Desmosomes have been studied extensively in recent years in consequence of diverse disease phenotypes including genetic and acquired diseases as well as autoimmune or microbial-mediated diseases which weaken intercellular adhesion (54).

Gene mutations involving obligate desmosomal proteins such as Dsg, Dsc, or Dp may bring about a wide spectrum of genetic diseases of the skin and other tissues in which the proteins are strongly expressed (55–57). In the skin, mutations of Dsg1 have been shown to cause severe dermatitis, multiple allergies, and metabolic wasting (SAM syndrome), in which loss of Dsg1 besides epidermal splitting causes a profound alteration of the epidermal barrier as well as the immune system, as shown in Dsg1-deficient animal models (58–61). In humans, mutations in desmosomal components are also associated with heart diseases. Mutations in Dsg2 and Dsc2 are known to cause arrhythmogenic right ventricular cardiomyopathy whereas those in Pg and Dp cause cardiomyopathy as well as various cardio-cutaneous syndromes (62–65). Some aspects in the pathophysiology of arrhythmogenic right ventricular cardiomyopathy may be similar to pemphigus due to common aspects in the structure and regulation of desmosomal contacts in cardiac intercalated discs and desmosomes of the epidermis (66). Moreover, mutations in desmosomal proteins have been shown to affect skin appendages. For example, homozygous and heterozygous mutations of Dsg4 have been reported to be associated with a spectrum of disease phenotypes such as hypotrichosis and monilethrix (67–69). Similarly, genetic alteration in Dsc3 was identified as the underlying cause for hair loss and vesicle formation in skin (70). Mutations in Dsg3, although not known in humans so far, caused diverse clinical features, including severe skin and mucosal lesions and hair loss as in squeaking (sqk) phenotype mice (71).

Microbial and viral agents are among the extrinsic factors that alter expression of desmosomal proteins (72). Exfoliative toxin A produced by staphylococcus bacteria is known to target Dsg1 and proteolyzes its adhesive ectodomain resulting in a PF-like lesion as in a bullous impetigo or staphylococcal-scalded skin syndrome (73) An adenovirus known for affecting the epithelial lining of the respiratory and urinary tracts was identified to bind to Dsg2 destabilizing cell-cell attachment (74). Studies in other acquired diseases such as cancer have revealed dysregulation of desmosomal proteins in tumor cells. For example, Dsg1 is downregulated in squamous cell carcinoma whereas Dsg3 is upregulated in head and neck carcinoma (57). Moreover, some molecular mechanisms underlying desmosome dysregulation in cancer cells have been reported (75, 76). Finally, alterations in desmosome ultrastructure have been detected in patients suffering from inflammatory bowel disease (77, 78). Because animal models deficient for Dsg2, Dsc2, and Dp have been shown to have intestinal epithelial barrier defects and disturbed wound healing and are prone to colitis (79–82), several lines of evidence indicate that disturbed desmosomal adhesion contributes to the pathogenesis of inflammatory bowel diseases (83). In this respect, a new function of desmosomes has been elucidated as they control tight junction structure and function (84, 85).

Isolated cell lines from humans and murine sources have been utilized to investigate the pathogenic effects of pemphigus autoantibodies in vitro. The most extensively used epidermal cells are immortalized human keratinocyte (HaCaT cells: Human adult high Calcium low Temperature) (116) and primary human or mouse keratinocytes. Both cells were utilized mainly as monocultures but were also used to construct a three- dimensional (3D) organ culture model representing a stratified and differentiated epidermis (117). Both cell types have pros and cons in terms of originality, reproducibility, cost, and so on (118). The models are vital to assess the level of pathogenicity of autoantibodies derived from PV patients through various functional dissociation assays as well as to determine the effects of different pharmacological mediators in response to the autoantibodies (119–121).

Animal models are used extensively in medical research. Several mouse models have been developed which rendered great insights into PV pathogenesis (122). Based on the method in which the disease is induced and persistence of clinical features, the models can be identified as active or passive. Passive animal model refers to administration of pemphigus autoantibodies into a healthy animal which results in production of a transient disease phenotype (123, 124). The active models represent the generation of animals which manifest the disease through genetic modification as in Dsg3 or Dsg1 knockout mice (9, 125) or immunization in which the mice actively produce antibodies against Dsg3 (90, 126). In the latter, splenocytes from Dsg3 knockout mice, following immunization with recombinant Dsg3, were transferred into Rag2 knockout mice expressing Dsg3 (126–128). This model is more relevant as it involves active production of anti-Dsg3 autoantibodies similar to the situation in PV patients, whereas Dsg3 knockout mice produce a disease phenotype which is an ultrastructural correlate of only the acute phase of the disease (129).

