Global multisensory reorganization after vestibular brain stem stroke

Abstract Objective Patients with acute central vestibular syndrome suffer from vertigo, spontaneous nystagmus, postural instability with lateral falls, and tilts of visual vertical. Usually, these symptoms compensate within months. The mechanisms of compensation in vestibular infarcts are yet unclear. This study focused on structural changes in gray and white matter volume that accompany clinical compensation. Methods We studied patients with acute unilateral brain stem infarcts prospectively over 6 months. Structural changes were compared between the acute phase and follow‐up with a group of healthy controls using voxel‐based morphometry. Results Restitution of vestibular function following brain stem infarcts was accompanied by downstream structural changes in multisensory cortical areas. The changes depended on the location of the infarct along the vestibular pathways in patients with pathological tilts of the SVV and on the quality of the vestibular percept (rotatory vs graviceptive) in patients with pontomedullary infarcts. Patients with pontomedullary infarcts with vertigo or spontaneous nystagmus showed volumetric increases in vestibular parietal opercular multisensory and (retro‐) insular areas with right‐sided preference. Compensation of graviceptive deficits was accompanied by adaptive changes in multiple multisensory vestibular areas in both hemispheres in lower brain stem infarcts and by additional changes in the motor system in upper brain stem infarcts. Interpretation This study demonstrates multisensory neuroplasticity in both hemispheres along with the clinical compensation of vestibular deficits following unilateral brain stem infarcts. The data further solidify the concept of a right‐hemispheric specialization for core vestibular processing. The identification of cortical structures involved in central compensation could serve as a platform to launch novel rehabilitative treatments such as transcranial stimulations.


Introduction
Acute central vestibular syndrome manifests with rotational vertigo, spontaneous nystagmus (SPN), tilts of the subjective visual vertical (SVV), and postural instability with lateral falls. Vestibular symptoms recover within weeks, a process known as central vestibular compensation. [1][2][3] Central compensation of a unilateral peripheral loss of vestibular function is based on multiple processes that occur in distributed neuronal networks at different locations and at different times. [4][5][6][7] Unique features within the vestibular system have to be taken into account: In contrast to other sensory systems, vestibular signals are integrated early with other sensory input in the lower brain stem, and thus cortical vestibular areas are always multisensory as they respond to several sensory stimuli. [8][9][10] The structural basis of compensation is the bilaterally organized vestibular system. The sensory signals are conveyed from the vestibular end organs to the vestibular nuclei in the pontomedullary brain stem and via several bilateral pathways with multiple crossings to the thalamus. From there, they reach the multisensory integration centers of the temporoparietal cortex. [11][12][13][14][15] The posterior insula with the parietal opercular cortex (OP2), the posterior insular and retroinsular cortex were reliably identified as the core regions of the multisensory vestibular network in humans. 14,[16][17][18][19] These regions show a right-hemispheric preponderance for vestibular signal processing in right-handed humans. 18 Some structural compensatory changes have been demonstrated for peripheral vestibular lesions. [20][21][22] Central compensation in central vestibular lesions has only been investigated in a few studies using PET and functional MRI. [23][24][25][26] The current study used voxel-based morphometry (VBM) to evaluate changes in gray matter volume (GMV) and white matter volume (WMV) over time in 24 patients with acute unilateral brain stem infarcts presenting with vestibular or ocular motor deficits, compared to the baseline (acute phase). 27 A group of healthy age-and gendermatched participants (HC) served as a control group.
There should be no differences in GMV and WMV between the HC and the patients in the acute phase, but changes were expected after 6 months. The following questions were addressed: (i) Are the structural changes dependent on the location of the infarct, that is, pontomedullary vs pontomesencephalic lesions? (ii) Which sensorimotor areas are particularly involved, those of the multisensory vestibular network only or also areas belonging to the visual and somatosensory system? This is important to evaluate the contribution of substitution to central compensation. (iii) Do the compensatory structural changes reflect the vestibular dominance of the right hemisphere and the upper brain stem as identified by functional imaging and functional connectivity MRI? Are the compensatory changes in the brain stem and cerebellum symmetric or asymmetric?

