Memory formation in the human medial temporal lobe

One of today's scientific challenges is to understand how memories are stored in the brain. Consider something as simple as remembering meeting a person for the first time. Our brain creates such memories effortlessly, but this involves complex neuronal processes that we still do not understand. One could, for example, ask: How is the neuronal representation of the new person formed, and how is this representation consolidated and stored for years to come?

To answer these questions, we would ideally like to record the activity of neurons while subjects perform memory tasks. However, there is a major limitation: we usually cannot record the activity of individual neurons in humans. The problem is that to record the activity of individual neurons we must introduce small electrodes inside the brain, something that cannot be done for obvious ethical reasons and, consequently, we have only access to recordings of neuronal activity from outside the skull, using techniques such as EEG or fMRI. However, these methods cannot give information about the activity of individual neurons and can therefore not unravel the precise neural mechanisms of how memories are formed. An alternative approach is to record individual neurons by implanting electrodes in animals' brains, but the types of experiments and questions that can be studied are limited, as animals cannot give feedback of their thoughts and recollections and need extensive training, far from the natural conditions of real-life memory formation.

In very specific cases, recordings of individual neurons can be performed in humans. This is the case of epileptic subjects, who are implanted with intracranial electrodes for clinical reasons. While studying the activity of neurons in these subjects, we discovered what have been named "Concept Cells" (a.k.a. "Jennifer Aniston neurons"): namely, neurons that respond in a remarkably selective and abstract manner to specific persons or objects, like Jennifer Aniston, Luke Skywalker or the Tower of Pisa. For example, one neuron responded to 7 different pictures of Jennifer Aniston and not to 80 pictures of other persons or objects. That means the neuron responded to the concept "Jennifer Aniston" and not to the different details of each of the pictures presented. Given that they are located in an area that is known to be critical for memory, we have argued that these neurons are involved in memory functions - in agreement with the fact that we tend to remember concepts and forget irrelevant details. However, we still do not know how they start responding to specific concepts (e.g. the person we meet for a first time) and how these new memory representations may eventually consolidate.

The project will exploit the unique opportunity of recording individual neurons in humans, who, in contrast to other animals, can give detailed feedback of their thoughts and recollections. We will explore the neural mechanisms of memory formation by tracking the firing of neurons while memories are created and consolidated. In particular, we will analyze the activity of neurons while the subjects familiarize themselves with initially unknown faces after repeated presentations. To study memory formation in more natural situations, we will also track the activity of neurons while the subjects watch initially unknown and very engaging movies (from a Hitchcock TV series from the 60s), thus studying how the neurons encode the different movie characters as they become familiar. Experiments will be repeated twice a day (after a few hours' break) and in consecutive days, to assess the stability of the neurons' responses. Furthermore, experiments will be complemented with specific questionnaires to determine the nature of memories encoded by the neurons, such as their emotional saliency or the triggering of specific recollections.

In summary, we will describe for the first time the neural machinery that is involved in creating and consolidating new memories.

For clinical reasons, patients with epilepsy may be implanted with intracranial electrodes, thus allowing recording from multiple single neurons in human subjects performing cognitive tasks. Electrodes are typically placed in the Medial Temporal Lobe (MTL), an area involved in declarative memory (memories of facts and events) where we have found "Concept Cells" (Nature 2005) - i.e. neurons selectively firing to specific concepts, such as a particular person or place.

Declarative memory relies on establishing associations between concepts. For example, the memory of meeting a person at a particular place involves establishing an association between the neural representation of this person and the one of the place. Given the well-established role of the MTL in memory, we have argued that Concept Cells provide a representation of concepts and their associations for memory functions (Nat. Rev. Neurosci. 2012). In recent experiments, we have shown that Concept Cells encode meaningful associations between concepts (Nat. Comm. 2016; Nat. Comm. 2018) and that they can very rapidly change their tuning to encode new associations (Neuron 2015). However, these experiments were performed using concepts that were already very familiar to the subjects and it is still unknown how the representation of new concepts is formed and consolidated. This is exactly the issue we will address in this project, by using repeated presentations of initially unknown faces and by showing unknown movies. In particular, we will track on a trial-by-trial basis the mechanism by which the repeated presentation of new concepts gives rise to the very sparse coding we have already found for familiar concepts.

In summary, this project will offer an unprecedented level of understanding of the neural mechanisms involved in memory formation and consolidation, exploiting the unique opportunity to record multiple single neurons in the human MTL, while the subjects perform memory tasks.

The primary impact of our work will be academic, as detailed in the Academic Beneficiaries section. However, the project also has the potential to have medium- to long-term clinical, economic and societal impact, contributing to the study of epilepsy and Alzheimer's Disease.

Epilepsy is one of the most common neurological disorders, affecting almost 1% of the population. The fact that we will perform 24/7 single neuron recordings in epileptic patients, including the recording of seizure, pre- and post-seizure activity, will enable to study the neural mechanism underlying epileptogenic activity. Intracranial EEGs (iEEGs) have given invaluable information to classify different types of seizures and to identify electrophysiological patterns, but cannot provide detailed mechanistic information because they reflect the activity of large neural populations and give only indirect and ambiguous evidence of the activity of single neurons generating these behaviours. In this respect, our project will provide unique data combining iEEGs with direct recordings from multiple single neurons in seizure-originating and neighboring areas.

A major problem faced by epileptic patients is that seizures usually have an abrupt onset, thus imposing severe risks and limitations that affect their everyday life. In contrast to these abrupt clinical manifestations, it has been argued that more progressive changes in the dynamics of the underlying brain activity could in principle be tracked from the iEEG signals. However, there is still no consensus on how to predict seizures from iEEGs and our project will provide unique data that may help seizure forecasts by considering the activity of multiple single neurons. This has an immense clinical potential, since it would lead to the development of systems that could give a warning of an impending seizure, or closed-loop systems that could preclude its onset, for example, by delivering a fast-acting anticonvulsive drug or electrical stimulation.

With respect to Alzheimer's Disease (AD), there are more than 0.5 million people suffering from this pathology in the UK alone, a number that will double in the next 30 years (Alzheimer's Society Factsheet 2014). AD is characterized by a progressive memory impairment due to degenerative changes of the area of the brain that we will focus on. In advanced stages, AD patients cannot even recognise a person they recently met. In this respect, understanding how the brain forms memories and consolidates them is a natural step towards understanding what goes wrong in a pathology affecting memory itself. This will inform clinicians and the health system about cognitive behavioural therapies and strategies to, for example, delay the progression of AD by stimulating and reinforcing the type of memories and memory mechanisms involved in normal memory functioning. In fact, studying the nature of memories encoded by neurons in the hippocampus is one of the key aims (AIM3b) of our project. The overall UK economic impact of dementia is £26.3 billion per year, so, should our studies lead to therapies that delay the progression of the disease and the need for assistive care, this would have major economical and societal implications. The project can also have an impact for policymakers, by providing information on new strategies and potential investments to be eventually implemented by the NHS and health visitors to treat Alzheimer's patients.

Finally, our work will also have an impact on the general public, as the topic of memory formation and consolidation is of great interest in modern society. The high attendance at, and interest shown by the layperson in, talks and seminars given by the PI in the past is testimony of the general interest in this area.