The Dsg3 extracellular domain (ECD) has an extensive surface area of 32133 Å ². This surface area is ~3% larger than the Dsg1 ECD atomic accessible surface (31093 Å ²). The general folds of Dsg1 and Dsg3 ECDs are similar (Figures 1 and 2). In particular, the ECD1, ECD2 and ECD3 of Dsg1 and Dsg3 are well conserved with Cα r.m.s.d. of 1.03 Å, 1.09 Å and 1.94 Å, respectively. The main differences between Dsg1 and Dsg3 ECDs lie in ECD4 (Cα r.m.s.d. = 3.76 Å) and ECD5 (Cα r.m.s.d. = 6.95 Å). In ECD4, the most obvious difference between the structures involve the loop 411–420 of Dsg1 and loop 405–422 of Dsg3. The backbone of Gln410, Ala411 and Ile412 in Dsg1 are intercalated more deeply into the ECD4 interface than the corresponding residues in Dsg3. In ECD5, the loop 468–479 curled upward toward Dsg1, with differences of up to 2.18 Å between the corresponding positions in Dsg3 (loop 468–475). It appears that these regions may not contain high concentration of T-cell epitopes and do not contribute directly to the differences in T-cell specificities of the two PV target antigens (Figure 3). Nonetheless, these structural differences are solvent exposed and may interact with pemphigus autoAb.

The locations of Dsg1 and Dsg3 immunological hotspots are highly conserved across the ECD, TM and intracellular domain (ICD), despite limited sequence identity (54% identical, 72% similar). There are substantial overlaps between predicted and known immunological (T-cell epitope) hotspots (Dsg3 145–192, Dsg3 240–303, Dsg3 570–614) at the threshold -123 kJ/mol (Figure 2). At this threshold, the number of predicted Dsg3 hotspots are ten (residues 21–88, 125–147, 173–221, 245–299, 330–391, 435–456, 495–522, 584–640, 772–797, 830–859). Among these, one spans the signal peptide (SP), propeptide (PP) and the ECD; six are present in the ECD; one extends across the ECD, TM and ICD; while two hotspots are predicted to exist in the ICD (Figure 2). Similarly, nine hotspots were predicted for Dsg1 (residues 1–62, 82–151, 173–240, 310–372, 423–455, 470–498, 550–569, 656–688, 732–760), all of which overlap with those from Dsg3 (Figure 2). Collectively, these results suggest that the potential T-cell epitope repertoire encoded in Dsg1 and Dsg3 is substantially overlapping, and may help to explain the molecular basis underlying the observed inter-molecular spreading from Dsg3 to Dsg1 targets during the course of PV.

We investigated the effect of multiple contact regions or binding registers in Dsg3 peptides specific to DRB1*0402 33. Among 985 Dsg3 peptide sequences (including SP and PP derived sequences), 658 were predicted high-affinity binders with 77% containing two or more registers that can be docked into the binding groove of DRB1*0402. A similar number of high-affinity binders are predicted to exist within the Dsg1 glycoprotein, with analogous distribution of binding registers as illustrated in Figure 4. Of 1035 Dsg1 peptides, 665 were predicted high-affinity binders with 129 (~23%), 112 (~19%), 130 (~17%), 99 (~15%) and 43 (~6%) peptides possessing two, three, four, five and six registers, respectively.

The predicted Dsg1 high-affinity binding sequences were examined for their similarity with the Dsg3 proteome. Each Dsg1 (15 mer) sequence was used to probe the entire Dsg3 proteome for the highest identity Dsg3 (15 mer) peptide with the minimal number of substitutions. Figure 5 details the degree of conservation of Dsg1 predicted high-affinity binders with Dsg3 sequences. All predicted Dsg1 and Dsg3 (15 mer) binding peptides share at least four common residues along their primary sequences. It has been reported that PV and PF autoAbs can cross-react with Dsg1 and Dsg3 peptides with 75% identity 38. Sequence alignment showed that 18% (or 122) of these peptides are highly conserved with at least 75% sequence identity. In this context, two peptides Dsg1_58–72 CREGEDNSKRNPIAK and Dsg1_59–73 REGEDNSKRNPIAKI near the N-terminus of ECD1 appear to be of particular interest, since they are fully conserved within the Dsg3 proteome and may represent the most likely antigenic link between self-directed responses to two distinct autoantigens (Dsg3 and Dsg1). A subset of patients with PV has been shown to have T cell reactivity to both Dsg3 and Dsg1 39. However, defined Dsg1 epitopes have not yet been determined.

The program MODELLER 45 was employed for comparative modeling of DRB1*0402, Dsg1 and Dsg3. Each model was constructed by optimally satisfying spatial constraints obtained from the alignment of the template structure with the target sequence and from the CHARMM-22 force field 46, and subsequently relaxed by conjugate gradient minimization, using the program Internal Coordinate Mechanics (ICM) 4748.

Overlapping 15 mer peptides are generated from the Dsg1 and Dsg3 sequences. An overlapping sliding window of size nine is applied to each 15 mer peptide to generate all combinations of binding registers to be modeled into the binding groove of DRB1*0402 (Figure 6). A total of 1035 Dsg1 15 mer peptides (6210 binding registers) and 985 Dsg3 15 mer peptides (5910 binding registers) were generated and used in the current analysis. Docking was performed using a standard protocol as previously described 333435, consisting of (i) pseudo-Brownian rigid body docking of peptide fragments to the ends of the binding groove, (ii) central loop closure by satisfaction of spatial constraints, (iii) refinement of the backbone and side-chain atoms of the core recognition residues and receptor contact regions within 4.00 Å radius, and (iv) extension of flanking peptide residues by satisfaction of spatial constraints (Figure 7).

The scoring function presented in this study is based on the free energy potential in ICM 4748. Computation of the binding free energy was performed according to previous similar works 3637 based on the difference between the energy of the solvated complex and the sum of the energy of the solvated receptor and that of the peptide ligand, followed by optimization using experimental IC₅₀ values. The DRB1*0402 model optimization, training and testing were described earlier 33.

In the present study, 'immunological hotspots' are defined as antigenic regions of up to 30 amino acids based on the sum of predicted binding energies of the top four binders within a window of 30 amino acids 4950. Where available, predicted hotspots were validated using available experimentally determined sites (Figure 3).

Solvent accessible surface areas were calculated with the program NACCESS 51 by tracing out the maximum permitted van der Waals' contact that is covered by the center of a water molecule (1.40 Å probe radius) as it rolls over the surface of the protein.