Mechanical stress induces profound remodelling of keratin filaments and cell junctions in epidermolysis bullosa simplex keratinocytes
David Russell, Paul D. Andrews, John James, E. Birgitte Lane


The outer epidermal layer of the skin is an epithelium with remarkable protective barrier functions, which is subject to pronounced physical stress in its day-to-day function. A major candidate component for absorbing this stress is the K5/K14 keratin intermediate filament network. To investigate the part played by keratins in stress resilience, keratinocyte cell lines were subjected to mechanical stress. Repeated stretch and relaxation cycles over increasing time produced reproducible changes in the configuration of the keratin network. When wild-type cells were compared with cells carrying a keratin mutation associated with severe epidermolysis bullosa simplex-type skin fragility, the mutant keratin filaments were unable to withstand the mechanical stress and progressively fragmented yielding aggregates and novel ring structures. The cell junctions into which the keratin filaments are normally anchored also progressively disassembled, with all components tested of the cytoplasmic plaques becoming relocated away from the membrane and onto the keratin rings, while integral membrane receptors integrins and cadherins remained at the plasma membrane. The results suggest that maintenance of desmosomes and hemidesmosomes may require some tension, normally mediated by keratin attachments.


As the flexible outer covering of the body, the properties of skin must accommodate continuous stretching forces. The skin must be strong enough to endure physical trauma yet be flexible enough to allow a large range of movement. A large contribution to the physical resilience of epidermal skin cells comes from the keratin intermediate filaments that they express, as was shown with the discoveries that the inherited skin fragility disorder epidermolysis bullosa simplex (EBS) is caused by mutations in keratin genes (Bonifas et al., 1991; Coulombe et al., 1991; Lane et al., 1992). In EBS, the basal layer cells of the epidermis exhibit fragility under mild physical trauma, resulting in blistering; in severe cases, electron microscopy characteristically reveals electron-dense keratin aggregates in basal keratinocytes of patients (Anton-Lamprecht and Schnyder, 1982).

Keratins belong to the superfamily of intermediate filament proteins. Heterodimeric filaments of type I and type II keratins are expressed in a tissue- and cell-specific manner in epithelial cells (reviewed by Fuchs and Weber, 1994). Stratified epithelia such as the epidermis express K5/K14 in the basal layer of cells and K1/K10 in the suprabasal layers. The keratins form a cytoplasmic network of filaments that are connected to the plasma membrane at complex cell-cell junctions (desmosomes) and cell-substrate junctions (hemidesmosomes). Tissue integrity in the epidermis is maintained by a complex cytoskeleton interaction between the keratin filaments and desmosomes and hemidesmosomes. Keratins are linked into desmosomes through desmoplakin and into hemidesmosomes via plectin; desmosomes are connected from cell to cell by the transmembrane cadherins (desmogleins and desmocollins), and hemidesmosomes are connected to the extracellular matrix substrate by integrins (usually α6β4) and BP180. Other proteins found in these zones include plakoglobin within the plaques of both junctions, plakophilin in desmosomes and BP230 in hemidesmosomes. The exact nature of interactions between all these components is not yet clear (for reviews see Borradori and Sonnenberg, 1996; Jones et al., 1998).

The mutations that cause the severe (Dowling-Meara) type of EBS are mostly found in the highly conserved ends of the α-helical rod domains of keratins K5 or K14. These regions are particularly important in filament assembly (Herrmann and Aebi, 1998). There is a particularly mutagenic codon at residue 125 of K14, in the helix initiation motif, leading to the substitution of the arginine residue at this position (Coulombe et al., 1991); mutations at this residue account for approximately 70% of Dowling-Meara EBS cases (Porter and Lane, 2003). Morphological effects of most keratin mutations are hardly detectable in unstressed cultured keratinocytes but spontaneous keratin aggregates are characteristically seen in cells containing a K14-R125 mutation; this is in keeping with the theory that the helix initiation motif is crucial in normal filament assembly (Herrmann and Aebi, 1998).

To understand exactly how keratins provide stress resilience in cells, cell lines have been generated from EBS patients with mutant keratins (Morley et al., 2003) and subjected to a variety of non-mechanical stresses (Morley et al., 1995; D'Alessandro et al., 2002; Morley et al., 2003). However, a direct mechanical stress is more likely to address the question of physical resilience of keratinocytes. It would also reproduce the in vivo physiological stimulus that leads to breakdown in the epidermis of EBS sufferers. In this paper we present the results of a study of the effects of mechanical stretch on the keratin cytoskeleton in EBS keratinocytes. It was observed that repeated stretch and relaxation caused compaction of the filament network and in EBS cells led to fragmentation of keratin filaments. This was accompanied by changes in the localisation of desmosome and hemidesmosome components. We show that keratin fragmentation in response to mechanical stretch in DM-EBS cells leads to the formation of novel keratin ring structures that contain many nontransmembrane components of desmosome and hemidesmosome junctions. On the basis of these results we present a model for stretch-induced progressive fragmentation of the keratin cytoskeleton and its associated junctions in DM-EBS cells, which has implications for our understanding of cell junction turnover in normal keratinocytes.

