Structural basis for reduced FGFR2 activity in LADD syndrome: Implications for FGFR autoinhibition and activation

Abstract

Mutations in fibroblast growth factor receptor 2 (FGFR2) and its ligand, FGF10, are known to cause lacrimo-auriculo-dento-digital (LADD) syndrome. Multiple gain-of-function mutations in FGF receptors have been implicated in a variety of severe skeletal disorders and in many cancers. We aimed to elucidate the mechanism by which a missense mutation in the tyrosine kinase domain of FGFR2, described in the sporadic case of LADD syndrome, leads to reduced tyrosine kinase activity. In this report, we describe the crystal structure of a FGFR2 A628T LADD mutant in complex with a nucleotide analog. We demonstrate that the A628T LADD mutation alters the configuration of key residues in the catalytic pocket that are essential for substrate coordination, resulting in reduced tyrosine kinase activity. Further comparison of the structures of WT FGFR2 and WT FGFR1 kinases revealed that FGFR2 uses a less stringent mode of autoinhibition than FGFR1, which was also manifested in faster in vitro autophosphorylation kinetics. Moreover, the nearly identical conformation of WT FGFR2 kinase and the A628T LADD mutant to either the phosphorylated FGFR2 or FGFR2 harboring pathological activating mutations in the kinase hinge region suggests that FGFR autoinhibition and activation are better explained by changes in the conformational dynamics of the kinase rather than by static crystallographic snapshots of minor structural variations.

Lacrimo-auriculo-dento-digital (LADD) syndrome is characterized by multiple congenital anomalies including aplasia of the lacrimal and salivary glands, cup-shaped, small, and low-set ears, hypo- and microdentia, hearing loss, and malformation of the digits, most commonly the thumb (1). Patients with LADD syndrome exhibit overlapping phenotypic features with aplasia (or hypoplasia) of lacrimal and salivary gland (ALSG) syndrome. Recent genetic studies have implicated mutations in fibroblast growth factor 10 (FGF10) and the “IIIb” isoform of FGFR2 (also designated FGFR2b) in LADD syndrome (2–4).

FGFR2 is a member of the FGFR family of receptor tyrosine kinases (RTK) that includes three additional receptors. FGF-receptors play an important role in the control of diverse cellular processes including cell proliferation, differentiation, migration, and maintenance of cellular homeostasis (5). The four members of the FGFR family (FGFR 1–4) each consist of an extracellular ligand-binding region composed of three Ig-like domains designated D1, D2, and D3, a single transmembrane-spanning domain, and an intracellular tyrosine kinase domain with additional regulatory sequences. FGFRs 1–4 exhibit different temporal and spatial expression patterns, and FGFRs 1–3 are additionally subject to alternative splicing in the C-terminal half of D3 to generate either the “IIIb” or the “IIIc” spliced isoforms. It was shown that the IIIb isoforms are expressed in epithelial cells, and the IIIc isoforms are expressed in the mesenchymal cells. Moreover, FGFs that activate the IIIb FGFR isoforms are expressed in mesenchyme, whereas FGFs that activate the IIIc FGFR isoforms are produced in epithelial cells (6).

After ligand binding, FGFRs undergo receptor-mediated dimerization and subsequent transautophosphorylation on numerous tyrosine residues in the tyrosine kinase core and in other parts of the cytoplasmic region (7). The tyrosine kinase domain catalyzes the transfer of the γ-phosphate of ATP to the hydroxyl group of tyrosine residues, which serves to both increase the intrinsic catalytic activity of the kinase and recruit a number of downstream signaling molecules via their conserved Src homology-2 (SH2) or phosphotyrosine-binding (PTB) domains (8). The x-ray crystal structure of the tyrosine kinase domain of FGFR1 revealed that in the inactive state, FGFR1 exists in an autoinhibited conformation in which the nucleotide-binding loop is open, but the catalytic loop is occluded by residues from the C terminus of the activation loop (9). Crystal structures of tyrosine kinase domains of other RTKs have shown a variety of autoinhibitory conformations (10), highlighting the requirement for precise regulation of the tyrosine kinase activity, a principle exemplified by the discovery of numerous gain-of-function and loss-of-function mutations in the kinase domain of FGFRs responsible for a variety of clinical entities including Crouzon syndrome, Pfeiffer syndrome, Kallmann syndrome, various cancers, and LADD syndrome, among others (11).

