Can Bioelectrical Signals Reverse Aging? | Dr. Michael Levin
Live Longer World Podcast Episode #19
Aastha Jain Simes is the creator of Live Longer World, a podcast and newsletter where she interviews scientists researching the frontiers of longevity science and biology.
"Asking the question: Why do we have Cancer is the wrong question. The real question is, why is there ever anything but cancer. Why are individual cells able to work together to build complex organs?"Live Longer World Podcast Episode #19 has been released!
My guest today is Dr. Michael Levin. He is an American developmental and synthetic biologist at Tufts University where he heads his own lab. He also directs the Tufts Center for Regenerative and Developmental Biology.
His groundbreaking research focuses on bioelectrical signals that dictate the collective intelligence of cells and direct it to form tissues and organs. His work on bioelectric signals has important applications for how we understand cancer, regenerative medicine and even aging. In fact, he is collaborating with David Sinclair to see how his work can be applied to reverse aging.
Michael Levin’s work is fascinating and will surely leave you thinking. I hope you enjoy the conversation!
Best, Aastha.
Listen to the Podcast:
Apple Podcasts: here
Spotify:
YouTube:
Google / All Platforms: here
Time Stamps:
0:40 How do cells decide what to build?
10:05 DNA is insufficient for dictating collective intelligence of cells
19:00 Anatomical Complier (End game)
22:24 History of Cellular Collective Intelligence & Ala Turing
26:30 Regenerative Medicine & Morphoceuticals
38:05 Cells forming new anatomies
44:53 Cancer as a morphogenetic problem
55:16 David Sinclair, Bioelectrics to reverse aging
57:44 Synthetic Wombs & Artificial Life
59:41 Support Live Longer World!
Transcript
Aastha Simes: Very excited to talk to you about your research today. Huge fan of it. I want to start by reading out a sentence from your website where you say that, "Most of the interesting questions in biology boil down to the control of shape. We all start life as a single cell, the egg, which somehow self-assembles into an incredibly complex organism, and the question of how it's able to achieve its intended pattern or morphology as you call it, is the main issue of developmental for biology."
I think this is fascinating because what you're saying is that cells come together to form these tissues and organs and entire complex organisms. A lot of your research focuses on the fact that cells have this collective intelligence that they use to form entire organisms or bioelectrical signals. I'll let you explain what you mean by collective intelligence or bioelectrical signals.
Michael Levin Levin: Sure. There's a couple of things going on here with the focus on embryonic development. One is, if we just really practically think about biomedicine, you can notice that with the exception of infectious disease, almost all of the problems of medicine would be solved if we could tell cells what to do. If we have the ability to determine what it is that cells build together, then birth defects, traumatic injury, cancer, aging, degenerative disease, all of these could be resolved because you could simply tell the cells to build healthy, new organs.
Morphogenesis is a really fundamental problem in biology and medicine in particular, in terms of this question of how do collective cells decide what they're going to build, because it's clear that they can build multiple things. They normally build the same default kinds of patterns, but how do we understand that? How do we control it? That's the first thing.
The other more deeper, more philosophical way to think about it is that a lot of times, people will look at themselves or other humans or even other what's so-called advanced animals, and they will say, "I am a cognitive system. I have true beliefs, memories, cognition, right preferences," that there is just physics. This other thing, whether it be some AI that somebody built or a synthetic construct, whatever, that's just physics, that's just mechanism.
We all in our life made that journey from just physics, which was a single cell with all the chemistry that goes on there, but very slowly step by step with no magical life flash of any kind when cognition shows up, we become this highly sentient being with a first person perspective and so on. That process is very slow and smooth, and we all journeyed through that. That helps us really think in an important way about where things come from, such as cognition, such as memory preferences, all these issues. I think it's very important to keep in mind where we come from, both evolutionarily and developmentally.
Aastha Simes: I see. There are two points you mentioned. One of course, once we understand how cells process this information, it can have important implications for regeneration, aging, cancer, as you mentioned, and we'll talk about some of that. Then the second one, super interesting from a philosopher's standpoint then, I guess is that a way of saying we're maybe programmed and once you figure out these electrical signals, you can essentially create these organisms with memory and cognition, or free will, or consciousness, or what have you. I think you've probably shown that maybe initially with some of your experiments around xenobots as well.
Michael Levin: Yes, I think that's unavoidable in a modern scientific worldview. I think it is unavoidable, and that in taking evolution seriously, taking developmental biology seriously, realizing that all of these changes are very slow and gradual. I think it is unavoidable to think that the processes that give rise to cognition, memory, and so on in us and in other animals are something that not only can arise through the processes of evolution, but of course can be engineered as well.
There's nothing really magical about evolution. It's this large-scale hill-climbing search through the space of possible bodies, which give rise to possible minds. There's zero reason why engineers couldn't do better than a more or less randomly guided process that basically just optimizes for biomass. It's not really optimizing for intelligence or anything like that.
Yes, I think it's absolutely likely that we will be able to engineer those things that raises many important responsibilities for us both ethically and scientifically to understand really what we're doing.
Yes and I think we can talk separately about the issue of programming, but just in general this idea that there's a deep continuity between things that we consider to be just chemistry and physics, and things that we consider to be cognitive systems. There is no sharp dividing line there. It's a very smooth continuum.
