Excitotoxic neurodegeneration is associated with a focal decrease in metabotropic glutamate receptor type 5 availability: an in vivo PET imaging study

Metabotropic glutamate receptors (mGluRs) have been proposed as promising therapeutic targets to correct the dysregulated glutamate signaling, associated with neurodegenerative pathologies. Of all mGluR subtypes, especially mGluR5 acts as a modulator of glutamate-induced excitotoxicity. To study the behavior of mGluR5 following localized excitotoxicity, we utilised a pharmacological model that portrays exacerbated neuronal glutamate release, mediated by the endogenous excitotoxin quinolinic acid (QA). Using longitudinal positron emission tomography (PET) with [18F]FPEB, we investigated cerebral changes in mGluR5 following striatal QA-lesioning. Behavioral tests were executed to monitor motor and cognitive performance. Decreased mGluR5 binding potential (BPND) was found in the affected striatum and globus pallidus of QA-lesioned rats at week 3, and further decreased at week 7, as compared to sham-injected controls. mGluR5 availability in the ipsilateral nucleus accumbens was significantly decreased at 7 weeks post-injection. QA rats performed significantly worse on motor coordination and balance compared to control rats. Correlation analysis indicated a positive correlation between striatal mGluR5 BPND and rotarod performance whereas print width of the unaffected forepaws showed a positive relation with mGluR5 BPND in the contralateral motor cortex. Together, our results suggest decreased mGluR5 availability to be related to excitotoxin-induced neurodegeneration and symptomatology although late stage effects do indicate possible cortical mGluR5-mediated effects on motor behavior.

Glutamate is the dominant excitatory neurotransmitter in the brain and plays a crucial role in the central nervous system (CNS) by acting through two types of receptors, the ionotropic (iGluR) and metabotropic (mGluR) glutamate receptors. Glutamatergic dysfunction, leading to excitotoxic neuronal death, is a common neuropathological pathway in several CNS disorders, including Parkinson's (PD) and Huntington's disease (HD) 1,2 . Chronic glutamate-mediated neurotoxicity through overstimulation of iGluRs (such as the N-methyl-D-aspartate (NMDA) receptor), but also mGluRs, has been implicated in the disease onset and progression of several neurodegenerative diseases [3][4][5] . To date, the mechanisms that underlie the disease progression of these neurodegenerative diseases remain unclear, complicating the quest for a disease-modifying drug, which would be able to cure or slow down the disorder.
Out of the 8 types of mGluRs, mGluR5 is a transmembrane G-protein-coupled receptor linked to phospholipase C and inositol 1,4,5-triphosphate and increases intracellular calcium levels following its activation 6 . mGluR5 has been targeted for potential treatment of various neurodegenerative disorders and possesses some favorable characteristics. First, mGluR5 is involved in the regulation of normal locomotor activity. Hence the receptor may be centrally involved in neurodegenerative movement disorders 7,8 . In a physiological state, glutamatergic signaling is a key factor in the normal functioning of the basal ganglia -thalamus -cortical loop, thereby mediating voluntary motor function 9 . mGluR5 modulation was shown to improve motor function in several movement disorders both clinically and in rodent models 10,11 . Second, mGluR5 may act as a co-receptor to mutant huntingtin, amyloid-β plaques and α-synuclein oligomers, providing a direct link between mGluR5 and proteinopathies [12][13][14] . mGluR5 activation may facilitate the activation of NMDA receptors through the Homer/PSD95/Shank protein complex in dendritic spines, and this way play a permissive role in neuronal excitotoxicity 15 . Accordingly, mGluR5 positive allosteric modulators (PAMs) were shown to be predominantly neurotoxic whereas negative allosteric modulators (NAMs) have shown neuroprotective properties against excitotoxic degeneration, although clinical trials have shown no major benefit so far [16][17][18][19] . Third, the topographical organization of neuronal mGluR5 expression is restricted to cerebral regions affected by neurodegenerative diseases, such as the striatum, cerebral cortex and hippocampus 20 .
