Structure and Binding Mechanism of Vascular Endothelial Cadherin: A Divergent Classical Cadherin

doi:10.1016/j.jmb.2011.01.031 J. Mol. Biol. (2011) 408, 57–73

Contents lists available at www.sciencedirect.com

Journal of Molecular Biologyj ourna l homepage: ht tp : / /ees.e lsev ie r.com. jmb

Structure andBindingMechanismof Vascular EndothelialCadherin: A Divergent Classical Cadherin

Julia Brasch1, Oliver J. Harrison1,2, Goran Ahlsen1, Stewart M. Carnally3,Robert M. Henderson3, Barry Honig1,2,4⁎ and Lawrence Shapiro1,5⁎1Department of Biochemistry and Molecular Biophysics, Columbia University, 701 West 168th Street, New York,NY 10032, USA2Howard Hughes Medical Institute, Columbia University, 1130 St. Nicholas Avenue, New York, NY 10032, USA3Department of Pharmacology, University of Cambridge, Tennis Court Road, Cambridge CB2 1PD, UK4Center for Computational Biology and Bioinformatics, Columbia University, 1130 St. Nicholas Avenue, New York,NY 10032, USA5Edward S. Harkness Eye Institute, Columbia University in the City of New York, 635 West 165th Street, New York,NY 10032, USA

Received 30 November 2010;received in revised form11 January 2011;accepted 13 January 2011Available online24 January 2011

Edited by I. Wilson

Keywords:cell–cell adhesion;N-glycosylation;cadherin adhesive binding;domain swapping

*Corresponding authors. B. Honig isNicholas Avenue, New York, NY 10Columbia University, 701 West [email protected] used: VE-cadherin,

N-cadherin, neural cadherin; EM, eltransferase I; AUC, analytical ultraccadherin; AFM, atomic force micros

0022-2836/$ - see front matter © 2011 E

Vascular endothelial cadherin (VE-cadherin), a divergent member of thetype II classical cadherin family of cell adhesion proteins, mediateshomophilic adhesion in the vascular endothelium. Previous investigationswith a bacterially produced protein suggested that VE-cadherin forms cellsurface trimers that bind between apposed cells to form hexamers. Here wereport studies of mammalian-produced VE-cadherin ectodomains suggest-ing that, like other classical cadherins, VE-cadherin forms adhesive transdimers between monomers located on opposing cell surfaces. Trimerizationof the bacterially produced protein appears to be an artifact that arises froma lack of glycosylation. We also present the 2.1-Å-resolution crystalstructure of the VE-cadherin EC1–2 adhesive region, which revealshomodimerization via the strand-swap mechanism common to classicalcadherins. In commonwith type II cadherins, strand-swap binding involvestwo tryptophan anchor residues, but the adhesive interface resembles type Icadherins in that VE-cadherin does not form a large nonswappedhydrophobic surface. Thus, VE-cadherin is an outlier among classicalcadherins, with characteristics of both type I and type II subfamilies.

© 2011 Elsevier Ltd. All rights reserved.

to be contacted at Howard Hughes Medical Institute, Columbia University, 1130 St.032, USA; L. Shapiro, Department of Biochemistry and Molecular Biophysics,h Street, New York, NY 10032, USA. E-mail addresses: [email protected];

vascular endothelial cadherin; EC, extracellular cadherin; BSA, buried surface area;ectron microscopy; HEK, human embryonal kidney; GNTI, N-acetyl-glucosaminylentrifugation; E-cadherin, epithelial cadherin; C-cadherin, Xenopus cleavage stagecopy; PDB, Protein Data Bank.

lsevier Ltd. All rights reserved.

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.jmb.2011.01.031

http://dx.doi.org/

http://www.sciencedirect.com

58 Structure and Binding Mechanism of VE-Cadherin

Introduction

Proteins of the cadherin superfamily are found inboth invertebrate and vertebrate animals and arethought to function primarily in intercellular adhe-sion. More than 350 cadherins have been identified,and these are classified into several distinctsubfamilies.1 The best characterized subfamily isthe vertebrate classical cadherin subfamily, includ-ing type I and type II classical cadherins, which aredefined by distinct sequence and structural charac-teristics.2 In mice and humans, 5 type I cadherinsand 13 type II cadherins have been described.3 Theseare expressed in specific and often overlappingpatterns throughout solid vertebrate tissues, wherethey function in calcium-dependent cell–celladhesion.4 Despite well-defined differences (seethe text below), type I and type II cadherins shareseveral common structural features. They bothcontain an ectodomain region, which is composedof five tandem extracellular cadherin (EC) domains(each of about 110 amino acids) and is rigidified bythe binding of calcium ions between successive ECdomains.5,6 Classical cadherins are anchored by asingle-transmembrane region and have a shortcytoplasmic domain with conserved binding sitesfor β/γ-catenins and p120 catenins,7–11 which helpto mediate attachment to the cytoskeleton and tocontrol cadherin trafficking.12,13

The mechanism by which cadherins mediate celladhesion through the interaction of their ectodo-mains has been intensively studied. It is nowunderstood that both type I and type II cadherinsbind via a three-dimensional domain-swappingmechanisminvolving theirN-termini.2,14,15Membrane-distal cadherin EC1 domains from apposed cellsbind across the intercellular space by a reciprocalexchange of N-terminal β-strands to form “strand-swapped” dimers. The principle of strand exchangeis shared by type I and type II cadherins, but it differsin detail at the molecular level.2 Structures of type Icadherin dimers reveal that residue Trp2 in theswapped strand is docked into a hydrophobic‘acceptor’ pocket in the partner EC1 domain,anchoring the swapped dimer.5,14,16–18 In type IIcadherins, two anchoring tryptophan residues areobserved: Trp2 and Trp4, which are accommodatedby an acceptor pocket that is larger in type IIcadherins than in type I cadherins.2 Type IIcadherins, unlike type I counterparts, also havea large nonswapped region in their adhesiveinterfaces, which is assembled by conservedhydrophobic residues, thus extending the hydro-phobicity of the acceptor pocket towards the baseof the EC1 domain. This region contributes almosta third more buried surface area (BSA) in thedimer interface compared with type I cadherins,which lack this extended hydrophobic interfacesurface.2

Vascular endothelial cadherin (VE-cadherin) isfound exclusively in the cells of the vascularendothelium in vertebrate species, where it isconcentrated in adherens junctions.19–22 Endothelialcells also express type I neural cadherin (N-cadherin)and placental cadherin; however, only VE-cadherinstrongly localizes to adherens junctions in theendothelium, while N-cadherin is dispersed over thecell surface.11,23–26 Gene deletion studies show thatVE-cadherin is essential for embryonic angiogenesis,vascular maintenance, and restoration of vascularintegrity after injury.11,27 VE-cadherin knockout micedie during development on E9.5 due to disintegrationof the vasculature.28 VE-cadherin gene null mutationsin murine embryonic stem cells lead to a dispersedendothelium that lacks an organized vasculature inembryonic bodies.22 Similarly, wild-type adult miceinjected with antibodies directed against VE-cadherindied within 24 h due to disassembly of thevasculature.29,30 These results suggest a specificnonredundant role for VE-cadherin-mediated adhe-sion in the vasculature. VE-cadherin also regulatescellular processes such as leukocyte trafficking31

and control of vascular permeability,32 playing acentral role in the formation of a controlled barrierbetween blood and underlying tissues.Early structural studies on VE-cadherin, using a

bacterially expressed ectodomain fragment com-prising domains EC1–4, identified a cylindricalhexameric structure in cryo-electron microscopy(EM) and a biophysical behavior that are consistentwith hexamer formation.33,34 Results from single-particle EM reconstructions led to a unique modelfor VE-cadherin binding in which three cadherinmolecules assemble on one cell surface to form a cistrimer via EC4 domain interactions. The cis trimerwould then form an EC1-domain-mediated transdimer with a trimer from the opposing cell, resultingin a hexameric structure.We set out to investigate the adhesive mechanism

of VE-cadherin. Biophysical characterization of thecomplete VE-cadherin ectodomain produced inmammalian cells reveals dimers in solution, butnot higher-order oligomers. Furthermore, analysisof nonglycosylated bacterially expressed fragmentsreveals nonbiological multimer formation involvingthe EC3–4 region, which is absent in similarmammalian-expressed fragments with natural gly-cosylation. Crystallographic analysis of the adhesiveEC1–2 domain fragment of chicken VE-cadherinreveals strand-swap binding involving two trypto-phan anchor residues, characteristic of type IIcadherins. However, the adhesive interface alsoresembles type I cadherins in that it lacks a largenonswapped hydrophobic surface. Together, thesefindings show that VE-cadherin forms strand-swapped dimers like other classical cadherins viaan adhesive interface with characteristics of bothtype I and type II subfamilies.

