Two 5-HT2C serotonin receptor structures were selected as protein targets for the molecular docking studies, as listed in Table 1
. These are referred to as “S1” (5-HT2C in an agonist-bound active conformation; chain A of PDBID: 6BQG [22
]) and “S2” (5-HT2C in an antagonist-bound inactive conformation; chain A of PDBID: 6BQH [22
We performed docking of 50 CS ligands to S1 and S2 in order to determine the ligands which are most likely to act as agonists, antagonists or as allosteric modulators which bind outside of the canonical agonist-binding pocket. Figure 1
shows the results of the dockings of all 50 ligands, with the ligands binding to a number of clusters in the vicinity of the protein.
For both S1 and S2, the vast majority of the CS ligands are predicted to bind within the known hydrophobic pocket, as indicated by the large cluster of docked ligands in both Figure 1
a,b. This pocket corresponds to the agonist binding site. CS ligands which bind directly to this pocket may potentially act as direct agonists or competitive antagonists. We examine relative differences in the binding affinity at S1 and S2 to assign our studied CS compounds as putative agonists and antagonists as follows.
We hypothesised that ligands which show a more negative binding energy to the agonist-bound 5-HT2C conformation (S1) are more likely to act as agonists; similarly, those that are predicted to exhibit a more negative binding affinity to the known antagonist-bound form of 5-HT2C (S2) are here proposed to serve as antagonists of the receptor.
shows the difference in predicted binding affinity values for each CS ligand when docked to S1 subtracted by that of the S2. These differences are described as ∆∆G values in Table 2
, where negative values of ∆∆G indicate preferential ligand binding to S2, and thus, hypothesised to act as antagonists, whereas positive values of ∆∆G indicate preferential binding to S1 and are here hypothesised to act as agonists. In addition, ligands which are predicted to bind outside of the agonist binding site may serve as allosteric modulators of 5-HT2C, and these are discussed in subsequent sections. An alternative version of this table, with compounds sorted by S1 binding affinity, is provided in Appendix A
2.1.1. Proposed 5-HT2C Agonists
Inspection of Table 2
indicates that roughly half of all of the 50 CS ligands examined may potentially act as agonists, with positive ∆∆G values. However, most exhibit relatively minor differences. Two compounds exhibit ∆∆G of greater than +1 kcal/mol: c34 (obtusifoliol, PubChem ID 65252), with ∆∆G = +1.3 kcal/mol, and c49 (cassiaside B2 2014, PubChem ID 85185010), with ∆∆G = +1.4 kcal/mol. These are highlighted in blue in Table 2
a shows the predicted most energetically favoured hypothetical binding location of the two putative agonists, c34 and c49, at the agonist-bound conformation of 5-HT2C, indicating, as noted above, that they are predicted to bind to the canonical agonist binding pocket (active site). Furthermore, Figure 2
illustrates the 2D ligand-receptor interaction diagrams for c34 (Figure 2
a) and c49 (Figure 2
b) bound at the agonist conformation.
For c34, inspection of the contacts indicates that there are 17 residues within 4 Å distance of the ligand, with a combination of nonpolar, polar and acidic residues. Within the active site (Figure 2
a), c34 maintained a number of interactions which are similar to those of lorcaserin [15
]. There is a close contact with the highly-conserved Asp 134, for which it is known that its involvement is vital to agonist activity [23
]. There are also contacts with a number of phenylalanine residues, including Phe 214, Phe 327 and Phe 328, some of which are mediated via π–π ring-stacking interactions with the polycyclic aromatic rings of c34. These ring-stacking interactions are also supported by contacts with a cluster of surrounding nonpolar sidechains within the binding pocket, including Ile 131, Val 135, Val 208 and Leu 209, as well as close contacts with polar residues, including Ser 138, Thr 139, Asn331 and Asn 351. The ligand c34 is also predicted to be sufficiently close to potentially form hydrogen bonding interactions with another acidic residue: Glu 347. This residue is proximal to the hydroxyl group of the ligand c34. The ligand is also predicted to be in close vicinity of Tyr 118, again with the possibility of potentially forming hydrogen bonding interactions with the c34 hydroxyl group. The combination of an additional interaction with Glu 347 and possible hydrogen bonding with Tyr 118 suggests that c34 may be capable of acting as a more potent agonist than LOR and other previously studied agonists. In particular, the roles of Glu 347 in 5-HT2C interactions with ligands is not presently well-characterised and may be worthy of further study.
For the other presently proposed agonist, c49, inspection of the contacts indicates that there are 26 residues within 6 Ang of the ligand, also with a combination of nonpolar, polar and acidic residues. Within the active site (Figure 2
b), c49 maintained a number of interactions which are similar to those of c34, described above, as well as those of lorcaserin [15
]. These interactions include the key conserved residue Asp 134 and Phe 214, Phe 327 and Phe 328, as well as a cluster of surrounding nonpolar sidechains within the binding pocket, including Ile 131, Val 135, Val 208 and Leu 209. Close contacts with polar residues, including Ser 138, Thr 139, Asn331 and Asn 351, are also present, with Ser 138 also in particular being known to be important to a number of 5-HT2C agonists [27
]. In addition to the above residues, c34 is also predicted to be sufficiently close to potentially form hydrogen bonding interactions with Tyr 118 and Glu 347 via the hydroxyl group on the ligand’s aromatic ring.
