Tools for mapping and repairing the brain [part 3] (19:58)
June 2, 2016
July 8, 2015
All Captioned Videos CBMM Summer Lecture Series
Ed Boyden, Professor of Biological Engineering and Brain and Cognitive Sciences at MIT, leads the Synthetic Neurobiology Group, which develops tools for analyzing and repairing complex biological systems such as the brain, and applies them systematically to reveal basic principles of biological function and to repair these systems. In this three-part lecture, he discusses tools for mapping and repairing neural circuitry using expansion microscopy (part 1), whole-brain imaging with light-field microscopy (part 2), and optogenetics (part 3).
Ed Boyden’s Lab website: The Synthetic Neurobiology Group Karagiannis, E.D. & Boyden, E. S. (2018)
Expansion microscopy: Development and neuroscience applications, Current Opinion in Neurobiology 50:56-63. Klapoetke, N. C., et al. (2014)
Independent optical excitation of distinct neural populations, Nature Methods 11:338–346. Prevedel, R. et al. (2014)
Simultaneous whole-animal 3D imaging of neuronal activity using light-field microscopy, Nature Methods 11:727-730. Tye, K. M. & Deisseroth, K. (2012)
Optogenetic investigation of neural circuits underlying brain disease in animal models, Nature Reviews Neuroscience 13:251-266.
ED BOYDEN: So controlling neurons has been done for a long time. People use electricity to stimulate neurons. People use pharmacology to modulate neurons with drugs, but drugs are slow, and electricity goes everywhere. So how could you activate or shut down just one neuron?
If you could activate or shut down neurons, then you could really figure out how they contribute to their circuits, right? You could turn on some cells and figure out what they can initiate. You could shut down cells and figure out what they are needed for.
And so light can be aimed, unlike electricity, and light is a lot faster than, well, everything else, actually, so it has enough speed that we can keep up with brain activity. And you can also put probes into the brain. We talked about electrodes earlier, but you could put optical fibers into the brain, too. The brain doesn't feel pain.
And the hard part, in a way, is knowing how to make neurons sensitive to light. So it turns out all over the tree of life you can find organisms that have molecules that convert light into electrical signals. If you've ever looked out into a salty water and seen sort of an orangish color, that's because of microbes that have a pigment that absorbs light, and when hit by light, moves a charged particle, a proton, from one side of the cell membrane to the other. So it's a way of storing energy.
There are other molecules that pump negatively charged ions like chloride. And then finally, other molecules that also sit in membranes that let positively charged ions into the membrane. So these are all proteins. They're encoded for by small genes. We can put those genes into the genome or into the cell using gene therapy vectors.
And over the years, what we and our colleagues and collaborators have found is that some of these molecules work really well. Some of some of the light-driven proton pumps can pump protons out of neurons, and when you do that you silence the neurons. Light-driven chloride pumps pump chloride into neurons, also shutting them down, and the light-driven ion channels let sodium, potassium, calcium, and protons into the cell, and that can activate the neuron.
So there's all sorts of things you can do with these molecules. There's a disease called narcolepsy, where people fall asleep at random and inopportune times, and in these patients, a small cluster of cells has died off, cells that make the peptide hypocretin or orexin. And so the [INAUDIBLE] group made a transgenic mouse that has a light-driven chloride pump in just these hypocretin or orexin neurons.
They then aimed an optical fiber at these cells connected to a laser. And when you turned the light on, this is the probability of being awake. The mice start out awake, and then they turned the laser on around here. And as you can see, the mice all completely fall asleep. And then the light turns off, and bam-- they all wake back up.
So you could really use this to prove that a specific cell is involved with a specific behaviorally relevant condition. So what we've been doing recently is figuring out how to make this even more powerful. Red light can go very deep into the brain because red light is not absorbed by blood.
And so Amy Chuong, who recently finished her PhD in our group, worked on a molecule that was more red light sensitive. We named it Jaws. It's a long story. And you know, we find these molecules by looking at genomes and then mutating the things we find.
