Report

Allosteric Activation of a Spring-Loaded Natriuretic Peptide Receptor Dimer by Hormone

See allHide authors and affiliations

Science  31 Aug 2001:
Vol. 293, Issue 5535, pp. 1657-1662
DOI: 10.1126/science.1062246

Abstract

Natriuretic peptides (NPs) are vasoactive cyclic-peptide hormones important in blood pressure regulation through interaction with natriuretic cell-surface receptors. We report the hormone-binding thermodynamics and crystal structures at 2.9 and 2.0 angstroms, respectively, of the extracellular domain of the unliganded human NP receptor (NPR-C) and its complex with CNP, a 22–amino acid NP. A single CNP molecule is bound in the interface of an NPR-C dimer, resulting in asymmetric interactions between the hormone and the symmetrically related receptors. Hormone binding induces a 20 angstrom closure between the membrane-proximal domains of the dimer. In each monomer, the opening of an interdomain cleft, which is tethered together by a linker peptide acting as a molecular spring, is likely a conserved allosteric trigger for intracellular signaling by the natriuretic receptor family.

The natriuretic peptides (NPs) are three homologous peptide hormones that play important roles in the maintenance of cardiovascular homeostasis, blood pressure, and body fluid regulation (i.e., natriuresis) (1). Collectively, these hormones function as an endogenous counterbalance to the renin-angiotensin/aldosterone system, as well as the hypothalamic/pituitary/adrenal axis. The three members of this family are atrial (ANP) and brain (BNP) natriuretic peptides, which are produced by the heart; and CNP, which is expressed in endothelial cells (1). ANP and BNP are thought to be the primary regulators of peripheral natriuretic activity; CNP is present mainly in the brain. ANP, BNP, and CNP are highly homologous (∼70% identical) and share as a common motif a 17–amino acid loop formed by a disulfide bond (Fig. 1). The lack of defined structure(s) in solution and the questionable relevance of “lowest energy” solution conformations to the receptor-bound conformation of peptide hormones mean that assessing the bioactive conformations of these peptides remains a general problem (2).

Figure 1

Thermodynamics and 1:2 stoichiometry of NP interactions with human NPR-C. The ITC of interactions of NPR-C with CNP, ANP, and BNP is shown from left to right. The sequences of CNP, ANP, and BNP are portrayed in schematic form showing the cyclic peptides with the invariant residues colored yellow and the different residues colored blue. The measured thermodynamic (ΔH and ΔS) and stoichiometric (N) values are indicated below. The top panel is the raw data and the bottom panel is the plot of heat changes against the ratio of peptide hormone to receptor for each injection. The solid curve is the best-fit curve with nonlinear least-squares fitting. The negative ΔH and TΔS (T, temperature) show that the interactions are strongly enthalpy-driven and entropically unfavorable. The N of 0.5 is evidence of that one peptide hormone binds to two receptor monomers. The overallK d (and ΔG) is too low (i.e., high affinity) for precise quantitation by ITC; however, approximate values calculated from these data are CNP, K d ∼ 0.45 nM; ANP, K d ∼ 0.95 nM; and BNP,K d ∼ 47.5 nM (actualK d values would be lower). This rank order of affinities matches the order of affinities measured on cells bearing NPR-C. Further details of the ITC are given in (30).

The actions of the NPs are mediated by three homologous single-transmembrane, glycosylated cell-surface receptors (NPR-A, -B, and -C) (3–5). These receptors possess about 30% homologous extracellular ligand-binding domains (ECDs) (∼450 amino acids) with conserved topologies but possess different downstream activation mechanisms. NPR-C is the most promiscuous of the receptors, binding to all NPs with high affinity, whereas NPR-A and NPR-B are more specific for ANP and CNP, respectively (6). In the cases of NPR-A and NPR-B, hormone binding to the ECDs results in the production of intracellular cyclic guanosine 3′,5′-monophosphate by a guanylyl-cyclase activity that resides in the intracellular domains (5, 7). For NPR-C, which represents over 95% of NPR in vivo, ligand binding results in both internalization and degradation (i.e., clearance), as well as signaling by heterotrimeric GTP-binding protein–mediated phosphoinositide hydrolysis and inhibition of adenylyl cyclase (8, 9). The clearance function serves to buffer NP-mediated changes in blood pressure, and the signaling function can suppress catecholamine release (10–12). The NPR-C cytoplasmic tail is phosphorylated upon hormone engagement of the ECD, activating Gi-like proteins in a fashion similar to the insulin-like growth factor–II receptor (13).

