The N-terminal Disulfide Linkages of Human Insulin-like Growth Factor-binding Protein-6 (hIGFBP-6) and hIGFBP-1 Are Different as Determined by Mass Spectrometry*

The actions of insulin-like growth factors (IGFs) are modulated by a family of six high affinity binding proteins (IGFBPs 1–6). IGFBP-6 differs from other IGFBPs in having the highest affinity for IGF-II and in binding IGF-I with 20–100-fold lower affinity. IGFBPs 1–5 contain 18 conserved cysteines, but human IGFBP-6 lacks 2 of the 12 N-terminal cysteines. The complete disulfide linkages of IGFBP-6 were determined using electrospray ionization mass spectrometry of purified tryptic peptide complexes digested with combinations of chymotrypsin, thermolysin, and endoproteinase Glu-C. Numbering IGFBP-6 cysteines sequentially from the N terminus, the first three disulfide linkages are Cys1-Cys2, Cys3-Cys4, and Cys5-Cys6. The next two linkages are Cys7-Cys9 and Cys8-Cys10, which are analogous to those previously determined for IGFBP-3 and IGFBP-5. The C-terminal linkages are Cys11-Cys12, Cys13-Cys14, and Cys15-Cys16, analogous to those previously determined for IGFBP-2. Disulfide linkages of IGFBP-1 were partially determined and show that Cys1 is not linked to Cys2 and Cys3 is not linked to Cys4. Analogous with IGFBP-3, IGFBP-5, and IGFBP-6, Cys9-Cys11 and Cys10-Cys12 of IGFBP-1 are also disulfide-linked. The N-terminal linkages of IGFBP-6 differ significantly from those of IGFBP-1 (and, by implication, the other IGFBPs), which could contribute to the distinctive IGF binding properties of IGFBP-6.

The insulin-like growth factor (IGF) 1 system plays an important role in normal physiology (1). Dysregulation of the system has also been implicated in many diseases including malignancy, atherosclerosis, and the development of diabetic complications. IGF-I and IGF-II promote proliferation and differentiation of many cell types. More recently, IGFs have been ascribed a potent role as anti-apoptotic agents.
The actions of IGFs are modulated by a family of six structurally related high affinity IGF-binding proteins (IGFBPs 1-6) (1)(2)(3). The N-and C-terminal domains of IGFBPs 1-6 share significant sequence homology, whereas the central regions linking these domains are not homologous with each other. Under differing circumstances, IGFBPs may inhibit or enhance IGF actions. The IGFBPs differ in their regulation, relative IGF binding affinities, susceptibility to proteolysis by specific proteases, and sites of synthesis. The IGFBPs therefore constitute a flexible system for the regulation of IGF activity.
IGFBP-6 is an O-linked glycoprotein, which differs from IGFBPs 1-5 in a number of significant respects. Of the IGFBPs, IGFBP-6 has the highest binding affinity for IGF-II (4). IGFBP-6 also has the highest specificity for IGF-II, exhibiting a 20 -100-fold higher binding affinity for IGF-II than IGF-I, whereas IGFBPs 1-5 bind to IGF-II with equal or slightly higher affinity than IGF-I (4 -7). A potentially significant structural difference is that, whereas human and rat IGFBPs 1-5 have 12 homologous N-terminal and 6 homologous C-terminal cysteines, 2 of the N-terminal cysteines are not present in human and rat IGFBP-6 and a short sequence containing another 2 N-terminal cysteines is missing from rat IGFBP-6 (8).
Disulfide linkages of cysteines are important for the correct folding and maintenance of the three-dimensional structure of many proteins. The N-terminal cysteines of the IGFBPs are disulfide-linked to each other, as are the C-terminal cysteines, with no disulfide linkages between the N-and C-terminal domains (2). To date, a complete set of disulfide linkages has not been reported for any of the IGFBPs. Recently, the three Cterminal disulfide linkages of bovine IGFBP-2 were determined (9). The C-terminal disulfide linkages of human IGFBP-6 were also partially determined (10) and are consistent with those of IGFBP-2. We recently suggested that the N-terminal region of IGFBP-6 consists of two disulfide-linked domains, the first including the six most N-terminal cysteines and the second including the next four cysteines in the sequence (10). A recent structural study of IGFBP-5 has shown that the second of these domains contains the high affinity IGF binding site (11); the two disulfide linkages within this domain are the same as those determined for the homologous cysteines of IGFBP-3 (12). The other N-terminal domain of IGFBP-5, containing the eight most N-terminal cysteines, is thought to form a separate domain that does not interact with the high affinity binding site of IGFBP-5 nor strongly influence IGF binding to this site (11).
