Conversation #1 with Diego Bohórquez and Mijail D. Serruya: how information moves levels in bodies
Neuroscientists Diego Bohórquez and Mijail Serruya discuss how information transitions from molecular to behavioral levels in biological systems and how technological interfaces could enable new biomedical applications.
Watch Episode Here
Listen to Episode Here
Show Notes
This is a ~59 minute working meeting where I introduce two very interesting neuroscientists to each other - Diego Bohórquez (https://medicine.duke.edu/profile/diego-v-bohorquez) and Mijail D. Serruya (https://www.jefferson.edu/academics/colleges-schools-institutes/skmc/departments/neurology/faculty/serruya.html) and we discuss ideas around how information moves from the molecular to the behavioral level in bodies and how we might be able to interface with that system via technologies that could lead to biomedical applications and deep insight into diverse intelligence. At the end we discuss one such application briefly - McConnell's professor-burgers.
CHAPTERS:
(00:00) Gut brain biology foundations
(13:28) Neural interfaces and organoids
(26:18) Information across biological levels
(32:41) Environmental sensing and placebo
(47:32) Translating internal biological dialogue
(54:23) Collaborative next steps
PRODUCED BY:
SOCIAL LINKS:
Podcast Website: https://thoughtforms-life.aipodcast.ing
YouTube: https://www.youtube.com/channel/UC3pVafx6EZqXVI2V_Efu2uw
Apple Podcasts: https://podcasts.apple.com/us/podcast/thoughtforms-life/id1805908099
Spotify: https://open.spotify.com/show/7JCmtoeH53neYyZeOZ6ym5
Twitter: https://x.com/drmichaellevin
Blog: https://thoughtforms.life
The Levin Lab: https://drmichaellevin.org
Transcript
This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.
Main Episode
[00:00] Diego Bohórquez: So, Misha, I grew up in Ecuador, in the Amazon of Ecuador, the Amazonian side of Ecuador. I came here around 2005, twenty-one years ago. I became interested in gut-brain biology because when I was in grad school—I got my PhD in nutrition—I heard this woman in 2006 saying that after bariatric surgery, not only she lost 40% of body weight and her diabetes was gone, but she said that previously, she had a repulsion for the yolk of the eggs, but after the surgery, she had a craving for the yolk. And I remember thinking that everybody should be studying this and it should be known. But I think the first clinical report of that phenomenon was in 2011. And now, obviously, it's pretty well established. So I started to study, first, the cells that secreted all of these hormones, including the GLP-1 that is so famous now. But when I entered the picture, since 1875, these cells that are endocrine cells or gut endocrine cells have been studied in an indirect manner because there were no mice or no models to extract the actual cells. So these models became available in 2008, the first mouse. In fact, it was to isolate the GLP-1-secreting cells that people know as L cells. And when they took them out and started to do sequencing, they found that those cells had taste receptors, cytokine receptors, nutrient receptors, and a bunch of things. The other thing that they found is that up until that point, it was thought that every cell only produced one type of hormone. So that's why we have this terminology of L cells, for instance, the GLP-1 ones are known as L cells. Then there are other ones like K, something like that. But it was only this outdated idea that there was one cell to produce one hormone. So when I began studying this system, because now it was possible to isolate the cells, very quickly, I realized that each one of these cells could produce or could at least express the genes of 12 different neuropeptides. They had synaptic proteins. They have axonal proteins. They have axon guidance proteins, and they had a lot of other things and nerve growth factor proteins. The big question in the field was that these cells only communicated with the nervous system through indirect means, through the secretion of hormones. And this is very important because you all have heard the term gut-brain axis. And the reason that they use axis is because axis is an imaginary line. But we never talk about the nose-brain axis or the tongue-brain axis or the ear-brain axis. We just talk about sensory systems. So then the question was, could these cells not be only releasing hormones, but actually be connected to neurons? Some people had tried in the seventies, but because the technology was not there, it was almost impossible to isolate the cells. So now that I was able to isolate the cells, I ran this experiment. Let me show you. There is one slide that I could show you. I'm going to show you this little video. Here are the cells isolated. This is one of these neuropod cells.
