Interactions Between the Corticospinal Tract and Premotor–Motor Pathways for Residual Motor Output After Stroke
Background and Purpose—Brain imaging has continuously enhanced our understanding how different brain networks contribute to motor recovery after stroke. However, the present models are still incomplete and do not fit for every patient. The interaction between the degree of damage of the corticospinal tract (CST) and of corticocortical motor connections, that is, the influence of the microstructural state of one connection on the importance of another has been largely neglected.
Methods—Applying diffusion–weighted imaging and probabilistic tractography, we investigated cross-network interactions between the integrity of ipsilesional CST and ipsilesional corticocortical motor pathways for variance in residual motor outcome in 53 patients with subacute stroke.
Results—The main finding was a significant interaction between the CST and corticocortical connections between the primary motor and ventral premotor cortex in relation to residual motor output. More specifically, the data indicate that the microstructural state of the connection primary motor–ventral premotor cortex plays only a role in patients with significant damage to the CST. In patients with slightly affected CST, this connection did not explain a relevant amount of variance in motor outcome.
Conclusions—The present data show that patients with stroke with different degree of CST disruption differ in their dependency on structural premotor–motor connections for residual motor output. This finding might have important implications for future research on recovery prediction models and on responses to treatment strategies.
In the past decade, functional1 and structural connectivity analyses2 have dramatically enhanced our understanding of network alterations after ischemic stroke. Noninvasive brain stimulation protocols have been developed and applied variously to use this knowledge to enhance motor recovery. Despite that our present models can explain more and more variance in residual motor output and recovery on the one hand,3 and brain stimulation has evidenced its potential in neurorehabilitation4 on the other hand, we are far from models and protocols that fit for every patient. The question arises if important information has been left unnoticed when looking at stroke-related network alterations: How can we further advance our network concepts? When and how do patients have to be recruited for specific brain stimulation protocols? Do we have to adjust our models and treatment assumptions and develop them toward individualization, for example, on specific clinical or imaging characteristics?
Structural analyses have predominantly investigated the influence of the integrity of the corticospinal tract (CST).2 Functional analyses have extensively investigated the importance of changes of local brain activity and inter-regional corticocortical connectivity between key areas of the human motor network that are the primary motor cortex (M1) and secondary motor areas in the frontal and parietal lobe. Particularly, the premotor areas of the frontal lobe, which are the dorsal and ventral premotor cortex (PMv), the supplementary motor area, and cingulate motor areas, might render one potential substrate for brain reorganization after stroke as they have direct access to M1, as well as to the spinal cord.5 For instance, studies have suggested reduced neuronal coupling between ipsilesional premotor cortices and M1 after stroke and that the reinstatement of normal connectivity over time is related to the amount of recovery.6 Likewise, the importance of the microstructural state of single corticocortical networks between premotor areas and M1 has also been demonstrated by structural imaging.7 Only few studies have investigated to what degree specific corticocortical structural8 or functional9–12 networks relate to motor output when taking the integrity of the CST as the main outflow tract of the human motor network into account: It has been shown that—independent from the influence of the CST—microstructural properties of specific premotor–motor connections, especially between M1 and the PMv, might be relevant for motor output in chronic stroke.8
Strikingly though, it has been largely neglected whether the influence of the corticocortical motor network on motor function might be a generalizable finding or whether it might depend from the state of the CST? The detection of such cross-network interactions might help to better understand brain reorganization patterns. Additionally, it might provide novel insights toward patient stratification, for example, brain stimulation protocols in the context of high intersubject variability of behavioral responses and responders and nonresponders.4,13–15 One previous study has already reported that interhemispheric coupling strengths between both M1 correlated with motor function only in patients with intact CST, whereas patients with disrupted CST did not show a similar association.16 Until now, structural cross-network interactions between corticocortical premotor–motor connections and the CST have not been investigated systematically in detail.
