Desmosomes are intercellular junctions whose primary function is strong intercellular adhesion, known as hyperadhesion. In the present review, we discuss how their structure appears to support this function as well as how they are assembled and down-regulated. Desmosomal components also have signalling functions that are important in tissue development and remodelling. Their adhesive and signalling functions are both compromised in genetic and autoimmune diseases that affect the heart, skin and mucous membranes. We conclude that much work is required on structure–function relationships within desmosomes in vivo and on how they participate in signalling processes to enhance our knowledge of tissue homoeostasis and human disease.
- cell adhesion
- tissue development
- tissue remodelling
Desmosomes are intercellular junctions whose primary function is strong adhesion, providing intercellular links in the desmosome–IF (intermediate filament) complex and lending great tensile strength to tissues. They are most abundant in tissues that are subject to mechanical stress, such as epidermis and myocardium. Their importance in maintaining tissue integrity is underlined by their contribution to a number of genetic diseases that affect skin and heart, and by the autoimmune blistering disease pemphigus. Desmosomes perform this essential function because, unlike other adhesive junctions, they are able to adopt a strongly adhesive state known as hyperadhesion. They have a characteristic and highly organized structure that has been recognized since the early days of EM (electron microscopy). This structure appears to be fundamental to both their strong adhesion and their interaction with the cytoskeleton. Although the major molecular components have been known for almost 30 years, details of how the molecules are spatially arranged within and interact to form the structure are only just beginning to emerge. This role of cementing together the cellular building blocks of tissue architecture appears rather static, yet the architectural analogy is essentially a poor one as desmosomes are as dynamic as the tissues they support. The cell turnover that occurs in the course of normal tissue homoeostasis and the more dramatic remodelling that occurs during embryonic development and wound re-epithelialization necessitate desmosome assembly, adhesive modulation and down-regulation, processes into which there are insights, but where enormous gaps in our understanding remain. In addition to their adhesive role, desmosomes appear to participate in cell signalling during development, tissue morphogenesis and wound healing. The present review examines our current knowledge of the structural and functional dynamics of desmosomes, as well as their role in genetic and autoimmune disease. Other excellent recent reviews relating to these topics are [1–6].
Desmosomes are 0.2–0.5 μm in diameter in human epidermis and approx. 0.22 μm in MDCK (Madin–Darby canine kidney) cells . They consist of dense plaques arranged symmetrically on the cytoplasmic faces of the PM (plasma membrane) of adjoining cells (Figure 1). The membranes are separated by an ECD (extracellular domain) approx. 30 nm wide and bisected by a dense midline. Each plaque comprises an ODP (outer dense plaque), 15–20 nm thick, adjacent to the PM (approx. 7 nm thick), an 8 nm electron-lucent zone and an IDP (inner dense plaque), less dense than the ODP, but also 15–20 nm thick. Isolated desmosomes are irregular in shape (Figure 2).
The plaques show periodic organization perpendicular to the PM and lamellar organization parallel to it [8,9]. Cryo-EM of rapidly frozen human epidermis showed the ODP as an 10–11 nm region of medium electron density adjacent to the PM followed by two electron-dense layers with a combined thickness of 6.7 nm, showing a transverse periodicity of 6.6 nm [10,11]. The midline and plaque can also be visualized by AFM (atomic force microscopy) (Figure 2A) .
Desmosomes have five major component proteins, the DCs (desmosomal cadherins) DSG (desmoglein) and DSC (desmocollin), the plakin family cytolinker DP (desmoplakin), and the arm (armadillo) proteins PG (plakoglobin) and PKP (plakophilin). DSC and DSG are the desmosomal adhesion molecules, DP links the desmosomal plaque to the IF cytoskeleton, and PG and PKP are adaptor proteins that link between the adhesion molecules and DP (Figures 3 and 4).
DSC and DSG share 30% amino acid identity with each other and with classical cadherins  and have five ECs (extracellular cadherin repeats) containing Ca2+-binding sites and a CAR (cell-adhesion recognition) site [14,15]. Homology models for the DSC2 and DSG2 EC domains were generated using the crystal structure of Xenopus C-cadherin as a template [16,17]. These models imply that the DCs, like the classical cadherins, undergo trans-adhesive binding by strand exchange between their N-terminal domains. This is supported by inhibition of desmosomal adhesion with CAR-site peptides and mutation of a conserved alanine residue (Ala80) within the adhesive interface [14,15] (Z. Nie, A.J. Merritt and D.R. Garrod, unpublished work).
The cytoplasmic domains of DSGs possess a membrane-proximal region containing an intracellular cadherin-typical region (catenin-binding domain) and a DSG-specific region containing a proline-rich region, a series of unique 29-amino-acid repeats and a terminal domain. The DSG-specific region is disordered in solution, but shows weak specific interactions with PG, the plakin domain of DP, PKP1 and the cytoplasmic domain of DSC1 . Alternative splicing produces ‘a’ and ‘b’ forms of DSCs, the ‘b’ form being shorter and generated by insertion of a mini-exon containing a stop codon. The ‘a’ form binds PG through its catenin-binding domain, which the truncated cytoplasmic domain of the ‘b’ form lacks .
There are four DSG isoforms in humans (six in mice) and three isoforms of DSC. DSG2 and DSC2 are ubiquitous, expressed in all desmosome-bearing tissues and the predominant isoforms in simple epithelia [20,21]. In stratified epithelia, the distribution of DSC2 mirrors that of DSC3 (see below), whereas expression of DSG2 is confined to the basal cells. DSG1, DSG3 and DSG4 and DSC1 and DSC3 are restricted to stratified epithelia [20,22–24]. Reflecting tissue differentiation, DSG1 and DSC1 are strongly expressed in the granular and spinous layers of the epidermis, and more weakly in the lower layers, whereas DSG3 and DSC3 are expressed strongly in the basal and immediate suprabasal layers, decreasing towards the stratum granulosum [3,13,25]. When co-expressed, different isoforms occur within the same desmosomes .