[04:29] Diego Bohórquez: We call them neuropods because they are not endocrine. They are sensory transducers. Because we were able to isolate them from fluorescence, we put them in culture with a neuron from the brain. On the left side, you have a neuropod, and on the right-hand side, you have a sensory neuron from the brain, and the time is in hours. This was an experiment in 2012. It was crazy at the time because not only do these cells know how to recognize each other in a plate, but they actually know how to find each other and reconstitute circuitry. If you play it frame by frame, there's actually movement across the physical connection. There are so many things in there that are happening. Of course, we were able to run a lot of other experiments like rabies tracing to show where the cells were connecting to, and so on and so forth. We showed that these cells were physically connected to the vagus nerve just like taste receptor cells or olfactory receptor neurons. Eventually, once we had the physical connection, the question was, what is the purpose of this sensory system? We realized that these cells were not only detecting the nutrients that we eat or the chemicals in the food, but they were very rapidly discerning. Then they are secreting signals in the form of neurotransmitters that stimulate the nervous system, and by that, they guide decision-making—meaning that that's how a mouse knows whether or not it's sugar or sweetener. Now we have evidence that these cells detect not only proteins in the diet, but the specific components of the protein, to specific amino acids. If you deprive that animal of a specific amino acid, it actually alters the sensory function of the cells, so they favor that specific amino acid even in the presence of something that is very rewarding like sugar. One of the pushbacks in this field, which sounds silly now, was that when we eat, we don't think or perceive; we're not cognitively aware of what we are eating. But we all know that we're not cognitively aware of what is going on in the back. We don't have to be cognitively aware. A lot of this is just running in the subconscious. And the pushback was because eating or digesting is so slow. Take a look at this video. This was done by a collaborator, Teresa Lever. This is a fluoroscopy video, and the mouse is fasted. The mouse is going to eat this food on the right that is labeled, and the time is in seconds on the bottom right. Here it goes, the mouse starts to eat.
[08:58] Diego Bohórquez: One, two, three. The food is already in the stomach. And then in less than ten seconds, boom, it's already in the intestine. So well before the animal has got any chance to give rise to thinking whether it's full, the gut has to make a raster scanning of what is in the food. Now we're realizing that it is giving rise to many of our behaviors, or it is influencing our behaviors. Last year, we published a paper in which the same cells in a specific part of the colon mediate a communication between the biota by detecting microbial patterns in the microbiota, and then giving a whole neural cascade or influencing meal patterns, like how much we eat and for how long we eat. A lot of this is in mice, but we have done some translation into humans using human intestinal organoids. At least for the sugar-sensing mechanisms, it looks to be very conserved, because the sensory system in the gut is the most ancient. Organisms were formed around food, and the sensory system seems to become a Velcro or a recognition system for the food that is available. That's on the biology side. Back to the Amazon. I grew up being exposed to native people, and my parents worked with a lot of native people. When I was five years old, I heard this story from one of the natives. That story was retold over the years by my mom, so it's more like a little bit of a remembrance, hearing that when his grandmother would cross a creek, she would become an anaconda. And I was like, how the heck can you become an anaconda? I also want to cross that creek. Fast-forward to when I was 35, this was 2018. I began going quite a bit back home. My father passed away in November 2018. But a week before he passed away, I went to visit a friend on a farm, and he lived with a native Quichuan. I remember that we started to chat, so I asked this question: Is it true that you guys can become anacondas when you cross a creek? And he said to me, "Yes, of course. If you drink ayahuasca, you can also become like a jaguar." Six months later, through that same friend, Mike—he's good friends with a lot of payés, the actual native payés, because there are very few of them now. That's a dying art because it requires at least one year of initial training, and then five years of training: it's dieting, it's celibacy, it's social isolation. And so I had ayahuasca, and I had it since then, 2019. It has really transformed my way of feeling, my interoceptive feeling, down to the level of being aware of a glass of water in the mornings. And then cognitively, a lot of other things, like following thoughts and so on and so forth. When I shared some of these ideas with Mike, Mike said, "We've got to bring Misha to the table." So I'll pass it on to you. I don't know if it was for "let's rescue Diego."
