Quantitation of the A2A Adenosine Receptor Density in the Striatum of Mice and Pigs with [18F]FLUDA by Positron Emission Tomography

The cerebral expression of the A2A adenosine receptor (A2AAR) is altered in neurodegenerative diseases such as Parkinson’s (PD) and Huntington’s (HD) diseases, making these receptors an attractive diagnostic and therapeutic target. We aimed to further investigate the pharmacokinetic properties in the brain of our recently developed A2AAR–specific antagonist radiotracer [18F]FLUDA. For this purpose, we retrospectively analysed dynamic PET studies of healthy mice and rotenone–treated mice, and conducted dynamic PET studies with healthy pigs. We performed analysis of mouse brain time–activity curves to calculate the mean residence time (MRT) by non–compartmental analysis, and the binding potential (BPND) of [18F]FLUDA using the simplified reference tissue model (SRTM). For the pig studies, we performed a Logan graphical analysis to calculate the radiotracer distribution volume (VT) at baseline and under blocking conditions with tozadenant. The MRT of [18F]FLUDA in the striatum of mice was decreased by 30% after treatment with the A2AAR antagonist istradefylline. Mouse results showed the highest BPND (3.9 to 5.9) in the striatum. SRTM analysis showed a 20% lower A2AAR availability in the rotenone–treated mice compared to the control–aged group. Tozadenant treatment significantly decreased the VT (14.6 vs. 8.5 mL · g−1) and BPND values (1.3 vs. 0.3) in pig striatum. This study confirms the target specificity and a high BPND of [18F]FLUDA in the striatum. We conclude that [18F]FLUDA is a suitable tool for the non–invasive quantitation of altered A2AAR expression in neurodegenerative diseases such as PD and HD, by PET.


Abstract:
The cerebral expression of the A 2A adenosine receptor (A 2A AR) is altered in neurodegenerative diseases such as Parkinson's (PD) and Huntington's (HD) diseases, making these receptors an attractive diagnostic and therapeutic target. We aimed to further investigate the pharmacokinetic properties in the brain of our recently developed A 2A AR-specific antagonist radiotracer [ 18 F]FLUDA. For this purpose, we retrospectively analysed dynamic PET studies of healthy mice and rotenonetreated mice, and conducted dynamic PET studies with healthy pigs. We performed analysis of mouse brain time-activity curves to calculate the mean residence time (MRT) by non-compartmental analysis, and the binding potential (BP ND ) of [ 18 F]FLUDA using the simplified reference tissue model (SRTM). For the pig studies, we performed a Logan graphical analysis to calculate the radiotracer distribution volume (V T ) at baseline and under blocking conditions with tozadenant. The MRT of [ 18 F]FLUDA in the striatum of mice was decreased by 30% after treatment with the A 2A AR antagonist istradefylline. Mouse results showed the highest BP ND (3.9 to 5.9) in the striatum. SRTM analysis showed a 20% lower A 2A AR availability in the rotenone-treated mice compared to the control-aged group. Tozadenant treatment significantly decreased the V T (14.6 vs. 8.5 mL · g −1 ) and BP ND values (1.3 vs. 0.3) in pig striatum. This study confirms the target specificity and a high BP ND of [ 18 F]FLUDA in the striatum. We conclude that [ 18 F]FLUDA is a suitable tool for the non-invasive quantitation of altered A 2A AR expression in neurodegenerative diseases such as PD and HD, by PET.

