Scientist Stories: Karl Deisseroth, Inventing Optogenetics & New Tools in Neuroscience

It's my distinct honor to introduce our plenary speaker, Dr. Karl Deisseroth. He's the Stanford Medicine D.H. Chen Professor and Professor of Bioengineering and of Psychiatry and Behavioral Sciences, and an investigator of the Howard Hughes Medical Institute. He's also a practicing psychiatrist at Stanford with specialization in affective disorders and autism spectrum disease, employing medications along with neural stimulation. So again, you have the privilege of hearing someone who really transfers from the basic fundamental sciences right through to clinical care. Over the last 16 years, his laboratory is recognized worldwide for creating and developing optogenetics, hydrogel tissue chemistry including Clarity, among so many other such enabling methods for deciphering the mechanistic basis of mental health disorders. Last year, Karl was a co-recipient of the Lasker Basic Medical Research Award for his development of optogenetics and for fundamental discoveries regarding channelrhodopsin's principles of function. He was most recently awarded the highly prestigious Horowitz Prize for the foundational contribution of optogenetics transforming our neuropsychiatric research. He has freely distributed these tools and models, training thousands of scientists from around the globe in putting optogenetics to work. Indeed, I would say that his generosity is something that we all here at Stanford so appreciate. He is a very giving investigator. Numerous researchers, including Karl, have used these technologies to study learning, the fundamentals of memory, perception, motivation, mood, and appetite, and detect neural circuit deficiencies causing or exacerbating so many of the mental health disorders we've been talking about today, from schizophrenia, autism, anxiety, depression, and addiction. There is simply no doubt that in the case of Dr. Desiroth, it will be our great privilege to say, "We knew him when." Thank you so much for joining us to hear from Carl today. Thank you.

