In vivo imaging of synaptic density with [11C]UCB-J PET in two mouse models of neurodegenerative disease

The positron emission tomography (PET) radioligand [11C]UCB-J binds to synaptic vesicle protein 2A (SV2A) and is used to investigate synaptic density in the living brain. Clinical studies have indicated reduced [11C]UCB-J binding in Alzheimer's disease (AD) and Parkinson's disease (PD) brains compared to healthy controls. Still, it is unknown whether [11C]UCB-J PET can visualise synaptic loss in mouse models of these disorders. Such models are essential for understanding disease pathology and for evaluating the effects of novel disease-modifying drug candidates. In the present study, synaptic density in transgenic models of AD (ArcSwe) and PD (L61) was studied using [11C]UCB-J PET. Data were acquired during 60 min after injection, and time-activity curves (TACs) in different brain regions and the left ventricle of the heart were generated based on the dynamic PET images. The [11C]UCB-J brain concentrations were expressed as standardised uptake value (SUV) over time. The area under the SUV curve (AUC), the ratio of AUC in the brain to that in the heart (AUCbrain/blood), and the volume of distribution (VT) obtained by kinetic modelling using the heart TAC as input were compared between transgenic and age-matched wild type (WT) mice. The L61 mice displayed 11-13 % lower AUCbrain/blood ratio and brain VT generated by kinetic modeling compared to the control WT mice. In general, also transgenic ArcSwe mice tended to show lower [11C]UCB-J brain exposure than age-matched WT controls, but variation within the different animal groups was high. Older WT mice (18-20 months) showed lower [11C]UCB-J brain exposure than younger WT mice (8-9 months). Together, these data imply that [11C]UCB-J PET reflects synaptic density in mouse models of neurodegeneration and that inter-subject variation is large. In addition, the study suggested that model-independent AUCbrain/blood ratio can be used to evaluate [11C]UCB-J binding as an alternative to full pharmacokinetic modelling.


Introduction
Alzheimer's disease (AD) and Parkinson's disease (PD) are the most common neurodegenerative disorders, and the number of patients is steadily increasing worldwide. Like most neurodegenerative disorders, the gradual loss of functional neurons and synapses is a key characteristic in the AD and PD brain ( Coleman and Yao, 2003 ;Esposito et al., 2012 ;Matuskey et al., 2020 ). In both disorders, neurotoxic protein aggregates are central to the pathogenesis. In the AD brain, amyloidbeta (A ) deposits as extracellular plaques ( Hardy and Higgins, 1992 ), whereas alpha-synuclein ( -syn) forms intracellular Lewy bodies, espe-Abbreviations: -syn, Alpha-synuclein; 1TCM, One-tissue compartment model; 2TCM, Two-tissue compartment model; AD, Alzheimer's disease; AIC, Akaike information criterion; APP or A PP, Amyloid precursor protein; AUC, Area under the curve; A , Amyloid-beta; A PP, Amyloid-beta protein precursor; BPND, Non-displaceable binding potential; BSA, Bovine serum albumin; CBS, Corticobasal syndrome; CER, Cerebellum; COV%, Coefficient of variation; CT, Computed tomography; CX, Cortex; ELISA, Enzyme-linked immunosorbent assay; EOS, End of synthesis; FA, Formic acid; FOV, Field of view. disease-related brain regions. [ 11 C]UCB-J binding, quantified as either the volume of distribution (V T , brain-to-blood concentration ratio at equilibrium) or as the binding potential (BP ND ), in the hippocampus was lower in AD patients than in healthy subjects ( Chen et al., 2018 ;Tuncel et al., 2020 ). [ 11 C]UCB-J binding was also reduced in substantia nigra in early or mild-moderate PD cases ( Delva et al., 2020 ;Matuskey et al., 2020 ). Another study found unilaterally decreased [ 11 C]UCB-J uptake in the hippocampus ipsilateral to the side with MRI confirmed sclerosis in patients with temporal lobe epilepsy ( Finnema et al., 2016 ). Moreover, patients suffering from two different tauopathies, progressive supranuclear palsy (PSP) and amyloid-negative corticobasal syndrome (CBS), also presented reductions of up to 50% in regions such as the hippocampus and amygdala ( Holland et al., 2020 ).
Several new drug candidates for AD, PD or related neurodegenerative disorders have entered into clinical trials. Still, most of them have failed to meet their clinical endpoints in late phase studies . These endpoints are usually related to cognitive function. Although amyloid-PET can verify that immunotherapy with antibodies against A can remove insoluble aggregates in AD patients, its effect on the cognitive decline has been modest ( Sevigny et al., 2016 ). Hence, quantification of synaptic density could serve as a better outcome measure for cognitive status than the A plaque load.
Genetically modified mouse models are used extensively in AD and PD research to investigate novel drug candidates and their ability to diminish intrabrain levels of A and -syn, respectively. Bertoglio and colleagues have demonstrated that [ 11 C]UCB-J binds specifically to SV2A in the mouse brain, and the radioligand binding can be quantified by kinetic modeling using an image-derived input function ( Bertoglio et al., 2019 ). Another study showed that [ 11 C]UCB-J binding in transgenic AD mouse model (APPswe/PS1DE9 [APP/PS1]) was reduced in the hippocampus, and that the reduction to some extent could be restored by saracatinib. This drug inhibits tyrosine kinase Fyn and interferes with the A signaling pathway ( Toyonaga et al., 2019 ).
The present study aimed to investigate [ 11 C]UCB-J binding in one AD (APP-ArcSwe) and one PD (Thy1-Syn "Line 61 ″ ) mouse model in comparison to healthy control mice.

