Ruthenium-Mediated 18F-Fluorination and Preclinical Evaluation of a New CB1 Receptor Imaging Agent [18F]FPATPP

Cannabinoid receptor 1 (CB1R) controls various physiological and pathological conditions, including memory, motivation, and inflammation, and is thus an interesting target for positron emission tomography (PET). Herein, we report a ruthenium-mediated radiolabeling synthesis and preclinical evaluation of a new CB1R specific radiotracer, [18F]FPATPP. [18F]FPATPP was produced with 16.7 ± 5.7% decay-corrected radiochemical yield and >95 GBq/μmol molar activity. The tracer showed high stability, low defluorination, and high specific binding to CB1Rs in mouse brain.


■ INTRODUCTION
Cannabinoid receptors, part of the endocannabinoid system, are expressed throughout the nervous system. Cannabinoid receptors are divided into two classes, cannabinoid receptor 1 (CB1R) and cannabinoid receptor 2 (CB2R). CB1R is a seven transmembrane domain G protein-coupled receptor (GPCR) and is one of the most abundant GPCRs in the brain. 1,2 In the brain, CB1Rs are localized, e.g., in cerebral cortex, especially the allocortex, basal ganglia and globus pallidus, substantia nigra, cerebellum, and medulla. Among vertebrate species, CB1Rs are highly conserved. 3,4 Until recently, unlike CB1Rs, the expression of CB2R in the CNS was thought to be very low. 1,2 However, recent studies have demonstrated the expression of the CB2 A isoform in the brain. These receptors can be upregulated under pathological conditions. 5,6 In the periphery and the brain, CB1Rs are mainly found in presynaptic neurons where they inhibit release of neurotransmitters, e.g., γ-aminobutyric acid (GABA) and glutamate. 1,2 Activation of CB1Rs leads to activation of various signal transduction pathways, such as adenylyl cyclase regulation and ion channel regulation. 1 CB1Rs control various physiological and pathological conditions such as brain development, memory and learning, appetite, motivation, sensation of pain, and inflammation. This result is in agreement with the high CB1R concentration found in the sensory, cognition, and motor regions in the brain. 2 Alterations in the endocannabinoid system and CB1R expression can be caused by various neuropsychiatric and neurodegenerative conditions, 7 e.g., Alzheimer's disease (AD) impairs the CB1R activity. 8 Positron emission tomography (PET) has been utilized to monitor CB1R expression in humans. Until this point, a few CB1R specific 11 C-or 18 Figure 1). 11,12 Fluorine-18 is an attractive radionuclide for PET applications due to its physical properties. Its relatively long half-life (109.8 min) enables even demanding synthesis and imaging protocols. The low positron energy (maximum ß + energy 635 keV) allows for high resolution imaging. In addition, aqueous [ 18 F]fluoride can be easily accessed with a cyclotron. However, the use of [ 18 F]fluoride in traditional nucleophilic labeling reactions, especially in the labeling of electron-rich structures, can be challenging. Late-stage 18 F-fluorination of arenes and heteroarenes with [ 18 F]fluoride has been widely studied, and various synthesis methods with various labeling precursors have been developed. 13−17 Metal-free approaches to 18 F-fluorination utilize aryliodonium salts, 18 arylsulfonium salts, 19 and iodonium ylides 20 as precursors. Metal-mediated 18 Ffluorination utilizes complex Pd and Ni compounds as labeling precursors in the Pd- 21 and Ni-mediated 22 labeling processes, respectively. In Cu-mediated 18 F-fluorination, arylboronic esters, 23 arylboronic acids, 24 and arylstannanes 25 have been used as labeling precursors. Cu-mediated 18 F-labeling has gained wide popularity due to the straightforward synthesis, relatively easy access to precursor materials, and wide substrate scope. However, Cu-mediated labeling of compounds containing unprotected alcohols or amines can lead to undesired side reactions. Recently, a Ru-mediated 18 F-deoxyfluorination reaction via an in situ prepared η 6 -coordinated ruthenium− phenol complex was introduced. 26 Ruthenium has been successfully applied in the 18 F-fluorination of small molecules containing electron-rich structures, including unprotected amines and heterocyclic scaffolds, 26 and in the site-specific 18 F-labeling of small peptides. 27 18 F-Deoxyfluorination is an appealing method due to the use of easily accessible phenols as starting materials.
