Synthesis and Characterization of Insulin-like Growth Factor (IGF)-1 Photoprobes Selective for the IGF-binding Proteins (IGFBPs) PHOTOAFFINITY LABELING OF THE IGF-BINDING ON

Elevated insulin-like growth factor (IGF)-1 levels are prognostic for the development of prostate and breast cancers and exacerbate the complications of diabetes. In each case, perturbation of the balance between IGF-1/2, the IGF-1 receptor, and the IGF-binding proteins (IGF-BPs) leads to elevated IGF-1 sensitivity. Blockade of IGF action in these diseases would be clinically significant. Unfortunately, effective IGF antagonists are currently unavailable. The IGFBPs exhibit high affinity and specificity for the IGFs and serve as natural IGF antagonists, limiting their mitogenic/anti-apoptotic effects. As an initial step in designing IGFBP-based agents that antag-onize IGF action, we have begun to analyze the structure of the IGF-binding site on IGFBP-2. To this end, two IGF-1 photoprobes, N a Gly1 -(4-azidobenzoyl)-IGF-1 (abG 1 IGF-1) and N a Gly1 -([2–6-(biotinamido)-2( p -azido-benzamido)hexanoamido]ethyl-1,3 * -dithiopropionoyl)-IGF-1 (bedG 1 IGF-1), selective for the IGFBPs were synthesized by derivatization of the a -amino group of Gly 1 , known to

Insulin-like growth factor (IGF) 1 -1 and IGF-2 play central roles in a number of cellular processes, including growth, proliferation, differentiation, survival, transformation, and metastasis (1,2). Enhanced activity of the IGFs has been implicated in diabetic complications and cancer. These effects are mediated by the IGF-1 receptor (IGF-1R), a member of the receptor tyrosine kinase family of cell-surface receptors. The IGF-2 receptor, which lacks signaling activity, plays a role in clearing IGF-2 from the cell surface (3,4). The IGFs are regulated at the extracellular level by a family of six IGF-binding proteins (IGFBPs), designated IGFBP-1-6 (5-7). These six proteins exhibit higher affinities for the IGFs than the IGF-1R, while having negligible affinity for insulin.
Renewed interest in the function of the IGF system stems from the observations that IGF-1 and IGF-2, acting through the IGF-1R, increase the tumorigenic potential of breast and prostate cancer cells (8). Accordingly, increased serum IGF-1 levels have been shown to be prognostic for the development of prostate and breast cancers (9,10). Alterations in IGFBP expression may also contribute to disease states. For example, IGFBP-3 is a target of the p53 tumor suppressor, and a common p53 mutation results in decreased IGFBP-3 secretion (11,12), which is likely to cause an increased proliferative response to IGF-1. Also, reduced IGFBP-2 expression resulting from the hyperglycemia of diabetes was recently shown to enhance the sensitivity of renal mesangial cells to the growth and secretory effects of IGF-1, pushing the cells toward a glomerulosclerotic phenotype (13). Because IGF-1 can suppress apoptosis, cells lacking IGF-1 receptors, cells with compromised IGF-1R signaling pathways, or cells treated with the IGFBPs may selectively die by apoptosis (8). Taken together, these findings suggest that the IGFBPs serve a role as natural IGF antagonists.
IGF-1 and IGF-2 are homologous protein hormones of 70 and 67 amino acids in length, respectively (14). Based on studies of chemically modified and mutated IGF-1, a number of residues have been identified as being part of the IGF-1R contact site, in particular the aromatic residues at positions 23-25 (15). Cooke et al. (16) used NMR and restrained molecular dynamics to elicit the solution structure of IGF-1; this model clearly illustrates an IGFBP-interacting domain on the surface of IGF-1 and its lack of overlap with the receptor-docking site (see Fig.  1). Specifically, this site consists of the N-terminal tripeptide Gly-Pro-Glu (17) and residues 49 -51 (18). These two regions come together to form an independent binding domain (see Fig.  1) (16). The analog des-1-3-IGF-1 binds to the IGF-1R with high affinity, but has dramatically reduced affinity for the IGFBPs, underscoring the importance of the N-terminal con-tact site on IGF-1 for binding protein specificity (17). In addition, mutations of Glu 3 or residues 49 -51 result in a ligand with severely reduced binding activity (19). In good agreement with these findings, insulin lacks these residues common in the IGFBP-binding region and thus does not bind to the IGFBPs with high affinity.
The IGFBPs are globular proteins containing 18 spatially conserved cysteine residues participating in the formation of nine disulfide bonds. They range in size from 200 to 300 amino acids and, based on their high degree of homology, can be divided into three distinct domains, each constituting about one-third of the protein (20). The N-and C-terminal regions, designated domains 1 and 3, respectively, share the highest homology (20,21), whereas the intervening region (domain 2) is highly variable (Ͻ30% homology) (6). Domains 1 and 3 contain 12 and 6 spatially conserved cysteine residues, respectively, except in the case of IGFBP-6, which is missing 2 cysteine residues in domain 1 (5); IGFBP-4 has 2 additional cysteine residues in domain 2. Because of the high homology of these proteins in domains 1 and 3 (ϳ70%) (6), the IGF-binding domain has been proposed to reside within one of these regions. Support for this notion is based on studies in which N-or C-terminal IGFBP fragments were found to retain high affinity binding activity for IGF-1 and/or IGF-2 (19). IGFBP-related protein-1 and -2, also known as IGFBP-7/Mac25 and IGFBP-8/ connective tissue growth factor, respectively, have homologies within their N termini to IGFBP-1-6 (23). These proteins have lower affinities for the IGFs compared with the IGFBPs and have been reported to interact with insulin (24). The precise mechanism by which the IGFBPs inhibit IGF-1 and IGF-2 action is presently unknown. It is thought to involve high affinity binding of the IGFs by the IGFBPs, thereby limiting their access to the IGF-1R. A complete understanding of this inhibitory action will come with the solution of the threedimensional structure of the IGFBPs. As of yet, the IGF-binding domain on the IGFBPs has not been defined.
