Radiofluorination and biological evaluation of N-aryl-oxadiazolyl-propionamides as potential radioligands for PET imaging of cannabinoid CB2 receptors

Background The level of expression of cannabinoid receptor type 2 (CB2R) in healthy and diseased brain has not been fully elucidated. Therefore, there is a growing interest to assess the regional expression of CB2R in the brain. Positron emission tomography (PET) is an imaging technique, which allows quantitative monitoring of very low amounts of radiolabelled compounds in living organisms at high temporal and spatial resolution and, thus, has been widely used as a diagnostic tool in nuclear medicine. Here, we report on the radiofluorination of N-aryl-oxadiazolyl-propionamides at two different positions in the lead structure and on the biological evaluation of the potential of the two tracers [18F]1 and [18F]2 as CB2 receptor PET imaging agents. Results High binding affinity and specificity towards CB2 receptors of the lead structure remained unaffected by the structural changes such as the insertion of the aliphatic and aromatic fluorine in the selected labelling sites of 1 and 2. Aliphatic and aromatic radiofluorinations were optimized, and [18F]1 and [18F]2 were achieved in radiochemical yields of ≥30% with radiochemical purities of ≥98% and specific activities of 250 to 450 GBq/μmol. Organ distribution studies in female CD1 mice revealed that both radiotracers cross the blood–brain barrier (BBB) but undergo strong peripheral metabolism. At 30 min after injection, unmetabolized [18F]1 and [18F]2 accounted for 60% and 2% as well as 68% and 88% of the total activity in the plasma and brain, respectively. The main radiometabolite of [18F]2 could be identified as the free acid [18F]10, which has no affinity towards the CB1 and CB2 receptors but can cross the BBB. Conclusions N-aryl-oxadiazolyl-propionamides can successfully be radiolabelled with 18F at different positions. Fluorine substitution at these positions did not affect affinity and specificity towards CB2R. Despite a promising in vitro behavior, a rather rapid peripheral metabolism of [18F]1 and [18F]2 in mice and the generation of brain permeable radiometabolites hamper the application of these radiotracers in vivo. However, it is expected that future synthetic modification aiming at a replacement of metabolically susceptible structural elements of [18F]1 and [18F]2 will help to elucidate the potential of this class of compounds for CB2R PET studies.


Background
Cannabinoid receptors (CBR) belong to the superfamily of G protein-coupled receptors (GPCR) and are involved in various physiological processes. Besides the CBR identified so far, the cannabinoid receptor type 1 (CB 1 R) [1] and type 2 (CB 2 R) [2], another GPCR natively interacting with endogenous cannabinoids (GPR55) has been proposed as a third type of CBR [3,4]. While the CB 1 R is primarily expressed in the central nervous system, the CB 2 R is predominantly located in the periphery, especially in tissues related to the immune system. In pathological conditions, the up-regulation of CB 2 R expression is mostly associated with inflammatory processes [5,6], neuropathic pain [7][8][9], Alzheimer's disease, and amyotrophic lateral sclerosis (ALS) [10][11][12][13]. It has also been shown that the activation of CB 2 R is connected with the induction of apoptosis in several cancer cell lines [14][15][16][17]. However, CB 2 R are also expressed in healthy brain under physiological conditions at very low expression levels [18,19]. Therefore, the non-invasive quantification of CB 2 R in the brain, possibly in general by application of highly sensitive imaging techniques such as positron emission tomography (PET), requires the availability of radioligands binding with high affinity and high specificity towards CB 2 R [20]. Apart from the numerous specific pharmacologically relevant ligands of CB 2 R reported so far (see [21,22]), only a limited number have been applied for the development of PET radiotracers for imaging of CB 2 R [23 -29]. As we reported earlier, the increase of brain uptake and metabolic stability and decrease of non-specific binding remained as a challenge for further development [29]. Even a follow-up study on [ 11 C]AZD1940 in nonhuman primate with PET confirmed a relatively low CNS exposure of this radioligand [28]. Mu et al. reported on N-(1-adamantyl)-8-methoxy-4-oxo-1-phenyl-1, 4-dihydroquinoline-3-carboxamide as a 11 C-labelled PET probe for imaging of CB 2 R [30]. Although the study showed a rather low brain uptake, and two blocking experiments with GW4058233 to demonstrate the specificity of brain uptake were not conclusive, the potential of the compound for PET imaging of amyotrophic lateral sclerosis has been proposed.
Further, a whole-body biodistribution and radiation dosimetry study of the CB 2 R ligand [ 11 C]NE40 has been performed in healthy subjects [31]. It showed the expected biodistribution being compatible with lymphoid tissue (spleen) uptake and an appropriate uptake and kinetics in the brain. This underscores the potential of this tracer for application in central and peripheral inflammation imaging. Despite this considerable progress in CB 2 R PET imaging, a suitable ligand [32] radiolabelled with the advantageous longer-lived isotope 18 F is still missing.
Recently, we reported on the synthesis, radiofluorination, and first biological investigation of the N-aryl-oxadiazolyl-propionamide [ 18 F]2 as a potential radioligand for the PET imaging of CB 2 R [29]. Although the initial radiofluorination approach using a nitro precursor proved to be unsatisfactory (yields ≤ 3%), the promising biological findings encouraged us to revise and upgrade the radiosynthesis of [ 18 F]2 and to perform a more detailed biological evaluation.

