Desmosomes are intercellular junctions specialised for strong adhesion that are prominent in the epidermis and heart muscle. Defective desmosomal function due to inherited mutations in the constitutive desmosomal gene desmoplakin (DSP) causes skin or heart disorders and in some instances both. Different mutations have different disease-causing molecular mechanisms as evidenced by the varying phenotypes resulting from mutations affecting different domains of the same protein, but the majority of these mechanisms remain to be determined. Here, we studied two mutations in DSP that lead to different dosages of the two major DSP splice variants, DSPI and DSPII, and compared their molecular mechanisms. One of the mutations results in total DSP haploinsufficiency and is associated with autosomal dominant striate palmoplantar keratoderma (PPK). The other leads to complete absence of DSPI and the minor isoform DSPIa but normal levels of DSPII, and is associated with autosomal recessive epidermolytic PPK, woolly hair and severe arrhythmogenic dilated cardiomyopathy. Using siRNA treatments to mimic these two mutations and additionally a DSPII-specific siRNA, we found striking differences between DSP isoforms with respect to keratinocyte adhesion upon cellular stress with DSPII being the key component in intermediate filament (IF) stability and desmosome-mediated adhesion. In addition, reduction in DSP expression reduced the amount of plakophilin 1, desmocollin (DSC) 2 and DSC3 with DSPI having a greater influence than DSPII on the expression levels of DSC3. These results suggest that the two major DSP splice variants are not completely redundant in function and that DSPII dosage is particularly important for desmosomal adhesion in the skin.
Desmosomes are intercellular junctions of epithelial and cardiovascular tissues that connect intermediate filaments (IFs) of adjacent cells, generating a large and mechanically resilient network. The desmosome is formed by the products of three gene superfamilies: the desmosomal cadherins, the armadillo family and the plakins. The transmembrane desmosomal cadherins interact in the extracellular space to couple the two halves of the desmosome and are linked to the IF network through plakoglobin (PG), plakophilins (PKPs), desmoplakin (DSP) and a number of other proteins (Yin and Green, 2004). The importance of desmosomes for maintenance of the strength and flexibility of these tissues is highlighted by natural and in-vitro-engineered mutations in desmosomal genes, which compromise skin or heart and in some instances both (Bolling and Jonkman, 2009; Chalabreysse et al., 2011; Brooke et al., 2012).
DSP is the most abundant component of the desmosome and a key linker plakin protein providing attachment for IFs. Additionally there is published literature highlighting the importance of DSP and other desmosomal components in other adhesion junctions such as complexus adherens junctions in vascular tissue (Zhou et al., 2004). Numerous human mutations in the gene encoding DSP have been identified to date, resulting in a range of clinical phenotypes designated as cardio-cutaneous syndromes. These phenotypes range from non-syndromic skin striate palmoplantar keratoderma (SPPK) through to early lethality due to severe skin blistering and subsequent water loss or sudden death from cardiomyopathy (Bolling and Jonkman, 2009; Brooke et al., 2012). Ablation of DSP in the germline of mice results in lethality at embryonic day 6.5 (E6.5) (Gallicano et al., 1998), and conditional targeting to the epidermis leads to postnatal lethality through skin blistering and extensive water loss (Vasioukhin et al., 2001). Conditional ablation of DSP in the heart of mice results in lethality from E11 onwards due to severe cardiac abnormalities (Garcia-Gras et al., 2006) confirming the earlier demonstration of the importance of DSP in the heart by partial embryonic rescue experiments (Gallicano et al., 2001).
Alternative DSP mRNA transcripts encode two major splice variants, desmoplakin I (DSPI) and desmoplakin II (DSPII), and a minor isoform, desmoplakin Ia (DSPIa), which we have recently identified and characterised (Cabral et al., 2010). DSPI and DSPII are expressed at nearly equivalent levels in stratified epithelia such as the epidermis, but DSPI is the predominant isoform in the heart (Angst et al., 1990; Uzumcu et al., 2006).
