X-ray structures of human ETB receptor provide mechanistic insight into receptor activation and partial activation

Endothelin receptors (ETA and ETB) are class A GPCRs activated by vasoactive peptide endothelins, and are involved in blood pressure regulation. ETB-selective signaling induces vasorelaxation, and thus selective ETB agonists are expected to be utilized for improved anti-tumour drug delivery and neuroprotection. The effectiveness of a highly ETB-selective endothelin analogue, IRL1620, has been investigated in clinical trials. Here, we report the crystal structures of human ETB receptor in complex with ETB-selective agonists, endohelin-3 and IRL1620. The 2.0 Å-resolution structure of the endothelin-3-bound receptor revealed that the disruption of water-mediated interactions between W6.48 and D2.50, which are highly conserved among class A GPCRs, is critical for receptor activation. These hydrogen-bonding interactions are partially preserved in the IRL1620-bound structure, and a functional analysis revealed the partial agonistic effect of IRL1620. The current findings clarify the detailed molecular mechanism for the coupling between the orthosteric pocket and the G-protein binding, and the partial agonistic effect of IRL1620, thus paving the way for the design of improved agonistic drugs targeting ETB.


Introduction
Endothelin receptors belong to the class A GPCRs, and are activated by endothelins, which are 21-amino acid peptide agonists 1 . Both of the endothelin receptors (the ETA and ETB receptors) are widely expressed in the human body, including the vascular endothelium, brain, lung, kidney, and other circulatory organs 2,3 .
Three kinds of endothelins (ET-1, ET-2, and ET-3) activate the endothelin receptors (ETRs) with sub-nanomolar affinities. ET-1 and ET-2 show similar affinities to both of the endothelin receptors, while ET-3 shows two orders of magnitude lower affinity to ETA 4-6 . The stimulation of the ETA receptor by ET-1 leads to potent and long-lasting vasoconstriction, whereas that of the ETB receptor induces nitric oxide-mediated vasorelaxation 7-9 . The human brain contains the highest density of endothelin receptors, with the ETB receptor comprising about 90% in areas such as the cerebral cortex 10 . The ETB receptor in neurons and astrocytes has been implicated in the promotion of neuroprotection, including neuronal survival and reduced apoptosis 11,12 . Moreover, the ET-3/ETB signaling pathway has distinct physiological roles, as compared to the ET-1 pathway. In the brain, ET-3 is responsible for salt homeostasis, by enhancing the sensitivity of the brain sodium-level sensor Nax channel 13 . The ET-3/ETB signaling pathway is also related to the development of neural crest cells, and plays an essential role in the formation of the enteric nervous system 14 . Thus, mutations of the ET-3 or ETB genes cause Hirschsprung's disease, a birth defect in which nerves are missing from parts of the intestine 15,16 . Overall, the endothelin system participates in a wide range of physiological functions in the human body.
Since the activation of the ETB receptor has a vasodilating effect, unlike the ETA -4 -receptor, ETB-selective agonists have been studied as vasodilator drugs for the improvement of tumour drug delivery, as well as for the treatment of hypertension 2,3 . IRL1620 (N-Suc-[E9, A11, 15] ET-18-21) 17 , a truncated peptide analogue of ET-1, is the smallest agonist that can selectively stimulate the ETB receptor, and currently no non-peptidic ETB-selective agonists have been developed. The affinity of IRL1620 to the ETB receptor is comparable to that of ET-1, whereas it essentially does not activate the ETA receptor, and thus it shows high ETB selectivity of over 100,000-fold. Due to its large molecular weight, IRL1620 is not orally active and thus requires intravenous delivery. Despite its pharmacokinetic disadvantages, IRL1620 is an attractive candidate for the treatment of various diseases related to the ETB receptor. Since the ETB-selective signal improves blood flow, IRL1620 could be utilized for the improved efficacy of anti-cancer drugs by increasing the efficiency of drug delivery, as shown in rat models of prostate and breast cancer [18][19][20][21] . Moreover, this strategy can also be applied to radiotherapy in the treatment of solid tumours, as the radiation-induced reduction in the tumour volume was enhanced by IRL1620 22 . IRL1620 also has vasodilation and neuroprotection effects in the brain. IRL1620 reduced neurological damage following permanent middle cerebral artery occlusion in a rat model of focal ischaemic stroke 23 .
