A role for a helical connector between two receptor binding sites of a long-chain peptide hormone.

The conformational freedom of single-chain peptide hormones, such as the 41-amino acid hormone corticotropin releasing factor (CRF), is a major obstacle to the determination of their biologically relevant conformation, and thus hampers insights into the mechanism of ligand-receptor interaction. Since N- and C-terminal truncations of CRF lead to loss of biological activity, it has been thought that almost the entire peptide is essential for receptor activation. Here we show the existence of two segregated receptor binding sites at the N and C termini of CRF, connection of which is essential for receptor binding and activation. Connection of the two binding sites by highly flexible epsilon-aminocaproic acid residues resulted in CRF analogues that remained full, although weak agonists (EC(50): 100-300 nM) independent of linker length. Connection of the two sites by an appropriate helical peptide led to a very potent analogue, which adopted, in contrast to CRF itself, a stable, monomer conformation in aqueous solution. Analogues in which the two sites were connected by helical linkers of different lengths were potent agonists; their significantly different biopotencies (EC(50): 0.6-50 nM), however, suggest the relative orientation between the two binding sites rather than the maintenance of a distinct distance between them to be essential for a high potency.

The biologically important peptide hormones corticotropin releasing factor (CRF) 1 , glucagon, secretin, vasoactive intestinal polypeptide, growth hormone releasing factor (GRF), calcitonin, parathyroid hormone (PTH), calcitonin gene-related peptide (CGRP), etc. have significant features in common. All exert their activity via binding to and activation of class 2 G proteincoupled receptors (GPCRs). They are polypeptides comprising about 25-40 amino acid residues without preferred conformation in aqueous solution and exhibit no documented biologically relevant secondary and tertiary structure. Under structureinducing conditions (e.g. in the presence of trifluoroethanol or membrane mimicking lipids), however, these peptide hormones (CRF (Ref.  16)). Based on these results, it has been assumed that almost the entire peptide is necessary for binding to and activation of the corresponding receptors.
In the case of CRF, which is the principal neuroregulator of the basal and stress-induced secretion of ACTH, ␤-endorphin and other proopiomelanocortin-related peptides from the anterior pituitary (see Ref. 17 for review), previously published structure-activity relationship studies of single-point substituted (18 -20) and terminally truncated CRF analogues (7,21) showed that the N-terminal sequence (9 -19) represents a receptor binding site, since substitutions in this region resulted in a significant decrease in receptor binding. The most Nterminal amino acid residues are thought to be responsible for receptor activation (7), since truncation of these residues produced antagonists. The N-terminal peptide sequence (6 -20) is highly conserved within the CRF family, peptides from different species that activate CRF receptors. In contrast, there is great sequence diversity within the C-terminal region . Substitutions of Arg-35 or Leu-38 in oCRF by alanine (18), conversion of the C-terminal carboxamide to a carboxyl group or truncation of the C-terminal dipeptide from oCRF, however, reduced biopotency dramatically (14), indicating an essential binding site to be located at the extreme of the C terminus. The existence of two receptor binding sites in peptide ligands of class 2 GPCRs was also suggested by studies using chimeric receptors and peptide ligands, but nothing has been described concerning the structural organization of the ligands (22,23).
We have investigated whether segregated receptor binding sites in CRF do exist and, if so, what role the connector unit between the two binding sites might play. CRF analogues with highly flexible, structurally simplified as well as conformationally stabilized connector units between the two sites were investigated to address these questions.

EXPERIMENTAL PROCEDURES
Preparation of Peptides-Peptides were synthesized automatically (MilliGen 9050 peptide synthesizer) by the solid-phase method using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry in the continuous flow mode as described previously for the synthesis of CRF analogs (20). Purification was carried out by preparative HPLC to give final products of Ͼ95% purity by reverse phase-HPLC analysis. The peptides were characterized by mass spectrometry, which gave the expected [MϩH] ϩ mass peaks and correct amino acid analyses.
