How a Brain of 170 Billion Cells Knows Where Each Cell Belongs

The human brain contains roughly 170 billion cells. Each of those cells, at some point during development, needed to land in exactly the right place and become exactly the right type. A neuron misplaced in the developing cortex can derail the formation of an entire circuit. The precision required across such an enormous scale has long posed a fundamental question: how does the brain pull this off?

A theory published in the journal Neuron by researchers at Cold Spring Harbor Laboratory, with collaborators at Harvard University and ETH Zurich, proposes a surprisingly compact answer. The brain does not need each cell to receive a long-range positional signal pointing it to the correct destination. Instead, cells solve the positional problem by staying near their relatives.

The Problem with Chemical Gradients Alone

The prevailing model of how cells acquire positional information during development centers on chemical gradients - molecules that diffuse through tissue and create concentration profiles that cells read to determine where they are. A cell near one end of an embryo encounters high concentrations of a particular signaling molecule; a cell at the other end encounters low concentrations. The cell uses that gradient as a coordinate system.

This works well for small numbers of cells over short distances. But the brain is not small, and it is not simple. Chemical signals can only diffuse so far before they fade to noise. Asking a gradient to span the full complexity of a developing human cortex - billions of cells, organized into dozens of distinct regions and hundreds of cell types - strains the mechanism to a breaking point.

"The only thing a cell 'sees' is itself and its neighbors," explained Stan Kerstjens, a postdoc in Professor Anthony Zador's lab at Cold Spring Harbor and lead author of the study. "But its fate depends on where it sits. A cell in the wrong place becomes the wrong thing, and the brain doesn't develop right. So, every cell must solve two questions: Where am I? And who do I need to become?"

Lineage as a Coordinate System

The solution Kerstjens and colleagues propose is that chemical gradients work in conjunction with a lineage-based mechanism. When cells divide, their daughters tend to remain near their mother. Those daughters divide again, producing granddaughters that cluster near their mothers. Over many generations, descendants of a given progenitor cell accumulate in a local region of tissue, creating a spatial structure that encodes lineage information.

Kerstjens draws an analogy to human population geography. "Consider how human populations spread across a country over generations. Descendants settle near their parents, so people who share ancestry end up in neighboring regions, producing large-scale geographic structures without long-range communication. We argue that a similar principle operates in the developing brain."

Because related cells cluster together, any cell can infer its approximate position from the molecular identity of its nearest neighbors - essentially reading the local ancestry pattern. This requires only local communication, not long-range signals, and it scales naturally with the size of the brain. A human brain with 170 billion cells uses the same mechanism as a zebrafish brain with far fewer, because the underlying logic - stay near your lineage - is independent of total brain size.

Testing the Theory Across Species

To test this model, Kerstjens and colleagues constructed what they call a "lineage-based model of scalable positional information." They began with theoretical computations, then validated the model by analyzing gene expression data from developing mouse brains, looking at both individual cells and clusters. Finally they confirmed the results in zebrafish, demonstrating that the model correctly predicts positional organization in a brain of very different size and complexity from the mouse.

The cross-species validation is particularly meaningful. A purely mathematical model might fit one dataset by coincidence. Fitting the same model to brains that evolved independently over hundreds of millions of years, and that differ dramatically in size and organization, provides much stronger evidence that the underlying principle is real.

Implications Beyond Neuroscience

Kerstjens notes that the lineage-based positional model could apply to developing tissues beyond the brain - including tumors, where cells also proliferate from a common origin and the spatial relationship between daughter cells may influence how a tumor grows and invades surrounding tissue.

There are also potential implications for artificial intelligence systems that learn from data passed hierarchically from one generation of models to the next, in a structure that echoes biological lineage. "How did the brain manage to accumulate this capability, not just over its developmental time, but over evolutionary time? This is one piece in that big puzzle," Kerstjens said.