2.1. MDMA Allosteric Occupation of the hSERT Overlaps the Binding Site of Escitalopram
High-performance induced fit docking of MDMA was carried out on S1 and S2 of the hSERT in an outward open conformation (PDB:5I73). At the allosteric site, this phenylethylamine adopts a conformation, perpendicular to the
z-axis (aligned to the lumen of the transporter), docking between transmembrane segments TM1, TM6 and TM10 (
Figure 1A). The simulation of the system with the reference interaction site model (3D-RISM) provided a quantitative thermodynamic description of the aqueous environment in which MDMA binds the extracellular vestibule of the hSERT.
The methylammonium moiety of MDMA simultaneously formed a salt bridge and a H-bond with the backbone and side chain of Glu494, in addition to a H-bond with a water molecule that favourably accommodates inside the binding pocket. While the benzodioxol group of MDMA forms a cation-π and a parallel π-π interaction with Arg104 and Phe335, respectively (insert in
Figure 1A). Of note, these three residues constitute key elements for the allosteric binding of escitalopram (
Figure 1B), as the substitution R104K decreases its affinity for this transporter [
17]. Likewise, mutant E494Q significantly reduced the potency of this antidepressant but did not affect that of imipramine, which also binds S2 [
18].
Glu494 is also fundamental for the binding of the high-affinity allosteric inhibitor Lu AF60097 and importantly, Phe335 is part of a motif that propagates the allosteric communication between S1 and S2 [
18]. Thus, the occupation of S2 by MDMA may be coupled to its binding on the orthosteric site and in turn may play a role in its mechanism of action on the hSERT.
Next, to validate our modelling results we tested the hypotheses of whether the 5-HT block activities of MDMA and a set of related psychoactive hSERT inhibitors could correlate with the estimated Gibbs free energies (ΔG) on either their central or their allosteric binding. We selected five benzofurans, one cathinone, methamphetamine, MDMA, and cocaine (
Figure 2) and carried out the simulations on an outward open hSERT structure.
Of note, our estimations of binding free energies go beyond classical docking scores by including entropic, enthalpic, polar ligand desolvation, and non-polar solvent contributions (see methods
Section 3.2).
2.2. Experimental-Computational Quantitative Correlation of 5-HT Reuptake Inhibitors to S1 and S2 of the hSERT
The ΔGs of the MDMA congeneric hSERT blockers, methamphetamine, and cocaine were calculated upon their molecular association to the central and the allosteric binding sites of the hSERT. In agreement with the affinity data for allosteric antidepressants [
17,
18], all compounds showed higher binding strength to S1 than to S2. The in vitro inhibitory activities of tritiated 5-HT via the hSERT, heterologously expressed in cell assays [
7] were linearly correlated to their theoretical free energy values. Of note, the correlations were possible despite the use of racemic mixtures in the experiments, while only the (R)-enantiomers were used for calculations.
In line with previous experimental findings on other MDMA analogues [
6], the occupation of S1, rather than S2 best explained the inhibitory potencies of this set of ligands (
Figure 3A,B). Of note, the computed binding modes of the selected compounds in S1 only have Glu98 in common with the binding of classical antidepressants and novel hSERT inhibitors [
21], with the exception of escitalopram that may interfere with Tyr176 and Ser336 [
22]. In particular, the orientation and modes coincide with the previously described binding mode for MDMA (on the paroxetine-induced conformation) [
8] with the addition of Ala96 and S336 side chain contributions that likely reflect the escitalopram-induced conformational rearrangement. Our proposed MDMA binding pose and binding residues were highly conserved among all compounds, even for cocaine and methamphetamine at S1 (
Figure 3C). In contrast, the binding modes for S2 were more heterogenous (
Figure 3D).
