Modern biology is advancing by leaps and bounds, not only in understanding how organisms work, but in learning how to modify them in interesting ways. One exciting frontier is the study of tiny "robots" created from living molecules and cells, rather than metal and plastic. Gizem Gumuskaya, who works with previous guest Michael Levin, has created anthrobots, a new kind of structure made from living human cells. We talk about how that works, what they can do, and what future developments might bring.
Support Mindscape on Patreon.
Gizem Gumuskaya received her Ph.D. from Tufts University and the Harvard Wyss Institute for Biologically-Inspired Engineering. She is currently a postdoctoral researcher at Tufts University. She previously received a dual master's degree in Architecture and Synthetic Biology from MIT.
Click to Show Episode Transcript
0:00:00.1 Sean Carroll: Hello everyone and welcome to the Mindscape Podcast. I'm your host Sean Carroll. You know there's a lot of excitement these days about artificial intelligence and computers and technology more generally possibly changing the world in a dramatic way. We talked about that in a recent solo podcast as well as with other people. But the other thing we mentioned in the solo podcast the Royal We... Of course me, is biology because biology has not gone away in its dramatic effects on what we can do to ourselves biologically not to mention how we can program and build and design biological organisms to do interesting things for us. The basic idea is that when you want to build some technology to either affect things on very very small scales or do things very very delicately or very precisely or to do them in a biological context within a person. To cure diseases to deliver drugs or something like that. Very often Mother Nature has been there first. Has figured out a way to do these things that you want to do and so you don't have to invent from scratch how to do them.
0:01:11.0 SC: DNA and the cells and metabolism with mitochondria could always be very very useful in these precise contexts. So it's very exciting to contemplate what's coming down the road in terms of ways that we can manipulate biology and use biology to help us in a bunch of ways. And a recent breakthrough along exactly those lines is the subject of today's podcast. Gizem Gumuskaya is a researcher who recently got her PhD. She's now a postdoc and she works in the lab same lab as Michael Levin who was on the podcast before. And Gizem's thing is something called anthrobots. This is a subset of something called biobots. Biobots are basically little robot-like things but made out of cells of one form or another. And as you'll see from the podcast it's still in a very very early form. It's not like we have many different pieces that we're designing and putting them together in an intricate way but the point is you can take cells and in the case of anthrobots you can take human cells and you can sculpt them.
0:02:20.0 SC: You can nudge them into a particular configuration and then some of them won't do what you wanna do but others will. And you can keep the ones you want and discard the ones you don't want. And what they can do is pretty amazing. They can heal things inside your body. They could like we said deliver drugs. They could be used as sensors to figure out what's going on inside you. And it is just the very beginning of this process. This is different than DNA robots or synthetic biology. When you are designing the genome. This is just using the cells that you already have and putting them together in very interesting ways. So I think that what comes through in this conversation is both that what's already been done is extremely amazing and exciting and that this is just the beginning. That we're gonna be going places with this kind of technology that it's hard to foresee the impact of. But certainly, the impacts are gonna be big. We do briefly talk on, are there any dangers here? But I think for the most part the possible impacts here are gonna be pretty awesome for humankind. It's a wonderful landscape of opportunities out there and we're gonna be exploring it. I'm very excited. So let's go.
[music]
0:03:51.7 SC: Gizem Gumuskaya. Welcome to the Mindscape Podcast.
0:03:55.1 Gizem Gumuskaya: Thanks so much for having me. It's so exciting to be here.
0:03:58.3 SC: Let's just start with the very big picture here so that people know how to orient themselves. What are you doing? How do you think of your task or your project biologically? Is it designing better robots or is it manipulating life or what is your self-conception there?
0:04:19.3 GG: I think there are a couple of different layers. Two big layers are science and engineering. So number one goal is really engineering new systems by leveraging what nature has to offer in terms of us being able to create new types of structures and living architectures that are not necessarily evolved but designed by us. So pushing the boundaries of what we can do in terms of harnessing nature's unique properties when it comes to construction. So this is regeneration healing replication of constituent building blocks self-construction ability to sense and respond to the environment. These are all really amazing properties we would love to have in artificial structures but they have been so far exclusively reserved to natural structures. So bringing those hallmarks into engineering through synthetic morphogenesis is the first layer. And then in doing that in the second the scientific layer, we're understanding how morphogenesis happens in nature. How these amazing structures arise in nature in the first place? What are the rules there? What are the knobs we can turn? So understanding system to create new things with it. So really those are the two legs of this pursuit of synthetic morphogenesis.
0:05:47.0 SC: Well you used the phrase synthetic morphogenesis which I think we're gonna have to define. Probably I can figure out what synthetic is. Genesis probably morpho is shape but what do you have in mind?
0:06:00.0 GG: So it's really starting with the concept of morphogenesis which is this amazing thing that happens in nature. Basically development of form in nature. This process of a single cell building itself into this functional multicellular organism. The process of morphogenesis is seen across all kingdoms essentially from simple biofilms like a single bacterium self-replicating and giving rise to this amazing biofilm all the way up to higher order mammals. So it is the development of form in nature. The business of morphogenesis. And the synthetic part is... Well as I unpacked at the beginning understanding how this happened and then steering it into new ends. So it's really bringing this goal-oriented design which is something that does not exist in nature. It is a human construct. Merging that with this trial and error and bottom-up construction that nature uniquely reserves bringing the two together to have nature build itself into end design by us humans.