59Structure and Binding Mechanism of VE-Cadherin

Results

Production of natively glycosylated wholeVE-cadherin ectodomains in mammalian cells

Previous work on a bacterially produced humanVE-cadherin EC1–4 fragment revealed the formationof hexameric structures in which two cis trimersappear to associate via strand-swap binding involv-ing each EC1 domain.33–35 Initially, we wished toexamine this novel hexameric structure at highresolution by X-ray crystallography. Since theproteins used in these experiments lacked domainEC5 and native glycosylation, andwere purified frombacterial inclusion bodies, we endeavored to use amore native-like protein in our study. We thereforeused a mammalian expression system to produce thecomplete soluble EC1–5 ectodomain of VE-cadherin,from human (Asp1-Asp542) and chicken (Asp1-Glu545), without the transmembrane and cytoplas-mic regions. These proteins were secreted fromhuman embryonal kidney (HEK) 293F cells, resultingin native proteins thatwere glycosylated and could bepurified from conditioned media.Both human and chicken VE-cadherin ectodo-

mains migrate on SDS polyacrylamide gels∼10 kDaabove their predicted masses, suggesting the pres-ence of significant glycosylation. We determined theprecise molecular masses for these proteins bymatrix-assisted laser desorption/ionization time-of-flight mass spectrometry and calculated themass of glycan by the difference from molecularmasses calculated from amino acid sequence. This

Fig. 1. N-linked glycosylation sites in the VE-cadherin ectoexperimentally as described in the text and are depicted on aEC1–5. Molecular surface is colored by residue type: negativelhydrophobic residues, gray; semipolar residues, yellow; polspheres mark the positions of N-linked glycosylation. (b) Closshowing N-linked glycosylation site Asn395 in the EC4 domaresidues. Residue side chains with solvent-exposed surface acolored as in (a). (c) Protein sequence alignments of VE-cadheconservation of the glycosylation site (N-X-S/T) at Asn395 in

procedure suggested the presence of 9503 Da and13,144 Da of glycan for the chicken and humanproteins, respectively. To assess the contributions ofN-linked and O-linked glycosylation, we thenproduced VE-cadherin in HEK293 N-acetyl-gluco-saminyl transferase I (GNTI)− cells lacking theenzyme GNTI, limiting N-linked glycans to minimalsugar trees of Man5GlcNac2, which can then beremoved by endoglycosidase H treatment. Massspectrometry analysis after treatment with endogly-cosidase H to remove the N-linked glycosylationsuggested that the VE-cadherin ectodomain fromchicken contained 1836 Da of O-linked sugar and7667 Da of N-linked sugar. Similarly, the human VE-cadherin ectodomain appeared to contain 2731 Daof O-linked sugar and 10,336 Da of N-linked sugar.Five sites of N-linked glycosylation were mapped onthe human VE-cadherin ectodomain by mass deter-mination of tryptic peptides: Asn14 and Asn65 inEC1, Asn110 in EC2, Asn315 in EC3, and Asn395 inEC4 (Fig. 1a). No N-linked glycosylation sites werefound in domain EC5.

Natively glycosylated whole VE-cadherinectodomain forms dimers, but nothigher-order multimers

To determine the oligomerization behavior ofhuman and chicken VE-cadherin ectodomains insolution, we used sedimentation equilibrium ana-lytical ultracentrifugation (AUC). Sedimentationequilibrium experiments, which yield accuratemasses independent of molecular shape, revealed a

domain. (a) N-linked glycosylation sites were determinedhomology model of VE-cadherin encompassing domainsy charged residues, red; positively charged residues, blue;ar residues, light blue; aromatic residues, rose. Magentae-up stereo image of the homology model of VE-cadherinin within a region with hydrophobic and aromatic surfacerea greater than 20% are shown as sticks. The surface isrin with type I cadherins, and type II cadherins showingthe EC4 domain.

Table 1. Dissociation constants (Kd) of cadherinectodomain fragments determined by equilibrium AUC

Protein Description Dimer Kd (μM)

VE-cadherinck VE-cadherin EC1–5 Wild type, glycosylated 1.14±0.04h VE-cadherin EC1–5 Wild type, glycosylated 1.03±0.22h VE-cadherin EC1–5 Wild type,

deglycosylatedAggregation

h VE-cadherin EC3–5 Wild type, glycosylated Monomerh VE-cadherin EC3–5 Wild type,

deglycosylatedAggregation

h VE-cadherin EC3–4 Wild type, E. coli 1.93±0.26a

h VE-cadherin EC1–2 Wild type, E. coli 4.38±1.2ck VE-cadherin EC1–2 Wild type, E. coli 1.63±0.19m VE-cadherin EC1–2 Wild type, E. coli 2.22±0.11ck VE-cadherin EC1–2

W2A W4AStrand-swapmutant, E. coli

Monomer

m VE-cadherin EC1–2W2A W4A

Strand-swapmutant, E. coli

Monomer

Type I cadherinsm N-cadherin EC1–2 Wild type, E. coli 25.8±1.5b

m E-cadherin EC1–2 Wild type, E. coli 96.5±10.6b

Type II cadherinsck cadherin-6b EC1–2 Wild type, E. coli 9.2±0.6m cadherin-6 EC1–2 Wild type, E. coli 3.13±0.09c

m cadherin-8 EC1–3 Wild type, E. coli 15±0.4m cadherin-9 EC1–2 Wild type, E. coli 17±1.1m cadherin-10 EC1–2 Wild type, E. coli 42.2±2.7m cadherin-11 EC1–2 Wild type, E. coli 33.8±0.2

a Human VE-cadherin domains EC3–4 were also found toaggregate in some experiments.

b Reported by Katsamba et al.36c Reported by Harrison et al.18


monomer–dimer equilibrium for VE-cadherin fullectodomains with dimerization dissociation con-stants (Kd) of 1.14 μM and 1.03 μM for chicken andhuman, respectively (Table 1, Fig. 2). These valuesare similar to those obtained for two-domain EC1–2fragments, which encompass the adhesive bindingregion (Table 1). No evidence for higher-orderstructures was found in any of the AUC experimentsperformed. A comparison of the Kd value for thedimerization of wild-type human and chicken VE-cadherin EC1–5 fragments with Kd values previous-ly reported for type I cadherins [mouse, human, andchicken epithelial cadherin (E-cadherin) andN-cadherin EC1–2, and frog Xenopus cleavagestage cadherin (C-cadherin) EC1–5] shows thatVE-cadherin dimerization is markedly stronger bybetween 20-fold and 150-fold (Table 1).18,36,37

To further analyze the biophysical behavior ofhuman VE-cadherin ectodomains, we employed size-exclusion chromatography and passed human VE-cadherin EC1–5 over an analytical Superose 6column. The elution profile shows two distinctpeaks, which, given our AUC results, are likely torepresent VE-cadherin dimer and monomer (Fig. 2b).The same experiment was performed with thedouble-tryptophan mutant W2AW4A, in which

strand-swap dimerization is ablated by mutation ofthe docking tryptophans. The mutant elutes as asingle peak with an elution volume corresponding tothat of the suggested monomer peak of the wild type(Fig. 2b).The binding properties of the wild-type and

W2AW4A mutant proteins were also assessed byliposome aggregation experiments. Both proteinsinclude a hexa-histidine tag at the C-terminus,which can bind by affinity to nickel NTA lipidsincorporated into liposomes that act as artificial cellsin the aggregation assay. The human VE-cadherinectodomain EC1–5 aggregates liposomes efficiently,whereas the W2AW4A mutation ablates binding,consistent with analytical size-exclusion chromatog-raphy results (Supplementary Fig. 1).Overall, our solution biophysics experiments

suggest that VE-cadherin ectodomain fragmentsform only dimers in solution with no evidence ofhigher-order structures, behaving identically toother classical cadherin adhesive dimers. Further-more, given that dimerization is ablated by thewell-characterized W2AW4A strand-swappingmutation, VE-cadherin appears to adopt a strand-swapped dimer in common with other classicalcadherins.Although extensive efforts to crystallize full-length

VE-cadherin ectodomain fragments did not yielddiffracting crystals, we were able to image the overallarrangement of human VE-cadherin complexes byatomic force microscopy (AFM). Images of VE-cadherin ectodomain on poly-lysine-coated micasurfaces revealed two distinct reoccurring forms(Fig. 3). A crescent-shaped form (Fig. 3a) with alength of 28±2 nm is strikingly similar to theindividual protomers observed in the publishedC-cadherin ectodomain crystal structure and is thereforelikely to represent the monomer (Fig. 3a).5 A largerstructure with a length of approximately 48±9 nmwas also frequently observed, resembling twocrescent-shaped forms overlapping at their terminiwith a corresponding increase in measured samplethickness at the overlapping region (Fig. 3b). Theoverall shape is reminiscent of the C-cadherin dimerobserved by X-ray crystallography.5,38,39 No struc-tures similar to the cylindrical hexameric structuresreported for the bacterially expressed EC1–4 frag-ment of VE-cadherin33 were observed. These datasuggest that full-length native VE-cadherin ectodo-mains form strand-swapped dimers similar to otherclassical cadherins.

