Structure of the N-terminal domain of a type B1 G protein-coupled receptor in complex with a peptide ligand

Abstract

The corticotropin releasing factor (CRF) family of ligands and their receptors coordinate endocrine, behavioral, autonomic, and metabolic responses to stress and play additional roles within the cardiovascular, gastrointestinal, and other systems. The actions of CRF and the related urocortins are mediated by activation of two receptors, CRF-R1 and CRF-R2, belonging to the B1 family of G protein-coupled receptors. The short-consensus-repeat fold (SCR) within the first extracellular domain (ECD1) of the CRF receptor(s) comprises the major ligand binding site and serves to dock a peptide ligand via its C-terminal segment, thus positioning the N-terminal segment to interact with the receptor's juxtamembrane domains to activate the receptor. Here we present the 3D NMR structure of ECD1 of CRF-R2β in complex with astressin, a peptide antagonist. In the structure of the complex the C-terminal segment of astressin forms an amphipathic helix, whose entire hydrophobic face interacts with the short-consensus-repeat motif, covering a large intermolecular interface. In addition, the complex is characterized by intermolecular hydrogen bonds and a salt bridge. These interactions are quantitatively weighted by an analysis of the effects on the full-length receptor affinities using an Ala scan of CRF. These structural studies identify the major determinants for CRF ligand specificity and selectivity and support a two-step model for receptor activation. Furthermore, because of a proposed conservation of the fold for both the ECD1s and ligands, this structure can serve as a model for ligand recognition for the entire B1 receptor family.

The ability of the body to adapt to stressful stimuli and the role of stress maladaptation in human diseases has been intensively investigated. Corticotropin releasing factor (CRF) (1), a 41-residue peptide, and its three paralogous peptides, urocortin (Ucn) 1, 2, and 3, play important and diverse roles in coordinating endocrine, autonomic, metabolic, and behavioral responses to stress (2, 3). CRF family peptides and their receptors are also implicated in the modulation of additional central nervous system functions including appetite, addiction, hearing, and neurogenesis and act peripherally within the endocrine, cardiovascular, reproductive, gastrointestinal, and immune systems (4, 5). CRF and related ligands initially act by binding to their G protein-coupled receptors (GPCRs). These belong to the peptide hormone B1 family (family B1 GPCRs), comprising receptors for growth hormone releasing factor, secretin, calcitonin, vasoactive intestinal peptide, glucagon, glucagon-like peptide-1, and parathyroid hormone. Two CRF receptors, CRF-R1 and CRF-R2, have been cloned in mammals (6, 7).

Structure activity studies of CRF showed that the first eight N-terminal residues of the hormone are necessary for GPCR signaling (1, 8), whereas the C-terminal (≈15) residues are important for binding (9, 10). A two-domain behavior for ligand binding was also observed for the receptors (11, 12). Analyses of mutant and chimeric receptors identified residues located mainly in the extracellular domains (ECDs) of CRF receptors that affect binding of peptide ligands (13–16). Furthermore, we identified the major peptide binding site to be the ECD1 of CRF receptors (16, 17) and determined the 3D NMR structure of the ECD1 of CRF-R2β (18). The polypeptide fold of the ECD1–CRF-R2β is known as a short-consensus-repeat (SCR), which is a common multiple protein-binding module in the Ig superfamily (19). To obtain a detailed molecular understanding of peptide–receptor interaction, we investigated the 3D NMR structure of the complex of ECD1 of CRF-R2β with the peptide hormone antagonist, astressin.

Results

Refined 3D Structure of ECD1 of CRF-R2β.

