Structural basis of adhesive binding by desmocollins and desmogleins Oliver J. Harrisona,b,1, Julia Braschc,1, Gorka Lassoa,d, Phinikoula S. Katsambaa,b, Goran Ahlsena,b, Barry Honiga,b,c,d,e,f,2, and Lawrence Shapiroa,c,f,2 a
Department of Systems Biology, Columbia University, New York, NY 10032; bHoward Hughes Medical Institute, Columbia University, New York, NY 10032; Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY 10032; dCenter for Computational Biology and Bioinformatics, Columbia University, New York, NY 10032; eDepartment of Medicine, Columbia University, New York, NY 10032; and fZuckerman Mind Brain Behavior Institute, Columbia University, New York, NY 10032 c
Contributed by Barry Honig, April 23, 2016 (sent for review January 21, 2016; reviewed by Steven C. Almo and Dimitar B. Nikolov)
Desmosomes are intercellular adhesive junctions that impart strength to vertebrate tissues. Their dense, ordered intercellular attachments are formed by desmogleins (Dsgs) and desmocollins (Dscs), but the nature of trans-cellular interactions between these specialized cadherins is unclear. Here, using solution biophysics and coated-bead aggregation experiments, we demonstrate family-wise heterophilic specificity: All Dsgs form adhesive dimers with all Dscs, with affinities characteristic of each Dsg:Dsc pair. Crystal structures of ectodomains from Dsg2 and Dsg3 and from Dsc1 and Dsc2 show binding through a strand-swap mechanism similar to that of homophilic classical cadherins. However, conserved charged amino acids inhibit Dsg:Dsg and Dsc:Dsc interactions by same-charge repulsion and promote heterophilic Dsg:Dsc interactions through oppositecharge attraction. These findings show that Dsg:Dsc heterodimers represent the fundamental adhesive unit of desmosomes and provide a structural framework for understanding desmosome assembly. desmosome
| cell adhesion | cadherin | heterophilic binding
D
esmosomes are spot-weld-like intercellular adhesive junctions that link the intermediate filament networks of adjacent cells to impart strength to the solid tissues of vertebrates (1–3). Dysfunction of desmosomes in inherited and acquired human diseases as well as in mouse genetic ablation studies causes characteristic defects in heart muscle and skin (3–5), demonstrating their importance in tissues that undergo mechanical stress. In electron micrographs, the hallmarks of mature desmosomes include a constant intermembrane distance of 280–340 Å, and apparently ordered electron-density in the intercellular space, often with a discrete midline connected by periodic cross-bridges to the cell membranes (6–9). The intercellular attachments of desmosomes are composed of transmembrane proteins from two specialized cadherin subfamilies: desmocollins (Dscs) and desmogleins (Dsgs). The human genome encodes three Dsc (Dsc1– Dsc3) and four Dsg (Dsg1–Dsg4) proteins, which share an overall domain organization comprising four to five extracellular cadherin (EC) domains, a single-pass transmembrane region, and an intracellular domain that binds to intermediate filaments via adaptor proteins desmoplakin and plakoglobin (1). Individual Dsgs and Dscs show differential expression patterns: Dsg2 and Dsc2 are expressed widely in all desmosome-forming tissues (1), whereas other desmosomal cadherins are expressed specifically in stratified epithelia with graded, overlapping patterns (1, 10). Notably, both Dscs and Dsgs appear necessary for adhesion in transfected cells (1, 11–13), and loss of either in genetic experiments causes loss of normal desmosomal adhesion (5, 14, 15). Although the ultrastructure of desmosomes is well characterized, a molecular-level understanding of the binding interactions between desmosomal cadherin extracellular regions that assemble these junctions has remained elusive. In particular, whether desmosomal cadherins have homophilic preferences or whether interactions occur between heterophilic pairs has been a matter of dispute (1, 11–13, 16– 18). Electron tomography studies of native desmosomes (7, 8) have revealed cadherins binding through their EC1 domains at the central 7160–7165 | PNAS | June 28, 2016 | vol. 113 | no. 26
dense midline, consistent with a strand-swap mode of interaction first characterized for classical cadherins (19, 20). Nevertheless, the identity of Dscs and Dsgs in these tomographic reconstructions could not be determined, and atomic-resolution structures of desmosomal cadherins have not been available, with the exception of an NMR structure of a monomeric EC1 fragment of mouse Dsg2 with an artificially extended N terminus (PDB ID code 2YQG). In addition, a small-angle X-ray scattering (SAXS) study showed a classical cadherin-like monomeric molecular envelope for mouse Dsg2, but with additional flexibility in the membrane-proximal region (21). Here we present crystal structures of ectodomains from human Dsgs and Dscs, along with solution biophysical analysis and coated-bead aggregation studies. Our results identify the central importance of heterophilic binding between Dscs and Dsgs and reveal the structural basis of trans-cellular adhesive binding between desmosomal cadherins. Results Heterophilic Adhesive Binding Between Dsgs and Dscs. Using a transient-transfection HEK-293 cell protein expression system (Materials and Methods), we produced full-length ectodomains for all seven human desmosomal cadherins Dsg1–Dsg4 and Dsc1–Dsc3. We first assessed the homo-association state of each of these proteins by sedimentation-equilibrium analytical ultracentrifugation (AUC) (SI
Significance Desmosomes are crucial for the integrity of tissues that undergo mechanical stress. Their intercellular attachments are assembled from desmogleins (Dsgs) and desmocollins (Dscs), two families of specialized cadherins whose structures and interactions have remained uncharacterized. Our study demonstrates family-wise heterophilic interactions between these proteins, with all Dsgs forming adhesive dimers with all Dscs. Crystal structures of ectodomains from Dsg2 and Dsg3 and from Dsc1 and Dsc2 show binding through a strand-swap mechanism similar to that of classical cadherins, which we show underlie heterophilic interactions. Conserved compatibly charged amino acids in the interfaces promote heterophilic Dsg:Dsc interactions. We show that Dsg:Dsc heterodimers represent the fundamental adhesive unit of desmosomes and provide a structural framework for understanding the extracellular assembly of desmosomes. Author contributions: O.J.H., J.B., B.H., and L.S. designed research; O.J.H., J.B., G.L., P.S.K., and G.A. performed research; O.J.H., J.B., G.L., P.S.K., G.A., B.H., and L.S. analyzed data; and O.J.H., J.B., B.H., and L.S. wrote the paper. Reviewers: S.C.A., Albert Einstein College of Medicine; and D.B.N., Memorial Sloan-Kettering Cancer Center. The authors declare no conflict of interest. Data deposition: The atomic coordinates have been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 5ERD, 5EQX, 5IRY, 5ERP, and 5J5J). 1
O.J.H. and J.B. contributed equally to this work.
2
To whom correspondence may be addressed. Email: [email protected] or [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1606272113/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1606272113
Appendix, Table S1). Each of the four Dsgs and Dsc2 and Dsc3 appeared to be either monomers or very weak dimers in solution with KD values higher than 400 μM, weaker than the weakest-binding classical cadherin (E-cadherin, KD = 156 μM; ref. 22). Only Dsc1 showed appreciable homodimerization, but this appeared to be a result of technical issues involving the formation of nonnative intermolecular disulfides and was not observed in other binding assays (see following text and Materials and Methods). We then used surface plasmon resonance (SPR) to measure all pairwise homophilic and heterophilic interactions among the seven human desmosomal cadherins (Fig. 1 A–C). Consistent with the AUC results, no significant homophilic interactions were observed, and in addition, no significant binding within families (Dsg:Dsg or Dsc:Dsc) was detected (Fig. 1A). In contrast, all experiments in which a Dsg was passed over a Dsc surface or a Dsc was passed over a Dsg surface showed binding responses indicative of strong heterophilic interaction (Fig. 1A). Optimal fits of SPR data were obtained with a 1:1 binding model (SI Appendix, Fig. S1), yielding heterophilic KDs ranging between 3.6 and 43.9 μM (Fig. 1C), similar to the affinities of other tight-binding cell adhesion molecules (22–28). Interestingly, the strongest-binding pairs (Dsg1:Dsc1 and Dsg4:Dsc1) correspond to those expressed in the outermost layers of human skin, and the weakest (Dsg3:Dsc3) to those restricted to basal layers (1, 10), consistent with a potential role for differential affinities in maintenance of cell arrangement in stratified epithelia (1). Dsgs and Dscs both conserve sequence features in EC1 critical to the strand-swap mechanism of adhesive binding in classical cadherins (20, 29), suggesting strand-swapping mediates heterophilic Dsg:Dsc interactions. Consistent with this mechanism, deletion of domain EC1 in Dsgs and Dscs abolished all heterophilic interactions, whereas deletion of the C-terminal EC4–EC5 domains had little effect on binding (SI Appendix, Fig. S2A). Furthermore, combined mutation of Trp2 (W2A), whose indole side chain anchors the swapping N-terminal β-strand (the A*-strand), and Ala80/82 [A80I (Dsc2) or A82I (Dsg2)], whose small side chain permits Trp2 docking in the dimer-partner acceptor pocket, ablated all heterophilic binding of Dsg2 and Dsc2 (Fig. 1B). Interestingly, Dsg2 containing only a W2A mutation on the swapping strand could still mediate strong heterophilic binding to Dscs, whereas an A82I acceptorpocket mutation ablated all binding (SI Appendix, Fig. S2B). This was reversed in Dsc2, where the W2A mutation was more severe. These mutants suggest that Dsg:Dsc interactions are asymmetric, with docking of the Dsc A*-strand in the Dsg acceptor pocket representing the dominant interaction.
