“Contralateral control” is the core tenet of motor functioning, supported by the anatomical arrangement of motor and sensory fibers. Notwithstanding, multiple pieces of evidence agree that this is only half the story, as the contribution of the ipsilateral hemisphere holds a fundamental role in successful motor planning and execution. The advent of more sophisticated methods to assess connectivity within and between hemispheres, such as new TMS protocols and fMRI data analysis techniques, has boosted knowledge on the topic.
While TMS is an ideal candidate to infer the causal impact of one brain region on an interconnected area and ensures high temporal resolution, it can hardly deal with the complexity of a network consisting of multiple hubs that exert a reciprocal influence. Indeed, the majority of studies using TMS to assess interhemispheric dynamics have focused on the M1 activity, as the increase or decrease in amplitude of the motor-evoked potential (MEP) recorded through electromyography has been used as a hallmark of neural inhibition or facilitation.
DCM applied to fMRI data can be a valid method to detect complex interactions within a wider network and to assess directional influences between brain areas besides M1. Since caution is necessary when relating BOLD signal increases or decreases to actual neural states, we will discuss our results in light of those deriving from both techniques, being aware that it is not possible to find strict homologies.
As we evaluated the effective connectivity within the grasping network during right-hand grasping execution and imagery, we will discuss our results separately for each experimental condition.
Grasping execution
The first evidence drawn from the “real” condition pointed toward striking similar network connectivity across hemispheres: the dynamic interactions occurring between right parieto-frontal areas were similar to those of the left hemisphere, even though connection strengths were overall lower, a result highly consistent with that found by Begliomini and colleagues [17]. The signal spanned from aIPs to PMv, was susceptible to modulations by PMd and SMA, and finally reached M1. Notably, only the forward excitatory connections from aIPs to PMv and from PMv to M1 were stronger in the left vs the right hemisphere.
Beyond confirming the wide involvement of a bilateral parieto-frontal network during grasping, our results suggest that very similar mechanisms occur across hemispheres in terms of network functionality. As we already discussed in our previous work relative to the left-hemisphere results [19], we suggest that a general motor program for grasping is planned by the aIPs-PMv circuit, whereas PMd and SMA encode high-level features of the movement. Notably, the excitatory input from the right PMv to the right M1 supports the view that the ipsilateral M1 provides additional resources when performing unilateral complex movements even if the amount of activation detected by fMRI is extremely low. Together, these findings support the key role of the ipsilateral hemisphere in the planning and execution of a complex unimanual action, whereby hand actions are supported by limb-invariant representations in parietal and frontal areas [7].
The crucial issue we revamp here is that interhemispheric connections may bridge the two hemispheres to allow sharing of resources and information for a successful motor plan. For this reason, the second step of our work sought to describe interhemispheric dynamics occurring during the task. We observed a massive interhemispheric inhibition spanning from the right to the left hemisphere, whereas interhemispheric facilitation occurred in the opposite direction. Noteworthy, this latter pattern reverses the inhibition occurring at baseline (A matrix), therefore being a hallmark of the connectivity modifications depending on the current motor state, i.e., active or passive. While a short and a long latency interhemispheric inhibition emerges from contralateral PMd to M1 at rest [52], some studies have pointed out that interhemispheric dynamics change during the transition from a motor state to another, reflecting the switch from motor planning to implementation. For instance, Liuzzi and colleagues [53], using a simple right-hand reaction time task, recorded a biphasic pattern of modulation exerted from the right PMd toward the left M1, namely early and late latency facilitation respectively evoked by movement selection and execution, while M1-M1 interactions were modulated only right before movement onset. With the current experimental design, we could not decompose grasping movements into planning and implementation phases, therefore future studies will be aimed at disentangling dynamic changes in interhemispheric connectivity according to each stage of the movement.
