PEDS Advance Access originally published online on June 13, 2006
Protein Engineering Design and Selection 2006 19(8):377-384; doi:10.1093/protein/gzl022
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Model for the complex between the insulin-like growth factor I and its receptor: towards designing antagonists for the IGF-1 receptor
CSIRO Molecular & Health Technologies 343 Royal Parade, Parkville, Victoria 3052, Australia
1To whom correspondence should be addressed. E-mail: Vidana.Epa{at}csiro.au
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResults and discussionReferences |
|---|
The type-1 insulin-like growth factor receptor (IGF-1R) is the cognate tyrosine kinase receptor for the insulin-like growth factor IGF-I and is expressed widely in many foetal and postnatal tissue cells. IGF-1R is overexpressed in a number of human tumour types and is a valid target for anti-cancer therapeutic efforts. Designing antagonists for IGF-1R would be greatly facilitated by the availability of structural information on the complex between IGF-I and IGF-1R. In the present work we model the three-dimensional structure of the complex between IGF-I and the first three domains of IGF-1R using a macromolecular docking method guided by selected experimental data. Interface metrics indicative of the binding affinity and reliability of the model are computed and compared with other biomolecular complexes. This model is consistent with experimental chimerical and mutagenesis data, provides a structural basis for understanding the primary interaction of IGF-I with its receptor and facilitates design of antagonist ligands.
Keywords: antagonist design/complementarity/IGF-1R/protein docking
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionReferences |
|---|
The insulin-like growth factors IGF-I and IGF-II are essential for normal foetal and postnatal growth and development. Cell proliferation as well as protection of cells from apoptosis are cellular responses following IGF-I binding and activating its cognate receptor, the type-1 insulin-like growth factor receptor (IGF-1R). IGF-1R is a member of the tyrosine kinase receptor superfamily, which also includes the insulin receptor (IR) and the epidermal growth factor receptor. [Adams et al. (2000)
and De Meyts and Whittaker (2002) have reviewed IGF-1R and IR recently.]
Both IGF-1R and its close homolog IR are disulfide-linked dimers, denoted (
ß)2, with each monomer composed of an extracellular chain and a ß chain which has domains both extra and intracellular. The extracellular portion of each ß monomer consists of two large domains L1 and L2 flanking a cysteine-rich domain CR, followed by three fibronectin type III domains, with a so-called insert domain inside the second fibronectin domain. The cysteines linking the two receptor monomers are located in the first fibronectin domain and the insert domain. The intracellular portion of each monomer consists of the transmembrane and juxtamembrane domains, followed by the tyrosine kinase domain and a cytoplasmic tail. While the atomic structure of the tyrosine kinase domain of IGF-1R (as well as IR) has been solved by X-ray crystallography (Favelyukis et al., 2001; Pautsch et al., 2001), an extracellular domain X-ray crystal structure is available only for the first three domains (L1-CR-L2) of IGF-1R (Garrett et al., 1998). Luo et al. (1999) have proposed a model for the quaternary structure of the insulin-IR complex based on cryo-electron microscopic studies. The L1 and L2 structural domains are right-handed ß-helices and the CR domain that connects these consists of eight disulfide-linked modules. The central space enclosed by these three domains is of size large enough to accommodate IGF-I, and many of the IGF-I-specific binding determinants can be mapped on to the L1 and CR molecular surface surrounding this space.
The whole IGF-1 Receptor, i.e. the (ß)2 dimer, binds IGF-I with high affinity, as does the whole IR to insulin. Schäffer (1994) proposed a cross-linking mechanistic model to explain the experimentally observed curvilinear Scatchard plots and the negative cooperativity. This was further developed by De Meyts (1994). In these mechanistic hypotheses, the two binding sites on one ligand molecule (site 1 and site 2) bind to two distinct sites on the two receptor monomers ß.
Recently there has been renewed interest in IGF-1R as a target for anti-cancer therapeutic efforts. LeRoith and Helman (2004) point out that epidemiological and experimental studies provide strong evidence for a critical link between IGF signalling and human cancer. Butler et al. (1998) showed that in cell culture systems inhibition of IGF-1R expression or activation was successful in inhibiting cancer cell growth. Most recently, Mitsiades et al. (2004) and Garcia-Echeverria et al. (2004) showed that small molecule inhibitors of the tyrosine kinase domain of IGF-1R have significant in vivo anti-tumour activity and validate IGF-1R as a therapeutic target for a broad spectrum of cancers.
