Cannabinoid receptor type 2 (CB2)-selective N-aryl-oxadiazolyl-propionamides: synthesis, radiolabelling, molecular modelling and biological evaluation

Background The endocannabinoid system is involved in many physiological and pathological processes. Two receptors (cannabinoid receptor type 1 (CB1) and type 2 (CB2)) are known so far. Many unwanted psychotic side effects of inhibitors of this system can be addressed to the interaction with CB1. While CB1 is one of the most abundant neuroreceptors, CB2 is expressed in the brain only at very low levels. Thus, highly potent and selective compounds for CB2 are desired. N-aryl-((hetero)aromatic)-oxadiazolyl-propionamides represent a promising class of such selective ligands for the human CB2. Here, a library of various derivatives is studied for suitable routes for labelling with 18F. Such 18F-labelled compounds can then be employed as CB2-selective radiotracers for molecular imaging studies employing positron emission tomography (PET). Results By varying the N-arylamide substructure, we explored the binding pocket of the human CB2 receptor and identified 9-ethyl-9H-carbazole amide as the group with optimal size. Radioligand replacement experiments revealed that the modification of the (hetero)aromatic moiety in 3-position of the 1,2,4-oxadiazoles shows only moderate impact on affinity to CB2 but high impact on selectivity towards CB2 with respect to CB1. Further, we could show by autoradiography studies that the most promising compounds bind selectively on CB2 receptors in mouse spleen tissue. Molecular docking studies based on a novel three-dimensional structural model of the human CB2 receptor in its activated form indicate that the compounds bind with the N-arylamide substructure in the binding pocket. 18F labelling at the (hetero)aromatic moiety at the opposite site of the compounds via radiochemistry was carried out. Conclusions The synthesized CB2-selective compounds have high affinity towards CB2 and good selectivity against CB1. The introduction of labelling groups at the (hetero)aromatic moiety shows only moderate impact on CB2 affinity, indicating the introduction of potential labelling groups at this position as a promising approach to develop CB2-selective ligands suitable for molecular imaging with PET. The high affinity for human CB2 and selectivity against human CB1 of the herein presented compounds renders them as suitable candidates for molecular imaging studies.


