Ep 39: Bioelectric Computation (with Michael Levin)
How do animals construct tissues, organs, and limbs in the right places during development? How do some animals manage to regenerate missing body parts?
On this episode of Big Biology, we talk with Michael Levin, a biologist at Tufts University who studies how electric fields inside animals guide cells during development and regeneration. His work shows that electric fields play fundamental roles in structuring body plans and, in some species, can even be inherited across generations.
-
AW = Art Woods
MM = Marty Martin
ML = Michael Levin
NS = Nick Strait (Student Spotlight)
PD = Paul Davies (intro)
EK = Ellen Ketterson (outro)
MM: As usual, we're starting off this episode with a Student Spotlight. This week, we're hearing from Nick Strait.
NS: My name is Nick Strait. I am a second-year master’s student at the College of Charleston, working in Dr. Heather Spalding's lab. For my thesis, I am looking at the dynamics of stable isotopes and tissue nutrients in Hawaiian mesophotic macroalgae. Mesophotic reefs extend from the middle to the end of the photic zone, roughly 30 to 150 meters in depth. What's really cool about these reefs, particularly in Hawaii, is their high abundance of macroalgae, which is unexpected since these areas are low light and oligotrophic. I am interested in understanding how nutrients are influencing the abundance of macroalgae. To do this, I am analyzing nitrogen stable isotopes and percent nitrogen for mesophotic and shallow water macroalgal samples across the Hawaiian archipelago. Another really cool part of this project is I will be able to detect any anthropogenic input of nitrogen such as sewage or fertilizer. Algal tissue with higher N-15 in percent in values can indicate these inputs, and this data could then be used to address management and pollution issues around Hawaii.
AW: Thanks for sending that in! If you're a student, send us a one-minute audio clip describing your research, and you might hear it on the show. Send a voice memo to info@bigbiology.org. Here's the episode.
[instrumental]
MM: At the very earliest stages of development, a human and a crocodile look pretty much the same -- just lumps of amorphous cells. Over the next several months though, those lumps develop into vastly different creatures.
AW: This process of development, and similar processes like limb regeneration are profound mysteries in biology. How do organisms know which cells go where? How do they make sure a limb or digit is long enough, but not too long? And how do we get from a string of information in the genome to an animal made up of billions of coordinated cells?
ML: I gave a talk to a bunch of 9-year-olds at a school once, and I showed them an egg and I asked what determines whether it's a dinosaur or a bird or a snake that comes out of this egg, and everybody, even the 9-year-olds, immediately said, oh, you know, it's the genes, it's the genome, right? So, so, in theory, everybody knows this, but in reality, of course, when you look at the genome, you don't see anything in there directly about symmetry type, or organ number, or size, or eyes, or any of that stuff.
MM: That's the voice of Mike Levin. He's a biologist at Tufts University who studies development in animal body plans. He's trying to understand how organisms use electric fields to direct cells to move to the places they belong in the body.
AW: All cells have passageways built into their membranes that allow ions like sodium (Na), potassium (K), and calcium (Ca) to flow in and out. These passageways are called voltage gated ion channels, because local electric fields determine when they open and when they shut.
MM: These electric fields also represent a kind of memory. The current state of the ion channels encodes something about the past state of the electrical field. Mike says that many organisms use these electric fields to map out their own bodies.
AW: Mike uses a lot of different species in his work, but especially flatworms called planaria. Flatworms reproduce asexually by latching onto something and then ripping themselves in half. The head portion typically grows a tail, and the tail portion grows a head, so that both halves become whole again. Remarkably, they use electric fields to figure out whether heads or tails are missing.
MM: When Mike and his colleagues prevented cells from talking to each other electrically, they found that the half worms grew back the wrong parts. When the electric fields were distorted, the head half could be made to grow a second head, and the tail to grow a second tail. His conclusion was that electric fields are a key component of the animal's body plan. The fields that cells experience determine what form of animal they make.
AW: This is an amazing example of how organisms encode critical information outside their genome. What's even more remarkable is that this encoded information gets passed on to offspring. When Mike cut a two-headed worm in half, it regenerated two more two-headed worms. As previous Big Biology guest Paul Davies says...
PD: And so, the information about their physical form, their morphology, two heads or two tails, is propagated down from one generation to the next. But they have identical DNA. The genes are exactly the same in the two-tailed worms or the two-headed worms. That information is inherited. We still don't fully understand the mechanism.
MM: Now you might be thinking, fine, but these are just weird, small worms. Complex organisms like us and other vertebrates don't do such things. Wrong. Mike has shown that tadpoles use electric fields to figure out where to put the eyes, the mouth, the ears, and other key parts of the face. And if he distorts the electric field near a frog's gut in just the right way, he can cause eyes to develop there.
AW: On this episode of Big Biology, we're talking to Mike about how electric fields affect the information organisms inherit, how that changes the way we think about evolution, and how organisms are and are not like computers.
MM: I'm Marty Martin.
AW: And I'm Art Woods.
MM: You're listening to Big Biology.
[instrumental]
05:14
AW: So, Mike, it's just a total thrill to have you on the show, and we, we want to talk about a couple of your papers and just sort of run some, some big ideas by you that we've been kicking around, and talk a little bit about how they relate to some of our, what our prior guests have said. But just to set the stage for what we're going to talk about today, I want to maybe try to articulate what I see as a, as a key biological problem that you, that you've been working on really hard for a number of years, and that's maybe the fundamental problem in evo-devo biology, and that is, how do developing bodies produce the right tissues in the right place at the right time? And, you know, so we take an organism like, like me, I'm a bilaterian, so I have left-right symmetry, and I have a head and a tail essentially, and some limbs along the way. And, as I was developing from a fertilized zygote, my, my embryo made all the right decisions, or, I guess maybe questionable decisions...
MM: Yeah, bit presumptuous there, be careful.
AW: ...about where to put the main parts of my body. And so, how do, how do embryos do that overall?
ML: Yeah, that's a great question, but it, the problem is actually even much tougher than that, because in standard embryogenesis, you're starting out from the same reliable state, meaning a single fertilized egg, and you're ending up at ideally a, the same invariant outcome, which is a normal human morphology. So, that's, that process is hard enough. But we actually have in regulative development and in regeneration, an even harder problem of being able to start at different starting states, let's say a salamander limb that's cut off at different positions, and the biggest problem of all is how does it know to stop? So, in a process like that, right, you have to, it can't be just, be just hardwired, because, because you have to recreate exactly what's missing, no more, no less, and then you have to stop when the correct structure has been, has been produced. So, this issue of how do they know when to stop and how do they know when to build is super critical, and it's actually even, even more complex than simple embryogenesis.
AW: So, so, you would say that, that, understanding regeneration is sort of a step beyond just understanding, just understanding embryogenesis.
ML: I would actually say that I think the regeneration is the fundamental thing. I think the ability of cells to get together to implement anatomical homeostasis, the ability to reduce error between whatever's going on now and the pattern that you would like to make, is the fundamental property. Everything else is an offshoot of that. So, normal development is an easy version of that problem, but I think fundamentally that is what multicellular life is all about. It's coordinating, you know, the, the goals of individual cells into one large morphogenetic goal of the body, and this automatically gives rise to both development and regeneration.
08:00
MM: So, maybe, can you walk us through one of the examples of regeneration? I mean, just sort of articulate the nuts and bolts of what you're learning about how this happens, and how animals or whatever example you want to use knows how to stop?
