Mechanistic insights into dopaminergic and serotonergic neurotransmission – concerted interactions with helices 5 and 6 drive the functional outcome

Brain functions rely on neurotransmitters that mediate communication between billions of neurons. Disruption of this communication can result in a plethora of psychiatric and neurological disorders. In this work, we combine molecular dynamics simulations, live-cell biosensor and electrophysiological assays to investigate the action of the neurotransmitter dopamine at the dopaminergic D2 receptor (D2R). The study of dopamine and closely related chemical probes reveals how neurotransmitter binding translates into the activation of distinct subsets of D2R effectors (i.e.: Gi2, GoB, Gz and β-arrestin 2). Ligand interactions with key residues in TM5 (S5.42) and TM6 (H6.55) in the D2R binding pocket yield a dopamine-like coupling signature, whereas exclusive TM5 interaction is typically linked to preferential G protein coupling (in particular GoB) over β-arrestin. Further experiments for serotonin receptors indicate that the reported molecular mechanism is shared by other monoaminergic neurotransmitter receptors. Ultimately, our study highlights how sequence variation in position 6.55 is used by nature to fine-tune β-arrestin recruitment and in turn receptor signaling and internalization of neurotransmitter receptors.


General binding mode of dopamine by unbiased molecular dynamics simulations
A first approximation of the dopamine binding mode was obtained by unbiased molecular dynamics simulation ( Figure S1). The initial pose of dopamine, was obtained using the structure of adrenaline bound to the active β2-adrenergic receptor (PDB code: 4LDO). The binding of dopamine is mediated via a salt bridge formed between the protonated nitrogen of dopamine and the highly conserved D3.32. In addition, meta and para hydroxyl groups establish polar contacts with TM5 and TM6. In particular, polar interactions with TM5 are in agreement with site directed mutagenesis [1][2][3][4][5] and computational [6] studies. Note that interactions with TM6 are mediated by a water molecule at times.
Previous site directed mutagenesis studies [7] propose that H6.  Figure S1. General binding mode of dopamine. Red dashed lines: salt bridge, red radar: polar contacts, blue surface: occupancy map of water is calculated over classical unbiased simulation using the volmap tool implemented in VMD software package. [8]

Site directed mutagenesis data
During the last decades, several mutagenesis studies have explored the impact of polar residues (S5.42, S5.43 and S5.46) in TM5 for binding of dopamine analogues. While they agree in TM5 as an important anchor point for binding of dopamine analogues, deviating results are obtained for individual residues. Here we provide a selection of references in Table S2. It is worth noting that mutation-induced alteration of ligand binding (Table   S2A) does not always correlate to an altered ligand efficacy even in the same experimental setup (summarized in Table S2B). For instance, a S5.46A mutation reduces only slightly dopamine binding while almost abolishing ligand efficacy. [2] One possible explanation is that the S5.46A mutation affects signaling in a manner independent of ligand-receptor contacts (i.e. by disturbing a conserved ionic lock within the receptor). In addition, the slight effect on the binding affinity can be induced by water-mediated indirect interaction or by a general perturbation of the hydrogen bonding network around the ligand. Table S2A. Impact of mutations of serine residues in TM5 on binding of dopamine and its analogues at the D2R. The impact of a mutation is reported as the fold change between the WT and MT. A decrease is denoted with an arrow facing downwards (↓).  [5] p-tyramine short Ki versus ['251]epidepride 2↓ insignificant [5] m-tyramine short Ki versus ['251]epidepride insignificant insignificant [5] Table S2B. Impact of mutations of serine residues in TM5 on the signaling response elicited by dopamine and its analogues at the D2R. The impact of a mutation is reported as the fold change between the WT and MT. A decrease is denoted with an arrow facing downwards (↓) and an increase with an arrow facing upwards (). Dopamine short

Single mutants
Ligand-stimulated binding of [ 35 S]GTPγS 1,03↓ 1,25↓ [4] p-tyramine short Inhibition of isoproterenol-stimulated cAMP accumulation [EC50] no significant change in respect to WT no significant change in respect to WT [5] p-tyramine short Inhibition of isoproterenol-stimulated cAMP accumulation [Emax] no significant change in respect to WT no significant change in respect to WT [5] m-tyramine short Inhibition of isoproterenol-stimulated cAMP accumulation [EC50] no significant change in respect to WT no significant change in respect to WT [5] m-tyramine short Inhibition of isoproterenol-stimulated cAMP accumulation [Emax] no significant change in respect to WT no significant change in respect to WT [5]

Metadynamics and applied bias
To exhaustively sample all potential binding modes, we biased contacts of m-and/or p-OH groups to polar residues in TM5 (S5.42 and S5.46) with a distance restraint of 12 Å. In addition, we applied a constraint to maintain the protonated nitrogen within a distance of 5 Å of the carboxylic group the D3.32. In initial experiments, we also biased the distance of aromatic substituents to S5.43. However, we did not obtain any binding peaks, therefore neglecting this interaction for following experiments.

