The NMR Solution Structure of the Relaxin (RXFP1) Receptor Lipoprotein Receptor Class A Module and Identification of Key Residues in the N-terminal Region of the Module That Mediate Receptor Activation*

The receptors for the peptide hormones relaxin and insulin-like peptide 3 (INSL3) are the leucine-rich repeat-containing G-protein-coupled receptors LGR7 and LGR8 recently renamed as the relaxin family peptide (RXFP) receptors, RXFP1 and RXFP2, respectively. These receptors differ from other LGRs by the addition of an N-terminal low density lipoprotein receptor class A (LDLa) module and are the only human G-protein-coupled receptors to contain such a domain. Recently it was shown that the LDLa module of the RXFP1 and RXFP2 receptors is essential for ligand-stimulated cAMP signaling. The mechanism by which the LDLa module modulates receptor signaling is unknown; however, it represents a unique paradigm in understanding G-protein-coupled receptor signaling. Here we present the structure of the RXFP1 receptor LDLa module determined by solution NMR spectroscopy. The structure is similar to other LDLa modules but shows small differences in side chain orientations and inter-residue packing. Interchange of the module with the second ligand binding domain of the LDL receptor, LB2, results in a receptor that binds relaxin with full affinity but is unable to signal. Furthermore, we demonstrate via structural studies on mutated LDLa modules and functional studies on mutated full-length receptors that a hydrophobic surface within the N-terminal region of the module is essential for activation of RXFP1 receptor signal in response to relaxin stimulation. This study has highlighted the necessity to understand the structural effects of single amino acid mutations on the LDLa module to fully interpret the effects of these mutations on receptor activity.

LDLa modules (17); however, the function of this receptor is unknown.
The LDLa modules of the RXFP1 and RXFP2 receptors are homologous to the ligand binding domains found in the LDL receptor and members of the LDL receptor family, thus acquiring their namesake. The structures of several of these LDLa modules have been solved by NMR spectroscopy (18 -24) and x-ray crystallography (25), revealing a number of conserved features. They are all small (ϳ4 kDa) and have a topology dominated by the three conserved disulfide bonds and the organization of the C-terminal region around a calcium ion ligated by a motif of acidic residues, Asp-X-X-X-Asp-X-X-Asp-X-X-Asp-Glu. All LDLa modules structurally characterized to date, including the RXFP1 receptor module (26), have displayed a distinct interdependence on the presence of both the disulfide bonds and a calcium ion to maintain a globular structure.
The discovery of the RXFP1 and RXFP2 receptors has significantly propelled understanding of relaxin signaling. Simplistically, ligand binding stimulates cAMP production through the G s stimulatory pathway. The molecular details of how relaxin binding induces signaling are not yet clear; however, relaxin has been shown to bind primarily to the LRRs and the second extracellular loop of the transmembrane region (27). There is no evidence to suggest that the LDLa module is involved in the binding of relaxin. Conversely, the LDLa module has been demonstrated to be essential to the activation of signaling of the RXFP1 receptor (28). Removal of the LDLa module has no effect on the affinity of relaxin binding but abolishes the capacity of the truncated receptor to activate the response of cAMP production. It is therefore hypothesized that binding of relaxin to the LRRs induces a conformational change in the receptor positioning the LDLa module to interact with the transmembrane region and inducing a second conformational change required for G-protein-activated cAMP production. This apparent mechanism entails a novel role for an LDLa module that is generally involved in protein-protein interactions. As the RXFP1 and RXFP2 receptors are the only human GPCRs to contain an LDLa module, this system represents a paradigm in our understanding of GPCR activation.
In this study we aim to understand how the LDLa module mediates RXFP1 receptor activation. We have solved to high resolution the solution structure of the module by NMR spectroscopy, and we demonstrate that the activity of the module is not interchangeable with another LDLa module of high sequence identity. We describe the effect of several side chain mutations on the structure of the module and signaling function of the RXFP1 receptor. Results of this work highlight that signaling of the RXFP1 receptor is mediated by specific residues of the LDLa module.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification of the RXFP1 LDLa Module-The method for sample preparation is essentially as reported elsewhere (26). Briefly, the RXFP1 LDLa module was cloned from the pcDNA3.1/Zeo plasmid containing the human RXFP1 sequence without the signal peptide (a kind gift from Aaron Hsueh). The DNA fragment encoding amino acid residues Gln-1 to Gly-41 of the mature receptor was amplified by PCR and ligated via the BamHI and XhoI restriction sites into the pGEV2 vector, a thrombin-cleavable GB1 fusion expression vector (the kind gift from Dr. A. Gronenborn, National Institutes of Health). The resulting plasmid pGEV-LDLa was transformed into the T7 bacterial expression strain BL21(DE3)trxB (Novagen) for the expression of the GB1-LDLa fusion protein.
Freshly transformed cells were used for all protein expression cultures. All cultures were grown and protein expression was induced at 37°C. Proteins were 15 N-or 13 C, 15 N-labeled by growing cultures in a 2-liter Braun Biostat fermenter containing 1 liter of minimal media with 15 NH 4 Cl and D-[ 13 C]glucose as the sole nitrogen and carbon sources. Fermentation was conducted as described by Cai et al. (29). Protein expression was induced by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 2.5 h after which cells were harvested, pelleted, and stored at Ϫ20°C.
