In Vitro Functional Analyses of Arrhythmogenic Right Ventricular Cardiomyopathy-Associated Desmoglein-2-Missense Variations

Background Although numerous sequence variants in desmoglein-2 (DSG2) have been associated with arrhythmogenic right ventricular cardiomyopathy (ARVC), the functional impact of new sequence variations is difficult to estimate. Methodology/Principal Findings To test the functional consequences of DSG2-variants, we established an expression system for the extracellular domain and the full-length DSG2 using the human cell line HT1080. We established new tools to investigate ARVC-associated DSG2 variations and compared wild-type proteins and proteins with one of the five selected variations (DSG2-p.R46Q, -p.D154E, -p.D187G, -p.K294E, -p.V392I) with respect to prodomain cleavage, adhesion properties and cellular localisation. Conclusions/Significance The ARVC-associated DSG2-p.R46Q variation was predicted to be probably damaging by bioinformatics tools and to concern a conserved proprotein convertase cleavage site. In this study an impaired prodomain cleavage and an influence on the DSG2-properties could be demonstrated for the R46Q-variant leading to the classification of the variant as a potential gain-of-function mutant. In contrast, the variants DSG2-p.K294E and -p.V392I, which have an arguable impact on ARVC pathogenesis and are predicted to be benign, did not show functional differences to the wild-type protein in our study. Notably, the variants DSG2-p.D154E and -p.D187G, which were predicted to be damaging by bioinformatics tools, had no detectable effects on the DSG2 protein properties in our study.


Introduction
Arrhythmogenic right ventricular cardiomyopathy (ARVC) is a heart muscle disease, characterised by non-ischemic ventricular arrhythmias and progressive dystrophy of the myocardium with fibrofatty replacement (reviewed in [1]). ARVC is a leading cause of exercise-related sudden death in people ,35 years old [2], with an estimated prevalence of 6:10,000 to 44:10,000 [3]. Although the initial reports of ARVC described the right ventricular disease, left ventricular involvement or even global dilated cardiomyopathy also occur [4]. The occurrence of biventricular disease is also supported by genomics data, which have revealed that only a small number of genes are differentially regulated between the right and left ventricle in ARVC [5]. Myocyte loss accompanied by inflammation, fibrosis and adiposis mainly within the right ventricular wall are the histological hallmarks of the disease [6][7][8].
Between 50% and 80% of ARVC cases are regarded as familial and have autosomal dominant inheritance in most cases [1,6,9]. Estimating the amount of familial cases remains difficult due to reduced and age-dependent penetrance combined with variable and incomplete disease expression within a family [10,11].
Desmosomes are highly organised anchoring junctions that are primarily abundant in tissues that are subject to frictional and shear stress (reviewed in [16]). As maculae adhaerentes they are intercellular adhering junctions which secure the structural and functional integrity of cardiac tissue [17][18][19]. Together with adherens junctions (fasciae adhaerentes) and gap junctions, they localise to the intercalated discs (ID), the main site of myocyte interconnection [17,20]. However, colocalisation of the usually distinct components of desmosomes and adherens junctions is observed in a complex and intimate structure termed area composita in the human ID [17].
Although between 30% and 40% of ARVC cases harbour sequence variants in one of the known disease-causing genes [11] the etiopathogenesis of the disease remains poorly understood. Rather than containing mutational hot spots, DSG2 is mutated in numerous and varied locations spread over the complete open reading frame. Therefore, interpretation of missense mutations/ variations for genetic counselling still remains challenging [11].
In this study, we recombinantly expressed the extracellular cadherin domains (EC1-4) of wild-type and mutant DSG2 (rECD) and subsequently analysed their secondary structure and prodomain cleavage. We further tested the homo-dimerisation properties of these molecules in solution and their cellular adhesion properties with a flow cytometry-based binding assay. Additionally, we examined the cellular localisation of full-length DSG2-EYFP (fl-DSG2-EYFP) chimeras in cells. For the analysis, we selected five ARVC-related DSG2-variants located within the extracellular domain (ECD; Table 1): p.R46Q [12], p.D154E [32], p.D187G [33], p.K294E [34], p.V392I [35]. Our goal was to analyse the functional impact of sequence variations that have been reported to be associated with ARVC.
