The first aim was to find out if the composition of desmosomes in cultured cells is comparable to that of epidermal desmosomes. The composition of desmosomes in the HaCaT cell line, as well as in primary NHEK cells, was mostly comparable to the composition of epidermal desmosomes located in the spinous layer (SL) as published previously (49). It was furthermore relatively similar between the two cell types, except for the colocalization of Dsg3 with Dp ( Figure 1 ).

To confirm the protective effect on cell adhesion of each mediator to be used in this study the dispase-based dissociation assay was performed in HaCaT cells. These include inhibitors for the cell stress kinase p38MAPK, SB202190 (SB20) blocking the ATP-binding site (52) and EO1428 (EO) blocking the hydrophobic pocket, important for substrate binding and autophosphorylation (53); PP2 a potent inhibitor of Src-family kinases via a complex interaction, partially competing with ATP-binding (54); U-73122 inhibiting the PLC by blocking the binding of its upstream activating G-protein ligant (55). U0126 which inhibits the activation of MEK1/2 and increasing its dephosphorylation rate (52). Forskolin and activator of adenylate cyclase increasing activity by binding inside the catalytic pocket (56). Rolipram which inhibits degradation of cAMP by blocking Phosphodiesterase-4 activity by competitively binding to the catalytic domain, but also via a more complex allosteric binding mechanism (57). All mediators were effective in ameliorating the loss of adhesion induced by PV-IgG. None of the mediators showed any effect on baseline adhesion when applied with C-IgG. In contrast, activation of p38MAPK using Anisomycin (Aniso), induced activating phosphorylation of p38MAPK, and downstream ERK1/2, most likely via a general stress response (58–60), for 1 h induced severe fragmentation with about 2000 fragments and thus a factor of 100 more than PV-IgG (Figures 2A, B). It was confirmed that the fragmentation is not caused by cytotoxicity, using MTT vitality staining before sheering, producing a bright purple coloration of the monolayers under all conditions ( Figure 2A ). We compared cell adhesion after autoantibody treatment with cell adhesion after Ca²⁺-depletion. Complete disruption of cadherin-mediated adhesion by Ca²⁺-depletion for either 30 min or 1 h yielded around 2000 fragments with no significant difference between 2.5 mM or 5 mM EGTA. Repletion of Ca²⁺ for 24 h normalized adhesion for both concentrations of EGTA ( Figures 2C–E ).

Evaluation of Dp plaque ultrastructure using STED microscopy revealed that PV-IgG-induced several alterations compared to desmosomes under control conditions ( Figures 3A, B ). Both, the number of desmosomes per µm of cell border length and desmosome size were significantly decreased ( Figures 3A , 4 , 5A–D ). The plaque thickness was increased ( Figures 3A , 4 ), and some desmosomes showed asymmetry between plaques of neighboring cells ( Figure 5E ) indicating plaque disorganization. The distance between the desmosomal plaque sub-cells was increased ( Figures 3A , 4 ). In regions where the cell borders had already drifted apart, half/split desmosomes were observed ( Figure 3A , green arrowheads). At some cell borders either single desmosome-like structures in the cytoplasm or linear streaks with desmosome-like structures arranged in lines were present inside the cytoplasm, indicating desmosome remodeling ( Figure 3A , red arrowheads/circles). Lastly, in some areas, clusters of Dp staining were present ( Figure 3A , blue arrowhead). After activation of p38MAPK with anisomycin, similar effects were observed for all parameters of desmosome ultrastructure. However, clustering of Dp staining was absent ( Figures 3C , 4B , 5A–D ).