Conformational variability of ligand binding
Typically, GPCR-ligand complexes obtained by X-ray crystallography reveal only one binding mode. In our study, enhanced molecular dynamics simulation shows that dopamine and analogues often adopt different binding modes within the orthosteric binding pocket. In particular, the energetic plot for m-tyramine ( Figure 1L in main manuscript) predicts two distinct binding modes which involve an aromatic ring rotation. The energetic barrier between both binding modes is less than 1.5 kcal/mol. This makes it possible to observe both modes in unbiased simulations ( Figure S3A). Such diversity in binding modes is not surprising and captured in several X-ray structures (Table S5). An example similar to m-tyramine is shown for the ovine cyclooxygenase-1 in complex with meloxicam (PDB 4O1Z).

Chirality-driven binding mode of (R)-and (S)-DPATs
A deeper structural analysis helps clarify why the S-enantiomer binds in an inverted position that allows only for TM5 contacts compared to the corresponding R-enantiomer of the 7-OH-DPAT ( Figure S4A-B). A common feature of ligands in aminergic GPCRs is that the protonated nitrogen faces D3.32. In this position the two Npropyl substituents are directed to TM7. We find that a steric requirement for DPAT binding is that the chiral center (red asterisk, Figure S4) of the DPAT scaffold points the smaller substituent, namely the hydrogen (highlighted in red in Figure S4) to the same direction as the bulky N-propyl substituents. This structural arrangement forces the aromatic OH group of (S)-7-OH-DPAT to extend to the bottom of the binding pocket allowing only for TM5 interaction ( Figure S4A, right). Our data shows that forming exclusively TM5 interactions is linked to G protein bias. Due to the same steric requirements, the 7-OH group of the R-enantiomer is restricted to point towards the top of the binding pocket. In turn, this binding mode promotes TM5/TM6 contacts ( Figure   S4B, right) that favors βarr2 recruitment and balanced signaling. The same observation is true for the 5-OH-DPATs. Here, the R-enantiomer points the OH-group down whereas the S-enantiomers orientates it to the top of the binding pocket due to described steric requirements. Figure S4. Chirality induces inverted binding modes for 7-OH-DPATs. Table S6. D2R MD simulations predict no ECL2 engagement for our set of compounds. Contacts have been computed over preferred binding modes extracted from metadynamics and are defined as distance < 4Å between centre of the ligand and the centre of Ile184. Preferred binding modes have been extracted according to detected energetic wells in Figure 1 and 2.

Contact heatmaps of ligand binding at energetic minimum
Contact heatmaps for ligand binding at energetic minimum provides a general overview of polar contacts between the studied ligands and the D2R ( Figure S5). As expected, we find strong interactions with D3.32 and S5.42 for all studied ligands. In contrast, polar interactions with H6.55 are preferentially observed for ligands with an unbiased dopamine-like coupling profile for G proteins and βarr2. Furthermore, the contact heatmap reveals additional contacts with residues located in TM7 which are only found for dopamine, p-and m-tyramine but not for any of the DPAT derivatives. The reason is that 5-OH-DPATs extend hydrophobic aliphatic substituents (di-propyl groups) towards TM7 which hamper the formation of polar contacts. Nevertheless, we can conclude that these polar contacts do not contribute to the coupling specificity as they are existent in compounds with a dopamine-like coupling profile (i.e. dopamine) as well as with a G protein biased profile (i.e. ptyramine). Metadynamics have been filtered for ligand binding poses that correspond to their energetic minimum. Polar contacts including direct and water-mediated interactions have been computed across the filtered simulations using the GetContacts pipeline on GPCRmd [9] . Colors correspond to the contact frequency expressed as percentage of formed contacts across simulation frames. Compounds with an unbiased dopamine-like coupling profile (highlighted with *) are typically characterized by simultaneous TM5 and TM6 interactions. Interestingly, m-tyramine (highlighted with **) co-exists in two binding modes with simultaneous TM5-TM6 (state 3) or exclusive TM5 (state 4). See main text for discussion on this topic. (B) Low energetic binding modes detected in metadynamics.  Table S3. E-H: Energetic plots of ligand binding obtained by metadynamics using as metrics the distance of the p-OH groups to S5.42 and S5.46. An energetic well at ~2.8 Å indicates a favorable distance for binding contacts with the corresponding residue. To ensure convergence of binding energetics, we monitored free energy profiles along simulation by plotting the profile every 20000 deposited Gaussians (graphs shown in different colors). I-J: Representative structures of the binding mode corresponding to the energetic wells identified in the energetic plots. K-L: Coupling ratios were approximated using the area under the curve (AUC) and its ratio for individual signaling effectors (e.g. βarr2 vs Gz, βarr2 vs GoB etc.). Note that to eliminate the observational bias linked to differences within different recruitment assays (e.g. βarr2 vs Gi), we use dopamine as internal standard for analyzing AUCs. The coupling profile of the reference compound dopamine is denoted by a coupling ratio of 1 for all pathway combinations and highlighted in all plots as a blue line. Preferential or disfavoured coupling are indicated by ratios > 1 or < 1, respectively. Dose-response curves were generated using data obtained from 3 independent experiments.