But Jaws, because it's red light sensitive, allows us to non-invasively shut down neurons. We can shine light from an optical fiber across the skull of an awake behaving mouse. And as you can see here, this is percent suppression plotted versus depth. We can shut down neurons many millimeters deep in the brain.
What about activators? So this is some work from the Fiorillo group. What they were doing is studying dopamine neurons. I think we've all heard of dopamine neurons. They're involved with reward, addiction, learning, all sorts of stuff, and also involved in their absence with Parkinson's disease and other pathological conditions.
So what they did was take a transgenic mouse and virally deliver the activating opsin, known as [INAUDIBLE] to the neurons, and then they would put a mouse in a box. And every time the mouse goes to this spot, it gets a pulse of light, and every time the mouse goes to this spot, nothing happens. And so you can see what happens.
The mouse goes, gets a pulse of light, pokes his nose, gets a pulse of light, pokes his nose again, gets a pulse of light. And so the mouse is basically working for light, and so that allows you to prove that activating these cells is enough to reinforce the behavior and make the brain do more of what it was just doing, at least in the context of this behavior test. So recently, on the tool side, we've developed a red, in fact, nearly infrared, activator neurons that we call Chrimson. You can-- this is, you know-- down here are action potentials from the mouse visual cortex, and you can fire spikes with 735 nanometer light, which is quite far to the red.
We also solve the problem, which is, can you make a very fast molecule that's also very light sensitive, a kind of good all-around opsin for activating neurons, and so we called this one Chronos. And interestingly, Chronos and Chrimson, these two molecules that we just talked about, if you use them together, you can do two color control. You can put Chronos in one pathway and activate it with blue light, Chrimson in another pathway to activate it with red light, and see how two sets of neurons work together. And the actual reason why this works is a little bit complicated to go through, but this is work that Nathan Klapoetke, who also recently graduated from our group, did, and in intact brain circuitry, you can activate two separate populations with blue and red light.
All of these tools, you can, again, find on our website, SyntheticNeurobiology.org. We've developed a whole pile of other tools to make these tools easier to use, including wireless devices, ways of using optogenetics in conjunction with functional MRI. Some of our collaborators have developed transgenic animals, and then finally, devices for 3D light delivery, and so we can be more naturalistic.
And just to end on a note that this is all really interdisciplinary if not omnidisciplinary endeavor. We're really working from all these different angles-- physics, and nanofabrication, and optics, and chemistry, and protein engineering, and genomics, and so forth trying to build these tools. But you know, the working model that we have in our group is, let's assume that sometime, maybe 10 years, maybe 20 years, maybe more from now, but not longer than 40 years, the brain is completely solved. How do we get there? Thank you.
Great question. Yeah, how do we get one kind of cell sensitive to light? Let's pick this example here. So what does this mean?
There's a molecule called Cre, and Cre will bind to certain sequences in DNA and delete or otherwise alter the stuff in between the Cre target sites, and there are hundreds and hundreds of transgenic mice that express Cre in different cell types. Basically, the way this works is-- oops. Suppose here is the cell that makes dopamine. It has a gene called DAT, which stands for dopamine transporter, and in the genome, that gene is regulated by a little regulatory sequence called a promoter.
And some factor-- we don't know what it is-- factor x, let's call it, binds to that promoter and turns on the DAT gene, and now this becomes a dopamine neuron. So what we do is we take that chunk of DNA that's the promoter. Usually people will just take some number of bases of DNA in front of the gene and hope that it works, and then you could put Cre after it, and then put this back into the genome. It just goes into some random part of the genome for the most part.
So now what happens? This is in this cell. This is in this cell. We have the promoter, and Cre, and here we have the promoter and Cre.
Well, here in this cell there's no factor x. We know that because if there was factor x, it would have turned on dopamine transporter. So no factor x, no promoter activity. Cre is off.
This cell, though, has factor x, so it binds to this promoter because it's the same as that promoter, right? So Cre is on. That's part 1. You can have Cre on in cells you like, basically by taking the piece of DNA in front of the gene that's on in those cells, pasting it in front of Cre, bring it back in.
