that this direct readout of 5-HT2AR conformational states fills a methodological gap left by assays that either capture only snapshots of binding or depend on slower secondary signalling readouts. The authors anticipate the psychLight cellular assay will be scalable (for example to 384-well format) and complementary to orthogonal GPCR assays for drug discovery. Using psychLight, the investigators identified AAZ‑A‑154, a non-hallucinogenic 5-HT2R ligand that promotes dendritic growth and produces rapid and sustained antidepressant-like and anti-anhedonic effects in rodents after a single administration. The authors position AAZ‑A‑154 alongside other non-hallucinogenic psychoplastogens such as tabernanthalog, noting comparable or potentially greater potency and duration in vivo. Limitations acknowledged by the authors include the need for structural determination of psychLight bound to activating versus inactivating ligands to illuminate molecular mechanisms; the unknown relationships between psychLight signals and other downstream signalling assays; and the requirement for a comprehensive pharmacological profile of AAZ‑A‑154 (mechanism, off-targets, pharmacokinetics, full dose–responses, toxicity, and orthogonal validation of efficacy). They recommend future work combining imaging, electrophysiology and behavioural tools to dissect the circuit-level mechanisms that distinguish hallucinogenic from antidepressant effects.