Organ culture has been an important experimental model for studying pemphigus since the pioneer work done by Schiltz and Michel (130) in which they placed a skin biopsy on a lens paper floating on a liquid medium containing unpurified PV sera. This model is of paramount importance because of the viability of the tissues at optimal time point, 24 h, as well as the potential of the method to reproduce all the major histological and clinical features of the disease. This model stands out to be better than animal models since it enables overcoming the genetic and immunological differences which otherwise would elicit respective species-specific autoreactions (120). Furthermore, it favors the assessment of the pathogenicity of a given autoantibody and helps to correlate to disease activity in patients to develop more specific therapeutic strategies (120, 121). Although organ culture models are not best suited for biochemical studies, they provide a physiologically relevant setup to investigate mechanisms causing altered expression of desmosomal proteins and the resulting acantholysis (131, 132) which closely reflects the human in vivo situation (115). Thus, the model is ideal to conduct experiments under controlled conditions with sizeable samples which otherwise are not feasible to apply to human subjects. Therefore, we employed a human skin organ culture model as well as a novel mucosa ex vivo model to investigate the role of various signaling molecules in PV pathogenesis (32, 59, 133–136). Large blisters with associated ultrastructural changes in desmosomes including reduction in desmosome density and size as well as formation of split desmosomes and keratin filament uncoupling from desmosomes was observed in samples treated with PV-IgG (Figures 2A–C) (135) which shows that the skin model reflects the ultrastructural hallmarks known from pemphigus patients’ lesions (137, 138). Interestingly, the outer and inner plaque could not be differentiated after treatment with PV-IgG, an observation which requires further attention, especially because it has recently been shown that reorganization of the desmosomal plaque occurs during desmosome maturation and it is conceivable that these events may be reverted in pemphigus pathogenesis (139). Also, it must be noted that PV-IgG containing both autoantibodies against Dsg1 and Dsg3 were required for acantholysis whereas mucosal-dominant PV-IgG with autoantibodies against Dsg3 but not against Dsg1 similar to AK23, which is specific for Dsg3, were not sufficient (135). Taken together, the model better reflects the situation in patients compared to mice where high concentrations of anti-Dsg3-specific IgG are sufficient to cause skin blisters (114, 140).

Although not fully unraveled, steric hindrance and signaling have been proposed as the major overarching pathomechanisms that drive loss of intercellular contacts downstream of PV-IgG binding to keratinocytes (52, 141). Both mechanisms are believed to be involved, but not strictly independent of each other (57), and the exact chronology of events as well as contribution of each remains unknown.

Passive transfer of IgG from PV patients or anti-Dsg3 monoclonal antibodies to a healthy neonatal mouse has been shown to induce epidermal blisters (90, 94) which has been reproduced in human skin organ culture, as well (142). These pathogenic immunoglobulins were shown to preferentially target the EC1 and EC2 adhesive regions of Dsg3 (90, 92, 94, 143). The latter has been indicated to predominatly contain epitopes recognised by PV autoantibodies (92, 144, 145). As a result, it has been suggested that PV-IgG autoantibodies directly interfere with the adhesive interaction of Dsg, thereby triggering the initial events leading to acantholysis (90, 127). AFM studies on cell-free surfaces (45, 46, 146) or on living keratinocyte cell surfaces, combined with bead assays, demonstrated direct inhibition of Dsg interaction for Dsg3 but not for Dsg1 (45, 53, 146), but suggested that direct inhibition is not sufficient to cause complete loss of keratinocyte adhesion. In addition, a recent study using bead assays under cell-free conditions showed that PV-IgG and PF-IgG blocked the heterophilic interaction between Dsg3/Dsc3 and Dsc1/Dsc1, respectively (43, 106).