Materials and Methods

Cell culture

KEB-7 cells express the severe Dowling-Meara type EBS mutation R125P in the K14 1A domain and NEB-1 cells express wild-type keratins (Morley et al., 2003); both were immortalised using HPV16. All experiments described here were carried out on cells between passage 10 and 20 post-immortalisation. Cell lines were cultured in 75% DMEM/25% Ham's F12 medium, containing 10% fetal calf serum (FCS) and additional growth supplements hydrocortisone (0.4 μg/ml), cholera toxin (10–10 M), transferrin (5 μg/ml), lyothyronine (2×10–11 M), adenine (1.9×10–4 M) and insulin (5 μg/ml). These cell lines are fibroblast feeder cell independent and were cultured at 37°C in 5% CO2.

Mechanical stretch

Flexplates (Flexcell International, USA) were coated with collagen (Sigma) for 60 minutes at room temperature and then incubated for 60 minutes with 5% bovine serum albumin in DMEM supplemented with 10% FCS. Cells were seeded onto six-well Flexplates, and grown to 80% confluence. Wells in flexplates not to be stretched were isolated using FlexStops (Flexcell International, USA). At appropriate times, FlexStops were removed and stretching was resumed in well. Stretching was carried out using FX-4000T™ Cell Stretcher (Flexcell International, USA). Stretch was carried out at a frequency of 4 Hz and an amplitude of 12% for times varying up to 180 minutes at 37°C in 5% CO2. Cells were then harvested or prepared for microscopy. Previous experiments have suggested that keratinocytes are resistant to shear stress (E.B.L. and G. B. Nash, unpublished data). To determine whether fluid shear stress was likely to be significant during stretching experiments we examined the dissipation of drops of a viscous coloured liquid in tissue culture medium using the stretch parameters used in our experiments. We found that when more than 4 ml of medium was placed in flexplate wells there was very little dissipation of the viscous drop. This suggests that the effect of shear force on cells during stretching is minimal and that the observed effects are due to mechanical stretch.


Silicone membranes were excised from Flexplate using a clean, sharp, new scalpel for each well. Cells were fixed and permeabilised using 100% methanol-acetone (1:1) for 5 minutes at –20°C. Cells were then washed twice with PBS. Cells were then treated in blocking buffer (5.5% normal goat serum in PBS) for 1 hour at room temperature. Cells were washed as before then incubated for 1 hour at room temperature with the following primary antibodies: rabbit polyclonal antibody BL-18 (anti-K5) (dilution 1:500) (Purkis et al., 1990), monoclonal primary antibodies LL001 (anti-K14) (Purkis et al., 1990), 11-5F (anti-desmoplakin) (dilution 1:100) (Parrish et al., 1987), HD121 (anti-plectin) (Hieda et al., 1992), CBL175 (anti-plakoglobin) (Cymbus Biotechnology), IE5 (anti-BP230), 233 (anti-BP180) (both gifts from K. Owaribe, University of Nagoya, Japan), PP1-5C2 (anti-plakophilin) (Cymbus Biotechnology Ltd), CD104 (anti-β4-integrin) (Novocastra Laboratories Ltd), 6D8 [anti desmoglein 2 (Dsg2)] (gift from M. Wheelock, University of Toledo, Ohio, OH). All monoclonal antibodies were used at dilution 1:100. Cells were then washed 3×5 minutes in PBS and LL001 primary antibody was detected with FITC-conjugated sheep anti-mouse serum (F3008, Sigma, dilution 1:50). Monoclonal antibodies were detected using Texas Red conjugated anti-mouse serum (N2031, Amersham Pharmacia, UK, dilution 1:50). BL-18 antibody was detected using an Alexa fluorescein-conjugated goat anti-rabbit serum (Molecular Probes, Leiden, Netherlands; dilution 1:400). All antibodies were diluted in DMEM supplemented with 10% FCS and incubations were carried out in the dark.

Fluorescence microscopy

Silicone membranes were mounted onto glass slides using CitiFluor (Sigma Aldrich, UK), air-dried and visualised with an Axiovert 200M system (Carl Zeiss). For higher resolution microscopy, images were acquired using a DeltaVision microscopy system (Applied Precision, USA), using a Nikon PlanApo100×/1.4 N.A. objective lens and a Roper Scientific Interline cooled CCD camera (5 MHz MicroMax1300YHS) using the standard DeltaVision filter set. Optical sectioning was performed at 0.2 μm intervals to encompass the entire cell. Binning was set at 1×1 to give an effective pixel size of 0.067 mm. Three-dimensional (3D) data sets were deconvolved using the SoftWorRx application (Applied Precision, USA). MediaRecorder software (Silicon Graphics Ind., USA) was used to generate TIFF files from either images of single optical sections or from 3D maximum-intensity volumetric projections (generated using SoftWoRx). TIFFS were processed for publication using Adobe Photoshop.

Transmission electron microscopy

KEB-7 cells stretched for 120 minutes were fixed using Karnovsky's fixative (5% glutaraldehyde, 4% paraformaldehyde) containing 1% osmium tetroxide, for 30 minutes. Cells were rinsed three times with phosphate buffer. Cells were then dehydrated through an ethanol series, ending with two 10-minute incubations in 100% ethanol. Cells were incubated overnight in a 100% ethanol/Durcupan (Sigma) solution (1:1). This was then replaced three times with undiluted resin over a 24-hour period. The silicone membrane was then excised and seven BEEM cups (Agar Scientific, Essex, UK) filled with resin were placed upside down onto membrane. Resin was then polymerised at 60°C for 48 hours. The silicone membrane was then peeled off leaving cells attached to resin. 70 nm sections were cut using an Ultracut UCT ultramicrotome (Leica, Vienna, Austria) using a diamond knife (Diatome, Biel, Switzerland). Sections were collected on hexagonal 100 mesh copper grids (Agar Scientific, Essex, UK) then stained with 3% uranyl acetate for 10 minutes followed by lead citrate for 10 minutes. Images were obtained using a JEOL 1200 EX transmission electron microscope (JEOL, Tokyo, Japan) equipped with a Fuji FDL5000 digital plate reader. Images were scanned using a high-resolution imaging plate scanner (DITABIS, Pforzheim, Germany) and processed for publication using Adobe Photoshop.


