An Investigation of the Ligand Binding Properties and Negative Cooperativity of Soluble Insulin-like Growth Factor Receptors*

To investigate the interaction of the insulin-like growth factor (IGF) ligands with the insulin-like growth factor type 1 receptor (IGF-1R), we have generated two soluble variants of the IGF-1R. We have recombinantly expressed the ectodomain of IGF-1R or fused this domain to the constant domain from the Fc fragment of mouse immunoglobulin. The ligand binding properties of these soluble IGF-1Rs for IGF-I and IGF-II were investigated using conventional ligand competition assays and BIAcore biosensor technology. In ligand competition assays, the soluble IGF-1Rs both bound IGF-I with similar affinities and a 5-fold lower affinity than that seen for the wild type receptor. In addition, both soluble receptors bound IGF-II with similar affinities to the wild type receptor. BIAcore analyses showed that both soluble IGF-1Rs exhibited similar ligand-specific association and dissociation rates for IGF-I and for IGF-II. The soluble IGF-1R proteins both exhibited negative cooperativity for IGF-I, IGF-II, and the 24-60 antibody, which binds to the IGF-1R cysteine-rich domain. We conclude that the addition of the self-associating Fc domain to the IGF-1R ectodomain does not affect ligand binding affinity, which is in contrast to the soluble ectodomain of the IR. This study highlights some significant differences in ligand binding modes between the IGF-1R and the insulin receptor, which may ultimately contribute to the different biological activities conferred by the two receptors.

Insulin-like growth factors (IGFs 2 ; IGF-I and IGF-II) mediate their biological functions primarily through the type 1 IGF receptor (IGF-1R). Both IGF-I and IGF-II are essential for normal growth and development. Targeted gene disruption of IGF-I or IGF-II results in mice with reduced birth weight compared with their wild type littermates (1,2), whereas disruption of the IGF-1R gene results in death at birth from respiratory failure (2). Conversely, overexpression of the IGF-1R or IGFs in cancer tissues leads to the potentiation of cancer cell growth and survival (3). Also, the IGF-1R is essential for transformation to a malignant phenotype (4). Interestingly, there is increasing evidence to support a role for IGF-II in a variety of cancers, where elevated IGF-II levels are associated with increased cancer risk (5)(6)(7)(8), reviewed in Ref. 3.
Despite their similarity in sequence and structure IGF-I and IGF-II can stimulate both overlapping and distinct biological functions (reviewed in Ref. 9). This is evident in patients with IGF-I deficiency, which results in severe growth and mental retardation, where IGF-II does not compensate for the loss of IGF-I activity (10 -12). Therefore, in order to understand how both IGF-I and IGF-II stimulate their respective biological outcomes, we first need to understand the mechanism by which both ligands bind and in turn activate the IGF-1R.
The IGF-1R is a member of the tyrosine kinase family of receptors and, together with the insulin receptor (IR) and insulin-related receptor, forms a subfamily with similar structural organization (reviewed in Ref. 9). The IGF-1R and IR are homodimers composed of two ␣ and two ␤ subunits and are synthesized as a single precursor polypeptide, which is then post-translationally processed by dimerization, proteolytic cleavage, and glycosylation.
The amino-terminal regions of the ␣ chain of these receptors are composed of three domains, two structurally homologous subdomains designated large homologous domain 1 (L1) and large homologous domain 2 (L2), which are separated by a cysteine-rich (CR) domain of ϳ160 amino acids (3) (see supplemental Fig. S1). The major ligand binding determinants of the structurally related IGF-1R and IR reside within the extracellular ␣ subunits of the receptors (reviewed in Refs. 9, 13, and 14). Studies performed with truncated IRs have demonstrated that dimerization of the ␤ chains is essential for receptor activation and signaling (15,16). Full-length disulfide-reduced half-receptors, dimeric receptors truncated at the transmembrane domain (ectodomain), and monomeric IR fragments all exhibit reduced affinity compared with the wild type receptor (17,18). High affinity binding is restored by the inclusion of IR dimerization domains (Fn1, Fn2 insert domain) in soluble IRs (19,20) and following fusion of the IR ectodomain with self-associating domains, such as immunoglobulin Fc (21) or a leucine zipper motif (22). A soluble IGF-1R extracellular-Z domain fusion protein binds IGF-I and IGF-II with high affinity (23), suggesting that all of the determinants required for high affinity ligand binding reside within the extracellular domain of the IGF-1R.
