Every time we reach for a cup of coffee, the movement is never quite the same. A team of neuroscientists at the University of Fribourg (Unifr) has shown that this variability relies on neurons that allow the brain to correct each movement in a unique way. Their results, published in the journal Current Biology, reveal that these neurons act as the brain's «optimal estimators», combining sensory input with internal predictions. When they are inactive, movements remain accurate, but they begin to follow a more stereotyped pre-planned path.

Try reaching for the same cup a hundred times, and your hand will never follow the same path twice. These variations are not due to clumsiness; rather, they show that the brain constantly adjusts our movements using proprioception, the «sixth sense» that tells us where our limbs are without needing to look at them. But which neurons perform this correction and what exactly do they compute? To find out, neuroscientists at UNIFR studied reaching movements of mice after selectively switching off a small group of proprioceptive neurons.

An essential, yet poorly understood sense
All our actions rely on continuous sensory feedback from sensors in our muscles, tendons and joints. These proprioceptors tell the brain where our body is and how it moves. We know that this information reaches the somatosensory cortex, a key region of the brain. However, how the cortex uses this information to control movement, rather than just to perceive it, remains unclear. «We wanted to understand what cortical proprioceptive neurons concretely do for our movements», explains Professor Mario Prsa. «We wanted to move beyond describing what perceptual signals they encode and instead ask what happens to a movement when we take them away.»

Fewer cells, more stereotyped movements
The researchers at Unifr studied mice trained to grasp water droplets, recording their three-dimensional paw trajectories with high-speed cameras. They then used two-photon microscopy to specifically identify proprioceptive neurons in the somatosensory cortex, which they managed to eliminate one by one by aiming a tightly focused laser at their soma, a target a fraction the width of a human hair. This approach is fundamentally different from traditional methods, which typically damage or disable an entire area of the brain.

Following this selective ablation, the mice kept grasping the droplets just as accurately as before, but their reaching paths changed in a surprising way. «It was an unexpected result», explains Mélanie Palacio-Manzano, the study's lead author. «The trajectories became more scattered in space, yet at the same time more stereotyped in their shape, as if the movements had become more robotic.»

A brain that predicts as much as it senses
To make sense of this result, the researchers turned to a mathematical framework called optimal feedback control. By systematically adjusting the model, they found that a single parameter change reproduced their empirical observations: lowering how much the system trusted its sensory feedback relative to its internal prediction. «These neurons therefore calculate an optimal estimate of the limb by weighing what the senses report against what the brain expects. When that estimate is degraded, the movements fall back on a nominal pre-planned path, instead of being uniquely corrected», concludes Mario Prsa, head of the Laboratory of Sensorimotor Neuroscience.

Discoveries that pave the way for better therapies
This study provides rare biological evidence for a central idea in motor neuroscience: the natural variations in our gestures are not errors, but the sign that the brain is constantly correcting itself. The authors also show that this specificity is lost when larger regions of the cortex are removed, underscoring the value of precise, cell-level manipulations for understanding how the brain works.

These results change how we think about disorders in which movement is disrupted, and how we might restore it. If we want prosthetic limbs to feel natural, feedback signals may need to support this same process of prediction and correction and not just represent where the limb is.

> Read the full article