Pharmacological evaluation of synthetic cannabinoids identified as constituents of spice

All compounds were obtained from Cayman Chemicals (Ann Arbor, MI, USA). According to the declaration by the manufacturer [liquid chromatography—tandem mass spectrometry (LC–MS/MS) data], the purity of all compounds was >95 %. We confirmed the purity in our laboratories by liquid chromatographic—mass spectrometry (LC–MS) measurements and found it to be >97 % for all compounds, except for two, RCS-8 (34; 92.3 %) and MAM-2201-4F-analog (32; 94.5 %). Compounds FUB-AKB48 (18) and A-834-735 (46) were synthesized in our laboratory at a multigram-scale and analyzed by LC–MS, ¹H nuclear magnetic resonance (NMR) and ¹³C NMR spectroscopy (for details, see the supplementary material).

Membranes of Chinese hamster ovary (CHO) cells recombinantly expressing the respective human CB receptor subtype, as described before [25], were prepared by scratching the cells off the previously frozen cell culture dishes in ice-cold hypotonic buffer (5 mM Tris-HCl, 2 mM EDTA, pH 7.4). The cell suspension was homogenized on ice for 1 min using an Ultra-Turrax (Ika, Higashiosaka, Japan) followed by further homogenization for 1 min with a dounce homogenizer, and subsequently spun down for 10 min at 4 °C and 1000g. The supernatant was centrifuged for 60 min at 48,000g. The obtained membrane pellets were resuspended and homogenized in the required amount of 50 mM Tris-HCl puffer, pH 7.4, to obtain a protein concentration at 5–7 mg/mL. Aliquots of the membrane preparation (1 mL each) were stored at −80 °C until being used [25].

Competition binding assays were performed using the CB agonist radioligand [³H](−)-cis-3-[2-hydroxy-4-(1,1-dimethylheptyl)phenyl]-trans-4-(3-hydroxypropyl)cyclohexanol (CP55,940, 4, final concentration 0.1 nM; Perkin-Elmer Life Siences, Rodgan-Ingesheim, Germany) as previously described [26]. As a source for human CB₁ and CB₂ receptors, membrane preparations of the CHO cells stably expressing the respective receptor subtype were used (30 μg of protein/well for CB₁ and 8 µg of protein/well for CB₂ receptor preparations). Stock solutions of the test compound were prepared in dimethyl sulfoxide (DMSO). The final DMSO concentration in the assay was 2.5 %. After addition of 15 μL of the test compound in DMSO, 60 μL of [³H]CP55,940 solution in assay buffer, and 60 μL of membrane preparation to 465 μL of assay buffer [50 mM Tris, 3 mM MgCl₂, 0.1 % bovine serum albumine (BSA), pH 7.4], the suspension was incubated for 2 h at room temperature. Total binding was determined by adding DMSO without a test compound. Nonspecific binding was determined in the presence of 10 μM of unlabeled CP55,940. Incubation was terminated by rapid filtration through a GF/C glass fibre filter (Perkin-Elmer, Boston, MA, USA) presoaked for 0.5 h with 0.3 % aq. polyethyleneimine solution, using a Brandel 96-channel cell harvester (Brandel, Gaithersburg, MD, USA). The filter was washed three times with ice-cold washing buffer (50 mM Tris, 0.1 % BSA, pH 7.4) and then dried for 1.5 h at 50 °C. Radioactivity on the filter was determined in a liquid scintillation counter (Topcount NXT, Packard/Perkin-Elmer, Boston, MA, USA) after 10 h of preincubation with 50 µl of scintillation cocktail (Multiscint 25, Perkin-Elmer). Data were obtained in three independent experiments, performed in duplicates. Data were analyzed using GraphPad Prism Version 4.02 (San Diego, CA, USA). For the calculation of K _i values, the Cheng-Prusoff equation and a K _D value of 2.4 nM ([³H]CP55,940 at CB₁) and 0.7 nM ([³H]CP55,940 at CB₂) were used [26].