Recently, [ 18 F]FMPEP-d 2 has been widely used in CB1R PET studies, e.g., for imaging in schizophrenia and changes in CB1Rs in first-episode psychosis, 28 to study differences in brain CB1Rs between the sexes, 29 in brown adipose tissue, 30 and in AD. 31 An advantage of [ 18 F]FMPEP-d 2 is its high selectivity toward CB1Rs over CB2Rs, especially in regard to the ongoing discussion 5 of CB2R expression in the brain. A shortcoming of this tracer is its tendency toward radiolytic decomposition. 10,32 In this study, our objective was to utilize the ruthenium-mediated radiofluorination pathway to produce a stable CB1R specific tracer (3R,5R)-5-(3- Figure 1), an analogue of [ 18 F]FMPEP-d 2 , and to evaluate its usefulness for imaging in vivo and ex vivo with wild-type mice.

■ RESULTS AND DISCUSSION
Chemistry. We developed a simple route to produce a nonradioactive reference standard 4 for [ 18 F]4, following the original precursor synthesis, 10 as presented in Scheme 1. Briefly, compound 5 was produced from ethyl pyruvate, 4aminobenzotrifluoride, and 3-methoxybenzaldehyde in acetic acid with 45% yield. Acid-catalyzed hydrolysis of 5 with hydrochloric acid and glacial acetic acid yielded compound 6.
Compound 7 was separated from the mixture of isomers formed in the reaction with (R)-(+)-1-phenylethylamine and was further reduced with sodium cyanoborohydride to yield the reference standard 4. The ruthenium complex, CpRu-(COD)Cl was synthesized as previously reported 26 with a total synthesis yield of 20 ± 4% (n = 3) Radiochemistry. We adapted a ruthenium-mediated labeling method 26 to produce [ 18 F]4 starting from the phenol precursor 8, CpRu(COD)Cl complex, and no-carrier added [ 18 F]fluoride ( Figure 2A). In situ prepared ruthenium−phenol complex was used for the 18 F-radiolabeling reactions without purification. We performed an optimization study for the 18 Fradiofluorination reaction. The effect of the activation method for aqueous [ 18 F]fluoride on radiochemical yield (RCY) was studied by using solid-phase extraction (SPE) (bars 1−2, Figure 2B) and traditional azeotropic distillation (bar 4, Figure  2B). With the SPE method, the reaction temperature was varied from 100 to 130°C (bar 1, Figure 2B) and further to 160°C (bar 2, Figure 2B) with 10, 20, and 30 min reaction times. With the azeotropic distillation method, the reaction was conducted at 160°C for 10−30 min (bar 4, Figure 2B). To study the effect of the base (potassium carbonate), we activated [ 18 F]fluoride with the SPE method, added potassium carbonate to the reaction mixture, and followed the reaction for 30 min at 160°C (bar 3, Figure 2B).
In the SPE method, the ruthenium−phenol complex was used for the [ 18 F]fluoride elution from the anion exchange cartridge with 93 ± 5% elution efficiency (n = 12). Using 100°C reaction temperature, no reaction occurred even after 30 min. Elevation of the temperature from 100 to 130°C afforded a RCY (based on HPLC analysis of the crude product) of 20.9 ± 0.3% (n = 3) after 30 min reaction (bar 1, Figure 2B). Further elevation of the temperature to 160°C provided 47.4 ± 3.5% (n = 3) RCY (based on HPLC analysis of the crude product) after 30 min reaction (bar 2, Figure 2B).
The reaction occurred faster when [ 18 F]fluoride was activated via azeotropic distillation compared to the activation with the SPE method. With azeotropic distillation activation    Figure 2B), the RCY was already 36.3 ± 5.6% after a 10 min reaction and 47.9 ± 8.5% after a 30 min reaction (values based on HPLC analysis of the crude product) at 160°C. Even though some of the initial [ 18 F]fluoride was lost during the azeotropic drying process via adherence to the glass vial, the RCY was considerably higher (36 vs 16%) after a 10 min reaction at elevated temperature compared to the SPE method.
