Role of the Activation Loop Tyrosines in Regulation of the Insulin-like Growth Factor I Receptor-tyrosine Kinase*

The tyrosine kinase activity of insulin-like growth factor I receptor (IGF1R) is under tight control. Ligand binding to the extracellular portion of IGF1R stimulates autophosphorylation at three sites (Tyr1131, Tyr1135, and Tyr1136) in the activation loop within the tyrosine kinase catalytic domain. Autophosphorylation at all three sites is required for maximum enzyme activity, and for IGF1-stimulated cellular activity of the receptor. Previous studies have not clarified the contributions of the individual tyrosines to enzymatic activation. Here, we produced single Tyr-to-Phe mutations at these positions, and compared activities of the purified kinase domains (unphosphorylated and phosphorylated) with wild-type IGF1R. Rates of autophosphorylation of the three mutants were more rapid than for wild-type IGF1R; this was most apparent for the Y1135F mutant. Substrate phosphorylation studies on the unphosphorylated forms of IGF1R confirmed that the value of Vmax for Y1135F was elevated relative to wild-type IGF1R, consistent with a disruption of an autoinhibitory interaction. In contrast, activity measurements on the fully phosphorylated enzymes indicated that kcat/Km values were lowered relative to wild-type IGF1R; this effect was most dramatic for Y1136F. We confirmed these findings using limited proteolysis and tryptophan fluorescence experiments. The results demonstrate that Tyr1135 plays a particularly important role in stabilizing the autoinhibited conformation of the activation loop, while Tyr1136 plays the key role in stabilizing the open, activated conformation of IGF1R.

The tyrosine kinase activity of insulin-like growth factor I receptor (IGF1R) is under tight control. Ligand binding to the extracellular portion of IGF1R stimulates autophosphorylation at three sites (Tyr 1131 , Tyr 1135 , and Tyr 1136 ) in the activation loop within the tyrosine kinase catalytic domain. Autophosphorylation at all three sites is required for maximum enzyme activity, and for IGF1-stimulated cellular activity of the receptor. Previous studies have not clarified the contributions of the individual tyrosines to enzymatic activation. Here, we produced single Tyr-to-Phe mutations at these positions, and compared activities of the purified kinase domains (unphosphorylated and phosphorylated) with wild-type IGF1R. Rates of autophosphorylation of the three mutants were more rapid than for wild-type IGF1R; this was most apparent for the Y1135F mutant. Substrate phosphorylation studies on the unphosphorylated forms of IGF1R confirmed that the value of V max for Y1135F was elevated relative to wild-type IGF1R, consistent with a disruption of an autoinhibitory interaction. In contrast, activity measurements on the fully phosphorylated enzymes indicated that k cat /K m values were lowered relative to wild-type IGF1R; this effect was most dramatic for Y1136F. We confirmed these findings using limited proteolysis and tryptophan fluorescence experiments. The results demonstrate that Tyr 1135 plays a particularly important role in stabilizing the autoinhibited conformation of the activation loop, while Tyr 1136 plays the key role in stabilizing the open, activated conformation of IGF1R.
Insulin-like growth factor 1 receptor (IGF1R) 2 is a transmembrane tyrosine kinase that is essential for fetal and postnatal growth and development (1)(2)(3)(4)(5). It is expressed in most human tissues and cell types (the only known exceptions are hepatocytes and mature B cells). The receptor is activated by binding of the secreted growth factor ligand IGF1 (or the lower affinity ligand IGF2) to the extracellular domain of IGF1R.
Ligand binding stimulates the intrinsic tyrosine kinase activity of the cytoplasmic domain. This in turn triggers a number of cellular signaling pathways, including the phosphatidylinositol 3-kinase (PI-3K)/Akt pathway (the main mechanism by which IGF1R protects cells from apoptosis), and the Erk/MAPK pathway leading to mitogenesis. Aberrant IGF1R signaling has been implicated in malignant transformation of cells, and the enzyme has emerged as a target for anticancer drug design (6 -10).
