Molecular and cellular approaches to memory allocation in neural circuits

Abstract

Although memory allocation is a subject of active research in computer science, little is known about how the brain allocates information within neural circuits. There is an extensive literature on how specific types of memory engage different parts of the brain, and how neurons in these regions process and store information. Until recently, however, the mechanisms that determine how specific cells and synapses within a neural circuit (and not their neighbors) are recruited during learning have received little attention. Recent findings suggest that memory allocation is not random, but rather specific mechanisms regulate where information is stored within a neural circuit. New methods that allow tagging, imaging, activation, and inactivation of neurons in behaving animals promise to revolutionize studies of brain circuits, including memory allocation. Results from these studies are likely to have a considerable impact on computer science, as well as on the understanding of memory and its disorders.

Fig 1. Model of memory allocation in neuronal populations

(A) Neurons in a naïve neural circuit (open circles) are recruited into encoding episode A (orange). This increases their excitability so that shortly thereafter they are very likely to also be involved in encoding episode B (purple). With time, the increase in excitability wanes and consequently episode Z (blue) no longer is stored in the same neurons. A consequence of this pattern of storage is that recall of episode A will also result in the recall of episode B (and vice versa), while recall of episode Z becomes unrelated to the other two episodes. (B) Training causes an temporary increase in the activity and levels of CREB. The CREB gene family of transcription factors includes both activators (they increase transcription) and repressors. Following the initial increase in activators (green section of the curve), CREB repressors are expressed that decrease the overall levels of CREB activity, thus eventually bringing them below basal levels (red section of the curve). (C) Higher levels of CREB in a specific cell population result in increases in the levels of specific proteins (e.g. Scn1b), which in turn increase the excitability of neurons. Thus, two memories acquired while CREB levels are high would be stored in overlapping populations of neurons. Since the change in excitability would involve transcription and require the stability of transcribed molecules, such as channels, the time scale of this memory allocation mechanism would be in the order of hours (or perhaps even days), so that one memory could affect the allocation of proceeding memories for many hours.

Fig 2. Model of memory allocation within dendritic trees

(A) Following learning and subsequent potentiation of synapses in encoding neurons, molecular components diffuse to nearby (10 microns) unpotentiated spines for a limited time (10 minutes), thus resulting in a temporary increase in the probability that these spines will participate in subsequent learning (become potentiated themselves). (B) Two memories acquired within minutes of each other may be stored in similar populations of cells (Fig. 1) and in nearby synapses (orange and purple circles), thus resulting in strong co-recall. (B) In contrast, two memories acquired within hours of each other may be stored in overlapping cellular populations, but perhaps not in nearby synapses, which would result in weaker co-recall. Lack of neuronal or dendritic co-localization would prevent automatic co-recall. Thus, unlike the cellular allocation mechanism discussed in fig. 1, the time-scale of synaptic allocation mechanisms would be in the order of minutes.