BIAcore analysis of bovine insulin-like growth factor (IGF)-binding protein-2 identifies major IGF binding site determinants in both the amino- and carboxyl-terminal domains.

In the absence of a complete tertiary structure to define the molecular basis of the high affinity binding interaction between insulin-like growth factors (IGFs) and IGF-binding proteins (IGFBPs), we have investigated binding of IGFs by discrete amino-terminal domains (amino acid residues 1-93, 1-104, 1-132, and 1-185) and carboxyl-terminal domains (amino acid residues 96-279, 136-279, and 182-284) of bovine IGFBP-2 (bIGFBP-2). Both halves of bIGFBP-2 bound IGF-I and IGF-II in BIAcore studies, albeit with different affinities ((1-132)IGFBP-2, K(D) = 36.3 and 51.8 nm; (136-279)IGFBP-2HIS, K(D) = 23.8 and 16.3 nm, respectively). The amino-terminal half appears to contain components responsible for fast association. In contrast, IGF binding by the carboxyl-terminal fragment results in a more stable complex as reflected by its K(D). Furthermore, des(1-3)IGF-I and des(1-6)IGF-II exhibited reduced binding affinity to (1-279)IGFBP-2HIS, (1-132)IGFBP-2, and (136-279)IGFBP-2HIS biosensor surfaces compared with wild-type IGF. A charge reversal at positions 3 and 6 of IGF-I and IGF-II, respectively, affects binding interactions with the amino-terminal fragment and full-length bIGFBP-2 but not the carboxyl-terminal fragment.

IGFs 1 are polypeptide hormones that elicit mitogenic and metabolic effects upon binding the IGF and insulin receptors (reviewed by Jones and Clemmons (1)). A family of at least six IGF-binding proteins (IGFBPs) modulates the bioavailability of IGF-I and IGF-II. IGFBPs play an important role in regulating IGF actions by increasing their circulating half-life and by affecting their tissue distribution and localization. IGFBPs have been shown to inhibit IGF action by preventing interactions between IGFs and IGF receptors. IGFBPs also potentiate IGF action in vitro by co-localizing the IGF⅐IGFBP complex with cell surface receptors and promoting the subsequent release of IGF through a number of mechanisms, including proteolysis (2). Tissue expression and serum levels of IGFBPs are under a variety of influences, such as developmental, hormonal, and physiological conditions including pregnancy, and several disease states including tumorigenesis (reviewed by Rajaram et al. (3)).
IGFBPs 1-6 are generally thought to consist of three structural domains of approximately equal length (4,5). All six IGFBPs are characterized by highly conserved cysteine-rich amino-and carboxyl-terminal domains, joined by a linker domain unique to each IGFBP species. Although the role of the IGFBPs in modulating IGF activity is widely accepted, the exact mechanism of interaction and the IGF binding sites on the IGFBPs have yet to be elucidated. The proposed IGF binding site on the IGFBPs is believed to be located within the highly conserved amino-and carboxyl-terminal domains, as studies have shown both contain residues important for IGF binding (reviewed by Hwa et al. (4) and Baxter (2)).
The earliest insights into IGF binding regions of IGFBPs arose from observations that proteolysed fragments often retained residual affinity for IGFs. IGF binding activity has been reported for amino-terminal fragments of IGFBP-1, -3, -4, and -5 (6, 8 -12). Although these fragments demonstrate IGF binding properties, there is a distinct loss of affinity, suggesting a requirement for additional components. Interestingly, IGFBPrelated proteins, which share sequence similarities with the amino-terminal domain of the high affinity IGFBPs, have also been reported to bind IGF-I, IGF-II, and insulin (4,(13)(14)(15).
Deletion and chemical modification studies have also contributed significantly to the elucidation of critical residues involved in IGF binding. Removal of either amino-or carboxylterminal residues from IGFBP-1 severely disrupted IGF binding (20,21), whereas deletion of amino acid residues 222-284 of bIGFBP-2 significantly altered the IGF binding capabilities of bIGFBP-2 (22). However, chemical iodination and sitedirected mutagenesis of bIGFBP-2 by Hobba et al. (23,24) showed the amino-terminal residue Tyr-60 contributed to IGF binding. NMR analysis and site-directed mutagenesis confirmed the involvement of the equivalent residue of hIGFBP-5 (10,25), whereas recombinant amino-terminal fragments of rIGFBP-3 (26) and hIGFBP-4 (27) have also been shown to retain residual binding affinity.