IGF1R, insulin receptor (IR) and insulin-related receptor (IRR) belong to the same receptor-tyrosine kinase (RTK) family (11,12). These receptors are composed of two extracellular ␣ subunits containing the ligand binding domain and two transmembrane ␤ subunits possessing tyrosine kinase activity. IGF1R and IR share 70% sequence identity overall, and 84% identity within the tyrosine kinase catalytic domains. The mechanism for receptor activation is thought to be similar for IR and IGF1R (2)(3)(4)13).
The structures of the IGF1R and IR catalytic domains display the typical two-lobed protein kinase fold (14 -18). ATP binding and catalysis takes place in a deep cleft between the two lobes. These RTKs contain a segment known as the activation loop (A-loop), which contains the three principal autophosphorylation sites (Tyr 1131 , Tyr 1135 , and Tyr 1136 in IGF1R) (Fig. 1). Ligand binding to the extracellular ␣ subunit promotes a conformational change that leads to autophosphorylation of these sites. The first site of autophosphorylation is predominantly Tyr 1135 , followed by Tyr 1131 and then by Tyr 1136 (16). The unphosphorylated, monophosphorylated, bisphosphorylated, and trisphosphorylated forms of IGF1R (IGF1R-0P, -1P, -2P, and -3P, respectively) have increasing levels of enzyme activity (16).
The three-dimensional structures of IGF1R-0P, -2P, and -3P have been solved (16 -18). In IGF1R-0P, the A-loop inhibits access to substrates (17). Phosphorylation of Tyr 1131 , Tyr 1135 , and Tyr 1136 triggers a large change in the A-loop conformation, which allows access of ATP and peptide substrate to the catalytic site. Thus, autophosphorylation of IGF1R is the crucial regulator of the catalytic activity. The details of how autophosphorylation controls enzymatic activity, however, are not completely understood. Steady state kinetic studies show that a significant decrease in the substrate K m occurs in the transition from the 1P state (mainly Tyr 1135 ) to the 2P state (Tyr 1135 and Tyr 1131 ) (16). The conformation of the trisphosphorylated activation loop in IGF1R-3P is stabilized by numerous interactions between the activation loop and other segments of the kinase. The purpose of the present study was to define the roles of Tyr 1131 , Tyr 1135 , and Tyr 1136 in regulation of autophosphorylation.
When expressed in fibroblasts, a mutant lacking all three A-loop tyrosines showed dramatically reduced signaling in response to IGF-1 (19,20). A Y1135F/Y1131F double mutant was incapable of further autophosphorylation in NIH3T3 cells, and did not promote downstream Shc and IRS-1 signaling (21). Mutation of Tyr 1131 alone significantly reduced the level of IGF1-induced receptor autophosphorylation, whereas it did not have an impact on cell proliferation or phosphorylation of the substrate IRS-1 (22). In a separate study, a Y1136F mutant showed a marked decrease in IGF1-stimulated receptor phosphorylation, DNA synthesis, and mitogenesis, whereas Y1135F and Y1131F mutations had smaller effects (20). The effects of these mutations on purified preparations of the IGF1R kinase domain have not previously been published.
We previously reported a method to separately purify each of the phosphorylated forms of IGF1R (IGF1R-0P, -1P, -2P, and -3P) (16). Steady-state kinetic analyses of the isolated phosphorylated forms of the IGF1 receptor showed that each autophosphorylation increases enzyme turnover number and decreases K m for ATP and peptide. In this study, we used similar methods to purify unphosphorylated and phospho-rylated forms of IGF1R mutants containing single mutations at Tyr 1131 , Tyr 1135 , and Tyr 1136 . We analyzed the conformation of the A-loop and the level of enzymatic activity in these forms of IGF1R. Our results indicate that each tyrosine is important for full activation of the IGF1R kinase domain. Furthermore, our results shed light on the individual roles of the three tyrosines. Autophosphorylation of Tyr 1135 and Tyr 1131 appears to destabilize the autoinhibitory conformation of the activation loop, whereas Tyr 1136 phosphorylation plays the key role in structural stabilization of the A-loop.

EXPERIMENTAL PROCEDURES
Materials-10% Tris-HCl gels were from Bio-Rad. Mass spectrometry grade trypsin was obtained from Promega. ADP and potassium iodide were obtained from Sigma.