Inhibition of adenylate cyclase activity was determined in CHO cells stably expressing the CB₁ or the CB₂ receptor subtype, respectively, using a competition binding assay for cAMP [25]. All details on the reagents and their origins were described in Ref. [25]. Cells were seeded into a 24-well plate at a density of 200,000 cells/well 24 h before performing the assays. After the incubation (see below), the cells were washed with Hank’s buffered saline solution (HBSS) consisting of NaCl (13 mM), HEPES (20 mM), glucose (5.5 mM), KCl (5.4 mM), NaHCO₃ (4.2 mM), CaCl₂·2 H₂O (1.25 mM), MgSO₄ (0.8 mM), MgCl₂ (1 mM), KH₂PO₄ (0.44 mM), and Na₂HPO₄ (0.34 mM) dissolved in deionized, autoclaved water. After addition of 190 μL of HBSS per well, cells were incubated for 2 h at 37 °C. After this period of time, the phosphodiesterase inhibitor Ro-20-1724 [4-(3-butoxy-4-methoxybenzyl)-2-imidazolidinone, Sigma-Aldrich, St. Louis, MO, USA], at a final concentration of 40 μM dissolved in HBSS, test compound, and forskolin (final concentration: 10 μM, Sigma-Aldrich), all dissolved in HBSS containing 10 % DMSO, were added to each well. The final DMSO concentration was 1.9 %. The suspension was incubated for 10 min after the addition of Ro-20-1724, for 5 min after the addition of test compound, and for another 15 min after adding forskolin. cAMP accumulation was stopped by removing the supernatant from the cell suspension and subsequently lyzing the cells with 500 μL of hot lysis buffer (100 °C; 4 mM EDTA, 0.01 % Triton X-100). Aliquots of 50 μL of cell suspension were transferred to 2.5-mL tubes, into which 30 μL of [³H]cAMP (3 nM) and 40 μL of cAMP-binding protein (50 µg) were added, followed by 1 h of incubation at room temperature. The cAMP binding protein was obtained from bovine adrenal cortex as previously described [25]. Bound and free radioligands were separated by rapid filtration through a GF/B glass fibre filter (Perkin-Elmer). Radioactivity on the filter was determined in a liquid scintillation counter (TRICARB 2900TR, Packard/Perkin-Elmer) after 6 h of preincubation with 3 mL of scintillation cocktail (LumaSafeplus, Perkin-Elmer). Data were obtained from three independent experiments, performed in duplicates.

Recruitment of β-arrestin to the respective receptor was detected by using β-galactosidase enzyme fragment complementation technology (β-arrestin PathHunter™ assay, DiscoverX, Fremont, CA, USA) as previously described [27]. Data were obtained from three independent experiments, performed in duplicates. Data were analyzed using Graph Pad Prism Version 4.02 (San Diego, CA, USA).

Properties of compounds were predicted with the program Stardrop 5.5 (Optibrium, Cambridge, UK) using the ADME QSAR (quantitative structure-activity relationship) tool predicting the Lipinski rule of five, oral central nervous system (CNS) scoring profile and intravenous CNS scoring profile with standard conditions.

All compounds were investigated in radioligand binding experiments in CHO cell membrane preparations stably expressing the human CB₁ or CB₂ receptor using [³H]CP55,940 as a radioligand. K _i values are presented in Table 1. CB₁ or CB₂ selectivity of compounds was calculated based on the K _i values, and can be found in the electronic supplementary material (Table S1). All of the investigated compounds share a common core structure: (aza)indole. Three different types of linkers between the (aza)indole ring system and a bulky, lipohilic residue are observed: an amide, an ester or a shorter carbonyl linker. The most potent compounds for the CB₁ receptor were found among the ester-linked subgroup [BB-22 (27), PB-22 (24) and 5F-PB-22 (25), NM-2201 (19)] with K _i values ranging from 0.217 to 0.468 nM. The only exception was the carbonyl-linked compound EAM-2201 (31), with a similarly low K _i value of 0.380 nM.