Fusion protein was purified using IgG-Sepharose (GE Healthcare) via the manufacturer's instructions. The eluted protein was buffer-exchanged into 50 mM Tris-HCl, 150 mM NaCl, pH 8.5, and concentrated to ϳ100 g/l using a 3-kDa cutoff Vivaspin centrifugal concentrator (Sartorius). Protein concentration was adjusted to 100 -300 g/ml in refolding buffer (3 mM GSH, 0.3 mM GSSG, 50 mM Tris-HCl, 150 mM NaCl, 2.5 mM CaCl 2 , pH 8.5). The sample was incubated overnight at 4°C with stirring to allow the complete formation of disulfide bonds. Oxidation was monitored by reversed phase (RP)-HPLC using a C18 analytical column (Vydac) on an Akta Explorer system (GE Healthcare) and a 20 (Buffer A) to 70% (Buffer B) gradient, where Buffer A consisted of 0.1% trifluoroacetic acid, and Buffer B consisted of 80% acetonitrile and 0.1% trifluoroacetic acid. Postoxidation, the GB1 fusion protein was cleaved from the LDLa module overnight by incubating with 1 unit of thrombin protease (GE Healthcare) per mg of fusion protein. The cleaved GB1 was separated from the LDLa module by passing the sample over IgG-Sepharose. The unbound LDLa module was further purified by RP-HPLC as described above but using a Jupiter Proteo 4-m 90-Å column. The eluted protein was lyophilized and stored at Ϫ20°C.
Expression and Purification of GB1-LDLa Mutants-Protein samples of GB1-LDLa mutants were expressed and purified as described above for wild type protein; however, expression of 15 N-labeled protein was achieved using the method described in Marley et al. (30). Briefly, 500-ml cultures were grown on LB media to an A 600 ϳ0.7 at 37°C with shaking. The culture was pelleted and the supernatant decanted, and cells were transferred to 125 ml of minimal medium containing 15 NH 4 Cl as the sole nitrogen source. The culture was equilibrated for 1 h at 16°C with shaking, after which protein expression was induced overnight at 16°C by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside.
Site-directed Mutagenesis-Site-specific mutants of the LDLa domain and the full-length receptor were prepared using the QuickChange PCR mutagenesis kit (Stratagene) following the manufacturer's instructions. Complementary forward and reverse primers were designed to introduce the minimum number of base pair changes necessary to alter the codon. Primer sequences were as follows: Mutations of the recombinant LDLa module used the pGEV-LDLa plasmid. For the preparation of full-length receptor mutants, the pcDNA3.1-RXFP1 plasmid was used. The sequences of all plasmids generated from the mutagenesis reactions were verified by sequencing.
LDLa mutant samples ϳ0.3-0.6 mM were prepared from lyophilized 15 N-labeled protein in 50 mM imidazole, pH 6.0, and 10 mM CaCl 2 . Two-dimensional 15 N, 1 H-HSQC data were routinely collected at 25°C on a 600-MHz Varian Inova spectrometer equipped with a triple resonance cold probe.
Structure Calculations-Distance constraints were obtained from three-dimensional 15 N-edited NOESY-HSQC and 13 Cedited HSQC-NOESY spectra. A total of 29 backbone and angle constraints were generated from TALOS, and 19 1 angles were derived from inspecting peak intensities of threedimensional HACAHB and HNHB spectra. NOE data were assigned using the CANDID (34) module of CYANA (35) (version 1.0.7) with rounds of both manual and automatic assignment. The NOE constraints were further added to and refined with calculations using XPLOR-NIH (36) (version 2.9.9) using a simulated annealing protocol with molecular dynamics in both torsion and Cartesian space. The final force constants used were 50 kcal mol Ϫ1 for experimental distance constraints, 200 kcal mol Ϫ1 rad Ϫ2 for dihedral angle constraints, and 1.0 kcal mol Ϫ1 for the Ramachandran data base potential of mean force (37,38). The quality of structures was validated using PRO-CHECK-NMR (39), and MOLMOL (40) was used to inspect structures, calculate ring current shifts, and produce figures.
Chemical Shift Mapping-Assignment of spectra of the mutant proteins was guided by the wild type LDLa module chemical shifts (26). Yields of folded Y9A and F10A were poor reducing the sensitivity of the spectra; therefore, spectra of the fusion proteins GB1-LDLa Y9A and GB1-LDLa F10A were compared with GB1-LDLa. Chemical shift differences were scaled according to Equation 1 (41), where ⌬␦ HN and ⌬␦ N are the chemical shift differences for the NH and 15 N resonances of the mutant and wild type proteins. Errors were calculated as a ratio of peak line width to signal to noise and scaled according to Equation 1.
Cloning of LB-RXFP1 Chimera-The LB2-RXFP1 chimera was created by amplifying the DNA segment encoding the LB2, bp 198 -312 of the LDL receptor, from the pGEX LB1-LB2 plasmid (kind gift of Ross Smith, the University of Queensland) with EcoRI sites at both the 5Ј and 3Ј ends. The primers designed were as follows: LB2EcoRIFwd, CATCATGAATTC-GTCACCTGCAAATCCGGG; RevLB2EcoRI, CATCAT-GAATTCACAGCCTTGCTCGTCTGAGCC.
The pcDNA3.1/Zeo plasmid containing the LGR7 short construct (28), missing the LDLa module, retained an EcoRI site following the bovine prolactin signal sequence and FLAG epitope, prior to the beginning of the LGR7 sequence. The LB2 insert and the pcDNA3.1 LGR7 short plasmid were therefore simultaneously digested with EcoRI and ligated together to create the LB2-RXFP1 chimeric receptor. The correct orientation of insertion and sequence was verified by sequencing. 33 P-Labeled H2 Relaxin Binding Assays-Human embryonic kidney (HEK) 293T cells grown in RPMI 1640 media (Sigma) supplemented with 10% fetal bovine serum, 100 g/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine were plated into 24-well poly-L-lysine-coated plates for whole cell binding assays. Upon reaching 80% confluency, cells were washed with 1 ml of PBS and transiently transfected with 1 g of pcDNA3.1/Zeo plasmid containing the receptor of interest in Lipofectamine 2000 (Invitrogen) and Opti-MEM serum-free media (Invitrogen). 5 h after transfection, 0.5 ml of RPMI 1640 media supplemented with 10% fetal bovine serum and 2 mM L-glutamine was added to each well and further incubated for 16 h. Competition binding assays were performed with 100 pM of 33 P-labeled H2 relaxin in the absence or presence of increasing concentrations of unlabeled H2 relaxin. Data were collected as triplicate points in three independent experiments and fitted to one-site binding curves in Graphpad Prism 4.
