Authors :
Presenting Author: Mohamed Rabieh Khalife, PhD – Nemours Children's Health
Patrick Jasinski, PhD – Nemours Children's Health
Khalil Abed Rabbo, MS – University of Vermont
Mohamed Ouardouz, PhD – Nemours Children's Health
Matt Mahoney, PhD – The Jackson Laboratory
Rod Scott, MD/PhD – Nemours Children's Health
Amanda Hernan, PhD – Nemours Children's Health
Rationale:
Cognitive flexibility relies on rapid reorganization of neuronal population codes in the medial prefrontal cortex (mPFC), but the circuit mechanisms that support this adaptability remain poorly defined. Early-life seizures (ELS) provide a controlled developmental perturbation that produces long-lasting impairments in learning and network plasticity. We tested whether adaptive network reorganization is a fundamental principle of mPFC function, whether it is disrupted by ELS, and whether it can be preserved by transient intervention with adrenocorticotropic hormone (ACTH), a clinically relevant neuropeptide acting via melanocortin-4 receptor (MC4R) signaling.
Methods: Male and female C57BL/6J mice underwent flurothyl-induced ELS from postnatal days 10–14, with daily pretreatment of ACTH or vehicle. At adulthood, animals received mPFC tetrode implants and performed a two-day fear conditioning and extinction paradigm. Single-unit activity was analyzed for firing rates, spike-timing dynamics via generalized linear model post-spike filters, and functional connectivity like graph-theoretic measures of centrality, clustering, and efficiency. Adaptive reorganization during extinction was quantified by comparing early vs. late session dynamics. Predictive modeling of freezing behavior was performed with an optimized graph attention network (GAT), benchmarked against linear regression and generalized estimating equations (GEE).
Results: All groups acquired fear learning normally, but vehicle-treated ELS animals showed impaired extinction with persistently high freezing, whereas ACTH-treated ELS mice extinguished normally. At baseline, ELS neurons exhibited reduced firing rates and exaggerated beta/theta spike-timing rigidity; ACTH preserved normal temporal flexibility. During extinction, controls showed robust increases in tone-evoked firing and strengthening of functional connectivity, with parallel gains in clustering, centrality, and network efficiency. ELS animals failed to show this adaptive reorganization, while ACTH treatment preserved it. GAT modeling accurately predicted freezing behavior (R² = 0.54 early, 0.81 late) and outperformed linear regression (R² = –0.24 early, 0.28 late). Feature attribution highlighted firing rate, inter-spike interval variability, and temporal coding as critical predictors of adaptive behavior.
Conclusions: Adaptive reorganization of mPFC population codes is a circuit-level requirement for fear extinction learning. Transient developmental disruption like ELS impairs this flexibility, but ACTH delivered only during seizure days rescues both network dynamics and cognitive performance. These results establish flexible network reorganization as a tunable substrate for learning, demonstrate the translational potential of early neuropeptide intervention, and highlight graph neural network modeling as a powerful tool for dissecting and predicting circuit mechanisms of cognition.
Funding: This work was supported by NIH NINDS 5K22NS104230 (AH), NIH NINDS 5R01NS134491 (AH), and R21NS117112 (RS).