ALK (anaplastic lymphoma kinase) is an RTK (receptor tyrosine kinase) of the IRK (insulin receptor kinase) superfamily, which share an YXXXYY autophosphorylation motif within their A-loops (activation loops). A common activation and regulatory mechanism is believed to exist for members of this superfamily typified by IRK and IGF1RK (insulin-like growth factor receptor kinase-1). Chromosomal translocations involving ALK were first identified in anaplastic large-cell lymphoma, a subtype of non-Hodgkin's lymphoma, where aberrant fusion of the ALK kinase domain with the NPM (nucleophosmin) dimerization domain results in autophosphosphorylation and ligand-independent activation. Activating mutations within the full-length ALK kinase domain, most commonly R1275Q and F1174L, which play a major role in neuroblastoma, were recently identified. To provide a structural framework for understanding these mutations and to guide structure-assisted drug discovery efforts, the X-ray crystal structure of the unphosphorylated ALK catalytic domain was determined in the apo, ADP- and staurosporine-bound forms. The structures reveal a partially inactive protein kinase conformation distinct from, and lacking, many of the negative regulatory features observed in inactive IGF1RK/IRK structures in their unphosphorylated forms. The A-loop adopts an inhibitory pose where a short proximal A-loop helix (αAL) packs against the αC helix and a novel N-terminal β-turn motif, whereas the distal portion obstructs part of the predicted peptide-binding region. The structure helps explain the reported unique peptide substrate specificity and the importance of phosphorylation of the first A-loop Tyr1278 for kinase activity and NPM–ALK transforming potential. A single amino acid difference in the ALK substrate peptide binding P−1 site (where the P-site is the phosphoacceptor site) was identified that, in conjunction with A-loop sequence variation including the RAS (Arg-Ala-Ser)-motif, rationalizes the difference in the A-loop tyrosine autophosphorylation preference between ALK and IGF1RK/IRK. Enzymatic analysis of recombinant R1275Q and F1174L ALK mutant catalytic domains confirms the enhanced activity and transforming potential of these mutants. The transforming ability of the full-length ALK mutants in soft agar colony growth assays corroborates these findings. The availability of a three-dimensional structure for ALK will facilitate future structure–function and rational drug design efforts targeting this receptor tyrosine kinase.

INTRODUCTION

ALK (anaplastic lymphoma kinase) is an RTK (receptor tyrosine kinase) implicated in multiple human cancers. Deregulation of ALK was first identified over 15 years ago in ALCL (anaplastic large-cell lymphoma), a non-Hodgkin's lymphoma subtype, where the 3′-end of the coding region for the cytoplasmic portion of ALK, containing the kinase domain, is fused to the 5′-end of the NPM (nucleophosmin) gene, a product of the recurrent chromosomal rearrangement t(2;5)(p23;q35) [1]. Subsequently, an increasing number of karyotypic aberrations involving oncogenic fusion of ALK to other fusion partners have been identified in inflammatory myofibroblastic tumours, some non-small-cell lung cancers and other solid tumours [2,3]. Fusion proteins resulting from oncogenic translocations possess constitutive tyrosine kinase activity that frequently results from fusion-partner-induced dimerization and kinase domain auto-activation. Although the exact function of full-length ALK is poorly understood, it is believed to be involved developmentally in neuronal cell differentiation and regeneration, as well as synapse formation and muscle cell migration [4].

In its native full-length single-chain receptor form, ALK consists of 1620 amino acids. A 1030-residue extracellular region encompasses multiple subdomains including an LDL-A domain (low-density lipoprotein class A domain), a MAM (meprin, A5, mu) domain, and a glycine-rich region (see Figure 1A) [5,6]. The cytoplasmic portion contains 563 residues and includes the kinase catalytic domain. This full-length form is implicated in malignancies where ALK promotes tumorigenesis via activation by autocrine and paracrine growth-promoting loops involving the putative endogenous ALK ligands PTN (pleiotrophin) and MK (midkine) [7,8]. Ligand binding leads to receptor dimerization and activation via trans-autophosphorylation of tyrosine residues within kinase domain A-loop (activation loop) segments. Phosphorylation of sites outside the A-loop within the cytoplasmic domain serve as docking sites for downstream SH2 (Src homology 2) and PTB (phosphotyrosine-binding domain) domain-containing effector and adapter proteins involved in the signal transduction cascade. Genetic abnormalities or disease states leading to ALK hyperactivation manifest in oncogenic transformation from overstimulation of downstream STAT (signal transducer and activator of transcription), Ras/Raf/MEK [MAPK (mitogen-activated protein kinase)/ERK (extracellular-signal-regulated kinase) kinase], PI3K (phosphoinositide 3-kinase)/Akt and PI3K/PLC (phospholipase C)-γ pathways involved in cell survival, differentiation and apoptosis [3]. Recently, the ALK locus was discovered to be a site of genetic alteration in the form of somatic and germline mutations that play a major role in advanced neuroblastoma [911]. Two kinase-domain-activating mutations, R1275Q and F1174L, were among the most frequent polymorphisms identified in full-length ALK. Thus ALK is emerging as an intensely pursued oncology target.

Domain architecture of full-length ALK, overall ALK crystal structure and A-loop details

Figure 1
Domain architecture of full-length ALK, overall ALK crystal structure and A-loop details

(A) Domain architecture of full-length human ALK. The extracellular domain of human ALK comprises two MAM domains (green) separated by an LDL-A domain (purple). A glycine-rich domain (G-rich; blue) precedes the transmembrane (TM)-spanning domain. The protein tyrosine kinase (PTK) domain lies in the cytoplasmic portion (red). The corresponding amino acids for each domain are labelled. (B) A representation of the apo crystal structure of ALK shown in two orthogonal orientations. The N- and C-termini are indicated by NT and CT. The NT lobe encompassing approximately residues 1093–1199 is comprises a twisted five-stranded antiparallel β-sheet (orange) and a single major helix αC (purple). The helices comprising the CT lobe (residues 1200–1399) are depicted in blue. The glycine-rich P-loop is shown in bright green, the hinge region connecting the NT and CT lobes is in yellow, the catalytic loop (C-loop) in salmon, and the A-loop in red. The β-turn found at the N-terminal-most portion of the ALK catalytic domain crystallography construct is shown in green adjacent to the αC (purple) and αAL (red) helices. The three bound glycerol molecules in the ALK structure are displayed in stick representation with carbons coloured yellow and oxygens coloured red. (C) Close-up view of the ALK A-loop interactions between the αAL helix, αC-helix and β-turn of the NT lobe. Left-hand panel: the ALK A-loop conformation is stabilized by electrostatic and polar packing interactions between the hydrophilic faces of the αAL and αC helices. The αAL helix and accompanying A-loop is shown in dark green, the αC helix is shown in purple and the catalytic loop is shown in salmon. Side chains are displayed in stick representation. Residues Arg1275 and Arg1279 of the hydrophilic face of the αAL helix co-ordinate with Asp1163 of the hydrophilic face of the αC helix from opposite sides. Arg1279 also hydrogen bonds to Gln1159 of the proximal end of the αC helix. Lys1285 downstream of αAL appears co-ordinated to both Asp1276 and Asp1160/Asp1163 side chains of the αAL helix and αC helix respectively. The positions of the three A-loop tyrosine residues (Tyr1278, Tyr1282 and Tyr1283) are shown with side chains displayed. The hydrophobic cluster also contributes to stabilizing the ALK A-loop conformation between the distal A-loop residues Leu1291/Pro1292 and Met1273 of the proximal A-loop is visible. The hydrophobic interaction between Tyr1283 and flanking Met1273 and Met1290 side chains is also shown. Side chains for catalytically important invariant Asp1249 and Asn1254 residues of the catalytic loop are also displayed. Right-hand panel: hydrophobic packing interactions between the αAL, αC helices and N-terminal β-turn also stabilize the ALK A-loop conformation. The view is shown from the perspective of the N-terminal β-turn and is rotated relative to the left-hand panel. Tyr1278 participates in van der Waals interactions with Tyr1096 of the N-terminal β-turn (grey) and Met1166 of the αC hydrophobic face. The aliphatic portions of Arg1275 and Arg1279 side chains are also directed towards the hydrophobic cluster. Residues Ile1170/Ile1171 of the C-terminal end of the αC helix, Phe1271 of the DFG motif, Phe1174 of the interconnecting strand between the αC helix and strand β5, Phe1245 of the proximal catalytic loop and Phe1098 of the N-terminal β-turn all contribute to this hydrophobic cluster. The hydrogen bond between Tyr1278 and the backbone amide nitrogen of Cys1097 is shown in addition to the hydrogen bond between Arg1279 of αAL and Gln1159 of αC. (B and C) of this Figure were generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Figure 1
Domain architecture of full-length ALK, overall ALK crystal structure and A-loop details

(A) Domain architecture of full-length human ALK. The extracellular domain of human ALK comprises two MAM domains (green) separated by an LDL-A domain (purple). A glycine-rich domain (G-rich; blue) precedes the transmembrane (TM)-spanning domain. The protein tyrosine kinase (PTK) domain lies in the cytoplasmic portion (red). The corresponding amino acids for each domain are labelled. (B) A representation of the apo crystal structure of ALK shown in two orthogonal orientations. The N- and C-termini are indicated by NT and CT. The NT lobe encompassing approximately residues 1093–1199 is comprises a twisted five-stranded antiparallel β-sheet (orange) and a single major helix αC (purple). The helices comprising the CT lobe (residues 1200–1399) are depicted in blue. The glycine-rich P-loop is shown in bright green, the hinge region connecting the NT and CT lobes is in yellow, the catalytic loop (C-loop) in salmon, and the A-loop in red. The β-turn found at the N-terminal-most portion of the ALK catalytic domain crystallography construct is shown in green adjacent to the αC (purple) and αAL (red) helices. The three bound glycerol molecules in the ALK structure are displayed in stick representation with carbons coloured yellow and oxygens coloured red. (C) Close-up view of the ALK A-loop interactions between the αAL helix, αC-helix and β-turn of the NT lobe. Left-hand panel: the ALK A-loop conformation is stabilized by electrostatic and polar packing interactions between the hydrophilic faces of the αAL and αC helices. The αAL helix and accompanying A-loop is shown in dark green, the αC helix is shown in purple and the catalytic loop is shown in salmon. Side chains are displayed in stick representation. Residues Arg1275 and Arg1279 of the hydrophilic face of the αAL helix co-ordinate with Asp1163 of the hydrophilic face of the αC helix from opposite sides. Arg1279 also hydrogen bonds to Gln1159 of the proximal end of the αC helix. Lys1285 downstream of αAL appears co-ordinated to both Asp1276 and Asp1160/Asp1163 side chains of the αAL helix and αC helix respectively. The positions of the three A-loop tyrosine residues (Tyr1278, Tyr1282 and Tyr1283) are shown with side chains displayed. The hydrophobic cluster also contributes to stabilizing the ALK A-loop conformation between the distal A-loop residues Leu1291/Pro1292 and Met1273 of the proximal A-loop is visible. The hydrophobic interaction between Tyr1283 and flanking Met1273 and Met1290 side chains is also shown. Side chains for catalytically important invariant Asp1249 and Asn1254 residues of the catalytic loop are also displayed. Right-hand panel: hydrophobic packing interactions between the αAL, αC helices and N-terminal β-turn also stabilize the ALK A-loop conformation. The view is shown from the perspective of the N-terminal β-turn and is rotated relative to the left-hand panel. Tyr1278 participates in van der Waals interactions with Tyr1096 of the N-terminal β-turn (grey) and Met1166 of the αC hydrophobic face. The aliphatic portions of Arg1275 and Arg1279 side chains are also directed towards the hydrophobic cluster. Residues Ile1170/Ile1171 of the C-terminal end of the αC helix, Phe1271 of the DFG motif, Phe1174 of the interconnecting strand between the αC helix and strand β5, Phe1245 of the proximal catalytic loop and Phe1098 of the N-terminal β-turn all contribute to this hydrophobic cluster. The hydrogen bond between Tyr1278 and the backbone amide nitrogen of Cys1097 is shown in addition to the hydrogen bond between Arg1279 of αAL and Gln1159 of αC. (B and C) of this Figure were generated with PyMOL (DeLano Scientific; http://www.pymol.org).

ALK is a member of the IRK (insulin receptor kinase) superfamily, which includes IGF1RK (insulin-like growth factor-1 receptor kinase), LTK (leucocyte tyrosine kinase) (ALK's closest relative), IRRK (insulin-receptor-related kinase) and c-ROS [12]. IGF1RK and IRK are the most well-characterized of the superfamily members and are the only ones whose structures have been elucidated. A common YXXXYY autophosphorylation motif is present within their A-loop, and catalytic activity is correlated with the degree of tyrosine residue phosphorylation [13,14]. Peptide mapping studies demonstrate A-loop autophosphorylation in IGF1RK and IRK is sequential, with the second tyrosine residue (Tyr1135/Tyr1162) preferentially phosphorylated first, followed by the first tyrosine residue (Tyr1131/Tyr1158), and lastly the third position tyrosine residue (Tyr1136/Tyr1163).

