Structural basis of Tie2 activation andTie2/Tie1 heterodimerizationVeli-Matti Leppänena,1, Pipsa Saharinena,b, and Kari Alitaloa,b,1

aWihuri Research Institute, Biomedicum Helsinki, Haartmaninkatu 8, 00290 Helsinki, Finland; and bTranslational Cancer Biology Program, Research ProgramsUnit, University of Helsinki, 00014 Helsinki, Finland

Contributed by Kari Alitalo, March 8, 2017 (sent for review September 28, 2016; reviewed by Joseph Schlessinger and Michel O. Steinmetz)

The endothelial cell (EC)-specific receptor tyrosine kinasesTie1 and Tie2 are necessary for the remodeling and maturationof blood and lymphatic vessels. Angiopoietin-1 (Ang1) growthfactor is a Tie2 agonist, whereas Ang2 functions as a context-dependent agonist/antagonist. The orphan receptor Tie1 modu-lates Tie2 activation, which is induced by association of angio-poietins with Tie2 in cis and across EC–EC junctions in trans.Except for the binding of the C-terminal angiopoietin domainsto the Tie2 ligand-binding domain, the mechanisms for Tie2activation are poorly understood. We report here the structuralbasis of Ang1-induced Tie2 dimerization in cis and providemechanistic insights on Ang2 antagonism, Tie1/Tie2 heterodi-merization, and Tie2 clustering. We find that Ang1-inducedTie2 dimerization and activation occurs via the formation of anintermolecular β-sheet between the membrane-proximal (third)Fibronectin type III domains (Fn3) of Tie2. The structures ofTie2 and Tie1 Fn3 domains are similar and compatible with Tie2/Tie1 heterodimerization by the same mechanism. Mutagenesisof the key interaction residues of Tie2 and Tie1 Fn3 domainsdecreased Ang1-induced Tie2 phosphorylation and increasedthe basal phosphorylation of Tie1, respectively. Furthermore,the Tie2 structures revealed additional interactions betweenthe Fn 2 (Fn2) domains that coincide with a mutation ofTie2 in primary congenital glaucoma that leads to defectiveTie2 clustering and junctional localization. Mutagenesis of theFn2–Fn2 interface increased the basal phosphorylation of Tie2,suggesting that the Fn2 interactions are essential in preformedTie2 oligomerization. The interactions of the membrane-proximaldomains could provide new targets for modulation of Tie recep-tor activity.

tyrosine kinase | angiopoietin | dimerization | homotypic interactions |crystallography

Receptor tyrosine kinases (RTKs) expressed in the endothelialcells (ECs) of blood and lymphatic vessels control the de-

velopment and function of the cardiovascular and lymphaticsystems. The VEGFs and their endothelial receptors (VEGFRs)are key regulators of angiogenesis and vascular integrity (1, 2).The angiopoietin ligand/Tie receptor pathway is necessary forblood and lymphatic vessel remodeling during embryonic andpostnatal development and for homeostasis of the mature vas-culature (3, 4). Recently significant interest has focused on tar-geting the VEGFR and Tie receptor pathways in antiangiogenicand antilymphangiogenic therapies (5).Ang1 activation of Tie2 is indispensable for embryonic cardiac

development and angiogenesis, and both Ang1 and Ang2 arenecessary for the development of lymphatic and ocular vascula-ture. In adult tissues, Ang1 is required for vessel stabilizationafter angiogenesis (6–8). Ang2, which is produced by ECs andstored in their Weibel–Palade bodies for rapid release, canfunction as a weak Tie2 agonist or as a context-dependent an-tagonist that inhibits Ang1-induced Tie2 activation and vascularstability (9–11). Tie2 is the major signal-transducing receptor ofthe angiopoietin/Tie signaling axis, and the homologous Tie1receptor modulates Tie2 signaling (12, 13). Although Tie1, firstidentified in human leukemia cells (14), is an orphan receptor,

mice lacking Tie1 develop severe edema around E13.5 becauseof compromised microvessel integrity and defects in lymphaticvasculature and die subsequently (15, 16). Furthermore, Tie1 hascritical functions in vascular pathologies, e.g., in tumor angio-genesis and atherosclerosis progression (12, 17).In EC monolayers, angiopoietins stimulate Tie receptor trans-

location to cell–cell junctions for Tie2 trans-association, whereasin the absence of cell–cell adhesion the Tie receptors are an-chored to the extracellular matrix (ECM) by Ang1-induced Tie2cis-association (10, 18). Integrins also have been implicated inTie2 signaling, and the α5β1–integrin heterodimer enhancesAng1-induced EC adhesion and Tie2 activation (13, 19, 20). TheTie1 and Tie2 heteromeric complexes are promoted by angio-poietin stimulation, resulting in Ang1-induced activation of bothTie1 and Tie2 (13, 21–23). Several studies have indicated Tie1 asa Tie2 inhibitor (22, 24, 25), whereas recent experiments showthat Tie1 association with Tie2 is required for Tie2 activation byAng1 and Ang2 in mice (13, 26). Tie1 expression inhibitsTie2 presentation at the cell surface in sprouting endothelial tipcells, (23), but Tie1 sustains Tie2 signaling in contacting cells(13, 23). Thus, Tie1 exerts its context-dependent effects bymodulating Tie2 activity (13, 22, 23, 26).Ligand-induced dimerization is regarded as a common, but

not the only, mechanism for activation of RTK signaling (27).Angiopoietin monomers form heterogeneous multimeric Tie2ligands, and it has been suggested that Tie2 activation requiresreceptor clustering (28–30). Also, Tie2 clustering without ligand

Significance

Tie1 and Tie2 receptor tyrosine kinases are key regulators ofblood and lymphatic vessel development and of pathologicalprocesses including tumor angiogenesis, atherosclerosis, andvascular leakage, e.g., in sepsis. Tie1 is essential for theTie2 agonist activity of angiopoietins, and the activated re-ceptors form heteromeric complexes in endothelial cell–celljunctions. However, little is known about the activationmechanisms of the Tie receptors. Here we demonstrate thatthe membrane-proximal domains of Tie2 mediate homotypicinteractions, which occur via intermolecular β-sheet formationand are necessary for Tie2 activation. The structural analysissuggests that Tie2/Tie1 heterodimerization occurs by the samemechanism. The crystal structures provide a model forangiopoietin-stimulated Tie2 ectodomain dimerization, clus-tering, and activation and insights into therapeutic targeting.