Standard protocols and procedures
The study was performed in accordance with the 1964 Declaration of Helsinki (latest applicable revision Fortaleza 2013) and approved by the institutional review board of LMU Munich, Germany (no.094-10). All patients gave informed written consent to participate in the study.

Patients
We included 24 patients with ischemic brain stem infarcts who presented to our tertiary referral center (University Hospital, LMU Munich, Germany) between 2012 and 2019. Inclusion criteria were as follows: Imaging confirmed unilateral brain stem infarct, ability to complete the detailed vestibular and ocular motor examination, completion of follow-up imaging, vestibular and ocular motor examination after 6 months. Exclusion criteria were as follows: Absence of an ischemic lesion on diffusion MRI, clinically confirmed peripheral vestibular deficit, bilateral or multifocal infarcts, prior stroke, tumor, cerebral hemorrhage, vascular malformation, edema (i.e., compression of CSF space, shift of midline structures), severe white matter hyperintensities (WMH, Fazekas grade> 1 for periventricular WMH and deep WMH), and if patients were unable to complete the neurological and neuro-ophthalmological examination due to cognitive impairment or impaired vigilance. 28 All patients received a complete clinical and radiological work-up in the acute (M0) and chronic (6-month follow-up, M6) stage. We had to exclude some patients due to loss to follow-up (n = 8); insufficient clinical data (n = 9); missing structural imaging/poor imaging data quality (n = 4).

Controls
We also examined a group of 38 age-and gendermatched right-handed healthy controls with an identical imaging protocol. The control population had no prior history of peripheral or central vestibular disorders.

Clinical examination
All patients received a thorough clinical and neuro-orthoptic examination, including measurements of the SVV. 11 In a subset of patients (n = 3 acute, n = 17 chronic) the SVV was determined using the bucket test (Table 1)  For the group with pontomedullary infarcts, analyses were conducted for those with spontaneous nystagmus (SPN, n = 9/15), deviation of the SVV (n = 9/15), and for patients with rotational vertigo (n = 9/15). We chose a dichotomous categorization for the analysis of SVV tilts (pathological/ not pathological). In the group of patients with pontomesencephalic infarcts only the patients with pathological tilt of the SVV (n = 7/9) were analyzed as a group because nystagmus and rotational vertigo were present in less than 50% of the cases and did not allow for further subcategorization. T contrasts were estimated to detect differences between the groups. The results were further analyzed using nonparametric permutation testing (threshold-free cluster enhancement, TFCE) as implemented in the CAT12 toolbox calculating 5000 permutations. 32 TFCE is threshold free, sensitive for high focal as well as widely distributed low effects (cluster enhancement), nonparametric, and does not interfere with focal changes in smoothness. 33 All results were corrected for multiple comparisons on the cluster level using family-wise error (FWE) correction; p < 0.05. Changes in GMV and WMV were projected onto a MNI152 template brain using MRICROGL (https://www.mccauslandcenter.sc.edu/ mricrogl/).

Interhemispheric differences of volumetric changes in homologous brain regions
To account for interhemispheric differences of signal changes, we compared cluster size (in voxels) and peak signal intensity (T score) on a whole brain level (all significant clusters in the left vs right hemisphere). In the second step, we compared only those homologous brain areas that process vestibular information in both hemispheres (cerebellum, brain stem, thalamus, insular and parietal opercular cortex (this includes areas Ig1, Ig2, retroinsular cortex, OP), cingulate cortex, intraparietal sulcus (IPS)/superior parietal lobule (SPL); Wilcoxon signed-rank test, P < 0.05).

Data availability statement
The dataset is not publicly available due to European Privacy laws and lack of consent for publication by the patients.

Sociodemographic
There was no significant difference in age or handedness between the patient and control groups: median 68 years (range: 28-86 years) in the patient group vs 68 years (51-79 years) in the control group; 100% right-handed in the patient group vs. 97.3% (HC).

