Miscellaneous
Structural insights into the ligand binding domains of membrane bound guanylyl cyclases and natriuretic peptide receptors1

https://doi.org/10.1006/jmbi.2001.4922Get rights and content

Abstract

Membrane bound guanylyl cyclases are single chain transmembrane receptors that produce the second messenger cGMP by either intra- or extracellular stimuli. This class of type I receptors contain an intracellular catalytic guanylyl cyclase domain, an adjacent kinase-like domain and an extracellular ligand binding domain though some receptors have their ligands yet to be identified. The most studied member is the atrial natriuretic peptide (ANP) receptor, which is involved in blood pressure regulation. Extracellular ANP binding induces a conformational change thereby activating the pre-oligomerized receptor leading to the production of cGMP. The recent crystal structure of the dimerized hormone binding domain of the ANP receptor provides a first three-dimensional view of this domain and can serve as a basis to structurally analyze mutagenesis, cross-linking, and genetic studies of this class of receptors as well as a non-catalytic homolog, the clearance receptor. The fold of the ligand binding domain is that of a bilobal periplasmic binding protein (PBP) very similar to that of the Leu/Ile/Val binding protein, AmiC, multi-domain transmembrane metabotropic glutamate receptors, and several DNA binding proteins such as the lactose repressor. Unlike these structural homologs, the guanylyl cyclase receptors bind much larger molecules at a site seemingly remote from the usual small molecule binding site in periplasmic binding protein folds. Detailed comparisons with these structural homologs offer insights into mechanisms of signal transduction and allosteric regulation, and into the remarkable usage of the periplasmic binding protein fold in multi-domain receptors/proteins.

Introduction

The family of mammalian membrane bound guanylyl cyclases is comprised of seven members. The ANP receptor (GC-A) and C-type natriuretic peptide (CNP) receptor (GC-B) are both activated by extracellular binding of homologous natriuretic peptides.1, 2, 3 Activation of the latter receptor produces a strong smooth muscle relaxation whereas the former receptor is more involved in diuresis and natriuresis, and thus blood pressure regulation. The heat-stable enterotoxin toxin receptor (GC-C) is activated by its name bearing Escherichia coli toxin as well as its physiological ligand, guanylin.4 Two other guanylyl cyclase receptors, retinal guanylyl cyclase I and II (GC-E and GC-F), are located inside the cones and rods of the eye. These receptors are involved in photoadaption and restoration of Ca2+ and cGMP levels after light excitation. The retinal receptors are activated intracellularly by changes in calcium levels that are sensed by GCAP proteins5 that bind to the intracellular domain of these receptors.6 Finally, there are two additional receptors, GC-D7 and GC-G,8 of which the former is found in certain olfactory sensory neurons and thought to be regulated by GCAP1 and calcium similar to retinal GCs.9 In addition to these seven guanylyl cyclases, there is a homologous non-catalytic clearance receptor (NPR-C).10 Whereas GC-A binds both ANP and brain natriuretic peptide (BNP) hormones and GC-B binds only C-type natriuretic peptide (CNP) hormone, the clearance receptor is less specific and binds all three hormones with similar affinity.

The hormone binding domain of the rat GC-A was expressed in COS-1 cells11 and is post-translationally modified with complex and high-mannose containing N-linked glycosylation that make up 16% of the total mass.12 This extracellular domain was crystallized as a dimer due to the high protein concentration (10 mg/ml) used for crystallization.13 The two GC-A hormone binding domain monomers are related by a 180° rotation (Figure 1(a)). Each monomer is comprised of a membrane distal and membrane proximal lobe/subdomain that are connected by a three cross-over hinge. The subdomains contain a mostly parallel β sheet flanked with α helices on both sides. The N terminus starts at the first β strand of the membrane distal domain and the C-terminal residues D435, being seven residues away from the transmembrane helix, ends in an irregular and protruding region at the bottom of the membrane proximal domain. The two monomers are very similar to each other except that they have a slight difference in hinge angle of about 2° between the membrane distal and membrane proximal domains indicating an inherent hinge flexibility. As noted earlier13 the hormone binding domain structure of GC-A is very similar to that of periplasmic binding proteins such as AmiC depicted in Figure 1(b). Further details of the hormone binding domain structure of GC-A and structural comparisons will be discussed below in the context of the sequence alignment and biochemical and genetic data available on GC-A and of the homologous GC-B, GC-C, and NPR-C receptors.

The crystal structure of the GC-A hormone binding domain can now be used to carry out a structure-guided multi-sequence alignment of homologous receptors (Table 1). The alignment indicates that insertions and deletions occur in loop regions and show in one case a deletion of a short one-turn helix. The calculated sequence identities in the extracellular domains for rGC-A with hGC-B, hGC-C, and hNPR-C are 41%, 19%, and 29%, respectively. The GC-A binding domain structure contains three disulfides, as determined previously,14 which are located in the membrane distal, the membrane proximal domain, and in the irregular C-terminal region that is close to the membrane at the bottom of the dimer (Figure 1(a)). The first disulfide bond C60-C86 is fully conserved in these four receptors. The second disulfide C164-C213 is mostly conserved but absent in GC-B (Table 1). The third disulfide bond is missing in NPR-C as well as in GC-C. This disulfide is part of a small loop protruding at the bottom of the GC-A dimer (Figure 1(a)). NPR-C uses its C-terminally located cysteine residues C428 and C431 to form inter-receptor disulfide linkages resulting in a covalent dimer.15

