A Soluble Form of the First Extracellular Domain of Mouse Type 2 (cid:1) Corticotropin-releasing Factor Receptor Reveals Differential Ligand Specificity*

The heptahelical receptors for corticotropin-releasing factor (CRF), CRFR1 and CRFR2, display different specificities for CRF family ligands: CRF and urocortin I bind to CRFR1 with high affinity, whereas urocortin II and III bind to this receptor with very low affinities. In contrast, all the urocortins bind with high affinities, and CRF binds with lower affinity to CRFR2. The first extracellular domain (ECD1) of CRFR1 is important for ligand recognition. Here, we characterize a bacterially expressed soluble protein, ECD1-CRFR2 (cid:1) , corresponding to the ECD1 of mouse CRFR2 (cid:1) . The K i values for binding to ECD1-CRFR2 (cid:1) are: astressin (cid:2) 10.7 (5.4–21.1) n M , urocortin I (cid:2) 6.4 (4.7–8.7) n M , urocortin II (cid:2) 6.9 (5.8–8.3) n M , CRF (cid:2) 97 (22–430) n M , urocortin III (cid:2) sauvagine > 200 n M . These affinities are similar to those for binding to a chimeric receptor in which the ECD1 of CRFR2 (cid:1) replaces the

The actions of CRF ligands are initiated by binding to their receptors, whose activation results in an increase of intracellular cAMP, hydrolysis of phosphoinositol, activation of mitogen-activated protein (MAP) kinases (8,9), and other signaling pathways (10). In mammals, two receptor types, CRFR1 and CRFR2, have been cloned (11)(12)(13)(14)(15)(16)(17); orthologous receptors have also been identified in many other species including chicken (18), fish (19), and Xenopus (20). A third receptor, CRFR3, with a high level of sequence identity to CRFR1, has been cloned in catfish (19). CRF receptors have been characterized in the central nervous system and various peripheral sites including pituitary, gastrointestinal tract, epididymis, heart, gonad, adrenal, skin, and skeletal muscle.
The CRF receptors are 7-transmembrane domain proteins with relatively large first extracellular domains (ECD1s). Both CRFR1 and CRFR2 exist as multiple splice variants and belong to the type B receptor family that includes receptors for growth hormone-releasing factor, secretin, calcitonin, vasoactive intestinal peptide, glucagon, glucagon-like peptide (GLP), and parathyroid hormone (2).
The ligand specificities in binding to CRFR1 and CRFR2 are markedly different. Although both CRF and urocortin I bind with equally high affinities to CRFR1, the affinity of CRF for CRFR2 is at least 10 times lower than that of urocortin I. There is no high affinity interaction of either urocortin II or urocortin III with CRFR1, whereas their affinities for CRFR2 are in the subnanomolar range. The agonist, sauvagine, and the peptide antagonist, astressin, have equally high affinities for both types of receptors (5,6,21).
The majority of differences between the sequences of the two receptors are found in their ECD1s. Mutagenesis studies have identified regions of the receptors that are important for differential recognition of agonists and small molecule antagonists, as well as for governing the ligand selectivity of the two types of receptors (22)(23)(24)(25)(26)(27). Binding data from chimeric receptors in which the ECD1 of CRFR1 is annealed to transmembrane domains of other receptors suggested the importance of the ECD1 in CRF/receptor interactions (28,29). Further, a soluble protein corresponding to the ECD1 of CRFR1 expressed either in bacteria or mammalian cells binds astressin and urocortin with moderately high affinity (28,30).
Recently, 5 amino acids have been identified in the ECD1 of the type 2 CRF receptor that appear to govern the differences in ligand recognition of the Xenopus and human receptors (31). Further understanding of the role of the ECD1 of CRFR2 in the binding of the CRF ligands may be gained from an investiga-tion of their binding to the ECD1 isolated from the rest of the receptor.
In this study, we present biochemical, biophysical, and functional characterization of a soluble protein corresponding to the ECD1 of mouse (m)CRFR2␤. We find that the relative affinities of the CRF ligands for the soluble protein are similar to those for a chimeric receptor in which the ECD1 of CRFR2␤ is anchored in the plasma membrane by a single transmembrane domain of the type 1 activin receptor (32). Further, we present the disulfide arrangement of the ECD1 and show that it is identical to that of the ECD1 for CRFR1. Finally, we show that ligand binding induces a conformational change in the protein.