OK, now what about part 2? How do we actually get the opsin in there? I'm going to draw a version that is easier to explain, but it's not exactly what is here.
Cre binds to a sequence called lox. Suppose you have a promoter, and this is a generic promoter, just some boring old promoter that drives gene expression, and then you have a stop sequence here. There are various stop sequences that you can get, that you find, that you can put them in.
And here's the opsin. Let's [INAUDIBLE], which is the one that was used for activating the dopamine neurons. Cre will bind to the lox and delete what's in between. So in all the cells that don't have Cre, like this cell down here, no Cre, the stop stays, and so something will try to make this gene, but it'll hit the stop sequence and stop. However, in the cells with Cre, Cre deletes the lox-stop-lox cassette, and now this promoter can drive the expression to the opsin, and now the cell is light activated.
So two parts-- this is a very common strategy right now in all of biology. You have what's called a driver, which you express Cre in a certain cell type, and a reporter, which is a little bit of a misleading concept right here because this is not just reporting. It also lets you control, but they call it reporters. And so by combining the driver and the reporter, you can then express any gene you want in any cell that has a well-known existing gene.
So nobody's used this in people because it's a gene therapy. There are no gene therapies approved in the US. There are people trying to see whether it could be a therapy for patients with blindness, for example. You know, neurotechnologies have been around for a while and haven't always been used properly, you could argue.
Over a third of a million people have had some kind of neural implant for stimulating their neurons so far. And then, of course, millions of people take pharmaceuticals that alter brain activity as well, so I think you're asking two questions. One is, could this be used in people? And the other is, could that go wrong? Right?
So the answer to that first is, for certain forms of blindness, there are three companies that are trying to see whether you could actually treat blindness with these molecules. People who have lost their ability to see, could you install these molecules and help them see again? As far as the "should people use it" aspect, you know, that's a question has to be asked for any biotechnology, I would argue.
You know, people can edit the genome now. People can make stem cells from almost any person. You know, all these technologies have to be thought about, openly discussed, and ethically used. The brain is really complicated, so if you look at this picture, they've turned off the neurons, and the mice fell asleep. They turn off the light, and the mice woke back up, but they're still bit groggy, right?
They go to sleep again, so there's very complicated time dynamics, right? All the cells in the brain are interacting in a very complex way, and people have shown that if you stimulate even one cell in the brain you can change the entire brain state, so it's a very complicated set of questions. The original activator that people used, as you recall, was using blue light, right?
So in this movie, the animal pokes his nose in and he gets a pulse of blue light. The reason that Jaws and Chrimson are important is because they can be activated with red light, and red light can go much deeper in the body than blue light. If you take a bright light and shine it through your finger or something, it looks red, right?
We don't look blue. The red light goes through. So you can perturb much bigger volumes of brain tissue with red light than with blue light.
ED BOYDEN: Chrimson is a molecule that we found through a massive genomic screen of looking at over 1,000 plants, and it's spectrally shifted. So this is current on the y-axis and color on the x-axis, and this is the Chrimson response curve. It's shifted to the red. Were you asking about how Chrimson is different from Jaws?
So jaws is a silencer, right? It pumps negative charge into the cell. Chrimson is an activator, and it brings positive charge into the cell. So you need both.
Jaws and Chrimson have nearly identical spectra, but if you wanted to do two-color perturbation, the way to do that would be to use Chronos to activate neurons with blue light and Chrimson to activate with red light or jaws to silence the red light. Yeah, I think that would work. There are groups already trying such experiments in vivo. Suppose that x years from now the mouse brain is solved, right?
We have a good map. We've recorded the whole thing. We can simulate it. You know, maybe that takes a couple of decades. Who knows, right?
It's really hard to know how much it will generalize the human brain. Certainly, some things will, right? Mice have emotions. They make decisions. They have memories.