Peptide-induced Testosterone Production-Leydig cells were prepared from adult, male NMRI mice as described previously (24) and allowed to attach to well plates (100,000 cells). Medium was removed and replaced with 1 ml of fresh incubation medium containing the phosphodiesterase inhibitor IBMX (2.5 mM) and CRF or CRF analogues (0.01 nM to 1 M). Incubations were performed in a shaking water bath (35 rpm) at 37°C for 30 and 60 min. 100 l of the medium were frozen for the determination of testosterone by radioimmunoassay (DPC Biermann GmbH, Bad Nauheim, Germany).
CRF Receptor Assay-Whole brains of male Wistar rats (220 -250 g) were homogenized with a Teflon-glass homogenizer (10 strokes at 800 rpm) in 0.32 M sucrose, 50 mM Tris/HCl (pH 7.2), 10 mM MgCl 2 , 2 mM EGTA, and 0.15 mM bacitracin (1 ml/50 mg wet weight). After centrifugation at 1000 ϫ g for 5 min, the supernatant was centrifuged at 26,000 ϫ g for 20 min. The pellet was resuspended in 50 mM Tris/HCl (pH 7.2), 10 mM MgCl 2 , 2 mM EGTA, 0.15 mM bacitracin, and 0.0015% aprotinin (assay buffer) and again centrifuged. The resulting pellet was resuspended in assay buffer containing 0.32 M sucrose and stored at Ϫ20°C. All steps were carried out at 4°C. Protein concentrations were determined by the method of Bradford (25) using bovine serum albumin as standard.
100 g of membrane protein in 300 l of assay buffer were incubated in quadruplicate with 0.1 nM [ 125 I]Tyr(0)-oCRF in the absence and presence of 12 different concentrations (0.2 nM to 1 M) of unlabeled peptide at 25°C for 2 h. Nonspecific tracer binding was determined in the presence of 1 M oCRH. At the end of incubation, 3 ml of ice-cold wash buffer (assay buffer without inhibitors containing 0.01% Triton X-100) was added to the assay tube, and the samples were immediately filtered through GF/C filter discs (Whatman), presoaked for 2 h in 0.1% polyethyleneimine, using a Brandel Harvester, followed by washing of the incubation tubes and filters with 3 ml of cold wash buffer. Triton X-100 in this buffer strongly reduced the nonspecific tracer peptide binding. Radioactivity retained on the filter was measured by ␥-counting.
Receptor affinities (K a , K d ϭ 1/K a ) and capacities (B max ) were estimated using the non-linear least squares curve fitting program RA-DLIG (Biosoft, Cambridge, United Kingdom) and a K d of 0.48 nM for the binding of the tracer peptide as determined from tracer saturation assays. The amount of total bound tracer was 5%, of which about 30% was nonspecific (26).
ACTH Releasing Activity of Peptides on Rat Anterior Pituitary Cells-Pituitary cells were obtained by enzymatic digestion of the anterior pituitary of male Wistar rats weighing 220 -250 g following the procedure by Denef (27). 200,000 cells in Dulbecco's modified Eagle's medium and 0.25% bovine serum albumin per well were seeded in cell culture plates and maintained at 37°C under 5% CO 2 , 95% air for 3 days. The culture medium was replaced by 0.5 ml of fresh medium and after 2 h by culture medium containing one of the peptides to be studied at different concentrations. After a stimulation period of 3 h, the medium samples were harvested and stored at Ϫ70°C. ACTH in the samples was determined by immunoradiometric assay (HS-ACTH-IRMA from the Nichols Institute Diagnostika GmbH, Bad Nauheim, Germany) using hACTH as standard. This assay uses two antibodies directed against the N-and C-terminal sequences of ACTH, which are identical in human and rat. EC 50 values were calculated from the dose-response curves by a four-parameter logistic curve-fitting program.
CD Spectroscopy-CD measurements were carried out on a Jasco 720 spectrometer from 185 to 260 nm.  (28). For determination of peptide concentration quantitative amino acid analysis was used. CD spectra were recorded of samples dissolved in water (pH 3.4 -3.6) and in 10 mM phosphate buffer (pH 7.1).