The first signaling pathway to be triggered by pemphigus autoantibodies was PLC-mediated influx of Ca²⁺ (147). Along the years, several studies have shown that steric hindrance alone does not adequately account for acantholysis (45, 111, 146, 148–150), which indicated the involvement of other mechanisms underlying the pathogenic effect of PV-IgG. Phosphorylation and activation of signaling cascades downstream of PV-IgG binding to target antigens in keratinocyte both in vitro and in vivo has highlighted the crucial role of signaling in PV pathogenesis (151–154). As a result, a great number of signaling molecules that are implicated in PV have been identified and characterized (22, 155). These include mitogen-activated kinases (MAPKs) such as p38MAPK, protein kinase C (PKC), extracellular signal-regulated kinases (ERK1/2), Rous sarcoma-related kinases (Src), phospholipase C (PLC), Epidermal growth factor receptor (EGFR), and other cellular responses that alter adhesive interactions (22, 48, 59, 110, 112, 131, 147, 151, 156–162). For several years, the plethora of signaling mechanisms appeared to be triggered without recognizable hirarchy upon binding of autoantibodies. However, it was shown that signaling molecules such as p38MAPK, PI4K, PLC, and PKC directly bind to desmogleins and that Dsg1 and Dsg3, together with Pg, organize overlapping yet distinct signaling hubs (136, 140, 163). These findings help to explain why autoantibodies against Dsg3 and Dsg1 were observed to cause different sets of signaling responses (48) and led us to propose that the different clinical phenotypes of pemphigus with respect to mucosal and skin involvement, as well as suprabasal versus superficial epidermal blistering, may at least in part be caused by the different signaling profiles observed in PV and PF (22).

p38MAPK has been thoroughly characterized due to its essential role in pemphigus pathophysiology. The different isoforms (α, β, γ, δ) display a species-specific expression pattern (164–166) among which the α subtype is the most commonly expressed isoform in adult tissues (165, 167). Activation of p38MAPK was detected in keratinocyte cell cultures treated with PV-IgG and in mouse skin (151). Moreover, p38MAPK was phosphorylated in perilesional skin of PV patients (153, 168) as well as in the skin of Dsg3-deficient mice (169). This signaling molecule has also been shown to be associated with Dsg3 (33, 112, 136, 140), Dsc3 (169), and Dsg1 (136). p38MAPK-mediated Dsg3 internalization followed by depletion from endosomes was also detected in keratinocyte cultures and patient skin (108, 112, 170). Interstingly, previous studies have shown that pharmacologic inhibition of p38MAPK was sufficient to avert cell dissociation in vitro (33, 48, 151, 171), rescue membrane-bound as well as cytoskeletal fractions of Dsg3 (112, 172, 173), and reorganize keratin cytoskeleton (49, 140, 151). Moreover, it sufficiently abolished blister formation after passive transfer of PV-IgG (152, 171) or PF-IgG (174, 175) in mice.

There are strata of protein kinase cascades functionally subordinate to p38MAPK (176). Mitogen-activated protein kinase 2 (MK2) regulates several cellular activities such as actin remodeling (177), an event which can be correlated to PV pathogenesis. Phosphorylation of MK2 has been detected upon p38MAPK activation by PV-IgG (173), the inhibition of which was protective both in vitro and in vivo. Rho A is crucial to maintain a strong keratin association with desmosomes, enhance cortical actin filaments, stabilize cytoskeletal bound Dsg1 and Dsg3, and was found to be inactivated following PV-IgG and PF-IgG in p38MAPK-dependent manner (131). Besides, toxin-mediated inhibition of Rho GTPases recapitulated the PV-IgG-induced suprabasal blistering in human skin (132).

We employed human organ culture models to assess the role of p38MAPK in mediating the ultrastructural changes of desmosomes in PV pathogenesis and found that inhibition of p38MAPK abolished blister formation in epidermis but not in mucosa, indicating that p38MAPK is crucial for the mechanisms causing epidermal blisters but not mucosal erosions (32, 135). Interestingly, loss of desmosomes as well as all ultrastructural alterations of desmosomes, including reduction in size, splitting, and keratin filament dissociation, were averted by inhibition of p38MAPK (Figures 3A–D).

These results suggest that modulation of this signaling pathway would be effective in treating pemphigus patients. However, a clinical trial using a p38MAPK inhibitor was terminated because of dose-limiting hepatotoxicity and did not show therapeutic benefits (178). This observation is important because it reveals that not all signaling pathways that are sufficient to stabilize desmosomal adhesion are druggable in patients.

