Structural basis for ligand reception by anaplastic lymphoma kinase

Abstract

The proto-oncogene ALK encodes anaplastic lymphoma kinase, a receptor tyrosine kinase that is expressed primarily in the developing nervous system. After development, ALK activity is associated with learning and memory¹ and controls energy expenditure, and inhibition of ALK can prevent diet-induced obesity². Aberrant ALK signalling causes numerous cancers³. In particular, full-length ALK is an important driver in paediatric neuroblastoma^4,5, in which it is either mutated⁶ or activated by ligand⁷. Here we report crystal structures of the extracellular glycine-rich domain (GRD) of ALK, which regulates receptor activity by binding to activating peptides^8,9. Fusing the ALK GRD to its ligand enabled us to capture a dimeric receptor complex that reveals how ALK responds to its regulatory ligands. We show that repetitive glycines in the GRD form rigid helices that separate the major ligand-binding site from a distal polyglycine extension loop (PXL) that mediates ALK dimerization. The PXL of one receptor acts as a sensor for the complex by interacting with a ligand-bound second receptor. ALK activation can be abolished through PXL mutation or with PXL-targeting antibodies. Together, these results explain how ALK uses its atypical architecture for its regulation, and suggest new therapeutic opportunities for ALK-expressing cancers such as paediatric neuroblastoma.

Extended Data Figure 1 |. Structural comparison of human and invertebrate ALK

a, Size exclusion chromatography of each protein was carried out using a Superdex 75 Increase 10/300 GL column. The non-glycosylated mass was calculated from the protein sequence. The mass corresponding to each SEC peak was determined based on the log (MW) versus elution volume plot of standards for the column. The SEC peak mass of each protein is consistent with a glycosylated monomer. b, ALK’s glycine helices project downward from the TNF-α like region. Unique to ALK, strand 10 crosses over strands 4 and 5 to terminate the fold, producing a surface terminal loop, “C-term loop”. The GRD has a distinct topology, and does not form the jelly-roll characteristic of TNF-α domains c, The GRD of human ALK. d, The hexagonal array of the Pole, and e, the order and topology of the Pole (red circles into the page, black out of the page). f, C. elegans ALK adopts a similar overall architecture to the human GRD. However, the invertebrate PXL is structurally different, forming a β-hairpin rather than helices (orange). g, The Pole of invertebrate ALK is also smaller, with 11 glycine helices that form 2 complete hexagons. h, The helical strands are in the same order and topology as in human ALK. Interestingly, the missing glycine helices of the invertebrate Pole are partially matched by non PG-II loops that interact with the hexagonal array (dashed circles). The invertebrate GRD structure additionally includes a C-terminal cysteine-rich region that leads up to the transmembrane domain. This region has 10 cysteines and forms 2 EGF-like domains (f, blue).

Extended Data Figure 2 |. Structure of invertebrate ALK’s EGF-like domains

a, The first EGF-like domain has canonical disulfide pairing. b, The second EGF-like domain is atypical in that it lacks the first 1–3 disulfide bond, the stabilizing role of which is replaced by hydrophobic interactions involving Y871 and F859. c, The EGF-like domains pack tightly to the TNF-α like domain with hydrophobic interactions. Y836 and F837 of the first EGF domain bind to the proximal C-terminal loop. F874 of the second EGF like domain is buried in a hydrophobic cavity.

Extended Data Figure 3 |. HDX analysis of ALKAL binding to ALK

HDX-MS percent exchange butterfly plot for ALKAL2-AD binding to human ALK GRD. Each peptide is assigned a peptide ID number (Extended Data Table 2) from the N- (left) to the C-terminus (right). Each grey bar shows the sum of Δ%Exchange at all labeling timepoints (Δ%sum) for each peptide. The dotted purple line corresponds to statistically significant Δ%sum (16%, or ± 0.77 Da difference in deuterium uptake between unliganded and ALKAL-bound GRD) with 98 % confidence limit calculated based on the measured standard deviation of deuterium uptake for each peptide. Regions with positive or negative Δ%Exchange become more stable or flexible, respectively, upon ALKAL binding. Statistics were derived from two independent biological repeats, each with three technical repeats. Data represent mean ± SD.

