Structure-activity relationship of prevalent synthetic cannabinoid metabolites on hCB1 in vitro and in silico dynamics
https://ror.org/05ynxx418grid.5640.70000 0001 2162 9922Division of Clinical Chemistry and Pharmacology, Department of Biomedical and Clinical Sciences, Faculty of Medicine and Health Sciences, Linköping University, SE 581 85 Linköping, Sweden
https://ror.org/00x69rs40grid.7010.60000 0001 1017 3210Department of Science and Engineering of Matter, Environment and Urban Planning, Polytechnic University of Marche, Ancona, Italy
https://ror.org/00x69rs40grid.7010.60000 0001 1017 3210Department of Biomedical Sciences and Public Health, Polytechnic University of Marche, Ancona, Italy
https://ror.org/04mq2g308grid.410380.e0000 0001 1497 8091Institute of Chemistry and Bioanalytics, School of Life Sciences, University of Applied Sciences and Arts Northwestern Switzerland, CH 4132 Muttenz, Switzerland
https://ror.org/05ynxx418grid.5640.70000 0001 2162 9922Department of Physics, Chemistry and Biology, Linköping University, SE 581 85 Linköping, Sweden
https://ror.org/02dxpep57grid.419160.b0000 0004 0476 3080Department of Forensic Genetics and Forensic Toxicology, National Board of Forensic Medicine, SE 587 58 Linköping, Sweden
https://ror.org/052tfza37grid.62562.350000 0001 0030 1493Center for Forensic Science Advancement and Application, RTI International, Research Triangle Park, NC 27709 USA
https://ror.org/05ynxx418grid.5640.70000 0001 2162 9922Department of Biomedical and Clinical Sciences, Science for Life Laboratory, Linköping University, SE 581 85 Linköping, Sweden
Abstract
Synthetic cannabinoids (SC) target the human cannabinoid receptor 1 (hCB1) and are extensively metabolized, but the metabolite activity on the hCB1 receptor after a SC intake is largely unknown. In this study we compared the in vitro hCB1 receptor activity of 26 metabolites of the synthetic cannabinoid receptor agonists (SCRA) JWH-018, AM-2201, THJ-018 and THJ-2201 as a model system for SC metabolite activity to elucidate their structure-activity relationships. The efficacy and potency of metabolites were assessed using an AequoScreen hCB1 receptor assay in triplicates and 7–8 concentration points (20 µg/mL–9.5 ng/mL) were used to construct dose-response curves and to determine EC50 and Emax. In silico docking and molecular dynamics were performed using a model of the active form of the hCB1 receptor with all the metabolites. Final poses were simulated to assess stability under physiological conditions. We showed that carboxylic acid metabolites and 2-hydroxyindole biotransformational products were inactive, while 5-hydroxypentyl SCRA metabolites decreased efficacy to <70%, qualifying them as partial agonists. Eighteen metabolites retained >70% efficacy of their parent compound. Metabolite potencies ranged from 13–3500 nM where the most potent were the 4-hydroxypentyl derivatives of THJ-2201 and THJ-018 and the 4-hydroxyindole derivatives of AM-2201 and JWH-018, also known to be prevalent in vivo metabolites. The efficacy data from in silico experiments were correlated with the in vitro results demonstrating a linear trend (R2 = 0.9457), significant (P < 0.0001) at the 95% confident interval between the binding energies and efficacies of the compounds investigated. In silico analysis with docking and molecular dynamics simulations showed that active metabolites maintained a minimum of six amino acid interactions involving all substructures. The in silico molecular dynamics simulations revealed that the efficacy and potency seemed to be driven by a complex network of hydrophobic weak amino acid-ligand interactions. Most prevalent were CH-π interactions and π-π stackings. This study demonstrates the clear structure-activity relationships well correlated to the molecular dynamics simulations, suggesting that metabolites, especially the 4-hydroxy pentyl metabolites, may contribute to the overall effect of SCs in vivo.