To date, extensive use of molecular techniques has been applied to assess the binding domain on the IGFBPs. However, only sparse structural information has been obtained. Currently, a debate exists as to which domain(s) of the IGFBPs are most crucial for IGF binding. This is further confounded by the paucity of precise structural information about the IGFBPs as recently reviewed by Baxter (25). To more precisely identify the points of contact between IGF-1 and IGFBP-2, we chose to pursue a photoaffinity labeling approach, which has been used to identify sites of interaction between a number of interacting proteins. On this basis, we derivatized the ␣-amino group of the N-terminal glycine of IGF-1 with two different photoaffinity reagents, N-hydroxysuccinimidyl 4-azidobenzoate (HSAB) and sulfosuccinimidyl [2-6-(biotinamido)-2-(p-azidobenzamido)hexanoamido]ethyl-1,3Ј-dithiopropionate (SBED), to generate IGF-1 photoprobes capable of selectively labeling the IGF-binding domain on IGFBP-2. The advantage of this approach is that the photoreactive group is inserted within the region shown to be essential for high affinity binding of IGF-1 to the IGFBPs. Since IGF-2, which has a 3-residue N-terminal extension, exhibits a higher affinity for IGFBP-2 than IGF-1, reduced binding affinity of IGF-1 for IGFBP-2 resulting from N-terminal substitutions was not anticipated. In this study, we describe the synthesis and characterization of N ␣Gly1 -(4-azidobenzoyl)-IGF-1 (abG 1 IGF-1) and N ␣Gly1 -([2-6-(biotinamido)-2-(p-azidobenzamido)hexanoamido]ethyl -1, 3Ј-dithiopropionoyl)-IGF-1 (bedG 1 IGF-1) and their successful application to photoaffinity label rhIGFBP-2. Based on these direct photoaffinity labeling analyses, our results indicate that the C terminus of IGFBP-2 contains the IGF-binding domain.
Synthesis and Purification of abG 1 IGF-1 and bedG 1 IGF-1-All derivatizations and handling of photoprobes were carried out under subdued lighting or under a red safety light. abG 1 IGF-1 was synthesized by reaction of recombinant human IGF-1 (1 mg, 130 nmol) with a 10-fold molar excess of HSAB (0.34 mg, 1.3 mol) for 1 h at 23°C. Unreacted ester was quenched by the addition of 30 l of ethanolamine for 30 min, and the entire reaction mixture was lyophilized. The lyophilized reaction mixture was dissolved in 0.5 M acetic acid and 10 l of trifluoroacetic acid and injected onto a C 18 column equilibrated in 0.1% trifluoroacetic acid and 24% acetonitrile at a flow rate of 1 ml/min. After 20 min, a linear gradient of 24 -60% acetonitrile was developed over 60 min to elute IGF-1 and the reaction products. Peak 4 was collected, dried in vacuo in a SpeedVac concentrator (Savant Instruments, Inc., Farmingdale, NY), reinjected onto the C 18 column equilibrated in 50 mM triethanolamine phosphate (pH 3.0) containing 27.5% acetonitrile, and eluted with a gradient of 27.5-38% acetonitrile over 60 min. The major peak was collected, dried, and further analyzed. Synthesis and purification of bedG 1 IGF-1 was carried out essentially as described for abG 1 IGF-1 using a 1:1 ratio of SBED to protein.
Pepsin Digestion of abG 1 IGF-1 and bedG 1 IGF-1-IGF-1 and the photoprobes were digested with pepsin at an enzyme/substrate ratio of 1:20 in 0.01 M HCl for 5 h at 23°C (28). Digests were then injected onto a C 18 column equilibrated in 0.1% trifluoroacetic acid, and the fragments were eluted using a linear gradient of 0 -60% acetonitrile over 90 min. For each photoprobe, the HPLC fractions containing fragments with potentially modified residues (AE, CG, and D fragments) (see Fig. 3) (28) were analyzed by matrix-assisted laser desorption/ionization timeof-flight mass spectrometry (MALDI-TOF-MS).
MALDI-TOF-MS-Dried HPLC fractions were dissolved in 0.1% trifluoroacetic acid containing 70% acetonitrile. 0.5-l aliquots of each fraction were mixed with 1 l of 50 mM ␣-cyano-4-hydroxycinnamic acid (Sigma) in 0.1% trifluoroacetic acid and 70% acetonitrile. The mixture was spotted onto a gold-coated, stainless steel sample plate and airdried. The samples were analyzed using a PerSeptive Biosystems Voyager-DE MALDI-TOF mass spectrometer equipped with a 337-nm nitrogen laser. A delayed extraction source was operated in linear mode (1.2-m ion flight path, 20-kV accelerating voltage), yielding an instrumental resolution of ϳ700 (full width at half-maximum) at m/z 1297.5. One mass spectrum was based on 256 averaged mass scans. External mass calibration was performed using angiotensin I (1297.5 Da) and bovine insulin (5734.54 Da) as standards. Mass accuracy was Ϯ0.1%.
