A Brain Organoid Is Not a Brain

Paola Arlotta is a professor of stem cell and regenerative biology at Harvard, interested in the molecular rules that govern how the human cerebral cortex is born, differentiated, and assembled. Lex Fridman opens with an out-of-the-box question: how unlikely is a brain like ours, and how hard is it to build one through evolution? Her answer is blunt. It has happened here, on this planet, once. That simply tells you it *could* happen again. Elsewhere is only a matter of probability — and we do not know how small that probability is.

The human brain is difficult to make, she says, but the deeper problem is how little we know about the process. Most of what biology knows about mammalian brains comes from other species, especially the mouse. Put a picture of a mouse brain next to a picture of a human brain and they do not look like each other at all. Studying the mouse has taught us a great deal. It is not the same as having studied the human brain.

Asked how often she finds herself in awe of the thing she studies, Arlotta does not hesitate. Pretty much all the time. At the start of embryogenesis, by means we still do not fully understand, a structure called the neural tube literally self-assembles. It happens in an embryo. It can also happen from stem cells in a dish.

From that tube, stem-like cells give rise over time to the thousands of cell types in the brain. The striking observation, Arlotta says, is that the time it takes to make a human brain is *human* time. Gestation is almost nine months; getting from a newborn to an adult brain that can hold a conversation like this one takes another twenty years of learning.

A mouse takes about twenty days from embryo to birth, and its brain is built on that shorter clock. Take mouse stem cells, put them in a culture dish, and the brain organoid forms faster than if you take human stem cells and let them make a human brain organoid. Same experimental setup, two species, two tempos.

The developmental process itself is controlled by the speed of the species. That is not an accident. It matters for getting the brain we have. Human beings need a long postnatal stretch — from baby through puberty to adulthood — to learn what an adult brain has to learn. Time scale is not a byproduct of complexity. It is part of the build.

Fridman asks her to walk through the stages. At the beginning you have to build a brain, and a brain is made of cells. While the embryo is also making muscle, heart, and blood, it builds the brain from the neural tube: nothing but a tube of cells spanning the embryo from head to tail. Over months of human gestation, stem-like cells in that tube produce the many other building blocks — the cell types that form specific structures of the brain.

Those stem cells are not all different at the start. They begin more homogeneous, more multipotent: one cell has the potential to make many others. With time the progenitors become more heterogeneous, spinning out distinct stem-cell types and then progeny that are not stem cells at all, but specific brain cells different from the mother cell. Embryonic development is the period when you are making the bricks, stacking them into buildings and regions, and wiring connections, especially among neurons that fire action potentials.

The order of that building matters. An engineer might imagine gathering every cell type and assembling them at the end. Development is not that. Neurons are made first; supportive cells such as glia come later, because they have to assemble in specific ways. You cannot simply dump them together at the finish. As they develop next to each other, they influence one another's development. A cell made alone in a dish is not the same as a cell made in an embryo surrounded by signals from its neighbors.

Fridman, coming from artificial intelligence, says the conversation is making him realize that the beautiful thing is not only the final product but the developmental process itself. Arlotta calls it an incredibly choreographed dance that happens the same way every time each of us builds a brain. That is why developmental neurobiologists never get tired of studying it.

What is the code that drives the dance? Part of it is millions of years of evolution fine-tuning gene-expression programs so that certain cells appear at a certain time and become a certain type. Another part is mechanical. The embryo does not stay a tube. At some point the front expands to make the primordium of the brain, and the forces the cells feel — different from a week earlier — tell a cell it is being squished, or sitting at a corner, or stretched in a band, and to start producing new genes.

Mechanical, physical force shapes the fate of the cell as well. It is not only chemical. For someone used to thinking of information as signals, sequences, or electricity, the claim is worth sitting with. Shape and pressure are not the background against which a program runs. Changing shape is how the program is rewritten.

Fridman asks how messy biological systems look compared with the machines humans build. Arlotta says they are much more prone to taking alternative routes. That difference becomes sharp when you contrast printing a brain with developing one.

Given the same building blocks, you could in principle print a brain the same way every time. That final brain may not work the way a brain built during development does, because the very cells you are using developed in a completely different environment — not the environment of the brain. By definition they are going to be different.

Watch development in a dish and you see organoids go wrong, one different from the next. Think about an embryo: it almost always goes right. Every time a baby is born, with very few exceptions, the brain is like the next baby's. In a dish you do not even develop a brain; you develop something much simpler, with more options for building things differently. Evolution, Arlotta says, has played a really tight game in vivo.

Fridman wonders whether the build is a dictatorship — some high-level mechanism that kills a cell that assembled itself the wrong way — or a strongly distributed process. There are a lot of possibilities, she says. A lizard brain, a chicken brain, and a human brain are each different, and still brains, built from stem cells that had pretty much the same potential. Evolution built different brains because they serve the purpose and well-being of that species. Nobody knows the entire code of development. We know bits of very specific aspects: which genes make a cell type, how two cells interact. The controlled entirety of it is what is mind-blowing.

