Altered Functional Connectivity Between Cortical Premotor Areas and the Spinal Cord in Chronic Stroke

Abstract

BACKGROUND:

Neuroscience research has contributed significantly to understanding alterations in brain structure and function after ischemic stroke. Technical limitations have excluded the spinal cord from imaging-based research. Available data are restricted to a few microstructural analyses, and functional connectivity data are absent. The present study attempted to close this knowledge gap and assess alterations in corticospinal coupling in chronic stroke and their relation to motor deficits.

METHODS:

In this cross-sectional study, patients with chronic stroke and healthy controls underwent corticospinal functional magnetic resonance imaging while performing a simple force generation task at the University Medical Center Hamburg-Eppendorf between September 2021 and June 2023. Task-related activation was localized in the ipsilesional ventral premotor cortex, the supplementary motor area, and the cervical spinal cord. Psycho-physiological interactions and linear modeling were used to infer functional connectivity between cortical motor regions and the cervical spinal cord and their associations with clinical scores.

RESULTS:

Thirteen well-recovered patients with stroke (1 woman, 12 men; mean age, 62.6 years; mean time after stroke: 47.6 months) and 13 healthy controls (5 women, 8 men; mean age, 64.5 years) were included. The main finding was that ventral premotor cortex and supplementary motor area showed topographically distinct alterations in their connectivity with the spinal cord. Specifically, we found a reduced coupling between the supplementary motor area and the ipsilateral ventral spinal cord and an enhanced coupling between the ventral premotor cortex and ventral and intermediate central spinal zones. Lower supplementary motor area and higher ventral premotor cortex–related spinal cord couplings were correlated with residual deficits.

CONCLUSIONS:

This work provides first-in-human functional insights into stroke-related alterations in the functional connectivity between cortical premotor areas and the spinal cord, suggesting that different premotor areas and spinal neuronal assemblies might be involved in coupling changes. It adds a novel, promising approach to better understanding stroke recovery and developing innovative models to comprehend treatment strategies with spinal cord stimulation.

Imaging-based systems neuroscience research has significantly contributed to our current understanding of alterations in brain structure and function after ischemic stroke and how they relate to deficits and recovery processes. Apart from cognitive and language functions, deficits in the motor domain, particularly affecting the upper limb, are important sequelae after stroke, impeding private, and professional rehabilitation.¹ Structural² and functional imaging^3,4 have built a robust body of evidence showing how acute stroke lesions affect brain activation and multisite communication patterns, as well as the structural integrity of the human motor network and its key areas and interconnecting pathways comprising the primary motor cortices (M1), frontal secondary motor areas including the ventral premotor cortex (PMV), the supplementary motor area (SMA), subcortical brain regions, and the cerebellum. The corticospinal tract (CST), the motor network’s central outflow system, has been chiefly addressed structurally at various cerebral levels via lesion load,⁵ fiber count,^6,7 and diffusion imaging–based analyses of white matter microstructure.^2,8 Recently, advanced structural magnetic resonance imaging (MRI) extended the field of view toward the brain stem and the cervical spinal cord regarding structural CST alterations.⁹ Such CST data were collectively reported to be critically linked to stroke recovery. Notably, studies showed that not only CST fibers originating from M1 but also secondary CST originating from nonprimary premotor areas, such as PMV and SMA,¹⁰ show structure-outcome associations.^8,11–15 It has been repeatedly discussed that such premotor CSTs might mediate corticospinal neurotransmission while bypassing lesioned CST components.^16–20 For instance, higher SMA activity in electroencephalography²¹ and stronger structure-outcome associations for M1-PMV pathways^17,22 were found mainly in patients with more CST damage. It has been speculated that these alternate CST trajectories might act on lower motor neurons directly or, more likely, indirectly via modulation of spinal interneuronal assemblies and circuits.^23,24 After earlier reports on the spinal cord and circuit dysfunctions after stroke²⁵ and the first report of successful epidural cervical spinal cord stimulation for improvements of chronic poststroke upper limb paresis just recently,²⁶ the interest in the spinal cord after cortical stroke experienced a remarkable renaissance.^27–29 Animal data had already suggested that the spinal cord might be involved in recovery after cortical motor stroke.^30–33 However, functional data on spinal cord activations and functional connectivity with cortical motor networks are unavailable. Such analyses would be of great value to better understand stroke recovery in general and shed light on such groundbreaking innovative treatment protocols with spinal cord stimulation, in particular.

