Abstract
Desmogleins are cadherin family adhesion molecules essential for epidermal integrity. Previous studies have shown that desmogleins associate with lipid rafts, but the significance of this association was not clear. Here, we report that the desmoglein transmembrane domain (TMD) is the primary determinant of raft association. Further, we identify a novel mutation in the DSG1 TMD (G562R) that causes severe dermatitis, multiple allergies, and metabolic wasting (SAM) syndrome. Molecular modeling predicts that this G to R mutation shortens the DSG1 TMD, and experiments directly demonstrate that this mutation compromises both lipid raft association and desmosome incorporation. Finally, cryo-electron tomography (cryo-ET) indicates that the lipid bilayer within the desmosome is ∼10% thicker than adjacent regions of the plasma membrane. These findings suggest that differences in bilayer thickness influence the organization of adhesion molecules within the epithelial plasma membrane, with cadherin TMDs recruited to the desmosome via establishment of a specialized mesoscale lipid raft-like membrane domain.
Summary Lewis, Caldara, and colleagues show that the length of the desmoglein transmembrane domain is important for lipid raft association, and that a mutation within the desmoglein-1 TMD that abrogates raft association causes a desmosomal skin disease (SAM syndrome). Cryo-ET indicates that the lipid bilayer within desmosomes is thicker than surrounding plasma membrane, suggesting that the desmosome is a specialized lipid raft-like membrane domain.
Introduction
A characteristic feature of epithelial cells is the assembly of specialized plasma membrane domains that mediate cell adhesion, communication, and barrier function (Garcia et al., 2018; Sumigray and Lechler, 2015). Among these structures, adherens junctions and desmosomes play overlapping but distinct roles in cell adhesion, signaling, and morphogenesis (Garcia et al., 2018). Desmosomes are particularly abundant in tissues exposed to mechanical stress, including the skin and heart (Amagai and Stanley, 2012; Delmar and McKenna, 2010; Hsu et al., 2018). These adhesive complexes are characterized by highly organized and dense arrangements of desmosomal proteins that can be visualized by electron microscopy (Al-Amoudi et al., 2007; Delva et al., 2009; Harmon and Green, 2013). Considerable progress has been made in identifying protein interactions that mediate adhesion in both adherens junctions and desmosomes, as well as the associations that anchor these adhesive structures to the cytoskeleton (Holthofer et al., 2007; Nekrasova and Green, 2013; Quinlan et al., 2017; Troyanovsky, 2012). However, the physical constraints imposed by the epithelial plasma membrane that contribute to the segregation of adherens junctions and desmosomal complexes into morphologically, biochemically, and functionally distinct structures are poorly understood.
The adhesive core of the desmosome is comprised of single pass transmembrane desmosomal cadherins termed desmogleins and desmocollins that mediate adhesion between adjacent cells (Al-Amoudi et al., 2007; Harrison et al., 2016; Saito et al., 2012b). In humans, there are four desmoglein genes (DSG1-4), along with three desmocollins (DSC1-3) (Simpson et al., 2011). The desmosomal cadherins are coupled to the intermediate filament cytoskeleton through adaptor proteins such as plakoglobin, plakophilins, and the cytolinker protein desmoplakin (Al-Amoudi et al., 2011; Delva et al., 2009; Jones et al., 2017). These interactions form an electron dense plaque that couples the adhesive interactions of the desmosomal cadherins to the intermediate filament cytoskeleton of adjacent cells, thus conferring tissue resilience to mechanical stress (Hsu et al., 2018; Thomason et al., 2010). Loss of desmosome function results in skin (Hsu et al., 2018; Payne et al., 2004) and heart (Delmar and McKenna, 2010; Samuelov and Sprecher, 2015) diseases characterized by tissue fragility. In the skin, loss of desmosomal adhesion manifests clinically as epidermal blisters and erosions (Brooke et al., 2012; Payne et al., 2004), and in some disorders, aberrant thickening of the epidermis (Has and Technau-Hafsi, 2016; Stahley and Kowalczyk, 2015). One example of such a disease is severe dermatitis, multiple allergies, and metabolic wasting (SAM) syndrome (Samuelov et al., 2013). This disease is typically caused by null mutations in DSG1, leading to epidermal fragility and barrier defects (Cheng et al., 2016; Has et al., 2015).
We and others have previously demonstrated that desmosomal proteins associate with lipid rafts and that this association regulates desmosome assembly and desmoglein endocytosis (Brennan et al., 2012; Delva et al., 2008; Nava et al., 2007; Resnik et al., 2011; Stahley et al., 2014). Lipid rafts are sphingolipid and cholesterol enriched membrane microdomains that introduce spatial heterogeneity into lipid bilayers (Lingwood et al., 2009; Pike, 2004; Pike, 2006; Simons and Sampaio, 2011). These domains are critical for protein trafficking, membrane organization, and signaling (Diaz-Rohrer et al., 2014a; Honigmann and Pralle, 2016; Levental and Veatch, 2016; Santos and Preta, 2018; Sezgin et al., 2017). The sphingolipids present in rafts feature long saturated acyl chains that, along with cholesterol, contribute to the more ordered, densely packed, and thicker membrane environment characteristic of lipid rafts (Brown and London, 2000; Honigmann and Pralle, 2016; Lingwood et al., 2009). Desmogleins and other desmosomal proteins have been shown to associate with lipid raft membrane domains as determined by detergent resistance and buoyancy on sucrose gradients (Brennan et al., 2012; Nava et al., 2007; Resnik et al., 2011; Stahley et al., 2014). In addition, disruption of lipid rafts by removal of cholesterol from cellular membranes results in weakened desmosomal adhesion and defects in desmosome assembly, suggesting that lipid rafts play a critical role in desmosome homeostasis (Resnik et al., 2011; Stahley et al., 2014).
