Characterization of Glycosaminoglycan-binding Domains Present in Insulin-like Growth Factor-binding Protein-3*

Matrix metalloproteinase 3 cleaves insulin-like growth factor-binding protein-3 (IGFBP-3) into six fragments, four of which bind heparin-Sepharose (Fowlkes, J. L., Enghild, J. J., Suzuki, K., and Nagase, H. (1994) J. Biol. Chem. 269, 25742–25746). Sequence analysis of IGFBP-3 heparin-binding fragments shows that all fragments contain at least one of two highly basic, putative heparin-binding consensus sequences present in IG-FBP-3. Epitope-specific antibodies generated against synthetic peptides containing these domains recognized IGFBP-3, yet were significantly inhibited from binding in the presence of heparin, demonstrating that these regions of IGFBP-3 contain functional heparin-binding domains. IGFBP-3 peptides containing one of the two heparin-binding consensus sequences bound heparin in a solid phase binding assay in a dose-dependent and saturable manner. However, the IGFBP-3 peptide containing the heparin-binding consensus sequence 149 KKGHA 153 bound heparin with (cid:59) 4-fold less affinity than the IGFBP-3 peptide containing the longer heparin-binding consensus sequence 219 YKKKQCRP 226 . Examination of several well characterized glycosaminoglycans to inhibit the binding of heparin to both heparin-binding IGFBP-3 peptides revealed that the most potent inhibitors were heparin, heparan sulfate, and dermatan sulfate; chondroitin sulfate A and hyaluronic acid were 95% pure by high pressure liquid chromatography, and sequence verification was performed by electrospray mass spectrometry. Production, Purification, and Characterization of Epitope-specific Antibodies— Three mg of each peptide was conjugated to keyhole limpet hemocyanin using m -maleimidobenzoyl- N -hydroxysuccinimide, utiliz-ing the terminal –SH group of each peptide for conjugation. The pep- tide-conjugate was mixed with an equal volume of complete Freund’s adjuvant and injected in a pair of New Zealand White rabbits. Repeat injections were performed on five separate occasions over (cid:59) 105 days. Crude antiserum contained high titer antiserum to each peptide when compared with preimmune serum as assessed by an enzyme-linked immunosorbent assay using BSA-conjugated peptide in the solid phase.


Matrix metalloproteinase 3 cleaves insulin-like growth factor-binding protein-3 (IGFBP-3) into six fragments, four of which bind heparin-
. Sequence analysis of IGFBP-3 heparin-binding fragments shows that all fragments contain at least one of two highly basic, putative heparin-binding consensus sequences present in IG-FBP-3. Epitope-specific antibodies generated against synthetic peptides containing these domains recognized IGFBP-3, yet were significantly inhibited from binding in the presence of heparin, demonstrating that these regions of IGFBP-3 contain functional heparin-binding domains. IGFBP-3 peptides containing one of the two heparin-binding consensus sequences bound heparin in a solid phase binding assay in a dose-dependent and saturable manner. However, the IGFBP-3 peptide containing the heparin-binding consensus sequence 149 KK-GHA 153 bound heparin with ϳ4-fold less affinity than the IGFBP-3 peptide containing the longer heparin-binding consensus sequence 219 YKKKQCRP 226 . Examination of several well characterized glycosaminoglycans to inhibit the binding of heparin to both heparin-binding IGFBP-3 peptides revealed that the most potent inhibitors were heparin, heparan sulfate, and dermatan sulfate; chondroitin sulfate A and hyaluronic acid were intermediate in their inhibitory activities; and chondroitin sulfate C caused no inhibition. These studies identify and characterize the glycosaminoglycan-binding domains in IGFBP-3, providing a basis for the better understanding of IGFBP-3-glycosaminoglycan interactions at the cellular and extracellular interface.