Author contributions: V.-M.L. and K.A. designed research; V.-M.L. performed research;V.-M.L. analyzed data; and V.-M.L., P.S., and K.A. wrote the paper.

Reviewers: J.S., Yale University School of Medicine; and M.O.S., Paul Scherrer Institut.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates, and structure factors reported in this paperhave been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 5MYA,5MYB, and 5N06).1To whom correspondence may be addressed. Email: kari.alitalo@helsinki.fi or veli-matti.leppanen@helsinki.fi.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1616166114/-/DCSupplemental.

4376–4381 | PNAS | April 25, 2017 | vol. 114 | no. 17 www.pnas.org/cgi/doi/10.1073/pnas.1616166114

Dow

nloa

ded

by g

uest

on

May

27,

202

1

binding has been reported (31). The Tie receptors have a uniqueextracellular domain (ECD) for ligand binding, a single-passtransmembrane domain, a two-partite cytoplasmic protein tyro-sine kinase domain, and a C-terminal tail. The ECDs consist ofIg, epidermal growth factor-like, and three fibronectin type III(Fn) domains (Fn1, Fn2, Fn3) (30, 32). The angiopoietins havean N-terminal region responsible for their multimerization, acoiled-coil domain for dimerization, and a C-terminal fibrinogen-like domain (FLD) that contains the Tie2-binding region (28, 29).Crystal structures of the angiopoietin/Tie2 ligand-binding domain(LBD) complexes demonstrate that Ang1 and Ang2 FLDs bindTie2 in a similar manner (30, 33). Ang1 is a strong Tie2 agonist,and Ang2 a weak Tie2 agonist, suggesting that the Tie2 agonismof the native angiopoietins resides in sequences outside the FLDs.Because multimerization of Ang1 and Ang2 is critical for Tie2phosphorylation, it has been suggested that the difference be-tween the agonistic activities of Ang1 and Ang2 is caused bydifferent oligomeric states (34, 35). Indeed, multiple studies in-dicate that the strongly activating Ang1 has higher tendency toform large oligomers than the weak agonist Ang2, although thesignificance of differential oligomerization is poorly understood(28, 29, 35, 36). The need for oligomerization of angiopoietinligands suggests that Tie2 clustering is required for its activation,but the mechanism(s) mediating Tie2 dimerization or higher-order clustering are not known.We report here the crystal structures of Tie2 and Tie1 mem-

brane proximal domains and their structural and functional anal-ysis, which provide mechanistic insight into ligand-inducedTie2 activation and to the mode of Tie1/Tie2 heterodimerization.

ResultsStructure of Tie2 Fn-Like Domains 1–3. To better understand themechanism of Tie2 dimerization and activation, we expressedTie2 Fn-like domains 1–3 (Fn1–3) (Fig. 1A) in insect cells andcrystallized the purified protein in space groups C2 and P21. Thecrystal structure in space group C2 was determined at 2.9-Åresolution using multiple isomorphous replacement with anom-alous scattering (MIRAS) phases, and the structure in spacegroup P21 was determined at 2.6-Å resolution with molecularreplacement using the structure in the C2 space group as asearch model (Table S1). Both structures revealed a homodimerof the Tie2 Fn-like domains in the asymmetric unit (Fig. 1B).

The dimerization is based on symmetrical interfaces between themembrane-proximal Fn3 domains, placing the C termini close toeach other at a distance of about 25 Å. The C termini point inthe same direction, whereas the N-terminal Fn-like domains aredirected away from each other. Dimerization is mediated mainlyby hydrogen bonds between the main-chain atoms of antiparallelβ-strands in residues Asp682 to Lys690. Other interactions occurbetween symmetry-related Lys700 and Gly701, and the inter-acting residues are largely conserved (Fig. 1C). According to thePisa server analysis, the interface buries a surface area of about580 Å2 per chain in space group C2; the solvation free energygains of −1.5 and −1.9 kcal/mol for each chain and the related Pvalue of 0.54 suggest considerable specificity (Tables S2 and S3).In space group P21, the interactions between the antiparallelβ-strands are almost identical, but, because of additional inter-actions, the buried surface area is larger, about 700 Å2 per chain(Tables S2 and S3).The Tie2 homodimers in the two crystal structures share the

same overall structure, except that only one of the two Tie2proteins in the C2 dimer has the Fn1 domain (Fig. S1). The 376Cα atoms in the two Tie2 Fn2–3 dimers superimpose with anrmsd of 1.445 Å. The crystal packing indicates that the Fn1 domainswere proteolytically removed before crystallization in space groupP21, whereas in space group C2 the other Fn1 domain is eitherdisordered or present at low occupancy, because crystal packing inC2 should accommodate it, but only weak residual density is seen(Fig. S2). Superposition of the Fn1 domain-bearing chain in spacegroup C2 with the chain lacking the Fn1 domain creates a model ofa Tie2 Fn1–3 homodimer (Fig. 1D). A recent EM analysis of full-length Tie2 ECDs did not indicate Tie2 dimerization, but the shapeand length of the extended structure of Tie2 Fn-like domains 1–3 inspace group C2 are consistent with those in the EM analysis(11.8 nm vs. 12 nm) (37). Although data from small-angle X-rayscattering (SAXS) of an Fn2–3 protein (at ∼80 μM) indicated astructure that fits best with our crystal structure of an Fn3 domain-mediated Tie2 dimer, the Tie2 ECD and Tie2 Fn domains weremonomeric in solution when analyzed by size-exclusion chroma-tography multiangle laser light scattering (SEC-MALLS) (Fig. S3A–E). This finding suggests that the unliganded receptor has arelatively weak tendency for the Fn-domain–mediated interactionsand that the formation of the putative homotypic interactions bythe full-length receptor requires spatial (and dimensional) con-straint in the plasma membrane and/or is ligand dependent.