In this study, we have investigated the possibility of reconstituting the high affinity binding interaction between IGF and bIGFBP-2 using truncated amino-and carboxyl-terminal fragments. On the basis of sequence alignments, exon coding regions (28), and naturally occurring IGFBP fragments (6,7,10,17,18,26,28), discrete modules of the native protein encompassing the amino-terminal domain (amino acid residues 1-93, 1-104, 1-132, and 1-185) and carboxyl-terminal domain (amino acid residues 96 -279, 96 -284, 136 -279, and 182-284) of bIGFBP-2 were recombinantly expressed and investigated using BIAcore analysis (Fig. 1). It is clear from the present investigation that the amino-terminal domain of bIGFBP-2 is capable of rapid association with both IGF-I and IGF-II but also exhibits increased dissociation rates compared with native bIGFBP-2. In contrast, the carboxyl-terminal fragment forms a somewhat more stable complex with IGF than does the aminoterminal domain, although it is clear that both the amino-and carboxyl-terminal domains act in combination to form the high affinity complex exhibited by the native IGFBP-2. Furthermore, we have shown that the amino-terminal tripeptide of IGF-I and hexapeptide of IGF-II are involved in binding both the amino-and carboxyl-terminal domains of bIGFBP-2.

Carboxyl-terminal Expression Constructs
To enable the secretion of carboxyl-terminal domains expressed in mammalian cells, an ApaI restriction site was introduced at the 3Ј end of the sequence encoding the bIGFBP-2 leader sequence of pGF8. This ApaI site allowed subsequent introduction of cDNAs encoding carboxylterminal domains into the expression vector. The ApaI restriction site was generated using the Bio-Rad in vitro mutagenesis kit and mutagenic oligonucleotide 5Ј-CAC CTC GGC GCG GGC CCC GCA GTC G-3Ј. PCR products encoding the carboxyl-terminal domain of bIGFBP-2 to be cloned into the pXMT2 vector were generated using the forward primers 5Ј-TTT TTT GGG GCC CGC GCC GCC AGG ACC CCC TGC CAG-3Ј ( 182-284 IGFBP-2), 5Ј-TTT TTT GGG GCC CGC GCC GCT GAG TAC AGC GCC AGC-3Ј ( 96 -284 IGFBP-2/ 96 -279 IGFBP-2HIS), and 5Ј-AAA AAT TGG GCC CGG AAG CCC CTC AAG TCC GGC ATG-3Ј ( 136 -279 IGFBP-2HIS), which incorporated 5Ј ApaI restriction sites (underlined) to facilitate ligation downstream of the signal sequence. The reverse primer 5Ј-GCC TTC ACA CGC TAG GAT T-3Ј, complementary to the pXMT2 vector, was used to generate the carboxyl-terminal construct 182-284 IGFBP-2, whereas the primer 5Ј-TTT AAG CTT GAA TTC TTA GTG GTG GTG GTG GTG GTG CAC CCC TCG AGC CCC CTG CTG CTC GTT GTA-3Ј was used to introduce a six-histidine tag at the end of the carboxyl-terminal constructs 96 -279 IGFBP-2HIS and 136 -279 IGFBP-2HIS. The reverse primers introduced a 3Ј EcoRI restriction site (underlined) for ligation into the parent pXMT2 vector. The cDNA encoding amino acid residues 136 -279 IGFBP-2HIS was also introduced into the pET32a (ϩ) vector (pET 136 -279 IGFBP2HIS) using the forward primer 5Ј-AAA GGT ACC CCA TGG GGT CGG AAG CCC CTC AAG-3Ј, which introduced a NcoI restriction site (underlined), and the reverse primer 5Ј-TTT AAG CTT GAA TTC TTA GTG GTG GTG GTG GTG GTG CAC CCC TCG AGC CCC CTG CTG CTC GTT GTA-3Ј, which introduced an EcoRI restriction site (underlined) and incorporated cDNA encoding a hexahistidine tag.
PCRs were performed routinely with 10% Me 2 SO as described by Sun et al. (30) to overcome secondary structure formation as a consequence of the high GC-rich content of the bIGFBP-2 cDNA. All PCRs were performed in a PerkinElmer Life Sciences thermal cycler (Norwalk, CT) using the following PCR conditions: initial denaturation at 94°C for 3 min, followed by 30 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 90 s. fmol DNA cycle sequencing (Promega) and ABI PRISM Dye Terminator Cycle Sequencing (PerkinElmer Life Sciences) confirmed generation of correct cDNA clones. Automated sequencing was performed at the Institute of Medical and Veterinary Science Sequencing Center (Adelaide, South Australia, Australia).