The NMR structure of ECD1–CRF-R2β comprising amino acids 39–133 at pH 5 at 30°C was determined by using the standard procedure [see supporting information (SI) Text]. The NMR structure of ECD1–CRF-R2β at pH 5 is better defined than at pH 7.4 (18) because several additional amide proton resonances (i.e., residues 84, 86, and 87) were observed, probably because of slower hydrogen exchange at lower pH. The good quality of the 3D structure is represented by the small rmsd of 0.75 Å for residues 57–83 and 98–119 (Fig. 1A and SI Table 1). The 3D structure of ECD1–CRF-R2β at pH 5 is an SCR motif similar to the fold at pH 7.4 (18) with three disulfide bonds (Cys-45–Cys-70, Cys-60–Cys-103, and Cys-84–Cys-118) (20), and two antiparallel β-sheet regions from residues 63–64 (β1-strand), 70–71 (β2-strand), 79–82 (β3-strand), and 99–102 (β4-strand) (18) (Fig. 1A). The central core consists of a salt bridge involving Asp-65–Arg-101, sandwiched between the aromatic rings of Trp-71 and Trp-109. The 3D structure has also two disordered segments comprising residues 45–58 (loop 1) and residues 84–98 (loop 2). Because of the small line-width of the cross-peaks in the [¹⁵N,¹H]-TROSY spectra and the small number of NOEs observed both at pH 5 and at pH 7.4 (data not shown), the disordered loop 1 seems to be of flexible nature and comprises a high sequence variability within the CRF receptor family (18). In contrast, the disordered loop 2 is, in part, highly conserved in the CRF family (Phe-88, Gly-90, Tyr-93, Asn-94, and Thr-96) (SI Fig. 4) and undergoes slow conformational exchange as evidenced by very broad (Tyr-87, Gly-90, and Ile-91) or missing (Phe-88, Asn-89, Lys-92, and Arg-97) cross-peaks in the [¹⁵N,¹H]-TROSY spectrum (Fig. 2B). The slow conformational exchange is estimated to be on the order of 10⁻² seconds corresponding to an energy barrier of 15 kcal/mol. Furthermore, two sets of resonances of almost equal intensities are observed for Pro-83 and Pro-85, indicating cis/trans isomerization, which is a slower exchange process on the order of 100 seconds and corresponds to a larger energy barrier of ≈20 kcal/mol (21). The slow dynamics of loop 2 is interesting, because it is involved in binding of the peptide hormone and the slow dynamics is suppressed in the complex (discussed below).

Fig. 1.

3D NMR structure of ECD1–CRF-R2β in complex with astressin. (A) Superposition of 20 energy-minimized conformers representing the 3D NMR structure of free ECD1–CRF-R2β (the backbone C^α atoms of residues 57–83 and 99–120 were superimposed). (B) Superposition of 20 energy-minimized conformers of ECD1–CRF-R2β in complex with astressin (the backbone C^α atoms of residues 57–120 of ECD1 and 30–41 of astressin were superimposed). The backbone of residues 44–122 of ECD1 is shown in magenta, and the backbone of residues Leu-27–Ile-41 of astressin is colored in green. In A and B the disulfide bonds are shown in yellow. (C) Ribbon diagram of the lowest energy conformer representing the 3D NMR structure of the ECD–CRF-R2β–astressin complex. The β-sheets are shown in cyan, and the side chains of the core residues Trp-71 and Trp-109 along with the disulfide bonds are shown in yellow. The salt bridge Arg-101 (in blue)–Asp-65 (in red) is shown as dashed spine in orange. The backbone of astressin from Leu-27–Ile-41 is shown in green. (D) Side view of the ribbon diagram shown in C. MOLMOL was used to generate the figures (49).

Fig. 2.

Structural and dynamical characterization of the ECD–CRF-R2β–astressin complex. (A) Superposition of the 20 conformers of free ECD1–CRF-R2β depicted as a cyan sausage with the lowest energy conformer of the complex ECD1–CRF-R2β–astressin shown as a magenta ribbon. The disulfide bonds are shown in yellow. (B) [¹⁵N,¹H]-TROSY spectra of the free (in red contours) and complex (in black contours) ECD1–CRF-R2β. Residues enclosed in rectangular boxes are from loop 2 (residues 84–98). The peaks from free ECD1–CRF-R2β are missing for most of the residues in loop 2 because of slow dynamics. (C) Surface representation of ECD1–CRF-R2β showing the residues in pink, whose chemical shift perturbations are >0.5 ppm upon complex formation with astressin (see also SI Fig. 4C). Astressin is shown in green along with the side chains interacting with the ECD1. (D) Amino acid sequence differences of mouse CRF-R1 and CRF-R2 are mapped in purple onto the ECD1 surface of the complex structure. The residues are labeled with single letter code for CRF-R1, followed by CRF-R2. Astressin is shown as a green–yellow ribbon. Highlighted in yellow are the sequence differences of CRF ligands (see also SI Fig. 6).