Purification of rhIGFBP-2-rhIGFBP-2 was purified from DHFR Ϫ Chinese hamster ovary cells (29) stably transfected with 30 g of pCMV-hIGFBP-2 and 2 g of pMT2 containing murine dihydrofolate reductase (30) using calcium phosphate. Cells were grown to confluence in roller bottles and subjected to a weekly cycle of 3 days growth in serum-containing medium, followed by 4 days in serum-free medium. Pooled conditioned medium (400 -600 ml) was acidified to pH 3-4 with glacial acetic acid and dialyzed against 10 mM acetic acid in Spectrapor 4 membranes (12,000 -14,000 M r cutoff; Spectrum, Laguna Hills, CA) to remove salt and IGFs. After lyophilization, IGFBP-2 was purified by sequential IGF-1-agarose affinity chromatography and reversed-phase HPLC on a C 4 column equilibrated in 0.1% trifluoroacetic acid. IGFBP-2 eluted between 30 and 40% acetonitrile.
IGF-1-Agarose Affinity Chromatography-2 ml of Affi-Gel 10-activated immunoaffinity support (Bio-Rad) were washed four times with 3 volumes of cold water, followed by the addition of 2 mg of IGF-1 in 100 mM HEPES (pH 7.4). The mixture was agitated overnight at 4°C, followed by two washes with 100 mM HEPES (pH 8.0). Unreacted sites were then quenched with 2 ml of 1 M Tris-HCl (pH 7.9) for 1 h at ambient temperature. The column was then washed with three cycles of alternating low pH/high pH salt washes (0.5 M NaCl and 0.1 M sodium acetate (pH 4.0) and 0.5 M NaCl and 0.1 M Tris (pH 8.0), respectively). Finally, the column was washed and stored in 10 ml of 150 mM NaCl and 50 mM HEPES (pH 7.4) with 0.05% sodium azide.
Dialyzed and lyophilized conditioned medium was dissolved in 40-45 ml of 50 mM HEPES (pH 7.4) containing 150 mM NaCl (Buffer A). Insoluble material was removed by centrifugation at 3000 ϫ g for 20 min. 2 ml of IGF-1-agarose in Buffer A were then added, and the slurry was incubated overnight at 4°C with gentle agitation. The column was subsequently washed with 100 ml of Buffer A, followed by 50 ml of 10% Buffer A. Proteins bound to the column were then eluted with 15 ml of 0.5 M acetic acid, and the column was washed with 20 ml of Buffer A plus 1 ml of 1 M HEPES (pH 7.4), followed by 40 ml of Buffer A. The eluate was dried in vacuo using a SpeedVac concentrator. The dried eluate was stored at Ϫ20°C until further purification as described above.
IGFBP-2 Binding Assay-Soluble IGFBP-2 binding assays were carried out using polyethylene glycol precipitation and centrifugation (31). 1 ng of rhIGFBP-2 was combined with various concentrations of IGF-1, abG 1 IGF-1, or bedG 1 IGF-1 ranging from 30 fM to 100 nM in binding assay buffer (100 mM HEPES (pH 7.4), 44 mM NaHCO 3 , 0.01% bovine serum albumin, 0.01% Triton X-100, and 0.02% NaN 3 ), followed by the addition of 10 nCi of 125 I-IGF-1 (Amersham Pharmacia Biotech). After a 4-h incubation at room temperature, 250 l of 0.5% bovine ␥-globulin were added, followed by 500 l of 25% polyethylene glycol (average M r of 8000; Sigma). The samples were incubated for 10 min at room temperature and centrifuged for 3 min at 15,000 ϫ g. The pellets were washed with 1 ml of 6.25% polyethylene glycol, and bound radioactivity was quantified in a Compugamma spectrometer (LKB-Wallac, Turku, Finland). Counts bound in the presence of 1 M or 100 nM IGF-1 (nonspecific binding) were subtracted to obtain specific binding. IC 50 values were calculated using the equation where B is the concentration of bound ligand and B max is the maximal binding observed. The Microsoft Excel 97 Solver was used to minimize the sum of the squares of the differences from the mean IC 50 values for each IGF-1 concentration by optimizing B restrained by the above equation. The calculated IC 50 values were used to generate smooth curves.
Trypsinization of IGF-1, IGFBP-2, and Photoaffinity-labeled Complexes-Trypsinization of reduced and alkylated proteins was performed according to Honegger and Humbel (34). Proteins (20 M) were dissolved in 100 mM N-ethylmorpholine acetate (pH 8.5) to which was added sufficient trypsin (1 g/l) to achieve a 1:50 enzyme/substrate ratio. Mixtures were then incubated for 2 h at 37°C, followed by a second addition of trypsin. After 2 additional h at 37°C, the reaction was stopped by the addition of an equal volume of 0.1% trifluoroacetic acid. The mixture was dried in vacuo or applied directly to a C 18 reversed-phase HPLC column.
Avidin Chromatography of bedG 1 IGF-1⅐IGFBP-2 Tryptic Peptides-Tryptic peptides generated from the bedG 1 IGF-1⅐IGFBP-2 complex were applied to an UltraLink TM monomeric avidin column. Flowthrough fractions were collected and pooled. The column was eluted with low pH buffer, and the eluate fractions were pooled. The flowthrough and eluate fractions were dried and further analyzed by reversed-phase HPLC on a C 18 column equilibrated in 0.1% trifluoroacetic acid and eluted with a linear gradient of acetonitrile. Eluted peaks were analyzed by MALDI-TOF-MS.
