Cortical and subcortical gray matter changes in patients with chronic tinnitus sustaining after vestibular schwannoma surgery

Tinnitus is attributed to partial sensory deafferentation resulting in a central maladaptive neuroplasticity. Unfortunately, the agent of deafferentation is usually unknown or irreversible. In patients with unilateral vestibular schwannoma (VS), however, the auditory nerve is affected by a benign tumor. Hence, removal of the tumor can cease the tinnitus. In turn, sustaining complaints after surgery indicate cortical neuroplasticity. The present study is a cross sectional study which aims to track cortical structural changes by surface-based morphometry in 46 VS patients with sustained (i.e. centralized) or ceased (i.e. peripheral) tinnitus after surgery. A volumetric analysis of cortical and subcortical gray matter (GM) anatomy was performed on preoperative high-resolution MRI and related to the presence of hearing impairment, pre- and/or postoperative tinnitus. Patients with sustained (i.e. chronic) tinnitus showed an increased GM volume of the bilateral caudate nucleus, the contralateral superior colliculus, the middle frontal and middle temporal gyrus, the fusiform gyrus as well as the ipsilateral pars orbitalis when compared to those patients in whom tinnitus ceased postoperatively. Chronic tinnitus in VS patients is associated with characteristic structural changes in frontal, temporal and subcortical areas. Notably, a significant GM change of the caudate nucleus was detected providing further support for the striatal gaiting model of tinnitus.


Results
Audiovestibular symptoms of the patient cohort. This retrospective cross section study enrolled 46 consecutive patients with unilateral sporadic VS. Basic patient characteristics are shown in Table 1. Preoperatively, 67% patients had a functional hearing (GR1/2). None of the patients suffered from deafness. 57% patients suffered from preoperative tinnitus (preopTN) which disappeared in 62% of the cases after surgical VS removal. www.nature.com/scientificreports/ 22% patients suffered from a preoperative tinnitus which then sustained for at least three months after surgery. Overall, there were 26% patients suffering from postoperative tinnitus (postopTN), indicating 2 new-onset tinnitus. 63% patients suffered from vestibulopathy (i.e. vertigo or dizziness), preoperatively. Patients with preoperative tinnitus where more likely to suffer from vestibulopathy (X 2 = 4.95, p = 0.26) and of the 10 patients with chronic tinnitus, only 2 did not suffer from a vestibulopathy.
Volumetric changes in patients with sustained VS-associated tinnitus after surgery. A total of 10/46 patients (22%) who suffered from a preoperative tinnitus reported postoperative tinnitus that sustained for at least three months. These latter patients are considered to suffer from chronic tinnitus after centralization making neuroplastic cortical changes more likely. To evaluate volumetric gray matter (GM) changes related to sustained VS-associated tinnitus (susTN), we performed a surface-based morphometry on patients' individual structural MRI. Whole-brain vertex-based statistical maps (corrected for TIV) indicated volumetric changes covering ipsilesional and contralesional frontal (e.g., opercular part of inferior frontal gyrus), temporal (e.g., medial temporal gyrus, fusiform gyrus) and mesial (e.g., cingulate gyrus and precuneus) structures ( Fig. 1). However, results did not withstand FDR-based multiple comparison correction. Subsequently, we performed a ROI-based analysis ( www.nature.com/scientificreports/

Discussion
The inconsistent and partially conflicting results of the hitherto existing morphometry studies are attributed to methodological and statistical drawbacks as well as the heterogeneity of the analyzed patients 11,12 . The present study aimed to address these problems by concentrating on a -to our opinion-more homogenous patient cohort with VS-associated tinnitus 15,19 . The unique characteristic of this patient group lies in the fact that surgery nowadays is able to remove peripheral tinnitus source (i.e. tumor), potentially without causing new harm to the cochlear nerve [16][17][18] . This technique, in turn, is ceasing tinnitus postoperatively in a large number of the patients [19][20][21] . Sustained tinnitus after VS surgery could indicate a centralization (i.e. cortical neuroplasticity) of the tinnitus 19 . In line with this hypothesis, the present study found significant volumetric changes only in patients with sustained tinnitus covering temporal, frontal and the caudate nucleus but not in patients whose tinnitus ceased after surgery.