Note that, although the variants differ from the wild-type protein only by one amino acid, the migration of rECD-variants in SDS-PAGE differed from that of the rECD-wt. In the case of rECD-R46Q, differing electrophoretic velocities could be explained by altered processing with MALDI-in-source decay (MALDI-ISD). For the other variants, the deviating mobility might result from different glycosylation patterns as compared to the wild-type molecule.
rECD-wt Secondary Structure, Consisting Mainly of bstrands, is Affected by Ca 2+ Since at least one of the protein variants (rECD-V392I) could not be purified under native conditions by affinity chromatography (data not shown), we purified all rECDs under denaturating conditions for a better comparability. As a control we evaluated the influence of the purification conditions on the secondary structure of rECD-wt by circular dichroism (CD). We did not find evidence for structural differences between denaturating and native conditions by CD-spectroscopy. Evaluation of CD data with DichroWeb (http://dichroweb.cryst.bbk.ac.uk/) revealed that the recombinant proteins mainly consisted of b-strands with only a small ratio of a-helices ( Figure 2).
The addition of 5 mM Ca 2+ led to a significant decrease of ahelical and to a significant increase of b-strand content (Figure 2), indicating the dependency of the secondary structure of the rECDwt on the presence of Ca 2+ . Additional CD data (molar circular dichroism; De) are presented in the Supporting Information ( Figure S1).
ARVC-associated DSG2-p.R46Q Inhibits the Prodomain Cleavage of rECD Prodomain cleavage was previously shown to be essential for the maturation of other DSG isoforms [27]. A consensus PC cleavage site in DSG2 is located at amino acids 46-49 (positions 1-4 of the cleavage site sequence logos; Figure S2). In the DSG2-p.R46Q variant, the arginine at position 1 is substituted by a glutamine, thereby eliminating the potential PC cleavage site.
Amino acid sequences of the N-terminus of the rECDs and prodomain cleavage were examined by MALDI-ISD. N-terminal fragmentation corresponding to SS-Pro-ECD-c-ions (DSG2-preproprotein) was not observed for any of the rECDs (data not shown), demonstrating proper signal sequence cleavage. Likewise, the prodomain was properly cleaved for all rECDs with the exception of rECD-R46Q, as demonstrated by the peaks corresponding to ECD-c-ions (matured DSG2) that were detected ( Figure 3B). The lack of ECD-c-ions and the presence of Pro-ECDc-ions (DSG2-proprotein) for rECD-R46Q ( Figure 3C) signify that cleavage of the rECD-R46Q-prodomain failed.
rECD-R46Q Shows an Increased Binding to HT1080 Cells as Compared to rECD Wild-Type and Other Variants in the Flow Cytometry-based Adhesion Assay The human fibrosarcoma cell line HT1080 [36], used here as a model system for the flow cytometry-based adhesion assay, is similar to the human cardiac area composita of the ID [17,37] in that it expresses the desmosomal cadherins DSG2 and DSC2 [22,38] as well as the classical cadherin N-cadherin ( Figure S3). To establish the cellular adhesion assay, the binding of 0.8 mM rECDwt to HT1080 was analysed in 5 mM CaCl 2 or in 2 mM EGTA ( Figure 4). rECD-wt-binding to HT1080 was detectable in samples incubated with 5 mM CaCl 2 , whereas EGTA led to a significant (p,0.01) decrease of flow cytometry-detectable rECD-binding. Further information about how the Ca 2+ -dependent cell adhesion model was established is available in the online Supporting Information ( Figure S4 and S5).
Statistical analyses on the ratio of rECD-binding as compared to that of controls revealed that no rECDs except for rECD-R46Q showed significant difference in binding to HT1080 ( Figure 5A, 5B). In contrast to the others, the ratio of rECD-R46Q-binding to HT1080 increased significantly (2.2960.08) as compared to the rECD-wt (1.2560.08; p,0.0001).
We assumed that rECDs bind to DSG2 expressed by HT1080 cells. Although siRNA knock-down led to a reduction of available DSG2 epitopes ( Figure S6), we could not detect a reduced binding of rECD-wt after DSG2 knock-down ( Figure S7). The identity of further rECD-binding partners on HT1080 remained unclear.

