Almost everyone knows how it feels to remember things for decades on end – and also how it feels to forget where you left your car keys. The answer to this strange paradox could be simple, according to a new study from Caltech. The researchers found that that the brain recruits a higher number of neurons which fire impulses in a synchronized manner to form and store strong memories over long periods.
Diagrams of neural activity in the hippocampus, recorded from a mouse as it learned about new surroundings. Colors correspond to unique locations within the new place. Over time and continued exposure to the arena, the mouse forms stable memories by recruiting teams of neurons to encode for the location. Credit: Caltech
The key word here is “redundancy”, which means the brain has backup to remember the exact occurrence, even if one or a few of the neurons involved fail or fall silent. The study adds the crucial bit of knowledge that this is achieved by forming stable groups of neurons with synchronous firing to represent an environment, even while individual neurons showed fluctuations from day to day. This is important in adding to our understanding of how memory works, and could help treat memory loss in conditions that affect the brain, like Alzheimer’s disease or strokes.
How was the study done?
Previous research has shown that the hippocampal region of the brain is responsible for forming memories, with different cells having the responsibility to store memories of space, time and other cues. In order to study these long-term, researchers in the current study made use of chronic implants with high sensitivity to pick up hippocampal activity, and software that could reliably store and analyze the recordings over time.
The equipment was used to detect and record activity from the CA1 group of pyramidal neurons over three sessions. The first 5 sessions were for “learning,” where mutant mice were removed from their home cage and were exposed to a new linear track 1.5 m long, and 12 cm wide, with white walls 15 cm high. Three groups of black lines in different shapes were drawn on the walls on both sides to provide space cues. Sugar water (a reward) was dispensed from devices at both ends, first when the mouse reached that end, as signaled by an infra-red sensor, and later 5 seconds after the mouse reached, to test for delayed reward recognition.
The second type of session was the “re-exposure” sessions after a 10-day gap. The third was the “damage” sessions, following hippocampal lesions induced by light.
The neuronal activity was observed for a long period, up to 35 days. The mice first wandered aimlessly until it got the reward by reaching the end of the maze. In this period, single neurons fired when it noticed one of the visual cues, but later on, multiple neurons began to fire in synchrony. This showed the evolution of recognition of each location with respect to the visual cues.
On re-exposure, mice which showed more synchronized neurons firing remembered the task more quickly compared to those which had fewer neurons.
What were the results?
The study showed that almost 90% of CA1 neurons were active in each session, and about half showed consistent activity across all sessions. About 95% neurons were active at home and in the new environment. However, between different environments (cage vs track) or different days in the same track, only about 40% of neurons retained the same response to the same stimulus on the second day. The field representation declined sharply from day to day, but only by small amounts thereafter.
Two types of cells were studied, those which were activated at particular locations of the maze, or “place cells”, and those which were activated when the mice were immobile, called “time cells”. After the “learning” phase, the more familiar the animal became with its task, the less the neural representations changed over time. Overall, the same set of CA1 neurons remained active from day to day.
The changes in field representation might be due to minor differences in the environment of the training site with each repeated exposure. Field changes occurred more often in the familiarization fields, that is, the first five days of “learning” and the first two days of “re-exposure.”
Researcher Carlos Lois says, “The more you practice an action, the higher the number of neurons that are encoding the action. The conventional theories about memory storage postulate that making a memory more stable requires the strengthening of the connections to an individual neuron. Our results suggest that increasing the number of neurons that encode the same memory enables the memory to persist for longer.”
Following light-induced brain damage, the first day post-lesion showed a burst of direction-specific abnormal firing like that seen between seizures. Most of the neurons were involved in these episodes.
At such times, a higher number of CA1 cells recorded time and place, but different neurons were involved on two consecutive days. This disruption ceased over 2 to 10 days following the lesion, and the time and place fields reverted stably to pre-lesion representations in a high percentage of cases.
What do we learn?
The study suggests that synchronized firing of neuronal groups is responsible for long-term stable memory. Such hippocampal cell assemblies responsive to time and space field cues are known to exist.
The study showed that as neurons begin to act in synchrony, the mice became familiar with the learning track, primarily due to the firing of time and place cells. The synchronized neurons typically acted together, whether right or wrong, which led to the hypothesis that such synchronization was the basis of stable representation of the occurrence over time.
Different neuron groups encoded information specific to the environment. Also, individual nerve cells responded quickly, within minutes, but cell groups took 2-4 days to develop. Subsequently, these cell groups retained the information relating to the task over time, even when not continuously exposed to that environment. Thus stable memories of the task are formed within CA1 cell groups even as individual neurons show fluctuating activity.
Using a graphical representation of neuronal activity, the investigators confirmed earlier research which showed that analysis of CA1 neurons can predict animal behavior and interactions with the environment. The graph can also tell how responsive the neuron will be over longer periods.
They were also able to show that neuronal graphs can predict the neuron field, whether time, place, or neither (even without access to any information about the animal’s behavior), and that these signatures do not vary between animals.
While individual neurons retained information for about 10 consecutive days, cell groups encoded field representations in a stable manner for 35 days, which was the longest period analyzed. These groups add to the information provided by individual neurons, thus forming a more detailed and complete memory. The presence of the neuron in the cell group added to the chance that it would continue to be responsive to the field in the future as well. Thus individual neurons showed shifts in their patterns of firing and field representations over time, but the group continued to show synchronous activity, which kept the memory stable and durable. The higher the number of such groups, the stronger the memory.
In other words, memory loss could be due to the involvement of fewer neurons in forming the memory. Thus conditions like Alzheimer’s, and even the memory loss that occurs with aging, could one day be treated by increasing the number of neurons to form more stable memories.
The study was published in the journal Science on August 23, 2019.
Persistence of neuronal representations through time and damage in the hippocampus. Walter G. Gonzalez, Hanwen Zhang, Anna Harutyunyan, and Carlos Lois. Science. Vol. 365, Issue 6455, pp. 821-825DOI: 10.1126/science.aav9199, https://science.sciencemag.org/content/365/6455/821.full