Abbreviations used : cAMP, cyclic AMP ; CRF, corticotropin-releasing factor ; CRF-R, corticotropin-releasing factor receptor ; CRF-R1 and CRF-R2, corticotropin-releasing factor receptor type 1 and 2, respectively ; EC, extracellular domain ; GPCR, G protein-coupled receptor ; hCRF-R1 and hCRF-R2, human corticotropin-releasing factor receptor type 1 and 2, respectively ; h/rCRF, human/rat corticotropin-releasing factor ; hVIP-R1 and hVIP-R2, human vasoactive intestinal polypeptide receptor type 1 and 2, respectively ; rCRF-R1, rat corticotropin-releasing factor receptor type 1 ; rUcn, rat urocortin ; TM, transmembrane domain ; Ucn, urocortin ; VIP, vasoactive intestinal polypeptide ; xCRF-R1, Xenopus corticotropin-releasing factor receptor type 1.
Address correspondence and reprint requests to Dr. F. M. Dautzenberg at Pharma Division, Preclinical Research, Hoffmann-La Roche, Ltd., Grenzacher Strasse, CH-4070 Basel, Switzerland.
Abstract : The aim of the present study was to identify the N-terminal regions of human corticotropin-releasing factor (CRF) receptor type 1 (hCRF-R1) that are crucial for ligand binding. Mutant receptors were constructed by replacing specific residues in hCRF-R1 with amino acids from the corresponding position in the N-terminal region of the human vasoactive intestinal peptide receptor type 2 (hVIP-R2). In cyclic AMP stimulation and CRF binding assays, it was established that two regions within the N-terminal domain were crucial for the binding of CRF receptor agonists and antagonists : one region mapping to amino acids 43-50 and a second amino acid sequence extending from position 76 to 84 of hCRF-R1. Recently, it was found that the latter sequence plays a very important role in determining the high ligand selectivity of the Xenopus CRF-R1 (xCRF-R1). Replacement of amino acids 76-84 of hCRF-R1 with residues from the same segment of the hVIP-R2 N terminus markedly reduced the binding affinity of CRF ligands. Mutation of Arg76 or Asn81 but not Gly83 of hCRF-R1 to the corresponding amino acids of xCRF-R1 or hVIP-R2 resulted in 100-1,000-fold lower affinities for human/rat CRF, rat urocortin, and astressin. These data underline the importance of the N-terminal domain of CRF-R1 in high-affinity ligand binding.
The ligand-selective domains of human CRF-R2 (hCRF-R2), a receptor that preferentially binds the CRF analogues sauvagine, urotensin I, and urocortin (Ucn), have been mapped to EC2 and EC3 but not to EC1 (Liaw et al., 1997 a,b). However, as previously shown, the ligand-selective domain of Xenopus CRF-R1 (xCRF-R1), a receptor with remarkable substrate specificity, resides in its EC1 domain (Dautzenberg et al., 1998). The ligand-selective region was located between residues 68 and 92 of xCRF-R1, a region that is highly conserved among all CRF-Rs known to date. Mutation of five amino acids only within this region to the sequence of human CRF-R1 (hCRF-R1) completely abolished the ligand selectivity of xCRF-R1, whereas the introduction of these five residues into hCRF-R1 created a receptor mutant with almost identical substrate specificity as xCRF-R1 (Dautzenberg et al., 1998). These data are in agreement with a finding by Perrin et al. (1998), who mapped the major ligand binding determinant of rat CRF-R1 (rCRF-R1) to its EC1 domain. Based on the construction of a chimeric receptor with EC1 of rat growth factor-releasing factor receptor substituted by EC1 of rCRF-R1, it could be shown that the resulting receptor chimera bound the potent CRF antagonist astressin (Gulyas et al., 1995) and the CRF agonist rat Ucn (rUcn) (for references, see Vaughan et al., 1995) with almost identical affinities as rCRF-R1.