Chemical cross-linking studies, the generation of chimeric IGF-I/IR, and alanine-scanning mutagenesis have been used to further define the ligand binding sites within the ectodomain (24 -29). Insulin binds to the IR via residues within the L1 domain of the IR, and a similar binding site for IGF-I and IGF-II exists on the L1 domain of the IGF-1R. In addition, the carboxyl-terminal region of the ␣-subunit is critical for ligand binding to both receptors and is provided by the opposite receptor monomer to the L1 domain binding site (28, 30 -34). Another binding determinant is present in the L2/Fn1 domains and is provided by the opposite receptor monomer (35,36). Interestingly, a unique binding site for IGF-I exists within the CR domain of the IGF-1R and provides the specificity for that ligand (37)(38)(39).
Recently, the crystal structure of the IR ectodomain has been determined in the absence of bound ligand (40) and reveals a folded over conformation, which could accommodate two potential ligand binding sites. This arrangement is consistent with the evidence that ligand binding involves interaction with both receptor halves.
Kinetic analysis of the ligand-receptor interaction with wild type IR and IGF-1R reveals high and low affinity ligand states that are involved in the characteristic negative cooperativity mode of binding (reviewed in Refs. [41][42][43]. The accelerated dissociation of bound ligand from the IR in the presence of unlabeled insulin suggests the presence of at least two interacting binding sites. In contrast, those soluble IR ectodomain fragments that have low affinity for ligand have a linear Scatchard plot and do not show negative cooperativity (41,42). Although the wild type full-length IGF-1R exhibits negative cooperativity for IGF-I (44), it is yet to be determined whether a similar mode of binding exists for the related IGF-II ligand. In addition, truncation of the IGF-1R at the transmembrane domain is assumed to remove the characteristic feature of the wild type IGF-1R, negative cooperativity, but this has not been formally demonstrated.
In the present study, we have assembled and expressed two soluble human IGF-1R variants, comprising the ectodomain of the IGF-1R and a soluble chimeric IGF-1R-Fc fusion protein of the IGF-1R ectodomain, fused to the Fc region of the heavy chain of mouse immunoglobulin (IgG1), designated sIGF-1R and sIGF-1R/Fc, respectively. These soluble IGF-1Rs exhibit high affinity IGF-I and IGF-II binding and negative cooperativity determined by immunocapture competition binding assays and BIAcore biosensor analysis. The sIGF-1R/Fc and sIGF-1R both exhibit high affinity IGF-I and IGF-II binding and negative cooperativity. We have also utilized a number of conformationdependent anti-IGF-1R antibodies (45) as additional probes for receptor conformation and have examined their effect on IGF-I and IGF-II binding to either solubilized wild type or sIGF-1Rs.

EXPERIMENTAL PROCEDURES
Construction of Soluble IGF-1 Receptor cDNA Plasmids-A schematic representation of the IGF-1R domain structure and the soluble receptors used in this study is provided in supplemental Fig. S1. Two soluble human IGF-1R variants were generated, comprising the ectodomain of the IGF-1R and a soluble chimeric IGF-1R-Fc fusion protein of the IGF-1R ectodomain fused to the Fc region of the heavy chain of mouse immunoglobulin (IgG1), designated sIGF-1R and sIGF-1R/Fc, respectively. The cloning strategy, the generation of stable cell lines expressing the soluble IGF-1 receptors, screening of transfected cell lines, metabolic labeling and immunoprecipitation assays, purification of soluble IGF-1 receptors, and SDS-PAGE and Western blot analysis are all described in the supplemental materials.