Although there are in vivo and/or in vitro data on some DSP mutations, the molecular mechanisms that lead to skin and/or heart disease associated with DSP are not fully understood and it is not clear why different mutations in the same gene can lead to such different phenotypes. Additionally, the functional significance of DSP splice variants remains to be elucidated. Here, we model two different classes of human DSP mutations in HaCaT keratinocytes to investigate their molecular mechanism of action and to gain further insight into the functional role(s) for the major DSPI and DSPII splice variants in keratinocyte biology and disease. Heterozygous premature termination codon (PTC)-generating mutations such as p.Q331X and c.939+1G>A, which lead to haploinsufficiency of both DSPI and DSPII causing dominant non-syndromic SPPK (Armstrong et al., 1999; Whittock et al., 1999), were modelled alongside the DSPI-specific nonsense mutation p.R1267X, which results in complete absence of the DSPI isoform (without affecting DSPII) and causes a recessive mild epidermolytic PPK, woolly hair and early lethal cardiomyopathy (Uzumcu et al., 2006). Heterozygous carriers of this DSPI mutation show no clinical phenotype. We investigated these mutations to gain further insight into the influence of DSPI and DSPII dosages on keratinocyte-associated desmosomes.
siRNA knockdown in HaCaT keratinocytes effectively models DSP haploinsufficiency and DSPI-specific ablation
The DSP haploinsufficiency mutations such as p.Q331X and c.939+1G>A, which cause striate PPK, were modelled by using a siRNA that was described previously (Wan et al., 2007; Cabral et al., 2010) to transiently downregulate all DSP isoforms to about 50% in HaCaT keratinocytes. Cells transfected with this siRNA were designated as siI/II cells. The p.R1267X nonsense mutation targets only DSPI mRNAs. To model this mutation, a pool of three siRNAs was designed to specifically target the DSPI-specific nucleotides (c.3861 to c.5659). The sequences and targeting sites of these three siRNAs (si2, si3 and si4) as well as the position of the DSPI-nonsense mutation (c.4079C>T; p.R1267X) are shown in Fig. 1A,B. Cells transfected with the pool of DSPI-specific siRNAs (siI) were designated as siI cells. A pool of four non-targeting (NT) siRNAs was used as a negative control (NT cells). A western blot showing downregulation of DSPI (which mimics the homozygous DSPI-nonsense mutation) in siI cells and partial downregulation of DSPI and DSPII (mimicking the haploinsufficiency mutation) in siI/II cells with a C-terminal DSP antibody (11-5F) (Parrish et al., 1987) is depicted in Fig. 1C and shows a DSP-depletion pattern indicating effective modelling of these specific mutations. Immunofluorescence analysis confirming the result is shown in Fig. 1D. Densitometry quantification of the western blots (n = 6) is shown in Fig. 1E,F and demonstrates that using this methodology we are able to effectively replicate the DSP levels produced by these specific mutations. The total amounts of DSP are reduced by 40–50% in both knockdowns, but the relative amounts of DSPI and DSPII vary. In the siI cells, DSPI is downregulated by 90% with DSPII unchanged, and in the siI/II cells DSPI and DSPII are both partially downregulated (by 45% and 81%, respectively).
HaCaT cells modelling DSP haploinsufficiency and the DSPI-specific nonsense mutation reveal DSP-isoform specific decreases in DSC3
Heterozygous nonsense and splice site mutations causing DSP haploinsufficiency have been previously shown, by ultrastructural analysis of patient skin, to result in loss of cell–cell contacts, under-developed desmosomes and retraction of the keratin IF network towards the nucleus (Armstrong et al., 1999; Whittock et al., 1999; Wan et al., 2004). In addition, immunohistochemical staining of skin biopsies from patients harbouring these mutations showed decreased expression of PKP1, DSC1 and DSC3, particularly in the spinous layer (Wan et al., 2004), but these three proteins have not been investigated in the skin from the patient homozygous for the DSPI-specific nonsense mutation.
To investigate whether we could reproduce these downstream effects of DSP haploinsufficiency mutations in keratinocytes and compare them with the effects of the DSPI-specific nonsense mutation, we first investigated the influence of these mutations on the expression levels of DSP-binding partners and other junctional proteins by western blot. Two independent siRNA knockdown experiments were conducted and two to four replicate western blots were carried out to assess protein levels for each protein in each experiment. Fig. 2A shows representative western blots for each analysed protein.