Immunoblots-Samples were dissolved in SDS sample buffer with or without dithiothreitol and resolved on a 10 or 12.5% SDS-polyacrylamide gel according to the procedure of Laemmli (35) using a Hoefer Scientific Instruments apparatus. The proteins were transferred to nitrocellulose and immunoblotted using a commercial antiserum against intact bovine IGFBP-2 (Upstate Biotechnology, Inc., Lake Placid, NY). Blots were developed with horseradish peroxidase-labeled secondary antibody (Chemicon International, Inc., Temecula, CA) and an enhanced chemiluminescence kit (Amersham Pharmacia Biotech). For sequential anti-IGF-1/anti-IGFBP-2 antibody analysis of the same blot, the membrane was stripped for 20 min at 60°C in 1 M Tris (pH 6.7) containing 10% SDS and 0.1 M ␤-mercaptoethanol. The stripped membrane was washed twice with Tris-buffered saline containing Tween for 10 min at 23°C, followed by blocking for 1 h with 5% nonfat dry milk and reprobed. Fig. 1, the three-dimensional structure of IGF-1 reveals the presence of distinct binding domains for the IGFBPs and the IGF-1R, which do not overlap. IGF-1 contains four primary amines (the ⑀-amino groups of 3 lysyl residues and the ␣-amino group of Gly 1 ), all of which are reactive with HSAB and SBED. Fig. 2 shows the structures of these reagents and highlights the relative lengths of the spacer arms from the site of covalent linkage to IGF-1 to the photoreactive azide group. In addition to a significantly longer spacer arm, SBED also contains a reduction-sensitive disulfide linkage and a biotin moiety. As a result, although we anticipated that use of HSAB as a photocross-linking agent would yield a reductionstable covalent IGF-1⅐IGFBP-2 complex, use of SBED was expected to produce a reduction-sensitive IGF-1⅐IGFBP-2 complex that, after reduction, would result in biotinylation of IGFBP-2 at the site of photoincorporation.

Synthesis and Purification of abG 1 IGF-1 and bedG 1 IGF-1-As shown in
As shown in Fig. 2, reaction of HSAB and SBED with IGF-1 as described under "Experimental Procedures" resulted in production of three major products (peaks 2-4) in addition to unreacted IGF-1 (peak 1). Analysis of the HSAB reaction products by acidic polyacrylamide gel electrophoresis in 8 M urea and by MALDI-TOF-MS revealed that these three peaks were monoderivatized forms of IGF-1 (data not shown). Peak 2 had previously been shown to represent the elution position of abK 27 IGF-1 (where K 27 is Lys 27 ) (36). Peaks 3 and 4 had been tentatively identified as abK 65 IGF-1 and abG 1 IGF-1, respectively, based on amino acid sequencing analysis (36). No abK 68 IGF-1 was detected. The three product peaks were further purified by chromatography on a C 18 column equilibrated in triethanolamine phosphate and acetonitrile. Structural assignments were confirmed using pepsin digestion as described below. After derivatization with SBED, peaks 2-4 were shown to similarly represent bedK 27 IGF-1, bedK 65 IGF-1, and bedG 1 IGF-1 (Fig. 2B).
Characterization of abG 1 IGF-1 and bedG 1 IGF-1-To verify that the "ab" and "bed" moieties were covalently bound to the ␣-amino group of Gly 1 for each photoprobe, it was necessary to isolate Gly 1 from the 3 Lys residues. Since reduction and alkylation might disrupt the azide moiety, we utilized pepsin digestion of nonreduced abG 1 IGF-1 and bedG 1 IGF-1. As reported by Forsberg et al. (28) and shown in Fig. 3, pepsin releases a disulfide-linked AE fragment containing Gly 1 and no Lys residues. Pepsin digestion of abG 1 IGF-1 and bedG 1 IGF-1 was FIG. 1. IGFBP-binding site on IGF-1. Left, the N-terminal residues Gly 1 , Pro 2 , and Glu 3 combine with Phe 49 , Arg 50 , and Ser 51 to form the IGFBP-binding domain (6). Note that the Gly 1 ␣-NH 2 group is the site chosen for derivatization with aryl azide-based photoaffinity reagents. Right, the IGF-1R-binding domain on IGF-1. The image was obtained by RasMol Version 2.6 using coordinates (6)

IGF-1 N-terminal Contact Sites on IGFBP-2
carried out in the dark to avoid photoactivation of the probe, and recombinant human IGF-1 was also digested in parallel to serve as a control for subsequent HPLC and mass spectrometric analyses.
For each photoprobe, HPLC purification of the pepsin digestion products revealed that the retention times for the derivatized AE fragments were significantly increased compared with the underivatized AE fragment (data not shown). This was predicted based on the added hydrophobicity of the additional functional groups. Fig. 4A shows the MALDI-TOF-MS of abG 1 AE. This fragment had the correct mass (predicted average mass of 2419.7 Da; observed mass of 2419.8 Da) and exhibited loss of nitrogen as a result of photoactivation by the 337-nm laser, to yield a second peak at 2394.3 Da. The identities of the CG and D fragments, containing Lys 65 /Lys 68 and Lys 27 , respectively, were also confirmed by MALDI-TOF-MS and were in good agreement with those of Forsberg et al. (28). Similar observations were made for bedG 1 IGF-1. As shown in Fig. 5, a peak at 2956.2 Da (predicted average mass of 2955.9 Da) was observed along with a second peak at 2929.4 Da, reflecting the loss of nitrogen resulting from photolysis. Again, the CG and D fragments exhibited masses consistent with a lack of derivatization. As an alternate method of validation of the derivatization, electrospray ionization MS 3 was also carried out (data not shown).
