Ligand-independent Oligomerization of Natriuretic Peptide Receptors IDENTIFICATION OF HETEROMERIC RECEPTORS A DOMINANT NEGATIVE MUTANT*

Activation of many single-transmembrane receptors requires ligand-induced receptor oligomerization. We have examined the oligomerization of the atrial natriuretic peptide receptor, NPR-A, using epitope-tagged receptor in a co-immunoprecipitation assay. Unlike other single-transmembrane receptors, NPR-A oligomerized in a ligand-independent fashion. Extracellular receptor sequences were both necessary and sufficient for oligomer formation. NPR-A was also able to oligo- merize with the related natriuretic peptide receptor, NPR-B. A truncated NPR-A lacking most of the cyto- plasmic domain blocked activation of the full-length receptor, presumably through formation of an inactive heteromer. These results indicate that oligomerization of this single-transmembrane receptor is important for the transduction of a conformational change across the plasma membrane but are not consistent with models in which natriuretic peptide receptor oligomerization serves merely to bring intracellular domains together. transmembrane receptors transduce extracellular into intracellular conformational is transmembrane domain rigid a-helix.

Several large families of receptors have single transmembrane domains (1, 2). The mechanism by which these receptors transduce an extracellular signal into intracellular conformational changes is controversial. The transmembrane domain connecting extracellular and intracellular domains is thought to be a rigid a-helix. In one model, movement of this helix into the membrane would be the driving force for a conformational change (3). This model does not appear to account for receptor activation in solution, however. A widely accepted model for tyrosine kinase receptors involves receptor activation by ligand-induced dimerization (1). In this model, ligand binds to a receptor's extracellular domain, changing its conformation to favor its dimerization. A resulting association between intracellular domains then leads to their activation. In support of such a model, a number of single-transmembrane receptors form dimers in a ligand-dependent fashion (1, 4). The observation that growth factor receptors with cytoplasmic truncations can form inactive heterodimers with wild-type receptors further supports the importance of associations between intracellular domains (5-8) . A recently described family of single-transmembrane receptors has intrinsic ligand-activated guanylyl cyclase activity (9). The receptor/guanylyl cyclases are similar in overall topology to the growth factor receptors, including a protein kinase-like domain that seems to function as a negative * This work was supported by National Institutes of Health Grant HL 47063 (to M. C.) and Medical Research Foundation of Oregon Grant 9053 (to M. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. regulator of guanylyl cyclase activity (10, 11). Two members of this family are the natriuretic peptide receptors, NPR-A and NPR-B (10,12-14). NPR-A is a receptor for the cardiac natriuretic peptides, ANP and BNP, while NPR-B is a receptor for the brain natriuretic peptide, CNP (15). Both ANP and BNP induce natriuresis, diuresis, and vasodilation upon release from the heart, but their secretion is differentially regulated (16, 17). CNP has been suggested to have central effects on fluid homeostasis (18-21). While intravenous administration of these peptides has dramatic physiological effects and the secretion of ANP and BNP is clearly elevated in cardiovascular disease states (22, 23), the normal physiological roles of the natriuretic peptides are unknown (24).
The observation that both adenylyl cyclases and soluble guanylyl cyclases require two catalytic subunits for activity (25-28) has led to suggestions that ligand-induced dimerization of receptor/guanylyl cyclases will lead to their activation by facilitating association of guanylyl cyclase catalytic domains (29, 30). Consistent with this idea, ANP receptors in tissue extracts have been reported to migrate both as monomers and as higher molecular weight complexes when analyzed by gel filtration chromatography (31, 32). Also, a partially purified ANP receptor has been shown to migrate in SDS' gels as a high molecular weight disulfide-linked complex (33); the disulfide linkages may be an artifact of purification (cross-linking studies have generally not observed disulfidelinked complexes of NPR-A (34)), but the high molecular weight NPR-A complex could be physiological. Opposing this idea, studies using combinations of gel filtration and sedimentation analysis have suggested that the receptor/guanylyl cyclases are monomeric, even in the presence of ligand (35, 36). These apparently conflicting results have been difficult to interpret because of the problems inherent in attempts to estimate molecular weight for proteins in detergent solution and because high molecular weight receptor complexes formed in crude tissue extracts or partially purified preparations may represent either receptor oligomers or complexes of receptor with other proteins. We have used epitope-tagged NPR-A and a co-immunoprecipitation assay to re-examine this question, and report here that NPR-A does in fact oligomerize, but in a ligand-independent fashion. We also describe the ability of a truncated NPR-A to form an inactive heteromer with the full-length receptor. These results substantiate the importance of receptor oligomerization in signal transduction but challenge the notion that the association of intracellular domains is sufficient for receptor activation. In addition, we describe the formation of a heteromeric receptor by NPR-A and NPR-B.

