Development of Fluorescent Probes that Target Serotonin 5-HT2B Receptors

Some 5-HT2B fluorescent probes were obtained by tagging 1-(2,5-dimethoxy-4-iodophenyl)-propan-2-amine (DOI) with a subset of fluorescent amines. Some of the resulting fluorescent ligands showed excellent affinity and selectivity profiles at the 5-HT2B receptors (e.g. 12b), while retain the agonistic functional behaviour of the model ligand (DOI). The study highlighted the most salient features of the structure-activity relationship in this series and these were substantiated by a molecular modelling study based on a receptor-driven docking model constructed on the basis of the crystal structure of the human 5-HT2B receptor. One of the fluorescent ligands developed in this work, compound 12i, specifically labelled CHO-K1 cells expressing 5-HT2B receptors and not parental CHO-K1 cells in a concentration-dependent manner. 12i enables imaging and quantification of specific 5-HT2B receptor labelling in live cells by automated fluorescence microscopy as well as quantification by measurements of fluorescence intensity using a fluorescence plate reader.

The 5-HT 2B subtype remains one of the most attractive and enigmatic receptors amongst the 5-HT receptor superfamily 12,13 , with key functional, signalling, and regulatory aspects remaining ambiguous. The involvement of this receptor in the development of migraine 14 , the modulation of the 5-HT transport system 15 , and the rewarding and reinforcing effects of the widely abused drug ecstasy (3,4-methylenedioxy-N-methylamphetamine, MDMA) has been validated 16 . 5-HT 2B receptor also participate in other relevant processes, particularly in the cardiovascular system where it regulates cardiac development and cardiomyocyte proliferation and survival 17, 18 . 5-HT 2B activation has been associated with diverse pathologies 19,20 (e.g., cardiac hypertrophy and pulmonary hypertension). It has been shown that 5-HT 2B activation, along with the inhibition of serotonin transporters, plays a significant role in the pathogenesis of serotonin-induced valvular abnormalities [20][21][22] . In line with these observations, it has been demonstrated that norfenfluramine (a metabolite of the antiobesity drug fenfluramine that exhibits potent 5-HT 2B agonist activity) and other 5-HT 2B agonists used in the treatment of migraine induce valvular heart disease [23][24][25] . Similarly, it was recently verified that the use of the antiparkinsonian dopaminergic agonists cabergolide and pergolide (both of which also exhibit high 5-HT 2B agonism) is associated with mitral, aortic, and tricuspid valvular heart disease 26,27 . Accordingly, drugs that are able to activate 5-HT 2B receptors and/ or increase circulating serotonin levels are considered to be potential valvulophatic inducers and, as a consequence, 5-HT 2B agonism is considered a dangerous off-target effect 22,[27][28][29] , thus hampering the promising therapeutic potential of 5-HT 2B receptor 28 . However, our understanding of the role of 5-HT 2B receptor and an in-depth knowledge of the processes triggered by ligand−5-HT 2B receptor interactions in living cells are still limited due to the lack of ad hoc molecular probes.
The introduction of fluorescence-based techniques has progressed the study of GPCR pharmacology to the single cell level 30 . Fluorescent molecular probes have proven to be valuable tools that offer a wealth of relevant evidence 31,32 , particularly the mapping or identification of ligand binding sites, ligand binding mechanisms, the physical nature of the binding pocket, the movement and internalization of receptors in living cells, and the localization and visualization of labelled receptors. Furthermore, fluorescent probes represent a safer, less expensive and faster alternative to radioligands. The extensive application of these techniques to GPCR research demands the development of fluorescently labelled GPCR ligands that have appropriate photochemical and pharmacological properties. A number of fluorescent probes have been described for 5-HT 1A 33 , 5-HT 3 34 and 5-HT 6 35 receptors but, to the best of our knowledge, reports concerning fluorescently-tagged 5-HT 2 ligands have not been published to date. As part of a project aimed at developing molecular probes for the study of the 5-HT 2 receptor family, we report here the development and optimization of fluorescent tools that selectively target 5-HT 2B receptors. The new ligands combine good fluorescence properties with satisfactory affinity and selectivity; accordingly, these molecular probes can contribute to a better understanding of the physiological and pathological implications of 5-HT 2B receptors.

