- Top of page
- Materials and methods
Rat corticotropin-releasing factor receptor 1 (rCRFR1) was produced either in transfected HEK 293 cells as a complex glycosylated protein or in the presence of the mannosidase I inhibitor kifunensine as a high mannose glycosylated protein. The altered glycosylation did not influence the biological function of rCRFR1 as demonstrated by competitive binding of rat urocortin (rUcn) or human/rat corticotropin-releasing factor (h/rCRF) and agonist-induced cAMP accumulation. The low production rate of the N-terminal domain of rCRFR1 (rCRFR1-NT) by transfected HEK 293 cells, was increased by a factor of 100 in the presence of kifunensine. The product, rCRFR1-NT-Kif, bound rUcn specifically (KD = 27 nM) and astressin (KI = 60 nM). This affinity was 10-fold lower than the affinity of full length rCRFR1. However, it was sufficiently high for rCRFR1-NT-Kif to serve as a model for the N-terminal domain of rCRFR1. With protein fragmentation, Edman degradation, and mass spectrometric analysis, evidence was found for the signal peptide cleavage site C-terminally to Thr23 and three disulfide bridges between precursor residues 30 and 54, 44 and 87, and 68 and 102. Of all putative N-glycosylation sites in positions 32, 38, 45, 78, 90, and 98, all Asn residues except for Asn32 were glycosylated to a significant extent. No O-glycosylation was observed.
Corticotropin-releasing factor (CRF), a peptide 41 amino acids long (Spiess et al. 1981), is released from the hypothalamus into the hypophyseal portal system and stimulates ACTH secretion from the pituitary (Vale et al. 1981) as an endocrine response to stress. In addition to CRF, the CRF-like 40-amino acid peptide urocortin (Ucn) has been characterized (Vaughan et al. 1995; Donaldson et al. 1996). CRF and Ucn are distributed widely throughout the CNS of rodents and humans (Eckart et al. 1999), where they modulate various central functions such as locomotor activity, food intake, anxiety, and learning (Eckart et al. 1999; Radulovic et al. 1999). Furthermore, pathophysiological changes in the CRF system have been associated with several neuropsychiatric disorders such as major depression, panic disorder, anorexia nervosa, and Alzheimer's disease (Behan et al. 1996).
CRF and Ucn exert their biological actions by binding to two CRF receptor (CRFR) subtypes, CRFR1 and CRFR2. CRF receptors belong to the class of G protein-coupled receptors (GPCR) which possess four extracellular, four intracellular, and seven transmembrane domains (Radulovic et al. 1999). They are coupled to G proteins mainly stimulating the production of cAMP as second messenger. In different species, the CRFR1 precursor consists of 415 to 420 amino acids (Chang et al. 1993; Chen et al. 1993; Perrin et al. 1993; Vita et al. 1993; Yu et al. 1996; Dautzenberg et al. 1997; Myers et al. 1998; Palchaudhuri et al. 1998) and is expressed mainly in the brain and pituitary (Potter et al. 1994). Several splice variants of CRFR2 have been ideied: CRFR2α, CRFR2β, and CRFR2γ. They consist, depending upon the species, of 410–413 (α), 431–438 (β), and 397 (γ) amino acids (Kishimoto et al. 1995; Lovenberg et al. 1995; Perrin et al. 1995; Stenzel et al. 1995; Liaw et al. 1996; Dautzenberg et al. 1997; Kostich et al. 1998; Palchaudhuri et al. 1999). CRFR2 is found in discrete regions of the brain and peripheral organs (Chalmers et al. 1995; Stenzel et al. 1995).