The EC3–4 fragment forms multimers only whenglycosylation is absent

The hexameric cryo-EM structures reported forthe bacterially expressed VE-cadherin EC1–4 frag-ment suggested the presence of a cis trimer interfacein or near the EC4 domain.33 To investigate the

Fig. 2. Biophysical characterization of the VE-cadherin ectodomain. (a) Sedimentation equilibrium AUC profilesshowing dimerization for human VE-cadherin EC1–5 and ideal monomeric behavior for the truncated EC3–5 fragment.Dissociation constants (Kd) from the analyses are listed in Table 1; calculated molecular masses are given in the plots. (b)Comparison of elution profiles from analytical size-exclusion experiments with human VE-cadherin EC1–5 (blue),double-tryptophan mutant W2AW4A (green), and EC3–5 (orange) at a concentration of 0.5 mg/mL. VE-cadherin EC1–5elutes as two peaks (a major peak of higher molecular size and a minor peak representing a smaller species), whereas forthe strand-swapped mutant W2AW4A, only the lower molecular size peak is observed. The EC3–5 fragment elutes as onepeak with the smallest molecular size of the three proteins. The void volume of the column is approximately 8 mL.


biological relevance of this finding, we producedhuman VE-cadherin ectodomain fragments encom-passing the predicted trimer interface region usingboth bacterial and mammalian expression systems.A VE-cadherin EC3–4 fragment produced in bacte-ria and thus lacking glycosylation behaved as adimer in size-exclusion chromatography and either

as a strong dimer with a Kd of 1.9 μM or as anisodesmic aggregate in AUC (Table 1). Althoughthis protein did not associate trimerically, it isknown that proteins with nonspecific hydrophobicassociations can sometimes form complexes withunexpected numbers of protomers.40 We then usedthe mammalian HEK293F cell system to produce a

image of Fig. 2

Fig. 3. AFM imaging of the fullectodomains of VE-cadherin depos-ited on poly-L-lysine mica revealsmonomer and dimer forms. (a)Recurring crescent forms with alength of 28±2 nm. These resembleC-cadherin ectodomain protomersin crystal structures (PDB code1L3W), depicted in orange surfacerepresentation at the bottom right.(b) Second recurring form in whichtwo crescent-shaped forms overlapat their tips with an overall lengthof 48±9 nm. These resemble transdimer structures observed in thecrystal structure of C-cadherin (bot-tom right). The color scale indicatessample thickness above the micasurface. The scale bar for all imagesrepresents 35 nm.


similar but fully glycosylated protein, human VE-cadherin EC3–5. We included the EC5 domain inthis construct, since expression of the EC3–4construct was poor in the HEK293 system. Remark-ably, the fully glycosylated EC3–5 protein behavesas a monomer in size-exclusion chromatography,liposome aggregation assays, and equilibrium AUC(Fig. 2; Supplementary Fig. 1). This result suggestedthat the multimeric association of the bacterialprotein and the high-affinity EC3–4 dimerizationarose from its lack of glycosylation. To test this idea,we produced VE-cadherin EC3–5 in GNTI− cells,which yield minimal sugars at N-linked glycosyla-tion sites that can be removed by treatment withendoglycosidase H. Consistent with our hypothesis,this minimally glycosylated protein associated asdimer and isodesmic aggregate in AUC experimentsand was able to efficiently aggregate liposomes(Table 1, Fig. 2a; Supplementary Fig. 1). WhenN-linked glycan was removed from this proteinwith endoglycosidase H, it behaved similarly. Theseresults strongly suggest that the trimeric associationpreviously observed in cryo-EM studies of thebacterially produced EC1–4 fragment represents anartifact arising from the lack of glycosylation.

Mapping of the glycosylation sites identified inFig. 1a onto the molecular surface of a VE-cadherinEC4 homology model reveals that the singleN-linked glycan site in the domain is centered ona molecular surface dominated by hydrophobicresidues (Fig. 1b). Furthermore, sequence compar-isons show that this glycosylation site is conservedamong classical cadherins (Fig. 1c). It is notunexpected that removal of the protruding hydro-philic glycan moiety to reveal a large hydrophobicsurface could result in nonspecific associations.Thus, structural data support the idea that trimer-ization of VE-cadherin in the EC4 domain, aspreviously proposed, is likely to be artifactual,supported by the high affinity for the EC3–4 proteinassociation in comparison to the glycosylatedcounterpart monomers.

Biophysical characterization of VE-cadherin EC1–2reveals strand-swap-dependent homodimerization

Since adhesive interfaces of classical cadherinshave been localized to the EC1 domain andadhesive binding is strongly affected by Ca2+

binding sites between the EC1 domain and the

image of Fig. 3

Table 2. Crystallographic data and refinement statistics

Data collectionSpace group P43212Cell dimensions a, b, c (Å); α, β, γ (°) 99.97, 99.97, 105.99; 90, 90,

90Resolution (Å) 80–2.1Rmerge 13.7 (41.8)I/σI 1395.2/64.4 (339.1/48.2)Completeness 100 (100)Redundancy 14.1 (14.0)Observed reflections 450,866Unique reflections 31,991

RefinementResolution (Å) 72.7–2.1Number of reflections 30,319Rwork 18.3Rfree 24.2Number of atoms 3841

Protein 3247Ion 6Water 588

RMSDBond length (Å) 0.016Bond angles (°) 1.59

Mean B-factors (Å2)Protein 19.4Ion 13.6Water 32.1

Ramachandran plotOutliers (%) 0Favored (%) 97.1


EC2 domain, we produced EC1–2 fragments ofVE-cadherins from human, mouse, and chicken.We used equilibrium AUC analysis to assess themultimerization of these proteins. The resultsreveal a monomer–dimer equilibrium for eachprotein with dimerization dissociation constants(Kd) of between 1.63 μM and 4.38 μM, values thatare very similar to those reported for the full-length EC1–5 constructs (Table 1).To test if VE-cadherin dimerization is ablated by –

2 proteins. AUC analysis shows that these mutantsfail to dimerize, consistent with the strand-swapbinding mode found for all classical cadherins testedto date (Table 1).The binding affinities determined for VE-

cadherin are at least 1 order of magnitude strongerthan those for type I cadherins characterizedpreviously.18,36,37 To address the question ofwhether higher-affinity binding is a property ofVE-cadherin or type II cadherins in general, weperformed sedimentation equilibrium analysis onthe wild-type EC1–2 fragments of mouse cadherin-9, cadherin-10, and cadherin-11, and on an EC1–3fragment of cadherin-8 (Table 1). The Kd values forthe dimerization of type II cadherins, including thepreviously reported value for cadherin-6,18 rangefrom 3.13 μM to 42.2 μM, suggesting that,consistent with the larger size of their bindinginterfaces, type II cadherin dimerization is, ingeneral, stronger than type I cadherin dimerization.Nevertheless, significant variation is found withinthe type II cadherin family. VE-cadherin exhibitsadhesive binding among the strongest of the type IIcadherins.

Crystal structure of chicken VE-cadherin EC1–2reveals a strand-swapped dimer

To investigate the structural features of theadhesive interface of VE-cadherin, we determinedthe crystal structure of chicken VE-cadherindomains EC1–2 (Asp1-Asp203 of the mature pro-tein) at 2.1 Å resolution. Data collection andrefinement statistics are summarized in Table 2.Two VE-cadherin molecules are present in the

crystallographic asymmetric unit, each adopting anoverall structure similar to that previously reportedfor the structures of other type I and type II classicalcadherins.2,5,17,18 Two tandem EC domains, eachadopting a seven-stranded β-barrel fold, areconnected via a short linker region (Fig. 4a). Threecalcium ions are coordinated in this interdomainregion by the side chains of residues Glu11, Glu12,Asp62, and Glu64 from the EC1 domain; Asp96,Asn98, and Asp99 in the linker region; and Asp129,Asp132 and Asp184 from the EC2 domain (Fig. 4a).These residues are completely conserved withintype II classical cadherins and are also conserved intype I cadherins, with the exception of Glu12, which

in type I family members is replaced by anasparagine that does not directly coordinate calci-um. In addition, the backbone carbonyl groups ofresidues Ile97, Asn100, and His139, as well as onewater molecule, contribute to the coordination ofcalcium in the VE-cadherin structure, precisely asobserved in other type II cadherins (Fig. 4a).2

The two VE-cadherin molecules in the crystallo-graphic asymmetric unit form strand-swappeddimers (Fig. 4b), as observed for other classicalcadherins. The dimer interface is symmetrical andformed between EC1 domains, which are orientedapproximately perpendicular relative to each otherand reciprocally swap a segment of the N-terminalβ-strand, designated the A⁎-strand. The Trp2 andTrp4 residues in the A⁎-strand serve as anchors thatare docked into a hydrophobic ‘acceptor’ pocket inthe EC1 domain of the partner protomer. Thetryptophan side chains partake in van der Waalsinteractions with hydrophobic residues lining theacceptor pocket: Leu24, Ala73, and Phe90 at thebase, and Tyr35 and Ile75 towards the top of thepocket (Fig. 4c).Additionally, Trp2 and Trp4 engage in two

intermolecular hydrogen bonds: one between theε1 nitrogen of the Trp2 side chain and the backbonecarbonyl of Pro86, and the other between the ε1nitrogen of the Trp4 side chain and the side chain ofSer88. While the hydrogen bond involving Trp2 has