Methods
We describe an improved synthesis of [ 18 F]2 using a trimethylammonium precursor for the radiolabelling at the aromatic site of the lead structure (Figure 1, compound 13, X 2 = NMe 3 + I). In parallel, we explored the aliphatic radiofluorination at the carbazole N-alkyl chain (Figure 1, compound 15, X 1 = OTs). The labelling of the lead compound at two different sites opens up the possibility to investigate the dependence of affinity, biodistribution and metabolism of these radiotracers on the site of radiolabelling.

Results and discussion
Synthesis of N-arylamide oxadiazoles: precursors for radiochemistry and reference compounds In a previous study, we have described the synthesis of >20 fluorinated N-arylamide oxadiazoles including labelling precursors for radiochemistry [29]. Here we describe a modified and improved route to obtain novel derivatives for high-yield radiochemistry. In brief, treatment of the nitriles 3 and 4 with excess of hydroxylamine hydrochloride under alkaline conditions delivered the (Z)-amidoximes 5 and 6 in 55% and 73% yield together with amides 7 and 8 (Scheme 1), which have been separated by column chromatography on silica. The Z configuration of the amidoximes was determined by two-dimensional nuclear magnetic resonance (NMR) spectroscopy. Treatment of the amidoximes 5 and 6 with excess of succinic anhydride delivered the acids 9 and 10 in a nearly quantitative manner. The N-aryl-oxadiazolyl-propionamides 11 and 2 were obtained in high yields (85% and 78%) by using N, N′-diisopropylcarbodiimide (DIC) as coupling reagent [33]. For the aromatic radiofluorination, trimethylammonium iodide has been used as a leaving group. Although methyl iodide (MeI) has a very high nucleophilicity, the choice of I − as a counter ion speeds up the displacement of the trimethylammonium salt in an aromatic ring during radiofluorination under mild conditions [34], thus lowering the amount of potential formation of non-radioactive and radioactive by-products. The trimethylammonium salt 13 was synthesized from the nitro derivative 11 by reduction, reductive methylation with paraformaldehyde and NaBH 4 [35], and quaternization by employing large excess of MeI (Scheme 2) [36].
The ethanol derivative 14 was synthesized in four steps and 28% overall yield starting from the commercially available carbazole (23). The synthesis started with deprotonation of carbazole with n-BuLi, reaction of the carbazolyl anion with ethylene sulfate, and subsequent hydrolysis with dilute H 2 SO 4 to afford the hydroxyethyl derivative 25 (see 'Experimental' section). The crystalline cyclic ethylene sulfate represents a non-toxic but reactive alternative to the gaseous and toxic oxirane. However, there are only few examples for the introduction of a 2-hydroxyethyl group using ethylene sulfate. Nitration of the hydroxyethylcarbazole 25 was performed with concentrated nitric acid at 5°C to 10°C to give the 3-nitro carbazole derivative [37] 24 in 54% yield (Scheme 1). Reduction of the nitrocarbazole 24 with H 2 in the presence of Pd/C provided the primary aromatic amine 26, which was precipitated as HCl salt. The final coupling of the carbazolamine 26 with the propionic acid 10 was induced by COMU® providing the amide 14 (Scheme 1).
The fluoroethyl reference compound 1 was prepared by treatment of 14 with the fluorinating agent diethylaminodifluorosulfonium tetrafluoroborate (XtalFluor-E®) (Scheme 2). In order to obtain the precursor for the radiosynthesis, the alcohol 14 was transformed into the tosylate 15 since the tosyloxy moiety represents a good leaving group for the nucleophilic substitution with [  For investigations regarding the main metabolite [ 18 F]10, the trimethylammonium salt 18 and reference compound 17 were synthesized. Intermediate 16 was obtained by acid protection followed by reduction from 9. To synthesize the reference compound 17, Sandmeyer reaction has been applied [38]. The use of tetrafluoroborate as fluorinating reagent led to the formation of a complex reaction mixture whereas the use of KF delivered 17 in 40% yields. Alternatively, 17 could also be obtained from 10 (not shown in Scheme 3, see 'Experimental' section) in improved yield. The trimethylammonium salt 18 was obtained as described above in 48% over two steps.

