Where memory traces remain

Where - and how - are memories stored in the brain? While our brain changes throughout the years, strengthening or weakening synapses, birthing or killing neurons, the traces of our most cherished memories survive these physical reconstructions and are recalled with a moment’s thought.

What we now know is that multiple brain regions are responsible for encoding, storing and retrieving memories. One common theory is that long-term memories are encoded by the hippocampus, and transferred to the neocortex for storage later on.  For example, in a popular task called “contextual fear conditioning”, in which rodents link a certain environment with a nasty electrical shock (evidenced by freezing when they’re re-introduced into the same environment), blocking hippocampal activity with drugs immediately after training results in an inability to retain the training, but inhibiting it weeks later does not affect the memory. This type of amnesia is also seen in some human patients with damage to the medial temporal lobe, where the hippocampus is located. However, there are also cases where the hippocampus is extensively damaged, but fear memory is spared. This has led to a conflicting theory, which states that the memory trace may not have transferred , but instead copied to a new location. Previous studies have tried to evaluate both hypotheses with pharmacological agents and genetic manipulations, but are subject to problems like non-specific drug actions, slow time scales and compensatory mechanisms. So the authors in this paper decided to use the cool new toy, optogenetics, look at the contribution of the hippocampus in storing and retrieving long-term (remote) memory.

They first made a transgenic mouse that expressed a light-activated channel in excitatory (CaMKIIalpha) neurons in the CA1 part of the hippocampus, which is known to play an important role in contextual fear conditioning (and called the mice eNpHR3.1 in the figures; I’ll abbreviate to 3.1). When they shine a light directly onto these cells, they inhibit their activity. After verifying their model, the authors trained the mouse with the light on (thus inhibiting hippocampal function), then tested the mouse the next day. Compared to the controls, the eNpHR3.1 mice showed significantly less freezing, indicating that they didn’t remember the training. To show that shining the light didn’t screw the brain up in non-specific ways, the authors retrained these mice with the lights off, and re-tested them the next day. The mice learned fine. There were no differences in the anxiety level or motor abilities between the normal and 3.1 mice. This suggests that hippocampal activity is required for encoding contextual fear memory.

The authors next turned to remote fear memory. After training the mice without inhibiting hippocampal activity, they tested long-term memory 28 days later with the light on. The 3.1 mice showed less freezing than the controls. This means that either the 3.1s didn’t learn the task, forgot the task or simply couldn’t retrieve the task with their hippocampus activity inhibited. So the authors retested them a day later with the light off – and they remembered fine. Even more impressively, the authors showed the same results for ultra-long memories, 9 and 12 weeks after training.  This suggests that the hippocampus has to be activated at the time of retrieval, acting either as an index to the memory stored somewhere else, or retain a copy of the memory itself. However, when the authors used pharmacological agents to inhibit hippocampal function 28 days after training, the mice had no problem remembering the task! This directly contradicts their optogenetics data, so what is going on?

One hypothesis is that – because drugs take time to act and don’t have the temporal specificity of light-induced inhibition- the brain is coming up with compensatory ways to retrieve the memory. The authors tried to mimic drug action by prolonging the timespan they keep the light on – and this abolished the need for hippocampal activation in retrieving long-term memories, just like the drugs. When the mice were retested 24 hours later with a short burst of light as before, they couldn’t remember again. This suggests that when the brain is trying to retrieve a remote memory, it first utilizes the hippocampus; but if something brings down the hippocampus, given enough time, the brain activates a backup system to retrieve the memory trace. So what is this backup system?

So the authors tried to see pick brain areas are activated when the mice try to retrieve long-term memory after a prolonged light episode (thus the hippocampus is shut off, and enough time is given to activate the backup). By staining for the immediate early gene c-fos – a marker for recent activation – the authors found significantly high levels in an area called the anterior cingulate cortex (ACC). While the ACC is well known for its function in mediating rational cognitive behaviour, it has previously been implicated in remote memory storage as well. So the authors targeted the ACC by directing the light there and inhibiting its activity, instead of that of the hippocampus. ACC inhibition a day after training didn’t affect the fear memory, but did 28 days later. So the ACC is not involved in short-term memory retrieval, but is crucial for long-term memory recall. In conclusion, it seems like the hippocampus serves as the default activator for long-term fear memory, but once inhibited, the ACC will take its place and retrieve the memory trace.

This study is extremely cool. Not only does it examine the question of where our memories are stored, it also shows how important both temporal and special manipulations are in probing neural functions. It supports the theory that multiple traces of the same memory remain in our brains over time, sheds light on which brain areas may be involved at different stages after encoding. It also shows that memory traces can be supressed by specific manipulation of a single brain region. While far off from clinical use, the methods presented in this paper might serve as the basis to examine the role of specific neuronal populations under normal and pathological (such as post-traumatic stress disorder) conditions.

Goshen et al (2011). Dynamics of Retrieval Strategies for Remote Memories Cell DOI: 10.1016/j.cell.2011.09.033