Ligand Binding Analysis-Wells of microtiter plates (Immulon 4 HBX from Dynex Technologies) were coated with anti-IGF-1R monoclonal antibody 24-31 (10 g/ml, 100 l/well in phosphate-buffered saline) (45), which does not interfere with receptor binding, 3 and blocked with 0.5% bovine serum albumin (radioimmunoassay grade from Sigma) in phosphate-buffered saline for 2 h at room temperature. Cell culture supernatant from soluble receptor-expressing cell lines or purified soluble receptor was incubated overnight at 4°C. The wild-type IGF-1R was obtained from the p6 BALB/c3T3 cell line overexpressing the IGF-1R (46), and cell membranes were solubilized as described by Denley et al. (47). Various concentrations of unlabeled receptor grade recombinant human IGF-I or human IGF-II (both prepared in house as previously described (48,49)) were diluted in binding buffer (100 mM HEPES, 100 mM NaCl, 10 mM MgCl 2 , 0.05% (w/v) bovine serum albumin, and 0.025% (w/v) Triton X-100), either 25 pM 125 I-IGF-I or 125 I-IGF-II (GE Healthcare) was added, and plates were then incubated overnight at 4°C. Receptor was diluted to bind 10 -15% of added 125 I-IGF-I or 125 I-IGF-II tracer in the absence of unlabeled ligand. Plates were washed three times with phosphate-buffered saline, and bound counts were determined using a ␥ scintillation counter. The background binding was 5% of the total counts added, and this was subtracted to give specific counts bound. The binding data were analyzed using GraphPad Prism 3.03 by curve-fitting with a one-site competition model.
Receptor binding affinities were also determined using europium-labeled IGF-I (Eu-IGF-I) and IGF-II (Eu-IGF-II) in microtiter plate assays as described by Denley et al. (47). Briefly, IGF-1R was immunocaptured with mAb 24-31 on white Greiner Lumitrac 600 plates and blocked with 0.5% bovine serum albumin/TBST (Tris-buffered saline, 0.1% Tween 20), and then supernatant containing soluble receptor or solubilized p6 cell lysate was added. Bound europium-labeled IGF-I or IGF-II was then incubated on immobilized receptors together with unlabeled IGF-I or IGF-II or 100 nM mAb, as specified for individual experiments. Wells were washed three times in TBST and once in water. 100 l/well enhancement solution (PerkinElmer Life Sciences) was added and then incubated for 10 min prior to determining the bound europium label by timeresolved fluorescence (47).

BIAcore Determination of IGF-I and IGF-II Affinities for Soluble IGF-1R-sIGF-1R/Fc
and sIGF-1R were coupled to the CM5 biosensor chip using a similar method to that previously described (50). Briefly, CM5 sensor chips were activated with 35 l of N-ethyl-NЈ-[(dimethylamino)propyl]carbodiimide and N-hydroxysuccinimide at 5 l/min. Receptor dialyzed against HBS running buffer (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant P20, pH 7.4) was coupled to the CM5 sensor chip by injecting 35 l of receptor (4 g) in 10 mM sodium acetate, pH 4.5. Uncoupled surfaces were then deactivated by 1 M ethanolamine, pH 8.5. A blank flow cell was used as a reference on all chips. Binding analyses were performed at 25°C using a BIAcore 2000 biosensor (BIAcore, Uppsala, Sweden). A sensor surface coupled with 8,000 -10,000 response units would routinely result in a response of ϳ100 response units with 200 nM IGF-I.
Kinetic analyses were performed using a range of IGF-I or IGF-II concentrations (100, 50, 25, 12.5, and 6.25 nM) at 30 l/min. Surfaces were regenerated using 0.3 M sodium citrate, 0.4 M NaCl, pH 4.5. Data were analyzed using BIAevaluation 3.2 software using a 1:1 Langmuir binding model. This model describes a simple reversible interaction between two molecules in a 1:1 complex. In each of at least two repeat experiments, samples were injected in duplicate, and the response on the reference flow cell was subtracted.
Dissociation of Eu-IGF Ligands from IGF-1R-Following the binding of Eu-IGF-I or Eu-IGF-II ligand to immobilized receptors, plates were washed in TBST before the addition of 100 nM unlabeled IGF-I, IGF-II, or 24-60 mAb in 100 l of europium-binding buffer at room temperature. At various time points (0 -2 h), wells were washed once in water, and then 100 l/well enhancement solution was added and incubated for 10 min prior to determining the bound europium label by time-resolved fluorescence (47). The dose-response experiments for examining the accelerating effect of unlabeled ligand or anti-IGF-1R mAb 24-60 were performed with an increasing concentration of IGF-I or IGF-II (0.01 nM to 5 M) and 24-60 mAb (0.01 nM-5 M), respectively.

