Learn more: PMC Disclaimer | PMC Copyright Notice
Dopaminergic mesolimbic structural reserve is positively linked to better outcome after severe stroke
Abstract
The concept of brain reserve capacity has emerged in stroke recovery research in recent years. Imaging-based biomarkers of brain health have helped to better understand outcome variability in clinical cohorts. Still, outcome inferences are far from being satisfactory, particularly in patients with severe initial deficits. Neurorehabilitation after stroke is a complex process, comprising adaption and learning processes, which, on their part, are critically influenced by motivational and reward-related cognitive processes. Amongst others, dopaminergic neurotransmission is a key contributor to these mechanisms. The question arises, whether the amount of structural reserve capacity in the dopaminergic system might inform about outcome variability after severe stroke. For this purpose, this study analysed imaging and clinical data of 42 severely impaired acute stroke patients. Brain volumetry was performed within the first 2 weeks after the event using the Computational Anatomy Toolbox CAT12, grey matter volume estimates were collected for seven key areas of the human dopaminergic system along the mesocortical, mesolimbic and nigrostriatal pathways. Ordinal logistic regression models related regional volumes to the functional outcome, operationalized by the modified Rankin Scale, obtained 3–6 months after stroke. Models were adjusted for age, lesion volume and initial impairment. The main finding was that larger volumes of the amygdala and the nucleus accumbens at baseline were positively associated with a more favourable outcome. These data suggest a link between the structural state of mesolimbic key areas contributing to motor learning, motivational and reward-related brain networks and potentially the success of neurorehabilitation. They might also provide novel evidence to reconsider dopaminergic interventions particularly in severely impaired stroke patients to enhance recovery after stroke.
Asmussen et al. report that larger volumes of the dopaminergic key areas amygdala and nucleus accumbens are associated with a more favourable outcome after severe ischaemic stroke. This suggests that structural reserve of the mesolimbic system might play a role in recovery processes after severe stroke.
Graphical Abstract
Introduction
Stroke is one of the leading causes of persistent functional impairment worldwide.1 The outcome after stroke crucially affects patients’ quality of life and plays an increasing role in the public health of aging societies.2 Despite a substantial improvement in acute stroke treatment, a relevant proportion of stroke survivors is left with severe deficits. The amount of functional improvement over time varies between individual patients, the translation of current recovery models, obtained in cohorts with rather mild to moderate deficits, to patients with larger symptom burden remains unsatisfactory.3,4 There is still a relevant gap of knowledge about why some patients with severe deficits recover more than others, despite comparable clinical or demographic properties. Previous studies have aimed at exploring potential imaging5 or electrophysiological biomarkers to better understand outcome variability after a severe stroke. For instance, they have focused on lesion topography,6 early brain network alterations7-9 or motor cortical excitability10 to improve outcome models.
Aside from such parameters which relate more directly to the acute brain lesion itself, more recently, the field has moved towards the concepts of brain reserve capacity as a novel, informative measure for correlative or predictive outcome inference.11-13 According to this concept, individual structural and functional properties of the brain modulate the degree of impairment caused by a brain lesion by influencing the actual clinical manifestation.14 For instance, it might explain why younger stroke patients usually show a remarkably larger potential for recovery despite large and strategically located lesions. Global radiomics-derived surrogates of the pre-stroke, i.e. the pre-morbid state of the brain, have been successfully linked to outcome variability. Examples are relative brain age15-17 or more directly accessible estimates of brain health, such as white matter hyperintensity burden or the degree of brain atrophy.18-20 Besides these global parameters, other studies have focused on measures of focal structural brain reserve of distinct regions, such as the cerebellum21 or the cortex of the contralesional hemisphere.22 These analyses of brain reserve capacity have significantly added novel mechanistic insights into how pre-stroke anatomy of specific brain regions or networks might influence recovery processes.