Aastha Simes: Fascinating. What you're saying is that if I understand, say, your bioelectrical signals, I can create another Michael Levin Levin just like you, maybe thinks like you, has similar intelligence, but I guess with your experiences that would evolve as well.
Michael Levin: Yes. Let's go back to where the bioelectrics comes into all of this. We are all collective intelligences. People often think of, "There's collective intelligence like ants and termites, and bees, and things like that, and then there's real intelligence like me. I'm a real intelligence in the sense that I have a centralized-- I don't feel like a collection of things. I feel like a single unified--" Obviously there's all kinds of data in cognitive science and psychology that says that's mistaken in many ways. The reality though is that we are all collective intelligences in the sense that we're all made of parts.
You and I are bags of neurons basically, and some other things. The mystery here of course is the scaling. How is it that a collection of individual cells, which themselves are cognitive agents, they're very competent in their local areas, they know how to solve problems that are metabolic problems, physiological and so on, how do they join together to form this new emergent entity that knows things that none of the individual pieces know, that have goals and preferences that the individual parts are unable to comprehend and so on?
That means that there has to be some mechanism that acts as a cognitive glue. It's something that will take individual sub-units and bind them together in a way that they're not just a pile of stuff, but in some way a new capability of following much larger goals ,of storing larger memories, of thinking about bigger things. Somehow that has to happen. Now in the brain, of course, we know how that happens at least. At least we know the pieces involved. We don't know most of the important stuff, but we know the pieces. The pieces, it's an electrical network.
Basically what evolution has discovered, which is that electrical networks amongst cells are an excellent mechanism for coordinating information across distance, for binding sub-units into decision-making circuits, for representing information as memory, all these kinds of things. I believe that bioelectricity is great for that. Then it's very simple. You just ask yourself, where did neurons come from? Neurons didn't just appear. This magical trick that the brain does didn't just appear out of nowhere. It basically just speed-optimized things that cells were doing long before neurons and brains appeared.
Bioelectrics was evolution-discovered bioelectricity around the time of bacterial biofilms, but that far back already, it was selectively advantageous to exploit that type of physics to coordinate information across space in a multicellular kind of structure. Now one can ask some other interesting questions there too. Before neurons and muscles appeared, where at that point you would be thinking about behaviors you're going to do in three dimensional space, what did tissues think about before there were brains? In the early evolutionary lineages when the bodies were using, which they still are, using electrical circuits to control anatomy, what information were they processing?
I think the answer is they were navigating morpho space. Morpho space is simply the space of all possible anatomical configurations. There are areas of that morpho space belonging to a particular shape of the face of a frog and the foot of a chicken, and all these things. All these things are regions of morpho space. Bodies navigate morpho space the way that modern animals navigate three dimensional space, and you need an information processing tool for that. That's what bioelectricity does.
Aastha Simes: I see. It's typically assumed though, that DNA is this information processing tool, and you say that, "No, DNA is just this hardware. There's bioelectricity, which is actually the cellular collective intelligence or information processing." Can you talk a bit about that? How do we know that this information does not come from the DNA?
Michael Levin: Sure. I'd like to take a step back and just again make a more or less of a philosophical point, which is that what I'm not arguing is that there is one correct way to look at this and that mine is the correct way, and that everybody else is wrong. I'm not saying that. What I'm saying is that there are multiple-- Basically, any kind of a model-- There's a famous saying, I forget who said it, but there's a famous saying that says, "All models are wrong, but some are useful." I think that's very true.
This question of when we look for a controller, when we look for information, when we look for a hardware, software, all of these things are formal models that we come up with, that we try to impose on the world. My point is, you impose them, you see how well that does for you, and then you just decide, "That model is good for these set of scenarios, it's maybe not so great for this other set of scenarios." What I'm going to propose is that there are biological questions in which it makes perfect sense to look at the DNA as the source of your information.
If you want to know where the proteins in your cell come from, and what determines the various hardware components that the cell has, for sure the genome as your source of information. It's a really good metaphor that served us very well for many decades. That's great. If you want to know where large-scale pattern comes from and how collections of cells make decisions about various anatomical outcomes, I'm going to argue that the genome is insufficient. The reason it's insufficient is for the same reason that if I gave you a device, a computer that played chess, it's one of those chess-playing games.
I gave you a device that plays chess and you said to me, "This thing is really just a collection of silicone and copper, and things like that," I'm going to study those and in the end, I'm going to know everything there is to know about this. I would say, in a certain sense, that might be true. If you had the age of the universe to manipulate every atom and whatever, you might have some insight into what's going on. Maybe you could even make some predictions about what it's going to do but none of that is practical in our lifetime.
Also, you would be completely missing the most important thing which is you would never recover the rules of the game of chess from this. You might be able to predict it, Laplace's demon kind of thing. You might be able to predict what it's going to do just by tracking the electrical fields and magnetic fields, but you would miss the most exciting thing about this, which is that it's a device with goals. It has goals in a really weird and interesting space. It has goals like, "Protect the king. Dominate the center of the board," get whatever it is that chess players do. You would be missing out on all of that.
Now the thing with DNA is we can read genomes now. We know that there is no direct encoding for size, shape, none of that is in there. The only thing that's in the genomes are protein sequences. Here's an example. In my lab, we are making something we call a frogolotl. Frogolotl is an embryo that's some percentage frog cells, some percentage axolotl cells. Now, here's the thing. Baby axolotls have legs, frog tadpoles don't have legs.
