Interview with Prof. David Berson
Pioneering Accessibility in Retinal Connectomics
David Berson, professor at Brown University, is best known for his pioneering research on retinal neurons, especially for discovering the cells that regulate our circadian rhythms and those that stabilize our view of the world. As he approaches the end of his career, David opens up about his final work: a gift to the scientific community. This legacy project aims to equip future generations of researchers with comprehensive and accessible retinal data at electron microscopic scale. With his lab now closed, he reflects on his journey, past achievements, and the ambitious initiative he has started together with other renowned researchers. Let’s dive in!
Hi David. Can you start by giving us a broad overview of your research?
Absolutely. My primary focus throughout my career has been to explore how the initial stages of the nervous system’s visual pathway process, filter, and select essential information from the visual environment, thus driving our adaptive visual behaviors.
Many people consider the retina merely as camera film, capturing images for further processing. However, the retina is much more complex: it’s where the brain’s first stage of visual information processing occurs. There are multiple synaptic connections in there, and many, many different cell types. If you want to understand anything about how the nervous system uses synapses, cells, and circuits to compute information, the retina is a superb place to do so. It’s all happening in a very thin sheet, with beautifully organized layers.
For example, my lab did a lot of research about the cells that synchronize our circadian rhythm by adjusting to environmental cues. In the 1990s, research by Russell Foster and his team demonstrated that the biological clocks in mice remained synchronized with the day-night cycle even when these mice appeared to lack rods and cones. However, removing the eyes disrupted this synchronization, indicating the presence of another type of photoreceptor. Our lab was the first to record from these cells, confirming that they respond to light independent of synaptic input. As many teams help to prove, these cells have their own photopigment — a substance capable of absorbing light and converting it into a neural signal. These cells are located in a different layer of the retina than where rods and cones are found. They play a crucial role in synchonizing the animals’ circadian rhythms with their environment.
This is just one of the 40 identified types of mouse ganglion cells. My lab has worked on many other types. One of them is responsible for image stabilization. As we turn our heads, our eyes counter-rotate with remarkable precision. This stabilizes our view of the world, a lot like a steadicam in filmmaking. To do this, the brain relies on visual motion signals computed in the retina by a specific type of neuron. Retinal systems for circadian regulation and image stabilization are highly conserved across species. These remarkably specific retinal circuits have evolved to extract useful information from the visual world that help the animal survive and reproduce. That’s all computed by synapses in this thin little sheet of retina. That’s my thing!
How did you first decide to study neuroscience and specifically focus on the retina?
Well, the science part was easy: my father, a chemist, and my mother, who was deeply interested in science as a young woman, filled our dinner conversations with scientific discussions. Their enthusiasm was contagious. As for my fascination with vision and the retina, I think a lot of it came from a series of illustrated books for egghead kids my folks got for us. One was filled with visual illusions. That captivated me and is probably how I first realized I was interested in visual perception.
So I went to grad school to do biological work on vision. I wanted to do neurophysiology, but I ended up doing neuroanatomy instead with the fabulous scientist Ann Graybiel, who was just starting out her career on the MIT faculty at that time. Ann inspired my love for brain structure. I recently closed my lab so I could get close to the science again and do the anatomy myself, just the way I did when I started out, spending hours and hours at the microscope. Now, I spend my time on the WEBKNOSSOS interface, just exploring, and exploring.
How has serial electron microscopy influenced your work?
Serial electron microscopy really came from Kevin Briggman’s beautiful work on direction-selectivity. I had never gone near EM before, relying instead on light-microscopic approaches. Kevin and colleagues beautifully demonstrated the enormous potential of SEM for working out retinal synaptic circuits on an unprecedented scale. It was perfect for linking specific cells and synaptic circuits to particular retinal computations. I was determined to get my hands on this approach, but at that time we didn’t have the necessary technology at Brown to acquire such volumes. Almost nobody did anywhere! So I basically told Kevin I would do anything to get my hands on material like this for work on ganglion cells and he offered to share with me a volume that hadn’t been published yet. And I have been working on it ever since.
I began searching for my favorite cells within this volume and I quickly realized that I could recognize them. My first goal was to get a look at some really peculiar synapses that carry rod and cone signals into the melanopsin ganglion cells. These were completely bizarre, like nothing I’d ever seen before. At that moment, I could immediately see the power of the method for answering questions about anatomical and functional specializations within these parallel channels that are doing the computation.