In addition to the above residues, c49 is also predicted to be sufficiently close to potentially form hydrogen bonding interactions with three acidic residues: Asp 134, Glu 347 (also predicted to be present for c34) and Glu 338. The numerous hydroxyl groups available on c49 provides ample opportunity for a hydrogen-bond formation between the ligand and these acidic residues, and c49 may therefore also be a more potent agonist compared to presently known 5-HT2C-active compounds. The compound c49 also forms far more numbers of contacts with other aromatic residues compared to c34, including two tryptophan residues: Trp 130 and Trp 355, which may further indicate the possibility of c49 being an especially potent agonist. The ligand binding roles of the acidic and Trp residues described above have yet to be fully elucidated, and their involvement in the predicted binding mode of putative 5-HT2C agonists is worthy of further investigation.
2.1.2. Proposed 5-HT2C Antagonists
For the proposed antagonists, an inspection of Table 2
indicates that approximately half of the 50 CS ligands examined may potentially act as antagonists, with negative ∆∆G values. As noted above, most exhibit relatively minor differences. However, three compounds which are predicted to bind within the known ligand binding pocket exhibit ∆∆G markedly less than −1 kcal/mol: c06 (2-hydroxymethylanthraquinone, PubChem ID 87014), with ∆∆G = −1.5 kcal/mol; c29 (chrysoobtusin, PubChem ID 155381), with ∆∆G = −2.2 kcal/mol and c47 (physicion-β-d
-gentiobioside, PubChem ID 100813), with ∆∆G = −1.5 kcal/mol. These are highlighted in red in Table 2
. It is noted that c32 (beta-sitosterol, PubChem ID 222284) exhibits the most negative ∆∆G value of all, with −3.0 kcal/mol, and although this might suggest that it could be hypothesised as a strong antagonist, it is not predicted to bind within the known ligand pocket and instead binds to an external site which we propose may interfere with cholesterol binding. Thus, we excluded c32 from the present discussion regarding possible antagonist.
a shows the predicted most energetically-favoured hypothetical binding location of the three putative agonists, c06, c29 and c47, at S1, indicating, as noted above, that they are predicted to bind to the canonical agonist-binding pocket (active site). Furthermore, Figure 3
displays the 2D ligand-receptor interaction diagrams for c06 (Figure 3
a), c29 (Figure 3
b) and c47 (Figure 3
c) bound at the antagonist conformation.
The pattern of interactions for c06 shares many of the same characteristics as those for the putative agonists, described in sections above (Figure 1
a). The key characteristics are summarised as follows. Inspection of the contacts indicates that there are 11 residues within 6 Ang of the ligand with a combination of nonpolar, polar, and acidic residues, many of which are similar to those involved in agonist interactions. In particular, nonpolar interactions include π-π ring-stacking interactions with Phe 327, Phe 328 and Trp 324, all of which also exist in agonist binding. Unlike the agonists, there is additional involvement of Phe 223 and Phe 320. These ring-stacking interactions are further supplemented by contacts with a cluster of other nonpolar residues, including Val 135, Ile 142, Leu 209 and Ala 222. Hydrogen-bonding is predicted to form between the hydroxyl group of c06 with Ser 138, and Thr 139 is also sufficiently close to potentially form a hydrogen bond with the ligand.
For c29 (Figure 3
b), there are likewise 11 residues within 6 Ang of the ligand. Interactions are formed with several aromatic residues. There is a ring-stacking interaction with Phe 327, and a hydrogen bond mediated via the ligand’s hydroxyl group to Tyr 358. This is again supported by a cluster of nonpolar residues: Val 208, Leu 209, Leu 350 and Val 354. Perhaps, most importantly, in addition to the interaction with the known key residue Asp 134, there is also a close contact formed with a second acidic residue, Glu 347. These residues are also predicted for the putative agonists described in the section above.
Most interestingly, the third predicted antagonist ligand, c47, forms close contacts with 21 nearby residues within the active site (Figure 3
c), but, apart from the typical residues described above, this ligand also presents several novel interactions in addition to all of the other presently proposed agonists and antagonists. Moreover, c47 is predicted to form two hydrogen bonding interactions with Ser 334 and Gln 343. Furthermore, unique to this ligand, it is also predicted to be in close vicinity of the basic residue Lys 344. The role of this residue in the ligand binding properties of 5-HT2C is presently unknown. While much attention has focused on the role of the acidic residue Asp 134 in 5-HT2C ligand binding, the impact of the presence of basic residues, such as Lys 344, near the active site is worth further experimental characterisation.
