Engram Neurons: A New Take on Memory Consolidation

As far back as Plato and Aristotle, people believed that our memories had to be physical somethings that were stored somewhere in the brain. But only in modern times have we learned much about what this something is. First, the something was given a name: memory engram. Then, as knowledge accumulated about what happens in neurons and their synapses as they become active in learning and remembering, it became clear that learning events that could be remembered were causing chemical and physical changes in the junctions (synapses) between neurons that participate in the learning experience. Participating neurons grow new dendritic branches (called spines) and the synapses on those spines enlarge and their neurotransmitter systems become enhanced. These changes constitute the engram. Post-learning reactivation of the synapses holding such an engram can produce recall of the original learning that created the engram.

In the early days of neuroscience, scientists believed that learning experiences assigned or recruited certain parts of the brain to hold the memory. An experimenter, Karl Lashley, taught certain tasks to lab animals, and then under anesthesia, destroyed different parts of the neocortex in the hopes of finding where the memory was stored. He couldn’t find any particular storage location. What he did find was that the more extensive he made the cortical lesions, the more likely he could erase the memory. In other words, memory of a given experience seemed to be deconstructed and parceled out into different regions.

Then came quantitative EEG studies by E. Roy John, in which he tracked the location of brain electrical evoked responses in different parts of the cortex during learning experiences. He saw that a given learning experience would produce electrical responses in several parts of the cortex, again suggesting a deconstruction and distribution of memory engrams. This led him to famously proclaim, “Memory is not a thing in a place, but a process in a population.”  Well, we know that this is a bit of over-statement. There is such a thing as a memory engram that is stored in specific places. Nonetheless, there is a distribution process for creating the engram in multiple locations and for orchestrating them into simultaneous and coordinated activity during recall of the memory.

Modern genetic engineering and neuron staining technology provide powerful new tools to examine neurons that participate in joining the neural circuits involved engrams. There are now ways to image and manipulate engrams at the level of neuronal ensembles.  Several lines of evidence show that engram neurons can be seen histologically and evaluated under various experimental approaches. For example, histological stains revealing neurons that are activated by a learning experience show that they are also active during memory retrieval of that experience. Second, loss-of-function studies show that impairing engram neuron function after an experience impairs subsequent memory retrieval. Third, studies show that memory retrieval can be triggered by optogenic stimulation of engram neurons in the absence of any natural sensory retrieval cues.

The basic approach used by investigators in the lab of Susumu Tonegawa was to teach mice to avoid walking into a chamber in which they would receive a mild electric shock. Neurons that are activated by this fear conditioning fluoresce in immunohistological stains of brain slices in mice that are sacrificed at various times after learning reveal a memory engram that resides in selected neurons in the amgydala (which processes fear information), in the hippocampus (which converts short-term memory to longer-term memory), and in multiple regions of neocortex (which holds long-term memory in the form of enhanced synaptic capability). Some of these cells still fluoresce when examined many days later, indicating that they have become part of an ensemble of engram neurons that hold a relatively lasting representation of the original learned experience.

Other mice were genetically engineered so that engram cells would fluoresce and be activated when exposed to light delivered via micro-fiber optic cables surgically implanted in various regions of neocortex. Such light stimulation of engram cells confirmed their engram status, because light stimulation alone triggered the previously learned behavior (freezing in place, rather than entering the shock chamber). A key finding was that engram neurons in the prefrontal cortex were “silent” soon after learning — they could initiate freezing behavior when artificially activated by light delivered via surgically implanted fiber optic filaments, but they did not fire during natural memory recall. In other words, the memory engram was formed right away in all three places (amygdala, hippocampus, and neocortex), but the engram cells in the neocortex had to mature over time to become fully functional.

Over the next two weeks, the engram neurons in the neocortex gradually matured, as reflected by changes in their anatomy and physiological activity. By the end of that same period, the hippocampal engram cells became silent and were no longer used for natural recall. At this point, the mice could recall the event naturally, without activation of neocortical cells by fiber-optic light. However, traces of the memory remained in the hippocampus, because reactivating those hippocampal neurons with light prompted the animals to freeze.

The past prevailing view was that learning experiences are temporarily held in circuits in the hippocampus and then later exported out to other parts of brain for final storage. Both in the past and now, all the evidence indicates that the hippocampus is crucial for forming lasting memories of experiences that do not involve motor learning, but the mechanisms had been uncertain. 

Neuroscientists did know that long-term memories were stored outside of the hippocampus, because people with hippocampal damage can lose the ability to form new long-term memories, but they are still able to recall old memories.

Now, the new research suggests that memory engrams are not transported from hippocampus to neocortex but are present in both places at the outset of learning. The memory engram in the neocortex just requires maturation for the memory to become more permanent. Moreover, the hippocampus cannot, and need not, hold long-lasting engrams.

Though this is a new way to think about the mechanisms of how temporary memories consolidate into longer-lasting ones, the conventional idea of consolidation remains confirmed. That is, the memory engram must mature over time in the form of biochemical and anatomical changes in the engram cells. Obviously, such maturation process would be disrupted if those same engram cells are recruited to serve other learning purposes before they have finished their maturation as a specific memory engram. This also helps to explain why subsequent rehearsals help make memories last longer, because each rehearsal re-engages engram neurons into the same kind of activity they performed during learning, thus strengthening the relevant synapses.

Once memories were formed in the fear-conditioned mice, the engram cells in the amygdala remained unchanged throughout the course of the experiment. Those cells, which are necessary to evoke the emotions linked with specific memories, like fear of entering the shock chamber in this case, communicate with engram cells in both the hippocampus and the prefrontal cortex.