Insulin-like growth factor (IGF) 1 -binding proteins (IGFBPs) are a group of six homologous, high affinity carrier proteins for IGF-I and IGF-II, which are produced by a wide variety of cells and tissues as soluble proteins (for recent reviews see Refs. [1][2][3][4]. However, at least two IGFBPs, IGFBP-3 and IGFBP-5, have been shown to associate with cell surfaces and/or extracellular matrix (1)(2)(3)(4). This association may facilitate the localization of IGF-IGFBP complexes into close proximity of IGF receptors, thereby enhancing IGF bioavailability. The mechanisms by which IGFBP-3, the major carrier of serum IGFs, interacts with cell surfaces and/or extracellular matrix remain unclear. Oh et al. (5) have provided evidence that IGFBP-3 may interact with a specific cell-surface receptor, yet other data suggest that IGFBP-3 may interact with glycosaminoglycan (GAG)-containing molecules (i.e. proteoglycans) present at the cell surface and/or in extracellular matrix (6,7). IGFBP-3 binds avidly to heparin-Sepharose (8 -10), and we have recently reported that proteolysis of recombinant human (rh) IGFBP-3 by matrix metalloproteinase 3, an IGFBP-3-degrading proteinase (11,12), produces six IGFBP-3 fragments, four of which bind heparin-agarose (13). Sequence analysis reveals that the nonheparin-binding fragments comprise the first 100 -110 amino acids of IGFBP-3, demonstrating that no heparin-binding domains reside in this segment of the binding protein. However, amino acid analysis of the heparin-binding fragments shows that each contains at least one of two heparin-binding consensus sequences present in IGFBP-3 (11,13,14). Herein, we examine whether these putative heparin-binding domains are involved in IGFBP-3-GAG interactions and define their affinities for heparin, as well as for other GAGs.
Production, Purification, and Characterization of Epitope-specific Antibodies-Three mg of each peptide was conjugated to keyhole limpet hemocyanin using m-maleimidobenzoyl-N-hydroxysuccinimide, utilizing the terminal -SH group of each peptide for conjugation. The peptide-conjugate was mixed with an equal volume of complete Freund's adjuvant and injected in a pair of New Zealand White rabbits. Repeat injections were performed on five separate occasions over ϳ105 days. Crude antiserum contained high titer antiserum to each peptide when compared with preimmune serum as assessed by an enzyme-linked immunosorbent assay using BSA-conjugated peptide in the solid phase.
Epitope-specific antibodies were purified from crude antiserum using affinity columns prepared from each peptide covalently linked via its terminal cysteine to SulfoLink TM according to the manufacturer's instructions. Three ml of peptide-affinity matrix was incubated with 20 ml of crude antiserum and 20 ml of PBS and mixed in a 50-ml conical tube at room temperature for 2-3 h. Antibodies were eluted from the column in 100 mM glycine buffer, pH 2.5, and neutralized with 1 M Tris-HCl, pH 9.5. Antibodies were dialyzed into 5 mM phosphate buffer, pH 7.4, and then lyophilized and stored at Ϫ20°C. Little or no crossreactivity was observed among the three antibody preparations when analyzed by dot blotting using all three peptides (data not shown).
Solid-phase Peptide Binding Assays-Biotinylated heparin (bHep) was prepared based on a method used by Yu and Toole (17)  The solid-phase peptide binding assay was performed using a modification of a method reported by Kost et al. (18). Fifty-microliter aliquots of various concentrations of synthetic peptides dissolved in carbonate buffer, pH 9.6, were absorbed onto 96-well tissue culture plates overnight at 4°C. The wells were then saturated for 1 h at room temperature with 100 l/well PBS, pH 7.4, containing 3% BSA, which had been denatured at 60°C for 30 min. The plate was washed with PBS, pH 7.4, containing 0.1% Tween 20 (PBST) and then incubated for 3 h at room temperature with bHep diluted in PBST, 0.2% BSA (final concentration, 5 g/ml) with or without various concentrations of unlabeled heparin, HS, CS-A, CS-C, DS, or HA. After washing the plate, 50 l of streptavidin-conjugated horseradish peroxidase (Amersham Corp.) diluted 1:1000 in PBST, 0.2% BSA was added to each well and incubated for 1 h at room temperature. After a final wash, the peroxidase substrate 3,3Ј,5,5Ј-tetramethylbenzidine dihydrochloride (Sigma) was added, and the reaction was terminated with the addition of 2 M H 2 SO 4 . The plate was read in an automated plate reader at A 450 .
Statistical Analysis-Statistical significance between groups was determined by paired Student's t test. Curve-fitting, EC 50 , and IC 50 values were calculated using PRISM Software (GraphPad Software, San Diego, CA).