Fig. 4. Crystal structure of the EC1–2 domain of chicken VE-cadherin showing a strand-swapped cadherin dimer. (a)Single protomer of VE-cadherin EC1–2 shown as a ribbon diagram, with calcium ions as green spheres. Boxed is a detailedview of calcium coordination in the linker region between EC1 and EC2. Three calcium ions (green spheres) arecoordinated by side chains of Asp, Glu, or Asn, and by backbone carbonyl groups (stick representation), in addition to onewater molecule (gray sphere). Coordinating interactions are shown as magenta dotted lines. (b) Ribbon diagram in tworotations showing the symmetrical strand-swapped dimer of VE-cadherin. One protomer is shown in orange, and itsbinding partner is shown in blue. Trp2 and Trp4 residues in the A-strand are exchanged between protomers in the dimerand are shown in stick representation. The dimer interaction involves only the EC1 domains. (c) Detailed view of thestrand dimer interface. The docking of the A-strand of one protomer (orange) into the pocket of the partnering molecule(blue) is mediated by the following: van der Waals interactions between the tryptophan side chains and Leu24, Tyr35,Ala73, Ile75, and Phe90 in the acceptor pocket; two hydrogen bonds between the ε nitrogen of Trp2 and Trp4, the backbonecarbonyl of Pro86, and the hydroxyl group of Ser88, respectively; and a salt bridge between the N-terminal amine group ofAsp1 and the carboxylate side chain of Glu85. Residues that mediate binding are shown in stick representation; hydrogenbonds are depicted as magenta dotted lines; and salt-bridge interactions are depicted as green dotted lines.


been observed in other cadherin structures, the Trp4hydrogen bond with Ser88 appears to be specific toVE-cadherin. The N-terminal amino group of theswapped A⁎-strand is held in place via a salt bridgewith the side chain of Glu85, a conserved interactionamong classical cadherins.VE-cadherin has been classified as a type II family

member, since intron/exon boundaries are similarto the genes of other type II cadherins, and it has twotryptophan anchor residues and a short prodomaincharacteristic of the type II subfamily.41,42 However,phylogenetic analysis indicates that it is an outlierwithin this group.42 Notably, mouse type II cadher-ins—cadherin-6, cadherin-8, cadherin-9, cadherin-10, and cadherin-11—share 65–84% amino acididentities within their EC1–2 regions, whereas VE-cadherin, in comparison, is only 43–47% identicalwith these more representative type II familymembers (Supplementary Table 1). To determinehow this sequence divergence impacts the structure

of the VE-cadherin adhesive region, we superposedour VE-cadherin structure with the EC1–2 adhesiveregion structures available for type I and type IIcadherins (Fig. 5). The overall structures, includingthe angle between the EC1 domain and the EC2domain, appear to be quite similar. The RMSDs are1.52 Å between superposed VE-cadherin and E-cadherin over 187 Cα atom pairs, and 1.86 Åbetween VE-cadherin and cadherin-11 over 191 Cα

atoms. When comparisons are restricted to EC1domains, VE-cadherin aligns more closely with typeII cadherins than with type I cadherins, and notablylacks the quasi-β-helix between strand C and strandD found in all type I cadherins but absent in type IIcadherins (Supplementary Table 2). However, typeII cadherin-8, cadherin-11, and motor neuroncadherin align more closely with each other thanwith VE-cadherin, indicating that VE-cadherin is, tosome extent, a structural outlier within the group(Supplementary Table 2).

image of Fig. 4

Fig. 5. Superposed α-carbon traces from the crystal structures of VE-cadherin and type I and type II classical cadherins.Single protomer (left) and dimer (right) superpositions of chicken VE-cadherin EC1–2 (blue) with E-cadherin, which is atype I cadherin (red, upper panel; PDB code 2QVF), and with cadherin-11, which is a type II cadherin (orange, lower leftpanel; PDB code 2A4E). The structures are represented as α-carbon traces. Note that the overall structure and interdomainangles are closely similar between the three different classical cadherins, but the relative orientation of EC2 domains in theVE-cadherin dimer differs remarkably from that in the dimers of the other cadherins. The angle between EC2 domains inVE-cadherin is 168°, compared to 129° for E-cadherin and 143° for cadherin-11.


Unique features of the VE-cadherinbinding interface

Despite a protomer structure that is similar overallto other classical cadherins, VE-cadherin showsremarkable differences in the arrangement of itsadhesive interface (Fig. 5, right). The VE-cadherindimer appears to be almost linear, with an angle ofapproximately 168° between the long axes of theopposing EC2 domains, whereas the E-cadherin andcadherin-11 dimers assemble with correspondingangles off 129° and 143°, respectively (Fig. 5, right).BSAs for type I cadherin adhesive interfacesaverage ∼850 Å2 per protomer, whereas the areafor type II cadherin protomers is significantlylarger (∼1250 Å2) (Supplementary Table 3). Inter-estingly, the VE-cadherin adhesive interface buries1067 Å2 per protomer, which is intermediatebetween type I and type II cadherins. Figure 6ashows the footprint of the adhesive interface forVE-cadherin and representative type I and type IIcadherins in a molecular surface representation.For the type I subfamily member E-cadherin, theadhesive interface is restricted to the “upper” halfof the EC1 domain around the swapped strandand the acceptor pocket. However, the interface oftype II cadherins, represented in Fig. 6 by

cadherin-11, extends over the whole face of theEC1 domain, owing to the participation of anextended region of nonswapped hydrophobic inter-action surface at the base of the EC1 domain, whichis restricted to type II cadherins (Figs. 6b and 7).2

Remarkably, the surface involved in the VE-cadherin strand-swapped dimer more closelyresembles that of type I cadherins, since theextended hydrophobic region is absent (Figs. 6band 7). The corresponding region in VE-cadherindoes not contribute to the dimer and remains solventexposed.The residues of VE-cadherin that map to the

nonswapped hydrophobic interface of more typicaltype II cadherins are distinct in character (Fig. 7a). Intypical type II cadherins, hydrophobic residues atpositions 8, 10, 13, 19, and 20 in the A and B strandspack together in the dimer interface (Figs. 6b and7a).2 These are conserved in character in all type IIcadherins, except for VE-cadherin, which, like type Icadherins, has hydrophilic residues at most of thesepositions (Fig. 7a). These differences are likely tounderlie the more type-I-like dimer interface con-figuration found for VE-cadherin.Overall, the structure of the VE-cadherin adhesive

dimer shows that VE-cadherin is an outlier amongthe classical cadherins. While interactions between

image of Fig. 5

Fig. 6. Comparison of the strand-swapped dimer interface of VE-cadherin with those of type I and type II cadherins. (a)EC1 domains of the single protomers of E-cadherin (left; PDB code 2QVF), cadherin-11 (middle; PDB code 2A4C), and VE-cadherin (right) are displayed as a molecular surface (orange). The “footprint” of the interaction partner in the strand-swapped dimer, determined as regions with a per-residue BSA in the dimer of N10 Å2, is shown in blue on the surface. Theinterface for type I cadherins (left) and VE-cadherin (right) is restricted to a smaller region compared to that in type IIcadherins (middle), in which the interface not only is localized around the A-strand but extends toward the base of theEC1 domain. (b) Ribbon diagrams of the EC1 domains of the same dimers as in (a), with residues involved in the interface(BSAN10 Å2) shown as sticks. One protomer is shown in orange, and the other protomer is shown in blue. Note that theadhesive interface of VE-cadherin shares the features of both subfamilies of classical cadherins.


the swapped A⁎-strand and the acceptor pocket arealmost identical between VE-cadherin and othertype II cadherins, the VE-cadherin interface moreclosely resembles type I cadherins outside of theswapping region.