[13:28] Mijail D. Serruya: Thank you. Thanks for that introduction. That's pretty amazing. It reminds me I was talking to György Buzsáki, who's a neuroscientist at NYU, who has these ideas that the frame rate of consciousness, at least of mammals, of rats and humans, is tied to two things: eye movements and then the gut. That the gut, peristaltic—there's nested oscillators all the way down to move food through. And he was convinced that that somehow was setting the clock of the brain for conscious perception. I don't know if that's a falsifiable hypothesis, but it's interesting. And your video of the rat showed that happening very fast. I should talk about me, but I'm trying to process everything you just said. There's obviously tons of work on vagus nerve stimulation, which does all kinds of things to the gut. It was just approved earlier this year for rheumatoid arthritis as an anti-inflammatory system, but there's a lot of work on ulcerative colitis. So this idea that stimulating electrically the vagus nerve is going to have this profound autonomic and immune change. Like most things in medicine, it's empirical. I don't think people really understand the mechanism. They're blundering around. But the point is that we have people that have these things implanted and they're perturbable, if you knew what to measure. My background is my parents are from Argentina, but I was born in Philly. That's why I speak like an American, I guess. Born in Philadelphia. And I did my doctoral work on brain-computer interfaces: electrode arrays in the primary motor cortex of macaques, and then created a company. We did the first human twenty-plus years ago, so a little too early for the current wave of brain-computer interfaces—decoding ensemble activity of brain cells to do something with them, rather than looking at one cell at a time, to look at an ensemble and say, "What does this population represent in terms of some sort of intended movement?" and then making it behaviorally useful to the macaque and then to a person who's paralyzed. I got clinical training as a neurologist, so I take care of patients who have all kinds of conditions: stroke, spinal cord injury, cerebral palsy, lots of chronic pain, chronic fatigue syndrome. Actually, in terms of the gut stuff, in the last few years, for whatever reason, I've seen a lot of people with autonomic problems that are manifest both cardiac, meaning they have orthostatic intolerance, but also a lot of GI dysmotility. Their food is not moving through them properly, and they have lots of bloating and pain and other things like dyssynergia—the sphincters aren't working right, they're incoordinated. Things that, frankly, I had not really been aware of in my medical training, but clinically is a thing. In fact, there's a whole field of neurogastroenterologists that I didn't even know existed. So I have a little bit of clinical exposure to that. The lab at the moment works on three things. One is wearable devices to help people who have paralysis or chronic pain: little braces that will move, or electrical stimulation, including on the neck. We can do vagus modulation that way indirectly; it's not specific. We also work with implants. We have people that have wires in their brain or in their spinal cord, and trying to explore what we can do with that, what we can learn to record activity and stimulate, usually with a mind towards restoring movement, but also other things, including cognitive problems with language issues. And then we have in vitro stuff, which maybe is where Mike will have to tell us the big reveal of where in his mind there's the Venn diagram overlap.
[17:14] Mijail D. Serruya: We're also working on neural specimens in the dish. We're growing them on electrode arrays—two-dimensional, three-dimensional electrode arrays—and we're working with both rat cells and human cells. They're neural tissue, so they're neurons. Usually, they're cortical neurons, but sometimes we do hippocampal neurons. We either have them dissociated, so they're just randomly growing however, or we aggregate them so they get little spheres, or we take organoids. The organoids we've been using are derived from humans. Jefferson, where I am at, has a big ALS research center. And so, obviously, it's not practical to biopsy the spinal cord of a person with ALS. So instead, they create autologous organoids. And when they have a bunch of them, they're like, "Hey, Misha, you want some?" And I'm like, "Yeah, of course. I'll take the organoid." So these are organoids both wild-type for human healthy adults as well as those who have neuromuscular disorders. We also get slices. We will cut up the brain either of a rat or, if we have a human being who is having an epilepsy surgery and we get the brain, we can cut that up. And then we put these tissues on the electrode arrays. Typically, when people are doing that, they're either trying to learn the physiology of that system as it normally functions or they are modeling a disease. Where we think a little different, we're actually asking a computational question, which is to say we know that a bunch of neurons outside of a brain is very abnormal. This is not how it normally behaves. It's not getting inputs. It's not getting outputs. The 3D structure is different. There's no CSF. There's no blood. We know it's an abnormal, artificial system. But even so, they still give rise to lots of complicated electrical activity. As you showed in your video, they're doing all kinds of crazy things. They look like little creatures. They're running around. They're interacting. Our cultures aren't quite as mobile. They're definitely moving. They're growing processes all over the place. And I'm sure at the level of the axons, they're moving, but the cell body gets stuck in place. What we're interested in is applying different spatiotemporal patterns and seeing how the system can distinguish one from another. What we're working on literally at this moment is an abstraction language to say that most people, when they stimulate, use these giant square-wave biphasic pulses that come from digital circuit design. They're not biological signals.
