Assignment of disulfide bonds in corticotropin-releasing factor-binding protein.

We have previously isolated, cloned, and characterized a protein that specifically binds and inactivates the peptide corticotropin-releasing factor. The integrity of the disulfide bonds in the binding protein is essential for this activity as reduction abolishes the protein's ability to bind corticotropin-releasing factor. The disulfide arrangement of the 10 cysteines present in the mature protein was established by analysis of proteolytically cleaved protein and sequence analysis of cystine containing fragments. A pattern is observed where each cysteine is connected to the next one in a sequential manner. Inspection of the genomic DNA encoding for this protein reveals that four of the domains defined by disulfide linkage coincide with four different exons.

W e have previously isolated, cloned, and characterized a protein that specifically binds and inactivates the peptide corticotropin-releasing factor. The integrity of the disulfide bonds in the binding protein is essential for this activity as reduction abolishes the protein's ability to bind corticotropin-releasing factor. The disulfide arrangement of the 10 cysteines present in the mature protein was established by analysis of proteolytically cleaved protein and sequence analysis of cystine containing fragments. A pattern is observed where each cysteine is connected to the next one in a sequential manner. Inspection of the genomic DNA encoding for this protein reveals that four of the domains defined by disulfide linkage coincide with four different exons.
Corticotropin-releasing factor-binding protein (CRF-BPI' was initially purified from human serum (Behan et al., 1989) and subsequently cloned from human liver and rat brain cDNA libraries (Potter et al., 1992). Both purified and expressed CRF-BPS bind to CRF with high affinity and neutralize the peptide's biological activity. In human beings, the only species thus far shown to express CRF-BP in the liver, circulating CRF-BP has been proposed to protect the maternal endocrine system from the high levels of CRF (Campbell et al., 1987;Golandet al., 1986;Linton et al., 1987;Sasaki et al., 1984) produced by the placenta (Petraglia et al., 1987) during pregnancy. Rats, primates, and other species examined express CRF-BP in the brain and pituitary gland, where it may serve to modulate the neuroendocrine a n d neural actions of CRF (Vale et al., 1981). From both human and rat cDNAs, 322 amino acid precursors and 298 amino acid mature proteins containing 10 cysteines are predicted, suggesting the presence of five disulfide bonds. To gain insight into the structural properties of the protein we have now determined the disulfide bond arrangement of the expressed human CRF-BF?
MATERIALS AND METHODS Expression of CRF-BP in Stably Dansfected Chinese Hamster Ovary Cells-The full-length 1.9-kilobase pair CRF-BP cDNA containing its own polyadenylation signal sequence was cloned into expression vector pSG5 (Stratagene) downstream from a strong SV40 promoter as described previously (Potter et al., 1991). Chinese hamster ovary cells * This work was supported by National Institutes of Health Grants DK 26741 and HD 13527. It was also supported by the Foundation for Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. were transfected with this construct and a Rous sarcoma virus neomycin construct by the calcium phosphate procedure. Stable cell colonies were selected over a period of 2 weeks by growing the cells in complete media (Dulbecco's modified Eagle's medium, 10% fetal calf serum, 2 m~ glutamine) containing 1 mg/ml G418. Colonies were subsequently isolated and grown in complete media as separate cell lines. Media from each of the lines were then screened for the presence of the CRF-BP by virtue of their ability to inhibit binding of a CRF antibody to radiolabeled CRF trace, and the one producing the highest quantity of CRF-BP was then aliquoted and stored in liquid nitrogen to be used for all subsequent expression studies. For large scale protein expression, cells were grown to confluence in 500 ml of complete medium on large tissue culture plates. Two liters of binding protein medium was then percolated through a 1-ml CRF-solid phase (which was prepared by coupling 1 mg of rathuman CRF to 1 ml of Affi-Gel-10) overnight with the aid of a peristaltic pump. Bound recombinant CRF-BP was then eluted in 50 m~ sodium acetate, formate, 20% acetonitrile buffer, pH 3.0, essentially as already described (Behan et al., 1989). Both the media from the cells and the purified CRF-BP were found to exhibit the same characteristics as the natural material from human plasma in that they inhibited CRF-induced ACTH secretion from rat anterior pituitary cells in vitro and hindered binding of a CRF antibody to radiolabeled CRF trace with identical dose-response characteristics.
Dypsin Digest ofCRF-BP-Affinity-purified CRF-BP (approximately 10 pg) was further purified by reversed phase HPLC. (A gradient from 18 to 54% acetonitrile was run within 30 min at a flow rate of 0.2 ml/min.) The protein eluting at 28 min was collected and concentrated to dryness in a Savant Speed-Vac concentrator. Trypsin (sequencing grade, 1 pg) dissolved in 50 pl of TES buffer (0.1 M TES, 5% acetonitrile, pH 8, adjusted with NaOH) was added. After 2 h of incubation at 37 "C, Tween 20 was added to a final concentration of 0.5%. The sample was divided into two equal aliquots. One aliquot was mixed with 75 pl of 0.5 M aqueous acetic acid and subjected to HPLC separation. To the other aliquot, 55 pl of ammonium acetate buffer (0.1 M, pH 4.5) and TCEP (0.5 M in dH,O, 20 p1) was added, followed by incubation for 30 min at room temperature. The sample was kept frozen at -20 "C until injected into the HPLC.
Endoproteinase Asp-N Digest ofCRF-BP-Protein (approximately 10 pg) was purified by HPLC and dried down as described above.'Endoproteinase Asp-N (sequencing grade, 1 pg) dissolved in sodium phosphate buffer (25 m~, 5% acetonitrile, 0.2% Nonidet P-40, pH 8) was added, and digestion was allowed to proceed for 16 h at 37 "C. The sample was divided into two aliquots. One aliquot was injected into the HPLC after mixing with 75 pl of 0.5 M acetic acid; the other aliquot was reduced using TCEP as described above.
Reversed Phase HPLC-AHewlett-Packard 1090L Liquid Chromatograph equipped with a Vydac C-18 column (2.1 x 150 mm; particle size: 5 pm; pore size: 300 A) was used. The flow rate was 0.2 mumin. Samples were collected manually based on absorbance at 210 nm.
Sequence Analysis-Sequence analysis was performed on an Applied Biosystems 470A protein sequencer equipped with an on-line phenylthiohydantoin-derivative analyzer (AB1 120A). Samples were applied to polybrene-coated glass fiber filters and sequenced using the standard program supplied by the manufacturer. Fig. 1) expressed by stably transfected Chinese hamster ovary cells was treated with two highly specific proteases, trypsin and endoproteinase Asp-N. The resulting fragments were separated by reversed phase HPLC. Ali- quots of the digests were reduced using the novel reducing agent TCEP (Burns et al., 1991;Fischer et al., 1993). This agent reduces disulfide bonds efficiently at room temperature and at an acidic pH. The reduced digestion mixtures were resolved on HPLC under conditions identical to the ones used for nonreduced samples. Fragments that shifted their elution positions upon reduction (Fig. 2) were deduced to contain a disulfide bond. The nonreduced fractions equivalent to the ones that shifted were thus subjected to Edman degradation in an automated gas phase sequencer. The results of the sequence analyses are summarized in Table I.

