Structural and Functional Neuroplasticity Following Bilateral Vestibular Loss: A Longitudinal [ 18 F]-UCB-H/[ 18 F]-FDG Dual Tracer Rat Study

Neuronal lesions trigger mechanisms of structural and functional neuroplasticity, which can support recovery. However, the temporal and spatial appearance of structure-function changes and their interrelation remain unclear. The current study aimed to directly compare serial whole-brain in vivo measurements of functional plasticity (by [ 18 F]-FDG-PET) and structural synaptic plasticity (by [ 18 F]-UCB-H-PET) before and after bilateral labyrinthectomy in rats and investigate the effect of locomotor training. Complex structure-function changes were found after bilateral labyrinthectomy: in brainstem-cerebellar circuits, regional cerebral glucose metabolism (rCGM) decreased early, followed by reduced synaptic density. In the thalamus, increased [ 18 F]-UCB-H binding preceded a higher rCGM uptake. In frontal-basal ganglia loops, an increase in synaptic density was paralleled by a decrease in rCGM. In the group with locomotor training, thalamic rCGM and [ 18 F]-UCB-H binding increased following bilateral labyrinthectomy compared to the no training group. Rats with training had relatively fewer body rotations. In conclusion, combined [ 18 F]-FDG/[ 18 F]-UCB-H dual tracer imaging reveals that adaptive neuroplasticity after bilateral vestibular loss is not a uniform process but is composed of complex spatial and temporal patterns of structure-function coupling in networks for vestibular, multisensory, and motor control, which can be modulated by early physical training.


Introduction
Neuroplasticity following neuronal lesions is an inherent resource of the brain to augment functional recovery 1,2 . Multiple mechanisms evolving on different temporal scales are engaged, starting with rapid onset alterations during a critical post-lesion time period and followed by slower processes of reorganization, which may sustain functional gains or losses 3,4 . At the neuronal level, changes in membrane excitability, synaptic, dendritic and axonal anatomy have been described both in animals and humans 5,6 . At the network level, recruitment of novel or alternative neural circuits and changes in the strength of connections between brain areas may foster adaptation to functional consequences of neuronal damage 7,8 . Previous studies have targeted the cerebral cortex to investigate adaptive plasticity. However, it is widely acknowledged that neural substrates of functional recovery are distributed over multiple sites at different brain levels and reorganization requires ne-tuned activity at cortical and subcortical sites 1,9 . Temporal dynamics and spatial coherence of structure-function relationships after a neuronal injury are not fully understood yet. Theoretically, the lesion-induced change in function could either be succeeded by a structural alteration in corresponding brain regions ('form follows function') 10 , or an adaptive reorganization of network structures could induce a possible gain of function ('function follows form').
The latter would resemble a developmental strategy, where structural plasticity generates new functional specialization 11 . Sensory deprivation models have been used extensively to investigate plasticity mechanisms, for example, in cortical representations [12][13][14] . In this line, animal models of inner ear damage are well suited to investigate lesion-induced cerebral plasticity in correlation to behavioural markers 15 . Bilateral vestibular deafferentation in rodents results in a typical clinical syndrome with gait ataxia and postural imbalance, which partially recovers over weeks due to adaptation and sensory substitution [16][17][18] . In contrast, spatial orientation de cits persist over time 19 . Structural changes such as altered receptor expressions, synaptic morphology and cell proliferation were reported in vitro in the brainstem, striatum, hippocampus, and frontal cortex after bilateral vestibular deafferentation in the rat 17,20−22 . In clinical practice, multimodal physical training of eye-head-coordination, postural balance, and locomotion are established and evidence-based treatment principles for patients with bilateral vestibular loss 23 . While these exercises are effective at improving mobility and reducing the risk of falls, their mode of action on cerebral structure and function is barely understood.
In the current dual tracer study, we performed serial [ 18 F]-UCB-H and [ 18 F]-uorodeoxyglucose (FDG) PET measurements from baseline to 9 weeks after bilateral labyrinthectomy (BL) in the rat, to directly compare post-lesion whole-brain changes in synaptic density and glucose metabolism and investigate the effect of immediate locomotor training on adaptive neuroplasticity. In this model, [ 18 F]-FDG uptake was interpreted as an equivalent of functional plasticity mechanisms (e.g., change in neuronal activity), while [ 18 F]-UCB-H binding was taken as a biomarker for structural remodelling (e.g., synaptic loss or synaptic sprouting) 24 . We hypothesized that 1) dynamic changes of structure and function would appear at multiple brain levels with a focus on ascending vestibular networks 25,26 , as well as cortical and subcortical sensorimotor networks involved in sensory substitution 27 and locomotor adaptation 28 ; 2) complex patterns of structure-function coupling would be found across different neural networks, which potentially represent distinct modes of lesion-induced plasticity (i.e., loss/gain of function); 3) locomotor training might augment the process of adaptive plasticity and partially improve behavioural de cits after BL. To our best knowledge, this study is the rst to apply [ 18 F]-UCB-H longitudinally and in comparison to [ 18 F]-FDG to investigate whole-brain adaptive plasticity in vivo following a neuronal lesion.