ARVC-associated DSG2 Variations do not Influence Homo-oligomerisation Properties of rECD
Protein interactions can be identified by crosslinking neighbouring proteins by suberic acid bis(3-sulfo-N-hydroxysuccinimide ester) sodium salt (BS 3 ) and subsequent SDS-PAGE. BS 3 treatment of rECDs in 5 mM CaCl 2 resulted in the formation of rECD dimers and oligomers ( Figure 5C). Reproducibility of rECD oligomer formation was verified for two independent rECD purifications. Since dimerisation and oligomerisation of rECDs were found for both the wild-type and the rECD-variants, we conclude that the investigated ARVC-related DSG2-variants have basically no influence on the multimerisation properties of rECD in solution ( Figure 5C).

Discussion
Since the initial characterisation of DSG2 as an ARVCassociated gene in 2006 further disease-linked variants have been identified in the desmosomal cadherins [12,[32][33][34][35]39]. As clinical diagnosis in ARVC is challenging, genetic screening plays an increasing role in the clinical classification of familial forms of ARVC. In the revised Task Force Criteria (TFC) for classic rightdominant subtype of ARVC, Identification of a pathogenic mutation categorized as associated or probably associated with ARVC/D in the patient under evaluation is even one of the major criteria [40]. But the high prevalence of ''private'' mutations in ARVC patients makes interpreting genetic test results highly challenging, since the pathogenic influence of each novel variation has to be confirmed. According to TFC a pathogenic mutation is defined as a DNA alteration associated with ARVC/D that alters or is expected to alter the encoded protein, is unobserved or rare in a large, non-ARVC/D control population, and either alters or is predicted to alter the structure or function of the protein or has demonstrated linkage to the disease phenotype in a conclusive pedigree [40]. A transgenic mouse model only exists for one gene variation (DSG2-p.N266S) so far [41], and molecular pathomechanisms in desmosomal cadherin variants have been investigated in only a few studies [42][43][44][45]. Important goals of our study were therefore to develop tools that can be directly applied to novel missense variations in DSG2-ECD, and to gain further insight into the functional consequences of sequence variations for the development of ARVC. The prevalence in controls for the DSG2-V392I was adapted to the results in our research group [35]. We chose five ARVC-associated DSG2-variations localised in different extracellular subdomains of DSG2 that were previously published by us and others [12,[32][33][34][35] (Table 1). The DSG2p.R46Q mutation [12] affects a conserved PC recognition motif. Variations DSG2-p.D154E [32] and DSG2-p.D187G [33] are localised within the conserved Ca 2+ binding sites ( Figure S9). In contrast, the two other variations, DSG2-p.K294E [34] and DSG2-p.V392I [35], are located in the extracellular cadherin domains 3 and 4, respectively.  The extracellular domain of desmogleins has been previously used as an independent functional unit, e.g. for nanomechanical interaction analysis [25,46,47]. We therefore consider the recombinantly expressed EC1-4 of DSG2 (rECD) to be an appropriate model for the biochemical studies of ARVCassociated DSG2-variants. The recombinantly expressed rECDwt was confirmed by CD to consist mainly of b-strands. This is in agreement with the data obtained by Sano et al. (PDB ID 2YQG [48]), who found 4% helical and 34% b-sheet content for the first extracellular domain of DSG2. Since our data on the secondary structures of the DSG2-EC1-4 were comparable (Figure 2), we concluded that the purified molecules were appropriate for subsequent experiments.
It is well known that Ca 2+ binds to desmosomal cadherins and induces conformational changes in the ECD, as has been previously shown by CD for the first two extracellular domains of DSG2 and DSC2 [24]. Of note, we show here for the first time that the addition of Ca 2+ indeed leads to an increase of the bstrand content in the extracellular cadherin domains EC1-4. However, the rECD consists mainly of b-strands even under Ca 2+free conditions, which indicates that the basic structure of the extracellular domain is not dependent on the presence of Ca 2+ . Rather Ca 2+ just induces further b-strand formation, presumably in proximity to the Ca 2+ coordination sites. These changes might also induce an alteration of the tertiary structure that has been detected for classical cadherins upon Ca 2+ -binding [49,50].
We should also mention that, although rECD-V392I could not be purified under native conditions, deconvolution of CD data on this variant showed no significant deviation of a-helix and b-strand content as compared to the wild-type protein (data not shown).