No detectable differences were observed between siI or siI/II and NT control cells in the expression levels of the desmosomal components PKP2, PKP3, PG and DSG3 (Fig. 2A), as well as DSG2, β-catenin and keratin 14 (data not shown). Densitometry analysis confirmed consistent and significant reductions in the expression levels of PKP1, DSC2 and DSC3 in both siI and siI/II cells compared with NT control cells. Total DSP expression levels were reduced, on average, to 50% in siI cells and to 40% in siI/II cells when compared to NT control cells (P<0.001, Fig. 1E). Although the difference in total DSP levels between siI and siI/II cells was not statistically significant, the DSPI:DSPII ratios changed, on average, from 1∶8 in siI cells to 4∶1 in siI/II cells (Fig. 2A). In a similar way, PKP1 and DSC2 levels are not significantly different between siI and siI/II cells. In contrast, DSC3 expression levels are significantly higher in siI/II cells than in siI cells (P≤0.05). These results suggest that PKP1 and DSC2 expression levels are not influenced by the DSPI:DSPII ratios, whereas, in contrast, DSC3 expression is influenced by the ratios of these DSP isoforms. Its expression is lower when expression of the DSPI isoform is lower, suggesting that DSPI has a greater influence than DSPII on DSC3.
HaCaT cells mimicking DSP haploinsufficiency demonstrate deficiencies in IF stability and an adhesion defect upon mechanical stress
The keratoderma of patients harbouring heterozygous premature termination codon (PTC) mutations leading to DSP haploinsufficiency is prominent in sites exposed to intense and constant mechanical pressure, such as palms and soles. To mimic such mechanical pressure we subjected HaCaT cell monolayers to oscillating mechanical stress for different periods of time.
Before any stretch was applied (0 hours), NT control, siI and siI/II cells exhibited a normal network of keratin IFs, which extended from the nucleus to the cell periphery (Fig. 3A–C). However, among some siI/II cells, widening of intercellular spaces were observed, which were more pronounced in regions where there was no expression of DSP at cell–cell junctions (Fig. 3C, arrowheads). [This increase in intercellular space was confirmed by low-power electron microscopy (EM) analysis (see Fig. 5, below), eliminating the possibility of it being a consequence of IF contraction.] Keratin IFs appeared slightly thicker and IF attachment appeared stronger in regions where DSP expression levels were higher (Fig. 3C, boxed area). Despite the wider intercellular spaces and loosening of IF attachment, these cells still appeared connected to each other through intercellular junctions.
Following a 30-minute stretch, NT and siI cells had slightly thicker and more compact keratin IFs, particularly around the nucleus (Fig. 3D,E, open arrows). These thicker IFs were even more prominent in the perinuclear region of siI/II cells (Fig. 3F, open arrows). However, close to the plasma membrane of siI/II cells keratin IFs were thinner, particularly along free edges (Fig. 3F, double arrow). In contrast with NT and siI cells, wider intercellular spaces were observed among these cells, particularly in regions where DSPI was absent (Fig. 3F, arrowhead). After a longer time period of mechanical stress (1 and 2 hours), NT cells showed thickening of the keratin IFs, which were denser around the nuclei (Fig. 3G, open arrow) as published elsewhere (Russell et al., 2004). Similar results were observed for siI cells (Fig. 3H,K, open and closed arrows) but, in contrast, although some siI/II cells still showed thick IFs, most siI/II cells had IFs that appeared thinner (Fig. 3I,L) and in some cases they were even deformed and retracted towards the nucleus (Fig. 3L, star). Large intercellular gaps were observed among some cells, particularly after 2 hours of stretch (Fig. 3L, arrowheads).
These results demonstrate that, in contrast with NT and siI cells, which showed thickening of keratin IF bundles with increased exposure to mechanical stretch, siI/II cells exhibited thinner filaments bundles, suggesting an IF stability defect, and wider intercellular spaces suggesting an adhesion deficiency. These observations were prominent in siI/II cells that expressed lower levels of DSP.
Subsequent experiments on the siI/II and NT cells using more severe mechanical stress (a frequency up to 5 Hz, amplitude range up to 0 to 18% and duration of up to 4 hours) revealed that the IF thickening observed in the NT cells, and the IF instability and cell adhesion defects in the siI/II cells were more severe and so these differences were dependent on the relative amount of cell stretch.
To verify these observations, automatic measurement of keratin filament bundle thickness of cells in images acquired from ten random high-power fields of view was performed using Metamorph software (Molecular Devices, CA). This confirmed that in NT control cells the average keratin filament bundle thickness significantly increased upon 2 hours of mechanical stress, from 2.6 pixels to 2.9 pixels (P≤0.05) (data not shown). The data suggested a similar increase in keratin filament bundle thickness in siI cells but this increase was not observed in siI/II cells, confirming the result that the cells mimicking the haploinsufficiency mutants were unable to respond to mechanical stress as observed by thickening of keratin bundles.