In all three subgroups, compounds with typical bioisosteric exchanges are found. Three features of the molecule are varied: the N1-substituent, which was originally a pentyl moiety in the lead compounds of the JWH group [13]; in the current compounds, it is fluorinated or exchanged for a para-fluorobenzyl residue. The effect of fluorination on binding affinity was moderate: in the nine examples included in our study, binding affinity for the CB₁ receptor was slightly enhanced for fluorinated compounds [compare MN-18 (8) and 5F-MN-18 (9); THJ (10) and 5F-THJ (11); APICA (15) and STS-135 (16); SDB-005 (22) and 5F-SDB-005 (23); THJ018 (28) and THJ2201 (29); and AB001 (35) and 5F-AB001 (36)] or slightly decreased [compare NNEI (4) and 5F-NNEI (5); SDB-006 (12) and 5F-SDB-006 (13); PB-22 (24) and 5F-PB-22 (25)]. Banister et al. [28] investigated the effects of fluorinated compounds and found that although the EC₅₀ value of the investigated compounds were lower in vitro, this was not translated to higher in vivo potencies, leading to the assumption that pharmacokinetic effects play a role [28]. In their study, they investigated, amongst others, the pairs UR-144 (37) and XLR-11 (38), PB-22 (24) and 5F-PB-22 (25), and also APICA (15) and STS-135 (16). They performed membrane potential measurements using a fluorometric imaging plate reader (FLIPR) assay kit and determined slightly higher EC₅₀ values for the compounds as compared to the radioligand binding data obtained in the present study. XLR-11-2-fluoropentyl-isomer (39), a derivative with a 2-fluoropentyl side chain, is the only compound in this series with a fluorine introduced at position 2 of the pentyl side chain. In comparison to the non-fluorinated analogue UR-144 (37), the affinity of 39 at the CB₁ receptor was almost the same, but it was not as potent as XLR-11 (38), the 5-fluoinated derivative. MAM-2201-4-fluoropentyl-substituted isomer (32), showed also slightly higher K _i values than the 5-fluoropentyl derivative MAM-2201 (30). In this series, only one compound contains of a 5-chloro-substitution: 5Cl-NNEI (6), which displayed about five-fold lower affinity for the CB₁ receptor than the unsubstituted derivative NNEI (4). Another bioisosteric replacement of the 5-fluoropentyl side chain is a para-fluorobenzyl residue. This variation is observed in four compounds of the present series [compare: 5F-AKB48 (17) and FUB-AKB48 (18); NM-2201 (19) and FDU-PB-22 (20); 5F-PB-22 (25) and FUB-PB-22 (26); XLR-11 (38) and FUB-144 (42)]. The affinity for both CB receptors was almost identical in three of the four pairs; only FDU-PB-22 (20) was not quite as potent as NM-2201 (19). Thus, a para-fluorobenzyl residue appears to be an optimal bioisosteric exchange for obtaining compounds with similarly high affinity as the 5-fluoropentyl-substituted parent compound.

Other side chains have been introduced at the indole nitrogen atom. Huffman et al. [13], who established alkylindoles as cannabinoid receptor ligands, already performed a comprehensive structure-activity relationship study introducing different side chains. They showed that a five-carbon side chain is preferred [13]. Thus, pentyl side chains and their bioisosteric analogs confer high potency and activity at the CB₁ receptor. Whenever the size is decreased, affinity for the CB₁ receptor is largely reduced. As this structural feature is crucial for high CB₁ affinity, it had previously been modified to design CB₂-selective compounds [29].

Another frequently observed variation is the replacement of the indole core by an indazole ring system. In the group of compounds with an amide linker (A), it could be observed that the affinity for the CB₁ receptor was quite similar for indoles and indazoles, while the affinity for the CB₂ receptor was slightly increased in indazole derivatives [compare NNEI (4) and MN-18 (8); 5F-NNEI (5) and 5F-MN-18 (9); STS-135 (16) and 5F-APINACA (17)]. In the group of compounds with an ester linkage (B), the indole derivative NM-2201 (19) showed lower K _i values at CB₁ and CB₂ receptors than the corresponding indazole derivative 5F-SDB-005 (23). In group C compounds containing a keto-group as a linker, XLR-11 (38) and its indazole analogue FAB-144 (42) displayed almost identical binding affinities. Thus, a variation of the heterocyclic core from indole to indazole is widely tolerated.

