Hyper-adhesion: A Unique Property of Desmosomes

Abstract

Hyper-adhesion is a unique, strongly adhesive form of desmosomal adhesion that functions to maintain tissue integrity. In this short review, we define hyper-adhesion, summarise the evidence for it in culture and in vivo, discuss its role in development, wound healing, and skin disease, and speculate about its molecular and cellular basis.

Keywords::

• desmosome
• cell–cell adhesion
• desmosomal cadherin
• development
• wound healing
• pemphigus

INTRODUCTION: WHAT IS HYPER-ADHESION?

The term “hyper-adhesion” was coined to describe the stronger of two adhesive states exhibited by desmosomes, intercellular junctions that are characteristic of the epithelia and cardiac muscle of vertebrates (Garrod & Kimura, 2008; Garrod et al., 2005). Hyper-adhesion is believed to be crucially important because the principal function of desmosomes is to mediate adhesion strong enough to resist shearing forces that can potentially disrupt tissues by separating cells (Figure 1). Alone, hyper-adhesion is not sufficient to prevent tissue disruption because this can occur via cell breakage as well as via cell separation. Cell breakage is resisted by the intermediate filament (IF) cytoskeleton, composed of keratins in epithelia and desmin in the heart. In order to provide a tissue-wide, force-resistant scaffolding, the IFs are bound to the inner cytoplasmic faces of desmosomes, thus creating the desmosome–IF complex (Garrod & Chidgey, 2008).

Figure 1. Breaking adhesion. An epithelial cell sheet can be disrupted by a shearing force in at least two distinct ways. Either the cells can be physically separated at A giving rise to Aˊ or the cells can be split, broken or fragmented at B giving rise to Bˊ. Both mechanisms will result in cell separation but, strictly speaking, only A-Aˊ is loss of adhesion because it involves direct unbinding of the cell-cell adhesion molecules. B-Bˊmay superficially appears as a loss of adhesion, but the mechanism is indirect. Electron microscopy is necessary to determine whether true loss of adhesion has occurred (Kimura et al., 2007). Red, desmosomes; green, keratin filaments.

2014 is the 150th anniversary of the first brief description of desmosomes as intercellular connections with the light microscope (Bizzozero, 1864). Since then much has emerged about their structure and composition and their role in tissue function, development, and disease. As this article deals only with specific aspects we refer the reader to other recent reviews for broader coverage of the topic. These include desmosome assembly and dynamics (Nekrasova & Green, 2013), mechanistic basis of heart diseases involving desmosomes (Al-Jassar et al., 2013), pemphigus (Kitajima, 2013; Waschke, 2008), desmosomes and cancer (Chidgey & Dawson, 2007), as well as more general reviews (Cirillo, 2014; Dubash & Green, 2011; Kowalczyk & Green, 2013; Thomason et al., 2010).

The operational test for hyper-adhesion involves a simple experimental assay (Garrod, 2013). Tissue culture media and body fluids generally contain calcium at concentrations of 1–2 mM and it has long been known that calcium is necessary to support cell adhesion (Herbst, 1900; Roux, 1894) quoted by Curtis (1967). Much later it was shown that mammalian keratinocytes can proliferate rapidly when the calcium concentration in the medium is reduced to 0.02–0.1 mM, but desmosomes do not form and the keratin filaments are collapsed around the nucleus. When the medium calcium concentration is raised to 1.2 mM, desmosomes with associated tonofilaments form within 1–2 h (Hennings & Holbrook, 1983). Subsequently, we found that such calcium-induced keratinocyte desmosomes rapidly become resistant to chelation of divalent cations with ethylene-diamine tetraacetic acid (EDTA) (Watt et al., 1984). Then we showed, using a simple epithelial cell line, Madin–Darby canine kidney (MDCK), that desmosomes clearly exist in two adhesive forms, which are characterised by the chelation of extracellular calcium with ethylene glycol tetraacetic acid (EGTA), which is effectively a specific calcium chelator (Mattey & Garrod, 1986). One form is calcium dependent because desmosomes in this form lose adhesion and split in half. The other form is operationally calcium independent because prolonged exposure to EGTA has no apparent effect upon desmosomal adhesion or structure. This calcium independent form is referred to as hyper-adhesive because its adhesive strength appears to be greater than that of the calcium-dependent form (Kimura et al., 2007). Thus the assay for hyper-adhesion is to expose cells or tissues to 3 mM EGTA for a minimum of 90 min. If the desmosomes separate into halves they are calcium dependent; if they remain intact they are hyper-adhesive. It was demonstrated many years ago that cells can possess both calcium-dependent and calcium-independent adhesion mechanisms (Beug et al., 1970). However, to our knowledge, ours was the first demonstration that a single adhesion mechanism in a single cell type can adopt both calcium-dependent and calcium-independent forms.