The inhibition of p38MAPK with either SB20190 (SB20) or EO1485 (EO) but also inhibition of PLC with U-73122 ameliorated all PV-IgG-induced ultrastructural alterations and even slightly increased the number of desmosomes ( Figures 4A, C , 5C, D ). Inhibition of MEK1/2, which is upstream of Erk1/2, with U0126, ameliorated the loss of number of desmosomes but only partially restored desmosome size. Increased plaque distance was rescued but increase of plaque thickness was only partially reduced ( Figures 4D , 5A–D ). Both, inhibition of Src via PP2 and increase of cAMP levels using F/R strongly increased the number of desmosomes and desmosome number was not decreased by PV-IgG. However, desmosome size was decreased to a similar degree as after incubation with PV-IgG. Both plaque distance and thickness were protected after addition of PP2 or F/R ( Figures 4E, F , 5A, B, G ). Under both conditions, clusters of delaminated cells, with even more desmosomes than the surrounding cells were present ( Figure 5G ). The desmosomal plaque asymmetry induced by PV-IgG and anisomycin was abolished by all mediators ( Figure 5E ).

PV-IgG- and p38MAPK activation-induced alterations, including decreased desmosome number, increased plaque distance and thickness and loss of adhesion, were comparable to features observed for newly formed immature desmosomes, after depletion of Ca²⁺ for 30 min or 1 h with 2.5 or 5 mM EGTA as described above. Repletion for 24 h completely restored the normal ultrastructure of desmosomes ( Figure 6 ).

Both PV-IgG and anisomycin induced retraction of keratin filaments from cell borders toward the nucleus ( Figure 5F ). Keratin retraction induced by PV-IgG was prevented by all mediators except U0126, which only reduced it. Alterations induced by PV-IgG and anisomycin furthermore included fragmented staining of Dsg1 and Dsg3 ( Figures 7 – 9 ), with mostly desmosomal Dsg staining remaining as visible by the keratin insertion ( Figures 8B , 9B , white arrowheads). Anisomycin as well as U0126, PP2 and F/R increased Dsg1 staining at the membrane and anisomycin ( Figures 7G, H ), PP2 and F/R also enhanced Dsg3 staining intensity both at the membrane and in the cytoplasm ( Figures 7G–J ). The fragmentation of Dsg1 and Dsg3 staining was normalized by all mediators, only exception was U-73122 which reduced Dsg3 fragmentation but did not normalize it ( Figures 7A, C, F , 8 , 9 ). U-73122 did also not normalize Dsg3 staining intensity ( Figure 7 ). p38MAPK activation by anisomycin caused an increased Dsg intensity both at cell borders and in the cytoplasm ( Figures 7G–J ).

First important step was to check whether the findings on desmosome ultrastructure in cultured cells are comparable to the situation in keratinocytes of human epidermis. For this we verified that the desmosomes in both the HaCaT cell line and primary NHEK cells are comparable to desmosomes found in epidermis of body doners and control sample from non-pemphigus patients ex vivo. Overall, desmosomes in both cell types investigated represented epidermal desmosomes found in the SL of the epidermis quite well. SL desmosomes showed the overall highest colocalization for Dsg3+Dp, Dsg1+Dp and Pg+Dp compared to the other layers, and very similar to the cells, while Dsg1+Pg was similar across all layers. The only difference was that both HaCaTs and NHEKs were more homogeneous with respect to Dsg1/Dsg3 composition whereas in the skin, desmosomes mostly contain either Dsg1 or Dsg3, with the highest degree of colocalization of the to in the basal layer (BL) (49). The only difference between HaCaTs and NHEKs was the colocalization of Dsg1 and Dsg3 with Dp. NHEKS showed a bit less colocalization of Dsg3 with Dp, while HaCaTs showed a bit lower colocalization of Dsg1 with Dp. Both were respectively significantly different to the SL composition. Since the higher colocalization of Dsg3 and lower colocalization of Dsg1 with Dp is more similar to the BL, HaCaTs apparently represent desmosomes at the basal-suprabasal interface of the epidermis. Because autoantibodies in PV induce splitting at this interface, all further experiments on regulation of the desmosome ultrastructure were performed with HaCaT cells (61, 62).