PG has 12 arm repeats each consisting of three α-helices (two in repeats 1 and 7) flanked by unstructured N- and C-termini . These arm repeats share 65% amino acid identity with β-catenin that associates with AJs (adherens junctions). PG can substitute for β-catenin in AJs because both bind E-cadherin with similar affinity, but PG has higher affinity for DSG suggesting why it, rather than β-catenin, locates to desmosomes . The arm repeats of β-catenin form a superhelical positively charged groove, the cadherin-binding site, and PG binds to E-cadherin in a similar manner .
PKP contains nine arm repeats flanked by an N-terminal head and a small C-terminal region [27,28]. The arm repeat region of PKP1 is sickle-shaped because of an insert between repeats 5 and 6, whereas PG and β-catenin are straight . The N-terminal head is responsible for all of the known binding interactions of PKP, including the DCs, DP and PG. PKP also interacts with keratin and actin [30,31]. There are three main isoforms, PKPs 1 and 2 having ‘a’ and ‘b’ splice variants. In both cases, the shorter ‘a’ variant predominates [32,33]. p0071 (PKP4) is a member of this group, but its association with desmosomes is debatable [34,35].
PKP1 is restricted to the differentiated layers of stratified epithelia [27,28]. PKP2 is found in simple, complex and stratified epithelia, and in cardiac myocytes, where it is the only isoform . PKP3 is found in most epithelia except for hepatocytes. PKP1 and PKP2 exhibit dual localization in the nucleus, as well as at desmosomes [33,36]. The alternative splicing of the ‘b’ form, 21 amino acids longer in PKP1b and 44 amino acids longer in PKP2b may target these proteins to the nucleus.
DP has two splice variants, I and II, both consisting of a coiled-coil α-helical rod domain, shorter in DPII, flanked by globular N- and C-terminal domains. DP is predicted to form homodimers via the coiled-coil region . The C-terminal domain contains three regions, each consisting of 4.5 copies of a 38-amino-acid plakin repeat. These C-terminal regions form discrete subdomains that bind IFs .
A systems biology approach has characterized the ‘desmo-adhesome’ . This includes 59 proteins of which 30 are intrinsic and 29 are accessory. Apart from those above, the intrinsic proteins include the plakins, envoplakin and periplakin, the epidermal late differentiation antigens, involucrin and loricrin, and several keratin isoforms. Among the accessory proteins are 14 protein kinases and four protein phosphatases, which may have a role in regulating desmosomal adhesion.
MAPPING THE MOLECULES ON TO THE STRUCTURE
The challenge to elucidating desmosomal plaque structure is its electron density, but immunogold EM has produced a molecular map of the plaque (Figure 5A) . This showed that the arm proteins interact with the N-terminus of DP and the cytoplasmic domains of the DCs in the ODP, whereas the C-terminus of DP binds to IFs in the IDP in accordance with functional data [40,41]. Mapping the localization data on to the cryo-EM structure suggests that the densities in the plaque correspond to this same region of substantial protein–protein interaction [9–11] (Figure 5B).
The structure of the ECD is clearer because it is less electron-dense. Lanthanum infiltration of guinea pig heart showed that the ECD contained a regular staggered array of side arms with a periodicity of 75 Å (1 Å=0.1 nm) linking the midline to the PM . In en face view, the side arms were arranged in regular rows, a quadratic array. The homology models of the DCs  were used to generate a three-dimensional array that appeared to account for the midline and showed remarkable agreement with the periodicity noted by Rayns et al. . Cryo-EM of human epidermis supported such an arrangement [10,11] and ET (electron tomography) indicated a regular array of 3 nm diameter densities at 7 nm intervals along the midline with a curved shape resembling the crystal structure of C-cadherin . Computational fitting of the C-cadherin crystal structure on to tomograms indicated that molecules interact at the midline, forming building blocks of alternate V-shaped cis dimers and W-shaped trans dimers, resulting in a highly packed regular organization. Immunogold labelling of the ECD with a specific antibody showed that the N-terminus of DSG3 is localized to the midline, consistent with current views of cadherin trans interaction and molecular models of the ECD .
THE SPECIFICITY OF DESMOSOMAL ADHESION
Desmosomal adhesion specificity has two aspects. First, can desmosomes form non-specifically between cells of different types and even from different species? The answer is definitely yes, in both respects [45,46]. Secondly, at the molecular level, does DSG bind to DSC and do different isoforms of DSG and DSC bind to each other? Studies with transfected cells or recombinant proteins gave equivocal answers suggesting either heterophilic or both heterophilic and homophilic interaction [47–49]. Amagai et al.  found that DSG3 can mediate weak homophilic adhesion. Single-molecule AFM using the tethered EC domain of DSG1 has revealed homophilic multivalent Ca2+-dependent low-affinity interactions . AFM has also shown a homophilic DSC3-binding and heterophilic interaction between DSC3 and DSG1, but not between DSC3 and DSG3 . Work with desmosome-forming cells seems to suggest homophilic adhesive binding between DCs. Thus anti-adhesion peptides of both DSG and DSC were required to block desmosome assembly by mammary epithelial cells and de novo desmosome formation has been demonstrated in non-desmosome-bearing cells expressing DSG only [15,53]. Cross-linking experiments with HaCaT cells expressing DSC2 and 3 and DSG2 and 3 suggested that adhesive binding is both homophilic and isoform-specific (Z. Nie, A. J. Merritt and D. R. Garrod, unpublished work)
HYPERADHESION: ITS SIGNIFICANCE AND REGULATION
A property of desmosomes that distinguishes them from other intercellular junctions is their ability to adopt a strongly adhesive state known as hyperadhesion [17,54]. This is characterized by Ca2+-independence and resistance to disruption by chelation of EC Ca2+ [55–57]. It is the normal state of desmosomes in mammalian tissues and appears to be of key functional significance, permitting desmosomes to perform their primary role of maintaining tissue integrity [4,17,57].