Animals
Animals had free access to food and water and were kept on a 12/12 h light/dark circle. All animal procedures were approved by the Uppsala County Animal Ethics Board (ethical approval 5.8.18-13,350/2017), carried out in accordance with the European Communities Council Directive of 22 September 2010 (2010/63/EU) and reported according to the ARRIVE guidelines. The number of animals included in the study was based on the previous report of an approximately 20% difference between WT and transgenic mice ( Toyonaga et al., 2019 ). Detection of such difference therefore required 6-10 animals per group (power = 0.90 and = 0.1) assuming a 10-15% standard deviation within the groups. The transgenic mouse model ArcSwe, maintained on a C57BL/6 J background, expressing the human amyloid-protein precursor ( A PP ) with the Arctic ( A PP E693G ) and Swedish ( A PP KM670/671NL ) mutations under the murine Thy1 promoter, was used as a model for ADlike A pathology. In this model, soluble A protofibrils are elevated in the brain from one month of age, while amyloid plaques are visible at around six months of age ( Codita et al., 2010 ;Lord et al., 2006 ). Plaques, with a high resemblance to human AD pathology, are especially abundant in cortical regions, thalamus and hippocampus, and increase in number and size until the age of 17-18 months ( Lillehaug et al., 2014 ;Lord et al., 2006 ). In the present study, we included transgenic mice aged 18-20 months to resemble brain pathology at an advanced disease stage. As the disease progression and pathology do not differ between genders ( Lillehaug et al., 2014 ), both male ( n = 7) and female ( n = 4) transgenic ArcSwe mice (Tg-ArcSwe) were used. As the model is heterozygous and therefore also produces 50% non-transgenic mice, age-matched C57BL/6 J wild type littermates (WT-ArcSwe) were used as controls (male: n = 1; female: n = 4).
As a model for PD, heterozygous Thy1-Syn mouse model Line 61 (L61) overexpressing human -syn maintained on a B6D2F1 background was used. The transgene expression starts at postnatal day 10, and transgenic mice display robust levels of -syn already at 2,3 months of age ( Chesselet et al., 2012 ;Roshanbin et al., 2021 ). Transgenic L61 mice (Tg-L61) also display early signs of behavioural changes and motor impairments ( Chesselet et al., 2012 ;Rabl et al., 2017 ). Pathology progression differs between males and females, with males showing a faster deterioration ( Chesselet et al., 2012 ;Roshanbin et al., 2021 ). Transgenic L61 mice (male: n = 10), aged 8,9 months with visually confirmed motor symptoms, were used in the PET study, while wild type age-matched B6D2F1 (WT-L61) littermates served as controls (male: n = 6; female: n = 2). In addition, a separate cohort of one male Tg-L61 and one male WT-L61 mouse, both 9 months, were used for the measurement of synaptic density with transmission electron microscopy.
All animals included in the study were randomly selected among the in-house bred mice of the correct age. There were no exclusion criteria, except that Tg-L61 mice that did not show motor symptoms were not included in the study.
All subsequent studies and analyses were conducted unblinded.

[ 11 C]UCB-J synthesis
[ 11 C]UCB-J was produced in a one-step synthesis method, as previously described ( Rokka et al., 2019 ). The final product was reformulated using solid phase extraction in approximately 10% ethanol in phosphate-buffered saline (PBS). The radiochemical yield was 19 ± 5% based on [ 11 C]methyl iodide, the molar activity was 314 ± 186 GBq/ mol ( n = 10) at the end of synthesis (EOS) and the average radiochemical purity was > 99%.

