Conversation with Buck Trible on his work on static allometry of ant castes and emergent cognition

This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.

[00:00] Buck Trible: Thanks so much for inviting me to talk. It's so fun. I really appreciate it.

[00:04] Michael Levin: It sounds like you're doing some very interesting work. I'd love to hear about it and then brainstorm on it with you.

[00:11] Buck Trible: Absolutely. So some background is to say that I've been studying ants for pretty much my whole career. And mostly from the perspective of molecular genetics. My undergrad advisor pictured there found these genetic elements that can switch fire ant colonies from having a single queen to a few hundred queens in them. This was a big story in the late '80s and early '90s, the first example of the genetic regulation of social behavior. And I've been continuing on that thread, trying to identify natural genetic polymorphisms and then use them to understand basic biological questions. I think of it as a natural mutagenesis screen, the same way that if you have a screen where you see a certain phenotype, you know that you've hit the pathway you're interested in. We're trying to do that with wild populations. We look for certain natural phenotypes that we think have to result from modification to some underlying process.

[01:35] Michael Levin: Can I ask a question about the phenotype that you just described with 100 queens? Is it a modification of ant behavior that allows more queens to develop? Or is it actually pushing specific animals to become queens and the rest of the colony—what's going on?

[01:56] Buck Trible: It's a profoundly beautiful discovery and it's very, very deep. It's one of these things where it's really a whole syndrome of features and it's more like any trait that you measure is a fingerprint or an indicator of which state of the syndrome you are, as opposed to a description of the actual phenomenon. The syndrome is that there are these two forms of colonies. One of them will have one queen in the colony, and one of them will have a few hundred queens. The colonies with a few hundred queens also lack territoriality, so they don't form discrete colonies with discrete territories, but instead they form carpets of ants with colonies all next to each other. They generate a very large number of queens, but they also accept new queens back into their colony. One of the big discoveries of this was confirmation of the green beard effect predicted by Hamilton and then later Richard Dawkins: that you could have selfish genetic elements that act by perceiving themselves and other individuals and then killing or benefiting them. This was the first demonstration of such a thing existing in nature. They found that these queens that form these multiple queen colonies all have a genetic element that induces the trait and their worker daughters have some kind of altered behavior, some kind of altered pheromone perception where if a queen has the super gene, she'll accept it, but if she lacks the super gene, she'll kill it. So all of the hundreds of queens in the colonies have to carry the genetic element that induces it. Since then, I had this joke where we were studying ant genetics in the late 2000s, early 2010s. At the time, there were about 10 people in the world doing this. Ken was one of the first, and I was trained under him. I had this joke that the people studying ant genetics had collectively discovered fewer Mendelian elements of any consequence than Gregor Mendel found by himself in his one pea plant, in his one greenhouse. It's really a new field to do this type of molecular genetics with ants. There are two reasons for it. One is technical challenges, but I think an equally important reason is that it has not been the historical focus of people studying ants. Since then, now that we have whole genome resequencing, people are mapping these genetic polymorphisms all over the place and it's growing exponentially. The second one, the one that Ken had, was the only one for decades. The second one was found in, I think, 2012. By 2020, we had six. This year, we'll probably have 20. We have tons and tons of genetic information now about different types of polymorphisms in any species.

[05:31] Michael Levin: And in terms of how many colonies do you need to look at to be able to find this? Because a lot of times in these screens, you would have to go through a lot of animals to actually see it. Is this just more prevalent or are you guys better at detecting it? How does that work?