IGFBPs 1-5 contain a GCGCC motif in their N-terminal sequences, which is replaced by GCAEA in human IGFBP-6 (8). Therefore, unless the two adjacent cysteines in IGFBPs 1-5 are disulfide-linked, which is unlikely on structural grounds, the disulfide linkages of IGFBP-6 in this domain must differ from those of the other IGFBPs. A chimera of the N-terminal and non-conserved middle regions of IGFBP-6 with the C-terminal region of IGFBP-5 retains preferential binding affinity for IGF-II over IGF-I (13). Since the non-conserved middle regions of IGFBPs are not thought to be directly involved in high affinity IGF binding, this suggests that the determinants of IGF-II binding preference are located in the N-terminal region. In order to further characterize the structural differences between IGFBP-6 and IGFBPs 1-5, we completely solved the disulfide linkages of human IGFBP-6 using electrospray ionization mass spectrometry (14) and reverse-phase HPLC of proteolytic digests. Since none of the N-terminal disulfide linkages and only a few other linkages of some of IGFBPs 1-5 are known, we also partially solved the disulfide linkages of IGFBP-1 to provide a basis for comparison. We found that at least two of the N-terminal disulfide linkages of IGFBP-1 are different from those of IGFBP-6.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant Human IGFBP-6 -Recombinant glycosylated human IGFBP-6 (hIGFBP-6) was previously expressed in Chinese hamster ovary cells and purified by IGF-II affinity chromatography and reverse-phase FPLC, but was contaminated with endogenously synthesized IGFBP-4, which copurified with hIGFBP-6, necessitating lectin chromatography for removal of the IGFBP-4 (10). For this reason, human embryonic kidney 293 cells, which synthesize only small amounts of IGFBP-2 (15), were stably transfected with phBP6-E3, the eukaryotic expression vector encoding human IGFBP-6 (16), using LipofectAMINE. Geneticin-resistant clones were selected, and the colony producing the highest levels of hIGFBP-6 were recloned and expanded for production of recombinant protein.
Proteolytic Digestion and Reverse-phase HPLC of hIGFBP-1-Proteolytic digestion of hIGFBP-1 and reverse-phase HPLC was carried out as described above for hIGFBP-6, except that 20 g of hIGFBP-1 was digested (20 h, 21°C) with 0.5 g of trypsin and 0.5 g of S. aureus V8 protease (endoproteinase Glu-C/Asp-C when in phosphate buffer) combined in 45 l of 50 mM sodium phosphate (pH 7.8), 0.8 M urea. After ESMS analysis of reverse-phase HPLC fractions, 5-pmol aliquots of the isolated 5-kDa N-terminal hIGFBP-1 fragment were digested (20 h, 21°C) with 1 pmol of thermolysin in 2 l of 100 mM ammonium bicarbonate (pH 7.8) and re-analyzed by ESMS.
Electrospray Ionization Mass Spectrometry (ESMS)-Samples (1-3 l) of HPLC fractions and their digests were mixed with 3-6 l of 50% acetonitrile, 0.1% formic acid prior to ESMS, which was performed as described previously (10, 18) on a Perkin-Elmer Sciex API-300 triple quadrupole mass spectrometer with micro-ion spray ion source, calibrated to an accuracy of Ϯ0.01% using singly charged poly(propylene glycol) ions. MS/MS spectra (Q3 scans) were obtained using nitrogen collision gas (4 millitorr pressure, 20.7 cm cell length) and optimized collision energies of 32-64 eV. Signal-averaged raw mass spectra were analyzed manually and transformed to a true mass scale using the PE-Sciex BioMultiview program Biospec Reconstruct. Peptide sequences inferred from observed masses were confirmed by analysis of MS/MS spectra manually and using the BioMultiview programs Predict Sequence and Peptide Fragments.
Amino Acid Sequencing-N-terminal amino acid sequencing was performed by sequential Edman degradation using a Hewlett-Packard G1005A automated protein sequencing system, calibrated with phenylthiohydantoin-amino acid standards prior to each sequencing run.