Two-dimensional 1 H NMR Spectroscopy and Molecular Modeling-NMR measurements were made on a 1 mM protein sample in 90% H 2 O, 10% D 2 O at pH 3.6. All NMR experiments were performed on a Bruker DRX600 spectrometer, operating at 600.13 MHz, at a temperature of 283 K. The three two-dimensional NOESY spectra were recorded with 1-K increments in the t 1 dimension, 64 scans, mixing times of 40, 100, and 200 ms, a relaxation delay of 1.7 s, a spectral width of 16.66 ppm, and 8000 data points in t 2 . The two-dimensional TOCSY (DIPSI spinlock sequence) was acquired with 1-K increments in the t 1 dimension, 32 scans, a spinlock time of 94.3 ms, a relaxation delay of 1.7 s, a spectral width of 16.66 ppm, and 8000 data points in t 2 . The twodimensional double quantum-filtered (DQF)-COSY spectrum was recorded with 1-K increments in the t 1 dimension, 40 scans, a relaxation delay of 1.7 s, a spectral width of 16.66 ppm, and 8000 data points in t 2 . Water suppression was achieved by WATERGATE gradients (NOESY, TOCSY), and presaturation during relaxation delay (DQF-COSY), respectively. Prior to Fourier transformation, the time-domain data were zero-filled to a final data matrix of 8000 ϫ 2000 multiplied by a shifted squared sine bell function.
Distance restraints for the structure calculation were collected from two-dimensional NOESY experiments and converted into distances. Because of peak overlapping, distances were not partitioned into categories. Structures were calculated using standard DG and SA protocols implemented in the program X-PLOR 3.81 (29) applying a large distance range (2-5 Å).

RESULTS
Two CRF receptor subtypes (CRFR-1 and CRFR-2) have been identified in vertebrates; CRFR-1, in contrast to CRFR-2, appears non-selective for human/rat CRF (h/rCRF), ovine CRF (oCRF), and the structurally related CRF analogues, rat urocortin (Uct), carp urotensin, and frog sauvagine. All these peptides stimulate ACTH release in an in vitro pituitary cell assay with similar potency (30). Analogous results were described for CRF-stimulated testosterone production via CRFR-1 from mouse Leydig cells (24) (Table I), which was used as the preferred biological assay in this work.
Structural Simplification of CRF-Intramolecular interactions between the N and C termini of long-chain peptide ligands, e.g. CRF (31), have been suggested as stabilizing a Ref.

Ovine CRF (oCRF) S Q E P P I S L D L T F H L L R E V L E M T K A D Q L A Q Q A H S N R K L L D I A-NH 2
4.82 (24) Carp urotensin

Role of Connector between Receptor Binding Sites
biologically active conformation. Therefore, amino acid variations within the N and C termini (non-conserved amino acid residues) within the CRF family may function interdependently in such interactions. In order to address the question, we synthesized chimeric peptides, combining the N and C termini of oCRF, urotensine, and sauvagine. The chimeric peptides exhibited a high biological potency (Table I), showing that these amino acid variations are, in fact, not interdependent with respect to stabilization of the biologically active conformation.
Assuming an ␣-helical conformation of CRF to be advantageous for receptor interaction, substitution of those amino acid residues of CRF that are not individually essential for receptor interaction, especially the non-conserved residues within the CRF family, by others with a high helical propensity, such as alanine, should be possible without loss of biopotency. Thus, while retaining arginine-35 and the hydrophobic residues at positions 36/37/38 (leucine), the remaining amino acid residues within the C-terminal portion (residues 22-41) of h/rCRF were replaced by alanine or glutamine residues. This modification yielded an analogue of high biological potency (Table II), demonstrating that the amino acid residues of the middle portion (residues 22-33) of CRF are not individually essential for receptor interaction. From these results, the question arose as to whether this middle portion is essential at all for receptor activation. Combination of the urocortin N terminus (residues 1-19) via Ile-Glu and a highly flexible ⑀-aminocaproic acid (acp) residue with a mixed C-terminal site (residues 34 -41) from members of the CRF family resulted in an analogue that exhibited full receptor activation (full intrinsic activity), but only at increased concentration (reduced biopotency) (Fig. 1). Surprisingly, connection of the N-terminal to the C-terminal site via 1, 2, 3, or 4 acp residues produced only a slight difference in biopotency (Table II), showing that the length of the flexible connectors has little effect on agonistic potency. Direct connection of the two sites to yield Uct (1-19)-Ile-Glu-Gln- (34 -41), again resulted in a full agonist (Table II). Replacement of the C-terminal site (residues 34 -40) by another acp residue, retaining only the C-terminal Val-amide, led, however, to a total loss of intrinsic activity (Table II). Furthermore, the two peptides representing the receptor binding sites, Uct (1-19)-Ile-Glu-OH and the C terminus (34 -41), either alone or as an equimolar mixture, revealed no intrinsic activity (Table II), showing connection of the two sites to be essential for potency.