Inhibition and deactivation of the IGF-1R can be done targeting not only the intracellular domains but also the extracellular domains of IGF-1R. This may be attempted with different approaches such as small molecule inhibitors that block access to the ligand-binding site, ligand blocking antibodies and bivalent ligand with site 2 inactivated (De Meyts and Whittaker, 2002). The last approach is based on the Schäffer (1994) and De Meyts (1994) mechanism of two-site binding of IGF-I to the IGF-1R dimer. All these approaches would be greatly facilitated by having a structural model of IGF-I in complex with IGF-1R. The objective of the present study was to use macromolecular docking to construct a structural model of the ligand IGF-I in complex with the first three domains of IGF-1R. Such a model would provide a structural basis for understanding the interaction of IGF-I with IGF-1R, guide mutagenesis experiments and help design antagonists for IGF-1R. It would also serve as an assessment of the macromolecular docking method used, especially after the X-ray crystallographic solution of the structure.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionReferences |
|---|
Preparation of structures
The receptor was prepared from the crystal structure of the first three domains (residues 1–459) of the human IGF-1R structure solved by Garrett et al. (1998) [PDB ID: 1IGR
[PDB]
]. Several side-chain atom coordinates are missing in this structure. These were modelled in with InsightII (Accelrys, San Diego) and the energy of modelled side-chains minimized with Discover (Accelrys, San Diego). For the docking studies only the first two domains (residue # 1–300) were used. The ligand was prepared from the NMR structures of human IGF-I solved by Cooke et al. (1991) (minimum average structure, PDB ID: 2GF1
[PDB]
and the ensemble of 10 NMR structures, PDB ID: 3GF1). The program NMRCLUST (Kelley et al., 1996) was used to group the ensemble of structures into three clusters, and a representative structure was selected from each cluster for the docking (labelled 3gf1a, 3gf1b and 3gf1c). In addition, the structure in 2GF1 was also used in the docking studies.
Docking
The docking studies were performed with the program suite 3D-Dock (including FTDock and RPScore) from the Biomolecular Modelling Laboratory of Cancer Research, UK (http://www.bmm.icnet.uk/docking). The macromolecular docking of IGF-I (2gf1, 3gf1a, 3gf1b and 3gf1c) on to IGF-1R was done with FTDock, v. 2.0 (Gabb et al., 1997). FTDock is a grid-based method performing global scans of translational and rotational space and scores and ranks the docked solutions by means of a surface complementarity score (SCscore). The default values of 12° for the search angle step, 1.3 Å for the surface thickness and –15.0 for the internal deterrent value were used. The global grid of 1823 corresponded to a grid unit (cell) size of 
0.7 Å, and there were a total of 9240 global rotations. This resulted in a total of 10 000 docked solutions.
These docked solutions were then rescored and ranked using RPScore. This program (Moont et al., 1999) scores the docked solutions using an empirical pair potential matrix. The potentials give the likelihood of a particular trans-interface pair of residues. The matrix i90_p05_d4.5_2dp.matrix was used in this work.
Next these solutions were subjected to three sequential filtering steps based on experimental constraints deduced from mutagenesis and binding data. First, the constraint that the residues Phe 90 and Asp 8 of IGF-1R be in the near vicinity (i.e. no further than 4.5 Å) of the ligand IGF-I was applied, i.e. only docking solutions that satisfy this criterion were accepted. Second, we required that Tyr 24 of IGF-I be in the near vicinity (no further than 4.5 Å) of IGF-1R. Third, the filtered docking solutions were then subjected to the constraint that Arg 36 or Arg 37 of the ligand IGF-I be within 8.0 Å of Glu 242 of IGF-1R.
Refinement and analysis
The above steps resulted in two docking solutions from the ligand structure 2gf1, two solutions from 3gf1a, six solutions from 3gf1b and four solutions from 3gf1c. Each of these solutions was then visually examined. This resulted in discarding the docking solutions from 2gf1, 3gf1a, 3gf1b and one solution from 3gf1c, leaving three solutions from 3gf1c as acceptable. Out of these three the solution with the highest RPScore was chosen as the model of the IGF-I/IGF-1R complex. After adding hydrogen atoms this model was energy minimized with Discover, v. 2.98 (CVFF force field, distance-dependent dielectric, 1000 steps of steepest descents and 1000 steps of conjugate gradients) keeping all heavy atoms of the receptor fixed and all heavy atoms of the ligand tethered with a force constant of 100.0 kcal/mol/Å2. (Application of strong restraints minimizes artefacts from the molecular mechanics force field used being introduced to the docked solutions.) The shape complementarity (Lawrence and Colman, 1993) and electrostatic complementarity (McCoy et al., 1997) metrics were computed for this model complex. Finally, by superimposing the C atoms of receptor residues 1–300 of the above docked model with the corresponding atoms of the crystal structure 1IGR, the model for IGF-I bound to the first three domains of IGF-1R (Figure 1) was generated.
|
|
Results and discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionReferences |
|---|
Construction of the model
In the present work we model the structure of the complex between IGF-I and the first three domains of IGF-1R using the macromolecular rigid body docking algorithm suite 3D-Dock. In the Critical Assessment of Prediction of Interactions (CAPRI) protein docking experiments concluded in 2003–2004, 3D-Dock was one of the better overall performing methodologies (Méndez et al., 2003, 2005). The values of the FTDock parameters used in this study are reasonable and have been seen to produce reasonably accurate results in the past (Smith and Sternberg, 2003). While even smaller values of grid separation, angle step and surface thickness are possible, it does not make sense to attempt such high-resolution docking for two reasons: first, we expect some degree of conformational rearrangement at the interface upon ligand binding (see for example, Méndez et al., 2005); using a finite non-zero surface thickness implicitly allows for some flexibility. Second, we are using structures from an NMR ensemble solution of IGF-I for the docking; again, this allows for some degree of flexibility during the ligand–receptor binding.