The purpose of the present study was to determine whether the amino acids involved in the ligand selectivity of xCRF-R1 were also critical for ligand binding of hCRF-R1. In addition, the hypothesis that other regions in the EC1 domain of hCRF-R1 were required to maintain the integrity of the ligand binding pocket of the CRF-R was investigated. Accordingly, a series of hCRF-R1 mutants with single or multiple residues in EC1 being replaced by the corresponding residues of the human VIP receptor type 2 (hVIP-R2) (Svoboda et al., 1994) was constructed to characterize the binding site of hCRF-R1. The sequence of hVIP-R2 was chosen to replace amino acids of hCRF-R1 because both receptors belong to the same subfamily of GPCRs and it was shown that hCRF-R1 does not bind VIP (Chen et al., 1993). In addition, it was reported that EC1 of human VIP receptor type 1 (hVIP-R1) and hVIP-R2 is important for high-affinity VIP binding. Thus, it can be expected that different amino acids are crucial for the formation of the binding pockets in both receptors. In addition to the hCRF-R1 mutants carrying residues from hVIP-R2, two receptor mutants with one (residue 76) or two (residues 76 and 81) amino acids from xCRF-R1 were analyzed. Here, we report on the identification of two regions and two specific residues within EC1 that are part of the binding pocket of hCRF-R1.
Materials, peptides, and reagents
All cell culture reagents were purchased from GibcoBRL, and aprotinin was obtained from Boehringer Mannheim. Human/rat CRF (h/rCRF), rUcn, and astressin were from Bachem (Bubendorf, Switzerland). The purity was >98%.
Construction of mutated or deleted receptors and nomenclature
Mutant receptors (multiple amino acid substitutions) were constructed from hCRF-R1 cDNA and inserted in the pcDNA3 vector (Invitrogen, San Diego, CA, U.S.A.) with the ExSite kit (Stratagene, La Jolla, CA, U.S.A.). Single and double amino acid exchanges or deletions were accomplished with the Quick-Change kit (Stratagene) according to the manufacturer's instructions. The sequences of the mutant oligonucleotides used in this study are given in Table 1. Sequences obtained by PCR and synthetic oligonucleotides were verified by DNA sequencing using an ABI 373 DNA sequencer (Applied Biosystems, Weiterstadt, Germany) ; the GCG software package (Madison, WI, U.S.A.) was used for analysis. In mutated receptors like hCl(R56W57) the origin of the receptor (namely, hCRF-R1) is indicated by hCl. Residues that were mutated to either the sequence of hVIP-R2 or xCRF-R1 are listed in parentheses, and the location is indicated by the number.
Table 1. Oligonucleotides used for mutagenesis of hCRF-R1Mismatches to the hCRF-R1 cDNA sequence (EMBL accession no. L23332 ; for references, see Chen et al., 1993) are underlined.
cDNAs of hCRF-R1 or mutant receptors, inserted into pcDNA3, were transfected into HEK293 cells with the calcium phosphate coprecipitation method (Sambrook et al., 1989) as described (Dautzenberg et al., 1997). Two days after transfection with 10 μg of plasmid DNA (Chen and Okayama, 1987), geneticin (600 μg/ml) selection was initiated.
Radioreceptor binding assays
Scatchard analysis using 0.1 nM125I-Tyr0-h/rCRF (NEN, Bad Homburg, Germany) was performed with 30-50 μg of membrane proteins from permanently transfected HEK293 cells expressing hCRF-R1 or mutant receptors as described previously (Dautzenberg et al., 1997, 1998). The Bmax values for all mutant receptors when tested with the three CRF analogues did not differ significantly from each other (1.1-3.0 pmol/mg of membrane protein). The dissociation constant, Ki, was calculated by the LIGAND program (Munson and Rodbard, 1980).
Cyclic AMP (cAMP) assay
Stably transfected HEK293 cells were plated at 105 cells per well into 24-well dishes. Stimulation for 30 min (37°C, 5% CO2) with increasing concentrations of peptides in stimulation buffer (2 mM 3-isobutyl-1-methylxanthine and 1 mg/ml bovine serum albumin in Dulbecco's modified Eagle's medium) and intracellular cAMP assays were performed as described (Dautzenberg et al., 1997). Statistical analysis was performed by two-way ANOVA. Significant differences between groups were determined by post hoc analysis using Dunnett's test.