Continuing previous structural work on parietofrontal motor pathways as an important network for human motor control in healthy participants17 and patients with stroke,8,18–20 this study sought to investigate structural cross-network interactions between the CST and corticocortical motor pathways between M1, PMv, and the intraparietal sulcus (IPS).8 Specifically, the question was addressed whether the importance of specific tracts of the parietofrontal ipsilesional corticocortical network after stroke might depend on the microstructural integrity of the CST or whether these factors are independent. A large group of patients with subacute stroke was analyzed applying diffusion tensor imaging, probabilistic tractography, and template-based tract analysis to assess tract-related biophysical properties of white matter microstructure21 of the CST and 3 predefined intrahemispheric parietofrontal tracts between M1, PMv, and IPS. We hypothesized that the microstructural state of the different corticocortical connections would be associated with residual motor output depending on the level of damage to the CST.
Materials and Methods
Participants and Clinical Testing
Fifty-three patients (aged 60±11.2 years, 30 men; 4 left handed) with first-ever ischemic strokes (dominant hemisphere, 23) were recruited and tested 3 months after onset (mean, 99±12.4 days). Figure 1 gives an overview of the distribution of stroke locations with cortical, subcortical, and pontine lesions. Patients were evaluated by means of hand grip force (calculated as the ratio between the affected and unaffected hand) and the upper limb score of the Fugl-Meyer assessment. See Table I in the online-only Data Supplement for an overview of the demographic and clinical characteristics. Both measures were combined to one composite motor output score (MO) by principal component analysis (first principal component).6,8,22 For the reconstruction of the motor tracts, we used structural brain imaging data of 26 healthy older participants (66±10.1 years, 15 men, all right handed) from a previous study.22 The present study was approved by the local ethics committees (Hamburg, PV3777; Seoul, Institutional Review Board No. 2015-07-02). All participants gave written informed consent according to the Declaration of Helsinki.
Brain Imaging Sequences
In the patients with stroke, diffusion–weighted images were acquired using a 3T Phillips ACHIEVA magnetic resonance imaging (MRI) scanner (Philips Medical Systems, Best, the Netherlands). For the controls, diffusion–weighted and high-resolution T1-weighted anatomic images were available from a previous study and were acquired using a 3T Siemens Skyra MRI scanner (Siemens, Erlangen, Germany). Details of the sequence parameters are given in Text I in the online-only Data Supplement.
Image Processing and Probabilistic Tractography
The FSL software package 5.1 (http://www.fmrib.ox.ac.uk/fsl) was used to analyze the imaging data of the patients with stroke. In brief, after correcting for eddy currents and head motion, brains were skull stripped and fractional anisotropy (FA) maps were calculated for each participant fitting the diffusion tensor model at each voxel. The FA maps were then registered nonlinearly to the Montreal Neurological Institute standard space applying FSLs flirt and fnirt commands. Herein, stroke lesions were masked out and were not considered during the registration process.
The characterization of the CST and the 3 intrahemispheric connections between M1, PMv, and IPS bilaterally in the patients with stroke was conducted using structural and diffusion–weighted imaging data of 26 healthy controls taken from a previous study.22 Herein, for the CST originating from M1, normalized and binarized group average tract masks of varying thresholds of the left and right average CST at the level from the mesencephalon to the cerebral peduncle (Montreal Neurological Institute coordinates, z=−25 to −20) were already available. These masks were used to calculate the mean FA for the ipsilesional and contralesional CST in the patients with stroke. CST integrity was finally reported as proportional FA values (ipsilesional/contralesional tract; see Texts II and III in the online-only Data Supplement for details on the mask creation of M1 and the CST tractography).
In addition to the CST, the available data set of the controls was also used to reconstruct 3 intrahemispheric corticocortical connections between M1, PMv, and IPS. The cortical seed mask for M1—biased to hand function and standardized in size and relation to the cortical gray matter/white matter boundary8—was taken from the previous study.22 The same algorithm was now used to calculate additional PMv and IPS masks for both hemispheres (Text II in the online-only Data Supplement). Probabilistic tractography was conducted in each of the 26 controls to reconstruct connections between M1 and PMv, PMv and IPS and between M1 and IPS within each hemisphere. An established pipeline, whose details are given in the Text IV in the online-only Data Supplement, was used to calculate common group average trajectories for each of the connections of 4 different spatial extents, that is, thresholds (Figure 2). These trajectories were used to calculate mean tract-related FA values across the whole tract in each patient and finally reported as proportional FA values (ipsilesional/contralesional hemisphere). When overlapping, the stroke lesion was not excluded in congruence with previous reports.8 Consequently, mean proportional FA values reflected a combination of direct injury, as well as postinjury distant changes in fiber integrity. Because group average template tracts were used to read out individual tract FA values, proper congruence of the location and topography of the average tracts and the individual anatomy in the FA map was carefully checked before statistical analysis. In some participants, gyral anatomy particularly around the hand knob, the inferior precentral gyrus, and the postcentral sulcus was different from the Montreal Neurological Institute template; hence, some average tracts occasionally resided in sulci in single patients. Consequently, focal brain morphology was considered out of the normal range despite good overall registration to standard Montreal Neurological Institute and the proportional FA values of such tracts were omitted from the final analysis. Table I in the online-only Data Supplement summarizes which tract-related proportional FA values of which patients with stroke could be finally used in the analyses.
