Star-Shaped Brain Cells May Hold Key to Memory Storage

The human brain’s massive memory storage capacity has puzzled scientists for decades.

With 86 billion neurons packed into our skulls, the brain can somehow store vastly more information than should be mathematically possible using neurons alone. Now MIT researchers believe they’ve found the missing piece: star-shaped cells called astrocytes that have been hiding in plain sight, quietly orchestrating memory formation alongside neurons.

The team’s new model, published in the Proceedings of the National Academy of Sciences, suggests that astrocytes don’t just support neurons—they actively participate in memory storage through an intricate network of cellular processes that can contact hundreds of thousands of synapses simultaneously.

Beyond Support Cells

“Originally, astrocytes were believed to just clean up around neurons, but there’s no particular reason that evolution did not realize that, because each astrocyte can contact hundreds of thousands of synapses, they could also be used for computation,” says Jean-Jacques Slotine, an MIT professor of mechanical engineering and brain and cognitive sciences.

Astrocytes have long been relegated to housekeeping duties in the brain—cleaning debris, delivering nutrients, and maintaining blood flow. But recent experiments have revealed something surprising: when connections between astrocytes and neurons in the hippocampus are disrupted, memory storage and retrieval fail.

This discovery prompted the MIT team to ask a fundamental question. What if memories aren’t stored solely in the connections between neurons, as scientists have assumed for decades?

The Calcium Connection

Unlike neurons, astrocytes can’t fire electrical signals. Instead, they communicate through calcium signaling—waves of calcium ions that flow through their branching processes like messages in a biological telegraph system.

Each astrocyte extends thin tentacles called processes that wrap around individual synapses, creating what scientists call “tripartite synapses”—three-way connections between two neurons and one astrocyte. When neurons become active, astrocytes detect this activity and respond by releasing chemical messengers called gliotransmitters back into the synapse.

“There’s a closed circle between neuron signaling and astrocyte-to-neuron signaling,” explains Leo Kozachkov, the study’s lead author. “The thing that is unknown is precisely what kind of computations the astrocytes can do with the information that they’re sensing from neurons.”

A New Memory Architecture

The researchers developed a mathematical model based on dense associative memory networks—sophisticated systems that can store and recall patterns far more efficiently than traditional neural networks. Their model treats each astrocyte process as an individual computational unit rather than viewing the entire astrocyte as a single entity.

This approach yields remarkable results. While conventional memory models store a fixed amount of information regardless of network size, the neuron-astrocyte model shows superior scaling. As the network grows larger, the ratio of stored memories to computational units increases linearly.

“By conceptualizing tripartite synaptic domains—where astrocytes interact dynamically with pre- and postsynaptic neurons—as the brain’s fundamental computational units, the authors argue that each unit can store as many memory patterns as there are neurons in the network,” notes Maurizio De Pitta, an assistant professor at the University of Toronto who wasn’t involved in the study.

Key Findings from the Research:

• Astrocytes can form over one million tripartite synapses per cell
• Memory capacity scales linearly with network size in neuron-astrocyte systems
• Calcium flow patterns within astrocytes may encode memory information
• The model bridges dense associative memories and transformer architectures
• Human astrocytes are larger and more active than those in rodents

The BCS-BEC Connection

A particularly intriguing aspect that extends beyond typical coverage involves the model’s connection to quantum physics principles. The researchers discovered their system operates in what’s called the “BCS-BEC crossover” regime—a frontier area where memory storage mechanisms transition between loosely coupled states (similar to how electrons pair in superconductors) and tightly bound quantum fluid-like states.

This crossover behavior suggests the brain may employ quantum mechanical principles for information storage that don’t apply to conventional artificial neural networks. The finding could explain why biological memory systems vastly outperform current computer architectures in both capacity and energy efficiency.

Rethinking Memory Storage

The implications extend far beyond neuroscience. The model suggests that memories might be encoded in the calcium transport machinery within astrocytes—a radical departure from the century-old belief that memories live in synaptic connections between neurons.

“By careful coordination of these two things—the spatial temporal pattern of calcium in the cell and then the signaling back to the neurons—you can get exactly the dynamics you need for this massively increased memory capacity,” Kozachkov explains.

Testing this hypothesis will require new experimental techniques. Researchers would need to selectively interfere with calcium diffusion within astrocytes and observe whether memory recall suffers as predicted.

From Brain to AI

The discovery could also revolutionize artificial intelligence. Current AI systems like large language models require enormous computational resources and energy. But the neuron-astrocyte architecture offers a more efficient alternative.

The model can smoothly transition between operating like a dense associative memory system and functioning like a transformer—the architecture behind ChatGPT and other language models. This flexibility could enable new AI designs that are both more powerful and more energy-efficient.

“While neuroscience initially inspired key ideas in AI, the last 50 years of neuroscience research have had little influence on the field,” Slotine observes. “In this sense, this work may be one of the first contributions to AI informed by recent neuroscience research.”

What started as a question about brain cell housekeeping may have unveiled a fundamental principle of biological computation. If the researchers are correct, the star-shaped cells scattered throughout our brains aren’t just cleaning up—they’re actively writing the story of our memories.