Structural basis of cytokine-mediated activation of ALK family receptors

Abstract

Anaplastic lymphoma kinase (ALK)¹ and the related leukocyte tyrosine kinase (LTK)² are recently deorphanized receptor tyrosine kinases³. Together with their activating cytokines, ALKAL1 and ALKAL2^4–6 (also called FAM150A and FAM150B or AUGβ and AUGα, respectively), they are involved in neural development⁷, cancer^7–9 and autoimmune diseases¹⁰. Furthermore, mammalian ALK recently emerged as a key regulator of energy expenditure and weight gain¹¹, consistent with a metabolic role for Drosophila ALK¹². Despite such functional pleiotropy and growing therapeutic relevance^{13, 14}, structural insights into ALK and LTK and their complexes with cognate cytokines have remained scarce. Here we show that the cytokine-binding segments of human ALK and LTK comprise a novel architectural chimera of a permuted TNF-like module that braces a glycine-rich subdomain featuring a hexagonal lattice of long polyglycine type II helices. The cognate cytokines ALKAL1 and ALKAL2 are monomeric three-helix bundles, yet their binding to ALK and LTK elicits similar dimeric assemblies with two-fold symmetry, that tent a single cytokine molecule proximal to the cell membrane. We show that the membrane-proximal EGF-like domain dictates the apparent cytokine preference of ALK. Assisted by these diverse structure–function findings, we propose a structural and mechanistic blueprint for complexes of ALK family receptors, and thereby extend the repertoire of ligand-mediated dimerization mechanisms adopted by receptor tyrosine kinases.

ALK is an evolutionarily ancient receptor tyrosine kinase (RTK). The vertebrate orthologues are uniquely endowed with an amino-terminal heparin-binding domain (HBD) (Fig. 1a, Extended Data Fig. 1a, b). Gene duplication in vertebrates resulted in LTK, a second ALK-like receptor¹⁵, which evolved divergently with loss of the HBD and the MAM-LDLa-MAM module in mammals (Extended Data Fig. 1c). The common architectural hallmark in the ectodomains of ALK and LTK comprises their cytokine-binding segment, which is unique among cytokine receptors and features an array of a TNF-like (TNFL) module, a glycine-rich (GR) region, and a membrane-proximal EGF-like module (EGFL) (Fig. 1a). Whereas ALKAL1 and ALKAL2 are both strong activators of LTK⁴, only ALKAL2 potently activates ALK^{5, 6}–this activation is coupled to additional regulation via glycosaminoglycan binding to its HBD¹⁶.

Fig. 1 |. Structure of the ALK family cytokine-binding domain.

a, Domain organization of ALK and LTK, and ligands. MAM, meprin–A5-protein–PTPμ domain; LDLa, low-density lipoprotein class A repeat; HBD, heparin-binding domain. b, c, Crystal structures of ALK_TG–EGFL (b) and LTK_TG (c). Disulfides (yellow spheres) and experimentally observed N-glycans in ALK (N808, N863, N864 and N986) are indicated. Helices are denoted by α. d, Topology diagram of the TG supradomain, showing β-strands (green), α-helices (blue), pGII helices (orange), central pGII helices d, k and l (vermillion). e, Cross-sections and packing of pGII helix lattices in ALK (left) and LTK (right). f, g, Hydrophobic interfaces between the TNFL and GR subdomains in the TG supradomain of ALK (f) and LTK (g).

Structure of ALK-LTK cytokine-binding domains

To characterize the cytokine-binding segments of ALK and LTK, we determined their crystal structures based on purified glycan-shaved human ALK_TG–EGFL (residues 648–1038), its complex with a non-neutralizing Fab fragment¹⁷, and human LTK_TG (residues 63–378) (Extended Data Fig. 2a–c, Extended Data Table 1).

Unexpectedly, the TNFL and GR regions in ALK and LTK do not form separate domains, but rather a fully globular TNFL–GR (TG) supradomain featuring a chimeric arrangement of two interwoven subdomains with distinct secondary structure (Fig. 1b–d, Extended Data Fig. 2d–f). The TNFL subdomain is an anti-parallel β-sandwich, and the GR subdomain contains 14 long polyglycine type II (pGII) helices¹⁸ tightly packed into a honeycomb-like lattice (Fig. 1e, Extended Data Fig. 2g, h). The boundary between TNFL and GR comprises an extensive hydrophobic interface lined by conserved residues (Fig. 1f, g). The N terminus of the TG supradomain maps to the first strand of its TNFL subdomain and the C terminus is on the adjacent strand (Fig. 1d), which in ALK_TG–EGFL connects to the membrane-proximal EGFL module (Cys987 to Prol025) through a short N-glycosylated linker (Fig. 1b). Additionally, the TG supradomain is decorated by four peripheral α-helices, with α1 tethered to the TNFL subdomain through a conserved disulfide, whereas the disulfide-linked α2 and α4 cluster at the tip of the GR subdomain together with α3 (Fig. 1b, c).

Notably, the GR core has three pGII helices (d, k and l) exclusively containing glycine residues, which are surrounded by six other pGII helices, establishing tight van der Waals contacts and hydrogen bonding via their main-chain amide and carbonyl groups (Fig. 1e, Extended Data Fig. 2g). Collectively, the observed interactions within the pGII helix lattices offer a rationale for the sequence conservation of the polyglycine sequences (Extended Data Fig. 2i) and for loss-of-function mutants in this region of Drosophila ALK¹⁹.

The TNFL subdomains adopt a markedly different chain topology compared with TNF/C1q-class folds, despite their spatial similarity (root mean square deviation (r.m.s.d.) = 2.8 Å against TNF/C1q, 72 Cα atoms) and the sequence homology of individual β-strands (Extended Data Fig. 2j, k). However, the shuffling of spatially equivalent β-strands between the ALK_TNFL/LTK_TNFL subdomain and TNF/C1q proteins goes far beyond a simple permutation and cannot be explained by a simple evolutionary path. The connected β-strands in the ALK_TNFL/LTK_TNFL subdomain break up the alternating sheet-to-sheet register of the TNF/C1q β-jellyroll, and instead permit the spatially contiguous sprouting of the three glycine-rich loop inserts towards the distinctive pGII helix lattice of the ALK_GR/LTK_GR subdomains (Extended Data Fig. 2j). The spatial coalescence of three otherwise intrinsically disordered stretches of glycine-rich insertions into a topologically tortured TNFL subdomain suggests that this subdomain has potential as a versatile scaffold for protein design²⁰.

Assembly of ALK– and LTK–cytokine complexes

To cover the breadth of possible ALK– and LTK–cytokine complexes, we investigated structures of ALK–ALKAL2 and LTK–ALKAL1 assemblies. We first established that the conserved C-terminal domains of the cytokines (ALKAL1 and ALKAL2) and full-length ALKAL1 (ALKAL1_FL) could drive ALK-dependent cellular proliferation (Extended Data Fig. 3a–c). Notably, purified complexes of ALK and LTK with cytokines displayed distinct stoichiometries. LTK_TG–EGFL–ALKAL1 and LTK_TG–EGFL–ALKAL2 formed 2:1 stoichiometric assemblies (2 receptors and 1 cytokine), whereas ALK_TG–EGFL–ALKAL1 and ALK_TG–EGFL–ALKAL2 exhibited 1:1 stoichiometries (Extended Data Fig. 4a–c). The same stoichiometric dissonance was observed for complexes of ALK_TG and LTK_TG lacking the membrane-proximal EGFL domain (Extended Data Fig. 4d, e). Thus, ALK/LTK–cytokine complexes deviated from the canonical 2:2 stoichiometry of RTK–cytokine complexes.

Diffraction-quality crystals could be obtained only from purified ALK_TG–ALKAL2 and LTK_TG–ALKAL1 stabilized by a non-neutralizing single-domain camelid antibody fragment (nanobody Nb3.16) (Extended Data Fig. 4f, g, Extended Data Table 1). Our structures revealed highly similar assemblies in which a single cytokine molecule is cradled proximal to the membrane by two receptor TG supradomains, resulting in receptor dimerization (Fig. 2a, b, Extended Data Fig. 4h). In hindsight, the inordinately high protein concentration reached during crystallization might have compensated for the low degree of cytokine-induced ALK_TG dimerization in solution. ALK– and LTK–cytokine complexes feature three compact interaction interfaces. Sites 1 and 2 correspond to receptor–cytokine interactions and site 3 describes receptor–receptor contacts (Fig. 2c–f). The cytokine ligands are asymmetric three-helix bundles and use helices B and C to contact one receptor molecule via site 1, and helix A to engage the second receptor via site 2. The dimerized ALK_TG and LTK_TG supradomains incline their GR subdomains at approximately 45° to establish site 3. Thus, ALK family receptors establish two-fold-symmetric assemblies mediated by asymmetric cytokines, which we envisage become fully encapsulated at the cell surface, given the expected position of EGFL (Fig. 1b, Fig. 2a, b).

Fig. 2 |. Cytokine-mediated dimerization of ALK and LTK.

a, b, Structures of the ternary assembly of ALK_TG with ALKAL2 (a) and LTK_TG with ALKAL1 (b). The putative positions of the EGF-like domains are outlined. c, d, Interaction footprints (blue) for site 1 (755 Ų) and site 2 (630 Ų) in the ALK_TG–ALKAL2 complex (c) and interaction footprints (green) for site 1 (740 Ų) and site 2 (580 Ų) in the LTK_TG–ALKAL1 complex (d). e, f, Sideviews with interaction footprints for site 3 (red) in ALK_TG–ALKAL2 (760 Ų) (e) and LTK_TG-ALKAL1 (880 Ų) (f). The front receptor footprint is represented by a dotted outline.