The data were analyzed using R statistical package 3.1.3. In a first analysis, linear mixed-effects modeling with repeated measures was used for the analysis of the proportional FA values of the different tracts of interest in the patients with stroke. Relevant white matter disintegrity was assumed when the proportional FA values were significantly lower than 1 (Text V in the online-only Data Supplement). In a second analysis, 3 separate multiple linear regression models were fitted using R’s lm for the proportional FA value of each of the 3 corticocortical tracts (TRACT), with MO (log-transformed) as the dependent variable and CST integrity, as well as the interaction CST*TRACT as independent variables. Nonsignificant interactions or main effects were kept given the primary focus of this study to investigate cross-network interactions after stroke. Cross-correlations between CST and corticocortical tract FA values were moderate (Pearson’s R, all ≤0.35). For both analyses, age, hemisphere (dominant or nondominant hemisphere lesioned), and time after stroke were included as additional independent variables to adjust the target effects.8,22 Estimated coefficients are given with 95% confidence interval. For visualization, mean correlations between TRACT and MO were estimated for 4 arbitrary CST integrities, that is, proportional FA values between 0.6 and 0.9 with higher values indicating better CST tract integrity. Behavioral scores are given as mean±SD. Statistical significance was assumed at P<0.05.
Probabilistic Tractography and Tract-Related FA
Probable trajectories between M1 and PMv, PMv and IPS and between M1 and IPS were successfully obtained in the healthy participants. Figure 2 shows the group averages of the tracts with reasonable spatial reproducibility across the participants. Highest spatial homogeneity of the tracts was found for M1–PMv with significant convergence of the trajectories toward M1 hand knob area and the anterior bank of the precentral gyrus for PMv. For PMv–IPS and M1–IPS, the variability in the trajectories toward IPS was higher, particularly for the left hemisphere. The trajectories for the CST have been already reported in a previous study.22 Tract-related white matter integrity was significantly affected by stroke in all 4 tracts (<1), estimated mean proportional FA were 0.82 (95% confidence interval, 0.78–0.86) for CST, 0.88 (0.84–0.93) for M1–PMv, 0.85 (0.80–0.90) for PMv–IPS, and 0.92 (0.87–0.97) for M1–IPS. M1–IPS was significantly less affected than CST (P=0.02; corrected by false discovery rate). The other proportional FA values did not show significant differences. Tract-related FA values for both hemispheres are given in Table II in the online-only Data Supplement.
Interactions Between Corticospinal and Corticocortical Networks and Their Impact on Residual Motor Output
The main finding was a significant cross-network interaction between CST integrity and the microstructural state of the connection between M1 and PMv for residual motor output in patients with stroke (P=0.014; Table). Figure 3 visualizes this interaction showing that M1–PMv integrity specifically contributes to MO in patients in which the CST is more affected by the stroke. In contrast, in patients with more preserved CST integrity, M1–PMv integrity does not explain a relevant amount of variance in MO. A similar strong trend was also observed for PMv–IPS (P=0.064). The model M1–IPS did not show significance in regard to this interaction between the CST and the corticocortical pathway.
The main finding of this study was a significant cross-network interaction between the CST and the corticocortical connection between M1 and PMv in relation to residual motor output after stroke. The data indicate that the microstructural state of M1–PMv seems to play a relevant role for residual motor output in patients with stroke with a significant damage to the CST. In contrast, in patients with slightly affected CST, the microstructural state of M1–PMv did not seem to explain a relevant amount of variance in motor outcome.