Clinical
Infarcts were termed as pontomedullary using the final MRI lesions (pontomedullary infarcts extended to the vestibular nuclei complex). Fifteen patients had pontomedullary infarcts without evidence of a concurrent cerebellar lesion and nine patients had pontomesencephalic infarcts. Patients with affection of the vestibular pathways to the ocular motor centers in the rostral midbrain and collateral paramedian thalamic lesions were included in the group of pontomesencephalic infarcts (n = 2). One pontomesencephalic and one pontomedullary infarct extended into the territory of the other group. In both cases > 80% of the infarct lay in the territory to which it was assigned. Clinical deficits were not significantly different between the groups and were compensated at M6 (no SPN and no rotational vertigo, one borderline pathological but significantly improved SVV score in each group of pontomedullary and pontomesencephalic infarcts without clinical symptoms, Table 1). Unilateral infarcts were distributed similarly on both sides in the group of pontomedullary infarcts (n = 8 left, n = 7 right). The majority of the nine pontomesencephalic infarcts were left sided (n = 6). There was no size difference between left-and right-sided infarcts (pontomedullary: Mann-Whitney U-Test P = 0.779; pontomesencephalic: P = 0.30, Fig. 1).

GMV
The patterns of changes in GMV and WMV for patients with SPN were similar to those with rotational vertigo. While there were no differences between patients in the acute phase (M0) and the HC, increases in GMV at M6 compared to the HC were found in the parietal opercular cortex and postcentral gyrus of the right hemisphere only. GMV decreases were located bilaterally in the cerebellar hemispheres (crus I, lobule VI, VIIa/b) and cerebellar vermis (lobules X, IX, VIIIa,b), the pulvinar, the anterior thalamic nuclei (ANT) extending to the mediodorsal nucleus (MD) bilaterally and in the premotor cortex (cytoarchitectonic areas 6d, 6mc, 6mr). Additional decreases were found along the ventral visual stream bilaterally (hOC 1-4v, area FG3 extending to the hippocampus; Fig. 2A, Fig. 3, for results of GMV in patients with rotational vertigo, see summary WMV WMV increases were located within the parietal opercular cortex (around cytoarchitectonic areas TE1, Ig1, Ig2, OP2, OP3 includes the retroinsular cortex, see above) and adjacent to the postcentral gyrus in both hemispheres with larger clusters in the right hemisphere. Clusters extended to the posterior parietal cortex in both hemispheres. Subcortical WMV increases were found from the flocculus and cerebellar hemispheres via the superior cerebellar peduncle (SCP) and the cerebello-thalamocortical tract to the parietal cortex of both hemispheres. While increases were rather symmetric in the cerebellum, there was a preponderance of right-sided increases in the upper brain stem and parietal opercular areas (Fig. 4C).   Figure 3. Pontomedullary infarcts. GMV decreases at follow-up after 6 months in the group of patients with spontaneous nystagmus compared to the control group (n = 9). GMV decreases were located in premotor cortex (6d, 6mc, 6mr) and ventral visual streams (hOC1-4v; thresholded at P < 0.001, FWE corrected for visualization). ANT, anterior thalamic nuclei; MD, mediodorsal nucleus; hOC, human occipital cortex. cytoarchitectonic area OP2, Ig1 and Ig2, TE1 (includes the retroinsular cortex, see above) of the right hemisphere, and the posterior cingulate cortex (CSv) as well as the corticospinal tract (CST) bilaterally. Additional clusters were found in the WM beneath the IFG and the MT + region bilaterally (see Fig. 5B for a detailed depiction).

Interhemispheric differences in cluster size and peak signal increases
Volumetric increases were larger in the right compared to the left hemisphere (P = 0.012, Wilcoxon signed-rank test) for pontomedullary and pontomesencephalic infarcts (Table 2A) on a whole brain level. However, when considering the known central vestibular sites alone, only the clusters in the parietal opercular cortex showed this effect (P = 0.012). When using the peak T score intensity, only the difference between right and left parietal opercular cortex was significant (Table 2B).

