Watch Episode Here
Listen to Episode Here
Show Notes
This is a ~45 minute keynote talk at the Bioelectricity and Cancer conference (https://cam.cancer.gov/research/bioelectricity_and_cancer_conference.htm) held at NIH in September 2024.
CHAPTERS:
(00:00) What is Cancer?
(13:00) Cognitive Light Cone & Cancer
(18:00) Bioelectric Layer: Cognitive Glue
(23:00) Bioelectricity in Cancer
(28:00) Tools: Imaging & Control
(32:00) Bioelectric Diagnostics & Induction
(37:00) Suppressing Cancer Bioelectrically
(42:00) Roadmap and Future Directions
(46:00) Conclusion & Acknowledgements
CONNECT WITH ME:
Transcript
Thank you to the organizers for setting up this remarkable meeting. I'm incredibly pleased that this is happening at NIH and to also have the opportunity to speak to all of you and share some ideas. What I think I'm going to do today is introduce... I'll go back to the very beginning and ask the question of what is, in fact, cancer and introduce the idea of cancer as a kind of dissociative identity disorder of the morphogenetic intelligence. And I'll explain what all of that means.
Let's ask a kind of a strange question. Why is cancer not a problem in the robotics field? So what you can see here is that we have now fairly complex engineering structures, very functional, doing all sorts of interesting things. But one thing that this field does not have is a problem that is similar to cancer. So why is that? And what I want to point out is that, at least for now, the major type of architecture that we deal with in engineering is very flat. In other words, the large-scale system as a whole might have various problem-solving capacities, but they're made of parts without agendas. The parts themselves are not smart.
Biology is not like that at all. What we have in biology is what I call a multi-scale competency architecture. That is, we are made of a material that all the way down from the genetic components to the subcellular organelles, tissues, organs, and all the way through groups and colonies and so on, have various competencies at every level. So at each level, this is not just a structural set of nested dolls, but actually at each level, there are competencies of solving certain problems in different spaces. And this has massive implications both for the remarkable plasticity and robustness of life, but also for certain failure modes. And so all of these are able to pursue specific states with various degrees of competency and ingenuity in different spaces that I will mention momentarily.
Now, the classic developmental biologist, Wilhelm Roux, wrote this amazing monograph called The Struggle of the Parts. And what he was talking about is precisely this idea that we are made of components that cooperate in some cases, but also have their own individual identities and that keeping the organism together, the sort of Ship of Theseus that is our body, where cells and molecular materials must be constantly replaced and fit into the functional architecture, a lot of these parts have autonomy. And that autonomy must be harnessed in order for global health to take place.
So I'll just show you a couple of examples. This is a flatworm called a planarian. And one thing you don't see here is a little tube called the pharynx that they use to eat. This pharynx is quiescent most of the time, but if you liberate that pharynx from the rest of the animal, here they are, and you will see that they have their own ability to move. They have their own little set of behaviors. This is a piece of liver that they're going to try to eat. Now, of course, they're just tubes, so what they end up doing is burrowing through it and the food comes out the other end. But basically, these components here have independent activity that are normally harnessed by the rest of the body towards specific behaviors.
Now, here's another example. This little creature here, if you see this for the first time, you might think that this is something that we got from the bottom of a pond somewhere. Actually, what it's made of are adult human tracheal epithelial cells. This is what we call an anthropod. And Gizem Gumusci in her PhD work in our lab showed that if you take these out of an adult human patient under specific circumstances, they join together to be a novel kind of moving construct with all sorts of fascinating behaviors, its own transcriptome that's quite different from that of the source tissue. And they have interesting behaviors. Here they are assembling into a super bot kind of cluster where what they'll be able to do in that configuration is heal neural scratch wounds. So this is a scratch wound through some neural tissue and these anthropods will sit there and try to heal the gap.
