Michael Levin: Bioelectricity, Aging, and the Future of Medicine

Yeah, happy to be here. How are you?

Yeah. So, you know, there's a couple of different aspects to this aging thing. We have to think about what the fundamental cause is and then also what are the downstream effects. So there's a ton of downstream effects that we can talk about, but one of the real fundamental questions is what causes it in the first place. And the standard theories are basically in two camps. It's either physics, which is like an accumulation of errors and over time, you know, entropic noise, or it's something evolutionary/ecological, that is for some reason it's beneficial to have us dead at a certain point, right? And so we've been working on both ends of it. So on the one hand we have a new theory of the origin of aging which places the problem neither at the physics end nor the biology end but actually at the cognition end, and we could talk about that. So it's basically a, you know, we call it the boredom theory of aging. That's one way to think about it.

And then on the downstream side, yeah, I think one of the things that happens is a kind of blurring of the information that's normally needed to upkeep a complex three-dimensional organism, because materials are always coming and going, cells are always coming and going, and you can't really maintain the ship of Theseus that is our body if you don't know where everything is supposed to go in that pattern, which we've been studying, it's at least partially bioelectrical, and that pattern we know gets more diffuse over time.

Both are possible. These are not mutually exclusive. What we do know, here's what we know, and of course there's a lot more that needs to be done, but the first thing we know is that we've looked at mortal and immortal hydra. So hydra that ages versus the ones that don't, a strain that doesn't, which is most of them. And we definitely see differences. There are definitely biological differences that come up when the hydra is aging. We've also done it in human cells, in vitro. So again, as cells senesce and die in vitro, there are bioelectrical changes. However, we also saw that when these human cells do senesce, then they get older, if you try to shift them using drugs, you try to shift them to a different voltage, they don't shift as well. So like they're more resistant. So it may be both. It may be that yes, you're losing some bioelectric information, but also the cells are becoming worse at acting on it, and so they're just not as responsive to various cues that cause them to acquire bioelectric states in the first place. So they might be losing plasticity or something like that. Again, you know, that's in vitro data. So all of this has to be done in vivo at some point.

Well, first of all, pushing the system into a different stable attractor is a pretty good description of memory. So that's not a bad description of how memory works at all. And so in Hopfield models and other kinds of models, that's exactly what, and we study memories in gene regulatory networks that work exactly the same way, that they're basically specific past experiences chase you into a region of the state space where you respond differently to future stimuli. I mean, that's basically a kind of memory.

But the other thing about the planaria, which is I think a deeper issue, is the entire question of is there a memory of a goal state? Because when you early on, you take your first biology class and they tell you about salamanders regenerating their limbs or embryos, you know, cut in half and they give the resulting complete body and the twins and so on. And the first thing that students always ask is how does it know? How does it know what to build? How does it know when to stop? Right? And what you're told immediately is don't ask that question. That's not a good question. We're going to do nothing here knows anything. We're going to explain all this stuff in terms of chemistry, which is specifically a formal model of objects that don't know anything and don't have any goals. That's what nobody tells you at that time is that this is not some sort of a fundamental result. This is an axiom. It's an assumption.

Now, you could sort of imagine at the start of when science was first getting off the ground that this was a pretty important assumption to do because the alternative was that people would just say, well, you know, God wanted it this way and that's why it does it. So that doesn't really help. And so we said, okay, we're going to try to get away from that and see how far we can get by making models where there is no master plan. Let's see how well we do. So that's great and that was useful at the time. But the thing is that back then you really only had two options. You could be dumb like rocks or you could be smart like humans and angels and that was it. Those were your two options. But since the 1940s we've had a good science of all the stuff in between and that's cybernetics and of course behavior science. So you don't have to be completely stupid and have no goals and know nothing in order to not have to be a human or above-level self-reflective intelligence. There are many things in between starting with thermostats and everything else. So this idea that nothing in the system knows anything is an axiom and I think it's very important to lose that axiom and actually do experiments instead of making assumptions about what other levels of the cybernetic hierarchy are appropriate. In other words, I think it's really limiting to assume you're going to do all of biology by only using the lowest level of that hierarchy where nothing knows anything and nothing has any goals.