0:07:14.0 SC: So I have a very long thought and I'm sure that I've said this many times in the podcast before that nature has a huge obvious advantage over technology because usually when we build things technologically we build them out of metal or whatever is the minimal thing to do with the shape we want to do. And as a result the things are very brittle. Like if my car breaks down I have to take it to someone to fix it. It doesn't fix itself whereas nature has the job of fixing itself. So in that sense, it's kind of an obvious place to look to build better things to piggyback off what nature does right?
0:07:49.6 GG: Yeah absolutely. And I think why this hasn't really been accessible to us I mean we all admire nature right? That's the genesis of science trying to understand nature trying to understand how all these processes happen. And it really looked to us like magic because it has this magical quality due to its bottom-up nature. But for the first time really in the history of science, we're really trying to understand now what's the logic behind that magic. And we've had the genetic code started being sketched out in the 20th century. So what we are really doing in the 21st century is looking at that and understanding that systemics behind it and then essentially recoding it in order to still retain those magical qualities that only nature has to offer like self-construction or regeneration but bringing that engineering coding mindset and trying to get it to do something specific. And biobots are just the first example or one of the examples of where we can take this.
0:09:10.2 SC: I could imagine that somebody could just take inspiration from nature but more or less design from the ground up. Maybe that's very hard and you don't wanna waste all your time rediscovering things. But if I understand correctly what you are doing is literally starting with cells and pushing them around to make robots.
0:09:30.2 GG: Yes. That's an excellent distinction and one I actually try to clarify often in my explanation of what we are doing. So there is being inspired by nature and that is something we've been doing for hundreds of years literally since the Asian times right? Looking at nature admiring it building architectures like ionic columns that look like nature. But in doing that we are still exclusively using the top-down methods that us humans have developed where everything needs to be manipulated one by one. It lacks that autonomy. And still to this day in the 21st century, we are doing things like this. I mean a lot of... In biodesign, there is molding things or even 3D printing to a degree is just a contemporary version of that building things top-down in order to approximate the final product to what we might be observing in nature. But what we are saying is let's take some of what we have developed as humans. Again goal-oriented thinking. All these engineering principles of modularity robustness and marry that to what nature uniquely has to offer. Mainly self-construction.
0:11:00.5 GG: So yeah I am literally working with cells and recognizing that in those cells there actually exists a morphogenetic code that we can work with and that is amenable to human design. So I think just that combination of the two is really a nascent field and it is going to go to really exciting places.
0:11:25.5 SC: Okay. Let's be more specific about this biobot idea that you've already mentioned. Apparently, the earlier thing before we get to your particular anthrobots there were xenobots. And again if we have a classical education we can probably figure out what those mean a little bit. But tell us about the xenobots if you can.
0:11:47.3 GG: Yeah so xenobots were the first fully cellular biological robots as we call them. So prior to this, the field of biobots existed. A lot of law was being done in the field though is precisely what we're talking about. A hybrid between gels or scaffolds or things that could support the cells. And then on top of that, you can add cells in order to optogenetically induce them or harness some of the biological qualities. So they were hybrid biobots. With xenobots which were developed in my PhD advisors Michael Levin's laboratory in collaboration with Josh Bongard's lab in the University of Vermont by Sam Kriegman and Doug Blackiston. So xenobots were the first fully cellular biological bots so there is no scaffold no support. It is created using frog embryos by extracting tissue from frog embryos and by surgically manipulating. So there is still that top-down manipulation aspect. It's not quite self construction but it's one step forward in my definition of where we're trying to go which is quite literally building self-constructing robots and structures. So it was definitely one step forward in that axis.
0:13:24.4 GG: How they were built is that cells were extracted from frog embryos and sculpted into spheroids with cilia covering their surface. It was discovered that these xenobots are able to move around in different patterns and able to do useful work like aggregating loose cells. And it was really exciting because now with that work it was really in the field of biobots we're able to see that we don't need any type of inner materials. We can use only nature and we can work with it to get it to show up as a new architecture.
0:14:09.6 SC: These were stem cells they were starting with I think. So maybe you should tell us what a stem cell is and why that mattered.
0:14:16.1 GG: Right. So these cells were from frog embryos of frog embryonic stem cells. And why stem cells are important here is because stem cells have huge potency in terms of the different types of tissues and structures they can become. So they have a high degree of anatomical plasticity as we call it. What that means is that a cell's or a tissue's ability to become different things. For example, as adults, your morphology is locked in. You don't really have a ton of plasticity but at the embryonic state the adult grows... So that adult has a lot of different tissues and every single cell in that adult's body is coming from one original cell. So every single cell in the body is sharing the exact same DNA but they have all different kinds of properties. So our cells in our eyes and in our kidney have the exact same genetic makeup in terms of DNA but have completely different functionalities shapes.
0:15:25.9 GG: Well this is because that original cell in the embryo had a huge degree of morphological plasticity. It could become a lot of different things. And during the process of morphogenesis through differentiation, as cells self-replicate one becomes arm, one becomes leg, one becomes hair, one becomes skin. So there is that ability to become different things in embryonic stem cells. And in xenobots, they have essentially leveraged that in order to build a new architecture that has a completely different shape and function than that of a frog embryo. So that was really exciting to see.
0:16:12.6 SC: And you used the word or the phrase sculpting cells. I would like to know what that means. Are you sculpting individual cells? Are you sculpting how they fit together? And how does one sculpt at the level of cells?
