A Novel Approach to Identify Two Distinct Receptor Binding Surfaces of Insulin-like Growth Factor II*

Very little is known about the residues important for the interaction of insulin-like growth factor II (IGF-II) with the type 1 IGF receptor (IGF-1R) and the insulin receptor (IR). Insulin, to which IGF-II is homologous, is proposed to cross-link opposite halves of the IR dimer through two receptor binding surfaces, site 1 and site 2. In the present study we have analyzed the contribution of IGF-II residues equivalent to insulin's two binding surfaces toward the interaction of IGF-II with the IGF-1R and IR. Four “site 1” and six “site 2” analogues were produced and analyzed in terms of IGF-1R and IR binding and activation. The results show that Val43, Phe28, and Val14 (equivalent to site 1) are critical to IGF-1R and IR binding, whereas mutation to alanine of Gln18 affects only IGF-1R and not IR binding. Alanine substitutions at Glu12, Asp15, Phe19, Leu53, and Glu57 analogues resulted in significant (>2-fold) decreases in affinity for both the IGF-1R and IR. Furthermore, taking a novel approach using a monomeric, single-chain minimized IGF-1R we have defined a distinct second binding surface formed by Glu12, Phe19, Leu53, and Glu57 that potentially engages the IGF-1R at one or more of the FnIII domains.

The mitogenic and metabolic activities of the IGFs and insulin result from their interaction with the type 1 IGF receptor (IGF-1R) and/or the exon 11-(IR-A) and exon 11ϩ (IR-B) isoforms of the insulin receptor (IR). These class II receptor tyrosine kinases exist at the membrane as preformed disulfidelinked homodimers composed of two ␣ and two ␤ subunits in a ␤-␣-␣-␤ arrangement (reviewed in Refs. 9 -11). The IR and IGF-1R share between 41 and 84% sequence similarity that is most pronounced in their tyrosine kinase domains. Despite the homology in sequence and structure between these receptors, each exhibits distinct ligand binding preferences. The IGF-1R and IR bind their cognate ligands with high affinity (IGF-I and insulin, respectively). Both the IGF-1R and IR-A bind IGF-II with high affinity and are capable of mediating IGF-II action (12). Interestingly, the IR-B has a low affinity for IGF-II. The discerning factors for this isoform discrimination are largely undefined but may involve steric hindrance between the 12 amino acids encoded by exon 11 of IR-B and the IGF-II C domain.
There are currently no structures of any of these ligand⅐receptor complexes. The binding of insulin to the IR is certainly the best characterized of these interactions. Insulin has two IR binding surfaces: the "site 1" binding surface lies within the insulin dimerization surface, whereas "site 2" overlaps its hexamer-forming surface. Insulin is proposed to crosslink opposite halves of the insulin receptor dimer through these two receptor binding surfaces (13,14). The stoichiometry of binding at physiological insulin concentrations is 1:1. However, each IR has two potential ligand binding pockets both also consisting of two ligand binding surfaces (site 1 and site 2 of one monomer and site 1Ј and site 2Ј of the other, which combine to give two identical binding pockets, site 1/2Ј and site 1Ј/2). This putative arrangement is consistent with the three-dimensional crystal structure of the IR ectodomain (15). Evidence from ligand binding studies suggests that insulin may cross-link only one pair of binding surfaces at a time, with binding of a second molecule to an unoccupied site accelerating the dissociation of the first in a negatively cooperative fashion (9). Site 1 on the IR is proposed to contain ligand binding determinants from the L1 and FnIII-2 insert domains (probably operating in trans between opposing monomers); site 2, though less well defined, is proposed to involve loops at the junction of the FnIII-1 and FnIII-2 domains (9, 16 -19).
The interaction of IGF-I with the IGF-1R uses a similar binding mechanism to the insulin-IR interaction, with equivalent binding surfaces on the ligands and receptors (1, 20 -22). An additional interaction of the IGF-I C domain with the cysteine rich (CR) domain of IGF-1R contributes to high affinity binding (20).
In contrast to our knowledge of IGF-I and insulin receptor interactions there is relatively little known about the interaction of IGF-II with the IGF-1R and IR. Sequence and structural similarities suggest that IGF-II shares similar IR and IGF-1R binding surfaces (and similar modes of binding) to insulin and IGF-I. Three IGF-II residues that correspond to insulin's site 1 binding surface, i.e. Val 43 (Val A3 ), Phe 26 (Phe B24 ), and Tyr 27 (Phe B25 ) (insulin homologues in parentheses) have previously been shown to be important for IR and IGF-1R binding (23). The IGF-II C domain is largely responsible for the differences in specificity between IGF-I and IGF-II for the IR and IGF-1R (24), although, unlike the IGF-I C domain, alanine scanning mutagenesis of the IGF-1R and IR suggests it does not contact the CR domain of either receptor (21,25).
