Alanine scanning mutagenesis of a type 1 insulin-like growth factor receptor ligand binding site.

The high resolution crystal structure of an N-terminal fragment of the IGF-I receptor, has been reported. While this fragment is itself devoid of ligand binding activity, mutational analysis has indicated that its N terminus (L1, amino acids 1-150) and the C terminus of its cysteine-rich domain (amino acids 190-300) contain ligand binding determinants. Mutational analysis also suggests that amino acids 692-702 from the C terminus of the alpha subunit are critical for ligand binding. A fusion protein, formed from these fragments, binds IGF-I with an affinity similar to that of the whole extracellular domain, suggesting that these are the minimal structural elements of the IGF-I binding site. To further characterize the binding site, we have performed structure directed and alanine-scanning mutagenesis of L1, the cysteine-rich domain and amino acids 692-702. Alanine mutants of residues in these regions were transiently expressed as secreted recombinant receptors and their affinity was determined. In L1 alanine mutants of Asp(8), Asn(11), Tyr(28), His(30), Leu(33), Leu(56), Phe(58), Arg(59), and Trp(79) produced a 2- to 10-fold decrease in affinity and alanine mutation of Phe(90) resulted in a 23-fold decrease in affinity. In the cysteine-rich domain, mutation of Arg(240), Phe(241), Glu(242), and Phe(251) produced a 2- to 10-fold decrease in affinity. In the region between amino acids 692 and 702, alanine mutation of Phe(701) produced a receptor devoid of binding activity and alanine mutations of Phe(693), Glu(693), Asn(694), Leu(696), His(697), Asn(698), and Ile(700) exhibited decreases in affinity ranging from 10- to 30-fold. With the exception of Trp(79), the disruptive mutants in L1 form a discrete epitope on the surface of the receptor. Those in the cysteine-rich domain essential for intact affinity also form a discrete epitope together with Trp(79).

The insulin-like growth factors I and II are essential for normal fetal and post-natal growth (1). They were originally identified as circulating polypeptides with potent mitogenic activity, which mediated many of the actions of growth hormone, and were later shown to be structurally homologous to proinsulin. It is now apparent that these growth factors are produced by many cell types and have paracrine and autocrine as well as endocrine functions. Targeted disruption of the gene for IGF-I 1 in transgenic mice results in both embryonic and post-natal growth retardation (2). In contrast, the effects of disruption of the IGF-II gene are confined to growth retardation during the embryonic period (2). In addition to being mitogens, it is now evident that these peptides play a crucial role in cell survival (3) and contribute to transformation and the maintenance of the malignant phenotype in many tumor systems (4). However, despite extensive study, the signal transduction mechanisms underlying the biological effects of these peptides remain to be elucidated.
The mitogenic effects of these growth factors appear to be mediated by receptors belonging to the insulin receptor subclass of receptor tyrosine kinases (for review see Ref. (5)). The type 1 IGF receptor binds both peptides with high affinity; the affinity for IGF-I being greater than that for IGF-II. Transgenic experiments indicate that the growth-promoting effects of both peptides can be mediated by this receptor (2,6). Such studies also point to the role of a second receptor in mediating the mitogenic effects of IGF-II (2,6), and recent in vitro studies indicate that this is the A isoform of the insulin receptor (7); this receptor binds IGF-II with high affinity and can mediate the growth-promoting effects of the peptide (8).
The receptors in this family are dimeric protein-tyrosine kinases with significant homology (5). In higher vertebrates there are three known members, the insulin receptor (9,10), the type 1 IGF receptor (11), and the orphan insulin receptorrelated receptor (12). They are dimeric M r 350,000 glycoproteins composed of two disulfide-linked monomers. Each monomer is in turn composed of an N-terminal ␣ subunit and a C-terminal ␤ subunit, which are linked by a single disulfide. The ␤ subunit is both extracellular and intracellular with a single ␣ helical transmembrane domain. The intracellular portion contains the tyrosine kinase catalytic domain. The structure of this domain of the insulin receptor has been determined at high resolution in both the basal and active state (13,14). Comparative homology modeling suggests that the extracellular portion of the receptors is composed of seven distinct structural domains (15)(16)(17). At the N terminus there are two homologous globular domains flanking a cysteine-rich domain. The remainder is formed from three fibronectin III repeats, the second of which contains a 100-amino acid insert of undetermined structure.