Introduction
Besides being a constituent of nucleic acids, the nucleoside adenosine also represents an important signalling molecule, modulating neurotransmission and physiological processes by activating at least four G-protein-coupled adenosine receptor subtypes: A 1 , A 2A , A 2B , and A 3 [1][2][3]. All these adenosine receptor subtypes are present in the brain, among which the adenosine A 2A receptor (A 2A AR) has the highest expression in the striatum [1]. In that region, the A 2A AR interacts with dopamine signalling by regulating the output of the extrapyramidal motor system [4]. Striatal A 2A ARs mainly occur in the medium spiny neurons of the dopamine D2 receptor expressing indirect striatal output pathways projecting to the subthalamic nucleus [5]. A 2A ARs frequently form heterodimers in complex with other G-protein coupled receptors such as the dopamine D 2 , metabotropic glutamate mGluR 5 , cannabinoid CB 1 , and adenosine A 1 receptors [6]. Hypoxanthine caffeine is an antagonist of all four adenosine receptor subtypes, with the highest affinity towards the A 2A AR (K i(human) = 9.5-23.4 mM) [7,8], which is thought to mediate its psychostimulant and nootropic effects [9].
The A 2A ARs modulate GABAergic, glutamatergic, and cholinergic responses in the striatum [10], and an altered receptor expression is implicated in neurodegenerative disorders such as PD [11], as well as HD [12] and Alzheimer's disease [13]. New treatment strategies in PD seek to potentiate the efficacy of dopamine-replacement therapy by targeting adenosine-dopamine interactions [10], especially in the context of levodopa-induced dyskinesias [14]. Additionally, A 2A AR antagonist treatment had neuroprotective effects attributed to its anti-inflammatory actions [15][16][17], in contrast to the first-line PD treatment with levodopa [18]. The A 2A AR antagonist istradefylline (KW-6002, Nouriast TM ) has recently received FDA approval for adjunctive treatment in patients with PD [19], while a phase III trial with preladenant (SCH 420814) was terminated due to a lack of efficacy [20]. The highly selective A 2A AR antagonist tozadenant (SYN-115) [21] was well-tolerated in a phase IIb study as a levodopa adjunct in PD patients [22], but was discontinued at phase III because of hematological toxicity [20].
Non-invasive receptor occupancy studies by positron emission tomography (PET) can serve to determine dose-dependent target engagement for optimisation of new medications and to provide non-invasive biomarkers for assessing neuroreceptor changes in PD and other progressive neurodegenerative diseases [23]. A number of PET tracers are available for assessing A 2A AR availability in the living brain, e.g., [ 11 C]KF17837 [10], [ 11 C]CSC [11], [ 11 [24]. We have recently reported that deuteration of the alkyl chain in [ 18 F]FLUDA led to improved metabolic stability and negligible cerebral uptake of radiometabolites compared to the isotopologue [ 18 F]FESCH in CD-1 mice [25,26].
In the present study, we evaluate the non-displaceable binding potential (BP ND ) of [ 18 F]FLUDA in healthy CD-1 mice and investigate its suitability for detecting striatal A 2A AR changes in a rotenone-induced murine PD model. Aiming towards clinical translation, we also characterise the binding of [ 18 F]FLUDA in pigs-a species with similar brain development as humans [27], which also offers a much larger brain size than rodents, thus allowing better quantitation and minimising partial volume effects.