ML: Yeah, so I'll give, I'll give one example that we found in frogs several years ago, and then, and then we can talk about planaria, where I think we actually know a little bit, even more about what's happening. So, the tadpoles of course have to transition to frogs, and if you look at their face, a tadpole face doesn't look like a frog face, and so, they have to rearrange their tissues. So, the eyes have to move, the jaws have to move, the nostrils. All the different organs of the head have to move around. And it used to be thought that what the genetic encodes is a hardwired set of movements, so, every tadpole looks the same, every frog looks the same. So, if you just remember where every, how much every organ is supposed to move in which direction, you should be okay, right? So, what we ended up doing is testing this, this hypothesis, because we suspect that there was more plasticity to it, and we made what we call Picasso tadpoles, and we'll tell you, well I could tell you in a minute about how, how we make them, but basically, everything is in the wrong place, so the eyes are on the back of the head, the nostrils are off to the side, the mouth is, you know, off to the side, like everything is scrambled. And the amazing thing is that those animals basically make quite normal looking frog faces, and that's because the organs move around in sort of novel and unnatural paths and continue to move around until they reach a correct frog face. Now, this is remarkable. It tells you that the genetics does not in fact encode a hardwired system that always moves the same way...
AW: Yeah, way more flexible.
ML: Instead, way more flexible. What it gives you is a system that A, knows what a correct frog face looks like, and B, has this kind of error minimization loop where it will continue to move around until it gets to a, a particular pattern. And obviously, you can overwhelm it, but that's what it does.
AW: Wow. So many questions. So, first of all, how do you scramble up the, the tadpole features? And then the second question is, what is this error minimization routine that happens?
09:57
ML: Yes, so, how do we scramble the face? Basically, one of the main sides of research that we have in our lab has to do with the understanding of bioelectric communication and how cell networks use electricity to make decisions about what they're going to build. It's, it's very similar to what happens in the brain when neural networks make decisions, store memories, and so on. And so, it turns out, this is my colleague Danny Adams and I discovered this some years back, that actually the major organs of the face in the developing frog, and it turns out other animals as well, are determined by the positions of specific bioelectrical pre-patterns. We call this the electric face. So, literally, you can look down on the surface of a developing frog embryo face, before any of the genes come on that pattern the eyes and the mouth and all of that, and if you use something that we developed called voltage sensitive fluorescent dye technology, which basically just tells you where all the voltage gradients are, you can see where the different organs are going to be formed. This, this electrical pre-pattern is a scaffold for what's going to happen. So, what we found out was that this electrical pre-pattern doesn't just tell you where things are going to be, but it's actually instructive. So, if you move it around, and by move it around I don't mean move the cells, I mean change the electrical distribution across the, across the nascent ectoderm, what you will be able to do is cause these organs to form elsewhere. So, one of our early applications was to take the normal eye-inducing electrical signal, put it somewhere else, for example on the gut, and make a complete eye on the tadpole's gut. And so, when you move these patterns, the cells, that, that pattern is a map that the cells use to know where to build stuff, and if the, if the map is off, the cells will build.
AW: And, and, is it as simple as, you know, this, this sort of voltage difference across a cell membrane specifies an eye, and some other voltage level specifies something else? Is it, is it that simple?
ML: We, that, that's a hypothesis we toyed with in the beginning. Turns out no, it's not that simple, and it, it's actually, the bioelectric code is not a single cell property. And if you think about it, it has to be that way, because an eye has all kinds of internal tissues, it has complex structures, there's simply not enough information in a single cell...
AW: So, it's electric coupling among cells that are going to make these things.
ML: Correct. It's, it's very much like the way neural networks store information by, by virtue of their electrical communication. It's a network property of many cells, it's not a single cell. And in the face, it's rather simple, which is why this is kind of an early example that, early success that we had, where the pre-pattern looks very much like a face. I mean, you can literally see where the eyes and the mouth are going to go. But that's, they're not all like that. Some are incredibly complicated, the same way that when you look at the electrical activity of the brain of somebody who's thinking about a cat, you don't see a little picture of a cat in there on the MRI, right? You have to, you have to do a lot of deconvolving of the data to figure out how the mapping works. And this is the same thing. In some cases, it's simple, in many cases it's not.
12:56
MM: Huh. So, can I ask you to take us back in history about where the sort of patterns used to be thought to exist, when, you know, people sort of, you know, it's in the genes in some way. Was there a specific articulation of, of where that information resided? Or was it just sort of the argument, it must be there because it's not obvious where else it would be?
ML: Yeah. It's, it's pretty much, it's pretty much what you just said, because there's this, there's this belief that everything is in genetics, and so, in some sense it has to be there. In fact, I gave a talk to a bunch of 9-year-olds at a, at a school once and I showed them an egg and I asked them what determines whether it's a dinosaur, a bird, or a snake that comes out of this egg? And everybody, even the 9-year-olds immediately said, it's the genes, you know, it's the genes, it's the genome, right? So, in theory, everybody knows this, but in reality of course, when you look at the genome, and most people know this but we don't really deal with the consequences of it, if you actually read the genome, you don't see anything in there directly about symmetry type, or organ number, or size, or eyes, or any of that stuff. What you see is a parts list. What you see is a specification of the proteins. Now, you could say that, well, that's the job of developmental biology to figure out how you get from a parts list to the anatomy, but actually a lot of people skip that part, and a lot of discussions of evolution for example, when they talk about gene frequencies and then traits that get selected, there's a whole part in between there that is very still poorly understood of how you get to specific anatomical structures, and especially this business of knowing, knowing how to rebuild structures in novel circumstances is, is very poorly understood, and this is something I hope we get into, because our new work on the, on the synthetic living machines and also the reprogramming of the worm body plan, all of these are telling us that there's a lot we don't understand about actually the input of the genome into the anatomy.
[instrumental]
14:58
AW: Let's just sort of continue on talking about bioelectricity for a little bit. We've, we want to hear about your experiments with planaria, these worms that have the amazing ability to regenerate parts. So, as I understand it, if you cut a planaria in half, then the tail will regrow a head, and the head will regrow a tail, and, and you can cut it into much smaller pieces, and it will regenerate often the entire planaria from even a very small piece. So, this is astonishing. I think a lot of people know this at some level, but the mechanisms that underlie it are just, just blew me away. And so, maybe talk about that phenomenon and the roles of bioelectricity in it.
ML: Sure. And in fact, there are two, two aspects of this that are even more astonishing that people hardly ever talk about. The first is that if, if you weren't amazed enough, here's a couple more things. So, the first thing is that if you think about the way we reproduce... So, so, if most animals, due to a thing called Weismann's Barrier, if we get a mutation in our body during our lifetime, that mutation does not propagate to our offspring, okay? So, so we have this sexual reproduction that basically cleans up the gene, you know, the genome from, from animal to animal. Now, planaria, at least the species we work with, most of the time, they don't do that. They can do it, but it's rare. What they mostly do is rip themselves in half and then regenerate. So, so...
AW: So, this is a typical way of reproducing.
ML: Correct, correct. The typical mode of reproduction is, is the worm, the back-end grabs onto the bottom of the dish, the front end takes off, the middle of...
AW: Oh, so they literally just rip themselves apart?