In the present study we have mapped by site-directed mutagenesis the IGF-1R and IR binding surfaces of IGF-II. This has allowed the first comprehensive comparison of the receptor binding surfaces of IGF-II, IGF-I, and insulin. Furthermore, we have taken a novel approach using a minimized IGF-1R (see supplemental Fig. S1) to define two receptor binding surfaces on IGF-II. This has allowed us to identify subtle differences between the ligand binding surfaces of all three ligands, which could contribute toward receptor binding specificity. In addition we have investigated the possible contribution of the IGF-1R FnIII domains to IGF-II binding. Table S1) were purchased from Geneworks Pty Ltd. (Adelaide, South Australia). Restriction enzymes were from New England Biolabs (Hitchin, UK) or Geneworks Pty Ltd. Long TM R 3 IGF-I was purchased from Novozymes GroPep Pty Ltd. (Adelaide, South Australia). Greiner Lumitrac 600 96-well plates were from Omega Scientific (Tarzana, CA). Human insulin was purchased from Novo Nordisk (Bagsvaerd, Denmark). The DELFIA europium-labeling kit and DELFIA enhancement solution were purchased from PerkinElmer Life Sciences. Eu-IGF-II was produced as described by Denley et al. (24) according to the manufacturer's instructions. IGF-I mono-iodinated at Tyr 31 ( 125 I-Tyr 31 IGF-I) (26) was a gift from Novo Nordisk A/S. P6 IGF-1R cells (BALB/c3T3 cells overexpressing the human IGF-1R) (27) were a kind gift from Prof. R. Baserga (Philadelphia, PA). IGF-1R-negative cells overexpressing the IR-A (R Ϫ IR-A) or the IR-B (R Ϫ IR-B) were generated as described (24).

Materials and Cell Lines-Oligonucleotides (supplemental
Construction of Expression Plasmids Encoding Human IGF-II Analogues-The IGF expression vector was developed by King et al. (28), and IGF-II cDNA was introduced into the vector as previously described (29). The QuikChange site-directed mutagenesis kit (Stratagene) was used to incorporate the 10 different mutations into IGF-II (supplemental Table S1). Resultant constructs were transformed into Escherichia coli JM101 (lac Iq) for expression. IGF-II mutants were expressed as fusion proteins with the first 11 amino acids of porcine growth hormone ([Met 1] pGH (1-11)) after isopropyl ␤-D-thiogalac-toside induction. Inclusion bodies were isolated as previously described (24).
Purification of IGF-II Analogues-The purification of the IGF-II analogues used in this study was as described in Delaine et al. (29). Purified proteins were analyzed by mass spectroscopy and N-terminal sequencing (Dr. Chris Bagley and Chris Cursaro, Adelaide Proteomics Facility) and were shown to have the correct masses and to be Ͼ95% pure. All IGF-II analogues maintained the same fold as native IGF-II as determined by far-UV CD spectral analysis as previously described (29) (see supplemental Fig. S2). Quantitation of analogues was performed by comparing analytical C4 highperformance liquid chromatography profiles with profiles of standard Long TM R 3 IGF-I preparations (24).
Immunocaptured IGF-1R and IR Binding Assays-Receptor binding was measured essentially as described by (24). Briefly, R Ϫ IR-A, R Ϫ IR-B, or P6 IGF-1R cells were serum-starved for 4 h before lysis in lysis buffer (20 mM HEPES, 150 mM NaCl, 1.5 mM MgCl 2 , 10% (v/v) glycerol, 1% (v/v) Triton X-100, 1 mM EGTA, and 1 mM phenylmethylsulfonyl fluoride, pH 7.5) for 1 h at 4°C. Lysates were centrifuged for 10 min at 2,200 ϫ g, then 100 l was added per well to a white Greiner Lumitrac 600 96-well plate previously coated with anti-IR antibody 83-7 (30) or anti-IGF-1R antibody 24-31 (31) as appropriate. Approximately 500,000 fluorescent counts of Eu-IGF-II (IGF-1R and IR-A binding) or Eu-insulin (for IR-B binding) were added to each well along with increasing concentrations of unlabeled competitor in a final volume of 100 l and incubated for 16 h at 4°C. Wells were washed four times with 20 mM Tris, 150 mM NaCl, and 0.1% (v/v) Tween 20 (TBST), then twice with water, and then DELFIA enhancement solution (100 l/well) was added. Time-resolved fluorescence was measured using 340 nm excitation and 612 nm emission filters with a BMG Lab Technologies Polarstar fluorometer (Mornington, Australia). Assays were performed in triplicate at least three times.