3D Structure of ECD1 of CRFR-2β in Complex with Astressin.

In the ECD1–CRF-R2β–astressin complex, the ECD1 retains the SCR fold observed in the free ECD1 (18) (Fig. 1 C and D) with its elliptical β-sandwich stabilized by three disulfide bridges and by the core comprising residues Trp-71 and Trp-109 and the salt bridge Asp-65–Arg-101. Furthermore, loop 1 in the complex continues to remain disordered. However, a significant difference between free and complexed ECD1 is observed for loop 2. In free ECD1, loop 2 is largely disordered and undergoes a slow conformational exchange in the millisecond and second time range and is structured upon astressin binding (Fig. 1 A and B). Furthermore, all of the resonances in the ¹⁵N-¹H backbone moieties of loop 2 were observed in the TROSY spectrum (Fig. 2B), and a single set of resonances was observed for Pro-83 and Pro-85 in the NMR spectra. Both prolines prefer to be in the trans conformation, as evidenced by the presence of NOEs between the δH protons of prolines to the αH protons of the preceding residues Arg-82 and Cys-84, respectively. Furthermore, the chemical shift perturbation data mapped onto the 3D structure of the complex shows that loop 2 is fully involved in astressin binding (Fig. 2C and SI Fig. 4C).

Astressin binds to the ECD1 in an α-helical conformation as predicted for CRF (22) (Fig. 1B). Several experimental data were observed to confirm the α-helical conformation of astressin. (i) In the [¹⁵N,¹H]-HMQC spectrum of labeled astressin bound to unlabeled ECD1, all of the nine labeled residues in the C terminus of astressin show a remarkable chemical shift change as well as line broadening upon ECD1 binding (SI Fig. 5A) indicating a conformational change from random coil to a folded entity. (ii) Continuous strong sequential and medium-range NOEs observed confirm the presence of an α-helix (23) (SI Fig. 5B). (iii) Chemical shifts observed for ¹³C^α, ¹³C^β, and ¹H^α for Leu-27 to Ile-41 are indicative of a helical conformation (24) (SI Fig. 5C). (iv) Plot of amide proton chemical shifts versus residue number (SI Fig. 5D) shows a wave pattern characteristic of an amphipathic helix for residues 27–41 of astressin (25). (v) Absence of exchange cross-peaks between the amide protons with water in the NOESY spectrum suggests that most of the amide protons of astressin are involved in hydrogen bonds. These data are reflected in the 20 conformers representing the 3D structure of ECD1–CRF-R2β–astressin complex for which continuous helical hydrogen bonds are observed from CO_i to NH_i+4 for residues 27–37 of astressin. The C-terminal four residues prefer to be in a 3₁₀-helical conformation, which is supported by the presence of αH_i-HN_i+2 NOEs and the missing Nle-38(αH)-Ile-41(βH) NOE (SI Fig. 5B). The preference for a 3₁₀ helix over an α-helix toward the C terminus of helical peptides and proteins is often reported (26). Of special interest are the hydrogen bonds between the carbonyl of Glu-39 and the C-terminal NH₂ group, and between the carbonyl of Ile-41 and the amide of Val-113 of ECD1. They stabilize the helical conformation by extending the helical hydrogen bonds toward the C terminus, which explains the presence of the C-terminal amidation for the peptide hormone to be bioactive (1).