Animals and housing
All animal experiments were approved by the ethical commission of the government of Upper Bavaria and performed in accordance with the guidelines for the use of living animals in scienti c studies and the German Law for the Protection of Animals (ROB-55.2-2532.Vet_02-21-32). The authors complied with the ARRIVE guidelines.
Male Sprague-Dawley rats (mean weight 400 +/-20 g, age 10-11 weeks at the time of surgery, Charles River, Sulzfeld) were housed two animals per cage in a temperature-and humidity-controlled room with a Page 5/24 systems from IntelliBio® (IntelliBio Innovations, France) with integrated running wheels. Another group (n = 12) was placed in double decker cages (GR1800, Tecniplast, Germany).

Running wheels
After baseline measurements, an RFID Chip (IntelliBio®) was transplanted between the shoulder blades of all animals to ensure uniform conditions. The rats were then randomized into two groups (training and no training). The training group could use the integrated running wheels in their cages freely. Use of the running wheel was detected automatically by the implanted RFID microchip. In this way, subject-speci c information about the distance travelled, running speed and exercise duration of the movement phases was obtained.

Chemical bilateral labyrinthectomy
Chemical labyrinthectomy was performed as described previously 80,81 : Perioperative analgesia was ensured by pre-emptive subcutaneous administration of meloxicam (1 mg/kg) s.c. 30 min before the procedure. After initiation of the anesthesia with 2-% iso urane in O2 (1-2 l/min) via a mask, local anesthesia with 0.5% bupivacaine solution (500 µl) was applied s.c. approximately 1 cm dorsomedial of the ear. A double-sided injection of 2.5 ml saline solution into the knee fold was applied to stabilize circulation during surgery. For infection prophylaxis, marbo oxacin was administered s.c. at a dosage of 2 mg/kg. With a paramedian incision, the surgical eld was opened, exposing the lamboidal ridge and the external auditory canal. After opening the external auditory canal anterior to the exit point of the facial nerve, the tympanic membrane was perforated caudally to the hammer shaft with a 26-gauge needle.
Afterwards 20% bupivacaine solution (150 µl) was injected into the tympanic cavity. The substance was then repeatedly applied and aspirated to avoid bagging into the Eustachian tube. The same procedure was repeated with 10% arsanilic acid (150 µl), which was previously shown to induce irreversible toxic damage to the primary sensory cells of the inner ear 82 . The wound closure was followed by skin suture. This procedure was carried out on both sides, starting on the left side. The analgesic and antibiotic supply was continued postoperatively for a further 3 days by administration of meloxicam (2 mg/kg) s.c. twice daily and administration of marbo oxacin (2 mg/kg) s.c. once daily.