RESULTS
Expression and Purification of hIGFBP-6 in 293 Cells-Approximately 300 g of recombinant human IGFBP-6 was purified from 500 ml of medium conditioned by stably transfected 293 human embryonic kidney cells. N-terminal Edman sequencing showed that ϳ80% of the protein commenced at Arg 28 with the sequence R(C)PG(C)GQG (numbering based on the sequence of the IGFBP-6 precursor protein, SWISS-PROT accession number P24592; Fig. 1). The remainder of the protein FIG. 2. Reverse-phase HPLC fractionation of a tryptic digest of hIGFBP-6. As described under "Experimental Procedures," hIGFBP-6 (40 g) was denatured without reduction, digested with trypsin, and subjected to C18 reverse-phase HPLC using a gradient of 0 -50% acetonitrile, 0.1% trifluoroacetic acid over 150 min. Upper trace, A 215 . Lower trace, A 280 . Peptides were identified by ESMS. Samples of fractions 1, 2, and 3, containing disulfide-linked tryptic peptide complexes, were further analyzed by ESMS after digestion with chymotrypsin or thermolysin (or both together), or S. aureus V8 protease (endoproteinase Glu-C), as described in Table I. FIG. 1. Amino acid sequences of hIGFBP-6 and hIGFBP-1. Amino acids are numbered according to the SWISS-PROT entries for proIGFBP-6 (accession number P24592) and proIGFBP-1 (accession number P08833).
commenced at Leu 26 with the sequence LAR(C)PG(C)G. This heterogeneity of the N-terminal sequence is consistent with previous studies of IGFBP-6 purified from natural and recombinant sources (10,19,20). The previously observed absence of Gly 240 from ϳ70% of IGFBP-6 (10) was also identified in Cterminal fragments analyzed in the present study. In the results that follow, cysteines are numbered 1-16 in hIGFBP-6 and 1-18 in hIGFBP-1 sequentially from the N terminus ( Fig. 1).
Purification of Disulfide-linked Tryptic Fragments of hIGFBP-6 -Disulfide-linked tryptic fragments of hIGFBP-6 were produced by tryptic digestion of 40 g of purified hIGFBP-6 followed by C18 reverse-phase HPLC fractionation, which yielded a HPLC profile ( Fig. 2) similar to that following a previous digestion using the same procedure on purified hIGFBP-6 expressed in Chinese hamster ovary cells (10). ESMS analysis of all major and some minor peak fractions confirmed that the three most intense peaks (Fig. 2, peaks 1-3) correspond to the same three hIGFBP-6 disulfide-linked tryptic fragments of hIGFBP-6 that were previously identified by ESMS and Edman sequencing (10) and that other major and minor peptide fragments were also as previously identified. As reported previously (10), attempts to release disulfide-linked peptides by reduction with 2-mercaptoethanol in concentrations compatible with electrospray ionization (less than a few percent, v/v) were unsuccessful at room temperature and resulted in peptide degradation and/or loss at elevated temperatures. Therefore, assigned sequences for all fragments with masses below 2300 Da were confirmed by MS/MS fragmentation analysis as described under "Experimental Procedures." Fractions 1, 2, and 3 were further digested and reanalyzed by ESMS, as described below. The peptide complexes in these three fractions account for 14 of the 16 cysteines in hIGFBP-6. By default, and as previously determined (10), the remaining two cysteines (Cys 11 and Cys 12 in the sequences below) are disulfide-linked, and connect glycosylated (140 -165)NPGTST-TPSQPNSAGVQDTEMGPC 11 RR to (181-193)GAQTLYVPNC-12 DHR. A number of glycoforms of this complex, with masses corresponding to carbohydrate contents of 3-7 N-acetylhexosamine-hexose disaccharides and 0 -5 sialic acids, were identified as minor components of peak fraction 1 (Fig. 2) and nearby fractions (data not shown), thereby confirming that Cys 11 is indeed linked to Cys 12 in hIGFBP-6 expressed in 293 cells.