You have the genome of the frog, you have the genome of the axolotl, can you tell me if frogolotls are going to have legs or not? There's no model that would let you just answer that question. If I have a planarian flatworm with a flat head, and one with a round head, and I mix the stem cells from one in the other in a single animal, and I cut the head off, what shape are we going to get?
We have all these great papers in science and nature about the molecular, the genetic components that control stem cell differentiation, that's great, but there's not a single model based on any of that stuff that will tell you what head shape they're going to make because that is a different level of question. In neuroscience, it's the equivalent of trying to resolve psychological issues at the level of synaptic proteins. Usually that might work, but generally speaking, that doesn't work.
We have many examples where the most efficient representation of what the system is going to do in terms of what shape it's going to regenerate, whatever, is very well controlled and read from the electrical information, but it's very poorly controlled and read from genetic information. That's, to me, the only marker or definition of what it means to have an answer to that question of where does the information come from.
Philosophically answers, things like, "I like all my explanations at the level of biochemistry. I want to be a reductionist." Those philosophical kinds of pronouncements are, to me, of very limited value. What you really want is, here's my metaphor. Here's what it enables me to do. Here are the experiments, the biomedical applications. Here's what enables me to do. Let's see yours. Let's see what yours allows us to do. Then we will know which one is more or less useful in different circumstances.
Aastha Simes: That's a great point, just working with different models and see which one's useful. On the point of if you combine two different heads, have you done any of those experiments, and have you shown what ends up happening?
Michael Levin: We're in the process of doing those experiments. I don't want to focus on the answer because the answer isn't the point. It's not the point where the frogolotls have legs, there's an infinite number of combinations that you can make. In fact, there's even a bigger point which is, never mind the combinations, if I just give you a genome, can you tell what the shape is going to be?
Now, you can cheat and compare that to a different genome where you already know what the shape is. That's fine. Comparative genomics, you can do that, but that's cheating. What you really want to do is be able to look at the genome and say, "Can I tell you what the symmetry type, the size, the shape, regenerative capacity, what is this thing going to be?" We have almost zero ability to do that.
It's not about the outcome of specific experiments. It's really the question of, what is it really that you're asking? What we're asking is not questions about specifically the hardware, like which proteins do you have? DNA is great for that. It's how do you make decisions in novel circumstances? We have lots of situations, and I can tell you about some, where cells and tissues are confronted with novel situations that they've never seen before, either during their own lifetime or during evolution. In fact, evolution never prepared them for that specific circumstance.
They're able to do problem-solving. What we see is that evolution doesn't, in fact, provide solutions for specific environments. What it provides are machines that are able to solve problems in a range of environments, very interesting. That plasticity and that generalization is super, super interesting. Those are the kinds of questions that we really want to understand, which is how do they make decisions? How do they generalize? How do they learn? How do they solve problems? None of that is usefully addressed at the level of the DNA, or the proteins for that matter.
Aastha Simes: It almost sounds like it's related to say, the theory of knowledge or epistemology, but more from a biochemistry or biophysical perspective.
Michael Levin: It is. I think that a lot of work gets done in this area where people think that they're avoiding philosophy. When you think you're avoiding philosophy, that just means you're doing very bad philosophy, meaning that you're neglecting some really important questions that you should be keeping track of. I think epistemology goes into it.
Some people will say, "I only believe in chemistry, I'm a reductionist. I don't believe in goals. I don't believe in purpose. I don't believe in any of these large-scale things that we think about." Of course, if somebody is a real reductionist, then they would really like to be talking about quantum foam not chemical gradients and so on. It gets to that question, what does it mean to believe in the existence of something like that?
Aastha Simes: Amazing. Before we go into maybe some of the applications of your research, I guess one last point is that you mentioned that the goal ultimately should be that we can create this anatomical compiler. Any sort of electrical signal you give it, it should be able to create whatever shape you want.
Michael Levin: To be clear, the anatomical compiler is not just about bioelectricity, it's a bit of a different claim. I happen to think that bioelectricity is very important, but the anatomical compiler is a much more general type of concept.
Aastha Simes: Maybe you can talk a bit about it. What is it then?
Michael Levin: The idea is that, if you ask-- I sometimes ask the people in my lab this when they join, I say, "What is the endgame here? When can we all go home and consider that we're done? I think it's important to look forward and to try to identify what is your actual goal for these things. I think the actual goal is the complete control of structure and function, and dysfunction. Basically, any shape that you want to make, you should be able to make.
What does that look like in practical terms? You should be able to sit down in front of a computer and you should be able to draw the way we draw up car parts and various other things in CAD, computer-aided design, you should be able to draw what you want. You're drawing anatomy, you're not drawing molecular biology. You don't know anything about pathways, you don't need to.
When this field is mature, somebody should be able to sit down and say, "I want an organ. I want a heart but one that looks like this," or, "I want a perfectly normal human eye," or, "I want a frog and I'd like it to have six legs and tentacles," or whatever. I'm making that up but you understand. You should be able to draw absolutely anything. Then if we knew what we were doing, the system would compile that description into a set of stimuli that would have to be given to cells to make them build it, whatever you drew.
Now, I'm sure partly will be bioelectrical but partly it won't. There will be chemical cues. There will be biophysical cues. There's evidence for utra-weak photon communication in the body. There's all kinds of other modalities. The point is that's what it should look like. If this field were solved, that's what the solution would look like. You should be able to draw anything you want and out would come a set of instructions for stimuli that here's what you do to the cells to get them to do this.