Crystal structures of IGF1RK and IRK in their unphosphorylated (PDB codes 1P4O/1IRK) [15,16] and in triphosphorylated forms in ternary complex with short synthetic substrate peptides and non-hydrolysable ATP analogues (PDB codes 1K3A/1IR3) [14,16,17] reveal the kinase conformations in their inactive and active forms (see Figure 2A). In the inactive unphosphorylated form, the A-loop binds as a pseudosubstrate with the ATP-binding site obstructed by the proximal A-loop DFG (Asp-Phe-Gly) motif (‘DFG-out’ conformation) and the P-site (phosphoacceptor site) of the substrate peptide-binding region bound by the second A-loop tyrosine residue (Tyr1135) in a cis-auto-inhibitory fashion. The preference for initial autophosphorylation of Tyr1135 is rationalized by the required release of this negative regulatory restraint for activation. In the activated triphosphorylated forms, the A-loop undergoes a dramatic conformational change to an extended form with unrestricted ATP (‘DFG-in’) and peptide substrate access [1417]. The fully activated A-loop conformation is stabilized by electrostatic interactions between the second and third phosphotyrosine residues and conserved basic A-loop residues, with the distal end adopting a β-strand secondary structure that participates in substrate peptide binding (see Figure 2B). The structure of unphosphorylated IGF1RK bound to an inhibitor exhibits the A-loop conformation during initial trans-auto-activation, where Tyr1135 is presented to the P-site of a symmetry related molecule for phosphorylation (PDB code 3D94). The A-loop conformation is also stabilized by a salt-bridge interaction. This IGF1RK structure will be referred to as the t0P form in the remainder of the text. These structures reveal snapshots of three distinct A-loop conformations that correspond to different stages of catalysis with the t0P A-loop adopting an intermediate conformation between the unphosphorylated and triphosphorylated forms. Given the degree of structural and sequence conservation among IRK superfamily members (80% sequence identity and 90% sequence conservation over 303 residues between IGF1RK and IRK; 45% sequence identity and 62% sequence conservation over 280–290 residues between ALK and IRK/IGF1RK respectively), it generally is accepted that a common regulatory and activation mechanism is shared. Together, these structures have laid the framework for typifying the conformational changes associated with kinase domain auto-activation and have provided mechanistic insights to autoregulation and catalysis for the IRK superfamily members.

ALK exhibits a distinct A-loop conformation compared with those observed for IGF1RK structures

Figure 2
ALK exhibits a distinct A-loop conformation compared with those observed for IGF1RK structures

(A) Left-hand panel: the backbone α-carbon trace of the unphosphorylated ALK structure is shown as a thick tube with β-strands of the P-loop coloured yellow, αC helix blue and A-loop green. The different A-loop conformations observed in the unphosphorylated cis-auto-inhibited inactive (unphosphorylated; 0P) (PDB code 1P4O) and fully activated (triphosphorylated; 3P) (PDB code 1K3A; the bound peptide substrate is not shown) IGF1RK structures are shown superposed and coloured purple and orange respectively. The A-loop conformation from the t0P IGF1RK structure (PDB code 3D94) representing the conformation during the initial trans-autophosphorylation event at the preferred second A-loop tyrosine residue is displayed in cyan. Trans-autophosphorylation of the three A-loop tyrosine residues in IGF1RK results in kinase activation and resulting translocation of the triphosphorylated A-loop to the other side of the lower CT lobe and conformational change to an extended form. In its unphosphorylated form, ALK exhibits a distinct A-loop conformation that does not mimic the cis-autoinhibitory pose as observed in the unphosphorylated IGF1RK structure. Rather, the proximal A-loop trajectory more closely resembles that observed in the t0P and triphosphorylated forms in that the ATP-binding site is not partially occluded. Right-hand panel: The ALK structure exhibits a ‘DFG-in’ proximal A-loop conformation and an intermediate lobe closure between inactive and fully active forms. Superposition of the ALK structure (green) with the inactive unphosphorylated (purple) and fully activated triphosphorylated (orange) IGF1RK structures with only the regions from the N-termini to the β3 strands of the N-terminal β-sheets, the αC helices and A-loops displayed. The bound peptide in the activated IGF1RK structure is shown as a yellow strand. The first two β-strands comprise the glycine-rich P-loop and strand β3 contains Lys1150/Lys1003 (ALK/IGF1RK) that forms the lysine–glutamate salt bridge with Glu1167/Glu1020 of the αC helix in active structures. The degree of lobe closure and αC-helix positioning, as well as the A-loop conformation and orientation of the DFG motif observed in the unphosphorylated ALK structure, is illustrated compared with that observed in the inactive and active IGF1RK structures. The ALK structure exhibits a non ATP-competitive ‘DFG-in’ conformation of the DFG motif at the proximal A-loop and is directly superposable with the ‘DFG-in’ conformation in the active triphosphorylated IGF1RK structure. The side chains for the aspartate and phenylalanine residues of the DFG motifs in the three structures (Asp1270/Phe1271 in ALK and Asp1123/Phe1124 in IGF1RK) are shown located centrally in the Figure but are not labelled, for clarity. Catalytically competent kinase conformations are also associated with a lobe closure resulting from rotation of the NT lobe and αC helix, relative to the CT lobe, to a more closed form. This brings a conserved glutamate residue of the αC helix into the active site where it can make a catalytically important salt-bridge interaction with the invariant β3 lysine residue of the N-terminal β-sheet (Glu1020/Lys1003 in IGF1RK). The ALK structure shows an intermediate lobe closure between the inactive ‘open’ ‘αC -helix out’ and fully active ‘closed’ ‘αC-helix in’ forms. Although side-chain atoms for Lys1150 in ALK are disordered in the apo- and ADP-bound structures (as shown), both Lys1150 and Glu1167 side chains are positioned appropriately for salt-bridge formation and approximate what is expected to be observed in the active form. The side chain density for Lys1150 in the staurosporine-bound ALK structure is fully interpretable and the salt-bridge is observed. The overall lobe closure and αC helix position in the t0P IGF1RK structure is similar to that observed for ALK except the αC helix is rotated away positioning Glu1020 away from Lys1003 (results not shown). (B) Upper-left-hand panel: view of the triphosphorylated A-loop conformation from the ternary fully activated IGF1RK structure (PDB code 1K3A). The main chain A-loop residues from Asp1123, of the DFG motif, to Glu1152 are depicted as a thick tube coloured orange with side chains displayed for Asp1123/Phe1124 at the proximal portion and Leu1143, Leu1144, Pro1145 and Met1149 at the distal portion. The phosphorylated tyrosine residues are displayed and labelled pY1131, pY1135 and pY1136. Side chains for Arg1128 and Arg1137, which stabilize the second and third phosphotyrosine residues via hydrogen bonding and electrostatic interactions (dotted green lines), are shown. The hydrophobic packing between Met1126 at the proximal A-loop and hydrophobic residues of the distal end of the A-loop is shown. Basic residues involved in the triphosphorylated IGF1RK A-loop stabilization are invariant in ALK. Upper-right-hand panel: view of the A-loop conformation from the IGF1RK t0P structure where the A-loop is stabilized for presentation of Tyr1135, the second A-loop tyrosine for trans-phosphorylation to its symmetry-related dimerization partner (PDB code 3D94). The A-loop is depicted as a thick tube coloured cyan. This A-loop conformer is stabilized by a salt-bridge interaction between Glu1132 of the ‘ETD’ sequence and Arg1137 of the downstream A-loop sequence beyond Tyr1135/Tyr1136. A similar hydrophobic interaction between Met1149 and the hydrophobic residues contribute to stabilization of the loop structure. Glu1132 is absent in ALK suggesting a similar A-loop conformation does not exist. These panels were generated with PyMOL (DeLano Scientific; http://www.pymol.org). Bottom panel: alignment of the IGF1RK, IRK and ALK A-loop sequences with every fifth residue labelled. The IGF1RK A-loop sequence is identical with IRK. The electrostatic interactions that stabilize the fully activated triphosphorylated A-loop conformation in the IGF1RK structure (PDB code1K3A) involving the second and third position phosphotyrosine residues and invariant arginine residues are shown by orange lines (pY1135–R1137 and pY1136–R1128). The stabilizing salt-bridge interaction between Glu1132 of the ‘ETD’ motif and Arg1137 that holds the IGF1RK A-loop in conformation, so that the preferred second A-loop tyrosine residue is presented optimally for trans-autphosphorylation in the t0P structure, is shown by cyan lines. Equivalent arginines, Arg1275 and Agr1284 are present in ALK suggesting the A-loop could adopt an analogous conformation in its triphosphorylated fully activated form though the extent of phosphorylation at the second and third A-loop tyrosine residues remains unclear. The presence of the corresponding ‘RAS’ motif in ALK would not provide an equivalent glutamate residue for an analogous Glu1132–Arg1137-stabilizing interaction as observed in the IGF1RK t0P A-loop structure. A hypothetical salt-bridge interaction between Arg1279 of the RAS motif and Asp1276 could be formed in ALK which would be predicted to present the preferred first A-loop tyrosine residue, Tyr1278, for the initial trans-autophosphorylation.

Figure 2
ALK exhibits a distinct A-loop conformation compared with those observed for IGF1RK structures

(A) Left-hand panel: the backbone α-carbon trace of the unphosphorylated ALK structure is shown as a thick tube with β-strands of the P-loop coloured yellow, αC helix blue and A-loop green. The different A-loop conformations observed in the unphosphorylated cis-auto-inhibited inactive (unphosphorylated; 0P) (PDB code 1P4O) and fully activated (triphosphorylated; 3P) (PDB code 1K3A; the bound peptide substrate is not shown) IGF1RK structures are shown superposed and coloured purple and orange respectively. The A-loop conformation from the t0P IGF1RK structure (PDB code 3D94) representing the conformation during the initial trans-autophosphorylation event at the preferred second A-loop tyrosine residue is displayed in cyan. Trans-autophosphorylation of the three A-loop tyrosine residues in IGF1RK results in kinase activation and resulting translocation of the triphosphorylated A-loop to the other side of the lower CT lobe and conformational change to an extended form. In its unphosphorylated form, ALK exhibits a distinct A-loop conformation that does not mimic the cis-autoinhibitory pose as observed in the unphosphorylated IGF1RK structure. Rather, the proximal A-loop trajectory more closely resembles that observed in the t0P and triphosphorylated forms in that the ATP-binding site is not partially occluded. Right-hand panel: The ALK structure exhibits a ‘DFG-in’ proximal A-loop conformation and an intermediate lobe closure between inactive and fully active forms. Superposition of the ALK structure (green) with the inactive unphosphorylated (purple) and fully activated triphosphorylated (orange) IGF1RK structures with only the regions from the N-termini to the β3 strands of the N-terminal β-sheets, the αC helices and A-loops displayed. The bound peptide in the activated IGF1RK structure is shown as a yellow strand. The first two β-strands comprise the glycine-rich P-loop and strand β3 contains Lys1150/Lys1003 (ALK/IGF1RK) that forms the lysine–glutamate salt bridge with Glu1167/Glu1020 of the αC helix in active structures. The degree of lobe closure and αC-helix positioning, as well as the A-loop conformation and orientation of the DFG motif observed in the unphosphorylated ALK structure, is illustrated compared with that observed in the inactive and active IGF1RK structures. The ALK structure exhibits a non ATP-competitive ‘DFG-in’ conformation of the DFG motif at the proximal A-loop and is directly superposable with the ‘DFG-in’ conformation in the active triphosphorylated IGF1RK structure. The side chains for the aspartate and phenylalanine residues of the DFG motifs in the three structures (Asp1270/Phe1271 in ALK and Asp1123/Phe1124 in IGF1RK) are shown located centrally in the Figure but are not labelled, for clarity. Catalytically competent kinase conformations are also associated with a lobe closure resulting from rotation of the NT lobe and αC helix, relative to the CT lobe, to a more closed form. This brings a conserved glutamate residue of the αC helix into the active site where it can make a catalytically important salt-bridge interaction with the invariant β3 lysine residue of the N-terminal β-sheet (Glu1020/Lys1003 in IGF1RK). The ALK structure shows an intermediate lobe closure between the inactive ‘open’ ‘αC -helix out’ and fully active ‘closed’ ‘αC-helix in’ forms. Although side-chain atoms for Lys1150 in ALK are disordered in the apo- and ADP-bound structures (as shown), both Lys1150 and Glu1167 side chains are positioned appropriately for salt-bridge formation and approximate what is expected to be observed in the active form. The side chain density for Lys1150 in the staurosporine-bound ALK structure is fully interpretable and the salt-bridge is observed. The overall lobe closure and αC helix position in the t0P IGF1RK structure is similar to that observed for ALK except the αC helix is rotated away positioning Glu1020 away from Lys1003 (results not shown). (B) Upper-left-hand panel: view of the triphosphorylated A-loop conformation from the ternary fully activated IGF1RK structure (PDB code 1K3A). The main chain A-loop residues from Asp1123, of the DFG motif, to Glu1152 are depicted as a thick tube coloured orange with side chains displayed for Asp1123/Phe1124 at the proximal portion and Leu1143, Leu1144, Pro1145 and Met1149 at the distal portion. The phosphorylated tyrosine residues are displayed and labelled pY1131, pY1135 and pY1136. Side chains for Arg1128 and Arg1137, which stabilize the second and third phosphotyrosine residues via hydrogen bonding and electrostatic interactions (dotted green lines), are shown. The hydrophobic packing between Met1126 at the proximal A-loop and hydrophobic residues of the distal end of the A-loop is shown. Basic residues involved in the triphosphorylated IGF1RK A-loop stabilization are invariant in ALK. Upper-right-hand panel: view of the A-loop conformation from the IGF1RK t0P structure where the A-loop is stabilized for presentation of Tyr1135, the second A-loop tyrosine for trans-phosphorylation to its symmetry-related dimerization partner (PDB code 3D94). The A-loop is depicted as a thick tube coloured cyan. This A-loop conformer is stabilized by a salt-bridge interaction between Glu1132 of the ‘ETD’ sequence and Arg1137 of the downstream A-loop sequence beyond Tyr1135/Tyr1136. A similar hydrophobic interaction between Met1149 and the hydrophobic residues contribute to stabilization of the loop structure. Glu1132 is absent in ALK suggesting a similar A-loop conformation does not exist. These panels were generated with PyMOL (DeLano Scientific; http://www.pymol.org). Bottom panel: alignment of the IGF1RK, IRK and ALK A-loop sequences with every fifth residue labelled. The IGF1RK A-loop sequence is identical with IRK. The electrostatic interactions that stabilize the fully activated triphosphorylated A-loop conformation in the IGF1RK structure (PDB code1K3A) involving the second and third position phosphotyrosine residues and invariant arginine residues are shown by orange lines (pY1135–R1137 and pY1136–R1128). The stabilizing salt-bridge interaction between Glu1132 of the ‘ETD’ motif and Arg1137 that holds the IGF1RK A-loop in conformation, so that the preferred second A-loop tyrosine residue is presented optimally for trans-autphosphorylation in the t0P structure, is shown by cyan lines. Equivalent arginines, Arg1275 and Agr1284 are present in ALK suggesting the A-loop could adopt an analogous conformation in its triphosphorylated fully activated form though the extent of phosphorylation at the second and third A-loop tyrosine residues remains unclear. The presence of the corresponding ‘RAS’ motif in ALK would not provide an equivalent glutamate residue for an analogous Glu1132–Arg1137-stabilizing interaction as observed in the IGF1RK t0P A-loop structure. A hypothetical salt-bridge interaction between Arg1279 of the RAS motif and Asp1276 could be formed in ALK which would be predicted to present the preferred first A-loop tyrosine residue, Tyr1278, for the initial trans-autophosphorylation.