The MAP kinase ERK, a downstream signal effector of EGFR, was phosphorylated upon activation by PV-IgG and specific inhibition of which prevented acantholysis in cell cultures (48, 158, 161). Interestingly, ERK was activated in the presence of Dsg1 autoantibodies only (48, 162) and its specific inhibition was protective against PV-IgG or PF-IgG in cultured keratinocytes (48). Dsg1 is required to suppress ERK1/2 signaling to promote keratinocyte differentiation in the suprabasal epidermis where it is strongly expressed (159) by interacting with ErbB2 binding protein (Erbin2) (179). Mek1 inhibition, an upstream target of ERK, was sufficient to avert epidermal blistering in human skin ex vivo (21, 134) and reduce in number of desmosomes primarily along the basal-suprabasal interface but did not prevent desmosome splitting and keratin detachment from desmosomes (Figures 3A–D) (134).

On the other hand, for PKC and Src, the human ex vivo skin model yielded results different from what was found in cultured keratinocytes in vitro or for in vivo mouse models. PKC has been shown to be involved in signaling, Dsg3 depletion, loss of cell adhesion, and blister formation in cultured keratinocytes as well as in mice (48, 110, 147, 156, 180). Nevertheless, PKC inhibition did not ameliorate skin blistering in the human skin model and failed to modulate ultrastructural alterations (181) (Figures 3A–D). This discrepancy may be caused by the fact that several PKC isoforms exist which might be involved differently in both desmosome assembly and disassembly (22, 64). Src regulates desmosome assembly via interaction of Dsg3 with E-cadherin (169, 182). Reduced phosphorylation of Src was detected along with decreased Dsg3 expression in basal keratinocytes surrounding blister cavities, linking this pathway to PV pathogenesis (182). Inhibition of Src prevented cell dissociation (48, 59, 183, 184) and abrogated manifestation of skin lesions in mice but not in human skin culture where it was also not sufficient to modulate ultrastructural alterations of desmosomes (Figures 3A–D) (59). Since Src, similar to PKC, appears both to participate in desmosome assembly, a process which may require cortactin (59, 162), and to cause loss of desmosome adhesion in response to PV-IgG, this may explain the discrepancy between in vivo data in mice and ex vivo studies in human skin. Alternatively, the role of Src in PV pathogenesis may depend on the autoantibody profile of patients because the PV-IgG fraction used in human skin included higher levels of antibodies targeting Dsg1. It was shown that Src-mediated EGFR activation is associated with anti-Dsg3 autoantibody-mediated signaling rather than with signaling caused by autoantibodies against Dsg1 or Dsc3 (162, 185). Therefore, it is possible that Src inhibition may be effective in treating some PV patients but not others.

However, for all studies using electron microscopy evaluation of human epidermis, the limitation is that they are feasible by using a very limited number of patients’ IgG fractions only. Thus, no preliminary dose-response characterization is possible, and it cannot be ruled out that, with autoantibody samples from other patients or using higher concentrations of pharmacological inhibitors, a protective effect may be found for other signaling molecules as well. Therefore, negative data must be taken with extreme care. On the other hand, if a protective effect under all conditions and in all models, including human skin, is observed for modulation of a specific signaling pathway, this is a strong indication for a potentially interesting treatment paradigm.

PV-IgG augments intracellular Ca²⁺ and inositol 1,4,5-trisphosphate (IP₃) levels (186) via activation of Phosphoinositid-Phospholipase C (PLC) (147) leading to cell dissociation. IP3, via activation of IP3 receptor (IP3R), causes Ca²⁺ release, which in turn activates PKC (187). Inhibition of PLC was protective against the dyscohesive effect of PV-IgG in cell culture (147), in vivo (157), and in human skin explant (133, 136). Interestingly, specific inhibition of PLC and Ca²⁺ signaling significantly ameliorated keratin dissociation and desmosome splitting (Figures 3A–D) (133).

It is well established that pemphigus is a desmosome turn-over disease (188) because the signaling pathways involved interfere with different steps of desmosome assembly and maturation on one hand, and desmosome internalization and disassembly on the other. The interesting question is whether the ultrastructural analyses as outlined above allow to allocate signaling pathways to specific steps of desmosome turnover. Loss of desmosomes, which was caused by mucocutaneous PV-IgG but not by mucosal dominant PV-IgG and AK23, is the ultimate consequence of a dysbalance between assembly and disassembly. Thus, depletion of extradesmosomal Dsg molecules, which serve as a pool for incorporation into existing desmosomes, may account for this phenomenon (109, 189). Similarly, it was shown that following depletion of Triton-soluble extradesmosomal Dsg3, Triton insoluble Dsg3 was reduced as well (109). This brings up the important question of whether the latter Dsg3 molecules were derived from desmosomes, which would be characterized best as desmosome disassembly, or whether this is just the consequence of impaired assembly. Experiments showing that Dsg3 internalization following autoantibody binding is a coordinated process involving endocytosis are compatible with both interpretations (111, 170, 171, 190–193).