Mini-IGF-1R Binding Assays-To construct an expression vector for the mini-IGF-1R (L1, CR, and L2 domains fused to the CTa peptide, residues 692-706, see supplemental Fig. S1) an oligonucleotide cassette encoding the CTa peptide preceded by an in-frame BamHI site and followed by a stop codon and an XbaI site was ligated into the BamHI and XbaI sites of pcDNA3.1zeo(ϩ) (Invitrogen). An in-frame BamHI site was introduced into the IGF-1R cDNA after the codon for Asp 461 , the C terminus of the IGF-1R L2 domain, by site-directed mutagenesis (32). The resulting HindIII-BamHI fragment, encoding the N terminus of the IGF-1R, including the L1, CR, and L2 domains, was subcloned into the pcDNA3.1zeo(ϩ) plasmid containing the CTa peptide cassette to produce pCDZ mini-IGF-1R.
The recombinant secreted mini-IGF-1R protein was expressed by transient transfection of 293PEAK cells as previously described (20). Competitive binding assays with 125 I-Tyr31 IGF-I and IGF-II analogues were performed on mini-IGF-1R from conditioned media harvested from transfected cells immobilized in microtiter plates coated with antireceptor monoclonal antibody 24 -31 as previously described (20). Dissociation constants were obtained by non-linear regression analysis of assay data by the method of Wang (33).
Whole Cell Binding Assays-R Ϫ IR-A or P6 IGF-1R cells were serum-starved for 4 h and then harvested using cell disrupting buffer (40 mM Tris, pH 7.5, 10 mM EDTA, 150 mM NaCl), washed twice in ice-cold Dulbecco's minimal essential medium (Invitrogen) and then resuspended in assay buffer (100 mM Hepes, 120 mM NaCl, 5 mM KCl, 1.2 mM MgSO 4 , 8 mM glucose, and 0.5% bovine serum albumin). 10 5 cells were incubated in a 1.5-ml centrifuge tube with Eu-IGF-II and competing IGF-II in a total volume of 300 l of assay buffer overnight at 4°C. Cells were pelleted at 1025 ϫ g for 1.5 min, and the supernatant was aspirated. Cells were washed with 500 l of cold TBS and pelleted, and the supernatant was aspirated again. Pellets were incubated with Enhancement solution (100 l) in the dark for 30 min and then transferred to a white Greiner Lumitrac 600 96-well plate. Time-resolved fluorescence was measured as described above. Assays were performed in triplicate at least three times.
We have demonstrated that the IGF-2R does not interfere in these binding assays as Thr 16 3 Ala IGF-II and Phe 19 3 Ala IGF-II behave identically to IGF-II (Table 1) and neither bind the IGF2R (29).
IGF-1R and IR-A Phosphorylation Assays-Receptor phosphorylation was detected essentially as described by (24). Briefly, P6 IGF-1R cells (2.5 ϫ 10 4 cells/well) or R Ϫ IR-A cells (5 ϫ 10 4 cells/well) were plated in a Falcon 96 well flat bottom plate and grown overnight at 37°C, 5% CO 2 . Cells were starved in serum-free medium for 4 h before treatment with IGF-II or IGF-II analogue in 100 l of Dulbecco's minimal essential medium with 1% bovine serum albumin for 10 min at 37°C, 5% CO 2 . Cells were lysed with ice-cold lysis buffer containing 2 mM Na 3 VO 4 and 100 mM NaF, and receptors were captured onto white Greiner Lumitrac 600 96-well plates pre-coated with anti-IR antibody 83-7 (500 ng/well) (30) or anti-IGF-1R antibody 24 -31 (250 ng/well) (31) as appropriate and blocked with 20 mM Tris, 150 mM NaCl, and 0.1% (v/v) Tween 20 (TBST)/ 0.5% bovine serum albumin (30, 31). Following overnight incubation at 4°C, the plates were washed three times with TBST. Phosphorylated receptor was detected by incubation at room temperature for 2 h with europiumlabeled antiphosphotyrosine antibody PY20 (76 ng/well, PerkinElmer Life Sciences). Wells were washed four times with TBST, twice with water, and then DELFIA enhancement solution (100 l/well) was added. Time-resolved fluorescence was detected as described above. Assays were performed in triplicate at least three times.

RESULTS
Production and Characterization of IGF-II Analogues-To define and compare the receptor binding surfaces of IGF-II with those of insulin, two groups of IGF-II analogues were produced (Fig. 1). The site 1 analogues, Val 14 3 Thr, Gln 18 3 Ala, Phe 28 3 Leu, and Val 43 3 Met IGF-II, are so called because they incorporate amino acid substitutions at residues corresponding to insulin's site 1 binding surface (Val B12 , Tyr B16 , Tyr B26 , and Val A3 in insulin, respectively). The site 2 analogues contain alanine substitutions at the residues Glu 12 , Asp 15 , Phe 19 , Asp 52 , Leu 53 , and Glu 57 , which correspond to insulin's second binding surface (His B10 , Glu B13 , Leu B17 , Ser A12 , Leu A13 , and Glu A17 , respectively). Following successful expression in E. coli, analogues were purified, and their structural integrity was analyzed by far-UV CD spectroscopy. All analogues had CD spectra indistinguishable from that of IGF-II (29) (supplemental Fig. S2), indicating correct folding and maintenance of structural integrity.