In contrast to other studies, we found no significant changes in GM volume due to the preoperative hearing impairment 25,26 . However, there was a significant negative correlation between the preoperative AEP measurement and the GM volume of the contralateral superior temporal gyrus. In contrast, there was no relationship between the AEP and the volume of the Heschl's gyrus. These results are presented in Fig. 3.
Notably, as one of the few volumetric studies handling VS patients exclusively, Wang et al. 25 showed a positive correlation between hearing loss and the decrease in the cortical thickness of the superior temporal gyrus and Heschl's gyrus. On the other hand, Profant et al. 27 reported a rather small effect of hearing loss on the auditory cortex. The present study could partially confirm these findings, however, likely attributable to the composition of the patient cohort, in which a great number of the analysed patients had good or at least functional hearing, whereas no patient was suffering from hearing loss. Considering the limited sample size, this could have influenced the present results.
In good agreement with other studies, our analysis provide additional evidence for an increase of the medial temporal gyrus in patients with persistent tinnitus 24,28-32 which has implicated the medial temporal lobe as a final common pathway for all tinnitus patients 33 . Same results were presented from patients with chronic tinnitus and unilateral hearing loss as well. Although the fusiform gyrus is not yet associated with tinnitus, many reports relate it to speech processing and tone recognition 34,35 . We hypothesize, that VS patients might need more effort to recognise and process the sound due to the VS-associated tinnitus and hearing impairment. On the other www.nature.com/scientificreports/ side, it is well known that the fusiform gyrus is affected in patients with vestibulopathy. As 8/10 patients with sustained tinnitus have been suffering from preoperative vertigo and/or dizziness, the detected changes might be related to the preoperative vestibulopathy 32 .
Most detected volumetric changes are covering a tinnitus-related fronto-temporal network including auditory and limbic-associated areas that have been described in prior studies 36 . Surprisingly, our analysis provides evidence for a volume reduction of the paracentral lobule. Volumetric changes in this region associated to tinnitus have been described before 13 . Positron emission tomography, however, have shown an increased activity of this region in tinnitus 28,37 . Nevertheless, the involvement of the paracentral lobule in tinnitus pathophysiology is still unclear. Besides, the present study revealed a volume increase of the contralateral medial frontal gyrus, the location of the ventromedial (vmPFC) and dorsomedial prefrontal cortex (dmPFC). However, in contrast to our findings previous studies have shown a volume reduction [38][39][40][41] correlating with perceptual loudness and awareness of the tinnitus sensation 40 . One possible explanation for this discrepancy might be the duration since the onset of tinnitus. While the previous studies evaluated patients suffering from tinnitus for several years or even decades, VS patients are usually treated surgically in a rather short time. In fact, concentrating on the patients with a duration since tinnitus onset < 5 year, vmPFC surface area and dmPFC curvature show a positive correlation to the tinnitus duration 40 . Initially, tinnitus is supposed to lead to a hyperactivity of the associated regions [42][43][44] . Chronic hyperactivity could, however, generate excitotoxic levels of glutamate receptor activation 45 . We hypothesize that after an initial volume increase of the medial frontal cortex prolonged exposure results in a cell death and successive volume reduction. The same pathophysiological concept is suggested for the volume increase of the pars orbitalis of the inferior frontal gyrus which is also in conflict to the volume reduction which was found in previous studies 22 .