Aastha Simes: Fascinating. Almost sounds like science fiction to me.
Michael Levin: The job of science fiction is to imagine what the science should look like going forward. From that perspective, sure it's science fiction. There's a difference between-- I love science fiction. I've read a lot of science fiction. There's a difference between things that are actually impossible, and are many reasons why things might be impossible.
Then there are things that are purely limited by existing resources, existing concept imagination. This is squarely in that camp. There is nothing impossible about this. We know that something as with the limited foresight of natural evolution is able to produce these amazing bodies. Can you imagine what we could do rationally if you knew what we were doing? It's like science fiction.
Aastha Simes: I'm curious then, is this relatively new then? Have people been working on it maybe for a long time and they just didn't have breakthroughs, or is there a limit to imagination?
Michael Levin: Specifically the compiler you're talking about?
Aastha Simes: The compiler or just talking about collective intelligence of cells and how they process information.
Michael Levin: Let's look at it this way. The question of how single cells become adults has been around since the time of Aristotle and probably before that. The interest in that question is very old. Developmental biology can probably be traced back to the 1700s, something like that. Scientists have been trying to study this process ever since. For sure, this is a very old question. Now, what's new here is this idea that using an interdisciplinary approach where you take things from physics, from computer science, from cognitive science and you use them to apply to developmental biology, it's new and it's not new.
It's new because the trend over the last 60 to 80 years has been avoiding all of that stuff, and really developing something very focused on molecular genetics, biochemistry. That's been the approach. It's been considered inappropriate to use tools from other fields like cognitive science. You're not supposed to be thinking about software or things like goal-directedness or learning in these systems because biologists are really terrified about this.
This is a kind of teleophobia that operates here, where as soon as you start thinking about goals and software, and things like that, the danger is that somebody is going to slip into, "If there's an algorithm, somebody must have written it." Then there's some semi-religious thing, and biologists really hate that. They try to be intentionally mind blind. There's this mind blindness that exists where people can't have a theory of mind about others. They see everything as a mechanism. Scientists are intentionally mind blind in that way.
It's new-ish because it's been that way. It's just starting to crack that, that paradigm is just starting to crack now. On the other hand, if you look back to the classical workers in this field, all of this has been said before. The forefathers of developmental biology, I think it was Driesch probably who said that the question of cognition and the question of development are the same question. I think that is profoundly important and interesting, and true, and it's been ignored for many decades. It's new and it's not new. Conceptually, I think people have seen this a long time ago though, really brilliant mindsets seen this a long time ago.
However, what's also really new here is that now we know a lot more about now there is a science of collective behavior. There is a science of complexity. There is a computer science which didn't exist before. Isn't that amazing that Alan Turing who many people consider the forefather of computer science, he was interested in intelligence. He also wrote some of the first papers on mathematical modeling of morphogenesis. He was interested touring patterns, right?
He had this incredible paper about how order arises in biology and Turing patterns. You ask yourself, why would he be interested in this? If you're into AI and computers especially in the '30s and '40s where there wasn't really any computational biology or synthetic biology, or anything like that, why would you be interested in both of those things? Because there is a profound symmetry here. I think if he had lived, we would've seen way-- This field would be way further along.
Aastha Simes: I'm very glad you're working in it. Very interesting work. I want to talk about some of the applications so it's more concrete for people. You talk about some of the applications being in regenerative medicine are potentially seeing cancer as a morphogenetic code. Maybe for the issue of regenerative medicine, if you could perhaps explain using some of your experiments with flatworms or tadpoles or whatever you think is easiest to explain it.
Michael Levin: There are three broad areas of application of our work in medicine; birth defects, regenerative repair, and cancer. Let's start with the regeneration. The basic issue is that all of your body organs were created at one point. They were made by cells. That information is still there. If somebody's injured and they lose an organ for whatever reason, presumably the information on how to build it is still there. The trick is then to convince the cells to do it again, to rebuild it. How would you convince them to rebuild it? You might start thinking about how did they know to build it in the first place?
If the answer to that is there's a bioelectrical pattern memory that helps them understand what to build, maybe you can reactivate that again. We have a lot of work. For example in planaria, what we've done is we've said the amazing thing about planaria is that you can chop them into pieces. Every piece knows exactly what's missing and regrows exactly the right parts. If you chop a planarian into thirds, that middle trunk fragment will grow a head at one end and a tail at the other end. Things like questions about how many heads of you supposed to have?
Where does the head go? Where does the tail go? Those questions are easily resolved by the pieces. In fact, the record is something like 275 pieces. It can be actually very tiny pieces. What we did was we simply looked at the bioelectric. We use the voltage-reporting fluorescent dye and we simply looked at those pieces to ask, what does the bioelectrics look like? We saw an amazing thing, that there's a pattern that you can literally read out that looks like what a normal worm should look like.
Now, the thing with regeneration, animals that can regenerate, the most amazing thing about regeneration is not that they rebuild these organs. The most amazing thing is that they stop. Because once you injure them or you cut off, let's say, you cut off the limb of a salamander or something like that, it will keep regenerating until a correct salamander limb or a correct planarian is formed, and then they stop. How do they know when to stop? If they know when to stop, there has to be some internal process that says the error isn't been reduced to the point where you are now correct. You correct enough and you can stop.