IGF-1R Binding Properties of IGF-II Analogues-Competition binding assays were used to measure the affinities of IGF-II and IGF-II analogues for IGF-1R immunocaptured from P6 IGF-1R cell lysates (Fig. 2, A and B) and for IGF-1R present on P6 IGF-1R cells (Fig. 2, C and D), because there is evidence that some analogues behave differently in the two assay formats (34,35). The affinities of IGF-II for either immunocaptured or membrane-bound IGF-1R were consistent with the literature (24, 36) ( Table 1). As has previously been reported (24, 36) the IGF-II affinity for immunocaptured receptors was lower than IGF-1R on intact cells while relative binding affinities of the different analogues generally remained the same. To allow direct comparison of IGF-II analogue affinities between the two assays we have chosen to present these as affinities relative to wild-type IGF-II (Table 1).
With the exception of Gln 18 3 Ala, which displayed only a modest decrease in affinity (41-52% of IGF-II) for the IGF-1R, the site 1 analogues bound to this receptor with drastically reduced affinities. This was most marked for Val 43 3 Met IGF-II with an affinity Ͻ1% that of wild-type IGF-II. Phe 28 3 Leu and Val 14 3 Thr IGF-II bound the IGF-1R with ϳ10% the affinity of IGF-II (less on the whole cell receptor). Among the FIGURE 1. The putative receptor binding surfaces of IGF-II. The amino acid sequence alignment of IGF-II, IGF-I, and insulin (lower panel) shows the conserved residues boxed in light gray, residues conserved between IGF-II and IGF-I boxed in dark gray, and the domain structure below. Insulin residues important for IR binding and defined as site 1 residues are highlighted in bold type, and those defined as site 2 residues are underlined and in italics (9). Those IGF-II residues mutated in this study are highlighted with an asterisk and are indicated as putative site 1 (dark gray) and site 2 (light gray) residues on the structure of IGF-II (upper panel). The coordinates for IGF-II are from the pdb file 1IGL. site 2 analogues, most detrimental was alanine substitution of residue Asp 15 , which caused a 10-fold decrease in affinity for the IGF-1R. Introducing alanine into positions 53 and 57, respectively, resulted in 2.2-to 2.9-fold reductions in affinity, but substitution of Asp 52 with Ala had essentially no effect (87-110% of IGF-II). Interestingly, Glu 12 3 Ala-IGF-II displayed only a 2.5-fold decrease in affinity for immunocaptured IGF-1R but a significantly greater 6-fold reduction in affinity for whole cells. The behavior of Phe 19 3 Ala IGF-II was also different across the two assay formats, with a 2.8-fold decrease in affinity observed on the immunocaptured IGF-1R but no effect in the whole cell assay (112% of IGF-II).
Mini-IGF-1R Binding Properties of IGF-II Analogues-To investigate if any of the IGF-II residues mutated in the present study interact with regions within the FnIII domains of the IGF-1R (proposed to contain the receptor's site 2 binding surface), the affinities of the analogues for a "mini-IGF-1R" were measured and are presented relative to IGF-II in Table 1. The mini-IGF-1R consists of only the L1, CR, and L2 domains fused to the 16 amino acids from the C terminus of the ␣ subunit (CTa 692-706) (see supplemental Fig. S1). As the mini-IGF-1R contains only the site 1 component of the receptor, we hypothesized that only analogues with a mutation in site 1 would have a decrease in affinity relative to IGF-II. The affinity of mini-IGF-1R for the wild-type IGF-II is shown in Table 1 and is consistent with previous reports (37).
All of the site 1 analogues displayed a similar decrease in affinity for the mini-IGF-1R as they did for the wild type IGF-1R (Table 1); as for the full-length receptor, Val 43 3 Met IGF-II was the worst of the analogues to compete with Eu-IGF-II for mini-IGF-1R binding. There again was little effect of alanine substitution at position 18 on mini-IGF-1R binding. Among the site 2 analogues, Asp 15 3 Ala showed a similar reduction in affinity (8-to 10-fold) for both the mini-and wild-type IGF-1R. As was seen with the wildtype IGF-1R, mutation of Asp 52 to Ala did not reduce binding to the mini-IGF-1R. Interestingly, however, alanine substitution into positions 12, 19, and 53 did not reduce binding to the mini-IGF-1R despite a 2-to 2.5-fold reduction in wildtype IGF-1R binding described above (Table 1). Glu 57 3 Ala also showed a significantly reduced affinity for the wild-type immunocaptured IGF-1R (2.9-fold) but only a 1.9-fold reduction in mini-IGF-1R affinity.