The rehabilitation after stroke is a complex process. Aside from functional adaptation, the re-learning of lost functions and the acquisition of novel skills are critically involved.23,24 On their part, these mechanisms are influenced by motivational and reward-related cognitive processes.25,26 In a recent randomized controlled trial, feedback and monetary reward in virtual-reality arm training positively influenced the improvement in motor scores in subacute stroke patients.27 One important neuronal system contributing not only to the motor domain, such as planning and execution of voluntary movements but also to the cognitive domain, including learning, motivation, reward and memory processes, is the dopaminergic system.26,28-32 It comprises of four major pathways (Fig. 1): the nigrostriatal, the mesolimbic, the mesocortical and the tuberoinfundibular pathway.33 The nigrostriatal pathway consists of dopaminergic neurones in the substantia nigra which project to the caudate nucleus (CAU) and putamen (PUT). Alterations in this pathway are not only fundamentally involved in Parkinson’s disease34 but also in motor planning, motor control and habit formation in healthy participants.35,36 The ventral tegmental area (VTA) projects via the mesolimbic pathways primarily to the amygdala (AMYG), the nucleus accumbens (Nacc) and the hippocampus (HIP),37 and via the mesocortical pathway to multiple cortical areas, predominantly the medial prefrontal cortex (mPFC).31,38-40 The mesolimbic and the mesocortical pathway together are commonly referred to as the mesocorticolimbic pathway.31 Finally, the tuberoinfundibular pathway comprises the dopaminergic innervation of the hypothalamic arcuate nucleus.31,41 Collectively, multiple deep subcortical nuclei, limbic brain regions and the medial prefrontal cortex are structural key areas of the dopaminergic system. They have been linked to motor and cognitive functions including motivational and reward processes which could exert an importance for neurorehabilitation.
Dopaminergic pathways. This illustration gives an overview of the main dopaminergic pathways. Dopaminergic neurotransmission derives mainly from the pars compacta of the substantia nigra and the ventral tegmental area, but also from hypothalamus. Neurones from the substantia nigra project to the dorsal striatum including the putamen and the caudate nucleus whereas neurotransmission from the ventral tegmental area reaches limbic structures, such as the amygdala, the hippocampus, the nucleus accumbens and the media preofrontal cortex.
The question arises whether the amount of structural reserve capacity in the dopaminergic system might inform about outcome variability after a severe stroke, i.e. in patients with the highest needs to regain lost neurological functions. For this purpose, this study analysed imaging and clinical data of 42 severely impaired acute stroke patients. Imaging data were collected within the first two weeks after stroke. Brain volumetry was performed to estimate grey matter volumes for seven key areas of the human dopaminergic system along the mesocortical, mesolimbic and nigrostriatal pathways of the contralesional hemisphere. Ordinal logistic regression models were used to relate regional volumes to the future functional outcome after 3–6 months after stroke, operationalized by the modified Rankin Scale (mRS). Models were adjusted for age, lesion volume and the initial impairment. We hypothesized that larger volumes of dopaminergic key areas are positively related to a more favourable outcome after a severe stroke.
Materials and methods
Participant data
Clinical and imaging data taken from two independent cohorts of acute ischaemic stroke patients were re-analysed.7,42 Sixty-one patients were recruited for the first cohort (C1),42 and 30 patients for the second cohort (C2).7 All 91 patients were admitted to, and treated at the University Medical Center Hamburg-Eppendorf between 2012 and 2020. For proper integration of the two cohorts, inclusion criteria of both studies were combined. The final sample comprised of 42 more severely impaired stroke patients (C1: 22 patients, C2: 20 patients). A flow chart of cohort integration is given in Fig. 2. Inclusion criteria for both studies were age ≥ 18 years, first-ever unilateral ischaemic stroke, persistent motor deficit of the upper extremity including hand function and no history of severe psychiatric or neurological disease. Patients of cohort C1 were only included if Barthel Index was ≤30 or mRS > 3 at the acute stage (days 3–5). This approach of cohort integration was successfully applied by our previous studies.21,22 All patients underwent structural (T1-weighted) brain imaging (MRI) within the first two weeks after stroke (C1: days 3–5; C2: days 3–14). Baseline symptom burden was assessed using the National Institutes of Health Stroke Scale (NIHSS) at study inclusion (T1). In cohort C1, the follow-up timepoint (T2) was set after three months in the late subacute stage of recovery.43 Follow-up time in cohort C2 was set accordingly, but as clinical follow-up data for seven patients were only available six months after the event, T2 for these patients was set at six months post-stroke.7 Functional outcome at T2 was operationalized by the mRS, a measure of persistent global disability, in line with our previous reports.21,22
Inclusion criteria and study integration. The flowchart shows the integration process. Of the 91 patients that were considered in the beginning, 49 had to be excluded from the present analyses due to missing or insufficient data.