Non-Compartmental Analysis and Determination of [ 18 F]FLUDA BP ND in Healthy CD-1 Mice
First, we retrospectively calculated the kinetic parameters in the mouse brains by non-compartmental analysis of the [ 18 F]FLUDA PET time-activity curves (TAC) for vehicle and blocking conditions (tozadenant or istradefylline pre-treatment) from a published data set [25]. As previously described, treatment with istradefylline, but not tozadenant, significantly reduced the area under the curve (AUC) in the murine striatum (Tables 1  and 2). The pre-treatment with tozadenant was without effect on the kinetic parameters calculated for the target region striatum or for the reference region cerebellum (Table 1). In contrast, the pre-treatment with istradefylline tended to shorten the time-to-peak and to diminish the TAC peak value in the striatum to a level comparable with the values in the cerebellum. Hence, the mean residence time (MRT) was significantly reduced in the striatum (MRT veh : 20 ± 2 min vs. MRT istra : 14 ± 0 min, p < 0.001) ( Table 2). In accordance with the observed AUC 0-60 min values, pre-administration of istradefylline did not alter the MRTs in the cerebellum, validating its use as a reference region (Tables 1 and 2).  Second, we estimated the BP ND of [ 18 F]FLUDA in the mice by simplified reference tissue modelling (SRTM), which does not require arterial input function. The low affinity of tozadenant towards the murine A 2A AR (K i = 246 nM, [25]) is reflected by the kinetic parameters derived from the brain TACs (Table 1). Therefore, we used the A 2A AR antagonist istradefylline (K i = 58 nM, [25]) as a blocking agent to determine the A 2A ARspecificity of [ 18 F]FLUDA. The mean parametric BP ND maps showed high total binding of [ 18 F]FLUDA ( Figure 1A) in the mouse striatum and a complete blocking by pre-treatment with istradefylline ( Figure 1B); the striatal BP ND declined from 3.9 ± 1.2 to zero (Table 3). Additionally, the parametric maps did not suggest any displaceable binding in regions other than the striatum.
PET estimates of radiotracer uptake in structures are inherently vulnerable to underestimation due the size of mouse striatum and net spillover of signal. Indeed, the mean BP ND values derived from a mouse brain atlas volume of interest (VOI) encompassing the entire mouse striatum (3.9 ± 1.2) were significantly lower compared to findings for a 1 mm spherical VOI (5.9 ± 1.7, p < 0.0001), placed within the striatum, centered on the peak activity ( Table 3). The R 2 and Akaike Information Criterion (AIC) provided similar values for the SRTM analyses using either the atlas-based or 1 mm spherical VOIs. Additionally, the parametric maps did not suggest any displaceable binding in regions other than the striatum. PET estimates of radiotracer uptake in structures are inherently vulnerable to underestimation due the size of mouse striatum and net spillover of signal. Indeed, the mean BPND values derived from a mouse brain atlas volume of interest (VOI) encompassing the entire mouse striatum (3.9 ± 1.2) were significantly lower compared to findings for a 1 mm spherical VOI (5.9 ± 1.7, p < 0.0001), placed within the striatum, centered on the peak activity ( Table 3). The R 2 and Akaike Information Criterion (AIC) provided similar values for the SRTM analyses using either the atlas-based or 1 mm spherical VOIs. Table 3. Striatal [ 18 F]FLUDA BPND (SRTM, using the Ma-Benveniste-Mirrione-T2 Atlas whole striatum template, or 1 mm diameter spherical VOI placed in the centroid of the target and reference region of vehicle (veh, n = 8) and istradefylline pre-treated (istra, n = 4) healthy CD-1 mice.

Non-Compartmental Analysis and Determination of the BPND in a C57BL/6JRj Murine Rotenone-Induced Parkinson Disease Model
The [ 18 F]FLUDA time-activity curves of the control and rotenone-treated mice were determined retrospectively in the striatum target region and the cerebellum reference region, using the atlas templates. The rotenone-treatment had no significant impact on the radiotracer uptake to the striatum and cerebellum ( Figure 2, Table 4). The TAC peak values were observed at an earlier time point in the cerebellum compared to the striatum (0.8 vs. 2.3 min) with lower magnitude (SUV of 0.8 ± 0.1 vs. 1.1 ± 0.2, p < 0.001), and lower AUC0-60 min (7 ± 1 vs. 19 ± 2 SUV · min, p < 0.001), for both groups as expected for a reference region.  The [ 18 F]FLUDA time-activity curves of the control and rotenone-treated mice were determined retrospectively in the striatum target region and the cerebellum reference region, using the atlas templates. The rotenone-treatment had no significant impact on the radiotracer uptake to the striatum and cerebellum ( Figure 2, Table 4). The TAC peak values were observed at an earlier time point in the cerebellum compared to the striatum (0.8 vs. 2.3 min) with lower magnitude (SUV of 0.8 ± 0.1 vs. 1.1 ± 0.2, p < 0.001), and lower AUC 0-60 min (7 ± 1 vs. 19 ± 2 SUV · min, p < 0.001), for both groups as expected for a reference region.  Table 4. Non-compartmental pharmaokinetic parameters derived from the time-activity curves of [ 18 F]FLUDA in target and reference regions in healthy (ctrl, n = 7) and rotenone-treated (rot, n = 6) C57BL/6JRj mice.

Brain Region
Nonetheless, the mean parametric BP ND maps suggested a 20% lower striatal BP ND in the rotenone-treated mice compared to the control group ( Figure 3, Table 5). The R 2 and AIC were better for the SRTM analysis using the T2-Atlas VOI, although both BP ND evaluation strategies showed good agreement. Furthermore, the BP ND was lower in the rotenone-treated group regardless of the method of VOI delineation, as suggested also by the TACs (Figure 2, Table 5).  Table 4. Non-compartmental pharmaokinetic parameters derived from the time-activity curves of [ 18 F]FLUDA in target and reference regions in healthy (ctrl, n = 7) and rotenone-treated (rot, n = 6) C57BL/6JRj mice.