ML: They rip themselves in half, which makes us feel less bad about doing regeneration experiments because that's what they normally do anyway. And then, and then they regenerate, so now you've got two worms. But the amazing thing about this if you think about it is that they have somatic inheritance. It means that any mutation that doesn't kill the, the neoblast, the stem cell in which it occurs, propagates to the next generation. So, for over 400 million years, these guys have been accumulating every mutation that happens to hit their bodies. Now, we see evidence of this. Their genomes are an incredible mess. They're in fact, mixoploid, which means that you can't even really ask how many chromosomes they have, because every cell might have a different number of chromosomes. They have, we don't even have a proper genomic assembly for these things, because being mixoploid, it's not terribly obvious what you're sequencing when you try to sequence these things. So, think about that. They have, they have this incredibly battered genome, but 100% perfect anatomical fidelity. Every time you cut them, with 100% efficiency, every single piece is going to give rise to a perfectly scaled, sized, and patterned planarian. Now, what is that telling us about, right? How little we understand about the relationship between anatomy and genome, when the genome can be, can be mutated to that level and the anatomy is rock solid, right? That, that's one thing. Another thing to think about is people talk a lot about gradients and, and we'll probably talk about this later on, but gradients in planaria and how, how a given fragment knows what's, which end is head and tail. But that makes a lot of sense when you sort of cut out a middle third, which is what the pictures in the textbooks always show you is kind of the middle third, and there's a gradient from one side to the other. Think about a single cut. So, you make, you bisect the animal. You make a single cut. The cells on one side are going to grow a tail. The cells on the other side are going, are going to grow a head.
AW: And this means that they were right next to each other, right?
ML: Correct. Bingo. Direct neighbors, same exact value of positional information, and yet, they go to have completely opposite anatomical endpoints. So, what that tells you is that you cannot make this decision, if you're a cell and you're trying to decide what you're going to build, you cannot make this decision locally. You cannot decide from your position what you're going to be. You have to talk to the rest of the tissue. And this became the logic behind our experiment that started back in 2005 to understand, okay, how does a head, a fragment know where the head and tail goes, or in fact how many heads and tails it's supposed to have? And it was all based on this idea that it cannot be local, it had to be global communication where the cells at the wound have to talk to the other tissues and ask, well, who's over there? Do we have a head, do we have a tail, which way are we facing, what are we doing? So, so, so, these are global decisions. And the way we got into this is that we were already studying from our earlier work on left-right asymmetry, electrical communication via these synapse, electrical synapses known as gap junctions. So, these are just little, you can think of them as, as little, almost docking ports or submarine hatches where, where two cells can line them up, and they dock, and they allow small molecules to go, or current in fact, to go from one cell to another. So, we asked the question of could planarian fragments be using this kind of electrical communication system to determine how many heads they were supposed to have? And so, we did some experiments and long story short, we found out that that seems to be the case, and that what we could do is if we prevent the cells from talking to each other electrically after we cut the fragments, they end up making two heads. So, not forked, you know, not forked two heads the way you sometimes see on snakes, but actually a kind of pull me, push you, animal that has heads on both ends. And so, so, that was sort of the first set of experiments we did in 2005. We then actually characterized some of the bioelectrics of this circuit and around 2011, we were able to show that, this is the work of Wendy Beane who is currently at the University of Western Michigan, she, working in my lab, we figured out that we can basically determine whether or not you get a head or a tail at any particular region by controlling the voltage, that, that the worm was using a kind of voltage gradient to determine where the head and the tail was going to go, but this is reprogrammable. If we change the voltage gradient, the cells are happy enough to build whatever, and it's certainly dovetailed nicely with, with other work that was going on in the lab at the same time.
20:47
AW: And, and just tell me again, how do you manipulate the voltage gradients?
ML: So, so, in planaria, so, planaria have a particular, another weird feature which may or may not relate to their other weird features like regeneration and mortality and so on. And this weird feature is that you cannot knock in foreign transgenes. There is no misexpression, so there are no GFP flood planaria, there are no, you know, there's nothing like that. So, in planaria, what we're left with is using ion channel targeting drugs to turn on and off the native ion channel that, that create the planarian electric map. In frog and in other models that we work with, it's much easier because we can actually put in new ion channels. We can put in optogenetic channels, whatever. In planaria we have to use drugs.
MM: Mike, in nature, are there sort of natural gradients, voltage gradients that these organisms have to cope with? Is this something that could've played a role in their ecology or evolution?
ML: That's a great question. Basically, yes and no. So, so, externally there really aren't any gradients in the environment, we don't think. However, recent, recent data that we've been working on suggests that some parasites, and in particular some microbiota, so, so, like a bacteria that live on these things have keyed into the system and actually produce compounds that trigger the hosts' ion channels to alter the body in ways that are beneficial to the, to the bacteria.
AW: Wow. Exploiting the bioelectricity, yeah.
ML: Yeah, absolutely yes. Certainly, there can be, but the majority of this is really internal. It comes from, evolution, basically long story short, evolution figured out early on that electricity is a super convenient way to process information. I mean, it's not an accident that brains do this, it's not an accident that all of our computer technology is based on this. Evolution discovered really early that electrical networks are incredibly convenient for memory, for decision making, for distributive processing. So, all of these are endogenous mechanisms that have been around since the time of bacterial biofilms.
AW: Let me ask, actually a related question about environmental influences on phenotypes. And I can imagine some environmental conditions altering these voltage gradients in ways that lead to plasticity and body shape. For example, a planaria at one temperature versus at another temperature, that might affect something about the leakiness of the membrane, or the way the voltage gated ion channels work that set up the, the bioelectric fields. So, so, is there known plasticity that originates via bioelectricity?
ML: Yeah. So, so, there's a couple of ways to think about this. On the one hand, these, these electric circuits have evolved to be really robust. So, every organism evolutionarily has been selected for the ability to keep doing its thing, despite the fact that the K levels are going up and down, right?
AW: Perturbed all the time, right.
ML: It's perturbed all the time. So, they're, they're extremely stable to these things. On the other hand, the plasticity is, is there to be exploited, as I said, by other creatures, but the plasticity also has a positive side. So, one of the papers we put out a few years ago showed that with a very simple, very brief perturbation of the electrical network, you can take worms that have a normal planarian genome and cause them to regenerate heads that belong to other species of planarian. So, their head shape, their brain shape, and the distribution of stem cells now become very similar to other species, about 150 million years distance. So, this is no doubt, I mean, we don't know how widely this has spread through evolution, but this is no doubt a potential mechanism for evolutionary change, because it means that, you know, you can, you can basically, the large-scale morphology is so modular that the organism itself can, can really use this as, during kind of the increases of evolvability and adaptation to new conditions, and so on. I mean, you can, you can radically change your body shape with a very simple trigger. So, overall, we think that one of the reasons that bioelectrics is so important is because it provides a modular organization. It means that with relatively little effort, you can swap out major aspects of the body plan, like create a whole eye, or you know, whatever. So, so, that I'm sure, that has implications for evolvability.
24:56
MM: I mean, how different are species of planaria? The body form, is it relatively consistent, to the behavior, the ecology...how diverse are we talking?