To compare brain volumetric data, 42 healthy controls of similar age and gender were included in this analysis. A pseudo-affected hemisphere was assigned to each control accordingly to account for the distribution of the lesion across right and left hemispheres. Informed consent was provided by all participants or their legal guardian in line with the ethical declaration of Helsinki. Original studies were granted permission by the local ethics committee of the Chamber of Physicians Hamburg (PV3777, PV5442 and PV5357).
Brain imaging and voxel-based morphometry
All participants underwent brain imaging at the same 3T Skyra MRI scanner (Siemens, Erlangen, Germany). A 32-channel head coil and a magnetization-prepared rapid gradient echo sequence was used [parameters: time repetition (TR) = 2500 ms, time echo (TE) = 2.12 ms, flip angle = 9°, 256 coronal slices with a voxel size of 0.8 × 0.8 × 0.9 mm3, and field of view (FOV) = 240 mm]. T2-weighted images were performed to help delineate stroke lesions by using a fluid-attenuated inversion recovery sequence applying TR = 9000 ms, TE = 86 ms, time to inversion = 2500 ms, flip angle = 150°, 43 transversal slices with a voxel size of 0.7 × 0.7 × 3.0 mm3, and FOV = 230 mm as parameters. The Computational Anatomy Toolbox (CAT12)44 was used for voxel-based morphometry of the imaging data. The pipeline includes the steps of normalization, segmentation, data-quality-estimation and smoothing. First, the data were segmented according to the CAT12 manual. Individual images were displayed in SPM12 to check for possible gross anatomical abnormalities. To improve spatial registration, the images were centred manually at the anterior commissure and oriented to the anterior–posterior commissure line. In the next step, the images were segmented into white matter, grey matter and cerebrospinal fluid. The DARTEL algorithm was used for spatial normalization and modulation of the segmented images.44,45 Next, the images were transformed to the Montreal Neurological Institute (MNI) standard space. Finally, the normalized, modulated grey matter segmentations were smoothed with an 8-mm full-width at half-maximum Gaussian kernel to improve the signal-to-noise ratio. The total intracranial volume (TIV) was estimated by summing the volumes of grey matter, white matter and cerebrospinal fluid. Grey matter volumes of the following seven key structures of the dopaminergic system along the mesocorticolimbic and the nigrostriatal pathways were estimated based on AAL3 atlas46 data included in CAT12: nucleus accumbens, medial prefrontal cortex, amygdala, hippocampus, substantia nigra (pars compacta) (SNpc), caudate nucleus and putamen. For the medial prefrontal cortex, the label of the medial frontal gyrus was used. Data were obtained from the unaffected, contralesional hemisphere only to exclude any direct lesion effects.