Anti-FLAG Cell Surface Expression Assay-HEK293T cells transfected with pcDNA3.1/Zeo plasmids containing the FLAG-tagged receptor of interest were plated into 6-well poly-L-lysine coated plates (1,000,000 transfected cells per well). 24 h later, cells were washed with 1 ml of PBS before incubation with 1 ml of a 5% BSA/PBS solution containing 5 g/ml anti-FLAG M2 antibody (Sigma) for 1 h at room temperature on a gently tilting platform. Cells were then lifted from wells via washing and resuspension to 1.5 ml in the above solution and centrifuged for 2 min at 14,000 rpm in a Heraeus Biofuge 13 (Sepatech). The antibody solution was aspirated, and the cell pellet was washed twice in ice-cold PBS prior to resuspension in 200 l of ice-cold solution of 5% BSA/PBS containing 2 ϫ 10 6 cpm/ml of 125 I-labeled anti-mouse IgG sheep antibody (Amersham Biosciences). The cell suspension was incubated on ice for 2 h and then subjected to two cycles of centrifugation, aspiration, and washing with 1 ml of 1% BSA/PBS. The final pellet was counted in a Cobra Auto-Gamma liquid scintillation analyzer (Packard Instrument Co.). All points were collected in triplicate in three independent assays. The level of nonspecific binding was determined using cells transfected with an empty vector.
cAMP Enzyme-linked Immunosorbent Assay-HEK293T cells in 48-well plates were transiently transfected upon reaching ϳ80% confluency with 0.5 g of pcDNA3.1/Zeo plasmid containing the receptor of interest in Lipofectamine 2000 (Invitrogen) and Opti-MEM serum-free media (Invitrogen). The transfection media were aspirated after ϳ24 h and replaced with RPMI 1640 media containing 0.25 mM 3-isobutyl-1-methylxanthine and the appropriate concentration of H2 relaxin and incubated for 30 min at 37°C. The maximum response capacity for cAMP production of the cells was determined by stimulating cells with 5 M forskolin in the presence of 0.25 mM 3-isobutyl-1-methylxanthine. Upon stimulation the media containing treatments were aspirated. Cell lysis and measurement of intracellular cAMP produced in response to treatments were performed using a cAMP Biotrak enzyme immunoassay kit (GE Healthcare) according to the manufacturer's protocol. Each treatment point was carried out in triplicate in three independent experiments. Data were analyzed using Graphpad Prism 4 and represented as the mean Ϯ S.E. Significance was determined using a one-way analysis of variance and Bonferroni post-test comparison.
Isothermal Calorimetry-Protein samples purified by RP-HPLC to remove residual calcium were reconstituted in 100 mM Tris-HCl, pH 7.4, 100 mM NaCl 2 buffer treated with Chelex. Experiments were conducted at 30°C on a Microcal MCS ITC unit (Microcal Inc.) based on methods described previously (23,42). Typically, 2-l injections of 2-10 mM CaCl 2 were made into 5 or 10 M samples of wild type protein and GB1-LDLa L7A and 50 M samples of GB1-LDLa F10A or GB1-LDLa Y9A with 180-s delays between injection to allow sufficient recovery. The peaks generated by heat of binding were integrated and data-fitted using the program MicroCal Origin version 2.9 assuming a single site model and fixing the value of N to 1 to generate the best fit. The reported apparent K d values are the means Ϯ S.E. averaged from three independent experiments.

NMR Solution Structure of the RXFP1 LDLa Module-The
NMR solution structure was calculated based on distance constraints from three-dimensional 15 N-and 13 C-edited NOESY spectra. A total of 257 distance constraints and 48 dihedral angles was used to calculate a good quality structure ( Table 1). The ensemble of 24 best structures presented in Fig. 1 indicate that the structure is well defined between residues 5 and 38 with a backbone r.m.s.d. value of 0.46 Å. The first two residues, Gly(Ϫ1) and Ser(Ϫ2), which are cloning artifacts, show no NOE data and are therefore unstructured. The next four residues are fully assigned, but equivalent side chain groups typically show degeneracy, for example Val-3 shows only one methyl peak in both 1 H and 13 C dimensions. Little NOE data are observed for these residues, consistent with structural disorder, and most likely reflect the state of these residues in the intact receptor.
The remainder of the module shows a general topology consistent with other LDLa modules in the Protein Data Bank. The structure of the module can be divided into two regions as follows: an N-terminal region consisting of a short anti-parallel ␤-sheet (strand 1, residues 9 -11, and strand 2, residues 18 -20) immediately followed by a 3 10 helix (residues 21-23); and a C-terminal region that is predominantly organized around a calcium ion (Fig. 1). NOE data, particularly ␣ i ␤ iϩ3 and ␣ i NH iϩ3 NOEs, suggest that residues 33-37 are organized into a 3 10 helix, consistent with the topology of the LB5 crystal structure (25), but this element of secondary structure is not represented by MOLMOL (Fig. 1). Based on sequence alignment of other LDLa modules and analysis of chemical shift changes during calcium titrations, the residues involved in calcium ligation are Asn-26, Asp-30, Asp-36, and Glu-37 (26). The 1 angles of Asn-26, Asp-36, and Glu-37 are well defined consistent with limited rotation of these side chains. As the H␤ protons of Asp-30 are degenerate, we could not determine its 1 . The orientation of these side chains in structures calculated in the absence of calcium were consistent with their role in calcium ligation (Fig. 1), and the addition of calcium in final structure calculations did not produce any violations of the experimental data and only a minor improvement in r.m.s.d. (from 0.51 to 0.46 Å for the backbone atoms of residues 5-38).