Radiochemistry
Reaction conditions, purification, and formulation procedures were optimized to achieve a high radiochemical yield (RCY) and a high radiochemical purity in a short synthesis time. Aliphatic and aromatic radiolabellings were carried out at 82°C in MeCN under no-carrieradded (NCA) conditions (Scheme 4). High labelling yields ( Figure 2) were obtained using 2 mg of precursor. For [ 18 F]1, the optimal conversion of the tosylate 15 was achieved after 10 min (53 ± 6%, n = 12). For [ 18 F]2, a maximum was reached after 10 min (63 ± 5%, n = 7), along with a steady loss of product (51 ± 6%, 15 min) Scheme  thereafter, probably due to the decomposition of the trimethylammonium salts under prolonged heating [39]. Altogether, evaluating the current results of labelling of [ 18 F]2, a great improvement was obtained by using the herein described new trimethylammonium precursor 13 instead of the formerly applied nitro compound 11 (see Scheme 4) where a radiochemical yield of only 3% has been achieved [29]. Reaction mixtures were separated by isocratic semipreparative HPLC, fractions were collected, and the identities confirmed by spiked analytical HPLC samples with the respective reference compounds of [ 18 F]1 and [ 18 F]2 ( Figure 3). Final purification was performed by solid phase extraction (SPE) using Sep-Pack cartridges, and for biological investigations, the products were formulated in isotonic saline containing 10% EtOH. The radiotracers were produced with 30% to 35% RCY, high radiochemical purities (≥98%), and high specific activities (250 to 450 GBq/μmol, n = 4).

In vitro stability and logD determination
Investigations on the in vitro stability of [ 18 F]1 and [ 18 F] 2 were performed prior to the experimental determination of logD values. [ 18 F]1 exhibited a slight tendency to de-fluorinate at 40°C in Tris-HCl, EtOH, 0.9% NaCl, and Dulbecco's phosphate buffer with 89%, 93%, 95%, and 94% of intact tracer after 30 min of incubation and no further decrease up to 90 min. [ 18 F]2 showed a very high stability with ≥98% of intact tracer at incubation in all selected media up to 90 min. Radio-thin-layer chromatography (TLC) and radio-HPLC analyses were in good agreement, and no further radioactive degradation products were observed.
Regarding the HPLC-based logD determination, differences were noticed between isocratic and gradient elutions using the same column. However, as compiled in Table 1, the logD values using Reprosil-Pur C18 AQ column under gradient conditions are in good agreement with those found with Prodigy 5 μm C8 column.