Statistical analysis
R version 4.2.1 was used to conduct the statistical analyses (r-project.org). Gender and handedness were compared between stroke patients and healthy controls using χ2 tests, mean age and size of the regional volumes were compared via unpaired two-sided t-tests. For volumetry-outcome inference in the stroke patients, we used ordinal logistic regression models fitted via ‘polr’ from the MASS package.47 The dependent variable was mRS at timepoint T2. The independent variables of interest were the regional volumes at timepoint T1 (one separate model for each region), dichotomized by median split into larger and smaller regional volumes, in line with our previous reports to improve statistical power and reduce the risk of influential points.21,22 Age, TIV, stroke lesion volume and NIHSS at timepoint T1 were added as covariates. Age and TIV were included after linear residualization against the volumes of interest to address multi-collinearities. Lesion volumes were log10-transformed to improve data distribution. Odds ratios (ORs) for the variables of interests were calculated along with the confidence intervals (CI, 95%) and the P-values. An OR < 1 would correspond to a lower risk for scoring one level higher in mRS in patients with larger volumes in comparison to patients with smaller volumes. All tests were conducted with a two-sided alpha level of 5% and corrected for multiple comparisons using the method of false discovery rate (noted as PFDR). A leave-one-out model analysis (LOOA) was finally applied to validate the robustness of the final results. The results were defined as robust if the model remained significant in all single LOOA iterations at a P < 0.05. Significances are indicated by asterisks. For sensitivity analyses, surviving correlative outcome models were re-computed while iteratively taking various potentially confounding factors into account: (i) side of the lesion including its interaction with the regional volume to account for hemispheric specialization of basal ganglia48-50 and possible lateralization effects51 and differential plasticity rates52 known from traumatic brain injury patients; (ii) peak width of skeletonized mean diffusivity (PSMD) of the unaffected hemisphere as a marker of cerebral small vessel disease burden53; and (iii) influence of initial symptom severity onto volumetry–outcome relationship as operationalized by the interaction between NIHSS (dichotomized by median split into higher and lower NIHSS scores) and regional volume. For specificity analyses, we computed additional outcome models which included only global—and not regional—parameters of structural reserve that are (i) TIV; (ii) averaged grey subcortical matter volume; and (iii) averaged cortical grey matter volume of the unaffected hemisphere.
Results
Demographical and clinical data
Data from 42 more severely impaired patients were analysed. The mean age of the stroke patients at examination was 70.8 years (standard deviation, SD = 12.4). Two (4.8%) patients were left-handed, 20 patients (47.6%) were female. The mean age in the control group was 70.3 years (SD = 10.1) and gender distribution was identical. There was no significant difference neither in age (P = 0.83), gender (P = 0.83) nor handedness (P = 0.79) between patients and controls. Twenty-six (61.9%) patients suffered from left-sided lesions. Mean lesion volume was 52.2 mL (SD = 68.8 mL, range: 0.6–303.3 mL). Median NIHSS at T1 was 8.5 (interquartile range, IQR 5.25–13.00), median mRS at T2 was 3 (IQR 1.25–4.00). Peak width of skeletonized mean diffusivity data for sensitivity analyses were available for 37 patients. Table 1 summarizes the demographic and clinical data. Individual patient data are given in Supplementary Table 1 and clinical characteristics by lesion side are shown in Supplementary Table 2. The topographic lesion distribution is shown in Fig. 3.
Lesion distribution. Heat map of stroke lesion distributions. Lesions overlay superimposed on a T1-weighted image in MNI standard space. Z values are shown for each axial slice. The colour represents the number (N) of patients affected per voxel. The letter R indicates the right side of the brain.
Table 1
Demographic and clinical data of the stroke patients
Parameter | Value |
---|---|
Age (mean, SD) | 70.8, 12.4 |
Handedness (left, right) | 2, 40 |
Gender (male, female) | 22, 20 |
Lesion side (left, right) | 16, 26 |
Lesion volume mL (mean, SD) | 52.2, 68.8 |
NIHSS at T1 (median, IQR) | 8.5, 5.25–13 |
mRS at T2 (median, IQR) | 3, 1.25–4 |
Systematic overview of demographic and clinical data of the included patients including mean or median dependent on the parameter.
Volumetry at baseline
Mean volumes of the seven regions of interest are shown in Table 2 along with the corresponding median for subsequent group separation into patients exhibiting larger and smaller volumes. Group comparison was conducted to assess any possible structural alterations present already at timepoint T1 which would complicate the interpretation of volumes as ‘pre-stroke’ brain reserve. There were no significant group differences in TIV, mesocorticolimbic structures, substantia nigra and caudate nucleus of the nigrostriatal pathway. Only the putamen showed a significant volume reduction at T1 in the stroke patients (mean 3.6 mL) when compared to controls (mean 4.1 mL, PFDR = 0.032, Table 2).