Having demonstrated the structure and purity of abG 1 IGF-1 and bedG 1 IGF-1, it was necessary to ensure that each probe retained high affinity binding for IGFBP-2 similar to that of native IGF-1. The results of IGFBP-2 competition binding assays are shown in Fig. 6. The IC 50 values for IGF-1, abG 1 IGF-1, and bedG 1 IGF-1 were 173, 131, and 168 pM, respectively, indicating that the photoprobes and IGF-1 had identical affinity for IGFBP-2. These data indicate that the addition of the photoreactive group does not significantly hinder the IGF-1/IGFBP-2 interaction.
Photoaffinity Labeling of IGFBP-2-Photoaffinity labeling of IGFBP-2 with abG 1 IGF-1 and bedG 1 IGF-1 was carried out as described under "Experimental Procedures." In a small-scale photolabeling study with abG 1 IGF-1, an aliquot of the reaction mixture was subjected to immunoblot analysis with anti-IGFBP-2 and anti-IGF-1 antisera under reducing conditions (Fig. 7A). In addition to rhIGFBP-2 migrating at ϳ32 kDa and detected by anti-IGFBP-2 antiserum, a second protein species migrating at ϳ40 kDa was detected. This band reacted with both anti-IGF-1 and anti-IGFBP-2 antibodies, indicating that the photoaffinity labeling reaction had successfully generated abG 1 IGF-1⅐IGFBP-2 complexes. These data confirm that photolabeling of IGFBP-2 with abG 1 IGF-1 results in a stable covalent IGF-1⅐IGFBP-2 complex that is not disrupted by reduction.
A similar analysis was carried out for bedG 1 IGF-1 (Fig. 7B). As shown in lane 1, two bands were detected with anti-IGFBP-2 antiserum: IGFBP-2 itself and the bedG 1 IGF-1⅐IGF-BP-2 complex. This complex bound to and was specifically eluted from an avidin-agarose column, whereas IGFBP-2 did not bind at all (lane 2). Under nonreducing conditions, NeutrAvidin TM -peroxidase was unable to detect the bedG 1  To verify derivatization of IGF-1 at its N-terminal glycine residue, purified probes were digested with pepsin. The primary sequence of IGF-1 is shown, with arrowheads indicating the primary sites of pepsin cleavage (28). Letters A-G indicate the principal peptides generated. The disulfide-linked AE fragment (shaded) contains the N-terminal glycyl residue and none of the lysyl residues. complex was first reduced and alkylated with tris(2-carboxyethylyl)phosphine and 4-vinylpyridine prior to electrophoresis. As discussed above, reduction and alkylation of bedG 1 IGF-1⅐ IGFBP-2 should yield IGFBP-2 biotinylated at the site of photoincorporation. As shown in lanes 4 -6, two bands were again detected with reduction and alkylation of the samples. The upper band had the same electrophoretic mobility as uncrosslinked, reduced, and alkylated IGFBP-2 and was detectable with anti-IGFBP-2 antibodies. However, roughly one-half of the content of this band was retained on the avidin-agarose column (lane 5) and reacted with NeutrAvidin TM -peroxidase (lane 6), indicating that the band in lane 4 contains both unreacted and biotinylated IGFBP-2. The lower band reacted with anti-IGFBP-2 antibodies, was retained by the avidinagarose column (lane 5), and could be labeled with NeutrAvidin TM -peroxidase (lane 6), indicating that it represents a biotinylated fragment of IGFBP-2. These results indicate that reduction of the bedG 1 IGF-1⅐IGFBP-2 complex allowed greater access of avidin-peroxidase to the biotin moiety on photolabeled IGFBP-2.
Identification of the abG 1 IGF-1 Photoincorporation Site-To identify the site of photoincorporation of abG 1 IGF-1, the abG 1 IGF-1⅐IGFBP-2 complex was isolated by HPLC following reduction and alkylation of the photolysis reaction mixture. Owing to the similar retention times of the free and crosslinked species, the two components overlapped significantly (Fig. 8A). Subsequent re-chromatography of each component provided sufficient resolution to attain a high level of purification of the abG 1 IGF-1⅐IGFBP-2 complex (Fig. 8B). Immunoblot analysis of the column fractions indicated that Ͼ40% photoincorporation was achieved.
Fractions containing purified abG 1 IGF-1⅐IGFBP-2 were pooled, dried, and trypsinized. It was anticipated that tryptic digestion of the abG 1 IGF-1⅐IGFBP-2 complex would yield a cross-linked peptide containing a tryptic fragment of IGFBP-2 (BP2T) and the N-terminal tryptic fragment of abG 1 IGF-1 (abIGF1T). Since abIGF1T alone (residues 1-21) is 2521 Da, this represents the minimum mass for the resulting crosslinked peptide (BP2T-abIGF1T). The tryptic digest was applied to a C 18 column and eluted with a shallow gradient of acetonitrile to obtain optimal separation of the tryptic peptides (data not shown). MALDI-TOF-MS was then performed on each fraction to locate BP2T-abIGF1T. The MALDI mass spectrum of the identified fraction is shown in Fig. 9A. Although not pure, we obtained a significant enrichment of BP2T-abIGF1T in this fraction with an observed mass of 5295.6 Da. These results suggest that abIGF1T (predicted mass of 2521 Da) was co-valently incorporated into the C-terminal tryptic peptide of IGFBP-2 corresponding to residues 266 -287 (BP2T predicted mass of 2772 Da), yielding a complex with a predicted mass of 5293 Da. To further validate this assignment, the fraction containing this complex was subdigested with Staphylococcus aureus V8 protease and analyzed by MALDI-TOF-MS. As illustrated in Fig. 9 (B and C), MALDI-TOF-MS of this fraction identified four distinct peptides consistent with the proposed structure. These data indicate that when attached to the ␣amino group of Gly 1 , the azidobenzoyl moiety contacts IGFBP-2 in its distal C-terminal end within the tryptic peptide corresponding to residues 266 -287 (see Fig. 11A).