So the question becomes how do you know if these things actually have a pattern memory to which they're going to build? Because the standard story of biology in these kinds of cases is not that. It's emergence and complexity theory. The idea is that everything is a feed-forward process. Everything follows local rules and eventually complex things happen and we call it emergence because you get some kind of complex outcome. So the reason I say that the planaria actually have a pattern memory is the way you know that there's a goal state, it's a homeostat, the way you know that it has a goal state is you have to be able to find the mechanism by which you store the set point because every goal-directed mechanism has to have a set point that is the memory, so it has to know what the goal is and then you have to be able to read and interpret that goal state, so you have to decode it and you have to be able to rewrite it and show that when you rewrite it the system executes the new goal. Yeah, those are the criteria. So that's exactly what we did in cases of the planaria and other cases.

That's what happened. So we showed that actually there's a biological pattern in the body that stores the number of heads. Apparently, it also stores the shape of the head, but we haven't decoded that very well yet. But we have decoded the number of heads and you can see it. You can see the pattern memory. You can decode it and you can rewrite it. That is how we make the two-headed worms is that we open and close ion channels and gap junctions to rewrite that pattern. And when you rewrite that pattern and turn it into a two-headed pattern, that's exactly what the cells build. So that's it. And furthermore, you know, it's not only a goal-driven process, but it actually is a long-term memory because when you take the two-headed worms and you cut them into pieces, you continue to get two-headed worms. It's permanent once you've, I mean, you could set it back if you want to. We can rewrite them back into a one-headed state. But as long as the bioelectric pattern is not changed by us, the tissue holds it. You know, it's stored. So it has all the features of memory. It's labile, meaning you can change it, but it's stable, right? And it has discrete outcome. So there are your attractors. I mean, reasonable enough.

Well, what it looks like is, and the broader effort is called cracking the bioelectric code. So the idea is that what we want to do in the general case is understand when you're looking at a particular kind of bioelectric pattern, what will be the interpretation of cells. When cells see that pattern, what are they going to build? Because we need to understand this so that when we want to build new things, whether it be to repair birth defects or to build healthy new organs for regenerative medicine or some kind of bioengineering, maybe you want to make a useful living machine that's never existed before or something, you want to know, okay, so what kind of pattern maps, so it's the mapping between the biological pattern and the resulting form and function. So what that looks like is taking lots of measurements of biological patterns, large-scale bioelectrical patterns under health and disease conditions and different species, different model systems, and then making hypotheses about what those are saying to the cells and then testing those hypotheses by making targeted changes in the bioelectrics and looking to see if you correctly predicted what the resulting anatomy is going to be and then building a computer model.

So this is where we are now. At the beginning when we first started doing this, so back in '98 or so when I first started doing this, we were still doing screens. We had no idea what we were, like I knew something interesting was going to happen. Everybody else thought it was nuts, but what I didn't have is a model where I could say this is how we're going to make a leg or an eye or a heart or anything like that. So we were just doing screens to see whether this even worked. And now we're to the point where we have computational platforms that predict specific drugs that repair very complex birth defects of the brain, the face, the heart, the gut, and you know, we're much better along now. We still, there's a million things that we don't know and there's still an incredible amount of work left to be done to acquire the kind of data sets in physiology that we have for omics. So proteomics, transcriptomics, genomics, all those things are much easier because you can do all that on dead material. So you can take your cell and fractionate it and kill it and sequence all the stuff that's in it. So that's great. But bioelectricity doesn't work like that. The minute you kill a cell, it's gone. All that information is gone. It's the software that disappears when the hardware stops running. So that's been much harder. Now we have somewhat better tools and now that information is flowing, but still we need way more.

Sure. Well, I mean the first thing is that we don't all become worse. Most of us become worse, but planaria don't and hydra don't and there may be other creatures that don't, but so we know it's possible not to succumb to that. So here's the boredom theory of aging. Let's, I like to start thinking about it this way. Imagine that your pet snake and your dog go to heaven, you know, like the Christian version of heaven where it's going to be around forever. There's no decay, there's no entropic processes, there's no bacteria, there's no decay, nothing is falling apart. So you're going to be there forever. So what happens? Well, the snake probably is going to be fine doing snake things for a trillion years and beyond. Probably fine. The dog, I don't know. We could assume maybe that if it's a nice environment and it gets to chase rabbits on the farm all day, then maybe every day is the same as every other day and it's fine. The human though, I don't know. It seems to me that you'll keep yourself busy for the first 10,000 years or whatever. But eventually, like a billion years in, are you sane? Is it possible for a high-level intelligence to, and if you think you can keep busy for a billion years, fine, let's talk about a trillion years or whatever, right? It seems reasonable to me that at some point a cognitive system like that is going to fall apart. Now what's interesting about it is that if that's true, then what's the problem? Organic disease, there's nothing physically wrong with your brain and we already know there's such a thing as the thought that breaks the thinker in a way, like you know, if you're certain high-IQ individuals, you know, they come across, I don't know, Hume's problem of induction or something and like, oh my god, it just sends them into a spiral of despair and then they get depressed and then the depression markers come up in their brain and all the physical stuff gets going, but the trigger for that wasn't any kind of disease process, it wasn't physics and it wasn't biology, it was thought itself, right? And so we already know that for a complex cognitive system, there doesn't have to be a physical issue. It doesn't have to be biological or physics. You can have other kinds of disease conditions that are purely at that level, that are purely cognitive in their origin.