Several mutated receptors with deletions in the EC1 domain were analyzed. Only the hCl Δ(H29-L42) mutant retained the ability to bind h/rCRF, rUcn, or astressin with affinities that were comparable to the native hCRF-R1 (Table 2). As hCRF-R1 mutants with deletions in other areas of EC1 lost their ability to bind CRF analogues normally (data not shown), it was concluded that these EC1 regions are necessary for proper folding of the receptor into the conformation required for ligand recognition.
Table 2. Binding of h/rCRF, rUcn, and astressin to membranes of HEK293 cells transfected with cDNAs coding for different hCRF-R1 mutantsData are mean ± SE values of three independent binding studies. Statistically significant differences are indicated :
Site-directed mutagenesis was used to pinpoint the specific regions and residues in EC1 that are critical for ligand binding of hCRF-R1 (Fig. 1). Mutation of specific residues C-terminal to Ala89 did not significantly influence ligand binding (data not shown). Replacement of residues 43-50 in the hCl (A43S45G46V47W48N50) mutant by the corresponding amino acids of hVIP-R2 markedly decreased the binding affinities for h/rCRF (Ki = 104 ± 12 nM), rUcn (Ki = 194 ± 20 nM), or astressin (Ki = 126 ± 21 nM) (Fig. 2 and Table 2). When the neighboring amino acids at positions 56 and 57 in the receptor mutant hCl(R56P57) or residues 58-66 in the receptor mutant hCl (A58N59V60E62T63T65V66) were replaced by the corresponding amino acids of hVIP-R2, ligand binding selectivity was changed only to a minor extent. hCl (R56P57) bound h/rCRF, rUcn, and astressin with an affinity that was comparable to hCRF-R1 (Table 2). The Ki values for the binding of h/rCRF, rUcn, and astressin to the hCl (A58N59V60E62T63T65V66) receptor were 14.1 ± 3.1, 12.3 ± 4.7, and 16.9 ± 5.1 nM, respectively, which is comparable to the binding of these CRF ligands to hCRF-R1. Replacement of residues 70-75 in hCRF-R1 by the amino acids in the same N-terminal region of hVIP-R2 to produce hCl (K70V71S73N74F75) had no effect on the binding of h/rCRF (Ki = 3.1 ± 0.4 nM) or rUcn (Ki = 2.4 ± 0.6 nM). In contrast, hCl (K70V71S73N74F75) bound astressin with an affinity (Ki = 1.8 ± 0.2 nM) that was even higher than that exhibited by hCRF-R1 (Ki = 13.1 ± 3.1 nM ; see Table 2). The same higher affinity with astressin (Ki = 1.9 ± 0.8 nM) was found for the hCl Δ(H29-L42) mutant. The mutation of residues 78-84 to form hCl-(S78K79A80I83S84) strongly affected binding of h/rCRF, rUcn, and astressin (Fig. 2 and Table 2). A dramatic loss (up to 100-fold) in affinity for h/rCRF (Ki = 184 ± 27 nM) was measured for hCl (S78K79A80I83S84). A similar change in CRF ligand binding affinities resulted from the mutation of Arg76 and Asn81 in the proposed CRF binding pocket. In the amphibian CRF-R1, Gln76 and Gly81 are crucial for ligand selectivity of xCRF-R1 (Dautzenberg et al., 1998). Like xCRF-R1, hVIP-R2 encodes Gly at position 81 rather than the Asn81 of hCRF-R1. No amino acid is encoded by hVIP-R2 in the position equivalent to Arg76 of hCRF-R1 or Gln76 of xCRF-R1, when the sequences of hCRF-R1 and hVIP-R2 were aligned (Fig. 1). If Gln76 was substituted for Arg76 to form hCl-(Q76) or Arg76 was deleted to synthesize hCl (ΔR76), CRF binding was strongly impaired (Fig. 2 and Table 2). Membranes prepared from HEK293 cells stably transfected with hCl (Q76) cDNA bound h/rCRF (Ki = 102 ± 11 nM), rUcn (Ki = 109 ± 19 nM), or astressin (Ki = 121 ± 10 nM) with lower affinities than the native hCRF-R1, whereas membranes from HEK cells expressing hCl (ΔR76) were incapable of binding these CRF ligands (Table 2). Mutation of Asn81 to Gly81 shifted the affinity of the resulting receptor mutant hCl (G81) significantly rightward (Table 2). However, the disruption of CRF ligand binding was not as great as in hCl (Q76) and hCl (ΔR76). When Arg76 was deleted and Asn81 was replaced by Gly81, which is present at the corresponding positions in hVIP-R2, CRF ligand binding at the receptor mutant hCl(ΔR76G81) was completely abolished (Table 2). The receptor mutant hCl(Q76G81), however, encoding the equivalent residues of xCRF-R1 bound h/rCRF, rUcn, and astressin with high affinity (Fig. 2 and Table 2).