Criteria for exclusion
Animals were excluded from the study if the following symptoms were observed: -loss of body weight equal to or more than 20% of the value before BL -ulcer of the cornea, which could occur due to an inadvertent lesion of the facial nerve during BL -bleeding from the tympanic cavity, which could prevent the diffusion of bupivacaine or arsanilic acid into the inner ear -circulatory failure or peracute apnoea with lethal consequences.
Based on these criteria, two animals had to be excluded.

Instrumental analysis of locomotion and spatial orientation
In all rats, locomotion and spatial exploration behaviour was recorded sequentially 14-times from baseline to 9 weeks post BL in an open eld (70 cm x 70 cm x 36 cm) using an automated video tracking system detecting nose, body center and tail (EthoVision® XT 16, Noldus®, Netherlands). Rats could move freely and were recorded for 10 min per run.
The following parameters were selected and evaluated based on the ndings of a previous study 63 : mean velocity (cm/s), total number of zone transitions from "border zone" to "center zone" and total number of rotations per run.

PET-CT-imaging
The animals were anesthetized with 2% iso urane in O2 (1 -2 l/min) via a mask. For the application of the tracers, the lateral tail vein was catheterized (24-gauge) and a bolus with 40 MBq of the tracer injected (in 0.5 ml saline). The animals were positioned in the PET-CT scanner (nanoPET/CT®, Mediso Medical Imaging Systems®, Budapest, Hungary) and were kept warm with a heating pad. To avoid any passive movement of the head, its position was xed using a custom-made head-holder. For individual attenuation correction, an x-ray CT scan (9 min in duration) was performed for each measurement. While v4.004). To achieve comparability, normalization to the whole brain mean activity was performed after applying a 0.4 mm isotropic Gaussian lter, using a brain mask in atlas space in a self-written Python script. Subsequently, the images were segmented into brain regions using Px Rat (W. Schiffer) atlas and mean normalized activity values for every brain region were extracted. In addition, regions of interest (ROI) for the left and right vestibular nucleus were de ned. Mean normalized activity values were extracted for the left and right vestibular nucleus ROI and included in the further analysis.

Statistics
Statistical analysis was performed with IBM SPSS 25 software and Microsoft Excel.

Statistical analysis of mean normalized activity in segmented brain regions in all animals
Mean normalized activity values per segmented brain region at 1, 3, 5, 7, 9 weeks post BL were compared to baseline in all rats separately for [ 18 F]-UCB-H and [ 18 F]-FDG to delineate spatial and temporal patterns of changes in synaptic density and regional cerebral glucose metabolism (rCGM). In general, synaptic density decreased signi cantly in 20.3% and increased in 28.8% of all brain regions, while a signi cant rCGM decrease was found in 40.6% and an rCGM increase in 5% of all brain regions at least twice post BL. In terms of anatomical distribution, serial [ 18 F]-UCB-H measurements indicated a main cluster with an overall decrease of synaptic density in brainstem-cerebellar networks (e.g., vestibular nuclei, cerebellar white matter, colliculus inferior, midbrain) ( 3.2 Statistical voxel-wise whole-brain analysis in all animals A marked decrease of rCGM (compared to baseline) started in the vestibular nuclei and adjacent vestibular cerebellum in week 1 and persisted up to week 9 post BL. In comparison, loss of synaptic density was observed with a temporal delay, beginning at 3 weeks and progressing until 9 weeks post BL ( Fig. 2A). Similar dynamics were found in the colliculus inferior, which display a reduced [ 18 F]-FDG uptake from week 1-9 post BL, while [ 18 F]-UCB-H binding decreased progressively from week 3 post BL. Parietal multisensory cortex areas showed an early rCGM decrease (week 1) and a delayed loss in synaptic density (week 3) (Fig. 2B).
Synaptic density in the posterolateral thalamus increased progressively from week 1 to 9 post BL. [ 18 F]-FDG uptake started to increase with a temporal delay at week 5 post BL at the same anatomical location and further advanced until week 9 post BL (Fig. 3).
In the frontal cortex-basal ganglia loops, voxel-based analysis showed an inverted pattern of [ 18 F]-FDG uptake and [ 18 F]-UCB-H binding. While rCGM in the frontal association cortex was reduced from week 1 post BL, synaptic density increased after week 3 post BL. In parallel, striatal rCGM decreased after week 3 post BL, while synaptic density bilaterally increased in this brain region (Fig. 4).

