HomeResearchLearning and Memory in Genetically Engineered Mice

Our Scientists

Learning and Memory in Genetically Engineered Mice

Research Summary

Susumu Tonegawa uses genetically engineered mice to investigate neural development and the molecular, cellular, and neural circuit mechanisms underlying learning and memory.

Our primary research interests are the molecular, cellular, and neural circuit mechanisms underlying acquisition, consolidation, and retrieval of hippocampus-dependent memory in rodents. To study these problems, we produce conditionally engineered (i.e., spatially targeted and/or temporally regulated) mice and analyze these mice by multifaceted methods, including molecular and cellular biology, in vitro and in vivo physiology, optical imaging, and behavioral studies. We attempt to identify deficits at each of the multiple levels of complexity in specific brain areas or cell types and to determine which deficits underlie specific aspects of learning and memory.

Memory Acquisition
Using the Cre/loxP system, we previously targeted a knockout of the obligatory N-methyl-D-aspartate (NMDA) receptor (NR) subunit, NR1, to the CA1 pyramidal cells of young adult mice. These mice displayed impairments in the Schaffer collateral (SC)-CA1 long-term potentiation (LTP) and in spatial learning using the standard Morris water maze. That this mutant mouse is defective in the acquisition rather than the retrieval of the memory was suggested by its inability to form normal memory representations as CA1 place cells. These findings remain the most cogent single evidence for the hypothesis that synaptic plasticity underlies memory.

The standard Morris water maze task tests animals' ability for "reference" memory, which is acquired incrementally over multiple trials and involves information that is constant across trials. Another type of memory supported by the hippocampus is "episodic" memory, which is acquired rapidly with one-trial or one-time experience and involves trial- or event-specific information. It is likely that different mechanisms underlie these two types of declarative memory, although both can be hippocampus-dependent. Little is known, however, about underlying differential mechanisms.

We tested the hypothesis that a recurrent network with robust synaptic modifiability like the one in area CA3 of the hippocampus plays a crucial role in episodic memory by rapidly encoding a novel event. For this purpose, we generated a knockout mouse (CA3-NR1 KO) in which the deletion of the NR1 gene is restricted to the CA3 pyramidal cells of an adult mouse. These mice are impaired in the spatial delayed-matching-to-place (DMP) task, which tests an episodic memory, but are normal when the platform location employed a few days earlier is reused. This behavioral deficit is highly specific, in that the mutants are normal in the acquisition of spatial reference memory. We monitored the activities of the pyramidal cells in CA1, the area downstream of CA3, and the site for the hippocampal output, before and after the animals entered a novel space from a familiar space (a collaboration with Matthew Wilson). The specificity of spatial tuning in the mutants was reduced during the first 15 minutes of exploration in the novel space compared to the same period in the familiar space. In contrast, no space shift–associated change of spatial tuning was observed when the mutant mice were returned one day later to the pair of spaces experienced on the previous day. The spatial tuning of CA1 place cells of control animals did not exhibit any space shift–associated changes. These results suggest that CA3 NRs, most probably those in the recurrent network, play a crucial role in rapid hippocampal encoding of a novel encounter and in one-trial- or one-experience-based rapid learning.

During the past year, the cell-type-restricted (NR) gene deletion approach was extended to the third major hippocampal area, the dentate gyrus (DG). The DG serves as the primary interface between the cortex and the hippocampus, utilizing sparse connectivity and a large number of cells to generate more unique hippocampal representations of activity arriving via the perforant path (PP). To understand the precise role of NR-mediated synaptic plasticity in the DG circuit, we have engineered a mouse line in which the NR1 subunit is deleted only in the DG granule cells (GCs) of adult mice. The mutant mice are defective in the in vivo induction of LTP at the PP-DG synapses. Despite this striking loss of plasticity in the DG, DG-NR1 KO mice are normal in several hippocampal-dependent learning and memory tasks, including the Morris water maze, contextual fear conditioning, and T-maze alteration.