PET acquisition and reconstruction
Mixed paired animals (any combination of the four studied groups) were induced and maintained under anaesthesia (1.5-3.0% isoflurane mixed with 0.2 L/min O 2 and air) during the experiment and placed side by side in the scanner. Except for pre-defined animal pairs, other confounders, e.g. sex, litter, weight etc., were not controlled. Scans were acquired between 9 am and 3 pm. Mice recived during 3-5 s an intravenous (i.v) bolus injection of 6.3 ± 3.2 MBq (0.05 ± 0.03 g) of [ 11 C]UCB-J and underwent a PET scan for 60 min followed by a 3 min CT examination in a prone position in a Triumph® II PET/SPECT/CT scanner (TriFoil Imaging, Inc., Northridge, USA, Field of View (FOV) = 8.0 cm). Directly after scanning, the mice were perfused with 40 ml 0.9% NaCl for 3 min.
The PET data was reconstructed on a 160 × 160 × 128 grid with 0.5 × 0.5 × 0.6 mm 3 voxels using 3-dimensional ordered-subsets expectation maximisation (20 iterations) into the following frames: 10 × 2 s, 2 × 5 s, 3 × 10 s, 2 × 30 s, 3 × 60 s, 5 × 300 s and 3 × 600 s. The CT raw files were reconstructed using filtered back-projection. Subsequent processing of the PET and CT images was performed in Amide version 1.0.4 ( Loening and Gambhir, 2003 ). The CT scans were manually aligned with a T2-weighted, MRI-based mouse brain atlas ( Ma et al., 2005 ) containing outlined regions of interests (ROIs). The PET images were then aligned with the CT image containing the brain atlas. Time-activity curves (TACs) based on concentrations normalised to injected dose per body weight (standardised uptake value, SUV) were extracted for hippocampus (Hipp), thalamus (Tha), striatum (Str), cortex (Cx), cerebellum (Cer), and whole brain (Wb). An image-derived input function (IDIF) was extracted from a manually drawn spherical ROI (1.3 × 1.3 × 1.3 mm 3 ) to estimate total radioactivity in the blood. The ROI was placed in the heart's left ventricle in early frames (frame 2-4) with a threshold of around 50% of the maximum intensity ( Verhaeghe et al., 2018 ).

Data analysis
Areas under the SUV curve (AUC) for each studied brain region and heart left ventricle were calculated based on the brain TACs and the TAC derived from the heart IDIF. Then, regional brain-to-blood ratios (AUC brain/blood ) based on the selected brain regions and blood were generated. This AUC-based ratio was further compared to the related modelderived estimates of the volume of distribution (V T ), which also describes the relation between brain and blood concentrations ( Innis et al., 2007 ). V T was obtained by kinetic modelling in PMOD 3.7 software (PMOD Technologies, Switzerland), fitting the regional TACs to either a one-tissue compartment model (1TCM), a two-tissue compartment model (2TCM), or the Logan plot ( Logan et al., 1990 ). The blood volume fraction (V B ) was set to 3.6% ( Julien-Dolbec et al., 2002 ) for the 1TCM and 2TCM. For the Logan plot, the linear phase (t * ) was set to 10 min based on visual inspection and previous studies ( Bertoglio et al., 2019 ). The IDIF was used without any further correction for metabolites or plasma protein binding, as described earlier ( Bertoglio et al., 2019 ;Finnema et al., 2017 ). Chi-square test ( 2 ) and Akaike Information Criterion (AIC) were generated to determine the goodness-to-fit for each model ( Akaike, 1974 ;Cunningham, 1985 ).

Brain tissue homogenisation
Brains were extracted after perfusion and divided into left and right hemispheres. Cortex (Cx), hippocampus (Hipp), and cerebellum (Cer) were further dissected out from the left hemisphere and immediately frozen at − 70 °C until analysis of A or -syn levels. Brain tissue was processed according to previously published protocols ( Gustafsson et al., 2018 ;Roshanbin et al., 2021 ). In short, brain tissue from the selected brain regions was homogenised at a 1:10 wt: volume ratio using Pre-cellys® Evolution (Bertin Technologies, France) through serial extractions in Tris-buffered saline (TBS), TBS-triton, and formic acid (FA) except pellet wash between the extractions. The FA fraction, representing insoluble protein deposits, was analyzed with sandwich enzyme-linked immunosorbent assays (ELISA).

A 1-40 and A 1-42 sandwich elisa
High-binding half area plates (Corning Inc. Corning, NY, USA) were coated with 50 ng/well polyclonal rabbit anti-A 40 (Agrisera, Umeå, Sweden) or anti-A 42 (Invitrogen) antibodies overnight followed by blocking with 1% bovine serum albumin (BSA) at room temperature (RT) with shaking for 2 h. The FA extracts were neutralised with a premix of 1 M Trizma base and 0.5 M Na 2 HPO 4 at 1:15 and then diluted in ELISA incubation buffer to a final dilution of 1:12,600 for A 40 and 1:150,000 for A 42. Diluted samples were added to the plate as duplicates and incubated at 4 °C overnight, followed by detection with biotinylated 3D6 (in-house production ( Gustafsson et al., 2018 )) and streptavidin-horseradish peroxidase conjugate (SA-HRP, Mabtech AB, Nacka Strand, Sweden). The plate was developed using K -blue aqueous substrate (TMB, 308,177, Neogen Life Sciences) and absorbance was read at 450 nm (Infinite M1000, Tecan, Männedorf, Switzerland). All dilutions were made in ELISA incubation buffer (PBS with 0.1% BSA, 0.05% Tween, and 0.15% proclin).