Extended Data Figure 4 |. ALK binding and signaling by ALKAL2-AD

a, Representative BLI sensograms traces (black) and kinetic fittings (red). Fitting was carried out using ForteBio Data Analysis 10.0 software using 1:1 model with Rmax linked global fitting. Where appropriate, kinetic fit parameters are included. b-c, NIH 3T3 cells stably expressing wild type (WT) or mutated full length ALK, stimulated with high concentrations (10 nM) of ALKAL2 to assess residual signaling ability. b, Single point mutations of conserved binding-site residues. E978R has the greatest impact on ALKAL induced ALK signaling in agreement with binding data. c, C-terminal conserved glutamates mutated to the residues found at the same position in invertebrate ALK. The double (E974L/E978Y) mutant fails to signal, consistent with it being the only mutation shown here that completely abolished ligand binding in (a) (Extended Data Table 3). For gel source data, see Supplementary Figure 2. The stimulation experiment was repeated three times with similar results.

Extended Data Figure 5 |. Ligand binding and receptor dimerization are coupled to Pole rotation and PXL changes.

a, View looking down the long axis of the Pole with apo ALK (gray) aligned to the complex structure (color). Compared to unliganded ALK, the ligand-bound ALK dimer undergoes a clockwise rotation of the Pole about the central glycine helix (number 4). b, Unliganded ALK (gray) aligned to both protomers of the complex dimer (color). The PXL residues surrounding Q788 adopt a helical structure upon ligand binding and dimerization. c, Upon ligand binding, two additional peptides – not directly involved in ligand binding and discussed in Fig. 2 – are significantly protected. Statistics were derived from two independent biological repeats, each with three technical repeats. Data represent mean ± SD. d, these peptides (cyan) (peptide #1 and #21, Extended data Fig. 3) correspond to the dimer interface including the disulfide linked helix of the PXL (#21) and the helix on the TNF-α like domain it forms a bond with (#1). e, The core helix-turn-helix of the AD is largely unaltered upon binding. ALKAL2 from the ALK-ALKAL fusion complex structure (green) is aligned to an un-complexed ALKAL1-AD MBP (Maltose Binding Protein) fusion (cyan).

Extended Data Figure 6 |. The PXL region is necessary for ALK signaling

a, Binding of ALKAL2-AD to GRD PXL mutations detected by BLI. The sensors were loaded with ALKAL2-AD. REQ is a mutant that alters the interface residues 795-“IGE”−797 (shown in Fig. 3b, c) to REQ. ΔPXL is a mutant that removes the entire “disulfided helix” 783–797 (shown in Fig. 3b, c). The relative binding is the binding normalized to the maximum responses. Data represent mean ± SD of four measurements. b, Expression of Halo-ALK on NIH/3T3 cell membranes. NIH/3T3 cells were untransfected, transfected with wildtype Halo-ALK or ALK mutants. Cells were stained with membrane impermeable dye, JF549i, before imaging. For each construct fluorescence (top row), brightfield (middle row), and a merge (bottom row) were shown. Scale bar 5 μm. The experiment was repeated two times with similar results. c, Binding of CDX123 or CDX125 to GRD or GRDΔPXL detected by BLI. Data represent mean ± SD from three independent experiments.

Extended Data Figure 7 |

Implications for GRD mutation and evolution a, Invertebrate ALK GRD structure with known glycine-to-acidic residue (D/E) mutations highlighted (purple spheres). b, Human ALK GRD structure with all of glycine-to-acidic residue mutations found in the COSMIC database (https://cancer.sanger.ac.uk) highlighted (purple spheres). c, Invertebrate ALK GRD does not bind to human ALKAL2-AD. The relative binding is the binding normalized to the maximum responses. Data represent mean ± SD of three measurements. d, The EGF-like domain of invertebrate ALK occupies a region of the C-terminal loop required for ALKAL binding. This would prevent binding of any similar helix-loop-helix like ligand.