Keratin fragmentation is induced by mechanical stretch in EBS keratinocytes

In the epidermis of patients with EBS, keratinocytes break down in response to mechanical stress. This occurs because of the mutation of keratin intermediate filaments, which in some way renders the cells more susceptible to cytolysis on physical trauma. We attempted to reproduce this fragility in patient-derived EBS cells by subjecting the cells to oscillating mechanical stretch at a frequency of 4 Hz (i.e. four cycles of stretch and relaxation per second) and a 12% stretch (i.e. a 12% increase in diameter across the silicone membrane). Changes in the keratin intermediate filament network were observed in both wild-type (NEB-1) and mutant keratin EBS (KEB-7) cultured keratinocytes, but no cell lysis was induced. Before any stretch was applied, the NEB-1 (control) cells showed a well-formed network of keratin tonofilaments (filament bundles) throughout the cytoplasm that appeared slightly more dense around the nucleus. Tonofilament-membrane linkages at points of cell-cell adhesion (desmosomes) were numerous and pronounced (Fig. 1A). KEB-7 cells appeared quite similar to the NEB-1 cells before stretch (Fig. 1B), except that the keratin network appeared denser in places as if thicker bundles were present. In KEB-7 cells before stretch, clusters of keratin particles or aggregates appeared spontaneously in places close to the cell periphery. Keratin aggregates are diagnostic of Dowling-Meara EBS (Anton-Lamprecht and Schnyder, 1982) and are seen in situ in both intact and lysed cells, but not in every cell.

Fig. 1.

Mechanical stretch induces keratin fragmentation in DM-EBS keratinocytes. Wild-type cells, NEB-1 (A,C,E) and DM-EBS cells, KEB-7 (B,D,F) were subjected to mechanical stress using an oscillating stretch with frequency of 4 Hz and amplitude of 12% for varying times. Cells were stained for keratin 14 using monoclonal antibody LL001. (A,B) Before stretch; (C,D) after 30 minutes stretch; (E,F) after 120 minutes stretch. All cells show concentric compaction and wrinkling of keratin filaments but only the EBS cells show accumulation of peripheral aggregates of keratin after stretch. By 120 minutes of stretch, the EBS cells show severely disrupted keratin networks with any remaining filaments concentrated around the nucleus. Bars, 10 μm.

After 30 minutes of cyclical stretch, NEB-1 cells started to exhibit thickening of filaments (Fig. 1C) and compaction of bundles became increasingly evident. KEB-7 cells exhibited a dramatic increase in this phenomenon at this time (Fig. 1D). KEB-7 cells showed increased amounts of filament fragmentation, again, particularly along free edges. The keratin network appeared progressively denser around the nucleus, as the whole network retracted from the cell edge. After 120 minutes stretch there was a pronounced collapse of the network in approximately 80% of KEB-7 cells (Fig. 1F). The keratin then mainly exhibited a perinuclear localisation, but there was still a large amount of fragmentation and keratin rings were present at the cell edge. Control NEB-1 cells showed dramatic bundling of filaments at this time point (Fig. 1E) but the network was still well dispersed throughout the cytoplasm. These cells did not exhibit the network collapse that was seen in KEB-7 cells.

The position of a cell within a colony of cells appeared to have an effect on the extent that the keratin network collapses. KEB-7 cells appeared to be afforded some protection from the effects of stretch when they were in the middle of a group of cells, as cells on the edge of a colony showed faster and more dramatic effects than those in the middle. Cells positioned internally, with no free edges, nevertheless did exhibit fragmentation of the keratin network but this was only evident after longer exposure to mechanical stretch (between 1 and 2 hours).

The pattern of keratin fragmentation in KEB-7 cells was uneven around the cells suggesting that uneven forces or uneven response to force was operating in different parts of the cell. Fig. 2 shows that by 10 minutes of oscillating stretch, keratin fragmentation was particularly pronounced in regions where three cells meet and/or at free edges of cells, i.e. areas where there was a dearth of desmosomes and few tonofilament bundles at right-angles to the cell membrane. Fragmentation was also observed in desmosomal areas in response to short periods of stretch. However, the frequency of this was much less than fragmentation observed in areas lacking desmosomes. This suggests that the presence of desmosomes somehow provides the keratin network with an element of resistance to mechanical stretch. As mechanical stress was applied, keratin fragmentation was seen first in these regions. After only 10 minutes of stretch, two distinct types of keratin structures were observed. Solid aggregates of keratin (Fig. 2, inset, arrow) were detected as would be expected in Dowling-Meara EBS keratinocytes. Second, novel ring-like keratin structures were also observed, easily distinguishable from the solid aggregates (Fig. 2, inset, asterisk). These hollow structures have not been previously reported.

Fig. 2.

