Scientists have uncovered a new kind of electrical process in the human brain that could play a key role in the unique way our brains compute.
Our brains are computers that work using a system of connected brain cells, called neurons, that exchange information using chemical and electric signals called action potentials. Researchers have discovered that certain cells in the human cortex, the outer layer of the brain, transmit signals in a way not seen in corresponding rodent cells. This process might be important to better understanding our unique brains and to improving programs that are based on a model of the human brain.
“Human neurons may be more powerful computational devices than previously thought,” study corresponding author Matthew Larkum at Humboldt University of Berlin told Gizmodo in an email.
Human brains have a thick cortex, especially the second and third layers (L2/3) from the surface. These layers contain brain cells with lots of branches, called dendrites, that connect them to and exchange information with other brain cells. The researchers acquired and analysed slices of L2/3 tissue from patients with epilepsy and tumours, focusing specifically on these dendrites. Larkum explained via email that epilepsy surgeries provided a sufficient amount of available cortex tissue, while the tumour patient tissue was used to ensure that the observations weren’t unique to people with epilepsy.
The team hooked the tissues to a patch clamp—essentially a system that constructs an electrical circuit from the cells and a measurement instrument—and used fluorescing microscope to observe the action of these L2/3 cells. The team noticed that inputted electrical currents ignited more action potentials than they would in rodent cells and that a chemical that should have blocked the dendrites’ activity did not completely do so.
The experiments revealed the presence of a new kind of action potential that travels with the help of calcium ions, rather than sodium and calcium ions. It’s a kind of action potential not seen before in mammalian cortex cells, according to the paper published in Science.
But the researchers took their experiment further. After studying the behaviour of these “calcium-mediated dendritic action potentials,” they modelled them in a computer simulation, and it turned out that they were able to perform a computational function (called the XOR gate) that scientists previously thought would require a network of neurons, rather than just one. They propose that an artificial neural network could incorporate these new kinds of action potentials to simplify calculations.
Studies like these have their limitations. The researchers couldn’t model the entire neuron, and the work wasn’t carried out in humans, just in human cells. Perhaps other mammals do fire these kinds of action potentials, but they’re not visible in tissue samples in the lab.
The researchers hope to continue studying this action potential and the behaviour of dendrites as a way to understand the cortex and why it works the way it does.