HYPER-ADHESION IN CELL CULTURE

We have investigated the acquisition of hyper-adhesion in cell culture using both the simple epithelial line MDCK and the human keratinocyte line HaCaT (Kimura et al., 2007; Wallis et al., 2000). Several important features have emerged:

1. In order to acquire hyper-adhesive desmosomes cells must be in confluent sheets. Cells maintained at sub-confluence do not become hyper-adhesive.

2. Hyper-adhesion develops slowly in confluent culture, both HaCaT and MDCK cells taking 6 days to become approximately 100% hyper-adhesive.

3. Disrupting confluence by wounding cell layers results in rapid (∼1 h) transition from hyper-adhesion to calcium dependence.

4. Hyper-adhesive desmosomes are more strongly adhesive than calcium-dependent desmosomes as shown by greater resistance to disruption of hyper-adhesive cell sheets.

5. There is no change, qualitative or quantitative, in the major desmosomal components Dsg 2 and 3, Dsc3, DP, PG and PKP 1, 2 and 3 as cells progress from calcium dependence to hyper-adhesion, but there is a slight increase in the amount of Dsc2.

The latter result together with the rapid transition from hyper-adhesion to calcium dependence on wounding suggests that a cell signalling process may be involved in regulating hyper-adhesion. Indeed we have shown that protein kinase C alpha (PKCα) plays a major role in the regulation such that activation of PKCα promotes calcium dependence while inactivation or inhibition promotes hyper-adhesion. The evidence for this is:

1. Treatment of cells with hyper-adhesive desmosomes with phorbol ester causes rapid transition to calcium dependence.

2. Treatment of cells with calcium-dependent desmosomes with PKC inhibitors, including the conventional PKC isoform inhibitor Gö6976, causes rapid transition to hyper-adhesion, even in sub-confluent cells.

3. PKC activity decreases in cells as they progress toward hyper-adhesion.

4. PKCα is located at the cell periphery in cells with calcium-dependent desmosomes.

5. Knockdown of PKCα promotes hyper-adhesion.

6. PKC may not be the sole signalling mechanism to regulate desmosomal adhesion as tyrosine phosphorylation may also play a role (Garrod et al., 2008).

Thus, although we cannot rule out the possibility that acquisition or loss of an unknown desmosomal component is responsible for the transition from calcium dependence to hyper-adhesion, the most likely explanation is that cell signalling, principally involving PKCα, is capable of switching desmosomes rapidly between two alternative adhesive states of different strength.

What could be the function of such a mechanism? Our suggestion is that hyper-adhesion functions to maintain tissue integrity and as such represents the “normal” adhesive state of desmosomes; calcium dependence is the “default” state, which can be rapidly signalled when weaker adhesion is required, for example in wound healing. This appears to be borne out by our studies in vivo.

DESMOSOMAL HYPER-ADHESION IN VIVO

Using electron microscopy (EM) and immunofluorescence of EGTA-treated mouse tissues, we showed that all, including epidermis, trachea, oesophagus, tongue, liver and cardiac muscle, possess calcium independent, hyper-adhesive desmosomes (Wallis et al., 2000). A subsequent quantitative EM study of mouse epidermis found that all desmosomes in normal mouse epidermis were hyper-adhesive (Garrod et al., 2005). This may not mean that calcium-dependent desmosomes are entirely absent but suggests that, if present, they are rare and not easily detectable using EM. Thus it may be that desmosomes are calcium dependent in the early stages of formation, but such early stages have not been detected by us in mature tissues. Earlier studies of frog (Rana pipiens) and mammalian (rat) tissues showed that desmosomes in stratified and glandular tissue were resistant to EDTA, but those in simple columnar epithelia were not (Borysenko & Revel, 1973; Sedar & Forte, 1964). It should be noted, however, that we have found hyper-adhesion in several simple epithelial cell lines, including MDCK, A459 (from airway epithelium), and Caco-2 (from colonic epithelium) (Wallis et al., 2000).