This finding extends the present knowledge about the interrelationships between the microstructural states of the motor networks at the corticospinal and corticocortical level and motor output after stroke and might have also implications for our understanding of recovery processes and ultimately treatment protocols by means of noninvasive brain stimulation.
Network Alterations After Stroke
Previous structural imaging studies have mainly investigated the influence of the CST for motor output after stroke showing that better CST integrity relates to better outcome. More recently, structural alterations of corticocortical connections between primary and secondary motor areas, such as M1, the dorsal and ventral premotor areas, and also posterior parietal brain regions, have been assessed in more detail as well.2,8
Importantly though, these analyses have largely neglected potential interdependencies between the different networks at the corticospinal and corticocortical level in relation to motor outcome. Thus, the isolated consideration of single networks will lead to findings whose interpretation might be limited. Therefore, the simultaneous consideration of corticocortical connections, especially together with the CST as the major output pathway might provide novel insights into brain reorganization after stroke. Relevant questions arising are (1) to what extent the microstructural state of premotor–motor connections would add to explain variance to the known involvement of the CST and (2) whether the relevance of these premotor–motor connections would depend on the magnitude of the CST damage or whether it would be a rather general phenomenon. To date, only the first question has been experimentally addressed in few trials by investigating corticocortical networks under simultaneous consideration of the CST. For instance, a positive contribution to residual motor output in patients with chronic stroke has been reported for the microstructural state of ipsilesional parietofrontal pathways between the anterior IPS and PMv and particularly between PMv and M1, in fact in addition to that of the CST.8 In line, it has been also shown that the combination of CST injury and functional connectivity between ipsilesional M1 and the premotor cortex was a better marker of the residual motor status than either measure alone.12 A similar additive effect has been obtained for the combined analysis of CST integrity and interhemispheric connectivity.9 However, to date, the question whether the magnitude of damage to the CST would determine the relevance of corticocortical structural connections for motor output has not yet been addressed in detail.
Cross-Network Interactions Between Corticospinal and Corticocortical Networks After Stroke
The goal of the present study was to tackle this open question. We hypothesized that the contribution of corticocortical connections to motor output would depend on the state of the CST, which would go beyond a simple additive effect in correlating brain structure of different motor tracts with behavior. Indeed, such cross-network interaction was present for M1–PMv and CST, providing a novel insight into the relevance of premotor–motor connections for motor output after stroke. This effect seemed to be specific to M1–PMv and not a general phenomenon for any parietofrontal connections which supports the view of M1–PMv as an important pathway for recovered hand function after stroke.8 Previous studies have consistently reported that ipsilesional premotor areas such as PMv and their interplay with M1 are contributing to motor function, spontaneous recovery,6 and also motor learning after stroke.23 It has been commonly assumed that this interplay was related to the reorganization of cortical motor networks in patients with stroke with varying degrees of motor impairment.6,8 However, the present data now suggest that such relationships between premotor–motor connectivity and motor output might not hold true for all patients independent of the impairment of their CST and the level of motor deficit, respectively. In contrast, the interplay between premotor areas and M1 might be rather relevant in patients with significant damage to the CST. Notably, this would further strengthen our concept that secondary motor areas in the frontal lobe undergo plastic changes after stroke to support motor output particularly in those patients with the biggest demand for supportive cortical motor areas. Prospective future studies will have to validate our model arguing that patients with high lesion load to the CST but preserved microstructural state of M1–PMv would show superior residual motor function compared with patients with disrupted M1–PMv connectivity. In patients with well-preserved CST integrity, most likely also less affected in average, the state of M1–PMv fiber tracts would not influence residual motor output. The present evidence of significant structural cross-network interactions is also supported by a functional imaging study, which has found that interhemispheric coupling strength correlated with motor function only in patients with intact CST, whereas patients with disrupted CST did not show a similar association.16
Potential Implications of Cross-Network Interactions
