Cysteinyl leukotrienes LTC4, LTD4, and LTE4 are lipid mediators of inflammation acting via two G protein-coupled receptors (GPCRs), cysteinyl leukotriene receptor type 1 (CysLT1R) and type 2 (CysLT2R)1. While LTD4 is the favored endogenous ligand for CysLT1R2, CysLT2R responds equally to LTC4 and LTD43. CysLTRs exhibit bronchoconstrictive and pro-inflammatory effects and, therefore, have been recognized for their role in asthma, allergic rhinitis, cardiovascular diseases, and cancers4,5,6,7. Several selective CysLT1R antagonists, such as zafirlukast, pranlukast, and montelukast, have been approved as antiasthmatic drugs, however, a large fraction of patients does not respond to this therapy8. The different expression profiles, tissue distribution, and sensitivity to endogenous ligands for CysLTRs, their heterodimerization and cross regulation9,10,11 as well as the prevalence of asthma-associated polymorphisms in CysLT2R12,13 suggest distinct roles for each receptor subtype in physiology and pathology. Based on an LTC4-induced animal asthma model, it was proposed that CysLT2R-selective or dual antagonists may improve treatments of severe asthma cases14. Furthermore, selective inhibition of CysLT2R predominantly expressed in cardiovascular and brain tissues has shown remedial effects in ischemic conditions and acute brain injuries15. The development of more efficient therapies against asthma and related diseases is hampered by the lack of specific knowledge about selectivity and functional mechanisms of CysLTRs, which requires high-resolution structural data. Here, we describe four crystal structures of CysLT2R in complex with three dual CysLT1R/CysLT2R antagonists (cpds 11a–c, Supplementary Fig. 1, Supplementary Methods) and the results of extensive mutagenesis and computer modeling studies. Along with recently reported structures of CysLT1R in complex with zafirlukast and pranlukast16, we now have a complete structural view of receptors mediating action of cysteinyl leukotrienes in their inhibited, inactive state.


CysLT2R structure determination

To facilitate crystallization, human CysLT2R was modified by truncating N- and C-termini, inserting a thermostabilized apocytochrome b562RIL17 into the intracellular loop 3 (ICL3), and introducing three stabilizing point mutations18: W511.45V, D842.50N, and F1373.51Y (superscript refers to the Ballesteros–Weinstein GPCR residue numbering scheme19). The engineered receptor was crystallized in lipidic cubic phase (LCP)20 in complex with three antagonists: ONO-2570366 (cpd 11a) (2.4 Å resolution in two different space groups), ONO-2770372 (cpd 11b; 2.7 Å), and ONO-2080365 (cpd 11c; 2.7 Å) (Supplementary Figs. 14 and Supplementary Table 1). To validate the structures and probe the role of key residues, involved in ligand binding and receptor function, we conducted cell surface expression and IP1 stimulation and inhibition assays with a set of 24 mutants (Supplementary Figs. 5 and 6 and Table 1).

Table 1 Signaling and cell surface expression data for CysLT2R.

Overall architecture of CysLT2R

All CysLT2R structures adopt the canonical seven-transmembrane helical bundle architecture (Fig. 1a) and are structurally similar to CysLT1R-pranlukast16 (Supplementary Table 2). Overall CysLT2R conformations are identical to each other (Supplementary Table 2), except for the structure with cpd 11c, which is described below. Our further analysis, therefore, is focused on the highest resolution CysLT2R-11a structure, unless noted otherwise. Extracellular loop 2 (ECL2) in CysLT2R is stabilized by the highly conserved disulfide bond21 between C1113.25 and C187ECL2. An additional disulfide bond is formed between C311.25 and C2797.27 (Fig. 1c). Notably, both TM1 and TM7 are about one helical turn shorter than in CysLT1R, resulting in a ~5 Å shift of ECL3 tip (Fig. 1b).

Fig. 1: Structure of CysLT2R.
figure 1

a Overall structure of CysLT2R-11a (C2221 space group). b Structural superposition of CysLT2R-11a (blue; C2221 space group) with CysLT1R-pranlukast (yellow). c Comparison of disulfide bridges between CysLT1R (yellow) and CysLT2R (blue). Comparison of functional motifs: NPxxY (d) P-I-F (e), and DRY (f). Membrane boundaries are shown as dashed lines in a and b.