HYPER-ADHESION IS DEVELOPMENTALLY REGULATED

Since desmosomes appear early in mouse tissue development and since developing tissues must remain malleable to participate in morphogenetic movements, we hypothesised that initial weak adhesion would be followed by acquisition of hyper-adhesion (Kimura et al., 2012). We showed previously that desmosomal proteins are expressed during the earliest stages, mesenchymal condensation, of kidney tubule development, but that the desmosomes present during tubule development are immature in form and therefore likely to be calcium dependent (Davies & Garrod, 1995; Garrod & Fleming, 1990). More recently we showed that epidermal desmosomes were calcium-dependent until embryonic day 12 (E12) and became hyper-adhesive by E14 (Kimura et al., 2012). Similarly, blastocyst trophectodermal desmosomes were calcium-dependent on E3 but became hyper-adhesive by E4.5. In both, development of hyper-adhesion was accompanied by the appearance of a midline supporting previous evidence that hyper-adhesiveness depends on the organised arrangement of desmosomal cadherins (DCs). By contrast, adherens junctions remained calcium dependent throughout development but tight junctions became calcium independent as desmosomes mature. Using protein kinase C (PKC) activation and PKCα−/− mice, we provided evidence suggesting that conventional PKC isoforms are involved in developmental progression to hyper-adhesiveness. Furthermore, regulation of desmosomal adhesion by PKC may be important in trophoblast migration during implantation. It appears that tissue stabilisation is one of several roles played by desmosomes in animal development (see Berika & Garrod, 2014 for further discussion).

HYPER-ADHESION IN WOUND HEALING

Hyper-adhesion, which locks epidermal cells together, seems incompatible with cell migration and therefore with wound re-epithelialisation. A series of studies have now shown that desmosomal adhesion changes from hyper-adhesive to calcium dependent in wound edge epithelium.

Firstly, wounding a hyper-adhesive monolayer of MDCK cells caused desmosomes at the wound edge to become rapidly calcium dependent and this change was propagated to cells hundreds of micrometres from the edge (Wallis et al., 2000). Furthermore, the change appeared to be signalled by PKCα. Secondly, on wounding mouse epidermis, desmosomes in the wound-edge epithelium lost hyper-adhesiveness, becoming calcium dependent (Garrod et al., 2005). Transition to calcium dependence was accompanied by translocation of PKCα from cytosol to desmosomal plaques suggesting its activation. Thirdly, we provided direct evidence of the role of PKCα in wound re-epithelialization (Thomason et al., 2012). Thus, PKCα−/− mice showed delayed re-epithelialisation whereas bitransgenic mice over-expressing constitutively active PKCα showed accelerated re-epithelialisation. Furthermore, these effects were associated with delayed loss of hyper-adhesion in PKCα−/− mice and accelerated loss of hyper-adhesion in constitutively active PKCα (CA-PKCα) mice. Moreover, in acute human epidermal wounds, PKCα localised to desmosomes at the wound edge. This was associated with switch of desmosomes to calcium-dependence. However, in chronic wounds desmosomes failed to switch from the hyper-adhesive state and PKCα remained cytoplasmic. These results suggest that manipulation of PKC signalling could provide a novel therapeutic approach for chronic wounds.

HYPER-ADHESION IN PEMPHIGUS VULGARIS

In pemphigus vulgaris (PV), a potentially fatal blistering disease of skin and mucous membranes, autoantibodies to Dsg1 and Dsg3 cause acantholysis (Amagai et al., 1991; Ding et al., 1999). Suggested alternative, but not necessarily mutually exclusive mechanisms for acantholysis, include direct disruption of desmosomal adhesion due to steric hindrance by the autoantibodies and activation of outside-in signalling by autoantibody binding (Muller et al., 2008; Waschke, 2008).