The striatum is one of the primarily affected structures in several movement disorders and is known to play a cardinal role in the pathophysiology of both motor and non-motor symptoms. To date, however, the effect of excitotoxicity and subsequent post-synaptic dysfunction of striatal medium spiny neurons (MSN) on mGluR5 availability has not been researched yet. Localized dysfunction and degeneration of MSNs has been modeled by neurotoxin models such as quinolinic acid (QA), a model for excitotoxic lesioning and HD. We therefore investigated the effect of QA lesioning on regional mGluR5 levels in vivo, employing 3-[ 18 F]-fluoro-5-(2-pyridinylethynyl) benzonitrile ([ 18 F]FPEB) microPET in relation to behavioral measures and immunohistochemistry.

Materials and Methods
Animal model. All animal experiments were performed in accordance with the European Communities Council Directive 2010/63/EU and approved by the local Animal Ethics Committee of the KU Leuven (P147/2014). In vivo experiments were conducted on 15 male Sprague-Dawley rats (on average 10 weeks old; body weight range at the start of the experiment 320.6 ± 13.1 g). Animals had free access to pellet food and tap water, and were under a 12 h light/dark cycle. Stereotactic quinolinic acid (QA) injections into the left striatum were executed in accordance to a protocol described previously 21 . In short, animals were injected with 2 µl containing either QA (n = 10; 120 nmol dissolved in 0.9% saline solution) or saline solution (control group, n = 5), using following coordinates for the striatum: anteroposterior (AP) +0.2, lateral (LAT) +2.8, dorsoventral (DV) −4.5. QA-and saline-injected rats will be mentioned hereafter as QA and control groups, respectively. For histology, an additional cohort of male Sprague-Dawley rats were included of the same age with a body weight range 296.1 ± 16.5 g.
Before and during PET imaging, rodents were anesthetized using 2.5% isoflurane in 100% oxygen (1.5 l/min flow rate) and temperature was maintained at ±37 °C. Tail veins were catheterized for injection of 18.2 ± 2.2 MBq [ 18 F]FPEB. Dynamic 60-min scans were initiated simultaneously with [ 18 F]FPEB injection. Scans were conducted at three time points -in the last time point (7 weeks), there was a technical issue with the scans of three QA rats and were not included in the PET data analysis at this time point.
PET image reconstruction and data processing. List-mode data were reconstructed in 21 frames (4 × 15, 4 × 60, 5 × 180 and 8 × 300 seconds) using an iterative maximum a posterior probability (MAP) algorithm with ordered subsets (18 iterations, 9 subsets; fixed resolution: 1.5 mm) and attenuation corrected by means of a 57 Co-transmission scan, conducted prior to the dynamic scan.
PET images were normalized to an in-house rat brain template in Paxinos stereotactic space 26 . Parametric non-displaceable binding potential (BP ND ) images of [ 18 F]FPEB were generated by using Ichise's original multi-linear reference tissue model (MRTM0), with the cerebellum as reference tissue 27 .
Voxel-wise analysis was performed using Statistical Parametric Mapping 12 (SPM12, Wellcome Department of Cognitive Neurology, London, United Kingdom). Spatially normalized images were masked to exclude extra-cerebral signal and smoothed with an isotropic Gaussian kernel of 1.6 mm. SPM analysis was performed using a 0.8 relative threshold of mean image intensity. T-maps were interrogated at a p height ≤ 0.005 (uncorrected) peak level and extend threshold of k E > 200 voxels (1.6 mm 3 ). A multifactor design including 'condition' (QA vs. control) × 'time' (baseline, 3-and 7-weeks) was used for voxel-wise analysis.
Behavioral testing. All behavioral tests were executed during the light phase of the 12 h light/dark cycle at baseline, 3 and 7 weeks post-injection. Tests were performed on two consecutive days prior to PET imaging, in the same order, and on similar time points throughout the day. All rats were evaluated for motor dysfunction with the rotarod 28 and CatWalk test 29 while cognitive performance was evaluated using the Novel Object Recognition Test 30 .