Moreover, the stimulation of the ETB receptors by IRL1620 reduces the cognitive impairment induced by beta amyloid (1-40), a pathological hallmark of Alzheimer's disease, in rat experiments 24,25 . These data suggest that ETB selective agonists might offer new therapeutic strategies for neuroprotection and Alzheimer's disease. The safety and maximal dose of IRL1620 were investigated in a phase I study. While a recent phase 2 study of IRL1620 in combination with docetaxel as the second-line drug -5 -reported no significant improvement in the treatment of advanced biliary tract cancer (ABTC) 26 , further trials for selected patients based on tumour types with various choices for the second-line drugs are still expected. Concurrently, the pharmacological properties of IRL1620 could be also improved for better clinical applications. However, little is known about the selectivity and activation mechanism of this artificially designed agonist peptide, although the ETB structures in complex with ET-1 and antagonists have been determined 27,28 .
In this study, we report the crystal structures of the ETB receptor in complex with two ETB-selective ET variant agonists, ET-3 and IRL1620, together with their detailed biochemical characterization. These results reveal the different activation mechanisms of these agonists, especially for the partial activation by IRL1620.

Functional characterization of ET-3 and IRL1620
We first investigated the biochemical activities of ET-3 and IRL1620 for the human endothelin receptors, by TGFα shedding (G-protein activation) and β-arrestin recruitment assays. The EC50 and Emax values of ET-3 for the ETB receptor were similar to those of ET-1 in both assays, while the EC50 value for ETA was about 5-fold lower . These data indicate that ET-3 functions as a full agonist for the endothelin receptors, with moderate ETB-selectivity. The EC50 value of IRL1620 for the ETB receptor was 0.11 nM in the TGF shedding assay, and was almost same as that of ET-1 ( Fig. 1b). By contrast, a 320 nM concentration of IRL1620 did not activate ETA in the TGF shedding assay (Fig. 1a). These data showed that IRL1620 is ETB-selective by over 3,000-fold, in excellent agreement with previous functional analyses 17,29 . However, despite its sub-nanomolar affinity, the Emax values of IRL1620 for the ETB receptor were 88% (TGFα shedding assay) and 87% (β-arrestin recruitment assay) of those of d), suggesting that IRL1620 functions as a partial agonist for the ETB receptor.
To obtain mechanistic insights into the different actions of these agonists, we performed X-ray crystal structural analyses of the human ETB receptor in complex with ET-3 and IRL1620. For crystallization, we used the previously established, thermostabilized ETB receptor (ETB-Y5) 27,30 . We confirmed that the thermostabilizing mutations minimally affect the binding of these agonists. The EC50 and Emax values of ET-3 and IRL1620 for ETB-Y5 were comparable to those for the wild type receptor in both assays (Fig. 1b, d). In addition, IRL1620 also functions as a partial agonist for the -7 -thermostabilized receptor, as the Emax values for ETB-Y5 were lower than those of ET-1 in both assays (84% and 85% in the TGF shedding assay and the -arrestin recruitment assay, respectively). To facilitate crystallization, we replaced the third intracellular loop (ICL3) of the receptor with T4 Lysozyme (ETB-Y5-T4L), and using in meso crystallization, we obtained crystals of ETB-Y5-T4L in complex with ET-3 and IRL1620 ( Supplementary Fig. 1a, b). In total, 757 and 68 datasets were collected for the ET-3-and IRL1620-bound receptors, respectively, and merged by a data processing system KAMO 31 . Eventually, we determined the ETB structures in complex with ET-3 and IRL1620 at 2.0 and 2.7 Å resolutions, respectively, by molecular replacement using the ET-1-bound receptor (PDB 5GLH) ( Table 1). The datasets for the ET-3 bound receptor were mainly collected with an automated data-collection system, ZOO (K.Y., G.U., K.H., M.Y., and K.H., submitted). This system allowed the convenient collection of a large number of datasets and the determination of the highest-resolution agonist-bound GPCR structures. The electron densities for the agonists in both structures were clearly observed in the Fo − Fc omit maps ( Supplementary Fig. 1c, d).

ETB structure in complex with the full agonist ET-3
We first describe the ETB structure in complex with ET-3. The overall structure consists of the canonical 7 transmembrane helices (TM), the amphipathic helix 8 at the C-terminus (H8), two antiparallel β-strands in the extracellular loop 2 (ECL2), and the N-terminus that is anchored to TM7 by a disulfide bond (Fig. 2a, b), and is similar to the previous ET-1-bound structure 27 (overall R.M.S.D of 1.0 Å for the Cα atoms) (Fig. 2c).