Natriuretic Peptide
Receptor Interactions

EXPERIMENTAL PROCEDURES
Construction of Plasmids-The pSVL vector (Pharmacia LKB Biotechnology Inc.) was used for all expression constructs. Structures of all constructs described below are summarized in Fig. 1. Full-length cDNA clones and deletions of the kinase-like (AKIN) or guanylyl cyclase (ACYC) domains of rat NPR-A have been described previously (10,11,14). To delete all but 5 amino acids of the intracellular domain of NPR-A, a synthetic linker containing a termination codon and XbaI site was inserted into the PuuII site at position 1709 of the rat NPR cDNA. This construct was named AKC to denote deletion of kinase-like and cyclase domains. To express the soluble intracellular domain of NPR-A with an amino-terminal epitope "tag," a synthetic linker/adaptor was inserted between the BamHI and PuuII sites at positions 1454 and 1709, replacing wild-type sequences. Digestion with BamHI and BglII excised sequences encoding a methionine followed by the FLAG epitope (DYKDDDDK) and all but the first amino acid of the intracellular domain. This construct was named INT. A construct encoding the FLAG epitope and the guanylyl cyclase catalytic domain, but lacking most of the kinase-related sequences, was prepared by inserting the FLAG linker/adaptor described above into the PuuII site at position 2549 of NPR-A. This construct was named CYC. A construct encoding the FLAG epitope and the kinaselike domain, but not the guanylyl cyclase domain, was prepared from the INT construct by inserting a synthetic linker containing a termination codon and NheI site into the FspI site at position 2619 of NPR-A. This construct was named KIN. Insertion of FLAG and myc epitopes carboxyl to the signal sequences of NPR-A, ACYC, and AKC involved several steps. The 0.3-kilobase MluIIApaI fragment of NPR-A was cloned into the corresponding sites of the plasmidpGEM7Zf(+) transmembrane domain (TM), protein kinase-like domain, and cyclase catalytic domain are shown for NPR-A and NPR-B. All mutant constructs represent modifications of NPR-A. Peptide epitopes inserted into the protein are indicated by a flag symbol (FLAG epitope) or circle (myc epitope). Functions of ACYC and AKIN have been described previously (11). The ability of AKC to bind lZ5I-labeled ANP is described under "Experimental Procedures"; the properties of similar constructs have also been described by others (15, 51) as have those of constructs similar to EXT (51, 53). The INT and CYC proteins have guanylyl cyclase specific activities similar to that of full-length NPR-A (M. Chinkers and E. M. Wilson, unpublished observations). Functional properties of the epitope-tagged NPR-A proteins do not appear to be altered relative to the untagged proteins (see Table I).
(Promega), and SphI and NruI sites were introduced at positions 305 and 313 of NPR-A by oligonucleotide-directed mutagenesis (37). A synthetic linker/adaptor encoding either the FLAG epitope or a myc epitope (EQKLISEEDL) was then inserted between the SphI and NruI sites. The resulting MluIIApaI fragments were used to replace the wild-type MluIIApaI fragment in pSVL-NPR-A, pSVL-ACYC, or pSVL-AKC. To express a secreted extracellular domain (EXT), a termination codon was introduced at position 1635 of NPR-A using the polymerase chain reaction. In all cases, the sequences of regions altered from the wild-type cDNA were confirmed by dideoxy sequencing (381, as was the absence of undesired mutations in DNA fragments synthesized during in vitro polymerase reactions. Preparation of a Recombinant Vaccinia Virus Expressing the Extracellular and Transmembrane Domains of NPR-A-The AKC mutant described above was cloned into the XhoI site of vaccinia transfer vector pZVneo, and a vaccinia virus expressing AKC (VV:AKC) was then prepared by homologous recombination with vaccinia virus strain WR DNA (39). Expression of the truncated NPR-A protein by the plaque-purified recombinant virus was confirmed by measuring specific binding of lZ5I-labeled ANP (40) to BSC40 cells (41) infected with wild-type vaccinia virus or VVAKC. Specific binding was approximately 5-fold higher to cells infected with VV:AKC than to cells infected with wild-type virus (approximately 50% versus approximately 10% of total counts/min specifically bound).
Antibodies-Monoclonal antibodies M1 and M2, recognizing the FLAG peptide sequence, were obtained from Immunex and IBI.