The N-and C-terminal regions pack against each other through modest hydrophobic contacts, consistent with the general structural features of other LDLa modules (43). The side chain of Phe-10, which is in the center of ␤-strand 1, packs against the imidazole ring of His-24, which is located at the C-terminal end of the first 3 10 helix and prior to Cys-25. This is consistent with the x-ray structure of LB5 (25) (Fig. 1) and the  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 NMR structures of LB1 (19) and LB2 (18) in which a Phe at the equivalent position packs against an unconserved residue (for example, Val and Arg) also prior to a cysteine. Of interest, Phe-10 in RXFP1 LDLa adopts a 1 of ϩ60, whereas in other LDLa modules it is ϩ180°. Signal intensities in HNHB (strong and weak) and HACAHB (weak and weak) are consistent with this orientation. Furthermore, no peak is observed in a 15 N-{ 13 C␥} spin-echo difference 1 H, 15 N HSQC experiment (44) confirming that the 1 of Phe-10 is not 180°. A hydrophobic patch near Phe-10 and His-24 that extends to the surface includes Leu-19 and Pro-20 (␤-strand 2) and Leu-22 and Leu-23 (first 3 10 helix). The C ␦1 H 3 of Leu-19 shows weak NOEs to the aromatic ring of Phe-10. The hydrophobic patch continues toward the N terminus, including Leu-7 and Tyr-9, and the latter is in ␤-strand 1, packs against Pro-20, and is oriented in the opposite direction to Phe-10.

Residues of the RXFP1 Receptor That Mediate Signaling
The surface charge of the module is predominantly acidic within the C-terminal region. However, most of these residues (Asp-30, Asp-36, and Glu-37) are involved in calcium binding. The remaining two acidic residues Asp-29 and Asp-38 are surface-exposed but distant from each other. At the N-terminal end of the protein are three basic residues Lys-4, Lys-17, and Tyr-9. Lys-4 is positioned among the largely disordered region prior to Cys-5, and therefore the orientation of Lys-4 is difficult to determine; however, structures suggest that it could cluster with Lys-17 to form a basic surface. Tyr-9 is not a part of this basic patch and packs against Pro-20 as a part of a largely hydrophobic surface patch. Its side chain OH, however, is oriented into the solvent, and there is no evidence of it forming a hydrogen bond as suggested for the equivalent residue in other LDLa modules (43).
Effects of Mutation to Leu-7, Tyr-9, Phe-10, Leu-22, Leu-23, and Cys-5/18 to the RXFP1 LDLa Module Refolding and Structure-Structure and sequence alignments of the LDLa module of RXFP1 with other LDLa modules highlights several residues and regions that may significantly contribute to the structure or function of the module. Of particular interest is a hydrophobic patch in the N-terminal region of the module consisting of residues Leu-7, Tyr-9, Leu-19, Pro-20, Leu-22, and Leu-23 (Fig. 2). Potentially this hydrophobic surface may be involved in a protein interaction that drives activation of the receptor. Leu-7, Tyr-9, and Leu-22 are significantly solventexposed (Ͼ30%) and were therefore mutated to alanine to test the hypothesis that these residues are involved in the unknown interaction that leads to receptor activation. Leu-19 (10%), Pro-20 (15%), and Leu-23 (18%) are all buried; however, Leu-23 was also mutated to alanine to test whether it has a role in recognition or is essential for structure. Phe-10, which is highly conserved in LDLa modules, is located on the opposite side of the module to the hydrophobic surface. It is largely buried and links the N-and C-terminal regions and was therefore mutated to alanine to test its structural and functional importance. In addition, the two cysteine residues, Cys-5 and Cys-18, which form the first disulfide bond and tether the structure of the N-terminal region, were mutated to serine. A similar mutation in LB5 of the LDL receptor did not disrupt the oxidation of the Although not represented by MOLMOL, NOE data suggest that there is a second 3 10 helix in the C-terminal region of RXFP1 LDLa as seen in the LB5 structure. D, comparison of side chain orientation of Phe-10 in relation to LB5. The side chains of RXFP1 LDLa are blue and those of LB5 are red. Both phenylalanine residues pack against an unconserved residue; however, the 1 of the RXFP1 Phe-10 is ϩ60, whereas in LB5 and other LDLa modules it is ϩ180. E, overlay of residues that bind the calcium ion necessary for structure. Residues of the RXFP1 LDLa module are colored blue and LB5 red. Numbering is by the RXFP1 sequence.
remaining two disulfide bonds, and ligation of the calcium ion was retained whereas the N-terminal region became disordered (25). If such a mutation could be made in RXFP1, the functional importance of both the N-and C-terminal regions could be tested.
All mutations were made to the pGEV-LDLa vector for the recombinant expression of the LDLa module as a fusion to GB1 harboring the desired mutation. The ability for these mutated proteins to refold and maintain structure under the same conditions as the wild type protein was assessed via RP-HPLC and NMR spectroscopy. RP-HPLC traces show that the RXFP1 LDLa module when oxidized in vitro in the presence of calcium (26) refolds to a single native structure that elutes as a single peak. In contrast, if the LDLa module is oxidized in the absence of calcium, the protein elutes as several broad peaks representing the formation of several disulfide-bonded isomers of the protein. When correctly folded, the LDLa module shows well dispersed two-dimensional 15 N, 1 H-HSQC spectra consistent with a globular structure (26).
The mutation of either Leu-7 or Leu-22 does not affect the folding of the protein to a single native-like species in the presence of calcium. Both proteins, GB1-LDLa L7A and GB1-LDLa L22A, when oxidized in the presence or absence of calcium produce RP-HPLC elution profiles similar to the wild type protein (Fig. 3). Both proteins give two-dimensional 15 N, 1 H HSQC spectra with chemical shifts similar to those of the wild type LDLa module (Fig. 4). The differences in the chemical shift of the peptide 15 N and NH of the LDLa mutants were compared with the wild type protein to identify structural perturbations caused by the mutation. Several mutants were studied as GB1 fusions, and in these cases chemical shifts were compared with wild type also as a GB1 fusion. As expected, the mutations cause a local perturbation that is most likely because of the change of the side chain group to an alanine. Other than this local effect, the chemical shift mapping of the LDLa L7A and LDLa L22A resonances compared with the wild type protein revealed that neither mutation inflicted large perturbations to the structure of the LDLa module distant from the site of mutation (Fig. 5).