The other family members are not included in the sequence alignment as these receptors have no known extracellular ligands and there have been few biochemical and/or mutagenesis studies. Nevertheless, these receptors are expected to have extracellular domains which adopt the same fold despite their low sequence similarity as indicated by similar disulfide bond pairing, as noted earlier.16, 14

The two monomers interact via their helices α8 and α9 that run parallel with helices α9′ and α8′, respectively, of the other monomer thereby forming a local four-helix bundle (Figure 1(a)). A few additional dimerization contacts are made with several residues in the loop formed by residues 262–269. This dimer 1 interface is the largest inter-molecule contact found in the crystal and contains by far the most inter-molecular hydrogen bonds. Dimerization buries a total of 1723 Å2 of solvent accessible surface as calculated using the program CNS17 and involves seven intermolecular hydrogen bonds formed by the side-chains of D174, Y196, K198, R201, N221, N228, S269, and main-chain atoms of P193, L200, D264, G265. As postulated previously,13 this dimer organization resulted in a relatively short distance between the C termini, which are in a protruding protease-susceptible conformation and is therefore likely to represent an activated state of the receptor as observed in previous studies.18, 19, 20

The sequence alignment in Table 1 shows that several of the dimer interface residues are reasonably conserved yet quite a few are more variable. The limited similarity of the dimerization could be intended, as too great a similarity would allow for the possibility of forming non-productive heterodimer receptors although GC-A and GC-B heterodimers have been observed by co-immunoprecipitation.21

Mapping of the hormone binding site to the receptor binding domain crystal structure was carried out using biochemical data that point to the location of where the hormone binds (Figure 2). Residues H185 and E169 cause loss of hormone binding when mutated and are solvent exposed13 (and unpublished results). The residue at position 185 in NPR-C receptor was also found to be responsible for hormone specificity.22 Furthermore, crosslinking studies with ANP substituted hormone at either its N terminus or in the middle lead to covalent cross-linking with residue M173 whereas C-terminally substituted hormone lead to a covalent cross-linking with H19513 and unpublished data. These cross-linking studies agree with previous experiments that localized the reacting amino acid residues to within residues 173–188 and 191–198.23 These residues all cluster in the same area on the membrane proximal domain near glycosylation residue N180, which was also found to be in close proximity to the hormone binding site.23 Additional biochemical studies using tetranitromethane tyrosine modification suggested that Y88 and Y120 are specifically protected by hormone binding and that they are at or near the hormone binding site13 (unpublished results), thereby extending the binding site to include part of the membrane distal domain (Figure 2). The mapping of likely binding residues to two non-overlapping sites in the dimer structure13 is consistent with a 2:2 stoichiometry in agreement with previously published observations11, 24, 25 and in contrast to a single report of a 2:1 complex.26 The position of the hormone binding surfaces within the dimer suggest that the C terminus of the ANP hormone binds near the dimer interface, since the adjacent loop 262–269 from the other monomer forms part of a postulated concave ANP binding site.13 Such an important role for the C terminus of ANP could explain why it is critical for natriuretic and vasorelaxant activities.27, 28 Each ANP interacting simultaneously with both receptor monomers provides a possible structural explanation for the ability of ANP to strongly enhance dimerization of the receptor hormone binding domain.11 The proposed ANP binding site on GC-A overlaps with the AmiR binding site on the structurally similar AmiC protein13 (Figure 1(b)) and both AmiC and the GC-A binding domain use several residues in helices α3 and α7 to interact with AmiR and ANP, respectively Figure 1, Figure 2. Further structural comparisons of AmiC and GC-A and the possible implication of this remarkable similarity in structure and binding sites will be discussed later.

A substantial amount of additional mutagenesis data for the different receptors and also a genetic analysis of a mice mutant for one of the receptors can now be discussed in light of the structural information generated by the crystal structure of the GC-A binding domain.

The GC-A receptor has been subjected to several additional mutagenesis studies. The double mutant H99L and W100L was observed to cause loss of hormone binding.29 Although residue H99 is fully exposed, W100 is substantially buried as it makes contacts with residues V70, F81, and F96 (Figure 3(a) and (b)). The W100L could therefore have drastic structurally destabilizing consequences and might indirectly lead to loss of hormone binding. H99 is located on helix α3 and, although perhaps a bit too distant from the proposed hormone binding site, its involvement in hormone binding cannot be ruled out and could also be indirectly responsible for hormone binding as will be discussed later. A different group observed that the mutant L364P also caused loss of hormone binding in GC-A.30 Inspection of the crystal structure reveals that L364 is located in a β-strand and is buried in the hydrophobic core (Figure 3(a) and (b)). Substitution with a proline will likely disrupt the β-strand, since a proline residue cannot make the necessary backbone hydrogen bond pairing. This could result in local or global destabilizing of the receptor binding domain and therefore indirectly explain the loss of hormone binding. Duda et al. also reported that the equivalent residue V358 in the GC-B receptor can be mutated into a Leu and observed a remarkable change in the hormone specificity of the receptor as it can now be activated in an ANP-dependent manner.31 The crystal structure, in which this particular residue is buried and distant from the proposed hormone binding site, cannot explain their data unless this residue, located somewhat near the hinge region, changes indirectly the hinge angle and which perhaps indirectly could lead to a change in hormone specificity.