Mutagenesis and Purification of IGF1R Kinase Domain-IGF1R mutants were produced with the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's protocols. The mutants were made in a pFastBac-1 baculovirus transfer vector (Invitrogen) encoding the kinase catalytic domain of IGF1R (IGF1RK). The mutations were verified by DNA sequencing on an ABI 373 automated DNA sequencer. Mutant forms of IGF1RK were expressed in Spodoptera frugiperda (Sf9) cells using the Bac-to-Bac Baculovirus Expression System (Invitrogen) as described previously (16). Sf9 cells were cultured in Excel 410 (JRH Biosciences) with 5% heat-inactivated serum (Sigma) and 1% pencillin/streptomycin/ amphotericin at 27°C. The cells were harvested after 72 h of infection.
The purification of the mutants was accomplished using similar procedures as for wild-type IGF1RK (16). Unphosphorylated forms of the mutants were purified in three chromatographic steps on an FPLC system (GE Healthcare): 1) a Source Q-15 column, 2) a Superdex-75 gel filtration column, and 3) a Mono-Q HR10/10 column. To produce the completely phosphorylated forms of IGF1RK (i.e. IGF1RK-3P for wild-type enzyme, or IGF1RK-2P for the single site mutants), enzymes were incubated with 10 mM ATP and 30 mM MgCl 2 for 5 min at room temperature. The autophosphorylation reactions were terminated by addition of 100 mM EDTA. Native gel analysis confirmed that these reactions contained predominantly the completely phosphorylated forms. Sodium orthovanadate (200 M) was added to prevent dephosphorylation. The reactions were then passed over Superdex 75 to remove nucleotides, and the phosphorylated forms were separated by Mono-Q chromatography (16). The pooled fractions were concentrated and stored at Ϫ80°C in 20 mM Tris-HCl, pH 7.5. All final protein concentrations were determined by the Bradford method (Bio-Rad).
Kinase Assays-The kinetics of IGF1R autophosphorylation and substrate peptide phosphorylation were measured by a continuous spectrophotometric assay (16,23). In this assay, the consumption of ATP is coupled to the oxidation of NADH, which is monitored as a reduction in NADH absorption at 340 nm. Reactions (50 l) contained 100 mM Tris (pH 7.5), 10 mM MgCl 2 , 1 mM phosphoenolpyruvate, 1. Native-PAGE autophosphorylation assays were carried out using 6 M enzyme, 10 mM ATP, and 30 mM MgCl 2 at room temperature (24). The reactions were stopped by addition of 100 mM EDTA at various time points. Autophosphorylation reactions were analyzed by 10% Tris-HCl native PAGE and visualized by Coomassie Blue staining.
Limited Proteolysis of IGF1RK-Limited proteolysis of wildtype and mutant forms of IGF1RK was carried out according to previous methods established for IRK (25). Proteins were diluted to 9 M in 50 mM Tris-HCl, pH 7.0, 30 mM MgCl 2 , and 2 mM dithiothreitol, with or without 10 mM ADP. Trypsin was added at a ratio of trypsin to IGF1RK of 1:25-30 by mass. The reactions proceeded at room temperature for 15 min and were stopped by addition of 5ϫ SDS-PAGE buffer. The digestion products were resolved by SDS-PAGE, and the fragments were visualized by Coomassie Blue staining.
Fluorescence Spectra and Iodide Quenching-Tryptophan fluorescence experiments were carried out in 100 mM Tris-HCl and 1 mM dithiothreitol using methods described for IRK (26). Experiments were carried out in volumes of 500 l using 0.5 M IGF1RK (wild-type or mutants). Potassium iodide was included at varying concentrations to quench emission of tryptophan 1148 of IGF1R. The ionic strength in the reactions was kept at 0.5 M by addition of KCl, and 0.1 mM sodium thiosulfite was included to inhibit the formation of I 3 Ϫ . The fluorescence emission spectra were obtained with a Cary Eclipse Fluorescence spectrophotometer. The excitation wavelength was 295 nm, and emission spectra were collected from 310 -420 nm. The solvent accessibility of fluorescent tryptophan residues was characterized by the Stern-Volmer constant, K SV , according to the relationship between the peak fluorescence intensity and the iodide quencher concentration [Q] in the solution, as defined by Equation 1, where F 0 is the fluorescence intensity in the absence of quencher.