Brain Region
Time-to-Peak (min) Nonetheless, the mean parametric BPND maps suggested a 20% lower striatal BPND in the rotenone-treated mice compared to the control group ( Figure 3, Table 5). The R 2 and AIC were better for the SRTM analysis using the T2-Atlas VOI, although both BPND evaluation strategies showed good agreement. Furthermore, the BPND was lower in the rotenone-treated group regardless of the method of VOI delineation, as suggested also by the TACs (Figure 2, Table 5).   Table 5. Striatal BP ND (SRTM) calculated using the Ma-Benveniste-Mirrione-T2 Atlas or use of a 1 mm spherical VOI within the target or reference region and R 2 for control (n = 7) and rotenonetreated (n = 6) C57BL/6JRj mice.

Plasma Metabolism of [ 18 F]FLUDA in Pigs
We quantified the parent and radiometabolite fractions for [ 18 F]FLUDA in plasma samples of pigs by radio-HPLC. As shown in Figure 4A, we detected up to four different radiometabolites ( Figure 4A). As shown in Figure 4B, the parent fraction of [ 18 F]FLUDA had declined to 50% at 15 min post-injection in the control pigs, and to 50% at 22 min in the tozadenant group, suggesting competitive inhibition of the enzymatic degradation of the radiotracer by the high plasma concentration of tozadenant. Indeed, the plasma AUC was higher in the tozadenant group (AUC 0-90, tozadenant = 82 SUV · min) than in the control group (AUC 0-90 = 58 SUV · min, vehicle), suggesting a 40% increase in bioavailablility of radiometabolites ( Figure 4A). As shown in Figure 4B, the parent fraction of [ 18 F]FLUDA had declined to 50% at 15 min post-injection in the control pigs, and to 50% at 22 min in the tozadenant group, suggesting competitive inhibition of the enzymatic degradation of the radiotracer by the high plasma concentration of tozadenant. Indeed, the plasma AUC was higher in the tozadenant group (AUC0-90, tozadenant = 82 SUV · min) than in the control group (AUC0-90 = 58 SUV · min, vehicle), suggesting a 40% increase in bioavailablility of

Kinetic Analysis of [ 18 F]FLUDA Uptake into Different Porcine Brain Regions
We used the standard T1 CH. Malbert pig brain atlas [28] integrated into the PMOD software for the definition of the cerebral subregions. The mean [ 18 F]FLUDA TACs for striatum ( Figure 5A), cerebellum ( Figure 5B), cerebral cortex ( Figure 5C), and midbrain ( Figure 5D) indicate substantial blockade by pre-treatment with tozadenant only in the striatum. Notably, the cerebellar [ 18 F]FLUDA uptake was unaffected by blocking, and thus meets an essential criterion for to serve as a reference region. The same figure also presents the corresponding area-under-the-moment curves (AUMC) used for the calculation of the MRT, along with the other non-compartmental kinetic parameters summarized in Table 6. The TAC peak in the striatum was observed earlier after tozadenant treatment as compared to the control group (1.6 vs. 5.5 min, p = 0.05), accompanied by a significantly lower peak TAC value (SUV of 0.9 ± 0.2 vs. 1.3 ± 0.1, p = 0.03), and a significantly reduced AUC 0-90 min and AUMC 0-90 min (p = 0.01), all indicating displaceable binding of [ 18 F]FLUDA in striatum. These parameters were unaffected by blocking in the other three investigated pig brain regions. Interestingly, the MRT in all brain regions did not differ under control and blocking conditions, indicating that tozadenant treatment did not alter the washout kinetics of [ 18 F]FLUDA from the brain. compared to the control group (1.6 vs. 5.5 min, p = 0.05), accompanied by a significantly lower peak TAC value (SUV of 0.9 ± 0.2 vs. 1.3 ± 0.1, p = 0.03), and a significantly reduced AUC0-90 min and AUMC0-90 min (p = 0.01), all indicating displaceable binding of [ 18 F]FLUDA in striatum. These parameters were unaffected by blocking in the other three investigated pig brain regions. Interestingly, the MRT in all brain regions did not differ under control and blocking conditions, indicating that tozadenant treatment did not alter the washout kinetics of [ 18 F]FLUDA from the brain.