0:16:28.1 GG: Right. So it's sculpting a multicellular tissue sculpting an aggregate of cells. It is a very fine surgical operation that our surgical genius in the lab Doug Blackiston is able to put together. I myself I'm obsessed with self-organization so I actually don't do any sculpting and if you gave those cells to me I cannot make you a xenobot. But yeah Doug has a really fine hand as you call it. And he was able to put together and he has these really beautiful videos. I encourage listeners to go look them up for how he builds them. And he can also build them in all kinds of different shapes different protrusions and inlets. Essentially it is very much like sculpting very much like kind of almost like playing with dough and giving it into... Build it with a scalpel under a microscope, but giving it the shape that you want it to have. So it has that advantage.
0:17:37.7 SC: So you're giving the collection of cells this shape or you're shaping an individual cell?
0:17:43.4 GG: No it's a collection of cells. You're shaping a multicellular aggregate.
0:17:48.3 SC: Okay very good. And so in the xenobots anyway, you basically are nudging the stem cells together to make a specific shape. Do you do any more than that? Do you edit the DNA or give it chemicals or how do you make a bunch of cells into a biobot?
0:18:09.0 GG: So in that case they also supply the system with different inputs. I believe they use a notch inhibitor. And so they have a matrix in the 2021 paper I believe it is. And based on the different environmental inputs you give to the system it is able to acquire different shapes and properties. They haven't really edited the DNA or integrated different genes to make the xenobot. Once you have your xenobot and same thing with anthrobots you can use it as a platform to add in different genes. So it is still very much... And I think we should get to these different approaches to synthetic morphogenesis and genetic circuit approach versus this more morphological and anatomical plasticity approach. The two are complementary. So once you have your biobot platform you can still add genetic circuits which has been the predominant approach in synthetic biology so far. And they have done this too. In xenobots, they added an optogenetic receptor that enabled them to communicate with xenobots through light. And I believe that's also in the second or third paper. But to make a xenobot you do not need genetic circuits or you do not need to change the DNA.
0:19:34.7 SC: Okay. Very very good. And one quasi-philosophical question before we jump into the anthrobots in particular. This biobot is a new kind of thing. It's not an organism but it's not a robot either right? I mean has anyone maybe philosophers or maybe someone else opened their eyes and said oh my goodness you've created not just a new thing but a new kind of thing?
0:20:05.5 GG: I don't quite know if it's a new kind of organism because this kind of goes into how do you define organism all the way down to how do you define life? I think it's more like expanding the pellets of stable morphologies that we can create using things that we call nature right? Using cells that are evolved. We did not build those frog embryos, they are natural in that sense. But by using them we are able to create a stable fully cellular morphological steady state structure. So I don't think it's a new organism quite and maybe that's not even the useful question to ask but in terms of why they are robots. In general I think the field of biobots is using the term robot maybe not with utmost rigor. I think in general starting with earlier hybrid biobots, xenobots, anthrobots we call them robots because there is some degree of programmability. So programmable anatomy I think is the justifying factor there to call it robot. But there is a lot of side discussion going on for terminology and how should we refer to them. Mike has a whole blog post about this. I also encourage readers to check out his blog and read on that if that's of interest.
0:21:41.6 SC: But the Biobots at least so far do not reproduce themselves, right? There's no natural selection going on just within the Biobots.
0:21:49.4 GG: Correct. So for xenobots for paper number three what they have done is to show that these xenobots that move around so far mindlessly because we haven't really seen otherwise. So doesn't mean they can't be, it's just that we haven't seen otherwise. Collected cells together aggregated individual cells that are free floating in the dish. And they were able to show that these cellular aggregates themselves were able to then move around because they also have cilia on their surface. And this went on for three generations. So in that sense I think there is a kinematic self-replication. Although that reminds me of machines that make... 3D printers that print parts for more 3D printers. So bots aggregating cells to create aggregating bots that aggregate cells. So it's kind of that recursive nature of it maybe is what's making it look like a software replication. But that's the... I think we've dipped our toe in the software application water in that way with xenobots but not with anthrobots. Anthrobots... They can aggregate cells together like xenobots however because their modus operandi is not through aggregation but through self-construction. That aggregation itself doesn't create a new anthrobot or we're not making that claim.
0:23:40.9 SC: Okay. I think it's probably then time to tell the audience what an anthrobot is. A xenobot. The xenobot experiments we'll link the papers in the blog posts. But they were made of stem cells from a certain type of frog. The anthrobots. You just dig up little cells from human beings.
0:23:58.4 GG: Yeah and not from embryos spoiler alert. So yeah so when I started in the lab 2018 we wanted to make mammalian versions of these xenobots. However, it is not... So two major differences. Difference number one you cannot just take... And we ideally want to make these from human cells because we're now interested in using these in medicine potentially. So you can't just take human embryos and pick cells and play with them and then put them together and see what they do. That is just not gonna happen. And even if that was allowed I actually personally think there is a lot of merits in trying to get these bots to build themselves from single cells to recapitulate like go even one step further. Not just use fully biological cells but also have them bring to the table what they can uniquely do. So I can sculpt a lot of different things but I can't make anything build itself unless it's of biological substance.
0:25:11.3 SC: And just to be clear when you say that you can't do that you're legally not allowed to do that in the United States.
0:25:19.0 GG: Right.