[21:00] Mijail D. Serruya: It's like hitting it with a hammer. We're trying to be a little more precise and exploring a wide range of stimulation patterns to see what will give us the biggest dynamic range of whatever we have growing in the culture, because we don't know a priori. Most of the research on stimulation in vitro and in vivo tends to use this cannonball of stimulation, which is fine, but drives the system into one state or another. When if you just listen without stimulating, we know the neural tissue goes through an enormous range of states. And so we want to do something that's a little more subtle and work with it, so to speak, rather than pushing it to be in a more simplified system. Why would we do this? One rationale is for my patients who have terrible stroke, and they've damaged too much tissue. If you put a brain-computer interface in and you wanted to decode their language, that's fine if they have the language areas of the brain. But what if that's gone? What are you going to record from? The idea here is to say, let's get a self-organizing system that we can link to the brain and use that. We have two hemispheres. Could you give someone a third hemisphere? How would you build that? What would it look like? How many states would that have to go through? Is it something you would implant in their bodies? Is it something that would be external? We can work on those things. But that's the ultimate goal: to get self-organizing systems that can be linked to a person to help them. There are groups that have tried to do other stuff. There are companies, in fact, that want to use biological tissue for energy-efficient computing with the idea that it's very energy inefficient to, every time you do a ChatGPT query, burn half the Amazon down with the energy expenses or something horrible. I think that premise is a little bit flawed. It's true that a human brain operates on a fraction of the energy that it takes to run a ChatGPT query, but that's because the brain is not floating by itself. It's in our body, and the body is an organized system. So the energy efficiencies aren't from just the neural tissue. They're from the whole integrated system. The idea that you can pull a brain out of a body, throw it in a dish, and maintain the energy efficiency ignores all the cost of culturing the tissue and all the other things. All that said, is it possible that we could find certain computing principles that could complement AI pipelines? Maybe. Obviously, as a neuroscientist, I know that you appreciate that each neuron itself is like a supercomputer. They're doing incredible things. The question is, if you're looking at it from a Wright Brothers practical engineering point of view, what can I do with it? What's it good for? How can I help this patient? Then you need some language to mediate what I want to achieve as a clinician versus what this cell likes to do. How do I talk to it in a way that it wants to be listened to, wants to be talked to, and that it will do something for me other than what it just wants to do for itself? That's my brief overview. I'm sure I could pull up some pictures. I'll show you one thing just to give you a flavor. Let's see if I can pull this off. Let's see. Share.
[24:50] Diego Bohórquez: So this is what I shared with Mike before. You can see this. Slideshow, play, current slide.
[25:02] Mijail D. Serruya: This is a guy who's paralyzed and he's using a brace we built. This is a little girl we have made a 3D-printed thing. This is a guy who we put an implant in his motor cortex over a stroke and then use it to move his hand. This is a grid from Precision Neuroscience we're working with now. Also, we have some patients who have implanted with these that we're trying to decode what they're thinking and seeing. With my colleagues at Penn, we've worked on living electrodes. These are microcolumns made of agar that are filled with extracellular matrix and then you seed a bunch of neurons on one end. The idea is that you transform them so they can respond to light or sound or magnetic fields or electricity, and then this whole thing can be implanted into the brain. It looks like that. And then this is these future ideas of can I make a third hemisphere somewhere and then take these organoids and do stuff with it to compute? That's a brief overview.
[26:15] Michael Levin: Yeah. Super. Well
[26:17] Diego Bohórquez: Wonderful.
[26:18] Michael Levin: I have some thoughts, but I think you guys can see why I thought you'd enjoy hearing from each other. One quick thing, and then I can talk about what I think the big theme here could be. The quick thing is, Misha, you were talking about stimulating these things in some more biological way than the square wave. Have you guys played back to them any recordings from other biologicals? Not necessarily neurons, but other datasets that are biological. Have you ever tried that?
[26:54] Mijail D. Serruya: We did a little bit. We've done a couple of pretty crazy things. We've actually played music, like Beethoven, obviously not just the sound waves, but as the electrical.
[27:06] Diego Bohórquez: Yeah. Yeah.
[27:06] Mijail D. Serruya: So we've only done a little bit. Part of the problem is what's your unit, your time measure, and what are you varying and what are you repeating? And that's what Alessandro, my engineering colleague, is trying to figure out. The short answer is yes. We've played with other kinds of signals, sine waves, and just other flavors of noise. It's a huge state space to explore. That's one of the ideas. In fact, what we tried to do is we have two separate specimens. We've tried to record, rather than recording one neuron's action potential, just what's the overall field saying? What's the chatter? Turn that into a continuously varying waveform and then talk to the other. It turns out, Mike, that in some ways, that's electrically harder to do. It would have been easier for an electrical engineer a hundred years ago just playing with analog circuitry. When everything is digital, you have to break it and then rebuild the analogs. But I'm definitely interested in that, and we've done a little bit.