Advancements in functional MRI (fMRI) of the spinal cord^34,35 and simultaneous cortical and spinal cord fMRI^36,37 opened a new window to investigate functional corticospinal networks in humans with high spatial resolution.³⁸ In the motor domain, studies have just started to explore corticospinal functional connectivity during finger or hand movements in healthy participants.^35,39,40 In our previous study, we could link ventral spinal activation during simple hand movements not only to M1 activation but also to PMV activation.⁴⁰ This added first fMRI data to sparse transcranial magnetic stimulation data,^41,42 which suggested that human premotor areas might be functionally connected to the lower cervical spinal cord, likely interneurons, and contribute to distal upper limb functions in humans.

The present study was designed to explore alterations in cortical and spinal cord activation and changes in corticospinal coupling during a simple visually guided force generation task for the first time in human stroke survivors. Task-related activation was localized in ipsilesional PMV, SMA, and the cervical spinal cord. Task-related spinal cord activation and psycho-physiological interactions were inferred to assess functional connectivity between cortical premotor regions and the cervical spinal cord. Linear modeling was used to address their associations with motor deficits.

Methods

Data Availability

All data needed to evaluate the conclusions in the article are present in the article and the Supplemental Material. Processed brain and spinal cord activation data and demographic data to reproduce the findings are available from the corresponding author upon reasonable request.

Subjects

Thirteen patients with stroke and 13 age-matched healthy controls without any neurological damage unrelated to healthy aging were included in the analysis. Participants were right-handed and provided informed consent following the Declaration of Helsinki. The study was approved by the local ethics committee of the Medical Association of Hamburg (PV6026). The measurements were performed between September 2021 and June 2023 at the University Medical Center Hamburg-Eppendorf. See Supplemental Material for more details.

Motor Task

A simple motor task was used comprising repetitive whole-hand grips in a block design, as previously introduced in detail.⁴⁰ The patients with stroke performed repetitive, visually guided, almost isometric whole-hand grips with their affected hand with 3 varying predefined force levels. The healthy controls performed the task with the right or left hand corresponding to the affected side of the matched patients (pseudoside). See Supplemental Material for details.

Behavioral Data

A JAMAR Hand Dynamometer (built by Patterson Medical, Warrenville, IL) measured the maximum whole-hand grip force of both hands. Further standardized tests included the National Institutes of Health Stroke Scale, the Fugl Meyer assessment of the upper extremity (UEFM), the modified Rankin Scale, and the Nine-Hole Peg Test (NHP). Healthy controls underwent grip force measurement and NHP testing. Relative NHP values and relative grip force (grip ratio) were calculated by dividing the values of the affected and unaffected sides or pseudosides.

MRI Data Acquisition

A 3T Prisma MRI scanner (Siemens Healthineers, Erlangen, Germany) and a 64-channel combined head-neck coil were used to acquire cerebral and spinal imaging data. The MRI protocol was identical to the previously described MRI protocol.⁴⁰ Detailed sequence parameters are described in the Supplemental Material.

MRI data preprocessing and first-, and second-level analyses are similar to our previously described methods.⁴⁰ Detailed information is available in the Supplemental Material (including Figure S1).

Psycho-Physiological Interaction

Psycho-physiological interaction models were implemented with seed regions in ipsilesional PMV and SMA to assess the functional connectivity between the brain and the spinal cord. The time course was extracted from spherical regions of interest with a radius of 2 mm around the individual peak voxel for each area in each subject (Table S1). Psycho-physiological interaction models were calculated in the spinal cord with the extracted time course of each brain region. See the Supplemental Material for further statistical details, including information for additional sensitivity analyses.

Further Statistical Analyses

The statistical package R 4.3.1⁴³ was used for statistical analysis. Linear models were used to analyze age, NHP ratio, grip ratio, and maximum grip force for group-specific differences, with side as an additional confound of no interest. Two-sided t tests were used to compare the exerted forces between the groups.