[05:58] Buck Trible: I would make the hesitant claim that most ant species lack the types of polymorphisms I'm referring to. That may be wrong. Maybe they mostly have them, but I think most of them appear to be monomorphic. But the key kind of observation is a gestalt or an intuition that you go into a population and within a freely interbreeding population, you have two discrete forms. You can even do high school level Mendelian statistics where if it's a continuous variation in the thing that determines them, you should have a unimodal distribution. Whereas if it's a Mendelian trait, you'll have a bimodal distribution with peaks on the Mendelian ratios. So you can get a vibe for if it looks like a genetic polymorphism. And even without doing those stats, by being in the field and looking at the ants, a lot of people can get that intuition. For years there was nothing you could do with that. And so it would be described in the literature as an interesting phenotypic polymorphism. It would be described as a polymorphism and assumed to be environmentally induced or induced by quantitative genetic variation. And you would call it "facultative." That was the prevailing assumption for decades. Ken was the one who questioned that assumption in his fire ants that were thought to be either polygenic or environmental and therefore facultative. He felt it seems discrete to me. He did all these allozyme markers and found that one of them perfectly correlated with the trait before they even had microsatellites. These days you find those two forms, sequence the whole genome, and look to see if there's a genetic element that consistently distinguishes the two. This is the background. A lot of these Mendelian elements—one thing worth knowing is that there's a double-edged sword to these things, which is that on the one hand, mapping them is incredible because you have a genetic element that's necessary and sufficient for a phenotype of interest. So far, every one of them that's been mapped has turned out not to be a SNP or a simple small genetic locus. They've all turned out to be what we call a social chromosome or a supergene. Though they're very large. You'll often have a chromosome with large inversions and a movement of a centromere. The chromosome will have an X-like form that's freely recombining and normal, but then it'll have a Y-like form. It has massively enriched transposons, repressed recombination, and is all screwed up. And that'll be the one that induces the alternative trait. And so it's almost like these discrete forms, a single versus multiple queen form of these fire ants. It's almost a sex in a sense. It's two alternative reproductive strategies, two alternative ways of being in the world, and it's not conferred by a single mutation, but it's a whole suite of things.

[09:47] Michael Levin: And in general, what kinds of genes tend to be involved? Are these neurotransmitters? Are they hormones? Are they structural elements, ion channels? What kinds of things?

[10:04] Buck Trible: We've never figured out the causality of any of them. They all have hundreds of genes in them. The work that I've been doing is to take one of these super genes and do laboratory evo-devo and really try to map the causal mutation or combination of mutations. Even for sex chromosomes, there were some key genes, the sex differentiation factor or the sex determination factor SRY, that were known. The rest of the sex chromosome, even those, were a black box until recent CRISPR methods in the last decade or so. This is a hard problem even in model systems. That's the background of this field that we're working in. It's very exciting. All kinds of crazy stuff is getting found every year. The one that I've become particularly enamored with is actually a mutation that arose in a clonal species. We have a species we keep in the lab called the clonal raider ant. My PhD was based around making the first CRISPR lines of ants in order to show how ants perceive pheromones, to test that they use these odorant receptors to perceive pheromones. This is a clonal species, but unbeknownst to us at the time, its ancestor probably had one of these supergene polymorphisms in the population. That clonal genome is a diploid female who just lays eggs that directly give rise to new diploid females. In that diploid clonal genome, one of the chromosomes looks like a sex chromosome. It has these two highly distinct forms. We found a mutant that has a large loss of heterozygosity along this chromosome. I often draw it like this. You go from a heteromorphic chromosome with two very different forms to a loss of heterozygosity, where in an interval of the chromosome you don't have the heterozygosity but have two copies of one form. This mutant does something terribly interesting: it switched the organism from a free-living state to an obligate parasite of its own species. Normally the small-bodied ants will develop into workers, and the large-bodied ants will develop into queens. In this mutant, the small-bodied larvae have the same size, developmental time, and external characteristics — they are completely indistinguishable. At the onset of metamorphosis, they have overgrowth of all their tissues and they differentiate into this parasitic queen phenotype. They don't resemble normal morphological queens. We showed that they're very different from queens of related species, but they aberrantly express queen-like traits in small-bodied larvae that would otherwise become workers.

[13:56] Michael Levin: Thinking about the morphogenesis for a second, are they going to a different—how different morphologically are they? Are they really quite different or is it just a different path to the same queen-like morphology?

[14:12] Buck Trible: Can I show you some slides? Projects completed. Can you see this?

[14:34] Michael Levin: Yeah.

Buck Trible: That's the wild type of this species. This is often: the evolution of clonality in ants is accompanied by the loss of the morphological queen caste, because the queen is the primary reproductive of an ant colony. If you can clone yourself, you can have workers that directly give rise to new workers and you don't need a queen anymore. We're working with the clade that ancestrally looked like this, which is typical for ancestors: a worker and a winged queen. This is what the species was like when we found it. This is the mutant that originated. That's the picture of them. This gives you a picture of what it looks like. Here are two close relatives of this species. I traced them in Illustrator and stretched and compressed them to make them aligned to each other. The queens are larger than the workers of these other two species. In our mutant we have an aberrant queen-like form that's broadly overlapping in body size with the wild type, which would resemble the worker of the other ants. This shows that the workers are smaller than the queens of these other two species, whereas the mutants and the wild types are broadly overlapping. This shows that if you do a principal components analysis with two principal components, the first one captures size, and the second one brings in this caste information. It pulls the wild type of our species with the workers of these other ants, and it pulls the mutant of our species with the queens of these other ants. This phenotype is really amazing, and it's the first intraspecific polymorphism, or what we call a workerless social parasite. These workerless social parasites are known to emerge directly from free-living ant species all over the phylogeny of the ants. I was talking to my colleague last week, and he said he has at least 83 independent evolutionary origins of this trait, which is a lot.

