Translational studies in humans and mice to test a circuit-level computational model of auditory hallucinations..

Auditory hallucinations (AH) are core symptoms of psychosis for which treatment is often ineffective or poorly tolerated. A first step towards developing more selective and safer biological interventions is to elucidate AH mechanisms at a neurobiological circuit level, a level at which AH are currently poorly understood. Here, we use a novel computational circuit-level model combined with translational experiments in humans and mice to identify circuit mechanisms underlying AH. Dorsal-striatal dopamine (DA) excess is implicated in AH, and AH severity correlates with a task behavioral phenotype consisting of increased false alarms (endorsing auditory sounds that are not present in signal-detection tasks) reported with high confidence. In mice, stimulating DA release in the dorsal striatum also induces this AH-like phenotype of high-confidence false alarms in a similar signal-detection task. These findings are consistent with computational models whereby AH result from exaggerated perceptual prior expectations and suggest a role for their implementation in dorsal striatum. However, the precise relationships between model-proposed cognitive computations and circuit neurobiology are unclear. Important gaps include how dorsal-striatal DA and medium spiny neuron activity contribute to perceptual learning and AH-like percepts, as well as potential additional roles of reward-based processes in ventral striatum. To address these gaps, we have developed a first-of-its-kind computational corticostriatal circuit model of AH which recapitulates documented behavioral and neural phenotypes associated with perceptual and reward tasks, and which additionally generates DA-dependent AH-like false alarms. Informed by this model, here we will use human data from antipsychotic-free patients with schizophrenia (Aim 1) and mouse data including a mouse model of genetic risk for schizophrenia (Aim 2), combined with a translational signal-detection paradigm, to test quantitative predictions from our AH circuit model. Aim 1 (humans) will use behavior, fMRI and neuromelanin-sensitive MRI to test for distinct contributions of perceptual learning to AH and their implementation by dorsal-striatal circuits and dopaminergic nigral regions innervating dorsal striatum. Aim 2 (mice) will use DA sensors, neuronal recordings, and optogenetic stimulation to parse the specific contributions of dorsal and ventral-striatal DA and medium spiny neurons to perceptual learning and AH-like false alarms. Exploratory Aim 3 will develop circuit-model extensions incorporating additional circuit elements (e.g., direct D1 and indirect D2 pathways, cholinergic interneurons) to help further explain circuit mechanisms of existing D2 and candidate non-D2 antipsychotic drugs. This multidisciplinary project will thus use translational and computational methods combining the strengths of clinical and preclinical research, and of theory- and data-driven methods, to advance our knowledge about circuit mechanisms of psychosis. By outlining circuit-level dopaminergic and non-dopaminergic targets, it will further pave the way for developing novel treatment approaches with enhanced selectivity and tolerability.