MATERIALS AND METHODS
ECD1-CRFR2␤ Expression in Escherichia coli-A cDNA encoding amino acids 27-134 of mCRFR2␤ was inserted into pET-32a(ϩ) (Novagen) with KpnI and XhoI (pET-ECD1-CRFR2␤), and the integrity of the construct was confirmed by automated sequencing. The E. coli strain, Origami trxB, gor (DE3)pLysS (Novagen) was chemically transformed according to manufacturer's directions using pET-ECD1-CRFR2␤. Bacteria were grown to an A 600 nm ϭ 0.6 -1.0 and induced with isopropyl-1-thio-␤-D-galactopyranoside (1 mM) for either 2.5 h at 37°C or 16 h at 22°C. The bacterial pellet was solubilized, sonicated in 10 mM Tris-HCl, pH 8, (10 ml/g of wet pellet), and centrifuged at 100,000 ϫ g for 30 min. The soluble ECD1-CRFR2␤, obtained as a thioredoxin fusion protein in the supernatant, was subjected to thrombin cleavage and enriched by affinity chromatography with S-protein-agarose (Novagen). After thrombin cleavage, ECD1-CRFR2␤ contains 24 additional amino acids at the N terminus derived, in part, from the S-(epitope) tag. The sequence of ECD1-CRFR2␤ is GSGMKETAAAKFERQHMDSPDLGT (mouse CRFR2␤(27-134)) (the S-tag is underlined); the additional amino acids, GS and DLGT, are part of the thrombin cleavage site and the KpnI cloning site, respectively.
Radioreceptor Assays-Transfections and binding to membrane fractions were performed as described (28). The data for the chimeric receptors are from crude membrane fractions of COS-M6 cells transiently expressing the receptors; the data for the wild-type receptor are from crude membrane fractions of Chinese hamster ovary cells stably expressing mCRFR2␤. For membrane binding, 5-10 g of membrane protein/50 l of assay buffer A (20 mM Na-HEPES, pH, 7.4, 0.1% bovine serum albumin, 10% sucrose, 2 mM EGTA) were incubated with increasing concentrations of unlabeled peptides (50 l of assay buffer) and [ 125 I-D-Tyr 0 ]astressin (ϳ200,000 cpm/50 l of assay buffer A ϩ 0.02% Triton) or [ 125 I-Tyr 0 ,Glu 1 ,Nle 17 ]sauvagine (ϳ100,000 cpm/50 l of assay buffer A) in a final volume of 200 l. For soluble protein binding, 0.2-0.5 g of ECD1-CRFR2␤/50 l of assay buffer B (20 mM Na-HEPES, pH, 7.6, 0.1% bovine serum albumin) were incubated with increasing concentrations of unlabeled peptides (50 l of assay buffer B) and [ 125 I-D-Tyr 0 ]astressin (ϳ200,000 cpm/50 l of assay buffer Bϩ 0.02% Triton) in a final volume of 200 l. Incubation proceeded for 90 min at room temperature in MAGV microtiter plates (Millipore) precoated with 0.1% polyethylenimine. The mixture was aspirated under vacuum, and the plates were washed twice with relevant assay buffer. The counts bound in the wells were quantified by ␥-counting. To obtain the K i values, the competitive displacement data were analyzed by a non-linear regression analysis (GraphPad Prism program (GraphPad Softwares, Inc., San Diego CA)). All assays were performed in triplicate at least three times.
