New map captures every connection in an insect’s brain and reveals secrets for better AI

Breakthroughs don’t come often in neuroscience, but we just had one. In a tour de force, an international team released the complete map of the young fruit fly’s brain connectivity, described in a paper published in last week Science. With 3,016 neurons and 548,000 synapses, the map – called the connectome – is the most complex circuit diagram of the entire brain to date.

“It’s a wow,” said Dr. Shinya Yamamoto of Baylor College of Medicine, who was not involved in the work.

Why should you care about a fruit fly? Far from uninvited guests at the dining table, Drosophila melanogaster is a darling of neuroscience. Although its brain is smaller than a poppy seed — a far cry from the 100 billion neurons that power the human brain — the fly’s neural system shares principles similar to those underlying our own brain.

This makes them excellent models for refining ideas about how our neural circuits encode memories, make difficult decisions, or navigate social situations like flirting with a potential partner or hanging out with a crush of new friends.

About the main author Dr. Marta Zlatic from the University of Cambridge, MRC Laboratory of Molecular Biology and Janelia Research Campus: “All brains are similar – they are all networks of interconnected neurons – and all brains of all types have many complex behaviors to perform: they all have to process sensory information, learning, choosing actions, navigating their environment, choosing food, escaping predators, etc.

With the new connectome map, “we now have a reference brain,” she said.

A Behemoth Atlas

Connectome are valuable resources. The maps popularized by Sebastian Seung establish neural connections within and between brain regions. Much like tracing computer wires to see how different chips and processors fit together, the connectome is a valuable resource for cracking the brain’s “neural code” — the algorithms that underlie its calculations.

In other words, the connectome is essential to understanding brain function. Because of this, similar work is being done in mice and humans, albeit on a much smaller scale or with much less detail.

So far, scientists have only mapped three whole-brain connectomes, all in worms — including the first animal to receive this honor, the nematode C. elegans. With just over 300 neurons, the project lasted over a decade, with an update for both sexes released in 2019.

Drosophila presents a far greater challenge with around ten times as many neurons C. elegans. But it’s also an ideal next candidate. For one, scientists have already sequenced their entire genome, making it possible to correlate genetic information with the fly’s neural wiring. This could be particularly useful, for example, in deciphering how genes that contribute to Alzheimer’s disease alter neural circuits. Secondly, fruit fly larvae have transparent bodies, which makes them much easier to image under a microscope.

Not all brain wiring maps are created equal. Here, the team opted for the highest resolution: mapping the entire brain at the synapse level. Synapses are connections between neurons where they connect: imagine two mushroom-shaped structures floating close together with a gap. Although neurons are often touted as the fundamental component of computing, synapses are where the magic happens—their connectivity helps functionally wire neural circuits.

Neuron connectivity in the brain. Each dot represents a neuron, and those with similar connectivity are closer. The lines show how neurons connect. Credit: Benjamin Pedigo

Slice and Dice and… robots?

To map synapses, the team turned to the big guns of microscopy: the electron microscope. Compared to microscopes used in university biology, this hardware can capture images at the nanoscale – about a tenth the width of a human hair.

The whole process sounds a bit like a wild dinner recipe. The team first soaked a single six-hour-old larval brain in a solution filled with heavy metals, which penetrated the membranes and proteins of the neurons in the synapses. The brain is then carefully sliced ​​into ultra-thin sections with a diamond blade — think of a delicatessen slicer — and placed under a microscope.

The resulting images – all 21 million – were stitched together using software. The entire process took over a year and a half, with many hours spent manually inspecting the reconstructed neurons and synapses.

The final brain map not only included the location of neurons and their synapses — it also highlighted wiring errors that could support highly efficient neural computations.


The beauty of the new map is that it provides a bird’s-eye view of brain connectivity, supercharged with the power to zoom and enhance.

“The most challenging aspect of this work was understanding and interpreting what we saw,” Zlatic said.

In one analysis, the team found that neurons can be classified into 93 different types based on their connectivity, even when they share the same physical structure. It’s a drastic departure from the most common way to categorize neurons. Instead of grouping them by appearance or function, it may make more sense to focus on their “social network” instead.

As the team dug for synapses, they stumbled upon another surprise. Let me explain: neurons have two main branches. One is the larger input cable – the axon – and the other is a tree-shaped exit – the dendrite. Neurons normally “wire” when synapses connect these two wires.

However, recent studies show that synapses on axons can connect to other synapses on axons; The same applies to dendrites. Analyzing the reconstructed brain, the team found evidence of these non-traditional connections.

“Now we have to rethink them: we probably have to think about creating a new computer model of the nervous system,” said Dr. Chung-Chuang Lo from National Tsing Hua University in Taiwan.

On a broader scale, the map revealed that neurons are eager to chat with others half a world away. Almost 93 percent of neurons are connected to a partner neuron in the other brain hemisphere, suggesting that long-range connections are incredibly common. Even more surprising was a peculiar population that didn’t extend: these neurons, dubbed Kenyon cells, populate primarily the fly’s learning and memory center. Why this happens is still unclear, but it illustrates the brain map’s ability to generate new insights and hypotheses.

Although the neurons and synapses are wired in a nicely compact, “nested” multi-layered structure, the connectome showed that some loved developing connections that jumped through layers — a shortcut that ties together otherwise separate circuits.

Even more fascinating was how much the brain “talks” to itself. Almost 41 percent of the neurons received repetitive input – feedback from other parts of the brain. Each region had its own feedback program. For example, information generally flows from sensory areas of the brain to motor regions, although the opposite also happens, creating a feedback loop.

But perhaps the most socially adept neurons are the ones that secrete dopamine. Known for encoding reward and driving learning, these neurons also had some of the most complex repetitive wiring compared to other types.

From shortcuts to recurring wiring, these biological hardware structures could increase the brain’s computing capacity, compensating for the finite number of neurons and their biological limitations.

“None of us expected that,” says study author Dr. Michael Winding.

From fly to AI

The study isn’t the first to map the Drosophila brain. Previously, a team led by Dr. Davi Bock at the Janella Research Campus pointed to a small node in the adult fruit fly brain that is responsible for learning and remembering smells with details at the synapse level. Zlatic’s team also traced a sensory circuit in the fruit fly larvae to make decisions by mapping just 138 neurons.

The whole brain connectome is a game changer. For one, scientists now have a sophisticated reference brain to test theories for neural computations. Second, the connectome map and its derived computation are similar to state-of-the-art machine learning.

“It’s actually quite nice because we know that recurrent neural networks are pretty powerful in artificial intelligence,” Zlatic said. “By comparing this biological system, we may be able to inspire better artificial networks as well.”

Credit: Michael Winding

Leave a Reply

Your email address will not be published. Required fields are marked *