We believe that current state-of-the-art macromolecular docking algorithms find it difficult in many cases to accurately predict the structure of the complex starting from the apo structures of the components without using any biochemical experimental data to guide the docking. This appears to be one of the conclusions of the CAPRI experiments (Méndez et al., 2003, 2005). Hence, we used the following experimental information to guide the docking and construct the model in this study:
There is no experimental (site-directed mutants or chimera) evidence that the second large domain, L2 (residues # 301–459), of IGF-1R contributes to binding of IGF-I in any significant way. Hence, we used only the first large domain, L1 (i.e. residues #1–150), and the cysteine-rich domain, CR (i.e. residues #151–300), in our docking. This also facilitates the computational task.
Although residues #692–693, 697–698 and 701 of the C-terminal region of the -chain (belonging to the insert domain) of the holoreceptor do contribute significantly to the ligand binding (Mynarcik et al., 1997), we neglected this region due to the lack of any structural information. This qualification limits our model. [Most of the insert domain is predicted to be intrinsically disordered using the DisProt web server (Peng et al., 2005), a method found to be the most reliable in predicting disorder in proteins at the recently concluded CASP6 experiment (Jin and Dunbrack, 2005)]. Likewise, we did not consider the binding of the second receptor monomer (linked by disulfide bonds to the first receptor monomer) to the IGF-I ligand.
We filtered the 10000 docked solutions produced by FTDock (and rescored by RPScore) in three sequential steps so that they satisfy constraints imposed by mutagenesis data.
First, alanine scanning experiments on the receptor by Whittaker et al. (2001) show that mutations at Phe 90 and Asp 8 of the IGF-1R L1 domain cause the largest reduction in ligand binding of any L1 domain residue mutation. (These two residues are located in a large hydrophobic patch on the second ß-sheet of L1). Therefore we required that Phe 90 and Asp 8 of IGF-1R be at a near-binding distance (i.e. 
4.5 Å) from the ligand IGF-I.
Second, Bayne et al. (1990) showed that Tyr 24
Leu mutation in the ligand IGF-I causes an 18-fold decrease in binding to the receptor, IGF-1R. According to Cascieri et al. (1988) this mutation causes a 32-fold reduced affinity to the human placental IGF-1R. Tyr 24 is part of a large hydrophobic patch on IGF-I that also includes Phe 23, Phe 25 and Tyr 31. Therefore we required that Tyr 24 of IGF-I be no more than 4.5 Å distant from IGF-1R.
Zhang et al. (1994) showed that Ala substitution of Arg 36 and Arg 37 of the C-domain in IGF-I led to a 15-fold loss of binding affinity to human IGF-1R. They also showed that the region with residues # 217–284 of the CR domain of IGF-1R is responsible for recognizing Arg 36 and Arg 37. The electrostatic potential properties of IGF-I and IGF-1R (Figure 2a and b) show that there is significant electrostatic complementarity between the C-domain of IGF-I and the CR domain of IGF-1R. This complementarity may have an orientational effect in the binding event of the ligand to the receptor. The alanine scanning experiments of Whittaker et al. (2001) show that the Glu 242Ala mutation in IGF-1R CR domain causes a 4-fold decrease in ligand binding. Therefore, keeping in mind the long-range nature of the Coulombic forces, we applied the filter constraint that either Arg 36 or Arg 37 of IGF-I be no more than 8.0 Å distant from Glu 242 of IGF-1R. (Note that all these residues that are used in the filtering process are conserved in IGF-I and IGF-1R.)
|
As a result of these filtering procedures, the 10 000 initial docking solutions were reduced to two, two, six and four solutions, respectively, for the ligand structures 2gf1, 3gf1a, 3gf1b and 3gf1c. These solutions, along with their respective RPscore and SCscore values, are listed in Table I. Each of these 14 solutions was then visually examined. This showed that apart from solutions #1, 2 and 3 from 3gf1c all the other solutions had the large hydrophobic patch on the ligand IGF-I surface consisting of Phe 23, Tyr 24, Phe 25 and Tyr 31 oriented towards the external solvent rather than the large hydrophobic patch on the underside (i.e. second ß-sheet) of the ligand-binding domain L1 of the receptor, and thus were discarded. Of the three remaining solutions, the first one, with the highest value of RPscore (i.e. the highest likelihood of occurrence of trans-interface residue pairs) of 5.23, was selected as the favoured docked solution. Our selection scheme of discarding solutions which did not have the large hydrophobic patches on IGF-I and on the underside of L1 domain of IGF-1R facing each other is supported by the work of Lijnzaad and Argos (1997)
. In a survey of multimeric protein complexes they found that in 90% of the cases the largest or the second largest hydrophobic patch was involved in the multimeric interface. More recently, Berchanski et al. (2004) found that using interface hydrophobic ‘complementarity’ as a criterion significantly improved their protein–protein docking results.