One other common feature of this group of compounds is the bulky lipophilic residue in position R¹. Huffman et al. [16] introduced mainly naphthyl residues in that position. A variation of this structural element represents the introduction of a quinoline found in some compounds such as THJ (10) and PB-22 (24) [30]. In group A compounds with an amide linker, the introduction of a quinoline led to 14- and 27-fold higher K _i values at CB₁ receptors, respectively [compare MN-18 (8) with THJ (10); and 5F-MN-18 (9) with 5F-THJ (11)], while the affinity for CB₂ receptors remained unaltered in the low nanomolar range. In the ester-linked compounds (B), the quinoline-substituted analogue of NM-2201 (19), 5F-PB-22 (25), showed comparable affinities for both receptors. FUB-PB-22 (26) is a quinoline derivative with somewhat higher affinity at CB₁ and CB₂ receptors as compared to its analogue FDU-PB-22 (20). The most potent compound in this series of cannabinoid ligands, BB-22 (27)—sometimes referred to as QUCHIC—is also a quinoline derivative, which was first described in illicit drug material in 2013 in Japan [30]. This compound has a cyclohexylmethyl residue in position R², which imitates the length of a pentyl chain that was previously described to be important for CB potency [13], and which was beneficial for CB₁ receptor affinity also in a series of magnolol derivatives [31].

Compounds MAM-2201 (30) and EAM-2201 (31) display substitution of the naphthyl residue, containing a methyl (MAM-2201 (30) or an ethyl (EAM-2201 (31) group in position 4 of the naphthyl ring. EAM-2201 (31) was highly potent at the CB₁ receptor with a K _i value of 0.380 nM without preference for any of the CB receptor subtypes. MAM-2201 (30), which had been described to cause severe toxicity in the cerebellum of rats [32], was found to be four times less potent at the CB₁ receptor.

The only compound which is not linked in the 1-position of the naphthyl residue but is linked in the 2-position, 5F-NNEI-2-naphthyl-isomer (7), was a much weaker CB₁ receptor ligand and also showed only partial inhibition of radioligand binding at the CB₂ receptor. NNEI (4), which was first described by Blaazer et al. [33] in 2011, showed a pK _i value of 8.9 in their binding experiments at the CB₁ receptor, which we have now confirmed. The authors also synthesized a non-fluorinated derivative of compound 7 (5F-NNEI-2-naphthyl-isomer) which displayed a lower pK _i value of 7.2 for the CB₁ receptor. The same relation could be shown in the present study [compare 5F-NNEI (5) and 5F-NNEI-2-naphthyl-isomer (7)]; if the naphthyl residue is linked in 2-position to the amide, the affinity was decreased by about 100-fold.

Huffman et al. [13] investigated the effects of substituting the naphthyl ring by smaller aromatic residues, which reduced affinity to the CB₁ receptor. This could also be observed for the benzyl-substituted compounds SDB-006 (12) and 5F-SDB-006 (13) investigated in the present study. They showed much lower affinity for both CB receptors as compared to the napthyl-substituted compounds with K _i values in the high nanomolar range. The phenyl-substituted derivative SDB-006-N-phenyl-analog (14) displayed even higher K _i values. In group C compounds, RCS-4 (33) and RCS-8 (34) also feature a phenyl or a benzyl residue. Wiley et al. [34] described that the substitution in the ortho-position is crucial for high affinity, which is realized in both compounds. RCS-8 (34), first described in 2012 in the USA [35], is benzyl-substituted in position 1 and has a cyclohexylethyl residue in position 2; it shows weaker affinity for both CB receptors than RCS-4. RCS-4 and isomers were investigated by Banister et al. [36] who found that RCS-4 (33) displayed EC₅₀ values of 145 nM for CB₁ and 46 nM for CB₂. In the present study, RCS-4 (33) with K _i values of 26.6 nM for CB₁ and 2.86 nM for CB₂ displayed higher binding affinities.