So we see that we are made of a material that has all kinds of interesting and often unexpected competencies. And those project into different spaces. So we as humans are pretty primed to observe intelligent behavior in three-dimensional space. So that's medium-sized objects moving at medium speeds in 3D space like birds and primates and so on, and maybe an octopus or a whale or something. But there are all of these other spaces. There's the space of possible transcriptional states. There's the space of physiological states and the space of possible anatomical configurations, which is the one we'll talk most about today.
In all of these spaces that are hard for us to visualize, we don't normally think of this as behavior, as moving through these spaces as behavior, or in fact problem solving behavior. But in all of these spaces, cells and tissues can do some very sophisticated things. For example, individual cells, even gene regulatory networks, never mind the whole cell, but just small gene regulatory networks can do Pavlovian conditioning and several other kinds of learning, and so on. And so this is the sort of thing that we must understand if we are going to be able to detect, prevent and reverse defections from this coordinated activity.
And what happens in multicellularity, both during our embryogenesis and during evolution, is a radical scale up of what we call the cognitive light cone. So the cognitive light cone is simply the size, both in space and time, of the largest set point that any kind of a system, be it cells, animals, artificial systems, whatever, any kind of a set point that these agents can pursue in a kind of homeostatic process. So the bigger the ability of the system to have a memory going backwards in time, anticipation forwards in time, spatial extent, the right kind of connectivity between sub-units allows them to expand their cognitive light cone and use much larger states as the goals that they pursue in these homeostatic kinds of loops.
And what I mean by that is this. This is a single cell, so this happens to be a free-living organism known as a lacrymaria. And you can see it's here. There's no brain. There's no nervous system. It's handling all of its local goals within the scale of a single cell. So the cognitive light cone of this system is roughly the size of the cell. Basically, it's doing whatever it needs to do to manage the conditions inside of the cell. The rest of the environment, it doesn't matter. It will dump entropy into the environment. It will eat what it wants. It will go where it wants.
What happens during embryogenesis and through evolution is that cells, individual cells get together and instead of pursuing very tiny physiological, metabolic, and so on goals, they end up taking on these massive grandiose construction projects. So here is a group of cells making a salamander limb. And what happens is if you deviate them from this state, meaning you cut the limb anywhere along this axis, they will very rapidly spring into action and rebuild, and they stop when it's done. So what's happening here is this kind of anatomical homeostasis where all of these cells are perfectly aligned on what their goal is. Their goal is to rebuild this. We know this because if we deviate them from this quiescent position, they will build it again, and then they will stop, and they will do exactly what's needed, no more, no less, to get back to this particular region of that anatomical morphospace. So by scaling up their cognitive light cones, individual cells are able to now have much larger set points as the target of these kind of homeostatic processes.
Now that system, that ability to take on larger set points by joining together has an obvious failure mode. It's inevitable. Unlike many other disease conditions which are very contingent on the details of evolution and physiology, cancer is fundamental because this process of joining together towards large scale construction projects is going to break down occasionally for a number of reasons. And that is when we see cancer.
So this happens to be human glioblastoma. But what's happened here is that these cells are not, unlike what's sometimes modeled in game theory kinds of models of cancer, I don't think these cells are any more selfish than these cells. They have smaller selves. So what's happened here is that the boundary between self and world, the size of the things you are actively trying to manage, here it's quite large. Here it's shrunk down back to the scale of individual cells. At this point, the rest of the body is just external environment to them. All they're trying to do, like their ancient unicellular ancestors, is manage their own internal state. And there's a lot of work that's been done on this kind of atavistic theory of cancer. But I think what we need to do is pay attention to the policies that normally allow this kind of scale-up to see if we can reverse this process.
So what is this cognitive glue? Individual cells have these little homeostatic loops like this. There's a particular scalar that they might be managing. Let's say pH or hunger level or something like that. And then by joining together... So here's this homeostatic kind of loop. You measure something, you compare it to a remembered set point, and then if the error is more than some acceptable level, then you act to reduce that error. So individual cells are able to measure and control fairly small things. By the time you have a large group of cells, let's say a tissue or an early embryo, you're able to do things like this.