Discussion

We employed the biophysical characterization ofnatively glycosylated VE-cadherin EC1–5 to test anovelmodel for the adhesive binding of VE-cadherinthat differed substantially from the known struc-tures of other classical cadherins. This model, basedon solution biophysics and cryo-EM analysis of abacterially produced VE-cadherin EC1–4 ectodo-main fragment, suggested that VE-cadherin formscis cell surface trimers via an interface in EC4, whichengage in EC1-mediated trans interactions betweencells to form hexamers.33–35,43,44 By contrast, a widearray of evidence for other classical cadherins

suggests that they form adhesive bonds throughstrand-swap interactions between monomers. It isthought that adherent dimers then assemble to formadherens junctions via weak cis interactions, whichhave recently been identified for type I cadherins,5,45

but strong cis interactions like the trimer proposedfor VE-cadherins have not been observed. While theknown cadherin strand-swap mechanism and theputative hexamermodel both involve EC1-mediatedtrans dimers that are broadly compatible with thestrand-swapped structure of chicken VE-cadherinEC1–2 reported here, they differ with respect to thepresence of strong cis trimer interactions in EC3–4.Our biophysical studies showed that “naked”bacterially produced EC3–4 exhibits a high affinityfor association, and minimally glycosylated EC3–5fragments of VE-cadherin aggregate liposomesthrough homophilic interactions. These findingsprovide an explanation for why a trimeric interfaceinvolving EC4, a phenomenon prevented by nativeglycosylation, could be observed in previous studies

image of Fig. 6

Fig. 7. VE-cadherin uses a set of residues for trans dimerization different from that used by type II cadherins. (a) Sequence alignment of the EC1 domains of chickenVE-cadherin and MN-cadherin (PDB code 1ZVN), and of mouse cadherin-8 (PDB code 1ZXK) and cadherin-11 (PDB code 2A4C), for which strand-swapped dimerstructures have been determined. Residues buried in the strand-swap interface (N5% buried) in all structures are boxed in purple. Residues buried in dimers ofcadherin-8, cadherin-11, andMN-cadherin, but solvent exposed in VE-cadherin, which comprise the type II cadherin specific hydrophobic interface, are boxed in green.(b) Comparison of BSA values for EC1 domain residues in the strand-swapped dimer interfaces of VE-cadherin and type II cadherins as a percentage of the residuesurface area. VE-cadherin shares the buried A-strand and acceptor pocket residues with type II cadherins (purple boxes in the upper and lower panels), but residues inthe type II cadherin hydrophobic interface (positions 10, 13, 19, 20, 21, and 22) are more than 95% solvent exposed in the VE-cadherin dimer (green boxes in the upperand lower panels).

67Structure

andBinding

Mechanism

ofVE-C

adherin

image of Fig. 7


using bacterially expressed constructs of VE-cadherin.33,34,35 In agreement with our data, Ahrenset al. found mammalian-expressed VE-cadherinectodomains to also associate only via their N-termini in rotary shadowing EM experiments.38 Thebiophysical experiments reported here clearly showthat VE-cadherin, although a phylogenetic outlier,shares an adhesive mechanism similar to that ofother classical cadherins.We tested human VE-cadherin EC1–5 domain

constructs, which were expressed in HEK293F cellsand purified from conditioned media, in equilibri-um AUC and size-exclusion chromatographyexperiments. In all cases, we observed exclusivelya monomer–dimer equilibrium with no evidence ofhexamer formation. By contrast, nonglycosylatedbacterially produced constructs of the EC1–4region behave as multimers.33–35 In our experi-ments, bacterially expressed EC3–4 regions, encom-passing the putative trimerization region, behaveas dimers or aggregate. Furthermore, whereasisolated natively glycosylated VE-cadherin EC3–5fragments were monomeric in solution, partiallyglycosylated or deglycosylated protein alsoshowedmultimer/aggregate behavior. The partiallyglycosylated EC3–5 fragments also showed bindingactivity in liposome aggregation assays, whereasequivalent natively glycosylated fragments did not.These results strongly suggest that the trimerizationobserved previously arises from the lack of glyco-sylation inherent to bacterial protein production,which exposes a hydrophobic surface on EC4 thatwould be shielded in the natural protein by aconserved N-linked glycan. Furthermore, previousstudies showed that the native glycan of humanVE-cadherin carries negatively charged sialic acid inalmost all branches that might contribute to repul-sion mediated by the glycan.46While our data exclude strong cis interactions

such as those observed for bacterially expressedVE-cadherin EC1–4 fragments,33,34,35 other potentialcis interactions cannot be excluded and couldcontribute to adhesive junction formation. Such cisinteractions, which function in the context ofmembrane-bound molecules, are often weak andcan be undetectable by standard solution biophysicsmethods. Examples ofweak cis interactions that havebeen characterized include type I cadherins5,45 andEph/ephrin complexes.47,48 In type I cadherins, acis-oriented interaction common to the crystal struc-tures of C-cadherin, E-cadherin, and N-cadherin hasrecently been shown to facilitate the assembly ofadherens junctions in concert with the well-characterized trans interaction.5,45 VE-cadherin isalso found in adherens junctions,49 but the cisinteraction observed in type I cadherins is not presentin the chicken VE-cadherin EC1–2 crystal lattice,possibly because it lacks the pseudo-β-helix, astructural element that contributes to the cis interface

in type I cadherins. It remains to be determinedwhether VE-cadherin employs an alternative cisinterface between neighboring proteins.Here we have also presented the 2.1-Å-resolution

crystal structure of the adhesive dimer of chickenVE-cadherin EC1–2. The structure shows that VE-cadherin adopts a strand-swapped dimer stabilizedby intermolecular docking of Trp2 and Trp4residues, as observed for the other more typicaltype II cadherins—cadherin-8, cadherin-11, andMN-cadherin.2 However, VE-cadherin lacks thenonswapped hydrophobic interaction region of theadhesive interface; thus, the overall arrangement ofthe VE-cadherin adhesive dimer is more similar tothat of type I cadherins. The absence of this contactregion likely arises from the substitution of hydro-philic residues in VE-cadherin for hydrophobicresidues near the base of EC1 in the more typicaltype II cadherins.It has previously been suggested that the larger

strand dimer interface in type II cadherins contrib-uted by the nonswapped hydrophobic region mayenhance binding affinities in comparison to those oftype I cadherins.36 Prior to our study, Kd valuesfrom an AUC analysis of type II cadherin dimeriza-tion were available only for EC1–2 fragments ofmouse cadherin-6.18 We report here the Kd valuesfor the homodimerization of five further type IIcadherins: mouse cadherin-8, cadherin-9, cadherin-10, and cadherin-11, and chicken cadherin-6b, inaddition to VE-cadherin from several species. TheKd values determined span a range from 1.6 μM to42.2 μM and, in general, indicate stronger adhesivebinding than for type I cadherins (19.7–156 μM forN-cadherin, E-cadherin, and C-cadherin36,37,45). It issurprising, however, that VE-cadherin is the stron-gest of these despite having a smaller BSA in thedimer interface. Other factors, such as strain in theclosed monomer form, must therefore play animportant role in the determination of homodimer-ization affinity.50,51 The high affinity of VE-cadherinbinding interactions, in comparison with otherclassical cadherins, may reflect its unique role inmaintaining intercellular adhesion in the vascularendothelium, which is exposed to substantialhemodynamic stresses.52

Classical cadherins often display homophilicdimerization but, at least within a subtype (e.g.,type I cadherins), can bind heterophilically as well.36Domain-shuffling experiments have shown that thedeterminants of cellular specificity reside in the EC1domains.2,36,53–55 Cell aggregation studies haveshown that cells expressing individual type IIcadherins, while displaying homophilic selectivityin general, bind to each other promiscuously todifferent degrees.53 Cells expressing E-cadherin andN-cadherin have been shown to display a qualita-tively similar behavior;36 however, consistent withthe differing structures of their adhesive regions,


cross-interactions between type I and type IIcadherins have not been observed either at themolecular level36 or at the cellular level.2,53 Accord-ingly, it has been demonstrated that transfected cellsexpressing VE-cadherin do not heterophilicallyinteract with transfected cells expressing type IN-cadherin.56 Furthermore, VE-cadherin excludesN-cadherin from adherens junctions in vascularendothelial cells, in which the two cadherins arecoexpressed in vivo.11,24–26 The differences that weobserve between the dimer interfaces of VE-cadherin and other classical cadherins, includingthe previously studied type II family members,suggest that VE-cadherin is unlikely to bind hetero-philically to other classical cadherin molecules dueto the incompatibility of the binding interfaces theypresent. The structural divergence of VE-cadherin,which has a critical nonredundant role in theassembly of the vascular endothelium, is thereforelikely to result in a high degree of homophilicspecificity, since cadherins with similar structuralfeatures have not been identified.