Background
The role of the endocannabinoid system in specific CNS disorders is related to the regulation of the temporal dynamics of neurotransmitter release by the retrograde cannabinoid signalling network [1]. Mediated by the G protein-coupled cannabinoid receptors [2], the release of endogenous or the administration of exogenous ligands affects both the long-term synaptic plasticity as well as the short-term regulation of synaptic transmission [1,3]. Two types of specific cannabinoid receptors have been cloned so far, termed cannabinoid receptor type 1 (CB1) and type 2 (CB2) [2,4]. The existence of additional cannabinoid-binding receptors has been suggested [5,6]. In contrast to classical neurotransmitters, endocannabinoids function as retrograde synaptic messengers which are released from postsynaptic neurons, diffuse across synapses, activate CB1 on presynaptic axons and eventually suppress neurotransmitter release [7]. In addition, endocannabinoids and their receptors control the decision about survival or death of neuronal cells [8], such that the pharmacological manipulation of this system might provide either neuroprotective or pro-apoptotic effects. A therapeutic role of cannabinoids has also been suggested for mood disorders [9], traumatic brain injury [10,11] and tumour treatment [12,13]. The development of CB2-selective anticancer agents is regarded to be advantageous in light of unwanted central effects exerted by binding of those agents to CB1 [14].
CB1 is abundantly expressed in the central nervous system and has been thoroughly investigated on cellular [15] and functional levels [16,17] in the brain. In contrast to this, only few and, to some extent, ambiguous data are available regarding CB2 expression and function in the brain. Initial reports stated that CB2 is mainly expressed peripherally in the cells and organs of the immune system such as the B lymphocyte-enriched marginal zone of the spleen or the cortex of lymph nodes [4,18]. However, more recent studies have proven structural and functional CB2 expression in primed glial (as reviewed in [5]) or neural progenitor cells [19]. Moreover, CB2 mRNA was detected in mouse [20] and rat cerebellum [21]. It could be shown in immunostaining studies that CB2 protein is abundant under basal conditions in neuronal and glial processes in the cerebellum and hippocampus of mouse and rat [22][23][24]. As reviewed recently by Atwood and Mackie [25], a growing number of reports substantiate a neuronal expression of CB2 through immunohistochemical data. Thus, it has been described that the activation of microglia is paralleled by an increase in CB2 immunoreactivity in the injured brain [26]. For instance, significantly increased CB2 expression levels were observed in severe Alzheimer's disease [27]. This indicates that CB2 expressed in the cerebellum might be involved in the pathogenesis of various neurodegenerative disorders and therefore might represent an interesting target for the diagnostics and therapy of such diseases [14,28]. Consequently, detailed investigation of the functional changes of both receptor subtypes CB1 and CB2 is needed to achieve a better understanding of the endocannabinoid system and the effects of potential cannabinoid therapeutics in the normal and diseased human brain [29].
Despite numerous in vitro approaches to directly identify CB2 in the brain, a non-invasive and quantitative analysis of these receptors in vivo has not been reported to date, probably due to the lack of ligands applicable for molecular imaging approaches such as positron emission tomography (PET). Recently, in a microPET study in a rat model, Evens et al. [30] observed specific binding of the type 2 cannabinoid receptor PET tracer [ 11 C]NE40. In this model, hCB2 receptors were locally overexpressed in the brain after stereotactic injection of an adeno-associated viral vector (AAV2/7) encoding hCB2R with a point mutation (Asp80Asn) in the right striatum, yet the quantitative analysis of these receptors in native brain tissue remains a challenge.
Although CB2 has been already mentioned as a valuable biomarker of, e.g. neuroinflammation [30], the quantitative imaging of this target requires PET radioligands with superior affinity and specificity towards the low-abundance cerebral CB2 [31]. The [ 11 C]-methoxylabelled potent inverse CB2 agonist Sch225336 [32] showed only poor brain uptake in mice under baseline conditions. Various 18 F-fluoroethoxy- [33] and [ 11 C]methoxy-labelled [34] 2-oxoquinoline derivatives were suggested as CB2 PET radioligands too, yet unfavourable brain-to-plasma ratios reflect the need for further improvement(s) regarding CB2 imaging in the healthy brain as recently reviewed [35].

Methods
Here, we extend the synthesis of various CB2-binding [36] N-aryl-((hetero)aromatic-oxadiazolyl-propionamides as potential PET imaging compounds. Guided by molecular docking studies employing a novel comparative model of the human CB2 receptor, the promising site for 18 F labelling was selected. In vitro binding experiments using human CB2-transfected Chinese hamster ovary (CHO) cells were performed. The obtained affinity data are in good agreement with predicted binding strength and could be verified by autoradiographic studies on mouse spleen slices.
As depicted in Scheme 2, the two-step synthesis starts from a properly substituted nitrile and hydroxylamine yielding the hydroxy-amidin derivative. Following the route described by Cheng at al. [36], the products were dried after dilution with methylene chloride, and succinic acid anhydride was added to give the 3-phenyl-subsituted 1,2,4-oxadiazole acids. To support the formation of the oxadiazole acids, 0.1 eq. KF was added except for the compound with R 1 = Br and R 2 = F. In this case, the free 3-phenyl-substituted 1,2,4-oxadiazol-5-yl propanoic acid was protected, employing (diazomethyl)trimethylsilane. Subsequently, Br was substituted with CN by addition of 4.2 eq. KCN. Deprotection with LiOH yielded then the product. In a final step, the so obtained differently substituted 3-(hetero) aromatic-1,2,4-oxadiazol-5-yl-propanoic acids were coupled to various arylamides using standard (dimethylamino)-pyridine (DMAP)-catalysed carbodiimide chemistry. The obtained products were isolated either by semi-preparative high-performance liquid chromatography (HPLC) or crystallization (details can be found in the 'Experimental' section).
To suggest potential radiolabelling schemes of the compounds suitable for molecular imaging studies by PET, several strategies were evaluated, which are given in Scheme 3. For instance, 5e was synthesized by methylation of 4e with methyl iodide. This approach opens the way to introduce, for example, 11 C or 18 F by N-alkylation into the final compound. A further approach is the use of prosthetic groups, such as the introduction of fluorine-substituted alkyl chains via hydroxyl groups. In this study, 6k was synthesized from 6j by adding 1-fluoro-2-iodoethane. 18 F can then be introduced into the molecule by using 18 F-labelled alkylating agents. Another approach for the synthesis of 18 F-labelled compounds is the replacement of fluorine with 18 F via [2.2.2]cryptand chemistry, which is shown in this study as an example in the 'Radiochemistry' section. Figure 1 shows the single-crystal X-ray structure of compound 6f. As expected, the amide bond of this molecule (C5-N3 in Figure 1) is in trans configuration. Though several bonds in 6f are freely rotatable, the structure is planar. Thus, the dihedrals of the planes 1a  1b  2a  2b  2c  3b  4e  5e  6a  6b  6c  6d  6e  6f  6g  6h  6i  6j  6k  6l  6m  7h Ar Scheme 1 N-aryl-((hetero)aromatic)-oxadiazolyl-propionamides synthesized in this study. defined by the ring systems are 8.16 (9)°between rings A and B and 11.2(1)°between rings B and C.