ESMS of Proteolytic Digests of hIGFBP-6 Disulfide-linked Tryptic Fragments-The three disulfide-linked tryptic complexes purified as HPLC fractions 1, 2, and 3 ( Fig. 2) were subjected to further digestion by S. aureus V8 protease (endoproteinase Glu-C), thermolysin, chymotrypsin, or a mixture of chymotrypsin and thermolysin, as detailed in Table I. Samples of fraction 2 ( Fig. 2), containing the 3634-Da N-terminal tryptic fragment (28 -65)RC 1 PGC 2 GQGVQAGC 3 PGGC 4 VEEEDGG-SPAEGC 5 AEAEGC 6 LR with three intramolecular disulfide linkages, were digested with endoproteinase Glu-C and reanalyzed by ESMS (Table I). This revealed digestion into two complementary fragments with masses of 1619 and 2032 Da (Table I), as shown in Fig. 3A. Minor fragments were present with extra or missing Glu (E), consistent with alternate Glu-C cleavages in the sequence EEED 49 (Fig. 3A), and minor fragments due to additional cleavage at one or more Glu residues (E) in the sequence EGCAEAE 61 were also observed at low mass (not shown). The identity of the 1619-Da fragment as (49 -65)DGGSPAEGC 5 AEAEGC 6 LR, with the disulfide linkage Cys 5 -Cys 6 , was confirmed by MS/MS collisional fragmentation (spectrum not shown) of the doubly protonated (m/z 811) ion, which generated the following b-type (N-terminal) and complementary y-type (C-terminal) sequence ions: b 2 -b 8 , y 9 -y 13 (singly charged), and y 13 -y 15 (doubly charged). The ions b 9 -b 14 and y 3 -y 8 were completely absent, as expected due to the presence of the disulfide link bridging the cysteines at positions 9 and 15 in the fragment sequence (21).
The identity of the 2031-Da fragment as (28 -48)RC 1 PGC-2 GQGVQAGC 3 PGGC 4 VEEE, with two intramolecular disulfide linkages, was confirmed and the disulfide linkage arrangement also determined by MS/MS collisional fragmentation of the doubly protonated (m/z 1017) ion (Fig. 3B). Prominent b-type (N-terminal) fragment ions b 5 -b 12 and b 17 -b 21 were TABLE I ESMS analysis of hIGFBP-6 disulfide-linked proteolytic fragments and deduced disulfide linkages Molecular masses of major observed components were determined by ESMS of disulfide-linked tryptic peptide complexes in HPLC fractions 1, 2, and 3 ( Fig. 2), with or without additional digestion (as indicated under "protease") by endoproteinase Glu-C (Glu-C), thermolysin (Therm), chymotrypsin (Chym), or a mixture of chymotrypsin and thermolysin (Ch&Th), as described under "Experimental Procedures." Sequence assignments of tryptic peptide complexes were determined previously by ESMS and Edman sequencing, as was the linkage C 11 -C 12 (10). Sequence assignments of fragments with masses less than 2300 Da from additional digestion were confirmed by MS/MS fragmentation analysis as in Fig.  3 (other data not shown). ESMS (observed) masses averaged from several charge states (typically 2ϩ to 4ϩ) are compared with calculated average molecular masses above 2000 Da, but with monoisotopic (m) masses below 2000 Da due to the lower observed charge states (1ϩ to 2ϩ). Uncertainties are mean deviations plus calibration uncertainty. Calculated masses assume disulfide linkage of all cysteines (C), which are numbered 1-16 sequentially (Fig. 1). Long sequences are abbreviated (-). Deduced disulfide linkages are summarized at the foot of the table and shown in Fig. 4 Deduced hIGFBP-6 linkages N-terminal: C 1 -C 2 , C 3 -C 4 , C 5 -C 6 , C 7 -C 9 , C 8 -C 10 C-terminal: C 11 -C 12 (see legend), C 13 -C 14 , C 15 -C 16 observed, while the ions b 13 -b 16 and b 2 -b 4 were conspicuously absent, as expected due to the presence of bridging disulfide linkages (21). The key observation is the presence of prominent b 5 -b 12 ions, which would not be observed unless the two disulide linkages are sequential (do not cross over), as shown schematically in Fig. 3B. The linkage arrangement Cys 1 -Cys 2 was independently determined by ESMS analysis of thermolysin digests of the 3634-Da N-terminal tryptic fragment, which were found to contain the complementary fragments (28 -35)RC 1 PGC 2 GQG and (36 -65)VQAGC 3 PGGC-4 VEEEDGGSPAEGC 5 AEAEGC 6 LR as major components (Table I). Combining this result with that from Glu-C digestion (Fig. 3A) independently implies the same disulfide linkages as those determined by MS/MS (Fig. 3B). It is concluded that the three most N-terminal disulfide linkages of hIGFBP-6 are Cys 1 -Cys 2 , Cys 3 -Cys 4 , and Cys 5 -Cys 6 .