To determine whether calcium binding has been affected, we have also analyzed apparent calcium affinities by ITC. The apparent K d values of calcium binding to wild type was estimated to be 20 Ϯ 0.38 M ( Table 2) and is within the range of K d values (0.5-40 M) reported for other LDLa modules (23,42,(45)(46)(47)(48)(49)(50). The apparent K d value for the mutant GB1-LDLa L7A is 13.5 Ϯ 0.8 M, which is also similar to wild type. As there is no evidence to suggest mutation of Leu-22 affects module structure or its activity (see below), the affinity of calcium binding of GB1-LDLa L22A was not determined. Collectively, characterization of LDLa L7A and LDLa L22A demonstrated that neither mutation disrupts refolding or stability of the module.
Mutation of Leu-23 to alanine disrupts the ability of the LDLa module to refold to a single isomer, and consequently GB1-LDLa L23A elutes as several broad peaks on RP-HPLC (Fig. 3). The mutation also appears to introduce an in vitro instability to the protein. Upon concentration of the oxidized isoforms of the protein separated by RP-HPLC, the protein was prone to precipitation both prior to and post-cleavage of the GB1 fusion protein. The carbonyl of Leu-23 is expected to contribute to calcium ligation (25), and these results may reflect an impairment of this function in vitro. Consequently it was not possible to generate enough material to further characterize GB1-LDLa L23A by either NMR spectroscopy or ITC.
Phe-10 is conserved throughout most LDLa modules; therefore, we anticipated that mutation of this residue may result in a protein domain of disrupted structure. Indeed, the RP-HPLC elution profiles of the mutant protein oxidized in the presence and absence of calcium indicate that the mutation reduces the capacity of the protein to refold to a single disulfide species (Fig.  3). Each eluted HPLC peak was collected, and two-dimensional 15 N, 1 H HSQC spectra were acquired on the fractions in the presence of calcium. Only one fraction gave spectra of dispersed chemical shifts, which in most cases are similar to those in spectra of the wild type protein (26) (Fig. 4) suggesting that this fraction is correctly folded. Two-dimensional 1 H, 15 N HSQC spectra collected for the other fractions showed resonances clustered in the random coil region indicating the protein in these fractions could not adopt a globular structure. Comparison of NH and 15 N resonances of GB1-LDLa F10A in the spectra of the fraction that gave dispersed resonances compared with those of wild type GB1-LDLa showed that the largest chemical shift differences (Ͼ0.1 ppm) occur for Lys-17, Asn-26, Val-28, Asp-30, Gln-34, Asp-36, and Glu-37. These residues, except for Lys-17, are from the C-terminal region, and the carboxylate-containing residues are all involved in the ligation of the calcium ion. Further inspection of spectra collected FIGURE 2. Top view of the RXFP1 LDLa module structure. The orientation of the side chains of residues Leu-7, Tyr-9, Leu-22, and Leu-23 suggest these residues cluster to create a hydrophobic surface on the molecule. N-Linked glycosylation of Asn-14 has been reported recently not to be essential for receptor activity (28), suggesting that this region is not involved in the primary activation of RXFP1 receptor signaling.  (Table 2), which is much higher than the K d of 20 M for wild type protein. Two factors may contribute to this higher K d value as follows: the possibility that the peak was not fully separated from other isomers of the protein via HPLC and the perturbation of the side chains of those residues that bind calcium.
Tyr-9 is positioned on the opposite side of the ␤-sheet to Phe-10 and forms a part of the hydrophobic surface (Fig. 2). Mutation of Tyr-9 to alanine also reduces the capacity of the LDLa module to refold to a single isomer. The GB1-LDLa Y9A oxidized in the presence of calcium elutes as several broad peaks with a similar profile to GB1-LDLa F10A (Fig. 3). Two-dimensional 15 N, 1 H HSQC spectra were acquired for each fraction resolved on RP-HPLC, but only one generated spectra that resembled those of the wild type protein. Chemical shift mapping of 15 N GB1-LDLa Y9A compared with wild type protein suggests that the mutation has not significantly affected the structure of the protein (Fig. 5). However, similar to GB1-LDLa F10A, resonances that were assigned to the LDLa Y9A module were of much lower signal intensity than those of the GB1 fusion protein. An apparent K d value for calcium binding of ϳ350 M was determined for the fractions corresponding to folded protein, which is much higher than the K d values observed for wild type and GB1-LDLa F10A. This infers that the reduced binding affinity of GB1-LDLa Y9A for calcium is contributed by residual unfolded protein and/or as suggested by the RP-HPLC traces reflects a role of Tyr-9 in the stabilization of the structure upon calcium binding. GB1-LDLa C5/18S oxidized in the presence or absence of calcium elutes predominantly as a single peak on RP-HPLC (Fig. 3). Two-dimensional 15 N, 1 H HSQC spectra acquired on the various fractions as either GB1 fusions or isolated LDLa modules and in the presence of calcium show that only the major peak has significant amounts of protein. However, the chemical shifts were clustered in the region of 8 -8.5 ppm indicative of disordered structure (supplemental Fig. S1). As the resonances in the spectra of this species of LDLa C5S/C18S could not be assigned, it was not possible to assess the global GB1-LDLa L7A and GB1-LDLa L22 generate elution profiles very similar to the wild type protein that elutes as a single peak when oxidized in the presence of calcium or as multiple peaks when oxidized in the presence of EDTA. Refolding of GB1-LDLa Y9A, GB1-LDLa F10A, and GB1-LDLa L23A appears to be disrupted by each mutation, as the proteins oxidized in the presence of calcium elute as several peaks. The fraction collected that appeared to be folded by NMR is indicated with shading and an arrow. The other fractions collected were assessed as unfolded by NMR. effect of the mutation using chemical shift difference mapping. These results suggest that in the absence of the first disulfide bond the protein is unstructured, and this differs with similar mutations of the fifth module of the LDL receptor that showed the C-terminal region could independently refold and retain its capacity to bind calcium, albeit lower (25).