0:25:19.5 SC: You're not allowed to take little human embryos and sculpt them [laughter]
0:25:23.5 GG: Exactly. Poke around. I mean I think up until day 14 I believe it's allowed not to poke them but to run experiments. But beyond a certain legally defined threshold you are obliged to shut down the experiments. And so it is more like a legality and ethical issue rather than science. There's no reason why you can't take cells from anywhere and there's no reason why you can't physically do that. But that's just not something that is done. And even then I think that as I said like sculpting we just want to try a new construction modality and I mentioned at the beginning I do have a background in architecture. So I actually came to synthetic biology from architecture or civil engineering space in order to explore some of these potentials that biological systems have in terms of how they build everything we see in and around us. So the idea of getting cells to build themselves was really interesting. So I think this is a really good case study actually for synthetic morphogenesis because... The building of anthrobots because our really first step was, what is our goal?
0:26:57.2 GG: What is our end morphology? Which is a very much engineering problem or even in design. You just start with what do I wanna see in the final system? What are my design requirements? So for anthrobots, the list of design requirements is as follows essentially. In terms of structure we wanted to basically replicate xenobots. So we already have a target structure right there. A spheroid with cilia on the surface so they can move around. Or actually that was another thing we were... So this is the side of that tension between science and engineering. This is a very much engineering goal. I wanna make a multicellular spheroid with cilia on the surface from human cells because I wanna make these bots that move around and that I can put into a human body. This is all... We're very completely talking about engineering here. But there's also this underlying scientific question. Well if I create a multicellular structure that looks exactly like a multicellular structure that is from a completely different kingdom right from amphibians?
0:28:02.0 GG: Will it move? Will it do the same thing? What will the similarities be? What will the differences be? This is also anthrobots versus xenobots and there is some minute differences like anthrobots are a little smaller. We can get into that later. But essentially in terms of their morphology, they're kind of like twins. They're both spheroids. Multicellular spheroids with Cilia on the surface. And I don't think we've ever seen in the history of life on this planet where two multicellular structures or two structures of any kind looking so similar to one another but from being in completely different kingdoms. So recreating that in the lab and investigating their behavioral and morphological properties also I think really helped. It was a really exciting question and really helped us understand some of these morphogenetic rules that we are trying to crack. So this is just kind of an example for science engineering tension but going through the list of requirements when we're building multi-cellular spheroids with cilia on the surface human cells and then self-constructing.
0:29:24.0 GG: So we have our target goal. Target engineering goal and we want to get cells to build this. So how do we go from here as a synthetic morphogenesis project? So the traditional way to approach this project would be through synthetic circuits which is the dominant modality that is currently used in synthetic biology. What that embeds... It's fascinating and learning about synthetic circuits and finding out that we can do these things was the thing that sucked me into biology from design to bring the two together. So it was massively inspiring but I was also realizing that maybe it was coming to its limits. So with anthrobots then we have our list of goals which is where any engineering project starts where am I trying to get to? And this is not how nature does it. Nature is just trial and error across millions of years that brings about something that works.
0:30:29.2 GG: But for us grant timelines are not millions of years. So we do need to get to what we are trying to build hopefully in the span of a PhD. So starting points okay we want spheroids we want spheroids covered with cilia. We want these from human cells and we want these to build themselves. With that goal list in mind if a synthetic biologist sits down for a synthetic morphogenesis project traditionally what we do and this is something... In my work in my master's prior to PhD, this is also the space where I learned all my synthetic biology knowledge is through synthetic circuits. So I'll just pause here and unpack a little bit what synthetic circuits are. So synthetic circuits are very much like electrical circuits. Instead of though having transistors connect to one another with copper wires you essentially have genes connect with one another through proverbially chemical wires. You have this one gene getting expressed and that could be a regulatory protein that goes and turns on or off another gene. And so through these interactions, you can generate Boolean logic in live cells.
0:32:03.3 GG: And that is very much an electrical engineering paradigm bringing this to biology. So in the past, there have been some amazing examples for synthetic morphogenesis creating synthetic patterns in space. A lot of things have been accomplished through genetic circuits. However though if you wanna create something... These all have been in 2D or 3D fields so it has not quite yet gone beyond just creating patterns and I don't wanna say just because even that's fascinating getting cells to create large-scale patterns. But if you wanna do something quite large as a multicellular spheroid with function so growing cilia on the surface and moving in one direction which means there is a symmetry-breaking event going on there is directionality there is axis. I mean this is just not something we're going to yet be able to accomplish using just genetic circuits because even if you can fathom what kind of a circuit would create this type of incredibly complex tissue organization event, for us to build that circuit and deliver it into cells? Just our technology is not there yet.
0:33:27.7 GG: And the biology community is a really fast-moving community. So I think we'll get there in 50 years which will be my lifetime. Would love... I love that idea of designing a structure and then compiling it into a genetic circuit and putting it into a single cell and have that cell execute that circuit and start self-replicating itself and proliferate into that structure. Very much like the growth of a tree. And in doing that also copy those instructions that we've put in there and propagate it into the progeny so everybody knows what we're trying to build. That is I think really the hundred-year view of synthetic morphogenesis. But we are not there yet. And I did have to get my PhD [laughter] So basically we started thinking about, is there another approach to synthetic morphogenesis that we can think about? Why are we trying to build everything from scratch? We don't even yet understand fully how ciliogenesis happens in nature. That is an incredibly complex process. The pathways are not all even worked out. How are we going to take this and put it into a synthetic circuit? Now Why don't we just look at cells that already know how to make cilia? So instead of trying...
0:34:53.6 SC: Sorry just by the way. The cilia are the little hairs that move around the right kind of cell. Right.
