Crossfrequency routing in a hippocampocortical circuit during probabilistic reversal learning

Abstract

Learning under uncertainty requires detecting latent changes in environmental contingencies and flexibly adjusting choice strategy. To model this process and identify associated circuit dynamics, we trained rats on a two-armed bandit task with uncued reward reversals while simultaneously recording local field potentials from the circuit comprised by dorsal hippocampus (CA1d), lateral entorhinal cortex (LEC), and prefrontal cortex (PFC). Behavioral performance improved with training and progressively shifted from outcome-reactive exploration to an exploitation-biased strategy, quantified using a model-free index. Mixed-effects modeling combining behavioral and neural metrics identified tonic cortical synchrony in theta and fast-gamma bands as a negative session-scale performance marker. Theta and gamma oscillations were coordinated, as cross-regional phase-amplitude coupling showed that CA1d theta phase gated fast-gamma activity in PFC, while biasing both slow- and fast-gamma bursts in LEC, which selectively increased at goal-reaching. Neuronal spiking showed cross-regional phase-locking to both theta and gamma rhythms, indicating synchronized timing across the hippocampo-cortical circuit. Finally, distributed neuronal spiking across the circuit represented goal-approach, yet only prefrontal neurons ramped during goal-reaching, suggesting a role in outcome assessment, while hippocampal and entorhinal units transiently suppressed, consistent with locomotor tracking and resetting. These results reveal a tonic cortical connectivity marker of ongoing performance during reversal learning and dissociate it from hippocampal theta mechanisms that selectively organize cortical gamma bursts and spike timing. Together, these results provide mechanistic insight into how frequency-specific hippocampo-cortical interactions dynamically reconfigure to support strategy transitions, offering a circuit-level framework for understanding adaptive decision-making under uncertainty. Significance StatementAnimals often learn in uncertain environments, where they must decide whether to keep exploiting a known option or explore alternatives. We trained rats on a probabilistic choice task while recording brain activity from hippocampus, entorhinal cortex, and prefrontal cortex. We found that performance improvements were linked to an ongoing interaction between entorhinal and prefrontal areas, whereas hippocampal signals changed mainly with training and controlled the timing of brief, fast activity bursts in cortex. Near reward, prefrontal neurons ramped up while hippocampal and entorhinal neurons were suppressed. Together, these results separate brain signals that track current performance from those that shape cortical dynamics during learning.