Introduction
Synthetic cannabinoids (SC) were initially synthesized for research purposes with the aim of finding new pain relief medicine by selectively targeting human synthetic cannabinoid 2 (hCB2) receptors [1]. The Huffman laboratory researched a series of alkyl indoles (JWH series) for binding affinities on CB1 and CB2 to find a common pharmacophore [1]. Makriyannis and colleagues investigated among other factors, the effect of an incorporated fluorine at the terminal carbon of the alkyl indoles (AM series). However, the compounds were found to be non-selective, also activating CB1 receptors, showing psychoactive effect and thus having abuse potential [2].
In 2008, JWH-018 (1-naphthalenyl(1-pentyl-1H-indol-3-yl)-methanone) was rediscovered as a drug of abuse when it was first identified as the active ingredient in Spice [3]. Since then, clandestine laboratories have produced a wide range of SC and at the end of 2023, the total number of cannabinoids monitored by European Union Drugs Agency (EUDA) were 254 [4]. The scheduling of JWH-018 paved way for similar SC. The most common structural elements of SC are indole or indazole cores with pentyl or 5-fluoropentyl tail groups [5]. Among the structural analogs of JWH-018 identified in forensic samples are the fluoro analog of JWH-018, AM-2210 ([1-(5-fluoropentyl)-1H-indol-3-yl]-1-naphthalenyl-methanone), the indazole analog THJ-018 (1-naphthalenyl(1-pentyl-1H-indazol-3-yl)-methanone) and the fluoro analog of THJ-018, THJ-2201 ([1-(5-fluoropentyl)-1H-indazol-3-yl]-1-naphthalenyl-methanone). These structurally related SC (Fig. 1) were first encountered by EUDA in 2011 (AM-2201), 2014 (THJ-018) and 2013 (THJ-2201) [6].
To date, JWH-018 is commonly used as reference, comparing its characteristics to novel SC in research literature and governmental reports [7–9]. The new role as reference increased the demand for more in-depth data to completely understand its pharmacology including metabolite hCB1 activity. Several research studies report JWH-018 metabolism, affinity and potency on the hCB1 receptor, mapping reported concentrations in authentic samples and adverse effects after intake [10–12]. Prevalent phase I metabolite pathways were monohydroxylation, resulting in positional isomers on the pentyl tail or on the indole core (for JWH-018 and AM-2201) as well as carboxylation on the pentyl tail (18, 23–25). These data enabled, e.g., synthesis of metabolite standards for forensic research [13–17].
Selected SC metabolites have been investigated for CB1 binding affinities and efficacy in vivo using a rat model and in vitro. 4- and 5-hydroxypentyl metabolites of JWH-018, JWH-122, JWH-210 and PB-22 were found to be active CB1 agonists, in contrast to their pentanoic acid metabolites or the carboxylic acid PB-22, which is inactive on the hCB1 receptor but activated the serotonin receptor 5-HT2A [13, 18, 19]. 4-, 5-, 6- and 7-Hydroxyindole metabolites of JWH-018 were found to be active on the CB1 receptor [19]. Moreover, beta-D-glucuronide 5-hydroxypentyl JWH-018 has been reported as a neutral antagonist on hCB1 [20]. It is hypothesized that active SC metabolites are responsible for prolonging the effects of SC or are implicated in SC adverse reactions [21, 22].
Although some pharmacological data on SC and their metabolites exist, there is a lack of knowledge regarding the effect of metabolites in humans and on the hCB1 receptor. SC metabolites are often position isomers formed in varying abundances. As it is laborious to elucidate the exact structures of the metabolites from each SC, synthesize all metabolite standards and determine receptor activity, larger pharmacological studies on SC metabolites are scarce. However, such investigations would provide further knowledge on the pharmacological effects of existing SC, which is crucial for tracking SC in drug screens and for enabling the synthesis of relevant metabolite standards as reference materials.
