Differential recognition of opioid analgesics by µ opioid receptors: Predicted interaction patterns correlate with ligand-specific voltage sensitivity

The µ opioid receptor (MOR) is the key target for analgesia, but the application of opioids is accompanied by several issues. There is a wide range of opioid analgesics, differing in their chemical structure and their properties in receptor activation and subsequent effects. A better understanding of ligand-receptor interactions and resulting effects is important. Here, we calculated the respective binding modes for several opioids and analyzed fingerprints of ligand-receptor interactions. We further corroborated the binding modes experimentally by cellular assays. As ligand-induced modulation of activity due to changes in membrane potential was displayed by MOR, we further analyzed the effects of voltage sensitivity of this receptor. With a combined in silico and in vitro approach, we defined discriminating interaction patterns for the ligand-specific voltage sensitivity. With this, we present new insights for interactions likely in ligand recognition and their specific effects on activation of the MOR.


Introduction 30
Opioids, agonists at the µ opioid receptor (MOR), are the most effective analgesics in clinical use.

42
However, based on the above-mentioned differences between the different opioids, it is important to 43 understand details of ligand-receptor interactions. We recently showed that ligand-induced MOR activity 44 was modulated by the membrane potential, and the effect and extent of this voltage sensitivity was 45 ligand specific (Ruland et al., 2020). Since the first report of voltage sensitivity of the muscarinic M2 46 receptor (Ben-Chaim et al., 2003), several other GPCRs have been observed to be modulated in their 47 activity depending on the membrane potential. Moreover, these effects were found to be ligand specific 48 expressed (Fig. S1A, B-G). Ligand dependence of the voltage sensitivity, mainly based on a change in 79 efficacy in receptor activation, was additionally reported previously for morphine, Met-enkephalin, 80 DAMGO and fentanyl (Ruland et al., 2020). Therefore, the MOR shows a strong ligand-specific voltage  (Ballesteros and Weinstein, 1995)). In contrast, the binding mode for methadone (Fig. 2B) indicated only 100 a salt bridge with D147 3.32 , and hydrophobic interactions and/or aromatic-aromatic stacking interactions 101 with V236 5.42 , H297 6.52 , W293 6.48 and Y326 7.43 . In contrast, fentanyl (Fig. 2C) was predicted to form only 102 one H-bond with Y326 7.43 via its amide carbonyl. Indeed, the charged amine was not predicted to form 103 an ionic interaction with the conserved D147 3.32 . In addition, N127 2.63 , D147 3.32 , C217 45.50 and H297 6.52 104 were possible interactions for fentanyl. As the D147 3.32 interaction was known to be important for opioids 105 (Surratt et al., 1994), and the formation of a salt bridge between the D147 3.32 and fentanyl was already

125
The principal component analysis therefore revealed diverse interaction patterns of the different opioid 126 ligands with MOR. The calculated fingerprint of the docked fentanyl and the best-scored fentanyl 127 conformer were nearly identical (Fig. 2F). The fingerprints of the second and third best scored 128 conformers ( Fig. S2L-M) were comparable as well. We further excluded our reference agonist DAMGO 129 from this analysis, as we were not able to calculate a reliable binding mode for DAMGO due to the high

149
To experimentally corroborate the observed docking poses, we performed site-directed mutagenesis of 150 several residues that were predicted to be important or not in the binding pocket of the MOR. These 151 mutations were first introduced in silico, and we repeated the docking for each ligand with the mutated 152 receptor, analyzed the new binding mode to make a prediction about the likely effect of the mutation on 153 binding and then evaluated the effects in functional assays in living cells. We identified Y148 3.33 as one 154 of the main interaction partners for morphine ( Fig. 2A). The replacement of this residue by F (and with 155 this, exclusion of the two water molecules from the docking calculation, as the F would prohibit this 156 interaction) led to a changed binding mode for morphine (Fig. 3A), with now only the H-bond to D147 3.32 157 left as interaction. We did not mutate D147 3.32 , as it had already been determined as the most important 158 amino acid for ligand binding and activation of MOR previously (Surratt et al., 1994). Next, we 159 determined concentration-response curves for G protein activation in single-cell FRET measurements          We were not only able to confirm binding modes of ligands by the voltage effect, we were also able to 275 do this vice versa by the prediction of the voltage effect of a certain ligand by analysis of its predicted 276 binding mode. The binding mode of etorphine (Fig. S5C) was comparable to the binding modes of 277 morphine ( Fig. 2A) and buprenorphine (Fig. S2D), and the fingerprint grouped well with these agonists, 278 which are displaying an activation upon depolarization (Fig. 5D). Based on this observation, we 279 predicted that etorphine would display a voltage sensitivity in the same direction. This was confirmed by 280 the analysis of voltage sensitivity of etorphine, revealing an increased Gαi activation upon depolarization 281 ( Fig. 5E and S5D).