Notably, the docking solution of cocaine correctly explains why the mutant Y176C significantly decreases its affinity [
23]. In stark contrast, methamphetamine with the lowest 5-HT inhibitory potency docked between S2 and S1 (in light green in
Figure 3D). Our interpretation is that this reflects the propensities of the drugs to be internalized through the transporter eliciting the efflux of 5-HT, e.g., MDMA and methamphetamine or to merely block the neurotransmitter reuptake, e.g., cocaine.
Thereafter, we focused on characterizing the allosteric binding of (R)-MDMA on the conformational landscape of the hSERT given by experimentally determined structures, by carrying out an ensemble binding space analysis. This approach mimics the dynamic ligand-protein cooperation, accounting for protein flexibility and ligand mobility within the active site by considering the best alternative binding modes on an ensemble of protein conformations [
24].
2.3. Ensemble Binding Space Analysis of MDMA on the Allosteric Site of the hSERT: 5-HT and Escitalopram Analogous Binding Modes
The simulations were carried out first on the six recently elucidated cryo-EM structures of the hSERT, which characterize the allosteric site of 5-HT. Importantly, these structures are believed to recapitulate all fundamental states of the neurotransmitter transport cycle from holo open outward, holo occluded, over holo inward open toward apo inward open states. Holo structures are in complex with two molecules of 5-HT at S1 and at S2, [
25].
Several low-energy ‘escitalopram-like’ binding modes similar to the previously proposed in
Figure 1A were identified on the outward open and occluded holo states (PDBs:7LIA and 7MGW), i.e., the conformations in NaCl in which the substrate internalization cycle begins. (
Figure 4A). In this binding solution, it is only Glu494 that binds both 5-HT and MDMA by their cationic amines and again, the thermodynamic calculations indicate that it is probable that this moiety interacts with water (insert in discontinuous lines,
Figure 4A). Intriguingly, at this conformation, MDMA lies deeper inside the extracellular vestibule of the transporter, protruding its ring system towards the sodium (~7.6 Å apart) located in between transmembrane helixes TM1a and TM1b, which also binds Asp98, a key residue for MDMA binding to the central site [
6,
8].
The ensemble binding space of MDMA at the allosteric site was analysed and parameterized with the 80 most energetically favourable binding solutions.
Figure 4B shows the number of solutions obtained at each state and the parameters based on the calculated Gibbs free energies. Statistical metrics (mean, standard deviation, mode, and range values) indicated little ΔG disparity and reflected a heterogeneous ensemble of solutions, implying relevancy to the MDMA allosteric mechanism of action. The sensitivity score encodes the capacity of the ligand to vary its free energy values by adjusting its own conformation [
24]. Compared with the ensemble binding space reported for MDMA at the central site, the sensitivity score for S2 denotes a higher intramolecular flexibility, as the range score was double of the one previously obtained for S1 [
8]. Meanwhile, the average rmsd at S2 reflected a considerable exploration of the binding space or ‘binding mode displacement’. It is noteworthy that the parameters obtained here originate from more precise ΔG terms (see methods
Section 3.2) rather than classical enthalpic docking scores previously obtained for the central site [
8].
Additionally, in the previous work, an equal number of poses per structure was obtained, here only the highest scoring poses were considered. As a consequence, it was clear to see that MDMA preferentially binds the occluded and the outward open states, suggesting that the affinity of the drug increases as the hSERT transits from the latter to the former, being sequestered similarly to the endogenous substrate. Furthermore, although no energetically favourable poses were found for the transporter in the occluded apo state (PDB:7LI7), some were found for the inward open apo states with Na+ and K+ cations, conditions under which MDMA could initiate the reverse translocation of 5-HT.
Remarkably, the top binding solution from this analysis, identified in the occluded state, coincides with the orientation and binding mode of 5-HT. All solutions were clustered and the pose from the most populated cluster (highest number of solutions) interacted with three of the six 5-HT binding residues in this state (
Figure 4C). The reduced distribution of thermodynamically stable water molecules at S2 in the occluded state compared with the outward open state accounts for the diminished solvent accessibility. However, in the occluded state, the ammonium groups of 5-HT and MDMA can directly interact with Asp328 or via H-bond networks of waters as it is predicted by the 3D-RISM to be heavily hydrated. In contrast with Glu494, which preserves the salt bridge with these monoamines, observed in the outward open state.
