Dendritic hold and read: a gated mechanism for short term information storage and retrieval.

Mariton D Santos,Sunggu Yang,Bradley E Alger,Joseph P Y Kao,Michael H Mohammadi,Cha-Min Tang,Conrad W Liang,Scott M Thompson,Ya-Ping Tang

doi:10.1371/journal.pone.0037542

Abstract

Two contrasting theories have been proposed to explain the mechanistic basis of short term memory. One theory posits that short term memory is represented by persistent neural activity supported by reverberating feedback networks. An alternate, more recent theory posits that short term memory can be supported by feedforward networks. While feedback driven memory can be implemented by well described mechanisms of synaptic plasticity, little is known of possible molecular and cellular mechanisms that can implement feedforward driven memory. Here we report such a mechanism in which the memory trace exists in the form of glutamate-bound but Mg2+-blocked NMDA receptors on the thin terminal dendrites of CA1 pyramidal neurons. Because glutamate dissociates from subsets of NMDA receptors very slowly, excitatory synaptic transmission can leave a silent residual trace that outlasts the electrical activity by hundreds of milliseconds. Read-out of the memory trace is possible if a critical level of these bound-but-blocked receptors accumulates on a dendritic branch that will allow these quasi-stable receptors to sustain a regenerative depolarization when triggered by an independent gating signal. This process is referred to here as dendritic hold and read (DHR). Because the read-out of the input is not dependent on repetition of the input and information flows in a single-pass manner, DHR can potentially support a feedforward memory architecture.

Highlights

One testable prediction of the dendritic hold and read (DHR) hypothesis is that the NMDA receptors and the dendrite on which they are located can be electrically silent during information storage
Photolysis is used to release a low concentration of free glutamate that is sufficient to bind to the high affinity NMDA receptors over a restricted region of a terminal dendrite, but below that needed to activate the low affinity AMPA receptors
The potentiation observed here represents a novel cellular mechanism of response potentiation and not a form of dendritic paired pulse facilitation (PPF). These experiments demonstrate that synaptic activity, as simulated with photoreleased glutamate, can leave a transient ‘‘memory trace’’ on individual dendritic branches of pyramidal neurons

Summary

Introduction

Without some form of short term memory buffer to hold together a temporal sequence, the individual bits of information that are received at any moment in time cannot be properly interpreted. In order to predict the trajectory of a flying ball, it is necessary to hold in mind a temporal sequence of recent positions of the ball. The cellular mechanism for such a short term memory buffer is poorly understood. Some in the neural computational field began to question whether the memory networks proposed to handle static signals can adequately handle dynamic signals [3,4,5]. Current models postulate that the memory trace is maintained by synaptic reverberations within a recurrent feedback network [1,2]. Mechanisms of synaptic plasticity such as paired pulse facilitation (PPF) (Figure 1A) and NMDA-stabilized recurrent synapses have been proposed to establish ‘attractor’

Methods

Results

Conclusion