All right. Thank you, Ruth, for that very gracious introduction. Really very also deeply appreciative of all you do for our community and the department and the medical school and beyond. As I was saying earlier over lunch, a lot of the things that we did really, I think, could only have happened at Stanford for a variety of reasons. From top to bottom, the openness of the community, the lack of hierarchy, the willingness of people to reach across boundaries has been critical in what we've been able to do. I want to share some work. In the first half, I'll talk about some, you know, really molecular type, you know, work. And it maybe is not the cup of tea of everybody here, although some people I know it will be. But I think it'll share the foundation that we can then get to the second half and think about these drug-altered states that are so interesting and so important And going from the molecular side to the systems cognitive and psychiatric side has been really crucial for what we've done. And we're not shy about tackling things that are very hard and mysterious. This is a painting by Andy Warhol that to me evokes dissociation. And this is a phenomenon, a cognitive phenomenon in which different parts of the self that are normally that are unitary become separated, like knowing something's happening and feeling a particular way about it, or one's sense of self and one's sense of one's body. We take for granted that these things are the same, and yet it turns out they're not. It actually takes work to keep them together, and that can break down. Very, very interesting. But to get to these questions in a real rigorous causal way, we have to you really take a very big step back, even billions of years ago, if you'll come with me on this journey to the ancient oceans from billions of years ago, there are really just two chemicals that life uses to harvest energy from light. And one of them is better known. It's the chlorophylls you can see at the lower right. These are what make plants green because they absorb everything but the green. And they're very complex. It's not just this chemical, but it's all the photosystems around it, the proteins that work with what comes out of the chlorophyll that are very complicated. A much simpler molecule is shown here. These are retinoids. These are like vitamin A. And not only are these simpler in structure, they need much less around them in order to work. In fact, a single protein with actually one other partner can make energy from light using these. And so it's thought that the ancient oceans billions of years ago were dominated by microbes, that their membranes were filled with green light activated hydrogen ion pumps or proton pumps. This is a proton right here. Getting pumped out of the cell. And the reason this was the case is, first of all, UV light doesn't penetrate very deeply. So a lot of the volume that's relevant, the green and the light around the green, is going to be available. And so these rhodopsins, as they're called, they were made by bacteria. And they have very broad spectra. They absorb actually through most of the visible spectrum, starting in the blue and going— in some cases into the red. And actually, this might be why the chlorophylls are pushed off to the side, because there wasn't much light left that hadn't been absorbed. And this may be why plants are green today. Now, this is pretty important for us because the breadth of this spectrum, this is not like an information transmission line where you might want to use a very narrow spectrum so you could use multiple different spectra in parallel to transmit information. This is an energy harvesting spectrum, so it's very broad. And this is good to some extent. It makes the proteins very efficient at using light, but it's also bad if, as you'll see, we want to use different wavelengths of light for different purposes. So it requires a little work on our end to make these as useful as they could be. Now the ancient original proteins were pumps. They moved one charged particle across the membrane for every photon that hit them. And that was pretty good, but along the way later came a souped-up version of these called channelrhodopsins, which instead of just moving one charged particle, one ion, actually opened a pore, a hole in the membrane, and let hundreds of ions across. And you've no doubt know, all of us know that the movement of ions is critical for how the nervous system works. The entry of positive ions into the cell will trigger neurons to fire action potentials. And there's a whole family, a whole set of families of these channelrhodopsins. The original Halobacterium salinarum, that's an ancient form of bacteria called an archaebacterium, made this green light-activated proton pump, bacteriorhodopsin. It was identified just up the highway at UCSF in 1970 by a guy named Dieter Oesterheld. These are the pumps, but all these channels, they were identified over a number of years later. Peter Hegemann, my friend here, identified the blue light activated channelrhodopsins from these algae. These are very interesting single-celled algae. They're plants. They photosynthesize, but they have flagella so they can swim through the water and find the optimal light level to photosynthesize. So they actually need to use light for information. And that's probably why they developed these souped-up forms of the proton pump. They need to react very quickly and avoid predators and also find the right light level. Unusual behavior for a plant, but they needed a lot of ions to be coming across the membrane quickly. But these are also just single cells. They don't have space for a lot of complex machinery, a lot of networks of signaling molecules. They had to be very efficient about how they detected and used light. So it was very useful for them to use this single protein that receives a photon and opens up a channel in the membrane. So very, very interesting. That was a blue light activated channel rhodopsin. A few years later, my group here at Stanford found a red light activated form of this from this multicellular algae called Volvox. And we figured out other ways. We did a lot of structural work that I won't dwell on here. We got the high-resolution structures of all of these classes of channelrhodopsin, including one just this year. And we used techniques like electron microscopy and cryo-EM. And this let us see the exact atomic positioning of all the atoms that made up all these classes of channelrhodopsin. Why do we do that? Well, I'll show you. It was really— first of all, it was just really interesting. These are just amazing natural proteins that do this all-in-one job of absorbing light and moving ions, but also getting to know how they worked really well let us change them in very useful ways and make them more useful for neuroscience. And here now we're zooming in as a result of all the structural work. This is that ancient retinal, that vitamin A-like molecule embedded within the membrane of the cell. These are water molecules, these gray spheres on the outside and inside of the cell.. And these blue arrows are the pathway through which the ions flow. And the retinal is embedded within this little pocket. And by changing the environment, by making mutations, we can change the color of light that it responds to. We can change the speed at which it reacts. We can even change the kinds of charged particles or ions that flow across the membrane. And we did this over many years with a great team of colleagues and collaborators.. And we did actually turn a blue light, we actually did take a blue light cation channel and turned it into an anion channel that moved negative ions, and that made it an inhibitor. So we could shut off cells instead of stimulating them. These little spiky deflections in this line here, these are action potentials. These are millisecond-scale deflections in the membrane voltage that are the currency of information flow in the brain. We made red light activated channelrhodopsins by 2011. And that led us by just 2019, just a couple years ago, we finally were able to do the following kind of experiment. We could take a living mouse and we could guide spots of light to individual neurons in its brain. And we could turn individual neurons and groups of them 10, 20, 30 individually selected neurons that we picked, we could turn them on while the animal was carrying out complex