“We teach mice to play video games!”

As the EPFL Life Sciences Independent Research (ELISIR) program enters its fifth year, we meet one of its recent scholars, Dr James Priestley, who joined EPFL in 2023 from Columbia University. Priestley’s research focuses on the neurobiology of memory, and particularly the computational role of the hippocampus and associated structures in memory formation.

Did you always want to be a scientist?

I've been involved in scientific research since pretty early on in my undergrad. I studied neurobiology as an undergraduate and also for my PhD. And I've been working in the field of memory research for about ten years now.

Were you always interested in memory research?

I took an interest in neuroscience pretty early in my education. And with memory in particular, I think it’s central to a lot of what we ascribe to cognitive behavior. It allows us to take advantage of our past experience and use it to guide adaptive behavior, or simply just reminisce about the past. It's an intrinsically interesting topic for me.

The hippocampus has been a consistent focus of my work throughout my career, from my research as an undergrad to my PhD and onwards. I studied for my bachelors at Boston University where I was working in the lab of Howard Eichenbaum, who was really a champion for a memory-centric view of hippocampal function, and his work really left a strong impression on my views of the field.

I should mention that there's a little bit of controversy regarding the functional purpose of hippocampal circuits. We know from the human neuropsychology literature that insults to the medial temporal lobe cause deficits in episodic memory but a lot of the early recordings of hippocampal neurons in rodents focus on the spatial selectivity of these neurons – individual neurons in the hippocampus appear to encode for specific locations in physical space, and collectively encode a “cognitive map” of the environment.

In my work with Howard [Eichenbaum], we explored variables beyond physical space that might be encoded in the rodent hippocampus. Specifically, my early research focused on the encoding of time, a critical component for episodic memory as it helps describe the progression of events in sequence. We investigated how hippocampal circuits encode temporal relationships and how the organization of sequence coding in the hippocampus depends on upstream cortical inputs to the circuit.

My PhD began as a natural continuation of this work, where I tested models of temporal coding in the hippocampus during the formation of fear memories in the labs of Attila Losonczy and Stefano Fusi at Columbia. In later work, I began delving into aspects of spatial coding and this was where we began using virtual reality (VR) to more flexibly control the spatial and sensory experience of animals during neural recordings. We used these techniques to study how novel experiences are initially encoded into memory in the hippocampus, by recruiting particularly rapid forms of synaptic plasticity. More recently, my research has centered on using VR to study how the hippocampus might take advantage of the patterns present in the sensory world to build more efficient codes for memory.

Where is research on memory now?

Studies on human patients since the 1950s have identified the hippocampus and medial temporal lobe as crucial for the formation of new episodic memories, our “mental autobiography". And there’s been a huge amount of work since then across many species that has really illustrated the diversity of different stimuli and facets of experience that drive neural activity in the hippocampus, both spatial and non-spatial. So the challenge now is to try to make sense of this rich and mixed neural code and try to find more general computations that could explain what the hippocampus is specifically contributing to the brain’s memory systems, or to other proposed functions like navigation.

One idea we’ve been pursuing recently is that the hippocampus may actively learn information about shared patterns that occur in the sensory content of different experiences, and use that structure to organize memories into a more efficient format, one that can optimize storage capacity and robustness. When you think about it, a lot of memories we encode are very similar, like remembering two conversations that happened in the same room. So the hippocampal circuit might contribute to learning and compressing these redundant features in memories, which can make the stored patterns much sparser, decorrelated, and easier to recall, and also possibly explain why we see “place cells” during exploratory behavior. That is, they could simply reflect how the brain parcellates the correlated components of memories that occur in the same location.

Is there any loss of information? Any tradeoffs?

