A metabolically stable PET tracer for imaging synaptic vesicle protein 2A: synthesis and preclinical characterization of [18F]SDM-16

Purpose To quantify the synaptic vesicle glycoprotein 2A (SV2A) changes in the whole central nervous system (CNS) under pathophysiological conditions, a high affinity SV2A PET radiotracer with improved in vivo stability is desirable to minimize the potential confounding effect of radiometabolites. The aim of this study was to develop such a PET tracer based on the molecular scaffold of UCB-A, and evaluate its pharmacokinetics, in vivo stability, specific binding, and nonspecific binding signals in nonhuman primate brains, in comparison with [11C]UCB-A, [11C]UCB-J, and [18F]SynVesT-1. Methods The racemic SDM-16 (4-(3,5-difluorophenyl)-1-((2-methyl-1H-imidazol-1-yl)methyl)pyrrolidin-2-one) and its two enantiomers were synthesized and assayed for in vitro binding affinities to human SV2A. We synthesized the enantiopure [18F]SDM-16 using the corresponding enantiopure arylstannane precursor. Nonhuman primate brain PET scans were performed on FOCUS 220 scanners. Arterial blood was drawn for the measurement of plasma free fraction (fP), radiometabolite analysis, and construction of the plasma input function. Regional time-activity curves (TACs) were fitted with the one-tissue compartment (1TC) model to obtain the volume of distribution (VT). Nondisplaceable binding potential (BPND) was calculated using either the nondisplaceable volume of distribution (VND) or the centrum semiovale (CS) as the reference region. Results SDM-16 was synthesized in 3 steps with 44% overall yield and has the highest affinity (Ki = 0.9 nM) to human SV2A among all reported SV2A ligands. [18F]SDM-16 was prepared in about 20% decay-corrected radiochemical yield within 90 min, with greater than 99% radiochemical and enantiomeric purity. This radiotracer displayed high specific binding in monkey brains and was metabolically more stable than the other SV2A PET tracers. The fP of [18F]SDM-16 was 69%, which was higher than those of [11C]UCB-J (46%), [18F]SynVesT-1 (43%), [18F]SynVesT-2 (41%), and [18F]UCB-H (43%). The TACs were well described with the 1TC. The averaged test–retest variability (TRV) was 7 ± 3%, and averaged absolute TRV (aTRV) was 14 ± 7% for the analyzed brain regions. Conclusion We have successfully synthesized a novel SV2A PET tracer [18F]SDM-16, which has the highest SV2A binding affinity and metabolical stability among published SV2A PET tracers. The [18F]SDM-16 brain PET images showed superb contrast between gray matter and white matter. Moreover, [18F]SDM-16 showed high specific and reversible binding in the NHP brains, allowing for the reliable and sensitive quantification of SV2A, and has potential applications in the visualization and quantification of SV2A beyond the brain. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-021-05597-5.