[28:22] Michael Levin: This is something for us to talk about as well because we're very interested in communication between very different types of entities, both living and nonliving. And so we think a lot about converting things in a way that we could make communication interfaces for weird sorts of hybrid things to talk to each other. For example, we could certainly send you recordings from Anthrobot signals, whether they were optical or electrical or whatever. Doesn't matter because you can convert them. In general, to have different organoids talk to each other, this is the kind of stuff that we're doing. And having done that, then going back to some of the stuff that Diego was talking about, hitting some of these things with whether they be psychedelics or other kinds of neuromodulators, to see whether you could treat on one end and then send the signals somewhere else, kind of like a long-range thing. All these systems like to receive information from far away, it seems to me, and so I think we could take advantage of that. The biggest thing that I wanted us to talk about, because I think there are real opportunities here for practical collaboration and so on, is this: The theme that I see here, which I'm very interested in, is how information moves across levels—levels of organization. What do your molecular networks know? What do your cells know? What do your organs and tissues know? And what does your brain know, and what do you know? And how information is or is not available along that route, and how much of it is apparently unavailable because we don't have the right interface to get it up. Misha and I have talked about, for example, hypnosis is one way to get at information that you may otherwise not have access to. Some of the stuff that Diego mentioned, it sounds to me like you might also use that to soften the boundaries between levels, things that normally try to stay very separate, to maybe get a little more mixing going, and in general, in both directions. Like what you were saying, not only to tell the cells what you want in therapeutic applications, but also to get information out. I would like to be able to ask them questions about the things they know about, which is physiology—physiological space, metabolic space, things like that. I think I sent this paper to Diego a few weeks ago. There was an early paper in frog where they injected the very early embryo intracellularly with some kind of odorant, and then eventually, the animal that resulted a long time later would seek out that chemical in its search for food. That kind of thing is super cool because the system has to take intracellular chemical-level information and eventually turn it into an organism-wide neural behavioral program to go look for whatever this was. Almost like a tiny mass spec to figure out what this is, and then how do we go find it by a completely different level of problem, from intracellular to organism. That's just an example. I think both of you guys have really interesting conceptual pieces and model systems for looking at how information crosses networks. Together, ranging from in vitro organoid bacterial cell culture things all the way up to human patients, I think we could put something together on that view. I wanted to throw that out and just discuss what you think about that—the aspect of what do you think the lower levels know, what are the barriers to getting that information in and out, and then what are some of the technological approaches that we can take to move the needle on that stuff?
[32:41] Diego Bohórquez: Mike and Mijail. Last week, I was at a conference of poultry nutritionists. I got my PhD at North Carolina State in the Department of Poultry Science, and they invited me 17 years later to give a lecture to a group of nutritionists. One thing that I had forgotten, but it really hit me, was that we know that in mammals, the rate-limiting amino acid is lysine. But they were telling me that in chickens, it's actually methionine, that you cannot overlook the relationship to lysine, but that the main one is methionine. And it made me think, obviously, they are formulating for absorption. They are not formulating for sensing yet because that's something that we haven't made the mental jump yet. Even drug development, drug discovery is formulated from the perspective of absorption, and the kinetics have to reach the brain. But we have already some evidence, and there's a published paper last year that SSRIs are actually working strictly from the wall of the gut. They don't need to get inside of the system. Also, that paper that Mike sent me made me think—and here's another piece of the puzzle: linguistics is often an expression of what is going on in the biology that we take as colloquial, but at some point, they come to make sense. For instance, "trust your gut." We have to trust the gut because once it's inside of the pipe, there's not much that you can do. "We are what we eat." This specifically hit me because I asked the nutritionist, "Why methionine?" And he said, "Because the chickens have more feathers, and lysine is for muscle. Now if you need a chicken that is putting a lot of muscle versus a hen that is putting eggs, you do need more lysine, but it's still methionine." But methionine is because of the feathers and the beak and the nails. So I was thinking, could it be that the entire organism is built around the ability to sense the digestible environment? So the moment that it senses and detects, then it's able to organize, attract, absorb, digest all of the building blocks. The organism is built around that ability to detect and discern that environment in the way and shape of food. And then, obviously, the microbiota will benefit from it. That was one thought. And the other one, Mijail, I was at MIT a few months ago. And this idea that the gut is driving the pulsing of the brain was actually floating over there too. Some people were trying to work on fMRIs and PET scans to demonstrate that. But it makes me think that now we have access to mini-guts and mini-brains or brain organoids, I think that they call it. And I was thinking, if there will be this mini-gut, which has a lumen, and this brain organoid, that they are connected by a vagal neuron to that level, if we pulse the nutrients over here, could we alter the whole electrical connectivity? And then if you pulse in a field from here, could you alter secretion or digestion or something in this mini-gut? Thinking from a reductionist perspective, this is more for the experiment itself, Mike. But I just have those two ideas. It hit me when the nutritionist said the chickens need methionine because of the beak and feathers. Could it be that the system is really built around the ability to sense?