Building blocks keep being added all the way through birth and even postnatally. Myelin is the insulation built around the cables of neurons so electricity can travel fast down axons. A six-year-old has barely started making the mature oligodendrocytes that will wrap axons, and the process continues, Arlotta says, until we are about twenty-five or thirty. Maturation, tweaking, and addition keep going, including in response to what we do.

The textbook story is right as far as it goes. Myelin insulates stretches of axon; naked gaps remain; the impulse jumps from gap to gap and therefore goes faster. The canonical inference was that a big, complex nervous system needs a lot of myelin, because you want information to get quickly from point A to point B.

A few years before this conversation — about five years, she estimates — her group found the opposite on some of the newest neurons humans and other primates have, at the top of the cerebral cortex, the cells that should be doing some of our most complex work. Those axons have very little myelin. A little piece here, a long stretch, another chunk there; some have none at all.

Looking at it as an electrical engineer might, she first thought evolution just had not finished: give it another few million years and these neurons would be heavily myelinated too. She now thinks the opposite. Less myelin may be the future of the brain. Less insulation means more flexibility in what you do with your axons, and therefore more complicated and unpredictable functions. Timing itself encodes information. If myelin can move, the same axon can mean different things at different moments. That is useful if you are trying to come up with a new capability.

How much of this astonishing capacity is hardware, and how much is nurture? Both, Arlotta says. We are born with most of the cells and most of the structures. Then we have twenty years of interacting with the environment. The brain that genetic cues built, perfectly or imperfectly, uses that stretch to incorporate experience into its further maturation. Different childhoods, schools, and things learned leave brains a little different, and we behave differently because of that.

We are born with a plastic brain: one able to change in response to stimuli, including sensory ones. Some of the most illuminating studies, she says, looked at sensory organs that did not work. If you are born with eyes that do not work, the piece of cortex that would normally process vision develops differently after birth, and it might be used to do something else. That is plasticity at its most extreme — not software on fixed circuits, but circuits rewiring in hardware.

Before she explains the tool, Arlotta makes a request: please keep this in the video. A brain organoid is not the same as a brain. It is a cellular system you can develop in a culture dish from stem cells, one that will mimic some aspects of brain development, but not all of it. At maximum they become about four to five millimeters in diameter — the size of a tiny raisin, or perhaps an apple seed. They are much simpler than our brain. They are also the only system where we can literally watch a process of human brain development unfold and study it.

We cannot study the actual development of the human brain because it all happens in utero. We will never have access to that process. Organoids are the next best thing: stem cells coaxed into starting neural-tube formation, then producing many of the cell types present in the brain. You can watch and study that. You can also think about diseases where development does not proceed properly — the many kinds of neurodevelopmental disease, the many types of autism spectrum disorder.

Take a blood or skin sample from a patient, make a stem cell, and watch formation of a brain organoid with that person's genetic code in it. You can ask what the code is doing to some aspects of development. For the first time, she says, you may come to answers like which cells are involved in autism. Reproducibility is the catch. There is much more variability in building organoids than in building a brain. Most of us, as babies, have brains that look a lot like each other's. That is the magic the embryo does when it builds a brain in the context of a body.

Scientists do not have the full code, so the organoid partly builds itself — some structures the cells know how to make — and partly depends on factors added to the media, mostly learned from the mouse and very limited. The product is reductionist. It mimics early events of development and, as of this conversation, does not build very complex anatomy. You can grow fifty to a hundred organoids in one bioreactor. The problem is not quantity. It is growing them so they are reproducible enough to study a biological process, because with too much variability you never know if what you see is the exception or the rule.

Different brain regions have different cell types, and scientists have grown organoids that mimic aspects of those regions, including cortex and striatum. Arlotta's group is especially interested in cerebral cortex, the part that most truly makes us human. If you keep cortical organoids growing for months, the way a real embryo does, many different types of cortical neurons appear, and at some point so do astrocytes, the glial cells of cortex.

Astrocytes are not neurons. They support neurons, and they play more active roles, including in making synapses, the points of contact between two nerve cells. Neurons themselves, mouse or human, have been cultured for a long time; even a young neuron that finds another cell to talk to will form a synapse. Put several neurons together and chances are they will connect. An organoid is still not a brain, but it can mimic properties of cortex well enough to study how this diversity of cell types is made, what they need, and what goes wrong.

If the stem cell came from a patient with a neurodevelopmental disease, you can watch, for the first time, what may have gone wrong years before, when that child's own brain was being made. Arlotta calls it a little tiny, rudimentary window into the past. Today we barely know which cell types or even which brain regions are affected in these diseases. The experimental system lets you ask which cells were hit, which of the many molecules that control development really messed up, and what the final product is: a weaker circuit, a missing cell type, something else. Then treatment and care can be informed rather than based only on the current diagnostic label.