A
To assess their adhesive properties on apposed surfaces, we coupled C-terminally biotinylated Dsg or Dsc ectodomains to NeutrAvidin-coated beads and monitored their aggregation by fluorescence microscopy (Fig. 2). As expected from the SPR results (Fig. 1A), robust aggregation was observed when beads coated with Dsgs were mixed with beads coated with Dscs, and no aggregation was observed for any Dsg:Dsg or Dsc:Dsc pair. Because Dscs and Dsgs were immobilized on separate surfaces, the bead aggregation assays confirm that Dsg:Dsc heterophilic interactions can form in a trans-orientation between apposed surfaces analogous to apposed cell membranes in desmosomes. Aggregation was dependent on the presence of Ca2+, and identical results were obtained when Dsg2 and Dsc2 were coated together on the same beads (SI Appendix, Fig. S3). Together, the AUC, SPR, and bead aggregation data show that Dsgs and Dscs form strong strand-swapped adhesive trans interactions with family-wise heterophilic specificity, which are likely to underlie intercellular adhesion in desmosomes. Overall Structures of Dsgs and Dscs. We determined crystal structures for full-length glycosylated ectodomains of human Dsg2, Dsg3, and Dsc1 (Fig. 3) with diffraction limits between 2.9 and 3.1Å (SI Appendix, Table S2). Diffracting crystals of human Dsc2 ectodomains were not obtained, but we determined crystal structures of an EC2−EC5 fragment (Dsc2ΔEC1) at 2.7 Å resolution and of the adhesive EC1 domain in the context of a chimeric protein Dsc2EC1Dsg2EC2–EC5 at 3.3Å resolution (Fig. 3 and SI Appendix, Table S2). Each desmosomal cadherin extracellular region has an overall structure comprising an elongated, curved arrangement of tandem EC domains connected by calcium-binding linker regions (Fig. 3). Strand-swapped homodimers, formed between EC1 domains, were observed for Dsg2, Dsc1, and Dsc2EC1Dsg2EC2-5 (Fig. 3). Evidence of extensive N- and O-linked glycosylation was also observed in the electron density maps. Notably, there are conserved, subfamily-specific N-linked glycosylation sites in the adhesive EC1 domains, although none are involved in adhesive contacts, and contiguous regions of conserved O-linked glycosylation in EC4 (Figs. 3–5 and SI Appendix, Fig. S4), corresponding to the novel mono-Omannosylation sites previously identified in Dsc2 and other cadherins by mass spectrometry (30, 31). Because inherited mutations in Dsg2 and Dsc2 are associated with weakened heart muscle desmosomes in arrhythmogenic right-ventricular cardio-myopathy (4, 32), we mapped pathogenic missense mutations associated with the disease onto the Dsg2 and Dsc2 ectodomain structures (SI Appendix, Fig. S5). The mutations were found to target structural elements including calcium-binding residues, glycosylation sites, and disulfide bonds in all
B KD [ M]
Dsg1
Dsg2
Dsg3
Dsg4
Dsc1
3.55 ±0.01
11.52 ±0.02
8.07 ±0.03
7.9 ±0.1
Dsc2
21.9 ±0.2
29.0 ±0.2
37.0 ±0.2
22.5 ±0.2
Dsc3
18.1 ±0.3
34.4 ±0.6
43.9 ±0.6
32.5 ±0.4
Fig. 1. Heterophilic interactions between Dsgs and Dscs in SPR. (A) Wild-type Dsg1–Dsg4, Dsc1–Dsc3, or (B) double-mutant Dsg2 W2A A82I and Dsc2 W2A A80I analytes (columns) were tested at concentrations of 3, 6, and 12 μM (12 μM omitted for Dsg4) over surfaces of wild-type Dsg2 (top row), Dsg3, Dsc1, Dsc2, and Dsc3 (bottom row). Responses were normalized for analyte molecular weight differences and are drawn to the same scale across rows. (C) Heterophilic binding affinities (KD) from fitting of SPR data to 1:1 interaction models (SI Appendix, Fig. S1). ±Errors of the fit.
Harrison et al.
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7161
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
C
Dsg1
Dsg2
Dsg3
Dsc1
the presence of all three canonical Ca2+ ions in the EC3–EC4 linker of Dsc1 and Dsc2, whereas only two Ca2+ ions were observed in Dsg2 and Dsg3, which lacked Ca3 that is normally coordinated primarily by residues in EC4 (Fig. 4A). Sequence analyses (Fig. 4C and ref. 21) show that all Dsgs lack two of the canonical Ca2+-binding residues in EC4 corresponding to Asn333 and Gln371 (N-cadherin numbering), explaining the lack of Ca3 coordination and suggesting Dsg1 and Dsg4 will adopt similarly bent conformations. In Dscs, most Ca2+-binding ligands are conserved; however, residues equivalent to Gln371 and Asp420, which coordinate Ca3 in classical cadherins, are positioned far from Ca3 in the Dsc structures (Fig. 4 A–C). In addition to the structural role of the EC3– EC4 bend, it is possible that these sites represent loci of flexibility, consistent with recent SAXS analysis of mouse Dsg2 (21), which could be important for the mechanical resilience of desmosomes.
Dsc3
Dsc2
Dsg1
Dsg2
Dsg3
Dsc1
Dsc2
Homodimer Interfaces Reveal Determinants of Heterophilic Specificity.
Three of the crystal structures reported here (Dsg2, Dsc1, and Dsc2EC1Dsg2EC2–EC5) reveal cadherins engaged in strand-swapped dimers typical of the adhesive bound-state of classical cadherins (Fig. 5). In contrast, the Dsg3 ectodomain structure reveals a monomeric auto-inhibited conformation in which the A*-strand of EC1 is self-docked into its own domain (Fig. 5A, Right). Both of these conformations represent distinct states in the 3D-domain swap binding mechanism (24, 33, 34) of desmosomal and type 1 classical cadherins (24), and observation of both forms is consistent with weak homodimerization affinities observed in our biophysical assays. The molecular interactions between the hydrophobic acceptor pocket and swapped strand in each of the desmosomal cadherin homodimers and the self-docked monomer form of Dsg3 are highly conserved and are similar to those described for type 1 classical cadherins (24) (SI Appendix, Fig. S6). The docked Trp2 side chain forms hydrophobic interactions with residues Ile/Val24, Tyr36, Ala80, and Leu92 (numbering according to Dsc1) that line the hydrophobic pocket, whereas the indole nitrogen engages in a hydrogen bond with the backbone carbonyl of residue 90. In addition, the amino terminus of the swapped strand engages in ionic bonds with the acidic side chains of Asp27 and Glu89 of the partner domain. Notably, the dimer structures reported here represent homodimers, which our binding data show to be of very low affinity compared with Dsg:Dsc heterodimers (Figs. 1 and 2 and SI Appendix, Table S1), and are presumably observed as a result of the high cadherin
Dsc3
Fig. 2. Heterophilic binding between Dsg- and Dsc-coated beads. Bead aggregation assay showing mixtures of green fluorescent Dsg- or Dsc-coated beads (columns) with similarly coated red fluorescent beads (rows) after 1 h of aggregation. (Scale bar: 0.5 mm.)
EC domains in addition to strand-swap interface residues in EC1. Notably, EC3 of Dsg2 contains a cluster of mutation sites, including four affecting the EC2–EC3 calcium binding linker and three surface-exposed residues in EC3 whose function is currently unknown. The overall structures of each Dsg and Dsc are similar to those of classical cadherins, except that a sharp bend is observed between domains EC3 and EC4 in Dsg2 and Dsg3, and to a lesser degree in Dsc1 and Dsc2 (Figs. 3 and 4). Dsg2 and Dsg3 have EC3–EC4 bend angles of ∼103° and ∼99°, whereas Dsc1 and Dsc2 each show a bend of ∼124° compared with ∼145° for classical N-cadherin (Fig. 4D). These conformational differences likely result from the incomplete EC3–EC4 Ca2+-coordination that is observed in the structures of desmosomal cadherins (Fig. 4A). We produced Bijvoet difference Fourier maps using low-energy X-rays to directly visualize bound Ca2+ by anomalous scattering (Fig. 4B). Anomalous density maps clearly indicated
EC4
EC3
C
C
EC5
N465 N413 T431/433/435
EC5 EC4
EC5
S424/T426
329 Å
N260
C
EC4
N412
N465 N413 T431/433/435
326 Å
EC3
EC3
N260
T321 N133
EC2
T223
EC2 N63 N
N133
EC1 EC1
N
EC1
N63
EC2
EC4
N260
N N
T203/205
EC2
N131
Asn31
EC2
N
EC2
T203/205
T223
EC3
EC3
T431/433/435 N413 N465
EC2
N494
EC4 EC5
C
Dsg2
Dsg3
Dsc1
T431/433/435 N413 N465
T423/425 N411
EC4
EC4
N31
N133
EC3
S424/T428
T427/429/431
N N
N260
EC3
C
EC5
N31
EC1 EC1
N257
EC3
N133
N61
N
EC1 EC1
EC2
T203/205
EC5
C
EC4 EC5
C
C
Dsc2 EC1
Dsc2EC1Dsg2EC2-5
Fig. 3. Overall structures of Dsg and Dsc extracellular regions. Crystal structures shown in ribbon representation of human Dsg2, Dsg3, Dsc1, Dsc2ΔEC1, and chimera Dsc2EC1Dsg2EC2-5 ectodomains (left to right). Ca2+ ions are shown as green spheres, and N-linked and O-linked glycans are shown in wheat and violet, respectively. Electron density for EC5 of Dsg3 was not observed, and its likely position is indicated with a dashed line. Predicted intermembrane distances derived from measurements between C-termini are indicated by gray arrows for the complete homodimer structures Dsg2 and Dsc1.
7162 | www.pnas.org/cgi/doi/10.1073/pnas.1606272113
Harrison et al.
A
Fig. 4. Calcium coordination in the EC3–EC4 linker of desmosomal cadherins. (A) Ribbon representation of EC3–EC4 linkers of desmosomal cadherins and classical N-cadherin. Side chains or backbone carbonyls coordinating Ca2+ ions (green spheres) are shown as sticks. Residues that fail to coordinate Ca2+ in desmosomal cadherins are highlighted with magenta text, and corresponding Ca2+-coordinating residues in classical cadherins are underlined. A Dscspecific disulfide bond in EC4 is shown as sticks. Violet spheres represent O-linked glycans, and gray spheres represent water. (B) Bijvoet anomalous difference maps, contoured at 3 σ (green mesh), showing Ca2+ ion positions in EC3–EC4 for desmosomal cadherins, as in A. (C) Multiple sequence alignment of EC3–EC4 linkers of human Dsgs, Dscs, and N-cadherin. Conserved and nonconserved Ca2+-binding residues are highlighted green and red; gray lines indicate Ca2+ coordination. (D) Superposition of Cα-traces over domain EC3, colored according to A.
