Karel Svoboda's lab records thousands of neurons in mice navigating virtual environments using whiskers. They built a two-photon microscope with a 5-millimeter field of view at single-neuron resolution, and partnered with computational neuroscientist Charles Druckmann to interpret the massive datasets they generate.
Chapters
Everyone is probably interested in the brain, whether or not they know it or not. It's what makes us move around the world in an adaptive manner. It also happens to be the most complex object in the universe currently under study.
You went the wrong way, right?
We're really the laboratory of brain dynamics.
So this is now on the top. Right, and these are all dendrites, the basal dendrites of the pyramidal cells.
Yeah. We're studying the somatosensory system of rodents. For the first time, we're looking at large populations of neurons while mice use their whiskers to navigate a virtual reality environment. The mouse, he's actually running on a treadmill, on a spherical treadmill, and when he's moving the sphere left or right, we can move walls closer or further away from him and thereby simulate corridors of various shapes.
He's navigating this maze, this zigzag shape. The way he navigates this corridor already tells you that he's very familiar with this environment.
By just filming what the whiskers are doing moment by moment, we can reconstruct all of the information that the mice have available to make judgments about the world.
These are the flashing neurons that correspond to place fields.
These are famous place fields and place cells, cells that code for particular locations in space, and it's a major issue that we can record these place fields in this virtual navigation system. The basic behavioral setup, in fact, it's a core part of a student's thesis, Nick Sofronieff, a very talented student, student develop the system. So why do you think—
I mean, I think one advantage, right, that the hippocampus has basically is that it's just a sheet of neurons.
It looks very different than the cortex. I mean, I feel like it has more activity there.
We look for people who have a passion for science and who want to do difficult experiments that potentially would not be done elsewhere. If someone proves their mettle, they can really get a lot of independence in my laboratory. One of the key projects in the lab is to build a new microscope. What we're trying to do with this new microscope is to image large parts of the brain, but with single neuron resolution. So not give up on the micro and not give up on the macro. So really what we want to do is to have our cake and eat it too. It turns out there's no reason, optical reason, why that can't be done.
We just made this piece.
Tanya is really the manager of the engineering group. She has a lot of technical know-how, especially in the area of mechanical design.
It's a complicated piece, yeah.
We've been developing a microscope that allows us to do microscopy at the level of single neurons, but essentially over fields of view that are matched to the, at least the mouse brain, 5 millimeters or so.
Could you pull this apart in distance?
Stretching this out another 5 inches, I mean, maybe that's possible, but it's going to make these objects way bigger. Figuring out how to do the mechanics of the transmission dichro is kind of tough.
This has been a very challenging project. There's optical design that comes into it, optical physics. We're converging on a design and moving into a phase that is called tolerancing and manufacturing. Tolerancing is effectively wiggling all these elements and seeing how quickly performance degrades with Dan's model.
And if that's catastrophic for many elements, then it will be very difficult to manufacture. And we have no idea how that converts into cost.
We'll have to come to a decision whether or not to go ahead with manufacturing. It is still only a computer model that tells us that in principle it can be made, but The major outstanding issue: how well can it actually be manufactured?
There's still some unsolved problems.
I'm around Grunwald, who is one of the pioneers of brain imaging. He came for a conference where we discussed the new microscope that we're building, and I was trying to pick his brain on design parameters and so on and so forth.
All of these changes in the two-photon microscopy— when you were at Bell Laboratory, did you imagine that it will go so far? So how far down the road is the completion of this innovation?
We are not yet certain whether or not it will go into, whether or not we'll even do it in terms of production. I think the implicit understanding is that this is only worth doing if it can be a commodity system, right? If we make one, you know, it will be maybe $250,000 and the second one will be $150,000 or so, so that lots of other people can use it, right?
No question, it's badly needed for everybody working on mice and above.
Especially for bigger animals.
Pretty good. Wonderful.
We have to build microscopes to acquire information, to acquire images from the brains. The GENIE project is a collaboration of multiple laboratories at Janelia, trying to develop better sensors for brain imaging. This was a very exciting meeting. We had really a string of, there were about 5 or 6 new sensors that had never been tested in vivo, and they all behaved roughly how we would have predicted based on in vitro measurements, which just happens very rarely. We criticize each other's data. That's really a very, very important part of these meetings, perhaps the most important part of these meetings. We give each other a hard time, and that's really what makes the science better. Looking under the rocks, trying to uncover flaws. Another very important challenge with microscopy is really what do you do with the data? How do you relate the activity of neurons to what an animal is doing? And that's where we collaborate with a theorist, Charles Druckmann here. He's very much interested in the output of the microscope. To learn about what neurons are doing and what circuits are doing.
So do you have a good idea how many neurons you need to describe the dynamics?
So you start, say, with 10 neurons, and you look at just the raw discriminability between the two.
If we can collect data with a microscope like this, he will have his hands full to help us analyze the data and make sense of it. That's really his role. He's a theorist, computational neuroscientist.
It would be really nice if we could come up with schemes perhaps to pull out something different about these two different neuronal populations.
So it takes collaboration on all of these levels, from developing molecules that make neurons light up, to building the microscopes that allow us to see the neurons light up, to theory to interpret these data. It takes all of that to really make progress in this field. In my view, the brain clearly makes us who we are. Everything that we perceive, all of our feelings, all of our actions. There are many mysteries. Each time we have a new set of tools that provide a new kind of window into what neurons are doing, how they interact with each other, the more we will learn, the more mysteries will be cracked.