Materials and Methods

Bacterial protein production

Coding sequences for mouse, chicken, and human VE-cadherin EC1–2; mouse cadherin-8 EC1–3; and mousecadherin-9, cadherin-10, and cadherin-11 EC1–2 wereamplified by PCR from cDNA libraries (Clontech) (chickenVE-cadherin, clone pgp1n.pk007.i4; Delaware Biotechnol-ogy Institute) and cloned in-frame with an N-terminalhexa-histidine-tagged SUMO protein into the BamHI/NotI sites of the pSMT3 vector. Cleavage of SUMO fusionproteins with Ulp1 (ubiquitin-like protease 1) after a Gly-Gly motif yields cadherin proteins with native N-termini.Extra amino acids occurring due to cloning after thecleavage site were removed using the QuikChange site-directed mutagenesis kit (Stratagene) to ensure the nativeN-termini of all proteins used in our studies, unlessspecifically altered. We introduced the W2AW4A muta-tion using the same method.For protein expression, Escherichia coli Rosetta 2(DE3)

pLysS cells (Novagen) were transformed with preparedpSMT3 vector and grown at 37 °C until an OD600 of 0.6had been reached, with shaking at 200 rpm. To induceprotein expression, we added 100 μM IPTG and loweredthe temperature to 18 °C. After 18 h, cells were harvestedby centrifugation at 6000g for 15 min. Pelleted bacteriawere resuspended in lysis buffer [500 mM NaCl, 20 mMTris–Cl (pH 8.0), 20 mM imidazole (pH 8.0), and 3 mMCaCl2] and lysed for 6 min by sonication in 15 s intervalswith a 45 s rest in between pulsing. Cell debris was spundown at 4 °C and 20,000g for 1 h, and His-tagged proteinswere extracted from the clear lysate by flowing overnickel-charged IMAC Sepharose 6 Fast Flow resin (GEHealthcare). Beads were subsequently washed with40 column volumes of lysis buffer to remove contami-nants, and His–SUMO fusion proteins were eluted with250 mM imidazole containing lysis buffer. The 6His–

SUMO tag of fusion proteins was cut enzymatically byadding Ulp1 to a final concentration of 2 μg/mL to theelution. Proteins were dialyzed into a low-ionic-strengthbuffer [100 mM NaCl, 20 mM Tris–Cl (pH 8.0), and 3 mMCaCl2]. For cadherin-8, cadherin-9, cadherin-10, andcadherin-11, we removed cleaved 6His–SUMO tags anduncut fusion protein on nickel-charged IMAC resinequilibrated in dialysis buffer. The cadherins were furtherpurified by anion-exchange chromatography (Mono Q10/10 HR; GE Healthcare) and size-exclusion chromatog-raphy (HiLoad 26/60 Superdex™ S75 prep-grade; GEHealthcare), leaving them in a final buffer of 150 mMNaCl, 20 mM Tris–Cl (pH 8.0), and 3 mM CaCl2. Proteinswere concentrated to a final concentration of approxi-mately 5–10 mg/mL using Amicon Spin concentrators(Millipore) and flash frozen.

Mammalian protein production

Coding sequences for human or chicken VE-cadherindomains EC1–5 (Asp1-Asp542 and Asp1-Glu545 of thenative protein, respectively) or the coding sequence forhuman VE-cadherin EC3–5 (Ile204-Asp542) were ampli-fied by PCR from cDNA libraries. All—preceded by aKozak sequence and the signal sequence for human CD33(MPLLLLWAGALA), with the transmembrane domainand intracellular domain replaced by a hexa-histidine tagat the C-terminus—were cloned into the episomalexpression vector pCEP4 (Invitrogen) using the restrictionsites KpnI/NotI. For protein expression, HEK293F cells orHEK293 GNTI− cells were transfected in six-well stageusing Lipofectamine 2000 (Invitrogen) according to themanufacturer's manual. During continuous selection with200 μM hygromycin B (Cellgro) for stable expression,proteins were secreted and purified from 4 L of collectedDulbecco's modified Eagle's medium/F12 medium sup-plemented with 10% newborn calf serum, 1:100 penicillin/streptomycin (Gibco), 1:50 L-glutamine (Gibco), and200 μM hygromycin B (Cellgro). After the addition of500 mM sodium chloride, 20 mM Tris–Cl (pH 8.0), 3 mMcalcium chloride, and 10 mM imidazole (pH 8.0) to thesupernatant, His-tagged proteins were extracted from themedium by batch-affinity purification by adding 20 mL ofnickel-charged IMAC resin (GE Healthcare) preequili-brated in 500 mM sodium chloride, 20 mM Tris–Cl(pH 8.0), 3 mM calcium chloride, and 10 mM imidazole(pH 8.0). After 3 h of incubation, the resin was collectedfrom the supernatant after passing through Kontescolumns and washed with 20 column volumes of bindingbuffer and 20 column volumes of the same buffer withincreased imidazole concentration (12.5 mM). Proteinswere eluted with 5 column volumes of 75 mM imidazolein the same buffer. The eluted protein was dialyzedovernight at 4 °C into a low-ionic-strength buffer of100 mM sodium chloride, 20 mM Tris–Cl (pH 8.0), and3 mM CaCl2. Further purification by flowing throughMono S HR 10/10 (GE Healthcare) andMono QHR 10/10(GE Healthcare) ion-exchange columns and by size-exclusion chromatography with HiLoad 26/60 Super-dex™ S200 prep-grade (GE Healthcare) leaves proteins ina buffer of 150 mM NaCl, 10 mM Tris–Cl (pH 8.0), and3 mM CaCl2. Proteins were concentrated and flash frozenat a concentration of approximately 7 mg/mL. Proteinswere N-terminally sequenced to ensure the proper


cleavage of the artificial signal sequence, in turn ensuringnative N-termini.

Analytical ultracentrifugation

Equilibrium AUC experiments were performed using aBeckman XLA/I ultracentrifuge with a Ti50An or Ti60Anrotor. Prior to each experiment, all proteins were dilutedwith buffer [150 mM NaCl, 10 mM Tris–Cl (pH 8.0), and3 mM CaCl2] and dialyzed for 16 h at 4 °C in the samebuffer. One hundred twenty microliters of proteins atthree different concentrations (0.7 mg/mL, 0.46 mg/mL,and 0.24 mg/mL) was loaded into six-channel equilibriumcells with parallel sides and sapphire windows. Weperformed all experiments at 25 °C and collected data atwavelengths of 280 nm (UV) and 660 nm (interference).Five-domain proteins were spun for 20 h at 8800g and 4scans (1 h−1) were collected; speed was increased to12,300g for 10 h and 4 scans (1 h−1) were collected; andspeed was increased to the highest speed of 16,400g foranother 10 h and 4 scans (1 h−1) were taken, yielding 72scans per sample. Three-domain proteins were analyzedusing the same protocol, except for the use of 14,200g,23,500g, and 35,200g, respectively. For two-domainproteins, 23,500g, 35,200g, and 49,100gwere used. Relativecentrifugal force are given at the measuring cell center at aradius of 65 mm. We calculated the buffer density andprotein V-bars using the program SednTerp (AllianceProtein Laboratories), and analyzed the retrieved datausing HeteroAnalysis 1.1.0.28†. We fitted data from allconcentrations and speeds globally by nonlinear regres-sion to either a monomer–dimer equilibrium model or anideal monomer model. All experiments were performed atleast in duplicate.

Analytical size-exclusion chromatography

Analytical size-exclusion chromatography was per-formed at 4 °C using a Superose 12 10/20 column (GEHealthcare). After the equilibration of the column in buffer[150 mMNaCl, 10 mM Tris–Cl (pH 8.0), and 3 mM CaCl2]at a flow rate of 0.3 mL/min, 150 μL of purified protein ata concentration of 6.7 μM was injected and eluted at thesame flow rate while the UV signal at 280 nm wasmonitored. We collected 0.5 mL fractions for SDS-PAGEanalysis with subsequent immunological detection (Tetra-His antibody; Qiagen) (anti-mouse antibody with horse-radish peroxidase; Santa Cruz Biotechnology) against theC-terminal hexa-histidine tag common to all proteins usedin this experiment. Void volume was determined by BlueDextran (GE Healthcare) to be 8 mL.

Liposome aggregation assay

We prepared liposomes in a 9:1 molar ratio of 1,2-dioleoyl-sn-glycero-3-phospohocholine to nickel(II) 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)imi-nodiacetic acid) succinyl]) using hydration and extrusion.Lipids were prepared according to the manufacturer's

†http://www.biotech.uconn.edu/auf

instructions (Avanti Lipids), suspended in liposomeaggregation buffer [25 mM Hepes (pH 7.4), 0.1 M KCl,10% (vol/vol) glycerol, and 3 mM CaCl2], and passedthrough a 100 nm filter membrane to yield liposomes ofequal size.The conduct of aggregation assays was based on the

method described by Harrison et al.45 and He et al.57

Liposomes at a final concentration of ∼5 mg/mL andhexa-histidine-tagged protein at a concentration of0.5 mg/mL were mixed at starting point 0, andaggregation was monitored every 20 s by changes inlight scattering at OD650 for a total time of 40 min. Asnegative control, a sample containing only liposomes andbuffer was used. Experiments were performed in tripli-cate for VE-cadherin EC1–5 and in duplicate for EC3–5proteins.