Table 2
Group comparison of averaged volumes of interest
Pathway | Region/variable | Stroke mean [Median] | Control mean [Median] | P FDR |
---|---|---|---|---|
Total intracranial volume | 1519.216 [1517.300] | 1479.242 [1454.415] | 0.301 | |
Mesocorticolimbic | Nucleus accumbens | 0.740 [0.758] | 0.773 [0.777] | 0.262 |
Medial PFC | 11.946 [11.900] | 12.483 [12.105] | 0.262 | |
Amygdala | 1.147 [1.202] | 1.219 [1.232] | 0.195 | |
Hippocampus | 3.583 [3.633] | 3.664 [3.572] | 0.590 | |
Nigrostriatal | Substantia nigra, pars compacta | 0.013 [0.010] | 0.013 [0.010] | 0.701 |
Caudate nucleus | 2.969 [2.978] | 3.201 [3.220] | 0.195 | |
Putamen | 3.619 [3.594] | 4.058 [4.162] | 0.032a |
Mean [with median] baseline volumes (mL) of the mesolimbic structures of the unaffected hemisphere and the TIV are compared between stroke patients and controls. Notably, median values were used for group separation in patients exhibiting larger and smaller volumes.
aIndicates that the volume of the putamen was found to be significantly different between stroke patients and healthy controls.
Volumetry–outcome relationship
Ordinal logistic regression models showed significant associations between baseline grey matter volumes of the nucleus accumbens and the amygdala, and functional outcome after stroke. Specifically, patients exhibiting larger volumes in both regions showed a reduced odds of scoring higher on mRS at follow-up 3–6 months after stroke when compared to patients with smaller volumes (Nacc: OR = 0.176, PFDR = 0.025, AMYG: OR = 0.183, PFDR = 0.025, Table 3, Fig. 4). This association was independent of age, TIV, lesion volume and NIHSS at baseline. For additional sensitivity analyses, we re-computed the two winning models while considering several potentially confounding variables. First, possible hemispheric specialization of basal ganglia48-50 and possible lateralization effects51 and differential plasticity rates52 known from traumatic brain injury patients were addressed by including the interaction lesion side × regional volume. This interaction was not significant, neither for the nucleus accumbens (P = 0.78), not for the amygdala (P = 0.53) which argues against side effects regarding both the mesolimbic network assessed and the affected stroke hemisphere. Second, cerebral small vessel disease burden53 was assessed by incorporating PSMD data into the model. This did not alter the significant results for the nucleus accumbens and the amygdala. Third, the two winning models were extended by an interaction between NIHSS at baseline (binarized into higher and lower NIHSS scores) and the regional volume. This interaction was not significant (both P = 0.22) which argues against a dependence of the associations on initial symptom severity within the present cohort of severely impaired patients. The interplay between the volumes of the amygdala and the nucleus accumbens is illustrated in Supplementary Figure 1. To address region specificity of the present findings for the nucleus accumbens and the amygdala, we additionally explored the influence of global parameters of structural reserve to functional outcome, such as TIV, averaged grey subcortical matter volume and averaged cortical grey matter volume of the unaffected hemisphere. However, none of these measures showed a significant association (all P > 0.25).
Volumes of the nucleus accumbens and the amygdala and outcome after stroke. Stacked histograms illustrate the distribution of mRS after 90 days for patients exhibiting bigger volumes (VOLBig, upper rows) and smaller volumes (VOLSmall, lower rows) for the nucleus accumbens (A1) and the amygdala (B1). The mRS ranges from 0 to 6: 0: no symptoms; 1: no disability despite symptoms; 2: slight disability, unable to carry out all activities; 3: moderate disability, help required; 4: moderately severe disability, unable to walk; 5: severe disability, bedridden; 6: dead. Raincloud plots depict individual volumes for the nucleus accumbens (A2) and the amygdala (B2) for both volume groups. For comparison, volumes for healthy controls are given in A3 and B3. Representative examples for big and small volumes for the nucleus accumbens (A4) and theamygdala (B4) are depicted by individual T1-weighted MRI images overlaid with the segmented regions of interest with the corresponding participant’s ID.