Replacement of the RXFP1 Receptor LDLa Module with the Second Ligand Binding Domain (LB2) of the LDL Receptor-
The LDLa module of RXFP1 is essential for receptor activation (28). To demonstrate that the function of the RXFP1 LDLa module depends on specific side chain interactions, we replaced it with the second ligand binding domain (LB2) of the LDL receptor producing the chimera LB2-RXFP1. LB2 was chosen as it has a high level of sequence identity (48%) with the RXFP1 receptor LDLa module, and the structure solved by NMR (18) is also a typical LDLa fold (Fig. 6). The LB2-RXFP1 receptor was transfected into HEK293T cells and tested for cell surface expression and affinity for relaxin. These assays confirmed that replacement of the LDLa module with LB2 had no effect on trafficking of the receptor to the cell surface or the binding of H2 relaxin (Fig. 6). To determine the potential of LB2 to mediate RXFP1 signaling, the receptor chimera was stimulated with H2 relaxin in a dose-dependent manner. Compared with wild type RXFP1, the LB2-RXFP1 chimera failed to stimulate the production of intracellular cAMP (Fig. 6). These data support the concept that specific residues in the LDLa module of the RXFP1 receptor are required for activation.
Cell Surface Expression of RXFP1 Receptor Mutants L7A, Y9A, F10A, L22A, L23A, and C5S/C18S-To further investigate which residues or regions are important in receptor activation, HEK293T cells were transfected with a pcDNA3.1 vector expressing the RXFP1 receptor with the mutations described above (L7A, Y9A, F10A, L22A, L23A, and C5S/C18S). Initially, the effect of bearing a potentially unstructured LDLa module on cell surface expression was determined. All the mutant receptors produced in this study were expressed on the cell surface at levels equal to the wild type receptor (Fig. 7), providing confidence that any loss of relaxin binding or signaling phenotypes was not because of low expression of the receptor at the cell surface. As in vitro GB1-LDLa C5S/C18S is unstructured, these data demonstrate that expression of an unstructured LDLa module does not reduce trafficking of the receptor to the cell surface.
Relaxin Binding to RXFP1 Receptor LDLa Module Mutants L7A, Y9A, F10A, L22A, L23A, and C5S/C18S-The affinity of the mutant receptors for H2 relaxin was next determined using a 33 P-labeled H2 relaxin competition assay ( Table 3). The results of these assays indicate that binding of relaxin is not altered by any of the mutations introduced in this study (supplemental Fig. S2), and correlate with previous reports that the LDLa module is not involved in primary binding of relaxin, because an RXFP1 receptor lacking the LDLa module binds relaxin with the same affinity as the full-length receptor (28).
Relaxin-induced cAMP Levels Generated by RXFP1 Receptor Mutants L7A, Y9A, F10A, L22A, L23A and C5S/C18S -To determine the signaling potential of the mutant RXFP1 receptors, transfected HEK293T cells were dose-dependently stimulated with increasing amounts of H2 relaxin, and cAMP produced intracellularly was measured using an enzyme immunoassay (Fig. 8). As expected the mutation C5S/C18S of RXFP1 showed no activation. The in vitro refolding experiments described above indicate that the LDLa module missing the first disulfide will not fold and consequently should not activate.
The mutations of Leu-7 and Leu-22 elicited responses of increasing cAMP production (Fig. 8). However, the dose-response curves of both mutant receptors appear to be rightshifted compared with those of the wild type receptor. The pEC 50 of RXFP1 L22A of 9.980 Ϯ 0.33 does not statistically differ from the wild type receptor of 10.24 Ϯ 0.32 (p Ͼ 0.05), whereas the pEC 50 of RXFP1 L7A is higher at 9.337 Ϯ 0.17 (Table 3) and represents a significant lower efficacy than the wild type receptor (p Ͻ 0.001). As the in vitro calcium-induced refolding and the NMR spectra analysis of the L7A mutant show minimal perturbation to the structure, we conclude that the decrease in RXFP1 receptor signaling is a direct perturbation of a protein-protein interaction that leads to receptor activation.
The in vitro refolding experiments described for the mutations Y9A, F10A, and L23A suggest that in vivo there could be multiple disulfide isomers or a significant population of  GB1-LDLa F10A (B). All spectra were collected at 25°C in 50 mM imidazole, pH 6.0, and 10 mM CaCl 2 . The spectrum of 15 N LDL-A L7A is similar to the wild type protein (27). Resonances in spectra of 15 N GB1-LDLa F10A reveal that the protein is folded; however, the intensity of the resonances that are assigned to the LDLa module are much lower than those that belong to the GB1 protein. (The spectrum of GB1-LDLa F10A has not been fully annotated for clarity. Example resonances belonging to GB1 are circled.) Notably, the chemical shifts of Asp-30 and Asp-36 indicate calcium ligation, but compared with those of wild type and LDLa L7A they are significantly shifted suggesting a structural perturbation near the calcium-binding site. FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 unfolded protein despite high concentrations of extracellular calcium; therefore, it may not be simple to interpret signaling experiments for these mutations. Nevertheless, both RXFP1 Y9A and RXFP1 L23A receptors retain the capacity to activate signaling; however, dose-response curves were significantly right-shifted with an pEC 50 of 8.836 Ϯ 0.24 (p Ͻ 0.01) and 9.116 Ϯ 0.20 (p Ͻ 0.001), respectively, whereas RXFP1 F10A fails to signal ( Fig. 8; Table 3). These data combined with the chemical shift mapping suggest that the inability of the RXFP1 F10A receptor to signal cAMP production could be a direct consequence of a perturbation to the C-terminal region. However, the lack of signaling may be due to few receptors at the cell surface containing folded LDLa modules.