Determination of the V T and BP ND of [ 18 F]FLUDA in the Pig Brain
The mean voxelwise total distribution volume (V T ) maps of [ 18 F]FLUDA were calculated by Logan plot analysis and the averaged parametric BP ND maps by SRTM ( Figure 6). We did not attempt to evaluate the microparameters (K 1, k 2, k 3, and k 4) because of biased plasma input functions in two animals. However, the compartmental analyses clearly showed complete displacement of the striatal binding of [ 18 F]FLUDA by tozadenant treatment. The V T maps indicate a global non-specific distribution volume (V D ) of about 7 mL · g −1 throughout the blocked pig brain. Notably, the V T was two to three-fold higher in the unblocked striatum, while the mean BP ND was 1.3 ± 0.4 (1.1 ± 0.3 in nucleus accumbens, 1.4 ± 0.4 in caudate nucleus, and 1.3 ± 0.4 in putamen according to VOI analysis) ( unblocked striatum, while the mean BPND was 1.3 ± 0.4 (1.1 ± 0.3 in nucleus accumbens, 1.4 ± 0.4 in caudate nucleus, and 1.3 ± 0.4 in putamen according to VOI analysis) ( Table 7). Tozadenant pretreatment decreased the striatal BPND to 0.31 ± 0.17, corresponding to 76% displacement of [ 18 F]FLUDA. The magnitude of BPND was not significantly different from zero in the other two brain regions examined, nor was there clear evidence for displacement by tozadenant. Figure 6. Mean parametric maps of the VT (mL · g −1 ) maps (left, n = 2) and the corresponding BPND maps (right, n = 3) of control and tozadenant-treated pigs. Figure 6. Mean parametric maps of the V T (mL · g −1 ) maps (left, n = 2) and the corresponding BP ND maps (right, n = 3) of control and tozadenant-treated pigs. Table 7. Mean estimates of total distribution volume (V T ; Logan plot) and BP ND (SRTM, cerebellum reference region) in pig brain volumes of interest.

Oral Rotenone Administration
Wild-type C57BL/6JRj mice (12 months) were divided into two groups and treated five days a week for two months. A 1.2 mm x 60 mm gavage tube (Unimed, Lausanne, Switzerland) was used to administer 0.01 mL/g bodyweight of rotenone (Sigma-Aldrich, Munich, Germany) solution corresponding to a 5 mg/kg daily dose to the rotenone-treated group (n = 6). The control group (n = 7) was treated only with the vehicle solution (2% carboxymethyl cellulose (Sigma-Aldrich, Munich, Germany) and 1.25% chloroform (Carl Roth, Karlsruhe, Germany) [26].
Reconstruction of the PET scans was done using filtered back projection with a Hanning filter, along with attenuation and further corrections as mandatory (scatter, dead time, decay). A transmission scan with three rotating 68 Ge rod sources performed prior to the emission scan was used for attenuation correction. After completing the dynamic PET recording, pigs were euthanised with an IV 5 mL dose of T61 (Intervet Deutschland GmbH, Unterschleißheim, Germany).

Blood Sampling of Pigs
The hematocrit was measured in an ear vein blood sample collected just prior to imaging. Blood samples of a volume between 0.5 and 1.0 mL were collected in intervals between 15 and 60 s by a peristaltic pump (P-1, Pharmacia Biotech Inc., Uppsala, Sweden) from a catheter placed in a femoral artery using an autosampler (Fraction Collector FRAC-100, Pharmacia Biotech Inc., Uppsala, Sweden). At circulation times after 40 min, blood samples were drawn by hand every ten min. Subsequently, plasma was obtained by centrifugation (15,000 rpm), and aliquots were counted in a Cobra gamma counter (Packard Instrument Company, Meriden, CT, USA) cross-calibrated to the scanner, and decay corrected for the fluorine-18 half-life. We obtained additional plasma samples for HPLC analysis of radiometabolites at 2, 4, 6, 8, 16, 30, 60 and 90 min post injection (p.i.) of the radiotracer.

