Memories allow us to survive and adapt in constantly changing environments. Fear memory especially warns us to avoid that jumpy hornet in the garden, or the slithering snake on the hiking trail. These memories aren’t very specific – this is evolutionarily beneficial as it allows us to respond to new but similar threats on the fly.
How specific or generalized a memory is relies on a balance in the neural circuit containing at least the hippocampus and the medial prefrontal cortex (mPFC). There’s still quite a lot of debate in the field, but the hippocampus seem to be critical in maintaining the SPECIFICITY (or the context) of the memory, while the mPFC gates how generalized it becomes. Global inactivation of transmission in the mPFC, for example, leads to overgeneralization of a fear memory. We know that information flows from the mPFC to the hippocampus. But is there anything in the middle that coordinates the activity of these two crucial brain regions?
Wei Xu and Thomas C. Sudhof. 2013. A Neural Circuit for Memory Specificity and Generalization. Science, 339, 1290
To answer this question, the authors first “mapped” projections from the mPFC using an engineered virus called SynaptoTag in the rat. This virus can express fluorescent proteins that “stain” the outgoing projections red, and the synapses those projections form with downstream targets green. The density of green dots illustrates how many synaptic connections are formed within a downstream brain area. mPFC projections turned out to be pretty promiscuous, but one region caught the authors’ eyes – the nucleus reuniens (NR). Anatomically, NR projects to the hippocampus and back to the mPFC, creating a closed loop. Functionally, the NR may be involved in hippocampus-dependent learning and memory. Looks promising – is NR that middle node?
The authors used a very smart approach to answer this. They injected a virus into the mPFC that could POTENTIALLY silence synaptic transmission, and another into the NR that provided a crucial component for the first virus to work. So transmission is stopped ONLY when mPFC projections are synaptically connected to the NR. They then trained infected rats on contextual fear conditioning, where rats learn to fear a sound in their training environment (usually a box with decorations). The authors then tested the rats to see if they learn to fear the tone, the training context, or an altered context similar to the training context. Remarkably, while all rats learned to fear the tone and the training context, rats with their mPFC/NR transmission halted were also scared of the SIMILAR context, meaning they have overgeneralized the fear memory.
To directly look at information gating in the NR, the authors then infected it with two different viruses. One expressed a toxin (TetTox) that silenced information outflow from the NR, while the other eliminated a protein (neuroligin-2) involved in neuronal inhibition, thus strengthening the output of NR (NL2-KD).
As you can see in D, when the viruses were injected BEFORE training, the TetTox group (with NR inhibited) showed markedly enhanced overgeneralization of the fear memory, while the NL2-KD group (with NR enhanced) showed LESS generalization than the control group. All groups responded to the tone and the training context similarly, meaning they learned the original fear association to the same degree (although NR-inhibited group showed less freezing for both tone and training context, but the effect was not significant). Interestingly, this difference disappeared when the rats were first trained then infected. This shows that NR gates memory generalization during the initial encoding of the memory, not retrieval.
As mentioned before, the NR directly projects to the hippocampus, and the hippo is crucial in maintaining memory specificity. So how much of the activity in NR neurons ACTUALLY influence activation of the hippocampus? In other words, is NR a crucial node in the circuitry? One way to look at neuronal activity is to stain for c-Fos, a protein immediately increased after activation. TetTox indeed significantly decreased c-Fos activation in regions of the hippocampus and mPFC associated with contextual fear conditioning, while NL2-KD enhanced the effect of training on c-Fos activation of these two regions. While this says nothing of the OUTCOME of hippocampus activation (i.e. is it mediating memory specificity), it does show there is a functional connection between NR, the hippo and the mPFC.
Finally, the authors showed that the TYPE of NR neuron activity also gates memory generalization. Using optogenetics, the authors forced NR neurons to fire either in a tonic way or in a “pulsing” phasic way WHILE the rats undergo fear conditioning. Surprisingly, tonic excitation inhibited memory generalization, while phasic burst firing promoted overgeneralization. This shows that, unsurprisingly, NR mediates gating in ways more than acting as an on-off switch. It would be extremely interesting to know how NR usually fires in the brain, and if this is changed in disordered states that are associated with memory overgeneralization, such as post-traumatic stress disorder and severe depression.
I love the feats of engineering that made this study possible, as well as the authors’ thorough approach in examining the role of NR, a relatively “indie” brain region, in gating memory generalization. Although the study can’t readily translate into medical therapies, just learning something new and cool about the brain is valuable in itself, isn’t it?
Xu W, & Südhof TC (2013). A neural circuit for memory specificity and generalization. Science (New York, N.Y.), 339 (6125), 1290-5 PMID: 23493706