Development of [18F]LU14 for PET Imaging of Cannabinoid Receptor Type 2 in the Brain

Cannabinoid receptors type 2 (CB2R) represent an attractive therapeutic target for neurodegenerative diseases and cancer. Aiming at the development of a positron emission tomography (PET) radiotracer to monitor receptor density and/or occupancy during a CB2R-tailored therapy, we herein describe the radiosynthesis of cis-[18F]1-(4-fluorobutyl-N-((1s,4s)-4-methylcyclohexyl)-2-oxo-1,2-dihydro-1,8-naphthyridine-3-carboxamide ([18F]LU14) starting from the corresponding mesylate precursor. The first biological evaluation revealed that [18F]LU14 is a highly affine CB2R radioligand with >80% intact tracer in the brain at 30 min p.i. Its further evaluation by PET in a well-established rat model of CB2R overexpression demonstrated its ability to selectively image the CB2R in the brain and its potential as a tracer to further investigate disease-related changes in CB2R expression.


Introduction
The psychoactive effect of the well-known drug marijuana is caused by the active ingredient (−)-∆9-trans-tetrahydrocannabinol (THC), which abundantly occurs in the cannabis plant [1]. The investigation of the mechanism through which THC manifests its biological activity led to the discovery of the endocannabinoid system. It comprises a class of transmembrane proteins that belongs to the superfamily of G-protein-coupled receptors; their endogenous ligands (endocannabinoids), such as anandamide (AEA) [2] and 2-arachidonoylglycerol (2-AG); the endocannabinoid-synthesising (diacylglycerol lipase N-acylphosphatidyl-ethanolamine phospholipase D) and -degrading enzymes (fatty acid amide hydrolase, monoacylglycerol lipase), as well as the endocannabinoid transporters [3,4]. Two types of cannabinoid receptors, namely cannabinoid receptors type 1 (CB1R) [5] and cannabinoid receptors type 2 (CB2R) [6], were identified so far and are being thoroughly investigated [7,8]. Both CB1R and CB2R transduce signals from extracellular to intracellular space by inhibiting the activity of adenylyl cyclase over G i/o proteins, thereby suppressing the subsequent cyclic adenosine monophosphate (cAMP) pathway [9,10]. While the CB1R is associated with the central nervous system and is abundantly expressed in the brain (highest density in the hippocampus, cerebellum and striatum), the CB2R is found in the spleen, tonsils and the thymus gland and modulates immune cell functions. In the brain, the CB2R are present at low levels [11][12][13][14][15], and their overexpression is associated with neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, Huntington's disease and cancer [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30]. More recently, it has been shown that CB2R is expressed in dopaminergic neurons in the midbrain and, in glutamatergic neurons in the red nucleus, modulates motor behaviour in mice [31]. Others have shown that the selective activation of the CB2R receptor results in cell apoptosis, the inhibition of tumor cell growth and the inhibition of neo-angiogenesis [32][33][34]. Moreover, selective CB2R agonists were successfully used to suppress inflammatory and neuropathic pain without psychotropic side effects caused by the use of THC or other CB1R-related pharmaceuticals [35,36]. The overexpression of CB2R receptors was correlated with microglial polarisation from a pro-inflammatory M1 to anti-inflammatory M2 state by a yet not fully elucidated mechanism [37]. Thus, CB2R are regarded as a highly valuable marker for the early detection and treatment of several neuropathological conditions [38,39]. The investigation of CB2R as a target to develop novel therapeutic and early diagnostic