The present findings might have interesting implications for the refinement of prediction models for stroke recovery, on the one hand, and, on the other hand, for research focusing on patient-tailored treatment protocols by means of noninvasive brain stimulation. First, there is already stimulating data for prediction models of motor recovery based on clinical scores and electrophysiological and structural measures of CST integrity.3,24,25 These models are designed to be workable in the clinical setting, as analyses of the CST as a potential biomarker can be easily conducted in automated frameworks. However, by focusing only on the CST, the prediction might be limited,25 and the inclusion of information on corticocortical networks and their interactions with the CST as additional biomarkers could improve these models, a matter that has to be tested prospectively. Second, noninvasive brain stimulation techniques have been increasingly developed to enhance motor recovery after stroke.4 To date, these approaches have been generally applied independently of specific patient characteristics. Neglecting them, which could have an impact on the selection of the stimulation site or the protocol, might be one possible explanation for the heterogeneous responses, even in the view of responders and nonresponders in brain stimulation trials in patients with stroke. Data based on the present analytic approach might provide a rational for patient stratification to predict individual responses to stimulation or to determine the best cortical site for an intervention. In fact, in terms of M1 stimulation—the most widely used approach to enhance motor recovery—responses are particularly weak in patients with serious impairment and high lesion load to the CST.26 Consequently, where should we stimulate severely affected patients with heavily disrupted CST? In line with the present results, it has been already discussed that premotor areas such as PMv might be alternative targets as these areas normally have a higher probability of survival, they might act via parallel corticospinal pathways bypassing the disrupted CST originating from M1 as defined at the beginning, and, for example, PMv might have a high potential for plastic reorganization.27 Indeed, first proof-of-concept studies have indicated that this type of stimulation might be a safe and reasonable approach to enhance motor recovery in patients with rather severe deficits.28 Based on the present data, we would hypothesize that facilitating stimulation of ipsilesional PMv, for example, by means of anodal transcranial direct current stimulation would result in significant behavioral effects particularly in patients with disrupted CST on the one hand and high tract-related FA of M1–PMv on the other hand. In these patients, only PMv stimulation might drive adaptive plastic alterations both at the level of preserved corticospinal fiber tracts originating from PMv and also at the level of residual premotor–motor fibers targeting intact M1 neurons. In contrast, in patients with either well-preserved CST or high lesion load to the CST and concurrently disconnected M1 and PMv, we hypothesize that PMv stimulation might only show marginal effects. Accordingly, facilitating ipsilesional M1 stimulation might be preferentially efficient in patients with well-preserved CST, independent from the microstructural state of M1–PMv. Prospectively tested, such models might pave the way of precision medicine by tailoring interventional strategies for motor recovery after stroke.29
Some important limitations of the present study are worth to consider. First, the present study was conducted in the subacute stage after stroke. Whether the present findings of relevant cross-network interactions also hold true for earlier stages or the chronic phase of recovery are open questions for future research. Second, group average templates of corticocortical and CSTs from healthy controls were used to infer tract-related white matter properties of different motor pathways in a large group of patients with stroke. Probabilistic and averaged in nature, these tracts might differ to individual brain anatomy on the one hand. On the other, such approaches will allow us to perform the necessary more complex analyses on larger samples. Third, based on clear a priori hypotheses,8 we focused on the CST and defined parietofrontal motor connections at the cortical level. The present findings are not exhaustive, and the state of other parts of the motor network, such as frontal or prefrontal parts, and other cortico-subcortical or cortico-cerebellar circuits might also influence comparable analyses.
Sources of Funding
This research was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government, Ministry of Science, ICT and Future Planning (MSIP, NRF-2017R1A2A1A05000730 to Dr Kim).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.016834/-/DC1.
- Received January 27, 2017.
- Revision received June 7, 2017.
- Accepted July 6, 2017.
- © 2017 American Heart Association, Inc.
- Koch P,
- Schulz R,
- Hummel FC
- Kang N,
- Summers JJ,
- Cauraugh JH
- Nicolo P,
- Ptak R,
- Guggisberg AG
- Schulz R,
- Frey BM,
- Koch P,
- Zimerman M,
- Bönstrup M,
- Feldheim J,
- et al
- Buch ER,
- Rizk S,
- Nicolo P,
- Cohen LG,
- Schnider A,
- Guggisberg AG
- Cunningham DA,
- Varnerin N,
- Machado A,
- Bonnett C,
- Janini D,
- Roelle S,
- et al
- Koch PJ,
- Hummel FC