-syn ELISA
The -syn ELISA was carried out according to the same procedure as described for the A ELISAs. First, plates were coated with 12.5 ng/well of rabbit monoclonal anti--syn antibody MJFR1 (ab138501, Abcam, Netherlands) diluted in PBS overnight. The FA extracts were neutralised as above and diluted to a final ratio of 1:250. Mouse monoclonal antibody against -syn (Clone 42, 610,787, BD Biosciences, San Jose, CA) and HRP-conjugated goat anti-mouse IgG F(ab) 2 antibody (115-036-006, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) were used for detection. Plates were developed and read as above.

Transmission electron microscopy (TEM)
Mice were anaesthetised with 3.5% isoflurane and perfused with 0.1 M phosphate buffer (5 ml) followed by 20 ml fixative containing 2.5% glutaraldehyde (18,427, Ted Pella, Redding, CA) and 1% paraformaldehyde (Merck, Darmstadt, Germany) in 0.1 M phosphate buffer (pH 7.4). Brains were isolated at 2 h post-perfusion and postfixed in fresh fixative overnight at 4 °C. 500 m thick coronal sections were cut on a vibratome (Leica, Germany). Subregions (about 2 mm 2 ) of hippocampus, cortex, and cerebellum were identified and dissected from Tg-L61 and WT-L61 under a light microscope (Leica, Germany) for comparison. Samples were rinsed with 0.1 M phosphate buffer for 10 min following incubation in 1% osmium tetroxide (O013, TAAB laboratories, England) in 0.1 M phosphate buffer for 1 h. After rinsing in 0.1 M phosphate buffer, samples were dehydrated with acceding gradient of ethanol (70% to 100%) for a total of 45 min, followed by 5 min incubation in propylene oxide (P021, TAAB laboratories, England) and then placed in a mixture of Epon Resin (18,012, Ted Pella, Redding, CA) and propylene oxide (1:1) for 1 h, followed with 100% resin and left overnight. Samples were then embedded in capsules in newly prepared Epon resin and left for 1,2 h and polymerised at 60 °C for 48 h. The specimens were cut into semi-thin sections (1,2 m) and stained in Toluidine Blue (Merck, Darmstadt, Germany) and examined in a light microscope (Leica, Germany) to search for the matching area of interests. After trimming, ultra-thin sections (60-70 nm) were cut in an EM UC7 Ultramicrotome (Leica, Germany) and placed on a grid. The grids were contrasted in 5% uranyl acetate (Merck, Darmstadt, Germany) for 10 min and 3% Reynolds lead citrate (Merck, Darmstadt, Germany) for 2 min and were examined by TEM (FEI Tecnai G2) operated at 80 kV.

Measurement of synaptic density
At least 9 images from each brain region were obtained per mouse. A synapse structure was identified by the presence of a pre-and postsynapse with a synaptic cleft. Images were photographed at 9900 × magnification in random grids ( Crowder et al., 1999 ). ImageJ (Fiji 1.53c, USA) was used for manually counting the number of the identified structures in each photograph (5.8 × 5.8 m 2 ). The cluster of vesicles was photographed at 26,500 × magnification. The synapse density was calculated as the number of synapses divided by the area and expressed as the number of synapses per m 2 .