Disulfide Arrangement-The ECD1-CRFR2␤ (10 g) was initially subjected to digestion by endoproteinase Asp-N (1 g) in 50 mM MES at pH 6.2 for 16 h. A large fragment containing all 6 cysteine residues was isolated after HPLC purification. This fragment was analyzed by chemical sequence analysis and yielded three sequences beginning with residues Glu 43 , Asp 65 , and Glu 105 . The endoproteinase Asp-N fragment was further subjected to digestion by chymotrypsin, and fragments were again isolated after HPLC separation. The HPLC fractions were analyzed by MALDI-MS and by chemical sequence analysis. Two chymotryptic fragments were identified that contained cysteine residues. Fragment CT-1 exhibited an intense signal at m/z ϭ 1655.71. Chemical sequence analysis yielded two major sequences corresponding to that of ECD1-CRFR2␤: 101-109 (containing Cys 103 ) and 60 -64 (containing Cys 60 ). The calculated monoisotopic mass (MH ϩ ) for this cystine-connected fragment is 1655.72 Da. The MALDI-MS spectrum of fragment CT-2 showed an intense signal at m/z ϭ 2868.33. Chemical sequence analysis yielded two major sequences corresponding to residues 80 -88 (containing Cys 84 ) and 116 -129 (containing Cys 118 ). The calculated monoisotopic mass (MH ϩ ) for the cystine-connected fragment is 2868. 35 Da. These fractions were reduced by treatment with tris(2-carboxyethyl)phosphine hydrochloride (TCEP) and reanalyzed by MALDI-MS. In both cases, the original signal diminished in intensity giving rise to signals corresponding to the individual cysteine-containing peptides.

RESULTS
In our previous work, we were able to express, in bacteria, milligram quantities of a functional soluble form of the ECD1 from CRFR1 (26). Further, we found that the amino acids corresponding to the putative signal peptide of the ECD1 were cleaved in the mature CRFR1, expressed in mammalian cells. We assume here that amino acids 1-26 of mCRFR2␤ encode its signal peptide and are absent in its mature protein. Accordingly, a cDNA encoding amino acids 27-134 of mCRFR2␤ was used to transform a strain of E. coli in which there is enhanced formation of disulfide bonds as a result of mutations in thioredoxin and glutathione reductase. The ECD1 was obtained as a soluble trx-fusion protein. Following thrombin cleavage, the soluble protein, ECD1-CRFR2␤, was obtained. The protein contains 132 amino acids and comprises 108 amino acids of the ECD1, 15 amino acids encoding the S-(epitope) tag, and other amino acids from the thrombin cleavage site and the KpnI cloning site. Analyses, by SDS-PAGE, of the fusion protein and its thrombin cleavage products indicate that the apparent molecular size of the fusion protein is ϳ29 kDa, the apparent molecular size of the trx tag is ϳ14 kDa, and the apparent molecular size of the ECD1-CRFR2␤ is ϳ15 kDa (data not shown). Mass-spectrometric analysis confirms the size of ECD1-CRFR2␤: measured m/z ϭ 14,944, calculated [MH] ϩ ϭ 14,950 Da.
An essential aspect of the biochemical characterization of ECD1-CRFR2␤ is an assessment of its affinity for the CRF family members. In Fig. 1A, we show the competitive displacements by astressin, CRF, urocortins I, II, III, or sauvagine of [ 125 I-D-Tyr 0 ]astressin bound to ECD1-CRFR2␤. For comparison, we also show the competitive displacements by the same ligands of [ 125 I-D-Tyr 0 ]astressin bound to the wild-type mCRFR2␤ stably expressed in CHO cells (Fig. 1B). The affinities of the agonists CRF, urocortin I, and urocortin II as well as of astressin for the soluble ECD1 are similar to their affinities for the complete receptor, whereas the affinities of urocortin III and sauvagine for ECD1-CRFR2␤ are much lower than their affinities for the wild-type mCRFR2␤.
To investigate the role of the ECD1 of mCRFR2␤ in the context of a membrane environment, we studied a chimeric receptor, ECD1-CRFR2␤/ALK4, in which the ECD1 of mCRFR2␤ replaced the ECD of the type 1B activin receptor (ALK4), a single transmembrane receptor (32). The competitive displacements of 125 I-[D-Tyr 0 ]astressin bound to this chimera by CRF family ligands are shown in Fig. 2. Again, the relative potencies of CRF, urocortin I, urocortin II, and astressin are similar to their potencies for the full-length receptor, whereas both urocortin III and sauvagine have very low potencies; the affinities of the CRF family ligands for this chimeric receptor are nearly the same as those for ECD1-CRFR2␤. Consistent with the lack of displacement of labeled astressin by nanomolar concentrations of unlabeled sauvagine is the absence of specific binding of [ 125 I-Tyr 0 ,Glu 1 ,Nle 17 ]sauvagine to either the ECD1-CRFR2␤ or the ECD1-CRFR2␤/ALK4. These data are summarized in the first four rows of Table I. For comparison, binding to a chimeric receptor in which the ECD1 of CRFR1 replaces the ECD of ALK4 is also given. The low affinity binding of urocortin II and urocortin III for the chimera expressing the ECD1 of CRFR1 is in accord with the low affinity binding of these ligands for the full-length type 1 CRF receptor.