behaviors, choosing to do a task like feeding, for example. And I'll give you— I'll show a movie here showing that in operation as we're stimulating these cells by guiding light. These are cells that we've made to express channelrhodopsins. We've given them the gene for this algal protein. Normally, Neurons in the brain don't respond to light. There's no reason to. But when we give them this channelrhodopsin gene from algae, they do. This pore opens up. Ions flow across. And we can see that we're successful because we also introduce a fluorescent molecule that changes its fluorescence when the neurons are active. Calcium rushes into cells, and that binds to this fluorescent reporter, and it makes them glow. So we know we're being successful. And what I'll do, I'll play this movie. And all these rainbow-colored cells, T1 to T10, these are cells that we are stimulating in synchrony with spots of light that we're guiding to them. And in real time in the brain of this living mouse, the 3 gray traces on the top, those are the non-targeted cells that we're not guiding light to. And you can see they're not coming along with this pattern of activity. They're doing their own thing. And what was amazing is we then could see, OK, which are the cells we want to stimulate now that we have this power? And so we thought, well, let's take neuroscience, which is primarily at this level had been a correlational specialty, and let's make it causal. Let's figure out which cells actually matter for behavior. And what you find if you do this sort of experiment is— Sometimes you'll see cells that respond in the brain to something like feeding. You'll see, particularly in this part of the brain that we're in, orbital frontal cortex, some cells start to glow when the animal feeds. Others start to glow when there's a social interaction. So there are feeding cells and social cells. And we were able to ask the question, what if we just stimulate the feeding cells? Or what if we just stimulate the social cells, what happens to behavior, getting to causality, getting to what actually matters and is not just correlation. And what we found is if we stimulated the feeding cells, just 20 or 25 of them, we could make the animal feed more vigorously. If we stimulated the social cells instead, it actually suppressed feeding. The animal ate much less. And so this was pretty interesting. It showed that it mattered that you could pick out the cells, that it mattered that you could guide light to them, and is one example of many things that we've been able to do. But that was in 2019. A lot's happened since then. We've worked on finding new versions of these channelrhodopsins that are even more potent. We've done mining of transcriptomes. We do a very broad collection of DNA from the ocean, and we look at thousands of genes that are rhodopsins. Just like those ancient oceans, there's DNA for rhodopsins that's just still filling the oceans, in fact, much more diverse even than it was back then. But so you can find all the rhodopsin genes you want from the ocean. How do you narrow things down a little bit to find the ones you study? Well, there we used our structural knowledge to say, OK, these particular ones are likely to be channelrhodopsins and likely to be high potency. And so we ended up discovering— a very interesting new channelrhodopsin that we called Carmine. We called it Carmine, the CHR is for channelrhodopsin, and we got it by mining a genome. But also carmine is a deep red color, and this particular channelrhodopsin responds to very deep red and made very big what we call photocurrents. If you flashlight, you get big fluxes of ions ions across the membrane much bigger than the other channelrhodopsins that were red light activated that we knew at the time. And it could be very light sensitive. We could go— this is the trace for carmine compared to the other ones at the time. We could go 100-fold lower in light power and still get very good actuation of neurons. And this opened up a whole new domain for us because we could even start doing non-invasive through skin and even through in the skull recruitment of activity of neurons, even deep in the brain with high speed. So now it's getting pretty interesting, right? You can imagine genetically targeted, deep, fast, precise actuation or inhibition of cells in ways where you don't have to bring in light hardware. You still have to put the gene in, but you don't have to bring in, in this case, light power. So that was the promise. That's why we were so interested and excited by this. We did a lot of tests of Carmine in various settings. We saw that indeed we could activate many cells here. We built basically hologram generators to play in hundreds of individual spots of light into three-dimensional swaths of the mouse brain. All the little red circles are cells that we are stimulating in visual cortex, part of the brain that controls the first steps of visual processing.. And we could reliably activate very large numbers of cells with safe light power. So that was exciting. We actually were even able to make a mouse act as if it was perceiving something that wasn't there. We could effectively play in a percept to the animal. And this was a pretty fun experiment. There was a mouse that was trained to lick for a reward if there was a vertical for visual stimulus, vertical bars, but not for horizontal bars. And so the mice got very good at this. They could make that visual discrimination very well. And then we said, OK, let's take away the visual stimulus and let's just stimulate optogenetically the right set of cells. And we found which cells naturally respond to vertical or horizontal bars. And we optogenetically stimulated them. And what we found is it worked very well. The animal, even without a visual stimulus there, that the animal was able to discriminate whether we were stimulating the vertical responsive cells or the horizontal cells just as well as if it was seeing it. And what was going on in the brain was similar to what we call the internal representation of the activity. We tracked the activity of thousands of neurons in visual cortex, and it was very natural, as if it were a visual stimulus, which we knew wasn't there because we weren't providing it. We were just kicking it off by stimulating a few precise cells. So as far as we can tell, it's a real percept that we've created in the mind of the animal. Now we've been advancing this. I won't dwell on this for the sake of time, but we've now made mouse lines that express Carmine at just the right ratio of the channelrhodopsin and this fluorescent indicator. And we've made and are now distributing mouse lines that make this sort of thing very robust. We can pick out populations of cells. Here we're stimulating 8 groups of cells, 24 at a time. And we can stimulate this 24-cell group, this 24-cell group, this 24-cell group. And at will, it's almost like we're playing the piano or playing the role of an orchestra and look at the behaviors that the animal is expressing. So now you might say, well, that's cool, but it sounds pretty complicated. Playing holograms into the brains of people, right? This is, this is a good science project. And that's really true. It is, it is primarily science. It's really about discovery. It's about understanding how the brain works. And once you understand the causal elements of behavior or of a symptom, that gives you a target. You could design any kind of medication that might target that cell type. Now that you know that cell type is causal. But at the same time, there are simpler forms of optogenetics that maybe are more readily translatable. You don't have to do single-cell resolution spots of light. In a way, the promise of optogenetics is, you know, you could bathe a whole region of the nervous system in light. If you've targeted your channelrhodopsin to the right cells, then only those will be the initial direct responders.. And you actually don't have to do much light guidance. In fact, you could do something pretty simple, like with a fiber optic. You could bring light into the right general area. And if the gene you've cleverly made to be expressed in only the cell type of interest, you could get pretty specific effects. And this is the more bread and butter optogenetics. It works really well. You can make these fiber optics. You can deliver them to— deep in a rodent brain, a primate brain. We've