More recent biochemical evidence suggests that FGFR2 mutations responsible for LADD syndrome exhibit decreased tyrosine phosphorylation as well as decreased ligand-induced recruitment of downstream signaling molecules (12). However, the precise mechanism governing how these point mutations in highly conserved regions of the FGFR kinase domain lead to partial inactivation or a loss-of-function phenotype is poorly understood.

In this article, we describe the crystal structure at 1.8 Å resolution of the tyrosine kinase domain of FGFR2 harboring a single missense mutation, A628T, described in a sporadic case of LADD syndrome, in complex with a nucleotide analog. The structure precisely revealed that the A628T LADD mutation altered the catalytic pocket, which would compromise the ability of the tyrosine kinase to coordinate its substrate and thus, lead to the partial FGFR2 inactivation found in LADD syndrome. Moreover, the structure provides more detailed insight into RTK-mediated phosphotransfer and provides a molecular mechanism at atomic resolution for how this single point mutation directly affects FGFR2 enzymatic activity. We also show that FGFR2 adopts a less autoinhibited conformation relative to FGFR1, suggesting that FGFR2 uses a somewhat different mode of autoregulation to maintain its tyrosine kinase domain in an inactive state. Finally, the nearly identical conformation of WT FGFR2 kinase to those of A628T LADD mutant and various forms of activated FGFR2 kinase underscores the significance of conformational dynamics in the control of FGFR autoinhibition and activation.

Results and Discussion

In a sporadic LADD case, a de novo missense mutation was found in the catalytic loop in the tyrosine kinase domain of FGFR2 resulting in the substitution of a highly conserved alanine to a threonine (A628T) residue (13). Although recent biochemical evidence has shown that this mutation leads to decreased tyrosine phosphorylation of FGFR2 and recruitment of downstream signaling molecules (12), we sought to precisely determine how this mutation affects the intrinsic properties of the FGFR2 kinase domain. We first purified the intact kinase domains of both wild-type FGFR2 (WT) and FGFR2 harboring a single A628T point mutation (A628T) [supporting information (SI) Fig. 5]. Both WT FGFR2 and A628T-FGFR2 migrated as a monomer of ≈36 kDa on an SDS/PAGE gel and had very similar elution profiles from an anion exchange column, suggesting that the A628T mutant remained intact and that the surface charge was comparable to the WT FGFR2 tyrosine kinase.

Reduced Kinase Activity of LADD Mutant Is Not Caused by Impairment of ATP Binding.

We next asked whether the A628T mutation had a direct effect on the intrinsic catalytic activity of FGFR2 kinase. The mutation resides in center of the catalytic loop, which is a highly conserved region within the FGFR family and among other RTKs (14). An in vitro kinase assay was performed by incubating 69.4 μM WT FGFR2 or A628T-FGFR2 with ATP and MgCl₂ to a final concentration of 10 mM and 25 mM at 4°C, respectively (Fig. 1A). The reaction was stopped at various time points upon addition of EDTA to a final concentration of 100 mM, and the formation of phosphorylated species was followed by native gel electrophoresis. For WT kinase, formation of a monophosphorylated species (1P) was detected as early as 3 min, and formation of a fully phosphorylated species (5P) was visible by 45 min. In contrast, the A628T mutant failed to undergo autophosphorylation even after a 5-h incubation with ATP/MgCl₂, suggesting that the A628T mutation greatly impaired FGFR2 catalytic activity. The kinase assay was also performed at room temperature (5 mM ATP and 10 mM MgCl₂, final concentration) to determine whether the mutation renders the kinase inactive (Fig. 1B). Although tyrosine phosphorylation of the A628T-FGFR2 mutant was observed, the kinetics of autophosphorylation were significantly slower than those observed for WT FGFR2 and were similar to the phosphorylation kinetics of WT FGFR2 observed at 4°C. These results show that the A628T mutation directly compromises the intrinsic catalytic activity of FGFR2 kinase, which would account for the decreased tyrosine phosphorylation of ectopically expressed A628T-FGFR2 mutant in transfected L6 cells (12).