Discussion
The main findings of the study are as follows: (i) Downstream structural volumetric changes following brain stem infarcts take place in multiple sensory and motor regions in both hemispheres. These changes accompany clinical compensation. This was evident in patients with pontomedullary and pontomesencephalic infarcts alike. (ii) The volume increases in multisensory vestibular cortical areas showed a right-hemispheric preference. (iii) Compensation of spontaneous nystagmus and rotational vertigo in pontomedullary brain stem infarcts was accompanied by GMV and WMV increases in the core cortical vestibular areas and ventral parts of the postcentral gyrus in the right hemisphere only. (iv) Compensation of graviceptive dysfunction (i.e., SVV tilts) led to large supplementary GMV and WMV increases bilaterally in the parietal and postcentral (somatosensory) cortex and along the white matter tracts that connect the parietal opercular cortex and intraparietal sulcus with the premotor cortex. (v) Volumetric increases were located primarily in multisensory areas in pontomedullary infarcts. In pontomesencephalic infarcts, additional increases were found in motor and middle temporal areas. (vi) Volume decreases after pontomedullary brain stem infarcts involved the visual and motor systems.

Central compensation of vestibular syndromes
A few studies have already demonstrated cortical changes following chronic peripheral vestibular lesions. 21,22 Studies on central compensation of central vestibular lesions are scarce. Bense and colleagues found signal decreases in visual cortex in pontomedullary infarcts and decreases in the premotor cortex in both pontomedullary and pontomesencephalic infarcts with functional imaging. 25,26 However, structural plasticity, that is, an increase of GMV and WMV in the cortical multisensory vestibular areas following unilateral brain stem infarcts has not been demonstrated before. Apart from the inherent differences in these two methods, we used a thorough state-of-theart preprocessing and analysis algorithm that allows subtle changes in GMV to be detected and is corrected for type 1 errors (TFCE with FWE correction for multiple comparisons) which could additionally explain the differences between ours and the former studies.
Significantly, we found a right-hemispheric dominance for GMV and WMV increases in the parietal opercular multisensory vestibular and somatosensory areas for patients with pontomedullary infarcts presenting with the unique vestibular symptoms of spontaneous nystagmus and rotatory vertigo. This is in line with the known lateralization of the vestibular system after caloric irrigation with a dominance of the right hemisphere in right handers. 18,34 For SVV tilts, signal increases were also more pronounced in the right cerebral cortex but involved additional bilateral cortical and subcortical regions. This implies that spontaneous nystagmus and rotatory vertigo represent core vestibular dysfunction, whereas the perception of verticality-an otolith and multisensory achievement-is compensated in the vestibular and somatosensory areas bilaterally with a predominance of some vestibular areas within the right hemisphere. [35][36][37][38] In other words, the structural changes following infarcts that lead to SVV tilts require the activation of bilateral cortical multisensory areas (See summary figure 6). The findings

Volume decreases in visual cortex
Decreases were present in GMV of the visual system (pulvinar, ventral visual stream). A decrease in glucose metabolism and BOLD signal in visual cortex areas has been repeatedly shown in the studies of vestibular stimulation in healthy volunteers and in patients with unilateral peripheral and central vestibular lesions. 17,18,39 This was attributed to the attempt to minimize a visuo-vestibular mismatch in visual perception caused by oscillopsia due to nystagmus or by the divergent input in the two sensory systems. 39 In the current analysis, these changes were also evident on a structural level in the chronic phase, that is, when nystagmus had ended much earlier. It seems that an acute lesion-induced visual vestibular mismatch and visual vestibular reciprocal interaction cause structural reorganization of visual and vestibular multisensory brain areas over time.

Volume decreases in premotor areas and increases in structural multisensory WM connectivity
Additionally, we found decreases in the anterior thalamic nuclei and the premotor cortex in close proximity to but not limited to the FEF (cytoarchitectonic areas 6v, 6d, 6mr, 6mc). A possible explanation could be a profound reduction of voluntary head and neck movements in the acute phase of severe vertigo and spontaneous nystagmus, since these symptoms are aggravated by head movements. While there was a reduction in GMV in these areas, the central vestibular multisensory cortical areas and WM pathways mediating perception of the body in space (superior longitudinal fascicle, SLF) showed a positive response. The SLF provides an anatomical link between the parietal lobe and premotor cortex and is involved in ocular motor coordination, attention, and visuospatial processing, all of which need vestibular and other sensory input to compute maps for spatial orientation. 40,41 After the infarct-induced partial loss of vestibular information, a strengthening of these multisensory links is required. This might represent a compensatory perceptual processing strategy for the patients' disturbance of stance and gait.

