Characterization of binding kinetics and intracellular signaling of new psychoactive substances targeting cannabinoid receptor using transition-based reweighting method
- Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-Champaign, United States
- Department of Chemistry, University of Illinois at Urbana-Champaign, United States
- Center for Biophysics and Quantitative Biology, University of Illinois at Urbana-Champaign, United States
- Department of Bioengineering, University of Illinois at Urbana-Champaign, United States
- Cancer Center at Illinois, University of Illinois at Urbana-Champaign, United States
Figures
Figure 1 with 2 supplements
Classification of cannabinoid agonists.
(A) Molecules derived from cannabis plants (phytocannabinoids) (B) endogenous agonists (Endocannabinoids) (C) synthetically designed molecules (Synthetic cannabinoids). Synthetic cannabinoids can be further classified based on scaffolds (phytocannabinoid analogues and endocannabinoid analogues or new psychoactive substances). Common pharmacophore groups of the ligands are shown in different colors. For phytocannabinoids and phytocannabinoid synthetic analogues, tricyclic benzopyran group and alkyl chains are colored in red and blue, respectively. Polar head group, propyl linker, polyene linker, and tail group of endocannabinoid and endocannabinoid analogues are colored with green, yellow, red, and orange, respectively. Linked, linker, core, and tail group of new psychoactive substances are colored with green, yellow, red, and orange, respectively.
Figure 1—figure supplement 1
Atom numbering scheme of classical cannabinoid.
(Δ9-Tetrahydrocannabinol).
Figure 1—figure supplement 2<bold>.</bold>
Pharmacophore components and representative new psychoactive substances (NPS) scaffolds.
Four pharmacophore components (Linked: Green, Linker: Orange, Core: Red, Tail: Cyan) of NPS synthetic cannabinoid MDMB-FUBINACA are shown in different colors (A). Existing common scaffolds of other NPS in four pharmacophore components (B).
Figure 2
Structural comparison between new psychoactive substances (NPS) bound and classical cannabinoid bound CB1.
NPS bound CB1 (PDB ID: 6N4B, Krishna Kumar et al., 2019 color: Blue) structure is superposed with the classical cannabinoid bound CB1 (PDB ID: 6 KPG, Hua et al., 2020 color: Purple). Both structures are in Gi bound active state. Proteins are shown in transparent cartoon representation. Structural comparison of conversed activation matrices (Toggle switch, DRY motif, and NPxxY motif) and ligand poses are shown as separate boxes. Quantitative values of the activation metrics for both active structures are compared as scatter points on 1-D line with the CB1 inactive structure (PDB ID: 5TGZ, Hua et al., 2016 color: orange). These quantitative measurements were discussed in Dutta and Shukla, 2023.
Figure 3 with 3 supplements
Ligand unbinding pathways for MDMB-FUBINACA and HU-210.
The ligands MDMB-FUBINACA (A) and HU-210 (C) are depicted in three distinct stages along their unbinding pathways, as determined by well-tempered metadynamics simulations. The ligands are illustrated using stick representations, with each stage represented by a different color to indicate the progression from the bound (color: red) to the unbound state (color: green). Representative ligand positions from an intermediate state are shown in yellow. Additionally, the superposition of representative frames of an intermediate stage of the unbinding process is shown, where MDMB-FUBINACA (B) and HU-210 (D) are dissociating from the receptor. The frames are obtained from two different well-tempered metadynamics simulation replicas and are shown with different colors (green and orange). Both transmembrane (left panel) and extracellular (right panel) views are displayed. Proteins are represented as cartoons.
Figure 3—figure supplement 1
Comparison of cryo-EM and docked structures of two classical cannabinoids.
Structural superposition of HU-210 (color: Orange) docked CB1 with active CB1 cryo-EM structure (PDB ID: 6 KPG, ligand: AM841). Protein is shown as a cartoon (color: purple). Ligands are shown as sticks.
Figure 3—figure supplement 2
Characterization of ligand binding pathways.
Unbinding ensemble of the MDMB-FUBINACA (A) and HU-210 (C) are projected as 2-D scatter plots where Z-component distance of ligand center of mass from W3566.48 is plotted against the ligand RMSD. These ensembles were obtained by running well-tempered metadynamics. Reweighted probability densities are plotted with respect to the Z-component distance of ligand center of mass from W3566.48 for MDMB-FUBINACA (B) and HU-210 (D). Measured qualities from two simulation replicas for each system are shown in different colors on the same plot.