As expected, CysLT2R structures with antagonists 11a and 11b are captured in a fully inactive state. Similar to inactive structures of CysLT1R and other receptors from the δ-branch of class A GPCRs, the P5.50-I3.40-F6.44 microswitch is found in a distinct conformation (Fig. 1e)16, previously associated with activation of receptors from other class A GPCR branches. The role of this microswitch in receptors from the δ-branch is apparently different and is likely linked to the substitution of the “toggle switch” W6.48 with F6.48, which prevents this microswitch from accessing its inactive conformation. The highly conserved D[E]R3.50Y motif, in which R3.50 is stabilized in an inactive conformation via a salt bridge with D[E]3.49, is replaced by VR3.50F in CysLT2R (Fig. 1f). As expected, restoring the canonical ionic lock by V1353.49D in CysLT2R decreases the potency of LTD4 while increasing the potency of antagonists through stabilization of the inactive conformation (Table 1). Restoration of Y in the D[E]RY motif via F1373.51Y mutation, which is also present in the crystallized construct, has no effect on the potency of LTD4 or antagonists. Similarly, the stabilizing mutation W511.45V in the crystallized construct has little effect on ligand binding and receptor signaling. Finally, the third crystallization construct mutation D842.50N, a known stabilizing mutation in the conserved in class A GPCRs sodium-binding pocket22,23,24, abolishes LTD4-stimulated IP1 production in CysLT2R, similar to its effect in other receptors25. Likewise, N2977.45C in the sodium-binding pocket results in a complete loss of signaling activity. Mutating N3017.49D in the conserved NP7.50xxY motif (Fig. 1d) (NPLLY in CysLT2R; DPLLY in CysLT1R) stabilizes the sodium-binding pocket and thus reduces LTD4 signaling potency 6-fold, while increasing receptor surface expression and Emax (Table 1).

Interestingly, the CysLT2R-11c structure shows a different orientation of the Y2215.58 microswitch along with a distinct conformation of the intracellular part of TM6, shifted ~5 Å outward compared with other CysLT2R structures (Supplementary Fig. 7a). Both changes are consistent with a partially active-like GPCR state26, which, however, lacks key activation-related changes in TM7 and sodium pocket. Molecular dynamics (MD) simulations show that this state is distinct from both active and inactive states and highly dynamic (Supplementary Fig. 7b, c), suggesting that CysLT2R-11c likely represents an intermediate conformational state, selected and stabilized by the crystal lattice.

Unlike CysLT1R structures, all CysLT2R structures, except for the complex with cpd 11c, possess a well-resolved intracellular amphipathic helix 8 (H8) running parallel to the membrane (Fig. 1b). While the function of H8 is not fully understood, a mounting evidence points toward its importance in the regulation of G protein and β-arrestin binding27,28. Notably, the junction between TM7 and H8 in CysLT1R contains a rare GG8.48 motif, which likely increases dynamics of H8. On the other hand, position 8.48 in CysLT2R is occupied by E3108.48, which stabilizes the junction and the inactive state by forming salt bridges with R1363.50 and K2446.32 (Fig. 1f). Removing these interactions by E3108.48A or E3108.48G results in a slightly increased potency of LTD4 in IP1 signaling assays (Table 1).

Ligand-binding pocket and ligand-receptor interactions

In all CysLT2R structures, a strong electron density for the ligand (Supplementary Fig. 4) is present inside the central cavity of the receptor that consists of residues from all seven TMs and ECL2. It has a narrow opening (~3 Å diameter) between ECLs into the extracellular space and a larger access cleft (~5 Å across) from the lipid bilayer between TM4 and TM5 (Fig. 2a). All antagonists cocrystallized with CysLT2R share the same 3,4-dihydro-2H-1,4-benzoxazine scaffold and bind in the pocket in similar conformations (root mean square deviation < 0.3 Å in the common scaffold, Fig. 2). A key anchoring residue Y1193.33, conserved in CysLTRs, forms multiple polar contacts with the benzoxazine part, carboxylic group, and amide linker of all ligands (Fig. 2b–d). Y1193.33F mutant shows decreased potencies for both LTD4 and antagonists in IP1 assay (Table 1). The N-linked carboxypropyl moiety makes salt bridges with K371.31 and H2847.32 that are specific to CysLT2R. Mutating these residues to their CysLT1R counterparts (K371.31R or H2847.32Q) drastically decreases potencies for LTD4 activation as well as inhibition by antagonists, suggesting distinct binding interactions of these ligands with CysLT1R and CysLT2R.