Results

Demographic, Clinical, and Task Data

Thirteen patients with chronic stroke (12 males, 1 female, all right-handed, aged 62.6±9.7 years [mean±SD]) and 13 healthy controls (8 males, 5 females, all right-handed, aged 64.5±11.9 years) were included in this work. Clinical characteristics are given in Table S2. There were no significant group differences for age (P=0.65) or maximum grip force (P=0.75). The grip ratio was significantly lower in the patients with stroke (P=0.03), and the NHP ratio was numerically higher among patients with stroke (P=0.056). A topographical map of the distribution of stroke lesions is shown in Figure 1. During the motor task, the exerted target forces for patients with stroke were 42.0±8.5%, 61.0±8.4%, and 76.7±10.97% for low, medium, and high levels, respectively. For controls, they were 43.7±12.2%, 60.2±8.0%, and 77.9±6.5%, respectively, with no group differences. Mean performance (stroke: 1.24±0.16, controls: 1.26±0.21) was also not significantly different (Figure S2). Please see the Supplemental Material for details.

Figure 1.

Topography of stroke lesions. All masks of stroke lesions are projected on the left hemisphere, overlaying a T1 template in Montreal Neurological Institute (MNI) space (z coordinates below each slice). The color intensity indicates the number of subjects whose lesion voxels lie within the colored region.

Spinal Cord and Cortical Activation During Force Generation

Force generation across all 3 force levels led to a significant activation on the group level, primarily in the ipsilesional (=ipsilateral to the measured/affected hand) spinal cord at the C6 and C7 vertebral level, corresponding to the C7 and C8 spinal segments in both groups (Figure 2A and 2B). The patients with stroke exhibited a higher activity than controls, localized in the middle of the spinal cord cross-section between vertebra C6 and C7 (Figure 2C). The cluster characteristics are listed in Table S3. In patients with stroke, increasing force levels increased the spatial extent of blood oxygenation level–dependent responses from a focal activation at the C7 vertebral level toward more distributed spinal activations between C5 and C7 (Figure S3A). In the control group, a similar increase was observed from a focal activation at the C6 vertebral level during low forces towards more distributed spinal activations between C5 and C7 (Figure S3B).

Figure 2.

Topography of spinal cord and cortical brain activation during force generation. A and B, Group mean spinal activation during the task with maximum (max.) grip force, side, and age as additional confound parameters: (A) control group and (B) stroke group. C, Group comparison [stroke>control], controlled for max. grip force. D and E, Group mean activation during the task with max. grip force, side, age as additional confound parameters: (D) control group and (E) stroke group. F, Group comparison (stroke>control) controlled for max. grip force. Z maps were thresholded by Z>2.4 (group mean) and Z>2.0 (group comparison), with a cluster significance threshold of P<0.05; spinal cord activations are overlaid on the PAM50_t2-template. Cerebral activations are rendered on a T1 template in Montreal Neurological Institute (MNI) space. A indicates anterior; au, arbitrary units; CL, contralesional; IL, ipsilesional; and P, posterior.

Cerebral activity on group level was detected in both groups across force levels primarily in the ipsilesional primary sensorimotor cortex comprising M1 and the primary sensory cortex S1, in bilateral SMA, bilateral dorsal premotor cortex, bilateral PMV, and a widespread activation in posterior parietal cortices along the intraparietal sulcus (Figure 2D and 2E). Group comparisons revealed higher SMA and bilateral intraparietal sulcus activity in the stroke group (Figure 2F). The results of the force-level specific analyses are shown in Figure S4. Cluster characteristics of the task activation are listed in Table S4. Additional analyses included potential relationships between motor impairment and mean brain and spinal activation data (Figure S5).