282
Altered binding modes influence agonist-specific voltage sensitivity of the MOR

283
As we already showed that site-directed mutagenesis alters the predicted binding mode of the ligands,

284
we evaluated the influence of these altered binding modes on the agonist-specific voltage sensitivity of 285 the MOR to gain more information on potential molecular determinants for the voltage sensitivity.   -90 mV. We did not analyze the effect of double-mutants, as these displayed only weak and not 301 evaluable Gαi activation (Fig. S6D). Overall, although the suggested binding mode of morphine 302 changed, depolarization always increased Gαi activation to a varying extent. For methadone and 303 fentanyl, the altered predicted binding mode was consistent with the change in direction of the voltage 304 effect of methadone-or fentanyl-induced Gαi activation, which was now increasing upon depolarization 305 in some cases. Overall, the strongest effects were induced by mutation of Y148 3.33 and H297 6.52 . As 306 already shown by the fingerprint analysis (Fig. 4C), the interactions with helices 3 and 6 seem to have 307 the largest influence on voltage sensitivity, probably because of the largest voltage-induced movements 308 upon depolarization. Especially mutations leading to more space close to these helices (Y148 3.33 A,

328
Depolarization converts the antagonist naloxone to an agonist 329 Naloxone is the classical antagonist for the MOR. We analyzed the binding mode of naloxone by 330 molecular docking, and, as the chemical structure of naloxone contains the morphinan scaffold and is 331 highly related to morphine overall, we compared their predicted binding modes (Fig. 7A). Their binding 332 modes were highly comparable, as expected due to their conserved scaffold. Only the direct interaction 333 with Y148 3.33 seems to be missing for naloxone. According to the fingerprint analysis naloxone belongs 334 to the group of ligands that shows increased activation upon depolarization (Fig. 7B) (Fig. S7B), resulting in nearly identical V50 and z-values for data fitted to a Boltzmann function (Fig.   350   S7C). Further, we checked if this effect is also visible in assays that show no amplification. For this, we 351 measured the direct interaction of MOR-sYFP and arrestin3-mTur2 (Fig. S7D, see also (Ruland et al., 352 2020)) under voltage clamp conditions. In this case, naloxone induced no arrestin recruitment to the 353 receptor, neither at -90 mV nor at +45 mV. This was comparable to effects of weak partial agonists like 354 tramadol, which induce no arrestin recruitment as well (Fig. S7E). In order to further verify the observed 355 voltage-induced conversion from antagonist to agonist for naloxone, we measured MOR-evoked inward 356 GIRK currents at different holding potentials, as previously described (Ruland et al., 2020). We applied 357 naloxone and compared the evoked K + current to a saturating concentration of DAMGO (Fig. 7E) at -90 358 mV and -20 mV. The response evoked by naloxone at -90 mV was approx. 8% of the response evoked 359 by DAMGO, whereas the response at -20 mV was approx. 16% of the response evoked by DAMGO 360 (Fig. 7F), indicating a significantly increased naloxone-induced current at -20 mV. To verify that the 361 measured currents were K + currents, we applied Ba 2+ before and after every agonist or antagonist 362 application.

363
All in all, this confirms the strong agonist specific voltage effect on the MOR, which is even able to 364 convert antagonists to agonists. All the effects seem to be determined by the interaction pattern of each        However, the additional interaction with D147 3.32 , which is postulated to be crucial for fentanyl (Surratt

463
could lose affinity for the receptor due to this movement or they might impede this movement, stabilizing 464 the receptor in a more inactive state. What contradicts this hypothesis is that the Y326 7.43 F mutation, 465 which, based on the docking calculations, altered the binding mode of fentanyl so that it interacted with 466 Y148 3.33 , did not affect the voltage effect induced by fentanyl, but did change the direction of the voltage 467 effect for methadone. It is possible that fentanyl can adopt different binding modes, as described by Vo 468 et al. as well, as we also saw a possible interaction of the charged amine with D147 3.32 .

469
These results suggest that ligand-specific voltage sensitivity of MOR activation is mechanistically based 470 on the interaction patterns between ligands and the receptor. Therefore, we propose that depolarization 471 influences the conformation (or probability to reach certain conformations) of MOR in a way that 472 increases the probability to activate receptors for ligands primarily interacting with helix 6 and decreases 473 it for those ligands interacting with motif 2 in helix 3. In light of the observed ligand specific voltage 474 sensitivity also seen with other receptors, this hypothesis might well apply to those receptors as well, if 475 not to those for which voltage sensitivity has not been described yet. Our approach, strongly involving 476 the opportunities enabled by in silico methods, allows the screening of a large number of predicted 477 interactions and helps to choose the most information-rich receptor mutants and ligands for the 478 subsequent in vitro analysis in a systematic and rational way. The MOR, with its diverse voltage 479 pharmacology, was a good model system to illustrate the potential of this approach.

480
As MOR is expressed in neuronal tissue, which is highly excitable, a pharmacological relevance of 481 voltage sensitivity of the MOR is very likely, albeit difficult to proof. We have already shown that the 482 voltage sensitivity of the MOR is also reflected in brain tissue (Ruland et            showed just a slight difference between the docked fentanyl ( Fig. 2C and F) and the best scored fentanyl 820 conformer ( Fig. 2D and F).