Recently a high resolution crystal structure of an N-terminal fragment of the insulin-like growth factor I receptor (amino acids 1-460) has been reported (18). The molecule is composed of an extended bilobed structure composed of the two globular L domains flanking the cysteine-rich domain with dimensions * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  of 40 ϫ 48 ϫ 105 Å. The N-terminal globular domain contacts the cysteine-rich domain along its length. In contrast there is minimal contact between the C-terminal domain and the cysteine-rich domain. In the crystal structure, L1 and L2 occupy very different positions relative to the cysteine-rich domain. However, this could be an artifact of crystal packing in this fragment, and the position of L2 may be very different in the native molecule (18). It is possible that it is rotated into a position similar to that of L1 in relation to the cysteine-rich domain. However, irrespective of this, a cavity of ϳ24-Å diameter occupies the center of the molecule and possibly represents a binding pocket.
Each L domain resembles a loaf of bread with dimensions 24 ϫ 32 ϫ 37 Å and is formed from a single right-handed ␤ helix capped at the ends by short ␣ helices and disulfide bonds. The base of the loaf is formed from a six-stranded ␤ sheet five residues in length. Both sides are formed from ␤ sheets three amino acids in length, and the top is composed of irregular loops connecting the short ␤ strands. As predicted from sequence comparisons, the cysteine-rich domain is composed of repetitive modules resembling parts of laminin and the tumor necrosis factor receptor. These form a rod-like structure connecting the two globular L domains from which a large mobile loop projects into the putative binding pocket.
Despite this wealth of structural detail very little is known about the precise location and nature of the ligand binding site(s). Studies with chimeric insulin/IGF-I receptors suggest that the C terminus of the cysteine-rich domain is a major determinant of IGF-I binding specificity (19 -22). Its location in the N-terminal fragment of the IGF-I receptor is consistent with it forming part of a ligand binding pocket (18). Furthermore, alanine mutagenesis of residues in the LI N-terminal globular domain indicate that this also forms part of the ligand binding site (23). These studies also demonstrated that a Cterminal peptide of the ␣ subunit, amino acids 692-702, is involved in IGF-I binding (23). Furthermore, fusion of this C-terminal fragment to the N-terminal 460 amino acids results in a recombinant protein, which binds IGF-I with an affinity similar to that of the full-length secreted recombinant extracellular domain (24,25), suggesting that these elements are sufficient to form an intact ligand binding site.
Alanine scanning studies of the structurally related insulin receptor have demonstrated that determinants in the L1 domain and in the C terminus of the ␣ subunit (amino acids 705-715) are sufficient to form a ligand binding site (26 -28). In the present study we have used alanine mutagenesis to localize the equivalent binding site of the IGF-I receptor. The results of the studies indicate that it is formed from three elements, the first in the L1 domain, the second predominantly in the cysteine-rich domain and the third at the C terminus of the ␣ subunit between amino acids 692 and 702.

MATERIALS AND METHODS
General Procedures-All molecular biological procedures, including agarose gel electrophoresis, restriction enzyme digestion, ligation, bacterial transformation, and DNA sequencing were performed by standard methods (29). All oligonucleotides were purchased from DNA Technology (Aarhus, Denmark). Restriction and modifying enzymes were from New England BioLabs (Beverly, MA). Recombinant IGF-I (receptor grade) was from GroPep (Adelaide, Australia). High performance liquid chromatography-purified mono-iodinated [ 125 I-Tyr 31 ]IGF-I (30) was from Novo Nordisc A/S. Protease inhibitors were from Roche Molecular Biochemicals (Mannheim, Germany). Medium and serum for tissue culture were from Life Technologies A/S (Tåstrup, Denmark). PEAK Rapid cells (293 cells constitutively expressing SV40 large T antigen) were purchased from Edge Biosystems (Gaithersburg, MD). The mammalian expression vector pcDNA3-zeo(ϩ) was from Invitrogen (San Diego, CA). The hybridoma secreting monoclonal antibody 24-31 directed toward the IGF-I receptor ␣ subunit (31) was a generous gift of Drs. M. Soos and K. Siddle (University of Cambridge, UK). Protein A-purified IgG from the hybridoma medium was kindly provided by by Dr. P. Jorgensen (Novo Nordisc A/S, Bagsvaerd, Denmark). cDNAs encoding both full-length and recombinant secreted extracellular domain of the IGF-I receptor were as previously described (23).