Monoclonal antibody 9E10, recognizing a c-myc peptide sequence (42, 43) was obtained from Oncogene Science. Polyclonal antibodies recognizing the extracellular domain of NPR-A were prepared by immunizing female New Zealand White rabbits with lo8 plaque-forming units of VVAKC at a single subcutaneous site, followed by a boost 1 month later. Anti-AKC sera were collected 10 days after each injection, as well as 3 weeks after the booster injection.
Determination of Cyclic GMP in Intact Cells-Incubation of cells for 5 min at 37 "C in the absence or presence of natriuretic peptides (10, 40) and preparation of cell extracts (45) were performed as described. Cyclic GMP determinations were performed by radioimmunoassay (46). One day after transfections, plates were trypsinized, and cells transfected with the same plasmids were pooled and seeded into 12-well plates. Assays were performed on the following day.
Metabolic Labeling and Immunoprecipitation-Metabolic labeling was performed by incubating cells for 4 h at 37 "C in methionine-free Dulbecco's modified Eagle's medium containing approximately 50 pCi/ml [35S]methionine. Cells were then placed on ice, washed twice with cold 20 mM Hepes pH 7.4, 150 mM NaCl, and lysed in 1 ml of cold buffer A (20 mM Hepes pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 5 pg/ml aprotinin). After 10 min on ice, lysates were clarified by centrifugation at 2 "C for 20 min at 18,500 X g. 0.5 pg of monoclonal antibody or 5 pl of polyclonal antiserum were then added to the clarified lysates, followed by incubation on ice for 1 h. Immune complexes were collected by different methods, based on the antibodies' abilities to bind to protein A. We used either an excess of goat anti-mouse IgG beads (Sigma, antibodies M1 and M2, 1-h incubation at 4 "C with continuous end-over-end mixing) or formalin-fixed Staphylococcus aureus cells (Boehringer Mannheim, anti-AKC, 10-min incubation on ice). Immune complexes were then washed 4 X 1 ml with buffer A (M1 or M2) or 3 X 1 ml with buffer A (anti-AKC) and eluted by heating in sample buffer (47). For immunoprecipitations with the Ca2+-dependent monoclonal antibody M1, 10 mM CaClz was included in lysis and wash buffers. Samples were analyzed by SDS-PAGE in 10 or 11% gels (47), followed by staining with Coomassie Blue, treatment with sodium salicylate (48), drying, and fluorography. Molecular weights were determined by comparlson with standards from Bio-Rad.
Western Blotting-To test the immunoreactivity of myc-NPR-A ( Fig. 2, center), COS-7 cells in 6-cm plates were lysed in 1 ml of 20 mM Hepes pH 7.4,40 mM KC1,l mM EDTA, 1 mM EGTA, 10 pg/ml aprotinin by passage 10 times through a 22-gauge needle. After centrifugation for 20 min at 18,500 X g at 2 "C, pellets were resuspended in 100 pl of 20 mM Hepes pH 7.4, 150 mM NaC1, 1% Triton X-100,10% glycerol, 10 pg/ml aprotinin. After incubation for 20 min on ice, centrifugation was repeated, and an equal volume of 2 x sample buffer (47) was added to the supernatants, followed by heating Left, immunoprecipitation of FLAG-NPR-A by anti-FLAG monoclonal antibodies M1 and M2. COS-7 cells were transfected with pSVL or with the indicated cDNA constructs in pSVL, followed by metabolic labeling with [3sS]methionine and immunoprecipitation with 0.5 pg of the indicated antibody, as described under "Experimental Procedures." When present, the FLAG peptide (DYKDDDDK) was used a t a concentration of 10 pg/ml. The position of the FLAG-NPR-A protein is indicated by a bullet. Center, Western blotting of myc-NPR-A using anti-myc monoclonal antibody 9E10. COS-7 cells were transfected with pSVL or with the indicated cDNA constructs in pSVL, and detergent extracts of particulate fractions were analyzed by Western blotting as described under "Experimental Procedures," using anti-myc monoclonal antibody 9E10 or anti-FLAG monoclonal antibody M2. Right, immunoprecipitation of NPR-A and AKC proteins by anti-AKC serum. COS-7 cells were transfected with pSVL vector or with the indicated cDNA constructs in pSVL, followed by metabolic labeling with ["S]methionine and immunoprecipitation with 5 pl of anti-AKC serum (left) or preimmune serum (right), as described under "Experimental Procedures." Positions of NPR-A and AKC are indicated by bullets.
for 5 min a t 95 "C. Thirty pl of each detergent extract was analyzed by SDS-PAGE on a 10% minigel followed by blotting to nitrocellulose membranes (49). Rainbow molecular weight markers (Amersham Corp.) were used as size standards. Blots were blocked overnight in 20 mM Tris, 137 mM NaC1, 0.1% Tween 20, pH 7.6 (TBST) supplemented with 5% nonfat dry milk. Blots were then incubated with the primary antibodies in blocking buffer, washed with TBST, and incubated with peroxidase-conjugated anti-mouse IgG in the blocking buffer. After washing with TBST, the bound antibodies were detected by chemiluminescence (Amersham ECL kit).