Two independent studies indicate that the N-terminal domain of CRFR1 is essential for ligand recognition. Dautzenberg et al. (1998) made use of the unusual binding properties of Xenopus leavis CRFR1 (xCRFR1) which binds ovine CRF (oCRF) and the amphibian CRF analog sauvagine (Svg) (Montecucchi and Henschen, 1981) with significantly lower affinity than hCRFR1 (Dautzenberg et al. 1997). In experiments with chimeric receptors of xCRFR1 and hCRFR1, it was shown that the N-terminal domain (NT) of xCRFR1 is responsible for the ligand selectivity of xCRFR1 (Dautzenberg et al. 1998). Perrin et al. (1998) constructed a chimeric receptor composed of the N-terminal part of rCRFR1-NT connected to the transmembrane and intracellular domains of the activin II B receptor (Perrin et al. 1998). This chimeric receptor bound rat Ucn (rUcn) and astressin (Ast), a peptidic CRFR antagonist (Gulyas et al. 1995). In the same study, it was observed that chimeras composed of rCRFR1 and the GPCR rat growth hormone-releasing factor receptor, which contained the N-terminal domain of rCRFR1, bound Ucn and Ast with high affinity. Therefore, it was suggested that only the N-terminal domain of rCRFR1 was required for high affinity binding of Ucn and Ast.
It is known that the extracellular cysteines of CRFR1 are critical for binding of CRF (Qi et al. 1997). Chemical reduction of the disulfide bonds of mouse CRFR1 (mCRFR1) decreased the specific binding of h/rCRF significantly (Qi et al. 1997). Additionally, several single and paired mutations of cysteine residues to serine or alanine were introduced and the biological activity of the mutated receptors was analyzed. On the basis of these data, a pattern of disulfide linkages was proposed (Qi et al. 1997).
The objective of this study was to develop a model of rCRFR1. Therefore, the N-terminal domain of rCRFR1 (rCRFR1-NT) was produced as a soluble protein in human embryonic kidney (HEK) 293 cells transfected with cDNA coding for rCRFR1-NT. The production of biologically functional full length rCRFR1 in these cells has been demonstrated (Dautzenberg et al. 1998). The yield of rCRFR1-NT produced by the transfected HEK 293 cells was increased significantly by the mannosidase I inhibitor kifunensine, which prevented formation of complex carbohydrate moieties. The suitability of the resulting high mannose glycosylated rCRFR1-NT (rCRFR1-NT-Kif) as a model for rCRFR1 was demonstrated by specific binding of rUcn and Ast. Furthermore, the role of the glycosylation type for high affinity binding and receptor functionality was investigated by two differently glycosylated forms of rCRFR1 produced in the presence or absence of kifunensine. We have used mass spectrometry coupled on-line to RP-HPLC for the analysis of the N-terminal processing sites, the disulfide linkages, and the glycosylation pattern of the purified protein rCRFR1-NT-Kif. Furthermore, the secondary structure domains of rCRFR1-NT were proposed by a prediction method.
- Top of page
- Materials and methods
Mannosidase I of HEK 293 cells was inhibited by kifunensine (Elbein et al. 1990) to prevent the formation of hybrid or complex oligosaccharide structures. Deglycosylation experiments revealed a complex glycosylation type for rCRFR1 and, as expected, a high mannose glycosylation type for rCRFR1-Kif. The binding data obtained with the SPA for rCRFR1 agreed with earlier observations for rCRFR1 (Perrin et al. 1993,, 1998,, 1999). Both receptors bound rUcn, h/rCRF, and Ast with high affinity, demonstrating that kifunensine treatment did not prevent the correct folding of the receptor and that the investigated glycosylation types of rCRFR1 did not influence the binding of the tested ligands. Furthermore, the presented cAMP data indicated that rCRFR1-Kif was fully functional. Alteration of the glycosylation type impaired neither the targeting of the receptor to the cell surface nor the intracellular coupling to G-proteins. Thus, the kifunensine treatment did not prevent the correct insertion of the receptor into the membranes.
Recently, it was shown that the molecular size of native CRFR1 varies not only between mouse and rat brain, but also between different brain regions (Radulovic et al. 1998). These differences are probably caused by alterations in the glycosylation of CRFR1. The presented binding and cAMP data for rCRFR1 and rCRFR1-Kif suggested that the different glycosylation of CRFR1 did not influence the binding affinities or the coupling to adenylate cyclase in vivo.