Mechanical stress leads to normal localisation of most junctional proteins but partial re-localisation of PG
We performed immunofluorescence staining of stretched and unstretched cells using a panel of desmosomal and some non-desmosomal proteins to investigate their cellular localisation in all three cell types. Before stretch was applied, NT cells showed characteristic PG immunofluorescent staining around the plasma membrane. In contrast, both siI and siI/II cells showed some discontinuous PG staining at the plasma membrane, mostly in regions with reduced DSP expression. Following stretch, siI/II cells showed a decrease in the number of cell–cell contacts, and PG staining at the plasma membrane was only present at the less frequent sites of cell–cell adhesion. A slight increase in intracellular PG staining was also observed in some siI/II cells compared with NT cells, predominantly following mechanical stretch (supplementary material Fig. S1).
Other desmosomal and adherens junction components, including DSG1/2, DSC2, DSC3, PKP2, PKP3, E-cadherin and β-catenin (data not shown) showed no major differences in cellular localization between NT and siI or siI/II cells either before or after stretch.
The role of DSPII was investigated further with a DSPII-specific siRNA knockdown
As the disease-modelling siI and siI/II knockdown cells suggest that there is a more significant role for DSPII in intercellular adhesion compared with that of DSPI, a DSPII-specific knockdown using a trans-splice site strategy was performed. Western blot analysis of siII cells showed near complete depletion of DSPII, whereas the larger DSPI isoform remained unaffected (Fig. 4A, lane 3). Immunofluorescence analysis of keratin filaments showed that there was a similar adhesion defect to that observed in siI/II cells, including thinner IFs (Fig. 4B, double arrow) and IFs deformed and retracted to the nuclei (Fig. 4B, star), and large intracellular gaps were also observed (Fig. 4B, arrowhead).
The levels of DSPII are important for HaCaT cellular adhesion
To independently confirm and quantify the effects of DSP isoform downregulation on cell adhesion, a dispase adhesion assay was performed (Fig. 4C). The cells simulating the DSPI-specific nonsense mutation showed only a small decrease in cell–cell adhesion (indicated by a small increase in the number of monolayer fragments produced upon agitation) in contrast to the siII and siI/II cells, which showed a much larger statistically significant decrease in intercellular adhesion (P≤0.01, Fig. 4C). The defects were greater than those observed in the siI/II haploinsufficiency-mimicking cells (P≤0.01). Additionally, intercellular adhesions could be empirically visualised upon detachment of the cell monolayer from the tissue culture dish by the degree of elastic condensation of the cell sheet. The siI cell monolayer acted identically to the NT control cell monolayer, condensing into a small clump. In contrast, the siII and siI/II cells showed a cell–cell adhesion defect as the monolayer sheet remained as a distinct sheet (data not shown). These results are consistent with the increased severity of the defects observed in the stretch-immunofluorescent assay in the siI/II cells compared with the siI cells (Fig. 3A–L), as well as similar changes in the IF structure in siII cells (Fig. 4B) and suggest that there is a greater role for DSPII compared with that of DSPI in cell adhesion.
Desmosome ultrastructural analysis reveals defects in siII and siI/II cells
We have performed ultrastructural analysis of desmosomes in both unstretched cells and cells that were mechanically stressed for 4 hours. Low power EM (5000×) confirms the findings from the immunofluorescence studies (e.g. the increase in intercellular spaces in siII and siI/II cells, indicative of an adhesion defect). Additionally, siII cells that had been stretched were consistently unable to be imaged owing to the cells becoming detached during the embedding process, presumably because of the adhesion defect caused by the depletion of DSPII.
High power EM (120,000×) gives insights into desmosome ultrastructure. Depletion of DSPI in siI cells does not appear to change the ultrastructure of desmosomes either in stretched or unstretched cells. In contrast, both specific depletion of DSPII in siII cells and haploinsufficiency of both isoforms in siI/II cells leads to an increase in the proportion of desmosomes that appear to have structural deficiencies, including less-well defined inner and outer dense plaques (IDPs and ODPs, respectively). This is particularly prominent in stretched cells (siI/II only) where a defined ODP is not observed and the IDP appears less electron-dense and rudimentary. Additionally, we observed an increase in the proportion of smaller desmosome junctions in both siII and siI/II cells compared with the NT control and siI cells and also a decrease in the total number of distinctive desmosome structures in these cells, particularly in stretched cells when available to analyse. This latter result was independently objectively quantified. Unstretched siII and siI/II cells have a statistically significant decrease in the number of desmosomes per unit area of cytoplasm compared with NT control and siI cells (Fig. 5H, P≤0.05 and P≤0.01, respectively).