|
Next, this structure was energy minimized as described in the Materials and methods section. (The energy minimization increased the RPscore value for the complex model from 5.23 to 5.75.) This, after superimposing the C
atoms of the receptor in the docked model with the corresponding atoms of the crystal structure 1IGR, resulted in the model of the complex between IGF-I and IGF-1R shown in Figure 1. Note that IGF-I does not make any contact with any atom of the L2 domain (residues #301–459) of the receptor in our model.
The interface metrics of shape complementarity Sc (Lawrence and Colman, 1993), interface area A (in Å2), electrostatic complementarity EC (McCoy et al., 1997) and RPscore (Moont et al., 1999) were calculated for this model and are shown in Table II. (Note that the metric Sc calculated here is a quantity different from SCscore in FTDock.) For purposes of comparison, these metrics were also computed for the X-ray crystallographic structures of the following complexes between some well-known cytokine/hormone/growth factors and their multi-domain receptors: interleukin-1ß and interleukin-1 receptor (PDB ID: 1ITB, resolution: 2.5 Å), erythropoietin and erythropoietin receptor (PDB ID: 1CN4, resolution: 2.8 Å), growth hormone and growth hormone receptor (PDB ID: 3HHR
[PDB]
, resolution: 2.8 Å), and the epidermal growth factor and epidermal growth factor receptor (PDB ID: 1IVO
[PDB]
, resolution: 3.3 Å). The degree of complementarity, both shape (i.e. geometry) and chemical (i.e. electrostatic potential), observed at the interface between the two components of a protein–protein complex is indicative of the binding affinity and in structure prediction is a good indicator of the reliability of the model. As can be seen from Table II, the values of these interface metrics for our docked model compare reasonably well with those computed for the crystal structures. (The very compute-intensive procedure of EC calculation makes it impracticable to use EC as a metric in filtering docked solutions or for a large number of protein–protein complexes.)
|
Schäffer (1994)
first proposed a mechanistic model where high-affinity binding of insulin is achieved by one insulin molecule cross-linking two IR monomers (ß) and involve two binding surfaces of the ligand insulin and two distinct sites (site 1 and site 2) on the two receptor monomers. This mechanism can reasonably be expected to hold for the high-affinity binding of IGF-I to the IGF-1 receptor. As De Meyts and Whittaker (2002) point out, antagonists can be designed for such a sequential two-site binding system by introducing a mutation in the ligand that decreases binding affinity of site 2, as was done in the case of the human growth hormone receptor (Fuh et al., 1992; de Vos et al., 1995). One could conceptualize the design of an IGF-1R antagonist by preserving the binding of the ligand IGF-I to site 1 on the receptor and ablating the binding of IGF-I to site 2 on the receptor (and thus inactivating the signal transduction—i.e. the tyrosine kinase phosphorylation and the subsequent cascade of intracellular events) by a suitable mutation of an IGF-I amino acid residue that binds to site 2. Now, which IGF-I residue should be mutated for this purpose? This would be a straightforward question to answer if a crystal structure of IGF-I bound to IGF-1R (1:2 stoichiometry, i.e. high affinity) was available. In the absence of such a crystal structure our structural model of IGF-I bound to site 1 of IGF-1R assists us in narrowing down the possible candidates.
LigPlot (Wallace et al., 1995) was used on the IGF-I–IGF-1R model to generate a list of the IGF-I residues that interact with the receptor. (Here, to compute the interactions, a maximum donor–acceptor distance of 4.5 Å and a maximum H-acceptor distance of 3.1 Å was used.) This gives the IGF-I residues that interact with the receptor as Gly 7, Ala 8, Leu 10, Val 11, Leu 14, Gln 15, Gly 19, Phe 23, Tyr 24, Phe 25, Lys 27, Thr 29, Gly 32, Ser 33, Ser 34, Arg 36, Arg 37 and Val 44. The interactions these residues make with site 1 of the receptor need to be preserved (or even enhanced) in the design of a mutant IGF-I antagonist. Furthermore, one would not attempt to mutate the residues that are in the interior of IGF-I, structurally sensitive glycines, prolines or cysteines, or the mobile first few residues at the N- and C-termini of IGF-I. Of the remaining residues, the sterically less accessible ones may be removed from consideration by neglecting the residues with fractional solvent accessible surface area (calculated with InsightII in the context of the receptor environment) less than 0.3. Taking all the above into consideration, we propose that mutagenesis of one or more of the finally remaining amino acid residues Glu 9, Asp 12, Phe 16, Asp 20, Arg 21, Tyr 31, Ser 35, Ala 38, Thr 41, Asp 45, Phe 49, Arg 50, Leu 54, Arg 55, Arg 56, Leu 64 and Lys 65 be done to design an antagonist of IGF-1R. (i.e. we predict that a subset of these amino acid residues would be among those in contact with the second receptor monomer or site 2.) These residues are coloured yellow in the molecular surface representation of the IGF-I-Receptor docked model shown in Figure 3. And finally, one should consider the physicochemical nature of each of the above amino acid residues and mutate it to one that is complementary in nature. For example, acidic residues can be mutated into basic ones and vice versa, and small polar residues into large hydrophobic residues and vice versa. As a caveat, we note that while sterically inaccessible and structurally sensitive residues have been removed from consideration in compiling this residue list, as in most mutagenesis experiments, there is potential for a particular mutation to cause tertiary structural changes that can affect the primary interactions with the receptor in an undesirable manner.