The ECD1 interacts with astressin through its hydrophobic face of the helix comprising residues Leu-27, Ala-31, Asn-34, Arg-35, Leu-37, Nle-38, Ile-40, and Ile-41, whereas the lactam bridge between Glu-30 and Lys-33 interacts partially with the ECD1. In contrast, residues of the hydrophilic face, namely Gln-29, His-32, Lys-36, and Glu-39, are solvent-exposed, and the side chains of Lys-36 and Glu-39 can form a solvent-exposed salt bridge (as several NOEs were observed between these residues) (Fig. 3A). As a consequence of the complex formation, eight residues of astressin and 15 residues of the ECD1 bury ≈1,500 Ų of surface. There are two prominent hydrophobic interaction sites between astressin and ECD1 sandwiched between the intermolecular salt bridge involving Arg-35 of astressin and Glu-86 of ECD1 (Fig. 3A). The most crucial hydrophobic interaction is composed of the cluster of residues Leu-37, Nle-38, Ile-40, and Ile-41 of astressin with Gln-66, Ile-67, Cys-84, Phe-88, Val-113, Tyr-115, and Cys-118 of the ECD1 (Fig. 3). Nle-38 is in the center of this hydrophobic patch in close proximity to Ile-67 and Tyr-115, whose ring current shifts upfield the proton resonances of Nle-38 and Ile-67. In addition, binding studies with astressin of mutant CRF-R2β highlights the importance of this patch [K_i^Astr(I67E) = 130 (80–190) nM; K_i^Astr(Y115R) > 200 nM] (18). The other hydrophobic interaction site involves Leu-27 and the highly conserved Ala-31 of the ligand with Ile-91 of the receptor (Fig. 3A). Furthermore, four intermolecular hydrogen bonds could be formed between astressin and ECD1 surrounding the hydrophobic patch: The hydrogen bond of the backbone carbonyl of Ile-41 of astressin with the amide proton of Val-113, which is consistent with the maximum chemical shift change upon astressin binding (Fig. 3A and SI Fig. 4C); the hydrogen bond between the highly conserved C-terminal amide group of astressin and the hydroxyl group of the conserved Tyr-115, which is supported by the high field shift of ε-protons of Tyr-115 (6.31 ppm); the side chain of the highly conserved Asn-34 of the ligand is involved in a hydrogen bond with the backbone carbonyl of Phe-88 (Fig. 3B and SI Figs. 4 and 6); and finally, the carbonyl group of the lactam bridge could be involved in a hydrogen bond with the amide group in the side chain of Asn-89 because these two groups are in close proximity (Fig. 3B).

Fig. 3.

Molecular anatomy of the residues in the interaction site between astressin and ECD1–CRF-R2β. Shown are front (A) and back (B) stereo views of the side chain interactions between the ECD1 and astressin. The backbones of astressin and ECD1 are shown as green and magenta ribbons, respectively. Hydrophobic side chains are shown in yellow, hydrogen bonds are shown in cyan, and the salt bridges Arg-35–Glu-86 and Lys-36–Glu-39 are shown in gray. All of the residues in the interaction surface are marked for clarity. “X” refers to norleucine residue (Nle).

Residue-Specific Contributions to Binding Affinities Based on an Ala Scan of Ovine CRF (oCRF).

The 3D structure of ECD1–CRF-R2β in complex with astressin shows a large intermolecular interaction interface along the entire hydrophobic face of the peptide helix with residues 27, 30, 31, 33, 34, 35, 37, 38, 40, and 41 whereas residues 29, 32, 36, and 39 are solvent-exposed (Fig. 3). The roles of the residues in the interaction interface are consistent with the effects of Ala substitutions in oCRF on the biological responses of CRF-R1 (27). Substitution of Arg-35 and Leu-38 in oCRF by alanine reduced the biological potencies of the ligand dramatically, whereas Ala substitution at positions 29, 32, 36, and 39 did not reduce significantly the biological potencies when compared with oCRF (SI Table 2). Because of this good correlation between the reported 3D structure and the biological activities of a series of oCRF analogs for CRF-R1, a quantitative description of the interface between peptide hormone and ECD1 of CRF-R2β was anticipated. Hence, we measured the binding affinities of all Ala analogs of oCRF for full-length CRF-R2β and calculated the difference in Gibbs free energy ΔΔG° between oCRF and the Ala analog (SI Table 2; note: for residues 31 and 41 ΔΔG° could not be determined, because oCRF has Ala at these positions). Similar measurements were carried out for astressin with amidated or carboxylated C terminus (SI Table 2). The values of ΔΔG° vary from 4 to 5 kcal/mol for NH₂ and L38A analogs to values <1 kcal/mol for several residues (SI Table 2). Also, we have predicted the corresponding ΔΔG° terms for individual residues from the 3D structure of the ECD1–CRF-R2β–astressin complex using energy terms established in the literature (28–30) (SI Table 2). This includes typical energy values for hydrogen bond formation, salt bridge, hydrophobic interactions and helical tendencies (see also SI Table 2). Predicted ΔΔG° values based on the 3D structure of the complex correlate well with the experimentally derived values except for a small deviation for the residues at 29 and 32 and a significant deviation for 38. The latter deviation indicates the structural role for the side chain of 38 is more complex than simple hydrophobic interactions.