B D
C
interface in Dsgs and a smaller negative patch in Dscs (Fig. 5D). The large positive Dsg potentials arise primarily from Arg10, Lys17, Lys/Arg18, and Lys97/99 (Fig. 5A), which are conserved specifically in Dsgs (SI Appendix, Fig. S4). Notably, Lys17 and Lys/Arg18 are positioned close to the interface in part as a result of a Dsg-specific shift of the AB-loop of ∼10 Å in Cα position relative to Dscs and classical cadherins (Fig. 5C). In the Dsg:Dsg dimers, the electropositive residues are brought together at Cα–Cα
concentrations in the crystals. Heterophilic Dsg:Dsc interactions are nonetheless likely to depend on the same binding mechanism, according to our mutational data (Fig. 1B and SI Appendix, Fig. S2). Structural features of the EC1 interface regions of Dsgs and Dscs suggest a molecular basis for their family-wise heterophilic specificity. Electrostatic potentials mapped to the molecular surfaces of the Dsg2, Dsg3, Dsc1, and Dsc2 EC1 domains reveal an electropositive patch near the base of the strand-swapped
Dsg2
A
C
Dsg3
A1
W2
W2
E1
W2 EC1
W2
EC1
EC1
EC1
N63 N-glycan P20 K99 N63 N-glycan
K18
K17
B
Dsc1
Y86 Y32
EC1 F35
N31 N-glycan Y90 R1
W2
R10
P18
P16 G15
Y86
Tyr77
F35
Dsc2
Y86
Y86 Y32
Y86 Y32 F79
W2
R1
N31 N-glycan
Y86
W2
Y32 EC1
R1
F79 Y35
K17
N31 N-glycan
N31 N-glycan
Y32 Y79 EC1 EC1
AB loop (Dsg2/Dsg3) K/R18
F17
K17
K17
R10
AB loop (Dsc1/Dsc2/ N-cad)
R18
N63 N-glycan
Y79 Y77
K97
N63 N-glycan
N19
Y35
Y32
F79
Y35
Y52
Y52 F61
F61 Y71
Y71
Y71
Y61
Y61 F20
F20
Y71 Y61
Y71 Y71
F20 E99
E99
E99
Y52
Y52
E99
D101
D -5kT
+5kT
Dsg2
Harrison et al.
Dsg3
Dsc1
Dsc2
Fig. 5. Adhesive interfaces of desmosomal cadherins. (A and B) EC1 domains of strand-swapped homodimer (Dsg2, Dsc1, Dsc2) or self-docked monomer (Dsg3) configurations shown in ribbon representation. Selected interface residues are shown as cyan sticks (see Results), N-glycans are shown in wheat, and clusters of aromatic residues in Dscs are shown as green sticks. (C) Superposition of Dsgs, Dscs, and N-cadherin over EC1, showing differences in the position of the ABloop. (D) Electrostatic surfaces showing the interface regions in EC1 for each desmosomal cadherin structure. (Scale bar: red, −5 kT; blue, +5 kT.)
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7163
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
Y35 Y52
A
B
C
D
F
E
G
Fig. 6. Dependence of Dsg2:Dsc2 binding on charged interface residues. (A and B) SPR affinity measurements of (A) Dsg2 or (B) Dsc2 analytes injected over surfaces captured with Dsc2 in 150, 300, or 600 mM NaCl. Black traces: experimental data at five concentrations from 12 to 0.75 μM. Red traces: fits to 1:1 interaction models. Kinetic constants: Ka 8.3(3) × 103 M−1·s−1, Kd 0.29(1)s−1 (150 mM NaCl); Ka 4(1) × 103 M−1·s−1, Kd 0.2 (5)s−1 (300 mM NaCl); Ka 1.4(4) × 103 M−1·s−1, Kd 0.10(2)s−1 (600 mM NaCl). Errors of last digit in parentheses. (C) Putative Dsg2:Dsc2 EC1 heterodimer model showing interface residues mutated in D–G. (D) Binding of wild-type or mutant Dsg2 over captured Dsc2 surfaces. (E) Binding of wild-type or mutant Dsc2 over Dsg2 surfaces. (F) Dsc2 E99Q E101Q mutant over a Dsc2 surface. (G) Dsg2 K17Q K18Q mutant over a Dsg2 surface. Proteins in D–G tested at 12, 6, and 3 μM; scale on D and E, Left.
distances of ∼10 Å, leading to moderate electrostatic repulsion (Fig. 5A and SI Appendix, Fig. S7), the exact magnitude of which would depend on side chain and EC1/EC1 orientation. Comparable repulsive interactions are also evident in Dsc:Dsc dimers, where they arise primarily from conserved Dsc-specific acidic residues Glu99 and Asp101 (Fig. 5B and SI Appendix, Fig. S7). We built models of putative Dsg:Dsc heterodimers by superposition over the homodimer structures (SI Appendix, Fig. S8). In each model, the same charged residues that repel in the homodimers (Fig. 5 and SI Appendix, Fig. S7) were positioned to form stabilizing ion pairs (Glu99 to Lys/Arg18; Asp/Glu101 to Lys17; SI Appendix, Fig. S8). This structural analysis suggests that the proximity of like charges in Dsg:Dsg and Dsc:Dsc dimers and opposite charges in Dsg: Dsc heterodimers provide a likely explanation for the family-specific heterophilic binding preferences we observe. We have reported previously that the same principle accounts for heterophilic preferences in the nectin family of adhesion proteins (25). We tested this mechanism by investigating the effects of ionic strength on Dsg2:Dsc2 binding, as high salt concentrations should decrease the affinity of interactions with favorable electrostatics (Dsg2:Dsc2) and increase the affinity of interactions limited by unfavorable electrostatics (Dsg:Dsg and Dsc:Dsc). Consistent with favorable electrostatic interactions in Dsg2:Dsc2 heterodimers, we observed that the affinity of Dsg2:Dsc2 binding in SPR decreased approximately twofold when NaCl concentrations were increased to 600 mM (Fig. 6A). No detectable increase in Dsc2 homophilic binding was observed (Fig. 6B), however, indicating that additional factors may disfavor these interactions. In a more targeted approach, we introduced point mutations in Dsg2 and Dsc2 ectodomains to neutralize or reverse the charges of Lys17 and Lys18 in Dsg2 and Glu99 and Glu101 in Dsc2, as these residues are predicted to form interacting pairs in our heterodimer models (Fig. 6C and SI Appendix, Fig. S8). Neutralization of Lys17 and Lys18 in Dsg2 by substitution with either glutamine (K17Q K18Q) or alanine (K17A K18A) markedly decreased binding to Dsc2 in SPR assays, and reversal of the charge by mutation to Glu (K17E K18E) further decreased the interaction to background levels (Fig. 6D). Similarly, in Dsc2, neutralization of residues Glu99 and Glu101 by substitution with glutamine (E99Q E101Q) or charge reversal by substitution with Lysine (E99K E101K) decreased or ablated Dsg2:Dsc2 binding (Fig. 6E). The neutralizing mutations did not measurably affect homodimerization of Dsg2 or Dsc2 (Fig. 6 F and 7164 | www.pnas.org/cgi/doi/10.1073/pnas.1606272113
G), suggesting, along with the salt-dependence data, that additional residues weaken within-family binding. Discussion The solution biophysics and bead aggregation data reported here demonstrate that all Dsgs form strong heterophilic interactions with all Dscs, whereas little or no specific binding is observed within the respective subfamilies (Figs. 1 and 2). The heterophilic Dsg:Dsc interactions show characteristics consistent with binding through the strand-swap mechanism (Fig. 1 and SI Appendix, Fig. S2) and can form in a trans orientation between apposed surfaces (Fig. 2). Ectodomain structures of representative Dsgs and Dscs show distinct features in the conserved strand-swap interface that suggest heterodimers may be favored in part by electrostatic compatibility between interface regions of Dscs and Dsgs (Fig. 5 and SI Appendix, Fig. S8), which is supported by mutational analysis and by the effects of ionic strength (Fig. 6). Together, these findings suggest that Dsg:Dsc trans heterodimers represent the basic adhesive unit of desmosomes, which is consistent with the observed coexpression of Dsgs and Dscs in tissues (1), their apparent colocalization on both apposed cell surfaces in desmosomes (1, 11, 35, 36), the requirement for expression of both subtypes for adhesion between transfected cells (1, 11–13), and previous evidence of Dsg2:Dsc2 interaction in solution (16, 17). Although desmosome formation by Dsc2 alone has been reported in a Dsg-depleted colon carcinoma cell line (37), the structure and integrity of these junctions compared with wild-type desmosomes remains to be determined. Preferential heterophilic binding between Dscs and Dsgs has important implications for the extracellular structure of desmosomes. Current models of the extracellular architecture of desmosomes are derived from electron tomography reconstructions of native desmosomes in plastic embedded or vitrified sections of mouse (7) or human skin (8). Both studies showed cadherin dimers interacting through their EC1 domains at the desmosome central dense midline. Desmosomes of mouse skin appeared dense, but structurally less ordered compared with those of human skin, which formed regular 2D arrays of cadherin dimers in the intercellular space. Despite these advances, the specific arrangement of Dsgs and Dscs in these structures could not be derived from the data, as Dsgs and Dscs could not be distinguished. However, the overall structures of desmosomal cadherins we report here suggest that, as a result of differential Ca2+ binding between domains EC3 and EC4, Dsgs and Dscs may adopt subtly different overall conformations that could potentially be Harrison et al.
