Characterization of binding kinetics and intracellular signaling of new psychoactive substances targeting cannabinoid receptor using transition-based reweighting method

Cannabinoid receptor 1 (CB₁), which is majorly expressed in the central nervous system (CNS) belongs to the class A G-protein coupled receptor (GPCR) family proteins (Hua et al., 2016; Mackie, 2008; Zou and Kumar, 2018; Dutta and Shukla, 2023). GPCRs are expressed in the cellular membrane and help transduce chemical signals from the extracellular to the intracellular direction with the help of the downstream signaling proteins (G-proteins and β-arrestin) (Rosenbaum et al., 2009; Latorraca et al., 2017; Weis and Kobilka, 2018). In addition, GPCRs are the largest family of drug targets due to their substantial involvement in human pathophysiology and druggability (Hauser et al., 2017; Yang et al., 2021). Significant research efforts have been invested in the discovery of drugs targeting CB₁, which helps to maintain homeostasis in neuron signaling and physiological processes (Smith et al., 2010; An et al., 2020).

Initial drug discovery efforts, especially the design of synthetic agonists, were based on modifying the scaffolds of phytocannabinoids (e.g. ${\mathrm{\Delta }}^9$-Tetrahydrocannabinol, cannabinol) and endocannabinoids (e.g. Anandamide, 2-arachidonoylglycerol) (Figure 1; Pertwee, 2006; Pertwee and Ross, 2002; Pertwee et al., 2010). The synthetic molecules, which maintain the aromatic, pyran, and cyclohexenyl ring of the most common psychoactive phytocannabinoid ${\mathrm{\Delta }}^9$-THC, are known as classical cannabinoids (Figure 1—figure supplement 1; Razdan, 2009 Madras, 2018; Dutta et al., 2022a). However, the pharmacological potential of these molecules was diminished due to their psychological and physiological side effects (‘tetrad’ side effect) (Moore and Weerts, 2022; Wang et al., 2020; Tummino et al., 2023). One such example of a synthetic cannabinoid is 1,1-Dimethylheptyl-11-hydroxy-tetrahydrocannabinol (commonly known as HU-210), which is a Schedule I controlled substance in the United States (Farinha-Ferreira et al., 2022).

Apart from the canonical structures of synthetic cannabinoids, molecules with diverse scaffolds were also synthesized through structure-activity studies (Wiley et al., 2016; Schoeder et al., 2018; Walsh and Andersen, 2020). However, these molecules also lacked any pharmacological importance due to psychological side effects (Akram et al., 2019; Worob and Wenthur, 2020). Due to the diverse structures and psychological effects, these molecules became unregulated substitutes for traditional illicit substances (Peacock et al., 2019). These synthetic cannabinoids belong to a class of molecules known as NPS as these molecules are not scheduled under the Single Convention on Narcotic Drugs (1961) or the Convention on Psychotropic Substances (1971) (Peacock et al., 2019; Madras, 2016). Synthetic cannabinoids make up the largest category of NPS molecules (Shafi et al., 2020; Alam and Keating, 2020). NPS creates a significant challenge for drug enforcement agencies, as they appeal to drug users seeking ‘legal highs’ to avoid the legal consequences of using traditional drugs and to be undetectable in drug screenings (Worob and Wenthur, 2020).

The molecular structures of NPS synthetic cannabinoids consist of four pharmacophore components: linked, linker, core, and tail groups (Worob and Wenthur, 2020; Potts et al., 2020). The core usually consists of aromatic scaffolds (e.g. indole, indazole, carbazole, benzimidazole) (Figure 1—figure supplement 2; Schoeder et al., 2018). The tail and linker groups are connected to the core. In the tail group, long alkyl chain-like scaffolds are ubiquitous in most NPSs; however, molecules with bulkier cyclic chains (e.g. AB-CHMINACA) are also present (Potts et al., 2020). Frequently encountered scaffolds in linker groups are methanone, ethanone, carboxamide, and carboxylate ester groups (Hill et al., 2018). The linker acts as a bridge between the core and the linked group. In the initial NPS synthetic cannabinoids, the linked group included polyaromatic rings; however, non-cyclic linked groups have also been identified in NPS recently (Schoeder et al., 2018; Potts et al., 2020). Structural diversity in every component, while maintaining high binding affinity and potency for CB₁ make these molecules easier for drug manufacturers and harder to ban by drug enforcement agencies (Banister et al., 2015a; Ametovski et al., 2020; Cannaert et al., 2020; Banister et al., 2015b).

The use of NPS synthetic cannabinoids has been found to cause more physiological side effects than traditional cannabinergic ‘tetrad’ side effects (Tai and Fantegrossi, 2014). These side effects include tachycardia, drowsiness, dizziness, hypertension, seizures, convulsions, nausea, high blood pressure, and chest pain (Tai and Fantegrossi, 2014; Finlay et al., 2019). For instance, Gatch and Forster have shown that the high concentrations of AMB-FUBINACA, the molecule which caused ‘zombie outbreak’ in New York, induced tremors (Gatch and Forster, 2019; Adams et al., 2017). A recent biochemical study has linked these discriminatory effects with the differential signaling of β-arrestin (Finlay et al., 2019). According to Finlay et al., NPS shows higher β-arrestin signaling compared to the classical cannabinoids, which has also been confirmed by other β-arrestin signaling studies (Finlay et al., 2019; Grafinger et al., 2021). However, a mechanistic understanding of these differential downstream signaling effects between NPS and classical cannabinoids is still missing.