Cadherins, including the desmosomal cadherins, are synthesised as inactive precursors, which are post-translationally processed by removal of the N-terminal prodomain. In a baculovirus overexpression system DSG1 and DSG3 were shown to be substrates for the PC furin [27]. The wild-type DSG2 precursor contains a dibasic recognition motif (RQKR) predicted to be appropriate for cleavage by furin and other PCs active in the early secretory pathway such as PACE4, PC5/6 and PC7/8 [51][52][53][54]. By analysing the N-terminal sequence of the recombinantly expressed extracellular domain of DSG2 with MALDI-ISD we revealed that the sequence corresponded to the mature DSG2. This proves that the signal sequence and the prodomain were cleaved from the DSG2 molecule in our HT1080 cell model system. However, for the DSG2-p.R46Q-variant, MALDI-ISD analysis of the Nterminus showed that the molecule still carried the prodomain but lacked the signal sequence. These findings support the assumption that DSG2 like DSG1 and DSG3 is matured by a PC, which specifically recognises a dibasic motif. Additionally, it is obvious that the signal sequence is cleaved independently and earlier during intracellular maturation, and that it is located after amino acid 23, in agreement with the cleavage site predicted by SignalP 4.0 [55] and the UniProtKB/Swiss-Prot database [56] (accession number Q14126).
The presence of the prodomain in the rECD-R46Q-variant explains its deviating electrophoretic mobility as compared to the wild type protein (Figure 1). For the other investigated variants it is speculated that the deviating electrophoretic mobilities might result from different glycosylation patterns between wild type and variants. The approval of this assumption is in the focus of future studies.
For the classical cadherins, maturation by prodomain cleavage has been predicted to be necessary for the adhesive functions of the molecules [26,57]. Only classical (type I) cadherins and desmocollins have prodomains of considerable length. It has been confirmed that the prodomain of murine N-cadherin (134 aa) contains a cadherin-like fold without homophilic interaction capabilities [57]. Since even an incomplete prodomain cleavage results in loss of the adhesion properties of E-cadherin [26] we predicted that even the comparably small prodomain of DSG2 (26 aa) would influence the adhesion of the p.R46Q-variant. Surprisingly, compared to the wild-type the mutant molecule showed an increase in binding of about two-fold; this could be due to yet unknown binding properties of DSG2's prodomain. Aside from the p.R46Q variant, no other rECDs investigated differed in their adhesion properties to HT1080 from the wild-type molecule under the investigated conditions. The reason for the increased cellular adhesion of the p.R46Q variant was not investigated in this study but should be analysed in future studies, for example by using nanomechanical measurements. Nevertheless, we postulate that the p.R46Q variant is a gain-of-function mutation, since it confers new or enhanced activity to the protein. The flow cytometry-based assays were performed under unphysiologically high Ca 2+ -concentrations (5 mM CaCl 2 ). However, this Ca 2+ -concentration was also previously used for the analysis of the first two extracellular subdomains of DSG2 [24]. Since two of the variants (p.D154E and p.D187G) are predicted to affect the Ca 2+ -binding regions ( Figure S9) the molecules might be sensitive to moderate Ca 2+ -changes.
We further analysed the rECD binding properties by crosslinking experiments and found that the rECDs of all variants and the wild-type formed homo-oligomers.
The human fibrosarcoma cell line HT1080 [36] can be regarded as an appropriate cell adhesion model for the study of ARVC-associated DSG2-variants. On the one hand HT1080 cells are convenient for the development of a robust and simple assay to investigate the binding properties of the extracellular domains of DSG2 to the surface of human cells. This is important since FACS analysis of cardiomyocytes is tricky, and human cardiomyocytes are difficult to handle and obtain. On the other hand HT1080 cells are comparable to cardiomyocytes [17,37] as they express the classical cadherin N-cadherin ( Figure S3) as well as the desmosomal cadherins DSC2 and DSG2 ( Figure S3; [38]). This is important considering the fact that siRNA knock-down of DSG2 on HT1080 did not lead to reduced binding of rECD-wt in our study. As heterophilic interactions between DSG and DSC have been previously shown [22,24], the rECD might therefore also bind to the DSC2 or even to N-cadherin present on the surface of HT1080 cells in the DSG2 knock-down experiments.