To examine co-immunoprecipitation of myc-NPR-A with FLAG-NPR-A ( Fig. 6), 10-cm cultures of co-transfected COS-7 cells were subjected to immunoprecipitation with 1 pg of M2 and 20 pl of antimouse IgG beads as described above, except that cells were not metabolically labeled. Equal aliquots of the immunoprecipitates were analyzed by Western blotting with 9E10 or M2 as described above, except that prestained molecular weight standards were from Bio-Rad.
Materials-FLAG peptide was purchased from Research Genetics. Rat ANP was from Boehringer Mannheim and Sigma; rat BNP and rat CNP were from Bachem. Synthetic oligonucleotides were from Research Genetics, Midland, and Promega. Anti-mouse IgG conjugated to horseradish peroxidase, chemiluminescence reagents, and ECL nitrocellulose membranes were from Amersham. [3sS]Methionine (Tran3'S-label) was from ICN, and '251-labeled succinylated cyclic GMP tyrosine methyl ester (2500 pCi/pg) was from Biomedical Technologies. Other radiochemicals were from Du Pont-New England Nuclear. Enzymes for restriction and modification of DNA were from GIBCO/BRL, Boehringer Mannheim, and New England Biolabs. Vaccinia virus, strain WR, and the pZVneo transfer vector were gifts of Dr. Gary Thomas (Vollum Institute). Reagents for DNA sequencing were from U. S. Biochemical Corp. (Sequenase kit). Other chemicals were of the best grade available from Sigma, VWR, GIBCO/ BRL, and Boehringer Mannheim.

RESULTS
Characterization of Epitope-tagged NPR-A-We inserted sequences encoding peptide epitopes, recognized by commercially available monoclonal antibodies, into the sequence of NPR-A. The "FLAG" epitope (DYKDDDDK) or a myc epitope (EQKLISEEDL) was positioned such that it would be present at the amino terminus of NPR-A after cleavage of the signal peptide. Addition of either epitope did not interfere with the ability of the receptor to bind or be activated by ANP, as judged by ANP-dependent formation of cyclic GMP in intact COS-7 cells transfected with wild-type and epitopetagged NPR-A (Table I).
We tested the abilities of two monoclonal antibodies to the FLAG epitope, M1 and M2, to immunoprecipitate FLAG-NPR-A from COS-7 cells after metabolic labeling with (35S] methionine ( Fig. 2, left). M1 recognizes the FLAG sequence only when it is precisely at the amino terminus of a protein.
A single residue preceding the FLAG sequence prevents recognition, as does the removal of the amino-terminal residue of the epitope (50). In contrast, M2 can recognize internal or amino-terminal FLAG sequences. Both M1 and M2 immunoprecipitated a protein of the expected size (130 kDa). The 130-kDa protein was not immunoprecipitated in the presence of excess FLAG peptide and was not immunoprecipitated from cells transfected with control vector or cells expressing NPR-A lacking the FLAG epitope (Fig. 2, left). These results confirmed the specificity of the FLAG antibodies and suggested that signal cleavage occurred at the predicted site (10, 51).
We also tested the ability of a monoclonal antibody to the myc epitope, 9E10, to react with myc-NPR-A in Western blots (Fig. 2, center). The antibody detected a 130-kDa band in a particulate fraction from cells expressing myc-NPR-A but not from cells expressing FLAG-NPR-A. Expression of FLAG-NPR-A was confirmed by Western blotting with M2. Thus, the expressed myc-NPR-A protein reacts specifically with monoclonal antibody 9E10.
To analyze untagged NPR-A, specific antisera were prepared by immunizing rabbits with a recombinant vaccinia virus expressing the extracellular and transmembrane domains of NPR-A (VVAKC). The antisera, but not the preimmune sera, immunoprecipitated metabolically labeled NPR-A (130 kDa) or AKC (62 kDa) from COS-7 cells transfected with the respective cDNAs but not from cells transfected with control vector (Fig. 2, right). The AKC protein migrates as a doublet on SDS gels; the basis for this heterogeneity has not yet been determined. None of the anti-AKC sera cross-reacted with NPR-B expressed in COS-7 cells (data not shown; see Fig. 8).