Receptor affinity
The substitution of the ethyl chain of the carbazole moiety in 2 with a fluorine-ethyl chain in 1 had little effect on target affinity and specificity as reflected by comparable K i values for the CB 2 R (1, K i 2.32 ± 2.12 nM; 2, K i 4.27 ± 3.03 nM) [29]. Based on the data published by Cheng et al. [40] (JCPDS [42][43][44], an agonistic profile of 1 and 2 can be assumed. Both compounds possess no affinity towards the CB 1 R (K i > 1 μM). By contrast, one major brain-penetrating metabolite, compound 10, shows almost no binding towards the CBRs (K i CB1R > 1 μM, K i CB2R > 1 μM). In accordance with the ex vivo autoradiographic studies on [ 18 F]2 in spleen tissue slices [29], compound [ 18 F]1 targets CB 2 R in vivo too. The autoradiograms are shown in Figure S1 of Additional file 1.

Organ distribution of [ 18 F]1 and [ 18 F]2 in mice
The biodistribution of compounds [ 18 F]1 and [ 18 F]2 in percentage of injected dose per gram (% ID/g) at 5, 30,   Tables 2  and 3. For both radiotracers, the highest uptake was detected at 5 min p.i. in the spleen. Comparable kinetics were observed in the brain and most other organs such as the lung, liver, kidney, thymus, and adrenals after administration of [ 18 F]1 and [ 18 F]2, respectively. Both compounds showed a high gastrointestinal excretion as reflected by the constant increase of activity uptake in the small intestine. A constantly low uptake of activity in the femur indicates no confounding defluorination during the experiment.
The highest accumulation of activity was found in the kidneys and liver for both compounds. While the activity uptake is higher in the kidneys for [ 18  Blocking experiments were performed with preadministration of the CB 2 R-specific inverse agonist SR144528. At 60 min, the animals were sacrificed and the percentage of injected dose per gram of activity uptake was calculated for the various organs. No significant reduction of percentage of injected dose per gram could be observed for both compounds in any of  the organs investigated including CB 2 R-expressing organs (details are compiled in Additional file 1: Table S1). For [ 18 F]2, the pre-administration of SR144528 led to an increase of percentage of injected dose per gram in some organs, which might be related to metabolic processes. Interestingly, major differences between the radioactivity uptakes in various organs were found for [ 18 Figure 4 shows all organs (plus the blood and liver) in which significant differences in radioactivity uptake were found at 60 min after injection of the two radiotracers. The measured radioactivity was higher after application of [ 18 F]1 in those organs expressing CB 2 R natively [18] in comparison to the data obtained after injection of [ 18 F]2. This relation is reversed in those organs associated to excretion. The cause of this is probably related to the much faster metabolism of [ 18 F]2 and its radiometabolites compared to that of [ 18 F]1, which is in agreement with the observation of significantly lower accumulation of radiolabelled compounds in the plasma and kidney. The significant differences in the pancreas and femur can be considered as a further indicator of different metabolic processes of [ 18 F]1 and [ 18 F]2 leading to different radiometabolites with diverse organ distribution patterns. Thus, the activity measured in the ex vivo biodistribution study is largely determined by metabolites bearing the 18 F label.