In the GC-B receptor, the same group also found that the mutation E332H or E332 K resulted in loss of C-type natriuretic peptide hormone binding.32 This residue is at a position equivalent to GC-A E338 (Table 1) and located distant from the proposed hormone binding site (Figure 3(a)). It is interesting to note that in the human GC-A sequence this residue is histidine33 and it is therefore puzzling why such a histidine residue in the GC-B receptor cannot be accommodated.

The NPR-C receptor has been studied by mutagenesis as well. Its residues 188 and 205 have been shown to be important for modulating hormone specificity for this receptor.22 These residues are equivalent to GC-A residues H185 and A202 (see Table 1) and are in the same binding region of the ANP binding site of its receptor Figure 2, Figure 3, suggesting that NPR-C binds its natriuretic peptide hormones in the same site as the homologous GC-A receptor. A different study found that a set of double mutants in the bovine NPR-C receptor lead to loss of hormone binding.34 These double mutants are D407 and R408 (corresponding to rGC-A residues D371 and R372, Table 1), and D411 and F412 (corresponding to rGC-A residues D375 & F376, Table 1). These residues are all buried and make a salt-bridge with R125 (D371), or form hydrogen bonds with T361 and S320 (R372) or a hydrogen bond with G396 (D375), or are embedded into the hydrophobic core (F376) (Figure 3(b)). These sets of double mutants will likely have drastic consequences for the stability of the protein and therefore indirectly lead to loss of hormone binding. There are also several allelic mouse mutations in NPR-C that all cause skeletal overgrowth.35 These have been pinpointed to the extracellular domain of this receptor: one mutation introduces a stop codon, another a 12 residue in-frame deletion, and one causes an H168N mutation. Although hormone binding studies have not been carried out with this particular mutant, the fact that the latter phenotype of the single site mutation is like that of the truncated receptor suggests that the H158N mutant has lost its binding ability. The equivalent residue in the rGC-A receptor is L122 (Table 1). This residue is partially buried and contacts the hydrophobic atoms of T349, P103, and R101 in the GC-A structure (Figure 3(b)). L122 is located in the loop comprising 109–122, which includes residue Y120, which is speculated to be at or near the hormone binding site (Figure 2),13 so the likely explanation for the observed phenotype is that the substitution causes a displacement of this loop, thus indirectly causing a defect in hormone binding.

The last of the receptors discussed here is the GC-C receptor. The residues R136 and D347 of the pGC-C receptor were implicated in ligand binding.36 These residues were either part of a double or triple mutation and the other neighboring residues that were mutated were speculated not to be critical.36 The equivalent residues in rGC-A are R125 and D367 (Table 1). These residues are buried and involved in several hydrogen bonding and salt-bridge interactions in the crystal structure (Figure 3(a) and (b)). Residue R125 participates in a salt-bridge interaction with D371 and a hydrogen bond with the backbone atom of A111 whereas residue D367 makes a hydrogen bond with backbone atoms of residues N369 and D371. Substitutions at these residues will likely lead to alterations in the local structure of the region that is in the vicinity of the hormone binding site (Figure 3(a) and (b)). Either these local disruptions or perhaps global misfolding of the extracellular domain could explain the observed lack of ligand binding. Another study demontrated that a modified heat-labile enterotoxin ligand was able to covalently react with an amino acid within residues S387-K393 indicating that this region is in close proximity to the ligand binding site.37 Additional mutagenesis studies indicated that several residues at the C terminus of the GC-C receptor were also found to be important for ligand binding. These residues are T389, F390, and W392.37 The “equivalent” residues in rGC-A are K408, L409, and W411, respectively (Table 1). These GC-A residues are located in the loopy region at the bottom of the monomer (Figure 3(a)). These and surrounding residues do not align very well between GC-A and GC-C (Table 1) and it is quite possible that the conformation and position of this region are different between the two receptors. It is however interesting to note that in the AmiC:AmiR crystal structure, several residues in the C terminus of AmiC are also observed to interact with AmiR38 (Table 1, Figure 1(b)). Since AmiR is thought to bind to its partner in the same region where ANP binds to its receptor13 (see below), such an interaction of the C terminus of the binding domains with the ligand might perhaps be a possibility for all the receptors discussed here. Note also that mutagenesis of nearby proline residues such as P412 in rat GC-A did not result in loss of hormone binding.20 This particular region in GC-A seems to be more involved in anchoring the C terminus to the rest of the PBP domain as mutations in the anchoring residue P417, heavily involved in interacting with the membrane proximal domain, caused a decoupling of the receptor yet maintained its hormone binding ability.20

The N-linked glycosylated residues in GC-A, GC-B, and NPR-C have been experimentally determined12, 15, 39 whereas in GC-C all N-linked glycosylation consensus sites have been mutated individually by site-directed mutagenesis to check for proper stability and ligand binding.40 From Table 1 it is immediately apparent that there is hardly any conservation of the precise position of these sites within this group of four receptors. The glycosylation sites from all four of these receptors are mapped onto the GC-A binding domain structure and are found to be scattered over the surface of the structure except for the hormone binding site and the dimer interface (not shown). None of these glycosylation sites have been implicated in direct ligand binding but have been found to be important for proper folding and stability.41, 42, 43 In particular, N24 of GC-B39 and N379 of GC-C40 have been shown to be important for stability. N180 in GC-A was speculated to be near the hormone binding site23 in agreement with its location in the crystal structure (Figure 2). In addition to proper folding and stability, the glycosylation of the extracellular domains of the GC receptors could be important for their orientation and packing on the cell surface as suggested for other membrane bound glycoproteins.44

