EunJung Hwang, Neuroscientist and Leading Researcher in the Mechanism of Long-Term Learning

Growing up in South Korea, EunJung Hwang loved watching the American television show The Six Million Dollar Man. This hit science fiction program starred Lee Majors as Steve Austin, a former NASA astronaut and test pilot who suffered serious injuries in flight and who was subsequently "rebuilt" with bionic body parts. Sparing no expense (hence the title of the show), Colonel Austin was soon endowed with capabilities that made him something of a bionic Superman who was far stronger and faster than any "natural" human. In the best traditions of science fiction, for Hwang, the show was not simply fantastical - it was inspirational, and it put her on a path where today she is one of the world's preeminent researchers studying the operations of brain circuitry and the underlying mechanisms that direct long-term learning. She operates at the forefront of neuroscience which approaches the brain as a network that can be both studied and manipulated according to certain bionic approaches that are not so far from the fantastical premise that brought her favorite childhood TV show to life.

At Seoul National University, Hwang embraced electrical engineering at a time when bioengineering had not yet matured to the point of being a discrete program. That limitation served Hwang well; mastering electrical circuitry proved to be exactly the training she needed to gain admission to Johns Hopkins for her doctoral work with Reza Shadmehr, whose lab was focused on theoretical motor control. During her dissertation research, Hwang developed the idea that the movements that one "desires" at the neuronal level should be thought of as a gain field, where the physical positioning of one's body centrally influences how neurons respond to external stimuli. That idea, she soon learned, aligned her closely with the pioneering work that Richard Andersen was doing at Caltech. As Hwang relates in the discussion below, from the moment of her arrival to interview as a postdoc in Andersen's group, she felt something quite common among scientific visitors experiencing Caltech for the first time: this was a place in which the joy of doing science was pure and unimpeded.

During Hwang's time at Caltech, she made key contributions to the Andersen lab's revolutionary work on neural prosthetics. At that juncture, Andersen was laying the basic science groundwork that would turn science fiction to reality: capturing the neuronal basis of intent and translating that to robotic arms that could perform manual tasks. This meant that for a paralyzed person, the act of robotically grasping and then drinking from a cup of coffee could be achieved just by thought. In her subsequent postdoc at Takaki Komiyama's lab at UC San Diego, Hwang arrived at a very exciting time. Technical advances in microscopy had allowed for in vivo imaging of neuronal activity; in other words, for the first time, researchers could study neurons in their natural living state. Hwang seized on the opportunity to "follow" 100 neurons for two months, and that experience was as fascinating as it was foundational.

Today, as an assistant professor as Rosalind Franklin University of Medicine and Science, Hwang leads a dynamic lab that crosses effortlessly between the cutting edge of basic neuroscience and a range of applications and techniques geared toward translational breakthroughs. Toward the end of the discussion, as Hwang reflects on the impact and value of her time at Caltech, she emphasizes Andersen's mentorship and the value of his focus on the posterior parietal cortex, which is a region of the brain that directs a wide range of neurosensory and decision-making functions. Hwang is open to new areas of research for her young career, but she is certain that the post parietal cortex will always remain her focus.

DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It is Wednesday, March 20, 2024. It's my great pleasure to be here with Dr. EunJung Hwang. EunJung, it's wonderful to be with you. Thank you so much for having me.

EUNJUNG HWANG: Thank you, thank you. It's an honor to speak with you.

ZIERLER: To start, would you please tell me your title and institutional affiliation?

HWANG: Sure. I'm an Assistant Professor in the Chicago Medical School at Rosalind Franklin University of Medicine and Science. We are located in North Chicago in Illinois.

ZIERLER: Tell me about Rosalind Franklin University and its connection with the Chicago Hospital.

HWANG: We do not have our own hospital. Our medical students receive clinical trainings in afflicated hospitals such as Northwestern Medicine, Captain James A. Lovell Federal Health Care Center, and Advocate Health Care. Our university started as the Chicago Medical School in the early 1900 to provide medical education to working class students. They had to work during the daytime, so the Chicago Medical School provided nighttime medical education for those students. And then, gradually, it incorporated other health professional education programs, such as Dr. Scholl's Podiatric Medicine, the Nursing school, and the Pharmacy school. In 2004, we changed our name to Rosalind Franklin University of Medicine and Science, in honor of Dr. Rosalind Franklin who's famous for her X-ray defraction images of the DNA helix.

ZIERLER: Being in a medical school, are you working at all in a clinical environment? Do you work with patients?

HWANG: I do not. I'm a basic science faculty member, so I do research in neuroscience. However, I have interactions with medical students. I haven't started teaching major courses yet because I joined only three and a half years ago, but I do have small group discussions with the medical students, talking about topics that are clinically relevant. But I do not have direct clinical work or collaborations yet.

ZIERLER: How big is the neuroscience program at the Chicago Medical School?

HWANG: I haven't counted, [Laugh] but we have an institute called the Brain Science Institute. And within that umbrella, we have about 15 faculty members who are working on neuroscience topics, including stress, anxiety, or the neurodegenerative diseases. And then, folks like me study the brain more at a circuit level, focusing on areas such as breathing, motor functions, or decision-making.

The Brain is a Connected Network

ZIERLER: Let's move onto the major research questions that you're focused on. First of all, anatomically, what areas of the brain do you study?

HWANG: The main brain area that I study is the posterior parietal cortex, which I started working on with Richard (A. Andersen). But I am expanding from the posterior parietal cortex, of course, because the posterior parietal cortex, PPC doesn't work alone. The brain is a network, so I'm studying other interconnected regions surrounding the posterior parietal cortex.