Identification of the bedG 1 IGF-1 Photoincorporation Site-To determine the site of photoincorporation of bedG 1 IGF-1 into IGFBP-2, the photolyzed reaction mixture obtained as described under "Experimental Procedures" was reduced and alkylated with tris(2-carboxyethylyl)phosphine and 4-vinylpyridine, resulting in cleavage of IGF-1 from the complex and biotinylation of IGFBP-2 at the site of photoincorporation. The mixture was then desalted on a C 4 column, dried in vacuo, and digested with trypsin. The tryptic peptides were loaded onto an UltraLink TM monomeric avidin column, which was washed with 6 column volumes of phosphate-buffered saline. To release retained biotinylated tryptic peptides, the column was eluted with regeneration buffer (0.2 M glycine (pH 2.8)). The flow-through/washes and eluate fractions were pooled separately and dried, and the peptides present in each were resolved by HPLC on a C 18 column. As expected, the majority of peptides eluted in the flow-through fraction of the avidin column (Fig. 10A), resembling a representative combined tryptic digest of uncross-linked IGF-1 and IGFBP-2 (data FIG. 6. Competition binding analyses of N ␣Gly1 -derivatized photoprobes. The binding activities of the IGF-1 photoprobes were analyzed with a polyethylene glycol precipitation-based assay using IGF-1 (q), abG 1 IGF-1 (f), or bedG 1 IGF-1(Ⅺ) to compete with 125 I-IGF-1 for binding to IGFBP-2 (see "Experimental Procedures"). Data points shown are average values of two independent experiments performed in triplicate. Binding curves were generated as described under "Experimental Procedures. "   FIG. 7. Analysis of photoaffinity-labeled complexes. A, abG 1 IGF-1-labeled complexes were resolved on an SDS gel and transferred to nitrocellulose. The blot was probed with anti-IGFBP-2 antibodies, followed by stripping and reprobing with anti-IGF-1 antibodies. B, products of photoaffinity labeling with bedG 1 IGF-1 were immunoblotted with a polyclonal anti-IGFBP-2 antibody or probed with Neu-trAvidin TM -peroxidase (Av-HRP) as indicated. Some aliquots were applied to an UltraLink TM avidin-agarose (Av-Ag) column; the low pH eluate was run on the gel. Red/Alk, reduction and alkylation. not shown). Two major peaks were observed in the low pH elution of the column. The first peak exhibited a mass of 2595.7 Da when analyzed by MALDI-TOF-MS (Fig. 10B). When corrected for the mass of the biotin and remaining cross-linker residues from the bed moiety (653.3 Da), this corresponded to a BP2T fragment of 1942.3 Da (Fig. 10B). This coincides with tryptic peptide 212-227 (predicted mass of 1942.2 Da) from the C terminus of IGFBP-2. MALDI-TOF-MS analysis of the second peak eluted from the avidin column revealed the presence of a major signal at 1667.4 Da. When corrected for cross-linker/ biotin, the BP2T fragment had a mass of 1014.1 Da, which did not correspond to any tryptic peptides present in rhIGFBP-2 (Fig. 10B). However, this mass did correspond to the first 9 residues of the BP2T fragment already identified (residues 212-220). These findings suggest that photoaffinity labeling with the bed reagent resulted in partial peptide bond cleavage at the site of covalent insertion. The potential for peptide bond cleavage during photoaffinity labeling has previously been suggested in the literature (37). This conclusion is further sup-ported by the data presented in Fig. 7B (lanes 4 -6), where we detected an IGFBP-2 fragment that was reactive with NeutrAvidin TM -peroxidase. Taken together, these results suggest that the site of insertion may be residue 220 of IGFBP-2 (Fig. 11A). DISCUSSION In this report, we present evidence for a C-terminal contact site on IGFBP-2 for IGF-1 based on direct photoaffinity labeling studies with two unique N-terminally modified photoaffinity derivatives of IGF-1. We interpret these findings as an indication of the presence of a high affinity binding site for IGF-1 within the C terminus of IGFBP-2. Consistent with this finding are a number of reports describing the C terminus of IGFBP-2 as the domain containing the high affinity IGF-binding site. Wang et al. (38) and Ho and Baxter (39) each identified C-terminal fragments of IGFBP-2 with high affinity IGF-binding activity. In preliminary studies, we have characterized a 15.8-kDa C-terminal fragment of rhIGFBP-2 from our transfected Chinese hamster ovary cell cultures that exhibits an affinity for IGF-1 identical to that of intact rhIGFBP-2. 2 This fragment lacks N-terminal and mid-region epitopes based on tryptic peptide mapping studies. Additional support for the C terminus as the site of IGF binding stems from studies of Brinkman et al. (40), who deleted the last 20 amino acids from the C terminus of IGFBP-1 and thereby abolished IGF-1-binding activity. Similarly, Forbes et al. (41) generated a series of four sequential C-terminal truncation mutants of bovine IGFBP-2 and concluded that residues 222-236 are required for high affinity IGF interactions. Finally, Schuller et al. (42) prepared monoclonal antibodies to residues 188 -196 and 222-227 of IGFBP-1 that blocked IGF-1 binding. These investigators concluded that the regions surrounding these epitopes were important for IGF binding.