Since the synthesis of DSG2 needs several posttranslational modifications such as glycosylation and vesicular trafficking (from the ER to the Golgi apparatus, and finally to the plasma membrane), ARVC-related variants might affect transport or modification of the molecule. As a type I transmembrane protein [21,58] the DSG2-ECD should be cotranslationally translocated into the ER. Within the ER, a strict quality control (QC) hinders incompletely folded proteins from leaving the ER and entering the cell-surface where they could be potentially damaging [59,60]. Chimeras of a desmosomal and a fluorescent protein were previously used to analyse the localisation and behaviour of desmosomal proteins within cells [42,61,62]. We used chimeras of fl-DSG2 C-terminally coupled to EYFP to analyse the localisation of wild-type and mutant molecules in living HT1080 cells. Since all chimeras were transferred to the plasma membrane, we concluded that the missense variations do not have a major impact on intracellular QC. Similarly, no influence of intracellular ARVC-associated DSG2-variations on the localisation of fl-DSG2-EGFP chimeras was shown by Gehmlich et al. in primary neonatal rat cardiomyocytes [42]. In contrast to these results we (data not shown) and others [44,63] have found that some DSC2fluorescent protein chimeras carrying other ARVC associated variations (DSC2-p.R203C, -p.I231T, -p.T275M, -p.D350Y,p.A897KfsX4) were mislocalised as compared to the wild-type protein.
While we can assume that rECD-R46Q is not properly processed, it is interesting to note that fl-DSG2-R46Q-EYFP appeared to be transported correctly to the plasma membrane. This is comparable to mutant E-cadherin [26] and N-cadherin [57] which are also fully integrated into the plasma membrane when expressed together with their prodomains.
In summary, we present here an expression system for analysing the extracellular domain and full-length human DSG2. To our surprise we could not find relevant differences in vitro in the binding properties or the cellular localisation, respectively, for the variants DSG2-p.D154E and -p.D187G. Figure 5. Adhesion properties of rECDs. A+B Flow cytometrybased assay for the binding of 0.8 mM rECD-wt or -variants to HT1080. A Representative histograms of FITC-fluorescence for binding of rECD-wtand rECD-R46Q (as indicated). Bound rECD was detected with anti-HisFITC. As a negative control, HT1080 cells were incubated with only anti-HisFITC (grey filled area). B Column plots representing the ratio of rECD-binding related to the negative control (ratio rECD-bound ) as detected by flow cytometry. Ratios rECD-bound are indicated as mean6 SEM of 7 independent measurements for rECD-variants and 9 independent measurements for rECD-wt with rECDs from at least 3 different purifications. Statistical analysis was performed by one-way ANOVA with Dunnett's posttest using rECD-wt as a control (GraphPad Prism 5.01). rECD-R46Q-binding to HT1080 is increased 1.8-fold as compared to rECD-wt. Other ARVC-associated variants have no influence on rECD-binding to HT1080. According to Table 1, these variants were classified by a predictive algorithm as possibly or probably damaging. In contrast the variants p.K294E and p.V392I were classified benign. Furthermore with respect to the clinical data the pathogenic impact of these variants remains debatable. Of note, DSG2p.R46Q, which was classified as probably damaging [64], was properly transported to the plasma membrane despite its aberrant post-translational processing. However, flow cytometry analysis suggested that this variant is a gain-of-function rather than a lossof-function mutation.

Cloning and Mutagenesis of DSG2 and DSC2b cDNA Constructs
Human full-length (fl) DSG2 and DSC2b cDNA were obtained by reverse transcription of RNA isolated from explanted cardiac tissue. The myocardial samples were obtained from patients undergoing cardiac transplantation and prepared according to standardised operating procedures (SOP). The investigation conforms to the principles outlined in the Declaration of Helsinki. The study was approved by the ethics committee of the Ruhr-University Bochum in Bad Oeynhausen, and all patients provided written informed consent for the use of tissue samples.

Primary Antibodies
The clone 3B11 (anti-DSG2-3B11 [65]; Figure S10) directed against the extracellular domain of DSG2 was used for indirect immunofluorescence detection either with a microscope or by flow cytometry. To detect His-tagged proteins by flow cytometry, a FITC-conjugated anti-6x His tagH antibody (anti-HisFITC; ab1206, Abcam) was used. The affinity-isolated rabbit anti DSC2 (anti-DSC2; HPA01911, Sigma) and a rabbit polyclonal antibody to N-cadherin IgG (anti-NCad; ab18203, Abcam) were used for indirect immunofluorescence detection either with a microscope or by flow cytometry.