RESULTS
Autophosphorylation of Single Site Mutants-To analyze the importance of the individual A-loop tyrosines, we produced mutant forms of IGF1RK in which each tyrosine was changed to phenylalanine (Y1135F, Y1131F, and Y1136F). We expressed wild-type and mutant IGF1RKs using the Sf9/baculovirus system, and purified the unphosphorylated forms of the enzymes. In wild-type IGF1R, the first A-loop tyrosine to be phosphorylated is Tyr 1135 , followed by Tyr 1131 , and then Tyr 1136 (16). To determine whether elimination of single A-loop tyrosines blocked the autophosphorylation of the remaining two sites, we incubated the enzymes with Mg-ATP and followed the time course of autophosphorylation by native PAGE analysis ( Fig.  2A). The various phosphoforms of IGF1R are readily resolved by native PAGE ( Fig. 2A and Ref. 16). All three mutants were capable of autophosphorylation at the remaining two sites, indicating that no one site is obligatory for full phosphorylation ( Fig. 2A).
To compare the rates of autophosphorylation for wild-type and mutant forms of IGF1RK, we used a continuous spectrophotometric assay. We incubated the enzymes with ATP in the absence of any exogenous substrate. As we observed previously (16), the enzyme progress curves were biphasic, because autophosphorylation activates IGF1RK. Autophosphorylation of all three single-site mutants was more rapid than for wild-type IGF1R (Fig. 2B), suggesting that to a certain extent all three tyrosines participate in autoinhibitory interactions. Activation was most apparent in the case of the Y1135F and Y1131F mutants; the final autophosphorylation rates for these mutants were 63.1 and 31.8 mOD/min., respectively, compared with 6.2 mOD/min for wild-type IGF1RK. The Y1136F mutant had a rate that was closer to wild-type IGF1R (12.2 mOD/min).
Substrate Phosphorylation Assays-We carried out steadystate kinetic measurements of phosphorylation of a synthetic peptide substrate by wild-type and single site mutants of IGF1RK. We first tested the unphosphorylated forms of the proteins. To ensure that the enzymes remained in their unphosphorylated states during the initial rate measurements, we analyzed samples from the reactions by native PAGE; there was no detectable conversion to 1P, 2P, or 3P forms in the peptide phosphorylation reactions (data not shown). The kinetic experiments showed that all three mutants had higher activity than wild-type IGF1RK (Table 1). This activation was driven primarily by changes in V max . The values of V max for Y1135F, Y1131F, and Y1136F were ϳ6-fold, 3-fold, and 2-fold higher than wildtype, respectively. Changes in K m for ATP were modest; the largest change was a 2-fold reduction for the Y1135F mutant. Values of K m (peptide) were increased for all three mutants, with the largest change (ϳ3-fold) in the Y1136F mutant.
Next, we wished to test the effects of the mutations on the fully-phosphorylated, activated states of the proteins. To produce the fully phosphorylated forms of wild-type and mutant IGF1RK, we incubated the enzymes with MgATP. We monitored autophosphorylation by native PAGE. After removal of excess ATP by gel filtration, we purified the fully-phosphoryla-ted forms (i.e. IGF1RK-3P, Y1131F-2P, Y1135F-2P, and Y1136F-2P) by MonoQ-FPLC. We then used the spectrophotometric assay to measure V max and K m for both ATP and peptide substrate. Each mutant showed a lowered value of k cat /K m relative to wild-type IGF1RK ( Table 2); values for Y1135F, Y1131F, and Y1136F were decreased by 9.3-fold, 2.2-fold, and 18.6-fold, respectively. These differences were due for the most part to increases in peptide substrate K m , especially for the Y1136F mutant (Table 2). These results indicate that Y1136 plays a particularly important role in stabilizing a conformation of IGF1R that is catalytically competent for phosphorylation of exogenous substrates.
