The tug-of-war of memory: pattern completion and pattern separation in the brain

Thumbnail image adapted from : https://www.tennessean.com/story/news/local/williamson/2018/06/15/two-kittens-better-than-one-animal-center-pet-adoption/706919002/

Have you ever heard someone humming a portion of a tune and immediately started singing along with them? This sort of memory recall is what neuroscientists call pattern completion. As the name suggests, pattern completion happens when we take one or a few details and use them to construct a complete memory. While completing patterns is important for any functional memory system, an equally important process is separating patterns. Pattern separation is what lets you differentiate memories. For instance, you probably have little difficulty distinguishing between the first and last days of class, despite extensive similarities between the two experiences (maybe they took place at the same school, perhaps in the same classrooms, with the same teacher and students, etc.).

Figure 1: Schematic of the mouse hippocampus. Dentate gyrus connects to CA3, which connects to CA1. Each region is hypothesized to perform a different task: DG may perform pattern separation, CA3 may perform pattern completion, and CA1 may consolidate these reports and send them off for a final decision. (Adapted from De Michele 2015.)

Both pattern completion and separation are believed to take place in the hippocampus, a brain region that plays an important role in memory formation, and different parts of the hippocampus are thought to specialize in each process. In particular, it is thought that CA3 is responsible for pattern completion, while dentate gyrus (DG) is responsible for pattern separation (Figure 1). The two regions have something akin to a tug-of-war relationship, where CA3 says “I’ve seen this before,” and DG says “this is different from what I’ve seen.” The final arbiter is the output region of the hippocampus, known as CA1. CA1 combines the processed information from the two brain regions and sends information to the rest of the brain to make appropriate decisions.

The idea of a tug-of-war between pattern completion in CA3 and pattern separation in DG makes for a clean story, but is it accurate? In a recent paper from the Pasteur Institute in France, Allegra and colleagues test the specific hypothesis that DG is responsible for pattern separation. How does one go about testing such a hypothesis? The authors used a combination of two techniques. First, they performed calcium imaging in the dentate gyrus. Calcium imaging is a method that labels particular neurons so that they light up every time they send out a signal. The stronger the signal, the brighter the labeled cells glow. Second, they used virtual reality to precisely control how different any two environments are. With this setup, the scientists could perform the following simple and elegant experiment. First, they trained mice in one virtual reality environment (the “familiar environment”, Figure 2). After many days in that environment, the mice were introduced to the “novel environment,” which featured a small change to the walls and the location where the mice could get a reward. Throughout the experiment, they performed calcium imaging over the DG. Would these DG cells engage in pattern separation by distinguishing between the two similar environments?

Figure 2: Familiar and novel environments from Allegra et al., 2019. Note the different stripe orientations and reward locations. These environments should be distinguishable via pattern separation.

It seems like the short answer is “yes”! When the mice entered a new, slightly different environment, most of the DG neurons that were active in the old environment stopped signaling, and a nearly completely new group of neurons took up the signal. Among those few cells that were active in both environments, most of these neurons changed where they were active. In other words, a given cell was active at one location in the familiar environment, and at a new location in the novel environment. This is consistent with the hypothesis that DG is responsible for pattern separation, because different experiences resulted in different neural activity.

All in all, Allegra and her colleagues provide tantalizing evidence in support of a pattern separation role for DG. This is exciting work that brings us one step closer to understanding how the brain encodes memories. The next step would be to learn about how pattern completion works in the brain by doing a similar experiment in CA3. With these two pieces, we’d have a pretty good overview of how the memory system works in the brain! This understanding would teach us about how our brains normally function, and could also provide insight into rarer instances of super-charged memory, like photographic memory, or memory failures, like dementia.

Read the source article here: https://www.biorxiv.org/content/10.1101/868794v1.full

Edited by: Isabel Low

References

Allegra, M., Posani, L., & Schmidt-Hieber, C. (2019). The hippocampus as a perceptual map: neuronal and behavioral discrimination during memory encoding. bioRxiv. doi: 10.1101/868794

De Michele, P. (2015). Analysis, tuning and implementation of neuronal models simulating hippocampus dynamics. PhD Thesis.