Statistical analyses
Data were analysed in GraphPad Prism version 9.0.0 for Windows (GraphPad Software, San Diego, USA). Results are reported as mean ± SD. The Bland-Altman plot with 95% limits was used to assess the three kinetic models. The coefficient of variation (COV%) for the kinetic models and selected brain regions was calculated as: COV% = SD ∕ mean × 100% Differences in [ 11 C]UCB-J uptake between groups were analysed using a two-way ANOVA with post hoc Bonferroni correction for multiple comparisons, thereby comparing the genotypes for selected brain regions with unequal sample size. The model was fitted with genotype × brain regions as fixed effects. V T generated from 1TCM and 2TCM were compared with paired t-tests for each animal model. Pearson's correlation (r) and simple linear regression (R 2 ) were used to investigate the correlation between V T and model-independent AUC brain/blood , A , or -syn concentration. Significance was set to 95% and indicated as * : p ⟨ 0.05, * * : p < 0.01, * * * : p < 0.001, * * * * : p < 0.0001, ns: p ⟩ 0.05. Fig. 1. Representative time-activity curves (TACs) in five brain regions and heart (A-F) displayed as standardised uptake value (SUV, g/ml). The area under the curves (AUC, g/ml × min) were extracted from the TACs in the hippocampus (Hipp), thalamus (Tha), striatum (Str), cortex (Cx), cerebellum (Cer), whole barin (Wb) and blood (heart). Regions of interest (ROI) of blood were acquired using the image-derived input function (G). There were no significant differences between transgenes (Tg) and their respective wild type (WT) littermates. However, 8, 9 months WT-L61 displayed higher AUC in all brain regions than 18-20 months WT-ArcSwe, and transgenic L61 mice displayed higher AUC than Tg-ArcSwe. The closed blue symbols indicate Tg-ArcSwe, while the open blue symbols indicate WT-ArcSwe. In parallel, the closed red symbols represent Tg-L61, and open red symbols represent WT-L61. Female animals are indicated as triangles and male animals are indicated as circles. Group differences were tested in two-way ANOVA with Bonferroni post hoc correction and genotype was considered as main variance. The significanct effect of genotype was indicated as * : p ⟨ 0.05, * * : p < 0.01, * * * : p < 0.001, * * * * : p < 0.0001, ns: p ⟩ 0.05. Values in figures are shown as mean with SD. Sagittal views of averaged PET images (5-60 min) in transgenic ArcSwe and L61 mice and their respective non-transgenic WT controls, WT-ArcSwe and WT-L61 (H) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Data availability statement
Imaging data are available in dicom or text format and can be transferred per request by the corresponding author. Processed data, i.e. SUV, AUC, V T , A and -syn concentrations, are available in table format in GraphPad Prism files.

Results
The [ 11 C]UCB-J radioligand was rapidly and universally distributed into the brain, and concentrations peaked (C max ) at around 10 min after administration ( Fig. 1 A-F). At 60 min, brain concentrations had dropped to about half of the C max . Transgenic animals appeared to display lower SUVs than their WT littermates, although differences were not significant. Furthermore, transgenic L61 and their non-transgenic littermates WT-L61 displayed higher SUVs than transgenic ArcSwe and their non-transgenic littermates WT-ArcSwe. AUC, calculated over the 60 min scanning time, was highest in central brain regions such as hippocampus, thalamus, and striatum. The results showed that both genotype and region had an impact on [ 11 C]UCB-J brain uptake (genotype: p < 0.0001, brain region: p < 0.0001, interaction: p < 0.0001, Supplementary Table 1). However, the post hoc test show no significant differences in AUC in any of the studied brain regions between transgenic and WT animals, although the WT animals tended to display somewhat Table 1 AUC brain/blood ratio and V T (ml/cm 3 , mean ± SD) determined by 1TCM, 2TCM, and Logan plot based on 60 min acquisition in Tg-ArcSwe ( n = 11) and WT-ArcSwe ( n = 5). Note: 2TCM did not converge for one animal in each group. Coefficient of variation (COV%) within the kinetic models and brain regions are shown in Supplementary Table 7.
Tg-L61 WT-L61 Region AUC brain/blood 1TCM 2TCM Logan AUC brain/blood 1TCM 2TCM Logan higher AUCs than their respective transgenic littermates, in line with the SUVs ( Fig. 1 G, Supplementary Table 2). In addition, WT-L61 displayed higher AUCs than WT-ArcSwe in all brain regions but not in blood ( Fig. 1 G, Supplementary Table 1). Representative PET images are presented in Fig. 1 H and Supplementary Fig. 1. Example of brain and heart ROI placement are given in Supplementary Fig. 2. The brain-to-blood partition coefficient was calculated based on AUC in the brain and AUC in the blood (AUC brain/blood ). No significant difference in AUC brain/blood was observed between the ArcSwe and their age-matched controls, WT-ArcSwe ( Table 1 , Supplementary Table 3). However, Tg-L61 mice showed a 11,12% lower AUC brain/blood than WT-L61 ( Table 2 , Supplementary Table 3). Post hoc analyses of genotype differences in the regions showed that this difference was significant in the hippocampus ( p = 0.04), while almost reaching significance also in thalamus ( p = 0.05) and striatum ( p = 0.07) ( Table 2 , Supplementary  Table 4). The older WT-ArcSwe mice demonstrated lower AUC brain/blood than the younger WT-L61 mice in hippocampus, thalamus and striatum ( Fig. 2 A, Supplementary Table 4).
[ 11 C]UCB-J retention in the brain was also explored by model-based estimation of the volume of distribution (V T ). V T was estimated using 1TCM, 2TCM, and Logan plot. In general, V T estimated with 1TCM was higher than the V T estimated using 2TCM in the Tg-ArcSwe and WT-ArcSwe mice ( Table 1 ). The opposite was found for Tg-L61 and their corresponding controls, WT-L61. Compared to 1TCM estimates, V T generated by 2TCM estimates were 1.7% higher in transgenic L61 mice ( p < 0.0001) and 4.2% higher in WT-L61 mice ( p < 0.0001). Still, Bland-Altman analysis indicated that the V T estimated from 1TCM and 2TCM were in good agreement ( Supplementary Fig. 3). Logan plot derived V T was generally lower than those derived by the 1TCM and 2TCM. To check goodness-to-fit, 2 and AIC indicated that 2TCM was more accurate for the L61 model while 1TCM fitted the data obtained in the ArcSwe model slightly better. However, the better goodness-to-fit value in the ArcSwe model may have been caused by 2TCM failing to converge for a few animals. Comparison of models and model diagnostics and representative kinetic fits are shown in the supplementary (Supplementary Figs. 4 and 5, Supplementary Tables 5-8).
Moreover, V T calculated by kinetic modelling using the three different models were all well-correlated with the model-independent AUC brain/blood directly derived from TACs ( Fig. 3 ). Although 2TCM fitted the data in the L61 model best, V T obtained by 1TCM has, in line with our data in the ArcSwe model, proved to be more robust in previous reports ( Bertoglio et al., 2019 ;Finnema et al., 2017 ;Mansur et al., 2020 ). Thus, V T generated from 1TCM was chosen for further evaluation and association with disease pathology. However, the use of 2TCM or Logan plot derived V T would not have changed conclusions and general interpretation of the results.
In all studied brain regions, Tg-ArcSwe tended to display lower V T than the WT-ArcSwe, but the difference was not significant ( Table 1 ). For example, in the cortex, a region with abundant A pathology, V T in Tg-ArcSwe mice was 94% of that in WT-ArcSwe mice. The same trend was also seen in the L61 model ( Table 2 ). V T (1TCM) of Tg-L61 and WT-L61 were significantly different (genotype: p < 0.001, interaction: p > 0.05), but this difference did not reach significance in any of the studied regions in the post hoc test ( Supplementary Tables 9 and 10). Moreover, the younger group of WT mice, WT-L61 , displayed higher V T values in almost all brain regions compared with the older WT-ArcSwe ( Fig. 2 B and Supplementary Table 10).
To assess A and -syn levels in the respective models, we performed A and -syn sandwich ELISA using the FA extracts of homogenates from the different brain regions. In ArcSwe animals, both A 1-40 and A 1-42 were assessed. As expected, measurement of A and -syn in brain homogenates verified that transgenic ArcSwe and L61 animals expressed high levels of A and -syn, respectively, while their WT littermates had no measurable levels ( Fig. 4 . A-C). However, there was no correlation between individual protein levels and V T in the transgenic mice ( Fig. 4 . D-F).
Next, we assessed synaptic density in selected brain regions in Tg-L61 and WT-L61 through TEM. An example of a synapse structure is presented in Fig. 5 . Synapse density was highest in the hippocampus, but differences between Tg-L61 and WT-L61 were not detected in any of the three analysed regions ( Table 3 ).