Ultrastructural analysis of human PV appears to show that direct disruption of desmosomal adhesion is not the primary event (Diercks et al., 2009). Rather there is extensive loss of cell–cell adhesion in inter-desmosomal regions and possible intracellular cleavage behind the desmosomal plaque that might indicate a weakening of the cytoskeleton, perhaps through a signalling mechanism involving plakoglobin (Diercks et al., 2009; Muller et al., 2008). By contrast abundant split desmosomes with inserted keratin filaments were found in a mouse model of pemphigus (Shimizu et al., 2004).

Work in which keratinocytes were treated with PV autoantibodies in culture showed that desmosomes were down-regulated and desmoglein 3 was depleted and internalised (Calkins et al., 2006; Cirillo et al., 2006, 2008; Delva et al., 2008; Yamamoto et al., 2007). However, all of this work was presumably carried out with keratinocytes possessing calcium-dependent rather than hyper-adhesive desmosomes. A study in which keratinocytes with calcium-dependent desmosomes and those with hyper-adhesive desmosomes were compared showed that hyper-adhesion inhibited PV autoantibody-induced acantholysis and internalisation of adhesion molecules including desmoglein 3 and E-cadherin (Cirillo et al., 2010). Furthermore, over-expression of plakophilin 1 (PKP1) in keratinocytes caused the desmosomes to become hyper-adhesive and protected them from dissociation by PV antibodies (Tucker et al., 2013). This result is intriguing in relation to previous observations that loss of PKP1 destabilises desmosomes (South et al., 2003). Moreover, PKP1 is predominantly expressed in the upper layers of epidermis where acantholysis does not occur in PV. However, Kimura et al. found that keratinocytes became hyper-adhesive in confluent culture without any change in the level of PKP1 expression (Kimura et al., 2007).

While the primary effect of PV autoantibodies does not appear to be direct disruption of existing desmosomal adhesion, they could inhibit de novo desmosome assembly, which must be a continuous process in the basal layers of stratified epithelia. During the progress of disease this would be expected to result in gradual down-regulation of desmosomes and loss of cell–cell adhesion. It would therefore be interesting to know whether the above-mentioned results obtained with calcium-dependent keratinocytes in culture have any counterpart in vivo. It could also be the case that the primary loss of inter-desmosomal adhesion found in PV might resemble epidermal wounding and thus cause activation of PKC and consequent weakening of desmosomal adhesion as we have described (Garrod et al., 2005). In this case inhibition of PKC could provide a novel therapy for pemphigus (Cirillo et al., 2010; Kitajima, 2013).

DESMOSOME COMPOSITION AND STRUCTURE

In order to discuss hyper-adhesion further, it is necessary to provide a brief background on desmosomes composition and structure. Desmosomes are composed of a small number of well-defined molecular components (Figure 2). These are their adhesion molecules, the desmosomal cadherins desmoglein (Dsg) and desmocollin (Dsc), the plakin desmoplakin (DP) that links the adhesion molecules to the IFs, and the armadillo proteins plakoglobin (PG) and plakophilin (PKP) that link the adhesion molecules to desmoplakin and appear to regulate desmosomal assembly and size. In humans there are four isoforms of Dsg, and three each of Dsc and PKP. For details please see recent reviews (Bass-Zubek et al., 2009; Dusek et al., 2007; Garrod & Chidgey, 2008; Green & Simpson, 2007; Hatzfeld, 2007; Holthofer et al., 2007; Kottke et al., 2006; Sonnenberg & Liem, 2007; Waschke, 2008).

Figure 2. Desmosome structure. (A) Electron micrograph of two desmosomes from human chronic wound epidermis showing the midline between the plasma membranes, the outer dense plaques, and the inner dense plaques with attached keratin filaments (Bar = 100 nm). (B) Diagram showing the quadratic array of desmosomal cadherins in the en face view (x,y plane) of the intercellular region of the desmosome based on the work of Rayns et al. (1969), who describe a periodicity of approximately 70 Å. (C) Diagram (not to scale) showing transverse (z plane) lamellar arrangement of desmosome. The locations of the molecules are based on immuno-gold labelling and electron tomography (See text). ODP, outer dense plaque; IDP, inner dense plaque; KF, keratin filaments; Dsc, desmocollin; Dsg, desmoglein; DP, desmoplakin; PG, plakoglobin; PKP, plakophilin.

