Henry Denny on聽Visual objects refine head direction coding
Read the article in .
If you ask anyone who knows me well, they will probably be able to tell you that I enjoy taking the long way around. While my academic career started at the University of Manchester in the United Kingdom, my research interests have taken me far and wide; from studying synaptic metabolism in the scrubby sagebrush of the Nevada desert, to studying endocannabinoid signaling dynamics in the bustling metropolis of Tokyo. As such, it seems fitting to me that I landed in the field of navigation; studying the head direction system under the supervision of Prof. Adrien Peyrache.
Invariably, every traveller receives both solicited and unsolicited advice during their journey. One piece that鈥檚 stuck with me is that 鈥淚f you want to go fast, go alone; if you want to go far, go together.鈥� I think our recent paper 鈥淰isual Objects Refine Head Direction Coding鈥�, published in Science, is a testament to this idea. Scientific work is often portrayed as heroic feats, achieved by individuals of great intellect. In contrast, it is my opinion that this paper is a testament to what can happen when friends come together and collaborate.
My arrival on this project was mainly due to a series of fortunate events on my part. During my rotation in the Peyrache lab in the spring of 2022, I had played around with a project in the head direction cortex, the post-subiculum. We were interested in how inhibition governed the ability of the head direction system to anchor to rotating cues. In pursuit of this, I worked with Dr. Adrian Duszkiewicz (who is now starting a faculty position at my alma mater, the University of Manchester! Good choice Adrian!) to develop a fully symmetrical LED arena for mice. This arena would allow us to precisely control cue structure and position while the animal freely navigates. When I joined the Peyrache lab full time, my focus shifted away from landmark control of head direction to how internal dynamics stabilise the head direction attractor. As such, the LED rig, affectionately named 鈥淢ouse Berghain鈥� by the lab, lay dormant in storage, gathering dust.
A few years later, in the summer of 2023, Adrien approached me with a proposal. Stuart Trenholm鈥檚 lab, in collaboration with Dr. Dominique Siegenthaler and Prof. Emilie Mac茅 at Universit盲tmedizin Gottingen, had used Functional Ultrasound (FUS) imaging to examine the mouse visual stream to test whether the mouse visual system is hierarchically organised like the primate visual cortex. They presented mice with images of objects, such as machines, mice, or human faces, while measuring blood-flow across large volumes of the brain. Significantly, they found that in the mouse brain, regions usually responsible for navigation responded more strongly to objects than images that had been scrambled. These scrambled images maintained the same statistics of the original images but no longer contained any semantic content. Critically, the region with the strongest response to visual objects was my old friend, the post-subiculum.
Recordings in anaesthetised animals by Dom and Johanna Mayer, a PhD student in Emilie鈥檚 lab, had shown that this preference was produced by heterogenous changes in the firing rates of post-subiculum neurons. However, the Mac茅 lab was not positioned to perform the experiments in freely moving animals that would untangle precisely which neuronal subtypes were responsible for this change in firing, and how these changes modified head direction coding. Emilie and Stuart had approached Adrien about performing these experiments and he suggested that my experience studying landmark control in the post-subiculum would be well applied to this project.
Around this time, I had been talking to my friend, mentor, and fellow PhD student Sofia Skromne Carrasco, whose primary project was in the post-subiculum. We had been talking for a year about how much we鈥檇 like to work on a project together, but since my project had moved to the thalamus, we had found limited scope for such a collaboration. But now Adrien was offering a prime opportunity where our skillsets could synergise; my experience with visual control of the head direction system, and Sofia鈥檚 experience doing electrophysiological recordings in the post-subiculum.
After getting the rig up and running again, Sofia and I began recording from neurons in the post-subiculum while showing the mice images of natural objects and their scrambled counterparts. What we found was striking. Some neurons lit up when the mouse looked at an image, while others fell quiet. When we separated these cells by type, it turned out that the positively modulated neurons were mostly head-direction (HD) cells and fast-spiking interneurons. Both cell types preferred visual objects over scrambled versions, meaning that real-world structure, rather than raw visual texture, mattered to them. In other words, the post-subiculum wasn鈥檛 simply responding to light, it was responding to meaning in the visual scene.