Fig. 1.

Reduced tyrosine kinase activity of the A628T LADD FGFR2 mutant. (A) FGFR2 WT or FGFR2-A628T (69.4 μM final concentration) were incubated with ATP and MgCl₂ at 4°C to a final concentration of 10 mM and 25 mM, respectively. Reactions were quenched at each time point upon addition of EDTA (100 mM final concentration), and the phosphorylation states were visualized by native gel electrophoresis. (B) Autophosphorylation reaction was similarly performed at room temperature upon addition of ATP and MgCl₂ to a final concentration of 5 mM and 10 mM, respectively. Reactions were quenched with EDTA (100 mM final concentration), and results were visualized by native gel electrophoresis. The positions of unphosphorylated (0P) and FGFR2 phosphorylated on 1 to 5 tyrosine residues are marked (1P–5P).

To examine the structural basis for the decreased catalytic activity of the A628T mutant, we crystallized the kinase domain of the A628T LADD mutant in complex with a nucleotide analog, AMP-PCP, in the inactive state. The 1.8 Å resolution structure revealed that the kinase domain adopts the classical RTK bilobate architecture, and clear density was observed for the activation loop, the catalytic loop including the A628T mutation, and the ATP analog (Fig. 2A and SI Fig. 6 A and C). AMP-PCP resides in the nucleotide-binding pocket between the N- and C-terminal lobes, and the adenine ring is hydrogen -bonded to the backbone of Glu-565 and Ala-567, analogous to the interactions observed between inactive FGFR1 and AMP-PCP and in other RTKs (Fig. 2B) (9, 15). The nucleotide analog does not appear to be in a productive binding mode because the γ-phosphates in both molecules in the asymmetric unit were disordered, and Mg²⁺ ions were not observed in coordination with AMP-PCP. The α-, and β-phosphate groups extend in a configuration parallel to the nucleotide-binding loop, in contrast to what was seen in the structure of inactive FGFR1 (9), but they do not make direct contacts with the protein backbone.

Fig. 2.

Overview of the structure of FGFR2 A628T LADD mutant and effect of LADD mutations on ATP binding. (A) Ribbon diagram of A628T FGFR2 structure in complex with nucleotide analog, AMP-PCP. The mutation in the catalytic loop is indicated. The nucleotide-binding loop is shown in blue, the hinge region in green, the catalytic loop in purple, and the activation loop in cyan. AMP-PCP is shown in stick representation with nitrogen atoms blue, carbon atoms white, oxygen atoms red, and phosphate atoms orange. (B) Close view of nucleotide-binding pocket. Hydrogen bonds are indicated by solid black lines. Coloring is as in A (C) Fluorescence titration experiments measuring FGFR WT or FGFR2 A628T intrinsic protein fluorescence at 340 nm in the presence of increasing concentrations of AMP-PNP from 0 to 50 μM. Data were fit to a quadratic equation by nonlinear regression (Kaleidograph). AMP-PNP-binding values for FGFR2 WT and FGFR2 A628T were K_D = 4.56 ± 1.27 μM and K_D = 6.21 ± 2.10 μM, respectively.