0:34:58.8 GG: Exactly. So cilia are little hairs exactly. They're in our... So there are three places in the body where cilia is found. So you have the mucociliary escalator which is in the trachea that helps inhaled pathogens and particles to be sent back up. You have it in the brain matrices and then you also have it in the oviductal epithelia helping the egg to move around. So it is essentially this locomotive appendage... One of the locomotive appendages. There are different ways of locomotion in the body. So very fascinating, very complex. So then really this new approach to synthetic morphogenesis is asking the question can we leverage what cells already know how to do and have them deliver as much of the final engineering goals as possible without really interfering at the genomic level but only by engineering their environments nudge them so to speak towards the designed goal.
0:36:05.0 GG: So this is really a shift from looking at how morphology develops in nature. I think the traditional view is very Gene-centric. And this is something Mike's lab has been working for a very long time. The traditional view is everything is encoded in the genome. If you read my work prior to Anthrobots it's also very much like okay which gene are we editing? How are we building this? And so the PhD has really gave me the room to think about this more extensively and really started realizing that the morphogenetic code encompasses more than just the genetics. It also has these added layers of environment and epigenetics. And by changing things. Tuning things at these levels we can also change the final morphology. Mike has this STAR setting work double-headed worm. So he took a worm... This is years before I joined the lab. If you cut the head head grows. If you cut the tail tail grows. So they cut the head and changed the bioelectrical signature at the cut point and managed to get a tail grow instead. So you have two-tailed worms and two-headed worms.
0:37:39.3 GG: And there are a bunch of different examples like this whereby just changing the environmental inputs you can change the morphology. So for then building anthrobots, we asked the question which cells in the human body already know how to make cilia? So we can go with brain cells, we can go with tracheal cells, we can go with oviductal epithelium. We've just decided to go with cells from trachea because the cells are just more available, because of the more widespread research on lung diseases. So started with a single cell from the epithelium in HVS and looked at different ways to get these cells to create cilia. So already there are protocols out there for getting ciliated epithelium to build itself so we can study. And this is again where science and engineering is splitting up. In science, we just want to create these things so we can better study and better understand the native tissues. So the goal is really... So in the field of organoids the goal is to recapitulate tissue architecture. We don't want multi-ciliated spheroids running around because that doesn't look anything like what's in the human body.
0:38:51.7 GG: So there wasn't anything... So basically just kind of looking at what type of culture methods are available. So there was one real interesting culture method that helped these cells to grow into spheroids but cilia is looking inside. So this is a traditional area organoid method. We take cells from human trachea culture them in an environment with a dense net matrix and then they grow something that's very similar to what's in the human body. This lumen with cilia lining it. So that's great except that's the exact opposite of what we're trying to do. Cilia is looking... It's a spheroid with cilia inside. We're like no this is not going to move around. We want a spheroid with cilia outside. But look at how close we are. I think that's why it's really powerful to look at what these cells can do on their own readily before we kind of jump in on genetic circuits and trying to build it from scratch.
0:39:51.8 SC: Can I ask just very quickly? When you say spheroid we have this collection of cells we're talking about I don't know 100 cells maybe?
0:40:02.1 GG: Hundred? So yeah a very varying number but... Yeah from 100 it could go up to thousands but in the order of hundreds yes.
0:40:11.3 SC: And the spheroid is empty inside? It's just the container or just the shell I suppose. It's not a solid ball.
0:40:20.5 GG: Exactly. So it is very much like the trachea. It's empty inside. There's a lumen and it's filled with mucus and other debris that the cilia is moving around. Some of them... It quite literally looks like a washing machine with an empty hole and then things are moving around in there. So this was something that was so... So it's basically like the first step is to, okay here's my engineering goal. Here are the things I needed in my system. What already exists out there that does something similar to this? And then how can I nudge it in different ways to explore all the different shapes and structures it can create? So I push it towards my engineering goal. So again as an example for anthrobots, we had the airway organoids. So we need to get these guys to flip. I think this is where thinking about cells and their morphogenetic functions what they like what they don't like, what environmental inputs. So it's known that the cilia grow in the middle and not on the surface because the surface is this thick matrix environment. And cilia likes to be closer to more kind of a lower phase more water-like environment.
0:41:42.4 GG: And that's the empty lumen or filled with mucus. So essentially trying a bunch of different things but the hypothesis that worked was that okay what if once we have these cilia in airway organoids if you remove the matrix from the environment if you just melt it out? I say just but even this is five months of research because you're like okay now how do I get just the matrix to dissolve but keep the cells intact? But ultimately the way to dissolve the matrix and keep the spirits intact. And when we put them then into a non-adhesive environment so not just in... Especially non-sticky environment to really force those ciliated cells to come outside while bombarding the system with retinoic acid which is also something ciliated cells are known to really thrive on. In a matter of seven days, I managed to get these spheroids to essentially flip inside out which is really fascinating because that almost looks like gastrulation which is an embryonic event but unfolding in a structure that has nothing to do with the embryo literally taken from elderly patients.
0:43:04.7 SC: Again just to be clear you have this spheroidal collection of tracheal cells and all the cilia want to be inside and you didn't convince the individual cells to get cilia outside.
0:43:19.5 GG: No.
0:43:20.5 SC: You just convinced the whole sphere to flip inside out.
0:43:23.3 GG: Exactly. Undergo this phenomenon called eversion. So essentially what a bioengineer does at that point okay here's my engineering goal. Here's what's already available to me. Here are the kinds of structures that are either developed by evolution, meaning these are the natural structures available to me or by other scientists, the airway organoid. So a survey of what's out there and then from there finding the closest point to where you're trying to go and then asking... We're not messing with the... We're not going sculpting and trying to get it inside out by hand none of that. Those are all very much top-down construction methods but asking ourselves, what kind of morphogenetic functions can I get these cells to engage in. What kind of functions can I execute in their morphogenetic code? So I get the system to build itself towards my target morphology.