The reaction mixture is more basic when using the azeotropic distillation method due to the addition of potassium carbonate. The effect of the additional base was further studied with the SPE method (bar 3, Figure 2B); the RCY was 17.9 ± 5.5% (n = 3) after a 10 min reaction time and 61.8 ± 4.9% (n = 3) after a 30 min reaction time. Thus, the addition of a base improved the RCY by ∼30% with a 30 min reaction.
For PET applications, late-stage fluorination approaches are highly desirable to minimize the radiosynthesis time and to avoid the multiple purification processes often needed in multistep radiosyntheses. Ruthenium-mediated 18 F-fluorination starting from a phenol precursor is a facile labeling method with reasonable synthesis time and good RCY. The present optimized reaction conditions, 30 min at 160°C, were even harsher than typical reaction conditions of traditional nucleophilic 18 F-fluorination. However, the good RCY and robustness of the present conditions (bar 3, Figure 2B), demonstrated by the low SD, are advantageous. For potential clinical applications, the ruthenium content in the formulated preparation must be considered. The maximum parenteral daily exposure of ruthenium according to ICH Q3D(R1) is limited to 10 μg/day. 33 In this material, the ruthenium content is rather high, limiting the administrable volume of [ 18 F]4. Thus, the purification process needs to be refined in order to reduce the ruthenium content before clinical tracer production can be conducted via ruthenium-mediated 18 F-labeling.
For PET tracers for clinical use, a generally acknowledged limit for RCP is 95%. A significant difference was found in the RCP at EOS and stability of [ 18  Animal Studies. The animal studies were divided into five sections; in vivo 120 min dynamic PET/computed tomography (CT) studies, ex vivo digital autoradiography of brain at 120 min, ex vivo biodistribution at 30, 60, and 120 min, and radiometabolite analysis at 30, 60, and 120 min. The fifth separate section was an ex vivo biodistribution study with [ 18 F]FMPEP-d 2 . The potency and receptor selectivity of [ 18 F]4 was not determined separately. However, blocking studies using the selective CB1R antagonist rimonabant (SR141716) were conducted for both in vivo PET and ex vivo autoradiography and biodistribution. Rimonabant is highly specific for CB1Rs in the CNS (K i 1.98 nM), and its selectivity for CB1Rs over CB2Rs is more than 1000 fold. 34 Rimonabant is thus suitable to demonstrate the specificity of the tracer to CB1Rs.
In Vivo PET Imaging. In vivo evaluation was performed for [ 18 F]4 with wild-type mice with (n = 3, 1 female) and without blocking (n = 3, 1 female). The [ 18 F]4 tracer uptake reached its maximum 30 min after the injection. This observation is comparable to a mice study by Takkinen et al. with [ 18 F]FMPEP-d 2 , which similarly reached its tracer uptake maximum at 30 min. 31 In a rat study by Casteels et al. with [ 18 F]MK-9470, the tracer uptake was considerably slower, reaching its maximum at 300 min post injection (p.i.). 11 The peak standardized uptake values (SUVs) of [ 18 F]4 were 1.92 for the whole brain, 1.80 for the neocortex, and 2.14 for the hippocampus ( Figure 3A−D). This outcome is somewhat higher than with [ 18 F]FMPEP-d 2 , where the whole brain SUV was 1.5, 31 and noticeably higher than with [ 18 F]MK-9470, where the whole brain SUVs were 0.6−1.6 in rats. 11  The pretreatment with 2 mg/kg rimonabant reduced the binding of [ 18 F]4 120 min after injection by 58, 59, and 62% for the whole brain, neocortex, and hippocampus, respectively ( Figure 3A−D), when the injected amount of [ 18 F]4 was 1.7 ± 0.7 μg/kg (n = 6). The moderate dose of rimonabant blocked a large percentage of the binding of [ 18 F]4 in specific murine CB1R-rich areas, such as the neocortex and hippocampus, and avoided CB1R poor areas, such as the thalamus.