Changes in rates of cell proliferation or cytotoxicity are not the cause of the siRNA phenotypes observed
Given that an increase in cell proliferation has been reported in cells mimicking DSP haploinsufficiency (Wan et al., 2007), an MTT-based cell proliferation assay was performed to eliminate differences in cell number being responsible for differences in adhesion (supplementary material Fig. S2). No differences in proliferation were observed between the NT cells compared with any of the DSP siRNA-treated cells, in contrast to the positive control Cbx7 siRNA-treated cells, which are known to display a decrease in proliferation rate (Bishop et al., 2010). This data also eliminates the possibility that differences in cytotoxicity are the cause of the siRNA phenotypes observed, as no decrease in cell number was observed.
The focus of this investigation was to gain insight into the mechanism by which the DSPI and DSPII isoforms contribute to tissue architecture and normal function, by studying the role of two isoform-specific DSP mutations which cause non-syndromic skin disease or cardio-cutaneous disease.
Over the past 13 years, an increasing number of monogenic diseases that are caused by mutations in genes encoding desmosomal proteins have been discovered. Although the genetic defects leading to these diseases are known, the molecular pathogenic mechanisms are, for the most part, not completely understood. Why mutations in DSP can give rise to different clinical presentations in patients is not clear. Some DSP mutations cause phenotypes restricted to skin and/or hair, others affect only the heart and some result in a combination of skin, hair and heart disease. Additionally, varying degrees of severity are observed in skin and heart manifestations depending on the type of mutation. Secondly, there is a lack of correlation between the mutation-harbouring functional domain of DSP and the resulting clinical outcome. For instance, mutations in regions encoding the central rod domain as well as the N- and C-terminal domains can lead to cardiomyopathies, whereas mutations in regions coding for the N-terminus of the protein can cause skin disease only, a combination of skin and hair abnormalities and also a combination of skin, hair and heart disease. This human mutation data suggests that different DSP splice variants might have different adhesion and/or signalling properties and that changes in DSPI:DSPII ratios might influence normal desmosomal function.
Previous reports have shown downregulation of desmosomal components in HaCaT cells upon DSPI/DSPII knockdown and in primary keratinocytes derived from SPPK skin (Wan et al., 2004; Wan et al., 2007). Here, we asked whether downregulation of DSPI (without targeting DSPII) in cells mimicking the DSPI-specific nonsense mutation would cause similar effects on HaCaT cells.
A reduction in expression levels of PKP1, DSC2 and DSC3 was observed upon DSPI knockdown in both siI and siI/II cells, confirming and extending the results published elsewhere (Wan et al., 2004). Previously published studies have shown that under low Ca2+ conditions or conditions that prevent the formation of cell–cell contacts, the turnover of desmosomal proteins is high and they are rapidly degraded. However upon cell–cell contact these proteins are stabilised within the desmosome and their turnover decreases (as evidenced by less degradation) (Pasdar and Nelson, 1988; Penn et al., 1989; South, 2004). We therefore believe the changes we see in expression levels of other desmosomal components occur either at the post-translational level or because DSP is not present in levels sufficiently high to enable their stabilisation in junctions.
Comparison between siI and siI/II cells revealed that, in contrast with PKP1 and DSC2, the DSC3 expression is influenced by the DSPI:DSPII ratios. This increase in DSC3 expression levels from siI to siI/II cells suggests that DSPI has a greater effect on DSC3 expression than DSPII. It is not clear why there is this apparent functional link between DSC3 and the DSPI isoform but it might relate to cell-type-specific expression because DSC3 is primarily a basal cadherin (North et al., 1996). Also, as the basal layer of epidermis has smaller and less electron-dense desmosomes than that of suprabasal keratinocytes (Skerrow et al., 1989; Green and Simpson, 2007), it is possible DSC3 is less important than the upper layer cadherins for adhesion, but might have other functions including cell signalling. This could in turn indicate that DSPI has a less important role than DSPII in maintaining cell–cell adhesion, although the fact that PKP1, DSC2 and DSC3 are all downregulated to some extent upon DSP knockdown suggests that DSPII can compensate partially, but not completely, for the lack of DSPI.