In the present study, we sought to determine the mechanisms by which raft association governs desmosome assembly, and the determinants of desmoglein partitioning to rafts. Our results indicate that the transmembrane domain (TMD) of the desmogleins is critical for raft association, whereas the E-cadherin TMD does not support raft targeting. This association appears to be essential for desmoglein function, as a novel mutation within the TMD of human DSG1 abrogates lipid raft association, impairs desmosome targeting, and causes the human skin disease SAM syndrome. Using chimeric proteins, raft association assays, and molecular modeling, we determine that the extended length of the desmoglein TMD is a key feature driving its raft association. Furthermore, cryo-electron tomography and sub-tomogram averaging demonstrates that the lipid bilayer within the desmosome is thicker than the adjacent plasma membrane, consistent with previous studies predicting that the lipid bilayer is thicker at raft domains compared to non-raft membranes. Thus, our results support a model in which the desmosome is a specialized type of lipid raft membrane microdomain, and that the lengthy desmoglein TMD enables efficient desmosome incorporation by facilitating desmoglein partitioning into the thicker desmosomal bilayer. These findings suggest that epithelial junctional complexes achieve plasma membrane domain specification not only through selective protein interactions, but also through constraints imposed by the biophysical characteristics of the plasma membrane. Collectively, these findings reveal a novel pathomechanism for a human disease and demonstrate the importance of desmosomal protein association with lipid rafts in epidermal homeostasis.
Results
Palmitoylation of Dsg3 is not required for lipid raft association
Desmogleins and other desmosomal components associate with lipid raft membrane microdomains (Delva et al., 2008; Nava et al., 2007; Resnik et al., 2011; Stahley et al., 2014). A number of raft associating proteins, including plakophilins, utilize palmitoylation as a membrane raft targeting mechanism (Greaves and Chamberlain, 2007; Levental et al., 2010; Roberts et al., 2014). Palmitoylation is a reversible post-translational modification that occurs when palmitoyltransferases add a 16-carbon fatty acid (palmitate) to cysteine residues (Charollais and Van Der Goot, 2009; Munday and Lopez, 2007). Sequence alignments (Fig. 1 A) reveal that desmosomal cadherins contain conserved cysteine residues at the cytoplasmic face of the transmembrane domain, and our previous studies have shown that these residues are critical for desmoglein palmitoylation (Roberts et al., 2016). We hypothesized that palmitoylation of desmogleins would mediate lipid raft association. Therefore, we mutated cysteines 640 and 642 to alanine residues in murine Dsg3 and used a lentiviral expression system to generate stable A431 cell lines expressing FLAG tagged wild type or mutant Dsg3(CC). Mass tag labeling confirmed that the mutation of these conserved membrane proximal cysteines eliminated Dsg3 palmitoylation (Fig. 1 B). Interestingly, the loss of Dsg3 palmitoylation had no discernable effect on lipid raft association as determined by Dsg3 incorporation into buoyant and detergent resistant membranes (DRM) (Lingwood and Simons, 2007) (Fig. 1 C). In addition, both WT Dsg3 and Dsg3(CC) localized to cell-cell borders as assessed by widefield immunofluorescence (Fig. 1 D). Furthermore, Dsg3 and Dsg3(CC) exhibited similar Triton-X 100 solubility, suggesting no defect in the desmosome or cytoskeletal association of Dsg3(CC) (Fig. 1 E-G). Collectively, these results indicate that palmitoylation is not required for lipid raft association of desmogleins or for normal Dsg3 subcellular distribution in quiescent A431 monolayers.
The transmembrane domain of desmogleins mediates lipid raft association
In addition to palmitoylation, emerging evidence indicates that lipid raft association of membrane spanning proteins is also regulated by the physiochemical properties of the transmembrane domain (TMD) (Lorent et al., 2017). In particular, TMD length is a critical determinant for targeting to lipid rafts (Diaz-Rohrer et al., 2014b; Garcia-Garcia et al., 2007; Lorent et al., 2017; Scheiffele et al., 1997). Sequence alignments (Fig.1 A) indicate that the TMDs of the desmogleins, which associate with rafts, are considerably longer (24 amino acids) than the corresponding TMDs of classical cadherins, such as E-cadherin (21 amino acids) and VE-cadherin (20 amino acids), that exhibit minimal raft association (Stahley et al., 2014). Recent studies indicate that the free energy of raft association can be calculated based on TMD length, surface area, and palmitoylation (Lorent et al., 2017). These parameters predict efficient WT Dsg1 raft partitioning (ΔGraft=0.17), with markedly lower raft affinity for the TMD of E-cadherin (ΔGraft=0.30) (Table 1). To directly test if the TMD is the principle motif conferring lipid raft association on the desmoglein family of proteins, we generated a chimeric cadherin in which the Dsg3 TMD was replaced with the E-cadherin TMD (Dsg3(EcadTMD)). Lentiviral transduction was used to generate stable A431 cell lines expressing either wild type Dsg3-FLAG or Dsg3(EcadTMD)-FLAG. Sucrose gradient fractionations demonstrated that the Dsg3(EcadTMD) chimera was virtually excluded from DRM fractions when compared to wild type Dsg3 (Fig. 2 A and B). Immunofluorescence localization indicated that Dsg3(EcadTMD) localized to cell-cell contact sites. However, Triton-X 100 extraction showed decreased insoluble pool partitioning as assessed by both immunofluorescence (Fig. 2 C) and western blot analysis (Fig. 2 D-F), suggesting decreased Dsg3(EcadTMD) association with cytoskeletal elements relative to wild type Dsg3. Expression of the Dsg3(EcadTMD) mutant caused no apparent changes in endogenous E-cadherin distribution (Fig. 2 A, D-F). To determine if the Dsg3 TMD is sufficient to confer lipid raft targeting, we constructed interleukin 2 receptor (IL2R) α chain-Dsg3 chimeric proteins comprising the IL2R extracellular domain coupled to the Dsg3 cytoplasmic tail with either the IL2R TMD or the Dsg3 TMD (Fig. 2 G and Saito et al., 2012a). The IL2R-Dsg3 chimera harboring the Dsg3 TMD partitioned to DRM fractions, whereas the chimera containing the IL2R TMD did not partition with DRM fractions. Collectively, these studies indicate that the Dsg3 TMD is the primary determinant of Dsg3 raft association.
