July 2025

A flexible hippocampal population code for experience relative to reward

Task & imaging: Using 2-photon calcium imaging in CA1, head-fixed mice navigated VR tracks where the hidden reward zone moved across sessions/environments, dissociating spatial cues from reward.
Reward-relative (RR) code: A subpopulation remapped to the same distance relative to reward when reward moved, forming behavioral-timescale sequences spanning the whole task; the spatial (track-relative, TR) map was simultaneously preserved in a parallel, dynamic subpopulation.
Scope & decoding: RR remapping rose above chance and eventually covered nearly the entire track; after a switch, only RR cells decoded RR position above shuffle, and RR firing order relative to reward was preserved across switches/environments.
Learning & behavior: The RR representation strengthened with experience (more RR neurons, denser near reward) and often preceded behavioral updates by ~2 trials after reward relocation.

In the Drosophila head direction (HD) system, cue salience and familiarity shape the width, amplitude, and accuracy of the network’s “bump” attractor representation, with brighter or more familiar cues producing narrower, higher-amplitude bumps and better HD encoding.
When cues conflict, the more informative cue (based on individual-specific encoding accuracy) exerts greater influence, and stable pairing of cues improves encoding through Hebbian plasticity at sensory-to-HD synapses.
A computational ring attractor model with plastic inhibitory inputs recapitulates experimental findings, explaining individual variability through differences in sensory response amplitude.
Results highlight a stability–flexibility trade-off in spatial representations, showing how continuous synaptic plasticity supports adaptation in dynamic environments.

In head-fixed mice navigating a precisely controlled visual–olfactory virtual reality, hippocampal CA1 place cells still exhibited significant representational drift over days, even when behavioral patterns and sensory inputs were highly reproducible.
Drift rate was unaffected by subtle differences in running speed, task engagement, or odor variability, indicating that drift is not solely a byproduct of environmental or behavioral noise.
Place cell excitability predicted stability: more excitable cells maintained place fields more consistently over time.
Findings suggest that representational drift is an intrinsic property of hippocampal coding, potentially supporting temporal separation of experiences and continual learning.

In the Restaurant Row foraging task, mice exhibited two distinct phenotypes: offer-sensitive (decisions scaled with offer delay) and offer-insensitive (accepted offers regardless of delay).
Nucleus accumbens dopamine levels tracked decision confidence during evaluation and encoded past and future value during change-of-mind events.
In offer-sensitive mice, optogenetic dopamine inhibition increased change-of-mind behaviors, while stimulation boosted confidence and reduced later reversals; offer-insensitive mice showed weaker dopamine–behavior coupling.
Results reveal that dopamine’s role in decision-making depends on individual valuation strategies, extending its function beyond simple reward prediction error to include confidence modulation and future-oriented evaluation.