ZIERLER: What are those regions? What are the connections?

HWANG: One of the interesting findings I discovered in the past is that there are multiple projection pathways from the posterior parietal cortex. It's not just going to a single brain region; it goes to many brain regions. Depending on which brain regions it projects to, it seems to route different information. It's an associative area, so it's involved in many different functions. I couldn't study all those connections, but I focused on two major connections, one going to the striatum, part of the basal ganglia, and the other to the frontal cortex, specifically the secondary motor cortex. I found the pathway from the posterior parietal cortex to the striatum is important for decision-making, while the other pathway is important for the control of movements. Both functions have been implicated as major functions of the posterior parietal cortex, but I found that downstream routes or pathways for the two different functions are segregated. I've been studying these pathways, elaborating on what kind of information is communicated between the posterior parietal cortex and the secondary cortex, and the posterior parietal cortex and the striatum, using some advanced optical techniques. I switched to mice from monkeys, so I can use all those fancy techniques now. [Laugh]

ZIERLER: You mentioned mice and monkeys. Do you also work with human specimens? Do you work on the human brain?

HWANG: Not at the moment, but recently, I found something very interesting that's potentially translatable to humans. I'd like to form collaborations with people who are working with humans so that we can connect what we're finding in rodents to humans. Particularly, about aging. We found that like humans, when mice become older, they become more stubborn when they're making decisions. When they are exposed to new tasks or a new environment, they don't explore much. They stick with one thing over and over. Consequently, they don't seem to learn a new task as well as young mice. We know that the posterior parietal cortex has been linked to the regulation of exploratory behavior, at least in decision-making contexts, so we thought, "Perhaps we can stimulate the posterior parietal cortex, then we can promote the older mice to explore. Let's see if that leads them to learn better."

Indeed, that's what we found. By stimulating the posterior parietal cortex in old mice, we made them explore more, and they learned just as much as the young mice. That was a really exciting result, and we wanted to see whether that was translatable to humans, which would promote aged individuals to explore more. In various ways, not just brain stimulation. Perhaps, we can change the environment or do some behavioral modification training such that they can explore more to learn new tasks better, then they can make better decisions in situations where their decisions should be adaptive rather than just doing the same thing over and over.

ZIERLER: The term brain circuit, what do we mean by that exactly? Is it really like an electrical circuit, or is it something different?

HWANG: It's not exactly the same, but there are a lot of analogies. There are elements of the brain circuit called neurons, neurons are brain cells and they are connected with each other through a structure called a synapse. There is information transferred from one neuron to another. In electrical circuits, there's current flowing from one element to another, such as transistors and resistors. There's a transformation that's happening depending on the way the elements are configured. The connections in the brain are also configured in a way that allows specific transformation of information. I think there's a structural analogy in the element-to-element connections, and also a functional analogy in the information processing and transfer between elements. I think, in that way, we call the brain a circuit.

Advanced Optics for Brain Circuitry

ZIERLER: What are the tools or technologies you use to study brain circuitry?

HWANG: In mice, I am lucky enough to use fancy advanced optical techniques. I use a technique called two-photon calcium imaging. When I was working with the monkeys, I used arrays, multiple electrodes. I got to use up to 16-channel arrays. So 16 small needles go into the brain, and we detect electrical signals at the tips of the electrodes. We learned a lot using this technique, but it was very limited in the sense that you can get only, if you're lucky, a few hundred neurons. But now, with the optical techniques like two-photon imaging, we use microscope to see the actual neurons. How we measure the activity is by using a calcium sensor, which changes its fluorescence signal depending on the concentration, the amount of calcium inside the cells. Calcium inside a neuron increases when the neurons is active. When a neuron is active, there's a lot of calcium inside the cell, which intensifies the fluorescence signal.

We just place a fluorescence microscope above the brain, and then we can see individual neurons, and detect their fluorescence changes. From these changes, we can infer their neural activity. When we're doing that, we can monitor up to thousands of neurons at the same time. Those neurons are all active, they're all talking to each other while aniamals are behaving. So the two-photon calcium imaging increases the throughput, in terms of the number of neurons simultaneously monitored. Another fascinating benefit of this optical technique is that we can follow the same set of neurons over days and months, so we can study how these neurons change with time, which is important for studyng learinng and memory, but very difficult to do with a traditional technique. That's because we often lose electrical signals from the same neurons over days when using just electrodes. With optical techqnique, we are able to study populations of neurons in a longitudinal way to study the function of the brain in a more holistic way. [Laugh]

The Projection of Neurons

ZIERLER: When you talk about neurons projecting information, physiologically, how does a neuron project? What does that mean?

HWANG: Neurons have long processes called axons. Axons can travel a long distance, not just within a small area. Not all neurons project long distance, but many neurons project long distances to different brain regions. They make synapses with neurons in a remote area, so they can excite or inhibit the neurons in the remote destination. Those are projection neurons. We can study projection neurons specifically. There are traditional ways to study or identify projection neurons in primates as well, but that was very, very difficult technically. It used antidromic stimulation. You put an electrode in a projection destination area and another one in the source area, then you stimulate the axon terminals in the destination, and see whether you can detect spikes that are backpropagating. Finding projection neurons that way was very time-consuming. You can find only one or two at a time. It takes a long time to study a substanal number of projection neurons.