Introduction
Proteins in the synaptic vesicle glycoprotein 2 (SV2) family located in presynaptic terminals are essential components of synaptic vesicles [1]. As one of the isoforms, SV2A is ubiquitously expressed in virtually all synapse terminals, and involved in the regulation of synaptic exocytosis and endocytosis [2,3]. SV2A is a known target for anti-epilepsy This article is part of the Topical Collection on Neurology drugs, such as levetiracetam (Keppra®, LEV) [4]. Positron emission tomography (PET) is a non-invasive quantitative imaging modality that provides functional and physiological information in living systems. SV2A PET tracers can be used to study target occupancy in the clinical development of new drug candidates targeting SV2A, and to measure changes of SV2A in neuropsychiatric diseases [5][6][7][8][9][10]. SV2A PET has potential applications beyond the brain, as SV2A is expressed in the all central nervous system (CNS) [11]. While the current metabolically labile SV2A PET tracers are suitable for brain PET imaging due to the BBB preventing their radiometabolites from entering the brain, a more metabolically stable and higher binding affinity radiotracer is desirable for the investigation of SV2A expression in the whole CNS, to minimize the potential confounding effect of radiometabolites. For example, the spinal cord expresses SV2A with a much lower B max than the brain dose [12], and is protected by blood-spinal cord barrier (BSCB) rather than BBB. The difference between the permeability of BBB and BSCB may render spinal cord potentially more susceptible to the confounding effect of radiometabolites of PET tracers [13,14].
Several SV2A PET tracers have been synthesized and evaluated in animals and human during the past few years by our group and others ( Fig. 1) [7,15]. [ 18 F]UCB-H (2) [16][17][18] was the first SV2A PET tracer tested in human [17], followed by [ 11 C]UCB-J (3) [5], [ (5) [19,20], and [ 18 F]SynVesT-2 (a.k.a. [ 18 F]SDM-2) (6) [21]. The isotopologue of 3, [ 18 F]UCB-J was evaluated in rhesus monkeys, but not pursued for clinical evaluation since the harsh labeling conditions and the production process were considered unsuitable for routine production supporting clinical studies [22]. [ 11 C]UCB-J is currently the SV2A PET tracer most widely used in PET imaging investigations of neuropsychiatric disorders, i.e., epilepsy, Alzheimer's disease, Parkinson's disease, schizophrenia, major depressive disorder, and posttraumatic stress disorder [5,[23][24][25]. Among the available SV2A PET tracers, [ 11 C]UCB-A was arguably the most metabolically stable, even though its prevalent radiometabolite species in the plasma were not identified yet [26,27]. However, the relatively short radioactive half-life (~ 20 min) of carbon-11 together with the slow kinetics in the brain limited the potential clinical application of [ 11 C]UCB-A [28,29]. We hypothesized that the slow kinetics of [ 11 C] UCB-A (as reflected in its low K 1 and k 2 values) was due to its relatively low membrane permeability, which is associated with its low hydrophobicity (LogD: 1.1). To develop a metabolically more stable analog of [ 11 C]UCB-A, with improved pharmacokinetics (PK), we synthesized and evaluated a novel 18

Chemistry
All compounds were prepared from commercially available starting materials. Details are described in the supplemental materials.

Competition radioligand binding assay
Competition binding assays were performed twice independently using separate assay materials, with 4 technical replicates from each independent assay. Compounds were dissolved in DMSO (10 mM), which was diluted in PBS pH 7.4 (Gibco) with 0.1% BSA assay buffer to give 12 halflog dilutions from 10 µM to 32 pM. Duplicate samples of human frontal cortex gray matter were homogenized in PBS buffer (10 mg/mL) for storage at − 80 °C and were diluted to a stock concentration of 4 mg/mL in PBS on the day of the assays. [ 3 H]UCB-J was obtained with a molar activity of 1.29 TBq/mmol (34.9 Ci/mmol) and radiochemical purity of 98.9%, diluted in duplicate to a stock concentration of 6.25 nM. Working stocks of brain homogenate (100 µL; final concentration of 2 mg/mL), blocking ligands (20 µL), and radioligand (80 µL; final concentration 2.5 nM) were combined in quadruplicate wells of 96-well plates, sealed, and incubated at room temperature for 90 min on an orbital shaker set to 250 RPM. Reaction plates were filtered, rapidly washed with cold PBS, and dried. Forty microliters of Microscint-20 scintillation cocktail (Perkin-Elmer) was added to each well, and the plate was counted using a Microbeta2 plate reader (Perkin-Elmer). GraphPad Prism was used for curve fitting using the one-site K i model.

Measurement of lipophilicity
The logP of [ 18 F]SDM-16 was determined by a method modified from previously published procedures [30]. Briefly, an aliquot of 70 kBq (10 µCi) of the radioligand was added to a 2-mL microtube containing 0.8 mL of octanol and 0.8 mL of 1 × phosphate buffered saline (1 × PBS, pH 7.4). The mixture was vortexed for 30 s and then centrifuged at 2000 g for 2 min. A subsample of the octanol (0.1 mL) and 1 × PBS (0.5 mL) layers was evaluated with a gamma counter. The major portion of the octanol layer (0.5 mL) was diluted with another 0.3 mL of octanol, mixed with a fresh portion of 0.8 mL of PBS, vortexed, centrifuged, and analyzed as described above. This process was repeated until consistent log P values were obtained, with five consecutive equilibration procedures being performed for each logP measurement. Four separate measurements were performed for [ 18 F] SDM-16 on different days.