Synthesis, Expression, and Purification of Soluble IGF-1
Receptor Constructs-Two recombinant soluble IGF-I receptor (sIGF-1R) constructs were cloned into the pEEL5.HCMV-GS mammalian expression vector (51) driven by the human cytomegaloviral promoter. The first expression plasmid, designated pEEL-sIGF-1R, encoded the ectodomain of the human IGF-1R (sIGF-1R; 1-906 amino acids). In addition, a chimeric human IGF-1R-mouse Fc immunoglobulin fusion protein was encoded by the plasmid pEEL-sIGF-1R/Fc (for more detail, see the supplemental materials). Both of these cDNA expression constructs were transfected into hamster kidney fibroblast BHK-21 cells, and stable cell lines were generated. Cell lines expressing high levels of soluble IGF-1Rs (ϳ5 mg/liter) were identified by ELISA using conformation-dependent anti-IGF-1R antibodies (45).
Receptor Binding Assays-We initially established that europium labeling of IGFs resulted in a tracer with identical charac- teristics in receptor binding assays as the traditional 125 I-labeled IGFs. Competition binding assays using captured solubilized IGF-1R with Eu-IGF-I and 125 I-IGF-I (Fig. 1A) resulted in binding curves that were superimposable. The same result was seen with Eu-IGF-II and 125 I-IGF-II (data not shown). Therefore, europium-labeled tracers were used for all subsequent binding assays.
Competition binding using solubilized wtIGF-1R, sIGF-1R/ Fc, and sIGF-1R showed that both soluble ectodomain-containing fragments were identical in their binding affinities for IGF-I ( Fig. 1B and Table 1). However, they showed a slightly lower binding affinity (IC 50 ϭ 0.4 nM) compared with the solubilized wild type IGF-1R (IC 50 ϭ 0.23 nM). A similar result was seen with Eu-IGF-II competition binding assays (Fig. 1C), where the ectodomain fragments bound IGF-II with a 1.8-fold lower affinity than solubilized IGF-1R. All receptors had a 2-fold lower affinity for IGF-II compared with IGF-I. BIAcore analyses confirmed the competition binding assays and showed that sIGF-1R/Fc and sIGF-1R have the same relative affinities for IGF-I or IGF-II ( Fig. 1D and Table 2). Dissociation of Eu-IGF Ligands Bound to sIGF-1Rs-To investigate negative cooperativity of IGF-I and IGF-II binding to the soluble IGF-1Rs, we initially examined the dissociation of bound Eu-IGF-I in the presence or absence of unlabeled IGF-I (Fig. 2). In the absence of unlabeled ligand, the rate of dissociation of bound IGF-I from both soluble receptors (Fig. 2, B and C) was similar and slightly faster than from the solubilized wild type IGF-1R ( Fig. 2A). However, the presence of excess unlabeled IGF-I accelerated the dissociation from the sIGF-1R, with ϳ80% dissociation seen at 30 min compared with 55% from sIGF-1R/Fc and 50% from the wild type receptor (Fig. 2).