In this study, we investigated and compared the potential activity of metabolites from JWH-018, AM-2201, THJ-018 and THJ-2201 on the hCB1 receptor. The parent compound and metabolite structures are shown in Fig. 1. By determining efficacy and potency of a large set of metabolites using a functional hCB1 receptor assay, we were able to investigate structure-activity relationships for the compounds. Moreover, we assessed the receptor-ligand interactions and binding characteristics in silico using docking and molecular dynamic simulations, with the aim of identifying the amino acids binding to the ligand and using the results to explain the differences observed in hCB1 activation induced by the metabolites.
Materials and methods
Drugs and chemicals
JWH-018, THJ-018, AM-2201, THJ-2201, 4- and 5-hydroxypentyl JWH-018, 4-hydroxypentyl AM-220, pentanoic acid JWH-018, 2-, 5-, 6- and 7-hydroxyindole AM-2201, 2-, 5-, 6- and 7-hydroxyindole JWH-018 were purchased from Cayman Chemicals (Ann Arbor, MI, USA). All other metabolite standards (presented in Fig. 1) were synthesized at Linköping University and their synthetic schemes and proof of purity in the form of NMR spectra, UV-Vis chromatogram and mass spectra are shown in Supplementary Material A. Stock solutions at 1 mg/mL were prepared in acetonitrile or methanol and stored at −20 °C. Transparent Dulbecco’s modified Eagle’s medium/Ham’s F-12 Nutrient Mixture (DMEM/Ham’s F12) supplemented with 15 mM HEPES and L-glutamine were purchased from Thermo Fisher (Gothenburg, Sweden). Trypsin was also purchased from Thermo Fisher Scientific (Gothenburg, Sweden). Adenosine triphosphate (ATP), digitonin, fetal bovine serum (FBS) and protease-free bovine serum albumin (BSA) were supplied by Fluka (Sigma-Aldrich, Stockholm, Sweden). Coelenterazine was from Nanolight Tech (Pinetop, AZ, USA). Stock solutions of 500 μM coelenterazine were prepared in methanol (and protected from light), 50 mM digitonin in DMSO and 10 mM ATP in Milli-Q water were stored at −20 °C.
Cell lines and receptor assay
Calcium sensitive AequoScreen recombinant CHO-K1 cells expressing the hCB1 receptor were purchased from Perkin Elmer (Waltham, MA, USA) and stored in liquid nitrogen prior to culturing. The cell line was in culture for less than two months before the analysis was performed. The receptor assay was carried out according to the manufacturer’s recommendations. Specifically, AequoScreen recombinant CHO-K1 cells were transferred in DMEM/Ham’s F12 with 0.1% BSA. The cell suspension was spun down and the supernatant was discarded. After resuspension with 1 mL DMEM/Ham’s F12 medium, cells were diluted in assay medium to a final concentration of 3 × 105 cells/mL. Coelenterazine, final concentration 2.5 µM, was added to the cell suspension and the cells were incubated for 3 h (dark, room temperature) prior to flash luminescence analysis (reading time 25 s/well) using a Spark 10 M with injector (Tecan, Switzerland).
In silico ligand-hCB1 docking
The structure of each metabolite and drug was generated and minimized using UCSF Chimera [23]. The three-dimensional structure of the hCB1 receptor was obtained from the 6N4B pdb file [24] and modified to include the N-domain by using I-Tasser [25]. The generated files were then used to perform molecular docking.
The ligand-hCB1 interactions were investigated to rationalize the binding mode of the ligands using AutoDock Suite 4.2 [26]. Autodock tools were used to add polar hydrogen atoms and partial charges to the receptor and ligands [27]. Atomic solvation parameters and fragmental volumes for the hCB1 were assigned using the Addsol tool, included in the program package. Flexible torsions of the ligands were allocated with the Autotors module and all dihedral angles were allowed to rotate freely. Affinity grid fields were generated using the auxiliary program Autogrid. A grid field of 50 Å × 48 Å × 48 Å and the resulting docked conformations were clustered into families of similar binding modes, with a root mean square deviation (RMSD) clustering tolerance of 2 Å. The most populated docking conformations with more negative binding energy were considered the most stable orientations of each compound in the hCB1 receptor pocket. The above-mentioned binding energy represents the sum of the intermolecular contributions and the internal energy, in terms of the intermolecular and the torsional energetic values [28]. The binding energies of docked complexes were calculated by an empirical free energy force field with a Lamarckian genetic algorithm (LGA), which provides a fast prediction of conformation and free energy. This calculated free binding energy is related to the inhibition constant (Ki) through the thermodynamic law ∆G = −RT ln Ki.