Keratin aggregate formation is delayed where desmosomes are numerous. DM-EBS cells (KEB-7) stretched for 10 minutes and stained for keratin 14 using monoclonal antibody LL001. Short periods of stretch reveal that desmosomes provide initial resistance to keratin fragmentation and highlight areas of the cell where the keratin network is least resistant to the force of stretch, i.e. at apices where three cells meet and along free edges. Inset shows the presence of two distinct types of keratin aggregate: solid keratin aggregates (arrow) and hollow ring structures (asterisk). Bars, 10 μm.

No effect of stretch on microtubules was observed, either in control or DM-EBS cells before or during stretch. As the microtubule network is highly dynamic it is possible that recovery could have occurred between cessation of stretch and fixation. We were aware of this possibility and endeavoured to keep this time period to a minimum, less than 90 seconds. We also examined the effect of mechanical stretch on the actin cytoskeleton: all cells examined only showed typical epithelial patterns of actin distribution and there was no fragmentation of actin during the cyclical stretch. The formation of actin stress fibres in response to stretch has been previously reported (Shirinsky et al., 1989) and is particularly evident in response to unilateral stretching (Shirinsky et al., 1989; Sugimoto et al., 1991; Takemasa et al., 1998). Radial stretching was used here, which may not lead to the formation of actin stress fibres and may explain the absence of stress fibres in our experiments. We conclude that cyclical stretch as imposed here has no major redistributing effect on the actin cytoskeleton. This suggests that it is unlikely that actin has a direct role in the observed keratin fragmentation.

Effect of mechanical stretch on desmosome morphology

Because keratin fragmentation always starts from the cell edge, we examined the point where keratin filaments attach to the cell membrane, the desmosome. Within the desmosome, desmoplakin is thought to form a direct link between the keratin network and the desmosome plaque (Green et al., 1990). Both NEB-1 and KEB-7 cells make abundant desmosomes and before the application of stretch, desmoplakin staining was similar in both (Fig. 3A,B). As the time of stretch was increased, changes in desmoplakin staining were seen. After 120 minutes of stretch, desmoplakin distribution in NEB-1 cells was unchanged but the shape of the desmoplakin patches at the cell-cell contacts became elongated, suggesting some elasticity in these desmosome cell-cell junctions (Fig. 3C).

Fig. 3.

Mechanical stretch causes relocation of desmoplakin from desmosomes in DM-EBS keratinocytes. KEB-7 cells were stretched using a cyclic stretch with a frequency of 4 Hz and amplitude of 12% for varying times. (A) Control cells before stretch; (B) DM-EBS cells before stretch; (C) Control cells after 120 minutes of stretch; (D) DM-EBS cells after 120 minutes of stretch. Elongation of desmoplakin staining is seen in the majority of control cells after stretch (C, arrow). Desmoplakin staining remains localized to desmosomes where the keratin network is maintained on at least one cell edge (D, arrow). Cells were stained using a polyclonal antibody against keratin 5 (BL18) and a monoclonal antibody against desmoplakin (11-5F). Bars, 10 μm. (E) Desmoplakin staining elongates in response to stretch in wild-type keratinocytes. Forty regions of desmoplakin staining were measured before and after stretch. Results show that desmoplakin staining elongates approximately threefold in response to stretch, suggesting an inherent elasticity within desmosomes that involves desmoplakin.

At this same time point of 120 minutes, the desmoplakin localisation in mutant keratin cells was highly abnormal. Desmoplakin staining was very dispersed in regions where keratin fragmentation was high (Fig. 3D), but remained localised to desmosomes in regions where the keratin network was well maintained on at least one cell edge (Fig. 3D, arrow). This suggests that an intact keratin network is needed to maintain desmoplakin at desmosomes. Furthermore, keratin attached to desmosomes appeared to have some initial resistance to fragmentation, and the desmoplakin staining suggests some elasticity at the desmosome during stretch. We measured desmoplakin staining in wild-type keratinocytes before and after 10 minutes of stretch, taking 40 measurements for each condition. To try and exclude the possibility of desmoplakin recruitment to desmosomes during stretch we only measured desmoplakin staining at the junction where we could see no desmoplakin on keratin filaments close to the junction. It was observed that in response to short periods of stretch, the average length of a desmoplakin staining patch increased approximately threefold (Fig. 3E). This supports the hypothesis that there is some intrinsic elasticity of desmosomes involving desmoplakin.

In these subconfluent cultured keratinocytes, hemidesmosome protein distribution is often virtually reciprocal to desmosome protein distribution, as was seen with antibodies to BP180. The hemidesmosome protein was more sparse in desmosome-rich areas but present in larger amounts where desmosomes were few (e.g. the apex between three cells; see Fig. 4A, arrows). Thus, BP180 localisation also correlates with areas of the cell where keratin fragmentation is seen earliest after stretch. If desmosomes between cells have elastic properties that provide damping of mechanical stress and initially stop filaments breaking, the situation at hemidesmosomes may be different. If there is no elastic capacity in the junction because it is linked to a less pliable surface, such as the dermis – or even worse, a plastic culture dish – it follows that there would be less capacity to absorb force. Thus, breakage of mutated keratin filaments would probably be initiated here.

Fig. 4.