Statistical voxel-wise analysis of training and no training group
In the thalamus, voxel-based analysis showed a signi cantly higher uptake of

Behavioural testing
All rats showed characteristic signs of bilateral vestibular deafferentiation, such as postural imbalance, gait ataxia, head instability ('bobbing') and opisthotonus. On clinical assessment, postural imbalance and gait ataxia became less apparent after day 5 post BL and head bobbing stopped completely until day 5 post BL. Phases with opisthotonus persisted over the course of the experiment but were observed less frequently with time. Analysis of locomotion in the open eld indicated that velocity increased signi cantly and progressively after day 2 post BL compared to baseline (see Supplementary Fig. 1).
Analysis of subgroups (training and no training) showed a signi cant increase in locomotor velocity after day 5 post BL compared to baseline in both subgroups, but no relevant differences relative to each other at any time point post BL (Fig.6A). The number of zone transitions between the border and center zone of the open eld increased signi cantly after week 1 post BL compared to baseline, but to an equal extent in both subgroups (Fig.6B). The number of rotations around the body axis per run increased in the no training group until week 3 post BL and decreased slowly thereafter. Rotations in the training group also increased up to week 3 post BL. However, in weeks 1 and 3 post BL rats in the no training subgroup displayed signi cantly more rotations than in the training subgroup (week 1: no training subgroup: 27.3 ± 14, training subgroup: 17.8 ± 12.7; week 3: no training subgroup: 27.7 ± 16.3, training subgroup: 18.4 ± 9.3; p < 0.05 respectively) (Fig. 6C).

Discussion
The major ndings of this study were as follows: 1) the combination of [ 18 F]-UCB-H and [ 18 F]-FDG measurements was a feasible approach to depict dynamic changes of structure and function following bilateral vestibular loss in the rat; 2) lesion-induced plasticity was not a uniform process, but involved complex structure-function relationships at various brain levels with different temporal scales; 3) a 'form follows function' type of plasticity was found in brainstem and cerebellar networks, which were most closely exposed to loss of vestibular input, while a 'function follows form' type of plasticity appeared in thalamic nuclei dedicated to multisensory processing and motor adaptation; 4) locomotor training had a mild effect on dynamic balance after BL, which was paralleled by changes in synaptic density and glucose metabolism in the thalamus.