Molecular dynamics modeling
Molecular dynamic (MD) simulations were conducted for 7-hydroxyindole AM-2201, pentanoic acid JWH-018 and 7-hydroxyindole JWH-018 with high (87%), medium (57%) and low (43%) population percentages, respectively, to preliminarily assess whether MD simulations were needed for all ligand-hCB1 complexes. As a result, all compounds with population percentages less than 74% were simulated through MD to achieve a dynamically stable pose. Details of all validity assessments performed in this study can be found in Supplementary Material B. A simulation box of 6.82 nm × 6.82 nm × 8.50 nm was generated using Charmm-GUI [29] and periodic boundary conditions (PBCs) along all axes were utilized. A membrane composed of 1-palmitoyl-2-oleoyl-sn-glyceo-3-phosphocholine (POPC, 146 lipids) was modeled and hCB1 was inserted inside the lipid membrane using the correct coordinates obtained by Positioning of Proteins in Membranes (PPM) server [30]. The simulation box was solvated by 16,582 TIP3 H2O molecules [31] and 44 Na+ ions with 52 Cl− counterions were included to reach the physiological condition of 0.15 M NaCl and to neutralize the net charge of hCB1.
For each ligand, a new model was generated, and a massive minimization step was performed followed by six cycles of equilibration. Thus, 200 ns of (MD) simulations were carried out for the production phase using a thermodynamic ensemble in which the number of molecules (N), the values of Pressure (P) and of the Temperature (T) were constant throughout the MD time.
For the equilibration step, a temperature of 298 K was applied whiles 310 K was employed during the 200 ns MD simulations. This approach, also called NPT ensemble, was used to treat each system in semi-isotropic conditions. All simulations were performed using GROMACS 2023.3 version and CHARMM36 force field [32]. Finally, the trajectories were analyzed using Visual Molecular Dynamics (VMD) [33] and UCS Chimera software. The population percentages for the binding poses of the ligands were all above 76% when the ligand-hCB1 complexes were analyzed for number and type of interactions [23]. The interaction pattern of each ligand was identified by considering all residues within 3 Å of the compound position, analyzing in detail the nature of the interactions based on the surfaces of the functional groups in contact. In this study, no intramolecular interactions were investigated. The details of the in silico study is presented in Supplementary Material B.
Statistical analysis
For data analysis, Excel by Microsoft version 16.24 and GraphPad Prism for Windows 64-bit version 8.1.1 (330) were used. The raw data from the luminescence assay were transferred into Excel, where signals from each well were summed up, blank signals subtracted and normalized to the reference compound JWH-018. Next, the concentration of the compounds was converted to molar and divided into sets according to when the experiments were performed. All data was transferred to GraphPad Prism where dose-response curves of JWH-018 were drawn and EC50 calculated for JWH-018 separately for all sets. Efficacy and EC50 values were normalized to JWH-018 to compensate for between-run variability, allowing comparisons across the whole data set. Parent drug and metabolites showing less than 20% efficacy were regarded as inactive and excluded from further testing. Efficacy and potency were statistically compared using One-Way ANOVA with multiple comparisons at the significance level of P = 0.05. The EC50 values of the metabolites were compared to the parent drugs using Dunnett’s multiple comparison test. Metabolite EC50 values were groupwise tested using Tukey’s multiple comparison test.