Mammalian Cell Expression
Recombinant bIGFBP-2 and the truncated mutants were expressed in COS-1 simian kidney cells as previously described (22). Culture medium was collected every 24 h for 5 days. Presiliconized Eppendorf tubes were used for protein handling and storage.

E. coli Protein Expression
The pET constructs were transformed into E. coli BL21 cells for large scale expression. Two-liter cultures were grown in Luria-Bertani broth containing carbenicillin (100 g/ml) at 37°C with shaking and induced with 1 mM isopropyl-␤-D-thiogalactopyranoside at an A 600 of ϳ0.6. Protein was expressed for a minimum of 4 h prior to pelleting the cells (5000 ϫ g for 5 min at 4°C). The cells were then resuspended to one-tenth of the original volume in 2.5 mM EDTA, 20 mM Tris, pH 8.0, with lysozyme (100 g/ml) and sonicated (Sonifier cell disrupter B-30) three times for 30 s each on ice. The soluble fraction was collected (10000 ϫ g for 10 min at 4°C) and filtered.

Purification
The expressed products of wild-type and amino-terminal bIGFBP-2 constructs were purified from transfected COS-1 cell conditioned medium as described previously (22). The carboxyl-terminal histidinetagged proteins were purified by Ni-IDA chromatography (Bioserve). Briefly, nonspecific interacting proteins were removed with 60 mM imidazole, and specifically bound protein was eluted with 1 M imidazole. Wild-type and truncated bIGFBP-2 preparations were further purified by rpHPLC as described by Hobba et al. (23). The eluted protein was detected by absorbance at 215 nm. The amino-terminal fusion protein was cleaved from the E. coli expressed proteins by enterokinase (0.5 units/mg of protein, Invitrogen EkMax), and the products were separated by rpHPLC. Purified protein samples were analyzed by separation on 12.5% polyacrylamide gels under nonreducing conditions and stained with Coomassie Blue R250 as previously described by Hobba et al. (23). All protein samples were lyophilized for storage.

BIAcore Analysis
Biosensor Surface Preparation-Two types of CM5 biosensor chips were generated. IGF-I and IGF-II were immobilized to the first chip, and 1-279 IGFBP-2HIS, 1-132 IGFBP-2, and 136 -279 IGFBP-2HIS to the second. A reference surface, to which no ligand was bound, was included on both chips. Ligands were immobilized to the CM5 gold surfaces by amine coupling as previously described (32). Briefly, the peptides to be immobilized (12.5 g/ml in 50 mM sodium acetate, pH 4.7) were injected onto the activated CM5 surface at 5 l/min (HBS running buffer: 10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant P20, pH 7.4). The surfaces were then deactivated by passage of 1 M ethanolamine. Biosensor surfaces were coupled to final resonance values of ϳ200 response units for IGF/insulin chips and 600 response units for IGFBP-2 chips.
Acquisition of Kinetic Binding Data-Various concentrations of analyte (200, 100, 50, 25, and 12.5 nM) were injected during the association phase for 5 min (40 l/min), with HBS as the running buffer. The dissociation phase, initiated by passage of HBS alone, was carried out over a period of 10 min. The biosensor surfaces were regenerated by a 60-s injection of 0.1 M HCl. Samples were injected in duplicate in random order in at least two separate experiments. Kinetic data were analyzed using the BIAevaluation software, version 3.0. All binding curves were corrected for background and bulk refractive index contribution by subtraction of the reference flow cells. Models were fitted both globally across the data sets and for a single concentration. Models used were the 1:1 Langmuir binding interaction describing 1:1 binding between analyte (A) and ligand (B) (A ϩ B 7 AB), and a two-state reaction (conformational change) model, based on a 1:1 binding of analyte to an immobilized ligand followed by a conformational change (A ϩ B 7 AB 7 AB*). The 1:1 stoichiometry for the interaction between IGFs and IGFBPs was previously established by Bourner et al. (33).