Interestingly, the present study depicted volumetric changes in the tectum and the basal ganglia. Previous studies might have been unable to detect these changes as VBM is not optimal for assessing subcortical structures 40,46 . Although the connection of inferior colliculus in the auditory pathway through the temporal role and its abnormal activation in tinnitus is already shown [47][48][49][50][51][52] , those of the superior colliculus remains to be elucidated. Animal and human research has shown a relation of the superior colliculus activation to sound perception and its involvement in auditory pathways 47,[53][54][55][56] . Furthermore, it is known that the deeper layers of superior colliculus receive auditory information from the inferior colliculus, in order to form audio-visual integration of speech 57 . In light of this, we observed a GM increase of the superior colliculus. Unfortunately, changes of the superior and inferior colliculus did not reach statistical significance. However, the interpretation of this result must consider the technical limitation of SBM with thin gyral stalks 58 . Finally, one of the most robust findings in our data, depicted by all three statistical evaluations, was a bilateral increase of GM volume of the caudate nucleus. The basal ganglia, in particular the caudate nucleus, is usually not in the principal focus of tinnitus research. However, there is evidence for an increased connectivity of the basal ganglia in chronic tinnitus patients 59,60 . This received further support by the recent findings that there is an increased connectivity between the caudate nucleus and the cortex in chronic tinnitus patients 61 . Similar to our results with changes in the limbic system and the bilateral peri-auditory regions in patients with chronic tinnitus were delivered from Meyer et al. 62 . Comparable to this study, volumetric changes were independent of age, hearing loss and sex 62 .
The caudate nucleus is strongly interconnected with the frontal limbic system 61,63-65 , including vmPFC which is related to the perception of tinnitus and is thus part of the gating system 36 . A recent development is the striatal gating model which hypothesizes the caudate nucleus to act as a gating mechanism for tinnitus awareness 66 . This hypothesis was initially based on the observation that direct stimulation of the dorsal striatal area LC, a locus of nucleus caudatus located at the junction of the head and body of the caudate nucleus, during deep brain stimulation (DBS) surgery in patients with movement disorders modulates auditory phantom loudness 67 and triggers auditory phantom percepts in patients with hearing loss 66 . Furthermore, vascular infarction of area LC results in enduring tinnitus loudness suppression 68 . However, the exact physiological mechanism is unclear. Due to its interconnection to both the auditory cortex and the frontal limbic system (e.g. vmPFC) 61,64,65 the caudate nucleus could act as a switch for auditory phantom representations to reach perceptual awareness and define its severity. In fact, a recent study has been able to show an increased functional connectivity between the left caudate nucleus and the left Heschl's Gyrus as well as the left and right Heschl's Gyrus and the left caudate nucleus in patients with bothersome tinnitus and single-sided deafness 65 . Notably, these connections correlated with the difficulty to relax due to the tinnitus and with the sense of reduced control over the tinnitus 65 . Furthermore tinnitus is shown to enhance the functional connectivity between the temporal cortex and the limbic system which explains the volumetric changes we found in these areas 27,69 . The present study is the first volumetric study to show GM changes in the caudate nucleus which could be a structural correlate of the augmented functional interactions of the caudate nucleus and cortical hearing and non-hearing centres in tinnitus.
This study comes however with some limitations. Although the SBM approach is supposed to be robust to different field strengths, and different scanner specifications 70 , combining scans from multiple centres introduces a bias. However, as the study design is balanced, scanner differences affect all patients and are not specific to one clinical subgroup. Although having a rather large patient cohort in comparison to related studies 12 , subgroup analysis within the present study is limiting the sample size of each group. Although sample size calculations for neuroimaging studies are challenging 12 , studies with larger patient cohorts are needed to verify structural changes in VS-associated tinnitus patients. A major limitation of the study is the dichotomization of the patients' tinnitus complaints. Although advantageous for the statistics, there was available systematic data on the tinnitus severity (e.g. Tinnitus Handicap Inventory) due to the retrospective design of the study. Finally, methodological aspects should be discussed. Several software packages offer automated voxel-/vertex-based morphometry solutions (e.g., Brainsuite, Freesurfer, SPM). While most of the available studies apply Freesurfer or SPM 12  www.nature.com/scientificreports/ data 12 . However, Brainsuite is known to provide results of similar accuracy 71,72 and might be more suitable for subcortical structures 62,[73][74][75] .