Introduction of photoactivable aryl azide moieties within the IGFBP-binding domain on IGF-1 (residues 1-3 and 49 -51) resulted in the selective photoaffinity labeling of two separate sites in the C terminus of IGFBP-2, within tryptic peptides 266 -287 (abG 1 IGF-1) and 212-227 (bedG 1 IGF-1) (Fig. 11A). The labeling of these different sites can be attributed to the different side chain lengths of the two IGF-1 photoprobes. abG 1 IGF-1 contains an azidobenzoyl moiety and thus lacks an appreciable side chain between the ␣-amino group of Gly 1 and the aryl azide. In this case, the aryl azide resides near the outer edge of the IGF-1-binding domain, defining the site of contact for Gly 1 . The longer, more flexible side chain present on the aryl azide in bedG 1 IGF-1 has the potential to define a contact site within the vicinity of the other residues of the IGFBPbinding domain, including Glu 3 , which plays an essential role in maintaining high affinity binding to the IGFBPs (43). Although we cannot rule out the possibility that this side chain results in the labeling of regions of the IGFBPs outside the IGF-binding domain, we believe that the likelihood of this occurring is minimal based on the previous identifications of C-terminal binding activity described above. Given the globular nature of the IGFBPs and the disulfide bonding pattern in this domain (Fig. 11B) (41), these two sites are likely to be closely apposed in three-dimensional space. Although the two photoprobes labeled sites within the C terminus of IGFBP-2 that are separated by ϳ40 amino acids in the primary sequence, these sites are likely to be much closer when the secondary structure of the protein is taken into consideration (Fig. 11B). Indeed, the two labeled sites cannot be farther apart than the total length of the two photoprobe spacer arms combined (ϳ300 nm).
Based on chemical modifications, site-directed mutagenesis 2 M. J. Horney and S. A. Rosenzweig, manuscript in preparation.
FIG. 8. Purification of IGF-1⅐IGFBP-2 complexes. A, following reduction and alkylation, the abG 1 IGF-1/IGFBP-2 photolysis reaction mixture was purified on a C 18 column equilibrated in 0.1% trifluoroacetic acid and 24% acetonitrile. B, the major components (a-c) of the elution profile obtained in A were re-chromatographed separately using a shallow acetonitrile gradient. Insets, fractions from each run were analyzed by immunoblotting with anti-IGFBP-2 antibody. B, IGFBP-2; BI, IGF-1⅐IGFBP-2 complex.

FIG. 9. Analysis of abG 1 IGF-1⅐IGF-BP-2 photoaffinity-labeled complex.
A, shown is a MALDI-TOF mass spectrum of a unique peak having a mass (5295.6 Da) that lacks correspondence to IGF-1 or IGFBP-2 tryptic peptides. B, the column fraction containing the 5295.6-Da fragment was dried and subdigested with S. aureus V8 protease. Shown is the MALDI-TOF mass spectrum of the digestion products. C, the peptides identified in the V8 digest in B are indicated.
FIG. 10. Analysis of bedG 1 IGF-1⅐IGFBP-2 photoaffinity-labeled complex. A, elution profile of the reduced, alkylated, and trypsin-digested photolysis reaction mixture following chromatography on a column of monomeric avidin. Shown are the HPLC column profiles of the column flow-through/washes and eluate. B, MALDI-TOF mass spectrum of the peaks identified by arrows in A. The first peak exhibited a mass of 2595.7 Da. After correcting the observed mass for the associated biotin and spacer, it was identified as tryptic peptide 212-227. The second biotinylated component (second arrow) was identified as having a mass of 1014.1 Da, which corresponds to a truncated form of the same tryptic peptide (residues 212-220) and suggests rupture of the peptide backbone of IGFBP-2 during photoincorporation. This finding explains the composition of the smaller band observed in Fig. 7B (loss of ϳ7 kDa; residues 221-289). Z, pyridylethylated cysteine. studies, and the identification of fragments with IGF-binding activity, the N terminus of the IGFBPs has also been described as the site of the IGF-binding domain. Iodination studies on bovine IGFBP-2 in isolation or as a complex bound to IGF-2 resulted in efficient iodination of Tyr residues at positions 71, 98, 213, 226, and 269; Tyr 60 was protected from iodination by bound IGF-2 (44). Huhtala et al. (45) isolated a 21-kDa Nterminal fragment of IGFBP-1 that retained some IGF-1-binding activity. More recently, Kalus et al. (46) reported that N-terminal fragments of IGFBP-5 representing residues 1-104 and 40 -92 exhibit weak IGF-binding activity. Based on NMR studies, they defined a hydrophobic patch comprising residues 49, 50, 62, and 68 -75 that potentially represents the primary IGF-binding site on IGFBP-5. This hypothesis was subsequently tested by the construction of full-length IGFBP-5 and IGFBP-3 mutants in which combined substitutions at residues 68, 69, 70, 73, and 74 resulted in a Ͼ1000-fold reduction in binding affinity (47). Contrary to these mutagenesis results, Stä ndker et al. (48) reported that the C terminus of IGFBP-5 contains the IGF-binding domain based on the isolation of a naturally occurring, biologically active truncation of IGFBP-5. The observation by Yamanaka et al. (24) that insulin can bind to (and be chemically cross-linked into) the N terminus of IGFBP-related protein-1 may provide further insight into the mechanism of IGF/IGFBP interactions, suggesting the notion that the N-and C-terminal domains may both interact with IGF-1, but with differing specificity and function (49).