There's basically this homeostatic process. I call this anatomical homeostasis where much like a thermostat that basically will act to reduce the error from a certain set point, the body will act in terms of all kinds of cell behaviors to reduce the error of the anatomy. Now, the thing about those kind of processes, those homeostatic processes is that they have to store a set point. Somewhere you have to record, what am I trying to reach? What's the information? In a thermostat, it's very simple. There are two numbers, it's your temperature range.
In regeneration, it has to be more complex. You have to store some level of a course-grade description of the anatomy, so that we know when it's right and when it's wrong. We had this idea that, this is a hypothesis that I made many years ago, which is that-- Actually even that, again, it was considered very new and very crazy when we first started talking about it originally. It was originally said by Harold Burr in 1936, who basically had no technology to work with except the voltmeter. He made the first good voltmeter. He used to go around measuring all kinds of living things.
On the basis of that, he already pulled out this theory that what if the tissue was storing a bioelectrical pattern that reminded it, what the shape should be in case it gets injured. In planaria, we saw that we could actually see it. It was amazing. We could actually read the set poin. Because the super important thing about all of this is that, here's what's cool about your thermostat. What's cool about your thermostat and all devices like it is that if you want to control the temperature of the room, you don't need to understand how your thermostat works. You don't need to rewire your thermostat.
All you need to know is A, that it is a thermostat, and B, how to read and write the set point. Once you understand that it will obey the set point and you know how to control the set point, the rest of it, you don't care about it. Doesn't matter. You don't need to change the wiring. You don't need to do any of that, because it's a good thermostat. This is an incredibly important path forward for regenerative medicine, because if it were true that there was an encoded set point, then you could manipulate the set point and you wouldn't have to rewire the genetics underneath.
Why is that important? Because we don't have any clue, for the same reason I told you before that genetics was not a good tool for understanding large-scale decision-making in terms of patterning, because other than single gene diseases, we don't have any idea how to solve this inverse problem of saying, "I want particular finger to have this particular shape," or, "I want the heel of the foot to have a different shape." What genes am I going to edit? All the CRISPR technology you want, you don't know what genes to edit to make that happen. It's very difficult.
If our hypothesis is true, it means that you could control the set point and let the cells do what they do best, which is build to the set point. Then you could go beyond the normal set points and code some other stuff that they weren't going to build, but that they will once you've encoded. You can control it. In planaria, we found this electrical pre-pattern.
Then we figured it out using ion channel drugs that open and close ion channels to manipulate the voltages, we figured out how to reset that pattern. What you can do is you can take a planaria, you can look at the patterns of having just one head, one tail. You can reset it to be bipolar symmetrical, meaning two heads. Then guess what the cells do. They will build a planarian with two heads.
Aastha Simes: You've shown this in labs. You've been able to build planaria with two heads.
Michael Levin: Yes, sure. We've published half a dozen papers on this. There's videos of these two-headed planaria and everything. Because, and this is very important, the question of how many heads should you have is not locked down by the DNA. What the DNA does lock down is the hardware that by itself reliably produces a pattern that makes one head. That's not the only thing it can make. If you rewrite that pattern, the cells are just as happy to build other patterns. That's where the analogy of hardware and software comes in.
It's not about who wrote the algorithm, there's no human writing this algorithm, it's about a machine that separates the data from the execution component. The data is how many heads should we have? It's separate, independent of the cells, which will consult that data and say, "It says two? Then we build two." That's the important piece here that it's not-- None of this would be reachable if you think about it from the perspective of DNA, then you would do what all of molecular medicine does today, which is drill down on the hardware. It's rewiring genomic, rewiring circuits and transcriptional circuits, genomic editing, protein structures.
None of this is apparent at that level. In fact, if you were to sequence these two-headed worms, you could get the genomics, you can get the proteomics, you could get anything you want. You would never know that they're two-headed because their genome's a wild type. You didn't edit the genome, there's nothing wrong with their genome. That's not where the information is. We can make two-headed worms. We can make no-headed worms. Then we show that actually in the anatomical, morpho space, guess what? There are other shapes, belonging to other species of planaria.
We show them that you can take a genetically normal piece of a worm and cause it to find a completely different-- Basically to make a different head shape that belongs to different species. The brain shape, the head shape, the distribution of stem cells becomes just like another species. There's nothing genetically wrong with them, it's the same genome. Again, if you were to sequence them, you would find that you would have zero clue that they have in fact had a completely different head that belong to a different species. That's the kind of stuff you can do in these systems. That's in planaria.
In frog, what we did was to look at organ regeneration and we said, already we can control whether you get a head or a tail without having to control directly the underneath molecular biology. The signal that we provide is very simple, so we don't need to micromanage the process by trying to guide all the hundreds of thousands of gene expression events that have to happen. We don't control any of that. We don't need to. The system already does that. We provide the upper level decision-making of, how many heads should you have? Then the system takes care of the rest because it knows how to make heads.
We said, "Could we go even upstream of that?" What if I don't even tell you what to build, I just activate a "build whatever normally goes here" signal? Just build whatever normally goes at this location. Presumably you already know what goes, I just build that. We started with tadpole tails, and this is important because tails have spinal cord and we were interested in getting muscle and spinal cord, and that kind of thing. Then, we did frog legs. Adult frogs unlike salamanders do not regenerate their legs. What we showed is that after amputation, you can provide a very short trigger that would convince the cells to just start building whatever they normally build.