Fig. 2.
Comparison of AUC brain/blood ratio (A) and 1TCM V T (B) between 18 and 20 months old WT-ArcSwe (blue) and 8-9 month old WT-L61 (red) suggested that [ 11 C]UCB-J uptake could be influenced by age, as age (genotype) is strongly associated with the AUC brain/blood ratio and V T (p ⟨ 0.0001, interaction: p ⟩ 0.05 for both readouts). Female animals are indicated as triangles, and male animals are indicated as circles. Significance level of age (genotype) is shown as * : p ⟨ 0.05, * * : p < 0.01, ns: p ⟩ 0.05, two-way ANOVA with post hoc Bonferroni correction for multiple tests. Hipp: hippocampus. Tha: thalamus. Str: striatum. Cx: cortex. Cer: cerebellum. Wb: whole brain (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.). Fig. 3. V T generated with 1TCM, 2TCM, and Logan plot correlated well with matching AUC brain/blood ratio in ArcSwe (A-C) and L61 mice (D-F). Correlations were first tested (p) and presented as Pearson's coefficient (r) and later fitted into a simple linear regression model shown as the black straight lines and the formulas at the bottom. Closed symbols represent transgenic mice, and open symbols represent wild type mice. Hippocampus (Hipp), thalamus (Tha), striatum (Str), cortex (Cx), cerebellum (Cer), whole brain (Wb).