Oligonucleotide-directed Mutagenesis-Oligonucleotide-directed mutagenesis was performed by the method of Kunkel (32). Uracil containing single-stranded DNA prepared from phage rescued from Escherichia coli CJ236 transformed with a cDNA encoding the full-length IGF-I receptor cloned into the phagemid pTZ18U. Restriction sites were deleted or introduced with the specific mutation to facilitate screening of the mutants. Successful mutagenesis was confirmed by DNA sequencing.
Expression of Mutant Receptor cDNAs-Recombinant mutant secreted IGF-I receptor cDNAs were reconstructed in the plasmid pcDNA3-zeo(ϩ) for expression. DNA for transfection was prepared from 10-ml overnight cultures by a boiling hexadecyltrimethylammonium bromide method (33) followed by purification using QIAwell strips. The mutant receptor cDNAs were expressed transiently in Peak Rapid cells (293 cells constitutively expressing SV40 large T antigen) by transfection using Fugene 6 (Roche Molecular Biochemicals, Mannheim, Germany) according to the manufacturers' directions. Conditioned medium was harvested 4 days post-transfection and, if necessary, concentrated prior to assay using Centriprep 30 centrifugal concentrators (Millipore, Bedford, MA).
Receptor Binding Assays-Soluble IGF-I receptor binding assays were performed using a modification of the microtiter plate antibody capture assay that we have described previously (23). Microtiter plates (Nunc Maxisorb, Roskilde, Denmark) were incubated overnight at 4°C with anti-IGF-I receptor antibody 24-31 IgG (100 l/well of 46 g/ml solution in phosphate-buffered saline). Washing, blocking, and receptor binding were as previously described. Competitive binding assays with labeled and unlabeled IGF-I were carried out as done previously, except that the incubation was for 16 h at 25°C.
Binding data were analyzed by computer fitting to a one-site model to obtain the K d of the expressed protein.
Western Blotting-Western blotting of conditioned media with an anti-IGF-I receptor ␣ subunit peptide (amino acids 31-50) antibody (Santa Cruz Biotechnology, Santa Cruz, CA) was performed using previously described procedures (27). Blots were visualized by chemiluminescence using reagents from Pierce (Rockford, IL).

L1 Domain
Mutagenesis-In previous alanine-scanning mutagenesis studies of the L1 domain of the structurally related insulin receptor, we have demonstrated that the residues critical for insulin binding are those exposed side chains located in (i) the ␤ sheet forming the wall of the central cavity, i.e. the base of the L1 domain and (ii) the adjacent third ␤ sheet (26). Thus we systematically mutated the exposed residues in the equivalent regions of the IGF-I receptor to alanine; glycines, prolines, and cysteines were not mutated to avoid potential structurally deleterious effects. The mutant IGF-I receptor cDNAs were transiently expressed in 293 PEAK cells. To confirm and evaluate expression, conditioned medium from transfected cells was analyzed by immunoblotting with an antibody directed toward the N terminus of the ␣ subunit of the IGF-I receptor. In conditioned medium from transfections with all mutant cDNAs except those with mutations of Leu 32 , Leu 33 , Ser 35 , Tyr 54 , and Thr 93 , a M r 135,000 protein, representing the IGF-I receptor ␣ subunit was detectable in comparable amounts to that in conditioned media from cells transfected with wild type receptor cDNA (data not shown). The failure to detect receptor in blots of the medium from cells transfected with the Leu 32 , Leu 33 , and Ser 35 mutant cDNAs despite the presence of IGF-I binding activity in the medium (see below) is presumably because the epitope of the antibody used for these blots is directed toward amino acids 31-50 of the receptor ␣ subunit. In contrast, in medium from cells transfected with Tyr 54 and Thr 93 , there was neither detectable IGF-I binding activity nor receptor detectable by blotting despite the presence of an immunoreactive M r 160,000 protein corresponding to receptor precursor in detergent lysates of transfected cells (data not shown), a finding that has previously been observed with mutations that impair appropriate folding of this form of the homologous insulin receptor (26).