The rate of catalysis can also be affected by a change in the affinity of the tyrosine kinase for ATP. To determine whether the A628T LADD mutation affected ATP binding, we performed fluorescence titration experiments where WT FGFR2 or A628T-FGFR2 were incubated with increasing concentrations of AMP-PNP in the presence of excess MgCl₂. The kinase was excited at 285 nm, and the change in intrinsic protein fluorescence due to binding of the ATP analog was monitored between 300 and 420 nm (Fig. 2C). Decrease in fluorescence intensity was observed with increasing concentrations of AMP-PNP, and the data were fit to a quadratic equation. The dissociation constants obtained for WT and A628T were K_D = 4.56 ± 1.27 μM and K_D = 6.21 ± 2.10 μM, respectively, indicating that the LADD mutation did not affect ATP binding and that the diminished FGFR2 catalytic activity in the A628T mutant was not caused by decreased ATP binding.

FGFR2 Kinase Is Less Autoinhibited Than FGFR1 Kinase.

Recent crystal structures have shown that, in the inactive state, RTKs adopt a number of autoinhibitory configurations often involving inhibition of the substrate-binding pocket or blockage of the nucleotide-binding pocket (10). The structure of the A628T-FGFR2 mutant revealed that in the inactive state, both the nucleotide binding site and the substrate-binding pocket appeared accessible, and the kinase adopted a configuration much different from that observed in the structure of inactive FGFR1 (Figs. 2A and 3A) (9). To determine whether this altered configuration was due to the A628T point mutation or whether it is an intrinsic property of FGFR2 kinase, the tyrosine kinase domain of WT FGFR2 (PDB ID code, 1GJO) and the A628T-FGFR2 mutant were compared (Fig. 3C). The two FGFR2 structures were nearly identical, with an overall r.m.s.d. value of 0.7 Å. The tyrosine kinase domains of FGFR1 and FGFR2 share >90% sequence identity; however, the divergence of their three-dimensional structures warranted a detailed comparison of the two structures (Fig. 3A). Overall, the conformation of the C-terminal lobe was nearly identical, whereas the N-terminal lobe of FGFR2 relative to FGFR1 (PDB ID code, 1FGK) appeared rotated toward the C-lobe with movement of αC 20^o downward in a configuration reminiscent of an active state. In addition, the activation loop of FGFR2 is flipped outward and adopts a conformation more favorable to αC downward movement as well as more conducive to substrate binding. This conformation was observed for both WT FGFR2 and the A628T-FGFR2, which were crystallized in two different space groups and had different crystal-packing arrangements, suggesting that the conformation is intrinsic to FGFR2 kinase. Comparison of FGFR2 to activated insulin receptor kinase (IRK-3P) (PDB ID code, 1IR3) (Fig. 3B) (15) showed that the positions of αC and the activation loop, especially near the P + 1 loop, were very similar, further suggesting that FGFR2 kinase is less autoinhibited than FGFR1 kinase. Furthermore, in FGFR1 kinase, substrate binding is occluded by Pro-663 and Arg-661, which appear to directly interfere with the binding of a substrate Tyr (P-site). In contrast, in FGFR2, the corresponding residues, Pro-666 and Arg-664, are away from the catalytic pocket (Fig. 4 A and I). Pro-666 is in a conformation very similar to that found for the equivalent residue in IRK-3P and appears poised to form favorable van der Waals interactions with a substrate Tyr(P) residue. Arg-664 forms an ion pair with Asp-530 in αC, preventing it from adopting the autoinhibitory conformation found in the structure of FGFR1 and furthermore, stabilizing the active configuration of the αC helix (Fig. 4I). To determine whether the less autoinhibited conformation may affect the intrinsic catalytic activity of FGFR2 relative to FGFR1, we next incubated either 69.4 μM purified FGFR2 WT kinase or FGFR1 WT kinase with ATP and MgCl₂ to a final concentration of 10 mM and 25 mM, respectively, at 4°C and quenched the reaction mixture with 100 mM EDTA at different time points. The experiment presented in Fig. 3D shows that formation of mono-phosphorylated and bis-phosphorylated FGFR2 was observed by 1 min, with formation of the fully phosphorylated species visible by 15 min. In contrast, by 1 min, only monophosphorylated FGFR1 was visible, with formation of the fully phosphorylated species visible by 30 min, suggesting that FGFR2 is modestly more catalytically active than FGFR1, most prominently in the initial stages of phosphorylation. Despite its less autoinhibited conformation, unphosphorylated FGFR2 kinase does not fully adopt an active conformation. The N-lobe of FGFR2 was not fully closed for proper coordination of the Mg²⁺ ions or full interaction with the nucleotide analog, AMP-PCP, as evidenced from the lack of clear electron density for both the Mg²⁺ ions and the γ-phosphate of AMP-PCP and also from lack of engagement of the α- and β-phosphate groups of AMP-PCP with the protein backbone (Figs. 2B and 3B). In addition, αC is oriented such that the highly conserved Lys-517 and Glu-534 (Lys-514 and Glu-531 in FGFR1), found ≈4 Å apart after activation of IRK, are hydrogen-bonded, which prevents proper orientation of the five-stranded β-sheets of the N-lobe relative to αC and may also be inhibitory for Mg²⁺ binding. In addition, structures of activated kinases in complex with peptide analogs have shown that, in addition to accommodating the substrate Tyr(P) moiety, residues C-terminal to Tyr(P) comprise a β-strand that forms an antiparallel sheet with β11 at the C terminus of the activation loop (Fig. 4B) (15, 16). In FGFR2, residues in the C terminus of the activation loop fail to adopt this configuration but, rather, are internally stabilized by a hydrogen-bonding network consisting of Arg-649, Thr-660, Asn-662, and Arg-625 (Fig. 4I). Transition from the inactive to the active state most likely involves rearrangement of these hydrogen-bond interactions upon autophosphorylation, including the hydrogen bond between Arg-664 with Asp-530 on αC, and rearrangement of the activation loop to form interactions more similar to those seen in the crystal structure of IRK-3P and other phosphorylated RTKs.