Results
Efficacy of AM-2201, JWH-018, THJ-018, THJ-2201 and their metabolites
Data on efficacy (Emax) are summarized in Table 1 and Fig. 2. For the active metabolites, dose-response curves used to determine the potency (EC50) on the hCB1 receptor are presented in Fig. 3. Calculated EC50 and P-values from statistical testing are presented in Table 1. Compared to the Emax of JWH-018 (102%, reference), the Emax of AM-2201 (99.67%, P-value 0.57) at 95% confidence level was not significantly different. In comparison, the Emax of THJ-018 (96.39%, P-value 0.03) and THJ-2201 (92.83%, P-value 0.0004) were significantly lower. The four drugs were found to have both hCB1 active pentyl- and indole-hydroxy metabolites with efficacies comparable to full agonists and their respective parent drugs. In total, 21/26 metabolites activated the hCB1 receptor.
All 2-, 3- and 4-hydroxypentyl metabolites retained hCB1 activity with Emax ranging from 82% to 105% of their respective parent SC (Fig. 2a–d and Table 1). In contrast, both 5-hydroxypentyl metabolites had reduced Emax (63% and 67%) and both pentanoic acid metabolites were inactive (Emax < 20%). Among the hydroxyindole metabolites (Fig. 2e, f), 4-hydroxy- and 5-hydroxyindole metabolites retained hCB1 activity with Emax 89%–95% of JWH-018 and AM-2201. On the contrary, both 2-hydroxyindole metabolites were inactive (Emax < 20%). Of interest, the 6- (Emax 59%) and 7-hydroxyindole metabolites (Emax < 20%) of JWH-018, showed less efficacy than those of AM-2201 (Emax 94% and 110%, respectively).
Molecular docking and dynamic simulation of drug/metabolite-hCB1
Receptor docking was simulated for all ligands (SC and metabolites); binding energy, population percentage and predicted Ki are presented in Table 2. For the metabolites with a chiral center (marked with * in Fig. 1), simulations were performed for both enantiomers separately; however, the racemates of the enantiomers were calculated and used for comparison with the in vitro data. Binding energies ranged from −10.89 kcal/mol for AM-2201 to −4.29 kcal/mol and the predicted Ki values were from 6.45 nM to 1599 nM. Only small differences in binding energy were observed for 2-, 3- and 4-hydroxypentyl metabolites (Table 2) and both the S- and R- forms generally exhibited similar affinities to the receptor. As seen in Fig. 4, a linear trend (R2 = 0.9457), significant (P < 0.0001) at the 95% confident interval, was found when efficacy data were plotted against the binding energies. The binding energy is a measure of binding affinity: the lower the binding energy, the higher the binding affinity for the compound binding to the receptor. A correlation was also found between the calculated Ki values and the in vitro measured log EC50 (R2 = 0.2656) significant (P < 0.0084) at the 95% confident interval (Supplementary Material B).Name Binding Energy (kcal/mol) % Population Predicted Ki (nM) AM-2201 −10.89 91 10.42 2-OH pentyl AM-2201 −10.05/−9.05 84 / 68 43.16/109.04 3-OH pentyl AM-2201† −9.91/−10.29 84 / 88 55.09/25.01 4-OH pentyl AM-2201† −9.99/−10.28 86 / 81 60.67/26.91 2-OH indole AM-2201† −4.49 89 1461.45 4-OH indole AM-2201† −10.01 83 49.22 5-OH indole AM-2201 −9.72 89 71.23 6-OH indole AM-2201 −9.79 83 67.17 7-OH indole AM-2201 −10.71 87 6.45 JWH-018 −10.44 90 22.27 2-OH pentyl JWH-018 †† −9.26/−8.56 90/85 93.42/128.91 3-OH pentyl JWH-018 †† −9.36/−9.01 93/87 109.68/139.02 4-OH pentyl JWH-018 †† −10.51/−10.45 93/91 17.50/20.56 5-OH pentyl JWH-018 † −8.14 90 142.17 Pentanoic acid JWH-018 −4.59 87 1371.86 2-OH indole JWH-018 † −4.29 82 1598.57 4-OH indole JWH-018 −9.45 87 119.21 5-OH indole JWH-018 −9.61 83 91.35 6-OH indole JWH-018 † −7.91 90 192.72 7-OH indole JWH-018 † −6.88 83 1109.88 THJ-018 −10.45 91 19.64 2-OH pentyl THJ-018†† −9.45/−9.59 82/84 118.89/91.76 3-OH pentyl THJ-018 † −10.09/−10.18 94/89 46.75/30.65 4-OH pentyl THJ-018 † −10.15/−10.16 87/80 36.18/35.12 5-OH pentyl THJ-018 † −8.28 87 138.34 Pentanoic acid THJ-018† −6.54 83 1165.12 THJ-2201† −9.48 83 88.95 2-OH pentyl THJ-2201 −9.25/−8.48 80/76 93.39/134.01 3-OH pentyl THJ-2201 −10.22/−10.18 89/84 31.9/30.64 4-OH pentyl THJ-2201 −10.47/−10.54 83/86 20.74/16.54