Transfection of HT1080 cells
For all transfections, Lipofectamine TM 2000 (LFA; Life Technologies) was used according to the manufacturer's protocol.
For DSG2 siRNA knock-down 2610 5 HT1080 cells were reversely transfected with 100 nM DSG2 SilencerH Select Predesigned siRNA (ID s2773; Life Technologies). As a control LFA mock transfected cells and cells transfected with SilencerH Select Negative Control #1 siRNA (Life Technologies) were used. Knock-down efficiency was analysed 96 h after transfection by flow cytometry and Western blot. For Western blot cells were lysed by subcellular fractionation [66] and the membranous fraction was analysed. Purification of Recombinant Proteins rECDs expressed in HT1080 and collected from cell culture supernatant were dialysed against native or denaturating (4 M urea) HisTrap binding buffer and purified by affinity chromatography on HisTrap TM HP Columns (GE Healthcare), according to the manufacturer's instructions. Proteins were eluted with HisTrap native or denaturating elution buffer, respectively, and dialysed against a urea-free buffer.
All spectra were acquired on an ultrafleXtreme with smartbeam-II TM solid state 1 kHz laser, FlashDetector TM , and PAN TM broadband resolution with FlexControl software v.3.3 (BrukerDaltonics).

Circular Dichroism Spectroscopy (CD)
For CD spectroscopy, the purified rECDs were displayed against 10 mM Tris-HCl, pH 7.5, and measured at a concentration of 0.09-0.42 mg/ml with or without 5 mM CaCl 2 . The CD spectra were recorded on a J-810 spectrometer (Jasco) in a 1 mm quartz cell at 22uC, using a scanning rate of 50 nm min 21 , a data pitch of 0.1 nm and three accumulations. All spectra were measured in the range of 350 to 190 nm and subjected to baseline correction. Deconvolution of the CD spectra was performed with DichroWeb [67]. For secondary structure estimation the average of all matching solutions obtained with CONTINLL based on the CONTIN algorithm [68,69] and the CRYST175 reference database were used [70].

Chemical Crosslinking
For crosslinking 0.8 mM rECD in 10 mM HEPES, pH 7.5, 150 mM NaCl, and 5 mM CaCl 2 were incubated with BS 3 (Sigma Aldrich) at a final concentration of 10 mM at 22uC for 60 min. The reaction was stopped by adding Tris-HCl, pH 7.5 to a final concentration of 90 mM. Negative controls were incubated under the same conditions but without BS 3 .

Flow Cytometry Binding Assay
HT1080 cells were collected with enzyme-free PBS-based cell dissociation buffer (Life Technologies). 2.5610 5 cells were incubated either with primary antibody or with rECD for 30 min at 4uC in a final volume of 50 ml with 5 mM CaCl 2 . After washing with ice cold PBS bound primary antibody and rECD were incubated with a fluorescence-dye conjugated antibody for 30 min at 4uC in the dark. After washing with ice cold PBS cells were resuspended in PBS and analysed by flow cytometry with FACScan (Becton Dickinson). As a negative control cells only incubated with the fluorescence-dye conjugated antibody were used.
For flow cytometric analysis after knock-down, siRNA and mock transfected cells were used 96 h after transfection.
For indirect immunofluorescence of HT1080 cells, see Supporting Information.

Statistical Analyses
Statistical analyses were performed with the unpaired student's t-test, one-way ANOVA with Dunnett's or Bonferroni's posttest, or two-way ANOVA using Prism v5.01 (GraphPad Software, San Diego California USA). All indicated values are presented as means6standard error of the mean (SEM).