In addition, both ligands bind Tyr495, albeit in a different manner i.e., 5-HT forms a H-bond with its backbone and MDMA an edge-to-face π-π with its side chain (
Figure 4C).
Moreover, the global results show that MDMA can interact with all six 5-HT binding residues in the occluded state and specify the individual bonding incidences along the whole conformational ensemble (
Figure 4D). Clearly, the aforementioned salt bridges with the two anionic residues at TM10 and TM6 are critical for MDMA allosteric binding throughout the hSERT transport cycle. The high prevalence of the amine-Glu494 bonding in the allosteric binding of MDMA to hSERT is explained by the often simultaneous interactions with the side chain and backbone atoms of this glutamate. Likewise, Thr497 may form two concomitant H-bonds, and finally alternating aromatic interactions, e.g., with Tyr495, Phe556 and Tyr579, may also play a role in the occupation of the allosteric site by MDMA. Of note, Phe556 is also involved in the recognition of escitalopram at the allosteric site [
26].
Taken together these results indicate that both the ‘escitalopram-like’ and the ‘5-HT-like’ MDMA binding modes at S2 may contribute to the allosteric mode of action of this psychoactive compound. In agreement with the literature, multiple binding modes, such as these two, can be exploited for the drug design of new allosteric compounds of the hSERT [
27].
In the pursuit of a compelling theoretical pathway for MDMA along the hSERT, we then wondered whether these two symmetrically opposed configurations of MDMA could alternate within the S2 binding pocket.
2.4. Intermediate Poses between the ‘5-HT-like’and the ‘Escitalopram-like’ Binding Modes of MDMA
Indeed, low-energy binding solutions in between the ‘5-HT-like’ and the ‘escitalopram-like’ orientations of MDMA were found in the ensemble binding space analysis. The polar cavity in which these intermediate binding modes lie (
Figure 5A,B) suggest the ligand is able to flip from one orientation to the other in situ. Most probably during the transition process from the outward open to the tightly bound occluded conformation. Since ‘the most intermediate conformations’ detected (in magenta,
Figure 5B), i.e., almost parallel to the substrate pathway (
z-axis) are exclusively present in the occluded state. The binding residues involved in such proposed reorientation from the outward open to the occluded state (at transmembrane helixes TM1b, TM6 and TM10) are shown in
Figure 5C. Of note, the allosteric residue Phe335 drastically rearranges its side chain along the transition from one state to the other (black arrow in
Figure 5C).
The following three steps provided a more thorough and dynamic characterization of the early allosteric MDMA binding events on the hSERT: (i) selection of the most energetically favourable MDMA/hSERT complexes, in the ‘5-HT-like’ and the ‘escitalopram-like’ configuration from the ensemble space; (ii) high-accuracy induced fit docking of this ligand to the central site yielding double-bound models occupying both S1 and S2; (iii) subjecting these models to stochastic energy perturbations with the MC method. These steps were followed with the aim of evaluating the stability of the MDMA-hSERT interactions quantitatively and to capture the protein rearrangements induced by occupation of the two binding sites.
Four energetically favoured poses from the ensemble binding space were chosen for the simulations, two in which MDMA is in the ‘5-HT-like’ binding mode and two in which it is in the ‘escitalopram-like’ orientation. A ‘control simulation’ was run with MDMA only at the central site to distinguish the protein intramolecular changes induced by the occupation of S2.
2.5. Monte Carlo (MC) Simulations on Double-Bound MDMA/hSERT Models to Identify Allosteric Determinants
The preservation and stability of the MDMA-hSERT intermolecular and intra-residue interactions along the MC trajectories in the lowest ΔG outward open models were assessed. To this end, the interatomic donor-acceptor, π-π or cation-π distances were monitored as a quantitative marker of bonding contacts (
Figure 6A–C and
Figure 7B,C).