strategies gained importance over the past years. In the search for novel agonists and antagonists acting at the CB2R, several classes of compounds were thoroughly investigated by systematic SAR (structure-activity relationship) studies [40]. The development of novel potent and selective CB2R ligands is supported by computer-based docking studies and accelerated by the recently reported crystal structure of the receptor [41]. The non-invasive imaging of CB2R with positron emission tomography (PET) is of great interest, as demonstrated by the large number of positron-emitting radioligands developed over the past years [42][43][44][45][46].
Despite tremendous efforts, the development of a suitable radiotracer for PET imaging of the CB2R in the brain remains a challenging task [47]. To gain eligibility for consideration as a potential PET radioligand for brain imaging, the first criterion to be fulfilled by a CB2R ligand is high receptor affinity and selectivity, which in many cases is hampered by the homologous similarity between CB1R and CB2R. Another important criterion is a balanced lipophilicity which is usually used as a predicting tool for (1) the ability to cross the blood-brain barrier and (2) non-specific binding. The development of CB2R ligands with optimal lipophilicity as a brain-targeting drug is also challenging due to the highly lipophilic nature of the CB2R binding pocket and its ability to primarily recognise highly lipophilic compounds (e.g., THC has a LogD 7.4 > 7) [48,49]. Figure 1 shows some of the most representative CB2R radioligands [42,47]. Among them, [ 11 C]NE40 [50,51] is the only CB2R radioligand investigated in humans so far [52]. The suitability of CB2R imaging in pathologic brain conditions was demonstrated by Horti and co-workers by using the 2-aminothiazole [ 11 C]A-836339 [53]. Further structurally related radioligands were developed by Moldovan and co-workers, who designed [ 18 F]JHU94620, [54,55] and Caille and co-workers, who synthesised [ 18 F]FC0324 [56,57]. However, all these radioligands suffer from low metabolic stability in vivo. A comprehensive work was performed by the group of Ametamey, which reported on a series of 11 C and 18 F-labelled quinolinones ( Figure  1) [58] and thiophenes (not shown in Figure 1) [59]. Further, labelled 2,3,5-oxodiazoles were prepared by Ahamed [60] and co-workers as derivatives of radioligands earlier reported by Teodoro [61] and co-workers and evaluated for their ability to detect CB2R in vivo. Recently, considerable progress was made by the discovery of the 11 C-and 18 F-labelled pyridine radioligands [ 11 C]RSR-056 and [ 18 F]3 [62,63] and, most importantly, the deuterated analogue [ 18 F]RoSMA-18-d 6 , [64], which appears to be the most promising CB2R radioligand for PET imaging so far.
In this work, we focus on the development and biological evaluation of a radiofluorinated compound belonging to the naphthyridin-2-one class, and we selected compound 5 (Scheme 1) due to the published high affinity and selectivity towards CB2R (K i (hCB2R) = 1.3 nM, SI > 700) and its favourable physicochemical properties for targeting the brain [34,65,66]. Compound 5 was characterised as a diastereomeric mixture; however, for similar compounds, a considerably higher binding affinity was described for the cis as compared to the trans diastereomer [66], and therefore, we pursued the development of the 18 F-labelled cis-5 (LU14, Scheme 1). In order to evaluate in vivo the ability of LU14 to image CB2R in the brain by PET, we selected a well-established rat model carrying an adeno-associated viral (AAV2/7) vector expressing hCB2R(D80N) at high densities in a striatal region [50,67,68].

