Distinct Patterns of Adaptive Decision-Making in Mouse Models of Neurodevelopmental Disorders

Abstract

Neurodevelopmental disorders (NDDs) arise from diverse genetic alterations that can affect adaptive decision-making, but quantifying specific components of cognitive processes remains challenging. We developed a sequence of deterministic and probabilistic reversal learning tasks using an automated operant platform and applied reinforcement-learning models to quantify latent cognitive processes in mice. We applied this framework to three genetic NDD models: Mecp2+/- (Rett syndrome), MECP2-Tg1 (MECP2 duplication syndrome), and Kmt2c+/fs (KMT2C haploinsufficiency) mice. Mecp2+/- mice showed heterogeneous impairment in acquiring deterministic reversal learning, accompanied by reduced task engagement and increased premature responses. MECP2-Tg1 mice exhibited mild impairment of deterministic reversal learning but preserved performance in probabilistic reversal learning. Although Kmt2c+/fs mice showed largely preserved behavioral performance, reinforcement-learning modeling revealed increased learning from unrewarded outcomes and faster forgetting of unchosen options, which were not apparent from overall performance. These findings demonstrate that combining quantitative reversal learning tasks with computational modeling can distinguish specific alterations in cognitive processes, providing an effective framework for characterizing cognitive dysfunction across mouse models of NDDs.