They would build a leg or a tail. They would never build an eye or a tumor or anything else, they would build exactly what happens. In the case of the tail, one hour exposure of the stump to the right bioelectric drugs, kick-starts a set of events including all the downstream molecular biology, everything else that you would otherwise have to micromanage. It gives you nine days of tail growth. Just from one hour. In the adult fraud, one day, so 24-hour application of this cocktail gives you a year and a half of leg growth, during which time we don't touch it at all.
Everything we do is in the first few minutes, it's all about making the decision, pushing those cells to a decision point and then leaving it alone, and letting the cells do what they need to do. This is probably a good time to do a disclaimer. There's a company called more Mophorceuticals Inc, which David Kaplan and I co-founded. We are now trying to push that into mice. The idea is, can we push it to mammals? Hopefully eventually human patients. That's the regeneration story.
Aastha Simes: I would love to hear how the experiments go there. I'm curious. You said that, frogs for example, tadpoles would build what normally goes there, or planaria already have the information to build ahead. Can you build something which it doesn't already have the information to do? For example, can you build a tail in humans even though humans and have tails? Can you build an extra limb in a flatworm even though it doesn't have this information?
Michael Levin: Yes, we've made planaria that have the wrong species heads. We've made tadpoles that have tails that belong to zebrafish or faces that belong to different kinds of frogs that are not the frog we started with. You can do that. The question of information that it doesn't already have, it's a tricky question. If you think from the point of view of the genetics, then absolutely, because we can absolutely make a head shape of an animal, let's say, a planarian from a genome that you don't have that genome, you have a different genome. From that perspective, sure, we can.
I want to be careful about that because that's not the right perspective. The better perspective is that of software and hardware. It's of a collective intelligence navigating through this morpho space. Once you have that perspective, it's not so clear anymore because once you have the ability to navigate morpho space, there are many, many things in this morpho space that you might be able to find. In fact, we've made planaria that look like hedgehogs, they're spiky, they don't look like a flatworm at all. They're three-dimensional tall tubes. All kinds of crazy stuff.
It's hard for me to say that they don't have that information. I think they don't have it in the sense that it's not internal, but I think they can reach that information by exploring morpho space. It's, again, a computational analogy. Once you've made a circuit that can do computations, let's say, it can do math, then you say, "Is the answer to five plus five in there?" It's not in there, but if you were to pose the question, if this thing knows the algorithm, you can certainly find it. Then you get to the interesting philosophical question of, well, where do the truths of mathematics live?
It sounds very philosophical, but it's actually quite practical here because when cells cooperate in novel ways to form new anatomy, like for example, xenobots, there's never been xenobots ever, there's never been this evolutionary selection to make xenobots, how come they know how to make xenobots, and with all kinds of new behaviors and everything else? Then the question is, where does that exist? I think in the end, it's the same question to where do the truths of mathematics live?
The fact that if you're evolution, and you're building a triangle, and you nail down the first two angles of that triangle, you don't need to evolve the third. How come? Because you already know the third. If you know two angles, you know the third. Where does that come from? There's no gene for that. It's a free gift from mathematics, from geometry, from where? That's where this comes in. You can make these devices. Evolution makes devices that exploit physics.
Once you exploit physics, you have laws of computation, you have laws of mathematics that help you do these kinds of things, it's a stupidly simple example with this triangle but that's the idea. You don't need to evolve that third angle, it's there for you for free. It's tricky just to say whether they have that information or they don't have it. You don't need to have it in the genome in order to-- The genome, those are the contingent details, but everything else you can actually find.
Aastha Simes: I see. Interesting. It actually reminds me, I was reading David Deutsch’s book last year. He talks about, you know how animals are actually restricted by the knowledge that's contained in the genome. A lot of your work might show otherwise now once we go live.
Michael Levin: If there are restrictions, if you have a genome that does not let you make voltage-sensitive ion channels, you are missing a very powerful way to interface with feedback loops, with the logic gates, things that transistors are good for. Our voltage-gated ion channels, so basically transistors. If you don't have that in your genome, there are things that are very hard or maybe impossible for you to do.
From that perspective, sure the genome constrains what you can do. The genome enables other things you can do. Once you have adhesion proteins, you can take advantage of all kinds of cool physics of adhesion and sorting, and all of that. Yes, it's true that it constraints, but it does not determine all of the possibilities. It's not a controversial claim. It's like hardware and software. When you buy a PC, does it determine what you can do with it? Yes or no, right?
Aastha Simes: Yes.
Michael Levin: Some things you're not going to be able to do with it based on the hardware, but there are so many things that you could do with it that are not directly described by that hardware.
Aastha Simes: Now we're just expanding the possibilities of this software, learning more about it.
Michael Levin: That's right. Because if you don't know what subroutines are available to you, you have very limited capacity. This is why when I give a talk, one of my first slides is what programming looks like in the '40s and '50s. It's great. It's this big, giant thing and this woman sitting there. She's literally plugging wires in and out, because in those days, programming had to be at the hardware level. Everybody laughs and I say to people, "How terrible would it be if on your laptop, if you wanted to switch from Microsoft Excel to PowerPoint, you would have to get out your soldering iron and start rewiring?"
