Reduced correlations in spontaneous activity amongst CA1 engram cells

Memories are believed to be stored in the strengthened synaptic connectivity between
distributed networks of neurons after learning. This biophysical memory trace is often referred to as
an engram, provisionally defined as the cells activated during learning, reactivated during memory
retrieval, and causally linked to the expression of memory. Recent work exploring the activity of these
ensembles in vivo demonstrates that after learning, c-Fos tagged populations consist of highly active,
correlated ensembles of hippocampal neurons. The current model for engram formation predicts
that during learning, c-Fos tagged cells undergo synaptic potentiation, resulting in their increased
propensity to be reactivated together during memory retrieval, which has yet to be demonstrated with
in vivo recording from engram cells. Given previous modeling work on assembly formation in
recurrent networks, we reasoned that if memory formation resulted in strengthened synaptic
connectivity between CA3 and CA1 engram cells, we would be able to detect this via increased
correlations of their spontaneous activity. Here, we tested this prediction with a spiking network
model, in which an ‘engram’ was directly simulated in a population of CA1 neurons by strengthening
their common inputs from CA3. We observed a relative increase in pairwise correlations of the
spontaneous activity of tagged CA1 engram neurons with one another. We then tested this model in
vivo by combining two-photon calcium imaging with the c-Fos driven TetTag system to identify IEG
tagged putative engram neurons and simultaneously record their population activity. Surprisingly, we
observed that after contextual fear conditioning, c-Fos tagged cells in CA1 displayed relatively lower
spontaneous correlations compared to neighboring non-tagged cells, which disagrees with our
proposed model. These results are inconsistent with a model of engram formation based on simple
Hebbian learning rules in excitatory synapses, and may indicate that engram activation is gated by
context-dependent or retrieval-dependent sensory inputs.