Previous studies suggest ALK exhibits a unique peptide substrate specificity and auto-activation mechanism [18]. In contrast with many RTKs, a narrower substrate peptide specificity profile is reported for ALK, where it phosphorylates peptide substrates corresponding to its own A-loop sequence with greater efficiency than generic and conventional peptide substrates frequently used in tyrosine kinase activity measurements [18]. Preferential phosphorylation of the first A-loop tyrosine residue Tyr1278, with little phosphorylation of the second and third tyrosine residues was observed for ALK, in contrast with with the preferential initial autophosphorylation of the second tyrosine and ultimate triphosphorylation in the fully activate forms for IGF1RK/IRK. Tyrosine-to-phenylalanine A-loop mutants in the context of NPM–ALK demonstrate that the first tyrosine residue (Tyr1278) is essential for auto-activation of the ALK kinase domain and transforming activity [19]. These findings suggest ALK exhibits a conformational difference in its unphosphorylated form distinct from IGF1RK/IRK despite the high overall sequence identity and sequence conservation within their A-loop regions.

In the present paper we report the first X-ray crystal structure of the unphosphorylated ALK catalytic domain in the apo, ADP-bound and staurosporine-bound forms at 1.85–2.1 Å (1 Å=0.1 nm) resolution. The structures reveal an inactive protein kinase conformation distinct from, and lacking many of the negative regulatory features observed in, inactive IGF1RK/IRK structures in their unphosphorylated forms. The A-loop adopts a unique inhibitory pose where a short helix at the proximal A-loop restricts mobility of the NT lobe (N-terminal lobe) and the distal portion of the A-loop sterically obstructs a portion of the predicted peptide-binding region. A single amino acid difference in the ALK peptide-binding P−1 site was identified that, in conjunction with A-loop sequence differences including the RAS motif, rationalizes the difference in A-loop tyrosine autophosphorylation preference between ALK and IGF1RK/IRK. The ALK structure provides the molecular framework to rationalize the unique peptide substrate specificity and auto-activation mechanism reported for ALK. In addition, the ALK structure provides insights into the kinase activation mechanism for the R1275Q, F1174L and other neuroblastoma mutations. Recombinant ALK harbouring these mutations were characterized in enzymatic and soft agar colony growth transformation assays. The enzymatic data coupled with the structural data suggest that enhanced activity of these mutants may result from destabilization of the observed inactive ALK conformation. The availability of the ALK structure enables the rational design of potent and specific ALK inhibitors for the treatment of ALK-derived malignancies.

MATERIALS AND METHODS

Plasmid construction and mutagenesis

A series of ALK expression constructs (GenBank® accession number NM_004304.3 and National Center for Biotechnology Information accession number NP_004295.2; UniProt acce-ssion number, Q9UM73) encompassing the catalytic domain were engineered for insect cell expression using BV (baculovirus). Domain boundaries were selected on the basis of Pfam analysis (http://www.sanger.ac.uk/Software/Pfam/search.shtml) of the ALK sequence together with structural and sequence alignments to IRK superfamily members. Standard cloning methods were used to subclone the PCR products into the pFastBacHTb vector (Invitrogen), which encodes for an N-terminal His6 fusion tag followed by a TEV (tobacco etch virus) cleavage site. TEV cleavage of the expressed fusion product removes the N-terminal His6 tag and leaves a 5-amino-acid extension at the N-terminus of the kinase domain (GAMGS). Sequencing of the constructs revealed a single point mutation derived from the cDNA PCR template that encodes a serine-to-glycine mutation at position 1281 (see the Supplementary Materials and methods section at http://www.BiochemJ.org/bj/430/bj4300425add.htm). These constructs were evaluated in small scale expression and scaled up for subsequent crystallization trials. The R1275Q and F1174L mutations were introduced by QuikChange® (Stratagene) site-directed mutagenesis, using the pFastbacHTb ALK crystal construct as template, according to the manufacturer's instructions.

Protein expression and purification of ALK catalytic domain

Recombinant bacmid DNA was purified from DH10Bac Escherichia coli transposed with pFastBacHTb constructs and used to transfect Sf9 (Spodoptera frugiperda) cells. After one round of amplification, BV was used to infect Sf9 cultures grown in suspension at a density of ~2×106 cells/ml, at an MOI (multiplicity of infection) of 10, and harvested at 48 h post-infection. Cell pellets were resuspended in five volumes of lysis buffer {50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 10 mM imidazole and 1 mM TCEP [tris-(2-carboxyethyl)phosphine], supplemented with complete protease inhibitor cocktail (Roche)}, lysed by sonication and centrifuged at 16000 g for 50 min at 4 °C. The clarified lysate was loaded on to an Ni-NTA (Ni2+-nitrilotriacetate) chelating resin (Qiagen), washed and eluted with buffer containing 300 mM imidazole. The His6 tag was removed by TEV digestion overnight at 4 °C, dialysed against 20 mM Tris/HCl, pH 8.0, 150 mM NaCl, 0.5 mM EDTA and 0.5 mM TCEP. Digested protein was rebound to Ni-NTA resin and flow-through, containing the cleaved recombinant ALK catalytic domain, was pooled, concentrated and subjected to a final size-exclusion chromatography step on a Superdex 75 column (Pharmacia). Purified ALK was concentrated to 6–7 mg/ml prior to crystallization trials.

Crystallization of the ALK catalytic domain

Crystal trials of the purified ALK catalytic domain were performed using submicrolitre sitting drops in a 96-well plate format (Greiner) utilizing a Phoenix crystallization robot (Art Robbins Instruments) and commercial sparse matrix coarse crystal screens derived from the JCSG (Joint Center for Structural Genomics) screening conditions [20]. Equal volumes of protein and reservoir solution (250 nl) were mixed in the set-ups. Plates were incubated at both 4 and 20 °C. Apo crystals were obtained at 4 °C, for the construct encompassing residues 1072–1410, in a condition containing 0.1 M Hepes, pH 7.5 and 20% (v/v) PEG [poly(ethylene glycol)] 8000. ADP- and staurosporine-bound ALK crystals were obtained in similar conditions by incubating purified protein with 1 mM and 0.5 mM ADP and staurosporine respectively, prior to crystal set-up.

Data collection and structure determination

Complete data sets at 100 °K were collected from single crystals and vitrified using the reservoir solution supplemented with 20–25% (v/v) glycerol prior to submerging in liquid nitrogen. Apo, ADP-bound and staurosporine-bound crystals were isomorphous and belong to the orthorhombic space group P 21 21 21 (Table 1). A single molecule exists in the asymmetric unit, and the solvent content is ~40%. Data to 1.8–2.1 Å resolution were collected in a single pass on single crystals at the ALS (Advanced Light Source) beamline 5.0.3 equipped with an ADSC Quantum-4 CCD (charge-coupled-device) detector. Data were processed using DENZO and SCALEPACK of the HKL2000 program suite [21].

Table 1
Diffraction data and structural refinement statistics

RMSD, root mean square deviation.

(a) Data collection 
Parameter ALK apo ALK–ADP ALK–staurosporine 
Space group P 21 21 21 P 21 21 21 P 21 21 21 
Molecules in asymmetric unit (n
Unit cell (Å) 52.06, 56.90, 103.61 51.86, 57.12, 104.43 52.12, 57.01, 104.36 
Wavelength (Å) 0.9765 0.9765 0.9765 
Resolution (Å) 50–1.83 50–2.1 50–1.95 
Total reflections (unique) 90463(27817) 61816 (17736) 119694 (24009) 
Completeness % (highest shell) 95.5 (85.1) 94.7 (75.0) 95.4 (95.7) 
Rmerge, % (highest shell) 0.065 (0.386) 0.077 (0.380) 0.084 (0.538) 
Highest resolution shell (Å) 1.83–1.87 2.10–2.14 1.94–1.99 
Mean I/σ(I) (highest shell) 21.9 (1.8) 17.5 (2.0) 21.6 (2.1) 
Wilson B-factor (Å2) 32.9 40.3 38.8 
(b) Structural refinement 
Parameter ALK apo ALK–ADP ALK–staurosporine 
Refs. working set (n25680 16755 20915 
Refs. test set (n1372 908 1131 
Rcryst/Rfree 0.1791(0.2151) 0.2049 (0.2553) 0.1940 (0.2295) 
RMSD bonds, Å 0.016 0.016 0.009 
RMSD angles, ° 1.524 1.833 1.727 
Average B-factor (Å215.776 24.471 22.506 
Protein/ligand/glycerol/water molecules 15.2 (2396)/22.3, 26.5, 22.4/24.5 (162) 24.0 (2424) 52.3 (27) 38.8, 45.4/28.0 (94) 22.3 (2420)/21.2 (35) 27.1, 35.5/26.9 (119) 
Estimated overall coordinate error based on Rfree value (Å) 0.129 0.217 0.163 
Ramachandran Plot    
 Most favoured (%) 92.5 87.5 90.9 
 Additional allowed (%) 7.5 11.7 8.3 
 Generously allowed (%) 0.0 0.4 0.8 
 Disallowed (%) 0.0 0.4 0.0 
(a) Data collection 
Parameter ALK apo ALK–ADP ALK–staurosporine 
Space group P 21 21 21 P 21 21 21 P 21 21 21 
Molecules in asymmetric unit (n
Unit cell (Å) 52.06, 56.90, 103.61 51.86, 57.12, 104.43 52.12, 57.01, 104.36 
Wavelength (Å) 0.9765 0.9765 0.9765 
Resolution (Å) 50–1.83 50–2.1 50–1.95 
Total reflections (unique) 90463(27817) 61816 (17736) 119694 (24009) 
Completeness % (highest shell) 95.5 (85.1) 94.7 (75.0) 95.4 (95.7) 
Rmerge, % (highest shell) 0.065 (0.386) 0.077 (0.380) 0.084 (0.538) 
Highest resolution shell (Å) 1.83–1.87 2.10–2.14 1.94–1.99 
Mean I/σ(I) (highest shell) 21.9 (1.8) 17.5 (2.0) 21.6 (2.1) 
Wilson B-factor (Å2) 32.9 40.3 38.8 
(b) Structural refinement 
Parameter ALK apo ALK–ADP ALK–staurosporine 
Refs. working set (n25680 16755 20915 
Refs. test set (n1372 908 1131 
Rcryst/Rfree 0.1791(0.2151) 0.2049 (0.2553) 0.1940 (0.2295) 
RMSD bonds, Å 0.016 0.016 0.009 
RMSD angles, ° 1.524 1.833 1.727 
Average B-factor (Å215.776 24.471 22.506 
Protein/ligand/glycerol/water molecules 15.2 (2396)/22.3, 26.5, 22.4/24.5 (162) 24.0 (2424) 52.3 (27) 38.8, 45.4/28.0 (94) 22.3 (2420)/21.2 (35) 27.1, 35.5/26.9 (119) 
Estimated overall coordinate error based on Rfree value (Å) 0.129 0.217 0.163 
Ramachandran Plot    
 Most favoured (%) 92.5 87.5 90.9 
 Additional allowed (%) 7.5 11.7 8.3 
 Generously allowed (%) 0.0 0.4 0.8 
 Disallowed (%) 0.0 0.4 0.0 

