Reinforcement learning and Gilles de la Tourette syndrome: dissociation of clinical phenotypes and pharmacological treatments.

CONTEXT Gilles de la Tourette syndrome (GTS) is a hyperkinetic movement disorder with heterogeneous phenotypic expression ranging from simple motor and vocal tics to more complex tics associated with psychiatric comorbidities. The heterogeneity of clinical phenotypes may relate to the dysfunction of distinct frontal cortex-basal ganglia circuits. OBJECTIVES To assess the hypothesis that simple motor tics and comorbid obsessive-compulsive disorders are associated with dysfunction of motor and reward circuits, respectively, and to assess the effects of various antipsychotic medications because they are known to reduce motor tics and interact with dopamine-related reward processing. DESIGN Sixty patients with GTS were divided into different subgroups depending on their clinical phenotypes and pharmacological treatments. The GTS patients and healthy control subjects underwent functional magnetic resonance imaging while they performed an instrumental learning task that involved adjusting choices between 2 responses (left- or right-hand movements) based on outcomes (reward or no reward). SETTING Reference center for GTS, Centre de NeuroImagerie de recherche (CENIR) Paris, France. Patients  Sixty GTS patients and 50 controls. RESULTS Movement-related activation in motor circuits was diminished in GTS patients with simple tics only. Reward-related activation in limbic circuits was independently reduced by the following 2 factors: the presence of associated obsessive-compulsive symptoms (mostly compulsions) and the presence of medication with typical antipsychotics (dopamine receptor antagonists). Computational modeling with standard reinforcement learning algorithms indicated that, for both factors, the diminished reward-related activation could account for the impaired choice performance. Reinforcement learning was not affected by aripiprazole, a recent medication that acts as a partial dopamine agonist. CONCLUSIONS These results support the hypothesized correspondence between clinical phenotypes and frontal cortex-basal ganglia circuits. Antipsychotic treatment effects comply with formal conceptions that dopamine serves as a teaching signal for reinforcement learning. Furthermore, we suggest that, unlike typical antipsychotics, aripiprazole may preserve reward sensitivity and hence avoid blunting motivational drives.