Recently, it was reported in Madin–Darby canine kidney (MDCK) cells that hepatocyte growth factor (HGF) induces internalization of intact desmosomes with no alteration in desmosome size and composition, which was interpreted as desmosome internalization (194). As outlined above, the situation in pemphigus is different and more complex. PV-IgG cause significant shrinkage of remaining desmosomes (Figure 3B). In addition, transmission electron microscopy and structured illumination microscopy investigations have revealed that desmosomes in neonatal mice were split prior to internalization (124) and were present at cell surfaces surrounding blister cavities in patients’ lesions (129, 138, 195–198) and human ex vivo skin (Figure 3C) (181). These data show that desmosome splitting occurs when desmosomes are weakened due to an imbalance of desmosome assembly and disassembly which is facilitated by mechanical stress (198). In parallel to desmosomes with reduced size, double-membrane structures together with desmosomes which were uncoupled from keratin filaments were also observed (135), indicating that desmosome internalization in skin occurs similar to MDCK cells, which represent a single-layer epithelium. However, internalized desmosomes were rare and thus could not be evaluated in quantitative terms.

These results show that desmosome disassembly and internalization may be dependent on both cell type and stimulus. For both events, uncoupling of Dsg molecules from the cytoskeleton was shown (Figure 3D) (32, 133, 181, 194). It remains unclear to which extent mechanical destabilization of desmosomes resulting in desmosome splitting under force results from desmosome shrinkage or keratin filament dissociation. The data from ultrastructural evaluation of desmosomes indicate that inhibition of p38MAPK was the only pharmacologic intervention to rescue desmosome size (Figure 3B). On the other hand, both p38MAPK and PLC/Ca²⁺ were effective to modulate keratin dissociation and splitting of desmosomes, which can be interpreted that cytoskeletal anchorage is sufficient to prevent desmosome splitting. The data are in line with experiments showing that phosphorylation of Dp by PKC, which is activated downstream of PLC/Ca²⁺, in PKP1-dependent manner regulates intermediate filament anchorage, and thereby causes desmosome hyperadhesion by trapping of desmosome components (199–201). In response to PV-IgG, PKC-mediated Dp phosphorylation reverts the hyperadhesive state and causes loss of desmosome adhesion (180, 202). Another conclusion is that steric hindrance alone cannot explain these ultrastructural alterations observed in patient lesions and ex vivo skin model because this mechanism would cause split desmosomes with intact size and keratin filament anchorage.

Taken together, p38MAPK and PLC/Ca²⁺/PKC were shown to be involved in depletion of Dsg1 and Dsg3 (110, 112, 136), as well as in keratin filament dissociation (133, 135), indicating that these signaling pathways are important for disturbed desmosome assembly as well as for desmosome disassembly and internalization in pemphigus pathogenesis. For Src and ERK, it remains less clear which mechanisms are involved in loss of keratinocyte adhesion. It is likely that they contribute to the same mechanisms impairing desmosome turnover like p38MAPK and PLC-mediated signaling but may be less central for these events.

The ultrastructural analysis in human epidermis revealed that for all signaling pathways where pharmacologic modulation was protective, an ultrastructural correlate in desmosomes was found (Table 1). Based on this, we conclude that investigations on desmosome composition are helpful to advance our understanding of the regulation of desmosome turnover and to ultimately decipher a signaling pathway which may be druggable in patients as an additional line of therapy until depletion of pathogenic autoantibodies is effective and to reduce potential side effects.

Electron microscopy is a gold standard method to define ultrastructural alterations of desmosomes in pemphigus patient lesions (138) and to investigate the underlying mechanisms of a given signaling molecule in human skin, the role of which has been suggested by studies in cultured keratinocytes or mouse models. By this approach, mechanisms occurring in cell culture and mouse skin, but not in human epidermis, such as apoptotic cell death, also can be evaluated (138, 203). In the long run, super-resolution microscopy will be applied to study the pathogenesis of pemphigus in more detail as has been started already (170).