|
Comparison of the model with other experimental data
estawski et al. (2001) solved the crystal structure of IGF-I in complex with the N-terminal IGF-binding domain of the IGF-binding protein 5 (mini-IGFBP-5). Using the structure of the complex they identified the IGF-I residues that make contact with mini-IGFBP-5 as Glu 3, Thr 4, Leu 5, Glu 9, Phe 16, Cys 52, Asp 53, Leu 54, Leu 57 and Glu 58. None of these residues makes contact with IGF-1R in our model. Hence it is possible (as estawski et al. also point out on the basis of experimental data on IGF-I) for IGF-I complexed with mini-IGFBP-5 to still bind to IGF-1R, as has been experimentally observed (Kalus et al., 1998).
Siwanowicz et al. (2005) determined the crystal structure of the complex of the N-terminal domain of the IGF-binding protein 4 (NBP-4) in complex with IGF-I and showed that IGF-mediated stimulation of the IGF-1R autophosphorylation is inhibited by NBP-4. Their structure shows NBP-4 interacting with IGF-I residues Asp 12, Glu 13, Gln 15 and Phe 23. The latter two residues interact with the receptor in our model, and hence explain their observation on inhibition.
Carrick et al. (2005), in an NMR study of the interaction of IGF-I with the IGF-binding protein 2, found that the IGF-I residues for which chemical shifts occurred upon binding to IGFBP-2 included Val 11, Leu 14, Gln 15, Gly 19, Tyr 24, Phe 25 and Lys 27, which in our structural model of the ligand–receptor complex are in contact with IGF-1R. This explains IGFBP-2 blocking ligand binding to the IGF-1R.
Sakano et al. (1991) showed that truncation of the D-domain region (residues # 62–70) of IGF-I had no effect on receptor binding. This again is in concordance with our model where none of these residues is in contact with IGF-1R.
Cascieri et al. (1989) showed that replacement of the A-domain residues 42–56 of IGF-I with A1–A15 of insulin, along with a Thr 41Ile mutation, had no effect on IGF-1R binding. This is quite reasonable in the light of our model, where out of the residues 42–56 only Val 44 is in contact with the receptor, and the analogous insulin position A3 is also a valine.
NMR studies by Denley et al. (2005) of the IGF-I mutant Val44Met showed retention of near native structure. However, this naturally occurring mutation causes growth retardation in childhood. In our structural model, Val 44 interacts with the IGF-1 receptor, and mutation to the longer side-chain Methionine sterically hinders the binding.
Whittaker et al. (2001) performed alanine scanning mutagenesis on IGF-1R and found that alanine substitution of residues Asp 8, Tyr 28, His 30, Leu 33, Phe 58, Trp 79, Phe 90, Glu 242, Asn 11, Leu 56, Arg 59, Arg 240 and Phe 241 reduces binding affinity at least 2-fold. The first eight residues mentioned are in contact with IGF-I in our model.
It is known that IGF-II binds to IGF-1R with almost one-fifth affinity as IGF-I (Denley et al., 2004). This is not surprising as the majority of the IGF-I residues that we predict from our model to make contact with IGF-1R, namely Gly 7, Leu 10, Val 11, Leu 14, Gln 15, Gly 19, Phe 23, Tyr 24, Phe 25, Ser 33, Arg 37 and Val 44, are completely conserved in IGF-II. (Figure 4). Also, Headey et al. (2004) determined that the binding site for the C-terminal domain of the IGF-binding protein 6 (CBP-6) on IGF-II includes Phe 28. This again is consistent with our structural model where the analogous residue on IGF-I, Phe 25, binds to IGF-1R, thus explaining inhibition of receptor binding by CBP-6.