Table 3
Structure–outcome relationship
Pathway | Region | OR (95% CI) | P FDR |
---|---|---|---|
Mesocorticolimbic | Nucleus accumbens | 0.176 (0.043, 0.634) | 0.025a |
Medial PFC | 0.401 (0.120, 1.277) | 0.142 | |
Amygdala | 0.183 (0.049, 0.619) | 0.025a | |
Hippocampus | 0.292 (0.080, 0.981) | 0.108 | |
Nigrostriatal | Substantia nigra, pars compacta | 1.957 (0.608, 6.485) | 0.260 |
Caudate nucleus | 0.397 (0.118, 1.268) | 0.142 | |
Putamen | 0.345 (0.100, 1.116) | 0.133 |
Estimated ORs with 95% CIs for one level scoring higher in mRS at follow-up after 3–6 months after stroke in patients with larger volumes in mesocorticolimbic and nigrostriatal dopaminergic regions compared to patients with smaller regional volumes. Results are corrected for age, lesion volume, total intracranial volume and NIHSS at baseline.
aIndicates that larger volumes were significantly associated to favourable outcome.
Discussion
The main finding of the present study was a significant positive association of larger volumes of the nucleus accumbens and the amygdala, key regions of the mesolimbic dopaminergic system, and a more favourable functional outcome 3–6 months after severe ischaemic stroke. This association was independent of the degree of initial impairment, age, lesion volume and TIV.
These findings add further evidence to the concept of structural reserve in the context of ischaemic stroke which proposes that larger reserve is associated with enhanced potential to recover.11,14 According to this concept, higher reserve equals increased neural substrate which in turn might influence both spontaneous neural repair, compensation and recovery mechanisms following stroke.11,54 Thus, the assessment of brain reserve may help to address and understand the differences in individual recovery from stroke and ease stroke outcome prediction.13
Structural reserve may relate to the global and focal level. On the global level, previous work has successfully linked the degree of white matter hyperintensities to worse functional outcome.18,19 On the focal level, other studies have shown that larger volumes of specific cerebellar lobules involved in motor and cognitive functioning, such as lobule IV, VI and VIIIB,21 or larger thicknesses of contralesional motor and non-motor cortices22 also carry valuable information regarding outcome variability after a severe stroke. The present results further add to these data by showing that structural reserve of distinct brain regions, the amygdala, and the nucleus accumbens, involved in dopaminergic neurotransmission across the motor and cognitive domain, particularly affecting motivational learning and reward processes, positively correlates with a more favourable outcome after stroke.
Our findings agree with previous animal data, studies in healthy participants and patient data. The nucleus accumbens is one of the key areas involved in motivation.55 In macaques, the recovery of dextrous finger movements after spinal cord injury at the cervical level was diminished by reversible pharmacological inactivation of the nucleus accumbens.29 In chronic stroke patients, smaller ipsilesional volumes of the nucleus accumbens correlated with worse sensorimotor functions.56 In acute stroke patients, lesions to this nucleus were associated with higher symptom burden and attenuated responses to specific training paradigms.57 Reward-related processes are another important domain influenced by dopaminergic neurotransmission via the mesocorticolimbic pathway, especially its mesolimbic part which includes the nucleus accumbens, the amygdala and the hippocampus.38,58 Reward correlates with better motor learning,59,60 procedural learning,61 and motor adaptation in healthy patients.62 In stroke patients, motor adaptation is also typically impaired. The addition of reward or punishment has been shown to improve motor adaptation in stroke patients.63 Finally, punishment has been reported to be strongly connected to fear memory formation, which depends on endogenous dopamine release in the amygdala.64 Collectively, these findings argue that reward and punishment-related processes could be influential for recovery after stroke. Our structural data extend these behavioural findings and show that surrogates of brain reserve in key areas along the underlying mesolimbic dopaminergic pathways might help to understand inter-subject variability in long-term outcome after stroke.