Identifying amino acids–ligand interactions
Analyzing the binding mode of each compound allowed the identification of amino acids involved in the interaction, within the binding pocket of the hCB1 receptor. Figure 5 depicts an overview of all interactions between amino acids of the hCB1 receptor and the ligands following molecular docking and dynamic simulations. Predominant interactions were observed for the hydrophobic amino acids, such as Val196, Leu193, Met363 and Iso271 via CH-π interactions, while minor differences were observed between R- and S-enantiomers of the hydroxylated pentyl metabolites (Fig. 5).
The total number and type of interactions between the hCB1 and ligands are depicted in Fig. 6. Arg182 had several interaction types with the parent drug and metabolites within the hCB1 binding pocket. Polar induced dipole-dipole interactions were found between Arg182 and AM-2201 and JWH-018, as well as their indole metabolites (5- and 6-OH indole AM-2201 and 7-OH indole JWH-018). However, 7-hydroxyindole AM-2201 had H-halogen-dipole interactions, a much stronger interaction, with this hydrophilic amino acid. H-bonds and H-halogen interactions with Arg182 were also identified for hydroxy-pentyl metabolites, except for 2R-OH-pentyl THJ-018, which showed a dipole-dipole interaction. Other interactions identified for the metabolites included H-H bonding and halogen interactions. For the parent drugs, halogen interaction involving the fluorine atoms of AM-2201 and THJ-2201 were with His178 and Phe177, respectively.
Compounds with low binding energy from the molecular docking approach reached ten or eleven explicit interactions with amino acids. In contrast, metabolites with higher kcal/mol showed no more than four amino acids interactions involved in the binding. At the end of the MD simulations, the number of amino acids involved was the same for each compound, even if accommodation in the binding site was always observed. All SC and their metabolites interacted with Val196 and Leu193 via CH-π interactions, with minor exceptions for enantiomeric pairs. Ser383 was found to interact with the core of active hydroxylated indole/indazole metabolites, except for 6-hydroxyindole JWH-018. Interestingly, this interaction was not observed for the parent compounds or pentyl metabolites. The amino acid patterns varied for sets of drugs/metabolites due to the presence of different functional groups. However, the molecules with related experimental efficacy exhibited similar interactions from our in silico docking and MD simulation study. The 2-, 3-, and 4-hydroxypentyl shared the overall configuration in the receptor but differed in amino acid interactions, which caused a slight rotation in comparison to one another (Fig. 7).
Discussion
In this study, JWH-018, AM-2201, THJ-018 and THJ-2201 and their metabolites were screened for hCB1 efficacy and potency using a cell-based flash luminescence assay. This hCB1 receptor assay has previously been used to report on SC effects for both research purposes and to provide governmental agencies timely data aimed at speeding up the scheduling process of novel SC [8, 9, 18, 34].