Results and Discussion
1-(2,5-Dimethoxy-4-iodophenyl)-propan-2-amine (DOI) 36 was selected as the reference ligand to develop the fluorescent probes designed within this study. DOI is a partial agonist derived from the amphetamine chemotype and it is one of the most useful and better pharmacologically characterized molecular tools for the study of the 5-HT 2 -receptor family 37,38 . In addition, the [ 125 -I]-R-DOI radioligand is recognized as a reference standard for high throughput screening campaigns at 5-HT 2 receptors 39,40 . Recent findings have shown that DOI produces a super-potent (≈15 pM) blockade of the pro-inflammatory effects of tumor necrosis factor alpha (TNF-α) in primary aortic smooth muscle cells and animal models 41,42 . These findings, which provided new evidence about the role of 5-HT 2A receptors in inflammation, increase the demand for DOI-based fluorescent probes. For this study, it was decided to employ racemic DOI as reference, accordingly all molecular probes were synthesized and tested as racemates.
DOI is a well-recognized specific 5-HT 2 ligand that, notwithstanding, exhibits weak selectivity within the 5-HT 2 receptor family (5-HT 2A , 5-HT 2B , 5-HT 2C ) [36][37][38] . Previous reports have evidenced the introduction of functionalized alkyl chains on the methoxy group at position 2 of the phenyl ring in DOI produces an increased 5-HT 2B selectivity 43 . This structure-activity relationship trend encouraged us to explore the use of DOI-based derivatives for the development of fluorescent ligands for the 5-HT 2B receptor. A set of acid-functionalized DOI derivatives (7a-d) bearing variable spacers on the oxygen atom at position 2 was synthesized ( Fig. 1) according to previously described procedures [44][45][46][47][48] . With the aim of carrying out a preliminary evaluation of the biological repercussions of the different spacers, precursors 7a-d were transformed into the corresponding fluorescent derivatives 10a-d by reacting with a model dansyl amine (8a) and subsequent cleavage of the Boc groups (Fig. 1).
The fluorescent probes 10a-d contain the 1-(2,5-dimethoxy-4-iodophenyl)-propan-2-amine scaffold and the dansyl chromophore linked through variable spacers that differ in both length and composition (Fig. 2). The results of the pharmacological evaluation of the model fluorescently-tagged amides 10a-d at the 5-HT 2 receptor family (5-HT 2A , 5-HT 2B , 5-HT 2C ) 49,50 are presented in Fig. 2. All compounds were evaluated in binding experiments (see experimental part) and the results were compared with the biological activity determined for the reference ligand (DOI). Additionally, the functional activities (IPs accumulation) 51 of the compounds at the 5-HT 2B receptor were also determined (Fig. 2).
In agreement with previous findings 43 , it was verified that the introduction of alkyl chains on the oxygen atom at position 2 of the phenyl ring of the amphetamine scaffold produces derivatives that exhibit incipient 5-HT 2B subtype selectivity (10a-d, Fig. 2). The drop observed in the 5-HT 2A affinity is particularly noteworthy as this is in the micromolar range for all of the dansyl probes tested. Another remarkable trend identified on considering the biological data (Fig. 2) obtained for DOI derivatives bearing dansyl as a fluorescent tag (10a-d) is the importance of the spacer group linking the pharmacophore and the fluorophore. It was observed that elongation of the spacer has a negative effect on both the affinity and activity at the 5-HT 2B receptor (Fig. 2), with the fluorescent probe bearing the propyl group (10a) displaying the most attractive affinity and selectivity profiles.  In an effort to shed light on the molecular determinants for the binding of the novel fluorescent probes, we used the recent crystallized structure of the 5-HT 2B receptor (PDB ID: 4IB4) 7 to create three-dimensional models of this receptor in complex with two fluorescent derivatives with different linker lengths: compound 10a (3 carbon linker) and 10b (5 carbon linker). The resulting complexes were embedded in a physiological environment consisting of a hydrated lipid bilayer and then subjected to extended molecular dynamics simulations with a total simulation time of 1.6 µs. A common binding pose was obtained for the DOI fragment for both simulated systems and this fragment is inserted deep into to the 5-HT 2B receptor (Fig. 3, inset top panel). The protonated nitrogen of DOI establishes a strong electrostatic interaction with D3.32 whereas the aromatic ring is sandwiched in a hydrophobic environment formed by a valine in position 3.33 and two phenylalanines in positions 6.51 and 6.52. This tight binding results in a stable pose along the simulation, as reflected by low RMSD values of 0.761 Å (compound 10a) and 0.703 Å (compound 10b) with respect to the average structure (Fig. 3, lower panel). In contrast, the fluorescent tag that is linked to position 2 of the aromatic ring and reaches towards the extracellular side of the 5-HT 2B receptor shows much higher dynamic variability around the average structure, with an RMSD of 2.144 Å for ligand 10a and 3.732 Å for compound 10b. The general higher fluctuation for both fluorescent probes is due to interaction with areas of higher flexibility in the extracellular loop region of the receptor. This situation is inevitable as the fluorescent tag needs to reach sufficiently far out of the receptor in order to avoid undesired quenching interactions with the receptor, which would give rise to a non-functional fluorescence probe.
Despite this higher fluctuation, it was found that during the course of the simulation compound 10a (3 carbon linker) formed frequent polar interactions with the backbone of ECL 2 and Q7.32 via the SO 2 group (Fig. 3, top left panel), which result in a stabilization of the fluorescent tag (RMSD 2.144 Å, bottom left panel). Such polar stabilization is reduced for compound 10b because the longer linker architecture (5 carbon linker) is not optimal for allowing this polar interaction. Such reduced stabilization could be a plausible explanation for the reduced binding affinity observed for compound 10b (5 carbon linker, K i : 167.19 nM) compared to compound 10a (3 carbon linker, K i : 111.70 nM). The model developed here predicts that compounds with a longer linker have reduced stabilizing interactions in the extracellular loop regions and therefore exhibit lower binding affinity. This prediction is consistent with the other experimental data. For example, compound 10c (9 carbon linker, K i 806.04 nM) has a markedly lower binding affinity compared to compound 10a (3 carbon linker, K i : 111.70 nM) and 10b (5 carbon linker, K i 167.19 nM).
The propyl spacer group present in the most pharmacologically appealing fluorescent probe identified (10a) was fixed for subsequent fluorophore exploration. Accordingly, the carboxylic acid 7a was combined with a set of 9 propylamine-functionalized fluorescent tags (Fig. 4) by employing classical coupling experimental conditions ( Fig. 4) [44][45][46][47][48] . The selection of fluorescent amines (8) embraced not only classical fluorescent tags (e.g., 8a,b or 8i) but also several new fluorescent scaffolds (e.g., 8c-h). The employed tags (8) incorporate fluorescent groups with relatively small to large molecular volumes, thus providing a collection of probes with diverse structural and photophysical properties. Some of the non-conventional fluorophores were chosen on the basis of their complementary absorption/emission fluorescence spectra, since they could be employed as donor and acceptor pairs for the implementation of fluorescence resonance energy transfer (FRET) studies in the near future. Amines 8d-8g were prepared from the corresponding dimethyl phenanthrene-9,10-dicarboxylates (or pentaphene-6,7-dicarboxylate, for 8g), these aromatic o-diesters were synthesized by palladium-catalyzed [2 + 2 + 2] cycloaddition of arynes 52, 53 with dimethyl acetylenedicarboxylate (see supporting information). The experimental details for the syntheses of compounds 12 are outlined in Fig. 4. Briefly, the carboxylic acid 7a was reacted with different fluorescent amines (8a-i) employing DCC as a coupling reagent to give the amides 11. Finally, the protecting group (BOC) was removed by treatment with TFA to afford the fluorescent DOI-based probes 12a-i. Details on the synthesis, structural and photo-physical characterization of the obtained fluorescent probes are described in the experimental part and in Fig As for the first series described above, the pharmacological characterization of the resulting fluorescent ligands (12) was performed at two levels. Firstly, the in vitro affinity was assessed by radioligand binding assays at the three 5-HT 2 receptor subtypes. Subsequently, the functional activity at the 5-HT 2B receptors was evaluated (IPs accumulation). The data are collected in Fig. 6 and some of the representative curves obtained are shown in Fig. 7. It can be observed that all compounds stimulated IPs accumulation in a concentration-dependent manner, with EC 50 values in the nanomolar range. These data confirm that the novel ligands retain the agonist behaviour exhibited by the model ligand (DOI).
Examination of the pharmacological data contained in Fig. 6 reveals the identification of some highly attractive fluorescent ligands (e.g., 12b-f). Most of the fluorescent probes retain the incipient 5-HT 2B selectivity profile observed for the former member of the series (10a). Taking advantage of the existing X-ray structure, we performed docking followed of molecular probes from the 12 series (12a to 12i) into the 5-HT 2B receptor (PDB ID: 4IB4). All probes have overlapping docking poses (Figs 3 and 8) forming common interactions of the DOI fragment with residues D3.32, V3.33, F6.51 and F6.52 in the orthosteric binding pocket, as previously observed in our molecular dynamics simulations. In addition, we find that two residues, Q7.32 and the backbone of K211 at the extracellular loop clamp the polar part of the fluorophore tag (highlighted as red dashed lines in Fig. 8). The linker length determines the position of the fluorophore and its polar region with respect to the polar extracellular loop region. As discussed above, it is likely that longer linkers limit such favourable polar interaction. The obtained binding modes (Fig. 8) provide a potential explanation for affinity differences of studied compounds. According to our models, it appears that solvent exposure is an important factor for binding affinity towards the 5-HT 2B receptor. Comparing dyes 12a to 12h (Fig. 8) suggests that increased and unfavourable exposure of complex and hydrophobic probes that reach out into the solvent seems to be responsible for a drop of 5-HT 2B affinity. The obtained data support the benefits of introducing the acid-functionalized linker on the oxygen atom at position 2 of the phenyl ring of the model ligand (DOI), but also highlight the need for the optimized alkyl spacer (X = CH 2 −CH 2 −CH 2 ). It should be noted that most of the fluorescent probes exhibited weak affinity (typically micromolar range) at the 5-HT 2A and 5-HT 2C receptors, while the functional experiments confirmed their agonistic behaviour (Fig. 7). Moreover, a progressive drop in 5-HT 2B affinity was observed when the molecular complexity of the dye increased (e.g. cpds 12g-i). Of the probes obtained, the 7-nitrobenz-2-oxa-1,3-diazole (NBD) labelled DOI ligand (12b) warrants particular attention as it is slightly less potent (K i = 90.99 nM) than the reference ligand (DOI, K i = 48.19 nM) while exhibit an excellent selectivity (≥20-fold) for the 5-HT 2B receptor subtype. Although the NBD ligand (12b) has the most attractive pharmacological profile, the rhodamine derivative 12i showed the most promising properties of the series from the photophysical point of view.
Having finished the preliminary photophysical and pharmacological characterization of the fluorescent derivatives 12a-i, their potential to specifically label 5-HT 2B receptors in live cells was evaluated. Chinese hamster ovary K1 (CHO-K1) cells and CHO-K1 cells stably expressing the human serotonin 5-HT 2B receptor (CHO-K1-5-HT 2B ), growing in 96-well plates, were incubated with at least three different concentrations of each compound for varying times and, after compound removal and washing of the cells, plates were subjected to fluorescence microscopy using a high content imaging instrument and the appropriate excitation and emission filter set for each compound.
Unfortunately, substantial specific labelling of 5-HT 2B receptors was not observed for compounds 12a-h under any of the conditions evaluated. For these compounds, the cellular fluorescence intensities were similar in CHO-K1-5-HT 2B and in parental untransfected control CHO-K1 cells (data not shown). This non-specific labelling of cells independently of 5-HT 2B expression might arise due to non-specific membrane binding or membrane penetration of the fluorescent ligands, as observed previously for other fluorescent probes for GPCRs 54 .
While exhibiting a moderate 5-HT 2B affinity/selectivity profiles compound 12i displayed a concentration-dependent specific labelling of CHO-K1-5-HT 2B cells and this resulted in brighter fluorescence images obtained from these cells in comparison with those from control CHO-K1 cells (Fig. 9). The best results were obtained for a 3 μM concentration of compound 12i, which yielded the maximum window of specific labelling of the cells for the concentrations assayed (Fig. 9B). Under these conditions, the fluorescence intensity of the images collected from CHO-K1-5-HT 2B cell wells was clearly superior to that observed for control CHO-K1 cell wells (***p < 0.001, two-way ANOVA and Bonferroni posttests), as quantified both by automated image analysis and by direct measurement of fluorescence emission on a fluorescence plate reader (Fig. 9C,D). In the absence of compound, no substantial fluorescence signal was detected in the cells (Fig. 9A), which indicates a lack of signal specificity due to cell autofluorescence. These results support the specific binding of 12i to 5-HT 2B receptors. Image acquisition at higher magnification (40x) and analysis retrieved similar quantitative results (Supplementary Figure S1). Furthermore, confocal microscopy imaging of CHO-K1-5-HT 2B cells labelled with compound 12i in the same conditions also allowed visualization of the probe fluorescence, mainly located in intracellular compartments after 10 min incubation at 37 °C (Supplementary Figure S2).
The specificity and sensitivity of the 12i probe towards 5-HT 2B receptors over other 5-HT 2 receptor subtypes was evaluated by performing similar labelling experiments in CHO cells stably expressing 5-HT 2A receptors (CHO-FA4-5-HT 2A ). In this case, no statistically significant specific labelling was detected when compound 12i was employed at the concentration of 3 μM (Supplementary Figure S3), and it was required to increase the 12i concentration to 20 μM in order to detect certain specific labelling of 5-HT 2A receptors over background. However, even in these conditions (20 μM concentration of 12i) the signal window over background achieved for 5-HT 2A labelling was considerably smaller than that detected in the case of 5-HT 2B labelling at a concentration of 12i more than 6 times lower (3 μM) ( Fig. 9 and Supplementary Figure S3). These results were in good agreement with the affinity values of compound 12i obtained from our radioligand binding assays. Hence, compound 12i allows direct visualization of 5-HT 2B receptors stably expressed in a cell line by live fluorescence microscopy imaging and constitutes a promising molecular probe for future studies.

Conclusions
A set of fluorescent ligands based on the 1-(2,5-dimethoxy-4-iodophenyl)-propan-2-amine (DOI) chemotype has been developed. Some of the novel fluorescent probes (e.g. 12b) show excellent affinity and selectivity profiles at the 5-HT 2B receptors, while retaining the agonistic functional behaviour of DOI. The study highlighted the most salient features of the structure-activity relationship in this series and these were supported by a molecular modelling study based on a receptor-driven docking model constructed on the basis of the crystal structure of the human 5-HT 2B receptor. One of the fluorescent ligands reported here enabled the visualization of 5-HT 2B receptors in live cells. Compound 12i specifically labelled 5-HT 2B receptors stably expressed in CHO-K1 cells in a concentration-dependent manner in fluorescence microscopy studies. Hence, the probes described here are the first examples of 5-HT 2B selective fluorescent ligands and their availability should help to boost the GPCR biomolecular imaging field, thus enabling direct visualization and monitoring of spatiotemporal changes associated with 5-HT 2B -related (patho)-physiological states. The ligand might be also useful for other assays where well characterized specific fluorescent probes and appropriate signal-to-background window are required, such as BRET-based binding assays.

Methods
Chemistry. Unless otherwise stated, all starting materials, reagents and solvents were purchased and used without further purification. The reactions were monitored by thin-layer chromatography (TLC) on 2.5 mm Merck silica gel GF 254 strips, and the purified compounds each showed a single spot; unless stated otherwise, UV light and/or iodine vapour were used to detect compounds. The purity and identity of all tested compounds were established by a combination of HPLC, elemental analysis, mass spectrometry and NMR spectroscopy   ) and compound 12i at a concentration of 3 μM, both by image analysis (C) and by direct fluorescence measurement using a plate reader (D). The graphs show mean ± SEM of 4 wells, 5 fields/well (image analysis) and mean ± SEM of the same wells (plate reader). ***p < 0.001, two-way ANOVA and Bonferroni post tests. as described below. Purification of isolated products was carried out by column chromatography (Kieselgel 0.040-0.063 mm, E. Merck) or medium pressure liquid chromatography (MPLC) on a CombiFlash Companion (Teledyne ISCO) with RediSep pre-packed normal-phase silica gel (35-60 µm) columns followed by recrystallization. Melting points were determined on a Gallenkamp melting point apparatus and are uncorrected. The NMR spectra were recorded on Bruker AM300 and XM500 spectrometers. Chemical shifts are given as δ values against tetramethylsilane as internal standard and J values are given in Hz. Mass spectra were obtained on a Varian MAT-711 instrument. High-resolution mass spectra were obtained on an Autospec Micromass spectrometer. Analytical HPLC was performed on an Agilent 1100 system using an Agilent Zorbax SB-Phenyl, 2.1 mm × 150 mm, 5 µm column with gradient elution using the mobile phases (A) H 2 O containing 0.1% CF 3 COOH and (B) MeCN and a flow rate of 1 mL/min. The purity of all tested compounds was determined to be >95%. The acid functionalized DOI precursors were synthesized by following previously described methods [44][45][46][47][48] . The experimental details of the synthesis of amines 8d-g are reported in the supplementary information.

(±)
Radioligand competition binding assays at human 5-HT 2C receptors. Serotonin 5-HT 2C receptor competition binding experiments were carried out in membranes from HeLa-5-HT 2C cells prepared in our group. On the day of assay, membranes were defrosted and resuspended in binding buffer (50 mM Tris-HCl, pH 7.5). Each reaction well of a 96-well plate contained 15 μg of protein, 4 nM [ 3 H]Mesulergine (83.5 Ci/mmol, Perkin-Elmer, Waltham, MA, USA), and different concentrations of the compounds in the range from 0.01 nM to 10µM. Non-specific binding was determined in the presence of 10 μM mianserin (Sigma Aldrich, Spain). The reaction mixture was incubated at 27 °C for 1 h, after which samples were transferred to multiscreen GF/C 96-well plates (Millipore, Spain) pretreated with 0.5% polyethylenimine (PEI, Sigma Aldrich, Spain), filtered, and washed four times with 250 μL wash buffer (50 mM Tris-HCl, pH 6.6). The filters were dried, 35 μL Universol (MP Biomedicals, Spain) per well were added and radioactivity was measured as described above. Fluorescence microscopy imaging. For cell labelling studies, cells were plated onto 96-well plates (Cell Carrier; Perkin-Elmer, Waltham, MA, USA) pre-treated with poly-D-lysine hydrobromide 0.1 mg/mL (Sigma Aldrich, Spain) at a density of 10000 cells/well and cultured for 24 h. For cell labelling with compound 12i, cultured media was removed and cells were washed with HBSS supplemented with 0.1% BSA and incubated in the same buffer for 30 min at 37 °C. The buffer was removed and cells were incubated in the absence or presence of the compound at the indicated concentrations in HBSS for 10 min at 37 °C. A 10 −2 M stock solution of the compound was prepared in DMSO and then diluted to the final assay concentration in HBSS. Hoechst 33342 (1 µg/mL) (Thermo Fisher, Spain) was used to stain nuclei where indicated. After the incubation time, the supernatant was removed, cells were washed twice with HBSS and plates were subjected to microscopy. Fluorescence images were acquired using an Operetta high content imaging instrument (Perkin-Elmer, Waltham, MA, USA) at 20x and 40x magnification, at excitation wavelengths of 520-550 nm and emission wavelengths of 560-630 nm (standard 5-TAMRA filter set) for compound 12i, and excitation wavelengths of 360-400 nm and emission wavelengths of 410-480 nm for Hoechst 33342 (standard Hoechst 33342 filter set).