Detailed protocols can be obtained from the Supporting Information.  [7,8]. For explanations on how to interpret the cleavage site sequence logo compare Crooks et al. [8]. (TIF) Figure S3 Analysis of cadherin expression on HT1080 cells with immunofluorescence microscopy. To verify the expression of cadherins on the human fibrosarcoma cell line HT1080 [9] unpermeabilised cells were labelled with anti-DSC2 (A), anti-DSG2-3B11 (B), and anti-NCad (C) antibodies and corresponding FITC-(A+C) or Cy TM 3-conjugated (B) secondary antibodies, respectively. Negative controls performed with the secondary antibody only (not shown) did not show any specific fluorescence under the same conditions. Nuclear staining was performed with DAPI (blue). Immunofluoresecence microscopy revealed that DSG2, N-cadherin and in traces DSC2, the cadherins of the area composita in the human ID, are also expressed on HT1080. N-cadherin localises especially at the cell borders whereas DSC2 and DSG2 are scattered over the cell surface. Scale (red or green bar) = 10 mm. (TIF) Figure S4 Flow cytometry analysis of HT1080. Cells were brought into suspension with enzyme free cell dissociation buffer to avoid the cleavage of cadherins from the cell surface and incubated with rECD and the appropriate antibodies at 4uC to inhibit epitope endocytosis. For the discrimination between live/ dead cell population cells were subsequently incubated with PI. Representative FL1/FL2 (A) dot plot shows one population (blue) with a low FL2 fluorescence intensity and two populations with high FL2 fluorescence intensity (pink, red circle). High FL2 fluorescence intensity corresponds to a PI uptake characteristic of dead cells. The dead cells in A correspond to the pink population in the representative SSC/FSC dot plot (B). The blue cell population in B corresponding to live cells was gated. Only the cells in the live cell gate (70-80% of all cells) were considered for the flow cytometry-based binding assay. (TIF) Figure S5 Flow cytometric detection of DSG2 on HT1080. Depletion of Ca 2+ has a negative effect on the detection of DSG2 on HT1080. A HT1080 cells were incubated with (black line) or without (negative control, grey filled area) anti-DSG2-3B11. Bound antibody was detected with anti-msFITC. Shown are histograms of FITC fluorescence for DSG2 detection with 5 mM CaCl 2 (1) or with 2 mM EGTA (2). B Column plots representing the ratio of DSG2 detection related to the negative control. Ratios are indicated as mean6SEM of 3 independent measurements. Fluorescence intensity ratio was significantly (p,0.0001) decreased from 1.5760.01 for samples incubated in 5 mM CaCl 2 (1) to 1.3160.02 for samples incubated with 2 mM EGTA (2). Statistical analysis was performed with unpaired student's t-test (GraphPad Prism 5.01). (TIF) Figure S6 Detection of DSG2 on HT1080 after siRNA knock-down. Treatment with DSG2-specific siRNA led to a reduction of DSG2 in HT1080. A Shown are representative histograms of FITC fluorescence for the detection of DSG2 with anti-DSG2-3B11+anti-msFITC on HT1080 in control cells (1, black line and 2, red line) and after DSG2-specific siRNA knockdown (3, green line). As negative control only anti-msFITC (grey filled area) was used. B Column plots represent the ratio of flow cytometric DSG2 detection (ratio DSG2 ; for calculation see formula S4) related to the negative control. Ratios DSG2 are indicated as mean6SEM of 3 independent knock down experiments. Statistical analysis was performed with one-way ANOVA with Bonferroni's posttest (GraphPad Prism 5.01). Even control siRNA treatment significantly decreased surface DSG2 on HT1080 cells. After DSG2-specific siRNA knock-down surface DSG2 was undetectable by flow cytometry. C Western blot analysis of the membranous fraction (5 mg/lane) of HT1080 cells with anti-DSG1+2-DG3.10 (red box) and anti-PDI (grey box, ca. 60 kDa) as ER marker and loading control. Full-length DSG2 (ca. 165 kDa) and a cleavage fragment (ca. 105 kDa) were detectable in all three samples. DSG2 expression was obviously reduced in the lysate derived from DSG2 siRNA treated cells. Loading marker PDI showed that variability in protein loading could not account for the observed DSG2 decrease in the DSG2 siRNA treated sample. Due to high non-specific binding antibodies against the extracellular domain of DSG2 (anti-DSG2-3B11 and anti-DSG2-10G11) were not suitable for Western blot analysis of cell lysates. (TIF) Figure S7 Flow cytometry-based assay for the binding of rECD-wt to HT1080 cells after siRNA knock-down. A Shown are representative histogram plots of FITC-fluorescence for rECD-wt (0.8 mM) binding to HT1080. Bound rECD-wt was detected with anti-HisFITC. As negative control only anti-HisFITC (grey filled area) was used. B Column plots represent the ratio of rECD-wt-binding (ratio rECD-bound ; calculation according to formula S3) to HT1080. Values represent the mean ratio6SEM of 3 independent knock-down experiments. Analysis was performed with one-way ANOVA and Bonferroni's posttest (GraphPad Prism 5.01). Since rECDs were able to form homooligomers in solution ( Figure 5C) and DSC2 for hetero-oligomer formation is only rarely present at the cell surface ( Figure S3) we assumed that rECDs bind to DSG2 expressed by HT1080 cells. Nevertheless rECD-wt-binding was not influenced by DSG2 siRNA knock-down. (TIF) Figure S8 DSC2b overexpression in HT1080 cells leads to increased surface expression of endogenous DSG2. In wild type HT1080 cells neither overexpression of fl-DSG2 nor the expression of chimeric fl-DSG2-wt-EYFP was possible (data not shown). A Immunofluorescence analysis of DSC2b-HT1080 cells.