It's possible that there's some tradeoff between learning the patterns shared between different memories and how much information we retain about specific details of individual episodes. We know the hippocampus tends to generate quite sparse codes, which can transform the representation of experiences to be less similar to one another and make it easier to retrieve distinct episodes. In the memory literature, this is referred to as “pattern separation” and this can reduce interference between memories during recall. But we think these effects are likely regulated by other factors like attention, novelty, motivation, and other cognitive factors, and so we’re also very interested in studying how the structure of these neural codes changes under diverse behavioral and sensory conditions. And these factors could definitely affect which information about experience is encoded into memory or forgotten.

Speaking of interference, neurodegenerative disorders like Alzheimer’s indicate that memory is more global rather than restricted in a part of the brain. Can you comment on that?

Memory is definitely a dynamic and distributed process in the brain. In my group, we’re focusing for now on the hippocampus, which seems to be very important in particular for the initial formation of episodic memory. We’re very interested in this initial encoding stage and what happens when an experience is novel: how rapid plasticity in the hippocampus might support the organization of this new information into memory. But it will also be valuable to study how this brain region interacts with other parts of the cortex both during initial encoding and later recall. Many theories of memory consolidation – how memory is organized and stored long-term – suggest that memory traces become less hippocampal-dependent over time as information is consolidated in the cortex for more long-term storage, but how this process unfolds is still largely unclear.

What kind of techniques do you use in the lab?

In our experiments, we want to be able to expose animals to rich sensory environments and play with the statistical structure of those experiences to see how the hippocampus may respond to learn and encode those patterns, beyond simply mapping space. We do this is by creating virtual reality environments for mice, kind of like a simple maze, where we can have really precise control over the sensory cues. VR is great because we can arbitrarily introduce novel information or manipulate the environment in real time. We can introduce new structures to the environment, like correlated patterns in different locations; we can move the animal to a completely new environment and study how these different scenes are encoded in the hippocampus. Essentially, we teach mice to play video games! And then we combine this with large-scale recording techniques like calcium imaging, which allows us to record the activity of up to thousands of neurons simultaneously during behavior.

What are some milestones from your research?

In my latest paper where we showed that when animals are exposed to new environments – with novel sensory stimuli – neurons in the hippocampus often recruit a very unique synaptic plasticity rule that enables them to form really strong selectivity for specific locations and environments with just a single exposure to the environment. So very rapid “one-shot” learning. And we think that this is particularly important for the initial encoding of new memories.

What are your research plans while at EPFL?

We’re hoping to take advantage of VR, which is really great because it allows us to arbitrarily manipulate the animal’s sensory world and ask precise questions about how details of experience are encoded into memory. So we’re exploring how representations in the hippocampus change when we use the VR to change the patterns in sensory experience, for example, by modifying how “compressible” an environment is.

We’re particularly interested in the dentate gyrus, the main input region to the hippocampal circuit, which is thought to be very important for generating sparse codes for memory. We want to gain some insight into how the circuit may learn sparse codes that are adapted to the specific statistics and structure of ongoing experience, and also study the role of local inhibitory neurons in these computations. We benefit a lot from our shared interests and technical approaches with Carl Petersen’s lab, which has made EPFL a great environment to begin research in systems neuroscience.

Our work so far has used relatively simple, linear environments, since our neural recording methods generally require animals to be head-restrained, but we’re also planning to explore more complex behaviors in freely moving animals. To do this we're using miniaturized, head-mounted microscopes that we can use to record neural activity during unrestrained movement in augmented reality environments.

Our work involves a mix of experiments and computational modeling, where we design VR tasks to record memory formation under different conditions, and we try to replicate these phenomena in artificial neural networks. We hope to draw parallels between efficient coding in artificial neural networks and what we observe in the brain during initial memory encoding and long-term memory recall and consolidation.

What do you think about the ELISIR program?

It's a really great and unique opportunity, to be able to start my own group. After my PhD I felt like I had a pretty good sense of the direction I wanted to go with my work, and so the opportunity to launch directly into the experiments and projects that I found most interesting, to have that independence – I think it was something that would be impossible to turn down. And certainly, the environment here at EPFL for neuroscience research is excellent.