The apo data set was solved by molecular replacement with Phaser (CCP4i) [22] using the apo IGF1RK structure as a search probe (PDB code 1P4O) [15]. Initial building of the 1.85 Å structure was followed by several rounds of manual building (with the COOT tool; [23]) and refinement with Refmac5 [24]. The final structure converged at an Rfactor and Rfree of 17.9 and 21.7% respectively, with excellent geometry; 91.7% of the residues were in the most favoured, 8.3% in additionally allowed, and no residues were in the disallowed regions of the Ramachandran plot (Table 1). ADP- and staurosporine-bound data sets were solved by rigid body refinement with the apo co-ordinates, and the presence of the co-crystallized ligands was observed in the initial FoFc maps.

ALK enzyme assays

Enzymatic assays for wild-type, R1275Q and F1174L ALK catalytic domains (1072–1410) were conducted using a HTRF (homogeneous time-resolved fluorescence) format with biotin–Lck peptide (New England Peptide) as the substrate. The assay mixture contained 15.7 nM wild-type ALK, 3.9 nM R1275Q mutant or 2.1 nM F1174L mutant, 1 μM biotin–Lck peptide and 1000 μM ATP in a buffer containing 50 mM Hepes, pH 7.1, 10 mM MgCl2, 2 mM MnCl2, 0.01% BSA, 2.5 mM DTT (dithiothreitol) and 0.1 mM Na3VO4, in a final volume of 10 μl. Reactions were carried out at room temperature (25 °C) in white ProxiPlate™ 384-Plus plates (PerkinElmer). At designated time points, 5 μl of EDTA (0.2 M) was added, to quench the reaction, and 5 μl of detection reagents were added at final concentrations of 3.4 ng/well for PT66K (Cis-Bio International) and 50 ng/well SAXL (Prozyme). Plates were incubated at room temperature for 1 h and analysed on an EnVision reader (Perkin Elmer).

For the apparent ATP Km determination, substrate concentration was fixed at 1 μM, with various ATP concentrations (from 1–500 μM for wild-type, 1–100 μM for R1275Q or 0.5–100 μM for F1174L ALK). Enzyme amounts used in the assays were from 15.7–78.5 nM for wild-type, 5.2–26 nM for R1275Q and 2.1-10.5 nM for F1174L ALK. The HTRF signal was corrected for the different enzyme concentrations, where applicable, and the linear portion of each curve was used to calculate initial velocity.

For the substrate biotin–Lck peptide Km determination, the ATP concentration was saturated (300 μM for wild-type and 100 μM for both R1275Q and F1174L ALK), while the biotin–Lck peptide concentration was from 10–1000 μM for wild-type, 10–500 μM for R1275Q or 10–400 μM for F1174L ALK; the amount of enzyme used was 15.7–31.4 nM for wild-type, 5.2–26 nM for R1275Q and 2.6–5.2 nM for F1174L ALK. Each reaction product was diluted to a final substrate/product concentration of 10 μM for detection as described previously [25]. The HTRF signal was corrected for the detection dilution factor, as well as the enzyme concentration, where applicable, and the linear portion of the curve was used to calculate the initial velocity. Plates were read on an EnVision multilabel plate reader.

In vitro soft agar colony formation assays

Full-length wild-type ALK or full-length ALK carrying the neuroblastoma mutation F1174L or R1275Q was cloned into a the retroviral expression vector pMSCV. Retroviruses containing empty vector, or wild-type or mutant ALK, were generated in HEK (human embryonic kidney)-293 cells and then were used to infect RIE (rat intestinal epithelial) cells. After puromycin selection, control RIE cells (RIE/vector, RIE cells carrying empty vector), RIE cells expressing wild-type ALK (RIE/WT ALK) or mutant ALKs (RIE/ALK F1174L or RIE/ALK R1275Q) were generated. For colony formation assay in soft agar, 0.6% agar in growth medium [DMEM (Dulbecco's modified Eagle's medium) containing 10% (v/v) FBS (fetal bovine serum)] was plated on to the bottom of wells (50 μl/well, 96-well plate). After the bottom layer settled, 1000 RIE cells/well were mixed with the middle layer of agar (0.36% agar in growth medium; 50 μl) and plated on top of the bottom layer. After the middle layer had settled, 50 μl of growth medium was added on top of the middle layer. After incubation at 37 °C for 8 days, the images of the cells in each well were taken under light microscope.

RESULTS AND DISCUSSION

Structure of the ALK catalytic domain and A-loop conformation

We have determined the crystal structure of the unphosphorylated ALK catalytic domain in the apo, ADP-bound and staurosporine-bound forms to 2.1 Å or better. ALK shares the canonical tyrosine kinase domain architecture and topology (Figure 1B). A smaller NT lobe is connected to a larger CT lobe (C-terminal lobe) by a loop referred to as the hinge region, which helps form the ATP-binding site at the interlobe cleft. The NT lobe predominantly comprises a five-stranded twisted antiparallel β-sheet and an adjacent single major helix, αC. The interconnecting loop between the first two β-strands of the N-terminal β-sheet contains the glycine-rich P-loop, which has the consensus GxGxΦG motif that co-ordinates the ATP phosphates. A portion of the juxtamembrane segment at the N-terminus of the ALK structure adopts a β-turn motif formed by residues 1096–1103 (-YCFAGKTS-) and packs against the distal end of the αC helix on the side opposite the β-sheet. The larger globular CT lobe is formed primarily by an eight-helix bundle (αD, αEF and αF–αI) with a small two-stranded β-sheet situated above αD and αE and adjacent to the hinge region. The A-loop encompasses residues 1270–1299 and begins with the DFG-motif and ends with residues Pro-Pro-Glu. The A-loop is fully defined in the electron density and emerges from the two-stranded β-sheet motif, adopting an overall relatively extended solvent exposed conformation. A short α-helix formed by two helical turns (αAL, residues 1272–1279; GMARDIYR) is visible at the proximal portion of the A-loop immediately following the phenylalanine of the DFG motif. The catalytic loop (residues 1247–1254) is positioned between αE and the first strand of the two-stranded β-sheet and contains invariant Asp1249 and Asn1254 residues.

αAL is positioned orthogonally below αC, and packs against it and the N-terminal β-turn motif through a network of polar and hydrophobic interactions (Figure 1C). Basic residues adorning the polar face of αAL and Lys1285 draw the proximal portion of the A-loop toward the lower and inner aspect of αC that is lined by acidic residues. Arg1275 and Arg1279 emerge from the same αAL face and flank Asp1163 on the lower aspect of αC, engaging it in electrostatic interactions. Asp1163 and Glu1167 on the same αC face in turn flank Arg1275. Lys1285 of the mid-A-loop is co-ordinated on three sides by Asp1163, and residues Asp1160 and Asp1276 of αC and αAL respectively. The Asp1276 carboxy group is directly hydrogen-bonded to the ε-nitrogen of Arg1275. Arg1279 makes an additional hydrogen bond to Gln1159 at the proximal end of αC. Arg1279 is the last residue of αAL and is part of the ‘RAS’ motif separating the first tyrosine residue (Tyr1278) from the second and third in the ALK A-loop sequence. This RAS sequence motif [Glu-Thr-Asp (ETD) in IGF1RK/IRK] has been predicted as a distinguishing A-loop feature contributing to ALK A-loop autophosphorylation efficiency and the preference for Tyr1278 [18]. Aliphatic portions of the Arg1275 and Arg1279 side chains are directed towards a dense hydrophobic cluster at the juncture of αC (Met1166 and Ile1170), the juxtamembrane N-terminal β-turn (Tyr1096 and Phe1098) and the up-stream catalytic loop region (Phe1245). Tyr1278, the first A-loop tyrosine residue reported to be preferentially autophosphorylated and critical for ALK transforming activity, is present in αAL and makes a hydrogen bond to the backbone amide nitrogen of Cys1097 of the N-terminal β-turn motif (Figure 1C, right-hand panel). Tyr1278 and Phe1271 of the DFG motif are positioned at opposite ends of αAL and together with Ala1274 pack with the hydrophobic cluster. Thus the proximal A-loop conformation is stabilized by significant αAL interactions with αC and the N-terminal β-turn. The N-terminal β-turn is not observed in the IGF1RK/IRK structures and appears unique to ALK.

Beyond αAL, the A-loop makes a sharp turn back toward the active site before descending into the CT lobe by αG (Figures 1B and 2A). Tyr1282, the second A-loop tyrosine residue, is the most solvent-exposed of the A-loop tyrosine residues and is poised below Arg1284. Tyr1283 is flipped nearly 180 ° back towards the active site and is sandwiched between hydrophobic residues of the proximal (Met1273 of αAL) and distal (Met1290/Leu1291) ends of the A-loop (Figure 1C). Despite the divergent paths observed in the unphosphorylated ALK A-loop and the three A-loop conformations observed in the IGF1RK/IRK structures, all converge at the conserved terminal hydrophobic portion of the A-loops (P+1 loop), proximal to αEF. The P+1 loop conforms to the LPXXW consensus motif (residues 1191–1195; LPVKW in ALK), which, in conjunction with the αEF helix, has been implicated in substrate recognition [26,27]. Met1273 of the proximal portion of the ALK A-loop packs against Leu1291 and Pro1292 forming a hydrophobic stem that stabilizes the A-loop structure (Figure 1C, left-hand panel). The distal portion of the ALK A-loop encompassing residues Lys1285 to Met1290 is involved in crystal contacts with primarily P-loop residues of the N-terminal β-sheet of a symmetry-related molecule in the ALK crystal lattice.

The majority of residues included in the ALK crystal construct were traceable in the 2 FoFc electron density maps with the exception of three disordered regions, which include residues of the N-terminus of the ALK construct and residual vector post-TEV cleavage, residues between strands β2/β3 of the N-terminal β-sheet and the kinase insert region residues between helices αD and αE. Residues 1072–1092 and the remaining vector sequence after TEV digestion are absent from the final models for the apo and ADP-bound structures (1072–1095 in staurosporine-bound ALK). Residues 1138–1142, corresponding to the loop region between β2 and β3, are poorly defined in the electron density maps and are excluded from the final model for the apo structure, and tentatively modelled in ADP- and staurosporine-bound structures. Residues 1214–1218 of the proline-rich kinase insertion segment were tentatively modelled in all three structures. The side chains for Asn1093 and Lys1101, and Arg1284 and Lys1285 (A-loop) were either partially or tentatively modelled in the structures. The side chain for Lys1150 of strand β3 is disordered beyond the δ-carbon in the apo and ADP-bound structures, but fully defined in the staurosporine–ALK structure.

In the apo ALK structure, three discretely bound glycerol molecules are observed from the cryopreservation (Figure 1B). One glycerol molecule is observed bound to the hinge region with the 1- and 2-position hydroxy group hydrogen bonding to the backbone amide and carbonyl of Met1199. Another glycerol molecule is observed bound below αAL and the interconnecting strand between the C-terminus of αE and αF. The 1-position hydroxy group of this glycerol participates in a shared hydrogen bond to the backbone carbonyl of Ile1246 and Nε of Arg1248 located in the proximal catalytic loop just distal to αE. The 3-position hydroxy bonds to the backbone amide -NH of Ile1246. A third glycerol molecule is visible situated near the C-terminal end of αC and beginning of strand β4 at the ‘back’ of the NT lobe. This glycerol is co-ordinated through shared hydrogen bonds between the 1-position hydroxy group and both the backbone carbonyl of Phe1174, in the distal αC helix, and the Arg1181 side chain -NH2 in β4. The 3-position hydroxy group is hydrogen-bonded to the backbone -NH of Cys1182 and participates in a water-mediated hydrogen bond with the backbone carbonyl of Val1180. Although the influence of alternate cryoprotectants was not explored, the presence of the bound glycerol molecules probably does not have an impact on the observed ALK conformation. The two non-active-site glycerols are located in similar positions in the ADP- and staurosporine-bound structures.