The GC-A binding domain crystal structure revealed an unexpected anion binding site occupied by a chloride ion, buried in the membrane distal domains of both monomers.13 The chloride ion makes three bonding interactions, with the hydroxyl of S53 and the backbone nitrogen atoms of residues G85 and C86. An extensive sequence alignment (unpublished data) reveals that the side-chain ligand for the chloride ion, S53, except for an occasional threonine, is conserved in all sequenced GC-A receptors (rat, mouse, human, Anguilla japonica), GC-B (rat, human, Bos taurus, A. japonica) and GC-C (human, rat, Cavia porcellus, B. taurus, Sas scrofa) receptors and several other GCs (Xenopus laevis, Oryzias latipes), but not in the non-coupled NPR-C, suggesting an important conserved role for an Oγ atom at residue 53 for guanylyl cyclase coupled receptors (see also Table 1 for a smaller subset of sequences). A possible role for chloride is indicated, since GC-A activity in cultured vascular endothelial cells has been shown to be affected by NaCl concentrations,45 and rats on a salt-depleted diet are found to be unresponsive to ANP as a mechanism for salt-conservation.46 Both these studies eliminated changes in GC-A receptor densities as the cause for the observed NaCl effect on GC-A activity.45 The observed chloride binding site could perhaps explain these experimental observations of NaCl dependence of GC-A receptor activity. Subsequent binding studies using purified hormone binding domain concluded that the chloride ion is an allosteric effector as it is required for ANP binding.13, 47 However, the observed affinity for chloride ions is in the low millimolar range,47 two orders of magnitude from physiological levels and suggesting that under normal physiological conditions this Cl-dependent allosteric regulation is not likely to take place if the chloride affinity in the full-length receptor is in the same range. In this case the chloride ion may be a structural ion or an evolutionary remnant: its predecessor receptor in fish was thought to be an osmo-regulatory receptor48 instead of the volume-depleting hormonal system it evolved into in vertebrates. A chloride allosteric effector seems more logical for an osmo-regulatory receptor as it can provide a negative feedback loop to warrant cellular salt conservation under low ionic stress conditions. It is also interesting to note that chloride ion concentrations have been shown to affect heat-stable enterotoxin binding to its receptor.49 Further studies are needed in order to investigate possible roles of chloride ions for guanylyl cyclases in perhaps the folding process, the receptor internalization, and/or the ligand release process.

A DALI structure similarity search50 revealed that the closest structural neighbors of the ANP binding domain are the Leu/Ile/Val binding protein (LIVBP) from E. coli51 and AmiC from Pseudomonas aereginosa38, 52 with Z-scores of 28 and 21, respectively (see Figure 1). More distantly related structurally similar proteins are the glucose/galactose-binding protein53 and d-ribose-binding proteins54 each with Z-scores of around 10. These proteins all belong to the periplasmic binding protein (PBP) type I fold as defined by SCOP.55 This similarity of the GC-A binding domain with PBP fold proteins was previously predicted in a sequence analysis study on metabotropic glutamate receptors56 yet remained unknown to the ANP receptor field; this published prediction was uncovered once the PBP fold was evident from crystallographic studies. The proteins all have two lobes or subdomains connected by a hinge region. Each of the subdomains is comprised of a mostly parallel β-sheet flanked by α-helices. The orientation of each of these subdomains is different among these proteins, making it difficult to compare their entire structures at once. However, with just the membrane proximal domain of the GC-A binding domain superimposed onto the corresponding subdomain in LIVBP and AmiC, the r.m.s.d. values are 1.66 Å (for 162 Cα atoms) and 1.83 Å (for 158 Cα atoms), respectively. With the membrane proximal domain of the GC-A binding domain superimposed onto the corresponding subdomain of LIVBP and AmiC, the r.m.s.d. values are 2.16 Å (for 110 Cα atoms) and 2.08 Å (for 117 Cα atoms), respectively. The sequence identity of the GC-A binding domain with AmiC is 9% for the entire structure and 11% for just the structurally conserved regions (Table 1). The similarity of the GC-A with LIVBP is somewhat higher, resulting in a 14% sequence identity for the entire LIVBP structure and 18% for just the structurally conserved regions including a conserved first disulfide bond (Table 1).