So what's critical here are these connections. The policies and the information exchange that allows this tiny loop to scale up to much bigger things that can then be modeled via landscapes whether it be morphogenetic or transcriptional landscapes, pattern completions such as here when part of this information is gone, the cells will rebuild. So there's a collective there that knows what the whole thing is supposed to look like and is able to reduce the error. So it becomes really important to understand what that cognitive glue is.
And we know what it is in the brain. So in the brain, you have groups of neurons. We also are more than just a pile of neurons because there is an electrophysiological communication system that binds those neurons into memories, goals, preferences, and various other capacities that do not belong to any of the individual neurons themselves. And so in the body, we have numerous kinds of signals, chemical, biomechanical, and so on. But today, we're going to talk about my favorite, which is the bioelectric layer. And so I think it's not that the bioelectricity does everything by itself. Obviously these other things are quite important, but I think the bioelectrical layer of this communication network has some very useful and very interesting properties that make it a distinct and very attractive target for regenerative medicine.
So let's compare this to what actually happens in the brain, because I think it's a very good analogy to start with, and then we have to break some of those assumptions to really understand developmental or cancer bioelectricity. So in the brain, we have this hardware architecture that's specified genetically where you've got individual cells. In their membrane, they have ion channels. These ion channels, by virtue of letting ions in and out under specific rules, will set up a voltage potential across this membrane. That voltage potential may or may not be communicated to the neighbors through these electrical synapses known as gap junctions. And all of these, both many kinds of ion channels and certain kinds of gap junctions are themselves voltage-sensitive. So this enables a really interesting kind of historicity. Meaning, what it's going to do now is dependent on whatever the physiological state was before that, so that gives you a kind of memory immediately, and feedback loops immediately already at the level of the single cell. But of course, things get much more complex in these networks.
So that's the hardware. And what neuroscience does is study how this network supports an interesting kind of software. So the software is the physiological dynamics that operate within the brain and nervous system. Here's an example. This group made this amazing video of the electric activity of a living zebrafish brain as the fish is thinking about whatever it is that fish think about. And in neuroscience, there's this project of neural decoding. So the commitment is that all of the animal's memories, preferences, goals, and behavioral competencies are in some way encoded in the electrophysiology that you're seeing here. And so the idea is that if we could record in the living state, we record this electrophysiology, we can translate it in some way. We mine it for the patterns that are there, and we can now decode and be able to, by reading this electrophysiology, we can tell what the animal has done before, so we can read the memories. And perhaps we can tell what it's going to do later, meaning that we can understand what are the goals and the incipient behaviors of the system in three-dimensional space, as it navigates through manipulation of muscle activity, how it's going to navigate three-dimensional space.
Well, it turns out that this system, and this is a remarkable system, as we all know. But it turns out this system is incredibly ancient, evolutionarily. It actually shows up around the time of bacterial biofilms. And so pretty much every cell in your body has ion channels. Most cells have these electrical synapses or gap junctions to their neighbors. And during evolution, we've been suggesting that what's happened here is that evolution took these ancient ways of processing information in cellular networks, which originally were for processing decisions in anatomical space, so deciding what the shape of the early embryo and the final body anatomy is going to be. And it basically did two things. First of all, it pivoted into a new space. So instead of navigating anatomical space, once muscle and nerve came on the scene, it started to use that system also for navigating three-dimensional space. And of course, it sped up the time. So instead of the kind of a long-term, let's say hours-long, timescale activity here, we went to milliseconds and enabled rapid motion.
But there's a fundamental symmetry here between neuroscience and developmental biology and its disorders such as cancer. And so that means that we can take many of the tools and concepts, so both the practical tools and the conceptual ways of thinking about things, from cognitive neuroscience and see what they allow us to do in this field. And so this is some technology that Patrick McMillan has been pushing in our group, which allows you now to look at non-neural cells in real time, both in vivo and in vitro, and characterize their bioelectrics and process them via interpretation pipelines similar to what neuroscientists are doing in the brain.