To test if the TMD of other desmoglein family members also functions in raft association, similar experiments were conducted in the context of Dsg1. Dsg1 WT and a Dsg1(EcadTMD) chimera were generated. Both proteins were tagged with a carboxyl terminal green fluorescent protein (GFP) and stably expressed in A431 cell lines as described above. Similar to the Dsg3(EcadTMD) chimera, the Dsg1(EcadTMD) chimera showed a marked decrease in association with DRM fractions as determined by sucrose gradient fractionation (Fig. 3A and B). Additionally, Dsg1(EcadTMD) was partially excluded from Triton insoluble fractions of cell lysates (Fig. 3 C-E), similar to the results seen with Dsg3(EcadTMD) (Fig. 2 D-F). Lastly, both the WT Dsg1 and Dsg1(EcadTMD) demonstrated border staining characteristic of desmogleins (Fig. 3 F). Together, these results demonstrate a central role for the TMDs of the desmoglein family in lipid raft association.
A mutation in the transmembrane domain of DSG1 causes severe dermatitis, multiple allergies, and metabolic wasting syndrome
Loss of DSG1 function is associated with a number of autoimmune, infectious, and genetic diseases (Payne et al., 2004; Samuelov and Sprecher, 2015; Stahley and Kowalczyk, 2015). One recently discovered desmosome associated disease is severe dermatitis, multiple allergies, and metabolic wasting (SAM) syndrome (Samuelov et al., 2013). Most instances of SAM syndrome are caused by homozygous functional null mutations in the desmosomal cadherin desmoglein 1 (DSG1) (Cheng et al., 2016; Has et al., 2015). Here, we report a novel and dominantly inherited heterozygous DSG1 missense mutation within the DSG1 TMD (Fig. 4). The probands presented with ichthyosiform erythrokeratoderma, diffuse palmoplantar keratosis and multiple allergies (Fig 4 A). Proband III-2 suffered metabolic wasting and died of status asthmaticus and recurrent infections. Hemotoxylin and eosin staining of skin biopsied from the proband revealed compact hyperkeratosis with parakeratosis, frequent detachment of the entire stratum corneum, and dissociation of individual corneocytes (Fig 4 B). Although these findings indicate an adhesion defect, we observed minimal alterations in desmosome ultrastructure when patient epidermis was examined by electron microscopy (Fig 4 C). These clinical and genetic observations led us to diagnose the patient with SAM syndrome. Unlike previously reported instances of DSG1 mutations in SAM syndrome (Cheng et al., 2016; Danescu et al., 2017; Has et al., 2015; Samuelov et al., 2013), this patient harbored a novel missense mutation in DSG1 which introduces a hydrophilic arginine residue (p.G562R) into the otherwise hydrophobic transmembrane domain of DSG1 (Fig 4 D-F). Subsequent to our characterization of this initial family, a second unrelated individual was identified with a G562R heterozygous mutation previously reported as a case of erythrokeratoderma variabilis (Ishida-Yamamoto et al., 2000). The parents of this patient lacked this mutation and were disease free. Together, these observations demonstrate that a heterozygous G562R mutation in the DSG1 TMD causes a human skin disease best characterized clinically as SAM syndrome.
To determine how the G562R mutation impacted DSG1 organization in patient skin, biopsies from the proband were processed for immunofluorescence microscopy. DSG1 levels were markedly reduced (∼40%) in the spinous and granular layers of patient epidermis (Fig 4 G and H), and DSG1 localized in cytoplasmic puncta and aberrant clusters at cell-cell borders. Interestingly, DSG1 staining in patient stratum corneum was markedly increased, perhaps reflecting increased antibody penetration. Desmoplakin levels were slightly reduced in patient epidermis, whereas DSG3 levels were markedly increased (Fig. 4 I-K). To further investigate alterations in DSG1 distribution, we performed structured illumination microscopy (SIM) on patient and control epidermis. DSG1 fluorescence intensity within patient and control desmosomes was comparable in basal keratinocytes, where DSG1 expression is low and other DSG isoforms (DSG2, DSG3) are expressed. However, DSG1 fluorescence intensity in patient desmosomes was significantly reduced (Fig. 4L and M) in the spinous and granular layers where DSG1 is prominently expressed. Thus, although morphologically normal desmosomes could be observed by electron microscopy (Fig. 4C), these desmosomes apparently lack sufficient DSG1 levels to support normal epidermal cohesion.