Residues of the RXFP1 Receptor That Mediate Signaling
The signaling experiments demonstrate that RXFP1 Y9A and RXFP1 L23A receptors can express folded and functional LDLa modules at the cell surface. The dose-response curve of the RXFP1 L23A receptor shows that the maximum levels of cAMP production are equivalent to the wild type receptor, as is the case for RXFP1 L7A and RXFP1 L22A. Hence, it is likely that the RXPF1 L23A receptor has a fully folded LDLa module at the cell surface and that Leu-23 is involved in a specific inter-   action that elicits cAMP signaling. As it was not possible to further characterize the recombinant GB1-LDLa L23A mutant because of instability, it was not possible to determine whether the reduction of receptor efficacy is the direct loss of interactions involving the leucine side chain or mutation of Leu-23 has perturbed other essential residues. In contrast, RXFP1 Y9A receptor can only stimulate ϳ40% of the maximum response of the wild type receptor suggesting that there is a significant proportion of unfolded LDLa module in the mutated receptor. The chemical shift mapping of GB1-LDLa Y9A indicated that the mutation does not significantly perturb the structure of the protein if it can fold correctly. If this is the case, we would anticipate that the RXFP1 Y9A receptors that express a folded LDLa module would have the same maximal cAMP response as the wild type receptor, and any pEC 50 changes would reflect a role for Tyr-9 in receptor signaling. However, the presence of RXFP1 Y9A receptors with unfolded LDLa modules will result in attenuated cAMP responses as well as a minor right shift in the dose-response curves, as these receptors will compete for relaxin binding with the functional receptor. However, as the efficacy of RXFP1 Y9A has been decreased 25-fold, equivalent to 3.8% of the wild type response, while retaining 40% of the maximum cAMP stimulation compared with wild type receptor, it is more than likely that this is a reflection of Tyr-9 being involved in a specific interaction involved with receptor signaling. This is consistent with our initial hypothesis that this resi-due is a part of a hydrophobic surface involved in signal activation and the observation that Leu-7, which is a part of that surface is important for LDLa module function.

DISCUSSION
The LDLa modules of the RXFP1 and RXFP2 receptors are critical for eliciting the stimulation of intracellular cAMP production upon ligand binding (28). The molecular details of the mechanism by which the LDLa module elicits cAMP signaling are unknown. The elucidation of this mechanism, however, would present a paradigm in the understanding of GPCR activation. In an effort to gain insight into the mechanism by which the RXFP1 receptor LDLa module mediates receptor signaling, the structure of the module was first determined via solution NMR spectroscopy. This was followed by analysis of LDLa mutants in both recombinant proteins produced in bacteria and in the full-length receptors expressed in mammalian cells. To comprehensively understand the local or global effect of mutation and the loss of function of the RXFP1 receptor, we tested calcium-dependent in vitro refolding, compared chemical shift data in NMR spectra, and determined calcium binding affinity of the recombinant proteins while in parallel assessing receptor cell surface expression, relaxin binding, and relaxininduced cAMP signaling in full-length receptor LDLa mutants.
The NMR structure of the RXFP1 receptor LDLa module has retained many of the features common to these domains, such as the general fold that is dominated by six conserved cysteine residues and the ligation of a calcium ion by a motif of acidic residues also largely conserved. LDLa modules are well defined for their function in lipoprotein binding and metabolism as repeating domains in members of the LDL receptor family. Apart from the Tva receptor LDLa module (24,51), the structure of the RXFP1 receptor LDLa module is the only other LDLa module that is not associated with the LDL receptor family to be structurally characterized to date. The solution structure of the RXFP1 receptor LDLa module can be considered as divided into N-and C-terminal regions. The N-terminal region of the structure includes a short anti-parallel ␤-sheet consisting of residues 9 -11 and 18 -20. The loop between these strands contains an N-linked glycosylation site at Asn-14 (28). Immediately following the ␤-sheet is a short 3 10 helix (residues 21-23). The C-terminal region contains the conserved acidic motif responsible for the ligation of the calcium ion. Because of the high level of sequence conservation over this region of the molecule, between all LDLa modules the C-terminal region is most structurally similar to the other modules characterized to date, including LB2 (18) and LB5 (25).
Despite the high level of sequence and structural similarity shared between the RXFP1 receptor LDLa module and the second ligand binding domain of the LDL receptor, LB2, the modules were not interchangeable. The LB2-RXFP1 receptor chimera bound relaxin with full affinity in comparison with the wild type receptor, but it was unable to elicit intracellular cAMP production. This observation implies that there are specific residues in the RXFP1 receptor LDLa module essential for receptor activation. Sequence and structural alignment of the two modules highlights that the majority of differences between the two are in the N-terminal region of the modules. From here it was hypothesized that should the acidic C-terminal region  organized around the calcium ion in fact be the region of the module that directs signal activation, then the modules should have been interchangeable to retain at least a modest level of receptor efficacy. As the LB2-RXFP1 receptor was inactive, this suggests that the N-terminal region of the LDLa module is primarily involved in the unknown interaction that leads to the activation of signaling. Further inspection of the structure of the LDLa module shows a clustering of hydrophobic residues, including Leu-7, Tyr-9, Leu-22, and Leu-23. The prominence of this area suggests a role for this region in signal activation. Importantly, most of these residues are conserved in RXFP1 sequences from other mammalian species (16). The converse side to the hydrophobic patch includes the loop between the two ␤-strands, which includes the N-linked glycosylation site at Asn-14. Because the absence of glycosylation at this site has been reported to not reduce receptor efficacy (28), it seems unlikely that this region of the module is directly involved in the interaction that leads to receptor activation. We therefore performed a series of mutations within the N-terminal region to assess their effects on LDLa structure and function.