Therefore we next examined directly the influence of DSPI and DSPII on intercellular adhesion and responses to mechanical stress by assessing IF morphology. In response to cyclic mechanical stretch in these experimental conditions, normal HaCaT cells exhibit thickening and wrinkling of its keratin IFs and condensation of bundles, particularly in perinuclear regions, presumably as an attempt to compensate for the stresses imposed by the mechanical stretch. Crucially, this effect is also observed in siI cells but although some siI/II cells also exhibit thickening of keratin IFs around the nucleus, close to the cell membrane these filaments are thinner. These results were confirmed by automatic scoring of keratin filament bundle thickness using image analysis software. Deformation and retraction of thin keratin IFs in siI/II cells is most prominent after 2 hours of stretch, especially in regions where no junctional DSP is observed, and coincides with increased intercellular gaps among these cells, which suggests that there is an adhesion defect.
It is interesting to note the lack or reduction of DSP leads to decreased expression and/or stabilisation of some desmosomal components, including PKP1, DSC2 and DSC3, but not the other components tested. Similarly, DSP depletion causes changes in PG intracellular localisation, but does not change the staining pattern of any of the other components tested. These subtle observed effects, rather than a more global disruption of all components, is consistent with a measurable effect on desmosome function but some degree of cell–cell adhesion being maintained.
Keratinocyte cell lines expressing a keratin 14 mutation that causes epidermolysis bullosa simplex were exposed to repeated stretch and relaxation cycles using the same mechanical stress system (Russell et al., 2004). Although the stretch conditions and cell lines in this study were different, these authors observed some cellular responses that resemble the ones observed here. First, normal cells after mechanical stretch showed thickening and compaction of keratin IFs. Second, keratin mutant cells showed IFs that were unable to withstand the levels of mechanical stretch used, and thus collapsed and fragmented. As a consequence, cell junctions started disassembling. In a similar way, the mutations in DSP studied here lead to decreased levels of DSP which cause weakened IF–desmosome attachment resulting in decreased cell–cell adhesion. It is possible therefore, as suggested by Russell and colleagues (Russell et al., 2004), that loss of IF tension is the driving force for desmosome disassembly because of its importance in maintaining desmosomes. Norgett and colleagues also support this idea by showing that a recessive mutation in the C-terminus of DSP which compromises IF attachment leads to loss of cell–cell adhesion and large intercellular spaces among keratinocytes (Norgett et al., 2006).
Another explanation for the different phenotypes for the two DSP isoforms might be due to their ability to regulate the actin network, possibly directly or possibly through plakophilin. Previously published studies demonstrate that PKP2 contributes to actin-dependent regulation of desmosome assembly (Godsel et al., 2010).
The predicted role of DSPII in cell–cell adhesion that emerged from the initial focus on the two disease-related DSP siRNA models (siI and siI/II) was confirmed by specifically ablating DSPII using a trans-splice site siRNA strategy. A similar adhesion defect to the siI/II cells was observed by way of increased intercellular gaps as well as a similar disruption to the IF network.
This DSPII role in adhesion was independently confirmed with a dispase assay demonstrating a reduction in cell–cell adhesion in both siII and siI/II cells. The adhesion defects observed specifically in cells expressing a reduced amount of DSPII (sII and siI/II cells) demonstrate the key role for DSPII in desmosomal adhesion in keratinocytes.
Our EM ultrastructural analysis revealed that haploinsufficiency of both isoforms (siI/II cells), as well as cells specifically lacking DSPII (siII cells) leads to an increase in the number of desmosomes that appear to have structural deficiencies. Given that DSPII is thought to determine the extent of the IDP (North et al., 1999), the lower levels of DSPII in siI/II cells are likely to be the cause of the more rudimentary and less electron-dense IDPs. IF attachment to these underdeveloped IDPs might subsequently be affected, possibly leading to retraction and collapse of the filament bundles that we observe. Similar results to ours have been shown in the skin of patients with DSP haploinsufficiency, including structural deficiencies in the desmosomes as well fewer cell–cell contacts (Armstrong et al., 1999; Whittock et al., 1999).