0:44:23.1 GG: And the morphogenetic function there was eversion is the terminological name to get it to flip inside out. And then you do a bunch of things to the spheroids to get them to do that. And then you have a really... Yes, developing the protocol is very painstaking but once you've done it then you have the system that... So anthrobots build themselves in a matter of two weeks from single cells to multicellular structures and then one more week for them to flip. So it's three weeks. And in the course of these three weeks we just feed them twice and then essentially melt out the matrix. So change your diaper. And we go from thousands of cells to thousands of bots. I mean there are thousands of them just running around. And that's the... I think real beauty of self-construction beyond that intellectual bleeding edge trying to... That architect side of me. Beyond that, it's just very canonical like high throughput. It's not like 3D printing. When you're 3D printing you have to print them one by one. It's not like molding or casting. You have to create individual molds for each thing you're trying to...
0:45:43.8 GG: So I think that's why it's really exciting to ask the question of can we really get this unique construction framework of biology to work for us towards our design ends?
0:46:00.8 SC: So you have these multicellular collections and you've nudged them chemically not mechanically...
0:46:09.2 GG: Correct.
0:46:09.8 SC: To make sure that they can move around. They have cilia outside. But they're not... Just to be fair they're not programmed to do any particular thing. I mean they're not robots that you designed for a purpose. You made them to fulfill some requirements and then you're going to watch them go and see what happens.
0:46:30.7 GG: Right. The goal there is to make a spheroid that moves around that in and of itself. And you can kind of argue that you've programmed the system to reach that goal. So that program... I think that creating artificial anatomy part is the reason why in general people call these structures robots. But here's the interesting thing. Sorry was there a question?
0:46:55.1 SC: I'm just trying to make sure that the audience understand so like what the current capabilities are. This is not a robot that has a computer memory that you can give it instructions to. You've basically built a thing and are just doing science on it asking how it moves around and what functions it could possibly serve.
0:47:16.2 GG: Yeah. And I think that's a really exciting thing. When you do that with biology by leveraging these emerging dynamics because in the traditional way, you would just expect that okay, now I have these spheroids that move around do this one thing that I wanted to do. Except that what we saw is that they're all doing different things. One is moving in circles, one is going straight, one is wiggling in place, one is making arcs. And I think there lies the cue to further programmability. If each one of these guys are doing something different there must be something categorically different about them. What is there? I mean that really became the next question for us. Because you look at that dish you made the mistake of getting them to self-construct. So now you have thousands of them and it's overwhelming. And you want to characterize the system. So what we did is to ask the question are there any emergent patterns here? Are there cues to... 'Cause all we wanted to do was to make these things kind of move around so they can go through live tissues and we could see what they can do. That was the degree of control we wanted to have. But then now we're seeing they're all doing different things. And now we want to understand well what are the differences between these different bots here.
0:48:50.8 GG: And the first thing we did was to do a large-scale time-lapse and look at the world population. This is again really interesting. So when you look at... So biology does this idea of variation on a theme. Yes, they're all ciliated. They're all multicellular. So they all look like each other but none two bots are the same. Sort of like fingerprints. They all kind of look similar but none two are the same. It turns out actually the FBI has a whole way to categorize fingerprints. So each one has a fingerprint type. I believe there are eight categories. We can double-check but it's really on their website. So I'm like wow this idea of character formation. So are there different categories in how these bots move? Was the first question. We collected this mass time-lapse data and then tracked them. And I did a PCA. So shout out to Simon Garnier from New Jersey Institute of Technology. He helped us a lot with the stats on this and my student Pranjal Srivastava.
0:50:00.6 GG: So the three of us tackled this question of can we find categories here? And it turns out there are four statistically significant distinct categories in this chaotic system. You have bots that go in circles. You have bots that go straight. You have bots that go in arcs. And then you have bots that are random and eclectic and just kind of noise. So that was really interesting because now getting to that question like okay, what else can we tune during their developmental trajectory that we can now start to program the population to maybe always go in circles. Always go in straight lines or do it 50-50. So what are the engineering knobs in the system? And our first hypothesis was well it's got to be about morphology because we know that in nature... Well in architecture school they tell you form follows function. In biology, I learned function follows form in nature. So if these functions are different something's going to be different about their form.
0:51:16.2 GG: So then we did something where people who work with confocal microscopy will understand something heroic. 3D scanned 300 anthrobots. So with the confocal microscopy, we basically shoot the tissue with lasers to create a 3D reconstruction without really harming the bots themselves. So shout out to Ben Cooper and Hannah Lesser two of my students. So we tackled that problem next and then created this massive data pool of different anthrobots and then collected also motility data from a subset. So what that enabled us to do is actually the next thing. But with this morphology pool, we asked the same question are there stable patterns? And it turns out there are three stable categories. Three different flavors of anthrobots. You have these bots that are a little smaller and cilia everywhere. Fully ciliated. And you have these other bots that are larger and sort of like a patchwork cilia. And then you have this third category again larger but only cilia on one side and not the other. And we keep getting these frequencies. The same frequency in the population.
0:52:37.7 GG: So it's really cool that the system is self-organizing into different attractor states in terms of its morphology and its behavior. And then next the million-dollar question is there a correlation between the two? Can we map the frame not necessarily causation but at least correlation? Because from there you can hypothesize about causation but is there any correlation between them? And it turns out the ones that are half-ciliated half-bold are the ones that always go in circles. Ones that are large and...
