Cell . . . The neurobiological basis behind the "one-on-one" and "local conditions" of memory.

The memories in our brains record our past and dominate our present - in a way, everyone's memory defines who we are.however, how are these total Zero Zero memories stored in the brain? What is the material basis of memory in the brain? Neuroscientists have recently discovered that each particular memory is stored in a group of corresponding brain cells.for example, in the dentate gyrus (DG) of the hippocampus, each memory segment is activated and eventually stored in 1-5% of the cells.these 1-5% of cells are almost randomly selected, and different combinations of cells allow us to store thousands of different memories.using photogenetic / chemical genetic methods, scientists have shown that artificially activating specific 1-5% cells can immediately recall a certain memory segment in mice; conversely, inhibiting this group of cells can block memory.these studies show that studying these 1-5% cells is the key to further understanding the neural basis of memory.however, the key question that has not been solved is whether these 1-5% of cells have the same function or have division of labor? In addition, what changes have taken place in these cells at the molecular, cellular and loop levels to enable memory to be written into the cell? On March 17, 2020, the research group of Lin Yingxi of MIT / Upstate Medical College of New York (the first author is sun Xiaochen) published a study entitled "functionally distinct neural assemblies within the memory engram" in the journal Cell.this work found that memory storage cells have different division of labor; among them, one group of cells makes the stored memory more easily extended to other situations (memory generalization, similar to "infer from one instance"), while another group of cells makes the memory less likely to be confused and can be properly used in specific situations (memory) Discrimination, memory discrimination, similar to "adjusting measures to local conditions").in the study, researchers used chemical genetics to inhibit two groups of cells in the dentate gyrus (DG) of mice, which are responsible for storing fear memory: FOS signaling pathway was activated in one group, and npas4 signaling pathway was activated in another group.Fos and npas4 are two genes closely related to learning and memory, both of which encode transcription factors and mediate downstream transcription pathways.when the cells corresponding to npas4 were inhibited, the mice could not distinguish the environment in which they had been shocked (context a) and a similar environment (context b); on the contrary, when the cells corresponding to FOS were inhibited, the response of mice to the two environments became more different.this shows that Fos and npas4 cells, which store the same memory segment, have opposite functions: FOS cells make memory generalization, and make the response of mice to similar environment more similar; while npas4 cells promote memory discrimination, which makes the response of mice to similar environment more different.so, is memory more "generalized" or better "discriminative"? In fact, the "generalization" and "discrimination" of memory are just like "Yin" and "Yang" in Taiji. They are in harmony and indispensable.the "generalization" of memory helps us to "learn by analogy" in our study and draw inferences from one instance. However, excessive generalization can cause the dilemma of "once bitten by a snake, afraid of the well rope for ten years", and even lead to anxiety disorder, post-traumatic stress disorder (PTSD) and other psychological diseases.similarly, appropriate "memory discrimination" enables us to clearly distinguish the different situations in memory, so as to adapt to local conditions; excessive "discrimination" is harmful, such as "overfitting" in the field of machine learning.therefore, "generalization" and "discrimination" need to be balanced.and the brain chooses two different groups of cells to be responsible for the generalization and discrimination of memory, which may be a clever design to balance the two. so, where does the division of labor between FOS and npas4 come from? Using electrophysiology, calcium imaging, molecular biology and other means, the researchers further found that these two groups of cells have different molecular, cell and loop characteristics. a key difference between the two is that FOS cells specifically receive stronger excitatory signals from the upstream medial entorhinal cortex (MEC), while npas4 cells receive stronger inhibitory signals from CCK + inhibitory neurons in the dentate gyrus. this suggests that the medial entorhinal cortex and CCK + inhibitory neurons may be the key to promote the difference in the function of these two groups of cells. therefore, this study further revealed that medial entorhinal cortex also promoted memory generalization, which was consistent with FOS cell function; CCK + inhibitory neurons promoted memory discrimination, which was consistent with the function of npas4 cells. therefore, the researchers believe that Fos and npas4 cells interact with different brain regions in the process of memory formation, leading to different division of labor. in conclusion, what does this study tell us? There are two main lessons we have learned: 1) memory is stored in specific cells, and these cells have different functions (generalization vs. discrimination). 2) the different functions of cells come from the different changes of molecules, cells and circuits. of course, these findings also raise many new questions: are there any other cell groups besides Fos and npas4 cells? Is the division of labor found in other brain regions other than hippocampus? These follow-up questions are under active research. we believe that further understanding the neural mechanism of memory is not only the basis for understanding many advanced cognitive functions, but also can provide practical reference for the design of artificial intelligence and machine learning algorithms in the future. What are the implications for our understanding of memory? One of the important implications, the researchers believe, is that the brain encodes memories more complex than we think. different functions of cells encoding specific memory may also contain different memory information. a memory is very complicated. For example, the simple memory of "I went downstairs to buy coffee yesterday" contains a series of information such as time, place, scene, characters, abstract concepts, etc. how does the brain store this information in different cells orderly and efficiently? What are the benefits of this? Similar questions need to be solved in the future. another interesting revelation is that we have to generalize and distinguish the same memory from time to time, which just shows that the purpose of memory is not to record facts objectively. Bi Shumin wrote in the female psychologist that "memory is the servant of the soul, not the real clerk". from an evolutionary point of view, the purpose of memory storage is to let it guide our future behavior. the ancestors who first learned to store memory were able to remember the location of food, the habits of natural enemies and new skills, which was a revolutionary progress in evolution. and it is these ancient learning and memory mechanisms that have been used in the brain to help human beings survive in modern society. remarks of the first author: 1. "How the brain stores memory" is not only my doctoral research topic, but also the fundamental motivation that initially prompted me to enter the field of neuroscience. A clear picture in my memory was in a neuroscience class in the afternoon of the fifth education in Tsinghua University in my sophomore year. I was puzzled about the way the brain stores memory. It is not an accident to select the cells related to Fos and npas4 pathways; the laboratory has focused on the study of immediate early genes closely related to learning and memory for many years. Interested students can refer to ramamoorhi et al., 2011, science and Wen et al., 2018, We would like to express our sincere thanks to Professor Polina anikeeva of MIT and Rao Siyuan, Professor Zhang Xiaohui of Beijing Normal University and Yao Li, doctoral student. original link: