The synthetic routes of 9-11 are shown in Figures 3-5 and described in detail in Methods S1 (Supporting information). Briefly, 9 was prepared starting from commercially available 2-bromo-4-methoxyphenol, epoxide 12 underwent 5-exo cyclisation to dihydrobenzofuranyl methanol 13 upon treatment with BuLi, as reported by Bradsher.[22] Introduction of the amino group in 14 was accomplished by a Mitsunobu reaction with phthalimide. This was followed by deprotection to give the free amine 15 and finally 4-bromination to yield 9.

Compounds 10 and 11 could not be prepared in the same manner as 9 because the required 6-exo/7-exo cyclisations onto the corresponding epoxides did not occur. We were thus forced to incorporate the bromine at an earlier stage to circumvent this problem. 10 was prepared as shown in Figure 4. The 7-bromochroman-4-one 16 was prepared as previously described[23] and reduced with sodium borohydride to alcohol 17. Reaction with trimethylsilyl cyanide afforded nitrile 18, which was reduced with diisobutylaluminiumhydride (DIBALH) to the amine in 10.

11 was synthesized as shown in Figure 5. Bromophenol 19[24] was alkylated using ethyl 3-bromobutyrate and Cs₂CO₃ in refluxing acetonitrile. Cyclization of the resulting acid 20 via treatment with polyphosphoric acid afforded dihydrobenzoxepinone 21, which in turn gave access to amine 11 following the same protocol utilized in the synthesis of 10: borohydride reduction, cyanation and DIBALH reduction.

Table 1 shows the binding affinities of published (1−8) and new (9−11) 2-oxygen- and/or amine-constrained phenethylamine ligands. The binding affinities of 9−11 against the 5-HT_2A-C receptors were determined in competition assays with [³H]-ketanserin, [³H]-LSD and [³H]-mesulergine as radioligands for 5-HT_2A, 5-HT_2B and 5-HT_2C, respectively. 9 and 10 have higher affinities in 5-HT_2B-C than 5-HT_2A. This was unexpected as the 5-HT_2A and 5-HT_2C affinities are typically the most similar. The highest affinity, 70 nM, is displayed by 10 in 5-HT_2B. 11 is inactive in 5-HT_2A and 5-HT_2C and displays only weak affinity (1.9 µM) for 5-HT_2B.

8, despite the 7-membered ring, appears to have somewhat higher affinity (422 nM) than 9−11. Of note however, 8 was tested in rat 5-HT_2A, in which the binding site bears more resemblance to that of 5-HT_2B-C as these three receptors contain an alanine in position 5.46 whereas human 5-HT_2A holds a more polar serine residue. Thus, until 8 has been tested in human 5-HT_2A or 5-HT_2B-C we consider it equipotent to 9−10. Also ligands 1b and 3a-b have been tested in rat receptors and may not be equipotent if tested in human receptors.

The 1,2-heterocyclized analogs 8−11 display at best a 480-fold lower affinity at 5-HT_2A than the unconstrained reference 2C-B (1a). However, pharmacological testing of 9−11 against all three 5-HT₂ subtypes, revealed significantly higher affinities for the 5-HT_2B and 5-HT_2C receptors. Below we set out to rationalize these two findings by ligand conformational analysis and, for the first time, ligand docking inside the 5-HT_2B crystal structure.[25] Specifically, the different sections have investigated ligand-receptor interactions, ligand conformational penalties of binding and the optimal positions of the 2-oxygen and amine functionalities in comparison to the highest affinity reference compounds, 4 and 5, respectively.

The 5-HT_2B receptor has been crystallized in complex with a partial agonist, ergotamine.[25] The reference ligands Bromo-DragonFLY (4) and 2C-TCB (5) could be docked directly into this crystal structure, but a small optimization of the binding pocket was needed to adapt it to the phenyl-ethylamine scaffold. The contacts for the charged amine, phenyl ring and 4-bromo functionalities were all in perfect alignment with the interaction map of the binding site. The triple-ring core of 2C-TCB (5) was better accommodated by tilting F6.52³⁴¹ slightly towards TMH5 to the same position as observed in the G protein-bound β₂ adrenergic structure.[26] The previously described[19] hydrogen bonding between the 2-oxygen functionality and S3.36¹³⁹ required rotation of the oxygen dihedral towards TMH5 until close (0.6 Å) to the most frequent state (42%) in the library. The proposed hydrogen bond[18] between the 5-oxygen and S5.43²²² hydroxyl cannot form as the oxygen atom pair distances are 5.8 and 5.7 Å for Bromo-DragonFLY (4) and 2C-TCB (5), respectively). Inspection of the crystal structure shows that the base (i.e. C-alpha to C-beta bond) of the S5.43²²² side chain projects towards TMH6 rather than TMH3 and that F6.52³⁴¹ blocks access.

Also the 1,2-cyclized compounds 8−11 could be docked directly into the 5-HT_2B structure (Figure 6a-d). Similar contacts were achieved for the charged amine and phenyl ring, whereas the 4-bromo pointed deeper and closer to A5.46²²⁵. Their 2-oxygen lone pairs are directed in opposite direction compared to the reference ligands and the optimal hydrogen bonding angle was found to be for the third rotameric state of S3.36¹³⁹ (21% frequency in rotamer library), which positions the hydroxyl deeper and just below the ligands. For compounds 8 and 9 both enantiomers fitted, although (S)-8 and (R)-9 formed more optimal receptor interactions. For 10 and 11 only the R-enantiomer fitted in a way that the 2- and 5-oxygens could be directed towards the corresponding receptor contacts. In conclusion, all compounds could be docked into the 5-HT_2B receptor. The 2-oxygen to S3.36¹³⁹ hydrogen bond could form, but required alternative rotamer shifts. A 5-oxygen to S5.43²²² hydrogen bond could not be formed. Arguably, it may form in another conformational receptor state, but it is unlikely that the helical backbones would move enough. If such as bond is formed it could however be indirect being bridged either by a water molecule or the proximal residue N6.55³⁴⁴.

We calculated the conformational energy penalties of binding for 8−11 by comparing the energies of the receptor-bound poses with their respective lowest energy conformation in solution (Fig. 6e-h). The inactive 11 exhibits a considerable penalty, 21.4 kJ/mol, whereas 8−10 display close to none. A closer inspection of 8 and 11, which both have 7-membered rings shows that their ring conformations are exactly the same in the global energy minimum, whereas their amine positions differ by 2.7 Å. In the docking, 11 displays a higher energy ring conformation, which is necessary to direct the amine in the proximity of D3.32 and also has a strained methyl amine linker. As the binding sites of 5-HT_2A-C are identical in the region around this ring, this observation provides a plausible explanation also to the lack of affinity of 11 in the 5-HT_2A and 5-HT_2C receptors.

We next investigated the orientation of the amine functionality. Figure 6i-l shows a superimposition of the minimum energy conformations of 8−11 to that of docked (R)-TCB-2 (5), which is the amine-constrained ligand with the highest affinity (5-HT_2A: 0.26 nM). In 8 the amine is slightly distanced whereas in 9−11 it is positioned closer towards the side of the interacting D3.32. The distances between the charged nitrogen atoms are 1.3, 1.5, 1.5 and 2.3 Å for 8−11, respectively, from that of 5. After docking, the distances are 0.5−3.2 Å, and the amine is shifted primarily upwards compared to TCB-2 (5). We next turned to the lone pair orientations of the 2-oxygen, which has been suggested based on mutagenesis to form a hydrogen bond with S3.36.[19] Figure 6m-p shows a superimposition of the minimum energy conformations of 8−11 to that of the docked Bromo-DragonFLY (4), which is the 2-oxygen-constrained ligand with the highest affinity (0.02−0.19 nM in 5-HT_2A-C). The distances between the 5 and 8−11 2-oxygen atoms are small (0.2 Å in 8,10−11 and 0.5 in 9). However, as expected from their 2D structures, the orientations of the lone pair vectors differ markedly. This has an effect on the docked poses (Fig. 1a-d), in which the 2-oxygen atoms of 8−11 have shifted 1.5−3.0 Å from that of Bromo-DragonFLY (4) away from THM5 and a somewhat higher.

The large changes in the amine orientations of 8−11 seem to be accommodated by the receptor as the interacting residue D3.32 offers a large contact area and there is some flexibility on both sides (one-carbon linkers in the amine and carboxylic acid). Maintaining the 2-oxygen hydrogen bond to S3.36 seems more challenging, as there is less flexibility at this point. Moreover, the amine and 2-oxygen both interact with residues on the same helix, TMH3, and a helical movement would therefore not relieve the combined constraint. The 1,2-cyclization is therefore compensated for by a translation of the ligand that shifts the positions of the methoxy, bromine and phenyl functionalities and, in particular, the 4-bromo and 5-oxygen substituents are located markedly deeper. Taken together, the constrained moieties may to some extent be compensated for by flexible receptor contact points, but alter the position and/or angle of the core scaffold and so modulate the remote 4-bromine and 5-oxygen functionalities.