0:53:11.8 SC: That makes sense.
0:53:13.4 GG: With sort of patchwork right? It's like if you have a rowboat and then if you're only rowing on one side you're going to go in circles. If you're sitting in a rowboat and have both sides are rowing you're going to go straight. So these larger bots with patchwork cilia checkerboard pattern always go straight. And then the small guys even though they have cilia all around in their surface actually are the ones that are wigglers. They don't really go anywhere because their surface is not large enough to have enough mini cilia to generate thrust. So yeah I mean once we kind of figured out that there is a correlative mapping between morphology and shape I think that's where the future efforts to program will flow because now you can kind of nudge a system to go one way or the other.
0:54:00.5 SC: And because I'm sure that the audience cares about this the anthrobots they don't reproduce and they have a finite lifespan right? They're not going to take over the world probably.
0:54:10.9 GG: No. Probably. No. So I have yet to see and I have seen thousands and thousands and thousands of anthrobots. It could be more than 20,000 now. Never seen one that never disintegrated. So every single time they disintegrate into individual cells after a month five weeks to two months period. So that differs from bot to bot. And basically it would just... If we are talking about inoculating these bots into the human body after a certain time they would just disintegrate and it wouldn't be different than not to be in your body. We're characterizing right now in a follow-up paper for their modes of disintegration. But yeah.
0:55:07.9 SC: But nevertheless we're burying the lead here because even though the anthrobots just wander off in different directions and move in different ways you found that they did have a potentially beneficial effect on nerve cells.
0:55:24.0 GG: Right. So once we did this characterization I was like okay we've built the thing we wanted to build. Let's do it. The whole reason why we wanted to use human cells was so that we can put these things into human tissues and then see if well what do they do? If they can at the very least go through the tissue traverse not navigate but move across. So we put these into human cortical neuronal monolayer differentiated tissues. So these are neuronal tissues or monolayer tissues differentiated from human-induced neural stem cells hiNSCs. After differentiation, we basically make a scratch. So damage the tissue in different ways to scratch outside standard. And then when we put the bots on there we saw not only that bots moved across but that the bots that had different motility profiles moved across differently. So the bots that had circular tendencies tended to explore the edges of the scratch more. And the bots that had more shape and tendencies just kind of went right through.
0:56:45.6 GG: So you can imagine a scenario where based on your application if you want your bot to for example dispense a drug. High doses of a specific drug in a particular tissue maybe you're gonna wanna use your circular bots so they spend more time there. Or if you want larger coverage like if you want these bots to patrol a tissue and collect information in a certain way maybe you're gonna wanna use the Linear Bot. So it's also useful in terms of application beyond morphological characterization. And then yeah when we put the bots into scratches they just went right through. However, in parallel we were experimenting with can we make larger structures using these bots. And we had discovered that at a very specific point, so when you first dissolve these from the matrix. Before they evert. Before they go inside out. If you constrain them a lot they actually start to fuse together and form this massive bot. And then that bot then actually continues local eversion at different points and then becomes motile. So you can have these very large kind of super bots that are motile.
0:58:00.1 GG: So when we put these things into the scratches they didn't of course move quite as well 'cause they're giants. They're not like the small guys that just go right across. But what they did was to function as a bridge on both sides of the scratch. So connecting the two sides of the tissue. And then what we saw is that at the end of a three-day period, they enabled the neurons underneath to fix themselves. We don't know if it's neurons like... I mean so the most likely explanation is that there's some migration event happening and that gap is closing in that way. And that migration event might be either due to cell tonin release from the bots or due to some electrical transport. We have different hypotheses that we're testing. So we don't know why they are fixing this tissue the neurons. But we know that when we put another type of I don't know a HEK cluster. A steroid made from these kind of HEK cells which are off the shelf we've tried that and we didn't see the same effect. And we also just try so that's controlling for different tissue types. And then we also just put a piece of Agros and also didn't see the same effect. So it's also not just mechanical loading. Yeah, we are investigating the mechanisms but that was very surprising.
0:59:32.6 SC: All you did was make some steroids that could move around and without even trying to program it too much they went in and started to heal some neurons which is pretty promising for what happens when we get better at it.
0:59:49.8 GG: Yeah I mean I think that's why the idea of anatomical plasticity and trying to leverage what cells already might be doing is very exciting. This is a complementary approach to synthetic biology. So from here, we can now add synthetic circuits to these bots and now we have the room for it to payload. Because we're not using any of that circuit budget on trying to get them to create this morphology and function. So I think the really exciting next step is to put different genetic circuits into these bots and see how we might be able to expand their abilities.
1:00:38.8 SC: So basically rather than trying to program the bots you can program something using synthetic biology or whatever and the bots are just easy to make and are I don't wanna say smart enough but are able to find and locate certain features in the body where you want to dump your payload.
1:01:00.8 GG: Right. I think finding where they go so that gets into navigation. I think actually that's where we might really make use of genetic circuits because right now their movement is random. We did very little chemotaxis studies to see if they have any taxis behavior towards anything. So we wanna expand those to see if they naturally gravitate towards anything. But I think navigation and actuation will be the areas where we can now incorporate genetic circuits and really expand the abilities of morphogenetic engineering as a whole.
1:01:42.4 SC: This is where your architectural background comes in and you try to fit different pieces together to do fun things?