ADP and staurosporine binding to ALK

ALK binds ADP and staurosporine at the hinge region of the interlobe cleft as previously described in other protein kinase co-structures [2832]. The binding of ADP and staurosporine do not induce significant conformational changes or side chain movements compared with the apo form. An exception is that in the staurosporine-bound structure the P-loop is drawn more tightly over the active-site cleft reflecting the induced-fit nature of binding. In the ALK–ADP structure, the planar adenine ring is sequestered between hydrophobic residues of the NT lobe (Leu1122, Val1130 and Ala1148), hinge (Leu1198 and Met1199) and CT lobe (Leu1256) (Figure 3A). Two hydrogen bonds are observed between the N6-amino and N1 nitrogens of the 6-aminopurine ring and the backbone carbonyl and amide of Glu1197 and Met1199 hinge residues respectively. A water-mediated hydrogen bond is visible between the O3 oxygen of the ribose sugar moiety of ADP and the side chain carboxylate oxygen of Asp1203 at the downstream hinge region. The O2 oxygen of the α-phosphate is within hydrogen-bonding distance of the backbone carbonyl of His1124 of the P-loop. The ADP β-phosphate is tilted downward and deeper into the active-site cleft relative to the α-phosphate and one of the oxygens hydrogen bonds to the Asp1270 side chain of the DFG motif. A second water-mediated interaction bridges another β-phosphate oxygen to the side chain amide nitrogen of Asn1254 of the catalytic loop. These contacts and water-mediated interactions serve to co-ordinate ADP, which is probably weakly bound, as reflected by the high b-factors (52.3 Å2) associated with the ADP atoms relative to the mean b-factor (24 Å2) for the protein (Table 1).

ADP and staurosporine binding to ALK

Figure 3
ADP and staurosporine binding to ALK

(A) Close-up view of ADP bound to the ALK active site with representative amino acid residues labelled that contribute to ADP binding; 2 FoFc electron density is shown for ADP in purple contoured at 1σ. The hydrogen bonds to the hinge region and water-mediated interactions described in the text are shown. (B) Close-up view of staurosporine bound to the ALK active site with representative amino acid residues indicated involved in staurosporine binding; 2 FoFc electron density is shown for staurosporine in purple. Hydrogen bonding interactions to the hinge region and Asp1203 are displayed as well as the water-mediated hydrogen bond to Arg1253. The Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Figure 3
ADP and staurosporine binding to ALK

(A) Close-up view of ADP bound to the ALK active site with representative amino acid residues labelled that contribute to ADP binding; 2 FoFc electron density is shown for ADP in purple contoured at 1σ. The hydrogen bonds to the hinge region and water-mediated interactions described in the text are shown. (B) Close-up view of staurosporine bound to the ALK active site with representative amino acid residues indicated involved in staurosporine binding; 2 FoFc electron density is shown for staurosporine in purple. Hydrogen bonding interactions to the hinge region and Asp1203 are displayed as well as the water-mediated hydrogen bond to Arg1253. The Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Staurosporine fits snugly in the ATP-binding site within the interlobe cleft, with the amide moiety of the β-lactam ring mediating the hydrogen bond donor–acceptor interaction with the hinge region of ALK (Figure 3B). The hydrogen bond donor -NH and acceptor keto oxygen of the β-lactam ring of the inhibitor make a pair of hydrogen bonds with the backbone carbonyl oxygen of Glu1197 and the backbone amide of Met1199 respectively. The N4 nitrogen of the methylamino moiety of the tetrahydropyran ring hydrogen bonds to the carboxylate of Asp1203 of the downstream hinge region. This N4 nitrogen also appears to engage in a water-mediated hydrogen bond with the backbone carbonyl oxygen of Arg1253 of the catalytic loop. The remainder of the non-bonded interactions are largely hydrophobic and involve van der Waals interactions between the carbons of the staurosporine indolocarbazole ring system and hydrophobic residues adorning the ‘floor’ and ‘ceiling’ of the ALK active-site cleft.

Unphosphorylated ALK does not adopt a fully inactive conformation

The ALK structure, although unphosphorylated, lacks the complete repertoire of negative regulatory structural elements consistent with a fully inactive kinase conformation, as is observed in the unphosphorylated IGF1RK/IRK structures. The ALK A-loop adopts a distinct conformation that approaches a more active A-loop conformation and does not exhibit the same pseudosubstrate cis-inhibitory pose (Figures 2A and 4A). The proximal A-loop containing the DFG motif does not obstruct the ATP-binding site, which is open to binding (Figures 1B and 2A). αAL packs orthogonally directly below helix αC, extending away from the interlobe cleft. Met1273 of αAL packs with Leu1291/Pro1292 of the hydrophobic P+1 portion of the distal A-loop, a loop-stabilizing interaction only observed in the active IGF1RK/IRK A-loop conformations. The ALK A-loop DFG motif orientation is directly superposable with the non-ATP competitive ‘DFG-in’ conformation observed in active IGF1RK/IRK structures, rather than the ‘DFG-out’ conformation observed in the inactive structures (Figure 2A, right-hand panel). The relative interlobe closure between the NT- and CT-lobes in ALK is intermedi-ate between the ‘closed’ catalytically active and ‘open’ inactive IGF1RK/IRK conformations [33,34]. The αC helix is sequestered by αAL, the N-terminal β-turn, and the last two strands of the N-terminal β-sheet and, along with the NT lobe, is restricted in mobility. Lobe closure in the IGF1RK/IRK active forms orients the αC helix closer to the active site and an accompanying rotation of the helix positions an invariant Glu1020/Glu1047 side chain in closer proximity to Lys1003/Lys1030, which extends from the β3 strand of the N-terminal β-sheet, forming an catalytically important salt-bridge interaction (Figure 2A, right-hand panel). In ALK, the αC helix is rotated into the active site, properly positioning residue Glu1167 for juxtaposition with Lys1150 of strand β3. Thus the catalytically important Lys–Glu salt bridge, usually observed in active kinase conformations, is observed in the unphosphorylated ALK structure, although lobe closure appears incomplete.

Close up view of the phosphoacceptor P-site interactions in the cis-auto-inhibited unphosphorlyated IGF1RK and in the fully activated triphosphorylated IGF1RK peptide-bound structures and comparison with ALK

Figure 4
Close up view of the phosphoacceptor P-site interactions in the cis-auto-inhibited unphosphorlyated IGF1RK and in the fully activated triphosphorylated IGF1RK peptide-bound structures and comparison with ALK

(A) Portions of the unphosphorylated IGF1RK structure (PDB code 1P4O) are displayed (grey cartoon) showing the cis-auto-inhibitory pseudosubstrate A-loop (purple) conformation occupying the peptide-binding site. The second A-loop tyrosine residue (Tyr1135) binds to the phosphoacceptor P-site and hydrogen bonds to the putative catalytic base of the phosphotransfer reaction Asp1105 and Arg1109 of the catalytic loop (salmon). The adjacent Asp1134 of the ETD motif occupies the P−1-binding site and co-ordinates both Lys1058 of the αF helix and Arg1109. Asp1134 makes a hydrogen bond to Gln1181 of the interconnecting strand between αF and αG. Hydrogen bonding interactions involving the first and third A-loop tyrosine residues, Tyr1131 and Tyr1136, with Ser1059 of the αD helix and Glu1189 of the αG helix respectively, are also shown. β-Strands of the glycine-rich P-loop are shown in yellow. Ser1059 is invariant in ALK (Ser1206) whereas Glu1189 is a glutamine residue (Gln1336) in ALK. The presence of Gln1336 in ALK provides a weaker interaction with the cognate tyrosine residue if ALK adopted a similar cis-auto-inhibited A-loop conformation The Thr1127 side chain is also displayed which makes a hydrogen bond to Asn1110. This residue corresponds to ALK Ala1274, which also would not provide a similar A-loop-stabilizing interaction. (B) The same view of the peptide-binding site as observed in the peptide-bound fully-activated triphosphorylated IGF1RK structure (PDB code1K3A) is shown highlighting the interaction between the activated A-loop (orange) and ordered residues of the synthetic bound peptide (slate) occupying the P−1 to P+5 pockets. The A-loop is shown without the phosphotyrosine residues of the A-loop displayed. The αC- helix is coloured purple with the catalytic loop shown in salmon. Residues 1132–1134 of the A-loop are displayed, and correspond to the ETD motif (the Glu1132 side chain is disordered). The bound peptide is shown corresponding to the residues GEYVNIEF of the KKKSPGEYVNIEFG synthetic peptide that were ordered in the structure (the phosphoacceptor tyrosine residue is in bold). The phosphoacceptor tyrosine residue bound to the P-site in the figure is labelled with a P and makes similar P-site interactions to Tyr1135 in the cis-auto-inhibited unphosphorylated IGF1RK structure. The preceding glutamate residue of the peptide binds to the P−1 site similarly to Asp1134 in the unphosphorylated IGF1RK structure. In the triphosphorylated activated IGF1RK A-loop, the terminal portion of the A-loop forms a β-strand consisting of residues 1140–1145 (GKGLLP) not observed in the inactivated IGF1RK structure, which binds to the β-strand formed by P+1 to P+5 residues of the bound peptide, forming a two-stranded antiparallel β-sheet. IGF1RK residues from αG, αEF and the A-loop forming the hydrophobic binding groove for the P+1 to P+5 peptide residues are displayed. (C) The same view of the of the peptide substrate binding region of the unphosphorylated ALK structure with the peptide as (B) overlayed based upon superposition of the two structures. The A-loop is shown in dark green. The position of the terminal hydrophobic portion of the ALK A-loop corresponding to residues 1288–1290 (Cys-Ala-Met) would be predicted to be incompatible with peptide substrate binding and sterically block the peptide-binding site near the P, P+1 and P+3 sites. Cys1288 is observed in two alternative conformations. Met1328 of ALK is shown (Gln1181 in IGF1RK) that renders the P−1 site in ALK more hydrophobic and probably decreases the binding efficiency and phosphorylation of peptide substrates with an acidic residue at the P−1 position. The increased hydrophobicity of the ALK P−1 site by Met1328 may influence the preferred autophosphorylation of its first A-loop tyrosine residue (Tyr1278), which is preceded by an isoleucine residue, as well as disfavouring a similar cis-inhibited A-loop conformation observed in unphosphorylated IGF1RK. This Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Figure 4
Close up view of the phosphoacceptor P-site interactions in the cis-auto-inhibited unphosphorlyated IGF1RK and in the fully activated triphosphorylated IGF1RK peptide-bound structures and comparison with ALK