done this with our colleague Krishna Shenoy, and you can start to discover some pretty interesting things. You can— we've dissected what it is to, to be anxious. As we know, with anxiety there are different components to it. There's respiratory rate changes and heart rate changes. There's physiological things that happen when you're anxious. There's also a subjective feeling. It feels bad, right? It's a negative valence, we say. That's a whole separate thing, but it comes with the physiological changes. And also it causes behavioral changes. We, when we're anxious, we avoid risky situations. That's a whole other thing. And using optogenetics, using just a fiber optic, we were able to pick apart these different features. We found that there were cells in a brain region that's part of the broader extended amygdala called the bed nucleus of the stria terminalis, or BNST. And it sends projections, long-range connections, to other parts of the brain— the ventral tegmental area where dopamine neurons live, the lateral hypothalamus that controls a lot of behaviors, the parabrachial nucleus deep in the pons, part of the brainstem where breathing and cardiac signals, among other things, are regulated. And we found that there were cells that lived here and projected here that we could recruit with a fiber optic that controlled the valence of the state, whether it was positive or negative, how the animal classified the stimulus as a good thing or a bad thing, reporting that to us by where it chose to spend its time, where it was getting the stimulus, or not. And another projection to the lateral hypothalamus controlled the risk-taking behavior, the risk tolerance of the animal. So the risk behavior was separable from the subjective, it seemed, valence. And then finally, a whole other projection controlled the respiratory rate changes. So the different components of anxiety could be assembled from their parts. And just using this simple fiber optic, we were able to piece that apart. This has turned out to be very versatile and useful. Our friend and colleague at Harvard, Catherine Dulac, in 2018 published a very interesting paper dissecting parenting, if you can believe that, but in mice. And she used the same basic principle and found that there's a brain structure called the MPOA. And it projects a couple different regions, one being the VTA, another being one called the PAG. Alphabet soup of names. Just think of them as different regions. But very interestingly, she found that one projection controlled the grooming behavior of mice, how they care for their young by licking and grooming, but didn't affect the willingness of the mouse to go and get the young that had been separated from the nest over a little barrier. It didn't affect that go and get the kids. Impulse at all. A different projection controlled that go and get the young stimulus, but didn't affect the grooming at all. So all the parents here appreciate these big fundamental two domains of parenting, keeping them clean and bringing them back when they wander out. We think of this as a unitary state of parenting. In fact, it's assembled from parts, just like anxiety. And these are things that are tunable and can be brought down to the physical elemental components using optogenetics. So that's pretty interesting. But then Carmine changes this also a little bit. We do like the fiber optic method. We can bring in light to a particular region. But the extreme light sensitivity of this new channelrhodopsin means we can even do non-invasive light delivery. And here's an example of that. We can have the fiber— optic completely outside the brain of the animal. And we can target deep structures like the ventral tegmental area dopamine neurons. And using an electrode, we can track how well we're doing. And we can actually drive these neurons, drive action potentials in these neurons without brain-penetrating hardware at all. In fact, you can even use non-invasive injection-based tricks to get the gene in too. We have ways of getting the gene into the brain through an intravenous injection. And this was a technique developed by a former postdoc, grad student and postdoc of mine named Viviana Gradinaro. She trained here at Stanford and is now at Caltech. So you can do completely non-invasive, deep, fast control of neurons now. And that raises all kinds of questions. What is, you know, what are the therapeutic possibilities. Happy to talk about those in detail. We're going to get to the dissociation experiments in a little bit, but I want to tell you about one other thing before we get to that, which comes to our never being quite satisfied. As amazing as Carmine is, we thought we could make it even better, and we thought we could narrow the spectrum with some clever engineering. And let us do very much more advanced simultaneous imaging and control, play-in and readout of neural activity. And it seems like a tall order, right? This is a pretty baked-in ancient spectrum. And indeed, we haven't made it as narrow as we'd like, but we have very consequentially made it more narrow. And we were enabled in doing this by getting the very high-resolution structure of Carmine. And we just published that this year. So we used the method cryo-electron microscopy. And it is a beautiful structure. It's a trimer. It has 3 components that come together. Each one of them has its own retinal. Each one has its own pore, but they come together. The ancient pumps also were trimers, so that's pretty interesting. We did a lot of investigation of how the pore works. I won't dwell on this here, but we were able to see aspects of how it's ion selective. It does move potassium ions very well, sodium also, but doesn't move the calcium and magnesium very well at all, which actually is good. That's useful in some ways. But what I want to tell you about is we were able to really change the spectrum. We were able to look in at that retinal binding pocket, all the residues around it, and we were able to come up with ideas on which ones to change. And we found a couple positions that we were able to change to mutate. And here's the outcome. This new mutant we call redshifted RSCarmine in the dark red. It's less responsive in the blue part of the spectrum, a little bit more in the red. That's not super impressive, but this actually is very useful. That opens up the blue for imaging light. It gives us a pretty free channel of light to use for getting that readout of activity. So we can play in and read out in a really unimpeded way. So getting the structure was really helpful for this. I won't dwell on it for the sake of time, but we've done a lot of work to show this is possible and really enhances this all-optical interrogation of neural circuitry. So what else does this new sensitivity give us? It gives us the ability to take non-invasiveness to new levels. We can now look at computations that cells are carrying out relatively non-invasively as well. So we can actually, in a pretty quantitative way, see what populations of cells are actually computing in real time in a mouse. And this work just came out last week. And we were looking at basically almost like a foraging task, where a mouse has 2 spouts that it can lick water from.. But it's not clear which one has the water. This is a slightly thirsty mouse, and it has to figure out which spout has the water. And there are blocks of time when it's the right spout that has the water and blocks of time when it's the left spout that has the water, and those switch at unpredictable times. So the mouse has to forage a little bit. It has to— when it makes a particular choice, it thinks, well, I think it's gonna be this spout. And that could be wrong, but that's the guess it's taking at the moment. And it's tracking how it's doing, and when it doesn't work, and there's an intermittent failure rate at one spout, so it's not as if it always works, so if you have a failure at one spout, you don't know, is it time to switch or not, or was that just an intermittent failure that happens? So there's some tough decision-making that has to happen. It's like hunting for, for reward, for food, for water in an uncertain environment. So this is obviously very complex and interesting. And we found there's a deep brain structure called the habenula, and in particular, a part of it called the medial habenula that's very much not well understood. But there are cells in there