Also, for comparison, we list in Table I (in the last row) the affinities derived from competitive displacement of labeled sauvagine bound to mCRFR2␤. It can be seen that the calculated affinities of some, but not all, of the analogs are dependent on the analog used as the radiolabel. When determined by displacement of labeled astressin, the K i of sauvagine is ϳ5 times greater, and that of urocortin III is ϳ10 times greater than their K i values determined from displacement of labeled sauvagine.
To further characterize the soluble protein at a molecular level, we determined the Cys-Cys connectivity pattern. The disulfide arrangement of the ECD1-CRFR2␤ was determined by employing a strategy similar to the one used previously for the ECD1-CRFR1 protein (26). The protein was digested by proteases that cleave between cysteine residues. The fragments were then isolated, analyzed by MALDI-MS, and sub-jected to chemical sequence analysis. Cystine-linked fragments exhibit a mass in MALDI-MS analysis corresponding to the sum of the cysteine-containing peptides. Chemical sequence analysis yields two sequences, again corresponding to the constituent peptides. Initial digestion with endoproteinase Asp-N was insufficient to cleave the cystine-containing core of the protein. Instead, three N termini were observed during sequence analysis.
The analysis of chymotryptic fragments allowed the determination of two of the disulfide bridges. The fragment CT-1 was found to consist of the peptides 101-109 and 60 -64 from mCRFR2␤. This established a connection between Cys 103 and Cys 60 . The other fragment, CT-2, consisted of peptides 80 -88 and 116 -129 and thereby established the connection between Cys 84 and Cys 118 . The remaining connection between Cys 45 and Cys 70 was established by sequence analysis of the original endoproteinase Asp-N digestion product. The presence of the three N termini at residues Glu 43 , Asp 65 , and Glu 105 in a single species can only be explained by a connection between cysteine residues 45 and 70. A schematic representation of the disulfide pattern is shown in Fig. 3. The overall pattern of cysteine connectivity is identical to that found in the ECD1 of CRFR1 (26) and also in the related parathyroid hormone receptor (35).
The behavior of ECD1-CRFR2␤, attending ligand interaction, was studied by circular dichroism. Difference CD spectra were collected in the far UV (185-260 nm) of ECD1-CRFR2␤ in the absence and presence of astressin. The CD spectra of ECD1-CRFR2␤ in low ionic strength, physiological pH aqueous solution, or more physiological ionic strength (0.15 M sodium chloride, not shown) are equivalent and are dominated by the random coil with additional notable contributions from the antiparallel sheet and turn conformations (Fig. 4). Deconvolution by the method of Bohm (34) resulted in contributions of 8.1, 16.3, 3.4, 24.8, and 46.8% for the helix, antiparallel, parallel, ␤-turn, and coil conformations, respectively. To determine the structural effects of ligand binding, ECD1-CRFR2␤ was incubated with equimolar astressin, and the spectrum of the resulting mixture was compared with the arithmetic sum of the spectra of the components Fig. 4. The resulting difference CD spectrum (Fig. 4, inset) clearly shows the formation of positive ellipticity with a peak near 192 nm and a slight increase in the magnitude of the negative ellipticity in the range 220 -225 nm. Deconvolution of the spectrum of the mixture gives conformational contributions of 11.4, 28.7, 3.4, 26.0, and 31.4% for the helix, antiparallel, parallel, ␤-turn, and coil conformations, respectively. Thus, the deconvolution and graphical data suggest that when combined with astressin, ECD1-CRFR2␤ loses significant coil contribution (46.8 -31.4%) with a concomitant increase in antiparallel sheet (16.3-28.7%) and to a much lesser extent helix (8.1-11.4%). Because the spectra shown in Fig. 4 were collected under equimolar conditions of protein and ligand, and because the residue ratio of ECD1-CRFR2␤ to

ECD1 of mCRFR2␤ Reveals Differential Ligand Specificity