[18:11] Buck Trible: Most of the independent origins only have one species associated with them. Some of them have radiated into clades with a handful of species, but they're evolutionarily young, a dead end strategy. In every case, except for this one, all of those origins are a species. It's a distinct evolutionary lineage with sexually reproducing organisms with no interbreeding between them. In this case, we're going in a single step within a nominal species from the one to the other. We think that's because this supergene probably conferred this as an intraspecific polymorphism in the sexual ancestor of the organism we're studying. We can't prove that, but that's the most parsimonious explanation. It gives a really nice picture about how these parasites evolve because it's been a big mystery for years and years; Darwin, the earliest ant natural historian William Morton Wheeler, and E.O. Wilson all extensively discussed the mystery of the evolution of these ant parasites, because they are obligate parasites of their host, and their host is often their closest relative. It looks like they have to speciate; they have to become a new species out from inside another species without ever having an isolated population and any obvious way of preventing gene flow between the two forms. Now we know about the supergenes, and there's an easy explanation: if the supergene confers these alternate phenotypes on a single chromosome, then there's no recombination of that genetic element that's inducing the alternative trait, and so you can get divergence of two forms inside a single population and then you can attach a gene that can cause reproductive isolation onto that social chromosome a million years later, if you want to, and speciate it, or you could retain it as an intraspecific polymorphism. That's what this paper was: a new form of natural history variation and a plausible explanation for this mystery about where these parasites come from. The real thing that I'm interested in about this mutant and the reason why I reached out to you is that the whole body is differentiating as if it were a large-bodied individual, an individual who's a small-bodied individual. It's a radical decoupling of the relationship between tissue growth and body size that occurs normally in ants. We have some other research showing that this relationship between caste and size is extremely deeply conserved. You might even be able to show that there's a single allometric function that relates size and morphology that is invariant across all of the ants that explains the worker-queen distinction, and that this mutation and the evolution of these other parasite species represents a breaking or a secondary loss of that relationship. What I've been trying to do is figure out why the development of ant castes in general shows this pervasive relationship with body size, why larger individuals differentiate in a queen-like manner and smaller individuals differentiate in a worker-like manner across the ants, but also what's broken or has changed in this mutant to cause these small individuals to differentiate in a queen-like way instead.

[21:57] Michael Levin: Do these species build any kind of a nest or a structure?

[22:08] Buck Trible: What are you thinking?

[22:11] Michael Levin: What does that look like? I'm interested in the relationship between morphology and behavior at multiple scales. Their cell behavior makes embryonic morphology, animal behavior. I've been looking at the behavior of making and upkeeping a nest structure in terms of, for example, regenerative response. In some scenarios where you damage a piece of it, they'll go and rebuild it, and the mistakes that are sometimes made in rebuilding it are very similar to morphogenetic errors that cells make when they rebuild the body. You'll see duplications, you'll see misperceptions of distance.

[22:59] Buck Trible: In nest building organisms? you're referring to.

[23:01] Michael Levin: I'm not 100% sure it was ants as opposed to termites. It might have been termites. But that kind of thing.