Although negative cooperativity of IGF-I binding to the wild type IGF-1R has been previously demonstrated, the effect of IGF-II on negative cooperativity has not been reported. Therefore, we also examined the dissociation of bound Eu-IGF-II in the presence or absence of IGF-II from the wild type and soluble IGF-1Rs (Fig. 2). In buffer alone, dissociation of bound Eu-IGF-II from sIGF-1R/Fc and sIGF-1R (Fig.  2, E and F) was similar and slightly faster than that seen for the solubilized wtIGF-1R (Fig. 2D). The presence of unlabeled IGF-II accelerated the dissociation from the solubi-

TABLE 2
BIAcore analysis of IGFs binding to soluble IGF-1Rs coupled to the biosensor surface The Langmuir 1:1 model was used to fit kinetic data in Fig. 1, D and E. Association rates (k a ) and dissociation rates (k d ) and dissociation constants (K D ϭ k d /k a ) are given. lized wtIGF-1R and both sIGF-1Rs (Fig. 2). The solubilized wtIGF-1R exhibited a slower dissociation rate in the presence of unlabeled IGF-II compared with sIGF-1R/Fc (ϳ30 and ϳ50% dissociation at 30 min, respectively), whereas dissociation from sIGF-1R was faster (ϳ80% dissociation at 30 min Fig. 2). In addition, the dissociation of bound IGF-II with unlabeled ligand from sIGF-1R was similar to that observed with bound IGF-I with unlabeled ligand from the sIGF-1R (ϳ80% dissociation at 30 min). It was concluded that both of the soluble IGF-1Rs exhibit negative cooperativity of IGF-I and IGF-II binding. Soos and Siddle 4 have previously demonstrated that the anti-IGF-1R mAb 24-60 accelerates the dissociation of 125 I-IGF-I from solubilized wtIGF-1R. This antibody binds to a similar epitope to that of ␣IR3 mAb on the IGF-1R (52) and inhibits IGF-I binding but not IGF-II binding (45). Therefore, we investigated the effect of the 24-60 antibody on the dissociation of bound Eu-IGF-I from the sIGF-1Rs (Fig. 2). As seen in Fig. 2, B and C, the 24-60 antibody accelerated the dissociation of bound IGF-I ligand from both the soluble sIGF-1R/Fc and sIGF-1R (ϳ25 and ϳ40% dissociation at 30 min, respectively), with a faster off rate observed for sIGF-1R. The effect of the 24-60 mAb on the dissociation rate of bound Eu-IGF-I to all receptors was less than that observed with unlabeled IGF-I. The addition of the 24-60 mAb did not affect the dissociation of bound IGF-II ligand from any of the IGF-I receptors (Fig. 2).

Analyte
In contrast to IR, the IGF-1R, does not exhibit a bell-shaped curve in dissociation experiments with increasing concentrations of IGF-I (53). To determine whether IGF-II behaved similarly to IGF-I, we derived dose-response curves for the dissociation of bound IGF-I and IGF-II to detergent solubilized wild type and soluble IGF-1Rs after 1 h of dissociation (Fig. 3). The accelerating effect of unlabeled IGF-I on bound Eu-IGF-I to either sIGF-1R/Fc or sIGF-1R, reached a maximum at 100 nM and remained stable at the highest concentration examined (5 M). In contrast to the soluble IGF-1Rs, a greater concentration of unlabeled IGF-I (1 M) was required to achieve maximal dissociation from the wtIGF-1R (Fig. 3A). A similar result was seen using Eu-IGF-II where the accelerating effect of unlabeled IGF-II bound to sIGF-1R and sIGF-1R/Fc reached a maximum effect at 100 nM and 1 M, respectively, and remained stable at 5 M (Fig. 3B). The maximum effect of unlabeled IGF-II on Eu-IGF-II bound to the wild type IGF-1R was observed at 1 M and also remained stable (Fig. 3B). The concentration at which 50% of the dissociable tracer is displaced from the receptor (EC 50 ) using IGF-I binding to sIGF-1R, sIGF-1R/Fc, and wtIGF-1R is ϳ2, ϳ6, and ϳ26 nM, respectively. Similarly, the EC 50 concentration for IGF-II binding to sIGF-1R, sIGF-1R/Fc, and wtIGF-1R is ϳ6, ϳ20, and ϳ100 nM, respectively. The faster accelerated dissociation observed with the soluble and wild type IGF-1Rs with unlabeled IGF-I compared with IGF-II is compatible with the higher affinity of the receptors for IGF-I.
We also derived dose-response curves for the dissociation of bound IGF-I from the wild type and soluble IGF-1Rs in the presence of mAb 24-60 at 4 h. The accelerating effect of mAb 24-60 on Eu-IGF-I bound to sIGF-1R reached a maximum at 100 nM, and the effect for sIGF-1R/Fc and wtIGF-1R reached a maximum at 1 M. All remained stable at the highest concentration (5 M) examined (data not shown).