Structure of ALKAL1 and ALKAL2

ALKAL1 and ALKAL2 adopt highly similar structures (r.m.s.d. = 0.54 Å, 56 Cα atoms) featuring a novel disulfide-stabilized three-helix bundle. (Extended Data Fig. 5a). The ALKAL1 and ALKAL2 fold is conspicuously open and lacks the classically buried hydrophobic core of helical cytokines. Rather, helix A merely buries Ile77 into a hydrophobic pocket formed by Leu117, Tyr110 and the Cys104–Cys113 disulfide in the BC loop (Extended Data Fig. 5a).

Despite their engulfment by the dimerized receptors, ALKAL1 and ALKAL2 harbour solvent-exposed hydrophobic cavities lined by conserved residues along the internal BC face (Leu117, Phe94 and Tyr98) and the AB loop (Val84 and Phe86) and hydroxyl-containing residues around the outer rim (Extended Data Fig. 5b–d). The strong conservation of all core residues suggests the universality of this fold among vertebrate ALKAL1 and ALKAL2 proteins (Extended Data Fig. 5e). The few sequence differences map to distinct patches contributed by the end of helix C and the AB loop, and might dictate cytokine specificity (Extended Data Fig. 5f).

ALK– and LTK–cytokine interactions

In site 1, ALKAL1 and ALKAL2 use a hydrophobic patch surrounded by arginine residues on helix B and helix C to contact one receptor molecule via its TNFL subdomain (Fig. 3a, b, Extended Data Fig. 6a, Extended Data Table 2). Here, LTK_TNFL accommodates ALKAL1 via a conserved hydrophobic platform (Phe143, Leu361, Leu364 and Val366) (Fig. 3b), whereas ALK_TNFL uses a rather amphipathic patch to interact with ALKAL2 (Ser758, Thr967, Leu970 and Val972) (Fig. 3a, Extended Data Fig. 6b).

Fig. 3 |. ALK– and LTK–cytokine complexes harbour three distinct interaction sites.

a–f Site 1 interactions in ALK–ALKAL2 (a) and LTK–ALKAL1 (b) complexes. Site 2 interactions in ALK–ALKAL2 (c) and LTK–ALKAL1 (d) complexes. Site 3 interactions in the LTK–ALKAL1 (e) and ALK–ALKAL2 (f) complexes. Residues investigated by mutagenesis are shown in red. g, h, Proliferation of Ba/F3 cells expressing wild-type (WT) ALK with wild-type or R123E/R136E, F97E or H100A mutants of ALKAL2 (g), or with wild-type or R102E/R115E or F76E mutants of ALKAL1 (h). n = 3 biologically independent experiments; mean ± s.d.; two-way analysis of variance (ANOVA) with Tukey’s multiple comparison test compared with the wild type. *P < 0.05, **P < 0.01 and ***p< 0.001; Exact P-values are provided in Source Data, i, Sequence motif analysis of ALK_T750–S758 and LTK_H135–F143 derived from 248 and 224 vertebrate sequences.

In site 2, the short helix A in ALKAL1 and ALKAL2 and the tip of the BC hairpin engage the second receptor molecule (Fig. 3c, d). In the LTK_TG–ALKAL1 complex, Phe80 and Phe76 protrude from helix A in ALKAL1 into a conserved hydrophobic pocket on LTK (Tyr124, Phe143 and Val366). However, site 2 in the ALK_TG–ALKAL2 complex is substantially less hydrophobic because ALK_TG projects Ser758 instead of a phenylalanine (Extended Data Fig. 6c). Thus, although the locations of sites 1 and 2 on LTK and ALK are identical, these sites are less hydrophobic in ALK.

Cytokine-mediated dimerization of ALK_TG and LTK_TG results in receptor–receptor contacts via site 3 by the locking of α1 of the TNFL subdomain in one receptor with α2 and α3 at the distal end of the GR subdomain in the second receptor (Fig. 2e, f, Extended Data Fig. 6d). However, the site 3 interface in ALK is markedly less extensive (Fig. 3e, f, Extended Data Fig. 6d–f).

Site 1 drives ALK– and LTK–cytokine complexes

To obtain insights into the contribution of the three interaction interfaces, we interrogated polar interactions in site 1 and the hydrophobic pocket in site 2, and receptor–receptor contacts in site 3.

Following identification of ALKAL2 as the high-affinity cytokine for ALK and ALKAL1 for LTK, we inserted charge-reversal mutations of two conserved polar contacts in the cytokines at site 1 (ALKAL1(R102E), ALKAL1(R115E), ALKAL1(R102E/R115E), ALKAL2(R123E), ALKAL2(R136E) and ALKAL2(R123E/R136E)), which markedly reduced the affinity for both receptors (Extended Data Fig. 7a, b). The effect of site 2 mutations ALKAL1(F76E), ALKAL2(F97E) and ALKAL2(H100A) on binding to LTK was less pronounced, although still substantial (Extended Data Fig. 7c, d). By contrast, ALKAL2(F97E) bound to ALK in a similar manner to wild-type ALKAL2 (Extended Data Fig. 7e). Nevertheless, all mutants markedly impaired Ba/F3 cell proliferation via ALK (Fig. 3g, h, Extended Data Fig. 7f, g). Of note, site 2 mutants retained their ability to form 1:1 stoichiometric complexes with both ALK and LTK, whereas site 1 mutants did not (Extended Data Fig. 7h), indicating that site 1 engagement might drive ALK/LTK–cytokine encounter complexes.

Site 3 mutants in LTK (LTK(R241A) and LTK(R241A/N369G)) did not undergo ALKALl-mediated dimerization (Extended Data Fig. 7i). To probe site 3 in ALK, we expressed ALK(M751T) in Ba/F3 cells, hypothesizing that this mutation would introduce an N-glycosylation site at Asn749. We found that ALK(M751T) did not drive cytokine-dependent cellular proliferation, although it was expressed at similar levels to wild-type ALK (Extended Data Fig. 7j).

ALK appears less well-equipped to fully harness contributions from sites 2 and 3 for its dimerization. Most notably, it lacks equivalents of Phe143 and His135 in LTK (Fig. 3i, Extended Data Fig. 8a) and a critical insert in the e–f loop (Extended Data Fig. 6e, f). Nevertheless, the conservation of interfacing residues at the three interaction sites supports a common binding mode among all vertebrate ALK family orthologues (Extended Data Figs. 5a, 8a, b), consistent with reported species cross-reactivity of zebrafish ALKAL2a and human ALKALs²¹. Finally, amino acids on the opposite side of the ALK cytokine-binding interfaces are well conserved, in contrast to those in LTK, suggesting their relevance for interactions with the N-terminal domains, which are absent in LTK (Extended Data Fig. 8c).

Mapping of somatic mutations in ALK

To broaden structure–function insights and to contextualize possible disease-relevant ALK and LTK variants, we structurally mapped mutations in ALK_TG and LTK_TG, combining documented oncogenic potential and frequently occurring missense single-nucleotide polymorphisms (Extended Data Fig. 9a, b). Mutations leading to constitutive receptor activation, enhancement of receptor–receptor contacts or stabilization of active receptor states are widely studied to evaluate oncogenic potential. LTK variant R243Q and ALK variants G685R, G747R and H694R are localized at positions that would be compatible with such roles. Notably, H694R is a gain-of-function mutation in lung adenocarcinoma leading to constitutive activation of ALK²². Mutations that might affect cytokine binding–for example, by increasing affinity–are less common. Here we identify two such candidate mutations: LTK variant P363L and ALK variant S737L, which map to interaction sites 2 and 1, respectively.

ALK(F856S), a gain-of-function mutation linked to acute myeloid leukemia²³, and ALK(R753Q) identified in histiocytic neoplasms²⁴, are roughly equidistant from the start of helix C of bound ALKAL2 (Extended Data Fig. 9a). Expression of ALK(F856S) and ALK(R753Q) in Ba/F3 cells resulted in markedly increased cytokine-dependent cell proliferation (Extended Data Fig. 9c–f). In contrast to an earlier study²³, we found that ALK^F856S was not constitutively active. Thus, these mutations might facilitate reorganization of key regions of the cytokine–receptor interfaces.

EGFL dictates ALK cytokine specificity

Given the structural similarity of ALK_TC– and LTK_TG–cytokine assemblies and the ligand-binding capacities of EGFL modules²⁵, we hypothesized that the membrane-proximal EGFL domain might underlie the preference of ALK for ALKAL2 over ALKAL1. Benchmarking of binding thermodynamics showed that even at micromolar concentrations, ALK_TG–EGFL formed enthalpically driven binary complexes with either cytokine characterized by a markedly higher affinity for ALKAL2 (dissociation constant (K_D) = 40 nM) than for ALKAL1 (K_D = 600 nM) (Fig. 4a, b). By contrast, LTK_TG–EGFL bound with high affinity and 2:1 stoichiometry to both cytokines (K_D,ALKAL1 = 10 nM, K_D,ALKAL2 = 55 nM) (Extended Data Fig. 10a, b) in agreement with stoichiometries derived from small-angle X-ray scattering (Extended Data Fig. 10c). The affinity of ALKAL1_FL for LTK_TG–EGEL was only modestly higher than that of the truncated form (Extended Data Fig. 10d), indicating that the conserved C-terminal part of ALKAL1 determines receptor engagement.

