Introduction
Memory is an emergent process of the brain, phylogenetically conserved, that allows us to retain biologically relevant information obtained from the environment for survival. In addition, it provides humans with the foundation to maintain a coherent relationship with our social reality and is the basis of our individuality1. Based on its content, memory is classified into two types: Explicit (episodic and semantic) and implicit memory (priming, skills, habits, reflexes, etc.). The former is the type most affected in diseases characterized by pathological forgetting, such as Alzheimer’s disease. As for the latter or non-declarative memory, it is the last to be affected in Alzheimer’s disease, as the structures supporting this type of memory do not degenerate until very advanced stages of the disease. On the other hand, when considering its temporality, memory is divided into short-term memory and long-term memory. The latter requires a process called consolidation, where information, which is labile and temporary during acquisition, becomes stable and persistent over time2. At present, two types of information consolidation have been described, both necessary for storing information in the brain in the long term:
Structural consolidation
This occurs after the process known as long-term potentiation (LTP) or long-term depression (LTD) (each involving different molecular mechanisms). Structural consolidation begins in the brain structure responsible for encoding a specific type of information3.
Systems consolidation
Like the former, it requires cellular, molecular, and structural mechanisms but involves subcortical connections with various cortical regions (which will be the final relays of the information)3.
Molecular loops
For both types of consolidation to occur, the reactivation (reverberation) of the same group of neurons initially activated for encoding is required4. In this regard, it has been demonstrated that protein synthesis is essential for molecular and cellular reactivation, enabling long-term information storage5. This occurs to maintain cellular metabolism and prolong neuronal communication, resulting in subsequent structural changes. Pharmacological blocking of long-term “molecular reactivations” prevents memory persistence and structural change6,7. Many authors have confirmed that dopamine activation loops, for example, are fundamental for memory persistence7,8. Wang et al. demonstrated that the reactivation of the calcium-calmodulin kinase II protein and NMDA-type glutamatergic receptors is also important9. In addition, the reactivation of proteins in the mitogen-activated protein kinase (MAPK) pathway and the transcription factor CREB during sleep has been observed10.
Structural consolidation
At the structural level, it has been proposed that morphological changes in neurons (shape, size, and number of dendrites) are required for the establishment of long-term memories11. To date, it is known that dendrites, and primarily their spines, are the sites of greatest excitatory contact in a neuron12; moreover, their intrinsic properties allow them to be recruited to form part of the memory trace or engram13. Semon proposed that experience in organisms activates a group of neurons, which undergo persistent chemical and/or physical changes that lead them to become an engram14. Therefre, it has been observed that the intrinsic electrical properties of neurons allow them to be selected in the neuronal group that will form the engram, whose formation depends on experiences15. The formation of these groups also requires molecular mechanisms, such as the participation of early-expression genes16. For the structural changes induced by learning to be maintained in neurons, increases in actin in the cytoskeleton are required17.
Similarly, studies have been conducted in different subcortical and cortical regions, confirming the existence of anatomical changes in dendrites and dendritic spines underlying different types of learning. In fear conditioning, for example, microstructural morphological changes have been observed in the amygdala18, or in the hippocampus following spatial memory19. In cortical areas, it has been observed that after a motor task and its repetition, dendritic spines increase in the motor cortex20. Furthermore, regarding the cellular mechanisms underlying these structural changes, both the induction of LTP21 and LTD22 can produce microstructural changes in neurons.
Systems consolidation and engrams
The theory of the physical state of memory was initially proposed in “The Physical Theory of Human Memory” by Semon and Simon in 1921. The engram law proposed by these authors postulates that changes (which form the engram) occur after the encoding of a stimulus. Semon proposed that experience in organisms activates a group of neurons, which undergo persistent chemical and/or physical changes that lead them to become an engram. Thus, the reactivation of the engram through signals from the specific moment of the experience would induce memory retrieval. In addition, in their work, they define ecphory as the activation by some influence of the engram or silent memory trace (which we now call retrieval or spontaneous recall)23. At present, various studies have delved into the study of memory persistence, especially regarding how and where long-term memory is stored. Canadian neuropsychologist Brenda Milner studied the case of H.M., a patient who underwent surgical removal of the hippocampus bilaterally. The results showed that patient H.M. could not store recent information. This situation was termed anterograde amnesia24. This study paved the way for several studies that corroborated that the hippocampus houses the temporal repertoire of information. Subsequently, it was postulated that memory is stored, according to its temporality, in two timeframes: (1) short-term memory, which appears at the beginning and is dependent on the hippocampus; (2) long-term memory, which occurs over time and repetitions and becomes relatively independent of the hippocampus and dependent on the cerebral cortex, leading to systems consolidation25. Later, we will see how hippocampal neurogenesis continues to play a role in the retrieval of information already stored in the cortex.
To recap, engrams refer to the structural and/or chemical changes that persist after learning and underlie the memories formed after such learning26. In this regard, the formation of local engrams-complex or neuronal group-will require the reactivation of this group of cells in subcortical structures26,27. This ensures the storage of short-term memory. Subsequently, continuous and reverberant communication with the cerebral cortex will allow structural reorganization, leading to the formation of the cortical engram, which is nothing more than the memory trace that will form in different cortical areas depending on the type of memory to be stored3,28 and will be “independent” of subcortical regions to persist and give rise to long-term memory3. In addition, it has been reported that the memory trace is not a simple process; it apparently requires not just one engram but a network of engrams distributed across various parts of the brain27.