To sum up, up to date volumetric studies in VS-associated tinnitus could help trace the pathophysiology of this disease due to the unique possibility of causal therapy in these patients. Hence, the present study represents, to our knowledge, the first volumetric study showing GM changes in VS-associated tinnitus. In fact, sustained tinnitus after VS surgery was associated with structural changes in frontal and temporal regions comparably to non-VS tinnitus studies. However, although using a rather homogeneous patient cohort the present study could not resolve all discrepancies which are obviously inherent in comparable studies. Nevertheless, it is noteworthy, that for the first time a significant GM volume increase was detected for the caudate nucleus providing further support for the striatal gaiting model in tinnitus. This knowledge provides further insight in the pathophysiology of tinnitus and might help to predict tinnitus outcome after VS surgery which improve preoperative counselling and might affect surgical strategy. Further prospective studies with a larger patient cohort and standardized MRI-scanner protocols should be performed, which will enlighten us about the VS-associated tinnitus development and persistence.

Methods
Patients. This retrospective cross section study enrolled 46 consecutive patients with unilateral sporadic VS who underwent a neurosurgical VS removal in the Neurosurgical Department of the University of Tuebingen, Germany between January 2008 and January 2015. The inclusion criteria covered an age range of 18-80 years old and the availability of preoperative high-resolution magnetic resonance imaging (MRI) T1 sequences, as the later is not part of the usual neuroradiological workup of VS patients. All patients received clinical evaluations of VS-associated symptoms (i.e. tinnitus and hearing impairment) prior to and three months after surgery. The principles of the surgical procedure were unchanged throughout all patients 17 . Patient characteristics are shown in Table 1. 26/46 patients (57%) suffered from preoperative tinnitus which disappeared in 16/26 (62%) of the cases. This study was approved by the ethics committee of the Eberhard Karls University Tuebingen (registration no. 513/2017B02). All participants gave informed consent. All methods were performed in accordance with the guidelines and regulations.
Clinical evaluation. All patients underwent a thorough clinical evaluation of VS-associated symptoms (i.e. hearing impairment, tinnitus, dizziness, balance problems, gait disturbance, headache, facial dysesthesia and/or palsy, swallowing difficulties, nausea, vomiting) by a semi-structured interview by experienced neurosurgeons. Finally, the presence of ipsilateral tinnitus symptoms was dichotomized for statistical analysis (0: no tinnitus, TN-; 1: tinnitus present, TN +). Hearing impairment was classified according to the Gardner & Robertson (GR) scale 76 based on the results of the pure tone audiometry (PTA) and speech discrimination (SDS) resulting in five classes: GR 1 (good, PTA 0-30 dB and SDS 70-100%), GR 2 (serviceable, PTA 31-50 dB and SDS 50-69%), GR 3 (non-serviceable, PTA 51-90 dB and SDS 5-49%), GR 4 (poor, PTA 51-90 dB and SDS 1-4%), GR 5 (deaf, PTA 0 dB and SDS 0%). According to previous publication, the GR score was reclassified in (i) functional hearing (GR1 and GR2), (ii) non-functional hearing (GR3 and GR4) and (iii) no hearing (GR5) 19,20,77 . Additionally, hearing function was classified based on the presence or absence of waves I, III and V in the auditory brainstem response (ABR) examination performed before surgery: Group I (Waves I, III and V are present and the latency I-III is normal or slightly increased), Group II (Waves I, III and V are present and the latency I-III is pathologically increased > 2.66 ms), Group III (Wave III is lost but Waves I and V are present), Group IV (only wave I or wave 5 is present) and Group V (all waves are lost). VS tumor size was graded according to Koos classification 78 into 4 classes: T1 (purely intrameatal), T2 (intra-and extrameatal), T3 (filling the cerebellopontine cistern), T4 (compressing the brain stem).