Fig. 2: Ligand-binding pocket of CysLT2R.
figure 2

a Sliced surface representation of the ligand-binding pocket in CysLT2R. b Binding pose of cpd 11a and details of ligand-receptor interactions. Schematic diagrams of CysLT2R interactions with cpds 11a and 11b (c) and cpd 11c (d). Residues are colored according to the effect of their mutations on the antagonist potency in IP1 signaling assays: light red—strong effect, blue—no effect, white—not tested. The outline color indicates the effect of mutations on LTD4 potency: red—strong effect, red dashed—nonresponsive mutants, blue—no effect.

The hydrophobic bottom part of the ligand-binding cleft containing the butoxybenzene group of cpd 11a is formed by side chains of TM3-TM5 and, in case of cpd 11c, extends to L1654.52 and I1664.53. Y1273.41 forms an interhelical hydrogen bond with the carbonyl oxygen of Val2085.45, stabilizing a Pro-induced kink in TM5 (Fig. 2c, d). An aromatic residue in position 3.41 at the intersection of TM3-TM5 was previously described to confer receptor stabilization29. Interestingly, mutation Y1273.41W slightly improves CysLT2R surface expression and potencies of LTD4 and cpd 11c, however, dramatically decreases the potency of cpd 11a to inhibit LTD4-induced IP1 accumulation, likely because of a clash between bulky tryptophan and 2-chloro-5-fluoro-phenyl group of cpd 11a. S1694.56 forms a hydrogen bond with the carbonyl group of L1654.52 and interacts with the fluorine atom of cpd 11c phenyl group. Mutation S1694.56A does not affect EC50 for LTD4 and IC50 for cpd 11a but moderately improves inhibition by cpd 11c.

Antagonist selectivity to CysLTR subtypes

To understand the mechanism of ligand selectivity, we performed docking of 18 derivatives of the common 3,4-dihydro-2H-1,4-benzoxazine-2-carboxylic acid scaffold30,31 with a large spectrum of CysLT1R/CysLT2R selectivity (Supplementary Table 3). Docking models of the most selective compounds in this structure-activity relationship (SAR) series30, cpd 13e (1,800-fold selective for CysLT1R) and cpd 15b (200-fold selective for CysLT2R), are shown in Fig. 3, alongside with cpd 11a (dual CysLT1R/CysLT2R), cocrystallized with CysLT2R, and pranlukast (4,500-fold selective for CysLT1R as shown in Supplementary Fig. 6a), cocrystallized with CysLT1R.

Fig. 3: Structural determinants of antagonist selectivity to CysLTR subtypes.
figure 3

a Examples of compounds used in the docking studies, with their IC50 values toward CysLT1R and CysLT2R shown in yellow and blue, respectively. IC50 values for pranlukast were obtained in this work (3.8 ± 0.7 nM (CysLT1R) and ~17,000 ± 12,000 nM (CysLT2R), expressed as mean ± s.d. of three independent experiments, tested in quadruplicate) and for other ligands were quoted from ref. 30. The common 3,4-dihydro-2H-1,4-benzoxazine-2-carboxylic acid scaffold is shown in gray. Overview of the ligand-binding pocket with the docked ligands for CysLT1R (b) and CysLT2R (c). Inserts show docking poses and details of ligand interactions with CysLT1R and CysLT2R.

SAR analysis revealed that the most important factor for CysLT2R selectivity is the length of the alkyl chain for the O-substituents (R1), where longer phenylpentyl group in cpd 15b achieves much higher CysLT2R selectivity than phenylbutyl in cpd 11a, cpd 13e, and pranlukast or phenylpropyl in some other compounds such as cpd 15a (Supplementary Table 3). Comparison of the contacts of these substituents in CysLT1R-pranlukast and CysLT2R-11a suggests that in CysLT1R the cleft opening to the lipid membrane is restricted by a hydrophobic ridge formed by F1504.52, F1123.41, and V1965.45, while in CysLT2R the replacement F4.52L removes this restriction, making the cleft more open. Accordingly, docking of cpd 15b into CysLT1R results in a strained alkyl chain and a clash of the terminal phenyl group with F1504.52, while in CysLT2R the phenyl group readily extends outside of the cleft (Fig. 3b, c). Moreover, the phenylbutyl group in this and other scaffolds tolerates methyl and halogen decorations in the ortho and meta positions, which enables tuning pharmacological properties of the ligand such as solubility and stability, as exemplified by the development of gemilukast32.