In vitro binding studies
The specific binding of [ 3 H]CP55,940 towards cell membranes obtained from CHO cells stably transfected with human CB1 (hCB1-CHO) or CB2 (hCB2-CHO) hCB-CHO accounted for 40% to 50% of total binding. Nonspecific binding was related mainly to binding to the glass fibre (data not shown). For the binding of [ 3 H]CP55,940 towards hCB1-CHO and hCB2-CHO cell membranes, K D values of 2.43 ± 1.89 nM (n = 3) and 1.48 ± 0.88 nM (n = 4) were determined in competitive binding experiments, which correspond with recently published data [37,38]. Kinetic analysis of the binding of [ 3 H] CP55,940 towards hCB2-CHO cell membranes revealed the rate constants k on = 0.266 × 10 9 M −1 min −1 and k off = 0.196 min −1 . The accordingly estimated K D value of 0.73 nM is consistent with the data obtained by saturation assay. The inhibition constants (K i ) for each of the human CB receptors are compiled together with reference compounds CP55,940, SR144528, and SR141716A in Table 1. In Figure 2, the inhibition curves of 6f and 6h on the hCB2-CHO cell homogenates obtained in the radioligand displacement experiments with [    The data on the hCB2 affinity and selectivity of the synthesized compounds are in good agreement with the results published by Cheng et al. [36] and are consistent with the therein reported potencies of similar compounds. For compound 6a (compound 44 in [36]), very high potency with an EC50 value of 0.85 nM was obtained by functional cAMP assays. Taken together with the high affinity measured for 6a (K i = 6.98 nM), it is obvious that the efficacy of this compound is mainly caused by the strength of the direct physical interaction with CB2. Thus, the K D values can be seen as a direct measure of the activity of the compounds under investigation.
As the data in Table 1 show, the variation of the arylamide moiety has a significant impact on the affinity towards the cannabinoid receptors. The compounds with carbazolamides (Y 4 in Scheme 1) show, on average, the highest affinity to hCB2. The 5-tert-butyl-isoxazol-3-yl derivative 1a shows high affinity too, while a quinoline substitution (Y 2 in Scheme 1) results in less favourable properties. The approximately 100-fold lower hCB2 affinity of the 3-substituted quinoline 2a is in good agreement with published efficacy data, where a 6-substituted quinoline shows an approximately 50-fold lower [36] efficacy in comparison to 6a. The situation is completely different for CB1. Here, replacement of the carbazole group with a smaller moiety results in a strong loss of affinity towards the hCB1 receptor. For compounds 2c, 2b and 2a, binding on the hCB1 receptor could not be observed. Weak affinity towards hCB1 could be observed for 7h, which bears a benzyl-substituted indole at the left-hand side. Remarkably, the affinity towards hCB2 is lost in comparison to 6h despite the identical 2,4-diflourophenyl moieties at the 3-position of the oxadiazole ring. While the majority of the compounds bind to hCB2, those with an unsubstituted nitrogen in the ring system (aryles Y 1 , Y 2 , Y 4a in Scheme 1) show no binding to hCB1 at all. In general, the affinity of all compounds towards hCB1 is much lower than that towards hCB2. This might indicate a different binding mode of 6h and 7h at hCB2.   Comparing the compounds with the highest affinity (6g) and the best specificity for hCB2 (6e), the cause of this seems to be encoded in the functional group at para-position of the phenyl group. In concordance with published data [36], a fluorine atom at this position increases the affinity and efficacy towards hCB2. Depending on its electronegativity, a second substituent at ortho-position shows a significant impact on the affinity as shown in Figure 3. While the introduction of -Br (6e) or -CN (6a) leads to a higher affinity, the -OCH 3 (6l) or -OCH 2 CH 2 F (6k) substituents decrease the affinity. Interestingly, replacement of the methoxy group with a hydroxyl moiety (6j) results in a further loss of affinity.