astressin is 120:32, ϳ80% of the signal observed in Fig. 4 is contributed by ECD1-CRFR2␤. Thus, although undoubtedly a small fraction of the difference CD spectrum (Fig. 4, inset) arises from changes in the astressin conformation, the major CD changes observed must reflect changes in the conformation of ECD1-CRFR2␤. DISCUSSION The CRF ligand/receptor system consists of the related ligands CRF, (frog) sauvagine, (fish) urotensin, urocortin I, the recently discovered mammalian orthologs urocortin II and urocortin III (similar to stresscopin-related peptide and stresscopin, respectively), and the receptors, CRFR1 and CRFR2, derived from two distinct genes. Each receptor exists in at least two forms as a result of alternative splicing. The CRF receptors belong to the type B G-protein-coupled receptor family whose members are characterized by comparatively large ECD1s with 6 conserved cysteines.
Given the distinct pharmacology and physiological roles of the type 2 CRF receptor, we wished to explore the ligand binding characteristics of a soluble form of the ECD1 for mCRFR2␤. By means of competitive displacement assays using radiolabeled astressin, we find that the soluble protein, ECD1-CRFR2␤, binds astressin, urocortin I, and urocortin II with high affinities (6 -10 nM), binds CRF with moderate affinity (ϳ100 nM), and binds urocortin III and sauvagine with low affinities (Ͼ200 nM). Under the conditions of our assay, there is no binding of labeled sauvagine to ECD1-CRFR2␤, and displacement of bound, labeled astressin by urocortin III or unlabeled sauvagine is seen only at high concentrations. This absence of binding of sauvagine was seen also for the ECD1 of CRFR1 (26). In summary, the relative affinities for ECD1-CRFR2␤ are: astressin Х urocortin I Х urocortin IIϾCRFϾ Ͼ urocortin III Х sauvagine. These data are supported by complementary data on a chimeric receptor, ECD1-CRFR2␤/ALK4, in which the ECD1 of mCRFR2␤ is anchored to the plasma membrane by means of the single transmembrane domain of a type 1 activin receptor. The relative potencies of the ligands in binding to the chimera are similar to those for binding to the ECD1-CRFR2␤.
The observation that sauvagine does not displace labeled astressin bound to the soluble ECD1 of either the type 1 or type 2 receptor suggests that other regions of the complete receptor are necessary for binding of sauvagine. These data are consistent with those showing that the affinity of sauvagine was ϳ500 nM and that the affinity of CRF was Ͼ1000 nM in binding to a chimeric receptor in which the ECD1 of CRFR1 replaced the corresponding ECD1 of the parathyroid hormone receptor (29). In another study, it was demonstrated that sauvagine crosslinks to the second extracellular loop of CRFR1 (36).
The large decrease in affinity of urocortin III for the isolated ECD1 of CRFR2␤ suggests that, also for this ligand, other regions of the receptor are required for high affinity binding. Another explanation for these observations is that sauvagine and urocortin III require correct coupling of the receptor to a G-protein, whereas urocortin I and urocortin II are able to bind with high affinity to receptors in absence of G-protein coupling. These data are consistent with the lack of effect of GTP on binding of urocortin I (21,37) and with the suggestion that urocortin I may possess intrinsic antagonistic properties (38).
The data in Table I show that the apparent affinities of urocortin III and of sauvagine for wild-type mCRFR2␤ depend on the nature of the labeled analog that is used in the radioreceptor assay. For both peptides, when labeled astressin is used, the K i values are higher than those obtained by displacement of labeled sauvagine. The affinities of urocortin I and of urocortin II do not appear to depend on the nature of the radioligand. These observations are similar to those in another study (38) in which a similar reduction in the affinity of sauvagine but not in the affinity of urocortin I was seen in competitive displacement of labeled astressin, as compared with their affinities determined from displacement of labeled sauvagine.