[23:14] Buck Trible: That's really pretty. So these ants aren't really complex mound building ants. The clonal raider ant, as with most of the ant species that people study in detail, is an invasive species. It's found all over the world from human activity. These invasive species tend to favor disturbed habitats and have relatively simplistic nest architectures. There's a secondary loss of some of the complexity that allows them to grow and spread faster. So the ancestors probably were doing something more sophisticated, but we don't have field studies of them. This one is forming little clumps in the ground. It's a cool point because I agree with you that there's physics papers about these fractal branching architectures and how an ant colony can be similar to a circulatory system. We have this mutant that's doing this really crazy thing. It's a detective story from the genetics to try to figure out the proximate cause. There is a proximate cause: some gene that is differentially expressed that will be responsible for this, or different in activity or a combination of that. There is at some level a genetic event that flips that switch. But answering that kind of proximate question might not be the satisfactory answer to this problem we're looking for. Because at a deeper level there's some kind of system or combination of systems that's coupling tissue growth and body size, and that could involve intelligence or collective behavior. I'm not in any way dogmatic about what it is, but there's always this nagging question: even if we solve the genetics, we might not solve the question we're interested in here. How do you go looking when you have this network and all these connections and layers? How do you look for the part that is necessary and sufficient for this as opposed to something hooked into it? Another way of saying it is: even if we saw the thing in the middle, the magic thing. Sometimes I call it the allometry machine. One of the things we discuss in lab meetings is: what did an allometry machine even look like? If we found one, would we know that we found it? That's what I wanted to discuss with you because I felt the way you're thinking about this, and I emailed you after watching one of your videos online about embodied cognition and hierarchical structures of intelligence. It's very much like what I think these ants are doing. Not from an aesthetic or personal view, but empirically there are reasons to think that some architecture like that is required to explain what we're seeing. But then the question is how do you get from that to the empiricism? What are the predictions you make and how do you try to pull this thing apart?

[27:30] Michael Levin: One thing, I'm going to share this. You've probably seen this. Have you seen these things before? These are Darcy.

[27:42] Buck Trible: Of course, yeah.

[27:43] Michael Levin: So that's interesting to me, this idea that what you're doing to get across these different species is deforming a grid according to mathematical rules. And this question of what the grid is, in real terms, trying to map this onto developmental mechanisms. Nobody has anything like that. I think that this issue of what the machine should look like: a lot of people assume the answer is going to be some kind of set of genes. It's going to be a genetic structure. You can go past that. We've recently published some things on the genome as a generative model. Specifically the idea that the genome is a set of prompts that are interpreted by the layer that sits between the genetics and the phenotype and the behavior. That developmental layer is a very active problem-solving computation. It's not a straight mapping from genotype to phenotype. Specifically, the mapping isn't just complexity, it isn't just emergent feed-forward emergence, but the middle layer is literally intelligent in the sense that it navigates a problem space with certain competencies to solve specific problems and therefore the interpretation of the genome is not fixed. It's from the perspective of an observer, in that case being the developmental machinery. There's that. You could go even further and say that many things in biology, the actual explanation is not to be found in physical space. It's found in some mathematical space. In other words, the actual answer to why this is like that is often a piece of mathematics. Often because that's how prime numbers are distributed, or because NAND is a specific kind of universal gate, or because Feigenbaum's constant is what it is. They're not the kinds of things over which genetics usually spans.

[30:16] Buck Trible: Another way of putting it is that genetics is constrained to work with certain kinds of properties of how systems operate.

[30:28] Michael Levin: A lot of people do think of it that way as constraints, but it's constrained by it, and I flip it in that. I think better than constraints, it's actually, you get a lot of free lunches that way. Because a lot of these facts of mathematics of computation and so on, they give you things that you otherwise would have had to do a lot of hard work evolving. They give you for free certain properties that otherwise you would have to micromanage and get for yourself. And so from that perspective, I think that what evolution is doing is taking advantage of these things. Yes, it's constrained by them, but it doesn't do it justice to think of it as just constraints. I think it's actually an incredibly rich pool of resources.

[31:20] Buck Trible: When I learned about Hox genes in the early 2000s, it was this idea. Maybe this isn't what the people who were studying it thought, but at least learning about it, this is what I was taught: it's like a program, inducing a program to make an eye or a wing. If you induce the wing program in the eye primordium, you can get a wing to try to grow out of an eye or a leg or whatever. My view on it now is that it's more like a place indicator. It's telling you you're in the tissue that makes a wing. Then knowing that you're in the tissue that makes a wing, there are lots of ways to respond to that information to build a wing. It's going to differ for different cells, but it's not really telling it what to do. It's more just telling it where it is and therefore giving it a goal of what to do. In ant colonies, pheromones were again thought to be like verbs, telling you to do something. There are a few examples of that, like a sex pheromone or an alarm pheromone. But the overwhelming majority of pheromones in ant colonies are really labels. They know who's the queen, who's the worker. Within the worker, they know if she's a nurse, a scout, or a forager. They know all the developmental stages of the larvae, their sex, their age, and what their future caste is going to be. It's mostly just a way of labeling everything so that the individual worker can decide how to engage with it more than it is telling them exactly what to do. Another way of putting it I always think is funny is a job: the job description is not very important, but the job title is pretty important. It's important to tell you what's your job title, and then you do what you think you need to do to fill out that role. But the description of the detail of what you should actually be doing on a day-to-day basis is hard to predict or know.