Dsg:Dsc trans-heterodimers, with the likelihood of a close-packed arrangement consistent with electron tomography observations that may involve cis hetero-interactions between alternating Dsgs and Dscs on each membrane surface. A complete understanding of desmosome extracellular architecture will require additional studies to determine whether extracellular interactions other than the adhesive interactions defined here play a role in desmosome assembly.
distinguished in future studies to shed light on how trans heterodimers are arranged in the desmosome. It is of interest to contrast our structural understanding of desmosomes to that of adherens junctions (24) formed by related type 1 classical cadherins. First, end-to-end distances seen in Dsc and Dsg homodimer structures predict intermembrane distances of ∼326– 329 Å, if perpendicular to the membrane (Fig. 3), which are consistent with measurements of native desmosomes by electron microscopy (280–340 Å; refs. 7 and 8). In contrast, structures of type 1 classical cadherin dimers (24) bend less sharply between EC3 and EC4 and are thus longer than Dsgs and Dscs, but paradoxically mediate shorter intermembrane spacing as a result of the oblique angle at which assembled classical cadherins protrude from the membrane in adherens junctions (24). Second, adherens junctions are formed by a self-assembling 2D molecular layer mediated by cis and trans interactions (24), and we have found that interplay between cis and trans interactions is also crucial to clustered protocadherin function (38). Given the highly ordered structures observed in cryo-EM tomograms (8), it seems likely that cis interactions will also be important for desmosomes. We do not observe cis interactions in our crystal structures. However, if such interactions were formed between Dsgs and Dscs, they would not be expected to be observed in the homomeric structures reported here. Interestingly, the Dsc1 and Dsc2 EC1 domains contain nine and seven conserved aromatic residues distal from the adhesive interface, respectively, which protrude from the surface to give the appearance of an aromatic “pin-cushion” (Fig. 5B and SI Appendix, Fig. S4). The function of these unusual surface-exposed aromatic residues is unknown, but it is possible that they mediate cis interactions by binding Dsgs in desmosome assembly. Overall, our results suggest a desmosome extracellular architecture composed of
ACKNOWLEDGMENTS. We thank Surajit Banerjee, Igor Kourinov, Frank V. Murphy, and Narayanasami Sukumar at APS for synchrotron data collection support. cDNA for Dsc3 was provided by Peter Koch (University of Colorado) and Aimee S. Payne (University of Pennsylvania). This work was supported by NIH Grants R01 GM062270 (to L.S.) and R01 GM030518 (to B.H.) and NSF grant Grant MCB-1412472 (to B.H.). X-ray data were collected at the NECAT beamlines, funded by NIH Grant P41 GM103403.
1. Dusek RL, Godsel LM, Green KJ (2007) Discriminating roles of desmosomal cadherins: Beyond desmosomal adhesion. J Dermatol Sci 45(1):7–21. 2. Garrod D (2010) Desmosomes in vivo. Dermatol Res Pract 2010:212439. 3. Stahley SN, Kowalczyk AP (2015) Desmosomes in acquired disease. Cell Tissue Res 360(3): 439–456. 4. Al-Jassar C, Bikker H, Overduin M, Chidgey M (2013) Mechanistic basis of desmosometargeted diseases. J Mol Biol 425(21):4006–4022. 5. Delva E, Tucker DK, Kowalczyk AP (2009) The desmosome. Cold Spring Harb Perspect Biol 1(2):a002543. 6. Rayns DG, Simpson FO, Ledingham JM (1969) Ultrastructure of desmosomes in mammalian intercalated disc; appearances after lanthanum treatment. J Cell Biol 42(1):322–326. 7. He W, Cowin P, Stokes DL (2003) Untangling desmosomal knots with electron tomography. Science 302(5642):109–113. 8. Al-Amoudi A, Díez DC, Betts MJ, Frangakis AS (2007) The molecular architecture of cadherins in native epidermal desmosomes. Nature 450(7171):832–837. 9. Holthöfer B, Windoffer R, Troyanovsky S, Leube RE (2007) Structure and function of desmosomes. Int Rev Cytol 264:65–163. 10. North AJ, Chidgey MA, Clarke JP, Bardsley WG, Garrod DR (1996) Distinct desmocollin isoforms occur in the same desmosomes and show reciprocally graded distributions in bovine nasal epidermis. Proc Natl Acad Sci USA 93(15):7701–7705. 11. Chitaev NA, Troyanovsky SM (1997) Direct Ca2+-dependent heterophilic interaction between desmosomal cadherins, desmoglein and desmocollin, contributes to cell-cell adhesion. J Cell Biol 138(1):193–201. 12. Getsios S, et al. (2004) Coordinated expression of desmoglein 1 and desmocollin 1 regulates intercellular adhesion. Differentiation 72(8):419–433. 13. Marcozzi C, Burdett ID, Buxton RS, Magee AI (1998) Coexpression of both types of desmosomal cadherin and plakoglobin confers strong intercellular adhesion. J Cell Sci 111(Pt 4):495–509. 14. Chidgey M, et al. (2001) Mice lacking desmocollin 1 show epidermal fragility accompanied by barrier defects and abnormal differentiation. J Cell Biol 155(5): 821–832. 15. Koch PJ, et al. (1997) Targeted disruption of the pemphigus vulgaris antigen (desmoglein 3) gene in mice causes loss of keratinocyte cell adhesion with a phenotype similar to pemphigus vulgaris. J Cell Biol 137(5):1091–1102. 16. Syed SE, et al. (2002) Molecular interactions between desmosomal cadherins. Biochem J 362(Pt 2):317–327. 17. Lowndes M, et al. (2014) Different roles of cadherins in the assembly and structural integrity of the desmosome complex. J Cell Sci 127(Pt 10):2339–2350. 18. Nie Z, Merritt A, Rouhi-Parkouhi M, Tabernero L, Garrod D (2011) Membraneimpermeable cross-linking provides evidence for homophilic, isoform-specific binding of desmosomal cadherins in epithelial cells. J Biol Chem 286(3):2143–2154. 19. Shapiro L, et al. (1995) Structural basis of cell-cell adhesion by cadherins. Nature 374(6520):327–337.
20. Brasch J, Harrison OJ, Honig B, Shapiro L (2012) Thinking outside the cell: How cadherins drive adhesion. Trends Cell Biol 22(6):299–310. 21. Tariq H, et al. (2015) Cadherin flexibility provides a key difference between desmosomes and adherens junctions. Proc Natl Acad Sci USA 112(17):5395–5400. 22. Katsamba P, et al. (2009) Linking molecular affinity and cellular specificity in cadherinmediated adhesion. Proc Natl Acad Sci USA 106(28):11594–11599. 23. Brasch J, et al. (2011) Structure and binding mechanism of vascular endothelial cadherin: A divergent classical cadherin. J Mol Biol 408(1):57–73. 24. Harrison OJ, et al. (2011) The extracellular architecture of adherens junctions revealed by crystal structures of type I cadherins. Structure 19(2):244–256. 25. Harrison OJ, et al. (2012) Nectin ectodomain structures reveal a canonical adhesive interface. Nat Struct Mol Biol 19(9):906–915. 26. Vendome J, et al. (2014) Structural and energetic determinants of adhesive binding specificity in type I cadherins. Proc Natl Acad Sci USA 111(40):E4175–E4184. 27. Seiradake E, et al. (2014) FLRT structure: Balancing repulsion and cell adhesion in cortical and vascular development. Neuron 84(2):370–385. 28. Özkan E, et al. (2014) Extracellular architecture of the SYG-1/SYG-2 adhesion complex instructs synaptogenesis. Cell 156(3):482–494. 29. Messenger AG, Bazzi H, Parslew R, Shapiro L, Christiano AM (2005) A missense mutation in the cadherin interaction site of the desmoglein 4 gene underlies localized autosomal recessive hypotrichosis. J Invest Dermatol 125(5):1077–1079. 30. Lommel M, et al. (2013) Protein O-mannosylation is crucial for E-cadherin-mediated cell adhesion. Proc Natl Acad Sci USA 110(52):21024–21029. 31. Vester-Christensen MB, et al. (2013) Mining the O-mannose glycoproteome reveals cadherins as major O-mannosylated glycoproteins. Proc Natl Acad Sci USA 110(52): 21018–21023. 32. van der Zwaag PA, et al. (2009) A genetic variants database for arrhythmogenic right ventricular dysplasia/cardiomyopathy. Hum Mutat 30(9):1278–1283. 33. Chen CP, Posy S, Ben-Shaul A, Shapiro L, Honig BH (2005) Specificity of cell-cell adhesion by classical cadherins: Critical role for low-affinity dimerization through betastrand swapping. Proc Natl Acad Sci USA 102(24):8531–8536. 34. Häussinger D, et al. (2004) Proteolytic E-cadherin activation followed by solution NMR and X-ray crystallography. EMBO J 23(8):1699–1708. 35. North AJ, et al. (1999) Molecular map of the desmosomal plaque. J Cell Sci 112(Pt 23): 4325–4336. 36. Allen E, Yu QC, Fuchs E (1996) Mice expressing a mutant desmosomal cadherin exhibit abnormalities in desmosomes, proliferation, and epidermal differentiation. J Cell Biol 133(6): 1367–1382. 37. Fujiwara M, et al. (2015) Desmocollin-2 alone forms functional desmosomal plaques, with the plaque formation requiring the juxtamembrane region and plakophilins. J Biochem 158(4):339–353. 38. Thu CA, et al. (2014) Single-cell identity generated by combinatorial homophilic interactions between α, β, and γ protocadherins. Cell 158(5):1045–1059.
Harrison et al.
Materials and Methods
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7165
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
Recombinant Dsg and Dsc ectodomains were expressed in HEK293 cells and purified by nickel affinity chromatography and gel filtration. Biotinylated proteins containing C-terminal AviTag sequences were cotransfected with biotin ligase. AUC experiments were performed in a Beckman XL-A/I analytical ultracentrifuge at 25 °C, using UV detection. For SPR, biotinylated proteins were captured over NeutrAvidin-immobilized surfaces, and binding of analytes at 3, 6, and 12 μM was assessed at 25 °C, using a Biacore T100 biosensor. For bead aggregation, biotinylated cadherins were captured on fluorescent NeutrAvidin-labeled microspheres, and aggregation was assessed by fluorescence microscopy after 1 h mixing. For structure determination, protein crystals were grown by vapor diffusion in hanging drops, and X-ray diffraction data were collected from single crystals at 100 K, using a wavelength of 0.979 Å at the 24-ID-C/E beamlines at Argonne National Laboratory. Structures were solved by molecular replacement using models 3Q2W, 3Q2V, and 1L3W. See SI Appendix, Materials and Methods for details.