Figure 6D shows the two resulting complexes after the stochastic energy jumps in the ‘5-HT-like’ configuration and
Figure 7D shows the models in the ‘escitalopram-like’ orientation.
The most frequently contacted amino acid in the full MDMA allosteric ensemble space (
Figure 4D) was Glu494 and the prevalence of its interactions with the cationic amine of MDMA was corroborated. Accordingly, in three of the four MC simulations, an intermolecular salt bridge was preserved or acquired (in red and orange,
Figure 6A), while, simultaneously in two of them, a backbone H-bond with this residue was conserved (in light and dark blue,
Figure 6A). Of note, these interactions distinctly enhanced the intracellular Glu494-Arg104 salt bridge, slightly pulling TM10 and TM1b together and briefly inducing an extracellular Glu494-Lys490 interaction. In contrast, in one case, the loss of the MDMA-Glu494 salt bridge precluded the formation of the Glu494-Arg104 bond and coincided with the optimization of the Glu494-Lys490 ionic interaction (
Figure 7B,C). Importantly, the conformational coupling between Glu494-Arg104 and Glu494-Lys490 is crucial to the allosteric inhibition of escitalopram [
26], as well as for the binding of the allosteric SSRI vilazodone [
19].
Likewise, the strengthening and stabilization of the cation-π attraction between Arg104 and the benzodioxol of the drug (
Figure 6B in blue) in the simulations with the escitalopram-like orientation, also optimizes the Glu494-Arg104 interaction compared with the control MC simulation, to the expense of losing the Glu494-Lys490 bond in one system, while it is preserved and optimized in the other with the aid of Glu493, that also binds Arg104 (
Figure 7B,C).
Aromatic bonds also prevailed along MC trajectories in both ligand orientations, the most solvent exposed ‘5-HT-like’ pose of MDMA formed a stable π-π stacking with Phe556 or Tyr495 and a short-lived π-π interaction with Tyr579, three residues involved in the allosteric binding of 5-HT (
Figure 6B,D) but not escitalopram. Subsequent energy minimization of one of the ‘5-HT-like’ systems allowed to retain these interactions while forming a cation-π tie between the methylammonium of MDMA and the phenyl ring of S2-S1 allosteric propagating residue Phe335 [
18]. Of note, this residue is also contacted by MDMA in one of the escitalopram systems albeit via a π-π contact displaced stacking, while retaining the electrostatic interactions with Ar104 that engages one of the oxygens of MDMA (
Figure 6B and
Figure 7A).
The comparison of the MC-induced conformational readjustments with those occurring in the occluded state (
Figure 5C) suggest that these interactions may not only be key to the allostery of this compound but may participate in the substrate-induced transition from the outward open to the occluded state [
18].
Lastly, the stability of the aforementioned Thr497 and particularly Gln332 H-bonds with the amine of MDMA was verified in the ‘5-HT-like’ and ‘escitalopram-like’ configurations, respectively. It is noteworthy that in the latter case, the salt bridge with Asp328, previously observed in the occluded conformation, may exist simultaneously in the outward open state (
Figure 6C and
Figure 7A). Aforementioned residues are also involved in the allosteric modulation of the hSERT, as the high affinity ligand Lu AF60097 but not escitalopram also H-bonds with the side chain of Gln332, while Thr497 caps the binding pocket of the allosteric SSRI vilazodone [
19].
In short, the MC trajectories show how the presence of MDMA at S2 in the two symmetrically opposed orientations affect the conformation of the hSERT at sites critically involved in the allosteric regulation of this transporter. In turn, these results suggest that the binding of MDMA to the allosteric site may synergistically affect that of the central site. In addition, we revealed that MDMA shares molecular allosteric features with 5-HT, antidepressants, and allosteric ligand Lu AF60097, reflecting the electrophysiological evidence of its role as a substrate and as an uptake inhibitor [
10]. Together, the ensemble binding space dockings and MC simulations imply that MDMA may constitute a unique chemotype for structure-based drug design.
