The study explores the neural mechanisms underlying physiological memory, focusing on how the brain encodes past experiences to generate corresponding physiological changes. While much research has focused on behavioral memory, relatively little is known about how physiological memories are formed and stored. The authors hypothesize that physiological and behavioral memories may not share the same neural architecture. They chose to investigate corticotropin-releasing hormone (CRH) neurons in the paraventricular nucleus of the hypothalamus (PVN) because these neurons control the hypothalamic-pituitary-adrenal (HPA) axis, a key player in the stress response, and exhibit bidirectional changes in activity in response to both aversive and rewarding stimuli. Previous research has shown that these neurons are both necessary and sufficient for contextual memory formation, making them strong candidates for investigating physiological memory. The researchers aim to determine how CRH PVN neurons encode and recall memories of both aversive and appetitive experiences and whether this encoding aligns with behavioral responses.

The authors review existing research on associative memory, particularly the role of engrams (specific neuron subsets activated during learning and retrieval) in brain regions such as the cortex, amygdala, and hippocampus. They note that even species with simpler nervous systems can form associative memories, highlighting the possibility of distinct memory mechanisms. The study builds upon previous findings demonstrating the bidirectional response of CRH PVN neurons to aversive and rewarding stimuli, their role in gating defensive behaviors, and their necessity for anxiety states. The authors also cite previous work demonstrating that repeated optogenetic manipulation of CRH PVN activity is sufficient to create contextual memory.

The study utilized male and female Crh-IRES-Cre;Ai14 and Crh-IRES-Cre;Ai148 mice. Stereotaxic surgery was performed to implant either optical fibers (for fiber photometry) or GRIN lenses (for miniature microscopy) targeting the PVN. A Cre-dependent AAV construct with GCaMP6s or GCaMP6f was used to express calcium indicators in CRH neurons. Mice were exposed to different contexts: a novel neutral context, a context paired with foot shocks (aversive), and a context paired with hazelnut spread (appetitive). Fiber photometry and miniature microscopy were used to record calcium transients in CRH PVN neurons during and after context exposure. Plasma corticosterone levels were measured to assess the HPA axis response. Behavioral responses (freezing) were also recorded. Data analysis included principal component analysis (PCA) to assess changes in neuronal activity state space, and data-driven clustering to identify neuronal subpopulations. Computational modeling using a network of leaky integrate-and-fire neurons was used to simulate the learning rules underlying the observed changes in neuronal activity. Statistical analyses involved paired and unpaired t-tests, ANOVA, linear regression and mixed effects models. Histology was performed post-experiment to verify GCaMP expression and fiber/lens placement.

Exposure to a novel, neutral context triggered sustained anticipatory activity in CRH PVN neurons. Following aversive foot shock conditioning, CRH PVN neurons showed a more robust increase in activity upon re-exposure to the context, reflecting contextual recall of negative valence. Analysis of individual neuron activity revealed a two-factor learning rule: the magnitude of the response to the foot shock and the weakness of the cell's activity before the shock were critical for determining activity increase during recall. Weakly responding neurons showed a disproportionately large increase in activity after the foot shock. Computational modeling confirmed that these two factors could explain the observed increase in anticipatory response. In contrast, appetitive conditioning with hazelnut spread led to a significant decrease in CRH PVN neuron activity upon re-exposure to the context. This was modeled successfully using a simpler, one-factor learning rule based only on pre-exposure activity. Importantly, the aversive memory in CRH PVN neurons outlasted the behavioral response (freezing), exhibiting no extinction even as freezing behavior extinguished with repeated exposure. The study demonstrated a striking disconnect between behavioral and physiological responses to aversive stimuli.

The findings demonstrate that hypothalamic CRH PVN neurons form enduring contextual memories of positive and negative experiences through distinct learning rules. The two-factor rule for negative experiences is unique, recruiting weakly active neurons into the response pool based on their pre-stimulus activity and shock response strength. This contrasts with typical Hebbian learning. The simpler, one-factor rule for positive experiences involves a scaled reduction in activity across the neuron population. The persistence of CRH PVN activity despite behavioral extinction underscores the independence of hypothalamic memory from cortical, amygdala, and hippocampal processes. The anticipatory activity in CRH PVN neurons in response to a decrease in safety is interpreted as a precautionary signal preparing the organism for potential threats. The disconnect between behavioral and physiological responses highlights that external behaviors may not accurately reflect internal physiological states, challenging common assumptions about using behavior as a proxy for emotional states.

This study provides crucial insights into the neural mechanisms of physiological memory, demonstrating that hypothalamic CRH PVN neurons form long-lasting contextual memories of both positive and negative experiences through distinct learning rules. The independence of this hypothalamic memory system from other brain regions involved in behavioral memory suggests a more complex interplay between physiological and behavioral responses to salient stimuli. Future research should explore the downstream targets of these CRH PVN neurons and how they contribute to specific physiological changes associated with positive and negative memories. The identified learning rules may also be further investigated to elucidate the underlying molecular and cellular mechanisms.

The study focused on a specific neuronal population (CRH PVN neurons) in mice. The generalizability of these findings to other species and neuronal populations warrants further investigation. The computational model is a simplification of the complex biological system, and the parameters used in the model may not perfectly capture the underlying biological processes. The study primarily focused on short-term memory; longer-term memory formation and maintenance should be further studied.