[ 18 F]-UCB-H/ [ 18 F]-FDG dual tracer PET imaging -a novel method to study neuroplasticity
Advances in non-invasive measurements (such as MRI, PET, NIRS, EEG, MEG) have massively contributed to our understanding of adaptive brain plasticity in vivo following neuronal lesions. While MRI has an excellent spatial resolution, EEG and MEG have a better temporal resolution. MR techniques may provide information on the relation of structural changes in grey and white matter (VBM, DTI) [29][30][31][32][33] , and functional activity of brain networks (e.g., rs-fMRI) during behavioural recovery processes after a neuronal lesion 34 . However, the cellular mechanisms underpinning adaptive structural changes in grey and white matter are not completely understood yet 1 . applied to study time dynamics and spatial distribution of neuroplasticity induced by acute lesions. In this study, we strived to depict changes in synaptic density and regional glucose consumption longitudinally in an established sensory deprivation model in the rat. For this methodological purpose, subgroups with and without training were pooled to increase statistical robustness. We think that this approach is reasonable, because subgroups were comparable for most behavioural parameters (see below) and underwent an identical longitudinal imaging protocol. We were able to show dynamic changes of synaptic density across six consecutive imaging time points within 10 weeks, which appeared in biologically plausible networks in the brainstem, cerebellum, thalamus, multisensory cortex and motorbasal ganglia circuits. Most importantly, rCGM changes were found in similar regions, but with different temporal scales. The direct comparison of time courses and spatial distribution of [ 18 F]-UCB-H and [ 18 F]-FDG signals may potentially allow conclusions to be drawn about the mode of plasticity ('form follows function' or 'function follows form').
Differential modes of structural and functional plasticity following complete vestibular loss The current study shows three major modes of structure-function coupling following complete vestibular loss, which appeared in different neural networks: i) In brainstem-cerebellar networks, early-onset decrease in rCGM after BL was followed by reduced synaptic density 3 weeks later in the same brain regions (Fig. 3). A rapid drop in neuronal activity in the vestibular and cochlear nucleus is expected after BL, because these nuclei receive primary afferents from the inner ear and are consequently affected most by loss of signal input 43 . Of note, rCGM also decreased in secondary hubs of vestibular and auditory processing (e.g., vestibular cerebellum, colliculus inferior) instantaneously after BL. Damage to inner ear inputs resulted in a delayed loss of synaptic density in the same regions (vestibular nuclei, vestibular cerebellum, colliculus inferior), which slowly increased until 9 weeks post BL. This observation can be interpreted as a persistent structural degeneration in brainstemcerebellar networks following irreversible loss of bilateral vestibular function. Our data seem to be in partial contrast to a previous in vitro study that showed a loss of synaptic density of about 35% in the medial vestibular nucleus 1 week after unilateral vestibular neurectomy in the cat, which was followed by a synaptic reoccupation from 3 weeks to 5 months after the lesion 44 . However, it has to be considered that static behavioural de cits after a unilateral vestibular lesion recover rapidly and completely due to central vestibular compensation, while a bilateral vestibular lesion induces symptoms that are more persistent and can in part not be fully compensated. It seems likely that the intact contralateral vestibular inputs contribute to synaptic reoccupation in unilateral vestibular lesions.
ii) In the thalamus, synaptic density started to increase bilaterally from 1 week after BL, which was succeeded by an rCGM increase 2 weeks later (Fig. 4). The thalamus is a well-known hub for integration of multisensory inputs from the vestibular, the visual, and the somatosensory system, as well as vestibular-motor interaction 45 . Various thalamic subnuclei with a functional role in multisensory processing (ventroposterior nuclei) 46 , motor control (ventrolateral nuclei) 47 , heading direction and spatial orientation (anterior nuclei) 48 receive ascending vestibular information directly from the vestibular nuclei or from vestibular cerebellar nuclei 25,26,49,50 . Changes in synaptic density and glucose metabolism in the posterior and lateral parts of the thalamus found in the current study likely represent a process of adaptive plasticity, which contributes to multisensory substitution and recalibration.
Accordingly, human neuroimaging studies have documented changes in cortical and subcortical multisensory networks in patients with chronic bilateral vestibular loss, which suggest an increased visual substitution 51,52 . From a neuroscienti c perspective, it is interesting that synaptic density in the thalamus increases before glucose consumption. It seems that during adaptive neuroplasticity novel synaptic connections are established rst, which in turn result in a gain of function and increase of neuronal activity. This example shows that function can follow form, similar to the processes in the developing brain during childhood 1 . This poses the question, whether the brain reactivates these developmental strategies in the case of a neuronal lesion. Interestingly, dynamic balance, which can be estimated by the number of body rotations, deteriorated until week 3 post BL and stabilized thereafter. This may be interpreted as a behavioural correlate of adaptive plasticity driven by the thalamus.