Mutagenesis studies have shown that the conserved NPxxY motif of CB₁ have a larger role in downstream β-arrestin signaling than G-protein signaling (Leo et al., 2023; Liao et al., 2023). Recently published MDMB-FUBINACA bound CB₁-β-arrestin-1 complex structure also points out the importance of the unique triad interaction (Y397^7.53-Y294^5.58-T210^3.46) involving NPxxY motif in β-arrestin-1 signaling (Liao et al., 2023). However, structural comparison of the classical cannabinoid (AM841) and NPS (MDMB-FUBINACA) bound active CB₁-G_i complex shows a conformationally similar NPxxY motif (Figure 2; Krishna Kumar et al., 2019; Hua et al., 2020). In light of these experimental observations, it can be inferred that higher β-arrestin signaling stems from higher dynamic propensity of triad interaction formation for NPS-bound CB₁. We hypothesized that distinct orthosteric pocket interactions for NPS and classical cannabinoids cause differential allosteric modulation of intracellular dynamics that facilitate triad interaction.

To study these distinct dynamic effects, we compared the (un)binding of the classical cannabinoid (HU-210) and NPS (MDMB-FUBINACA) from the receptor binding site. These molecules have nanomolar affinities. Obtaining the initial pathway of ligand unbinding from unbiased sampling will be computationally expensive. Therefore, a well-tempered metadynamics approach was used to sample the unbinding event, where a time-dependent biased potential is deposited for the faster sampling of the metastable minima along the pathway (Barducci et al., 2008). However, a detailed characterization of the unbinding processes is only possible through the thermodynamics and kinetics estimation of intermediate states. Thus, a transition operator-based approach is needed, which helps to estimate the transition timescale between the states and the stationary density of each state. Estimation from these approaches usually depends on the equilibrium between the local states, which can only be maintained by reversible sampling. For high-affinity ligands like MDMB-FUBINACA and HU-210, reversible sampling is expensive as ligands move from high energy unbound states to lower energy bound states irreversibly. Hence, we implemented a transition operator approach named the transition-based reweighting analysis (TRAM) method, which can tackle this lack of local equilibrium between states by combining unbiased and biased approaches (Wu et al., 2016). TRAM has been used in in different simulation studies for estimating thermodynamics and kinetics of processes that have high free energy barriers. For example, TRAM have been utilized for characterization of small molecule and peptide (un)binding processes (Wu et al., 2016; Paul et al., 2017; Ge et al., 2021; Spiriti et al., 2022; Ge and Voelz, 2022), protein dimerization (Meral et al., 2018), ion transportation (Hu et al., 2019). To implement TRAM for our study, extensive sampling of the (un)binding process of both ligands was performed using a combination of umbrella sampling and unbiased simulations from the pathway obtained using metadynamics (see Methods section) (Kästner, 2011). We showed that TRAM can produce consistent kinetic estimation with less unbiased simulation data compared to traditional methods like the Markov state model (Prinz et al., 2011).

Based on estimates of thermodynamics and kinetics, it was observed that both NPS and classical cannabinoids have similar unbinding pathways. However, their unbinding mechanisms differ due to the aromatic tail of the MDMB-FUBINACA compared to the alkyl side chain of HU-210. Furthermore, dynamic interaction calculations reveal a major difference with TM7 between NPS and classical cannabinoid. Specifically, the hydroxyl group in the benzopyran moiety of HU-210 forms much stronger polar interactions with S383^7.39 compared to the carbonyl oxygen of the linker group in MDMB-FUBINACA. MD simulations of other classical cannabinoids and NPS molecules bound to CB₁ also support these significant interaction differences. The ligand binding effect in intracellular signaling was estimated by measuring the probability of triad formation in the intracellular region. NPS-bound CB₁ shows higher probability of forming triad interaction compared to the classical cannabinoids, which supports the experimental observations of high $\beta$-arrestin signaling of NPS-bound receptors. To validate that the triad formation is indeed caused by the binding pocket interaction differences between the two ligands, allosteric strength binding pocket residues and NPxxY motif was estimated with the deep learning technique, Neural relational inference (NRI) (Zhu et al., 2022a). NRI network revealed that binding pocket residues of NPS-bound ensemble have higher allosteric weights for the NPxxY motif compared to classical cannabinoids. These analyses validate our hypothesis that the differential dynamic allosteric control of the NPxxY motif might lead to the $\beta$-arrestin signaling for different ligands. This study provides a foundation for additional computational and experimental research to enhance our understanding of the connection between ligand scaffolds and downstream signaling. This knowledge will assist drug enforcement agencies in proactively banning these molecules and inform policies that can protect individuals from the effects of abuse.