Homotypic Fn3 Interactions Are Required for Ligand-Induced Tie2Activation. We tested the model of Fn3-mediated homotypic in-teractions of Tie2 by targeting the dimerization interface with aV685Y/V687Y double mutation and by expressing the WT andmutant receptors in HeLa cells for ligand-stimulation experi-ments. Two conserved valine residues were changed to bulkiertyrosine residues to introduce steric hindrance across the Fn3–Fn3 interface (Fig. 2 A and B). A model of the Tie2 Fn3 mutantindicates that the neighboring Tyr674 and Tyr697 residues forcethe tyrosine mutant conformations toward the Fn3–Fn3 interface(Fig. 2B). Although Comp-Ang1 stimulation of the WT Tie2 in-duced strong Tie2 phosphorylation and colocalization with Tie1 incell–cell junctions of receptor-transfected HeLa cells, phosphory-lation of the mutant Tie2 was much reduced, strongly suggestingthat the mutation interfered with Tie2 dimerization (Fig. 2 C andD) (10, 18). Neither these Fn3 mutations nor the other Tie2 orTie1 mutations described below had any overt effect on receptorfolding or expression in HeLa cells (Fig. S3 F and G). The Fn1–Fn2 junction in Tie2, unlike the Fn2–Fn3 junction, has no specificinteractions between the two domains, consistent with its apparentflexibility and susceptibility to proteolysis (Fig. S4 A–C). In com-parison with the analysis of the Fn1 domain of the Ang1–Tie2LBD complex at 4.5-Å resolution (33), we observed a sequenceregister difference of one to three residues between the Fn1structures (Fig. S4D). However, by superimposing the Fn1 do-mains in the Tie2 Fn1–3 homodimer and Ang1–Tie2 LBD com-plex structures (33), we were able to construct a structural model

Fig. 1. Homodimerization of Tie2 Fn-like domains. (A) A schematic modelof Tie2 with the LBD and the three membrane-proximal Fn-like domainslabeled. (B) Comparison of the homodimer structures of Tie2 Fn-like domainsin space groups C2 and P21. The structures are shown as cartoon diagramswith the Fn-like domains labeled and the N and C termini labeled whereapplicable. (C) A surface-and-stick representation of the Tie2 Fn3 interactions.The Tie2 residues in the Fn3 surface model are colored according to the rateof evolutionary conservation from cyan (variable) to magenta (conserved).(D) A model of homodimer of the Tie2 Fn-like domains 1–3. Asn-linkedN-acetylglucosamine moieties are shown as cyan spheres.

Leppänen et al. PNAS | April 25, 2017 | vol. 114 | no. 17 | 4377

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

27,

202

1

of the Ang1-bound Tie2 ECD homodimer that provided mecha-nistic insight for Tie2 activation in cis (Fig. 3 A and B). Interestingly,the Tie2 LBDs are too far apart for a dimeric angiopoietin ligand tobridge the receptors for successful dimerization and activation.Instead, a dimeric ligand would compete with multimeric angio-poietin for Tie2 binding (Fig. 3B).

A Model of Tie1/Tie2 Heterodimerization. To study the mechanismof Tie1 and Tie2 heterodimerization, we expressed the membrane-proximal Tie1 Fn3 in insect cells and crystallized the purifiedprotein for structure determination. The structure was solved at2.6-Å resolution by molecular replacement using the Tie2 Fn3domain as a search model (Table S1). Interestingly, the structure isa strand-swapped homodimer in which the two Fn-like domainsexchange their C-terminal β-strands (Fig. S5 A and B). Strandswapping occurs by extension of the sixth β-strand toward itscounterpart to form an antiparallel β-sheet that extends overIle722–Gly724 in the bridge region where the two strands meet.This swapping may be an important feature of Tie1 function, buta model derived from SAXS data indicates that Tie1 Fn3 is amonomer rather than a dimer in solution (Fig. S5C). Thus, thestrand-swapped dimerization may instead be a crystallizationartifact of an isolated domain.Comparison of the Tie1 and Tie2 Fn3 domain structures re-

veals very similar folds in both, despite their low sequenceidentity (Fig. 4A), suggesting that structural homology betweenthe Tie receptors extends to the Fn3-mediated dimerization andactivation. Superposition of Tie1 Fn3 domain with an Fn3 domainin the Tie2 homodimer indicates that Tie1/Tie2 heterodimerizationalso could be mediated by interactions between the antiparallelβ-strands in the Fn3 domains (Fig. 4B). The model also suggestsadditional favorable interactions, including a salt bridge betweenTie2 Asp682 and Tie1 Arg697 (Fig. 4B). We tested the model ofFn3-mediated Tie2/Tie1 heterodimerization by analyzing Tie1phosphorylation in Tie1/Tie2-transfected HeLa cells. We created a

Tie1 tyrosine mutant (3Y; V681Y, L691Y, and I693Y) in ananalogous manner to the Tie2 Fn3 double mutant. Val681 wasmutated to tyrosine to mimic the Tie2 Fn3 domain and to directthe side-chain conformers of I693Y and L695Y toward theinterface. Another mutant (5M) with additional D689R andD696R substitutions was designed to disrupt the interactionsfurther. The Tie1 mutants increased the basal phosphorylation ofTie1 and made Tie1 insensitive to Tie2-mediated Comp-Ang1stimulation (Fig. 4 C and D). Interestingly, the 5M mutant did notaffect Tie1/Tie2 heterodimerization according to the cell-surfacecrosslinking assay, and Tie1 baseline phosphorylation was not af-fected in Tie1 singly transfected cells, suggesting that the Tie1Fn3 mutations affected only Tie1/Tie2 kinase crosstalk (Fig. S6A–C).