Adoption of hyperadhesion is a remarkable property of the DCs. During the early stages of culture, desmosomal adhesion is Ca2+-dependent [54,56,57], chelating agents inducing loss of adhesion by splitting desmosomes into halves. This is what would be expected from classical cadherin-mediated adhesions. Maintained in confluent culture, epithelial desmosomes acquire hyperadhesion without changing their molecular composition with respect to their major components [54,57]. How can the change from Ca2+-dependence to Ca2+-independence occur? A possible answer comes from studies of epidermal wound healing, as follows.
Hyperadhesive mouse epidermal desmosomes have a prominent midline . Wounding the epidermis causes desmosomes near the wound edge to revert to Ca2+-dependence. They then lose the midline and the intercellular space narrows by 10% . This suggests that hyperadhesion is associated with the ordered arrangement of the DCs, but that reversion to Ca2+-dependence involves the loss of this order. Order is associated with strong adhesion and disorder with weak adhesion.
We speculate that the DCs become locked into an ordered arrangement possibly involving entrapment of Ca2+ ions [17,54]. Furthermore, the desmosomal plaque may be crucial in regulating the adhesive state by transducing signals that order or disorder the DCs. In this regard, desmosomal adhesiveness may be regulated by PKCα (protein kinase Cα), activation promoting Ca2+-dependence and inhibition promoting hyperadhesion [17,57]. Phosphorylation of plaque components may cause rearrangement within the plaque and transmit a signal to the EC domains. However, desmosomal adhesiveness may be capable of being regulated by different cytoplasmic signals because inhibition of tyrosine phosphatases (i.e. promotion of tyrosine phosphorylation) also promotes hyperadhesion .
The first essentials for desmosome assembly are cell–cell contact and specific adhesive interaction. Prevention of these by maintaining cells at low EC Ca2+ concentration, or by blocking with antibodies or anti-adhesion peptides, inhibits desmosome assembly [15,55,59]. Intercellular adhesion is normally initiated by AJs, then stabilized by desmosomes . PG, which binds to both E-cadherin and DCs, may be crucial for interaction between AJs and desmosomes during assembly . PKP2 and p120 catenin also associate with both desmosomal and classical cadherins and may be involved in sorting AJ and desmosomal components during junction assembly . Nectin-1, which has been shown to participate in AJ assembly, has now also been shown to be involved in desmosome assembly in the enamel-forming epithelium of mouse incisors .
The DSC ‘a’ domain supports desmosomal plaque assembly, recruiting both PG and DP to the PM . The ‘b’ form, however, cannot, leaving the function of the ‘b’ form unresolved. The ‘b’ form has been shown to bind PKP3 , but zebrafish lacks a ‘b’ form, yet can assemble apparently normal desmosomes (X.-M. Luan, A. Hurlstone and D. R. Garrod, unpublished work). DSG also binds PG, and this binding is crucial for its incorporation into desmosomes . DSG, like DSC ‘a’, can support desmosomal plaque assembly because desmosomes assemble in cells lacking DSC . All mammalian desmosomes appear to contain at least one isoform of DSG and DSC ‘a’ and ‘b’.
PG and PKP are crucial for normal plaque assembly. Deletion or loss-of-function of either gives rise to diminished plaques with loss of DP and IF attachment [66–68] (Tables 1 and 2), demonstrating that these proteins link between the DCs and IFs via DP. In the fatally defective hearts of mid-gestation PG−/− mice, desmosomes were absent, but extended AJs contained desmosomal components  (Table 1). The C-terminus of PG plays a role in regulating desmosome size; its deletion results in the formation of extremely long junctions .
The arm repeat domain of PKP1 contains a basic patch that could be responsible for ligand binding . Nevertheless, expression studies suggest that the unstructured N-terminus recruits DP to the desmosome, whereas the C-terminus recruits PKP itself to the PM [30,71,72]. A role for PKP in mediating lateral interactions between desmosomal components has been postulated .
DP mediates attachment between the plaque and the IFs, its N-terminus being associated with the ODP and its C-terminus with IF binding at the IDP [9,40,41,73]. During desmosome assembly, DP can be recruited to contact zones from both diffusible and particulate cytoplasmic pools by an actin-dependent mechanism and involving association with PKP2 .
DESMOSOME STABILITY AND TURNOVER
Maintenance of tissue integrity demands that desmosomes should be both strongly adhesive and stable. Their remarkable stability in cultured cells was demonstrated by monitoring the behaviour of fluorescently labelled DSC2a incorporated into desmosomes . Desmosomes were relatively immobile and maintained their structural integrity for long periods throughout the cell cycle. Only minor destabilization was evident during mitosis, confirming earlier work . However, significant turnover of DSC2a was possible as FRAP (fluorescence recovery after photobleaching) showed a half-life of approx. 30 min .
However, there is a conflict between maintaining stability and tissue integrity and the continual need for tissue remodelling (e.g. embryonic development, wound re-epithelialization and cell renewal in the epidermis and the intestinal mucosa). Cells at different levels within the epidermis have desmosomes with differing composition of DCs . Given the FRAP results of Windoffer et al. , this could occur by molecular exchange within existing desmosomes or may require completed desmosomal turnover.