Autoradiographic studies
For further investigation of the CB2 affinity of the test compounds, a preliminary autoradiographic investigation of the displacement potential of reference and test compounds on mouse spleen tissue co-incubated with [ 3 H]CP55,940 has been performed. Figure 4 shows representative autoradiograms of coronal mouse spleen sections incubated with 6 nM [ 3 H]CP55,940 in the presence of 1 μM of test compounds 6a, 6e and 6h. For comparison, the autoradiograms for CP55,940, the CB1selective antagonist SR141716A, and the CB2-selective antagonist SR144528 are shown in Figure 4b. As visible in Figure 4a, the distribution of [ 3 H]CP55,940 was heterogeneous within the coronal spleen sections, showing a pattern similar to that reported previously [18,39]. Accordingly, high-density (HD) and low-density (LD) regions were assumed to reflect binding sites in the white or the red pulp, respectively.
To estimate the displacement efficacy of the reference and test compounds, the intensity of the radioligand binding in the HD regions was chosen to assess total binding of [ 3 [18]. The assignment of the [ 3 H]CP55,940 binding sites in mouse spleen to either CB1 or CB2 has been obtained by co-incubating the tissue with [ 3 H]CP55,940 and either the CB1-selective SR141716A or the CB2-selective SR144,528 antagonist. As visible in the panels of Figure 4, co-incubation with 1 μM SR144528 displaced 78% of the specific binding of 6 nM [ 3 H]CP55,940, while nearly no displacement has been detected in the presence of 1 μM SR141716A. In the latter, 94% of the specific CP55,940 binding remained. Thus, the basal binding of [ 3 H]CP55,940 in mouse spleen was identified to reflect mainly CB2 binding, and the experimental conditions in this autoradiographic study are suitable to assess the CB2 binding potential of the test compounds. Figure 3 Influence of substitution at ring C (cf. Figure 1) on affinity towards hCB2. The pK i values were calculated according to pK i = −log (K i ). High pK i values represent high affinity. At 1 μM, 6a, 6e, and 6h displaced 83%, 71%, and 66% of the specific binding of 6 nM [ 3 H]CP55,940 respectively, and these data correlate to the rank order of CB2 affinity obtained on hCB2-CHO cells (K i = 4.27, 6.98 and 39.4 nM, respectively). However, the displacement potential of the three compounds in mouse spleen is slightly lower than that calculated according to the Gaddum equation, based on which a fractional occupancy of 98%, 97% and 83%, respectively, of [ 3 H]CP55,940 binding sites has been estimated (cf. 'Experimental' section). This difference between experimental and calculated data might reflect the previously reported non-significant variances between affinity values obtained on native tissue by in vitro autoradiography and in radioligand displacement studies using transfected cells [40]. Alternatively, as shown for the binding of WIN55,212-2 towards [ 3 H] CP55,940-labelled mouse or human CB2, species differences might also explain the slightly distinct affinity of the studied compounds [41].

