2.2. In Silico Studies
In our previous studies, we addressed mechanisms of allosteric modulation of MOR by negative and positive modulators known from literature [
22,
23,
24]. Results of these simulations were further validated by in silico and in vitro studies of independent research groups [
25,
26,
27]. Therefore, we decided to use this validated framework for investigation of putative allosteric action of compounds
1 and
2, taking advantage of considerable in-house frame of reference available.
The first step was flexible docking of investigated compounds to the extracellular part of the receptor. The docking was performed in presence and in absence of an orthosteric ligand. Morphine was chosen for this purpose, since it was used in the in vivo part of the study. Subsequently, best-scoring representatives of docking clusters of compound
1 (
Figure 4A) were used for short, duplicated 50 ns molecular dynamics (MD) simulations, in order to assess stability of the poses. In most cases, ligand molecules drifted to the same region of the receptor (
Figure 4B). In particular, final poses of all simulations were located in direct proximity of TM2, and most of them (4 of 6 short simulations) showed the investigated compounds being anchored by the phenyl moiety interacting with the hydrophobic pocket composed of Trp 135, Val 3.28, Ile 3.29 and cysteine bridge between Cys 3.25 ad Cys 219, below the extracellular loop 2 (ECL2), and ketone groups interacting with arginine or aspartate from ECL2, either directly or through water molecules (
Figure 4C). Since these poses were also energetically favorable, they were chosen for longer simulations. Docking results for compound
2 were less diverse and mostly consisted of poses located below ECL2, in agreement with MD results for compound
1. Therefore, the best scored pose was used in further steps.
In the next step, 400 ns simulations were performed in various ligand configurations. To provide appropriate reference, we performed simulations of MOR in complex with an orthosteric partial agonist (morphine) and an orthosteric antagonist (naloxone). Additionally, trajectories of MOR in complex with a full agonist ((R)-methadone) from the previous study [
24] were added to analysis for comparison. Complexes of MOR with compounds
1 and
2 were simulated in presence of morphine in the orthosteric site. Additionally, due to interesting preliminary in silico results and relatively low doses needed for nociceptive activity in vivo, additional simulations of compound
1 in complex with MOR without any orthosteric compounds were performed. Visual analysis performed on trajectories revealed both similarities and differences in behavior of compound
1 and compound
2 in the allosteric binding pocket. Both compounds established frequent direct contacts with Tyr 2.64 through the heterocyclic rings, and preferred location of the phenyl moiety was the pocket between Trp 135, Val 3.28, Ile 3.29. However, chlorine substituent seems to prevent the phenyl moiety from stable binding in the hydrophobic aromatic pocket under ECL2. The compound
1 is therefore more shallowly anchored. Visual analysis of trajectories shows increased motility of the compound
1 in the allosteric pocket, especially regarding the heterocyclic part. To confirm this observation, we measured the ligand root mean square fluctuation (RMSF), with 7TM bundle Cα atoms as fitting reference. RMSF values, presented in
Table S1, confirm higher motility of compound
1 in presence of morphine in the orthosteric pocket, when compared to values obtained for compound
2. To further investigate this phenomenon, we measured root mean square deviation (RMSD) values for both investigated compounds, calculated between a given pose and a pose 500 ps before in trajectory. The evolution of RMSD in time reveals several peaks corresponding to changes in compound 1 conformation, as shown in
Table S2.
Principal component analysis (PCA) was used to sift relevant relationships. Trajectories of 7-transmembrane (7TM) domain obtained from simulations were extracted and concatenated, so that PCA of all simulations was done in the same conformational space. Several PCA rounds were performed, for the whole 7TM domain and for particular helices separately. Heavy atoms of all residues of 7TM bundle except the hydrophobic residues located at the protein-membrane interface were considered in the analysis, including sidechains. Sidechains of residues located at the protein-membrane interface were ignored to remove noise. PCA was performed on whole trajectories. To clarify presentation of the results, conformational space explored through 50 last ns only is presented in figures.