So we asked the question, what happens to a goal-seeking system after it's completed its goal? If it's really simple, so your thermostat has only one loop and as long as the hardware isn't broken, it presumably will keep doing that forever, right? No problem. But if you're a little bit more complex than that, if you're a kind of collective intelligence that has multiple components that are working together, and my story of how collective intelligences are held together is from this 2019 paper that basically says that collective systems are held together by goals. So they are upheld to the degree that the collective can represent a goal that's compelling to all the parts. Basically the whole has to align all of the parts to do things that they otherwise wouldn't do and it's the overall goal that holds it together. So if you're a little bit more complex than that, what happens if you complete your goal and then you don't have a new goal?

So we had a simulation where we had individual cells working together and the collective had a goal. It had a goal to build a frog face. This is the work of Ben Hardle and Leo Pio-Lopez in my group. So we were simulating development exactly the way that I suggested that it works, which is it's a collective intelligence operating in morphospace and it was fine. And you look, there's this kind of curve over time and you see that, okay, so here's the developmental process and the thing begins to look like a frog face and eventually it makes a really good frog face. And then there's a period of adulthood and it's sort of flat and it looks like it upkeeps the frog face for a while and then the whole thing starts to fall apart. And at first, we were looking at it like, why is this thing suddenly, like nothing new has happened? We haven't, so here's the important thing, in our model there is no entropic decay and there is no evolutionary pressure to fall apart, none of those things are there. So where is this coming from? And what we noticed is that after a while, we gave it a goal during development of forming a good structure, but we didn't tell it what to do after that. And this is very different from theories. There are theories that say that the genes that are important early on in development end up being deleterious later on. They end up being harmful later on. This isn't that. There are no harmful components here. It's not that the things you need early eventually take you apart. It's that you have a compelling goal early on, but then there isn't a new one. And so what happens after a while is that all of the individual cells, the thing disbands basically, the cells stop trying to make a frog face and the whole thing falls apart. So that's really interesting in the absence of selection forces or noise. This demonstrates that there's a goal-seeking system that eventually will disband for no reason other than that the goal has stopped being compelling.

Now we since then got some interesting data from, we analyzed human tissues from different age subjects and we looked at, we used a technique called phyllostratigraphy which basically looks at the gene expression and these are public data that anybody can analyze. Gene expression in tissues in the body from young and old patients and phyllostratigraphy basically tries to estimate the evolutionary age of each gene. So when during evolution, not the organismal age, but evolutionary age. So each gene, roughly when during our evolutionary history did this gene first show up, you know. And so what we found is that when you're young, all of your tissues and all of the cells in your tissues are expressing genes mostly from the position that we have on the evolutionary tree, which is modern Homo sapiens. And all of the cells agree that that's what you are and that's the genes you should be expressing. But in some, not all, but in some tissues in older patients, first of all, the phylogenetic age of these genes starts to go backwards. So they start expressing genes that are way back in evolutionary history and they stop agreeing with each other. So your cells in many, many tissues, your cells simply don't agree anymore on where you are on the tree of life. So again it's this dissociation and we called it atistic dissociation because it's basically they're dissociating from each other. They don't have a shared vision of what they were.

And if you think about it, it's really a mirror of development because in development you look at an embryo and there's like, I don't know, a million cells at a certain point and you look at that and you say, oh, there's an embryo, one embryo. What is there one of? Like you're looking at a million cells. What exactly is there one of? What there's one of in that case is a commitment. All the cells have a shared vision of where in anatomical morphospace they're going to go. They have the same set point. They all agree. They've all been aligned to agree that this is, and so they all have to get aligned and so they make an embryo. And then eventually they no longer are compelled by this shared memory of what it is that they're supposed to be doing and they all fall apart. And you can see this morphologically in our model and you can see this transcriptionally in this data from human tissues.