|
Ottensmeyer et al. (2000)
proposed a model for insulin in complex with an IR dimer by fitting crystal structures of insulin and IGF-1R (first three domains) to cryo EM data. It may be interesting to see how our docked model of IGF-I–IGF-1R complex compares with their insulin-IR model, in terms of predicted ligand and receptor binding amino acid residues. (Note that the coordinates of the Ottensmeyer et al. model are not available.) However, in any comparison with experimental data on insulin binding to IR, an important caveat must be kept in mind: direct one-to-one correspondence between the two models (insulin/IR and IGF-I/IGF-1R) is contingent on binding modes of insulin (to IR) and IGF-I (to IGF-1R) being the same and there being no significant conformational changes in insulin or IGF-I accompanying their binding. However, Hua et al. (1991) solved the structure of an active insulin mutant with the B-chain partially unfolded and suggested that detachment of the C-terminal region of the B-chain occurs when native insulin binds to IR. This has been further confirmed by photo-cross-linking studies by Xu et al. (2004) and Wan et al. (2005). Unlike IGF-I, insulin consists of two distinct chains linked by two disulfide bonds. Also, insulin lacks the C-domain and D-domain loops of IGF-I. Furthermore, alanine scanning mutagenesis experiments [for example that of Whittaker et al. (2001)] for IGF-1R have implicated the CR domain in IGF-I binding, unlike the case for insulin and IR. All these indicate that despite the obvious homology care must be taken in comparing IGF-I binding to IGF-1R with that of insulin binding to IR: while the binding of the two ligands to their respective receptors is broadly similar (for example, in each case, the ‘classical’ hydrophobic patch on the ligand binding to the hydrophobic patch on L1 domain of the receptor), there may be differences in the detailed specifics of the binding. This is known from the comparative studies of Whittaker's group [for example, Mynarcik et al. (1997)], who showed that the relative importance of equivalent residues to insulin and IGF-I binding was not the same. The Ottensmeyer et al. model has the major hydrophobic interactions between insulin and the IR monomer, which they designate as ‘monomer II’, due to the following binding residues of insulin (corresponding IGF-I residues in parentheses): Ser B9 (Ala 8), His B10 (Glu 9), Glu B13 (Asp 12), Val B12 (Val 11), Leu B17 (Phe 16), Tyr B16 (Gln 15), Glu B21 (Asp 20), Arg B22 (Arg 21), Phe B24 (Phe 23), Tyr B26 (Phe 25), Gln A5 (Glu 46), Gln A15 (Arg 56). Out of these residues, Ala 8, Val 11, Gln 15, Phe 23, Phe 25 are in contact with IGF-1R in our model.
Ottensmeyer et al. also list the residues Glu A4 (Asp 45), Thr A8 (Phe 49), Lys B29 (Pro 28), Glu A17 (Glu 58), Asn A21 (Ala 62) as the insulin residues that contact the other IR monomer (which they name ‘monomer I’). None of these residues is involved in binding IGF-1R in our docked model. (In fact, Glu 58 and Ala 62 are buried in the core of IGF-I).
Recently, Xu et al. (2004) using photoactivatable derivatives of insulin showed that while Phe B25 cross-links the C-terminal domain of the subunit of IR, Phe B24 cross-links the L1 domain. The corresponding IGF-I residues, Phe 23 and Tyr 24, are both in contact with the second ß-sheet of the L1 domain in our IGF-I–IGF-1R complex model. Fabry et al. (1992) found that insulin derivatized at Phe B1 cross-linked an IR fragment ranging from Gly 390 (in the L2 domain) to Arg 488 (in the first Fibronectin III domain). In our model, the N-terminus of IGF-I is oriented towards the L2 domain of IGF-1R.
Schäffer (1994) analysed anomalous binding properties of insulin analogues substituted at positions Leu A13 and Leu B17 and concluded that these residues must form part of a second binding site for the IR. The analogous IGF-I residues are Leu 54 and Phe 16, two of the residues that we have nominated for possible mutagenesis. In addition to the above two insulin residues, De Meyts (2004) proposed that, on the basis of mutagenesis studies, His B10 (Glu 9), Glu B13 (Asp 12), Ser A12 (Asp 53) and Glu A17 (Glu 58) also form part of a second binding site for the IR. (Analogous IGF-I residues are indicated in parentheses.) Of these, Glu 9 and Asp 12 are residues that we have nominated for mutagenesis while Asp 53 and Glu 58 we removed from consideration due to low solvent accessibility.
In conclusion, in the present work we have constructed a three-dimensional model of the complex between IGF-I and the first three domains of the IGF-1 receptor using macromolecular docking methods guided by some known experimental data as constraints. This model explains and provides a structural basis for other experimental observations and suggests the IGF-I amino acid residues that may be mutated in order to design antagonists of the receptor. This structural model will also be of use in designing experiments to understand better the interaction of IGF-I with its receptor. Finally, when the X-ray crystal structure of the IGF-I–IGF-1R complex is solved, the details of the deviations of this model from the experimental structure should be valuable in improving the protein–protein docking algorithms used.