CRF-BP (sequence in
Digestion with endoproteinase Asp-N yielded two such fragments which were designatedAl andA2. Fraction A1 contained two peptides, the sequences of which could be mapped to CRF-BP 35-74 and CRF-BP 75-97, establishing the connectivity of CysGo and Cys81. The sequence of fraction A2 could be mapped to CRF-BP 177-218 containing 2 cysteines (Cysla3 and Cys205), thus establishing the connectivity of the two. The shift of elution position for this fragment upon reduction is minimal (0.5 min) as expected for a fragment with an intramolecular disulfide bond.
In the tryptic digest of CRF-BP, four fragments (designated Tl-T4) were observed that changed their elution behavior upon reduction. Fraction T1 could be mapped to CRF-BP 57-111 and CRF-BP 137-146. The first of these two sequences contains 3 cysteines (CysGo, Cysa', Cys1O4), the second sequence contains 1 cysteine (Cys141). With the connectivity of CysG0 to CysS1 established previously, it follows that Cyslo4 and Cys141 form a disulfide bond. Sequence analysis of fraction T2 revealed the presence of two sequences both of which contain 1 cysteine. The sequences could be mapped to CRF-BP 233-256 and CRF-BP 257-274, thus establishing the presence of a disulfide bond between Cys237 and Cys2-. Fraction T3 mapped to the same area of the molecule but in one of the peptides a chymotrypsin-like cleavage occurred between Leu246 and Leu247. The last tryptic fragment (T4) was found to contain four sequences containing 1 cysteine each. The sequences were mapped to CRF-BP fragments 176-193, 194-216, 275-283, and 300-322. As the connectivity of Cysla3 and Cys205 had already been established by Fragment A2, it could be concluded that cysteines in position 277 and 318 are joined by a disulfide bond. These results are summarized schematically in Fig. 3.
The disulfide bonding pattern of all 10 cysteines could thus be assigned unambiguously. A pattern is observed where each cysteine is joined to the next one in a sequential manner. The fragments that shifted their elution position upon reduction and that were sequenced are labeled.
To assess the biological significance of disulfide bonding in CRF-BP, the ability of the reduced protein to bind CRF was tested. CRF-BP was reduced by TCEP treatment and isolated by reversed phase HPLC. The ability of the reduced protein to bind CRF was compared with that of nonreduced CRF-BP. As shown in Fig. 4, no significant binding activity can be detected for the reduced binding protein. This demonstrates the importance of the correct disulfide arrangement for biological activity.