Organic Chemistry
The synthesis of the chosen reference compound 5 is depicted in Scheme 1 and was performed based on the synthetic route reported by Lucchesi et al. [66]. In brief, naphtyridin-2-one-3-ethyl ester 3 was condensed with cis-/trans-4-methylcyclohexylamine under thermal reaction conditions to give amide 4, which was further N-alkylated under alkaline reactions conditions to provide 5. For compounds similar to 5, Lucchesi and coworkers [66] reported a strong discrepancy between the binding affinities of the cis and trans diastereomers towards the CB2R caused by a different binding modality/interaction within the CB2R binding pocket. This encouraged us to perform the diastereomeric separation of 5 achieved by flash chromatography on silica 40-63 Micron eluting with isohexane(45-60)/ethyl acetate (IH/EA), gradient 9:1 to 3:2. The purity determination and identification of the two diastereomers were performed by HPLC, 1D and 2D-NMR spectroscopy (see Supporting Information) and by comparing the chemical shifts with published data.

In Vitro Binding Assay
In the present work, we evaluated the binding affinity of the separated diastereomers cis-5 and trans-5 towards the CB1R and CB2R and determined the K i (hCB2R) of 6 nM for cis-5 and of 52 nM for trans-5 ( Figure 2). For 5, a K i (hCB2R) of 1.4 nM was reported [66]. The binding affinity of cis-/trans-5 was previously determined by Lucchesi and co-workers by using isolated membranes from HEK-293 cells transfected with CB2R and [ 3 H]CP-55,940 as a competitor. It is highly likely that the slight difference between the reported and herein determined CB2R binding affinities is related to different experimental protocols. Our data confirm the high affinity of 5 towards the CB2R as well as a nearly 10-fold higher affinity of the cis isomer, as reported for similar members of this class of compounds [66]. Consistent with the reported binding affinity for 5, both diastereoisomers are highly selective towards CB1R (K i > 1 µM) [66]. In view of the remarkable affinity and selectivity of cis-5 (LU14), the radiofluorination, and subsequently, the pre-clinical characterisation of [ 18 F]LU14 was carried out.

Radiochemistry
The mesylate precursor for radiosynthesis, 7 was prepared with a similar synthetic route as described for the fluorinated compounds, depicted in Scheme 1. Briefly, the carboxamide 4 was coupled with 1-bromo-4-butanol followed by the separation of the diastereomers and the mesylation of the hydroxyl group. For the optimisation of the reaction and the efficient formation of [ 18 F]LU14, the reaction conditions were varied concerning the amount of precursor 7, solvent, base and temperature. Choosing a less basic environment by applying TBAHCO 3 in MeCN at 100 • C resulted in non-isolated radiolabelling yields lower than 10%. Use of K2.2.2./K 2 CO 3 in DMSO at 150 • C with a very low amount of 7 (1-2 mg) resulted in a two-fold increase of the non-isolated radiolabeling yields.
Based on these preliminary results, an automated radiosynthesis for [ 18

In Vitro Evaluation of [ 18 F]LU14
As a first step to assess the ability of [ 18 F]LU14 to label the CB2R, the equilibrium dissociation constant K D (hCB2R) was determined by homologues competition using hCB2R-CHO cell membrane homogenates ( Figure 4). A value of 2.9 nM was experimentally determined. Next, the in vitro evaluation by autoradiography on spleen tissue was performed (Supplementary Figure S1), as the spleen is regarded as a suitable reference organ, having high densities of CB2R (rat spleen CB2R B max = 0.71 pmol/mg) [70,71]. Cryosections of spleen tissue (10 µm; SPRD rats, female; n = 3-4) were incubated with [ 18 F]LU14. However, as reflected by the images in Supplementary Figure S1, the spleen autoradiography revealed no or low specific binding of [ 18 F]LU14. The inability to visualise CB2 receptors in rodent spleen tissue was previously reported also for other highly affine CB2R agonists [72][73][74], e.g., [ 18    Human CB2 receptors were expressed in CHO cells, and affinities were determined by competitive radioligand binding experiment using [ 18 F]LU14 at different molar activities. The inhibition curve was obtained by non-linear regression of percent specific binding (mean ±SD; one experiment run in triplicate) plotted versus the log concentration of the test compound.

In Vivo Metabolism
The metabolism of [ 18 F]LU14 in vivo was investigated by the use of blood plasma, brain and spleen tissue obtained from female CD-1 mice at 30 min after radiotracer injection.
As shown in Figure 5, a rather strong metabolic conversion of [ 18 F]LU14 into more polar metabolites was observed. In plasma, only 17 ± 2% (n = 3) of unchanged [ 18 F]LU14 was measured. The radioactive signal detected between 2 and 4 min represented about 10% of total activity. It might be caused by fluoride salts formed by the defluorination of the tracer or similar [ 18 F]N-alkyl chain oxidative degradations [75]. A similar analytical radio-HPLC chromatogram was also observed when investigating the spleen tissue, but with a higher amount of intact [ 18 F]LU14 (50 ± 10% of the total activity detected), which might be a hint of a CB2R-specific retention of the tracer in this organ. In the brain, 82 ± 5% intact [ 18 F]LU14 was detected ( Figure 5).

