A unifying motor control framework for task-specific dystonia

Abstract

Task-specific dystonia is a movement disorder characterized by a painless loss of dexterity specific to a particular motor skill. This disorder is prevalent among writers, musicians, dancers and athletes. No current treatment is predictably effective, and the disorder generally ends the careers of affected individuals. Traditional disease models of dystonia have a number of limitations with regard to task-specific dystonia. We therefore discuss emerging evidence that the disorder has its origins within normal compensatory mechanisms of a healthy motor system in which the representation and reproduction of motor skill are disrupted. We describe how risk factors for task-specific dystonia can be stratified and translated into mechanisms of dysfunctional motor control. The proposed model aims to define new directions for experimental research and stimulate therapeutic advances for this highly disabling disorder.

Figure 1. Motor hierarchy in skill learning.

a | In early learning, explicit or cognitive processing of task instructions occurs at the selection level. At the execution level, the most appropriate set of motor elements is mapped to task requirements. b | Later in learning, task performance is largely automatic. Skill elements become encoded within the dynamic neural network as an intermediate level, and the flow of motor elements requires little explicit or cognitive control. Motor chunking refers to the linking of elemental execution representations within the intermediate level. For example, a specific motor goal at the selection level (s1) might initiate two motor chunks at the intermediate level (c1, c2) that lead to distinct motor sequences (e1–e3 in the case of c1 and e4–e5 in the case of c2). A second motor goal (s2) has a different order requirement but is built from the same components; initiating the two chunks in reverse order (c2 then c1) still produces the required motor behaviour (e4–e5 from c2 and e1–e3 from c1). These intermediate-level representations link elementary units, enabling them to be activated in a fluent manner without the need for direct mapping from the selection level. Permission obtained from Elsevier © Diedrichsen, J. & Kornysheva, K. Trends Cogn. Sci. 19, 227–233 (2015).

Figure 2. Evidence for a motor hierarchy.

a | Motor synergies are fragments of movement sequences encoded within the motor cortex. Complex gestures such as grasping or licking can be reproducibly evoked by electrical stimulation of the shaded regions of primary motor and premotor cortices in monkeys b | The results of an experiment in which humans learned to do 10 sets of two button presses. Each point on the graph represents one set, and the spacing between two adjacent lines corresponds to the time interval between the onsets of each set. During early learning (trials 1–5) the total time taken for the sequence to be executed was longer and dispersion of the 10 sets through time was approximately even. During late learning (trials 76–80) the total time taken for the entire sequence has decreased and motor chunking is evident — the different sets are grouped into three chunks (set 1–4, set 5–7, set 8–10). c) Evidence for a modular representation of rhythm or ‘temporal chunking’. The timing and finger order of four different sequences are shown: trained (green); temporal transfer (red); spatial transfer (blue); and novel (grey). Behavioural benefits (reaction time decreases) of new sequences retaining either the trained temporal or spatial features are seen in comparison to entirely novel sequences. Multivariate analysis of functional MRI data reveals independent representations (red, blue) of these spatial and temporal features, some of which occur in overlapping (pink) regions of the premotor (PM) cortex. The primary motor cortex (M1), by contrast, contains integrated (that is, non-separable) representations of the two sequence features (green). Part a adapted with permission from Elsevier © Graziano, M. S. Trends Cogn. Sci. 20, 121–132 (2016). Part b adapted with permission from Springer © Sakai, K. Exp. Brain Res. 152, 229-242 (2003). Part c adapted with permission from Elsevier © Diedrichsen, J. & Kornysheva, K. Trends Cogn. Sci. 19, 227–233 (2015).

Figure 3. Components required for skill performance and the dynamic interactions between risk factors.

Risk factors can be comprehensively identified by considering all the components that interact in the performance of a given skill (central nervous system, periphery, task and tool). For example, case A exemplifies an illustrator that had taken a prestigious but demanding new job in animation. The patient was required to use a tablet and stylus rather than the usual paint brush (tool) and the work required 1000s of dots demanding forceful demarcation with the stylus (task). The patient also worked for many hours until a painful forearm overuse injury occurred (periphery) and was highly stressed attempting to make imposed deadlines (psychology). Here multiple risk factors seemed to interact in the development of task-specific dystonia. Others patients present with fewer risk factors. The pianist in case B was a classical pianist that took a job playing in a musical. This required a subtle change in instrument such that the pianist was using a smaller keyboard with keys that were less responsive than a piano. In this case the change in tool was the dominant risk factor with repercussions for task parameters and sensorimotor control (indicated by the arrows linking skill components).

Figure 4. Vulnerabilities of highly skilled representations.

a | The neural architecture of overlearned skills is not well defined, but here the highly trained skill is represented as a rigid synergy-like execution pattern, which has poor flexibility due to limited use of motor chunking at the intermediate level. b | Stable everyday tasks (grey region) require a moderate range of spatial and temporal accuracy, and are encoded at approximately the midpoint between floor (F) and ceiling (C) values. By contrast, performers strive to improve time and accuracy functions within movement goals that are tightly prescribed (requiring minimal motor variability and error rates that approach zero). Thus, highly optimised skill representations (red region) require encoding of spatial and temporal accuracy at very close to ceiling values. The cost of such optimization is largely unknown. A high degree of optimizationis likely to limit the flexibility to respond to new task requirements (narrow diameter of skill representation) and other parameters may not be optimized (shown close to floor). c | With increasing skill expertise, the magnitude of the precipitant of task-specific dystonia decreases. In part, this association may be due to the reduced generalizability of highly optimized skill representations.

Figure 5. The development of task-specific dystonia.

a | If the existing hierarchical representation can no longer accommodate task requirements, novel motor control options must be sought. Solutions are likely to require mechanisms comparable to early learning states, in which task requirements are explicitly mapped to basic execution elements. Such mechanisms are ill-equipped to immediately reinstate previous levels of skill performance, which were encoded by a hierarchy of neuronal elements optimised over many years of practice. Movements that are either inappropriate or non-physiological might start to be produced, which can be classed as dystonic as they no longer attain task goals. b | If dystonic movements are rehearsed they are likely to become encoded in a manner similar to any other learned sequence of movements, with a shift of motor control towards increased automaticity and reduced explicit cognitive monitoring of movement sequences. In this situation dystonic movement sequences become increasingly difficult to correct.