[37:08] Michael Levin: For the technology part, over here at the Wyss Institute, they have human on a chip and organs on a chip and all of this. At one point, we made something like a mouse without the mouse. Basically, various mouse organs in a dish, and they were able to connect to each other and so on. This sensing thing is very interesting, and beyond that, the analysis of the data. I'll give you two examples, and you can tell me if you know what's going on there. First of all, this phenomenon known as pica. This is when either pregnant women or other animals, when they need certain trace elements, they'll go and eat the weird stuff that isn't food. It was always interesting to me: how do you know it's there? So it's vanadium or it's some crazy element in the wallpaper or whatever that the cats will eat and things like that. How do you have any idea what is actually in there? I'm really interested in that, again, the ability of the system to have this information about its environment. Furthermore, there's one other step here that I think is really interesting that I haven't heard anybody talk about. The people who study placebo effects, in particular, Fabrizio Benedetti gave this talk at our group. He had this amazing talk where he said, "Words and drugs have the same mechanism of action." Because what he was saying, I think it's really profound for the kind of stuff we're interested in: how thoughts become chemistry, from voluntary motion all the way to all this digestive stuff. I think it's really profound. What he was saying is that with his test subjects, he tells them that they're getting some drug, and then he does measurements of blood chemistry and things like this. Sure enough, you've given them placebo, but the right molecular markers are coming up as if they got SSRIs and all that. I asked him afterwards, "Are your patients scientists?" And he said, "No. They're regular people who come off the street." I said, "Great. How do they know what molecular markers are supposed to come up when you take an SSRI? Who actually knows in the general population that your Wnt or your BMPs, whatever the heck is supposed to come up?" And then I forget if it was him or somebody else was pointing out that there's actually animal data on this too. There's placebo effects in animals too, which even further raises the question of how do they know what's supposed to come up? I think what we're circling around here is some of the physics and physiology of sensing, but then the deeper issue is what does the sensing actually tell them? What do they know about it that's strategic and adaptive?
[40:02] Diego Bohórquez: Mhmm.
[40:02] Michael Levin: So that they can make decisions and run the control, whether the control goes up or down or no doubt both. That's the kind of stuff that I'm very interested in is to figure out: can we figure out some assays where we can unequivocally show what the system knows about what's around it? And can we show that it knows more than we think it does and then figure out how it knows and what is it using to process? Viewing all of these systems as doing informational exploration and decision-making of their physiological and metabolic space. We can do in vitro stuff. We can connect these things, whether by neurons or by artificial interfaces. So you can imagine connecting it. If the neurons don't want to connect them, that's fine. We can make synthetic kinds of things that would send and transduce information back and forth. We could even start with some kind of instrumental conditioning. Somebody had made a thing where the plants signal when they need water and they make a sound, and they sort of be crying—made it do with a baby crying sound, which is just really brutal when it starts crying: water.
[41:28] Diego Bohórquez: Even to humans.
[41:29] Michael Levin: It's a little much, but you can imagine. Here's what I would like: I would like cells in culture to be able to say to us, "Hey, your medium doesn't have enough taurine or lysine," or your liver would say, "Hey, the potassium levels are low, and also, you don't realize, but I need more whatever rare mineral."
[41:58] Diego Bohórquez: That's fascinating, because—Mijail, you probably have heard these stories too or have worked with it—I've heard people with schizophrenia, for instance, some of them smoke really hard because nicotine dumping some receptors, but they also crave at certain times a steak because it's very rich in certain compounds. So could it be that there is this internal sensing system that is making it crave, and out of that is making it seek or making it communicate, trying to close that loop?
[42:44] Mijail D. Serruya: I guess there are two ways to think about it. One is to think what are some model systems in vitro that let us perturb and ask these questions about interlevel communication? And then the other extreme is, are there natural experiments that are already happening in an intact rat or an intact person? One of the things we give for depression is acetyl methionine. You give people supernormal doses of this stuff. There's placebo-controlled trial data that says that it has a benefit. Based on what you're saying, Diego, it probably doesn't even reach their brain directly. Maybe it's doing everything in the gut. My point is that there are examples where humans are—pica is the example Mike gave or, for that matter, any pregnant woman. With my wife, what she wants to eat changed with each baby. The fetus is communicating through—I don't know the mechanism—very precise changes in dietary preference that presumably have to do with what the fetus needs. And this is, of course, before it has a fully formed brain; there's no conscious decision-making, obviously. I think there are a lot of natural examples where this communication must be happening. The question is how do we want to query it and have some falsifiable experiment to say what's going on. This idea of having a person deficient in something and then we replenish it, or a person is not deficient, but they're overexposing themselves to something—anytime you're in an animal experiment, it gets complicated because if the blood-brain barrier has a certain number of transporters and they get saturated, then the timing matters. But there are ways to do it. The cleanest thing would be to have a series of experiments that combine them: you have an in vitro model, and then the rat, and then the human, so that they can mutually inform what you learn from the perturbation. It certainly seems like we don't have to be conscious, Mike, for this information transfer to happen. Consciousness seems like the last thing to come along.
