Brain Organoids Learn to Distinguish Signals in Connected Networks

The human brain is an extraordinarily intricate system for processing information, yet certain forms of research involving it remain firmly out of bounds.

Brain research endured the ethical failures that marked much of the 20th century. For very sound moral reasons, strict regulations now determine what researchers may do with human brains that are still being used by living people.

What researchers are able to do, however, is cultivate small pieces of brain tissue in laboratory dishes.

These minute, three-dimensional clusters of human neural tissue, known as brain organoids, are not complete brains recreated at a reduced scale.

Even so, they replicate sufficient cellular structure and electrical activity to let scientists examine processes that could never ethically be probed, stimulated, fitted with electrodes, and repeatedly shocked inside a living person's head.

A team headed by University of Tokyo neuroscientist Yoshiho Ikeuchi has now grown several of these organoids, connected them into a network, and observed them apparently ‘learning’ to recognise which of the other organoids was transmitting signals.

“Together, these results demonstrate that repeated input to an organized modular network can rewire organoids generated under identical conditions into functionally differentiated modules,” the researchers write, “thereby generating task-relevant heterogeneity that underlies consistent functional enhancement in vitro.”

No, these organoids are not creating sentient networks of brain blobs poised to rise up against their flesh-and-blood tormentors.

Instead, the study indicates that neural tissue may need a particular modular arrangement before it can learn to distinguish between separate signals.

Brain organoids and the brain's modular structure

The human brain is not a single uniform mass of interconnected neurons. It comprises distinct regions with their own local networks, linked to one another through longer-distance connections. This modular layout enables individual brain areas to carry out specialised functions while still operating as part of an interconnected system.

Yet the minimum conditions needed for neural tissue to form functional circuits are not fully known. As brain organoid models have become more advanced, Ikeuchi and colleagues turned to these brain blobs to explore that question.

The blobs are produced from human induced pluripotent stem cells (hiPSCs): mature human cells reprogrammed into a stem-cell-like condition, allowing them to become many cell types when guided appropriately.

Ikeuchi's team directed the hiPSCs to form brain tissue, producing a collection of small neural-tissue samples for experimentation.

They set these pieces onto specialised chips coated with tiny electrodes, enabling the team both to deliver electrical stimulation to the organoids and to measure their neural activity.

Certain chips held one brain blob, others contained two, and some carried three.

The researchers then allowed them to develop for more than a fortnight. This gave organoids sharing a chip time to extend axons - the long projections through which nerve cells carry signals - and establish connections.

Recordings made afterwards revealed synchronised electrical activity among organoids placed together, confirming that functional networks had formed.

Testing connected brain organoid networks

Next came The Zappening (why anyone might imagine the organoids would seek revenge remains a complete mystery).

For two weeks, researchers electrically stimulated the neural networks daily at two separate points, delivering 100 pulses at each site and recording neural activity elsewhere on the chip.

On chips containing a single organoid, the two sets of pulses targeted two areas of that same organoid. With two organoids, both stimulation sets were applied to one organoid while activity was measured in the other. For the three-organoid chips, stimulation was divided between two organoids, and recordings came from the third organoid, which received no pulses.

The aim was to determine whether repeated stimulation could ultimately train neural tissue to react differently according to the signal's origin.

To test this, the scientists used machine-learning algorithms to assess the activity patterns generated after every pulse, asking the systems to determine which of the two sites had been stimulated.

At the experiment's outset, the team intentionally chose stimulation points that triggered responses which were hard to differentiate. The algorithms identified the correct source only about half the time, matching the result expected from guessing.

After The Zappening, however, a different pattern started to appear.

This was not true of the individual organoids or the pairs, whose responses to the two signals continued to be difficult to distinguish.

In networks of three organoids, by contrast, the signals became progressively easier to tell apart from the activity recorded in the third organoid. Their responses also grew quicker, more consistent, and more spatially distinct.

Three-organoid networks that didn't receive The Zappening showed no equivalent improvement, indicating that the change was not merely because the blobs had spent a further fortnight growing in a dish.

Rather, the repeated signals themselves seemed to have prompted the neural networks to reorganise.

What the organoid learning results suggest

This represents a highly rudimentary type of learning, and remains vastly removed from the intricate information processing of a real, fully developed human brain.

Nevertheless, the researchers say the findings imply that arranging neural tissue into connected modules with separate functions could support the development of an ability to distinguish different information sources.

Because basic signal discrimination is essential for more sophisticated information processing, the finding gives scientists a new route for studying how progressively complex neural functions arise.

“Testing organoids with more complex signals and sophisticated tasks will be critical for future research,” the researchers write.

“Future studies should focus on identifying potential 'node cells' or 'hub neurons' that facilitate connections within and between organoids – combining this with biochemical and molecular analyses could reveal the mechanistic basis of changes in synaptic plasticity and connectivity.”

The study has been published in Communications Biology.

This article was fact-checked and edited by Rebecca Dyer. Although we take pride in our process, we are only human. If you notice an error, please tell us.