Corticospinal Connectivity During Force Generation

Psycho-physiological interactions were computed to investigate task-specific functional corticospinal coupling. Compared with controls, the connectivity was enhanced between ipsilesional PMV and an intermediate spinal zone at C5 and ventral areas at the C6 vertebral levels (Figure 3B). The correlation analysis between functional connectivity and motor impairment revealed 1 cluster in the ventral spinal cord at the C6 level, in which an increase in connectivity was correlated with lower UEFM scores and, therefore, more severe motor impairment (Figure 3C). For the more force-related impairment, assessed via grip ratio, 1 cluster was found in the area of the intermediate zone at the upper C6 level, in which more severe motor impairment was related to an increase in connectivity. In addition, for the correlation with fine motor skills, assessed via NHP ratio, 1 additional cluster was found in the area of the intermediate zone, also at the C6 level (Figure 3C). A sensitivity analysis excluding the 3 most severely affected patients with stroke (UEFM<55) supports the results with higher connectivity in the patients with stroke in the ventral/intermediate zone at C5 and an increasing and even positive coupling strength in patients with worse fine motor skills and worse force-related motor impairment, that is, higher NHP ratios, and lower grip ratio, at C6 (Figure S6). Cluster characteristics are listed in Tables S5 and S6.

Figure 3.

Alterations in corticospinal coupling during force generation for ventral premotor cortex (PMV). Psycho-physiological interaction analysis between ipsilesional PMV (A) and the spinal cord and overview of the slices displayed along the z axis. B, Group comparison stroke (S) > control (C), controlled for maximum grip force. C, Patients with stroke: correlation with motor impairment, controlled for maximum grip force, age, side; (Z maps were thresholded by Z>2.0 [group comparison] and |Z|>2.4 [correlation analysis], cluster significance threshold of P<0.05). Spinal cord activations are overlaid on the PAM50_t2-template, and Montreal Neurological Institute (MNI) coordinates are given. All images are in radiological orientation. The coupling strengths in spinal cluster voxels (MNI coordinates as indicated) were extracted and displayed in B and C. au indicates arbitrary units; NHP, Nine-Hole Peg Test; and UEFM, Fugl Meyer assessment of the upper extremity.

For SMA, patients with stroke exhibited significantly reduced corticospinal connectivity comprising ventral horn areas and surrounding white matter at ipsilesional lower C6 vertebral level and contralesional C5 vertebral level (Figure 4B). In the correlation analysis with motor impairment, we detected an ipsilesional ventral location in which reduced, that is, negative connectivity, was correlated with more severe deficits (Figure 4C). This correlation was particularly evident for the NHP ratio and grip ratio. For UEFM, this correlation was also evident, but only when the Z threshold was reduced. The sensitivity analysis without the 3 most severely affected patients with stroke supports the results with lower connectivity in the patients in the ventral horn area at C6 (Figure S7). The correlation analysis with grip ratio and UEFM showed even a significant relationship in the described regions without these severely affected patients. Additional cluster results are listed in Tables S5 and S6.

Figure 4.

Alterations in corticospinal coupling during force generation for supplementary motor area (SMA). Psycho-physiological interaction (PPI) analysis between ipsilesional SMA (A) and the spinal cord and overview of the slices displayed along the z axis. B, Comparison control (C) >stroke (S), controlled for maximum grip force. C, Patients with stroke: correlation with motor impairment, controlled for maximum grip force, age, side; Z thresholds, and image presentations are identical to Figure 3 (exception: |Z|>1.9 for the correlation between PPI-SMA and Fugl Meyer assessment of the upper extremity [UEFM]). au indicates arbitrary units; and NHP, Nine-Hole Peg Test.

Discussion

The main finding of the present study was that patients with chronic stroke show alterations in activation and connectivity between cortical motor areas and the spinal cord while performing a simple force generation task with the affected hand. Concerning connectivity, topographically different and clinically relevant changes were found, which comprised the connectivity between PMV, SMA, and the spinal cord. They were opposite in the group comparison: for SMA, we found a reduced coupling with voxels localizing in the ventral spinal cord at lower C6 vertebral levels. For PMV, an enhanced coupling was detected with ventral regions at the C6 vertebral level and intermediate central spinal zones at the C5 level. Finally, lower SMA–related and higher PMV–related spinal cord coupling located at the C6-level, most likely contributing to the ipsilateral ventral areas, were directly correlated with persistent motor deficits. These data provide first-in-human functional insights into possible corticospinal network alterations after stroke, adding a novel, promising functional perspective to the emerging field of spinal cord analyses and stimulation in stroke recovery research.