In addition to the single domain (with two subdomains) proteins mentioned above, the binding domain of GC-A is also structurally similar to larger multi-domain proteins such as the purine nucleotide synthesis repressor PurR57 and the lactose repressor LacR58, 59 both of which give a DALI Z-score score of ∼9 (Figure 4). In addition, the ionotropic glutamate receptor60 and the recently reported metabotropic glutamate receptor61 have a similar extracellular domain (Figure 5). The former is structurally more distantly related, as it is a member of the periplasmic binding protein type II class: in these structures some of the β strands are going in the opposite direction.62 The metabotropic glutamate receptor has a sequence identity of 9% with the GC-A binding domain and, remarkably, contains two “conserved” glycosylation sites, one that is also found in GC-B and the other is in the same position as predicted in GC-C (Table 1). In these structurally related multi-domain proteins, the position of the PBP fold domain is variable, as it can be found at the N terminus (GCs and mGlutR), at the C terminus (PurR), or located in the middle (LacR). The dimer interfaces in the DNA repressor and metabotropic glutamate receptor are dramatically different. In the DNA repressor both subdomains of its PBP domain participate in dimerization interactions. The C-terminal subdomain interface remains intact upon ligand binding and those residues lie mostly on the helix that corresponds to the α9 GC-A binding domain (Table 1, Figure 4) and the region in between the following β strand and subsequent helix corresponding to β9 and α10 in the GC-A binding domain. GC-A uses these same two stretches of residues in the proposed dimer interface (Table 1) yet the packing is somewhat different as the GC-A dimer interface also includes additional residues from helix α8 yielding a tightly packed 2 × 2 helix bundle made up of helices α8, α9, α8′, and α9′ (Figure 1(a)). The dimer interface in the metabotropic glutamate receptor is comprised of only residues in the N-terminal part of the PBP domain fold (Figure 5(b)). These residues lie on two helices corresponding to helices α2 and α3 of the GC-A binding domain (Table 1). The relative orientation of these helices with respect to the same helices in the other monomer changes by a dramatic 70° upon glutamate binding.61 As this dimer interface is located on the membrane distal subdomains whereas the dimer interface of the GC-A binding domain is located on the membrane proximal subdomains, no similarity is obviously found. However, earlier analysis of the crystal structure of the GC-A binding domain for crystal contacts and possible dimerization surfaces indicated a dimer 2 interface that is similar to the metabotropic glutamate receptor Figure 5, Figure 6. This interface is less extensive as it contains only two hydrogen bonds and buries only 1083 Å2 of solvent accessible surface whereas the dimer 1 interface contains seven hydrogen bonds and buries 1724 Å2 of solvent accessible surface.13 The dimer 2 interface is below the observed range of 1600(±400) Å2 buried surface in standard protein-protein complexes,63 and is made up of residues from helix α2 and α3 interacting with the same helices from the other monomer. The dimer 2 is remarkably similar to that observed in one of the dimer orientations of the recently determined glutamate receptor bound to glutamate61 (Table 1; Figure 5(b)) suggesting that this smaller interface could also be functionally important. The dimer 1 is made up of two non-crystallographically related molecules whereas dimer 2 is generated by 2-fold crystallographic symmetry. The possible relevance of the dimer 2 similarity with that of the metabotropic glutamate receptor will be discussed below.

In all the above-mentioned examples, a small molecule binds to the PBP fold domain and interacts with both subdomains to cause a change in hinge angle. In the DNA repressors this hinge motion induced by either hypoxanthine or IPTG binding to PurR and LacR, respectively, leads to a repositioning of the N-terminal subdomains as the inter-molecular interactions between the C-terminal subdomains remain relatively fixed58 (Figure 7(a)).64 The change in position and orientation of the N-terminal subdomain affects the DNA ligand binding domains for the DNA repressors as it is situated N-terminal to the PBP domain. This then leads to an intended change in DNA binding affinity and thus repressor activity. In the metabotropic glutamate receptor the glutamate ligand induces a change in hinge angle which affects the relative position of the termini at the other end of the PBP domain61 (Figure 7(b)). This is achieved by anchoring the N-terminal subdomains by creating an inter-receptor disulfide bond by means of a cysteine in one of the protruding loops, instead of fixing the C-terminal subdomains as in the DNA repressors (Figure 5(b)). The hinge motion of 31° induced by glutamate binding and the 70° change in the dimer interface orientation translates into a change of 25 Å in the position of the C termini of the second subdomains61 (Figure 7(b)). This shift is speculated to be the first step in signaling for the metabotropic glutamate receptor that could then lead to changes in position of the transmembrane domains and/or the cysteine-rich domain.61 In the ionotropic glutamate receptor it is not quite clear how the hinge motion will open up the tetrameric ion channel thus affecting its ion channel conductive properties.65

The similarity of the GC-A dimer 2 with one of the metabotropic glutamate receptor dimers suggests that this interface may also be physiologically relevant. Dimer 2 as such likely does not represent the hormone activated state in the full-length receptor as the two C termini are now 55 Å apart whereas experimental evidence suggests that they should be ∼7 Å apart,18, 19 much more consistent with dimer 1.13 Nevertheless, if the hormone ANP were indeed to bind to dimer 2, its binding site would be located in the wedge between the two membrane proximal domains, which is around 15 Å wide, and ANP will likely have extensive inter-receptor contacts. The cleft is only wide enough for one ANP hormone but perhaps two could fit in the cleft if they are packed along the length of the cleft. Interestingly, the dimer 2 interface also contains an inter-receptor hydrogen bond made by the side-chains of H99 and D71′, the former speculated to be important for hormone binding as it was part of a double mutant H99L and W100L.29 As discussed above, this mutagenesis data could be explained by destabilization of the structure or, alternatively, perhaps this mutation could have an effect on dimer 2 interface and therefore indirectly cause loss of hormone binding. Besides the striking similarity of the dimer 2 interface with that of one of the metabotropic glutamate receptor dimers, there is additional similarity with the metabotropic glutamate receptor dimer in one of the natriuretic peptide receptors. The NPR-C receptor cloned from eel lacks the fifth and sixth cysteine residues66 that mammalian NPR-Cs use for inter-receptor disulfide bond linked dimerization15 but instead has an additional cysteine residue between the standard first and second cysteine residues that form the first disulfide. A sequence alignment indicates that this additional cysteine residue would be located in a small loop between a helix and a strand corresponding to the GC-A α2 helix and β3 strand (Table 1). This cysteine residue was shown to be involved in inter-receptor dimer formation of eel NPR-C66 and is located in the homologous but larger loop where a similar inter-receptor disulfide bond is present in the metabotropic glutamate receptor.61 For the cysteine residues to form an inter-receptor disulfide bond in a dimer 2 organization, the distance between the Cα atoms needs to be around 6–7 Å. The cysteine residues’s equivalent position in the GC-A structure is W74 and its Cα-Cα distance to W74′ in the alternative dimer organization is 8 Å, just a few ångström units too long to form an inter-chain disulfide. In order to accommodate such an inter-chain receptor disulfide while maintaining the same dimer interface at the membrane distal domains, only a slight reorientation of the dimer interfaces or slight partial unwinding of the helix is needed.