[33:41] Michael Levin: In morphogenesis we see all of this. We see creative problem solving where you have a goal and the system, under different novel scenarios and perturbations, will reach that goal using different molecular mechanisms if it has to. It actually dips into the affordances, the genetically provided affordances that it has and picks what it needs to solve the problem. We have other examples like the Hox thing that you were just talking about. If you induce an eye somewhere else in a tadpole, which, by the way, can't be done with a master eye gene Pax6, it only works in the anterior and ******, but with a particular voltage state you can actually do that anywhere. What happens is that it's a very simple signal. It doesn't remotely say how to build an eye. It's a very simple trigger. It says, this is where the eye goes. Then there's this conversation between the cells that we've injected and the surrounding tissue, because we could inject very few cells. There's not enough of them to make an eye. What they do is they talk to their neighbors and they try to recruit them. Sometimes when you take a section through these eyes, you see that it's only about 10% of the cells — they're the ones that we injected. Everything else was recruited by them to do this. At the same time, all the surrounding cells are telling these cells, no, you're wrong, you should be making skin or muscle or whatever it's going to be, because that's basically a cancer suppression mechanism: for the navy, if you come across an aberrant blue voltage cell, you try to get it to normalize. They have this debate back and forth about what they are going to do. Eventually it settles down one way or the other. Either the whole eye disappears, or conversely, it wins and it recruits a bunch of its neighbors and makes a nice eye that's only partially modified by us. There's this active interpretation. What's central to being a collective intelligence is that you've got parts that are aligned to the same story. That's in these processes: trying to figure out what the decision making is at the level of the subunits, and then, collectively, what they are aligned toward. How do they become aligned to specific outcomes? In development you see this at the very beginning, because when we say there's one embryo, what are we actually counting? What is there one of? There are hundreds of thousands of cells at some point. What you're counting when you say there's one embryo is that they're all aligned to the same morphogenetic goal that they're going to try to reach, even though you're cutting it and pasting it and introducing all kinds of weird stuff. They're all aligned. That's what you're counting.

[36:33] Buck Trible: You're speaking my language. I don't have any issues with that. The idea that the one cell needs to recruit the others, either they do or they don't. I think there's something like that happening during allometric scaling during development as well, where you can have a developing fly and you damage the wing disc, and then it sends out this modified insulin-like peptide that can basically arrest the growth of the rest of the animal and give the wing disc time to catch up.

[37:22] Michael Levin: Oh, that's awesome.

[37:23] Buck Trible: Right.

Michael Levin: I didn't know that example. Could you, when we break, e-mail that to me?

[37:28] Buck Trible: It's a beautiful finding. That's very nice. This is a little bit abstract, but the organ-body scaling is a spectrum. You go from small body size with small tissues to large body size with large tissues. And it's really interesting to me that it's a one-dimensional spectrum across all animals. You take a principal components analysis of individuals of the same sex and genotype and PC1 always explains a lot of the data, 80 or 90% or more, which is weird because there are a lot of tissues, and body size is just one variable. You could have an n-dimensional space. It does not have to be largely one-dimensional. The way that I think about it is almost like a harmony: imagine that you are at a certain body size and you have a wing that's a little bit bigger than it should be, or smaller, it's damaged. The wing's a little bit smaller than it should be for that body size. It has dissonance with the rest of the system and says, "I'm off by this delta." Then it doesn't just fix itself. The whole system also is like, "oh crap, you're off by that delta." Maybe everything moves down a little bit while that one tissue moves up a lot. And then everything is back into balance now. The point is I'm trying to figure out, and again, I think that this stuff is testable. It's an empirical question, and I've got this mutant, but I'm trying to figure out what is the right way to approach this because you are talking from the starting point of having this electrical signal that seems to be doing this stuff. Maybe you'll say, "I think I need to check electrical stuff in this context." But there's an arbitrary number of things I could check, and so it's hard to know. That's the kind of thing I wanted to discuss with you.