Fig. 1 |. Structure of the ALK-ALKAL complex and ALK’s unusual glycine rich domain

a, The (2:2, ALK:ALKAL) complex. One of two complexes in the crystal’s asymmetric unit (ASU) is shown. ALK’s GRD is colored by regions of predominant secondary structure (β-strands, red; α-helices, orange; strands, yellow). ALKAL is colored green. b, ALK’s GRD shown in surface representation. View orthogonal to (a). c, ALK’s GRD has a TNF-α like region (red) with an unusual Poly-Glycine Extension (Pole, yellow). Displayed on one end of the Pole are Poly-glycine eXtension Loops (PXL, orange). The primary loop helix is stabilized by a single disulfide, “disulfided helix” (asterisk). d, An enlarged orthogonal view details the Pole’s hexagonal lattice of rare glycine helices. Hydrogen bonds (dashed lines) link the polypeptide backbone to neighboring helices.

Fig. 2 |. Ligand binding is directed by ALK’s C-terminal loop

a, ALKAL binds to the TNF-α like region of ALK. b, The primary ligand-interacting residues on ALK involve the C-terminal loop of the TNF-α like region. The distal segment of the loop, including glutamates E974 and E978 is detailed. An additional glutamate, E859, outside the TNF-α like region interacts with the ALKAL2-AD. The positively charged arginine-containing surface of ALKAL faces the negatively charged glutamate cluster of ALK. c, Interactions involving the non-polar initial segment of ALK’s C-terminal loop (Y966-L970) are detailed. d, Statistically significant protection seen in HDX-MS studies mapped onto the crystal structure of human ALK GRD. Regions that show statistically significant protection when bound by ALKAL are colored blue. e, Percent exchange plots for the C-terminal loop peptides 964–970 and 967–977 with or without ALKAL2-AD as a function of exposure time. Data represent mean ± SD of two biologically independent measurements of which three technical repeats were carried out.

Fig. 3 |. Dimerization of ALK is directed by the PXL

a, The ALK-ALKAL fusion complex forms a dimer (2:2, ALK:ALKAL). For clarity, the second receptor is shown in surface representation (left) with both bound ligands in green cartoon. b, An enlarged view of the boxed dimer interface from (a) to show details of the interaction. The PXL of one receptor (orange, Complex A) makes contacts with an epitope that combines ALKAL (green) and the TNF-α like region (red) of a second ligand-bound receptor (Complex B). c, An orthogonal view of the interface looking down the long axis of the Pole. d, NIH/3T3 cells stably expressing Halo-tagged ALK or ALK mutants were stimulated with a low concentration of purified ALKAL2-AD (0.5 nM for 10 min) to assess ligand responsiveness. REQ is a mutant that alters the three interface residues 795–797 (IGE, shown in b, c) to REQ. ΔPXL is a mutant that removes the entire “disulfided helix” 783–797 (shown in b,c). Data represent five independent measurements. For gel source data, see Supplementary Figure 1.

Fig. 4 |. Antibodies to the C-terminal loop or PXL prevent ALK activation

a, Percent exchange plots for residues 790–798 of the PXL as a function of exposure time. b, Percent exchange plots for the residues 966–972 of the C-term loop as a function of exposure time. c, Binding of GRD or a GRD-antibody complex to ALKAL2-AD detected by BLI. d, Neuroblastoma (IMR32) cells preincubated with the indicated concentrations of fab123 were stimulated with purified ALKAL2-AD. Phosphorylation of endogenous ALK at Y1604 (pALK) and total ALK were assessed by immunoblotting and quantified. Data represent mean ± SD of three independent measurements. For gel source data, see Supplementary Figure 1.