[45:18] Michael Levin: Well, I would flip that upside down. I would say that we don't have to be conscious, but I'm, in general, very skeptical of this notion of unconscious processes because they're always calibrated against the patient reports. If you want to know whether it was unconscious, you ask the person, and if they say, "Nope, I didn't see it, didn't feel it, whatever," you say, "Good, it was unconscious." Well, that's great. You've communicated with the language user in the body. Nice. But who's to say that, actually, the other processes didn't have some other kind of first-person experience that didn't have access to language anyway? It's not an advanced version of consciousness. We don't have to resolve that ancient puzzle, but I think saying that the high-level linguistic human consciousness is not aware of it, to me, doesn't cut off the possibility that—I don't think we need to have a strong claim on this, but I'm not sure we know what other processes are or are not accompanied by conscious experience. And also, I'm not sure at all that unconscious processes are even—we think it's possible to have an unconscious process. It seems obvious, but there are plenty of situations, for example, in mathematics, of like a round square, that I could tell you that it seems perfectly plausible to trisect an angle with a compass and straightedge. You would think, sure, it's possible. And it turns out, no, actually, that was a failure of imagination. We thought it was possible, but it's really not possible at all. It's a logical impossibility. So it may well turn out that we think we can imagine unconscious processes, but, actually, there is no such thing. And I sort of lean that way. I don't know why we're so convinced that it's even possible to have unconscious processes. But it doesn't matter. We don't need to resolve that here, other than
[47:24] Diego Bohórquez: Right.
[47:24] Mijail D. Serruya: I I I guess I
[47:25] Diego Bohórquez: just
[47:25] Mijail D. Serruya: Things are being communicated across levels without it requiring a linguistic report.
[47:32] Michael Levin: For sure. But having said that, now my question is, what transduction interface can we offer that would eventually enable a linguistic report such that we, the external observers, can more easily interact with it?
[47:47] Diego Bohórquez: Yeah.
[47:48] Michael Levin: There are people who read dog behaviors and so on. So it doesn't have to be linguistic, but it would sure be nice if, especially now in the age of language models, we could slap a language-using layer on top of this whole thing. We have a system that's very good at doing things and knowing things and not great at talking about it. We now have systems that are great at talking about stuff, unclear what, if anything, they know. But I feel like we could put those together. We've already started this in our group trying to talk to gene regulatory networks and things like that, and we've had some success. You can imagine the basic understanding of what these things know is critical, and then the opportunities for life-improving technologies is, I think, massive if we hear the correct
[48:47] Mijail D. Serruya: So are you saying that whether it's in a rat or an in vitro organ, body on a chip, that we could sub some set of sensors and basically read out in English, saying, "Body or whatever structure X needs amino acid Y, and it's currently being requested"?
[49:06] Michael Levin: Please deliver. What I'm envisioning once we know what we're doing is on your cell phone instead of, "Hey, Siri," it's going to be, "Hey, liver, why do I feel like crap today?" And it should be able to tell you that, given your eating history and the fact that you've been doing XYZ, and, of course, it can talk to your fridge so it knows what you've been buying. At some point, you should be able to have a conversation with it. You won't be talking about movies or politics. You'll be talking about physiological states. We're now trying to figure out how to incorporate the things that it natively is able to do with an understanding of the literature. Ideally, you will also have a layer that's like, "Well, there's a paper from Hopkins from ten years ago that shows this and that, and therefore, I think you should do this." That's, I think, within reach of the technology now. That's what I'd like to work on.
[50:02] Diego Bohórquez: Mike, take a look at this statistic. I was exactly thinking one of the implications of this. Can you imagine that 40% of the entire food supply is wasted?
[50:24] Michael Levin: Damn. That's wild. That's completely wild.
[50:29] Diego Bohórquez: No wonder why we are destroying everything. If we are to reimagine efficiency, we have to imagine it from the organism, from the financial supply chain.
[50:46] Diego Bohórquez: You know? Yeah.
[50:48] Michael Levin: I'm generally focused on the clinical patient issues, but the other side that I think is closer to what you just showed us is the lab meat production stuff, lab-grown meat. Imagine if we could tune those processes to what you actually need. Down the line, maybe it's not even—I don't know if this is crazy, but maybe it's not even some centralized thing that grows these cell burgers. Maybe it's in your house. Maybe there's a 3D cell printer, a meat printing thing—doesn't have to be meat, of course, but some kind of a cell printing thing. It's in your house, next to your fridge, some kind of thing. All of this is a system where you don't know this yet, but you should have more of this or that, and this is how the profile is going to shift. Or conversely, maybe you do know it. And again with the waste thing, the people who are growing—David Kaplan is a close collaborator of mine; he's big into growing—they're not sure what exactly they should be growing. The nutritional profile of these cells, who knows? You grow these muscle cells or other cells from various animals. Do they still have the same metabolic content as they did in the animal? Who the hell knows? So maybe monitoring that and specifically...