PET Imaging
As the CB2R are not expressed in a healthy brain at a sufficient level to be visualised by PET, the ability of [ 18 F]LU14 to image CB2R expression in the spleen was investigated. Dynamic small animal PET/MR scans with [ 18 F]LU14 under control and blocking conditions were performed in healthy male Wistar rats. The time-activity curves (TACs, Supplementary Figure S2) revealed a rather low uptake in the spleen, comparable to muscle tissue used as reference (SUV ratio spleen-to-muscle of 1.1 to 1.4 between 20 and 60 min p.i.). This is in agreement with the results from the autoradiography study described above. In order to evaluate the ability of [ 18 F]LU14 to penetrate the blood-brain barrier and potentially image the human CB2R (hCB2R) in pathological conditions, PET imaging experiments were performed with [ 18 F]LU14 in a rat model with local overexpression of hCB2 receptors in the right striatum, as described by Vandeputte et al. [67] and validated as a suitable model for evaluation of both CB2R agonists and antagonists by Attili [68]. As shown in Figure 6 and Table 1, analysis of the TACs revealed a significantly higher uptake into the striatal region overexpressing the hCB2R(D80N), compared to the contralateral region and the cerebellum after i.v. injection of [ 18 F]LU14. The initial uptake was followed by a washout with a significantly longer mean residence time (MRT) in the target region compared to non-target regions. Based on these findings, we conclude that (i) these experiments confirm the ability of [ 18 F]LU14 to penetrate the BBB, (ii) that the model clearly reveals a specific uptake of [ 18 F]LU14 under pseudopathological conditions of elevated hCB2R expression and (iii) that, due to the negligible uptake of radioactivity in the cerebellum after i.v. of [ 18 F]LU14 during the time span of the analysis, this region could be used as a reliable reference region in our analysis (Table 2). Taking the aforementioned assumptions into account, the SUVr (hCB2R-to-contralateral ratio) reached a plateau (6.6 ± 1.7 and 7.0 ± 1.7) at 45 min p.i. and the SUVr (hCB2R-to-cerebellum ratio) an earlier plateau (5.8 ± 1.1 to 6.5 ± 1.4) at 30 min p.i. Thus, both reference regions demonstrated favourable kinetics of this tracer to image cerebral CB2R. The uptake of [ 18 F]LU14 in the brain region with CB2R overexpression was about five-fold higher compared to the previously reported radioligand [ 18 F]MA3 [68] and about two-fold higher compared to [ 11 C]NE40 in the same rat model [68]. In order to assess whether the binding of [ 18 F]LU14 to the CB2R is specific and reversible, displacement experiments were performed by the intravenous administration of the known CB2R agonist GW405833 at 20 min p.i ( Figure 6B). The amount of activity in the brain region with CB2R overexpression was considerably displaceable by the competitor, as shown by the significant reduction of the AUC 20-60min by 73% (compared to the contralateral region) and 76% (compared to the cerebellum) ( Table 1). In summary, these experiments demonstrate that [ 18 F]LU14 is a specific CB2R radioligand with reversible receptor binding, favourable kinetics and high potential to be used for the detection of cerebral CB2R overexpression in pathological conditions like neuroinflammation or brain cancer.

General Information
All chemicals and reagents were purchased from commercially available sources and used without further purification. Moisture-sensitive reactions were conducted under dry argon with oven-dried glassware and anhydrous solvents. Reaction progress was monitored by thin-layer chromatography (TLC) using Alugram SIL G/UV 254 pre-coated plates (Macherey-Nagel; Düren, Germany). The spots were identified by using a UV lamp or by dipping the plates into a potassium permanganate solution (3 g KMnO 4 , 20 g K 2 CO 3 , 0.25 mL glacial acid, 300 mL water). For the purification of products, flash column chromatography was used with silica gel 40-63 µm (VWR International Chemicals, Darmstadt, Germany). The purity of all the tested compounds was ≥95% as determined by an LC-MS system including a DAD detector (Dionex Ultimate 3000, Thermo Fisher Scientific Inc., Waltham, MA, USA) system incorporating a LPG-3400SD pump, a WPS-3000 TSL autosampler, a TCC-3000SD column compartment, a DAD 3000 diode array detector and an MSQ 3000 low-resolution mass spectrometer (Thermo Fisher Scientific  Inc.