Mechanistically, there are two very interesting questions: First, would better reserve in dopaminergic key areas might enhance motivation, susceptibility to reward or punishment, and the compliance to rehabilitation on a behavioural, psychological level? Previous studies have shown that the degree of motivation in stroke patients promotes engagement and adherence, and it may increase the therapeutical effect of neurorehabilitation therapy.25,26 In patients with chronic traumatic brain injury, bilateral deep brain stimulation of the nucleus accumbens has shown to primarily impact on behavioural and emotional adjustment through prefrontal cortex modulation, which in turn improved functional independence even years after the injury.65 Second, would dopamine more directly act on cerebral reorganization and plastic processes? Ischaemic strokes alter the function of the dopaminergic system.31 For instance, dopamine release into the ipsilesional striatum shortly after stroke has been reported by some studies in rodents.66,67 The amount of released neurotransmitters has been found to correlate with stroke severity, outcome and age.68 Studies in gerbils and primates have suggested a negative link between higher releases of dopamine and a worse outcome.68,69 However, other authors have observed that spontaneous recovery in mice was linked to increased levels of dopamine in the contralesional hemisphere. Thus, they have argued that dopamine could be a promotor of recovery after stroke.70 This argumentation would be in line with animal data on motor cortex plasticity and motor skill learning71 or reports on traumatic brain injury which highlighted that dopaminergic neurotransmission is critically altered after the injury and dopamine levels and dopaminergic interventions have been linked with reductions in oxidative stress and improvement in cellular functions, neuronal survival and functional outcome.72,73 Given these considerations, the present data might indicate that enhanced dopamine reserve within the unaffected hemisphere may protect against stroke-induced damage and promote neuro-plastic processes on the cellular level.
So far, the translation of these promising preclinical data into the clinical application has been proven to be difficult. Dopaminergic interventions by drug therapy in clinical stroke cohorts have provided inconclusive results rendering a dopamine medication to improve recovery still a treatment approach which lacks clear evidence.74-76 Our present findings in more severely impaired stroke patients suggest that the response to dopaminergic interventions might depend on the level of impairment. Thus, future clinical trials might be warranted to include pre-specified subgroup analyses to explore the importance of impairment for treatment responses systematically.
Against our hypothesis, we did not find any significant associations between regional volumes along the mesocortical and nigrostriatal pathways and outcome. In rats, lesions to the medial prefrontal cortex have been reported to lead to selective but not generalized deficits in executive function.77 In humans, dopaminergic projections have been linked with higher cognitive functions such as top-down executive functions, self-knowledge, updating goal-directed behaviour and prediction of reward and motivation.78-80 On a speculative note, the outcome measure mRS as a global marker of disability could explain the absence of further associations. Future, prospective research focusing on more elaborated measures of activity of daily living or cognitive functioning are needed to investigate the importance of brain reserve of the dopaminergic systems for functional recovery after stroke in more detail. The same might hold true for the regions of the nigrostriatal pathway. Intuitively, this pathway appears even more relevant for motor functioning. It is involved in motor planning,35 habit formation36 and reward-driven motor learning.28,55 Direct lesions to the basal ganglia have been shown to associate with poor sensorimotor behaviour and recovery after stroke.81,82 One study in well-recovered stroke patients found that the amount of damage to nigrostriatal white matter tracts, quantified three months after stroke, correlated with worse fine motor skills after one year.83 Hence, a more detailed motor assessment including hand, finger and reach-to-grasp functions might be particularly needed to uncover structure–outcome associations for key areas along this pathway. A combined analysis of severely and mildly impaired patients would be needed to further address an impairment-specific importance of nigrostriatal pathways for stroke recovery.