What projection neurons are encoding is the information that's moving from the source to destination area. That's information transfer. With optical techniques, you can study many projection neurons at a time because you can label all these projection neurons using retrograde tracers. We inject, say, a retrograde virus that is taken up by the axon terminals in the destination area. The viruse moves up from the axon terminal to the soma of the projection neurons in the source and label them with flouresence. So we can visually identify which are projection neurons under a microscope. Let's say we inject retrograde virus in the striatum, and then the virus moves up through the axons to the somas so that these somas glow with flouresence. When we look at the PPC, we can see which neurons are projecting their axons to the striatum. Then, we can measure their calcium activities using the same microscope to examine what kind of information these neurons are encoding in comparison to the surrounding non-projection neurons. That way, we know what specialized information is going to the striatum. That's how we found projection target-dependent functional division in the PPC. We found different information is encoded by striatum projecting PPC neurons and secondary motor cortex projecting PPC neurons.

In addition, we can manipulate these projection neurons selectively to examine their causal functions. What if you silence these neurons selectively? What happens to the animal's behavior? By answering these kinds of questions, we could conclude that striatum-projecting PPC neurons are more involved in decision-making, wheras those projecting to the secondary motor cortex neurons are more related to the control of movements.

It's a bit of an abstract level of description, but those are the techniques to study information transfer from one brain area to another, linking the propertpies of information coding from the imaging experiment to causal effects observed in the manipulation experiment.

ZIERLER: When you talk about neurons projecting information upstream and downstream, what does that mean? Upstream and downstream relative to what?

HWANG: I talked about the PPC and its downstream areas, the striatum and the secondary motor cortex. I'm glad you asked me this question because my lab has been studying an upstream area of the PPC, expanding on the earlier studies about its downstream processes. What we have found so far is that the PPC neurons projecting to the striatum receive prominent axonal projections from the cingulate cortex, CC. Axons from the cingulate cortex do not make synapses with all PPC neurons equally, but they are more likely to go to PPC neurons projecting to the striatum. Now, we have sort of a neural arc connected in a chain, from the CC to the PPC to the striatum. Information flows from the CC, to the PPC, and to the striatum, like a stream. The PPC is downstream to the CC, but upstream to the striatum.

I am hypothesizing this neural arc is important for decision-making, and I have a specific idea in what sense it's important for decision-making. When we make decisions, many things affect our decisions, such as external sensory stimuli, past experiences, cognitive strategies, internal physiological states and moods. One of the decision factors that I study is recent experience, our history of choices and outcomes in the past affects how we make decisions. Different people use this past experience in different ways. I think that's one of the reasons why you and I make different decisions when encountering very similar situations. I hypothesize that the neural arc from the CC to the PPC to the striatum is involved in this idiosyncratic history-dependent decision-making; this arc determines how an individual uses their past experience when making decisions in a certain situation. More specifically, in this neural arc, information about past choice and outcome flows and then fuses together, forming history-dependent decision bias. But we don't fully understand how such information flows and is processed in this neural arc. To answer that question, we're currently studying CC neurons that project to the PPC using similar techniques we used to study PPC neurons projecting to the striatum.

ZIERLER: Is modeling or computer simulation important for your experimental work?

HWANG: Modeling is important because not all factors that affect decision-making are external variables we can observe. We have to indirectly infer those factors from the animal's choice behaviors, in addition to what we can control or observe outside. Those are internal variables that are not directly observable. One way to infer internal variables is to create a model of how we make decisions with internal variables such as choice history and outcome history. We put these internal variables in a functional model that integrates history inforation, counting how many choices we made in the past, what are the outcomes of those choices, and fusing them together to form a decision bias. By tunining parameters of such a model, we can build a decision-making model that reproduces the choice sequences observed in animals or humans. And if we have a model that reproduces these choice sequences very well, then one way to use that model is to find whether and how neurons encode the internal variables of that model.

The Centrality of Optogenetics

ZIERLER: Can you explain optogenetics and why it's so important for your work?

HWANG: Optogenetics has been very important for examining the causal roles of speicifc brain regions or specific neuron types such as the PPC neurons projecting to the striaum. Even before optogenetics were avaible, people were inactivating or stimulating brain regions eletrically or pharmacologically, to study the consequences of neural manipulation on the behavior. But those traditional methods have limitations.

HWANG: For instance, we can inject a GABA agonist in the brain to inacativate neurons. But we cannot precisely control the timing of inactivation. Once the GABA agonist is in the brain, its effect lasts for a couple of hours, so typically behavior in a whole session is affected. Your control data may be from the day before or after, but not within the same session. So the day-by-day behavioral fluctuations can contaminate your interpretation. That's one issue with the pharmacological manipulation. Another issue is that you can't really silence specific neuron types. All neurons in the injected area are silenced. You can't silence just the PPC neurons that project to the striatum, for instance. Given that different neuron types within the same region play different roles, neuron-type specific manipulation has become very important.

With electrical stimulation, we can control the timing of manipulation precisely. You can manipulate selective trials within the same session and eliminate day-to-day fluctuations as an explanation for the change in behavior. But the issue with electrical stimulation is that you're not stimulating only the neuronal somas in the region. You're stimulating all the processes as well, including incoming axons in the region. So, the results are ambiguous. Is it because you are manipulating the neurons in the region or manipulating the axons coming into the region?

One example of such ambiguity is the effects of deep-brain stimulation, DBS. For Parkinson's disease, the common DBS target is the subthalamic nucleus, STN, so electrodes are implanted in the STN, to alleviate motor symptoms of Parkinson's disease. While DBS has been used as a viable therapy, its mechanisms have been elusive. It was found that therapeutic effects of DBS is not through the neuronal stimulation in the STN but the stimulation of the axons that are transferring information to the STN. This finding was made poassible through optogenetics which enabled the investigators to stiimulate selective neuron types and compartments.

