Result

The binding modes of µ-opioid receptor with morphine, fentanyl and sufentanil

The 3D model of homo sapiens active µ opioid receptor was constructed based on the crystal structure of mouse active µ opioid receptor (PDB code: 5c1m)(15 ) and shown inFigure S1A , 91.2% amino acids (270aa/296aa) located in the additional allowed regions in Ramachandran plot (Figure S1B ). The molecular docking followed by molecular dynamics simulations identified the accurate binding modes of µ opioid receptor with morphine, fentanyl and sufentanil. These three binding modes were representative structures which accounted for the largest group in cluster analysis.
As shown in Figure 1A~C, the protonated nitrogen atom of three drug molecules forms a strong salt bridge with the carboxyl oxygen atom of residue D3.32 in μ opioid receptor, though the structure of morphine is completely different from that of fentanyl and sufentanil. In addition, the role of H6.52 was identified: the epoxy group of morphine interacted with the imidazole side chain of H6.52 through the water bridge. At the same time, the phenolic hydroxyl group of morphine formed a hydrogen bond with the nitrogen atom of the imidazole ring in H6.52. In terms of the binding mode of fentanyl and sufentanil with MOR, the oxygen of amide forms the hydrogen bond network with the side chains of H6.52 through water bridges. Therefore, these two residues were selected to mutate to investigate their role in the binding and biased agonism. Another important residue W6.48 was identified to be a molecular switch in the activation of MOR. W6.48 was relatively far from the active pocket and does not have direct interaction with the ligand, while its role was identified by MD simulations in adjusting the signal transduction of G-protein or β-arrestin. Therefore, D3.32, H6.52, and W6.48 were selected to mutate and test their biased signals in order to investigate their roles in the biased mechanism.

W6.48 is a critical site influencing ligand bias at μ-opioid receptor

We developed CHO stably expressing MOR (wild-type and D3.32A, W6.48L, H6.52L mutant) cell lines to explore the role of these three sites in the interaction between ligand and μ-opioid receptor.
The D3.32A and H6.52L mutation disrupted the binding of fentanyl with MOR in radioligand binding assay (Figure 1D ), as well as the binding of the fluorescent ligand with MOR in HTRF binding assay (Figure 1E ). Both in the cAMP assay and β-arrestin recruitment assay, the effects of test agonists to stimulate D3.32A and H6.52L mutant MOR were completely or notably decreased in comparison with wild-type MOR (Figure S5C, D, G, H ). It did not contradict the results of binding assays and was in agreement with the binding mode that these two residues form hydrogen bond network with morphine, fentanyl, and sufentanil molecules. The D3.32A, and H6.52L mutations result in the breakdown of the hydrogen bond network, which interfered the ligand-receptor binding, thus led to completely loss in activity of MOR.
W6.48L mutation increased the affinity of fentanyl derivatives in binding with MOR. As shown in radioligand saturation binding curves (Figure 1D ), the affinity of [3H]fentanyl in binding with the W6.48L mutant MOR (Kd = 0.23 ± 0.03 nM) was stronger than that of the wild-type MOR (Kd = 1.07 ± 0.21 nM). Then we performed HTRF binding assay to test the affinity of a series of fentanyl derivatives in binding with MOR and determined the differences between the affinity of W6.48L mutant MOR and wild-type MOR. In HTRF saturation binding, there was no significant difference observed between the Kd value of W6.48L mutant MOR and wild-type MOR via unpaired, two-tailed t-test, which enabled the following determined Ki and Bmax values of wild-type and W6.48L mutant MOR to be statistically comparable. The competition binding curves showed that W6.48L mutation did not block the binding of test agonists with MOR (Figure 1G ). The corresponding Ki values of test agonists for wild-type and W6.48L mutant MOR were shown in Table 1 . Unlike classic MOR ligands DAMGO and morphine, fentanyl derivatives showed stronger affinity at W6.48L mutant MOR over the wild-type. Collectively, W6.48L mutation promoted the binding of fentanyl derivatives with MOR.
We carried out the commercially available cAMP assay (Figure 2A, C ) and β-arrestin recruitment assay (Figure 2B, D ) to further investigate if W6.48L mutation influenced the functional activity of fentanyl derivatives to stimulate MOR-mediated G-protein-dependent signaling and β-arrestin-dependent signaling. As shown in Table 1 , the potencies of morphine, DAMGO and fentanyl to activate cAMP assay for W6.48L mutant MOR were lower than wild-type MOR. The efficacies of fentanyl, sufentanil and remifentanil for W6.48L mutant MOR were significantly reduced versus the wild-type. It is suggested that W6.48L mutation decreased the activities of fentanyl derivatives to stimulate G-protein-dependent signaling. However, remifentanil and 3-methyfentany showed improved potency and efficacy to promote β-arrestin recruitment for W6.48L mutant MOR over the wild-type (Table 1 ), which indicated that W6.48L mutation increase the effects of fentanyl derivatives to activate β-arrestin-dependent signaling. Altogether, W6.48L mutation exerted contrary influences to G-protein-dependent signaling and β-arrestin-dependent signaling, suggesting that W6.48L changes the ligand bias of fentanyl derivatives at MOR.
To quantify the ligand bias of fentanyl derivatives, the operational model was applied to calculate the bias factor ΔΔlog(τ/KA)(21 ,22 ). We found that W6.48L mutation drove most test agonists relatively bias toward β-arrestin-dependent signaling except sufentanil (Table 2 ). It is important to note that W6.48L mutation had comparatively considerate but not identical impacts on morphine, fentanyl and sufentanil (Figure 2E ). The mutation reduced the degree of G-protein-bias for morphine, made the G-protein-bias convert to β-arrestin-bias for fentanyl, while it made the β-arrestin-bias convert to G-protein-bias for sufentanil.

W6.48 acts as a structural determinant in adjusting the 3-7 lock

Above studys show that W6.48 is a critical site affecting ligand bias at MOR, but how the W6.48L mutation changes the ligand bias and why the mutation have varying impact on the three different ligands remain unknown. The molecular docking and molecular dynamics (MD) simulations were used to investigate the biased mechanism of morphine and fentanyl derivatives to activate MOR-mediated signaling.
The MD simulations showed that, for all of these systems, the temperature, mass density and volume are relatively stable after 2 ns, the fluctuations scale became much smaller for both the RMS deviations of the Cα atoms and potential energy of all of the systems (Figure S2 ), indicating that all of the molecular systems were well behaved.
It is reported that the states of 3-7 lock, which was formed by the hydrogen bond between D3.32 and Y7.43, that impacts the activation of MOR. The distance of 3-7 lock was monitored to examine if the close or open state of 3-7 lock is relative to the activation of wild and mutation MOR (Figure 3 ). In terms of wild type MOR-drug systems, the distance in 3-7 lock of sufentanil fluctuated from 0.4 to 0.6 nm after simulation of 10 ns, which is slightly longer than that of morphine and fentanyl’s bound MORs. While in terms of drug-bound W6.48L mutant MORs, all distances among three systems tend to be smaller than 0.3 nm, i.e., the 3-7 lock is tightly closed, i.e. the strong interaction in TM3-TM7 was formed. Combining the biased signaling assay (Figure 2 ) and the time-evolved distance changes in 3-7 lock (Figure 3 ), the tight pairs were formed in morphine and fentanyl bound mutation systems. Correspondingly, the biased signalling is propitious to β-arresin; while the distance in 3-7 lock is larger than that of morphine and fentanyl’s mutation systmes, the G-protein biased signaling is enhanced. In this process, 3-7 lock was adjusted by W6.48, that stabilized Y7.43 by forming paired stable hydrophobic interaction (Figure 4 ). It was indicated that W6.48 acted as a structural determinant in adjusting the 3-7 lock that functions as a molecular switch to activate the downstream signaling.

W6.48 and Y7.43 were paired activation switches of ligand bias at μ-opioid receptor

By analyzing the processes of morphine, fentanyl and sufentanil interacting with MOR during the MD simulations as well as the above results of functional assays, we found that W6.48 and Y7.43 are the paired activation switches in the biased signal transduction. The interactions of these two residues in MOR with morphine, fentanyl and sufentanil were shown in Figure 1A-C . The distances between W6.48 and the three ligands, together with the distance between Y7.43 and the three ligands were monitored to disclose the molecular mechanism of biased agonism and elucidate the role of these two residues played in the biased activation. The time evolved distances as well as the distance distributions were shown in Figure 4A-F . Furthermore, the conformational states of the side chains in W6.48 and Y7.43 were monitored to see how these two residues adjust the biased function of MOR (Figure 6 ).
In terms of morphine-bound MOR systems, morphine is far away from Y7.43 and does not interact with it in both wild type and mutant MORs (Figure 1A ). All of the distances were kept stable during the MD simulations except the distance between W6.48 and Y7.43 in wild-type morphine-MOR complex, which changes from 0.6 nm to 1.0 nm after the simulation of 40 ns, and the distance was stable at ca. 0.9 nm upon the mutation W6.48L. Combining the functional assays, the W6.48L mutation reduced the degree of G-protein-bias for morphine, W6.48 is the key to adjust the G-protein bias.
In terms of fentanyl-bound MOR complex, the phenyl group in fentanyl was inserted into a small pocket between W6.48 and Y7.43, its phenyl group forms strong π-π interaction and this kind of interaction was kept during the whole simulations in both the wild-type and mutant MORs (Figure 1B ). The distance between W6.48 and Y7.43 is stable at ca. 1.2 nm after the simulation of 40 ns in wild type MOR system, while it became stable (ca. 0.8 nm) after the simulation of 40 ns in the mutant MOR system. Upon the mutation W6.48L MOR system, Y7.43 approached to L6.48 and formed the stable hydrophobic network with the phenyl group of fentanyl. Such stable hydrophobic network is conducive to the biased β-arrestin signal, as demonstrated by the functional assays, which showed that the biased signal was strengthened (Figure 4 H and C ).
In terms of sufentanil-bound MORs system, the phenyl group stretches into the right part of the active site in MOR since the introduction of methoxymethyl group in sufentanil results in the reorientation of phenyl group (Figure 1C ). In addition, the phenyl is far away from Y7.43 in both the wild type and mutant systems. It is quite different from that of the fentanyl bound MOR system (Figure 4B, C ). In contrast to the wild type MOR system, W6.48 approached to Y7.43 due to the formation of hydrophobic network in the W6.48L mutant system (Figure 4I, L ). The functional assays showed that the signal in β-arrestin became stronger upon the W6.48L mutation within fentanyl-bound system and the signal in G-protein pathway is much stronger than that of in β-arrestin pathway in sufentanil bound W6.48L MOR system.
All evidences indicated that the stabilization of Y7.43 by ligand and the stable interactions of ligand with W6.48 and Y7.43 are propitious to the β-arrestin signal, while the stabilization of W6.48 is propitious to biased G-protein signal. Correspondingly, both the Y7.43 and W6.48 existed in the gauch- conformational state upon fentanyl modulated β-arrestin biased signal and existed in the gauch+conformational state upon the sulfentanil modulated G-protein signal (Figure 6 ).

W6.48L-induced conformational change adjusts MOR signaling bias

Through further analyzing the states of transmembrane helix in ligand-MOR systems, we found that morphine, fentanyl, sufentanil induced the different movement of helices due to the effect of 3-7 lock and the interactions with D3.32, H6.52, W6.48 and Y7.43. In morphine bound wild system, Helix 6 (H6) moved toward outside significantly (9.5 Å) and was far away from Helix 6 (H6). In fentanyl bound W6.48L mutant MOR system, H6 moved toward H5 and it moved 2.9 Å, while in sufentanil bound W6.48L mutant MOR, H6 was far away from H5 and it moved outside for 5.9 Å (Figure 5 ). Collectively, the fentanyl induced the closure of H6 towards H5 in W6.48L mutant MOR system, correspondingly, it thus transduced the β-arrestin-bias signal. In contrast to fentanyl, sufentanil made H6 away from H5, correspondingly, it transduced the G-protein-bias signal. MD simulation showed that sufentanil induced larger conformational change on MOR than that of fentanyl. It indicated that the G-protein biased agonist can induce the larger helix movement in MOR.