ML: Yeah. So, so we specifically looked at planaria that have very different head shapes, so they could be round, they could be flat, they could be triangular, they could be, you know, all different, all different head shapes. The brain shape is quite different Behavior is interesting. We haven't published it, but we have a bunch of unpublished work trying to ascertain whether these worms in fact take on the behavior of these other species, and you know, I can't say too much because it's, we don't have a published peer-review paper on this, but that's certainly a very important, important topic. There's a lot of diversity. We've done similar things in vertebrates. So, so, at one point we made tadpoles with tails that look like zebrafish tails, and in general it brings up this issue of what's a defect and what's speciation, right? So, so, a birth defect in one sense is a speciation event in another sense, because it might be a perfectly reasonable morphology, especially if you look at, you know, craniofacial shapes, right? You know, what, what's considered a birth defect from the perspective of one standard species might simply be a very workable organism in a different context, so, that raises, that raises those kinds of interesting questions.
[instrumental]
26:21
MM: Can we circle back a little bit to, you know, what you had said about bioelectricity and especially its relationship to information, how it's so useful, how evolution figured out a long time before we did its utility for moving around information? Maybe speak a little bit about how specifically it encodes information about the body shape.
ML: Sure, sure, yep, and then I can, the rest of the planaria story sort of gives you a good example of how this actually plays out. So, if you think about what an ion channel is. An ion channel is, at least some ion channels, are voltage gated current conductants. So, a voltage gated ion channel is a protein that allows different kinds of voltage depending, different kinds of current depending on the cell voltage. So, that's basically a transistor. That means that the history, the past electrical state of the cell, determines the future ion flow through the cell. So, that's immediately a kind of memory, and it immediately enables feedback loops, both positive and negative. So, this is, this is, from computer science we know that if you have this kind of computational element, you can immediately make [logic gates?], and you can make very complex networks with really interesting self-organizing behavior and computational processes. So, bioelectrics really enables computation. It enables distributed computation, because it means you can very easily connect networks into, cells into networks. It also, you know, bioelectricity is a really great example of a, of a central node in a kind of bowtie architecture. So, there's been a lot of talk in biology about bowtie architectures where a lot of things feed in, and then a lot of things come out, but there's one sort of functional node in the middle. Bioelectrics is great for that because if you think about it, you can get to a certain voltage by lots of different ion channels, right? It's not only one, and so, so, it's not a 1:1 relationship between the voltage and any particular gene. This becomes super important in a lot of our work. And conversely, any given voltage can activate a bunch of other, let's say, transcriptional events, other cell behaviors downstream. So, voltage is, is a really nice kind of control hub for a lot of these things, and evolution has used this really to a large extent. And if you want, I can, I can talk about how this plays out in planaria.
28:33
AW: But before you do that, could you just give an example of what you mean by computation? So, so, how would a cell network use bioelectricity to do a computation?
ML: Sure. So, so, one example, and we just, we just put out a, a kind of a theory paper explaining this, is you might want a cell network to be able to determine whether the pattern of the body or of a certain organ is correct or not. So, this is a pattern recognition task. Now, I'll back up by saying that the question of what exactly is a computation is something that people argue about vigorously. I've been to conferences where we spend all day arguing what really is a computation.
AW: Sounds exciting/distressing.
ML: It really was. We, we at one point had a guy who was actually a computer scientist who was arguing that there was no such thing as a computation, which was really interesting. But anyway, so, so you could imagine a cell network that as input, receives information from the spatial organization of some other organ, and then its job is to decide whether that pattern is correct or not. So, am I, do I have the right number of fingers, or are my fingers the right length, or, you know, is this kidney shaped correctly, whatever, and the cells have to decide that, and then they will have to issue an output that is either do nothing because it's fine, or grow some more in this particular pattern because you're not correct. So, that is an example of a computation, and that kind of a thing can be nicely handled by an electric circuit. Another example is a very, a much simpler example, happens with normalization. So, if I take a cell, a small group of cells in the body, and depolarize them, one of the things that they're going to do is, this is basically an early form of neoplastic transformation, that's how cancer begins. And what the neighboring cells do is with their gap, their active gap junctions, they constantly try to override that. They basically sort of try to bully outlier cells into having a generally correct overall voltage. And this is their attempt to maintain a coherent sheet of cells and not let any of them sort of go off on their own and revert to a cancer-like phenotype. So, that, that is a, that is a very simple kind of example of how bioelectrics tries to maintain global state over single cell state.
AW: So, so are you saying that, that cancer involves, in part, cells disconnecting themselves from local bioelectric networks and sort of assuming their own electrical phenotypes? Is that...
ML: Yes. Yeah, exactly. And so, if you think about it, you know, the question we've done a lot of work on this in cancer. The question isn't really why do we get cancer, the question is why is there anything but cancer? Because what you're, well, because think, think about how, and I've, I've got some movies that I usually show at talks of these amazing single cell organisms and all of the stuff that they do, and it's quite clear that individual cells are very competent, and they handle them, their own little local goals, with their physiological, anatomical, structural, and behavioral goals, all in one cell. No brain, no other cells needed.
AW: So, why don't they all do that all the time, right?
ML: Correct. The question is, what could possibly possess a group of these cells to get together and form, you know, work towards a much larger structure, and we have some thoughts on why that is, but long, you know, long story short, what happens is by, by assembling into these electrical networks, they are able to take on larger scale goals. Morphogenetic set points instead of single cell homeostatic set points like temperature and you know, food level, and so on. So, by taking on these goals, they, they form large-scale patterns that basically keep a lot of cells quiescent in the body, and this, but this process can break down, and when it breaks down, what happens to a cell that's become electrically isolated from its neighbors, it basically decides that the rest of the body is just environment, you know? At this point, it's on its own, it reverts back to its unicellular self, and then as somebody said, the dream of every amoeba is to become two amoebas, so, at that point, it, you know, I forget who said it, but at that point, it, it, you know, these cells basically treat the rest of the body as outside world, so that, that boundary, you know, that boundary of the self shrinks from, from the boundary of the whole organism to the, to the surface of the single cell. And they just, they go where they want, they eat what they want, they metastasize, you know, that's... And we've done a lot of work on this.
AW: So, so, it strikes me that, that these same issues might arise in the context of the evolution of multicellularity from unicellular ancestors, and so, so, are there sort of, you know, bioelectric kinds of switches that you can imagine that might facilitate multicellularity?
32:49
ML: Yeah, yeah, you should, you should come work in our lab, that's, that's a great idea, we've...
AW: I'd love to.
ML: Yeah, yeah, that's a great idea. We, we are going to try it. We're working now with William Ratcliff at Georgia Tech, and this is exactly one of the things we're trying to do is kickstart multicellularity using, by, by providing cells that are still unicellular with a little bit of novel bioelectric machinery, and I think, I think it's extremely reasonable to expect something like that to happen, and, and we, we've done it, we've done it in a slightly different context where we've shown that A, you can basically trigger metastatic melanoma in a, in a genetically normal, no carcinogens, no ANCA [?] genes, genetically normal tadpole just by deranging the bioelectrics a little bit, but better yet, we've shown that if you do throw a human ANCA gene like KRAS, P53 mutations, into a tadpole, where they normally make tumors, you can actually prevent the tumors if you artificially force those cells into a normal bioelectrical state, despite what the ANCA gene is telling them to do, which is to depolarize and disconnect. And then, and then there's a very impressive tumor suppression effect. So, that's, that's kind of a forced multicellularity on cells that are trying to, you know, defect.