TABLE 1 Relative binding of IGF-II and the IGF-II analogues for the IGF-1R and IR isoforms
Binding affinities of the analogues to IGF-II were derived from the IC 50 values for all but the mini-IGF-1R binding data, which were derived using K d values. Relative binding is expressed as a percent of IGF-II Ϯ S.E. Data is derived from at least 3 separate experiments performed in triplicate.  IR Binding Properties of IGF-II Analogues-Further competition assays were used to measure the affinities of IGF-II analogues for the IR-A present on intact R Ϫ IR-A cells (Fig. 3, C and D, and Table 1) and immunocaptured IR-A (Fig. 3, A and B, and Table 1) and IR-B (Table 1). As seen for the IGF-1R, the affinities of the wild-type ligands for the IR-A were higher in the whole cell receptor assay than those for the immunocaptured receptor (Table 1), and these are consistent with previous reports (24,36).

IGF-II
The relative binding affinities for the immunocaptured IR-A are IGF-II ϭ Gln 18 3 Ala ϭ Asp 52 3 Ala ϭ Glu 12 3 Ala Ͼ Phe 19 3 Ala Ͼ Leu 53 3 Ala Ͼ Glu 57 3 Ala ϭ Asp 15 3 Ala Ͼ Val 14 3 Thr Ͼ Phe 28 3 Leu (Table 1). A similar order of relative affinities was seen for the IGF-1R, although Gln 18 3 Ala, in contrast to its 2-fold decrease in affinity for the IGF-1R, bound the IR-A with wild-type affinity (129 -144%). In addition, the IR-A and IR-B binding properties of the IGF-II analogues were essentially identical; no analogue exhibited Ͼ2-fold difference in its binding to the IR-A compared with the IR-B ( Table 1).
The affinities of the analogues relative to IGF-II for the immunocaptured IR-A and IR-A on cells were, for the most part, comparable ( Fig. 3 and Table 1). Glu 12 3 Ala, however, was an exception as this analogue exhibited a 10-fold decrease in affinity for the IR-A in the R Ϫ IR-A whole cell receptor assay but had a very similar affinity (128%) to IGF-II for the immunocaptured receptor. The opposite was true for Leu 53 3 Ala IGF-II (89.7% of IGF-II for R-IR-A cells and 25.2% of IGF-II for immunocaptured IR-A). Asp 15 3 Ala and Val 14 3 Thr also showed a slight difference (ϳ2-fold) across the two assay formats.
Receptor Phosphorylation Properties of IGF-II Analogues-To correlate receptor binding with receptor activation, the abilities of IGF-II and the analogues to stimulate IGF-1R and IR-A phosphorylation were assessed in kinase receptor activation assays (Fig. 4). Maximal IGF-1R phosphorylation was achieved following 10-min treatment of P6 IGF-1R cells with 100 nM IGF-II; higher concentrations resulted in less than maximal phosphorylation (Fig. 4, A and B). In contrast, IGF-II was a less potent activator of the IR-A than IGF-1R and IGF-II-induced stimulation of the IR-A was still approaching its maximum at 1 M (Fig. 4, C and D).

. Activation of the IGF-1R and IR-A by IGF-II and IGF-II analogues. P6 IGF-1R cells and R Ϫ IR-A cells were serum starved for 4 h and then incubated with increasing concentrations of IGF-II (E), Val
affinities. Accordingly, Val 14 3 Thr, Phe 28 3 Leu, Val 43 3 Met, and Asp 15 3 Ala were the least potent of the analogues on both the IGF-1R and IR-A. In agreement with their modest changes in binding affinity, the dose-response curves of Phe 19 3 Ala, Asp 52 3 Ala, and Leu 53 3 Ala were similar to that of IGF-II for both receptors. Gln 18 3 Ala IGF-II also had doseresponses curves similar to IGF-II but did not stimulate phosphorylation to the same maximum (80% of IGF-II at 100 nM on IGF-1R; 75% of IGF-II at 1 M on IR-A). Glu 57 3 Ala displayed reduced potency compared with IGF-II on both receptors, however, this was more marked on the IR-A. Interestingly, Glu 12 3 Ala IGF-II displayed receptor phosphorylation properties that were not in accordance with its affinities for both the IGF-1R and IR-A. Maximal Glu 12 3 Ala IGF-II stimulation of IGF-1R phosphorylation was achieved at 100 nM (greater than at 300 nM) as was seen for IGF-II; however, its maximal response was Ͻ40% of that achieved by IGF-II. On the IR-A, Glu 12 3 Ala IGF-II stimulated phosphorylation to Ͻ30% that of wild-type IGF-II.