AFM imaging

AFM sample pucks were prepared as follows: 10-mm-diameter disks of muscovite mica (Agar Scientific) werestamped out from as-received sheets. Each was fixed to a13 mm steel puck (Agar Scientific) with cyanoacrylatesuperglue and allowed to dry for 16 h. Poly-L-lysine(Sigma-Aldrich) was diluted 1:100 into biotechnologyperformance certified-grade water (Sigma-Aldrich), and45 μL was pipetted onto freshly cleaved mica, incubatedfor 30 min at room temperature, washed 10× with 1 mL ofBPC-grade water, and dried under a gentle and steadystream of nitrogen.Samples were diluted in 150 mM NaCl, 20 mM Tris–Cl

(pH 8.0), and 3 mM CaCl2 to a final concentration of 2 nMand stored at 4 °C, which yielded workable and consistentsurface concentrations of protein. Forty-five microliters ofthe sample was pipetted onto poly-L-lysine-coated micaand incubated at room temperature for 10 min, thenwashed 10× with 1 mL of BPC-grade water and driedunder a gentle and steady stream of dry nitrogen.Imaging was performed by a multimode atomic force

microscope with an attached Nanoscope IIIa controller(both from Veeco) under ambient conditions in tappingmode, using OMCL0AC160TS single-crystal siliconprobes (Olympus, Japan) with a resonant frequency of∼300 kHz, a nominal spring constant of ∼42 nm, and aradius of curvature of b10 nm. Images were collected at3 Hz with an integral gain of 0.2, a proportional gain of0.35, a look-ahead gain of 0, and a set point of ∼0.85 (tominimize vertical probe-induced sample deformation).Length determination of molecules was performed usingSPIP (Scanning Probe Image Processor), version 3.3, fromthree-dimensional AFM data using full width at half-maximum.

Crystallization and structure determination

We expressed and purified chicken VE-cadherin EC1–2as described above and used it for crystallization studies at8.6 mg/mL. Using the vapor diffusion method, we grewprotein crystals at 20 °C after combining 1 μL of proteinsolution with 1 μL of well solution composed of 18% (wt/vol) polyethylene glycol 8000, 200 mM calcium acetate,and 100 mM sodium cacodylate (pH 6.5). The crystalsgrew within 20 h and were flash frozen in liquid nitrogen

http://www.biotech.uconn.edu/auf


after being immersed briefly in cryoprotectant [30% (wt/vol) glycerol, 18% (wt/vol) polyethylene glycol 8000,200 mM calcium acetate, and 100 mM sodium cacodylate(pH 6.5)].Data on a single frozen crystal were collected at the X4C

beamline of the National Synchrotron Light Source,Brookhaven National Laboratory, at a wavelength of0.979 Å. We processed the data using the HKL suite58 andsolved the structure by molecular replacement usingmouse cadherin-11 [Protein Data Bank (PDB) code2A4E] as search model. Refinement was carried out bymanual building in Coot,59 followed by automatedrefinement in Refmac.60 The Ramachandran plot statisticsfor the final model are as follows: 96.1% in favoredregions, 100% in allowed regions, and 0% in disallowedregions. Data collection and refinement statistics aresummarized in Table 2.

Accession number

Coordinates and structure factors for chicken VE-cadherin EC1–2 domains have been deposited in thePDB under accession number 3PPE.

Acknowledgements

This work was supported by US National Insti-tutes of Health grant R01 GM062270 (L.S.) and USNational Science Foundation grant MCB-0918535(B.H.). X-ray data were acquired at the X4C beamlineof theNational Synchrotron Light Source, BrookhavenNational Laboratory. The X4 beamlines are operatedby the New York Structural Biology Center.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2011.01.031

References

1. Hulpiau, P. & van Roy, F. (2009). Molecular evolutionof the cadherin superfamily. Int. J. Biochem. Cell Biol.41, 349–369.

2. Patel, S. D., Ciatto, C., Chen, C. P., Bahna, F.,Rajebhosale, M., Arkus, N. et al. (2006). Type IIcadherin ectodomain structures: implications forclassical cadherin specificity. Cell, 124, 1255–1268.

3. Posy, S., Shapiro, L. & Honig, B. (2008). Sequenceand structural determinants of strand swapping incadherin domains: do all cadherins bind throughthe same adhesive interface? J. Mol. Biol. 378,954–968.

4. Takeichi, M., Hatta, K., Nose, A. & Nagafuchi, A.(1988). Identification of a gene family of cadherin celladhesion molecules. Cell Differ. Dev. 25, 91–94.

5. Boggon, T. J., Murray, J., Chappuis-Flament, S., Wong,E., Gumbiner, B. M. & Shapiro, L. (2002). C-cadherinectodomain structure and implications for cell adhe-sion mechanisms. Science, 296, 1308–1313.

6. Nagar, B., Overduin, M., Ikura, M. & Rini, J. M. (1996).Structural basis of calcium-induced E-cadherin rigid-ification and dimerization. Nature, 380, 360–364.

7. Ozawa, M., Baribault, H. & Kemler, R. (1989). Thecytoplasmic domain of the cell adhesion moleculeuvomorulin associates with three independent pro-teins structurally related in different species. EMBO J.8, 1711–1717.

8. Lampugnani, M. G., Corada, M., Caveda, L., Bre-viario, F., Ayalon, O., Geiger, B. & Dejana, E. (1995).The molecular organization of endothelial cell to celljunctions: differential association of plakoglobin,beta-catenin, and alpha-catenin with vascular endo-thelial cadherin (VE-cadherin). J. Cell Biol. 129,203–217.

9. Huber, A. H. & Weis, W. I. (2001). The structure of thebeta-catenin/E-cadherin complex and the molecularbasis of diverse ligand recognition by beta-catenin.Cell, 105, 391–402.

10. Ishiyama, N., Lee, S. H., Liu, S., Li, G. Y., Smith, M. J.,Reichardt, L. F. & Ikura, M. (2010). Dynamic and staticinteractions between p120 catenin and E-cadherinregulate the stability of cell–cell adhesion. Cell, 141,117–128.

11. Gentil-Dit-Maurin, A., Oun, S., Almagro, S., Bouillot,S., Courcon, M., Linnepe, R. et al. (2010). Unravelingthe distinct distributions of VE- and N-cadherins inendothelial cells: a key role for p120-catenin. Exp. CellRes. 316, 2587–2599.

12. Liu, H., Komiya, S., Shimizu, M., Fukunaga, Y. &Nagafuchi, A. (2007). Involvement of p120 carboxy-terminal domain in cadherin trafficking. Cell Struct.Funct. 32, 127–137.

13. Reynolds, A. B. & Carnahan, R. H. (2004). Regulationof cadherin stability and turnover by p120ctn:implications in disease and cancer. Semin. Cell Dev.Biol. 15, 657–663.

14. Shapiro, L., Fannon, A. M., Kwong, P. D., Thompson,A., Lehmann, M. S., Grubel, G. et al. (1995). Structuralbasis of cell–cell adhesion by cadherins. Nature, 374,327–337.

15. Chen, C. P., Posy, S., Ben-Shaul, A., Shapiro, L. &Honig, B. H. (2005). Specificity of cell–cell adhesion byclassical cadherins: critical role for low-affinity dimer-ization through beta-strand swapping. Proc. NatlAcad. Sci. USA, 102, 8531–8536.

16. Parisini, E., Higgins, J. M., Liu, J. H., Brenner, M. B. &Wang, J. H. (2007). The crystal structure of human E-cadherin domains 1 and 2, and comparison with othercadherins in the context of adhesion mechanism. J.Mol. Biol. 373, 401–411.

17. Haussinger, D., Ahrens, T., Aberle, T., Engel, J.,Stetefeld, J. & Grzesiek, S. (2004). Proteolytic E-cadherin activation followed by solution NMR andX-ray crystallography. EMBO J. 23, 1699–1708.

18. Harrison, O. J., Bahna, F., Katsamba, P. S., Jin, X.,Brasch, J., Vendome, J. et al. (2010). Two-step adhesivebinding by classical cadherins. Nat. Struct. Mol. Biol.17, 348–357.

19. Breier, G., Breviario, F., Caveda, L., Berthier, R.,




Schnurch, H., Gotsch, U. et al. (1996). Molecularcloning and expression of murine vascular endothe-lial-cadherin in early stage development of cardiovas-cular system. Blood, 87, 630–641.

20. Dejana, E., Zanetti, A. & Del Maschio, A. (1996).Adhesive proteins at endothelial cell-to-cell junc-tions and leukocyte extravasation. Haemostasis, 26,210–219.

21. Lampugnani, M. G., Resnati, M., Raiteri, M., Pigott,R., Pisacane, A., Houen, G. et al. (1992). A novelendothelial-specific membrane protein is a marker ofcell–cell contacts. J. Cell Biol. 118, 1511–1522.

22. Vittet, D., Buchou, T., Schweitzer, A., Dejana, E. &Huber, P. (1997). Targeted null-mutation in thevascular endothelial-cadherin gene impairs the orga-nization of vascular-like structures in embryoidbodies. Proc. Natl Acad. Sci. USA, 94, 6273–6278.

23. Liaw, C. W., Cannon, C., Power, M. D., Kiboneka,P. K. & Rubin, L. L. (1990). Identification and cloningof two species of cadherins in bovine endothelial cells.EMBO J. 9, 2701–2708.

24. Salomon, D., Ayalon, O., Patel-King, R., Hynes, R. O.& Geiger, B. (1992). Extrajunctional distribution of N-cadherin in cultured human endothelial cells. J. CellSci. 102, 7–17.