Ex Vivo Autoradiography. In addition to the in vivo PET studies, we performed ex vivo brain autoradiographic studies 120 min after [ 18 F]4 injection to examine smaller brain areas in more detail. Figure 4A shows specific [ 18 F]4 binding in CB1Rrich areas, like striatum, globus pallidus, hippocampus, cortex, and cerebellar gray matter. The binding of [ 18 F]4 even allowed us to distinguish different hippocampal and cortical layers and the cerebellar gray matter with similar resolution to the in vitro studies in mice using the CB1R agonist [ 3 H]CP55,940 37,38 or showing similar hippocampal distribution to that observed in CB1R immunohistochemical studies. 39 Previous autoradiography studies in mice with the PET radioligand [ 18 F]FMPEPd 2 did not discriminate the cortical and hippocampal layering with as high a resolution. 31 The blocking of [ 18 F]4 with rimonabant was also evident in the autoradiographic images ( Figure 4A). In addition, we quantified the binding of [ 18 F]4 in the different CB1R-rich regions using the thalamus as a reference region given the very low level of CB1Rs in that brain region in mice ( Figure 4B). The ratios with respect to the thalamus were 2.3 for the parietotemporal cortex, 2.8 for the striatum, 2.5 for the frontal cortex, 2.6 for the whole hippocampus, 3.7 for the cortex and cerebellar gray matter, and 4.9 for the globus pallidus. The region-to-thalamus ratios in the mice pretreated with rimonabant were close to 1, indicating complete displacement of specific tracer binding. The P values were less than 0.001 for all analyzed regions.
In a separate study, we investigated the ex vivo biodistribution of [ 18 F]FMPEP-d 2 in healthy mice ( Figure  5B). It is worth noting that [ 18 F]FMPEP-d 2 had lower 18 F-  Metabolite Analysis. [ 18 F]4 was metabolized slowly. In the plasma and cortex samples collected at 120 min p.i. (n = 5), two polar radioactive metabolites in plasma and one in cortex were observed. The amount of residual [ 18 F]4 in plasma was approximately 21.3 ± 3.5% of the total 18 F-radioactivity at 120 min p.i. The amount of residual [ 18 F]4 in the cortex at 120 min p.i. was 75.0 ± 4.3% ( Figure 6). The amount of residual [ 18 F]FMPEP-d 2 has been shown to be 86% at 180 min p.i., which is somewhat higher than the value obtained with [ 18 F] 4. 31 However, the higher brain uptake of [ 18 F]4 and much lower formation of [ 18 F]fluoride as compared to [ 18 F]FMPEPd 2 will presumably be an advantage for imaging.
Chromatographic Methods. NMR analyses were performed with a 500 MHz Bruker AVANCE-III NMR-system. The multiplicities are abbreviated as follows: s = singlet, d = doublet, t = triplet, m = multiplet, br = broad signal, dd = doublet of doublets, etc. MS analyses were performed with a linear ion trap quadrupole mass spectrometer (QTRAP, Applied Biosystems SCIEX) equipped with a Turbo Ion Spray source. Purity of the reference standard 4 was determined with HPLC. Ruthenium content in the final preparation was analyzed with inductively coupled plasma mass spectrometry (ICP-MS; PerkinElmer, Elan DRC Plus). Commercial standards were used for instrument calibration.  (7). (R)-(+)-1-Phenylethylamine (5.28 mL, 46 mmol) was added to a solution of 6 (8 g, 23 mmol) in CH 2 Cl 2 (58 mL). The solution was stirred at room temperature for 18 h. A silica gel column was prepared in hexane. The solution was evaporated to dryness and dissolved with a small volume of EtOAc and was then poured onto a silica gel column. The material was purified by silica gel chromatography (5−15% EtOAc/hexane) to afford 7, the second eluting isomer, as a yellow foam (2.8 g, 28%). 1  Radiochemistry. General Radiochemistry Methods. RCYs are decay-corrected to the start of the synthesis, which in this case corresponds to EOB. The RCYs (based on HPLC analysis) are based on the overall radioactivity eluted from the HPLC column, whereas the product fraction was collected in a separate vial. We determined that there was no leftover radioactivity in the injector or in the HPLC column after the analytical HPLC run. A m calculation is decaycorrected to EOS. The concentration of the 4 in the formulated solution was low, below the detection limit of the UV-detector. Thus, the A m value is based on the radioactivity collected at the analytical HPLC column outlet and the detection limit for the reference compound 4 (1 μg/mL).