Equilibrium binding studies were performed on conditioned media from transfected cells to characterize IGF-I binding to wild type and mutant receptors. As previously described (23), IGF-I binding to recombinant secreted receptor displayed simple kinetics and could be best fitted to a single-site model (data not shown). Computer analysis indicated a single population of binding sites with a K d of 0.67 Ϯ 0.06 ϫ 10 Ϫ9 M (mean Ϯ S.E., n ϭ 8). It should be noted that this value is higher than we have previously reported (23) and probably reflects changes in the source of IGF-I, assay conditions, and computerized analysis of binding data. Because studies utilizing alanine-scanning mutagenesis have demonstrated that meaningful changes in affinity, produced by a single alanine substitution, range from 2to 100-fold (34), in the experiments described below we regarded any mutant with a greater than 2-fold increase in K d , i.e. K d greater than 1.3 ϫ 10 Ϫ9 M, as exhibiting a significant disruption of IGF-I-receptor interactions.
The results of our analyses of the IGF-I receptor L1 domain alanine mutants are shown in Fig. 1. Data are expressed as a ratio of the dissociation constant of the mutant receptor to that of the wild type receptor. Conditioned medium from cells transfected with two mutant cDNAs discussed above, Tyr 54 and Thr 93 , failed to exhibit sufficient [ 125 I-Tyr 31 ]IGF-I binding to permit accurate quantitative analysis of the expressed receptors' binding properties. As discussed, immunoblotting failed to reveal any evidence of secretion of these receptors from the cell, indicating that the mutant proteins were malfolded.
Of the other 27 alanine mutants, 10 caused a significant impairment of IGF-I binding, i.e. greater than 2-fold increase in K d . Eight of these 10 mutants (Asp 8 , Asn 11 , Tyr 28 , His 30 , Leu 33 , Leu 56 , Phe 58 , and Arg 59 ) are located in the N-terminal half of the L1 domain and result in increases in K d for IGF-I ranging from 3-fold (Phe 58 ) to 9-fold (Asp 8 ). One mutant (Trp 79 ), which results in a 3-fold increase in K d for IGF-I is located in the bulge region, amino acids, in the fourth turn of the ␤ helix. The last mutant (Phe 90 ) is located in the C-terminal half of the domain and results in a 23-fold increase in K d .
Cysteine-rich Domain Mutagenesis-With the exception of glycines, prolines, and cysteines, we mutated to alanine all the residues, in the cysteine-rich domain, that are predicted to be accessible to ligand, on the basis of the published structure of the receptor N-terminal fragment (18), i.e. in the region between amino acids 240 and 284. When expressed in 293 PEAK cells, all alanine mutants in this region appeared to be secreted normally, on the basis of immunoblotting of conditioned media and cell lysates as described above (data not shown). Thus it is likely that there was no major perturbation of receptor structure attributable to the mutations.
Equlibrium binding assays were performed on conditioned media from the transfected cells to characterize the IGF-I binding properties of the mutant receptors. Fig. 2 summarizes these results. Only 4 out of a total of 26 mutations produced significant decreases in affinity for IGF-I. These are all located at the N terminus of the region analyzed and produce decreases of 2to 6-fold; the largest decreases of 6-and 4-fold are produced by the mutations of Phe 241 and Glu 242 , respectively, to alanine.
Mutagenesis of the C Terminus of the ␣ Subunit-We systematically mutated amino acids 692 to 702 to alanine and expressed the resulting mutant cDNAs in 293 PEAK cells. As previously reported, when analyzed by immunoblotting as described above, all mutants appeared to be folded and secreted normally (data not shown) (23). Equilibrium binding studies were performed on conditioned media to evaluate the IGF-I binding properties of the mutants. The results of these experiments are shown in Fig. 3. Three mutants Phe 695 , Ser 699 , and Val 702 appeared to be without effect on affinity for IGF1. Mutation of Phe 701 produced a receptor that had no detectable IGF-I binding activity. Mutation of Phe 692 , Glu 693 , Asn 694 , Leu 696 , His 697 , Asn 698 , and Ile 700 to alanine results in decreases in affinity for IGF-I ranging from 10-fold (Asn 698 ) to 29-fold (Phe 692 ).