A Model for Tie2 Clustering. In addition to the very similar inter-actions between the Fn3 domains in the C2 and P21 structures,we observed structurally analogous interactions between theFn2 domains of neighboring homodimers in the crystal, sug-gesting that both types of interactions exhibit natural proteincontacts (Fig. 5A). Combining Fn3-guided homodimerization ofTie2 with Fn2-guided interactions between homodimers createsan array of Tie2 molecules that could represent the structuralbasis of Tie2 clustering (Fig. S7A). The symmetrical interactionsbetween the Fn2 domains involve multiple hydrogen bonds be-tween highly conserved Arg-Trp-Arg motifs and the main-chainatoms of residues 582–585 in a hairpin loop of the neighboringchain (Fig. S7 B and C). Interestingly, in a primary congenitalglaucoma (PCG) family, a TEK mutation (Y611C) is locatedclose to this Fn2′–Fn2′′ interface of neighboring homodimers,resulting in haploinsufficiency because of the loss of Tie2 func-tion (Fig. 5B) (38). Tyr611 is a buried residue that packs againstthe hairpin loop, making interactions with the Arg-Trp-Argmotif. Mutation of Tyr611 to cysteine would remove a poten-tially important hydrogen bond to His606 in the neighboringFn2 loop. We next created a Q588R/V612R double mutant totarget the Fn2′-Fn2′′ interface and analyzed the effect of theFn2 mutations on Tie2 activation in Tie2 singly transfected andTie1/Tie2 doubly transfected HeLa cells (Fig. 5 C and D and Fig.S7 D and E). The mutant showed a significant increase in basal

Fig. 2. Tie2 Fn3-mediated homotypic interaction. (A) A bottom view of the di-merization interface in Tie2 Fn3. The dimerization is mediated by hydrogen bondsbetween the main-chain atoms of antiparallel β-strands between two neighbor-ing Tie2 molecules. (B) As in A with Val685 and Val687 mutated to tyrosines. Thetyrosine side-chain conformations are the most common ones in the backboneconformation-dependent rotamer library in PyMOL. (C) GFP (green) and phospho-Tie2 (Tyr992; red) immunofluorescence staining of HeLa cells transfected withTie1-GFP and Tie2-WT or the Tie2-Fn3 mutant. Note the junctional colocalizationof Tie1 and phospho-Tie2 in Comp-Ang1 (cA1) stimulated WT Tie2 transfectedcells and that the phospho-Tie2 (Tyr992) antibody is known to bind also to thehomologous region in Tie1. (D) Western blot analysis of Tie2 phosphorylation inthe transfected HeLa cells. Cells were stimulated with COMP-Ang1 (cA1) for 1 h orwere left unstimulated. Error bars indicate SEM. **P ≤ 0.01. Student’s t test; n = 3.

Fig. 3. Structural basis of Tie2 dimerization and activation in cis. (A) Amodel of the ligand-bound homodimer of Tie2 ECDs based on the structureof the Ang1-bound Tie2 LBD–Fn1 complex (PDB ID code 4K0V) and thehomodimer of the Tie2 Fn-like domains. The Ang1 receptor-binding domain(RBD), colored in red, and the two chains of Tie2 ECDs, colored in blue andorange, are shown as cartoon and semitransparent surface models. Note thedistance between the Tie2 LBDs. (B) A schematic model of Ang2 antagonism.The model of the Ang1 tetramer was adopted from angiopoietin tetramersanalyzed by EM (29). The Ang2-activating antibody ABTAA binds to theAng2 fibrinogen-like domains and creates a tetrameric Ang2 suitable forTie2 dimerization and activation (43).

4378 | www.pnas.org/cgi/doi/10.1073/pnas.1616166114 Leppänen et al.

Dow

nloa

ded

by g

uest

on

May

27,

202

1

Tie2 phosphorylation in both singly and doubly transfected cells.Interestingly, the Fn2 mutant could be stimulated with Comp-Ang1 in the presence of Tie1 but not in the absence of Tie1.Alignment of Ang1-bound homodimers of the Tie2 ECDs cre-ates a model for Tie2 clustering (Fig. 5E). Notably, multipleTie2 homodimers can interact without steric clashes, and themodel brings the neighboring LBDs close to each other.

DiscussionAlthough the structural features of the angiopoietin receptor-binding domain interactions with the Tie2 LBD have been de-scribed, the structural basis for Ang1-induced Tie2 activation,Tie1/Tie2 heterodimerization, and the context-dependent an-tagonism by Ang2 has been largely lacking. Our crystal structuresand functional analysis of mutant receptors show that themembrane-proximal Fn3 domains in Tie2 and Tie1 contributeto Tie2 homodimerization and Tie2/Tie1 heterodimerization,whereas the Tie2 Fn2 domain seems to mediate Tie2 clustering.In addition, our results provide a structural explanation for theneed of Ang1 oligomers in Tie2 activation.Crystal structures of the Tie2 Fn-like domains revealed a

mechanism of Fn3-mediated Tie2 dimerization in which theFn3 domains bring their C termini close to each other in a waysimilar to that described for the membrane-proximal Ig-like

domains in KIT and VEGFR-2 (39, 40). In VEGFR-2, the Ctermini also are about 25 Å apart, and in these three RTKs themembrane-proximal domains contribute symmetrical interactionsaround an axis with twofold symmetry (39, 40). Interestingly, theinteractions between the Tie2 Fn3 domains occur via in-termolecular hydrogen bonding between antiparallel β-strands.Although this binding creates an intermolecular β-sheet with onlya few side-chain–mediated interactions, the interacting residuesare largely conserved. Mutation of two conserved valines in theinteracting β-strands in Fn3 to bulkier tyrosines was made todisorient the transmembrane and kinase domains for optimaltransphosphorylation upon ligand-stimulated Tie2 activation.These mutations reduced Comp-Ang1–stimulated Tie2 phosphor-ylation, thus confirming the importance of the intermolecularβ-sheet formation for Tie2 activation. In the accompanying study inthis issue of PNAS, Moore et al. describe the same intermolecularinteractions in an independent crystal structure of Tie2 Fn-likedomains and show reduced Ang1-induced Tie2 phosphorylationwith the Y697A mutation located in the interface (41). Receptoractivation via intermolecular β-sheet formation, as seen in this di-mer, is unique to RTKs, although intermolecular β-sheet formationrepresents a common mode of protein–protein interactions (42).We constructed a model of the ligand-bound Tie2 ECD