iii) In frontal-basal ganglia loops, we found an increase of synaptic density, which was paralleled by a relative rCGM decrease especially after 3 weeks post BL (Fig. 5). as an adaptive response to this functional imbalance of direct-indirect pathways, or as a secondary maladaptive effect due to the overactive direct pathway. As the hyperlocomotion does persist after irreversible vestibular loss, it seems that the synaptic plasticity in the basal ganglia is not su cient to compensate for this behavioural dysregulation.
Impact of locomotor training on adaptive brain plasticity following bilateral vestibular loss Multimodal physical training is the mainstay of therapy in patients with bilateral vestibular loss 23,66 . The applied principles are based on neurophysiological considerations and tailored to training of the vestibular-ocular re ex (habituation training), the interplay of multisensory sources to control posture (sensory pertubation training), and dynamic balance (locomotor training) 67,68 . While the effectiveness of these exercises is highly evidence based 69 , their mode of action on cerebral function and structure has been completely neglected so far. Cortical reorganization by a motor training is well documented for patients with stroke 70,71 . From this literature, we have learned that new training methods based on basic neuroscienti c principles such as motor imagery 72,73 or action activation 74 may add to the therapeutic success. In the current study, we aimed to start with a comparably simple training program based on the voluntary use of running wheels installed in the cages, to understand how this intervention may in uence behaviour, brain structure and function after bilateral vestibular loss in the rat. This selection was motivated by the fact that active locomotion stimulates multisensory feedback and trains motor control.
Furthermore, a bene cial effect of voluntary running wheel training has been shown in various neurological disease models 75,76 . In the current study, we found an effect of running wheel training on dynamic balance, as measured by the number of body rotations, with a peak 1-3 weeks post BL, while other markers of locomotor behaviour such as velocity or movement duration remained unchanged (Fig.   6). Accordingly, transient and subtle changes of synaptic density and regional glucose metabolism were found in the thalamus (Fig. 5). Given the role of the thalamus for multisensory adaptation and postural control, it is reasonable that the therapeutic effect of running wheel training was mediated by this brain region. Alterations of [ 18 F]-UCB-H/ [ 18 F]-FDG binding in brainstem-cerebellar circuits and frontal-basal ganglia circuits were not affected by training after BL. The nding of a training effect on adaptive plasticity and balance control in BL is promising, because it gives rst evidence for a potential neurobiological correlate of vestibular rehabilitation. However, the rather mild and transient effects documented in this study show that a stereotyped training such as running wheel locomotion likely is not su cient to induce persistent and functionally relevant effects on balance and locomotor control in a three-dimensional space. Future studies might test the effect of more elaborated training protocols (e.g., visual cueing, vibration feedback, more complex and varying movements) on both adaptive brain plasticity and behavioural recovery in vestibular animal models 77,78 .
In conclusion, the current study shows the potential of a [ 18 F]-UCB-H/[ 18 F]-FDG dual tracer application to investigate structural and functional plasticity mechanisms after a neuronal injury and the effect of treatment on this process. Lesion-induced neuroplasticity is a complex process with different modes of structure-function coupling, which arise in various neural networks and at different temporal courses. An improved understanding of these processes could help to tailor interventions to the optimal time window and to target speci c networks by therapeutic strategies at different times.

Declarations ACKNOWLEDGEMENT
We thank Prof. Dr. Ammer and Dr. Ungern-Sternberg for advice and discussion of study results. We thank Dr. Simon Lindner for synthesis of radiotracers. Furthermore, we thank Katie Göttlinger for copyediting the manuscript.

COMPETING INTERESTS
The authors declare no competing interests.

ANIMAL ETHICS STATEMENT
In this study, we observed humane principles, respected the welfare of animals and excluded situations when animals were in pain.

FUNDING
The study was performed as a project of the German Center for Vertigo and Balance Disorders (DSGZ) with the support of the German Federal Ministry of Education and Research (BMBF) (grant number 01 EO 1401).

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.      Values are depicted as mean + standard deviation; *signi cant difference (p<0.05). d: day, w: week.