To investigate the mechanism by which the DSG1(G562R) mutation causes SAM syndrome, GFP-tagged murine wild type Dsg1α and a mutant harboring the equivalent G-to-R substitution, Dsg1(G578R), were expressed in A431 epithelial cells. Widefield fluorescence imaging revealed that both WT Dsg1 and Dsg1(G578R) were present at cell-to-cell borders (Fig 5 A). Interestingly, the Dsg1(G578R) mutant also exhibited a prominent perinuclear staining pattern. There were no obvious differences in desmoplakin localization in the two cell lines (Fig. 5 A), and plakoglobin generally co-localized with both cell-cell border and perinuclear pools of WT Dsg1 and Dsg1(G578R) (Fig. 5 B). To determine if the Dsg1(G578R) mutant was defective in desmosome targeting, SIM was performed and Dsg1 fluorescence intensity was measured at cell borders both within and outside of individual desmosomes to control for possible variations in Dsg1 levels at different cell-cell contact sites. We observed that Dsg1(G578R) displayed a decreased association with desmosomes when compared to WT Dsg1 (Fig. 5 C and D). Furthermore, parallel bands of GFP fluorescence signal overlapping with DP were routinely observed for WT Dsg1-GFP but not for Dsg1(G578R)-GFP (Fig. 5 C and E). Lastly, WT Dsg1 efficiently entered a detergent resistant pool, consistent with incorporation into insoluble desmosome and cytoskeletal-associated complexes, whereas mutated Dsg1(G578R) remained predominantly soluble (Fig. 5 F-H). Together, these findings indicate that the G-to-R TMD mutation reduces DSG1 incorporation into desmosomes in both cultured cells and in patient epidermis.
In addition to being deficient in desmosome targeting, we also observed that DSG1 was present in cytoplasmic puncta in SAM patient epidermis (Fig. 4 G) and that Dsg1(G578R) was concentrated in perinuclear compartments in A431 cell lines (Fig. 5 A and B). To determine if the Dsg1(G578R) mutant exhibited membrane trafficking defects, cell surface proteins were biotinylated and pulse chase experiments conducted to measure Dsg1 turnover rates. These experiments revealed no difference in the rate of Dsg1(G578R) turnover from the plasma membrane compared to WT Dsg1 (Fig. 6A and B). To measure rates of delivery to the plasma membrane, cell surface proteins were cleaved using trypsin and the rate of Dsg1 recovery at the cell surface was monitored by biotinylation (Fig. 6 C and D). While the surface pool of WT Dsg1 recovered within 3-6 hours, Dsg1(G578R) exhibited a severe reduction in the rate of plasma membrane accumulation. To determine if Dsg1(G578R) was being retained in secretory compartments, A431 cell lines were grown in low calcium medium overnight to internalize all cadherins, and subsequently switched to high calcium medium to allow Dsg1 to traffic out to cell-cell borders. These experiments revealed that Dsg1(G578R) was retained in GM130-labeled compartments (Fig. 6 E and F), indicating that the G-to-R mutation causes retention of Dsg1 in the Golgi apparatus, delaying its trafficking through the secretory pathway.
Disease causing mutation abrogates lipid raft association of Dsg1
Based on our findings that the TMD of the desmogleins is critical for lipid raft association, we hypothesized that the G-to-R mutation observed in SAM patients would prevent Dsg1 from partitioning to lipid rafts. Calculations based on parameters which predict free energy of raft association from TMD sequences (Lorent et al., 2017) revealed that introduction of the SAM-causing G-to-R mutation into the DSG1 TMD dramatically alters the energetics of raft association (Table 1). Consistent with these calculations, sucrose gradient fractionations revealed that Dsg1(G578R) was virtually absent from lipid raft (DRM) fractions (Fig. 7A and B). Interestingly, we observed a notable reduction in plakoglobin association with DRM in cell lines expressing Dsg1(G578R), suggesting that the DSG1 mutant also recruited plakoglobin out of raft domains. To further test the ability of WT Dsg1 and Dsg1(G578R) to associate with lipid rafts, these proteins were transiently expressed in rat basophilic leukemia cells and giant plasma membrane vesicles (GPMV) were chemically isolated (Levental and Levental, 2015a; Levental and Levental, 2015b). Non-raft membrane domains were labelled with F-DiO, a dialkylcarbocyanine dye. WT Dsg1 efficiently partitioned into areas of vesicles lacking F-DiO, indicating partitioning to the liquid ordered, raft domain (Fig. 7 C and D). In contrast, Dsg1(G578R) was almost entirely co-segregated with F-DiO and excluded from the liquid ordered membrane domain, indicating minimal raft affinity. Together, these findings reveal that the G-to-R TMD mutation reduces DSG1 association with lipid rafts.
The lipid bilayer within desmosomes is thicker than non-desmosomal membranes
The results above illustrate a critical role for the desmoglein TMD in the association of this family of cadherins with lipid raft membrane microdomains and for its crucial role in epidermal homeostasis. To understand how the physiochemical properties of the desmoglein TMD confer raft and desmosome targeting, structural models of the TMDs of wild type Dsg1, the Dsg1(G578R) SAM mutant, and E-cadherin were generated by the Robetta structure prediction server (Kim et al., 2004). The modeling predicts that the SAM-causing G-to-R mutation interrupts the Dsg1 TMD helix and significantly shortens the run of helical hydrophobic residues (Table 1 and Fig. 8 A), potentially deforming the lipid bilayer as phospholipids position to maintain energetically favorable interactions between hydrophilic and hydrophobic amino-acid residues (de Jesus and Allen, 2013; Nezil and Bloom, 1992). These findings are consistent with the notion that the SAM causing mutant disrupts lipid raft association by shortening the DSG1 TMD, thereby increasing the energy cost of entering the thicker lipid bilayer characteristic of lipid raft domains (Lorent et al., 2017).