Fig. 4 |. The EGF-like domain dictates cytokine specificity in ALK.

a–d, Isothermal titration calorimetry (ITC) thermograms for the titration of ALK_TC–EGFL (10 μM) with ALKAL2 (100 μM) (a), ALKAL1 (25 μM) with ALK_TC–ECFL (210 μM) (b), ALKAL2 (33 μM) with ALK_TC (330 μM) (c), ALKAL1 (33 μM) with ALK_TG (330 μM) (d). Data are mean ± s.d. from three measurements. DP, differential electrical power, e, Proposed assembly mechanism of ligand-mediated complexes in the ALK receptor family.

Sequence differences in ALKAL1 and ALKAL2 map to distinct patches proximal to EGFL (Extended Data Figs. 5b, 10e). Notably, removal of EGFL from ALK reduced the receptor’s affinity for ALKAL2 by 30-fold, but reduced the affinity for ALKAL1 by only 4-fold (Fig. 4d, c); by contrast, the LTK_TG domain retained its high affinity to both ALKAL1 and ALKAL2 (Extended Data Fig. 10f–h). Thus, the EGFL domain of ALK drives cytokine specificity and might orchestrate the orientation of the membrane-proximal domains and their connected transmembrane helices to fine tune signal transduction.

Finally, we measured the binding affinity of full-length ALK ectodomain (ALK_FL) for ALKAL2. We obtained a similar affinity and 1:1 stoichiometry to the ALK_TG–EGFL–ALKAL2 complex, indicating that the four N-terminal domains of ALK are not important for cytokine binding (Extended Data Fig. 10i). Nevertheless, consistent with the interaction of canine ALK with heparin¹⁶, heparin induced dimerization of a large fraction of human ALK_FL in the presence or absence of bound cytokine (Extended Data Fig. 10j).

Mechanistic considerations

Despite the substantial structural similarity of ALK– and LTK–cytokine complexes, their assembly mechanism appears to be distinct and critically dependent on initial receptor–cytokine interactions via site 1 and the extent of receptor–receptor contacts. LTK–cytokine complexes are fully cytokine-driven, whereas ALK–cytokine assemblies might require the synergy of glycosaminoglycan binding⁴ and the reduced dimensionality of membrane-proximal engagement (Fig. 4e). The currently known modes of ligand-induced activation of RTKs range from receptor dimerization exclusively driven by the ligand (for example, TRKA–NGF)²⁶ to complexes mediated fully via receptor contacts (for example, EGFR–EGF)²⁷. However, several cytokine–RTK assemblies feature a combination of ligand- and receptor-mediated contributions (for example, CSF-1–CSF-1R)²⁸, including the involvement of accessory molecules (for example, FGF–FGFR)²⁹ or co-receptors, and the use of multiple cytokine copies (for example, insulin receptor)³⁰. The use of a single monomeric cytokine by ALK family receptors to dimerize with two-fold receptor symmetry introduces a novel cytokine-driven assembly mechanism among RTKs (Extended Data Fig. 10k). Our findings are poised to facilitate interrogation of ALK and LTK signalling in physiology and disease, and in the therapeutic targeting of ALK and LTK ectodomains¹⁴ and their cognate cytokines¹³.

Online content

Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-021-03959-5.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. All cell lines used have been tested for mycoplasma contamination.

Uncropped images of protein gel electrophoresis and western blot analyses reported in Extended Data Figs. 2, 4, 7, 9 are provided in Supplementary Fig. 1.

Protein expression constructs for recombinant protein production in mammalian cells

Sequence-optimized DNA for full-length wild-type ALK (Uniprot ID Q9UM73), LTK (Uniprot ID P29376), ALKAL1 (Uniprot ID Q6UXT8) and ALKAL2 (Uniprot ID Q6UX46) were purchased from Genscript. DNA encoding different human ALK constructs comprising amino acids 19–1038 (ALK_FL), 648–1038 (ALK_TG–EGFL) and 648–985 (ALK_TG), and human LTK constructs comprising amino acids 63–420 (LTK_TG–EGFL) and 63–378 (LTK_TG) were cloned in the pHLsec vector³¹ in frame with a N-terminal chicken RTPμ-like signal peptide sequence and a C-terminal caspase-3-cleavabIe Fc–His₆ tag at the C-terminus.

Sequences encoding ALKAL1_FL (residues 28–129), ALKAL1 (residues 57–129) and ALKAL2 (residues 78–152) were cloned in the pHLsec vector in frame with a N-terminal chicken RTPμ-like signal peptide sequence followed by a caspase-3-cleavable Sumostar-tag at the C-terminus. Sequence-optimized DNA encoding the light and heavy chains of Fab324 were purchased from IDT as GBlocks. The N-terminal signal peptide sequences were exchanged for a chicken RTPμ-like signal peptide sequence. The heavy chain was cloned in frame with a C-terminal caspase-3 site followed by an Avi–His₆ tag, while the light chain was cloned without a purification tag.

Protein expression in HEK 293 cells and purification from conditioned medium

Production of all ALK_TG–EGFL and ALK_TG constructs was performed in adherently grown HEK 293 MGAT1^−/− cells³² (obtained from P. Reeves, University of Essex, UK) maintained in DMEM supplemented with 10% FCS. Cells with a confluence of 80% were transiently transfected using branched polyethylenimine 25 kDa as transfection reagent in DMEM with 3.6 mM valproic acid without FCS.

Expression of the Fab fragment was achieved in adherent HEK 293T cells (obtained from N. Callewaert, VIB-Medical Biotechnology Center) using the same method. For the heterodimeric Fab fragment, plasmids encoding each chain were co-transfected in a 1:1 ratio.

Protein production for ALK_FL, ALKAL1, ALKAL2 and LTK constructs was performed in HEK 293S (obtained from N. Callewaert) cells grown in suspension and maintained in a mixture of 50% Freestyle (Thermofisher), 50% Ex-Cell (Sigma-Aldrich) growth medium. Transient transfection was performed with linear polyethylenimine 25 kDa (Polysciences) as transfection reagent. One day after transfection valproic acid was added³³ to a final concentration of 1.5 mM.

For expression in suspension cells, conditioned medium was collected after 5 days and subsequently clarified by centrifugation for 12 min at 8,000g while medium from adherently grown cells was collected after 6 days and centrifuged for 15 min at 6,000g. After centrifugation media were filtered through a 0.22-μm filter prior to chromatographic purification steps.

ALK and LTK proteins were captured via their Fc tag on a protein A column (HiTrap Protein AHP, Cytiva) followed by on-column digestion with caspase-3 for 1 h at 37 °C and an additional 2-h incubation at room temperature, and were finally eluted with HBS (20mM HEPES, pH 7.4, 150 mM NaCl). The tag-free proteins were concentrated and injected onto a HiLoad 16/600 SD200 (Cytiva) size-exclusion chromatography column pre-equilibrated with HBS. Purified proteins were stored at −80 °C until further use.

ALKAL-containing medium was diluted fourfold with 20mM HEPES pH 7.4 before loading on a cation-exchange column packed with SP Sepharose Fast Flow resin (Cytiva) equilibrated in 20 mM HEPES, pH7.4, 50 mM NaCl. ALKALs were eluted using a NaCl gradient from 50 mM–750 mM for 20 min. Fractions containing ALKALs were immediately diluted with 20 mM HEPES pH7.4 to a NaCl concentration of 200 mM and further supplemented with 0.1% (w/v) CHAPS. The C-terminal Sumo-tag was cleaved with caspase-3 overnight at 20 °C. To remove undigested protein as well as the cleaved Sumo tag, the digestion mixture was loaded on a MonoQ 5/50 GL column (Cytiva). Flowthrough containing the ALKALs was concentrated and injected on a Superdex 75 Increase 10/300 GL equilibrated in HBS supplemented with 0.1% (w/v) CHAPS. Purified ALKALs were stored at −80 °C at a concentration of 1 mg ml⁻¹ until further use.

For ALKALs used in BaF/3 assays, endotoxin levels were measured using a Endosafe PTS limulus amoebocyte lysate assay (Charles River) and were below 5 EU mg⁻¹.

Fab fragments were captured using complete His-tag purification resin (Roche) and eluted using HBS supplemented with 250 mM imidazole followed by buffer exchange to HBS on a HIPrep 26/10 desalting column. Caspase-3 was added to the purified Fab fragment to remove the Avi-His6 tag of the heavy chain by overnight digestion at 20 °C. The sample was loaded on an immobilized metal affinity column (IMAC) to remove the enzyme and undigested protein. The flow-through containing tagless Fab fragments was concentrated and injected on a Superdex 200 Increase 10/300 GL (Cytiva) column pre-equilibrated in HBS.

Production of a non-neutralizing single domain camelid nanobody against LTK

Single domain camelid nanobodies against LTK were raised by immunizing llamas with LTK_TG–EGFL and were selected for specific binding to LTK_TG–EGFL via ELISA and BLI. Epitope binning via biolayer interferometry (BLI) led to the identification of candidate nanobodies with non-neutralizing behaviour with respect to cytokine binding.

The sequence of non-neutralizing nanobody Nb 3.16 was cloned in a MoClo-derived yeast expression vector in frame with an N-terminal preproMF secretory leader sequence followed by the N-terminal His₆ tag and a caspase cleavage site. Komagataella phaffii cells were transformed by electroporation and grown on yeast extract peptone dextrose agar with sorbitol, containing 500 μg ml⁻¹ zeocin. One colony was picked to inoculate 500 ml of buffered glycerol-complex medium supplemented with 100 μg ml⁻¹ zeocin and grown at 28 °C for 24 h. Next, cells were pelleted by centrifugation at 500g for 7 min and resuspended in 500 ml buffered methanol-complex medium without zeocin and incubated overnight at 28 °C. Expression was further induced by adding 2.5 ml of 50% methanol, followed by addition of the same volume of methanol after 8 h and 24 h. The cells were subsequently incubated for another 8 h at 28 °C and conditioned medium was collected by centrifugation for 10 min at 6,000g.