9−10 display 5−12 fold higher affinities in 5-HT_2B-C than 5-HT_2A. Arguably, this difference is too large to only be due to the use of different radioligands; an antagonist for 5-HT_2A (ketanserin) and agonists for 5-HT_2B (LSD) and 5-HT_2C (mesulergine). An additional factor is the difference in the binding sites. As noted above human 5-HT_2A holds a polar serine residue in 5.46, whereas 5-HT_2B-C have an alanine (Fig. 7d-f). In our ligand docking, the 4-bromo substituent is closer to 5.46 in the 1,2-cyclized compounds 8−11. Thus, it is plausible that the lower affinity of these ligands in 5-HT_2A is caused by a less favorable environment for the 4-bromine in the presence of S5.46. In future studies it would therefore be interesting to exchange the 4-bromine for a polar substituent, for example a hydroxyl or nitrile, too see if the affinity profile is inverted (i.e. higher affinity in 5-HT_2A than 5-HT_2B-C).

5.39, located approximately two helical turns higher on TMH5, is close to the 5-oxygen of the docked phenethylamine ligands. This position is occupied by a methionine in 5-HT_2B, but an alanine in 5-HT_2A and 5-HT_2C (Fig. 7a-c). This could give more room for large ligands in the latter receptors tentatively explaining the somewhat lower affinities of Bromo-DragonFLY (4) and DOB-fly (2b) in 5-HT_2B (Table 1). Taken together, the two subtype differences could explain some of the observed affinity profiles, but many ligands have only been tested in 5-HT_2A making it difficult to define any general relationships for all constrained phenethylamines. Future ligand design aiming at selectivity, could exploit the subtype differences.

K_i determinations were generously provided by the National Institute of Mental Health's Psychoactive Drug Screening Program, Contract # HHSN-271-2008-00025-C (NIMH PDSP). The NIMH PDSP is Directed by Bryan L. Roth MD, PhD at the University of North Carolina at Chapel Hill and Project Officer Jamie Driscol at NIMH, Bethesda MD, USA.

The 5-HT_2B receptor crystal structure[25] in complex with the partial agonist ergotamine (4IB4) was downloaded from the protein data bank[27] and prepared with the Maestro protein preparation workflow[28]. A map of the interaction features and areas of the binding site was generated by SiteMap.[29] Accordingly, as a first step to adapt the binding pocket to phenylethylamine ligands, the hydroxyl hydrogen atoms of S3.36¹³⁹ and S5.43²²² were rotated towards the center of the binding site to constitute hydrogen bond donors. In opposite, the hydroxyl hydrogen of S3.37¹⁴⁰ rotated away to enlarge a hydrophobic portion of the binding site. The triple-ring core of 2C-TCB (5) was better accommodated by tilting F6.52³⁴¹ slightly towards TMH5 to the same position as observed in the G protein-bound β₂ adrenergic structure.[26] The binding sites of the 5-HT₂ receptor subtypes deviate only in two residue positions: 5.46 (5-HT_2A: S, 5-HT_2B: A and 5-HT_2C: A) and 5.39 (5-HT_2A: A, 5-HT_2B: M and 5-HT_2C: A). Models of the 5-HT_2A and 5-HT_2C binding sites were simply derived by mutation of these two residues.

Glide was used to generate the receptor grid, in which the S3.36 hydroxyl was allowed to rotate.[30] Glide SP was used for the docking, including sampling of ligand ring conformations and increasing the maximum number of output poses from 1 to 10. In a first docking round, (R)-Bromo-DragonFLY (4), (R)-2C-TCB (5) and 8−11 were docked with no further adjustments. In a second subsequent docking, separate grids were prepared for the reference and 1,2-constrained ligands, after shifting of the S3.36 rotamer to provide optimal fit (see Results). Binding poses were evaluated for placement of all interacting moieties as well as strain on the amine linker. For the reference ligands the top scoring poses proved satisfactory, whereas for 8−11 the selected poses were among the first 2−5 presented.

The lowest energy (global minimum) conformations of 8−11 were calculated using MacroModel conformational searches with exhaustive settings (maximum iterations: 5000 and convergence threshold 0.01) and applying the OPLS2005 force field (used in all calculations). The energies of the bound conformations (all poses with a docking score within one unit of the highest scoring) were also calculated with MacroModel (current energy) after a mild minimization that restricted the movement of the heavy atoms to 0.3 Å. Finally, the conformational energy penalty of binding was calculated as the energy difference between the bound and global minimum conformations.