DISCUSSION
In the present study, using structure-directed and alaninescanning mutagenesis, we have identified 22 amino acids in the ␣ subunit of the IGF-I receptor, which appear to be functional determinants of its ligand binding site. Ten of these are located in the L1 domain, four in the cysteine-rich domain, and eight at the C terminus of the ␣ subunit. The amino acids in the L1 and cysteine-rich domains are organized into two discontinuous epitopes. The first of these is located in the N-terminal part of the L1 domain and is composed of the amino acids Asp 8 , Asn 11 , Tyr 28 , His 30 , Leu 33 , Leu 56 , Phe 58 , Arg 59 , and Phe 90 . They form a footprint on the ␤ sheet, which forms the base of the domain and the wall of the putative ligand binding cavity (Fig. 4). Mutation of each of these residues to alanine produces a 2-to 10-fold decrease in affinity with the exception of Phe 90 , which produced a 23-fold decrease. The second epitope is formed from the L1 residue Trp 79 , which is located in the loop/bulge (amino acids 78 -85) between the ␤ sheets forming one of the sides and the base of the domain and the cysteine-rich domain residues Arg 240 , Phe 241 , Glu 242 , and Phe 251 ; Trp 79 is in contact with Glu 242 (Fig. 4). These form a small patch on the cysteine-rich domain adjacent to the base of the L1 domain. Mutation of each of these residues results in a 2-to 6-fold decrease in affinity. The observation that the disruptive mutants in both domains form contiguous patches on the protein surface surrounded by non-disruptive mutations provides strong evidence that they are contact sites for IGF-I (34). The third element of the binding site is formed from amino acids Phe 692 , Glu 693 , Asn 694 , Leu 696 , His 697 , Asn 698 , Ileu 697 , and Phe 701 , located at the C terminus of the ␣ subunit, a region of the receptor for which there is no structural information. Thus, whether all the side chains forming this binding element are directly involved in interaction with IGF-I or whether the effects of the mutations are indirect cannot be ascertained, until the structure of this region of the receptor is determined. Nonetheless, the magni-

FIG. 2. Alanine-scanning mutagenesis of the cysteine-rich domain.
293PEAK cells were transfected with cDNAs encoding alanine mutants of ligand-accessible amino acids of the cysteine-rich domain of the recombinant secreted IGF-I receptor (amino acids 240 -284) prepared by oligonucleotide-directed mutagenesis. Four days after infection conditioned medium from the cells was harvested and the expression and IGF-I binding of the mutant receptors were evaluated as described under "Materials and Methods." The dissociation constant was determined by computer fitting to a single-site model. The dissociation constant of the wild type receptor determined under these conditions was 0.67 Ϯ 0.06 ϫ 10 Ϫ9 M (mean Ϯ S.E., n ϭ 8). The results are expressed as the ratio of the dissociation of the mutant to that of the wild type and represent the mean Ϯ S.E. of three independent determinations. The amino acids mutated to alanine are designated by the single-letter code.

FIG. 3. Alanine-scanning mutagenesis of amino acids 692-702.
293PEAK cells were transfected with cDNAs encoding alanine mutants of amino acids 692-702 of the recombinant secreted IGF-I receptor prepared by oligonucleotide-directed mutagenesis. Four days after infection conditioned medium from the cells was harvested and the expression and IGF-I binding of the mutant receptors were evaluated as described under "Materials and Methods." The dissociation constant was determined by computer fitting to a single-site model. The dissociation constant of the wild type receptor determined under these conditions was 0.67 Ϯ 0.06 ϫ 10 Ϫ9 M (mean Ϯ S.E., n ϭ 8). The results are expressed as the ratio of the dissociation of the mutant to that of the wild type and represent the mean Ϯ S.E. of three independent determinations. The amino acids mutated to alanine are designated by the single-letter code. *, receptor affinity too low to be accurately determined (mutant K d (MUT)/wild type K d (WT) Ͼ 250).

FIG. 4. Structure of the functional epitopes of the L1 and cysteine-rich domains.
The C␣ backbone of the the L1 and cysteine-rich domains is shown as a ribbon representation. The amino acids mutated are shown in space-filling representation. Alanine mutants of amino acids in yellow had no effect on affinity. Alanine mutants of those in orange produced a 2-to 10-fold reduction in affinity, and those in red had Ͼ10-fold reduction. Alanine mutations of residues in white resulted in receptors that were not secreted in detectable amounts. This figure was prepared with the Swiss PDB Viewer (53). tude of the observed effects of the mutations of amino acids in this region on affinity for IGF-I indicate that this element of the binding site appears to provide the majority of the free energy of the interaction with IGF-I, as we have previously observed for the homologous insulin receptor (28).