Department of Systems Biology, Columbia University, New York, NY 10032; bHoward Hughes Medical Institute, Columbia University, New York, NY 10032; Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY 10032; dCenter for Computational Biology and Bioinformatics, Columbia University, New York, NY 10032; eDepartment of Medicine, Columbia University, New York, NY 10032; and fZuckerman Mind Brain Behavior Institute, Columbia University, New York, NY 10032 c
Contributed by Barry Honig, April 23, 2016 (sent for review January 21, 2016; reviewed by Steven C. Almo and Dimitar B. Nikolov)
Desmosomes are intercellular adhesive junctions that impart strength to vertebrate tissues. Their dense, ordered intercellular attachments are formed by desmogleins (Dsgs) and desmocollins (Dscs), but the nature of trans-cellular interactions between these specialized cadherins is unclear. Here, using solution biophysics and coated-bead aggregation experiments, we demonstrate family-wise heterophilic specificity: All Dsgs form adhesive dimers with all Dscs, with affinities characteristic of each Dsg:Dsc pair. Crystal structures of ectodomains from Dsg2 and Dsg3 and from Dsc1 and Dsc2 show binding through a strand-swap mechanism similar to that of homophilic classical cadherins. However, conserved charged amino acids inhibit Dsg:Dsg and Dsc:Dsc interactions by same-charge repulsion and promote heterophilic Dsg:Dsc interactions through oppositecharge attraction. These findings show that Dsg:Dsc heterodimers represent the fundamental adhesive unit of desmosomes and provide a structural framework for understanding desmosome assembly. desmosome
| cell adhesion | cadherin | heterophilic binding
D
esmosomes are spot-weld-like intercellular adhesive junctions that link the intermediate filament networks of adjacent cells to impart strength to the solid tissues of vertebrates (1–3). Dysfunction of desmosomes in inherited and acquired human diseases as well as in mouse genetic ablation studies causes characteristic defects in heart muscle and skin (3–5), demonstrating their importance in tissues that undergo mechanical stress. In electron micrographs, the hallmarks of mature desmosomes include a constant intermembrane distance of 280–340 Å, and apparently ordered electron-density in the intercellular space, often with a discrete midline connected by periodic cross-bridges to the cell membranes (6–9). The intercellular attachments of desmosomes are composed of transmembrane proteins from two specialized cadherin subfamilies: desmocollins (Dscs) and desmogleins (Dsgs). The human genome encodes three Dsc (Dsc1– Dsc3) and four Dsg (Dsg1–Dsg4) proteins, which share an overall domain organization comprising four to five extracellular cadherin (EC) domains, a single-pass transmembrane region, and an intracellular domain that binds to intermediate filaments via adaptor proteins desmoplakin and plakoglobin (1). Individual Dsgs and Dscs show differential expression patterns: Dsg2 and Dsc2 are expressed widely in all desmosome-forming tissues (1), whereas other desmosomal cadherins are expressed specifically in stratified epithelia with graded, overlapping patterns (1, 10). Notably, both Dscs and Dsgs appear necessary for adhesion in transfected cells (1, 11–13), and loss of either in genetic experiments causes loss of normal desmosomal adhesion (5, 14, 15). Although the ultrastructure of desmosomes is well characterized, a molecular-level understanding of the binding interactions between desmosomal cadherin extracellular regions that assemble these junctions has remained elusive. In particular, whether desmosomal cadherins have homophilic preferences or whether interactions occur between heterophilic pairs has been a matter of dispute (1, 11–13, 16– 18). Electron tomography studies of native desmosomes (7, 8) have revealed cadherins binding through their EC1 domains at the central 7160–7165 | PNAS | June 28, 2016 | vol. 113 | no. 26
dense midline, consistent with a strand-swap mode of interaction first characterized for classical cadherins (19, 20). Nevertheless, the identity of Dscs and Dsgs in these tomographic reconstructions could not be determined, and atomic-resolution structures of desmosomal cadherins have not been available, with the exception of an NMR structure of a monomeric EC1 fragment of mouse Dsg2 with an artificially extended N terminus (PDB ID code 2YQG). In addition, a small-angle X-ray scattering (SAXS) study showed a classical cadherin-like monomeric molecular envelope for mouse Dsg2, but with additional flexibility in the membrane-proximal region (21). Here we present crystal structures of ectodomains from human Dsgs and Dscs, along with solution biophysical analysis and coated-bead aggregation studies. Our results identify the central importance of heterophilic binding between Dscs and Dsgs and reveal the structural basis of trans-cellular adhesive binding between desmosomal cadherins. Results Heterophilic Adhesive Binding Between Dsgs and Dscs. Using a transient-transfection HEK-293 cell protein expression system (Materials and Methods), we produced full-length ectodomains for all seven human desmosomal cadherins Dsg1–Dsg4 and Dsc1–Dsc3. We first assessed the homo-association state of each of these proteins by sedimentation-equilibrium analytical ultracentrifugation (AUC) (SI
Significance Desmosomes are crucial for the integrity of tissues that undergo mechanical stress. Their intercellular attachments are assembled from desmogleins (Dsgs) and desmocollins (Dscs), two families of specialized cadherins whose structures and interactions have remained uncharacterized. Our study demonstrates family-wise heterophilic interactions between these proteins, with all Dsgs forming adhesive dimers with all Dscs. Crystal structures of ectodomains from Dsg2 and Dsg3 and from Dsc1 and Dsc2 show binding through a strand-swap mechanism similar to that of classical cadherins, which we show underlie heterophilic interactions. Conserved compatibly charged amino acids in the interfaces promote heterophilic Dsg:Dsc interactions. We show that Dsg:Dsc heterodimers represent the fundamental adhesive unit of desmosomes and provide a structural framework for understanding the extracellular assembly of desmosomes. Author contributions: O.J.H., J.B., B.H., and L.S. designed research; O.J.H., J.B., G.L., P.S.K., and G.A. performed research; O.J.H., J.B., G.L., P.S.K., G.A., B.H., and L.S. analyzed data; and O.J.H., J.B., B.H., and L.S. wrote the paper. Reviewers: S.C.A., Albert Einstein College of Medicine; and D.B.N., Memorial Sloan-Kettering Cancer Center. The authors declare no conflict of interest. Data deposition: The atomic coordinates have been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 5ERD, 5EQX, 5IRY, 5ERP, and 5J5J). 1
O.J.H. and J.B. contributed equally to this work.
2
To whom correspondence may be addressed. Email: [email protected] or [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1606272113/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1606272113
Appendix, Table S1). Each of the four Dsgs and Dsc2 and Dsc3 appeared to be either monomers or very weak dimers in solution with KD values higher than 400 μM, weaker than the weakest-binding classical cadherin (E-cadherin, KD = 156 μM; ref. 22). Only Dsc1 showed appreciable homodimerization, but this appeared to be a result of technical issues involving the formation of nonnative intermolecular disulfides and was not observed in other binding assays (see following text and Materials and Methods). We then used surface plasmon resonance (SPR) to measure all pairwise homophilic and heterophilic interactions among the seven human desmosomal cadherins (Fig. 1 A–C). Consistent with the AUC results, no significant homophilic interactions were observed, and in addition, no significant binding within families (Dsg:Dsg or Dsc:Dsc) was detected (Fig. 1A). In contrast, all experiments in which a Dsg was passed over a Dsc surface or a Dsc was passed over a Dsg surface showed binding responses indicative of strong heterophilic interaction (Fig. 1A). Optimal fits of SPR data were obtained with a 1:1 binding model (SI Appendix, Fig. S1), yielding heterophilic KDs ranging between 3.6 and 43.9 μM (Fig. 1C), similar to the affinities of other tight-binding cell adhesion molecules (22–28). Interestingly, the strongest-binding pairs (Dsg1:Dsc1 and Dsg4:Dsc1) correspond to those expressed in the outermost layers of human skin, and the weakest (Dsg3:Dsc3) to those restricted to basal layers (1, 10), consistent with a potential role for differential affinities in maintenance of cell arrangement in stratified epithelia (1). Dsgs and Dscs both conserve sequence features in EC1 critical to the strand-swap mechanism of adhesive binding in classical cadherins (20, 29), suggesting strand-swapping mediates heterophilic Dsg:Dsc interactions. Consistent with this mechanism, deletion of domain EC1 in Dsgs and Dscs abolished all heterophilic interactions, whereas deletion of the C-terminal EC4–EC5 domains had little effect on binding (SI Appendix, Fig. S2A). Furthermore, combined mutation of Trp2 (W2A), whose indole side chain anchors the swapping N-terminal β-strand (the A*-strand), and Ala80/82 [A80I (Dsc2) or A82I (Dsg2)], whose small side chain permits Trp2 docking in the dimer-partner acceptor pocket, ablated all heterophilic binding of Dsg2 and Dsc2 (Fig. 1B). Interestingly, Dsg2 containing only a W2A mutation on the swapping strand could still mediate strong heterophilic binding to Dscs, whereas an A82I acceptorpocket mutation ablated all binding (SI Appendix, Fig. S2B). This was reversed in Dsc2, where the W2A mutation was more severe. These mutants suggest that Dsg:Dsc interactions are asymmetric, with docking of the Dsc A*-strand in the Dsg acceptor pocket representing the dominant interaction.