that we thought could be interesting. We were able to find cells that express a peptide called Tac1. This is a small neuropeptide. And these tracked, as time went on, as the animal was more and more successful with a particular strategy, the activity of these cells increased. And this was different from what any other cells were doing. Other cells reported on other signals, like you failed to get a reward. There's a population of cells that signal failure. There's another set of cells that signal when a cue is coming, that the reward might be coming.. But this sort of integration, this tracking of how well things have been done over time, was pretty novel and interesting. Here's an example of what that looks like. Here we're recording the population activity of either these Tac1 cells or another population. These are tyrosine hydroxylase cells, or TH cells. The TH cells are always reporting the same thing. Every time the water droplet comes across many trials, their activity goes up in much the same way. In this kind of plot, the redder or orange, more orange the color is, the more active the cell is. And so these cells always report that the water is available. This other population did, they showed this very interesting pattern over time. Initially they didn't even respond at all to the water, as if they didn't even care that it was present. But over many successful trials, they built up their activity more and more. So they were sort of tracking and integrating how successful a strategy had been over time.. And this is a pretty important parameter to have represented somewhere in the brain. You want to know how well a strategy has been working for a while. And this was very unusual, though. We hadn't seen this before in other cells. So we were able to then test that these cells were causally involved in this behavior. We did some optogenetics to control them. And we found that they were indeed causally involved in making the right decision. And then we did what I mentioned earlier. We did this non-invasive tracking of the computation of the cells. And what we did, we used this non-invasive Carmine approach. So light outside the brain. There's not a penetrating light source at all. But we put in electrodes deep in the brain to track the activity that was going on. And we could actually see using these electrodes in the medial habenula, these TAC1 cells ramped up their activity. As the animal got more and more rewards. And we were able to use Carmine to identify the cells using a very interesting method. Normally with this kind of recording that is shown here, we're using electrodes. Electrodes don't tell you what kind of cell you're recording from. You're just getting a little blip of activity, an electrical pulse of activity. All neurons generate these little pulses. But we did something called optotagging. We were able to stimulate the cells noninvasively using carmine. Carmine had been targeted to the medial habenula TAC1 cells. And we could measure the latency, the delay between our light pulse to stimulate the carmine cells and when we saw those first spikes appear. If that latency was very, very small, we knew that those were directly carmine-expressing cells and not some downstream cell that had been recruited later because there wouldn't have been time for activity to propagate through the synaptic delays in the circuitry. So we were able to actually know for certain that these were the right cells, the TAC1 medial condyloid cells. And we were able to see their computation over time, their tracking of this reward success over time. So that was the exciting thing that Carmine enabled, finding these cells among the thousands of other cells that we could record from. And the first author on the paper, my former postdoc Emily Sylvestrak, she's now at University of Oregon, but she also painted the COVID of the journal. She kind of did a pointillist-style representation of the complexity of cells in the medial habenula. So this is what developing these new methods really helps us do. We can do new kinds of measurements that we were not able to do before. So here I like to often— highlight something that Francis Crick wrote in 1999. So you'll know Francis Crick of DNA double helix fame. And he got interested in neuroscience later in his career. And he actually was quite prophetic about this. Before any of this work started, in 1999, he wrote a very interesting paper in the Proceedings of the Royal Society. He said, what we need in neuroscience is to to be able to turn the firing of one or more types of neurons on and off in a rapid manner. The ideal signal would be light, he said, and probably at an infrared wavelength, much like we're doing with our most advanced methods. He said, this seems rather far-fetched, but it's conceivable. Okay, so, but he was right about all of this, except it wasn't really far-fetched. We were able to do it just a few years later. And the cool thing about this, about his perception, is it kind of conceptually unifies different fields of biology. In genetics and biochemistry, people are steeped in this idea of causation. You do gain and loss of function interventions. You, if you want to know that a gene's important, you knock it out or you provide more of it. You do adding or taking away, and that's how you determine what matters. And geneticists and biochemists are comfortable with this, even though they work with very complex systems like cells, like tissues, like developing organisms, like systems like the immune system. If you do a precisely targeted intervention, even in a complex system, you get— the logic of the system is accessible and can be interpreted. And he was right. Principle translates to neural systems just as well. Even though the brain is complex, it has a lot of feedback and nonlinearities and redundancy. That's true of all these other biological systems too. And if you have precisely targeted gain or loss of function, as we can do now with optogenetics, you can understand the underlying logic of the system. And optogenetics gives us an ability to work even on more time scales than genetic methods provide. Gene knockouts or overexpressing genes, you work on the timescale of hours or days. We can do things on that timescale with optogenetics, but we can also get down to the millisecond scale if we like. And that's crucial to be able to operate on any temporal scale. And just to highlight the complexity, I often like to show this sort of a— horror movie-like picture. This is a fly with antennae— well, legs growing out from its head where the antennae should be. And this fly is called antennapedia. This is a famous mutation that came from developmental biology. People found that there's a position on the fly chromosome— and humans have homologous regions— that if you do gain or loss of function, knock-in, knockout, you can turn up or down the likelihood of forming legs where the antenna should be, or antenna where the legs should be. And this revealed a fundamental principle. This revealed that there's pattern formation that happens, that antennae really are evolutionarily modified legs. And each segment of the animal has an identity, segment identity, and there's a Coordination of genes that work together to define what the identity is, and cells in that segment respond to those signals, know what segment they're in, and then make the right structure. And that whole pattern formation concept was clarified by gain and loss of function interventions. Even though today we don't know everything that causes a leg structure to be formed at the finest level, the logic of pattern formation was elucidated by precisely targeted gain and loss of function, just like we can do now with optogenetics. So that is the background. But now to get to the main topic is the question is what, how can we use this capability to understand some of these very interesting deep questions that relate to our patients, that relate to ancient questions that have puzzled people, intrigued people for a very long time, including dissociation. And I've been working— I see patients who are, who have depression or autism. That's my clinical specialty. We do brain stimulation treatments, but we also give medication and we do therapy. And I've encountered, as many of you will appreciate, that there's We still need to do a lot of outreach in psychiatry to help people understand how physical and material