When we looked closer, we noticed a pattern. The HD cells that pointed toward the object in the environment became more active, while those whose preferred direction was away from it became quieter. If you imagine all the HD cells arranged around a compass, the ones aligned with the visual landmark formed a 鈥渂ump鈥� of heightened activity that sharpened the overall sense of direction (Figure 1). We brought on Dr. Dan Levenstein (who is now starting a position at Yale. Congrats Dan!) to build a network model to hypothesize about the potential mechanism behind this phenomenon. His model was powerful but remarkably simple; once the network has a ring-like architecture, adding untuned visual drive naturally strengthens activity for the currently faced direction and suppresses the rest. In essence, seeing an object stabilizes the compass needle of the HD system.
These findings posed a tantalising hypothesis; if meaningful cues enhance head direction neuron firing when the animal is facing them, then we should see a difference in firing rates when the animal is facing in the same direction while either looking at or not looking at the landmark. For us, this was an absolute moonshot experiment. The changes we were observing were small and captured over an hour and a half recording. Would it be possible to see them in a 15-minute exploration? But against all odds, in freely moving recordings, HD cells fired more when the animal faced a wall-mounted cue and less when it faced away. When we replaced the object with its scrambled version, this modulation disappeared. From this we could conclude that the HD representation sharpened only in the presence of a meaningful visual cue. Together, these results show that visual objects don鈥檛 just anchor the head-direction system; they actively refine it, tightening the population code that keeps the brain鈥檚 internal compass steady as the animal explores.
For me, this project was as much about people as it was about neurons. The findings themselves are beautiful in their simplicity; an elegant mechanism by which visual meaning sharpens our sense of direction. But the process of discovery was equally enlightening. Every step involved a different person鈥檚 expertise: Sofia and Johanna鈥檚 mastery of surgery and recordings, Adrien鈥檚 deep knowledge of navigation, Stuart鈥檚 complementary understanding of visual neuroscience, Dan鈥檚 expertise in computational modelling, and Emilie and Dominique鈥檚 mastery of frontier techniques that allowed them to uncover a very subtle phenomenon. My role felt like steering between all these worlds: implementing Dom鈥檚 setup here in Montreal, coordinating data collection and analysis via zoom call, helping the story of the data unfold. What stands out most is that none of us could have done this alone, and the work is all the better for it.
If you ask anyone who knows me well, they will probably be able to tell you that I enjoy taking the long way around. I鈥檝e learned that the detours (the unexpected projects, the late-night ideas over beers, the collaborations that start as hallway conversations) are often where the best science happens. This paper was one such detour, and it reminded me just how much joy there is in doing science together. So, here鈥檚 my unsolicited advice to future travelers: work with people who challenge you, make you laugh, and make your ideas better. The long road may take more time, but it鈥檚 a lot brighter when you walk it with friends.
Figure 1: Presentation of a visual landmark increases firing rates in a restricted segment of the head direction compass. Each point on the ring corresponds to an individual head direction neuron in the recording. The position on the ring indicates the relative tuning direction of the neuron in the horizontal plane. Orange indicates an increase in firing upon landmark presentation. Green indicates a decrease in firing upon landmark presentation. Two neuron鈥檚 head direction tuning curves and visual responses are displayed as examples.
Figure 2: The Peyrache lab on retreat, January 2024. Bottom right is me! Behind, from left to right: Dr. Dhruv Mehrota, Nour Chahine, Sofia Skromne Carrasco, Yuxiao Wang, Prof. Adrien Peyrache (above), Adel Halawa (below), Prof. Dan Levenstein, Dr. Sara Mahallati, Aleksei Efremov.