Effect of Anti-IGF-1R Antibodies on Europium-labeled Ligand Binding to Soluble IGF-1Rs-Soos et al. (45) have previously characterized the effect of a number of conformation-dependent anti-IGF-1R antibodies on IGF-I binding to either membrane-associated or detergent-solubilized wild type IGF-1R. The binding epitopes for these antibodies are schematically presented in supplemental Fig. S1. However, to date, there have been no published data examining the effect of these antibodies on IGF-II binding. These antibodies are an extremely useful tool in examining the ligand binding properties of these soluble receptors. Therefore, we performed a study to investigate the effect of a number of stimulatory  and inhibitory (24-55, 17-69, and 24-60) anti-IGF-1R antibodies on both IGF-I and IGF-II binding to our soluble IGF-1Rs in immunocapture assays using 24-31 mAb and europium-labeled ligands (Fig. 4).
The anti-IGF-1R antibody 26-3 did not affect the binding of either Eu-IGF-I or Eu-IGF-II to the detergent-solubilized wild type or soluble IGF-1Rs in these assay conditions, not surprisingly, since this antibody binds to the same region on the IGF-1R as the capture antibody (Fig. 4). In contrast, the 16-13 antibody marginally reduced binding of IGF-I and IGF-II to 4 M. A. Soos and K. Siddle, unpublished data.  FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9 wild type and sIGF-1R/Fc (80 -90% of control, respectively) (Fig. 4). The antibodies 24-55 and 17-69 both inhibited Eu-IGF-I binding to the wild type and sIGF-1R/Fc receptors (40 -45% of control for sIGF-1R/Fc) but only marginally inhibited binding of IGF-I to sIGF-1R (80% of control) (Fig. 4A). These antibodies also inhibited IGF-II binding to wild type and sIGF-1R/Fc but to a lesser extent than that seen with IGF-I, whereas 24-55 and 17-69 reduced Eu-IGF-II binding to sIGF-1R/Fc to 64 and 53%, respectively (Fig. 4B). Although the binding of IGF-II to sIGF-1R was unaffected by 24-55, the anti-IGF-1R antibody 17-69 did reduce the binding of Eu-IGF-II to sIGF-1R (79% of control) (Fig. 4B). As expected, the 24-60 antibody did reduce the binding of Eu-IGF-I to all three IGF-1Rs, with the greatest effect observed with the soluble IGF-1Rs (sIGF-1R and sIGF-1R/Fc, 62 and 63%, respectively) but did not affect the binding of Eu-IGF-II to any of the IGF-1Rs examined (Fig. 4).

DISCUSSION
In the present study, we have recombinantly expressed and purified two soluble IGF-1R proteins (sIGF-1R and sIGF-1R/ Fc) to examine the requirements for the generation of a high affinity ligand binding receptor and negative cooperativity. Although the IGF-1R has been extensively studied in regard to IGF-I binding, similar studies performed with IGF-II are lacking in the literature. Without the availability of a crystal structure of the high affinity receptor-ligand complex, these sIGF-1Rs provide an invaluable tool to elucidate the requirements for high affinity ligand binding and negative cooperativity and also allow a comparison with the extensive studies performed on the structurally related IR.
We have generated a chimeric fusion protein composed of the extracellular IGF-1R truncated at the transmembrane domain fused to the mouse immunoglobulin Fc, designated sIGF-1R/Fc. This strategy has been successfully used to characterize the ligand binding determinants for a number of receptors, including the IR (21) and TNFR (54). Metabolic labeling experiments demonstrated that the sIGF-1R and sIGF-1R/Fc were synthesized as a proreceptor and then cleaved into the ␣and ␤oor ␤o-Fc subunits, respectively (supplemental Fig. S2). Secreted IGF-1R protein was produced and purified by onestep affinity chromatography to yield a homogeneous receptor sample, and gel filtration analysis demonstrated that these proteins were not subject to aggregation. These soluble IGF-1Rs were used to investigate the binding of IGF-I and IGF-II.