It is possible that both dimers 1 and 2 represent different physiologically relevant states of the receptor. If we presume that the C termini remain relatively fixed, the dimer 2 organization may resemble the latent inactive state in which the intracellular domains force residues 435 of both receptor chains to be separated in distance.18 The dimer 1 organization in which the interface is comprised of part of the membrane proximal domain, somewhat similar to the fixed dimer interface of the DNA repressors, could then represent the hormone activated state13 (Figure 7(c)). This D435-D435′ distance of 14 Å between their Cα atoms is much smaller than the 55 Å for dimer 2 yet not quite close enough to the 6–7 Å needed for disulfide bond formation in a D435C mutant which is observed to occur upon hormone binding.18 However, this small distance discrepancy can be reconciled, since the C-terminal disulfide loop region of one monomer is merely held down by a crystal contact and could easily become somewhat more flexible beyond residues 425 as observed in the other monomer13 and therefore residues 435 and 435′ can each move a few ångström units to be within disulfide bonding distance (while maintaining the C423-C432 disulfide bond). Both dimer organizations are possible for mammalian NPR-C dimeric receptors to accommodate the observed inter-receptor disulfide cross-linking using C428 and C431,15 since NPR-Cs contains no short disulfide loop that restricts the position of the C-terminal region. Therefore even in the dimer 2 organization, the roughly ten C-terminal residues in front of C428 can each extend and still reach across to bridge the ∼50 Å distance for inter- receptor disulfide bond formation.

The postulated extracellular domain reorientation upon hormone binding, from dimer 2 to 1, is dramatically different from the hinge based conformational changes observed in the metabotropic glutamate receptor and the DNA repressors LacR and PurR (Figure 7). Such a small molecule mediated hinge-based conformational change is not expected, since the ANP hormone is not a small molecule but a 28 residue peptide that binds distant from the hinge effector pocket in both dimer organizations. Two different extracellular domain organizations could perhaps explain the observed low and high-affinity state for ANP of the GC-A receptor which was found to be regulated by the kinase-like domain.67 The hypothesized extracellular events in GC receptor signal transduction involving a rearrangement of the extracellular domains may perhaps be similar to the erythropoietin receptor which is also found to be predimerized with its extracellular domains undergoing a rearrangment leading to juxtaposition of the transmembrane helices.68 Juxtapositioning of the C terminus of the extracellular domain in GC-A by hormone binding could thereby overcome the kinase-like domain inhibition, thus leading to activation of the guanylyl cyclase domain.

Although this postulated dimer reorganization upon hormone binding seems consistent with the listed biochemical data, there are still inconsistencies with some other published results. First, in both dimer organizations the C-terminal disulfide loop region is protruding and flexible which does not explain the fact that this region is found to be proteolytically protected in the basal state of the receptor.20, 69 In fact, disruption of this disulfide bond by mutagenesis, which presumably results in increased C-terminal loop flexibility, yields a constitutively active receptor.20 This latter observation is not easy to explain. Too much flexibility is however not desired in the C-terminal region of the GC-A binding domain as mutation of P417, a likely important anchor residue,13 into alanine caused decoupling of the receptor.20 Second, the assignment of dimer 1 to the activated state and dimer 2 to the latent state of the receptor is solely based on the distance between the C termini with the assumption that this region does not move dramatically beyond residue C423 and that the disulfide bond remains intact thus restricting the position of C432. However, the region in GC-C near residue 390, as discussed above, is not in the immediate vicinity of its ligand which would likely require a substantial shift in the position of the entire C terminus from that seen in the GC-A structure unless the GC-C ligands bind in a totally different region to its receptor. In addition, there is some evidence that suggests that disulfide(s) might not always stay intact in full-length GC-A.70, 71 If the assumptions on the integrity and position of the C terminus are incorrect, it is conceivable that the hypothesized functional assignment of the physiological state for each of the dimers is incorrect. Thirdly, ANP binding to human full-length receptors in intact cells increased inter-receptor DSS crosslinking.70 This observed (weak) cross-linking presumably occurs extracellularly yet in both dimers 1 and 2 there are no lysine amine groups within the 11 Å needed to structurally explain this cross-linking in the human GC-A receptor unless perhaps additional mobility occurs in the C-terminal region including residue K422 upon hormone binding. Fourth, it was shown that the hormone binding characteristics of the crystallized extracellular domain are slightly different from that of the full-length receptor11 and also different from the GC-A-ΔKC construct in which the whole intracellular domain is deleted yet still contains the transmembrane helix.18 This indicates that the current GC-A binding domain structure and future hormone co-crystal structures cannot explain all the observed hormone binding intricacies if the membrane helix is not included. Either the membrane or membrane spanning helix may have a direct effect on the structure of the hormone binding domain or an indirect effect, by changing the local concentration of the hormone binding domains at the membrane and thereby the oligomeric state. Clearly, a co-crystal structure with a hormone bound to a GC or NPR receptor binding domain is very much needed to sort out the correct structural changes upon ligand binding.