[39:51] Michael Levin: The reason I would say to check electrical is not that bioelectricity is magic on its own, but what makes it really, really interesting and important is that it's a very convenient cognitive glue. It's a really convenient way of binding subunits towards common purpose. Evolution caught onto this as far as the brain and nervous systems, and we caught onto this for computer architectures. It's an incredibly convenient piece of biophysics for doing the job of scaling individual subunits into networks that can store goals, preferences, memories, decision-making algorithms better than the individuals can. There are many other ways. Biochemical and biomechanical signals can do some of that. Maybe they can do all of it for all we know, but bioelectricity is a very convenient way to do that. When you're looking for mechanisms that underlie problem solving, collective intelligence, reprogrammability, novelty, bioelectrics is a good place to look for that reason. This is one of the things when I was asking about the genes: I'd be interested in these super genes. How many of these things are ion channels, for example?

[41:15] Buck Trible: Our mutant, it's been this 10-year detective hunt. First, we found this weird chromosome and mapped the mutation to it. We figured out that early development is identical between the genotypes and that the genotypes start to diverge basically at the onset of metamorphosis, when the tissue proliferation is happening. At that time point, we find that there are all these genes involved in steroid hormone signaling that are differentially expressed. It's a really interesting finding because they're genes that are known to be differentially expressed in the context of the key steroid, in fact, the only known steroid hormone of insects, which is called ecdysone, basically the insect version of estrogen. If you agonize the receptor in other systems, or you look at the upstream targets of the receptor, none of those genes are differentially expressed. Even though we're associating with the pathway of this hormone, it's not via the canonical receptor that this hormone signals through. We had this good link that it's probably the hormone. I spent a year or two thinking it would be the receptor and then ruled that out. We find that the most differentially expressed genes at our relevant time point inside the supergene are cytochrome P450 enzymes that are only expressed in the synthesis gland of ecdysone, the steroid hormone synthesis gland called the prothoracic gland. I've got this big model now that the normal function of this hormone would be to signal via its canonical receptor to regulate developmental timing and body size. I think that it might be signaling via some non-canonical receptor to regulate tissue growth. If you build this type of feedback loop like I have here, we can explain a bunch of different stuff that we've observed. I haven't seen anything bioelectrical looking. It gets to this idea that the proximate genetic cause that we're finding could be hooking into some larger system. Again, with your frog example, maybe a mutant Hox gene knockout or Hox gene substitution—there's just no genetic route to getting that electrical perturbation that you got. It might be.

[44:40] Michael Levin: The two-headed planaria line is an example. In planaria there are no stable mutant lines at all. Unlike Drosophila and every other model system, I think we finally understand why that is. It's an amazing thing that there is no aberrant line of planaria except for two lines, and we made them, and they're not genetic at all. There's nothing wrong with their genomes. There's nothing different about them. One has two heads, and one is confused about how many heads it's supposed to have; it's randomly one or two. In both of those cases, if we had put them in the river somewhere and 100 years later some scientists came along and scooped them up, they'd say let's sequence the genomes and see where the speciation event went. That's not where it is. I'm sure that's just the example we've worked up, but there are many others. It could be things like that. I'm curious.

[45:50] Buck Trible: Part of the reason why I think that our queen-like mutant is so interesting is because I can't find a mutant in any model organism that produces this phenotype. A small-bodied fly. There are many mutants that can make a small-bodied fly, but I'm not aware of any where it's a small-bodied fly, but all of the tissues grow as if it were a large-bodied fly, right? In human genetic disorders, there are genetic disorders where you'll have overgrowth of a certain limb, but a systemic thing where your body size is small, but all of your organs grow as if you're big. I don't think that there are any human genetic disorders that look like that, or mouse mutants, or even C. elegans. I don't know how you'd know. I think maybe people have seen mutants like this, and they just didn't analyze it through this way of thinking. So maybe there are mutants like this. That's why I think that this is so interesting: this coupling of size and phenotype is a pervasive thing across the tree of life. When you find a mutant that breaks a phenomenon of interest, that's how you find a circadian clock or that's how you find these fundamental things. I think this could be something like that. This could be hitting something that's really conserved and deeply found across biology, but for whatever reason it's not easy to find it with mutagenesis screens or the traditional approaches people are using.

[47:35] Michael Levin: Yeah, totally.

[47:39] Buck Trible: Yeah.

[47:40] Michael Levin: When you make ants of different shapes, what do the other ants make of this? To what extent is it all based on chemical? Is it visual? Do they know when something's a weird shape and when do they care?