Samples of fraction 3 (Fig. 2), containing the 3680-Da tryptic   (Table II, Fig. 6) and the 5275-Da N-terminal disulfide-linked peptide complex in peak fraction 16 additionally analyzed by Edman sequencing (Table III). complex (67-95)EGQEC 7 GVYTPNC 8 APGLQC 9 HPPKDDEAP LR disulfide-linked to (104 -108)C 10 LPAR (Table I), were analyzed by ESMS after digestion with thermolysin, chymotrypsin, or a mixture of thermolysin and chymotrypsin. Digestion of the fraction 3 complex with chymotrypsin or thermolysin alone resulted in an 18-Da mass increase of the complex (or with thermolysin, the complex minus C-terminal LR), consistent with peptide bond hydrolysis at V-Y (or G-L with thermolysin) but with fragments still disulfide-linked (data not shown). Digestion of the complex with a mixture of thermolysin and chymotrypsin resulted in the generation of two complementary fragments with masses of 2231 and 1215 Da, corresponding to (67-74)EGQEC 7 GVY linked to (82-93)LQC 9 HPPKDDEAP and (75-81)TPNC 8 APG linked to (104 -108)C 10 LPAR, respectively (Table I). These and other assignments were confirmed by MS/MS analysis (data not shown). It is concluded that Cys 7 is linked to Cys 9 and Cys 8 is linked to Cys 10 .
Samples of fraction 1 (Fig. 2) containing the C-terminal disulfide-linked tryptic complex (200 -202)QC 13 R linked to (209 -217)RGPC 14 WC 15 VDR, which is in turn linked to (221-239)SLPGSPDGNGSSSC 16 PTGSS (Table I), were digested with chymotrypsin at protease-to-peptide molar ratios that were slightly (2-3-fold) higher than those used with other digests in order to obtain complete digestion. ESMS analysis showed that two major fragments with masses of 1021 and 2183 Da were generated (Table I) (Table I), all three C-terminal disulfide linkages are therefore sequential. The complete disulfide linkages determined for hIGFBP-6 are summarized at the foot of Table I and shown in Fig. 4.
ESMS and Edman Sequencing of Proteolytic Fragments of hIGFBP-1-hIGFBP-1 was digested with a mixture of trypsin and S. aureus V8 protease and subjected to C18 reverse-phase HPLC fractionation as shown in Fig. 5. All major and most minor HPLC peak fractions (Fig. 5) were analyzed by ESMS (Table II) and initially identified by comparison of observed masses with those expected on the basis of the hIGFBP-1 sequence (Fig. 1) and the protease specificities of trypsin (Arg-  Fig. 5) from reverse-phase HPLC of hIGFBP-1 digested with a mixture of trypsin and S. aureus V8 protease (endoproteinase Glu-C/Asp-C) were analyzed by ESMS as described under "Experimental Procedures," yielding observed masses as indicated. Uncertainties are mean deviations plus calibration uncertainty. Calculated molecular masses based on the hIGFBP-1 sequence (Fig. 1) are monoisotopic below 2000 Da and average (av) above 2000 Da. Assigned sequences for all fragments with masses below 2300 Da were confirmed by MS/MS fragmentation analysis (data not shown) as in Fig. 3. The sequence assignment for the 5275-Da fraction 16 fragment (see Fig. 6 for mass spectrum) was confirmed by Edman sequencing (Table III). Disulfide-linked peptides (indicated by *) identified in HPLC fractions 4, 13, 14, and 16 (indicated in bold) account for all cysteines except C 15 , C 16 , C 17 , and C 18 . Digestion of the 5275-Da fraction 16 complex by thermolysin (Th) generated a major 3771-Da fragment (as indicated), which corresponds to removal of (42-52)LGAAC 9 GVATAR linked to (56 -60)GLSC 11 R, demonstrating that C 9 is connected to C 11 . Deduced disulfide linkage information for hIGFBP-1 is summarized at the foot of the table. All cysteines (C) are numbered from 1 to 18 sequentially (Fig. 1) Deduced hIGFBP-1 linkage information C 1 not linked to C 2 ; C 3 not linked to C 4 ; one or both of C 1 , C 2 , and one or both of C 3 , C 4 linked to two or four of C 5 , C 6 , C 7 , C 8 ; C 9 -C 11 (see legend); C 10 -C 12 ; C 13 -C 14 . By default, C 15 -C 16 (or C 17 ); C 18 -C 17 (or C 16 ).