(A) Portions of the unphosphorylated IGF1RK structure (PDB code 1P4O) are displayed (grey cartoon) showing the cis-auto-inhibitory pseudosubstrate A-loop (purple) conformation occupying the peptide-binding site. The second A-loop tyrosine residue (Tyr1135) binds to the phosphoacceptor P-site and hydrogen bonds to the putative catalytic base of the phosphotransfer reaction Asp1105 and Arg1109 of the catalytic loop (salmon). The adjacent Asp1134 of the ETD motif occupies the P−1-binding site and co-ordinates both Lys1058 of the αF helix and Arg1109. Asp1134 makes a hydrogen bond to Gln1181 of the interconnecting strand between αF and αG. Hydrogen bonding interactions involving the first and third A-loop tyrosine residues, Tyr1131 and Tyr1136, with Ser1059 of the αD helix and Glu1189 of the αG helix respectively, are also shown. β-Strands of the glycine-rich P-loop are shown in yellow. Ser1059 is invariant in ALK (Ser1206) whereas Glu1189 is a glutamine residue (Gln1336) in ALK. The presence of Gln1336 in ALK provides a weaker interaction with the cognate tyrosine residue if ALK adopted a similar cis-auto-inhibited A-loop conformation The Thr1127 side chain is also displayed which makes a hydrogen bond to Asn1110. This residue corresponds to ALK Ala1274, which also would not provide a similar A-loop-stabilizing interaction. (B) The same view of the peptide-binding site as observed in the peptide-bound fully-activated triphosphorylated IGF1RK structure (PDB code1K3A) is shown highlighting the interaction between the activated A-loop (orange) and ordered residues of the synthetic bound peptide (slate) occupying the P−1 to P+5 pockets. The A-loop is shown without the phosphotyrosine residues of the A-loop displayed. The αC- helix is coloured purple with the catalytic loop shown in salmon. Residues 1132–1134 of the A-loop are displayed, and correspond to the ETD motif (the Glu1132 side chain is disordered). The bound peptide is shown corresponding to the residues GEYVNIEF of the KKKSPGEYVNIEFG synthetic peptide that were ordered in the structure (the phosphoacceptor tyrosine residue is in bold). The phosphoacceptor tyrosine residue bound to the P-site in the figure is labelled with a P and makes similar P-site interactions to Tyr1135 in the cis-auto-inhibited unphosphorylated IGF1RK structure. The preceding glutamate residue of the peptide binds to the P−1 site similarly to Asp1134 in the unphosphorylated IGF1RK structure. In the triphosphorylated activated IGF1RK A-loop, the terminal portion of the A-loop forms a β-strand consisting of residues 1140–1145 (GKGLLP) not observed in the inactivated IGF1RK structure, which binds to the β-strand formed by P+1 to P+5 residues of the bound peptide, forming a two-stranded antiparallel β-sheet. IGF1RK residues from αG, αEF and the A-loop forming the hydrophobic binding groove for the P+1 to P+5 peptide residues are displayed. (C) The same view of the of the peptide substrate binding region of the unphosphorylated ALK structure with the peptide as (B) overlayed based upon superposition of the two structures. The A-loop is shown in dark green. The position of the terminal hydrophobic portion of the ALK A-loop corresponding to residues 1288–1290 (Cys-Ala-Met) would be predicted to be incompatible with peptide substrate binding and sterically block the peptide-binding site near the P, P+1 and P+3 sites. Cys1288 is observed in two alternative conformations. Met1328 of ALK is shown (Gln1181 in IGF1RK) that renders the P−1 site in ALK more hydrophobic and probably decreases the binding efficiency and phosphorylation of peptide substrates with an acidic residue at the P−1 position. The increased hydrophobicity of the ALK P−1 site by Met1328 may influence the preferred autophosphorylation of its first A-loop tyrosine residue (Tyr1278), which is preceded by an isoleucine residue, as well as disfavouring a similar cis-inhibited A-loop conformation observed in unphosphorylated IGF1RK. This Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Despite the conformational similarities of the proximal A-loop with active A-loop conformations, the ALK A-loop is not in a fully active conformation capable of peptide substrate binding (Figure 2). In the ALK structure, the position of the distal end of the ALK A-loop encompassing residues 1288–1290 (-Cys-Ala-Met-) is incompatible with substrate peptide binding (Figures 4B and 4C). These residues would be predicted to occupy the P+1 and P+3 hydrophobic peptide-binding sites as identified in IGF1RK/IRK triphosphorylated structures crystallized with KKKSPGEYVNIEFG and KKKLPATGDYMNMSPVGD synthetic peptides (the Cys1288 side chain occupies the hydrophobic P+3 site and is observed in two alternate conformations). Thus although possessing structural elements observed in active kinase conformations, ALK, in its unphosphorylated form, is inactive due to the restricted lobe closure and obstruction of the peptide-binding site by the distal A-loop and only approximates an active kinase conformation.

Insights to ALK autoactivation, autoregulation and peptide substrate binding specificity

The importance of Tyr1278 phosphorylation in ALK auto-activation and transforming activity can be rationalized by the crystal structure. Phosphorylation would release the A-loop allowing it to adopt a fully activated conformation and remove the restriction of the NT lobe mobility. Additional insights into the auto-activation and peptide substrate specificity for ALK can be inferred by inspection of the corresponding ALK peptide-binding region and phosphoacceptor P-site as defined in IGF1RK/IRK structures. In IGF1RK/IRK, the second A-loop Tyr (Tyr1135 in IGF1RK) is bound in cis to the phosphoacceptor P-site as a negative regulatory element in the inactive unphosphorylated structure (Figure 4A) and in trans during initial autophosphorylation in the t0P IGF1RK structure (results not shown). The P-site in the ternary complex structures is bound by the phosphoacceptor tyrosine residues of the substrate peptide (Figure 4B). The P-site-binding determinants are essentially identical and are defined by polar and hydrophobic interactions of the bound tyrosine residue with the catalytic loop and conserved hydrophobic residues of the terminal portion of the A-loop. The tyrosine -OH is precisely positioned by hydrogen bonds with both the putative catalytic base Asp1105, and Arg1109 of the catalytic loop. The phenyl ring is cradled by hydrophobic contacts with conserved Leu1143, Leu1144 and Pro1145 at the end of the A-loop [15,16] and makes contacts with Met1126/Met1153 of the proximal A-loop in active IGF1RK/IRK structures [15,17]. The adjacent more polar P−1 site in these structures is occupied in cis or trans by the acidic residue Asp1134 that precedes the bound second A-loop tyrosine residue (Tyr1135) or the aspartate or glutamate residue preceding the phosphoacceptor tyrosine residue of the peptide in active ternary complex structures. The P−1 site largely comprises Lys1058 of the αD helix, Arg1109 of the distal end of the catalytic loop and Gln1181 positioned on the strand connecting αF and αG, which co-ordinate the acidic P−1 residue through polar and electrostatic interactions. This P−1 site interaction appears important for proper tyrosine residue positioning in the P-site and is likely to be a specificity determinant in autophosphorylation and peptide substrate specificity.

In ALK the identical P- and P−1 peptide-binding site residues are conserved with one exception. A methionine residue (Met1328) is present in the ALK P−1-binding site instead of Gln1181 in IGF1RK, rendering the site more hydrophobic (Figure 4C). A more hydrophobic P−1-binding site rationalizes the reported narrow peptide substrate specificity and preferred autophosphorylation of the first A-loop Tyr1278 in ALK [18,19]. As this site accommodates the adjacent N-terminal residue to the phosphoacceptor tyrosine residue in the substrate peptides, peptide substrates where the phosphoacceptor tyrosine residue is preceded by an acidic residue, such as aspartate or gluatmate, would bind less efficiently. The increased hydrophobicity allows a more hydrophobic residue to be accommodated at the P−1 site such as Ile1277 that directly precedes Tyr1278 in ALK, explaining the preferred autophosphorylation of this tyrosine residue. If the second Tyr were to be autophosphorylated, the polar serine residue (Ser1281) of the RAS sequence preceding Tyr1282 would occupy the P−1 site and may bind more weakly. This single amino acid difference may disfavour ALK from adopting the same cis-inhibitory pseudosubstrate A-loop conformation observed in inactive unphosphorylated IGF1RK/IRK structures. Thr1127 at the IGF1RK proximal A-loop between the DFG motif and the first A-loop tyrosine residue makes a hydrogen bond to Asn1110 of the catalytic loop and contributes to stabilization of the inhibitory A-loop conformation in the unphosphorylated structure. The sequence deviation of the ALK A-loop at this position (Ala1274) would not provide this same interaction. Another amino acid difference exists in Gln1336 located on αG of ALK. In inactive IGF1RK/IRK structures, Glu1189 exists in the analogous position and hydrogen bonds the third A-loop tyrosine hydroxy group (Figure 4A). The presence of Gln1136 may provide a weaker interaction. These amino acid differences suggest ALK is less likely to adopt the same auto-inhibited conformation as IGF1RK/IRK in its unphosphorylated form and rationalizes the unique unphosphorylated ALK structure observed.

The ordered peptide residues in bound IGF1RK/IRK triphosphorylated structures have defined the conserved P−2 to P+5 pockets of the peptide substrate binding region (Figure 4B). Side chains for residues P+1, P+3 and P+5 of the peptides occupy a hydrophobic groove located on the CT lobe formed by the hydrophobic A-loop terminus, the strand connecting to the N-terminus of αEF, αEF and αG [14,17]. These residues (IGF1RK numbering) Leu1154, Met1149, Leu1144, Gly1157, Val1146, Leu1192 and Asn1188 are invariant in ALK except for the conserved substitution of Phe1301 in ALK for Leu1154. The first three to five residues after the phosphoacceptor tyrosine residue in the bound peptides form a short β-strand that aligns in an antiparallel fashion to a short β-strand formed by residues 1141–1144 (-Lys-Gly-Leu-Leu-) of the distal A-loop region. The corresponding ALK residues are 1288–1291 (-Cys-Ala-Met-Leu-). Cys1288 was identified as a distinguishing ALK A-loop sequence feature influencing autophosphorylation efficiency and, along with the following alanine and methionine, are unique to ALK [18]. Assuming that the nature of peptide substrate binding is conserved and ALK adopts a similar A-loop conformation in its fully activated form, the -Cys-Ala-Met-Leu- residues are likely to contribute to the ALK peptide substrate specificity by adopting a similar complementary β-strand structure for peptide binding. The residues observed in phosphotyrosine-stabilizing interactions in IGF1RK/IRK ternary structures are invariant in the ALK A-loop (Figure 2B).

A key salt-bridge interaction between Glu1132 and Arg1137 on opposing sides of the IGF1RK A-loop stabilizes the loop conformation for presentation of the second tyrosine residue (Tyr1135) for trans-autophosphorylation (Figure 2B, right-hand panel). Glu1132 is in the ‘ETD’ sequence corresponding to the ‘RAS’ sequence in ALK. The presence of Arg1279 in the ‘RAS’ motif instead of Glu would not allow a similar salt bridge to be formed although the cognate arginine residue (Arg1284) is present. In order for ALK to preferentially phosphorylate the first tyrosine residue (Tyr1278), a different A-loop conformation for trans-autophosphorylation must exist that presents Tyr1278 favourably to the P-site. It would be tempting to speculate that Arg1279 of the RAS motif may participate in a similar stabilizing salt-bridge interaction with an upstream acidic residue, such as Asp1276 in the ALK A-loop trans-autophosphorylation conformation (Figure 2B, lower panel). Thus the presence of the RAS motif does not allow the necessary A-loop conformation to be stabilized for phosphorylation of the second A-loop tyrosine residue, as observed for IGF1RK, but may contribute to a distinct A-loop conformation for ALK where the first A-loop tyrosine residue is optimally presented to the phosphoacceptor-binding site for autophosphorylation.

Structural and enzymatic basis for neuroblastoma mutations

Recently, a series of ALK germline and somatic SNPs (single nucleotide polymorphisms) were identified [911] within the catalytic domain from neuroblastoma patient samples, many of which were shown to deregulate ALK and render it constitutively active (Figure 5). Availability of the three-dimensional co-ordinates of the ALK crystal structure allows for an accurate mapping of these missense mutations and provides a framework for the rationalization of the neuroblastoma mutations on ALK function. As can be seen in Figure 5, many of the mutations in the catalytic domain cluster within the N-terminal lobe, in particular, near the αC-helix. This localization of neuroblastoma mutants to this region makes sense in light of what is known about the functional role and importance of this helix in contributing to a catalytically competent protein kinase conformation. Of the nine mutants, two of the neuroblastoma-activating mutations, M1166R and I1171N, map to the αC-helix with F1174L/I located on the loop just distal to αC C-terminus. Phe1174 in the structure packs with Phe1271 of the DFG motif. The F1245C and I1250T mutations reside on the catalytic loop. Met1166, Ile1171, Phe1174 and Phe1245 are all part of the hydrophobic core at the juncture of αC, αAL and the N-terminal β-turn. Three of the mutations (G1128A, T1151M and R1192P) correspond to positions on strands of the same end of the N-terminal β-sheet, closest to αC. The G1128A mutation maps to the P-loop immediately upstream from β2 and makes hydrophobic interaction with Pro1153 located between β3 and αC. The T1151M mutation is located at the end of strand β3 and the R1192P mutation is found at the beginning of β5. Thr1151 makes a hydrogen bond to the backbone amide nitrogen of Glu1129 of the P-loop and is involved in a hydrophobic cluster of residues located at the face of the N-terminal β-sheet facing away from the active-site cleft. This residue is also adjacent to β3 Lys1150 that forms the catalytically important salt bridge to Glu1167 of αC. The G1128A and T1151M activating mutations probably result in unrestricted P-loop mobility. The R1192P mutation would be particularly disruptive to the local structure due to Pro1191 that immediately precedes it. The R1275Q mutation is found on the αAL helix of the proximal portion of the A-loop and in the structure Arg1275 provides stabilizing electrostatic and hydrogen-bonding interactions with the αC helix, contributing to A-loop stabilization in its unphosphorylated conformation (see above). Only two of the nine mutants (M1166R and R1192P) appear to map to amino acid locations that are not conserved among IRK superfamily members implicating a functional or structural role unique to ALK. In general, many of the neuroblastoma mutants probably play structural roles in stabilizing the pseudoactive ALK conformation that upon mutation allows unrestricted αC, P-loop and NT lobe mobility as well as the A-loop to adopt the necessary active conformation. This unbridled αC and NT lobe mobility translates to kinase activation. The two particular neuroblastoma mutations R1275Q and F1174L, which were identified with significantly higher frequency [911], were chosen for further biochemical characterization and structural pursuits.