Rotarod. Prior to lesioning, rats were trained for 5 consecutive days on an accelerating rotarod (4 to 40 rounds per minute (RPM) in 5 minutes). Every day, each rat was trained 3 times, with a minimum of 20 minutes between each run. Rodents were considered trained if they could remain for >90 sec on the accelerating rod, after which baseline was taken. After surgery, rats were subjected to the same accelerating protocol and latency to fall off (in seconds) was recorded.
Catwalk. The Catwalk ™ (Noldus Information Technology, Wageningen, The Netherlands) has been proven useful to detect gait disturbances in animal models of neurodegenerative diseases and allows for semi-automated quantification of a number of locomotor features, such as stride length and interlimb coordination. Animal training and performance of the catwalk test was executed as described previously 25 .
Novel object recognition Test (NORT). To investigate cognitive impairment involving memory and spatial navigation, we performed the NORT, following a previously described protocol 25 . The recognition index (RI) was calculated by (T B/C /(T A + T B/C ))x100, with T A as the time spent exploring the familiar object A and T B/C as time spent exploring novel object B (short-term memory) or C (long-term memory).
Histology and stereological quantification. An additional cohort of 8 rats (QA: n = 4; saline: n = 4) was included for immunohistochemistry and sacrificed at 3 weeks post-injection.
General statistics. Reported values are described as the mean ± standard deviation. Conventional statistics of VOI and behavioral data were executed using 2-way repeated-measures analysis of variance (ANOVA) and non-parametric Mann-Whitney U tests (PRISM 7, GraphPad Software, Inc., CA, USA). Tukey post-hoc tests were utilized to correct for multiple comparisons. A correlation analysis was performed using Spearman rank correlation on significantly altered findings from [ 18 F]FPEB PET, behavioral tests and histological analysis. P-values < 0.05 were accepted as statistically significant.

Small-animal [ 18 F]FPEB PET imaging.
Mean cross-sectional PET images of [ 18 F]FPEB binding in the rat brain over time are shown in Fig. 1. Regional [ 18 F]FPEB uptake is high in the striatum, hippocampus, and cortex, in accordance with the known brain distribution of mGluR5 20,27 .
Behavioral outcome. QA-lesioning resulted in altered motor function, as demonstrated by detailed gait analysis utilizing the CatWalk test. A general increase in print width of the ipsilateral, unaffected forepaw was noted in QA rats post-lesioning, in an attempt by the animals to stabilize gait (% increase print width: +44.8 ± 18.2% and +56.0 ± 8.9%, respectively; p < 0.01 and p < 0.001, respectively; Fig. 3a).
Lesioning of the striatum resulted in a decreased latency to fall in the QA rats, suggesting reduced endurance and/or motor-planning. Significance was reached only at the 3-week evaluation when comparing to controls (latency to fall: 148.5 ± 20.9 vs. 223.3 ± 43.6 sec; p < 0.01; Fig. 3b). Additionally, rotarod performance worsened significantly for QA rats over time (baseline: 198.1 ± 42.4 sec vs. week 7: 139.1 ± 13.6 sec; p = 0.02).
www.nature.com/scientificreports www.nature.com/scientificreports/ The novel object recognition test did not indicate any differences in short-or long-term memory between QA-lesioned and sham-lesioned animals (Fig. 3c).
[ 18 F]FPEB correlation analysis. VOI-based analysis showed a positive correlation between reduced mGluR5 BP ND values in the affected striatum and decreased rotarod performance of QA rats at 3 weeks after lesioning (VOI: r = 0.62; p = 0.03), which was confirmed by voxel-wise analysis in a cluster in the anterolateral striatum (r = 0.92; p peak = 2.02 * 10 −5 uncorrected, p cluster = 0.002; Fig. 4a,b). The compensatory increased print width of the non-affected (ipsilateral) forepaw was correlated to mGluR5 BP ND values in the contralateral motor cortical region (SPM: r = 0.89; p peak = 4.12 * 10 −4 uncorrected, p cluster < 0.001; Fig. 4c,d). No other correlations were observed within VOI-or voxel-based mGluR5 BP ND values, behavioral tests, and lesion volume.