homodimer based on our Tie2 Fn1–3 homodimer and the pub-lished Ang1–FLD/Tie2–LBD complex (33) structures. The modelindicates that the ligand-binding sites are too far apart for adimeric angiopoietin to promote the formation of a Tie2 dimerin the cis orientation. Since Ang2 is mainly a covalent dimer,whereas Ang1 is multimeric in nonreducing conditions, themodel provides an interesting possible explanation why Ang2functions as a weak agonist (28, 35, 36). In EM, both Ang1 andAng2 exist from dimers up to hexamers, but although the ma-jority of Ang2 showed low-order “mushroom-like” states, themajority of Ang1 existed as high-order oligomers (29, 35). Unlikethe native Ang2, tetrameric Bow–Ang2 (Ang-F2-Fc-F2) is a po-tent Tie2 agonist (26, 29), suggesting that the extended “reach”of receptor-binding sites in these multimers is crucial for pro-moting Tie2 dimerization in cis. This dimerization may be fa-cilitated by the flexibility of the Fn1–Fn2 junction revealed bycomparing our structures with that in the accompanying paper byMoore et al in this issue of PNAS (41). Furthermore, the recentlypublished Ang2-binding and Tie2-activating antibody ABTAApresumably binds two Ang2 low-order oligomers, possibly homo-dimers, in an optimal angle for Tie2 activation (Fig. 3B) (43).Tie1 is an orphan receptor, but Ang1 activates Tie1 in Tie2/

Tie1 heteromeric receptor complexes (21–23). Prompted by ourobservation of Fn3-mediated homotypic Tie2 interactions, wesolved the Tie1 Fn3 crystal structure for comparison. Structuralcomparison of the Tie2 and Tie1 Fn3 domains revealed verysimilar folds despite their low sequence identity, suggesting thatTie1 also interacts with Tie2 through intermolecular β-sheetformation. To analyze the role of Tie1 Fn3 in Tie2/Tie1 hetero-dimerization, we created the Tie1 Fn3 tyrosyl mutants 3Y and 5Min which the mutations correspond to those in Tie2. Both alter-ations increased the basal phosphorylation level of Tie1, makingTie1 insensitive to Comp-Ang1 stimulation in cells coexpressingTie2. Interestingly, the Tie1 Fn3 mutations did not affect Tie1/Tie2heterodimerization in the crosslinking assay, and the increasedbaseline activation was Tie2-dependent. These results indicate thatthe Tie1/Tie2 heterodimerization is likely to involve interactionsbetween the membrane-proximal domains and that these interac-tions restrict aberrant Tie2–Tie1 cross-activation. We cannotexclude alternative interactions for Tie1 and Tie2 Fn3 domains,and additional interactions, such as putative electrostatic inter-actions between the Tie2 LBD and the corresponding domain inTie1 (25), may stabilize the heterodimerization despite the Tie1Fn3 mutations.In addition to Fn3-mediated Tie2 homodimerization, the two

crystal structures of the Tie2 Fn-like domains revealed sym-metrical interactions between the Fn2 domains that could me-diate Tie2 clustering. The interactions are almost identical in

Fig. 4. Model of Tie1/Tie2 heterodimerization. (A) Structural comparison ofthe Fn3 domains in Tie1 and Tie2. (B) A Tie2 homodimer-guided model ofTie1/Tie2 heterodimer of the Fn3 domains. (C) A model of the Tie1Fn3 mutant 5M in the Tie1/Tie2 heterodimer. Three aliphatic residues werechanged to tyrosine residues in the Tie1 Fn3 3Y mutant (magenta). The 5Mmutant also has two aspartates changed to arginines (red). (D) Western blotanalysis of Tie1 phosphorylation in HeLa cells doubly transfected withTie2 and Tie1-WT or the Tie1-Fn3 mutant 3Y or 5M. Cells were stimulatedwith COMP-Ang1 (cA1) for 1 h or were left unstimulated. Error bars indicateSEM. *P ≤ 0.05; **P ≤ 0.01; ns, not significant; Student’s t test, n = 3.

Leppänen et al. PNAS | April 25, 2017 | vol. 114 | no. 17 | 4379

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

27,

202

1

both structures, and residues around the Arg-Trp-Arg motif,which interacts with a hairpin loop of the neighboring chain, areconserved. Site-directed mutagenesis of the Fn2–Fn2 interactionsincreased the basal phosphorylation of Tie2, suggesting that theFn2–Fn2 interactions may be involved in Tie2 oligomerization inthe absence of ligands rather than in ligand-induced Tie2 activa-tion. There is growing evidence that various transmembrane re-ceptors have a preformed, but inactive, oligomeric structures onthe cell surface, and Tie2 also has been shown to exist as pre-formed oligomers in the absence of ligands (27, 31, 44). Ligandbinding then will induce a conformational change for receptoractivation. The difference between the Fn2–Fn2 interface in ouranalysis and dimer 1 in the paper by Moore et al. (41) could rep-resent such a change, because mutational analysis also implicatesdimer 1 interactions in Ang1-stimulated Tie2 activation. Althoughthe Fn3-mediated dimers in both papers are identical, a lateral shiftof Fn3-mediated homodimers by only a few nanometers is requiredfor the adjustment of the Fn2–Fn2 and dimer 1 interfaces [ProteinData Bank (PDB) ID codes 5MYB and 5UTK].Interestingly, a mutation in the human TEK gene in the se-