Measurement of plasma free fraction (f p )
The unbound fraction of [ 18 F]SDM-16 in plasma (f p ) of rhesus monkey was measured in triplicate using the ultrafiltration method [19,21]. Briefly, [ 18 F]SDM-16 solution was added to 3 mL of whole blood. After incubation at ambient temperature for 5 min, the blood sample was centrifuged at 3900 rpm for 5 min. A sample of the supernatant plasma (0.3 mL) was loaded onto the reservoir of a Centrifree® Ultrafiltration device (Merck Millipore Ltd. Tullagreen, Carrigtwohill, Co. Cork, IRELAND) in triplicate and centrifuged at 1228 g for 20 min. The f p value was calculated as the ratio of radioactivity in the filtrate to that in the plasma.

PET imaging experiments in rhesus monkeys
A total of 10 PET imaging experiments with [ 18 F]SDM-16 were performed in rhesus monkeys (Macaca mulatta) according to a protocol approved by the Yale University Institutional Animal Care and Use Committee (IACUC). Four monkeys were studied. One monkey (8 years old, male, 9.5 kg) underwent two baseline scans and one blocking scan; one monkey (15 years old, female, 11.3 kg) underwent two baseline scans; one monkey (12 years old, male, 17.0 kg) underwent one whole-body scan; and the other monkey (13 years old, female, 9.5 kg) underwent two baseline scans, one displacement scan and one whole-body scan. Rhesus monkeys were fasted overnight and sedated using intramuscular injection of alfaxalone (2 mg/kg), midazolam (0.3 mg/ kg), dexmedetomidine (0.01 mg/kg), and anaesthetized with 0.75-2.5% isoflurane approximately 2 h before the PET scan. Anesthesia was subsequently maintained with isoflurane (1.5-2.5%) for the duration of the imaging experiments. Body temperature was maintained by a water-jacket heating pad. The animal was attached to a physiological monitor, and vital signs (heart rate, blood pressure, respirations, SPO 2 , EKG, ETCO 2 , and body temperature) were continuously monitored. A venous line was inserted in one limb for administration of radiotracer, displacement, and blocking drugs. A catheter was placed in the femoral artery in the other limb for blood sampling. Dynamic PET brain scans were performed on a Focus 220 system (Siemens Medical Solutions, Knoxville, TN, USA) with a reconstructed image resolution of approximately 1.5 mm. After a 9-min transmission scan, the radioligand was injected i.v. over 3 min. by an infusion pump. Dynamic PET scans were performed for 3 h (baseline and blocking scans) or 4 h (displacement scan). For the blocking scan LEV (30 mg/kg) was administered intravenously at 10 min before tracer injection, while in the displacement scan, the same dose of LEV was infused at 120 min after tracer injection.
PET images were reconstructed with built-in corrections for attenuation, normalization, scatter, randoms, and deadtime. PET brain images were registered to the animal's MR image, which was subsequently registered to a brain atlas to define the regions of interest. Dynamic images were reconstructed using a Fourier rebinning and filtered back projection algorithm. A rhesus monkey brain atlas was used for generation of regions of interest (ROIs) and time − activity curves (TACs) for the following ROIs: amygdala, brain stem, caudate nucleus, centrum semiovale (CS), cerebellum, cingulate cortex, frontal cortex, globus pallidus, hippocampus, insula, nucleus accumbens, occipital cortex, pons, putamen, substantia nigra, temporal cortex, and thalamus.