Keratin aggregate formation begins in areas rich in hemidesmosome proteins. (A) Unstretched DM-EBS (KEB-7) keratinocytes stained with polyclonal antibody BL18 against keratin 5 (green) and monoclonal antibody 233 against BP180 (red). Hemidesmosome components are usually concentrated in regions lacking many desmosomes at the periphery. These regions (arrows) coincide with aggregate formation early in stretch as seen in Fig. 2. Bar, 10 μm. (B) Localisaton of keratin aggregates with hemidesmosomal and desmosomal components after 10 minutes stretch. Cells stretched for 10 minutes were scored for colocalisation of keratin aggregates with desmoplakin, BP180 or plectin. 300 cells were counted for each antibody and results expressed as the average number per 100 cells. This shows that after short periods of stretch keratin aggregates form more commonly in areas of the cell high in hemidesmosomal components.

To test this further, KEB-7 cells were stained for either desmoplakin (desmosomes), BP180 (hemidesmosomes) or plectin (hemidesmosomes) and stretched for 10 minutes. One hundred cells were selected at random and scored for colocalisation between keratin aggregates and the above proteins. The only prerequisite for selecting cells was that they exhibited keratin aggregation. Cell scoring was repeated three times so that 300 cells were scored for each protein. The results were plotted to show the average number of cells per 100 showing colocalisation (Fig. 4B). This showed that in response to short periods of stretch an average of only 19/100 cells showed colocalisation between desmoplakin and aggregates. For the hemidesmosomal proteins this was greatly increased; 75/100 for BP180 and 77/100 for plectin. This supports the idea that keratin aggregation begins in areas of the cell rich in hemidesmosomes and that some intrinsic desmosomal elasticity may provide initial resistance to keratin fragmentation in response to stretch.

Stretch-induced ring structures contain keratins plus junction proteins

The altered location of desmoplakin during stretch was examined using deconvolution microscopy. After 120 minutes of stretch, desmoplakin can be seen to associate with small fragments of keratin at points of cell-cell contact (Fig. 5A). Looking at the cell from the cell edge towards the nucleus it is clear that the number of small keratin fragments decreases and the number of ring structures increases (Fig. 5A). This suggests that small keratin fragments are precursors of ring structures. It can be also be seen that desmoplakin is colocalised to these keratin ring structures (Fig. 5A, arrow). The colocalisation of desmoplakin and keratin rings can be seen more clearly in Fig. 6B.

Fig. 5.

Fragmentation of the keratin network leads to relocation of desmoplakin and plectin, but not desmogleins, from desmosomes and hemidesmosomes. DM-EBS (KEB-7) keratinocytes were stretched for 120 minutes and stained for keratin 5 (green) using rabbit polyclonal antibody BL18 and mouse monoclonal antibodies against desmoplakin (11-5F) (A); desmoglein 2 (Dsg2) (B); HD1/plectin (HD121) (C) (all red). 3D maximum-intensity volumetric projections were generated by deconvolution microscopy. Desmoplakin is associated with small fragments of keratin in areas close to desmosomal junctions (A). In areas towards the nucleus there is a decrease in the number of small fragments and an increase in the number of ring structures, which have desmoplakin (A, arrow) and plectin (C) associated with them. Desmoglein 2, a major transmembrane component of desmosomes, does not relocalise in response to stretch (B). Bars, 5 μm.

Fig. 6.

Keratin rings are associated with components of hemidesmosome and desmosome junctions in cells with mutant keratin. Fragmentation of the mutant keratin network in response to stretch leads to small fragments of keratin associated with desmoplakin (A). Keratin fragmentation results in the formation of ring structures, which are intercalated with patches of desmoplakin (B). HD1/plectin associates with these ring structures and is particularly located at adhesion points between rings, associated with the formation of chains of rings (C). BP180 (D), plakoglobin (E), BP230 (F) and plakophilin (G) are also associated with keratin rings. The hemidesmosomal transmembrane protein, β4 integrin, showed no specific localisation with keratin fragments or keratin rings (H and I). Single optical slices were obtained using deconvolution microscopy after 120 minutes of stretch. Cells were stained using polyclonal antibody BL18 against keratin 5 (green) and monoclonal antibodies against desmoplakin (11-5F), HD1/plectin (HD121), BP180 (233), plakoglobin (CBL175), BP230 (IE5), plakophilin (PP1-5C2) (all red) and β4 integrin (CD104) (blue). Bars, 1 μm.

To determine whether the cells were literally being pulled apart, cells were stretched for 120 minutes and stained for desmoglein 2 (Fig. 5B). Desmoglein 2 is one of the specialised desmosome cadherins, a major desmosomal transmembrane protein in cultured epidermal cells that is involved in maintaining the desmosome connection between two cells. Complete fragmentation of the keratin network within a cell did not lead to relocalisation of Dsg2, and there was no association of Dsg2 with keratin rings. The cells were also incubated, before stretch, with a lipophilic membrane dye (DiI, Molecular Probes, Leiden, Netherlands) known to stain internal and external membrane structures. After stretch, no changes in membrane staining were observed and there was no colocalisation of the lipophilic dye with keratin rings (results not shown). Together, these observations confirmed that the cell membranes at desmosomal junctions do not separate during stretch and that keratin rings are not associated with any membrane, as would be the case if they were contained within membrane vesicles.