Identification of Heparin-binding Domains within IGFBP-3
Using Epitope-specific Antibodies-In order to determine if regions of IGFBP-3 containing putative heparin-binding sequences do indeed bind heparin, we prepared an epitope-specific antibody to an IGFBP-3 peptide, which contains no heparin-binding consensus sequence (peptide I), and antibodies to two other IGFBP-3 peptides (peptides IV and VI), each of which contains one of two putative heparin-binding consensus sequences (14). Peptide IV contains the amino acid sequence 149 KKGHA 153 , which resembles the short heparin-binding consensus sequence XBBXBX (where B is basic amino acid and X is any non-basic amino acid), while peptide VI contains the longer heparin-binding consensus sequence XBBBXXBX ( 219 YKKKQCRP 226 ). All three anti-peptide antibodies recognized rhIGFBP-3 E. coli (Fig. 1) and rhIGFBP-3 CHO (data not shown) under reducing conditions. Heparin had little or no effect on the binding of the antibody to peptide I (Fig. 1). In contrast, binding of antibodies to peptides IV and VI was significantly inhibited by heparin, using either rhIGFBP-3 E. coli ( Fig. 1) or rhIGFBP-3 CHO (data not shown). These data were examined by densitometry, and the results are presented in Table I. Because heparin markedly inhibited the binding of antibodies to both peptide IV and peptide VI, these findings suggested that within these regions of IGFBP-3 reside functional heparin-binding domains.
Characterization of Heparin Binding to IGFBP-3 Heparinbinding Domains-A solid-phase binding assay using immobilized peptides IV and VI and bHep as ligand was used to characterize the relative affinities of these IGFBP-3 domains for heparin. As shown in Fig. 2A, both peptides bound bHep in a dose-dependent fashion. Peptide IV bound bHep (5 g/ml) with an EC 50 of 15 g/ml peptide (750 ng/well), while peptide VI bound bHep with an EC 50 of 3.6 g/ml peptide (180 ng/well), demonstrating that peptide VI bound bHep with ϳ4-fold higher affinity than peptide IV. Furthermore, binding of bHep to both peptides was saturable at concentrations of ϳ30 g/ml (1.5 g/well) for peptide IV and ϳ10 g/ml (500 ng/well) for peptide VI. As shown in Fig. 2B, binding of bHep to both peptides was specific. When peptides were coated onto plates at a maximal concentration (1.5 g/well for peptide IV and 500 ng/well for peptide VI), heparin displaced bHep in a dose-dependent manner with an IC 50 of ϳ3 g/ml heparin for both peptides. Together, these data demonstrated that heparin bound both peptides, yet it bound peptide VI more avidly than peptide IV.
Binding of GAGs to Heparin-binding IGFBP-3 Peptides-Since GAGs constitute a diverse group of complex macromolecules, we examined the ability of several well characterized GAGs to inhibit the binding of bHep to both heparin-binding FIG. 1. Binding of epitope-specific antibodies to rhIGFBP-3 E. coli . rhIGFBP-3 E. coli (500 ng/lane) was immunoblotted with antibodies to peptide I (lanes 1 and 2), antibodies to peptide IV (lanes 3 and 4), or antibodies to peptide VI (lanes 5 and 6) in the absence (lanes 1, 3, and  5) or presence of heparin (lanes 2, 4, and 6) as described under "Experimental Procedures." Immunoreactive rhIGFBP-3 E. coli is denoted with an 4 on the right side of the figure. Molecular weight markers are indicated on the left.

IGFBP-3 GAG-binding Domains 14677
IGFBP-3 peptides. As shown in Fig. 3, the most potent inhibitors of bHep binding to both peptides were heparin, HS, and DS. CS-A and HA were intermediate in their overall inhibitory activities toward bHep binding to both peptides, while CS-C caused no inhibition.

DISCUSSION
Herein, we demonstrate that IGFBP-3 contains at least two heparin-binding domains that conform to highly basic, heparinbinding consensus sequences proposed by Cardin and Weintraub (14). One is found in the non-homologous midregion of IGFBP-3, and it conforms to the heparin-binding consensus sequence XBBXBX, while the other is found in the highly homologous, C terminus of the binding protein, and it conforms to the longer heparin-binding consensus sequence XBBBXXBX. Similar heparin-binding consensus sequences have now been identified in a variety of proteins including protease inhibitors such as antithrombin III and heparin cofactor II, structural proteins such as fibronectin, vitronectin, and laminin, and growth factors such as fibroblast growth factor (for reviews see Refs. 19 -21). Such GAG-protein interactions appear to mediate a variety of physiological functions, including altering protease activity, enhancing growth factor action, and regulating gene expression (19 -21). While the consequences of IGFBP-3-GAG Localization of IGFBP-3 to cell surfaces and/or extracellular matrix is likely important for the enhancement of IGF action (22)(23)(24), possibly by reducing the affinity of IGFBP-3 for IGFs, thus making IGFs more available to interact with IGF receptors (4,24,25). Furthermore, this association may be important for IGF-independent actions of IGFBP-3 on cellular proliferation (5,26), for the cellular uptake of IGFBP-3 (4, 7), as well as for modulating the inhibitory effects of IGFBP-3 on IGFBP-4 proteolysis (13).