We next wondered how likely it was for MDMA to navigate from the allosteric to the central binding site, in view of the small tunnel there is between them in the open outward state [
25] that, nevertheless, is proposed to be a gateway for the bulkier escitalopram to reach the orthosteric site [
26] and since some putative allosteric ligands elongate between the two sites [
28].
2.6. Ensemble Binding Space Analysis and Electrostatic Complementarity of the Pathway of MDMA from S2 to S1
Pursuing the idea of a path for MDMA between the allosteric and the central site, we first carried out a binding space analysis sampling both active sites simultaneously on a double-bound complex subjected to MC simulations. Low-energy intermediate binding poses were identified (
Figure 8A in grey thin sticks).
Figure 8B shows how often these poses involved residues from the central site (in asterisks) and from S2.
In addition to previously identified ‘allosteric residues’ Arg104, Gln332 and Phe335, the contribution of Tyr95 along the proposed S2–S1 pathway stands out. The high contact frequency of this residue is due to alternating or simultaneous cation-π contacts and backbone H-bonding with the ionic head group of MDMA. Moreover, Tyr95 may play a dual role, in the allostery of MDMA and in its transport as a substrate, since it is a binding residue for the occupation of vilazodone at S2 [
19] and its displacement is necessary for the release of 5-HT into the cytoplasm [
25]. This binding space analysis revealed that the amine of MDMA persistently interacts with Asp98 of the central site, which seems to be the driving force for the entry of MDMA into the central site.
To account for protein conformational freedom, we next carried out an ensemble binding space docking, probing a more extensive region on all 26 available hSERT structures. The preferential ‘high-affinity’ binding of this drug to S1 was corroborated and intermediate poses were detected (
Figure 8C).
On the one hand, the MC results in one of the systems with the escitalopram-like orientation suggest MDMA could access the orthosteric site adopting this orientation without visiting the ‘5-HT-like’ binding pathway (arrow in
Figure 7A). On the other hand, the occurrence of an extracellular energetically favoured binding pose found in the occluded state (insert of
Figure 8C) and some found in complexes bound to sertraline (PDBs: 6AWQ and 6AWO) reinforces the notion of the reorientation of the ligand at the vestibule of S2 (
Figure 5B and
Figure 8C). It is noteworthy that in the outward conformation of the hSERT, S2 and S1 are fused into one cavity. However, both sites are separated in the occluded state by the Tyr176-Phe335 gate [
29].
Fundamentally, it is clear from the electrostatic complementarity calculations of the models subjected to MC energy perturbations (
Figure 8D) that: (i) MDMA in an escitalopram-like orientation can traverse the narrow cavity between the allosteric and the central site, without the energetic cost of changing its orientation, perhaps driven by the protein conformational rearrangements that ensure the closing of the extracellular vestibule in the occluded state before the sequestering of the substrate [
25]. (ii) Two MDMA molecules may concurrently occupy the allosteric site, one with the ‘5-HT’ and the other with the ‘escitalopram-like’ orientations (
Figure 8D, insert). This way, the passage of one molecule from S2 to S1 is possible, while another remains bound, stabilizing allosteric changes, e.g., Glu494-Arg104 and Glu494-Lys490 coupled salt bridges [
26], and possibly facilitating substrate internalization.
Finally, to challenge our allosteric models, we screened a set of hSERT active compounds [
30] (
Figure 9A). All 5-HT releasers fitted the MDMA ‘5-HT-like’ conformation (
Figure 9B) and the ‘escitalopram-like’ binding mode of S2, albeit with statistically different ΔG mean values (
Table 1).
These results emphasise the dual allosteric binding mode of 5-HT releasers via the hSERT, suggesting the contact with 5-HT binding residues is particularly important for the reverse flow of 5-HT.