So now, let's get specifically into cancer. What I've just told you is that all cells, not just neurons, communicate as electrical networks that process information that's important for cancer suppression. So the continuous essential process of taking new cells and integrating them into the anatomical structure of the body so that they can maintain healthy tissues against aging, against degenerative disease, and against carcinogenic transformation. And this leads to the hypothesis. The hypothesis that we've been working on for some years is that cancer can be detected, it should be able to be induced, and perhaps even normalized by manipulation of bioelectrical signaling that normally enables cell cooperation towards anatomical goals.
And so, to put the whole talk in one slide, what I'm going to show is that, like the brain, somatic tissues form electrical networks that make decisions about anatomy. And then we can target this control system of large-scale homeodynamics with many applications in cancer medicine.
Alright. So this leads to a suggestion, which is that perhaps what we can do is normalize cancer by rebooting a patterning program, basically by connecting cells to the set of cues that normally keep them harnessed towards some kind of morphogenesis. This would be an alternative to necessarily killing those cells, which might allow us to avoid a compensatory proliferation response, evolution of tumor resistance, and so on. Instead of trying to kill those cells, we're going to try to reconnect them to the large-scale homeostatic set points that they used to have.
So let's see how we might do that. Well, one thing that's been known for a really long time is that if you just take measurements of cells in different states, highly proliferative embryonic cells, stem cells, and also cancer cells, tend to be depolarized, okay? And quiescent to mature terminally differentiated cells tend to be hyperpolarized. Liver is an interesting exception, of course, highly regenerative, but even the mature tissue hangs out around down here with this group. So what this has suggested is a kind of axis of plasticity where resting potential might actually allow you to move from this state to this state and vice versa. And this is something that Clarence Cone first postulated back in the '70s.
And if you track more recent data, so this is our meta-analysis, if you track more recent data, the exact same cell types, which can be quite hyperpolarized in normal cells, are actually depolarized in cancer, right? So the exact cell types. So since then, there's been a large database of results showing that individual resting potential is responsible for controlling many things: cell differentiation, apoptosis, migration, changes in cell shape. But as interesting as this is at the single cell level, I actually think that the true import of bioelectricity in cancer is going to really shine at the multicellular scale as, again, as I was just talking about, the role of these potentials in being part of an electrical network that makes decisions.
So in order to study those things, we developed some tools. These include just, for example, voltage dye approaches to be able to read the information that these cells are exchanging with each other. Lots of computational modeling goes on in our lab. So this is a simulator, actually, a bioelectrical simulator made by Alexis Biatik in our center, that allows us to take voltage information like this, both the slowly and the rapidly. And you can see a mix here of very rapid dynamics, but then also slow ones, in this kind of voltage imaging that Patrick has developed. And so we are able to now simulate it to understand why the patterns change in the way that they do.
Now, of course, even more important than simply characterizing these patterns, we need functional tools. So what we've developed are ways to control the bioelectrical states without having to use older tools such as electrodes. So we don't use any electric field application. We don't use... there's no electromagnetics. There are no EM waves, nothing like that. What we are doing is targeting the native interface that the cells are already using to communicate with each other. So that would be the ion channels in the membrane and the gap junctions.
So this allows you to change individual voltage states of cells and this can be with mutations, it can be with pharmacology to open and close these channels, it can be via optogenetics, and that allows you to put down specific bioelectrical patterns. And this is more like, by manipulating the gap junctions, what you're doing is changing the topology of the electrical network, which cells talk to which other cells. So it's quite a different type of change. And then, of course, you can go a little bit downstream and look at, for example, some of the signals that are actually downstream of this, second messengers like serotonin and other neurotransmitters.
So when you do this, and I'm not going to take the time to show the many examples, I'll just show you one, that when you ask the question, "Okay, so what can you actually achieve using this? What does bioelectricity control in the large scale body besides individual cell properties? What can you actually do by controlling bioelectricity?" So this is one example. It's an old one from our lab and it's one of my favorites. This was by Vaibhav Pai and Sherry Au was discovered that one thing you can do is you can set up particular voltage states in the body that correspond roughly to endogenous patterns that dictate the position of certain organs. So if you inject specific potassium channels that set up a little voltage spot, that's very similar to the eye spot in, probably, Dani Adams when she gives a talk will show you the electric face in those eye spots. But you can introduce that anywhere else in the body by injecting this RNA. And when you do that, it makes eyes, and it can make eyes all over the place, even in locations that the developmental biology textbook will tell you are not competent to induce eyes. So even outside the anterior neuroectoderm, you can still form eyes if you use the right prompt.