Differences between pontomedullary and pontomesencephalic infarcts
Structural reorganization following graviceptive deficits in pontomedullary infarcts was confined to somatosensory and multisensory cortical areas bilaterally. In contrast, pontomesencephalic infarcts with tilts of the SVV produced far more heterogeneous adaptive changes including frontal, parietal, and middle temporal areas, as well as the striatum. The mean deviation of the SVV between both groups was similar as has been demonstrated before. 42

RH
A P LH P Figure 6. Differing GM reorganizational response size and location in vestibular subtypes of pontomedullary infarcts. Patients with a deviation of the SVV at stroke onset (green) compared to the depiction of the structural follow-up response in patients with pathological SVV deviation and SPN (yellow), and the response of patients with rotatory vertigo as the initial symptom (blue). Areas associated with the compensation of "pure" vestibular symptoms were located in the right parietal opercular cortex only while areas associated with multisensory integrative function (SVV) showed a bihemispheric distribution along somatosensory cortex and intraparietal sulcus. The different volumetric changes could be due to the more "integrated" nature of the vestibular pathways in the upper brain stem where the vestibular signals are transformed from a velocity to a position signal. [43][44][45] This signal is further integrated in the thalamus and cortex where it is needed for spatial orienting and navigation and the modulation of motor output.
With respect to lateralization, an effect of lesion site (R vs L) has to be accounted for, because the majority of pontomesencephalic infarcts was left sided.

Limitations
We were not able to differentiate between compensatory changes following right-sided vs left-sided brain stem infarcts. However, in the case of pontomedullary infarcts, we found a strong right-hemispheric dominance of volumetric changes where the infarcts were equally distributed between both sides. Therefore, we do not expect significant effects of lesion side in the brain stem on cortical vestibular compensation. Still, a sufficiently powered statistical analysis of this effect would be an interesting topic for further analysis. Furthermore, we were not able to compare our data with infarcts that did not elicit vestibular or ocular motor deficits separately, probably due to the high degree of interconnection of the two systems at the level of the brain stem. Based on our clinical data in which all patients included in the study suffered from some degree of oculomotor or vestibular dysfunction, this problem may even turn out to be impossible to separate in future research, since these pathways are running through the whole brain stem. Despite the long period of patient recruitment for our study, we were not able to include a substantial enough number of left-handed patients to warrant a dedicated analysis. The effect of handedness on central vestibular compensatory processes therefore remains unclear.
Further, the use of different scanners presents a bias that is owed to the long recruiting period. However, patients and controls were evenly balanced over the scanners and all patients received their longitudinal MRIs on the same scanner. Furthermore, the raw 3D resolution of the sequences and field strength was identical. We obtained high data quality estimates over the complete sample and good to very good signal homogeneity for the different tissue types in the quality control evaluation as part of the CAT12 toolbox. A side effect to further reduce the role of a scanner-effect bias is our application of rigorous permutation testing which would have resulted in a null-finding if the signal quality (noise level) between the scanner types had differed much. Therefore, while the use of different scanners represents a potential limitation of the study, we are confident that it did not bias our results in a negative way.
To correct for the moderate sample sizes in the respective groups we applied rigorous methodological scrutiny in the chosen methods (TFCE, FWE correction for multiple comparisons).
Lastly, we had to exclude patients with bilateral or multiple infarcts within the vertebrobasilar territory and those who needed prolonged mechanical ventilation. Therefore, there is an inherent bias to smaller infarcts and lesser clinical symptoms in our patient sample.

Conclusions
This study, for the first time, demonstrates substantial neuroplasticity in both hemispheres along with the clinical compensation of vestibular deficits following unilateral brain stem infarcts.
For patients with incomplete remission of vestibular symptoms, for instance, noninvasive brain stimulation of the right parietal opercular cortex could be an interesting treatment option to boost cortical compensation.