The data summarized above imply that this ligand binding site of the IGF-I receptor is formed from 14 -22 amino acids and thus implies that a similar number of IGF-I side chains form the cognate receptor binding site. Although this may seem surprising, because it indicates a quarter to a third of the surface of IGF-I is involved in binding to the receptor, a similar proportion of the surface of the homologous insulin molecule forms the receptor binding site (35,36). This is consistent with the crystal structure of the N terminus of the IGF-I receptor; the binding pocket is large enough to completely accommodate either molecule (18). This finding is also supported by recent molecular reconstructions of the insulin-receptor complex from electron microscopic studies that indicate that the binding cavity engulfs the ligand molecule, with at least a third of the ligand side chains being involved in interactions with the receptor (37).
Both the insulin and IGF-I receptors and insulin and IGF-I exhibit high homology (9 -11), although each receptor is highly specific for its cognate ligand (19,23,38). It is therefore pertinent to consider whether the findings of the present study provide any clues either as to whether similar binding mechanisms are employed by each receptor/ligand pair or to the molecular basis for this specificity. Comparison with the results of previous studies from this laboratory (26 -28), using directed alanine scanning in the absence of structural information to characterize the ligand binding site of the insulin receptor, reveals that, in L1 in each receptor, there is an overlapping epitope that is located in the three N-terminal strands of the ␤ sheet forming the base of the domain and some of the residues at the N terminus of the corresponding adjacent ␤ strands forming the side of the domain. This is somewhat larger in the IGF-I receptor and is composed of the side chains of Asp 8 , Asn 11 , Tyr 28 , His 30 , Leu 33 , Leu 56 , Phe 58 , Arg 59 , and Phe 90 (Fig.  5). In the insulin receptor (26), it consists of the side chains of Asp 12 , Arg 14 , Asn 15 , Gln 34 , Leu 36 , and Phe 64 (Fig. 5). Arg 10 , Asn 11 , and Phe 58 , corresponding to Arg 14 , Asn 15 , and Phe 64 in the insulin receptor, are conserved in both proteins but exhibit strikingly different functional properties when mutated to alanine; the IGF-I receptor mutations only result in a 2-to 10-fold decrease in affinity, whereas those in the insulin receptor lead to greater than 100-fold decrease (28).
In the insulin receptor two other groups of side chains in the L1 domain, Phe 39 and Tyr 67 , and Phe 89 , Asn 90 , and Tyr 91 , have been implicated in ligand binding (26). Although mutation of Phe 39 and Tyr 67 to alanine produce a significant decrease in affinity for insulin (10-to 20-fold for Phe 39 and 2-fold for Tyr 67 (26)), it is unlikely, on the basis of the IGF-I structure (18) and that of the homologous model of the insulin receptor that we have produced, that they play a direct role in binding. They are both located in ␤ sheets forming the side of the L1 domain and are two residues away from its base, in which all the other binding determinants are located (see Fig. 5). This is somewhat surprising because Phe 39 has been implicated in conferring insulin specificity on the insulin receptor (39). It must therefore be concluded that this is an indirect effect and that further studies, including a high resolution structure of the ligand receptor complex will be necessary to resolve this issue.
The insulin receptor residues Phe 89 , Asn 90 , and Tyr 91 form a second epitope on L1 (26). These amino acids are located in the loop/bulge just N-terminal to the fourth strand of the ␤ sheet forming the base of the domain (18). Asn 90 and Tyr 91 are conserved in the IGF-I receptor, but they do not appear to play a role in ligand binding. The only residue of the corresponding region of IGF-I receptor that we have shown to be important for binding is Trp 79 . However, this participates in an epitope that is formed largely from cysteine-rich domain residues.
The IGF-I receptor cysteine-rich domain epitope is composed of the side chains of Arg 240 , Phe 241 , Glu 242 , and Phe 251 in addition to Trp 79 from the L1 domain. Of these residues, only Phe 241 and Phe 251 are conserved in the insulin receptor (Phe 247 and Phe 257 , respectively (9, 10)). Although we have not formally evaluated whether this region of the insulin receptor is involved in ligand binding, we feel it is unlikely. On the basis of the IGF-I structure, the distance from the L1 residues involved in ligand binding is greater than 30 Å, which is significantly larger than the largest dimension (20 Å) of the putative receptor binding site of the insulin molecule (35,36).