We examined the localisation of the hemidesmosome protein plectin during stretch. Before stretch, plectin was seen predominantly around the distal areas of the cells, again in regions where one would predict from the above results that keratin fragmentation is initiated. After 120 minutes stretch, plectin was significantly redistributed (Fig. 5C): diffuse staining was seen at the cell periphery and plectin was seen throughout the cell. Plectin was colocalised with fragmented keratin at the cell periphery and could also be seen associated with keratin rings. The images in Fig. 5 were generated as a volume projection and could not confirm that desmoplakin and plectin are directly associated with keratin rings. Single optical slices (Fig. 6) revealed that small keratin fragments were directly associated with desmoplakin (Fig. 6A). Desmoplakin was also directly associated with keratin rings (Fig. 6B). Staining shows that desmoplakin was contained within the ring, giving rings a studded appearance. As desmoplakin associates directly with small fragments of keratin, we predict that these small fragments join together by linkage of keratin and desmoplakin, resulting in a ring structure.

Plectin was also directly associated with keratin rings (Fig. 6C), but unlike desmoplakin staining the plectin appeared to be peripheral, and not integral, to the rings. Chains of rings can be seen in stretched KEB-7 cells and plectin was commonly detected between the rings. This suggests that plectin may be involved in the lateral adhesion of keratin rings. This would be in keeping with its properties as a linker protein with the ability to cross-link keratin filaments (reviewed by Steinbock and Wiche, 1999).

The location of other desmosomal and hemidesmosomal proteins was investigated for association with keratin rings. We found that BP180 (Fig. 6D), plakoglobin (Fig. 6E), BP230 (Fig. 6F) and plakophilin (Fig. 6G) all associate with keratin rings, and staining suggests that these proteins too may interact mainly with the surfaces of keratin rings, unlike the integral interaction of desmoplakin. BP180 is a transmembrane component of the basal membrane of keratinocytes that has been localised to hemidesmosomes and shown to be important for the stabilisation of the hemidesmosome (Hopkinson et al., 1998; Hopkinson and Jones, 2000).

In contrast to the cytoplasmic hemidesmosome components and similar to the desmosome membrane cadherins, staining for the hemidesmosome integrin β4 subunit failed to show any localisation with keratin rings (Fig. 6H,I). This suggests a similar picture as seen for desmosomes: on disruption of the keratin filaments, the transmembrane β4 integrin maintains its connection with the extracellular matrix and stays at the plasma membrane, as the desmosomal cadherins also stay at the membrane in desmosomes, while all the cytoplasmic proteins are relocalised during stretch and interact with fragmented keratin. BP180 was the only transmembrane protein of the junctions to become associated with keratin rings, suggesting that its association with the plasma membrane is less strong than that of the integrins and cadherins. We examined whether both desmosomal and hemidesmosomal proteins were present in the same rings. However, it was difficult to resolve these images to gain useful information. While our observations (not shown) suggest they do co-exist within rings it was difficult to prove this conclusively.

Keratin rings are identifiable by transmission electron microscopy

Transmission electron microscopy of cells stretched for 2 hours revealed structures similar to those seen by immunofluorescence. Rings and whorls of keratin bundles were observed as well as small fragments of keratin (Fig. 7A). More solid aggregates of keratin close to the basal membrane of the cells were observed within stretched cells (Fig. 7B,C), which may be homologous to the aggregates previously observed in EBS keratinocytes. Examination of single keratin rings revealed that they are often angular, as if composed of stiffer segments of greater electron density, which could reflect the areas associated with junctional proteins (Fig. 7D-F). This would fit with the small fragments being precursors of keratin rings. Keratin rings were observed with regions dipping out of the plane of section confirming that keratin rings are not hollow cylindrical structures or hollow spheres of keratin that appear as rings after sectioning.

Fig. 7.

Keratin rings are composed of small fragments of keratin. DM-EBS keratinocytes were stretched for 120 minutes and prepared for transmission electron microscopy. Whorls and rings of keratin filament bundles were frequently observed in the cell periphery (A). Solid aggregates of keratin, reminiscent of the diagnostic hallmark of DM-EBS, were also seen (B,C). Keratin rings were observed and the irregular form of the rings suggests they may be formed of annealed stretches of filament bundles complexed with some associated proteins (D,E). In some cases these structures dipped in or out of the plane of section (E,). Bars, 500 nm.

The keratin filament network of DM-EBS keratinocytes recovers rapidly

After 2 hours of stretch cycles, wild-type and mutant cells were released and allowed to recover, and then fixed and examined at various time points. Even the mutant keratin filament network recovered quickly. After only 2.5 minutes, the mutant cells showed bands of keratin running parallel with the cell edge (Fig. 8A, arrows); these bands were still present after 10 minutes recovery (Fig. 8B, arrows). After 30 minutes (Fig. 8C), a sparse network was already rebuilt; this network continued to develop and after 1 hour, cells appeared fully recovered (Fig. 8D,E). By 24 hours after cessation of mechanical stress, the cells appeared to have a network almost indistinguishable from that of the wild-type cells before stretch (Fig. 8F). Also, there were very few cells containing the low-level aggregates normally seen in mutant keratin keratinocytes. There was also little evidence of the bundling of keratin filaments that we observed in these cells before stretch. Mechanical stretching of both wild-type (NEB-1) and DM-EBS (KEB-7) cells resulted in increased levels of apoptosis (D. Russell, unpublished). Both cell lines exhibited similar levels of apoptosis before stretch, and stretch resulted in increased levels of apoptosis in both cell lines. Interestingly, KEB-7 cells show delayed induction of apoptosis during stretch and show reduced numbers of apoptotic cells during recovery compared with NEB-1 cells (D. Russell, unpublished).