My mantra these days is: ‘the anatomy of the circuit is the computation’. Of course, that’s overstated. But having a complete connectome of the chemical synapses in a chunk of retina gives you an enormous boost in understanding how the retinal translates the image at the photoreceptors into 40 different pre-processed representations at the retinal output.
Does this have to do with what you are currently working on? Can you tell me more about your present project?
My dream now is to leave the field with a resource, a complete connectome of the mouse retina, accessible to every scientist who has a laptop and a decent internet connection. Together with Rachel Wong, Wan-Qing Yu, and many other talented researchers interested in retinal connectomics, we would like to image, segment and reconstruct a volume of mouse retina and make it available for the community. This platform would be a gradually evolving curated description of the major synaptic layers in the retina and be large enough to encapsulate all significant types of circuits, adopting a community-enhanced, wiki-style approach akin to the FlyEM Hemibrain project.
We are working right now with scalable minds on the segmentation and have planned an automatic synapse detection. There are two types of chemical synapses to find. Most are conventional chemical synapses like ones you see in the brain; in the retina, they come from retinal interneurons and are mostly (but not always) inhibitory. The other synapse type, the ribbon synapse, is highly specialized. It’s a plate-like organelle that tethers multiple synaptic vesicles near the release site, creating this large pool of vesicles ready to go. This is an excitatory synapse. It’s the workhorse for the main ‘vertical’ signaling pathway in the retina, from photoreceptors to output neurons. We would like to find out where all those synapses of both types are, and which neurons are using them to talk to each other.
Our goal is to lay a solid foundation for this project, so that when people gain access — hopefully within a year — they’ll find substantial work already done that can be used as a starting point for their own projects. Even with merge and split errors, the groundwork should be advanced enough for researchers who study a particular type of cell to know where to look and how to get started. Many people are not trained in anatomy but would gain very valuable insights from it if the process were less daunting. We want to lower the energy barrier.
It sounds like you are deeply convinced about the potential of connectomics for research on the retina. Do your peers in this field share this conviction?
Many people are really excited about the work that I’m doing yet are sure that under no circumstances would they ever get involved in connectomic research themselves. They’re not convinced that the process is as accessible as I think it is now. At times, I think of myself as a lonely prophet in the desert trying to get people to pay attention to connectomics. I am confident they’re going to understand eventually, but it might take a while.
The field needs to be able to produce, analyze, and share SEM datasets more quickly. Right now, we have to settle for incomplete analysis of a tiny collection of volumes. But imagine a day when it will be possible to make detailed connectomic comparisons of the retinas at specific stages of development… or of degeneration. When we gain the means to understand these processes at connectomic resolution, we will see a major leap forward.
You said you want this volume to be accessible online to everyone. Do you know which visualization and annotation platform you will be using?
This is a very good question, for which we haven’t got an answer yet. We’re exploring whether there is a single platform that could host all this retinal connectomic data while also facilitating data refinement in a collaborative, wiki-style manner. I think a key element is interoperability, meaning, for instance, if someone prefers to trace in WEBKNOSSOS but wants to address merge and split errors using Neuroglancer, it should be straightforward to transfer data between these tools. In the end, you need one repository that has the latest definitive curated version. This way, if somebody wants to look at a particular cell type, without doing any proofreading, they get the latest and greatest.
I understood you recently closed your lab. Are you still doing some teaching?
Yes, I continue to teach, mainly an advanced neuroanatomy course at Brown. Most of the people who take it are seniors majoring in neuroscience, and about half of those are pre med, going off to medical school. The goal of the course is to give them an overview of all the dimensions of structure in the nervous system.
For example, I designed an exercise using WEBKNOSSOS. I assign a lab project where students work with provided skeletons in WEBKNOSSOS that identify mitochondria, synapses, and more within an actual volume of tissue. This allows them to navigate through the sections and contextualize what they’re seeing. They used to look through a microscope at a piece of tissue. Now, they open WEBKNOSSOS from anywhere and can look at these large datasets. The exposure they get this way gives them a sense of possibility and shows what the evidence could look like, for example in a paper. I think that’s fabulous.