As far as we are aware, the only documented mechanism for desmosome down-regulation in tissues is the internalization by cells of entire desmosomes. One cell of a pair internalizes the whole desmosome, including the plaque and a small amount of immediately adjacent cytoplasm of its neighbour. This occurs in several situations, including wound healing and neoplasia . This may involve a type of phagocytosis where one cell of a pair produces cytoplasmic processes that engulf the desmosome and internalize it into a vacuole . Consistent with this is the observation that internalization of desmosomal halves formed by loss of desmosomal adhesion by cultured cells following Ca2+ depletion, a process which also involves engulfment in vacuoles, requires actin .
The fate of such desmosomes or desmosomal halves once internalized is far from clear. They are presumably degraded, although by what route is uncertain . Although the evidence is not entirely conclusive it seems unlikely that they disassemble and their components are reutilized [75,80].
Because of its importance in tissue remodelling, it seems that further work on the mechanism of desmosome down-regulation in vivo is urgently required.
PHOSPHORYLATION AND DESMOSOME REGULATION
There are several ways in which phosphorylation is believed to regulate aspects of desmosome function. PKCα is a conventional PKC isoenzyme and serine/threonine kinase. It localizes to the ODP of wound-edge epidermal desmosomes, most of which are Ca2+-dependent, but is absent from hyperadhesive desmosomes . The PKC target proteins involved in this regulation are currently unknown. The ProteinPredict program of ExPASy indicates many potential PKC-target motifs, as follows: four DSC2a cytoplasmic domain; 13 DSG2 cytoplasmic domain; 13 PG; 21 PKP1; 64 DPI. All of these proteins (not necessarily the same isoforms) have been shown to be serine-phosphorylated under certain circumstances [41,81–85]. To provide evidence of how phosphorylation regulates hyperadhesion, we mutated conserved sites in the cytoplasmic domains of DSG2 and DSC2a and expressed the mutants in MDCK cells (see  for more details). This produced a modest increase in hyperadhesion, which was enhanced by depletion of the endogenous protein. It seems likely that the regulation is complex, possibly involving multiple targets.
Activation of PKC has been reported to enhance both desmosome assembly and disassembly [87,88]. Recently, formation of a complex between PKP2, DP and PKCα was demonstrated during desmosome assembly in cultured cells . This may involve phosphorylation by PKCα of Ser2849 of DP, a site that had been shown previously, when phosphorylated, to negatively regulate the DP–IF interaction . It is not clear whether this is relevant in vivo as PKCα−/− mice, which are viable, assemble normal desmosomes with attached IFs (T. E. Kimura and D. R. Garrod, unpublished work).
Phosphorylation of PG has been reported to affect its distribution between the soluble and insoluble pools, that in the soluble pool being both serine- and threonine-phosphorylated, whereas insoluble PG was primarily phosphorylated on serine .
AJ adhesion is generally down-regulated by tyrosine phosphorylation by Src family kinases . The situation with desmosomes is less clear. Tyrosine phosphorylation can cause a modest destabilization of desmosomes. Treatment of A431 cells with EGF (epidermal growth factor) induces tyrosine phosphorylation of PG and DSG2, together with a modest decrease in cytoskeletal association of PG [91,92]. Furthermore, tyrosine-phosphorylated PG does not associate with DP. These effects are inhibited by treatment with EGFR (EGF receptor) inhibitor . In contrast, treatment of cells with the tyrosine phosphatase inhibitor sodium pervanadate caused tyrosine phosphorylation of the major desmosomal components DSG2 and PG in both the soluble and the insoluble cell fractions, without affecting their complex formation and, surprisingly, induced hyperadhesion . Collectively, these observations suggest that the role of tyrosine phosphorylation in regulating desmosomal adhesion is complex. Indeed tyrosine phosphorylation of PG at different sites by different kinases causes differential effects . Thus Src phosphorylates Tyr643, decreasing PG binding to E-cadherin and α-catenin, whereas Fer phosphorylates Tyr549, increasing PG binding to α-catenin. Furthermore, PG suppresses Src activity and cell motility .
Phosphorylation of desmosomal components may be important in skin disease. Treatment of squamous carcinoma cells with PV (pemphigus vulgaris)-IgG induced serine phosphorylation of DSG3 and its dissociation from PG . Furthermore a rapid sustained increase in PKC activity occurred upon PV-IgG binding to DSG3, suggesting that PV-IgG is able to induce PKC activation . However, PMA activation of PKC did not cause phosphorylation of DSG3 and the use of PKC inhibitors caused only a partial inhibition of DSG3 phosphorylation at best, suggesting that another kinase may be responsible for PV-IgG-induced phosphorylation of DSG3 .
DISEASES AND MOUSE MODELS: THE FUNCTION OF DESMOSOMES AND THEIR COMPONENTS
As the primary function of desmosomes is that of strong adhesion, it is not surprising that mutations in genes encoding desmosomal proteins are responsible for diseases in which cell adhesion is compromised . The identification of these mutations has highlighted the importance of desmosomes in ectodermal integrity, hair follicle development and cardiac function. Also, striking similarities manifest between abnormalities in patients and the phenotypes of mutant mice (Tables 1 and 2).
The first disorder found to result from a desmosomal protein gene mutation was ED-SFS (ectodermal dysplasia-skin fragility syndrome) (OMIM 604536), characterized by skin fragility, inflammation and abnormalities in ectodermal development, including scant hair, reduced sweating and astigmatism . Ten cases of ED-SFS have now been reported, all resulting from nonsense, frameshift or splice site mutations in PKP1 [99,100].
ED-SFS patients' epidermis shows fewer, poorly formed, desmosomes . Analysis of PKP1 mutations in cultured keratinocytes indicates a role for PKP1 in desmosomal plaque formation and stability . Also, PKP1 may have a nuclear signalling role, but whether defective signalling results in ED-SFS remains unclear [33,72].