Although rCRFR1-NT was detected as a soluble glycoprotein in the medium, the extremely low production level prevented the pharmacological and protein chemical characterization of rCRFR1-NT. The addition of kifunensine to the serum-free medium increased the rCRFR1-NT yield by approximately two orders of magnitude and changed the glycosylation type of this protein. This increased yield may be a result of impaired cytosolic proteasomal degradation of rCRFR1-NT-Kif. Recently, the fate of terminally misfolded α1-antitrypsin was studied in the presence of different glycosidase inhibitors (Liu et al. 1999). It was demonstrated that inhibition of mannosidase I prolonged the retention phase of misfolded α1-antitrypsin in the endoplasmic reticulum and impaired proteasomal degradation, but did not affect the secretion of misfolded α1-antitrypsin. In analogy, it can be speculated that kifunensine extended the retention phase of rCRFR1-NT in the endoplasmatic reticulum and thus enhanced the folding process to generate a CRFR1-like spatial structure that might be more resistant to proteasomal degradation.
The binding constants obtained for rCRFR1-NT-Kif were probably similar to those of rCRFR1-NT in view of the observation that for the full length receptor the glycosylation type altered by kifunensine did not change the binding affinities to rUcn, h/rCRF, and Ast significantly. rCRFR1-NT-Kif bound rUcn and Ast specifically with relatively high affinity, whereas the CRFR2-selective antagonist antisauvagine-30 (Rühmann et al. 1998) did not compete with radiolabeled rUcn. These findings showed that the membrane interaction of the full length receptor was not required for specific interactions of rUcn and Ast with the soluble N-terminal domain of rCRFR1.
The observation that rCRFR1-NT-Kif did not bind radiolabeled h/rCRF, in contrast to rUcn and Ast, indicated that h/rCRF required more than the N-terminal domain of rCRFR1 for specific binding. Thus, CRF in comparison to Ucn and Ast interacted in a different manner with the full length receptor. This observation was supported by the finding that binding of Ucn and Ast was independent of the G protein-coupling state of CRFR1, whereas binding of h/rCRF and oCRF was impaired by uncoupling of CRFR1 from G proteins (Spiess et al. 1998; Perrin et al. 1999). In agreement with this observation, the importance of the fourth extracellular domain (EC4) for binding of oCRF to rCRFR1 was demonstrated recently (Sydow et al. 1999). The specific binding of rUcn and Ast to rCRFR1-NT-Kif indicated that this soluble protein was a valuable model for the corresponding domain of the full length receptor.
rCRFR1-NT and rCRFR1-NT-Kif were found to have identical start sequences, demonstrating that the kifunensine treatment did not influence the signal peptide processing in HEK 293 cells. The major form starting with Ser24 and the minor form starting with Leu25 were predicted with the highest probability using an established algorithm for the ideication of signal peptides and their cleavage sites (Nielsen et al. 1997). In view of these results, it was suggested that both isolated forms of rCRFR1-NT and rCRFR1-NT-Kif which started either with Ser24 or Leu25 were products of the precursor protein cleaved by signal peptidase which removed the first 23 or 24 amino acids. Alternatively, the possibility has to be considered that the smaller species was derived from the larger species by action of an amino peptidase. By using the above algorithm, we found for human, mouse, and sheep CRFR1 the same signal peptides of 23 and 24 amino acids as most probable. It is proposed that the full length rCRFR1 which was expressed in HEK 293 cells was N-terminally processed in a similar manner as rCRFR1-NT-Kif.