Accordingly, these regions form similar interactions with the receptor in both structures ( Supplementary Fig. 3a, b). In contrast, all of the residues, except for the disulfide bond-forming C1 and C3, are replaced with bulkier residues in ET-3 (Fig. 2a). Despite these sequence differences, the N-terminal regions are similarly accommodated in the orthosteric pocket in both structures, because these bulky residues are exposed to the solvent and poorly interact with the receptor (Fig. 2d). These structural features explain the similar high affinity binding of ET-3 to the ETB receptor, as compared with ET-1.
Previous studies demonstrated that the N-terminal residues of the ETB receptor play a critical role in the virtually irreversible binding of the endothelins 32 . As in the ET-1-bound structure, the N-terminal tail is anchored to TM7 via a disulfide bond between C90 and C358 in the ET-3-bound structure, constituting a lid that prevents agonist dissociation. The high-resolution ET-3-bound structure allowed more accurate tracing of the elongated N-terminal residues (Fig. 2e, f, Supplementary Fig. 1e), as compared with the ET-1-bound structure and revealed more extensive interactions with the agonist peptide. P88, I94, Y247 ECL2 , and K248 ECL2 form a lid over ET-3, which is stabilized by a water-mediated hydrogen bonding network among the carbonyl oxygen of P93, the side chains of Y247 ECL2 and K248 ECL2 , and D8 of ET-3 (Fig. 2e). In addition, three consecutive prolines (P87, 88, 89) stretch over the N-terminal region of ET-3.
-9 -Moreover, ECL1, 2 and the N-terminal residues form an extended water-mediated hydrogen bonding network over ET-3. These extensive interactions strongly prevent the agonist dissociation.
ETB structure in complex with the partial agonist IRL1620 Next we describe the ETB structure in complex with the partial agonist IRL1620, a linear peptide analogue of ET-1 17 (Fig. 3a). Previous mutant and structural studies revealed that the N-terminal region contributes to the stability of the overall bicyclic structure by the intramolecular disulfide bonds, and thus facilitates the receptor interaction 17,27,33 . IRL1620 completely lacks the N-terminal region, and consists of only the α-helical and C-terminal regions (Fig. 3b). Two cysteines in the α-helical region are replaced with alanines, and negative charges are introduced into the N-terminal end of the helix, by replacing lysine with glutamic acid (E9) and modifying the N-terminal amide group with a succinyl group (Fig. 3c). The consequent cluster of negative charges on the N-terminal end of IRL1620 (succinyl group, D8, E9, and E10) reportedly plays an essential role in IRL1620 binding to the ETB receptor 17 . This cluster electrostatically complements the positively charged ETB receptor pocket, which includes K346 6.58 and R357 ECL3 (Fig. 3d). Moreover, these negative charges neutralize the N-terminal dipole moment of the α-helical region of IRL1620, and thus probably contribute to the stability of the α-helical fold 34 . Due to these interactions, IRL1620 adopts a similar helical conformation, even without the intramolecular disulfide bonds ( Supplementary Fig. 2c). IRL1620 thus forms essentially similar interactions with the receptor, as compared with the endogenous agonists, ET-1 and ET-3 ( Supplementary Fig. 3a-c). These structural -10 -features are consistent with the high affinity of IRL1620 to the ETB receptor, which is comparable to those of ET-1 and ET-3 ( Fig. 1).
By contrast, IRL1620 does not bind to the ETA receptor at all in the same concentration range, confirming its high selectivity for the ETB receptor (Fig. 1). To elucidate the mechanism of this selectivity, we compared the amino acid compositions of the IRL1620 binding sites between the ETB and ETA receptors ( Fig. 4a and Supplementary Fig. 4). While the transmembrane region is highly conserved, the residues in ECL2 are diverse. In particular, the hydrophobic residues L252 ECL2 and I254 ECL2 are replaced with the polar residues H236 and T238 in the ETA receptor, respectively ( Fig. 4a, b). These residues form extensive hydrophobic interactions with the middle part of IRL1620. However, the double mutation of L252H and I254T only reduced the EC50 values of IRL1620 by 2-fold in the TGFα shedding and β-arrestin recruitment assays (Fig. 4c), suggesting that these residues are not the sole determinants for the receptor selectivity of IRL1620. Therefore, we focused on other residues of ECL2. In the ETB receptor , P259 ECL2 and V260 ECL2 generate a short kink on the loop between the β-strand and TM5,, but The ETA receptor has a truncated loop region and completely lacks these residues. (Fig. 4a). In addition, the ETA receptor has a proline (P228) in the first half of ECL2, which should disturb the β-strand formation as in the ETB receptor (Fig. 4b). These observations suggest that ECL2 adopts completely different structures between the ETA and ETB receptors, which may account for their different selectivities. ECL2 of the ETA receptor may form interactions with the N-terminal region of the endothelin peptides, while such interactions are not formed in the ETB receptor, and thus the truncation of the N-terminal region results in the -11 -drastically reduced affinity for only the ETA receptor.

Structural insight into receptor activation and partial activation
To elucidate the mechanism of the partial activation by IRL1620, we compared the IRL1620-bound structure with the full-agonist ET-3-bound structure (Fig. 5). The intracellular portions of the receptors are quite similar between the ET-3-and IRL1620-bound structures, in which TM7 and H8 adopt active conformations, while the remaining parts of the receptors still represent the inactive conformation of GPCRs ( Fig.   5a, Supplementary Fig. 5). On the extracellular side, IRL1620 induces similar conformational changes to those observed in the ET-1 and ET-3 structures; namely, the large inward motions of TM2, 6, and 7, which are critical for receptor activation (Fig.   5b, c). However, the extent of the inward motions of TM6-7 is smaller by about 1 Å in the IRL1620-bound structure, as compared with the ET-3-bound structure, due to the different ligand architectures between IRL1620 and ET-3. Since IRL1620 lacks the N-terminal region, the orthosteric pocket of the receptor has more space, and consequently the α-helical region of IRL1620 is tilted differently toward TM6 (Fig. 5b).
In addition, while the N-terminal region of ET-3 interacts with TM6 of the receptor, by forming a hydrogen bond between the carbonyl oxygen of T2 and K346 6.58 (superscripts indicate Ballesteros-Weinstein numbers), IRL1620 lacks this interaction, resulting in the different orientation of TM6-7. As TM6 plays an especially important role for the cytoplasmic G-protein binding, this difference is probably related to the partial agonist activity of IRL1620.
A comparison of the intermembrane parts revealed further differences in the -12 -allosteric coupling between the orthosteric pocket and the intermembrane part. Previous studies have shown that the agonist binding induces the disruption of the hydrogen-bonding network around D147 2.50 , which connects TMs 2, 3, 6, and 7 and stabilizes the inactive conformation of the ETB receptor 28 . The present high resolution ET-3-bound structure provides a precise mechanistic understanding of this rearrangement ( Supplementary Fig. 6a-c). In particular, the water-mediated hydrogen bonds involving D147 2.50 , W336 6.48 , and N378 7.45 in the inactive conformation collapse upon ET-3 binding, by the inward motions of TMs 2, 6, and 7 (Fig. 5d). The W336 6.48 side chain moves downward by about 2.5 Å, resulting in the disruption of the water-mediated hydrogen bond with D147 2.50 , and consequently, the D147 2.50 side chain moves downward by about 3 Å and forms hydrogen bonds with the N382 7.49 and N119 1.50 side chains. The N378 7.45 side chain also moves downward by about 1.5 Å and forms a hydrogen bond with the nitrogen atom of the W336 6.48 side chain.
The downward movements of the W336 6.48 and N378 7.45 side chains consequently induce the outward repositioning of the F332 6.44 side chain, by about 1 Å. W6.48 and F6.44 are considered to be the "transmission switch" of the class A GPCRs 35 , which transmits the agonist-induced motions to the cytoplasmic G-protein coupling interface. Overall, our results show that the collapse of the water-mediated hydrogen-bonding network involving D147 2.50 , W336 6.48 , and N378 7.45 propagates as the structural change in the transmission switch, and probably induces the outward displacement of the cytoplasmic portion of TM6 upon G-protein activation ( IRL1620 induces a similar but slightly different rearrangement of the -13 -hydrogen bonding network in the intermembrane part (Fig. 5e). Due to the smaller inward shift of TM6, the downward shift of the W336 6.48 side chain is smaller in the IRL1620-bound structure, and it still forms a water-mediated hydrogen bond with the D147 2.50 side chain. Consequently, the D147 2.50 side chain forms a direct hydrogen bond with N378 7.45 , thereby preventing the downward motion of N378 7.45 and the hydrogen bond formation between N378 7.45 and W336 6.48 . Overall, the downward motions of the W336 6.48 and N378 7.45 side chains are only moderate, as compared to those in the ET-3-bound structure, and the hydrogen-bonding network involving D147 2.50 , W336 6.48 , and N378 7.45 is partially preserved in the IRL1620-bound structure ( Fig. 5e, Supplementary Fig. 6). Accordingly, in the IRL1620-bound structure, the position of the "transmission switch" residue F332 6.44 is in between those of the active (ET-3 bound) and inactive (K-8794-bound) structures ( Fig. 5f).
This intermediate position of F332 6.44 should partly prevent the outward displacement of the cytoplasmic portion of TM6 that is required for G-protein activation, thus accounting for the partial agonistic effect of IRL1620 (Fig. 6).

Discussion
Previous studies have suggested that the α-helical and C-terminal regions of endothelins are critical elements for receptor activation 36-38 , while the N-terminal region is only responsible for the ETR selectivity 6 . Indeed, the N-terminal region-truncated analog IRL1620 has similar EC50 values, as compared with ET-1 17 . However, our pharmacological experiments for the first time proved that IRL1620 functions as a partial agonist for the ETB receptor, rather than a full agonist, suggesting the -14 -participation of the N-terminal region in the activation process of the ETB receptor. To clarify the receptor activation mechanism, we determined the crystal structures of the human ETB receptor in complex with ET-3 and IRL1620. The high-resolution structure of the ET-3-bound ETB receptor revealed that the large inward motions of the extracellular portions of TMs 2, 6, and 7 disrupt the water-mediated hydrogen bonding network at the receptor core, leading to receptor activation. The IRL1620-bound ETB structure revealed that the IRL1620-induced inward motions of TMs 6 and 7 are smaller by about 1 Å, as compared with those caused by ET-3, due to the lack of the N-terminal region. Consequently, the hydrogen-bonding network at the receptor core is partially preserved, thus preventing the transition to the fully active conformation adopted upon G-protein binding. These observations suggest that the interactions between the N-terminal regions of endothelins and TM6 also participate in receptor activation, while the extensive interactions of the α-helical and C-terminal regions with the receptor primarily contribute to this process ( Supplementary Fig. 3c). This activation mechanism is different from that of the small-molecule activated GPCRs (e.g., 2 adrenaline and M2 muscarinic acetylcholine receptors), in which only a small number of hydrogen-bonding interactions between the agonist and the receptor induce receptor activation, by affecting the receptor dynamics 39,40 .
D2.50 is one of the most conserved residues among the class A GPCRs (90%).
Recent high-resolution structures have revealed that a sodium ion coordinates with D2.50 and forms a water-mediated hydrogen bonding network in the intermembrane region, which stabilizes the inactive conformation of the receptor 41 , and its collapse leads to receptor activation. Our previous 2.2 Å resolution structure of the -15 -K-8794-bound ETB receptor revealed that a water molecule occupies this allosteric sodium site and participates in the extensive hydrogen-bonding network, instead of a sodium ion 28 , and this hydrogen-bonding network is collapsed in the 2.8 Å resolution structure of the ET-1-bound ETB receptor, indicating its involvement in the receptor activation 27 . Nevertheless, the precise rearrangement of this network still remained to be elucidated, due to the limited resolution. The current 2.0 Å resolution structure of the ET-3-bound ETB receptor revealed that the collapse of the water-mediated interaction between W336 6.48 and D147 2.50 is critical for receptor activation. This network is still partly preserved in the IRL1620-bound structure, thus preventing the transition to the fully active conformation upon G-protein coupling (Fig. 6). W6.48 is also highly conserved among the class A GPCRs (71%) 42 , and the association between W6.48 and D2.50 plays a critical role in the GPCR activation process, as shown in the previous nuclear magnetic resonance (NMR) study of the adenosine A2A receptor 43 . Given the importance of W3.36 and D2.50 in the activation of GPCRs, our proposed model of the partial receptor activation by IRL1620 is consistent with the previous functional analyses of GPCRs. To date, the 1 adrenergic receptor is the only receptor for which agonist-and partial agonist-bound structures were reported 44 . However, these structures are both stabilized in inactive conformations by the thermostabilizing mutations and thus revealed only slight differences ( Supplementary Fig. 7). Therefore, our study provides the first structural insights into the partial activation of class A GPCRs. Our current study further suggests possible improvements in clinical studies using ETB-selective agonists. IRL1620 is the smallest among the ETB-selective agonists, and thus is expected to be useful for the treatment of cancers and other diseases [18][19][20][21][22]24,25 .
-16 -While its effectiveness has been proven in rat experiments, a recent phase 2 study has failed 26 , and thus further improvement of IRL1620 is required for clinical applications.
Our cell-based assays and structural analysis revealed the partial agonistic effect of IRL1620 on the ETB receptor, suggesting the possible enhancement of its efficacy. The development of an ETB-selective full agonist based on the IRL1620-bound ETB structure will be beneficial for clinical applications.

Expression and purification
The haemagglutinin signal peptide, followed by the Flag epitope tag (DYKDDDDK) and a nine-amino-acid linker, was added to the N-terminus of the receptor, and a tobacco etch virus (TEV) protease recognition sequence was introduced between G57 and L66, to remove the disordered N terminus during the purification process. The C-terminus was truncated after S407, and three cysteine residues were mutated to alanine (C396A, C400A, and C405A) to avoid heterogeneous palmitoylation.
The thermostabilized construct ETB-Y5-T4L was subcloned into a modified pFastBac vector, with the resulting construct encoding a TEV cleavage site followed by a GFP-His 10 tag at the C-terminus. The recombinant baculovirus was prepared using the Bac-to-Bac baculovirus expression system (Invitrogen). Sf9 insect cells were infected with the virus at a cell density of 4.0 × 10 6 cells per millilitre in Sf900 II medium, and grown for 48 h at 27 °C. The harvested cells were disrupted by sonication, in buffer containing 20 mM Tris-HCl, pH 7.5, and 20% glycerol. The crude membrane fraction was collected by ultracentrifugation at 180,000g for 1 h. The membrane fraction was solubilized in buffer, containing 20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1% DDM, 0.2% cholesterol hemisuccinate, and 2 mg ml −1 iodoacetamide, for 1 h at 4 °C. The supernatant was separated from the insoluble material by ultracentrifugation at 180,000g for 20 min, and incubated with TALON resin (Clontech) for 30 min. The -18 -resin was washed with ten column volumes of buffer, containing 20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.1% LMNG, 0.01% CHS, and 15 mM imidazole. The receptor was eluted in buffer, containing 20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.01% LMNG, 0.001% CHS, and 200 mM imidazole. The eluate was treated with TEV protease and dialysed against buffer (20 mM Tris-HCl, pH 7.5, and 500 mM NaCl). The cleaved GFP-His10 tag and the TEV protease were removed with Co 2+ -NTA resin. The receptor was concentrated and loaded onto a Superdex200 10/300 Increase size-exclusion column, equilibrated in buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.01% LMNG, and 0.001% CHS. Peak fractions were pooled, concentrated to 40 mg ml −1 using a centrifugal filter device (Millipore 50 kDa MW cutoff), and frozen until crystallization. During the concentration, ET-3 or IRL1620 was added to a final concentration of 100 μM.

Data collection and structure determination
X-ray diffraction data were collected at the SPring-8 beamline BL32XU with 1×10 to 8 × 25 μm 2 (width × height) micro-focused beams and an EIGER X 9M detector (Dectris). For the IRL1620 data, we manually collected 68 data sets (10° to 180° per crystal), and the collected images were automatically processed with KAMO 31 (https://github.com/keitaroyam/yamtbx). Each data set was indexed and integrated with XDS 47 and then subjected to a hierarchical clustering analysis based on the unit cell parameters using BLEND 48 . After the rejection of outliers, 46 data sets were finally merged with XSCALE 47 . From the ET-3-bound crystals, various wedge data sets (3-180°) per crystal were mainly collected with the ZOO system, an automatic data-collection system developed at SPring-8 (K.Y., G.U., K.H., M.Y., and K.H., submitted). The loop-harvested microcrystals were identified by raster scanning and subsequently analyzed by SHIKA 49 . The collected images were processed in the same manner, except that correlation coefficient-based clustering was used instead of BLEND, and finally 483 datasets were merged. The ET-3-bound structure was determined by molecular replacement with PHASER 50 , using the ET-1-bound ETB structure (PDB 5GLH). Subsequently, the model was rebuilt and refined using COOT 51 and PHENIX 52 , respectively. The IRL1620-bound structure was determined by molecular replacement, using the ET-1-bound ETB structure, and subsequently rebuilt and refined as described Emax values were obtained.

β-Arrestin Recruitment Assay
For the NanoBiT-β-arrestin recruitment assay 54 , a receptor construct was and Emax values were obtained as described above.