April 2026

Enterohinal cortex represents task-relevant remote locations independently of CA1

Is the medial enterohinal cortex (MEC) merely a real-time GPS, or can it mentally project do distant locations? During movement, MEC population activity indeed accurately decodes to the mouse’s actual location. But when the animal pauses, the neural representation frequently “jumps” to distant, rewarded locations on the opposite site of the maze, in a phenomenon the authors dub “nonlocal” coding. This nonlocal coding is highly structured and breaks several classic assumptions.

• Dual representation: Cells tuned to the remote location fire alongside cells tuned to the current location.
• Independance from hippocampus: Most MEC nonlocal coding events were found to occur outside of CA1 ripples (or replay)

Even more remarkably, when the researches changed the rules of the game, the MEC’s remote representations dynamically flipped to align with the new logical associations. Thus the MEC is an active associative engine, capable of supporting spatial planning that operates in parallel to the hippocampus.

Dual computational systems in the development and evolution of mammalian brains

Mammals exhibit a massive evolutionary trade-off between the neocortex and the olfactory-limbic system. This trade-off can be explained by two fundamentally distinct computational architectures competing for the same space.

When artificial neural networks were trained to classify different types of sensory modalities, two distinct architectures emerged:

• Spatiotopic networks: Networks optimized for visual, somatosensory, and auditory representations were found to be spatiotopic—meaning they form highly localized and ordered maps.
• Distributed networks: Networks optimized for olfactory and relational memory representations were found to be distributed, forming fractured patterns of information convergence.

When researchers trained a multimodal network where all types of sensory modalities had to be resolved at once, these two architectural domains were locked in a zero-sum competition.

Furthermore, they found that modalities coupled to the same architecture (e.g., relational memory and olfaction) would expand simultaneously. This paper thus demonstrates pleiotropic evolution: selecting for one trait (like smell) drags along developmentally tethered systems, causing network-wide shifts in cognition (like relational memory) even without direct evolutionary pressure.

The prefrontal cortex controls memory organization in the hippocampus

How does the brain decide whether to weave a new experience into an old memory, or file it away separately? When mice explored two different contexts separated by 7 days, the memories are kept naturally apart. If the contexts are the same, we would expect the memories to integrate.

Surprisngly, when the ventromedial prefrontal cortex (vmPFC) was inhibited, this separation collapsed, and the mice could no longer separate the same contexts. At the neural level, it was demonstrated that when the vmPFC was inhibited the two memories had overlapping neuronal ensembles. Conversely, when the vmPFC was activated the memory overlap was reduced, separating two memories with identical context. This gate was also shown to be reserved for distant memories. Inhibition of the vmPFC had little effect when two memories were formed closely apart (5 hours).

From a neuroanatomic perspective, it was further found that vmPFC does not dicctate memory allocation by talking to the CA1 directly; instead it projects to the MEC. The MEC → CA1 signal modulates the recruitment of neurogliaform-like inhibitory neurons. The takeaway is the memory organization is not governed soley by the hippocampus. It is an actively regulated top-down process where prefrontal input affects hippocampal memory allocation, ensuring related memories integrate while unrelated memories remain distinct.