Metabolism of [ 18 F]1 and [ 18 F]2 in mice
Typical HPLC radiochromatograms of brain extracts are shown in Figure 5. In general, radio-TLC analytics of plasma, urine, and brain samples obtained at 30 and 60 min p.i. of [ 18 F]1 or [ 18 F]2 are consistent with the data obtained by radio-HPLC. As shown in Table 4, unmetabolized [ 18 F]1 accounted for 60% of the recovered activity in plasma samples at 30 min p.i., while 36% of the total radioactivity can be addressed to a more hydrophilic metabolite M1 a . The amount of intact radiotracer [ 18 F]1 is further reduced at 60 min p.i. to 7% with a concomitant increase of M1 a to 92%.
Radiochromatograms of brain samples at 30 min p.i. shown in Figure 5 revealed that unmetabolized radiotracer corresponded to 68% and 88% of the total activity after administration of [ 18 F]1 and [ 18 F]2, respectively. Paralleled by the constantly increasing accumulation of radiometabolites in the brain, these values decreased to 35% and 43%, respectively, at 60 min p.i. In particular, these radiometabolites account for 13% (M1 a ) and 10% (M2 a ) as well as 8% (M1 b ) and 3% (M2 b ) of total activity at 30 min p.i. of [ 18 F]1 and [ 18 F]2, respectively, with the main radiometabolites M1 a and M1 b reaching values of 58% and 51% at 60 min p.i., respectively. The radiometabolite M2 a of [ 18 F]1 corresponds most likely to the amine, which is formed by enzymatic hydrolysis of the amide bond in the compound. Blocking was performed with pre-administration of SR144528 (3 mg/kg i.p.). Values are mean ± SD. a n = 4, b n = 5, c n = 6.
The total amount of activity measured in, e.g. the plasma at particular times p.i. also depends on the kinetics of the excretion of radiotracer/radiometabolites. For [ 18 F]2, this is more pronounced than for [ 18 F]1 with unmetabolized [ 18 F]2 accounting for only 2% of the total radioactivity in the plasma at 30 min p.i. At this time, more than 90% of the recovered activity in plasma samples for [ 18 F]2 consisted of the single radiometabolite M1 b . As this value decreases only slightly up to 60 min p.i. (88%), M1 b can be assumed as a rather metabolically stable compound. This radiometabolite of [ 18 F]2 probably corresponds to the N-aryl-oxadiazolylpropanoic acid resulting from enzymatic hydrolysis of the amide bond of 2 [40]. To prove this assumption, the trimethylammonium salt 18 was synthesized as the precursor of this metabolite. In order introduce 18 F into the molecule, the carbon acid moiety was protected as a methyl ester resulting in the labelled compound [ 18 F]17 (Scheme 5). It was expected that after injection of this compound in mice, the ester will be converted into the free acid [ 18 F]10. Those metabolic pathways are common for drug inactivation and excretion [41] as well as for the prodrug concept providing pharmacologically active metabolites in vivo from pharmacologically inactive compounds [42]. After intravenous injection of [ 18 F]17 in CD1 mice, samples of the plasma, brain, and spleen were analyzed at 30 and 60 min p.i. ( Table 5).
that the radiometabolite M1 b co-elutes with compound 10 ( Figure 6). Furthermore, in combination with the time-activity data presented in Table 5, this result indicates that the main radiometabolite M1 b of [ 18 F]2 derived from peripheral metabolism ([ 18 F]10) penetrates the blood-brain barrier (BBB). In addition, analyses of brain and spleen sample data revealed that no further metabolic transformation of [ 18 F]10 could be observed up to 60 min p.i. as shown in Figure 7.
In vitro investigation of the affinity of 10 towards both hCB 2 R and hCB 1 R revealed no specific binding (K i > 1 μM). Altogether, these data suggest that non-target binding radiometabolites of [ 18 F]1 and [ 18 F]2 account for the vast majority of ex vivo activity measured in the spleen and other organs known to express CB 2 R. This corresponds very well with the insignificant displacement of activity after administration of [ 18 F]1 and [ 18 F]2 by the CB 2 R-specific SR144528. Hence, radiometabolites generated in the periphery penetrate the BBB leading to a pronounced accumulation of the main radiometabolites in the brain, hampering the application of [ 18 F]1 and [ 18 F]2 for imaging of CB 2 R in the brain.