Ablation of Pkp2 in mice resulted in embryonic lethal alterations in heart morphogenesis, caused by reduced association of DP and PG with junctions in cardiomyocytes . Mutant PKP2 is unable to locate to the PM and, consequently, to recruit DP to the desmosomal plaque and IFs to cell–cell contacts . Subsequently, nonsense, frameshift, splice site and missense mutations in PKP2 have been identified as the major cause of ARVC (arrhythmogenic right ventricular cardiomyopathy), a disease which has a prevalence of 1/1000–1/5000 and is characterized by ventricular arrhythmias, syncope and sudden death [103,104]. Some 10% of deaths occur before the age of 19, and 50% before the age of 35, making ARVC a leading cause of cardiac sudden death in people under 35 years of age.
Pkp3-deficient mice develop hair abnormalities and increased susceptibility to cutaneous inflammation  despite compensatory up-regulation of PKP1, PKP2 and DP. Whereas Pkp3−/− mice raised in SPF (specific pathogen-free) facilities showed limited skin alterations, Pkp3−/− mice raised in conventional facilities developed heightened inflammation, with large areas of epidermal hyperplasia, spongiosis, neutrophil-filled pustules, severe dermal and epidermal infiltration by immune cells, hair loss, scaling and wasting . Culture experiments showed that PKP3 associates with ribonucleoprotein particles and under environmental stress incorporates into ‘stress granules’ to stall translation initiation complexes . Thus Pkp3−/− mice may serve as a model for particular forms of dermatitis. However, no human PKP3 mutations have yet been reported.
In 1999, Armstrong et al.  mapped a disease-causing mutation of an autosomal dominant pedigree with SPPK (striate palmoplantar keratoderma) (OMIM 125647) to 6p24.3, where the DSP gene resides. Sequence analysis revealed heterozygous nonsense and splice site mutations in DSP leading to haploinsufficiency . These findings indicate that a 50% loss of DP is sufficient for normal development and function of non-palmoplantar skin, but insufficient to maintain normal skin on the palms and soles.
The first reported autosomal recessive mutation in DSP, a deletion at the 3′ end of the gene, resulted in a DP protein lacking its C-terminal domain . This gave rise to Carvajal syndrome (OMIM 605676), a rare disorder comprising woolly hair, keratoderma and left ventricular cardiomyopathy [108,109]. Subsequently, compound heterozygous nonsense–missense mutations were found in patients exhibiting woolly hair and keratoderma, but no cardiomyopathy, termed skin fragility woolly-hair syndrome (OMIM 607655) . A compound heterozygous nonsense frameshift mutation that truncates DSP has been identified in a single case of lethal acantholysis epidermolysis bullosa (OMIM 609638) . In addition, homozygous missense mutation in DSP have been identified in patients with isolated cardiomyopathy, whereas homozygous missense mutations and compound heterozygous nonsense mutations have been identified in patients with a combination of woolly hair, keratoderma and ARVC [111–113]. Thus the genotype–phenotype correlation of DP mutations is complex and poorly understood.
The lethality of Dsp−/− mice at E (embryonic day) 6.5 precludes their use in determining the functional importance of DP in heart and skin development . Partial rescue of Dsp−/− embryos up to E10 using the tetraploid aggregation approach revealed major defects in heart muscle, neuroepithelium and skin epithelium, as well as in the microvasculature . Conditional ablation of Dsp using K14-Cre highlighted the role of DP in desmosomal adhesion in the epidermis. At birth, the epidermis of these mutant mice peels, leaving areas of denuded skin . Desmosomes in affected regions lacked an IDP and failed to attach to the keratin cytoskeleton, indicating the importance of DP in maintaining cytoskeletal architecture to reinforce stable intercellular adhesion.
As with DP, PG also plays an important role in heart, skin and hair development. Pg−/− mice display severe cardiac defects, causing death at E10.5. The few mutant pups that survive to birth (on a C57BL/6 background) display skin blistering and subcorneal acantholysis . In addition, Coonar et al.  mapped Naxos disease (OMIM 601214), an autosomal recessive disorder characterized by ARVC, woolly hair and keratoderma in a family originating from the Greek island of Naxos, to 17q21, a locus containing the JUP, the gene encoding PG. Subsequently mutations in JUP were described; a homozygous 2 bp deletion was found in 19 patients diagnosed with Naxos disease . The prevalence of Naxos disease on the Greek islands may be as high as 1/1000, but patients with Naxos disease have also been identified in Turkey, Israel and Saudi Arabia.
A number of pathogenic splice site and frameshift DSG1 mutations, believed to cause haploinsufficiency, cause SPPK (OMIM 148700) [120–123]. The patients' skin exhibits fewer, smaller, desmosomes and loss of IF attachment.
In contrast with DSG1 mutations, patients harbouring mutations in DSG2 exhibit cardiac defects, but no skin abnormalities. DSG2 mutations may account for 5–10% of ARVC and include dominantly inherited frameshift, splice site, nonsense and missense mutations, mostly in the EC domain [124,125]. Transgenic mice overexpressing the ARVC mutation DSG2 N266S recapitulate the ARVC phenotype, including spontaneous ventricular arrhythmias, cardiac dysfunction, biventricular dilatation, aneurysms and sudden premature death.
Global knockout of Dsg2 in mice is embryonic lethal shortly after implantation, indicating a non-desmosomal role for DSG2 in embryonic stem cell proliferation and embryonic survival . In contrast, overexpression of Dsg2 under control of the involucrin promoter results in epidermal hyperplasia and hyperkeratosis, which developed into intra-epidermal skin lesions .