Potentially ␣-Helical Connector Units between the Two Receptor Binding Sites-Circular dichroism (CD) spectroscopy studies of the potent polyalanine-substituted h/rCRF agonist in aqueous solution revealed strong helix induction by alanine incorporation, indicating that the N-and C-terminal binding sites of CRF might be advantageously connected by an ␣-helical structure. The helical content of the analogue was found to be concentration-dependent, i.e. the formation of secondary structure is at least partially caused by association (data not shown). In order to prevent such association, we created an analogue in which the two binding sites were connected by a highly charged peptide consisting exclusively of glutamic acid and lysine residues arranged such that helix stabilization could occur by salt bridge formation (32) between side chains at positions i and iϩ4 (EKEEKEKKRKE). The resulting analogue, urocortin-EK (UEK), was found to be at least as potent as the most active member of the CRF family, urocortin (Table III). The result was confirmed in an in vitro pituitary cell assay for peptide induced ACTH release (EC 50 for Uct, 0.06 nM, and for UEK, 0.04 nM). CD studies revealed an ␣-helical content for UEK of 45% at 4°C in aqueous solution, independent of concentration (1 M to 1 mM), indicating the acquisition of a monomer conformation. The existence of a monomer solution in water at a concentration of 1 mM was confirmed by light-scattering studies (data not shown). To demonstrate that the contribution of individual amino acid residues of the EK-connector is not re-   sponsible for the high biopotency of UEK, we substituted the EK-connector (residues [22][23][24][25][26][27][28][29][30][31] in UEK by the alanine-rich sequence RAAAQAAKKA. The corresponding analogue was at least as potent as UEK itself (Table III).

Poly-alanine-substituted h/rCRF S Q E P P I S L D L T F H L L R E V L E M A R A A A Q A A Q A A A N R L L L A A
In order to investigate whether the length of the connector unit (22)(23)(24)(25)(26)(27)(28)(29)(30)(31) in UEK plays a specific role with respect to biopotency, we synthesized the corresponding deletion analogues.
Stepwise shortening of the connector unit of UEK by deletion of one (des (23) (Table III). The deletion of five amino acid residues led to a drastic decrease in biopotency of about 2 orders of magnitude, but the deletion of seven residues resulted in a relatively minor loss (Table III). Moreover, a corresponding deletion of three amino acid residues led again to a high biopotency comparable to that of UEK itself. Furthermore, elongation of the EK connector in UEK by seven amino residues (UEK-t5) resulted in a significant decrease in biopotency (Table  III), possibly due to there being too great a distance between the two sites. Thus, in contrast to the acp analogues, which showed a similar biopotency while bearing different numbers of acp residues in the linker between the two receptor binding sites, the variation of the connector length in UEK strongly affected biopotency. CD measurements of the analogues (50 M) in aqueous buffer solution (pH 7.1) at room temperature revealed that the acp analogues possess no preferred conformation (Fig. 2), while the shortened UEK-analogues revealed a significant helical conformation (des(23-25): 36%, des(23-26): 23%, des (23)(24)(25)(26)(27): 38%, des(23-28): 21%, des(23-29): 36%). After deletion of 14 amino acid residues of the central part of UEK, des(19 -32)-UEK, no preferred conformation was observed, in contrast to UEK itself which revealed under these conditions an ␣-helical content of 39% compared with 34% at pH 3.4. Thus, linkers that consisted of ␣-amino acid residues as in UEK and its centrally shortened analogues induced and/or stabilized an ␣-helical ligand conformation whereby the different lengths of the linkers, in contrast to highly flexible acp connector units, showed a strong effect on receptor activation. To investigate the effect of single amino acid deletions in the suggested connector unit of the wild-type ligand oCRF on biopotency, we synthesized des-Thr(22)-oCRF and des-Gln(29)-oCRF. Indeed,