Fig. 3.

Comparison of the structure of the tyrosine kinase domains of FGFR1, FGFR2, and insulin receptor. (A) Overlay of inactive FGFR2 kinase (PDB ID code, 1GJO) and inactive FGFR1 kinase (PDB ID code, 1FGK). FGFR2 WT is shown in green, its activation loop in red, and its catalytic loop in yellow. FGFR1 WT is illustrated in gray, its activation loop in cyan, and its catalytic loop in purple. (B) Overlay of inactive FGFR2 WT kinase (PDB ID code, 1GJO) and activated IRK (PDB ID code, 1IR3). Colors for FGFR2 WT are as in A. Activated IRK is shown in gray, its activation loop in cyan, and its catalytic loop in purple. (C) Overlay of inactive FGFR2 WT (PDB ID code, 1GJO) and FGFR2 A628T LADD mutant. Colors for FGFR2 WT are as in A. FGFR2 A628T is illustrated in gray, its activation loop in cyan, and its catalytic loop in purple. (D) FGFR2 WT or FGFR1 WT (69.4 μM final concentration) were incubated with ATP and MgCl₂ at 4°C to a final concentration of 10 mM and 25 mM, respectively. Reactions were quenched at each time point upon addition of EDTA (100 mM final concentration), and the phosphorylation states were visualized by native gel electrophoresis.

Fig. 4.

A628T LADD FGFR2 mutation compromises substrate binding. (A) Comparison of A-loop orientations of IRK-3P in complex with peptide substrate and A628T FGFR2 mutant. A-loop of IRK-3P is shown in magenta and the catalytic loop in orange. The peptide substrate is shown in pink, and the tyrosine phosphorylation site is indicated by a stick representation. The A-loop of A628T FGFR2 is colored in cyan, and the catalytic loop is shown in purple. The corresponding proline residue in FGFR2 that plays an autoregulatory role in the inactive structure of FGFR1 is indicated. (B) Catalytic pocket of IRK-3P with peptide substrate. Colors are as in A, and key hydrogen-bonding interactions are indicated by solid black lines. (C) Catalytic pocket of phosphorylated FGFR2 WT with peptide substrate (PDB ID code, 2PVF). The peptide substrate is shown in white, and the tyrosine phosphorylation site is indicated by stick representation. The A-loop of activated FGFR2 WT is shown in green, and the catalytic loop is shown in teal. Hydrogen-bonding interactions are indicated by solid black lines. (D–F) Catalytic loop of FGFR2 WT (D, yellow), A628T FGFR2 LADD mutant (E, purple), and overlay (F). Key residues involved in catalysis are indicated. (G and H) Catalytic pocket of phosphorylated WT FGFR2 (G, PDB ID code, 2PVF) and IRK-3P (H) in complex with peptide substrate. Residues involved in catalysis are indicated. (I) Interactions within the A-loop of A628T-FGFR2 LADD mutant. The activation loop is shown in cyan, the catalytic loop is shown in purple, and key hydrogen bonds are indicated by solid black lines.