Another possible consequence of the two dimer organizations is that they may provide the interfaces for the formation of trimers or tetramers of guanylyl cyclase receptors as some investigators have suggested exists for GC-C72, 73 and GC-A,70, 74 respectively. The dimer 1 and dimer 2 interfaces are not mutually exclusive in the crystal as they are both present as non-crystallographic and crystallographic dimers, respectively. In the full-length receptor, in which the extracellular domain is of course restricted by the transmembrane helix, such a tetrameric arrangement might not be possible as such, as all four C termini do not quite end up in a plane although the transmembrane helix does not start until residue 442 and the construct ends at 435, so there is perhaps some flexibility that could allow this possibility.

In all the proteins discussed that are structurally related to the GC-A binding domain, the pocket that is formed between the two subdomains binds a small molecule ligand. Binding of the ligand to these PBP fold-containing proteins induces a conformational change, resulting in a hinge closure that can range from 5° to ∼50°. The protein AmiC deserves special attention, since it binds two ligands, the small molecule acetamide as well as the protein ligand AmiR. While acetamide binds in the usual small molecule binding pocket, AmiR binds off to the side of AmiC and its binding is dependent on the absence of acetamide.38 As noted earlier, the AmiR binding site of AmiC overlaps with that of the postulated ANP binding site of its receptor binding domain (Figure 1, Figure 2, Table 1). AmiC and ANP have no global or local sequence conservation yet have overlapping binding sites to their respective binding partners so one could therefore postulate that since AmiR binding to AmiC is allosterically regulated by acetamide, ANP binding to its receptor could also be allosterically regulated by a yet unknown effector.13 The GC-A does have a putative binding pocket also present in the structurally related PBP proteins distant from the ANP binding site (Figure 8). Table 1 contains highlighted residues of AmiC, LIVBP, and the metabotropic glutamate receptor which have been found to interact with their small molecule ligands acetamide, leucine, and glutamate, respectively. The sequence alignment reveals that there is relatively high conservation in the position of these interacting residues among these three PBPs yet the precise nature of these amino acids is different as each of the ligands is very different (Table 1). These residues interacting with the small molecule ligand are located on both subdomains of the PBP. Extrapolation of these small molecule liganding residues to that of the GC and NPR receptor sequences (Table 1) reveals that at several positions there is a high degree of conservation/homology among the GC receptors (see GC-A residues G108, D242, I317) and at other positions there is considerable variation. The presence of some significant sequence similarity in these putative binding residues suggests that the currently unknown ligands could share some similar characteristics among the GC receptors. If these GC receptors are indeed allosterically regulated as AmiC is, the small molecules will likely affect hormone binding, and thus in the case for GC-A affect blood pressure. One possible (non-physiologically relevant) candidate for binding to this pocket is the natriuretic peptide receptor inhibitor HS142-1.75 It is interesting to note that in certain pathophysiological conditions such as congestive heart failure, the GC-A response is blunted despite an elevation in ANP levels.76, 77 This blunted response could perhaps be mediated via either the chloride binding site or the putative effector binding pocket.

The detailed structural analysis and structural interpretation of published biochemical and genetic data presented here on the binding domains of GC and natriuretic peptide receptors provide insight into the mechanism of function for this class of receptors. In addition, the comparisons with structurally related proteins suggest mechanisms of domain rearrangements upon ligand binding and allosteric regulation that may serve as a basis for designing future experiments to test these hypotheses. Since many of the speculations are based on the uncomplexed GC-A binding domain crystal structure which does not have the hormone bound and does not have the transmembrane helix and intracellular domains present, a more complete picture of the mechanism of GC receptor signal transduction awaits future structural studies that include these missing parts.

Section snippets

Acknowledgements

This work was supported by the NIH (no. 1 F32 GM19665-01). I thank Vivien Yee for critical reading of the manuscript, George Stark for support, Paula Pearson for secretarial assistance, and Eldon Walker and the C.C.F. LRI computer core for facilities support.

References (84)

  • M Sugiyama et al.

    Synthesis of atrial natriuretic peptides and studies on structural factors in tissue specificity

    Biochem. Biophys. Res. Commun.

    (1984)
  • T Duda et al.

    Single amino acid residue-linked signaling shifts in the transduction activities of atrial and type C natriuretic factor receptor guanylate cyclases

    Biochem. Biophys. Res. Commun.

    (1995)
  • M Hasegawa et al.