[48:10] Buck Trible: This isn't a direct answer to your question, but related. There's a common phenomenon in ant colonies that's called reproductive policing. For example, a lot of species will only tolerate a single queen in the colony. If another queen tries to start laying eggs, maybe one of her daughters, the workers can smell, and there's decades of chemical ecology work showing that when you become reproductively active, it changes your hydrocarbon profile on the surface of the cuticle. That's perceived as a reproductive pheromone that, in the context of the good queen, coordinates behavior and so forth. But in the context of one of these extra queens, it'll lead to getting killed. It's like this feminist idea that gender is the social meaning of sex, that to some extent, your gender in society is dictated by how society interprets your sex, how you're treated. We have these morphological castes of ants that also have distinct behavioral profiles, right? There's the question about to what extent their distinct behavioral profile is programmed by their morphology, and to what extent their behavioral profile is the social meaning of their morphology, right? It could be that even if you're a worker ant or a soldier ant, some component of your behavior is dictated by your morphology, but some component of your behavior is dictated by the way that other ants in the colony treat you.

[50:10] Michael Levin: Yeah, that's the thing I'm getting at: some sort of a morphogenetic and/or behavioral chimera where if we do introduce strange new phenotypes, let's say you bioengineer an ant that's completely different, what are the responses going to be? We're working on this large review on collective intelligences and hacking of systems to try to infiltrate them, mostly from the biomedical perspective. I think there's a lot of interesting things here in these model systems as to how they detect invaders, how they know what's normal, and what the outcomes are when you have a mix of different things that normally don't go together under evolutionary selection. We've done another review on chimeric organisms. What happens when you put two different sets of parts together that have all made assumptions or not about what their neighbors should be like? And biologically, you can imagine altering these things to have different shapes, but also technologically, if you made artificial robotic ants that did specific things, would they be ejected? Would there be a new emergent mode of cohabitation?

[51:49] Buck Trible: One example of this is that there's a thriving community of ant keepers, ant farms, and hobbyists who do really sophisticated rearing. Some of these people have even figured out how to get ants to mate in captivity with weird combinations of temperature and UV light inspired by studies of reptiles that no biologist has found or published, but they can get them to mate in captivity. Some of these people are really cool. One of the things that a lot of people will do is they'll coerce ants into forming chimeric colonies of two different species. There are videos on YouTube of two very unrelated species that would never mix under normal conditions living together. And to my knowledge, this hasn't been studied scientifically as to what the behavioral consequences of this are. But you might imagine that your chimeric mixture of cells maybe goes one way or the other, but who knows, right? One thing I would say is that most ant colonies are inactive most of the time. And it's not the mental model where every ant is running around and perceiving stimulus and doing stuff continuously is how people would think about it, but it's not really what it looks like. There are a few ants that are doing things and the rest of them are just sitting there. So observing these dynamics can be subtle because you need to take a long view and look at places where there actually are variable dynamic behaviors that you could quantify. Not to say that you can't do it, but it might not jump off the page.

[54:06] Michael Levin: I wish I knew this stuff better. I think there should be a lot of parallels here with the kinds of things that we see in morphogenesis.

[54:20] Buck Trible: I guess one thing I'm hearing from you is that there's an interest in finding bioelectrical signals provide a convenient glue to bind subunits to a common purpose. So looking for the signal that is being used to act upon. And that might be one of your recommendations for me with this allometric scaling problem.

[55:02] Michael Levin: There's a couple of things that I could suggest in practical terms. One is keep an eye out in all the genetics and all the sequencing and everything else that you guys do. Keep an eye for anything that's an ion channel, ion pump, or an innexin, which is a gap junction component in vertebrates. If you see any of those, let me know and I'll help you make some sense of it. That's one thing. Another thing that you might do: are you able to drug these guys? Can you put compounds in the water?

[55:39] Buck Trible: So we have our cytochrome enzyme B as a good candidate. We've tried titrations of about five or six different agonists and antagonists of cytochrome signaling, and none of them were methyl, and we didn't get anything cool out of it.

[55:57] Michael Levin: We should talk offline about different options, and I don't know how onerous these experiments are, so maybe you can't try too many things. But I could suggest some cognitive modulators, some bioelectric and some not, that you could put in the water to communicate with the collective intelligence of the colony. We had a project a while back to try to communicate not with the individual ants, but with the colony itself. What we wanted to do was train it, and I can describe how that was going to be. We weren't going to train the individual ants; we wanted to train the colony. I think from that perspective, doing the kinds of things to the collective intelligence that you would do in neuroscience to probe the neural collective. There are some good tools that you can use. I think that would be interesting. I can suggest some specific reagents.