The biochemical characterization revealed that the pattern of disulfide bonds in the ECD1 of CRFR2␤ is the same as that  determined for the ECD1 of CRFR1 (26), which in turn is the same as that of the ECD1 of the parathyroid hormone receptor (35). The observations of high affinity ligand binding and of a unique disulfide pattern with no scrambling suggest that the bacterially expressed protein assumes a conformation close to that of the native receptor. This adds further support to the suggestion that the 3 pairs of conserved cysteines in the ECD1 of this subgroup of receptors are arranged in a unique manner that may characterize the family of receptors. A similar kind of disulfide arrangement is found in a number of small bioactive peptides including cocaine-and amphetamine-regulated transcripts (CARTs) whose NMR structures have been analyzed (39). Despite the variations in the loop sizes, the basic fold is identical in all these structures. It is possible to speculate that the ECD1 of the CRF receptors assumes a similar folding pattern. It is interesting to note that the ECD of receptors for the TGF-␤ and related ligands assumes the cardiotoxin fold that had been associated previously with a family of small peptides (40).
The observation that the CD spectra of ECD1-CRFR2␤ at pH 7.5 in both low and physiological ionic strength media are primarily random coil (ϳ47% by deconvolution) but that this coil conformation is significantly reduced by the interaction with astressin suggests that ECD1-CRFR2␤ exists as a mobile structure capable of large shifts in conformation without irreversible denaturation at room temperature. This behavior has been documented in the interaction of astressin with the soluble ECD1 of CRFR1 (26).
Biochemical characterizations of soluble ECD1s include those for the G-protein-coupled receptors for follicle-stimulating hormone (41), luteinizing hormone/human chorionic gonadotropin (42), calcium (43), parathyroid hormone (35), GLP-1 (44), and glutamate (45). A conformational change, as determined from CD measurements, was found for the soluble ECD1 of the follicle-stimulating hormone receptor following binding of follicle-stimulating hormone (41). Data obtained from an analysis of the CD spectra of the soluble ECD1-luteinizing hormone receptor/human chorionic gonadotropin receptor-hormone complex showed that it was characterized by more than 25% ␣-helices (42), and similar data from the ECD1 of the parathyroid hormone receptor showed ϳ25% ␣-helical and 23% ␤-sheet secondary structures (35). Structural data from x-ray crystallography have been obtained, thus far, only for ECD1 of the metabotropic glutamate receptor, which has been characterized to 2.2-Å resolution (46).
In conclusion, we have found that a soluble protein, expressed in bacteria, corresponding to the first extracellular domain of the mCRFR2␤, binds with nanomolar affinity the CRF family members, urocortin I and urocortin II, as well as a synthetic peptide antagonist, astressin. The agonist, CRF, binds with moderately high affinity, whereas the agonists urocortin III and sauvagine display only low affinities for the soluble ECD1. The disulfide pattern of the ECD1 of CRFR2␤ is the same as that of the ECD1 of CRFR1 and suggests a common pattern for the receptor family. Further, there is a significant conformational change in the ECD1 following binding of astressin.
It is possible to speculate that the ligands of the CRF family have co-evolved with the receptors in such a manner that the contribution to high affinity ligand binding of the ECD1, relative to that of other receptor sites, is specific for each ligand. For example, the ECD1 is a major target of high affinity interaction for CRF, urocortin I, and urocortin II, whereas interactions with other domains of the receptor appear to be important for high affinity binding of urocortin III and sauvagine. A model describing the activation of CRF receptors was sug-gested by our previous study of the sempiternal signaling of a chimeric receptor in which the ECD1 of CRFR1 was replaced by the first 16 amino acids of CRF. Taken together, our data suggest that the ECD1 of the CRF receptors interacts initially with the C-terminal half of ligands such as CRF, urocortin I, and urocortin II, which are then positioned in the correct proximity to present the N-terminal portion of the ligand to the remainder of the receptor for subsequent activation. It should now be possible to refine these models with data from molecular structure determinations using soluble proteins such as the ECD1 of the CRF receptors.