Data Analysis and Model Description
Image registration and brain volume of interest (VOI) analysis for mouse experiments were performed with PMOD software (PMOD Technologies LLC, v.4.202, Zurich, Switzerland). The time-activity data are expressed as the mean standardised uptake value (SUV) of the entire VOI. Non-compartmental analysis of achieved time activity curves (TACs) were performed with Microsoft Excel to determine the time-to-peak, the TAC peak value, the area under the curve (AUC): where c (radioactivity) is expressed as standardized uptake value normalized to the body weight in g (SUV), the area-under-the-moment curve (AUMC): and the mean residence time (MRT): Voxelwise maps of [ 18 F]FLUDA BP ND in the mouse brains were calculated in PMOD by simplified reference tissue model (SRTM) with cerebellum as reference tissue, as previously validated for the related radiotracer [ 18 F]FESCH [30] and used as preferred reference region in A 2A AR PET studies [31,32]. For the evaluations, we compared two VOI delineations of mouse striatum. First, the whole mouse striatum and whole cerebellum VOI from the Ma-Benveniste-Mirrione-T2 atlas template [33], and second, a 1 mm diameter sphere centred on the "hottest" voxels of the left and right striatum left and right and one positioned in the centre of the cerebellum to avoid potential signal spill-in from adjacent structures (Figure 7).  For pig analysis in PMOD, the summed PET brain images were co-registered to the standard T1 C.H. Malbert pig brain atlas [28] and time-activity curves were extracted from the striatum, cerebellum, midbrain and cortex VOIs. The non-compartmental analysis of the pig brain TACs was performed as stated above. Parametric maps of total distribution volumes (VT; mL · g −1 ) from two control and two tozadenant-treated pigs were calculated by Logan analysis using the metabolite-corrected arterial input function (the plasma curves from the two other scans were corrupted due to technical difficulties during the blood sampling). We calculated the BPND with an SRTM using the cerebellum as a reference region for all six pigs (Table 8). Parametric maps are presented as mean images For pig analysis in PMOD, the summed PET brain images were co-registered to the standard T1 C.H. Malbert pig brain atlas [28] and time-activity curves were extracted from the striatum, cerebellum, midbrain and cortex VOIs. The non-compartmental analysis of the pig brain TACs was performed as stated above. Parametric maps of total distribution volumes (V T ; mL · g −1 ) from two control and two tozadenant-treated pigs were calculated by Logan analysis using the metabolite-corrected arterial input function (the plasma curves from the two other scans were corrupted due to technical difficulties during the blood sampling). We calculated the BP ND with an SRTM using the cerebellum as a reference region for all six pigs (Table 8). Parametric maps are presented as mean images from two V T and three BP ND analyses for control and blocking groups.