C/Lys-C) and S. aureus V8 protease (Glu-C/Asp-C under the digest conditions employed), taking into account potential missed cleavages and possible disulfide linkage combinations. Use was also made of UV absorbance at 280 nm compared with that at 215 nm (Fig. 5) to gauge the fractional content of tryptophan and tyrosine in peptide fragments. Assignments of all peptides and disulfide-linked peptide complexes with masses below 2300 Da (Table II) were subsequently confirmed by MS/MS fragmentation analysis (data not shown).
Major disulfide-linked peptide fragments were identified in HPLC peak fractions 4, 13, 14, and 16 (Table II). Fraction 4 contained the fragment (53-55)Cys 10 AR linked to (75-82) GQGACys 12 VQE, demonstrating that Cys 10 is linked to Cys 12 . Fraction 13 contained the fragment (146-154)KWKEPCys 13 RIE linked to (176-183)FYLPNCys 14 NK, demonstrating that Cys 13 is linked to Cys 14 . A related fragment with the same disulfide linkage, but in which K 146 had been removed during digestion, was identified in Fraction 14, and additional related fragments due to further removal of WK 148 and/or IE 154 were identified as minor components of fractions 10 and 11 (data not shown).
Fraction 16 contained a 5275-Da fragment (Fig. 6), the mass of which corresponds precisely with that calculated for the N-terminal disulfide-linked complex consisting of the following fragments: (1-12)APWQC 1 APC 2 SAEK, (13-25)LALC 3 PPVSA-SC 4 SE, (29 -52)SAGC 5 GC 6 C 7 PMC 8 ALPLGAAC 9 GVATAR, and (56 -60)GLSC 11 R (Table II, fragments are those predicted for cleavage after R or K by trypsin, and after E by endoproteinase Glu-C). This assignment was confirmed by N-terminal Edman sequencing (Table III), the results of which are consistent with the apparent purity of fraction 16 when analyzed by ESMS (Fig. 6). Samples of fraction 16 were digested with thermolysin and re-analyzed by ESMS, which revealed an additional major 3771-Da fragment ( Table II) that was not present before digestion (Fig. 6) and was not present in a zero digestion-time control sample (data not shown). This fragment corresponds to removal of (42-52)LGAAC 9 GVATAR linked to (56 -60)GLSC 11 R, thereby demonstrating that Cys 9 is linked to Cys 11 . Fragments corresponding to the removed disulfidelinked section (or products of its further digestion by thermolysin on the N-terminal side of L, V, or A) may have been present in digests but were not clearly identified. Further digestion of the 5275-Da N-terminal complex to determine the remaining four disulfide linkages was precluded by the lack of suitable cleavage sites.
The 5275-Da N-terminal disulfide-linked complex incorporates the fragments (1-12)APWQC 1 APC 2 SAEK and (13-25)LALC 3 PPVSASC 4 SE (Tables II and III), each of which contains only two cysteines; therefore, in order to be part of the 5275-Da complex, Cys 1 cannot be connected to Cys 2 and Cys 3 cannot be connected to Cys 4 . For the same reason, Cys 1 , Cys 2 , Cys 3 , and Cys 4 cannot all be interconnected. Given the result from thermolysin digestion of the 5275-Da complex (Cys 9 linked to Cys 11 ), it is therefore concluded that one or both of Cys 1 and Cys 2 , and one or both of Cys 3 and Cys 4 must be connected to two or four of the cysteines in the sequence (29 -41)SAGC 5 GC 6 C 7 PMC 8 ALP (Table II).
The identified disulfide-linked peptides (Table II) account for all hIGFBP-1 cysteines except the four C-terminal cysteines, Cys 15 to Cys 18 . No peptide complexes containing any of these cysteines were identified and, in particular, the single complex expected to contain all four cysteines (Table II) was not observed. All other expected fragments containing more than four amino acids were recovered (Table II), and all fragments in all major and minor peak fractions analyzed were successfully identified. For example, fractions 10 and 15 (not listed in Table  II) contained peptides that were identified as oxidized (methi-onine sulfoxide) forms of those in fractions 12 and 16, respectively (respective masses 16 Da higher; data not shown). The C-terminal peptides unaccounted for are: (191-193)QC 15 E, (200 -212)AGLC 16 WC 17 VYPWNGK, and (222-234)GDPNC-18 QIYFNVQN. While it is possible that these are not linked to form a single complex, the loss of a single large peptide complex (possibly due to excessive hydrophobicity or other factors leading to aggregation) is more probable than the simultaneous loss of two or more smaller entities. Therefore it is concluded that Cys 15 is most likely linked to Cys 16 or Cys 17 , and Cys 18 is most likely linked to Cys 17 or Cys 16 . Deduced disulfide linkage information for hIGFBP-1 is summarized at the foot of Table II. DISCUSSION The IGFBPs are important modulators of IGF actions (1)(2)(3). Although the actions of IGFBPs have been the subject of intense scrutiny for over a decade, only limited information regarding their structure is available. The results of this study demonstrate for the first time the complete disulfide linkages of an IGFBP. In addition, the disulfide linkages of IGFBP-1 have been partially characterized and compared with those of IGFBP-6, revealing that the N-terminal disulfide linkages of these IGFBPs differ significantly. In contrast, the C-terminal disulfide linkages of bovine IGFBP-2 (9) and human IGFBP-6 are identical, and the disulfide linkages of IGFBP-5 within its putative high affinity IGF binding domain (11) are identical to those in the corresponding domains of IGFBP-6, IGFBP-1 (both shown in the present study) and IGFBP-3 (12).