25. Navarro, P., Ruco, L. & Dejana, E. (1998). Differentiallocalization of VE- and N-cadherins in humanendothelial cells: VE-cadherin competes with N-cadherin for junctional localization. J. Cell Biol. 140,1475–1484.

26. Jaggi, M., Wheelock, M. J. & Johnson, K. R. (2002).Differential displacement of classical cadherins by VE-cadherin. Cell Commun. Adhes. 9, 103–115.

27. Gulino, D., Delachanal, E., Concord, E., Genoux, Y.,Morand, B., Valiron, M. O. et al. (1998). Alteration ofendothelial cell monolayer integrity triggers resynthe-sis of vascular endothelium cadherin. J. Biol. Chem.273, 29786–29793.

28. Carmeliet, P., Lampugnani, M. G., Moons, L., Bre-viario, F., Compernolle, V., Bono, F. et al. (1999).Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediatedendothelial survival and angiogenesis. Cell, 98,147–157.

29. Corada, M., Mariotti, M., Thurston, G., Smith, K.,Kunkel, R., Brockhaus, M. et al. (1999). Vascularendothelial-cadherin is an important determinant ofmicrovascular integrity in vivo. Proc. Natl Acad. Sci.USA, 96, 9815–9820.

30. May, C., Doody, J. F., Abdullah, R., Balderes, P., Xu,X., Chen, C. P. et al. (2005). Identification of atransiently exposed VE-cadherin epitope that allowsfor specific targeting of an antibody to the tumorneovasculature. Blood, 105, 4337–4344.

31. Navarro, P., Caveda, L., Breviario, F., Mandoteanu, I.,Lampugnani, M. G. & Dejana, E. (1995). Catenin-dependent and -independent functions of vascularendothelial cadherin. J. Biol. Chem. 270, 30965–30972.

32. Corada, M., Liao, F., Lindgren, M., Lampugnani,M. G., Breviario, F., Frank, R. et al. (2001). Monoclonalantibodies directed to different regions of vascularendothelial cadherin extracellular domain affect ad-hesion and clustering of the protein and modulateendothelial permeability. Blood, 97, 1679–1684.

33. Hewat, E. A., Durmort, C., Jacquamet, L., Concord, E.& Gulino-Debrac, D. (2007). Architecture of the VE-cadherin hexamer. J. Mol. Biol. 365, 744–751.

34. Legrand, P., Bibert, S., Jaquinod, M., Ebel, C., Hewat,E., Vincent, F. et al. (2001). Self-assembly of thevascular endothelial cadherin ectodomain in a Ca2+-dependent hexameric structure. J. Biol. Chem. 276,3581–3588.

35. Bibert, S., Jaquinod, M., Concord, E., Ebel, C., Hewat,E., Vanbelle, C. et al. (2002). Synergy betweenextracellular modules of vascular endothelial cadherinpromotes homotypic hexameric interactions. J. Biol.Chem. 277, 12790–12801.

36. Katsamba, P., Carroll, K., Ahlsen, G., Bahna, F.,Vendome, J., Posy, S. et al. (2009). Linking molecu-lar affinity and cellular specificity in cadherin-mediated adhesion. Proc. Natl Acad. Sci. USA, 106,11594–11599.

37. Chappuis-Flament, S., Wong, E., Hicks, L. D., Kay, C.M. & Gumbiner, B. M. (2001). Multiple cadherinextracellular repeats mediate homophilic binding andadhesion. J. Cell Biol. 154, 231–243.

38. Ahrens, T., Lambert, M., Pertz, O., Sasaki, T.,Schulthess, T., Mege, R. M. et al. (2003). Homoassocia-tion of VE-cadherin follows a mechanism common to“classical” cadherins. J. Mol. Biol. 325, 733–742.

39. Tomschy, A., Fauser, C., Landwehr, R. & Engel, J.(1996). Homophilic adhesion of E-cadherin occurs bya co-operative two-step interaction of N-terminaldomains. EMBO J. 15, 3507–3514.

40. Weis, W. I., Kahn, R., Fourme, R., Drickamer, K. &Hendrickson, W. A. (1991). Structure of the calcium-dependent lectin domain from a rat mannose-bindingprotein determined by MAD phasing. Science, 254,1608–1615.

41. Suzuki, S., Sano, K. & Tanihara, H. (1991). Diversity ofthe cadherin family: evidence for eight new cadherinsin nervous tissue. Cell Regul. 2, 261–270.

42. Nollet, F., Kools, P. & van Roy, F. (2000). Phylogeneticanalysis of the cadherin superfamily allows identifi-cation of six major subfamilies besides several solitarymembers. J. Mol. Biol. 299, 551–572.

43. Al-Kurdi, R., Gulino-Debrac, D., Martel, L., Legrand,J. F., Renault, A., Hewat, E. & Venien-Bryan, C. (2004).A soluble VE-cadherin fragment forms 2D arrays ofdimers upon binding to a lipid monolayer. J. Mol. Biol.337, 881–892.

44. Lambert, O., Taveau, J. C., Him, J. L., Al Kurdi, R.,Gulino-Debrac, D. & Brisson, A. (2005). The basicframework of VE-cadherin junctions revealed by cryo-EM. J. Mol. Biol. 346, 1193–1196.

45. Harrison, O. J., Jin, X., Hong, S., Bahna, F., Ahlsen, G.,Brasch, J. et al. (2011). The extracellular architecture ofadherens junctions revealed by crystal structures oftype I cadherins. Structure, 19, 244–256.

46. Geyer, H., Geyer, R., Odenthal-Schnittler, M. &Schnittler, H. J. (1999). Characterization of humanvascular endothelial cadherin glycans. Glycobiology, 9,915–925.

47. Seiradake, E., Harlos, K., Sutton, G., Aricescu, A. R. &Jones, E. Y. (2010). An extracellular steric seedingmechanism for Eph–ephrin signaling platform assem-bly. Nat. Struct. Mol. Biol. 17, 398–402.

48. Himanen, J. P., Yermekbayeva, L., Janes, P. W.,


Walker, J. R., Xu, K., Atapattu, L. et al. (2010).Architecture of Eph receptor clusters. Proc. NatlAcad. Sci. USA, 107, 10860–10865.

49. Uehara, K. (2006). Distribution of adherens junctionmediated by VE-cadherin complex in rat spleen sinusendothelial cells. Cell Tissue Res. 323, 417–424.

50. Vendome, J., Posy, S., Jin, X., Bahna, F., Ahlsen, G.,Shapiro, L. & Honig, B. (2010). Molecular designprinciples underlying affinity and specificity in classi-cal cadherins. Nat. Struct. Mol. Biol. doi:10.1038/nsmb.2051.

51. Bennett, M. J., Schlunegger, M. P. & Eisenberg, D.(1995). 3D domain swapping: a mechanism foroligomer assembly. Protein Sci. 4, 2455–2468.

52. Baumgartner, W. & Drenckhahn, D. (2002). Plasma-lemmal concentration and affinity of mouse vascularendothelial cadherin, VE-cadherin. Eur. Biophys. J. 31,532–538.

53. Shimoyama, Y., Takeda, H., Yoshihara, S., Kitajima,M. & Hirohashi, S. (1999). Biochemical characteriza-tion and functional analysis of two type II classiccadherins, cadherin-6 and -14, and comparison withE-cadherin. J. Biol. Chem. 274, 11987–11994.

54. Nose, A., Tsuji, K. & Takeichi, M. (1990). Localization

of specificity determining sites in cadherin celladhesion molecules. Cell, 61, 147–155.

55. Shan, W. S., Tanaka, H., Phillips, G. R., Arndt, K.,Yoshida, M., Colman, D. R. & Shapiro, L. (2000).Functional cis-heterodimers of N- and R-cadherins. J.Cell Biol. 148, 579–590.

56. Breviario, F., Caveda, L., Corada, M., Martin-Padura,I., Navarro, P., Golay, J. et al. (1995). Functionalproperties of human vascular endothelial cadherin(7B4/cadherin-5), an endothelium-specific cadherin.Arterioscler. Thromb. Vasc. Biol. 15, 1229–1239.

57. He, Y., Jensen, G. J. & Bjorkman, P. J. (2009). Cryo-electron tomography of homophilic adhesion mediat-ed by the neural cell adhesion molecule L1. Structure,17, 460–471.

58. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode.Macromol. Crystallogr. Part A, 276, 307–326.

59. Emsley, P. & Cowtan, K. (2004). Coot: model-buildingtools for molecular graphics. Acta Crystallogr. Sect. D,60, 2126–2132.

60. Bailey, S. (1994). The CCP4 suite—programs forprotein crystallography. Acta Crystallogr. Sect. D, 50,760–763.

http://dx.doi.org/10.1038/nsmb.2051

http://dx.doi.org/10.1038/nsmb.2051

Documents

Structure and Binding Mechanism of Vascular Endothelial Cadherin: A Divergent Classical Cadherin