In conventional ligand binding assays, the sIGF-1R/Fc exhibited binding affinities for IGF-I and IGF-II that approached that of the wild type receptor (ϳ0.2-0.3 nM) (55,56). Also, the binding affinity of the sIGF-1R for IGF-I was similar to that previously reported (ϳ0.4 nM (57,58) and was 2-fold lower for IGF-II. The data obtained using BIAcore biosensor technology for immobilized sIGF-1R/Fc-binding IGF-I (5.3 nM) and IGF-II (10 nM) were similar to our previously reported binding affinities (50) and to that of a soluble IGF-1R/Z variant (23). Amine coupling of sIGF-1R/Fc to the sensor surface appeared to be limiting interactions between sIGF-1R/Fc and the IGFs or hindering the flexibility of the receptor. However, as previously seen, the relative binding affinities for IGF-I and IGF-II reflect those seen with conventional binding assay formats (50).
Using a panel of monoclonal antibodies that recognize specific conformational epitopes on the IGF-1R, we showed that the sIGF-1R and sIGF-1R/Fc maintained an overall structure similar to the wild type IGF-1R. Subtle differences in these receptor fragments could be detected when assessing the effect of the monoclonal antibodies on ligand binding. For example, the mAb 16-13, directed to the L1 domain of the IGF-1R, reduced ligand binding to the detergent-solubilized wild type and sIGF-1R/Fc but had little effect on IGF-I binding to the sIGF-1R. However, this difference may merely result from the lower affinity of the sIGF-1R for IGF-I than the sIGF-1R/Fc. This finding is consistent with the previous report by Soos et al. (45), who demonstrated that mAb 16-13 inhibits IGF-I binding to solubilized wild type IGF-1R. Interestingly, they also showed that the mAb 16-13 was stimulatory on membrane-associated receptors, suggesting that there is a difference in the conformation of the receptor between membrane-bound and solubilized receptors.
The mAbs 24-55 and 17-69 bind to a similar epitope located within the Fn1 domain of the IGF-1R (45). Although both mAbs exhibit inhibitory effects on IGF-I binding to the solubilized wtIGF-1R, they also reduce IGF-I binding to sIGF-1R/Fc and marginally affect binding to sIGF-1R. In addition, these two mAbs also inhibit IGF-II binding to both wtIGF-1R and sIGF-1Rs but to a lesser extent than that seen with IGF-I, suggesting that the ligand binding sites for IGF-I and IGF-II in that domain may be slightly different. The epitope of the inhibitory 24-60 mAb located within the CR of the IGF-1R (45) is similar to that of ␣IR3 and has recently been shown to overlap with residues important for IGF-I but not IGF-II binding (52). The mAb 24-60 affects the binding of IGF-I to both membrane-associated and solubilized receptors, with a more pronounced effect on membrane-associated receptors (45). In the current study, we have shown that 24-60 also inhibits IGF-I but not IGF-II binding to the sIGF-1Rs.
Our kinetic analyses using IGF-I as the competing ligand with the wtIGF-1R confirmed previous reports that a negative cooperativity mode of binding exists. Interestingly, IGF-II also induced negative cooperativity with the wtIGF-1R, an observation that has not previously been reported. This would suggest that both IGF-I and IGF-II binding to the IGF-1R uses a binding mechanism similar to that described previously for insulin binding to the IR (41). The mechanism proposed by DeMeyts (41) and Schäffer (59) to account for the properties of the IR suggested that ligand would need to contact binding sites on both ␣-subunits and that the IR dimer would have internal symmetry. At higher concentrations, a second molecule of ligand would be able to bind and thereby induce a structural change to accelerate the dissociation of the first molecule of ligand. Determination of the crystal structure of the IR ectodomain indicates that there are indeed two potential binding sites for insulin. These sites are composed of a binding region located within the L1 domain, which alone has low affinity for insulin (Site 1) and also includes the COOH-terminal region of the ␣-subunit, and a second region, which is located within the carboxyl-terminal surface of Fn1 that is required for high affinity binding (Site 2) (40). Therefore, according to the model and the IR crystal structure, high affinity ligand binding and negative cooperativity would require movement of the L1-CR module of one monomer toward the bottom of the Fn1 domain of the second monomer (40) (reviewed in Ref. 60).