So this bioelectric state, if you section these eyes, they can have normal, all the normal components, so the retina, the optic nerve, the lens, all of that stuff. And there's a couple of interesting things about this. First of all, it shows that the bioelectric state is instructive. So this is not just about causing defects in existing structures. You can call up entire new structures elsewhere in the body. It's extremely modular. So this is a very low information content stimulus. We don't tell them how to build eyes. We have no idea how to build an eye with all of its many components and all of the genes that have to come on and all of the spatial relationships. All we have to do is say, "Build an eye here." It's a very high level, almost a subroutine call that takes tissues that, in this case, for example, were going to be gut, and it says to them, "Build an eye." And everything else is handled below that. In other words, the shape of the eye, we don't need to micromanage that.
The other interesting thing about it is that actually again, as I started out talking about cellular competencies that we can take advantage of, observe here that this is a section of a lens is sitting out in a tail of a tadpole somewhere. And the blue cells are the ones that we actually injected. But all of these other cells that are participating in this morphogenesis were never targeted by us. What's happened here is that we tell these blue cells, "Build an eye." They determine that there's not enough of them to actually complete the task and they recruit, through some sort of secondary instruction, they recruit their neighbors to help them do this. So that ability, other collective intelligences do this, of course, ants and termites and so on do exactly that. But the idea is that that ability to scale your influence to the needed task and to communicate to the other cells that we as the regenerative medicine workers don't need to worry about is already in the material. That is part of the competency of the material that you can harness when you use this bioelectrical interface.
Okay. So the specific claims here are basically these, that if bioelectrical signals were to be important for cancer, you can expect a few things. First of all, there should be molecular data that implicate channel and pump proteins in cancer. And then we should be able to use this as a diagnostic tool for incipient tumorigenesis. We should be able to induce these kind of phenotypes in the absence of, for example, DNA damage by modulating the VMEM signals. And hopefully we can actually suppress or normalize them.
So let's look at those one at a time. So actually, there have been numerous ion channels and this is a fairly old list now. There are many more now that are known as either oncogenes or tumor suppressors. And so, for example, from the work of Emily Bates and many other people, lots of channels have now been identified that contribute to this phenotype. Now one thing that we have to note is that this is a serious underestimate because actually because of the properties of these physiological networks that are very robust where many, many different channel genes can compensate for each other, what we see from genetic knockdowns, for example, in genetic screens are probably an underestimate of what's actually there. You knock out one thing, it often happens that other things, other channels will actually take their role in the physiological circuit and we can miss that. So I'm sure there are many others. But this is a good starting list.
And you can see for many of these things, if you look through the GEO database, that through the progression through cancer, there are significant changes that you can track. Already you can track this just in the bioinformatics. But again, I think that the bioinformatics is really a drastic underestimate of what's actually going on because these channels open and close post-translationally. That is, you cannot infer from the presence of the gene or the RNA or the protein, you cannot infer directly what the physiological state is going to be. You can't guess the bioelectrics just from their presence, at least not reliably so. And that means that we have to go beyond the existing omics profiling of molecular entities to physiomics and the actual functional parameter which is the distribution of voltage.
So this is an example of the diagnostics modality. This was worked out by my grad student years ago, Brook Chernet. And what he found is that by injecting a variety of human and other oncogenes, they will form tumors. The tumors will eventually... This is in the tadpole model. The metastasis process where they'll spread. But before the tumor, even before the tumor becomes histologically apparent, we can already see using a voltage sensor, the fluorescent dye technique, we can already see that the region where the tumor is going to be already has an abnormal voltage potential.