Because ligand selectivity of xCRF-R1 was determined by three additional residues, Val83, His88, and Leu89 (Gly83, Leu88, and Ala89 in hCRF-R1), two additional receptor mutants were constructed. In the first mutant, hCl (I83), the Gly83 of hCRF-R1 was replaced by the corresponding amino acid of hVIP-R2. In the second mutant, hCl (N86T88S89), three residues of hCRF-R1 were replaced by the corresponding amino acids of hVIP-R2. hCl(I83) bound h/rCRF (Ki = 0.5 ± 0.1 nM), rUcn (Ki = 0.9 ± 0.4 nM), and astressin (Ki = 3.1 ± 1.9 nM) with affinities that were higher than their Ki values at hCRF-R1 (Table 2). At hCl (N86T88S89), the binding affinities of these three CRF ligands were decreased. However, this rightward shift was not significant (Table 2).
To determine whether the binding affinity changes were also accompanied by altered cellular signaling, the stimulation of cAMP accumulation by CRF agonists in HEK293 cells stably transfected with cDNAs encoding the various CRF-R1 mutants was investigated. The finding that intracellular cAMP accumulation was stimulated by low concentrations (<4 nM) of h/rCRF or rUcn in HEK cells expressing hCRF-R1, hCl Δ(H29-L42), hCl(R56P57), hCl (A58N59V60E62T63T65V66), hCl(K70V71S73N74F75), hCl(I83), hCl(Q76G81), and hCl(N86T88S89) supports the results of the binding experiments (Fig. 3 and Table 3). Intracellular cAMP accumulation in HEK293 cells stably transfected with several mutants, however, was only achieved at high (~100-200 nM) or very high (~1-10 μM) h/rCRF or rUcn concentrations (Fig. 3 and Table 3). In HEK cells transfected with the CRF-R1 mutant hCl(A43S45G46V47W48N50), the EC50 values for the stimulation of cAMP accumulation were >100-fold higher for h/rCRF (EC50 = 93 ± 14 nM) or rUcn (EC50 = 205 ± 26 nM) than in cells producing hCRF-R1. A similar impairment (50-200-fold higher EC50 values) was observed in CRF agonist-stimulated cAMP production in HEK293 cells stably transfected with hCl(S78K79A80I83S84) cDNA (Fig. 3 and Table 3).
Table 3. Stimulation of cAMP production in HEK293 cells, stably transfected with cDNAs coding for hCRF-R1 or different mutant receptors by h/rCRF and rUcnData are mean ± SE values of three stimulations.Statistically significant differences are indicated :
In HEK293 cells expressing hCl(G81), intracellular cAMP accumulation could only be stimulated by h/rCRF or rUcn concentrations >10-7M (Fig. 3 and Table 3). In HEK293 cells transfected with hCl(Q76) cDNA, even higher agonist concentrations were needed to stimulate cAMP production. In these cells, h/rCRF (EC50 = 897 ± 82 nM) or rUcn (EC50 = 717 ± 44 nM) only stimulated cAMP production at concentrations close to 1 μM. The hCl(▵R76) and hCl(▵R76G81) mutants only responded to micromolar concentrations of both peptides (Table 2).