The aromatic residue R¹ may be replaced by a more bulky lipophilic group, namely an adamantyl or a tetramethylcyclopropyl residue. Comparing the naphthyl derivatives NNEI (4) and 5F-NNEI (5) with the adamantyl derivatives APICA (15) and STS-135 (16), it can be observed that CB₂ affinity was increased. Also, the tetramethylcyclopropyl derivatives of group C displayed, independently of the side-chain variations, a CB₂ preference. Compounds UR-144 (37), A-796,260 (45), A-834,735 (46) and XLR-12 (41) were first described by Frost et al. [29] in the search for selective CB₂ agonists. We could confirm the reported K _i values, but only XLR-12 (41) displayed a 10-fold higher K _i value in our hands as compared to the literature data. From this group of compounds, some derivatives emerged on the illicit drug market, mainly in Sweden [37, 38]. FAB-144 (40), the indazole and 5-fluoropentyl analogue of UR-144 (37), showed slightly increased affinity for both CB receptors, and FUB-144 (42), the para-fluorobenzyl derivative displayed similar affinity. Also, compound M-144 (47), which is substituted in position 2 of the indole ring system with a methyl group, displayed a similar profile. AB-005 (43), a chimeric compound with the CB₂ selectivity-increasing tetramethylcyclopropyl residue for R¹ and N-methyl-2-piperidinylmethyl substitution as R² which retains CB₁ affinity, was first introduced by Frost et al. in 2010 [29]. A derivative with an azepane ring (44) appeared on the illicit drug market, but as we found, it displayed no affinity for the CB₁ receptor at concentrations up to 10 µM. If it should exert any psychotropic effect, it would not be mediated via this receptor. At CB₂ receptors, a moderate affinity was observed for 44. A structurally related but more potent compound is MN-25 (48), which was introduced by Wrobelenski et al. [39]; it was reported to be abused in previous years [39].

In summary, almost all investigated compounds showed high affinity for CB receptors. Some compounds displayed K _i values in the subnanomolar range and, thus, are many times more potent than the psychoactive drug Δ⁹-THC.

To investigate the functional properties of the compounds, cAMP accumulation assays were performed. Both CB receptors are G_i-coupled receptors, whose activation results in decreased cAMP levels in the cell. For comparison, the full agonist CP55,940 and the partial agonist Δ⁹-THC were investigated, and results were normalized to maximal receptor activation by the full agonist CP55,940 (see Fig. 2). Compounds were tested at a concentration where maximal binding was observed, either at 1 µM for the more potent compounds or at 10 µM for the less potent compounds.

In the utilized recombinant cell lines, Δ⁹-THC behaved as a partial agonist, at both CB₁ and CB₂ receptors, with 60–70 % activation as compared to the full CB₁/CB₂ agonist CP55,940 (2). Almost all compounds showed a high degree of activation of both receptor subtypes. Exceptions were 3-CAF (21) and AB-005 azepane isomer (44), which did not activate the CB receptors at all. As both compounds showed affinity for the CB₂ receptor, they may be characterized as moderately potent CB₂-selective antagonists. The only agonistic compounds with lower efficacy than Δ⁹-THC were NNEI-2-naphthyl isomer (7), MN-18 (8), XLR-12 (41) and AB005 (43). Most of the compounds had similar efficacies at both receptor subtypes; only 5F-APINACA (17) activated CB₁ receptor more efficaciously than CB₂.

For the CB₂-selective derivative XLR-12 (41), full concentration response curves were recorded and EC₅₀ values were determined (Fig. 3). It showed a 30-times lower EC₅₀ value of 0.391 nM at the CB₂ receptor than at the CB₁ receptor; thus, the compound’s preference could also be observed in the functional assays.