Plasma radiometabolite analysis
Arterial blood samples were collected during the PET scans to measure the radioactivity in plasma for generation of the metabolite-corrected arterial plasma input function. Plasma radiometabolite analysis was performed using the column-switching method, following a published protocol [31]. Briefly, arterial blood samples were collected at 3-, 8-, 15-, 30-, 60-, 90-, 120-, and 180-min post-injection (p. i.), treated with urea (8 M), filtered, and injected onto a self-packed short column (4.6 × 19 mm) eluting with 1% MeCN in water at a flow rate of 2 mL/min. The sample was then back flushed onto a Gemini-NX column (5 μm, 4.6 mm × 250 mm) eluting with 40% MeCN/60% 0.1 M ammonium formate (pH 6.4) at a flow rate of 1.2 mL/min. The eluent was fraction-collected using an automated spectrum chromatography CF-1 fraction collector. Activity in the whole blood, plasma, filtrated plasma-urea mix, filters, and HPLC fractions were counted with automatic gamma well-counter (Wizard 2, PerkinElmer). The sample recovery rate, extraction efficiency, and HPLC fraction recovery were monitored. The unmetabolized [ 18 F]SDM-16 parent fraction was determined as the ratio of the sum of radioactivity in fractions containing the parent compound to the total amount of radioactivity collected and fitted with inverted Gamma4 approaches.

Kinetic modeling
Volume of distribution (V T , mL·cm −3 ) values and the firstorder kinetic rate constants of tracer (K 1 ) were derived through 1-tissue (1T) compartment kinetic modeling with the metabolite-corrected arterial plasma input function, which was calculated as the product of the fitted total plasma curve and the parent radiotracer fraction curve. Nondisplaceable volume of distribution (V ND ) and SV2A occupancy by LEV was calculated using the Lassen plot [32]. Nondisplaceable binding potential (BP ND ) values were calculated from V T values using CS as reference region, or the V ND derived from the blocking study, i.e., BP ND = (V T, ROI − V T, CS )/ V T, CS or BP ND = (V T /V ND ) -1.

Radiation dosimetry study
Two whole-body biodistribution studies were performed in two rhesus monkeys (9.4 kg female and 17.0 kg male) to estimate human organ radiation dosimetry. Scans were carried out on a Biograph mCT (hybrid PET/CT, Siemens Medical Systems, Knoxville, TN) scanner following i.v. injections of 187.6 MBq (5.1 mCi) and 173.9 MBq (4.7 mCi) [ 18 F]SDM-16, and mass doses of 0.27 µg and 0.24 µg at time of injection, respectively. Monkeys were scanned for about 4 h in a sequence of 22-24 passes from top of the head to the midthigh. Scans were reconstructed and visually inspected for organ activity concentrations exceeding background level. The organs included were the brain, heart, liver, gall bladder, spleen, kidneys, and urinary bladder contents. Regions of interest were delineated on these organs, and mean activity values were computed to form TACs.
Within-pass decay correction was removed to reflect the actual activity in each organ, and cumulative activity (Bq . h/cm 3 ) computed by integration of the data from the scan. The tail portions beyond the end of the scan were extrapolated assuming only physical decay of the tracer. These values were multiplied by the organ volumes of a standard 55-kg adult female reference and 70-kg adult male mathematical phantom, and then normalized to injected activity to obtain organ residence times (N, h). Final values were then entered into the OLINDA software to obtain absorbed doses in all organs, which were computed with different voiding assumptions.

In vitro competition-binding assay
Rac SDM-16 possessed high binding affinity to human SV2A, with K i of 3.7 nM in our radioligand competitionbinding assay using

PET imaging experiments in rhesus monkeys
The injected radioactivity ranged from 183 to 188 MBq (n = 10), corresponding to 0.646-0.926 µg of SDM-16. At this microdose level, no adverse events were observed throughout the imaging study. Adverse event was also not observed following LEV (i.v., 30 mg/kg) administration, including the displacement and blocking scan.

Plasma analysis
After the administration of [ 18 F]SDM-16, the tracer concentration in the plasma showed a sharp increase within 5 min, followed by a fast distribution phase and a slow clearance phase. [ 18 F]SDM-16 had higher metabolite-corrected plasma SUV than the other SV2A radiotracers, indicating slower plasma clearance and higher metabolic stability (Fig. 2a).
Whole-blood and plasma-input functions were highly consistent between the two animals, with a stable plasma to whole-blood ratio of 0.93 ± 0.13 over the entire 180-min acquisition period (Fig. S1). In rhesus monkeys, [ 18 F] SDM-16 was metabolized slowly, with 89 ± 5% and 80 ± 7% intact radiotracer present in the plasma at 30-and 120-min post-injection (p.i., n = 9, Fig. 2b , respectively. All observed radiometabolite fractions in the plasma had shorter retention times than the parent tracer, indicating that they were more hydrophilic and less likely to penetrate the BBB (Fig. 2c).