quence encoding Tie2 Fn2 domain was discovered in a cohort ofPCG patients who did not carry mutations in other knowndisease-causing genes (Fig. 5B) (38). This mutation, Y611C,occurs in the symmetrical interface between the Fn2 domains ofthe neighboring Tie2 homodimers in our model of Tie2 clus-tering and in the dimer 1 interface in the Tie2 structure byMoore et al. (41). Although the mutant Tie2 was properly lo-calized at the plasma membrane in resting ECs, it did not undergonormal Ang1-stimulated Tie2 clustering or junctional localization(38). Although the nature of the impaired interactions remainselusive, the disease-causing mutation supports the evidence for theFn2-mediated Tie2 clustering. Similarly, in the Eph family tyrosinekinase receptors, the membrane-proximal Fn-like domains areknown to stimulate intermolecular interactions (45).Angiopoietins provide a rare example of soluble ligands that

engage receptor oligomers on the same cell in cis or bridge theirreceptors across cell–cell contacts in trans (10, 18). Ang1 stimula-tion of Tie2 in trans vs. in cis seems to induce partly different sig-naling pathways. Ang1 binding to Tie2 in EC–ECM contactsrepresents the cis association mode, which preferentially acti-vates the Erk and DokR pathways (10, 18). In contacting ECs,both angiopoietin ligands induce Tie2 translocation to cell–celljunctions, but Ang2 induces only weak Tie2 activation (10, 18).

Tie2 association may represent distinct ligand-binding modes intrans vs. in cis. Consistent with our model of Tie2 cis association,several studies indicate that artificial angiopoietin dimers, such asFc-tagged dimers, are inactive as Tie2 ligands (29, 30, 34). How-ever, Oh et al. have reported a specific Tie2-activating Ang1 dimer(CA1-3), which is comprised of the linker and the FLD domains ofAng1 fused to a dimeric Comp domain (46). CA1-3 stimulationshows prominent Tie2 activation in EC–EC junctions, indicatingthat a dimer of Ang1 FLD and the linker domain is capable of FLDdomain presentation in the correct angle for receptor in trans in-teraction (Fig. S7F). Comp-Ang1 stimulates strong Tie2 activationin EC–EC junctions in trans and in the rear of migrating cells in cis,although the pentameric bundle of Comp-Ang1 is only about 10 nmin diameter (10, 34, 35). Therefore, it is likely to represent a ligand-binding mode whereby the multimericity may bridge neighboringTie2 homodimers in cis in addition to Tie2 interactions in trans.Decreased junctional Tie2 phosphorylation upon site-directed

mutagenesis of the Fn3 interface suggests that Tie2 clustering inEC–EC junctions may involve arrays of Tie2 homodimers. Onthe other hand, Tie1 associates with Tie2 in EC–EC junctionsand thereby can regulate the context-dependent differencesin Tie2 signaling during angiogenesis (13, 23, 26). We show thatTie1/Tie2 heterodimerization may involve interactions betweenthe Fn3 domains, but it is not clear how Tie1 interacts with thearrays of Tie2 homodimers in trans or how angiopoietins bridgeTie2 association across EC–EC junctions. The context-dependentdifferences in angiopoietin-activated Tie2 signaling pathways maydepend on differences in other subcellular protein constituents,such as integrins in the ECM contacts and VE-PTP in EC–ECcontacts (10, 13, 19, 47). Further structural and functional in-vestigation is required to understand how Tie1 acts to modulatethe effects of Ang1 and Ang2 on Tie2 and the mechanism ofAng2 antagonism in the Tie2 trans association.Ligand-mediated dimerization of RTKs involves weak homo-

typic interactions between the membrane-proximal domainswhich allow precise positioning of the C-terminal regions and thetransmembrane domains in the correct orientation that enablesactivation of the cytoplasmic tyrosine kinase domains (27, 39, 40,48, 49). We have shown previously that targeting such interac-tions in VEGFR-3 with an antibody that does not block ligandbinding can still block the formation of VEGFR-3 homodimersand VEGFR-3/VEGFR-2 heterodimers, and signaling (50). Sim-ilarly, antibodies targeted to the membrane-proximal domain of

Fig. 5. Model for Tie2 clustering. (A) A cartoonpresentation of the Fn2′–Fn2′′ interaction betweenthe neighboring Tie2 chains. (B) A close-up view ofthe symmetrical interactions between the Fn2 do-mains centered on Arg581. Tyr611 is highlighted inred. Y611C corresponds to the TEK variant in a PCGfamily, which results in haploinsufficiency becauseof protein loss of function (38). (C ) Western blotanalysis of Tie2 phosphorylation in the WT or Fn2-mutant (Gln588Arg, Val612Arg)-transfected HeLacells. Tie2 single-transfected and Tie1/Tie2 double-transfected cells were stimulated with COMP-Ang1(cA1) for 1 h or were left unstimulated. (D) Quanti-tative analysis of C. Error bars indicate SEM. *P ≤0.05; **P ≤ 0.01; ***P ≤ 0.001: ns, not significant;Student’s t test. n = 3. (E) A model for clustering ofAng1 RBD-bound homodimers of Tie2 ECDs basedon the array of the Fn-like homodimers in A and inFig. S7A.

4380 | www.pnas.org/cgi/doi/10.1073/pnas.1616166114 Leppänen et al.

Dow

nloa

ded

by g

uest

on

May

27,

202

1

KIT inhibit its activation (51). Our results here indicate thatligand-induced Tie2 activation also involves homotypic inter-actions of the membrane-proximal domains. Importantly, site-directed mutagenesis of the Tie2 Fn3–Fn3 interactions inhibitedligand-stimulated Tie2 phosphorylation, whereas mutagenesisof the Fn2–Fn2 interactions increased the basal phosphor-ylation of Tie2. Also, site-directed mutagenesis of the Tie1Fn3 interactions in Tie1/Tie2 heterodimerization experimentsincreased the basal Tie1 phosphorylation. The targeting of Tiereceptor membrane-proximal domains thus may provide uniquetherapeutic approaches for the modulation of Tie receptoractivation.