Co-immunoprecipitation Assay for Receptor Oligomerization-We reasoned that if NPR-A normally forms dimers or oligomers, heteromers might form between truncated NPR-A proteins and full-length FLAG-NPR-A. In that case, an antibody to the FLAG epitope would immunoprecipitate a complex containing the untagged truncated protein as well as the FLAG-tagged full-length NPR-A. The truncated protein would be distinguished from FLAG-NPR-A by its faster migration during SDS-PAGE. Structures of the truncated proteins used in this study are shown in Fig. 1. As shown in Fig.  3 (top), M2 immunoprecipitates of cells co-expressing FLAG-NPR-A and various truncated NPR-A proteins lacking portions of the intracellular domain contained the truncated proteins (ACYC, AKIN, AKC; 115, 97, and 62 kDa, respectively) as well as FLAG-NPR-A. The truncated proteins were not immunoprecipitated by M2 in the absence of FLAG-NPR-A, indicating that they are immunoprecipitated due to their association with FLAG-NPR-A, and not due to reactivity with M2. Control immunoprecipitations with anti-AKC confirmed the presence of the truncated proteins in the lysates lacking FLAG-NPR-A (Fig. 3, bottom). Thus, NPR-A proteins lacking part or all of the cytoplasmic domain can form complexes with full-length NPR-A. No other proteins were detected that were specifically co-immunoprecipitated with In contrast, FLAG-tagged soluble intracellular NPR-A constructs did not form immunoprecipitable complexes with untagged full-length NPR-A; NPR-A was not co-precipitated by M2 from lysates containing FLAG-tagged intracellular proteins (INT, KIN, CYC; 60, 35, and 33 kDa, Fig. 3, top), and the soluble intracellular proteins were not co-precipitated by anti-AKC from lysates containing NPR-A (Fig. 3, bottom). Control immunoprecipitations with anti-AKC confirmed the presence of untagged NPR-A protein in the co-transfected cells (Fig. 3, bottom). Thus, extracellular and/or transmembrane sequences were necessary for receptor oligomerization.
Ligand Independence of NPR-A Oligomerization-Addition of ANP did not affect the formation of receptor complexes (Fig. 4). The amounts of untagged proteins containing cytoplasmic deletions (ACYC, AKIN, AKC) co-immunoprecipitated with FLAG-NPR-A by M2 were similar in the absence or presence of ANP (Fig. 4, left). In this experiment, cells were treated with ANP before lysis, and ANP was included in the lysis buffer. Thus, NPR-A differs from other singletransmembrane receptors that undergo ligand-induced dimerization (1, 4); NPR-A appears to be a constitutive dimer or oligomer.

FLAG-NPR-A.
b ? X r-

45.
anti- We had not expected to find that oligomerization required extracellular/transmembrane but not cyclase catalytic domain sequences; others have observed dimerization of the purified catalytic core expressed in bacteria (52). It seemed possible that the cyclase domain of the full-length protein might be unable to dimerize in the absence of ligand but that conformational changes after ANP binding might lead to cyclase domain dimerization and activation. Therefore, we examined the ability of full-length NPR-A and of AKIN (which is a constitutively active guanylyl cyclase (11)) to associate with the 33-kDa FLAG-tagged catalytic domain (CYC) in cells treated with ANP (Fig. 4, right). Neither NPR-A nor AKIN was co-precipitated by M2 with the FLAGtagged CYC protein, in the absence or presence of ANP (Fig.  4, right). Based on these experiments, NPR-A oligomerization is ligand-independent and requires extracellular and/or transmembrane sequences.

AKC
Extracellular Sequences Are Sufficient for Oligomerization-To test whether the extracellular sequences alone were sufficient for NPR-A oligomerization, M2 immunoprecipitates were prepared from lysates of cells expressing both FLAG-NPR-A and the secreted extracellular domain of NPR-A, Peptide lacking the transmembrane sequences (EXT). Others have shown that similar constructs are secreted and retain ANP binding activity (51, 53). We reasoned that formation of heteromers between NPR-A and the secreted EXT protein might anchor a fraction of the soluble protein to the cell. As shown in Fig. 5, EXT does associate with FLAG-NPR-A, as evidenced by M2 co-immunoprecipitation of a 56-kDa/57-kDa doublet. The amount of EXT co-precipitated with FLAG-NPR-A is similar to the amount of AKC co-precipitated (Fig.  5). In control immunoprecipitations, EXT was not immunoprecipitated by M2 in the absence of FLAG-NPR-A but was immunoprecipitated by anti-AKC in the absence or presence of FLAG-NPR-A (Fig. 5). Thus, the extracellular domain of NPR-A is sufficient for receptor oligomerization; transmembrane sequences are not required.