For me, using optogenetics, I could inhibit only the PPC neurons that project to the striatum or to the secondary motor cortex by expressing opsins only in those neurons. Light goes everywhere in the PPC, but only those neurons that have opsins, which are PPC neurons that project to the specific target, were inhibited. So I could infer differential functions of the two downstream pathways.

ZIERLER: Of course, you're involved in basic science. I wonder if you can describe, though, what are the connections between this research and neurodegenerative diseases of the brain?

HWANG: As I said earlier, we found that PPC stimulation in aged mice helped them to be more exploratory, and then we were able to make them learn better in a cognitive decision-making task. And then, there's evidence out there that the human brain also undergoes changes in the posterior parietal cortex, suggesting that the human PPC circuit is also compromised with aging. We try to translate what we've found in mice to human aging. That's one condition that we'd like to expand on. Additionally, many neurological or psychiatric diseases present with decision-making problems, such as addiction.

Addiction is a decision-making problem. I talked about how different people make different decisions. We know that even if people have the same first-time exposure to an addictive substance, not everybody becomes an addict. Some people are more prone than others. Perhaps the way they make decisions even before the drug exposure is very different. We can study the decision-making patterns and brain activity to find the factors that seem to be linked to addiction. If we can pre-screen those found behavioral traits, we may help individuals in a preventive manner, warning and heling those who have this tendency to develop addiction to substances or other things. I don't know if you want to go into the manipulation of brains already.

ZIERLER: Please.

HWANG: Perhaps those with the behavioral traits prone to additction may have pathological features in their decision-making circuits even before the significant addiction is developed. We can think about how to modify their brain circuits such that it doesn't become more pathological, or we can reverse the pathological features.

Applied Research and Implications for Task Learning

ZIERLER: What are some of the drugs or therapies where your research might ultimately contribute to helping people?

HWANG: There are special things about humans we can't really do with animals. Humans can attempt behavior modifications through mental exercises and communications. I think that's one thing we should pursue with humans first. I talked about exploration. Maybe we can devise a mental exercise or environmental modifications that promotes exploration in old individuals suffering from maladaptive decisions and learning deficit. If that's not enough, another thing we can consider that's less invasive is something like transcranial stimulation, magnetic stimulation or current stimulation targeting the posterior parietal cortex. When they have to learn a very critical new task, maybe we can stimulate their PPC to help them learn better. People don't have to be stimulated all the time, but there might be some important tasks they would like to learn and choose to get stimulation therapy. That's a less invasive procedure we can try.

And drug-wise, we were expanding our aging study to find the gene expression changes in the PPC that happen with aging. We'd like to find the molecular signatures of aging in the PPC. If we find signatures that are robust with aging and affect the way PPC neurons function, then we may be able to develop drugs that are targeted to that found signatures. I think it's a little too early for me to say what those are right now, but that's one thing that we are trying to do. We will be profiling the gene expression changes in the posterior parietal cortex with age so that we can find the molecular targets we can use in the future.

ZIERLER: What is your interest in long-term learning, and what does that mean by long term? What is the time scale?

HWANG: Many things in life involve long-term learning. For instance, when you are learning tennis, it takes many, many practices, over if not months, then years. These kinds of things are long-term learning. The one thing I've found that's very interesting in motor learning is that in the beginning, when you're learning a new motor skill, the motor cortex is very important. But once you learn a new motor task really well, after, say, two months, the control of the well-learned motor skill can move out of the motor cortex to somewhere else. It's stored somewhere else, such that when the motor cortex is inactivated after a motor skill is established, it doesn't do anything. While in the beginning, the first or second day of learning, without the motor cortex, the animals couldn't perform that task. The motor cortex seems to be very important for learning a new task in the beginning, but once the skill is established, it doesn't seem to be necessary.

Meaning that gradually, the control of the learned movements move out of the motor cortex. I think that's very fascinating in terms of the translational opportunities. For instance, when people have strokes near the motor cortex, they lose control of their body. But there might be some well-learned skills residing outside the motor cortex. We may have just lost the way to trigger those motor programs that are intact. If we knew how to retrieve those motor skill programs in the brain that aren't affected by the stroke, maybe that could be used for rehabilitation in therapeutics. We don't know where exactly the motor program is. We don't know how exactly we will retrieve them. But that's another avenue that I'd like to pursue, to study where these well-learned motor programs are stored and how we can trigger these motor programs, even if the motor cortex is damaged by strokes or tumors. The implication of long-term learning in the brain circuit is that there might be multiple copies of motor programs for well-learned behaviors across different regions and maybe we can use this for therapeutic purposes.

Dreaming of Bionic Humans

ZIERLER: Let's now go back and discuss your education at Seoul National University. I asked you about the potential metaphor connections between electrical circuits and brain circuits because you studied electrical engineering in Korea. I wonder if you can explain if you saw a connection ultimately between electrical engineering, and that's what got you into neuroscience.

HWANG: The interesting thing is that I actually went into electrical engineering because I was interested in neural prosthetics. [Laugh] My childhood dream was to build the Six Million Dollar Man, bionics.

ZIERLER: You had that television show in South Korea?

HWANG: Exactly, growing up watching that American show, I thought it was all about engineering. You see prosthetic arms with cables, electronics, and circuit boards. That's how I got into electrical engineering, to build bionics. And then, I learned that we don't know much about the brain to drive these devices. That got me into neuroscience, and then I became more and more interested in the basic science aspects of the brain, moving a bit away from the implementation of neural prosthetics. But I'd like to go back to that domain when I feel I know enough about the brain.