Figure 3—figure supplement 3
The unbinding simulation with well-tempered metadynamics with different parameters.
The well-tempered metadynamics runs with different parameters were projected as a scatter plot on top of transition-based reweighting analysis (TRAM) weighted two-dimensional projection of unbinding free energy landscape for MDMB-FUBINACA (A, C, E) and HU-210 (B, D, F). For MDMB-FUBINACA, distance between TM5 (W2795.43-Cα) and tail part of the ligand is plotted against the distance between TM7 (S3837.39-Cα) and ligand-linked part. For HU-210, distance between the TM5 (W2795.43-Cα) and tail is plotted against the TM7 (S3837.39-Cα) and cyclohexenyl ring of the ligand.
Figure 4 with 7 supplements
Comparison of thermodynamics and kinetics estimation of the unbinding process between MSM and transition-based reweighting analysis (TRAM).
(A) The bar plot represents standard binding free energy for HU-210, MDMB-FUBINACA, and difference of standard binding free energy between the ligands. MSM and TRAM estimations are shown as blue and orange bars, respectively. Experimentally predicted values are shown as dotted line. (B, C) Binding (B) and dissociation (C) time for HU-210 and MDMB-FUBINACA are shown as box plots. (D) Difference in dissociation time of the two ligands is plotted as box plot against fraction of unbiased trajectories used for the estimation. These timescales were obtained from the mean free passage time calculation using transition path theory (TPT) with transition probabilities estimated from MSM (color: blue) and TRAM (color: orange). Errors were calculated using bootstrapping method with three bootstrapped samples.
Figure 4—figure supplement 1
Distance used to cluster the metadynamics sampled (un)binding pathway.
Modeled CB1 holo structures are shown as cartoon representations with the ligands (MDMB-FUBINACA (A) and HU-210 (B)) in the orthosteric bound pose. TM5 distances from the ligands are presented as red dotted lines. Ligands and residues in TM5 are represented as sticks. Atoms of interest are shown as vdw representation.
Figure 4—figure supplement 2
Binding pocket residues considered in MSM and TRAM analysis.
Binding pocket residues are shown as sticks. Residues wereselected from TM2, TM3, TM5, and TM7 and were used for feature calculation inbuilding the MSM and TRAM models. The same residues were also used tocompute the coordination number in the metadynamics simulations.
Figure 4—figure supplement 3
Implied timescale convergence with MSM lag time.
Top three implied timescales were plotted against different lag times for unbinding simulations of MDMB-FUBINACA (A) and HU-210 (B). MSM lag times were selected to be 35 ns for both ligands. For MDMB-FUBINACA, these calculations were performed with 700 clusters and 7 tIC dimensions. For HU-210, these calculations were performed with 800 clusters and 6 tIC dimensions. Errors in the implied timescale were calculated using three bootstrapped samples, where each sample containing 95% of the original unbiased data.
Figure 4—figure supplement 4
Optimization of VAMP-2 scores.
VAMP-2 scores of MSMs built with different cluster numbers are shown for MDMB-FUBINACA (A) and HU-210 (B) unbinding simulations. Different number of tICs used for MSM building were shown with different colors. The optimal VAMP-2 score in each case is marked with a star symbol.
Figure 4—figure supplement 5
Chapman–Kolmogorov (C–K) test to judge Markovianity of MSM.
Absolute differences between and are shown as bar plots for MDMB-FUBINACA (A) and HU-210 (B) unbinding simulations. Different values of k are considered where is 35 ns.
Figure 4—figure supplement 6
Sampled raw probability density vs estimated weighted probability from MSM and transition-based reweighting analysis (TRAM).
Raw probability and MSM weighted probabilities of the clustered states are plotted against each other for MDMB-FUBINACA (A) and HU-210 (B) unbinding simulations. Raw probability and TRAM weighted probabilities of the clustered states are plotted against each other for MDMB-FUBINACA (A) and HU-210 (B) unbinding simulations.
Figure 4—figure supplement 7
Convergence of calculation.
Difference in the of the two ligands is plotted as box plot against fraction of unbiased trajectories used for the estimations. MSM and transition-based reweighting analysis (TRAM) estimations are shown using blue and orange colors, respectively.
Figure 5
Transition-based reweighting analysis (TRAM)-weighted two-dimensional projection of unbinding free energy landscape for MDMB-FUBINACA (A) and HU-210 (B).