Using deletion analysis, it was found that CRF ligand binding affinity and signaling were normal in hCRF-R1 mutants truncated at residues 24-42. Because amino acids 1-23 represent the signal peptide of the hCRF-R1 protein (authors' unpublished data), this N-terminal domain was not deleted. In contrast to the deletion of residues Ser24-Leu42, truncation of residues C-terminal to Leu42 resulted in a loss of binding affinity for CRF agonists and antagonists as well as in a loss in the capability of CRF agonists to activate the cAMP cascade in HEK293 cells stably transfected with these hCRF-R1 deletion mutants (data not shown). The observed affinity loss most likely resulted from the destruction of tertiary structures of the hCRF-R1 protein. It has been shown that mutagenesis of several extracellular cysteines in mouse CRF-R1 prevented a proper folding of the receptor protein, thereby resulting in the elimination of ligand binding (Qi et al., 1997). The susceptibility of hCRF-R1 for deletions has further been demonstrated by the isolation of a hCRF-R1 deletion mutant (Ross et al., 1994) lacking 40 amino acids (residues 41-80). This variant of hCRF-R1 did not bind CRF analogues, and cAMP accumulation in COS-1 cells stably transfected with the corresponding cDNA was only activated at high CRF (>300 nM) concentrations.
Based on the experiments presented here, two regions that are crucial for the integrity of the CRF ligand pocket were delineated by mutational analyses. The first region was mapped to residues 43-50 in EC1 of hCRF-R1. When six amino acids were mutated in this region of hCRF-R1 to form hCl(A43S45G46V47W48N50), the binding of CRF ligands was significantly impaired, whereas mutation of the neighboring residues was without effect. CRF ligand binding and second messenger signaling of the deletion mutant hClΔ(H29-L42) and the receptor mutants hCl(R56P57) and hCl (A58N59V60E62T63T65V66) were similar to those of the native hCRF-R1. Residues 51-55 of hCRF-R1 (Ile51Gly52Thr53Cys54Trp55) were not mutated because four of them are conserved between hCRF-R1 and hVIP-R2. Only the glycine residue in position 52 of hCRF-R1 is not present in hVIP-R2. However, it cannot fully be excluded that the deletion of Gly52 has an effect on CRF binding. Thus, it was concluded that CRF ligands bind to the region between Gln43 and Trp55 in EC1 of hCRF-R1.
In general, it was observed that hCRF-R1 mutations frequently altered the binding of the CRF agonists h/rCRF and rUcn to a greater extent than the binding to astressin. This observation may be explained by the approximately eightfold lower affinity of astressin for hCRF-R1 compared with h/rCRF or rUcn. Binding analyses indicated a two-site model for agonist binding with 30-40% uncoupled receptors. Astressin, however, was bound by the high-affinity receptors with a one-site model, indicating that the peptide did not discriminate between G protein-coupled and uncoupled forms of the receptor molecules. This finding was in agreement with data reported earlier (Maguire et al., 1976). Therefore, the reduced effect of astressin binding in comparison with binding of h/rCRF and rUcn as assessed with several receptor mutants could be due to a smaller fraction of uncoupled receptors.