In the context of the present findings, we propose the following model (Fig. 12) based on the idea that two binding sites for IGF-1 exist on the IGFBPs. The first is a high affinity, high specificity site (C terminus) responsible for binding selectivity for IGF-1 via its IGFBP-binding domain. The second low affinity site (N terminus) binds to IGF-1 via its IGF-1R-binding domain and thus plays a role in blocking IGF-1 binding to the IGF-1R. This model would likely require a conformational change to take place within the N terminus following IGF binding to the C terminus. Evidence to support this comes from NMR analyses of IGF⅐IGFBP complexes (50,51). It is possible to speculate that such conformational changes may reflect the ability of the IGFBP, following initial binding to the N terminus (IGFBP-binding domain) of IGF-1, to interact with the IGF-1R domain of IGF-1. This could provide a steric mechanism through which inhibition of IGF-1R activation is accomplished. This in turn explains low affinity insulin binding to the N terminus of IGFBP-related protein-1 (9). In addition, a conformational change explains why des-1-3-IGF-1 does not bind the IGFBPs well, even though it still contains an intact IGF-1R-binding domain.
This scheme may serve to explain how N-terminal fragments and truncation mutants of the IGFBPs are able to bind to IGF-1. This would also support a steric hindrance-based model of IGFBP inhibition of IGF-1 action at the IGF-1R. Following high affinity binding between the IGFBP-binding domain on IGF-1 and its contact site within the C terminus of IGFBP-2, the ensuing conformational rearrangement enables high affinity interaction of the N terminus of IGFBP-2 with the IGF-1R binding domain on IGF-1. Obviously, this is one of several potential models one might propose; confirmation of any such model will require a more definitive understanding of the three-dimensional structure of the IGFBPs. Future photolabel- FIG. 11. Sites of photoincorporation into IGFBP-2. A, depiction of sites of photoincorporation of abG 1 IGF-1 and bedG 1 IGF-1 into IGFBP-2. White lines represent sites of trypsin cleavage. B, detailed view of the C-terminal region (residues 210 -289) showing the known disulfide bonding pattern (41). Underlined peptides represent tryptic peptides identified as sites of photoincorporation.

FIG. 12. Model of IGF-1/IGF-1R binding inhibition by IGFBP-2.
Based on the data presented, the IGFBP-binding domain on IGF-1 interacts with the C terminus of IGFBP-2. Once this complex is formed, the IGF-1R-binding domain on IGF-1 then binds to the N terminus of IGFBP-2. IGFBP-rps, IGFBP-related proteins.
ing studies with abK 27 IGF-1 and bedK 27 IGF-1 may aid in addressing the issue of whether there are N-terminal contact sites on the IGFBPs for the IGF-1R-binding domain on IGF-1.
Based on their ability to block IGF actions, we have chosen to pursue the design of IGF antagonists based on the structure of the IGF-binding domain on the IGFBPs. This clearly requires more detailed information than is currently available concerning IGFBP structure. We have employed a photoaffinity labeling approach as a means of identifying the IGF-1 site of contact on IGFBP-2. To this end, a photoreactive derivative of IGF-1 was prepared, exploiting the reactivity of primary amines in the protein to covalently attach an aryl azide moiety within the IGFBP-binding domain. This approach has been used for many photoaffinity labeling studies, including the insulin (52) and IGF-1 (36) receptors. It has the drawback of incorporating a bulky group into the ligand, which may alter its binding characteristics. However, in the case of abG 1 IGF-1 and bedG 1 IGF-1, we did not observe a significant alteration in binding affinity for IGFBP-2. An alternative approach would be to utilize an intrinsically labeled IGF-1 derivative containing a photoactivable moiety at an aromatic amino acid. This has been reported for the identification of an insulin contact point on the insulin receptor (22). To accomplish this, Phe B25 in the receptor-binding domain of insulin was replaced with p-azidophenylalanine to generate (B25-p-azidophenylalanine-␣-carboxamide)-insulin. To apply a similar strategy to studies with IGF-1 would require considerable peptide semi-syntheses. Instead, we chose the alternative of incorporating an extrinsic photoactivable moiety into the ␣-amino group of Gly 1 , a residue that constitutes part of the IGFBP-binding domain on IGF-1. Of the two photoprobes employed, the use of SBED to biotinylate the site of IGF-1 photoincorporation provided a more direct approach to the identification of the photocross-linked tryptic peptide. This represents the first report of the successful use of this reagent in photolabeling a ligand-binding site.
In conclusion, this study represents the first report defining a contact site between IGF-1 and IGFBP-2. We propose to exploit this structural information to develop unique antagonists of the IGFs as important therapeutic adjuncts in the prevention/treatment of various cancers and the complications of diabetes. Small peptides comprising the proposed IGF-binding site would serve as stable, protease-resistant analogs of the IGFBPs (IGFBP mimetics). These compounds will be designed to block IGF action and will serve as important templates for the future design of peptidomimetics. To this end, IGF-1 antagonists should provide important therapeutic adjuncts in the prevention/treatment of these diseases.