RAAAQAAKKA (22-31)-UEK D D P P L S I D L T F H L L R T L L E I E R A A A Q A A K K
both analogues exhibited reduced biopotencies (des-Thr(22)-oCRF (EC 50 : 146 nM), des-Gln(29)-oCRF (EC 50 : 57 nM)) compared with oCRF (EC 50 : 3.82 nM) as determined by the testosterone assay, a result which is similar to that obtained for corresponding deletion analogues of UEK. Assuming a two-site receptor binding mode in the case of UEK and oCRF, corresponding amino acid deletions in the suggested binding sites should strongly affect receptor interaction. Consistently, the truncation of UEK by the two C-terminal amino acid residues, resulting in UEK (1-38)-amide, led to a drastic decrease in biopotency (Table III), showing the Cterminal binding site to be much more sensitive than the connector unit toward deletion/truncation with respect to biopotency. Shortened analogues of the wild-type ligand oCRF des-Asp(39)-oCRF (EC 50 : Ͼ 1 M) or des-Thr(11)-oCRF and des-Glu (17)-oCRF (both were found to be inactive at a peptide concentration of 1 M)) showed again strongly reduced biopotencies deleting single amino acid residues of the binding sites as determined by the testosterone assay.
In order to strengthen the indication that an ␣-helical linker connects the two receptor binding sites in the biologically very potent UEK, we investigated the structure of UEK in aqueous solution by 1 H NMR spectroscopy.
Solution Structure of UEK Determined by Two-dimensional NMR Spectroscopy and Molecular Modeling-The 1 H NMR assignments of UEK were made using standard procedures (33) and the main chain-directed approach (34) on the basis of double-quantum filtered DQF-COSY, TOCSY, and NOESY spectra at 5°C in 90% H 2 O, 10% D 2 O. In the amide region of the two-dimensional NOESY spectrum shown in Fig. 3A, nearly all sequential NH(i)-NH(iϩ1) cross-peaks between Thr-10 and Lys-30 were resolved. Extensive unambiguous medium-range NOEs ␣N(i,iϩ3) and ␣␤(i,iϩ3) NOEs, characteristic of an ␣-helical conformation, were observed for UEK 10 -30 and confirmed helix formation, at least from Thr-10 to Lys-30. Besides NOEs, the chemical shifts for C␣-protons (35) are commonly used for secondary structure assignments in peptides. In particular, the chemical shifts of C␣ protons within ␣-helices or ␤-sheets tend to exhibit an upfield or downfield shift respectively, relative to the chemical shift values typical for a random-coil conformation. The plot of the C␣-proton chemical shift deviations versus the position along the sequence of UEK is shown in Fig. 4. Applying the criteria for secondary structure, the differences between observed and random-coil C␣ proton chemical shifts from Thr-10 to Lys-35 show negative values, consistent with an ␣-helical arrangement. In addition to the helical domain 10 -30, a limited number of long-range NOEs suggests a preferred loop conformation for the C terminus. Particularly those NOEs between the ␦-NH of Asn-33 and and the ␣-, ␤-, and ␥-protons of Val-40 indicated the close proximity of both amino acid residues. Furthermore, NOEs between the NH of Asp-8 and the ␣H of Ser-6 as well as the NH of Asp-8 and the ␣and ␤-protons of Leu-5 suggested a kinked structural arrangement of the N terminus.