For MDMB-FUBINACA, distance between TM5 (W2795.43-Cα) and tail part of the ligand is plotted against the distance between TM7 (S3837.39-Cα) and ligand-linked part. For HU-210, distance between the TM5 (W2795.43-Cα) and tail is plotted against the TM7 (S3837.39-Cα) and cyclohexenyl ring of the ligand. Measured distances are shown as red dotted lines in the inset figures. Macrostate positions are shown on the landscapes. Different mechanisms of (un)binding are shown with arrow on top of the free energy landscape.
Figure 6 with 2 supplements
Mechanism of new psychoactive substances (NPS) MDMB-FUBINACA unbinding from CB1.
(A) The contact probabilities with binding pocket residues of MDMB-FUBINACA are shown as a heatmap for different macrostates, where ligand maintains contact with the receptor. Residues in different structural elements (loops and helices) are distinguished by distinct color bars. (B) Representative structures are shown where ligand (color: orange) and four residues (color: green) with highest interaction energies are shown as sticks. Proteins are shown as purple cartoon. Timescales between interstate transitions are shown as arrows. Arrow thickness is inversely proportional to the order of magnitude of the timescale. (C) Per residue K-L divergences between different states are shown with color (blue to red) and thickness (lower to higher) gradient. K-L divergences calculated on the inverse distance feature distributions were converted by residue basis by summing all the pair contributions corresponding to the residue. Thickness gradients are shown as rolling average to highlight a region of high K-L divergence. Errors in MFPT calculations were estimated based on three bootstrapped transition-based reweighting analysis (TRAM) calculations with randomly selected 95% of unbiased trajectories.
Figure 6—figure supplement 1
Contact probability and interaction energy calculations between MDMB-FUBINACA and CB1 binding pocket residues for each macrostate.
The contact probabilities with binding pocket residues and corresponding interaction energies of MDMB-FUBINACA are shown for different macrostates, where ligand maintains contact with the receptor. The macrostates presented here are in the order of I2 (A), Bound (B), I1 (C), and I3 (D) to clearly distinguish the two unbinding mechanisms. For each macrostate, ligand and binding pocket residue interaction probabilities (color: blue) and energies (color: orange) are plotted as bar plots.
Figure 6—figure supplement 2
Root mean square deviation of the MDMB-FUBINACA in different macrostates.
RMSD was shown as a density plot for Bound and I2 macrostates. Errors are calculated from five bootstrapped samples, where each sample contains 1000 conformations representing the macrostate. These 1000 conformations are selected based on the probability density of the microstates belonging to the macrostate.
Figure 7 with 1 supplement
Mechanism of classical cannabinoid HU-210 unbinding from CB1.
(A) The contact probabilities with binding pocket residues of HU-210 are shown as a heatmap for different macrostates, where ligand maintains contact with the receptor. Residues in different structural elements (loops and helices) are distinguished by distinct color bars. (B) Representative structures are shown where ligand (color: orange) and four residues (color: green) with highest interaction energies are shown as sticks. Proteins are shown as purple cartoons. Timescales between interstate transitions are shown as arrows. Arrow thickness is inversely proportional to the order of magnitude of the timescale. Errors in MFPT calculations were estimated based on three bootstrapped transition-based reweighting analysis (TRAM) calculations with randomly selected 95% of unbiased trajectories. (C) Per residue K-L divergences between different states are shown with color (blue to red) and thickness (lower to higher) gradient. K-L divergences calculated on the inverse distance feature distributions were converted by residue basis by summing all the pair contributions corresponding to the residue. Thickness gradient are shown as rolling average to highlight a region of high K-L divergence.
Figure 7—figure supplement 1
Contact probability and interaction energy calculations between HU-210 and CB1 binding pocket residues for each macrostate.
The contact probabilities with binding pocket residues and corresponding interaction energies of HU-210 are shown for different macrostates, where ligand maintains contact with the receptor. The macrostates presented here are in the order of I1 (A), Bound (B), I2 (C), and I3 (D) to clearly distinguish the two unbinding mechanisms. For each macrostate, ligand and binding pocket residue interaction probabilities (color: blue) and energies (color: orange) are plotted as bar plots.
Figure 8 with 5 supplements
Dynamic conformational change in intracellular region of the CB1 during ligand (un)binding.