Reconstitution of the High Affinity IGF⅐IGFBP Complex on IGF Biosensor Surfaces-Reconstitution of the wild-type binding interaction between bIGFBP-2 and IGF ligands was investigated by two mechanisms. Initially, the amino-and carboxylterminal fragments of bIGFBP-2 were preincubated in HBS running buffer prior to injection across IGF biosensor surfaces. Secondly, IGF-II and the truncated IGFBP fragments were sequentially injected across IGFBP biosensor surfaces. The 1-185 IGFBP-2 and 96 -279 IGFBP-2HIS fragments could not be used concurrently in the same BIAcore experiment due to nonspecific interactions with each other, presumably through the common hydrophobic central region (amino acid residues 96 -185). Interestingly, the carboxyl-terminal fragment 96 -279 IGFBP-2HIS bound not only the amino-terminal biosensor surface but also to itself on the biosensor surface. This was not observed for 1-132 IGFBP-2, 1-185 IGFBP-2, or 136 -279 IGFBP-2HIS. Preincubation of 1-132 IGFBP-2 and 136 -279 IGFBP-2HIS failed to reconstitute the high affinity binding interaction of bIGFBP-2 with IGF (Fig. 4). This would suggest the requirement for the two halves of the molecule to be linked to each other to produce the high affinity complex.
Orientation of the IGF⅐IGFBP Molecules in the High Affinity State Complex-The kinetic data obtained from binding of IGF to immobilized bIGFBP-2 mutant biosensor surfaces were evaluated using both the 1:1 Langmuir model and the two state conformational change model. Binding kinetics derived from both models were similar. Interestingly, the two state conformational change model was a better fit than the 1:1 Langmuir binding model when comparing a single concentration of IGF over IGFBP biosensor surfaces (previously reported by M. Lucic (49)). Therefore, the kinetic analyses derived using the two state conformational change model for 50 nM concentrations of IGF-II, des(1-6)IGF-II, and R 6 IGF-II over immobilized 1-279 IGFBP-2HIS, 1-132 IGFBP-2, and 136 -279 IGFBP-2HIS biosensor surfaces are presented here (Table II). Due to the significantly poorer response of the IGF-I analogues and IGF-II analogues over 136 -279 IGFBP-2HIS biosensor surfaces, kinetic parameters could not be accurately derived; however, biosensorgrams have been presented for qualitative comparisons (Fig. 6).
Native IGF-I and IGF-II biosensorgrams have been presented showing binding to each of the immobilized bIGFBP-2 analogue biosensor surfaces (Fig. 6). R 3 IGF-I and R 6 IGF-II bind 1-279 IGFBP-2HIS, 1-132 IGFBP-2, and 136 -279 IGFBP-2HIS with higher affinity than the truncated analogues, consistent with previously reported solution binding assays (34,35). Relative to native IGF-II, R 6 IGF-II exhibits 3.4-and 4.0-fold decreased equilibrium affinities for 1-279 IGFBP-2HIS and 1-132 IGFBP-2 biosensor surfaces, respectively (Table II). In contrast, R 3 IGF-I and R 6 IGF-II bind 136 -279 IGFBP-2HIS biosensor surfaces as well as do their respective native IGFs (Fig.  6, c and f). These results suggest that the glutamate residue is specifically involved in binding the amino-terminal domain of bIGFBP-2. In contrast, des(1-3)IGF-I and des(1-6)IGF-II exhibited reduced binding to 1-279 IGFBP-2HIS, 1-132 IGFBP-2, and 136 -279 IGFBP-2HIS biosensor surfaces. The amino-tripeptide of IGF-I and hexapeptide of IGF-II, therefore, influenced binding to both the amino and carboxyl termini of bIGFBP-2. It is unclear whether the low affinity interaction between des-IGFs and bIGFBP-2 is due to structural perturbations in the IGF molecule (to date, no structure analysis has been performed, although receptor binding is not affected (34, 36)) or is due specifically to loss of the amino-terminal residues of IGF.

DISCUSSION
We have shown that both the amino-and carboxyl-terminal domains of bIGFBP-2 bind IGF-I and IGF-II. The amino-terminal fragments 1-132 IGFBP-2 and 1-185 IGFBP-2 exhibit rapid association but lack the components essential for maintaining a stable complex. In contrast, 136 -279 IGFBP-2HIS binds IGFs less rapidly than either amino-terminal fragments or 1-279 IGFBP-2HIS but retains the ligand in a more stable complex than 1-132 IGFBP-2. These findings suggest that both the highly conserved amino-and carboxyl-terminal domains of bIGFBP-2 are required for sequestering IGF ligands. Indeed, they appear to act in a coordinated fashion in the high affinity interaction. Furthermore, we have shown that the aminoterminal residues of IGF are critical binding determinants for both 1-132 IGFBP-2 and 136 -279 IGFBP-2HIS, whereas the glutamate at positions 3 and 6 of IGF-I and IGF-II, respectively, is important for the interaction with the amino-terminal domain of bIGFBP-2.
