Memories are thought to be stored as enduring physical changes in ensembles of neurons, ensuring that activity patterns present at the time of encoding are recapitulated at the time of retrieval. But which neurons become part of this memory trace? Is allocation a random or non-random process? Our previous studies suggest the latter. In these studies we focused in particular on how fear memories are encoded in the amygdala. In our fear conditioning experiments, mice learn to associate a tone with a footshock and this association depends critically on neurons in the lateral amygdala (LA) . We found that manipulating the levels of the transcription factor cAMP responsive element binding (CREB) within this population of LA neurons influenced the likelihood of a neuron becoming part of a fear memory trace.
In this new study we explored the mechanisms underlying this CREB-mediated allocation. Previous studies had shown that one consequence of CREB over-expression is that cells become more excitable, and here we found that artificially increasing neuronal excitability produced exactly the same pattern of results as over-expressing CREB– more excitable neurons were more likely to become part of the memory trace. Moreover, co-expressing a factor that blocked these CREB-induced changes in excitability blocked these effects. Together these results suggest that relative levels of excitability determine which populations of cells encode the memory. The paper is published in Neuron, and Sheena and Adelaide explain all here:
Our knowledge of the world is based on the sum of our experiences, but it is not clear how multiple, distinct experiences are combined by the brain. Neuroscientists have hypothesized that the brain amalgamates different memories based on their commonalities, or patterns, long after the memories are formed, and possibly while we sleep (David Marr called this process the identification of statistical regularities across multiple experiences). This leads to the prediction that as time passes we lose the details from specific memories, but we gain a more accurate understanding of our cumulative experience.
To test this prediction, Blake Richards designed an experiment to measure how well mice remembered specific experiences versus a general pattern of experience. In these experiments, we trained the mice over many days to find a hidden platform in a pool of water. The location of this platform changed each day, though, according to a pattern, such that some locations were more likely than others. We then tested the mice by putting them in the pool without the platform. We found that if we tested them one day after training they spent most of their time searching in the last platform location, but if we gave them thirty days to rest and consolidate their memories, they spent more time searching for the more likely locations. Our study shows that the brain can ‘combine’ multiple memories even when an animal is not learning anything new and is simply resting.
Form the lab Frances Xia, Adam Santoro and Jana Husse contributed to this study. The paper is published in Nature Neuroscience, and a pdf is available here.
Last week the Keystone meeting on Adult Neurogenesis took place in Stockholm, Sweden. Bringing together leading neurogenesists from around the world, discussion focused on pattern separation (and pattern separation-like) processing in the dentate, neural stem cell lineage, human neurogenesis and even some forgetting. Looking like they are part of some weird parabiosis experiment, below are Rene Hen and Alejandro Schinder. More pictures from the meeting are posted here…