The role of the right PMd in driving the inhibition toward the left premotor and motor cortex deserves a spotlight. PMd covers a key role in decision making and is a central structure for the selection and initiation of voluntary actions [54,55]. In agreement with that, it has been suggested that PMd is responsible for the top-down processing of movement control both within and across hemispheres [56]. Beyond PMd, our data show that in the case of grasping movements, the left PMv is the central hub of the network as it groups inputs from the ipsilateral and the contralateral hemispheres.
Worthy of mention is the role of the left SMA which, together with the left PMv, exerts a positive influence on its right homologue. Notably, a similar modulation was found by Gao and colleagues [16] during a finger tapping task, supporting the idea that the information transfer between bilateral SMAs plays a crucial role in both unimanual and bimanual movements [15,57,58]. Still, a different role of left and right SMA, independently of the effector (left or right hand), was suggested by White and colleagues [22] in a TMS study on grip force. According to these authors, the left SMA is crucial for predicting the required grip force, overall encoding object dynamics, whereas the right SMA is more likely involved in the translation of object representation into motor commands. In line with this view, our results suggest that object dynamics encoded in the left SMA may converge in the right SMA, integrated with the generated motor command, and transferred back to the left PMv to drive the last stages of motor planning and execution.
TMS studies have consistently detected an interhemispheric inhibition between the two motor cortices, a result that we failed to replicate. Methodological and theoretical issues can account for this discrepancy. First, we did observe M1-M1 inhibition at rest, therefore our results must be interpreted by accounting for the evidence that this baseline inhibitory pattern is not strengthened or reduced during motor execution or imagery, in other words no task-dependent modulation emerged in the B matrix. Moreover, it has been suggested that iM1 shapes the muscular command in different ways depending on the stage of the motor execution, according to which this area can drive inhibitory or facilitatory inputs [59,60] and shows unspecific activity for right- or left-hand movements during planning, which turns to be specific during execution [7]. Although here we could not account for this distinction, on a deeper analysis the comparison between present and previous findings suggests that, during simple movements, a strengthening of the cM1-iM1 crosstalk is necessary (see the DCM study by [15]), whereas complex movements might rely upon a premotor rather than a direct motor interhemispheric modulation (see also [17]).
Grasping imagery
When imagining the grasping movement, a different scenario emerged. A first striking difference uncovered by the whole-brain activation maps is the only subtle involvement of the right hemisphere during grasping imagery, a result that suggests a contralateral hemisphere dominance in motor imagery vs the bilateral involvement of the network during motor execution.
In line with that, only partial similarities in the effective connectivity across the two hemispheres were revealed by the DCM. Differently from what we described for grasping execution, the connection from aIPs to PMv was equally modulated in both hemispheres. Furthermore, SMA and PMd cooperated to inhibit PMv only in the left hemisphere. Conversely, in the right one a loop directly linking SMA and PMv emerged, as well as an inhibition from the right PMd to the left M1. Presumably, these dynamic interactions both drive motor imagery and concur to prevent excitation from the left PMv toward M1, resulting in the suppression of the motor output. This latter interpretation is in line with the concept of an “impulse-control” mechanism aimed at preventing overt activity in the right hand [61].
Overall, these results point toward a different role of the right and left hemispheres during motor imagery. A previous study on motor imagery in stroke patients found that only right hemisphere damage impaired the timing estimation of imagined movements, sparing that of real movements [62]. These authors suggested that the right hemisphere is crucial for maintaining spatial information over time when internally simulating the motor pattern. As SMA is a key node in temporal processing [63], our results may be interpreted from the perspective that temporal encoding of the imagined movement may be processed by the right SMA and then this information is transferred to the right PMv to coordinate the imagined movement.
Together, we can conclude that ME and MI share neural effector- and task-independent representations in high-order areas such as aIPs, whereas lower-order areas as SMA and PMd deem with effector- and task-dependent representation [7,8], a result confirmed by the different connectivity patterns we found across hemispheres and condition. Future studies may be designed with the specific aim to address the different functional contributions of the two hemispheres to motor imagery vs execution with compelling implications for rehabilitation practice, for instance guiding the choice of target areas for brain computer interfaces (BCI) protocols using MI on post-stroke patients.