The present investigation shows in real time the high affinity association of the amino-terminal fragments 1-132 IGFBP-2 and 1-185 IGFBP-2 with IGF-I (k a ϭ 6.36 ϫ 10 5 and 7.94 ϫ 10 5 Ms Ϫ1 , respectively) and IGF-II (k a ϭ 2.89 ϫ 10 5 and 1.35 ϫ 10 5 Ms Ϫ1 , respectively). The amino-terminal domains of IGFBPs 1-6 (usually incorporating all 12 cysteine residues) have been reported to bind IGF (8,9,26,(37)(38)(39)(40). Interestingly, the central linker region appears to contribute to the structural integrity and hence the IGF binding potential of these amino-terminal fragments. Hashimoto et al. (26) highlighted the importance of inclusion of the linker region because a fragment of rat IGFBP-3 (amino acid residues 1-93) retained only 4.2% IGF binding, whereas addition of linker peptide residues (94 -186) increased IGF binding to 12% of native rat IGFBP-3. A similar effect was seen in the present study. In fact, expression of 1-94 IGFBP-2, 1-104 IGFBP-2, and 182-284 IGFBP-2 was not successful in the present investigation, presumably due to increased susceptibility to proteolysis resulting from a lack of structural integrity.
The present study has confirmed the ability of the carboxylterminal half of bIGFBP-2 (amino acid residues 136 -279) to bind IGF-I (K D ϭ 23.8 nM) and IGF-II (K D ϭ 16.3 nM). Previously, there was some debate about the ability of the carboxyl- terminal domain to bind IGF. Although carboxyl-terminal fragments of hIGFBP-4 (27) and hIGFBP-5 (10) had no demonstrable IGF binding, similar fragments, including a naturally occurring truncation of hIGFBP-2 (amino acid residues 169 -289) (7,16,40), retained partial IGF binding. Furthermore, carboxyl-terminal deletion studies of bIGFBP-2 clearly indicated the importance of the carboxyl-terminal domain in IGF binding (22). Removal of 14, 36, and 48 amino acid residues from the carboxyl terminus appeared to have no effect on IGF binding, whereas removal of a further 14 amino acid residues dramatically reduced IGF binding. Other studies have shown that removal of most or all of the carboxyl-terminal domains of IGFBP-3 (16,41), , and IGFBP-5 (8) reduces IGF binding significantly.
It is clear that the amino-and carboxyl-terminal domains are required to work together to form the high affinity complex. However, whether the amino-or carboxyl-terminal domain of IGFBP is the major binding site is a matter of interpretation. The present investigation clearly shows the amino-terminal domain is required for rapid association of IGFBP-2 with IGF, and the carboxyl-terminal domain is required for the ability of IGFBP-2 to maintain IGF in a complexed form. Recently, Galanis et al. (42) reported that the recombinant expression of IGFBP-3 fragments encompassing amino-and carboxyl-terminal residues (1-88 and 165-264) bind IGF-I and IGF-II with reduced affinity. Consistent with the current findings, the amino-terminal fragment (N-88) dissociated 50-and 75-fold faster from IGF-I and IGF-II biosensor surfaces compared with the carboxyl-terminal fragment (C-165). In addition, Kalus et al. (10) purified two amino-terminal fragments of hIGFBP-5 (amino acid residues 1-94 and 40 -92) that exhibited similar affinity for IGF, both with 200-fold reduced affinity compared with wild-type. Imai et al. (25) confirmed the importance of corresponding amino-terminal residues in IGFBP-3, and Qin et al. (27) did the same for IGFBP-4. Furthermore, as amino acid residues 1-39 of hIGFBP-5 did not contribute to maintaining the complex, it was concluded that additional components required for IGF⅐IGFBP-5 complex stability would be derived from the carboxyl-terminal residues of IGFBPs (10). It is feasible, therefore, that the IGF binding sites located in the ami-  (Fig. 6) fits were equal to or less than 0.8 except R 6 IGF-II over 136 -279 IGFBP-2HIS ( 2 ϭ 3.5). Dissociation constants were derived using both association (k a ) and dissociation (k d ) rates (K D ϭ (k d1 /k a1 ) (k d2 /k a2 )).  no-and carboxyl-terminal domains of IGFBPs combine with each other to create one high affinity binding site. Indeed, Lys-222 to Asn-236 of bIGFBP-2 may be located in close enough proximity to the amino-terminal domain to allow both domains to interact with IGF (22). Site-directed mutagenesis of Gly-203 and Gln-209 in the equivalent region of rIGFBP-5 resulted in an 8-and 6-fold reduction in IGF-I binding affinity (43).
