The ascending and descending pathways for the control of action inhibition

We are often required to cancel an action after it has been initiated, e.g. suddenly stopping because of an approaching car. Here, crossing the road is the 'go process', while cancelling the movement is the 'stop process'. Go and stop processes race from brain to muscle. The stop process takes a direct route, known as the hyperdirect pathway, allowing it to beat the go process and prevent movement if needed. Impaired stopping harms wellbeing and socioeconomic status, as is seen with impulse control disorders like addiction.

The hyperdirect pathway cuts excitatory drive from the motor cortex to the muscle. But a number of key questions remain. Firstly, it is unclear how the hyperdirect pathway contributes to the stopping of muscle relaxations. Relaxing a muscle, e.g. letting an outstretched arm fall, is an important part of movement control, and research has shown that movements caused by relaxations can be stopped through active contractions of the muscles. This does not fit with existing models of the hyperdirect pathway, which stipulate that stopping consists of terminating drive to muscles, not 'adding' muscle activity. We aim to determine the level at which muscle activity needed for stopping muscle relaxations originates, by comparing frontal and motor cortex activity in humans using magnetoencephalography during the stopping of muscle contractions and relaxations. This will reveal whether the hyperdirect signaling causes the motor cortex to activate the muscle, or if such activity originates beyond the motor cortex, perhaps constituting a peripheral braking mechanism.

We will also address the related but more general question of whether the hyperdirect pathway only outputs via the motor cortex, or has branches to the muscle that bypass the motor cortex. One hypothesis is that the stop process can take a subcortical route, known as peripheral braking. We will test this using our own recent advances in simultaneous imaging of the activity of the human spinal cord and brain. If alternate routes are utilized, the stop process will be detected in the spinal cord before it is detected in the motor cortex. Confirming alternate routes involving rapid modulation of spinal cord output will help explain how disparate patterns of muscle activation are automatically controlled by the stop process.

Muscle activity for given a movement changes with context, such as when a limb gets heavier from altered posture. How actions are cancelled under these dynamic conditions is not well captured by current models of the hyperdirect pathway, which do not include muscle feedback. We will address this using robotic devices that perturb the arm during stopping, and measuring the response on the current and future trials. This will reveal how the brain uses ascending information from the muscle to modify and improve the stopping process.

We hypothesize that feedback from muscle to brain during stopping also improves future movement accuracy. Movements are rarely fully cancelled. The resulting partial muscle activity could be used for learning, without the negative consequences of making a movement at the wrong time (i.e. after a stop signal). We will test this by having participants repeat reaching movements after cancelations. The pattern of the muscle activity during partial responses is predicted to be fed back to the brain to modify the motor cortex, thereby improving subsequent reach accuracy. As such, we will show how ascending signals related to the stop process actually modify future go processes.

Thus, with four sets of experiments we aim to develop a new, comprehensive account of how actions are stopped. This account will go beyond viewing stopping as a single, static and unidirectional process, instead emphasizing how stopping is underpinned by multiple pathways that can be dynamically adjusted by feedback, allowing disparate patterns of muscle activity to be controlled.

In life we often must reactively cancel actions in response to stop signals. Stop commands must be dynamically adjusted according to muscle state. Stopping depends on the hyperdirect pathway from the frontal cortex to subthalamic nucleus (STN), which suppresses primary motor cortex (M1) output, ending muscles contractions. However, this account does not explain how relaxing muscles are stopped, which is known to involve muscle activation, termed active braking. Hyperdirect models assume the pathway acts solely on M1, but peripheral changes often occur before M1 changes, suggesting alternate routes. Likewise, the role of peripheral afferent feedback is uncertain. Evidence suggests that active braking is scaled according to muscle state, implying feedback is critical, possibly involving the dorsoposterior parietal cortex (DPPC). Such feedback may be used to 'covertly' learn about cancelled actions, facilitating improved movement accuracy.

We will investigate if active braking originates in M1 using TMS and MEG. An M1 locus will manifest as increased excitability and decreased beta bursts during muscle relaxation stopping. Spinal cord MEG will show if the stop process is detected in the spinal cord before M1, supporting putative hyperdirect pathway branching, and whether signaling indicates suppression of subcortical drive or excitation of spinal interneurons, forming a peripheral braking mechanism. Perturbations during stopping will determine how active braking is updated in response to environmental change, while disruption of the DPPC with TMS will probe the region's control of such adjustments. Finally, analysis of muscle activity and M1 beta will confirm how stop-related afferent feedback is used by the brain to improve movement performance. Together these experiments will show the bi-directional flow of information during stopping, explaining how the brain controls movement despite the dynamic state of the muscles and environment.