Somatic- and germline-activating ALK mutations implicated in childhood neuroblastoma mapped on to the ALK crystal structure

Figure 5
Somatic- and germline-activating ALK mutations implicated in childhood neuroblastoma mapped on to the ALK crystal structure

Nine somatic and germline ALK-activating point mutations recently identified in the literature [911] are shown mapped on to the ALK crystal structure. Two orthogonal views are shown. The location of the individual point mutations are displayed as spheres and are labelled. The N- and C-termini of the structure are labelled accordingly. Regions shaded in purple correspond to those residues in structure and sequence alignments between ALK, IGF1RK and IRK that are not well conserved, whereas regions coloured in yellow represent sequence-conserved regions. Three of the activating mutations G1128A, T1151M and R1192P are coloured in red and map on the same edge of the N-terminal β-sheet closest to the αC helix. The activating mutations M1166R, I1171N and F1174L/I map to the αC helix and are shown as blue spheres. The R1275Q mutation maps to the αAL helix and is displayed as a green sphere. The I1250T and F1245C mutations map to the catalytic loop and are shown as orange spheres. The M1166R and R1192P mutations are the only two, of the nine, ALK neuroblastoma mutations that map to residue positions that are not conserved among ALK, IGF1RK and IRK. This Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Figure 5
Somatic- and germline-activating ALK mutations implicated in childhood neuroblastoma mapped on to the ALK crystal structure

Nine somatic and germline ALK-activating point mutations recently identified in the literature [911] are shown mapped on to the ALK crystal structure. Two orthogonal views are shown. The location of the individual point mutations are displayed as spheres and are labelled. The N- and C-termini of the structure are labelled accordingly. Regions shaded in purple correspond to those residues in structure and sequence alignments between ALK, IGF1RK and IRK that are not well conserved, whereas regions coloured in yellow represent sequence-conserved regions. Three of the activating mutations G1128A, T1151M and R1192P are coloured in red and map on the same edge of the N-terminal β-sheet closest to the αC helix. The activating mutations M1166R, I1171N and F1174L/I map to the αC helix and are shown as blue spheres. The R1275Q mutation maps to the αAL helix and is displayed as a green sphere. The I1250T and F1245C mutations map to the catalytic loop and are shown as orange spheres. The M1166R and R1192P mutations are the only two, of the nine, ALK neuroblastoma mutations that map to residue positions that are not conserved among ALK, IGF1RK and IRK. This Figure was generated with PyMOL (DeLano Scientific; http://www.pymol.org).

Apparent steady-state enzyme kinetic constants (Km and kcat/Km)

The apparent ATP Km values for the wild-type and mutant (R1275Q and F1174L) ALK enzymes were obtained under a fixed unsaturated peptide substrate concentration of 1 μM. The results are shown in Table 2. Wild-type ALK showed the highest ATP Km value (24 μM), whereas a 3–5-fold decrease in the Km value was observed for the R1275Q and F1174L mutants (7.6 and 5.3 μM respectively; Supplementary Figure S1 at http://www.BiochemJ.org/bj/430/bj4300425add.htm). The substrate biotin-Lck-peptide Km for the mutant ALK enzymes was obtained under a fixed saturated ATP concentration (Supplementary Figure S2 at http://www.BiochemJ.org/bj/430/bj4300425add.htm.

Table 2
Apparent ATP and peptide substrate Km and relative kcat/Km values for wild-type ALK, and the R1275Q and F1174L mutants

Note that, owing to the substrate peptide solubility limitation, a true substrate Km value for wild-type ALK could not be obtained. The kcat/Km values listed were obtained at a low substrate biotin–Lck peptide concentration of 1 μM, whereas the ATP concentration was at 1 mM. The relative values for the three constructs were compared by setting the wild-type value at 1.

Construct Apparent ATP Km (μM) Biotin–Lck peptide Km (μM) Relative kcat/Km 
ALK (wild-type) 24.0±0.7 >1300 
ALK R1275Q 7.6±0.3 114±17 4.2 
ALK F1174L 5.2±0.2 51±10 8.2 
Construct Apparent ATP Km (μM) Biotin–Lck peptide Km (μM) Relative kcat/Km 
ALK (wild-type) 24.0±0.7 >1300 
ALK R1275Q 7.6±0.3 114±17 4.2 
ALK F1174L 5.2±0.2 51±10 8.2 

The true biotin–Lck peptide Km for wild-type ALK could not be determined due to limitations imposed by the peptide substrate solubility, but appears to be much higher (>1300 μM) than the Km value for the mutants. There is an apparent decrease in the peptide Km value for the mutants, with F1174L showing the lowest value (Km=51 μM).

The catalytic efficiency of the three constructs was evaluated by comparing the kcat/Km value obtained at a low peptide substrate concentration (1 μM) and physiologically relevant ATP concentration (1 mM). The relative values for the R1275Q and the F1174L ALK mutant enzymes were normalized to the wild-type enzyme (Table 2). Both mutants showed accelerated catalytic efficiency compared with the wild-type, with R1275Q exhibiting ~4-fold and F1174L exhibiting ~8-fold higher catalytic efficiency than the wild-type enzyme.

These kinetic results demonstrate that the accelerated catalytic efficiency for the ALK mutant enzymes is at least partially due to their enhanced binding affinity for both ATP and peptide substrate. The roles that Arg1275 and Phe1174 play in stabilizing the observed auto-inhibited ALK conformation suggest mutation would destabilize these interactions, biasing the conformational equilibrium to a more active unimpeded form. This effect is most pronounced for the F1174L mutant, which demonstrated the highest catalytic efficiency (8-fold higher kcat/Km value compared with wild-type ALK enzyme). The determined ATP Km value for the wild-type ALK construct of 24 μM is approx. 30-fold lower than the Km value reported for unphosphorylated (0P) IGF1RK of 720 μM [14] and is consistent with the unobstructed ATP-binding site observed in the unphosphorylated ALK structure.

Evaluation of transforming ability of ALK and neuroblastoma mutants R1275Q and F1174L

In order to assess the transforming ability of the R1275Q and F1174L neuroblastoma mutations in ALK, the full-length wild-type and full-length ALK harbouring the R1275Q and F1174L independent mutants were evaluated for transforming ability in an anchorage-independent soft agar colony formation assay in RIE cells (Figure 6). In this assay, ALK was only able to confer transforming growth potential to RIE cells with the R1275Q and F1174L mutations, whereas vector-only controls and wild-type ALK-expressing RIE cells failed to grow. These results are consistent with these mutations enhancing ALK catalytic activity, resulting in uncontrolled cell proliferation.

Soft agar colony formation assay

Figure 6
Soft agar colony formation assay

Control RIE cells (RIE/vector) and RIE cells expressing wild-type ALK (RIE/ALK wt), ALK neuroblastoma mutant F1174L or R1275Q were plated in soft agar. After incubation at 37 °C for 8 days, colony formation was visualized under a light microscope. No colonies are observed in the vector-only control and the wild-type full-length ALK RIE wells, whereas the F1174L and R1275Q mutants afforded anchorage-independent colony formation indicative of the transforming ability of these mutants.

Figure 6
Soft agar colony formation assay

Control RIE cells (RIE/vector) and RIE cells expressing wild-type ALK (RIE/ALK wt), ALK neuroblastoma mutant F1174L or R1275Q were plated in soft agar. After incubation at 37 °C for 8 days, colony formation was visualized under a light microscope. No colonies are observed in the vector-only control and the wild-type full-length ALK RIE wells, whereas the F1174L and R1275Q mutants afforded anchorage-independent colony formation indicative of the transforming ability of these mutants.

Concluding remarks

The present study reports the structure of the unphosphorylated ALK catalytic domain and provides the first molecular details of ALK regulation and activation. The structure approaches an active RTK conformation with most of the catalytic machinery properly positioned, except for the A-loop, which adopts a unique inhibitory pose. The proximal A-loop possesses a short helix (αAL) formed in part by Arg1279 of the RAS motif that together with a novel N-terminal β-turn and the N-terminal β-sheet sequester αC and limit the interlobe mobility necessary for catalysis. The distal portion of the A-loop obstructs portion of the peptide-binding site. Residues involved in these unique A-loop regulatory interactions are sequence divergent. A single amino acid difference rendering the P−1 site more hydrophobic (Met1338) was identified that is likely to influence peptide substrate specificity, preference for autophosphorylation of the first A-loop tyrosine residue (Tyr1278), and along with the presence of Ser1281 in the RAS motif disfavours unphosphorylated ALK from adopting a similar inactive conformation to that observed in unphosphorylated IGF1RK/IRK structures. Arg1279 also may disfavour phosphorylation of the second A-loop tyrosine residue by not providing the necessary A-loop stabilizing salt-bridge interaction observed in the t0P IGF1RK structure. The A-loopstabilizing interactions that Tyr1278 provides in the structure rationalizes its reported importance in ALK activation and NPM–ALK transforming ability. Tyr1278 phosphorylation or an undescribed binding event would be required to release the restrictions imposed by the αAL and associated A-loop conformation. Curiously, Tyr1096 of the N-terminal β-turn has been identified as an ALK autophosphorylation site that may contribute to this release [35].

Previously, RTKs possessing the YXXXYY autophosphorylation motif within their A-loop peptides were considered to share a similar pseudosubstrate cis-inhibitory conformation, but the ALK structure shows that subtle sequence differences within the phosphoacceptor P−1 site and A-loop may disfavour this same conformation in ALK [14,16,36]. Negative regulatory juxtamembrane segments of several RTKs have been described for Flt-3 (FMS-related tyrosine kinase 3), c-KIT, InsR (insulin receptor) and EphB2 (ephrin receptor B2) [27]. The presence of an equivalent short αAL helix is also observed in the structures for inactive MEK (PDB code 1S9J), inactive CDK2 (cyclin-dependent kinase 2) (PDB code 1HCK), FAK (focal adhesion kinase) bound to the inhibitor TAE266 (PDB code 2JKK) and inactive c-Src (PDB code 2SRC), and it has been proposed that this small helix may serve as a site for protein–protein interactions and represent a negative regulatory motif [3739]. The presence of αAL and the accompanying structurally unique A-loop conformation may potentially be exploited in the development of ALK-specific inhibitors. The unphosphorylated ALK is amenable to co-crystallization of small molecules, as shown for ADP and staurosporine in the present study, and will facilitate current and future structure-assisted drug discovery efforts. The structure provides a framework to rationalize many of the reported neuroblastoma mutations, corroborated by our studies with the R1275Q and F1174L mutants, and catalyses future structural characterization of ALK.

Abbreviations

     
  • ALK

    anaplastic lymphoma kinase

  •  
  • A-loop

    activation loop

  •  
  • BV

    baculovirus

  •  
  • CT lobe

    C-terminal lobe

  •  
  • ERK

    extracellular-signal-regulated kinase

  •  
  • FBS

    fetal bovine serum

  •  
  • HTRF

    homogeneous time-resolved fluorescence

  •  
  • IGF1RK

    insulin-like growth factor-1 receptor kinase

  •  
  • IRK

    insulin receptor kinase

  •  
  • IRRK

    insulin-receptor-related kinase

  •  
  • LDL-A domain

    low-density lipoprotein class A domain

  •  
  • LTK

    leucocyte tyrosine kinase

  •  
  • MAM

    meprin, A5, mu

  •  
  • MAPK

    mitogen-activated protein kinase

  •  
  • MEK

    MAPK/ERK kinase

  •  
  • Ni-NTA

    Ni2+-nitrilotriacetate

  •  
  • NPM

    nucleophosmin

  •  
  • NT lobe

    N-terminal lobe

  •  
  • PI3K

    phosphoinositide 3-kinase

  •  
  • PLC

    phospholipase C

  •  
  • P-site

    phosphoacceptor site

  •  
  • RIE

    rat intestinal epithelial

  •  
  • RTK

    receptor tyrosine kinase

  •  
  • SNP

    single nucleotide polymorphism

  •  
  • TCEP

    tris-(2-carboxyethyl)phosphine

  •  
  • TEV

    tobacco etch virus

AUTHOR CONTRIBUTION

Kenneth Ng performed the insect cell expression, and Eileen Ambing and Mu-Yun Gao conducted the protein purification work. Enzyme kinetics was performed by Yong Jia. Nanxin Li, Xiuying Sun, Su Hua and Sungjoon Kim performed the soft agar colony transformation assays. Connie Chen executed the crystallization trial setups. Christian Lee designed and generated the molecular biology constructs, collected the majority of the data, solved the structures and wrote the paper. Pierre-Yves Michellys, Jennifer Harris, Scott Lesley and Glen Spraggon provided conceptual input, critical advice and helped write the paper.

We wish to thank the Advanced Light Source (ALS) Berkeley for their support on BL 5.03 where X-ray diffraction data was collected. We also thank Novartis and Peter Schultz for their continued support. We also thank Michael DiDonato for assistance in data collection.