The amino-terminal residues of IGF are known to be essential for interaction with IGFBPs as IGF analogues des(1-3)IGF-I and des(1-6)IGF-II have significantly reduced binding affinity (34). Ho and Baxter (17) orientated the IGF molecule in complex with hIGFBP-2 using the analogue des(1-6)IGF-II. des(1-6)IGF-II was found to displace tracer from intact hIGFBP-2 with 10% the activity of native ligand; however, it was totally inactive in displacing IGF-II tracer from a carboxylterminal fragment (amino acid residues 169 -289). These results suggest that the amino terminus of IGF-II interacts with the carboxyl-terminal domain of rhIGFBP-2. Photoaffinity labeling experiments by Horney et al. (19) further supported this conclusion. However, the present study showed that the aminohexapeptide of IGF-II influenced binding not only to the carboxyl-terminal half of bIGFBP-2 but also to the amino-terminal half. This indicates the existence of binding sites for the hexapeptide of IGF within both the amino-and carboxyl-terminal domains of IGFBP-2. A summary of these findings is represented schematically in Fig. 7. Furthermore, Glu-6 of IGF-II appeared to specifically affect IGF binding only to 1-132 IGFBP-2 and did not alter binding to 136 -279 IGFBP-2HIS. It will be interesting to determine the ability of 1-132 IGFBP-2 and 136 -279 IGFBP-2HIS to bind other IGF analogues known to influence interaction with IGFBPs (viz. IGF-I: Glu-3, Thr-4, Gln-15, Phe-16 (44); Thr-4, Glu-9, Phe-16 (45); Phe-49, Arg-50, and Ser-51 (46); IGF-II: Phe-26, Phe-48, Arg-49, and Ser-50 (47)). We are also identifying the specific residues of IGF-I and IGF-II involved in binding to either the amino-or carboxylterminal domains by NMR.
Based on the present findings, we propose a model for the modulation of IGF activity (Fig. 7). Classical high affinity IGF binding only occurs when both the amino-and carboxyl-terminal domains of IGFBPs are linked together, as is evident also from partial proteolysis studies (reviewed by Conover (48)). Two binding sites exist, one in the amino-terminal domain that has a fast association component and one in the carboxylterminal domain that is essential for the stability of the com-plex. These sites must be combined to form one high affinity binding site. Glu-3 of IGF-I or Glu-6 of IGF-II interacts directly with the amino-terminal domain of IGFBP-2, whereas the amino terminus of IGF is critical for binding of both the aminoand carboxyl-terminal fragments of IGFBP-2. Although the objective of this study was to elucidate the IGF binding site on IGFBP-2, it will be interesting to investigate the modulation of IGF activity by IGFBP fragments in biological assays.
The present experiments further emphasize the requirement for the two halves of the IGFBP to exist concomitantly in order to establish a high affinity binding complex with IGF, but we are not sure by which means the two domains come together to form the high affinity binding interaction. The IGF binding site of IGFBPs obviously requires further investigation. Determining the residues involved in the interaction between IGF and IGFBP will ultimately allow researchers to develop a comprehensive insight into the molecular dynamics of the binding interaction, making it possible to produce mutant binding proteins for therapeutic applications. Not until we can derive a three-dimensional structure of the complex, by NMR or x-ray crystallography, will we truly understand this interaction. FIG. 7. Model of the orientation of the IGF and IGFBP-2 molecules in the binding interaction. This model is based on the data presented. The amino termini of IGF-I and IGF-II interact with both the highly conserved amino-and carboxyl-terminal domains of bIGFBP-2. The amino-terminal glutamate of IGF interacts specifically with the amino-terminal domain of bIGFBP-2, shown here for IGF-I. This is a high affinity complex preventing interaction with the type 1 IGF receptor. Proteolysis of the IGFBPs yields amino-and carboxyl-terminal fragments, similar to those produced recombinantly in this study, which exhibit reduced IGF binding affinities compared with the native protein.