Cadherin flexibility provides a key difference between desmosomes and adherens junctions

Abstract

Desmosomes and adherens junctions are intercellular adhesive structures essential for the development and integrity of vertebrate tissue, including the epidermis and heart. Their cell adhesion molecules are cadherins: type 1 cadherins in adherens junctions and desmosomal cadherins in desmosomes. A fundamental difference is that desmosomes have a highly ordered structure in their extracellular region and exhibit calcium-independent hyperadhesion, whereas adherens junctions appear to lack such ordered arrays, and their adhesion is always calcium-dependent. We present here the structure of the entire ectodomain of desmosomal cadherin desmoglein 2 (Dsg2), using a combination of small-angle X-ray scattering, electron microscopy, and solution-based biophysical techniques. This structure reveals that the ectodomain of Dsg2 is flexible even in the calcium-bound state and, on average, is shorter than the type 1 cadherin crystal structures. The Dsg2 structure has an excellent fit with the electron tomography reconstructions of human desmosomes. This fit suggests an arrangement in which desmosomal cadherins form trans interactions but are too far apart to interact in cis, in agreement with previously reported observations. Cadherin flexibility may be key to explaining the plasticity of desmosomes that maintain tissue integrity in their hyperadhesive form, but can adopt a weaker, calcium-dependent adhesion during wound healing and early development.

Keywords: adhesion; cadherins; desmosomes; small-angle X-ray scattering; structure.

Fig. 1.

Structural and biophysical analyses of mDsg2. (A) Fitted distribution of the sedimentation coefficients for Dsg2-EDTA (blue) and Dsg2-Ca (red). The higher sedimentation coefficient in the absence of Ca²⁺ indicates a more compact particle, consistent with the lower R_h of 4.1 nm and a f/f₀ of 1.54 compared with the sample with Ca²⁺ (R_h, 4.6 nm; f/f₀, 1.72). Representative EM images of Dsg2-EDTA (B) and Dsg2-Ca (C) samples after negative staining. The Dsg2-EDTA particles appear as compact globules, whereas the Dsg2-Ca particles display a range of morphologies, some of them bent or curled at the ends. (Scale bar, 20 nm.) (D) The p(r) distribution plot for Dsg2-EDTA (blue) and Dsg2-Ca (red) shows D_max of 140 and 175 Å, respectively. SAXS profiles showing the log of X-ray scattering intensity (logI) as a function of the scattering vector q for the experimental scattering data Dsg2-EDTA (blue) (E) and Dsg2-Ca (red) (F). The normalized fit to the experimental data (from the program GNOM) is superimposed as a black line). (Inset, top right) Guinier plot (logI vs. q²) of the low q region of the X-ray scattering data where the radius of gyration (R_g) can be measured from the gradient of the slope $\left(\mathrm{-}{R}_g^2/3\right)$. (Inset, bottom left) Average DAMMIN models for Dsg2-Ca (red) and Dsg2-EDTA (blue).

Fig. 2.

EOM ensemble analysis. (A) R_g distribution of the random pool and selected ensembles for Dsg2-EDTA (blue) and Dsg2-Ca (red) generated with EOM. Ab initio shape restoration of Dsg2-Ca (B) and Dsg2-EDTA (C); superimposed are the models of the final ensemble.

Fig. 3.

Pool filtering analysis. Plots of R_g vs. sedimentation coefficients (s) as calculated with SoMo for the Ranch model pool (A) and the TAMD model pool (B). Dots represent the pairwise values for each individual model generated. The range of experimentally determined values for s (±1σ) is shown as red boxes for Dsg2-Ca and blue boxes for Dsg2-EDTA. (C) R_g distribution of EOM ensembles for the filtered subsets generated by Ranch (solid bars) and by TAMD with CNS (open bars).

Fig. 4.

Calcium coordination in Dsg2 is not conserved. (A) Table listing the residues involved in EC3–EC4 and EC4–EC5 interdomain calcium coordination. Nonconservative substitutions in mDsg2 and hDsgs are in red; residues located too far for proper coordination are in brackets. (B) Side chain coordination of one of the three Ca²⁺ ions at the EC3–EC4 region; residues coordinating the same Ca²⁺ ion (purple) in N-cadherin (PDB ID code 3Q2W, orange chain) and in the mDsg2 homology model in green. (C) Coordination of one of the three Ca²⁺ ions at the EC4–EC5 region; the two Ca²⁺ binding residues D470 and D525 located in loop regions are too distant from the calcium ion, leaving only residues N444 and D472 available for coordination. Gray spheres are the other two Ca²⁺ ions.

Fig. 5.

Fitting of the ab initio model into the desmosome tomography maps. (A) The electron density maps derived from the ET of the human native epidermal desmosome (14) were used to fit the DAMMIN model for Dsg2-Ca. Excellent fit was obtained, as shown in B, where four DAMMIN models illustrate the trans dimer interactions across the midline (MD). (C) Side view of the four DAMMIN models shown in B. Side (D) and top (E) view projections of an idealized array generated with the DAMMIN models from the fitting into the ET maps. A midline (MD) is clearly visible on the side view projection.