Everybody says, "Ha, ha, ha, would be terrible." How come you don't have to do that? Isn't it amazing? Why aren't we just as outraged when you do that in molecular medicine? That's all you have in molecular medicine is direct-- You're going to direct attempts to control the hardware. I truly think that some number of years from now, we're going to look back and we're going to have the same picture of somebody doing CRISPR on a generic circle and go, "Ha, ha, ha, ha, look at that."
They're having to program at the level of hardware. It's going to be the same level of amusement.
Aastha Simes: That's a great analogy. Another application you mentioned was cancer. Cancer is, say, just disorganization at the morphogenetic code level. I have two questions, that one, can you explain this? Then two, do we have any sense of why this disorganization at the information processing amorphous genetic level starts to occur?
Michael Levin: Yes. Let's go back and ask ourselves. Let's just think about what cancer is. I want to be clear by saying that cancer is a very complex set of diseases. There's all kinds of clinical manifestations. I'm once again not claiming that I have the answer to all of this, I'm providing what I think is a useful way to think about it. Asking the question, why do we have cancer is the wrong question. The real question is, why is there ever anything but cancer? Why are individual cells able to work together to build complex organs? Why do they not all act like amoebas, which is basically metastasis?
We used to be amoebas, all of us, and bacteria? How come there's anything else? Once you realize that, yes, our native state was single cell, eventually we got together and started working on cooperative projects like, build a kidney or build an arm, whatever mechanisms we have for coordinating that, inevitably, at some point, there's just going to be breakdowns of that mechanism. That tells you that cancer is the occasional price we pay for being multicellular organisms, for being made of parts.
Aastha Simes: When you say inevitably there will be any breakdowns, I guess, why do you think that's inevitable?
Michael Levin: Anytime you have a mechanism that does anything, there's a chance of it breaking down, right? Is there going to be a breakdown of gravity? No, because gravity doesn't require a machine to keep it going. Would there be a breakdown of a bicycle? There's a good chance because there are particular things that have to happen for this to be a workable bicycle. At some point, there's a chance that it can go wrong. I'm not putting any statistics on it. I'm not saying every human at some point has to get cancer. I'm saying that anytime there's an active mechanism required, there's some chance that it's going to break down.
Let's think about what actually happens during that breakdown. When you have cells connected to each other specifically in electrical networks, I don't have time to go into all the details, but we've described this in papers, what's important about that is there's a loss of individual identity in the cells. Because when two cells are connected to each other very tightly, then all the memories are shared, all the measurements that they take are shared. It's like a breakdown of individual identity. Instead of a bunch of little cells, you have a collective that has this emergent goal. In that case, it might be building some sort of organ.
The point is that it's impossible for the cells to defect, or in the language of gain theory, not cooperate with each other, because they are not unique individuals at that point. They are so tightly coupled, that the network is able to store goals that individual cells don't have. We know how this works in artificial neural networks, to some extent anyway. That's what happens. Now think of what happens during the breakdown. One way things break down is there are specific kinds of oncogenes, like KRAS mutations, where one of the first thing that happens after that mutation is that cells become electrically disconnected from their neighbors.
As soon as the cell becomes disconnected from its neighbor, the boundary between itself and the world starts to shrink. For a collection of cells, the boundary is quite large. It might be a whole finger, a whole arm, or the whole body. It's quite large. Once the cell becomes disconnected, that boundary between self and world shrinks. As far as that cell is concerned, the rest of the body is just external environment. It's no longer plugged into the goals of that network. It's now, "I'm now an amoeba like I used to be a billion years ago. What are my goals? My goals are to become to amoebas and to go wherever life is good. That's my goal. That's metastasis."
What you have is, once you disconnect from the network, the purpose of the network is to raise the IQ of the collective so that it can work towards goals that are bigger than single-cell level goals. Single cells can have metabolic goals, they can have some local shape goals and things like that. They can't contemplate the goal of making a finger or an arm or something like that. Only the collective can do that. When they disconnect, they don't become more selfish than normal cells. A lot of times people talk about cancer cells being more selfish, they're not more selfish, they just have smaller cells.
All the cells are equally selfish. It's just that when they're merged, it's like a mind meld. When they're merged into one group, that selfishness goes towards, "I'm making this organ in the body." By the way, that's a very selfish process because organs compete for energy and information with each other. Inside the same body, organs compete. That self is large but now when the self shrinks, now it's a single cell. Single cell doesn't care what happens to the environment. That's cancer. What we've been able to do basically, we've been able to do three things. Most of our work has been in frog models now.
In fact, just this morning, we published a paper on this in human glioblastoma. We're slowly moving to mammals in medicine. We've shown three things. Number one, that we can use voltage-sensitive dye technology to image and detect when cells are about to defect from the electrical network. Basically we inject oncogenes into a tadpole. When the cells get to-- When they become isolated, that voltage, that aberrant voltage potential, we can image it. You can see where the tumor is going to form. It's a basis of very obvious diagnostic kinds of applications. That's the first thing.
The second thing we found is that the standard story of cancer is that it starts with mutations, and that there's a set of mutations that what we are able to show is that you don't need any of that. What you can do is you can disrupt the electrical communication between cells, and you can trigger-- We've shown that you can trigger metastatic melanoma in a frog model with no mutations whatsoever. No mutations. No oncogenes. No carcinogens. No disruption of the genetics at all. The hardware is perfectly fine. It's like, again, it's new now.