Brain PET image analysis
Summed SUV images from the baseline and blocking scans of [ 18 F]SDM-16 are shown in Fig. 3. At baseline, high contrast between gray matter and ventricles was clearly visible (Fig. 3a) while blocking with LEV significantly reduced the tracer uptake in gray matters (Fig. 3b). [ 18 F]SDM-16 had an apparently slow kinetic profile, with tracer uptake increasing gradually till the end of the scan to an SUV of about 10 (for frontal cortex and putamen, Fig. 3c), which was higher than for [ 11 C]UCB-A (SUV about 4 in Fig. 3d).
Nevertheless, the binding of [ 18 F]SDM-16 was reversible, as demonstrated by the LEV displacement experiment, in which the tracer uptake was reduced by 58 ± 8% (averaged from 5 brain regions), based on the SUV values at the end of the displacement scan (4 h p.i.) and those at 180-min p.i. at the baseline scan of the same monkey (Fig. 3e, Fig. S2). In the pre-blocking study, the preinjected LEV (i.v., 30 mg/kg), also resulted in 57 ± 10% reduced tracer uptake in gray matter regions, based on the normalized terminal SUV values in the blocking and baseline scans (Fig. 3f) tracers, we observed more variability in these plots than what has previously been shown in plots using data from the same subject [36].

Lassen plot
To examine the in vivo binding specificity, SV2A occupancy, and the extent of nonspecific binding in the monkey brain, we performed the Lassen plot analysis using data from the two baseline scans and one blocking scan in the same monkey. The preinjected SV2A ligand LEV (30 mg/ kg, i.v.) blocked 79% of the available SV2A binding sites in all gray matters (R 2 = 0.99), indicating high in vivo binding specificity of [ 18 F]SDM-16 (Fig. 5). The degree of SV2A occupancy by LEV was similar to previously reported with other SV2A PET tracers [19,21,26]. Based on the Lassen plot, the V ND of [ 18 F]SDM-16 in the monkey we imaged was 2.54 mL/cm 3 , which was lower than we previously determined for [ 18

Binding potential
The specific to nonspecific binding signal, as reflected by the non-displaceable binding potential (BP ND ), was calculated either using the nondisplaceable volume of distribution (V ND ) obtained from the blocking study or using the CS V T (V T(CS) ) as the reference. With the V ND method, regional BP ND values ranged from 2.9 to 15.1 (Table 3). Using CS as the reference region, regional BP ND values ranged from 0.69 to 3.90, which was in average 67% lower than those calculated using V ND values. This difference is expected due to the substantial partial volume effect in CS, resulting in an overestimation of V ND when using CS V T . The regional

Test-retest reproducibility
For a preliminary evaluation of the reproducibility of the PK parameter estimation, we scanned one monkey twice with 161 days in between, using [ 18 F]SDM-16. The metabolite-corrected plasma input functions and SUV TACs were highly consistent between the two scans.

Dosimetry
In preparation for the evaluation of [ 18 F]SDM-16 in humans, we performed whole-body distribution studies in two rhesus monkeys (one female and one male). Organ residence   Table S1, while the absorbed doses estimated for the female and male phantom were listed in Table S2. The organ receiving the largest dose was the urinary bladder wall (0.1368 mGy/MBq and 0.0762 mGy/MBq for female and male, respectively), followed by the brain (0.1032 mGy/MBq, 0.0732 mGy/MBq), liver (0.0538 mGy/ MBq, 0.0454 mGy/MBq), kidneys (0.0454 mGy/MBq, 0.0478 mGy/MBq), and the gallbladder wall (0.0441 mGy/ MBq, 0.0484 mGy/MBq). Based on the urinary bladder wall as the critical organ, the maximum permissible singlestudy dosage of [ 18