Discussion
In the present study, we investigated synaptic density assessed by [ 11 C]UCB-J PET in two animal models of neurodegeneration -one AD and one PD model. The L61 mice showed a lower signal compared to non-transgenic WT-L61 using the model-independent AUC brain/blood read-out, 1TCM and Logan analysis, but the post hoc test showed significance only for the AUC brain/blood read-out in the hippocampus although a similar trend was seen for all brain regions using all studied outcome measures. Pathology in transgenic animals displays a rather large variation, and therefore small differences are difficult to detect Fig. 4. A 1-40 (A) and A 1-42 (B) levels in formic acid (FA) extractions from Tg-ArcSwe ( n = 6) and wild-type control WT-ArcSwe ( n = 3) brain homogenates, as measured with ELISA. FA-soluble -syn (C) in Tg-L61 ( n = 6) and WT-L61 ( n = 4) mice. Protein levels were expressed as ng per mg brain tissue (mean with SD). V T did not correlate with total brain A 1-40 (D), A 1-42 (E), or -syn (F) concentrations in the different brain regions of transgenic animals. Hipp: hippocampus. Cx: cortex. Cer: cerebellum. and therefore require rather large groups of animals. Data obtained in this study, which included 34 PET-scanned animals divided into four groups, should therefore be interpreted with caution, but several points are still interesting. First, the most considerable difference between Tg-ArcSwe and WT-ArcSwe mice in all of the model derived V T estimates (1TCM, 2TCM, and Logan) was related to the cortex, a region with abundant pathology, whereas the most minor difference was found for the cerebellum, a region with a later start of A accumulation and a lower pathology level. The same relationship between [ 11 C]UCB-J retention and pathology rich brain regions was also observed in the L61 model. This indicated that the decrease in [ 11 C]UCB-J retention is likely related to synaptic degeneration in these models. Second, the study indicated that synaptic density decreased with age regardless of pathology progression. The older WT mice displayed lower brain uptake of [ 11 C]UCB-J than the younger WT mice. This finding was not initially expected but is consistent with recently reported results using [ 11 C]UCB-J in a mouse model of obsessive-compulsive disorder ( Glorie et al., 2020 ).
It should be noted that the mouse genetic background was partly different for the two age groups of WT mice used in the present study. This was a consequence of the animal models used, as the PD model is maintained on a B6D2F1 background, which is a 50%:50% cross between C57BL/6 J and DBA/2 J, while the AD model is maintained on a pure C57BL/6 J background. Still, both mouse lines were kept in the same animal facility under the same housing conditions. Two previously conducted studies in 8-month respective 14-month old C57BL/6 J mice support the finding that [ 11 C]UCB-J brain uptake decreases with age ( Bertoglio et al., 2019 ;Toyonaga et al., 2019 ). In the 18-month old C57BL/6 J mice used in the present study, [ 11 C]UCB-J brain was even further decreased compared to the 8-month and 14month old mice. Opposite to this, a recently published study reported no decline in [ 11 C]UCB-J retention with age in healthy human subjects ( Michiels et al., 2021 ). However, it should be pointed out that in this study, although it included individuals from a very large age range (18-85 years), only few subjects (~10) above the age of 70 were investigated, and in this old cohort it appeared as [ 11 C]UCB-J retention was somewhat declined between ages 70 and 85.
A previous study in another AD mouse model, APP/PS1, showed that binding potential (BP ND ) obtained by the simplified reference tissue model (SRTM) ( Lammertsma and Hume, 1996 ) using the brain stem as the reference region was decreased about 26% compared to WT mice ( Toyonaga et al., 2019 ). We did not use the SRTM or any other referencebased read-out because the PD model used in our study does not have a suitable reference region. Especially the brain stem and cerebellum are regions affected by -syn pathology ( Chesselet et al., 2012 ). It should be noted that APP/PS1 mice show abundant amyloid deposits at 8 months of age in the neocortex, hippocampus and brain stem ( Radde et al., 2006 ), but the potential impact of this was not discussed in the paper by Toynaga et al. Also the Tg-ArcSwe model, in line with the ma-jority of all AD models, shows brain-wide pathology distribution without a pathology free reference region at old age ( Meier et al., 2018 ;Sehlin et al., 2017 ). Still, the 26% difference is greater than the approximately 5% Tg-ArcSwe versus WT-ArcSwe and 11, 12% Tg-L61 versus WT-L61 differences seen in the present study. The obvious explanation for the different degree of [ 11 C]UCB-J decrease observed in transgenic animals compared to WT is the different models used. The above mentioned previous study was performed in an animal model with more aggressive A pathology mainly caused by the inclusion of a presenilin mutation in addition to mutations in the amyloid-beta precursor protein ( A PP or APP ). The transgene expressed in the ArcSwe model does not harbour any mutation of presenilin (PSEN) , but only mutations in A PP . Previous studies have indicated modest cognitive defects in the ArcSwe model ( Lord et al., 2009 ). A recent study using another SV2A radioligand, [ 18 F]SynVesT-1, that did not explicitly report APP/PSEN1 versus WT differences due to a low number of animals ( n = 6), presented SUV curves that indicated a 5-10% difference between the transgenic and WT mice ( Sadasivam et al., 2020 ). Furthermore, the earlier mentioned [ 11 C]UCB-J study in a model of obsessive compulsory disorder also reported a difference of about 10-15% between Sapap3 knock-out mice and WT ( Glorie et al., 2020 ). To our knowledge, this is the first [ 11 C]UCB-J study performed in a mouse model of PD ( Thomsen et al., 2020 ), and hence, no comparisons to previous studies in PD models can be made.