His-tagged camelid single-domain nanobodies were captured by addition of 2 ml cOmplete resin (Roche) to 500 ml conditioned medium followed by overnight incubation at 4 °C while shaking. Nanobodies were eluted in HBS supplemented with 250 mM imidazole and buffer exchanged to HBS on a HiPrep 26/10 desalting column (Cytiva). The N-terminal His₆ tag was removed by an overnight digestion with caspase-3 at 20 °C. Undigested protein and the enzyme were removed by IMAC. The flowthrough containing tagless nanobody was concentrated and injected on an SD75 Increase 10/300 GL column (Cytiva) pre-equilibrated in HBS. Purified nanobody was concentrated to a concentration of 4 mg ml⁻¹ and flash-frozen and stored at −80 °C.

Crystallization and crystal structure determination

For ALK_TG–EGFL–Fab324 crystals, a 1.5 molar excess of Fab324 was added to ALK_TG–EGFL and N-linked glycans were digested overnight with EndoH. The complex was purified by size-exclusion chromatography (SEC) on a Superdex 200 Increase 10/300 GL (Cytiva) and concentrated to 13.5 mg ml⁻¹. Commercial sparse-matrix crystallization screens were set up using a Mosquito liquid handling robot (TTP Labtech) in sitting-drop format using 100 nl protein mixed with 100 nl mother liquor in SwissSci 96-well triple drop plates incubated at 287 K. A first hit in the Morpheus II screen³⁴ was further optimized to a condition consisting of (40 mM polyamines, 0.1 M Gly–Gly/2-amino-2-methyl-1,3-propanediol pH 8.5, 11% w/v PEG4000, 19 % w/v 1,2,6-hexanetrioI) in sitting-drop format with a 100 nl protein mixed with 200 nl mother liquor geometry. Crystals were cryoprotected in mother liquor containing 25% w/v 1,2,6-hexanetriol prior to being cryocooled in liquid nitrogen. Diffraction data was collected at 100 K at the ID23-2 beam line at ESRF, Grenoble. The datasets were processed using XDS³⁵. Initial phases were calculated by molecular replacement with PHASER³⁶ using the coordinates of a Fab fragment exhibiting the highest sequence identity (PDB: 5NUZ, chain A) followed by rigid body refinement in Buster³⁷. A partial polyalanine model was built into the visible electron density in Coot³⁸ followed by density modification via Resolve³⁹. The density-modified map showed density for several aromatic side chains allowing for tracing of the correct ALK sequence. Additional refinement steps were carried out in PHENIX⁴⁰ using individual B-factor refinement in combination with TLS, XYZ refinement, optimizing the X-ray/geometry weights as well as local torsion angle non-crystallographic symmetry restraints.

For crystals of ALK_TG–EGFL, glycans were trimmed by overnight enzymatic digestion with EndoH in HBS. The complex was polished via SEC and concentrated to 10.5 mg ml⁻¹. Commercial sparse-matrix crystallization screens were set up as described above. One hit was obtained in the Morpheus II screen and optimized to 0.5 mM manganese (II) chloride tetrahydrate, 0.5 mM cobalt (II) chloride hexahydrate, 0.5 mM nickel (II) chloride hexahydrate, 0.5 mM zinc acetate dihydrate, 13% w/v PEG 3000, 28% v/v 1,2,4-butanetriol, 1% w/v non-detergent sulfobetaine (NDSB) 256 in hanging-drop format. Crystals were cryoprotected in mother liquor containing 25% 1,2,4-butanetriol prior to being flash frozen in liquid nitrogen. Diffraction data were collected at 100 K at the P14 microfocus beam line at PETRA III, Hamburg and integrated using XDS with standard parameters except for the BEAM_DIVERGENCE parameter which was doubled. Initial phases were obtained using maximum likelihood molecular replacement in Phaser using the structure of the ALK_TG domain. The structure was refined using Phenix.refine followed by manual building in COOT. The EGF-like domain was manually built into the electron density. Refinement in phenix followed the same protocol as for the ALK_TG–EGFL–Fab324 structure except for the absence of non-crystallographic symmetry restraints and implementation of additional reference restraints based on the structure of ALK_TG in complex with Fab324.

For crystals of LTK_TG, purified LTK_TG was concentrated to 10 mg ml⁻¹ and used to set up sparse matrix screens as previously described. Crystals appeared in a condition of the Morpheus II screen with the final optimized condition (MOPSO/Bis-Tris pH 6.3, 12% PEG 20000, 26% trimethyl propane 1% w/v NDSB 195 5 mM yttrium (III) chloride hexahydrate, 5 mM erbium (III) chloride hexahydrate, 5 mM terbium (III) chloride hexahydrate, 5 mM ytterbium (III) 1 chloride hexahydrate) set up in sitting-drop format with a 100 nl protein mixed with 200 nl mother liquor geometry. For data collection, single crystals were cryoprotected in mother liquor and cryo-cooled in liquid nitrogen. Diffraction data were collected at 100 K at the P13 microfocus beam line at PETRAIII, Hamburg and processed using XDS as previously described. Phases were obtained by single-wavelength anomalous dispersion making use of the anomalous signal from lanthanide atoms. Determination of the lanthanide substructure for four sites was performed by the hybrid substructure search as implemented in Phenix. Phases were calculated using Phaser-EP. The experimental electron density was readily interpretable, and a model was manually built in Coot and further refined in Phenix implementing an anisotropic individual B-factor model.

For LTK_TG–ALKAL1–Nb3.16 crystals, a threefold molar excess of Nb3.16 was added to the LTK_TG–ALKAL1 complex, concentrated and injected into a Superdex 200 Increase 10/300 GL (Cytiva) equilibrated in HBS. Eluted fractions were concentrated to 13.5 mg ml⁻¹ and used to set up sparse-matrix crystallization screens in sitting-drop format as described above. Crystals were obtained in a condition consisting of MOPSO/Bis-Tris pH 6.3, 13% PEG 8000, 22% w/v 1,5-pentanediol, 5 mM sodium chromate tetrahydrate, 5 mM sodium molybdate tetrahydrate, 5 mM sodium tungstate tetrahydrate, 5 mM sodium orthovanadate tetrahydrate. Crystals were cryoprotected in mother liquor containing 25% 1,2,4-butanetriol and were cryo-cooled in liquid nitrogen. Diffraction experiments were performed at 100 K Proxima 2 (Soleil Synchrotron). Initial phases were obtained by maximum-likelihood molecular replacement using Phaser with the previously obtained LTK_TG structure as a search model. A model for Nb3.16 was automatically built using Arp Warp^{41, 42} followed by manual building of ALKAL1 in Coot. Refinement was performed in Phenix with individual anisotropic ADP parameters with a TLS model.

For ALK_TG–ALKAL2 crystals a threefold molar excess of ALKAL2 was added to ALK_TG and subjected to an overnight EndoH digest, concentrated and injected into a Superdex 200 Increase 10/300 GL (Cytiva) equilibrated in HBS. Eluted fractions were concentrated to 8 mg ml⁻¹ and used to set up sparse-matrix crystallization screens in sitting-drop format as described above. Initial hits were obtained in a condition consisting of 0.1 M MES pH 6.5, 15% w/v PEG 6000, 5% w/v MPD. A single crystal was used to prepare a seed stock⁴³ using the PTFE seed bead (Hampton Research). The best diffracting crystals were obtained by seeding with a 1:1,000 dilution of the seed stock into the optimization screen. Crystals were cryoprotected in 78% (v/v) mother liquor supplemented with 22% (v/v) ethylene glycol before flash freezing in liquid nitrogen. Diffraction data were collected at 100 K at the Proxima 2 microfocus beam line at the Soleil synchrotron. Data as processed as described above with the difference that anisotropy correction was implemented by the UCLA diffraction anisotropy server⁴⁴. Initial phases were obtained by molecular replacement in Phaser using the previously determined ALK_TG and ALKAL1 structures. Initial refinement cycles were performed in Buster followed by iterative refinement using Coot and Phenix. B-factors were refined using two isotropic atomic displacement parameters complemented by TLS. During refinement structures of ALK_TG and ALKAL1 provided reference model restraints. For Fab324 crystals, tag-free Fab324 was concentrated to 18.5 mg ml⁻¹ and a pH versus (NH₄)₂SO₄ concentration screen was set up in sitting drop format resulting in crystals in a condition consisting of (1.5 M (NH₄)₂SO₄,40 mM glycine pH 9.5). The crystal was cryoprotected using a saturated (NH₄)₂SO₄ solution. Diffraction experiments were performed at the P13 beamline at PetraIII, Hamburg and data were processed using XDS. Phases were obtained by molecular replacement using our structure of Fab324 obtained from the ALK_TG–EGFL–Fab324 complex. The structure was refined using Coot and Phenix.

All display items containing structures were generated using the PyMOL Molecular Graphics System, version 2.0.5 (Schrödinger).