Molecular modelling
In order to gain more insight into the binding of the herein synthesized compounds towards hCB2, we performed molecular modelling studies. As no experimentally determined structures of cannabinoid receptors are available yet, we created a three-dimensional (3D) structure model of this G protein-coupled receptor (GPCR) by comparative modelling. Based on the results obtained by a sequence search with the BLAST server, the 3D structure of the human beta1-adrenergic receptor (bAR1) [PDB:2Y00] [42] was identified as the structure with the highest sequence similarity to the hCB2 sequence. However, during the course of this study, a GPCR structure with a bound agonist was published [43]. In comparison to the bAR1 structure [PDB:2Y00], which is crystalized with a bound inverse agonist, this novel structure is more opened and thus represents a better description of a GPCR in an active state. Thus, the X-ray structure of the human A2A adenosine receptor with UK-432097 ([PDB:3QAK]) [43] was chosen as the template for comparative modelling. The cocrystalized ligand of this structure was kept in place during the building process of the receptor model to ensure a state ready to host an agonist. Figure 5 shows the constructed 3D model of the human CB2 with the proposed binding pocket. In agreement with mutagenesis data compiled by Poso and Huffman [44], residues known to be involved in the binding of CP55,940 are accessible in our model.
The results of the docking of 6a, 6e and 6h are shown in Figure 6. The pose with the lowest predicted binding energy is shown in green. In the majority of the predicted poses, the molecules are bent and deviated significantly from a planar structure, which was observed for 6f in the single crystal. However, the bound conformation of a ligand is not necessarily identical to that of the free molecule in solution or in a packed crystal. Moreover, due to thermal effects, several differently populated conformations might exist. Typically, molecular docking procedures deliver multiple suggestions for the binding geometry of a ligand in a protein-ligand complex. Identifying the native pose out of these suggestions is often challenging, especially, if no experimentally determined reference structure is available. By taking into account the computed binding energy, it is possible to weight the suggested poses employing Boltzmann statistics.
In Figure 6 this is illustrated by encoding the energy in the visibility of the molecule according to the method implemented in the program PostDock [45]. While for 6h only one representative pose was predicted, for 6a and 6e, two and three lower populated binding modes were found, respectively. This is illustrated by the blue shades in Figure 6, which represent the less favourable poses of these compounds. Nevertheless, in all cases, the three compounds bind primarily with their arylamide moiety inside the binding pocket. A detailed analysis of the binding site revealed a hydrophobic microdomain at the bottom of the pocket. The comparison of 6a with 5e indicates that a substituent at the N atom of the ring system increases the affinity towards CB2. If such a substituent is missing like in 4e, no binding at hCB2 could be observed (data not shown). However, a replacement with a longer or bulkier moiety reduces the binding affinity too, as it is visible from the data obtained for 7h. As illustrated in Figure 7, the preferred binding of 7h is with the arylamide moiety outside the binding pocket, which, together with the determined binding data, substantiates a different binding mode of this compound. Thus, the 9-ethyl-9H-carbazole group seems to have an optimal size for fitting into the CB2 binding pocket.
The low nanomolar affinity of 6a (K i = 6.98 nM), 6e (K i = 4.27 nM) and 6h (K i = 39.4 nM) towards the human CB2 and the more than 1,000-fold selectivity over hCB1 (K i > 1 μM) make these fluorine-substituted analogues the currently most promising candidates for further development of PET imaging agents within the series investigated. A proposed labelling at p-position of the phenyl group at the right-hand site seems to be favourable, e.g. the replacement of fluorine by 18 F would allow applying these compounds for molecular imaging of CB2 receptors.

Radiochemistry
The labelling with 18 F was performed exemplarily on 6g, yielding [ 18 F]-6e, since 6e has been identified as highly selective for hCB2 with a high affinity of K i = 4.27 nM (Scheme 4). 18 F was introduced by microwave-assisted nucleophilic substitution of the 4-NO 2 moiety employing [2.2.2]cryptand (Kryptofix ® , VWR International GmbH, Darmstadt, Germany). The electron-withdrawing effect (−I effect) of the Br substituent in meta-position to the leaving group -NO 2 enables the nucleophilic attack.
However, this electron-withdrawing effect might not be strong enough as the reaction delivered only poor yields (3% radiochemical yield). An introduction of nitrogen into the aromatic ring should facilitate the nucleophilic substitution. Thus, the pyridine equivalent 6c was labelled with 18 F under comparable conditions (Scheme 4). Though not supported by a further electron-withdrawing substituent, the radiochemical yield is significantly higher with 28%. Unfortunately, compounds with a pyridine moiety as ring C show only low affinity towards hCB2, as visible in Table 1 (compounds 1b, 2b, 2c, 3b, 6b, 6c).