So we start with the voltage monitoring techniques that Danny Adams worked out in our group a while back, for looking at embryos and looking at how these voltage gradients change during embryogenesis. And we can take that into tissue and organ maintenance and ask, "What does it look like when these cells contract their cognitive light cones and basically start treating the rest of the body as just external environment?" And you can already see that they depolarize, and in fact not just the tumor itself but there's actually plenty of other cells out here that are going to have this aberrant behavior. And so this is an obvious beginning of a diagnostic technology.
And I point out this is an artist's rendering so we don't actually have this working yet, but we're working on something like this where the idea is that you should be able to in real time using some sort of augmented reality goggles, and this is actually... Surgeons already use this for various other indicators, but we should be able to have a voltage indicating channel in there where the surgeon is going to be able to see where the margins are, how much they need to take, what are the straggler cells they might need to get, and so on. So the idea is of using real time bioelectric voltage imaging. Ultimately I think this technology is coming and we'll be able to use it for diagnostic applications and for surgical applications.
The second prediction we talked about is the ability to induce a cancer-like state in the absence of carcinogens, oncogenes, DNA damage and so on. And this was found by Doug Blakiston and Maria Lobikin in my group where what we were able to do is take animals with normal melanocytes, and so these little pigment cells here are normal melanocytes in this frog embryo. And by targeting a specific ion channel, in this case we targeted a chloride, a glycine-gated chloride channel. By targeting that chloride channel, we disrupted the ability of a population of cells... This is actually a specific population. We call them instructor cells because what they do is they talk to the melanocytes. They normally keep the melanocytes in order and doing this. When you silence the ability of those instructor cells to communicate, what happens is that the melanocytes, the brakes are off, these melanocytes and they go completely crazy. They over-proliferate. They enter these regions where they shouldn't be, and you can see the difference. This is MMP-dependent. Their migration is MMP-dependent.
And if you look through a section what's happening here... So these are sections of the neural tube here. These are sections, anterior and posterior through tadpoles. This is what normal melanocytes are supposed to look like, little round things. There's quite few of them. This is what ivermectin, which is an opener of these chloride channels... This is what ivermectin treated animals look like. The melanocytes, first of all, there's way too many of them. Second of all, they have this crazy projection that, you know, they're long. They're almost like neurons here. This is what they normally look like. Once those bioelectrical signals from the instructor cells are squelched, they transform and they do this, and then they start digging into all the other tissue. So here they are digging into the brain. Here they are in the neural tube. The blood vessels here, right? And so what happens is that basically an animal-wide metastatic melanoma phenotype, which you can pick up initially, there are no oncogenes. There is no DNA damage. All there is is temporary interference with the normal bioelectrical signals that keep order. But once this all starts, they turn on markers of epithelial-mesenchymal transition and all the other things you would see in cancer. But that comes later. The first event is physiological, not genetic.
And what's interesting here is that, again, I'll just point out, the effect is not cell-autonomous. So this is a cross-section through that tadpole. These are the cells right here. This is an injected dominant negative ion channel. These are the cells here that we target, but the melanocytes that change their phenotype are at some distance. And we've actually worked out how this works, and that takes place through, again, a serotonergic kind of signaling pathway. But what's apparent here is that it isn't the voltage of the cells themselves that determine what they're going to do. It's a voltage change in the environment as a physiological switch towards metastatic behavior. And in fact, it's not even necessarily the microenvironment because in vivo, in a tadpole, it only takes a few cells, a few targeted cells to kickstart this, and they can be quite far away, in fact, on the other side of the animal. And we have all kinds of data looking at where you can inject and where this property turns on. And the entire tadpole can be transformed by just a small number of cells that have an aberrant electrophysiological signature at one location. So this is a story of how you can kickstart a cancer process with the physiology.