Experiments in model membranes suggest that the high cholesterol content (Kucerka et al., 2008; Nezil and Bloom, 1992) and long, saturated acyl chains (Lewis and Engelman, 1983; Niemela et al., 2007) that constitute lipid raft domains contribute to significant thickening of raft phospholipid bilayers relative to non-raft regions of the membrane. A prediction derived from such observations, and from the implication of rafts in desmosome composition and function, is that the lipid bilayer within desmosomes would be thicker than non-desmosomal regions of the plasma membrane, thereby accommodating the lengthy desmoglein TMD. To test this possibility, cryo-electron tomography and sub-tomogram averaging was performed on mouse liver samples enriched in the plasma membrane fraction (Fig. 8 B). The thickness of lipid bilayers measured in the sub-tomogram averages within desmosomal and non-desmosomal regions of the plasma membrane was then determined. This analysis revealed that desmosomal bilayers were 10% thicker (4.5±0.4 nm) than regions immediately adjacent to the desmosome (4.0±0.3 nm, p = 2.2E-122) or at arbitrary regions of membrane visible within the tomograms (4.0±0.3 nm, p = 6.4E-104) (Fig. 8 C). Together, these findings suggest that desmosomes represent a highly specialized plasma membrane domain that is characterized by lipid raft associated proteins and a thickened phospholipid bilayer characteristic of lipid raft-like model membranes.
Discussion
Lipid rafts have emerged as important membrane microdomains that regulate membrane organization, endocytosis, and signaling (Honigmann and Pralle, 2016; Levental and Veatch, 2016; Lingwood et al., 2009; Pike, 2006; Sezgin et al., 2017). Desmosomal proteins have been shown to associate with lipid rafts in a variety of epithelial cell types (Brennan et al., 2012; Delva et al., 2008; Nava et al., 2007; Resnik et al., 2011; Stahley et al., 2014), but the mechanisms and physiological relevance of this association are poorly understood. Here, we report that the transmembrane domains of the desmogleins are the key determinants for targeting these cadherins to lipid rafts. A mutation within the DSG1 TMD that shortens this domain abrogates both lipid raft partitioning and desmosome association, and leads to the human skin disease SAM syndrome. Cryo-electron tomography reveals that the lipid bilayer within the desmosome is markedly thicker than the adjacent lipid bilayer, thereby favoring incorporation of the longer desmoglein TMDs into this plasma membrane domain. Collectively, our results suggest that desmosomes are a specialized mesoscale lipid raft-like membrane domain.
Essential functions for desmogleins have been exposed by human diseases in which desmogleins are targeted by autoantibodies, infectious agents, or genetic mutation (Amagai and Stanley, 2012; Stahley and Kowalczyk, 2015). DSG1 is the primary desmoglein expressed in the outermost layers of the epidermis, and DSG1 loss of function mutations lead to at least two different types of epidermal disorders. Haploinsufficiency of DSG1 causes palmoplantar keratoderma (Lovgren et al., 2017), whereas complete loss of DSG1 leads to SAM syndrome (Cheng et al., 2016; Has et al., 2015; Samuelov et al., 2013). Most individuals afflicted with SAM syndrome succumb to chronic infection in early childhood (Samuelov et al., 2013). Here, we report two separate instances of SAM syndrome, one inherited and one sporadic, caused by a glycine to arginine substitution (G562) within the hydrophobic transmembrane domain (Fig. 4). Arginine residues play an important role in terminating TMDs and establishing TMD orientation within the lipid bilayer (Parks and Lamb, 1993; Reddy et al., 2014), suggesting that the glycine to arginine substitution shortens the DSG1 TMD. This notion is supported by molecular modeling (Fig 8).
Our data indicate that shortening of the DSG1 TMD by insertion of an arginine residue disrupts DSG1 function in SAM syndrome patients by preventing lipid raft association. TMD length correlates positively with raft association (Lorent et al., 2017), and our structural predictions and molecular modeling predict that desmoglein TMDs confer raft association (Table 1 and Figure 8A). This notion is consistent with our findings using both classical DRM fractionation experiments (Fig. 2 and 3) and direct observations of partitioning into liquid ordered domains (Fig. 7). Surprisingly, palmitoylation does not appear to be required for desmoglein raft association (Fig. 1 and Roberts et al., 2016), although it does impact desmoglein dynamics at the plasma membrane (Roberts et al., 2016). Recently, the desmosomal component plakophilin also was shown to be palmitoylated (Roberts et al., 2014), further demonstrating an important role for this reversible posttranslational modification in regulating desmosome assembly dynamics. Further studies will be needed to assess how palmitoylation is utilized in combination with other physiochemical properties of the desmoglein TMD to modulate the trafficking and adhesive properties of these unique cadherins.
A prediction based on our finding that the desmoglein TMD is responsible for partitioning to lipid rafts is that the lipid bilayer within desmosomes should be thicker than surrounding non-desmosomal membrane. Indeed, cryo-electron tomography revealed that the lipid bilayer within the desmosome is substantially thicker than non-desmosomal regions of the plasma membrane (Fig. 8 B-D). These observations indicate that it would be energetically costly for the DSG1 G-to-R SAM mutant to enter the thicker bilayer present in desmosomes due to a hydrophobic mismatch between phospholipids and Dsg1 TMD amino-acid residues (de Jesus and Allen, 2013; Marsh, 2008). Therefore, it is likely that shortening of the TMD in the SAM mutant and failure to enter the thicker lipid bilayer domain of the desmosome represents a central pathophysiological mechanism of this disease-causing mutation. Indeed, we observed that the DSG1 G-to-R mutant is deficient in entering desmosomes both in patient epidermis (Fig. 4) and when expressed in cultured epithelial cell lines (Fig. 5). We also find that Dsg3 and Dsg1 polypeptides harboring the shorter E-cad TMD are unable to associate with lipid rafts (Fig. 2, Fig. 3). The predicted E-cadherin TMD is 21 amino acids, compared to the 24 amino acid TMD of desmogleins (Table 1). Although these chimeras do not effectively enter lipid rafts as assessed by DRM fractionation assays, we do find that these Dsg(EcadTMD) chimeras can be seen within desmosomes as assessed by SIM (not shown). It is likely that for these chimeras, protein-protein interactions mediated by the desmoglein cytoplasmic and extracellular domains can partially overcome the energy cost of incorporating into the thicker bilayers present in the desmosome. In addition, mismatch of TMD length and hydrophobic thickness of the bilayer can be accommodated by changes in TMD tilt within the membrane and by local bilayer deformation (de Jesus and Allen, 2013; Nezil and Bloom, 1992). In contrast, the predicted 16 amino acid TMD of the Dsg1(G578R) mutant is significantly shorter than the TMD of both desmogleins and E-cadherin, and therefore its entry into desmosomal membranes is apparently energetically prohibitive.