Despite their different downstream activation cascades, the molecular mechanisms of hormone recognition and receptor activation are likely to be conserved throughout this family because of the high degree of similarity found among the receptors and ligands (8,14). A number of key issues, such as the ligand:receptor stoichiometry, the roles of oligomerization and conformational change, and the importance of glycosylation remain controversial (1,4, 5, 15–19). It is clear that a simple ligand-induced oligomerization is not sufficient for activation. The signaling of the NPR appears highly sensitive to mutations of the membrane-proximal portion of the ECD; it is generally assumed that hormone binding induces conformational changes in this region as a means of activation (3,20–22). Another unique feature of this receptor family is that glycosylation appears to be important for modulating the binding activity of the extracellular domains (23–26).

We have attempted to resolve some of these structural issues by carrying out biophysical studies of a soluble form of the NPR-C ectodomain that recapitulates the ligand pharmacology of the receptor on the cell surface. ANP, BNP, and CNP bind to NPR-C with a 1:2 hormone:receptor stoichiometry and enthalpy-driven energetics. We have also determined the crystal structures of both the unliganded and CNP-complexed forms of the receptor ECD, which reveal a large-scale conformational change that is likely a conserved extracellular activation trigger for members of the NPR family.

We expressed the extracellular domains of both human NPR-A and NPR-C receptors in Drosophila melanogaster cells so that glycosylated receptors were used for biochemical and crystallographic studies (27, 28). Both soluble receptor ECDs eluted as dimers on gel filtration, and both exhibited quantitative band-shifting by substoichiometric (∼1:2) amounts of NP hormones in a nondenaturing native gel experiment. We measured the ligand:receptor stoichiometry and binding thermodynamics (enthalpy ΔH, entropy ΔS, and free energy ΔG) for the interaction of NPR-C with ANP, BNP, and CNP using isothermal titration calorimetry (ITC) (Fig. 1) (29). NPR-C binds to all NPs with a clear 2:1 stoichiometry, and all the hormones bind with strongly enthalpy-driven thermodynamics and unfavorable entropy (Fig. 1) (30). The affinities are K d ≤ 1 nM for CNP and ANP (K d, dissociation constant), and the rank order of affinities is CNP ≈ ANP ≫ BNP (Fig. 1), which mimics the pharmacology of the naturally occurring receptor (6). The strong enthalpy-driven binding (–10 to –18 kcal/mol) implies the formation of high-energy polar contacts between hormone and NPR-C, and the unfavorable (negative) entropic energy barrier is likely due to the substantial conformational chain entropy of the flexible peptide in solution (31).

We crystallized NPR-C both unliganded and bound to CNP (Table 1). The unliganded structure was determined to 2.9 Å with one receptor ECD, or a half-dimer in the asymmetric unit, with the second half of the dimer related by a crystallographic twofold axis (Fig. 2A). The structure was determined by molecular replacement with a truncated model derived from a monomer of the unliganded rat NPR-A structure (22). Each monomer is dumbbell-shaped, with each domain adopting an αβ fold consisting of a central β sheet surrounded by α helices (Fig. 2, A and B). This fold is similar to that found in bacterial periplasmic-binding proteins (bPBPs) (e.g., the leucine-isoleucine-valine binding protein LIVBP), and to the unliganded rat NPR-A structure (22, 32). The membrane-distal and membrane-proximal (NH2 and COOH) domains are connected by three short connecting peptides, which form a hinge with an “elbow angle” (∼120°) analogous to that seen in type I bPBPs. The hinge forms a cleft between the two domains (Fig. 2A), which is the ligand binding site for bPBPs (32). In the NPR-C structure, we see ordered, high-mannose carbohydrate chains emanating from two out of the four possible N-linked glycosylation sites (Asn248 and Asn349) (Figs. 2 and 3) [Web fig. 1 (33)].

Figure 2

Three-dimensional structures of human NPR-C and the NPR-C/CNP complex. Shown are ribbon diagrams of (A) the unliganded NPR-C dimer (top, in magenta) and of (B) NPR-C bound to human CNP (bottom, in cyan with orange CNP hormone in space-filling view). The cell surface is shown in schematic at the base of the COOH-termini of both receptors (on the cell surface, the receptor ECD would enter the membrane bilayer at about residue 430). The twofold axis relating each monomer in both the free and liganded complex runs directly through the center of the dimer, parallel to the page. The carbohydrates are drawn as yellow sticks. The large Asn248-linked glycan (six moieties) is especially clear in the unliganded structure between the membrane-proximal (NH2-terminal, N-ter) and membrane-distal (COOH-terminal, C-ter) domains of the dumbbell-shaped monomers. The figure was drawn with BOBSCRIPT and rendered with RASTER3D (41).