these disorders are and the value of biological understanding to help us understand, to empathize, and eventually to treat with new kinds of methodology these disorders. So one thing I've done is I wrote a book last— it was published last year— where I took some stories of patients I'd seen here at Stanford in the emergency room and different domains of psychiatry, and I put them in the context of modern neuroscience. And there is chapters that deal with autism, with mania, with depression, but also we touch on dissociation. And the different chapters are shown here. The first chapter deals with the grief of bereavement. There's one about mania, one about autism, borderline personality, schizophrenia. And the eating disorder chapter touches on some of these questions surrounding dissociation, which is very common in personality disorders and in eating disorders. So let's talk about this and put it in the context of the human beings who suffer from it. And that was the strategy I took in the book too. Each story starts with the human story and then gets into some of the neuroscience and what optogenetics has taught us about the disorder. So dissociation, as I mentioned earlier, this is where where two normally unified cognitive processes become dissociated, separated. And this is potentially a real problem. This is why people who have, you know, who take PCP, for example, can get into a lot of trouble. They can be having damage to their body, which they're aware of, they're fully aware of it. It's not as if they're unconscious, but they don't care about it. It's not necessarily linked to the typical emotional or affective response. Likewise, people can have these experiences, particularly in trauma or severe stress, where they dissociate. They feel separated from their body. They're not unconscious. They know perfectly well exactly what is happening. It's not attributed to the self anymore. And how common this is— some studies suggest more than 70% of people who have severe trauma will dissociate. It suggests it has perhaps some adaptive value, that it separates the suffering individual from the storm of pain and, and allows still processing of the situation, which may in some settings be adaptive.. And that's, I think, the biggest clue is how common it is in trauma. But it shows up very commonly in PTSD. There are dissociative disorders where it's an integral part of the illness. Certainly borderline personality, it's common. Drugs like PCP, ketamine, and others can cause it. And we've seen patients, as I'll talk about, who experience dissociation as part of their aura of epilepsy, the period of time right before a seizure where they start to feel unusual. But what's interesting is across causes, across etiologies, you can have people express very similar experiences. They will say, I'm in the passenger seat of the car looking at myself driving, or I'm in the audience of a movie Watching it, aware of it, judging it, but without an emotional reaction. So again, suggesting that there may be different causes, but they may be accessing shared circuitry. All right, how could you tackle something like this as a scientist, as a clinician, if you wanted to get down to what was actually happening and lend some some physical understanding to it, which is a major goal. People who have these experiences, they're often very much not understood. People don't— I think people haven't experienced this, don't necessarily relate to it very well, have difficulty empathizing with it. But after our work on this came out, I got contacted by people around the world just just thanking us for showing there was a physical basis to this, which is a big step forward, as you know, in all the domains of psychiatry. So what we did was we said, let's look, we have no idea how this is working. So let's take an unbiased screen. Let's image across the brain and see what we see. And we had to think, okay, there's going to be some trade-offs. What are we going to image? So we said, well, we're going to look over as much of the brain as we can. And this is the dorsal upper part of the mouse cortex. So we developed very— over a number of years, we developed very wide-field optics that let us track across the entire brain with true simultaneity, not scanning something, but really looking everywhere at once at activity. Because I had a suspicion there was something about synchrony that was relevant. So I didn't want to lose simultaneity by scanning some kind of imaging modality. And we gave dissociative agents to see what we would see. And we used these genetically encoded calcium indicators, these fluorescent reporters of activity, to tell us what was going on in the brain. And it was really good that we took this unbiased brain-wide approach because it's something completely unexpected that nobody, including us, predicted, which is that there was an oscillation, a rhythm, that was happening in one restricted patch of cortex called the retrosplenial cortex, RSP. And it showed up at this sort of 1 to 3 Hertz rhythm. It was well delimited by this anatomical patch of cortex, this surface sheet of cells on the top of the brain. It was really only in retrosplenial. Other cortical regions didn't show it. The dissociative drugs caused it— PCP, MK-1, ketamine— but not very potent hallucinogens like LSD, which is known to not have the dissociative qualities that ketamine and PCP have. It was in these deep layers. The cortex actually is not a single layer. It's like a stack of pancakes or crepes, really, very thin layers stacked on top of each other. And we saw it was just in the deep, one of the deep layers, layer 5, and not in the upper layers like layer 2, 3. So it was very specific, very localized, and that intrigued us even more. So we thought, OK, what can we do to understand behaviorally what this means for the animal? So we had to come up with a dissociation measure for the mice. And to do that, we have other methods now that also work. But this was the first thing we came up with, which was a very safe, easy-to-use, non-damaging test. But you had to do something that mattered to the animal. And so we have— there's a platform. Plate, it's too warm for the animal. There's a reflexive paw flick when the paw is opposed to it. And then there's licking of the paw to cool it. So that's the response. And we thought this— you can't control reflexes, but maybe this more prolonged response would give us something analogous to what people experience on PCP or ketamine. And that's pretty much what we saw. So even at pretty good doses of ketamine where the animal was still conscious, this paw flick, this reflexive stimulus detection was always present, was unchanged. But at a particular dose and higher of ketamine, the prolonged self-protective, what we call an affective or emotional response, was absent. And this was the same dose at which the oscillation appeared, where this rhythm appeared in retrosplenial. —to us suggested these were related, possibly causally. This was still correlation at this point. We didn't know that the rhythm was causing or was playing some causal role, but it was definitely suggestive. And so then we did gain and loss of function interventions of the rhythm in order to see what happens to this behavior. And so here, this logic of adding and taking away, what really matters, was crucial. To take away the rhythm, we actually did a little exploratory work. We found that there was a particular pacemaker ion channel. It's also expressed in the heart. That's part of where the heart gets its rhythm. It's called the HCN1 channel. And we noticed in our screening that it was highly expressed in retrosplenial cortex, but not in nearby visual cortex. So we said, OK, maybe this is acting like a rhythm generator. And we did some— genetics to knock it out. And if you knocked out the HCN1 channel from these cells, you no longer had the ketamine-induced rhythm. This red is the post-ketamine rhythm, and that was gone if you knocked out this, this pacemaker. So you take away the rhythm, you can do it. What happens to behavior? Well, amazingly, this removed the dissociation behavior. Animals on ketamine recovered their self-protective affective response. So on ketamine, normally not carrying out the self-care behavior. Now, if you knock out this pacemaker channel that generates the rhythm, the self-care behavior is back. So that was interesting. So loss