MM: So, I want to, I want to probe that a little bit more, because it, it sort of suggests that whatever, however the bioelectric field is generated, that it is, and you alluded to it earlier, that it is a pretty important thing to maintain within limits. It's not something that can really bounce around a whole lot, otherwise, you know, things just plain won't work. So, so, maybe, is it really that way? But before, before we go too far down that road, can we go back to what you had said just a minute ago about your ideas about what drove the transition from single-celled life to multicellularity? Are you willing to, to sort of share what your thinking is there, or is this still a work in progress?
ML: No, no, I can, I can share. I mean, we're obviously doing more work on it, but the first paper on this just came out in, in one of the Frontiers journals, and it's kind of this very wild paper on the origin of the self, and what in fact is a self, and where you put the boundary of the self, because I think that's critical. You know, it's not a philosophical issue, it's critical in both multicellularity and cancer. And so, we can, I can tell the following story, and whether or not this story is correct will only be seen through experiment and so we'll find out, but I'll give you the hypothesis of this story. There's this, there's this, these ideas largely emerging in, currently from, from someone named Karl Friston in the UK that have to do with active inference and surprise minimization. So, the idea is that all living systems in order to survive try to minimize surprise. Basically, they try to build an internal model of the world so that when things happen, they expect it. They know what to expect. They infer patterns from their life experience. So, if you think about a single cell trying to minimize surprise, one of the least surprising things around is a copy of yourself. If you would like to not be surprised by your environment, one thing you might want to do is surround yourself with copies of yourself so that you have a much better idea of what's going to happen. And so, one of the drivers of multicellularity might be this kind of, this balance between surprise minimization and infotaxis [spelling?], infotaxis being the desire for more information, for more measurements that will refine your model of the world. So, so, we have, and I do this work with somebody named Chris Fields. We have a bunch of papers on this where we, we look at these dynamics and we ask, how do cells join up in order to handle their information tasks? So, you can see this is a little bit of a different approach than a lot of people that might take this on a, on a kind of, focused on specific molecules, or focused on maybe game theory and competition and things like this. I'm really interested in the information-computation aspects, and what are the informational forces that drive cells towards multicellularity, what are they measuring, and what are they minimizing that enables them to most effectively work together? So, I think that is, and now, I'm also not saying there's only one, obviously we know there are multiple origins of multicellularity. There could be lots of forces at work, but I think this is an important and under-appreciated thing, that it's really about what measurements cells are making and what set points are they trying to maintain through some sort of homeostasis?
36:54
AW: So, you mentioned the frog example and planaria, but, but do you think that, that bioelectric fields play a fundamental role in organizing, you know, development, regeneration, and wound healing in all animals? I mean, is this, is this a general phenomenon, or are you cherry-picking, you know, some of the species that do this in especially profound ways?
ML: No, I think, I think it's extremely general, and I think it's very, very ancient. I think it was discovered at the time of bacterial biofilms as the work of Arthur Prindle shows, and we know it exists in humans as well because of the human channelopathy. So, so there are long, long lists now of human channelopathies, which are ion channel mutations that have specific morphogenetic defects, and, and it would be just completely bizarre if this is something that evolution only utilized in, in certain cases, especially given the high level of conservation of both, both functional and molecular steps. Now, that's not to say that there might not be animals that have reduced their dependence on this signaling system. So, for example, not much work has been done in C. elegans on this, and I could easily imagine that something as mosaic as C. elegans might have decided that, evolutionarily decided that bioelectrics is just not crucial for its body plan, you know, some sort of chemical cell-cell interaction and [word?] may be enough to, for what it needs to do, and so on. I used to think that about Drosophila as well, although now, as of the last couple of years, there have been some really nice papers finding a new role of bioelectrics in the fly, so, so I can't even cite that as an example anymore. So, it may be universal, or there may be species that don't use it.
MM: So, how do you put that together with, I mean, the fact that, you know, in humans for the most part, amputations are bad news for us, expect when we're young, you know, lose very distal tips of our fingers and such, maybe we can regenerate. But regeneration is not, I mean, that is a component of what we're talking about, and that's an advantage that's been lost by most at least large organisms. Why? I mean, is it, does it become a complexity issue when you reach a certain size threshold?
38:54
ML: Right. I don't believe it's a complexity issue, and I think, you know, just to preface this by saying that bioelectrics does a lot more than regeneration, so, so, for example regulative development, the fact that you can cut early embryos into pieces and each one will give you a normal embryo, and this is how we get monozygotic twins, and so on. I mean, so, there's a lot more than adult regeneration at play here. But, but so, so, let's talk about human regeneration. Nobody really knows why humans are less regenerative than other animals, although people oftentimes think mammals are not regenerative, but we can talk about deer, where they regenerate antlers at a rate of a centimeter and a half of new bone per day when this is happening, so, so highly regenerative adult mammal. But I can tell you a story about why I think humans are, are not regenerative, but it's just a story, we don't, we don't have proof of this.
AW: Well, we love stories. Stories are good
ML: Well, I'll tell you the story. So, the story might go something like this. So, imagine that you are a mammalian ancestor. You're something like a mouse, you're running around in the forest, and somebody bites your leg off. So, unlike a salamander, the problem you have is much more important than what happens over the next two months if you regenerate, is immediate. You might bleed out, you might get an infection, and by the way, when you try to regenerate that delicate blastema, you're going to be grinding it into the forest floor once you start putting weight on it. And so, it may be that what mammals have decided to do is to basically go to scarring instead of, instead of regeneration, that the important thing right now for survival is to seal the wound, scar it, and close it off so that you don't die right away, where that regeneration is just not going to happen anyway, so you're better off scarring it. It, it might be telling that the deer, the one example, good example we have, the, the antlers are not load-bearing, you know, you don't have to put weight on it. Maybe, you know... So, again, this is, this is pure, pure speculation on my part, but I think that's not an unreasonable... Now there's another component to this which is that most good regenerators are aquatic, and one thing that might be involved here is that in dry air, it's extremely difficult to drive the kind of electric currents out of the wound epithelium that you need that, that salamanders and other animals have no problem doing. I mean, regeneration is actually extremely broadly sprinkled. In fact, one of my most favorite examples is that even single cells regenerate. So, there are people who work on [word?] and single cell organisms. I mean, it's unbelievable. You can, you can go online and see some of these, some of these videos where you take one of these highly patterned single cells, you cut it in half, and you would think, or at least I naively thought that well, all the cytoplasm is just going to flow out and that'll be the end. So, it comes out a little bit and then it goes whoop, it comes back, it gets sucked back in, and, and the thing will actually regenerate, and this has been known for a really long time. And then it has to solve all of the same problems of patterning and knowing when to stop and knowing what's missing, but it does it at a single cell level. So, I think there is an amazing and still poorly understood scale-free aspect to this, that it's not just large cell networks that do this.
[instrumental]
42:02
AW: Mike, so I want to switch now and talk a little bit about a non-bioelectric thing, or at least I think is non-bioelectric, and that's your recent work on vector transport of morphogens along nerve cells, and the roles that those play in patterning development regeneration, and sort of where, where body parts are. And I'm thinking specifically of this 2019 paper by Alexis Pietak. So, so, can you just talk a little bit about what, what you mean by vector transport, and what's getting transported, and what are those morphogens doing?