So it was actually the other way around, I was interested in neuroscience without knowing the name of the discipline. [Laugh] That's why I went into electrical engineering. But being educated in electrical engineering was very helpful in terms of seeing things at a system level. I studied a lot of circuit diagrams to identify how the current/information moves along circuits, which was very useful to analyze the brain circuit from the computational point of view. Also, my background in electrical engineering, knowing the capacity, resistance, current, things like that was helpful for understanding electrophysiology.

And I was trained in instrumentation, so I was able to build a lab or a rig to run experiments. For instance, electrical amplifier and connecting that to the other equipment and all the analysis, it was absolutely useful to know electrical engineering. That was good. But there's a big difference, of course, between electrical circuits and brain circuits. For instance, in brain circuits, you can silence specific types of cells across the brain or some brain region. Yet, aniamls still behave. For an electrical circuit, you do something like that, nothing would work. You blow all the capacitors. The circuit is just not going to work. [Laugh] I can't really make a parallel between electrical circuits and brain circuits in terms of component to component. There are big differences there.

ZIERLER: And in recognizing the importance to study the biological aspects of engineering, was that what naturally got you interested in biomedical engineering?

HWANG: That's right. I wanted to do what I did in Richard's lab immediately, but I didn't have an opportunity. Those who were doing those kinds of studies had moved out of the school that I went to, Hopkins. I was looking for the next closest thing to do. Then, I joined the lab that built computational models of motor control that looks like a brain. We were building models to understand, "What kind of computational units should be in the brain to explain this human motor behavior?" That's what I studied at Hopkins as a PhD student. But I always wanted to work with actual physiological data, not just building a model that might be doing this and that. And I also felt I could always build a model that does what I wanted, but there was no ground truth to validate my model. I wanted to get my hands on physiological data, and that's why I decided to go to Richard's lab.

ZIERLER: In pursuing biomedical engineering, did you specifically want to come to the United States? Could you have stayed in South Korea?

HWANG: I don't think biomedical engineering excisted back then in Korea. That's another reason why I went into electrical engineering rather than biomedical engineering. After electrical engineering, I went into biomedical engineering in the US. From there, I became more and more interested in understanding the brain itself.

ZIERLER: Did you have a good experience at Johns Hopkins? Did you enjoy it there?

HWANG: Yeah, I made a lot of friends. I fell in love with science at Hopkins. My mentor, Reza Shadmehr held a weekly lab meeting. Instead of telling someone to present something, he would be presenting his own research. He would still have his own lab notebook, do experiments, then he'll pitch his idea to have discussion with us. I sit there in the discussion, watching postdocs, Reza, and graduate students arguing, often really loud. I loved science, this collective, the collective work. We don't necessarily get to a conclusion at the end, but sometimes we do understand better together through these discussions and exchanges of ideas such as what one learned from the literature that the others didn't know. Reza used to tease me that when I first joined the lab, I was very shy, I wouldn't open my mouth for many months, but all of a sudden, he couldn't stop me from talking. [Laugh] I had a big transition at Hopkins when I was working with Reza.

Motor Control at Johns Hopkins

ZIERLER: Tell me about Reza's work. What is Reza known for?

HWANG: He's known for theoretical motor control. He was also trained in engineering. Basically, he's one of the people who inferred what kinds of computational processes might be happening in the brain to explain how people learn a new motor skill or adapt to a new environment. He's the one who basically found the internal model concept. The brain builds a new model about the environment and what should be the proper motor command to generate in a new environment to have a desired motor output. He developed an experimental paradigm to prove that. Basically, his big concept is to perturb movements and see how people adapt. The way they adapt will reveal what's happening in the brain. The limitation of learning and generalization patterns will reveal what information is encoded and used in which way. That's what I did with him, designing specific perturbations and seeing how people adapt or don't adapt to a specific perturbation and how they generalize. But he's actually gotten into physiology now. [Laugh] He's doing neurophysiology like Richard in marmosets these days. He wants to get physiological data, too, now. [Laugh]

ZIERLER: What were the main conclusions of your thesis at Hopkins?

HWANG: I found that from the patterns of people's motor learning and generalization, we may have computational units in the brain that try to map what the intended movement should be, then what the motor command should be. What the intended movement should be is represented in terms of what the velocity should be, what the position should be, what the acceleration should be. What we found is that the most prominent representation in the brain seems to be velocity, then position. The representation of the position seems to be more gradually changing than velocity, based on the way people generalize their learning.

That's where I intersect with Richard. What I proposed for my thesis is that maybe the representation of desired movements at the neuronal level is more like a gain field. Meaning that neurons are mostly representing desired velocity, but they are modulated by the desired position linearly or monotonically. That's what Richard proposed and discovered in the posterior parietal cortex in terms of eye movement direction and eye position. That's where our intersection happens, gain field coding of information at the neuronal level.

ZIERLER: Did you know of Richard Andersen's work while you were at Hopkins, and did you specifically want to get fully involved in neuroscience for your postdoc?

HWANG: I became familiar with Richard's work on gain field because we were proposing gain field coding in different contexts. Not eye movement, but in arm movement. I was familiar with his gain field work, and was following up on it. I think it had been just a few years since Richard's lab started the neural prosthetic work, and I was following it back then. I wanted to see whether there was room for me to contribute. They had just published a really big paper in Science on neural prosthetics, so I knew the topic was growing fast in his lab. I was familiar with his work through the papers, but I never had a chance to see his talk or meet him in person.