In silico ligand-hCB1 simulations gave further indications on the behavior of the parent drug and metabolites within the receptor pocket. The molecular docking approach successfully reproduced the active conformation of the human CB1 receptor (hCB1) bound to the agonist methyl N-(1-((4-fluorobenzyl)-1H-indazole-3-carbonyl)-3-methyl-L-valinate (KCA), a co-crystallized ligand in the CB1 receptor structure (PDB ID: 6N4B), following remodeling to include the N-terminal domain [24]. The hCB1 receptor was remodeled in this study to ensure that likely interaction(s) between SC/metabolites and the N-terminal were not missed. RMSD value following simulation of the remodeled hCB1 and KCA residue was less than 0.9 Å demonstrating structural congruence.
It was found that the binding energy on the active form of the hCB1 receptor inversely correlates with the in vitro efficacy (Fig. 4). This finding indicates that the results from these two independent methods overlap, making it possible to extrapolate the in vitro results based on the in silico results. A weaker correlation between Ki and EC50 values was also observed (Supplementary Material B) which further strengthens this hypothesis. Thus, the predicted poses of the ligand in the hCB1 receptor after MD simulations could be used to further explain the differences in vitro efficacy and potency for each group of metabolites by looking at the number and type of interactions between the ligand and individual amino acids of the receptor. The found correlations between the in silico model and the experimental data suggest that this approach could be used for in depth studies of hCB1 activity of other SC and their metabolites too.
Structure-hCB1 activity relationship of JWH-018, AM-2201, THJ-018, THJ-2201
Comparing ligand configuration in the hCB1 receptor for the SC reveals only minor changes. All SC were full agonists with nanomolar-range potencies in vitro. Krishna et al., 2024, previously reported key interactions between JWH-018, and other derivatives from the JWH-018 family, and hCB1 using Schrödinger’s docking approach and molecular dynamics simulations. Similar results were obtained for JWH-018, THJ-018, AM-2201, and THJ-2201, depicting interactions with the amino acids, Phe170, Phe174, Phe200, and Trp279 are pivotal in SC stability in the hCB1 receptor pocket [35]. The largest difference observed in pairwise comparisons between AM-2201 and JWH-018 configurations was that AM-2201 had a more linear orientation in the hCB1 receptor pocket due to the fluorine atom at the 5th carbon of the pentyl chain. The addition of the electronegative fluorine on the tail interacts with His178, resulting in one additional amino acid interaction and perhaps anchoring AM-2201 in the pocket, contributing to the overall stability. This slight difference does not affect the configuration of the core and head, as the orientation of the substructures overlap. Similarly, the tail of THJ-2201 was more linear as compared to that of THJ-018. In general, the introduction of a fluorine atom stabilizes the molecule by enabling stronger binding to the receptor due to its high electronegativity and lipophilicity, thereby influencing the pharmacological effects of a drug [36].
Comparing the configuration of the two 5-fluoro-pentyl SC reveals that AM-2201 and THJ-2201 have overlapping orientation in the hCB1 pocket. The difference in their binding energy likely stems from the presence of two additional amino acid interactions for AM-2201 and better stabilization of the binding by the indole core of AM-2201 compared to the indazole core of THJ-2201 (Fig. 7). Additionally, Yano et al., 2023, demonstrated the interaction with His178 may not only stabilize 5-fluoro-pentyl SCRA derivatives in the hCB1 pocket, but also may be important in activating the CB1 receptor, in essence acting as an on/off switch for 5-fluoro-pentyl SC analogues [37]. All in all, chemical modifications more important than those above-mentioned need to be introduced for the SC to give a different effect in vitro.