Fig. 8.

DM-EBS keratinocytes recover from stretch by resorption of keratin rings and rebuilding of the keratin network. DM-EBS keratinocytes were stretched for 120 minutes followed by continued incubation at 37°C. Cells were fixed at various time points. Cells quickly begin to recover from stretch-induced malformations of the keratin network by formation of thickened bands of keratin close to the cell edge (arrows) (A,B and G). By 30 minutes after cessation of stretch, the majority of rings and aggregates have gone and the keratin network is being reformed (C). This is completed by 1 hour (D), and a degree of remodelling continues with time (E,F). We examined the keratin band, which often forms during early recovery (G, arrow), and found that keratin rings are resorbed in this zone (H). Cells were stained with monoclonal antibody LL001 against keratin 14 (A-G) and polyclonal antibody BL18 against keratin 5 and monoclonal antibody 11-5F against desmoplakin (red) (H). Images G and H were obtained from 3D maximum-intensity volumetric projections using deconvolution microscopy. Bars, 5 μm.

We examined cells after 10 minutes recovery from stretch to attempt to determine the nature of the keratin band frequently observed early in recovery (Fig. 8G). This band is usually found close to the cell edge and the filaments leading to this band from the nuclear side are very straight. On closer examination, this band of keratin has a diffuse appearance (Fig. 8H), not consistent with normal keratin filament staining. In areas of cells where this diffuse band of keratin is present, keratin rings were not observed on the nuclear side of the band. Keratin rings could be seen within the band with many giving the appearance of `opening up' within the band. The images suggest that keratin rings are resorbed into this band of keratin, which may provide the site at which keratin fragments are `rescued' and remodelled into a filament network. Desmoplakin staining (as well as other junction proteins, not shown), was concentrated within this band of keratin, presumably brought there with the keratin rings.


There is a need for greater understanding of the effects of physical force and the mechanisms of mechanical signalling in epidermis and keratinocytes. As well as providing information to help understand conditions such as EBS, it would also give an insight into processes such as wounding where there is altered skin mechanics around the wound site. Mechanical stretch has been shown to have a direct effect on cells. It can increase DNA synthesis in keratinocytes (Brunette, 1984) and induce signalling via mitogen-activated protein kinase pathways (Kippenberger et al., 2000). Stretch has also been linked to control of proliferation (Chen et al., 1997; Takei et al., 1997). It follows that skin keratinocytes must be able to sense mechanical forces and somehow relay this external signal into a cellular signal. The importance of altered cell shape in the control of growth and apoptosis has also been shown in endothelial cells (Chen et al., 1997). In this paper we have described preliminary experiments aimed at assessing the physical resilience of epidermal keratinocytes, and the way in which these properties depend on the integrity and function of the keratin intermediate filament network. To date, this is the most physiologically relevant stress that has been used to test the function of keratinocytes from EBS patients, which in vivo are known to lyse when the patients' skin is subjected to physical trauma. We have shown that radial stretching of keratinocytes, adhering to a collagen IV-coated flexible membrane, can induce breakdown of the keratin cytoskeleton in cells expressing a dominant negative mutation in K14 of the type that leads in vivo to severe phenotype epidermolysis bullosa simplex cell fragility. Several useful points can be drawn from these results.

DM-EBS keratin filaments fragment on mechanical stress

First, the demonstration that Dowling-Meara mutant keratins do indeed fragment as a result of mechanical stress would support the model that keratins provide internal reinforcement for cells and that the pathogenic mutations lead to filament network breakdown, hence compromising the reinforcing capacity of the keratin cytoskeleton. Cells expressing the K14-R125 mutation, but not cells with wild-type keratin, show fragmentation of the keratin network on mechanical stress in this experimental situation. Mutations at this residue are the most common type found in Dowling-Meara EBS and account for approximately 70% of all DM-EBS cases investigated to date worldwide (Porter and Lane, 2003). What is not directly seen in these experiments of course is the final cytolytic step, which in the patients' epidermis is what actually leads to epidermal blisters. Monolayer cultures of keratinocytes are widely held to mimic basal layer cells in the epidermis, but in situ these cells would be attached to the overlying layers of stratified keratinocytes, directly linked by desmosomes to the epibasal layer and thence indirectly coupled to the higher suprabasal layers, culminating in the hardened layers of the stratum corneum. These layers would provide another dimension to the physical stress acting on the basal cells, which if acting in opposition to lateral displacement in the basal layer could be the final force component that leads to cell lysis. In situ, mechanical stress to the skin would usually involve stress from displacement of the overlying cells and not just shear or compression against the underlying basal lamina and the neighbouring basal cells. In this experimental situation therefore we are only able to examine part of the stress pattern of skin, but this does make it somewhat easier to analyse the consequences of this stretch.

Stretch model allows dissection of the breakdown process

Second, the assay enables the pathogenic process of cytoskeleton breakdown to begin to be analysed. Short periods of cell stretching induce clear deformation of the keratin network irregularly around the cell periphery. The earliest stages examined of this deformation show the appearance of small fragments and aggregates at the cell periphery, to varying degrees in different areas, plus the formation of novel keratin filament rings further into the cell body. Aggregates and ring structures can be seen by electron microscopy to have different internal structures. Ring structures consist of loops or rings of small bundles of keratin filaments, often with an angularity that indicates patches of differential stiffness in the ring, possibly where desmosomal or hemidesmosomal filament-associated proteins are retained. Short fragments of keratin filament bundles are also seen in early stages or in the extreme periphery of the cell and the ring structures probably arise by the annealing of sticky ends of sheared keratin bundles or tonofilaments; keratin filaments are known to assemble extremely rapidly (reviewed by Strelkov et al., 2003).