SAR analysis of the N-substituent (R2) suggests that its length as well as the presence of a carboxyl group in this scaffold has critical influence on IC50 values for both CysLTRs. Indeed, docking of cpd 13e, the most selective antagonist in this series, shows that the oxo-pentanoic-acid moiety of this ligand forms a hydrogen bond with Y261.35, while CysLT2R has F411.35 at this position and cannot form a hydrogen bond with the ligand (Fig. 3b, c). Further elongation of this derivative chain is limited by the size of this subpocket. Interestingly, removal of the carbonyl group, as in cpds 14a-c and 15b, shifts selectivity toward CysLT2R, suggesting that a flexible carboxy-alkyl chain is favored for this receptor30. Altogether, CysLT1R and CysLT2R crystal structures provide atomic level insights into the mechanisms of ligand recognition and subtype selectivity. This knowledge should contribute to the rational design of more efficient antagonists with improved affinity/efficacy or subtype selectivity profiles.

Structural insights into CysLT2R disease-related mutations

Finally, our structures provide rational explanations of the two most common disease-associated single-nucleotide variants (SNVs) in CysLT2R: M2015.38V, related to atopic asthma13,33, and the oncogenic L1293.43Q mutation34,35. M2015.38 together with M1724.59, L1734.60, and L1985.35 define the shape of the hydrophobic part of the ligand-binding pocket. Substitutions of L1985.35 with alanine or M2015.38 with alanine or leucine result in nonresponsive mutants that bind LTD4 but fail to stimulate IP1 production. In contrast to the alanine or leucine substitution, the atopic asthma-associated variant M2015.38V still responds to LTD4 stimulation. However, this mutation significantly decreases LTD4 potency and efficacy to induce IP1 accumulation when compared with the wild-type CysLT2R (Table 1). These results along with a similar effect of N2025.39H suggest the importance of ligand-dependent TM5 displacement in CysLT2R activation. Indeed, all three TM5 residues (L1985.35, M2015.38, and N2025.39) that are important for potency interact with the benzamide core of antagonists, which distinguish them from agonists, and thus likely modulate TM5 conformation and dynamics that control activation (Fig. 4a).

Fig. 4: Naturally occurring missense SNVs, mapped on the CysLT2R structure.
figure 4

a M2015.38V polymorphism, associated with atopic asthma. b L1293.43Q mutation, related to uveal melanoma and blue nevi. c SNVs from the ExAC database and L1293.43, colored according to their location: ligand-binding pocket (red), microswitches (blue), sodium site (green), and G protein and β-arrestin-binding interface (yellow).

The second disease-relevant SNV, L1293.43Q, has been associated with uveal melanoma and blue nevi34,35,36. A hydrophobic amino acid is present in this position in 97% of class A receptors, most frequently L3.43 (73%), but also M3.43 as in CysLT1R. Located at the bottom of the sodium pocket, a large hydrophobic side chain in position 3.43 is part of a hydrophobic layer, which is important for stability of the inactive state. Mutation of L3.43 to a polar residue or to a small alanine residue can disrupt the hydrophobic layer (Fig. 4b, c), facilitating water and sodium passage37,38 and leading to receptor activation. Indeed, it was shown that mutation in position 3.43 to R, K, A, E, or Q induces constitutive activation in several receptors39, often resulting in distinct physiological disorders. In CysLT2R, we found that L1293.43Q displays constitutive activity for the Gq pathway with a fourfold increase in basal IP1 accumulation and is unresponsive to LTD4 stimulation (Supplementary Fig. 6b, c).

Further, we evaluated naturally occurring missense SNVs in CysLT2R from over 60,000 healthy individuals assembled in the exome aggregation consortium (ExAC) database40. Structural mapping of 117 SNV positions revealed that nine of them belong to the ligand-binding pocket, seven are activation-related microswitches or located in the sodium-binding site, and nine reside on the G protein and β-arrestin-binding interface (Fig. 4c), all of which could dramatically affect the receptor function41. Unlike the relatively frequent polymorphisms M2015.38V or L1293.43Q, most ExAC mutations are very rare (minor allele frequency < 10−4), and, therefore, it has not been possible yet to associate them with higher risk of asthma or other pathologies.


Compared with CysLT1R, which was successfully targeted by antiasthmatic drugs 20 years ago, the role of CysLT2R in physiology and pathogenesis of inflammation related processes is more complex and remains less understood4,42. Recently accumulated results suggest that CysLT2R-selective or CysLT1R/CysLT2R dual antagonists may offer more efficient alternatives to currently used CysLT1R-selective antagonists, especially for the treatment of severe asthma43,44. In addition, CysLT2R is arising as a promising drug target against brain injury and neurodegenerative disorders5,45. High constitutive Gq signaling activity of CysLT2R mutants has been associated with occurrence of uveal melanoma and other cancer types35, however, the role of CysLT2R in cancer remains controversial as its high expression levels have been correlated with antitumorigenic activity46.