Discussion
In the present study, we performed non-compartmental analysis, with additional compartmental analysis to determine the BP ND of [ 18 F]FLUDA (1) in healthy CD-1 mice, (2) in a rotenone mouse model of PD, and (3) in healthy pigs. We confirmed the A 2A ARspecific striatal uptake of [ 18 F]FLUDA in mice and pigs and the suitability of the cerebellum as a reliable reference region for SRTM analysis. Non-compartmental analysis in the A 2A AR antagonist-treated animals revealed no impact on the peak time, TAC peak value, MRT, and accumulated activity over time of [ 18 F]FLUDA in the reference region, or in any brain regions other than striatum in mice and pigs. In the pig studies, the time-to-peak, TAC peak value, and accumulated activity in the striatum were significantly lower in the group with tozadenant pretreatment, whereas no such effects were detectable in mice with tozadenant pretreatment. However, pre-treatment of mice with istradefylline resulted in significantly lower values for these pharmacokinetic parameters, including a significantly decreased MRT of the tracer in the striatum. The SRTM analysis demonstrated mouse strain and species (mouse vs. pig) differences in the striatal [ 18 F]FLUDA BP ND . Interestingly, we revealed a reduction in BP ND in the striatum of the rotenone-treated mice compared to control mice. In pigs, a receptor blockade with tozadenant evoked significantly decreased V T and BP ND values in the striatum relative to the baseline condition, thus validating the pharmacokinetic results from the non-compartmental analysis.   Table 8 shows BP ND values in striatum ranging from 0.74 to 9.6 for other radiotracers used for A 2A AR imaging in different species [5,21,30,49]. The K i values of those radiotracers in vitro are in the range of 0.05 nM to 12 nM for the human A 2A AR and 0.5 to 18.9 nM for the A 2A AR of rodents. Hence, FLUDA possesses a high affinity towards the human A 2A AR (K i = 0.7 nM), as shown by competition assays with [ 3 H]ZM241385 [25]. Analyses by the SRTM method have determined the cerebral cortex, midbrain, and cerebellum to serve as reference regions for the calculation of BP ND in striatum [21,30,44]. In the present study, we used the cerebellum as a reference region, with the VOI positioned and scaled to avoid significant partial volume effects, even in the small mouse brain. A 2A AR agonist treatment evoked an increase of cerebral blood flow in rats [49] and the A 2A AR antagonist tozadenant decreased regional cerebral blood flow in humans [50,51]. While treatmentevoked perfusion changes might conceivably alter [ 18 F]FLUDA uptake, we saw no effects of A 2A AR blockade in the non-compartmental analysis of reference regions in healthy CD-1 mice and pigs. The reductions in the time-to-peak, TAC peak values, and AUCs in the striatum under blocking conditions compared to baseline reflect the A 2A AR-specific binding in this brain region. The striatal BP ND values of [ 18 F]FLUDA determined in healthy CD-1 mice under baseline (3.9) and istradefylline blocking condition (0.0) indicate high specificity of the radiotracer towards the A 2A AR. The apparent magnitude of BP ND in the mouse striatum using a 1 mm spherical VOI placed near the centroid of activity (5.9) was considerably higher compared to the BP ND estimation by the atlas-based VOI for whole striatum (3.9). This is indicative of the penalty in accuracy due to spillover of signal from the mouse striatum, and may favour the use of a more stringent VOI in rodent PET studies. Similarly, the limited spatial resolution of PET led to systematic underestimation of the true BP ND of the D 2 R radiotracer [ 18 F]fallypride in the mouse striatum relative to gold standard ex vivo determination [52].
[ 18 F]FLUDA presents a more favourable BP ND (3.9-5.9 in CD-1 mice), compared to its isotopologue [ 18 F]FESCH (BP ND of 1.6-3.4 in rat striatum [30] vs. 2.7-3.8 in CD-1 mice (data not shown). This difference might reflect methodological factors, or inherent effects of deuteration on the ligand affinity. Furthermore, the enhanced stability of [ 18 F]FLUDA (parent fraction of > 99% in the mouse brain at 15 min p.i.) compared to [ 18 F]FESCH (parent fraction of 71% [26]) reduces the bias in quantitation due to brain-penetrant ra-diometabolites, which may be a factor explaining the higher BP ND of [ 18 F]FLUDA. In the present study, the differing [ 18 F]FLUDA BP ND between C57BL/6JRj mice and CD-1 mice (2.5 ± 0.4 vs. 3.9 ± 1.2 respectively, p = 0.005) suggests an important effect of strain on A 2A AR availability in vivo. While the two mouse strains also differed with respect to age, clinical PET studies in humans did not indicate important age-dependent changes in A 2A AR availability [53,54].