The Joy of Pure Science in the Andersen Lab

ZIERLER: Tell me about arriving at Richard's lab. What was that like for you?

HWANG: Can I tell you about the interview first?

ZIERLER: Of course, please.

HWANG: Maybe it was the second day of my visit to Richard's lab. After my presentation, the next day, I went back. Richard said, "When do you think you can start?" I think that was one of the happiest days of my life. I really wanted to join his lab, and I thought he'd say, "I'll get in touch with you in a month or so." But he didn't, he said, "When do you think you can start?" Right after that meeting, I just ran out of the building. I called my best friend and told her, "This is my happiest day. I got into Richard Andersen's lab." [Laugh] I was very excited to join. And when I joined, I don't remember the exact number, but there were maybe 10 or so postdocs already and a few graduate students. It was a really big group. And everybody was so mature, and they all had their own ideas and very independent projects. I was initially paired with a senior postdoc so that I could get started, and that was really good. He gave me a lot of ideas. But I was so motivated to have my own project. And soon, it happened. I think the years I spent in Richard's lab were the time I could appreciate the privilege of being in science without the burden of having responsibility. [Laugh]

ZIERLER: Pure science.

HWANG: I could just do science. I didn't have to worry about, "I have to write a grant. I have to support my people," things like that. Richard gave me full resources. I just had to have a good idea to test. I was able to do dream science. I could have my independent ideas, I could test the ideas, almost without limitation. I never had that experience ever again. [Laugh] I always had to take care of other things. It was a dream, I guess, sort of most postdocs could hope for. You can do science without worrying about anything else. Then, I also met colleagues who are still in touch with me. I still talk to one of my colleagues I met at Richard's lab, weekly. [Laugh]

ZIERLER: What was Richard focused on at that point?

HWANG: He was leading multiple projects. Imagine having 10 or so postdocs. But I think he was becoming more and more serious about neural prosthetics. We started with monkeys, but by the end of my period there, he was already thinking about doing human subject testing. I could see his focus was moving towards human applications more and more. But that doesn't mean he spent his whole time on that. He was still supporting basic sciences, like coordinate transformation and visual feature representation in the parietal cortex. But I would say about half of the folks in Richard's lab, by that time, were interested in doing some kind of neural prosthetics. Some of them were eager to work for human applications. And Richard was very serious about that.

The Signals that Drive Neural Prosthetics

ZIERLER: What about you? Were you working on neural prosthetics in Richard's lab?

HWANG: Yes, I was. I was doing both basic science and neural prosthetics. My focus was, in terms of neural prosthetics, evaluating alternative neural signals to use to drive commands for neural prosthetics. Before, we were mostly using single neuron spiking activities. We were trying to extract information from the spiking activities. But not all electrodes yield spiking activities. But all electrodes have local field potentials. Even if there's no spiking activity, there are background neural signals. When we filter specific frequency bands, we can extract some information about motor intentions. I was working on the local field potential information extraction.

ZIERLER: Were you interested in working with human subjects? Was that part of your work in Richard's lab?

HWANG: No, I was not involved in human subjects. I became more and more interested in basic science. I was a little hesitant back then. I felt, "I am not ready to go to human subjects yet." I was more on the cautious side. [Laugh] I wanted to learn more about the brain before I moved onto humans. I wanted to gain more understanding about the primate brains back then. But the lab was already designing and going through, I believe, some FDA approval process.

ZIERLER: What were the most important things you learned about the human brain in Richard's lab?

HWANG: Well, there were people who did their postdoc training in Richard's lab and transitioned to working on human brains. They were mostly trying to identify what is common between primate brains and humans. They compared what we found in the spiking activity and local field potentials from monkeys, to fMRI data from humans. They were able to find some common properties between the human and monkey posterior parietal cortex such as effector specificity and coordinate transformation. They found differences as well. For instance, the anatomical location of the reach region differ between humans and monkeys. Nevertheless, because we ultimately want to understand the human brain, we focus on commonalities rather than discrepancies. By identifying shared features across species, we aim to apply our findings to human brains effectively, which ultimately drives our research efforts.

ZIERLER: Tell me about your decision to move then to Takaki Komiyama's lab at UC San Diego.

HWANG: That all started from Caltech, too. I went to a seminar. We had a weekly seminar in the basement of BBB, inviting prominent scientists to present their recent data, and Takaki was one of them. He was showing two-photon imaging of neurons. That's what I was dreaming about back then.

ZIERLER: Why was that so important for you?

HWANG: When I was using electrodes to study neuronal circuits, I always wanted to record a lot of neurons at the same time. Individual neurons are noisy, so we can extract only limited information from each neuron. But when we record multiple neurons simultaneously, we can overcome the noise of individual neurons and decode information that the neuronal population represents collectively. When we're studying neurons in Richard's lab, we're only listening to neural activity. We listen to the spiking activity, tick-tick-tick-tick-tick, in multi channels. I always wanted to know what neurons in action look like. Finally, Takaki was showing neurons in action, not just one or two, hundreds of neurons that I could see.

Neuronal Activity in Real Life

ZIERLER: Was this a recent technological breakthrough, to be able to see neurons in action, as you say?

HWANG: That's right. Back then, maybe 15 years ago, it was just starting to be available. Takaki was one of the first people from Svoboda's lab at Janelia Farm that made this in vivo imaging of neuronal activity feasible. It was the very, very beginning, and I was just blown away, and I really wanted to learn the technique. I learned that from Takaki's lab, and now I'm using it in my own lab.