In fact, it was so new people said it was outrageous when we first-- Our first PNS paper, the reviewer is like, "What's the genetic mutation here? Where's your founder cell?" We said, "That's exactly the point. There is no founder cell. There is no genetic mutation." He said, "Then it's not a cancer." That's a very funny way to do this. You're defining it the way you'd like to define it, but look at the phenotype. The animal is packed with these metastatic transformed melanocytes. In the early stages, there's nothing wrong with them. You couldn't tell that genetically you would never find anything wrong with them.
Later on, of course they turned on all the response genes, slug and snail, and all that stuff. We showed that just by disruption of electrical signaling, you can induce cancer. It doesn't have to start. It's like, in fact, again, it's one of these things, it was considered new and strange, and whatever, but in the '60s, this guy whose name I'm blanking on at the moment had this quote that said that you'll never find the cause of a traffic jam by looking at the internal combustion engine of the car. That's not where problem is. There's nothing wrong with these cars.
That's exactly the thing here. You can sequence these cells going, you go blue. There's nothing wrong with them. It's not a problem with the hardware. There's nothing wrong with the gene, nothing wrong with the hardware. The third thing we did, which is the most exciting, is we could go in the opposite direction. We could take embryos that were injected with oncogenes, like nasty KRAS mutations and things like that. We could artificially force their voltage to stay normal so that they stay in electrical connection with their neighbors. If you do that, you can suppress the tumors.
You can suppress, or even normalize tumors that already exist. You can do that with drugs. You can do that with optogenetics. You can do that with injected channel RNA. We were able to show that. Those are the stories in cancer. You can detect it, you can induce it, and you can normalize it. You don't kill the cells, the cells don't die. It's a different approach going forward. It's not some sort of chemotherapy where you're trying to poison one set of cells. It's not that you're trying to re-inflate the size of the self again so that these cells just get harnessed back to the histogenesis programs, not cancer.
Aastha Simes: That's very different. You're giving signals to these cells to once again become part of the collective machinery and the collective goal instead of whatever caused the disorganization that occur for them to go off on their own path.
Michael Levin: Yes.
Aastha Simes: Amazing. I'm curious, what did you show in humans? You said published something.
Michael Levin: It just came out this morning. This is the work of a staff scientist in my group called Juanita Matthews. She showed that human glioblastoma cell lines, and this is in culture. The next, of course we have to go in vivo. In culture, using already human-approved ion channel drugs, these are things people already take for other reasons, can prevent those cells. They can stop them from proliferating and induce a level of normalization. Basically, a degree of differentiation into normal neurons, things like that. It's a very long paper, but that's the idea. Normalization of these cells in culture.
Aastha Simes: Amazing. A few last questions. I know you have to go at 12:00. On the topic of cancer, I guess aging is something that comes up as well. I know David Sinclair recently posted that your team and his team sat together and brainstormed how maybe some of your research can have potential applications for aging. I don't know if you can share it publicly, talk about how you're even thinking about some of these applications applying to aging.
Michael Levin: They're clearly connected problems, planaria that are so highly regenerative do not age. There's no such thing as an old planaria, they're immortal. That's not an accident. The question is what's going on? How are they able to continuously regenerate any senescent tissues? By the way, going back to your original question about the DNA, because some species of planaria, because the way they divide, they just tear themselves in half and regenerate, so they don't necessarily go through sperm and egg,
it means that any cell that gets a mutation and doesn't die repopulates the new worm, that's the second half of the worm, then after they split will come from these cells.
They're continuously accumulating somatic mutations. That doesn't happen for other creatures, because the somatic we get in our body don't get passed on to our offspring. In planaria they do. Planaria have an incredibly messy genome. In fact, a single worm, all the cells are mixaploid, meaning they're all different number of chromosomes like a tumor. If you were to sequence that, you'd say, "This is a mess, this should definitely be a tumor." They are 100% the best regenerators around the rock solid anatomical control. The genome is all over the place. It's a total mess.
That already tells you how little we understand about what genomes actually do, and what's actually controlling things. It should be, it's not anything you see in any textbooks, but it should be a scandal of, why can this happen? With David, we don't know exactly what model system we're going to look at. The bottom line is they have some really amazing examples of the control of aging in their various mammalian systems in mice, and human cells, so on. We are going to look to see what the bioelectric component of that is, how much of that control, what happens to the bioelectric storing aging? Can you use the bioelectrics to try to reverse aging? Those kinds of things. We'll see. It's very early. We just starting. It's very early days.
Aastha Simes: Very exciting. Curious to see what happens. Last question on, I guess this has the application to synthetic biology as well. More recently, I think Elon Musk tweeted about something and then synthetic wombs became a hot topic of debate. I guess your research could also apply to creating synthetic wombs or just artificial life, right?
Michael Levin: There's a lot there to talk about. We don't really make artificial life per se. We make novel organisms, and it's because we start with existing cells in my group. We don't address the origin of life issue really. There are other people who do a very nice job of making these minimal systems that are like living cells, but not really. We don't do that.
We start with existing cells, and then we make novel organisms to test the ability of life to adapt to novel configurations and try to understand something about evolution by doing that. Wombs in the sense of humans, we don't have anything to do with that issue, but really to understand the intelligence of cells and their ability to adjust to novel perturbations.
Aastha Simes: If you enjoyed the podcast, you might also enjoy my newsletter, livelongerworld.com, where I share practical longevity tips and also upcoming releases of my podcast episodes. Thank you for listening and I will see you next time.