Oligomerization of Full-length NPR-A-Although the above experiments showed that NPR-A could form hetero-oligomeric complexes with truncated NPR-A proteins, we were concerned that the ligand independence of this oligomerization, which is somewhat unusual, could be an artifact of receptor truncation. Therefore, we tested the ability of full- .~ length NPR-A to oligomerize in the absence or presence of ANP. Cells were co-transfected with full-length FLAG-NPR-A and full-length NPR-A tagged with a myc epitope (myc-NPR-A). Lysates of these cells were immunoprecipitated with the anti-FLAG monoclonal antibody M2, and the immunoprecipitates were analyzed by Western blotting with M2 or the anti-myc monoclonal antibody 9E10 (Fig. 6). A 130-kDa band recognized by the myc antibody was present in M2 immunoprecipitates from cells co-expressing FLAG-NPR-A and myc-NPR-A, but not from cells expressing either protein alone. ANP had no effect on the amount of myc-NPR-A coimmunoprecipitated with FLAG-NPR-A (Fig. 6). Thus, oligomerization of full-length NPR-A is independent of ligand binding.

NPR-A Oligomerizes in Intact
Cells-A potential artifact of the co-immunoprecipitation experiments was that receptor complexes might be formed only after cell lysis. Therefore, we performed a mixing experiment to test whether the formation of heteromeric NPR-A proteins could occur after cell lysis (Fig. 7). In the control lane, M2 precipitated both FLAG-NPR-A and ACYC proteins from a lysate of cells co-expressing these proteins (Fig. 7). However, when lysates from cells transfected with only FLAG-NPR-A were mixed with lysates from cells transfected with only ACYC, only FLAG-NPR-A was immunoprecipitated from the mixture by M2 (Fig. 7). This was not due to lack of expression of ACYC, as immu- noprecipitation with anti-AKC demonstrated high levels of ACYC protein in the mixture (not shown). The inability of FLAG-NPR-A and ACYC to form immunoprecipitable complexes in lysates indicated that the receptor associations we observed in the co-immunoprecipitation assay formed in the intact cell prior to lysis. Similar results were obtained using AKC in place of ACYC (not shown).

Formation of Heteromeric Natriuretic Peptide Receptors by NPR-A and NPR-B-The observation that NPR-A oligomer-
izes raised the possibility of a heteromeric NPR-A/NPR-B receptor. Such a receptor might have novel ligand specificity or altered responsiveness to the known natriuretic peptides. We tested the ability of untagged NPR-B to associate with an epitope-tagged NPR-A construct in a co-immunoprecipitation assay. COS-7 cells were transfected with NPR-B and a FLAG-tagged truncated NPR-A (FLAG-ACYC), followed by metabolic labeling and immunoprecipitation with M2 or anti-AKC. As shown in Fig. 8 (left), the 130-kDa NPR-B and NPR-A associate with the 115-kDa FLAG-ACYC to similar extents. In addition, anti-AKC immunoprecipitated NPR-B from cells co-expressing FLAG-ACYC, but not cells expressing NPR-B alone (Fig. 8, right). Thus, NPR-A and NPR-B can form heteromers. Similar results were obtained using FLAG-AKC in place of FLAG-ACYC (not shown), indicating that intracellular sequences are not required for association of NPR-A and NPR-B.

Inhibition of Signal Transduction by a Truncated NPR-A-
Others have shown, using other single-transmembrane receptors, that cytoplasmic deletion mutants that dimerize with wild-type receptor can function as dominant negative mutants, apparently due to formation of inactive heterodimers (5-8). Therefore, we examined whether AKC, when co-expressed with NPR-A, could inhibit the ability of the receptor to produce cyclic GMP in response to ANP. Cells were co- transfected with plasmids encoding NPR-A and AKC or with a plasmid encoding NPR-A and control vector. As shown in Fig. 9A, co-expression of AKC shifted the ANP concentration response to the right. At low concentrations of ANP, a dramatic inhibition of cyclic GMP formation was observed in cells expressing AKC. At very high pharmacological concentrations of ANP, the same maximal response was observed in the absence or presence of AKC. Therefore, at physiological concentrations of ANP, AKC is a potent inhibitor of NPR-A function. This shift of the ANP concentration response was not simply due to ANP binding by AKC reducing the effective ANP concentration available to wild-type NPR-A. When cells expressing AKC are co-cultured with cells expressing wildtype NPR-A, we have never observed any shift in the ANP concentration response (Fig. 9B, inset). Curiously in three of four such co-culture experiments, we did observe significant decreases in maximal cyclic GMP production ( Fig. 9 B ) that have never been observed in the co-transfection experiments (Fig. 9A).