The second region that was involved in the formation of the CRF binding pocket was located between residues Arg76 and Tyr84 of hCRF-R1. Gly83, however, did not appear to be important because replacement of this amino acid with either Ile83 of hVIP-R2 or Val83 of xCRF-R1 (data not shown) had no effect on the binding of h/rCRF, rUcn, or astressin. When single or multiple amino acids in this region other than Gly83 were mutated or Arg76 was deleted, CRF ligand binding was substantially impaired or completely abolished. The strongest effects on CRF binding and signal transduction were observed when Arg76 and Gln81 in hCRF-R1 were replaced by the amino acids in positions 76 and 81 of either hVIP-R2 or xCRF-R1, both of which encode a glycine residue at position 81. At position 76, the Arg of hCRF-R1 becomes a glutamine residue in xCRF-R1. No equivalent amino acid occurs in the sequence of hVIP-R2 when its sequence is aligned with hCRF-R1. Recently, residues 76, 81, and 83 were shown to be critical for the remarkable ligand selectivity of xCRF-R1 (Dautzenberg et al., 1998). Therefore, it was surprising that the mutation of Arg76 or Gly81 had such a strong effect on CRF-R1 binding affinity. However, if these two amino acids were replaced by the corresponding amino acids of xCRF-R1, high-affinity binding of several CRF ligands was completely rescued (Dautzenberg et al., 1998 ; present study). Thus, it seems likely that Arg76 and Gly81 are both crucial for the integrity of the CRF ligand binding pocket of CRF-R1. Because hVIP-R2 did not encode an amino acid equivalent to Arg76 of hCRF-R1 or Gln76 of xCRF-R1, no compensation effect was observed with the hCl(ΔR76G81) receptor mutant. The importance of amino acids 76-84 in the formation of the binding site of hCRF-R1 was further confirmed by the effects of mutations located N- or C-terminally to this region. No affinity loss in the case of the hCl(K70V71S73N74F75) or only a small affinity shift in the case of the hCl(N86T88S89) mutant was observed. A contribution of Arg85 to the formation of the ligand pocket of hCRF-R1 cannot be ruled out because it was not mutated in the present study. However, as hVIP-R2 encodes a basic residue (Lys91) at the position equivalent to Arg85, it seems unlikely that a mutation to this lysine residue would strongly impair CRF binding.
It is interesting that mutation of residues 43-50 impaired binding and signal transduction of rUcn to a greater extent than h/rCRF. Conversely, mutation of residues 78-84 resulted in a CRF receptor that preferred rUcn over h/rCRF. These mutated CRF receptors also bound astressin with affinities comparable to those of h/rCRF. Based on these results, it was suggested that the point of contact of h/rCRF and astressin with the receptor molecule may differ from the point where rUcn binds.
It is important to note that the N-terminal regions that have been identified in the present study to be critical for CRF ligand binding are located in the segment of EC1 that displays the highest homology with hVIP-R2. Within this region, 14 of 40 residues are conserved among hCRF-R1, hCRF-R2, and hVIP-R2. In contrast, only 13 and 10% homology, respectively, was found for residues 1-38 and 85-123 in hCRF-R1 and hVIP-R2. Consequently, it was concluded that the N-terminal regions from amino acids 43-50 and 76-84 of hCRF-R1 contains a sequence motif that is highly conserved within the subfamily of GPCRs that are activated by large peptides [the class II GPCR family (for references, see Horn et al., 1998)] such as the receptors for CRF, VIP (Ishihara et al., 1992 ; Svoboda et al., 1994), secretin (Ishihara et al., 1991), or growth factor-releasing factor (Mayo, 1992). Therefore, it is tempting to speculate that this structural motif may be important for ligand binding of all members of this GPCR subfamily. Indeed, for hVIP-R1 (Couvineau et al., 1995 ; Gaudin et al., 1995) it has been reported that this domain and residues therein are important for high-affinity VIP binding. However, adjacent residues within EC1 as well as other EC domains of hVIP-R1 and hVIP-R2 have been shown to be important for binding of VIP (Nicole et al., 1998) or VIP analogues (Couvineau et al., 1996). Furthermore, binding of secretin to its receptor, another member of the class II GPCR family, requires specific residues within EC2 and EC3 in addition to EC1 (Holtmann et al., 1995, 1996). These results are consistent with data on the ligand selectivity of hCRF-R2 (Liaw et al., 1997a,b) and indicate that the entire binding pocket of many class II GPCRs is built up by regions within several EC domains.
In conclusion, we have identified two regions within the EC1 domain of hCRF-R1 that are crucial for high-affinity CRF binding and give insights in receptor-ligand interactions of CRF-Rs.