Detecting the abnormal is hard because things can go wrong at multiple levels: a cell that is built but does not work, a cell that is never born, cells that interact differently. Single-cell technology that did not exist even five years earlier can read the molecular picture of one cell in a sea of cells with high precision, comparing gene expression between a control and a patient. Organoids also let you ask functional questions: how neurons communicate, how they respond to a stimulus, where the abnormality sits.

Fridman romanticizes a thousand-year horizon: if you are building something with properties of a brain, how far is the understanding? Arlotta's opinion is that even ten or twenty years before this 2019 conversation, we did not have technology to experimentally investigate human brain development. The field had studied many other organisms. Now stem-cell biology has changed the map. Figuring out how to turn a skin cell into an embryonic stem cell — work that won a Nobel Prize — let that cell begin development again, and started the field of using stem cells to build organs.

Brain organoids are a younger field, moving fast in labs with an ethical framework, building on solid science about what is real and what is not. Single-cell methods can now profile millions of cells at high resolution, including inside an organoid. It is not going to take a thousand years. She sees exponential growth enabled by tools we did not have, and a transformation we did not see in the prior millennium.

Building a whole human brain from a stem cell is not the goal for the majority of serious scientists in the field, she says, because you do not have to build the whole thing to understand development or disease. Properties that emerge from the real human brain — mind, perhaps consciousness, intelligence — are not understood even there, and therefore cannot be measured or studied in an organoid. The work may lead, many years on, to better neural circuits built from real understanding. She does not think we are close to a human brain. We will be building a worse version that may still have powerful properties, which is exactly why the science has to happen inside an ethical framework that scientists do not decide alone.

Every discussion of the ethics, Arlotta says, needs to be based on actual data from the models we have today and the ones we will have tomorrow. It is a continuous conversation, not something you decide once. The group in the room should not be only scientists. Bioethicists, lawyers, philosophers, psychiatrists, and psychologists should help decide, as a society, what we should and should not do.

She also claims a moral responsibility as a scientist. Being able to watch, study, and treat with drugs the developmental process of a neuropsychiatric patient's own brain, with that person's genetics, is transformative in a way that was impossible five years earlier. You could use the organoid as a cellular platform to screen drugs or to fix a process. We could not do it then. We can now. Should we not? She does not see a downside in the work itself.

Fridman asks for the devil's advocate. The worry, she agrees, is enabling a process of brain formation that could be misused, or that could show properties we do not want to see in a tissue. Today, she repeats, that is not the issue: the system is simple enough, and different enough from the actual brain, that you gain dramatically from the science. Because it *is* brain tissue, there is still an obligation to keep considering it, with disease and human betterment on the same plate.

Scientists should have a voice because they bring the reality of data, and they should be good at communicating that data to non-scientists. A lot of fights happen just because of how a topic is described. Call the same cellular system a brain organoid, or call it a human brain, and the reaction is very different. Scientists and journalists alike, she says, need to be careful that the debate is a real debate, informed by real data. That is all she is asking. Fridman notes the same pattern in autonomous vehicles: language can flip how people feel about the right way forward.

Raising two children, Arlotta says she has learned more than she could ever love: that children have amazingly plastic minds, that parents have a responsibility to foster growth in healthy ways, keep them curious and adventurous, and not raise them in fear, while also respecting who they are, which comes in part from genetics. Her children are very different from each other despite coming from the same two parents. They are genetically different. Each mixing of gametes creates a new mix of genes.

The genes you have play that orchestrated symphony of development; different genes play it slightly differently. It is like the same piece of music with a different orchestra and a different character. It is still a work by the same author, but it comes out differently if it is played by a high-school orchestra instead of La Scala in Milan. You are born, superficially, with the same brain: the same cell types, similar patterns of connectivity. The properties of the cells, and how those cells then react to the world, are also shaped by who you are genetically. Both the genetics and the nurturing are needed.

Studying the brain has made it almost impossible for her to separate what people do from how a part of the brain is working. A teenager being funny and impossible in the way they think sends her immediately to critical periods of plasticity, synapses being eliminated here and there, a brain that is simply confused. That gives her a different take on the behavior. She also looks at humanity, is amazed by the ideas we can come up with, and cannot stop thinking about how the brain is continuing to evolve. It is not over. She thinks about the next brain.

Will artificial intelligence be that next evolutionary step? In a way, she thinks it will, not necessarily as a synthetic implant, but because we evolve in the world we created. She half-jokes that someday a region of cortex will be attuned to the cell phone. Modern biology already shows cortex adapting to what you must do: precise finger movement enlarges a corresponding area; a world that must be seen at a distance gets a bigger visual cortex. Put on virtual-reality goggles and the brain places you in that ocean, with real emotion, without thousands of years of adaptation. Recent evolution has put a lot of chips on cortex, and the evolution of cortex *is* plasticity. Technology will be integrated, she says, whether we want it or not.