ML: Yeah, so, so, the big picture here was to try and understand how patterning information spread across the body, but one of the things that we wanted to, to, our model to reproduce is something that we've discovered that's basically pattern memory, and the idea is that we, we found in around 2009, 2010, that if you take a two-headed flatworm with a perfectly normal genome and you chop off the primary head, you chop off this crazy induced secondary head so all those tissues are gone, you have a nice normal gut fragment in the middle. That thing will still regenerate two heads even though it has a wild type genome. And, and in fact in perpetuity, it will keep doing that, as far as we can tell, forever. So, this at the time was kind of an amazing finding because it's a new kind of epigenetics, beyond the single cell level. It suggests that some information is in fact not in the genome, and you can, you can reprogram it. We then more recently discovered a way to set them back, so we can take a two-headed worm and change it back to permanently being a one-headed worm.
AW: You sort of erase the memory of the...
ML: Correct, correct, erase the two heads, yep. But basically, we wanted to understand this process of, of memory, so why do two-headed animals give rise to two-headed animals after cutting? So, we wanted to make a model that recapitulates most of the known experiments in the field and does so quantitatively, in other words, no magic. Every single step is specified to the point where you could just...
AW: A no magic model.
ML: No magic, which is surprisingly rare, I mean.
AW: So, so, you can predict novel cuts and how those will result in, in new planaria, and the model gets that, gets that right?
ML: Exactly. And in fact, some very surprising reversals of polarity for some unusual cuts that hadn't been done, and so on. So, that's, that's pretty cool, and I think kind of unique in the field. And so, so, so Alexis, Alexis did this, she worked with a very talented post-doc in my lab Johanna Bischof, who together, they, they then were able to show that a proper model that explains all of these previous results has to do with not only diffusion, but actually has this facilitated vector transport that we have a lot of data about the nerves, but actually my gut feeling is that this is not exclusively about nerves. I think it's more about cell polarity, like planar cell polarity that's probably present in a lot of the tissues.
AW: So, so, two quick follow-up questions. One is just the mechanical details of, of what do you mean by vector transport? So, what, what's doing the transport and what's getting transported, first of all?
ML: Yeah. Those are great questions. So, so, we think what's doing the transport is the cytoskeleton. So, we think that basically, I mean we know that in neurons there's this, and in fact in, in polarized cells, there are cyto-, microtubule tracks that are used by motor proteins and so on. So, we think it's a motor, and the reason that it's active is because it's not just diffusion, it takes energy, and it takes this kind of motor protein...
AW: Yeah. So, these are things like dynein that are sort of crawling along the cytoskeletal elements and carrying packages of morphogens which they then distribute, yeah? Okay got it.
ML: Correct, exactly. Exactly, exactly. In the end, what's redistributed are well-known factors like Wnt and beta-catenin [did I spell these correctly?] and so on. Whether those are themselves being moved we don't, we haven't shown that yet. It, it could be that something else is being moved and then that, you know, sort of transduces into, into relocalization of these other things. But the simplest model is that, is that in fact, these, these important morphogens are literally being shuttled along.
AW: Yeah, okay. And then the second thing is, you mentioned that this, this vector transport idea interfaced with bioelectricity. So, what's the connection there?
ML: We discovered, so I'll just tell you how, how this came about. So, if you treat a bunch of planaria with a gap junction blocker, some portion of them will make two heads in the next round and those two heads are permanent. The, for years we've been taking the ones that didn't make two heads, and we've been calling them escapees. Basically, we thought that for some reason they just weren't affected by the reagent, maybe their skin is thicker, who knows, right? And so, we've just, you know, every experiment has some percentage efficacy and we thought, okay well that's just how it is. So, so, a, a very talented student in my lab, she was a PhD student Fallon Durant, she's now a post-doc at Harvard, she did work on asking the following question: are those single-headed animals that arise from these, these cuts, are they in fact escapees? Are they wild type, are they normal? And so, what she did was she recut them in plain water, no more manipulation. And when she cut them in plain water, what she found out was that in fact, they were not wild type, because the same proportion of them became two-headed as originally became two-headed from the octanol treatment. So, it meant that there are, before, we thought there were two kinds of worms. There were normal one-headed worms, and then there were two-headed worms and that's it. There's actually a third kind. The third kind is a really weird phenotype called destabilized or cryptic worms. Cryptic worms, what that means is they look normal, they look one-headed. They only have one head, they only have nervous tissue on the right side, you know, on the anterior end, and so on. They look perfectly normal, but if you cut them, they have a, a solid chance of becoming two-headed. Once they become two-headed, that's it. Those will always be two-headed afterwards, but the ones that don't are still cryptic, and if you cut them again, once again there you have a chance of becoming two-headed. So, there's this kind of, you can imagine drawing, and in fact in the paper we have this state transition diagram that shows you what kind of worm can become what kind of worm. And so, there's this, and so, and so, it turns out that the reason that these cryptic worms can become two-headed is because they have a destabilized bioelectric pattern memory of what they're supposed to do. And there's a couple of different ways to think about it. One paper that we just published thinks about this as overlapping inconsistent memories in the visual system, like the rabbit-duck illusion, you know, that kind of thing, where you can have, you know, an electric network that finds it hard... It has to settle into one or the other, but they're sort of both there, and it's kind of bi-stable, right, that like, visual illusions like that. So, that, that may be part of how it works, but the important thing is that there's this, there's this bioelectric circuit that basically tells the, the, we already knew this, this bioelectric circuit that decides head or tail. But the bioelectric circuit can be put in a state where its bi-stable. It can go one way or the other after injury. And this is really important to me. This is, this is where this really becomes like a memory, because if you think about a one-headed cryptic worm, and you look at the bioelectrics and you see that it's not normal, it's you know, let's say it's got depolarization on both sides even though there's no head on the, on the posterior side. You, what you see right away is that the bioelectrics is not a readout of what the anatomy is right now. It doesn't match what the anatomy is right now, unlike normal worms and two-headed worms, the bioelectrics tells you exactly what the anatomy is now. This is much more subtle. This is a, this is a readout of what the anatomy will be if you injure the thing. It's like a latent memory that's, that's in there but it's not being expressed. So, so, now, now we're to the point now where we can rewrite these memories in any direction, so we can, we can wipe it, we can, we can induce it. And it's, it's rather remarkable that a single planarian body can hold more than one memory of what a proper worm is supposed to look like. And it doesn't act on it until you injure it. That, that thing, that information is sort of latent. So, when people ask where is the memory, we now know where it is, at least this. I mean, no doubt it has other roles, no doubt there are other things about head shape and eye number we, we have, you know, we have to, to do more work on that. But at the very least for the number of heads, you can see where the memory is and you can see that it's a kind of latent memory that doesn't get consulted until the animal needs to, needs to repair itself. And so, so, the future, the important future goal of this research is to merge the two models. So, we have a really good understanding from, from Johanna and Alexis' work of how the two-head memory is propagated, okay, and sort of this, this permanent two-head state. And now, we have from, from Fallon's work and from modeling work that we've done with Chris Fields, we have a pretty good understanding of how the electric circuit works. The electric circuit by the way functions within three hours of regeneration. It is the earliest known step in planarian regeneration. All the stuff that happens after that are the things like, you know, Wnt and beta-catenin signaling, and, and [word?] and all these things, they happen downstream. So, the electric circuit makes a decision quite quite rapidly, and then sometimes these things get cannulized into a long-term memory that is implemented via neural directionality and this kind of long-range transport. But in a short term it's an electric, it's almost like a short, I mean, one is tempted to think of long, short-term electric memory and long-term chemical memory as in the brain. But that's the deal. And so, now what we need to do is merge those two things into one complete model that, that has both in it, and, and that's, that's work ongoing right now, we're in the process of putting that together.