So that gives us some hope. That gives us the idea that we should be able to reverse this, prevent and maybe even reverse this. So we tried this, and this, again, is the work of largely of Brook Chernet, when he was a PhD student and a postdoc in our group. And what he did was, once again, he would inject oncogenes into various blastomeres. They are labeled with a tomato fluorescent tracer, so here you see it. We tried all kinds of oncogenes, so it's a really nasty, KRAS mutations, p53 dominant-negative, p53 mutations, all sorts of things that cause these tumor structures. And what we see is that they're quite efficient at causing this. But if you co-inject an ion channel, that will prevent the cells from depolarizing.
So one of the first things that oncogenes do is they disconnect cells, and this has been known since the '80s from mammalian cell studies, that oncogenes tend to disconnect the expressing cell from their neighbors, and this happens through a depolarization. And so if you prevent, artificially prevent that depolarization, then... So this is the same animal. Here's what happens. The oncoprotein is still blazingly expressed. It's, in fact, it's all over the place, but it's very strong at the site of injection. We don't repair the mutation. We don't kill the mutated cells. They're still here. You can see them. But there's no tumor, and the tumor incidence is greatly suppressed, because these cells, despite the genetic defect that they have, they remain physically connected to their neighbors. And so instead of crawling off and doing their own thing, they continue to operate as part of that network, which has large-scale set points. It continues to make skin, muscle, whatever other organs. And so this tells you that you can override... Much like some of our work on birth defects, it shows that some hardware issues, such as the really a dominant mutation in this protein, some hardware issues can be overridden, quote-unquote, "in software" by manipulating the voltage, not repairing the original mutation.
And this also, Brook and Danny did some work showing that this can actually also happen via optogenetics. It doesn't have to be any one modality. It really is the voltage pattern that does it, and you can trigger this effect with light that turns these channels on and off.
The role of the environment in this is really important and it goes against this mainstream idea that these phenotypes are very tightly linked to the genetics and ultimately to the clonal expansion of one particular set of mutations.
Here's a tadpole. And this tadpole can have a tumor or an ectopic eye, right? And what sets the difference between having a tumor or an eye is the amount of sodium that's getting in. This is following Mustafa's work, which I'm sure he'll talk about on this amazing NAV channel. But basically, how much sodium is in your medium and how much of it is getting in, and some other subtleties, will really switch between two radically different outcomes: an ectopic organ or, in fact, a mutant.
So we have to start to understand the primary role of the physiology, the collective decision-making in this process, and see these ion channels as a really important interface to that process.
Our roadmap looks roughly like this. What we would like to do is scrape the existing profiling data on what channels and pumps and gap junctions exist in various human tissues. That defines the interface. That shows us which are the control knobs that we can use through extensive physiomics, which we are just now getting started on. These data largely do not exist except in a few cases, in a few model systems. But this really needs to be done in a very coherent way.
We can use simulators such as the original BETSY platform made by Alexis. This is Danny's electric face preview. So a pattern like this where we say, "Okay, this is the pattern that we want." And we can simulate which of these channels and pumps needs to be opened or closed. And then that enables us to choose the specific reagents to then get what we want. There's a review here that you can take a look at of existing links between various ion channels and the cancer phenotype that can also be used by this and other manual strategies.
So the cancer pipeline looks... and you can start to play with this. This is online, of course. It's not remotely finished yet, but there are pieces of it already here where you can start to choose specific tissues and specific cell types. It will tell you what are the knobs that you have to play with, so what are the channels and pumps that are there. And using the simulator to ask if we know what electric state we want, which of these do we need to trigger? This is collaborative work with Jack Tuszynski, and the software is built by Philip Winter.
We can then begin to choose electroceuticals, which are either existing or novel drugs targeting ion channels that can have specific effects predicted by this computational platform.
Our latest work, and this is some work by Juanita Matthews and there's much more coming shortly, is now moving all of this from the frog model into human cells. First in 2D culture and now in cancer spheroids, and eventually of course in vivo. We're starting to look at, in this case, glioblastoma. We're also looking at colon cancer and breast cancer. There's some really nice data on not only affecting the individual cell behaviors but even looking for signatures of normalization. The idea is that when we use these drugs, not only do these cells stop many of their cancer-like behaviors, but they actually start to turn on some markers of normal multicellular tissues.