Immunohistochemical analysis of the QA-induced striatal lesion. The degeneration of striatal neurons was visualized by the loss of NeuN immunostaining at the site of injection and a quantitative histological analysis was performed to determine the final lesion volume (Fig. 5a,b). QA-lesioned rats depicted an average striatal lesion volume of 52.5 ± 10.7% as compared to 4.3 ± 2.6% in sham-lesioned rats at week 7 (Fig. 5b). In the QA lesion, mGluR5-positive cells were seen in the region of the injection site, but only at week 3 (Fig. 5c) and co-localized predominantly with cells of astrocytic origin (Fig. 5d).

Discussion
First used by Olney et al. 32 , the term excitotoxicity refers to the ability of L-glutamate to induce neuronal dysfunction and, upon chronic exposure, neurodegeneration of post-synaptic structures including both dendrites and cell bodies. Even though this process is largely mediated by ionotropic glutamate receptors, mGluR5 has also been proposed as an important co-factor since chronic activation of the receptor can potentiate NMDA receptor activation and thereby amplify excitotoxicity. Accordingly, targeting of mGluR5 could provide neuroprotection and alleviate cognitive/motor symptoms associated with different neurodegenerative disorders 33 .
In this study, we investigated the link between mGluR5 status, neurodegeneration, and motor symptoms using the pharmacological model employing QA, which is a well-known NMDA receptor agonist with excitotoxic  www.nature.com/scientificreports www.nature.com/scientificreports/ properties. In short, we evaluated mGluR5 receptor alterations in QA-lesioned rats using [ 18 F]FPEB longitudinal small-animal PET in combination behavioral assessment and looked at possible correlations between [ 18 F]FPEB and disease-associated motor deficits. In addition, we studied QA lesion size and mGluR5 immunoreactivity within different cell types present in the QA lesion.
Compared to sham-injected rats, QA-lesioned rats depicted an in vivo decrease in mGluR5 receptor binding in the affected striatum at both time points following stereotactic surgery. This decrease extended towards the ipsilateral globus pallidum and nucleus accumbens at week 7, but not to the contralateral hemisphere or distal anatomical regions. These findings are in line with Ishiwata et al., who studied at the adenosine 2A (A2A) receptor, as well as the dopamine D1 and D2 receptors using in vivo PET imaging 34 . The A2A receptor is predominantly located at the striatopallidal GABAergic D2 projection neurons, similar to mGluR5 35 . In this study, QA-lesioning led to a reduced uptake of the A2A tracer, comparable to D1 and D2 binding, though the D1 receptor showed a more prominent decline than D2. Previous reports have shown that QA-lesioning targets D1 MSNs of the direct www.nature.com/scientificreports www.nature.com/scientificreports/ pathway more prominently than D2 MSNs 36 . This could explain why mGluR5 availability remained proportionally higher as compared to the lesion volume, since mGluR5 is predominantly located post-synaptically on the striatopallidal and striatonigral D2 neurons. Striatal QA lesioning has also been utilized as a model for HD since the toxin spares a specific population of striatal neurons that are also spared in HD patients 37 . In accordance with our findings, [ 11 C]ABP-688 BP ND was significantly reduced in the striatum of Q175 transgenic HD mice from 6-13 months old 38 .