(A) Transition-based reweighting analysis (TRAM) weighted probabilities of triad interaction (Y3977.53-Y2945.58-T2103.46) formation are plotted for HU-210 (color: purple) and MDMB-FUBINACA (color: blue) unbinding ensemble. If side-chain oxygen atoms of all three residues are within 5 Å of each other, triad interaction is considered to be formed. (B) TRAM weighted probability densities of TM3 (R2143.50) and TM6 (K3436.35) distance distribution are plotted for HU-210 (color: purple) and MDMB-FUBINACA (color: blue) unbinding ensemble. Error in the probability densities is estimated using a bootstrapping approach, where TRAM was built for three bootstrapped samples with 95% of total data.
Figure 8—figure supplement 1
Ligand interaction of other classical cannabinoids and new psychoactive substances (NPS) molecules with TM7.
(A) Classical cannabinoids and NPS molecules that have been used to perform unbiased ligand-bound simulations. (B) Equivalent polar interaction distances for NPS and classical cannabinoids are shown as bar plot. For each NPS, the distance from S3837.39(Oγ) is calculated from the linker oxygen atom. For each classical cannabinoid the distance from S3837.39(Oγ) is calculated from hydroxyl oxygen (or equivalent hydrogen) bound to C1 carbon. The error bar is calculated using a bootstrapping approach with 80% of all trajectories. Three bootstrapped samples are used for the error calculations.
Figure 8—figure supplement 2
Training and validation losses during neural relational inference (NRI) network training.
The reconstruction errors in the training (solid line) and validation (dotted line) data after per epoch training of NRI network for MDMB-FUBINACA (A) and HU-210 (B) bound trajectories.
Figure 8—figure supplement 3
Neural relational inference (NRI)-based allosteric weight estimation.
Allosteric weights between the binding pocket residues and the NPxxY motif are plotted as bar plots for HU-210 (color: purple) and MDMB-FUBINACA (color: blue) bound trajectories. Allosteric weights are estimated from the posterior probability of the NRI network. Error in the allosteric weights is calculated by training the network on three different training data.
Figure 8—figure supplement 4
Mutual information-based allosteric weight estimation.
Allosteric weights between the binding pocket residues and the NPxxY motif are plotted as bar plots for HU-210 (color: purple) and MDMB-FUBINACA (color: blue) bound trajectories. Error in the allosteric weights is calculated by training the network with three different training data.
Figure 8—figure supplement 5
Probability densities of pairwise distances of residues involved in triad interaction.
Transition-based reweighting analysis ransition-based reweighting analysis (TRAM) weighted probability densities of pairwise distances between the residues involved in the triad interaction (Y3977.53, Y2945.58, T2103.46) are plotted for HU-210 (color: purple) and MDMB-FUBINACA (color: blue) unbinding ensemble.
Tables
Table 1
Asymmetric average brain membrane composition used in MD simulation.
| Head | Lipid | Upper | Lower | Total |
|---|---|---|---|---|
| Group | Group | Leaflet | Leaflet | |
| phosphatidylcholine | DPPC | 8 | 5 | 13 |
| POPC | 14 | 7 | 21 | |
| DOPC | 4 | 2 | 6 | |
| PAPC | 7 | 4 | 11 | |
| PDoPC | 1 | 0 | 1 | |
| phosphatidylethanolamine | POPE | 2 | 4 | 6 |
| PAPE | 5 | 9 | 14 | |
| PDoPE | 8 | 15 | 23 | |
| sphingolipid | SSM | 9 | 3 | 12 |
| OSM | 1 | 0 | 1 | |
| NSM | 2 | 0 | 2 | |
| phosphatidylserine | DPPS | 0 | 1 | 1 |
| POPS | 0 | 6 | 6 | |
| PAPS | 0 | 6 | 6 | |
| Glycolipid | GM1 | 2 | 0 | 2 |
| GM3 | 2 | 0 | 2 | |
| phosphatidylinositol | POPI | 0 | 5 | 5 |
| PIPI | 0 | 2 | 2 | |
| Ceramide | CER180 | 1 | 1 | 2 |
| Sterol | Cholesterol | 62 | 58 | 120 |
| Total | 128 | 128 | 256 | |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Characterization of binding kinetics and intracellular signaling of new psychoactive substances targeting cannabinoid receptor using transition-based reweighting method
eLife 13:RP98798.
https://doi.org/10.7554/eLife.98798.3