Although no human DSG3 mutations have been identified, autoantibodies against DSG3 are the cause of PV (discussed below). Dsg3−/− mice and the PV mouse model present similar suprabasal acantholysis [128,129]. Dsg3−/− mice exhibit reduced PG levels in the desmosomal plaque. The acantholysis observed in Dsg3−/− mice is therefore believed to arise through fragility of desmosomes lacking DSG3 . In contrast, the levels of desmosomal components of the PV mouse model are unaltered, indicating a different mechanism of acantholysis possibly involving altered DP localization and thus IF retraction . In addition to its role in epidermal integrity, transgenic mouse models of Dgs3 have identified clear roles for this protein in hair-shaft anchorage and epithelial differentiation [131–133].
DSG4 mutations cause the rare disorders autosomal recessive hypotrichosis (OMIM 607903) and recessive monilethrix. Although listed as distinct entities, both result from mutations in the DSG4 EC domain and there is considerable phenotypic overlap. A homozygous intragenic deletion in a patient with hypotrichosis, a disease characterized by abnormal hair growth, has since been found in several Pakistani families [23,134]. In addition, a homozygous missense mutation has been identified in a conserved motif known to be important for cell adhesion . Following on from this, a variety of mutations in DSG4, including frameshift, splice site, missense and nonsense mutations were identified. In these cases, the patients were reported to suffer from autosomal recessive monilethrix, a disease caused by structural defects of the hair shaft, usually inherited in an autosomal dominant fashion by mutations in the keratin genes hHb1, hHb3 and hHb6 [136–138].
Mutations in mouse and rat Dsg4 result in the lanceolate hair phenotype, characterized by sparse, fragile, broken hair shafts, ichthyosiform dermatitis and follicular dystrophy [139,140]. Furthermore, a mutation in Dsg4 causes hypotrichosis in EODhage mice, an inbred strain characterized by the complete absence of hair on the trunk and mast cell hyperplasia . These observations highlight an important role for DSG4 and desmosomes in the regulation of hair growth.
Dsc1 knockout in mice causes defects of the skin which become apparent 2 days after birth . Affected skin is flaky with punctate barrier defects that develop acantholysis in the granular layer leading to localized lesions and epidermal fragility. By 6 weeks of age, these lesions resemble chronic dermatitis and localized hair loss is observed. Unlike the global loss of Dsc1, mutant mice carrying a C-terminally truncated form of Dsc1a/b do not display an abnormal phenotype despite the fact that the mutant protein integrates into desmosomes, suggesting that the C-terminus is not required for desmosomal stability. No human DSC1 mutations have been described.
DSC2 is the only DSC isoform expressed in cardiac tissue, and heterozygous mutations in DSC2 have been identified in patients with isolated ARVC. Thus ARVC can result from mutations in at least five desmosomal protein components, JUP, DP, PKP2, DSG2 and DSC2 [143,144]. The N-terminal DSC2 mutant protein expressed in cultured cells resides predominantly in the cytoplasm and is unable to incorporate into desmosomes. However, it is yet to be determined whether these mutations result in loss-of-function null alleles through cytoplasmic degradation or in a dominant gain-of-function manner . More recently it was shown that the triad of ARVC associated with SPPK and woolly hair can also result from frameshift mutations in DSC2 .
Dsc3−/− mice, which are embryonic lethal, die before the formation of mature desmosomes and therefore fail to provide an insight into the importance of DSC3 in desmosomal adhesion . To overcome this problem, Chen et al.  utilized K14-Cre to ablate Dsc3 in the epidermis and found that mutant pups develop severe epithelial blisters induced by mild mechanical stress and cyclic hair loss in adulthood. Acantholysis was observed between basal–basal and basal–suprabasal keratinocytes, with desmosomes splitting in the midline. No differences were observed in the levels and localization of other desmosomal protein components, suggesting that DSC3 in the epidermis is required for desmosome stability . A homozygous nonsense mutation was identified in a family affected with hereditary hypotrichosis. Affected individuals exhibited sparse and fragile hair on the scalp, absent eyebrows and eyelashes and large skin vesicles filled with watery fluid . Thus DSC3 plays a role in epidermal integrity and hair development.
PF (pemphigus foliaceus) and PV
Autoantibodies against DSG1 appear to cause PF and fogo selvagem, and those against DSG3 or DSG3 and DSG1 cause PV [149–151]. Patients with PF and fogo selvagem have epidermal blisters within the granular layer or beneath the stratum corneum, whereas PV blisters occur just above the basal layer. These defects correlate with the expression profile of the cadherins, with increased DSG1 in the upper epidermis and DSG3 highest in the basal layer .
How the autoantibodies cause acantholysis remains controversial. One argument is that autoantibodies bind to the EC domains of the DCs impeding their adhesive function [51,153]. Heupel et al.  demonstrated that PV-IgG directly inhibits DSG3 trans-interactions. Upon binding of PV-IgG, both DSG3 and PG were rapidly internalized in a complex which does not contain DP and retraction of the IFs from the cell borders occurred . Caldelari et al.  demonstrated that PG is essential for PV-IgG to elicit this effect. Although both PG−/− and WT (wild-type) keratinocytes could bind the DSG IgG, only WT keratinocytes lost desmosomal adhesion and IF attachment.
In contrast, PF autoantibodies against DSG1 alter intracellular signalling before reducing cell-surface levels of DSG1 . Waschke et al.  demonstrated that p38 MAPK (mitogen-activated protein kinase) phosphorylation facilitates the retraction of IFs and loss of adhesion observed following treatment with DSG3 autoantibodies. Investigations of PV murine models have shown that inhibitors of tyrosine kinases, phospholipase C, calmodulin and PKC all impaired the development of acantholysis in mice, indicating that a broad range of signalling cascades are involved . Further studies have demonstrated that PG-mediated inhibition of c-Myc transcription is prevented by PV-IgG binding, resulting in increased intracellular c-Myc concentrations .