[instrumental]
51:28
MM: Wow. I have about a thousand questions to, to carry on with this, this line of thinking, but in the interest of time, let's try to zoom out a bunch and what do you think about the evolutionary implications of such things? I think the skeptics would say planaria are just freaks, everything that you're learning in these, these guys, it doesn't really transfer, they're all great stories, but, you know, doesn't, doesn't work for plants, doesn't work for elephants... I mean, what's your, what's your thought about the implications of what you're doing in bioelectricity specifically for the way that we're thinking about evolution now?
ML: Yeah, I think, I think that's factually wrong. I mean, people, people have shown this kind of stuff does work in plants. I don't know that anybody's used an elephant, but for human, you know, for human medicine, these channelopathies are really, really important. And for example, Anderson Tawil syndrome was, was known to be a Kir 2.1 mutation, and these patients had cardiac syndromes and craniofacial dysmorphias. And why they had cardiac arrythmias was pretty obvious, it's a K channel expressed in the heart, fine, but nobody had any clue why they had craniofacial dysmorphias until, again, Danny Adams and I figured this out, and we published a paper on, on this explaining how this works. And so, so again, don't know about elephants, but from bacteria to man it seems to be a constant factor, and I think it's very clear now from the work on basal cognition and the evolution of nervous systems where brains got their tricks. I mean, brains basically speed optimize dynamics that were happening in cells and tissues long before neurons showed up, and what happened is that bodies used to make electrical networks that thought about the configuration of the body in sort of [phrase?] development, and later this became adapted to thinking about the configuration of the body in three-dimensional space during running around and you know, trying to escape predators and so on. So, this is a, this is an ancient system that is absolutely conserved from, from, from bacteria to man, and we are now using this thing for biomedical approaches. So, we have work in our lab looking at limb regeneration in rodents, we're looking at cancer suppression in human tumors, we're looking at ways to repair birth defects in mice. So, this is absolutely not a worm or a frog-specific thing. This is meant to be medicine.
53:42
ML: And in the, and in the sort of more, I mean I'm going to challenge you to, to put the work that you're doing in the traditional context, you know, really because this is something that comes up on so many episodes, the modern synthesis and evolutionary biology, and relative roles of genes versus other forces in, in shaping the phenotype and influencing evolution. I mean, I think that the argument, that the point that you just made about the timing at which the bioelectric field is really driving change versus, you know, the action that the developmental, expression of developmental genes. I mean, what do you think is a sort of promising step in the merger of the types of work you're doing in the traditional evolutionary thinking, or just generally where we are in evolutionary biology now?
ML: Yeah, yeah. So, I heard two different questions there. So, so, one question is where does, do bioelectric controls sit with respect to the known signaling factors, the gene regulatory networks and everything else, like during the life scale of an animal? So there, while it is true that bioelectric states turn on and off all of the important genes that we know about, we have frzls and Wnts and BMPs and all of that, it is also true that bioelectric states are themselves produced by ion channels which are encoded by the genome. So, I think more usefully than trying to ask what's on top, because, because it's a whole cycle, right? And so, so, the physiology and the, and the genetics sort of cycle back and forth. I think a better question which is often asked by network scientists these days is where are the best control nodes, and by control nodes I mean where does the least amount of influence or input give you the most complex control of what happens? And I think for, for my money, bioelectrics is an incredibly convenient set of control nodes, because with very simple inputs, we've been able to make eyes and trigger limb regeneration and reprogram tumors and fix brain defects and all of these other things, which suggests with, with very simple controls, that what we're finding here is the natural modular structure of development. We're finding the subroutines in the software of pattern formation. And this is to me, again, now, so, my background is computers and so everything looks like this to me, but I think it's rather, I think it's rather apropos that this is very much a reverse engineering task. We've given these, these animals and plants which have amazing plasticity, our goal is to understand the software which in part means to identify the hooks, the, the control points. You know, you could try to assemble a hand from stem cell derivatives and try to pattern the growth factors everywhere they go, but you know, how many, how many hundreds of years is it going to be before we can do that? It's much nicer if we understood the causal architecture of the system itself and said okay, this is how the system understands where to build the arm and what an arm is, and here's the simple signal that it's going to cause it to do that. And so, so, that's, so that's one. The other issue you brought up was kind of the evolutionary context. So, I think a good way to think about this is via the distinction of software and hardware. And a lot of people go crazy when I say this and it is not because I think living things are computers in the way that we have, you know, computers right now that you and I are using right now. That is not what I mean. What I mean is that there's a fundamental and deep insight of computer science that if your hardware is good enough, you are much better off programming it by inputs and by experiences, by signals, not by rewiring. In the '40s, when to reprogram a computer in the '40s, you literally had to move wires around. And now, if I told you that in order to switch from Photoshop to Microsoft Word you had to get your sautering iron out and get in there and start moving...you know, you'd say that's crazy. But why is it crazy? Because the computer hardware is good enough that it is reprogrammable, and I make the argument that biology is absolutely reprogrammable. Evolution discovered this very early on. And so, what we're better off with is this idea of, of thinking about the genome is not the software of the cell by any means. The genome is what specifies the hardware. The genome is what tells every cell what proteins it gets to have. But once you've made that hardware, much like any of the good digital hardware that we make, it has a default behavior which is what it'll normally do when you turn the juice on, but it is also reprogrammable, meaning it, by, by [word?] choice of inputs and signals and experiences that it has, it will...
AW: Upgrading your app.
57:57
ML: Exactly. It will move to other, you know, it has lots of capabilities that it will go to other modes of, of being besides the, the default that normally just happens when you, when you turn on the power. So, this is, and this is the kind of thing that's seen, I don't know if you wanted to talk about this at all, but our latest paper on the, on the xenobots, this is exactly kind of where this work is going is to take these cells, perfectly normal genome, no genomic editing, and to ask what can we reboot you into making with appropriate signals? Are you happy enough to make something completely different that's not in our case frog-like, and the answer is absolutely. And this is, this is I think, this is I think the right way to think about this. I think evolution targets the hardware, but the hardware is amazing and it, it, it encodes systems that have massive plasticity that actually help evolvability, and we could talk about this for a while. I'm actually writing a piece right now with Rafael [what is the rest of this name?], Dan Dennett, and David Higg [spelling?] about the use of this kind of plasticity to really make things evolvable and modular.
AW: Awesome. Well, super quick comment. I can tell you've been collaborating with Paul Davies, because we had a conversation with him about six weeks ago in which we talked about this sort of software-hardware distinction, and he made some, some, you know, very parallel points. It was really great. But, yeah. I do in the last ten minutes or so want to talk about your, your, yet another mind-blowing paper that you published recently, this one in PNAS on xenobots, and if I can just sort of paraphrase how I understand it. So, you used stem cells from Xenopus frogs, and you, you put them together in particular ways that were specified by a sort of computational algorithm that you designed to find structures that would be the most functional in, in the environments in which they were operating. And then you built these things, and they worked, right? So, tell me, tell me how wrong I am about, about that description of xenobots.