Automated Radiosynthesis of [ 18 F]LU14
Remote-controlled radiosynthesis was performed using a Synchrom R&D EVO III automated synthesiser (Elysia-Raytest, Straubenhardt, Germany). Briefly, [ 18 F]fluoride (4-6 GBq) was trapped on a Waters QMA cartridge and eluted with a solution containing 5.6 mg K222 and 40 µL K 2 CO 3 dissolved in a mixture of H 2 O/MeCN (1 mL, 1:4, v/v) into the reaction vessel and dried via azeotropic distillation. An additional 1.5 mL of dried MeCN was added. After complete dryness, a solution containing 1 mg of precursor 7 in 800 µL DMSO was added, and the reaction mixture was stirred at 150 • C for 10 min. Upon cooling to 40 • C, the reaction mixture was diluted with 4 mL H 2 O, and the resulting solution was transferred to the semi-preparative HPLC. [ 18 F]LU14 was collected in the HPLC collection vial containing 40 mL of H 2 O and trapped in the Sep-Pak ® C18 light cartridge (Waters GmbH, Eschborn, Germany). The cartridge was washed with 2 mL H 2 O, and [ 18 F]LU14 was eluted with 1.2 mL EtOH. This ethanolic solution was transferred outside of the shielded cell, and the solvent was evaporated at 70 • C in a gentle stream of nitrogen for 5-10 min, and [ 18 F]LU14 was reconstituted in isotonic saline solution for further biological characterisation. Total synthesis time was about 70 min.
Analytical chromatographic separations were performed on a JASCO LC-2000 system, incorporating a PU-2080Plus pump, AS-2055Plus auto-injector (100 µL sample loop), and a UV-2070Plus detector coupled with a γ-detector (GABI Star, Raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany). Data analysis was performed with the Galaxie chromatography software (Agilent Technologies, Santa Clara, CA, USA) using the chromatograms obtained at 254 nm.

Biological Experiments
All studies involving animals were carried out according to the national laws on the protection of animals and were approved by the responsible authorities (Landesdirektion Sachsen, No. DD24.1-5131/446/19; TVV 18/18).

Determination of Binding Affinities by Homogenate Assays
The binding affinities towards CB 1 and CB 2 receptors were determined according to a protocol published previously [54]. In brief, membrane preparations obtained from CHO cell lines stably transfected with either human CB 1 (hCB 1 -CHO; obtained from Euroscreen, Gosselies, Belgium) or human CB 2 (hCB 2 -CHO; obtained from Paul L. Prather, Department of Pharmacology and Toxicology, College of Medicine, University of Arkansas for Medical Sciences, Fayetteville, AR, USA) were used along with