Discussion

The ECD1–CRF-R2β–Astressin Complex.

The 3D NMR structure of ECD1 of CRF-R2β in complex with astressin, together with the Ala scan analysis, shows that the most important interaction is between the hydrophobic residue 38 of the peptide with the hydrophobic core Ile-67, Cys-84, Val-113, Tyr-115, and Cys-118 of ECD1 (Fig. 3 and SI Table 2). The other important group is the C-terminal NH₂, which is involved in two hydrogen bonds (i.e., one intramolecular helical hydrogen bond to Glu-39 and an intermolecular hydrogen bond to Tyr-115). Other interactions are listed in the decreasing order of importance: The intermolecular hydrogen bond between the side chain of the highly conserved Asn-34 (peptide) and the backbone of Phe-88 (ECD1); the intramolecular solvent-exposed salt bridge between Glu-39 to Lys-36 of the peptide; the intermolecular salt bridge between Arg-35 (peptide) and Glu-86 (ECD1); and the intermolecular hydrophobic interaction between Leu-27 (peptide) and Ile-91 (ECD1). Furthermore, the intermolecular backbone hydrogen bond between Ile-41 (peptide) and Val-113 (ECD1), the hydrophobic interaction of Ile-41 (peptide) with Tyr-115 and Val-113 (ECD1), as well as the hydrophobic interaction of Ala-31 (peptide) with Ile-91 (ECD1) are important binding determinants, although their ΔΔG° values could not be evaluated (see above). The quantitative description shows that a large interface contributes to the stability of the complex, while the hydrophobic interactions and the hydrogen bonds of the four C-terminal residues are the most important binding determinants (SI Table 2).

In contrast to the peptide, which folds into a helix upon complex formation, the conformational changes of the ECD1 are of a local nature comprising residues 65–69 and 84–95 of loop 2 and residues 111–115 (Fig. 2A). This rearrangement is accompanied by the suppression of extensive slow motional processes of the loop residues 84–98 observed in the free ECD1. Such slow dynamics at a protein–protein interaction site in the free state of the protein have been reported in proteins with a SCR fold (31) as well as in enzymes (32). For the latter, there is evidence that the free state of the protein undergoes a conformational exchange between the bound and unbound conformation optimizing the catalysis. Applied to the ECD1, this suggests that loop 2 of the ECD1, in the absence of ligand, undergoes a slow conformational exchange between the ligand-bound and an unbound conformation. This interpretation is reasonable because the disordered conformational space of loop 2 in free ECD1 covers the conformation of loop 2 in the complex (Fig. 2A). Assuming that this slow conformational exchange is also retained in the full-length CRF-R2β, it is proposed that the nature of the conformational exchange would allow the ECD1 to bind the peptide hormone in an induced fit mechanism, without paying high entropy costs at the receptor.

CRF Ligand–Receptor Interactions Have a Common Binding Mode.

Although astressin is a peptide antagonist analog based on human CRF, we propose that the 3D structure of the complex between ECD1–CRF-R2β and astressin represents the 3D structure of CRF agonists with the ECD1 of CRF-R1 or CRF-R2. This is based on the following: There is a high sequence homology for the C-terminal 15 residues of CRF family agonists, as well as for the ECD1s of CRF receptors, including the peptide hormone binding site (Fig. 2D and SI Figs. 4 and 6). In addition, several CRF peptide agonists and antagonists have a similar bioactive-like helical conformation in DMSO (SI Fig. 6 C–E) including the amphipathicity, the positioning of Arg-35 and the hydrophobic side chain of residue 38, and the helical hydrogen bond of the C-terminal NH₂. Furthermore, chemical shift perturbation data suggest that the binding sites of antagonist and agonist on ECD1–CRF-R2β are similar (18). In a recent study (33), the conformation of a shorter CRF antagonist (9) was shown to be helical when bound to the ECD1 of CRF-R1. From these observations, we propose that CRF family ligands bind in a similar manner to the ECD1 of the CRF receptors.

Insights into the Receptor-Selective Binding Specificities of CRF Ligands.