Together, our observations support a model in which adherens junctions and desmosomes assemble into distinct plasma membrane microdomains based not only on protein interactions, but also due to the biophysical nature of the epithelial plasma membrane and the TMD characteristics of different cadherin subfamilies (Fig. 9). Interestingly, early studies of desmosomal composition found that these junctions are enriched in sphingolipids and cholesterol, key components of what are now referred to as lipid rafts (Drochmans et al., 1978; Skerrow and Matoltsy, 1974). In addition, most of the major desmosomal proteins are palmitoylated, including the desmosomal cadherins and plakophilins, whereas adherens junction components lack this modification (Roberts et al., 2014; Roberts et al., 2016). Given the key role for palmitoylation in lipid raft association, these findings further suggest that affinities for different lipid domains of the plasma membrane are central features that distinguish adherens junction and desmosomal proteins. Further studies will be needed to determine the precise structural and functional characteristics of different cadherin TMDs and how they selectively dictate lipid raft association. In addition, it will be important to discern how TMD characteristics are used in conjunction with lipid modifications such as palmitoylation to sort desmosome and adherens junction components into distinct plasma membrane domains with unique morphologies and functions. These features appear to be of fundamental importance for skin physiology, as our findings reveal that a mutation altering the structure of the desmoglein transmembrane domain is a novel pathomechanism of a desmosomal disease, and raise the possibility that other human disorders may result from alterations in lipid raft association or raft homeostasis. Indeed, loss of lipid raft targeting may be an under-appreciated pathomechanism in human diseases which were previously conceived as generalized protein trafficking defects.
The authors declare no competing financial interests.
Author Contributions
Materials and methods
Subjects
All affected and healthy family members or their legal guardians provided written and informed consent in accordance with the guidelines of the Institutional Review Board of Keio University and Emory University School of Medicine in adherence to the Helsinki guidelines. The investigators were not blinded to the allocation during experiments and outcome assessment.
Mutation analysis
Whole-exome sequencing was performed using genomic DNA isolated from the probands (II-2 and III-2) and their parents (I-1, I-2 and II-1). Whole exome sequencing libraries were constructed using SureSelect Human All Exon V5 (Agilent) and sequenced by HiSeq2500 (Illumina). Sequencing reads were mapped to a human reference genome sequence (hs37d5) by BWA software (0.7.12-r1039). The mapped reads were realigned and variation sites were detected by GATK-3.30 software. The detected variation sites were annotated by SnpEff/SnpSift 4.1d software. Since the phenotype appeared in the proband II-2 (delivered from healthy parents) and transmitted to the proband III-2 (Fig. 4D), we searched for genetic variations that de novo mutated in the proband II-2 and transmitted to the proband III-2. Only one variation was identified to fulfill the criteria, which was c.1684G>A (p.G562R) of DSG1, coding for the desmosomal cadherin desmoglein 1. Sanger sequencing confirmed the mutation was identified in the probands but not from other healthy family members (Fig. 4D, 4E). The mutation had not been identified in cohort studies (Genomes Project et al., 2015; Higasa et al., 2016; Nagasaki et al., 2015; Sudmant et al., 2015). The whole exome sequencing of the probands II-2 and III-2 revealed no other variations in the exons and exon-intron boundaries of DSG1.
Immunohistochemistry and electron microscopy of patient samples
Biopsies were embedded in optimum cutting temperature (OCT) solution and stored at −80°C. Prior to immunostaining, 5 µm cryosections were prepared on glass microscope slides. Primary and secondary antibodies are described below. Sections were sealed using mounting medium (ProLong Gold by ThermoFisher Scientific) and a coverslip. For electron microscopic studies, the biopsied sample was fixed in an ice-cold 2% glutaraldehyde/60 mM Hepes (pH 7.4) buffer followed by fixation with 1% osmium tetroxide, staining with 1% uranyl acetate, and embedding in Epon812. Ultrathin sections were stained with 1.5% uranyl acetate and Reynolds lead citrate and examined with an electron microscope (JEM-1010, JEOL) at the accelerating voltage of 80 kV.
Construction of mutants
Constructs were cloned using PCR and mutagenesis by the Cloning Division within Emory Integrated Genomics Core or purchased through Cyagen vectorbuilder services.
Structural Predictions
Sequences for transmembrane domains were analyzed using the Robetta structure prediction server (Kim et al., 2004)
Cell line generation, culture, and reagents
A431 cells were cultured in DMEM (Corning 10-013-CV) with 10% fetal bovine serum (Hyclone SH30071.03) and 1% penicillin/streptamycin (Corning 30-004-CI). Cells were stably infected with lentiviruses expressing the various murine desmoglein constructs. 5 µg/mL blasticidin was used to select for infected cells. No clonal isolation was performed. Cell lines expressing wild type and mutant DSG1-GFP were subjected to fluorescence activated cell sorting in order to obtain populations with roughly equal DSG1-GFP expression levels. For experiments utilizing a calcium switch, low calcium medium was prepared as described previously (Wilson, 2014): no calcium DMEM (Gibco/Molecular Probes 21068028), 10% fetal bovine serum, calcium chelating BT Chelex 100 resin (Biorad 143-2832), and 1% penicillin/streptamycin.