Previous research showed a clear astroglial reaction in the affected striatum at 30 days post-injection, with limited extrastriatal astrocytic reaction [39][40][41] . In line with our histochemical findings in the area of the lesion core, we showed mGluR5-positive cell presence with immune cell-like morphology, which was limited to the 3-week  www.nature.com/scientificreports www.nature.com/scientificreports/ time point. This finding may explain the dynamics of mGluR5 PET since mGluR5 BP ND continued to decline from week 3 to week 7. It is plausible that astrogliosis, associated with (elevated) mGluR5 expression, could partially mask the decreased mGluR5 signal from neuronal origin. Since mGluR5 is expressed on both microglia and astrocytes, invasion of these cell types during excitotoxicity-induced immune reactions could dampen the  NeuN (b) and mGluR5 (C) staining patterns in coronal sections surrounding the QA injection site in the striatum. The excitotoxic lesion can be visually appreciated by the loss of NeuN immunostaining in the striatum. The striatal lesion volume was determined by delineating NeuN immunoreactivity in controls (n = 5) and QA-lesioned rats (n = 10) and is represented in box plots. ****p < 0.0001; Mann-Whitney U test. Additionally, mGluR5 staining of the lesion core showed mGluR5positive cells at week 3 (white arrows), which was absent at week 7. (d) Representative slices show that the remaining mGluR5 immunostaining largely coincided with GFAP reactivity (inset). Scale bar: 25 µm. www.nature.com/scientificreports www.nature.com/scientificreports/ reduced mGluR5 contribution from striatal neurons [42][43][44] . In line with these findings, mGluR5 activation was shown to inhibit microglial NADPH oxidase activity and reduce LPS-induced ROS production, thereby attenuating microglial activation 45,46 . Arguably, we cannot exclude a relative mGluR5 upregulation in the surviving neurons as a compensatory mechanism. In contrast to a previous study from Casteels et al., we did not observe any receptor-related alterations in the intact hemisphere, albeit their study studied different targets (using [ 18 F] FDG and [ 18 F]fallypride) and was performed at different time points and a higher toxin concentration 21 . The authors reported increased glucose metabolism and D2 receptor availability in the contralateral striatum 16-25 weeks post-lesioning. Given that we performed PET imaging at around 2 months following surgery, we might not have been able to detect these adaptive changes in the basal ganglia circuitry, though previous data indicated the degenerative process to be stable at approximately the 2-month time point 40 .
In line with previous findings, QA rats developed significant motor impairment, but no memory deficit. Compared to other studies (>200 nmol toxin), we employed a 120 nmol intrastriatal injection of QA, which leads to partial lesioning and consequently more limited locomotor abnormalities and sensorimotor neglect 37,47,48 . We did notice a worsening of rotarod performance in both treatment groups, which might be due to reduced motivation in the rodents to perform the behavior test 49 . Notwithstanding, we chose to include Sprague-Dawley rats because they are considerably more susceptible to neurological damage and memory impairment as compared to the Wistar Han strain 50 .
Our observations indicated a linear relationship between decreases in mGluR5 receptor binding and reduced rotarod performance at week 3. This relation suggests that higher mGluR5 binding is linked to preserved motor performance, in accordance with our immunohistochemistry findings, which showed that reduced mGluR5 levels coincided with absent NeuN immunostaining. Thus, mGlu5 receptor imaging may reflect the striatal degeneration. However, mGluR5 alterations may also originate from astroglial origin, as suggested by our immunohistochemistry and alternatively, could also be caused by changes in receptor affinity or conformational state. As mentioned before, microglial mGluR5 activation may attenuate activation of the microglia, and potentially limit the extent of the neuronal damage. A limitation to our histological analysis is that vibratome floating sections are not recommended to perform immunofluorescence co-localization experiments because of their slice thickness. However, mGluR5-positive cells in the QA lesion clearly do not have a microglia-like morphology.
We also showed that increased print width of the unaffected forepaw correlated with higher mGluR5 binding in the contralateral motor cortex. VOI and voxel-wise analysis however did not show significant mGluR5 BP ND differences in this region as compared to saline controls. The role of cortical mGluR5 in locomotor activity was investigated recently and indicates that mGluR5 blockage reduces locomotor activity and rotarod performance 7 . Accordingly, increased print width might be interpreted as a compensatory mechanism namely that the animal supports most of its body weight at the side of the body unaffected by the striatal lesion, possibly leading to an excessive spreading of the digits. Alternatively, the observed correlation may be an (in)direct locomotor effect of the QA-induced striatal lesion. We did not detect a reduced paw intensity or print area of the affected paws.
In conclusion, we found regional changes in mGluR5 availability in the area affected by QA lesioning and limited involvement of extrastriatal regions including the cerebral cortex. We showed a highly significant relation between mGluR5 binding potential and excitotoxicity-mediated neurodegeneration, suggesting that at least part of the mGluR5 alterations were due to the necrotic effect of QA on striatal medium spiny neurons. Altogether, we showed a limited role for mGluR5 in compensatory or plasticity mechanisms to correct post-synaptic distress in the unilateral striatum.

Data Availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.