Quantification of fluorescence intensity. The fluorescence intensity of the cells labelled with compound 12i was quantified by image analysis of the acquired images using Harmony v3.5 software (Perkin-Elmer). Object identification was performed in the Hoechst channel followed by intensity measurement of the cell area of the objects in the 5-TAMRA channel and the results are expressed as the sum of the 5-TAMRA intensity in the cell region divided by the sum of cell area in each well. Five fields of view/well and 4 wells/condition were analyzed. Fluorescence intensity of the labeled cells was also quantified in the same plates and wells by measuring fluorescence emission (excitation wavelength 550 nm, emission wavelength 580 nm) on an Infinite M1000Pro fluorescence plate reader (TECAN, Männedorf, Switzerland). Data (mean ± SEM) obtained from both quantification methods were represented as bar graphs using Prism version 4.0 (GraphPad Software).

In Silico Methods.
Receptor preparation from the X-ray crystal structure. To simulate the interactions of 5-HT 2B with fluorescent probes, we used the recently crystallized structure of the receptor (PDB ID: 4IB4) 55 . The area of the structure, corresponding to the BRIL protein, was cut out. The protonation states were calculated at pH 7.4 using PROPKA 56 software implemented in MOE (http://www.chemcomp.com/software.htm).
Ligand Docking. First, DOI was docked into the 5-HT2B receptor defining a centroid point in the residue Asp3.32 and expanding it to 20 Å around this residue using the GOLD software 57 . One hundred genetic algorithm runs were submitted and further scored employing the ASP scoring function. The ligands were restricted to form a salt bridge between their positively charged nitrogen and the carboxylate of Asp3.32. The best poses from this docking procedure were used as inputs to explore the conformational space of the ligands with the Low Mode Search function of MOE, which is a short MD simulation that uses velocities with little kinetic energy on the high-frequency vibrational modes, using the AMBER12:EHT force field, Born solvation, 300 K and default settings. The lowest energetic pose was then used as a starting point to link the dansyl probe using MOE. In order to explore stability and the conformational space, we carried out extended molecular dynamics (MD) simulations for the 5-HT 2B receptor in complex with DOI-linked fluorescence probes.
Extended MD simulations. The protein with the ligand docked was then embedded in a 1-palmitoyl-2 -oleoyl-sn-glycero-3-phosphocholine bilayer using Charmm-gui 58 . Information obtained from the OPM database was employed to ensure proper alignment of the protein in membrane 59 . The system was solvated with TIP3 water. The ionic strength of the system was set to 0.15 M using NaCl ions. Parameters were derived from the Charmm36 forcefield. Missing parameters for the ligands were obtained from CGenFF 60 . Simulations were carried out in ACEMD 61 and the hydrogen mass-repartitioning scheme employed in ACEMD allowed us to use a 4 fs timestep 62 . The system was minimized for 3000 steps, and then subjected to 2 ns of NPT equilibration at a constant pressure of 1.01325 bar to ensure proper lipid packing in the membrane. Afterwards we proceeded with NVT simulations at a temperature of 300 K. For each complex (complex 1: 10a-5-HT 2B and complex 2: 10b-5-HT 2B ) we generated 4 individually built starting structures and simulated them for 200 ns each, yielding a total time of 800 ns per complex.