It was previously shown that the composition of desmosomes was not drastically altered regarding Dsg1, Dsg3, Pg and Dp in PV patient skin, compared to controls, neither in lesional, nor in non-lesional regions. However, the number of desmosomes per membrane length was significantly reduced, causing reduced adhesion and blister formation (19, 49). Electron microscopy and recently also STED and dSTORM microscopy further revealed PV-IgG-induced alterations of the desmosomal plaque. These also include reduced desmosome number (19, 63, 64) but furthermore reduced plaque size (19, 20, 33, 63–66), increased inter plaque distance (19, 20), increased plaque thickness and/or disorganization (4, 19), formation of split desmosomes (33, 63, 64) as well as altered keratin organization and insertion (4, 19, 63). These alterations can be concluded to be both a good indicator for, as well as a cause for the reduced desmosome adhesion. Using STED microscopy, we demonstrated that most alterations of the desmosome ultrastructure were comparable after incubation with PV-IgG or following activation of p38MAPK via anisomycin conforming a previous study (19). We found that the plaques of affected desmosomes are often asymmetrical. In contrast, clustering of desmosomal components was observed only after treatment with PV-IgG, identifying this effect to be p38MAPK-independent. Most of these alterations, including decreased desmosome number, increased interplaque distance and thickness as well as loss of adhesion closely resemble features found in newly assembled immature desmosomes as observed after Ca²⁺ depletion, also confirming previous reports (20, 67, 68). This strongly suggests that PV-IgG-induced signaling stimulates desmosome assembly pathways and reverts the desmosome ultrastructure back to a state similar to that of immature desmosomes or desmosome precursors. This also fits to previous reports about the importance of several PV-IgG-induced signaling pathways responsible for desmosome assembly and maturation, including PKC (69–72), RhoA (39, 73–75), Src (25, 35, 39), and Rap1 (76).

Next we wanted to investigate in more detail how modulation of different signaling pathways involved in pemphigus pathogenesis affected the desmosome ultrastructure. It is well known that p38MAPK is involved in pemphigus pathogenesis (18). Similar to previous reports, our results demonstrate that p38MAPK is directly involved in loss of adhesion (31, 32). We also confirm the involvement of p38MAPK in keratin retraction (28–30, 65) and alterations of Dsg3 distribution (21, 26, 27). Additionally, we observed that p38MAKP affected the distribution of Dsg1 in a similar fashion. p38MAKP inhibition ameliorated all alterations to the desmosomal plaque ultrastructure, which fits to previous findings using electron microscopy, reporting reduced number of desmosomes, decreased desmosome size, interdesmosomal widening, split desmosomes and altered keratin insertion (65). The slight increase in the number of desmosomes further indicates that also under basal conditions p38MAPK negatively regulates desmosome assembly.

We further investigated the role of PLC. Similarly to p38MAPK, PLC seems to regulate all parameters of desmosome ultrastructure as well as keratin filament organization. This mostly fits to previous reports showing that inhibition of PLC and its downstream targets was protective against keratin retraction and loss of adhesion (34, 63). For Desmosome number, a previous study reported a similar trend in human skin ex vivo. However, with not significant difference, possibly due to the weak effect and high interindividual variability in the investigated ex vivo skin (17, 63). This may also explain why desmosome length was not rescued. Interestingly, inhibition of PLC fully prevented fragmentation of Dsg1 staining whereas effects on Dsg3 staining were less pronounced. That might indicate that PLC impacts extradesmosomal Dsg1 more than extradesmosomal Dsg3. However, the at least partial rescue of both confirms previous findings (34).

The data are in line with the hypothesis that both p38MAPK and PLC are relatively far upstream in the signaling mechanisms involved in desmosome dysregulation in PV (18). In contrast, inhibition of MEK1/2, downstream of Src, p38MAPK and EGFR (18), improved but did not fully rescue desmosome size, plaque thickness or keratin retraction, which may indicate that MEK1/2 is further downstream in the signaling mechanisms after autoantibody binding and may regulate desmosome turn-over at specific steps only. While loss of number of desmosomes and plague distance were rescued, the desmosome size and plaque thickness were only partially normalized. This fits well with results from a previous electron microscopy study showing that neither desmosome size, desmosomal splitting, nor keratin dissociation were rescued in human skin ex vivo (17, 66).