The CysLT2R structures described in this study along with the structures of CysLT1R16 reveal important determinants of ligand binding and selectivity between these two receptors. Thus, our docking studies recapitulate binding of dozens of known ligands and allow to explain SAR for a series of 3,4-dihydro-2H-1,4-benzoxazine-2-carboxylic acid scaffold derivatives. These structures will serve as templates for rational design of a new generation of potent antagonists with desired selectivity profiles (receptor selective or dual), which could be further developed into efficient drug candidates or tool compounds, helping to decipher the specific role of each of the CysLT receptor subtype in various physiological processes and pathologies.

Our study also provides a key insight into structure and function of the intracellular H8 in CysLT receptors. While both receptors possess a canonical H8 amphipathic motif, this helix is well resolved in CysLT2R structure, but not observed in CysLT1R (Fig. 1). The difference is that the junction between TM7 and H8 in CysLT1R contains a very flexible GG8.48 motif, while CysLT2R has GE8.48 in the same position. Importantly, G8.48S mutation in CysLT1R is a known disease mutation that increases efficacy of the receptor signaling16,47, likely due to the improved stability of H8, known to be involved in regulation of G protein and β-arrestin binding27,28. Interestingly, the E3108.48 side chain in CysLT2R serves a special role, forming salt bridges with R1363.50 and K2446.32 and thus stabilizing the inactive receptor state. Introduction of Glu in positions 8.48 and 8.49 has been recently shown beneficial effects on stability of the inactive state in several GPCRs, including CB248 and CCR549.

Another promising application for structural information obtained in this study is the ability to rationalize effects of specific SNVs on receptor function. We mapped naturally occurring missense SNVs from 60,000 healthy individuals on the CysLT2R structure and observed that about quarter of them are located in functionally important regions, which may affect signaling40. Continuing increase in structural coverage of the GPCR superfamily combined with rapid accumulation of genome sequencing data and structure-function studies should enable reliable predictions of disease associations and effects of natural missense variants on drug efficacy and safety profiles, advancing us toward the realm of personalized medicine.


Protein engineering for structural studies

The wild-type DNA encoding human cysteinyl leukotriene receptor 2 (UniProt Q9NS75) was purchased from cDNA Resource Center ( and cloned into a modified pFastBac1 vector (Invitrogen) containing an expression cassette with a haemagglutinin signal sequence, FLAG tag, 10 × His tag followed by TEV protease cleavage site on the N-terminus. Amino acids 1–16 from the N-terminus and 323–346 from C-terminus were deleted by overlap extension PCR. Thermostabilized apocytochrome b562RIL (BRIL) from Escherichia coli with mutations M7W, H102I, and R106L was inserted into the ICL3 between the residues E232 and V240 by overlap extension PCR. Three point mutations, W511.45V, D842.50N, and F1373.51Y, designed using a sequence dissimilarity approach18, were further introduced to improve receptor surface expression in Spodoptera frugiperda Sf9 cells (Novagen, cat. 71104) as well as its stability and yield. Sequences of all primers used in this work are listed in Supplementary Table 4. The full DNA sequence of the CysLT2R crystallization construct is provided in Supplementary Table 5.

Protein expression and purification

Bac-to-Bac baculovirus expression system (Invitrogen) was used to obtain high-titer recombinant baculovirus (>3 × 108 viral particles per ml). Sf9 insect cells were infected at densities (2–3) × 106 cells per ml culture at multiplicity of infection of 5–10. BayCysLT2 ligand (Cayman Chemical) was dissolved in DMSO to 25 mM and added to the cell culture at the final concentration of 3 µM at the time of infection. Cells were harvested 48–50 h post infection by gentle centrifugation at 2,000 × g and stored at −80 °C until use.