The keratin rings are usually observed at the distal margins of the filament network; the small aggregates are observed further out than this. Electron-dense aggregates are one of the diagnostic hallmarks of Dowling-Meara EBS and aggregates can occasionally be seen in these KEB-7 cells in a resting state without subjecting the cells to any additional stress, in agreement with observations of Fuchs and colleagues (Coulombe et al., 1991; Fuchs and Weber, 1994). The nature of these aggregates is still unclear and undetermined, but they have been shown to contain both K5 and K14. However, the aggregates generated de novo by stretching in these experiments were noticeably smaller than the spontaneous aggregates seen in biopsies of patients' skin, or spontaneously in cultures of KEB-7 cells, suggesting that they might increase in size with time. This is reinforced by our observation that following recovery of KEB-7 cells from stretch-induced keratin disruption, the newly redistributed keratin initially shows no sign of aggregate formation.

A role for keratin filament tension in maintaining desmosomes

We have observed that in the EBS cells used here with mutant keratin K14 (R125P), the initiation of breakage of the keratin filaments is followed by progressive disassembly of desmosomes and hemidesmosomes. The cytoplasmic proteins of the junctions gradually become relocalised and associate with keratin fragments in the cytoplasm, which form characteristic rings, presumably because breakage of the filaments leaves `sticky' ends. A simple model for this is presented in Fig. 9. Our observations suggest that the disassembly of junctions is initiated by the loss of tension in the filament network. Observations from other pathological conditions may support this idea. A recessive mutation in desmoplakin that disrupts the desmoplakin-intermediate filament interaction has been identified that results in cardiomyopathy with large intercellular spaces, suggesting that intermediate filament attachment is important for maintaining desmosomal junctions (Norgett et al., 2000). Such tension would also be dissipated (1) on extensive phosphorylation of keratins (as occurs during mitosis in lamins and probably also in keratins, judging by the formation of keratin aggregates during mitosis) (Lane et al., 1982; Horwitz et al., 1981; Chou and Omary, 1994; Liao et al., 1997; Toivola et al., 2002) or (2) on breaking of the epithelial barrier (as in epidermal wounding). Both these circumstances call for cytoskeleton remodelling, cell shape change and repositioning of cells with respect to their neighbours in the epithelium. In the first case this would be necessary to accommodate the new daughter cell, and in the second case (more dramatically), to change the shape of the epithelial sheet and mobilize cells for epithelial migration and wound healing. In both cases, and in other physiological situations that can be imagined, disassembly of keratin-subtended junctions on loss of tension in the cell would be a strategy that would make disassembly of unwanted junctions a fast and automatic process.

Fig. 9.

Model for the formation of ring structures in DM-EBS keratinocytes in response to mechanical stretch. Breakage of mutant keratin results in small fragments of keratin attached to hemidesmosomal and desmosomal junctions. The subsequent loss of tension within the keratin filaments leads to the progressive disassembly of these junctions. Cytoplasmic proteins (coloured squares) from these junctions relocalise to the cytoplasm where they interact with keratin filaments which form ring structures presumably as breakage results in `sticky-ends'. Transmembrane components (green) do not relocalise and remain associated with the membrane.

Desmoplakin as an elastic component of desmosomes

We have shown that keratin aggregation is initiated in areas of cells where there are fewer desmosomes. Desmosomes may offer some initial resistance to fragmentation of keratin attached to desmosomes. Stretching of wild-type keratinocytes revealed that after stretch, desmoplakin staining showed an elongated appearance. Membrane staining and staining for transmembrane components of desmosomes showed that there is no cell-cell separation in response to stretch. Mapping of desmosomal components has suggested that desmoplakin may be folded or coiled at desmosomes (North et al., 1999). Our results suggest that in response to stretch, desmoplakin may unfold or result in elongation of coiled desmoplakin, implying some elasticity across the desmosome. The increasing length of patches of desmoplakin staining observed during mechanical stretch supports this idea. It is easy to imagine that this would be physiologically advantageous. Elasticity across the desmosome would enable a force applied to the epidermis to be absorbed across the tissue and would ultimately reduce the force experienced by a single cell. Desmosomes are also abundant in heart tissue and desmosome elasticity could reduce the stress on a single cell during the contraction/relaxation cycle of the heart.

Mechanical stretch is certain to have more to teach us about the function of intermediate filaments in tissues. The availability of this stretch assay, which reproduces at least part of the pathology of EBS in a tissue culture situation, will be useful for analysing the disease process and any hypothetical measures for ameliorating symptoms of this and related disorders. These experiments stress the need for much further analysis of the role of mechanical forces in regulating biological responses at the cellular level.


This work was supported by the Medical Research Council (G78/6810, studentship to D.R.), the Wellcome Trust (068046/Z/02/Z supporting P.D.A.) and Cancer Research UK (C26/A1461 to E.B.L.). The Centre for High Resolution Imaging and Processing (CHIPS) is funded by the Wellcome Trust JIF award scheme and the Medical Research Council co-operative group.

  • Accepted July 9, 2004.


View Abstract