Discussion
A quantitative tool to image the whole CNS synapses will open the opportunity to study the interplay between the brain and spinal cord, under normal and disease conditions. We have previously reported the synthesis and evaluation of a series of fluorine-18-labeled SV2A PET tracers, which all showed excellent brain imaging properties, and some have been translated into first-in-human studies [37,42,43]. However, these PET tracers are metabolically labile, with less than 50% parent fraction at 30-min post injection. While our data indicate that the radiometabolites are not brain penetrant and would not interfere with the quantitative analysis of brain SV2A expression levels, these radiotracers are not suitable for imaging of SV2A beyond the brain, due to the difference in permeability between BBB and BSCB.
For [ 11 C]UCB-J and [ 18 F]UCB-H, the prevalent radiometabolites in plasma are the corresponding N-oxidation products, which do not enter the brain to a significant extent as reported in LC/MS/MS and small animal PET imaging studies [26] [27]. Thus, we designed a new SV2A radiotracer, based on the structure of UCB-A, which possesses an imidazole ring and lacks the formation of a pyridinyl N-oxide radiometabolite [44,45]. We modified the structure of UCB-A, in a way to fine-tune the physicochemical properties and further improve its in vivo stability and brain kinetics, because UCB-A's PK in human brain is too slow to allow for the reliable estimation of PK parameters using data from a C-11 PET scan with reasonable length [46]. Based on the ChemDraw (Version 20.1.0.112)-predicted LogP values of SDM-16 (2.06) and UCB-A (0.96), we expected to see higher membrane permeability of SDM-16 over UCB-A; as in general, within the same series of compounds, higher lipophilicity is associated with higher cell membrane permeability [47]. However, since a higher fraction of SDM-16 is expected to be protonated at physiological pH than UCB-A, the delivery of SDM-16 from plasma to brain could potentially be hampered if the positively charged molecule does not enter the brain as effectively as the uncharged molecule. The newly designed SV2A ligand SDM-16 binds to human SV2A with high affinity. Based on our experience with the synthesis of [ 18 F]SynVesT-1 [19] and [ 18 F] SynVest-2 using organotin precursors [21], we decided to apply the same radiolabeling strategy for [ 18 F]SDM-16. To our satisfaction, [ 18 F]SDM-16 was synthesized with high radiochemical yield, radiochemical and chemical purities, and molar activities. The relatively higher hydrophilicity of SDM-16 than UCB-J, SynVesT-1, and SynVesT-2 was expected to increase its free fractions in plasma and the brain. Indeed, the plasma free fraction (f P ) of [ 18 F]SDM-16 is 69%, which is slightly lower than that of [ 11 C]UCB-A (75%), but much higher than of [ 11  The relatively slower plasma clearance (Fig. 2a) and higher plasma parent fraction (Fig. 2b) indicate the higher metabolic stability of [ 18 F]SDM-16 over the other existing SV2A radiotracers.
Because of the high metabolic stability and consistently high tracer concentration in the plasma (Fig. 2a), the brain TACs of [ 18 F]SDM-16 appear similar to tracers with irreversible binding kinetics (Fig. 3c). As for other SV2A PET tracers, the 1TC model provided good fits and reliable estimates of PK parameters of [ 18 F]SDM-16 (Fig. 3c). The excellent 1TC fitting and the efficient displacement by LEV  [48][49][50], N-protonated lutidine has higher pK a than N-protonated dimethyl imidazole (Fig. 7). in the plasma as free bases, based on the experimental pK a data of lutidine (6.77) and dimethyl imidazole (7.42, 7.85), and the pH value of plasma being 7.4. The acid/base property is a critical factor to consider in drug discovery, especially for CNS drugs that have special requirement for BBB penetration [51]. The pK a values of all these SV2A ligands are within the commonly accepted range for pK a of CNS drugs, i.e., from 4 to 10 [34]. While the higher plasma free fraction of SDM-16 favors its delivery from plasma into the brain, the more extensive protonation of SDM-16 and its relatively higher polar surface area could contribute to its lower K 1 values than [ 11 (Table S3). In average, the brain equilibration half-life of [ 11 C]UCB-A is about 1.7-fold longer than that of [ 18 F]SDM-16. Although [ 18 F]SDM-16 displaced relatively slow kinetics in the rhesus monkey brain where SV2A expression level is high (SV2A B max of Baboon' brain ranged from 2.2 pmol/mg protein in the pons to 19.9 pmol/mg Fig. 7 The influence of pK a on the properties of PET tracers protein in the temporal cortex) [5], its kinetics is expected to be faster in tissues with relatively lower SV2A expression, such as spinal cord [12].