Radiochemistry Optimization Study. Prior to the production of [ 18 F]FPATPP for preclinical studies, the Ru-mediated radiofluorination method was tested with temperatures from 100 to 160°C and reaction times from 10 to 30 min. In addition, for drying of [ 18 F]fluoride, a SPE method and an azeotropic distillation method were tested.
In the SPE method, a PS-HCO 3 − anion exchange cartridge (Synthra GmbH, Hamburg, Germany) was used. Aqueous [ 18 F]F − was trapped in a cartridge preconditioned with potassium oxalate (3 mL, 10 mg/mL) and water (2 mL). The cartridge was washed with CH 3 CN (1 mL), and [ 18 F]F − was eluted to a conical vial with mixture A. The cartridge was rinsed with DMSO (150 μL) and 1:1 DMSO/ MeCN solution (50 μL). The reaction mixture was heated at 100, 130, or 160°C for 30 min, and the reaction was followed with radio-HPLC. In the base-added SPE method, K 2 CO 3 (40 μL, 14.8 ± 4.2 μmol, 45 mg/mL) was added to DMSO (150 μL) used for cartridge rinsing after [ 18 F]fluoride elution. The reaction mixture was heated at 160°C for 30 min, and the reaction was followed with radio-HPLC.
In the azeotropic distillation method, aqueous [ 18 F]F − solution was dried using azeotropic distillation with MeCN, K 2 CO 3 (14.8 ± 4.2 μmol), and Kryptofix K 222 (16.6 ± 0.6 μmol) at 120°C with helium flow to form the dry [ 18 F]KF/K 222 -complex. Mixture A was added to the reaction vessel, and the reaction was heated at 160°C for 30 min. The reaction was followed with radio-HPLC.

Synthesis of [ 18 F]4 for Preclinical Studies. Aqueous [ 18 F]F − was trapped in a PS-HCO 3
− anion exchange cartridge (Synthra GmbH) preconditioned with potassium oxalate (3 mL, 10 mg/mL) and water (2 mL). The cartridge was washed with 1 mL CH 3 CN, and [ 18 F]F − was eluted to a conical vial with the mixture A as described above. The cartridge was rinsed with DMSO (150 μL) and 1:1 DMSO/ MeCN solution (50 μL). This mixture was heated at 160°C for 30 min. After heating, the mixture was diluted with HPLC eluent (1 mL) and purified by semipreparative HPLC. The product fraction was collected, diluted with water (20 mL), and loaded on a C18 Plus Light Sep-Pak cartridge (Waters corp, Milford, Massachusetts) preconditioned with ethanol (7 mL) and water (10 mL). The cartridge was washed with water (20 mL), and the product was eluted with ethanol (300 μL) followed by 9 mg/mL NaCl (aq) (2.7 mL).
In Vivo PET Imaging with [ 18 F]4. Dynamic PET scans were performed for 5 month old mice with an Inveon Multimodality PET/ CT scanner (Siemens Medical Solutions USA, Knoxville, Tennessee). Mice were anesthetized with 2.5% isoflurane/oxygen gas, and body temperature was maintained with a heating pad and a bubble-wrap cover. CT transmission scans were performed to correct for attenuation. PET scans were initiated in parallel with an intravenous (iv) bolus injection of [ 18 F]4 to the mouse tail vein (4−5 month old, n = 6, 2 females; injected radioactivity 4.0 ± 0.4 MBq; injected mass 56 ± 25 ng, 1.7 ± 0.7 μg/kg).
Dynamic emission scans were acquired for 120 min in 3D list mode, with an energy window of 350−650 keV. Scanning times were 120 min (54 frames 30 × 10 s, 15 × 60 s, 4 × 300 s, 2 × 600 s, 3 × 1200 s). The list mode data were stored in 3D sinograms. Data were reconstructed with OSEM3D/MAP software (2 OSEM3D iterations and 18 MAP iterations). The PET/CT images were preprocessed in Matlab R2017a (The MathWorks, Natick, Massachusetts) with an inhouse semiautomated pipeline for preclinical images that uses SPM12 (Wellcome Department of Cognitive Neurology, London, UK) preprocessing functionalities and analysis routines. Images were first cropped to a bounding box containing the heads, and individual PET images were coregistered through a rigid-body transformation to their corresponding CT scan. Subjects were spatially normalized through a two-step registration (a rigid followed by an affine transformation) of each subject's CT to a template CT that was previously constructed as an average of several subjects and was aligned with an atlas T2weighted MRI template. The combination of transformations was then applied to the PET images, which were also resampled to a voxel size of 0.2 × 0.2 × 0.2 mm 3 (trilinear interpolation), matching the anatomical atlas dimensions.