Although both IGF binding to the IGF-1R and insulin binding to the IR use a mechanism involving negative cooperativity, there is a difference in binding seen between these receptors in dose-response curves for dissociation at high ligand concentrations. At very high insulin concentrations, more than one molecule can bind to the IR, and this results in a bell-shaped doseresponse curves for dissociation, whereas IGF-IR competition binding assays with IGF-I binding to the IGF-IR do not display this characteristic (53) (Fig. 3). Christoffersen et al. (53) concluded that IGF-I also has a bivalent binding mode and crosslinks two ␣-subunits at sites that are distinct from the IR binding sites. In this study, we show for the first time that the same is true for IGF-II binding to the IGF-IR, since we do not see a bell-shaped dose-response curves for dissociation.
An interesting observation made in this study is that both the sIGF-IR and the sIGF-IR/Fc exhibit negative cooperativity using both IGF-I and IGF-II as the competing ligand, although they both have lower binding affinities compared with wtIGF-IR. This result was unexpected, since sIR does not exhibit a negative cooperativity binding mode and only binds insulin with low affinity (41). Interestingly, despite all of the necessary components required for negative cooperativity being present in both the sIGF-IR and sIGF-IR/Fc, there are slight differences between these soluble receptors seen in dissociation assays with unlabeled ligands. sIGF-1R/Fc has a dissociation rate similar to that of wild type receptor in the presence of either unlabeled IGF-I or IGF-II, which is slower compared with sIGF-1R. These findings suggest that there are subtle differences in the kinetics of these two soluble receptors that are only revealed in these dissociation experiments. Further evidence for the difference between the sIGF-1Rs is seen from the dose-response dissociation curves, where more IGF-I or IGF-II is required to accelerate the dissociation from the sIGF-1R/Fc or solubilized wtIGF-1R than from sIGF-IR.
Similarly, studies on these sIGF-1Rs with the inhibitory anti-IGF-1R mAb 24-60 also showed that this antibody specifically affects the dissociation of bound IGF-I from both sIGF-1R/Fc and sIGF-1R. This finding suggests that this mAb can also exert effects on negative cooperative ligand-receptor interactions. The epitope for mAb 24-60 lies within the CR domain of the IGF-1R, where a specific IGF-I binding region is found. Therefore, our data suggest that an interaction with this domain is important for the negative cooperativity mode of binding IGF-I.
Although there are differences seen in the ligand binding and negative cooperative properties of IR ectodomain and the IGF-1R ectodomain, it is difficult to explain a possible mechanism for these differences. One clue comes from studies looking at the effect of disulfide bond reduction on ligand binding. Interestingly, the IR when reduced has a significantly lower affinity for ligand. However, studies of IGF-1Rs have shown conflicting data as to whether disulfide reduction of IGF-1R does affect ligand binding (61,62). The findings from our current study are in agreement with that of Feltz et al. (62), where, in contrast to that seen with the IR, disulfide reduction of the IGF-1R does not significantly affect ligand binding. The soluble IGF-1R proteins retain all of the properties associated with the holoreceptor (i.e. high affinity ligand binding and negative cooperativity). Therefore, the differences between these receptors may reside within the conformation of the domains relative to each other.
The importance of domain organization has been observed when comparing the binding properties of an artificial soluble IR protein (mIR.Fn0/Ex10), which exhibits high affinity ligand binding and negative cooperativity (19,20), whereas the sIR does not show negative cooperativity but contains all of the same domains. Presumably, the arrangement of these domains in the two receptor fragments is different and influences negative cooperativity. Receptor-specific sequence differences between the IR and IGF-1R do occur, and this may also influence the relative positioning of the receptor domains relative to each other and may be important in negative cooperativity. It has recently been shown that the L2/Fn1 region of the IGF-1R cannot replace the IR L2/Fn1, although this region appears to be similar in structure for both receptors based on homology predictions (36). Either subtle structural changes or specific amino acid differences account for the specificity within this region. The structural determination of the ligand-receptor complex of both receptors will hopefully explain these differences.
The current findings suggest that although some regions of the IGF-1R contribute to binding of both IGF-I and IGF-II, there are also differences in the interactions of the IGF-1R with these two ligands. These findings are consistent with the previously reported biological functions of IGF-I and IGF-II, which are overlapping but also can be distinct (3, 7, 9 -12). Elucidation of the identity of these IGF-I and IGF-II-specific residues on the IGF-1R is currently in progress and will be essential in the design of therapeutic agents directed at specifically targeting the action of these ligands.