Cells were thawed and lysed by repetitive washes in hypotonic buffer (10 mM HEPES pH 7.5, 20 mM KCl, and 10 mM MgCl2) and high osmotic buffer (10 mM HEPES pH 7.5, 20 mM KCl, 10 mM MgCl2, and 1 M NaCl) with addition of protease inhibitor cocktail (500 µM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (Gold Biotechnology), 1 µM E-64 (Cayman Chemical), 1 µM leupeptin (Cayman Chemical), 150 nM aprotinin (A.G. Scientific)). Membranes were then resuspended in 10 mM HEPES pH 7.5, 20 mM KCl, 10 mM MgCl2, 2 mg ml−1 iodoacetamide, protease inhibitors, and 25 µM ligand for 30 min at 4 °C and then solubilized by addition of 2× buffer (300 mM NaCl, 2% of n-dodecyl-β-D-maltopyranoside (DDM; Avanti Polar Lipids) 0.4% of cholesteryl hemisuccinate (CHS; Sigma), 10% glycerol) and incubation for 3.5 h at 4 °C. All further purification steps were performed at 4 °C. Supernatant was clarified by centrifugation and bound to TALON IMAC resin (Clontech) overnight in presence of 20 mM imidazole and NaCl added up to 800 mM. The resin was then washed with ten column volumes (CV) of wash buffer I (8 mM ATP, 100 mM HEPES pH 7.5, 10 mM MgCl2, 500 mM NaCl, 15 mM imidazole, 10 μM ligand, 10% glycerol, 0.1/0.02% DDM/CHS), then with five CV of wash buffer II (25 mM HEPES pH 7.5, 500 mM NaCl, 30 mM imidazole, 10 μM ligand, 10% glycerol, 0.015/0.003% DDM/CHS), then buffer was exchanged into buffer III (25 mM HEPES pH 7.5, 500 mM NaCl, 10 mM imidazole, 10 μM ligand, 10% glycerol, 0.05/0.01% DDM/CHS) and the protein-containing resin was treated with PNGase F (Sigma) for 5 h. Resin was further washed with five CV of wash buffer III and eluted with (25 mM HEPES pH 7.5, 250 mM NaCl, 400 mM imidazole, 10 μM ligand, 10% glycerol, 0.05/0.01% DDM/CHS) in several fractions. Fractions containing target protein were desalted from imidazole using PD10 desalting column (GE Healthcare) and incubated with 50 µM ligand and a His-tagged TEV protease (homemade) overnight to remove the N-terminal tags. Reverse IMAC was performed the following day and protein was concentrated up to 40–60 mg ml−1 using a 100 kDa molecular weight cut-off concentrator (Millipore). The protein purity was checked by SDS-PAGE, and the protein yield and monodispersity were estimated by analytical size exclusion chromatography.

LCP crystallization

Purified and concentrated CysLT2R was reconstituted in LCP, made of monoolein (Nu-Chek Prep) supplemented with 10% (w/w) cholesterol (Affymetrix) in 2:3 protein:lipid ratio using a lipid syringe mixer20. Transparent LCP mixture was dispensed onto 96-wells glass sandwich plates (Marienfeld) in 25–40 nl drops and covered with 800 nl precipitant using an NT8-LCP robot (Formulatrix). All LCP manipulations were performed at room temperature (RT, 20–23 °C), and plates were incubated and imaged at 22 °C using an automated incubator/imager (RockImager 1000, Formulatrix). Crystals of CysLT2R-11a_C2221 grew to their full size within 3 weeks in a precipitant containing 100–200 mM NH4 tartrate dibasic, 28–32% v/v PEG400, and 100 mM HEPES pH 8.0; CysLT2R-11a_F222 for 3 weeks in a precipitant containing 30 mM NH4 tartrate dibasic, 24% PEG400, and 100 mM HEPES 7.0; CysLT2R-11b for 3 weeks in a precipitant containing 210 mM NH4 tartrate dibasic, 29% PEG400, and 100 mM HEPES 7.0; and CysLT2R-11c for 1 week in a precipitant containing 100 mM K formate, 30% v/v PEG400, and 100 mM TRIS-HCl pH 8.0. Crystals were harvested from LCP using 75–200 µm MiTeGen micromounts and flash-frozen in liquid nitrogen.

Diffraction data collection and structure determination

X-ray diffraction data were collected at the European Synchrotron Radiation Facility (ESRF, Grenoble, France) beamlines ID23–1, ID29, ID30b, and ID30a3, equipped with PILATUS3 6M, PILATUS3 6M-F, or Eiger X 4M detectors, using the X-ray wavelengths in range 0.96770–1.07234 Å and the beam size between 15 and 30 μm. In case of CysLT2R-11c, four partial (70–80°) datasets with oscillation 0.2° and three partial 20° datasets with oscillation 0.1° were collected and combined to obtain a complete final dataset. The exposure was calculated using RADDOSE50 based on a dose of 20 MGy per dataset, as implemented in BEST51. For CysLT2R-11a and CysLT2R-11b, partial datasets of 5–15° per crystal with oscillations of 0.1–0.15° per image and the exposure time set to reach 20 MGy dose for each partial dataset were collected following a raster scanning of each crystal and selection of best diffraction spots using DOZOR scoring52 and manual inspection of diffraction images. Data were integrated using XDS, scaled and merged with XSCALE53, nonisomorphous datasets were rejected using CC1/2-based clustering as previously described52. The structure was determined by molecular replacement using phenix.phaser54 with the receptor portion of CysLT1R-pran (PDB ID 6RZ4) and BRIL of A2AAR (PDB ID 4EIY) as models for the initial cpd 11a structure, and this model was subsequently used as the molecular replacement search model for the three other structures. Initial refinement rounds were performed using autoBUSTER55 and at later stages with phenix.refine56, followed with manual examination and rebuilding with COOT54 using both 2mFo-DFc and mFo-DFc maps. Final data collection and refinement statistics are shown in Supplementary Table 1.