Disulfide bridges are important determinants for protein conformations by stabilizing tertiary structures. Since we could demonstrate that rCRFR1-NT-Kif interacted specifically with rUcn and Ast, it was concluded that it probably possessed the tertiary structure of the respective domain of the full length receptor. Therefore, the disulfide linkages were established using protein chemical methods. It was demonstrated that rCRFR1-NT-Kif did not contain free cysteine residues. Thus, the six cysteine residues of rCRFR1-NT-Kif formed three disulfide bonds. It was concluded that neither of the cysteine residues Cys188 and Cys258 of rCRFR1 located in the extracellular domains 2 (EC2) and 3 (EC3), respectively, formed a disulfide bond with a Cys residue in the N-terminal domain. In almost all known GPCRs, the two Cys residues located in EC2 and EC3 are highly conserved. It has been proposed that they form a disulfide bridge stabilizing the tertiary structure of the receptor (Strader et al. 1994). This proposal agrees with our findings. Site-directed mutagenesis was performed on several GPCRs (Karnik and Khorana 1990; Savarese et al. 1992; Ohyama et al. 1995; Perlman et al. 1995; Cook and Eidne 1997) including the secretin receptor (Vilardaga et al. 1997) and suggestive evidence was found for the linkage between these two conserved Cys residues. This disulfide bond is also proposed for mCRFR1 (Qi et al. 1997).
Two disulfide bridges connecting residue Cys44 with Cys102 and residue Cys68 with Cys87 of the mCRFR1 precursor protein were proposed on the basis of mutations of single and paired Cys residues to Ser residues (Qi et al. 1997). In addition, it was found that mutating residue Cys30 did not affect the function of mCRFR1. These results contrasted our finding for rCRFR1 showing three disulfide bridges connecting residues Cys30 and Cys54, Cys44 and Cys87, and Cys68 and Cys102 of rCRFR1. However, site-directed mutagenesis provides only indirect evidence for protein structure. In addition to local changes, point mutations may influence the protein structure even in remote regions. In contrast, the disulfide structure of the functional rCRFR1-NT-Kif was elucidated by analyzing the protein structure. The disulfide bridge arrangement determined for rCRFR1-NT-Kif may represent the pattern of disulfide linkages of the full length CRFR1. CRFR belongs to the secretin-like GPCR family which is characterized by at least five conserved Cys residues in the N-terminal domain of its members. rCRFR1 contains an additional Cys residue located N-terminally to the conserved Cys residues. It is conceivable that these residues form a pattern of disulfide bridges which is also conserved within this receptor family. Thus, the receptors of the secretin-like GPCR family may contain two of the three disulfide linkages shown for rCRFR1-NT-Kif.
It is noteworthy that Cys30, which is missing in rCRFR2α but not in rCRFR2β, was located in the predicted α-helical domain of rCRFR1. Therefore, it is concluded that the tertiary structure of the N-terminal domain of rCRFRs is stable without the formation of a disulfide linkage of the Cys residue located in the α-helical part. This conclusion agrees with the site-directed mutagenesis of the first Cys residue of mCRFR1, which led to a functional receptor (Qi et al. 1997).
Asn98, which is part of the most C-terminally-located glycosylation site, was almost fully glycosylated, whereas Asn90 was glycosylated to an extent of only 70%. This lower glycosylation may be explained by the neighborhood of Trp93 to the glycosylation site Asn90-Gly91-Ser92. It has been demonstrated that tryptophan residues following glycosylation sequons impair the glycosylation efficiency (Mellquist et al. 1998), probably due to poor accessibility of these residues to oligosaccharyl transferase. It was probable that truncation of rCRFR1 did not influence the degree of glycosylation in view of the observation that the more terminally located residue Asn98 was almost fully glycosylated. The first potential glycosylation site (Asn32) was barely glycosylated. This site is located in the predicted α-helical structure. The remaining glycosylation sites are located in regions where no specific secondary structure was found by the prediction method used. Only CRFR1 of the rat contains in this position a potential glycosylation m. Therefore, we propose that the glycosylation of rCRFR1 in position 32 is not important for ligand binding and receptor function. Since it has been reported that kifunensine does not influence protein glycosylation even at concentrations leading to full inhibition of mannosidase I (Elbein et al. 1990), we concluded that glycosylation of rCRFR1-NT-Kif was not affected by kifunensine and thus probably resembled the glycosylation of full length CRFR1.