The C-terminal ␣ subunit element of the ligand binding site is highly conserved in both receptors yet the residues composing this epitope appear to play very different roles in ligand binding in each receptor (Fig. 6), as we have discussed in a previous study (23). However, despite these findings, chimeric minireceptors composed of the amino acids 1-470 of the insulin receptor and the C-terminal epitope of the insulin receptor or IGF-I receptor bind insulin with nearly identical affinity (24). Minireceptors formed from amino acids 1-460 of the IGF-I receptor and the C-terminal epitope of either receptor also exhibit the same behavior, i.e. nearly identical affinity for IGF-I (24).
Both receptors bind their cognate ligands with similar affinities, but as we have discussed, there appear to be significant quantitative differences between the alanine scanning results for each receptor; there appear to be significantly more amino acids in the functional epitopes of the insulin receptor whose mutation to alanine results in a profound impairment of insulin binding (26). One possible explanation for this finding is that many of these mutants of the insulin receptor are producing their quantitative effects both indirectly by causing intramolecular perturbation of the structure of the binding site as well as directly perturbing side-chain interactions with the  (53). Both receptor domains are shown viewed from the base, and the C␣ backbone is shown in ribbon representation. Amino acids whose mutation to alanine compromises affinity for ligand are shown in space-filling representation. Residues are colored according to the magnitude of the effect of the mutation on affinity; yellow corresponds to a 2-to 10-fold decrease, orange to a 10-to 100-fold decrease, and red to Ͼ100-fold decrease.
ligand. This appears to be likely, because we have previously demonstrated in a quantitative characterization of these alanine mutants, that the sum of the changes in free energy of binding attributable to each mutant is far in excess of the free energy of the interaction of the receptor with ligand (28). Whereas we have not been able to obtain similar data for the IGF-I receptor, because we have been unable to accurately quantitate the free energy change attributable to the mutation of Phe 701 to alanine, this does not appear to be the case for the IGF-I receptor, where the sum of the free energy changes attributable to the other mutations is more commensurate with the free energy of the interaction with IGF-I (data not shown). A second possibility is that main-chain interactions play a significant role in the IGF-I/receptor interaction; their contribution to the interaction would not be detectable by the methods used here. It is perhaps noteworthy in this context that recent alanine scanning studies of the interaction of IGF-I with IGF binding protein-3 failed to implicate any IGF-I side chains in this interaction (40).
From the above it is clear that, despite the homologies between insulin and IGF-I and between their cognate receptors and despite the quantitative similarities between their interactions, each ligand-receptor pair seems to employ a distinct binding mechanism. This is surprising because of the close homology of the ligands (41), and particularly so, in view of the degree of conservation of the amino acids forming insulin's receptor binding site (35,36) in IGF-I (41), 8 out of 13 residues were absolutely conserved and, of the remainder, only one was non-conservatively substituted. Mutational studies of IGF-I structure and function have been limited; the only conserved or conservatively substituted residues, corresponding to those critical for insulin binding, that have been studied are Val 11 (42,43), Phe 23 (44), Tyr 24 and Tyr 60 (45), which are equivalent to Val B12 , Phe B24 , Phe B25 , and Tyr A19 , respectively, in insulin. In addition, systematic mutation of these residues to alanine has not been undertaken. However, despite these limitations, these studies do provide some relevant insights. Mutation of Val 11 to alanine reduces the affinity of IGF-I by only 60% (42), whereas the equivalent mutation of Val B12 of insulin reduces its affinity by 99% (36). In contrast modification of Phe 23 of IGF-I to glycine reduces its affinity by 98% (44), whereas the affinity of the equivalent insulin analogue, Gly B24 insulin for the insulin receptor is nearly that of native insulin (46). This clearly confirms that the molecular mechanisms underlying interaction with the receptor are different for each ligandreceptor pair.
Studies with chimeric insulin/IGF-I receptors have indicated that the specificity of the receptors for IGF-I is mediated by the sub-domain of the cysteine-rich domain between amino acids 190 and 290 (19,22). The cysteine-rich domain binding epitope that we have identified resides in this region. As already discussed, two of the five residues forming this epitope Phe 241 and Phe 251 are conserved in the insulin receptor and mutation of the Trp 79 , Arg 240 , and Glu 242 to alanine have only relatively small effects on affinity for IGF-I. This is certainly insufficient to account for the 100-fold difference in affinity observed for the non-cognate ligand (19). However, recent experimental evidence suggesting another mechanism has been presented (47). Hoyne et al. (47) have demonstrated that the loop 255-265, which is adjacent to the cysteine-rich domain functional epitope that we have identified, modulates insulin/IGF-I affinity. Substitution of this loop for the equivalent loop of the insulin receptor increases the affinity of that receptor for IGF-I. It is quite clear from the results of the present study that this loop does not participate directly in the binding of IGF-I, and by implication its role in modulating affinity must be indirect. The loop in the insulin receptor exhibits significant charge differences from that of the IGF-I receptor: Two lysine and two arginine residues in the insulin receptor loop compared with two glutamate and one aspartate residues in the IGF-I receptor loop. It has been suggested that it might therefore produce an unfavorable charge environment in the putative binding cavity for IGF-I binding (18). Also, in the insulin receptor this loop is significantly longer than that of the IGF-I receptor and could thus possibly sterically impair access of the bulkier IGF-I molecule to the binding site. This mechanism is supported by the finding that reduction in the length of the C-domain of IGF-I, reducing the bulk of the molecule, specifically increases its affinity for the insulin receptor (48). Further support is provided by the finding that shortening the insulin receptor loop significantly increases its affinity for IGF-I. 2 Further experimental study will be necessary to fully elucidate this mechanism. Several reports suggest that the full-length IGF-I receptor, like the related insulin receptor, has a higher affinity for IGF-I than we have reported here for the secreted recombinant receptor (49) and exhibits complex binding kinetics with curvilinear Scatchard plots (49,50) and negative cooperativity (50,51). Two models have been proposed to explain such behavior in this family of receptors and ligands (49,52). Both propose that there are two distinct ligand binding sites of differing 2 J. Whittaker and A. V. Groth, unpublished observations. FIG. 6. Comparison of the C-terminal epitopes of the IGF-I and insulin receptors. The effects of alanine mutations of amino acids 692-702 of the IGF-I receptor and of amino acids 705-715 of the insulin receptor on affinity for their cognate ligands are compared. Results are presented as ratios of the dissociation constant of the mutant receptor to that of the wild type receptor. Data for the insulin receptor mutants were taken from Ref. 27. *, receptor affinity too low to be accurately determined (mutant K d (MUT)/ wild type K d (WT) Ͼ 250). affinities on each ␣ subunit and that there are two receptor binding sites on each ligand molecule. Ligand sequentially binds to one site on the first ␣ subunit and then to the second site on the other ␣ subunit, cross-linking the two heterodimers and generating the high affinity component of the receptorligand interaction. Subsequent binding of a second ligand molecule disrupts the cross-linking of the first and accelerates its dissociation (negative cooperativity). In the recombinant secreted form of the receptor, ligand only binds to one of the binding sites (the higher affinity binding site) and thus displays a lower affinity than the native receptor and simple binding kinetics. Data presented by Schaffer (49) for the holoreceptor indicate that the affinity of the high affinity binding site of the solubilized receptor is significantly greater than that of the low affinity, and thus it would be expected that mutations compromising the affinity of the soluble receptor, i.e. of the higher affinity binding site, would also compromise the affinity of the native receptor. Although we have not yet formally compared the effects of mutations on the affinities of the secreted and native IGF-I receptors, our previous study (26) with the insulin receptor indicate that this is indeed the case.
In summary, we have demonstrated by mutational analysis that the ligand binding site of the secreted extracellular domain of the IGF-I receptor is composed of three elements. The first of these, an epitope at the N terminus of the molecule, overlaps the corresponding epitope of the insulin receptor and exhibits significant identity to it. The second is in the cysteinerich domain and probably has no equivalent in the insulin receptor. The third, located at the C terminus of the ␣ subunit, is the most conserved and provides the majority of the free energy of the interaction. Further structural studies will be necessary to determine how these elements interact in the interaction of the receptor with IGF-I.