The disulfide bond between Cys-5 and Cys-18 that tethers the N-terminal region of the structure was removed via mutation of these residues to serine in order to probe the importance of global structure in this region for signal activation. The equivalent mutation in the LB5 module did not disrupt the oxidation of the remaining two disulfide bonds, and the C-terminal region retained calcium ion ligation (25). However, in the presence of this mutation the recombinant GB1-LDLa protein was unable to refold or bind calcium. In two-dimensional 15 N, 1 H HSQC spectra of this protein, acquired in the presence of calcium, the 1 H resonances were unassignable as they clustered between 8 and 8.5 ppm, implying LDLa C5S/C18S is unable to adopt a defined globular structure that may reflect the lowered affinity of calcium binding that the RXFP1 receptor has (ϳ20 M) in comparison with LB5 (0.5 M) (23). Despite this fact, RXFP1 C5S/C18S receptors, assumed to express with unfolded LDLa modules, are trafficked to the cell surface at levels equivalent to the wild type receptor with folded modules.
Mutation of Leu-7 and Leu-22 does not impact the refolding of the LDLa module in vitro. Additionally, both the RXFP1 L7A and the RXFP1 L22A receptors elicit a maximum cAMP response similar to the wild type RXFP1 receptor, highlighting that they have fully folded LDLa modules at the cell surface. The dose-response curve of both receptors is rightshifted in comparison with the wild type receptor; however, only the shift of the RXFP1 L7A receptor dose-response curve was significant. These data demonstrate the L22A does not make a significant contribution to either structural integrity or receptor function while supporting the hypothesis that Leu-7 is involved in a specific interaction that leads to cAMP signaling.
Tyr-9 and Leu-23 are structurally close to Leu-7 and Leu-22, but the results from their mutation are not simply interpreted. The refolding profile of the GB1-LDLa fusions of both mutants demonstrates that Tyr-9 and Leu-23 are essential for stabilizing the structure during in vitro calcium-dependent refolding. GB1-LDLa Y9A is unable to refold to a single disulfide isoform; however, the RXFP1 Y9A cell surface receptor retains the capacity to direct signal activation in response to relaxin stimulation. The dose-response curve of the RXFP1 Y9A receptor shows that the receptor is only capable of eliciting a response that is equivalent to 40% of the maximum RXFP1 receptor response to relaxin. This would support that the receptors are expressed on the cell surface in a mixed population of those with folded LDLa modules and those with unfolded LDLa modules. Results of the RXFP1 C5S/C18S receptor mutation show that the presence of the unfolded LDLa module does not effect receptor cell surface expression (Fig. 7) or binding of relaxin (supplemental Fig. S2). Therefore, in this situation where a population of receptors with folded and unfolded LDLa modules exists, those receptors with unfolded LDLa modules will bind relaxin but be unable to signal cAMP production, resulting in a lowered and right-shifted cAMP response. Chemical shift difference mapping indicates that for correctly folded modules the Y9A mutation does not introduce significant perturbations to the structure. The observation that the RXFP1 Y9A receptor retains 40% of the maximal cAMP response while having only 3.8% of the wild type RXFP1 receptor efficacy is further evidence that a specific interaction involving Tyr-9 in the population of receptors with correctly folded modules has been disrupted. In the RXFP1 receptors of most mammalian species, tyrosine is conserved in this position except for mouse and rat where a serine residue is observed (52). The presence of a serine residue at this position would reduce the size of the side chain but not the overall polarity.
In contrast to the RXFP1 Y9A receptor the dose-response curve of the RXFP1 L23A receptor reveals that it is capable of stimulating the production of cAMP to levels equivalent to wild type. From the RP-HPLC elution profile of the recombinant GB1-LDLa L23A mutant that demonstrates that the protein is unable to fold to a single isomer and the tendency of the protein to precipitate, we expected that the RXFP1 L23A receptor would have lowered signaling capacity. So it appears that despite this in vitro instability, RXFP1 L23A receptors at the cell surface express correctly folded LDLa modules. However, the dose-response curve is significantly right-shifted in comparison with the wild type receptor thus questioning whether the reduction in receptor efficacy is the consequence of the specific loss of the Leu-23 side chain or if mutation of Leu-23 has perturbed other residues. In the absence of further structural and folding studies, it is not possible to determine the conformational effects of this mutation on the LDLa module.
Phe-10 is conserved at the equivalent position in almost all LDLa modules and is absolutely conserved in all mammalian RXFP1 sequences, suggesting a crucial structural or functional role. Notably, the orientation of Phe-10 in the RXFP1 receptor differs from other modules studied previously. The significance of this may lie in the observation that in the presence of the F10A mutation the RXFP1 receptor is unable to elicit cAMP signaling. Although in vitro, the recombinant LDLa F10A module is able to refold to a conformation similar to the wild type protein, it loses the ability to fold only to a single correct isoform. Chemical shift mapping of the species that does fold to a globular structure indicates that in comparison with the wild type protein residues in the C-terminal region, primarily those of the calcium binding motif, are significantly perturbed by the mutation of Phe-10. Chemical shift perturbations are not simply because of loss of ring current effects. For example, calculation of ring current effects shows only small contributions for Asp-30 and Glu-37, which show large chemical shift changes in LDLa F10A compared with the wild type LDLa module. Asp-36 also shows significant chemical shift differences, but these may be due to the loss of ring current effects for the mutant. Asp-30, Asp-36, and Glu-37 are all calcium ligands; therefore, the mutation F10A may have perturbed the structure of this region. These differences suggest that Phe-10 plays a critical role in the arrangement of the residues that bind calcium and therefore the overall stability of the structure.
The RXFP1 and RXFP2 receptors are novel GPCRs, and their exclusive activation by the N-terminal LDLa modules represents a paradigm in understanding of GPCR activation. It also presents a unique opportunity to develop highly specific agonist/antagonists to the receptors by mimicking or blocking LDLa module function. Our studies have highlighted the key role that specific residues in the N-terminal region of the LDLa module play in receptor activation. Furthermore, we have demonstrated the necessity to study the structural effects of single amino acid mutations on the module to fully interpret their effects on receptor activity.