Yeah. I mean, yes, I think that both cancer and aging are fundamentally diseases of multicellularity. More sort of taking one step further, I think they're diseases of collective intelligence. If you're going to be a collective intelligence, you're susceptible to certain failure modes, ways that those processes go wrong. So what keeps you together as a collective intelligence is a kind of mind meld where all your components have to be sharing information in a way that doesn't let them have too much individuality. You know, everybody has kind of the same vision of what's going on. And so there's that connection and then there's the informational linkage between the pieces and some other stuff, stress sharing and some other things, and yeah, there are several ways for that to go wrong and one way is what you see in cancer and another way is what you see in aging. But I think the process of continued regeneration, which I think all of this is regeneration, so development is just a kind of regeneration from one cell. That's what I think development is. It's just regenerating the whole body from one cell. All of those kind of processes. So metamorphosis, regulative development, actual regeneration from damage, all of those things are self-reinforcing processes that resist aging and cancer for as long as they can.

Yeah, it's a good question. I don't know what it is, of course. People work on it, so not us in particular, but there are other labs that work on this. Yes, I think it's a super important point that babies are not born at the age of their parents, they start off at zero. So that's really important and I really think it speaks against this entropic theory of aging, you know, because you would expect errors to accumulate. We do have to keep in mind that the germplasm isn't exactly unchanged. I mean, the process of evolution is telling you that over time, you know, you certainly don't have the same germplasm as you did a billion years ago. So it does change over time, but it's not quite the same as aging. So whether you could keep it exactly the same, I'm actually not sure. One of the interesting questions is whether without evolutionary change, whether the germplasm would fizzle out as well.

The reason I say that is because what we found in our model is that, so okay, take a step back. So I've told you the story of the boredom theory of aging. Why can planaria apparently live forever? You know, what happens to them? Why isn't their organism bored with their body plan? Partially, I think it's because every two weeks or so, they give themselves this massive challenge. They rip themselves in half and they regenerate, right? So we actually tested this in our model and we showed that yeah, if you induce damage and ask the system to regenerate, that puts off aging, you know. So basically like refreshing the goal, refreshing the commitment to the goal sort of puts off the decay process. So maybe in the long term, maybe evolution plays that role for the germplasm in the long term. I don't know. This is just off the top of my head. I'm thinking now out loud, but it's possible that evolution plays a similar role on a much longer time scale, but yeah, I don't know if we'd be able to roll it back that way.

That's my prediction. So yeah, Dave and I have talked, we've talked long ago as well about doing some experiments to kind of look at the interactions between the system that he sees and ours. I don't know which one is sort of upstream of the other. My strong suspicion is it's the bioelectrics and that's simply because that's in the whatever 25, 30 years I've been doing this, every time we look it's turned out that there's a bioelectric switch upstream of the downstream molecular biology. So that would be my bet if I had to bet right now, that's what I would say, but we don't know and it's entirely possible that it's the other way around, but I suspect that's what it is.

Yeah, we'll find out. We do know that bioelectric change can trigger, for example, can regulate HDAC activity and can regulate certain kinds of chromatin modification. So I wouldn't be surprised if we find that kind of epistasis.

Yeah, we're doing it. We're doing it now. And actually I know at least one other person in a different lab who's doing it too. Basically, if what's happening during aging is that the bioelectric pattern is becoming more blurry, it's becoming less distinct, then one thing you might imagine doing is sharpening it, trying to maintain the pattern more crisp and sharp and the boundaries between tissue compartments is sharper. So the interesting thing is that we have about, I don't know, five or so papers showing how to do that in a birth defects context. In other words, there are birth defects that are caused by a blurry bioelectric pattern. And if you artificially strengthen that pattern, then despite like really nasty mutations like Notch mutations which are crazy strong for brain defects and face defects and so on, if you

You sharpen the pattern, then you have normal morphogenesis, normal behavior, and so on. So we are applying that same strategy to aging, yes, and this is in collaboration with Mitch McVey in the biology department here at Tufts, who also studies aging. He studies it in fruit flies. So we've made some fruit flies that are bioelectrically altered, and we're studying their lifespan. So stay tuned. I don't have a full answer yet, but I'm very hopeful. And October 7, I think, or the 6th or maybe the 6th, I'm supposed to be at an aging conference at the Harvard Medical School, and I'm going to hopefully show some of this data by then. We should have some answers by then. Yeah.