Alternative views are that the primary cause of acantholysis may be induction of apoptosis or loss of interdesmosomal adhesion. Culture studies have shown that exposure to pemphigus autoantibodies induces apoptosis resulting in acantholysis [160–162]. In vivo studies of PF found apoptotic cells in the epidermis, increased expression of pro-apoptotic factors and reduced anti-apoptotic factors. Activated caspase 3 was detected in the epidermis before blister development and a peptide-based caspase inhibitor blocks blister development .
A review of the literature on the EM of PV shows that direct disruption of desmosomal adhesion is not the primary event . Rather, there is loss of cell–cell adhesion in interdesmosomal regions and possible intracellular cleavage behind the desmosomal plaque that might indicate a weakening of the cytoskeleton, perhaps through a signalling mechanism involving PG [68,164].
Treatment of keratinocytes with PV autoantibodies in culture showed that desmosomes are down-regulated and that DSG3 is depleted and internalized [124,165,166]. However, this work was presumably carried out with keratinocytes possessing Ca2+-dependent desmosomes, which do not represent the in vivo situation. A study in which keratinocytes with Ca2+-dependent or hyperadhesive desmosomes were compared showed that hyperadhesion inhibited PV autoantibody-induced acantholysis and internalization of adhesion molecules, including DSG3 and E-cadherin . These observations are consistent with the view that down-regulation of desmosomes is not the primary event.
It has been shown that PV autoantibodies activate Src . We hypothesize that autoantibodies may activate Src causing loss of inter-desmosomal adhesion in the epidermis by disrupting AJs, and that this may cause secondary loss of desmosomal adhesion and acantholysis.
OTHER ROLES FOR DESMOSOMES: DEVELOPMENT, DIFFERENTIATION, PROLIFERATION AND APOPTOSIS
In 1996, having demonstrated the remarkable reciprocal distributions of DSC1 and DSC3 in epidermis, we suggested that they may be responsible for regulating cell positioning and/or generating positional information that regulates cell differentiation, the latter possibly involving PG or β-catenin . In the same year, targeted expression of ΔN-DSG3 was shown to cause changes in the epidermis, including hyperproliferation . Subsequently, there have been several indications that desmosomes contribute to cellular development and behaviour through signalling processes.
Desmosomes are essential for embryonic development. Deletion of the genes for PG, PKP2 or DP in mice results in embryonic death principally because of failure of intercellular adhesion (see Table 1 and above). On the other hand, the early embryonic lethal effects caused by deletion of Dgs2 or Dsc3 are more probably due to signalling defects. Deletion of Dsg2 causes death at around the time of implantation because of a defect in embryonic stem cell proliferation, whereas deletion of Dsc3 causes death before E2.5, which precedes the time of first desmosome assembly in the trophectoderm of the blastocyst [126,146]. Of further developmental significance may be the demonstration that desmosomal adhesion regulates cell positioning during mammary gland morphogenesis .
Several mouse desmosomal mutations give rise to cell proliferation changes. Thus failure of egg cylinder elongation in Dsp−/− mice was associated with hypoproliferation, whereas epidermal hyperproliferation was noted in Dsc1−/−, K1-Dsg3, K1-Dsc3 and Inv-Dsg2 mice [114,127,133,142,170] (Table 1). Whether such effects are primary, a direct consequence of the desmosomal changes, or secondary, a response to the effect on tissue caused by adhesive changes, is not certain, but culture assays are consistent with the idea that primary Dsc1−/− keratinocytes are hyperprolifative (M.A.J. Chidgey, personal communication). Such mutations are also associated with differentiative changes, including hyperkeratosis, abnormality of hair follicles and altered keratin expression [114,127,133,142,170] (Table 1). Inv-Dsg3 mice have an abnormal stratum corneum and lethal epidermal barrier defects, whereas Inv-Dsg2 keratinocytes show resistance to apoptosis [127,132]. The latter mice also showed increased susceptibility to chemically induced carcinogenesis .
The above epidermal changes have been shown to be associated with activation of various signalling pathways. Hardman et al.  found increased β-catenin signalling in K1-Dsc3 mice and similar changes were subsequently found in Dsc1−/− and K1-Dsg3 mice (A.J. Merritt, M.J. Hardman and D.R. Garrod, unpublished work). Brennan et al.  found enhanced activation of multiple growth and survival pathways in Inv-Dsg2 mice, including PI3K (phosphoinositide 3-kinase)/Akt, MEK [MAPK/ERK (extracellular-signal-regulated kinase) kinase]/MAPK, STAT3 (signal transducer and activator of transcription 3) and NF-κB (nuclear factor κB). In a culture model, DSG1 promotes keratinocyte differentiation in an adhesion-independent manner and through suppression of EGFR–ERK1/2 signalling . It will be interesting to learn whether this mechanism is relevant in vivo. DP has been shown to play a role in differentiation-related reorganization of microtubules in epidermis through recruitment of the centriolar protein ninein 
Inv-Dsg2 keratinocytes showed enhanced survival dependent on EGFR activation and NF-κB activity . This may represent an in vivo demonstration of previous results implicating DSGs in the regulation of apoptosis. Both DSG1 and DSG2 are caspase 3 targets and play apoptosis-promoting roles [173,174]. PG has also been implicated in the regulation of apoptosis, but it is unclear whether by promotion or inhibition. Dusek et al.  showed that deletion of PG protected keratinocytes from apoptosis by suppressing Bcl-XL expression, whereas Li et al.  showed that overexpression of PG inhibited apoptosis by inducing Bcl-2 expression.