Sequence analysis across dopamine, serotonin and adrenergic receptors
A complete sequence analysis of positions involved in neurotransmitter binding for dopamine, serotonin and adrenaline receptors (26 GPCRs) provides an overview of their conservation (3.32, 5.42, 5.46 and 6.55, Figure S7A). Overall, it appears that most diversity is found in the serotonin receptor family compared to dopamine and adrenergic receptors. Across the studied set, position 3.32 and 5.42 are highly conserved whereas more differences are found in position S5.46 and 6.55. We envisage that such variation determines the coupling and signaling outcome for different subtype receptors within a family and beyond.
For instance, differences in the polarity of position 6.55 will likely impact βarr coupling and in turn receptor internalization. It is tempting to speculate that the group of a1 adrenergic receptors with a nonpolar residue (L,M6.55) is likely to have an innately reduced arrestin coupling compared to the β adrenergic receptors with a conserved polar residue in position 6.55 (N6.55). A similar tendency is expected for the 5-HT1AR (A6.55) in comparison to the group of 5-HT2, 5-HT4, 5-HT6 and 5-HT7 (N6.55). This goes along with our finding that the 5-HT1AR can be converted into a 5-HT2-like receptor with enhanced βarr coupling properties by introducing a polar residue into position 6.55 (A6.55N, Figure 5). At times, substitution can be also polar-to-polar as seen in the dopaminergic receptor family (H6.55N). Such polar-to-polar substitutions should be able to mostly preserve the key interaction with the neurotransmitter and thus receptor coupling properties which is supported by experimental validation of the D2R H6.55N mutant ( Figure 3).
Finally, also natural genetic variations of these positions will alter the functional outcome of a specific receptor. Analysis of healthy individuals (data extracted from the GPCRdb database [10,11] and the Exome Aggregation Consortium [12] ) reveal primarily rare events (frequency < 1%) ( Figure S7B) with the highest occurrence in the serotonin receptor family when compared to dopamine and adrenergic receptors ( Figure   S7B). Nevertheless, genetic variances within all three studied families seem to be well tolerated as they are observed in healthy individuals.  [10,11] and the Exome Aggregation Consortium. [12] A B 21

Experimental Procedures
Radar graph generation. Radar graphs were produced using the area under the dose response curves (AUC; calculated using GraphPad Prism 6 software). For every compound tested, the calculated pathway-specific AUCs were divided by one another (e.g., AUC GoB/AUC Gi2, AUC βarr2/AUC GoB, etc.) resulting in "relative AUCs". To eliminate the influence of system (i.e. different coupling sensitivity of partners), we use dopamine as internal standard to obtain standardized coupling ratios for all compounds (Table S7). Relative AUCs for dopamine were used as reference values to which corresponding relative AUCs for other compounds were normalized (e.g., [AUC βarr2/AUC GoB] (R)-5-OH-DPAT / [AUC βarr2/AUC GoB] Dopamine). These normalized values, referred to as "normalized relative AUCs", were plotted as radar graphs. Normalization of AUCs for mutant receptor radar graph generation was conducted using an identical approach. However, relative AUCs for the mutant receptors were always normalized to the corresponding relative AUCs for the WT receptor. For hD2R mutants (for which multiple ligands were tested), the relative AUCs for dopamine for the WT receptor were used as reference values to which all other corresponding relative AUCs were normalized. Table S7. Quantification of the coupling ratios using area under the curves (AUC) from dose-response curves. AUCs for each pathway were obtained using PRISM 7.0 (Pathway-specific AUCs). The relative coupling ratios for each compound (Relative AUCs) were calculated by dividing corresponding pathway-specific AUCs (e.g. GoB /Gi2). Then, normalized coupling ratios were obtained from relative coupling ratios using dopamine as a reference compound (Normalized Relative AUCs).  Figure S8. Schematic representation of the area under the curve (AUC) for dose response curves calculated using GraphPad Prism 6 software. AUC1 from dose response curves 1 is greater than the AUC2 from dose response curves 2 reflecting its higher potency and efficacy.