A
To assess their adhesive properties on apposed surfaces, we coupled C-terminally biotinylated Dsg or Dsc ectodomains to NeutrAvidin-coated beads and monitored their aggregation by fluorescence microscopy (Fig. 2). As expected from the SPR results (Fig. 1A), robust aggregation was observed when beads coated with Dsgs were mixed with beads coated with Dscs, and no aggregation was observed for any Dsg:Dsg or Dsc:Dsc pair. Because Dscs and Dsgs were immobilized on separate surfaces, the bead aggregation assays confirm that Dsg:Dsc heterophilic interactions can form in a trans-orientation between apposed surfaces analogous to apposed cell membranes in desmosomes. Aggregation was dependent on the presence of Ca2+, and identical results were obtained when Dsg2 and Dsc2 were coated together on the same beads (SI Appendix, Fig. S3). Together, the AUC, SPR, and bead aggregation data show that Dsgs and Dscs form strong strand-swapped adhesive trans interactions with family-wise heterophilic specificity, which are likely to underlie intercellular adhesion in desmosomes. Overall Structures of Dsgs and Dscs. We determined crystal structures for full-length glycosylated ectodomains of human Dsg2, Dsg3, and Dsc1 (Fig. 3) with diffraction limits between 2.9 and 3.1Å (SI Appendix, Table S2). Diffracting crystals of human Dsc2 ectodomains were not obtained, but we determined crystal structures of an EC2−EC5 fragment (Dsc2ΔEC1) at 2.7 Å resolution and of the adhesive EC1 domain in the context of a chimeric protein Dsc2EC1Dsg2EC2–EC5 at 3.3Å resolution (Fig. 3 and SI Appendix, Table S2). Each desmosomal cadherin extracellular region has an overall structure comprising an elongated, curved arrangement of tandem EC domains connected by calcium-binding linker regions (Fig. 3). Strand-swapped homodimers, formed between EC1 domains, were observed for Dsg2, Dsc1, and Dsc2EC1Dsg2EC2-5 (Fig. 3). Evidence of extensive N- and O-linked glycosylation was also observed in the electron density maps. Notably, there are conserved, subfamily-specific N-linked glycosylation sites in the adhesive EC1 domains, although none are involved in adhesive contacts, and contiguous regions of conserved O-linked glycosylation in EC4 (Figs. 3–5 and SI Appendix, Fig. S4), corresponding to the novel mono-Omannosylation sites previously identified in Dsc2 and other cadherins by mass spectrometry (30, 31). Because inherited mutations in Dsg2 and Dsc2 are associated with weakened heart muscle desmosomes in arrhythmogenic right-ventricular cardio-myopathy (4, 32), we mapped pathogenic missense mutations associated with the disease onto the Dsg2 and Dsc2 ectodomain structures (SI Appendix, Fig. S5). The mutations were found to target structural elements including calcium-binding residues, glycosylation sites, and disulfide bonds in all
B KD [ M]
Dsg1
Dsg2
Dsg3
Dsg4
Dsc1
3.55 ±0.01
11.52 ±0.02
8.07 ±0.03
7.9 ±0.1
Dsc2
21.9 ±0.2
29.0 ±0.2
37.0 ±0.2
22.5 ±0.2
Dsc3
18.1 ±0.3
34.4 ±0.6
43.9 ±0.6
32.5 ±0.4
Fig. 1. Heterophilic interactions between Dsgs and Dscs in SPR. (A) Wild-type Dsg1–Dsg4, Dsc1–Dsc3, or (B) double-mutant Dsg2 W2A A82I and Dsc2 W2A A80I analytes (columns) were tested at concentrations of 3, 6, and 12 μM (12 μM omitted for Dsg4) over surfaces of wild-type Dsg2 (top row), Dsg3, Dsc1, Dsc2, and Dsc3 (bottom row). Responses were normalized for analyte molecular weight differences and are drawn to the same scale across rows. (C) Heterophilic binding affinities (KD) from fitting of SPR data to 1:1 interaction models (SI Appendix, Fig. S1). ±Errors of the fit.
Harrison et al.
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7161
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
C
Dsg1
Dsg2
Dsg3
Dsc1
the presence of all three canonical Ca2+ ions in the EC3–EC4 linker of Dsc1 and Dsc2, whereas only two Ca2+ ions were observed in Dsg2 and Dsg3, which lacked Ca3 that is normally coordinated primarily by residues in EC4 (Fig. 4A). Sequence analyses (Fig. 4C and ref. 21) show that all Dsgs lack two of the canonical Ca2+-binding residues in EC4 corresponding to Asn333 and Gln371 (N-cadherin numbering), explaining the lack of Ca3 coordination and suggesting Dsg1 and Dsg4 will adopt similarly bent conformations. In Dscs, most Ca2+-binding ligands are conserved; however, residues equivalent to Gln371 and Asp420, which coordinate Ca3 in classical cadherins, are positioned far from Ca3 in the Dsc structures (Fig. 4 A–C). In addition to the structural role of the EC3– EC4 bend, it is possible that these sites represent loci of flexibility, consistent with recent SAXS analysis of mouse Dsg2 (21), which could be important for the mechanical resilience of desmosomes.
Dsc3
Dsc2
Dsg1
Dsg2
Dsg3
Dsc1
Dsc2
Homodimer Interfaces Reveal Determinants of Heterophilic Specificity.
Three of the crystal structures reported here (Dsg2, Dsc1, and Dsc2EC1Dsg2EC2–EC5) reveal cadherins engaged in strand-swapped dimers typical of the adhesive bound-state of classical cadherins (Fig. 5). In contrast, the Dsg3 ectodomain structure reveals a monomeric auto-inhibited conformation in which the A*-strand of EC1 is self-docked into its own domain (Fig. 5A, Right). Both of these conformations represent distinct states in the 3D-domain swap binding mechanism (24, 33, 34) of desmosomal and type 1 classical cadherins (24), and observation of both forms is consistent with weak homodimerization affinities observed in our biophysical assays. The molecular interactions between the hydrophobic acceptor pocket and swapped strand in each of the desmosomal cadherin homodimers and the self-docked monomer form of Dsg3 are highly conserved and are similar to those described for type 1 classical cadherins (24) (SI Appendix, Fig. S6). The docked Trp2 side chain forms hydrophobic interactions with residues Ile/Val24, Tyr36, Ala80, and Leu92 (numbering according to Dsc1) that line the hydrophobic pocket, whereas the indole nitrogen engages in a hydrogen bond with the backbone carbonyl of residue 90. In addition, the amino terminus of the swapped strand engages in ionic bonds with the acidic side chains of Asp27 and Glu89 of the partner domain. Notably, the dimer structures reported here represent homodimers, which our binding data show to be of very low affinity compared with Dsg:Dsc heterodimers (Figs. 1 and 2 and SI Appendix, Table S1), and are presumably observed as a result of the high cadherin
Dsc3
Fig. 2. Heterophilic binding between Dsg- and Dsc-coated beads. Bead aggregation assay showing mixtures of green fluorescent Dsg- or Dsc-coated beads (columns) with similarly coated red fluorescent beads (rows) after 1 h of aggregation. (Scale bar: 0.5 mm.)
EC domains in addition to strand-swap interface residues in EC1. Notably, EC3 of Dsg2 contains a cluster of mutation sites, including four affecting the EC2–EC3 calcium binding linker and three surface-exposed residues in EC3 whose function is currently unknown. The overall structures of each Dsg and Dsc are similar to those of classical cadherins, except that a sharp bend is observed between domains EC3 and EC4 in Dsg2 and Dsg3, and to a lesser degree in Dsc1 and Dsc2 (Figs. 3 and 4). Dsg2 and Dsg3 have EC3–EC4 bend angles of ∼103° and ∼99°, whereas Dsc1 and Dsc2 each show a bend of ∼124° compared with ∼145° for classical N-cadherin (Fig. 4D). These conformational differences likely result from the incomplete EC3–EC4 Ca2+-coordination that is observed in the structures of desmosomal cadherins (Fig. 4A). We produced Bijvoet difference Fourier maps using low-energy X-rays to directly visualize bound Ca2+ by anomalous scattering (Fig. 4B). Anomalous density maps clearly indicated
EC4
EC3
C
C
EC5
N465 N413 T431/433/435
EC5 EC4
EC5
S424/T426
329 Å
N260
C
EC4
N412
N465 N413 T431/433/435
326 Å
EC3
EC3
N260
T321 N133
EC2
T223
EC2 N63 N
N133
EC1 EC1
N
EC1
N63
EC2
EC4
N260
N N
T203/205
EC2
N131
Asn31
EC2
N
EC2
T203/205
T223
EC3
EC3
T431/433/435 N413 N465
EC2
N494
EC4 EC5
C
Dsg2
Dsg3
Dsc1
T431/433/435 N413 N465
T423/425 N411
EC4
EC4
N31
N133
EC3
S424/T428
T427/429/431
N N
N260
EC3
C
EC5
N31
EC1 EC1
N257
EC3
N133
N61
N
EC1 EC1
EC2
T203/205
EC5
C
EC4 EC5
C
C
Dsc2 EC1
Dsc2EC1Dsg2EC2-5
Fig. 3. Overall structures of Dsg and Dsc extracellular regions. Crystal structures shown in ribbon representation of human Dsg2, Dsg3, Dsc1, Dsc2ΔEC1, and chimera Dsc2EC1Dsg2EC2-5 ectodomains (left to right). Ca2+ ions are shown as green spheres, and N-linked and O-linked glycans are shown in wheat and violet, respectively. Electron density for EC5 of Dsg3 was not observed, and its likely position is indicated with a dashed line. Predicted intermembrane distances derived from measurements between C-termini are indicated by gray arrows for the complete homodimer structures Dsg2 and Dsc1.
7162 | www.pnas.org/cgi/doi/10.1073/pnas.1606272113
Harrison et al.
A
Fig. 4. Calcium coordination in the EC3–EC4 linker of desmosomal cadherins. (A) Ribbon representation of EC3–EC4 linkers of desmosomal cadherins and classical N-cadherin. Side chains or backbone carbonyls coordinating Ca2+ ions (green spheres) are shown as sticks. Residues that fail to coordinate Ca2+ in desmosomal cadherins are highlighted with magenta text, and corresponding Ca2+-coordinating residues in classical cadherins are underlined. A Dscspecific disulfide bond in EC4 is shown as sticks. Violet spheres represent O-linked glycans, and gray spheres represent water. (B) Bijvoet anomalous difference maps, contoured at 3 σ (green mesh), showing Ca2+ ion positions in EC3–EC4 for desmosomal cadherins, as in A. (C) Multiple sequence alignment of EC3–EC4 linkers of human Dsgs, Dscs, and N-cadherin. Conserved and nonconserved Ca2+-binding residues are highlighted green and red; gray lines indicate Ca2+ coordination. (D) Superposition of Cα-traces over domain EC3, colored according to A.