Experimental
Chemistry All reagents were commercially obtained a and used without further purification unless otherwise stated. Anhydrous solvents were transferred via an oven-dried syringe or cannula. Flasks were oven-dried under vacuum and cooled under a constant stream of nitrogen. 1 H, 13     MHz for 19 F NMR; Billerica, MA, USA). The chemical shifts are reported relative to the residual solvent peak, which was used as an internal reference (chemical shifts in δ values, J in Hz). The following abbreviations were used to describe peak splitting patterns when appropriate: br = broad, s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, dd = doublet of doublet, ddd = triplet of doublet, pt = pseudo-triplet. High-resolution mass spectrometry (MS) was performed on a Bruker Daltonics APEX II FT-ICR, and low-resolved mass spectra (using electrospray ionization (ESI)) were recorded on a Bruker ESQUIRE. Reactions involving moisture-sensitive reactants were performed in oven-dried glassware under an atmosphere of nitrogen, reactants being added via a syringe. Flash column chromatography was performed on silica gel (VWR, 60 Å, 40 to 63 μm, VWR GmbH, Darmstadt, Germany) and analytical thin-layer chromatography (TLC) on pre-coated silica gel plates (Roth, 60 Å, F254, 0.25 mm, Carl Roth GmbH & Co. KG, Karlsruhe, Germany). HPLC was performed on silica gel (ReproSil-Pur 120 ODS3, C18, end-capped (endc.)) from JASCO (Jasco Labor-& Datentechnik GmbH, Groß-Umstadt, Germany) with a HPLC from Nordantec GmbH (Bremerhaven, Germany). All compounds were obtained with 95% purity.

General amide coupling method A
The coupling reaction was initiated by dropwise addition of N-ethyl-N0-(3-dimethylaminopropyl)carbodiimide (EDAC; 50 mg, 322 μmol, 1.1 eq.) in dry DCM (2.5 mL) to a mixture of 1.0 eq. of the deoxygenated propanoic acid derivate and 1.1 eq. amine in dry DCM (2.5 mL) at 0°C. The reaction mixture was stirred for 1 h at 0°C followed by stirring for 1 h at room temperature. A tip of spatula of 4-DMAP was added at 0°C, and the reaction mixture was stirred for 1 h at room temperature. EDAC (1.1 eq.) in dry DCM (2.5 mL) was added at 0°C, and the reaction mixture was again stirred for 14 h at room temperature followed by evaporation to dryness after concentration of the solvent in vacuo.

Animals
Animal experiments were performed under procedures approved by the State of Saxony Animal Care and Use Committee and conducted in accordance with the German Law for the Protection of Animals. Female CD-1 mice (10 to 12 weeks old, 20 to 25 g), obtained from the Medizinisch-Experimentelles Zentrum der Universität Leipzig (Leipzig, Germany), were used for the experiments.

Autoradiographic binding study
The animals were anesthetized and sacrificed, and their spleens were rapidly removed and frozen in isopentane at −35°C. The spleens were brought to −15°C in a cryostat (Microm International GmbH, Walldorf, Germany), and a series of 12-μm-thick coronal sections obtained from the spleens of three animals were collected in parallel on microscopic glass slides. The sections were dried at room temperature and stored at −35°C until they were processed for receptor binding autoradiography.
For autoradiographic experiments, the slides were brought to room temperature, dried in a stream of cold air, pre-incubated in a binding buffer (50 mM TRIS (pH 7.4) at 21°C, 5% BSA) for 15 min at room temperature and dried again in a stream of cold air. To account for total binding, samples were incubated for 2 h at room temperature with 6 nM [ 3 H]CP55,940 (5,328 GBq/mmol, PerkinElmer Life and Analytical Sciences, Rodgau, Germany). Sections were incubated in the presence of 1 μM CP55,940 (non-specific binding), SR144528 (CB2selective inverse agonist), SR141716A (CB1-selective antagonist), or a test compound in binding buffer. Samples were washed twice for 30 min at 4°C in 50 mM TRIS (pH 7.4) at 4°C, containing 1% BSA, dipped briefly in ice-cold deionized water (5 s), dried in a stream of cold air and exposed along with [ 3 H] standards (Amersham/ GE Healthcare, Piscataway, NJ, USA) for 14 days to [ 3 H]-sensitive screen plates (Fuji Photo Film, Co. Ltd., Tokyo, Japan).
The image plates were analysed using a BAS-1800II system bioimaging analyser (Fuji Photo Film, Co. Ltd., Tokyo, Japan). Scan data were visualized and processed by computer-assisted microdensitometry (Aida version 2.31, Raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany). Irregular regions of interest (ROIs) were drawn on selected areas of the spleen labelled as red pulp (low-density binding) and white pulp (high-density binding) according to Lynn and Herkenham [18] and Massi et al. [39]. The intensity of the radioligand binding was assessed by measuring the background-corrected optical density in each ROI expressed as photostimulated luminescence per square millimetre The theoretical fractional occupancy (f ) of [ 3 H]CP55,940 and the reference and test compounds at CB2 was calculated using the Gaddum equation: [50].