Isothermal titration calorimetry

Experiments were performed using a MicroCal PEAQ-ITC instrument at 310 K. Proteins used in ITC experiments were expressed in HEK 293S cells grown in suspension. As a final purification step all proteins were buffer exchanged to the same HBS buffer via size-exclusion chromatography. Titrations were preceded by an initial injection of 0.5 μl. The injection spacing was optimized per experiment to allow for the signal to get back to a stable baseline. Throughout the titration the sample was stirred at a speed of 750 rpm. Data were analysed using PEAQ-ITC analysis software (version 1.1.0.1262, Malvern) and fit using a ‘one set of sites’ model.

Multi-angle laser light scattering

Protein samples of 100 μl at approximately 1.0 mg ml⁻¹ were injected onto a Superdex 200 increase 10/300 GL column (Cytiva) connected in line to a UV-detector (Shimadzu), a miniDawnTReos (Wyatt) Multi-angle laser light scattering (MALLS) detector and an Optilab T-Rex (Wyatt) refractometer. The refractive increment value (dn/dc) was 0.185 ml g⁻¹. Band broadening was corrected for using reference measurements of BSA (Pierce). Data analysis was carried out using the Astra 6.1.6 software and standard deviations were calculated using Prism8 (Graphpad Software).

Biolayer interferometry

Benchmarking of binding kinetics and screening of mutant ALKALs were performed by immobilizing wild-type and mutant ALKAL1/2 variants. To this end, wild-type ALKAL1 (residues 57–129) and mutant variants (R102E, R115E, R102E/R115E, F76E and F80E) as well as wild-type ALKAL2 (residues 78–152) and interface mutants (R123E, R136E, R123E/R136E, F97E and H100A) were cloned in the pHLsec vector in frame with a C-terminal Avi tag. All constructs were transiently co-transfected in suspension-grown HEK 293S cells together with a BirA expression plasmid (pDisplayBirA-ER⁴⁵) as previously described and supplemented with 100 μM biotin upon transfection. After 4 days of expression, excess biotin was removed by desalting the conditioned media to HBS on a HIPrep 26/10 desalting column (Cytiva).

All measurements of binding kinetics and dissociation constants were performed using an Octet Red 96 (Forté Bio) in assay buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 0.02% (w/v) BSA, 0.002% (v/v) Tween 20) at 298 K. ALKALs were immobilized to a level of 0.5 nm on streptavidin-coated biosensors (Forté Bio). To verify that no nonspecific binding was present during the assay, non-functionalized biosensors were used as a control by measuring in parallel all ligand concentrations as well as running buffer. For all mutants a two fold dilution series from 6.4 μM–400 nM was employed. Data analysis was performed using Data Analysis software 9.0.0.14 (Forté Bio) and binding curves were exported to Prism8 (Graphpad Software) for plotting of curves.

Small angle X-ray scattering

SEC–small angle X-ray scattering (SAXS) data were collected at the SWING beamline at Soleil (France) using an integrated online HPLC set-up. Purified samples of ALK_TG–EGFL–ALKAL2 (18.5 mg ml⁻¹) and LTK_TG–EGFL–ALKAL1 (19 mg ml⁻¹) expressed in HEK 293S MGAT1^−/− cells were injected on a Biosec-3 column (Agilent) with HBS as a running buffer. The scattering data were collected in continuous flow mode with a flow speed of 0.3 ml min⁻¹ and a 1-s exposure time per frame. Buffer and sample frames were selected and subtracted using the program RAW⁴⁶. Theoretical scattering curves and fitting to experimental scattering data was performed with AllosMod-FoXS.

In brief, a model for the C-terminal EGFL domain of LTK was generated by homology modelling starting from the crystal structure of the ALK EGF-like domain using the SWISS-MODEL server⁴⁷. This model was manually placed and connected to the C terminus of the LTK_TG domain in Pymol based on the ALK_TG–EGFL structure. Missing regions in the ALK and LTK structures were added using MODELER⁴⁸. The resulting models were subsequently energy-minimized using Rosetta-Relax, and used as an input for AllosMod to add N-linked glycans at positions Asn709, Asn808, Asn886 and Asn986 for ALK and Asn380 and Asn412 for LTK, and the resulting model energy landscapes were calculated. The output of AllosMod was then used in AllosMOD-FoXS to calculate fits with theoretical scattering curves during fast AllosMod simulations at 300 K.

Cell culture and retroviral transduction

Ba/F3 (mouse pro-B cell line) cells were cultured in RPMI/10% FCS supplemented with 1 ng ml⁻¹ mouse interleukin-3 (IL-3). Ba/F3 cell line was not listed in the database of commonly misidentified cell lines maintained by ICLAC and NCBI Biosample. Ba/F3 cells were transduced with murine stem cell virus (MSCV) supernatant from MSCV expressing ALK(WT), ALK(R753Q), ALK(F856S) or EV-IRES-GFP (where EV is empty vector) for 2 days in RPMI, 10% FCS supplemented with mouse IL-3 as previously described⁴⁹. GFP-sorted cells were used for the cell growth assays and western blots. After removal of IL-3 from the medium, the cells were stringently washed with PBS for three times. The cell growth curves and heat maps were made using GraphPad Prism 9 software as mean values, with error bars representing standard deviation.

Reagents

For western blotting, the following antibodies were used: ALK (Purchased from Cell Signaling Technologies; catalogue no. 3633; dilution: 1:1,000), phospho-ALK (Tyr1278) (Cell Signaling Technologies; 6941; 1:250), phospho-ALK (Tyr1604) (Cell Signaling Technologies; 3341; 1:250), phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (Cell Signaling Technologies; 4370; 1:2,000), p44/42 MAPK (Erk1/2) (Cell Signaling Technologies; 9102; 1:1,000), β-actin (Sigma-Aldrich; A-5441; 1:2,000), GAPDH (GENETEX; GTX100118; 1:2000). For the cell growth assay, Crizotinib (Sigma-Aldrich; PZ0191-5MG), DMSO (Sigma-Aldrich; D8418-100ML), and IL-3 (Peprotech; AF-213-13) were used.

Statistics and reproducibility

Statistical significance was determined by two-way ANOVA followed by Tukey’s multiple comparison test where multiple comparisons should be adjusted. Data were plotted using GraphPad Prism 9 software as mean values, with error bars representing s.d. Heat maps were also made using GraphPad Prism 9 based on mean values. *P < 0.05, **P < 0.01 and ***P < 0.001, unless otherwise specified.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability

Coordinates and structure factors for the complexes have been deposited in the Protein Data Bank (PDB) under accessions 7NWZ (ALK_TG–ALKAL2), 7NX0 (LTK_TG–ALKAL1–Nb3.16), 7NX1 (LTK_TG), 7NX2 (unbound Fab324), 7NX3 (ALK_TG–EGFL–Fab324), 7NX4 (ALK_TG–EGFL). Most common single-nucleotide polymorphisms were obtained from the COSMIC database (https://cancer.sanger.ac.uk/cosmic) and gnomAD database (https://gnomad.broadinstitute.org). The Ba/F3 cell lines expressing ALK(WT), ALK(R753Q) or ALK(F856S) can be obtained from the authors upon request. Source data are provided with this paper.

Extended Data

Extended Data Fig. 1 |. Phylogenetic analysis and evolution of ALK and LTK.

a, Phylogenetic analysis of ALK (a). Multiple sequence alignment was performed with MAFFT 7 using the E-INS-I algorithm for refinement. The phylogenetic tree was visualized using Phyl.io tool as implemented in MAFFT. ALK is an evolutionarily ancient RTK in C. elegans and D. melanogaster, where it is activated by FIEN-1⁵⁰ and Jeb^{51, 52}. b, Sequence alignment of ALK sequences focusing on the heparin binding motif. Positively charged residues are highlighted in blue. The multiple sequence alignment was performed with MAFFT 7 using the E-INS-I algorithm for refinement. c, Phylogenetic analysis of LTK performed as for ALK. In mammals the LTK ectodomain consists of the TG and EGF-like domains while other vertebrates contain the additional MAM-LDLa-MAM domains at the N-terminus.

Extended Data Fig. 2 |. Purification of ALK_TG-EGEL and LTK_TG and structural details of the TNFL and GR subdomains of the novel TG supradomain fold.