A three-dimensional structure of UEK was calculated following the molecular dynamic protocol described under "Experimental Procedures" using a final set of 268 NOE-derived distance constraints (113 intra, 93 sequential, 59 medium-range, and 3 long-range Ͼ 5 residues) at 5°C. Because of peak overlapping, NOE intensities were not classified. After minimization 10 structures, based on low residual distance violations and low dihedral angle violations, were selected and used to compute the solution structure of UEK. Fig. 3B displays the 10 final structures superimposed for the minimum backbone deviations between residues 10 and 30, 1 and 9, and 31 and 40. Whereas an ␣-helix is well defined in the central part of UEK between residues Thr-10 and Lys 30, the N and C termini themselves exhibit some evidence of local order. The central helix comprises the N-terminal binding site (Fig. 3C, violet) and an ␣-helical connector (Fig. 3C, green) to the C-terminal binding site, i.e. an ␣-helix of about 3 turns forms the linker between the two sites. DISCUSSION The structural requirements of linear, long-chain peptide ligands for receptor interaction are more poorly understood than those of small molecules because of their numerous potential receptor interaction sites and great conformational freedom, which disfavors a stable conformation in water, a general obstacle to the determination of their biologically relevant conformation. The fact that N-and C-terminal truncations of various long-chain peptide ligands of class 2 GPCRs lead to a loss of biopotency had suggested that almost the entire peptide is essential for receptor activation.
We demonstrated here that stepwise structural simplification of a peptide ligand of this receptor class is a powerful approach to gain insight into the structural requirements of a peptide ligand for its receptor interaction. Multiple substitution of amino acid residues by alanine indicated that the amino acid residues of the middle portion (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33) of CRF are individually not essential for receptor activation. Indeed, deletion of the middle portion led to a full agonist at nanomolar concentrations (acp0, Fig. 5). Substitution of this middle portion by 1, 2, 3, or 4 acp residues resulted in analogues which exhibited a similar biopotency (Table II, Fig. 5). Therefore, in the case of flexible acp linkers the linker length between the two receptor binding sites had only little effect on receptor activation. In contrast, truncation of the C-terminal residues (34 -40) in the acp4 analogue led to a total loss of intrinsic activity, consistent with the existence of a second segregated receptor binding site in the peptide C terminus. Furthermore, the peptides that represent the two receptor binding sites, either alone or as an equimolar mixture, exhibited no receptor activation, demonstrating that connection of the two is essential for biopotency.
Because of the high helix forming propensity of alanine, the fact that an alanine-rich connector unit between the two binding sites led to an analogue showing a significantly increased biopotency compared with the corresponding acp analogue suggested an ␣-helical connector to be more appropriate for receptor interaction with the two sites. Alanine-rich peptides tend, however, to self-associate which hampers determination of their monomer solution conformation. On the contrary, the EK model peptide has been shown to be well suited not only to incorporate a stable ␣-helix in the peptide, but also to prevent self-association at millimolar concentrations, as observed with the native peptide ligands of CRF receptors. Therefore, incorporation of the highly charged EK helix allowed determination of the monomer structure of a biologically very active analogue (UEK) in water by means of NMR analysis. Although the solution structure of UEK was not totally resolved, the results clearly show the existence of a central ␣-helix (10 -30) comprising the amino acid residues of the N-terminal receptor binding site and the connector unit to the C-terminal site. Incorporation of the stable EK helix led to an exact distance between both binding sites in the ligand and, thereby, a high biopotency, in contrast to the acp analogues with highly flexible linkers, which exhibited a drastic loss of biopotency compared with that of UEK. On the other hand, this shows that an appropriate distance between both binding sites in UEK, although advantageous, is not essential for receptor binding and activation. Deletions of an increasing number of amino acid residues from the connector helix in UEK led to an interesting pattern with respect to biopotency of the analogues. While the deletions resulted generally in a decrease in biopotency compared with that of UEK, a relatively minor decrease was observed by deletion of either three or seven amino acid residues. There seems to be no correlation between biopotencies of the deletion analogues and net charge of linker units or position of individual amino acid residues in the connector to explain the differ-  (Table III). Assuming flexible connector units between the two receptor binding sites in the deletion analogues of UEK, they should exhibit a similar biopotency as in the case of the acp analogues. However, CD investigation of the deletion analogues of UEK revealed, in contrast to the acp analogues, a significant ␣-helical content when the middle part (20 -32) was not completely deleted, suggesting ␣-helical connector units in the UEK analogues. In this case, the number of amino acid residues in the connector helix will determine not only the distance between the two receptor binding sites but also their relative orientation (Fig. 5). Since the stepwise shortening of the UEK connector helix did not directly correspond with biopotency of the analogues, the relative orientation of the two binding sites rather than the maintenance of a distinct distance between them seems to be essential for a high potency. A high biopotency of analogues bearing a helical connector may also be explained by induction and/or stabilization of an appropriate conformation of one or both receptor binding sites. The fact, however, that des(23-27)-UEK revealed an ␣-helical content, which was at least comparable to that of UEK, but UEK is 80-fold more potent than des(23-27)-UEK, suggests that this conformational stabilization of the binding sites may play a minor role.