FUNDING

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

References

References
1
Morris
S. W.
Kirstein
M. N.
Valentine
M. B.
Dittmer
K. G.
Shapiro
D. N.
Saltman
D. L.
Look
A. T.
Fusion of a kinase gene, ALK, to a nucleolar protein gene, NPM, in non-Hodgkin's lymphoma
Science
1994
, vol. 
263
 (pg. 
1281
-
1294
)
2
Kutok
J. L.
Aster
J. C.
Molecular biology of anaplastic lymphoma kinase-positive anaplastic large-cell lymphoma
J. Clin. Oncol.
2002
, vol. 
20
 (pg. 
3691
-
3702
)
3
Palmer
R. H.
Vernersson
E.
Grabbe
C.
Hallberg
B.
Anaplastic lymphoma kinase: signalling in development and disease
Biochem. J.
2009
, vol. 
420
 (pg. 
345
-
361
)
4
Chiarle
R.
Voena
C.
Ambrogio
C.
Piva
R.
Inghirami
G.
The anaplastic lymphoma kinase in the pathogenesis of cancer
Nat. Rev. Cancer
2008
, vol. 
8
 (pg. 
11
-
23
)
5
Iwahara
T.
Fujimoto
J.
Wen
D.
Cupples
R.
Bucay
N.
Arakawa
T.
Mori
S.
Ratzkin
B.
Yamamoto
T.
Molecular characterization of ALK, a receptor tyrosine kinase expressed specifically in the nervous system
Oncogene
1997
, vol. 
14
 (pg. 
439
-
449
)
6
Morris
S. W.
Naeve
C.
Mathew
P.
James
P. L.
Kirstein
M. N.
Cui
X.
Witte
D. P.
ALK, the chromosome 2 gene locus altered by the t(2;5) in non-Hodgkin's lymphoma, encodes a novel neural receptor tyrosine kinase that is highly related to leukocyte tyrosine kinase (LTK)
Oncogene
1997
, vol. 
14
 (pg. 
2175
-
2188
)
7
Stoica
G. E.
Kuo
A.
Aigner
A.
Sunitha
I.
Souttou
B.
Malerczyk
C.
Caughey
D. J.
Wen
D.
Karavanov
A.
Riegel
A. T.
Wellstein
A.
Identification of anaplastic lymphoma kinase as a receptor for the growth factor pleiotrophin
J. Biol. Chem.
2001
, vol. 
276
 (pg. 
16772
-
16779
)
8
Stoica
G. E.
Kuo
A.
Powers
C.
Bowden
E. T.
Scale
E. B.
Riegel
A. T.
Wellstein
A.
Midkine binds to anaplastic lymphoma kinase (Alk) and acts as a growth factor for different cell types
J. Biol. Chem.
2002
, vol. 
277
 (pg. 
35990
-
35998
)
9
Chen
Y.
Takita
J.
Choi
Y. L.
Kato
M.
Ohira
M.
Sanada
M.
Wang
L.
Soda
M.
Kikuchi
A.
Igarashi
T.
, et al. 
Oncogenic mutations of ALK kinase in neuroblastoma
Nature
2008
, vol. 
455
 (pg. 
971
-
974
)
10
Mosse
Y. P.
Laudenslager
M.
Longo
L.
Cole
K. A.
Wood
A.
Attiyeh
E. F.
Laquaglia
M. J.
Sennett
R.
Lynch
J. E.
Perri
P.
, et al. 
Identification of ALK as a major familial neuroblastoma predisposition gene
Nature
2008
, vol. 
455
 (pg. 
930
-
935
)
11
George
R. E.
Sanda
T.
Hanna
M.
Frohling
S.
Luther
W.
II
Zhang
J.
Ahn
Y.
Zhou
W.
London
W. B.
McGrady
P.
, et al. 
Activating mutations in ALK provide a therapeutic target in neuroblastoma
Nature
2008
, vol. 
455
 (pg. 
975
-
978
)
12
Manning
G.
Whyte
D. B.
Martinez
R.
Hunter
T.
Sudarsanam
S.
The protein kinase complement of the human genome
Science
2002
, vol. 
298
 (pg. 
1912
-
1934
)
13
Zhang
B.
Tavaré
J. M.
Ellis
L.
Roth
R. A.
The regulatory role of known tyrosine autophosphorylation sites of the insulin receptor kinase domain: an assessment by replacement with neutral and negatively charged amino acids
J. Biol. Chem.
1991
, vol. 
266
 (pg. 
990
-
996
)
14
Favelyukis
S.
Till
J. H.
Hubbard
S. R.
Miller
W. T.
Structure and autoregulation of the insulin-like growth factor 1 receptor kinase
Nat. Struct. Biol.
2001
, vol. 
8
 (pg. 
1058
-
1063
)
15
Munshi
S.
Hall
D. L.
Kornienko
M.
Darke
P. L.
Kuo
L. C.
Structure of apo, unactivated insulin-like growth factor-1 receptor kinase at 1.5 Å resolution
Acta Crystallogr. D Biol. Crystallogr.
2003
, vol. 
59
 (pg. 
1725
-
1730
)
16
Hubbard
S. R.
Wei
L.
Ellis
L.
Hendrickson
W. A.
Crystal structure of the tyrosine kinase domain of the human insulin receptor
Nature
1994
, vol. 
372
 (pg. 
746
-
754
)
17
Hubbard
S. R.
Crystal structure of the activated insulin receptor tyrosine kinase incomplex with peptide substrate and ATP analog
EMBO J.
1997
, vol. 
16
 (pg. 
5573
-
5581
)
18
Donella-Deana
A.
Marin
O.
Cesaro
L.
Gunby
R. H.
Ferrarese
A.
Coluccia
A. M. L.
Tartari
C. J.
Mologni
L.
Scapozza
L.
Gambacorti-Passerini
C.
Pinna
L. A.
Unique substrate specificity of anaplastic lymphoma kinase (ALK): development of phosphoacceptor peptides for the assay of ALK activity
Biochemistry
2005
, vol. 
44
 (pg. 
8533
-
8542
)
19
Tartari
C. J.
Gunby
R. H.
Coluccia
A. M. L
Sottocornola
R.
Cimbro
B.
Scapozza
L.
Donella-Deana
A.
Pinna
L. A.
Gambacorti-Passerini
C.
Characterization of some molecular mechanisms governing autoactivation of the catalytic domain of the anaplastic lymphoma kinase
J. Biol. Chem.
2008
, vol. 
283
 (pg. 
3743
-
3750
)
20
Page
R.
Deacon
A. M.
Lesley
S. A.
Stevens
R. C.
Shotgun crystallization strategy for structural genomics II: crystallization conditions that produce high resolution structures for T. maritima proteins
J. Struct. Funct. Genomics
2005
, vol. 
6
 (pg. 
209
-
217
)
21
Otwinowski
Z.
Minor
W.
Processing of X-ray diffraction data collected in oscillation mode
Methods Enzymol.
1997
, vol. 
276
 (pg. 
307
-
326
)
22
McCoy
A. J.
Grosse-Kunstleve
R. W.
Adams
P. D.
Winn
M. D.
Storoni
L. C.
Read
R. J.
Phaser crystallographic software
J. Appl. Cryst.
2007
, vol. 
40
 (pg. 
658
-
674
)
23
Emsley
P.
Cowtan
K.
Coot: model-building tools for molecular graphics
Acta Crystallogr. Sect. D Biol. Crystallogr.
2004
, vol. 
60
 (pg. 
2126
-
2132
)
24
Murshudov
G. N.
Vagin
A. A.
Dodson
E. J.
Refinement of macromolecular structures by the maximum-likelihood method
Acta Crystallogr. Sect. D Biol. Crystallogr.
1997
, vol. 
53
 (pg. 
240
-
255
)
25
Jia
Y.
Quinn
C. M.
Gagnon
A. I.
Talanian
R.
Homogeneous time-resolved fluorescence and its applications for kinase assays in drug discovery
Anal. Biochem.
2006
, vol. 
356
 (pg. 
273
-
281
)
26
Nolen
B.
Taylor
S.
Ghosh
G.
Regulation of protein kinases: controlling activity through activation segment conformation
Mol. Cell
2004
, vol. 
15
 (pg. 
661
-
675
)
27
Cowan-Jacob
S. W.
Structural biology of protein tyrosine kinases
Cell. Mol. Life Sci.
2006
, vol. 
63
 (pg. 
2608
-
2625
)
28
Nowakowski
J.
Cronin
C. N.
McRee
D. E.
Knuth
M. W.
Nelson
C. G.
Pavletich
N. P.
Rodgers
J.
Sang
B.-C.
Scheibe
D. N.
Swanson
R. V.
Thompson
D. A.
Structures of the cancer-related Aurora-A, FAK and EphA2 protein kinases from nanovolume crystallography
Structure
2002
, vol. 
10
 (pg. 
1659
-
1667
)
29
Westwood
I.
Cheary
D. M.
Baxter
J. E.
Richards
M. W.
Van Montfort
R. L.
Fry
A. M.
Bayliss
R.
Insights into the conformational variability and regulation of human Nek2 kinase
J. Mol. Biol.
2009
, vol. 
386
 (pg. 
476
-
485
)
30
Kinoshita
T.
Matsubara
M.
Ishiguro
H.
Okita
K.
Tada
T.
Structure of human Fyn kinase domain complexed with staurosporine
Biochem. Biophys. Res. Commun.
2006
, vol. 
346
 (pg. 
840
-
844
)
31
Ikuta
M.
Kornienko
M.
Byrne
N.
Reid
J. C.
Mizuarai
S.
Kotani
H.
Munshi
S. K.
Crystal structures of the N-terminal kinase domain of human RSK1 bound to three different ligands: implications for the design of RSK1 specific inhibitors
Protein Sci.
2007
, vol. 
16
 (pg. 
2626
-
2635
)
32
Kuglstatter
A.
Villaseñor
A. G.
Shaw
D.
Lee
S. W.
Tsing
S.
Niu
L.
Song
K. W.
Barnett
J. W.
Browner
M. F.
Cutting edge: IL-1 receptor-associated kinase 4 structures reveal novel features and multiple conformations
J. Immunol.
2007
, vol. 
178
 (pg. 
2641
-
2645
)
33
Zheng
J.
Knighton
D. R.
Xuong
N. H.
Taylor
S. S.
Sowadski
J. M.
Ten Eyck
L. F.
Crystal structures of the myristoylated catalytic subunit of cAMP-dependent protein kinase reveal open and closed conformations
Protein Sci.
1993
, vol. 
2
 (pg. 
1559
-
1573
)
34
Hanks
S. K.
Hunter
T.
Protein kinases 6. The eukaryotic protein kinase superfamily: kinase (catalytic) domain structure and classification
FASEB J.
1995
, vol. 
9
 (pg. 
576
-
596
)
35
Wang
P.
Wu
F.
Ma
Y.
Li
L.
Lai
R.
Young
L. C.
Functional characterization of the kinase activation loop in nucleophosmin (NPM)-anaplastic lymphoma kinase (ALK) using tandem affinity purification and liquid chromatography–mass spectrometry
J. Biol. Chem.
2009
, vol. 
285
 (pg. 
95
-
103
)
36
Till
J. H.
Becerra
M.
Watty
A.
Lu
Y.
Ma
Y.
Neubert
T. A.
Burden
S. J.
Hubbard
S. R.
Crystal structure of the MuSK tyrosine kinase: insights into receptor autoregulation
Structure
2002
, vol. 
10
 (pg. 
1187
-
1196
)
37
Xu
W.
Doshi
A.
Lei
M.
Eck
M. J.
Harrison
S. C.
Crystal structures of c-Src reveal features of its autoinhibitory mechanism
Mol. Cell
1999
, vol. 
3
 (pg. 
629
-
638
)
38
Ohren
J. F.
Chen
H.
Pavlovsky
A.
Whitehead
C.
Zhang
E.
Kuffa
P.
Yan
C.
McConnell
P.
Spessard
C.
Banotai
C.
, et al. 
Structures of human MAP kinase kinase 1 (MEK1) and MEK2 describe novel noncompetitive kinase inhibition
Nat. Struct. Mol. Biol.
2004
, vol. 
11
 (pg. 
1192
-
1197
)
39
Lietha
D.
Eck
M. J.
Crystal Structures of the fak kinase in complex with TAE226 and related bis-anilino pyrimidine inhibitors reveal a helical DFG conformation
PLoS ONE
2008
, vol. 
3
 (pg. 
1
-
7
)

Author notes

2

Present address: Celgene Signal Research, 4550 Towne Centre Ct., San Diego, CA 92121, U.S.A.

The final models and structure factors for the apo, ADP-bound and staurosporine-bound crystal structures have been deposited in the PDB under codes 3L9P, 3LCS and 3LCT respectively.

Supplementary data