Although our studies demonstrate that IGFBP-3 contains at least two GAG-binding domains, it is unclear if both domains participate in binding GAGs under physiologic conditions. While both domains demonstrate specific binding of heparin, the heparin-binding domain present in the C terminus of the molecule demonstrates the highest affinity for heparin. Interestingly, IGFBP-5 and IGFBP-6 also contain C-terminal sequences, which are homologous to the C-terminal heparinbinding consensus sequence present in IGFBP-3 (1-4, 5, 10, 27), suggesting that the C-terminal heparin-binding sequence of all three IGFBPs may be important in GAG interactions. This is supported by the findings that synthetic peptides containing the C-terminal heparin-binding domains from IG-FBP-3, -5, and -6 inhibit IGFBP-3 and IGFBP-5 binding to endothelial cell monolayers (27,28). Consistent with these observations, Andress (29) has recently shown heparin prevents the association of intact IGFBP-5 with mouse osteoblasts, yet heparin does not interfere with the binding of a C-terminally truncated form of IGFBP-5 lacking the heparinbinding domain. Furthermore, Arai et al. (30) have demonstrated that point mutations of basic amino acids present in the C-terminal heparin-binding domain of IGFBP-5 can significantly reduce its affinity for heparin. Data from our laboratory suggest that the C-terminal heparin-binding sequence from IGFBP-5 has a similar affinity for heparin as does the homologous heparin-binding domain from IGFBP-3, while the homologous IGFBP-6 consensus sequence binds heparin much less

IGFBP-3 GAG-binding Domains 14678
avidly, 2 possibly explaining why IGFBP-6 has not been shown to associate with cell monolayers (27,28). Further studies using epitope-specific antibodies as described herein and sitedirected mutagenesis of IGFBPs should provide further insights into the specificity of IGFBP-GAG interactions.
The specificity involved in GAG-protein interactions has only recently been appreciated. For instance, both antithrombin III and heparin cofactor II bind heparin, but only heparin cofactor II binds DS (21). Our data suggest that both heparin-binding domains identified in IGFBP-3 bind several different GAGs but with different affinities. Both IGFBP-3 peptides containing heparin-binding domains demonstrated the greatest overall binding to heparin, HS, and DS, suggesting that common features shared among these three GAGs might provide clues as to the specificity involved in this interaction. A major structural similarity among heparin, HS, and DS is that each contains ␣-L-iduronic acid residues, suggesting that the higher affinity of these GAGs for IGFBP-3 heparin-binding domains may be dictated, at least in part, by the disaccharide backbone of the GAG. Consistent with our data, Arai et al. (31) demonstrated that heparin, HS, and DS were the most potent GAGs in inhibiting IGFBP-5-IGF interactions. Although these authors did not examine directly the binding of GAGs to IGFBP-5, their data suggested that GAG interactions which inhibited the formation of the IGFBP-5-IGF complex contained primarily Osulfate groups in either the 2-or 3-carbon positions. While sulfation of GAGs may affect protein binding, it is unclear from our studies to what extent the degree of sulfation or the position of the sulfate group modulates GAG binding to IGFBP-3. For instance, CS-A and CS-C contain the same disaccharide unit and both are sulfated GAGs. Nevertheless, CS-A is principally sulfated at the 4-carbon position, while CS-C is sulfated primarily at the 6-carbon position, suggesting that sulfation at the 4-carbon position promotes binding to IGFBP-3 heparinbinding domains, especially the short heparin-binding domain (see Fig. 3). In contrast, HA, which is not sulfated, bound both peptides with similar affinities to CS-A. Taken together, these data would suggest that sulfation at the 6-carbon position may inhibit IGFBP-3-GAG binding. Because GAGs are commonly covalently attached to protein cores (i.e. proteoglycans), it is possible that protein-protein interactions may also modulate IGFBP-3-GAG binding. For instance, several proteoglycans including decorin, fibromodulin, and biglycan have core proteins containing leucine-rich repeats, which are homologous to sequences found in the acid-labile subunit that binds IGFBP-3-IGF complexes in serum (32).
In conclusion, while other reports have established that IG-FBP-3 binds heparin-Sepharose, the studies herein are the first to localize and characterize the specific domains within the IGFBP-3 molecule that bind heparin. Furthermore, they suggest that IGFBP-3 may interact in a selective way with certain GAG moieties. Thus, these studies should provide essential information for the better understanding of IGFBP-3-GAG interactions and how these interactions mediate the effects of IGFBP-3 at the cellular and extracellular interface.