For social animals and humans to thrive, they must exhibit social competence, a key component of which is adjusting behavior based on social history. However, most neuroscience studies do not measure variables related to social history (e.g., familiarity level and social rank). Therefore, there is a large gap in our understanding of how social history affects the neural basis of social behaviors, and how this changes with autism spectrum disorder (ASD).
Individuals with ASD show atypical hypothalamic activity during social competitions11 and atypical prefrontal cortical activity during cognitive tasks2. Yet no study to date has tested if ASD mouse models result in functional changes in the prefrontal-hypothalamic pathway.
In this project, Nancy Padilla-Coreano and co-PI Shreya Saxena plan to use two etiologically distinct ASD mouse models (Shank3 and 16p11.2) to identify common pro-social behavioral motifs that are affected across both models and to understand the role of the medial prefrontal-lateral hypothalamic (mPFC-LH) pathway in pro-social behaviors. To close the current critical gaps, they also plan to use artificial intelligence (AI) tools to quantify how ASD models change behaviors across social history and identify the role of the mPFC-LH circuit in pro-social behavior.
Mouse behavior can be categorized into a set of interpretable, discrete “behavioral motifs”: brief recurring patterns of behavior defining the scope of the animal’s behavioral repertoire. However, it is yet unknown how sociability-related behavioral motifs change with social history. Existing tools for analyzing behavior use pose estimation, which is limited to the supervised labels provided and excludes important temporal and postural information that contains relevant behavior. Unsupervised techniques circumvent these problems but are currently designed for single-animal data.
To quantify social behavior in wild-type and ASD model mice reliably and automatically, Padilla-Coreano and colleagues recently developed a computer vision machine learning pipeline that combines both supervised and unsupervised approaches and dynamical modeling to characterize behavioral motifs in multi-animal settings3-4. In Aim 1, her team plans to use AI tools to characterize social behavioral motifs in these two ASD mouse models across social history states. Next, in Aim 2, they plan to use optogenetics and AI to test the hypothesis that the mPFC-LH circuit gates pro-social behavior and that decreasing its activity will lead to an increase in passive and active pro-social behavior in the ASD mouse models.
Knowing which social motifs and sequences differ in ASD mouse models will allow future closed-loop experiments to record and manipulate neural activity that could rescue these differences. These experiments will ultimately facilitate a better understanding of the neural mechanisms of prosocial behaviors to inform new therapies.