[56:57] Buck Trible: That's one thing I gleaned from what you were saying. The other one is this idea of introducing gross perturbations and seeing what state the system returns to. It seems like that's one of your other major inductive tools that you advocate.

[57:24] Michael Levin: There's the problem solving and then there's the creativity. The problem solving is, we hypothesize a goal, we put barriers between the system and its goal, and we see how good it is in reaching that goal despite all the weird things you're going to do to it. That's the kind of stable decision making and problem solving. But there's the flip side of that: I put you in a situation that you've never seen before, and now the question is, can you pick a new goal and implement something completely different? We've seen this — that would be like our Zenobot or our Anthrobot kinds of systems — where you do something to the, the genetics are still in, or wild type, you haven't touched the genes at all, but you put the cells in a novel environment. They will create some kind of autonomous, coherent living being that has structure, behavior. In fact, now we've studied the transcriptomes that have never been under selection per se, because this thing has never existed before. There's never been pressure to be a good Zenobot. The question is, how good are you at meeting the goals you already had that have an evolutionary history? How good are you at improvising on the fly, novel ways of being that are totally new. I love these kinds of AI models for the possibility to really probe that. Chimerism — we had a review with Vasili Nanos, who's my grad student — basically looks at chimerism as a tool to probe the nature of the collective intelligence. You introduce components that come from a system with different goals, or that have different properties. Then you look to see — in the cases where they don't get immediately rejected — what are the properties of the collective now, of the chimeric collective?

[59:26] Buck Trible: I know that we're short on time, but I think the thing that I'm really curious about is evolution, and you had this paper about how these types of intelligent processes shape evolutionary outcomes. So the question for me is, if you have an evolutionary question or a deeply conserved morphological or evolutionary pattern, like this allometric scaling pattern in the ant castes, how do you use these types of tools and approaches to address that question?

[1:00:19] Michael Levin: Do me a favor, and send some links to some of the key papers.

[1:00:36] Buck Trible: Yeah, I will.

[1:00:37] Michael Levin: I'll put it up. I'll show it to my group. We have a good machine learning group that's now looking at a lot of these basic questions and collecting questions, or the invariance between learning and evolution.

[1:00:58] Buck Trible: Yeah.

Michael Levin: There might be some useful stuff. I'll show it to them. I'll also put the links up with the video so that other people can see.

[1:01:07] Buck Trible: Absolutely. If there's anyone in your group that's interested in talking, I live pretty close by and I'd love to come out and just see, because I've been at Harvard for five years, and I'm starting a lab at the University of Georgia next year. I'm taking a bit of a pause, getting stuff published, but really trying to get set up in a new way and rebuild in an intelligent way that moves forward and builds on what we have now, so it's a great time to receive input and design experiments and make plans, because I'm not nearly as constrained as I would be if I were in the middle.

[1:01:47] Michael Levin: Do you know Will Radcliffe?

[1:01:53] Buck Trible: I do. I can't remember why. I know the name, though.

[1:01:56] Michael Levin: I don't know if it's Gio. He's in Georgia too. I don't know how close physically he'll be, but he's another really good person to talk to. He's a brilliant dude, and he's got a lot of good work on evolution of multicellularity, which ultimately will impact what you're doing.

[1:02:19] Buck Trible: Hey man, I really appreciate it. I learned a lot and this is really great. Thank you so much for taking the time.

[1:02:24] Michael Levin: Thank you so much. I learned a lot too. It's a really interesting system to think about. I'll keep thinking and I'll show this to a bunch of my lab people.

[1:02:33] Buck Trible: Sounds good, dude. Rock on.

Conversation with Buck Trible on his work on static allometry of ant castes and emergent cognition

Watch Episode Here

Listen to Episode Here

Show Notes

Transcript

Related episodes

“The Unreasonable Effectiveness of the Behavioral Sciences in Developmental Biology and Biomedicine”

Discussion #1 with Katrina Schleisman and David Burke

Conversation with David Resnik 1

Conversation with Buck Trible on his work on static allometry of ant castes and emergent cognition

Watch Episode Here

Listen to Episode Here

Show Notes

Transcript

Related episodes

“The Unreasonable Effectiveness of the Behavioral Sciences in Developmental Biology and Biomedicine”

Discussion #1 with Katrina Schleisman and David Burke

Conversation with David Resnik 1

Thoughtforms Life Podcast