Desmosomes have an extremely regular structure (Figure 2). In the transverse or “z” direction desmosomes have a symmetrical, layered structure (Odland, 1958). The space between the plasma membranes of the two cells contributing to the desmosome is sometimes called the desmosomal core or desmoglea. Halfway between the membranes lies a density, the midline, where adhesive binding occurs because the N-termini of the DCs are located there (Al-Amoudi et al., 2007; Shimizu et al., 2005). Lanthanum infiltration and electron tomography of vitreous sections have demonstrated that the DC EC domains are arranged in the en face or “x-y” plane in a quadratic array with a repeat of approximately 70 Å, which is very close to that found the C-cadherin crystal structure (Al-Amoudi et al., 2007; Boggon et al., 2002; Garrod et al., 2005; Rayns et al., 1969). The inter-membrane distance is approximately 35 nm, also very close to the 38.5 nm determined from the C-cadherin crystal structure (Al-Amoudi et al., 2007).

The desmosomal plaques also have a layered structure, resolved into an outer dense plaque (OPD), with its inner face about 20 nm from the membrane, and an inner dense plaque (IDP), which joins the IFs about 50 nm from the membrane (Al-Amoudi et al., 2011; North et al., 1999). Immuno-gold labelling showed the ODP as a region of multiple protein–protein interactions (North et al., 1999). The key observations were: (i) PKP lies close to the plasma membrane; (ii) PG and the N-terminus of DP are further from the membrane and overlap with the C-terminus of Dsc “a”, as well as with the entire cytoplasmic domain of Dsg3, which lies in the ODP; (iii) the C-terminus of Dsc “b” is closer to the membrane and spatially separated from PG and DP; (iv) the C-terminus of DP lies in the IDP, roughly 40 nm from the membrane. This is consistent with the predicted length of the shorter spliced form, DPII, and suggests that DPI is coiled or folded. Broadly speaking, the locations of these molecules are consistent with their interactions determined in vitro. We have since determined that the repeat unit domain (RUD) of Dsg1/2 (which have larger cytoplasmic domains than Dsg3) is located in the ODP near its inner face and the terminal domain (TD) internal to the ODP but not extending to the IDP (Scothern and Garrod; unpublished).

Electron tomography of the ODP produced a molecular map with a resolution of 3.2 nm (Al-Amoudi et al., 2011). This showed a 2D interconnected quasiperiodic lattice with a similar spatial organisation to the extracellular side. Based on the above immuno-gold labelling, the transverse organisation was resolved into an outer 4 nm-thick plakophilin (PKP) layer and an inner, denser, 8 nm-thick plakoglobin (PG) layer, which also contains the N-termini of DP molecules. At present it is not clear how the cytoplasmic domains of the DCs fit into the ODP.

DESMOSOME STRUCTURE VARIES WITH ADHESIVE STATE

Comparison of the ultrastructure of desmosomes in normal and wound-edge epidermis showed that those of normal epidermis, which were hyper-adhesive, exhibited very prominent midlines whereas those at the wound edge, many of which were calcium dependent, generally lacked midlines and the intercellular space appeared amorphous (Garrod et al., 2005; Thomason et al., 2012). Moreover, comparison of the width of the intercellular space showed that the plasma membranes of wound-edge desmosomes were slightly closer together than those in normal epidermis (21.9 ± 0.5 nm and 23.9 ± 0.5 nm, respectively, Garrod et al., 2005). Also, during development, calcium-dependent desmosomes in both the blastocyst and the epidermis lacked midlines but midlines were present when hyper-adhesiveness had developed (Kimura et al., 2012). The onset of hyper-adhesion in the developing epidermis also precedes establishment of the adult pattern of desmosomal cadherin expression and no changes in expression of desmosomal cadherin isoforms occur between E12 and E14 (Chidgey et al., 1997; King et al., 1996). Thus it appears that in hyper-adhesive desmosomes the extracellular domains of the DCs have a more regular or ordered arrangement than they do in calcium-dependent desmosomes.