Figure 3

Conformational changes in the NPR-C complex and the molecular spring. (A) Backbone representations of bound (cyan) versus unliganded (purple) NPR-C (the peptide in the middle is shown in red). At the base of the structures, the width of the gap separating the COOH-terminal domains of the dimer in bound versus free form is indicated. The identical amino acid closest to the COOH-terminal at the base of the gap (Ala208) was used in both structures as the point from which to measure the gap to the dimeric-related residue (Ala208*). For the elbow angle of the structures, identical reference points (a vector defining an α helix in the membrane-distal and -proximal domains) were chosen in bound versus free structures from which to measure an interdomain angle. (B) The spring tethering the membrane-distal and -proximal domains in each monomer is stretched and lengthened by 2.5 Å in the bound structure (40). A ribbon representation is shown of the linker peptide, along with the secondary structure elements leading up to and away from the peptide. The loose structure of the unbound peptide is obvious as compared with the straightened peptide in the complex. (C) The N-linked glycan at Asp248 forms extensive interactions with the linker peptide, which are broken upon hormone binding and conformational change (40). A stick representation of the peptide, the N-linked glycan, and the surrounding amino acids is shown. We have superimposed the Fo-Fc SIGMAA-weighted omit maps, at 2.9 Å (left) and 2.0 Å (right) of the NH2-linked glycan, to demonstrate the clarity of the carbohydrate conformational change.

Table 1

Crystallographic statistics for unliganded NPR-C and the NPR-C/CNP complex. The human NPR-C receptor extracellular domain (430 amino acids) was secreted from D. melanogaster cells, and the COOH-terminal hexa-histidine tag was removed with carboxypeptidase-A (40). Crystals of unliganded NPR-C were grown from 1.5 M sodium/potassium phosphate or 1.5 M potassium phosphate (pH 7.5) and cryo-cooled in the presence of 20% ethylene glycol. For the complex, NPR-C was incubated with an excess of human CNP and then purified by gel filtration chromatography. Crystals of the complex were grown from 0.8 M sodium citrate (pH 7.5) and were cryo-cooled in 20% glycerol. Data from a single crystal for both unliganded and complex were collected at SSRL BL9-2 and 9-1, respectively, and processed with MOSFLM and SCALA (41). The higher resolution NPR-C/CNP complex was solved first by molecular replacement (MR) with the coordinates of the unliganded ECD of rat NPR-A (Protein Data Bank accession code, 1DP4) (22). Only MOLREP (41) identified correct solutions with a polyalanine model. The two receptor monomers were refined in the early stages with NCS restraints, which were then released to allow for differences on surface loops. The hormone CNP was traced in a single unique conformation in two equally occupied twofold symmetric orientations, reflecting its location along the receptor dimer twofold. The peptide was refined in two orientations with equal occupancy (0.5) to satisfactory stereochemical criteria, with all but one (Met17) residue falling within favored or allowed regions of the Ramachandran plot as determined by PROCHECK (41). The unliganded NPR-C structure was solved by MR with the separate NH2- and COOH-terminal domains of the refined NPR-C complex. Both liganded and unliganded structures were refined with a maximum-likelihood target function, rigid-body refinement, cycles of simulated annealing with torsion angle molecular dynamics, and iterations between positional and B-factor minimization. SIGMAA-weighted 2Fo-Fc and omit Fo-Fc maps (CNS) were manually rebuilt in the program O, and stereochemical analysis was performed with PROCHECK (41). All data were used without truncation, except for 5% data randomly selected for cross validation. In the unliganded structure, two asparagine-linked glycosylation sites contained electron density for a total of 11 glycan residues (6 for Asn248, 5 for Asn349). In the NPR-C/CNP complex, we modeled a total of five N-linked glycan moieties on two N-linked sites in each monomer (Asn248 and Asn349). In both structures, one chloride ion was clearly defined in each monomer.

