The extracellular HXXCD/E motif may be involved in activation of some receptors.

The function of the HXXCD/E motif in extracellular loop 2 of many ORs has been debated. One study found that this sequence makes for a strong zinc-coordinating locus, which folds into an alpha helix as it binds the Zn²⁺ ion. Reasonably, that study postulated that the motif may bind zinc to the extracellular end of the OR in vivo, perhaps aiding in ligand capture or ligand binding. But the models from AlphaFold contradict this idea, in that they show the region more or less unfurled, with a disulfide bond between its highly conserved C^45.50 and the also highly conserved C^3.25 residue. Since the cryo-EM model of OR51E1 bound to propionate agrees well overall with AlphaFold, the case for a zinc-binding helix in the EXR2 loop seems doomed. But if that is not the motif’s function, then what is? The H^45.47 residue of the motif is often replaced by N, Q, or even Y, and may have an important role involving pi-stacking. But the D/E^45.51 residue is almost as well conserved as C^45.50. Additionally, many ORs have an Arg or Lys near the extracellular end of TMR6, and the Arg^6.59 of OR51E2 forms a strong ionic bond to the propionate ligand. This post presents evidence that the D/E^45.51 residue may in some ORs form a salt bridge with TMR6 in the active state, helping stabilize this configuration of the receptor.

The cryo-EM model of active OR51E2 differs from the AlphaFold model, presumed inactive, in having shifts and rotations of its transmembrane helices, as well as a slightly different arrangement of the loops between the helices. The biggest change is in TMR6, which is rotated to bring the extracellular end closer to TMR5 and the cytoplasmic end farther from the center of the protein, the latter point being consistent with the known importance of exactly that motion of TMR6 for GPCR activation in general. A terminal helix of the G protein is inserted into the space left by the outward motion of the cytoplasmic end of TMR6. The acid group of the propionate ligand stabilizes this structure by forming an ionic bond with R^6.59, while the aliphatic end of the ligand fits snugly into a pocket formed between TMR3, TMR4, TMR5, and EXR2. Noticeably, the D/E^45.51 residue is not preserved in OR51E2, being valine instead. Indeed, both OR51E2 and its close relative OR51E1 are receptors for short chain fatty acids (SCFAs), and OR51E1 also lacks an acidic group in position 45.51, having leucine there instead.

OR5A1 is a receptor for β-ionone and has D^45.51 normally. Jaeger et al (2013) found that the variant of OR5A1 with N^45.51 is insensitive to this odorant, and that individuals with two copies of the N^45.51 variant have a reduced sensitivity and a different perceptual experience of β-ionone. The authors made the assumption that a change in that part of the receptor would probably not affect activation itself, but only ligand binding. But that is a flawed assumption, since position 45.51 of an OR is not far from the extracellular end of TMR6, especially in the active configuration, and if an acid group at position 45.51 is important to some receptors’ active states, then the change from an acidic side chain to a polar uncharged residue such as Asn could raise the threshold of required binding energy and thereby exclude β-ionone as an agonist, or might even prevent the receptor from maintaining a stable active configuration at all. Unfortunately, the article does not discuss assays of any other potential ligands for OR5A1.

Here is a model showing the distance between Asp183^45.51 and Arg264^6.57 of the AlphaFold model of OR5A1. In this presumably inactive state, the two residues are too far apart to form an ionic bond, and the side chain of Arg264 is predicted to extend out into the aqueous extracellular medium.

To test the hypothesis that these two residues may be brought close enough to contact each other in the active state of the receptor, a dock calculation was executed using PrimaryDock’s new homology feature. Though not a complete homology modeling tool, it accepts a “template” model of a protein and attempts to conform the TM helices of the input protein as closely as possible to the template. The Asp183 and Arg264 side chains were then flexed as close together as possible, resulting in an ionic bond of between -55 and -57 kJ/mol, indicating that they are indeed near enough in the active configuration to make contact. These two residues were then frozen into place, and a “soft” dock was performed of β-ionone into the resulting active OR5A1 model. In a soft dock, the TM helices are allowed to move and rotate in small increments during the conformational space search in order to maximize ligand binding and minimize clashes. The relevant portion of the most energy-favorable docked pose is shown below; Asp183 and Arg264 are in contact and maintain a salt bridge with the ligand (partially visible at bottom) occupying the binding pocket.