a,b,c Representative chromatograms and SDS-PAGE gels for the purification of ALK_TG-ECFL (a), ALK_TG–GFL–Fab324 (b) and LTK_TG (c). The arrow indicates the shift in elution volume after EndoH digest of ALK_TG-ECFL. Each protein was purified several times, chromatograms and SDS-PAGE analysis of each sample are representative for different protein batches. Uncropped gels are included in source data, d, ALK_TG-ECFL structure colored according to secondary structure elements. α-Helices (blue), β-strands (green), pGII-helices (orange), loops (grey), e, Structure of the ALK_TG-ECFL–Fab324 complex with ALK coloured according to its secondary structure elements. CDR loops of Fab324 are coloured yellow. The constant domains of Fab324 are omitted for clarity. f, LTK_TG structure colored according to secondary structure elements. g, Hexagonal pGII-helix arrangement surrounding pGII-helix d in LTK. Vermillion pGII-helices consist exclusively of glycine residues. pGII-helix d shown as sticks, hydrogen bonds to other residues in LTK are indicated as dotted lines. The glycine-rich segment has complicated detection of a globular fold but has led to its sequence-based classification as Glycine-rich PFAM domain PF12810⁵³. h, Schematic representations of pGII-helix arrangements in reported structures. Drastically less extensive pGII-helix arrangements than the one displayed in the GR subdomain of ALK and LTK have been observed in synthetic polyglycines²⁰ and four functionally diverse proteins^54–57. Full circles indicate pGII-helices coming out of the plane of the page while empty circles indicate helices going into the plane of the page. S16 adhesin (pdb: 6F45) Ape complex (pdb: 5L9W) obg (pdb: 5M04) Sf antifreeze protein (pdb: 3BOI). i, Sequence alignment performed with Clustal Omega of human ALK, human LTK, C. elegans ALK (SCD-2) and D. melanogaster ALK covering pGII-helices j,k and l. Residues conserved across all four species are indicated with an orange background. Conserved hydrophobic residues involved in the hydrophobic groove between the TNF-like and glycine rich region are indicated by a green sphere, j, The β-sheet subregions of ALK and a trimmed view of TNF (pdb: 1TNF) are coloured in a N-(blue) to C-terminus (red) gradient and shown side by side after structural superposition. Topology diagram for the TNFL domain of ALK and the jelly-roll fold of TNF follow the same colour scheme. Jelly-roll fold nomenclature starts with strand B according to convention. For ALK_TNFL the nomenclature in black is according to the TG domain notation used in this study while the nomenclature according to the TNF convention (first β-strand labeled B) is shown in grey. Structural queries⁶³ using the TFNL subdomains retrieved TNF/C1q-class folds (r.m.s.d =2.8 Å against C1q and TNF, 72 Cα atoms). Topology-independent searches⁵⁸ covered an additional −20 residues in the canonical TNF fold, and structure-based sequence alignments clarified the sequence homology between the A, D and E β-strands in ALK/LTK_TNFL and β-strands B, E and F in TNF or TRAIL. The distinctly connected β-strands in the ALK_TNFL/LTK_TNFL subdomain break up the alternating sheet-to-sheet register of the TNF/C1q β-jellyroll, and instead permit the spatially contiguous sprouting of the three glycine-rich loop inserts (between β-strands D and E, F and G, and H and H’) towards the distinctive pGII-helix lattice of the ALK_GR/LTK_GR subdomain. The sequential B to 1 β-strands of the TNF/C1q β-jelly roll smoothly sew together the two β-sheets (that feature characteristic B’BIDG and FEHC faces) whereas the ALK/LTK_TNFL subdomain has AIDEH” and H’HGFCB faces (primed small caps denote additional, edge β-strands). k, Annotated alignment of selected β-strands of human ALK, LTK, TNF and TRAIL. Conserved hydrophobic residues are indicated in red.

Extended Data Fig. 3 |. ALKAL1- and ALKAL2-dependent proliferation of Ba/F3 cells expressing ALK.

a, Molecular mass determination of ALKAL2, ALKAL1 and ALKAL1_FL by size-exclusion chromatography and in-line multi-angle laser light scattering (SEC-MALLS). The differential refractive index (left vertical axis) is plotted against the determined molecular weight (right vertical axis). ALKAL2 (blue trace), ALKAL1 (green trace) and ALKAL1_FL (vermillion trace). b, Cell proliferation of ALK^WT or EV (empty vector) expressing Ba/F3 cells treated with 10 nM ALKAL2 with and without addition of crizotinib (n = 3 biologically independent experiments; mean ± s.d). c, Cell proliferation of Ba/F3 cells expressing ALK^WT or EV upon stimulation by a concentration series of ALKAL2, ALKAL1 or ALKAL1_FL at indicated concentrations. (n = 3 biologically independent experiments; mean ± s.d.; two-way ANOVA with Tukey’s multiple comparison test compared with EV; *P < 0.05, **P < 0.01 and ***P < 0.001; Exact P-values are provided in source data. The ratio of the observed ALKAL-induced cell growth with the IL-3 induced cell growth is shown in a heatmap representation for the measurements on day 4). Volume (ml)

Extended Data Fig. 4 |. Biophysical characterization and purification of ALKAL-mediated complexes of ALK and LTK.

a, Calculated molecular masses for the ALK, LTK and ALKAL constructs under study. b-e, Experimental molecular mass determination of ALKAL-mediated complexes with LTK_TG-EGFL (b), ALK_TG-EGFL (deglycosylated) (c), LTK_TG (d), ALK_TG (deglycosylated) (e) by SEC-MALLS. The differential refractive index (left vertical axis) is plotted against the experimentally measured molecular mass (right vertical axis). Unbound LTK and ALK are in orange and pink respectively. Complexes with ALKAL1 or ALKAL2 are in green and blue respectively. The reported molecular mass represents the average molecular mass ± s.d. across the elution peak. f,g, Representative SEC elution profiles and SDS-PAGE analysis for the purification of the ALK_TG-EGFL–ALKAL2 complex (f) and the LTK_TG-EGFL–ALKAL1–Nb3.16 complex (g). Each protein was purified several times, chromatograms and SDS-PAGE analysis of each sample are representative for different protein batches. Uncropped gels are included in source data, h, Surface representation of the LTK_TG-EGFL–ALKAL1–Nb3.16 complex illustrating the location of the binding sites of the non-neutralizing Nb3.16 to LTK_TG far from any cytokine–receptor or receptor–receptor binding site.

Extended Data Fig. 5 |. Structure of ALKAL1 and ALKAL2.

a, Structural superposition of the conserved C-terminal segments of ALKAL1 and ALKAL2 (coloured in an N- (blue) to C-terminus (red) gradient) as observed in their complexes with cognate receptors. ALKAL1 and ALKAL2 share high sequence identity (66%) in their C-terminal domain, but have variable N-terminal regions. ALKAL1 residues in the interface between helix A and helices B and C are shown as sticks. αA connects via a conserved short loop to a helical hairpin constructed from αB and αC, which in turn are tethered by two conserved disulfides. A conserved stretch of −10 residues preceding αA was not ordered in the reported structures, suggesting they might help to stabilize the soluble forms of these cytokines rather than contribute to direct receptor engagement or might help reduce the entropic cost of binding, b, Surface representation of ALKAL1 coloured according to the Eisenberg hydrophobicity scale. Hydroxyl groups surrounding the central cavity are shown as red spheres, c, Sequence alignment by ClustalOmega of human ALKAL1 with ALKAL2. Residues in the interface between helix A and helices B and C are coloured according to their position. Cysteines are coloured yellow and disulfide bonds are shown as yellow lines, d, Surface representation of the LTK_TG–ALKAL1 complex with LTK in ALKAL1 coloured according to the Eisenberg hydrophobicity scale (white is more hydrophobic). The black circle denotes the ALKAL1 hydrophobic cavity, e, Multiple sequence alignments of various vertebrate ALKAL1 and ALKAL2 sequences using the ESPripT server (http://espript.ibcp.fr/ESPript/ESPript/) and structural annotation according to secondary structure elements. Symbols indicate residues participating in interaction sites 1 and 2 according to the graphical legend. Hs (Homo sapiens), Mm (Mus musculus), Gg (Gallus gallus) Xl (Xenopus laevis), Xt (Xenopus tropicalis) Dr (Danio rerio). f, ALKAL1 shown in green transparent surface and ribbon representation. Residues differing with ALKAL2 are colored Vermillion and labeled with the ALKAL1/ALKAL2 numbering.

Extended Data Fig. 6 |. Structural details of receptor–cytokine and receptor–receptor interactions in ALK/LTK–cytokine complexes.

a, 2Fo-Fc electron density maps contoured at +1 r.m.s.d. showing details of site 1,2 and 3 of the LTK–ALKAL1 and ALK–ALKAL2 complexes, b, Transparent surface of ALKAL1 according to the Eisenberg hydrophobicity scale illustrating similarities and differences in site 1 of LTK/ALK–cytokine complexes. Shown is the central conserved hydrophobic patch formed by leucines (L97, L116 and L120) and the interacting residues of LTK (orange). The equivalent ALK residues (pink) are shown after alignment with LTK. c, View of Site 2 in the LTK–ALKAL1 and ALK–ALKAL2 complexes. ALKALs are coloured according to the Eisenberg hydrophobicity scale. Receptor residues surrounding the hydrophobic triad of helix A (L72, F76, F80 in ALKAL1 and M93, F97, L101 in ALKAL2) are shown as sticks for LTK (orange) and ALK (pink). d, Superposition of unbound ALK (dark gray), unbound LTK (light gray), bound ALK (pink, only helices shown) and bound LTK (orange, only helices shown), e, Superposition of the ALK–ALKAL2 and LTK–ALKAL1 complexes, zoomed in on the region around the e-f loop. f, View of the site 3 groove of ALK (top) and LTK (bottom). In LTK, site 3 centers on His153, which stacks against Gly74 and Arg241 on an LTK-specific loop with Asn369 residues hydrogen bonding across the twofold axis of the complex.

