Mechanism for flexible thinking and learning identified in the brain

Dendrites, the tree-like extensions of nerve cells in the brain, are characterised by a wide range of functional capabilities, the role of which has hitherto been unclear. Researchers at HU have now discovered that they enable flexible learning.

The ability to adapt behaviour flexibly to changing environments is a hallmark of intelligence. However, the biological mechanisms in the brain that are responsible for this cognitive flexibility remain poorly understood. Eduardo Maristany de las Casas, a PhD student in Prof. Matthew Larkum’s research group “Neural Plasticity” at the Department of Biology at Humboldt-Universität zu Berlin, in collaboration with Prof. Dieter Jaeger from Emory University (Atlanta, USA), has investigated the extent to which the biological basis of this intelligent behaviour lies in the long, branched extensions of nerve cells, the dendrites. To this end, in a technically sophisticated experimental setup, they combined a behavioural test with mice – in which the animals had to switch back and forth between easy and difficult tasks – with microscopic imaging (two-photon fluorescence microscopy) of the relevant brain regions as well as targeted manipulation of individual neurons. “In this study, we show that dendrites and their interaction with interneurons lie at the core of this ability, and that different forms of learning are implemented differently in the brain,” says Eduardo Maristany. The results of the study were recently published in the renowned journal Science.

Interaction with interneurons and cluster formation enables behavioural adaptation

Maristany and his team discovered that the tree-like extensions of a specific type of neuron – pyramidal neurons – in the cerebral cortex of the mouse play a crucial role in flexible learning. These pyramidal neurons are located in the part of the cerebral cortex responsible for controlling voluntary movements (frontal motor cortex). Flexible learning takes place through the interaction of these dendrites with a specific class of nerve cells (NDNF interneurons), which are located in the same outermost layer of the cerebral cortex and act as a kind of gatekeeper: when these cells are active, they suppress calcium signalling in the dendrites, thereby preventing the brain from updating learned rules. During a learning process, however, they reduce their activity. This opens a window of opportunity for the animal to adapt its behaviour. At the cellular level, this is accompanied by a change in the shape of the dendrites: their spines group together, according to their function thereby forming functional clusters. It is only through this dendritic plasticity that the dendrites can transmit information effectively and thus enable the learning process. 

^{A microscopic image of dendrites, the tree-like extensions of nerve cells, in a mouse’s brain. 
Photo: Kris Killmann, Larkum Lab/ HU}

In his experimental setup, Maristany combined a behavioural test with mice and microscopic images of the relevant brain regions (two-photon fluorescence microscopy) with targeted manipulation of individual neurons in the mice’s brains. For the test, the mice had to learn two different behavioural patterns of varying difficulty and were trained to switch back and forth between them (rule-switching paradigm). The complex task involved learning that, whenever their whiskers were stimulated on the left or right side, they should lick sugar water (as a reward) on the corresponding side. In the simple task, they were required to lick the sugar water always on the left-hand, regardless of which side their whiskers had been stimulated. Switching back and forth between these tasks revealed whether and how learnt behaviours are retained, overwritten or updated. 

Dendritic plasticity is only required in complex learning situations

During the behavioural tests, the researchers observed that the dendritic plasticity described above was not required for the execution of the simple behaviour once the mice mastered it, nor for the switch from the more difficult to the simpler rule, but only when the mice had to perform the complex behavioural rule again. Clusters formed by the spines on the dendrites disintegrated as soon as the animals performed the simple task.

This study provides some of the first direct evidence that the elaborate dendritic trees of cortical neurons act as active computational units, essential for adaptive behaviour. The extent to which these findings can be generalised to other brain regions or to more real-life learning conditions remains to be investigated.

Insights for the treatment of cognitive disorders and a source of inspiration for AI

The biological mechanisms of learning uncovered in the study could, in future, play a role in understanding disorders of cognitive flexibility – such as those seen in autism spectrum disorders, schizophrenia, and age-related cognitive decline, where the ability to adapt to changing rules and environments is impaired. Additionally, the principles uncovered here — particularly the gating of plasticity by inhibitory circuits — may inspire new architectures in artificial intelligence and machine learning. “One of the aspects I find most fascinating is the brain’s ability to remain cognitively flexible and to keep learning with relatively limited resources,” says Eduardo Maristany.

Source: Press release HU Berlin

Original publication

Contact:

Prof. Dr. Matthew Larkum
HU Berlin
NeuroCure PI
Faculty of Life Sciences
Department of Biology
Neuronal Plasticity
Charité Universitätsmedizin Berlin
Charité Cross Over - Campus Mitte
Charitéplatz 1 (intern: Virchowweg 6)
D-10117 Berlin

Larkum Lab