of function, taking away the rhythm, takes away the dissociation. And then we used optogenetics to provide the rhythm, to provide gain of function. And add the rhythm in to see what happened to the behavior. And we used a combination of optogenetic tools to stimulate with blue light and inhibit with orange light. And that gave us the ability to give rhythmic activity to these cells in the absence of ketamine, just providing the rhythm. And what we found was that the stimulus detection was maintained if we did this to retrosplenial cortex. But the self-care behaviors, the affective response was substantially reduced. And that only happened if we did this in retrosplenial cortex and not in somatosensory cortex. So this was gain of function. Gain and loss of function both suggesting this is causal in the behavior. OK, so now comes to the human case. And this is another kind of only at Stanford moment. We were, for a number of years, I've, I've run a little, uh, what I call is a clinical imaging and stimulation subgroup. And we invite everybody who's interested to come to this group. Pre-pandemic, we had sandwiches, and we're hopefully, we'll get back to that soon. But neurosurgeons come, neurologists, anesthesiologists, psychiatrists, and we and our students were talking about these mouse ketamine results, and one of our neurosurgeons here, Jamie Henderson, said, you know, we've got a patient on our epilepsy monitoring unit here at Stanford Hospital who's dissociating as part of the aura of his seizures. Before his seizures, he's experiencing symptoms. And it was actually— that's where this quote comes from, is from that patient. He's— when he Just before his seizures, he experiences dissociation. He described it in one case this way: I'm listening to two parts of my brain speak to each other in a way that the third part of my brain, which I considered to be me, was able to listen. Okay, how cool is that? And this is, you know, just amazing to be able to talk to such a person, to know they're getting the right clinical care. What happens is people come on to the epilepsy monitoring unit to map the seizure origins. These are patients who suffer from intractable epilepsy. Medications aren't working, so they come in to get these electrodes placed deep in their brain, all across the brain. It's kind of an unbiased electrical screen to see where the seizure is starting. That guides the neurosurgeon to know what to take out. So a person was getting the clinical care needed, and yet at the same time we were able to talk to this person and then go and look at the recordings that were already being taken clinically in that patient. So we did that. And we saw a pretty interesting thing. There was a 3-hertz rhythm, and it was in the human homologue of retrosplenial cortex, the same ancestral region that was oscillating in these patients, in these mice that had gotten ketamine. And it's deep. This is probably why surface EEG had not picked this up before, but deep, deep in the brain. And just because there were electrodes in this patient that happened to be close enough to the spot, we were able to see it. And the rhythm appeared only in these— in this particular patient, an epilepsy patient. This rhythm appeared when the patient was dissociating. Also, interestingly, talking about gain of function, part of epilepsy mapping is to not just record but also stimulate, because you want to see, can you trigger seizure-like activity? When these patients come in the hospital, actually you want them to have a seizure. You want to see where the seizure is starting. And so they actually come in and they spend a week sometimes in the hospital to have a few seizures. We want this to happen. And also there's some stimulation that's done electrically to see what can be elicited. And in this patient, when stimulation was carried out, this triggered dissociation, but only when these 3 Hz oscillating regions were stimulated electrically. And there was a beautiful description the patient gave in response to stimulation at one of these sites. He said, it's the same way a pilot can lose control of a plane, forced out of the cockpit still see what's happening to the whole plane. That's kind of what happened. I got pulled out of the pilot's chair. I could see all the gauges. Okay, that's stimulation at this, at this site. It was only when we stimulated, when the epileptologist Joseph Parfisi simulated these particular sites with this 3 Hz rhythm that were capable of generating this 3 Hz rhythm, that's when the symptoms appeared. So this gave us very substantial confidence that we were really working with ancestral conserved circuitry, again pointing to the likely adaptive value of this kind of thing. And what we did then is we brought it back to the mouse. And we said, let's figure out what's going on. And this is work we're doing now. We're doing a lot of mechanistic work. We have some ideas. We think the retrosplenial cortex is connected connected to some, we know it's connected to some deep structures but not others. Deep in the middle of our brain there's a structure called the thalamus, which has a lot of different parts to it. Some connect heavily to retrosplenial cortex, others don't, and they connect to other parts of cortex. And there's a lot of lateral inhibition in the thalamus. This is part of how perceptions and actions become unitary. We only do one thing at one time, we only experience one thing at one time.. And we think these thalamic regions are coming along with retrosplenial cortex because they're tightly wired to it, and then others are kind of pushed out of synchrony because they're not tightly connected. And then you have a situation where whole swaths of the brain are not active at the same time and not able to form joint representations of the same thing at the same time. So we said, OK, interesting hypothesis. Can we test that? Well, we came in with these electrodes. In mice using this hypothesis, guiding our electrodes to record from these different nuclei. And it looks like that's what's going on. So this is a cross-correlation plot, different brain regions correlated with each other before ketamine, on ketamine. On ketamine, retrosplenial cortex, red is more correlated, blue is less. Retrosplenial cortex becomes more correlated with itself, which makes sense. As all the cells are oscillating together. These two thalamic structures, AV and LD, that are known to be tightly wired to retrosplenial cortex also came along with it in that rhythm. But this one, AM, that's known to not be tightly connected actually was forced into an inverse correlation. And we could see this in the raw data. Here are these AV and LD structures coming right along with retrosplenial in green in the rhythm. But this one actually, it inverts correlation, it actually is anti-correlated. And so parts of the brain are actually forced into different regimes of different temporal regimes of being active. Nothing's being shut off. The brain is— the person's not unconscious, the animal's not unconscious, but you no longer have the ability to form a joint representation. And this insight could be general. This could be relevant not just to dissociation, but it could be relevant to more broadly, what becomes part of a conscious state, a state of perception, a state of perception, a state of action. There's a lot of complexity at the thalamocortical circuitry, but this principle we think, and we're testing this, could underlie a lot of what goes into making us have an internal state of the moment. So I'll end. In the next minute. This is to be able to come to this level of having some ideas for how this mysterious state can come about and to have it be rooted in these very beautiful ancient proteins and coming to get high-resolution structures of them and understand them as molecules is something that's been a really amazing journey. Really grateful to all our supporters here, many of whom are in the room, and colleagues and friends also, many of whom are in the room, and especially the students and postdocs, just brilliant folks. A lot of them are highlighted here. For example, Yoon Kim is a bioengineering student who led the structural work. On the dissociation side, Samvasuna, Isaac Calvar, and Ethan Richman, and on the Carmine side, Tim Machado, Jim Marshall, and Sean Kieran. Emily Silvestrak did the pointillist art as well as leading the cell paper.