DISCUSSION
The noncovalent, ligand-independent oligomerization of NPR-A is unusual among single-transmembrane receptors. Receptors with tyrosine kinase activity and growth hormone receptors, which represent another large family of singletransmembrane receptors, both undergo ligand-dependent oligomerization (1, 4). Considering the topological similarity of NPR-A to tyrosine kinase receptors which undergo liganddependent dimerization, including the preservation of a tyrosine kinase-like domain, one might have expected to observe similar results with NPR-A. In one model for receptor activation, oligomerization functions merely to bring together intracellular domains which then induce conformational changes in each other (1). In another model, oligomerization produces a protein unit having two or more transmembrane helices that can then transduce a signal through liganddependent movements of the helices relative to each other (54, 55). Our data are not consistent with the former model, since oligomerization of NPR-A is constitutive and is clearly not sufficient for receptor activation. We cannot rule out, however, that ligand-induced rotations of oligomerized sub- units relative to each other bring together intracellular domains that are separated in the basal state. In addition, we cannot exclude the possibility that ANP induces higher order oligomers than are present in the basal state, such as a conversion of dimers to tetramers. Finally, it is possible that ligand-induced dimerization of NPR-A is ordinarily responsible for cyclase activation but that overexpression in COS cells led to artifactual ligand-independent oligomerization. We typically observe an average of about 30,000 ANP binding sites per cell for a plate of transfected COS-7 cells, and preliminary immunofluorescence experiments suggest a transfection efficiency on the order of 10%'; thus, transfected cells may express on the order of 300,000 receptors at their surfaces. We do not favor this hypothesis because: 1) it would be surprising if overexpression drove oligomerization to completion so that it could not be further increased by ANP binding (Fig. 4) in the absence of ANP (Table I, Fig. 9), overexpression/ oligomerization has not led to receptor activation. Nevertheless, it will be important to repeat these experiments in stable cell lines expressing more moderate numbers of receptors before concluding that ligand-independent oligomerization of native NPR-A occurs in its natural environment.
The detection of interactions between extracellular, but not intracellular, portions of NPR-A was unexpected. It has been argued that the requirement of adenylyl and soluble guanylyl cyclases for two apparent catalytic domains (25-28) is likely to extend to the receptor/guanylyl cyclases, implying ligandinduced dimerization involving the cyclase catalytic domain (29,30). Consistent with that hypothesis, Thorpe et al. (52) described the dimerization of the purified catalytic core of NPR-A expressed in bacteria. This hypothesis has certain weaknesses. It is not clear that both of the cyclase homology domains in the heterodimeric adenylyl and soluble guanylyl cyclases are catalytically active; one of the domains might play a regulatory role. Also, a direct interaction between these domains has not yet been demonstrated. In the case of the bacterially expressed protein, an amphipathic region upstream of the catalytic domain was retained in the construct and could potentially mediate the observed dimerization. While we failed to observe interactions between intracellular sequences, we suspect that such interactions occur in the cell but are disrupted by the detergent buffer used in our assays. It is difficult to envision generation of an intracellular conformational change in the absence of interactions between cytoplasmic domains.
The size of the NPR-A oligomer is not yet known. The standard sedimentation and cross-linking approaches used to examine this question with tyrosine kinase receptors have not been successful with NPR-A. 1) The variable and sometimes low amounts of oligomer observed in the co-precipitation assays (e.g. Fig. 8) suggested either a low percentage of oligomeric receptor or instability of the receptor complexes in detergent buffer. In attempts to estimate the sizes of the receptor complexes by density gradient centrifugation, we found that all NPR-A migrated as an apparent monomer after overnight centrifugation (data not shown). This suggested that the receptor complexes might be unstable to long incubations in detergent solution. Meloche et al. (36) also observed only monomers when ANP receptor solubilized from adrenal zona glomerulosa was analyzed by density gradient centrifugation and gel filtration chromatography. 2) In preliminary experiments, we were unable to cross-link the receptor complexes with various bifunctional reagents differing in spacer length and reactive groups (disuccinimidyl suberate, bis(sulfosuccinimidy1) suberate, l-ethyl-3-(3-dimethylaminopropyl)carbodiimide, disulfosuccinimidyl tartarate, ethylene glycolbis(sulfosuccinimidylsuccinate), dimethyl pimelimidate, difluorodinitrobenzene), precluding size determination by this method (data not shown). However, during fast protein liquid gel filtration chromatography, which may be performed rapidly enough to avoid problems with stability of the receptor complexes, NPR-A guanylyl cyclase activity expressed in COS cells migrated with an M, of 450,000 in the absence of ANP (data not shown), consistent with NPR-A being tetrameric. Further characterization of NPR-A in the absence and presence of ANP will be necessary to determine an accurate molecular weight for the receptor oligomers.