The primary modelling based read-out was V T derived from the 1TCM and an IDIF based on total radioactivity. This set-up was recommended by previous publications that methodically and thoroughly investigated different modelling paradigms for [ 11 C]UCB-J PET in mice ( Bertoglio et al., 2019 ;Verhaeghe et al., 2018 ). The rationale for the use of a non-metabolite corrected IDIF was based on the observation that metabolite profiles did not differ between disease models and WT in the above-mentioned studies. However, this may not per se be true for all animal models, and therefore one potential limitation of the present study is the lack of metabolite data. The IDIF was based on radioactivity measured in the left ventricle of the heart as recommend by previous studies. An IDIF based on activity measured in vena cava was initially also investigated for a sub-set of the animals. However, for some animals the model fits were much worse compared to the IDIF based on the radioactivity in the left ventricle (Supplementary Figs. 2 and 6, Supplementary Table 11). One explanation could be that due to animal breathing, the vena cava ROI also included radioactivity from surrounding organs at the later parts of the scan. The scan time in the present study was 60 min. The above-mentioned studies acquired PET data for 90 min, but the authors concluded that 60 min scanning was sufficient to estimate V T using 1TCM if a systematic 10% over-estimation was acceptable. It should be noted that regional time-activity curves were generated using a non-normalised atlas, and thus inter-animal variation in brain shape and size, could influence the V T values especially in small regions such as hippocampus. Still, the V T values reported in the present study in central regions (5.0 -7.0 ml/cm 3 ) align with the V T estimates reported in the previous study. Also, a second study showed that BP ND estimates based on 60 min and 90 min data were in excellent agreement with each other ( Toyonaga et al., 2019 ).
We found that the AUC brain/blood ratio correlated strongly with V T derived from 1TCM. This indicated that the ratio, AUC brain/blood , may be used as a model-independent outcome measure and as an alternative to full pharmacokinetic modelling. We used an early frame, usually number 2, 3, or 4 (2-8 s after injection), to define the heart ROI used to extract the IDIF. This non-invasive method is similar to a previously described process ( Verhaeghe et al., 2018 ). It should be noted that when the injection solution contained less activity ( < 3 MBq) or when the activity concentration was lower, it was more challenging to place the IDIF ROI. Also, as sometimes induced by isoflurane anaesthesia, heavy breathing may hamper this procedure in small animals. Still, using a non-invasive method to estimate the blood input is essential when scan-ning preclinical models lacking an adequate pathology-free reference region.
Even though [ 11 C]UCB-J PET is thought to reflect synaptic density, it should also be pointed out that it is actually a measure of synaptic vesicle protein SV2A. Synaptic density and SV2A have been shown to correlate with each other ( Crowder et al., 1999 ), but studies suggest that SV2 proteins mainly regulate cytoplasmic Ca 2 + dependent neurotransmitter release but do not directly change brain structure and synapse composition ( Crowder et al., 1999 ;Janz et al., 1999 ;Wan et al., 2010 ). Because SV2A may function before the final fusion of vesicles, the changes in synaptic strength and altered presynaptic Ca 2+ attributed to SV2 loss can occur at different time points during disease progression ( Wan et al., 2010 ;Xu and Bajjalieh, 2001 ). Although the L61 mice used in our study showed obvious motor impairment, we did not observe a decreased number of synapses with TEM. It has been reported that despite reduced synaptic transmission, the number of dopaminergic neurons in substantia nigra is the same in Tg-L61 and WT-L61 mice ( Chesselet et al., 2012 ). Moreover, other synaptic proteins, such as SV2C, another member of the SV2 family, could also disrupt dopamine homeostasis and lead to motor disturbance in PD ( Dunn et al., 2017 ).

Conclusion
In general, [ 11 C]UCB-J binding appeared to be lower in transgenic mice than in WT, especially in regions with abundant pathology, but inter-animal variation was large. Thus, preclinical [ 11 C]UCB-J studies designed to investigate the effects of novel drug candidates, should ideally include large groups of same-age animals as the study also indicated that age might have a more important effect than the presence of A and -syn pathology on the synapse density measured with [ 11 C]UCB-J.

Data statement
Imaging data are available in Dicom or text format and can be transferred per request by the corresponding author. Processed data, i.e. SUV, AUC, V T A levels and -syn levels, are available in table format in GraphPad Prism files.

Data statement
Imaging data are available in dicom or text format and can be transferred per request by the corresponding author. Processed data, i.e. SUV, AUC, V T A levels and -syn levels, are available in table format in GraphPad Prism files.