Science: How does the hippocampus selectively recruit neurons to consolidate memory?

The hallmark of neuronal network activity is the selective recruitment of neurons into active collections, forming a temporarily stable pattern
of activity.
In the hippocampus of mammals, this collection of neurons is repeatedly activated during ripples (~200Hz) oscillations, supporting the consolidation
of spatial memory and declarative memory.

Information processing in neuronal networks involves recruiting selected neurons into coordinated patterns of spatiotemporal activity
.
This sparse activation is the result
of extensive synaptic inhibition combined with neuron-specific synaptic excitation.
A central question, however, is how individual neurons are selected to participate in these modes of cooperation
.

Recently, Martin Both's research team at the University of Heidelberg in Germany published a study in Science, reporting a new mechanism
by which hippocampal pyramidal cells are selectively recruited into model network activity.
During oscillations of ripples in vivo

Preferential recruitment of AcD cells

Recent studies have shown significant morphological and functional inhibition between host cells in cortical networks, and found non-cellular origin (i.
e.
, derived from dendrites) of pyramidal cell sub-concentrated axons [Fig.
1A-C].

The morphological characteristics of axons determine the differentiation and recruitment
of individual neurons.
The researchers began to investigate whether ripple oscillations in the hippocampus also have such characteristics
.

In CA1, about 50% of pyramidal cells were found to have axons originating from dendrites, so there is a possibility
that axons carry functional differences between dendritic cells (AcD cells) and non-AcD cells.
In vivo recordings found that AcD cells had a ~4.
5-fold higher probability of firing during ripples than non-AcD cells, and the firing frequency was about ~2.
5 times higher [Fig.
1F, G].

Therefore, AcD cells produce particularly strong inhibitory effects
during the cycle compared to non-AcD cells.

Figure 1 Preferential recruitment of AcD cells during in vivo ripples oscillation

Morphological differences between AcD and non-AcD cells

Are the firing properties of AcD cells related to their morphology? That is, excitatory input into AcD cells escapes perisomatic inhibition
.
The researchers conducted a series of experiments
in acute brain slices of the hippocampus.
First, there was no difference in the relative time of excitation and inhibition between AcD cells and non-AcD cells [Fig.
2C].

In addition, the number of dendrites at the base of AcD and non-AcD cells was similar, and there was no difference
in branching pattern, total dendritic length, and dendritic spine density.
However, AcD is longer than the basal dendrites of non-AcD cells, suggesting that synaptic inputs of specific dendrites have significant weights [Fig.
2I].

In addition, synaptic inhibition, I/E ratio, and dendritic complexity were basically similar
between the two cell types.
Thus, synaptic input may be a factor determining the preferred activation of AcD cells [Fig.
2].

Figure 2 AcD and non-AcD cells receive equivalent synaptic input and have similar dendritic morphology

Differences in firing between AcD and non-AcD cells

Next, to assess the effect of axon origin on the probability and threshold of discharge under different synaptic input conditions, the researchers used a fine multicompartment cell computational model
.
In this model, cells consist of a soma and three dendrites, with axons being somato-derived non-AcD and dendritic-derived AcD,
respectively.
AcD cells are more likely to excite APs (action potentials) than non-AcD cells [Fig.
3A].

A strong increase in local (dendritic) AMPA conductance may be accompanied by small changes
in apparent I/E ratio.
Thus, APs in AcD cells may be caused by particularly strong excitation of AcD, whereas somatos record similar I/E ratios for AcD cells and non-AcD cells [Fig.
3B].

This fully suggests that the difference in excitability of inputs to AcD and non-AcD branches increases with increasing axon-somal distance and perisomal inhibition [Fig.
3C].

Figure 3 Single-cell multi-chamber computational model predicts the difference between AcD and non-AcD cells

Functional differences between AcD and non-AcD cells

The above results suggest that the different recruitment mechanisms of pyramidal neurons into network activity depend on their axon origin
.
Even when typical (non-AcD) pyramidal cells are inhibited by GABA in large quantities [Fig.
4E], AcD cells retain the ability to
issue APs.
In this network state, activation of AcD cells is largely limited to excitatory input
to AcD.

The special function of this dendritic structure makes the network functional connection state-dependent, that is, when the perisomal inhibition is strong, the excitatory input has a more effect on AcD, and when the perisomal inhibition is weak, the effect is similar for all dendritic inputs [Fig.
4E].

This morphological-functional mechanism explains how specific cells are preferentially activated during the ripples oscillations of the hippocampal network
.
Figure 4 Perisomal inhibition of different controls on information processing by perisomal inhibition of AcD and non-AcD cells

Summary

In this study, it is proposed that axon position is a key determinant
of asymmetric recruitment in oscillating network states.
Moreover, the recruitment of neurons into active neuronal clusters is determined by the morphological characteristics of axons
.