1:01:50.4 GG: Yeah so I think number one is to try to see how we can create new ways of building new ways of creating structures. Because in architecture it's all top-down. You have to stack the bricks to build the thing but here we actually have this technology just sitting out there right nature. Where the bricks are self-replicating and stacking themselves. So it's just a question of speaking their language to get them to build the thing that we want them to build. So that's one. And then really I'm also interested... So I mean obviously medical applications are one avenue but I'm also interested in climate tech, whether we can use biological self-construction and morphogenetic engineering towards building more sustainable building blocks. Because right now close to half of all greenhouse gases causing global warming is coming from construction industry, just humans trying to build things. Versus you have this national paradigm that builds amazing things and sequesters carbon and now we can get it to build the things that we want built. I think there's a great opportunity there for sustainability as well.
1:03:15.0 SC: That sounds very good. I suspect that you would've told me if you thought that it was going to be able to cure cancer but it does sound like there's therapeutic medical applications as well.
1:03:25.5 GG: Right. Right right. So neurodegenerative diseases are I think one area where we could start testing this healing behavior and see in actual disease models what happens but... Right so beyond that I think this is more like a... And this is... It's not just anthrobots. I mean honestly, a lot of different biobots can be envisioned. The anthrobots are just one example. And you don't need cilia for biobots to move you can also do crawling behavior. So really the idea is to use nature as a palette to draw these features from, to build engineered systems. So once you have that system based on what your therapeutic goal is, you can envision different properties. So if you wanna chase some bacteria in the gut, you would engineer your system such that maybe you want it to move in 3D 'cause bots only move in 2D. But if you're trying to bulldoze mucus from cystic fibrosis patients. Then this ciliary crawling is a better locomotive paradigm.
1:04:45.3 GG: So really the idea is to create these platforms and adapt them to different types of therapeutic applications. So this paradigm shift of... We think of drugs as these inanimate chemicals that just do things in the body but what about living drugs? What about living medicine? So I think that's what morphogenetic engineering will enable us to experiment with in the therapeutic space.
1:05:18.9 SC: I guess it's my job to just ask what are the dangers of this. It sounds... Whenever we do something dramatic and biological you have to worry that it's gonna fall into the wrong hands or that a terrible mistake will be made. You said that they actually self-dissolve after a while. But maybe that's just the first generation I don't know.
1:05:39.8 GG: Exactly. Yeah, I think this is where if they're just able to explore their native abilities and leverage what they do naturally and deploy it in different and unique ways like in the neuro healing case I think the concern's less because then the bot is carrying the human DNA 100%. But if you're getting into integrating genetic circuits here, and how bots carry payloads. I think that's where maybe some off-target effects might be observed. But this is something the synthetic biology community tackles with a lot. And there are a lot of different strategies for this. For example, kill switches. The bots or whatever genetic payload is being deployed in the body could be deactivated by a molecule that's orthogonal to everything else in the body and specifically targets that genetic circuit and basically deactivates the whole thing. So there are definitely some fail-safe mechanisms that we could bring into the bots. But that's why I think bots are really exciting. Especially ones that don't require genetic editing because then you're able to just put a piece of tissue that has the exact same DNA as the patient.
1:07:13.3 SC: Human cells. Yeah. [laughter]
1:07:15.1 GG: Exactly. So if we were to make your cells your bots we would just take a cheek swab and then go from there. And then we would never touch the DNA. And then when we put those Sean bots into Sean once the bots are done what they're doing they would just kind of disintegrate and the body would never even know. And that's the advantage of staying under the radar of the immune system. But yeah once we expand this to include genetic circuits I think we're gonna have to think a bit more about off-target effects with the genes and incorporate some of that synthetic biology safety literature and protocols.
1:07:53.7 SC: So this last question is not even a question but just a statement that it seems like this is just the beginning of something truly big. I mean when you combine ideas from gene editing and synthetic biology to this robot building that you're doing it almost makes a physicist like me think that the 21st century will be the century of biology. [laughter]
1:08:17.5 GG: Yeah I mean I think it's just really exciting to see how much we can do. I think there is this... And we still haven't been able to shake it I feel like. There is this view of nature as this thing sitting outside waiting to be understood and mapped and decoded. And of course, we need to do that but as we do that, we're also discovering that this is a whole active design medium that we can build completely new things with. So yeah it is definitely... I mean to a degree that as a designer it just sucked me in and before I knew it I became a biologist. So I think there's huge opportunities for design and engineering to create things that are truly unique and have the whole works of biology. 21st century and beyond. [laughter]
1:09:09.4 SC: Yeah we always like to leave some... Give some ideas out there to youngsters who might be listening to Mindscape Podcast thinking about what they could do research-wise for a living. And this is certainly a very very exciting field in which to do them. So Gizem Gumuskaya thanks so much for being in the Mindscape Podcast.
1:09:27.1 GG: Thanks so much for having me, Sean. I've been a... I remember listening to your podcast late at night in the hood running experiments so it's full circle for me.
1:09:37.3 SC: Hope I didn't ruin your evening.
1:09:39.1 GG: I'm happy to be here.
1:09:40.2 SC: With what was happening then.
1:09:40.4 GG: No no you motivated them. [laughter]
1:09:42.6 SC: Excellent. All right thanks.
1:09:44.4 GG: Thank you.
[music]
I’ve been following Dr Levin’s work for quite a while and find it fascinating. He always makes sure to credit his student. collaborators so it was exciting to hear this one-on-one with one of them. More people should be talking about this field, as it’s absolutely mind-blowing.
It took me a while to understand anthro-bot as opposed to ant-robot. I had to double-check with the transcript.