Extended Data Fig. 7 |. Functional interrogation of site 1, site 2, and site 3 interfaces in ALK/LTK–cytokine complexes.

a,b Representative response curves as measured by biolayer-interferometry (BLI) for the interaction of wild type ALKAL1 and ALKAL1 mutants (containing charge-reversal mutations of residues involved in site 1) (a) and WT ALKAL2 and ALKAL2 mutants (b) with ALK_TG–EGFL and LTK_TG–EGFL. For wild type ALKALs LTK curves were fitted with a 2:1 binding model (red) while for ALK a 1:1 model was used. Start and end concentrations of the 2-foid dilution series used for the WT measurements is shown as an inset while for all mutants a 2-fold dilution series from 6.4 μM-400nM was used, c, BLI response curves for the interaction of the site 2 ALKAL1^F76E mutant with LTK_TG–EGFL. d, BLI response curves for the interaction of the site 2 ALKAL2^F97E and ALKAL2^H100A with LTK_TG–EGFL. e, BLI response curves for the interaction of the site 2 ALKAL2^F97E with ALK_TG–EGFL. f, SDS-PAGE analysis of purified ALKAL1 and ALKAL2 mutants used in Ba/F3 and SEC-MALLS assays. Each protein was purified several times, SDS-PAGE analysis of each sample are representative for different protein batches. Uncropped gels are included in source data, g, Western blot analysis of phosphorylated ALK (Y1278 and Y1604) after stimulation with ALKAL2^WT, ALKAL2^R123E/R136E, ALKAL2^F97E and ALKAL2^H100A. Uncropped western blot scans are provided in source data. h, Capacity of ALKAL1 and ALKAL2 mutants to form complexes with LTK_TG-EGFL and ALK _TG-EGFL respectively as characterized by SEC-MALLS. Differential refractive index (left axis) is plotted against the determined molecular weight (right axis). LTK_TG-EGFL (orange trace), LTK_TG-EGFL–ALKAL1^R102E/R115E (green trace) and LTK_TG-EGFL–ALKAL1^F76E (blue trace). ALK_TG-EGFL (pink trace), ALK_TGEGFL–ALKAL2^R123E/R136E (green trace) and ALK_TG-EGFL-ALKAL2^F97E (blue trace). The ALKAL1 site 1 mutant is unable to form a complex with ALK while the site 2 mutant still forms a binary complex. The reported molecular mass represents the average molecular mass ± s.d. across the elution peak, i, Capacity of the LTK_TG-EGFL^R241A LTK_TG-EGFL^R241A/N369G site 3 mutants to form complexes with ALKAL1 as characterized by SEC-MALLS. LTK_TG-EGFL^R241A (red trace), LTK_TG-EGFL^R241A/N369G (cyan trace), LTK_TG-EGFL^R241A–ALKAL1 (green trace), LTK_TG-EGFL^R241A/N369G–ALKAL1 (blue trace). LTK_TG-EGFL (orange trace) and LTK_TG-EGFL–ALKAL1 (pink trace) are shown for comparison. The reported molecular mass represents the average molecular mass ± s.d. across the elution peak, j, Cell proliferation of Ba/F3 cells expressing ALK^WT or ALK^M7S1T upon stimulation with 50 nM ALKAL1 or 50 nM ALKAL2. Western blot analysis of ALK^WT or ALK^M7SIT expression is a representative of three biologically independent experiments with similar results. Uncropped western blot scans are provided in source data.

Extended Data Fig. 8 |. Conservation of TG supradomains in ALK and LTK.

a, Structurally annotated multiple sequence alignments of the TG and EGFL domains of ALK and LTK using the ESPripT server (http://espript.ibcp.fr/ESPript/ESPript/). Hs (Homo sapiens), Mm (Mus musculus), Gg (Gallus gallus) Xl (Xenopus laevis), Xt (Xenopus tropicalis) Dr (Danio rerio). Symbols indicate residues participating in the different interfaces according to the graphical legend. b, Bottom view of dimerized LTK (left) and ALK (right) in surface representation coloured according to residue conservation. The ALKAL binding sites are shown as an outline of ALKAL1 (green) and ALKAL2 (blue) are shown as dashed lines. Conservation analysis was performed using the Consurf server (https://consurf.tau.ac.il) based on an alignment of 248 vertebrate sequences for ALK and 225 for LTK by COBALT. c, Top view of dimerized LTK (left) and ALK (right) coloured according to sequence conservation levels.

Extended Data Fig. 9 |. Mapping of missense mutations on the structures of the ALK–ALKAL2 and LTK–ALKAL1 complexes.

a, Mapping of most frequent SNPs (GnomAD) to the ALKAL2–ALK_TG complex shown in top view. SNPs also found in COSMIC database (https://cancer.sanger.ac.uk/cosmic) are also indicated on the bottom view. Mutations further characterized in this study are coloured red. Inset shows the detailed position of R753 in the ALK_TG-ALKAL2 complex. b, Mapping of most frequent SNPs (GnomAD) to the ALKAL1–LTK_TG complex shown in top view. SNPs also found in COSMIC database (https://cancer.sanger.ac.uk/cosmic) are also indicated on the bottom view, c, Western blot analysis of the expression levels of ALK^WT and the ALK^R753Qand ALK^F856S mutants and their ERK phosphorylation (left). On the right-side western blot analysis of phosphorylated ALK^WT, ALK^R753Qand ALK^F856S is shown. Representative results from three biologically independent experiments with similar results. Uncropped western blot scans are provided in source data, d, Sanger sequencing of cDNA showing WT or mutant ALK expression in isogenic Ba/F3 cells. e, Cell proliferation of ALK expressing Ba/F3 cells treated with 10 nM ALKAL2 with and without addition of crizotinib for ALK R753Q and F856S mutants. Data for EV and WT ALK are repeated from Extended Data Fig. 3 for direct comparison. Crizotinib is also able to inhibit ALKAL2 induced proliferation for mutant ALK, indicating ALK dependent signalling. (n = 3 biologically independent experiments; mean ± s.d.; two-way ANOVA with Tukey’s multiple comparison test compared with DMSO control). f, Proliferation of Ba/F3 cells expressing ALK carrying the R7S3Q or F856S mutation upon stimulation by a concentration series of ALKAL2, ALKAL1 or ALKAL1_FL at indicated concentration (n=3 biologically independent experiments; mean ± s.d.; two-way ANOVA with Tukey’s multiple comparison test compared with ALK^WT. ALKAL-induced cell growth relative to that of cells cultured with IL-3 is shown in a heatmap representation. E V and ALK^WT controls are included for comparison. *P < 0.05, **P < 0.01 and ***P <0.001. Exact P-values are provided in source data).

Extended Data Fig. 10 |. Mechanistic insights into the assembly of ALK/LTK–cytokine complexes derived from microcalorimetry, SAXS, and SEC-MALLS.

a,b ITC experiments for the titration of LTK_TG-EGFL (5 μM) with ALKAL1 (56 μM). (a) Titration of LTK_TG-EGEL (12 μM) with ALKAL2 (56 μM). (b). ITC titration curves, the left panel shows the raw data with the differential electrical power (DP) plotted against time. The right panel represents the binding isotherm obtained from the integration of the raw data and fitted to a one-site model. mean ± s.d were calculated based on 3 measurements, c, Small-angle X-ray scattering analysis and calculated FoXS fits of the binary ALK_TG-EGFL–ALKAL2 (pink), ternary ALK_TG-EGFL–ALKAL2 (purple), binary LTK_TG-EGFE–ALKAL1 (orange) and ternary LTK_TG-EGEL–ALKAL1 (light orange) to experimental SAXS data (black curves), d, ITC experiments for the titration of LTK_TG-EGFL (7.2 μM) with ALKAL1_FE (55 μM) Data representative of 2 independent experiments, e, Ternary ALK_TG:ALKAL2 complex with regions differing with ALKAL1 coloured Vermillion. C-termini of the TG domains leading towards the EGF-like domains are coloured red. f, ITC experiments for the titration of LTK_TC (10 μM) with ALKAL1 (70 μM) Data representative of 2 independent experiments, g, ITC experiments for the titration of LTK_TG (10 μM) with ALKAL2 (100 μM) Data representative of 2 independent experiments, h, ITC experiments for the titration of LTK_TG (10 μM) with ALKAL1_FL (40 μM) Data representative of 2 independent experiments, i, ITC experiments for the titration of ALK_FL (8 μM) with ALKAL2 (82 μM) Data representative of 2 independent experiments, j, Characterization of heparin-induced ALK_FL dimerization by SEC-MALLS. Differential refractive index (left vertical axis) is plotted against the determined molecular weight (right vertical axis). ALK_FL (pink), ALK_FL complexes with heparin dp20 (green) and ALK_FF complexed with heparin dp20 and ALKAL2 (blue trace). The reported molecular mass represents the average molecular mass ± s.d. across the elution peak, k, Overview of the different ligand-mediated extracellular assemblies across RTKs. Trka–NGF (PDB: 21FG), EGFR–EGF (PDB: 1IVO), INSR–INS (PDB: 6PXW), CSF1R–CSF1 (PDB: 4WRM), FGFR–heparin–FGF (PDB: 1FQ9), ALK_TC–ALKAL2 (PDB: 7NWZ).

Extended Data Table 1 |.