A628T LADD Mutation Alters the Conformation of Key Residues Essential for Substrate Coordination.

Despite the structural differences observed within the FGFR family and the conformational changes associated with the transition from the inactive to active state, the catalytic loop is a highly conserved region in the kinase, and superimposition of the catalytic loops revealed a high degree of structural similarity among both active and inactive RTKs (Figs. 3B and 4H) (17). In the structures of the activated IRK (15) and IGFR1 (18) in complex with peptide substrate, both the catalytic base (D1132 in IRK) and the conserved arginine residue within the catalytic loop (R1136 in IRK) are hydrogen-bonded to the hydroxyl group of the substrate Tyr(P), presumably in position for proton abstraction and charge neutralization, respectively (Fig. 4B). The A628T LADD mutation resides in the center of the catalytic loop and results in substitution of a small hydrophobic residue with a more bulky, polar residue (Fig. 4E). The A628T-FGFR2 structure revealed that the threonine substitution resulted in movement of the conserved Arg-630 ≈160^o away from the catalytic base, Asp-626, in a position stabilized by interaction with Asp-521 from the neighboring molecule in the asymmetric unit. Structures of both inactive and active RTKs in the presence and absence of substrate have shown that the hydrogen bond between Arg-630 and Asp-626 in the catalytic pocket is highly conserved (Fig. 4H), and in the A628T-FGFR2 mutant, movement of Arg-630 is most likely due to steric hindrance between Arg-630 and Thr-628, because superimposition of the catalytic loops of WT FGFR2 and the A628T-FGFR2 mutant showed that the Arg-630 cannot be accommodated in the catalytic pocket upon substitution of Ala-628 with threonine (Fig. 4 D–F). The position of Asp-626 remains roughly unchanged (Fig. 4F) and seemingly, would still be in position to interact with the substrate Tyr(P). The position of Arg-630, however, is too far from the catalytic pocket to contact the substrate Tyr(P), and this conformation would hinder phosphotransfer by failing to stabilize the substrate interaction and, thus, result in severely compromised catalytic activity as seen in LADD syndrome. The crystal structure of the A628T-FGFR2 mutant presented in this study is also in agreement with the crystal structure of WT FGFR2 kinase that was recently described, most notably in the P + 1 loop (SI Fig. 7) (19). Furthermore, despite similar crystal-packing arrangements, the two structures differ in the orientation of the Arg-630 side chain in the catalytic loop (Fig. 4C, E, and G), further supporting the notion that the A628T LADD mutation is responsible for disruption of the catalytic pocket.

Inactive LADD Mutant Adopts a Virtually Identical Conformation to both Phosphorylated FGFR2 and Activated Pathological FGFR2 Mutants.

