An interview with Professor Florian Mormann, University of Bonn, Germany conducted by April Cashin-Garbutt.

It is captivating to imagine neurons that only respond to images or written references to Jennifer Aniston, and not to any other actress. These so-called “Jennifer Aniston neurons” or “concept cells” were first discovered in 2005, and since then researchers have been studying these cells further to understand their role in cognitive functions. In his SWC seminar, Professor Florian Mormann delved into his work characterising the role of concept cells in perception and memory. In this Q&A, he shares some key findings so far and potential clinical implications for Alzheimer’s.

What are concept cells and when were they first discovered?

Concept cells represent the semantic contents of a presented stimulus. When you see a picture of a face, you instinctively look to see if you recognise it. If you do, you will almost inevitably think of the person’s name. These cells, originally published in Nature in 2005 by Rodrigo Quian Quiroga together with Christof Koch from Caltech and Itzhal Fried from UCLA and others, exhibit remarkable invariance in representation. 

The first examples included a cell that responded to pictures of Jennifer Aniston, showing a response to various photos of her but not to other actresses. An even more interesting example was a cell that responded to both pictures and the written name of Halle Berry. Later examples included responses also to the spoken name by a computer voice, illustrating the convergence of different sensory pathways in the medial temporal lobe. These representations are reminiscent of the infamous grandmother cell, with semantic invariance defining them.

How much is known about the way concept cells can respond to very different pictures of a given person and even their written or spoken name?

We don't really know exactly how concept cells function. However, we do know that their responses are very late, with typical response latencies to a picture being around 400 milliseconds, which is much longer than the time required for object recognition in humans and other primates. 

These cells become active only after a stimulus has been recognised and its semantic content has reached our conscious awareness. At least four different studies show that these concept cells, particularly those in the entorhinal cortex, hippocampus, and amygdala, follow the conscious subjective perception. If someone is tricked into seeing something that isn't there, these cells will be tricked in the same way. 
There is a lot of processing going on before these cells fire, as all the object recognition machinery has already been passed by the time conscious awareness is reached. Understanding this exact machinery in humans remains a topic for further research.

What role are concept cells thought to play in episodic memory?

Episodic memory is one of two types of declarative memory; the other is semantic memory. Semantic memory is knowing that Paris is the capital of France, while episodic memory involves remembering when you first learned that fact. For example, you may not recall the exact moment someone told you Paris is the capital of France, but at some point, that information was conveyed to you, and the memory of that event would be episodic.

A good illustration of this is how infants and toddlers learn language. They repeatedly see an object with a flat surface and four legs that their parents call a table. Through episodic experiences and identifying the invariant aspects of these experiences, they develop a semantic concept of what a table is. Once they have this concept, they can recognise any table as a table, even if they have never seen that particular one before. This transition from episodic to semantic knowledge involves abstract thinking—while a table is a concrete object, the concept of a table is abstract, allowing them to match any concrete table to this abstract concept.

This process is related to the function of concept cells, which so far have been found exclusively in the human medial temporal lobe and are believed to play a crucial role in forming episodic memories. This area is known to be essential for episodic memory formation, as evidenced by cases like patient H.M., who, after having his medial temporal lobes removed, could no longer form new episodic memories. He lived for more than 55 years without being able to encode any new experiences into memory.

Rodrigo Quian Quiroga was the first to formulate the hypothesis that these concept neurons represent the semantic building blocks needed to form episodic memories. Experiences can lead to semantic representations, and these representations can, in turn, be used to form new memories. 

For example, when you meet someone for the first time about whom you have already heard, you initially have just a name. During the encounter, you add details like their face and over time and through repeated encounters other characteristics to this initial concept, forming a more complete semantic concept by integrating experiences across various mnemonic episodes.

Episodic memory involves associating various aspects of an experience, such as who was there, where and when it took place, and what happened. Concept neurons are believed to help in representing these aspects, enabling the brain to activate different representations independently and associatively, making sense of our experiences.

It's interesting to note that the brain fills in gaps in our memories, often making up details to create a complete picture. This phenomenon is well-documented in psychological studies, where only a few core facts are reliably remembered, and much of the rest is fabricated. 

What were the key findings of your recent research on concept cells?

Apart from describing many features and characteristics of these concept cells, we have now published the first study that proves the activity of concept neurons predicts successful associated memory formation, which we see as a proxy for episodic memory. 

We are also in the revision process for a study showing that these cells play a role in memory consolidation during sleep. Specifically, during slow-wave sleep and sharp wave ripples, concept cells related to a previous experience from that day are reactivated. Cells involved in the same experience are co-reactivated within a time window that allows for spike-timing-dependent plasticity. This is reminiscent of place cell sequence replay in rodents, although we do not find the order of events preserved in humans.

Initially, this lack of order preservation surprised us until we considered that place cells, as proposed by researchers like György Buzsáki, are seen as content-free pointers that follow a pre-established sequence in rodents. For the rodent hippocampus, a new environment triggers a sequence where place cells are assigned to place fields as needed. In contrast, human concept neurons have specific content they retain at least over weeks, which defines them and limits their flexibility. This fundamental difference implies that episodic memory in humans works quite differently from rodent models.

Understanding these differences is crucial because it highlights the unique nature of human episodic memory and the specific role of concept neurons in forming and consolidating these memories. These studies not only advance our knowledge of memory processes but also underline the complex interplay between different types of neurons and the stages of memory formation and consolidation.
 

Do these findings have any clinical implications?

An indirect clinical implication of our research is its potential impact on neurodegenerative diseases such as Alzheimer's. By gaining a better understanding of how episodic memory works, we might be able to develop new therapeutic approaches or exercises for people with Alzheimer's.

Patients with Alzheimer's are typically not implanted with electrodes; hence, we conduct this research on epilepsy patients. However, the hippocampus is the first region typically affected in Alzheimer's, and episodic memory is impaired very early in the disease. Finding people with both Alzheimer's and epilepsy is challenging, as the conditions typically manifest at different ages. 

The people with epilepsy we work with are generally too young for Alzheimer's, but as early detection methods for Alzheimer's improve, we may eventually see overlap between the two conditions. This could provide new opportunities for research and treatment development, especially if we can identify and test these conditions earlier.

What are the next questions you are looking to tackle in your research?

I would like to gain a better understanding of the semantic tuning of these concept neurons. Currently, we're working on a study to demonstrate the existence of semantic tuning. If cells respond to multiple stimuli, there is typically a semantic or episodic relationship between these stimuli.

For example, a cell may respond narrowly, such as only to Jennifer Aniston, or broadly, such as to all the characters from a TV show like "Friends." This variety in responsiveness illustrates the range of semantic tuning that exists.

Another intriguing question we're exploring is how quickly these representations emerge. Based on our observations, sometimes within 24 hours of initial contact between a member of my group and one of our patients, we observe a representation of that group member in the form of a concept neuron. This rapid emergence raises questions about the mechanisms underlying this process and how quickly semantic representations can form.

Lastly, we know that episodic memories eventually become less dependent on the hippocampus and are transferred to the neocortex. This transfer is believed to occur during sleep and memory consolidation, where we observe reactivation of concept neurons. Understanding the mechanisms that facilitate this transfer between the hippocampus and neocortex would provide valuable insights into memory consolidation processes.

About Professor Mormann

Florian Mormann studied physics and medicine at the Universities of Karlsruhe, Bonn, and Cologne. After a postdoc term at the California Institute of Technology and the University of California Los Angeles with Dr Christof Koch and Dr Itzhal Fried, he returned to Bonn where he became Lichtenberg Professor of Cognitive and Clinical Neurophysiology.