View this table:

The uncomplexed NPR-C dimer orientation (Fig. 2A) is different from a tail-to-tail dimer that was observed in the uncomplexed rat NPR-A crystal structure and was proposed to be physiologically relevant (22) [(Web fig. 2 (33)]. The NPR-C dimer is mediated through the NH2-terminal domains, which form a complementary four-helix bundle hydrophobic dimer interface (1378 Å2 of the buried surface area), with the dimer twofold running through the center of the four helices (Fig. 2) [Web figs. 1 and 2 (33)]. The dimeric interface has a large, continuous hydrophobic core, including the residues Phe69, Leu68, Leu101, Trp105, Val72, Ala76, and Val75; and the aliphatic atoms of His105, Asn65, and Arg100 from each monomer [Fig. 2, Web fig. 1 (33)]. The helical bundle is flanked by six symmetric hydrogen bonds (three H bonds from each side), consisting of Asp73Oδ2-Trp104 Nɛ1, Asp73Oδ1-His105 Nδ2, and Asn65Nδ2-Glu125 Oɛ2. The H-bonding residues Asp73, Trp104, and His105 are completely conserved across the NPR family, and the hydrophobic core residues in the dimer interface are also either conserved or have conservative substitutions. This dimer places the ligand-binding sites within the interdimer interface, consistent with the measured stoichiometry. The NPR-A tail-to-tail dimer is mediated through an interface between the COOH-terminal domains, which results in the proposed ligand-binding regions facing opposite directions and thus accommodating a proposed 2:2 stoichiometry (22). Because the stoichiometry of the NPR appears to be 1:2, the NPR-A tail-to-tail dimer is probably not physiologically relevant [Web fig. 2 (33)].

We crystallized the complex of NPR-C with CNP and determined the structure to 2.0 Å resolution (Table 1). The conformation of the NH2-terminal domain dimeric interface is unchanged (Figs. 2B and 3). A single CNP molecule is bound in the center of the twofold symmetric dimeric receptors (Figs. 2B and4), confirming the ITC determination of the 1:2 stoichiometry (Fig. 1). Because the hormone is bound exactly along the center of the receptor dimeric twofold axis, the electron density shows the average of two equally occupied, twofold symmetric orientations of this conformation imposed by the dimeric receptor (Table 1) [Web fig. 1 (33)].

Figure 4

Asymmetry of the hormone/receptor interfaces and the conformation of CNP. (A) Stick representation of the bound CNP peptide (orange) and the interacting amino acids from each NPR-C monomer (cyan and green) (40). The yellow spheres represent the bound chloride ions in each monomer. Ile188, which has been shown to modulate the ligand pharmacology of NPR-C (37), is next to the CNP residue Phe7, and is labeled in black. (B) This interface is then shown in an “open-book” view of the molecular surface of each receptor monomer. The CNP peptide is shown as a yellow backbone-and-stick model projected onto the respective buried surfaces (red patches) of each NPR-C monomer. The figures were drawn with BOBSCRIPT, RASTER3D, and VMD (41).

The peptide adopts a disk-like conformation, in which the narrowest dimension intercalates between the two halves of the dimer. The gap between the two COOH-terminal domains closes by 20 Å, so that the two monomers make intimate contact with opposing faces of the peptide (Figs. 2 and 3). The closure of the membrane proximal domains onto the peptide causes the elbow angle between the NH2-terminal and COOH-terminal domains of each monomer to open by ∼13.5° (Fig. 3). Thus, the receptor appears to have “open” and “closed” states, which are regulated by hormone binding. This is analogous to, but the converse of, the soluble bPBPs' hinge-movement transition to a closed state upon ligand binding within the cleft (34). Here, CNP does not bind within the cleft between the domains to close the structure but rather opens the interdomain cleft by binding to the opposite side of the monomer (Fig. 3). Hence, it appears that the malleable architecture of the bPBPs has been used by the NPR family to transduce a ligand-induced allostery in a cell-surface receptor, but with a different switching mechanism. Mutants of NPR-A that are disulfide-bridged into homodimers through the juxtamembrane stalk are constitutively active, suggesting that the closure of the interdimer gap is associated with activation (21). Numerous other functional studies have also implicated the COOH-terminal region of the ECD as being critical for ligand-induced signaling in this family (15,21, 35, 36).

In the unliganded structure, the linker peptide between the NH2- and COOH-terminal domains is stabilized by nine H bonds [five with the Asn248-linked carbohydrate moieties and four with the surrounding protein structure (Lys284O-Glu287N; Lys284NZ-Glu286Oɛ2; Asn281N-372Oγ; Asp370O-Leu280N)] (Figs. 2 and 3).