Supporting Information
To verify the expression of DSC2 unpermeabilised HT1080 cells stably transfected with DSC2b (DSC2b-HT1080) were labelled with anti-DSC2 (+) and anti-rbFITC as a secondary antibody (green). Nuclear staining was performed with DAPI (blue). Shown are representative immunofluorescence images for three immunolabelling experiments. The controls were treated with anti-rbFITC only (-). DSC2 expression was increased in DSC2b-HT1080 as compared to HT1080 (see also Figure S3); protein localised particularly at the cell borders. Scale (white bars) = 10 mm. B Flow cytometric analysis of DSG2 and DSC2 expression on HT1080 or DSC2b-HT1080 cells. Cells were treated with anti-DSG2-3B11+anti-msFITC and anti-DSC2+anti-rbFITC, respectively and analysed by flow cytometry. Column plots represent the ratio of DSG2 or DSC2 on HT1080 cells (ratio desmosomal cadherin ; for calculation compare ratio DSG2 , formula S4) as compared to the negative control (secondary antibody only). Values represent the mean ratio6SEM of 3 independent flow cytometry experiments; statistical analysis was performed with two-way ANOVA (GraphPad Prism 5.01). The analysis reveals that stable transfection of HT1080 cells with a human DSC2b construct (fl-DSC2b-pLPCX) and overexpression of DSC2b in DSC2b-HT1080 lead to an increased expression of endogenous DSG2 as compared to wild type HT1080. This is in agreement with the assumption that DSC-surface expression is necessary for DSG2-membrane transport [10]. Since DSC2b overexpression promotes DSG2 expression, DSC2b-HT1080 cells were used for fl-DSG2-EYFP expression. (TIF) Figure S9 Protein alignment of murine N-cadherin (mNCad), murine E-cadherin (mECad) and human desmoglein-2 (hDSG2). The Ca 2+ binding motifs as annotated for murine N-cadherin (PDB ID 3Q2W) and murine E-cadherin (PDB ID 3Q2V) are highlighted in yellow [11]. The amino acids in hDSG2 corresponding to D154 and D187 are indicated in red and green, respectively. It is obvious that the ARVC-associated variations D154E and D187G concern conserved Ca 2+ -binding motifs. The multiple sequence alignment was performed with ClustalW. '*' indicates positions which have a single, fully conserved residue, ':' indicates that one of the 'strong' groups is fully conserved,'.' indicates that one of the 'weaker' groups is fully conserved. (TIF) Figure S10 Detection of rECDs with anti-DSG2-3B11 by Western blot. Although rECD-wt and all variants except rECD-R46Q were shown to lack the prodomain by MALDI-ISD ( Figure 3) recombinantly expressed proteins were stained with anti-DSG2-3B11 by Western blot analysis. anti-DSG2-3B11 was claimed to be a DSG2-prodomain specific antibody, previously [12] but it seems to detect also DSG2 fragments lacking the DSG2-Variations in ARVC prodomain. We used anti-DSG2-3B11 as a DSG2 extracellular domain specific antibody in our study. 1-6 = rECDs as labelled in Figure 1.

(TIF)
Table S1 Oligonucleotides used for cloning and the corresponding sequences. The bases used for the addition of a particular tag or for mutagenesis are written in capital letters. (XLSX) Methods S1 Supporting Information: Material and Methods. (DOC)