Implications, effects, and toxicity
To the best of our knowledge, this study is the first investigating potency of a large set of metabolites from a SC with the same in vitro hCB1 activity method and in combination with docking and molecular dynamics simulations. In total, 81% of the metabolites were found to activate the hCB1 receptor. The 4-hydroxypentyl, 5-hydroxypentyl and pentanoic acid metabolites are reported as major metabolites of JWH-018, AM-2201, THJ-018 and THJ-2201 [10, 44, 45]. In this study, the 4-hydroxypentyl metabolites were among the most potent metabolites, whilst acting as full agonists, whereas the 5-hydroxypentyl metabolites were less potent and were partial agonists. This finding suggests that metabolites contribute to the pharmacodynamic effects of SC and possibly prolonging their effects.
As previously mentioned, 5-hydroxypentyl JWH-018 is a metabolite whose structure can be formed from both JWH-018 and AM-2201, the latter undergoing defluorination during metabolism [45]. Similarly, 5-hydroxypentyl THJ-018 and the pentanoic acid metabolite of THJ-018 can be formed from both THJ-018 and THJ-2201 [44]. According to Wolfarth et al. 2015, there is a distinct difference in the metabolic patterns of SC with a pentyl tail versus a 5-fluoro pentyl tail, with the ratio of formed 4-hydroxypentyl/5-hydroxypentyl metabolites differing. During metabolism, SC with a pentyl tail seem to be predominantly hydroxylated at the fourth position over the fifth carbon of the pentyl tail [46]. Together with our results, this might indicate a possible prolonged effect of the pentyl tailed SC over the 5F-analogs. However, this needs to be confirmed by further studies as 5-hydroxypentyl metabolites are readily oxidized to the pentanoic acid via an aldehyde intermediate [47]. This metabolic route seems to be one of the major detoxification pathways to eliminate both pentyl and 5-flouro pentyl SC, as the pentanoic acids are inactive at the hCB1 receptor. To fully understand the pharmacological effect of the SC, other factors such as frequency of use, amount of dosage and clearance among other factors should also be considered.
In summary, we show that several prevalent phase I metabolites of JWH-018, AM-2201, THJ-018 and THJ-2201 activated the hCB1 receptor in vitro as agonists with efficacies and potencies comparable to the respective SC. Structure and activity relationship of positional isomers show that metabolic pathways resulting in 5-hydroxypentyl metabolites and pentanoic acid metabolites lead to a decrease in hCB1 activity, with the former acting as partial agonist and the latter being inactive. The efficacy data from in silico experiments correlated with the in vitro results demonstrating a linear trend between the binding affinity and efficacy of the compounds investigated. This correlation as well as the ability to explain the experimental data based on shifting binding poses validates the in silico model as a useful tool to model hCB1 binding to SC and their metabolites. Our data show that the efficacy and potency of the SC and their metabolites seem to be driven by a complex network of hydrophobic weak amino acid-ligand interactions. This study highlights that oxidation to 5-hydroxypentyl and the inactive pentanoic acid metabolites is likely an important mechanism for SC detoxification. In contrast, 4-hydroxypentyl metabolites retain both efficacy and potency and likely contribute to overall SC effects upon intake and possibly the duration of these cannabinergic effects. Additionally, the present study not only expound our understanding of SCRAs and their metabolites’ activity at the molecular level, but also presents a rapid and comprehensive model to enable clinical and forensic toxicologists, and public health advocates to respond timely to the constantly evolving and dynamic SCRA landscape, and NPS in general.
Supplementary information
Supplementary information
The online version contains supplementary material available at 10.1038/s41401-025-01678-5.
Acknowledgements
This research was funded by the Strategic Research Area in Forensic Sciences, grant number 2016:6 (Strategiområdet forensiska vetenskaper) at Linköping University. We also acknowledge the CINECA award under the ISCRA initiative (name of the project: ATOM-HMV, code: HP10CEE3EH), for the availability of high-performance computing resources and support to perform MD simulations. The authors would like to thank M.Sc. Anders Johansson (Department of Physics, Chemistry and Biology, Linköping University) and M.Sc. Karin Lindbom (Department of Medical and Health Sciences, Linköping University) for their help during the project.
Funding
Open access funding provided by Linköping University.
Competing interests
The authors declare no competing interests.