And I'll wrap up there. So thank you. Thank you so much. You've captivated all of us. Absolutely tour de force. I want to add another title to your many, which is poet, poet, psychiatrist, bioengineer. Will you accept that? Beautiful way that you phrase the science. But thank you for joining us for such a spectacular plenary. We will have time for questions, and we may, again with your indulgence, shorten the break so that our next Speaker also has time. So let's look at the microphones at the back.

Who has questions?

And then I'll look to Rachel to see if we have any on chat. Yes, we have an attendant that first said, fascinating talk. Do you see alteration of rhythms linked to dissociation persisting to some degree after the acute acute effect of ketamine.

Could this alteration be linked to therapeutic effect of ketamine?

Sorry, alteration of what? Can you just restate that? Yes.

Do you see alteration of rhythms linked to dissociation persisting? Persisting.

I think it's the retrosplenial rhythm persisting after ketamine. Oh, okay. I see. The question is, I think, do we see after ketamine, is there persistent change in the retrosplenial rhythm. We have not seen that in the mice, no. It seems to be a very acute effect. And actually, one thing we're doing now is we're actually going on the epilepsy monitoring unit and giving human beings ketamine. And this is pretty interesting, and we're not seeing any lasting effects, no persistent oscillation there either. Okay. Walter, you have a question? I think you just answered my question. I was going to ask what's next. So after the ketamine study, what's after that, I guess? Well, so we hope to— ketamine is particularly good because it's a very safe way to start this kind of work. It's FDA-approved for multiple indications. It has a very short duration of action. We would like to— explore other relevant agents in this regard that alter internal states in a similar context on the epilepsy monitoring unit. We also want to extend beyond dissociation as our measure. I'm interested in autism and depression. There are wonderful colleagues that we have in the pediatric hospital. And autism and epilepsy are highly comorbid, so we're interested in looking at what the internal processes may be relevant to social cognition. We can do this in adults who are autistic as well as in kids. So the paradigm could extend well beyond dissociation. Incredibly exciting. Hi, thank you so much for the talk. One question, I guess my understanding of ketamine is not everybody who takes it reports experiencing dissociation. I was wondering whether the rhythmicity that you're noting, particularly in the human subjects, is sort of related to the degree of, like, experience of dissociation, and maybe relatedly, whether the rhythmicity is, like, specific to the dissociation or some other aspect of the experience of the drug. Yeah, great questions. That's essentially what we're testing now. We've done a few patients. We have to do many more. We don't control where the electrodes go. So we have to build up, we think, probably about 50 patients before we get enough coverage to really answer that question well. But I will say, you're right that if you look at the package insert for the nasal spray ketamine that's given for depression, about 2/3 of patients, even with that mode of delivery, experience dissociative symptoms. Um, and, and, uh, so it's— and we've kind of seen that there are a couple patients, we've done about 10, about 2 or 3 did not have significant dissociation. We're seeing, yeah, probably 70 or 80% are common enough, but it gives us enough that we can— there's a range of symptom expression, and it'll give us the leverage to do exactly what you're saying. What correlates best with the symptoms? What which brain activity motifs are most informative about the symptoms. Thanks.

Do you want one more? Yeah?

You got to go?

Rachel, do we have one quick— I know you need to head out as well.

So can we take one more quick one? We do. Does Carmine show potential promise for non-invasive stimulation in brain surgery, or groups like children who have more potential for the light making it through the skull?

Yeah. So carmine is so light sensitive that in mice, the deep structures are activated without any light guidance hardware. In people, of course, the brains are much bigger. The skulls are much thicker. I don't think we'll have that concern in the brain per se. But actually, we're thinking about using that for maybe a peripheral nerve indication. You could imagine turning down pain fibers in the digits or limbs, for example, using the non-invasiveness of light delivery as a potential strategy. Thank you so much again. Fantastic. Thank you.

Thank you, Tom. Good work.