Volume of interest (VOI) analysis of the whole brain, neocortex, and hippocampus was performed on each subject by averaging the signal inside a slightly modified version of the Ma et al. 40 atlas delineated VOIs, and data were obtained as standardized uptake values (SUVs). SUVs were calculated as follows: SUV = tissue activity concentration (Bq/mL)/(injected dose (Bq)/body weight (g)); it was assumed that 1 mL of tissue equals 1 g. Tissue activity and dose were decay corrected to the same time point.
For autoradiography, brains were further frozen and cut into 20 μm coronal cryosections for digital autoradiography, and the slices were exposed on an imaging plate (BAS-TR2025, Fuji Photo Film Co.) for approximately 2 half-lives and scanned with a Fuji BAS5000 phosphorimager (FUJIFILM Life Science, Stamford, Connecticut). Digital autoradiographic images were analyzed for count densities with the Aida 4 program (Raytest Isotopenmessgeraẗe GmbH, Straubenhardt, Germany). For brain slices, regions of interest (ROIs) were drawn over the parietotemporal cortex, striatum, frontal cortex, hippocampus, cerebellar gray matter, and thalamus. The autoradiography was quantified as ratios of the different ROIs relative to the thalamus, which was used as the reference region. F]4 on mouse cerebral CB1Rs was examined in a 120 min in vivo study, and then, the same mice were used for ex vivo biodistribution and brain autoradiography studies. An inverse CB1R agonist rimonabant (90 μL, 2 mg/kg in 10% EtOH in Kleptose β-cyclodextrin; rimonabant, Merck; Kleptose β-cyclodextrin, APL pharma specials, Stockholm, Sweden) was administered iv into the mouse tail vein (n = 3, 1 female) 10 min prior to injection of [ 18 F]4. Control mice (n = 3, 1 female) were the given vehicle (90 μL, 10% EtOH in Kleptose βcyclodextrin). The binding specificity of [ 18 F]4 was estimated by comparing the binding of [ 18 F]4 in the presence and absence of rimonabant. Mice were sacrificed 120 min after iv injection of [ 18 F]4, and the 18 F-radioactivity distribution differences between pretreated and control mice were examined with in vivo PET imaging, ex vivo autoradiography, and 2480 Wizard Gamma Counter, and the data were analyzed as described in the previous sections.
Metabolite Analyses with [ 18 F]4. The 2 month old animals (n = 11, used only for ex vivo studies) were used for metabolite analysis. Plasma and brain samples were collected at 30 (n = 4), 60 (n = 2), and 120 min (n = 5) p.i. Plasma samples were treated with MeCN and centrifuged to remove the proteins from the solution. Brain tissue samples were treated with the 40/60 (v/v) 1% TFA (aq)/MeCN solution and then centrifuged. The supernatants were applied to HPTLC silica gel 60 RP-18 plates (Art. 1.05914.0001, Merck, Darmstadt, Germany), and the plates were developed in 40/60 (v/v) 1% TFA (aq)/ MeCN. After drying, the TLC plates were exposed to an imaging plate for approximately 4 h and the imaging plate was scanned with a Fuji Analyzer BAS5000. The amount of [ 18 F]4 and its radioactive metabolites were analyzed with the Aida 4 program, and the curves were fitted with GraphPad Prism 6.0 (GraphPad Software, San Diego, California) using a double exponential decay equation.
Statistical Analyses. Values are expressed as mean ± standard deviation (SD). For ex vivo brain autoradiography and biodistribution, statistical analyses were performed with the Mann−Whitney U test (GraphPad Prism 6.0). Author Contributions ▲ S.L. and N.A.R. contributed equally. All authors have given approval to the final version of the manuscript.

Funding
The study was funded by the Academy of Finland (grants no. 266891 and 307924) and State Funding for University Level Health Research (project no. 11017).