DISCUSSION
Unlike insulin and IGF-I, with which it shares structural homology and a number of biological activities, IGF-II binds both the IGF-1R and IR-A with high affinity. However relatively little was previously known about the IGF-II receptor binding interfaces. The present study systematically identifies specific residues of IGF-II required for these interactions using a series of IGF-II analogues inspired by the known receptor binding surfaces of insulin.
Two distinct receptor binding surfaces have been described for insulin with each proposed to contact opposite halves of the insulin receptor dimer. Site 1 comprises (IGF-II homologues in parenthesis): Gly A1 (Gly 41 28 ) and, as the first identified of insulin's receptor binding surfaces, is often referred to as the "classic" binding surface. A further six residues, Ser A12 (Asp 52 ), Leu A13 (Leu 53 ), Glu A17 (Glu 57 ), His B10 (Glu 12 ), Glu B13 (Asp 15 ), and Leu B17 (Phe 19 ), collectively known as site 2, were later shown by alanine scanning mutagenesis to contribute to IR binding (9,14,35,38). We recombinantly expressed and purified a series of IGF-II analogues with amino acid substitutions at positions corresponding to the two receptor binding surfaces of insulin and determined their receptor binding affinities. This allowed a direct comparison between the binding surfaces of insulin and IGF-II for the IR, and between IGF-I and IGF-II binding surfaces for the IGF-1R.
Given evidence in the literature that some analogues may behave differently in the two assay formats (22,35), the affinities of all IGF-II analogues were measured using both immunocaptured receptors (IGF-1R, IR-A, or IR-B) and whole cell receptor binding assays with cells expressing IGF-1R or IR-A. This observation was affirmed in the present study where receptor binding by the Glu 12 3 Ala analogue, for instance, was more affected when binding whole cells than immunocaptured receptor, an observation also made for the equivalent analogue of IGF-I (Glu 9 3 Ala IGF-I) (22). The opposite was true for Phe 19 3 Ala on the IGF-1R and Leu 53 3 Ala on the IR-A where a greater effect was seen in immunocapture assays than whole cell binding assays. The interaction of the IGFs and insulin with these receptors is a dynamic process that almost certainly involves flexible regions of the receptor and structural change, which is apparently not fully recapitulated in the absence of membrane anchorage. It has been shown that steric interactions between the two ␣␤ receptor halves (apart from specific ligand binding determinants) are necessary to produce high affinity binding of ligand to the IR (39). It is possible that detergent lysis of receptors prior to immunocapture could disturb these interactions and thus disrupt the receptor dimer interface and in turn ligand binding. Alternatively, although the antibodies used to immunocapture the receptor do not directly compete with ligand binding (30,31), immunocapture of the receptor may impose restrictions on how the receptor is able to change its conformation upon ligand binding. Consistent with either mechanism, and in agreement with previous reports (24,36), in our hands IGF-II had a higher affinity for the IGF-1R and IR-A on whole cells compared with immunocaptured receptor.
Among the site 1 residues analyzed in the present study Val 14 , Phe 28 , and Val 43 all proved critical to the binding of IGF-II to both the IGF-1R and IR isoforms, with Ն10-fold decreases in affinity resulting from their mutation. These residues form a continuous hydrophobic patch together with the aromatic sidechains of Phe 26 and Tyr 27 previously shown to be important for IR and IGF-1R binding (23). This hydrophobic cluster appears to provide the major contribution to the free energy of binding to both receptors; the corresponding residues are also important for the binding of both insulin to the IR and IGF-I to the IGF-1R. Indeed, mutation at Val A3 (Val 43 in IGF-II) in insulin is associated with diabetes mellitus (Wakayama insulin) (40), whereas Val mutation to Met at the equivalent residue 44 in IGF-I was identified in a patient with severe growth retardation (41,42). In the case of insulin, Val A3 at the N terminus of the A chain is largely buried within a cleft between the A and B chains. It has been proposed that displacement of the C terminus of the B-chain upon insulin binding exposes Val A3 (and Ile A2 ) and thus allows it to directly contact the IR (43). Photoaffinity cross-linking studies at Val A3 point to its interaction with the FnIII-2 insert domain of the IR (40,43), which forms part of the IR site 1 binding region along with residues of the L1 domain. The NMR structure of IGF-II shows that the corresponding residue Val 43 in IGF-II is largely surface-exposed (5, 6) and so is already in a position to interact with the IR (and IGF-1R) and thus would not necessarily require an equivalent structural change. Consistent with the importance of Val 14 to the IGF-1R and IR binding of IGF-II, IR binding is highly sensitive to mutation at the equivalent Val B12 residue of insulin with extensive mutagenesis pointing to the importance of both side-chain bulk and polarity (44,45). The equivalent Val 11 in IGF-I however appears less critical to its interaction with the IGF-1R. Val 11 3 Thr IGF-I binds the IGF-1R with 3.8-fold lower affinity than IGF-I (46) compared with the 10-to 30-fold decrease in affinity for Val 14 3 Thr IGF-II analyzed in the present study (Table 1).