A distinctive feature of IGFBP-6 compared with IGFBPs 1-5 is its high binding affinity for IGF-II and its marked preferential binding of IGF-II relative to IGF-I. A chimera of the Nterminal and middle regions of IGFBP-6 with the C-terminal region of IGFBP-5 retains the binding preference for IGF-II (13). Since the non-conserved middle regions of the IGFBPs are not thought to be directly involved in IGF binding, the structural determinants of the IGF-II binding preference are therefore likely to reside in the N terminus. It may be of structural significance that rat IGFBP-6 shares the IGF-II binding preference with human IGFBP-6 (22) despite lacking the AGC 3 PGGC 4 VE peptide segment present in human IGFBP-6  Tables II and III) and adjacent minor peaks corresponding to oxygen, sodium, potassium, and ferrous ion adducts. (8). Notably, the two N-terminal cysteines (Cys 3 and Cys 4 ) within this segment are disulfide-linked in human IGFBP-6. It is therefore likely that the other N-terminal linkages in rat IGFBP-6 are the same as those in human IGFBP-6.
The N-terminal disulfide linkages of IGFBP-6 differ substantially from those of IGFBP-1, and, by implication, the other IGFBPs. The N-terminal disulfide linkages of IGFBP-6 are the sequential linkages Cys 1 -Cys 2 , Cys 3 -Cys 4 , and Cys 5 -Cys 6 , whereas in IGFBP-1, Cys 1 is not linked to Cys 2 and Cys 3 is not linked to Cys 4 . Although these observations suggest that a different local fold in the N-terminal region is likely, it is possible that this is not the case. For example, the presence of proline-containing sequences in the N-terminal regions of IGFBPs 1-5 could result in a similar fold to that stabilized by the disulfide linkages in IGFBP-6.
Human and rat IGFBP-6 lack the adjacent cysteine (CC) pair that is part of the GC 5 GC 6 C 7 motif present in the N-terminal domains of IGFBPs 1-5 (8). Although it is possible that these two adjacent cysteines in IGFBPs 1-5 are disulfide-linked, this is unlikely since disulfide linkage of adjacent cysteines requires a relatively rare cis peptide bond between the cysteines (23). Additionally, linkage of these adjacent cysteines would serve no obvious structural purpose, which seems inconsistent with their being conserved in all IGFBPs other than IGFBP-6. It is therefore probable that these cysteines are disulfide-linked to other N-terminal cysteines in IGFBPs 1-5.
While further investigation is clearly needed to determine the N-terminal disulfide linkages of IGFBP-1 (and IGFBPs 2-5), it is unlikely that this can be accomplished by peptide mapping techniques relying on proteolysis alone, due to lack of suitable cleavage sites in the conserved sequence C 5 GC 6 C 7 and elsewhere. Indeed, the N-terminal disulfide linkages of IGFBP-3 could not be determined by this approach (12). Appropriate cleavage sites could conceivably be introduced by site-directed mutagenesis. Peptide mapping methods could also be coupled with selective mutagenesis of cysteines in order to determine disulfide linkages one at a time, provided correct formation of remaining disulfide linkages was not affected.