PCA calculations made for the transmembrane helix 2 (TM2) and 7 (TM7) yielded interpretable results. As shown in
Figure 5, presence of compound
1 or compound
2 affects conformation of TM2. In all complexes without investigated compounds bound, area of the low first principal component (PC1) values was occupied. On the other hand, introduction of these compounds in several cases resulted in exploration of completely new conformational space. Comparison of projections of extreme values of PC1 on trajectories revealed that the main difference is conformation of Tyr 2.64 (Ballesteros-Weinstein notation [
28]). While in absence of any putative modulators this residue tends to protrude outwards of the receptor, in their presence it frequently points toward protein interior.
Results of PCA calculations for TM7 are presented in
Figure 6. In this case, relationships could be observed across both PC1 and PC2. Complexes containing investigated compounds tend to assume conformations described by extreme values of PC1, and PC2 values around zero. The same tendency is observed for two of three systems containing full agonist, (R)-methadone. Meanwhile, complexes of MOR with morphine or naloxone assumed PC1 values near zero, and a broad range of PC2 values, with naloxone prevailing in above-zero region, and morphine in below-zero area. Projections of extreme values of these principal components on trajectories revealed that these PCs mostly describe TM7 bending and rotation, as well as Trp 7.35 and Tyr 7.53 conformation, which is in line with our previous simulations of MOR allosteric modulators [
23,
24]. Interestingly, most simulations of compound
1 bound alone (without orthosteric ligands) tend to occupy the same conformational space as simulations of full agonist, which remains consistent with relatively low doses required for antinociceptive effect in vivo.
To obtain a more general view on statistics of ligand-dependent conformational changes, PCA for TM2 and TM7 together was calculated (
Figure 7). Topology of conformational space described by first two PCs exhibits apparent similarity to PCA calculated for TM7 alone. Projections of extreme values of these PCs on trajectories show that the most pronounced conformational changes involve Tyr 2.64 and Tyr 7.53 residues.
Another relationship was observed in PCA calculated for TM6 alone (
Figure 8). Most complexes containing compound
1 or compound
2 assumed a well-defined conformation in terms of PC1 and PC2. Interestingly, the values of PC2 characteristic for these complexes correspond to values obtained for two (R)-methadone-containing complexes.
PCA was calculated on covariance matrix of Cartesian coordinates. To check how the statistical analysis corresponds to actual conformations in the trajectories, angles and dihedrals pointed by PCA were measured in unmodified trajectories. The most apparent relationship between binding of compounds
1 or
2 and the receptor conformation was observed in case of χ
1 dihedral of Tyr 2.64 residue (
Figure 9,
Table S3) and rotation of intracellular part of TM7 (
Figure 10,
Table S4). In the former, in the absence of any modulator, χ
1 of the residue oscillates around −180° for most of the simulation time, while in presence of compound
1 it is shifted to around −60° in two of three simulations. Moreover, compound
1 was able to induce such change and keep it stable even in absence of morphine in one of three simulations. This effect occurs probably due to direct interactions of the compounds with Tyr 2.64 residue, as they remain in spatial proximity as depicted in
Figure 4. In case of TM7 rotation, interpretation of the results is less straightforward, but careful analysis can lead to some conclusions. Apparently, compound
2 altered behavior of morphine and made it very similar to (R)-methadone, especially considering particular random seed series. Compound
1 alone also shifted TM7 rotation to values more characteristic to the full agonist. In turn, compound
1 bound together with morphine made TM7 assume ‘intermediate’ conformation, with angle values between those seen in full agonist- and antagonist-bound receptor. Hypothetically, as TM2 and TM7 are adjacent, ligand interactions with Tyr 2.64 could be the trigger of the allosteric signal migrating downwards to the intracellular part of TM7, affecting activation.
In PCA, residue Tyr 7.53 also seemed to be affected, similarly as in our previous studies on MOR modulation, so its dihedrals were also measured (
Figure S1 in Supplementary Materials). Its conformation has shown that compound
1 is able to induce agonist-bound-like changes in the receptor, even in absence of orthosteric ligands, while in presence of morphine, compound
2 provides more stable arrangement of the residue.