B D
C
interface in Dsgs and a smaller negative patch in Dscs (Fig. 5D). The large positive Dsg potentials arise primarily from Arg10, Lys17, Lys/Arg18, and Lys97/99 (Fig. 5A), which are conserved specifically in Dsgs (SI Appendix, Fig. S4). Notably, Lys17 and Lys/Arg18 are positioned close to the interface in part as a result of a Dsg-specific shift of the AB-loop of ∼10 Å in Cα position relative to Dscs and classical cadherins (Fig. 5C). In the Dsg:Dsg dimers, the electropositive residues are brought together at Cα–Cα
concentrations in the crystals. Heterophilic Dsg:Dsc interactions are nonetheless likely to depend on the same binding mechanism, according to our mutational data (Fig. 1B and SI Appendix, Fig. S2). Structural features of the EC1 interface regions of Dsgs and Dscs suggest a molecular basis for their family-wise heterophilic specificity. Electrostatic potentials mapped to the molecular surfaces of the Dsg2, Dsg3, Dsc1, and Dsc2 EC1 domains reveal an electropositive patch near the base of the strand-swapped
Dsg2
A
C
Dsg3
A1
W2
W2
E1
W2 EC1
W2
EC1
EC1
EC1
N63 N-glycan P20 K99 N63 N-glycan
K18
K17
B
Dsc1
Y86 Y32
EC1 F35
N31 N-glycan Y90 R1
W2
R10
P18
P16 G15
Y86
Tyr77
F35
Dsc2
Y86
Y86 Y32
Y86 Y32 F79
W2
R1
N31 N-glycan
Y86
W2
Y32 EC1
R1
F79 Y35
K17
N31 N-glycan
N31 N-glycan
Y32 Y79 EC1 EC1
AB loop (Dsg2/Dsg3) K/R18
F17
K17
K17
R10
AB loop (Dsc1/Dsc2/ N-cad)
R18
N63 N-glycan
Y79 Y77
K97
N63 N-glycan
N19
Y35
Y32
F79
Y35
Y52
Y52 F61
F61 Y71
Y71
Y71
Y61
Y61 F20
F20
Y71 Y61
Y71 Y71
F20 E99
E99
E99
Y52
Y52
E99
D101
D -5kT
+5kT
Dsg2
Harrison et al.
Dsg3
Dsc1
Dsc2
Fig. 5. Adhesive interfaces of desmosomal cadherins. (A and B) EC1 domains of strand-swapped homodimer (Dsg2, Dsc1, Dsc2) or self-docked monomer (Dsg3) configurations shown in ribbon representation. Selected interface residues are shown as cyan sticks (see Results), N-glycans are shown in wheat, and clusters of aromatic residues in Dscs are shown as green sticks. (C) Superposition of Dsgs, Dscs, and N-cadherin over EC1, showing differences in the position of the ABloop. (D) Electrostatic surfaces showing the interface regions in EC1 for each desmosomal cadherin structure. (Scale bar: red, −5 kT; blue, +5 kT.)
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7163
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
Y35 Y52
A
B
C
D
F
E
G
Fig. 6. Dependence of Dsg2:Dsc2 binding on charged interface residues. (A and B) SPR affinity measurements of (A) Dsg2 or (B) Dsc2 analytes injected over surfaces captured with Dsc2 in 150, 300, or 600 mM NaCl. Black traces: experimental data at five concentrations from 12 to 0.75 μM. Red traces: fits to 1:1 interaction models. Kinetic constants: Ka 8.3(3) × 103 M−1·s−1, Kd 0.29(1)s−1 (150 mM NaCl); Ka 4(1) × 103 M−1·s−1, Kd 0.2 (5)s−1 (300 mM NaCl); Ka 1.4(4) × 103 M−1·s−1, Kd 0.10(2)s−1 (600 mM NaCl). Errors of last digit in parentheses. (C) Putative Dsg2:Dsc2 EC1 heterodimer model showing interface residues mutated in D–G. (D) Binding of wild-type or mutant Dsg2 over captured Dsc2 surfaces. (E) Binding of wild-type or mutant Dsc2 over Dsg2 surfaces. (F) Dsc2 E99Q E101Q mutant over a Dsc2 surface. (G) Dsg2 K17Q K18Q mutant over a Dsg2 surface. Proteins in D–G tested at 12, 6, and 3 μM; scale on D and E, Left.
distances of ∼10 Å, leading to moderate electrostatic repulsion (Fig. 5A and SI Appendix, Fig. S7), the exact magnitude of which would depend on side chain and EC1/EC1 orientation. Comparable repulsive interactions are also evident in Dsc:Dsc dimers, where they arise primarily from conserved Dsc-specific acidic residues Glu99 and Asp101 (Fig. 5B and SI Appendix, Fig. S7). We built models of putative Dsg:Dsc heterodimers by superposition over the homodimer structures (SI Appendix, Fig. S8). In each model, the same charged residues that repel in the homodimers (Fig. 5 and SI Appendix, Fig. S7) were positioned to form stabilizing ion pairs (Glu99 to Lys/Arg18; Asp/Glu101 to Lys17; SI Appendix, Fig. S8). This structural analysis suggests that the proximity of like charges in Dsg:Dsg and Dsc:Dsc dimers and opposite charges in Dsg: Dsc heterodimers provide a likely explanation for the family-specific heterophilic binding preferences we observe. We have reported previously that the same principle accounts for heterophilic preferences in the nectin family of adhesion proteins (25). We tested this mechanism by investigating the effects of ionic strength on Dsg2:Dsc2 binding, as high salt concentrations should decrease the affinity of interactions with favorable electrostatics (Dsg2:Dsc2) and increase the affinity of interactions limited by unfavorable electrostatics (Dsg:Dsg and Dsc:Dsc). Consistent with favorable electrostatic interactions in Dsg2:Dsc2 heterodimers, we observed that the affinity of Dsg2:Dsc2 binding in SPR decreased approximately twofold when NaCl concentrations were increased to 600 mM (Fig. 6A). No detectable increase in Dsc2 homophilic binding was observed (Fig. 6B), however, indicating that additional factors may disfavor these interactions. In a more targeted approach, we introduced point mutations in Dsg2 and Dsc2 ectodomains to neutralize or reverse the charges of Lys17 and Lys18 in Dsg2 and Glu99 and Glu101 in Dsc2, as these residues are predicted to form interacting pairs in our heterodimer models (Fig. 6C and SI Appendix, Fig. S8). Neutralization of Lys17 and Lys18 in Dsg2 by substitution with either glutamine (K17Q K18Q) or alanine (K17A K18A) markedly decreased binding to Dsc2 in SPR assays, and reversal of the charge by mutation to Glu (K17E K18E) further decreased the interaction to background levels (Fig. 6D). Similarly, in Dsc2, neutralization of residues Glu99 and Glu101 by substitution with glutamine (E99Q E101Q) or charge reversal by substitution with Lysine (E99K E101K) decreased or ablated Dsg2:Dsc2 binding (Fig. 6E). The neutralizing mutations did not measurably affect homodimerization of Dsg2 or Dsc2 (Fig. 6 F and 7164 | www.pnas.org/cgi/doi/10.1073/pnas.1606272113
G), suggesting, along with the salt-dependence data, that additional residues weaken within-family binding. Discussion The solution biophysics and bead aggregation data reported here demonstrate that all Dsgs form strong heterophilic interactions with all Dscs, whereas little or no specific binding is observed within the respective subfamilies (Figs. 1 and 2). The heterophilic Dsg:Dsc interactions show characteristics consistent with binding through the strand-swap mechanism (Fig. 1 and SI Appendix, Fig. S2) and can form in a trans orientation between apposed surfaces (Fig. 2). Ectodomain structures of representative Dsgs and Dscs show distinct features in the conserved strand-swap interface that suggest heterodimers may be favored in part by electrostatic compatibility between interface regions of Dscs and Dsgs (Fig. 5 and SI Appendix, Fig. S8), which is supported by mutational analysis and by the effects of ionic strength (Fig. 6). Together, these findings suggest that Dsg:Dsc trans heterodimers represent the basic adhesive unit of desmosomes, which is consistent with the observed coexpression of Dsgs and Dscs in tissues (1), their apparent colocalization on both apposed cell surfaces in desmosomes (1, 11, 35, 36), the requirement for expression of both subtypes for adhesion between transfected cells (1, 11–13), and previous evidence of Dsg2:Dsc2 interaction in solution (16, 17). Although desmosome formation by Dsc2 alone has been reported in a Dsg-depleted colon carcinoma cell line (37), the structure and integrity of these junctions compared with wild-type desmosomes remains to be determined. Preferential heterophilic binding between Dscs and Dsgs has important implications for the extracellular structure of desmosomes. Current models of the extracellular architecture of desmosomes are derived from electron tomography reconstructions of native desmosomes in plastic embedded or vitrified sections of mouse (7) or human skin (8). Both studies showed cadherin dimers interacting through their EC1 domains at the desmosome central dense midline. Desmosomes of mouse skin appeared dense, but structurally less ordered compared with those of human skin, which formed regular 2D arrays of cadherin dimers in the intercellular space. Despite these advances, the specific arrangement of Dsgs and Dscs in these structures could not be derived from the data, as Dsgs and Dscs could not be distinguished. However, the overall structures of desmosomal cadherins we report here suggest that, as a result of differential Ca2+ binding between domains EC3 and EC4, Dsgs and Dscs may adopt subtly different overall conformations that could potentially be Harrison et al.