[52:25] Diego Bohórquez: Distilling it to the bare minimum, I think that the three of us are in that area. It can have some very profound implications. For instance, right now, I think 30% of the tomato supply in the United States comes from Canada. I recently was on the other side of—where is it? Close to Toronto. They have hectares and hectares of greenhouses, and a lot of the tomatoes are produced in sterile conditions. All of the nutrients are just going into this pot, and then you have these tomatoes that are huge. They have the size. They just don't have the flavor, and they don't have the nutrient content. So we're producing the volume, but we're not producing the actual molecules that are making the human thrive. And I think a lot of the waste comes from that. So the Velcro is not Velcro right now. It's not sticking. The same thing is with chickens, cattle, everything, but just talking from the food supply perspective. I think it can have very profound implications because in the future, it could be just a pouch or something that people take to boost not only themselves, but actually the bacteria that can also turn on some of these mechanisms, including pleasurable mechanisms where you actually feel that closer to what a steak is.
[53:58] Michael Levin: That third hemisphere idea, we've been kicking this around too. You can make one for the brain, but you could also make one for the gut, and you could make one for other, right? So increasing the IQ and the processing power of various parts of the body is a big, big opportunity.
[54:23] Diego Bohórquez: This is fascinating. Mike, how do you see the next step?
[54:29] Michael Levin: Here's what I think we should do. I think we should go off, and each of us writes down one or more minimal models that would answer one specific thing that we're all interested in. And let's look at them together and see if they're already overlapping or if they're dovetailing or what. I think from this conversation, we have a good idea of what our interests are here. So let's do that. Let's write down some very basic things, and we'll connect them together, and then we'll see what the early experiments should be.
[55:08] Diego Bohórquez: Fantastic. Mijail, where are you located? Sorry, I missed that. Philadelphia. And if at some point, I like face-to-face, and if I have to come up there, I'll be happy to go in at some point, visit you guys.
[55:26] Mijail D. Serruya: Michael, happy to visit you in Boston. Sure.
[55:29] Diego Bohórquez: So we roll it in a little bit more, and then we bring the next layer. It will be postdoctoral fellows or somebody or senior scientist. Then, usually, they make the magic happen.
[55:45] Michael Levin: We can just fantasize about stuff, and then we got some poor postdoc and be like, "Here. Make this thing work."
[55:53] Mijail D. Serruya: That's right. And then the postdocs, as they're doing their tasks: wait, why am I doing this? How is this communicating to me? I'm just doing these, and I don't know how someone controlled my decision point. Exactly.
[56:05] Diego Bohórquez: that's it.
[56:06] Michael Levin: That's the actual invention here. Is this a power bar or something that gets people to invent whatever you want them to invent?
[56:14] Mijail D. Serruya: That's that's right.
[56:17] Diego Bohórquez: This conversation into a power bar.
[56:21] Michael Levin: Just to finish this off, the closest thing to that is back when Jim McConnell in the sixties was doing the stuff where he would train the planaria and cut their heads off and show that they regenerate. That part is true. We replicated all of that, and that's true. He did one other thing, which I didn't replicate, and I don't know, but I have no reason to think it's wrong. He also did this thing where he would grind up the trained planaria and feed them to the others. And he reported memory transfer, which from Glanzman's work, we know that you can transfer RNA directly. Glanzman would train these animals, extract the RNA, inject it into a new host, and show the memory was transferring. The coolest part to me is that I asked him, "Do you inject in a particular neuron?" He goes, "Nah. I just inject it somewhere in the brain." That's the amazing thing is that information, you don't have to be micromanaging this. These systems are so good at making up for our sloppiness, and I think that's because biology is often all over the place. You can't assume pretty much anything if you're a living system. So they're good at picking information up from anywhere. McConnell, having shown that apparently planaria can pick up this information from their digestive tract, jokingly suggested—I don't know how much he was joking, but he suggested this thing he called professor-burgers. The idea that at some point, you would take your knowledge and make it into some sort of a hamburger, and then the students could just eat it, and they would take it up. I don't know how far this goes in mammals, although people did later synthesize fear-of-the-dark peptides and things like that from trained rats. Anyway, that's it. Maybe that's where we're going. Maybe this is back to professor-burgers.
[58:31] Diego Bohórquez: Professor-burgers. Man, this is a nice closing image. I actually heard this story in the Amazon that cannibalism occurs because of specific nutrient deficiencies.
[58:45] Michael Levin: What if it's an informational deficiency?
[58:48] Diego Bohórquez: Well, food is information. Right?
[58:52] Michael Levin: So what can I say? Why don't you guys come here, and we'll have a nice dinner, and whoever survives will be much smarter than he was before. Right?
[59:05] Diego Bohórquez: I don't expect to be boiled in that pot.
[59:07] Michael Levin: I promise nothing. You guys are very smart. I can certainly use this, so there will be only one key, Michael.
[59:20] Diego Bohórquez: That will be handed and where we get