Gln 18 in IGF-II lies at the periphery of the hydrophobic patch described above and appears to play a role in modulating the MARCH 20, 2009 • VOLUME 284 • NUMBER 12 specificity of IGF-II for the IGF-1R and IR. Mutation of this residue to alanine was without effect on IR binding but caused a 2-fold decrease in IGF-1R binding, indicating a difference in the way in which Gln 18 interacts with these two receptors. Mutation of the corresponding Gln 15 in IGF-I to alanine has a similar effect on IGF-1R binding as seen here for IGF-II (48).

Defining Two IGF-II Receptor Binding Surfaces
That IGF-II can accommodate Gln and Ala equally in its interaction with the IR represents a small point of difference between the interactions of insulin and IGF-II with the IR. The corresponding insulin residue, Tyr B16 , has been extensively mutated, and there is a preference for an aromatic side chain and alanine over glutamine for IR binding (44,45,47). Mutation of Tyr B16 to glutamine results in an 11-fold decrease in affinity for the IR-A, whereas mutation to alanine lead to only a 3-fold decrease (45). Tyr B16 cross-links to the L1 domain beta helix that forms part of site 1 (44). Mutation of Gln 15 in IGF-I to Tyr produces a 3.8-fold increase in IR-A binding and when mutated in combination with Thr 4 3 His, Phe 49 3 Thr, Ser 51 3 Ile this analogue has an equal affinity to insulin for the IR-A (34). Accordingly, we would predict that substitution of Gln 18 in IGF-II with Tyr would result in an increase in IR-A binding.
Five of the six site 2 analogues (Glu 12 3 Ala, Asp 15 3 Ala, Phe 19 3 Ala, Leu 53 3 Ala, and Glu 57 3 Ala) had significant (Ͼ2.5-fold) decreases in affinity for the IGF-1R and IR isoforms. Only alanine substitution of Asp 52 was without effect on receptor binding as was the case for the equivalent IGF-I analogue (Asp53 3 Ala IGF-I) (22); the equivalent insulin analogue produced a ϳ2.5-fold decrease in affinity for the IR-A on IM9 cells (35). Of these five, it was substitution of the charged residues Glu 12 , Asp 15 , and Glu 57 that caused the most significant decreases in affinity for both IGF-1R and IR-A, as is the case for IGF-I binding the IGF-1R (22). In insulin it is Glu B13 and the hydrophobic Leu B17 and Leu A13 of the six site 2 residues that contribute most to IR binding (equivalent to Asp 15 , Phe 19 , and Leu 53 in IGF-II) (35). This suggests that within site 2 IGF-II shares more in common with IGF-I, both in terms of sequence identity and contribution of specific residues to binding, than with insulin and, in agreement with the work of Benyoucef et al. (19), may explain why IGF-I and IGF-II can bind hybrid IGF-1R/IR receptors, whereas insulin does so with only low affinity (49).
All of the analogues showed similar affinities, relative to IGF-II, for both the IR-A and IR-B. We can therefore conclude that these residues do not contribute to the differing affinity of IGF-II for these receptors.
The designation of insulin residues (and by inference IGF-I and IGF-II) into sites 1 and 2 is largely historical with site 1 identified first, and site 2 not until years later (9). Direct evidence for interaction of site 1 and site 2 residues with two distinct receptor binding regions is fundamentally lacking. In this study we decided to define our IGF-II analogues based upon their abilities to bind two different forms of the IGF-1R: the full-length IGF-1R and the mini-IGF-1R construct containing only the site 1 component of IGF-1R. The mini-IGF-1R comprises the first three domains (L1-CR-L2) fused to residues 692-706 of the ␣ subunit (CTa) and thus lacks the loops of the FnIII-1 and FnIII-2 domains that are proposed to accommodate site 2 on the receptor (9, 17, 37). We hypothesized that residues that interact with site 2 of the receptor (or indeed any part of the receptor outside of the L1-CR-L2-CTa) would, when mutated, only show a decrease in affinity for the full-length and not the mini-IGF-IR. Conversely, residues that interact with site 1 of the receptor would show similar decreases in affinity for both the full-length and mini-IGF-1R. Of the nine analogues in this study that had a significant (Ͼ2-fold) decrease in affinity for the full-length IGF-1R (in either assay format), five showed a similar decrease in affinity for the mini-IGF-1R. These were Val 14 3 Thr, Gln 18 3 Ala, Phe 28 3 Ala, and Val 43 3 Met of the site 1 analogues and, perhaps unexpectedly, Asp 15 3 Ala from site 2. It is thus unlikely that Asp 15 is part of the site 2 surface but is rather an extension of site 1 and more likely interacts with the L1 or CTa. Asp 15 is part of the B chain ␣ helix and is adjacent to Val 14 , Gln 18 , and the hydrophobic cluster of residues that form site 1.