Sleep, consolidation, and the engram
As mentioned earlier, the systems consolidation hypothesis supports the existence of continuous communication between the hippocampus and various cortical areas14. In early studies on the topic, the relationship between sleep and memory initially focused on the hypothesis that memories were formed during REM (paradoxical) sleep. At present, it has been proposed that memory formation also occurs during slow-wave sleep29. It has been demonstrated that the visual cortex and hippocampus remain in constant communication during sleep30. Furthermore, it has recently been shown that during sleep, information is processed by category, which is how it becomes consolidated31. Finally, it has been observed that sleep facilitates structural remodeling in the hippocampus after learning and promotes the reactivation of the engram32.
Neurogenesis and the engram
Neurogenesis is the ability of the nervous system to produce new neurons. In the adult brain, new neurons are produced in the dentate gyrus and the subventricular zone of the hippocampus. These new neurons exhibit high excitability, a condition that facilitates their recruitment to form an engram33. The formation of the engram also requires the participation of interneurons34. In addition, it has been observed that new cells provide protection against information interference35. Interestingly, studies in rodent brains have shown that increasing the rate of neurogenesis in an Alzheimer’s model restores memory36. This suggests that new neurons may facilitate information retrieval through a mechanism involving a hippocampus-cortex engram network.
The prediction error (PE) theory and the engram
Our brain enables us, as organisms, to adapt to the environment in which we operate at any given moment. To meet this challenge, we need to predict the future in very short timeframes based on the regularities (familiarity) and irregularities (unfamiliarity) of the environment perceived by the sensory system during encoding. The PE model is based on the hypothesis that, when a prediction is not met, neural mechanisms are activated to restore homeostasis37. Therefore, perceptual states, motivations, and behaviors of an organism can be redirected38. Information is updated by the hippocampus when it is retrieved spatiotemporally, and the prefrontal cortex compares it with potentially familiar information for final integration39. Finally, this theory is related to studies aimed at measuring spontaneous retrieval (in response to cues) or ecphory, as a convergence between perceptual experience and the engram has been proposed in a process called ecphoric interaction40.
Forgetting and the engram
Forgetting is a process that affects the cognitive process of memory and is the inability to retrieve previously learned information. Natural forgetting is distinct from amnesia, as the latter is considered a pathological process41. In this regard, the engram theory has also contributed interesting proposals about the mechanisms of pathological forgetting and supports the theory of altered information retrieval.
It was previously reported that a group of neurons in the CA1 area of the hippocampus needs to be reactivated for information retrieval42. Similarly, when this neuronal group is inhibited, activity in a specific cortical cell group responsible for learning is also reduced43. This suggests that retrieval may depend on the cooperation of hippocampus-cortex engrams. However, it is necessary to consider how long after encoding engram manipulation studies are conducted. When hippocampal engram dependence is evaluated twelve days after learning, it becomes independent of the hippocampus and dependent on the cortical engram in the medial prefrontal cortex44. Poll et al. studied the participation of the local engram in the CA1 region of the hippocampus in an Alzheimer’s model, finding that competition can occur between two engrams with very similar information, making it difficult to retrieve the correct information (interference)45. Another study demonstrated that in the Alzheimer’s model, rodents have lower dendritic spine density in the dentate gyrus, which correlates with a deficit in memory retrieval27. In addition, studies demonstrate that, under natural and pathological conditions, the engram-cortical or final relay-is resilient; this suggests that the engram remains physically intact and can be functional, indicating that forgetting may occur due to problems in information retrieval46. It is also worth considering that there appear to be distributed engram networks in the brain whose activation depends on the stimulus, which may or may not facilitate retrieval27. Furthermore, in Alzheimer’s disease, one of the roles of the hippocampus in pathological forgetting is related to a low rate of neurogenesis; however, increasing the number of new neurons can reverse memory retrieval impairments36.
Human memory and the engram
The measurement of human memory dates back to the studies of Ebbinghaus. The studies conducted by Ebbinghaus demonstrated that retention depends on practice; the more we practice, the more we remember the stimuli we have learned47. In neuropsychology, pathological forgetting had been detected through neuropsychological tests or batteries (a group of tests aimed at pathology) only when changes were significant, that is, when consolidation and, therefore, information retrieval (delayed recall) were already compromised. Recently, there has been an emphasis on the pre-clinical study of dementias to detect changes prior to extreme deterioration in the subject, allowing for early intervention. The application of tests involving cognitive stress has been proposed to reveal subtle cognitive changes that correlate with neuroanatomical changes related to Alzheimer’s disease48. According to the engram theory, in Alzheimer’s dementia, the structural alteration leading to hippocampal deterioration, although not limited to this structure due to its neurodegenerative nature, will impair proper information consolidation, resulting in rapid pathological forgetting. Recent findings on engrams may open future opportunities to prevent severe cognitive decline in patients with Alzheimer’s disease.
Conclusion
We have updated information on the molecular, cellular, and structural mechanisms that contribute to long-term memory formation. We revisited the engram theory and its importance in explaining not only the maintenance and persistence of memory but also in understanding the mechanisms underlying pathological forgetting caused by Alzheimer’s disease. We also highlighted the relevance of two physiological processes, sleep and neurogenesis, and their roles in memory formation and retrieval. Finally, we provided a brief summary of predictive behavior in the PE theory and its connection to the engram theory.
Funding
The authors declare that this work was supported by funds from the Dirección General de Asuntos del Personal Académico (DGAPA) of the Universidad Nacional Autónoma de México (UNAM), through the Programa de Apoyo a Proyectos para la Innovación y la Educación (PAPIME): Project PE308423 to L.E.R.L.
Conflicts of interest
The authors declare no conflicts of interest.
Ethical considerations
Protection of humans and animals. The authors declare that no experiments involving humans or animals were conducted for this research.
Confidentiality, informed consent, and ethical approval. The study does not involve patient personal data nor requires ethical approval. The SAGER guidelines do not apply.
Declaration on the use of artificial intelligence.The authors declare that no generative artificial intelligence was used in the writing of this manuscript.