Immunofluorescence
A431 cells were cultured to ∼70% confluence on glass coverslips. In experiments in which pre-extraction is explicitly used, cells were treated with PBS+ containing 0.2% Triton X-100 and 300 mM sucrose on ice for 1 min prior to fixation. Cells were fixed in 3.7% paraformaldehyde in PBS+ on ice for 10 min. Cells were permeabilized in PBS+ containing 0.1% Triton X-100 and 3% bovine serum albumin for 10 min. Non-specific antibody binding was prevented with a blocking step in PBS+ containing 3% bovine serum albumin and 0.05% Triton X-100. Primary and secondary antibodies (listed below) were diluted into blocking solution. For rinse buffer, we used PBS+ containing 0.2% bovine serum albumin and 0.05% Triton X-100. Cells were mounted to glass microscope slides using prolong gold mounting medium (described above).
Antibodies
Mouse anti-DSG3 AK15 was described previously (Tsunoda et al., 2003). Rabbit anti-calnexin (Enzo Life Sciences ADI-SPA-860). Mouse anti-desmoplakin1/2 (Fitzgerald 10R-D108AX). Rabbit anti-desmoplakin NW6 was a kind gift from Dr. Kathleen Green (Northwestern University). Mouse anti-plakoglobin (gamma catenin) (BD TransLabs 610253). Mouse anti-E-cadherin (BD Biosciences 610252). Mouse anti-flotillin 1 (BD 610820). Mouse anti-flotillin 2 (BD 610383). Rabbit anti-Green Fluorescent Protein Life A11122). Rabbit anti-FLAG (Bethyl A190-102A). Secondary antibodies conjugated to Alexa Fluors were purchased from Invitrogen. Horseradish peroxidase-conjugated secondary antibodies were purchased from Biorad.
Image acquisition and processing
Widefield fluorescence microscopy was performed using a DMRXA2 microscope (Leica, Wetzler, Germany) equipped with a 100X/1.40 NA oil immersion objective and narrow band pass filters. Images were acquired with an ORCA digital camera (Hamamatsu Photonics, Bridgewater, NJ) and processed using Fiji ImageJ. Super-resolution microscopy was performed using a Nikon N-SIM system on an Eclipse Ti-E microscope system equipped with a 100X/1.49 NA oil immersion objective, 488- and 561-nm solid-state lasers in 3D structured illumination microscopy mode. Images were captured using an EM charge-coupled device camera (DU-897, Andor Technology) and reconstructed using NIS-Elements software with the N-SIM module (version 3.22, Nikon). All microscopy was performed at room temperature and imaging results are representative for at least two independent experiments containing at least 10 cells each.
Desmosome targeting analysis using SIM
To quantify desmosome targeting in cultured cells, Dsg1.GFP fluorescence was measured within regions of interest (ROI) drawn around desmoplakin railroad track staining at cell-cell borders. This Dsg1.GFP fluorescence intensity was compared to adjacent ROI at regions of cell borders lacking desmosomes. For both wild type and mutant Dsg1, targeting to desmosomes was measured as a fold-enrichment of Dsg1.GFP fluorescence in desmosomes compared to non-desmosomal regions. For SAM patient and control tissue, desmosomal ROIs were defined using desmoplakin railroad tracks and DSG1 fluorescence was measured therein.
Triton solubility/insolubility
A431 cells were cultured until confluent in 6 well tissue culture plates. Cells were washed twice with ice cold phosphate buffered saline. The triton soluble pool was isolated by incubating cells with triton buffer (1% Triton X-100, 10 mM Tris, pH 7.5, 140 mM NaCl, 5 mM EDTA, 2 mM EGTA, with protease inhibitor) for 10 min on ice. Lysate was then centrifuged at 16,000 x g for 10 min at 4°C to pellet triton insoluble fraction. Triton-soluble supernatant was collected and mixed 1:1 with 2x laemmli sample buffer containing 5% B-mercaptoethanol. The triton-insoluble pellet was resuspended in 2X laemmli sample buffer (Biorad 161-0737) sample buffer containing 5% β-mercaptoethanol. All samples were heated to 95°C for 10 minutes, vortexed for 30s half way through, prior to being run on a gel for western blotting.
Isolation of detergent resistant membranes
Detergent resistant membranes were isolated as described previously (Lingwood and Simons, 2007). Briefly, cells were cultured in 25 cm2 flasks (two per gradient) and washed with PBS+. Cells were collected by scraping in TNE buffer supplemented with protease inhibitors (Roche) and pelleted by centrifugation at 0.4 x g at 4°C for 5 min (5415R, Eppendorf). Cells were re-suspended in TNE buffer and homogenized using a 25-guage needle. TNE buffer containing Triton X-100 was added (final concentration of 1%) and cells were incubated on ice for 30 min. 400 µL of detergent extract was mixed with 800 µL of 56% sucrose in TNE and placed at the bottom of a centrifuge tube. 1.9 mL volumes of 35% and 5% sucrose were layered on top of the sample. Following an 18 hour centrifugation at 4°C (44,000 rpm, SW55 rotor, Beckman Optima LE-80 K Ultracentrifuge), 420 µL fractions (1–11, remaining volume combined to make up fraction 12) were removed from top to bottom of the gradient and stored at −20°C until processed for western blot analysis. Flotillin-1 and Flotillin-2 were used as raft markers while calnexin was used as a non-raft marker. Unless otherwise stated, all films shown are representative for at least three independent experiments.