PG has important and diverse roles in transcriptional regulation in addition to its above role in apoptosis. The role of its close relative, β-catenin, is well known [177,178]. Activation of the Wnt signalling pathway or mutation of members of the β-catenin degradation complex, e.g. in adenomatous polyposis coli, cause elevation of cytoplasmic levels of β-catenin, which then enters the nucleus and, in complex with Tcf (T-cell factor)/Lef-1 (lymphoid enhancer factor 1) transcription factors, regulates gene expression, for example up-regulation of the proto-oncogene c-Myc . PG can play a similar role. It is a strong activator of the c-Myc promoter, indicating a possible role in cancer . However, in other situations, PG represses c-Myc transcription [159,181]. This is important in relation to the initiation of keratinocyte terminal differentiation, with suppression of c-Myc promoting cell cycle exit . Transcriptional repression by PG is mediated through binding to Lef-1 and involves a Tcf/Lef-1-binding site in the c-Myc promoter distinct from those involved in c-Myc activation, which is mediated by PG binding to Tcf-4 [159,181]. Which function PG performs may depend upon the relative levels of Tcf/Lef-1 transcription factors .
PKP has both structural and signalling roles. In addition to their desmosomal localization, PKPs 1 and 2 localize to the nucleus. The nuclear localization of PKP2 is regulated by Cdc25C-associated kinase 1, which phosphorylates it, promoting its association with a 14-3-3 protein and nuclear entry . The nuclear role of PKPs is not entirely clear. PKP2 occurs in complexes with RNA polymerase III in nuclear particles and could play a role in regulating transcription of rRNA and tRNA . PKP3 has been identified in cytoplasmic particles containing the RNA-binding proteins poly(A)-binding protein 1, fragile-X-related protein and Ras-GAP (GTPase-activating protein)-SH3 (Src homology 3)-binding protein. In addition, these proteins were found to be associated with PKP3 or PKP1 in stress granules following heat shock along with eIF4 (eukaryotic initiation factor 4) E and the ribosomal protein S6 . The role of PKP in the regulation of translation initiation has recently been observed by Wolf et al.  who show that PKP1 associates directly with eIF4A and stimulates eIF4A-dependent translation. Overexpression of PKP2 also up-regulates Wnt/β-catenin signalling in cultured cells .
Taken together, the above considerations indicate an important role for several desmosomal components in signalling events and it will be fascinating to discover how they contribute to tissue development and homoeostasis.
Since the biochemical and molecular characterization of desmosomes began in the early 1980s, much has been learned about the interactions and functions of their component molecules as well as their assembly, down-regulation and signalling roles, and their contributions to human disease. However, many questions remain. Of crucial importance is the realization that desmosomes in tissues exhibit Ca2+-independent adhesion, which is strongly adhesive or ‘hyperadhesive’. This is fundamental to tissue strength. Almost all studies in culture are done on weakly adhesive Ca2+-dependent desmosomes, although hyperadhesion can be readily obtained in confluent cell culture. Ca2+-dependence is a default condition in vivo, found in wounds and embryonic development.
In order to understand the molecular basis of hyperadhesion, it will be necessary to determine how the structure and molecular interactions within desmosomes give rise to this adhesive strength. Hyperadhesion appears to be associated with an ordered arrangement of the ECDs of the DCs, which gives rise to the intercellular midline identified in ultrastructural studies. We need to determine the specific properties of desmosomal components that enable them to generate hyperadhesion and how these properties differ from those of the components of AJs, which do not become hyperadhesive. Two technical problems will have to be solved in order to gain this information: first, how to resolve the structure of desmosomes with molecular resolution, and, secondly, how to analyse regulatory changes, primarily phosphorylation events, that occur within desmosomes, given their extreme insolubility. Both of these will be helped by greater knowledge of the structure of desmosomal molecules. In order to understand tissue dynamics, much more needs to be known about the mechanisms of desmosome assembly and down-regulation in vivo.
With regard to the signalling roles of desmosomes, for which evidence appears to be accumulating, we need to know whether such signals are generated from within intact hyperadhesive desmosomes, or whether individual desmosomal components have a life outside of desmosomes. We also know virtually nothing about the mechanisms involved. As with considerations of adhesion, it is essential to focus on the situation in vivo. Increased knowledge of structure–function relationships, its regulation and signalling mechanisms will enhance our understanding of the roles of desmosomes in human disease leading in due course to the development of novel therapies.
We thank the Medical Research Council [grant numbers G0700074 and G08000041] and the Wellcome Trust [grant number 086167/Z/08/Z] for financial support.
Abbreviations: AFM, atomic force microscopy; AJ, adherens junction; arm, armadillo; ARVC, arrhythmogenic right ventricular cardiomyopathy; CAR, cell-adhesion recognition; DC, desmosomal cadherin; DP, desmoplakin; DSC, desmocollin; DSG, desmoglein; E, embryonic day; EC, extracellular cadherin repeat; ECD, extracellular domain; ED-SFS, ectodermal dysplasia-skin fragility syndrome; EGF, epidermal growth factor; EGFR, EGF receptor; eIF4, eukaryotic initiation factor 4; EM, electron microscopy; ERK, extracellular-signal-regulated kinase; FRAP, fluorescence recovery after photobleaching; IDP, inner dense plaque; IF, intermediate filament; Lef-1, lymphoid enhancer factor 1; MAPK, mitogen-activated protein kinase; MDCK, Madin–Darby canine kidney; NF-κB, nuclear factor κB; ODP, outer dense plaque; PF, pemphigus foliaceus; PG, plakoglobin; PKC, protein kinase C; PKP, plakophilin; PM, plasma membrane; PV, pemphigus vulgaris; SPPK, striate palmoplantar keratoderma; Tcf, T-cell factor; WT, wild-type
- © The Authors Journal compilation © 2010 Biochemical Society