ML: Sure, sure. Yeah, it's not too wrong. I, you know, the big picture of course was, was two-fold. We really wanted to understand plasticity of cells, and we wanted to understand what they could be willing to build when you liberate them from the constraints and the boundary conditions of a typical frog embryo. And the next, and the, and the sort of bigger utility of it beyond making little frog robots to do specific things is to really understand how to motivate cells to build things in regenerative medicine, but also to adapt these kind, lessons learned into robotics and, and, you know, communication systems and so on. So, what we did, and this is not so much about stem cells because the frog, the cells that we took were basically destined to become, to become skin and muscle, and this is, this is work that we did very closely with Josh Bongard's lab at University of Vermont, and also it should be noted that this, this particular, the stuff reported in this paper is just the tip of the iceberg. There's lots more, lots more there, so some of the conclusions I'm making rest not just on what's physically in this paper, but some of the other things we've seen that, that you'll be seeing later this year, hopefully. So, Doug Blackiston who's a staff scientist in our lab, very talented developmental biologist and microsurgeon, what he was able to do is basically take some, take, take embryos, take some cells, take some cells out of them, mix them up and put them in the little, in the little hole, and let the cells decide overnight what they're going to do. And the cells, probably for reasons that I described before of, of you know, there is sort of desire to build multicellular bodies that, that, that minimize surprise and all of that, they get together and they form functional novel creatures which we call xenobots for Xenopus laevis, that's the name of the frog, and biobots for this kind of idea of making synthetic living machines.
AW: And, and these cells get together just sort of, of their own volition, I mean, just by the fact of being close together? You don't have to chemically treat them to make them do that?
ML: Absolutely not, they get together on their own. We, we can provide, in this paper what we, what we show is simply sculpting them, so after they get together we sort of cut out different, different regions to make the whole thing look a certain way, and then, but actually we can give them other, other signals to make them do various things, and Sam Creedman [spelling?] who's a grad student with Josh at UVM, what he did was to make a computational platform that evolves in a virtual world, evolves designs for robots to specific specifications. Let's say you want it to walk or something like that. And so, then what we had was a very nice back and forth that was basically a cycle of they evolve a design that they say will, will walk a certain way, Doug actually builds it, we see, we characterize its behavior, we tell them what happened, they refine their model, they give us back a new design, and so, and it goes back and forth, and the goal is two-fold. One is to understand the rules. So, what we really want to know is how do we determine what is going to happen and what shapes are these things going to make, and two is to make a computational pipeline that enables us to make things that we're actually interested in for specific applications. And it's kind of amazing the, you know, so, so, I think one of the most amazing things about these things besides the plasticity of, of these cells is that unlike most organisms on Earth, well I guess all other organisms on Earth, these do not have a biological evolutionary history. This, the individual cells do, but they were selected for the ability to sit quietly on the surface of a, of a tadpole and keep the bacteria out. They were not selected for the ability to run around and work together and, and do all of the things that they do, whereas instead they have a digital evolutionary history. They have an evolutionary history that occurred in a virtual world. They were, they are the first living machines designed basically by a computer. And so, their evolutionary history did not occur on, on Earth in a, in a, in a, you know, in a biological niche, they occurred in, in a digital simulation, and I think that's a, that's a totally novel scenario.
AW: So, so, what do you envision as the sort of long-term utility of, of xenobots? Like, you know, if we were to imagine a pipeline that is, is producing things that are, you know, has some economic or medical use, what, what are they going to do?
ML: Yeah. I think we, we can certainly come up with short-term uses for specifically xenobots. So, so, you can imagine things like they could go into the waterway and collect microplastic or report on toxins, or be used in a dish to microsculpt some, something you're going to transplant into the body, you know, some, some tiny thing. Maybe they go in the arteries and scrape the plaque off the walls, or maybe they hunt down cancer cells. We, we can come up with these kinds of things. I think the much bigger picture here is, is not that we're going to necessarily be surrounded by robots made of frog cells, you know, taking care of all of our needs, I think, I think the bigger picture here is that once we understand how small, highly capable goal-seeking units arrange into larger structures that do specific things, this will have an impact in regenerative medicine, of, of you know, regeneration of body organs and so on. It will have impact on new robotics, on communication networks, [word?] of things... I think that's the big picture. The big picture is really to understand how small things scale up into large goals. I think that's, that's the big picture. But in the meantime, there are plenty of really useful things that I think these, these types of robotics can do, and will certainly be trying to, trying to make them.
MM: So, I think the last thing Art, unless you have any more questions, Mike, the last thing that we always give our guests the chance for is what else would you like to say? Have we left off anything that you'd really want to make sure you communicate to our listeners?
ML: No, I think the, the, the only, the only thing I would say is that you know, my lab focuses on bioelectricity. That is not because I think other things are unimportant. Certainly, biomechanics, chemical, you know, genetic networks and all of that are important, and I think the magic here is not bioelectricity per se, it's the fact that bioelectricity really enables evolution to exploit the laws of computation. So, I think what evolution does is it exploits physical forces to do very interesting kinds of computation, and that is really the future of biology. I don't think we're going to be able to micromanage a lot of the outcomes that we want, let's say in regenerative medicine. I think it's much better to understand how the system regulates itself, and these questions of unification of selfhood, of computation, are not, they're no longer philosophy. They are, they're here with us in biology, and now these new techniques from, from synthetic biology, and better yet synthetic morphology are allowing us to actually start to construct some of these things and watch them happen in front of us, and I think, I think that's, that's super exciting and I'm just, you know, thrilled.
[instrumental]
1:06:48
MM: Mike's work spans disciplines. Before he was a biologist, he was a software engineer and owned his own software company. That background gives him a unique perspective in biology, and maybe even unburdens him of the baggage that many of we traditionally trained biologists bring to our own research.
AW: To Mike, biological problems often look like computer problems, and he argues that we should think more about genomes as readable and writable software, not hardwired programs that alone determine the form and functions of life.
[instrumental]
AW: Thanks for listening to this episode of Big Biology. We hope you're staying healthy during this coronavirus pandemic, and we hope that this episode provided a healthy dose of distraction from never-ending coronavirus coverage.
MM: We've got more regular Big Biology episodes coming up this spring, but we're also trying to bring you a few episodes about COVID-19. If you have questions about the virus, or any other Big Biology topic you'd like to hear about, send your ideas to info@bigbiology.org.
AW: And please don't forget that we're a volunteer operation. If you'd like to become a patron, go to patreon.com/bigbiology, or go to our website www.bigbiology.org to make a one-time donation. Or, you could just spread the word about the show over your social media channels and give us a rating on iTunes.
MM: On the next regular episode of Big Biology, we talk with Ellen Ketterson, a biologist at the University of Indiana who studies how organisms adapt to changing environments.
EK: With a bird like the junco, and really anything else in eastern North America, it, where it lives now is covered by ice, and that was only 15,000 years ago. So, how did that bird manage to occupy regions down on the gulf coast during the last glacial maximum, and how did its biology enable it to move northward and reoccupy all of the regions that were ice covered? And then what can we extrapolate from that to try to predict how rapidly the junco and other organisms can respond to environmental change now? They don't have 15,000 years. They have a whole lot less than 15,000 years, but whatever those mechanisms were will be put to the test to see whether they can keep up.
[instrumental]
AW: Thanks to Matt Blois for producing this episode. Michael Levin manages our social media accounts and produces the Student Spotlights, and Dana Baxter helps with background research. As always, Steve Lane manages the website.
MM: Thank you to the College of Public Health at the University of South Florida, and the College of Humanities and Sciences at the University of Montana for support.
AW: Music on the episode is from Poddington Bear.