A similar case is computed for residues Arg260^6.57 and Asp180^45.51 of OR1A1 forming a salt bridge in the active conformation as shown below docked with its strong agonist allyl phenylacetate:

PrimaryDock predicts that OR2M3 is also capable of an active-state salt bridge, in this case between Asp180^45.51 and Arg261^6.57. Activation by its ligand, 3-mercapto-2-methylpentan-1-ol, a key food odorant (KFO) with an onion aroma, has been shown by Haag et al (2019) to be enhanced by a copper ion that coordinates to one of two candidate binding sites. The team’s site directed mutagenesis and molecular docking identify C202^5.42 and C203^5.43 as important for ligand sensitivity as a binding site for copper, and T105^3.33 as an important hydrogen-bonding site for the ligand’s hydroxy group.

Our docking of the predicted active state with PrimaryDock, using C202 and C203 as metal binding residues, does show the ligand’s hydroxyl group in close proximity to Thr105, but interestingly it never makes a hydrogen bond. Instead, the most energy-favorable pose consistently has Tyr259^6.55 forming a hydrogen bond to the ligand. This happens when docking without water molecules, and the Haag et al dock does include water. The local hydrophilic environment of Thr105^3.33, Ser106^3.34, Asp155^4.56, and Asp159^4.60 might create a “wet” pocket full of water molecules that could hold the hydroxyl group in place so that Tyr259 can make the active-state hydrogen bond and cannot simply pull the ligand away from TMR3 and back into the protein’s inactive conformation.

We also investigated the enantiomers of 3-mercapto-2-methylpentan-1-ol, since the molecule has two chiral centers and therefore four possible stereoisomers. All references to the molecule on the Internet seem to describe the racemic mixture, and it can be assumed that the cell response assays most likely used the racemate. The molecular docking in Haag et al appears to be using the 2S,3R enantiomer. Our results, shown in the video below, predicted that the 2R,3R enantiomer made the strongest hydrogen bonds to Tyr259 and Thr105, while the 2S,3R isomer performed slightly more poorly; the 2R,3S enantiomer hydrogen-bonded only to Tyr259, and the 2S,3S isomer performed worst of the four. Generally, the 2R isomers placed the methyl group in the direction of Leu108, allowing the hydroxyl group access to the posited “wet pocket”, while 2S isomers placed the methyl group in the direction of Met206, excluding the hydroxy group from that hydrophilic area. According to Leffingwell, of the two enantiomers of the related 3-mercapto-2-methylpropanol, both isomers smell the same, and the odor description is similar to that of the mercapto-methylpentanol, but the 2R isomer has a stronger odor and a considerably lower detection threshold. If OR2M3 is the receptor for mercapto-methylpropanol, then all of this agrees with the results of site-directed mutagenesis showing the influence, but not necessity, of Thr105 for agonism of OR2M3.

The identification of Tyr259^6.55 as a likely binding residue for OR2M3’s agonist is really not surprising, and there is precedent for a possible role of Tyr^6.55 in OR activation. Ahmed et al (2018), using molecular docking, identify this residue in OR5AN1 as forming a hydrogen bond to the lone oxygen of isomuscone and other macrocyclic musks. Tyr^6.55 is conserved in the vast majority of Class II (“tetrapod-like”) ORs, and the tyrosine side chain is capable of both h-bonding and pi stacking. When Tyr^6.55 is not conserved, it is often replaced by another amino acid such as Asn or His that can also form both of these types of coordinations. Its location on TMR6, two turns in the extracellular direction from the fulcrum-like FYG motif, makes it a convenient “handle” that a functional group of a ligand can “hold” and maintain TMR6 in an active configuration. Even the cryo-EM model of OR51E2, a Class I (“fish-like”) OR, involves Ser258^6.55 forming a hydrogen bond with the ligand.

Not all ORs have such features as a tyrosine “handle” or an R or K near the extracellular end of TMR6. A good example is OR1G1, a broadly tuned receptor with D^45.51 and D^6.55 that would presumably repel each other by virtue of their like charges. But perhaps these two acidic side chains coordinate to the amide nitrogen of N^45.53. OR1G1 is notable for having Phe^6.48 instead of the more usual Tyr^6.48, and olfactory receptors when mutated from Y to F in this position generally exhibit a significant loss in overall sensitivity to their agonists. (de Marche et al, 2015) Further, OR1G1’s agonists include many aldehydes, ketones/lactones, and esters that would not form a strong bond to the D^6.55 side chain. The fact that OR1G1 exhibits good sensitivity to a large number of agonists shows that it must function via a mechanism where Phe^6.48, a lack of Tyr^6.55, and a lack of a TMR6-EXR2 salt bridge, are not impediments.

But for ORs that do have the potential for the extracellular salt bridge, that potential allows more accurate predictions of active configurations, since the exact rotation and location of TMR6 in the 3D model can be fine tuned to optimize the salt bridge.