DISCUSSION
The strategy utilized here to determine the disulfide arrangement in CRF-BP involved digestion of the protein with specific proteolytic enzymes and resolution of the fragments by reversed phase HPLC with and without treatment with reduc- The sequence signal observed for the proteolytic fragments is given and mapped to the known sequence of hCFRBF? The numbering of the residues follows the one published (Potter et al., 1991). For each fragment the residues observed in sequencing cycle are listed in the order they elute on the phenylthiohydantoin analyzer. The fragments of CRF-BP to which these residues could be mapped are listed underneath the observed sequence.

D P N L F P C N V I S Q X P N G K F X L V V P X Q H Residues observed
Mapped to hCRF-BP

60-111
137-147 ing agent. Two enzymes, trypsin and endoproteinase Asp-N, were chosen based on the following considerations: both enzymes are highly specific, i.e. they only cleave at well defined sites. Inspection of the protein sequence predicted from the cDNA showed that cleavage sites were distributed between cysteine residues in CRF-BP such that small fragments containing 1 cysteine each could be expected. Fragments that changed their elution behavior upon reduction were concluded to contain cystine. Thus, the corresponding nonreduced fragments were subjected to chemical sequence analysis. In cases where two cysteine-containing peptides are connected by a disulfide bond, two sequence signals can be expected. The expected HPLC elution pattern for such a frag- Qn-OIhCRF to nonreduced (0) and reduced (0) CRF-BP is expressed as the ratio of the counts bound to the total counts (50,000 cpm) added. The complex of CRF-BP and bound CRF was precipitated with a specific anti-CRF-BP antibody. Details of this assay are described elsewhere (Behan et al., 1993b). ment shows one compound before and two compounds after reduction. If the disulfide bond is within the fragment, only one sequence should be observed. In the latter case only a minor change in elution behavior is expected as the net change in hydrophobicity is minute. When the ability of reduced CRF-BP to bind CRF was tested, it was found that this ability was abolished. It is concluded that the stabilization of the structure by disulfide bonds is essential for biological activity. The reducing agent used in this study, TCEP (Burns et al., 1991;Fischer et al., 1993), has several advantages over conventional reducing agents such as dithiothreitol or P-mercaptoethanol. It is highly efficient at low temperatures (21-37 "C), and incubation times are short (30 min or less). The reagent is specific toward disulfide bonds in proteins; it is not reactive to other functional groups found in proteins. Another major advantage is the absence of contaminants absorbing at 210 nm, which simplifies comparison of HPLC profiles of reduced and nonreduced materials.
It was found that the disulfide arrangement in CRF-BP follows a pattern where each cysteine is joined to the next one in a sequential manner. This sequential disulfide arrangement is reminiscent of that found in members of the immunoglobulin superfamily (Williams, 1987). One characteristic of this class of protein is the presence of domains defined by 5 sequential cystines. A sequence comparison between CRF-BP and members of the immunoglobulin superfamily, however, did not reveal any significant homologies at the amino acid sequence level. The loop size, i.e. the number of amino acid residues between 2 cysteines, is smaller in CRF-BP (20-46 residues) than that of the members of the immunoglobulin family (approximately 70 residues). The difference in loop size and the absence of sequence homology make a prediction of tertiary structure based on the known structures of immunoglobulins infeasible.
For the immunoglobulin superfamily it has been proposed that the domains defined by disulfide linked loops arose from gene duplication of a single ancestral gene (Williams, 1987). One argument in favor of this hypothesis is the fact that in most members of that superfamily these domains are found on separate exons. Analysis of the genomic CRF-BP sequence reveals that four of the domains defined by disulfide loops are found on single exons, whereas the domain defined by cysteine 104 and cysteine 141 is disrupted by an intron (Behan et al., 1993a).
To test whether the loops formed by the disulfide bonds in CRF-BP define domains that might have arisen from gene duplication as has been postulated for the immunoglobulin family, a comparison for internal homologies was carried out by the Dotplot program (University of Wisconsin Genetics Computer Group (UWGCG) program package). This program compares the sequence to itself at a given stringency marking homologous regions by a dot. Two regions were identified that exhibited moderate homology to each other. A computer alignment analysis of this region using the GAP program (UWGCG program package) revealed that CRF-BP 101-112 and CRF-BP 234-246 are 50% identical and show 83% similarity allowing a 1 amino acid gap. These two regions are defined by the second and fourth disulfide loop and are each contained on a separate exon (exon 3 and exon 6 (Behan et al., 1993a1.1. It is speculated that the five-loop motif found in CRF-BP, in analogy to the immunoglobulin structure, provides a framework in which the disulfide bonds are at the core of the structure and the loops are arranged to form pockets or clefts on the surface for substrate binding. Considering the size of the loops found in CRF-BP, it is hypothesized that at least two loops are involved in binding interactions, whereas additional loops stabilize the structure. In the absence of three-dimensional structure information, this hypothesis can now be tested by expression of shortened forms of CRF-BP containing pairs of loop domains and by determining the affinity for CRF of these analogs.