Comparing the very potent agonists UEK, des(23-25)-UEK and des(23-29)-UEK (Fig. 5), the corresponding distances between the two sites may be very different. Regarding receptor activation, biopotencies decreased in the order UEK (EC 50 : 0.6 nM) Ͼ des(23-25)-UEK (0.9 nM) Ͼ des(23-29)-UEK (4.6 nM). With respect to receptor binding, the same order was observed (UEK (K d : 0.4 nM) Ͼ des(23-25)-UEK (4.1 nM) Ͼ des(23-29)-UEK (8.1 nM), indicating that the appropriate distance of the two sites in UEK is responsible for high receptor affinity and, thereby, for high biological potency. Furthermore, an analogue in which the two sites are connected directly via Ile-Glu-Gln (acp0) still exhibited full intrinsic potency (Fig. 5). Assuming that the analogues bearing the identical receptor binding sites bind to the same sites of the receptor, it appears that the distance between the hormone binding sites of the receptor may vary remarkably (by approximately 7 Å/helix turn). This large difference requires a remarkable flexibility of the receptor domains for ligand binding, which may be explained if one binding site of the ligand binds to the receptor N terminus and the other site to the remaining C-terminal portion of the receptor. For chimeric constructs of CRF and GRF receptors, it was shown that the extracellular domains, particularly the N terminus and the third loop of the CRF receptor (CRFR-1), are responsible for high affinity for urocortin (36).
In the case of the members of the CRF family, which do not adopt a preferred conformation in water, helix formation may result from interaction with the receptor or its surroundings.  Under structure-inducing conditions, CRF was shown to adopt a largely ␣-helical conformation, particularly in the middle part of the molecule. NMR spectroscopic investigations confirmed that oCRF and h/rCRF comprise a well defined ␣-helix between residues 6 -36 in a mixed solvent system (TFE/H 2 O) (1,37). Both peptides bind to detergent micelles associated with an increase in the ␣-helical content (37) and, on interacting with hydrophobic interfaces, oCRF revealed an ␣-helical conformation comprising residues 6 -32 (38). Furthermore, connector units between the two receptor binding sites consisting of amino acid residues with a high helical propensity resulted in a very potent agonist (Table II), and it was recently reported that incorporation of two adjacent D-amino acid residues, which causes a significant destabilization of ␣-helices (39), in the connector region of oCRF led to a drastic loss of biopotency (24). Moreover, single-point amino acid deletions in the corresponding connector unit of oCRF at positions 22 or 29 led to a decrease in biopotency which is comparable to the effect of corresponding deletions in UEK, while amino acid deletions in the suggested receptor binding domains of oCRF resulted in almost complete loss of biopotency. Therefore, it seems very likely that our results for UEK may be relevant also in the case of the wild-type ligands of the CRF receptor.
Similarly to UEK, the only peptide ligand of class 2 GPCRs that has a preferred structure in water, the 35-amino acid peptide helodermin possesses a stable central ␣-helix (9 -23) (40). NMR studies of peptide hormones of this receptor class suggested N-and C-terminal helices to be separated by a highly flexible or "U-shaped" region (3,4,6). However, in the case of PTH-related protein, it was shown that incorporation of an amphipathic helix at the peptide C terminus resulted in a stabilization of a central helix comprising the postulated Ushaped region, and provided a very potent agonist (41,42). Therefore, a helical domain may commonly function as a defined linker between receptor binding sites at the N and C termini of such peptide ligands. Proof of the existence of two segregated receptor binding sites in a long-chain peptide ligand of class 2 GPCRs, which can be linked by different connector units to give biologically fully active analogues, opens a way for the development of novel, shortened agonists for this class of hormones which will also stimulate studies on ligand-receptor interactions.