K _i values measured in radioligand binding in many cases (Table 1) correlated quite well with reported and the EC₅₀ values determined in cAMP accumulation assays (data not shown). CB₂-selectivity of compound XLR-12 (41) could be confirmed, but in our hands it was lower (only 30-fold, Fig. 3) than previously reported one (167-fold) [40]. It should be emphasized that EC₅₀ values depend on receptor expression levels while radioligand binding data are independent of receptor density or G protein expression. They directly reflect the affinity of compounds for the binding site on the receptors.

Compounds that activate the CB₁ receptor to a comparable extent as Δ⁹-THC and that can cross the blood-brain barrier will likely cause similar psychotropic effects as Δ⁹-THC. Some compounds showed even higher efficacy than the full agonist CP55,940, including the very potent compounds EAM-2201 (31), NM-2201 (19) and BB-22 (27). Their toxicity may be much higher than that of Δ⁹-THC due to their high potency and full efficacy. PB-22 (24), a CB₁/CB₂ partial agonist with similar efficacy as Δ⁹-THC but with higher subnanomolar affinity (Table 1), had previously been reported to even cause lethal intoxications [41, 42].

As a precondition to achieve psychoactive effects, brain penetration of the compounds is required. This property can be determined in animal studies. Alternatively, an in silico prediction based on established data sets can be used to gain an idea whether a set of compounds is able to cross the blood-brain barrier. For the investigated compounds, this was accomplished using the QSAR software Stardrop 5.4 (Optibrium). In Fig. 4, affinities of the investigated compounds were compared to their lipophilicity, which is one of the major determinants for crossing biomembranes. As can be observed, all compounds share a rather high logP value between 3 and 7. All highly potent compounds exceeded a logP of 4.5. The standard CB agonists displayed similarly high logP values of 6.50 (Δ⁹-THC), and 5.36 (CP55,940). The compounds’ potency is not directly correlated with their lipophilicity (see Fig. 4). Based on calculations to estimate lipophilicity (logP), topographical polar surface area and other parameters, a prediction whether compounds are able to cross the blood-brain barrier is made by the program. The compounds could thus be divided into two groups, blood-brain barrier-penetrant and non-penetrant compounds. THJ (10) and 5F-THJ (11), both of which are 3-(8-quinolinyl)amido-indazoles, were predicted not to cross the blood-brain barrier. Based on in silico predictions it is, however, likely that the majority of the investigated compounds has the ability to cross the blood-brain barrier.

The orphan GPCRs GPR18 and GPR55 have been shown to be targeted by a range of cannabinoid receptor ligands [19, 29, 43]. Therefore, we investigated whether the investigated spice constituents also interact with these cannabinoid-related receptors (Table 2). None of the compounds was able to activate GPR18 or to inhibit GPR18 activation up to a concentration of 10 µM. At GPR55, some compounds were found to be moderate antagonists, namely, APICA (15) and STS-135 (16) with IC₅₀ values of 3–5 µM, as well as several compounds from group C. EAM-2201 (31) was the most potent GPR55 antagonists of this series with an IC₅₀ value of 1.86 µM. Interestingly, none of the ester compounds (B) showed any inhibitory effect, and most of the active compounds were tetramethylcyclopropyl-substituted derivatives containing the CB₂-preferring structure. UR-144 (37), XLR-11-2-fluoropentyl-isomer (39) and XLR-12 (41), which feature a lipophilic aliphatic or fluoropentyl side chain, were more potent than A-769,260 (45) or A-834,735 (46) with a morpholine or tetrahydropyran substituent, respectively. A typical functional behavior of cannabinoids at GPR55 can also be observed here: although all of the identified GPR55 ligands were agonists at the CB receptors, they showed inhibitory effects at GPR55. The same had been demonstrated for the CB agonist CP55,940 (2) as well as other CB receptor agonists [27, 44]. On the other hand, CB₁ receptor antagonists, such as rimonabant, are agonists of GPR55 [22, 27, 45]. Both receptors, CB₁ and GPR55, were reported to be co-localized in the brain, and receptor heteromerization has been postulated [23, 24].