Ligands of the CRF peptide family have distinct specificities for the two CRF receptors. The affinities of CRF and Ucn 1 for CRF-R1 are comparable but Ucn 1 is ≈10 times more potent than CRF for CRF-R2 (34). Both Ucn 2 and 3 are highly selective in binding for CRF-R2 compared with CRF-R1 (35). Insights into this ligand specificity for CRF-R1 and CRF-R2 may be gained by a detailed analysis of the ECD1–CRF-R2β–astressin structure. This is reasonable, because the 3D structure identifies the binding interface, which is the determinant of the binding affinity and selectivity. However, other regions of the peptide and the receptors could also be involved in determining binding specificity, as has been proposed for the parathyroid hormone system (36).

When the sequence variation between CRF-R1 and CRF-R2 is mapped onto the 3D structure of the ECD1–CRF-R2β–astressin complex, only minor differences are observed in the binding interface. These include the pair G99A (Gly in CRF-R1 versus Ala in CRF-R2) and L66Q and the cluster A86E, F87Y, V91I, and E120P on the ECD1 (Fig. 2D). These variations have their counterparts in the ligands. Ligand residues close in space to the G99A/L66Q pair are variable, i.e., residue 38 is Met, Leu, or Phe, residue 40 is Ile, Leu, Ser, Arg, or Gln, and residue 41 is either Ile or Val (comparing only the mouse ligands for simplicity). It is noteworthy to mention that there is enough space around ligand position 38 in the complex to accommodate the large aromatic ring of Phe, which is present in Ucn 1 (SI Fig. 6). In addition, residues 86–91 of the ECD1 interact with the ligand residues Arg-35, Gln-30, and Leu-27. These residues are not conserved and vary as follows: residue 27, which is either Leu, Gln, or Ala, residue 30, which is either Gln or Arg, and residue 35, which is Arg or Ala (only the mouse ligands are given) (SI Fig. 6A).

The observation that the variable residues for the ECD1 and the ligands are in close spatial proximity suggests that, the binding specificities of the ligands are, at least in part, governed by these two local interaction sites. In particular, residue 35 in the ligand seems to be of importance for CRF-R2 selection. Arg-35 is highly conserved in the family of CRF ligands but is replaced by Ala in Ucn 2 and 3, which selectively bind to CRF-R2 (SI Fig. 6). Furthermore, sauvagine analogs with a R35A mutation showed enhanced selectivity for CRF-R2 (37). Because Arg-35 is involved in an intermolecular salt bridge to Glu-86 of ECD1, the relatively modest energy cost (SI Table 2) of replacing Arg by Ala is attributed to the idea that internal salt bridges are often energetically not very favorable (38). In CRF-R1, Glu-86 is replaced by Ala and hence the salt bridge between residues 86 and Arg-35 is not possible. However, in our complex structure, Arg-35 is also close to Pro-120, which in CRF-R1 is a Glu. It is thus possible that Arg-35 could be involved in a partially solvent-exposed intermolecular salt bridge with Glu-120. Because partially solvent-exposed salt bridges are energetically more favorable than internal ones, Arg-35 might be more important for CRF-R1 binding than it is for CRF-R2. Indeed, the ΔΔG° of (R35A)-oCRF costs an additional ≈1 kcal/mol in CRF-R1 binding compared with the corresponding ΔΔG° for CRF-R2 (data not shown).

The Modular Nature of the CRF Peptide Family and Receptors Suggest a Two-Step Binding Mechanism.

The two-step model for ligand-binding and signaling of B1 GPCRs (18, 39–41) proposes that the ligand's C-terminal segment binds to the ECD1, which then positions the amino-terminal portion of the peptide in close proximity to the serpentine regions of the receptor to initiate signaling. The ECD1 is therefore the major peptide-binding domain and the C-terminal segment of the ligand is important for high affinity binding and selectivity. The 3D structure of the ECD1–CRF-R2β–astressin complex presented here is consistent with this model, because the C-terminal residues 27–41 of the peptide ligand interact with the ECD1. Hence, short peptide antagonists, lacking the N-terminal signaling domain, bind with high affinity to the ECD1 (SI Table 3). They are antagonists because they occupy the major binding site, thus blocking agonist binding, but they cannot induce receptor signaling because the first eight residues are absent. Such peptide antagonists are astressin (42) and the short CRF-based analogs comprising residues 30–41 (9), which bind to the ECD1 of CRF-R1 in a similar conformation as astressin (Fig. 1B and SI Table 3) (33). Therefore, CRF ligands and their receptors are of modular nature, comprising a binding domain and a signaling domain. However, the binding domain is also crucial for signaling because it controls the binding affinity of the ligand and determines the orientation of the ligand for receptor signaling. Some determinants of binding specificity and affinity are possibly also located in the signaling domains, as suggested by the observation that Ucn 3 and sauvagine bind to CRF-R2β with high affinity, but not to the ECD1 of CRF-R2β (SI Table 3) (20). It is noteworthy to mention that membrane–peptide interactions seem to be less important in receptor binding (see SI Text).