ZIERLER: What is so important about this technique? What can you do with it that simply wasn't possible before?

HWANG: We can track or follow the same neurons over the course of weeks and months. For instance, for two months, every day, I was tracking the same neurons in the motor cortex when mice were learning a new task. From this longitudinal data, I was able to find some signatures of disengagement of the motor cortical neurons when mice were performing the learned motor skill. This inference was corroborated by optogenetics. In fact, the need for the motor cortex went away when mice became very good at performing a new task. I don't think I was able to follow the same neurons day-to-day for two months with the traditional electrode recording technique, but using the two-photon imaing with high yield, I was able to follow 100 or so neurons for two months. It's really, really fascinating.

Another advantage is studying projection neurons. I was able to study not just all PPC neurons, but specific PPC neurons that project to the striatum or to the secondary motor cortex. Using this two-photon technique, we are able to visually identify which are projection neurons and how they're different from the other population of PPC neurons. That was something that I couldn't do easily with the traditional method. I may have been able to, but the yield would've been so low, it could've taken 10 or 20 years to collect the kind of dataset I had. [Laugh]

ZIERLER: You mentioned you learned about the Komiyama lab from Caltech. Were Richard and Takaki collaborators? Did they rely on each other's work?

HWANG: No, not really. Takaki was just starting his lab at UCSD. I'm almost sure Richard knew Svoboda who trained Takaki, but I don't think he would've been aware of Takaki at that time. I don't think it was on his radar, the new lab. [Laugh]

ZIERLER: So the plan for you in San Diego was to learn this new technique, and that would be how you'd start your own career as a professor.

HWANG: That's right. Initially, I was thinking, "Maybe if I learn the two-photon imaging technique very well, I can move it to primates." It turns out, there are significant technical challenges to move it to primates. There are many differences between rodent brains and, say, macaques. The primate dura is thick and opaque. Mouse dura is transparent and thin, making the brain optically accessible. The inflammatory responses are very different. Primates are very vulnerable to these kinds of surgical procedures and external devices implaned for imaging. In addition, the genetics of primates are not well-studied, and are more complex. All the vital technologies that allow this optical interrogation with two-photon imagine are not very easily translatable to macaques. Now, here and there, a little bit of breakthrough is happening in primate two-photon imaging, but it's taking much longer. I feel that I may not be the right person to spend time in refining the technology. [Laugh] I decided to focus on mouse models with this technique and ask the questions that I can ask.

ZIERLER: A more personal question. After all of this education, your graduate research, two postdocs, did you want to stay in the United States? Did you ever think about returning to Korea?

HWANG: When I first came, that was more of a preference. I wanted to go back to Korea and share what I learned in a more developed country with fellow students. But I happened to marry an American. [Laugh]

ZIERLER: That'll keep you here.

HWANG: That's right. It was realistically difficult to move back. And then, I was pretty happy here as well, so the decision was easy.

ZIERLER: When you were thinking about faculty positions, did you specifically want an affiliation with a medical school? Was that important to you?

HWANG: No, it was not. It just happened to be that there was a medical school. But I wouldn't have rejected being in say, a neurobiology department to do the kind of work that I do. I wasn't really selective about applying for medical school.

Becoming a PI in the Covid Era

ZIERLER: What year did you join Rosalind Franklin University?

HWANG: August 2000, right in the middle of the pandemic.

ZIERLER: That must've been extremely difficult, setting up a lab with all of the restrictions.

HWANG: Yes and no. Because it was COVID, nobody was asking me to do anything but set up my lab. [Laugh] I was focused on just building the lab. And surprisingly, the unversity was still operating, with people working remotely. For instance, I joined on the 17th of August as a faculty member, and I was able to place an order for half a million dollars of equipment on the 29th of August. Things were moving. [Laugh] I was pretty happy, although I wished there were more people around sometimes. But it also gave me time to focus on just building the lab.

ZIERLER: What was the most important instrumentation for you to get right at the beginning?

HWANG: Two things. One is, as I said, a two-photon setup to study the neurons in the circuit that I'm interested in, and two, a behavioral apparatus. I don't study intrinsic/innate behaviors, I study learned behaviors. I had to set up a system that trains mice to learn a new decision-making task. I built that by the end of December of 2020 and got it to run. We were collecting data from those learned decision-making behaviors.

ZIERLER: Do you see yourself mostly as a user of the techniques you learned in the Komiyama lab, or are you also trying to improve those techniques?

HWANG: I would say I'm more of a user. I focuse on answering scientific questions using those tools, rather than developing them. I would be more of a user.

ZIERLER: What is the most important funding for your work when you started up your lab? Are these mostly NIH grants?

HWANG: When I first started, I didn't have NIH funding. Usually, faculty members get startup funding from the university. I had a startup fund, and soon after that, I was able to get some foundational grants and fellowship by proposing studies here and there. Then, I recently got an NIH grant, so I can run larger scale projects now.

ZIERLER: Does the lab, as we continue to work our way out of COVID, feel built up, or are you still building it?

HWANG: I think finally, I'm in the right shape to suppor individual lab members to pursue their independent projects. I have two full-time research assistants, and one research assistant who's been working with me in the last three years and wants to do a PhD in my lab. And I just hired a full-time postdoc who's joining the lab in two weeks.

ZIERLER: Very exciting.

HWANG: It is very exciting. We can do the research I proposed in an NIH grant and other foundational grants, so I'm very excited to see the end results of all these projects. People are very motivated to work hard. [Laugh]

ZIERLER: We'll bring the story right up to today. What are you working on currently?