Both, inhibition of Src and increase of cAMP levels showed very distinct effects compared to the signaling pathways outlined above. Both signaling pathways rescued most desmosomal parameters and keratin insertion which is similar to inhibition of p38MAPK and PLC. However, under control conditions Src inhibition and cAMP increase resulted in a higher number of smaller desmosomes and increased staining of Dsg1 and Dsg3 along the membrane and in the cytoplasm. A possible interpretation would be that desmosome assembly was accelerated which prevented desmosomes to fully mature. The increased Dsg3 staining could result from an increased amount of Dsg3 containing desmosomal precursors. This fits well with previous reports on the protective role of Src and cortactin described to regulate the integration of desmosome precursors into desmosomes (25, 35, 39). In contrast to that, Src was not protective in preventing blister formation in human skin ex vivo and did not resuce desmosome number, size nor inderdesmosomal widening, but was effective in preventing blister formation in neonatal mouse skin in vivo (17, 39). This could possibly indicating, that for Src inhibition to be protective more dynamic reorganization needs to be in process. Src for example acts via the epidermal EGFR (36–38), which might show a different regulation in differentiated human skin than in cultured cells. It was furthermore reported, that Src was no longer active in skin after 24 h and Src inhibition was only protective during the timeframe, while it was active (39). This partially confirms the need for active reorganization processes to be going on for Src inhibition to be protective. For cAMP, these observations also conform to a previous report demonstrating its protective effects in mice in vivo and keratinocytes in vitro, for which keratin insertion were rescued, while desmosome size was not rescued, but in contrast to in vitro findings also not reported to be reduced (42). That the size of desmosomes was not reduced might, similarly as for Src, indicate, that differentiated adult human skin shows less dynamic desmosome reorganization than cells in vitro. Another study demonstrated that Rap1 downstream of cAMP affects plaque formation and thus desmosome assembly via Pkp3 (76), which also fits well to this hypothesis.

From our data, it can be concluded that modulation of the pemphigus signaling network downstream of autoantibody binding at different points may allow to rescue desmosome adhesion as has been shown in many studies before (18). However, in the recent study we observed that the effects of signaling pathways on the desmosomes ultrastructure may be very different which may indicate the role of these signaling mechanisms in the different steps of desmosome turn-over and the hierarchical position of the respective molecules in the pemphigus signaling network. Molecules more upstream in the corresponding signaling chain and thus activated earlier, upon autoantibody binding, such as p38MAPK and PLC affect all parameters of desmosome ultrastructure whereas others, including MEK1/2, may be more downstream of autoantibody binding or involved only in distinct steps of desmosome turn-over. For example, MEK1/2 appears to stabilize mature desmosomes but is less important for the regulation of cytoskeletal anchorage of desmosomes. In contrast, we conclude that both Src and cAMP most likely are involved in the regulation of desmosome assembly.

These new insights may be important in the future to establish new treatment options in pemphigus. At this stage, treatment is limited to immunosuppression and B cell depletion which is associated with side effects and considerable relapse rates and new experimental strategies also focus on the reduction of autoantibody levels (77). Therefore, new treatment options are required (78) and we proposed to stabilize keratinocyte adhesion by promoting desmosome adhesion for example via apremilast to enhance cAMP levels or erlotinib to inhibit EGFR (19, 42). Since in this context, Src families associated with EGFR activation were found to be activated by autoantibodies, our new data indicate that therapy options to enhance desmosome assembly may be effective to treat pemphigus in the future.

The human epidermal keratinocyte cell line HaCaT was cultivated in a humidified, 5% CO₂ atmosphere at 37°C in Dulbecco’s Modified Eagle Medium (DMEM) (Life Technologies, CA) with 10% FCS (Merk, DE) and 50 U/ml penicillin and 50 µg/ml streptomycin (both AppliChem, DE). Primary normal human epidermal keratinocytes (NHEK) in passages 2-6 in CnT-Prime Basal Medium 1 at low calcium (0.06 mM) until confluency and 1.8 mM calcium for differentiation for 24 h after reaching confluency. Passaging and seeding was performed, using trypsin-EDTA solution (Merck, DE). For Ca²⁺ depletion the cells were treated with EGTA 2.5 or 5 mM for 30 min or 1 h. Repletion was performed by washing with PBS 1x and exchanging to new medium.