Quantification of Radiometabolites
About 30 MBq [ 18 F]LU14 dissolved in about 150 µL isotonic saline was administered intravenously as a bolus in the tail vein of an awake female CD-1 mice weighing about 33 g (n = 3). At 30 min p.i., the animals were anaesthetised and blood was withdrawn by retrobulbar bleeding using glass capillaries. Immediately afterwards, the animals were euthanised by cervical dislocation, and released urine was sampled. Blood plasma was obtained from the whole blood sample by centrifugation (2 min, 8000 rpm, room temperature). In addition, the brain and spleen were isolated and homogenised in 1 mL demineralised water on ice (1000 rpm, 10 strokes; glass vessel, PTFE plunger; Potter S, B. Braun Biotech International, Göttingen, Germany).
The samples were further processed for subsequent radio-chromatographic analyses. Two consecutive extractions were performed as duplicates for plasma and brain determinations. Plasma, brain and spleen samples were added to an ice-cold MeOH/H 2 O mixture (9:1, v/v). The samples were vortexed for 3 min, incubated on ice for 5 min and centrifuged at 10,000 rpm for 5 min. Supernatants were collected, and the precipitates were re-dissolved in 100 µL of extraction solvent, and the extraction procedure was repeated. The activities of supernatants and precipitates were measured in a γ-counter (1480 WIZARD, Fa. Perkin Elmer, Turku, Finland), and the extraction efficiencies were calculated as the ratio of radioactivity in the supernatant to the radioactivity in the original sample (supernatant + precipitate). The supernatants from both extractions were combined, concentrated at 70 • C under argon stream up to a remaining volume of 100 µL, and subsequently analysed by analytical radio-HPLC with a gradient system (see quality control section).

PET Studies
In vivo biodistribution of [ 18 F]LU14 in rats was assessed by dynamic small animal PET (Nanoscan, Mediso, Budapest, Hungary) 60 min recordings, followed by T1-weighted (GRE, TR/TE = 15.0/2.4 ms, 252/252, FA = 25 • ) MR imaging with whole-body coils for anatomical correlation and attenuation correction. For the uptake studies into the spleen three male Wistar Hun rats (Janvier Laboratories, Le Genest Saint Isle, France) weighing 268 ± 56 g were used. For the uptake studies into the brain, five female Wistar rats weighing 263 ± 13 g carrying the stereotactically injected AAV 2/7-CaMKII0.4-intron-hCB2R D80N (right striatum) and AAV2/7-CaMKII0.4-intron-3flag-eGFP (contralateral, left striatum). Animals were initially anaesthetised with 5% isoflurane and placed on a thermostatically heated animal bed where anaesthesia was maintained with 2% isoflurane in 60% oxygen/38% room air. They were pre-treated by i.v. injections of vehicle solution only, containing DMSO: Kolliphor EL: saline in a composition of 1:2:7 (control group) or GW405833 (1.5 mg/kg bodyweight for spleen-scan experiments and 5 mg/kg for the displacement experiments with local overexpression of hCB2R) 10 min prior [ 18 F]LU14 administration. [ 18 F]LU14 (22.4 ± 3.5 MBq) was injected i.v. into the lateral tail vein (bolus within 5 s) at the start of the PET acquisition. List-mode PET data were binned as a series of attenuation-corrected sinogram frames (12 × 10 s, 6 × 30 s, 5 × 60 s and 10 × 300 s) and were reconstructed by ordered subset expectation maximisation (OSEM3D) with four iterations, six subsets and a voxel size of 0.4 mm 3 (Nucline v2.01, Mediso, Hungary). The analysis of reconstructed datasets was performed with PMOD software (v4.103, PMOD Technologies LLC, Zurich, Switzerland). Non-parametrical analyses of achieved timeactivity curves (TACs) were performed with Microsoft Excel to determine the time to peak, the TAC peak value and the area-under-the-zero-momentum curve (AUC): where c (radioactivity) is expressed as a standardised uptake value normalised to the bodyweight in g (SUV), the area-under-first-momentum curve (AUMC): and the mean residence time (MRT): GraphPad Prism (v9, San Diego, CA, USA) was used for graphical presentation.

Conclusions
To evaluate the suitability of the naphtyridin-2-one-3-carboxamide class of compounds to target CB2R in the brain, we have chosen the literature-known compound cis-/trans-5 and developed [ 18 F]LU14 in stereochemically pure form. The first biological evaluation of [ 18 F]LU14 revealed low nanomolar binding affinity towards human CB2R and adequate metabolic stability in mouse with only trace amounts of radiometabolites detected in the brain at 30 min after radioligand injection. PET experiments in a rat model of human CB2R overexpression revealed that [ 18 F]LU14 is able to reach and reversibly label cerebral CB2R, demonstrating its potential as a radiotracer for detecting CB2R alterations in pathologic conditions.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request.