Plasmids for functional assays

For CysLT2R functional assays, the initial CysLT2R wild-type gene with an N-terminal 3 × HA tag cloned into pcDNA3.1+ (Invitrogen) at EcoRI(5′) and XhoI(3′) was purchased from All further gene modifications (point mutations, truncations, or partner protein fusion) were introduced by overlapping PCR. Sequences of all primers used in this work are listed in Supplementary Table 4.

IP1 production assay

The Cisbio IP-One kit was used according to the manufacturer’s instructions. HEK293 cells (ATCC CRL-1573) were seeded onto poly-L-Lysine-coated 384-well plates at 20,000 cells per well and transfected with 40 ng of DNA coding for the wild-type CysLT2R or for the CysLT2R mutants using the X-treme-Gene HP (Roche) agent. At 48 h post transfection, the media was removed and the cells were washed with fresh Hank’s Balanced Salt Solution. Cells were either stimulated directly with a range of LTD4 concentrations (10−12–10−6 M) prepared in IP1 stimulation buffer, or sequentially stimulated with a range of antagonist concentrations (10−11–10−5 M), and LTD4 concentrations corresponding to the EC80 for each mutant. No LTD4 degradation was observed by mass spectrometry (Supplementary Fig. 8). After equilibration for 30 min at 37 °C, the cells were lysed with IP1-D2 and Ab-Crypt reagents in lysis buffer and then incubated for 1 h at RT. Fluorescence signal was recorded on a Tecan GENios Pro plate reader using an HTRF filter set (λex 320 nm, λem 620 and 655 nm). Data were plotted using the three parameters EC50/IC50 fit in GraphPad Prism 7 (San Diego, CA) and represent the mean ± s.d. of at least two independent experiments performed in quadruplicate.

Quantification of LTD4 degradation in IP1 assay

Potential conversion of LTD4 into LTC4 or LTE4 was checked by ultra-performance liquid chromatography coupled to mass spectrometry (UPLC/MS). HEK293 cells were seeded in a 6-well plate at a density of 300,000 cells per well. Forty-eight hours after seeding, medium was removed, and cells were washed twice with PBS. Then cells were incubated in stimulation buffer used for IP-One assays (Krebs buffer containing LiCl as an inhibitor of IP1 degradation) alone or containing 10 µM LTD4 or 10 µM LTD4 and 10 mM l-Cysteine (used as an inhibitor of LTD4 conversion) for 30 min at 37 °C. After incubation, supernatant was filtered through a 0.22 µm PVDF filter and an internal standard was added before injection on UPLC/MS (Waters UPLC system coupled with a SQ detector 2 and a PDA eλ detector, using an Acquity UPLC BEH C18 column, 2.1 mm × 50 mm, 1.7 μm spherical size). UPLC chromatograms were recorded using the following gradient: water + 0.1% TFA and acetonitrile (0 → 0.2 min, 5% acetonitrile; 0.2 → 1.5 min, 5% → 95%; 1.5 → 1.8 min, 95%; 1.8 → 2.0 min, 95% → 5%; and 2.0 → 2.5 min, 5%). Quantification was done by determining the area under the curve (AUC) ratio of the tested compound over AUC of the internal standard. 0% was determined by using results from stimulation buffer alone and 100% was determined by using results from stimulation buffer containing LTD4 but without incubation over the cell monolayer. Quantification results are expressed as mean ± s.e.m. of three independent experiments.