MethodsCrystallization and Structure Determination. Human Tie2 Fn1–3 (residues 443–735) and human Tie1 Fn3 (residues 641–738) were expressed in insect cells,

and the purified proteins were crystallized for structure determination. Fulldetails of protein purification and X-ray and SAXS data collection andanalysis are described in SI Methods.

Cell Culture and Phosphorylation of Tie1 and Tie2. WT and mutant Tie1 andTie2 were cloned into pMXs retroviral expression vector that was used totransduce HeLa cells as described by Saharinen et al. (10). Full details of cellculture and analysis are described in SI Methods.

ACKNOWLEDGMENTS. We thank Tapio Tainola, Seppo Kaijalainen, TanjaLaakkonen, Kirsi Mänttäri, and Jarmo Koponen for excellent technical assistanceand Drs. Kathryn Ferguson andMark Lemmon for valuable comments. This workwas funded by the Jenny and Antti Wihuri Foundation, the Academy of Finland(Grants 307366, 292816, and 273817), the Sigrid Juselius Foundation, the CancerSociety of Finland, the Leducq Foundation (Grant 11CVD03), the Helsinki Uni-versity Hospital’s Government Special State Subsidy for Health Sciences, theNovo Nordisk Foundation, and the Jane and Aatos Erkko Foundation.

1. Chung AS, Ferrara N (2011) Developmental and pathological angiogenesis. Annu RevCell Dev Biol 27:563–584.

2. Koch S, Claesson-Welsh L (2012) Signal transduction by vascular endothelial growthfactor receptors. Cold Spring Harb Perspect Med 2:a006502.

3. Augustin HG, Koh GY, Thurston G, Alitalo K (2009) Control of vascular morphogenesisand homeostasis through the angiopoietin-Tie system. Nat Rev Mol Cell Biol 10:165–177.

4. Saharinen P, Jeltsch M, Santoyo MM, Leppänen VM, Alitalo K (2015) The TIE receptorfamily. Receptor Tyrosine Kinases: Family and Subfamilies, eds Wheeler DE, Yarden Y(Springer International Publishing, Cham, Switzerland), 861:743–775.

5. Saharinen P, Eklund L, Pulkki K, Bono P, Alitalo K (2011) VEGF and angiopoietinsignaling in tumor angiogenesis and metastasis. Trends Mol Med 17:347–362.

6. Suri C, et al. (1996) Requisite role of angiopoietin-1, a ligand for the TIE2 receptor,during embryonic angiogenesis. Cell 87:1171–1180.

7. Thurston G, et al. (2000) Angiopoietin-1 protects the adult vasculature against plasmaleakage. Nat Med 6:460–463.

8. Jeansson M, et al. (2011) Angiopoietin-1 is essential in mouse vasculature duringdevelopment and in response to injury. J Clin Invest 121:2278–2289.

9. Maisonpierre PC, et al. (1997) Angiopoietin-2, a natural antagonist for Tie2 thatdisrupts in vivo angiogenesis. Science 277:55–60.

10. Saharinen P, et al. (2008) Angiopoietins assemble distinct Tie2 signalling complexes inendothelial cell-cell and cell-matrix contacts. Nat Cell Biol 10:527–537.

11. Yuan HT, Khankin EV, Karumanchi SA, Parikh SM (2009) Angiopoietin 2 is a partialagonist/antagonist of Tie2 signaling in the endothelium. Mol Cell Biol 29:2011–2022.

12. D’Amico G, et al. (2014) Tie1 deletion inhibits tumor growth and improves angio-poietin antagonist therapy. J Clin Invest 124:824–834.

13. Korhonen EA, et al. (2016) Tie1 controls angiopoietin function in vascular remodelingand inflammation. J Clin Invest 126:3495–3510.

14. Partanen J, Mäkelä TP, Alitalo R, Lehväslaiho H, Alitalo K (1990) Putative tyrosinekinases expressed in K-562 human leukemia cells. Proc Natl Acad Sci USA 87:8913–8917.

15. Puri MC, Rossant J, Alitalo K, Bernstein A, Partanen J (1995) The receptor tyrosinekinase TIE is required for integrity and survival of vascular endothelial cells. EMBO J14:5884–5891.

16. D’Amico G, et al. (2010) Loss of endothelial Tie1 receptor impairs lymphatic vesseldevelopment-brief report. Arterioscler Thromb Vasc Biol 30:207–209.

17. Woo KV, et al. (2011) Tie1 attenuation reduces murine atherosclerosis in a dose-dependent and shear stress-specific manner. J Clin Invest 121:1624–1635.

18. Fukuhara S, et al. (2008) Differential function of Tie2 at cell-cell contacts and cell-substratum contacts regulated by angiopoietin-1. Nat Cell Biol 10:513–526.

19. Cascone I, Napione L, Maniero F, Serini G, Bussolino F (2005) Stable interaction be-tween alpha5beta1 integrin and Tie2 tyrosine kinase receptor regulates endothelialcell response to Ang-1. J Cell Biol 170:993–1004.

20. Dalton AC, Shlamkovitch T, Papo N, Barton WA (2016) Constitutive association ofTie1 and Tie2 with endothelial integrins is functionally modulated by angiopoietin-1and fibronectin. PLoS One 11:e0163732.

21. Saharinen P, et al. (2005) Multiple angiopoietin recombinant proteins activate theTie1 receptor tyrosine kinase and promote its interaction with Tie2. J Cell Biol 169:239–243.

22. Yuan HT, et al. (2007) Activation of the orphan endothelial receptor Tie1 modifiesTie2-mediated intracellular signaling and cell survival. FASEB J 21:3171–3183.

23. Savant S, et al. (2015) The orphan receptor Tie1 controls angiogenesis and vascular re-modeling by differentially regulating Tie2 in tip and stalk cells. Cell Reports 12:1761–1773.