It was also apparent that there were fewer desmosome structures in both siII and siI/II cells in comparison with the siI and NT control cells, particularly in stretched cells when available. There are two mechanistic possibilities to explain this. First, depletion of DSPII could directly cause a decrease in the number of assembled desmosomes owing to a failure to recruit the other desmosomal components. As previously published, DSP is crucial to organise the desmosomal complex by recruiting the desmosomal cadherins to the plasma membrane through interactions with PG (Kowalczyk et al., 1997) and/or PKPs (Kowalczyk et al., 1999). This would explain why both the siII and siI/II cells appear to have fewer and structurally defective desmosomes compared with siI cells and control cells. This suggests an obvious mechanism to explain the defective cell–cell adhesion. Additionally, owing to the more rudimentary desmosomes in siII and siI/II cells (as indicated by the increase in the proportion of smaller, less electron-dense junctions), there could be an increase in the number of failed junctions that are being pulled apart, therefore becoming harder to identify. Although we believe it is likely that both possibilities are contributing to the decrease in number of observed junctions in siII and siI/II monolayers, our EM data lends direct weight to the second possibility. We have observed structures indicative of half-desmosomes in stretched siI/II cells, for example as marked by the arrow in Fig. 4, but not in siI or NT control cells.
Our results suggest that DSP haploinsufficiency mutations affect the skin, in particular palmoplantar epidermis, by causing reduced inter-keratinocyte adhesion owing to a dosage defect of DSP splice variants. This amount of DSP seems to be enough, however, for normal heart function. The observation that the DSPI-specific nonsense mutation still affects the skin, although 100% of DSPII is still expressed in this tissue, suggests that compensation of DSPII for the lack of DSPI is only partial.
This study has provided novel insights into the functional molecular mechanism of two isoform-specific mutations which cause non-syndromic skin disease or cardio-cutaneous disease demonstrating differing mechanisms by which the DSP isoforms contribute to desmosomal adhesion and therefore tissue structure and function. While our paper was under review, a report detailing a non-adhesion function for DSP in the epithelium of the murine intestine was published (Sumigray and Lechler, 2012), which is consistent with and supports the data in this manuscript.
Materials and Methods
Cell culture and transient siRNA-mediated DSP knockdown
The HaCaT spontaneously immortalised human keratinocyte cell line was cultured using standard protocols. siI and siI/II siRNA knockdown of DSP isoforms was performed as previously described (Cabral et al., 2010). siII siRNA knockdown was performed using a trans-splice site strategy, with two siRNA duplexes being designed across the DSPII-specific c.3861–c.5659 RNA junction. The Cxb7 siRNA was used as described elsewhere (Bishop et al., 2010).
All antibodies used for western blot and immunocytochemistry are available commercially with the exception of the 11-5F mouse monoclonal antibody against DSPI and DSPII, a generous gift from David Garrod (Faculty of Life Sciences, University of Manchester, UK) (Parrish et al., 1987). The ‘GP’ DSPI-specific guinea pig polyclonal antibody was purchased from Progen (Heidelberg, Germany) along with the mouse monoclonal antibodies against PG, DSC3, PKP1 and PKP2 and the rabbit polyclonal antibody against DSC2. The PKP3 and DSG3 mouse monoclonal antibodies were purchased from Abcam (Cambridge, UK), the monoclonal antibody against vinculin was purchased from Sigma (Poole, UK) and pan-cytokeratin polyclonal antibody was purchased from Dako (Glostrup, Denmark). Alexa Fluor 488- (green) and 568- (red) conjugated anti-mouse or -guinea pig IgG secondary antibodies were purchased from Invitrogen (Paisley, UK).
Cells were plated at high density and grown overnight to reach confluency. Cells were fixed in 1∶1 ice cold methanol:acetone for 5 minutes and immunocytochemistry was performed according to standard procedures. DAPI was used as a nuclear stain at 100 ng/ml. Immunfluorescence images were acquired with a Zeiss LSM 510 laser scanning confocal microscope (Carl Zeiss Ltd, Hertfordshire, UK) and processed using LSM image browser (Zeiss, Germany) and Adobe Photoshop and Illustrator.
HaCaT keratinocytes were seeded at high density to promote junction formation (7×105 cells per well of a 6-well dish) and total protein was extracted 24 hours later using 0.125 M Tris-HCl pH 6.8, 4% (w/v) SDS, 20% (v/v) glycerol, 0.001% (w/v) Bromophenol Blue and 1.44 M β-mercaptoethanol. Lysates were boiled for 5 minutes and resolved on NuPAGE Novex 3–8% Tris-acetate Mini gels (Invitrogen) when western blotting for DSP, or hand-cast 10% SDS-PAGE gels using the BioRad Mini-PROTEAN system for all other proteins, according to the manufacturer's instructions. Proteins were immobilised onto nitrocellulose membrane and immunoblotted using standard procedures.