Dsg:Dsc trans-heterodimers, with the likelihood of a close-packed arrangement consistent with electron tomography observations that may involve cis hetero-interactions between alternating Dsgs and Dscs on each membrane surface. A complete understanding of desmosome extracellular architecture will require additional studies to determine whether extracellular interactions other than the adhesive interactions defined here play a role in desmosome assembly.
distinguished in future studies to shed light on how trans heterodimers are arranged in the desmosome. It is of interest to contrast our structural understanding of desmosomes to that of adherens junctions (24) formed by related type 1 classical cadherins. First, end-to-end distances seen in Dsc and Dsg homodimer structures predict intermembrane distances of ∼326– 329 Å, if perpendicular to the membrane (Fig. 3), which are consistent with measurements of native desmosomes by electron microscopy (280–340 Å; refs. 7 and 8). In contrast, structures of type 1 classical cadherin dimers (24) bend less sharply between EC3 and EC4 and are thus longer than Dsgs and Dscs, but paradoxically mediate shorter intermembrane spacing as a result of the oblique angle at which assembled classical cadherins protrude from the membrane in adherens junctions (24). Second, adherens junctions are formed by a self-assembling 2D molecular layer mediated by cis and trans interactions (24), and we have found that interplay between cis and trans interactions is also crucial to clustered protocadherin function (38). Given the highly ordered structures observed in cryo-EM tomograms (8), it seems likely that cis interactions will also be important for desmosomes. We do not observe cis interactions in our crystal structures. However, if such interactions were formed between Dsgs and Dscs, they would not be expected to be observed in the homomeric structures reported here. Interestingly, the Dsc1 and Dsc2 EC1 domains contain nine and seven conserved aromatic residues distal from the adhesive interface, respectively, which protrude from the surface to give the appearance of an aromatic “pin-cushion” (Fig. 5B and SI Appendix, Fig. S4). The function of these unusual surface-exposed aromatic residues is unknown, but it is possible that they mediate cis interactions by binding Dsgs in desmosome assembly. Overall, our results suggest a desmosome extracellular architecture composed of
ACKNOWLEDGMENTS. We thank Surajit Banerjee, Igor Kourinov, Frank V. Murphy, and Narayanasami Sukumar at APS for synchrotron data collection support. cDNA for Dsc3 was provided by Peter Koch (University of Colorado) and Aimee S. Payne (University of Pennsylvania). This work was supported by NIH Grants R01 GM062270 (to L.S.) and R01 GM030518 (to B.H.) and NSF grant Grant MCB-1412472 (to B.H.). X-ray data were collected at the NECAT beamlines, funded by NIH Grant P41 GM103403.
1. Dusek RL, Godsel LM, Green KJ (2007) Discriminating roles of desmosomal cadherins: Beyond desmosomal adhesion. J Dermatol Sci 45(1):7–21. 2. Garrod D (2010) Desmosomes in vivo. Dermatol Res Pract 2010:212439. 3. Stahley SN, Kowalczyk AP (2015) Desmosomes in acquired disease. Cell Tissue Res 360(3): 439–456. 4. Al-Jassar C, Bikker H, Overduin M, Chidgey M (2013) Mechanistic basis of desmosometargeted diseases. J Mol Biol 425(21):4006–4022. 5. Delva E, Tucker DK, Kowalczyk AP (2009) The desmosome. Cold Spring Harb Perspect Biol 1(2):a002543. 6. Rayns DG, Simpson FO, Ledingham JM (1969) Ultrastructure of desmosomes in mammalian intercalated disc; appearances after lanthanum treatment. J Cell Biol 42(1):322–326. 7. He W, Cowin P, Stokes DL (2003) Untangling desmosomal knots with electron tomography. Science 302(5642):109–113. 8. Al-Amoudi A, Díez DC, Betts MJ, Frangakis AS (2007) The molecular architecture of cadherins in native epidermal desmosomes. Nature 450(7171):832–837. 9. Holthöfer B, Windoffer R, Troyanovsky S, Leube RE (2007) Structure and function of desmosomes. Int Rev Cytol 264:65–163. 10. North AJ, Chidgey MA, Clarke JP, Bardsley WG, Garrod DR (1996) Distinct desmocollin isoforms occur in the same desmosomes and show reciprocally graded distributions in bovine nasal epidermis. Proc Natl Acad Sci USA 93(15):7701–7705. 11. Chitaev NA, Troyanovsky SM (1997) Direct Ca2+-dependent heterophilic interaction between desmosomal cadherins, desmoglein and desmocollin, contributes to cell-cell adhesion. J Cell Biol 138(1):193–201. 12. Getsios S, et al. (2004) Coordinated expression of desmoglein 1 and desmocollin 1 regulates intercellular adhesion. Differentiation 72(8):419–433. 13. Marcozzi C, Burdett ID, Buxton RS, Magee AI (1998) Coexpression of both types of desmosomal cadherin and plakoglobin confers strong intercellular adhesion. J Cell Sci 111(Pt 4):495–509. 14. Chidgey M, et al. (2001) Mice lacking desmocollin 1 show epidermal fragility accompanied by barrier defects and abnormal differentiation. J Cell Biol 155(5): 821–832. 15. Koch PJ, et al. (1997) Targeted disruption of the pemphigus vulgaris antigen (desmoglein 3) gene in mice causes loss of keratinocyte cell adhesion with a phenotype similar to pemphigus vulgaris. J Cell Biol 137(5):1091–1102. 16. Syed SE, et al. (2002) Molecular interactions between desmosomal cadherins. Biochem J 362(Pt 2):317–327. 17. Lowndes M, et al. (2014) Different roles of cadherins in the assembly and structural integrity of the desmosome complex. J Cell Sci 127(Pt 10):2339–2350. 18. Nie Z, Merritt A, Rouhi-Parkouhi M, Tabernero L, Garrod D (2011) Membraneimpermeable cross-linking provides evidence for homophilic, isoform-specific binding of desmosomal cadherins in epithelial cells. J Biol Chem 286(3):2143–2154. 19. Shapiro L, et al. (1995) Structural basis of cell-cell adhesion by cadherins. Nature 374(6520):327–337.
20. Brasch J, Harrison OJ, Honig B, Shapiro L (2012) Thinking outside the cell: How cadherins drive adhesion. Trends Cell Biol 22(6):299–310. 21. Tariq H, et al. (2015) Cadherin flexibility provides a key difference between desmosomes and adherens junctions. Proc Natl Acad Sci USA 112(17):5395–5400. 22. Katsamba P, et al. (2009) Linking molecular affinity and cellular specificity in cadherinmediated adhesion. Proc Natl Acad Sci USA 106(28):11594–11599. 23. Brasch J, et al. (2011) Structure and binding mechanism of vascular endothelial cadherin: A divergent classical cadherin. J Mol Biol 408(1):57–73. 24. Harrison OJ, et al. (2011) The extracellular architecture of adherens junctions revealed by crystal structures of type I cadherins. Structure 19(2):244–256. 25. Harrison OJ, et al. (2012) Nectin ectodomain structures reveal a canonical adhesive interface. Nat Struct Mol Biol 19(9):906–915. 26. Vendome J, et al. (2014) Structural and energetic determinants of adhesive binding specificity in type I cadherins. Proc Natl Acad Sci USA 111(40):E4175–E4184. 27. Seiradake E, et al. (2014) FLRT structure: Balancing repulsion and cell adhesion in cortical and vascular development. Neuron 84(2):370–385. 28. Özkan E, et al. (2014) Extracellular architecture of the SYG-1/SYG-2 adhesion complex instructs synaptogenesis. Cell 156(3):482–494. 29. Messenger AG, Bazzi H, Parslew R, Shapiro L, Christiano AM (2005) A missense mutation in the cadherin interaction site of the desmoglein 4 gene underlies localized autosomal recessive hypotrichosis. J Invest Dermatol 125(5):1077–1079. 30. Lommel M, et al. (2013) Protein O-mannosylation is crucial for E-cadherin-mediated cell adhesion. Proc Natl Acad Sci USA 110(52):21024–21029. 31. Vester-Christensen MB, et al. (2013) Mining the O-mannose glycoproteome reveals cadherins as major O-mannosylated glycoproteins. Proc Natl Acad Sci USA 110(52): 21018–21023. 32. van der Zwaag PA, et al. (2009) A genetic variants database for arrhythmogenic right ventricular dysplasia/cardiomyopathy. Hum Mutat 30(9):1278–1283. 33. Chen CP, Posy S, Ben-Shaul A, Shapiro L, Honig BH (2005) Specificity of cell-cell adhesion by classical cadherins: Critical role for low-affinity dimerization through betastrand swapping. Proc Natl Acad Sci USA 102(24):8531–8536. 34. Häussinger D, et al. (2004) Proteolytic E-cadherin activation followed by solution NMR and X-ray crystallography. EMBO J 23(8):1699–1708. 35. North AJ, et al. (1999) Molecular map of the desmosomal plaque. J Cell Sci 112(Pt 23): 4325–4336. 36. Allen E, Yu QC, Fuchs E (1996) Mice expressing a mutant desmosomal cadherin exhibit abnormalities in desmosomes, proliferation, and epidermal differentiation. J Cell Biol 133(6): 1367–1382. 37. Fujiwara M, et al. (2015) Desmocollin-2 alone forms functional desmosomal plaques, with the plaque formation requiring the juxtamembrane region and plakophilins. J Biochem 158(4):339–353. 38. Thu CA, et al. (2014) Single-cell identity generated by combinatorial homophilic interactions between α, β, and γ protocadherins. Cell 158(5):1045–1059.
Harrison et al.
Materials and Methods
PNAS | June 28, 2016 | vol. 113 | no. 26 | 7165
BIOPHYSICS AND COMPUTATIONAL BIOLOGY
Recombinant Dsg and Dsc ectodomains were expressed in HEK293 cells and purified by nickel affinity chromatography and gel filtration. Biotinylated proteins containing C-terminal AviTag sequences were cotransfected with biotin ligase. AUC experiments were performed in a Beckman XL-A/I analytical ultracentrifuge at 25 °C, using UV detection. For SPR, biotinylated proteins were captured over NeutrAvidin-immobilized surfaces, and binding of analytes at 3, 6, and 12 μM was assessed at 25 °C, using a Biacore T100 biosensor. For bead aggregation, biotinylated cadherins were captured on fluorescent NeutrAvidin-labeled microspheres, and aggregation was assessed by fluorescence microscopy after 1 h mixing. For structure determination, protein crystals were grown by vapor diffusion in hanging drops, and X-ray diffraction data were collected from single crystals at 100 K, using a wavelength of 0.979 Å at the 24-ID-C/E beamlines at Argonne National Laboratory. Structures were solved by molecular replacement using models 3Q2W, 3Q2V, and 1L3W. See SI Appendix, Materials and Methods for details.