The present work is based on previous animal and human imaging and electrophysiology data, arguing that the spinal cord may be considered an essential, however largely neglected, factor in modulating recovery after cortical stroke. For instance, previous animal work has evidenced an upregulation of spinal structural plasticity, neurotrophins, and cytokines.^44,45 Recent transcriptome analyses found gene upregulations related to neurite outgrowth in stroke-denervated spinal gray matter, particularly in its intermediate laminae.³⁰ One study reported spinal axonal sprouting and microglial activation to support presynaptic site formation,³¹ potentially occurring in central spinal regions where axonal outgrowth has been consistently documented in stroke animal models.^32,46 Interestingly, at the group level, we also found increased spinal cord activation in similar central voxels. Furthermore, stronger PMV coupling in central voxels was correlated with greater deficits. Speculative in nature, such areas might correspond to intermediate zones such as lamina X or, even more specifically, interneurons from V0 families, which were reported to contribute to complex corticospinal circuits connecting motor and sensory cortical influences with spinal neuronal assemblies.^23,24 Animal studies suggest that these assemblies are involved, for example, in left-right coordination⁴⁷ or aberrant excitation during the progression of amyotrophic lateral sclerosis.⁴⁸

As a second topographically interesting finding, alterations in corticospinal coupling were detected for areas localizing towards the ipsilateral ventral spinal cord. Particularly for SMA, the close spatial proximity between voxels exhibiting reduced connectivity and voxels exhibiting a correlation with behavior was notable. Ipsilateral ventral horn areas would align with the location of lower motor neurons or V1 interneuron populations comprising multiple cell types, including Renshaw cells and Ia inhibitory interneurons.^49–51 It remains elusive if any, or which, neuronal component might be mainly related to the associations in the present fMRI datasets. At least from a simplified clinical view, the patients included in this analysis did not show hand spasticity in hand/finger flexion and extension. Larger samples with varying degrees of spasticity could allow for regression modeling of this valuable covariate. Combined approaches with electrophysiology, such as single-fiber electromyography, could add to the regression of trans-synaptic degeneration of spinal motor neurons after stroke.⁵²

Recent human data have convergingly evidenced, potentially against previous assumptions derived from nonhuman primates,^53–56 that PMV and SMA might be structurally⁵⁷ and functionally connected^40,41 not only with upper segments of the spinal cord but also with lower cervical spinal cord contributing to distal upper limb functions. These connections will likely be maintained by poly synaptic connections along spinal interneural routes, including the propriospinal system.⁵⁸ The present data builds on this knowledge. They confirm that premotor areas may be functionally connected to the lower spinal cord and provide evidence that poststroke connectivity is subject to changes associated with persistent clinical deficits. Unfortunately, we can only speculate about the meaning of the directions of alterations and linear associations. For instance, the negative coupling between SMA and the ventral spinal cord could be interpreted as an inhibitory process during the execution of the task.⁵⁹ How this coupling might mechanistically explain impairment remains open for discussion and controversy.^3,4,21,60

Functional analyses of corticospinal coupling and spinal cord activation might help to better understand stroke recovery in general. This approach might be promising to comprehend the effects of spinal cord stimulation after stroke, especially in light of recent breakthrough results demonstrating improvements in arm and hand paresis in 2 patients with chronic stroke.²⁶ It has been speculated that spinal cord stimulation might act via the recruitment of primary afferents, which provide excitatory input to motoneurons and interneurons directly connected to such afferents. Thereby, spinal stimulation might increase the responsiveness to residual cortical inputs, for example, from preserved primary and nonprimary CST or alternate descending pathways, including the reticulospinal tract.^2,9,61,62 Enabling researchers to assess changes in corticospinal activation and connectivity with sufficient spatial resolution, as illustrated in our study, corticospinal fMRI could contribute to disentangling mechanisms, such as increases in motoneuron excitability, postsynaptic inhibition, or sensory gating, which have been discussed as potential mechanisms for spinal cord stimulation.²⁷

It can be argued that the most severely affected patients significantly influenced the analysis results. For this reason, we performed a sensitivity analysis in which patients with UEFM <55 were excluded. However, this analysis showed that the group comparison results in the ventral areas for SMA and PMV remained stable, although the ventral region was slightly at a higher vertebral level for PMV in the sensitivity analysis. The correlation for PMV in the intermediate zone and SMA in the ipsilateral ventral spinal cord with grip ratio was also stable, indicating that the connectivity changes also occur in less severely affected patients and that grip ratio seems to be a stable parameter to assess the dependency between connectivity and motor impairment even in mildly affected patients during a force generation task.