CRF Receptor–Peptide Ligand Complex as a Model for Type B1 GPCRs.

The CRF ligand-receptor system is a prototype for peptide hormone GPCRs. The modular behavior of the CRF receptors with the ECD1 as the major peptide binding domain is also reported for other members of the peptide hormone GPCR family such as parathyroid hormone (12, 40, 41, 43, 44) and the SCR motif of the ECD1 is predicted to be conserved in all of the ECD1s of the B1 family of GPCRs (18). The modular property is also present for various peptide hormones such as parathyroid hormone (45). Typically the N-terminal segment of the ligand is responsible for receptor signaling, whereas the C-terminal segment is an important determinant for the receptor binding (45, 46) and may interact with the ECD1 of the receptor as evidenced by cross-linking (47). Furthermore, the bioactive conformations of various peptide hormones are proposed to be of amphipathic helical nature (48), similar to the conformation observed for astressin. The 3D structure of ECD1 of CRF-R2β in complex with astressin elucidates the major determinants and the physical basis for the ligand binding affinity and specificity for CRF receptors, which are crucial components toward an understanding of diverse biological functions of the CRF ligand/receptor system. More generally, the structure may be relevant to the entire B1 family of GPCRs, serving as a model to describe the nature of the binding mode between the SCR motif of the ECD1 and the helix of the peptide hormone. These structural studies, therefore, describe atomic resolution details of the action of the CRF ligand family and in addition establish ties within the peptide hormone family of GPCRs.

Materials and Methods

Structure Determination of ECD1 of CRFR-2β in Complex with Astressin.

A stable 1:1 complex of astressin with ECD1 of CRF-R2β for NMR studies was obtained at 30°C at pH 5 at 0.3 mM concentration, well above the K_d of the system (SI Table 3). Astressin was used as a representative ligand of the CRF family (SI Fig. 6) because it is a very potent peptide antagonist analog (42) and has high binding affinity for both CRF receptors and also their isolated ECD1s (20). The sequential assignment of the ECD1–CRF-R2β–astressin complex was obtained by assigning the resonances of the ¹³C-,¹⁵N-labeled ECD1–CRF-R2β bound to unlabeled astressin and the resonances of astressin with selected residues labeled with ¹³C and ¹⁵N bound to unlabeled ECD1–CRF-R2β using triple resonance experiments and 3D ¹⁵N- or ¹³C-resolved [¹H,¹H]-NOESY experiments (see SI Text).

For the 3D structure calculation, upper limit distance restraints were collected from NOESY spectra. For residues 44–122 of ECD1–CRF-R2β and residues 27–41 of astressin a total of 417 angle restraints and 1,524 distance restraints were collected. This includes 1,153 distance restraints for ECD1–CRF-R2β, 189 distance restraints for astressin, and 185 intermolecular distance restraints (SI Table 1). To obtain a well defined interface between astressin and ECD1, the large number of intermolecular distance restraints was crucial. The 3D structure was calculated for residues 44–122 of ECD1–CRF-R2β and residues 27–41 of astressin. The good quality of the 3D structure is represented by the small rmsd of 0.6 Å of residues 57–83 and 99–120 of ECD1 and residues 30–41 of astressin (Fig. 1B), by the small residual constraint violations in the 20 refined conformers, and by the small deviations from ideal geometry (SI Table 1).

Additional Details.

Further experimental procedures such as peptide synthesis, protein expression, radioreceptor assays, NMR experiments, and structure calculations are given in SI Text.

Abbreviations

CRF

corticotropin releasing factor

oCRF

ovine CRF

CRF-Rn

CRF receptor n

ECD

extracellular domain

GPCR

G protein-coupled receptor

Ucn

urocortin

SCR

short-consensus-repeat.