The fraction of receptor in the oligomeric state is also unknown and is somewhat difficult to address due to the apparent instability of receptor complexes after cell lysis. It is possible that all NPR-A in the plasma membrane is oligomeric or that oligomerization occurs co-translationally. Alter-natively, only a portion of the receptor population may oligomerize, and modulation of oligomerization by unknown factors may regulate responsiveness to natriuretic peptides. In this case, oligomeric NPR-A could represent the higher affinity form of the receptor seen in Scatchard plots (32).
The formation of heteromeric receptors by NPR-A and NPR-B adds a new level of complexity to the biology of the natriuretic peptides. Based on the similar amounts of NPR-A and NPR-B co-immunoprecipitated with an epitope-tagged, truncated NPR-A protein (Fig. 8), NPR-A and NPR-B may associate with each other as readily as with themselves. Koller et al. (15) have postulated that NPR-A represents a receptor for both ANP and BNP, although a 10-fold higher concentration of BNP is required for half-maximal receptor activation (14, 15). They have also suggested that NPR-B is the CNP receptor, although the EGO they reported for stimulation of NPR-B guanylyl cyclase by CNP (100 nM) (15) was surprisingly high for a physiological receptor. It seemed possible that the heteromeric receptor might be more responsive to BNP or CNP than homomeric NPR-A or NPR-B. However, while we were able to reproduce the results of Koller et al., we observed similar cyclic GMP responses to ANP, BNP, and CNP in cells co-expressing NPR-A and NPR-B and in cells expressing either receptor alone (data not shown). This could indicate either that the heteromeric receptor responds to the peptides similarly to the homomeric receptor or that altered responses were undetectable due to the background of homomeric receptors in the co-transfected cells. In order to address this question properly, it may be necessary to generate complementary NPR-A and NPR-B mutants that are unable to form functional homo-oligomers but that can form functional heteromers. The observation by Wilcox et al. (56) that NPR-A and NPR-B are co-localized in adrenal medulla, anterior pituitary, and cerebellum is consistent with a biological role for the heteromeric receptor. One could envision that the heteromeric receptor might show increased or decreased cyclic GMP formation in response to the known natriuretic peptides, could be desensitized or otherwise regulated differently than the homo-oligomeric receptor, or could be specific for an as yet undiscovered natriuretic peptide. Alternatively, the heteromeric receptor may be a laboratory artifact that does not occur in vivo.
The inhibition of NPR-A activation by AKC argues that receptor oligomerization is required for signal transduction. However, some of the details of the experimental observations are puzzling. First, assuming ligand independence of receptor oligomerization, one would expect co-expression of wild-type and truncated receptors to lead to a reduction of maximal ANP-dependent cyclic GMP production, rather than a shift in the concentration response. The observed shift in concentration response, with no change in maximal cyclic GMP production, suggests that ANP can shift the equilibrium to favor formation of wild-type oligomers, seemingly inconsistent with the ligand independence of oligomerization observed in the co-immunoprecipitation assays. A similar shift in concentration response was observed when EGF receptor (which dimerizes in a ligand-dependent fashion) was co-expressed with a truncated EGF receptor (57). Another explanation for the shift in concentration response could be that overexpressed AKC might bind a substantial percentage of the ANP in the medium at low, but not high, ANP concentrations, thus reducing the amount of ANP available to the wild-type receptor. This explanation would seem to be ruled out by the observation (Fig. 9 B ) that co-culturing cells expressing AKC with cells expressing NPR-A did not result in a shift in the ANP concentration response. Curiously, however, it did result in a small but reproducible reduction in the maximal response. Since the co-cultured cells compete for growth in the same plate, this decrease in maximal response could be an artifact of more rapid growth by AKC-transfected cells than by the control vector-transfected cells. It is difficult to imagine other mechanisms by which AKC on one cell may reduce signaling by NPR-A on another cell at high ANP concentrations.
The observation that AKC inhibits signaling by NPR-A could be important not only in elucidating the biochemical mechanisms of signaling but in exploring the normal physiological roles of the natriuretic peptides in cardiovascular homeostasis, which are poorly understood (24). If this dominant negative mutant can be used to block the function of natriuretic peptide receptors in transgenic animals (7), we could test directly the importance of natriuretic peptides in specific physiological responses in vivo.