Giant plasma membrane vesicle (GPMV) isolation and partitioning measurements
GPMVs were isolated and imaged as described (Levental et al., 2009; Sezgin et al., 2012). Before GPMV isolation, cell membranes were stained with 5 µg/mL of FAST-DiO (Invitrogen), a fluorescent lipidic dye that strongly partitions to disordered phases because of double bonds in its fatty anchors (Levental et al., 2011).
Biotin labeling in pulse-chase experiments
For Dsg1 cleavage and recovery experiments, cells were grown to confluence in 35 mm cell culture plates (Corning 430165). Cells were trypsinized using TrypLE (Gibco 12605-010) for ∼8 min and suspended. After the indicated refractory period, surface proteins were biotinylated. For experiments monitoring protein turnover from the plasma membrane, surface proteins were biotinylated before the indicated period. Biotinylation was achieved using PBS+ containing 0.5 mg/ml EX-Link sulfo NHS SS Biotin (Thermo Scientific 21331) for 30 min at 37°C. Unbound biotin was quenched in PBS+ containing 50 mM NH4Cl for 1 min. Cells were lysed in RIPA (PBS+ containing 1% Triton X-100, 0.1% sodium dodecyl sulfate, 0.1% sodium deoxycholate, 10 mM Tris-HCl, 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and protease inhibitor cocktail (Roche 11836170001)), scraped to transfer from culture plate to an Eppendorf tube, and incubated for 10 min on ice. Lysate was cleared via centrifugation at 16,000 x g at 4°C for 10 min. Biotinylated protein was captured on streptavidin-coated beads (manufacturer) during overnight incubation at 4°C. Beads were collected via centrifugation at 2,500 x g at 4°C for 1 min. Protein was released from beads using Laemmli buffer containing 5% β-mercaptoethanol.
Mass-tagging of palmitoylated proteins
For mass-tag labeling, we followed the procedure described by (Percher et al., 2016). Lysates from A431 cells expressing the indicated constructs were prepared in TEA buffer (50 mM triethanolamine; pH7.3, 150 mM NaCl, and 5 mM EDTA) containing 4% SDS. 200 μg of total cellular protein was treated with a final concentration of 10 mM neutralized TCEP for 30 min with end over end rotation. NEM was added to a final concentration of 25 mM and rocking continued for 2 hours. NEM was removed by 3 rounds of chloroform/methanol/H2O precipitation. The final pellet was resuspended in TEA buffer containing 0.2% Triton X-100. Samples are treated with 0.75 M NH2OH (+HA) or without hydroxylamine (-HA) and incubated at room temperature for 1 hour. Excess hydroxylamine was removed with one round of chloroform/methanol/H2O precipitation and the pellet was resuspended in TEA buffer containing 0.2% Triton X-100 supplemented with 1 mM mPEG-Mal (10 kDa; Sigma). Samples were incubated with rocking for 2 hours and reactions were terminated by 1 round of chloroform/methanol/H2O precipitation. The final pellet was suspended in 1x Laemmli sample buffer and resolved by SDS-PAGE.
Isolation, freezing, and imaging of desmosomes
To isolate desmosomes from mouse liver, a method based on the protocol of Tsukita and Tsukita (1989) was used, in which a desmosomal fraction was obtained by sucrose density gradient centrifugation, followed by NP-40 detergent treatment. The fraction should contain only bile canaliculi derived plasma-membranes, as the homogenization and centrifugation steps are designed to free the preparation from contaminating cell fragments and nuclear membranes due to their higher densities (Neville, 1960; Song et al., 1969; Tsukita and Tsukita, 1989). The desmosomal fraction was immediately plunge-frozen on holey carbon grids, which were subsequently inserted into the column of a FEI Titan Krios at liquid nitrogen temperatures. Tilt series of the sample (+60 to −60°) were recorded and subsequently reconstructed into 3D tomograms.
Membrane thickness measurements
For the thickness of the desmosomal membranes, 1768 selected positions (derived from 3 desmosomes in 2 tomograms) with visible cadherins were selected. As a comparison (control), 668 randomly selected positions at membranes adjacent to desmosomes (derived from 3 membranes in 3 tomograms) and 515 randomly selected positions from arbitrary membranes in the tomograms (derived from 2 membranes in 1 tomogram) were selected. Each position was cross-correlated with multiple references of a simplified membrane model of the two leaflets (dark lines representing phospholipid head groups and are included in the measurements) with varying bilayer distance (3.08, 3.52, 3.96, 4.4, 4.84 and 5.28 nm spacing) using sub-tomogram averaging routines with limited rotational freedom (±30° in 5° steps for all three Euler-angles) after rough pre-alignment using the overall membrane orientation. The reference with the highest cross-correlation score then provides the bilayer spacing of each single sub-volume
Statistics
Error bars represent standard error of the mean. Significance was determined using a student’s t-test (two tailed, heteroscedastic) and p-values have been indicated. Statistical analysis of immunofluorescence results was conducted on at least two independent experiments with ten images per condition per replicate. Statistical analysis of western blotting was conducted on results from three independent experiments.
Acknowledgements
The authors would like to thank Dr. Kathleen Green and members of the Kowalczyk lab for comments and insights during the preparation of this manuscript. We would also like to thank Joseph Lorent for help generating the transmembrane domain models. This work was supported by grants (R01AR048266, R01AR048266-13S1, and R01AR050501 to A.P.K.), (LOEWE Dynamem to A.S.F) and fellowships (F31AR066476 and T32GM008367 to J.D.L.) from the National Institutes of Health, and by the Practical Research Project for Rare/Intractable Diseases (16ek0109067h0003 to Y.M. and 16ek0109151h0002 to A.K.) from the Japan Agency for Medical Research and Development. Additional support was provided by core facilities at Emory University, including the Integrated Cellular Imaging core, the Emory Flow Cytometry Core, and the Cloning Division within Emory Integrated Genomics Core.