In vitro CB receptor affinity assay
For binding experiments, Chinese hamster ovary (CHO) cell lines stably transfected with human CB 1 R human and CB 2 R were used according to the procedures previously described [29]. Briefly, displacement of CB 1 R/ CB 2 R-specific radioligand [ 3 H]CP55,940 (6,438 GBq/ mmol; PerkinElmer Life and Analytical Sciences GmbH, Rodgau, Germany; working concentration, 0.2 to 0.5 nM) by test compounds in the range of 0.1 nM to 10 μ was assessed, and IC 50 values were estimated by non-linear regression (GraphPad Prism; version 3.0, GraphPad Software, Inc., San Diego, CA, USA). K D values of 2.4 and 1.5 nM were previously determined for [ 3 H]CP55,940 binding to hCB 1 R and hCB 2 R, respectively, and used for calculation of K i values of the test compounds according to Cheng et al. [45]. The binding experiments were performed in triplicates, and data were given as mean values from independent experiments.

Ex vivo biodistribution studies
Animals for in vivo studies were obtained from the Medizinisch-Experimentelles Zentrum, Universität Leipzig. All procedures that include animals were approved by the respective State Animal Care and Use Committee and conducted in accordance with the German Law for the Protection of Animal.
Female CD1 mice (10 to 12 weeks old, 20 to 25 g) received an injection of 300 to 400 kBq of [ 18 F]1 or [ 18 F]2 with specific activities of >450 GBq/μmol in 200 μL of 0.9% NaCl/10% EtOH into the tail vein. The animals were anesthetized (CO 2 /O 2 mixture) for blood and urine sampling and euthanized by luxation of the cervical spine at 5, 30, and 60 min after injection (p.i.) (n = 2 to 5 per time). The organs of interest were removed and weighed, and the activities were measured by γ counting using a calibrated γ counter Wallac Wizard 1470 (Perkin Elmer Inc., Waltham, MA, USA). The percentage of injected dose per gram of wet tissue (% ID/g wet weight) was calculated.
To verify the specificity of [ 18 F]1 and [ 18 F]2 towards CB 2 R, blocking experiments were performed with preadministration of the highly selective CB 2 R inverse agonist SR144528 [46] (3 mg/kg i.p. in 0.9% saline, 10 min before the injection of the radiotracer) at 60 min p.i. (n = 2 to 5). The unpaired two-tailed t test was used to compare the results between the groups. We considered differences to be significant at a p value <0.05.
Ex vivo metabolite studies [ 18 F]1 and [ 18 F]2 (100 to 150 MBq, 250 to 450 GBq/μmol in 150 μL NaCl 0.9%/10% EtOH) were injected via the tail vein in CD1 male mice (10 to 12 weeks old, 20 to 25 g). Blood and urine samples were obtained at 30 and 60 min p.i. (n = 3 per time point). Twofold extractions of plasma and brain samples (n = 3) were performed using ice-cold MeCN according to the standard protocol established in our group (see [47,48]). Briefly, plasma samples were obtained by centrifugation of the blood at 4,000×g at 4°C for 10 min, and brain samples were homogenized in ice-cold 50 mM Tris-HCl (pH = 7.4). The samples were vortexed, incubated on ice, and centrifuged at 10,000×g for 3 min. Supernatants were collected, and the precipitates were re-dissolved in ice-cold MeCN for the second extraction. The supernatants from the two extractions were combined, concentrated under a gentle argon stream at 65°C, and analyzed by radio-TLC and gradient analytical HPLC (see 'Radiochemistry' section). Aliquots from each extraction supernatant and the precipitates were also taken and quantified by γ counting (Wallac Wizard 1470, Perkin Elmer Inc., Waltham, MA, USA) along with the respective aliquots of intact plasma samples and brain homogenates. A moderate recovery of radioactivity was obtained from plasma samples and brain homogenates (60% to 70%). Radiometabolites of [ 18 F]17 in the plasma, spleen, and brain were assessed at 30 and 60 min p.i.

Conclusions
In conclusion, N-aryl-oxadiazolyl-propionamides were successfully radiolabelled with 18 F at different positions. Fluorine substitution at these positions did not affect affinity and specificity towards CB 2 R. However, the radiotracers investigated in this study undergo a fast metabolism in vivo with the main radiometabolites crossing the blood-brain barrier. Therefore, structural changes in the enzymatic cleavage sites of the evaluated candidates have to be performed to enhance their potential as CB 2 R PET imaging agents for the brain.