The high affinity IGF binding site of IGFBP-5 has recently been localized to the sequence containing Cys 9 to Cys 12 of that protein (equivalent to Cys 7 to Cys 10 of IGFBP-6) (11). Since IGFBPs -1, -3, and -6 share the same disulfide linkages in this region as IGFBP-5 and there is substantial homology between the IGFBPs of critical hydrophobic amino acids involved in IGF binding, it is likely that this region confers high IGF binding affinity to all IGFBPs. It was suggested that the N-terminal domain of IGFBP-5 does not significantly interact with the high affinity binding site of IGFBP-5 nor influence IGF binding by this site (11). With IGFBP-6, however, the substantially different disulfide linkages of the N-terminal domain compared with those of IGFBP-1 (and presumably the other IGFBPs) could result in the stabilization of a different N-terminal domain structure which may be responsible for or contribute to the distinctive IGF binding characteristics of IGFBP-6. An alternate possibility is that the structures of the N-terminal domains of IGFBP-6 and other IGFBPs are folded in a similar way that is stabilized by quite different disulfide linkages. In that case, assuming the remainder of the structure also to be similar in IGFBPs 1-6, the IGF-II binding preference of IGFBP-6 could depend more on differences in the primary sequence compared with other IGFBPs than to differences in three dimensional structure, although both are likely to be influential.
It has been claimed that the C-terminal domain of the IGFBPs is also involved in IGF binding (11,24). Although many studies have suggested that the C-terminal domains of IGFBPs do not independently bind IGFs (11,25,26), the presence of this domain substantially increases IGF binding affinity, possibly by stabilizing the IGF⅐IGFBP complex once the IGF has bound. In contrast, two studies have shown that Cterminal fragments of IGFBP-2 bind IGFs with low affinity (27,28). Interestingly, deletion of the sequence containing the two C-terminal disulfide linkages eliminated the 12-fold IGF-II binding preference of bovine IGFBP-2 over IGF-I (9). IGFBP-6 and IGFBP-2 share a preference for IGF-II and have the same three C-terminal disulfide linkages. It has not been demonstrated that these linkages are different in the other IGFBPs, which lack substantial IGF-II binding preference; but, if present, such a difference in the C-terminal disulfide linkages (such as those connected to the conserved CWCV sequence) could conceivably influence IGF-II binding specificity. However, the sequences of IGFBPs 1-6 are homologous with respect to Cterminal cysteines, suggesting that their disulfide linkages are likely to be the same. Unfortunately, the peptide complex containing the four C-terminal cysteines of IGFBP-1, which would provide considerable information on this matter, was not recovered in the present study. In any case, it would be difficult to reconcile the IGF-II binding preference of the IGFBP-6/ IGFBP-5 chimera mentioned above (13) with a C-terminal determinant of IGF-II binding preference.
In conclusion, we have determined all of the disulfide linkages of human IGFBP-6. As suggested previously (10), the N-terminal linkages fall into two domains comprising Cys 1 to Cys 6 (the N-terminal domain) and Cys 7 to Cys 10 (the IGF binding domain), respectively. The six C-terminal cysteines of IGFBP-6 form three disulfide bonds, and there are no linkages between the N-and C-terminal regions. The N-terminal linkages differ significantly from those of IGFBP-1 and therefore,  Fig. 5) isolated as described in Fig.  5 by reverse-phase HPLC of hIGFBP-1 digested with trypsin and S. aureus V8 protease. The 5275-Da complex was initially identified as consisting of the four disulfide-linked peptides shown in Table II on the basis of the hIGFBP-1 sequence (Fig. 1), the specificities of trypsin (Arg-C/Lys-C) and S. aureus V8 protease (Glu-C/Asp-C), and the ESMS mass (Table II). Corresponding amino acid signals expected in each Edman cycle are compared with the observed signals from 16 cycles of Edman sequencing carried out as described under "Experimental Procedures." Multiple signals in respective Edman cycles are arranged vertically according to the assigned sequences, therefore some expected or observed signals are repeated for ease of comparison. Dashes (-) represent cysteine (not expected to be observed under Edman sequencing conditions) or expected amino acid signal changes that were not observed. Observed Edman signals that are underlined were equivocal due to carryover of the same signal from a previous cycle, C-terminal amino acid washout, or weak signals from serine or in cycles following proline. The Edman sequencing data confirm the ESMS-based identification (Table II). by inference, those of the other IGFBPs. In contrast, the remaining disulfide linkages of IGFBP-6 are the same as those reported in other IGFBPs; all of these linkages are therefore likely to be identical in IGFBPs 1-6. Knowledge of the disulfide linkages of IGFBP-6 will facilitate further structural studies of this protein as well as aiding in the rational design of mutants for structure-function correlations.