MRI acquisition and processing. MRI data was acquired for preclinical diagnostic purposes on different MR-tomographs (43 cases with 1.5 T, 3 cases with 3 T). However, only patients with preoperative high-resolution 3D T1-weighted sequences in the sagittal plane with Gadolinium-based contrast agents were enrolled in the study (isovoxel of 0.5-1.0 mm in 36 patients; isovoxel of 1.2-2.2 mm in 10 patients). To ensure high MR quality (e.g. from movement artifacts), all images were carefully inspected. Volumetric analysis. MRI data was flipped according to the affected AN (i.e., right hemisphere represents ipsilateral to the tumor) with the MRIcron software (http:// www. nitrc. org/ proje cts/ mricr on). Subsequent analysis steps were performed by the Brainsuite software package for SBM (http:// www. brain suite. org) 73 . In brief, Brainsuite performs a sequence of image analysis steps including skull and scalp removal, nonuniformity correction, tissue classification, topology correction, and surface generation to produce triangular surface mesh models of the inner and outer boundaries of the cerebral cortex. Next, the surfaces for each subject were registered to a reference atlas surface using Brainsuite's surface/volume registration software (SVReg, http:// brain suite. org/ proce ssing/ svreg/) [79][80][81] . The SVReg-results were proofed manually to ensure proper segmentation and surface/ volume registration (see Fig. 5). This results in a spatial alignment of the white/gray matter (WM/GM) cortical surfaces across all subjects. The GM volume of each region-of-interest (ROI) was averaged and used for ROIbased analysis (Table 2). For, vertex-vice (i.e., whole-brain) analysis data was smoothed using a 2.5-mm kernel compensating for registration inaccuracies.
The GM volume of each ROI was normalized by division of the total cortical (GM + WM) volume 82,83 in order to correct for global effects such as head size, age and gender 84,85 . Most commonly, an estimate of intracranial volume is including the cerebrospinal fluid (CSF + GM + WM) [86][87][88][89] . In the present study, however, CSF volume, in contrast to GM and WM, was not normally distributed (p = 0.023, Kolgomorov-Smirnov-Test) and www.nature.com/scientificreports/ was consequently not included in the normalization process. This could be explained by the fact that vestibular schwannomas affect CSF circulation by obstruction or malreabsorption 90,91 . Finally, the presence of tinnitus did not affect CSF volume in the present study (p > 0.756, Kruskal-Wallis).
Statistics. All statistical tests were performed using R (http:// www.r-proje ct. org) and SPSS (IBM SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp.). For whole-brain analysis, GM surface was analyzed with a General Linear Model (http:// brain suite. org/ bss/) to test for the effect of susTN. TIV was included in the model as covariate. Vertex-based p values were corrected using a false discovery rate correction (FDR < 0.05) based on the Benjamini and Hochberg (BH) procedure 92 . For ROI-based analysis, separated multivariate analysis of variance (MANOVA) tests were applied to evaluate the impact of preoperative (preopTN) and sustained/ chronic tinnitus (susTN) as well as the preoperative hearing impairment (preopHI) on the normalized ROI volumes. Given the known dependency between preopTN, susTN and preopHI 19 , including these factors in a single MANOVA could bias the results due to multicollinearity. Multivariate outlier in ROI data were excluded using Mahalanobis Distance. MANOVA was follow-up by univariate ANOVAs. Resulting p-values were corrected for multiple comparisons based on FDR (< 0.05). Correlation analysis were based on Pearson's correlation. Data are shown as mean ± standard deviation (SD). Statistical significance was considered with p < 0.05 for each frequentist statistical test. This study followed the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) checklist.

Data availability
The dataset used and analyzed in this study is available from the corresponding author upon request.
Received: 17 October 2020; Accepted: 6 April 2021 Figure 5. (A) Exemplary MR image of a patient with a T3 vestibular schwannoma of the right side. Data processing included skull striping and brain extraction process followed by the automated brain atlas registration (B,C).