A conversation with Prof. Michael Levin about his work on the “bioelectric code” and its role in biological development, limb regeneration and cancer — with some bonus discussion about the role of electric fields in consciousness
Michael Levin has been studying bioelectricity since the 1980s. He got hooked on the stuff after he read The Body Electric: Electromagnetism and the Foundation of Life, by the late electrophysiologist and surgeon Robert Becker. Becker spent decades researching the role of the body’s electric fields in development, wound healing, and limb regrowth. It’s a fascinating deep dive into how, indeed, the body is electric through and through — despite our inability to see or sense most of these fields with our unaided senses (visible light is a tiny part of the full spectrum, most of which we can’t sense directly).
But Becker’s work was far from complete. Levin has been working on “cracking the bioelectric code,” as a 2013 paper of his put it, ever since. His team at Tufts University develops new molecular-genetic and conceptual tools to probe large-scale information processing in regeneration, embryo development, and cancer suppression — all mediated by bioelectric fields in varying degrees. This work involves examining, for example, how frogs, which normally don’t regenerate whole limbs (like salamanders and some other creates can) can regrow limbs, repair their brains and spinal cords, or normalize tumors with the help of “electroceuticals” — therapies that target their endogenous bioelectric circuits instead of or together with chemical-based therapies.
What we are learning from this work is that bioelectric fields are more powerful than suspected and perform many roles in the human body and all animal bodies. Nature seems to have figured out that electric fields, similar to the role they play in human-created machines, can power a wide array of processes essential to life and, more generally, act as robust information carriers at various levels of granularity throughout the body and brain.
Endogenous electric networks may even have a starring role in the dynamics of consciousness. Life seems to be tied to bioelectricity at every level. “Evolution,” Levin has said, “really did discover how good the biophysics of electricity is for computing and processing information in non-neural tissues,” the many kinds of cells that make up the body.
I interviewed Levin via email. Nautilus’s Brian Gallagher assisted in editing the interview.
What role does electricity play in what cells choose to do?
Consider the tadpole. To become frogs, tadpoles have to rearrange their faces during metamorphosis. It used to be thought that these movements were hardcoded, but our “Picasso” tadpoles — which have all the organs in the wrong places — showed otherwise. Amazingly, they still largely became normal frogs! All the organs move around, in abnormal paths, to end up in a correct frog configuration, despite their weird starting positions. The final frog face pattern is stored as a setpoint, like your thermostat has — in effect, a pattern memory, enables the cells to keep moving and remodeling to progressively reduce the error between the current novel or unexpected state and the target morphology. This ability to know what the current shape is, compare it to a stored memory setpoint, issue commands to subunits to try to reduce the distance between current and target states, and stop when a “normal frog face” is completed, is a computational process. Bioelectrics is one important layer of such computations, uniquely excellent at implementing memory and long-range homeostatic mechanisms, which is why evolution uses it in brains.
Where is the final frog face pattern stored, in a bioelectric circuit or somewhere else?
Yes, in a multicellular electric circuit, spatial patterns can be stored in a way that is likely very similar to how behavioral memories (of seeing a specific shape for example) are stored in a neuronal network.
How do multicellular electric circuits store pattern memories when it is pattern memories that explain the development of the multicellular electric circuits? This seems kind of circular or chicken-and-eggy.
Well, first it’s absolutely a chicken and egg issue — everything in biology is. Bioelectricity and genetics work in a cycle, with genes coding for the production of ion channel and gap junction proteins, which then act in cellular and tissue-level bioelectric circuits; meanwhile, the activity of these bioelectric circuits eventually control downstream gene expression. The question isn’t what’s the “ultimate upstream” cause (because everything is always caused by something else), but which nodes in the causal network — among bioelectric states, specific genes, biomechanical forces, and such — are the most tractable for controlling the network. We’ve shown that the bioelectric state is an amazingly potent and convenient control node, able to trigger complex, self-limiting morphogenetic “subroutines” via relatively simple triggers.
So how do these bioelectric circuit patterns arise?
These patterns are partly emergent from spontaneous symmetry breaking and amplification even in cell fields of totally identical ion channel expressing cells — for example, if one region’s ion channels open a little bit due to small random differences in local ion levels, they will cause the cell voltage to change which in turn opens those channels further, and so on — in a positive feedback loop, which could spread as activation or inhibition to other cells, triggering a pattern to appear. So called “Turing patterns”, responsible for example for the spots and stripes on zebras, fish, and leopards, are an example of this kind of phenomenon in chemical networks. Also, bioelectric patterns can come from differential expression of ion channels across the tissue. Ion channels and gap junctions (electrical synapses) are the cellular hardware that enables spontaneous symmetry breaking and pre-existing signals to establish large-scale electric circuits across tissue.
A good analogy is this. If you have a list of electronic parts (that’s the genome) and you connect them and turn on the juice, something happens — they work as a calculator, for example. The pattern of current through the device, how the whole thing processes information and performs specific algorithms and computes — none of this is specified in the parts list itself, but the parts embody and constrain some laws of physics that implement these dynamics, which could be really simple and fixed, or could be reprogrammable.
How promising is manipulating the bioelectric code as a possible therapeutic pathway for cancer?
Very promising. We’ve published a bunch of papers in frog models showing a few things. You can detect cell defections from the bioelectric network, or precancer, using voltage-sensitive fluorescent dyes. You can induce full-on metastatic melanoma — a kind of skin cancer — in perfectly normal animals with no carcinogens or nasty chemicals that break DNA. And best of all, you can use misexpression of ion channels, for example light-gated ones (a.k.a., optogenetics), to normalize existing tumors or prevent them from forming.