The atomic coordinates of the model can be obtained from the author on request.
|
Footnotes |
|---|
Edited by Bruce Tidor
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionReferences |
|---|
Adams T.E., Epa V.C., Garrett T.P.J., Ward C.W. (2000) Cell. Mol. Life Sci. 57:1050–1093.[CrossRef][Web of Science][Medline]
Bayne M.L., Applebaum J., Chicchi G.G., Miller R.E., Cascieri M.A. (1990) J. Biol. Chem. 265:15648–15652.
Berchanski A., Shapira B., Eisenstein M. (2004) Proteins: Struct. Funct. Bioinform. 56:130–142.
Butler A.M., Blakesley V.A., Tsokos M., Pouliki V., Wood T.L., LeRoith D. (1998) Cancer Res. 58:3021–3027.
Carrick F.E., Hinds M.G., McNeil K.A., Wallace J.C., Forbes B.E., Norton R.S. (2005) J. Mol. Endocrinol. 34:685–698.
Cascieri M.A., Chicchi G.G., Applebaum J., Hayes N.S., Green B.G., Bayne M.L. (1988) Biochemistry 27:3229–3233.[CrossRef][Medline]
Cascieri M.A., Chicchi G.G., Applebaum J., Green B.G., Hayes N.S., Bayne M.L. (1989) J. Biol. Chem. 264:2190–2202.
Cooke R.M., Harvey T.S., Campbell I.D. (1991) Biochemistry 30:5484–5491.[CrossRef][Medline]
De Meyts P. (1994) Diabetologia 37:Suppl. 2, S135–S148.[CrossRef][Web of Science][Medline]
De Meyts P. (2004) Bioessays 26:1351–1362.[CrossRef][Web of Science][Medline]
De Meyts P. and Whittaker J. (2002) Nature Rev. Drug Disc. 1:769–783.[CrossRef][Web of Science][Medline]
De Vos A.M., Ultsch M.H., Kossiakoff A.A., Fuh G., Cunningham B.C., Wells J.A. (1995) In Shiverick K.T. and Rosenbloom A. (Eds.). Human Growth Hormone Pharmacology. (CRC press, Boca Raton, FL) pp. 29–41.
Denley A., Bonython E.R., Booker G.W., Cosgrove L.J., Forbes B.E., Ward C.W., Wallace J.C. (2004) Mol. Endocrinol. 18:2502–2512.
Denley A., Wang C.C., McNeil K.A., Walenkamp M.J.E., van Duyvenvoorde H., Wit J.M., Wallace J.C., Norton R.S., Karperien M., Forbes B.E. (2005) Mol. Endocrinol. 19:711–721.
Fabry M., Schaefer E., Ellis L., Kojro E., Fahrenholz F. (1992) J. Biol. Chem. 267:8950–8956.
Favelyukis S., Till J.H., Hubbard S.R., Miller W.T. (2001) Nat. Struct. Biol. 8:1058–1063.[CrossRef][Web of Science][Medline]
Fuh G., Cunningham B.C., Fukunaga R., Nagata S., Goeddel D.V., Wells J.A. (1992) Science 256:1677–1680.
Gabb H.A., Jackson R.M., Sternberg M.J.E. (1997) J. Mol. Biol. 272:106–120.[CrossRef][Web of Science][Medline]
Garcia-Echeverria C., et al. (2004) Cancer Cell 5:231–239.[CrossRef][Web of Science][Medline]
Garrett T.P.J., McKern N.M., Lou M.Z., Frenkel M.J., Bentley J.D., Lovrecz G.O., Elleman T.C., Cosgrove L.J., Ward C.W. (1998) Nature 394:395–399.[CrossRef][Medline]
Headey S.J., Keizer D.W., Yao S., Wallace J.C., Bach L.A., Norton R.S. (2004) FEBS Lett. 568:19–22.[CrossRef][Web of Science][Medline]
Hua Q.X., Shoelson S.E., Kochoyan M., Weiss M.A. (1991) Nature 354:238–241.[CrossRef][Medline]
Jin Y. and Dunbrack R.L. (2005) Proteins: Struct. Funct. Bioinform. 61:Suppl. S7, 167–175.[CrossRef]
Kalus W., et al. (1998) EMBO J. 17:6558–6572.[CrossRef][Web of Science][Medline]
Kelley L.A., Gardner S.P., Sutcliffe M.J. (1996) Protein Eng. 9:1063–1065.
Kraulis P. (1991) J. Appl. Crystallog. 24:946–950.[CrossRef][Web of Science]
Lawrence M.C. and Colman P.M. (1993) J. Mol. Biol. 234:946–950.[CrossRef][Web of Science][Medline]
LeRoith D. and Helman L. (2004) Cancer Cell 5:201–202.[CrossRef][Web of Science][Medline]
Lijnzaad P. and Argos P. (1997) Proteins: Struct. Funct. Genet. 28:333–343.[CrossRef][Web of Science][Medline]
Luo R.Z.-T., Beniac D.R., Fernandes A., Yip C.C., Ottensmeyer F.P. (1999) Science 285:1077–1080.
Méndez R., Leplae R., Ge Maria L., Wodak S.J. (2003) Proteins: Struct. Funct. Genet. 52:51–67.[CrossRef][Web of Science][Medline]
Méndez R., Leplae R., Lensink M.F., Wodak S.J. (2005) Proteins: Struct. Funct. Bioinform. 60:150–169.
McCoy A.J., Epa V.C., Colman P.M. (1997) J. Mol. Biol. 268:570–584.[CrossRef][Web of Science][Medline]
Merritt E.A. and Bacon D.J. (1997) Methods Enzymol. 277:505–524.[Web of Science][Medline]
Mitsiades C.S., et al. (2004) Cancer Cell 5:221–229.[CrossRef][Web of Science][Medline]
Moont G., Gabb H.A., Sternberg M.J.E. (1999) Proteins: Struct. Funct. Genet. 35:364–373.[CrossRef][Web of Science][Medline]
Mynarcik D.C., Williams P.F., Schäffer L., Yu G.Q., Whittaker J. (1997) J. Biol. Chem. 272:18650–18655.
Nicholls A., Sharp K.A., Honig B. (1991) Proteins: Struct. Funct. Genet. 11:281–296.[CrossRef][Web of Science][Medline]
Ottensmeyer F.P., Beniac D.R., Luo R.Z., Yip C.C. (2000) Biochemistry 39:12103–12112.[CrossRef][Medline]
Pautsch A., Zoephel A., Ahorn H., Spevak W., Hauptmann R., Nar H. (2001) Structure 9:955–965.[Medline]
Peng K., Vucetic S., Radivojac P., Brown C.J., Dunker A.K., Obradovic Z. (2005) J. Bioinform. Comput. Biol. 3:35–60.[CrossRef][Medline]
Sakano K., Enjoh T., Numata F., Fujiwara H., Marumoto Y., Higashihashi N., Sato Y., Perdue J.F., Fujita-Yamaguchi Y. (1991) J. Biol. Chem. 266:20626–20635.
Schäffer L. (1994) Eur. J. Biochem. 221:1127–1132.[Web of Science][Medline]
Siwanowicz I., Popowicz G.M., Wisniewska M., Huber R., Kuenkele K.-P., Lang K., Engh R.A., Holak T.A. (2005) Structure 13:155–167.[Medline]
Smith G.R. and Sternberg M.J.E. (2003) Proteins:Struct. Funct. Genet. 52:74–79.
Wallace A.C., Laskowski R.A., Thornton J.M. (1995) Protein Eng. 8:127–134.
Wan Z.-L., Huang K., Xu B., Hu S.-Q., Wang S., Chu Y.-C., Katsoyannis P.G., Weiss M.A. (2005) Biochemistry 44:5000–5016.[CrossRef][Medline]
Whittaker J., Groth A.V., Mynarcik D.C., Pluzek L., Gadsbøll V.L., Whittaker L.J. (2001) J. Biol. Chem. 276:43980–43986.
Xu B., Hu S.-Q., Chu Y.-C., Huang K., Nakagawa S.H., Whittaker J., Katsoyannis P.G., Weiss M.A. (2004) Biochemistry 43:8356–8372.[CrossRef][Medline]
estawski W., Beisel H.-G., Kamionka M., Kalus W., Engh R.A., Huber R., Lang K., Holak T.A. (2001) EMBO J. 14:3638–3644.[CrossRef]
Zhang W.G., Gustafson T.A., Rutter W.J., Johnson J.D. (1994) J. Biol. Chem. 269:10609–10613.
Received December 16, 2005; revised April 5, 2006; accepted May 14, 2006.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. L. Alvino, K. A. McNeil, S. C. Ong, C. Delaine, G. W. Booker, J. C. Wallace, J. Whittaker, and B. E. Forbes A Novel Approach to Identify Two Distinct Receptor Binding Surfaces of Insulin-like Growth Factor II J. Biol. Chem., March 20, 2009; 284(12): 7656 - 7664. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Gauguin, C. Delaine, C. L. Alvino, K. A. McNeil, J. C. Wallace, B. E. Forbes, and P. De Meyts Alanine Scanning of a Putative Receptor Binding Surface of Insulin-like Growth Factor-I J. Biol. Chem., July 25, 2008; 283(30): 20821 - 20829. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Inagaki, A. Tiulpakov, P. Rubtsov, P. Sverdlova, V. Peterkova, S. Yakar, S. Terekhov, and D. LeRoith A Familial Insulin-Like Growth Factor-I Receptor Mutant Leads to Short Stature: Clinical and Biochemical Characterization J. Clin. Endocrinol. Metab., April 1, 2007; 92(4): 1542 - 1548. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||