Crystallographic data and refinement statistics

	ALKTG-EGFL	ALKTG-EGFL—Fab324	LTKTG	ALKTG—ALKAL2	LTKTG—ALKAL1—Nb3.16	Fab324
Crystallization conditions	0.1M Gly-Gly/AMPD pH 8.5 15% w/v PEG 3000 20% v/v 1,2,4-Butanetriol 1% w/v NDSB 256 5mM NiCl2.6H2O 5mM CoC12.6H2O 5mM ZnCl2.6H2O 5mM MnCl2.6H2O	40mMPolyamines, 0.1M Gly-Gly/AMPD pH8.5 12.5% w/v PEG4000 20% w/v 1,2,6-Hexanetriol	MOPSO/Bis-Tris pH 6.3 12% PEG 20000 25% Trimethyl propane 2% w/v NDSB 195 5mM YC13.6H2O 5mM ErCl3.6H2O 5mM TbCl3.6H2O 5mM YbCl3.6H2O	0.1M MES pH 6.5 15% w/v PEG 6000 5% w/v MPD	MOPSO/BIS-TRIS pH 6.5 13% PEG 8000 22% w/v 1,5-pentanediol 5mM Na2CrO4.4H2O 5mM Na2MoO4.4H2O 5mM Na2WO4.4H2O 5mM Na2VO4.4H2O	1.5M (NH4)2SO4 40 mM Glycine pH 9.5
Cryoprotectant	25% v/v 1,2,4-Butanetriol	25% w/v 1,2,6-Hexanetriol	25% Trimethyl propane	22% ethylene glycol	25% 1,5-pentanediol	Saturated (NH4)2SO4
Data collection
X-ray source (beamline)	PETRAIII (P14)	ESRF (ID-23-2)	PETRAIII (P13)	SOLEIL (Proxima 2A)	SOLEIL (Proxima 2A)	PETRAIII (P13)
Wavelength (Å)	1.033	0.8731	0.9763	0.9800	0.9800	0.9762
Space group	I41	P 21 21 21	P21	P 43 21 2	P 61	P 21 21 21
Cell dimensions
a, b, c (Å)	84.93 84.93 148.67	109.7 137.47 144.26	55.04 54.46 55.01	97.57 97.57 355.35	129.11 129.11 109.75	81.24 87.09 126.6
α,β,γ (°)	90.00, 90.00, 90.00	90.00, 90.00, 90.00	90 111.284 90	90.00, 90.00, 90.00	90.00 90.00 120.00	90.00, 90.00,, 90.00
Resolution (Å)	60.05 – 3.0 (3.11 – 3.0)	48.04 – 2.81 (2.91 – 2.81)	51.26 – 1.3 (1.38 – 1.3)	48.96 – 4.18 (4.32 – 4.18)	42.3 – 1.95 (2.02 – 1.95)	71.76 – 1.47 (1.52 – 1.47)
Rmeas (%)	18.9 (114.9)	25.4 (120.1)	6.9 (125.5)	20.6 (168.1)	7.6 (147.1)	4.2 (35.8)
<I/σ>	10.15 (1.26)	11.11 (1.91)	11.5 (1.0)	13.9 (2.0)	26.6 (1.9)	17.32 (2.4)
CC ½ (%)	100 (75.8)	99.9 (34.1)	99.9 (54.6)	99.9 (89.2)	99.9 (88.5)	99.7 (42.1)
Completeness (%)	99.90 (99.90)	99.19 (93.06)	98.90 (98.5)	99.50 (94.5)	99.7 (96.3)	96.20 (73.3)
Redundancy	14.3 (14.4)	5.8 (4.5)	3.4 (3.2)	25.6 (24.6)	20.8 (19.8)	3.8 (3.2)
Wilson B factor (Å2)	104.20	39.46	14.23	105.89	38.80	16.20
Refinement
Resolution (Å)	60.05 – 3.0	48.04 – 2.81	51.26 – 1.3	48.78 – 4.18	42.3 – 1.95	71.76 – 1.47
No. reflections	10550	53379	145183	13500	75501	146924
Rwork / Rfree (%)	23.4/26.0	22.6 / 26.3	16.2 / 18.1	25.3 / 26.8	16.9 / 19.6	16.9 / 20.5
No. atoms
Protein	2496	10653	2098	9763	6502	6486
Ligand/ion	31	-	46	98	4	-
Water	-	-	232	-	691	917
B-factors (Å2)
Protein	103.45	52.21	21.08	162.0	53.25	19.60
Ligand/ion	116.68	-	25.67	195.3	116.64	24.54
Water	-	-	36.65	-	57.86	30.49
R.m.s. deviations
Bond lengths (Å)	0.002	0.005	0.007	0.002	0.006	0.008
Bond angles (°)	0.53	0.87	1.09	0.50	0.90	1.03
Pdb id	7NX4	7NX3	7NX1	7NWZ	7NX0	7NX2

Each data set was collected from a single crystal. Values in parentheses correspond to the highest resolution shell.

Extended Data Table 2 |.

Interaction interface analysis of ALK—ALKAL2 and LTK—ALKAL1 complexes

a
LTK—ALKAL1 site 1			LTK—ALKAL1 site 2			LTK—ALKAL1 site 3
ALKAL1 Chain A	LTK Chain B	dist (Å)	ALKAL1 Chain A	LTK Chain C	dist (Å)	LTK Chain B	LTK Chain C	dist (Å)
Hydrogen-bonds and salt bridges			Hydrogen-bonds and salt bridges			Hydrogen-bonds and salt bridges
GLU 103[OE2]	TYR 124[OH]	2.57	HIS 79[O]	ARG 376[NH2]	3.43	ARG 241[NE]	GLY 88[O]	3.57
ARG 102[NH2]	TYR 124[OH]	3.51	HIS 79[NE2]	SER 122[OG]	2.8	ARG 241[NH2]	GLY 88[O]	3.46
ARG 119[NH2]	SER 251[O]	3.28	LYS 111[NZ]	TYR 124[OH]	3.0	ARG 241[NH2]	ALA 89[O]	2.86
ARG 119[NH2]	GLU 253[OE1]	3.09	ARG 112[NH2]	GLU 368[OE1]	3.06	ARG 241[NH2]	ALA 91[O]	2.62
ARG 119[NE]	GLU 253[OE2]	2.80	van der Waals contacts			GLN 85[NE2]	CYS 179[O]	3.37
HIS 93[ND1]	GLN 362[OE1]	3.46				THR 84[OG1]	GLU 182[OE2]	2.77
ARG 115[NH2]	VAL 366[O]	2.89	PHE 80	TYR 124		ARG 138[NH1]	ASN 369[OD1]	3.51
ARG 115[NE]	VAL 366[O]	2.84		PHE 143		ASN 369[ND2]	ASN 369[OD1]	3.22
ARG 112[NH1]	GLU 372[OE1]	3.59	PRO 107	LEU 364		GLY 88[O]	ARG 241[NH2]	3.16
ARG 102[NH1]	GLU 372[OE2]	3.28		VAL 366		ALA 89[O]	ARG 241[NH2]	2.94
ARG 102[NH2]	GLU 372[OE2]	3.20	ALA 108	ALA 365		ALA 91[O]	ARG 241[NH2]	2.89
ASN 100[O]	ARG 376[NH1]	3.61	PHE 76	LEU 361		ALA 91[O]	ARG 241[NH1]	2.86
				PHE 143		CYS 179[O]	GLN 85[NE2]	3.31
van der Waals contacts				GLN 362		ASN 369[OD1]	ARG 138[NH1]	3.14
ARG 96	PHE 360			LEU 364		ASN 369[OD1]	ASN 369[ND2]	3.17
LEU 97	LEU 361		TYR 110	LEU 364		van der Waals contacts
SER 123	GLN 362		HIS 79	LEU120
GLU 103	TYR 124			PHE143		HIS 135	CYS 73
LEU116	VAL 266			ARG376			GLY 74
	LEU 364						THR 84
	PHE 143					LEU 180	GLN 85
LEU 120	LEU 364					GLU 182	GLY 92
	GLN 362					ARG 243	HIS 135
						CYS 73
						GLY 74
						THR 84	LEU 180
						GLN 85	GLU 182
						GLY 92	ARG 243
b
ALK—ALKAL2 site 1			ALK—ALKAL2 site 2			ALK—ALKAL2 site 3
ALKAL2 Chain C	ALK Chain A	dist (Å)	ALKAL2 Chain C	ALK Chain B	dist (Å)	ALK Chain A	ALK Chain B	dist (Å)
Hydrogen-bonds and salt bridges			Hydrogen-bonds and salt bridges			Hydrogen-bonds and salt bridges
TYR 739[OH]	ASP 134[OE2]	2.8	HIS 100[O]	LYS 982[NZ]	3.53	THR 697[OG1]	ILE 795[O]	3.54
HIS 857[O]	ARG 140[NH2]	3.90	LYS 96[NZ]	TYR 966[O]	3.30	GLN 700[N]	ILE 795[O]	3.83
HIS 857[O]	ARG 140[NH2]	3.77	HIS 100[NE2]	SER 737[OG]	3.49	ASN 749[ND2]	GLN 706[OE1]	3.20
GLU 859[OE2]	ARG 140[NE]	2.96	LYS 132[NZ]	GLU 978[OE1]	2.91	ILE 795[O]	THR 697[OG1]	3.38
TYR 966[OH]	ASN 121[ND1]	3.02				CYS 794[O]	GLN 700[NE2]	3.84
VAL 972[O]	ARG 136[NH2]	3.23				ASN 749[OD1]	GLN 706[NE2]	3.31
GLU 974[OE2]	ARG 136[NE]	3.08
GLU 978[OE1]	ARG 133[NH1]	2.8
van der Waals contacts			van der Waals contacts			van der Waals contacts
ARG 117	TYR 966		MET 93	THR 967		GLY 689	THR 750
LEU 118	LEU 970			PRO 968		THR 750	GLY 689
LEU116	THR 967			LEU 970		ILE 795	GLN 706
	LEU 970		LEU 101	TYR 739
	VAL 972		PHE 97	THR 967
LEU 120	LEU 970			LEU 970
			LYS 132	VAL 972
				MET 973

List of contacting amino acids at interaction sites 1, 2, and 3 of the ALKAL2—ALK_TG and ALKAL1—LTK_TG complexes. Hydrogen bonding residues were determined using the PISA server at EBI (https://www.ebi.ac.uk/msd-srv/prot_int/pistart.html) and confirmed by analysis in ChimeraX. van der Waals contacts were analyzed using the ‘find contacts’ function in ChimeraX. The molecular components involved in each described interface are shown above each column.