Of memories and ripples

The hippocampus is one of the regions in the mammalian brain that
is associated with memory of events in their spatiotemporal
context. Sequences of neuronal activity in the hippocampus are the
chief candidate for a neurophysiological correlate of such
contextual, or episodic memory. Simultaneously to replaying these
behaviorally-related activity sequences, the hippocampus engages in
a powerful and fast oscillation known as sharp-wave ripples (SWR).
Ripples in turn participate in a brain-wide pattern of activity and
may orchestrate the local strengthening of memories and their
broadcasting to the cortex. In this Thesis, both memory sequences
and ripple oscillations are studied in the light of the unifying
hypothesis that the coordinated activation of a neuronal assembly
represents an individual memory item in the sequences, and is at
the same time responsible for the individual cycles in the
oscillations. To test the hypothesis, we investigated SWR in vitro
and in vivo in the mouse, using intracellular recordings of
currents in CA1 pyramidal cells referenced to the local field
potential. Expanding current hypotheses on SWR generation, we found
powerful, well ripple-locked and spatially pervasive but CA1-local
excitatory inputs, indicative of presynaptic assemblies of CA1
principal neurons. Combining a novel peeling reconstruction
algorithm for synaptic currents with recordings at different
holding potentials, we could for the first time unravel individual
synaptic contributions during ripples. Analysis of the strikingly
precise timing of currents demonstrated that inhibition aligns its
phase to excitation over the course of a ripple. We carried on the
dissection of ripples to the theoretical domain by incorporating
the effect of inhibition into a mean field model of sequence
replay. Using this model, we inquired what are the neuronal
assembly size and inhibitory feedback strength that maximize the
capacity of a hippocampal network to store memories, so that those
memories can be successfully retrieved during ripple episodes. We
found that a linearly coupled inhibitory population indeed helps
increase storage capacity by dynamically stabilizing replay in an
oscillatory manner for lower assembly sizes than in absence of
inhibition. The findings about the temporal structure of neuronal
activation during ripples complement our experimental observations.
Collectively, they offer new insights on the physiology and
function of sharp-wave ripples, paving the way for an integrated,
continuous-time model of large networks of sparsely connected
neurons that replay activity sequences concomitant to transient
ensemble oscillations.