In the complex, the opening of the interdomain elbow angle results in a lengthening of this six-residue peptide by ∼2.5 Å, which substantially straightens its structure (Fig. 3B). Moreover, all but one of the H-bonding contacts stabilizing this peptide in the unliganded structure are broken, and four new H bonds form to give the set of H bonds Lys284O-Glu353N, Thr282Oγ1-Ala356O, Arg281Nη2-Val309O, Asp370O-Leu280N, and Lys284N-Glu288N (Fig. 3C). The four H-bonding contacts with the Asn248-linked glycan are broken and the sugar becomes largely disordered (Fig. 3C). Thus, the linker peptide appears to act as a “molecular spring.” In concert with the N-linked glycan, it may help constrain the receptor in the closed conformation in the absence of ligand. The strong enthalpy of the hormone interaction (–19.8 kcal/mol) may be required to provide sufficient binding energy to destabilize the spring from its basal closed state.

The CNP conformation is nearly a circle in the bound state, with 18 of 22 hormone residues clearly visible (Figs. 2 and 4). The CNP NH2-terminal amino acids are largely disordered, suggesting a tenuous interaction of the CNP NH2-terminus with NPR-C. The peptide chain is fairly extended, with no intrapeptide H bonds formed, suggesting it will not be an energetically stable solution conformation in the unbound state. The conformation of the bound CNP does not resemble a solution structure of a conformationally constrained variant of ANP determined by nuclear magnetic resonance (2). In the bound state, the invariant bulky side chains of Phe7, Met17, and Arg13 fix the interactions with NPR-C. The peptide is oriented with the disulfide bond and with NH2- and COOH-termini pointing downward (Fig. 4) toward the receptor COOH-terminal region, leaving ample space to accommodate the side chains of additional COOH-terminal ANP and BNP residues that occur in these hormones and that contribute to their receptor specificity (35).

There are numerous polar and hydrophobic interactions between CNP and each of the two NPR-C monomers (Fig. 4) (Web table 1) (hereafter designated site I and site II for receptor monomers 1 and 2, respectively). Although the dimeric receptor has twofold symmetry, the site I and site II hormone contacts are entirely different on both receptor and peptide, reflecting the asymmetry of the hormone sequence and structure (Fig. 4). The site I interface is more extensive than site II, but both sites are primarily composed of the COOH-terminal domains (Figs. 2 and 4), consistent with the allosteric activation mechanism transduced through movement of the juxtamembrane domains. The CNP Phe7 side chain is in a large hydrophobic cavity bordered by NPR-C Ile188, which has been mapped in functional studies to determine the hormone pharmacology of NPR-C (8, 37). The key site I contact residues, CNP Phe7, Gly8, and Arg13, are conserved across all NP hormones (Fig. 1) and so likely represent common contacts. Site I and site II NPR-C residues contact CNP through numerous conserved residues shared between NPR-A, NPR-B, and NPR-C (Fig. 4) [Web table 2 (33)]. Collectively, the subset of hormone-contacting residues in NPR-C has a higher homology than the overall sequence homologies between the entire NPR receptor ECDs, indicating common ligand interaction modes [Web table 2 (33)].

NPR-C binds to all NPs, whereas NPR-A and NPR-B are more specific for ANP and CNP, respectively. The invariant receptor-contacting side chains of the three NP hormones would fix the basic binding geometry to the receptors, whereas the residues that differ between ANP, BNP, and CNP (mostly polar side chains with H-bonding potential) (Fig. 1) will likely form receptor subtype–specific contacts. The peptide interactions of the NPR appear similar in character to those between major histocompatibility complex (MHC) antigens and antigenic peptides, which bind to a single MHC though a small subset of invariant “anchor” residues, whereas the nonanchor residues vary and dictate the T cell receptor specificity (38).

Two important issues that the NPR-C structures illuminate are (i) the role of glycosylation and (ii) the role of chloride ions. It has been demonstrated that glycosylation is necessary for full binding activity of NPR, although this fact remains controversial (23–26, 28). None of the N-linked glycans seen in the structure make direct contact with the ligand. It is most probable that the N-linked glycans serve an overall role in structural stabilization, and the effect on binding activity is indirect. Chloride has also been suggested to be required for receptor activity (39). We see a deeply bound chloride ion in each monomer in both the unliganded and complexed structures. This does not interact with the ligand and exhibits no surrounding conformational changes upon hormone binding, so this chloride is likely integral to the structural stability of the protein.

The NPR family has used an ancient protein fold, structurally predisposed to conformational flexibility, as a cell-surface signal transduction module. However, the mode of hormone interaction and allostery is without precedent. The presence of two nonidentical hormone interaction sites on the receptors substantially expands the scope of chemistries possible for the development of NPR-C antagonists for the treatment of cardiovascular diseases.

  • * To whom correspondence should be addressed. E-mail: kcgarcia{at}stanford.edu

REFERENCES AND NOTES

View Abstract

Stay Connected to Science


Editor's Blog

Navigate This Article