Densitometric quantification of western blots
Protein levels were calculated from densitometry measurements of western blots using an image analysis program (Image J, NIH, Bethesda, MD) according to the manufacturer's recommendations. These protein levels were normalised to the loading control (vinculin) for each individual sample, and are presented as a fold change of the corresponding protein level of cells transfected with a non-targeting siRNA pool (NT cells). For the DSPI/DSPII graph, DSPI and DSPII protein levels are presented as a fold change of the DSPI level of NT control cells. Total DSP levels were calculated from the sum of the densitometry measurements of DSPI and DSPII. The t-test for matched pairs was used to compare the means of two samples in each western blot.
Flexcell adhesion assay
The Flexcell FX-4000 Tension System (Flexcell, Hillsborough, NC), a computer-regulated bioreactor that uses vacuum pressure to apply cyclic or static strain to cells cultured on flexible-bottomed culture plates, was used to subject a cell monolayer to mechanical stress. HaCaT cells were grown to a high density on BioFlex 6-well plates coated with pronectin (Flexcell) which contain a rubber membrane in each 35-mm well. Each plate was placed over the loading station containing 6 planar faced posts. Cells were subject to cyclic mechanical stretch with a frequency of 2 Hz and an elongation of amplitude ranging from 11 to 14%. Cells were stretched for different periods of time, between 0 hours (unstretched control) and 2 hours, and then prepared for immunocytochemistry as described above. Additionally (including for the EM analysis), more severe stress was applied to the monolayer (frequency up to 5 Hz, amplitude range up to 0 to 18%) for a duration of up to 4 hours.
To test the strength of cell–cell attachment, a dispase assay was performed as described elsewhere (Huen et al., 2002). Cells were grown to a high density in a 60-mm dish and washed once in PBS containing Ca2+ and Mg2+ before being incubated in 5 mg/ml dispase until the monolayer detached. Monolayers were transferred to a 15 ml tube containing PBS and subjected to mechanical stress by agitation (10–20 rapid inversions of the tube). The number of monolayer fragments was counted under a dissecting microscope.
Statistical analyses were carried out using the two-tailed, paired Student's t-test. *P≤0.05, **P≤0.01, ***P≤0.001.
Specimens were fixed in phosphate buffered 4% glutaraldehyde, post-fixed in 1% osmium tetroxide and dehydrated through a graded ethanol series. They were then cleared in propylene oxide and infiltrated with Araldite. The cells were embedded by inverting ‘BEEM’ capsules filled with partly cured Araldite over the monolayer and incubating them at 60°C for 48 hours. After cutting away the silicone membrane, semi-thin sections (0.5 mm) for light microscopy were cut and stained with Toluidine Blue. Specific areas where the cell density was highest were targeted for electron microscopy. Ultrathin sections (60–80 nm) were cut, mounted on copper grids and stained with uranyl acetate and lead citrate. They were examined in a J.E.O.L. JEM 1230 electron microscope and images collected with an Olympus ‘Morada’ 2K × 2K digital camera.
To measure desmosome density, electron micrographs were taken at 5,000× magnification and 10 arbitrary fields per sample were examined. The negatives were scanned into a personal computer using Adobe Photoshop image software. Only DMs displaying recognisable opposing plaques were scored. A coherent single square lattice over was applied to each EM negative to produce a total of 144 test points (1 square is 0.55 µM wide = 220 pixels for measurement purposes). Estimation of DM density was achieved by counting the total number of DMs per negative relative to the total number of test points falling on keratinocyte cytoplasm only. From these counts the DM density was calculated and expressed as a ratio of DM per cytoplasm test points. For the siRNA-treated cells that showed a decrease in desmosome density (siII and siI/II), the outliers that showed a density greater than one standard deviation from the mean, presumably associated with cells in the monolayer with poor knockdown efficiency, were discounted for statistical analysis.
The authors would like to thank Hong Wan (Institute of Dentistry) for technical assistance particularly with the siI/II siRNA, and Ann Wheeler of the Blizard Advanced Microscopy Facility for help with immunofluorescent imaging. We would also like to thank Claire Scott for critical reading of the manuscript.
This study was funded by a grant from the Barts and the London Charity and a MRC studentship. Deposited in PMC for release after 6 months.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.084152/-/DC1
- Accepted February 23, 2012.
- © 2012. Published by The Company of Biologists Ltd