There are several other significant limitations to note. First, the small sample size must be considered when interpreting the results. The correlation results, in particular, should be interpreted cautiously as larger sample sizes are generally required for reliable estimates. Despite technical advances, spinal cord imaging remains a time-consuming challenge, particularly for older participants. Inclusion criteria were rather strict; more severely impaired patients could not perform the task correctly; thus, they could not participate in this study. Future studies on larger sample sizes are needed to verify the present results and may help to generalize them to the broader population. It should also be mentioned that there was only 1 woman in the group of patients with stroke. This might have biased the results of the group comparisons and connectivity-outcome associations and complicate the generalizability of the findings to women. Second, the design of this study was cross-sectional. A longitudinal approach could give further insights into the temporal dynamics of stroke-related corticospinal coupling and activation changes. Third, due to the technically limited resolution, smoothing, and structure of the spinal cord, the spatial localization of the activated spinal regions can only be indicated approximately; further studies will be necessary to verify the results presented here. Congruent with our previous work in younger, healthy participants⁴⁰ and given broad evidence suggesting that PMV and SMA are key areas that undergo stroke-related changes in brain structure and function,^2–4 we focused our analyses on these cortical areas. Analyses of corticospinal connectivity changes related to other cortical regions, such as insular and sensory cortices or posterior parietal areas along the intraparietal sulcus that may contribute to the corticospinal system could be interesting topics for future analyses. Subcortical nodes of the human motor network, such as the basal ganglia, the cerebellum, and the brainstem, could not be included as seed areas in this work because of the reduced fields of view due to technical restrictions. Fourth, the Z thresholds used under 3.1 can increase the false positive rate above 5%. However, reduced Z thresholds have already been used in previous publications.^35,39 Fifth, the spinal resolution was limited, complicating the inference of precise topographical interpretations regarding spinal neuronal assemblies.

Conclusions

Collectively, this work provides first-in-human insights into stroke-related alterations in the functional connectivity between cortical motor areas and the spinal cord. The analysis of the results suggests that different premotor areas and spinal neuronal assemblies might be differentially prone to and involved in coupling changes. This study adds a novel, promising approach to better understanding stroke recovery in general and developing innovative models to comprehend groundbreaking treatment strategies with the spinal cord.

Article Information

Acknowledgments

Drs Braaß and Gerloff contributed to conceptualization. Dr Braaß, J. Feldheim, Dr Wolf, and Dr Chu participated in investigation. J. Feldheim, and Drs Braaß, Finsterbusch, Chu, Büchel, and Tinnermann developed the methodology. Drs Braaß and Schulz contributed to formal analysis. Drs Braaß and Schulz contributed to writing—original draft preparation. Drs Braaß, Schulz, Tinnermann, Büchel, Gerloff, Finsterbusch, Chu, and Wolf contributed to writing—review and editing.

Sources of Funding

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) SFB (Sonderforschungsbereich) 936-178316478-C1 (Dr Gerloff) and A6 (Dr Büchel), DFG Grant TI 1110/1-1 (Dr Tinnermann), and DFG Grant 459728725 (Dr Schulz). Dr Schulz was additionally supported by an Exzellenzstipendium from Else Kröner-Fresenius-Stiftung (2020_EKES.16).

Disclosures

Outside of the submitted work, Dr Gerloff reports consulting fees from Abbott Canada, Amgen, AstraZeneca, Bayer Healthcare, Boehringer Ingelheim, Daiichi Sankyo Europe, Novartis, and Prediction Biosciences; he reports funding outside of the submitted work from the Deutsche Forschungsgemeinschaft (DFG), German Aerospace Center (DLR), European Union, Gemeinnützige Hertie-Stiftung. Dr Tinnermann reports funding from the DFG (German Research Foundation). The other authors report no conflicts.

Supplemental Material

Supplemental Methods

Tables S1–S6

Figures S1–S7

References 63–76

STROBE Checklist