Comparison of the structure of the kinase domain of the inactive A628T-FGFR2 LADD mutant to the crystal structures of either phosphorylated or activated FGFR2 mutants (19) that have been implicated in severe bone disorders or other pathological conditions revealed nearly identical structures (SI Fig. 7). The virtually identical conformations of the kinase domain of the inactive A628T LADD mutant, unphosphorylated WT FGFR2, phosphorylated FGFR2, and several pathological activating FGFR2 mutants (SI Fig. 7) support the argument that autoinhibition is likely not caused by a static and crystallographically defined “molecular brake” in the hinge region of FGFRs and other RTKs (19), but, instead, it seems likely that the activating mutations in the hinge region exert their effects by altering the conformational dynamics. Interestingly, despite the compromised catalytic activity observed with the A628T LADD mutant, the triad of residues that comprise the molecular brake in the LADD mutant are partially disengaged in this structure and participate in a different network of hydrogen bonding interactions, suggesting that minor variations in the hinge and other flexible regions may be influenced by crystallization conditions and crystal lattices. The surprising similarity of activated and inactive forms of FGFR2 when viewed as static crystal structures suggest that both autoinhibition and activation of FGFRs and other RTKs may be dictated, to a large extent, by conformational dynamics rather than by well defined minor changes observed in FGFR2 kinase (19).

Materials and Methods

Expression, Purification, and Crystallization of A628T FGFR2 Tyrosine Kinase Domain.

A bacterial expression vector was engineered to encode residues 461–768 of the human fibroblast growth factor receptor 2 (FGFR2). For crystallization studies, a tobacco etch virus (TEV) enzymatic cleavage site was introduced after the N-terminal His₆-tag by PCR, and a single point mutation, A628T, was introduced by site-directed mutagenesis (Stratagene). Bacteria were grown at 37°C to an OD₆₀₀ of 1.3 and induced with 1 mM IPTG overnight at 20°C. The cells were lysed by French press (Thermo Scientific), and the protein was purified by using a Ni-NTA-His resin (Novagen). After elution from the column, the protein was incubated with a 1:100 molar ratio of TEV protease to A628TFGFR2 and simultaneously dialyzed in buffer containing 20 mM Tris, pH 8, and 5 mM DTT overnight at 4°C. The reaction mixture was subsequently passed over a Superdex200 HR10/30 (GE Healthcare) and further purified by using an anion exchange column on a Mono Q HR16/10 (GE Healthcare). Purified A628T FGFR2 was concentrated to 20 mg/ml in 20 mM Tris, pH 8, 200 mM NaCl, and 5 mM DTT.

Crystals were grown at 4°C by vapor diffusion in hanging drops containing 1 μl of protein solution (20 mg/ml A628T-FGFR2, 5 mM AMP-PCP, 10 mM MgCl₂) and 1 μl of reservoir solution (0.6 M NaH₂PO₄, 0.6 M KH₂PO₄, 0.1 M Hepes, pH 8). The crystals belonged to the orthorhombic space group P2₁2₁2 and had unit cell dimensions a = 67.7 Å, b = 80.3 Å, c = 118.6 Å, α = β = γ = 90°. There were two molecules per asymmetric unit, and the solvent content was ≈45% (SI Materials and Methods and SI Tables 1 and 2).

Data Collection, Structure Determination, and Analysis.

One cryocooled crystal was used for data collection. The crystal was transferred stepwise into a cryosolution containing 0.6 M NaH₂PO₄, 0.6 M KH₂PO₄, 0.1 M Hepes, pH 8, with increasing concentrations of glycerol (5%, 15%, 20%, 25%, and 30%). After ≈1 min in the final cryosolution, the crystal was flash-cooled in liquid nitrogen and transferred to the goniostat, which was bathed in a dry nitrogen stream at −180°C. Data were collected at beamline X29 at Brookhaven National Laboratory. All data were processed by using DENZO and SCALEPACK (20). A molecular replacement solution was found by using the program PHASER (21) with the FGFR2 tyrosine kinase domain (PDB ID code, 1GJO) as the search molecule. The A628T structure was subject to rigid body refinement from 50 to 3 Å by using REFMAC (22), resulting in a R_cryst value of 33.9%. Model building and refinement were carried out from 50 to 1.8 Å by using COOT (23) and CNS (24), respectively, to an R_cryst and R_free value of 20.1% and 23.1%, respectively. For calculation of R_free, 7% of the data were omitted. Figures were generated by using PyMOL (25). The stereochemistry of the model was analyzed with PROCHECK (26).