    Determination of the binding site on the extracellular domain of guanylyl cyclase C to heat-stable enterotoxin

    J. Biol. Chem.

    (1999)
  • D.G Lowe et al.

    Human natriuretic peptide receptor-A guanylyl cyclase. Hormone cross- linking and antibody reactivity distinguish receptor glycoforms

    J. Biol. Chem.

    (1992)
  • K.J Koller et al.

    Proper glycosylation and phosphorylation of the type A natriuretic peptide receptor are required for hormone-stimulated guanylyl cyclase activity

    J. Biol. Chem.

    (1993)
  • M.R Wormald et al.

    Glycoproteinsglycan presentation and protein-fold stability

    Structure Fold. Des.

    (1999)
  • T Katafuchi et al.

    Modulation by NaCl of atrial natriuretic peptide receptor levels and cyclic GMP responsiveness to atrial natriuretic peptide of cultured vascular endothelial cells

    J. Biol. Chem.

    (1992)
  • L Holm et al.

    Dalia network tool for protein structure comparison

    Trends Biochem. Sci.

    (1995)
  • J.S Sack et al.

    Periplasmic binding protein structure and function. Refined X-ray structures of the leucine/isoleucine/valine-binding protein and its complex with leucine

    J. Mol. Biol.

    (1989)
  • J.Y Zou et al.

    The 1.7 Å refined X-ray structure of the periplasmic glucose/galactose receptor from Salmonella typhimurium

    J. Mol. Biol.

    (1993)
  • A.J Bjorkman et al.

    Probing protein-protein interactions. The ribose-binding protein in bacterial transport and chemotaxis

    J. Biol. Chem.

    (1994)
  • A.G Murzin et al.

    SCOPa structural classification of proteins database for the investigation of sequences and structures

    J. Mol. Biol.

    (1995)
  • P.J O’Hara et al.

    The ligand-binding domain in metabotropic glutamate receptors is related to bacterial periplasmic binding proteins

    Neuron

    (1993)
  • K Fukami-Kobayashi et al.

    Domain dislocationa change of core structure in periplasmic binding proteins in their evolutionary history

    J. Mol. Biol.

    (1999)
  • L.L Conte et al.

    The atomic structure of protein-protein recognition sites

    J. Mol. Biol.

    (1999)
  • M.A Schumacher et al.

    Mechanism of corepressor-mediated specific DNA binding by the purine repressor

    Cell

    (1995)
  • N Armstrong et al.

    Mechanisms for activation and antagonism of an AMPA-sensitive glutamate receptorcrystal structures of the GluR2 ligand binding core

    Neuron

    (2000)
  • H.J Fulle et al.

    Functional modulation of ANF-sensitive particulate guanylate cyclase by redox mechanisms

    Eur. J. Pharmacol.

    (1991)
  • A.B Vaandrager et al.

    Heat-stable enterotoxin receptor/guanylyl cyclase C is an oligomer consisting of functionally distinct subunits, which are non-covalently linked in the intestine

    J. Biol. Chem.

    (1994)
  • Y Morishita et al.

    Microbial polysaccharide, HS-142-1, competitively and selectively inhibits ANP binding to its guanylyl cyclase-containing receptor

    Biochem. Biophys. Res. Commun.

    (1991)
  • E.A Merritt et al.

    Raster3D photorealistic molecular graphics

    Methods Enzymol.

    (1997)
  • S Singh et al.

    Isolation and expression of a guanylate cyclase-coupled heat stable enterotoxin receptor cDNA from a human colonic cell line

    Biochem. Biophys. Res. Commun.

    (1991)
  • J.G Porter et al.

    Isolation and functional expression of the human atrial natriuretic peptide clearance receptor cDNA

    Biochem. Biophys. Res. Commun.

    (1990)
  • S Schulz et al.

    The guanylyl cyclase family of natriuretic peptide receptors

    Vitam. Horm.

    (1999)
  • D.C Foster et al.

    Mechanisms of regulation and functions of guanylyl cyclases

    Rev. Physiol. Biochem. Pharmacol.

    (1999)
  • M.G Currie et al.

    Guanylinan endogenous activator of intestinal guanylate cyclase

    Proc. Natl Acad. Sci. USA

    (1992)
  • H.J Fulle et al.

    A receptor guanylyl cyclase expressed specifically in olfactory sensory neurons

    Proc. Natl Acad. Sci. USA

    (1995)
  • C Moon et al.

    Calcium- sensitive particulate guanylyl cyclase as a modulator of cAMP in olfactory receptor neurons

    J. Neurosci.

    (1998)
  • K.S Misono et al.

    Expression and purification of the extracellular ligand-binding domain of the atrial natriuretic peptide (ANP) receptormonovalent binding with ANP induces 2:2 complexes

    Biochemistry

    (1999)
  • M Miyagi et al.

    Glycosylation sites in the atrial natriuretic peptide receptoroligosaccharide structures are not required for hormone binding

    Eur. J. Biochem.

    (2000)
  • F van den Akker et al.

    Structure of the dimerized hormone-binding domain of a guanylyl-cyclase- coupled receptor

    Nature

    (2000)
  • J.T Stults et al.

    The disulfide linkages and glycosylation sites of the human natriuretic peptide receptor-C homodimer

    Biochemistry

    (1994)
  • Cited by (0)

    1

    Edited by P. E. Wright

    View full text