Conformational Changes in the A-loop-To confirm that the single site mutations destabilized the autoinhibited conformation of the A-loop, we carried out limited proteolysis experiments. This approach was developed by Kohanski and coworkers (25) to analyze the conformation of the IRK activation loop; the rates of limited trypsin proteolysis were used to estimate the differences in global conformation. While wild-type, unphosphorylated IRK was relatively stable to trypsin, addition of adenine nucleotide (to generate autophosphorylated IRK) gave rise to fragments of Ϸ25 kDa and Ϸ16 kDa. These fragments were generated by cleavage at an Arg-Lys sequence within the activation loop (25). The Arg-Lys sequence is conserved in IGF1R (Fig. 1). We applied the limited proteolysis strategy to purified preparations of unphosphorylated and phosphorylated wild-type and mutant IGF1RK. First, we carried out limited proteolysis experiments on purified IGF1RK-0P, Y1135F-0P, Y1131F-0P, and Y1136F-0P (Fig. 3A). After quenching the reactions with SDS-PAGE loading buffer, we analyzed the products on SDS-PAGE. As was the case for IRK, IGF1RK-0P was cleaved into fragments of Ϸ24 kDa and Ϸ14 kDa only in the presence of ADP. The Y1131F and Y1136F mutants behaved similarly to wild-type IGF1R. In contrast, the Y1135F mutant showed significant A-loop cleavage in the absence of ADP (Fig. 3A), suggesting that this mutation promotes the "open" conformation of the A-loop.
We carried out similar experiments on the fully phosphorylated forms of the proteins (IGF1RK-3P, Y1135F-2P, Y1131F-2P, and Y1136F-2P). Surprisingly, IGF1RK-3P was relatively resistant to trypsin cleavage in the A-loop, even in the presence of ADP (Fig. 3B). This result suggests that the triple phosphorylation of the A-loop stabilizes an "open" conformation that renders the Arg-Lys sequence inaccessible to trypsin. We were able to exploit this effect to compare the fully phosphorylated forms of the enzymes. In contrast to wild-type IGF1R, the phosphorylated mutants showed significant trypsin cleavage in the presence or absence of ADP (Fig. 3B). The highest extent of cleavage was seen for the Y1136F mutant, consistent with a role for Tyr 1136 in stabilizing the open conformation of the A-loop.   As a complementary approach to study conformational changes in the IGF1RK A-loop, we measured tryptophan fluorescence. This method has been previously applied to IRK; the solvent accessibility of Trp 1175 differs between IRK-0P and IRK-3P, as detected by solute quenching (26). The catalytic domains of both IR and IGF1R have five tryptophan residues. One is on the surface of the small lobe, and the other four are in the interior of the large lobe. Mutagenic analysis demonstrated that the fluorescence emission spectrum of IRK was dominated by Trp 1175 , a residue that is near the C terminus of the A-loop in the IRK active site (26). The solvent accessibility of IRK Trp 1175 increases significantly upon autophosphorylation. For IRK, unphosphorylated enzyme gave a Stern-Volmer constant of 0.3 M Ϫ1 , whereas phospho-IRK gave a value of 1.5 M Ϫ1 . IR and IGF1R are highly conserved in this region, and the residue corresponding to IRK Trp 1175 in IGF1R is Trp 1148 (Fig. 1). We employed a similar fluorescence quenching strategy to monitor solvent accessibility in the active site of purified IGF1RK-0P and IGF1RK-3P. We carried out fluorescence experiments using an excitation wavelength of 295 nm, because tyrosine absorbance is negligible at this wavelength. We measured emission spectra over 310 -420 nm, using increasing concentrations of potassium iodide as quencher. The fluorescence intensity over 310 -420 nm was integrated and analyzed on Stern-Volmer plots (Fig. 4). As reported previously for IRK, the solvent accessibility of IGF1RK increases upon autophospho-rylation: the Stern-Volmer constant for unphosphorylated IGF1RK was 0.29 M Ϫ1 , whereas the value for IGF1R-3P was 1.72 M Ϫ1 (Fig. 4 and Table 3). Next, we performed similar studies on the three A-loop mutants of IGF1R, in both their unphosphorylated and fully phosphorylated (2P states). All unphosphorylated mutants showed higher values of K SV than wild-type IGF1RK-0P, with the most significant increase for Y1135F (Table 3). These results are consistent with our other experiments, and suggest that Tyr 1135 plays a particularly important role in stabilizing the autoinhibited conformation. Tyr 1131 and (to a lesser extent) Tyr 1136 are also involved in autoinhibition. The Stern-Volmer constants for the doubly phosphorylated forms of the A-loop mutants were less than the value of K SV for fully phosphorylated wild type ( Table 3). The most significant decrease was observed for Y1136F-2P, consistent with a role for Tyr 1136 in stabilizing the "open" conformation.

DISCUSSION
The autophosphorylation of IGF1R at three sites within the activation loop is crucial for full catalytic activity and important for the physiological roles of the receptor in normal and transformed cells (13,19,20). Autophosphorylation of the A-loop tyrosines serves two functions: it disrupts the autoinhibitory contacts in the closed form of the receptor, and it introduces phosphate groups that stabilize the open conformation, in which ATP and protein substrates gain access to the active site (14,16). Previous studies have shown that there is a preferred sequence of A-loop phosphorylations (Tyr 1135 , followed by Tyr 1131 , then Tyr 1136 ), and that each phosphorylation increases   Our autophosphorylation assays (Fig. 2) showed that removal of individual A-loop tyrosines did not block the autophosphorylation of the remaining sites. Furthermore, for each mutant, rates of autophosphorylation at the remaining two sites exceeded the rate for wild-type IGF1RK (Fig. 2B). This effect was most pronounced for the Y1135F mutant. The Y1131F mutant also showed a significantly higher autophosphorylation rate than wild-type, while the Y1136F mutant was only slightly increased. The data suggest that the three A-loop tyrosines are involved in IGF1R autoinhibition in the following order of importance: Tyr 1135 ϾTyr 1131 ϾTyr 1136 . This possibility was borne out by steady-state kinetic measurements using the unphosphorylated forms of the enzymes (Table 1). For the mutants (particularly for Y1135F), values of V max for phosphorylation of a synthetic peptide substrate were elevated as compared with wild-type IGF1R kinase.
We also investigated how these three autophosphorylation sites contribute to the activity of the fully phosphorylated forms of the enzyme toward an exogenous substrate. If destabilization of the autoinhibited conformation is sufficient for full enzyme activity, we would expect the fully phosphorylated forms of the mutants to reach a level of activity that is comparable to wildtype IGF1RK. Our kinetic studies on the purified phosphorylated forms of the mutants ( Table 2) showed that this is not the case. The k cat /K m values for Y1135F, Y1131F, and Y136F were decreased by 9.3-fold, 2.2-fold, and 18.6-fold, respectively. For Y1136F, the K m for peptide substrate was raised by over 20fold. These results indicate that the phosphorylated A-loop tyrosines are required for full enzyme activity, in the order of importance Tyr 1136 ϾTyr 1135 ϾTyr 1131 .
We confirmed the enzyme activity results using two probes of the A-loop conformation: limited trypsin proteolyis and fluorescence quenching. Unphosphorylated IGF1RK behaved similarly to insulin receptor in these studies (the Stern-Volmer constant for unphosphorylated IR was 0.3 M Ϫ1 , compared with 0.29 M Ϫ1 for IGF1RK). In the proteolysis experiments on unphosphorylated forms of IGF1RK, we found that only the Y1135F mutant showed significant cleavage in the absence of ADP (Fig. 3A). Similarly, fluorescence experiments showed the largest degree of iodide quenching for Y1135F (Table 3). Y1131F and Y1136F also gave greater fluorescence quenching than wild-type, with the same rank order as was observed for the enzyme activity assays. The data suggest that removal of these tyrosine residues (particularly Tyr 1135 ) results in a more flexible conformation of the A-loop.
Limited proteolysis experiments on purified IGF1RK-3P showed that the A-loop was inaccessible to trypsin cleavage under these conditions, even in the presence of ADP (Fig. 3A). Phosphorylation appears to anchor the A-loop in such a way as to inhibit trypsin cleavage. In the structure of IGF1RK-3P, the phosphate group of Tyr 1135 is salt bridged to Arg 1137 within the trypsin cleavage site and the phosphate group of pY1136 is saltbridged to Lys 1138 (16). This might explain, at least in part, why trypsin can access the Arg-Lys site only when the A-loop has a flexible conformation. The phosphorylated forms of the mutants all showed enhanced cleavage relative to wild-type IGF1R, suggesting that the open form of the A-loop was destabilized in the mutants. The effect was most pronounced for Y1136F. Iodide quenching experiments showed the expected increase in quenching for IGF1RK-3P as compared with IGF1RK-0P, due to a change in solvent accessibility in the open conformation (26). The phosphorylated forms of the Y1135F, Y1131F, and Y1136F mutants showed progressively larger decreases in iodide quenching relative to IGF1RK-3P (Table 3). The results suggest that the A-loop in the single-site mutants adopts a conformation that is intermediate between IGF1RK-0P and IGF1RK-3P. The results are consistent with enzyme activity assays and point to a key role for Tyr 1136 in stabilizing the activated conformation.
Our results on the roles of the A-loop tyrosines are consistent with the available structural data. In the unphosphorylated forms of IGF1RK (17) or IRK (15), the A-loop adopts an autoinhibitory conformation with Tyr 1135 (IRK residue 1162) bound in the active site. Tyr 1135 of IGF1R interacts with Asp 1105 , Arg 1109 , and Pro 1145 (17). The side chain of Tyr 1131 is hydrogen-bonded to Ser 1059 . Phosphorylation of Tyr 1135 and Tyr 1131 would be expected to incrementally destabilize the autoinhibited conformation of the A-loop. In the structure of IGF1R-3P (Fig. 1), the phosphotyrosine side chains interact with numerous residues in the A-loop and in other regions of the enzyme (16). The phosphate group of Tyr 1135 interacts with Arg 1137 , whereas the phosphate group of Tyr 1136 is salt-bridged to Lys 1138 and Arg 1128 . The latter interaction is conserved in the structure of the tris-phosphorylated form of insulin receptor. In the crystal structure of IGF1R-2P, in which residues Tyr 1131 and Tyr 1135 are phosphorylated but Tyr 1136 is not, the central part of the activation loop is disordered (18). Tyr 1136 appears to trigger the final lobe closure in which ATP and substrate are brought close together, facilitating catalysis. Our data indicate that Tyr 1136 autophosphorylation is necessary to stabilize the activated state of IGF1R.
Our results are relevant for the development of conformationally selective inhibitors of IGF1R. Most of the clinically tested inhibitors of tyrosine kinases are compounds that target the ATP-binding site. There were initial concerns that the conserved architecture of kinase ATP binding sites would preclude the development of specific inhibitors. A large body of evidence now demonstrates that different ATP-binding sites have distinct features (shape of the pockets, composition of amino acids) that can serve as unique drug targets. In particular, it has been argued that the inactive conformations of protein kinases may differ more dramatically than the fully active conformations (27). Gleevec is an example of a tyrosine kinase inhibitor that targets the unphosphorylated, inactive conformation (28). Thus, it is possible to design inhibitors that not only discriminate among related tyrosine kinases, but also between conformational states of a given tyrosine kinase. Among other advantages of targeting the unphosphorylated form of IGF1R, there are greater differences in the ATP-binding pocket between IGF1R-0P and IR-0P than between their triply phosphorylated counterparts (17). Furthermore, the active site of IGF1R-2P is not as tightly closed as that of IGF1R-3P, suggesting that it could accommodate inhibitors that are too large to fit in the relatively small, closed IGF1R-3P active site (18). The variety of conformational states of IGF1R present potential targets for inhibitor design. For example, the cyclolignan PPP was recently shown to block the phosphorylation of Tyr 1136 , but not the other two A-loop tyrosines (29). A full understanding of the enzymatic and conformational changes that occur with each A-loop phosphorylation is an important prelude to the development of conformation-specific inhibitors.