On the other hand, it's not clear about language, and higher-order cognition, and all sorts of stuff, so there's a couple of strategies that could be brought to bear. One is, at MIT we're starting to try to get the marmoset going. It's a really small, like 1 kilogram, new world primate, not as smart as the old world primates like macaques, and so on, but you know, it's really tiny.
And so one idea is, can we adopt more model organisms into neuroscience that we can try to bridge those gaps? And that's important, too, because if you-- and even your therapeutics, if you develop a therapy, you only try it on a mouse, and then you go to humans, there's a very high failure rate. Right? We can barely cure anything as a brain disorder goes.
So I think that one strategy is to have more intermediate species, and it could be that the massive convergence of neuroscience over the last 15 years on mice is a temporary thing. We might eventually-- you know, now that it's so easy to put viruses into different animals, and this whole CRISPR business for editing the genome, and so on, maybe what's going to happen is that the mouse convergence between 1995 and, I guess, now will expand back out, and suddenly people will start working on different species.
There's a second question, a second point, though, which is, maybe we should try to develop better human neurotechnologies. So we have a project in its really early stage to try to develop neuroimaging strategies that could image-- without requiring neurosurgery, and without damaging the brain of a human at all, can we record neural activity with single cell, single spike resolution? And this is very early days. I can describe one of the ideas if there's time, but--
All right, here's the idea. Suppose we could put an optical fiber into the bloodstream. It will not enter the brain. It will just go through the bloodstream and sit in the bloodstream.
Suppose we send a pulse of light down the optical fiber. Now, suppose that every time a neuron fires the optical fiber changes so it becomes a little bit more like a mirror and reflects light. So let's walk through that, and then we'll get to the implementation.
Send a pulse of light. Time equals 1. Time equals 2. Times equals 3. Time equals 4.
Oops, a neuron fired-- reflection. 5, 6, 7, 8. So from the time of flight of light, you can figure out where the neuron was. Now, the question is, how do you have a dynamic change that response to a neural activity?
To be honest, we're not exactly sure how to do this. So we wrote a manuscript, and it's posted on the internet, and if anybody wants to join in and collaborate, the more the merrier, but here's our concept so far. This is a cross section of the optical fiber. There's a thin insulator.
The insulator is very thin so that all the neuron's voltage is concentrated over a very thin area. You get a very steep electric field. Then we have our optical fiber, but it's actually made out of silicon, and when you have a very strong electric field next to a silicon piece, you get an accumulation of charge. And when you accumulate charge, you change their factor of index, and when you change their factor of index, it reflects light.
So that's one concept that we're trying to get out there, but you can imagine that you could-- somebody could have an injection. These fibers would be staked into the brain. You could record the neural activity all along the fiber because the time that the light takes to get there, it tells you which neuron it was. Now, we just have to figure out if it's possible to make this, but at least we're trying. They're trying to get into the human neurotechnology changing. Yes?
AUDIENCE: Can we also find a way to turn it on some cells?
ED BOYDEN: As opposed to?
AUDIENCE: Turning it off.
ED BOYDEN: Oh yeah, so that mouse that was working for light, its neurons were getting activated by light. Yeah, let me go back to that slide. So in this slide, the light-driven proton pumps pump protons out, and that shuts the neurons down. The light-driven chloride pumps pump chloride in. That shuts neurons down.
The light-driven ion channels let positive ions in. That turns neurons on, and that's the molecule that we used in the "mouse working for light" experiment. Yeah.
AUDIENCE: My second question is, can we use some of these technologies for learning, for helping learning?
ED BOYDEN: Oh yeah, I mean, many groups, even at MIT, have used these molecules to study learning in mice, and yeah, there's a group here at MIT that they have their mice learn something, and they turn on the opsin only in cells that were used during that task. And then later, they activate those cells with light, and they can cause the learning to transfer.
Then can cause the memory to be recalled. They can cause false memories. Yeah, that's been done. Yeah, literally hundreds of groups have used these tools to study the brain, and they've even found neurons that trigger memory recall, change, shut down, epileptic seizures, all sorts of stuff. All right, well, thanks.
Associated Research Thrust: