Abstract

Endothelin receptors (ETRs) have crucial roles in vascular control and are targets for drugs designed to treat circulatory-system diseases and cancer progression. The nonpeptide dual-ETR antagonist bosentan is the first oral drug approved to treat pulmonary arterial hypertension. Here we report crystal structures of human endothelin ETB receptor bound to bosentan and to the ETB-selective analog K-8794, at 3.6-Å and 2.2-Å resolution, respectively. The K-8794-bound structure reveals the detailed water-mediated hydrogen-bonding network at the transmembrane core, which could account for the weak negative allosteric modulation of ETB by Na+ ions. The bosentan-bound structure reveals detailed interactions with ETB, which are probably conserved in the ETA receptor. A comparison of the two structures shows unexpected similarity between antagonist and agonist binding. Despite this similarity, bosentan sterically prevents the inward movement of transmembrane helix 6 (TM6), and thus exerts its antagonistic activity. These structural insights will facilitate the rational design of new ETR-targeting drugs.

Main

ETRs are class A G-protein-coupled receptors (GPCRs) activated by the vasoactive peptide hormones endothelins1. Two subtypes of ETRs (ETA and ETB receptors)2,3 are involved in various functions, such as regulation of blood pressure, sodium excretion, cell proliferation, and neural crest development4,5,6,7. Increased production of the endogenous agonist endothelin-1 (ET-1) is significantly related to diseases of the circulatory system8, including pulmonary arterial hypertension (PAH)9,10, ventricular dysfunction, chronic heart failure, and chronic kidney disease11. In addition, autocrine and paracrine signaling of ET-1 through the ETRs has crucial roles in tumor growth and survival12. Therefore, ETR antagonists have been developed for the treatment of circulatory-system diseases and cancers.

The dual ETA–ETB receptor antagonist bosentan, a nonpeptide sulfonamide compound developed by high-throughput screening13,14,15, is the first oral drug approved for the treatment of for PAH9,10,13,14. Although PAH is an orphan disease16, worldwide sales of bosentan have surpassed $1 billion dollars per year17. Bosentan is also used as a treatment for digital ulcers associated with systemic sclerosis18. Bosentan has been used in numerous pharmacological and clinical studies on ETRs. However, its pharmacological properties are not optimal: bosentan's moderate affinity and fast dissociation kinetics13,19,20 both limit its activity in vivo. In addition, the initial phase 2 study of bosentan for the treatment of essential systemic hypertension suggested possible dose-dependent hepatotoxicity21. Hence, bosentan-based antagonists have been developed to improve drug efficacy19,22,23,24,25. In particular, macitentan, a high-affinity analog of bosentan, has an improved adverse-effect profile compared with bosentan and is used clinically to treat PAH26. Despite the numerous structure–activity relationship (SAR) studies, little is known about the binding modes of bosentan analogs to ETRs and their subtype selectivities, as there is no structural information available for the complex of an ETR and its antagonist.

We previously reported the structures of human ETB receptor in the ligand-free form and in complex with the endogenous agonist ET-1, which revealed the activation process of the receptor by ET-1 (ref. 27). However, those structures provided little information about the binding mode of the antagonist, because there was no apparent structural similarity between the small chemical compound bosentan and the 21-amino-acid peptide ET-1. Here we report X-ray crystal structures of the human ETB receptor bound to bosentan and its high-affinity analog K-8794 (ref. 28) at 3.6-Å and 2.2-Å resolution, respectively. These structures provide mechanistic insights into the binding modes and the antagonistic effects of the drugs, which might facilitate the development of ETR ligands with improved efficacy.

Results

Overall structures

We first investigated the pharmacological properties of the thermostabilized receptor (ETB-Y5) used in the previous structural study (Supplementary Fig. 1a). The five thermostabilizing mutations improved the receptor stability in the ligand-free conformation, and did not affect the ability of the endogenous agonist ET-1 to activate G proteins29,30 (Table 1, Supplementary Fig. 1b, and Online Methods). Although ETB-Y5 showed slightly reduced affinity for the antagonists in radiobinding assays, it elicited responses similar to those of the wild-type ETB receptor in cell-based assays with bosentan and K-8794 (Table 1 and Supplementary Fig. 1b,c). For crystallization, we inserted minimal T4 lysozyme (mT4L) at the third intracellular loop (ICL3) of ETB-Y5 and crystallized this construct (ETB-Y5-mT4L) in complex with the ETB-selective antagonist K-8794. We solved the structure at 2.2-Å resolution by molecular replacement, using the ligand-free structure of ETB-Y5-mT4L (PDB 5GLI) (Table 2 and Fig. 1a). The crystal packing and overall structure are similar to those of the ligand-free, inactive form (r.m.s. deviation = 0.66 Å) (Fig. 1b and Supplementary Fig. 1d).

Table 1: Pharmacological characterization of thermostabilized ETB receptors
Table 2: Data collection and refinement statistics
Figure 1: Overall structures of antagonist-bound ETB receptors.
Figure 1

(a) Overall structures of the ETB receptor bound to K-8794 (left; turquoise) and bosentan (right; orange), viewed parallel to the membrane plane. The ETB receptors are depicted by ribbons. Bosentan and K-8794 are represented by CPK models and are colored cyan and magenta, respectively. The waters and lipids in the K-8794-bound structure are shown as red dots and white sticks, respectively. (b) Superimpositions of the inactive ETB structures, viewed from the extracellular (left) and intracellular (right) sides. The ligand-free structure is colored green. (c) Intracellular views of the ETB receptors bound to K-8794 (left) and bosentan (middle), and in the ligand-free form (right). Sulfate ions are shown as sticks, and hydrogen-bonding interactions are indicated by black dashed lines. ECL, extracellular loop; H, helix.

We also tried to crystallize the ETB-Y5-mT4L construct bound to bosentan, but we did not observe the electron density of the bosentan in the structure. Because one of the thermostabilizing mutations, Asp1542.57Ala (throughout, superscript numbers in mutations and residues represent Ballesteros–Weinstein numbers), slightly reduces the affinity for bosentan as well as for K-8794, we made another construct, ETB-Y4-mT4L, in which this mutation was reverted to the wild-type residue (Table 1). The crystals of the bosentan-bound ETB-Y4-mT4L were quite tiny (Supplementary Fig. 1e), but an automated data-collection and processing system (K.Y., G. Ueno, K. Hasegawa, M. Yamamoto and K.H., unpublished data) allowed us to collect and merge data sets from numerous crystals. Eventually we solved the structure at 3.6-Å resolution by molecular replacement, using the K-8794-bound ETB structure (Table 2 and Fig. 1a). The overall structure of the bosentan-bound ETB-Y4-mT4L is similar to those of the K8794-bound and the ligand-free, inactive conformations (r.m.s. deviation = 1.36 and 1.35 Å, respectively) (Fig. 1b), except for slight structural differences in ICL2, which probably reflect its innate flexibility (Fig. 1c).

Weak allosteric effect of Na+ ions

The high-resolution structure of the K-8794-bound ETB receptor allowed us to determine the precise assignments of the water molecules and the hydrogen-bonding network around the allosteric Na+ site that is conserved among class A GPCRs31,32 (Fig. 2). Na+ ions work as negative allosteric modulators for many class A GPCRs31,32, and recent crystal structures have revealed the conserved binding site within the transmembrane bundle33,34,35,36. This Na+-binding site is also conserved in ETB and includes Asp1472.50, a key residue that forms ionic interactions with the Na+ ions in other GPCR structures (Fig. 2a). The previous ligand-free ETB structure suggested that the water-mediated hydrogen-bonding network around the putative allosteric Na+ site probably stabilizes the inactive conformation of the ETB receptor. However, the medium-resolution (2.5-Å) structure did not allow for a precise understanding of the hydrating waters and their interactions at the putative Na+-binding site (Fig. 2b). The high-resolution structure of the K-8794-bound ETB receptor revealed an additional water-mediated hydrogen-bonding network (Fig. 2c), as well as an electron density at the putative Na+-binding site surrounded by the five oxygen atoms of the Asp1472.50, Ser1843.35, and Thr1883.39 side chains and two water molecules (Fig. 2d). This resembles the Na+ ion coordination in the human δ-opioid receptor (PDB 4N6H)35 (Fig. 2e). However, the coordination distances between the electron density and the oxygen atoms are all within 2.5–3.2 Å, which is somewhat longer than those for the Na+ coordination in other GPCR and protein structures33,34,35,36,37 (2.2–2.6 Å). To investigate the allosteric effect of Na+ ions on the ETB receptor, we measured the half-maximal inhibitory concentration (IC50) of Na+ ions on ET-1 binding, and found that the IC50 value was over 1 M, which is not in the physiological range (Fig. 2f). This value is much greater than those for agonist binding to NTSR1 (43 mM)38 or to the ETA receptor (245 mM). These results suggest the possibility that a water molecule, instead of Na+, occupies the allosteric Na+-binding site of the ETB receptor and weakly stabilizes the inactive conformation.

Figure 2: Hydrogen-bonding network in the transmembrane core.
Figure 2

(a) A comparison of the sequences of the putative Na+-binding sites of ETB, ETA, NTSR1, and other GPCR structures. BW, Ballesteros-Weinstein numbers. (bd) Water-mediated hydrogen-bonding networks at the transmembrane core, observed in the ligand-free (b) and K-8794-bound (c) ETB structures, shown as green and turquoise ribbons and sticks, respectively. Water molecules are shown as red spheres, and hydrogen-bonding interactions are indicated by black dashed lines. (d) A close-up view of the putative Na+-binding site in the K-8794-bound structure. The lengths between the waters and coordinating oxygens are indicated by red numbers. The electron density is indicated by the 2FoFc map, contoured at 1.5σ. (e) The Na+-binding site in the 1.8-Å resolution structure of the human δ-opioid receptor. The Na+ ion is shown as a large purple sphere. (f) Competition binding curves of Na+ ions for [125I]ET-1 binding to ETRs. Data are mean ± s.e.m. from two independent experiments. [125I]ET-1 binding to the ETB receptor was not 50% inhibited by 2 M NaCl.

Antagonist-binding site

In the current structures, both bosentan and K-8794 are buried within the positively charged transmembrane cleft of the ETB receptor with clear electron densities (Supplementary Fig. 2a–c). The depth of the binding site is comparable to those in peptide-activated GPCRs, such as the orexin OX2 receptor39 and the opioid receptor NOP40 (Supplementary Fig. 3a–c), which shows that bosentan occupies the typical drug-binding site of the peptide-activated GPCRs. The bosentan-binding site is open toward the extracellular solvent (Supplementary Fig. 3d), in contrast to the previous ET-1-bound structure in which the N-terminal tail and the second extracellular loop of the ETB receptor form a lid and prevent the dissociation of ET-1 (Supplementary Fig. 3e). This structural feature could account for the fast dissociation kinetics of bosentan19 (t1/2 < 1 min), in contrast to the virtually irreversible binding of ET-1 (refs. 29,41) (t1/2 > 1 week).

Bosentan and its high-affinity analog K-8794 contain many aromatic moieties composed of a central pyrimidine template with four substituents: a 2-pyrimidyl, a 4-sulfonamide, a 5-phenol, and a 6-hydroxyl (Fig. 3a). We first describe the interactions of K-8794 observed in the high-resolution structure. The sulfonamide of K-8794 forms ionic interactions with Lys1823.33, Lys2735.38, and Arg3436.55 (Fig. 3b and Supplementary Figs. 1, 2d, and 4a). The 4-t-butyl phenyl group linked to the sulfonamide forms hydrophobic contacts with Val1773.28, Pro1783.29, Phe2404.64, and Cys255ECL2. The methoxyphenoxy group of K-8794 fits within the hydrophobic pocket, consisting of Val1853.36, Leu2775.42, Tyr2815.46, Trp3366.48, Leu3396.51, and His3406.52 at the transmembrane core. The pyrimidine at the 2-position of the central pyrimidine forms a hydrophobic contact with Ile3727.39. The N-(2,6-dimethylphenyl)-3-propionamide at the 6-position of the pyrimidine fills the space within the transmembrane cleft toward TM2. The dimethylphenyl group makes van der Waals interactions with the side chains of Asn1582.61 and Val1773.28. The carbonyl oxygen of the amide bond forms a hydrogen bond with His1502.53 and water-mediated hydrogen bonds with Ser3797.46 and Gln1813.32. Overall, the sulfonamide moiety of K-8794 is specifically recognized by the positively charged residues of the ETB receptor, and the other moieties fill the space within the transmembrane binding pocket surrounded by TM2–7 and facilitate the interactions with the receptor.

Figure 3: Drug-binding site.
Figure 3

(a) Chemical structures and conformations of K-8794 and bosentan. The antagonists in the current complex structures are shown as sticks, color-coded as in Figure 1. (b,c) Detailed interactions of K-8794 (b) and bosentan (c) with the receptor, viewed from the extracellular side (left) and within the membrane plane (right). The ETB receptors are shown in ribbon representations, and the antagonists are shown as sticks color-coded as in Figure 1. Hydrogen-bonding interactions are indicated by black dashed lines. (d) Cross-sectional views of the bosentan-bound and K-8794-bound ETB structures, color-coded as in Figure 1. (e) Superimposition of the bosentan- and K-8794-bound ETB structures, viewed from the extracellular side. (f) Sequence conservation of the bosentan-binding sites between the ETA and ETB receptors, mapped on the bosentan-bound ETB structure. Conserved and nonconserved residues are colored pink and gray, respectively. The pharmacophore of bosentan is highlighted. (g) Superimposition of the bosentan- and K-8794-bound ETB structures, with the extracellular portion of TM2 color-coded as in Figure 1. His1502.53, Asp1542.57, and Pro1562.59 are represented by sticks. Hydrogen-bonding interactions are indicated by black dashed lines. The sequences of TM2 for the ETA and ETB receptors are shown at the bottom of the panel.

The chemical structure of bosentan is similar to that of K-8794, except that bosentan has an ethylene glycol at the 6-position of the central pyrimidine (Fig. 3a). Consistently, bosentan and K-8794 superimposed well in these complex structures, and the binding mode of bosentan was shown to be essentially the same as that of K-8794 (Fig. 3b–d and Supplementary Figs. 2e,f and 4b). However, the smaller substituent on the 6-position of the central pyrimidine in bosentan results in weaker interactions with the receptor: the hydroxyl group of the ethylene glycol forms a hydrogen bond only with Asp1542.57 on TM2 (Fig. 3c,e). This difference may explain the weaker affinity of bosentan compared with that of K-8794.

A sequence comparison of the ETA and ETB receptors showed that the residues that interact with the common moieties between bosentan and K-8794 are highly conserved, which suggests that their binding modes could be essentially similar in the ETA receptor (Fig. 3f and Supplementary Fig. 5). In contrast, the amino acid sequences of the extracellular portions of TM2, which is adjacent to the moieties where bosentan and K-8794 bind, are mostly divergent (Fig. 3g). A comparison of the bosentan- and K-8794-bound structures showed that the extracellular portions of TM2 and the adjacent helices (TM1 and TM7) adjust to fit the respective components of the ligands and adopt slightly different conformations. This observation proves that the helix orientation of TM2 can change and modulate the size and shape of the binding pocket, as well as form direct/indirect interactions with the substituents attached to the 6-position of the central pyrimidine (we discuss this point further in Supplementary Note 1). As revealed by the current structures, the pocket of the ETB receptor is capable of accommodating the bulky moieties of K-8794. In contrast, biochemical experiments42, in combination with the sequence comparison, indicate that this pocket is smaller in the ETA receptor than in the ETB receptor, and thus the former receptor is not able to accommodate the bulky ethylene glycol substituent of K-8794.

It should be noted that the ETB-Y5-mT4L construct used for the K-8794-bound structure exhibits about 8- and 21-fold decreased affinity for bosentan and K-8794, respectively, owing to the thermostabilizing mutation Asp1542.57Ala (Table 1). However, this construct still shows 40-fold (120-fold in the wild type) higher affinity for K-8794 compared with that for bosentan. Therefore, these data indicate that the obtained structure retains the ligand-binding preference of the wild-type ETB receptor.

Conserved binding mode of ETR antagonists

Bosentan analogs consist of pyrimidine as the central template substituted with four substituents at positions 2, 4, 5, and 6, with sulfonamide attached to the 4-position of the central pyrimidine (Supplementary Fig. 6). Previous SAR studies demonstrated that substitution of the sulfonamide leads to the total abrogation of binding to ETRs15. The residues constituting the binding sites for bosentan and K-8794, including the positively charged residues that coordinate the sulfonamide in the current structures, are highly conserved between ETA and ETB receptors (Fig. 3f and Supplementary Figs. 4 and 5), which suggests that bosentan and its analogs basically adopt common binding modes for both ETRs.

Bosentan analogs share additional features in their molecular components25 (Supplementary Fig. 6). The first similarity is the aromatic moiety that modifies the 5-position of the central pyrimidine ring, which corresponds to the 2-methoxyphenoxy group in bosentan. The second similarity is the aromatic or alkyl moiety that modifies the sulfonamide, which corresponds to the 4-t-butylphenyl group in bosentan. Previous SAR studies have shown that substitution of these moieties with smaller hydrophobic groups (e.g., methyl groups) cause up to a 100-fold reduction in the affinity for ETRs22, which suggests that the van der Waals interactions between these moieties and ETRs are important for the binding of bosentan analogs (Fig. 3b,c). In contrast, the 2-pyrimidyl group is not conserved among the bosentan analogs (Supplementary Fig. 6). The interaction of the 2-pyrimidyl group with the ETB receptor is relatively poor compared with that of the other moieties (Fig. 3b,c). This is consistent with previous SAR studies showing that modification or substitution of the 2-pyrimidyl group has less influence on these affinities compared with those of the other moieties of bosentan22. Those SAR studies and the current structures suggest that the region of bosentan highlighted in Figure 3f forms a pharmacophore that plays a critical role in receptor binding.

This model also suggests the putative binding mode of non-bosentan-like ETR antagonists. Such antagonists (e.g., the ETA-selective antagonists ambrisentan23 and antrasentan43,44) also have a negatively charged group, such as a sulfonamide or a carboxylate, that is surrounded by two or three hydrophobic (aromatic or acyl) groups (Supplementary Fig. 6). The negatively charged groups might occupy the center of the transmembrane pocket and form ionic interactions with the positively charged residues of the endothelin receptors, whereas the other hydrophobic groups might fit into the pocket and facilitate interactions, as observed in the current ETB structures.

Structural insights into antagonism

Bosentan was not developed through mimicry of endogenous agonist peptides, and thus it possesses little similarity to the agonist endothelin peptides13,14,15. However, structural comparison with ET-1-bound ETB revealed an unexpected resemblance in receptor binding. Bosentan occupies the bottom of the orthosteric site of the ETB receptor, which corresponds to the binding site for the C-terminal tripeptide of ET-1 (Ile19, Ile20, and Trp21) (Fig. 4a,b and Supplementary Fig. 7a,b). The position of the sulfonamide of bosentan superimposes well with that of the C-terminal carboxylate of ET-1, where they are coordinated by the same positively charged residues of the ETB receptor, Lys1823.33, Lys2735.38, and Arg3436.55 (Supplementary Fig. 7c). In addition, the aromatic and hydrophobic moieties of bosentan, including the 2-pyrimidyl group, the 4-t-butylphenyl group, and the 5-methoxyphenoxy group, occupy the positions corresponding to the side chains of Ile19, Ile20, and Trp21 of ET-1, respectively, and essentially the same residues of the ETB receptor are involved in binding these ligands (Supplementary Fig. 7c). Thus, the receptor interactions of bosentan and the C-terminal tripeptides of ET-1 are quite similar, which suggests that the ionic and hydrophobic interactions within the transmembrane core are crucial determinants for both antagonist and agonist binding to ETRs.

Figure 4: Comparison of the binding modes of ET-1 and bosentan.
Figure 4

(a) Superimposition of the ET-1-bound (pink) and bosentan-bound (orange) ETB receptors. ET-1 is depicted as a magenta ribbon, with the side chains of ET-118–21 shown as sticks, and bosentan is represented by cyan sticks. TM5 is omitted. (b) Superimposition of bosentan and ET-119–21, represented by sticks, viewed parallel to the membrane plane. (c) A close-up view of TM6, showing the different orientations of the three structures. The critical residues for receptor activation, Arg3436.55 and Trp3366.48, are represented by sticks. (d,e) Comparison of the ligand-induced conformational changes of ET-1 (d) and bosentan (e). Positively charged residues that recognize the negatively charged groups of the ligands (Lys1823.33, Lys2735.38, and Arg3436.55) are represented by sticks. Black arrows indicate the conformational changes in the extracellular halves of TM6 and TM7 upon ligand binding. (f) Schematic representations of the ET-1- and bosentan-induced conformational changes of the ETB receptor. Color-coding in ce is as described in Figure 1.

Despite the similarity between the binding modes of bosentan and of the C-terminal tripeptide of ET-1, these ligands induce strikingly different structural changes in the extracellular portion of the ETB receptor. Previous studies have shown that the ET-1-induced 4-Å inward movements of TM6 and TM7 are critical for receptor activation27 (Fig. 4c,d). These conformational changes are mediated mainly by the negative charges of Asp18 and the C-terminal carboxylate of ET-1, which attract the Arg3436.55 side chain27,45. Bosentan lacks the negatively charged moiety corresponding to Asp18 of ET-1 (Fig. 4e). Accordingly, Arg3436.55 is attracted toward the sulfonamide, and hence bosentan induces only a slight inward movement of TM6 (Fig. 4c,e). In addition, the 2-methoxyphenoxy group of bosentan protrudes farther toward TM6 compared with the Trp21 side chain of ET-1, thereby sterically preventing the inward movement of TM6 (Fig. 4c,f). Furthermore, bosentan does not induce any inward movement of TM7 because it only poorly interacts with TM7, as compared with ET-1 (Figs. 3b and 4e, and Supplementary Fig. 7d,e). These structural differences explain the antagonistic effects of bosentan and its analogs.

Discussion

The current crystallographic studies elucidated the interactions between bosentan and the ETB receptor, which are likely to be conserved in the ETA receptor. A comparison of the chemical structures of ETR antagonists suggested that the negatively charged groups of the antagonists are commonly recognized by the positively charged residues of the ETB receptor. This structural information will promote the development of new therapeutic drugs targeting ETRs. Moreover, we unexpectedly discovered that bosentan shows some resemblance to the C-terminal tripeptide of ET-1 in terms of its recognition by the receptor. In addition to the endothelin receptors, peptide-activated GPCRs often require the specific sequences of their agonist peptides for receptor binding46,47,48,49,50. Therefore, structure-guided mimicry of agonist peptides might facilitate the design of new small-molecule drugs for peptide-activated GPCRs51,52.

Methods

Expression and purification.

The crystallization construct was based on ETB-Y5-mT4L, which was used in a previous study27. In brief, we added the hemagglutinin (HA) signal peptide, followed by the Flag epitope tag and a 9-amino-acid linker, to the N terminus of the receptor, and introduced a tobacco etch virus (TEV) protease recognition sequence between Gly57 and Leu66. The C terminus was truncated after Ser407, and three cysteine residues were mutated to alanine (Cys396Ala, Cys400Ala, and Cys405Ala). In addition, five thermostabilizing mutations were introduced. We introduced mT4L, containing the Cys54Thr and Cys97Ala mutations, into ICL3, between Lys3035.68 and Lys3116.23. For the crystallization of the bosentan-bound ETB receptor, the thermostabilizing mutation Asp1542.57Ala was reverted to Asp (ETB-Y4-mT4L). These constructs were cloned into a modified pFastBac1 vector containing the TEV protease recognition sequence followed by an EGFP-His10 tag.

The recombinant baculovirus was prepared with the Bac-to-Bac baculovirus expression system (Invitrogen). Sf9 insect cells were infected with the virus at a cell density of 4.0 × 106 cells ml−1 in Sf900 II medium (Invitrogen) 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 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 0.5% DDM, 0.1% cholesterol hemisuccinate and 2 mg ml−1 iodoacetamide for 2 h at 4 °C. Antagonists were added to a 10 μM final concentration throughout the purification procedure. The supernatant was separated from the insoluble material by ultracentrifugation at 180,000g for 30 min, and incubated with TALON resin (Clontech) for 30 min. The resin was washed with 10 column volumes of buffer containing 20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.1% LMNG, 0.01% CHS, and 20 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 dialyzed against buffer containing 20 mM Tris-HCl, pH 7.5, and 500 mM NaCl. The cleaved GFP-His10 tag and the TEV protease were removed with Ni+-NTA resin. The receptor was concentrated and loaded onto a Superdex 200 10/300 Increase size-exclusion column and equilibrated in buffer containing 10 mM HEPES-NaOH, pH 7.5, 200 mM NaCl, 0.01% LMNG, and 0.001% CHS. Peak fractions were pooled, concentrated to 40 mg ml−1 with a centrifugal filter device (Millipore 50-kDa molecular weight cutoff), and frozen until crystallization. During the concentration step, the antagonists were added to a final concentration of 100 μM.

Crystallization.

The purified receptors were reconstituted into the lipidic cubic phase (LCP) of monoolein (Nucheck), supplemented with cholesterol at a ratio of 8:9:1 (w/w) protein:monoolein:cholesterol. The protein-laden mesophase was dispensed onto 96-well glass plates in 30-nl drops and overlaid with 800 nl of precipitant solution, using a mosquito LCP (TTP LabTech)53. Crystals of ETB-Y5-mT4L bound to K-8794 were grown at 20 °C in the precipitant conditions containing 25% PEG500DME, 100 mM MOPS-NaOH, pH 7.0, and 100 mM (NH4)2SO4. The crystals of ETB-Y4-mT4L bound to bosentan were grown in the precipitant conditions containing 20–25% PEG500DME, 100 mM MOPS-NaOH, pH 7.0, and 100 mM Na2SO4. The crystals were harvested directly from the LCP with micromounts (MiTeGen) or LithoLoops (Protein Wave) and frozen in liquid nitrogen, without any additional cryoprotectant.

Data collection and structure determination.

X-ray diffraction data were collected at SPring-8 beamline BL32XU with a 10 × 18 or 9 × 14 μm2 (width × height) micro-focused beam and an MX225HS CCD (charge-coupled device) detector (Rayonix, LLC). For the K-8794 data, wedges of 20–180° were collected from several crystals. The collected images were automatically processed with KAMO (https://github.com/keitaroyam/yamtbx). Each data set was indexed and integrated with XDS54 and then subjected to a hierarchical clustering analysis based on the correlation coefficients between data sets. The data sets from two crystals were finally merged with XSCALE54 after the rejection of outliers. From the bosentan-bound crystals, small wedge data sets of 10° per crystal were collected with the ZOO system, an automatic data-collection system developed at SPring-8 (K.Y., G. Ueno, K. Hasegawa, M. Yamamoto and K.H., unpublished data). The loop-harvested microcrystals were identified by raster scanning and subsequently analyzed by SHIKA55. The collected images were processed in the same manner, and finally 14 crystals were merged. In the space group (P3221) of the bosentan-bound crystal, there was two-fold indexing ambiguity, which was resolved before merging with the selective-breeding algorithm56 implemented in KAMO. The K-8794-bound structure was determined by molecular replacement with PHASER57, using the ligand-free ETB structure (PDB 5GLI). Subsequently, the model was rebuilt and refined with COOT58 and phenix refinement implemented in the PHENIX program suite, with the individual anisotropic ADP refinement59. The bosentan-bound structure was determined by molecular replacement, using the K-8794-bound structure, and subsequently rebuilt and refined as described above. The final model of K-8794-bound ETB-Y5-mT4L contained residues 85–303 and 311–404 of the ETB receptor, all residues of mT4L, K-8794, one cholesterol, 5 sulfate ions, 11 monoolein molecules, and 106 water molecules. The model of ETB-Y4-mT4L contained residues 90–303 and 311–403 of the ETB receptor, residues 1–10 and 23–117 of mT4L, four sulfate ions, and bosentan. The model quality was assessed with MolProbity60. Figures were prepared with cuemol (http://www.cuemol.org/en/).

TGF-α shedding assay.

For the functional study of the ETB constructs, residues 58–65 of the codon-optimized human ETB and the thermostabilized ETB constructs (ETB-Y4 and ETB-Y5) were replaced with the Flag epitope tag (DYKDDDDK) and inserted into a pCAG-neo expression plasmid vector (Wako Pure Chemical). For the ETA construct, the Flag epitope tag was inserted into residues 20 and 21 of the human ETA receptor and cloned into the pCAGGS expression plasmid vector (a kind gift from Dr. Jun-ichi Miyazaki, Osaka University). To monitor the G-protein-dependent signals of the endothelin receptors, we carried out a TGF-α shedding assay as described previously, with minor modifications61, including the use of codon-optimized alkaline phosphatase (AP)-tagged TGF-α (AP-TGF-α) and the use of a polyethylenimine transfection reagent. Briefly, HEK293A cells were seeded in a 10-cm dish at a density of 2 × 105 cells per well with 10 ml of Dulbecco's modified Eagle's medium supplemented with 10% FCS, 100 U ml−1 penicillin, and 100 μg ml−1 streptomycin (complete DMEM) and cultured for 1 d. The cells were transfected with the AP-TGF-α-encoding plasmid and the ETR-encoding plasmid (2.5 μg and 1 μg, respectively, per dish), combined with 25 μl of 1 mg ml−1 PEI solution (Polyethylenimine 'Max' (MW 40,000), Polysciences) together with 1 ml of Opti-MEM I reduced serum medium (Thermo Fisher Scientific). After a 24-h incubation, the cells were trypsinized, neutralized with the complete DMEM, collected in a 50-ml tube, centrifuged, and suspended in 30 ml of Hank's Balanced Salt Solution (HBSS) containing 5 mM HEPES, pH 7.4. After a 15-min incubation at room temperature to settle the spontaneous AP-TGF-α release caused by trypsinization, the cells were centrifuged and suspended in 30 ml of the HEPES-containing HBSS. The cell suspension was seeded in a 96-well plate (cell plate) at a volume of 80 μl per well. After a 25-min incubation in a CO2 incubator, 10 μl of vehicle (0.01% BSA and HEPES-containing HBSS) or 10 μl of either bosentan (final concentrations ranging from 10 nM to 32 μM) or K-8794 (final concentrations ranging from 100 pM to 10 μM) were added. We note that concentrations higher than 32 μM bosentan or 10 μM K-8794 induced nonspecific AP-TGF-α release responses, and thus we used the indicated ranges of the antagonists. After a 5-min incubation at room temperature, 10 μl of titrated concentrations of ET-1 (final concentrations ranging from 3.2 pM to 320 nM; for measurement of the agonist activity) or 10 μl of a set concentration of ET-1 (final concentration of 0.2 nM; for measurement of the antagonist activity) or vehicle (for normalization of the background antagonist activity) was added. After a 1-h incubation in the CO2 incubator, the plates were centrifuged at 190g for 2 min, and the supernatant (80 μl) was transferred to an empty 96-well plate (conditioned media (CM) plate). The AP reaction solution (10 mM p-nitrophenylphosphate (p-NPP), 120 mM Tris-HCl, pH 9.5, 40 mM NaCl, and 10 mM MgCl2) was dispensed into the cell plates and the CM plates (80 μl per well). The optical density at 405 nm (OD405) of the plates was measured with a microplate reader (SpectraMax 340 PC384, Molecular Devices) before and after a 1-h incubation at room temperature. For each well measurement, the increase in the OD405 during the 1-h incubation with p-NPP (ΔOD405) was used as the AP activity, and the value from the CM plate was normalized by the total ΔOD405. We calculated the AP-TGF-α release by subtracting the spontaneous AP-TGF-α accumulation under the vehicle-treated conditions from that under the compound-stimulated conditions. To assess the inhibitory effect of the endothelin antagonist, we subtracted the AP-TGF-α release signal in the antagonist treatment alone from the ET-1-stimulated signal for each antagonist concentration, and we normalized the signal in the ET-1 treatment alone at 100%. The AP-TGF-α release signals were fitted to a four-parameter sigmoidal concentration–response curve with Prism 7 software (GraphPad Prism). From the fitted curve, we obtained the values for pEC50 (equal to −log10 EC50; for agonist activity) and pIC50 (equal to −log10 IC50; for antagonist activity).

Ligand-binding experiments.

For competitive ligand-binding assays, wild-type (6hNETBR)62 and mutant receptors in the pFastBac1 vector and the pcDNA3.1 vector were used for expression in insect cells and mammalian cells, respectively. Membranes from Sf9 or HEK293 cells expressing 6hNETBR or its mutants were prepared, and the expressed ETB receptors were quantitated as described previously62,63. Peptide competition binding was initiated by the addition of membranes from Sf9 cells (0.1–1.2 μg) or HEK293 cells (1–5 μg) to the assay mixture, composed of 0.1% BSA, 0.03–0.05 nM [125I]ET-1 (2,200 Ci mmol−1; PerkinElmer Life Sciences), and eight concentrations of unlabeled bosentan (ranging from 10 pM to 1 mM) or K-8794 (ranging from 100 pM to 1 mM) in 50 mM HEPES-NaOH, pH 7.5, 10 mM MgCl2 (Mg-HEPES)64. Binding reactions were incubated at 37 °C for 1 h, terminated by dilution with cold Mg-HEPES, and filtered onto glass fiber filters in 96-well plates (multiscreen HTS FB, Merck Millipore) to separate the unbound [125I]ET-1. After three washes with cold Mg-HEPES, the radioactivity captured by the filters was counted with a γ-counter. Filters were pretreated with 0.1% BSA in Mg-HEPES. In the competitive binding assays with NaCl, NaCl concentrations ranging from 1 mM to 2 M were included in the Mg-HEPES reaction buffer containing 0.03–0.05 nM [125I]ET-1. The assays were repeated two to four times. The results were analyzed by nonlinear regression with GraphPad Prism 6 software.

A Life Sciences Reporting Summary for this article is available.

Data availability.

Coordinates and structure factors have been deposited in the Protein Data Bank under the accession numbers 5XPR for the bosentan-bound and 5X93 for the K-8794-bound ETB structures. All other data are available from the authors upon reasonable request.

Additional information

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Accessions

Primary accessions

References

  1. 1.

    et al. A novel potent vasoconstrictor peptide produced by vascular endothelial cells. Nature 332, 411–415 (1988).

  2. 2.

    , , , & Cloning and expression of a cDNA encoding an endothelin receptor. Nature 348, 730–732 (1990).

  3. 3.

    et al. Cloning of a cDNA encoding a non-isopeptide-selective subtype of the endothelin receptor. Nature 348, 732–735 (1990).

  4. 4.

    & Endothelin system: the double-edged sword in health and disease. Annu. Rev. Pharmacol. Toxicol. 41, 851–876 (2001).

  5. 5.

    , , & Regulation of blood pressure and salt homeostasis by endothelin. Physiol. Rev. 91, 1–77 (2011).

  6. 6.

    & Endothelins: molecular biology, biochemistry, pharmacology, physiology, and pathophysiology. Pharmacol. Rev. 46, 325–415 (1994).

  7. 7.

    , , & The temporal requirement for endothelin receptor-B signalling during neural crest development. Nature 402, 496–501 (1999).

  8. 8.

    , & New therapeutics that antagonize endothelin: promises and frustrations. Nat. Rev. Drug Discov. 1, 986–1001 (2002).

  9. 9.

    et al. Effects of the dual endothelin-receptor antagonist bosentan in patients with pulmonary hypertension: a randomised placebo-controlled study. Lancet 358, 1119–1123 (2001).

  10. 10.

    et al. Bosentan therapy for pulmonary arterial hypertension. N. Engl. J. Med. 346, 896–903 (2002).

  11. 11.

    , & The endothelin system and its antagonism in chronic kidney disease. J. Am. Soc. Nephrol. 17, 943–955 (2006).

  12. 12.

    , & Endothelin 1 in cancer: biological implications and therapeutic opportunities. Nat. Rev. Cancer 13, 637–651 (2013).

  13. 13.

    et al. Pharmacological characterization of bosentan, a new potent orally active nonpeptide endothelin receptor antagonist. J. Pharmacol. Exp. Ther. 270, 228–235 (1994).

  14. 14.

    et al. Pathophysiological role of endothelin revealed by the first orally active endothelin receptor antagonist. Nature 365, 759–761 (1993).

  15. 15.

    et al. The discovery of nonpeptide endothelin receptor antagonists. Progression towards bosentan. Chimia 50, 519–524 (1996).

  16. 16.

    Pulmonary arterial hypertension: on the way to a manageable disease. Curr. Opin. Investig. Drugs 9, 957–962 (2008).

  17. 17.

    Pulmonary arterial hypertension: a rare disease that encourages the development of multiple treatments. Expert Opin. Orphan Drugs 2, 1137–1145 (2014).

  18. 18.

    et al. Digital ulcers in systemic sclerosis: prevention by treatment with bosentan, an oral endothelin receptor antagonist. Arthritis Rheum. 50, 3985–3993 (2004).

  19. 19.

    , , , & Distinct ETA receptor binding mode of macitentan as determined by site directed mutagenesis. PLoS One 9, e107809 (2014).

  20. 20.

    , , , & Slow receptor dissociation kinetics differentiate macitentan from other endothelin receptor antagonists in pulmonary arterial smooth muscle cells. PLoS One 7, e47662 (2012).

  21. 21.

    , , , & The effect of an endothelin-receptor antagonist, bosentan, on blood pressure in patients with essential hypertension. N. Engl. J. Med. 338, 784–790 (1998).

  22. 22.

    et al. The discovery of N-[5-(4-bromophenyl)-6-[2-[(5-bromo-2-pyrimidinyl)oxy]ethoxy]-4-pyrimidinyl]-N′-propylsulfamide (macitentan), an orally active, potent dual endothelin receptor antagonist. J. Med. Chem. 55, 7849–7861 (2012).

  23. 23.

    & Ambrisentan, a non-peptide endothelin receptor antagonist. Cardiovasc. Drug Rev. 24, 63–76 (2006).

  24. 24.

    et al. Sitaxsentan therapy for pulmonary arterial hypertension. Am. J. Respir. Crit. Care Med. 169, 441–447 (2004).

  25. 25.

    , & From bosentan (Tracleer®) to macitentan (Opsumit®): the medicinal chemistry perspective. Bioorg. Med. Chem. Lett. 26, 3381–3394 (2016).

  26. 26.

    , & Macitentan for the treatment of pulmonary arterial hypertension. Vasc. Health Risk Manag. 10, 665–673 (2014).

  27. 27.

    et al. Activation mechanism of endothelin ETB receptor by endothelin-1. Nature 537, 363–368 (2016).

  28. 28.

    et al. Chronic effects of an orally active selective endothelin-B-receptor antagonist in experimental congestive heart failure. J. Cardiovasc. Pharmacol. 36, S323–S326 (2000).

  29. 29.

    , , , & Thermostabilization of the human endothelin type B receptor. J. Mol. Biol. 428, 2265–2274 (2016).

  30. 30.

    , & A fluorescence-detection size-exclusion chromatography-based thermostability assay for membrane protein precrystallization screening. Structure 20, 1293–1299 (2012).

  31. 31.

    , & An ion-responsive motif in the second transmembrane segment of rhodopsin-like receptors. Amino Acids 35, 1–15 (2008).

  32. 32.

    et al. Allosteric sodium in class A GPCR signaling. Trends Biochem. Sci. 39, 233–244 (2014).

  33. 33.

    et al. Structural basis for allosteric regulation of GPCRs by sodium ions. Science 337, 232–236 (2012).

  34. 34.

    et al. The 2.1 Å resolution structure of cyanopindolol-bound β1-adrenoceptor identifies an intramembrane Na+ ion that stabilises the ligand-free receptor. PLoS One 9, e92727 (2014).

  35. 35.

    et al. Molecular control of δ-opioid receptor signalling. Nature 506, 191–196 (2014).

  36. 36.

    et al. High-resolution crystal structure of human protease-activated receptor 1. Nature 492, 387–392 (2012).

  37. 37.

    Metal-ligand geometry relevant to proteins and in proteins: sodium and potassium. Acta Crystallogr. D Biol. Crystallogr. 58, 872–874 (2002).

  38. 38.

    et al. Structure of the agonist-bound neurotensin receptor. Nature 490, 508–513 (2012).

  39. 39.

    , , & Crystal structure of the human OX2 orexin receptor bound to the insomnia drug suvorexant. Nature 519, 247–250 (2015).

  40. 40.

    et al. Structure of the nociceptin/orphanin FQ receptor in complex with a peptide mimetic. Nature 485, 395–399 (2012).

  41. 41.

    , , , & Human endothelin receptor ETB. Amino acid sequence requirements for super stable complex formation with its ligand. J. Biol. Chem. 269, 7509–7513 (1994).

  42. 42.

    et al. Aspartate mutation distinguishes ETA but not ETB receptor subtype-selective ligand binding while abolishing phospholipase C activation in both receptors. FEBS Lett. 361, 243–249 (1995).

  43. 43.

    et al. Pharmacological characterization of A-127722: an orally active and highly potent ETA-selective receptor antagonist. J. Pharmacol. Exp. Ther. 276, 473–481 (1996).

  44. 44.

    et al. 2,4-Diarylpyrrolidine-3-carboxylic acids—potent ETA selective endothelin receptor antagonists. 1. Discovery of A-127722. J. Med. Chem. 39, 1039–1048 (1996).

  45. 45.

    et al. Endothelin-1 analogues substituted at both position 18 and 19: highly potent endothelin antagonists with no selectivity for either receptor subtype ETA or ETB. J. Med. Chem. 36, 4087–4093 (1993).

  46. 46.

    et al. Alanine scan of endothelin: importance of aromatic residues. Peptides 15, 703–708 (1994).

  47. 47.

    , & Structure-activity studies on endothelin (16-21), the C-terminal hexapeptide of the endothelins, in the guinea-pig bronchus. Br. J. Pharmacol. 101, 232–234 (1990).

  48. 48.

    , , & A structure-affinity study of the amino-acid side-chains in neurotensin—N and C-terminal deletions and Ala-scan. Bioorg. Med. Chem. Lett. 3, 949–952 (1993).

  49. 49.

    et al. Address and message sequences for the nociceptin receptor: a structure-activity study of nociceptin-(1-13)-peptide amide. J. Med. Chem. 40, 1789–1793 (1997).

  50. 50.

    et al. Structure-activity relationship of endothelin: importance of charged groups. Biochem. Biophys. Res. Commun. 163, 424–429 (1989).

  51. 51.

    , & Peptidomimetics, a synthetic tool of drug discovery. Curr. Opin. Chem. Biol. 12, 292–296 (2008).

  52. 52.

    Designing peptide receptor agonists and antagonists. Nat. Rev. Drug Discov. 1, 847–858 (2002).

  53. 53.

    & Crystallizing membrane proteins using lipidic mesophases. Nat. Protoc. 4, 706–731 (2009).

  54. 54.

    Xds. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2010).

  55. 55.

    et al. Remote access and automation of SPring-8 MX beamlines. AIP Conf. Proc. 1741, 050021 (2016).

  56. 56.

    Processing of X-ray snapshots from crystals in random orientations. Acta Crystallogr. D Biol. Crystallogr. 70, 2204–2216 (2014).

  57. 57.

    et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).

  58. 58.

    , , & Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010).

  59. 59.

    et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).

  60. 60.

    et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21 (2010).

  61. 61.

    et al. TGFα shedding assay: an accurate and versatile method for detecting GPCR activation. Nat. Methods 9, 1021–1029 (2012).

  62. 62.

    , , , & Interactions of endothelin receptor subtypes A and B with Gi, Go, and Gq in reconstituted phospholipid vesicles. Biochemistry 38, 3090–3099 (1999).

  63. 63.

    et al. Purification of an endothelin receptor from human placenta. Biochem. Biophys. Res. Commun. 167, 251–257 (1990).

  64. 64.

    et al. Molecular cloning and characterization of the major endothelin receptor subtype in porcine cerebellum. Mol. Pharmacol. 41, 465–473 (1992).

Download references

Acknowledgements

We thank the members of the Nureki lab and the beamline staff at BL32XU of SPring-8 (Sayo, Japan) for technical assistance during data collection. We also thank Kowa Co., Ltd., for providing K-8794. pCAGGS expression plasmid vector was a kind gift from J. Miyazaki (Osaka University, Osaka, Japan). The diffraction experiments were performed at SPring-8 BL32XU (proposals 2015A1024, 2015A1057, 2015B2024, and 2015B2057). This work was supported by JSPS KAKENHI grants 16K07172 (T.D.), 26640102 (T.D.), 16H06294 (O.N.), 15H05775 (F.Y.), 15J09780 (S.W.), 17J30010 (S.W.), 17H05000 (T.N.) and 15H06862 (K.Y.), the Core Research for Evolutional Science, PRESTO from the Japan Science and Technology (JST) Technology Program; the Platform for Drug Discovery, Information, and Structural Life Science from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; the Japan Agency for Medical Research and Development (AMED); and the National Institute of Biomedical Innovation. A.I. was funded by JST, PRESTO (grant JPMJPR1331), and the PRIME from AMED. J.A. received funding from AMED-CREST and AMED, and a MEXT Grant-in-Aid for Scientific Research on Innovative Areas (grant 15H05897).

Author information

Author notes

    • Akiko Okuta

    Deceased.

Affiliations

  1. Department of Basic Medicinal Sciences, Graduate School of Pharmaceutical Sciences, Nagoya University, Nagoya, Japan.

    • Wataru Shihoya
    •  & Yoshinori Fujiyoshi
  2. Cellular and Structural Physiology Institute, Nagoya University, Nagoya, Japan.

    • Wataru Shihoya
    • , Akiko Okuta
    • , Kazutoshi Tani
    •  & Yoshinori Fujiyoshi
  3. Department of Biophysics and Biochemistry, Graduate School of Science, The University of Tokyo, Tokyo, Japan.

    • Wataru Shihoya
    • , Tomohiro Nishizawa
    •  & Osamu Nureki
  4. Precursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency, Kawaguchi, Japan.

    • Tomohiro Nishizawa
    • , Asuka Inoue
    •  & Kunio Hirata
  5. RIKEN SPring-8 Center, Sayo, Japan.

    • Keitaro Yamashita
    •  & Kunio Hirata
  6. Graduate School of Pharmaceutical Sciences, Tohoku University, Sendai, Japan.

    • Asuka Inoue
    • , Francois Marie Ngako Kadji
    •  & Junken Aoki
  7. Japan Agency for Medical Research and Development, Core Research for Evolutional Science and Technology (AMED-CREST), Tokyo, Japan.

    • Junken Aoki
  8. Department of Biophysics, Graduate School of Science, Kyoto University, Kyoto, Japan.

    • Tomoko Doi

Authors

  1. Search for Wataru Shihoya in:

  2. Search for Tomohiro Nishizawa in:

  3. Search for Keitaro Yamashita in:

  4. Search for Asuka Inoue in:

  5. Search for Kunio Hirata in:

  6. Search for Francois Marie Ngako Kadji in:

  7. Search for Akiko Okuta in:

  8. Search for Kazutoshi Tani in:

  9. Search for Junken Aoki in:

  10. Search for Yoshinori Fujiyoshi in:

  11. Search for Tomoko Doi in:

  12. Search for Osamu Nureki in:

Contributions

W.S. designed experiments; expressed, purified, and crystallized the antagonist-bound ETB receptor; collected data; and refined the structures. T.N. initially crystallized the K-8794-bound ETB receptor, assisted with the structural determination, and designed the construct ETB-Y4-mT4L. K.Y. and K.H. developed a pipeline for data collection and processing, and assisted with the structural determination. A.I., F.M.N.K., and J.A. performed and oversaw the cell-based assays. A.O. introduced K-8794 in the experimental design and characterized its pharmacology. K.T. initially designed the T4L-fused construct. T.D. performed the radiobinding assays. The manuscript was prepared by W.S., T.N., K.Y., A.I., K.H., K.T., Y.F., T.D., and O.N. Y.F., T.D., and O.N. supervised the research.

Competing interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to Tomoko Doi or Osamu Nureki.

Integrated supplementary information

Supplementary information

PDF files

  1. 1.

    Supplementary Text and Figures

    Supplementary Figures 1–7 and Supplementary Note 1

  2. 2.

    Life Sciences Reporting Summary

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nsmb.3450

Further reading