2.1.3. Hypothetical Positive Allosteric Modulators of 5-HT2C
A small number of CS compounds are predicted to bind favourably to 5-HT2C outside of the known agonist binding pocket, all of them being present in the S1 (agonist-bound) form of the receptor with none of the ligands being predicted to bind outside the agonist site for the S2 (antagonist-bound) form of 5-HT2C. These results may suggest that such ligands may serve as allosteric modulators of the 5-HT2C receptor function, and the fact that such novel binding locations are predicted only for the agonist-bound conformation of the receptor may suggest that these compounds act as positive allosteric modulators. Referring to Figure 1
a for S1, c32 (beta-sitosterol, PubChem ID 222284) binds to the external surface of 5-HT2C at helical regions that would normally face the hydrophobic tails of the lipid bilayer in which the transmembrane region of the receptor is embedded. In particular, c32 is located within a region corresponding to the upper leaflet of the bilayer. Furthermore, another ligand, c33 (juglanin, PubChem ID 5318717), is also predicted to bind at an external site in a region corresponding to the lower leaflet of the bilayer. These predicted binding locations may suggest that these particular ligands could partition into a bilayer and subsequently interact with the receptor via these external sites.
Many G-protein-coupled receptors exhibit significant responses to membrane cholesterol with regard to the ligand-binding affinity and functional properties, including the 5-HT receptor family. Both locations described above, which we predict to bind CS compounds, suggest the possibility that both beta-sitosterol and juglanin may compete with cholesterol binding sites within the membrane. Previous work has shown the requirement of membrane cholesterol in the organization, dynamics and function of the 5-HT1A receptor [29
] and that membrane cholesterol binds preferentially to specific sites on the serotonin 1A receptor. A highly conserved cholesterol recognition/interaction amino acid consensus (CRAC) motif on transmembrane helix V was previously identified as one of the sites with high cholesterol binding capacity, lending support to the notion that it plays a role in the binding of membrane cholesterol. Similarly, another recent computational simulation study [30
] identified cholesterol interaction sites in the 5-HT1B and 5-HT2B receptors, with elevated cholesterol density predicted to lie near transmembrane helix 4. Interestingly, in that study, interactions of the endogenous agonist, serotonin with the receptor, is influenced by cholesterol binding at transmembrane helix 4, further highlighting the capability of membrane cholesterol to affect receptor function.
Although cholesterol binding has only been studied in detail, thus far, for the 1A, 1B and 2B subtypes of the 5HT receptor, our present results indicate possible binding locations for compounds c32 and c33, which are polycyclic compounds resembling cholesterol, and raise the intriguing possibility that 5-HT2C may likewise be modulated via the lipid membrane by compounds from cassia seed, perhaps involving competitive binding of these herbal compounds with cholesterol.
displays the 2D ligand-receptor interaction diagrams for c32 (Figure 4
a) and c33 (Figure 4
b) bound at the agonist conformation. For c32, inspection of the contacts (Figure 4
a) indicates that there are seven residues within 6 Ang of the ligand, and the interaction is largely mediated via contacts with nonpolar residues within the hydrophobic transmembrane region of 5-HT2C. These residues include ring-stacking contacts with Phe 220, in concert with other nonpolar residues: Leu 216, Leu 332 and Leu 336. It is noted that the hydroxyl group of c32 is predicted to form no close contacts with any of the residues, suggesting the possibility that the ligand may be aligned such that the polar hydroxyl group protrudes outwards, towards the extramembrane region. As noted in the above sections, c32 binds in a region on 5-HT2C close to those predicted to bind cholesterol in other 5HT receptor subtypes, and it is possible that this ligand could serve to displace cholesterol.
For the other presently proposed positive allosteric modular, c33 (Figure 4
b), inspection of the contacts indicates that there are 14 residues within 6 Ang of the ligand with a combination of nonpolar, polar and acidic residues. This ligand contains ample hydroxyl groups, and unsurprisingly, given that its binding position is near a region on 5-HT2C that should lie near the lipid-water interface, it is predicted to form contacts with several polar residues, including Thr 88, Asn 89, Asn 306 and Asn 372. It is also predicted to be close to both an acidic residue, Asp 151, and a basic residue, Arg 152. Its predicted close interactions, in particular with the basic residue Arg 152 close to the lipid-water interface, may suggest that c33 could competitively bind or displace anionic lipids (such as phosphotidylserines or phosphatidylinositols) that normally bind to 5-HT2C.
Both c32 and c33, then, are proposed to affect the 5-HT2C function. Both bind favourably to the predicted positions on the exterior of 5-HT2C only for the agonist-bound conformation of the receptor and are therefore proposed to be positive allosteric modulators. While c32 may also displace cholesterol, the location of c33, closer to the lipid-water interface at the inner leaflet region of the membrane, suggests it may displace anionic lipids within the bilayer in the vicinity of the 5-HT2C receptor.