Although in the field of PET neuroimaging, the rule of thumb is that the B max /K d of the PET tracer needs to be greater than 10, the ratio of two tracers' BP ND is determined partially by their degree of nonspecific uptake, which is reflected in V ND and brain tissue free fraction (f ND ). While V ND can be obtained only through in vivo blocking studies, f ND can be obtained either from in vivo blocking study or from in vitro assays using brain homogenates or slides, and f ND is considered to be consistent among different species [52]. While decreasing the tracer's K d value may eventually leads to undesired slow kinetics (low k 2 and long brain-toplasma equilibrium half-life), increasing f ND is an alternative but potentially more challenging approach to boost the specific PET signal, based on the equation BP ND = f ND *B max /K d . Using the averaged V ND and f P values, we calculated the f ND value of [ 18 F]SDM-16 to be 27%, which was higher than that of [ 18 F]UCB-H (6.1%, calculated from K 1 /k 2 using 2TCMc) [35], [ 11 C]UCB-J (7.3%), [ 18 F]SynVesT-1 (10.1%), and [ 18 F]SynVesT-2 (19.5%), assuming that these SV2A ligands enter the brain mainly through passive diffusion and are not subject to active influx or efflux transport, i.e., f ND = f P /V ND . We did not calculate the V ND and f ND values of [ 11 C]UCB-A because of the lack of blocking data for [ 11 C]UCB-A. [ 18 F] SDM-16 has the highest f ND value among all the current SV2A PET tracers and maintains the brain penetration ability.
To compare the in vivo K d and BP ND of [ 18 F]SDM-16 with those of [ 18 F]SynVesT-1 [19], [ 11 C]UCB-J [26], and [ 11 C]UCB-A in monkey brain, we adopted the Guo plot using their baseline V T values (Fig. 4). The K d ratios are Because we used different monkeys in the evaluations of these SV2A PET tracers, the BP ND ratios or in vivo K d ratios could be influenced by animal differences.
Next, we calculated the BP ND values of the SV2A PET tracers using either CS as reference region or using the V ND derived from blocking studies. We noticed that the BP ND values calculated using the V T values of CS are 67.2 ± 4.4% lower than the true BP ND derived from V ND values. Contributing factors to the underestimation of BP ND using CS as reference region are the spill-in effect of the PET signal from the gray matter surrounding CS and the presence of SV2A specific uptake in CS. The ranking order of [ 18 (Table 3 and Fig. 6). However, since the monkeys used in each tracer's evaluation are different, further studies using the same cohort of monkeys are needed to confirm if [ 18 F] SDM-16 possesses higher specific binding than the other SV2A PET tracers in NHP brains. An SV2A PET tracer with high specific binding signals will be advantageous in the imaging and quantification of SV2A in tissues with relatively low SV2A expression, e.g., spinal cord [12] and pancreas [53]. In fact, the BP ND values of [ 18 F]SDM-16 in the LEV blocking scan are relatively high in the gray matters (up to 2.11 in cingulate cortex and nucleus accumbens), even with 79% of the SV2A being occupied by LEV, indicating that [ 18 F]SDM-16 is advantageous in the imaging and quantification of SV2A at much less densities than the cerebrum.

Conclusions
We have successfully synthesized a new 18 F-labeled SV2A PET tracer [ 18 F]SDM-16 and evaluated its imaging characteristics in rhesus monkeys. [ 18 F]SDM-16 is metabolically more stable than the current SV2A PET tracers and displayed reversible and high specific binding in NHP brain with relatively low nonspecific binding in white matter. The TACs fitted well with 1TC to allow for reliable estimation of PK parameters. [ 18 F]SDM-16 may have potential applications in the quantification of SV2A in the whole CNS.