24. Marron MB, et al. (2007) Regulated proteolytic processing of Tie1 modulates ligandresponsiveness of the receptor-tyrosine kinase Tie2. J Biol Chem 282:30509–30517.

25. Seegar TC, et al. (2010) Tie1-Tie2 interactions mediate functional differences betweenangiopoietin ligands. Mol Cell 37:643–655.

26. Kim M, et al. (2016) Opposing actions of angiopoietin-2 on Tie2 signaling andFOXO1 activation. J Clin Invest 126:3511–3525.

27. Lemmon MA, Schlessinger J (2010) Cell signaling by receptor tyrosine kinases. Cell141:1117–1134.

28. Procopio WN, Pelavin PI, Lee WM, Yeilding NM (1999) Angiopoietin-1 and -2 coiledcoil domains mediate distinct homo-oligomerization patterns, but fibrinogen-likedomains mediate ligand activity. J Biol Chem 274:30196–30201.

29. Davis S, et al. (2003) Angiopoietins have distinct modular domains essential for re-ceptor binding, dimerization and superclustering. Nat Struct Biol 10:38–44.

30. Barton WA, et al. (2006) Crystal structures of the Tie2 receptor ectodomain and theangiopoietin-2-Tie2 complex. Nat Struct Mol Biol 13:524–532.

31. Bogdanovic E, Coombs N, Dumont DJ (2009) Oligomerized Tie2 localizes to clathrin-coated pits in response to angiopoietin-1. Histochem Cell Biol 132:225–237.

32. Partanen J, et al. (1992) A novel endothelial cell surface receptor tyrosine kinasewith extracellular epidermal growth factor homology domains. Mol Cell Biol 12:1698–1707.

33. Yu X, et al. (2013) Structural basis for angiopoietin-1-mediated signaling initiation.Proc Natl Acad Sci USA 110:7205–7210.

34. Cho CH, et al. (2004) COMP-Ang1: A designed angiopoietin-1 variant with nonleakyangiogenic activity. Proc Natl Acad Sci USA 101:5547–5552.

35. Kim KT, et al. (2005) Oligomerization and multimerization are critical for angio-poietin-1 to bind and phosphorylate Tie2. J Biol Chem 280:20126–20131.

36. Fiedler U, et al. (2003) Angiopoietin-1 and angiopoietin-2 share the same bindingdomains in the Tie-2 receptor involving the first Ig-like loop and the epidermalgrowth factor-like repeats. J Biol Chem 278:1721–1727.

37. Macdonald PR, et al. (2006) Structure of the extracellular domain of Tie receptortyrosine kinases and localization of the angiopoietin-binding epitope. J Biol Chem281:28408–28414.

38. Souma T, et al. (2016) Angiopoietin receptor TEK mutations underlie primary con-genital glaucoma with variable expressivity. J Clin Invest 126:2575–2587.

39. Yuzawa S, et al. (2007) Structural basis for activation of the receptor tyrosine kinaseKIT by stem cell factor. Cell 130:323–334.

40. Yang Y, Xie P, Opatowsky Y, Schlessinger J (2010) Direct contacts between extracel-lular membrane-proximal domains are required for VEGF receptor activation and cellsignaling. Proc Natl Acad Sci USA 107:1906–1911.

41. Moore JO, Lemmon MA, Ferguson KM (2017) Dimerization of Tie2 mediated by itsmembrane-proximal FNIII domains. Proc Natl Acad Sci USA 114:4382–4387.

42. Maitra S, Nowic JS (2000) β-Sheet interactions between proteins. The Amide Linkage:Structural Significance in Chemistry, Biochemistry, and Materials Science, edsGreenberg A, Breneman S, Liebman JF (Wiley, New York), pp 495–518.

43. Han S, et al. (2016) Amelioration of sepsis by TIE2 activation-induced vascular pro-tection. Sci Transl Med 8:335ra55.

44. Maruyama IN (2015) Activation of transmembrane cell-surface receptors via a com-mon mechanism? The “rotation model”. BioEssays 37:959–967.

45. Xu K, et al. (2013) Insights into Eph receptor tyrosine kinase activation from crystalstructures of the EphA4 ectodomain and its complex with ephrin-A5. Proc Natl AcadSci USA 110:14634–14639.

46. Oh N, et al. (2015) A designed angiopoietin-1 variant, dimeric CMP-Ang1 activatesTie2 and stimulates angiogenesis and vascular stabilization in N-glycan dependentmanner. Sci Rep 5:15291.

47. Pietilä R, et al. (2012) Ligand oligomerization state controls Tie2 receptor traffickingand angiopoietin-2-specific responses. J Cell Sci 125:2212–2223.

48. Leppänen VM, et al. (2013) Structural and mechanistic insights into VEGF receptor3 ligand binding and activation. Proc Natl Acad Sci USA 110:12960–12965.

49. Arkhipov A, et al. (2013) Architecture and membrane interactions of the EGF re-ceptor. Cell 152:557–569.

50. Tvorogov D, et al. (2010) Effective suppression of vascular network formation bycombination of antibodies blocking VEGFR ligand binding and receptor dimerization.Cancer Cell 18:630–640.

51. Reshetnyak AV, et al. (2013) Structural basis for KIT receptor tyrosine kinase inhibitionby antibodies targeting the D4 membrane-proximal region. Proc Natl Acad Sci USA110:17832–17837.

52. Kabsch W (2010) Xds. Acta Crystallogr D Biol Crystallogr 66:125–132.53. Vonrhein C, Blanc E, Roversi P, Bricogne G (2007) Automated structure solution with

autoSHARP. Methods Mol Biol 364:215–230.54. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-

molecular structure solution. Acta Crystallogr D Biol Crystallogr 66:213–221.55. Petoukhov MV, Svergun DI (2007) Analysis of X-ray and neutron scattering from bi-

omacromolecular solutions. Curr Opin Struct Biol 17:562–571.

Leppänen et al. PNAS | April 25, 2017 | vol. 114 | no. 17 | 4381

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

27,

202

1