It is ironical that when the crystal structure of Type 1 cadherin extracellular (EC) domains was first elucidated it was hailed as a desmosome-like adhesion zipper, rather than as a model for adherens junctions (Laskey, 1995). In our experience adherens junctions always remain calcium dependent, so one suspects that they are unable to become hyper-adhesive. Moreover, although there has been one report of densities within the adherens junction EC region (Miyaguchi, 2000), they generally exhibit no ordered structure in this region. This suggests that Type 1 cadherins do not adopt an ordered arrangement in vivo even though they do so in crystals. In our experience the EC domains of DCs are difficult to crystallise, yet they can adopt an ordered arrangement in vivo. This implies a fundamental difference between Type 1 and desmosomal cadherins, even though the two cadherins types show many apparent similarities of sequence and function.

Thus both Type 1 and DCs probably have a similar structure of their EC subdomains (Garrod et al., 2005). We have shown that the Dsg EC domain, like Type 1 cadherins, has a calcium-dependent structure: in the absence of calcium they are globular, but in its presence they are more extended (Tariq, H., Rouhi, M., Nie, Z., Jowitt, T., Holmes, D., Bella, J., Baldock, C., Garrod, D., and Tabernero, L., manuscript in preparation). Both cadherin types seem to exhibit predominantly homophilic adhesion mediated by strand exchange between the EC1 domains, involving insertion of the hydrophobic side chain of a conserved tryptophan (Trp2) into a hydrophobic pocket on the adhesion partner (Nie et al., 2011). Given these apparent similarities, what may be the key differences between them?

We have shown by molecular dynamic simulation that in the presence of calcium, that is, under physiological condition, DCs appear more flexible than Type 1 cadherins (Tariq, H., Rouhi, M., Nie, Z., Jowitt, T., Holmes, D., Bella, J., Baldock, C., Garrod, D., and Tabernero, L., manuscript in preparation). While we are not sure of the reasons for this difference, it may suggest a crucial difference in in vivo behaviour and crystallisability. Thus in vivo, where they are constrained by their membrane insertion and cytoskeletal attachment, Type 1 cadherin EC domains may be unable to adopt and ordered arrangement. By contrast, desmosomal cadherin EC domains, because of their flexibility, can become ordered despite their basal attachments. An analogy may be the constraint imposed on packing of unsaturated lipids in layers compared to saturated lipids. The presence of double bonds generates and element of rigidity making unsaturated lipids less flexible than saturated.

It seems probable that transmembrane signals emanating from the desmosomal plaque are involved in this regulation. Thus desmosomal adhesiveness is regulated by protein kinase C alpha (PKCα), activation promoting calcium dependence and inhibition promoting hyper-adhesion (Garrod et al., 2005; Kimura et al., 2007; Thomason et al., 2012; Wallis et al., 2000). Multiple signals may be involved since inhibition of tyrosine phosphatases (i.e., promotion of tyrosine phosphorylation) also promotes hyper-adhesion (Garrod et al., 2008).We suggest that phosphorylation of one or more plaque components causes rearrangement within the plaque and transmits a signal to the EC domains. Desmosomes lacking PKP1 or PG do not seem to acquire hyper-adhesion (South et al., 2003; McHarg, Mueller and Garrod: unpublished) so these plaque components may be important in its regulation. On the other hand, desmosomes lacking DP or carrying DP mutations that prevent IF attachment seem to acquire a midline (Jonkman et al., 2005; Sumigray & Lechler, 2012). Phosphorylation does, however, regulate DP–keratin binding and keratins have been shown to regulate PKCα activity and DP phosphorylation (Hobbs and Green, 2012; Kroger et al., 2013). Enhanced association of keratin with DP through blocking DP phosphorylation provides another mechanism for strengthening desmosomal adhesion (Hobbs & Green, 2012).

CONCLUSION

We believe that hyper-adhesion is an important concept. It is important functionally because it helps to explain how intercellular junctions contribute to the strength of vertebrate tissue. It is also important because it distinguishes desmosome from the other major type of adhesive junction, the adherens junction. The adherens junction appears more malleable than the desmosome and is likely to have a more important signalling function, but probably cannot develop the same adhesive strength. This may suggest an explanation of why both types of junction are essential components of vertebrate tissues; they have complementary roles that combine to support tissue homeostasis. Elucidation of the molecular basis of hyper-adhesion and its role in development and disease will make an essential contribution to our understanding of tissue function.

Declaration of interest: The authors report no declarations of interest. The authors alone are responsible for the content and writing of the paper.