Today, an international team of researchers shared an extraordinarily detailed atlas of human brain cells, mapping its staggering diversity of neurons. The atlas was published as part of a massive package of 21 papers in the journal Science, each taking complementary approaches to the same overarching questions: What cell types exist in the brain? And what makes human brains different from those of other animals?
With hundreds of billions of cells tangled together, mapping the whole brain is like trying to plot every star in the Milky Way. (The inner workings of each cell are mini worlds of their own.) But just as better telescopes make the universe clearer to astronomers, the analytical tools presented here give neuroscientists “unprecedented resolution looking at brain cells, which will open up new windows for understanding brain function,” says Andrea Beckel-Mitchener, deputy director of the US National Institutes of Health’s BRAIN Initiative, which funded the cell atlas projects.
With a comprehensive map of cell types, understanding how neurons work—and how brain disorders cause them to malfunction—is within reach. “This is a first step towards defining the cellular complexity of the brain,” says Bing Ren, a professor of cellular and molecular medicine at UC San Diego, and a lead investigator on the atlas project. “The results have been nothing but astonishing.”
This isn’t the first brain cell atlas, and it won’t be the last. But it is incredibly detailed. The 21-study collection reports the findings of the BRAIN Initiative’s last five-year funding program, BICCN (BRAIN Initiative Cell Census Network). The NIH allocated $100 million for this endeavor, aiming to catalog brain cell types in more depth than ever before. “The only other large-scale biology problem that we have thought about of this scope is the Human Genome Project,” says Beckel-Mitchener. “The cell atlas project is the biggest team science effort in neuroscience.”
Historically, it’s been nearly impossible to get a handle on the complexity of the human brain. With so many interconnected pieces, “it’s not really a single organ—it’s like a thousand organs,” says Ed Lein, a senior investigator at the Allen Institute for Brain Science who helped lead the atlas project.
“Prior to this data set, it was just a hypothesis that the brain was really complicated,” adds Amy Bernard, the director of life sciences at the Kavli Foundation, who was not involved in this project. “Now, we can see the cellular diversity and wrap our arms around the problem.”
Neuroscientists often think about the brain in terms of connections between cells, like a wiring diagram. But the brain’s wiring doesn’t say anything about what its individual units are made of. To understand what makes brain cells diverse, Lein says that neuroscientists are borrowing tricks from the genomics world.
All cells within a given brain share the same DNA, but different cells use different sets of genes, which determine what proteins each one makes. This, Lein says, “relates very strongly to all the other properties of the cell,” shaping what it looks like, how it develops, and what other cells it will connect with.
In an earlier phase of the BRAIN Initiative, scientists developed methods to create a cellular map of the mouse brain. But bringing these tools to human brains is no small feat. Our brains are about 15 times the size of a mouse’s, with a thousand times as many neurons. One major goal of this work was to expand methods used in mice to “create an atlas that tackles the problem of scale,” Lein says.
This was a massive undertaking, relying on collaboration between 250 researchers across 45 institutions worldwide. “People are familiar with big teams like this in fields like astrophysics, but it’s new in neuroscience,” says Bernard. “We took a divide-and-conquer approach,” Ren says, divvying processed tissues from three donated human brains across labs. From there, molecular biologists sequenced DNA, then passed the results off to computational biologists for analysis.
In one study led by Ren, researchers analyzed the molecular switches turning different genes off and on—the internal configuration defining what type of cell a neuron becomes—within more than a million human brain cells. They identified over 100 distinct cell types across 42 different brain regions, far more than the team expected.
With this expansive data set, the team trained deep-learning models to read long strings of genetic code and predict how noncoding sequence variants—hard-to-read chunks of DNA that don’t contain instructions for specific proteins—shape cell identity. Ren compares it to reading a book in a foreign language. “Initially, you know nothing,” Ren says. But using a dictionary built with machine-learning tools, “you can start at least making sense of words within that long string of characters.” Many of these gene sequences were indecipherable to researchers before, but their deep-learning model was able to extract hidden patterns and “learn something that our human mind hasn’t yet been able to grasp,” Ren says.
This paper brings scientists closer to being able to identify how someone’s cells work—and how they might falter—from how their genes are regulated. The researchers highlighted several cell types that appear to be strongly linked to neuropsychiatric disorders like schizophrenia and Alzheimer’s disease. They hope that by understanding the brain at this level of detail, they’ll someday be able to trace brain diseases back to their genetic roots, and find treatments that target them. This is “the holy grail for human genetics research,” says Jennifer Erwin, a molecular geneticist and neuroscientist at the Lieber Institute for Brain Development, who was not involved in this project.
While this grail is still out of reach, it’s within sight—and the BRAIN Initiative has years more research queued up and underway. This effort focused on translating methods developed for mouse brains to the brains of humans and monkeys, characterizing cell types, and figuring out what is unique to humans on a molecular level. Currently, many clinical trials fail because they can’t replicate promising results from mouse studies. With a more nuanced understanding of where the brains of mice and humans are alike, and where they’re not, scientists will be better equipped to predict whether a drug will fail in humans before getting too deep into testing.
As much as it reveals, no brain cell atlas can tell you anything about connectivity, or how neurons form networks and communicate across brain regions. Researchers first attempted to create a map of the brain’s neural fiber pathways over a decade ago with the Human Connectome Project, but much more work needs to be done to understand how these connections are formed, how they change over time, and how they generate thought and behavior.
Future BRAIN Initiative programs plan to study neural diversity across humans, but the slate of projects published today did not. Most of these studies analyzed tissue from the same three brains, all donated by neurotypical men of European ancestry. Given the time, effort, and tax dollars required to run experiments at this scale, researchers have to choose between molecular detail and human diversity. “You can either go broad or you can go deep, but you can’t do both at the same time,” Lein says.
Funding agencies like the NIH tend to prioritize the generation of new data over the reuse of existing data, but reusing this data is going to be very important. “Once data is published, it’s not dead. It’s there to be used,” says Bernard. She believes that, now that this massive atlas is online, funding should be funneled to people who want to dig into it—not just researchers who want to add to the pile. “It should be sexy to rediscover things from old data,” she says. Ren’s team made their atlas of genetic switches publicly available, hoping that scientists will mine it to fuel drug discovery, basic science, and clinical research.
These findings lay the groundwork for a new era of neuroscience, where finding personalized treatments for brain disorders is a little less impossible. “Science is somewhat incremental, but people always want to advertise it as groundbreaking,” Bernard says. “This is both.”