There are several limitations worth noting. First, besides brain reserve, cognitive reserve is considered a potential predictor of individual stoke outcome, commonly operationalized by years of education.14,17 In this study, relevant cognitive impairment was an exclusion criterium. However, most patients, particularly of cohort C2, had very severe deficits early after stroke. A detailed cognitive assessment was neither included, nor would it have been feasible in this population. To what extent unknown pre-morbid cognitive dysfunction might have biased the present findings remains unclear and further distinction between brain reserve and cognitive reserve was limited in this work. Second, we have argued that inter-subject variability in brain reserve of motivational and reward networks could influence engagement and adherence to therapy throughout the rehabilitation period after stroke. Unfortunately, we do not have any information regarding the achieved amount of rehabilitation between T1 and T2. The inclusion of such additional information might alter the present modelling results. Future studies could extend our models by including detailed information, e.g. regarding the amount of inpatient rehabilitative training after discharge, the access to therapy in the outpatient or home-based setting, or even regarding the family background including daily support and motivation. Third, the primary lines of argumentation followed the concepts of dopaminergic motivational and reward processes. However, particularly the amygdala has been also linked, while exhibiting smaller volumes, to major depression84 and post-stroke depression.85 Unfortunately, to what extent features of depression might have influenced the present global recovery trajectories cannot be answered as detailed information was not acquired in the present cohorts. In turn, such structure–behaviour relationships might also indicate that early volume assessments in key areas of psychiatric diseases might enable personalized management targeting comorbidities, such as the occurrence of post-stroke depression and treatment regimes. Fourth, the small sample size could influence specificity and sensitivity negatively. To address this, we aimed at increasing the statistical power by dichotomization of the volume data in line with previous reports and fully corrected the results for multiple testing. Fifth, due to technical limitations, we could not include all areas of the dopaminergic system. The ventral tegmental area was excluded as it was not possible to extract the volumes properly. Sixth, the present analysis only focused on the unaffected hemisphere to exclude any direct lesion effects, in line with our previous analyses of contralesional cortices.22 Finally, 35 patients (83.33%) had lesions directly affecting ipsilesional dopaminergic regions, thus key regions of the mesolimbic pathways (Nacc: N = 12; mPFC: N = 13; AMYG: N = 24; HIP: N = 17; SNpc: N = 16; CAU: N = 34; PUT: N = 35). Consequently, in these patients, the ipsilesional mesolimbic pathway was directly affected. The influence of these direct lesion effects for the outcome inference remains unclear. This topic might be explored by future analyses systematically, potentially including tract-based markers of structural brain reserve of the dopaminergic system of the ipsilesional and contralesional hemisphere.
Conclusion and future perspectives
The present study shows that larger volumes of distinct key areas of the dopaminergic network are positively associated with a better outcome after severe stroke. The data might suggest a link between the structural state of mesolimbic key areas contributing to motor learning, motivational and reward-related brain networks, and the success of neurorehabilitation, thereby promoting the emerging concepts of brain reserve capacity for stroke recovery to better understand outcome variability. Finally, the data could also provide novel evidence to reconsider dopaminergic interventions particularly in severely impaired stroke patients to enhance recovery after stroke given inconsistent effects in clinical trials until now.
Contributor Information
Liv Asmussen, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Benedikt M Frey, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Lukas K Frontzkowski, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Paweł P Wróbel, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
L Sophie Grigutsch, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Chi-un Choe, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Marlene Bönstrup, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany. University Medical Center Leipzig, Department of Neurology, 04103 Leipzig, Germany.
Bastian Cheng, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Götz Thomalla, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Fanny Quandt, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Christian Gerloff, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Robert Schulz, University Medical Center Hamburg-Eppendorf, Department of Neurology, 20246 Hamburg, Germany.
Supplementary material
Supplementary material is available at Brain Communications online.
Funding
This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) SFB (936—178316478—projects C1 to C.G., C2 to G.T.) and the Else Kröner-Fresenius-Stiftung (2016_A214 to R.S.) R.S. and C.C. are supported by an Else Kröner Exzellenzstipendium from the Else Kröner-Fresenius-Stiftung (2020_EKES.16 to R.S., 2018_EKES.04 to C.C.). F.Q. is supported by the Gemeinnützige Hertie-Stiftung (Hertie Network of Excellence in Clinical Neuroscience). We finally acknowledge financial support from the Open Access Publication Fund of UKE—Universitätsklinikum Hamburg-Eppendorf and DFG—German Research Foundation.
Competing interests
The authors report no competing interests.
Data availability
Data will be made available by the authors upon reasonable request.