HWANG: Currently, we are refining the findings from our aging study. I told you about the stimulation effect. Now we want to see what changes with age in PPC activity that drives the reduction of exploration that might be linked to reduced learning ability. We're imaging the brain activity in old versus young mice in the PPC. That's one.

The other one is, we tried to delineate this neural arc from the cingulate cortex, to the posterior parietal cortex, to the striatum, in terms of what kind of information leads to idiosyncratic experience-dependent decision biases. We're studying each neural path in the neural arc right now, the CC to the PPC, PPC to the striatum, both functionally and anatomically.

Those are two big things that I think I know how to do, and then we just started another one, a Parkinson's disease project. I've been studying motor control, and one of the most prominent impairments in Parkinson's disease is motor function. We wanted to find the neural basis of movement impairment in Parkinson's disease outside the striatum, which has been the hot focus in the community.

ZIERLER: Now that we've worked right up to the present, for the last part of our talk, I'd like to ask a few retrospective questions, and then we'll end looking to the future. Your educational trajectory, from electrical engineering, to biomedical engineering, to neuroscience, and now doing work that does touch on and ultimately will help, hopefully, into translational research, how do you understand that progression? What is the value of you getting those degrees in the sequence that you got them?

HWANG: It happened that way. I didn't really plan it that way. [Laugh] But at every step, there were very important skill sets and knowledge that helped me in the next step. We talked about electrical engineering. There, I think mostly the skill set I acquired during that training was critical for my experimentation and analysis. The computational work with Reza made me think at the system level of the brain rather than just the cellular or biological level. I was able to build a behavioral model. "What kind of latent variables would be important to explain this complex behavior when we can observe only the external stimulus? What's happening inside of the brain?" I can build computational models that could explain what we cannot explain when using only observable variables, and I can link the modeling results to physiology.

And then, you asked me what's good about having a medical school. Perhaps I can find someone who will be interested in my basic science findings in translational research with human subjects, without having to look for collaborators outside. I think these days, we have a lot of technologies that allow us to have remote collaboration. Nevertheless, I think it's also important to be physically close. I'm hoping that's going to happen, but it's still on my to-do list. [Laugh]

ZIERLER: Just a fun question, your childhood interest in the Six Million Dollar Man and making people bionic. Does that seem closer or farther away not that you know all that you know?

HWANG: On a personal level, I think I moved away a little bit in terms of implementation, but I think we're getting closer with Richard's work and others' work. We are beyond the proof-of-concept level, for sure. Based on what I'm finding from all this longitudinal imaging, perhaps one limitation might be that the brain circuit keeps changing. If we relies on ever-changing brain areas for decoding motor commands to control bionic devices, we're aiming at a moving target. I didn't appreciate that before I studied the brain. The brain always changes. It's plastic. The circuit that controls one behavior could move out of one region to another. I think understanding these fluidity of the brain will definitely be useful for building robust bionic devices or neural prosthetics. Maybe I can find a neural substrate that might be more applicable or better than the conventional regions we've been targeting. [Laugh] I feel that conceptually, I'm getting closer. At some point, I hope I will feel confident enough to move what I learned from animal models to human subjects or human patients.

The Long Game of Posterior Parietal Cortex Research

ZIERLER: I'm always curious how Caltech alumni feel about their time at Caltech. For you, what has been the impact, intellectually and scientifically, from your time in Richard's lab? How has that stayed with you?

HWANG: One of the big impacts is that I am stuck with the posterior parietal cortex. If I didn't go to Richard's lab, I would have not studied the posterior parietal cortex. [Laugh] I would not have discovered the projection-specific information routing and so on. Luckily, the posterior parietal cortex is a region that a lot of things happen, so I don't have to compete with Richard. That training, learning about the posterior parietal cortex in terms of motor control and even decision-making in Richard's lab was very pivotal. Knowingly or unknowingly, that became the foundation of my scientific questions these days, I believe.

And as I said, I am still in touch with some of my colleagues that I met at Caltech in Richard's lab, and we're still having scientific communication, talking about collaboration. I have a very pleasant and fun memory of Caltech. I still remember the time we were having lunch at the cafeteria, and I ran into Stephen Hawking going by. [Laugh] It was just a very fun and inspiring time to do science.

ZIERLER: Finally, looking to the future, now that you have your lab built up, now that you have students joining, if you think about things in terms of a one-year plan, a five-year plan, a ten-year plan, what are the things you're working on now that will lead you to the questions that you haven't yet worked on? What does that look like for you?

HWANG: I have hypotheses about many things. [Laugh] I may find very unexpected results that may make me think differently. But that's unforeseen by definition, so I can't tell what that will be. Or it might pan out in a way that validates my hypothesis. One thing I truly don't know about is this Parkinson's project, a new line of research in my lab. The motivation for this specific study is because it has not been studied, meaning that we don't really know what's going to happen. I'm excited to see what comes out, however it comes out.

We're studying particular neurons in the motor cortex that project to the subthalamic nucleus, which is the main target of DBS in Parkinson's disease. We want to know how these neurons change before and after Parkinson's disease. There might be a way to intervene with these neurons, so we can help alleviate the motor symptoms. That's a truly exciting unknown area, and I think in a year, we'll have some idea. In five years, I hope we'll have some kind of suggestion as to how we might manipulate these neurons for translational purposes.

ZIERLER: Good luck!

HWANG: Thank you.

ZIERLER: It's been a great pleasure spending this time with you. I want to thank you so much for doing this.

HWANG: Thank you, David.

[END]