Just a couple of things left to look into the future. I've made the claim that cancer is not just a genetic disease but actually a disorder of the scaling of cellular competencies in navigating anatomical space. I think that's fundamentally an important way to look at this problem. I've suggested that bioelectric properties can be used to detect, induce, and normalize neoplastic cell behavior. We now know that the behavior of these electric circuits can be modified in useful ways, just like we do in the nervous system with stimuli. That is not with hardware rewiring necessarily, but with various kinds of pharmacological, physiological, and probably many other kinds of stimuli.
I think the future involves pharmacological, optical, and other strategies guided by computational simulation platforms. One of the important things that needs to happen in the future is the knowledge of which bioelectrical states occur in different disease states and what are the healthy states that we want.
Much like the history of imaging, this is what Pluto looked like in '96 and this is what it looked like a few years ago, and who knows what they could do now. The idea is that imaging technologies are very important. The breakthroughs that we had made with Danny's work back in the day started to get for the first time an actual video. The first pictures of voltage in embryos were done by Thorleif Thorlin around 2000 in Ken Robinson's lab. But then the first time-lapse movies of actually watching all the cells interact with each other electrically and watching the patterns such as this electric face that Danny will show, which presages where and determines actually where all the organs such as the eye and so on are going to be.
We can now, people like Patrick McMillan in our group are working on novel ways to track in real time many different parameters: cytoskeletal voltage, and so on. This is what that same region now looks like. So we can start to get much more complex patterns, and hopefully get enough data here to deploy state-of-the-art machine learning tools to try to infer these patterns and apply all kinds of interesting metrics from computational neuroscience to understand how to manage this behavior.
I think in the future, what I would like to see is better technology and better data for developing these physiological signatures. We are working towards control methods in mammals because we'd like to move this into patients soon. The big idea is basically cracking this bioelectric code and using normalization via electroceuticals as stimuli to the cellular collective guided by computational tools.
This is just a reminder for all of us. Mostafa Jamgoz and I edit the Bioelectricity journal, so if you have any papers that are forthcoming on any of this, I encourage you to submit to our journal.
What I'd like to do at this point is thank all of you for listening and thank the people who did the work. Juanita Matthews is doing all of the human cancer bioelectricity in our group, and Patrick McMillan is studying the bioelectrics of cell collectivity and also developing ways to understand how these electric properties relate to single cell versus group cell behavior. Brooke and Maria here did all of the early in vivo work on the cancer that I just showed you. This is Gizem Gumuskaya who did the Anthrobots, and that's a model that we are going to be using in the cancer field shortly. Dani Adams and her pivotal early work on looking at bioelectric patterns in real embryos. Lots of other students, lots of support staff without whom we couldn't have done this work. Many collaborators. Here are some funders that supported some of this work.
I need to do a disclosure. Astonishing Labs is a company that supports a lot of this work in our group, and we're moving forward together towards various hopefully therapeutic avenues in patients.
The last thing I'll just pitch here, kind of the very end of this, is this idea that I really think the way to think about bioelectricity is not just another piece of biophysics that we micromanage the way we do with transcription factors and signaling pathways. I think when we study bioelectricity, what we're really looking at is a communications interface. It's an interface to the root of the problem, which is the boundary that active agents set between themselves and the outside world.
This is pulling back beyond cancer; this is a larger view of how I see biomedicine developing. Whereas currently most of the progress has been around these bottom-up kinds of interventions, I think there is massive room for top-down approaches. Many of these things we and other people are doing in their groups are looking at behavior shaping and taking advantage of the various competencies of cells and tissues, various kinds of agential implants such as Anthrobots for healing in the body, different kinds of morphaceuticals, and electroceuticals.
The cancer problem, I think, is going to be powerfully addressed in some of these ways if we use new techniques, both in terms of AI and powerful concepts that are being developed in computational neuroscience, to address the cancer problem in this context. Some of the details are here.
I think that's it. Again, I'll just thank all of these people who did this amazing work and I'll thank you for listening.