Rotenone treatment evokes behavioral parkinsonism and about 75% depletion of striatal dopamine content in rodent [55]. Interestingly, we found reduced striatal A 2A AR availability in the rotenone model mice as compared to the control-aged group (Table 5). Similarly, Zhou et al. showed a small decrease of striatal BP ND with [ 11 C]preladenant in 6-OHDA-induced parkinsonian rats compared to sham rats (BP ND 4.3 vs. 4.6), suggesting post-synaptic effects of dopamine depletion on A 2A AR availability [56]. Indeed, a loss of A 2A AR on striatal medium spiny neurons stands in contrast to the increased expression of dopamine D 2 Rs reported in a model where rotenone was directly administered to the substantia nigra [57], and in postmortem human brain studies [58]. However, Bhattacharjee et al. found elevated striatal uptake of [ 18 F]MRS5425 ([ 18 F]FESCH) in the 6-OHDA-induced PD model of rats [59]. The inconsistent A 2A AR PET findings in PD model animals may be due to lack of standardisation in the treatment protocols, and time dependence of the phenotypical changes [60]. Thus, the shorter rotenone treatment of two month in our present study may have induced transient receptor changes, which were not observable in the earlier study with rotenone treatment of four months [26]. Furthermore, A 2A AR expression on glial cells in the rodent brain may contribute to the PET signal [61,62]. Hence, further investigation is required to establish and explain the effects of rotenenone-induced parkinsonism on striatal A 2A AR, and the relationship with dopamine D 2 Rs coexpressed on medium spiny neurons.
A 2A ARs on medium spiny neurons are also implicated in the neurochemical pathology of Huntington's disease (HD). In autoradiographic studies with [ 3 H]CG21680, Martinez-Mir et al. detected a decrease of the A 2A AR density in the basal ganglia from patients with HD, but that finding in vitro has yet to be confirmed using A 2A AR PET in living HD patients [63]. Thus, PET imaging of A 2A AR with [ 18 F]FLUDA could prove to be a valuable tool for the staging of HD and intervention studies, as seen in pre-clinical models [64,65]. Activation of A 2A AR on striatal or extrastriatal neurons had opposite effects on psychomotor activity [66]. However, neither [ 18 F]FLUDA, nor other available radiotracers, are able to detect the low A 2A AR density in extrastriatal regions.
In terms of scale, the pig brain presents a distinct advantage over the rodent brain for molecular imaging by PET. On the other hand, the in vivo metabolite analysis of [ 18 F]FLUDA in pigs revealed faster biotransformation of the radiotracer over time (Figure 4), as compared to CD-1 mice, in which the parent fraction of [ 18 F]FLUDA in plasma collected at 15 min p.i. was still 71% [25] vs. only 50% in the pig. Additionally, we have already reported on the formation of at least two additional metabolites in pigs [25]; it remains unknown if the hydrophobic metabolites seen in Figure 4A can cross the blood-brain barrier, thus contributing to brain activity. The non-compartmental analysis did not indicate any effect of tozadenant pretreatment on the striatal [ 18 F]FLUDA uptake in CD-1 mice. However, continuous infusion of tozadenant throughout the pig recording resulted in an almost complete displacement of the striatal binding [ 18 F]FLUDA. The present estimate of striatal V T of [ 18 F]FLUDA in pigs (14.6 mL ·g −1 , Logan graphical analysis) is comparable to the [ 18 F]MNI-444 V T in monkeys (12.4-30.3 mL · g −1 , Logan graphical analysis) [21]. Human striatum shows a regionally heterogeneous distribution of A 2A ARs, with higher levels in the putamen compared to the head of the caudate nucleus [14,35]. We see some hint of gradients in [ 18 F]FLUDA uptake in pig striatum, although less than in a similar sized non-human primate brain (Table 8). Remarkably, the primate and pig results suggest lower BP ND than what we estimated in the mouse striatum, despite its small size. This is consistent with the previously determined receptor density in vitro with [ 18 F]FLUDA in murine (B max = 556 ± 143 fmol/mg wet weight) and pig striata (B max = 218 fmol/mg wet weight) [25]. In quanti-tative autoradiographic studies with the A 2A AR ligand [ 3 H]ZM241385, Villar-Menéndez et al. determined a B max of 730 fmol/mg protein in putamen of patients dying with PD vs. only 330 fmol/mg protein in controls [58]; we would expect a BP ND of [ 18 F]FLUDA comparable to our studies in mice. On the other hand, findings of increased A 2A AR binding in post-mortem brain from PD patients is at odds with our present findings in the acute rotenone model.

Conclusions
Our study supports the suitability of the SRTM using the cerebellum as a reference region for the evaluation of the BP ND of [ 18 F]FLUDA in healthy mice, a mouse PD model, and healthy pigs. [ 18 F]FLUDA kinetics in pigs differs from that in mice with respect to the greater number and formation rate of plasma radiometabolites, some of which may contribute to brain signals. The magnitude of BP ND in the striatum is higher in mice than in pigs, irrespective of the method for quantitation, and despite the greater vulnerability of quantitation in the small mouse striatum to underestimation. However, our investigation in a larger-brained species supports the translatability of [ 18 F]FLUDA for the non-invasive PET imaging of A 2A AR in the human basal ganglia.