Receptor modelling
The receptor model was created employing the 'Homology Modeling' module of the program Molecular Operation Environment (MOE, version 2010.11, Chemical Computing Group Inc. Montreal, Canada). For this purpose, the sequence of the human CB2 [Swiss-Prot: P34972] [4] was subjected to a PDB search on the protein structure database implemented in MOE 2010.11 with the Z cut-off set to 2.0 while leaving the other parameters at their default values. The search suggested the human adenosine receptor A2a (hAA2R) as the template for the comparative modelling procedure. Since the recently published structure of the agonist-bound hAA2 receptor [43] is not included in the MOE protein database, the hAA2R structure [PDB:3EML] [51] was chosen as reference for the sequence alignment. The alignment of the human CB2 sequence with that of hAA2R is given as Additional files 1, 2, and 3. After downloading the 3D structure of [PDB:3QAK] [43] from the RCSB database, the structure was superimposed on the crystal structure 3eml. After removal of all lipid molecules, ions and water molecules, the residues between 209 and 222, which belong to lysozyme T4, were removed from the structure. In order to improve the modelling accuracy, the co-crystalized agonist UK-432097 of [PDB:3QAK] remained in the system during the modelling procedure. Out of ten independently generated models employing the default parameters of the Homology Modeling module, a consensus model was computed. To remove clashes and optimize the overall geometry of the protein, the system was subjected to an energy minimization employing the implemented AMBER89 force field using the default values. The quality of the model was verified by running the Protein Report module of MOE. After removal of the template ligand UK-432097, the apo receptor was energy-minimized to convergence, giving the final hCB2 model structure.

Docking studies
The small compounds were built in MOE with subsequent geometry optimization based on the MMFF force field as incorporated in the program package with a distance-dependent dielectric constant. The residues of the proposed binding pocket were defined based on the Site Finder module of MOE. The docking site included residue Y190 5.39 (numbering scheme according to Ballesteros and Weinstein [52]), which is known to be involved in ligand binding (cf. [53] and references cited therein). For the molecular docking studies, the program GOLD [54] (version 5) with parameters set to their default values was employed. All single bonds of the small ligands were allowed to rotate freely during the docking simulation. The side chains of the residues composing the binding pocket were treated as flexible, and the atoms of the protein backbone were kept at fixed positions. This approach is considered to represent the main characteristics of an induced fit [55]. The predicted poses were clustered by the GOLD software with a cutoff of 1.0 Å (heavy atoms only), giving representative geometries of the small molecules. For each of the compounds, all final poses were subjected to a further analysis with the program PostDock [45].

Conclusions
Various N-aryl-3-((hetero)aromatic-oxadiazolyl-propionamides have been synthesized. Among those, 9ethyl-9H-carbazole-substituted compounds showed the highest affinity towards CB2 and good selectivity against CB1 receptors. The study revealed strong relationships between the structure of the CB2 ligands and their CB2 affinity and selectivity. The data of this study indicate that a modification of the compounds at the amide moiety shows a stronger impact on binding and specificity of the compounds toward CB2 than variation at the (hetero)aromatic ring located at the 3position of the oxadiazoles. Thus, introduction of functional groups for subsequent 18 F labelling at this moiety is a promising strategy for the creation of CB2selective radioligands suitable for (neuro)imaging by PET. Such imaging studies are currently in process at our laboratory. Endnotes a