We are currently moving this work to human clinical models. The key will be to learn to efficiently reconnect cells to the bioelectric cues that propagate across normal tissues. This allows cells to regain their multicellular cooperation (being part of a cellular swarm that actively works to build or maintain a complex tissue or organ). It’s an important complement to trying to kill cancer cells with chemotherapy.
You’ve spoken about the body’s electric fields possessing some form of intelligence. Do you stop there, or do you suspect the body’s electric fields might also possess a type of consciousness, too?
There is definitely decision-making and intelligence in the ability of living systems to achieve the same anatomical outcome despite novel perturbations. If we had collections of robots that could continue their morphogenetic functions in novel circumstances, engineers would certainly be calling it a kind of swarm intelligence. I usually stay away from discussions of consciousness, though, because it is very hard to study. Correlates of consciousness are easier, but consciousness itself is hard to approach definitively. But I can say one thing. There are very few fundamental differences between neural networks and other tissues of bioelectrically communicating cells. If you think that consciousness in the brain is somehow a consequence of the brain’s electrical activity, then there’s no principled reason to assume that non-neural electric networks won’t underlie some primitive, basal (ancient) form of nonverbal consciousness.
Are neurons special in any sense compared to other cells? How would we know if the body’s electric fields are conscious in any way?
I’m not sure how we would test for the presence of consciousness in somatic tissues, so I don’t spend a ton of time on it, but it’s really hard to define what’s special about neurons — almost all cells do the things neurons do, just more slowly. In fact, I often ask neuroscientists to consider the following thought experiment. Imagine that back in the 1950s or 60s, the biologists came to Warren McCulloch and other fathers of connectionism (the idea that mental operations like memory and logic can be explained by the dynamics of neural-like networks) and said, “Hey, we made a mistake, thought is not in the brain. It’s in the liver.” Then the folks trying to show logic and memory in models of neural networks would have two possible moves. They could say, “Fine. Who cares? Our models are of generic and deep dynamics, and it makes no difference what cells they’re made of.” Or they could say, “Sorry, you must be mistaken. It’s got to be the brain, because neurons uniquely do X.” Most people intuitively believe that thought has to be in neural networks in the brain. But no one has yet specified what principles would be the X that backs up that answer.
That’s interesting. There’s debate about the degree to which electric fields are causally potent in consciousness, as opposed to being mere epiphenomena akin to whistles on a steam-powered train. What’s your response to people who claim that electric fields, in the context of development and regeneration, are no more than “whistles” coming from the neural activity across synapses?
I hold to a very practical, empirical definition of what’s an epiphenomenon and what’s not. It’s important to take this issue beyond philosophy and people’s conceptual preferences of whether they prefer to think in terms of particles (chemicals) or fields (bioelectric circuits) and use a testable, objective definition: does targeting it give you better prediction and control than is available with competing targets or not? By learning to manipulate bioelectric states as control knobs, we have achieved anatomical outcomes in the lab that were not reachable by prior methods focused on biochemical signaling. So bioelectric states are not epiphenomenal.
Think of a simple analogy: If you want to manually put every tiny dipole into alignment, you can, but it’s a lot of effort. If you control an electric field, they all align automatically. In physics, fields are certainly not epiphenomena — in many ways, they run the show. But of course there’s feedback between the particles and the fields. Same with computer science. Could you understand a chess-playing computer from the chemical perspective, tracking the chemistry of its transistors? Not really. Can you do better by tracking the energy flow in its circuits? Sort of. But you can do even better by tracking and learning to re-write the information that is being represented by the energy flows — the algorithm that controls what it does.
If you want to be an effective programmer, we all know the level we have to work at — the level of the algorithm. Philosophers could pretend there’s not really any such thing as an algorithm controlling things, only the chemistry of the transistors. But when it’s time to get it to do something, it’s clear that programming the algorithm is your best control strategy. Whatever the engineers are using for optimal control — that’s where the primary action is. In computer science, and in our cases of anatomical control, it’s the information flows, mediated by electric dynamics working through suitable synapses (gap junctions or transistors). As far as the needs of regenerative medicine, or indeed of evolvability of complex body forms in the first lace, bioelectric circuits are an important, instructive control element.
Some researchers have begun to catalog very interesting hierarchies of resonating/synchronized structures in the brain and body. For example, gastric neuron rhythms (0.05 Hz) correlate with theta rhythms in the brain (4–7 Hz) and it’s well-known that cardiac neurons sync up in various ways with the brain’s electrical fields. Are these various types of synchronized oscillations considered part of the bioelectric code at this point? Do we know what function these inter-organ coordinated oscillations play?
We don’t know yet. I’ve been focusing on the spatial patterns, not temporal, so it’s an open question whether those kinds of oscillations play a role in pattern control during embryogenesis, regeneration, or cancer suppression.
Why did the study of bioelectric fields in development start to languish in the 1990s?
The reason it languished is because molecular biology and genetics took off and temporarily eclipsed physiology. Everyone flocked to biochemistry because it’s so much easier: gene products like proteins and mRNAs can be studied in a dead, preserved, or fractionated sample. Bioelectrics has to be done in the living state — it’s much harder. The bioelectric gradients disappear as soon as the cell is dead. In fact, bioelectricity is an excellent marker of life itself. Also there were significant conceptual challenges to understand how bioelectricity really contributes to anatomical control, and these are only now being sorted out by importing deep concepts from physics, computer science, and cognitive science.
When will fully cracking the bioelectric code allow us to regrow limbs and organs in humans?
I wouldn’t dare to give a precise time estimate, because it depends on so many unknown factors — the science, funding, human factors. But my personal feeling is that we will see it in our lifetime. We’re working as fast as we can!