Cell surface expression determined by ELISA

HEK293 cells were seeded in 24-well plates coated with poly-L-Lysine (Sigma) at 100,000 cells per well and transfected with 375 ng of plasmid coding for the wild-type or mutant CysLT2R using X-treme-Gene HP (Roche). Forty-eight hours after transfection, cells were fixed with 3.7% (v/v) formaldehyde in Tris-buffered saline (TBS, 20 mM Tris-HCl, pH 7.5, and 150 mM NaCl) for 5 min at RT. Cells were washed three times with TBS and incubated for 1 h in TBS supplemented with 3% (w/v) fat-free milk in order to block nonspecific binding sites. A mouse monoclonal anti-HA antibody coupled to HRP (Roche) was added at 1:1000 dilution in TBS-3% fat-free dry milk for 3 h at RT. Following incubation, cells were washed twice with TBS before the addition of 250 µl of 3,3′,5,5′-Tetramethylbenzidine (Sigma). Plates were incubated for 15 min at RT and the reaction was stopped by the addition of 250 µl of 2N HCl. Two hundred microliters of the yellow reaction was transferred into a 96-well plate and the absorbance was read at 450 nm on GENios Pro plate reader (Tecan). Cells transfected with the empty pcDNA3.1+ vector (mock) were used to determine background. Data were plotted using GraphPad Prism 7 and represent the mean ± s.d. of at least two independent experiments performed in quadruplicate.

Molecular docking

We collected 18 O- and N-derivatives of the common 3,4-dihydro-2H-1,4-benzoxazine-2-carboxylic acid scaffold from previous studies30,32, assigned charges for the ligands at pH 7.0, and generated 3D ligand structures from their 2D representations, using Monte Carlo optimization and the MMFF-94 force field. We preprocessed each protein structure (CysLT1R-pranlukast, PDB ID 6RZ4; CysLT2R-11a, PDB ID 6RZ6) by adding missing residues, optimizing side-chain rotamers, and removing water molecules. Rectangular boxes enclosing ligand-binding sites of pranlukast in CysLT1R and cpd 11a in CysLT2R with an additional 8 Å margin were used as the sampling space for docking. Receptors were presented as smoothened grid potentials, while the docking simulations sampled ligand conformations in the internal coordinate space using biased probability Monte Carlo optimization57 with the sampling parameter (docking effort) set to 50. We performed at least two independent docking runs for each ligand and selected binding poses with the lowest docking score. All docking simulations were done using the ICM-Pro v3.8–6 software package (MolSoft).

MD simulations

The initial CysLT2R models for MD simulations were prepared based on the crystal structures (CysLT2R-11a, PDB ID 6RZ6, for the inactive state; CysLT2R-11c, PDB ID 6RZ8 for the intermediate state) using the ICM-Pro molecular modeling package (v3.8–6). First, BRIL-fusions and all hetero atoms were removed, followed by the assignments of protonation states and modeling missing side-chain residues using internal coordinate mechanics force field. Then missing loops were modeled using the loop modeling and regularization protocols available in ICM-Pro58. These preprocessed CysLT2R models were used to prepare input files for MD simulations as previously described59. Briefly, the input files were generated using the CHARMM-GUI server60. The receptor orientation was calculated by superimposing the CysLT2R structures on the CB1 receptor coordinates (PDB ID 5XRA) obtained from the OPM database61. The input simulation box had 157 POPC lipids, 11,908 water molecules, and 31 sodium and 46 chloride ions. The system was first energy minimized and then equilibrated for 10 ns, followed by ten independent production runs of 500 ns each using Gromacs (v.2018.1) simulation package62. The analysis and plotting were performed using Gromacs and matplotlib plotting packages available in Python. The MD simulations were performed on GPU enabled nodes with P100 NVIDIA cards made available by the High-Performance Computing Center at the University of Southern California.

Ligand synthesis and characterization

The overall ligand synthesis scheme is shown in Supplementary Fig. 1 and described in Supplementary Methods. Analytical samples were homogeneous as confirmed by TLC, and afforded spectroscopic results consistent with the assigned structures. Proton and carbon nuclear magnetic resonance spectra (1H and 13C NMR) were taken on a Varian Mercury 300 spectrometer using deuterated chloroform (CDCl3) and deuterated dimethylsulfoxide (DMSO-d6) as the solvent. Fast atom bombardment mass spectra were obtained on a JEOL JMS-DX303HF spectrometer. Electrospray ionization (HRMS) mass spectra was obtained on a Thermo Fisher Scientific LTQ Orbitrap XL system. Column chromatography was carried out on silica gel (Merck Silica Gel 60, Wako gel C-200, or Fuji Silysia FL60D). Thin layer chromatography was performed on silica gel (Merck TLC or HPTLC plates, Silica Gel 60 F254).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.