In contrast, site 2 analogues Phe 19 3 Ala and Leu 53 3 Ala bound the mini-IGF-1R with essentially wild-type affinity (Ͻ2fold change). Glu 12 3 Ala did not show a decrease in affinity relative to IGF-II for the mini-IGF-1R, whereas Glu 57 3 Ala had only a 1.9-fold decrease. However, all of these site 2 analogues had a significant (Ͼ2-fold) decrease in affinity for the full-length IGF-1R. These results support the notion that Glu 12 , Phe 19 , Leu 53 , and Glu 57 form part of a second binding surface, which might interact with the FnIII domains of the IGF-1R.
The IGF-II site 1 and 2 residues are mapped onto a model of IGF-II in complex with the IGF-1R L1-CR-L2 domains in Fig. 5. The model is adapted from the Epa et al. model (50) of the IGF-I⅐IGF-1R (L1-CR-L2) complex, which assumes that Tyr 24 of IGF-I contacts the L1 domain and Arg 36 and Arg 37 contact the CR domain, and is consistent with insulin:IR cross-linking studies, which suggest that the IR site 1 consists of the L1 and FnIII-2 insert domains (40,44). This orientation of IGF-II within the L1-CR-L2 domains leaves the IGF-II site 2 residues identified in the present study (Glu 12 , Phe 19 , Leu 53 , and Glu 57 )  accessible to bind to the proposed FnIII site 2 on the receptor and is similar to that proposed by Gauguin et al. (22) for IGF-I.
Unfortunately, the affinity of IGF-II for the equivalent mini-IR construct is not sufficiently high to enable the analogous set of experiments with the insulin receptor. 3 Given, however, the similarity in effect on binding affinity of the majority of analogues (all but Gln 18 3 Ala-IGF-II) on both the IGF-1R and IR, similar results would be expected. At this stage this approach to define two binding regions using the two receptor forms has not been reported for insulin and IGF-I analogues. It will be interesting to see whether the equivalent analogues behave the same way or whether small differences in the ligandreceptor interactions exist.
With the exception of Glu 12 3 Ala, all the IGF-II analogues showed potency in stimulating IGF-1R and IR-A phosphorylation that correlated with their ability to bind these receptors. Glu 12 3 Ala in contrast showed reduced potency in IGF-1R and IR-A phosphorylation that was unexpected in light of its affinity. As well as reduced potency, Glu 12 3 Ala also showed a reduction in maximum response, stimulating IGF-1R phosphorylation to just 40% that of wild-type IGF-II. The mechanism for this loss of efficacy is the subject of ongoing investigation in our group; however, it is tempting to speculate that mutation at this position alters the ability or efficiency of that ligand to crosslink the two receptor halves and to thus activate the tyrosine kinase activity of the receptor. Glu 12 3 Ala was also interesting in that it showed a much greater decrease in affinity for the IGF-1R and IR in the whole cell binding assay than when binding to immunocaptured receptor. A similar observation was recently made with the equivalent IGF-I analogue Glu 9 3 Ala IGF-I, which interestingly also was unable to accelerate dissociation of IGF-I from the IGF-1R (22). The equivalent His B10 residue in insulin (which plays a role in coordinating Zn 2ϩ into the insulin hexamer) is important for insulin receptor binding and insulin activity. Substitution with Asp, which makes it more "IGF-like" at that position gives rise to a "super-active" insulin (51). Therefore this residue plays a vital role for all three ligands in receptor binding and subsequent activation.
With the elucidation of the IGF-1R and IR binding surfaces of IGF-II the present study allows a comparison of the receptor binding surfaces of insulin, IGF-I and IGF-II. We conclude that IGF-II uses almost identical residues for binding the IGF-1R and IR-A. Furthermore, IGF-II has broadly similar IGF-1R and IR binding surfaces to IGF-I and insulin with subtle differences in the relative contribution of particular amino acids (1,9,22,35). Similar observations have been made on the receptor side of the interaction, with alanine scanning mutagenesis of the IGF-1R and IR demonstrating that, within broadly similar ligand binding regions of the receptor, there are subtle differences in the contribution of certain residues to the binding of the different ligands (20,21,25,(52)(53)(54). Evidence in the literature suggests that any of IGF-I, IGF-II, and insulin can with small modifications have the specificity of one of the other ligands suggesting that there is significant plasticity within the IR and IGF-1R to allow high affinity binding of ligand to be variously achieved (24,34). It is an important distinction, however, that affinity does not necessarily correlate with biological outcome. For example, despite binding equipotency, insulin and the synthetic peptide mimetic S597 initiate divergent signaling outcomes through the IR (55,56). The ultimate goal now is to understand how different ligands bind the same receptor to give different biological outcomes. An understanding of this process might then allow the design of analogues to produce a desired biological outcome.