We don’t know what happens to memory-specific engram cells in the hippocampus. Maybe as they gradually lose their engram status, they become available for processing new kinds of learning experience. Perhaps some traces of engram remain in hippocampus and are accessible for reactivation if highly relevant inputs are received, as could be the case with strong memory cues. Perhaps the important point is that these new techniques for labeling engram cells open the door for new ways to study of memory retrieval, the long-neglected aspect of memory mechanisms.

Another potentially relevant finding of this kind of research is that memory engrams may become damaged but may still exist in a form that cannot be retrieved by natural means. The fact that such “silent” engrams can be retrieved with direct optogenetic stimulation indicates that failures to recall do not necessarily indicate that the memory is lost. The problem may lie in an inadequacy of the natural memory cues used to triggger memory retrieval.

The door is also now open for experiments that might advance our understanding of the maturation of engram neurons in the neocortex. What is known so far is that maturation requires initial communication with engram cells in the hippocampus. Disrupting hippocampal connections between hippocampus and frontal cortex prevents the maturation of neocortical engram cells.

Sources

Takashi Kitamura, Takashi, et al., (2017). Engrams and circuits crucial for systems consolidation of a memory,” Science, 356(6333), 73-78; DOI: 10.1126/science.aam6808

Josselyn, Sheena A., and Tonegawa, Susumu (2020). Memory engrams: Recalling the past and imagining the future. Science. 367 (6473), eaaw4325. Doi: 10.1126science.aaw4325. https://science.sciencemag.org/content/367/6473/eaaw4325

How Does Learning Change the Brain?

Learning programs the brain. It is nature’s way to create simultaneously both “hardware” and “software” for the brain. Neuroscientists have long known that learning experiences change the functional circuitry that is used to process and remember a given learning event. The circuit change is anatomical: electron microscope photographs reveal that the synaptic changes take the form of little blebs located on dendrites. These blebs are called “dendritic spines,” and their size and number change in response to learning and memory formation. You might think of the induced change as a physical location for information storage. What is stored at any given spine is an increase in the probability that the spine will participate in activations that generate recall of the stored memory from all the spines that were enhanced in creating the memory.

Functionally, the change operates as a template that resides more or less permanently that is available not only to recall the original learning event but to respond to similar events in the future. Of course, no one synapse accounts for these recall and programming effects. However, the learning and memory results from the collective enhancement of all the enhanced dendritic spines in the participating circuitry. The templates thus created provide a way for the brain to program itself for future capabilities. Learning how to solve one kind of task makes it easier to learn new tasks that are similar. This is the basis for the so-called “learning-set,” a concept introduced many decades ago by Harry Harlow.

We tend to think about such matters in the education of children. However, the principle applies at all ages, including seniors. The aging brain responds to learning the same way a child’s brain does: it grows new task-specific synapses that can be recruited for other uses.

The learning effect is manifest in the growth of existing synapses and the formation of new synapses. In the absence of mental stimulation, the spines degenerate. Indeed, a typical effect of aging is that the brain actually shrinks as a consequence of the cumulative shrinkage of spines. Many neurological diseases are characterized by reduced synaptic density: schizophrenia, autism, and dementia. An opposite problem occurs with drug addiction, where certain synaptic connections are strengthened by the drug, essentially creating a way to store a memory for the addiction.

Until now, we have not learned as much about the chemical changes that occur with learning. A recent study from Thomas Jefferson University reveals that new patterns of molecular organization develop as connections between neurons strengthen during learning. The researchers basically asked the question, “What does learning look like at the molecular level? Using super-resolution live-cell microscopy, the researchers confirmed the enlargement of synapses previously reported by others. But they also saw that that during a learning experience the molecules involved in sending and receiving the signals between neurons appeared to be organized in clumps or "nanomodules" that both dance and multiply when stimulated by learning-like signals.

The researchers developed a novel technique wherein they could visualize the chemical involved in transmitting signal from one neuron to another. Chemicals on the pre-synaptic side appeared green and those on the receiving postsynaptic sign appeared as red. Then while observing the colors, they observed live neurons in real time as they sent signals to each other via their neurotransmitter chemical systems. The color changes indicated that during signaling activity, the presynaptic chemicals clumped together and bound to clumped molecules on the postsynaptic side. The clumps appeared to have a uniform size. When the presynaptic neurons were stimulated in a way that promoted spine enlargement, they saw that the number of chemical clumps increased. Such stimulation caused non-moving clump to jiggle and move around the synaptic spine, with pre- and post-synaptic chemicals moving in lock-step. Maybe this jiggling helps to trigger the biochemical cascade that neurotransmission causes to change activity in the post-synaptic neuron.

Perhaps we don’t think enough about the movement of molecules in living tissue. All chemicals in solution bounce around randomly. Clumping together and jiggling in lock-step creates new ways for chemicals to produce their effects. Apparently, the clumping is the triggering event for the enlargement of dendritic spines that creates a structural basis for memory.

This kind of chemical interaction is not only relevant to learning. It also applies in unknown ways to the altered function in addictions and other neurological diseases in which strong interneuronal connections become too strong. Research on neurotransmitter clumping in disease states has not yet begun, but here we may find clues on how to treat some of these conditions. Clearly, disrupting the clumping would disrupt the ability to strengthen synapses. While we want our synapses strengthened to promote normal learning, we don’t want this to happen for example, for opiate pain relievers, which create addiction. At the molecular level, addiction is a learned condition.

Sources

Hruska, M. et al., "Synaptic nanomodules underlie the organization and plasticity of spine synapses," Nature Neuroscience, DOI: 10.1038/s41593-018-0138-9, 2018.

Klemm, W. R. Core Ideas in Neuroscience. https://www.smashwords.com/books/view/390780