Cells were pre-incubated with PP2 for 2 h or all other mediators for 1 h before IgG treatment for 24 h. Anisomycin was incubated for 1 h before fixing the cells. For concentration of mediators refer to Table 1 , for staining antibodies refer to Table 2 .

PV-IgG with ELISA scores for anti-Dsg1-IgG: 1207.00 U/ml and anti-Dsg3: 3906.00 U/ml was obtained as immunoapharesis purified material from the Lübeck Institute of Experimental Dermatology. Protein A agarose beads (Thermo Fisher Scientific, USA) were used for affinity purification to obtain IgG fractions from sera of healthy volunteers as control-IgG (C-IgG), as described previously (24). IgGs were used 1:50 in cell culture experiments and about 50 µl/cm² of skin ex vivo.

Confluent HaCaT monolayers were washed with Hank´s buffered saline solution (HBSS). 150 µl Dispase II, 2.4 U/ml (Sigma Aldrich, USA) in HBSS was added for 10 min at 37°C, 95% humidity. After monolayer detachment the dispase reaction was stopped by diluting with 200 µl HBSS. The monolayers were stained with MTT (VWR, DE) to confirm viability and produce a better contrast. An electrical pipette was used to induce defined shear stress. Fragment numbers were counted using images taken, using a binocular microscope (Leica, DE) and an EOS 600D camera (Canon, JP).

Skin biopsies from body donors deceased for less than 24 h were used. Each epidermis piece ~6 cm² was excised from the shoulder region and divided into ~1 cm² pieces. The skin was cultivated in floating culture, for 24 h after the second injection, in 6-well-plates in 4 ml medium under the same conditions as the keratinocytes. After 24 h, the samples were cut into shape about ~0.4 cm² embedded in tissue freezing medium (Leica, DE). 7 μm sections were produced, using a CryoStar™ NX70 Kryostat (Thermo Fisher, USA).

Treated cells grown on high performance glass cover slips or skin slices were fixed using -20°C ethanol for 30 min followed by -20°C acetone for 3 min. Primary antibodies were incubated for 3 h at RT and secondary goat-anti-mouse or -rabbit antibodies coupled to STAR-RED or Alexa 594 fluorophores (Abberior GmbH, Göttingen, DE) for 1 h. DAPI (Roche, DE) was added for the last 15 min of secondary antibody incubation. Using Prolong™ Diamond Antifade Mountant, cover slips were mounted to glass slides (Thermo Fisher Scientific, DE). An Abberior 3D Stimulated emission depletion (STED) confocal microscope with IMMOIL-F30CC (Olympus, JP) was used to image the samples. Star Red was excited at 638 nm and Alexa594 at 594 nm, using a pulsed diode lasers (PDL 594, Abberior Instruments; PiL063X, Advanced Laser Diode Systems). Fluorophores were depleted using a pulsed fiber laser (PFL-P-30-775B1R, MPB Communications) at 775 nm. Emission was detected with an avalanche photodiode detector at 605-625 nm for Star Red and 650-720 nm for Alexa 594.

Image processing was performed with Photoshop CS5. Image J was used to quantify fluorescence staining (IF). Each data point represents an average per STED image. All images were evaluated, using Fiji Image J software v 2.9.0. Data were analyzed using either two-way-ANOVA followed by Sidak-post-hoc-test for multiple comparison or if applicable one-way-ANOVA followed by Dunnett-post-hoc-test for multiple comparison, using GraphPad Prism v8.4.3 (GraphPad Software, USA). Significance was assumed for p ≤ 0.05. Data are shown as mean ± standard error of the mean. Each N represents one independent experiment, each n a technical replicate.