To investigate the effect of the NR1 mutation in more subtle aspects of hippocampus-dependent learning, we subjected these animals to an incremental fear-conditioning task of context discrimination. The DG-NR1 KO mice exhibited a deficit in the early phase of the trials, although their ability to distinguish the contexts developed slowly to the normal level as the trials were repeated. Thus, the mutant mice are normal in spatial and contextual learning per se but have a problem forming distinct memories of similar contexts rapidly, which the control littermates accomplish with no problem. These results suggest that the NMDA receptors in DG granule cells and probably NMDA receptor-dependent synaptic plasticity at the perforant path DG granule cell synapses play an important role in fast (with one or two trials) "pattern separation." We have also obtained corroborative evidence at the level of CA3 place cell activities. Thus, while a given CA3 pyramidal cell of the control mice fires at a rate unique to a particular context, the mutants' CA3 cells fire indiscriminately at similar rates in two contexts.

We have an opportunity to learn something when we encounter something new to us. That is, we cannot acquire new memories by encountering something we already know. This somewhat obvious day-to-day experience must be based on our ability to rapidly compare the online sensory information with the numerous pieces of information we have already stored in the brain network as memory from past experiences. Which part of the hippocampal network is responsible for this novelty detection? One hypothesis is that this is accomplished when CA1 pyramidal cells compare input from the direct entorhinal cortex (EC)→CA1 circuit with input from the CA3 cells thought to be loaded with memory information prestored in the recurrent CA3 network. To test this hypothesis, we have produced a new transgenic mouse line in which the neurotransmitter release from CA3 pyramidal cells can be specifically and exclusively inhibited in a temporally controllable manner. Our behavioral and place cell–recording experiments carried out so far support the comparator role of CA1 pyramidal cells in novelty detection.

Memory Recall
In day-to-day life, recall of associative memory almost always occurs under the constraints of limited cues. For instance, recalling the rich content of interesting conversations with someone can be triggered by the mere subsequent sighting of that person. In the past, a study of the mechanism underlying this fundamental feature of memory recall, referred to as "pattern completion," has been limited to computational modeling. These theoretical studies hypothesized that a recurrent network with modifiable synaptic strength such as that in hippocampal area CA3 could provide this pattern completion capability. We addressed this issue with the CA3-NR1 KO mice. The evoked NMDA currents and LTP were entirely missing specifically at commissural/association (C/A)-CA3 synapses. The mutant mice were normal in the acquisition and retrieval of spatial memory tested in the standard hidden platform version of the Morris water maze. However, when the memory of the location of the hidden platform was tested following removal of three of the four major extramaze cues (partial cue conditions), the mutants exhibited a clear deficit of memory retrieval compared to the control animals.

To investigate the neural mechanisms that might underlie the specific recall deficit, we examined the neurophysiological consequences of the CA3-NR1 deletion by analyzing CA1 place cell activity. We found that spatial information within CA1 is relatively preserved, despite the loss of CA3 NRs, providing a physiological correlate of the intact spatial performance of the CA3-NR1 KO mice in the Morris water maze under full-cue conditions. To investigate the effect of partial-cue removal on CA1 output, we allowed mice to explore a familiar arena for 20 to 30 minutes under full-cue conditions and then removed them to their home cage. Following a 2-hour delay, mice were returned to the arena with either the same four major extramaze cues present (full-cue conditions) or with three of the four cues removed (partial-cue conditions). In the control mice, there were no significant changes in place field properties associated with the change in the cue conditions, while mutant CA1 cells showed significant reduction in spatial tuning properties. These physiological impairments may underlie the inability of mutants to recall the location of the hidden platform when only partial distal cues are available.

This study, along with our previous study with CA1-NR1 KO mice, illustrates the power of cell-type-restricted, adult-onset gene manipulations in the study of molecular, cellular, and neuronal circuit mechanisms underlying cognition. This degree of spatial targeting—not to mention cell-type specificity—is difficult to accomplish by pharmacological manipulation.

This work received support from the National Institute of Neurological Disorders and Stroke, the National Institute of Mental Health, RIKEN (Institute of Physical and Chemical Research, Japan), and the Picower Foundation.

As of June 14, 2007